Documente Academic
Documente Profesional
Documente Cultură
Dierk Raabe
Computational Materials Science
ISBN 3-527-29541-0
Gernot Kostorz (Ed.)
Phase Transformations
in Materials
This book was carefully produced. Nevertheless, authors, editor and publisher do not warrant the informa-
tion contained therein to be free of errors. Readers are advised to keep in mind that statements, data, illus-
trations, procedural details or other items may inadvertently be inaccurate.
All rights reserved (including those of translation in other languages). No part of this book may be reproduced
in any form – by photoprinting, microfilm, or any other means – nor transmitted or translated into machine lan-
guage without written permission from the publishers. Registered names, trademarks, etc. used in this book, even
when not specifically marked as such, are not to be considered unprotected by law.
Composition, Printing and Bookbinding: Konrad Triltsch, Print und digitale Medien GmbH, 97199 Ochsenfurt-
Hohestadt
To the memory of Peter Haasen
(1927 – 1993)
In 1991, the late Peter Haasen, who had set out to edit a comprehensive treatment of “Ma-
terials Science and Technology” together with Robert W. Cahn and Edward J. Kramer, a
very successful series of up-to-date coverages of a broad range of materials topics, wrote
the following in a Preface to Volume 5.
“Herewith we proudly present the first volume of this Series, the aim of which is to pro-
vide a comprehensive treatment of materials science and technology. The term ‘materials’
encompasses metals, ceramics, electronic and magnetic materials, polymers and compos-
ites. In many cases these materials have been developed independently within different dis-
ciplines but are now finding uses in similar technologies. Moreover, similarities found among
the principles underlying these various disciplines have led to the discovery of common
phenomena and mechanisms. One of these common features, phase transformations, con-
stitutes the topic of this volume and rates among the fundamental phenomena central to the
Series. A phase transformation often delivers a material into its technologically useful form
and microstructure. For example, the major application of metals and alloys as mechanical-
ly strong materials relies on their multi-phase microstructure, most commonly generated by
one or more phase transformations.”
Peter Haasen, who passed away in 1993, did not live to see the overwhelming success of
this volume. A revised edition was planned as early as 1996 and finally, work on the indi-
vidual chapters began in 1998. Almost all of the original authors agreed to update their ear-
lier work, in many cases they arranged for the participation of younger colleagues who had
made major contributions to the field. It is thus with similar pride the present editor sub-
mits the second edition of “Phase Transformations” to the scholarly public. It was with some
hesitation that he assumed the task of editor, as he did not like giving the impression of pla-
giarizing a successful work. All the credit for the idea, the original chapter definitions and
the selection of the initial authors remains with Peter Haasen. He would certainly have liked
to work on a new edition himself and, assuming that he would have judged it timely to ac-
complish it about ten years after the first edition, this editor tried to help reaching this goal
in the original spirit.
Thus, all the chapters kept their original titles. The contents have been thoroughly re-
edited and updated and reflect the progress in the field up to about the middle of the year
2000. As before, the book starts with the foundations of phase transformations (“Thermo-
dynamics and Phase Diagrams” by A. D. Pelton). The sequence of the following Chapters
has been slightly modified. As most of the volume concerns the solid state, Chapter 2 (by
H. Müller-Krumbhaar, W. Kurz and E. Brener) is devoted to solidification, a subject of great
basic and technological relevance. Chapter 3 by G. E. Murch covers the most important
ideas and methods of diffusion kinetics in solids, an indispensible ingredient to many phase
transformations. Statistical theories of phase transformations are presented by K. Binder in
Chapter 4, featuring phenomenological concepts and computational methods. Diffusion con-
trolled homogeneous phase transformations are treated in Chapter 5 (“Homogeneous Sec-
2 Solidification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
H. Müller-Krumbhaar, W. Kurz, E. Brener
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 697
Arthur D. Pelton
Symbol Designation
ai activity of component i
C number of components
cp molar heat capacity
E electrical potential of a galvanic cell
F degrees of freedom/variance
G Gibbs energy in J
g molar Gibbs energy in J/mol
gi partial molar Gibbs energy of i
Gi0 standard Gibbs energy of i
gi0 standard molar Gibbs energy of i
Dgi relative partial Gibbs energy i
gE excess molar Gibbs energy
giE excess partial Gibbs energy of i
DG Gibbs energy change
DG 0 standard Gibbs energy change
Dgm molar Gibbs energy of mixing
Dgf0 standard molar Gibbs energy of fusion
Dgv0 standard molar Gibbs energy of vaporization
H enthalpy in J
h molar enthalpy in J/mol
hi partial enthalpy of i
Hi0 standard enthalpy of i
hi0 standard molar enthalpy of i
Dhi relative partial enthalpy of i
hE excess molar enthalpy
hEi excess partial enthalpy of i
DH enthalpy change
DH 0 standard enthalpy change
Dhm molar enthalpy of mixing
Dh0f standard molar enthalpy of fusion
Dhv0 standard molar enthalpy of vaporization
K equilibrium constant
kB Boltzmann constant
n number of moles
ni number of moles of constituent i
Ni number of particles of i
N0 Avogadro’s number
pi partial pressure of i
P total pressure
P number of phases
qi general extensive variable
4 1 Thermodynamics and Phase Diagrams of Materials
R gas constant
S entropy in J/K
s molar entropy in J/mol K
si partial entropy of i
Si0 standard entropy of i
si0 standard molar entropy of i
Dsi0 relative partial entropy of i
sE excess molar entropy
siE excess partial entropy of i
DS entropy change
DS 0 standard entropy change
Dsm molar entropy of mixing
Ds0f standard molar entropy of fusion
Dsv0 standard molar entropy of vaporization
T temperature
Tf temperature of fusion
Tc critical temperature
TE eutectic temperature
U internal energy
vi molar volume of i
vi0 standard molar volume of i
Xi mole fraction of i
Z coordination number
gi activity coefficient of i
e bond energy
h empirical entropy parameter
mi chemical potential of i
n number of moles of “foreign” particles contributed by a mole of solute
x molar metal ratio
s vibrational bond entropy
fi generalized thermodynamic potential
w empirical enthalpy parameter
standard molar Gibbs energies, Gi0. The action continues with the production of
Gibbs energy of the system then varies as: H2S, pH2S will increase while pH2 and pS2
will decrease. As a result, DG will become
9 0 13 0
0
dG / dn = g Al 6 Si 2 O13
+g − g progressively less negative. Eventually an
2 Si 2 SiO2 equilibrium state will be reached when
− 6 g Al
0
= DG 0 = − 830 kJ (1-3) dG/dn = DG = 0.
where DG 0 is the standard Gibbs energy For the equilibrium state, therefore:
change of reaction, Eq. (1-2), at 500 °C. DG 0 = – R T ln K (1-7)
Since DG 0 < 0, the formation of mullite 2 – 2 –1
= – R T ln ( pH2S pH2 pS2 )equilibrium
entails a decrease in G. Hence, the reaction
will proceed spontaneously so as to mini- where K, the “equilibrium constant” of the
mize G. reaction, is the one unique value of the ra-
tio ( p2H2S pH– 22 pS–12 ) for which the system will
1.2.2 Chemical Equilibrium be in equilibrium at the temperature T.
If the initial partial pressures are such
The partial molar Gibbs energy of an that DG > 0, then the reaction, Eq. (1-5),
ideal gas is given by: will proceed to the left in order to minimize
gi = gi0 + R T ln pi (1-4) G until the equilibrium condition of Eq.
(1-7) is attained.
where gi0 is the standard molar Gibbs en- As a further example, we consider the
ergy (at 1 bar), pi is the partial pressure in possible precipitation of graphite from a
bar, and R is the gas constant. The second gaseous mixture of CO and CO2 . The reac-
term in Eq. (1-4) is entropic. As a gas ex- tion is:
pands at constant T, its entropy increases.
Consider a gaseous mixture of H2 , S2 2 CO = C + CO2 (1-8)
and H2S with partial pressures pH2 , pS2 and Proceeding as above, we can write:
pH2S . The gases can react according to
dG/dn = gC + gCO2 – 2 gCO
2 H 2 + S 2 = 2 H 2S (1-5)
= ( gC0 + gCO
0
2
0
– 2 gCO 2
) + R T ln ( pCO2 /pCO )
If the reaction, Eq. (1-5), proceeds to the 0
= DG + R T ln ( pCO2 /pCO ) 2
(1-9)
right with the formation of 2 dn moles of = DG = – R T ln K + R T ln ( pCO2 /pCO 2
)
H2S, then the Gibbs energy of the system
2
varies as: If ( pCO2 /pCO ) is less than the equilibrium
constant K, then precipitation of graphite
dG/dn = 2 gH2S – 2 gH2 – gS2 will occur in order to decrease G.
= (2 gH0 2S – 2 gH0 2 – gS02 ) Real situations are, of course, generally
+ R T (2 ln pH2S – 2 ln pH2 – ln pS2 ) more complex. To treat the deposition of
= DG 0 + R T ln ( p2H2S pH– 22 pS–12 ) solid Si from a vapour of SiI4 , for example,
we must consider the formation of gaseous
= DG (1-6)
I2 , I and SiI2 so that three reaction equa-
DG, which is the Gibbs energy change of tions must be written:
the reaction, Eq. (1-5), is thus a function of SiI4 (g) = Si (sol) + 2 I2 (g) (1-10)
the partial pressures. If DG < 0, then the re-
action will proceed to the right so as to SiI4 (g) = SiI2 (g) + I2 (g) (1-11)
minimize G. In a closed system, as the re- I2 (g) = 2 I (g) (1-12)
1.3 Predominance Diagrams 7
The equilibrium state, however, is still that solid compounds of Cu, S and O. For ex-
which minimizes the total Gibbs energy of ample, at point Z, where pSO2 = 10– 2 and
the system. This is equivalent to satisfying pO2 = 10–7 bar, the stable phase is Cu2O.
simultaneously the equilibrium constants The conditions for coexistence of two and
of the reactions, Eqs. (1-10) to (1-12), as three solid phases are indicated respectively
will be shown in Section 1.12 where this by the lines and triple points on the diagram.
example is discussed further. For example, along the univariant line
(phase boundary) separating the Cu2O and
CuSO4 domains the equilibrium constant
1.3 Predominance Diagrams –2
K = pSO 2
pO– 23/2 of the following reaction is
satisfied:
1.3.1 Calculation of Predominance
Diagrams Cu2O + 2 SO2 + –23 O2 = 2 CuSO4 (1-13)
mation of each solid phase, always from Fig. 1-3. For the formation reaction:
one mole of the base element Cu, and in-
4 Cu + O2 = 2 Cu2O (1-19)
volving the gaseous species whose pres-
sures are used as the axes (SO2 and O2 in we can write:
this example):
DG 0 = – R T ln K = R T ln ( pO2 )equilibrium
1
Cu + – O2 = CuO
2 = DH 0 – T DS 0 (1-20)
DG = DG 0 + R T ln pO–1/2 (1-15)
2 The diagonal line in Fig. 1-3 is thus a plot
1 1 of the standard Gibbs energy of formation
Cu + – O2 = – Cu2O
4 2
DG = DG 0 + R T ln pO–1/4 (1-16) of Cu2O versus T. The temperatures indi-
cated by the symbol M and 씲
2
M are the melt-
Cu + SO2 = CuS + O2 ing points of Cu and Cu2O respectively.
DG = DG 0 + R T ln ( pO2 pSO
–1
2
) (1-17) This line is thus simply a line taken from
the well-known Ellingham Diagram or
Cu + SO2 + O2 = CuSO4
DG 0 vs. T diagram for the formation of
DG = DG 0 + R T ln ( pSO
–1
2
pO–12 ) (1-18) oxides. However, by drawing vertical lines
and similarly for the formation of Cu2S, at the melting points of Cu and Cu2O as
Cu2SO4 and Cu2SO5 . shown in Fig. 1-3, we convert the plot to a
The values of DG 0 are obtained from ta- true phase diagram. Stability domains for
bles of thermodynamic properties. For any Cu (sol), Cu (l), Cu2O (sol), and Cu2O (l)
given values of pSO2 and pO2 , DG for each are shown as functions of T and of imposed
formation reaction can then be calculated. pO2 . The lines and triple points indicate
The stable compound is simply the one
with the most negative DG. If all the DG
values are positive, then pure Cu is the
stable compound.
By reformulating Eqs. (1-15) to (1-18) in
terms of, for example, S2 and O2 rather
than SO2 and O2 , a predominance diagram
with ln pS2 and ln pO2 as axes can be con-
structed. Logarithms of ratios or products
of partial pressures can also be used as
axes.
and tangent constructions similar to that of For the reaction, Eq. (1-38), at 1400 K,
Fig. 1-4 can be used to relate the relative DG 0 = – 68.35 kJ (Barin et al., 1977). The
partial enthalpies and entropies Dhi and activity of Cu in the liquid alloy at
Dsi to the integral molar enthalpy of mix- XCu = 0.6 is aCu = 0.43 (Hultgren et al.,
ing Dhm and integral molar entropy of mix- 1973). Substitution into Eq. (1-39) with
ing Dsm respectively. pO2 = 10– 4 bar gives:
dG/dn = DG = – 50.84 kJ
1.4.6 Activity Hence under these conditions the reaction
entails a decrease in the total Gibbs energy
The activity of a component in a solution
and so the copper will be oxidized.
was defined by Eq. (1-32).
Since ai varies monotonically with gi it
1.4.7 Ideal Raoultian Solutions
follows that when two or more phases are
in equilibrium the activity of any compo- An ideal solution or Raoultian solution
nent is the same in all phases, provided that is usually defined as one in which the ac-
the activity in every phase is expressed tivity of a component is equal to its mole
with respect to the same standard state. fraction:
The use of activities in calculations of
aiideal = Xi (1-40)
chemical equilibrium conditions is illus-
trated by the following example. A liquid (With a judicious choice of standard state,
solution of Au and Cu at 1400 K with this definition can also encompass ideal
XCu = 0.6 is exposed to an atmosphere in Henrian solutions, as discussed in Sec.
which the oxygen partial pressure is 1.5.11.)
pO2 = 10– 4 bar. Will Cu2O be formed? The However, this Raoultian definition of
reaction is: ideality is generally only useful for simple
substitutional solutions. There are more
2 Cu (l) + –12 O2 (g) = Cu2O (sol) (1-38) useful definitions for other types of solu-
where the Cu (l) is in solution. If the reac- tions such as interstitial solutions, ionic so-
tion proceeds with the formation of dn lutions, solutions of defects, polymer solu-
moles of Cu2O, then 2 dn moles of Cu are tions, etc. That is, the most convenient def-
consumed, and the Gibbs energy of the inition of ideality depends upon the solu-
Au – Cu solution changes by tion model. This subject will be discussed
in Sec. 1.10. In the present section, Eq.
– 2 (dG l /dnCu ) dn (1-40) for an ideal substitutional solution
will be developed with the Au – Cu solution
The total Gibbs energy then varies as: as example.
dG/dn = gCu2O – –12 gO2 – 2 (dG l /dnCu ) In the ideal substitutional solution model
it is assumed that Au and Cu atoms are
= gCu2O – –12 gO2 – 2 gCu nearly alike, with nearly identical radii and
0
= (gCu – –12 gO0 2 – 2 gCu
0
) electronic structures. This being the case,
2O
there will be no change in bonding energy
– –12 R T ln pO2 – 2 R T ln aCu or volume upon mixing, so that the en-
= DG 0 + R T ln ( pO–1/2 –2
aCu ) thalpy of mixing is zero:
2
ideal
= DG (1-39) Dhm =0 (1-41)
1.4 Thermodynamics of Solutions 13
Furthermore, and for the same reason, the Hence Eq. (1-40) has been demonstrated
Au and Cu atoms will be randomly distrib- for an ideal substitutional solution.
uted over the lattice sites. (In the case of a
liquid solution we can think of the “lattice
1.4.8 Excess Properties
sites” as the instantaneous atomic positions.)
For a random distribution of NAu In reality, Au and Cu atoms are not iden-
gold atoms and NCu copper atoms over tical, and so Au – Cu solutions are not per-
(NAu + NCu ) sites, Boltzmann’s equation fectly ideal. The difference between a solu-
can be used to calculate the configurational tion property and its value in an ideal solu-
entropy of the solution. This is the entropy tion is called an excess property. The ex-
associated with the spatial distribution of cess Gibbs energy, for example, is defined
the particles: as:
(1-42)
S config = kB ln (NAu + NCu ) ! / NAu ! NCu ! gE = Dgm – Dgmideal (1-47)
where kB is Boltzmann’s constant. The con- Since the ideal enthalpy of mixing is zero,
figurational entropies of pure Au and Cu the excess enthalpy is equal to the enthalpy
are zero. Hence the configurational entropy of mixing:
of mixing, DS config, will be equal to S config.
hE = Dhm – Dhm
ideal
= Dhm (1-48)
Furthermore, because of the assumed close
similarity of Au and Cu, there will be no Hence:
non-configurational contribution to the en-
tropy of mixing. Hence, the entropy of gE = h E – T s E
mixing will be equal to S config. = Dhm – T s E (1-49)
Applying Stirling’s approximation, which Excess partial properties are defined simi-
states that ln N ! = [(N ln N) – N ] if N is larly:
large, yields:
giE = Dgi – Dgiideal
DSmideal = S config = − kB ( NAu + NCu ) (1- 43) = R T ln ai – R T ln Xi (1-50)
⎛ NAu NCu ⎞ siE = Dsi – Dsiideal = Dsi + R ln Xi (1-51)
× ⎜ NAu ln + NCu ln ⎟
⎝ NAu + NCu NAu + NCu ⎠
Also:
For one mole of solution, (NAu + NCu ) = N 0,
giE = hiE – T siE
where N 0 = Avogadro’s number. We also
note that (kB N 0 ) is equal to the ideal gas = Dhi – T siE (1-52)
constant R. Hence: Equations analogous to Eqs. (1-33),
(1-44)
DSmideal = – R (XAu ln XAu + XCu ln XCu ) (1-36) and (1-37) relate the integral and
partial excess properties. For example, in
Therefore, since the ideal enthalpy of mix- Au – Cu solutions:
ing is zero:
(1-45) gE = XAu g Au
E E
+ XCu g Cu (1-53)
Dgmideal = R T (XAu ln XAu + XCu ln XCu )
s E = XAu s Au
E E
+ XCu s Cu (1-54)
By comparing Eqs. (1-33) and (1-45) we
Tangent constructions similar to that of
obtain:
Fig. 1-4 can thus also be employed for ex-
Dgiideal = R T ln aiideal = R T ln Xi (1-46) cess properties, and an equation analogous
14 1 Thermodynamics and Phase Diagrams of Materials
to Eq. (1-29) can be written: Eq. (1-50), Dgi < Dgiideal. That is, the com-
ponent i is more stable in the solution than
giE = gE + (1 – Xi ) dgE /dXi (1-55) it would be in an ideal solution of the same
The Gibbs – Duhem equation, Eq. (1-31), composition. If g i > 1, then giE > 0 and the
also applies to excess properties: driving force for the component to enter
into solution is less than in the case of an
E E
XAu dg Au + XCu dg Cu =0 (1-56) ideal solution.
In Au – Cu alloys, gE is negative. That is, 1.4.10 Multicomponent Solutions
Dgm is more negative than Dgmideal and so
the solution is thermodynamically more The equations of this section were de-
stable than an ideal solution. We say that rived with a binary solution as an example.
Au – Cu solutions exhibit negative devia- However, the equations apply equally to
tions from ideality. If gE > 0, then the solu- systems of any number of components. For
tion is less stable than an ideal solution and instance, in a solution of components
is said to exhibit positive deviations. A– B – C – D …, Eq. (1-33) becomes:
Dgm = XA DgA + XB DgB + XC DgC
1.4.9 Activity Coefficient + XD DgD + … (1-59)
The activity coefficient of a component
in a solution is defined as: 1.5 Binary Phase Diagrams
g i = ai /Xi (1-57)
1.5.1 Systems with Complete Solid
From Eq. (1-50): and Liquid Miscibility
tion, expressed as mole fraction of MnO, (XMnO = 0.28). As the temperature is de-
XMnO . Note that XMnO = 1 – XCaO . Phase di- creased further, solid continues to precipi-
agrams are also often drawn with the com- tate with the compositions of the two
position axis expressed as weight percent. phases at any temperature being given by
At all compositions and temperatures in the liquidus and solidus compositions at
the area above the line labelled liquidus, a that temperature and with their relative
single-phase liquid solution will be ob- proportions being given by the lever rule.
served, while at all compositions and tem- Solidification is complete at 2030 °C, the
peratures below the line labelled solidus, last liquid to solidify having composition
there will be a single-phase solid solution. XMnO = 0.60 (point C).
A sample at equilibrium at a temperature The process just described is known as
and overall composition between these two equilibrium cooling. At any temperature
curves will consist of a mixture of solid during equilibrium cooling the solid phase
and liquid phases, the compositions of has a uniform (homogeneous) composition.
which are given by the liquidus and solidus In the preceding example, the composition
compositions at that temperature. For ex- of the solid phase during cooling varies
ample, a sample of overall composition along the line APC. Hence, in order for the
XMnO = 0.60 at T = 2200 °C (at point R in solid grains to have a uniform composition
Fig. 1-5) will consist, at equilibrium, of a at any temperature, diffusion of CaO from
mixture of liquid of composition XMnO = the center to the surface of the growing
0.70 (point Q) and solid of composition grains must occur. Since solid-state dif-
XMnO = 0.35 (point P). fusion is a relatively slow process, equi-
The line PQ is called a tie-line or co- librium cooling conditions are only ap-
node. As the overall composition is varied proached if the temperature is decreased
at 2200 °C between points P and Q, the very slowly. If a sample of composition
compositions of the solid and liquid phases XMnO = 0.60 is cooled very rapidly from the
remain fixed at P and Q, and only the rela- liquid, concentration gradients will be ob-
tive proportions of the two phases change. served in the solid grains, with the concen-
From a simple mass balance, we can derive tration of MnO increasing towards the sur-
the lever rule for binary systems: (moles of face from a minimum of XMnO = 0.28 (point
liquid) / (moles of solid) = PR /R Q. Hence, A) at the center. Furthermore, in this case
at 2200 °C a sample with overall composi- solidification will not be complete at
tion XMnO = 0.60 consists of liquid and solid 2030 °C since at 2030 °C the average con-
phases in the molar ratio (0.60 – 0.35) / centration of MnO in the solid particles
(0.70 – 0.60) = 2.5. Were the composition will be less than XMnO = 0.60. These con-
axis expressed as weight percent, then the siderations are discussed more fully in
lever rule would give the weight ratio of Chapter 2 of this volume (Müller-Krumb-
the two phases. haar et al., 2001).
Suppose that a liquid CaO – MnO solu- At XMnO = 0 and XMnO = 1 in Fig. 1-5 the
tion with composition XMnO = 0.60 is liquidus and solidus curves meet at the
cooled very slowly from an initial tempera- equilibrium melting points, or tempera-
ture of about 2500 °C. When the tempera- tures of fusion of CaO and MnO, which are
ture has decreased to the liquidus tempera- T f0(CaO) = 2572 °C, T f0(MnO) = 1842 °C.
ture 2270 °C (point B), the first solid The phase diagram is influenced by the
appears, with a composition at point A total pressure, P. Unless otherwise stated,
16 1 Thermodynamics and Phase Diagrams of Materials
T – X diagrams are usually presented for Since, to a first approximation, Dhf0 and
P = const. = 1 bar. For equilibria involving Dsf0 are independent of T, Dgf0 is approxi-
only solid and liquid phases, the phase mately a linear function of T. If T > Tf0 ,
boundaries are typically shifted only by the then Dgf0 is negative. If T < Tf0 , then Dgf0 is
order of a few hundredths of a degree per positive. Hence, as seen in Fig. 1-6, as T
bar change in P. Hence, the effect of pres- decreases, the g s curve descends relative to
sure upon the phase diagram is generally g l. At 1500 °C, g l < g s at all compositions.
negligible unless the pressure is of the or- Therefore, by the principle that a system
der of hundreds of bars. On the other hand, always seeks the state of minimum Gibbs
if gaseous phases are involved then the ef- energy at constant T and P, the liquid phase
fect of pressure is very important. The ef- is stable at all compositions at 1500 °C.
fect of pressure will be discussed in Sec. At 1300 °C, the curves of g s and g l cross.
1.5.3. The line P1 Q1 , which is the common tan-
gent to the two curves, divides the compo-
sition range into three sections. For compo-
1.5.2 Thermodynamic Origin
sitions between pure Ge and P1 , a single-
of Phase Diagrams
phase liquid is the state of minimum Gibbs
In this section we first consider the ther- energy. For compositions between Q1 and
modynamic origin of simple “lens-shaped” pure Si, a single-phase solid solution is the
phase diagrams in binary systems with stable state. Between P1 and Q1 , a total
complete liquid and solid miscibility. Gibbs energy lying on the tangent line
An example of such a diagram was given P1 Q1 may be realized if the system adopts
in Fig. 1-5. Another example is the Ge – Si a state consisting of two phases with com-
phase diagram in the lowest panel of Fig. positions at P1 and Q1 and with relative
1-6 (Hansen, 1958). In the upper three pan- proportions given by the lever rule. Since
els of Fig. 1-6, the molar Gibbs energies of the tangent line P1 Q1 lies below both g s and
the solid and liquid phases, g s and g l, at g l, this two-phase state is more stable than
three temperatures are shown to scale. As either phase alone. Furthermore, no other
illustrated in the top panel, g s varies with line joining any point on g l to any point on
composition between the standard molar g s lies below the line P1 Q1 . Hence, this line
Gibbs energies of pure solid Ge and of pure represents the true equilibrium state of the
0 (s)
solid Si, g Ge and g 0Si(s), while g l varies system, and the compositions P1 and Q1 are
between the standard molar Gibbs energies the liquidus and solidus compositions at
0 (l)
of the pure liquid components g Ge and 1300 °C.
0 (l)
g Si . As T is decreased to 1100 °C, the points
0 (l)
The difference between g Ge and g 0Si(s) is of common tangency are displaced to
equal to the standard molar Gibbs energy higher concentrations of Ge. For T < 937 °C,
of fusion (melting) of pure Si, Dg f0(Si) = g s < g l at all compositions.
(g 0Si(l) – g 0Si(s) ). Similarly, for Ge, Dg f0(Ge) = It was shown in Fig. 1-4 that if a tangent
0 (l) 0 (s)
(g Ge – g Ge ). The Gibbs energy of fusion of is drawn to a Gibbs energy curve, then the
a pure component may be written as: intercept of this tangent on the axis at Xi = 1
is equal to the partial Gibbs energy or
Dgf0 = Dhf0 – T Dsf0 (1-60)
chemical potential g i of component i. The
where Dhf0 and Dsf0 are the standard molar common tangent construction of Fig. 1-6
enthalpy and entropy of fusion. thus ensures that the chemical potentials of
1.5 Binary Phase Diagrams 17
Ge and Si are equal in the solid and liquid If we rearrange Eq. (1-61), subtracting
phases at equilibrium. That is: the Gibbs energy of fusion of pure Ge,
l s Dg f0(Ge) = (g Ge
0 (l) 0 (s)
– g Ge ), from each side, we
g Ge = gGe (1-61)
get:
l s
g Si =g Si (1-62) l 0 (l) s 0 (s)
(g Ge – g Ge ) – (gGe – g Ge )
This equality of chemical potentials was 0 (l)
= – (g Ge 0 (s)
– g Ge ) (1-63)
shown in Sec. 1.4.2 to be the criterion for
phase equilibrium. That is, the common Using Eq. (1-32), we can write Eq. (1-63)
tangent construction simultaneously mini- as:
mizes the total Gibbs energy and ensures l
Dg Ge s
– DgGe = – Dg f0(Ge) (1-64)
the equality of the chemical potentials,
thereby showing that these are equivalent or
l s
criteria for equilibrium between phases. R T ln aGe – R T ln aGe = – Dg f0(Ge) (1-65)
18 1 Thermodynamics and Phase Diagrams of Materials
l
where aGe is the activity of Ge (with re- Dgf0 = 0 when T = Tf0. Calculated phase dia-
spect to pure liquid Ge as standard state) in grams for Dsf0 = 3, 10 and 30 J/mol K are
the liquid solution on the liquidus, and aGes
shown in Fig. 1-7. A value of Dsf0 ≈ 10 is
is the activity of Ge (with respect to pure typical of most metals. However, when the
solid Ge as standard state) in the solid solu- components are ionic compounds such as
tion on the solidus. Starting with Eq. (1- ionic oxides, halides, etc., Dsf0 can be sig-
62), we can derive a similar expression for nificantly larger since there are several
the other component: ions per formula unit. Hence, two-phase
l s “lenses” in binary ionic salt or oxide phase
R T ln aSi – R T ln aSi = – Dg f0(Si) (1-66)
diagrams tend to be “fatter” than those
Eqs. (1-65) and (1-66) are equivalent to the encountered in alloy systems. If we are
common tangent construction. considering vapor – liquid equilibria rather
It should be noted that absolute values of than solid – liquid equilibria, then the
Gibbs energies cannot be defined. Hence, shape is determined by the entropy of
0 (l) 0 (l)
the relative positions of g Ge and g Si in vaporization, Dsv0 . Since Dsv0 is usually an
Fig. 1-6 are completely arbitrary. However, order of magnitude larger than Dsf0 , two-
this is immaterial for the preceding discus- phase (liquid + vapor) lenses tend to be
sion, since displacing both g 0Si(l) and g 0Si(s) by
0 (l)
the same arbitrary amount relative to g Ge
0 (s)
and g Ge will not alter the compositions of
the points of common tangency.
It should also be noted that in the present
discussion of equilibrium phase diagrams
we are assuming that the physical dimen-
sions of the single-phase regions in the
system are sufficiently large that surface
(interfacial) energy contributions to the
Gibbs energy can be neglected. For very
fine grain sizes in the sub-micron range,
however, surface energy effects can notice-
ably influence the phase boundaries.
The shape of the two-phase (solid + liq-
uid) “lens” on the phase diagram is deter-
mined by the Gibbs energies of fusion,
Dgf0 , of the components and by the mixing
terms, Dg s and Dg l. In order to observe
how the shape is influenced by varying
Dgf0 , let us consider a hypothetical system
A– B in which Dg s and Dg l are ideal Raoul-
tian (Eq. (1-45)). Let T f0(A) = 800 K and
T f0(B) = 1200 K. Furthermore, assume that
Figure 1-7. Phase diagram of a system A – B with
the entropies of fusion of A and B are equal
ideal solid and liquid solutions. Melting points of A
and temperature-independent. The enthalp- and B are 800 and 1200 K, respectively. Diagrams
ies of fusion are then given from Eq. (1-60) are calculated for entropies of fusion ∆Sf(A)
0
= ∆Sf(B)
0
=
by the expression Dhf0 = Tf0 Dsf0 since 3, 10 and 30 J/mol K.
1.5 Binary Phase Diagrams 19
very wide. For equilibria between two solid common tangent construction then gives
solutions of different crystal structure, the the equilibrium vapor and liquid composi-
shape is determined by the entropy of tions. The phase diagram depends upon the
solid – solid transformation, which is usu- Gibbs energies of vaporization of the com-
ally smaller than the entropy of fusion by ponents Dg v (Zn) and Dg v (Mg) as shown in
approximately an order of magnitude. Fig. 1-8.
Therefore two-phase (solid + solid) lenses To generate the isothermal pressure –
tend to be very narrow. composition (P – X ) phase diagram in the
lower panel of Fig. 1-8 we require the
1.5.3 Pressure – Composition Phase Gibbs energies of vaporization as functions
Diagrams of P. Assuming monatomic ideal vapors
and assuming that pressure has negligible
Let us consider liquid – vapor equilib-
effect upon the Gibbs energy of the liquid,
rium with complete miscibility, using as an
we can write:
example the Zn – Mg system. Curves of gv
and g l can be drawn at any given T and P, Dg v(i) = Dg0v(i) + R T ln P (1-67)
as in the upper panel of Fig. 1-8, and the
where Dg v(i) is the standard Gibbs energy
of vaporization (when P = 1 bar), which is
given by:
Dg0v(i) = Dh0v(i) – T Ds0v(i) (1-68)
lowing general equation, which gives the that the solid state is the stable state for
effect of pressure upon the Gibbs energy both pure components. However, in this
change for the transformation of one mole system gE (l) < gE (s), so that g s presents a
of pure component i from an a-phase to a flatter curve than does g l and there exists a
b-phase: central composition region in which gl < gs.
P Hence, there are two common tangent
b
Dga→ b = Dga0 → b + ∫ (vi − via ) dP (1- 70 ) lines, P1 Q1 and P2 Q2 . Such a situation
P =1 gives rise to a phase diagram with a mini-
where Dg 0aÆ b is the standard (P = 1 bar) mum in the two-phase region, as observed
Gibbs energy of transformation, and vib and in the Na2CO3 – K 2CO3 system (Dessu-
via are the molar volumes. reault et al., 1990) shown in Fig. 1-10. At a
composition and temperature correspond-
ing to the minimum point, liquid and solid
1.5.4 Minima and Maxima
of the same composition exist in equilib-
in Two-Phase Regions
rium.
As discussed in Sec. 1.4.8, the Gibbs en- A two-phase region with a minimum
ergy of mixing Dg m may be expressed as point as in Fig. 1-10 may be thought of as a
the sum of an ideal term Dgmideal and an ex- two-phase “lens” which has been “pushed
cess term gE. As has just been shown in down” by virtue of the fact that the liquid is
Sec. 1.5.2, if Dg ms and Dg ml
for the solid relatively more stable than the solid. Ther-
and liquid phases are both ideal, then a modynamically, this relative stability is ex-
“lens-shaped” two-phase region always re- pressed as gE (l) < gE (s).
sults. However in most systems even ap- Conversely, if gE (l) > gE (s) to a sufficient
proximately ideal behavior is the exception extent, then a two-phase region with a
rather than the rule. maximum will result. Such maxima in (liq-
Curves of g s and gl for a hypothetical uid + solid) or (solid + solid) two-phase re-
system A– B are shown schematically in gions are nearly always associated with the
Fig. 1-9 at a constant temperature (below existence of an intermediate phase, as will
the melting points of pure A and B) such be discussed in Sec. 1.5.10.
Figure 1-9. Isothermal Gibbs energy-composition Figure 1-10. Phase diagram of the K2CO3– Na2CO3
curves for solid and liquid phases in a system A – B in system at P = 1 bar (Dessureault et al., 1990).
which g E(l) > g E(s). A phase diagram of the type of
Fig. 1-10 results.
1.5 Binary Phase Diagrams 21
Below the critical temperature the curve gap at any temperature. Suppose that gE (s)
s
of Dgm exhibits two inflection points, indi- is so positve that Tc is higher than the min-
cated by the letter “s” in Fig. 1-11. These imum in the (solid + liquid) region. The re-
are known as the spinodal points. On the sult will be a phase diagram such as that of
phase diagram their locus traces out the the MgO – CaO system shown in Fig. 1-12
spinodal curve (Fig. 1-11). The spinodal (Doman et al., 1963; Wu, 1990).
curve is not part of the equilibrium phase The lower panel of Fig. 1-12 shows the
diagram, but it is important in the kinetics Gibbs energy curves at 2450 °C. The two
of phase separation, as discussed in Chap- common tangents define two two-phase re-
ter 6 (Binder and Fratzl, 2001). gions. As the temperature is decreased be-
low 2450 °C, the g s curve descends relative
to g l and the two points of tangency P1
1.5.6 Simple Eutectic Systems
and P2 approach each other until, at T =
The more positive gE is in a system, the 2374 °C, P1 and P2 become coincident at
higher is Tc and the wider is the miscibility the composition E. That is, at T = 2374 °C
there is just one common tangent line con- the liquid has solidified, and during the re-
tacting the two portions of the g s curve at action the compositions of the three phases
compositions A and B and contacting the gl will remain fixed at A, B and E. For this
curve at E. This temperature is known as reason the eutectic reaction is called an in-
the eutectic temperature, TE , and the com- variant reaction. More details on eutectic
position E is the eutectic composition. For solidification may be found in Chapter 2
temperatures below TE , gl lies completely (Müller-Krumbhaar et al., 2001).
above the common tangent to the two por-
tions of the g s curve and so for T < TE a
1.5.7 Regular Solution Theory
solid – solid miscibility gap is observed.
The phase boundaries of this two-phase re- Many years ago Van Laar (1908) showed
gion are called the solvus lines. The word that the thermodynamic origin of a great
eutectic is from the Greek for “to melt many of the observed features of binary
well” since the system has its lowest melt- phase diagrams can be illustrated at least
ing point at the eutectic composition E. qualitatively by simple regular solution
This description of the thermodynamic theory. A simple regular solution is one for
origin of simple eutectic phase diagrams is which:
strictly correct only if the pure solid com-
gE = X A X B ( w – h T ) (1-72)
ponents A and B have the same crystal
structure. Otherwise, a curve for g s which where w and h are parameters independent
is continuous at all compositions cannot be of temperature and composition. Substitut-
drawn. ing Eq. (1-72) into Eq. (1-29) yields, for
Suppose a liquid MgO – CaO solution of the partial properties:
composition XCaO = 0.52 (composition P1) (1-73)
E 2 E 2
is cooled from the liquid state very slowly gA = XB (w – h T ) , gB = XA (w – h T )
under equilibrium conditions. At 2450 °C Several liquid and solid solutions con-
the first solid appears with composition Q1 . form approximately to regular solution be-
As T decreases further, solidification con- havior, particularly if gE is small. Examples
tinues with the liquid composition follow- may be found for alloys, molecular solu-
ing the liquidus curve from P1 to E and the tions, and ionic solutions such as molten
composition of the solid phase following salts and oxides, among others. (The very
the solidus curve from Q1 to A. The rela- low values of gE observed for gaseous solu-
tive proportions of the two phases at any T tions generally conform very closely to Eq.
are given by the lever rule. At a tempera- (1-72).)
ture just above TE , two phases are ob- To understand why this should be so, we
served: a solid of composition A and a liq- only need a very simple model. Suppose
uid of composition E. At a temperature just that the atoms or molecules of the compo-
below TE , two solids with compositions A nents A and B mix substitutionally. If the
and B are observed. Therefore, at TE , dur- atomic (or molecular) sizes and electronic
ing cooling, the following binary eutectic structures of A and B are similar, then the
reaction occurs: distribution will be nearly random, and the
liquid Æ solid1 + solid2 (1-71) configurational entropy will be nearly
ideal. That is:
Under equilibrium conditions the tempera-
ture will remain constant at T = TE until all gE ≈ Dhm – T S E (non-config) (1-74)
24 1 Thermodynamics and Phase Diagrams of Materials
More will be said on this point in Sec. tively weak, then the solution is rendered
1.10.5. less stable, (w – h T ) > 0 and gE > 0.
We now assume that the bond energies Simple non-polar molecular solutions
eAA , e BB and eAB of nearest-neighbor pairs and ionic solutions such as molten salts of-
are independent of temperature and com- ten exhibit approximately regular behavior.
position and that the average nearest- The assumption of additivity of the energy
neighbor coordination number, Z, is also of pair bonds is probably reasonably realis-
constant. Finally, we assume that the en- tic for van der Waals or coulombic forces.
thalpy of mixing results mainly from the For alloys, the concept of a pair bond is, at
change in the total energy of nearest-neigh- best, vague, and metallic solutions tend to
bor pair bonds. exhibit larger deviations from regular be-
In one mole of solution there are (N 0 havior.
Z / 2) neareast-neighbor pair bonds, where In several solutions it is found that
N 0 is Avogadro’s number. Since the distri- | h T | < | w | in Eq. (1-72). That is, gE ≈ Dhm
bution is assumed random, the probability = w XA XB , and to a first approximation
that a given bond is an A–A bond is equal gE is independent of T. This is more often
to X A2 . The probabilities of B – B and A– B the case in non-metallic solutions than in
bonds are, respectively, X B2 and 2 XA XB . metallic solutions.
The molar enthalpy of mixing is then equal
to the sum of the energies of the nearest-
1.5.8 Thermodynamic Origin
neighbor bonds in one mole of solution,
of Simple Phase Diagrams Illustrated
minus the energy of the A–A bonds in XA
by Regular Solution Theory
moles of pure A and the energy of the B – B
bonds in XB moles of pure B: Figure 1-13 shows several phase dia-
grams, calculated for a hypothetical system
Dhm = (N 0 Z /2)
A– B containing a solid and a liquid phase
¥ (X A2 eAA + X B2 e BB + 2 XAB eAB ) with melting points of T f0(A) = 800 K and
– (N Z /2) (XA eAA) – (N 0 Z /2) (XB e BB )
0
T f0(B) = 1200 K and with entropies of fusion
= (N 0 Z ) [eAB – (eAA + e BB )/2] XA XB of both A and B set to 10 J/mol K, which is
= w XA XB (1-75) a typical value for metals. The solid and
liquid phases are both regular with temper-
We now define sAB , sAA and s BB as the ature-independent excess Gibbs energies
vibrational entropies of nearest-neighbor
pair bonds. Following an identical argu- gE (s) = w s XA XB and gE (l) = w l XA XB
ment to that just presented for the bond The parameters w s and w l have been varied
energies we obtain: systematically to generate the various pan-
s E (non-config) (1-76) els of Fig. 1-13.
0 In panel (n) both phases are ideal. Panels
= (N Z ) [sAB – (sAA + s BB )/2] = h XA XB
(l) to (r) exhibit minima or maxima de-
Eq. (1-72) has thus been derived. If A– B pending upon the sign and magnitude of
bonds are stronger than A–A and B – B (gE (l) – gE (s) ), as has been discussed in Sec.
bonds, then (eAB – h AB T ) < [(eAA – h AA T )/2 1.5.4. In panel (h) the liquid is ideal but
+ (e BB – h BB T ) / 2]. Hence, (w – h T ) < 0 positive deviations in the solid give rise to
and gE < 0. That is, the solution is rendered a solid – solid miscibility gap as discussed
more stable. If the A– B bonds are rela- in Sec. 1.5.6. On passing from panel (h) to
1.5 Binary Phase Diagrams 25
Figure 1-13. Topological changes in the phase diagram for a system A–B with regular solid and liquid phases,
brought about by systematic changes in the regular solution parameters ω s and ω l. Melting points of pure A and
B are 800 K and 1200 K. Entropies of fusion of both A and B are 10.0 J/mol K (Pelton and Thompson, 1975).
The dashed curve in panel (d) is the metastable liquid miscibility gap (Reprinted from Pelton, 1983).
panel (c), an increase in gE (s) results in a sultant lowering of the eutectic tempera-
widening of the miscibility gap so that the ture.
solubility of A in solid B and of B in solid Eutectic phase diagrams are often drawn
A decreases. Panels (a) to (c) illustrate that with the maximum solid solubility occur-
negative deviations in the liquid cause a ring at the eutectic temperature (as in Fig.
relative stabilization of the liquid with re- 1-12). However, panel (d) of Fig. 1-13, in
26 1 Thermodynamics and Phase Diagrams of Materials
which the maximum solubility of A in the by sufficiently rapid cooling, then a meta-
B-rich solid solution occurs at approxi- stable liquid – liquid miscibility gap is ob-
mately T = 950 K, illustrates that this need served as shown in Fig. 1-13 (d). For exam-
not be the case even for simple regular so- ple, in the Na2O – SiO2 system the flattened
lutions. (or “S-shaped”) SiO2 liquidus heralds the
existence of a metastable miscibility gap of
importance in glass technology.
1.5.9 Immiscibility – Monotectics
In Fig. 1-13 (e), positive deviations in
1.5.10 Intermediate Phases
the liquid have given rise to a liquid – liquid
miscibility gap. The CaO – SiO2 system The phase diagram of the Ag – Mg
(Wu, 1990), shown in Fig. 1-14, exhibits system (Hultgren et al., 1973) is shown in
such a feature. Suppose that a liquid of Fig. 1-15 (d). An intermetallic phase, b¢, is
composition XSiO2 = 0.8 is cooled slowly seen centered approximately about the
from high temperatures. At T = 1815 °C the composition XMg = 0.5. The Gibbs energy
miscibility gap boundary is crossed and a curve at 1050 K for such an intermetallic
second liquid layer appears with a compo- phase has the form shown schematically in
sition of XSiO2 = 0.97. As the temperature is Fig. 1-15 (a). The curve g b¢ rises quite rap-
lowered further, the composition of each idly on either side of its minimum, which
liquid phase follows its respective phase occurs near XMg = 0.5. As a result, the b¢
boundary until, at 1692 °C, the SiO2-rich phase appears on the phase diagram only
liquid has a composition of XSiO2 = 0.99 over a limited composition range. This
(point B), and in the CaO-rich liquid form of the curve g b¢ results from the fact
XSiO2 = 0.74 (point A). At any temperature, that when XAg ≈ XMg a particularly stable
the relative amounts of the two phases are crystal structure exists in which Ag and Mg
given by the lever rule. atoms preferentially occupy different sites.
At 1692 °C the following invariant bi- The two common tangents P1 Q1 and P2 Q2
nary monotectic reaction occurs upon cool- give rise to a maximum in the two-phase
ing: (b¢ + liquid) region of the phase diagram.
(Although the maximum is observed very
Liquid B Æ Liquid A + SiO2 (solid) (1-77)
near XMg = 0.5, there is no thermodynamic
The temperature remains constant at reason for the maximum to occur exactly at
1692 °C and the compositions of the phases this composition.)
remain constant until all of liquid B is con- Another intermetallic phase, the e phase,
sumed. Cooling then continues with pre- is also observed in the Ag – Mg system,
cipitation of solid SiO2 with the equilib- Fig. 1-15. The phase is associated with a
rium liquid composition following the liq- peritectic invariant ABC at 744 K. The
uidus from point A to the eutectic E. Gibbs energy curves are shown schemati-
Returning to Fig. 1-13, we see in panel cally at the peritectic temperature in Fig.
(d) that the positive deviations in the liquid 1-15 (c). One common tangent line can be
in this case are not large enough to produce drawn to gl, g b¢ and g e.
immiscibility, but they do result in a flat- Suppose that a liquid alloy of composi-
tening of the liquidus, which indicates a tion XMg = 0.7 is cooled very slowly from
“tendency to immiscibility”. If the nuclea- the liquid state. At a temperature just above
tion of the solid phases can be suppressed 744 K a liquid phase of composition C and
1.5 Binary Phase Diagrams 27
Figure 1-14. CaO – SiO2 phase diagram at P = 1 bar (after Wu, 1990) and Gibbs energy curves at 1500 °C illus-
trating Gibbs energies of fusion and formation of the stoichiometric compound CaSiO3 .
28 1 Thermodynamics and Phase Diagrams of Materials
mum in Fig. 1-15 (d). Because the CaSiO3 Eq. (1-78), to form a liquid of composition
single-phase region is so narrow, we refer C and another solid phase, b¢, of composi-
to CaSiO3 as a stoichiometric compound. tion A.
Any deviation in composition from the Another example of an incongruent
stoichiometric 1 : 1 ratio of CaO to SiO2 compound is Ca3Si2O7 in Fig. 1-14, which
results in a very large increase in Gibbs melts incongruently (or peritectically) to
energy. form liquid and Ca2SiO4 at the peritectic
The e phase in Fig. 1-15 is based on the temperature of 1469 °C.
stiochiometry AgMg3 . The Gibbs energy An incongruent compound is always as-
curve, Fig. 1-15 (c), rises extremely rapidly sociated with a peritectic. However, the
on the Ag side of the minimum, but some- converse is not necessarily true. A peritec-
what less steeply on the Mg side. As a re- tic is not always associated with an inter-
sult, Ag is virtually insoluble in AgMg3 , mediate phase. See, for example, Fig. 1-
while Mg is sparingly soluble. Such a 13 (i).
phase with a narrow range of homogeneity For purposes of phase diagram calcula-
is often called a non-stoichiometric com- tions involving stoichiometric compounds
pound. At low temperatures the b¢ phase such as CaSiO3 , we may, to a good approx-
exhibits a relatively narrow range of stoi- imation, consider the Gibbs energy curve,
chiometry about the 1 : 1 AgMg composi- g 0.5 (CaSiO3 ) , to have zero width. All that
tion and can properly be called a com- is then required is the value of g 0.5 (CaSiO3 )
pound. However, at higher temperatures it at the minimum. This value is usually
is debatable whether a phase with such a expressed in terms of the Gibbs energy
wide range of composition should be called of fusion of the compound, Dg 0f (0.5 CaSiO3 )
a “compound”. or the Gibbs energy of formation
From Fig. 1-14 it can be seen that if stoi- Dg 0form (0.5 CaSiO3 ) of the compound from
chiometric CaSiO3 is heated it will melt the pure solid components CaO and SiO2
isothermally at 1540 °C to form a liquid of according to the reaction: 0.5 CaO (sol) +
the same composition. Such a compound is 0.5 SiO2 (sol) = 0.5 CaSiO3 (sol). Both these
called congruently melting or simply a con- quantities are interpreted graphically in
gruent compound. The compound Ca2SiO4 Fig. 1-14.
in Fig. 1-14 is congruently melting. The b¢
phase in Fig. 1-15 is also congruently melt-
ing at the composition of the liquidus/sol- 1.5.11 Limited Mutual Solubility –
idus maximum. Ideal Henrian Solutions
It should be noted with regard to the con-
gruent melting of CaSiO3 in Fig. 1-14 that In Sec. 1.5.6, the region of two solids in
the limiting slopes dT /dX of both branches the MgO – CaO phase diagram of Fig. 1-12
of the liquidus at the congruent melting was described as a miscibility gap. That is,
point (1540 °C) are zero since we are really only one continuous g s curve was assumed.
dealing with a maximum in a two-phase re- If, somehow, the appearance of the liquid
gion. phase could be suppressed, then the two
The AgMg3 (e) compound in Fig. 1-15 is solvus lines in Fig. 1-12, when projected
said to melt incongruently. If solid AgMg3 upwards, would meet at a critical point
is heated it will melt isothermally at 744 K above which one continuous solid solution
by the reverse of the peritectic reaction, would exist at all compositions.
30 1 Thermodynamics and Phase Diagrams of Materials
Henrian activity coefficients can usually This is the Gibbs Phase Rule. F is called
be expressed as functions of temperature: the number of degrees of freedom or vari-
ance of the system and is the number of pa-
R T ln g i0 = a – b T (1-83)
rameters which can and must be specified
where a and b are constants. If data are lim- in order to completely specify the state of
ited, it can further be assumed that b ≈ 0 so the system.
that R T ln g i0 ≈ constant. Binary temperature – composition phase
diagrams are plotted at a fixed pressure,
usually 1 bar. This then eliminates one de-
1.5.12 Geometry of Binary Phase gree of freedom. In a binary system, C = 2.
Diagrams Hence, for binary isobaric T – X diagrams
The geometry of all types of phase dia- the phase rule reduces to:
grams of any number of components is F=3–P (1-86)
governed by the Gibbs Phase Rule.
Consider a system with C components in Binary T – X diagrams contain single-
which P phases are in equilibrium. The phase areas and two-phase areas. In the sin-
system is described by the temperature, the gle-phase areas, F = 3 – 1 = 2. That is, tem-
total pressure and the composition of each perature and composition can be specified
phase. In a C-component system, (C – 1) in- independently. These regions are thus
dependent mole fractions are required to called bivariant. In two-phase regions,
describe the composition of each phase F = 3 – 2 = 1. If, say, T is specified, then the
(because S Xi = 1). Hence, the total number compositions of both phases are deter-
of variables required to describe the system mined by the ends of the tie-lines. Two-
is [P (C – 1) + 2]. However, as shown in Sec. phase regions are thus termed univariant.
1.4.2, the chemical potential of any compo- Note that the overall composition can be
nent is the same in all phases (a, b, g, …) varied within a two-phase region at con-
since the phases are in equilibrium. That is: stant T, but the overall composition is not a
parameter in the sense of the phase rule.
gia (T, P, X1a , X2a , X3a , …) Rather, it is the compositions of the indi-
= gib (T, P, X1b, X2b, X3b, …) vidual phases at equilibrium that are the
parameters to be considered in counting the
= gig(T, P, X1g, X2g, X3g, …) = … (1-84)
number of degrees of freedom.
where gia (T, P, X1a , X2a , X3a , …) is a func- When three phases are at equilibrium in
tion of temperature, of total pressure, and a binary system at constant pressure,
of the mole fractions X1a , X2a , X3a , … in F = 3 – 3 = 0. Hence, the compositions of
the a phase; and similarly for the other all three phases, as well as T, are fixed.
phases. Thus there are C (P – 1) indepen- There are two general types of three-phase
dent equations in Eq. (1-84) relating the invariants in binary phase diagrams. These
variables. are the eutectic-type and peritectic-type
Let F be the differences between the invariants as illustrated in Fig. 1-16. Let
number of variables and the number of the three phases concerned be called a, b
equations relating them: and g, with b as the central phase as shown
in Fig. 1-16. The phases a, b and g can be
F = P (C – 1) + 2 – C (P – 1) solid, liquid or gaseous. At the eutectic-
F=C–P+2 (1-85) type invariant, the following invariant re-
32 1 Thermodynamics and Phase Diagrams of Materials
Figure 1-17. Fe – Mo
phase diagram at P = 1 bar
(Kubaschewski, 1982).
shown in Fig. 1-17. The invariants in this and 2051 °C; a monotectic at 1692 °C; and
system are peritectics at 1540, 1488 and a peritectic at 1469 °C. The compound
1450 °C; eutectoids at 1235 and 1200 °C; Ca3SiO5 dissociates upon heating to CaO
peritectoids at 1370 and 950 °C. The two- and Ca2SiO4 by a peritectoid reaction at
phase (liquid + g) region passes through a 1789 °C and dissociates upon cooling to
minimum at XMo = 0.2. CaO and Ca2SiO4 by a eutectoid reaction at
Between 910 °C and 1390 °C is a two- 1250 °C. Maxima are observed at 2130 and
phase (a + g) g-loop. Pure Fe adopts the 1540 °C. At 1470 °C there is an invariant
f.c.c. g structure between these two temper- associated with the tridymite Æ cristobalite
atures but exists as the b.c.c. a phase at transition of SiO2 . This is either a peritec-
higher and lower temperatures. Mo, how- tic or a catatectic depending upon the rela-
ever, is more soluble in the b.c.c. than tive solubility of CaO in tridymite and cris-
0 (b.c.c.-Fe)
in the f.c.c. structure. That is, g Mo tobalite. However, these solubilities are
0 (f.c.c.-Fe)
< g Mo as discussed in Sec. 1.5.11. very small and unknown.
Therefore, small additions of Mo stabilize
the b.c.c. structure.
In the CaO – SiO2 phase diagram, Fig.
1-14, we observe eutectics at 1439, 1466
34 1 Thermodynamics and Phase Diagrams of Materials
where i = LiF or NaF. Along the LiF-rich aLiF = 1 with respect to pure solid LiF as
liquidus, the liquid is in equilibrium with standard state. In the Henrian solution at
s
essentially pure solid LiF. Hence, X LiF =1 saturation,
E (s)
and gLif = 0. Eq. (1-99) then reduces to:
aLiF = g 0LiF XLiF = g 0LiF (1 – 0.915) = 1
l E (l) 0
R T ln X LiF +g Lif = – Dg f (LiF) (1-100)
Hence, the Henrian activity coefficient
l
From experimental values of X LiF on the in the NaF-rich solid solution at 649 °C
liquidus and with Eq. (1-97) for Dg f0(LiF) , is g 0LiF = 11.76. Since no solubilities have
E (l)
values of gLif at the measured liquidus been measured at other temperatures, we
points can be calculated from Eq. (1-100). assume that:
Along the NaF-rich solidus the solid so-
lution is sufficiently concentrated in NaF R T ln g 0LiF = R (922) ln (11.76) (1-104)
that Henrian behavior (Sec. 1.5.11) can be = 18 900 J/mol = constant
E (s)
assumed. That is, for the solvent, gNaF = 0. Using the notation of Eq. (1-82):
Hence, Eq. (1-99) becomes:
(l) (s) E (l)
g 0LiF
(s, NaF)
= g 0LiF
(s)
+ 18 900 J/mol (1-105)
R T ln X NaF – R T ln X NaF + gNaF
= – Dg f0(NaF) (1-101) where g 0LiF
(s)
is the standard Gibbs energy of
solid LiF, and g 0LiF (s, NaF)
is the hypothetical
Thus, from the experimental liquidus and standard Gibbs energy of LiF dissolved in
solidus compositions and with the Gibbs solid NaF.
energy of fusion from Eq. (1-98), values of The phase diagram drawn in Fig. 1-18
E (l)
gNaF can be calculated at the measured liq- was calculated from Eqs. (1-97) to (1-104).
uidus points from Eq. (1-101). Complete details of the analysis of the
Finally, enthalpies of mixing, hE , in the LiF – NaF system are given by Sangster and
liquid have been measured by calorimetry Pelton (1987).
by Hong and Kleppa (1976). As a second example of thermodynamic/
Combining all these data in a least- phase diagram optimization, consider the
squares optimization, the following expres- Cd – Na system. The phase diagram, with
sions for the liquid were obtained by Sang- points measured by several authors (Math-
ster and Pelton (1987): ewson, 1906; Kurnakow and Kusnetzow,
hE (l) = XLiF XNaF (1-102) 1907; Weeks and Davies, 1964) is shown in
Fig. 1-19.
¥ [– 7381 + 184 (XNaF – XLiF )] J/mol From electromotive force measurements
s E (l) = XLiF XNaF (1-103) on alloy concentration cells, several au-
¥ [– 2.169 – 0.562 (XNaF – XLiF )] J/mol thors have measured the activity coeffi-
cient of Na in liquid alloys. The data
Eqs. (1-102) and (1-103) then permit all are shown in Fig. 1-20 at 400 °C. From
other integral and partial properties of the the temperature dependence of g ENa =
liquid to be calculated. R T ln g Na , the partial enthalpy of Na in the
For the NaF-rich Henrian solid solution, liquid was obtained via Eq. (1-52). The re-
the solubility of LiF has been measured by sults are shown in Fig. 1-21. Also, hE of the
Holm (1965) at the eutectic temperature liquid has been measured by Kleinstuber
where the NaF-rich solid solution is in (1961) by direct calorimetry. These ther-
equilibrium with pure solid LiF. That is, modynamic data for g ENa , hENa and hE were
1.6 Application of Thermodynamics to Phase Diagram Analysis 37
Figure 1-19. Cd – Na phase diagram at P = 1 bar calculated from optimized thermodynamic parameters (Re-
printed from Pelton, 1988 a). 쑗 Kurnakow and Kusnetzow (1907), 왕 Mathewson (1906), × Weeks and Davies
(1964).
Figure 1-22. Projection of the liquidus surface of the Bi – Sn –Cd system onto the ternary composition triangle
(after Bray et al., 1961 – 1962). Small arrows show the crystallization path of an alloy of overall composition at
point a. (Reprinted from Pelton, 1996.)
The A- and B-liquidus surfaces in Fig. 1- pure Cd. (Note that this fact cannot be de-
23 intersect along the line e1 E. Liquids duced from Fig. 1-22 alone.) Therefore, as
with compositions along this line are there- solidification proceeds, the liquid becomes
fore in equilibrium with A-rich and B-rich depleted in Cd, but the ratio XSn /XBi in the
solid phases simultaneously. That is, P = 3 liquid remains constant. Hence, the compo-
and so F – 1. Such “valleys” are thus called sition path followed by the liquid (its crys-
univariant lines. The three univariant lines tallization path) is a straight line passing
meet at the ternary eutectic point E at through point a and projecting to the Cd-
which P = 4 and F = 0. This is an invariant corner of the triangle. This crystallization
point since the temperature and the compo- path is shown on Fig. 1-22 as the line a b.
sitions of all four phases in equilibrium are In the general case in which a solid solu-
fixed. tion rather than a pure component or stoi-
chiometric compound is precipitating, the
crystallization path will not be a straight
1.7.3 Polythermal Projections
line. However, for equilibrium cooling, a
of Liquidus Surfaces
straight line joining a point on the crystal-
A two-dimensional representation of the lization path at any T to the overall compo-
ternary liquidus surface may be obtained as sition point a will extend through the com-
an orthogonal projection upon the base position, on the solidus surface, of the solid
composition triangle. Such a polythermal phase in equilibrium with the liquid at that
projection of the liquidus of the Bi – Sn – Cd temperature.
system (Bray et al., 1961 – 62) is shown in When the composition of the liquid has
Fig. 1-22. This is a simple eutectic ternary reached point b in Fig. 1-22 at T ≈ 435 K,
system with a space model like that shown the relative proportions of the solid Cd and
in Fig. 1-23. The constant temperature liquid phases at equilibrium are given by
lines on Fig. 1-22 are called liquidus iso- the lever rule applied to the tie-line d a b:
therms. The univariant valleys are shown (moles of liquid)/(moles of Cd) = d a /a b.
as heavier lines. By convention, the large Upon further cooling, the liquid composi-
arrows indicate the directions of decreas- tion follows the univariant valley from b to
ing temperature along these lines. E while Cd and Sn-rich solids coprecipitate
Let us consider the sequence of events as a binary eutectic mixture. When the
occurring during the equilibrium cooling liquidus composition attains the ternary eu-
from the liquid of an alloy of overall com- tectic composition E at T ≈ 380 K the invar-
position a in Fig. 1-22. Point a lies within iant ternary eutectic reaction occurs:
the field of primary crystallization of Cd.
liquid Æ s1 + s2 + s3 (1-111)
That is, it lies within the composition re-
gion in Fig. 1-22 in which Cd-rich solid where s1 , s2 and s3 are the three solid
will be the first solid to precipitate upon phases and where the compositions of all
cooling. As the liquid alloy is cooled, the four phases (as well as T ) remain fixed un-
Cd-liquidus surface is reached at T ≈ 465 K til all liquid is solidified.
(slightly below the 473 K isotherm). A In order to illustrate several of the fea-
solid Cd-rich phase begins to precipitate at tures of polythermal projections of liquidus
this temperature. Now, in this particular surfaces, a projection of the liquidus of a
system, Bi and Sn are nearly insoluble in hypothetical system A – B – C is shown in
solid Cd, so that the solid phase is virtually Fig. 1-24. For the sake of simplicity, iso-
42 1 Thermodynamics and Phase Diagrams of Materials
therms are not shown, only the univariant field, since e is a congruent compound,
lines with arrows to show the directions of while the composition of the d phase lies
decreasing temperature. The binary sub- outside of its field since d is incongruent.
systems A – B and C – A are simple eutectic Similarly for the ternary compounds, h is a
systems, while the binary subsystem B – C congruently melting compound while z is
contains one congruent binary phase, e, incongruent. For the congruent compound
and one incongruent binary phase, d, as h, the highest temperature on the h liquidus
shown in the insert in Fig. 1-24. The letters occurs at the composition of h.
e and p indicate binary eutectic and peritec- The univariant lines meet at a number of
tic points. The e and d phases are called bi- ternary eutectics E (three arrows converg-
nary compounds since they have composi- ing), a ternary peritectic P (one arrow en-
tions within a binary subsystem. Two ter- tering, two arrows leaving the point), and
nary compounds, h and z, with composi- several ternary quasi-peritectics P¢ (two
tions within the ternary triangle, as indi- arrows entering, one arrow leaving). Two
cated in Fig. 1-24, are also found in this saddle points s are also shown. These are
system. All compounds, as well as pure points of maximum T along the univariant
solid A, B and C (the “a, b and g” phases), line but of minimum T on the liquidus sur-
are assumed to be stoichiometric (i.e., there face along a section joining the composi-
is no solid solubility). The fields of pri- tions of the two solids. For example, s1 is at
mary crystallization of all the solids are in- a maximum along the univariant E1 P¢3 , but
dicated in parentheses in Fig. 1-24. The is a minimum point on the liquidus along
composition of the e phase lies within its the straight line z s1 h.
1.7 Ternary and Multicomponent Phase Diagrams 43
Let us consider the events occurring dur- univariant valley until the liquid reaches
ing cooling from the liquid of an alloy of the peritectic composition P. The invariant
overall composition a in Fig. 1-24. The pri- ternary peritectic reaction then occurs iso-
mary crystallization product will be the e thermally:
phase. Since this is a pure stoichiometric
liquid + e + b Æ z (1-113)
solid the crystallization path of the liquid
will be along a straight line passing Since there are three reactants, there are
through a and extending to the composition three possible outcomes: (i) the liquid is
of e as shown in the figure. consumed before either e or b and solidifi-
Solidification of e continues until the cation terminates at P; (ii) e is consumed
liquid attains a composition on the univari- first, solidification then continues along
ant valley. Thereafter the liquid composi- the path P P¢3 ; or (iii) b is consumed first
tion follows the valley towards the point P¢1 and solidification continues along the path
in co-existence with e and z. At point P¢1 P P¢1 . Which outcome occurs depends on
the invariant ternary quasi-peritectic reac- whether the overall composition b lies
tion occurs isothermally: within the tie-triangle (i) e b z; (ii) b z P, or
(iii) e z P. In the example shown, outcome
liquid + e Æ d + z (1-112)
(i) will occur.
Since there are two reactants in a quasi-
peritectic reaction, there are two possible
1.7.4 Ternary Isothermal Sections
outcomes: (i) the liquid is completely con-
sumed before the e phase; in this case, so- Isothermal projections of the liquidus
lidification will be complete at the point P¢1 ; surface do not provide information on the
(ii) e is completely consumed before the compositions of the solid phases at equilib-
liquid. In this case, solidification will con- rium. However, this information can be
tinue with decreasing T along the univari- presented at any one temperature on an iso-
ant line P¢1 E1 with co-precipitation of d and thermal section such as that shown for the
z until, at E, the liquid will solidify eutecti- Bi – Sn – Cd system at 423 K in Fig. 1-25.
cally (liquid Æ d + z + h). To determine This phase diagram is a constant tempera-
whether outcome (i) or (ii) occurs, we use ture slice through the space model of Fig.
the mass balance criterion that, for three- 1-23.
phase equilibrium, the overall composition The liquidus lines bordering the one-
a must always lie within the tie-triangle phase liquid region of Fig. 1-25 are identi-
formed by the compositions of the three cal to the 423 K isotherms of the projection
phases. Now, the triangle joining the com- in Fig. 1-22. Point c in Fig. 1-25 is point c
positions of d, e and z does not contain the on the univariant line in Fig. 1-22. An alloy
point a, but the triangle joining the compo- with overall composition in the one-phase
sitions of d, z and liquid at P¢1 does contain liquid region of Fig. 1-25 at 423 K will
the point a. Hence, outcome (ii) occurs. consist of a single liquid phase. If the over-
An alloy of overall composition b in Fig. all composition lies within one of the two-
1-24 solidifies with e as primary crystal- phase regions, then the compositions of the
lization product until the liquid composi- two phases are given by the ends of the tie-
tion contacts the univariant line. There- line which passes through the overall com-
after, co-precipitation of e and b occurs position. For example, a sample with over-
with the liquid composition following the all composition p in Fig. 1-25 will consist
44 1 Thermodynamics and Phase Diagrams of Materials
Figure 1-25. Isothermal section of the Bi – Sn – Cd system at 423 K at P = 1 bar (after Bray et al., 1961–1962).
Extents of solid solubility in Bi and Sn have been exaggerated for clarity of presentation. (Reprinted from Pel-
ton, 1996.)
Figure 1-28. Isopleth (constant composition section) of the Bi – Sn – Cd system at P = 1 bar following the line
AB at XSn = 0.45 of Fig. 1-22. (Reprinted from Pelton, 1996).
ted with one or more compositional vari- at equilibrium for an overall composition
ables held constant. Hence these sections lying on the section. Tie-lines do not, in
are similar to the ternary isopleths dis- general, lie in the plane of the diagram, so
cussed in Sec. 1.7.5. In certain cases, sec- the diagram does not provide information
tions at constant chemical potential of one on the compositions or amounts of the
or more components (for example, at con- phases present. As a phase boundary is
stant oxygen partial pressure) can be use- crossed, one and only one phase appears or
ful. These are discussed in Sec. 1.8. disappears (Law of Adjoining Phase Re-
Two sections of the Fe – Cr –V– C system gions). If temperature is an axis, as in Fig.
(Lee and Lee, 1992) are shown in Figs. 1- 1-29, then horizontal invariants like the
29 and 1-30. The diagram in Fig. 1-29 is a line AB in Fig. 1-29 can appear. These can
T-composition section at constant Cr and V be considered as degenerate infinitely nar-
content, while Fig. 1-30 is a section at con- row phase fields of (C + 1) phases, where C
stant T = 850 oC and constant C content of is the number of components (for isobaric
0.3 wt.%. The interpretation and topologi- diagrams). For example in Fig. 1-29, on the
cal rules of construction of these sections line AB, five phases are present. Three or
are the same as those for ternary isopleths, four phase boundaries meet at intersection
as discussed in Sec. 1.7.5. In fact, the same points at which Schreinemakers’ Rule ap-
rules apply to a two-dimensional constant- plies. It is illustrated by the extrapolations
composition section for a system of any in Fig. 1-29 at points a, b and c and in Fig.
number of components. The phase fields 1-30 at points b, c, n, i and s (see discus-
on the diagram indicate the phases present sions in Sec. 1.9).
Figure 1-29. Section of the Fe – Cr – V – C system at 1.5 wt.% Cr and 0.1 wt.% V at P = 1 bar (Lee and Lee,
1992).
1.7 Ternary and Multicomponent Phase Diagrams 49
log pO2 – x diagram. This underscores the log pO2 – x diagram. For example, the (spi-
fact that every tie-line of a ternary isother- nel + (Fe, Cr) O + alloy) triangle with cor-
mal section corresponds to a constant chem- ners at points a, b and c corresponds to the
ical potential of each of the components. “eutectic-like” or eutecular invariant with
Another example of a log pO2 – x diagram the same phase compositions a, b and c at
is shown for the Fe – Cr – O system at log pO2 ≈ –10.7. We can see that within a
1573 K in the lower panel of Fig. 1-33 three-phase tie-triangle, pO2 is constant.
(Pelton and Schmalzried, 1973). The corre- An example of yet another kind of phase
sponding ternary isothermal section in diagram of this same geometrical type is
Jänecke coordinates is shown in the upper shown in Fig. 1-34. For the quaternary
panel. Each of the invariant three-phase Fe – Cr – O2 – SO2 system at T = 1273 K and
tie-triangles in the isothermal section at constant pSO2 = 10–7 bar, Fig. 1-34 is a
corresponds to an invariant line in the plot of log pO2 versus the molar metal ratio
Figure 1-34. Calculated phase diagram of log pO2 versus molar metal ratio at T = 1273.15 K and pSO2 = 10–7 bar
for the Fe – Cr – SO2 – O2 system.
x . Since log pO2 varies as –1/2 log pS2 when geometrical type. Often for ceramic
pSO2 and T are constant, Fig. 1-34 is also a systems, we encounter “binary” phase dia-
plot of log pS2 versus x . grams such as that for the “CaO – Fe2O3”
Plotting T versus x at constant pO2 in the system in Fig. 1-35, which has been taken
Fe – Cr – O system, or at constant pO2 and from Phillips and Muan (1958). How are
pSO2 in the Fe – Cr – SO2 – O2 system, will we to interpret such a diagram? How, for
also result in phase diagrams of this same instance, do we interpret the composition
Figure 1-37. Corresponding phase diagrams for the Fe – O system at PTOTAL = 1 bar (after Muan and Osborn,
1965).
1.8 Phase Diagrams with Potentials as Axes 55
diagrams (Figs. 1-32 to 1-34) are useful in of Figs. 1-1 to 1-3 are of this same geomet-
the study of hot corrosion, metallurgical rical type.
roasting processes, chemical vapor deposi- As yet another example of this geomet-
tion, and many aspects of materials pro- rical type of diagram, a plot of R T ln pO2
cessing. versus T for the Fe – O system is shown in
Another important geometrical type of Fig. 1-37 (b). Again, one-, two- and three-
phase diagram is exemplified by P – T phase regions are indicated by areas, lines
phase diagrams for one-component sys- and triple points respectively. Fig. 1-37 (a)
tems, as shown for H2O in Fig. 1-36. In is the binary T – composition phase diagram
such diagrams (see also Chapter 10 by for the Fe – O system. The correspondence
Kunz (2001)) bivariant single-phase re- between Figs. 1-37 (a) and 1-37 (b) is evi-
gions are indicated by areas, univariant dent. Each two-phase line of Fig. 1-37 (b)
two-phase regions by lines, and invariant “opens up” to a two-phase region of Fig.
three-phase regions by triple points. An 1-37 (a). Each tie-line of a two-phase re-
important rule of construction is the exten- gion in Fig. 1-37 (a) can thus be seen to
sion rule, which is illustrated by the broken correspond to a constant pO2 . Triple points
lines in Fig. 1-36. At a triple point, the in Fig. 1-37 (b) become horizontal invari-
extension of any two-phase line must pass ant lines in Fig. 1-37 (a).
into the single-phase region of the third Yet another type of phase diagram is
phase. Clearly, the predominance diagrams shown in Fig. 1-38. This is an isothermal
Figure 1-38. Phase diagram of log pS2 versus log pO2 at 1273 K and constant molar metal ratio nCr /(nFe + nCr) =
0.5 in the Fe – Cr – S2 – O2 system.
56 1 Thermodynamics and Phase Diagrams of Materials
section at constant molar metal ratio Table 1-1. Corresponding pairs of potentials f i and
nCr /(nFe + nCr ) = 0.5 for the Fe – Cr – S2 – O2 extensive variables qi .
system. This diagram was calculated ther- fi : T P m1 m 2 … mC
modynamically from model parameters.
qi : S –V n1 n2 … nC
The axes are the equilibrium sulfur and
oxygen partial pressures. Three or four
boundary lines can meet at an intersection
point. Some of the boundary lines on Fig. where U is the internal energy of the
1-38 separate a two-phase region (a + b) system. The corresponding potentials and
from another two-phase region (a + g). extensive variables are listed in Table 1-1.
These lines thus represent the conditions It may also be noted that the corresponding
for three-phase (a + b + g) equilibrium. pairs are found together in the terms of the
general Gibbs – Duhem equation:
S dT – V dP + S ni dm i = 0 (1-115)
1.9 General Phase Diagram
Geometry
1.9.1 General Geometrical Rules
Although the various phase diagrams for All True Phase Diagram Sections
shown in the preceding sections may ap-
The Law of Adjoining Phase Regions ap-
pear to have quite different geometries, it
plies to all true sections. As a phase boun-
can be shown that, in fact, all true phase di-
dary line is crossed, one and only one
agram sections obey the same set of geo-
phase either appears or disappears.
metrical rules. Although these rules do not
If the vertical axis is a potential (T, P,
apply directly to phase diagram projections
m i ), then horizontal invariant lines like the
such as Figs. 1-22, 1-24 and 1-31, such di-
eutectic line in Fig. 1-12 or the line AB in
agrams can be considered to consist of por-
Fig. 1-29 will be seen when the maximum
tions of several phase diagram sections
number of phases permitted by the phase
projected onto a common plane.
rule are at equilibrium. However, if these
By “true” phase diagram we mean one in
are considered to be degenerate infinitely
which each point of the diagram represents
narrow phase fields, then the Law of Ad-
one unique equilibrium state. In the present
joining Phase Regions still applies. This is
section we give the general geometrical
illustrated schematically in Fig. 1-39 where
rules that apply to all true phase diagram
the three-phase eutectic line has been
sections, and we discuss the choices of
“opened up”. Similarly, if both axes are po-
axes and constants that ensure that the dia-
tentials, then many phase boundaries may
gram is a true diagram.
be degenerate infinitely narrow regions.
We must first make some definitions. In
For example, all phase boundaries on Figs.
a system of C components we can define
1-1 to 1-3, 1-36 and 1-37 (b) are degenerate
(C + 2) thermodynamic potentials f i . These
two-phase regions which are schematically
are T, P, m 1 , m 2 , …, m C , where m i is the
shown “opened up” on Fig. 1-40.
chemical potential defined in Eq. (1-23).
All phase boundary lines in a true phase
For each potential there is a corresponding
diagram meet at nodes where exactly four
extensive variable qi related by:
lines converge, as in Fig. 1-41. N phases
f i = (∂U/∂qi )q j ( j ≠ i ) (1-114) (a1 , a2 , …, aN ) where N ≥ 1 are common
1.9 General Phase Diagram Geometry 57
ception to Schreinemakers’ Rule. How- Fig. 1-1, all lines are actually two coinci-
ever, the extremum in such a case is not ac- dent ZPF lines.
tually a node where four phase boundaries The ZPF line concept is very useful in
converge, but rather a point where two the development of general algorithms for
boundaries touch. Such extrema in which the thermodynamic calculation of phase di-
two phase boundaries touch with zero agrams as discussed in Sec. 1-12.
slope may occur for a C-phase region in a
phase diagram of a C-component system 1.9.2 Choice of Axes and Constants
when one axis is a potential. For example, of True Phase Diagrams
in an isobaric temperature – composition
In a system of C components, a two-
phase diagram of a four-component sys-
dimensional diagram is obtained by choos-
tem, we may observe a maximum or a min-
ing two axis variables and holding (C – 1)
imum in a four-phase region separating two
other variables constant. However, not all
three-phase regions. A similar maximum
choices of variables will result in a true
or minimum in a (C – n)-phase region,
phase diagram. For example on the P – V
where n > 0, may also occur, but only for
diagram for H2O shown schematically in
a degenerate or special composition path.
Fig. 1-42, at any point in the area where the
For further discussion, see Hillert (1998).
(S + L) and (L + G) regions overlap there
are two possible equilibrium states of the
1.9.1.1 Zero Phase Fraction Lines system. Similarly, the diagram of carbon
activity versus XCr at constant T and P in
All phase boundaries on true phase dia-
the Fe – Cr – C system in Fig. 1-43 (Hillert,
gram sections are zero phase fraction
1997) exhibits a region in which there is no
(ZPF) lines, a very useful concept intro-
unique equilibrium state.
duced by Gupta et al. (1986). There are
In order to be sure that a diagram is a
ZPF lines associated with each phase. On
true phase diagram, we must choose one
one side of its ZPF line the phase occurs,
and only one variable (either fi or qi ) from
while on the other side it does not. For ex-
ample, in Fig. 1-30 the ZPF line for the a
phase is the line a b c d e f. The ZPF line for
the g phase is g h i j k l. For the MC phase the
ZPF line is m n c i o p q. The ZPF line for
M7C3 is r n b h s p k e t, and for M23C6 it is
u d j o s v. These five ZPF lines yield the en-
tire two-dimensional phase diagram. Phase
diagram sections plotted on triangular co-
ordinates as in Figs. 1-25 and 1-27 also
consist of ZPF lines.
In the case of phase diagrams with de-
generate regions, ZPF lines for two differ-
ent phases may be coincident over part
of their lengths. For example, in Fig. 1-12,
line CABD is the ZPF line of the liquid,
while CEBF and DEAG are the ZPF lines Figure 1-42. Schematic P – V diagram for H2O. This
for the a and b phases respectively. In is not a true phase diagram.
1.9 General Phase Diagram Geometry 59
This defines another set of pairs of conju- case the model is formally the same as that
gate potentials and extensive variables: of a simple substitutional solution, because
(1/T, – H ), (P, – V/T ), ( m i /T, ni ). Choosing the site fractions XMg and XCa of Mg2 + and
one and only one variable from each pair, Ca2 + cations on the cationic sublattice are
we can construct a true phase diagram by numerically equal to the overall component
the procedure described above. However, mole fractions XMgO and XCaO . Solid and liq-
these diagrams may be of limited practical uid MgO – CaO solutions have been shown
utility. This is discussed by Hillert (1997). (Wu et al., 1993) to be well represented by
simple polynomial equations for g E .
As a second example, the intermetallic e-
1.10 Solution Models FeSb phase exhibits non-stoichiometry to-
ward excess Fe. This phase was modeled
In Sec. 1.4.7, the thermodynamic expres- (Pei et al., 1995) as a solution of Fe and
sions for simple ideal substitutional solu- stoichiometric FeSb by assuming two sub-
tions were derived and in Secs. 1.5.7 and lattices: an “Fe sublattice” occupied only
1.6.2, the regular solution model and poly- by Fe atoms and an “Sb sublattice” occu-
nomial extensions thereof were discussed. pied by both Fe and Sb atoms such that, per
For other types of solutions such as ionic gram atom,
mixtures, interstitial solutions, polymeric Dgm = 0.5 R T ( yFe ln yFe + ySb ln ySb )
solutions, etc., the most convenient defini-
+ a yFe ySb (1-118)
tion of ideality may be different. In the
present section we examine some of these where ySb = (1 – yFe ) = 2 XSb is the site frac-
solutions. We also discuss structural order- tion of Sb atoms on the “Sb sites” and a is
ing and its effect on the phase diagram. For an empirical polynomial in ySb .
further discussion, see Pelton (1997).
1.10.1.2 Ionic Solutions
1.10.1 Sublattice Models
Let us take as an example a solution,
The sublattice concept has proved to be solid or liquid, of NaF, KF, NaCl and KCl
very useful in thermodynamic modeling. as introduced in Sec. 1.7.8. If the cations
Sublattice models, which were first devel- are assumed to mix randomly on a cationic
oped extensively for molten salt solutions, sublattice while the anions mix randomly
find application in ceramic, interstitial so- on an anionic sublattice, then the molar
lutions, intermetallic solutions, etc. Gibbs energy of the solution can be mod-
eled by the following equation which con-
1.10.1.1 All Sublattices Except One tains an ideal mixing term for each sublat-
Occupied by Only One Species tice:
In the simplest limiting case, only one g = (XNa XCl g 0NaCl + XK XF g 0KF
sublattice is occupied by more than one + XNa XF g 0NaF + XK XCl g 0KCl ) (1-119)
species. For example, liquid and solid + R T (XNa ln XNa + XK ln XK )
MgO – CaO solutions can be modeled by + R T (XF ln XF + XCl ln XCl ) + g E
assuming an anionic sublattice occupied
only by O2 – ions, while Mg2 + and Ca2 + where the factor (XNa XCl ), for example, is
ions mix on a cationic sublattice. In this the probability, in a random mixture, of
1.10 Solution Models 63
finding a Na ion and a Cl ion as nearest accounts for the fact that the number
neighbors. Differentiation of Eq. (1-119) of Na+ – F – and K+ – Cl – nearest-neighbor
gives the following expression for the ac- pairs will be higher than the number of
tivity of NaF: such pairs in a random mixture. This term
is usually not negligible.
R T ln aNaF = – XK XCl DG exchange (1-120)
The phase diagram in Fig. 1-31 (a) was
+ R T ln (XNa XF ) + g ENaF calculated by means of Eqs. (1-119) and
(1-122) solely from optimized excess
where DG exchange is the Gibbs energy
Gibbs energies of the binary systems and
change for the following exchange reaction
from compiled data for the pure salts.
among the pure salts:
Agreement with the measured diagram is
NaCl + KF = NaF + KCl; (1-121) very good.
DG exchange
=g 0
+g 0
–g 0
– g 0KF Eqs. (1-119) and (1-122) can easily be
NaF KCl NaCl
modified for solutions in which the num-
In this example, DG exchange < 0. The salts bers of sites on the two sublattices are not
NaF and KCl are thus said to form the equal, as in MgCl2 – MgF2 – CaCl2 – CaF2
stable pair. The first term on the right of solutions. Also, in liquid salt solutions the
Eq. (1-120) is positive. The members of the ratio of the number of lattice sites on one
stable pair thus exhibit positive deviations, sublattice to that on the other sublattice
and in Fig. 1-31 this is reflected by the flat can vary with concentration, as in molten
liquidus surfaces with widely spaced iso- NaCl – MgCl2 – NaF – MgF2 solutions. In
therms for NaF and KCl. That is, the mix- this case, it has been proposed (Saboungi
ing of pure NaF and KCl is unfavorable be- and Blander, 1975) that the molar ionic
cause it involves the formation of K+ – F – fractions in all but the random mixing
and Na+ – Cl – nearest-neighbor pairs at the terms of these equations should be replaced
expense of the energetically preferable by equivalent ionic fractions. Finally, the
Na+ – F – and K+ – Cl – pairs. If DG exchange is equations can be extended to multicompo-
sufficiently large, a miscibility gap will be nent solutions. These extensions are all dis-
formed, centered close to the stable diago- cussed by Pelton (1988 b).
nal joining the stable pair. For solutions such as liquid NaF – KF –
Blander (1964) proposed the following NaCl – KCl for which DG exchange is not too
expression for g E in Eq. (1-119): large, these equations are often sufficient.
For solutions with larger exchange Gibbs
g E = XNa g ENaCl–NaF + XK g EKCl–KF (1-122) energies, however, in which liquid immis-
E E
+ XF g NaF–KF + XCl g NaCl–KCl cibility appears, these equations are gener-
– XNa XK XF XCl (DG exchange )2/ZRT ally unsatisfactory. It was recognized by
Saboungi and Blander (1974) that this is
where, for example, g ENaCl–NaF is the excess due to the effect of non-randomness upon
Gibbs energy in the NaCl – NaF binary the second nearest-neighbor cation – cation
system at the same cationic fraction XNa as and anion – anion interactions. To take
in the ternary, and where Z is the first coor- account of this, Blander proposed addi-
dination number. That is, g E contains a tional terms in Eq.(1-122). Dessureault and
contribution from each binary system. The Pelton (1991) modified Eqs. (1-119) and
final term in Eq. (1-122) is a first-order (1-122) to account more rigorously for
correction for non-random mixing which non-random mixing effects, with good re-
64 1 Thermodynamics and Phase Diagrams of Materials
sults for several molten salt systems with ions occupy both sublattices. The solu-
miscibility gaps. See also section 1.10.4. tion can be represented as (Fe21–+ x Ti3x + )A
4+ 3+
(Ti1– x Tix )B where x is the overall mole
fraction of Ti2O3 . The ions are assumed to
1.10.1.3 Interstitial Solutions
mix randomly on each sublattice so that:
As an example of the application of the
sublattice model to interstitial solutions we Ds ideal = – 2 R [(1 – x) ln (1 – x) + x ln x]
will take the f.c.c. phase of the Fe –V– C (1-124)
system. Lee and Lee (1991) have modeled Deviations from ideal mixing are as-
this solution using two sublattices: a metal- sumed to occur due to interlattice cation –
lic sublattice containing Fe and V atoms, cation interactions according to
and an interstitial lattice containing C
(FeA2+ – TiB4+ ) + (TiA3+ – TiB3+ )
atoms and vacancies, va. The numbers of
sites on each sublattice are equal. An equa- = (FeA2+ – TiB3+ ) + (TiA3+ – TiB4+ )
tion identical to Eq. (1-119) can be written DG = a + b T (1-125)
for the molar Gibbs energy:
where a and b are the adjustable parameters
g = (XFe Xva g 0Feva + XFe XC g 0FeC of the model. The probability that an A – B
+ XV Xva g 0Vva + XV XC g 0VC ) (1-123) pair is an (FeA2+ – TiB3+ ) or a (TiA3+ – TiB4+ )
+ R T (XFe ln XFe + XV ln XV) pair is equal to x (1 – x). Hence, g E =
x (1 – x) (a + b T ).
+ R T (XC ln XC + Xva ln Xva ) + g E
Similar models can be proposed for
where XFe = (1 – XV) and XC = (1 – Xva ) are other ceramic solutions such as spinels,
the site fractions on the two sublattices and pseudobrookites, etc. These models can
“Feva” and “Vva” are simply pure Fe and rapidly become very complex mathemati-
V, i.e., g 0Feva = g 0Fe . An expression for g E as cally as the number of possible species on
in Eq. (1-122), although without the final the lattices increases. For instance, in
non-random mixing term, was used by Lee Fe3O4 – Co3O4 spinel solutions, Fe2+, Fe3+,
and Lee with optimized binary g E parame- Co2+ and Co3+ ions are all distributed over
ters. Their calculated Fe –V– C phase dia- both the tetrahedral and octahedral sublat-
gram is in good agreement with experimen- tices. Four independent equilibrium con-
tal data. The sublattice model has been stants are required (Pelton et al., 1979) to
similarly applied to many interstitial solu- describe the cation distribution even for the
tions by several authors. ideal mixing approximation. This com-
plexity has been rendered much more tract-
able by the “compound energy model”
1.10.1.4 Ceramic Solutions (Sundman and Ågren, 1981; Hillert et al.,
Many ceramic solutions contain two or 1988). This is not actually a model, but is
more cationic sublattices. As an example, rather a mathematical formalism permit-
consider a solution of Ti2O3 in FeTiO3 (il- ting the formulation of various models in
menite) under reducing conditions. There terms of the Gibbs energies, g 0 , of “pseu-
are two cationic sublattices, the A and B docomponents” so that equations similar to
sublattices. In FeTiO3 , Fe2 + ions and Ti 4 + Eq. (1-119) can be used directly.
ions occupy the A and B sublattices, re-
spectively. With additions of Ti2O3 , Ti 3 +
1.10 Solution Models 65
0 0
where Dg1 = (g AA – g AB ) and Dg2 = 40)) are observed. This large discrepancy
0 0
(g BB – g AB ) are the Gibbs energies of for- can be attributed to the fact that the indi-
mation of the majority defects and where vidual segments of the polymer molecule
0 0
g BA has been set equal to g AB . At a given have considerable freedom of movement.
composition d = (x – y), and for given val- Flory (1941, 1942) and Huggins (1942)
ues of the parameters Dg1 and Dg2 , Eq. proposed a model in which the polymer
(1-129) can be solved to give x and y, segments are distributed on the solvent
which can then be substituted into Eq. sites. A large polymer molecule can thus be
(1-128) to give g. The more positive are oriented (i.e. bent) in many ways, thereby
Dg1 and Dg2 , the more steeply g rises on ei- greatly increasing the entropy. To a first ap-
ther side of its minimum, and the narrower proximation the model gives an ideal mix-
is the range of stoichiometry of the com- ing term with mole fractions replaced by
pound. volume fractions in Eq. (1-45):
Consider another model in which the
⎛ XA v A0
majority defects are vacancies on the B- Dg mideal = RT ⎜ XA ln
sublattice and B atoms on interstitial sites. ⎝ XA v A0 + XB v B0
We now have three sublattices with occu- XB v B0 ⎞
pancies (A)A (B1– y vay )B (Bx va1– x )I where + XB ln 0 ⎟
(1-131)
XA v A + XB v A ⎠
0
“I” indicates the interstitial sublattice. The
A-sublattice is occupied exclusively by A Lewis and Randall (1961) have com-
atoms. A vacancy is indicated by va. Stoi- pared the Flory – Huggins equation with
chiometric defect-free AB is represented experimental data in several solutions.
by (A) (B) (va) and (x – y) = 2 d /(1 – d ). In general, the measured activities lie be-
Per mole of A1– d B1+ d , the Gibbs energy is: tween those predicted by Eq. (1-131) and
0 by the Raoultian ideal equation, Eq. (1-45).
g = (1 – d ) {[(1 – x) (1 – y) gABva
0 0 0
A recent review of the thermodynamics
+ (1 – x) y gAvava + x (1 – y) gABB + x y gAvaB of polymer solutions is given by Trusler
+ R T [x ln x + (1 – x) ln (1 – x) (1999).
+ y ln y + (1 – y) ln (1 – y)]} (1-130)
1.10.3 Calculation of Limiting Slopes
Eq. (1-130) is identical to Eq. (1-128) apart
of Phase Boundaries
from the factor (1 – d ), and gives rise to an
equilibrium constant as in Eq. (1-129). From the measured limiting slope
Other choices of majority defects result in (dT/dX )XA =1 of the liquidus at the melting
very similar expressions. The model can point of a pure component A, much infor-
easily be modified to account for other mation about the extent of solid solubility,
stoichiometries Am Bn , for different num- as well as the structure of the liquid, can be
bers of available interstitial sites, etc., and inferred. Similar information can be ob-
its extension to multicomponent solutions tained from the limiting slopes of phase
is straightforward. boundaries at solid-state transformation
points of pure components.
1.10.2 Polymer Solutions Eq. (1-65) relates the activities along the
liquidus and the solidus to the Gibbs en-
For solutions of polymers in monomeric
ergy of fusion:
solvents, very large deviations from simple
l
Raoultian ideal behaviour (i.e. from Eq. (1- R T ln aA – R T ln a As = – Dg0f(A) (1-132)
1.10 Solution Models 67
In the limit XA Æ 1, the liquidus and solidus Agreement with the measured limiting liq-
converge at the melting point Tf (A) . Let us uidus slope is very good, thereby showing
assume that, in the limit, Raoult’s Law, Eq. that the solid solubility of NaF in LiF is not
(1-40), holds for both phases. That is, large.
l
aA = X Al and a As = X As . Furthermore, from In the general case, the solute B may dis-
Eq. (1-60), solve to form more than one “particle”. For
example, in dilute solutions of Na2SO4 in
Dg0f(A) Æ Dhf (A) (1 – T /Tf (A) )
MgSO4 , each mole of Na2SO4 yields two
Finally, we note that moles of Na+ ions which mix randomly
with the Mg2+ ions on the cationic sublat-
lim (ln XA ) = lim (ln (1 – XB )) = – XB tice. Hence, the real mole fraction of sol-
XA Æ 1 XA Æ 1
vent, XA , is (1 – n XB ) where n is the number
Substituting these various limiting values of moles of foreign “particles” contributed
into Eq. (1-132) yields: by one mole of solute. In the present exam-
lim (dX Al /dT – dX As /dT ) ple, n = 2.
XA Æ 1 Eq. (1-133) now becomes:
= Dh0f(A) /R (Tf (A) )2 (1-133)
lim (dX Al /dT – dX As /dT )
XA Æ 1
If the limiting slope of the liquidus,
lim (dX Al /dT ), is known, then the limiting = Dh0f(A) / n R (Tf (A) )2 (1-134)
XA Æ 1
slope of the solidus can be calculated, or The broken line in Fig. 1-44 is the limiting
vice versa, as long as the enthalpy of fusion liquidus slope calculated from Eq. (1-134)
is known. under the assumption of negligible solid
For the LiF – NaF system in Fig. 1-18, solubility.
the broken line is the limiting liquidus It can be shown (Blander, 1964) that Eq.
slope at XLiF = 1 calculated from Eq. (1- (1-134) applies very generally with the fac-
133) under the assumption that there is no tor n as defined above. For example, add-
solid solubility (that is, that dX As /dT = 0). ing LiF to NaF introduces only one foreign
tion with X11 = Y12, X22 = Y22 and 2 Y1 Y2 , and the following optimized (Kongoli et al.,
Eq. (1-140) reduces to the ideal Raoultian 1998) polynomial:
entropy of mixing. When (w – h T ) be- (1-142)
comes very negative, 1– 2 pairs predomi- (w – h T ) = – (70 017 + 9 T ) – 74 042 YS
nate. A plot of Dhm or s E (non-config) versus – (798 – 15 T ) YS3 + 40 791 YS7 J/mol –1
composition then becomes V-shaped and a
config
plot of Dsm becomes m-shaped, with Far fewer parameters are required than if a
minima at Y1 = Y2 = 1/2, which is the com- polynomial expansion of gE (as in Sec.
position of maximum ordering, as illus- 1.6.2) were used. Furthermore, and more
trated in Fig. 1-45. When (w – h T ) is quite importantly, the model permits successful
negative, the plot of gE also has a negative predictions of the properties of multicom-
V-shape. ponent systems as illustrated in Fig. 1-47,
For Fe – S liquid solutions, the activity where measured sulfur activities in quater-
coefficients of sulfur as measured by sev- nary liquid Fe – Ni – Cu – S solutions are
eral authors are plotted in Fig. 1-46. It is compared with activities calculated (Kon-
clear in this case that the model should be goli et al., 1998) solely from the optimized
applied with Z Fe = Z S . The curves shown in model parameters for the Fe – S, Ni – S
Fig. 1-46 were calculated from the quasi- and Cu – S binary systems. A pair exchange
chemical model with (w – h T ) expanded as reaction like Eq. (1-135) was assumed
for each M – S pair (M = Fe, Ni, Cu), and
an optimized polynomial expansion of
(wMS – h MS T ) as a function of YS , similar to
Eq. (1-142), was obtained for each binary.
Three equilibrium constant equations like
Eq. (1-141) were written, and it was as-
sumed that (wMS – h MS T ) in the quaternary
system was constant at constant YS . No ad-
justable ternary or quaternary parameters
were required to obtain the agreement
shown in Fig. (1-47), although the model
permits the inclusion of such terms if re-
quired.
Silicate slags are known to exhibit such
short-range ordering. In the CaO – SiO2
system, Dhm has a strong negative V-shape,
as in Fig. 1-45, but with the minimum at
XSiO2 = 1/3 which is the composition corre-
sponding to Ca2SiO4 . That is, the ordering
is associated with the formation of ortho-
silicate anions SiO44 –. In the phase diagram,
Fig. 1-14, the CaO-liquidus can be seen to
Figure 1-45. Molar enthalpy and entropy of mixing descend sharply near the composition
curves for a system AB calculated at 1000 °C with
Z1 =Z2 =2 from the quasichemical model for short- XSiO2 = 1/3. The quasichemical model has
range ordering with (ω – η T ) = 0, – 21, – 42, and been extended by Pelton and Blander
– 84 kJ. (1984) to treat silicate slags. The diagram
70 1 Thermodynamics and Phase Diagrams of Materials
Figure 1-46. Activity coefficient of sulfur in liquid Fe – S solutions calculated from optimized quasichemical
model parameters and comparison with experimental data (Kongoli et al., 1998).
Figure 1-47. Equilibrium partial pressure of sulfur at 1200 oC over Fe – Ni – Cu – S mattes predicted by the
quasichemical model from binary data (Kongoli et al., 1998) and comparison with experimental data
(Bale and Toguri, 1996).
1.10 Solution Models 71
Figure 1-49. Projection of liquidus surface of the SiO2 – MgO – MnO system calculated from optimized binary
parameters with the quasichemical model for the liquid phase.
(such as fixed T, P and overall composi- has been asked to calculate the equilibrium
tion), the free energy minimization algo- state when 1 mol of SiI 4 is held at 1400 K
rithms find the set of mole numbers nA , nB , in a volume of 104 l. The thermodynamic
etc., na , nb , etc. (some may be zero) as well properties of the possible product species
as the compositions of all solution phases have been automatically retrieved from
which globally minimize G. This is the the database and the total Gibbs energy
equilibrium phase assemblage. Other con- has been minimized by the Solgasmix
straints such as constant volume or a fixed algorithm. At equilibrium there will be
chemical potential (such as constant pO2 ) 2.9254 mol of gas of the composition
may be applied. shown and 0.11182 mol of solid Si will
A discussion of the strategies of such al- precipitate. The total pressure will be
gorithms is beyond the scope of the present 0.0336 bar.
chapter. One of the best known general Although the calculation was performed
Gibbs energy minimization programs is by minimization of the total Gibbs energy,
Solgasmix written by Eriksson (1975) and substitution of the results into the equilib-
constantly updated. rium constants of Eqs. (1-10) to (1-12) will
When coupled to a large thermodynamic show that these equilibrium constants are
database, general Gibbs energy minimiza- satisfied.
tion programs provide a powerful tool for Another example is shown in Table 1-3
the calculation of phase equilibria. Several (Pelton et al., 1990). Here the program has
such expert database systems have been been asked to calculate the equilibrium
developed. They have been reviewed by
Bale and Eriksson (1990).
An example of a calculation performed
Table 1-3. Calculation of equilibrium state when re-
by the F * A * C * T (Facility for the Analy-
actants shown (masses in g) are held at 1873 K at a
sis of Chemical Thermodynamics) expert pressure of 1 atm. Calculations performed by mini-
system, which the author has helped to de- mization of the total Gibbs energy.
velop, is shown in Table 1-2. The program
100. Fe + 0.08 O + 0.4 Fe + 0.4 Mn + 0.3 Si + 0.08 Ar =
0.30793 litre ( 99.943 vol% Ar
+ 0.24987E – 01 vol% Mn
Table 1-2. Calculation of equilibrium state when + 0.24069E – 01 vol% SiO
1 mole SiI4 is held at 1400 K in a volume of 104 l. + 0.82057E – 02 vol% Fe
Calculations performed by minimization of the total + 0.79044E – 07 vol% O
Gibbs energy. + 0.60192E – 08 vol% Si
+ 0.11200E – 08 vol% O2
SiI4 =
+ 0.35385E – 15 vol% Si2 )
2.9254 ( 0.67156 I
(1873.0, 1.00 ,G)
+ 0.28415 SiI2
+ 0.24835E – 01 I2 + 0.18501 gram ( 49.948 wt.% SiO2
+ 0.19446E – 01 SiI4 + 42.104 wt.% MnO
+ 0.59083E – 05 SiI + 7.9478 wt.% FeO)
+ 0.23039E – 07 Si (1873.0, 1.00 ,SOLN 2)
+ 0.15226E – 10 Si2 + 100.99 gram ( 99.400 wt.% Fe
+ 0.21462EE – 11 Si3) + 0.33630 wt.% Mu
(1400.0,0.336E – 01,G) + 0.25426 wt.% Si
+ 0.11182 Si + 0.98375E – 02 wt.% O2)
(1400.0,0.336E – 01,S1, 1.0000) (1873.0, 1.00 ,SOLN 3)
76 1 Thermodynamics and Phase Diagrams of Materials
state when 100 g Fe, 0.08 g oxygen, 0.4 g calculate all types of phase diagrams with
Fe, 0.4 g Mn, 0.3 g Si and 0.08 g Ar are any properly chosen variables as axes or
brought together at 1873 K at a total pres- constants.
sure of 1 bar. The database contains data
for a large number of solution phases as
well as for pure compounds. These data 1.13 Bibliography
have been automatically retrieved and the
total Gibbs energy has been minimized. At 1.13.1 Phase Diagram Compilations
equilibrium there are 0.30793 l of a gas
phase, 0.18501 g of a molten slag, and The classic compilation in the field of bi-
100.99 g of a molten steel of the composi- nary alloy phase diagrams is that of Hansen
tions shown. (1958). This work was continued by Elliott
The Gibbs energies of the solution (1965) and Shunk (1969). These compila-
phases are represented as functions of com- tions contain critical commentaries. A non-
position by various solution models (Sec. critical compilation of binary alloy phase
1.10). As discussed in Sec. 1.11, these diagrams is supplied in looseleaf form with
models can be used to predict the thermo- a continual up-dating service by W.G. Mof-
dynamic properties of N-component solu- fatt of the General Electric Co., Schenec-
tions from evaluated parameters for binary tady, N.Y. An extensive non-critical compi-
(and possibly ternary) subsystems stored in lation of binary and ternary phase diagrams
the database. For example, in the calcula- of metallic systems has been edited by
tion in Table 1-3, the Gibbs energy of the Ageev (1959 – 1978). An index to all com-
molten slag phase was estimated by the pilations of binary alloy phase diagrams up
quasichemical model from optimized pa- to 1979 was prepared by Moffatt (1979). A
rameters for the binary oxide solutions. critical compilation of binary phase dia-
grams involving Fe has been published by
Kubaschewski (1982). Ternary alloy phase
1.12.1 Phase Diagram Calculation
diagrams were compiled by Ageev (1959 –
Gibbs energy minimization is used to 1978).
calculate general phase diagram sections From 1979 to the early 1990s, the Amer-
thermodynamically using the zero phase ican Society for Metals undertook a project
fraction line concept (Sec. 1.9.1.1), with to evaluate critically all binary and ternary
data retrieved from databases of model co- alloy phase diagrams. All available litera-
efficients. For example, to calculate the di- ture on phase equilibria, crystal structures,
agram of Fig. 1-30, the program first scans and often thermodynamic properties were
the four edges of the diagram to find the critically evaluated in detail by interna-
ends of the ZPF lines. Each line is then fol- tional experts. Many evaluations have ap-
lowed from beginning to end, using Gibbs peared in the Journal of Phase Equilibria
energy minimization to determine the point (formerly Bulletin of Alloy Phase Dia-
at which a phase is just on the verge of be- grams), (ASM Int’l., Materials Park, OH),
ing present. When ZPF lines for all phases which continues to publish phase diagram
have been drawn, then the diagram is com- evaluations. Condensed critical evalua-
plete. Because, as shown in Sec. 1.9, all tions of 4700 binary alloy phase diagrams
true phase diagram sections obey the same have been published in three volumes
geometrical rules, one algorithm suffices to (Massalski et al., 1990). The ternary phase
1.13 Bibliography 77
Elliott, R. P. (1965), Constitution of Binary Alloys. Katsura, T., Muan, A. (1964), TMS-AIME 230, 77.
First Supplement. New York: McGraw-Hill. Kellogg, H. H., Basu, S. K. (1960), TMS-AIME 218,
Eriksson, N. G. (1975), Chemical Scripta 8, 100. 70.
Findlay, A. (1951), The Phase Rule, 9th edn., revised Kleinstuber, T. (1961), Ph.D. Thesis, Univ. Munich,
by Campbell, A. N., Smith, N. A. New York: Germany.
Dover. Kohler, F. (1960), Monatsh. Chemie 91, 738.
Flory, P. J. (1941), J. Chem. Phys. 9, 660. Köster, W., Dullenkopf, W. (1936), Z. Metallk. 28,
Flory, P. J. (1942), J. Chem. Phys. 10, 5. 309.
Fowler, R. H., Guggenheim, E. A. (1939), Statistical Kongoli, F., Dessureault, Y., Pelton, A. D. (1998),
Thermodynamics. Cambridge: Cambridge Univ. Metall. and Mat. Trans. 29B, 591.
Press, p. 350. Kubaschewski, O. (1982), Iron-Binary Phase Dia-
Garrels, R. M., Christ, C. L. (1965), Solutions Miner- grams. New York: Springer Verlag.
als and Equilibria. New York: Harper and Row. Kubaschewski, O., Alcock, C. B., (1979), Metallur-
Ginsberg, A. S. (1909), Z. Anorg. Chem. 61, 130. gical Thermochemistry, 5th ed. New York: Per-
Glasser, E. P., Osborn, E. F. (1960), J. Amer. Ceram. gamon Press.
Soc. 43, 136. Kurnakow, N. S., Kusnetzow, A. N. (1907), Z. Anorg.
Guggenheim, E. A. (1935), Proc. Roy. Soc. A 148, Chem. 52, 173.
304. Kunz, M. (2001), in: Phase Transformations in Ma-
Gulbransen, E. A., Jansson, S. A. (1970), Proc. terials: G. Kostorz (Ed.). Weinheim: Wiley-VCH,
Symp. Electrochem. Soc. Fall meeting 1969. New Chap. 10.
York: The Electrochem. Soc., p. 3. Lantratov, M. F., Mikhailova, A. G. (1971), Zh. Prikl.
Gupta, H., Morral, J. E., Nowotny, H. (1986), Scripta Khimii 44, 1778.
Metall. 20, 889. Lee, B.-J., Lee, D. N. (1991), Calphad 15, 293.
Hansen, M. (1958), Constitution of Binary Alloys, Lee, B. J., Lee, D. N. (1992), J. Phase Equilib. 13, 349.
2nd edn. New York: McGraw-Hill. Levin, E. M., Robbins, C. R., McMurdie, H. F. (1964),
Harshe, G., Venkatachalam, S. (1984), J. Metals 36, and 11 supplements to (1996), Phase Diagrams for
34. Ceramists; Columbus, Ohio: Am. Ceramic Soc.
Hauffe, K. (1940), Z. Elektrochem. 46, 348. Lewis, G. N., Randall, M. (1961), Thermodynamics,
Hensel, F. (1979), Adv. Phys. 28, 555. revised by Pitzer, K. S., Brewer, L. New York:
Hillert, M. (1980), Calphad Journal 4, 1. McGraw-Hill.
Hillert, M. (1985), Int. Metall. Rev. 30, 45. Lin, P.-L., Pelton, A. D., Bale, C. W. (1979), J. Am.
Hillert, M. (1997), J. Phase Equilib. 18, 249. Ceram. Soc. 62, 414.
Hillert, M. (1998), Phase Equilibria, Phase Dia- Lipson, H., Wilson, A. J. C. (1940), J. Iron Steel Inst.
grams and Phase Transformations. Cambridge: 142, 122.
Cambridge University Press. Lukas, H. L., Henig, E.-Th., Zimmermann, B.
Hillert, M., Jansson, B., Sundman, B. (1988), Z. (1977), Calphad Journal 1, 225.
Metallk. 79, 81. MacChesney, J. B., Rosenberg, P. E. (1970), in Phase
Holm, J. L. (1965), Acta Chem. Scand. 19, 638. Diagrams – Materials Science and Technology,
Hong, K. C., Kleppa, O. J. (1976), J. Chem. Thermo- Vol. 1: Alper, A. M. (Ed.). New York: Academic,
dyn. 8, 31. Ch. 3.
Huggins, M. L. (1942), Ann. N. Y. Acad. Sci. 43, 1. Maiorova, E. A., Morachevskii, A. G., Kovalenko, S.
Hultgren, R., Desai, P. D., Hawkins, D. T., Gleiser, G. (1976), Elektrokhimiya 12, 313.
M., Kelly, K. K., Wagman, D. D. (1973), Selected Massalski, T. B., Okamoto, H., Subramanian, P. R.,
Values of the Thermodynamic Properties of the Kacprzak, L. (1990), Binary Alloy Phase Dia-
Elements and Binary Alloys. Metals Park, Ohio: grams, 2nd ed. Metals Park, OH: Am. Soc. Metals.
Am. Soc. of Metals. Mathewson, C. H. (1906), Z. Ang. Chem. 50, 180.
Hume-Rothery, W., Christian, J. W., Pearson, W. B. Mills, K. C. (1974), Thermodynamic Data for Inor-
(1952), Metallurgical Equilibrium Diagrams. ganic Sulphides, Selenides and Tellurides. Lon-
London: Inst. Phys. don: Butterworths.
Inden, G. (1982), Bull. Alloy Phase Diag. 2, 412. Miodownik, A. P. (1982), Bull. Alloy Phase Diag. 2,
Inden, G. (2001), in: Phase Transformations in Mate- 406.
rials: G. Kostorz (Ed.). Weinheim: Wiley-VCH, Moffatt, W. B. (1979), The Index to Binary Phase Col-
Chap. 8. lections, Schenectady. N.Y.: General Electric Corp.
Ingraham, T. R., Kellogg, H. H. (1963), TMS-AIME Muan, A., and Osborn, F. (1965), Phase Equilibria
227, 1419. Among Oxides in Steelmaking. Reading, MA.:
Ingraham, T. R., Kerby, R. (1967), Canad. Metall. Addison Wesley.
Quart. 6, 89. Müller-Krumbhaar, H., Kurz, W., Brener, E. (2001),
Ivanov, A. I. (1953), Izv. Sekt. Fiz.-Khim. Anal. Inst. in: Phase Transformations in Materials: G. Kostorz
Obshchei Neorg. Khim. Akad. Nauk SSSR 23, 197. (Ed.). Weinheim: Wiley-VCH, Chap. 2.
80 1 Thermodynamics and Phase Diagrams of Materials
Palatnik, L. S., Landau, A. I. (1964), Phase Equilib- Robertson, W. D. (Ed.) (1966), Binary Phase Dia-
ria in Multicomponent Systems. NY: Holt, Rine- grams of Halide Salts, U. S. At. En. Comm. Con-
hart and Winston. tract AT (30-1) 2723, Washington: Clearinghouse
Pehlke, R. D. (1973), Unit Processes of Extractive for Federal Scientific and Techn. Info.
Metallurgy. New York: Elsevier. Robie, R. A., Hemingway, B. S., Fisher, J. R. (1978),
Pei, B., Björkman, B., Sundman, B., Jansson, B. Thermodynamic Properties of Minerals and Re-
(1995), Calphad 19, 1. lated Substances at 298.15 K and 1 Bar Pressure
Pelton, A. D. (1983), in: Physical Metallurgy: Cahn, and at Higher Temperatures. Washington: U.S.
R. W., Haasen, P. (Eds.). Amsterdam: Elsevier, Government Printing Office.
Chap. 7. Roos, G. D. (1916), Z. Anorg. Chem. 94, 329.
Pelton, A. D. (1988 a), Bull. Alloy Phase Diag. 9, 41. Saboungi, M.-L., Blander, M. (1974), High Temp.
Pelton, A. D. (1988 b), Calphad Journal 12, 127. Sci. 6, 37.
Pelton, A. D. (1995), J. Phase Equilib. 16, 501. Saboungi, M.-L., Blander, M. (1975), J. Am. Ceram.
Pelton, A. D. (1996), in: Physical Metallurgy: Cahn, Soc. 58, 1.
R. W., Haasen, P. (Eds.). Amsterdam, North-Hol- Saboungi, M.-L., Herron, S. J., Kumar, R. (1985),
land, Chap. 6. Ber. Bunsenges. Phys. Chem. 89, 375.
Pelton, A. D. (1997), in: Advanced Physical Chemis- Sangster, J., Pelton, A. D. (1987), J. Phys. Chem.
try for Process Metallurgy: Sano, N., Lu, W.-K, Ref. Data 16, 509.
Riboud P. (Eds.). New York, Academic, Chap. 3. Schenk, H., Frohberg, M. G., Nunninghof, R. (1964),
Pelton, A. D., Bale, C. W. (1986), Metall. Trans. Arch. Eisenhütten 35, 269.
A 17, 1057. Schreinemakers, F. A. H. (1915), Proc. K. Akad.
Pelton, A. D., Blander, M. (1984), Proc. AIME Symp. Wetenschappen, Amsterdam (Section of Sciences)
on Molten Salts and Slags. Warrendale, Pa.: The 18, 116.
Metall. Soc. AIME, p. 281. Shatynski, S. R. (1977), Oxidn. Metals 11, 307.
Pelton, A. D., Blander, M. (1986), Metall. Trans. Shunk, F. A. (1969), Constitution of Binary Alloys.
B 17, 805. Second Supplement. New York: McGraw-Hill.
Pelton, A. D., Chartrand, P. (2000), Metall. Mater. Spencer, P. J., Barin, I. (1979), Mater. Eng. Appl. 1,
Trans., in press. 167.
Pelton, A. D., Degterov, S. A., Eriksson, G., Robelin, Stringer, J., Whittle, D. P. (1977), Rev. Int. Htes.
C., Dessureault, Y. (2000), Metall. Mater. Trans., Temp. Réfract. 14, 6.
B 31, 651. Sundman, B., Ågren, J. (1981), J. Phys. Chem. Solids
Pelton, A. D., Schmalzried, H. (1973), Metall. Trans. 42, 297.
4, 1395. Trusler, J. P. M. (1999), in: Chemical Thermodynam-
Pelton, A. D., Schmalzried, H., Sticher, J. (1979), ics: Letcher, T. (Ed.). Abingdon, Oxon, UK:
J. Phys. Chem. Solids 40, 1103. Blackwell Science, Chap. 16.
Pelton, A. D., Thompson, W. T. (1975), Prog. Solid Van Laar, J. J. (1908), Z. Phys. Chem. 63, 216; 64,
State Chem. 10, part 3, 119. 257.
Pelton, A. D., Thompson, W. T., Bale, C. W., Erik- Villars, P., Prince, A., Okamoto, H. (1995), Hand-
son, N. G. (1990), High Temp. Sci. 26, 231. book of Ternary Alloy Phase Diagrams. Metals
Petric, A., Pelton, A. D., Saboungi, M.-L. (1988 a), Park, OH: Am. Soc. Metals.
J. Electrochem. Soc. 135, 2754. Voskresenskaya, N. K. (Ed.) (1970), Handbook of
Petric, A., Pelton, A. D., Saboungi, M.-L. (1988 b), Solid-Liquid Equilibria in Systems of Anhydrous
J. Chem. Phys. 85, 5070. Inorganic Salts, Vol. 1, TT69-55079/1, AEC-tr-
Phillips, B., Muan, A. (1958), J. Am. Ceram. Soc. 41, 69831/3, UC-4, Washington: U.S. At. En. Comm.
448. and Nat’l. Sci. Foundation.
Pilgrim, R. F., Ingraham, T. R. (1967), Canad. Weeks, J. R. Davies, H. A. (1964), AEC Report,
Metall. Quart. 6, 333. Conf. 660712-1, BNL-10372.
Polyakov, V. D. (1940), Izv. Sektora Fiz.-Khim Anal- West, D. R. F. (1965), Ternary Equilibrium Dia-
iza Inst. Obshchei Neorg. Khim. Akad. Nauk SSSR grams. New York: McMillan.
13, 308. Wisniak, J. (1981), Phase Diagrams: A Literature
Prince, A. (1963), Metall. Rev. 8, 213. Source Book, Vols. 1 and 2. New York: Elsevier.
Prince, A. (1966), Alloy Phase Equilibria. Amster- Wisniak, J., Tamir, A. (1978), Mixing and Excess
dam: Elsevier. Thermodynamic Properties: A Literature Source
Raynor, G. V. (1970), in: Physical Metallurgy, 2nd Book. New York: Elsevier.
edn: Cahn, R. W. (Ed.). Amsterdam: North Hol- Wu. P. (1990), Ph. D. thesis, Ecole Polytechnique,
land, Chap. 7. Montreal.
Rhines, F. N. (1956), Phase Diagrams in Metallurgy. Wu, P., Eriksson, G., Pelton, A. D. (1993), J. Am.
New York: McGraw-Hill. Ceram. Soc. 76, 2065.
Ricci, J. E. (1964), in: Molten Salt Chemistry: Yokokowa, H. (1999), J. Phase Equilib. 20, 258.
Blander, M. (Ed.). New York: Interscience, Chap. 4.
2 Solidification
Heiner Müller-Krumbhaar
Wilfried Kurz
Efim Brener
R0 radius of curvature
rc critical radius
S position on the interface, entropy
DS change in entropy
T temperature
t time
Dt change in time
T0 reference temperature
TE eutectic temperature
TI interface temperature
Tm melting temperature
u dimensionless temperature field or concentration field
U energy
V speed
Va maximal speed, absolute stability
Vc critical velocity
vR growth rate
w probability of fluctuation
x space coordinates
Xi , Xj extensive variables
Yi , Yj intensive variables
Z partition function
z space coordinate
b interface kinetics
b4 4-fold anisotropy of the kinetic coefficient
g surface tension or surface free energy
G ratio of S – L interface energy to specific melting entropy
G2 Green’s function integral
D supercooling (negative temperature field u at infinity)
d local concentration gap, surface tension anisotropy
e relative strength of anisotropy of capillary length
z (x, t) deviation of the interface
h coordinate
J angle of orientation relative to the crystallographic axes
Q function of (x, t)
k (s) mobility
l interface spacing, wavelength
L ( s, e ) solvability function
l2 wavelength of side branch
lf wavelength of the fastest mode
ls stability length
m chemical potential
n0 effective kinetic prefactor
x variable
84 2 Solidification
xi position vectors
Dxi small normal displacement
r Ivantsov radius
s scaling function
t variable
W atomic volume, atomic area
V and the length scale L of the resulting pler than if we attempted to carry along all
pattern: prefactors. The second and more important
reason is that the scaled form allows for a
V L2 ≈ constant (2-1)
much simpler comparison of experiments
Surprisingly enough, we have arrived at for different sets of parameters which usu-
about the most celebrated equation for pro- ally appear only in some combinations in
cesses of dendritic, directional and eutectic the equations, thereby leading to cancella-
solidification, without even defining any of tions and compensations.
these processes! Two remarks, therefore, Section 2.2 gives a quick summary of the
may be in order here. First, detailed studies ingredients for a theory, starting at nuclea-
of the different solidification schemes in tion, then deriving boundary conditions for
recent years have attempted to extract the a propagating interface between two
common and universal aspects of these phases, and finally discussing some gen-
processes. Such common features indicate eral aspects of the diffusion equation with a
a common basis of rather general nature, as propagating boundary. This is followed by
outlined above. Second, we have of course an introduction to basic experimental tech-
ignored most of the specific aspects of each niques in Sec. 2.3. In Sec. 2.4, the case of a
individual process. In directional solidifi- simple solid growing in a supercooled melt
cation, for example, a band of possible is discussed in some detail, leading to the
wavelengths for stationary patterns are present understanding of dendritic growth.
found and up to now it is not clear if and In Sec. 2.5 the technically important pro-
how a specific wavelength from that band cess of directional solidification is pre-
is finally selected. The assumption of just sented. The evolution of cellular patterns
two independent length scales in many above a critical growth rate can in principle
cases is also a rather gross simplification of be understood without any knowledge of
the actual situation. We will therefore leave dendritic growth. Actually, the parameter
this line of general arguments and look at range for simple sinusoidal cells is very
some concrete models that are believed to narrow so that we usually operate in the
capture at least some essentials for the fas- range of deep cells or even dendrites,
cinating patterns produced during solidifi- which suggests our sequence of presenta-
cation. tion. Finally, these concepts are extended
Some remarkable progress has been in Sec. 2.6 to alloys with a eutectic phase
achieved in the theoretical treatment of diagram and the resulting complex phe-
these phenomena during the recent years. nomena. As this field is currently in rapid
In the list of references, we have concen- theoretical development, our discussion
trated our attention on recent developments here necessarily is somewhat preliminary.
since there are some good reviews on older The chapter is closed by a summary with
work (for example, Langer, 1980 a; Kurz complementary remarks.
and Fisher, 1998).
An experimentalist may finally wonder
why we have expressed most of the equa- 2.2 Basic Concepts
tions in a non-dimensionally scaled form in First-Order Phase Transitions
rather than writing all material parameters
down explicitly at each step. One reason is The different possible phases of a mate-
that the equations then appear much sim- rial existing in thermodynamic equilibrium
2.2 Basic Concepts in First-Order Phase Transitions 87
mation of a spherical solid nucleus of ra- The extremal value of Eqs. (2-4) and (2-5)
dius r within a slightly supercooled liquid. with respect to the variation of r gives the
Here W is the atomic volume, P the pres- critical radius
sure, m L > m S are the chemical potentials of
liquid and solid in a homogeneous system, 2g
rc = W (3-dim) (2-6 a)
and g is the (here isotropic) surface ten- Dm
sion, or surface free energy. or
A few remarks should be made here.
g
When deriving Eq. (2-4), we use the con- rc = W (2-dim) (2-6 b)
Dm
cept of small deviations from equilibrium,
while Eqs. (2-2) and (2-3) are more gener- so that for r < rc , the nucleus tends to
ally valid (Landau and Lifshitz, 1970). In shrink, while for r > rc it tends to grow and
the estimation of the range of validity of at r = rc it is in an unstable equilibrium.
Eq. (2-4), it is apparent, however, that it The same thermal fluctuations causing
should be applicable to even very large such a nucleus to appear also produce devi-
supercoolings for most liquids, since the ations from the average spherical shape.
thermal transport is either independent of, This leads to power-law corrections in the
or faster than, the kinetics of nucleus for- prefactor of Eq. (2-5) or logarithmic cor-
mation (Ohno et al., 1990). The range of rections in the exponent (Voronkov, 1983;
validity of Eq. (2-4) is then typically lim- Langer, 1971).
ited by the approach to the “spinodal” re- So far these considerations have dealt
gion, where metastable states become un- with static aspects only. Since the fluctua-
stable, even when fluctuations are ignored tions vary locally with time, Eq. (2-5) may
(see the Chapter by Binder and Fratzl, 2001). be interpreted as the rate at which such
Assuming, therefore that we are still in fluctuation occurs, and, consequently, with
the range of well-defined metastable states, r = rc we obtain the rate for the formation
we may write Eqs. (2-2) and (2-3) as of a critical nucleus which, after appear-
w = n0 exp (– DG/T0 ) (2-5) ance, is assumed to grow until the new
phase fills the whole system. This is the
identifying the change in Gibbs energy DG classical nucleation theory. A very elegant
by Eq. (2-4), with an undetermined prefac- formulation was given by analytic continu-
tor n0 . Here we do not discriminate ation into the complex plane (Langer,
between surface tension and surface free 1971). Further additions include the defini-
energy, despite the fact that the first is a tion of the prefactor in Eq. (2-5) and a more
tensorial quantity, and the latter only scalar detailed analysis of the kinetics near r = rc
(although it may be anisotropic, which is (Zettlemoyer, 1969, 1976).
ignored here). A difference between sur- Considering the many uncertainties en-
face tension and surface free energy arises tering from additional sources such as the
when the system does not equilibrate range of atomic potentials and the change
between surface and bulk so that, for exam- of atomic interaction in the surface, we will
ple, the number of atoms in the surface ignore all these effects by absorbing them
layer is conserved. Throughout this chap- into the effective kinetic prefactor n0 in Eq.
ter, we will assume perfect local equilibra- (2-5), to be determined experimentally.
tion in this respect, and we may then ignore
the difference.
2.2 Basic Concepts in First-Order Phase Transitions 89
and the divergence taken in the same ab- from which the variational derivative in
stract space. The first term in Eq. (2-10) is Eq. (2-13) is formally obtained by
a drift, the second term the constitutive re-
lation with transport coefficients Di , and ⎡ dGˆ ⎤
dGtot = ∫ dS ⎢ ⎥ dx ( S ) (2-15)
the derivative is taken normal to the inter- S ⎣ dx ⎦
90 2 Solidification
where d x (S) is a small arbitrary displace- infinite solid or liquid without influences
ment of the interface in the normal direc- from curvature terms. For a single-compo-
tion. Explicitly this is written as nent system, m = g W, where m is the chemi-
cal potential of the respective phase. The
dGtot = ∫ {g d (dS ) + dS dg } generalization to three dimensions adds an-
S
other curvature term into Eq. (2-21), which
− ∫ dS d x ( S ) [ g L ( S ) − gS ( S )] (2-16) then corresponds to two curvatures and an-
S
gles in the two principal directions (for iso-
with Gibbs energy densities g given here tropic g , the curvature 1/R is simply
per atomic “volume” (or area) W. As usual, changed to 2/R).
we have to extract terms d x out of the Two useful observations can be made at
terms d (dS) and d g . Assuming g to depend this stage. There is a solution with V^ = 0 if
only on local orientation, the term {…} in Eq. (2-21) is zero. For fi-
nite radius of curvature R, this corresponds
dg precisely to the critical nucleus, Eq. (2-6 b),
dg (J ) = dJ (2-17)
dJ but now with anisotropic g (J ). This equa-
and with tion therefore determines the shape of the
(2-18) critical nucleus in agreement with the Wulff
d d dJ d d 1 construction (Wulff, 1901). Second, for
dJ = dx ; = ; J =−
dS dS dS dJ dS R very large mobility k (J ) the deviation
from equilibrium {…} may be very small
we obtain in order to produce a desired normal veloc-
dg 1 d ity V^ . We will use this simplification of
dg (J ) = − dx (2-19) equilibrium at the interface
dJ R dS
1
The other term simply gives Dg ≅ (g + g ′′ ) ( 2-dim) (2-22)
R
1 wherever possible, but we will comment on
d ( dS ) = dS dx (2-20)
R the changes due to finite k (J ). The double
Incorporating Eqs. (2-19) and (2-20) into prime means derivatives as in Eq. (2-21).
Eq. (2-16) and integrating Eq. (2-19) by In many cases this seems to be justified
parts, we obtain through comparison with by experimental conditions. On the other
Eq. (2-15) hand, very little is known quantitatively
(2-21) about k (J ). A last point to mention here is
∂x ⎧ 1⎛ d g ⎞⎫
2
our assumption that the interfaces are not
≡ V⊥ = k (J ) ⎨Dg ( S ) − ⎜ g + ⎟⎬
∂t ⎩ R⎝ dJ 2 ⎠ ⎭ faceted at equilibrium. If they are faceted,
as crystals typically are at low tempera-
as an explicit local equation for the ad- tures in equilibrium with their vapor, the
vancement of an interface in two dimen- situation is considerably more complicated
sions with normal velocity V^ , anisotropic and not completely understood (Kashuba
kinetic coefficient k (J ), surface Gibbs en- and Pokrovsky, 1990).
ergy g (J ), and jump Dg = gL – gS of Gibbs It is now generally believed that faceted
energy density at position S along the inter- surfaces undergo a kinetic roughening tran-
face (Burkhardt et al., 1977). Here, the sition even at small driving forces, so that a
Gibbs energy density g corresponds to an rough surface is present under growth con-
2.2 Basic Concepts in First-Order Phase Transitions 91
ditions. The concept outlined above then where cp is the specific heat of the liquid at
generally applies in an effective fashion. constant pressure. If, instead of tempera-
Actually, facets may persist over experi- ture or heat, a second chemical component
mentally relevant length scales, which are is diffusing, a similar transformation to
not covered by this analysis in the kinetic the same dimensionless equations can be
region. A summary of theoretical develop- made. This is described in Section 2.5.2.
ments can be found in the article by Krug We would like to stress the importance of
and Spohn (1992). such a scaled representation as it allows us
To complete this section on the basic to compare at a glance experimental situa-
theoretical ingredients, we will now dis- tions with different sets of parameters.
cuss the influence of diffusive transport of In this dimensionless form, the equation
heat or matter on the propagation of a of motion is
solid – liquid interface (see also Langer,
∂
1980 a). u = DT ∇ 2 u (2-25)
∂t
2.2.3 Growth of Simple Crystal Forms and the boundary conditions are
∂ ∂
T ( x, t ) = DT ∇ 2 T ( x, t ) (2-23) V = − DT u ( z, t ) at interface (2-28)
∂t ∂z
with thermal diffusion coefficient DT and The growth rate is therefore proportional to
appropriate boundary conditions. At infin- the gradient of the diffusion field at the
ity in the supercooled liquid the tempera- interface in the liquid. As the interface is
ture T• < Tm is prescribed, and Tm is the moving, we conveniently make a coordi-
temperature at the interface. It is now very nate transformation from {z, t} for {z¢, t ¢}
convenient to replace the temperature field into a frame of reference moving along
T by a dimensionless u through the trans- with the interface at z¢ = 0:
formation t
z ′ = z − ∫ V (t ) dt
T ( x, t ) − Tm 0 (2-29)
u ( x, t ) = (2-24)
( Lm / c p ) t′ = t
92 2 Solidification
Performing that transformation, we obtain rate V (or l) is not specified, but seems to
the following for the diffusion equation be arbitrary. This degeneracy is practically
(after having again dropped the primes for eliminated by other effects such as inter-
convenience): face kinetics (Collins and Levine, 1985) or
the density difference between solid and
1 ∂u 2 ∂u
= ∇2 u + (2-30) liquid (Caroli et al., 1984), so that in prac-
DT ∂t l ∂z tice a well-defined velocity will be selected.
which defines a diffusion length
b) Growth of a sphere
l ∫ 2 DT /V (2-31)
The influence of the surface tension g
The interface is now always at z = 0, which when it is incorporated through Eq. (2-21)
makes the boundary condition Eq. (2-26 b) is most easily understood by looking at a
and the continuity Eq. (2-28) definite. spherical crystal. This will not lead to sta-
We can obtain a stationary solution by tionary growth. In order to make the analy-
combining Eq. (2-30) with Eqs. (2-26) and sis simple, we will invoke the so-called
(2-28) when ∂u /∂t = 0: “quasistationary approximation”, by set-
ting the left-hand side in Eq. (2-25) equal
u = exp ⎛ − ⎞ − 1; z Û 0
2z
(2-32) to zero. The physical meaning is that the
⎝ l ⎠
diffusion field adjusts itself quickly to a
This equation is consistent with the as- change in the boundary structure which is,
sumption that u = 0 (T = Tm ) everywhere in however, still evolving in time because the
the solid. continuity equation is velocity-dependent.
This equation describes the diffusion field Of course, this approximation reproduces
ahead of the interface. It varies exponen- the stationary solutions precisely (if they
tially from its value at the interface toward exist), and in addition, it exactly identifies
the value far inside the liquid, so that the the instability of a stationary solution as
diffusion field has a typical “thickness” of l. long as it is not of the Hopf type (i.e., the
Note that we have not used Eq. (2-26 a) critical eigenvalue is not complex). It is
as a boundary condition at infinity, but find generally assumed that the approximation
from Eq. (2-32) that u = – 1 implies the so- is good as long as the diffusion length is
called “unit supercooling”, which corre- large compared with other lengths of the
sponds to evolving pattern.
In a spherical coordinate system, the
T (z = •) = Tm – (Lm /cp ) (2-33) equation of motion then becomes
This basically says that the difference in 1 ∂u ⎧ ∂2 2 ∂⎫
melting enthalpy Lm between liquid and ≈0=⎨ 2 + ⎬u (2-34)
DT ∂t ⎩ ∂r r ∂r ⎭
solid must be compensated by a tempera-
ture difference in order to globally con- as a simple Laplace equation.
serve the energy of the system during this The interface is at radius R0 (t) and is ad-
stationary process. In other words, if Eq. vancing with time. The continuity equation
(2-33) is not fulfilled by Eq. (2-26 a), the requires
process cannot run with a stationary profile
of the thermal field, Eq. (2-32). On the dR0 ∂u
vR ≡ = − DT (2-35)
other hand, a particular value of the growth dt ∂r r = R0 ( t )
2.2 Basic Concepts in First-Order Phase Transitions 93
have to consider deviations from a flat where we have linearized the exp (…) with
interface as we move parallel to the inter- respect to z . Instead of the velocity in Eq.
face in the x-direction. This is provided by (2-27), we must now use Vz = 2 D/l + ∂z /∂t.
Eq. (2-37) as the boundary condition, With this equation and the same lineariza-
where we set b = 0 for simplicity. tion as before
Denoting the deviation of the interface (2-49)
ˆ 2 ˆ
W k z k = − D ( 2 / l ) z k + Dq uˆk + D′ q ′ uˆk′
from z = 0 as z (x, t), we designate the cur-
vature for small amplitudes simply as is obtained for the conservation law, Eq.
K = – ∂ 2 z /∂x 2, where, in agreement with (2-27) (and for small values of z , practi-
Eq. (2-38), the sign of K allows for the cur- cally Eq. (2-28)). Eliminating ûk here using
vature to be positive for a solid protrusion Eq. (2-48) and eliminating q using Eq. (2-47)
into liquid. We follow here the notation of in the limit k l Ⰷ1 or k ≈ q ≈ q¢, we obtain
Langer (1980 a).
In a quasistationary approximation, for W k = V ( q − 2 / l ) − ( Dq + D′ q ′ ) d0 k 2
the sake of simplicity we set the l.h.s. in ≈ k V [1 − 12 (1 + D′ / D) l d0 k 2 ] (2-50 )
Eq. (2-30) equal to zero, which of course
allows for the stationary solution, Eq. (2- This formula describes the basic mecha-
32). It is not difficult to treat the fully time- nism of diffusion-controlled pattern forma-
dependent problem here, but the modifica- tion in crystal growth. Althoug in general,
tions do not change the results substan- diffusion tends to favor homogeneous
tially. We now perturb that solution by structures, in the present case it works in
making a small sinusoidal perturbation of the opposite direction! This is easy to
the flat interface: understand; the foremost points of a sinu-
soidal deformation of the interface can dis-
z (x, t) = zˆk exp (i k x + W k t) (2-45) sipate the latent heat of freezing by a larger
Similarly, a perturbation of the diffusion space angle into the liquid than the points
field is made in the liquid and in the solid inside the bays. The latter points will there-
fore be slowed down. The tip points will be
u ( x , z, t ) = exp ⎛ − ⎞ − 1
2z accelerated in comparison with the average
(2-46 )
⎝ l ⎠ rate of advancement.
+ uˆk exp (i k x − q z + W k t ); The formula therefore consists of a de-
u ′ ( x , z, t ) = uˆk′ exp (i k x + q ′ z + W k t ) stabilizing part leading to positive Wk and
a stabilizing part controlled by the capillar-
where the unprimed form is for the liquid, ity d0 . The stabilization is most efficient
and the primed values are for the solid. at large k values or short wavelengths; at
Inserting Eq. (2-46) into Eq. (2-30) with longer wavelengths, the destabilization due
∂u /∂t = 0, we obtain to diffusion into the supercooled liquid
dominates.
– 2 q /l + q 2 – k 2 = 0
The dividing line is marked by the com-
2 q¢/l ¢ + q¢ 2 – k 2 = 0 (2-47) bination of d0 l, which is most conveniently
expressed as
Replacing z in Eq. (2-46) by z (x, t) from Eq.
(2-45), we can insert this into the boundary l S = 2 p /kS = 2 p d0 l (2-51)
condition Eq. (2-37) for b = 0 to obtain
assuming D = D¢. This will be called the
uˆk′ = − 2 zˆk / l + uˆk = − d0 k 2 zˆk (2-48) “stability length”, which gives a measure
2.3 Basic Experimental Techniques 95
melting. In case i), nucleation is essential ferent, and some of its important elements
in controlling the microstructure (the grain will be described below.
size), while in case ii), nucleation is of mi-
nor importance.
Typical technical processes with respect 2.3.1 Free Growth
to these two classes of solidification are as Slow growth rates
follows:
In free growth, the undercooling DT or
i) casting into low conductivity molds
the temperature of the melt T0 = (Tm – DT )
(ceramics), producing small temperature
is imposed on the system, and the growth
gradients, or adding inoculating agents or
rate V and the microstructural scale are the
stirring the melt, thereby producing many
dependent variables. For low undercool-
heterogeneous or homogeneous nuclei;
ings, and therefore slow growth rates, all
ii) processes with high heat flux imposed
three variables can be measured precisely
through strong cooling of the solid such as
if the substance is transparent. The crystal-
continuous casting, welding, laser treat-
lization of non-transparent metals, how-
ment, or through heating of the melt and
ever, is the major issue in solidification
cooling of the solid, such as in single crys-
studies relevant to technical applications.
tal growth or Bridgman type directional so-
Instead of investigating the crystalliza-
lidification experiments (see also Flem-
tion of metallic systems directly, suitable
mings, 1991).
model substances have to be found. These
In free growth, the undercooling DT of are generally organic “plastic crystals”
the melt is given and controls the growth which, like metals, have simple crystal
rate V and the scale (spacing l or tip radius structures and low melting entropies (Jack-
R) of the forming microstructure. In direc- son, 1958). One of the substances which
tional growth, the rate of advance of the behaves very similarly to metals and also
isotherms is imposed by heat flux leading has well characterized properties is succi-
to an imposed growth rate (constrained nonitrile (SCN) (Huang and Glicksman,
growth). This, in turn, controls the inter- 1981). Numerous results have been ob-
face temperature (undercooling) and the tained with this material by Glicksman and
microstructural scales. These three vari- co-workers. Their careful experimental ap-
ables (temperature, growth rate, and micro- proach not only produced the most precise
structural size) have to be measured experi- measurements known at that time but has
mentally for a quantitative comparison of also stimulated new ideas about possible
theory and experiment. A number of mate- mechanisms of structure formation through
rial parameters of the alloy system have to the discrepancy found between the obser-
be known, such as the solid – liquid interfa- vations and the predictions of previous the-
cial energywith its anisotropy, the diffusion ories.
coefficient, the stable and metastable phase Fig. 2-2 shows the experimental setup
diagram, etc. developed by Glicksman et al. (1976) to
The experimental techniques are divided study free dendritic solidification. After
into two classes; those which produce rela- zone melting, the material is introduced
tively small growth rates and those which into the growth chamber (C), which is then
aim for high solidification rates. The corre- slowly undercooled with the heaters (A)
sponding experimental setup is quite dif- and (B) in order to avoid premature crystal-
2.3 Basic Experimental Techniques 97
fore metastable, liquid. A relatively straight- Saito, 1990; Xiao et al., 1988). For these
forward stability analysis (Mullins and Se- processes, some specific assumptions are
kerka, 1963, 1964; Langer, 1980 a) shows made about the incorporation of atoms into
that the solid tends to deviate from spheri- the advancing interface which are not eas-
cal form as soon as its radius R has become ily carried over into the notion of surface
larger than seven times the critical radius tension, etc. Another line of closely related
Rc = 2 d0 /D . A time-dependent analysis has problems concerns the viscous flow of two
also been carried out (Yokoyama and Ku- immiscible liquids (Saffmann and Taylor,
roda, 1988). During the further evolution, 1958; Brener et al., 1988; Dombre and Ha-
these deformations evolve into essentially kim, 1987; Kessler and Levine, 1986 c). We
independent arms, the primary dendrites. will briefly refer to this in Sec. 2.5.
The growth of such dendrites is a very
widespread phenomenon, as will become
2.4.1 The Needle Crystal Solution
clearer in Sec. 2.5 on directional solidifica-
tion. There are also close relations to pro- In this subsection, we will look at an iso-
cesses called diffusion-limited aggregation lated, needle-shaped crystal growing under
(DLA) (for a review see Meakin, 1988; for stationary conditions into a supercooled
relation to crystal growth, see Uwaha and melt. A stationary condition, of course, only
102 2 Solidification
holds in a frame of reference moving at ve- lattice, but they are not directly related
locity V in the positive z-direction. For a one- (Burkhardt et al., 1977). For the moment,
component crystal, the latent heat of freez- we will assume b = 0, i.e., the interface ki-
ing must be diffused away into the liquid. netics should be infinitely fast. Even then,
The process is therefore governed by the the curvature K in Eq. (2-54) is a compli-
heat diffusion Eq. (2-25) cated function of the interface profile. The
simplest approximation for the moment,
2 ∂u
0 = ∇2 u + (2-52) therefore, is to ignore both terms on the
l ∂z r.h.s. of the boundary condition, which cor-
where we have used a dimensionless form, responds to setting the surface tension to
Eq. (2-30), for a stationary pattern together zero. The boundary condition is then sim-
with the definition of the diffusion length, ply a constant (= 0 in our notation).
Eq. (2-31). If the diffusion constants in the A second-order partial differential equa-
crystal and the liquid are different, we tion such as the diffusion equation can be
would have to use two diffusion lengths. decomposed in the typical orthogonal coor-
The continuity Eq. (2-27) at the interface is dinate systems, and we therefore obtain a
for normal velocity closed form solution for the problem with a
boundary of parabolic shape: a rotational
V^ = – DT n̂ · —u L + DT n̂ · —u S (2-53) paraboloid in three dimensions and a
simple parabola in two dimensions. This is
where the subscripts L and S denote gra-
the famous “Ivantsov” solution (Ivantsov,
dients taken on the liquid side and on the
1947).
solid side of the interface. This is known as
The straightforward way to look at this
the “two-sided” model (Langer and Turski,
problem is as a coordinate transformation
1977) and, because of identical diffusion
from the cartesian {x, z} frame, where z is
constants, as the “symmetrical” model. The
the growth direction, to the parabolic coor-
generalization to different diffusion con-
dinates {x , h }:
stants is simple.
The boundary condition at the interface is x = (r – z)/r ; h = (r + z)/r
------
u I = – d0 K – b V (2-54) where r is the radial distance ÷x 2 + z 2 from
the origin, and r is a constant. After trans-
and at infinity forming the differential operators in Eqs.
u• = – D (2-55) (2-52) and (2-53) to {x , h } (Langer and
Müller-Krumbhaar, 1977, 1978), it can be
as introduced before in Eqs. (2-37) and (2- seen immediately that h (x ) = h S = 1 for the
41), respectively. Note that a constant like interface is a solution to the problem con-
D may be added on the right-hand side of firming the parabolic shape of the inter-
both equations without changing the results face, with r being the radius of curvature at
apart from this additive constant in the u- the tip.
field, as frequently appears in the litera- This Ivantsov radius, r , is now related to
ture. The definition of D follows easily the supercooling D and the diffusion length
from Eqs. (2-24) and (2-41). l by the relation
In general, the capillary length d0 and
the kinetic coefficient b are anisotropic be- ⎧ P eP E1 ( P ) 3-dim (2-56 a)
D=⎨ P
cause of the anisotropy of the crystalline ⎩ p P e erfc ( P ) 2-dim (2-56 b)
2.4 Free Dendritic Growth 103
∞
u ( x , z, t ) = ∫ dt ′ ∫ d x ′ G [ x − x ′, z − z ( x ′, t ′ )
e−t −∞
E1 ( P ) = ∫ t
dt (2-58) ⎧ ∂z ⎫
P + V (t − t ′ ), t − t ′ ] ⎨V + ⎬
⎩ ∂t′ ⎭
and erfc is the complement of the error
function. (Langer, 1987 b), with the Green’s function
Eq. (2-56) may be interpreted as an ex-
1 ⎛ | x |2 + z 2 ⎞
pression of supercooling in terms of the G ( x , z, t ) = exp ⎜− ⎟
Péclet number. This explanation is impor- ( 4 p D t )3 / 2 ⎝ 4 Dt ⎠
tant because the following considerations (2-60)
of surface tension give only very small cor- for diffusion in an infinite three-dimen-
rections to the shape of the needle crystal. sional medium (symmetrical case), while z
Therefore, Eq. (2-56) will be also valid is the z coordinate of the interface. The
with non-zero surface tension at typical ex- source term {…} in this equation is obvi-
perimental undercoolings of D Ó 10–1. An ously the interface in the frame of refer-
important consequence is that the scaling ence moving at velocity V in the z-direc-
results derived below then become inde- tion. This equation already contains the
pendent of the dimensionality (2 or 3), if conservation law or the continuity equation
the supercooling is expressed through the at the interface. Furthermore, it is valid
Péclet number. everywhere in space and, in particular, at
The basic result of Eqs. (2-57) and (2- the interface z = z (x, t), where the l.h.s. of
48) is that the growth rate of this parabolic Eq. (2-59) is then set equal to Eq. (2-54).
needle is inversely proportional to the tip Here u is assumed to vanish at infinity, so
radius, but no specific velocity is selected. D must be added on the l.h.s.
For experimental comparison, we make a In the two-dimensional case and for
fit of the tip shape to a parabola. The tip stationary conditions, this can be rewritten
radius of that parabola is then compared as
with the Ivantsov radius r. The actual tip
radius will be different from the Ivantsov d0
D− K {z} = P G2 {P, x , z ( x )} (2-61)
radius of the fit parabola, because of sur- r
104 2 Solidification
period of the evolution. This initial devia- portant intermediate asymptotic exists if
tion can be handled by a linear theory start- the size of the 2D pattern is still much
ing from a mode expansion which takes the smaller than a diffusion length (D t)1/2.
same form as Eq. (2-80) with, in principle, This 2D Laplacian problem (with a fixed
arbitrary coefficients. These amplitudes flux from the outside) was solved, both nu-
then have to be chosen according to the merically and analytically, by Almgren et
predictions of selection theory (Ben Amar al. (1993) who were interested in aniso-
and Brener, 1993) in order to provide the tropic Hele – Shaw flow. They found that
matching to the tip region. As ‘time’ goes after some transition time the system
on, the deviation increases owing to the shows an asymptotic behavior which is in-
Mullins – Sekerka instability and a nonlin- dependent of the initial conditions and in-
ear theory must take over. We can guess volves the formation of four well-devel-
what the long-time behavior of the system oped arms. The length of these arms in-
will be. Indeed, this two-dimensional prob- creases in time as t 3/5 and their width in-
lem is precisely the same as that which creases as t 2/5. The basic idea that explains
leads to two-dimensional dendritic struc- these scaling relations is that the stability
tures. Four well-developed arms (for cubic parameter s 2 = 2 D d0 /(r 22 v2 ) is supposed to
symmetry) grow with a constant speed in be equal to s 2* (e ) even though both v2 and
the directions favored by the surface en- r 2 depend on time. Moreover, the growing
ergy anisotropy. Each arm has a parabolic self-similar shape of the arms was deter-
shape, its growth velocity, v2 = 2 D P22 (D ) mined. In terms of the dendritic problem it
¥ s 2* (e )/d0 , and the radius of curvature of reads (Brener, 1993)
its tip, r 2 = d0 /(P2 s 2* ), are given by 2D se- 1/ 5 2/3
⎛ s*⎞ ⎛ x ⎞
lection theory, where the anisotropic sur- y ( x , z ) = ( 5 | z | / 3) 2/5
⎜ *⎟ ⎜x ⎟
face energy again plays a crucial role. The ⎝s2 ⎠ ⎝ tip ⎠
Péclet number P2 = r 2 v2 /2 D is related to 1
ds
the undercooling D by the 2D Ivantsov for- × ∫ (2-83 a)
mula, which for small D gives x / x tip s 2/3
1 − s4
P2 = (D ) = D2/p (2-81) where the tip position x tip of the arm is
The selected stability parameter s 2 = given by
d0 /(P2 r 2 ) depends on the strength of the x tip (z) = (5 |z |/ 3)3/5 (s 2*/s * )1/5 (2-83 b)
anisotropy e , and s 2* (e ) µ e 7/4 for small e
as was explained above. Replacing t by The ratio s 2* (e )/s * (e ) is independent of e
|z |/v and reducing all lengths by r, we can in the limit of small e. This means that
present the shape of one-quarter of the the shape, Eq. (2-83), in the tail region is
interface (except very close to the dendritic almost independent of the material and
backbone) in the form growth parameters, as well as shape, Eq.
(2-80), in the tip region (if all lengths are
v2 y 2 r
x ( y, z ) = | z | − (2-82) reduced by r ).
v 2 r2 Recent experiments on 3D dendrites
This asymptotic describes the strongly an- (Bisang and Bilgram, 1995) and numerical
isotropy interface shape far behind the tip simulations (Karma and Rappel, 1998) are
and does not match, for small D , the shape in very good agreement with these theoret-
described by Eq. (2-80). In this case an im- ical predictions.
108 2 Solidification
There are, of course, important differ- uid interface. In Fig. 2-8, we show a typical
ences between two and three dimensions, dendrite with side branches resulting from
as a three-dimensional needle crystal is not the time-dependent calculations (Saito et
necessarily rotationally symmetric around al., 1987, 1988) (compare with the experi-
its axis. Snow crystals, for example, show mental result by Glicksman et al. (1976) in
large anisotropies in directions orthogonal Fig. 2-9). The profile is symmetric around
to the growth direction of the primary den- the axis by definition of the calculation. An
dritic needle (Yokoyama and Kuroda, approximately parabolic tip has been
1988). In the immediate neighborhood of formed from which side branches begin to
the tip, however, the deviation from this ra- grow further down the shaft (only the early
tional symmetry is often small. Therefore, stage of side-branch formation was consid-
we may work with this two-dimensional ered). They have a typical distance which,
model by using an “effective” capillary however, is not strictly regular.
length. The scaling relations given below As a first result, the scaling relation, Eq.
are expected to be insensitive to this apart (2-68), was checked using the Péclet num-
from a constant prefactor of order unity in ber from Eq. (2-56 b). Experimentally, this
the s (e )-relation (Kessler and Levine, requires the anisotropic capillary length
1986 b, d; Langer, 1987 a). and the supercooling to be varied indepen-
The numerical simulations were per- dently. In Fig. 2-10, the scaled numerical
formed for a two-dimensional crystal – liq- results are shown as symbols for two dif-
Figure 2-11. Tip radius of free dendrite over Ivant- Figure 2-12. Numerical scaling result confirming
sov radius plotted versus anisotropy as a function of V R 2 = const. for free dendritic growth independent
dimensionless supercooling. The numerical results of supercooling D, depending on anisotropy e only
(see also Fig. 2-8, 2-10) are consistent with the pre- (Saito et al., 1988).
dictions from the needle solution (Ben Amar and
Moussallam, 1987).
order to check the scaling form Eq. (2-68) only. This is precisely the relation, Eq. (2-
involving the radius rather than the Péclet 1), derived from qualitative considerations
number. The Ivantsov parabola and its ra- in the introduction to this chapter. This re-
dius r basically originate from a global lation has been confirmed by the analysis
conservation law for the quantity (heat) re- of many experiments (Huang and Glicks-
leased at the interface, while the tip radius man, 1981).
R is a local geometric quantity. In practical The final point to be discussed here con-
experiments, by fitting a parabola to the cerns the side branches and their origin,
tip, we can interpolate between these two spacing and amplitudes. This issue is theo-
numbers, the result depending on how far retically not completely resolved, because
down the shaft the fitting parabola is used. none of the available analytical approxima-
Using the actual radius R rather than the tions can correctly handle the long-wave-
Ivantsov radius r , perfect scaling can be length limit of side-branch perturbations.
seen in Fig. 2-12 with respect to supercool- Moreover, the subject is somewhat tech-
ing D , even up to the very large value of nically involved. Therefore, we will only
D = 0.5. Since for smaller supercoolings, summarize the main arguments below and
D Ù 0.1, the difference between r and R be- refer to the above-mentioned numerical
comes negligible, as shown in Fig. 2-11, simulations (Saito et al., 1988) for compar-
and we may safely use Eq. (2-68) as ison with experiments.
An important quantity which charac-
2 Dd0
V R2 ≈ V r 2 = = constant (2-90) terizes the stability of flat moving interface
s (e ) ripples is the so-called stability length
independent of supercooling D , to interpret
l S = 2 p l d0 (2-91)
experiments and to make predictions. The
term “constant” here means that the prod- where d0 is the capillary length, and l the
uct V R 2 depends on material parameters diffusion length. Perturbations of wave-
112 2 Solidification
lengths l longer than l S will grow, while mental anisotropy of the material succino-
shorter wavelengths will decay with time. nitrile (Huang and Glicksman, 1981),
This quantity characterizes the competition e ≈ 0.1, gave the same s -value as the theo-
between the destabilizing diffusion field retical concept, and since e cannot be var-
through l against the stabilizing surface ied easily, there was no discrepancy.
tension through d0 . A derivation of this To summarize these results, it appears
Mullins – Sekerka instability was given in that the scaling relation, Eq. (2-92), shown
Sec. 2.2.4. in Fig. 2-13 from the numerical solution of
It is natural to assume that this length the model in two dimensions, is in agree-
scale is related to the formation of side ment with the experimental results.
branches. A direct estimate of the typical We will now give a somewhat qualitative
wavelength l 2 of the side branches is explanation of the mechanism of side-
branch formation as far as this can be de-
l 2 ⲏ l S = 2 p l d0 = 2 p r s (2-92) duced from the theoretical approaches. A
The remarkable result of the numerical linear stability analysis (Langer and Mül-
simulation is shown in Fig. 2-13. Appar- ler-Krumbhaar, 1978, 1980; Kessler and
ently, the ratio l 2 /l S is a constant of ap- Levine, 1986 a; Barber et al., 1987; Bar-
proximately 2.5, which is independent of bieri et al., 1987; Bensimon et al., 1987;
supersaturation and anisotropy. This seems Caroli et al., 1987; Kessler et al., 1987;
to be in quite good agreement with experi- Pelce and Calvin, 1987) indicates that the
ments (Glicksman et al., 1976; Dougherty relevant modes for side-branch formation
et al., 1987; Honjo et al., 1985; Huang and in the frame of reference moving with the
Glicksman, 1981). tip consist of an almost periodic sinusoidal
The experimental comparison was made, wave travelling from the tip down the shaft
in fact, with an older theoretical concept such that they are essentially stationary in
(Langer and Müller-Krumbhaar, 1978), the laboratory frame of reference (Langer
which did not correctly consider aniso- and Müller-Krumbhaar, 1982; Deissler,
tropy. By accident, however, the experi- 1987). The amplitude of these waves is not
constant in space, but first grows exponen-
tially in the tip region (Barbieri et al.,
1987; Caroli et al., 1987). The exponential
increase of that envelope in the tip region
depends on the “wavelength” of the oscil-
latory part (Bouissou et al., 1990).
In contrast to the earlier analysis by
Langer and Müller-Krumbhaar, all these
modes are probably stable, so that without
a triggering source of noise, they would
decay, and a smooth needle crystal would
result. Some driving force in the form
of noise due to thermal or hydrodynamic
Figure 2-13. Scaled sidebranch spacing l = l 2 , nor-
fluctuations is needed to generate side
malized with Ivantsov radius R0 and s (e )1/2, plotted
versus capillary anisotropy for two supercoolings. branches, but apparently this is usually
No dependence on e or D is found, as expected (Saito present. Estimates of the strength of these
et al., 1988). fluctuations (Langer and Müller-Krumb-
2.4 Free Dendritic Growth 113
haar, 1982; Barbieri et al., 1987; Langer, tual non-axisymmetric shape of the needle
1987 a) are still somewhat speculative. crystal, defined by Eqs. (2-83 a, b), was
Given such a small noise at the tip, the given by Brener and Temkin (1995). They
exponentially increasing envelope over the found that the root-mean-squared ampli-
side branches into the direction of the tail tude for the side branches generated by
then amplifies that noise so that the side thermal fluctuations is
branches become visible. This happens
over a range of about two to ten side- ⎧ 2 x 3/ 2 ⎫
〈x12 ( Z , Y )〉1/ 2 ~ Q exp ⎨ 1 / 2 ⎬ (2-93)
branch spacings. The actual selected wave- ⎩ 3 3s | z | ⎭
length of the side branches in that tip re-
gion (assuming a white noise, triggering all where the function x (z) is given by the
modes equivalently), according to these underlying “needle” solution, Eq. (2-83 b),
–
considerations, is defined by the mode with and the fluctuation strength Q is given by
–2
the largest amplitude at a distance of about Langer (1987 a), Q = 2 kB T 2 cp D/(L2m v r 4 ).
one “wavelength” away from the tip. This The root-mean-squared amplitude for
is the product of the average amplitude due the side branches increases with the dis-
to noise at the tip and the amplification fac- tance from the tip, | z |. This amplitude
tor from the envelope. grows exponentially as a function of
Langer (1987 a) described the side- (|z | 2/5/s 1/2 ). The important result is that
branching deformation as a small (linear) the amplitude of the side branches for the
perturbation moving on a cylindrically anisotropic needle grows faster than for
symmetric needle crystal (Ivantsov parabo- the axisymmetric paraboloid shape. In
loid). The noise-induced wave packets the latter case x (z) = 2 |z |1/2 and the ampli-
generated in the tip region grow in ampli- tude grows exponentially as a function
tude, spread and stretch as they move down of (|z |1/4/s 1/2 ). This effect resolves the
the sides of the dendrite producing a train puzzle that experimentally observed side
of side branches. In the linear approxima- branches have much larger amplitudes than
tion, the amplitude grows exponentially can be explained by thermal noise in the
and the exponent is proportional to |z | 1/4. framework of the axisymmetric approach.
These results are in approximate, qualita- Agreement with experiment now is indeed
tive agreement with available experimental very good (Bisang and Bilgram, 1995).
observations (Huang and Glicksman, 1981; Far down from the tip the side-branching
Dougherty et al., 1987; Bisang and Bil- deformations grow out of the linear regime
gram, 1995), but experimentally observed and eventually start to behave like den-
side branches have much larger amplitudes drites themselves. It is clear that the
than explicable by thermal noise in the branches start to grow as free steady-state
framework of the axisymmetric approach. dendrites only at distances from the tip
This means that either the thermal fluctua- which are of the order of the diffusion
tion strength is not quite adequate to pro- length which, in turn, is much larger than
duce visible side-branching deformations, the tip radius r in the limit of small P. This
or agreement with experiment would re- means that there is a large range of z,
quire at least one more order of magnitude 1 Ⰶ|z | Ⰶ 1/P, where the side branches al-
in the exponential amplification factor. ready grow in the strongly nonlinear re-
The description of the side-branching gime, but they do not yet behave as free
problem, which takes into account the ac- dendrites. We can think of some fractal ob-
114 2 Solidification
into two parts about its central axis with a his co-workers made an essential contribu-
narrow liquid groove between the two tion to our understanding of dendritic
parts, and triplons in three dimensions are growth in pure undercooled melts (Glicks-
split into three parts. For a finite aniso- man et al., 1976; Huang and Glicksman,
tropy, however, these structures only exist 1981). This research was initially con-
above a critical undercooling (or supersat- cerned mainly with highly purified succi-
uration for the isothermal solidification of nonitrile (SCN). It was extended to cyclo-
an alloy), such that standard dendrites hexanol (Singh and Glicksman, 1989), wa-
growing along ·100Ò directions are indeed ter (Fujioka, 1978; Tirmizi and Gill, 1989),
the selected structures in weakly aniso- rare gases (Bilgram et al., 1989), and to
tropy materials at low undercoolings, in other pure substances with a crystal aniso-
agreement with most experimental obser- tropy different from SCN such as pivalic
vations in organic and metallic systems. acid (PVA) (Glicksman and Singh, 1989).
From a broad perspective, the existence of Work on free growth of alloys includes
doublons and triplons is of fundamental NH4Cl – H2O (Kahlweit, 1970; Chan et al.,
importance because it has provided a basis 1978), NH4Br – H2O (Dougherty and Gol-
on which to classify the wide range of pos- lub, 1988), SCN with acetone and argon
sible growth morphologies that can form as (Glicksman et al., 1988; Chopra et al.,
a function of undercooling and anisotropy 1988), PVA – ethanol (Dougherty, 1990),
(Brener et al., 1996). and others.
We have so far ignored the influence of The specific merit of the work of Glicks-
the kinetic coefficient b in Eq. (2-88). This man et al. was that the systems for which
omission is not likely to be very important they characterized all the properties, in-
for low growth rates, but for fast growth cluding surface energy, diffusion coeffi-
rates, as in directional solidification, b cient, phase diagram etc., have been exam-
should be taken into account. We will re- ined. This led to clear evidence in the mid-
turn to this point in Sec. 2.5. 1970s that the theory of that time (using
extremum arguments for the operating
point of the tip) was not able to describe
2.4.3 Experimental Results
the results quantitatively.
on Free Dendritic Growth
At the same time, Müller-Krumbhaar
The answer to the question of whether and Langer worked on precisely the same
dendritic growth is diffusion-controlled problem and proposed a theory based on
or controlled by anisotropic attachment ki- the stability of the growing dendrite tip,
netics, was sought by Papapetrou (1935), called the marginal stability criterion (Lan-
who was probably the first to make system- ger and Müller-Krumbhaar, 1977, 1978).
atic in situ experiments on free dendritic Most of the existing experimental data
growth. He examined dendritic crystals could be fitted using this criterion. Despite
of transparent salts (KCl, NaCl, NH4Br, the fact that this theory incorrectly ignored
Pb(NO3)2 , and others) under a microscope the important role of anisotropy (as we
in aqueous solutions and proposed that the know now), it inspired a number of new
tip region should be close to a paraboloid experiments and also attracted the interest
of rotational symmetry. of other physicists.
Many years later, the extensive and As has been said before, today’s theory
systematic experiments by Glicksman and is consistent with the older approximate
116 2 Solidification
theory if we allow for a s (e ) value that var- perturbed tip of PVA is longer, with a
ies with the anisotropy of the capillary sharper delineation of the crystallographic
length. The corresponding central equation orientations. Glicksman and Singh (1989)
for dendritic growth (Eq. (2) in Kurz and found that PVA has a ten-fold larger sur-
Trivedi, 1990) should therefore still apply. face energy anisotropy than SCN (see Ta-
ble 2.1). The tip radii and growth rates as a
function of undercooling for both sub-
Pure substances (thermal dendrites) stances scale well when using the values
Fig. 2-15 shows dendrites of two differ- 0.22 and 0.195 for s , respectively (Fig. 2-
ent transparent materials with cubic crystal 16). According to solvability theory, the
structure: face-centered cubic PVA and great difference in the anisotropy constant
body-centered cubic SCN (Glicksman and e should make a larger difference in s (e )
Singh, 1989). Qualitatively, both dendrites (compare with Fig. 2-10). The reason for
look similar, but their branching behavior this discrepancy is not known, and we have
shows some important differences. The un- to leave this point to future research.
(a) (b)
Figure 2-15. Dendrite morphologies of two transparent materials with small melting entropies and cubic crys-
tal structures (plastic crystals); (a) pivalic acid (PVA) and (b) succinonitrile (SCN) (Glicksman and Singh,
1989).
2.4 Free Dendritic Growth 117
* Due to differences in the definitions of s * these values, as given in the corresponding literature, are smaller
by a factor of 2 with respect to the values used by Dougherty and Gollub (1988) and defined in this paper. The
values given here have been obtained by multiplying the original data by a factor of 2 in order to compare with
the same (one-sided) model. (See also Fig. 2-10.)
Koss et al. (1999) have shown that even behind the tip is delayed up to about seven
in microgravity environment there is a radii in PVA. This is quite consistent with
small but significant difference between the recent calculations discussed in Sec.
transport theory and experiment. Again 2.5.4. The ratio of initial secondary arm
this has to be left to the future. spacing l 2 over tip radius R is also indi-
The secondary branch formation which cated in Table 2-1.
starts in SCN at a distance of three tip radii
118 2 Solidification
Large undercoolings
Interesting experiments have also been
performed with pure and alloyed systems
under large driving forces, which reach
values beyond unit undercooling (for ex-
(b)
ample by Wu et al. (1987) and especially by
Figure 2-16. Effect of undercooling on (a) tip radius Herlach et al. (1990 – 1999), see Table 2-2).
and (b) growth rate of the two organic materials Some of these are reproduced in Fig.
shown in Fig. 2-15 (a = D). The experimental results
2-18 together with predictions from IMS
superimpose as they are plotted with respect to dimen-
sionless parameters (Glicksman and Singh, 1989). (Ivantsov – marginal stability) theory (Lip-
ton et al., 1987; Trivedi et al., 1987; Boet-
tinger et al., 1988) (with s (e ) = 0.025) and
including interface attachment kinetics.
2.4 Free Dendritic Growth 119
(a) (b)
Figure 2-17. Effect of dimensionless composition at constant undercooling of 0.5 K (2.1% of unit undercool-
ing) on (a) dimensionless growth rate and (b) on dimensionless tip radius for free dendritic growth in SCN – ace-
tone alloys (Lipton et al., 1987). Points: experiments (Chopra et al., 1988); solid line: LGK model (Lipton et al.,
1987); interrupted line: Karma and Langer (1984) model.
These results are specifically of interest be- Table 2-2. Dendritic growth velocity measurements in
cause two different techniques, one for the highly undercooled melts; comparison between experi-
ment and theory (Herlach and coworkers, 1999–2000).
determination of the diffusive speed and the
other for the undercooling – growth rate re- Metals and alloys
lationship of the dendrites, have been cou- Co – Pd Volkmann et al. (1998)
pled in order to make the analysis as free Co – V Tournier et al. (1997)
from adjustable parameters as possible. Up Cu Li et al. (1996)
Ni Eckler and Herlach (1994)
to undercoolings of 200 K there is reason-
Ni – Al Barth et al. (1994),
able agreement between experiment and Assadi et al. (1998)
theory using the best fit for the measured Ni – B Eckler et al. (1991 a, 1992, 1994)
diffusive speed and the liquid diffusivity, Ni – C Eckler et al. (1991 b)
VD = 26 m/s and DL = 2.7 ¥ 10– 9 m2/s (Ar- Ni – Si Cochrane et al. (1991)
Ni – Zr Schwarz et al. (1997),
nold et al., 1999). The lower V (∆T )-curve
Arnold et al. (1999)
is for a constant (equilibrium) value of the
distribution coefficient k (V) = ke showing Intermetallics
the importance of the appropriate velocity CoSi Barth et al. (1995)
FeAl Barth et al. (1997)
dependence of the distribution coefficient. FeSi Barth et al. (1995)
At higher undercoolings other phenom- Ni3 Al Assadi et al. (1996)
ena take over. Nix Sny Barth et al. (1997)
One example is the grain refinement be- NiTi Barth et al. (1997)
yond a certain undercooling. This structure Nix Tiy Alz Barth et al. (1997)
has been explained by dendrite fragmenta- Semiconductors
tion due to morphological instability of the Ge Li et al. (1995 a, 1996)
fine dendrite trunks (Schwarz et al., 1990; Ge – Cu Li and Herlach (1996)
Ge – Si Li et al. (1995 b)
Karma, 1998).
Ge – Sn Li and Herlach (1996)
120 2 Solidification
tracted. At low solidification rates, the flat The Helmholtz energy F is then
interface is stable; above a critical speed, it
F (T, V, Ni , …) = U – T S ;
becomes unstable against the formation of
ripples, cells and dendrites. dF = – S dT – P dV + S m i dNi (2-97)
i
In the next section, a few thermody-
and the often-used Gibbs energy G is
namic questions related to interface prop-
erties in two-component systems are con- G (T, P, Ni , …) = U – T S + P V = S m i dNi ;
sidered, before describing patterns in direc- i
dG = – S dT + V dP + S m i dNi (2-98)
tional solidification. i
number of NS particles in the solid and NL lar, the spatial continuity of chemical po-
particles in the liquid, whose composition tentials together with continuous tempera-
is still undetermined. Define the number n ture and pressure guarantees local thermal
as the number of B particles in the solid. equilibrium, which we will assume to hold
Keeping NS and n initially fixed, the Gibbs in most of the following discussions.
potential is obtained as We now come to the discussion of a typ-
ical phase diagram for a two-component
G (NA , NB , NS , n) = NS gS (CS ) + NL gL (CL ) system (Fig. 2-19). The vertical axis de-
+ 4 p R2 g (2-101) notes the temperature, the horizontal axis
the relative concentration of “solute” in a
where g is the surface free energy density, “solvent”, or, more generally, B atoms rel-
R is the radius of the solid sphere ative to A atoms.
NS = –43 p R 3 (unit atomic volume), and gS At high temperatures, T > T0 , the system
and gL are free energy densities for homo- is liquid, regardless of concentration. As-
geneous solids or liquids at concentrations suming a concentration C• to be given in-
CS and CL . Removing the constraints on NS itially, we lower the temperature to T1 . At
and n, we obtain thermodynamic equilib-
rium by minimizing G with respect to NS
and n, so that G = G (NA , NB ). Together
with the chemical potentials from Eq. (2-
99), this gives
g
m − m0 = − K (2-102)
( CL0 − CS0 )
this temperature, we first hit the liquidus concentration profile ahead of the interface
line C̃L (T ), and the system begins to solid- which decays exponentially from C L0 at the
ify, producing a solid of very low concen- interface to C• far away from the interface.
tration marked by the solidus line C̃S (T ). But why should it decay to C• (or why
When the temperature is slowly lowered, should the interface choose a temperature
solidification becomes complete at T0 . At position such that C S0 = C• )?
lower temperatures, the whole system is The answer is quite simple, and again
solid. was given in similar form in Sec. 2.2.3, Eq.
The region between C̃S (T ) and C̃L (T ) is (2-33) for the pure thermal case: if C S0 were
the two-phase region: if we prepare a not identical to C• , then during the solid-
system at a concentration between C S0 and ification process there would be either a to-
C L0 at high temperatures and then quickly tal increase (or decrease) of concentration
quench it to T0 , the system starts to sepa- – which clearly is impossible – or at least
rate into a solid phase of concentration C S0 the concentration profile could not be sta-
and a liquid phase of C L0 . In practice, this is tionary.
a very slow process, with lengths varying This is a rather strict condition, which
with time t approximately as t 1/3 (Lifshitz we can reformulate as follows: if we im-
and Slyozov, 1961; Wagner, 1961). pose a fixed temperature gradient ∂T /∂z
In the case of directional solidification, a and move this at fixed speed V0 over a
thermal gradient in the system defines a di- system of concentration C• at infinity in
rection such that the liquid is hot and the the liquid (toward the liquid in the positive
solid is cold. A flat interface may then be z-direction), then the interface will choose
present at a position in space at tempera- a position such that its temperature is at T0 ,
ture T0 . For equilibrium between solid and the concentration in the solid will be C S0
liquid at that temperature, the concentra- (averaged parallel to the interface), and the
tion in the solid must be at C S0 = C• , and in liquid concentration at a flat interface will
the liquid at C L0 = C̃L (T0 ). We now assume be at C L0 . This follows from global conser-
that the liquid at infinity has concentration vation of matter together with the imposed
C• . Clearly there must be a decrease in stationary solidification rate.
concentration as we proceed from the inter- As a final point, we can even derive a
face into the liquid. In order to maintain condition for the stability of the interface.
such an inhomogeneous concentration, the The concentration profile in the liquid will
interface must move toward the liquid. decay exponentially with distance z away
In other words, when the liquid of com- from the interface as
position C L0 freezes, the solid will only
CL (z) = (C L0 – C• ) e – 2 z /l + C• (2-104)
have a concentration C S0 . The difference in
concentrations by analogy to Eq. (2-32). Since we assume
DC = C L0 – C S0 (2-103) the temperature gradient
movement, together with the global con- with segregation coefficient k (equilibrium
servation of matter. We therefore normalize values assumed) defined as
the diffusion field in the liquid to
dT dT
k= (2-113)
C (Liquid) − C∞ dC˜ L dC˜ S
uL = (2-108)
DC
through the slopes of the liquidus and sol-
so that it varies from one to zero in the pos- idus lines. When they intersect at {Tm ,
itive z-direction from the interface at z = 0 C = 0} this is equivalent to the conventional
to z = •. If the interface is not at position definition k = CS /CL , but in the above for-
z = 0 but at z , we must require uL = 1 – z /l T , mulation, k = 1 may also be true for a con-
because at a distance stant jump in concentrations, independent
of temperature.
DC dT
lT = (2-109) The conservation law at the interface
GT dC˜ L z = z finally becomes
the liquidus concentration has reached the V^ {1 + (1 – k) (uL – 1)}
asymptotic value. This is the thermal = – DL n̂ · —uL + DS n̂ · —uS (2-114)
length which we assume to be fixed by the
thermal gradient GT , the concentration where V^ is the interface velocity in direc-
jump DC, and the liquidus line C̃ (T ), tion n̂ normal to the interface. For k = 1, the
which is here assumed to be a straight line brackets {…} give 1, corresponding to a
in the T vs. C diagram. constant concentration jump, while for
The equation of motion in quasi-station- k = 0, they give uL , since for a solid in Eq.
ary approximation then becomes, in anal- (2-112), uS = 0.
ogy to Eq. (2-52) with l = 2 DL /V This standard model for directional so-
lidification (Saito et al., 1989) therefore
1 2 ∂uL consists of Eqs. (2-108) to (2-114), to-
∂ t uL ≈ 0 = ∇ 2 uL + (2-110)
DL l ∂z gether with an additional diffusion equa-
tion as Eq. (2-110) inside the solid phase.
This equation applies equivalently to the
The open point is finally the relation of
solid but with a different chemical diffu-
the capillary length d (Eq. (2-111)) with ex-
sion length l ¢ due to different chemical
perimentally measurable material parame-
diffusion constants. The boundary condi-
ters. As a first step, we interpret the u-field,
tion in analogy to Eq. (2-88) obviously be-
Eq. (2-108), as a scaled form of the chemi-
comes
cal potential m (see Sec. 2.5.1) near T0
uL (z ) = 1 – d K – z (x, t)/l T – b V^ (2-111)
m − m∞
uL = (2-115)
where we now have D = 1 as the first term DC ( ∂m / ∂C˜ L )
on the r.h.s., with curvature K being posi-
assuming that linearizing m around its equi-
tive for the tip of a solid nose pointing into
librium value at the liquidus line C̃L (T0 ) is
the liquid. The capillary length d is dis-
sufficient to describe its dependence upon
cussed below and interface kinetics with
C. By the definitions in Eq. (2-102) to-
b ≠ 0 will be discussed in Sec. 2.5.4. The
gether with Eq. (2-111), we now obtain the
solid boundary condition is simply
capillary length in the form given in Eq.
uS = k (uL – 1) (2-112) (2-86) for the chemical case. Here we have
126 2 Solidification
The result of these investigations is that curve does not define such a normal bifur-
in a diagram of pulling speed V versus cation but rather an inverse bifurcation.
wavelength l there exists a closed curve of This means that immediately above Vc
neutral stability, Fig. 2-20. At fixed V, a large amplitude cells with deep grooves are
small amplitude perturbation of the inter- formed. A time sequence of the evolution
face at a wavelength on that curve neither of a sinusoidal perturbation into elongated
grows nor decays. Perturbations at wave- cells at 1% above Vc due to inverse bifurca-
lengths outside that curve decay, inside the tion is shown in Fig. 2-21. This can be
curve they grow to some finite amplitude. understood theoretically (Wollkind and
This is similar to periodic roll patterns in Segel, 1970; Langer and Turski, 1977; Ca-
the Rayleigh – Benard system of a fluid roli et al., 1982) by means of an amplitude
heated from below (Chandrasekhar, 1961), equation valid near Vc :
but here a maximal speed, Va , is present,
above which a flat interface is absolutely ∂A ⎛ V − Vc ⎞
⎟ A − a1 | A | A + …
2
=⎜ (2-120)
stable. For normal alloys, this speed is very ∂t ⎝ Vc ⎠
high, while for liquid crystals, it is more
easily accessible in controlled experiments where A is the (possibly complex) ampli-
(Bechhoefer et al., 1989). The diffusion tude of a periodic structure exp (i k x) with
length at Va is of the order of the capillary k = 2 p /lc . The coefficient a1 is called the
length. Landau coefficient. If it is positive, we have
At low speeds, but slightly above the
critical velocity, Vc sinusoidal “cells” will
be formed for systems with segregation co-
efficients k near unity. For small segrega-
tion coefficients, however, the neutral
-----
a normal bifurcation with | A | ~ ÷V – Vc , has been made (Brattkus and Misbah,
while for a1 < 0, the third order term does 1990). A phase-diffusion equation has been
not stabilize the pattern but allows very derived describing the temporal evolution
large amplitudes leading to elongated cells of a pattern without complete periodic vari-
(Fig. 2-21), which will be stabilized by ation of the interface. The basic idea is to
some higher-order effects. replace the periodic trial form exp (i k x)
A second phenomenon is usually asso- by a form exp (i Q (x, t)) so that q (x, t) =
ciated with this inverse bifurcation, namely, ∂Q /∂x is now no longer a constant but is
the splitting of the wavelength l c Æ l c /2. slowly varying in space along the interface
Qualitatively, this is understandable from and evolving with time. We can derive a
nonlinear corrections since the squaring of nonlinear phase diffusion equation
the original pattern ~ exp (i k x) produces
∂t q = ∂x {D̃ (q) ∂ x q} (2-122)
terms ~ exp (i 2 k x). This effect has clearly
been observed in experiments (de Che- with a diffusion coefficient D̃ (q) depend-
veigne et al., 1986). ing in a complicated way on q. The proce-
We will discuss some aspects of the very dure is well known in hydrodynamics and
high speed region in Sec. 2.5.7 but devote it is associated there with the so-called
the main part of the following discussions Eckhaus instabilility (Eckhaus, 1965). This
to the most interesting region for practical instability eventually causes an (almost)
purposes, which is not too close to the periodic spatial structure to lose or gain
upper and lower bounds of the growth rate one “period”, thereby slightly changing the
Va and Vc . average wavelength. In directional solidifi-
Approximating by straight lines the neu- cation, the result (Brattkus and Misbah,
tral curve of the logarithmic plot Fig. 2-20 1990) is shown in Fig. 2-22, where velocity
in the intermediate velocity region, we find is plotted against wavenumber in a small
for both the small and the large l limits the interval above the critical velocity. The full
relation line is the neutral curve, the full triangles
mark the Eckhaus boundary of phase
V l 2 ≈ constant (2-121)
stability. A periodic (sinusoidal) pattern is
Again we have recovered the form of Eq. stable against phase slips only inside the
(2-1) mentioned in the introduction as a region surrounded by triangles, thereby al-
scaling law where l here is the cell spac- lowing for an Eckhaus band of stationary
ing. This suggests that the cellular pattern periodic solutions with a substantially re-
formed in actual experiments would also duced spread in wavenumbers as compared
follow this behavior. Unfortunately, this to the linear stability results. Note also that
problem has not yet been settled to a satis- the results for phase stability (dashed line)
factory degree from a theoretical point of from the amplitude equation only hold in
view. This is partly due to the difficulty of an extremely small region above Vc , while
finding good analytical approximations to only 20% above Vc , it shows no overlap
the cellular structures, which makes nu- with the result from the present analysis
merical calculations necessary to a large (triangles). The short-wavelength branch
degree. We will return to this point in Sec. has a very complicated structure, while the
2.5.5. long-wavelength branch far from the
For small amplitude cells obtainable threshold scales as l Eck ~ V –1/2, again like
under normal bifurcation, some progress the neutral curve. This also seems to be in
2.5 Directional Solidification 129
Figure 2-22. Stability diagram V vs. 2 p/l near the lower critical threshold for a flat moving interface in direc-
tional solidification. The solid line is the neutral stability curve Fig. 2-20, the dotted curve is the most danger-
ous mode, the dashed curve is the limit of the Eckhaus stability from the amplitude equation. The triangles mark
the Eckhaus limit as obtained from the full nonlinear analysis (Brattkus and Misbah, 1990), with stable cellular
interfaces possible only inside that region. The band width of possible wavelengths for cellular interfaces ac-
cordingly is smaller by a factor of ≈ 0.4, as compared to the band width given by the neutral (linear) stability
curve.
agreement with experimentally observed narrow, similar to Fig. 2-21 (Ungar and
results, as discussed later (Billia et al., Brown, 1984 a, b, 1985 a, b, Karma, 1986,
1987, 1989; Somboonsuk et al., 1984; Kessler and Levine, 1989, McFadden and
Esaka and Kurz, 1985; Eshelman and Tri- Coriell, 1984, Pelce and Pumir, 1985).
vedi, 1987; Faivre et al., 1989; Kurowsky,
1990).
At higher velocities, the cells quickly be-
come elongated (Fig. 2-23) with deep
grooves forming bubbles. This was first
obtained through numerical calculations by
Ungar and Brown (1984 a, b, 1985 a, b).
Calculations with a dynamical code in
quasi-stationary approximation (Saito et
al., 1989) confirmed the stability of these
structures with respect to local deforma-
tions and short-wavelength perturbations.
The long-wavelength Eckhaus stability has
not been investigated yet for these cells.
All calculations were made in two dimen-
sions, which are believed to be appropriate
for experiments of directional solidifica-
tion in a narrow gap between glass plates.
At higher velocities and wavelengths (or Figure 2-23. Computed example for a deep cellular
cell sizes) not much smaller than the dif- interface at V ≈ 5 Vc with bubble formation at the bot-
fusion length, the grooves become very tom of the groove (Saito et al., 1990).
130 2 Solidification
If the velocity is fixed and the wavelength For comparison with experiments, it is
l is reduced significantly below the diffu- useful to draw a V vs. l diagram (Fig. 2-
sion length l, the Saffmann – Taylor limit is 24). Here the full line is again the neutral
reached (Brener et al., 1988; Dombre and curve, the broken line is the most danger-
Hakim, 1987; Kessler and Levine, 1986 c), ous (or most unstable) mode, and the dot-
which is equivalent to a low-viscosity fluid ted line denotes the relation l = l , where the
being pushed into a channel of width l diffusion length is equal to the imposed
filled with a high-viscosity fluid. The low- wavelength. The asterisks mark some de-
viscosity fluid forms a finger just like the tailed numerical investigations (Saito et al.,
solid in directional solidification. Near the 1989). The asterisk furthest to the left is
tip, the width of the finger l f corresponds to close to the above-mentioned Saffmann –
Taylor limit. At slightly higher wave-
lf = D l
lengths, where l < l still holds, we are in a
with cell spacing l , where D < 1 is the ac- scaling region, where the radius of curva-
tual supercooling at the tip (remember that ture at the tips of the cells is about 1/5 of
D = 1 for a flat interface at z = 0, and D = 0 the cell spacing, as also found experimen-
for a flat interface at z = l T ). This serves tally (Kurowsky, 1990). All these consider-
to verify the consistency of numerical ations give sufficient confidence that the
calculations (Saito et al., 1989). An even numerical calculations may also provide
more detailed analysis was carried out by insight into the mechanism of directional
Mashaal et al. (1990). solidification for the most interesting case
Figure 2-24. “Phase”-diagram log (V ) vs. log (l) for interface patterns in directional solidification. The solid
and dashed curves denote the neutral stability curve, and the dashed-dotted curve the most dangerous mode.
Asterisks mark fixed parameter values discussed hereafter. The lower critical threshold here is Vc ≈ 1, lc ≈ 0.5 for
velocity and cell spacing. For other parameters see text. At low pulling speeds and high wavelengths cellular
patterns with narrow grooves are found (a). At very short wavelengths and moderate speeds cellular patterns
with wide grooves are found, consistent with theories for viscous fingering. At high pulling speeds, such that
the cell spacing l is significantly wider than the diffusion length l, side-branching dendritic arrays are formed
(c) (Saito et al., 1990). The speeds are still much smaller than the absolute stability limit Va .
2.5 Directional Solidification 131
of dendritic arrays formed at higher growth other, until finally they may behave like in-
rates. dividual isolated dendrites.
A few words on numerical methods and In order to test this hypothesis, a series
system parameters may be in order before of numerical experiments were performed
we discuss the dendritic region. The nu- at a fixed cell spacing and increasing pull-
merical code is equivalent to the one used ing velocity (Saito et al., 1989). The nu-
for the free dendritic case with the modifi- merical parameters of the model were rep-
cation that it is necessary to integrate over resentatively taken to correspond to steel
several cells to arrive safely outside the with Cr – Ni ingredients (Lesoult, 1980). In
diffusion length. Furthermore, in principle, dimensionless units, the critical velocity
diffusion has to be considered in both the and wavelength for the plane-front in-
liquid and the solid. Since the diffusion co- stability were Vc = 1.136, l c = 0.514. The
efficient for material in the solid is usually anisotropy of the capillarity length was not
much lower than it is in the liquid, it is known and was taken as e = 0.1 to allow for
found that diffusion in the solid alloy can comparison with the previously mentioned
usually be neglected on time scales for the calculations on the free dendritic case. The
formation of cells. For long durations, of cellular wavelength was fixed to l = 0.36,
course, microsegregation takes place, and corresponding to the asterisks at increasing
solid diffusion then becomes important velocity and constant l in Fig. 2-24.
(Kurz and Fisher, 1998). At the lowest velocity still below the
A more serious difficulty in directional l = l dividing line, rounded cells were ob-
solidification is the large number of param- served; the tip was not well approximated
eters defining the system. We will concen- by a parabola. At higher speeds (V = 12) the
trate here on typical parameter values used parabolic structure of the tip became vis-
in experiments performed for some trans- ible, Fig. 2-25, and at even higher speeds
parent materials between glass plates. Sev- (V = 20) the dendritic structure with side
eral tests and specific calculations also branches was fully developed, Fig. 2-26.
done for alloys, however, indicate that a We can now compare the resulting tip ra-
large part of the results can be carried over dius with the predictions from free den-
to these more relevant situations from a dritic growth. Note that in the present case
metallurgical point of view without qual- the velocity is fixed rather than the super-
itative changes. cooling, so that the dendrite now uses a
supercooling corresponding to the given
velocity. This means that the tip of the den-
2.5.4 Directional Dendritic Growth
drite is no longer at a position in the tem-
The diagram in Fig. 2-24 showing veloc- perature-gradient field like a flat interface,
ity versus l in logarithmic form indicates but has advanced toward the warmer liq-
that qualitatively different behavior may be uid.
expected depending on whether the diffu- Fig. 2-27 contains the ratio of the tip ra-
sion length is larger or smaller than the cell dius divided by the radius from scaling, Eq.
spacing. In the previous section, we dis- (2-68) (where the Péclet number P was
cussed the first case. When the diffusion used in the original form as the ratio of tip
length becomes smaller than the cell spac- radius to diffusion length). Furthermore,
ing we expect that the individual cells be- this figure gives the ratio of the tip radius
come more and more independent of each to the Ivantsov radius, which comes from
132 2 Solidification
Figure 2-25. Transition from needle-shaped to dendritic cells at increasing pulling speeds. V = 4 is below the
dotted line in Fig. 2-24, and V = 12 above it. Parabolas adjusted to the tip radius are not a good fit to the profiles.
Figure 2-27. Ratios R tip /R x of tip radius computed numerically (Fig. 2-24) over two theoretical predictions,
where R x is either the Ivantsov radius (circles) or the radius from solvability theory (asterisks). See also Fig.
2-12. The result is in nearly perfect agreement with the solvability theory down to very low speeds in the cellular
region. The Ivantsov radius (for free growth) is not a good approximation there, as the diffusion fields of neigh-
boring cells strongly overlap. At high speeds, essentially free dendritic growth is confirmed (Saito et al., 1990).
Figure 2-28. Supercooling at the tip of a cell or dendrite vs. pulling speed as obtained from numerical simula-
tion. For a flat interface at low speeds, the global conservation law forces D = 1, then it decreases until l ≈ l , and
finally approaches the slowly increasing relation D (V ) obtained for the free dendritic case (see also Fig. 2-40).
The expected dependence on capillary anisotropy e is also recovered.
134 2 Solidification
A final point to be kept in mind is that Kurz and Fisher (1998) and references
the tip supercooling D in directional solid- therein).
ification is not small, as required by the ap- In a rigorous sense the answer is still
proximations used for the derivation of the negative, but at least arguments can be
scaling relation. On the other hand, P = 1 given for the existence of some boundaries
corresponds to a supercooling as large as on the wavelengths (or cell spacings)
D ≈ 0.75; 0.6 for 2-dim and 3-dim, respec- which can be estimated with the use of sim-
tively, and the scaling relations can be ex- plifications.
pected to hold over a large range of veloc- The situation here shows some similarity
ities, as already indicated from the other to the formation of hydrodynamic periodic
free dendritic case, Fig. 2-12. roll patterns (Newell and Whitehead, 1969;
In summary, these investigations have Kramer et al., 1982; Riecke and Paap,
shown that there appears to be a smooth 1986). In a laterally infinite system, a
transition from cellular to dendritic struc- whole band of parallel rolls is present
tures. The dendritic growth laws are very above the threshold for roll formation, the
well represented by the scaling relations so-called Eckhaus band. This was men-
for the free dendritic case. This scaling tioned in Sec. 2.5.3 for directional solidifi-
should hold in the region cation.
The reason for the stability of these rolls
d0 < R ≈ l < l < l T (2-124)
is that an infinitesimal perturbation is not
where l is the primary cell spacing. It was sufficient to create or annihilate a roll, but
proposed by Karma and Pelce (1989 a, b) a perturbation must exceed a threshold
that the transition from cells to dendrites value before such an adjustment can occur.
could occur via an oscillatory instability, In directional solidification, the situation
for which the present investigations under is different insofar as the envelope over the
quasi-stationary approximation have shown tips of the cells could make a smooth defor-
no evidence. A fully time-dependent calcu- mation of very long wavelength, thereby
lation is possible in principle with Green’s building up enough deformation energy so
function methods (Strain, 1989). that a cell could be created or annihilated at
isolated points. One indication for such a
process is the oscillatory instability of cells
2.5.5 The Selection Problem
postulated by Karma and Pelce (1989) and
of Primary Cell Spacing
Rappel and Brener (1992).
An important question from an engineer- The only hard argument for the selection
ing point of view appears to be the follow- of a unique wavelength comes from an
ing: suppose we know all the material pa- analysis of a spatially modulated thermal
rameters and the experimentally controlla- gradient acting on a cellular pattern of
ble parameters like thermal gradient and small amplitude (normal bifurcation)
pulling speed for a directional solidifica- which imposes a ramp on the pattern
tion process – can we then predict the dis- (Misbah, 1989; Misbah et al., 1990). The
tance between the cells and dendrites? idea originally proposed for the hydrody-
A positive answer to this question is de- namic case (Kramer et al., 1982) is to have
sirable because the mechanical properties a periodically varying thermal gradient
of the resulting alloy are improved with a parallel to the interface, which keeps the
decrease in the primary cell spacing (see interface flat in some regions and allows
136 2 Solidification
surface tension interacting between neigh- comes important there. Of course, a num-
boring tips. Taking this into account, the ber of rough approximations were used
destabilizing force is Fd ≈ W̃k(d) z˜k with specifically in the treatment of the destabi-
W̃k(d) = V |k | for a sinusoidal perturbation of lizing force, but this argument should at
amplitude z˜k of a plane interface without least qualitatively capture the competition
surface tension moving at velocity V and mechanism between neighboring den-
wavenumber k. The maximum lies at drites. A more detailed analysis (Warren
k = 2 p/l . The actual area under this pertur- and Langer, 1990) is quite promising at
bation z˜k contained in the solid cells is large velocities in comparison with experi-
smaller by a factor ≈ 2 R/l . We thus arrive ments (Somboonsuk et al., 1984) (see also
at a maximal destabilizing force of Kessler and Levine, 1986 c; Bechhoefer
and Libchaber, 1987).
Fd ≈ W (d) z l ; W (d) ≈ 4 p V R /l2 (2-125) Let us now look at the large-wavelength
corresponding to a depression or enhance- limit l . The initial growth conditions are
ment of every second dendrite. assumed to be just as before, Eq. (2-124),
On the other hand, each of these individ- but now at possibly large cell spacings l . In
ual dendrites knows its operating point, the numerical calculations, it was found
and through the given velocity its super- (Saito et al., 1989) that for fixed cell spac-
cooling at the tip. We approximate this ing l at increasing velocity, a tail instabil-
by the asymptotic form of Eq. (2-56) ity occurs (Fig. 2-31). A side branch in the
D ≈ 1 – 1/P since basically the variation
of D with P enters below, even through P
may not be very large compared to unity.
Capillary effects do not seem to be very
important in this region and are thus ne-
glected here for simplificity. By definition,
D = 1 – z tip /l T , and the two expressions for
D can be evaluated: z tip ≈ l T l /R. The stabi-
lizing force Fs ≈ W (s) z l follows from the
obvious relation W (s) = dVtip /dz tip as
VR
Fs ≈ W ( s ) z l ; W ( s ) ≈ − 2 (2-126)
lT l
As expected, the sign of this “force” is op-
posite to the destabilizing force, Eq. (2-
125). Setting the sum of the two equal to
zero, we expect an instability to occur first Figure 2-31. After a sudden increase in the growth
at cell spacings rate in the dendritic region at fixed cell spacing, tail
instability occurs. One of the side branches near the
tip produces a proturberance in the forward direction,
l ⱗ lT l (2-127) which then becomes a new primary cell (the imposed
for segregation coefficients around one. mirror symmetry is not present in reality of course).
In accordance with the stability of dendritic cells
We cannot say much about small segrega- against tip splitting (compare Fig. 2-29), this tail in-
tion coefficients because the nonlinearity stability appears to be an important selection mecha-
in Eq. (2-114) replacing Eq. (2-56) be- nism for primary cell spacing. See also Fig. 2-43.
138 2 Solidification
groove between two dendrites splits off a between the two critical values and again
ternary branch, which then moves so fast in recovers Eq. (2-1) by noting that l ~ V –1
the strong supersaturation in the groove and therefore V l 2 ≈ constant. The results,
that it finally becomes a new primary Eqs. (2-127) and (2-128), seem to be in
branch. In the plot, the imposed mirror agreement with experiments (Somboonsuk
symmetry is, of course, artificial, but it et al., 1984; Kurowsky, 1990; Kurz and
even acts opposite to this effect, making Fisher, 1981, 1998) concerning the scaling
the process more plausible in reality. In with respect to diffusion length l and ther-
fact, this is also observed experimentally mal length l T . The limit k Æ 0 for the seg-
(Esaka and Kurz, 1985), in particular when regation coefficient as a singular point is
the solid consists of slightly misaligned re- not reliably tractable here.
gions separated by grain boundaries, so The previously given relation l ~ l 1/4
that the growth direction of two neighbor- (Hunt, 1979; Trivedi, 1980; Kurz and
ing dendrites is slightly divergent. Our ba- Fisher, 1981) seems to be valid in an inter-
sic assumption now is that this tail instabil- mediate velocity region (Fig. 2-28) where
ity occurs when the intersection of para- D does not vary significantly, so that
bolic envelopes over neighboring dendrites z tip /lT ≈ 1/2 (see also Sec. 2.6.1).
occurs at a point z ⱗ0, where, theoretically, A serious point is the neglect of surface
D > 1. In this case, there is no need for tension and anisotropy in these derivations.
long-range diffusion around a side branch, In the experiments analyzed so far the rela-
for its dynamics become local. Of course, tion V l 2 ≈ const. seems to hold approxi-
this assumption ignores geometrical com- mately, but what happens for the capillary
petition between neighboring side branches anisotropy e going to zero? Numerically,
to a certain extent, but for the moment tip splitting occurs at lower velocities for
there seems to be no better argument to smaller e . In a system with anistropy e = 0
hand. (and zero kinetic coefficient) the structures
Taking into account the point that neigh- probably show chaotic dynamics at veloc-
boring parabolas with tip position at z tip > 0 ities where the diffusion length l is smaller
cannot intersect further down to the cold than the short wavelength limit of the neu-
side than at z = 0, we obtain z tip = l T l /R tral stability curve (Fig. 2-24), but this is
from the two relations for D, just as in the rather speculative (Kessler and Levine,
previous case of small wavelengths. But 1986 c).
now we must use the parabolic relation In considering whether the tail instabil-
z tip = l2/8 R for the intersection of two pa- ity (large l) or the competition mechanism
rabolas of radius R at z = 0, which are a dis- (small l) will dominate in casting pro-
tance l apart. The tail instability is then ex- cesses, we tend to favor the former. If the
pected to occur for solidification front consists of groups of
dendrites slightly misoriented against each
l ⲏ lT l (2-128)
other due to small-angle grain boundaries,
again with a prefactor roughly of the order cells will disappear at points where the lo-
of unity. In comparison with Eq. (2-127), it cal growth directions are converging and
can be seen that in both cases the same new cells will appear through the tail in-
scaling relation results. The scaling with stability at diverging points at the front.
the inverse growth rate l follows the neu- To summarize, the most likely scaling
tral stability curve at velocities safely in behavior of the primary cell spacing l , de-
2.5 Directional Solidification 139
pending on pulling velocities, follows Eqs. Mullins – Sekerka (MS) instabilities (also
(2-1) and (2-128) as a consequence of the called limit of constitutional undercool-
arguments presented in this section. This ing), to rates where plane front growth
conclusion is supported by a number of ex- again appears at velocities above Va , the
periments (Billia et al., 1987; Somboonsuk absolute stability limit.
et al., 1984; Kurowsky, 1990; Esaka and Morphological instabilities. The onset
Kurz, 1985), but more work remains to be of plane front interface instability is ob-
done. served to start at defects such as grain
boundaries, subgrain boundaries and dislo-
cations, forming a more or less pronounced
2.5.6 Experimental Results
depression at the intersection with the
on Directional Dendritic Growth
solid – liquid boundary, as shown in Fig.
Since 1950, in situ experiments on direc- 2-32 (Fisher and Kurz, 1978). It is inher-
tional solidification (DS) of transparent ently difficult to quantitatively observe the
model systems have been performed (Kof- break-down because the growth rates are
ler, 1950). However, it was some time be- small, and a long period of time is required
fore such experiments were specifically to reach steady state.
conceived to support microstructural mod- The amplitude and wavelength of the
els developed in the 1950s and early 1960s. perturbations as a function of V developing
The work of Jackson and Hunt (1966) is a in systems with small distribution coeffi-
milestone in this respect (see also Hunt et cients k (of the order of 0.1) are shown in
al., 1966). Their experimental approach to Fig. 2-33. In CBr4 – Br2 (de Cheveigne et
dendritic growth has been developed fur- al., 1986) and SCN – ACE (Eshelman and
ther by several groups: Esaka and Kurz Trivedi, 1987), the bifurcation is of a sub-
(1985), Trivedi (1984), Somboonsuk et al. critical type, i.e., there are two critical
(1984), Somboonsuk and Trivedi (1985), growth rates, one for increasing growth
Eshelman et al. (1988), Seetharaman and rate, Vc+ , and another lower value for de-
Trivedi (1988), Seetharaman et al. (1988), creasing growth rate, Vc– . Therefore, a peri-
de Cheveigne et al. (1986), Akamatsu et al. odic interface shape with infinitesimally
(1995), Akamatsu and Faivre (1998), and small amplitudes cannot form in these
others. Substantial progress has been made systems. As has been discussed in Sec.
during these years, especially due to results 2.5.3, only systems with k near unity will
obtained by Faivre and coworkers in very give supercritical (normal) bifurcation
thin (~ 15 µm) cells that constrain the pat- with a single, well-defined critical growth
terns to two dimensions. This research is rate Vc . The evolution of the wavelength
still producing interesting new insights into for two temperature gradients (70 and
the dynamics of interface propagating dur- 120 K/cm) is shown in Fig. 2-33 b. At the
ing crystallization. onset of instability the experimentally de-
The specific interest of DS is that growth termined wavelength is smaller than the
morphologies can be studied not only for critical wavelength by a factor of 2 – 3 com-
dendritic growth but also for cellular and pared with linear stability analysis. In-
plane front growth. We will discuss these creasing the rate above the threshold leads
phenomena in the sequence of their appear- to a decrease in the wavelength proportional
ance when the growth rate is increased to V – 0.4 (de Cheveigne et al., 1986; Kurow-
from Vc , the limit of first formation of sky, 1990). Once stability has started, the
140 2 Solidification
j J
(a) (b)
Figure 2-32. Morphological instabilities of a planar solid/liquid interface as seen on an inclined solidification
front between two glass plates (arrows). Photograph (a) was taken at an earlier stage than (b). The beginning of
the breakdown at defects such as dislocations, subgrain boundaries, or grain boundaries intersecting with the
solid/liquid interface is evident. The widths of the photographs correspond to 100 µm.
(a) (b)
Figure 2-33. Amplitude, A, and wavelength, l, of a periodic deformation of the solid/liquid interface of
CBr4 – Br2 solution versus pulling speed (de Cheveigne et al., 1986). Diagram (a) shows the hysteresis between
appearance and disappearance of perturbations (typical for an inverse bifurcation) for a temperature gradient of
12 K/mm. The lines in diagram (b) represent the calculated neutral stability curves for the two gradients indi-
cated. Open circles are experimental results for 7 K/mm, and crosses, for 12 K/mm. See also Fig. 2-21.
structure evolves to a steady-state cellular or Cells and dendrites. Losert et al. (1998 a)
dendritic growth mode. Which one of these developed an interesting technique in
structures will finally prevail is a question of which a spatially periodic UV laser pulse is
the growth conditions. directed onto the solidification front of the
2.5 Directional Solidification 141
Figure 2-34. Time evolution of the solid/liquid interface morphology when accelerating the growth rate from
0 to 3.4 µm/s at a temperature gradient of 6.7 K/mm. Magnification 41 ¥. (a) 50 s, (b) 55 s, (c) 65 s, (d) 80 s,
(e) 135 s, (f) 740 s (Trivedi and Somboonsuk, 1984).
142 2 Solidification
mental cell, which had approximately the et al., 1995). With the ·100Ò direction
same size as the diffusion distance, when oriented along the heat flow direction,
l = l. The corresponding growth rate also characteristic cellular/dendritic arrays such
marks the transition from two- to three-di- as shown in Fig. 2-36 are obtained. With
mensional growth of dendrites, as can be other orientations, where growth is ren-
seen in Fig. 2-36. Under conditions A to C, dered “effectively isotropic” a “seaweed”
no secondary arms are observed perpendic- structure is observed, the tips of which
ular to the plane of observation, while con- form what has been called on the basis of
ditions D to F show well-developed 3 D theoretical predictions a doublon (Ihle and
dendrites even far behind the tip. There- Müller-Krumbhaar, 1994).
fore, the gap might somewhat influence the The formation of doublons has also been
values of the transition rates but not the suggested to play an important role in the
qualitative behavior of the transition. As formation of “feathery” grains in technical
mentioned above, the theory (in quasi-sta- aluminum alloys. Recent detailed observa-
tionary approximation) only indicates a tions with electron back-scattered diffrac-
very gradual change in morphology. It is tion (EBSD) combined with optical and
interesting to compare Fig. 2-36 with Figs. scanning electron microscopy (Henry et
2-25 and 2-26. It can be seen that both the al., 1998) have clearly shown that feathery
theoretical and the experimental ap- grains are made of ·110Ò columnar den-
proaches show qualitatively the same be- drites, whose primary trunks are aligned
havior, even if the material constants used along and split in their center by a (111)
are not the same. coherent twin plane. The impingement of
The critical role of crystalline anisotropy secondary ·110Ò side arms gives rise to in-
in interface dynamics has been demon- coherent wavy twin boundaries. The switch
strated experimentally in directional growth from ·100Ò to ·110Ò growth direction was
of transparent CBr4 – C2Cl6 alloys grown as attributed to the small anisotropy of the
single crystals in thin samples (Akamatsu solid – liquid interfacial energy of alumi-
2.5 Directional Solidification 145
num, which can be changed by the addition The ratio l 2 /R obtained by Esaka and
of solute (Henry et al., 1998). Kurz (1985) for SCN – 1.3 wt.% ACE is in-
The scaling of the initial side-branch dependent of growth rate according to the
spacing l 2 with respect to the tip radius is precision of the measurements and takes a
shown in Fig. 2-38. Both quantities scale value of 2.1 ± 0.2 (see Fig. 2-39 and Table
closely with L 2 V = const. or L 2 C• = const. 2-1). This is in good agreement with the
(where L ~ R or l 2 ). Fig. 2-38 b also shows measurement of Somboonsuk and Trivedi
the quantity l p , the distance from the tip (1985) who found a value of 2.0 on the
down the shaft, where the first signs of tip same system over a wide range of growth
perturbations in the SCN – ACE system can rates, compositions, and temperature gra-
be observed. Here l p is of the same order dients (Trivedi and Somboonsuk, 1984).
as l 2 . This ratio increases with crystal anisotropy
(a)
A subject of appreciable practical impor- al. (1988). Other sources of convective in-
tance concerns the late stages of growth, stability (Corriell et al., 1980; Sahm and
where coarsening of the side branch struc- Keller, 1991) cannot be discussed here
tures occurs together with segregation in any detail, as the literature is too exten-
(Kurz and Fisher, 1998). If elastic forces sive in order to be covered within this
are not of primary importance, it is now chapter.
generally accepted (see Chapter 6, Sec.
6.4.1) that a relation L* = A + B t 1/3 (Huse,
Phase-field models
1986) is a good representation of typical
length scales L* varying with time t during In recent years, the phase-field metho-
diffusional coarsening processes. This con- dology has achieved considerable impor-
firms the classical Lifshitz – Slyozov – tance in modeling and numerically simulat-
Wagner theory (Lifshitz and Slyozov, ing a range of phase transitions and com-
1961; Wagner, 1961). The result, however, plex growth structures that occur during
is not specific about the geometric details, solidification. Phase-field models have
for example, in directional solidification, been applied to a wide range of materials
and it also assumes constant temperature. such as pure melts, simple binary alloys,
We have only briefly mentioned the tran- eutectic and peretectic phase transitions
sition from dendritic arrays back to a plane and grain growth structures in situations as
front during directional solidification at diverse as dendritic growth and rapid solid-
very high speeds. The importance of ki- ification.
netic coefficients was demonstrated by In the phase-field formulation a mathe-
Brener and Temkin (1989), and recently, matically sharp solid – liquid interface is
the possibility of chaotic dynamics in this smeared out or regularized and treated as
region similar to those described in the Ku- a boundary layer, with its own equation
ramoto – Sivashinsky equation was sug- of motion. The resulting formulation no
gested by Misbah et al. (1991). longer requires front tracking and the im-
A richness of dynamic phenomena was position of boundary conditions, but must
obtained in a stability analysis of eutectics be related to the sharp interface model by
(Datye et al., 1981). The possibility of an asymptotic analysis. In fact, there are
tilted lamellar arrays in eutectics was dem- many ways to prescribe a smoothing and
onstrated by Caroli et al. (1990) and, simi- dynamics of the sharp interface consistent
larly, for directional solidification at high with the original sharp-interface model. So
speeds by Levine and Rappel (1990). This there is no unique phase-field model, but
was observed in experiments on nematic rather a family of related models. The first
liquid crystals (Bechhoefer et al., 1989) phase-field model was developed by Lan-
and in eutectics (Faivre et al., 1989). ger (1986) in ad hoc manner. Subsequently,
Finally, the density difference between models have been placed on a more secure
liquid and solid should have some marked basis by deriving them within the frame-
influence on the growth mode (Caroli work of irreversible thermodynamics (Pen-
et al., 1984, 1989). For dendritic growth, rose and Fife, 1990; Wang et al., 1993).
forced flow was treated in some detail by Caginalp (1989) presented for the first time
Ben Amar et al. (1988), Ben Amar and Po- the relation between phase-field models
meau (1992), Bouissou and Pelce (1989), and sharp interface models or free boun-
Bouissou et al. (1989, 1990) and Rabaud et dary problems by considering the sharp
152 2 Solidification
interface limit in which the interface thick- Elder et al., 1994; Wheeler et al., 1996) and
ness is allowed to tend to zero. peritectic reactions (Lo et al., 2000).
However, the technique also has some
disadvantages. The first is related to the ef-
fective thickness of the diffuse interface of 2.6 Eutectic Growth
alloys (1– 5 nm) which is three to four or-
ders of magnitude smaller than the typical 2.6.1 Basic Concepts
length scale of the microstructure. Since
the interface must spread over several Eutectic growth is a mode of solidifica-
points of the mesh, this limits considerably tion for a two-component system. Operat-
the size of the simulation domain. The sec- ing near a specific point in the phase dia-
ond problem arises from the attachment ki- gram, it shows some unique features (Kurz
netics term, which plays a significant role and Fisher, 1998; Lesoult, 1980; Hunt and
in the phase-field equation, unlike micro- Jackson, 1966; Elliot, 1983).
structure formation of metallic alloys at The crucial point in eutectic growth is
low undercooling. These two factors have that the solidifying two-component liquid
so far limited phase-field simulations of at a concentration near CE (Fig. 2-19) can
alloy solidification with realistic solid- split into two different solid phases. The
state diffusivities to relatively large super- first phase consists of a high concentration
saturations. Recent mathematical and com- of A atoms and a low concentration of B at-
putational advances, however, are rapidly oms; the second solid phase has the oppo-
changing this picture. Some of the recent site concentrations. These two phases ap-
advances include: 1) a reformulated asymp- pear alternatively as lamellae or as fibers of
totic analysis of the phase-field model one phase in a matrix of the other phase.
for pure melts (Karma and Rappel, 1996, One condition for the appearance of a
1998) that has (i) lowered the range of ac- eutectic alloy is apparently a phase dia-
cessible undercooling by permitting more gram (as sketched in Fig. 2-19) with a tem-
efficient computations with a larger width perature TE . The two-phase regions meet at
of the diffuse interface region (as com- (TE , CE ), and the two liquidus lines inter-
pared to the capillary length), and (ii) made sect before they continue to exist as meta-
it possible to choose the model parameters stable liquidus lines (dash-dotted line) at
so as to make the interface kinetics vanish; temperatures somewhat below TE . This is a
2) a method to compensate for the grid an- material property of the alloy (Pb – Sn for
isotropy (Bösch et al., 1995); 3) an adaptive example). The other condition is the ex-
finite element method formulation that re- perimental starting condition for the con-
fines the zone near the diffuse interface and centration in the high-temperature liquid.
that has been used in conjunction with the Assume that we are moving a container
reformulated asymptotics to simulate 2D filled with liquid at concentration C• in a
dendritic growth at low undercooling (Pro- thermal gradient field where the solid is
vatas et al., 1998); 4) the implementation cold, the liquid is hot, and the solidification
of the method for fluid flow effects (Diep- front is proceeding towards the liquid in
ers et al., 1997; Tonhardt and Amberg, the positive z-direction. As a stationary so-
1998; Tong et al., 1998), and 5) the extension lution, we find a similar condition to the
of the technique to other solidification phe- simple directional solidification case; that
–
nomena including eutectic (Karma, 1994; is the concentration in the solid CS aver-
2.6 Eutectic Growth 153
aged across the front must be equal to C• to Diffusion of excess material not incorpo-
maintain global mass conservation together rated into one of the lamellae does not have
with a stationary concentration profile near to continue up to infinity but only to the
the front. neighboring lamella, which has the oppo-
Assume now that the liquid concentra- site composition relative to C• . Assuming
tion at infinity, C• , is close to CE and the again for simplicity a symmetric phase dia-
temperature at the interface is at TE . The gram, we may write this flux balance as
solid may split into two spatially alternat-
DC˜
ing phases now on the equilibrium (bino- V CE (1 − k ) = D (2-129)
dal) solidus lines C̃S (TI ), one located near (l / 2)
C Æ 0, and one located near C Æ1 at where the left-hand side describes the flux
TI < TE . To see what happens, let us look J of material to be carried away from each
at the situation with the concentration in lamellar interface during growth at veloc-
the liquid C• being precisely at eutectic ity V, which then must be equal to the
composition C• = CE . At TI < TE , there may diffusion current (concentration gradient)
now be alternating lamellae formed of over a distance l /2 to the neighboring la-
solid concentration C̃Sa and C̃Sb (Fig. 2-47), mella (across the liquid, as we ignore diffu-
the corresponding metastable liquidus con- sion in the solid). DC̃ = C̃L(a) – C̃L(b) is the
centrations being at C̃La and C̃Lb . For both concentration difference in the liquid at the
solid phases now the liquid concentration interfaces of the lamellae, and k < 1 is the
C• = CE is in the metastable two-phase segregation coefficient.
regions at TI : C̃Sa < C• < C̃La , C̃Lb < C• < C̃Sb . With the help of the liquidus slopes
In principle, we have to consider the pos- dT/dC̃L as material parameters, we can ex-
sibility of metastable solid phases (Tem- press the undercooling DTI = TE – TI as
kin, 1985), which we ignore here for sim-
plicity. dT
DTI = DC˜ 2 (2-130)
dC˜ L
dT l
DTI = 2 CE (1 − k ) (2-131)
dC˜ L l
solid – liquid interfaces relative to a flat imum undercooling would serve as the op-
interface. Accordingly, the growth front erating point of the system with spacing
will have local curvature. What is worse, given by Eq. (2-134).
there is no guarantee that the a – b solid – This stationary calculation was extended
solid interfaces are really parallel to the by Datye and Langer (1981) to a dynamic
growth direction. Making a symmetry as- stability analysis, where the solid – solid –
sumption for simplicity again, we can liquid triple points could move parallel and
ignore this problem for the moment. The perpendicular to the local direction of
curvature K of the growth front at each la- growth, coupled however to the normal di-
mella is then proportional to l–1. This cur- rection of the local orientation of the front.
vature requires, through the Gibbs – Thom- It was found that the marginal stability co-
son relation, Eq. (2-102), another reduc- incided exactly with the point of minimum
tion, DT K , of the interface temperature be- undercooling.
low T I : The basic model equations for eutectic
growth in a thermal gradient field can be
DT K = TE d K (2-132)
written in scaled form as follows (Brattkus
where d is an effective capillary length, de- et al., 1990; Caroli et al., 1990). Neglecting
pending on surface tensions, with K now diffusion in the solid (one-sided model)
taken as K ~ l–1. and assuming a single diffusion coefficient
The total reduction of temperature below D for solute diffusion near eutectic concen-
TE during eutectic growth can thus be writ- tration CE (Fig. 2-47), we assume a sym-
ten as (2-133) metric phase diagram about CE for simplic-
⎧ l d⎫ ity.
DT = DTI + DTK = TE ⎨aI + aK ⎬ We define a relative (local) concentra-
⎩ l l⎭
tion gap as
with dimensionless constants aI and aK .
Plotting this supercooling as a function DCa
d= (2-135)
of lamellar spacing, we find a minimal DC
supercooling ∂DT /∂l = 0 at
with DCa = CE – C̃Sa (T ), DCb = C̃Sb(T ) – CE
l = aK / aI ld (2-134) and DC = DCa + DCb .
The dimensionsless composition is then
Again, this is the fundamental scaling rela-
defined as
tion conjectured in Eq. (2-1) and encoun-
tered in various places in this chapter, C − CE
u= (2-136)
where the origin is a competition between DC ( TE )
driving force and stabilizing forces, most
intuitively expressed in Eq. (2-133). and (in contrast to the previous definitions)
A more thorough theoretical analysis of we express lengths and times in units of a
eutectic growth was given in the seminal diffusion length l˜ and time t̃
paper by Jackson and Hunt (1966), which l˜ = D/V , t˜ = D/V 2 (2-137)
is still a standard reference today. One ba-
sic approximation in this paper was to av- (Note that l˜ here differs by a factor of 2
erage the boundary conditions on flux and from previous definitions.) Restricting our
temperature over the interface. This led to attention to two dimensions corresponding
Eq. (2-133), and it was argued that the min- to lamellar structures, we now define the
2.6 Eutectic Growth 155
limited interval near the wavelength cor- system, revealing the similarities in the
responding to minimal supercooling (or structure of solution space between eutec-
maximal velocity in an isothermal process) tic and dendritic growth by showing the
and, accordingly, is described by Eq. (2- existence of a discrete set of solutions
134). for each parameter set (Kassner and Mis-
More recent extensions of the theory bah, 1991 b). Axisymmetric solutions were
(Coullet et al., 1989) gave indications that shown to cease to exist beyond a certain
the orientation of the lamellae (under iso- wavelength by merging in a fold singular-
tropic material parameters) are necessarily ity. For larger wavelengths, extended tilted
parallel to the growth direction of the front states appear. They were characterized in
but may be tilted and travel sideways at detail, both numerically (Kassner and Mis-
some specific angles (Caroli et al., 1990; bah, 1991 c) and analytically (Misbah and
Kassner and Misbah, 1991). Finally, it was Temkin, 1992; Valance et al., 1993). The
found (Kassner and Misbah, 1991 b) that supercritical nature of the tilt bifurcation
the standard model of eutectic solidifica- was demonstrated and a prediction was
tion has an intrinsic scaling structure made as to how extended tilted domains
could be produced rather than the localized
l ~ l d f (l / lT ) (2-146)
structures found so far. This prediction was
with a scaling function f depending only on borne out by subsequent experiments by
l /l T so that l ~ V –1/2 for 2 l ⱗ l T or for high the Faivre group (Faivre and Mergy,
enough velocities, while at lower velocities 1992 a). Amplitude equations were derived
the exponent should be smaller: l ~ V – 0.3. that explained how a supercritical bifurca-
This is in good agreement with the argu- tion could give rise to localized tilted do-
ments given for ordinary directional solid- mains and predicted a dynamical wave-
ification and also explains a lot of experi- length selection mechanism (Caroli et al.,
mental data (Lesoult, 1980). 1992; Kassner and Misbah, 1992 a) as well
Numerical simulations of the full contin- as describing the influence of crystalline
uum model of eutectic growth were subse- anisotropy (Kassner and Misbah, 1992 a).
quently carried out by Kassner and Misbah Again, this was verified in detail experi-
(1990, 1991 a, b, c). They gave a fairly mentally (Faivre and Mergy, 1992 b). A
complete description of the steady-state similarity equation was derived both
structures to be expected in the eutectic for axisymmetric (Kassner and Misbah,
2.6 Eutectic Growth 157
1991 b) and tilted states (Kassner and Mis- objection was countered (Baumann et al.,
bah, 1992 b), giving the corrections to the 1995 b) by pointing out that there is quite a
aforementioned scaling law. Anomalous number of steady states on the high-wave-
asymmetric cells bearing resemblance to length side of this point and that the ques-
asymmetric cells in experiments on dilute tion of irrelevance for pattern formation is
alloys (Jamgotchian et al., 1993) were a question of time-scales. Since these be-
found in the numerical solution of the eu- come slow near the accumulation point, the
tectic growth problem and described ana- relevance of a density of additional states
lytically (Kassner et al., 1993). cannot be easily discarded. It should be
To conclude the discussion of steady kept in mind that tilted lamellar states are
states, it was shown numerically that a also unstable with respect to phase diffu-
number of periodicity-increasing bifurca- sion near the tilt bifurcation (Fauve et al.,
tion exist, greatly enlarging the space of 1991). Nevertheless, they survive experi-
possible stationary solutions (Baumann et mental time-scales, because phase diffu-
al., 1995 a). These results strongly suggest sion is very slow.
that the question of wavelength selection The first dynamic simulation of the con-
should be considered from a different point tinuum model of eutectic growth (in the
of view. It is known that a band of steady- quasi-stationary approximation) appears to
state solutions exists in directional solidifi- have been done by Kassner et al. (1995),
cation beyond the instability threshold and but the only extensive study so far is due to
a similar statement holds for eutectic (even Karma and Sarkissian (1996). This beauti-
though there is no threshold in this case). ful work was done in close collaboration
Moreover, careful experiments (Faivre, with experimentalists and thus validated
1996) have clarified that the observed quantitively. It pinpointed a number of new
wavelength depends on the history of the short-wavelength oscillatory instabilities
sample. Therefore, there is no wavelength and made quantitative predictions for the
selection in the strict mathematical sense. CBr4 – C2Cl6 system allowing these in-
This is at least true in the absence of noise, stabilities to be reproduced in experiments
i.e. for the deterministic equations of mo- (Faivre, 1996; Ginibre et al., 1997). There-
tion. However, the investigations of Bau- fore, a pretty complete picture exists today
mann et al. (1995 a) render it plausible that of the possible stationary and oscillatory
a whole attractor of steady-state solutions patterns in lamellar eutectic growth and
with accumulation points exists. One of their stability range.
these accumulation points is the point of A recent development is the stability
marginal stability. Since the temporal evo- analysis of ternary eutectics by Plapp and
lution of any structure in the vicinity of this Karma (1999), based on an extension of the
attractor will become slow, it seems likely Datye – Langer theory. It shows how the
that observed wavelengths will correspond presence of a third component in the sys-
to points of increased density of this attrac- tem leads to the formation of eutectic colo-
tor. There have been objections (Karma nies; an effective description of the eutectic
and Rappel, 1994) that many of the newly front on length scales much larger than the
found steady-state solutions are on the low- lamellar spacing is provided, which yields
wavelength side of the point of marginal a simple means of calculating an approxi-
stability, hence they are unstable and there- mate stability spectrum.
fore irrelevant for pattern formation. This
158 2 Solidification
(a)
(b)
Figure 2-50. Lamellar spacings of (a) directionally solidified eutectics and (b) directionally transformed eutec-
toids as a function of growth rate (Kurz and Sahm, 1975).
160 2 Solidification
presented here. Fig. 2-51 indicates that in manent creation and movement of faults
eutectics the spacings are also not at all (see below). This process is three-dimen-
uniquely defined. There is a rather wide sional and cannot be realistically simulated
distribution around a mean value for each in two-dimensional calculations.
rate (Trivedi et al., 1991). The operating In Fig. 2-52, the mean spacings are plot-
range of eutectics is determined by the per- ted as points, and the limits of the distribu-
Figure 2-51. Eutectic spacing distribution curves as a function of velocity for Pb – Pd eutectics (Trivedi et al.,
1991).
(a)
(b)
Figure 2-54. Eutectic fault spacing distribution Figure 2-55. Periodic oscillations in hypoeutectic
curves for Al – CuAl2 directionally solidified with Al – CuAl2 eutectic under rapid laser resolidification
different growth rates and spacing values as indi- conditions: (a) experimental observation and (b) sim-
cated (Riquet and Durand, 1975). ulation (Zimmermann et al., 1990; Karma, 1987).
2.6 Eutectic Growth 163
three-phase junctions at the solid – liquid Barber, M. N., Barbieri, A., Langer, J. S. (1987),
Phys. Rev. A 36, 3340.
interface enter as additional conditions and Barbieri, A., Langer, J. S. (1989), Phys. Rev. A 39,
further details of the phase diagram be- 5314.
come important. The dynamics of the sys- Barbieri, A., Hong, D. C., Langer, J. S. (1987), Phys.
Rev. A 35, 1802.
tem seem to show a richer structure than Barth, M., Wei, B., Herlach, D. M., Feuerbacher, B.
ordinary directional solidification. The se- (1994), Mater. Sci. Eng. A 178, 305.
lection of spacings between the different Barth, M., Wei, B., Herlach, D. M. (1995), Phys. Rev.
B 51, 3422.
solid phases in materials of practical im- Barth, M., Wei, B., Herlach, D. M. (1997), Mater.
portance occurs through three-dimensional Sci. Eng. A 226 – 228, 770.
defect formation. In addition, nucleation Baumann, R., Kassner, K., Misbah, C., Temkin, D.
(1995 a), Phys. Rev. Lett. 74, 1597.
and faceting of the interfaces should be Baumann, R., Kassner, K., Misbah, C., Temkin, D.
considered. (1995 b), Phys. Rev. Lett. 75, 2445.
A number of problems common to all Bechhoefer, J., Libchaber, A. (1987), Phys. Rev. Lett.
35, 1393.
of these growth modes have only been Bechhoefer, J., Simon, A. J., Libchaber, A., Oswald,
touched upon so far. These problems in- P. (1989), Phys. Rev. A 40, 3974.
clude, for example, the redistribution of Beckermann, C., Wang, C. Y. (1996), Metall. Mater.
Trans. 27 A, 2784.
material far behind the tip regions, the Ben Amar, M. (1990), Phys. Rev. A 41, 2080.
treatment of elastic effects, and the interac- Ben Amar, M. (1998), Physica D 31, 409.
tion with hydrodynamic instabilities due to Ben Amar, M., Brener, E. (1993), Phys. Rev. Lett. 71,
589.
thermal and compositional gradients. Ben Amar, M., Brener, E. (1995), Phys. Rev. Lett. 75,
In summary, we expect the field to re- 561.
main very active in the future, as it is at- Ben Amar, M., Moussallam, B. (1987), Physica 25 D,
tractive from a technological point of view. 155.
Ben Amar, M., Pelce, P. (1989), Phys. Rev. A 39, 4263.
It will certainly provide some surprises and Ben Amar, M., Pomeau, Y. (1968), Europhys. Lett. 2,
new insights into the general concepts of 307.
pattern formation in dissipative systems. Ben Amar, M., Pomeau, Y. (1992), in: Solids Far
from Equilibrium: Godreche, C. (Ed.). Cambridge:
Cambridge University Press.
Ben Amar, M., Bouissou, Ph., Pelce, P. (1988), J.
Crystal Growth 92, 97.
2.8 References Ben Jacob, E., Garik, P. (1990), Nature 343, 523.
Ben Jacob, E., Goldenfeld, N., Langer, J. S., Schön,
Abel, T., Brener, E., Müller-Krumbhaar, H. (1997), G. (1983), Phys. Rev. Lett. 51, 1930.
Phys. Rev. E 55, 7789. Ben Jacob, E., Goldenfeld, N., Langer, J. S., Schön,
Adda Bedia, M., Ben Amar, M. (1991), Phys. Rev. A G. (1984), Phys. Rev. A 29, 330.
43, 5702. Bensimon, D., Pelce, P., Shraiman, B. I. (1987), J.
Akamatsu, S., Faivre, G. (1998), Phys. Rev. E 58, Phys. France 48, 2081.
3302. Bilgram, J. H., Firmann, M., Hürlimann, E. (1989), J.
Akamatsu, S., Faivre, G., Ihle, T. (1995), Phys. Rev. Crystal Growth 96, 175.
E 51, 4751. Billia, B., Jamgotchian, H., Capella, L. (1987), J.
Almgren, R., Dai, W.-S., Hakim, V. (1993), Phys. Crystal Growth 82, 747.
Rev. Lett. 71, 3461. Billia, B., Jamgotchian, H., Capella, L. (1989), J.
Arnold, C. B., Aziz, M. J., Schwarz, M., Herlach, Crystal Growth 94, 987.
D. M. (1999), Phys. Rev. B 59, 334. Binder, K. (2001), in: Phase Transformations in Ma-
Assadi, H., Barth, M., Greer, A. L., Herlach, D. M. terials: Kostorz, G. (Ed.), Weinheim: Wiley-VCH,
(1996), Mat. Sci. Forum 215 – 216, 37. Chap. 4.
Assadi, H., Barth, M., Greer, A. L., Herlach, D. M. Binder, K., Fratzl, P. (2001), in: Phase Transforma-
(1998), Acta Mater. 46, 491. tions in Materials: Kostorz, G. (Ed.), Weinheim:
Baker, J. C., Cahn, J. (1971), in: Solidification: Hu- Wiley-VCH, Chap. 6.
ghel, T. A., Bolling, G. F. (Eds.). Metals Park, OH, Bisang, U., Bilgram, J. H. (1995), Phys. Rev. Lett. 75,
USA: American Society of Metals. 3898.
166 2 Solidification
Boettinger, W. J., Coriell, S. R., Trivedi, R. (1988), Caroli, B., Caroli, C., Roulet, B., Faivre, G. (1989),
in: Rapid Solidification Processing: Principles J. Crystal Growth 94, 253.
and Technologies IV: Mehrabian, R., Parrish, P. A. Caroli, B., Caroli, C., Roulet, B., Brattkus, K. (1990),
(Eds.), Baton Rouge, LA: Claitor’s, p. 13. J. Phys. France 51, 1865.
Boettinger, W. J., Coriell, S. R., Greer, A. L., Karma, Caroli, B., Caroli, C., Fauve, S. (1992), J. Phys. I
A., Kurz, W., Rappaz, M., Trivedi, R. (2000), Acta (Paris) 2, 281.
Mater. 48, 43. Carpay, F. M. A. (1972), in: Handbuch der Mikrosko-
Bösch, A., Shochet, O., Müller-Krumbhaar, H. pie in der Technik, Vol. VI: Freund, H. (Ed.).
(1995), Z. Physik B97, 367. Frankfurt: Umschau, p. 355.
Bouissou, Ph., Pelce, P. (1989), Phys. Rev. A 40, Carrard, M., Gremaud, M., Zimmermann, M., Kurz,
6673. W. (1992), Acta Metall. Mater. 40, 983.
Bouissou, Ph., Perrin, B., Tabeling, P. (1989), Phys. Chan, S.-K., Reimer, H.-H., Kahlweit, M. (1978), J.
Rev. A 40, 509. Crystal Growth 43, 229.
Bouissou, Ph., Chiffaudel, A., Perrin, B., Tabeling, P. Chandrasekhar, C. (1961), Hydrodynamics and Hy-
(1990), Europhys. Lett. 13, 89. drodynamic Stability. Oxford. Clarendon.
Bowley, R., Caroli, B., Caroli, C., Graner, P., No- de Cheveigne, S., Guthman, C., Lebrun, M. M.
zieres, P., Roulet, B. (1989), J. Phys. France 50, (1986), J. Physique 47, 2095.
1377. Chopra, M. A., Glicksman, M. E. Singh, N. B.
Brattkus, K., Misbah, C. (1990), Phys. Rev. Lett. 64, (1988), Metall. Trans. 19 A, 3087.
1935. Classen, A., Misbah, C., Müller-Krumbhaar, H.
Brattkus, K., Caroli, B., Caroli, C., Roulet, B. (1990), (1991), Phys. Rev. A 43, 6920.
J. Phys. France 51, 1847. Cochrane, R. F., Greer, A. L., Eckler, K., Herlach,
Brazovsky, S. (1975), Sov. Phys. JETP 41, 85. D. M. (1991), Mater. Sci. Eng. A 133, 698
Brener, E. (1993), Phys. Rev. Lett. 71, 3653. Collins, J. B., Levine, H. (1985), Phys. Rev. B 31,
Brener, E. A., Levine, H. (1991), Phys. Rev. A 43, 6119.
883. Corriell, S., Sekerka, R. F. (1983), J. Crystal Growth
Brener, E. A., Melnikov, V. I. (1990), J. Phys. France 61, 499.
51, 157. Corriell, S., Cordes, M., Boettinger, M., Sekerka, R.
Brener, E. A., Melnikov, V. I. (1991), Adv. Phys. 40, (1980), J. Crystal Growth 49, 13.
53. Coullet, P., Goldstein, R. E., Gunaratne, G. H.
Brener, E. A., Temkin, D. E. (1989), Europhys. Lett. (1989), Phys. Rev. Lett. 63, 1954.
10, 171. Datye, V., Langer, J. S. (1981), Phys. Rev. B 24,
Brener, E. A., Temkin, D. (1995), Phys. Rev. E 51, 4155.
351. Datye, V., Mathur, R., Langer, J. S. (1981), J. Stat.
Brener, E. A., Geilikman, M. B., Temkin, D. E. Phys. 29, 1.
(1987), Sov. Phys. JETP 65, 93. Deissler, R. J. (1987), Physica 22 D, 233.
Brener, E. A., Geilikman, M. B., Temkin, D. E. Diepers, H.-J., Beckermann, C., Steinbach, I. (1997),
(1988), Sov. Phys. JETP 67, 1002. in: Solidification Processing 1997: Beech, J.,
Brener, E. A, Müller-Krumbhaar, H., Temkin, D. Jones H. (Eds.), Sheffield, UK: University, p. 426.
(1996), Phys. Rev. E 54, 2714. Dombre, T. Hakim, V. (1987), Phys. Rev. A 36, 2811.
Burden, M. H., Hunt, J. D. (1974), J. Crystal Growth Double, D. D. (1973), Mater. Sci. Eng. 11, 325.
22, 99. Dougherty, A. (1991), J. Crystal Growth 110, 501.
Burkhardt, T., Kroll, D., Müller-Krumbhaar, H. Dougherty, A., Gollub, J. P. (1988), Phys. Rev. A 38,
(1977), J. Crystal Growth 38, 13. 3043.
Caginalp, G., (1989), Phys. Rev. A 39, 5887. Dougherty, A., Kaplan, P. D., Gollub, J. P. (1987),
Callen, H. B. (1960), Thermodynamics. New York: Phys. Rev. Lett. 58, 1652.
Wiley. Eckhaus, W. (1965), Studies in Nonlinear Stability
Caroli, B., Caroli, C., Roulet, B. (1982), J. Phys. Theory; Springer Tracts in Natural Philosophy 6.
France 43, 1767. Berlin: Springer.
Caroli, B., Caroli, C., Roulet, B. (1984), J. Crystal Eckler, K., Cochrane, R. F., Herlach, D. M., Feuer-
Growth 66, 575. bacher, B. (1991 a), Mater. Sci. Eng. A 133, 702.
Caroli, B., Caroli, C., Roulet, B., Langer, J. S. Eckler, K., Herlach, D. M., Hamerton, R. G., Greer,
(1986 a), Phys. Rev. A 33, 442. A. L. (1991 b), Mater. Sci. Eng. A 133, 730.
Caroli, B., Caroli, C., Misbah, C., Roulet, B. Eckler, K., Cochrane, R. F., Herlach, D. M., Feuer-
(1986 b), J. Phys. France 47, 1623. bacher, B., Jurisch, M. (1992), Phys. Rev. B 45,
Caroli, B., Caroli, C., Roulet, B. (1987), J. Phys. 5019.
France 48, 1423. Eckler, K., Herlach, D. M. (1994), Mater. Sci. Eng.
Caroli, B., Caroli, C., Roulet, B. (1988), in: Solid A 178, 159.
State Phenomena: Kubin, L. P., Martin, G. Eds.). Eckler, K., Herlach, D. M., Aziz, M. J. (1994), Acta
Aedermannsdorf, CH: Trans. Tech. Publ. Metall. Mater. 42, 975.
2.8 References 167
Elliot, R. (1983), Eutectic Solidification Processing Hogan, L. M., Kraft, R. W., Lemkey, F. D. (1971),
– Crystalline and Glassy Alloys. London: Butter- Adv. Mater. Res. 5, 83.
worths. Honjo, H., Ohta, S., Sawada, Y. (1985), Phys. Rev.
Esaka, H. (1986), Sc.D. thesis, EPFL, Lausanne, CH. Lett. 55, 841.
Esaka, H., Kurz, W. (1985), J. Crystal Growth 72, Huang, S. C., Glicksman, M. E. (1981), Acta Metall.
578. 29, 701; 717.
Esaka, H. Kurz, W., Trivedi, R. (1988), in: Solidifica- Hunt, J. D. (1979), in: Solidification and Casting of
tion Processing 1987. London: The Institute of Metals. London: TMS, p. 1.
Metals, p. 198. Hunt, J. D. (1987), private communication.
Eshelman, M. A., Trivedi, R. (1987), Acta Metall. 35 Hunt, J. D., Jackson, K. A. (1966), Trans. Metall.
2443. Soc. AIME 236, 843.
Eshelman, M., Seetharaman, V., Trivedi, R. (1988), Hunt, J. D., Jackson, K. A., Brown, H. (1966), Rev.
Acta Metall. 36, 1165. Sci. Instr. 37, 805.
Faivre, G. (1996), J. Crystal Growth 166, 29. Hunt, J. D., Lu, S.-Z. (1996), Metall. Mater. Trans.
Faivre, G., Mergy, J. (1992a), Phys. Rev. A 45, 7320. 27A, 611.
Faivre, G., Mergy, J. (1992b), Phys. Rev. A 46, 963. Hunziker, O., Vandyoussefi, M., Kurz, W. (1998),
Faivre, G., de Cheveigne, S., Guthmann, C., Kurow- Acta Mater. 46, 6325.
sky, P. (1989), Europhys. Lett. 9, 779. Huse, D. (1986), Phys. Rev. B 34, 7845.
Fauve, S., Douady, S., Thual, O. (1991), J. Phys. II 1, Ihle, T., Müller-Krumbhaar, H. (1994), Phys. Rev.
311. E 49, 2972.
Fisher, D. J., Kurz, W. (1978), unpublished. Ivantsov, G. P. (1947), Dokl. Akad. Nauk SSSR 58,
Fisher, D. J., Kurz, W. (1980), Acta Metall. 28, 777. 567.
Flemings, M. C. (1991), in: Materials Science and Jackson, K. A. (1958), in: Liquid Metals and Solidifi-
Technology Vol. 15, Processing of Metals and Al- cation. Cleveland: ASM, p. 174.
loys: Cahn, R. W. (Ed.). Weinheim: VCH Verlags- Jackson, K. A., Hunt, J. D. (1966), Trans. Metall.
gesellschaft, p. 1. Soc. AIME 236, 1129.
Fujioka, T. (1978), Ph.D. Thesis, Carnegie-Mellon Jamgotchian, H., Trivedi, R., Billia, B. (1993), Phys.
Univ. Pittsburgh (unpublished). Rev. E 47, 4313.
Geilikman, M. B., Temkin, D. (1984), J. Crystal Jones, H., Kurz, W. (1981), Z. Metallkde 72, 792.
Growth 67, 607. Kahlweit, M. (1970), J. Crystal Growth 6, 125.
Gill, S., Kurz, W. (1993), Acta Metall. Mater. 41, Karma, A. (1986), Phys. Rev. Lett. 57, 858.
3563. Karma, A. (1987), Phys. Rev. Lett. 59, 71.
Gill, S., Kurz, W. (1995), Acta Metall. Mater. 43, Karma, A. (1994), Phys. Rev. E 49, 2245.
139. Karma, A. (1998), Intern. J. Non-Equilib. Process.
Ginibre, M., Akamatsu, S., Faivre, G. (1997), Phys. 11, 201.
Rev. E 56, 780. Karma, A., Langer, J. S. (1984), Phys. Rev. A 30,
Glickman, M. E., Singh, N. B. (1986), in: Rapidly 3147.
Solidified Powder Alloys: Fine, M. E. Stark, E. A. Karma, A., Kotliar, B. G. (1985), Phys. Rev. A 31,
(Eds.). Philadelphia: ASTM, p. 44. 3266.
Glickman, M. E. Singh, N. B. (1989), J. Crystal Karma, A., Pelce, P. (1989 a), Phys. Rev. A 39, 4162.
Growth 98, 277. Karma, A., Pelce, P. (1989 b), Europhys. Lett. 9, 713.
Glicksman, M., Shaefer, R. J., Ayers. J. D. (1976), Karma, A., Rappel, W.-J. (1994), Phys. Rev. Lett. 75,
Metall, Trans. A 7, 1747. 2444.
Glickman, M. E., Singh, N. B., Chopra, M. (1988), Karma, A., Rappel, W.-J. (1996), Phys. Rev. E 53,
J. Crystal Growth 92, 543. 3017.
Goetzinger, R., Barth, M., Herlach, D. M. (1998), Karma, A., Rappel, W.-J. (1998), Phys. Rev. E 57,
Acta Mater. 46, 1647. 4323.
Gremaud, M. Carrard, M., Kurz, W. (1990), Acta Karma, A., Sarkissian, A. (1993), Phys. Rev. E 47,
Metall. Mater. 39, 1431. 513A.
Ha, H. P., Hunt, J. D. (1997), in: Proc. 4th Dec. Int. Karma, A., Sarkissian, A. (1996), Metall. Trans. 27,
Conf. on Solidification Proc. (SP’97) Beech, J., 635.
Jones, H. (Eds.), Sheffield, UK: University, p. 444. Karma, A., Rappel, W. J., Fuh, B. C., Trivedi, R.
Henry, S., Minghetti, T., Rappaz, M. (1998), Acta (1998), Metall. Trans. 29 A, 1457.
Mater. 46, 6431. Kashuba, A., Pokrovsky, V. (1990), Z. Physik 78, 289.
Herlach, D. M. (1994), Mater. Sci. Eng. A 179 – 180, Kassner, K., Misbah, C. (1990), Phys. Rev. Lett. 65,
147. 1458.
Herlach, D. M. (1997), Mater. Sci. Eng. A 226 – 228, Kassner, K., Misbah, C. (1991 a), Phys. Rev. Lett. 66,
348. 445.
Herlach, D. M., Cochrane, R. F., Egry, I., Fecht, H. J., Kassner, K., Misbah, C. (1991 b), Phys. Rev. Lett.
Greer, A. L. (1993), Intern. Mater. Rev. 38, 273. A 44, 6513.
168 2 Solidification
Kassner, K., Misbah, C. (1991 c), Phys. Rev. A 44, Landau, L. D., Lifshitz, E. M. (1970), Statistische
6533. Physik, Berlin: Akademie-Verlag.
Kassner, K., Misbah, C. (1992 a), Phys. Rev. A 45, Langer, J. S. (1971), Ann. Phys. 65, 53.
7372. Langer, J. S. (1980 a), Rev. Mod. Phys. 52, 1.
Kassner, K., Misbah, C. (1992 b), J. Phys. Rev. A 25, Langer, J. S. (1980 b), Phys. Rev. Lett. 44, 1023.
3213. Langer, J. S. (1980 c), Physico-Chem. Hydrodyn. 1,
Kassner, K., Valance, A., Misbah, C., Temkin, D. 41.
(1993), Phys. Rev. E 48, 1091. Langer, J. S. (1986, in: Directions in Condensed Mat-
Kassner, K., Misbah, C., Baumann, R. (1995), Phys. ter. Grinstein, G., Mazenko, G. (Eds.). Singapore:
Rev. E 51, 2751. World Scientific, p. 164.
Kerr, H. W., Kurz, W. (1996), Int. Mater. Rev. 41, Langer, J. S. (1987 a), Phys. Rev. Lett. A 36, 3350.
129. Langer, J. S. (1987 b), in: Le hasard et la matiere/
Kessler, D., Levine, H. (1986 a), Phys. Rev. Lett. 57, Chance and matter: Souletie, J., Vannimenus, J.,
3069. Stora, R. (Eds.). Amsterdam: Elsevier Science
Kessler, D., Levine, H. (1986 b), Phys. Rev. A 33, Publ.
3352. Langer, J. S., Müller-Krumbhaar, H. (1977), J. Crys-
Kessler, D., Levine, H. (1986 c), Phys. Rev. A 33, tal Growth 42, 11.
3325. Langer, J. S., Müller-Krumbhaar, H. (1978), Acta
Kessler, D., Levine, H. (1986 d), Phys. Rev. A 36, Metall. 26, 1681; 1689; 1697.
4123. Langer, J. S., Müller-Krumbhaar, H. (1980), Acta
Kessler, D., Levine, H. (1988), Acta Metall. 36, Metall. 29, 145.
2693. Langer, J. S., Müller-Krumbhaar, H. (1982), Phys.
Kessler, D., Levine, H. (1989), Phys. Rev. A 39, 3041. Rev. A 27, 499.
Kessler, D. Koplik, J., Levine, H. (1987), in: Pattern, Langer, J. S., Turski, L. A. (1977), Acta Metall. 25,
Defects and Microstructures in Nonequilibrium 1113.
Systems: Walgraef, D. (Ed.). Dordrecht: Martinus Lesoult, G. (1980), Ann. Chim. Fr. 5, 154.
Nijhoff. Lesoult, G. (1990), private communication.
Kessler, D., Koplik, J., Levine, H. (1988), Adv. Phys. Levine, H., Rappel, W. (1990), unpublished.
37, 255. Li, D. Eckler, K., Herlach, D. M. (1995b), Europhys.
Kofler, A. (1950), Z. Metallkde 41, 221. Lett. 32, 223.
Koss, M. B., La Combe, J. C., Tennenhouse, L. A., Li, D. Eckler, K., Herlach, D. M. (1996), J. Crystal
Glicksman, M. E., Winsa, E. A. (1999), Metall. Growth 160, 59.
Mater. Trans. 30 A, 3177. Li, D. Herlach, D. M. (1996), Phys. Rev. Lett. 77,
Kramer, L., Ben Jacob, E., Brand, H., Cross, M. 1801.
(1982), Phys. Rev. Lett. 49, 1891. Li, D., Volkmann, T., Eckler, K., Herlach, D. M.
Krug, J., Spohn, H. (1992), in: Solids Far from (1995 a), J. Crystal Growth 152, 101
Equilibrium: Growth, Morphology and Defects: Li, Q., Beckermann, C. (1998), Phys. Rev. E 57, 1.
Godreche, C. (Ed.). Cambridge: Cambridge Uni- Lifshitz, I. M., Slyozov, V. V. (1961), J. Phys. Chem.
versity Press, p. 479. Solids 19, 35.
Kruskal, M., Segur, H. (1985), Asymptotics Beyond Lipton, J., Glicksman, M. E., Kurz, W. (1987), Met-
All Orders in a Model of Dendritic Crystals, all. Trans. 18 A, 341.
A.R.A.P. Tech. Memo. Lipton, J., Kurz, W., Trivedi, R. (1987 a), Acta Met-
Kurowsky, P. (1990), Thesis, GPS-Univ. Paris VII. all. 35, 957.
Kurz, W., Fisher, D. J. (1981), Acta Metall. 29, 11. Livingston, J. D., Cline, H. E., Koch, E. F., Russel, R.
Kurz, W., Fisher, D. J. (1998), Fundamentals of So- R. (1970), Acta Metall. 18, 399.
lidification. Aedermannsdorf, CH: Trans. Tech. Lo, T. S., Karma, A., Plapp, M. (2001), Phys. Rev.
Publ. E63, 1504.
Kurz, W., Trivedi, R. (1990), Acta Metall. Mater. 38, Losert, W., Mesquita, O. N., Figueiredo, J. M. A.,
1. Cummins, H. Z. (1998b), Phys. Rev. Lett. 81, 409.
Kurz, W., Trivedi, R. (1991), Metall. Trans. 22 A, Losert, W., Stillman, D. A., Cummins, H. Z., Kop-
3051. czynski, P., Rappel, W.-J. Karma, A. (1998 a),
Kurz, W., Trivedi, R. (1996), Metall. Mater. Trans. Phys. Rev. E 58, 7492.
27A, 625. Ludwig, A., Kurz, W. (1996a), Scripta Mater. 35,
Kurz, W., Sahm, P. R. (1975), Gerichtet erstarrte eu- 1217
tektische Werkstoffe. Berlin: Springer Verlag. Ludwig, A., Kurz, W. (1996b), Acta Mater. 44, 3643
Kurz, W., Giovanola, B., Trivedi, R. (1986), Acta Mashaal, M., Ben Amar, M., Hakim, V. (1990), Phys.
Metall. 34, 823. Rev. A 41, 4421.
Kurz, W., Giovanola, B., Trivedi, R. (1988), J. Crys- Maurer, J., Bouissou, P., Perrin, B., Tabeling, P.
tal Growth 91, 123. (1988), Europhys. Lett. 8, 67.
2.8 References 169
Mazumder, P., Trivedi, R., Karma, A. (2000), Metall. Rappel, W.-J., Brener, E. A. (1992), J. Phys. I France
Mater. Trans. A31, 1233. 2, 1779.
McFadden, G. B., Coriell, S. R. (1984), Physica Raz, E., Lipson, S. G., Polturak, E. (1989), Phys. Rev.
12 D, 253. A 40, 1088.
Meakin, P. (1988), in: Phase Transitions and Critical Riecke, H., Paap, H. G. (1986), Phys. Rev. A 33, 547.
Phenomena, Vol. 12: Domb, C., Lebowitz, J. Riquet, J. P., Durant, F. (1975), Mat. Res. Bull. 10,
(Eds.). London: Academic, p. 336. 399.
Meiron, D. I. (1986), Phys. Rev. A 33, 2704. Saffmann, P. G., Taylor, G. I. (1958), Proc. R. Soc.
Misbah, C. (1987), J. Phys. France 48, 1265. London Ser. A 254, 312.
Misbah, C. (1989), J. Phys. France 50, 971. Sahm, P. R., Keller, M. H. (1991), in: Materials
Misbah, C., Temkin, D. (1992), Phys. Rev. A 46, Science and Technology, Vol. 15, Processing of
4497. Metals and Alloys. Weinheim: VCH Verlagsge-
Misbah, C., Müller-Krumbhaar, H., Saito, Y. (1990), sellschaft, p. 539.
J. Crystal Growth 99, 156. Saito, Y., Goldbeck-Wood, G. Müller-Krumbhaar, H.
Misbah, C., Müller-Krumbhaar, H., Temkin, D. E. (1987), Phys. Rev. Lett. 58, 1541.
(1991), J. Physique I 1, 585. Saito, Y., Goldbeck-Wood, G. Müller-Krumbhaar, H.
Mordike, B. L. (1991), in: Materials Science and (1988), Phys. Rev. Lett. A 38, 2148.
Technology Vol. 15, Processing of Metals and Al- Saito, Y., Misbah, C., Müller-Krumbhaar, H. (1989),
loys: Cahn, R. W. (Ed.). Weinheim: VCH Verlags- Phys. Rev. Lett. 63, 2377.
gesellschaft, p. 111. Schumacher, P., Greer, A. L., Worth, J., Evans, P. V.,
Mullins, W. Sekerka, R. (1963), J. Appl. Phys. 34, Kearns, M. A., Fisher, P., Green, A. H. (1998),
323. Mater. Sci. Technol. 14, 394.
Mullins, W. Sekerka, R. (1964), J. Appl. Phys. 35, Schwarz, M., Arnold, C. B., Aziz, M. J., Herlach,
444. D. M. (1997), Mater. Sci. Eng. A 226–228, 420.
Nakaya, U. (1954), Snow Crystals. Cambridge: Har- Schwarz, M., Karma, A., Eckler, K., Herlach, D. M.
vard Univ. Press. (1990), Phys. Rev. Lett. 56, 324.
Nash, G. E., Glicksman, M. (1974), Acta Metall. 22, Seetharaman, V., Trivedi, R. (1988), Metall. Trans.
1283. 19 A, 2955.
Nelson, D. (1983), in: Phase Transitions and Critical Seetharaman, V. Eshelman, M., Trivedi, R. (1988),
Phenomena: Domb, C., Lebowitz, J. (Eds.). New Acta Metall. 36, 1175.
York: Academic Publ. Singh, N. B., Glicksman, M. E. (1989), J. Crystal
Newell, A. C., Whitehead, J. A. (1969), J. Fluid. Growth 98, 534.
Mech. 38, 279. Somboonsuk, K., Trivedi, R. (1985), Acta Metall. 33,
Ohno, K., Trinkaus, H., Müller-Krumbhaar, H. 1051.
(1990), J. Crystal Growth 99, 68. Somboonsuk, K., Mason, J. T., Trivedi, R. (1984),
Papapetrou, A. Z. (1935), Z. f. Kristallograph. 92, 89. Metall. Trans. A15, 967.
Park, J. S., Trivedi, R. (1998), J. Crystal Growth 187, Strain, J. (1989), J. Comput. Phys. 85, 342.
511. Temkin, D. E. (1985), Kristallografia 30, 1055.
Pelce, P., Clavin, P. (1987), Europhys. Lett. 3, 907. Temkin, D. E. (1990), unpublished.
Pelce, P., Pomeau, Y. (1986), Studies Appl. Math. 74, Tirmizi, S. H., Gill, W. N. (1989), J. Crystal Growth
1283. 96, 277.
Pelce, P., Pumir, A. (1985), J. Crystal Growth 73, 357. Tong, X., Beckermann, C., Karma, A. (1998), in:
Penrose, O., Fife, P. G. (1990), Physica D 43, 44. Modeling of Casting, Welding, and Advanced So-
Perepezko, J. H., Uttormark, M. J. (1996), Metall. lidification Conference VIII: Thomas, B. G., Beck-
Mater. Trans. 27 A, 533. ermann, C. (Eds.), Warrendale, PA: TMS, p. 613.
Perepezko, J. H., Mueller, B. A., Ohsaka, K. (1987), Tonhardt, R., Amberg, G. (1998), J. Crystal Growth
in: Undercooled Alloy Phases: Collings, E. W., 194, 406.
Koch, C. C. (Eds.). Warrendale: TMS, p. 289. Tournier, S., Barth, M., Herlach, D. M., Vinet B.
Plapp, M., Karma, A. (1999), Phys. Rev. E 60, 6865. (1997), Acta Mater. 45, 191.
Provatas, N., Goldenfeld, N., Dantzig, J. (1998), Trivedi, R. (1980), J. Crystal Growth 49, 219.
Phys. Rev. Lett. 80, 3308. Trivedi, R. (1984), Metall. Trans. 15 A, 977.
Quested, P. N., McLean, M. (1984), Mater. Science Trivedi, R., Karma, A., Lo, T. S., Mazumder, P.,
Eng. 65, 171. Plapp, M., Park, J. S. (1998), in: Zermatt Work-
Rabaud, M., Couder, Y., Gerard, N. (1988), Phys. shop on “Solidification Microstructures”, CD,
Rev. A37, 935. Swiss Federal Institute of Technology Lausanne,
Rappaz, M. (1989), Inter. Mater. Rev. 34, 93. 1015 Lausanne EPFL, Switzerland. This CD (PC
Rappaz, M., Desbiolles, J.-L., Gandin, C.-A., Henry, format) may be obtained from the Laboratory
S., Smoroz, A. (2000), Mater. Sci. Forum 329 – of Physical Metallurgy, EPFL through: arlette.
330, 389. blatti@epfl.ch
170 2 Solidification
Trivedi, R., Kurz, W. (1986), Acta Metall. 34, 1663. Wagner, C. (1961), Z. Elektrochem. 65, 581.
Trivedi, R., Laorchan, V. (1988), Acta Metall. 39, Wang, C. Y., Beckermann, C. (1996), Metall. Mater.
1941; 1951. Trans. 27 A, 2765.
Trivedi, R., Lipton, J., Kurz, W. (1987), Acta Metall. Wang, C. Y., Ahuja, S. Beckermann, C., de Groh III,
35, 965. H. C. (1995), Metall. Trans. 26 B, 111.
Trivedi, R., Mason, J. T., Verhoeven, J. D., Kurz, W. Wang, S.-L., Sekerka, R. F., Wheeler, A. A., Murray,
(1991), Metall. Trans. 22 A, 2523. B. T., Corriel, S. R., Braun, R. J., McFadden, G. B.
Trivedi, R., Somboonsuk. K. (1984), Mater. Sci. Eng. (1993), Physica D 69, 189.
65, 65. Wheeler, A. A., McFadden, G. B., Boettinger, W. J.
Trivedi, R., Somboonsuk, K. (1985), Acta Metall. 33, (1996), Proc. Royal Soc. London A 452, 495.
1061. Warren, J. A., Langer, J. S. (1990), Phys. Rev. A42,
Ungar, L. H., Brown, R. A. (1984 a), Phys. Rev. B 29, 3518.
1367. Wollkind, D., Segel, L. (1970), Philos. Trans. R. Soc.
Ungar, L. H., Brown, R. A. (1984 b), Phys. Rev. B 30, London 268, 351.
3993. Wu, Y., Piccone, T. J., Shiohara, Y., Flemings, M. C.
Ungar, L. H., Brown, R. A. (1985 a), Phys. Rev. B 31, (1987), Metall. Trans. 18, 915; 925.
1367. Wulff, G. (1901), Z. Kristallogr. Mineral. 34, 449.
Ungar, L. H., Brown, R. A. (1985 b), Phys. Rev. B 31, Xiao, R. F., Alexander, J. I. D., Rosenberger, F.
5931. (1988), Phys. Rev. A38, 2447.
Uwaha, M., Saito, Y. (1990), J. Crystal Growth 99, Yokoyama, E., Kuroda, T. (1988), in: Dynamics of
175. Ordering Processes in Condensed Matter: Ko-
Valance, A., Misbah, C., Temkin, D., Kassner, K. mura, S., Furukawa, H. (Eds.). New York: Plenum
(1993), Phys. Rev. E 48, 1924. Publ., p. 95.
Vandyoussefi, M., Kerr, H. W., Kurz, W. (2000), Acta Zettlemoyer, A. (Ed.) (1969), Nucleation I. New
Mater. 48, 2297. York: Dekker.
Versnyder, F. L., Shank, M. E. (1970), Mater. Sci. Zettlemoyer, A. (Ed.) (1976), Nucleation II. New
Eng. 6, 213. York: Dekker.
Volkmann, T., Wilde, G., Willnecker, R., Herlach, D. Zimmermann, M., Carrard, M., Kurz, W. (1989),
M. (1998), J. Appl. Phys. 83, 3028. Acta Metall. 37, 3305.
Voronkov, V. V. (1983), in: Modern Theory of Crys- Zimmermann, M., Karma, A., Carrard, M. (1990),
tal Growth: Chernov, A. A., Müller-Krumbhaar, Phys. Rev. B 42, 833.
H. (Eds.). Heidelberg: Springer Verlag. p. 75.
3 Diffusion Kinetics in Solids
Graeme E. Murch
x distance, coordinate
Xi , Xv driving force on species i, vacancies
Zi number of charges on i
g activity coefficient
G, Gq , Gi jump frequency (of charge q; of species i)
m chemical potential
n vibration frequency
s ionic conductivity
w exchange frequency
ally impossible to measure an atomic flux solution, C (x, t), of the diffusion equation
unless steady state is reached. Further, be- has been established, the diffusion coeffi-
cause solid-state diffusivities are generally cient itself is obtained as a parameter by
small, the attainment of steady state in fitting the experimental C (x, t) to the ana-
a macroscopic specimen can take a very lytical C (x, t). In the following we shall fo-
long time. In a few cases, where the solid- cus on some analytical solutions for some
state diffusivity is high, for example car- well-known initial and boundary condi-
bon diffusion in austenite (Smith, 1953), tions used in experimental diffusion stud-
the steady-state flux and the concentration ies.
gradient can be measured and the diffusiv- Let us first examine some solutions for
ity obtained directly from Eq. (3-1). Eq. (3-5). In a very common experimental
In order to produce a basis for measuring arrangement for “self” and impurity diffu-
the diffusion coefficient, Eq. (3-1) is usu- sion a very thin deposit of amount M of ra-
ally combined with the equation of con- dioactive isotope is deposited as a sand-
tinuity: wich layer between two identical samples
of “infinite” thickness. After diffusion for a
∂J ∂C
=− (3-3) time t the concentration is described by
∂x ∂t
M
to give Fick’s Second Law: C ( x, t ) = exp ( − x 2 / 4 Dt ) (3-8)
2 p Dt
∂C ∂ ⎛ ∂C ⎞
= ⎜D ⎟ (3-4) which is illustrated in Fig. 3-1. If the de-
∂t ∂x ⎝ ∂x ⎠ posit is left as a surface layer rather than a
sandwich, C (x, t) is doubled.
If the diffusion coefficient is independent
In another experimental arrangement,
of concentration and therefore position,
the surface concentration Cs of the diffus-
then Eq. (3-4) reduces to
ing species is maintained constant for time
∂C ∂2 C t, perhaps by being exposed to an atmo-
=D 2 (3-5) sphere of it. Again, the substrate is thick or
∂t ∂x
mathematically infinite. The solution of the
The second-order partial differential equa- diffusion equation is
tion Eq. (3-5) (or (3-4)) is sometiems called
the “diffusion equation”. Eq. (3-2) can also C ( x , t ) − Cs
= erf ( x / 2 Dt ) (3-9)
be developed in the same way to give C0 − Cs
In order to obtain a solution to the diffu- This function is now frequently available
sion equation, it is necessary to establish as a scientific library function on most
the initial and boundary conditions. Once a modern computers. Accurate series expan-
3.2 Macroscopic Diffusion 177
Figure 3-1. Time evolution of Eq. (3-8). Figure 3-2. Time evolution of Eq. (3-12).
clean section of surface. Another is the dif- of its practical significance the interdiffu-
fusion between two metal samples differ- sion coefficient is the one often quoted in
ing only in their relative concentrations of metal property data books (Brandes, 1983;
a highly mobile interstitial species such as Mehrer, 1990; Beke 1998, 1999). See also
H. A further example is diffusion between the following section for further discussion
two nonstoichiometric compounds, e.g., of D̃.
Fe1– d O and Fe1– d ¢ O. In all of these cases
only one species of atom is involved in the
3.2.2.4 Intrinsic or Partial Diffusion
diffusion process which brings the system
Coefficients
to a common composition (which is why it
can be considered a pseudo-one-compo- In contrast to the pseudo-one-component
nent system). Often the process can be pic- systems described above where the dif-
tured as the interdiffusion of vacant sites fusion rates of the atoms and vacant sites
and atoms. The diffusion coefficient de- are necessarily equal, in the substitutional
scribing this process is usually called the binary alloy the individual diffusion rates
chemical diffusion coefficient, sometimes of A and B are not generally equal since the
the interdiffusion coefficient, and occasion- corresponding self-diffusion coefficients
ally, the collective diffusion coefficient. are not. In the interdiffusion experiment
Generally the symbol used is D̃. For these this implies that there is a net flux of atoms
pseudo-one-component systems, the pre- across any lattice plane normal to the diffu-
ferred name is chemical diffusion coeffi- sion direction. If the number of lattice sites
cient. In general, the chemical diffusion co- is conserved, each plane in the diffusion re-
efficient does not equal the self-diffusion gion must then shift to compensate. This
coefficient because of effects arising from shift with respect to parts of the sample
the gradient of chemical composition, see outside the diffusion zone, say the ends of
Eq. (3-99). the sample, is called the Kirkendall effect.
Chemical diffusion in binary substitu- This effect can be measured by observing
tional solid solutions is frequently called the migration of inert markers, usually fine
interdiffusion. In a typical case pure metal insoluble wires which have been incorporat-
A is bonded to pure metal B and diffusion ed into the sample before the experiment.
is permitted at high temperature. Although The assumption is that the wires follow the
both A and B atoms move, only one con- motion of the lattice in their vicinity.
centration profile, say of A, is established The intrinsic diffusion coefficients of A
(the profile from B contains no new infor- and B, DAI and DBI , are defined with refer-
mation). The resulting diffusion coefficient ence to the fluxes of A and B relative to the
which is extracted from the profile, by the local lattice planes:
Boltzmann – Matano analysis (e.g., see Sec.
∂CA
3.2.1) is termed the interdiffusion coeffi- JA′ = − DAI (3-15)
cient and is given the symbol D̃. Not infre- ∂x
quently, the diffusion coefficient for this and
situation is loosely called the chemical dif- ∂CB
fusion coefficient or the mutual diffusion JB′ = − DBI (3-16)
∂x
coefficient. This single diffusion coeffi-
cient is sufficient to describe the concentra- The diffusion coefficients are sometimes
tion profile changes of the couple. Because also termed partial diffusion coefficients.
180 3 Diffusion Kinetics in Solids
The set of equations represented by Eq. e.g., the tracer correlation factor and the
(3-22) are normally called the “phenomen- vacancy-wind effect, can be conveniently
ological equations” or Onsager equations expressed in terms of the Ls. This can
and were postulated as a central part of the sometimes aid in understanding the nature
theory of irreversible processes. Details of of these complex correlations. A most im-
the theory as it applies to diffusion can be portant contribution to this was made by
found in many places but see especially the Allnatt (1982), who related the Ls directly
reviews by Howard and Lidiard (1964), to the microscopic behavior – see Sec.
Adda and Philibert (1966), and Allnatt and 3.3.1.8.
Lidiard (1993). In the next sections we shall restrict our-
The phenomenological coefficients are selves to isothermal diffusion and focus on
sometimes said to have a “wider meaning” the relations which can be derived among
than quantities such as the diffusion coeffi- the Ls and various experimentally acces-
cients or the ionic conductivity. The wider sible transport quantities.
meaning comes about because the pheno-
menological coefficients do not depend on 3.2.3.1 Tracer Diffusion
driving force but only on temperature and
composition. In principle, armed with a full Let us first consider tracer diffusion in a
knowledge of the Ls, the technologist pure crystal. The general strategy is (1) to
would have the power, if not to control the describe the flux of the tracer with the On-
diffusive behavior of the material, at least sager equations (Eq. (3-22)) and (2) to de-
to predict the diffusive behavior no matter scribe the flux with Fick’s First Law (Eq.
what thermodynamic force or forces or (3-1)) and then equate the two fluxes to
combination thereof were acting. find expressions between the diffusion co-
Unfortunately, the experimental determi- efficient and the phenomenological coeffi-
nation of the Ls is most difficult in the solid cients. We shall deal with this in detail to
state. This is in contrast, incidentally, to show the typical procedure (see, for exam-
liquids, where, by use of selectively perme- ple, Le Claire (1975)). We can identify two
able membranes, measurement of the Ls is components, A and its tracer A*. We write
possible. The reader might well ask then for the fluxes of the host and tracer atoms,
why the Ls were introduced in the first respectively,
place when they are essentially not amen- JA = LAA XA + LAA* XA* (3-26)
able to measurement! There are several
and
reasons for this. First, relations can be de-
rived between the Ls and the (measurable) JA* = LA*A* XA* + LA*A XA (3-27)
tracer diffusion coefficients – we will dis-
Strictly, the vacancies should enter as a
cuss this further in Sec. 3.2.3.4. Further,
component but it is unnecessary here be-
the Onsager equations provide a kind of ac-
cause of the lack of a vacancy gradient.
counting formalism wherein an analysis
Since the tracer atoms and host atoms
using this formalism ensures that once the
mix ideally (they are chemically identical),
components and driving forces have all
it can easily be shown that the driving
been identified nothing is overlooked, and
forces are
that the whole is consistent; this will be
discussed in Secs. 3.2.3.1 and 3.2.3.2. k T ∂cA
XA = − (3-28)
Finally, many “correlations” in diffusion, cA ∂x
3.2 Macroscopic Diffusion 183
For a pure crystal and where the vacancy Following the same kind of procedure as
concentration is very low, HR can be iden- before (see Sec. 3.2.3.1), but recognizing
tified directly with the tracer correlation that generally A and B do not ideally mix,
factor f (see Sec. 3.3.2) and we find that rather than by Eq. (3-28), Xi is
now given by
LA*A*
HR = f = (3-42) k T ∂ci ⎛ ∂ ln g ⎞
LA*A* + LA*A Xi = − ⎜1 + ⎟ (3-49)
ci ∂x ⎝ ∂ ln c ⎠
Note that if LA*A = 0, i.e., there is no inter-
ference between tracer and host, then where g is the activity coefficient of either A
HR = 1. More generally, when the defect or B. We find that the diffusion coefficient
concentration is high and there are corre- of, say A, actually an intrinsic diffusion co-
lations in the ionic conductivity (see Sec. efficient (see Sec. 3.2.2.4), is given by
3.3.1.7), we find that k T V ⎛ LAA LAB ⎞ ⎛ ∂ ln g ⎞
DAI = ⎜ − ⎟ ⎜1 + ⎟ (3-50)
LA*A* N ⎝ cA cB ⎠ ⎝ ∂ ln c ⎠
HR = f / f I = (3-43)
LA*A* + LA*A If we want to find a relation between DAI
and f I is the physical correlation factor. and the tracer diffusion coefficient, say
DA* , we would need to introduce A* for-
mally into the phenomenological equations
3.2.3.2 Chemical Diffusion which would now become (Stark, 1976;
For chemical diffusion between A and B Howard and Lidiard, 1964)
when, say, the vacancy mechanism is oper- JA = LAA XA + LAA* XA* + LAB XB (3-51)
ating (see Sec. 3.3.1.1), the Onsager equa-
tions are written as (Howard and Lidiard, JA* = LA*A* XA* + LA*A XA + LA*B XB (3-52)
1964) JB = LBB XB + L BA XA + L BA* XA* (3-53)
JA = LAA XA + LAB XB (3-44) These equations can be developed to give
JB = LBB XB + L BA XA (3-45) k TV ⎛ LA*A* LA*A ⎞
DA* = ⎜ − ⎟ (3-54)
Strictly, the vacancies should enter here as N ⎝ cA* cA ⎠
another species, so that we would write and
JA = LAA XA + LAB XB + LAV XV (3-46) k TV ⎛ LAA + LA*A LAB + LA*B ⎞
DAI = ⎜ − ⎟
JB = LBB XB + L BA XA + L BV XV (3-47) N ⎝ cA cB ⎠
This is in fact the so-called Darken equa- placement of the system of species i as if
tion (Darken, 1948), relating the intrinsic that system were a particle itself. The diag-
diffusion coefficient to the tracer diffusion onal phenomenological coefficients derive
coefficient but with the addition of the term from correlations of the system of species i
in brackets. This term is sometimes called with itself, i.e., self-correlations. This does
the “vacancy-wind-term” and does not not mean tracer correlations here: they are
vary much from unity. Note that it results correlations in the random walk of an atom.
from non-zero cross terms LA*A and LAB . The correlations here are in the random
The approach from the phenomenological walk of the system “particle”. The off-diag-
equations has told us only of the formal ex- onal terms derive from interference of the
istence of this term, but an evaluation of it system of species i with the system of spe-
requires consideration of the detailed mi- cies j. In effect it is mathematically equiva-
croscopic processes which generate the lent to two systems i and j being treated
correlations contained in the cross terms. like two interfering “particles”.
This is considered in Sec. 3.3.1.8. Eq. (3-57) has been very useful for cal-
These examples suffice to show how the culating the L ij by means of computer sim-
phenomenological equations are useful in ulation – see the pioneering calculations
presenting a consistent and unified picture by Allnatt and Allnatt (1984). Much of that
of the diffusion process, no matter how com- material has been reviewed by Murch and
plex. For further information on the subject, Dyre (1989). A brief discussion is given in
the reader is directed to the classic review by Sec. 3.3.1.8.
Howard and Lidiard (1964) and also many
other treatises, such as those by Adda and 3.2.3.4 Relating Phenomenological
Philibert (1966), Le Claire (1975), Stark Coefficients to Tracer Diffusion
(1976), Kirkaldy and Young (1987), Phili- Coefficients
bert (1991), and Allnatt and Lidiard (1993).
For the case of multi-component alloys,
Manning (1968, 1970, 1971) derived expres-
3.2.3.3 Einsteinian Expressions for sions relating the phenomenological coeffi-
the Phenomenological Coefficients cients to tracer diffusion coefficients. The ex-
An important development in the area of pressions were developed for a particular
solid-state diffusion was the fact that the model for concentrated alloys, the so-called
phenomenological coefficients can be ex- random alloy model. In this model the atomic
pressed directly in terms of atomistic Ein- components are randomly distributed over
steinian formulae (Allnatt, 1982). the available sites, the vacancy mechanism is
assumed and the exchange frequencies of the
Lij = lim lim (6 V k T t ) −1 atomic components with the vacancies de-
V→ ∞ t→ ∞
pend only on the nature of the atomic compo-
× 〈 DR (t ) ⋅ DR( j ) (t )〉
(i )
(3-57) nents and not on their environment.
where DR i (t) is the total displacement of For a binary system the relations are
species i in time t, V is the volume, and the N
Dirac brackets denote a thermal average A. LAA = cA DA* (3-58)
kT
It is important to note that DR(i) is the sum
⎡ (1 − f0 ) cA DA* ⎤
of the displacements of the individual par- × ⎢1 + ⎥
ticles of type i. In effect, DR(i) is the dis- ⎢⎣ f0 ( cA DA* + cB DB* ) ⎥⎦
186 3 Diffusion Kinetics in Solids
cuss short-circuit diffusion from a pheno- volume diffusion, then this condition
menological point of view. D¢/D > 106 is usual for temperatures below
Fisher (1951) was probably the first to about half the melting temperature. This is
introduce a distinct diffusion coefficient why experiments where lattice diffusion
for material migrating along the short-cir- only is of interest tend to be made at tem-
cuit path. For grain boundaries, each one is peratures above about half the melting tem-
conceived as a slab of width d for which perature.
the average diffusion coefficient is D¢. For Impurities are often bound to short-cir-
dislocations, each one is conceived as a cuit paths, in which case the Hart Equation
pipe of diameter 2 a for which the average (Eq. (3-60)) is written as
diffusion coefficient is D¢.
D¢ = ᑠ D¢ exp (Eb /k T ) + (1 – ᑠ ) Dl (3-62)
For conventional diffusion experiments
it has been usual to distinguish three dis- where Eb is the binding energy of the im-
tinct regimes, depending on the magnitude purity and the diffusion coefficients refer
---
of the lattice diffusion length ÷Dl t, where to impurities. The Hart Equation can be
Dl is the lattice diffusion coefficient and t is shown to be fairly accurately followed
--- ---
time (Harrison, 1961). when l /÷Dl t ≤ 0.3 and also for l /÷Dl t ≥ 100
---
Regime A: ÷Dl t is much larger than the (this corresponds to Regime C – see the
spacing l between the short-circuit paths. next paragraph) (Murch and Rothman,
For grain boundaries l is the average diam- 1985; Gupta et al., 1978).
eter of the grains and for dislocations l is Regime C: It is convenient to discuss the
---
the distance between dislocation pinning other limit out of sequence. When ÷Dl t is
points. Diffusion from adjacent short cir- much smaller than the distances l between
cuits overlap extensively. This condition the short-circuit paths, we have Regime C
is met for small-grained materials or very kinetics. In this instance, all material
long diffusion times. Hart (1975) proposed comes down the short-circuit paths and the
that an effective diffusion coefficient Deff measured diffusion coefficient is given en-
can be introduced which still satisfies solu- tirely by D¢.
tions to Fick’s Second Law and can be Regime B: In this intermediate case, it is
---
written as assumed that ÷Dl t is comparable to l so
that material which is transported down a
Deff = ᑠ D¢ + (1 – ᑠ ) Dl (3-61)
short-circuit path and which diffuses out
where ᑠ is the fraction of all sites which into the lattice is unlikely to reach another
belong to the short-circuit paths. Eq. (3-61) short-circuit path.
is usually called the Hart Equation. Le There are various solutions of Fick’s
Claire (1975) has given a rough calculation Second Law available to cope with tracer
using Eq. (3-61). For typical dislocation diffusion in the presence of short-circuit
densities in metals, ᑠ = 10– 8. The disloca- paths, but space prevents us from dealing
tion contribution to a measured or effective with these in any detail: they are discussed
D will then exceed 1% when D¢/D > 106. at length by Kaur, Mishin and Gust (1995).
Because the activation energy for lattice Suzuoka (1961) and Le Claire (1963) have
diffusion in metals is about 34 TM. Pt. , given a solution for the grain boundary
where TM.Pt. is the melting temperature problem for the usual case when there is a
and the activation energy for short-circuit finite amount of tracer originally at the sur-
diffusion is generally about half that of face, see Eq. (3-8). It is found, among other
188 3 Diffusion Kinetics in Solids
gies in close-packed solids suggest that this plus higher order terms. In Eq. (3-65), C (x0 )
mechanism is unlikely. It may, however, is the concentration of the diffusing atoms
occur in liquids and in quenched-liquids at x0 , ·XÒ is the mean displacement or drift,
such as amorphous alloys (Jain and Gupta, and ·X 2 Ò is the mean squared displacement.
1994). Ring versions of the exchange The Dirac brackets · Ò denote an average
mechanism certainly have lower theoreti- over all possible paths taken in time t.
cal activation energies but require substan- In a situation where diffusion properties
tial cooperation among the atoms, which do not depend on position, e.g., diffusion
seems unlikely. of tracer atoms in a chemically homogene-
ous system, the third term in Eq. (3-65) is
Surface Diffusion Mechanisms zero. The term containing the drift ·XÒ is
A number of mechanisms for surface also zero because in a chemically homoge-
diffusion on metals (by far the most studied neous system the probability that, say, a
class of material) have been postulated. tracer atom migrates some distance + X in
They include activated hopping of ad- time t (starting from x0 ) equals the prob-
sorbed atoms from one surface site to an- ability that an atom migrates a distance – X.
other, where the jump distance is simply Similarly, other odd moments are zero.
the distance between sites. Similarly, a va- Provided that other higher-order terms can
cancy in a terrace, i.e., surface vacancy, be neglected, Eq. (3-65) reduces to
can also move in the same sort of way ∂C 〈 X 2 〉
within the terrace. At low temperatures and Jx = (3-66)
∂x 2 t
rough surfaces, exchange between an ad-
sorbed atom and an atom in the substrate is By comparison with Fick’s First Law, Eq.
predominant. At high temperatures non-lo- (3-1), Eq. (3-66) immediately gives
calized diffusion and surface melting are
also possible. The mechanisms for surface 〈X2 〉
Dx* = (3-67)
diffusion are discussed further in the book 2t
edited by Vu Thien Binh (1983) and in the Eq. (3-67) is called the Einstein Equation,
reviews by Rhead (1989) and Bonzel probably the single most important equa-
(1990). tion in the theory of diffusion kinetics. The
superscript * indicates that the diffusion
3.3.1.2 The Einstein Equation coefficient refers to tracer atom diffusion
Let us consider atoms (of one chemical in a chemically homogeneous system.
type) diffusing in their concentration gradi- Equations for Dy* and Dz* have the same
ent in the x direction. It can readily be form as Eq. (3-67). For three-dimensional
shown (see, for example, Adda and Phili- isotropic crystals the tracer diffusion coef-
bert, 1966; Manning, 1968; Le Claire, ficient is the same in every direction and
1975) that the net flux of atoms across a
〈 R2 〉
given plane x0 is given by D* = (3-68)
6t
〈 X 〉 ∂C 〈 X 2 〉
J x = C ( x0 ) − where R is the vector displacement of an
t ∂x 2 t atom in time t. For two-dimensional situa-
∂C ∂ 〈 X 2 〉 tions the factor 6 in Eq. (3-68) is replaced
− C ( x0 ) (3-65)
∂x ∂C 2 t by 4.
3.3 Microscopic Diffusion 193
cancy remains in the vicinity of the tracer It should be noted that the limit n Æ • is
and can approach the tracer from a non- applied to Eq. (3-74) to ensure that all pos-
random direction. sible correlations are included.
Let us consider Fig. 3-8 b, where the col- There have been numerous publications
linear interstitialcy mechanism is assumed. concerned with the calculation of the tracer
Let us assume that the tracer is in the inter- correlation factor. The earlier work has been
stitial position. The first jump (a pair of reviewed in detail by Le Claire (1970),
atoms move) will take the tracer immedi- more recent work has been covered by All-
ately to a regular lattice site and a new natt and Lidiard (1993). Table 3-1 gives
atom, formerly at a regular lattice site, some values of f for various mechanisms
takes up a position as the interstitial. The and lattices. It should be noted from Table
pair of atoms which next moves possibly 3-1 that the correlation factor for the inter-
again involves the tracer atom and in doing stitial mechanism is unity. For this mecha-
so will cancel out the previous tracer jump. nism the interstitials, which are considered
Like the vacancy mechanism, the memory to be present at a vanishingly small concen-
or correlation between successive tracer tration, move on an uncorrelated random
atom jump directions comes about purely walk, very much like the vacancy in the va-
because of the proximity of the defect. Un- cancy-assisted diffusion mechanism. Accord-
like the vacancy diffusion mechanism, ingly, the second term in Eq. (3-71) drops out
however, in this case the correlation comes and f = 1. If the interstitial concentration is
about only for consecutive pairs of tracer increased, the interstitials impede one an-
jumps of the type interstitial site Æ regular other and correlations are introduced; as a
lattice site Æ interstitial site. result, the tracer correlation factor de-
In the presence of correlation between creases from unity. In fact, it continues to
jump directions, the sum of the dot prod- decrease until only one interstice is left va-
ucts ·ri · ri + j Ò no longer averages out to cant. The situation now is identical to that
zero. A convenient way of expressing these for vacancy-assisted diffusion. The varia-
correlations quantitatively is to form the tion of f with vacancy (or vacant interstice)
ratio of the actual ·R2 Ò to the ·R2 Ò result- concentration in the f.c.c. lattice is shown
ing from a complete random walk, i.e., in Fig. 3-9. Of course, in the unphysical sit-
·R2 Òrandom : uation where the interstitials do not “see”
one another, i.e., multiple occupancy of a
〈 R2 〉 site is permitted, the interstitials continue
f= (3-73)
〈 Rrandom
2
〉 to move on a complete random walk at all
concentrations, and f always equals unity.
where the quantity f is called the tracer cor-
For cubic lattices all the jumps are of the
relation factor or simply the correlation
same length. Then we have that | ri | = r :
factor. Sometimes the terminology “corre-
n
lation coefficient” is used, but this is to be
discouraged. With Eqs. (3-71) and (3-72) ∑ 〈 ri2 〉 = n r 2 (3-75)
i =1
we have
n −1 n − i Then we have, with Eqs. (3-68), (3-71),
2 ∑ ∑ 〈 ri ⋅ ri + j 〉 and (3-75)
i =1 j =1
f = lim 1 + (3-74)
n→∞ n n r2 f G r2 f
∑ 〈 ri2 〉 D* =
6t
=
6
(3-76)
i =1
3.3 Microscopic Diffusion 195
Table 3-1. Some correlation factors (at infinitely low defect concentrations) from Le Claire (1970), Manning
(1968), and Murch (1982 d).
Lattice Mechanism f
First, as can be seen from Table 3-1, the impurity. We shall deal with this subject in
tracer correlation factor is quite sensitive to the following section.
the mechanism of diffusion operating. Al-
though f, by itself, is not measurable, mea- 3.3.1.4 Impurity Correlation Factor
surements of the Haven ratio (see Sec.
We first consider that the impurity con-
3.3.2) and the isotope effect (see Sec. 3.3.3)
centration is sufficiently dilute that there is
(which are closely related to f ) in favorable
not a composition-dependent impurity dif-
cases can throw considerable light on the
fusion coefficient. Accordingly, each im-
mechanism(s) of diffusion that are operat-
purity atom is considered to diffuse in pure
ing. Identification of the diffusion mecha-
host. The jump frequency of the impurity is
nism is surely the most important ingredi-
given by w 2 whereas the jump frequency of
ent in understanding the way atoms mi-
the host atoms is given by w 0 . The signifi-
grate in solids and how it can be controlled.
cance of the subscripts will be apparent
Much of our discussion so far has been
later. The vacancy mechanism is assumed.
concerned with pure solids with few de-
The “tracer” now is the impurity. The
fects. There are, however, a large number
result for the impurity correlation factor f 2
of solids where the apparent defect concen-
in the f.c.c. lattice as a function of w 2 /w 0
tration can be fairly high, e.g., highly non-
is shown in Fig. 3-10. When w 2 < w 0 (upper
stoichiometric compounds, fast ion con-
curve), the impurity motion becomes
ductors, and certain concentrated intersti-
considerably decorrelated since, after an
tial solid solutions. In such cases the de-
impurity/vacancy exchange, the vacancy
fects interact and correlation effects tend to
does not trend to remain in the vicinity of
be magnified and become highly tempera-
the impurity. When the impurity is next
ture dependent. Then the apparent activa-
tion energy for tracer diffusion includes an
important contribution directly from the
correlation factor. Clearly, an understand-
ing of f is a very important part of the
understanding of the diffusion process in
such materials. We shall deal further with
this subject in Sec. 3.3.1.6. In ordered
binary alloys, the correlation factor can
become very small since atoms which
make a jump from the “right” lattice to the
“wrong” lattice tend to reverse, i.e., cancel
that jump. As a result, the tracer diffusion
coefficient can become abnormally small,
largely because of this strong memory ef-
fect. We shall deal further with this subject
in Sec. 3.3.1.5.
Another important case of correlation ef-
Figure 3-10. Impurity correlation factor in the f.c.c.
fects is impurity diffusion. The correlation
lattice as a function of w 2 /w 0 : points by computer
factor for the impurity is highly dependent simulation (Murch and Thorn, 1978), solid lines
on the relative jump frequencies of the im- from the formalism of Manning (1964); after Murch
purity and host atoms in the vicinity of the and Thorn (1978).
3.3 Microscopic Diffusion 197
Other more accurate expressions for F nearest neighbor by means of the assump-
have been reviewed by Allnatt and Lidiard tions w 3 = w 3¢ = w ≤3 and w 5 = w 6 = w 0 . This
(1987, 1993). For many practical purposes also implies that w 4 = w 4¢ = w ≤4 .
the extra rigor is probably unwarranted. Model I leads to the expression (Man-
Extension to the case where there are va- ning, 1964; Le Claire, 1970)
cancy – impurity interactions at second
7w 3′ F
nearest neighbors has been considered by f2 = (3-84)
Manning (1964). 2 w 2 + 7w 3′ F
In the b.c.c. lattice the second nearest with F given by
neighbor is close to the nearest neighbor
(within 15% in fact). In a formal sense we 2 a 2 + 5.1817 a + 2.476
7F = (3-85)
need to consider dissociative jumps for a a + 0.8106
vacancy, jumping from the first to the sec- and a = w 3 /w 4¢ .
ond, third (not the fourth), and fifth nearest Model II leads to the expression
neighbors: these are denoted by w 3 , w 3¢ ,
and w ≤3 . 7w 3 F
f2 = (3-86)
The corresponding reverse associative 2 w 2 + 7w 3 F
jumps are w 4 , w 4¢ , and w ≤4 . We also need to
where F is given by
consider dissociative jumps from the sec- (3-87)
ond nearest neighbor to the fourth (w 5 ) and 3a + 33.43a + 97.38 a + 66.06
3 2
This form for the impurity correlation fac- troduced which limit the number of jump
tor is especially relevant to the isotope ef- frequencies but only as a result of some
fect in diffusion (see Sec. 3.3.3) where for loss of realism.
self-diffusion the isotopes are tracers of the The first of these is the “random alloy”
host but, being isotopes, actually have model introduced by Manning (1968,
slightly different jump frequencies from 1970, 1971). The random alloy is of con-
the host and can be classified as “impur- siderable interest because, despite its sim-
ities”. We might also mention that for im- plicity, it seems to describe fairly well the
purity diffusion experiments the isotopes diffusion behavior of a large number of al-
can additionally be of the impurity. loys. The atomic components (two or
As a closing remark for this section, it more) are assumed to be ideally mixed and
should be remembered that because of its the vacancy concentration is assumed to be
mathematical form, the impurity correla- very low. The atomic jump frequencies,
tion factor will be temperature dependent. e.g., wA and w B , in a binary alloy are expli-
Over a fairly small temperature range the citly speficied and neither changes with
temperature dependence can frequently be composition or environment. In the very
approximated by an Arrhenius-like expres- dilute limit this model formally corre-
sion, i.e., sponds, of course, to specifying a host
jump frequency w 0 and an impurity jump
f 2 ≈ f 20 exp (– Q¢/k T ) (3-90)
frequency w 2 with all other host jump fre-
where Q¢ is some activation energy for the quencies equal to w 0 . However, it should
correlation process (this has no particular be appreciated for this model that physi-
physical meaning), k is the Boltzmann con- cally, although not mathematically, the
stant, and f 20 is a “constant”. It should be jump frequencies wA and w B are conceived
noted that Q¢ will be unavoidably included to be “average” jump frequencies. The
in the activation energy for the entire im- average jump frequency of the vacancy, wv ,
purity diffusion process and is not neces- was postulated to be given by
sarily unimportant.
wv = cA wA + cB w B (3-91)
3.3.1.5 Correlation Factors for where cA and cB are the atomic fractions
Concentrated Alloy Systems of A and B. Manning (1968, 1970, 1971)
finds that
The impurity correlation factors dis- (3-92)
S + ( 2 − 3 x0 + cB ) w B /w A + ( cA − x0 )
cussed in the previous section are probably fA =
appropriate for impurity concentrations up 2 {S + (1 − x0 ) w B /w A + x0 }
to about 1 at.%. At higher concentrations where
the impurities can no longer be considered
independent in the sense that the correla- S = {[(cB – x 0 ) wB /wA + (cA – x 0 )]2
tion events themselves are independent. + 4 x 0 (1 – x 0 ) wB /wA }1/2 (3-92 a)
This presents a special difficulty because
there is no easy way of extending, say, the and
five-frequency model into the concentrated x0 = 1 – f (3-92 b)
regime without rapidly increasing the num-
ber of jump frequencies to an unworkable and f is the correlation factor for vacancy
number. As a result, models have been in- diffusion in the pure metal A or B.
200 3 Diffusion Kinetics in Solids
Early Monte Carlo calculations indi- not given explicitly as in the random alloy
cated that Eq. (3-92) was a very good ap- model but are expressed in the following
proximation over a wide range of wA /w B way in terms of the interaction energies
but more recently it has been found that (bonds) for a given atom in a given config-
this was largely illusory (Belova and uration:
Murch, 2000). Eq. (3-92) is actually only a
reasonable approximation when the ex- wi = n i exp (– Ui /k T )
change frequency ratio is within about an ¥ exp [(gi Eii + g j Ejj )/k T ], (3-94)
order of magnitude of unity. A much better i = A, B i( j
approximation to the problem is that given
by Moleko et al. (1989), but unfortunately where Eij < 0, n i is the lattice vibration fre-
the equations are much more cumbersome quency, Ui is the reference saddle point en-
to use. ergy in the absence of interactions, gi is the
A different approach to diffusion in con- number of atoms of the same type which
centrated systems has been taken by Kiku- are nearest neighbors of the given atom and
chi and Sato (1969, 1970, 1972). They de- g j is the number of atoms of the other type
veloped their Path Probability Method which are nearest neighbors of the given
(PPM) to cope specifically with the prob- atom. Other forms of wi are also possible
lems of diffusion in concentrated systems. but have not been developed in this con-
The method can be considered to be a time- text.
dependent statistical mechanical approach. In the PPM a path probability function
They started with the so-called binary alloy is formulated and maximized, a process
analogue of the Ising antiferromagnet which is said to be analogous to mini-
model. In this model, sometimes called the mizing the free energy in equilibrium sta-
“bond” model, interactions EAA , EBB , and tistical mechanics. The first calculations
EAB are introduced between nearest neigh- were made using what was called a time-
bor components of the type A-A, B-B, and averaged conversion. The results were in
A-B, respectively. Equilibrium properties poor agreement with later Monte Carlo
of this model are well known; see for ex- computer simulation results. Subsequently,
ample, Sato (1970). It is convenient to fo- substantial improvements were made
cus on the ordering energy E defined by (Sato, 1984) and the PPM results are now
in fairly good agreement with computer
E = EAA + EBB – 2 EAB (3-93)
simulation results. A typical comparison of
(note that in the literature there are other the earlier results with Monte Carlo results
definitions of the ordering energy which from Bakker et al. (1976) is shown in Fig.
differ from this one by either a negative 3-12. Note the bend at the order/disorder
sign or a factor of 2). When E > 0, there is transition. The lower values of the correla-
an ordered region in the b.c.c. and s.c. lat- tion factors on the ordered side can be
tices (the f.c.c. lattice is more complicated ascribed generally to the higher probability
but has not been investigated by the PPM). of jump reversals as an atom which jumps
The ordered region is symmetrically placed from the “right” sublattice to the “wrong”
about cA = 0.5. When E < 0, there is a two- sublattice tends to reverse the jump while
phase region at lower temperatures (this the vacancy is still present. This cancella-
side has not been investigated by the PPM). tion of jumps obviously leads to small val-
The exchange frequencies wA and wB are ues of f.
3.3 Microscopic Diffusion 201
to a greater or lesser extent in real materi- from the right lattice to the wrong lattice it
als. Surprisingly, some materials such as tends to reverse that jump, thereby giving
interstitial solid solutions are described low values of the correlation factor. As in
fairly well by the lattice gas model even the ordered alloy, there are sequences of
with only nearest neighbor interactions. jumps which lead to long-range diffusion.
This must be a result of short-range inter- By and large these sequences are based
actions being by far the most important. around interstitialcy progressions, but
Sato and Kikuchi (1971) have employed rather than two atoms moving in unison as
the PPM to good effect in the lattice gas in the actual interstitialcy mechanism, here
model. Extensive calculations have been they are separated in time.
made, especially for the honeycomb lattice. The temperature dependence of f in the
Equally extensive Monte Carlo calcula- lattice gas, like the alloy, is fairly strong.
tions have also been made both for this lat- Again, it is usual to write f in an Arrhenius
tice and many others; see the review by fashion, e.g., Eq. (3-90), but there is no
Murch (1984 a). A typical result is shown known physical reason for assuming that f
in Fig. 3-14. The pronounced minimum in must really take this form. Fig. 3-15 shows
the correlation factor is due to the effects of the behavior in the square planar lattice
the ordered structure. This ordered struc- gas. The Arrhenius plot is in fact curved
ture consists of atoms and vacant sites ar- above and below the order/disorder tem-
ranged alternately. The wide composition perature, though in an experimental study
variability of the ordered structure at this this would be overlooked because of a
temperature (indicated by the arrows) is ac- much smaller temperature range. This acti-
commodated by vacant sites or “intersti- vation energy will be included in the ex-
tials” as appropriate. When an atom jumps perimental tracer diffusion activation en-
ergy. This is likely to be a significant con-
tribution which (like in the ordered alloy)
cannot be ignored.
The subject of tracer correlation effects
in defective materials with high defect con-
Figure 3-14. PPM results for the dependence of the Figure 3-15. Arrhenius plot of the tracer correlation
tracer correlation factor on ion site fraction ci in the factor gained from Monte Carlo simulation of a
honeycomb lattice with nearest neighbor repulsion. square planar lattice gas with 50% particles and 50%
T * = k T /fnn ; after Murch (1982 d). vacant sites (Zhang and Murch, 1990).
204 3 Diffusion Kinetics in Solids
where gA is the activity coefficient of A. where f is the tracer correlation factor for
However, it is straightforward to show that diffusion in the pure crystal, i.e., pure A or
the rigorous equation is Eq. (3-56) which pure B. It turns out that the maximum in rA
can be rewritten in terms of the correlation in this model occurs when cA Æ 1. In the
functions (Le Claire, 1975; Murch, 1982 a; extreme case when the jump frequency of
Stark, 1976): A is much larger than that of B, then
⎛ ∂ ln g A ⎞ rA Æ f0–1. Therefore, when f0 = 0.78146 (va-
DAI = DA* ⎜ 1 + ⎟ cancy diffusion on the f.c.c. lattice, see Ta-
⎝ ∂ ln cA ⎠
ble 3-1), rA can only enhance chemical dif-
⎡⎛ (A) cA ⎞ ⎤ fusion by a factor of 1.28.
× ⎢⎜ fAA − fAB ⎟ fA ⎥ (3-116)
⎣⎝ cB ⎠ Manning’s equation (Eq. (3-117)) seems
⎦
to have more general significance than
The fA in the denominator is the correlation application to the random alloy. Murch
factor for the actual composition of the al- (1982 a) showed that this equation performs
loy and not some impurity limit. In the lit- very well indeed for an interacting alloy
erature the term derived from correlations model that can exhibit short and long-range
[ ] is sometimes called a vacancy-wind fac- order. This suggests that this equation can
tor and is given the symbol rA . As it turns be used almost with impunity. This subject
out, however, this term cannot vary much arises again in the “Manning relations”
from unity since the original Darken equa- which relate the phenomenological coeffi-
tion is reasonably well obeyed. The behav- cients to the tracer diffusion coefficients
ior of rA is not especially transparent. The (see Sec. 3.2.3.4).
vacancy flux in chemical diffusion is al- The subject of the experimental check of
ways in the same direction as that of the rA has been discussed by Bocquet et al.
slower moving species (and opposite to (1996). They pointed out that in most cases
that of the faster moving species). The va- the Manning formulation for rA and rB
cancy-wind effect always tends to provide (Eq. (3-117)) improves the agreement be-
a given atom with an enhanced probability tween experimental and calculated values
of flowing in a direction opposite to that of of the Kirkendall shift and the ratio of
the vacancy flow. When A is less mobile the intrinsic diffusion coefficients DAI /DBI .
than B, the vacancy-wind factor rA effec- However, the individual experimental val-
tively makes DAI even smaller. Conversely, ues of DAI and DBI often tend to be quite a
when A is more mobile than B, the va- bit higher than the calculated values. The
cancy-wind factor rA effectively makes DAI experimental Kirkendall shift also has a
even larger. But how much? This can be tendency to be higher than the calculated
partially answered in the following. value. Carlson (1978) points out that the
Although we have given Manning’s problem could be due to the random alloy
(1968) expressions for the individual fij assumption in the Manning formulation.
and the correlation factor fA for the random However, the success of the Manning for-
alloy (Eqs. (3-108) and (3-109)), it may not mulation for the interacting alloy model
be immediately obvious that rA is in fact mentioned above seems to vindicate the
given in this model by (see also Sec. random alloy assumption and so the reason
3.2.3.4) for the discrepancy probably should be
f DB* + cA ( DA* − DB* ) sought elsewhere.
rA = (3-117)
f ( cA DA* + cB DB* )
3.3 Microscopic Diffusion 211
3.3.2 The Nernst – Einstein Equation When the external force is the result of an
and the Haven Ratio electric field E we have
The Nernst – Einstein equation relates Fd = Z e E (3-123)
the d.c. ionic conductivity to a diffusion
where Z is the number of charges (ionic
coefficient. Probably no other equation in
valence) and e is the electronic charge.
diffusion has generated more misunder-
The mobility u is defined as the velocity
standing than this one. Let us consider the
per unit field and so we have
standard derivation (see, for example,
Batchelor (1976)), and discuss its implica- u Ze
= (3-124)
tions in detail. ˜ kT
D
We consider a pseudo-one-component
system in a situation where the flux result- In the solid-state diffusion literature Eq.
ing from an applied force on the particles (3-124) is generally called the Nernst – Ein-
(which are completely noninteracting) ex- stein relation. Because the d.c. ionic con-
actly counterbalances the flux due to diffu- ductivity is related to the mobility by s =
sion. That is, from Eq. (3-2) C Z e u, Eq. (3-124) can be rewritten as
dC s C Z 2 e2
〈v 〉 C = D
˜ (3-118) = (3-125)
dx D̃ kT
It is important to note that the diffusion co- More generally, interactions are present
efficient here refers to a chemical composi- between the particles and it can be shown
tion gradient and is most definitely con- that the general form of the Nernst – Ein-
ceived as a chemical diffusion coefficient, stein equation is in fact (Murch, 1982 b)
not a tracer or self-diffusion coefficient.
Lack of appreciation of this fact leads to s C Z 2 e 2 ⎛ ∂ ln c ⎞
= ⎜ ⎟ (3-126)
misunderstandings and inconsistencies. ˜
D k T ⎝ ∂m ⎠
The external force Fd is a result of a po-
where m is the chemical potential of the
tential so that
particles and c is the site fraction.
df Now let us discuss this equation in detail
Fd = − (3-119)
dx by exploring some particular cases. When
At equilibrium the distribution of com- the distribution of particles is completely
pletely noninteracting particles follows a ideal, meaning that the particles do not feel
Boltzmann distribution such that one another, not even site blocking, the
thermodynamic factor drops out of Eq. (3-
C (x) = C0 exp [– f (x)/k T ] (3-120) 126) and s /D̃ reduces to Eq. (3-125). In this
very special case, and only in this case, the
Eq. (3-120) must be the solution of Eq. (3-
tracer diffusion coefficient D* equals D̃, so
120) at steady state. We then have that
that
dC C df C Fd
=− = (3-121) s C Z 2 e2
dx k T dC k T = (3-127)
D* kT
With Eq. (3-118) we find that
When the particles are ideally distributed
〈v 〉 Fd but subject to the condition that no more
= (3-122)
D̃ k T than one particle can occupy one site, then
212 3 Diffusion Kinetics in Solids
the thermodynamic factor in Eq. (3-126) i.e., formally the same as Eq. (3-129). This
equals (1 – c)–1. However, for this situation route to Eq. (3-132) is clearly a case of two
the tracer diffusion coefficient is related to wrongs ending up making a right. How-
D̃ by (see, for example, Murch (1982 c)) ever, the real danger lies in the fact that Eq.
D* = D̃ (1 – c) f (3-128) (3-132) obtained in this way blinds us to its
limitations, limitations which are clearly
where f is the tracer correlation factor, so stated in the derivation of Eq. (3-129).
that In practice, Eq. (3-132) has often been
s C Z 2 e2 used to describe situations where the par-
= (3-129) ticles are interacting and many sites are
D* kT f
vacant, such as in fast ion conductors. In
Ionic solids having virtually a perfect lat- such cases, we cannot necessarily expect a
tice of particles (charge carriers) fall into meaningful interpretation. What is the cor-
this category and Eq. (3-129) is appropriate rect way to proceed? We already have the
to such solids. Similarly, when the particles Nernst – Einstein equation to cover the situ-
are extremely dilute, Eq. (3-129) is again ation of interacting particles (Eq. (3-126)).
appropriate. However, this normally cannot be applied
In the solid-state diffusion literature we to real materials because local charge neu-
very often see Eq. (3-125) used directly trality prevents composition gradients be-
to calculate another diffusion coefficient, ing set up in ionic conductors by the con-
sometimes called the “charge” diffusion co- ducting ions themselves. What can be done
efficient and given the symbol Ds . Thus we is to return once again to Eq. (3-130) and use
encounter the following equation, which is it purely as a definition of Ds , recognizing
also called the Nernst – Einstein equation at the same time that Ds has no Fickian
s C Z 2 e2 meaning. Eq. (3-129) is then being used
= (3-130) purely as a means of changing s to a quan-
Ds kT
tity which has the dimensions of a diffusion
Ds is dimensionally correct but it does not coefficient. It is clear that it would be inap-
correspond to any diffusion coefficient that propriate in these circumstances to call this
can actually be measured by way of Fick’s equation the Nernst – Einstein equation.
laws for a solid system. Recall that the We can now define the Haven ratio,
identical Eq. (3-125) requires that the parti- which is simply the ratio of D* to Ds :
cles are completely noninteracting for it to
be meaningful. We frequently then see the D*
HR ≡ (3-133)
following equation relating D* and Ds Ds
D* = f Ds (3-131) In view of what has been said above about
with Ds often being called a self-diffusion Ds , it is appropriate to ask whether HR has
coefficient or the diffusion coefficient of any physical meaning. This can be partially
the (random walking) charge carriers. Eq. answered by examining hopping models
(3-131) is generally given as if it is self-ev- for conduction. It can be shown quite gen-
ident or similar to Eq. (3-73). Eqs. (3-130) erally for hopping models using the va-
and (3-131) are then combined to give cancy mechanism that
s C Z 2 e2 f f
= (3-132) HR = = (3-134)
D* kT f fI f + c g
3.3 Microscopic Diffusion 213
where f I is the physical or conductivity cor- Much less satisfactorily interpreted are
relation factor and g is a two-particle corre- fast ion conductors where the defect con-
lation factor (see Sec. 3.3.1.7). For specific centration is high and f I must be included
models of interacting particles f I ≤ 1 (see in the analysis of HR . A review of HR for
Murch and Dyre (1989) for a detailed re- such materials has been provided by Murch
view). At the limits of an almost full or (1982 d). Further comments on the subject
empty lattice of charge carriers, f I Æ 1 and can be found in the review on correlation
HR = f. This is compatible, of course, with effects in ionic conductivity by Murch and
Eq. (3-131) for these conditions. Dyre (1989). An introduction to the subject
Accordingly, a measurement of the may be found in the book by Philibert
Haven ratio (obtained by measuring s and (1991).
D*, preferably in the same sample) can in
some cases give f alone and therefore the 3.3.3 The Isotope Effect in Diffusion
mechanism of diffusion can be exposed.
The isotope effect, sometimes called the
However, the interstitialcy mechanism adds
mass effect, is of considerable importance
a minor complication because the “charge”
in diffusion because it provides one of the
moves two jump distances whereas the
two experimental means of gaining access
tracer moves only one (see Fig. 3-6). This
to the tracer correlation factor, f (the other
is easily taken care of in the analysis so that
is the Haven ratio; see Sec. 3.3.2). Since f
HR is now written generally as (see for ex-
depends on mechanism (see Sec. 3.3.1.3), a
ample, Murch (1982 d))
2
measurement of the isotope effect can, in
f ⎛r⎞ principle, throw light on the diffusion
HR = ⎜ ⎟ (3-135)
fI ⎝ rq ⎠ mechanism operating.
The measurement of the isotope effect
where r and rq are the distances moved by
depends on accurately measuring a small
the tracer and charge respectively. Pro-
difference between the diffusion coeffi-
vided we focus on cases where f I = 1 (either
cients of two tracers a and b (Peterson,
almost empty or full lattice of charge car-
1975). This small difference can be related
riers), then HR still has a unique value for
to f in the following way. For isotropic
each mechanism. The classic example of
tracer and impurity diffusion, the diffusion
the value of HR in identifying the mecha-
coefficients of a and b can be written as
nism is Ag motion in AgBr (Friauf, 1957).
It was found that HR varied from 0.5 at low Da = A wa fa (3-136)
temperature to 0.65 at high temperature. Db = A wb fb (3-137)
Frenkel defects (see Sec. 3.4.4) are ex-
pected and therefore a temperature depen- where A contains geometrical terms and
dence of HR . But HR is not consistent with defect concentrations which do not depend
a combination of vacancy and direct inter- on the atom/defect exchange frequency w.
stitial jumps. This would require that HR The correlation factors fa and fb have the
varies from 0.78 to unity. The lower values mathematical form of impurity correlation
of HR obtained require a contribution from factors because the tracers (i.e. isotopes) in
interstitialcy jumps. It was found that both real experiments are formally impurities in
collinear and noncollinear interstitialcy the sense that their jump frequencies differ
jumps were required to fit the behavior of from the host atoms. Most impurity corre-
HR . lation factors for impurity diffusion in oth-
214 3 Diffusion Kinetics in Solids
erwise pure and nearly perfect crystals From Eqs. (3-140) and (3-141), we find
have the mathematical form that
fa = u /(wa + u) (3-138) ( Da − Db )/ Db
f DK = (3-142)
( mb / ma )1/ 2 − 1
fb = u /(wb + u) (3-139)
Eq. (3-142) refers to a process where only
where u contains exchange frequencies
one atom jumps. The more general case of
other than the tracer. After taking the ratio
n atoms jumping simultaneously, e.g., the
of Eqs. (3-136) and (3-137) and eliminat-
interstitialcy mechanism (where n = 2) (see
ing u and fb (say) from Eqs. (3-136), (3-
Fig. 3-6), is described by
138), and (3-139), we find that
(Da – Db ) /Db = f [(wa – wb ) /wb ] (3-140) f DK (3-143)
[( Da − Db )/ Db ]
At this point, the formal distinction be- =
[( mb + ( n – 1) m0 ) /( ma + ( n – 1) m0 )]1 / 2 – 1
tween fa , fb and the tracer correlation fac-
tor f, the latter referring to a hypothetical where m0 is the average mass of the non-
tracer with the same jump frequency as the tracers. It is common to refer to f DK in
host, can be dropped since they are very Eqs. (3-142), (3-143) as the “isotope ef-
similar in numerical value. Knowledge of fect” E (n). Because f and DK are ≤ 1 then
the experimental jump frequencies, wa and this also applies to E (n). Accordingly, al-
wb , is usually not available, of course. Ac- though a measurement of the isotope effect
cordingly, they are expressed in terms of may not uniquely determine the mecha-
quantities which are known, such as the nism, it can be invaluable in showing
masses of the tracers. which mechanisms (through their values of
Application of classical statistical me- f and n) are not consistent.
chanics to describe the dynamics of the There are some mechanisms for which
jumping process leads to the result the appropriate correlation factor does not
have the simple form of Eq. (3-136), for
⎡⎛ mb ⎞ 1/ 2 ⎤ example a mechanism which has several
(wa − w b )/w b = DK ⎢⎜ ⎟ − 1⎥ (3-141) jump frequencies. Nonetheless, it is always
⎢⎣⎝ ma ⎠ ⎥⎦
possible to derive equations equivalent
where ma and mb are the masses of the to Eqs. (3-142), (3-143) for such cases (see
tracers a and b, DK is the fraction of the the review by Le Claire (1970) for details).
total kinetic energy, associated with the In the discussion above, there has per-
whole motion in the jump direction at the haps been the implication that the theory
saddle point of a jump, that actually be- described here applies only to tracer diffu-
longs to the diffusing atom. The conversion sion in “simple” materials, e.g., pure met-
of jump frequencies to masses has unfortu- als, and alkali metal halides. Certainly the
nately led to the introduction of this new theory was originally developed with these
quantity DK, about which detailed informa- cases in mind, but it can be used, with cau-
tion is difficult to obtain. If the remainder tion, in other more complex situations. One
of the lattice is not involved in the jump, of the earliest of these was application to
DK = 1 (this is the upper limit for DK). impurity diffusion. In this case the tracers
More usually, there is a certain amount of a and b now refer to isotopes of an impur-
coupling between the diffusing atom and ity which is chemically different from the
the remainder of the lattice and DK < 1. host. When the mechanism is not in any
3.3 Microscopic Diffusion 215
With a little manipulation it is soon found “wait” until a defect arrives. The probabil-
that ity of a defect, say a vacancy, arriving at a
particular neighboring site to a given atom
⎛C ⎞ is simply the fractional vacancy concentra-
ln ⎜ a ⎟ = const. − ln Ca [( Da − Db )/ Db ]
⎝ Cb ⎠ tion cv . However, in certain situations such
(3-144) as alloys showing order, or fast ion conduc-
This means that by plotting ln (Ca /Cb ) vs. tors showing relatively high defect order-
ln Ca , (Da – Db )/Db can in fact be obtained ing, the availability of vacancies to the
very accurately, avoiding errors due to time atoms can be enhanced or depressed from
and temperature of the diffusion anneal and the random mixing value which cv signi-
errors in sectioning. The separation of the fies. This topic is discussed further in Secs.
isotopes is normally achieved by half-life, 3.4.2 and 3.4.3.
energy spectroscopy, or the use of different The probability of a given atom being
kinds of radiation. Details of how to carry next to a vacancy is thus g cv where g is the
out careful experiments in this area can be coordination number. The jump frequency
found in the very comprehensive review of G can we decomposed to
experimental techniques in diffusion by
G = g cv w (3-145)
Rothman (1984). A more detailed review
of the isotope effect in diffusion has been where w is usually called the “exchange”
provided by Peterson (1975); see also Le frequency to signify the exchange between
Claire (1970). an atom and a neighboring vacant site and
to distinguish it from the actual jump fre-
quency G . It is permissible at low defect
3.3.4 The Jump Frequency
concentrations to call w the defect jump
In Sec. 3.3.1.2 we showed that a tracer frequency, since the defect does not have to
diffusion coefficient is, in essence, a prod- “wait” for an atom.
uct of an uncorrelated part containing In, say, f.c.c. crystals where the jump
–
the jump frequency and a correlated part distance r is given by a /÷2, where a is the
containing the correlation factor, see Eq. lattice parameter and g = 12, the tracer dif-
(3-76). The Arrhenius temperature depen- fusion coefficient from Eq. (3-76) can now
dence of the diffusion coefficient arises be written as
largely from the jump frequency, although
D* = 12 cv w (a 2/2) f /6 = cv w a 2 f (3-146)
some temperature dependence of the corre-
lation factor in some situations, e.g., al- The same equation is also valid for b.c.c.
loys, cannot be ignored (see Secs. 3.3.1.4 crystals. Other examples have been given
to 3.3.1.6). In this section we will study the by Le Claire (1975).
make-up of the atomic jump frequency. In many solids, such as metals, alloys
First, for an atom to jump from one site and ionic crystals, cv is said to be “intrin-
to another, the defect necessary to provide sic” with an Arrhenius temperature depen-
the means for the jump must be available. dence. The vacancy formation enthalpy is
For the interstitial diffusion mechanism contained in the measured activation en-
there is essentially always a vacancy thalpy for the tracer diffusion process (see
(really a vacant interstice) available except Eq. (3-149)). In some solids, notably ionic
at high interstitial concentrations. For other crystals and certain intermetallic com-
mechanisms, however, the atom has to pounds, apart from the inevitable tempera-
3.3 Microscopic Diffusion 217
ture-dependent intrinsic concentration, the In Sec. 3.3.4.2 we show that the defect
vacancy concentration can also be manipu- concentration in many cases depends on
lated by doping or by changing the degree temperature in an Arrhenius fashion (Eq.
of nonstoichiometry. In such cases this va- (3-159)). In such cases, the atomic jump
cancy concentration is sometimes said to frequency G, being the product of w , the
be “extrinsic”, and this contribution to the defect concentration, and the coordination
total vacancy concentration can easily number, is written as from Eq. (3-145)
swamp the intrinsic vacancy contribution. G = g n– exp [(S m + S f )/k]
v
The extrinsic vacancy concentration is in-
¥ exp [– (H m + Hvf )/k T ] (3-149)
dependent of temperature. However, in the
m f
case of a change in stoichiometry, tempera- The sum of H and H is more accurately
ture independence of cv requires adjust- called the activation enthalpy. We empha-
ment of the external partial pressure (see size again that the activation enthalpy
Sec. 3.4.4). measured in a diffusion experiment can, for
some solids such as ordered alloys, contain
other contributions such as from correla-
3.3.4.1 The Exchange Frequency
tion effects or defect availability terms.
With the defect immediately available, The enthalpy of migration and the en-
the atom can jump when it acquires suffi- thalpy of defect formation are now rou-
cient thermal energy from the lattice for it tinely calculated by Lattice or Molecular
to pass over the energy barrier between its Statics methods; see Mishin (1997) and
present site and the neighboring site. The references therein for examples of such
probability of the atom having this thermal calculations in grain boundaries.
energy is given by the Boltzmann probabil- Although it is straightforward to couch
ity exp (– G m/k T ), where G m (a Gibbs en- the argument above concerning w in statis-
ergy) is the barrier height, k is the Boltz- tical mechanical terms, this would still be
mann constant, and T is the absolute tem- inadequate because we have neglected the
perature. The attempt frequency, i.e., the participation of the remainder of the lattice,
number of times per second the atom on its especially the neighbors of the jumping
site is moving in the direction of the neigh- atom, in the jump process. A more satisfac-
boring site, is the mean vibrational fre- tory treatment has been provided by Vine-
quency n–. Accordingly, the “jump rate”, yard (1957). This puts n– on a sounder basis
i.e., number of jumps per second, w , is and gives physical significance to the en-
given by tropy of migration S m.
Vineyard (1957) took a classical statisti-
w = n– exp (– G m/k T ) (3-147)
cal mechanical approach by considering
Because G m is a Gibbs energy, the ex- the phase space of the system comprising
change frequency can be partitioned as all of its N atoms and one vacancy. Each
point in this phase space represents an
w = n– exp (S m/k) exp (– H m/k T ) (3-148)
atomic configuration of the system. All
where S m is the entropy of migration and possible configurations are represented.
H m is the enthalpy of migration. H m is We are concerned with two neighboring
sometimes loosely called the “activation” energy minima in this phase space. These
enthalpy, but this can lead to misunder- minima are centered on two points, P and
standings, as we shall now see. Q. Point P is the phase space point repre-
218 3 Diffusion Kinetics in Solids
senting an atom adjacent to a vacancy; all where ni0 is the frequency of the ith normal
atoms are in their equilibrium positions. mode with the system constrained to move
Point Q is the corresponding point after the normal to the direction joining point P to
atom – vacancy exchange. Between the the saddle point.
points P and Q there is a potential energy There have been many further considera-
barrier. In order to jump, the atom must tions of the detailed dynamics of the jump
pass over this barrier at its lowest point, a process, the most notable of these being the
saddle point. work carried out by Jacucci, Flynn and co-
Vineyard (1957) was able to calculate workers (see reviews by Jacucci (1984)
the jump rate, i.e., the number of crossings and Pontikis (1990)).
per unit time at the “harmonic approxima- For general purposes it is often sufficient
tion”. This entailed expanding the potential to focus on Eq. (3-148) and to note that the
energy at point P and also at the saddle Debye frequency is an adequate represen-
point in a Taylor series to second order to tation of the mean vibration frequency n–.
give for w
⎛ N ⎞ 3.3.4.2 Vacancy Concentration
⎜∏ ni ⎟
i =1
w = ⎜ N ⎟ exp ( − U / k T ) (3-150) The change in Helmholtz energy FVf as-
⎜ n ′⎟ sociated with the formation of one vacancy
⎜∏ i⎟
⎝ i =1 ⎠ is given by
where the ni are normal frequencies for vi- FVf = EVf – T SVf (3-154)
bration of the system at point P, ni¢ are the
where EVf and SVf are the energy and en-
normal frequencies at the saddle point and
tropy of formation, respectively. The for-
U is the difference in potential energy
mation process itself is conceived to be the
between the saddle point and the equilib-
removal of an atom from the interior of the
rium configuration point, P.
crystal to the surface. The entropy part has
It is possible to transform Eq. (3-150)
two contributions: a vibrational or thermal
into the form of Eq. (3-148) by writing f
part S vib arising from the fact that atoms
N
close to the vacancy have a different vibra-
∏ ni tional frequency from those far from the
i =1
N −1
= n exp ( S m / k T ) (3-151) f
vacancy, and a configurational part S config
∏ n i′ which is usually thought to be an ideal mix-
i =1
ing entropy, at least for pure metals – but it
where the entropy of migration can be is rather more complicated for alloys. The
identified with configurational part for pure metals is eas-
3N ily found from elementary classical statisti-
Sm = k ∑ ln (n i0 /n i′ ) (3-152) cal mechanics (see, for example, Peterson
i =2 (1978)). The number of different ways W
and of putting n vacancies and N atoms on N + n
3N sites assuming indistinguishability within
∏ ni each group is
i =1
n= (3-153)
3N ( N + n )!
∏ n i0 W=
N! n!
(3-155)
i=2
3.4 Diffusion in Materials 219
The configurational entropy is given by ation techniques with computer codes such
as DEVIL and CASCADE (these tech-
f ⎡ ( N + n )! ⎤
Sconfig = k ln W = k ln ⎢ ⎥ (3-156) niques are described in the book edited by
⎣ N! n! ⎦ Catlow and Mackrodt, 1982) and lattice
which, with Stirling’s approximation ln N ! dynamics and Monte Carlo methods (see,
= N ln N – N for large N, results in for example, Jacucci, 1984).
Vacancy concentrations in metals are con-
⎛ n ⎤⎞
= − k ⎜ N ln ⎡⎢
N ⎤
f
Sconfig ⎥ + n ln ⎡⎢ ⎟ ventionally measured by quenching, dila-
⎝ ⎣N + n⎦ ⎣ N + n ⎥⎦⎠ tometry, and positron annihilation. An intro-
(3-157) duction to these techniques has been pro-
The total change in free energy when n va- vided by Borg and Dienes (1988), and a de-
cancies are produced is tailed review is also available (Siegel, 1978).
In Eq. (3-166) D2 (0) arises from isolated w 0 , the solvent exchange frequency far
solute atoms, while the term containing B1 from the solute. For the f.c.c. lattice the re-
arises from pairs of solute atoms which are sult for b1 is (Howard and Manning, 1967)
sufficiently close that the solute jump fre-
w 4 ⎡ c 1 w1 c ⎤
quency differs from the isolated solute – va- b1 = − 18 + ⎢ 4 + 14 2 ⎥ (3-168)
cancy exchange frequency w 2 . B2 arises w 0 ⎣ f0 w 3 f0 ⎦
from triplets of solute atoms.
where c 1 and c 2 are termed partial correla-
The solute enhancement factor B1 has
tion factors and are known functions of the
been calculated for the f.c.c. lattice by ex-
ratios w 2 /w 1 , w 1 /w 3 and w 4 /w 0 , and f0 is
tending the five-frequency model (see Sec.
the correlation factor in the pure lattice.
3.3.1.4) to include three new frequencies.
Eq. (3-168) can be re-expressed as
These frequencies describe solute jumps
which create, i.e., associate, a new solute ⎡ c w c w ⎤
pair (w 23 ), dissociate a solute pair (w 24 ), b1 = − 18 + ⎢ 4 1 1 + 14 2 3 ⎥
⎣ f w0 f0 w 0 ⎦
and reorient an existing pair (w 21). Pairs of
solute atoms do not occur randomly when × exp( − E2B / k T ) (3-169)
solute atoms interact. When an interaction thereby showing the effect of an altered va-
energy E 22 between solute atoms is defined cancy concentration near a solute by virtue
it is straightforward to show that of the solute – vacancy binding energy E 2B .
w 24 exp (– E 22 /k T ) = w 23 exp (– E 2B /k T ) There are similar relations for the b.c.c.
lattice. For the impurity model described
where E 2B is the impurity – vacancy bind- in Sec. 3.3.1.4 (Model I), b1 is given by
ing energy. If it is assumed that the impur- (Jones and Le Claire, 1972)
ity correlation factor does not depend on
u1
solute concentration (Stark, 1972, 1974), b1 = − 20 + 14
then B1 is given by (Bocquet, 1972; Le f
w3 n 1
Claire, 1978) +6 exp( − E2B / k T ) (3-170)
w0 f
B1 = − [6 + 12 exp( − E22 / k T )] (3-167)
− exp( − E2B / k T ) and for Model II (see Sec. 3.3.1.4)
⎡ w w ⎤ u2
× ⎢ 4 21 exp( − E22 / k T ) − 14 23 ⎥ b1 = − 20 + 6
⎣ w2 w2 ⎦ f
w n
If it is assumed that E 22 = 0 and that + 14 3 2 exp( − E2B / k T ) (3-171)
w 21 = w 23 = w 2 then B1 is reduced to w0 f
where u 1 , n1 (w 2 /w 3¢ , w 3 /w 3¢ ) and u 2 , n 2
B1 = 18 [exp (– E 2B /k T ) – 1] (3-167 a)
(w 2 /w 3 , w 4 /w 0 ) are called mean partial
There have been no calculations along sim- correlation factors and are known functions
ilar lines of B2 . of the frequencies given.
The solvent enhancement factor b1 can It is possible to express b1 in terms of the
be calculated by noting the number of jump ratio of the solute and solvent diffusion co-
frequencies close to the solute. For in- efficients. This aids in the discussion of the
stance, in the five-frequency model for the numerical values taken by the enhance-
f.c.c. lattice we need to count the number ment factors. For the f.c.c. lattice, b1 is
of solvent frequencies which differ from given by Le Claire (1978).
3.4 Diffusion in Materials 223
Table 3-2. Some values of solute and solvent enhancement factors, from Le Claire (1978).
mole fraction of vacancies in the ith com- enhanced in a nonlinear way (Hehenkamp
plex cvi et al., 1980; Hehenkamp and Faupel, 1983).
Moreover, independent measurements of
⎛ g⎞
cvi = cv ( 0 ) ⎜ ⎟ c0g − i c2i exp( GBi / k T ) (3-173) the vacancy concentration on alloying have
⎝ 0⎠
shown a linear dependence between nor-
where cv (0) is the mole fraction of vacan- malized vacancy concentration and the
cies in the pure metal (A) and c0 and c2 are normalized solvent diffusion coefficient
the mole fractions of solvent (A) and solute (Hehenkamp et al., 1980; Hehenkamp and
(B) respectively, and g is the coordination. Faupel, 1983). The nonlinear enhancement
Bérces and Kovács (1983) modified Eq. (3- of solvent diffusion upon alloying is thus
173) to take into account possible configu- apparently due to a corresponding nonlin-
rations of the complexes. ear increase in the vacancy concentration
Within this complex model framework on alloying. These findings are very nicely
Faupel and Hehenkamp (1986, 1987) have described by the complex model.
introduced average effective jump fre- Despite its obvious limitations, the com-
A plex model can be useful quantitatively up
quencies per solvent (A) atom ·w eff Òi and
average effective jump frequencies per to 5 at.% of solute and qualitatively even
B higher; see Le Claire (1992).
solute (B) atom ·w eff Òi in the ith complex
(the latter specifically includes correlation
effects – which are always important for 3.4.2.2 Interstitial Alloys
the solute). The normalized solvent diffu-
sion coefficient can now be written as Elements such as H, N, O, and C dis-
solve interstitially in metals and diffuse by
g −1
D0 ( c2 ) 〈w eff
A
〉 i ⎛ g − 1⎞ the interstitial mechanism. The diffusion
= c0g −1 − ∑ ⎜ ⎟
D0 ( 0 ) i =1 f0 w 0 ⎝ i ⎠ coefficients here are often measured by re-
laxation techniques or outgassing. Tracer
× c0g −1 − i c2i exp( GBi / k T ) (3-174)
diffusion is difficult to measure in the case
Similarly, the normalized solute diffusion of N and O because of the lack of suitable
coefficient can be written as radioisotope; however, other techniques
are available (see Sec. 3.5.1).
D0 ( c2 ) g
i 〈w eff
B
〉i ⎛ g⎞
= c0g −1 − ∑ ⎜ ⎟ (3-175) For interstitials at infinite dilution, the
D0 ( 0 ) i = 2 g f2 w 2 ⎝ i ⎠ interstitials move independently (by the
× c0g −1 c2i −1 exp[( GBi − GB1 )/ k T ] interstitial mechanism, see Fig. 3-5). The
vacancy concentration (vacant interstice) is
These equations for the normalized solute unity, the correlation factor is unity (a com-
and solvent diffusion coefficients can be plete random walk), and the tracer diffu-
expanded in terms of c2 to give explicit ex- sion coefficient is given simply by (from
pressions for the various solute and solvent Eqs. (3-76) and (3-145))
enhancement factors by way of comparison
D* = g w 0 r 2/6 (3-176)
with Eqs. (3-165) and (3-166); see Le
Claire (1992). where w 0 is the interstitial/vacant interstice
It has been found, in a number of careful “exchange” frequency.
studies in f.c.c. systems (Ag – Sb, Ag – Sn) In a number of cases the interstitial con-
where the impurities have excess positive centration can become sufficiently high for
charge, that solvent diffusion on alloying is the interstitials to interfere with one an-
3.4 Diffusion in Materials 225
where V is the vacancy availability factor some loss of realism. The first of these is
(the average availability of neighboring va- the random alloy model introduced by
cant interstices to an interstitial solute Manning (1968, 1970, 1971). The second
atom, and is a more general quantity than is the interacting bond model which has
cv ), and W represents the effect of the envi- been extensively developed by Kikuchi
ronment on the jump frequency. It is in ef- and Sato (1969, 1970, 1972).
fect the statistical average of the first term In the random alloy model, the atomic
in Eq. (3-179). The same lattice gas model components are assumed to be mixed ran-
has also been applied extensively to fast domly and the vacancy mechanism is as-
ion conductors (see, for example, Murch sumed. The atomic jump frequencies wA
(1984 a)). and wB for the two components A and B
Chemical diffusion can readily be ex- are specified and these do not change with
pressed along similar lines (Murch, 1982 c) composition or environment. The tracer
as diffusion coefficient of, say, A is given by
D̃ = g r 2 V W n f I (3-181) * = g wA r 2 cv fA /6
DA (3-182)
¥ exp (– U/k T ) (∂m /k T/∂ ln c)/6 where g is the coordination, r is the jump
distance, and cv is the vacancy site fraction
where m is the chemical potential of the
(see Sec. 3.3.4.2). The only theoretical re-
solute and f I is the physical correlation fac-
quirement here is to calculate the tracer
tor (see Sec. 3.3.1.7). The physics for f I
correlation factor fA ; this has been de-
are not contained in the first approach (Eq.
scribed in Sec. 3.3.1.5 on correlation ef-
(3-178)).
fects in diffusion. The random alloy model
These approaches seem to describe C
performs very well in describing the diffu-
tracer and chemical diffusion in g -Fe fairly
sion behavior of many alloy systems which
well (McKee, 1980 a, b, 1981; Murch and
are fairly well disordered. This subject is
Thorn, 1979) but other applications have
discussed extensively in the review by
not been made because of the lack of suit-
Bakker (1984) to which we refer. A much
able data, especially for D*.
more accurate method for dealing with fA
Quantum effects are an important ingre-
has been given by Moleko et al. (1989).
dient in the description of H diffusion in
For alloys which exhibit order, Man-
metals; see the reviews by Völkl and Ale-
ning’s theory, based on the random alloy
feld (1975), Fukai and Sugimoto (1985),
model, still performs reasonably well in
and Hempelmann (1984).
describing certain aspects of diffusion be-
havior, notably correlation effects which
3.4.3 Diffusion in Concentrated Binary can be expressed in terms of the tracer dif-
Substitional Alloys fusion coefficients, see Sec. 3.3.1.8 and the
review by Murch and Belova (1998). A
Dilute alloy models, see Sec. 3.4.2.1, wholly different approach pioneered by
cannot be extended very far into the con- Kikuchi and Sato (1969) is the PPM (see
centrated regime without rapidly increas- Sec. 3.3.1.5) in which the tracer diffusion
ing the number of jump frequencies to an coefficient is expressed in terms of quan-
unworkable level. As a result, models have tities which are statistically averaged over
been introduced which limit the number of the atomic configurations encountered in
jump frequencies but only as a result of the alloy. The tracer diffusion coefficient
3.4 Diffusion in Materials 227
D* = g w – r 2 p– –f /6 (3-183)
with temperature. We also have contribu-
A A Av A
tions from p–Av and the tracer correlation
–
where wA is an averaged exchange fre- factor. In order to make sense of the experi-
quency which includes the effect of the en- mental activation energies, we need to
vironment on the jump frequency, p–Av is the model the system in question; there seems
vacancy availability factor, i.e., the prob- to be no alternative. Few detailed applica-
ability of finding a vacancy next to an A tions of this model have yet been made.
–
atom (see below), and f A is the tracer corre- One detailed application has been made to
lation factor. We have discussed the corre- b -CuZn (Belova and Murch, 1998). Fur-
–
lation factor f A in Sec. 3.3.1.5, to which we ther details can be found in the review by
refer. The quantity w – is the average of the
A Bakker (1984), see also the review by
exchange frequency for an A atom with a Murch and Belova (1998).
vacancy in particular configuration: Interdiffusion in binary concentrated al-
wA = nA exp (– UA /k T ) (3-184) loys has been dealt with largely in Secs.
3.2.3.2 and 3.3.1.8. Some recent calcula-
¥ exp [(gA EAA + gB EAB )/k T]
tions (which use the “bond model” de-
where nA is the vibration frequency, UA is scribed above) have been performed by
the migration energy (referred to some ref- Zhang et al. (1988) and Wang and Akbar
erence) for a jump, gA is the number of A (1993), see also the review by Murch and
atoms that are nearest neighbors to a given Belova (1998).
A atom (itself next to a vacancy), gB is the
number of B atoms which are likewise
3.4.4 Diffusion in Ionic Crystals
nearest neighbors to the A atom, and EAA
and EAB are nearest neighbor interactions In this section diffusion in ionic crystals
(assumed negative here). This equation is (exemplified here by oxides) will be briefly
essentially the binary analogue of Eq. (3- discussed. A more detailed discussion has
179). We see that the neighbors of the atom been provided by Kofstad (1972) and
A can increase or decrease the apparent mi- Schmalzried (1995). Diffusion in other
gration energy. ionic crystals, especially alkali metal ha-
The use of cv alone in the expression for lides and silver halides, has been reviewed
*
DA in Eq. (3-182) implies that either com- by, for example, Fredericks (1975), Laskar
ponent “sees” the vacancy equally. This (1990, 1992), Monty (1992).
obviously cannot generally be true. The in-
troduction of the quantity p–Av in Eq. (3-
3.4.4.1 Defects in Ionic Crystals
183) (and p–Bv ) is a recognition that vacan-
cies are somewhat apportioned between the In order to discuss diffusion in ionic
two atomic components. This is not a small crystals, we need first to discuss at an ele-
effect, particularly in alloys which show mentary level the types of defects which
separate sublattices. arise. Of almost overriding concern in
The overall activation energy for diffu- ionic crystals is the requirement of charge
sion is now rather complicated. Apart from neutrality.
the usual contribution from the vacancy Let us first consider stoichiometric crys-
formation energy and reference migration tals, say the oxide MO, and “intrinsic” de-
energy, we have a complicated contribution fect production. The Schottky defect (actu-
228 3 Diffusion Kinetics in Solids
ally a pair of defects) consists of a vacant where MM is a metal atom (M) on a metal
anion site and a vacant cation site. It arises site (M), M ••
i is an effectively doubly posi-
as a result of thermal activation and not tively charged metal ion on interstitial i site
through interaction with the atmosphere. and V≤ M is an effectively doubly negatively
The Schottky defect generation is written charged metal ion vacancy. We have as-
as a chemical reaction, i.e., sumed double charges here purely for illus-
0 [ V≤ •• trative purposes.
M + VO (3-185)
The equilibrium constant for the Frenkel
where 0 refers to a perfect crystal, V≤ M is a defect reaction can be written as
vacant (V) metal (M) site, the primes refer
to effective negative charges (with respect KF = [M •• M]
i ] [V≤ (3-189)
to the perfect crystal), VO•• is a vacant (V)
provided that the defect concentration is
oxgen (O) site and the dots refer to effec-
low. This equation is often called the Fren-
tive positive charges. This is the Kröger-
kel product. Again, the equilibrium con-
Vink defect notation (see, for example,
stant KF can be expressed as exp (– GFf /k T )
Kofstad (1972)).
where GFf is the Gibbs energy of formation
We can write an equilibrium constant Ks
of the Frenkel defect, which again can be
for the reaction in the following form, valid
partitioned into its enthalpy H Ff and en-
for low defect concentrations:
tropy SFf parts.
Ks = [V≤ ••
M ] [VO ] (3-186)
where the brackets [ ] indicate concentra-
tions. Eq. (3-186) is usually called the 3.4.4.2 Diffusion Theory in Ionic Crystals
Schottky product. Ks can be expressed in
In most mechanisms of diffusion except
the usual way
the interstitial mechanism, an atom must
Ks = exp (– Gsf /k T ) (3-187) “wait” for a defect to arrive at a nearest
where Gsf is the Gibbs energy of formation neighbor site before a jump is possible (see
of the Schottky defect, which can be parti- Sec. 3.3.4). Thus the jump frequency in-
tioned into its enthalpy Hsf and entropy Ssf cludes a defect concentration term, e.g., cv ,
parts. the vacancy concentration. Let us examine
The other type of defect occurring in the an example for diffusion involving the
stoichiometric ionic crystal is the Frenkel Frenkel defect. Although both an intersti-
defect. The Frenkel defect (actually a pair tial and a vacancy are formed, in oxides
of defects) consists of a cation interstitial one of them is likely to be much more mo-
and cation vacancy or an anion interstitial bile, i.e., to have a lower migration energy,
and anion vacancy. In the latter case it has than the other. (In the case of stoichiomet-
also been called an anti-Frenkel defect, al- ric UO2 , for example, theoretical calcula-
though this nomenclature is now relatively tion of migration energies suggests a much
uncommon. Like the Schottky defect, the lower migration energy for the oxygen va-
Frenkel defect is thermally activated. The cancy than for the interstitial (by either
Frenkel defect generation is also written as interstitial or interstitialcy mechanisms)
a chemical reaction, for example for cat- (Catlow, 1977).) At the stoichiometric
ionic disorder: composition,
MM [ M •• i + V≤ M (3-188) [M •• M]
i ] = [V≤ (3-190)
3.4 Diffusion in Materials 229
so that the cation vacancy concentration is Mn1– y O. As an example we will deal with
given by an anion excess accommodated by metal
f vacancies.
M ] = cv = exp (– G F / 2 k T )
[V≤ (3-191)
The degree of nonstoichiometry and
= exp (– H Ff /2 k T ) exp (SFf /2k) (3-192) therefore the defect concentration accom-
panying it are functions of temperature and
The measured activation enthalpy for dif- partial pressure of the components. The de-
fusion will be the migration enthalpy plus fects produced in this way are sometimes
half the Frenkel defect formation enthalpy. said to be “extrinsic”, but this terminology
The reader might well ask how we know is probably to be discouraged since they
we are dealing with a Frenkel defect and are still strictly intrinsic to the material. In
not a Schottky defect, and if we do know it an oxide it is usual, at the temperatures of
is the Frenkel defect then how do we know interest, to consider only the partial pres-
it is the vacancy mechanism that is operat- sure of oxygen since it is by far the more
ing and not the interstitialcy mechanism? volatile of two components. In carbides,
In general, we have to rely on independent where the temperatures of diffusion of
information, principally computer calcula- interest are much higher, the metal partial
tions of defect formation and migration en- pressure can be comparable to the carbon
thalpies, but structural information such as partial pressure and either can be manipu-
is provided by neutron diffraction and ther- lated externally.
modynamic information is also useful. The simplest chemical reaction generat-
Another process of interest here is the in- ing nonstoichiometry is
trinsic “ionization” process whereby an
electron is promoted from the valence band –12 O2 (g) [ OO
¥ ¥
+ VM (3-194)
to the conduction band leaving behind a where OO ¥
is a neutral (¥) oxygen ion on an
hole in the valence band. In the Kröger- oxygen site and VM ¥
is a neutral metal va-
Vink notation, we write for the intrinsic cancy. Physically, this corresponds to oxy-
ionization process (the electrons and holes gen being adsorbed on the surface to form
may be localized on the metal atoms) ions and more lattice sites, thereby effec-
0 [ e¢ + h• (3-193) tively making metal vacancies. These va-
cancies diffuse in by cation – vacancy ex-
All ionic crystals are capable, in princi- change until the entire crystal is in equilib-
ple, of becoming nonstoichiometric. The rium with the atmosphere. Excess oxygen
limit of nonstoichiometry is largely dic- nonstoichiometry corresponds also to oxi-
tated by the ease with which the metal ion dation of the metal ions. Here in this exam-
can change its valence and the ability of the ple, the holes produced are conceived to be
structure to “absorb” defects without re- associated very closely with the metal va-
verting to some other structure and thereby cancies. The holes can be liberated, in
changing phase. Nonstoichiometry can be which case the reaction can now be written
achieved by either 1) an anion deficiency,
–12 O2 (g) [ V¢M + OO
¥
+ h• (3-195)
which is accommodated by either anion va-
cancies, e.g., UO2 – x , or metal interstitials, If the hole is then localized at a metal ion,
e.g., Nb1+ y O2 or 2) an anion excess, which we can consider this as the chemical equiv-
is accommodated by either anion intersti- alent of having M 3+ ions formally present
tials, e.g., UO2 + x or metal vacancies, e.g., in the sublattice of M 2+ ions.
230 3 Diffusion Kinetics in Solids
Assuming reaction Eq. (3-194) is the gen partial pressure will expose the actual
¥
preferred one, we find that the equilibrium charge state of the vacancy, i.e., VM or V¢M .
constant for this reaction is In practice, this kind of differentiation
between charge states and even defect
[VM× ]
K1 = 1/ 2 (3-196) types does not often occur unambiguously.
pO 2 The result of plotting log D* versus log pO2
thereby immediately showing that the often does not show precisely one slope
metal vacancy concentration depends on and often there is also some curvature.
the partial pressure of oxygen in the fol- These effects can sometimes be associated
lowing way: with contributions to the diffusion from the
¥
¥
three types of vacancy, VM , V¢M , and V≤M,
[VM ] = K1 p1/2
O2 (3-197) e.g., Co1– d O (Dieckmann, 1977). Probably
Accordingly, the metal vacancy concentra- the most successful and convincing appli-
tion is directly proportional to p1/2 cation of this mass-action law approach has
O2 .
Assuming that the reaction in Eq. (3- made the defect types in Ni1– d O compat-
195) is the preferred one, we find that the ible with data from tracer diffusion, electri-
equilibrium constant for this reaction is cal conductivity, chemical diffusion, and
deviation from stoichiometry/partial pres-
[ h•] [O O× ] [VM
′] sure (Peterson, 1984).
K2 = 1/ 2
(3-198)
pO2 At higher partial pressures of oxygen,
which in some cases such as Co1– d O,
¥
The concentration of [OO ] is essentially Mn1– d O, and Fe1– d O can result in large
constant and is usually absorbed into K2 . deviations from stoichiometry, we can log-
The condition of electrical neutrality re- ically expect that defect interactions could
quires that play an important role. First, the concentra-
[h• ] = [V¢M ] (3-199) tions in the mass-action equations should
be replaced by activities. This so greatly
thereby showing that the metal vacancy complicates the analysis that progress has
concentration depends on the partial pres- been fairly slow. The retention of concen-
sure of oxygen in the following way: trations rather than activities may be per-
[V¢M ] = K21/2 p21/4 (3-200) missible, however, because there is some
evidence that non-ideal effects tend to can-
Accordingly, the metal vacancy concentra- cel in a log D/log pO2 plot (Murch, 1981).
tion is directly proportional to p1/4
O2 . At high defect concentrations, defect clus-
Let us assume that the migration of tering to form complex defects which
metal ions will principally be via vacancies themselves might move or act as sources
in the nonstoichiometric region, since they and sinks for more mobile defects, is pos-
are the predominant defect. Since the tracer sible. Two very well-known examples,
diffusion coefficient of the metal ions de- identified by neutron scattering, are the
pends directly on vacancy concentration Koch – Cohen clusters in Fe1– d O (Koch
(see Sec. 3.3.4), then the tracer diffusion and Cohen, 1969) and the Willis cluster in
coefficient will depend on oxygen partial UO2 + x (Willis, 1978). Although attempts
¥
pressure in the same way that either [VM ] have been made to incorporate such defects
or [V¢M ] does. A measurement of the tracer into mass-action equations, it is probably
diffusion coefficient as a function of oxy- fair to say that the number of adjustable pa-
3.5 Experimental Methods for Measuring Diffusion Coefficients 231
rameters thereby resulting and the lack of stoichiometry (cation vacancies) in the cat-
any precise independent information on the ion sublattice) is generally well understood
mobility or lifetime of the defect complexes except at large deviations from stoichiome-
make any conclusions reached rather spec- try (see the review by Peterson (1984)).
ulative. New statistical mechanical ap- Other oxides such as MgO and Al2O3 , al-
proaches have considerable promise, how- though apparently simple, are less well
ever (see the review by Murch (1995)). understood because of the low intrinsic de-
Somewhat analogous to changing the de- fect population and the presence of extrin-
fect concentration, i.e., deviation from stoi- sic defects coming from impurities (Subba-
chiometry by changing the oxygen partial Rao, 1985). An understanding of oxygen
pressure, is the doping of ionic crystals diffusion in oxides has been hampered by
with ions of a different valence from the the lack of suitable radioisotopes, but
host metal ion in order to produce defects. methods using 18O and secondary ion mass
As an example, consider the solubility of spectrometry (SIMS) have rapidly changed
an oxide ᑧ2O3 in an oxide MO2 . The im- the situation (Rothman, 1990). Ionic con-
plication in the stoichiometry ᑧ2O3 is that ductivity (where possible) (see Sec. 3.3.2
the ᑧ ions have a valence of + 3, compared for the relation to diffusion) has tradition-
with + 4 for M in the oxide MO2 . For ally been the measurement which has given
charge neutrality reasons the doped oxide much more information on oxygen move-
adopts either oxygen vacancies, metal in- ment; see, for example, the review by
terstitials, or some electronic defect, the Nowick (1984).
choice being dictated by energetics. Let us
consider the first of these:
ᑧ2O3 Æ VO•• + 2 ᑧ M
¢ + 3 OO (3-201) 3.5 Experimental Methods for
Measuring Diffusion Coefficients
where ᑧ M ¢ is a ᑧ ion on a M site. The site
fraction of oxygen vacant sites is not quite In this section we briefly discuss the
directly proportional to the dopant site more frequently used methods of measur-
fraction because extra normal oxygen sites ing diffusion coefficients.
are created in the process. The diffusion
coefficient of oxygen here is proportional
3.5.1 Tracer Diffusion Methods
to the concentration of vacant oxygen sites.
For very large deviations from stoichiom- By far the most popular and, if per-
etry, or high dopant concentrations, those formed carefully, the most reliable of the
extrinsic defects greatly outnumber the in- experimental methods for determining
trinsic defects. However, at low dopant con- “self” and impurity diffusion coefficients
centrations (or small deviations from stoi- is the thin-layer method. Here, a very thin
chiometry), all the various mass-action laws layer of radiotracer of the diffusant is de-
must be combined. This complicates the posited at the surface of the sample. The
analysis, and space prevents us from dealing deposition can be done by evaporation,
with it here. For a more detailed discussion electrochemical methods, decomposition
of this and the foregoing, we refer to Kof- of a salt, sputtering, etc. The diffusant is
stad (1972) and Schmalzried (1995). permitted to diffuse for a certain time t at
Generally, cation diffusion in the oxides high temperature, “high” being a relative
of the transition metals (which show non- term here. If the thickness of the layer de-
232 3 Diffusion Kinetics in Solids
---
posited is much smaller than ÷D t , then Eq. hard radiation the method can give com-
(3-8) describes the time evolution of the parable accuracy to the counting of sec-
concentration – depth profile. After diffu- tions.
sion, the concentration – depth profile in A few elements do not have convenient
the sample is established by sectioning and radioisotopes. The diffusion part of the ex-
counting the radioactivity in each section. periment is still performed in much the
A number of techniques are available for same way but now with stable isotopes. In
sectioning. For thick sections ≥ 3 µm, me- some cases, e.g., 18O, the diffusant source
chanical grinding is the standard method. is in the gas phase, although a thin source
Microtomes can be used for sections of of oxide containing 18O can be deposited in
about 1 µm, electrochemical methods for some cases. After diffusion, nuclear reac-
≥ 5 nm, and sputtering for ≥ 1 nm. The dif- tion analysis of 18O can be used (see, for
fusion coefficient is obtained from the example, the review by Lanford et al.
slope of the ln C (x, t) versus x 2 plot (see (1984)), in order to establish the concentra-
Eq. (3-8)). Fig. 3-22 shows an exemplary tion profile. Much more popular recently,
tracer concentration profile of this type especially for oxygen, is SIMS (see, for ex-
(Mundy et al., 1971). ample, the reviews by Petuskey (1984),
A variant of the method is the “residual Kilner (1986) and Manning et al. (1996)).
activity” or “Gruzin (1952) method”. Excellent detailed (and entertaining) ex-
Rather than counting the activity in each positions and critiques of all methods
section, the activity remaining in the sam- available for measuring tracer diffusion co-
ple is determined. This requires an integra- efficients in solids have been provided by
tion of Eq. (3-8) and a knowledge of the ra- Rothman (1984, 1990).
diation absorption characteristics of the
material. At the limits of very soft or very
3.5.2 Chemical Diffusion Methods
The chemical or interdiffusion coeffi-
cient can be determined in a variety of
ways. For binary alloys, the traditional
method has been to bond together two sam-
ples of different concentrations. Interdiffu-
sion is then permitted to occur for a time t
and at high temperature. The concentration
profile can be established by sectioning
and chemical analysis, but since about the
late 1960s the use of an electron micro-
probe has been the usual procedure for ob-
taining the concentration profile directly. If
the compositions of the starting samples
are relatively close, and the interdiffusion
coefficient is not highly dependent on com-
position, then analysis of the profile with
Figure 3-22. A tracer concentration profile for self- Eq. (3-12) can lead directly to the interdif-
diffusion in potassium at 35.5 °C (Mundy et al., fusion coefficient D̃ at the average compo-
1971). sition. The more usual procedure is to use
3.5 Experimental Methods for Measuring Diffusion Coefficients 233
the Boltzmann –Matano graphical integra- of 18O (exchanging with 16O) can be moni-
tion analysis (see Eqs. (3-13) and (3-14)) to tored in the gas phase with a mass spec-
obtain the interdiffusion coefficient and its trometer in order to lead to the tracer diffu-
composition dependence. If fine insoluble sion coefficient of oxygen (see, for exam-
wires (to act as a marker) are also incorpo- ple, Auskern and Belle (1961)). A draw-
rated at the interface of the two samples, back of methods which do not rely on sec-
then the marker shift or Kirkendall shift tioning is the necessary assumption of
with respect to the original interface can rapid gas/surface reactions.
be measured. This shift, in association
with the interdiffusion coefficient, can be
3.5.3 Diffusion Coefficients
used to determine the intrinsic diffusion
by Indirect Methods
coefficients of both alloy components at
the composition of the marker (see Sec. There are a number of methods for ob-
3.2.2.4). taining diffusion coefficients that do not
The rate of absorption or desorption of depend on solutions of Fick’s Second Law.
material from the sample can also be used Unfortunately space does not permit a de-
to determine the chemical diffusion coeffi- tailed discussion of these. We shall men-
cient. This method is useful only where one tion them here for completeness and direct
component is fairly volatile, such as oxy- the reader to the appropriate sources. Often
gen in some nonstoichiometric oxides and these methods extend the accessible tem-
hydrogen in metals. The surface composi- perature range for diffusion measurements.
tion is normally assumed to be held con- In all cases specific atomic models need to
stant. The concentration profile can be de- be introduced in order to extract a diffusion
termined by sectioning or electron micro- coefficient. Inasmuch as there are inevita-
probe analysis and can be analysed with bly approximations in a model, the result-
the aid of Eq. (3-9) to yield the interdiffu- ing diffusion coefficients may not be as re-
sion coefficient. The Boltzmann – Matano liable as those obtained from a concentra-
analysis can be used if there is a composi- tion profile. In some cases additional corre-
tion dependence. lation information can be extracted, e.g.,
The amount of material absorbed or de- nuclear methods. The diffusion coefficients
sorbed from the couple, e.g., weight found from these methods are essentially
change, can be used with Eq. (3-11) to “self” diffusion coefficients but chemical
yield the interdiffusion coefficient. Eqs. (3- diffusion coefficients for interstitial solutes
---
9) and (3-11) refer to situations when ÷D t are obtained from the Gorsky effect.
is very small compared with the geometry
of the sample, i.e., diffusion into an infinite
3.5.3.1 Relaxation Methods
sample is assumed. When the sample must
be assumed to be finite, e.g., fine particles, In the relaxation methods net atomic mi-
other, generally more complex, solutions gration is due to external causes such as
are required; see Crank (1975). A closely stress of magnetic field (Nowick and Berry,
related experiment can be performed where 1972). The best known phenomena are (a)
the gas phase is chemically in equilibrium the Gorsky effect (Gorsky, 1935); for a re-
with the solid, but is dosed with a stable view see Alefeld et al. (1970) and for an in-
isotope. An example is 18O, a stable iso- troduction see Borg and Dienes (1988); (b)
tope of oxygen. The absorption by the solid the Snoek effect (Snoek, 1939); for an ex-
234 3 Diffusion Kinetics in Solids
position see Wert (1970) and for an intro- Most applications have been to solids in-
duction see Borg and Dienes (1988); and corporating the rather ideal isotope 57Fe.
(c) the Zener effect (Zener, 1947, 1951); See Mullen (1984) for a detailed discus-
for an introduction see Bocquet et al. sion of MBS and Vogl and Feldwisch (1998)
(1996). for several recent examples of applica-
tions.
3.5.3.2 Nuclear Methods
3.5.4 Surface Diffusion Methods
A number of nuclear methods have be-
come of increasing importance for deter- The techniques for measuring surface
mining diffusion coefficients in solids. The diffusion are generally quite different from
first of these is quasielastic neutron scat- solid-state techniques. For short distance,
tering (QNS), which has often been used microscopic or “intrinsic” diffusion (see
for situations where the diffusion coeffi- Sec. 3.2.2.5), the method of choice is the
cients are larger than about 10–11 m2 s –1 field ion microscope. Much elegant work
and the diffusing species exhibits a reason- has been carried out with this technique by
ably high scattering cross-section. Most Erlich and co-workers; see for example Er-
of the applications have been to hydrogen lich and Stott (1980) and Erlich (1980).
(protonic) in metals (see, for example, With this method single atoms can be im-
Janot et al., 1986). For a detailed introduc- aged and followed. Another method, this
tion to the subject see Lechner (1983) and time using the field electron microscope
Zabel (1984). The second method is nu- (Lifshin, 1992), correlates fluctuations of
clear magnetic resonance (NMR). NMR is an emission current from a very small area
especially sensitive to interactions of the with density fluctuations arising from
nuclear moments with fields produced by surface diffusion in and out of the probe
their local environments. Diffusion of a nu- area (Chen and Gomer, 1979). Other meth-
clear moment can cause variations in these ods include quasi-elastic scattering of
fields and can significantly affect the ob- low energy He atoms (formally analogous
served resonance. In particular, diffusion to quasi-elastic neutron scattering) and re-
affects a number of relaxation times in laxation measurements, making use of dep-
NMR. In favorable cases diffusion coeffi- osition of the adsorbate in a non-equilib-
cients between 10–18 and 10–10 m2 s –1 are rium configuration, followed by annealing
accessible. For a detailed introduction to which permits relaxation to equilibrium.
the subject see Stokes (1984) and Heitjans Techniques useful for following the relaxa-
and Schirmer (1998). We make special tion process at this microscopic level in-
mention of pulsed field gradient (PFG) clude pulsed molecular beam combined
NMR, which has been found to be espe- with fast scanning IR interferometry
cially useful for studying anomalous diffu- (Reutt-Robey et al., 1988) and work func-
sion (Kärger et al., 1998). Finally, we men- tion measurements (Schrammen and Hölzl,
tion Mössbauer spectroscopy (MBS), which 1983).
shows considerable promise for under- Many methods exist for long distance or
standing diffusion processes in solids. The macrosopic diffusion. When the diffusing
general requirement is that the diffusion species is deposited as a source and the ap-
coefficient of Mössbauer active isotope propriate geometries are known then scan-
should be larger than about 10–13 m2 s –1. ning for the concentration profile followed
3.7 References 235
Bonzel, H. P. (1990), Diffusion in Metals and Alloys, Gomer, R. (1958), Discussions Faraday Soc. 28, 23.
in: Landolt-Börnstein, Vol. 13: Mehrer, H. (Ed.). Gorsky, W. S. (1935), Z. Phys. Sowjetunion 8, 457.
Berlin: Springer-Verlag. Gösele, U., Frank, W., Seeger, A. (1980), Appl. Phys.
Borg, R. J., Dienes, G. J. (1988), An Introduction to 23, 361.
Solid State Diffusion. New York: Academic. Gruzin, P. L. (1952), Dokl. Acad. Nauk. SSSR 86,
Brandes, E. A. (Ed.) (1983), Smithells Metal Refer- 289.
ence Book, 6th edn. London: Butterworths. Gupta, D., Oberschmidt, J. (1984), in: Diffusion in
Carlson, P. T. (1978), Met. Trans. A 9, 1287. Solids: Recent Developments: Dayananda, M. A.,
Carslaw, H. S., Jaeger, J. C. (1959), Conduction of Murch, G. E. (Eds.). Warrendale: TMS/AIME,
Heat in Solids. New York: Oxford University Press. p. 121.
Catlow, C. R. A. (1977), Proc. Roy. Soc. London, Ser. Gupta, D., Campbell, D. R., Ho, P. S. (1978), In: Thin
A 353, 533. Films – Interdiffusion and Reactions: Poates, J.
Catlow, C. R. A., Mackrodt, W. C. (1982), Computer M., Tu, K. N., Mayer, J. W. (Eds.). New York: Wi-
Simulation of Solids (Lecture Notes in Physics, ley, p. 61.
Vol. 166). Berlin: Springer. Harrison, L. G. (1961), Trans. Faraday Soc. 57,
Chemla, M. (1956), Ann. Phys. Paris 13, 959. 1191.
Chen, J. R., Gomer, R. (1979), Surface Science 79, Hart, E. W. (1957), Acta Metall. 5, 597.
413. Hehenkamp, Th., Faupel, F. (1983), Acta Metall. 31,
Chen, W. K., Peterson, N. L. (1972), J. Phys. Chem. 691.
Solids 33, 881. Hehenkamp, Th., Schmidt, W., Schlett, V. (1980),
Chen, W. K., Peterson, N. L. (1973), J. Phys. (Paris) Acta Metall. 28, 1715.
34, C 9. Heitjans, P., Schirmer, A. (1998), in: Diffusion in
Crank, J. (1975), The Mathematics of Diffusion. New Condensed Matter: Kärger, J., Heitjans, P., Haber-
York: Oxford University Press. landt, R. (Eds.). Wiesbaden: Vieweg, p. 116.
Darken, L. S. (1948), Trans. AIME 175, 184. Hempelmann, R. (1984), J. Less-Common Met. 101,
De Bruin, H. J., Murch, G. E., Bakker, H., Van der 69.
Meij, L. P. (1975), Thin Solid Films 25, 47. Ho, Y. K. (1982), Thesis. London: Imperial College.
Dieckmann, R. (1977), Z. Phys. Chem. N. F. 107, Howard, R. E. (1966), Phys. Rev. 144, 650.
189. Howard, R. E., Lidiard, A. B. (1964), Rep. Prog.
Dorn, J. E., Mitchell, J. B. (1966), Acta Metall. 14, Phys. 27, 161.
70. Howard, R. E., Manning, J. R. (1967), Phys. Rev.
Elcock, E. W., McCombie, C. W. (1958), Phys. Rev. 154, 561.
109, 605. Huntington, H. B. (1975), in: Diffusion in Solids: Re-
Erlich, G. (1980), J. Vac. Sci. Technol. 17, 9. cent Developments: Nowick, A. S., Burton, J. J.
Erlich, G., Stott, K. (1980), Ann. Rev. Phys. Chem. (Eds.). New York: Academic, p. 303.
31, 603. Huntington, H. B., Ghate, P. B. (1962), Phys. Rev.
Faupel, F., Hehenkamp, Th. (1986), Phys. Rev. B 34, Lett. 8, 421.
2116. Jacucci, G. (1984), in: Diffusion in Crystalline Sol-
Faupel, F., Hehenkamp, Th. (1987), Acta Metall. 35, ids: Murch, G. E., Nowick, A. S. (Eds.). New
771. York: Academic, p. 431.
Fisher, J. C. (1951), J. Appl. Phys. 22, 74. Jain, H., Gupta, D. (Eds.) (1994), Diffusion in Amor-
Flynn, C. P. (1972), Point Defects and Diffusion. Ox- phous Materials. Warrendale: TMS.
ford: Clarendon Press. Janot, C., Petry, W., Richter, D., Springer, T. (Eds.)
Foiles, S. M., Daw, M. S. (1987), J. Mater. Res. 2, 5. (1986), Atomic Transport and Defects in Materials
Frank, F. C., Turnbull, D. (1956), Phys. Rev. 104, by Neutron Scattering. Berlin: Springer-Verlag.
617. Jones, M. J., Le Claire, A. D. (1972), Phil. Mag. 26,
Frank, W., Gösele, U., Mehrer, H., Seeger, A. (1984), 1191.
in: Diffusion in Crystalline Solids: Murch, G. E., Kaur, I., Mishin, Y., Gust, W. (1995), Fundamentals
Nowick, A. S. (Eds.). New York: Academic, p. 64. of Grain and Interphase Boundary Diffusion. Chi-
Fredericks, W. J. (1975), in: Diffusion in Solids: Re- chester: Wiley.
cent Developments: Nowick, A. S., Burton, J. J. Kärger, J., Fleischer, G., Roland, U. (1998), in: Diffu-
(Eds.). New York: Academic, p. 381. sion in Condensed Matter: Kärger, J., Heitjans, P.,
Freer, R. (1980), J. Mater. Sci. 15, 803. Haberlandt, R. (Eds.). Wiesbaden: Vieweg, p. 40.
Friauf, R. J. (1957), Phys. Rev. 105, 843. Kelly, S. W., Scholl, C. A. (1987), J. Phys. C 20,
Fukai, Y., Sugimoto, H. (1985), Adv. Phys. 34, 263. 5293.
Ghate, P. B. (1964), Phys. Rev. A 113, 1167. Kikuchi, R., Sato, H. (1969), J. Chem. Phys. 51, 161.
Gjostein, N. A. (1970), in: Techniques of Metals Re- Kikuchi, R., Sato, H. (1970), J. Chem. Phys. 53,
search, Vol. IV, Part 2: Bunshah, R. F. (Ed.). New 2707.
York: Wiley, p. 405. Kikuchi, R., Sato, H. (1972), J. Chem. Phys. 57, 4962.
3.7 References 237
Kilner, J. A. (1986), Mat. Sci. Forum 7, 205. Matano, C. (1933), Jpn. Phys. 8, 109.
Kirkaldy, J. S., Young, D. J. (1987), Diffusional in McKee, R. A. (1980 a), Phys. Rev. B 21, 4269.
the Condensed State. Brookfield: The Institute of McKee, R. A. (1980 b), Phys. Rev. B 22, 2649.
Metals. McKee, R. A. (1981), Phys. Rev. B 23, 1609.
Koch, F., Cohen, J. B. (1969), Acta Crystallogr., Mehrer, H. (1990), Diffusion in Metals and Alloys,
Sect. B 25, 275. Landolt-Börnstein New Series, Berlin: Springer.
Kofstad, P. (1972), Nonstoichiometry, Diffusion and Mehrer, H. (1998), in: Diffusion in Condensed Mat-
Electrical Conductivity in Binary Metal Oxides. ter: Kärger, J., Heitjans, P., Haberlandt, R. (Eds.).
New York: Wiley. Wiesbaden: Vieweg, p. 1.
Kwok, T., Ho, P. S., Yip, S. (1984), Phys. Rev. B 29, Mishin, Y. (1997), Defect Diffusion Forum 143 –147,
5363. 1357.
Lam, N. Q., Rothman, S. J., Mehrer, H., Nowicki, L. Mishin, Y., Herzig, C., Bernardini, J., Gust, W.
J. (1973), Phys. Status Solidi B 57, 225. (1997), Int. Mat. Reviews 42, 155.
Lanford, W. A., Benenson, R., Burman, C., Wielun- Mohan Rao, M., Ranganathan, S. (1984), Mat. Sci.
ski, L. (1984), in: Nontraditional Methods in Dif- Forum 1, 43.
fusion: Murch, G. E., Birnbaum, H. K., Cost, J. R. Moleko, L. K., Allnatt, A. R., Allnatt, E. L. (1989),
(Eds.). Warrendale: TMS/AIME, p. 155. Phil. Mag. A 59, 141.
Laskar, A. L. (1990), in: Diffusion in Materials: Monty, C. (1992), in: Diffusion in Solids: Unsolved
Laskar, A. L., Bocquet, J. L., Brébec, G., Monty, Problems: Murch, G. E. (Ed.). Zürich: Trans.
C. (Eds.). Dordrecht: Kluwer, Academic, p. 459. Tech. Publications, p. 259.
Laskar, A. L. (1992), in: Diffusion in Solids: Un- Mullen, J. G. (1984), in: Nontraditional Methods in
solved Problems: Murch, G. E. (Ed.). Zürich: Diffusion, Murch, G. E., Birnbaum, H. K., Cost, J.
Trans. Tech. Publications, p. 207. R. (Eds.). Warrendale: TMS/AIME, p. 59.
Lechner, R. E. (1983), in: Mass Transport in Solids: Mundy, J. N. (1992), in: Diffusion in Solids: Un-
Bénière, F., Catlow, C. R. A. (Eds.). London: solved Problems: Murch, G. E. (Ed.). Zürich:
Plenum, p. 169. Trans. Tech. Publications, p. 1.
Le Claire, A. D. (1963), Br. J. Appl. Phys. 14, 351. Mundy, J. N., Miller, T. E., Porte, R. J. (1971), Phys.
Le Claire, A. D. (1970), in: Physical Chemistry – An Rev. B 3, 2445.
Advanced Treatise, Vol. 10: Eyring, H., Henderson, Murch, G. E. (1975), J. Nucl. Mater. 57, 239.
D., Jost, W. (Eds.). New York: Academic, p. 261. Murch, G. E. (1981), J. Phys. Chem. Solids 42, 227.
Le Claire, A. D. (1975), in: Treatise in Solid State Murch, G. E. (1982 a), Phil. Mag. A 46, 575.
Chemistry, Vol. 4, Reactivity of Solids: Hannay, N. Murch, G. E. (1982 b), Phil. Mag. A 45, 685.
B. (Ed.). New York: Plenum, p. 1. Murch, G. E. (1982 c), Solid State Ionics 6, 295.
Le Claire, A. D. (1978), J. Nucl. Mater. 69/70, 70. Murch, G. E. (1982 d), Solid State Ionics 7, 177.
Le Claire, A. D. (1981), Phil. Mag. A 43, 911. Murch, G. E. (1984 a), in: Diffusion in Crystalline
Le Claire, A. D. (1992), in: Diffusion in Solids: Un- Solids: Murch, G. E., Nowick, A. S. (Eds.). New
solved Problems: Murch, G. E. (Ed.). Zürich: York: Academic, p. 379.
Trans. Tech. Publications, p. 19. Murch, G. E. (1984 b), Phil. Mag. A 49, 21.
Le Claire, A. D., Rabinovitch, A. (1984), in: Diffu- Murch, G. E. (1995), Rad. Effects and Defects in Sol-
sion in Crystalline Solids: Murch, G. E., Nowick, ids 134, 1.
A. S. (Eds.). New York: Academic, p. 259. Murch, G. E., Belova, I. V. (1998), in: Diffusion
Lidiard, A. B. (1986), Acta Metall 34, 1487. Mechanisms in Crystalline Materials: Mishin, Y.,
Lifshin, E. (1992), Characterization of Materials, Vogl, G., Cowern, N., Catlow, R., Farkas, D.
Part I, Materials Science and Technology: R. W. (Eds.). MRS Symposium Proceedings 527. War-
Cahn, P. Haasen, E. J. Kramer (Eds.), Vol. 2A. rendale: MRS, p. 135.
Weinheim: VCH. Murch, G. E., Catlow, C. R. A. (1987), J. Chem. Soc.,
Lim, S. H., Murch, G. E., Oates, W. A. (1990), Phil. Faraday Trans. II 83, 1157.
Mag. B 61, 337. Murch, G. E., Dyre, J. C. (1989), CRC Crit. Rev.
Manning, J. R. (1959), Phys. Rev. 113, 1445. Solid State Mat. Sci. 15, 345.
Manning, J. R. (1964), Phys. Rev. 136, 1758. Murch, G. E., Rothman, S. J. (1979), Phil. Mag. A 43,
Manning, J. R. (1968), Diffusion Kinetics for Atoms 229.
in Crystals. Princeton: Van Nostrand. Murch, G. E., Rothman, S. J. (1985), Diffusion De-
Manning, J. R. (1970), Met. Trans. 1, 499. fect Data 42, 17.
Manning, J. R. (1971), Phys. Rev. B 4, 1111. Murch, G. E., Thorn, R. J. (1978), Phil. Mag. A 38,
Manning, J. R. (1975), in: Mass Transport in Ceram- 125.
ics: Cooper, A. R., Heuer, A. H. (Eds.). New York: Murch, G. E., Thorn, R. J. (1979), J. Phys. Chem.
Plenum, p. 75. Solids 40, 389.
Manning, P. S., Sirman, J. D., Kilner, J. A. (1996), Murray, A. D., Murch, G. E., Catlow, C. R. A.
Solid State Ionics 93, 125. (1986), Solid State Ionics 18/19, 196.
238 3 Diffusion Kinetics in Solids
Nowick, A. S. (1984), in: Diffusion in Crystalline Schrammen, P., Hölzl, J. (1983), Surface Science
Solids: Murch, G. E., Nowick, A. S. (Eds.). New 130, 203.
York: Academic, p. 143. Seeger, A. (1997), Defect Diffusion Forum 143 –147,
Nowick, A. S., Berry, B. S. (1972), Anelastic Relax- 21.
ation in Crystalline Solids. New York: Academic. Siegel, W. (1978), J. Nucl. Mater. 69/70, 117.
Peterson, N. L. (1975), in: Diffusion in Solids: Re- Smith, R. P. (1953), Acta Metall. 1, 578.
cent Developments: Nowick, A. S., Burton, J. J. Snoek, J. L. (1939), Physica 6, 591.
(Eds.). New York: Academic, p. 116. Stark, J. P. (1972), J. Appl. Phys. 43, 4404.
Peterson, N. L. (1978), J. Nucl. Mater. 69/70, 3. Stark, J. P. (1974), Acta Metall. 22, 1349.
Peterson, N. L. (1983), Int. Met. Rev. 28, 65. Stark, J. P. (1976), Solid State Diffusion. New York:
Peterson, N. L. (1984), Mater. Sci. Forum 1, 85. Wiley.
Petuskey, W. T. (1984), in: Nontraditional Methods Stokes, H. T. (1984), in: Nontraditional Methods in
in Diffusion: Murch, G. E., Birnbaum, H. K., Cost, Diffusion, Murch, G. E., Birnbaum, H. K., Cost, J.
J. R. (Eds.). Warrendale: TMS/AIME, p. 179. R. (Eds.). Warrendale: TMS/AIME, p. 39.
Philibert, J. (1991), Atom Movements: Diffusion and SubbaRao, E. C. (1985), Diffusion Defect Data 41, 1.
Mass Transport in Solids, Paris: Les Editions de Suzuoka, T. (1961), Trans. Jpn. Inst. Met. 2, 25.
Physique. Vineyard, G. H. (1957), J. Phys. Chem. Solids 3, 121.
Pontikis, V. (1990), in: Diffusion in Materials: Viswanathan, R., Burgess, Jr., D. R., Stair, P. C.,
Laskar, A. L., Bocquet, J. L., Brébec, G., Monty, Weitz, E. (1982), J. Vac. Sci. Technol. 20, 605.
C. (Eds.). Dordrecht: Kluwer, Academic, p. 37. Vogl, G., Feldwisch, R. (1998), in: Diffusion in Con-
Reutt-Robey, J. E., Doren, D. J., Chabel, Y. J., densed Matter: Kärger, J., Heitjans, P., Haberlandt,
Christman, S. B. (1988), Phys. Rev. Lett 61, 2778. R. (Eds.), Wiesbaden: Vieweg, p. 144.
Rhead, G. E. (1989), Int. Materials Reviews 34, 261. Völkl, J., Alefeld, G. (1975), in: Diffusion in Solids:
Rothman, S. J. (1984), in: Diffusion in Crystalline Recent Developments: Nowick, A. S., Burton, J. J.
Solids: Murch, G. E., Nowick, A. S. (Eds.). New (Eds.). New York: Academic, p. 231.
York: Academic, p. 1. Wang, C. C., Akbar, S. A. (1993), Acta Metall.
Rothman, S. J. (1990), in: Diffusion in Materials: Mater. 41, 2807.
Laskar, A. L., Bocquet, J. L., Brébec, G., Monty, Warburton, W. K., Turnbull, D. (1975), in: Diffusion
C. (Eds.). Dordrecht: Kluwer, Academic, p. 269. in Solids: Recent Developments: Nowick, A. S.,
Rothman, S. J., Peterson, N. L. (1967), Phys. Rev. Burton, J. J. (Eds.). New York: Academic, p. 171.
154, 552. Wert, C. A. (1970), J. Phys. Chem. Solids 31, 1771.
Rothman, S. J., Peterson, N. L., Robinson, J. T. Willis, B. T. M. (1978), Acta Crystallogr., Sect. A 34,
(1970), Phys. Status Solidi 39, 635. 88.
Sato, H. (1970), in: Physical Chemistry – An Ad- Zabel, H. (1984), in: Nontraditional Methods in Dif-
vanced Treatise, Vol. 10: Eyring, H., Henderson, fusion, Murch, G. E., Birnbaum, H. K., Cost, J. R.
D., Jost, W. (Eds.). New York: Academic, p. 579. (Eds.). Warrendale: TMS/AIME, p. 1.
Sato, H. (1970), in: Nontraditional Methods in Diffu- Zener, C. (1947), Phys. Rev. 71, 34.
sion, Murch, G. E., Birnbaum, H. K., Cost, J. R. Zener, C. (1951), J. Appl. Phys. 22, 372.
(Eds.). Warrendale: TMS/AIME, p. 203. Zhang, L., Oates, W. A., Murch, G. E. (1988), Phil.
Sato, H. (1989), in: Superionic Conductors: Laskar, Mag. A 58, 937.
A. R., Chandra, S. (Eds.). New York: Academic, p. Zhang, L., Oates, W. A., Murch, G. E. (1989 a), Phil.
439. Mag. B 60, 277.
Sato, H., Kikuchi, K. (1971), J. Chem. Phys. 55, Zhang, L., Oates, W. A., Murch, G. E. (1989 b), Phil.
677/702. Mag. A 59, 171.
Schilling, W. (1978), J. Nucl. Mater. 69/60, 465. Zhang, L., Murch, G. E. (1997), Defect Diffusion
Schmalzried, H. (1995), Chemical Kinetics of Solids, Forum. 141, 59.
Weinheim: VCH.
4 Statistical Theories of Phase Transitions
Kurt Binder
e ik strain tensor
h exponent related to the decay of correlations at Tc
l wavelength
l label of the phonon branch
L factor changing the length scale
m chemical potential
m magnetic moment per spin
Dm chemical potential difference
m coex chemical potential at phase coexistence
n exponent related to the correlation length
x correlation length
xi concentration difference (Eq. (4-126))
r density
r (x) charge density distribution function
t characteristic time
j phase shift in Friedel potential
F order parameter
F (x) order parameter density
Fi concentration difference between two sublattices
F ms order parameter of metastable state
Fs spinodal curve
c (q) wavevector-dependent susceptibility
c el dielectric tensor
cT response function of the order parameter
cT isothermal susceptibility
cT staggered susceptibility
y concentration difference between two sublattices
y I, II ordering parameter for D03 structure
ya amplitude of mass density waves
w characteristic frequency (for fluctuations)
w̃ scaling function
AF antiferromagnetic
cg coarse-grained (as index)
coex coexistent (as index)
CV cluster variation method
DAG dysprosium aluminum garnet
ESR electron spin resonance
f.c.c. face-centered cubic
LRO long-range order
MC Monte Carlo (method)
MD molecular dynamics
MFA mean field approximation
MFA molecular field approximation
244 4 Statistical Theories of Phase Transitions
of first-order phase transitions (Gunton are ubiquitous, e.g., diamond is the meta-
et al., 1983; Koch, 1984; Binder, 1984 b, stable modification of solid carbon whereas
1989; Haasen et al., 1984; Zettlemoyer, graphite is the stable phase, some com-
1969; Abraham, 1974; Gunton and Droz, ments about the statistical mechanics of
1983; Herrmann et al., 1992). The latter metastability are made in Sec. 4.4.
subject is treated in two other chapters
in this book (see the Chapters by Wagner
et al. (2001) and Binder and Fratzl (2001)) 4.2 Phenomenological Concepts
and will not be considered here.
In this chapter, we shall briefly discuss In this section, the main facts of the the-
the statistical thermodynamics of phase ory of phase transitions are summarized,
transitions on a phenomenological macro- and the appropriate terminology intro-
scopic level (Sec. 4.2), i.e., the Landau the- duced. Rather than aiming at completeness,
ory of first- and second-order transitions, examples which illustrate the spirit of the
critical and multicritical phenomena, and main approaches will be discussed, includ-
also the dynamics of fluctuations at phase ing a discussion of critical phenomena and
transitions, and the effects of quenched scaling laws.
disorder in solids. The second part (Sec.
4.3) is devoted to the complementary “mi-
4.2.1 Order Parameters and the Landau
croscopic” approach, where we start from
Symmetry Classification
“model Hamiltonians” for the system under
consideration. The statistical mechanics of Table 4-1 lists condensed matter systems
such models naturally leads to the consider- that can exist in several phases, depending
ation of some computational methods, i.e., on external thermodynamics parameters
methods based on the molecular field the- such as pressure, p, temperature, T, electric
ory and its generalizations, and methods or magnetic fields (E, H). We assume that
based on computer simulation techniques. an extensive thermodynamic variable can
Since in solid materials metastable phases be identified (i.e., one that is proportional
y I = ma − mc + mb − md
y II = − ma + mc + mb − md
mm ≡ (1 / N ) ∑ (2 ci − 1) (4-7)
iŒm
is taken by Al and ci = 0 otherwise) and the the change of sign of the order parameter
critical concentration, F i = ci – ccrit . In all for H = 0 and thus odd powers such as
of these cases F (x) is considered to be an F 3 (x) do not occur; this is true for magnets
order parameter field defined in continuum (no direction of the magnetization is pre-
space: ferred without magnetic field in a ferro-
magnet) and for sublattice ordering of al-
F ( x ) = ∑ Fi / Ld (4-9)
i Œ Ld
loys such as b-CuZn (since whether the Cu
atoms preferentially occupy sublattices a, c
x being the center of gravity of the “vol-
in Fig. 4-3 or sublattices b, d is equivalent),
ume” Ld . The appropriate magnitude of the
but it is not true in general (e.g., third-order
linear dimension L of the coarse-graining
terms do occur in the description of the
cell will be discussed later; obviously it
Cu3Au structure, or in the ordering of rare
must be much larger than the lattice spac-
gas monolayers adsorbed on graphite in
ing in order for the continuum description –
the ÷3 structure, as will be discussed be-
to make sense. Then a Helmholtz energy
low).
functional F [F (x)] is assumed:
In order to understand the meaning and
1 F use of Eq. (4-10), consider first the fully
F [F ( x )] = 0 (4-10)
kB T kB T homogeneous case, —F (x) ∫ 0, F (x) ∫ F 0 ;
then F [F ] is the standard Helmholtz en-
+ ∫ dx ⎧⎨ r F 2 ( x ) + uF 4 ( x )
1 1
ergy function of thermodynamics, which
⎩2 4
needs to be minimized with respect to F in
H 1 ⎫
− F (x) + [ R ∇F ( x )]2 ⎬ order to determine the thermal equilibrium
kB T 2d ⎭ state, and Ú dx = V the total volume of the
where F0 is the background Helmholtz en- system. Thus,
ergy of the disordered phase, and r, u, and (4-12)
1 ⎛ ∂F ⎞
R are phenomenological constants. (In fact, ⎜ ⎟ = r F0 + u F0 = 0
3
Figure 4-6. Schematic variation of the Helmholtz energy in the Landau model at transitions of (a) second or-
der and (b) first order as a function of the (scalar) order parameter F. Cases (a) and (b) assume a symmetry
around F = 0, whereas case (c) allows a cubic term.
order case F (F ) has two minima for T < Tc this value, i.e.,
which continuously merge as T Æ Tc and
only one minimum at F = 0 remains for [ F (F0 ) − F ( 0 )]/(V kB T )
menological coefficient, w. For u > 0, F (F ) only in mean-field theory, but lacks any
may have two minima (Fig. 4-6 c); again fundamental justification in statistical me-
the transition occurs when the minima are chanics. Schemes such as those shown in
equally deep. For r = r¢ (T – T0 ) this hap- Fig. 4-6 make sense for a local “coarse-
pens when grained free energy function” only (which
depends on the length scale L introduced in
Tc = T0 + 8 w 2/(81 u r¢) (4-18)
Eq. (4-9)), see Sec. 4.2.2, but not for the
the order parameter jumping there from global free energy.
F 0 = – 9 r/w to F 0 = 0. Again a stability After this digression we return to the
limit of ordered state in the disordered generalization of the Landau expansion,
phase occurs, i.e. Eq. (4-10), the case where the order param-
eter has vector or tensor character. How
T1 = T0 + w 2/(4 u r¢) (4-19)
can we find which kinds of term appear in
At this point, an important caveat should be the expansion?
emphasized: free energy curves involving Basically, there are two answers to this
several minima and maxima as drawn in question. A general method, which follows
Fig. 4-6 are used so often in the literature below, is a symmetry classification based
that many researchers believe these con- on group theory techniques (Landau and
cepts to be essentially rigorous. However, Lifshitz, 1958; Tolédano and Tolédano,
general principles of thermodynamics indi- 1987). A very straightforward alternative
cate that in thermal equilibrium the ther- approach is possible if we consider a par-
modynamic potentials are convex func- ticular model Hamiltonian Ᏼ (for a brief
tions of their variables. In fact, F (F 0 ) discussion of some of the most useful mod-
should thus be convex as a function of F 0 , els of statistical mechanics for studying
which excludes multiple minima! For Fig. phase transitions, see Sec. 4.3.1). We can
4-6 a and b this means that for T < Tc in then formulate a microscopic mean-field
states with – F 0 < F < F 0 (where F 0 is the approximation (MFA), such as the effec-
solution of Eqs. (4-13) or (4-14), respec- tive field approximation of magnetism
tively) the thermal equilibrium state is (Smart, 1966) or the Bragg-Williams ap-
not a pure homogeneous phase: rather, the proximation for order – disorder phenomena
minimum free energy state is given by the in alloys (De Fontaine, 1979), where we
double-tangent construction to F (F ) and then expand the MFA Helmholtz energy di-
this corresponds to a mixed phase state (the rectly.
relative amounts of the coexisting phases As an example of this approach, we con-
are given by the well known lever rule). sider a specific model of a ternary alloy
Now it is standard practice, dating back to where each lattice site i may be taken by ei-
van der Waals’ interpretation of this equa- ther an A, a B, or a C atom, assuming con-
tion of state for fluids, to interpret the part centrations cA = cB = cC = 1/3, and assuming
of F (F ) in Fig. 4-6 which lies above the that an energy J is achieved if two neigh-
F (F ) given by the double-tangent con- boring sites are taken by the same kind of
struction as a metastable state provided atom. This is a special case of the q-state
that c T = (∂2 F/ ∂F 2 )T < 0, whereas states Potts model (Potts, 1952; Wu, 1982), the
with c T < 0 are considered as intrinsically Hamiltonian being
unstable states. As will be seen in Sec. 4.4,
Ᏼ Potts = − ∑ J dSi S j , Si = {1, 2, …, q} (4-20)
this notion is intrinsically a concept valid 〈 ij 〉
256 4 Statistical Theories of Phase Transitions
ments of G0 fall away (spontaneously This approach also carries over to cases
broken symmetry); the remaining symme- where the order parameter is a tensor. For
try elements form a subgroup G of G0 . example, for elastic phase transitions
Now the invariance of F must hold separ- (Cowley, 1976; Folk et al., 1976) the order
ately for terms F k of any order k and this parameter is the strain tensor {e ik}. Apply-
requirements fixes the character of the ing the summation convention (indices oc-
terms that may be present. curring twice in an expression are summed
Rather than formulating this approach over), the Landau expansion is
systematically, which would require a
F [e ik (x)]
lengthy and very mathematical exposi-
tion (Tolédano and Tolédano, 1987), we = F0 + Ú dx ( –12 ciklm e ik e lm
rather illustrate it with a simple example. + –13 ciklmrs
(3)
e ik e lm e rs
Suppose a cubic crystal exhibits a transi- + –14 ciklmrsuv
(4)
e ik e lm e rs e uv
tion from a para-electric to a ferroelectric
+ … + gradient terms) (4-27)
phase, where a spontaneous polarization
P = (P1 , …, Pn ), n = 3, appears. F is then Here the ciklm are elastic constants and
(3) (4)
given as follows: ciklmrs and ciklmrsuv analogous coefficients
of higher order (“anharmonic”) terms. For
= 0 + ∫ dx ⎡⎢ c el−1 P 2
F [ P ( x )] F 1
most elastic transitions symmetry permits
kB T kB T ⎣2 some nonzero ciklmrs (3)
and hence leads to
1 ⎛ n 4 n ⎞ first-order transitions. Examples of such
+ ⎜ u ∑ Pm + u ′ ∑ Pm2 Pj2 ⎟ + … systems are the “martensitic transitions” in
4 ⎝ m =1 m < j =1 ⎠
Nb3Sn and V3Si; there the elastic distortion
R2 n ⎤ jumps at the transition from zero to a very
+ ∑ (∇Pm )2 + …⎥ (4-25)
small value (e ik ≈ 10– 4 ) and a Landau ex-
2d m =1 ⎦
pansion makes sense. Note that the Landau
Whereas the quadratic term of a general theory is not very useful quantitatively for
dielectric medium would involve the in- transitions which are very strongly first or-
verse of the dielectric tensor, ∑ ( c el−1 )ij Pi Pj , der, since in that case high-order terms in the
ij
this term is completely isotropic for cubic Taylor expansion are not negligible. More
crystals. Inversion symmetry requires in- about martensitic phase transformations can
variance against P Æ – P and hence no be found in the Chapter by Delaey (2001).
third-order term occurs. The fourth-order Just as in a ferromagnetic transition the
term now contains the two “cubic invari- inverse magnetic susceptibility c –1 van-
( )
ants” ∑ Pi4 and ∑ Pi2 . Here we invoke
2 ishes at Tc (Eq. (4-4)) and in a ferroelectric
i i transition the inverse dielectric susceptibil-
the principle that all terms allowed by sym- ity c el–1 vanishes at Tc (Eq. (4-25)), in a
metry will actually occur. Now Eq. (4-25) “ferroelastic” transition one of the elastic
leads, in the framework of the Landau the- constants ciklm vanishes at T0 . For example,
ory, to a second-order transition if both in KCN such a softening is observed for the
elastic constant c44 . Note that this material
u>0 and u + u¢ > 0 (4-26)
has already been mentioned as an example
whereas otherwise we have a first-order of a phase transition where the order pa-
transition (then terms of sixth order are rameter is the electric quadrupole moment
needed in Eq. (4-25) to ensure stability). tensor (Eq. (4-8)), describing the orienta-
258 4 Statistical Theories of Phase Transitions
and ya and y– a are the (complex!) order structures in terms of „spin density waves“
parameter amplitudes. In constructing the (Villain, 1959; Kaplan, 1959). We shall
free energy expansion with the help of Eq. outline the connection between this ap-
(4-28), note that the periodicity of the proach and the MFA, generalizing the ap-
underlying graphite lattice allows invariant proach of Eqs. (4-21) – (4-23) slightly, in
“umklapp” terms (the phase factors of Sec. 4.3. Here we only mention that these
third-order terms add up to a reciprocal lat- concepts are closely related to the descrip-
tice vector of the graphite lattice). Keeping tion of structural transitions in solids,
only those terms in the Landau expansion, where the order parameter can often be
–
and which are nonzero in the ÷3 structure, considered as a “frozen phonon”, i.e., a dis-
with the real order parameter components placement vector wave (Bruce and Cowley,
––
(i = ÷–1) 1981). Note also that the approach of con-
3 centration waves is not restricted to solids,
F1 = ∑ (ya − y − a ) /( 2 3 ) but can also be used to describe meso-
a =1 phases in fluid block-copolymer melts
3
(Leibler, 1980; Fredrickson and Helfand,
F2 = ∑ (ya − y − a ) /( 2 i 3 ) (4-30)
a =1 1987; Fredrickson and Binder, 1989; Bin-
der, 1994), see Fig. 4-7, in liquid crystal-
the expression line polymers, etc.
F 1 As a final remark in this section we men-
= − r (F12 + F22 ) (4-31) tion that not for all phase transitions in sol-
V kB T 2
ids there does exist a group – subgroup rela-
1 1
+ w (F13 − 3F1 F22 ) + u (F12 + F22 )2 tionship between the two groups G1 and G2
3 4 describing the symmetry of the phases co-
is obtained. This is equivalent to the result existing at the transition. These phases can-
for the three-state Potts model, Eq. (4-24) not be distinguished by an order parameter
(Alexander, 1975), after the quadratic form which is zero in one phase and becomes
is diagonalized. A three-dimensional ana- nonzero in the other. Such transitions must
log of this order (where the planes exhibit- be of first order. Examples of this situation
–
ing ÷3 structure are stacked together to are well known for structural phase transi-
form a hexagonal lattice) occurs in the tions, e.g., the tetragonal – orthorhombic
intercalated compound C6Li (Guerard and transition of BaTiO3 or the “reconstruc-
Herold, 1975; Bak and Domany, 1979). tive” transition from calcite to aragonite
Similarly to the description of the den- (Guymont, 1981). For the so-called “non-
sity modulation in the superstructure of disruptive transitions” (Guymont, 1981),
adsorbed layers (in two dimensions) or the new structure can still be described in
interstitial compounds (in three dimen- the framework of the old structure (i.e., its
sions) in terms of mass density waves of symmetry elements can be specified, the
the adsorbate (or interstitial, respectively), Wyckoff positions can be located, etc.).
the superstructure ordering in binary alloys Landau-type symmetry arguments still
can be described in terms of concentration yield information on the domain structures
waves (Khachaturyan, 1962, 1963, 1973, arising in such phase transitions (Guymont,
1983; De Fontaine, 1975, 1979). The same 1978, 1981). The tetragonal – orthorhombic
concept was used even earlier to describe transition of BaTiO3 is considered to be an
the magnetic ordering of helimagnetic spin example of such a non-disruptive transi-
260 4 Statistical Theories of Phase Transitions
is solved in terms of the Euler – Lagrange For T < Tc , using Eq. (4-13) the term 5 v F04
equation near Tc is still negligible, and hence
( ∂ᏸ / ∂x ) −
d
[ ∂ᏸ / ∂ ( dx / dt )] = 0 −1 1
cT = = (4-37)
dt 2 kB T r 2 r ′ ( Tc − T )
we conclude here that x = R2 /( − 2 rd ) = ( R / 2 r ′ d ) (T − Tc ) −1/ 2
( ∂ f / ∂y ) − ∇ [ ∂ f / ∂ (∇y )] = 0 Eqs. (4-36) and (4-37) simply imply the
in order that the Helmholtz energy func- well known Curie – Weiss law, i.e., the ex-
tional is a minimum. Eq. (4-10) thus yields ponents g, g ¢, defined in Eq. (4-4), are, in
the framework of the Landau theory,
r F ( x ) + uF 3 ( x ) + v F 5 ( x ) g = g¢ = 1 (4-38)
R2 2 H (x)
− ∇ F (x) = (4-33) Comparing Eqs. (4-35) and (4-37) we also
d kB T
conclude that
Using
n = n ¢ = 1/2 (4-39)
F (x) = F 0 + DF (x) = F 0 + DFq exp (i q · x) The physical meaning of the correlation
Eqs. (4-32) and (4-33) yield, on linearizing length x is easily recognized if the well
in DFq for small Hq , the wavevector-de- known fluctuation relationship
pendent susceptibility 1
c (q) = ∑ exp( i q ⋅ x )
DFq kB T x
c (q) ≡ × [ 〈F ( 0 ) F ( x )〉 − F02 ] (4-40)
Hq
−1 is expanded to second order in q as
1 ⎡ R2 2 ⎤
⎢ r + 3 u F0 + 5 v F0 + d q ⎥
2 4
=
kB T ⎣ ⎦ 1 ⎧
c (q) ≈ ⎨∑ [ 〈F ( 0 ) F ( x )〉 − F0 ]
2
cT kB T ⎩ x
= (4-34 )
1 + x 2 q2 1 ⎫
− ∑ ( q ⋅ x )2 [ 〈F ( 0 ) F ( x )〉 − F02 ]⎬ (4-41)
From Eq. (4-34) we can simply read off
2 x ⎭
the temperature dependence of the suscep- which can be written as
tibility c T (Eq. (4-4)) and the correlation
1
length x of order parameter fluctuations, c (q) = ∑ [〈F ( 0 ) F ( x )〉 − F02 ]
which is generally expected to behave as kB T x
⎧ q2
⎧xˆ ( T / Tc − 1) − n , T > Tc × ⎨1 − ∑ x 2 [〈F ( 0 ) F ( x )〉 − F02 ]
x =⎨ˆ −n′
(4-35) ⎩ 2d x
⎩x ′ (1 − Tc / T ) , T < Tc ⎫
∑ [〈F ( 0 ) F ( x )〉 − F02 ]⎬ (4-42)
Above Tc , the terms 3 u F02 + 5 v F04 = 0, x ⎭
hence Comparing Eqs. (4-34) and (4-42), we find
1 1 that in terms of the order parameter corre-
cT = = (4-36 )
kB T r r ′ ( T − Tc ) lation function G (x),
x = R2 /( rd ) = ( R / r ′ d ) (Tc − T ) −1/ 2 G( x ) = 〈F ( 0 ) F ( x )〉 − F02 (4-43)
262 4 Statistical Theories of Phase Transitions
and c T and x 2 are related to the zeroth and Tc . For T < Tc , we have instead, from Eqs.
second moment of the correlation function: (4-10) and (4-13)
F − F0
= F02 ⎛ + F02 ⎞
1 r u
cT = ∑ G (x) (4-44)
⎝2 4 ⎠
kB T x kB T V
− r2 (1 − T / Tc )2
1 ≈ − r′2 (4-52)
2
x = ∑ x G (x) ∑ G (x)
2
(4-45) =
4u 4 kB2 u
2d x x
which implies C = r¢ 2 (T /Tc ) / (4 kB u) for
Eq. (4-44) clearly shows that the critical di-
T < Tc . This jump singularity of the specific
vergence of c T occurs because the correla-
heat at Tc instead of the power law in Eq.
tions G (x) become long ranged: whereas
(4-5) is formally associated with vanishing
off Tc the correlation function for large | x|
critical exponents:
decays exponentially,
a = a¢ = 0 (4-53)
ln G (x) Æ – | x|/ x , | x| Æ • (4-46)
The question must now be asked whether
at Tc we have a power-law decay of the cor- this description of critical phenomena is
relation function: accurate in terms of the Landau theory. The
G (x) = Ĝ | x| – (d – 2 + h ) , T = Tc (4-47) free energy functional F [F (x)] in Eq.
(4-10) should be considered as an effective
The exponent describing this critical decay Hamiltonian from which a partition func-
has been defined such that tion Z can be obtained so that the true
h=0 (4-48) Helmholtz energy becomes:
where in suitable units the constant on the and Laursen, 1980), the essential part of
left-hand side of the equality is of order the magnetic Hamiltonian is a magneto-
unity. Using Eqs. (4-4), (4-6), and (4-35) static dipole – dipole interaction,
we obtain (anticipating g = g ¢ and n = n ¢)
3 z2 − x2
– n d +2 b + g Ᏼ = − m2 ∑ Si S j , Si = ± 1 (4-64)
const. O t (4-61) i≠ j | x |5
Inserting the Landau values for the critical where m is the magnetic moment per spin
exponents b = 1/2, n = 1/2, g = 1 and know- and z the coordinate of x in the direction of
ing the critical amplitude of x ~ R, the inter- the uniaxial anisotropy. Unlike the iso-
action range, Eq. (4-60) implies tropic Ising Hamiltonian in Eq. (4-56), here
the sign of the interaction depends on the
const. O (R/a)d t (d – 4)/2 (4-62) direction in the lattice. Therefore, fluctua-
This condition for the validity of the Lan- tions display an essential anisotropy: in-
dau theory for d < 4 always breaks down stead of the isotropic result (Eq. (4-34)) of
as t Æ 0 (T Æ Tc ). In fact, for d < 4 close the well known Ornstein-Zernike type,
enough to Tc a regime occurs where fluctu- the wavevector-dependent susceptibility
ations dominate the functional integral (Eq. c (q) = S (q)/kB T (S (q) = structure factor)
(4-54)). The “crossover” from the mean- becomes anisotropic:
field regime (where the Landau description S (q) ~ [x – 2 + q2 + g (qz /q)2 ]–1 (4-65 a)
is essentially appropriate) to the non-mean-
field regime occurs at a reduced tempera- where qz is the z-component of q and g is a
ture distance t = tcr , where Eq. (4-62) is constant. Whereas Eq. (4-32) implies that
treated as an equality: for d = 3, S (q x = 1) = S (q = 0)/2, the equation q x = 1
which for Eq. (4-32) defines a sphere of ra-
tcr ~ (R/a)– 6 (4-63) dius x –1 in q-space now defines an aniso-
tropic surface. From
Hence systems with a large but finite range
of interaction behave in an essentially Lan- x – 2 = q2 + g (qz /q)2 (4-65 b)
dau-like manner. An example of such a be-
we obtain an object having the shape of a
havior is the unmixing critical point in
flat disc, with radius 1/x in the qx – qy plane
polymer fluid mixtures with a high degree
but maximum extension of order 1/x 2 in
of polymerization N : since each coil has a
the qz direction. This object, defined by
radius rc ~ N 1/2, the monomer density in-
S (q) = –12 max {S (q)}, can be interpreted as
side the sphere taken by a polymer coil is
the Fourier transform of the correlation
only of order r ~ N/rc3 ~ N – 1/2, which im-
volume. This implies that the correlation
plies that each coil interacts with N 1/2
volume in real space is a long ellipsoid,
neighbor coils, and the effective interaction
with linear dimensions x in the qx , qy di-
volume (R/a)d in Eqs. (4-62) and (4-63)
rection but with linear dimension x 2 in the
should be taken as N 1/2, i.e., tcr ~ N – 1 (De
qz direction. In d-dimensions, this argu-
Gennes, 1979; Binder 1984 c, 1994).
ment suggests for uniaxial dipolar systems:
At this point, we emphasize that using
N (t) = [x (t)/a]d in Eq. (4-60) is not valid in N (t) = [x (t)/a]d +1 (4-66)
systems with long-range anisotropic inter-
which yields in Eq. (4-60)
actions, such as uniaxial dipolar magnets.
For a system such as LiTbF4 (Als-Nielsen const. O t – n (d +1) + 2 b + g = t – (d – 3)/2 (4-67)
4.2 Phenomenological Concepts 265
where in the last equality, the Landau ex- well over a wide temperature range (Salje,
ponents have been used. Comparing Eqs. 1990).
(4-61) and (4-67) shows that d-dimensional For all other systems we do not expect
uniaxial dipolar systems somehow corre- the Landau critical exponents to be an ac-
spond to (d + 1)-dimensional systems with curate description of the critical singular-
isotropic short-range forces. Therefore the ities. This is borne out well by the exact so-
marginal dimension d*, above which the lution for the two-dimensional Ising model
Landau theory predicts the values of criti- (Onsager, 1944; McCoy and Wu, 1973),
cal exponents correctly, is d* = 3 for uniax- where the critical exponents have the values
ial dipolar systems, unlike the standard iso-
a = 0 (log), b = 1/8, g = 7/4,
tropic short-range case where d* = 4.
For elastic phase transitions in cubic d = 15, n = 1, h = 1/4 (4-71)
crystals where the combination of elastic Obviously, these numbers are a long way
constants c11 – c12 softens as the critical from the predictions of the Landau theory.
point is appproached, such as the system Note that a = 0 in Eq. (4-71) has the mean-
PrAlO3 , the structure factor S (q) has the ing C ~ | log | t ||, unlike the jump singular-
form: (4-68) ity of the Landau theory.
S (q) ~ [x – 2 + q2 + A (qz /q)2 + B (q^ /q)2 ]– 1 Even in the marginal case of uniaxial di-
polar ferromagnets (or uniaxial ferro-
where q = (qz , q^ ) and A and B are con- electrics; see, e.g., Binder et al. (1976) for
stants. In this case the correlation vol- a discussion), the Landau theory is not
ume defined from the condition S (q) = completely correct; it turns out (Larkin and
–12 max {S (q)} in real space is a flat disk Khmelnitskii, 1969; Aharony, 1976) that
with radius x 2 and diameter x , i.e., for elas- the power laws have the Landau form but
tic systems: are modified with logarithmic correction
N (t) = [x (t)/a]d +2 (4-69) factors:
In this case, the Ginzburg criterion, Eq. cT = Ĝ± t – 1 | ln t |1/3, tÆ0 (4-72 a)
(4-60), yields 1/3
C = ± | ln t | , tÆ0 (4-72 b)
– n (d +2) + 2 b + g – (d – 2)/2
const. O t =t (4-70) Ms = B̂ (– t)1/2 | ln (– t)|1/3, t Æ 0 (4-72 c)
instead of Eqs. (4-62) or (4-67), respec- Ms | Tc = D̂ H 1/3 1/3
| ln H | , HÆ0 (4-72 d)
tively. Since the marginal dimensionality
for such systems is d* = 2, three-dimen- Although for uniaxial ferroelectrics the ex-
sional systems should be accurately de- perimental evidence in favor of Eq. (4-72)
scribed by the Landau theory, and this is is still scarce, convincing experimental ev-
what is found experimentally for PrAlO3. idence does exist for dipolar ferromagnets
Examples for elastic phase transitions such as LiTbF4 (see, e.g., Als-Nielsen and
which are of second order are scarce, as is Laursen (1980) for a review).
understandable from the symmetry consid- Eq. (4-72) results from including fluctu-
erations which lead to first-order transi- ation contributions to the functional inte-
tions in most cases (Cowley, 1976; Folk gral in Eq. (4-54) systematically, which can
et al., 1976). However, for many first-order be done in a most powerful way be renor-
ferroelastic transitions the Landau descrip- malization group theory (Aharony, 1973,
tion also fits the experimental data very 1976; Fisher, 1974; Wilson and Kogut,
266 4 Statistical Theories of Phase Transitions
1974; Ma, 1976; Amit, 1984; Yeomans, “universality”. Again we wish to convey
1992). A description of this theory is be- only the flavor of the idea to the reader,
yond the scope of this chapter; we only rather than to present a thorough discus-
mention that it is this method which yields sion. Consider, for example, the decay of
highly accurate approximations for the val- the correlation function of order parameter
ues of critical exponents of three-dimen- fluctuations at the critical point, Eq. (4-47):
sional systems. For example, for the uniax- changing the length scale from r to r¢ = L r
ial magnets (as described by the Ising Ha- would change the prefactor Ĝ but leave the
miltonian, Eq. (4-56), experimental exam- power-law invariant. The physical inter-
ples being the antiferomagnets MnF2 or pretation of this fact is the “fractal struc-
FeF2) or the ordering alloy b-brass, the ex- ture” of critical correlations (Mandelbrot,
ponents are predicted to be (LeGuillou and 1982): just as a fractal geometric object
Zinn-Justin, 1980): looks the same on every length scale, the
pattern of critical fluctuations looks the
a ≈ 0.110, b ≈ 0.325, g ≈ 1.240,
same, irrespective of the scale. Slightly off
d ≈ 4.82, n ≈ 0.63, h ≈ 0.032 (4-73) Tc , then, there should be only one relevant
Whereas within the Landau theory the “or- length scale in the problem, the correlation
der parameter dimensionality” n does not length x , which diverges as T Æ Tc , and de-
matter, as far as the values of the critical tails such as the precise behavior of the
exponents are concerned, n does matter if correlation functions for interatomic dis-
we go beyond the Landau theory. For ex- tances should not matter. As a result, it is
ample, for isotropic magnets as described plausible that the correlation function
by the well known Heisenberg model of G (x, x ) should depend only on the ratio of
magnetism, the two lengths x and x , apart from a scale
factor:
Ᏼ Heis = − ∑ Jij Si ⋅ S j − m H ∑ Siz (4-74)
i≠ j G (x, x ) = Ĝ x – (d – 2 + h ) G̃ (x /x ) (4-76)
Si being a unit vector in the direction of the Obviously, the condition G̃ (0) = 1 for the
magnetic moment at lattice site i, we have “scaling function” G̃ (x /x ) ensures that Eq.
(LeGuillou and Zinn-Justin, 1980): (4-76) crosses over smoothly to Eq. (4-47)
a ≈ – 0.116, b ≈ 0.365, g ≈ 1.391, as T Æ Tc , where x Æ ∞.
d ≈ 4.82, n ≈ 0.707, h ≈ 0.034 (4-75) From the “scaling hypothesis” in Eq.
(4-76) we immediately derive a “scaling
Again these results are in fair agreement law” relating the exponents g , n and h. Us-
with available experimental data, such as ing Eq. (4-44) and denoting the surface of
the isotropic antiferromagnet RbMnF3 (see a d-dimensional unit sphere as Ud (Ud = 4 p
Als-Nielsen (1976) for a review of experi- in d = 3) we obtain
ments on critical point phenomena). Note
1 G˜
that the negative value of a in Eq. (4-75) cT = ∑ G (x) ≈ ∫ G ( x , x ) dx
implies that the specific heat has a cusp of kB T x kB Tc
finite height at Tc . ∞
Ud Gˆ 1−h
The renormalization group theory also = ∫x G˜ ( x / x ) dx
kB Tc 0
provides a unifying framework and justifi-
∞
cation for two important concepts about 2 −h Ud Gˆ 1−h
=x ∫ᒗ G˜ (ᒗ) dᒗ (4-77)
critical phenomena, namely “scaling” and kB Tc 0
4.2 Phenomenological Concepts 267
Since the integral over ᒗ = x /x in Eq. (4-77) cancel out. On the other hand, this also
yields a constant, we conclude yields
c T ~ x (2 – h ) ~ | t | – n (2– h ) ~ | t | – g F | T = Tc ~ H 1/(1 + g /b ) = H1/d ,
cf. Eqs. (4-4) and (4-35). Thus the scaling i.e., d = 1 + g /b (4-83)
relation results in
The scaling assumptions also imply that
g = n (2 – h ) (4-78) exponents above and below Tc are equal,
i.e, a = a ¢, g = g¢, and n = n¢. Now there is
Eq. (4-76) indicates that G (x, x ) depends
still another scaling law which results from
on the two variables x and x in a rather spe-
considering all degrees of freedom inside a
cial form, namely, it is a homogeneous func-
correlation volume x d to be highly corre-
tion. A similar homogeneity assumption is
lated with each other, while different corre-
also true for the singular part of the free
lation volumes can be considered as essen-
energy, (4-79) tially independent: this argument suggests
F = Freg + t 2 – a F̃ (H̃ ), H̃ = H Ĝ t – g – b/B̂ that the singular part of the Helmholtz en-
ergy per degree of freedom can be written
where the singular temperature dependence
as Fsing ≈ [1/Ns (t)] ¥ const., since indepen-
for H̃ = 0 is chosen to be compatible with
dent degrees of freedom do not contribute
Eq. (4-5), Freg is a background term which
any singular free energy. Since
is analytic in both T and H even for T = Tc ,
while F̃ (H̃ ) is another “scaling function”. Ns (t) ~ x d ~ |t | – dn
At this point, we present Eq. (4-79) as a
postulate, but it should be emphasized that we conclude, by comparison with Eq.
both Eqs. (4-76) and (4-79) can be justified (4-79), that
from Eq. (4-54) by the renormalization dn = 2 – a (4-84)
group approach.
Combining Eqs. (4-1) and (4-79) we ob- Note that this so-called “hyperscaling rela-
tain: tion” with the Landau-theory exponents is
(4-80)
Gˆ ( 2 − a − g − b ) ˜ ˜ only true at the marginal dimension d* = 4,
F = − ( ∂F / ∂H )T = − t F′ ( H ) whereas the other scaling relationships ob-
Bˆ
viously are fulfilled by Landau exponents
Defining – (Ĝ /B̂) F̃ ¢ (H̃ ) ∫ B̂ M̃ (H̃ ) with independent of the system dimensionality.
M̃ (0) = 1, we obtain: However, all scaling relationships (includ-
ing Eq. (4-84)) are satisfied for the two-
F = B̂t b M̃ (H̃), b = 2 – a – g – b (4-81)
and three-dimensional Ising model (Eqs.
since for H̃= 0, Eq. (4-80) must reduce to (4-71) and (4-73)) and the three-dimen-
Eq. (4-6). Taking one more derivative we sional Heisenberg model, Eq. (4-75).
find Let us consider the two-dimensional
cT = (∂F /∂H )T = Ĝ t – g M̃¢(H̃) (4-82) Heisenberg model. For this model the ef-
fects of statistical fluctuations are so strong
hence Eq. (4-79) is compatible with Eq. that they destabilize the ordering alto-
(4-4), as it should be. The condition that gether: no spontaneous ordering exists in
Eq. (4-81) reduces to the critical isotherm d = 2 for any system with order parameter
for t Æ 0 requires that M̃ (H̃ Æ •) ~ dimensionality n ≥ 2, and a critical point
H̃ b /(g + b ), in order that the powers of t occurs only at zero temperature, Tc = 0, for
268 4 Statistical Theories of Phase Transitions
n > 2. Thus d = 2 is the “lower critical di- model of magnetism, Eq. (4-74), since the
mensionality” d1 for n ≥ 2, whereas for ferroelectric ordering would also be truly
n = 1 we have d1 = 1: quasi-one-dimensional isotropic. However, in the presence of a
orderings are always unstable. The case nonvanishing “cubic anisotropy”, u¢ ≠ 0, in
n = 2, d = 2 is very special: a phase transi- general a different universality class re-
tion still occurs at a nonzero Tc , the so- sults. The effect of such higher order invar-
called “Kosterlitz – Thouless” transition iants in the Landau expansion are particu-
(Kosterlitz and Thouless, 1973); for T < Tc larly drastic again in systems with reduced
we have Ms (T ) ∫ 0 while at the same time dimensionality; e.g., the model with n = 2
x (T ) = •, and the correlation function but cubic anisotropy no longer exhibits a
shows an algebraic decay, Eq. (4-47), with Kosterlitz – Thouless transition, but rather
a temperature-dependent exponent h . At a nonzero order parameter is stabilized
first sight, this behavior may appear fairly again. Owing to the “marginal” character
esoteric, but in fact it is closely related of the cubic anisotropy in d = 2, however,
to the “roughening transition” of crystal the behavior is not Ising-like but rather the
surfaces (Weeks, 1980). Such roughening pathological case of a “universality class”
transitions have been observed for high-in- with “nonuniversal” critical exponents oc-
dex crystal faces of various metals (Sala- curs. The latter then do depend on “micro-
non et al., 1988). The power-law decay of scopic” details, such as the ratio of the
correlations, Eq. (4-47), can be related to interaction strengths between nearest and
the behavior of the height – height correla- next nearest neighbors in the lattice (Krin-
tion function of a crystal surface in the sky and Mukamel, 1977; Domany et al.,
rough state (Weeks, 1980). As is well 1978; Swendsen and Krinsky, 1979; Lan-
known, the roughness or flatness of crystal dau and Binder, 1985). Again such prob-
surfaces has a profound effect on their ad- lems are not purely academic, but relevant
sorption behavior, crystal growth kinetics, for order – disorder transitions in chemi-
etc. (Müller-Krumbhaar, 1977). sorbed layers on metal surfaces, such as O
The other important concept about criti- on W (110) (see Binder and Landau (1989)
cal phenomena is “universality”: since the for a review of the theoretical modeling of
only important length scale near a critical such systems).
point is provided by the correlation length One important consequence of univer-
which is much larger than all “micro- sality is that all gas – fluid critical points in
scopic” lengths such as lattice spacing and three-dimensional systems, all critical
interaction range, it is plausible that “de- points associated with unmixing of fluid or
tails” on the atomic scale do not matter, and solid binary mixtures, and all order – disor-
systems near a critical point behave in the der critical points involving a single-com-
same way provided that they fall in the ponent order parameter (such as b-brass)
same “universality class”. It turns out that belong to the same universality class as the
universality classes (for systems with three-dimensional Ising model. All these
short-range interactions!) are determined systems not only have the same critical ex-
by both spatial dimensionality d and order ponents, but also the scaling functions
parameter dimensionality n, and in addi- M̃ (H̃ ), G̃ (ᒗ) etc. are all universal. Also,
tion, the symmetry properties of the prob- critical amplitude ratios are universal,
lem: e.g., Eq. (4-25) for n = 3 and u¢ = 0 such as Ĉ/Ĉ +, (Eq. (4-4), Â/¢ (Eq. (4-5)),
falls in the same class as the Heisenberg or more complicated quantities such as
4.2 Phenomenological Concepts 269
D̂ Ĝ B̂d –1. The theoretical calculation and netization (due to the coupling F 2 (x) M (x)
experimental estimation of such critical in the Hamiltonian, as mentioned above).
amplitude ratios are discussed by Privman However, if we now consider the variation
et al. (1991). of F with M, combining both laws we get
An interesting problem concerns the (assuming a > 0)
critical singularities associated with nonor-
dering “fields”. These singularities occur
F ~ [Mc (T ) – M ]b /(1– a ) (4-85)
because the ordering field F (x) considered Considering now the phase transition in the
only in Eqs. (4-10) and (4-54) couples to space of {M, T} as independent thermody-
another quantity: e.g., for an antiferromag- namic parameters, we then find that on
net in a magnetic field the order parameter crossing the critical line Mc (T ) the expo-
F (the sublattice magnetization) couples to nent describing the vanishing of the order
the uniform magnetization M, to the lattice parameter is b /(1 – a ) rather than b. This
parameters, to the electron density, etc. effect generally occurs when the critical
Since the Hamiltonian is invariant against line depends on extensive (rather than in-
interchange of the sublattices which im- tensive) thermodynamic variables and is
plies a sign change of F , this coupling is called “Fisher renormalization” (Fisher,
quadratic in F 2. Whereas in mean field 1968). This is very common for order – dis-
theory the resulting critical behavior of the order phenomena in adsorbed layers at
nonordering “field” is then proportional to fixed coverage, or in alloys at fixed con-
·F Ò2 ~ t 2 b , taking fluctuations into account, centration, for unmixing critical points in
the actual critical behavior is proportional ternary mixtures, etc. Other exponents also
to ·F 2 Ò ~ t (1– a ), i.e., an energy-like singu- become “renormalized” similarly, e.g., g is
larity. Such energy-like singularities are replaced by g /(1 – a ), etc. A detailed anal-
predicted for the electrical resistivity r el at ysis (Fisher, 1968) shows that no such
various phase transitions (Fisher and “renormalization” of critical exponents oc-
Langer, 1968; Binder and Stauffer, 1976 a), curs for systems with a fixed concentration
for the refractive index n r (Gehring and c if the slope of the critical line vanishes,
Gehring, 1975; Gehring, 1977), etc. Look- dTc (c)/dc = 0, for the considered concen-
ing for anomalies in the temperature deriv- tration. This is approximately true for b-
ative dr el (T )/dT, dn r (T )/dT etc., which brass, for instance.
should exhibit specific-heat-like singular-
ities, is often a more convenient tool for lo-
4.2.3 Second-Order Versus First-Order
cating such phase transitions than measure-
Transitions; Tricritical and Other
ments of the specific heat itself.
Multicritical Phenomena
An important phenomenon occurs when
a nonordering field which is coupled to the An important question for any phase
order parameter in this way, such as the transition is deciding a priori whether it
magnetization in an antiferromagnet in an should be a second- or first-order transi-
external field, is held fixed. Suppose we tion; general principles are sought in an-
first study the approach to criticality by let- swering this question, such that there
ting the uniform magnetic field h tend to its would be no need for specific experimental
critical value hc (T ): we then have the law data.
F ~ [hc (T ) – h]b for the order parameter In the Landau theory, general symme-
and M – Mc (T ) ~ [hc (T ) – h]1– a for the mag- try conditions exist which allow second-
270 4 Statistical Theories of Phase Transitions
order transitions (Lifshitz, 1942). We here by Baxter (1973), these models have sec-
merely state them, without even attempting ond-order transitions. The critical expo-
to explain the group-theoretical language nents for these models are now believed
(see, e.g., Tolédano and Tolédano, 1987, to be known exactly, e.g., (q = 3) a = 1/3,
for a thorough treatment): b = 1/9, g = 13/9 and (q = 4) a = 2/3,
b = 1/12, g = 7/6 (Den Nijs, 1979; Nienhuis
(i) The order parameters F transform as
et al., 1980). As mentioned in Sec. 4.2.1, an
a basis of a single irreducible repre-
experimental example for the three-state
sentation X of the group G0 character- –
Potts model is the ÷3 superstructure of var-
izing the symmetry properties of the
ious adsorbates on graphite. Convincing
disordered phase.
experimental evidence for a specific heat
(ii) The symmetric part of the representa-
divergence of He4 on grafoil described by
tion X3 should not contain the unit
a = 1/3 was presented by Bretz (1977). The
representation.
system O on Ru (001) in the p (2 ¥ 2) struc-
(iii) If the antisymmetric part of X2 has a
ture (Piercy and Pfnür, 1987) falls into the
representation, the wavevector q asso-
class of the four-state Potts models and
ciated with X is not determined by
again the data are in reasonable agreement
symmetry. In this case q is expected
with the predicted critical exponents.
to vary continuously in the ordered
In d = 3 dimensions, however, the three-
phase.
states Potts model has a weak first-order
If these conditions are met, a transition transition (Blöte and Swendsen, 1979) so
can nevertheless be first order, because a that the Landau rule (ii) is not violated.
fourth-order term can be negative (see Sec. This is of relevance for metallurgical sys-
4.2.1). If they are not met, the transition tems such as the CuAu ordering (Fig. 4-8)
must be first order, according to the Lan- on the f.c.c. lattice, which, according to
dau rules. Domany et al. (1982), belongs to the class
The first of these rules essentially says of the three-states Potts model, while the
that if in the ordered phase two quantities Cu3Au structure belongs to the class of the
essentially independent of each other (not four-state Potts model. Another example of
related by any symmetry operation, etc.) the three-state Potts model is the structural
play the role of order parameter compo- transition of SrTiO3 stressed in the [111]
nents F1 , F2 , there is no reason why in the direction (Bruce and Aharony, 1975). The
quadratic term r1 (T ) F12 + r2 (T ) F22 of the order – disorder transition for all systems
Landau expansion the coefficients r1 (T ), described by these types of ordering are al-
r2 (T ) should change their sign at the same ways of first order, so there is no contradic-
temperature. Thus, if F1 , F2 are “primary” tion with this Landau rule, except for the
order parameter components, they should transition to a charge density wave state in
appear via a first-order transition. 2H-TaSe2 (Moncton et al., 1977) which ap-
The second condition essentially implies parently is second order although there is a
the absence of third-order terms (F 3 ) in the third-order invariant (Bak and Mukamel,
Landau expansion. It turns out, however, 1979). It is possible, of course, that the
that in d = 2 dimensions there are well- transition is very weakly first order so that
known counter-examples to these rules, experimentally it could not be distin-
namely the Potts model (Potts, 1952) with guished from second order. This remark
q = 3 and q = 4 (see Eq. (4-20)); as shown also holds for other examples of apparent
4.2 Phenomenological Concepts 271
free-energy functional does not attain its which implies a specific heat divergence
minimum at wavevector q = 0 in reciprocal C ~ (– t)–1/2, i.e., a t = 1/2. We note that this
space but on a surface given by the equa- set of exponents also satisfies the scaling
tion | q | = q* = 2 p/l , l being the wave- laws in Eqs. (4-78), (4-81), (4-83), and
length of the lamellar pattern. This reflects (4-84). Moreover, using this set of expo-
the degeneracy that there is no preferred di- nents in the Ginzburg criterion, Eqs. (4-58)
rection and hence the lamellae may be to (4-61), we find that, with the Landau ex-
oriented in any direction. Owing to the ponents for a tricritical point, this condi-
large “phase space” in q-space where the tion is marginally fulfilled, unlike the case
inverse of the wavevector-dependent sus- of ordinary critical points (Eq. (4-62)). It
ceptibility goes “soft” (c –1 ( | q | = q*) Æ 0 turns out that d* = 3 is the marginal dimen-
as T Æ Tc ), the mean-square amplitude of sion for tricritical points, and the mean-
the local fluctuation of the order parameter field power laws are modified by logarith-
·F 2 Ò – ·F Ò2 would diverge as T Æ Tc , mic correction terms similar to those noted
which is physically impossible. It can be for dipolar systems (Eq. (4-72)). Experi-
shown that this difficulty is avoided be- mental examples for tricritical points in-
cause the strong local order parameter fluc- clude strongly anisotropic antiferromagnets
tuations turn u negative near Tc (Brazov- in a uniform magnetic field (e.g., FeCl2 ,
skii, 1975; Fredrickson and Helfand, 1987; (Dillon et al., 1978), and dysprosium alu-
Fredrickson and Binder, 1989), producing minum garnet (DAG) (Giordano and Wolf,
a free energy, as shown in Fig. 4-6 b, for the 1975), see Fig. 4-9 for schematic phase
case of “symmetric” diblock copolymers diagrams), systems undergoing structural
where mean field theory would predict a phase transitions under suitable applied pres-
second-order transition. Experiments seem sure, such as NH 4Cl which has a tricritical
to confirm the first-order character of the point at Tt = 250 K and pt = 128 bar (Yelon et
transition (Bates et al., 1988). al., 1974) or the ferroelectric KDP at Tt =113
For certain systems, for which u > 0 and K and pt = 2.4 kbar (Schmidt, 1978), etc. A
thus a second-order transition occurs, it is model system which has been particularly
possible by variation of a non-ordering carefully studied is He3 –He4 mixtures: the
field to change the sign of the coefficient u transition temperature Tl (x) of the normal
at a particular point. This temperature Tt , fluid–superfluid He4 is depressed with in-
where u vanishes in Eq. (4-14), is called a creasing relative concentration x until, at a
tricritical point. From Eq. (4-14) we then tricritical point Tt = Tl (x t ), the transition be-
immediately find comes first order. This then implies a phase
(4-86)
separation between a superfluid phase with
F0 = (– r /v)1/4 = (r ¢/v)1/4 (Tt /T – 1)1/4 a lower He3 content and a normal fluid He3-
Hence the tricritical order parameter expo- rich phase. Most common, of course, are
nent b t = 1/4, while Eqs. (4-32) to (4-48) re- tricritical phenomena in fluid binary mix-
main unchanged and thus g t = 1, n t = 1/2, tures arising from the competition of gas –
h t = 0. The critical isotherm, however, be- liquid transitions and fluid – fluid phase
comes v F 05 = H/kB T, i.e., d t = 5. From Eq. separations in these systems (Scott, 1987).
(4-15) we finally find Just as the vanishing coefficient u in Eq.
(4-10) leads to a multicritical point, the tri-
F (F0 ) − F ( 0 ) t − t 1/ 2
=− ⎛ ⎞
critical point, another multicritical point is
(4-87)
V kB T 3⎝ v ⎠ associated with the vanishing of the coeffi-
4.2 Phenomenological Concepts 273
ed to magnetic systems, and many exam- such as pressure, magnetic field (for mag-
ples of dielectric incommensurate phases netic structures), concentration (for alloys;
have been identified (e.g., NaNO2 (Tani- see Fig. 4-10), etc.). Theoretically under
saki, 1961), Rb2ZnCl 4 , BaMnF4 , thiourea certain conditions even a staircase with an
[SC(NH2 )2 ], see Tolédano and Tolédano infinite number (mostly extremely small!)
(1987)). Whereas Rb2ZnCl 4 and BaMnF4 of steps can be expected (“devil’s stair-
also involve the existence of Lifshitz invar- case”, see Selke (1988, 1989, 1992) for a
iants, thiourea is an example of the case discussion and further references). We also
discussed in Eqs. (4-88) to (4-90). Various emphasize that the modulation need not in-
examples of modulated superstructures ex- volve only one direction in space (one
ist in metallic alloys; see Selke (1988, wavevector), but can involve several wave-
1989, 1992) and De Fontaine and Kulik
(1985) for reviews of the pertinent theory
and experimental examples such as Al3Ti
(Loiseau et al., 1985) and Cu3Pd (Broddin
et al., 1986).
Whereas for TeTc the ordering of the
incommensurate phase can be described in
terms of a sinusoidal variation of the local
order parameter density F (x), the nonlin-
ear terms present in Eq. (4-10) at lower
temperatures imply that higher-order har-
monics become increasingly important.
Rather than a sinusoidal variation, the
structure is then better described in terms
of a periodic pattern of domain walls (or
“solitons”), and we talk about a “multi-
soliton lattice” or “soliton staircase”. An-
other important fact about modulated
phases is that the wavevector characteriz-
ing the modulation period is not fixed at
qmax , but varies with temperature or other
parameters of the problem. This is ex-
pected from the third of the general Landau
rules formulated at the beginning of Sec.
4.2.3. Usually this variation of q stops at Figure 4-10. Period M (in units of the lattice spacing
some commensurate value where a “lock- in the modulation direction) for the alloy Al3 – xTi1+ x
as a function of the annealing temperature, as ob-
in transition” of the incommensurate struc- tained from visual inspection of high-resolution elec-
tures occurs. In the incommensurate re- tron microscopic images. Each of the fifteen different
gime rational values of the modulation commensurate superlattices observed is composed of
(“long-period superstructures”) may have antiphase domains of length one or two (1-bands or
an extended regime of stability, leading to 2-bands, as indicated by sequences ·211Ò, ·21Ò etc.)
based on the L12 structure. Very long annealing times
an irregular staircase-like behavior of the were needed to produce these long-period super-
modulation wavelength l as a function of structures supposedly at equilibrium. From Loiseau
temperature (or other control parameters et al. (1985).
4.2 Phenomenological Concepts 275
vectors. An example of the case of a two- While experimental examples for Lif-
rather than one-dimensional modulation is shitz points are extremely scarce, e.g., the
LiKSO4 (Pimenta et al., 1989). structural transition in RbCaF3 under (100)
Here we discuss only the special case stress (Buzaré et al., 1979), more common
of Eqs. (4-88) – (4-90) briefly where, by multicritical points are bicritical points
changing external control parameters, it where two different second-order transi-
is possible to reach a multicritical point tion lines meet. Consider, for example, the
where K1 vanishes, the so-called “Lifshitz generalization of Eq. (4-10) to an n-compo-
point” (Hornreich et al., 1975). From Eq. nent order-parameter field, but with differ-
(4-88) we then find that for T = Tc (= TL ) we ent coefficient ri for each order-parameter
have c (q) ~ q – 4, and comparing this with component F i :
the result following from a Fourier trans-
1 F
formation of Eq. (4-47), c (q)Tc ~ q – 2 + h , Ᏺ [F ( x )] = 0 (4-92 )
we conclude that h L = – 2 for the Lifshitz kB T kB T
+ ∫ dx ⎧⎨ r1 F12 ( x ) + r2 F22 ( x ) + …
point, whereas for T > Tc we still have 1 1
c (q = 0) = Ĝ0 t0–1, i.e., g L = 1. Since for ⎩ 2 2
K1 = 0 Eq. (4-88) can be rewritten as 1 u
+ rn Fn2 ( x ) + [F 2 ( x )]2
2 4
c (q) = Ĝ0 (1 – TL /T )–1/ [1 + x 4 q 4 ] ⎫
R2
x = [K2 /(1 – TL /T )]1/4 (4-91) + [(∇F1 )2 + … + (∇Fn )2 ]⎬
2d ⎭
we conclude that nL = 1/4 for the correla- For simplicity, the fourth-order term has
tion length exponent at the Lifshitz point. been taken as fully isotropic. If r1 = r2 = …
Also for the Lifshitz point we can ask = rn , we would have the isotropic n-vector
whether the Landau description presented model discussed in Secs. 4.2.2 and 4.2.3.
here is accurate, and whether or not statis- We now consider the case where some of
tical fluctuations modify the picture. The these coefficients differ from each other,
result is that the critical fluctuations are but a parameter p exists on which these co-
very important. A marginal dimensionality efficients depend in addition to the temper-
d* = 8 results in the case of an “isotropic ature. The nature of the ordering will be
Lifshitz point”, where in determined by the term ri F i2 , for which
K1x (∂F /∂x)2 + K1y (∂F /∂y)2 + K1z (∂F /∂z)2 the coefficient ri changes sign at the high-
est temperature. Suppose this is the case
we have simultaneous vanishing of all co- where i = 1 for p < pb , then we have a one-
efficients K1x , K1y , and K1z . Note that the component Ising-type transition at Tc1 ( p)
more common case is the “uniaxial Lifshitz given by r1 ( p, T ) = 0 (the other compo-
point”, where only K1z vanishes while K1x nents F i being “secondary order parame-
and K1y remain nonzero. In this case the ters” in this phase). If, however, for p > pb
multicritical behavior is anisotropic, and the coefficient r2 ( p, T ) = 0 at the highest
we must distinguish between the correla- temperature Tc2 ( p), it is the component F 2
tion length x || describing the correlation which drives the transition as the primary
function decay in the z-direction and the order parameter. The point p = pb , Tc1 ( p) =
correlation length x ^ in the other direc- Tc2 ( p) = Tb is then called a bicritical point.
tions. Eq. (4-91) then holds for x || only, An example of this behavior is found for
whereas Eq. (4-39) still holds for x ^. weakly uniaxial antiferromagnets in a uni-
276 4 Statistical Theories of Phase Transitions
form magnetic field H|| (Fig. 4-9). A model order parameter, due to the perpendicu-
Hamiltonian for such a system can be writ- lar components Msx and Msy of the stag-
ten as gered magnetization). In the framework of
Eq. (4-92), this would mean n = 3 and
Ᏼ = − ∑ {J (1 − D ) [ Six S jx + Siy S jy ]
〈 i, j 〉
r2 ∫ r3 ≠ r1 , with H|| being the parameter p.
+ J Siz S jz } − H|| ∑ Siz (4-93) Both lines TN|| (H|| ) and TN^ (H|| ) and the
i
first-order line between the two antiferro-
where we assume J < 0 (antiferromagne- magnetic structures join in a bicritical
tism) and for D > 0 the easy axis is the z- point. A well known example of such a be-
axis. For small H|| , we have a uniaxial anti- havior is GdAlO3 with Tb = 3.1242 K (Roh-
ferromagnetic structure and the order pa- rer, 1975). Another example, in our opin-
rameter is the z-component of the stag- ion, is the Fe – Al system where the ferro-
gered magnetization Msz . For stronger paramagnetic critical line joins the second-
fields H|| , however, we have a transition order A2 – B2 transition (Fig. 4-11). In the
to a spin-flop structure (two-component metallurgical literature, the magnetic tran-
Figure 4-11. The Fe – Al phase diagram, as obtained from a mean-field calculation by Semenovskaya (1974)
(left), and experimentally by Swann et al. (1972) (right). The ferro-paramagnetic transition is shown by the
dash-dotted line. The crystallographically disordered (A2) phase is denoted as a n in the nonmagnetic and a m in
the ferromagnetic state. The ordered FeAl phase having the B2 structure (Fig. 4-3) is denoted as a 2 , and the or-
dered Fe3Al phase having the D03 structure (Fig. 4-3) is denoted as a 1n or a 1m , depending on whether it is non-
magnetic or ferromagnetically ordered, respectively. Note that first-order transitions in this phase diagram are
associated with two-phase regions, since the abscissa variable is the density of an extensive variable, unlike the
ordinate variable in Fig. 4-9. From Semenovskaya (1974).
4.2 Phenomenological Concepts 277
We do not give a detailed account of crit- to establish its precise value (Wansleben
ical dynamics here, but discuss only the and Landau, 1987).
simplest phenomenological approach (van It is important to realize that not all fluc-
Hove, 1954) for non-conserved order pa- tuations slow down as a critical point is ap-
rameters. We consider a deviation of the or- proached, but only those associated with
der parameter DF (x, t) from equilibrium at long-wavelength order parameter varia-
space x and time t, and ask how this devia- tions. This clearly expressed in terms of the
tion decays back to equilibrium. The stan- dynamic scaling relationship of the charac-
dard assumption of irreversible thermody- teristic frequency:
namics is a relaxation assumption; we put
w (q) = q z w̃ ( ᒗ), ᒗ = qx (4-99)
∂ ∂Ᏺ
DF ( x, t ) = − G0 (4-95) where w̃ (ᒗ) is a scaling function with the
∂t ∂DF ( x ) properties w̃ (•) = const. and w̃ (ᒗ O 1) ~ ᒗ– z ,
where G0 is a phenomenological rate factor. and, hence, is consistent with Eq. (4-94).
Using Eq. (4-10) and linearizing F (x, t) = Most experimental evidence on dynamic
F0 + DF (x, t) around F0 yields critical phenomena comes from methods
such as inelastic scattering of neutrons or
∂
∂t
{
DF ( x, t ) = − G0 ( r + 3 u F02 ) DF ( x, t ) light, NMR or ESR spectroscopy, and
ultrasonic attenuation. Although the over-
R2 2 ⎫ all agreement between theory and experi-
− ∇ [ DF ( x, t )]⎬ (4-96) ment is satisfactory, only a small fraction
d ⎭
of the various theoretical predictions have
Assuming that DF (x, t) is produced by been thoroughly tested so far. For reviews,
a field H (x) = Hq exp (i q · x) which is see Hohenberg and Halperin (1977) and
switched off at time t = 0, we obtain (see Enz (1979).
also Sec. 4.2.2) Finally, we also draw attention to nonlin-
(4-97) ear critical relaxation. Consider, for exam-
DFq (t ) ple, an ordering alloy which is held well
= exp [ − w ( q ) t ], w ( q ) = G0 / c ( q )
DFq ( 0 ) below Tc in the ordered phase and assume
that it is suddenly heated to T = Tc . The or-
where Eq. (4-34) has been used. Since c (q) der parameter is then expected to decay to-
diverges for q = 0, T Æ Tc , the characteris- wards zero with time in a singular fashion,
tic frequency w (q = 0) ∫ t –1 vanishes as F (t) ~ t – b /n z (Fisher and Racz, 1976). Sim-
w (q = 0) ~ c T–1 ~ x –g /n = x – (2 – h ), and by ilarly, if the alloy is quenched from a tem-
comparison with Eq. (4-94) the result of perature T > Tc to T = Tc , the superstructure
the “conventional theory” is obtained: peak appears and grows in a singular fash-
z=2–h (4-98) ion with the time t after the quench,
S (q, t) ~ t g /zn (Sadiq and Binder, 1984). Re-
Although Eq. (4-98) suggests that there is a lated predictions also exist for the nonlin-
scaling relationship linking the dynamic to ear relaxation of other quantities (ordering
the static exponents, this is not true in gen- energy, electrical resistivity) (Binder and
eral if effects due to critical fluctuations are Stauffer, 1976 a; Sadiq and Binder, 1984).
taken into account. In fact, for noncon- A very important topic is the dynamics
served systems, z slightly exceeds 2 – h and of first-order phase transitions, which will
extensive Monte Carlo studies were needed not be discussed here because it is dis-
4.2 Phenomenological Concepts 279
cussed elsewhere in this book (see Chap- rameter space of the corresponding multi-
ters by Binder and Fratzl (2001) and by component system show up like a rounding
Wagner et al. (2001)). of the first-order transition.
In solids, however, diffusion of impurity
atoms is often negligibly small and such a
4.2.5 Effects of Surfaces and of Quenched
disorder due to frozen-in, immobile impur-
Disorder on Phase Transitions:
ities is called “quenched disorder” (Brout,
A Brief Overview
1959). Other examples of quenched disor-
It a system contains some impurity at- der in solids are due to vacancies, in addi-
oms which are mobile, we refer to “an- tion to extended defects such as disloca-
nealed disorder” (Brout, 1959). From the tions, grain boundaries, and external sur-
point of view of statistical thermodynam- faces. Quenched disorder has a drastic effect
ics, such mobile impurities act like addi- on phase transitions, as will be seen below.
tional components constituting the system Since the disorder in a system is usually
under consideration. For a second-order assumed to be random (irregular arrange-
transition, small fractions of such addi- ment of frozen-in impurities, for example),
tional components have rather minor ef- we wish to take an average of this random
fects: the transition point may be slightly disorder. We denote this averaging over the
shifted relative to that of the pure material, disorder variables, which we formally de-
and as the concentration of these impurities note here as the set {xa }, by the symbol
is normally strictly conserved, in principle […]av in order to distinguish it from the
the so-called “Fisher renormalization” thermal averaging ·…ÒT implied by statisti-
(Fisher, 1968) of critical exponents is ex- cal mechanics. Thus the average free en-
pected, as discussed in Sec. 4.2.2. For ergy which needs to be calculated in the
small impurity contents, the region around presence of quenched disorder is
the transition point where this happens is
extremely narrow, and hence this effect is F = − kB T [ln Z {xa}]av (4-100 )
not important.
The effect of such impurities on first-or-
[ ( {F }
= − kB T ln tr exp( − Ᏼ {xa ,Fi }/ kB T )
i
) ]av
der transitions is usually more important, whereas in the case of annealed disorder it
e.g., for a pure one-component system, the would be the partition function rather than
melting transition occurs at one well-de- the free energy which is averaged over the
fined melting temperature, whereas in the disorder:
presence of mobile impurities this transi-
Fann = − kB T ln[ Z {xa}]av (4-101)
tion splits up into two points, correspond-
ing to the “solidus line” and “liquidus line”
in the phase diagram of a two-component
( {F } i
)
= − kB T ln tr [exp( − Ᏼ {xa ,Fi }/ kB T )]av
metry, off-center positions may produce for a review). Of course, the impurities will
both a random electric field and a random always produce some shift of the critical
strain field. Also, symmetry generally per- temperature, which decreases as p in Eq.
mits random fields to act on tensorial order (4-106) decreases. When p becomes
parameters in diluted molecular crystals smaller than a critical threshold value pc ,
(Harris and Meyer, 1985). For diluted mo- only finite clusters of spins are still mutu-
lecular crystals (such as N2 diluted with Ar, ally connected by nonzero exchange bonds
or KCN diluted with KBr, etc.) it is still un- J > 0, and long-range order is no longer
clear whether the resulting “orientational possible. The phase transition at T = 0 pro-
glass” phase is due to these random fields duced by variation of p in Eq. (4-106) can
or to random bonds (for reviews, see Loidl hence be interpreted purely geometrically
(1989); Höchli et al. (1990); Binder and in terms of the connectivity of finite clus-
Reger (1992); Binder (1998)). ters or an infinite “percolating” network of
The “random bond” Hamiltonian differs spins (Stauffer and Aharony, 1992). This
from Eq. (4-102) by the introduction of dis- transition is again described by a com-
order into the exchange terms rather than pletely different set of exponents.
–
the “Zeeman energy”-type term, For the case where | J | O D J in Eq.
(4-105), a new type of ordering occurs,
Ᏼ = − ∑ Jij Si ⋅ S j − H ∑ Siz (4-104) which is not possible in systems without
i≠ j i
quenched disorder; a transition occurs to a
Whereas in Eq. (4-102) we have assumed state without conventional ferro- or antifer-
the pairwise energy to be translationally in- romagnetic long-range order but to a “spin
variant (the exchange energy J depends glass phase” where the spins are frozen-in
only on the distance ai – aj between the in a random direction (Binder and Young,
spins but not on their lattice vectors ai and 1986; Young, 1998). The order parameter
aj separately), we now assume Ji j to be a of the Ising spin glass was introduced by
random variable, e.g., distributed accord- Edwards and Anderson (1975) as
ing to a Gaussian distribution
– qEA = [·Si Ò2T ] av (4-107)
P (Ji j ) µ exp [– (Ji j – J )2/2 (DJ)2 ] (4-105)
The nature of the phase transition in spin
or according to a two-delta function distri- glasses and the properties of the ordered
bution phase have been the subject of intense re-
P (Ji j ) = p d (Ji j – J) + (1 – p) d (Ji j ) (4-106) search (Binder and Young, 1986). This
great interest in spin glasses can be under-
–
If J o DJ (Eq. (4-105)) or if 1 – p O 1 (Eq. stood because many different systems show
(4-106)), the ferromagnetic order occurring spin glass behavior: transition metals with
–
for Eq. (4-104) if J > 0 or J > 0 is only a small content x of magnetic impurities
weakly disturbed, both for d = 3 and for such as Au1– x Fex and Cu1– x Mnx , diluted
d = 2 dimensions. Following an argument insulating magnets such as Eux Sr1– x S, var-
presented by Harris (1974), we can see that ious amorphous alloys, and also mixed
the critical behavior of the system remains dielectric materials such as mixtures of
unaltered in the presence of such impurities RbH 2PO4 and NH 4 H 2PO4 , where the spin
provided that the specific heat exponent represents an electric rather than magnetic
a < 0, whereas a new type of critical behav- dipole moment. A related random ordering
ior occurs for a ≥ 0 (see Grinstein (1985) of quadrupole moments rather than dipole
282 4 Statistical Theories of Phase Transitions
moments if found in “quadrupolar glasses” glasses) the critical dynamics again have a
(also called “orientational glasses”) such very different character, and dynamics
as K(CN)1– x Brx and (N2)x Ar1– x , (Loidl, characteristic of thermally activated pro-
1989; Höchli et al., 1990; Binder and cesses with a broad spectrum of relaxation
Reger, 1992; Binder, 1998). However, it times are often observed.
is still debated (Franz and Parisi, 1998) It is also interesting to discuss the effect
whether or not concepts appropriate for the of quenched impurities on first-order tran-
description of spin glasses are also useful sitions (Imry and Wortis, 1979). It is found
for the structural glass transition of under- that typically “precursor phenomena” near
cooled liquids (Jäckle, 1986). the first-order transition are induced, and
Finally we mention systems with ran- the latent heat associated with the transi-
dom anisotropic axes which can be mod- tion in the pure systems can be signifi-
eled by the following Hamiltonian (Harris cantly reduced. It is also possible that such
et al., 1973): impurity effects may completely remove
(4-108) the latent heat discontinuity and lead to a
Ᏼ = − ∑ J ( ai − a j ) Si ⋅ S j − ∑ ( ei ⋅ S j )2 rounding of the phase transition.
i≠ j i
Such rounding effects on phase transi-
where the ei are vectors whose components tions also occur when extended defects
are independent random variables with a such as grain boundaries and surfaces are
Gaussian distribution. This model is also considered. These disrupt long-range order
believed to exhibit destruction of long- because the system is then approximately
range order due to break-up in domains homogeneous and ideal (i.e., defect-free)
similar to the random field systems. It is only over a finite region in space. While
also suggested that spin glass phases may the description of the rounding and shifting
occur in these systems. Again Eq. (4-108) of phase transitions due to finite size ef-
is expected to be relevant, not only for dis- fects has been elaborated theoretically in
ordered magnetic materials but also for detail (Privman, 1990), only a few cases
dielectrics where the spin represents elec- exist where the theory has been tested ex-
tric dipole moments, or displacement vec- perimentally, such as the normal fluid –
tors of atoms at structural phase transitions, superfluid transitions of He4 confined to
etc. pores, or the melting transition of oxygen
Very small fractions of quenched impur- monolayers adsorbed on grafoil where the
ities which do not yet have an appreciable substrate is homogeneous over linear di-
effect on the static critical properties of a mension L of the order of 100 Å (see Marx,
second-order phase transition can already 1989). While at critical points, a rounding
affect the critical dynamics drastically. An and shifting of the transition normally set
example of this behavior is the occurrence in when the linear dimension L and the cor-
of impurity-induced “central peaks” for relation length x are comparable (Fisher,
structural phase transitions in the scattering 1971; Binder, 1987b), at first-order transi-
function S (q, w ) at frequency w = 0 in addi- tions the temperature region DT over
tion to the (damped) soft phonon peaks which the transition is rounded and shifted
(Halperin and Varma, 1976). In the above is inversely proportional to the volume of
systems where the impurities disrupt the the system (Imry, 1980; Challa et al., 1986).
conventional ordering more drastically An important effect of extended defects
(such as random field systems and spin such as grain boundaries and surfaces is
4.3 Computational Methods Dealing with Statistical Mechanics 283
that they often induce precursor effects to Transitions occur where the basic degree of
phase transitions, e.g., the “wetting transi- freedom is the (thermally activated) diffu-
tion” (Dietrich, 1988; Sullivan and Telo da sion process of atoms between various lat-
Gama, 1985) where a fluid layer condens- tice sites. This happens for unmixing alloys
ing at a surface is a precursor phenomenon such as Al – Zn or for ordering alloys such
to gas – fluid condensation in the three- as b-CuZn or the Cu –Au system. The mod-
dimensional bulk volume. Similarly “sur- eling of such systems will be discussed in
face melting” and “grain-boundary melt- Sec. 4.3.1. Many structural transitions are
ing” phenomena can be interpreted as the of a very different nature: we encounter pe-
intrusion of a precursor fluid layer at a sur- riodic lattice distortions where atomic dis-
face (or grain boundary) of a crystal (Di placements are comparable to those of lat-
Tolla et al., 1996). Similarly, at surfaces of tice vibrations. Short wavelength distor-
ordering alloys such as Cu3Au, the effect tions may give rise to “antiferrodistortive”
of the “missing neighbors” may destabilize and “antiferroelectric” ordering, as exem-
the ordering to the extent that “surface-in- plified by the perovskites SrTiO3 , PbZrO3 ,
duced disordering” occurs when the system etc. Long wavelength distortions corre-
approaches the transition temperature of sponding to acoustic phonons give rise to
the bulk (Lipowsky, 1984). Finally, we “ferroelastic” ordering. Sec. 4.3.2 will deal
draw attention to the fact that the local crit- with the mean field treatment of such phase
ical behavior at surfaces differs signifi- transitions where the order parameter is a
cantly from the critical behavior in the bulk phonon normal coordinate, and Sec. 4.3.3
(Binder, 1983 b). Such effects are outside is devoted to numerical methods going be-
the scope of this chapter. yond mean-field theory.
where vAA , vAB , and vBB are interactions where Dm is the chemical potential differ-
between pairs of AA, AB, and BB atoms. ence between the two species.
In fact, terms involving three- and four- The same mapping applies for the lat-
body interactions may also occur, but are tice-gas model of fluids, which at the same
not considered here. Also, effects due to time can be considered as a model of ad-
vacancies may easily be included but are sorbed layers on crystalline surface sub-
neglected here. strates (in d = 2 dimensions) (see Binder
As is well known, Eq. (4-109) can be re- and Landau (1989) for a review) or as a
duced to the Ising model, Eq. (4-56), by the model of interstitial alloys, such as hydro-
transformation Si = 1 – 2 ci = ± 1, apart from gen or light atoms such as C, N and O in
a constant term which is of no interest to us metals (Alefeld, 1969; Wagner and Horner,
here. The “exchange interaction” Jij 1974; Alefeld and Völkl, 1978). If we
between spins i and j and “magnetic field” interpret B in Eqs. (4-109) – (4-111) as an
H in Eq. (4-56) are related to the interac- occupied site and A as a vacant site, then
tion parameters of Eq. (4-109) by usually vAB = vAA= 0, such that
(4-112)
J (xi – xj) ∫ Jij = [2 vAB (xi – xj ) (4-110) Ᏼ = Ᏼ 0 + ∑ ci c j v ( xi − x j ) + e ∑ ci
– vAA (xi – xj ) – vBB (xi – xj )]/4 i≠ j i
Ᏼ = Ᏼ0 ( 4-115)
+ ∑ {ci c j [v BB ( xi − x j ) − Jm ( xi − x j ) s i ⋅ s j ]
i≠ j
+ 2 ci (1 − c j ) v AB ( xi − x j )
+ (1 − ci ) (1 − c j ) v AA ( xi − x j )} + …
where Jm (xi – xj) denotes the strength of Figure 4-13. Partial phase diagram of copper – gold
the magnetic interaction, which we have alloys in the temperature – concentration plane indi-
assumed to be of the same type as used in cating the existence regions of the three ordered
the Heisenberg model (Eq. (4-102)) with phases Cu3Au, CuAu, and CuAu3 (cf. Fig. 4-8 a and
n = 3 and Hi = 0. c). These phases are separated from each other (and
from the disordered phase occurring at higher tem-
We will not go further into the classifica- peratures) by two-phase coexistence regions. The
tion of the various models here, but note boundaries of these regions are indicated by full and
one general property of both models in broken lines. Region II is a long-period modulated
Eqs. (4-109) and (4-112) which becomes version of the simple CuAu structure occurring in re-
evident when mapping the Ising Hamilton- gion I. Note that for strictly pairwise constant inter-
actions (of arbitrary range!) in a model such as Eq.
ian Eq. (4-56): for H = 0, this Hamiltonian (4-109) the phase diagram should have mirror sym-
is invariant against a change of sign of all metry around the line cAu = 50%. From Hansen
spins. For H ≠ 0, it is invariant against the (1958).
286 4 Statistical Theories of Phase Transitions
tion c (q) describing the response to a (cf. Sec. 4.2.2); a more accurate procedure
wavevector-dependent field (cf. Eqs. (4-32), to deduce J (x) from experimental data on
(4-34), and (4-40)), if we apply the molec- a (x) is the “inverse Monte Carlo method”
ular field approximation (see the next sub- (Gerold and Kern, 1986, 1987; Schweika,
section). The result is (Brout, 1965): 1989), but this method also relies on the as-
(4-118)
sumption that an Ising model description as
1− M 2
S ( q ) = kB T c ( q ) = written in Eq. (4-109) is appropriate. On
1 − [ J ( q )/ kB T ] (1 − M 2 )
˜
the level of the molecular field approxima-
where J̃ (q) is the Fourier transform of the tion (for alloys this is usually referred to as
“exchange interaction” J (x), the Bragg – Williams (1934) approxima-
(4-119) tion) or the Bethe (1935) approximation, it
˜J ( q ) = ∑ J ( xi − x j ) exp [ i q ⋅ ( xi − x j )] is possible to avoid models of the type of
j (≠ i) Eq. (4-109) and include the configurational
degrees of freedom in an electronic struc-
In Eq. (4-118), we have once again invoked
ture calculation (Kittler and Falicov, 1978,
that there is a “fluctuation relation”
1979). Although such an approach sounds
(= static limit of the so-called “fluctuation-
very attractive in principle, the results are
dissipation theorem”) relating c (q) to the
not so encouraging in practice, as shown in
structure factor S (q) which is just the
Fig. 4-15. The results of this method for
Fourier transform of the correlations ap-
Cu3Au are compared with Monte Carlo data
pearing in Eq. (4-117):
(4-120) of a nearest neighbor Ising model (Binder,
S (q) = ∑ exp [ i q ⋅ ( ri − rj )] ( 〈 Si S j 〉 − M 2 ) 1980), with the cluster-variation treatment
j (≠ i) of the same model (Golosov et al., 1973)
and with experimental data (Keating and
Combining now Eqs. (4-118) and (4-120)
Warren, 1951; Moss, 1964; Orr, 1960).
we see that the reciprocal of this diffuse
The conclusion of this subsection is that
scattering intensity S (q) in q-space is sim-
the development of microscopic models for
ply related to the Fourier transformation of
the description of order – disorder phenom-
the interactions as
ena in alloys is still an active area of re-
4 c (1 − c ) 4 J˜ ( q ) search, and is a complicated matter, be-
=1− c (1 − c ) (4-121) cause the validity of the models can only be
S (q) kB T
judged by comparing results drawn from
For the model defined in Eqs. (4-109) to the models with experimental data. How-
(4-111), this expression even is exact in ever, these results also depend on the ap-
an expansion of 1/S (q) in a power series in proximation involved in the statistical-me-
1/T to leading order in 1/T, and is one of chanical treatment of the models (e.g., the
the standard tools for inferring information full and broken curves in Fig. 4-15 refer to
on interactions in alloys from diffuse scat- the same nearest-neighbor Ising model,
tering data (Clapp and Moss, 1966, 1968; whereas the dash-dotted curve refers to a
Moss and Clapp, 1968; Krivoglaz, 1969; different model). In the next subsections,
Schweika, 1994). An example is shown we consider various sophistications of the
in Fig. 4-14. Close to the order – disorder statistical mechanics as applied to various
phase transition, we expect corrections to models for phase-transition phenomena.
the molecular field expression Eq. (4-121) More details about all these problems can
due to the effects of statistical fluctuations be found in the Chapter by Inden (2001).
288 4 Statistical Theories of Phase Transitions
The MFA can now be defined by factor- These probabilities can be expressed in
izing the probability P ({Si}) of a spin con- terms of the multi-spin correlation func-
figuration of the whole lattice into a prod- tions gnc (i) ∫ ·Si Sj 1 … Sjn Ò, where the set of
uct of single-site probabilities pi which vectors xj 1 – xi , …, xjn – xi defines the n-
can take two values: p+ = (1 + M )/2 is the point cluster of type c located at lattice site
probability that the spin at site i is up i. The Helmholtz energy functional to be
and p– = (1 – M )/2 is the probability that it minimized in this cluster variation method
is down, p+ – p– = M is the magnetization. (Kikuchi, 1951; Sanchez and De Fontaine,
Now the expression Jij Si Sj pi pj (cf. Eq. 1980, 1982; Finel, 1994) is a more compli-
(4-56)) summed over the possible values cated approximation of Eq. (4-122) than
p+ and p– simply yields Jij M 2, and hence Eq. (4-123). If the largest cluster consid-
Eq. (4-122) reduces to, using Eq. (4-119) ered exceeds the interaction range, the en-
ergy term in F = U – TS is treated exactly;
1 MFA 1 ˜
Ᏺ = J (q = 0) M 2 − H M (4-123) unlike Eq. (4-123), the entropy is approxi-
N 2 mated. We find
⎡1 + M ⎛ 1 + M ⎞
+ kB T ⎢ ln
⎣ 2 ⎝ 2 ⎠ 1
Ᏺ= ∑ ∑ Jij g2, rj (i ) + kB T ∑ ∑ g nc
1 − M ⎛1 − M⎞⎤ 2 i j i n, c
+ ln
2 ⎝ 2 ⎠ ⎥⎦ 2n
× ∑ pn, c ( k , i ) (4-125)
Minimizing Ᏺ MFA
with respect to M now k =1
yields the elementary self-consistency where the coefficients gnc are combinato-
equation: rial factors depending on the lattice geome-
1 ˜ try and the clusters included in the approx-
M = tanh [ J (q = 0) M + H ] (4-124) imation (Kikuchi, 1951), e.g., in the tetra-
kB T
hedron approximation for the f.c.c. lattice,
As is well known, Eq. (4-124) implies a the sum over c in Eq. (4-125) includes the
second-order transition at Tc ∫ J̃ (q = 0)/kB (nearest-neighbor) tetrahedron, the near-
with the same exponents as in the Landau est-neighbor triangle, the nearest-neighbor
theory. pair, and the single site.
Clearly, in factorizing P ({Si}) into a Assuming the ordered structure to be
product of single-site probabilities and known, the symmetry operations of the as-
solving only an effective single-site prob- sociated group can be applied to reduce
lem, we have disregarded correlation in the the number of variational parameters in
probabilities of different sites. A system- Eq. (4-125). In the MFA, there is a single
atic improvement is obtained if we approx- non-linear self-consistent equation (Eq.
imate the probability of configurations not (4-124)) or a set of equations involving the
just by single-point probabilities but by us- order-parameter components if a problem
ing “cluster probabilities”. We consider more complicated than the Ising ferromag-
probabilities pnc (k, i) that a configuration k net is considered. In the CV method, a
of the n spins in a cluster of geometric con- much larger set of coupled non-linear
figuration c occurs (c may be a nearest- equations involving the short-range order
neighbor pair, or a triangle, tetrahedron, etc.). parameters gnc (i) is obtained when we min-
Note k = 1, … , 2n for Ising spins whereas imize Eq. (4-125). Therefore, whereas the
k = 1, … , qn for the q-state Potts model. simple MFA is still manageable for a wide
290 4 Statistical Theories of Phase Transitions
dered structures (De Fontaine, 1975, 1979; of an order-parameter component for the
Khachaturyan, 1973, 1983). transition: in mean-field theory, the asso-
This formulation of order – disorder tran- ciate eigenfrequency vanishes at a temper-
sitions in alloys is analogous to the treat- ature Tc (“soft phonon”). If this happens for
ment of structural transitions, for which a a phonon with wavevector k0 at the Bril-
model Hamiltonian similar to Eq. (4-127) louin zone edge, we have an antiferro-
can be written: rather than concentration electric order, provided that the phonon is
deviations xi we now have displacement polar, i.e., it produces a local dipole mo-
vectors ul (x) associated with a lattice vec- ment. For non-polar phonons, such as for
tor x for the l th atomic species in the unit the transition in SrTiO3 at Tc = 106 K where
cell. Just as it is useful to relate xi to the k0 = p (–12 , –12 , –12 )/a (the soft phonon there
Fourier transform X (q) of the concentra- physically corresponds to an antiphase ro-
tion deviation, it is useful to relate ul (x) to tation of neighboring TiO6 octahedra, as
the phonon normal coordinate Qk, l , de- indicated schematically in Fig. 4-12), the
fined as transition leads to “antiferrodistortive” or-
(4-134) der. Long-wavelength distortions corre-
1
ul ( x ) = ∑ exp( i k ⋅ x ) el ( k , l ) Qk, l sponding to optical phonons give rise
N Ml k, l
to ferroelectric ordering (an example is
where Ml is the mass of the atom of type l Pb5Ge3O11, where k0 = 0 and Tc = 450 K
at site Rli in the i th unit cell, el (k, x) is a (Gebhardt and Krey, 1979)). There are also
phonon polarization vector, l labels the paraelectric – ferroelectric transitions of
phonon branch and k its wavevector (Fig. first order, e.g., the cubic – tetragonal tran-
4-16). In this case, ·Qk0 , l0 ÒT plays the role sition of BaTiO3, such that no soft mode
occurs.
Just as the “macroscopic” ferroelectric
ordering can be associated with the normal
coordinate Qk0 = 0, l of the associate optical
phonon as a microscopic order parameter
characterizing the displacements on the
atomic scale, the “macroscopic” ferroelas-
tic orderings, where in the phenomenologi-
cal theory, Eq. (4-27), a component of the
strain tensor emn is used as an order pa-
rameter, can be related to acoustical pho-
nons. Examples are the martensitic mate-
rial In-25 at.% Tl where the combination
c11 – c12 of the elastic constants nearly sof-
tens at Tc = 195 K, and LaP5O14 where c55
Figure 4-16. (a) Schematic temperature variation of softens at Tc = 400 K and the structure
order parameter F = ·Q k0 , l 0 Ò T and square of the changes from orthorhombic to monoclinic
“soft-mode” frequency w 2 (k0 , l 0 ) at a displacive (Gebhardt and Krey, 1979). Of course, this
structural transition. (b) Schematic phonon spectrum
correspondence between the microscopic
of the solid at T > Tc . Note that either optical or
acoustic phonons may go soft, and for optical pho- description of displacements in crystals in
nons a softening may often occur at the boundary of terms of phonons and the phenomenologi-
the first Brillouin zone rather than at its center. cal macroscopic description in terms of po-
292 4 Statistical Theories of Phase Transitions
come clear that all these structural transi- different structures can be handled by per-
tions with a one-component order parame- forming this calculation for both phases
ter acquire characteristics of order – disor- and identifying the temperature Tc where
der transitions close to Tc, as expected from the free energy branches of the two phases
the “universality principle”, and therefore cross. Since the quasi-harmonic theory is a
the distinction between the character of a calculation of the mean-field type, as
structural transition as being “order – disor- pointed out above, first order transitions
der” or “displacive” is not a sharp one also show up via stability limits of the
(Bruce and Cowley, 1981). phases, where the soft modes vanish; thus
It should also be noted that the softening we are not locating Tc but rather tempera-
of w (k0, l0) near Tc does not mean that dis- tures T0 or T1 (cf. Fig. 4-6 b), which are
placements ul (x) become very large. In often not very far from the actual transi-
fact, the mean square displacement of an tion temperature. This quasi-harmonic ap-
atom at a structural transition is only ex- proach to structural phase transitions has
pected to have an energy-like singularity been tried for many materials. Typical ex-
(Meißner and Binder, 1975) amples include RbCaF3 (Boyer and Hardy,
1981) and the systems CaF2 and SrF2
·u2l ÒT – ·u2l ÒTc ~ (T /Tc – 1)1 – a (4-137)
(Boyer, 1980, 1981 a, b), which show phase
where a is the specific heat exponent. This transitions to a superionic conducting state.
is important because ·u2l ÒT is easily de-
duced experimentally from the Mössbauer
effect, from the Debye – Waller factor de- 4.3.3 Computer Simulation Techniques
scribing the temperature variation of Bragg In a computer simulation, we consider a
peaks in X-ray or neutron scattering, etc. finite system (e.g., a cubic box of size L3
We end this subsection with a comment with periodic boundary conditions to avoid
on the theory of first-order structural tran- surface effects) and obtain information on
sitions. The common approach is to restrict the thermodynamic properties, correlation
the analysis entirely to the framework of functions, etc. of the system (as specified
the quasi-harmonic approximation, in by its model Hamiltonian) which is exact,
which the Helmholtz energy at volume V apart from statistical errors. However, this
and temperature T is written as approach is restricted to classical statistical
mechanics (including quasi-classical mod-
F (T , V ) = U − T S
els such as Ising or Potts models), although
1
= U0 ( V ) + ∑ w V ( k , l ) (4-138) remarkable progress on application to
2 k, l quantum problems has been made (Kalos,
+ kB T ∑ ln [1 − exp( − w V ( k , l )/ kB T )] 1985; Kalos and Schmidt, 1984; De Raedt
k, l and Lagendijk, 1985; Suzuki, 1992; Ceper-
Therefore, if the effective potentials speci- ley, 1995). The principal approaches of this
fying the dynamical matrix ∂2 U/[(∂xil )a type are the molecular dynamics (MD)
∂xjl )b ] in Eq. (4-135) are known, the pho- technique and the Monte Carlo (MC) tech-
non frequencies wV (k, l) for a given vol- nique (for reviews of MD, see Ciccotti
ume and the free energy F (T, V) are ob- et al., 1987; Hoover, 1987; Hockney and
tained. Of course in this approach, knowl- Eastwood, 1988; Binder and Ciccotti,
edge of the structure of the material is 1996; of MC, see Binder, 1979, 1984 a;
assumed. First-order transitions between Mouritsen, 1984; Binder and Heermann,
294 4 Statistical Theories of Phase Transitions
1988; Binder and Ciccotti 1996). In the principle, this problem is well understood
MD method, we numerically integrate New- (Fisher 1971; Challa et al., 1986; Binder,
ton’s equation of motion which follows 1987 b; Privman, 1990). In practice, this
from the chosen Hamiltonian, assuming er- makes it difficult to distinguish between
godic behavior, and the quantities of inter- second-order and weakly first-order transi-
est are obtained as time averages from the tions. For example, extensive MD work
simulation. This method requires that the was necessary to obtain evidence that the
relevant physical time scales involved in melting transition of two-dimensional sol-
the problem are not too different from each ids with pure Lennard – Jones interaction is
other. The MD approach has often been ap- first order (Abraham, 1983, 1984; Bakker
plied successfully to liquid – solid transi- et al., 1984), and that the suggested two
tions. It would not be suitable to study or- continuous transitions involving the hexa-
der – disorder phenomena in solid alloys, tic phase do not occur in these systems
since the time step in the MD method for a (Nelson and Halperin, 1979). However,
solid must be much less than a phonon fre- this conclusion has been called into ques-
quency, and this time scale is orders of tion by recent simulations for hard-disk
magnitude smaller than the time between fluids (Jaster, 1998) providing evidence for
diffusive hops of atoms to neighboring va- continuous two-dimensional melting.
cant sites, which is the process relevant to Another difficulty is that the periodic
equilibration of configurational degrees of boundary condition (for a chosen shape of
freedom (Fig. 4-12). the box) prefers certain structures of a solid
For problems of the latter type, the MC and suppresses others which do not “fit”:
method clearly is to be preferred. In the this is particularly cumbersome for incom-
MC method, random numbers are used to mensurate modulated structures (Selke, 1988,
construct a random walk through the con- 1989; 1992) and for off-lattice systems,
figuration space of the model system. Us- such as studies of the fluid – solid transition
ing the Hamiltonian, transition probabilities or phase transitions between different lat-
between configurations are adjusted such tice symmetries. For example, particles in-
that configurations are visited according to teracting with a screened Coulomb potential
their proper statistical weight (Binder, (this is a model for colloidal suspensions or
1979, 1984 a; Binder and Ciccotti, 1996). colloidal crystals (see Alexander et al.,
Again averages are obtained as “pseudo- 1984)) may exhibit a fluid phase in addi-
time averages” along the trajectory of the tion to an f.c.c. and a b.c.c. crystal, and the
system in phase space, the only difference determination of a complete phase diagram
from the MD method being that the tra- is correspondingly difficult (Kremer et al.,
jectory is now stochastic rather than deter- 1986, 1987; Robbins et al., 1988). The tra-
ministic. Both methods have been exten- ditional approach to dealing with such prob-
sively reviewed (see the references quoted lems is to repeat the calculation for differ-
above); therefore, we do not give any de- ent box shapes and compare the free ener-
tails here, but only briefly mention the dif- gies of the different phases. An interesting
ficulties encountered when phase transi- alternative method has been proposed by
tions are studied, and discuss a few typical Parrinello and Rahman (1980) and Parri-
examples of their application. nello et al. (1983), who generalized the MD
One principal difficulty is the finite-size method by including the linear dimensions
rounding and shifting of the transition. In of the box as separate dynamic variables.
4.3 Computational Methods Dealing with Statistical Mechanics 295
Figure 4-19. Phase diagram of a uniaxial classical Figure 4-20. Critical line Kc–1 of the F 4 model
Heisenberg antiferromagnet on the simple cubic lat- (Eq. (4-136)) on the square lattice, shown in the
tice, as a function of temperature T and field H|| ap- space of couplings K –1, L–1 defined as K = – r C/u,
plied in the direction of the easy axis. The anisotropy L = r (r + 4 C)/(4 u). Note that for L–1 = 0, and K finite,
parameter D in the Hamiltonian Eq. (4-93) is chosen the square Ising model results, the transition temper-
as D = 0.2. Both Monte Carlo results (crosses, circles, ature of which is exactly known (Onsager, 1944; ar-
full curves) and the Landau theory fitted to the phase row). (쐌) Monte Carlo simulation of Milchev et al.
diagram in the region off the true bicritical point is (1986); (+) MD results of Schneider and Stoll
shown (dash-dotted straight line). The Landau theory (1976); (왌, ¥) from a real space renormalization
would overestimate the location of the bicritical tem- group calculation of Burkhardt and Kinzel (1979).
perature Tb (Tb* > Tb ), and fails to yield the singular The disordered phase occurs above the critical line.
umbilicus shape of the phase diagram lines Tc^, Tc|| From Milchev et al. (1986).
near the bicritical point. The broken straight lines de-
note the appropriate choices of “scaling axes” for a
crossover scaling analysis near Tb . The triangles nearest neighbors only, further-neighbor
show another phase diagram, namely when the field interactions being negligibly small, with
is oriented in the perpendicular direction to the easy
axis. The nature of the phases is denoted as AF (anti-
Jnnn/ Jnn = – 1/2. Thus Fig. 4-21 exhibits a
ferromagnetic), SF (spin-flop) and P (paramagnetic). remarkable agreement between computer
See Fig. 4-9 a for a schematic explanation of this simulation and experiment (Maletta and
phase diagram. From Landau and Binder (1978). Felsch, 1979); there are no adjustable pa-
rameters involved whatsoever. Note that
where Si are unit vectors in the direction of analytical methods for such problems with
the Eu magnetic moment, and from spin quenched disorder would be notoriously
wave measurements (Bohn et al., 1980) it difficult to apply.
is known that EuS essentially exhibits To conclude this subsection, we state
superexchange between nearest and next- that computer simulation techniques are
298 4 Statistical Theories of Phase Transitions
that otherwise dominate the partition func- F = Fs (T ) acts like a line of critical points,
tion. Langer (1974) suggested that phase as discussed above, we require (cf. Eq.
separation into two phases with order pa- (4-58))
rameters – F 0 and + F 0 coexisting at the
·[F (x) – F]2 ÒT, L O [F – Fs (T )]2 (4-142)
first-order transition at H = 0 in Fig. 4-1
and Fig. 4-22 is suppressed if we consider i.e., the mean-square fluctuations in a vol-
a system (a) at fixed F between – F 0 and ume Ld must be much smaller than the
+ F 0 but (b) constrain the system and di- squared distance of the order parameter
vide it into cells of size Ld, and (c) require from its value at criticality, which in this
that the order parameter is not only glo- case is the value at the spinodal. Now, the
bally fixed at F but also inside each cell. maximum permissible value for L is the
If L is small enough, namely a O L O xcoex correlation length x in the metastable state,
(this is the same condition as noted for which becomes (4-143)
the construction of a continuum model
from the microscopic Hamiltonian, see x = {R/[6 d u Fs (T )]1/2} [F – Fs (T )]– 1/2
Sec. 4.2.2, and Eqs. (4-9) and (4-57)), Eq. (4-143) exhibits the critical singularity
phase separation inside a cell cannot occur, of x at Fs (T ), corresponding to that of cT
and hence a coarse-grained Helmholtz en- noted above. With a little algebra we find
ergy density fcg (F ) of states with uniform from Eqs. (4-142) and (4-143) (Binder,
order parameter F is obtained, which has 1984 c)
exactly the double-well shape of the Lan-
dau theory (Fig. 4-6 a). Since fcg (F ) then 1 O R2 x d − 2 [F − Fs ( T )]2
necessarily depends on this coarse-grained ~ R d [Fs ( T )]( 2 − d )/ 2 [F − Fs ( T )]( 6 − d )/ 2
length scale L, it is clear that the metastable
~ R d ( Hc − H )( 6 − d )/ 4 (4-144)
states of fcg (F ) are not precisely those
observed in experiments, unlike the dy- This condition can only be fulfilled if the
namic definition. While fcg (F ) defines a interaction range is very large. Even then
“limit of metastability” or “spinodal curve” for d = 3 this inequality is not fulfilled very
Fs where (∂2 fcg (F )/∂F 2 )T = c T– 1 vanishes close to the spinodal curve.
and changes sign, this limit of metastability In order to elucidate the physical signifi-
also depends on L and hence is not related cance of this condition further and to dis-
to any physical limit of metastability. In the cuss the problem of the lifetime of meta-
context of spinodal decomposition of mix- stable states, we briefly discuss the mean
tures, the problem of the extent to which a field theory of nucleation phenomena. In
spinodal curve is relevant also arises and is practice, for solid materials, heterogeneous
discussed in detail in the chapter by Binder nucleation (nucleation at surfaces, grain
and Fratzl (2001). boundaries, dislocations, etc.) will domi-
In some systems where a mean-field de- nate; and this may restrict the existence of
scription is appropriate, the lifetime of metastable states much more than expected
metastable states can be extremely long, from Eq. (4-144). However, we consider
and then the mean-field concept of a limit here only the idealized case that homoge-
of metastability may be very useful. This neous nucleation (formation of “droplets”
fact can again be understood in terms of of the stable phase due to statistical fluctu-
the Ginzburg criterion concept as outlined ations) is the mechanism by which the met-
in Sec. 4.2.2. Since the spinodal curve astable state decays. Since there is no
4.4 Concepts About Metastability 301
such a model Hamiltonian does not exist. Baxter, R. J. (1982), Exactly Solved Models in Statis-
tical Mechanics. London: Academic.
Many of the more sophisticated methods Bean, C. P., Rodbell, D. S. (1962), Phys. Rev. 126,
(CV method, computer simulation, etc.) 104.
are restricted in practice to relatively sim- Belanger, D. P. (1998), in: Spin Glasses and Random
Fields: Young, A. P. (Ed.). Singapore: World Sci-
ple models. Hence, there still remains a lot entific, p. 251.
to be done to improve our understanding of Berlin, T. H., Kac, M. (1952), Phys. Rev. 86, 821.
phase transitions. Bethe, H. (1935), Proc. Roy. Soc. A 150, 552.
Bieber, A. Gautier, F. (1984 a), J. Phys. Soc. Japan
53, 2061.
Bieber, A., Gautier, F. (1984 b), Z. Phys. B 57, 335.
4.6 References Binder, K. (1973), Phys. Rev. B 8, 3423.
Binder, K. (Ed.) (1979), Monte Carlo Methods in
Abraham, F. F. (1974), Homogeneous Nucleation Statistical Physics. Berlin: Springer.
Theory. New York: Academic. Binder, K. (1980), Phys. Rev. Lett. 45, 811.
Abraham, F. F. (1983), Phys. Rev. Lett. 50, 978. Binder, K. (1983 a), Z. Physik B 50, 343.
Abraham, F. F. (1984), Phys. Rev. B 29, 2606. Binder, K. (1983 b), in: Phase Transitions and Criti-
Aharony, A. (1973), Phys. Rev. B 8, 1973. cal Phenomena, Vol. 8: Domb, C., Green, M. S.
Aharony, A. (1976), in: Phase Transitions and Criti- (Eds.). London: Academic Press, p. 1.
cal Phenomena, Vol. 6: Domb, C., Green, M. S. Binder, K. (1984 a), Applications of the Monte Carlo
(Eds.). London: Academic, p. 358. Method in Statistical Physics. Berlin: Springer
Alefeld, G. (1969), Phys. Stat. Sol. 32, 67. Verlag.
Alefeld, G., Völkl, J. (Eds.) (1978), Hydrogen in Binder, K. (1984 b), in: Condensed Matter Research
Metals I, II. Berlin: Springer Verlag (Topics in Using Neutrons: Lovesey, S. W., Scherm, R.
Appl. Physics, Vols. 28, 29). (Eds.). New York: Plenum Press, p. 1.
Alexander, S. (1975), Phys. Lett. 54 A, 353. Binder, K. (1984 c), Phys. Rev. A 29, 341.
Alexander, S., Chaikin, P. M., Grant, P., Morales, G. Binder, K. (1986), in: Atomic Transport and Defects
J., Pincus, P., Hone, D. (1984), J. Chem. Phys. 80, in Metals by Neutron Scattering: Janot, C., Petry,
5776. T., Richter, D., Springer, T. (Eds.). Berlin: Sprin-
Als-Nielsen, J. (1976), in: Phase Transitions and ger, p. 12.
Critical Phenomena, Vol. 5 A: Domb, C., Green, Binder, K. (1987 a), Rep. Progr. Phys. 50, 783.
M. S. (Eds.). New York: Academic, p. 88. Binder, K. (1987 b), Ferroelectrics 73, 43.
Als-Nielsen, J., Laursen, I. (1980), in: Ordering of Binder, K. (1989), in: Alloy Phase Stability: Gonis,
Strongly Fluctuating Condensed Matter Systems: A., Stocks, L. M. (Eds.). Dordrecht: Kluwer Acad.
Riste, T. (Ed.). New York: Plenum Press, p. 39. Publ., p. 232.
Amit, D. J. (1984), Field Theory, the Renormaliza- Binder, K. (1994), Adv. Polymer Sci. 112, 181.
tion Group and Critical Phenomena. Singapore: Binder, K. (1998), in: Spin Glasses and Random
World Scientific. Fields: Young, A. P. (Ed.). Singapore: World Sci-
Angell, C. A., Goldstein, M. (Eds.) (1986), Dynamic entific, p. 99.
Aspects of Structure Change in Liquids and Binder, K., Fratzl, P. (2001), in: Phase Transforma-
Glasses. New York: N.Y. Acad. of Sciences. tions in Materials: Kostorz, G. (Ed.). Weinheim:
Ausloos, M., Elliott, R. J. (1983), Magnetic Phase Wiley-VCH, Chap. 6.
Transitions. Berlin: Springer Verlag. Binder, K., Heermann, D. W. (1988), Monte Carlo
Bak, P., Domany, E. (1979), Phys. Rev. B 20, 2818. Simulation in Statistical Physics: An Introduction.
Bak, P., Mukamel, D. (1979), Phys. Rev. B 19, 1604. Berlin: Springer.
Bak, P., Mukamel, D., Villain, J., Wentowska, K. Binder, K. Landau, D. P. (1981), Surface Sci. 108, 503.
(1979), Phys. Rev. B 19, 1610. Binder, K., Landau, D. P. (1989), in: Molecule – Sur-
Bakker, A. F., Bruin, C., Hilhorst, H. J. (1984), Phys. face Interactions: Lawley, K. P. (Ed.). New York:
Rev. Let. 52, 449. Wiley & Sons, p. 91.
Bardham, P., Cohen, J. B. (1976), Acta Cryst. A 32, Binder, K., Stauffer, D. (1976 a), Z. Phys. B 24,
597. 407.
Bates, S. F., Rosedale, J. H., Fredrickson, G. H., Binder, K., Stauffer, D. (1976 b), Adv. Phys. 25, 343.
Glinka, C. J. (1988), Phys. Rev. Lett. 61, 229. Binder, K., Young, A. P. (1986), Rev. Mod. Phys. 58,
Baus, M. (1987), J. Stat. Phys. 48, 1129. 801.
Baus, M., Rull, L. F., Ryckaert, J. P. (Eds.) (1995), Binder, K., Meißner, G., Mais, H. (1976), Phys. Rev.
Observation, Prediction and Simulation of Phase B 15, 267.
Transitions in Complex Fluids. Dordrecht: Kluwer Binder, K., Kinzel, W., Stauffer, D. (1979), Z. Phys.
Acad. Publ. B 36, 162.
Baxter, R. J. (1973), J. Phys. C 6, L445. Binder, K. Reger, J. D. (1992), Adv. Phys. 41, 547.
4.6 References 305
Binder, K., Ciccotti, G. (Eds.) (1996), Monte Carlo Cusack, N. E. (1987), The Physics of Structurally
and Molecular Dynamics of Condensed Matter. Disordered Matter: An Introduction. Bristol:
Bologna: Società Italiana di Fisica. Adam Hilger.
Blinc, R., Levanyuk, A. P. (1986), Incommensurate De Fontaine, D. (1975), Acta Met. 23, 553.
Phases in Dielectrics, Vols. 1, 2. Amsterdam: De Fontaine, D. (1979), in: Solid State Physics, Vol.
North-Holland. 34: Ehrenreich, H., Seitz, F., Turnbull, D. (Eds.).
Blinc, R., Zeks, B. (1974), Soft Modes in Ferro- London: Academic, p. 73.
electrics and Antiferroelectrics. Amsterdam: De Fontaine, D., Kulik, J. (1985), Acta Metall. 33, 145.
North-Holland. De Gennes, P. G. (1974), The Physics of Liquid Crys-
Bloch, D., Mauri, R. (1973), Phys. Rev. B 7, 4883. tals. Oxford: Clarendon Press.
Blöte, H. W., Swendsen, R. H. (1979), Phys. Rev. De Gennes, P. G. (1979), Scaling Concepts in Poly-
Lett. 43, 799. mer Physics. Ithaca: Cornell University Press.
Blume, M. (1966), Phys. Rev. 141, 517. De Jongh, J. L., Miedema, A. R. (1974), Adv. Phys.
Blume, M., Emery, V., Griffiths, R. B (1971), Phys. 23, 1.
Rev. A 4, 1971. Delaey, L. (2001), in: Phase Transformations in Ma-
Bohn, H. G., Zinn, W., Dorner, B., Kollmar, A. terials: Kostorz, G. (Ed.). Weinheim: Wiley-VCH,
(1980), Phys. Rev. B 22, 5447. Chap. 9.
Born, M., Huang, K. (1954), Dynamical Theory of Den Nijs, M. P. M. (1979), J. Phys. A 12, 1857.
Crystal Lattices. Oxford: Oxford Univ. Press. De Raedt, H., Lagendijk, A. (1985), Phys. Repts.
Boyer, L. L. (1980), Phys. Rev. Lett. 45, 1858. 127, 235.
Boyer, L. L. (1981 a), Phys. Rev. Lett. 56, 1172. De Raedt, B., Binder, K., Michel, K. H. (1981), J.
Boyer, L. L. (1981 b), Ferroelectrics 35, 83. Chem. Phys. 75, 2977.
Boyer, L. L., Hardy, J. R. (1981), Phys. Rev. B 24, Diep, H. T., Ghazali, A., Berge, B., Lallemand, P.
2577. (1986), Europhys. Lett. 2, 603.
Bragg, W. L., Williams, E. J (1934), Proc. Roy. Soc. Dietrich, S. (1988), in: Phase Transitions and Criti-
A 145, 699. cal Phenomena, Vol. 12: Domb, C., Lebowitz, J. L.
Brazovskii, S. A. (1975), Soviet Phys. JETP 41, 85. (Eds.). New York: Academic, p. 1.
Bretz, M. (1977), Phys. Rev. Lett. 38, 501. Dillon, J. F., Chen, E. Y., Guggenheim, H. J. (1978),
Broddin, D., Van Tendeloo, G., Van Landuyt, J., Phys. Rev. B 18, 377.
Amelinckx, S., Portier, R., Guymont, M., Loiseau, Di Tolla, D. F., Tosatti, E., Ercolessi, F. (1996), in:
A. (1986), Phil. Mag. A 54, 395. Monte Carlo and Molecular Dynamics of Con-
Brout, R. (1959), Phys. Rev. 115, 824. densed Matter Systems: Binder, K., Ciccotti, G.
Brout, R. (1965), Phase Transitions. New York: Ben- (Eds.). Bologna: Società Italiana di Fisica, p. 345.
jamin Press. Domany, E., Schick, M., Walker, J. S., Griffiths, R.
Bruce, A. D., Aharony, A. (1975), Phys. Rev. B 11, B. (1978), Phys. Rev. B 18, 2209.
478. Domany, E., Shnidman, Y., Mukamel, D. (1982), J.
Bruce, A. D., Cowley, R. A. (1981), Structural Phase Phys. C 15, L495.
Transitions. London: Taylor and Francis. Domb, C., Green, M. S. (Eds.) (1972 – 1976), Phase
Burkhardt, T. W., Kinzel, W. (1979), Phys. Rev. B 20, Transitions and Critical Phenomena, Vols. 1 – 6.
4730. New York: Academic.
Burkhardt, T. W., van Leeuwen, J. M. J. (Eds.) (1982), Domb, C, Green, M. S. (Eds.) (1983 – 1997), Phase
Real-Space Renormalization. Berlin: Springer. Transitions and Critical Phenomena, Vols. 6 – 17.
Buzaré, J. Y., Fayet, J. C., Berlinger, W., Müller, K. New York: Academic.
A. (1979), Phys. Rev. Lett. 42, 465. Dünweg, B., Binder, K. (1987), Phys. Rev. B 36, 6935.
Cahn, J. W., Hilliard, J. E. (1958), J. Chem. Phys. 28, Edwards, S. F., Anderson, P. W. (1975), J. Phys. F 5,
258. 965.
Cahn, J. W., Hilliard, J. E. (1959), J. Chem. Phys. 31, Enz, C. P. (Ed.) (1979), Dynamic Critical Phenom-
688. ena and Related Topics. Proceedings. Berlin:
Ceperley, D. M. (1995), Rev. Mod. Phys. 67, 279. Springer Verlag (Springer Lecture Notes in Phys-
Challa, M. S. S., Binder, K., Landau, D. P. (1986), ics, Vol. 104).
Phys. Rev. B 34, 1841. Ferrell, R. A., Menyhard, N., Schmidt, H., Schwabl,
Chandrasekhar, S. (1992), Liquid Crystals. Cam- F., Szepfalusy, P. (1967), Phys. Rev. Lett. 18, 891.
bridge: Cambridge University Press. Finel, A. (1994), in: Static and Dynamics of Alloy
Ciccotti, G., Frenkel, D., McDonald, I. R. (Eds.) Phase Transformations: Turchi, P. E. A., Gonis, A.
(1987), Simulations of Liquids and Solids. Amster- (Eds.). New York: Plenum Press, p. 495.
dam: North-Holland. Fisher, M. E. (1968), Phys. Rev. 176, 257.
Clapp, P. C., Moss, S. C. (1966), Phys. Rev. 142, 418. Fisher, M. E. (1971), in: Critical Phenomena: Green
Clapp, P. C., Moss, S. C. (1968), Phys. Rev. 171, 754. M. S. (Ed.). London: Academic, p. 1.
Cowley, J. M. (1950), Phys. Rev. 77, 669. Fisher, M. E. (1974), Rev. Mod. Phys. 46, 587.
Cowley, J. M. (1976), Phys. Rev. B 13, 4877. Fisher, M. E. (1975 a), AIP Conf. Proc. 24, 273.
306 4 Statistical Theories of Phase Transitions
Fisher, M. E. (1975 b), Phys. Rev. Lett. 34, 1634. Halperin, B. I., Hohenberg, P. C. (1967), Phys. Rev.
Fisher, M. E., Langer, J. S. (1968), Phys. Rev. Lett. Lett. 19, 700.
20, 665. Halperin, B. I., Varma, C. (1976), Phys. Rev. B14, 4030.
Fisher, M. E., Nelson, D. R. (1974), Phys. Rev. Lett. Halperin, B. I., Hohenberg, P. C., Ma, S. K. (1974),
32, 1350. Phys. Rev. B 10, 139.
Fisher, M. E., Racz, Z. (1976), Phys. Rev. B 13, 5039. Halperin, B. I., Hohenberg, P. C., Siggia, E. (1976),
Fishman, S., Aharony, A. (1979), J. Phys. C 12, L 729. Phys. Rev. B 13, 4119.
Flory, P. J. (1953), Principles of Polymer Chemistry. Hansen, M. (1958), Constitution of Binary Alloys.
Ithaca: Cornell University Press. New York: McGraw Hill.
Folk, R., Iro, H., Schwabl, F. (1976), Z. Phys. B 25, 69. Harris, A. B. (1974), J. Phys. C 7, 1671.
Franz, S., Parisi, G. (1998), Physica A 261, 317. Harris, A. B., Meyer, H. (1985), Can. J. Phys. 63, 3.
Frazer, B. C., Shirane, G., Cox, D. E. (1965), Phys. Harris, R., Plischke, M., Zuckermann, M. J. (1973),
Rev. 140 A, 1448. Phys. Rev. Lett. 31, 160.
Fredrickson, G. H., Binder, K. (1989), J. Chem. Heermann, D. W., Klein, W., Stauffer, D. (1982),
Phys. 91, 7265. Phys. Rev. Lett. 49, 1262.
Fredrickson, G. H., Helfand, E. (1987), J. Chem. Herrmann, H. J., Janke, W., Karsch, F. (Eds.) (1992),
Phys. 87, 697. Dynamic of First Order Phase Transitions. Singa-
Friedmann, L. F., Tunstall, D. P., (Eds.) (1978), The pore: World Scientific.
Metal – Non Metal Transition in Disordered Hockney, R. W., Eastwood, J. W. (1988), Computer
Systems. Edinburgh: SUSSP. Simulation Using Particles. Bristol: Adam Hilger.
Fujimoto, M. (1997), The Physics of Structural Höchli, U. T., Knorr, K., Loidl, A. (1990), Adv. Phys.
Phase Transitions. Berlin: Springer. 39, 405.
Gahn, U. (1986), J. Phys. Chem. Solids 47, 1153. Hohenberg, P. C., Halperin, B. I. (1977), Rev. Mod.
Gebhardt, W., Krey, U. (1979), Phasenübergänge Phys. 49, 435.
und kritische Phänomene – Eine Einführung. Hoover, W. G. (1987), Molecular Dynamics. Berlin:
Braunschweig: Vieweg. Springer.
Gehring, G. A. (1977), J. Phys. C 10, 531. Hornreich, R. M., Luban, M., Shtrikman, S. (1975),
Gehring, G. A., Gehring, K. A. (1975), Rep. Progr. Phys. Rev. Lett. 35, 1678.
Phys. 38, 1. Hulliger, F., Siegrist, T. (1979), Z. Phys. B 35, 81.
Gerold, V., Kern, J. (1986), in: Atomic Transport and Imbrie, J. Z. (1984), Phys. Rev. Lett. 53, 1747.
Defects in Metals by Neutron Scattering: Janot, C., Imry, Y. (1980), Phys. Rev. B 21, 2042.
Petry, T., Richter, D., Springer, T. (Eds.). Berlin: Imry, Y., Ma, S. K. (1975), Phys. Rev. Lett. 35, 1399.
Springer Verlag, p. 17. Imry, Y., Wortis, M. (1979), Phys. Rev. B 19, 3580.
Gerold, V., Kern, J. (1987), Acta Met. 35, 393. Inden, G. (2001), in: Phase Transformations in Mate-
Ginzburg, V. L. (1960), Sov. Phys.-Solid State 2, rials: Kostorz, G. (Ed.). Weinheim: Wiley-VCH,
1824. Chap. 8.
Giordano, N., Wolf, W. P. (1975), Phys. Rev. Lett. 35, Jäckle, J. (1986), Rep. Progr. Phys. 49, 171.
799. Jaster, A. (1998), Europhys. Lett. 42, 277.
Golosov, N. S., Popov, L. E., Rudan, L. Y. (1973), Jona, F., Shirane, G. (1962), Ferroelectric Crystals.
J. Phys. Chem. Solids 34, 1149 – 1159. New York: Pergamon Press.
Gonis, A., Stocks, L. M. (Eds.) (1989), Alloy Phase Joyce, J. S. (1972), in: Phase Transitions and Criti-
Stability. Dordrecht: Kluwer Acad. Publ. cal Phenomena, Vol. 2: Domb, C., Green, M. S.
Grazhdankina, N. P. (1969), Sov. Phys.-Usp. 11, 727; (Eds.). New York: Academic, Chapter X.
Usp. Fiz. Nauk. 96, 291. Kämmerer, S., Dünweg, B., Binder, K., D’Onorio De
Grinstein, G. (1985), in: Fundamental Problems in Meo, M. (1996), Phys. Rev. B 53, 2345.
Statistical Mechanics, VI: Cohen, E. G. D. (Ed.). Kalos, M. H. (Ed.) (1985), Monte Carlo Methods in
Amsterdam, North-Holland, p. 147. Quantum Problems. Dordrecht: Kluwer Acad. Publ.
Guerard, D., Herold, A. (1975), Carbon 13, 337. Kalos, M. H., Schmidt, K. E. (1984), in: Applications
Gunton, J. D., Droz, M. (1983), Introduction to the of the Monte Carlo Method in Statistical Physics:
Theory of Metastable and Unstable States. Berlin: Binder, K. (Ed.). Berlin: Springer Verlag, p. 125.
Springer. Kaplan, T. A. (1959), Phys. Rev. 116, 888.
Gunton, J. D., San Miguel, M., Sahni, P. S. (1983), Keating, D. T., Warren, B. E. (1951), J. Appl. Phys.
in: Phase Transitions and Critical Phenomena, 22, 286.
Vol. 8: Domb, C., Lebowitz, J. L. (Eds.). London: Kehr, K. W., Binder, K., Reulein, S. M. (1989), Phys.
Academic, p. 267. Rev. B 39, 4891.
Guymont, M. (1978), Phys. Rev. B 18, 5385. Khachaturyan, A. G. (1962), Fiz. Metallov i Metallo-
Guymont, M. (1981), Phys. Rev. B 24, 2647. vedenie 13, 493.
Haasen, P., Gerold, V., Wagner, R., Ashby, M. F. Khachaturyan, A. G. (1963), Fiz. tverd. Tela 5, 26;
(Eds.) (1984), Decomposition of Alloys: The Early ibid 5, 750.
Stages. New York: Pergamon Press. Khachaturyan, A. G. (1973), Phys. stat. sol. (b) 60, 9.
4.6 References 307
Khachaturyan, A. G. (1983), Theory of Structural Michel, K. H., Naudts, J. (1977), J. Chem. Phys. 67,
Transformation in Solids. New York: J. Wiley & 547.
Sons. Michel, K. H., Naudts, J. (1978), J. Chem. Phys. 68,
Kikuchi, R. (1951), Phys. Rev. 81, 998. 216.
Kikuchi, R. (1967), J. Chem. Phys. 47, 1664. Milchev, A., Heermann, D. W., Binder, K. (1986),
Kittel, C. (1967), Quantum Theory of Solids. New J. Stat. Phys. 44, 749.
York: J. Wiley & Sons. Moncton, D. E., Axe, J. D., Di Salvo, F. J. (1977),
Kittler, R. C., Falicov, L. M. (1978), J. Phys. B18, 2506. Phys. Rev. B 16, 801.
Kittler, R. C., Falicov, L. M. (1979), Phys. Rev. B 19, Morita, T. (1972), J. Math. Phys. 13, 115.
291. Moss, S. C. (1964), J. Appl. Phys. 35, 3547.
Kleemann, W., Schäfer, F. J., Tannhäuser, D. S. Moss, S. C., Clapp, P. C. (1968), Phys. Rev. 171, 764.
(1980), J. Magn. Magnet. Mater. 15 – 18, 415. Mott, N. F. (1974), Metal-Insulator Transitions. Lon-
Klein, W., Unger, C. (1983), Phys. Rev. B 28, 445. don: Taylor and Francis.
Koch, S. W. (1984), Dynamics of First-Ordered Mouritsen, O. G. (1984), Computer Studies of Phase
Phase Transitions in Equilibrium and Nonequilib- Transitions and Critical Phenomena. Berlin:
rium. Berlin: Springer. Springer.
Kötzler, J. (1984), Z. Phys. B 55, 119. Müller, K. A., Berlinger, W. (1975), Phys. Rev. Lett.
Kötzler, J., Raffius, G., Loidl, A., Zeyen, C. M. E. 22, 1547.
(1979). Z. Phys. B 35, 125. Müller-Krumbhaar, H. (1977), in: Crystal Growth
Kosterlitz, J. M., Thouless, D. J. (1973), J. Phys. C 6, and Materials: Kaldis, E. (Ed.). Amsterdam:
1181. North-Holland, p. 79.
Kremer, K., Robbins, M. O., Grest, G. S. (1986), Mukamel, D., Krinsky, S. (1976), Phys. Rev. B 13,
Phys. Rev. Lett. 57, 2694. 5065.
Kremer, K., Grest, G. S., Robbins, O. (1987), J. Mukamel, D., Fisher, M. E., Domany, E. (1976 a),
Phys. A 20, L181. Phys. Rev. Lett. 37, 565.
Krinsky, S., Mukamel, D. (1977), Phys. Rev. B 16, Mukamel, D., Krinsky, S., Bak, P. (1976 b), AIP
2313. Conf. Proc. 29, 474.
Krivoglaz, M. A. (1969), Theory of X-Ray and Ther- Nattermann, T. (1998), in: Spin Classes and Random
mal Neutron Scattering by Real Crystals. New Fields: Young, A. P. (Ed.). Singapore: World Sci-
York: Plenum Press. entific, p. 277.
Landau, D. P., Binder, K. (1978), Phys. Rev. B17, 2328. Nelson, D. R., Halperin, B. I. (1979), Phys. Rev. B19,
Landau, D. P., Binder, K. (1985), Phys. Rev. B31, 5946. 2457.
Landau, L. D., Lifshitz, E. M. (1958), Statistical Nienhuis, B., Riedel, E. K., Schick, M. (1980), J.
Physics. Oxford: Pergamon Press. Phys. A 13, L189.
Langer, J. S. (1974). Physica 73, 61. Nishiyama, Z. (1979), Martensitic Transformation.
Larkin, A. I., Khmelnitskii, D. E. (1969), Zh. Eksp. New York: Academic.
Fiz. 56, 2087; Sov. Phys. JETP 29, 1123. Onsager, L. (1944), Phys. Rev. 65, 117.
Lebowitz, J. L., Penrose, O. (1966), J. Math. Phys. 7, 98. Orr, R. L. (1960), Acta Met. 8, 489.
LeGuillou, J. C., Zinn-Justin, J. (1980), Phys. Rev. Orr, R. L., Luciat-Labry, J., Hultgren, R. (1960), Acta
B 21, 3976. Met. 8, 431.
Leibler, L. (1980), Macromolecules 13, 1602. Parrinello, M., Rahman, A. (1980), Phys. Rev. Lett.
Lifshitz, E. M. (1942), J. Physique 6, 61. 45, 1196.
Lipowsky, R. (1984), J. Appl. Phys. 55, 213. Parrinello, M., Rahman, A., Vashishta, P. (1983),
Loidl, A. (1989), Ann. Rev. Phys. Chem. 40, 29. Phys. Rev. Lett. 50, 1073.
Loiseau, A., Van Tendeloo, G., Portier, R., Ducas- Patashinskii, A. Z., Pokrovskii, V. I. (1979), Fluctua-
telle, F. (1985), J. Physique 46, 595. tion Theory of Phase Transitions. Oxford: Per-
Lynden-Bell, R. M., Michel, K. H. (1994), Rev. Mod. gamon Press.
Phys. 66, 721. Pershan, P. S. (1988), Structure of Liquid Crystal
Ma, S.-K. (1976), Modern Theory of Phase Transi- Phases. Singapore: World Scientific.
tions. Reading: Benjamin Press. Piercy, P., Pfnür, H. (1987), Phys. Rev. Lett. 59, 1124.
Maletta, H., Felsch, W. (1979), Phys. Rev. B 20, 1079. Pimenta, M. A., Echegut, P., Luspin, Y., Hauset, G.,
Mandelbrot, B. B. (1982), The Fractal Geometry of Gervais, F., Abélard, P. (1989), Phys. Rev. B 39,
Nature. San Francisco: Freeman. 3361.
Marx, R. (1989), Phys. Rev. B 40, 2585. Potts, R. B. (1952), Proc. Camb. Phil. Soc. 48, 106.
McCoy, B. M., Wu, T. T. (1973), The Two-Dimen- Privman, V. (Ed.) (1990), Finite Size Scaling and Nu-
sional Ising Model. Cambridge, Mass: Harvard merical Simulation. Singapore: World Scientific.
University Press. Privman, V., Hohenberg, P. C., Aharony, A. (1991),
McGuire, T. R., Gambino, R. J., Pickart, S. J., Alpe- in: Phase Transitions and Critical Phenomena,
rin, H. A. (1969), J. Appl. Phys. 40, 1009. Vol. 14: Domb, C., Lebowitz, J. L. (Eds.). London:
Meißner, G., Binder, K. (1975), Phys. Rev. B 12, 3948. Academic Press, Chapter 1.
308 4 Statistical Theories of Phase Transitions
Pynn, T., Skjeltorp, A. (Eds.) (1983), Multicritical Suzuki, M. (Ed.) (1992), Quantum Monte Carlo
Phenomena. New York: Plenum Press. Methods in Condensed Matter Physics. Singapore:
Riste, T., Sherrington, D. (Eds.) (1989), Phase Tran- World Scientific.
sitions in Soft Condensed Matter. New York: Ple- Swann, P. S., Duff, W. R., Fisher, R. M. (1972),
num Press. Metallurg. Trans. 3, 409.
Robbins, M. O., Kremer, K., Grest, G. S. (1988), J. Swendsen, R. H., Krinsky, S. (1979), Phys. Rev. Lett.
Chem. Phys. 88, 3286. 43, 177.
Rohrer, H. (1975), Phys. Rev. Lett. 34, 1638. Tanisaki, S. (1961), Phys. Soc. Japan 16, 579.
Rohrer, H., Gerber, Ch. (1977), Phys. Rev. Lett. 38, 909. Tolédano, J. C. (1981), Ferroelectrics 35, 31.
Roth, W. L. (1958), Phys. Rev. 110, 1333. Tolédano, P., Pascoli, G. (1980), Ferroelectrics 25,
Sadiq, A., Binder, K. (1984), J. Stat. Phys. 35, 517. 427.
Salanon, B., Fabre, F., Lapoujoulade, J., Selke, W. Tolédano, J. C., Tolédano, P. (1987), The Landau
(1988), Phys. Rev. B 38, 7385. Theory of Phase Transitions. Singapore: World
Salje, E. (1990), Phase Transitions in Ferroelastic Scientific.
and Co-elastic Crystals. Cambridge: Cambridge Tsakalakos, T. (1984), Phase Transformations in Sol-
Univ. Press. ids. Amsterdam: North-Holland.
Sanchez, J. M., De Fontaine, D. (1980), Phys. Rev. Turchi, P. E. A., Gonis, A. (Eds.) (1994), Statics and
B 21, 216. Dynamics of Alloy Phase Transformations. New
Sanchez, J. M., De Fontaine, D. (1982), Phys. Rev. York: Plenum Press.
B 25, 1759. van Hove, L. (1954), Phys. Rev. 93, 1374.
Schick, M. (1981), Progr. Surf. Sci. 11, 245. Villain, J. (1959), J. Phys. Chem. Solids 11, 303.
Schmid, F. (1999), in: Computational Methods in Villain, J. (1985), in: Scaling Phenomena in Disor-
Colloid and Interface Science: Borowko, M. (Ed.). dered Systems: Pynn, R., Skjeltorp, A. (Eds.). New
New York: M. Dekker, Chapter 10. York: Plenum Press, p. 423.
Schmidt, V. H. (1978), in: High Pressure and Low Wagner, H., Horner, H. (1974), Adv. Phys. 23, 587.
Temperature Physics: Chu, C. W., Wollam, J. A. Wagner, R., Kampmann, R., Vorhees, P. W. (2001),
(Eds.). New York: Plenum Press, p. 237. in: Phase Transformations in Materials: Kostorz,
Schneider, T., Stoll, E. (1976), Phys. Rev. B 13, 1216. G. (Ed.). Weinheim: Wiley-VCH, Chap. 5.
Schwartz, L. H., Cohen, J. B. (1965), J. Appl. Phys. Wansleben, S., Landau, D. P. (1987), J. Appl. Phys.
36, 598. 61, 3968.
Schweika, W. (1989), in: Alloy Phase Stability: Stocks, Weeks, J. D. (1980), in: Ordering in Strongly Fluctu-
G. M., Gonis, A. (Eds.). Dordrecht: Kluwer, p. 137. ating Condensed Matter Systems: Riste, T. (Ed.).
Schweika, W. (1994), in: Statics and Dynamics of New York: Plenum Press, p. 293.
Alloy Phase Transformations: Turchi, P. E. A., Wiechert, H., Arlt, S. A. (1993), Phys. Rev. Lett. 71,
Gonis, A. (Eds.). New York: Plenum Press, p. 103. 2090.
Schweika, W., Hauboldt, H.-G. (1986), in: Atomic Wilding, N. B. (1997), J. Phys.: Condens. Matter 9,
Transport and Defects in Metals by Neutron Scat- 585.
tering: Janot, C., Petry, T., Richter, D., Springer, T. Wilding, N. B., Schmid, F., Nielaba, P. (1998), Phys.
(Eds.). Berlin: Springer Verlag, p. 22. Rev. E 58, 2201.
Scott, J. F. (1987), in: Phase Transitions and Critical Wilson, K. G., Kogut, J. (1974), Phys. Repts. 12 C, 75.
Phenomena: Domb, C., Lebowitz, J. L. (Eds.). Wu, F. Y. (1982), Rev. Mod. Phys. 54, 235.
London: Academic Press, Chapter 2. Yelon, W. B., Cox, D. E., Kortmann, P. J. (1974),
Selke, W. (1988), Phys. Repts. 170, 213. Phys. Rev. B 9, 4843.
Selke, W. (1989), in: Alloy Phase Stability: Gonis, A., Yeomans, J. (1992), Statistical Mechanics of Phase
Stocks, L. M. (Eds.). Dordrecht: Kluwer Akad. Publ. Transitions. Oxford: Clarendon Press.
Selke, W. (1992), in: Phase Transitions and Critical Young, A. F. (Ed.) (1998), Spin Glasses and Random
Phenomena, Vol. 15: Domb, C., Lebowitz, J. L. Fields. Singapore: World Scientific.
(Eds.). London: Academic Press, p. 1. Zallen, T. (1983), The Physics of Amorphous Solids.
Semenovskaya, S. V. (1974), Phys. Stat. Sol. (b) 64, New York: J. Wiley and Sons.
291. Zarzycki, J. (Ed.) (1991), Glasses and Amorphous
Smart, J. S. (1966), Effective Field Theories of Mag- Materials (Materials Science and Technology,
netism. New York: W. B. Saunders. Vol. 9). Weinheim: VCH.
Stanley, H. E. (1971), Introduction to Phase Transi- Zettlemoyer, A. C. (Ed.) (1969), Nucleation. New
tions and Critical Phenomena. Oxford: Oxford York: Marcel Dekker.
University Press. Zunger, A. (1994), in: Statics and Dynamics of Alloy
Stauffer, D., Aharony, A. (1992), An Introduction to Phase Transformations: Turchi, P. E. A., Gonis, A.
Percolation Theory. London: Taylor and Francis. (Eds.). New York: Plenum Press, p. 361.
Sullivan, D. E., Telo da Gama, M. M. (1985), in:
Fluid Interfacial Phenomena: Croxton, C. A.
(Ed.). New York: J. Wiley and Sons, p. 45.
5 Homogeneous Second-Phase Precipitation
Richard Wagner
Reinhard Kampmann
Peter W. Voorhees
p pressure
R radius of precipitates
–
R mean radius of precipitates
R* radius of a critical nucleus or a particle being in unstable equilibrium
with the matrix
Rg molar gas constant
R (k); R (l) amplification factor in the CH spinodal theory
S entropy
S(k) structure function
Sm = S (km) maximum of structure function
T temperature
TA annealing temperature
TH homogenization temperature
t aging time
U internal energy
V volume
Va , Vb molar volume of a or b phase
Z Zeldovich factor
a, a ¢ matrix phase
b equilibrium precipitated phase
b¢ metastable precipitated phase; transition phase
g shear strain
d misfit parameter
h = (1/a0) (∂a/∂c) atomic size factor
k, k scattering vector and its magnitude
km wavenumber of the maximum of the structure function S (k)
l wavelength
m shear modulus
m ij chemical potential of component i in phase j
n Poisson’s ratio
W atomic volume
s ab specific interfacial energy
t incubation period or scaled time in the MLS and KW models
– –
a) 2 min.; 2 R = 2,4 nm b) 180 min.; 2 R = 8 nm
–
c) 420 min.; 2 R = 11 nm d)
Figure 5-2. a) – c) Neon field ion image of g ¢-precipitates (bright images in dark matrix) in Ni–36.8 at.% Cu–8
at.% Al aged for the given times at 580 °C (Liu and Wagner, 1984). d) Three-dimensional distribution of Mg and
Zn atoms in a commercial Al–5.5 at.% Mg–1.35 at.% Zn alloy analyzed with the tomographic atom probe. Each
point represents an atom. The T ¢ precipitates are enriched in Mg and Zn (Bigot et al., 1997).
The terminal solute concentration of the During the first step of the precipitation
solvent (A)-rich matrix a¢ is given by the sequence a metastable precipitate (transi-
equilibrium solubility limit cae (T ); the ter- tion phase b¢) is formed, frequently with
minal composition cp of the precipitated large associated coherency strains and a
phase is given by the solubility at TA at the small interfacial energy s ab. Often the
B-rich side of the phase diagram (Fig. 5-1). metastable phase b¢ is an intermetallic
Frequently, the interfacial energy s ab compound with a lower solute concentra-
(J/m2) between the matrix and the equilib- tion (c¢be ∫ c¢p) than the equilibrium precipi-
rium precipitate is rather high, particularly tate (cp). Because of the large coherency
if the width of the miscibility gap is large strains, the metastable (‘coherent’) solvus
(see Sec. 5.7.4.4). In this case, the decom- line (dashed line in Fig. 5-1) is shifted to-
position of a proceeds via the sequence wards higher solute concentrations. This
a Æ a≤ + b¢ Æ a¢ + b leads to a reduction of the supersaturation
5.2 General Considerations 317
precipitating phase b already has the final Hence, if, for example, F (c, T ) is known
bulk composition cbe ∫ cp rather than cpnuc for the stable solid solution, this value can
(which is only a reasonable assumption if readily be extrapolated into the adjacent
the supersaturation is not too large (Cahn two-phase region to yield F¢(c) (see Fig.
and Hilliard, 1959 a, b)), DFch has been de- 5-5) of the supersaturated homogeneous
rived for a unit volume of b phase with mo- solid solution or the related chemical po-
lar volume Vb as (Aaronson et al., 1970 a, tentials and activities. Of course, physi-
b): cally this is only meaningful if we assume
Rg T (as, in fact is done for the derivation of Eq.
DFch = – (5-9)
Vβ (5-9)) that the free energy of the non-equi-
librium solid solution can be properly de-
⎡ a (c ) a (c ) ⎤ fined within the miscibility gap.
× ⎢cβe ln B 0e + (1 – cβe )ln A 0e ⎥
⎣ aB ( cα ) aA ( cα ) ⎦ For many binary alloys and for some ter-
nary alloys of technological significance,
where ai (c) is the activity of the solvent the thermodynamic functions have been
(i = A) or solute component (i = B) in the evaluated by means of the CALPHAD
parent phase a for the given composition. method and are compiled in volumes of the
For most alloy systems the activity data CALPHAD series (Kaufman, 1977).
required for a computation of DFch by Hitherto, due to the lack of available ac-
means of Eq. (5-9) are not available. In tivity data, the activities entering Eq. (5-9)
principle, they can be derived from a com- were frequently replaced by concentra-
putation of the thermodynamic functions tions, i.e.,
by means of the CALPHAD method. Orig-
Rg T
inally this method was developed for the DFch = – (5-10)
calculation of phase diagrams by Kaufman Vβ
and coworkers (Kaufman and Bernstein, ⎡ c ⎛ 1 – c0 ⎞ ⎤
1970) on the basis of a few accurately × ⎢cβe ln 0e + (1 – cβe ) ln ⎜ ⎟⎥
measured thermodynamic data to which ⎣ cα ⎝ 1 – cαe ⎠ ⎦
suitable expressions for the thermodynam- If b is almost pure B, then cbe ≈ 1 and the
ic functions had been fitted. For the sake of chemical driving force is approximated as
a simplified mathematical description, the
stable solid solutions of the particular alloy Rg T c
DFch ≈ – ln 0e (5-11)
system are frequently described in terms of Vβ cα
the regular solution model, whereas the For many alloys the condition cbe ≈ 1 is
phase fields of the intermetallic com- not met. Nevertheless, it has been used for
pounds are approximated by line com- computation of DF. As discussed by Aa-
pounds. The thermodynamic functions ob- ronson et al. (1970 a), the resulting error
tained thereby are then used to reconstruct can be rather large. If the nucleating phase
the phase diagram (or part of it). The de- is a solvent-rich intermetallic compound
gree of self-consistency between the recon- (e.g., Ni3Al or Cu4Ti), the computation of
structed and the experimentally determined DFch is particularly difficult, since up to
phase diagram (or the agreement between now the corresponding activities are nei-
the measured thermodynamic data and the ther known nor furnished by the CAL-
derived data) serves as a measure of the ac- PHAD method, and Eqs. (5-10) and (5-11)
curacy of the thermodynamic functions. are no longer valid.
324 5 Homogeneous Second-Phase Precipitation
ii) Reduction of the driving force by the The problem of calculating DFel for co-
elastic strain energy, D Fel. The formation herent inclusions is rather complex. It has
of a composition fluctuation is associated generally been treated within the frame-
with the expenditure of elastic strain Helm- work of isotropic elastic theory for ellip-
holtz energy, DFel , if the solvent and the so- soidal precipitates with varying axial ra-
lute atoms have different atomic radii. Ac- tios, e.g. Eshelby (1957), and has been ex-
cording to Cahn (1962), the Helmholtz en- tended to the anisotropic case (Lee and
ergy of a system containing homophase Johnson, 1982). Basically, these treatments
fluctuations is raised by show that the strain energy depends on the
particular shape of the new phase. Only if
DFel = h2 Dc2 Y (5-12)
the transformation strains are purely dilata-
where h = (1/a0) (∂a/∂c) denotes the change tional and if a and b have about the same
in lattice parameter (a0 for the homogene- elastic constants will DFel become inde-
ous solid solution of composition c0) with pendent of the shape. It is then given as
composition, and Dc the composition am- (Eshelby, 1957):
plitude. Y is a combination of elastic con-
1+n ⎞ 2
stants and depends on the crystallographic DFel = 2 g ⎛ d (5-14)
direction of the composition modulation. It ⎝1–n ⎠
reduces to E/(1 – v), E and v being Young’s with g being the shear modulus.
modulus and Poisson’s ratio, respectively, As is shown schematically in Fig. 5-7,
if the elastic anisotropy A ∫ 2 c44 + c12 – with respect to the Helmholtz energy of the
c11 is zero; otherwise, in order to minimize incoherent equilibrium phase b, the Helm-
DFel , the composition fluctuations are ex- holtz energy of the coherent phase b¢ is
pected to grow along the elastically soft di- raised by the elastic energy to match the
rections, which for cubic crystals and A > 0 two lattices. The driving force DF inc ∫ ––
xy
are the ·100Ò directions.
In the context of heterophase fluctuations,
the barrier against nucleation of the new
phase is dominated by the matrix/ nucleus
interfacial energy (see Sec. 5.5.1). Since co-
herent interfaces have a lower energy than
incoherent ones, a precipitate is usually co-
herent (or at least semi-coherent) during the
early stages of nucleation and growth. Often
its lattice parameter, ab , is slightly different
from that of the parent phase, aa. The result-
ing misfit between the (unstrained) matrix
and the (unstrained) precipitate:
aα – aβ
d=2 (5-13)
aα + aβ
can be accommodated by an elastic strain
if both d and the particles are sufficiently Figure 5-7. Helmholtz energy curves and associated
small; this is commonly the case during the driving forces for precipitates of the coherent meta-
early stages of nucleation and growth. stable (Fb¢) and incoherent equilibrium (Fb) phases.
5.2 General Considerations 325
amount interest for optimum mechanical with respect to both the detection limit and
properties (Bratland et al., 1997). How- the spatial resolution of microanalysis.
ever, due to its inherent complexity, the
precipitation sequence of these commercial
5.3.1.1 Direct Imaging Techniques
alloys is still a matter of controversy (Dutta
and Allen, 1991; Gupta and Lloyd, 1992; Field ion microscopy (FIM) (Wagner,
Edwards et al., 1998). 1982) as well as conventional (CTEM)
(Hobbs et al., 1986) and high resolution
(HREM) transmission electron microscopy
5.3 Experimental Techniques for (Smith, 1983) allow for direct imaging of
Studying Decomposition Kinetics the second-phase particles, provided the
contrast between precipitate and matrix is
sufficient.
5.3.1 Microanalytical Tools
In CTEM both the bright field and the
In general, the course of a decomposi- dark field contrast of particles less than
tion reaction, including the early stages ≈ 5 nm in diameter are often either too
(during which composition fluctuations weak or too blurred for an accurate quanti-
and second-phase nuclei are formed, see tative determination of the relevant structu-
Sec. 5.2.1) and the coarsening stages, can- ral precipitate parameters. Hence, CTEM
not be followed continuously by any one does not provide access to an experimental
microanalytical technique. The progress of investigation of the early stages of decom-
the reaction is usually reconstructed from position but remains a technique for study-
the microstructure that develops at various ing the later stages. In contrast, HREM al-
stages of the phase transformation. Thus it lows solute clusters of less than 1 nm diam-
is necessary to analyze the spatial exten- eter to be imaged, as was demonstrated for
sion and the amplitude of composition Ni3Al precipitates in Ni–12 at.% (Si-Qun
fluctuations of incipient second-phase par- Xiao, 1989) and for silver-rich particles in
ticles, as well as the morphology, number Al–1 at.% Ag (Ernst and Haasen, 1988;
density, size and chemical composition of Ernst et al., 1987). As is shown in Fig. 5-9
individual precipitates at various stages of for the latter alloy, the particles are not
the phase transformation. For this purpose ideally spherical but show some irregular-
microanalytical tools are required that are ities which, however, are small in compari-
capable of resolving very small (typically a son with their overall dimensions. Hence,
few nm) solute clusters, and which allow describing their shapes by spheres of radius
(frequently simultaneously) an analysis of R is still a rather good approximation.
their chemical composition to be made. Prior to measuring the particle sizes di-
The tools that, in principle, meet these rectly from the HREM micrographs, it
requirements can be subdivided into two must be established via computer image
groups: direct imaging techniques and simulations that there exists a one-to-one
scattering techniques. correspondence between the width of the
It is beyond the scope of this chapter to precipitate contrast (dark area in Fig. 5-9)
discuss any one of the techniques belong- and its true size. This was verified for the
ing to either group in any detail. We shall HREM imaging conditions used in Fig.
only briefly summarize the merits and the 5.9. The evolution of the size distribution
shortcomings of the various techniques in Al–1 at.% Ag with aging time at 413 K
5.3 Experimental Techniques for Studying Decomposition Kinetics 327
ysis of composition fluctuations during scale (Fig. 5-2 d). The spatial distribution
early-stage decomposition. of each chemical species can be directly
Analytical field ion microscopy (AFIM) observed with a spatial resolution better
(Wagner, 1982; Miller et al., 1996) can fa- than 0.5 nm (Auger et al., 1995; Pareige et
vorably be applied for an analysis of ultra- al., 1999; Al-Kassab et al., 1997). The vol-
fine solute clusters in the early stages of ume typically sampled is about 15 × 15
decomposition, as both the imaging resolu- × 100 nm3 and, thus, about a hundred times
tion of the FIM (see Fig. 5-2) and the spatial larger than the volume analyzed with the
resolution of microanalysis of the integrated AFIM.
time-of-flight spectrometer (‘atom probe’)
are sufficient, i.e., <1 nm and ª 2 nm, re-
5.3.1.2 Scattering Techniques
spectively. (Although the atom probe de-
tects single atoms from the probed volume, The time evolution of the structure of
the requirement for the analysis to be sta- supersaturated alloys, as well as of oxides
tistically significant confines the microana- and polymer blends undergoing phase sep-
lytical spatial resolution to ≥ 2 nm.) As is aration, can be analyzed by means of small
shown in Figs. 5-2 a – c and 5-19 c, both the angle scattering of X-rays (SAXS), neu-
morphologies of small particles and their trons (SANS), and light. Light scattering is
three-dimensional arrangements can also of course confined to transparent speci-
be determined in the FIM, at least for pre- mens, in which the domain sizes of the
cipitates with sufficient contrast. evolving second phase must be of the order
The volume sampled during an atom of ≈ 1 µm. It has been successfully applied
probe FIM analysis is rather small (typi- in studies of decomposition, e.g., of a poly-
cally about 200 nm3). Hence, in order to mer mixture of polyvinyl methyl ether
obtain statistically significant data con- (PVME) and polystyrene (PS) (Snyder and
cerning the average size (R̄) and the num- Meakin 1983 a, b, 1985), and of various
ber density (Nv) of the precipitates, the lat- glass-forming oxide systems (Goldstein,
ter ought to exceed 1023 m–3. Since after 1965; Rindtone, 1975).
the early stages of precipitation Nv de- In principle, SAXS (Glatter, 1982) and
crease with time (see Sec. 5.2.1), AFIM SANS (Kostorz, 1979; Sequeira et al.,
shows its full potential as a microanalytical 1995) provide access to a structural analy-
tool in studies of the early stages of precip- sis of unmixing alloys, in both the early
itation, during which R̄ is commonly small stages where the composition fluctuations
and Nv sufficiently large (Haasen and Wag- can be small in spatial extension and in am-
ner, 1985). plitude, as well as in the later stages of de-
With respect to analyses of the spatial ar- composition where R̄ and Nv may have at-
rangement and the composition of precipi- tained values which are quite unfavorable
tated phases, the versatility of the AFIM for a quantitative analysis by TEM or FIM.
was considerably improved by the devel- In addition, evaluation of the Laue scatter-
opment of the tomographic atom probe ing allows the remaining supersaturation of
(TAP) (Blavette et al., 1993, 1998; Miller solute atoms in the matrix to be deter-
et al., 1996). This instrument allows for mined.
the three-dimensional reconstruction of a The metallurgist is left with the problem
small volume of the microscopically heter- of extracting in a quantitative manner all
ogeneous material on a sub-nanometric the structural data contained in scattering
5.3 Experimental Techniques for Studying Decomposition Kinetics 329
a) b)
Figure 5-21. a) Strain energy vs. aspect ratio of an Ag precipitate in Cu and Al matrices. The numbers 1, 4, 5
and 7 designate different orientation relationships between matrix and precipitate. The lowest strain energy is
obtained if the cubic directions of both particle and matrix are parallel to each other (curves 1), d is the linear
misfit. After Lee et al. (1977). b) The variation in a parameter, aR2, as a function of L. aR2 is used to characterize
the symmetry of a given family of equilibrium shapes: it is zero if the shape is four-fold symmetric and non-zero
if it is two-fold symmetric (Thompson et al., 1994).
particular, this holds true if the nucleating sions they found that at a critical size the
phase is a metastable transition phase. shape of a circular precipitate will change
In essence, however, the adopted shape to an ellipse in a smooth manner as the size
is determined by the balance between the of the precipitate is increased; the shape bi-
interfacial Helmholtz energy D Fa/b and the furcation is supercritical. A supercritical
elastic Helmholtz energy DFel. D Fa/b is bifurcation is analogous to the change in
minimized for particles with spherical or the order parameter on passing through a
faceted shape (e.g., Figs. 5-9 and 5-19) second-order phase transition. In contrast,
whereas DFel , which is related to the parti- in three dimensions, transcritical bifurca-
cle volume, commonly prefers a platelet- tions (first-order transitions) are possible.
like morphology (e.g., Fig. 5-21 a). Thus, Such a bifurcation implies that discontinu-
during the early stages of precipitation, ous changes in particle shape will occur as
D Fa/b is the dominating term, whereas the particle increases its size.
DFel prevails in the limit of large particles. The work of Johnson and Cahn clearly
The nature of this elastic stress-induced showed the importance of particle shape
precipitate–shape transition has been in- bifurcations in understanding microstruc-
vestigated by Johnson and Cahn (1984) us- tural evolution in systems with misfitting
ing bifurcation theory. Assuming an elasti- particles. Their study, however, was re-
cally isotropic inhomogeneous system and stricted to ellipsoidal particle shapes. This
by restraining the morphology of the parti- was remedied in the work of Thompson and
cle to be ellipsoidal it was possible to ad- Voorhees (1999) where the equilibrium
dress analytically the structure of the shape shape of a particle in an elastically aniso-
bifurcation. For example, in two dimen- tropic homogeneous system with isotropic
338 5 Homogeneous Second-Phase Precipitation
interfacial energy was determined numeri- cally anisotropic system with a tetragonal
cally, see Fig. 5-21 b. Here the equilibrium misfit, and Schmidt and Gross (1997)
shapes were found as a function of the di- found strongly non-convex global energy
mensionless parameter L = d 2 C44 l/sab minimizing, pincushion-like shapes when
where C44 is an elastic constant in a system the particle is sufficiently soft in an elasti-
with cubic elastic anisotropy, and l is the cally anisotropic system. Similar pincush-
size of a particle, e.g., l = 3 V1/3/4 p where ion-like precipitates have been observed
V is the volume of a precipitate, L is the ra- experimentally, but in a system with a
tio of the characteristic elastic to interfacial much smaller difference in elastic con-
energy. When the particle is small, interfa- stants (Maheshwari and Ardell, 1992). Em-
cial energy is the dominant factor in setting phasizing the difficulty in obtaining true
the particle shape and as L increases, the equilibrium shapes experimentally, Wang
effects of elastic stress become important. and Khachaturyan (1995) have suggested,
When L is small and the equilibrium shape however, that the shapes observed by Ma-
has a fourfold symmetry, a cube-like mor- heshwari and Ardell are kinetic in origin.
phology with rounded corners and rounded In an internally nitrided Fe 3 at.% Mo al-
sides will result. As L increases, at a criti- loy, (Fe, Mo)16N2-type precipitates were
cal size there is a supercritical bifurcation found to nucleate as thin platelets with
to one of two plate-like shapes oriented K ≈ 0.1 and with an undistored {100}a-Fe
along the elastically soft directions. How- plane common to both the a-matrix and the
ever, due to the presence of interfacial en- nitrides. The observed morphology, as well
ergy, the equilibrium shapes above the bi- as the {100}a-Fe habit planes, have been
furcation point are not plates for any finite accurately predicted by applying macro-
L. In the limit of infinite L the equilibrium scopic linear elastic theory to tetragonal
shape is an infinitely long plate of zero Fe16N2 precipitates in ferritic iron, and as-
thickness (Khachaturyan, 1983). The loca- suming the interfacial energy to be iso-
tion of the bifurcation point is a function of tropic (Khachaturyan, 1983). The latter as-
both the elastic anisotropy and the differ- sumption, however, is debatable, since dur-
ence in elastic constants between the ing aging at 600 °C (Fe, Mo)16N2 platelets
phases (Schmidt and Gross, 1997). Three- only grow markedly in diameter; their
dimensional calculations have also been thickness remains almost constant (Fig.
performed where the particles are not con- 5-22). This has been interpreted in terms of
strained to be of a simple geometric shape the large difference between the interfacial
(Mueller and Gross, 1998, Lee, 1997; energy of the habit plane (≈0.05 J/m2) and
Thompson and Voorhees, 1999). that of the peripheral plane (≈0.3 J/m2)
Another qualitative difference between (Wagner and Brenner, 1978).
the possible equilibrium shapes of a parti- Under certain conditions, the elastic or
cle in the presence of stress and that in the magnetic interaction of precipitates with
absence of stress is that the morphology of external stress fields or magnetic fields al-
misfitting particles can be non-convex. In lows for the generation of anisotropic,
the absence of stress, regardless of the type highly oriented precipitate microstructures
of interfacial energy anisotropy, the equi- which are sometimes of technological im-
librium shape must be convex. In contrast, portance. If the transformation strain is
Thompson and Voorhees (1999) found non-spherical, such as for Fe16N2 in ferritic
metastable non-convex shapes in an elasti- Fe–N, the particles may interact with an
5.5 Early Stage Decomposition Kinetics 339
tude and spatial extension (or number of at- scientist. For more recent developments,
oms) of a solute fluctuation which becomes see the chapter by Binder and Fratzl (2001).
critical and, hence, stable against decay.
Once formed, the critical fluctuations ren- 5.5.1 Cluster-Kinetics Approach
der the supersaturated alloy unstable with
5.5.1.1 Classical Nucleation –
respect to further unmixing.
Sharp Interface Model
According to Sec. 5.2.4 and referring to
Fig. 5-6 a, nucleation theories consider the Suppose the homogenized solid solution
formation rate of stable nuclei, the latter is quenched not too deeply into the meta-
representing spatially localized solute-rich stable regime of the miscibility gap (e.g., to
clusters (‘particles’ or ‘droplets’) with point 1 in Fig. 5-5 a). There it is isother-
large concentration amplitudes. In this con- mally aged at a temperature sufficiently
text, a distinction is made between classi- high for solute diffusion. After a certain
cal (Sec. 5.5.1.1) and non-classical nuclea- time, a distribution of microclusters con-
tion theory (Sec. 5.5.1.4) depending on taining i atoms (i-mers) will form in the
how the Helmholtz energy of the non-uni- matrix.
form solid solution containing the cluster Generally, classical nucleation is now
distribution is evaluated. Once the Helm- based on both a static and a dynamic part.
holtz energy has been specified, the equi- In the static part the changes of Helmholtz
librium distribution of heterophase fluctua- energy associated with the formation of an
tions and the nucleation barrier (Sec. 5.2.3) i-mer and the cluster distribution f (i ) must
can be calculated. The nucleation rate is be evaluated. In the dynamic part the kinet-
then obtained as the transfer rate at which ics of the decay of the solid solution which
smaller clusters attain the critical size. now is described by the given distribution
A ‘composition wave’ picture rather of non-interacting microclusters, are calcu-
than a ‘discrete droplet’ formalism is em- lated in terms of the time evolution of f (i );
ployed in the spinodal theories (Sec. 5.5.4) ultimately this will furnish the formation
which, as another approach, describe the rate of stable clusters, i.e. the nucleation
early-stage decomposition kinetics in rate.
terms of the time evolution of the ampli- Classical nucleation theory treats the so-
tude and the wavelength of certain stable lute fluctuations as droplets which were cut
‘homophase’ fluctuations (Fig. 5-6 b). from the equilibrium precipitate phase b
There are several papers dealing with the and embedded into the a matrix; in this
early-stage unmixing kinetics. Russel capillarity or droplet model the interface
(1980) and Aaronson and Russel (1982) between a and b is assumed to be sharp,
consider both classical and non-classical e.g., Fig. 5-6 a. In essence, the approxima-
nucleation phenomena from a more metall- tion reduces the number of independent
urgical point of view; the formulation of a variables which characterize a cluster, and
microscopic cluster theory of nucleation which may vary during the nucleation pro-
can be found in the article by Binder and cess (e.g., the solute concentration, the
Stauffer (1976). The comprehensive arti- atomic distribution within the cluster, its
cles by Martin (1978) and by Gunton and shape, the composition profile across the
Droz (1984) disclose developments of both interface, etc.) and, hence, determine the
nucleation and spinodal concepts not only Helmholtz energy of the system, to virtu-
to the theoretican but also to the materials ally a single one. That is the number i of
5.5 Early Stage Decomposition Kinetics 341
tion process to terminate itself under the expressions for D F *, Z, b* and t given
steady-state conditions if the incubation above for spherical clusters, have merely to
time t is shorter than some critical time tc. be multiplied with numerical factors but
If, however, t > tc , phase separation will be otherwise remain unchanged. The magni-
completed without steady-state ever having tude of these numerical factors depends on
been achieved. This situation, which in fact the particular equilibrium shape of the nu-
is met in most decomposing alloys studied cleus. Ignoring volume strain energy, the
up to now (see Sec. 5.7.4.1), is referred to equilibrium shape of the nucleus can be de-
as catastrophic nucleation; under such termined by means of the Wulff construc-
conditions D F *, Z and t become time de- tion on a three-dimensional sab-plot (Her-
pendent. Assuming that nucleation ceases ring, 1953). The latter is obtained from a
once the diffusion fields and around pre- radial plot of different vectors in every di-
cipitates with radius R and composition cbe rection e.g., v 1, v 2, v 3 in Fig. 5-25), the
embedded in a matrix with initial composi- length of which is proportional to the en-
tion c0 begin to overlap (Fig. 5-6 a), and ergy of the interface perpendicular to the
furthermore that nucleation becomes un- particular vector. The surface connecting
likely within a region R/e (e ≈1/10) around the tips of all vectors represents the polar
the particle where the solute content has sab-plot, wherein the cusps indicate inter-
decreased below (1 – e) c0, the critical time faces with good atomic matching , i.e., low
tc is estimated to be: interfacial energies. Subsequently planes
1/ 5
⎡ ⎛ ce ⎞ 3 ⎤
β 1 1 ⎥
tc = ⎢e 6 ⎜ ⎟ (5-31)
⎢ ⎝ c0 ⎠ ( 2 D)3 J s ⎥
2
⎣ ⎦
where D is the solute diffusion coefficient.
For a quantitative assessment of classi-
cal nucleation theory, the atomic impinge-
ment rate b* in Eq. (5-27) must be known.
For spherical nuclei it was evaluated to be
(Russell, 1970):
4 p R *2 Dc0
b* = (5-32)
a4
where a is the lattice parameter. b* is pro-
portional to the nucleus surface, as was as-
sumed in the original Becker–Döring the-
ory.
According to Sec. 5.4.2 the nucleus
shape with minimum energy may deviate
from the spherical one due to different en- Figure 5-25. Polar sab-plot for particles with aniso-
ergies of the interfaces (some may be co- tropic interfacial energies sab. The normals drawn at
herent, some semi- or incoherent, Sec. the tip of each vector v are the Wulff planes, the inner
envelope of which gives the equilibrium shape of the
5.2.1) bounding the nucleus in different precipitate. If the sab-plot displays deep cusps (e.g.,
crystallographic directions. It has been for v 1, v 3, and v 5) the Wulff construction yields a fac-
shown (Chan et al., 1978) that in this case eted polyhedron.
5.5 Early Stage Decomposition Kinetics 345
are drawn (the so-called Wulff planes) per- face between Al-rich matrix and Ag-rich
pendicular to the vectors v 1, v 2, v 3, etc. GP zones are found to be equal within 5%,
which intersect the sab-plot (Fig. 5-25). and to increase from zero to about
The inner envelope of the Wulff planes 30 mJ/m2 if the temperature is lowered
yields the equilibrium shape of the nucleus from Tc to ~ 0.5 Tc. This result is compat-
with facets at the cusps of the sab-plot ible with the spherical particle morphology
(Martin and Doherty, 1976). resolved in the HREM micrographs (Figs.
LeGoues et al. (1982) employed a near- 5-9 and 5-13).
est-neighbor interaction, regular solution
model for a computation of the sab-plot as
5.5.1.3 Experimental Assessment of
a function of T/Tc (Tc is the critical temper-
Classical Nucleation Theory
ature) within the miscibility gap of an f.c.c.
solid solution. At low temperatures they A quantitative assessment of classical
found the nucleus shape to be fully faceted nucleation theory in solids is inherently
by {100} and {111} planes whereas at 0.5 difficult and thus has prompted only a few
Tc and near-zero supersaturation, the nu- studies. First of all, the range of supercool-
cleus shape can be rather well approxi- ing (or supersaturation) has to be chosen
mated by a sphere. They furthermore such that nucleation is homogeneous and
showed that beyond T ≈0.4 Tc the parame- the nucleation rates are neither unmeasure-
ters DF *, Z, t, and b* which enter the time- ably slow nor beyond the limits of catas-
dependent (classical) nucleation rate (Eqs. trophic nucleation. Secondly, the quench-
(5-27) and (5-29)) need no longer be cor- ing rate must be sufficiently high in order
rected for deviations from spherical shape. to avoid phase separation during the
Recently the question of to what extent quench but also sufficiently low in order to
the energy of the coherent interphase boun- avoid excess vacancies to be quenched in.
dary depends on its crystallographic orien- Servi and Turnbull (1966) studied in a
tation and on temperature has been read- well-designed experiment homogeneous
dressed by employing the cluster variation nucleation kinetics of coherent Co-rich
method (CVM) with effective cluster inter- precipitates in Cu–1 … 2.7 wt.% Co al-
action parameters (Sluiter and Kawazoe, loys. By using electrical resistivity meas-
1996; Asta, 1996). The latter are specific urements, they could determine rather ac-
for the particular two-phase alloy system curately the precipitated volume fraction.
under consideration. They can be derived Assuming that the growth is diffusion-con-
from the results of ab initio total energy trolled (Sec. 5.5.2), from the latter the par-
calculations as was demonstrated by com- ticle density at the end of the precipitation
puting the thermodynamic properties of process could be derived as a function of
interphase boundaries between disordered aging temperature and composition. The
f.c.c. matrices and ordered (L12 type) pre- thus indirectly obtained number density of
cipitates in Al–Li (Sluiter and Kawazoe, Co-rich particles, which was later corrobo-
1996) and Al–Sc (Asta et al., 1998) as well rated by CTEM studies, agreed within one
as disordered GP zones (cf. Fig. 5-13) in order of magnitude with the value pre-
Al–Ag alloys (Asta and Hoyt, 1999). dicted by classical nucleation theory if the
Between Tc and T = 0.5 Tc the CVM-calcu- interfacial energy was taken as ≈0.19 J/m2;
lated energies for the cube {100} and octa- this value was derived from discrete lattice
hedral {111} orientations of the flat inter- calculations (Shiflet et al., 1981).
346 5 Homogeneous Second-Phase Precipitation
core of the nucleus where the composition Although some caution is advisable in
is about constant (Fig. 5-27). treating these tiny particles (critical nuclei
If we assume that a perceptible nuclea- in solid-state transformation are tens to
tion rate is obtained for D F * ≤ 25 kT (in hundreds of atoms in size) in terms of a
the literature, DF * ≈ 60 kT is frequently continuum theory1, and assigning them a
stated for the maximum barrier at which macroscopic surface and thermodynamic
nucleation becomes measurable; this value, bulk properties, classical nucleation theory
however, is much too high for homogene- seems to be appropriate for an estimation
ous nucleation to occur in solids) and em- of the nucleation rate. This conclusion will
ploys Eqs. (5-19) and (5-20), classical nu- be corroborated in Sec. 5.7.4.
cleation theory may be applied if (Cahn
and Hilliard, 1959 a, b): 5.5.2 Diffusion-Controlled Growth
of Nuclei from the Supersaturated Matrix
4 p s ab
l2 Ⰶ1 (5-38) Suppose that after nucleation the stable
75 kT nucleus is embedded in a still supersatu-
rated matrix. As is illustrated in Fig. 5-28,
For Cu–Ti (s ab = 0.067 J/m2; T = 623 K), the particle will then be surrounded by a
Fe–Cu (s ab = 0.25 J/m2; T = 773 K), and concentration gradient which provides the
Cu–Co (s ab = 0.17 J/m2; T = 893 K) this driving force for solute diffusion, and thus
means that the interfacial width has to be gives rise to its growth.
smaller than ≈ 0.56, 0.33, and 0.42 nm, re- The growth rate can be controlled either
spectively, and, hence, has to be rather by the rate at which atoms are supplied to
sharp. So far it has not been possible to the particle/matrix interface by diffusion or
measure the composition profile across a by the rate at which they cross the interface
nucleus/matrix interface. However, analyt- (Tien et al., 1973). It may be rationalized
ical field ion microscopy of Cu–1.9 at.% Ti that for small particles the interface reac-
aged at 350 °C for 150 s revealed the Ti tion is likely to be the rate-controlling step
concentration of particles with radii of only since the diffusion distances are rather
≈1 nm to decrease from 20 at.% (corre- short; once the particles have grown to a
sponding to Cu4Ti, cf. Sec. 5.2.1) to that of certain size, the matrix will be depleted
the matrix within one to two atomic (111)- from solute atoms and the associated re-
layers (von Alvensleben und Wagner, duction of the driving force makes diffu-
1984). For the chosen aging conditions nu- sion likely to be the slower and, thus, rate-
cleation was evaluated to terminate within controlling step (Shewmon, 1965). The
about 60 s (Kampmann and Wagner, 1984), transition from one step to the other de-
thus after aging for 150 s the analyzed par- pends upon the relative magnitudes of so-
ticles must have already experienced some lute diffusion and interface mobility.
growth beyond the original size of the nu- After termination of nucleation it is com-
cleus. Nevertheless, these results provide monly assumed that the mobility of the
rather good evidence that during nucleation interface is sufficiently high in order to al-
the Cu4Ti clusters may be considered drop-
lets with sharp interfaces rather than spa-
1
tially extended (long-wavelength) fluctua- Many FIM studies of two-phase alloys (e.g., Cu–1
at.% Fe) revealed the shapes of even tiny clusters
tions with diffuse interfaces, even though with as little as about twenty atoms to be already
DF * is only about 10 kT. compact rather than ramified (Wagner, 1982).
5.5 Early Stage Decomposition Kinetics 351
By assuming a monodispersive particle into a bulk and a surface term appears rather
distribution and cR ≈ c̄ (t Æ •) = cae (or c¢ae), debatable. To avoid some of these deficits,
and c̄ (t) ≈ c0 , integration of Eq. (5-43) yields there have been several attempts to develop
1/ 2 more accurate descriptions of clusters, tak-
⎛ c – ce ⎞
R (t ) = ⎜ 2 0 ae ⎟ ( D t )1 / 2 (5-47) ing into account their different sizes and
⎝ cp – ca ⎠ shapes (e.g., Binder and Stauffer, 1976;
It should, however, be pointed out that, so Stauffer, 1979), and to derive an equilibrium
far, there is little experimental evidence for distribution of clusters which is more realis-
the existence of a precipitation regime dur- tic than the one given by Eq. (5-23). As in
ing which particle growth strictly follows the classical nucleation theory, the latter is
Eq. (5-47) (Kampmann and Wagner, 1984). an important quantity for a computation of
In Sec. 5.7.4.3 the reason for this lack of the cluster formation rate in terms of the
experimental evidence will be provided, as more recently elaborated theories of cluster
well as a guideline to the design of experi- dynamics which, in essence, are extensions
ments that allow a verification of the exis- of the Becker–Döring theory (Binder and
tence of diffusion-controlled particle Stauffer, 1976; Penrose and Lebowitz,
growth with the predicted kinetics to be 1979; cf. Gunton et al., 1983 and the chapter
made. Most quantitative experimental ob- by Binder and Fratzl (2001) for a general
servations of growth rates which yielded discussion of the various developments).
good agreement with the diffusion-con- Unlike Becker and Döring, however,
trolled models outlined above, were con- who confined the microscopic mechanism
fined to large particles, frequently with of cluster growth or shrinkage to the con-
sizes in the micrometer range (Aaronson et densation or evaporation of monomers
al., 1970 a, b; Doherty, 1982). (Sec. 5.5.1.1), in their generalized nuclea-
tion theory Binder and coworkers consider
the time evolution of the cluster size distri-
5.5.3 The Cluster-Dynamics Approach
bution function f (i, t) more generally in
to Generalized Nucleation Theory
terms of a cluster coagulation or cluster
As has been pointed out in Sec. 5.5.1.1, splitting mechanism, i.e., in a single cluster
the cluster-dynamics approach chosen by reaction step i-mers (i = 2, 3, …) are also
Becker and Döring (1935) is based on the allowed to be added or separated from an
assumption that the equilibrium cluster dis- existing cluster (e.g., Binder, 1977; Mirold
tribution function can be specified by Eqs. and Binder, 1977). Thus, the related kinetic
(5-23) and (5-18). Even though the result- equation for the time evolution of f (i, t)
ing predictions of classical nucleation the- (Binder, 1977),
ory ought to become valid asymptotically ∞
d
for large droplets at small supersaturations, f (i, t ) = ∑ ai + i ′, i ′ f (i + i ′, t ) (5-48)
dt i′ = 1
there remain some inherent deficits
i –1 i –1
(Binder, 1980; Gunton et al., 1983). 1 1
– ∑ ai, i ′ f (i, t ) + ∑ bi – i ′, i ′
For instance, for small clusters the sur- 2 i ′= 1 2 i ′= 1
face energy entering Eq. (5-18) should ∞
contain size-dependent corrections to the × f ( i ′, t ) f ( i – i ′, t ) – ∑ bi, i ′ f (i, t ) f (i ′, t )
i ′= 1
macroscopic equilibrium energy of a flat
interface. Furthermore, the separation of contains four different terms. The first ac-
the droplet formation Helmholtz energy counts for an increase of i-mers due to
5.5 Early Stage Decomposition Kinetics 353
The other symbols in Eq. (5-51) were de- where S ( k, 0) denotes the equal-time
fined in Sec. 5.5.1.4. structure function of the system equili-
As M is always positive, D̃ takes the sign brated at the homogenization temperature.
of ∂2 f /∂c2 and so is negative inside the spi- Inside the spinodal region (∂2 f /∂ c2)|c0 is
nodal regime (Sec. 5.2.2), giving rise to an negative. Thus for a given negative value
‘uphill diffusion’ flux of solute atoms (cf. of the latter, R ( k) becomes positive for all
Fig. 5-6 b). wavenumbers k satisfying
The linearized version of the more gen- ⎞
⎛ ∂2 f
eral nonlinear diffusion equation holds if k 2 < k c2 = ⎜ 2 – 2h 2 Y ⎟ 2 K * (5-58 a)
the amplitude c (r) – c0 of the composition ⎝ ∂c c0 ⎠
fluctuation is rather small and both M and or, correspondingly, for all fluctuation
(∂2 f /∂ c2)|c0 are independent of composi- wavelengths l larger than lc with
tion. These approximations inherently con-
fine the Cahn–Hilliard theory (CH theory) ⎛ ∂2 f ⎞
to the earliest stages of phase separation. l2c = 4 p K * ⎜ 2 – 2h 2 Y ⎟ (5-58 b)
⎝ ∂c c0 ⎠
By expanding the atomic distribution
c (r, t) into a Fourier series, Eq. (5-51) can Long wavelength fluctuations with l > lc
be written as thus will be amplified exponentially
∂A (k ) whereas short wavelength fluctuations will
= R (k ) A (k ) (5-53) decay with aging time. For lm ∫ 앀옽2 lc , R (l)
∂t
attains a maximum giving rise to fastest
with the amplitude growth of fluctuations with wavelength lm.
On approaching the coherent spinodal
A (k ) = ∫ [c ( r ) – c0 ] exp (i k r ) dr (5-54) curve where the denominator in Eq.
(5-58 b) becomes zero (cf. Sec. 5.5.3), the
of a Fourier component with wavenumber
critical wavelength lc diverges in a similar
k = 2p /l if l denotes the wavelength of the
manner as does the critical radius R* (Fig.
composition fluctuation. Eq. (5-51) has a
5-30). Thus both the linear theory on spino-
solution for every A ( k, t) with
dal decomposition and the non-classical
A (k , t ) = A (k , 0 ) exp [ R (k ) t ] (5-55) nucleation theory (Sec. 5.5.1.4) predict
abrupt changes in the decomposition kinet-
where A ( k, 0) is the initial amplitude of ics on crossing the spinodal line from ei-
the Fourier component with wavenumber ther side. This has prompted several ex-
k . The so-called amplification factor R ( k) perimental studies to determine the spino-
358 5 Homogeneous Second-Phase Precipitation
from kinetic studies of decomposition in curve exists for these solid mixtures (cf.
various polymer mixtures, some of which Binder and Fratzl (2001), Sec. 6.2.5, this
are cited in Table 5-1 (for reviews, cf. Gun- volume). Thus in general, it is currently not
ton et al., 1983; Binder and Fratzl (2001), feasible to predict on theoretical grounds
Chapter 6 of this volume), were found to be whether a metallic, glass or oxide solid
consistent with the predictions of the line- mixture has been quenched into the meta-
arized spinodal CHC theory (e.g., Okada stable or unstable region of the miscibility
and Han, 1986). In contrast, apart from a gap. In this context, it remains debatable
SANS study on the kinetics of phase separ- whether each of the metallic, glass or oxide
ation in Mn–33 at.% Cu (Gaulin et al., mixtures that were studied by means of
1987), none of the scattering experiments SAXS or SANS, and which have been re-
on metallic alloys or oxides which are listed ported to undergo spinodal decomposition
in Table 5-1 and which probably had been (Table 5-1), were truly quenched into and
quenched into the unstable region of the aged within the spinodal region of the
miscibility gap, corroborated the predic- phase diagram. Nevertheless, extensive ex-
tions of the CHC theory. SANS curves perimental SAXS or SANS studies, in par-
taken from Mn–33 at.% Cu during short- ticular on Al–Zn alloys and Fe–Cr (Table
term aging (between 65 s and 210 s at 5-1), did not reveal any evidence for a dras-
450°C) could be fitted to the CHC structure tic change in the time evolution of the S (k, t)
function with km being time-independent curves with a change in the initial supersat-
during this period. Thus, it was concluded uration. For the Al–Zn systems in particu-
that phase separation in Mn-33 at.% Cu is of lar, it was shown (Simon et al., 1984) that
the spinodal type and the kinetics follow the the clustering rate increases rapidly with
CHC predictions (Gaulin et al., 1987). The increasing supersaturation, i.e. both km and
S (k) curve of the as-quenched state already Sm are larger the deeper the quench.
revealed a peak, in fact, at a smaller value of Extensions of the Cahn–Hilliard theory
km than for the aged samples. This is indic- of spinodal decomposition to ternary or
ative of the fact that phase separation oc- multicomponent systems have been elab-
curred during the quench owing to a finite orated by de Fontaine (1972, 1973) and
quench rate. In this case, as has been exten- Moral and Calm (1971). However, to our
sively discussed by Hoyt et al. (1989), km knowledge the kinetic predictions from
may indeed become initially time-indepen- these extensions have not yet been com-
dent, thus pretending an experimental con- pared with experimental results, probably
firmation of the CHC linear approximation. because of difficulties in determining the
In the framework of the mean-field theo- partial structure functions of ternary alloys
ries the spinodal curve is well defined as (cf. Sec. 5.3.1.2).
the locus where ∂2 f /∂ c2 vanishes (Sec. As has been pointed out in Sec. 5.2.3,
5.2.2). It can then be easily evaluated, e.g., the formulation of a ‘unified theory’ com-
from an extrapolation of the thermody- prising both nucleation and growth as well
namic data which are known for the single- as spinodal decomposition can also be at-
phase solid solution, into the (‘mean-field’) tacked on grounds of a spinodal theory.
spinodal regime (Hilliard, 1970). Since, This has been attempted in the statistical
however, neither metallic alloys nor glass model of Langer, Bar-On and Miller
or oxide mixtures behave as mean-field- (1975), which takes into account thermal
like systems, no uniquely defined spinodal fluctuations and nonlinear terms which are
360 5 Homogeneous Second-Phase Precipitation
Table 5-1. Some small angle scattering experiments – sometimes jointly employed with other microanalytical
tools for studying the morphology – primarily designed for an investigation of spinodal decomposition. If not
otherwise stated the concentrations are in at.% (N.d.: not determined.)
Metallic alloys
Al–4 wt.% Cu SAXS Modulated structure Naudon et al. (1976)
Al–22% Zn SAXS Gerold and Merz (1967)
Al–22% Zn SAXS, TEM Modulated structure Agarwal and Herman (1973)
Al–20.7 … 49.1% Zn SAXS Bonfiglioli and Guinier (1966)
Al–5.3 … 6.8% Zn SANS GP-zones Hennion et al. (1982)
Al–5.3 … 12.1% Zn SANS, TEM Guyot and Simon (1981)
Simon et al. (1984)
Al–22% Zn–0.1 Mg SAXS, TEM Interconnected spherical Forouhi and de Fontaine
clusters at early times; (1987)
regularly spaced platelets
at later stages
Al–12 … 32 Zn SAXS (synchro- Hoyt et al. (1987, 1989)
tron radiation)
Al–2.4% Zn–1.3 Mg SANS Blaschko et al. (1982, 1983)
Al–3.8 … 9% Li with SANS N.d. Pike et al. (1989)
additions of Cu, Mn
Au-60% Pt SANS Singhal et al. (1978)
Fe-28% Cr–10% Co SANS, TEM, AFIM ‘Sponge-like’ structure Miller et al. (1984)
Fe–34% Cr SANS N.d. Katano and Iizumi (1984)
Fe–30 … 50% Cr SANS N.d. Ujihara and Osamura (2000)
Fe–29.5% Cr–12.5% Co Anomalous SAXS N.d. Simon and Lyon (1989)
(synchrotron source)
Fe–52% Cr SANS N.d. La Salle and Schwartz (1984)
Fe–20 … 60% Cr SANS N.d. Furusaka et al. (1986)
Cu–2.9% Ti SANS, TEM Modulated structure Eckerlebe et al. (1986)
CuNiFe SANS Modulated structure Aalders et al. (1984)
CuNiFe SANS, TEM, AFIM Mottled structure Wagner et al. (1984)
CuNiFe Anomalous SAXS N.d. Lyon and Simon (1987)
(synchrotron source)
Cu–2% Co SANS N.d. Steiner et al. (1983)
Mn–25 … 52% Cu SANS (analysis of N.d. Vintaikin et al. (1979)
integrated intensity)
Mn–33% Cu SANS N.d. Gaulin et al. (1987)
Ni–13% Al SANS Anisotropic clustering Beddoe et al. (1984)
Ni–12.5% Si SANS, SAXS, TEM Modulated structure Polat et al. (1986, 1989)
(‘side bands’)
Ni–11.5% Ti SANS, TEM Modulated structures Cerri et al. (1987, 1990)
5.5 Early Stage Decomposition Kinetics 361
Glasses, oxides a
B2O3–(15 wt.% PbO SAXS Craievich (1975)
– 5 wt.% Al2O3) Acuña and Craievich (1979)
(quasi-binary system) Craievich and Olivieri (1981)
B2O3 – (27 wt.% PbO SAXS Craievich et al. (1986)
– 9 wt.% Al2O3) (synchrotron source)
Vycor glass SANS ‘Sponge-like’ structure Wiltzius et al. (1987)
consisting of a SiO2-rich
and a B2O3 -alkali
oxide-rich phase
SiO2–13 mole% Na2O SAXS, TEM ‘Sponge-like’ structure Neilson (1969)
TiO2–(20–80 mole%) SnO2 SAXS, TEM, X-ray Lamellar modulations Park et al. (1976)
diffraction (‘side along [001] (Fig. 5-38)
band’ analysis)
Polymer mixtures
Critical mixture of per- Light scattering Wiltzius et al. (1988)
deuterated and protonated Bates and Wiltzius (1989)
1,4-polybutadiene
polybutadiene and Izumitani and Hashimoto (1985)
styrene–butadiene
copolymer mixtures
polystyrene–polyvinyl- Light scattering Sato and Han (1988)
methylether (PS–PVME) Sato et al. (1989)
Snyder et al. (1983 a, b)
Hashimoto et al. (1986 a, b)
Okada and Han (1986)
a
For a comprehensive survey on oxides and glasses up to 1978, see Jantzen and Herman (1978).
neglected in the CH theory. The LBM non- With certain approximations, nonlinear ef-
linear theory of spinodal decomposition fects and thermal fluctuations are con-
also accounts for the later stages of phase tained in the term A* (t) and in the last
separation, though not for the coarsening term, respectively. Setting A* (t) ∫ 0 repro-
regime. If coherency strains are neglected duces the corresponding equation of mo-
it approximates the kinetic evolution of the tion of the CHC theory, and, if the fluctua-
structure function as tion term is also omitted, Eq. (5-57) of the
original linear CH theory is regained.
dS (k , t ) M 2
=–2 k Langer et al. (1975) proposed a computa-
dt nv tional technique for solving Eq. (5-60).
⎡⎛ ∂2 f ⎞ ⎤ This required several approximations to be
× ⎢⎜ K * k 2 + 2 ⎟ S (k ) + A * (t )⎥ made, e.g., on f (c) and on M, which inher-
⎢⎣⎝ ∂c c0⎠ ⎥⎦
ently leaves the LBM approach with some
M
+2 k T k2 (5-60) fundamental shortcomings. Nevertheless,
nv with respect to the features displayed in the
362 5 Homogeneous Second-Phase Precipitation
time evolution of S (k, t), the predictions of region of the miscibility gap. Furthermore,
the LBM theory (Fig. 5-35) are in rather the various spinodal theories are difficult to
good agreement with both experimental handle and many parameters in the result-
studies (e.g., Fig. 5-32 a and 5-34) and ing kinetic equations are often not avail-
Monte Carlo studies (Sec. 5.5.6, Fig. 5- able for most solid mixtures. Hence, like
41). the non-classical nucleation theory, the
In their study of phase separation in various elegant theories that describe the
Fe–(20, 30, 40 and 60) at.% Cr alloys, Fu- kinetics of phase separation of a ‘spinodal
rusaka et al. (1986) compared the experi- alloy’ are of little use for the practical met-
mental S (k, t) curves with Eq. (5-60) of allurgist.
the LBM theory. For shorter aging times at Because of these problems, materials
515 °C, the experimental data points taken scientists have up to now employed mor-
from Fe–40 at.% Cr could well be fitted by phological criteria for the definition of a
Eq. (5-60) (Fig. 5-36) if k T/K*, M K*, and ‘spinodal alloy’. These are simply related
A*/K* were employed as three independent to the predictions of the linear spinodal CH
fitting parameters. On the grounds of the theory concerning the morphological evo-
agreement with the spinodal theory of lution of a solid mixture undergoing spino-
LBM, and additionally, as this particular dal decomposition.
alloy was evaluated to lie within the Most crystalline solid solutions show a
(mean-field) spinodal region, it has been variation of lattice parameter with compo-
concluded that Fe–40 at.% Cr is a ‘spino- sition leading to coherency strains. The as-
dal alloy’. sociated strain energy fel = h 2 Y (c – c0)2
Based on a similar LBM analysis of (Eq. (5-34)), which is accounted for in Eq.
SANS data from Fe–30 … 50 at.% Cr, Uji- (5-51), reduces the driving force for phase
hara and Osamura (2000) corroborated this separation. This effect shifts the locus of
conclusion recently. the original chemical (mean field) spinodal
The nonlinear theory of spinodal decom- to lower temperatures, yielding the coher-
position developed by Langer (1971) for ent spinodal curve (Eq. (5-58)). If the pa-
binary alloys has also been extended to ter- rameter Y (cf. Sec. 5.2.4), which is a com-
nary substitutional systems (Hoyt, 1989) bination of various elastic constants, de-
by deriving the time-dependent behavior of pends on the crystallographic direction, fel
the three linearly independent partial struc- also becomes anisotropic. Therefore, the
ture functions (cf. Sec. 5.3.1.2). An experi- locus of the coherent spinodal may also
mental examination of the kinetic predic- vary with the crystallographic direction.
tions, however, is still lacking. This becomes particularly discernible for
the tetragonal TiO2–SnO2 oxide system
5.5.5 The Philosophy of Defining (Park et al., 1975, 1976), where fel attains a
a ‘Spinodal Alloy’ – minimum for composition waves along
Morphologies of ‘Spinodal Alloys’ [001] and is larger for waves along [100]
and [010]. Hence, the coherent spinodal
Referring to the previous section it is splits up into a [001], a ·101Ò, and a ·100Ò
rather difficult or even impossible to assess branch (Fig. 5-37). As a consequence the
on a thermodynamic basis whether an al- SnO2-rich modulations form preferentially
loy, glass or oxide system was truly along [001], giving rise to a lamellar struc-
quenched into and aged within the spinodal ture at later aging stages (Fig. 5-38). For
364 5 Homogeneous Second-Phase Precipitation
most cubic metallic systems the elastic an- less steel that undergoes phase separation
isotropy parameter A (Sec. 5.2.4 ii) is posi- during tempering between 350 °C and
tive, rendering the minimum fel along the 450 °C with an associated embrittlement
elastically soft ·100Ò directions. According (Auger et al., 1989). Increasing the Cr con-
to the CH theory the growth rate will thus tent of the ferritic Fe–Ni–Cr solid solution
be highest along the three ·100Ò directions, leads to the formation of Cr-rich a¢-precip-
giving rise to the frequently observed mod- itates during tempering. As the a¢-phase is
ulated precipitate microstructures (Table discernible as individual particles by
5-2, Figs. 5-18, 5-19 b). In isotropic materi- CTEM, it is concluded that phase separa-
als such as polymers, glasses, or, for in- tion occurred via nucleation and growth. A
stance, Fe–Cr and Fe–Cr–Co alloys, the further increase in the Cr content, to about
modulations do not grow along any prefe- 25 wt.%, yielded a ‘sponge-like’ micro-
rential directions. The resulting two-phase structure after tempering below 400 °C.
microstructure is of the ‘sponge-like’ type This was attributed to a spinodal mecha-
and sometimes referred to as a ‘mottled’ or nism. Thus the spinodal is drawn as the line
‘interconnected’ precipitate microstructure that separates the two morphologies (Fig.
(e.g., Fig. 5-16, 5-19 c, 5-20). Based on 5-39).
these considerations, two-phase alloys dis- It must be pointed out, however, that
playing either lamellar, modulated or interconnected or modulated structures
interconnected precipitate microstructures represent two-phase microstructures in the
are commonly termed ‘spinodal alloys’ by later stages of the reaction (e.g., Fig. 5-18).
metallurgists. Furthermore, the morphol- Even though they are widely believed to
ogy of a two-phase material is sometimes result from spinodal decomposition in the
employed to derive the locus of the spino- sense of the CH theory, their formation
dal. This is illustrated in Fig. 5-39 for the might be of a rather different origin. For in-
Cr-rich ferrite phase of a cast duplex stain- stance, a strong elastic interaction of a high
5.5 Early Stage Decomposition Kinetics 365
Table 5-2. TEM and AFIM studies on phase separating solids that revealed ‘spinodal precipitate
microstructures’. If not otherwise stated the concentrations are in at.%.
Metallic alloys
Al–4 wt.% Cu TEM (‘side-bands’) Modulated structure Rioja and Laughlin (1977)
Al–(2.4 –3%) Li HREM, Radmilovic et al. (1989)
X-ray diffraction
Cu3–4MnxAl TEM (‘side-bands’) Modulated structure; Bouchard and Thomas (1975)
at later stages
Cu2MnAl + Cu3Al-plates
Cu–(1.5–5.2 wt.%) Ti TEM (‘satellite Modulated structure Laughlin and Cahn (1975)
analysis’)
Cu–2.7% Ti AFIM Modulated structure Biehl and Wagner (1982)
Ni–12% Ti AFIM Modulated structure Grüne (1988)
Ni-29% Cu–21% Pd TEM Modulated structure Murata and Iwama (1981)
Ni-base superalloys AFIM Short wavelength Bouchon et al. (1990)
(≈ 2.5 nm) Cr fluctuation
in g-phase
Nimonic 80 A TEM (‘side-bands’) Modulated structure Wood et al. (1979)
Fe–25% Be AFIM Modulated structure Miller et al. (1984, 1986)
CuNiFe AFIM Interconnected percolated Piller et al. (1984)
structure (cf. Fig. 5-19 c)
CuNiFe TEM (‘side-bands’) Modulated structure Wahi and Stager (1984)
(cf. Fig. 5-18) Livak and Thomas (1974)
Co–10% Ti FIM Modulated structure Davies and Ralph (1972)
Co–3 wt.% Ti TEM (‘side-bands’) Modulated structure Singh et al. (1980)
– 1 … 2 wt.% Fe
Mn–30 wt.% Cu TEM, magnetic Interconnected Yin et al. (2000)
susceptibility structure
Steels
Cast Duplex stainless TEM, AFIM After decomposition Auger et al. (1989)
original ferritic
Fe–Cr–Ni shows
‘sponge-like’ structure
Ferrous martensite
Fe–15 wt.% Ni–1 wt.% C TEM Tweed structure Taylor (1985)
Fe-25 wt.% Ni TEM Tweed structure Taylor (1985)
– 0.4 wt.% C
Fe–13 … 20 wt.% Mn X-ray diffraction Modulated structure at Miyazaki et al. (1982)
(‘side-bands’), TEM early times; isolated
Fe2Mo particles at later
stages
Amorphous alloys
Ti50Be40Zr10 AFIM Wavy composition profile Grüne et al. (1985)
5.5 Early Stage Decomposition Kinetics 367
Glasses, ocides
TiO2–60 mole% SnO2 TEM Tetragonal system with Stubican and Schultz (1970)
lamellar modulations
along [001]
TiO2–50 mole% SnO2 HREM Tetragonal system with Horiuchi et al. (1984)
lamellar modulations
along [001]
SiC–(50–75 mole%) AlN TEM (‘satellites’) Modulated structure Kuo and Virkar (1987)
the atomic exchange is assumed to occur close to the spinodal line within the meta-
directly rather than indirectly via the va- stable regime, evolves similarly to that of
cancy mechanism. Nevertheless, MC simu- ‘alloy’ # 5 quenched into the center of the
lations based on Kawasaki dynamics on an spinodal region (Fig. 5-41c). Thus, in
atomic scale yielded the time evolution of agreement with the generalized nucleation
the cluster configuration and the structure theory and the LBM spinodal approach, but
function S (k, t) of model alloys quenched unlike the predictions of linearized CH the-
into any region below the solubility line ory of spinodal decomposition, MC simu-
(Fig. 5-40) without worrying about com- lations again reveal i) no evidence for any
plicating factors, such as an insufficient abrupt change of the decomposition kinet-
quenching rate, excess vacancies, or lattice ics on crossing the spinodal curve, ii) no
defects, which we commonly face when in- common cross-over point of the S (k)
vestigating real alloys (Sec. 5.3.2). Fur- curves taken after different aging times, iii)
thermore, MC simulations have allowed a no exponential growth of the scattering in-
critical examination of the various theoret- tensity for a certain time-independent wave
ical approaches in terms of cluster dynam- vector in any time regime, and, further-
ics models or spinodal models to be made. more, iv) the peak position of S (k, t) at km
For practical computational limitations, is not found to be time-independent but is
however, the maximum size of the model shifted towards smaller values of k indicat-
alloy has commonly been restricted to ing the immediate growth of clusters.
about 50 ¥ 50 ¥ 50 lattice sites. Because of Frequently, the MC data for the time ev-
the size limitation, in general MC kinetic olution of the peak position (km) and the
experiments can only cover the earlier peak height (Sm) of the structure function
stages of a precipitation reaction in alloys have been fitted to simple power laws (cf.
where the supersaturation is sufficiently Sec. 5.8.2), such as km (t) µ t –a and Sm (t)
high for the formation of a large number d t b ; a and b were estimated to range from
density of clusters or nuclei, and where the 0.16 to 0.25 and 0.41 to 0.74, respectively,
cluster sizes are still considerably smaller depending on the initial supersaturation of
than the linear dimension of the model the ‘alloy’ (Marro et al., 1975, 1977; Sur et
system. This is frequently not the case in al., 1977). Lebowitz et al. (1982) pointed
real alloys. out, however, that due to the finite (small)
In essence, the results from MC simula- size of the system it is difficult to extract
tions based on Kawasaki dynamics have re- from computer simulations precise and re-
covered the predictions both from the gen- liable information about the analytical
eralized nucleation theory of Binder and form of km (t), for example and that it is
coworkers (Sec. 5.5.3) and from the non- possible to fit the same MC data with other
linear spinodal LBM approach. In particu- functional forms than the power laws given
lar, the time evolution of the structure func- above. (In Sec. 5.8.2 we will show, in fact,
tion in the kinetic Ising model displays that apart from the late stages of coarsen-
qualitatively the same features as the corre- ing, it is usually not feasible to interpret ex-
sponding ones obtained from the latter the- perimental kinetic data over an extended
ories (cf. Figs. 5-31 and 5-35 b) or from ex- aging period in terms of a power-law be-
periment (e.g., Fig. 5.32 a). havior with a time-independent exponent.)
For instance, S (k, t) of ‘alloy’ # 4 (Fig. The Kawasaki dynamics are based on an
5-41 b), which according to Fig. 5-40 lies unrealistic exchange mechanism between
5.5 Early Stage Decomposition Kinetics 369
neighboring atoms. In order to account for and Penrose, 1994). Comparing the growth
the more realistic case of atomic diffusion rates of clusters in the Ising model of a
being based on the vacancy mechanism, two-dimensional model alloy showed that
the Ising model was more recently ex- the asymptotic regime of cluster coarsen-
–
tended to binary alloys comprising vacan- ing (R ~ t1/3; cf. Sec. 5.6.2) is approached
cies (Yaldram and Binder, 1991 a,b; Fratzl much faster with vacancy dynamics than
370 5 Homogeneous Second-Phase Precipitation
with Kawasaki dynamics, in particular at bility cR (R) in the presence of small parti-
larger supersaturations; the cluster shapes cles with a large ratio of surface area to
are found not to depend on the particular volume is larger than that for larger ones.
dynamics model (Fratzl and Penrose, With reference to Eq. (5-43), this leads to a
1994). size-dependent growth rate, which is posi-
In a first attempt to interpret experimen- tive for larger particles with c̄ > cR and neg-
tal data with kinetic data obtained from ative for smaller ones with c̄ < cR. The
MC simulations, Soisson et al. (1996) sim- growth rate becomes zero for particles with
ulated phase separation in a Fe–1.34 at.% c̄ = cR which are in unstable equilibrium
Cu alloy. The alloy was modeled as a rigid with the matrix. Their radius R* is derived
b.c.c. crystal with 2 ¥ L3 lattice sites (L var- from Eq. (5-46) as (5-61)
ying from 32 to 64) containing one single
2 s ab Vb 1 1
vacancy; the energetics and the parameters R* ≈ ≡ K ′ s ab
of the atomistic kinetic model were ad- Rg T ln ( c /cae ) ln ( c /cae )
justed to the thermodynamic and diffusion Hence, driven by the release of excess
data of the Fe–Cu system. The MC results interfacial energy, larger precipitates will
confirmed earlier results from magnetic grow at the expense of smaller ones which
SANS studies (Beaven et al., 1986) that the dissolve again given rise to a change in the
precipitating clusters are pure copper; fur- precipitate size distribution. This process,
thermore the time evolution of the precipi- which is commonly referred to as coarsen-
tated volume fraction agreed quite well ing or Ostwald ripening2, frequently re-
with the one experimentally determined duces the precipitate number density form
from resistivity measurements (Lê et al., ≈ 1025 m–3 to less than 1019 m–3 in typical
1992). In the later stages dynamical scaling two-phase alloys during aging (cf. Fig.
behavior was shown to hold, indicating that 5-3). Usually, the coarsening process is
the cluster pattern remains similar as aging considered to be confined to the latest
time increases (cf. Sec. 5.8.1). stages of a precipitation reaction. However,
as will be shown in Sec. 5-7, coarsening
may accompany the growth process out-
5.6 Coarsening of Precipitates lined in Sec. 5.5.2, or may even start while
the system is still in its nucleation period,
5.6.1 General Remarks depending on the initial supersaturation of
the solid solution.
For most two-phase alloys, the simple
model of diffusional growth of isolated 5.6.2 The LSW Theory of Coarsening
non-interacting particles with uniform size,
on which Eq. (5-47) is based, frequently In essence, the coarsening of randomly
does not give a realistic description of the dispersed second-phase particles is a multi-
further dynamic evolution of the precipi- particle diffusion problem which is diffi-
tate microstructure beyond its nucleation cult to handle theoretically. In their classic
stage. In reality, towards the end of the nu- LSW coarsening theory, Lifshitz and Slyo-
cleation period a more or less broad parti-
2
cle size distribution f (R) is established In its original meaning, Ostwald ripening is con-
fined to a coarsening reaction where the second-
(Fig. 5-10 and 5-29). According to the phase particles act as the only sinks or sources of so-
Gibbs–Thomson equation (5-45), the solu- lute atoms.
5.6 Coarsening of Precipitates 371
Ga≤ is the second derivative of the molar longer feasible to follow minor changes of
Helmholtz energy of the a phase; f3 is the Dc with time quantitatively by any of the
third moment of the time-independent experimental techniques frequently em-
scaled particle radius distribution function ployed, such as CTEM, AFIM, SAXS or
–
f (R/R ) which in the zero volume fraction SANS (Sec. 5.3.1); in the asymptotic limit
limit (i.e. f = fLSW; see Sec. 5.6.3) is 1.129. Dc Æ 0, it even appears difficult to measure
For an ideal solution and the dilute solu- Dc (t) in alloys containing a ferromagnetic
tion limit (i.e. cae Ⰶ1, cbe ≈ 1) the rate con- phase with magnetic techniques (Ardell,
stant KR (Eq. 5-64 a) adopts the form 1967, 1968), though these direct methods
are certainly more sensitive than CTEM,
8 D Vb cae s ab AFIM or SAS. On the other hand, in those
K RLSW = (5-65)
earlier decomposition stages, where Dc (t)
9 Rg T
is experimentally accessible, an LSW anal-
of the original LSW theory (Calderon et ysis cannot be performed, and, in particu-
al., 1994). lar, the rate constant KRLSW (Eq. (5-65)) is
Numerous experimental studies on a no longer valid (see Sec. 5.7.4.3). Thus, in
wide variety of two-phase alloys attempted practice, an LSW analysis based on Eqs.
to examine the coarsening kinetics pre- (5-63) and (5-65), at its best, can be per-
dicted by the LSW theory. Regardless of formed only in dilute systems and yields
the particular alloy system and the micro- only a value for the product Ds ab . This fact
analytical technique employed (e.g., was ignored in most LSW analyses based
CTEM on Ni–Al, Ni–Ti: Ardell, 1967, on the independent measurement of both
–
1968, 1970; and on Ni–Si: Cho and Ardell, R (t1/3) and Dc (t –1/3); hence, the values of
1997; AFIM on Ni–Al: Wendt and Haasen, D and s ab derived from this type of analy-
1983; SANS on Fe–Cu: Kampmann and sis must be regarded with some reserva-
Wagner, 1986), these studies frequently re- tion. However, frequently the interfacial
vealed the experimental size distribution energy was derived exclusively from Eq.
function to be considerably broader than (5-65) under the assumption that the effec-
– –
fLSW (R/R ) (Fig. 5-42), whereas a plot of R 3 tive diffusivity D is identical with that ob-
or Nv–1 vs. t yielded more or less straight tained from an extrapolation of available
lines (due to limited statistics, the error high-temperature data to the aging temper-
bars are usually rather large). From the ature, thus neglecting the influence of
slopes of these LWS plots the product s ab D quenched-in vacancies (Sec. 5.3.2.1). (In
can be derived with cae usually taken from Sec. 5.7.4.4 two different methods will be
the known phase diagram. Frequently Dc discussed which allow a separate deriva-
was also measured and plotted versus t–1/3 tion of s ab and D to be made from experi-
2
(Eq. (5-63 c)) in order to determine D/s ab mental data, one method without even
(e.g., Ardell, 1967, 1968, 1995; Wendt knowing Dc (t).)
and Haasen, 1983). Thus apparently abso- For non-ideal solid solutions, s ab must
lute values for both s ab and D have been be derived from Eq. (5-64 a). Apart from
determined from values of s ab D and knowing D, this requires detailed informa-
2
D/s ab . However, the rate constants KRLSW tion on the thermodynamics of the a matrix
and KCLSW in Eq. (5-63) were derived phase in order to derive Ga≤. For many bi-
with the assumption c): Dc ≈ 0 or fp = const. nary alloys the necessary thermodynamic
Once this condition is fulfilled, it is no functions have become available via the
5.6 Coarsening of Precipitates 373
each other causes this small particle to dis- formed in a system wherein the materials
appear. At larger interparticle separation parameters, such as the interfacial energy
distances, however, large particles tend to and diffusion coefficient, are known (inde-
be surrounded by small particles as they pendent of a coarsening experiment) and
are feeding solute to the growing large par- where this system satisfies all the assump-
ticles (Akaiwa and Voorhees, 1994). Thus, tions of theory. Therefore, it is unclear if
there are both spatial and particle size cor- the disagreement between theory and ex-
relations (Marder, 1987). Information on periments is due to an artifact of the system
the spatial correlations between particles is employed in the experiments or a defect in
needed to determine the structure function the theories. Recent experiments that em-
that is measured using small angle scatter- ploy a system in which the materials pa-
ing. The agreement between the predic- rameters are known and satisfy all the as-
tions of Akaiwa and Voorhees (1994) for sumptions of theory have been performed
the structure function and the experimen- using solid Sn-rich particles in a Pb-rich
tally measured structure function in Al–Li eutectic liquid (Alkemper et al., 1999).
alloys is reasonable (Che et al., 1997). (c) There was no convective motion of the par-
Another effect of interparticle diffusional ticles because the experiments were per-
interactions is that the center of mass of the formed in the microgravity environment of
particles is not fixed, but moves in a man- the Space Shuttle. These experiments show
ner consistent with the concentration gra- that there can be very long transients asso-
dients in the system. This particle motion ciated with the coarsening process. In par-
has been determined theoretically (Mar- ticular, during the course of the experiment
der,1987; Akaiwa and Voorhees, 1994), and the initially broad particle size distribution
has been observed in experiments on coars- evolved slowly towards that predicted by
ening in transparent solid–liquid mixtures theory. Even with a factor of four change in
(Voorhees and Schaefer, 1987). the average particle size the distribution
Experiments in which the coarsening ki- never reached the steady-state distribution
netics of solid particles in a liquid (Fig. predicted by theory. This may be the reason
5-45) have been measured show clearly why the experimentally measured particle
that the coarsening rate increases with vol- size distributions (measured using other
ume fraction (Hardy and Voorhees, 1988, systems) rarely agree with theory. Never-
and references therein). In two-phase solid theless, the evolution of the distribution
systems, however, there are reports that the was found to correspond quite well with
rate constant is independent of the volume the predictions of two theories for transient
fraction (Cho and Ardell, 1997). The rea- Ostwald ripening.
son for these contradictory results may be Coarsening of two-phase alloys with
the elastic stress that is present in the two- modulated microstructures consisting of
phase solid systems. Although the pre- large volume fractions ( fp ≥ 30%) of iso-
dicted particle size distributions of the non- lated coherently misfitting particles, e.g.,
zero volume fraction theories are broader Cu–Ni–Fe and Cu–Ni–Cr (Fig. 5.19),
than those of LSW, nearly all experimen- Cu–Ni–Si (Yoshida et al., 1987), or mod-
tally measured particle size distributions ulated Co–Cu (Miyazaki et al., 1986), was
are broader than the predictions of the non- reported to be rather sluggish. The time ex-
zero volume fraction theories. Until re- ponent a of the coarsening rate, defined by
–
cently there have been no experiments per- R µ t a, of these alloys was consistently
376 5 Homogeneous Second-Phase Precipitation
Figure 5-45. Evolution of Sn-rich particles with fp = 0.64 in a liquid Pb–Sn matrix during coarsening at 185 °C.
The top row is at constant magnification and illustrates a typical coarsening process. For scaling the absolute
size of the microstructure, in the bottom row the magnification has been multiplied by the ratio of the average
– –
intercept length L at time t to L at t = 75 min. (From Hardy and Voorhees, 1988.)
5.6 Coarsening of Precipitates 377
found to be less than 1/3, the value ex- ing the intermediate stages of coarsening.
pected from conventional coarsening the- Depending on the specific (local) micro-
ory on the basis of Ostwald ripening. Fur- scopic diffusional mechanism which is as-
thermore, it was shown that coarsening of a sumed to contribute to the shift of the cen-
·100Ò modulated structure in less concen- ter of gravity of the particles and their
trated Co–Cu alloys (≤ 20 at.% Cu; fp ≈ likely coagulation, the time exponents a for
20%) still follows the t1/3 kinetics, whereas the related coarsening rate were evaluated
the coarsening rate becomes extremely between a = 1/6 and a = 1/4 and hence are
small in a Co–50 at.% Cu alloy ( fp ≈50%), smaller than predicted by the LSW theory
as indicated by the small time exponent (a = 1/3).
a < 1/50 (Miyazaki et al., 1986). According to Fig. 5-32 b the position of
As was pointed out in Sec. 5.5.5 “mot- the SANS peak intensity at k m varies in
tled structures” or “interconnected micro- proportion to t –0.25. This might, in fact, be
structures” can result from spinodal de- interpreted in terms of a cluster-diffusion-
composition of alloys with vanishing cohe- coagulation mechanism being dominant
rency strains. Unlike for modulated struc- prior to LSW-type coarsening. In Sec.
tures, elastic interactions can therefore be 5.8.2 it will be shown, however, that even
ruled out as a likely reason for the stability if single atom evaporation or condensation
of interconnected microstructures against in the LSW sense is assumed to be the only
coarsening. Nevertheless, the coarsening microscopic mechanism contributing to
rate of interconnected phases is frequently particle growth, the time exponent a al-
also found not to follow the t1/3 kinetics. ready displays a strong time dependence.
Experimental studies on the coarsening ki- Depending on the initial supersaturation of
netics of interconnected phases in hard the alloy, a (t) may then vary between al-
magnet materials such as Fe–Cr–Co (Zhu most zero and 0.5 during the course of a pre-
et al., 1986), Al–Ni–Co (Hütten and Haa- cipitation reaction. The question of whether
sen, 1986; Katano and Iizumi, 1982) in- a even reaches its asymptotic value 1/3 de-
stead yielded time exponents a between 1/4 pends on whether the experiment spans a
and 1/10. Evidently, an adequate theoreti- sufficiently long range of aging times.
cal description of the coarsening kinetics of From the experimental point of view it
both modulated and mottled microstruc- thus does not appear feasible to decide
tures has to go beyond the mere modeling merely on grounds of the measured time
of diffusional interaction between an en- exponent whether the cluster-diffusion-
semble of isolated spherical particles. coagulation mechanism influences the
growth or coarsening kinetics of a precipi-
tate microstructure.
5.6.4 Other Approaches Towards
Coarsening
5.6.5 Influence of Coherency Strains
In addition to the LSW-type coarsening
on the Mechanism and Kinetics
mechanism, which is based on the evopora-
of Coarsening – Particle Splitting
tion and condensation of single atoms from
dissolving and growing precipitates, The LSW theory, as well as the exten-
Binder and Heermann (1985) also con- sions to finite volume fractions, assume
siderred a cluster-diffusion-coagulation that the coarsening process is driven en-
mechanism likely to become operative dur- tirely by the associated release of interfa-
378 5 Homogeneous Second-Phase Precipitation
cial energy. In the presence of elastic tion energy is also a function of the shape
stresses, however, such as those induced by of the particles, the importance of which
coherency strains, the coarsening process has been emphasized by Onuki and Nishi-
can be driven by the release of both the mori (1991).
interfacial and elastic energy. The elastic In the presence of misfit strains, inter-
energy can be decomposed into an elastic particle elastic interactions can give rise to
self-energy, the elastic energy of an iso- pronounced spatial correlations between
lated particle in an infinite medium, and an the precipitates. Ardell et al. (1966) pro-
interaction energy that is due to the pres- posed that the elastic interactions between
ence of other particles in the system. The g¢–Ni3Al particles are responsible for the
elastic self-energy is a strong function of alignment of the particles along the elasti-
the shape of a particle and is responsible cally soft directions of the crystal during
for the shape bifurcations mentioned in coarsening. The resulting structure of
Sec. 5.4.2. Moreover, the elastic interac- many nickel-based alloys reveals that the
5.6 Coarsening of Precipitates 379
g¢-type precipitates are rather uniformly on the morphology of the particles and
distributed (Fig. 5-19 b), which is some- hence on the magnitude of the elastic inter-
times reminiscent of modulated structures. action energy itself. In fact, two-dimen-
These strong spatial correlations develop sional calculations (Thompson et al., 1994)
by two basic mechanisms. The first is that wherein the morphology of a particle was
particles aligned as stringers along the not constrained found that a misfitting
elastically soft directions will grow at the fourfold symmetric particle in an elasti-
expense of particles not so aligned (Ardell cally anisotropic homogeneous crystal was
et al., 1966). The second is that alignment stable with respect to interfacial perturba-
will occur by particles migrating through tions at least up to L = 26 (cf. Sec. 5.4.2).
the matrix (Voorhees and Johnson, 1988). Thus the conclusion reached on the basis of
This migration is a result of the elastic elastic energy considerations that splitting
interactions inducing a non-uniform inter- is possible appears to be due to the assump-
facial concentration and, hence, non-uni- tion that the morphology of the precipitate
form concentration gradients along the par- is invariant upon splitting. Miyazaki et al.
ticle interface. (1982) proposed that the splitting process
It has been observed (Miyazaki et al., begins with the formation of the matrix
1982; Doi et al., 1984; Kaufmann et al., phase in the middle of the precipitate. This
1989) that during coarsening, individual idea was verified through the diffuse inter-
g¢-type precipitates can split into groups of face calculations of Wang et al. (1991).
two or eight smaller cuboidal particles. Us- However, the experimental results of Yoo
ing a deep etching technique and a scan- et al. (1995) confirmed the hypothesis of
ning electron microscope, the recent experi- Kaufmann et al. (1989) that the splitting
ments of Yoo et al.(1995) have illustrated process is initiated via the amplification of
clearly the three-dimensional morphology perturbations along the precipitate–matrix
of the particles undergoing the splitting interface and not by the formation of the
process (Figs. 5-46 and 5-47). Such a split- matrix phase in the center of the particle.
ting process clearly cannot be driven by a Thus, the cause of the splitting remains, at
reduction in interfacial energy and thus the this point, unexplained. However, the
cause of the splitting has been ascribed to a stability of a cuboidal shaped particle in
decrease in the elastic energy. Although the three dimensions has not been examined,
elastic self-energy does not change on and the effects of elastic inhomogeneity are
splitting, assuming that the particle mor- still to be explored fully. Recent work by
phologies are the same before and after Lee (2000) has shown that splitting is pos-
splitting, if the elastic interaction energy is sible for certain differences in elastic con-
negative, e.g., as is the case for particles stants between particle and matrix.
aligned along the elastically soft directions In addition to inducing particle migra-
of an elastically anisotropic crystal, the to- tion and selective coarsening, elastic inter-
tal elastic energy will decrease upon split- actions may alter the kinetics of the coars-
ting (Khachaturyan and Airapetyan, 1974; ening process. Calculations employing
Khachaturyan et al., 1988; Johnson and fixed (spherical or circular) particle mor-
Cahn, 1984; Miyazaki et al., 1982). How- phologies have shown that in a two-particle
ever, constraining the morphology of the system inverse coarsening is possible,
particles to be invariant upon splitting ne- wherein a small particle will grow at the
glects the effects of the elastic interaction expense of a large particle (Johnson,1984;
380 5 Homogeneous Second-Phase Precipitation
a value of 1/2. This result does not contra- tion, growth and coarsening. In contrast to
dict Onuki’s, as the onset of stabilization the considerations of the preceding sec-
clearly violates the statistically time invar- tions we now ask how experimental kinetic
iant assumption of Leo et al. data can be interpreted if they refer to very
There have been many investigations of early decomposition stages which include
the kinetics of coarsening in elastically nucleation, growth and coarsening as con-
stressed solids. Experiments performed in comitant rather than consecutive processes
Ni–Cu–Si alloys by Yoshida et al. (1987), on the time scale. In attempting such inter-
and Miyazaki and Doi (1989) show a clear pretations we have to recall that these early
departure from the classical t1/3 predictions decomposition stages especially are char-
for the average particle radius with the acterized by dramatic changes in supersat-
growth rate of the average particle radius uration and, thus, in the driving forces for
and the width of the particle size distribution nucleation and growth processes. Further-
decreasing in time (Figs. 5-48 and 5-49). A more, during these early stages the size dis-
similar departure from the t1/3 function has tribution function evolves and is subjected
also been observed in the Ti–Mo system by to drastic alterations with rather short ag-
Langmayr et al. (1994). In contrast, in ing periods where nucleation, growth and
Ni–Al (Ardell, 1990), Ni–Mo–Al (Fähr- coarsening must be seen as competing and
mann et al., 1995) and Ni–Si (Cho and Ar- overlapping processes. Evidently these
dell, 1997) alloys the exponent for the aver- rather complicated phenomena are not
age particle size appears to be 1/3. The rate properly accounted for by splitting the
constant, however, appears to be indepen- course of decomposition into a nucleation
dent of the volume fraction of precipitate. regime, a growth regime, and a coarsening
Although it has been speculated that elastic regime. Moreover, the kinetic theories de-
interactions are responsible for this result, veloped for either regime (Secs. 5.5.1, 5.5.2
this has not been confirmed due to the diffi- and 5.6, respectively) are based on ideal-
culty of performing calculations in systems ized assumptions which are frequently not
that are sufficiently large to yield accurate expected to be close to reality. This holds
statistical information. true for the classical nucleation theories of
Volmer and Weber and Becker and Döring
(Sec. 5.5.1). In these theories the supersat-
uration is assumed to be constant. This may
5.7 Numerical Approaches only be fulfilled – if at all – during the ear-
Treating Nucleation, Growth liest nucleation stage. Furthermore, these
and Coarsening as Concomitant theories are based on artificial assumptions
Processes of the cluster size distribution in the vicin-
ity of the critical radius (Fig. 5-24), which
are not consistent with the fact that during
5.7.1 General Remarks on the
the nucleation process, many growing pre-
Interpretation of Experimental Kinetic
cipitates slightly larger than the critical
Data of Early Decomposition Stages
size are formed. The theory of diffusional
In this section we turn our attention growth by Zener and Ham (Sec. 5.5.2) de-
to decomposition studies of alloys with scribes only the time evolution of precipi-
sufficiently high nucleation barriers tates with uniform size. However, towards
(DF */k T t 7) decomposing via nuclea- the end of nucleation as well as at the be-
382 5 Homogeneous Second-Phase Precipitation
ginning of the LSW coarsening regime we mann and Wagner (MLS model; 1984) in
certainly have to deal with a polydispersed such a way that it could be applied for the
precipitate microstructure. Thus, even if description of the kinetics of precipitate
nucleation and LSW coarsening were to formation and growth in metastable alloys
be well separated on the time scale, in of rather high degrees of supersaturation.
between these two regimes the Zener–Ham The MLS model is still based on the same
theory cannot be expected to correctly pre- assumptions as the original LS theory. In
dict the measured growth kinetics quantita- particular, the explicit form of the size dis-
tively. Finally, the coarsening theories of tribution is not accounted for and the long-
Lifshitz, Slyozov and Wagner are based on time coarsening behavior is assumed to
the linearized version of the Gibbs–Thom- match the LSW results, i.e., is described by
son equation (Eq. (5-46)) and on the as- Eqs. (5-63 a, 5-65). A priori, it is not pos-
sumption that the supersaturation is close sible to foresee the influence of these as-
to zero. These restrictions also hold for the sumptions on the precipitation kinetics.
more recent theories (Sec. 5.6.3) which Therefore, Kampmann and Wagner
take into account finite precipitated vol- (KW; 1984) have devised an algorithm that
ume fractions and overlapping concentra- accurately describes the entire course of
tion profiles between precipitates. precipitation within the framework of clas-
Thus, any theory for the kinetics of pre- sical nucleation and growth theories. This
cipitation that can be employed either for a algorithm is accurate in the sense that, un-
more realistic interpretation of experimen- like the LS and the MLS theories, it con-
tal data or for a prediction of the dynamic tains no simplifying assumption; in partic-
evolution of a second-phase microstructure ular, in this algorithm, termed the Numeri-
under elevated temperature service condi- cal model (N model), the time evolution of
tions, has to treat nucleation, growth and the size distribution is computed without
coarsening as concomitant processes. This any approximations. From a comparison of
was accounted for in the cluster dynamics the N model with experimental data it is
theories of Binder and coworkers (Secs. possible to determine some crucial precipi-
5.5.3, 5.6.4 and Binder and Fratzl (2001), tation parameters of the particular alloy
Chapter 6 of this volume). However, as system as well as to scrutinize the existing
was pointed out by these authors (Binder et nucleation and growth theories with re-
al., 1978; Mirold and Binder, 1977), the spect to their applicability to decomposi-
theory developed yields a reasonable qual- tion reactions in real materials. Further-
itative prediction of the features of experi- more, the N model allows an evaluation to
mentally observable quantities but does not be made of how realistic the various ap-
attempt their quantitative interpretation. proximations are which occur in both the
A further decomposition theory treating LWS theory and the MLS model.
nucleation, growth, and coarsening as con- In this section we briefly introduce the
comitant processes was developed by MLS and the N models. An attempt is
Langer and Schwartz (LS model; Langer made to demonstrate their ability to de-
and Schwartz, 1980). This theory was for- scribe the entire course of the decomposi-
mulated for describing droplet formation tion reaction; we will also show how some
and growth in near-critical fluids. It was essential parameters of the decomposing
later modified by Wendt and Haasen alloy, such as the diffusion constant and the
(1983) and further improved by Kamp- specific interfacial energy, can be evalu-
5.7 Numerical Approaches to Concomitant Processes 383
Figure 5-51. Evolution of various scaled precipitation parameters with scaled aging time t according to both the
N model and the MLS model. The chosen values of x 0, sab correspond to those for Cu–1.9 at.% Ti aged at 350 °C.
Figure 5-52. Evolution of the scaled size distribution function in Cu-1.9 at.% Ti with aging times as computed
with the N model. For comparison, the distribution functions fLSW (r, t) of the LSW theory with the known
–
values of Nv (t) and R (t) are also shown. D = 2.5 × 10–15 cm2/sec.
During the early nucleation period (t Ù 1) growth rate of nucleated precipitates is also
both J̃ and n increase, whereas x and, close to zero, i.e., r– remains about con-
hence, r* remain roughly constant. At this stant. After (t Û 1, those precipitates nucle-
stage, the N model gives a rather narrow ated first become considerably larger than
size distribution function f (r, t) (Fig. r* and f (r, t) becomes much broader. This
5-52). Since cR is still close to c– ≈ c0, the is the beginning of the growth period
5.7 Numerical Approaches to Concomitant Processes 387
which is characterized by (i) the highest served in Fig. 5-51. During the later
growth rate ever observed during the growth period, dissolution of particles with
course of precipitation; (ii) a ratio r–/r* r < r* commences. At this stage, the MLS
which becomes significantly larger than 1; model only counts particles with r > r* as
(iii) the maximum number density (nmax, belonging to the second phase (Eq. (5-67)),
Nv, max) of particles which remains about and, furthermore, assumes fa (r = r*) to be
constant; (iv) a stronger decrease of x; and, proportional to 1/(r– – r*) (Eq. (5-66)). At
consequently, (v) a decrease of J̃ from its this stage, this is a rather poor approxima-
maximum value; in this particular case, J̃ tion because the N model yields a steep
never reaches its steady-state value. At the slope for f (r Æ r*, t) with a rather small
end of the growth regime (t ≈ 15) the density n (r = r*). These facts make the
supersaturation has dropped significantly MLS model predict considerably larger rate
and the growth rate becomes small. This constants for the decrease of the particle
effect causes r* to converge towards r– and number density and, hence, for the growth
makes dn/dt < 0 (Eq. (5-68)). During the of r– at the end of the growth period (t ≈ 15).
subsequent transition period (t ≈ 15), the Depending on the particular choice of b
growth rate of r– is primarily controlled by (Eq. (5-66)), the coarsening rate of the
the dissolution of particles with r < r* and MLS model approaches asymptotically the
only to a lesser extent by the uptake of so- value KRLSW (Eq. (5-65)) from the LSW the-
lute atoms from the matrix, the supersatu- ory. In this asymptotic limit the mean radii
ration of which is still about 20%. During from both models and, hence, their coars-
the transition period, the true distribution ening rates become identical. At this stage
function f (r, t) continuously approaches the N model yields r– Ó r* as predicted by
the one predicted by LSW (Fig. 5-52). At LSW, whereas the MLS model yields r– –
–
this stage, however, fLSW is still a rather r* = const., i.e., R LS > R*, as required by
poor approximation for f (r, t). This simply Eq. (5-67).
reflects the influence of the linearization of We can conclude that the MLS model,
the Gibbs–Thomson equation (Eq. (5-46)) which requires much less computing time
on which the LSW theory is based. For than the N model, provides a good survey
3
t Æ •, dr– /dt approaches asymptotically of the general course of precipitation.
a constant value, i.e., the reaction is within However, due to the simplifying assump-
the asymptotic limit of coarsening where tions made, it does not predict the precipi-
the supersaturation is almost zero. At this tation kinetics with the same accuracy as
stage f (r, t) is well approximated by fLSW the N model. This is particularly evident
with only minor deviations for small parti- for those precipitation stages where the
cle radii. shape of f (r, t) is extreme, as, for in-
stance, during the later growth stages in the
example discussed above.
5.7.4.2 Comparison Between the MLS
Model and the N Model
5.7.4.3 The Appearance and
The precipitation reaction starts with an
Experimental Identification
identical nucleation rate and identical par-
of the Growth and Coarsening Stages
ticle sizes in both the MLS and the N mod-
els. Therefore, both models are expected to In the following, the results from the
yield identical results which, in fact, is ob- MLS and the N models are compared with
388 5 Homogeneous Second-Phase Precipitation
–
Figure 5-53. Variation of R ,
–
R*, RLS and of the relative
supersaturation with aging
time for Cu–1.9 at.% Ti as
computed with the N model
and the MLS model for the
given set of thermodynamics
data; also shown are the ex-
–
perimental data for R (t) from
von Alvensleben and Wagner
(1984).
Figure 5-54. Variation of R2 and R*2 with time during the growth regime for Cu–1.9 at.% Ti. During the period
–
marked by the two arrows the kinetics follow the power-law R ⬃ t1/2.
–3
b) Variation of the coarsening rate dR /dt with aging time; KRLSW is the value predicted by the LSW theory,
Eq. (5-65).
5.7 Numerical Approaches to Concomitant Processes 389
experimental data obtained from a FIM theory (Eq. (5-65)) is only reached for ag-
–
atom probe study of early-stage precipita- ing times beyond ≈ 104min! At this stage R
tion in Cu–1.9 at.% Ti at 350 °C (von Al- has already grown to ≈ 6.4 nm. From this
vensleben and Wagner, 1984). result, which reflects the influence of the
Fig. 5-53 shows the time evolution of R* linearization of Eq. (5-46), KW inferred
–
and or R in physical units. Again these ex- that the LSW theory predicts the correct
perimental data points are well described coarsening rate once
by the N model. Fig. 5-53 also reveals that Vm s ab
the experimental data cover neither the R ⭌13 (5-79)
early nucleation period nor the growth re- Rg T
gime, but rather start (t = 2.5 min) at the If this relation holds, f (R, t) is almost iden-
end of the latter region. From Fig. 5-54 a,
– tical to fLSW (Fig. 5-52) and c̄/cae ≈ 1.
where R 2 is plotted as a function of t, it is According to the results from the N
recognized that during the period t ≈ model, during the early coarsening stages
–
0.7 min to 1.2 min, R 2 varies linearly with t the precipitation kinetics deviate signifi-
as suggested by Eq. (5-40). Since this time cantly from those predicted by the LSW
period extends only over 0.5 min, KW con- theory (e.g., Fig. 5-54 b). Thus the widely
cluded that for Cu–1.9 at.% Ti the time used LSW analyses resulting in a determi-
window is too short for revealing the nation of D and s ab on the basis of Eq.
–
R ⬃t1/2 kinetics experimentally; in fact, the (5-63 a) and Eq. (5-63 c) should not be ap-
same holds true for many other alloy plied to early coarsening stages during
systems analyzed by KW using the N which the relation does not hold.
model. Furthermore, from Fig. 5-54 a the
slope of the straight line has been evaluated
5.7.4.4 Extraction of the Interfacial
to be 0.88 nm2/min, whereas the corre-
Energy and the Diffusion Constant
sponding growth rate from Eq. (5-47) is
from Experimental Data
computed (cae = 0.22 at.%, cp = 20 at.%,
D = 2.5 ¥ 10–15 cm2/sec) to be 2.5 nm2/ KW determined s ab and D separately by
–
min. This result clearly demonstrates that fitting R (t) and Nv (t) as obtained from the
no growth regime exists which is ade- N model to the corresponding experimental
quately accounted for by Eq. (5-47). In data. Fig. 5-55 shows the variation of Nv
other words, the idealizations made in the and J* with aging time as computed with
derivation of Eq. (5-47) do not approxi- the N and the MLS models, together with
mate the true situation in Cu–1.9 at.% Ti. the corresponding experimental data. The
However, if c̄ (t) and cR in Eq. (5-43) are peak number density Nv, max of particles in
replaced by their mean values respectively, a precipitation reaction is essentially gov-
–
during the period for which R ⬃t1/2 holds, erned by the value of DF* via the nuclea-
then Eqs. (5-40 and 5-42) with li = 앀앙 k* tion rate equation (5-27). Since D F* ⬃ s ab
3
yield a value for the growth rate (0.104 nm2/ (Eq. (5-20)), Nv, max depends sensitively on
min) which is only 16% higher than the the value of the interfacial energy s ab. The
true value. very strong dependence of Nv, max on s ab is
–
In Fig. 5-54 b the rate constant dR 3/dt for clearly revealed by Fig. 5-55, showing a
coarsening is plotted versus t. It is evident good fit of Nv (t) for s ab = 0.067 J/m2 and
that for Cu–1.9 at.% Ti the rate constant only poor agreement for s ab = 0.071 J/m2.
KRLSW= 1.2 ¥ 10–24 cm3/sec from the LSW On the other hand, a variation in D mani-
390 5 Homogeneous Second-Phase Precipitation
Figure 5-55. Variation of Nv and J* with aging time for Cu–1.9 at.% Ti as computed with the N model and the
MLS model for sab = 0.067 J/m2; for this value of sab the computed Nv (t) curve agrees well with the experi-
mental data; poor agreement is obtained for sab = 0.071 J/m2.
Table 5-3. Correlation between the width of the miscibility gap and the coherent interfacial energies sab for
various two-phase alloys as determined from a fit of the N model to experimental kinetic data.
Ni–14 Al 1
550, 500 g ¢-Ni3Al ≈ 15 ≈ 0.016
Ni–26 Cu–9 Al 2
550, 500 g ¢-(Cu, Ni)3Al ≈ 20 ≈ 0.052
540, 500 Not determined ≈ 20 ≈ 0.050
580, 500 Not determined ≈ 20 ≈ 0.052
Cu–1.9 Ti 3
350, 500 b¢-Cu4Ti ≈ 20 ≈ 0.067
Cu–2.7 Ti 4
350, 500 b¢-Cu4Ti ≈ 20 ≈ 0.067
Cu–1.5 Co 5
500, 500 > 95 at.% Co ≈ 95 ≈ 0.171
Fe–1.4 Cu 6
400, 500 > 98 at.% Cu ≈ 100 ≈ 0.250
Fe–0.64 Cu 6
400, 500 > 98 at.% Cu ≈ 100 ≈ 0.250
fests itself in a parallel shift of the entire From a variety of different two-phase al-
–
curve on the time scale. Thus, a fit of R (t) loys, the available kinetic data have been
and of Nv (t) as obtained from the N model interpreted in terms of the N model. Table
to the experimental curves allows both s ab 5-3 presents the interfacial energies s ab to-
and D to be determined quite accurately. gether with the width of the (coherent) mis-
5.7 Numerical Approaches to Concomitant Processes 391
cibility gap for each given alloy. It is evi- ones. After aging for ≈ 100 min at 350 °C
dent that there is a pronounced correlation the metastable b¢-solvus line is nearly
between s ab and the compositional width, reached at c¢ae≈0.22 at.% Ti (cf. Sec. 5.2.1).
i.e., the broader the gap, the larger is s ab. The experimentally determined kinetic
This is consistent with various theoretical behavior of the precipitate number density,
predictions on the interfacial energy (see of their mean radius, and of the supersatu-
Lee and Aaronson, 1980, for a comprehen- ration are displayed in Fig. 5-57 a – c and
sive discussion of this aspect). compared with the predictions of the N
model. For the computations, the result
from the SANS evaluation was taken into
5.7.5 Decomposition Kinetics in Alloys
account, which yielded the homogenized
Pre-Decomposed During Quenching
sample to already contain ≈2 ¥ 1025 clus-
–
The versatility of the N model is further ters/m3 with R ≈ 0.7 nm. These could im-
illustrated by its ability to predict the pre- mediately grow by further depleting the
cipitation kinetics in alloys which have ex- matrix from solute atoms. Moreover, at
perienced some phase separation during t = 0 the supersaturation was still large
quenching. This is exemplified for Cu–2.9 enough for nucleating new clusters with
at.% Ti, the decomposition reaction of smaller radii at a nucleation rate J* (Fig.
which was studied by Kampmann et al. 5-57 a). Thus, during the first minutes of
(1987) by means of SANS. They found that aging, the cluster number density in-
the cooling rate of their specimen was not creased. At this stage, the alloy contained a
sufficient to suppress the formation of sort of bimodal cluster distribution: the
Cu4Ti precipitates during the quench; in larger ones formed at a smaller supersatu-
fact, as is shown in Fig. 5-56, the solute ration during quenching, and the smaller
concentration decreased from c0 = 2.9 at.% ones resulted from nucleation at 350 °C.
to 2.2 at.% Ti during the quench. With fur- Due to both nucleation of new clusters and
ther aging, the supersaturation decreased growth of pre-existing ones, the supersatu-
continuously through the formation of ad- ration and, hence, the nucleation rate de-
ditional clusters and the growth of existing creased rapidly; after aging for ≈ 3 min, nu-
precipitation. At present, it is unclear adequate for describing the growth (R > R*)
whether the various theories dealing with or dissolution (R < R*) of clusters.
finite volume effects (Sec. 5.6.3) are still To overcome these shortcomings, Kamp-
insufficient or whether some heterogenous mann et al. (Kampmann et al., 1992; Staron
precipitation at lattice defects accounts for and Kampmann, 2000 a, b) extended the N
the observed discrepancy. model by a cluster-dynamics (C–D) simu-
At first glance the experimental kinetic lation of the kinetics during the nucleation
–
data for R (t) and Nv (t) in Fig. 5-58 might and growth stages (cf. Sec. 5.5.3). To save
be seen as being amenable to an LSW anal- computation time, in the later growth and
ysis in terms of Eq. (5-63). Analyses of the coarsening stages where the stochastic pro-
SANS data for t Ⰷ103 min, however, cess is no longer relevant, the cluster-dy-
showed the width of the particle size distri- namics approach is linked to the determin-
bution to be much broader (standard devia- istic description of the original N model.
tion: s ≈ 0.31) than expected from the As in the latter, the input parameters enter-
LSW theory or its modification (s ≈ 0.23). ing the extended C–D model are c0, cae , s
Furthermore, the measured supersaturation and D with variations in D leading only to
–
was still far from being close to zero. shifts of the computed curves (e.g., R (t) or
Hence, the conditions for an LSW analysis J (t)) parallel to the time scale. Therefore,
of the experimental data are not at all ful- this C–D simulation can be used in as ver-
filled. satile a manner as the original N model
for the interpretation of experimental data
(Staron and Kampmann, 2000 b).
5.7.7 Combined Cluster-Dynamic
The cluster-dynamics algorithm was de-
and Deterministic Description
vised such that its corresponding Helm-
of Decomposition Kinetics
holtz energy functional approximates
There remain two main shortcomings of closely that of the regular solution model.
the N model outlined in Sec. 5.7.3. Firstly, As proved by CALPHAD studies, this is
the stochastic nature of the nucleation pro- a rather good approximation for Cu–Co.
cess is solely accounted for by using the Fig. 5-59 shows the computed time evolu-
time- and concentration-dependent nuclea- tion of the cluster size distribution during
tion rate (i.e., Eq. (5-29)), whereas the the nucleation period and the nucleation
growth (and shrinkage!) is computed in a rate with input parameters corresponding
deterministic manner on the basis of Eq. to a Cu–0.8 at.% Co model alloy aged at
(5-43). Thus, the stochastic nature of early- 500 °C. At t = 5 sec all clusters are still be-
stage growth which, in particular, becomes low the critical size4 and, thus, Fig. 5-59 a
effective in systems with small nucleation provides insight into the incubation process.
S
energies, is not adequately accounted for in The nucleation rate JB–D calculated ac-
the N model. cording to Eq. (5-27) with c̄ ∫ c0 exceeds
Secondly, as in all previous theories, the the maximum computed one (Fig. 5-59 b)
matrix concentration cR at the cluster–ma- by more than a factor of 1000. This dis-
trix interface was also calculated on the ba- crepancy cannot be attributed to incubation
sis of the Gibbs–Thomson equation, Eq.
(5-46). This relation, which describes an 4
In the C –D approach the critical radius is defined
equilibrium between the cluster size R and as that size for which the probabilities of cluster
the matrix concentration c̄, is generally not growth and shrinkage are identical.
5.8 Self-Similarity, Dynamical Scaling and Power-Law Approximations 395
5.8 Self-Similarity,
Figure 5-59. a) Evolution of the cluster size distri- Dynamical Scaling
bution during the nucleation stage for a Cu–0.8 at.% and Power-Law Approximations
Co model alloy aged at T = 500 °C, with pair ex-
change parameter W = 6.24, cae = 0.2 at.%, DCuCo =
1¥ 10–19 m2/s. 5.8.1 Dynamical Scaling
b) Time evolution of the nucleation rate J (t) (solid
According to the LSW theory and its ex-
line) according to the cluster-dynamics calculation.
S
JB–D S
is based on Eq. (5-27), Ĵ B–D is the modified tensions to finite volume fractions (Sec.
Becker–Döring nucleation rate (Eq. (5-80)). (After 5.6), the distribution of relative particle
–
Kampmann et al., 1992.) sizes R /R evolves during extended aging
(t Æ •) towards an asymptotic, time-invar-
iant form, the particular shape of which de-
396 5 Homogeneous Second-Phase Precipitation
pends on the precipitated volume fraction was first provided by Monte Carlo simula-
(Fig. 5-44). The final time invariance of tions of the time evolution of binary model
–
f (R /R ) reflects the fact that once the pre- alloys (Sec. 5.5.6). From these studies it
cipitated volume fraction has reached its was concluded that there is a small though
equilibrium value, consecutive configura- systematic dependence of F̃ (x) on the in-
tions of the precipitate microstructure are itial supersaturation, at least for small x
geometrically similar in a statistical sense, (Lebowitz et al., 1982); for large values of
i.e., all consecutive configurations are sta- x, the scaling function appears to be uni-
tistically uniform on a scale that is consid- versal in that it becomes independent of
erably larger than some characteristic temperature and precipitated volume frac-
–
length such as the mean particle size R or tion and even of the investigated material
–
the mean center-to-center distance L = (Fratzl et al., 1983). By analogy to the
Nv–1/3. The self-similarity of the microstruc- Porod law of small-angle scattering, in this
tural evolution has found its expression in regime F̃ (x) decays in proportion to x–4.
the dynamical scaling of the structure func- In experiments designed to test the valid-
tion S (k, t) (Binder and Stauffer, 1974; ity of the scaling behavior, l (t) is com-
Binder et al., 1978). Furukawa (1981) pro- monly related to either the radius of gyra-
posed S (k, t) to satisfy (after some tran- tion RG , the mean particle radius, or to ei-
sient time t0) a scaling law of the form: ther km–1 (t) or k1–1 (t) if km and k1 denote
the maximum and the first moment of S (k,
S (k, t) = l 3 (t) F̃ [k l (t)]; t ⭌ t0 (5-81)
t), respectively. The scaling function F̃ (x)
where F̃ [k l (t)] ∫ F̃ (x) is the time-inde- is then simply obtained, for instance, by
pendent scaling function. As the scaling plotting km3 S (k, t) versus k /km. If scaling
parameter, l (t) denotes some characteristic holds, F̃ (k /km) is time independent 5. After
length and contains exclusively the time
5
dependence of S (k, t). In order to test whether experimental data satisfy
Strong theoretical support for the valid- the scaling law, Fratzl et al. (1983) have proposed
– a
direct method by which the evaluation of RG, R, km
ity of the scaling hypothesis, Eq. (5-81), or k1 can be avoided and by means of which F̃ (x)
during the later stages of decomposition can be determined graphically.
5.8 Self-Similarity, Dynamical Scaling and Power-Law Approximations 397
chosen characteristic length, e.g., km–1 (t), corresponding data from computer simula-
entering the scaling law, Eq. (5-81). As the tions.
self-similarity of a precipitate microstruc- Inspired by the theoretical predictions,
ture and, hence, the scaling law is impli- many scattering experiments on alloys
citly contained in the LSW theory of coars- were interpreted in terms of power-law ap-
ening, the region of validity of dynamical proximations. Frequently the existence of
scaling coincides with the LSW regime of two well-defined kinetic regimes with dis-
coarsening. km–1 (t) is thus expected to show tinct values of a has been reported (e.g.,
the simple power-law behavior, Fig. 5-32 b). At earlier times, a ranges
between Ù0.1 and ≈ 0.2, whereas at later
km–1 (t) ⬃ t a (5-82)
stages a and b are found close to the values
with a = 1/3. Accordingly, if scaling holds, 1/3 and 1 predicted by the LSW theory.
the maximum of the structure function Sometimes this has been taken as evidence
must evolve in time as (e.g., Katano and Iizumi, 1984) for the first
regime to be dominated by the cluster–
Sm (t) ⬃ t b (5-83)
diffusion–coagulation mechanism (Sec.
with b = 3 a (Eq. (5-81)). Such a power-law 5.6.4), whereas in the second regime, the
behavior was frequently corroborated by evolution proceeds according to the LSW
scattering experiments on materials that mechanism via the evaporation and con-
were aged in the scaling region, and also by densation of single solute atoms.
computer simulations (e.g., Lebowitz et al., However, the interpretation of and Sm (t)
1982), which indicated that scaling would and km (t) in terms of two distinct kinetic
hold. regimes, each of which is well described by
The more recent theoretical develop- a power law, seems rather debatable. A
ments on the kinetics of phase separation closer examination of the Sm (t) and km
have predicted various other values for the curves (e.g., Fig. 5-32 b) always reveals
exponent a. On the basis of their cluster– some curvature prior to reaching the scal-
diffusion–coagulation model (Sec. 5.6.4), ing region. This clearly shows that the ex-
for intermediate times Binder and cowork- ponents a and b are time dependent; thus,
ers predict a = 1/6 and a = 1/5 or 1/4 for apart from the LSW regime, a power-law
low and intermediate temperatures, respec- approximation must be seen as a rather
tively (Binder and Stauffer, 1974; Binder poor description of the dynamic evolution
1977; Binder et al., 1978). As is shown in of a decomposing solid and commonly
Fig. 5-31 b, approximation of Sm (t), which does not disclose the specific growth
displays some curvature, by a power-law mechanism dominating at a certain aging
(Eq. (5-83)) yields b = 0.7 rather than 0.48 regime. This becomes particularly evident
as implied by scaling. The LBM theory of by employing the Numerical Model (Sec.
spinodal decomposition, which accounts 5.7.3) for a derivation of the exponent a (t)
–
for some coarsening at earlier stages (Sec. = ∂ log R /∂ log t. As the N model com-
5.5.4), yields a = 0.21. As outlined in Sec. prises nucleation, growth and coarsening
5.5.6, these values agree quite well with as concomitant processes on the basis of
the corresponding values (a = 0.16 to 0.25 just one growth mechanism – single-atom
and b = 0.41 to 0.71, depending on the evaporation or condensation in the LSW
supersaturation and ‘aging temperature’) sense – a plot of a (t) versus t allows a
obtained from fitting power laws to the closer examination to be made of the valid-
400 5 Homogeneous Second-Phase Precipitation
Figure 5-63. Dynamical scaling of the structure function of Cu-2.9 at.% Ti beyond t = 250 min; SANS results
ity of power-law approximations at any in- sharply from ≈ 0.15 to its maximum value a
stant. This is shown in Fig. 5-62 for Cu–1.9 = 0.56 which is indicative of a diffusion-
at.% Ti, the experimental kinetic data of controlled growth of the particles. The du-
which are well described by the N model 6
Accounting for modified boundary conditions, the
(cf. Figs. 5-53 and 5-55). Towards the end cluster-dynamics approach even yields a maximum
of the nucleation regime, a (t) increases value a = 0.7 (Kampmann et al., 1992).
5.9 Non-Isothermal Precipitation Reactions 401
–
ration of the growth regime where R 5.9 Non-Isothermal Precipitation
evolves according to the parabolic power Reactions
–
law R ⬃t1/2 is rather short for this alloy
(cf. Sec. 5.7.4.3). At the end of the growth For many age-hardening commercial al-
regime, where the particle number density loys, the microstructure is established dur-
has reached its maximum value (Fig. ing continuous cooling from the processing
5-55), a (t) drops within about 250 s to a temperature, i.e., the solidification or hot-
value of less than 0.1. In the subsequent working temperature in cast or wrought al-
transition regime at intermediate times, loys, respectively, or during cooling after
a (t) increases continuously and approaches laser surface treatments or welding (Brat-
only slowly the LSW-coarsening regime land et al., 1997, and references therein).
where dynamical scaling holds. Evidently, The resulting non-isothermal transforma-
there is no time regime between the appear- tion usually involves nucleation and
ance of the growth regime and the LSW re- growth of the precipitates as concomitant
gion where power-law behavior is ob- processes. Modeling of non-isothermal
served. On the other hand, it may be in- transformation rates thus ought to account
ferred from Fig. 5-62 that the kinetic evo- for the independent variations of the nucle-
–
lution of R during intermediate aging ation and the growth rates with temperature
stages might be artificially interpreted in and with the temperature-dependent solu-
terms of power laws if the time window bilities in the matrix and the precipitated
covered by the experiment was too short; phase. Furthermore, heterogeneous nuclea-
in this case any exponent between 0.15 and tion at lattice defects often must also be
0.33 may be derived. Thus, if the time win- taken into account, as usually the driving
dow for the kinetic experiment is not prop- force for homogeneous nucleation be-
erly chosen, a ≈ 0.2 may be obtained, but comes sufficient only once the alloy has
without the cluster–diffusion–coagulation been cooled sufficiently deep into the mis-
mechanism being operative. As has been cibility gap.
pointed out in Sec. 5.6.4.3, with increasing Whereas previous models, which have
supersaturation, the growth regime with been based on isokinetic behavior and the
–
R µ t 1/2 disappears completely and a (t) additivity concept, are limited to a descrip-
takes values only between zero and 1/3. tion of only diffusional growth (or dissolu-
Furthermore, the transition period (where tion during reheating) (e.g., Shercliff
a < 1/3) becomes shorter. In this case, LSW and Ashby, 1990a, b; Myhr and Grong,
coarsening and dynamical scaling are ob- 1991a, b; Onsoien et al., 1999), Grong and
served after rather short aging times. This is Myhr (2000) considered the non-isother-
illustrated in Fig. 5-63 for Cu–2.9 at.% Ti mal precipitation reaction as a coupled nu-
which satisfies dynamical scaling after ag- cleation and growth process in terms of a
ing for ≈ 250 min at 350 °C, whereas for the numerical solution and two analytical mod-
less concentrated Cu–1.9 at.% Ti alloy scal- els. As validated through a comparison
ing only holds after about 5 ¥ 104 min. with the numerical results, both analytical
models, i.e., a simplified state variable so-
lution and a solution based on the Avrami
equation, yielded an adequate description
of the overall non-isothermal transforma-
tion behavior comprising the variation of
402 5 Homogeneous Second-Phase Precipitation
the nucleation and growth rates as well as Ardell, A. J. (1990), Scripta Metall. Mater. 24, 343.
Ardell, A. J. (1995), Interface Sci. 3, 119.
the precipitated volume fraction with tem- Ardell, A. J., Nicholson, R. B. (1966), Acta Metall.
perature and cooling rate. 14, 1793.
Ardell, A. J., Nicholson, R. B., Eshelby, J. D. (1966),
Acta Metall. 14, 1295.
Asta, M. (1996), Acta Mater. 44, 4131.
5.10 Acknowledgements Asta, M., Hoyt, J. J. (1999), Acta Mater. 48, 1089.
Asta, M. Foiles, S. M., Quong, A. (1998), Phys. Rev.
B 57, 11265.
The authors would like to thank Prof. R. Auger, P., Danoix, F., Menand, A., Bourgoin, J.,
Bormann and Dr H. Mertins for their criti- Guttmann, M. (1990), Mat. Sci. Techn. 6, 301.
cal comments on this chapter and Mrs E. Auger, P., Pareige, P., Akamatsu, M., Blavette, D.
(1995), J. Nucl. Mat. 225, 225.
Schröder, A. Conrad-Wienands and B. Bartel, T. L., Rundman, K. B. (1975), Metall. Trans.
Feldmann for their kind endurance and as- 6 A, 1887.
sistance in preparing the manuscript. The Bates, F. S., Wiltzius, P. (1989), J. Chem. Phys. 91,
3258.
support of this work by the Deutsche For- Beaven, P. A., Frisius, F., Kampmann, R., Wagner, R.
schungsgemeinschaft (Leibniz-Programm) (1986), in: Atomic Transport and Defects in Metals
is also gratefully acknowledged. by Neutron Scattering: Janot, C., Petry, W., Rich-
ter, D., Springer, T. (Eds.). Berlin: Springer Verlag,
p. 228.
Becker, R., Döring, W. (1935), Ann. Phys. 24, 719.
Beddoe, R., Haasen, P., Kostorz, G. (1984), in: De-
5.11 References composition of Alloys: the Early Stages: Haasen,
P., Gerold, V., Wagner, R., Ashby, M. F. (Eds.).
Aalders, C., Van Dijk, S., Radelaar, S. (1984), in: De- Oxford: Pergamon Press, pp. 233 – 238.
composition of Alloys: The Early Stages: Haasen, Biehl, K.-E., Wagner, R. (1982), in: Solid – Solid
P., Gerold, V., Wagner, R., Ashby, M. F. (Eds.). Phase Transformations: Aaronson, H. I., Laugh-
Oxford: Pergamon Press, pp. 149–155. lin, D. E., Sekerka, R. F., Wayman, C. M. (Eds.).
Aaron, H. B., Fainstain, D., Kotler, G. R. (1970), J. Warrendale: The Metallurgical Society of AIME,
Appl. Phys. 41, 4404. pp. 185 – 189.
Aaronson, H. I., Russell, K. C. (1982), in: Solid-State Bigot, A., Auger, P., Chambreland, S., Blavette, D.,
Phase Transformations: Aaronson, H. I., Laugh- Reeves, A. (1997), Microsc. Microanal. Microstr.
lin, D. E., Sekerka, D. F., Wayman, C. M. (Eds.). 8, 103.
Warrendale: The Metallurgical Society of AIME, Binder, K. (1977), Phys. Rev. B 15, 4425.
pp. 371–399. Binder, K. (1980), J. Phys. (Paris) C 4, 51.
Aaronson, H. I., Kinsman, K. R., Russell, K. C. Binder, K. (1983), J. Chem. Phys. 79, 6387.
(1970 a), Scripta Metall. 4, 101. Binder, K. (1984), Phys. Rev. A 29, 341.
Aaronson, H. I., Laird, C., Kinsman, K. R. (1970 b), Binder, K. (1987), Advances on Phase Transitions
in: Phase Transformations: Aaronson, H. I. (Ed.). and Disorder Phenomena: Busiello, G., De Ce-
Metals Park, Ohio: Am. Soc. of Metals, pp. 313 – sare, L., Mancini, F., Marinaro, M. (Eds.). Singa-
396. pore: World Scientific, pp. 1– 71. (Kinetics of First
Acuna, R., Craievich, A. F. (1979), J. Non-Cryst. Order Transitions)
Solids 34, 13. Binder, K., Heermann, D. W. (1985), in: Scaling
Agarwal, S. C., Herman, H. (1973), in: Phase Transi- Phenomena in Disordered Systems: Pynn, R.,
tions and Their Application in Materials Science: Skjeltorp, T. (Eds.). New York: Plenum Press, pp.
Heinisch, H. K., Roy, R. (Eds.). New York: Per- 207 – 230.
gamon, pp. 207 – 222. Binder, K., Stauffer, D. (1974), Phys. Rev. Lett. 33,
Akaiwa, N., Voorhees, P. W. (1994), Phys. Rev. E 49, 1006.
3860. Binder, K., Stauffer, D. (1976), Adv. Phys. 25, 343.
Al-Kassab, T., Wollenberger, H., Blavette, D. (1997), Binder, K., Fratzl, P. (2001), in: Phase Transforma-
Z. Metallkd. 88, 2. tions in Materials: Kostorz, G. (Ed.). Weinheim:
Alkemper, J., Snyder, V. A., Akaiwa, N., Voorhees, Wiley-VCH, Ch. 6.
P. W. (1999), Phys. Rev. Lett. 82, 2725. Binder, K., Billotet, C., Mirold, P. (1978), Z. Physik
Ardell, A. J. (1967), Acta Metall. 15, 1772. B 30, 183.
Ardell, A. J. (1968), Acta Metall. 16, 511. Binder, K., Kalos, M. H., Lebowitz, J. L., Marro, J.
Ardell, A. J. (1970), Met. Trans. 1, 525. (1979), Colloid Interface Sci. 10, 173.
Ardell, A. J. (1972), Acta Metall. 20, 61. Blaschko, O., Fratzl, P. (1983), Phys. Rev. Lett. 51, 288.
5.11 References 403
Blaschko, O., Ernst, G., Fratzl, P., Auger, P. (1982), Danoix, F., Auger, P., Bostel, A., Blavette, D. (1991),
Acta Metall. 30, 547. Surf. Sci. 246, 260.
Blavette, D., Bostel, A., Sarran, J. M., Deconihout, Davies, C. K. L., Nash, P., Stevens, R. N. (1980),
B., Menand, A. (1993), Nature 363, 432. Acta Metall. 28, 179.
Blavette, D., Deconihout, B., Chambreland, S., Bos- Davies, D. M., Ralph, B. (1972), J. Microscopy 96,
tel, A. (1998), Ultramicroscopy 70, 115. 343.
Bonfiglioli, A. F., Guinier, A. (1966), Acta Metall. de Fontaine, D. (1969), Trans. AIME 245, 1703.
14, 1213. de Fontaine, D. (1971), J. Appl. Crystallogr. 4, 15.
Bouchard, M., Thomas, G. (1975), Acta Metall. 23, de Fontaine, D. (1973), J. Phys. Chem. Sol. 34, 1285.
1485. de Fontaine, D. (1982), in: Metallurgical Treatises:
Brailsford, A. D., Wynblatt, P. (1979), Acta Metall. Tien, J. K., Elliott, J. F. (Eds.). Warrendale, PA:
27, 489. The Metallurgical Society of AIME, pp. 423 – 444.
Bratland, D. H., Grong, O., Shercliff, H., Myhr, (Kinetics of First Order Transitions)
O. R., Tjotta, S. (1997), Acta Mater. 45, 1. Doherty, R. D. (1982), Metal. Science 16, 1.
Brenner, S. S., Camus, P. P., Miller, M. K., Soffa, W. Doherty, R. D. (1983), in: Physical Metallurgy:
A. (1984), Acta Metall. 32, 1217. Cahn, R. W., Haasen, P. (Eds.), Vol. 2. Amsterdam:
Cahn, J. W. (1962), Acta Metall. 10, 179, 907. North Holland Physics Publishing, pp. 933 – 1030.
Cahn, J. W. (1965), J. Chem. 42, 93. Doi, M., Miyazaki, T. (1986), Mater. Sci. Eng. 78,
Cahn, J. W. (1966), Trans. AIME 242, 166. 87.
Cahn, J. W., Charles, R. J. (1965), Phys. Chem. Doi, M., Miyazaki, T., Wakatsuki, T. (1984), Mater.
Glasses 6, 181. Sci. Eng. 67, 247.
Cahn, J. W., Hilliard, J. E. (1958), J. Chem. Phys. 28, Doi, M., Fukaya, M., Miyazaki, T. (1988), Phil. Mag.
258. A 57, 821.
Cahn, J. W., Hilliard, J. E. (1959 a), J. Chem. Phys. Dutta, I., Allen, S. M. (1991), J. Mater. Sci. Lett. 10,
31, 539. 323.
Cahn, J. W., Hilliard, J. E. (1959 b), J. Chem. Phys. Eckerlebe, H., Kampmann, R., Wagner, R. (1986),
31, 688. Atomic Transport and Defects in Metals by Neu-
Calderon, H. A., Voorhees, P. W., Murray, J. L., Kos- tron Scattering: Janot, C., Petry, W., Richter, D.,
torz, G. (1994), Acta Metall. Mater. 42, 991. Springer, T. (Eds.). Berlin: Springer Verlag,
Cerri, A., Schmelczer, R., Schwander, P., Kostorz, pp. 66 – 72.
G., Wright, A. F. (1987), in: Proc. MRS Symp., Vol. Edwards, G. A., Stiller, K., Dunlop, G. L., Couper,
82: Characterization of Defects in Materials: Sie- M. J. (1998), Acta Mater. 46, 3893.
gel, R. W., Weertman, J. R., Sinclair, R. (Eds.). Enomoto, Y., Kawasaki, K., Tokuyama. M. (1987),
Pittsburgh: Materials Research Society, p. 169. Acta Metall. 35, 907, 815.
Cerri, A., Schönfeld, B., Kostorz, G. (1990), Phys. Epperson, J. E. (1988), in: Dynamics of Ordering
Rev. B 42, 958. Processes in Condensed Matter: Komura, S., Fu-
Chan, K. S., Lee, J. K., Shiflet, G. J., Russell, K. C., rukawa, H. (Eds.). New York: Plenum Press, pp.
Aaronson, H. I. (1978), Metall. Trans. 9 A, 1016. 245 – 250.
Che, D. Z., Spooner, S., Hoyt, J. J. (1997), Acta Ernst, F., Haasen, P. (1988), Phys. Stat. Sol. (a), 104,
Mater. 45, 1167. 403.
Chen, H., Polat, S., Epperson, J. E. (1988), in: Dy- Ernst, F., Xiao, S. Q., Haider, F., Wilbrandt, P.-J.
namics of Ordering Processes in Condensed Mat- (1987), Scripta Metall. 21, 1189.
ter: Komura, S., Furukawa, H. (Eds.). New York: Eshelby, J. D. (1957), Proc. R. Soc. A 241, 376.
Plenum Press, pp. 245 – 250. Fährmann, M., Fratzl, P., Paris, P., Fährmann, E.,
Cho, J. H., Ardell, A. J. (1997), Acta Mater. 45, Johnson, W. C. (1995), Acta Metall. Mater. 43,
1393. 1007.
Christian, J. W. (1975), The Theory of Transforma- Feder, J., Russell, K. C., Lothe, J., Pound, G. M.
tions in Metals and Alloys, 2nd edn. Oxford: Per- (1966), Adv. Phys. 15, 111.
gamon Press. (General Diffusive Phase Transfor- Ferguson, P., Jack, K. H. (1984), Phil. Mag. A 50,
mations in Solids) 221.
Cockayne, D. J. H., Gronsky, R. (1981), Phil. Mag. Ferguson, P., Jack, K. H. (1985), Phil. Mag. A 52,
A 44, 159. 509.
Cook, H. E. (1970), Acta Metall. 18, 297. Forouhi, A. R., de Fontaine, D. (1987), Acta Metall.
Craievich, A. (1975), Phys. Chem. Glasses 16, 133. 35, 1863.
Craievich, A. F., Olivieri, J. R. (1981), J. Appl. Cryst. Fratzl, P., Penrose, O. (1994), Phys. Rev. B 50, 3477.
14, 444. Fratzl, P., Penrose, O. (1996), Acta Metall. 44, 3025.
Craievich, A., Sanchez, J. M. (1981), Phys. Rev. Lett. Fratzl, P., Lebowitz, J. L., Marro, J., Kalos, M. H.
47, 1308. (1983), Acta Metall. 31, 1849.
Craievich, A. F., Sanchez, J. M., Williams, C. E. Frenkel, J. (1939), J. Phys. (USSR) 1, 315.
(1986), Phys. Rev. B 34, 2762. Furukawa, H. (1981), Phys. Rev. A 23, 1535.
404 5 Homogeneous Second-Phase Precipitation
Furusaka, M., Ishikawa, Y., Yamaguchi, S., Fujino, Hirata, T., Kirkwood, D. H. (1977), Acta Metall. 25,
Y. (1986), J. Phys. Soc. Japan 55, 2253. 1425.
Gaskell, D. R. (1983), in: Physical Metallurgy, Vol. Hobbs, L. W., Westmacott, K. W., Williams, D. B.
1: Cahn, R. W., Haasen, P. (Eds.). Amsterdam: (Eds.) (1986), Proc. of the Materials Research
North Holland Physics Publishing, pp. 271 – 326. Society Symposium, Vol. 62: Materials Problem
(Metallurgical Thermodynamics) Solving with the Transmission Electron Micro-
Gaulin, B. D., Spooner, S., Morii, Y. (1987), Phys. scope. (Experimental Techniques; CTEM)
Rev. Lett. 59, 668. Horiuchi, S., Izumi, F., Kikuchi, T., Uchida, K.
Gerling, R., Schimansky, F.-P., Wagner, R. (1988), (1984), Phil. Mag. A 50, L 29.
Acta Metall. 36, 575. Hornbogen, E. (1967), Aluminium 43, 115.
Gerold, V., Merz, W. (1967), Scripta Metall. 1, 33. Hoyt, J. J. (1989), Acta Metall. 37, 2489.
Glatter, O. (1982), in: Small Angle X-Ray Scattering: Hoyt, J. J., de Fontaine, D. (1989), Acta Metall. 37,
Glatter, O., Kratky, O. (Eds.). London: Academic, 1611.
p. 167. Hoyt, J. J., Sluiter, M., Clark, B., Kraitchman, M., de
Gleiter, H. (1983), in: Physical Metallurgy: Cahn, R. Fontaine, D. (1987), Acta Metall. 35, 2315.
W., Haasen, P. (Eds.), Vol. 1. Amsterdam: North Hoyt, J. J., Clark, B., de Fontaine, D. (1989), Acta
Holland Physics Publishing, pp. 649 – 713. Metall. 37, 1597.
Goldstein, M. (1965), J. Am. Ceram. Soc. 48, 126. Hütten, A., Haasen, P. (1986), J. Phys. (Paris) C 7,
Grong, O., Myhr, O. R. (2000), Acta Mater. 48, 205.
445. Izumitani, T., Hashimoto, T. (1985), J. Chem. Phys.
Gronsky, R., Okada, M., Sinclair, R., Thomas, G. 83, 3694.
(1975), 33rd Ann. Proc. E.M.S.A., 22. Jantzen, C. M. F., Herman, H. (1978), in: Phase Dia-
Grüne, R. (1988), Acta Metall. 36, 2797. grams: Materials Science and Technology, Vol. 5:
Grüne, R., Oehring, M., Wagner, R., Haasen, P. Alper, A. M. (Ed.). New York: Academic Press,
(1985), in: Rapidly Quenched Metals, Vol. 1: pp. 128 – 185.
Steeb, S., Warlimont, H. (Eds.). Amsterdam: North Johnson, W. C. (1984), Acta Metall. 32, 465.
Holland, pp. 761 – 765. Johnson, W. C., Cahn, J. W. (1984), Acta Metall. 23,
Gunton, J. D. (1984), in: Decomposition of Alloys: 1839.
the Early Stages: Haasen, P., Gerold, V., Wagner, Johnson, W. C., Abinandanan, T. A., Voorhees, P. W.
R., Ashby, M. F. (Eds.). Oxford: Pergamon Press, (1990), Acta Metall. 38, 1349.
pp. 1–10. Kalos, M., Lebowitz, J. L., Penrose, O., Sur, A.
Gunton, J. D., Droz, M. (1984), Lecture Notes in (1978), J. Stat. Phys. 18, 39.
Physics, Vol. 183 – Introduction to the Theory of Kampmann, R., Wagner, R. (1984), in: Decomposi-
Metastable and Unstable States. Berlin: Springer tion of Alloys: the Early Stages: Haasen, P., Ge-
Verlag. (Kinetics of First Order Transitions) rold, V., Wagner, R., Ashby, M. F. (Eds.). Oxford:
Gupta, A. K., Lloyd, D. J. (1992), in: Proc. 3rd Int. Pergamon Press, pp. 91 – 103.
Conf. on Alluminium Alloys: Their Physical and Kampmann, R., Wagner, R. (1986), in: Atomic Trans-
Mechanical Properties, Vol. 2: Arnberg, L. (Ed.). port and Defects in Metals by Neutron Scattering:
Trondheim: Norwegian Institute of Technology Janot, C., Petry, W., Richter, D., Springer, T.
and SINTEF Metallurgy, p. 21. (Eds.). Berlin: Springer Verlag, pp. 73 – 77.
Guyot, P., Simon, J. P. (1981), in: Solid – Solid Phase Kampmann, R., Eckerlebe, H., Wagner, R. (1987),
Transformation: Aaronson, H. I., Laughlin, D. E., in: Phase Transitions in Condensed Systems – Ex-
Sekerka, R. F., Wayman, C. M. (Eds.). New York: periments and Theory: Cargill, G. S., Spaepen, F.,
Metallurgical Society of A.I.M.E., p. 325. Tu, K.-N. (Eds.). MRS-Symp. Proceedings, Vol.
Haasen, P., Wagner, R. (1985), Ann. Rev. Mat. Sci. 57: Materials Research Society, Pittsburgh, PA, pp.
15, 43. 525 – 542.
Hardy, S. C., Voorhees, P. W. (1988), Metall. Trans. Kampmann, R., Ebel, T., Haese, M., Wagner, R.
19 A, 2713. (1992), Phys. Stat. Sol. B. 172, 295.
Hashimoto, T., Itakura, M., Hasegawa, H. (1986 a), Katano, S., Iizumi, M. (1982), J. Phys. Soc. Japan
J. Chem. Phys. 85, 6118. 51, 347.
Hashimoto, T., Itakura, M., Hasegawa, H. (1986 b), Katano, S., Iizumi, M. (1984), Phys. Rev. Lett. 52, 835.
J. Chem. Phys. 85, 6773. Kaufman, L. (Ed.) (1977), CALPHAD, 1. Oxford:
Hennion, M., Ronzaud, D., Guyot, P. (1982), Acta Pergamon Press. (Volumes are published annually,
Metall. 30, 599. from 1977 onwards.)
Herring,, C. (1953), in: Structure and Properties of Kaufman, L., Bernstein, H. (1970), Computer Calcu-
Solid Surfaces: Gomer, R., Smith, C. S. (Eds.). lations of Phase Diagrams. New York: Academic
Chicago, IL: Univ. of Chicago Press, pp. 5 – 149. Press.
Hilliard, J. E. (1970), in: Phase Transformations: Kaufmann, M. J., Voorhees, P. W., Johnson, C.,
Aaronson, H. I. (Ed.). Metals Park, Ohio: Am. Soc. Biancaniello, F. S. (1989), Metall. Trans. 20 A,
of Metals, pp. 497 – 560. 2171.
5.11 References 405
Kawasaki, K. (1972), in: Phase Transitions and Crit- LeGoues, F. K., Lee, Y. W., Aaronson, H. I. (1984 b),
ical Phenomena, Vol. 2: Domb, C., Green, M. S. Acta Metall. 32, 1845.
(Eds.). New York: Academic Press, p. 2. LeGoues, F. K., Wright, R. N., Lee, Y. W., Aaronson,
Khachaturyan, A. G. (1983), Theory of Phase Trans- H. I. (1984 c), Acta Metall. 32, 1865.
formations in Alloys. New York: John Wiley. Leo, P. H., Mullins, W. W., Sekerka, R. F., Vinals, J.
Khachaturyan, A. G., Airapetyan, V. M. (1974), (1990), Acta Metall. Mater. 38, 1573.
Phys. Stat. Sol. 26, 61. Lifshitz, I. M., Slyozov, V. V. (1961), Phys. Chem.
Khachaturyan, A. G., Semenovskaya, S. V., Morris, Solids 19, 35.
J. W. (1988), Acta Metall. Mater. 36, 1563. Liu, Z. G., Wagner, R. (1984), J. Phys. Colloq. C 9,
Kirkwood, D. H. (1970), Acta Metall. 18, 563. 441.
Kostorz, G. (Ed.) (1979), Neutron Scattering. New Livak, R. J., Thomas, G. (1974), Acta Metall. 22,
York: Academic, p. 227. (Experimental Techniques; 589.
SANS) Lyon, O., Simon, J. P. (1987), Phys. Rev. B 35, 5164.
Kuo, S. Y., Virkar, A. V. (1987), J. Am. Ceram. Soc. Lyon, O., Simon, J. P. (1988), J. Phys. F, Metal Phys-
70, C-125. ics 18, 1787.
Langer, J. S. (1971), Ann. Phys. 65, 53. Maheshwari, A., Ardell, A. J. (1992), Scripta Metall.
Langer, J. S. (1973), Acta Metall. 21, 1649. 26, 347.
Langer, J. S. (1975), in: Fluctuations, Instabilities, Mandyam, H., Glicksman, M. E., Helsing, J., Marsh,
and Phase Transitions: Riste, T. (Ed.). New York: S. P. (1998), Phys. Rev. E 58, 2119.
Plenum Press, pp. 19 – 42. Marder, M. (1987), Phys. Rev. A 36, 858.
Langer, J. S., Schwartz, A. J. (1980), Phys. Rev. A 21, Marqusee, J. A., Ross, J. (1984), J. Chem. Phys. 79,
948. 373.
Langer, J. S., Bar-On, Miller, H. D. (1975), Phys. Marqusee, J. A., Ross, J. (1984), J. Chem. Phys. 80,
Rev. A 11, 1417. 536.
Langmayr, F., Fratzl, P., Vogl, G. (1994), Phys. Rev. Marro, J., Bortz, A. B., Kalos, M. H., Lebowitz, J. L.
B 49, 11759. (1975), Phys. Rev. B 12, 2000.
La Salle, J. C., Schwartz, L. H. (1984), in: Decompo- Marro, J., Bortz, A. B., Kalos, M. H., Lebowitz, J. L.
sition of Alloys: the Early Stages: Haasen, P., Ge- (1977), Phys. Rev. B 15, 3014.
rold, V., Wagner, R., Ashby, M. F. (Eds.). Oxford: Marsh, S. P., Glicksman, M. E. (1996), Acta Mater.
Pergamon Press, pp. 104 – 109. 44, 3761.
Laughlin, D. E. (1976), Acta Metall. 24, 53. Martin, G. (1978), in: Solid State Phase Transforma-
Laughlin, D. E., Cahn, J. W. (1975), Acta Metall. 23, tions in Metals and Alloys: Aussois: Les Editions
329. de Physique, pp. 337 – 406. (Kinetics of Phase
Lê, T. N., Barbu, A., Maury, F. (1992), Scripta Met- Separation in Solids)
all. Mater. 26, 771. Martin, J. W., Doherty, R. D. (1976), Stability of Mi-
Lebowitz, J. L., Kalos, M. H. (1976), Scripta Metall. crostructure in Metallic Systems. Cambridge: Uni-
10, 9. versity Press, p. 163.
Lebowitz, J. L., Marro, J., Kalos, M. H. (1982), Acta Miller, M. K., Brenner, S. S., Burke, M. G., Soffa, W.
Metall. 30, 297. A. (1984), Scripta Metall. 18, 111.
Lee, J. K. (1997), Mater. Sci. Eng, A 238, 1. Miller, M. K., Horton, L. L., Spooner, S. (1986), in:
Lee, J. K. (2000), Theor. Appl. Fract. Mech. 33, Proc. 32nd Intl. Field Emission Symp.: J. Phys.
207. (Paris) 47, Coloque C 2: Miller, M. K., Brenner,
Lee, J. H., Johnson, W. C. (1982), in: Solid-State S. S. (Eds.). Paris: Les Editions de Physique, pp.
Phase Transformations: Aaronson, H. I., Laugh- 409 – 416.
lin, D. E., Sekerka, D. F., Wayman, C. M. (Eds.). Miller, M. K., Cerezo, A., Hetherington, M. G.,
Warrendale: The Metallurgical Society of AIME, Smith, G. D. W. (1996), Atom-Probe Field Ion Mi-
pp. 127 – 150. croscopy, Oxford: Oxford Science Publications.
Lee, J. K., Barnett, D. M., Aaronson, H. I. (1977), (Tomographic Atom-Probe)
Metall. Trans. 8 A, 963. Mirold, P., Binder, K. (1977), Acta Metall. 25, 1435.
Lee, Y. W., Aaronson, H. I. (1980), Acta Metall. 28, Miyazaki, T., Doi, M. (1989), Mater. Sci. Eng. A 110,
539. 175.
LeGoues, F. K., Aaronson, H. I. (1984), Acta Metall. Miyazaki, T., Imamura, H., Kozakai, T. (1982), J.
32, 1855. Mater. Sci. Eng. 54, 9.
LeGoues, F. K., Aaronson, H. I., Lee, Y. W., Fix, G. Miyazaki, T., Seki, K., Doi, M., Kozakai, T. (1986),
J. (1982), in: Solid – Solid Phase Transformations: J. Mater. Sci. Eng. 77, 125.
Aaronson, H. I., Laughlin, D. E., Sekerka, R. F., Morral, J. E., Cahn, J. W. (1971), Acta Metall. 19,
Wayman, C. M. (Eds.). Warrendale: The Metallur- 1037.
gical Society of AIME, pp. 427 – 431. Mueller, R., Gross, D. (1998), Comp. Mat. Sci., 11,
LeGoues, F. K., Lee, Y. W., Aaronson, H. I. (1984 a), 25.
Acta Metall. 32, 1837. Mullins, W. W. (1986), J. Appl. Phys. 59, 1341.
406 5 Homogeneous Second-Phase Precipitation
Murata, Y., Iwama, Y. (1981), Trans. Japan. Inst. of Sato, K., Tagawa, K., Inoue, Y. (1989), Mater. Sci.
Metals 22, 433. Engin. A 111, 45.
Myhr, O. R., Grong, O. (1991 a), Acta Metall. 39, Sauthoff, G., Pitsch, W. (1987), Phil. Mag. B 56, 471.
2693. Schmidt, I., Gross, D. (1997), J. Mech. Phys. Solids
Myhr, O. R., Grong, O. (1991 b), Acta Metall. 39. 45, 1521.
2703. Schmidt, I., Mueller, R., Gross, D. (1998), Mech.
Neilson, G. F. (1969), Phys. Chem. Glasses 10, 54. Mater. 30, 181.
Nishi, T., Wang, T. T., Kwei, T. K. (1975), Macro- Schmitz, G., Haasen, P. (1992), Acta Metall. Mater.
molecules 8, 227. 40, 2209.
Nishimori, H., Onuki, A. (1990), Phys. Rev. B 42, 980. Sequeira, A. D., Calderon, H. A., Kostorz, G., Peder-
Okada, M., Han, C. C. (1986), J. Chem. Phys. 85, sen, J. S. (1995), Acta Metall. Mater. 43, 3427.
5317. Servi, I. S., Turnbull, D. (1966), Acta Metall. 14, 161.
Onsoien, M. I., Grong, O., Gundersen, O., Skaland, Shercliff, H. R., Grong, O., Myhr, O. R., Ashby,
T. (1999), Metall. Mater. Trans. 30 A, 1069. M. F. (1992), in: Proc. 3rd Int. Conf. on Aluminium
Onuki, A., Nishimori, H. (1991), Phys. Rev. B 43, Alloys – Their Physical, Mechanical Properties,
13649. Vol. 3: Arnberg, L. (Ed.). Trondheim: Norwegian
Pareige, P., Soisson, F., Martin, G., Blavette, D. Institute of Technology and SINTEF Metallurgy,
(1999), Acta Mater. 47, 1889. p. 357.
Paris, O., Langmayr, F., Vogl, G., Fratzl, P. (1995), Z. Shercliff, H. R., Ashby, M. F. (1990 a), Acta Metall.
Metallkd. 86, 860. 38, 1789.
Park, M. W., Mitchell, T. E., Heuer, A. H. (1975), J. Shercliff, H. R., Ashby, M. F. (1990 b), Acta Metall.
Am. Ceram. Soc. 58, 43. 38, 1803.
Park, M. W., Mitchell, T. E., Heuer, A. H. (1976), Shewmon, P. G. (1965), Trans. TMS-AIME 233, 736.
J. Mater. Sci. 11, 1227. Shiflet, G. J., Lee, Y. W., Aaronson, H. I., Russell, K.
Penrose, O. (1978), in: Lecture Notes in Physics, Vol. C. (1981), Scripta Metall. 15, 719.
84: Garrido, L., Seglar, P., Shepherd, P. J. (Eds.). Si-Qun Xiao, Haasen, P. (1991), Acta Metall. 39, 65.
Heidelberg: Springer Verlag, pp. 210 – 234. Simon, J. P., Lyon, O. (1989), Acta Metall. 37, 1727.
Penrose, O., Lebowitz, J. L. (1979), in: Studies in Simon, J. P., Guyot, P., DeSalva, A. G. (1984), Phil.
Statistical Mechanics, Vol. 7: Lebowitz, J. L., Mag. A 49, 151.
Montroll, E. (Eds.). Amsterdam: North Holland, Sinclair, R., Thomas, G. (1974), 32nd Ann. Proc.
pp. 293 – 340. (Kinetics of First Order Transitions) E.M.S.A., 500.
Penrose, O., Lebowitz, J. L., Marro, J., Kalos, M. H., Singh, J., Lele, S., Ranganathan, S. (1980), J. Mater.
Sur, A. (1978), J. Stat. Phys. 19, 243 – 267. Sci. 15, 2010.
Pike, B. C., Messoloras, S., Stewart, R. J. (1989), Singhai, S. P., Herman, H., Kostorz, G. (1978), J.
Scripta Metall. 23, 983. Appl. Cryst. 11, 572.
Piller, J., Wagner, W., Wollenberger, H., Mertens, P. Skripov, V. P., Skripov, A. V. (1979), Sov. Phys. Usp.
(1984), in: Decomposition of Alloys: the Early 22, 389.
Stages: Haasen, P., Gerold, V., Wagner, R., Ashby, Sluiter, M., Kawazoe, Y. (1996), Phys. Rev. B 54,
M. F. (Eds.). Oxford: Pergamon Press, pp. 156 – 10381.
164. Smith, D. J. (1983), Helv. Phys. Acta 56, 463. (Ex-
Polat, S., Marsh, C., Little, T., Ju, C. P., Epperson, J. perimental Techniques; HREM)
E., Haydn Chen (1986), Scripta Metall. 20, 1759. Snyder, H. L., Meakin, P. (1983 a), J. Chem. Phys.
Polat, S., Haydn Chen, Epperson, J. E. (1989), Met. Soc. 79, 5588.
Trans. 20 A, 699. Synder, H. L., Meakin, P. (1983 b), Macromolecules
Purdy, G., Bréchet, Y. J. M. (2001), in: Phase Trans- 16, 757.
formations in Materials: Kostorz, G. (Ed.). Wein- Snyder, H. L., Meakin, P. (1985), J. Polym. Sci. 73,
heim: Wiley-VCH, Chap. 7. 217.
Radmilovic, V., Fox, A. G., Thomas, G. (1989), Acta Soisson, F., Barbu, A., Martin, G. (1996), Acta
Metall. 37, 2385. Mater. 44, 3789.
Rikvold, P. A., Gunton, J. D. (1982), Phys. Rev. Lett. Staron, P., Kampmann, R. (2000 a), Acta Mater. 48,
49, 286. 701.
Rindone, G. E. (1975), Bull. Cent. Glass Ceram. Re- Staron, P., Kampmann, R. (2000 b), Acta Mater. 48,
search Inst. Calcutta 22, 119. 713.
Rioja, R. J., Laughlin, D. A. (1977), Metall. Trans. Stauffer, D. (1979), Phys. Rep. 54, 1.
8 A, 1257. Steiner, D., Beddoe, R., Gerold, V., Kostorz, G.,
Rundman, K. B., Hilliard, J. E. (1967), Acta Metall. Schmelzcer, R. (1983), Scripta Metall. 17, 733.
15, 1025. Stubican, V. S., Schultz, A. H. (1970), J. Am. Ceram.
Russell, K. C. (1980), Adv. Colloid Interface Sci. 13, Soc. 53, 211.
205. Su, C. H., Voorhees, P. W. (1996), Acta Metall.
Sato, R., Han, C. C. (1988), J. Chem. Phys. 88, 2057. Mater. 44, 1987.
5.11 References 407
Sur, A., Lebowitz, J. L., Marro, J., Kalos, M. H. morrow – Technology, Products, Research: Cere-
(1977), Phys. Rev. B 15, 3014. sara, S. (Ed.). Barga die Lucca, pp. 119 – 127.
Taylor, K. A. (1985), Sc.D. Thesis Aging Phenomena Wahi, R. P., Stajer, J. (1984), in: Decomposition of
in Ferrous Martensites. MIT, Cambridge, MA. Alloys: the Early Stages: Haasen, P., Gerold, V.,
Thompson, M. E., Voorhees, P. W. (1999), Acta Wagner, R., Ashby, M. F. (Eds.). Oxford: Per-
Mater. 47, 983. gamon Press, pp. 165 – 169.
Thompson, M. E., Su, C. S., Voorhees, P. W. (1994), Wang, Y., Khachaturyan, A. G. (1995), Acta Metall.
Acta Metall. 42, 2107. Mater. 43, 1837.
Tien, J. K., Shewmon, P. G., Foster, J. S. (1973), Wang, Y., Chen, L.-Q., Khachaturyan, A. G. (1991),
Scripta Metall. 7, 1171. Scripta Metall. Mater. 25, 1387.
Tokuyama, M., Kawasaki, K. (1984), Physica A 123, Wang, Y., Chen, L.-Q., Khachaturyan, A. G. (1992),
386. Phys. Rev. B 46, 11194.
Tsumuraya, K., Miyata, Y. (1983), Acta Metall. 31, Weins, J. J., Cahn, J. W. (1973), in: Sintering and Re-
437. lated Phenomena, Kuczynski, G. C. (Ed.). Lon-
Ujihara, T., Osamura, K. (2000), Acta Mater. 48, don: Plenum, p. 151.
1629. Wendt, H., Haasen, P. (1983), Acta Metall. 31, 1649.
Vintaikin, Y. E., Dmitriyev, V. B., Udovenko, V. A. West, A. W., Kirkwood, D. H. (1976), Scripta Metall.
(1979), Phys. Met. Metall. 46, 97. 10, 681.
Volmer, M., Weber, A. (1926), Z. Phys. Chem. 119, Williams, D. B., Carter, C. B. (1996), Transmission
277. Electron Microscopy. New York and London: Ple-
von Alvensleben, L., Wagner, R. (1984), in: Decom- num Press, p. 683.
position of Alloys: the Early Stages: Haasen, P., Wiltzius, P., Bates, F. S., Dierker, S. B., Wignall, G.
Gerold, V., Wagner, R., Ashby, M. F. (Eds.). Ox- D. (1987), Phys. Rev. A 36, 2991.
ford: Pergamon Press, pp. 143 – 148. Wiltzius, P., Bates, F. S., Heffner, W. R. (1988), Phys.
Voorhees, P. W., Glicksman, M. E. (1984), Acta Rev. Lett. 60, 1538.
Metall. 32, 2001, 2013. Wood, J. V., Mills, P. F., Bingham, J. K., Bee, J. V.
Voorhees, P. W., Johnson, W. C. (1988), Phys. Rev. (1979), Metall. Trans. 10 A, 525.
Lett. 61, 225. Yaldram, K., Binder, K. (1991 a), Z. Phys. B 82, 405.
Voorhees, P. W., Schaefer, R. J. (1987), Acta Metall. Yaldram, K., Binder, K. (1991 b), Acta Metall. Mater.
35, 327. 39, 707.
Wagner, C. (1961), Z. Elektrochem. 65, 581. Yao, J. H., Elder, K. R., Guo, H., Grant, M. (1993),
Wagner, R. (1982), Field Ion Microscopy in Ma- Phys. Rev. B 47, 14110.
terials Science, Vol. 6, Crystals – Growth, Proper- Yin, F., Oksawa, Y., Sato, A., Kawakara, K. (2000),
ties and Applications: Freyhardt, H. C. (Ed.). Ber- Acta Mater. 48, 1273.
lin: Springer Verlag. (Experimental Techniques; Yoo, Y. S., Yoon, D. Y., Henry, M. F. (1995), Metals
AFIM) and Mater. 1, 47.
Wagner, R., Brenner, S. S. (1978), Acta Metall. 26, Yoshida, S., Fukaya, M., Miyazaki, T. (1987), J. Ja-
197. pan. Inst. Metals 51, 18.
Wagner, W., Poerschke, R., Wollenberger, H. (1984), Zeldovich, J. B. (1943), Acta Physicochim. (USSR)
in: Decomposition of Alloys: the Early Stages: 18, 1.
Haasen, P., Gerold, V., Wagner, R., Ashby, M. F. Zener, C. (1949), J. Appl. Phys. 20, 950.
(Eds.). Oxford: Pergamon Press, pp. 170 – 179. Zhu, F., Haasen, P., Wagner, R. (1986), Acta Metall.
Wagner, R., Kampmann, R., Jiang, B., Beaven, P. A. 34, 457.
(1988), in: Proc. Int. Conf. Cu ’86 – Copper To-
6 Spinodal Decomposition
Kurt Binder
Peter Fratzl
hT random force
Q temperature at which behavior of polymer in solution is ideal
k factor describing screening of Coulomb interaction
l wavelength
lc critical wavelength in Cahn’s theory
lijmn elastic stiffness tensor
L generalized Onsager coefficient
m chemical potential difference
m elastic modulus
n critical exponent of the correlation length
nP Poisson coefficient in elasticity theory
x interfacial width; correlation length
xˆ critical amplitude of the correlation length
xcoex correlation length at the coexistence curve
r1, r 2 , … probability distributions
s subunit length of a polymer
s ij elastic stress tensor
t rescaled coordinate (time)
t time constant
j pair interaction energy
F volume fraction
Fc volume fraction at the critical point
c Flory – Huggins parameter
cs Flory – Huggins parameter at the spinodal curve
c n–1 phenomenological coefficient
y complex order parameter
Figure 6-1. Schematic diagram of unstable thermodynamic fluctuations in the two-phase regime of a binary
mixture AB at a concentration cB (a) in the unstable regime inside two branches cBs of the spinodal curve and (b)
in the metastable regime between the spinodal curve cBs and the coexistence curve c coex (1)
. The local concentration
c (r) at a point r = (x, y, z) in space is schematically plotted against the spatial coordinate x at some time t after
the quench. In case (a), the concentration variation at three distinct times t1 , t2 , and t3 is indicated. The diame-
ter of the critical droplet, whose cross-section is shown in case (b), is denoted by 2 R*, and the width of the inter-
facial region by x . Note that the concentration profile of the droplet reaches the other branch of the coexistence
(2) (1)
curve c coex in the droplet center only for weak “supersaturations” of the mixture, where cB – c coex O cBs – cB and
R* o x ; for the sake of clarity, the figure is therefore not drawn to scale (Binder, 1981).
6.1 Introduction 415
stricts the quantitative validity of the theo- chapter, and the spontaneous growth of
retical concepts developed in the present thermal fluctuations dominating phase
chapter (see also Binder, 2000). This chap- changes of unstable but nearly ideal (i.e.,
ter will exclusively consider the mecha- defect-free) systems will be emphasized.
nism of Fig. 6-1 a, where in a thermody- Although most of the discussion refers to
namically unstable initial state long-wave- binary mixtures, an extension to ternary
length delocalized small-amplitude statis- mixtures is often possible.
tical fluctuations grow spontaneously in Unmixing of binary or ternary mixtures
amplitude as the time after the quench in- is not the only phase change where sponta-
creases. For a binary mixture, this mecha- neous growth of statistical fluctuations oc-
nism is called spinodal decomposition curs. Consider, for example, an alloy A – B,
(Cahn, 1961, 1965, 1966, 1968); the rele- which in thermal equilibrium undergoes an
vant fluctuations can then be considered as order – disorder phase transition form a dis-
a wavepacket of “concentration waves”, ordered state, where the two species of at-
and one such wave is shown in Fig. 6-1 a. oms A and B are distributed at random over
As will be discussed below, this mecha- the available lattice sites, to an ordered ar-
nism should not occur inside the whole rangement. There A and B atoms preferen-
two-phase coexistence region of the phase tially occupy sites on sublattices (e.g., b-
diagram of the mixture, but rather only in- CuZn, Cu3Au, CuAu, Fe3Al, and FeAl). If
side a smaller region, the boundary of we quench such an alloy from the disor-
which is given by the spinodal curve (Cahn dered regime to a state which in equilib-
and Hilliard, 1958, 1959). Between the spi- rium should be ordered, the unstable disor-
nodal curve cBs and the coexistence curve dered initial state may also decay by spon-
(1)
c coex (or in between the other branch of the taneous growth of fluctuations. The dis-
(2)
spinodal and c coex the mechanism of drop- tinction from Fig. 6-1 a is that the wave-
let nucleation and growth is involved (Fig. length of the growing concentration wave
6-1 a). This latter mechanism is discussed is not large but coincides with the lattice
in the chapter by Wagner et al., 2001). spacing of the superstructure. Neverthe-
Apart from these phase transformation less, the theoretical treatment of this spino-
mechanisms triggered by spontaneous ther- dal ordering (De Fontaine, 1979) is similar
mal fluctuations, heterogeneous nucleation to the theory of spinodal decomposition,
processes must also be considered (Zettle- and will also be discussed in this chapter,
moyer, 1969); near inhomogeneities in which emphasizes the theoretical aspects
solids such as grain boundaries, disloca- (see Wagner and Kampmann (2000) for a
tions, external surfaces, or point defects complementary treatment emphasizing the
such as substitutional or interstitial impur- point of view of the experimentalist).
ities or clusters thereof, microdomains of One of the basic questions in the applica-
the new phase may already be formed be- tions of materials is to understand the pre-
fore the quench, or at least their formation cipitated microstructure that forms in the
after the quench is greatly facilitated. Gen- late stages of a phase separation process.
eral theoretical statements about these One of the main characteristics of this mi-
heterogeneous mechanisms, however, are crostructure concerns its morphology: the
hardly possible without a detailed discus- minority phase may either form compact
sion of the nature of such defects. Such islands well separated from each other or it
problems will not be discussed in this may form an irregular interconnected net-
416 6 Spinodal Decomposition
Figure 6-3. (a) Order parameter y of a second-order phase transition vs. temperature, assuming a two-fold
degeneracy of the ordered state (described by the plus and minus signs of the order parameter). The quenching
experiment is indicated. (b) Phase diagram of a binary mixture with a miscibility gap ending in a critical point
(Tc , cBcrit ) of unmixing, in the temperature – concentration plane. Again the quenching experiment is indicated,
and the quenching distances from the coexistence curve (dT ) and from the critical point (DT ) are indicated.
(c) Free energy plotted versus composition at temperature T (schematic). For further explanations, see text
(Binder, 1981).
defines the spinodal curve c = cBs (T ). This (Binder et al., 1978). As will be discussed
distinction is now linked to the two trans- in more detail in Sec. 6.2.5, the spinodal
formation mechanisms shown in Fig. 6-1; curve cannot be unambiguously defined.
it is the unstable regime inside the two At this point, we recall that the “weak delo-
branches of the spinodal curve, where calized long-wavelength fluctuations” can-
long-wavelength fluctuations spontane- not be identified in terms of the atomic
ously grow rather than decay. At a later concentration variable ci , which undergoes
stage, the inhomogeneous concentration rapid large-amplitude (ci = 0 to ci = 1!) vari-
distribution thus generated will coarsen. In ations from one lattice site to the next, but
the metastable regime, the system is stable implies the introduction of the coarse-
against such weak (small-amplitude) fluc- grained concentration field c (r), Eq. (6-1).
tuations, and localized large-amplitude The free energy F¢ of “homogeneous”
fluctuations (“droplets” of the new phase) states in the two-phase region depends on
must form in order to start the transforma- the length scale L over which short-wave-
tion. length concentration fluctuations have
This idea implies, as will be explained been integrated (Fig. 6-4). This “coarse-
below, a singular transition in the kinetic grained” free-energy density fcg is not pre-
transformation mechanism at the spinodal cisely identical to the true free-energy den-
curve. However, this spinodal singularity is sity f in the single-phase region, since con-
really only an artefact of an over-simplified centration variations with wavelengths ex-
theoretical picture: apart from the very ceeding L contribute to f but are excluded
special limit of infinitely weak, infinitely from fcg . However, this difference is minor,
long-ranged forces in which mean-field and in the single-phase region we may con-
theory becomes correct (Penrose and Lebo- sider the limit L Æ ∞ and then fcg tends to-
witz, 1971; Lebowitz and Penrose, 1966; wards f uniformly. This is not possible in
Binder, 1984 b), the transition from the nu- the two-phase region, however, where fcg
cleation mechanism to the spinodal decom- describes homogeneous states only if
position mechanism is completely gradual LÁ x , the interfacial width; if L Ô x then
the states that yield dominating contribu- The term –12 r2 [—c]2 in Eq. (6-4) accounts
tions to fcg are phase separated on a local for the free energy cost of inhomogeneous
scale, and therefore fcg tends smoothly to- concentration distributions. Here the
wards the double-tangent construction as coarse-grained free-energy density fcg (c)
L Æ ∞ . Hence there is no unique theoreti- resulting from Eqs. (6-3) and (6-1) by car-
cal method of calculating a spinodal curve; rying out the restricted trace over the {ci}
remember that mean-field theories, which for a fixed concentration field {c (x)} is
yield spinodal curves easily, are only inac- very difficult to obtain in practice; qualita-
curate descriptions of real systems. Simi- tively the behavior of fcg (c) should be sim-
larly, extrapolation procedures by which ilar to the mean-field (MF) result for the
spinodal curves are extracted from experi- free energy of a binary mixture,
mental data also involve related ambigui-
1
ties, as will discussed in Sec. 6.2.5. fMF ( c ) = c ln c + (1 − c ) ln (1 − c )
Hence the coarse-graining alluded to kB T
T MF
above means that a microscopic Hamil- + 2 c c (1 − c ) (6-6)
tonian Ᏼ {ci} of the binary mixture is re- T
placed by a so-called free-energy function with
DᏲ {c (x)}. For example, the Hamiltonian
may correspond to a standard Ising-type kB TcMF ≡ ∑ Dj ( xi − x j ) (6-7)
j
pairwise interaction model (Binder, 1986;
De Fontaine, 1979): Near TcMF we can replace fMF (c) by its Lan-
dau expansion,
Ᏼ {ci } = 1
2 ∑ {j BB ( xi − x j ) ci c j (6-3)
i≠ j
fMF ( c ) = f0 + A ( c − ccrit )2
+ j AB ( xi − x j ) [ci (1 − c j ) + c j (1 − ci )]
+ j AA ( xi − x j ) (1 − ci ) (1 − c j )} + B ( c − ccrit ) 4 + … (6-8)
where j is the interaction energy between a where A ~ (T/TcMF – 1) < 0 for T < TcMF and
pair of atoms. B > 0. It is well known that the actual
Carrying out the coarse-graining defined critical temperature Tc does not coincide
in Eq. (6-1) we expect that Ᏼ {ci}/kBT will with the mean-field prediction. Therefore,
be replaced by it is usually assumed that the parameters
appearing in the actual coarse-grained
DᏲ {c ( x )} free energy fcg (c) are not the mean-field
(6-4)
kB T parameters A, B, … , but rather these pa-
= ∫ d d x { fcg [c ( x )]/ kB T + 12 r 2 [∇c ( x )]2 } rameters are “renormalized” due to short-
wavelength fluctuations, and also r (Eq.
where d is the spatial dimensionality of the (6-5)) may thus be modified. Therefore,
system and r is the range of the effective fcg (c) and r are not calculated from micro-
interaction in Eq. (6-3) (Dj (xi – xj) ∫ j AA scopic models such as Eq. (6-3), but are
+ j BB – 2 j AB): treated as phenomenological input parame-
ters of the theory, which are fitted to ex-
∑ ( xi − x j )2 Dj ( xi − x j ) periment.
j
r2 = (6-5)
2 d ∑ Dj ( xi − x j )
j
420 6 Spinodal Decomposition
6.2.2 The Cahn – Hilliard – Cook and using this result in the continuity rela-
Nonlinear Diffusion Equation tion, Eq. (6-10), we obtain the Cahn – Hil-
liard nonlinear diffusion equation (Cahn,
Since in the total volume V the average
1961):
concentration,
c– = (1/V ) Ú dx c (x, t) ∂c ( x, t )
(6-9) = M ∇2 (6-14)
∂t
is conserved, the time-depent concentra-
⎧⎪⎛ ∂ fcg [c ( x, t )] ⎞ ⎫⎪
tion field c (x, t) satisfies a continuity equa- × ⎨⎜ ⎟ − r kB T ∇ c ( x , t ) ⎬
2 2
tion, ⎪⎩⎝ ∂c ⎠T ⎪⎭
∂c ( x, t ) One immediately obvious defect of this
+ ∇ ⋅ j ( x, t ) = 0 (6-10)
∂t equation is its completely deterministic
where j (x, t) is the concentration current character, which implies that random sta-
density. Following standard nonequilib- tistical fluctuations are disregarded (apart
rium thermodynamics (de Groot and Ma- from fluctuations included in the initial
zur, 1962), j (x, t) is assumed to be propor- condition, the state at temperature T0 where
tional to the gradient of the local chemical the quench starts). This defect can be rem-
potential difference m (x, t): edied, following Cook (1970), by adding a
random force term h T (x, t) to Eq. (6-14):
j (x, t) = – M —m (x, t) (6-11)
∂c ( x, t )
where M is a mobility that is discussed be- = M ∇2 (6-15)
low. ∂t
In the thermal equilibrium the chemical ⎧⎪⎛ ∂ fcg [c ( x, t )] ⎞ ⎫⎪
× ⎨⎜ ⎟ − r kB T ∇ c ( x , t ) ⎬
2 2
potential difference is given as a partial de-
⎩⎪⎝ ∂c ⎠T ⎭⎪
rivative of the Helmholtz energy F (c, T ):
+ h T ( x, t )
m = (∂F/∂c)T ; (6-12)
Here h T (x, t) is assumed to be delta-corre-
remember that the condition for two-phase lated Gaussian noise, and the mean-square
coexistence, the equality of chemical po- amplitude ·h 2T ÒT is then linked to the mobil-
tential differences ity M via a fluctuation-dissipation relation.
m1 = ( ∂F / ∂c )T | ccoex
( 1 ) = m 2 = ( ∂F / ∂c ) T | ( 2 )
ccoex ·h T (x, t) h T (x¢, t¢)ÒT
is the physical content of the double-tan- = ·h T2 ÒT —2 d(x – x¢) d(t – t¢) (6-16)
gent construction shown in Fig. 6-3 c. Eq. ·h T2 ÒT = 2 kB T M (6-17)
(6-12) is generalized to an inhomogeneous
nonstationary situation far from equilib- Eqs. (6-15) – (6-17) constitute the main re-
rium, where both c (x, t) and m (x, t) depend sults of this section, on which all further
on space and time, by defining m (x, t) as treatment presented here is based. At this
a functional derivative of the Helmholtz point we stress the main assumptions that
energy functional DᏲ in Eq. (6-4): have been made either explicitly or tacitly:
m (x, t) ∫ d (DᏲ {c (x, t)})/dc (x, t) (6-13) (i) Effects due to the lattice anisotropy of
the solid have been ignored. In a crystalline
Inserting Eq. (6-14) into Eq. (6-13) yields
solid the interfacial free energy between
m (x, t) = (∂ fcg /∂c)T – r2 kB T —2c (x, t) co-existing A-rich and B-rich phases will
6.2 General Concepts 421
see below (Secs. 6.2.5 and 6.2.6), but nev- periment (Fig. 6-5 a, b) nor in simulations
ertheless it is instructive to study it! (Figs. 6-5 c, d) is such a time-independent
Given this assumption, we may linearize intersection point observed. Also in the
Eq. (6-15) (or Eq. (6-14) if we also neglect Cahn plot, R (k)/k2 plotted versus k2, in-
the thermal noise) in dc (x, t). Then Eq. (6- stead of the predicted linear behavior
14) becomes (Fig. 6-6 a), 2 R (k)/k2 = 2 D0 (1 – k2/kc2) with
a negative diffusion constant
∂
d c ( x, t ) = M ∇ 2 (6-18)
∂t D0 = – M(∂2 fcg /∂c2 )T, c– (uphill diffusion)
⎧⎛ ∂ 2 f ( c ) ⎞ ⎫
⎪ cg 2⎪ curvature is typically observed (Fig. 6-6 b).
× ⎨⎜ ⎟ − r 2
k B T ∇ ⎬ d c ( x, t )
⎪⎩⎝ ∂c
2
⎠ T, c ⎪⎭ There are several possible reasons why
the simple linearized theory of spinodal de-
and, introducing Fourier transformations composition as it has been outlined so far is
dck (t) ∫ ∫ dd x exp (i k · x) dc (x, t) (6-19) invalid:
(i) Fluctuations in the final state at the
Eq. (6-18) is solved by a simple exponen-
temperature T must be included (Cook,
tial relaxation,
1970). This problem will be considered at
dck (t) ∫ dck (0) exp [R (k) t] (6-20) the end of this subsection.
(ii) Nonlinear effects are important dur-
with the rate factor R (k):
(6-21) ing the early stages of the quench. This
R (k) ∫ M k2 [(∂2 fcg /∂c2 )T, c– + r 2 kB T k2 ] problem will be discussed in Secs. 6.2.5
and 6.2.6.
The equal-time structure factor S (k, t) at (iii) There is already an appreciable re-
time t after the quench: laxation of the structure factor occurring
S (k, t) ∫ ·dc– k (t) dck (t)ÒT (6-22) during the quench from T0 to T if the cool-
ing rate g = – dT/dt is finite (see Sec. 6.2.7).
where ·…ÒT denotes a thermal average, (iv) The concentration field is coupled to
then also exhibits a simple exponential re- another slowly relaxing variable (Binder et
laxation: al., 1986; Jäckle and Pieroth, 1988). Here
S (k, t) ∫ ST0 (k) exp [2 R (k) t] (6-23) we only very briefly outline the idea of the
approach of Binder et al. (1986). Accord-
Here the prefactor ing to Eqs. (6-23) and (6-21), the maximum
(6-24)
growth rate of the structure factor occurs at
ST0 (k) ∫ ·dc– k (0) dck (0)ÒT ∫ ·dc– k dck ÒT0
–
Rm = R (km), km = kc / ÷2 (6-26)
is simply the equal-time structure factor in
thermal equilibrium at temperature T0 be- Suppose now the concentration couples
fore the quench. Note that R (k) is positive linearly to a non-conserved variable a (t)
for 0 < k < kc , with describing, for example, structural relaxa-
(6-25)
tion, whose fluctuations in the absence of
kc ∫ 2 p/lc = [– (∂ fcg /∂c )T, c– /(r kB T )]1/2
2 2 2
any coupling would decay exponentially,
Thus, whereas the structure factor should proportional to exp (– g t). The decay of
exhibit exponential growth in this region, concentration fluctuations will be affected
for k = kc it should be time-independent, (i) if the coupling between the variables
S (kc , t) = S (kc , 0). However, neither in ex- a (t) and c (x, t) is sufficiently strong, and
6.2 General Concepts 423
Figure 6-5. (a) Neutron small-angle scattering intensity vs. scattering vector k (b : scattering angle) for an
Au – 60 at.% Pt alloy quenched to 550 °C (Singhal et al., 1978). (b) Time evolution of the structure factor
at 541 K in Al – 38 at.% Zn after subtraction of the prequench scattering. Solid lines are the best fit to the data
using the LBM theory, Eq. (6-60) (Mainville et al., 1997). (c) Time evolution of the structure factor S (k, t) ac-
cording to a Monte Carlo simulation of a three-dimensional nearest-neighbor Ising model of an alloy at critical
concentration and temperature T = 0.6 Tc . Due to the periodic boundary condition for the 30 ¥ 30 ¥ 30 lattice, k is
only defined for discrete multiples of (2 p)/30; these discrete values of S (k, t) are connected by straight lines
(Marro et al., 1975). (d) Time evolution of the normalized structure factor S̃ (k, t) vs. k for the discrete version
of the Ginzburg – Landau model (Eq. (6-4)), namely
Ᏼ / kB T = ∑ [ A ( cl − ccrit ) 2 + B ( cl − ccrit ) 4 ] + ∑ 1
2 C ( cl − c m ) 2
l 〈l m〉
where cl is a continuous variable representing the average concentration in the l th cell of size L ¥ L, and the
-
phenomenological constants A, B, and C have been chosen as A/C = – 2.292, ÷B/C = 0.972 (C can be scaled out
by redefining the cl s). Data are for an N = 40 ¥ 40 lattice with periodic boundary conditions at times t = 0, 10,
20, …, 90 Monte Carlo steps (MCS) per site. The arrow indicates the estimate for the wavenumber of maximum
- ----
growth, km (0) = kc / ÷2 = ÷ – A/C (Milchev et al., 1988).
424 6 Spinodal Decomposition
(ii) if the rates Rm and g are of the same or in fluid polymer mixtures near their
order. In the absence of any coupling glass tansition (e.g., Meier and Strobl,
between these variables their fluctuations 1988). The slow variables are then ex-
would decay independently of each other, pected to relax with a broad spectrum of
with decay rates being given as G+ (k) = g rates rather than with a single rate g. Fluc-
(relaxation of a (t) is then independent of k)
and G– (k) = – |D0 | k2 (1 – k2/kc2) (spinodal
decomposition). This mode spectrum of
uncoupled structural and concentration
fluctuations is shown by the dashed curves
in Fig. 6-7). However, both relaxation rates
become strongly modified if these vari-
ables are coupled. The strength of this
coupling can be related to 1 – D∞ /D0 , where
D0 and D∞ are the low- and high-frequency
limits of the diffusion constant. Full curves
in Fig. 6-7 show for three (arbitrary)
choices of parameters, the relaxation rates
G+ (k) and G– (k) for the case of coupled
variables. It is seen that a “mixing” of
interdiffusion and relaxation occurs; for
small k the interdiffusion (or spinodal de-
composition) is given by G– (k), for large k
by G+ (k), and for intermediate k both expo-
nentials exp [– G+ (k) t)] and exp [– G– (k) t)]
contribute to the growth and/or decay of
concentration fluctuations. Plotting the
mode G– (k) which describes spinodal de-
composition for small k in the form of a
Cahn plot, we encounter pronounced cur- Figure 6-7. Mode spectrum {G+ (k), G– (k)} (full
vature. curves) of an unmixing system coupled to a slow var-
Such a coupling where Rm and g are of iable plotted vs. k2/k c2 for three parameter choices. In
the absence of any coupling, the slow variable would
comparable size might occur if we study
relax with a rate G+ (k) = g (broken horizontal straight
spinodal decomposition in glasses (see for lines) and the unmixing system would relax with a
example Yokota, 1978; Acuña and Craie- rate G– (k) = 4 Rm (k2/k c2 ) (1 – k2/k c2 ) (broken curves).
vich, 1979; Craievich and Olivieri, 1981) All rates are normalized by Rm (Binder et al., 1986).
6.2 General Concepts 425
D̃eff (q, t) following from Eqs. (6-27) and distance and gyration radius of these coils
(6-29) are plotted against the (rescaled) then is (Flory, 1953; De Gennes, 1979):
square of the scattering vector, q2 (where
q = k/kc , t = 2 M r 2 T kc4 t, S̃ = r2 kc2 S). It is ·RA2 Ò = sA2 NA , ·RB2 Ò = s B2 NB , (6-31)
2
seen that on the level of the Cahn – Hilli- ·Rgyr, AÒ = –16 sA2 NA , 2
·Rgyr, BÒ = –16 s B2 NB
ard – Cook (CHC) approximation there is
initially some shift in the position at which In the Flory – Huggins (FH) approximation
S (k, t) has its maximum, although this shift (Flory, 1953; Koningsveld et al., 1987) the
is more pronounced when nonlinear effects expression corresponding to Eq. (6-4) be-
are taken into account, as is approximated comes (De Gennes, 1980)
by the theory of Langer et al. (1975), which DᏲ {F ( x )}
we shall refer to as the Langer – Baron – kB T
{
= ∫ d d x fFH [F ( x )]/ kB T (6-32)
Miller (LBM) approximation. It is seen that 1 1 ⎫
the main distinction between the CHC and + a2 [∇F ( x )]2 ⎬
2 F (1 − F ) ⎭
LBM approximations during the early stages
is the lack of a common intersection point in where F (x) is the volume fraction of A seg-
the LBM approximation. Also the growth of ments, 1 – F (x) is the volume fraction of B
S̃(q, t ) with t is generally slower (note the segments, the mixture is assumed to be
difference in the ordinate scales!). Qualita- incompressible, and the lattice spacing a of
tively, however, the behavior is similar, and the Flory – Huggins lattice model is given
this is also true for the behavior of D̃eff (q, t ). by a2/[F (1 – F )] ∫ sA2 /F + s B2 /(1 – F ). Note
Note that D̃eff (q, t) is distinctly curved in that unlike Eq. (6-4) the parameter a is not
both cases (apart from the limit t Æ 0). related to the range of interaction, but
rather reflects the random coil structure of
the polymer chains. Owing to the connec-
6.2.4 Spinodal Decomposition
tivity of the polymer chains, the entropy of
of Polymer Mixtures
mixing term is much smaller than in Eq. (6-
Here we consider the modifications nec- 6) (Flory, 1593):
essary for describing binary mixtures of
long flexible macromolecules. In a dense 1 F ln F
fFH (F ) = (6-33)
polymer melt, the configurations of these kB T NA
linear chain molecules (we disregard here (1 − F ) ln (1 − F )
+ + c F (1 − F )
star polymers, branched polymers, copol- NB
ymers, and also interconnected networks,
where the Flory – Huggins parameter c con-
although some aspects of the theory can be
tains all enthalpic contributions that lead to
extended to these more complex situations,
unmixing. If c does not depend on volume
see Binder and Frisch (1984)) are random
fraction, then the spinodal curve resulting
Gaussian coils interpenetrating each other.
from Eq. (6-33) is given by
In terms of subunits (“Kuhn segments”) of
lengths sA and sB for the two types of 2 cs (F ) = (F NA)– 1 + [(1 – F ) NB]– 1 (6-34)
polymers A and B, the degree of polymer-
ization XA and XB is expressed in terms of and the critical point occurs at
chain lengths NA and NB as XA = NA gA and
XB = NB gB, if each subunit contains gA and Fc = ( NA / NB + 1) −1
gB monomers. The mean square end-to-end c c = ( NA−1/ 2 + NB−1/ 2 )2 / 2
6.2 General Concepts 427
These results assume well-defined chain (6-6)) only in the long wavelength limit,
2 1/2
lengths (“monodisperse polymers”), but where kc ·Rgyr Ò < 1 (for both types of
can be generalized to a distribution chains). In this limit L k can be replaced
of chain lengths (“polydispersity”; see by a constant L 0 , and the results are just
Joanny, 1978). special cases of Eqs. (6-9) – (6-30); e.g., for
Another modification concerns the con- a symmetric mixture (NA = NB , sA = sB = a)
stitutive kinetic equation: instead of Eq. (6- we simply find:
11) the connectivity of the chains requires 1/ 2
a non-local relationship for the current den- ⎛ 18 ⎞
2 p / l c = kc = ⎜ ⎟ (1 − c s (F )/ c )1/ 2 ,
sity JF (x, t) relating to volume fraction F ⎝ N a2 ⎠
(De Gennes, 1980; Pincus, 1981), k m = kc / 2 (6-39)
JF (x, t) = – Ú L (x – x¢) —m (x¢, t) dx¢ (6-35) Hence the critical wavelength lc is rescaled
where L (x – x¢) is a generalized Onsager by a large prefactor, namely the coil gyra-
coefficient describing polymer – polymer tion radius, similar to the correlation length
interdiffusion. The rate R (k) in Eq. (6-21) xcoex of concentration fluctuations at the
is then expressed in terms of the Fourier coexistence curve,
transform L k of L (x – x¢), i.e., x coex = (N a2/36)1/2 (c /cc – 1)–1/2 (6-40)
2
R (k) = k L k [STcoll (k)]– 1 (6-36) Eq. (6-39), however, is only applicable for
where the expression for the effective col- “shallow” quenches, where c does not
lective structure factor STcoll (k) can be ex- greatly exceed c s (F). A different behavior
pressed in terms of the structure factors occurs (Binder, 1983) for deep quenches,
SA (k) and SB (k) of the single chains as where c o c s (F) (thus describing the un-
(Binder, 1983, 1984 c, 1987 b) mixing of two incompatible polymers):
1/ 2
[STcoll (k)]– 1 = [F SA (k)]– 1 ⎛ 6 ⎞
kc ≈ ⎜ ⎟ [2 c / c s (Φ )]1/ 2
+ [(1 – F ) SB (k)] – 2 c–1 eff
(F, T, k) (6-37) ⎝ N a2 ⎠
1/ 2
eff
where 2 c (F , T, k) is a wavevector- ⎛ 6 ⎞
km ≈ ⎜ ⎟ [2 c / c s (Φ )]1/ 4 (6-41)
dependent generalization of the term re- ⎝ N a2 ⎠
sulting from the second derivative of
Now the initial wavelength of maximum
∂2 [c F (1 – F )]/∂f2 in Eq. (6-33). The sin-
growth is smaller than the coil radius,
gle-chain structure functions SA (k) and
and km O kc , since the gradient-square ex-
SB (k) are expressed for Gaussian chains by
pansion, Eq. (6-32), is no longer appli-
the Debye function fD (x) as
cable.
SA (k) = NA fD (k2 ·Rgyr,
2
A Ò) As is obvious from Eq. (6-36), the relax-
SB (k) = NB fD (k2 ·Rgyr,
2 ation rate R (k) is the product of two fac-
BÒ) (6-38)
tors, a static factor [STcoll (k)]– 1 which con-
where fD (x) ∫ 2 [1 – (1 – e– x )/x]/x. tains the thermodynamic singularities at
Eqs. (6-36) and (6-37) result (Binder, the critical point (and along the spinodal
1983) from random phase approximations curve; cf Eq. (6-39)), and a kinetic factor
(De Gennes, 1979) and agree with the k2 L k, to which we now turn. Here the fac-
treatment as given by Eqs. (6-9) – (6-30) tor k2 simply reflects the conservation law
(but using Eq. (6-32) instead of Eqs. (6-4) – for the concentration, while L k reflects the
428 6 Spinodal Decomposition
when the concentration c– approaches the tion, which according to Eqs. (6-21), (6-25)
concentration csp (T ) of the spinodal curve, and (6-26) can be written as
defined by
Rm = (1/4) M r2 kB T (2 p /lc )4 (6-47)
2 2
(∂ fcg /∂c )T, c– = csp (T ) = 0 (6-44) vanishes as c– Æ c (T ), and a similar “criti-
sp
Here the critical wavelength lc diverges to cal slowing down” (Hohenberg and Halpe-
infinity. For example, if we adopt Eq. (6-8) rin, 1977) would occur in the growth rate of
with A ~ (T/TcMF – 1), near the spinodal we a supercritical droplet. Thus, within the
obtain, omitting prefactors of order unity, framework of this theory, the spinodal
−1/ 2
curve plays the role of a line of critical
⎡ c − csp ( T ) ⎤ points. We now wish to investigate whether
lc ~ r (1 − T / TcMF ) −1/ 2 ⎢ ⎥
⎣ ccrit − csp ( T ) ⎦
there is a physical signficance to this singu-
c Ê csp ( T ) (6-45) lar behavior.
In Sec. 6.2.1 we emphasized that the def-
On the metastable side of the spinodal inition of fcg is not unique but really in-
curve a similar divergence occurs for both volves a length scale L over which a coarse-
the correlation length x (c– ) of concentra- graining of short-range fluctuations is per-
tion fluctuations and the radius R* of a crit- formed. This is best seen from the attempts
ical droplet, namely (Binder, 1984 b) to calculate fcg explicitly, which can be
approximated by using renormalization
x (c ) ~ R* ~ r (1 − T / TcMF ) −1/ 2 (6-46) group methods (Kawasaki et al., 1981) or
−1/ 2
⎡ csp ( T ) − c ⎤ by Monte Carlo simulation (Kaski et al.,
× ⎢ (2) (1) ⎥ , c Á csp ( T ) 1984). These treatments show that the posi-
⎣ ccoex − ccoex ⎦
tion of csp (T ) depends strongly on the
Although R* diverges as c– approaches length scale L (see Fig. 6-10 a), and there-
csp (T ), the associated free-energy barrier fore for systems with short-range forces, to
DF* against the formation of a critical which these treatments apply, there is no
droplet vanished there (Fig. 6-9). The max- physical significance to the spinodal singu-
imum growth rate of spinodal decomposi- larity (Eqs. (6-44) to (6-47)) whatsoever.
Figure 6-10. (a) Monte Carlo results for the dependence of the relative distance of the “spinodal” from the co-
existence curve on the size of the coarse-graining cell L. The results refer to a simple-cubic nearest-neighbor
Ising magnet in the critical region, and are obtained from sampling the magnetization distribution function
PL (M) in L ¥ L ¥ L sub-blocks of a 243 system. Here, Smax is the value at which ln PL (M) has its maximum, and
corresponds to the coexistence curve if we assume ln PL (M) ~ L3 fcg /kB T, with c = (1 – M)/2, and the “spinodal”
is estimated as inflection point Ms of ln PL (M). By scaling L with the correlation length x , all temperatures
superimpose on one “scaling function” (Kaski et al., 1984). (b) Extrapolation of the inverse collective structure
function versus inverse temperature to locate the spinodal temperatures Tsp (c–) or their inverse (arrows). Here
Monte Carlo simulation data for a polymer mixture are used, the polymers (A, B) being modelled as self- and
mutually avoiding random walks on the simple cubic lattice with NA = NB = N = 32 steps, at a concentration of
vacancies fv = 0.6. Values on the curves are the reduced volume fraction FA /(1 – Fv ) of monomers of A chains.
If two neighboring lattice sites are taken by monomers of the same kind, an energy e is obtained, and thus an en-
thalpic driving force for phase separation is created. From Sariban and Binder (1989 b). (c) Phase diagram of the
model for a polymer mixture as described in (b), displaying both the true coexistence curve (binodal) and the
extrapolated spinodal. From Sariban and Binder (1989 b).
6.2 General Concepts 431
Another way of showing this comes from Heermann, 1984 a, b). Although very close
a closer examination of the procedures to the critical point Tc such system behave
by which experimentalists locate spinodal qualitatively like short-range systems (this
curves. Eqs. (6-30) and (6-37) imply that is to be expected from the so-called “uni-
versality principle” (Kadanoff, 1976)), far-
[S coll (k Æ 0)]– 1 ~ (∂2 f /∂c2) – ~ T – T (c– )
T cg T, c sp
ther from Tc a well-defined mean-field crit-
where Tsp (c– ) is the inverse function of ical region exists. It is this region where
csp (T ) in the c – T plane. Thus a plot of both the coarse-graining defined in Eq. (6-
[STcoll (k Æ 0)] vs. temperature should allow 1) and the linearization approximation of
a linear extrapolation to locate Tsp(c– ) from Sec. 6.2.2 make sense, as we will now dis-
the vanishing of [STcoll (k Æ 0)] (cf. Fig. 6- cuss.
10 b). However, the extrapolated spinodal The argument is simply an extension of
determined in this way (e.g., Sariban and the Ginzburg (1960) criterion for the valid-
Binder, 1989 b) (Fig. 6-10 c), is not physi- ity of the mean-field theory for critical
cally meaningful as it crosses the true co- phenomena to nucleation and spinodal de-
existence curve (binodal) near Tc , which is composition (Binder, 1984 b): nonlinear
physically impossible. This happens be- terms in dc (x, t) can be neglected if their
cause, in reality, mean-square amplitude in a coarse-grain-
ing cell is small in comparison with the
[STcoll (k Æ 0)]c––1= ccrit ~ (T – Tc )g
concentration difference squares in the
with g ≈ 1.24, a critical exponent different system, over which relevant nonlinear ef-
from the mean-field result g = 1. Hence fects are felt:
a linear extrapolation fails near c– = ccrit . At
·[dc (x, t)]2 Ò [c– – c (T )]2
T, L sp (6-48)
strongly off-critical concentrations this ex-
trapolation is also ambiguous, because usu- We estimate ·[dc (x, t)]2 ÒT, L as
ally actual data cannot be taken deep in the
·[dc (x, t)]2 ÒT, L ≈ ·[dc (x, 0)]2 ÒT, L exp[2 Rm t]
metastable phase for temperatures close to
Tsp(c– ). where Rm is the maximum growth rate de-
This situations is different, however, for fined in Eq. (6-26), and the initial mean-
systems with infinitely weak but infinitely square amplitude is related to the correla-
long-range interactions (Penrose and Lebo- tion function of concentration fluctuations
witz, 1971); then the mean-field theory is in the initial state at temperature T0 . Thus,
correct because statistical fluctuations are using Eq. (6-1):
suppressed. At the same time, the lifetime
〈[d c ( x, 0 )]2 〉 T , L = 〈[d c ( x )]2 〉 T0 , L
of metastable states is infinite because ho-
1
mogeneous nucleation is no longer pos- = 6 ∑ [ 〈 ci c j 〉 T0 − c 2 ]
sible (“heterophase fluctuations” are su- L i, j
pressed at the same times as “homophase 1
= 3 ∫ dx [ 〈 c ( 0 ) c ( x )〉 T0 − c 2 ] (6-49)
fluctuations”). The spinodal curve is the L
limit of metastability here.
the summations over i and j being restrict-
While such a system with infinitely
ed to sites within the cell of size L3 cen-
long-range interactions is clearly artificial,
tered at x and in the last step the sums are
it makes sense to consider systems with
converted into integrals (one sum  is can-
long but finite range r of the interactions i
(Heermann et al., 1982; Binder, 1984 b; celled against a factor L3 making use of the
432 6 Spinodal Decomposition
related in spirit to the Langer – Schwartz clusters of size l and l¢, and cl, l¢ is the rate
(1980) and Kampmann – Wagner (1984) factor for the inverse “coagulation” reac-
treatments of concomitant nucleation, tion. These rate factors are assumed to be
growth, and coarsening (see Wagner et al., independent of time t and hence Eq. (6-54)
2001). Each state of the system is charac- also describes the concentration fluctua-
terized by a cluster size distribution n–l (t), tions in thermal equilibrium, where de-
where l is the number of B atoms contained tailed balance must hold between splitting
within a cluster (Fig. 6-11 a) and the bar and coagulation reactions:
represents an average over other “cluster
Sl + l¢, l¢ nl + l¢ = cl, l¢ nl nl¢ ∫ W (l, l¢) (6-55)
coordinates” (cluster surface area, shape,
etc.) which are not considered explicitly. where nl is the cluster concentration in
The time evolution of the cluster size equilibrium and W (l, l¢) is a cluster reac-
distribution in a quenching experiment tion matrix in equilibrium. While nl (t = 0)
is described by a system of kinetic equa- for a random distribution of atoms in the
tions: alloy is just the cluster distribution in the
∞ well-known percolation problem (Stauffer,
d
nl (t ) = ∑ Sl + l ′, l ′ nl + l ′ (t ) 1985), nl and W (l, l¢) are not explicitly
dt l ′ =1 known, but can be fixed by plausible as-
l −1
1 sumptions (Binder, 1977; Mirold and
− ∑ Sl, l ′ nl (t ) Binder, 1977). Then Eqs. (6-54) and (6-55)
2 l ′ =1
l −1 can be solved numerically (see Fig. 6-11 b).
1
+ ∑ cl − l ′, l ′ nl ′ (t ) nl − l ′ (t ) What must happen is that n–l (t) for t Æ ∞
2 l ′ =1 develops towards nl , the cluster size distri-
∞
bution in the state at the coexistence curve
− ∑ cl, l ′ nl (t ) nl ′ (t ) (6-54) ∞
l ′ =1 (1)
c coex (1)
, with c coex = ∑ l nl . The excess
l =1
Here Sl + l¢, l¢ is a rate factor for a splitting ∞
reaction of a cluster of size l + l¢ into two concentration c– – c coex
(1)
= ∑ l [nl (0) – nl ] is
l =1
redistributed into macroscopic B-rich do- lations (Marro et al., 1975; Rao et al.,
(2)
mains (of concentration c coex ), i.e., occurs 1976; Sur et al., 1977).
at cluster size l Æ ∞ for t Æ ∞. For interme- The description in terms of Eqs. (6-54)
diate times, we find a nonmonotonic clus- and (6-55) contains both nucleation and
ter size distribution (Fig. 6-11 b); a mini- growth and coagulation as special cases
mum occurs at the critical size l* of nucle- (Binder, 1977), and it can be used as a ba-
ation theory (Binder and Stauffer, 1976 a), sis for understanding the gradual transi-
while the broad maximum at larger sizes is tion from nucleation to spinodal decompo-
due to growing supercritical clusters which sition (Binder et al., 1978). In the meta-
have already been nucleated. For l < l* the stable regime the density of unstable fluc-
cluster size distribution is basically the tuations (“critical” and “supercritical” clus-
equilibrium size distribution of a slightly ters with l < l*) is very small (Fig. 6-12 a),
supersaturated solid solution. As time goes because the energy barrier D F* p kB T.
on, the peak in n–l (t), representing the Near the spinodal curve, on the other hand,
supercritical growing clusters, shifts to D F* ≈ kB T, and hence there is a high den-
larger and larger cluster sizes, and at the sity of unstable fluctuations: DF* here no
same time the supersaturation is dimin- longer limits the growth, but rather the
ished, until for t Æ ∞ the peak has shifted conservation of concentration. Near grow-
to l Æ ∞. This separation of clusters into ing clusters, c (x, t) locally decreases, and
two classes – small ones describing con- then no other cluster can grow there. Ow-
centration fluctuations in the supersatu- ing to this “excluded volume” interaction
rated A-rich background and large ones of clusters, a quasiperiodic variation of
describing the growing B-rich domains – concentration results (Fig. 6-12 a, bottom),
can also be shown analytically from Eqs. roughly equivalent to a wavepacket of
(6-54) and (6-55) (Binder, 1977). This pic- Cahn’s concentration waves. But the latter
ture of the phase separation process also are not growing independently, rather they
emerges very clearly from computer simu- are strongly interacting; hence it is more
Figure 6-13. Various regimes in the phase diagram of a binary mixture AB, showing part of the plane formed
by temperature T and volume fraction F of the B component (only volume fractions F < Fcrit are shown, since
this schematic phase diagram is symmetric around the axis F = Fcrit ). The horizontal broken line separates the
non-mean-field critical regime (top) from the mean-field critical regime (bottom). The two solid curves are the
coexistence curve (left) and the spinodal curve (right). The two dash-dotted curves on both sides of the spino-
dal limit represent the regime where a gradual transition from nucleation to spinodal decomposition occurs. The
linearized theory of spinodal decomposition (Fig. 6-1 a) should hold to the right of these dash-dotted curves. The
regime between the coexistence curve and the left of the two broken curves is described by classical nucleation
theory (compact droplets, Fig. 6-1 b). The regime between the right broken curve and the left dash-dotted curve
is described by “spinodal nucleation” (ramified droplets). From Binder (1984 b).
Figure 6-14. Schematic plots of the nucleation free energy barrier for (a) the mean-field critical region
of a d-dimensional alloy system, i.e., r d (1 – T/ Tc ) (4 – d )/2 o 1 and (b) the non-mean-field critical region,
i.e., r d (1 – T/ Tc ) (4 – d )/2 < 1. The gradual transition from nucleation to spinodal decomposition occurs for
DF*/kB Tc ≈ 1. From Binder (1984 b).
6.2 General Concepts 437
So far the most popular (if approximate) that is similar to Eq. (6-27), namely,
approach is the decoupling approximation
d
suggested by Langer et al. (1975). Exact S ( k, t ) = − 2 M k 2 (6-60)
equations of motion are derived for the dt
probability distributions r1 [c (x)], r2 [c (x1), ⎧ ⎡⎛ ∂ 2 f ⎞ ⎤ ⎫
⎪ ⎪
c (x2)], etc. Here r1 is the probability den- × ⎨ ⎢⎜ 2 ⎟ + a (t ) + r 2 T k 2 ⎥ S ( k, t ) − kB T ⎬
⎪⎩ ⎢⎣ ⎝ ∂c ⎠ T, c ⎥⎦ ⎪⎭
sity that, at point x, the concentration c (x)
occurs, and r2 is the corresponding two-
where all nonlinear effects are now con-
point function. As expected, the equation
tained in a correction term a (t), which it-
of motion for r1 involves r2, and the equa-
self depends on S (k, t) in a nonlinear way.
tion of motion for r2 involves the three-
As noted above, the coefficients A, B,
point function r3 , etc., so that an infinite
and r in Eqs. (6-5) and (6-8) are adjusted
hierarchy of equations of motion is gener-
such that the critical behaviors of the co-
ated.
existence curve, critical scattering inten-
This hierarchy is decoupled by the fol-
sity, and correlation length (at the coexis-
lowing approximation for the two-point
tence curve) are reproduced:
function [d c (x) = c (x) – c– ]:
(6-59) (1)
(c coex – ccrit )/ccrit = B̂ (1 – T /Tc ) b (6-61 a)
r2 [c ( x1 ), c ( x2 )] = r1 [c ( x1 )] r1 [c ( x2 )] –g
ccoex = Ĉ (1 – T /Tc ) (6-61 b)
⎧ 〈d c ( x1 ) d c ( x2 )〉 T d c ( x1 ) d c ( x2 ) ⎫
× ⎨1 + ⎬ xcoex = x̂ (1 – T /Tc ) –n
(6-61 c)
⎩ 〈 (d c ) 2 〉 T 〈 (d c ) 2 〉 T ⎭
The motivation for Eq. (6-59) is the fol- where B̂, Ĉ, and x̂ are the appropriate criti-
lowing: if there were no correlation be- cal amplitudes and b, g, and n the asso-
tween concentrations at points x1 and x2, ciated critical exponents (Stanley, 1971;
the probability r2 would just be the product Binder, 2001). It now turns out that the
of the one-point probabilities. Therefore, strength of nonlinear effects is controlled
the correction of this factorization approxi- by the inverse of a parameter f0 which near
mation is made proportional to the two- Tc is expressed as (Billotet and Binder,
point correlation function ·dc (x1) dc (x2 )Ò T . 1979)
In this way, Eq. (6-59) yields a closed equa- f0 ~ x̂ d B̂ 2 Ĉ – 1 (1 – T /Tc )g + 2 b – dn (6-62)
tion of motion for the probability r1 . This
equation is then solved approximately, as- for a d-dimensional system. In the non-
suming that the coarse-grained free energy mean-field critical region, the hyperscaling
has the Landau form (Eqs. (6-5) and (6-8)). relation dn = g +2 b (Kadanoff, 1976) elimi-
The coefficients A, B and r are adjusted nates the temperature dependence from Eq.
self-consistently such that the resulting (6-62). In addition, two-scale factor uni-
state equilibrium is correctly described in versality (Stauffer et al., 1972) implies that
the non-mean-field critical region. the critical amplitude ratio xˆd B̂ 2/Ĉ and
This approach has also been worked out hence f0 is a universal constant of order
for the dynamics of non-conserved order unity (Billotet and Binder, 1979), i.e.,
parameters (Billotet and Binder, 1979) and f0 ≈ 9.45. On the other hand, if we consider
the validity of this approximation has been a quench into the mean-field critical re-
studied carefully (see also Binder et al., gion, again adjusting the coefficients A and
1978). The final result can be cast in a form B in Eq. (6-8) but now using mean-field
6.2 General Concepts 439
results in Eqs. (6-61 a – c), namely, b = –12 , However, for large values of f0 the linear
g = 1, and n = –12 , xˆ ~ r, we obtain theory does hold initially, consistent with
the simulations (Fig. 6-15). The same con-
f0 ~ r d (1 – T /Tc )(4 – d)/2 clusion emerges from the time dependence
for r d (1 – T /Tc )(4 – d)/2 p1 (6-63) of the rescaled effective second derivative
m – (T, t) in Eq. (6-60):
It is seen that f0 is now simply proportional
to the parameter appearing in the Ginzburg m – (T, t) ∫ 1 + a (t)/[∂2 f /∂c2]T, c (6-64)
criterion, and hence very large. For (see Fig. 6-18). Nonlinear effects are negli-
r d (1 – T /Tc )(4 – d)/2 ≈ 1 a crossover to the uni- gible as long as m – = 1, while the decrease
versal constant f0 ≈ 9.45 in the non-mean- in m – is a signal of nonlinear effects, since
field critical region occurs. a (t) is always negative. The physical
Fig. 6-17 shows the time evolution for significance of m – (T, t) is that it describes
the structure factor for two choices of f0. the ratio between the actual kc2 (t), where
Here again a rescaling of the structure fac- the growth rate in Eq. (6-60) changes sign,
tors Ŝ = r 2 kc2 S and of time t = 2 M r 2 T kc4 t and the corresponding prediction of the
was used, while q = 40 k /(2 p kc) here. It is Cahn – Hilliard theory. Fig. 6-18 implies
seen that for f0 ≈ 9.45 the linear theory is that even if the nonlinear effects are very
indeed invalid from the start, as expected. strong the Cahn – Hilliard prediction for kc2
differs from the actual kc2 (t) by about a fac-
tor of two at most; therefore, the Cahn –
Hilliard theory is certainly useful for esti-
6.2.7 Effects of Finite Quench Rate al., 1966; Carmesin et al., 1986). Here we
only quote a few results from the model
So far both the theory and the discussed calculation of Carmesin et al. (1986), since
simulations have always assumed instanta- the work by Houston et al. (1966) only
neous quenches from an initial temperature considers fluctuations in the initial state
Ti to a final temperature Tf . With regard to and not in the intermediate states visited in
actual experiments, this is extremely ideal- the quench, which is an approximation that
ized. Often the early stages of spinodal de- usually cannot be justified.
composition have already been passed dur- As an example, Fig. 6-19 shows a situa-
ing such a continuous quench where the tion similar to the quench treated in Fig. 6-
temperature is gradually lowered. While it 8, but here the quench is not carried out in-
is believed that the late stages are not af- stantaneously from infinite temperature to
fected by the “quenching history”, the lat- T /Tc = 4/9, but takes several steps: at t = 0
ter can have a drastic effect on both early the system is cooled instantaneously to
and intermediate stages of phase separa- T1 /Tc = 0.75185, at t = 1 to T2 /Tc = 0.67667,
tion. Unfortunately, in general, the problem at t = 2 to T3 /Tc = 0.60148, at t = 3 to
is complicated – the behavior of both the T4 /Tc = 0.5263, and at t = 4 to T /Tc = 4/9,
thermodynamic functions and of the mobil- where the system is later maintained. This
ity M (T ) in the full regime from Ti to Tf stepwise quenching is more readily acces-
may affect the phase separation behavior. sible to calculation than a fully continuous
Thus, relatively little theoretical effort has quench. A further simplification of the cal-
been devoted to this problem (Houston et culation is to neglect any temperature de-
tion transition of a correlated percolation tion is reached where t1 = t2 (for still lower
problem at cp(corr) (T ). This line starts out for concentrations there is no percolation at
T Æ ∞ at the percolation concentration for all), whereas for increasing concentration
random percolation (cp(random) ≈ 0.312 for another critical concentration is reached
the simple cubic lattice (Stauffer, 1985)) where t2 Æ ∞. For still higher concentra-
and bends over to the left, until it hits the tions the system percolates as the time
coexistence curve at the point T /Tc ≈ 0.96, exceeds t1 and then remains percolating
c– ≈ 0.22 (Müller-Krumbhaar, 1974). This throughout.
line continues in the two-phase region as A different behavior occurs if a system
a transient time-dependent phenomenon, with a much coarser resolution is studied;
cp(corr) (T, t) (see Fig. 6-22). Hayward et al. at late stages, where the system is phase-
(1987) and Lironis et al. (1989) have separated on a length scale km–1 (t) into the
shown that for certain concentrations the coexisting phases with concentrations
configuration is percolating for a time given by the two branches of the coexis-
(1) (2)
interval t1 < t < t2 , whereas it does not per- tence curve, c coex and c coex , respectively. It
colate for 0 & t1 and for t 7 t2 . If the con- now makes sense to consider percolation
centration decreases, a critical concentra- phenomena on much larger length scales
than the lattice spacing a. Suppose we di-
vide our system again into cells of linear
dimension L (Eq. (6-1)), with a L km–1 (t).
Most of these cells will then have concen-
(1) (2)
trations close to either c coex or c coex . We
may now define clusters consisting of
neighboring cells with concentrations in a
(2) (2)
given interval [c coex – dc/2, c coex + dc/2]
and we may ask whether these clusters are
well separated from each other or if they
form an infinite percolating network. Since
l (t Æ ∞) Æ ∞ we may also take L Æ ∞ in
this limit and therefore dc Æ 0. Hence there
is no longer any ambiguity in this coarse-
grained percolation problem. We expect
that this “macroscopic percolation” will
occur at a critical volume fraction f p of the
minority phase, which does not depend on
temperature. Therefore, the line of macro-
scopic percolation concentrations is simply
cp(macro) = c coex
(1) (2)
+ (c coex (1)
– c coex ) fp
and hence must end in the critical point.
Figure 6-22. Part of the phase diagram of the simple This line is also shown schematically in
cubic, nearest-neighbor lattice-gas model, showing
Fig. 6-21. Note that this line will be ob-
the percolation transition line cp(corr) (T, t), where the
time t refers to an average over the time interval from served experimentally by techniques which
t = 80 to t = 240 MCS per site during phase separa- are sensitive to the “contrast” (i.e., differ-
tion. From Hayward et al. (1987). ence in refractive index) between the two
6.2 General Concepts 445
sence of any elastic interactions (see Sec. Binder, 1977; and Binder et al., 1978) and
6.2.10), a result originally derived by Lif- various extensions (e.g., Langer and
shitz and Slyozov (1961) and Wagner Schwartz, 1980; Kampmann and Wagner,
(1961) holds 1984; Wagner et al., 2001) incorporate the
LSW growth law, nucleation and coagula-
x = 1/3 (6-68)
tion in a phenomenological way. But they
This LSW theory is essentially a mean- do not take into account the statistical fluc-
field theory valid in the limit of zero tuations and the correlations in the diffu-
volume fraction F of the new phase sion field around growing clusters.
[F ∫ (c– – c coex
(1) (2)
)/(c coex (1)
– c coex )] and consid- Eq. (6-68) has now been confirmed by
ers the cluster size distribution n–l (t) that approaches based on scaling ideas (Furu-
was discussed in Sec. 6.2.5, showing that kawa, 1978, 1984, 1985 a, 1988; Bray,
for F = 0 there exists a solution nl (t) = l –1 1998), renormalization group concepts
ñ (l t – d/3 ), which implies that the mean (e.g., Lai et al., 1988; Bray, 1990, 1994),
–
cluster “size” (i.e., volume) l (t) scales as and theories considering fluctuating ran-
–
l (t) ~ t d/3 for t Æ ∞ (for a more detailed dom interfaces (Mazenko, 1994; Mazenko
outline of this theory, see Wagner et al., and Wickham, 1995). However, the most
2001). Thus the cluster linear dimension convincing evidence that Eq. (6-68) is true
–
scales as [l (t)]1/d ~ t1/3, and if we can ex- both for critical and for off-critical
tend this result to nonzero F we would ex- quenches in d = 2 and in d = 3 comes from
–
pect km (t) ~ [l (t)]1/d and hence Eq. (6-68) computer simulations (Amar et al., 1988;
results. However, despite numerous at- Gunton et al., 1988; Gawlinski et al., 1989;
tempts (e.g., Tokuyama and Enomoto, Huse, 1986; Rogers et al., 1988; Rogers
1993; Akaiwa and Voorhees, 1994) even and Desai, 1989; Chakrabarti et al., 1993).
the extension of LSW theory to the case of We emphasize that Eq. (6-68) holds both
small F is only approximately possible, in d = 2 and in d = 3 dimensions. The situa-
and the accuracy of these extensions is tion is different for fluid binary mixtures:
open to doubt (Mazenko and Wickham, the droplet diffusion – coagulation mecha-
1995). Different power laws for not so late nism (also called “Brownian coalescence”)
stages have also been proposed (e.g., predicts (Binder and Stauffer, 1974)
Binder, 1977, invoking a cluster diffusion
km (t) ~ t –1/d (6-69)
and coagulation mechanism, and To-
kuyama and Enomoto, 1993), and observed while mechanisms invoking hydrodynamic
in computer simulations where atomic dif- flow of interconnected structures imply
fusion is mediated by a single vacancy (Siggia, 1979)
moving through the lattice (Fratzl and Pen-
km (t) ~ t –1 , d=3 (6-70)
rose, 1994, 1997). Only for deep quenches
where we must take into account that the and km (t) ~ t –1/2, d = 2 (San Miguel et al.,
mobility M in Eq. (6-11) is itself concen- 1985), in the “viscous hydrodynamic re-
tration-dependent and (almost) vanishing gime” (Bray, 1994), while km (t) ~ t – 2/3 in
in the pure phases, diffusion along inter- the “inertial hydrodynamic regime” (Furu-
faces results in a slower growth law, x = 1/4 kawa, 1985 c).
(Puri et al., 1997). A theoretical problem that is still out-
“Cluster dynamics” approaches such as standing is the calculation of the scaling
Eq. (6-54) (see Binder, 1977; Mirold and function S̃ (z ) in Eq. (6-67) (see Bray
6.2 General Concepts 447
(1994), Puri et al. (1997) and Mazenko call that for fluid binary mixtures even the
(1994, 1998) for recent discussions). We exponent x is different in these two regimes
mention here only a few approaches very (Eqs. (6-69), (6-70)). For the percolative
briefly. Rikvold and Gunton (1982) used regime of fluid mixtures, Furukawa (1984)
Eq. (6-56), following Binder et al. (1978) has proposed an approximate relation for
but treating the depletion zones around qm vs. t
growing clusters more realistically. Clearly,
their model is only qualitative. More ambi- ( qm−1 − 1) [ A*/ B* arctan ( B*/ A* qm )
tious early attempts are due to Furukawa − arctan B*/ A*] = B* t (6-73)
(1984, 1985 a, b), Ohta (1984), Tomita
(1984) and Tokuyama et al. (1987). An where A* and B* are adjustable constants.
equation due to Furukawa (1984) has been Again a more precise treatment has to rely
extensively compared with experimental on numerical calculations (Koga and Ka-
data on Al – Zn alloys (Komura et al., 1985) wasaki, 1991, 1993; Puri and Dünweg,
1992; Valls and Farrell, 1993; Shinozaki
˜
S̃ (z ) ~ z 2/(g̃ 2/2 + z 2 +g ) (6-71) and Oono, 1993; Alexander et al., 1993;
where g̃ = d + 1 for strongly off-critical Bastea and Lebowitz, 1995).
quenches (cluster regime) and g̃ = 2 d for
critical quenches (percolative regime). 6.2.10 Effects of Coherent Elastic Misfit
S̃ (z ) thus exhibits “Porod’s law” (see e.g.,
When phase separation occurs on a crys-
Glatter and Kratky (1982))
talline lattice it is often the case that the
S (k, t) ~ k – (d +1) (6-72) two phases differ slightly in their crystal
in the cluster regime only. Eq. (6-71) also structure or their lattice constants, thus in-
fails to reproduce the exactly established troducing elastic strains in the crystal. The
(Yeung, 1988) behavior for small z , resulting elastic interaction is long-range
S̃ (z ) ~ z 4. For systems without hydrody- and typically anisotropic and may consid-
namics, good empirical forms for S̃ (z ) erably change the phase separation pro-
describing the limiting behavior both for cess, e.g., in metal alloys (for a recent re-
small z and for large z correctly and ac- view, see Fratzl et al., 1999; an extensive
counting both for simulation and experi- treatment of this problem is also given in
ment have been constructed by Fratzl et al. Khachaturyan, 1983).
(1991), but the derivation of such functions The coherent misfit strain tensor eij0 is
from more fundamental theories is still not the strain required to transform the (undis-
fully solved (Mazenko, 1994, 1998; Ma- torted) lattice of one phase into the (undis-
zenko and Wickham, 1995; Bray, 1994). torted) lattice of the other. In metals, parts
Fig. 6-27 shows a comparison of some of this strain can be relaxed by misfit dislo-
theoretical predictions for the halfwidth of cations disrupting the continuity of the lat-
the scaling function S̃ with corresponding tices between the two phases. This process
experiments (Kostorz, 1991). It is intri- will not be discussed here. In cases of non-
guing to note that the behavior of S̃ (z ) does zero misfit strain, eij0 ≠ 0, an influence on
not change much when the volume fraction spinodal decomposition is expected if one
f changes from the “cluster regime” at of the following conditions is violated:
smaller f through a percolation transition (i) both phases have the same elastic
to the interconnected regime at large f . Re- stiffness tensor;
448 6 Spinodal Decomposition
If all five conditions are satisfied, then The first term arises from a dependence of
phase separation can proceed indepen- the elastic constants on composition and
dently of the misfit strains (Bitter – Crum the second one from the composition de-
theorem, see, for example, Cahn and pendence of the misfit strain.
Larché (1984)). Otherwise, we can expect The main difficulty in the use of Eq. (6-
changes in the shape of the single-phase 77) is the determination of Deij . Since the
domains, e.g., from spherical to cuboidal or strains in the alloy can be assumed to relax
plate-like shapes but also in their spatial ar- much faster to their equilibrium values
rangement and coarsening kinetics. Typi- than the concentration profiles, Deij will al-
cally the tendency towards shape changes ways be given by the elastic equilibrium
increases when single-phase droplets be- condition
come larger because the elastic energy (be- ∂tij
ing proportional to the droplet volume) in- ∑ =0 (6-78)
j ∂x j
creases faster with the radius than the sur-
face energy. where the stress tensor is defined by
The introduction of elastic misfit effects Hooke’s law
into the theory of spinodal decomposition
means substituting fcg by fcg + w in Eq. (6- tij = l ijmn Demn (6-79)
15). The function w is the elastic energy As a consequence, the equilibrium strain at
density stored in the lattice each point in the material is a (non-local)
1 functional of the entire concentration pro-
w= ∑ l ijmn Deij ( x ) Demn ( x ) (6-74) file c (x, t). In the special case when the
2 ijmn
elastic constants l ijmn are independent of
where l ijmn is the elasticity tensor or stiff- composition, the elastic problem can be
ness tensor (which may depend on the alloy solved by Fourier transformation (Khacha-
composition). turyan, 1966, 1983) to give
1
Deij (x) = eij (x) – eij0 (x) (6-75) w=
2
∫ ∫ Vel ( x − y ) (c ( x, t ) − c )
is the difference between the strain at posi- × ( c ( y, t ) − c ) d 3 x d 3 y (6-80)
tion x in the elastic equilibrium, eij (x), and or
the misfit strain, eij0 (x). Inserting this into (6-81)
∂w
the nonlinear Cahn – Hilliard equation (Eq. ( x, t ) = ∫ Vel ( x − y ) ( c ( y, t ) − c ) d y
3
age concentration within the specimen is and Onuki, 1990). A discretized version of
called c–. similar equations was used by Khachatur-
In the case of isotropic elasticity and yan and coworkers to study the behavior of
misfit strain, Eq. (6-81) reduces to (Onuki, single precipitates (Wang et al., 1991) or
1989 a – c) the evolution of precipitate morphologies
(Wang et al., 1992, 1993). Moreover,
∂w 2K 2
= h a [c ( x, t ) − c ] (6-82) Monte Carlo simulations of phase separa-
∂c 1 − n P tion have also been performed, including
where h a is the change of lattice constant elastic misfit interactions between un-
with concentration c, that is equally sized atoms on a lattice (Fratzl and
Penrose, 1995, 1996; Laberge et al., 1995,
eij0 (c) = h a (c – c– ) d ij (6-83) 1997; Lee, 1997, 1998; Gupta et al., 2001).
A further approach is the simulation of the
K and nP are Young’s modulus and the
coarsening of particles with sharp inter-
Poisson coefficient of the elastic matrix,
faces to the matrix, describing the elastic
and d ij = 0 if i ≠ j, d ii = 1 (Kronecker delta).
misfit interaction in the framework of mac-
Eq. (6-82) has been introduced by Cahn
roscopic elasticity theory. Recent examples
(1961). The consequence of this equation is
are the studies by Su and Voorhees (1996),
a shift of the spinodal line towards lower
Abinandanan and Johnson (1993, 1996) or
temperatures because Eq. (6-44) has to be
Jou et al. (1997) (see also the review by
replaced by
Fratzl et al. (1999)). The common observa-
∂ 2 fcg ∂ 2 w ∂ 2 fcg 2K 2 tions in all these approaches are that the
2 + 2 = 2 + ha = 0 domains become very anisotropic and typi-
∂c ∂c ∂c 1 − nP
cally orient parallel to crystallographic di-
Hence, the coherent elastic misfit between rections of the alloy crystal (Fig. 6-24 d, e).
the phases may stabilize the solid solution, Moreover, the spatial arrangement of the
even though the demixing into two inco- domains becomes progressively more peri-
herently separated phases would decrease odic. Despite these enormous changes with
the total energy (Cahn, 1961). The reason respect to the case without elastic misfit
is that the coherency condition forces the interactions, the growth law of a typical do-
alloy to store a considerable amount of main size is often still described by the re-
elastic energy in the lattice, which could be sult of the LSW theory (Eq. (6-68)). If ex-
released by creating an incoherent boun- ternal stress is applied, an additional reor-
dary between two regions both having the ientation of the domains either parallel or
(undistorted) equilibrium lattice structure. perpendicular to the applied stress is ob-
For anisotropic stiffness constants l ijmn served (Laberge et al., 1995, 1997; Weinka-
or anisotropic misfit strains eij0 , the chemi- mer et al., 2000). Many of these effects are
cal potential due to elastic interactions, observed in real alloys, most notably for the
∂w/∂c, will depend on the direction in the technically important nickel-based superal-
crystal. This has already been recognized loys (Maheswari and Ardell, 1993; Conley
by Cahn (1962) (see also Khachaturyan et al., 1989; Sequeira et al., 1995; Paris et
(1983) and Onuki (1989 a – c)). Numerical al., 1995, 1997; Fährmann et al., 1995).
solutions of Eq. (6-76) combined with Eq. If the elastic stiffness constants l ijmn de-
(6-81) have shown the development of pend on alloy composition (see the first
strongly anisotropic domains (Nishimori term in Eq. (6-77)), this can result in anom-
450 6 Spinodal Decomposition
that occurs in diverse systems; second, Paris et al., 1995, 1997; Sequeira et al.,
there has been much interest from a theo- 1995). In all these systems, the structure
retical point of view in this phenomenon, factor looks very similar to the early data
and therefore many experimental studies on Au – Pt alloys (Singhal et al., 1978) and
have been carried out in an attempt to test those of Al – Zn (Mainville et al., 1997) re-
some of the theoretical concepts. produced in Fig. 6-5. These results are in
This section cannot present a complete qualitative agreement with the nonlinear
overview of all these experiments. More LBM theory and the computer simulations.
detailed reviews of various aspects of the The hallmarks of the linearized Cahn the-
experimental work can be found in Haasen ory (exponential increase in intensity with
et al. (1984), De Fontaine (1979), Gerold time, time-independent intersection point
and Kostorz (1978), Gunton et al. (1983), of S (k, t), maximum position km of S (k, t)
Goldburg (1981, 1983), Beysens et al. independent of time) are never found. In a
(1988), Kostorz (1988, 1994), Hashimoto few cases it has been found that at early
(1987, 1988, 1993), and Nose (1987), times km (t) is nearly independent of time,
among others. Here we attempt only to e.g., in Mn67Cu33 (Morii et al., 1988), and
give a few representative examples to illus- this fact has been taken as evidence for the
trate some of the points discussed in the validity of the Cahn–Hilliard–Cook (CHC)
theoretical section, and at the same time approximation. In view of the fact that
show similarities – as well as differences – km (t) almost independent of t can also re-
between different systems. sult as an effect of continuous rather than
instantaneous quenching (Sec. 6.2.7) from
a nonlinear theory, we feel that the question
6.3.1 Metallic Alloys
of whether the CHC theory applies to any
Some systems in which spinodal decom- of the metallic systems quantitatively is
position has been very extensively studied still unanswered. Certainly we expect that
are Al – Zn alloys (e.g., Hennion et al., the CHC theory gives a useful order of
1982; Guyot and Simon, 1982, 1988; Si- magnitude estimate of km (0) and the initial
mon et al., 1984; Osamura, 1988; Komura growth rate Rm , as it does for the theoreti-
et al., 1985, 1988; Mainville et al., 1997), cal models (Fig. 6-5 d), provided a spinodal
ternary Al – Zn – Mg alloys (Komura et al., curve is not close. Experiments by Guyot
1988; Fratzl and Blaschko, 1988), Ni- and Simon (1982) and Hennion et al.
based alloys such as Ni – Al, Ni – Ti, Ni – Cr, (1982) were deliberately carried out choos-
and Ni – Mo (Kostorz, 1988; see also ing temperatures both above and below the
Kampmann and Wagner, 1984), Al – Li al- spinodal curve of Al – Zn, for several con-
loys (Furusaka et al., 1985, 1986, 1988; Li- centrations. The structure factor in both
vet and Bloch, 1985; Tomokiyo et al., cases is qualitatively the same, as was pre-
1988; Che et al., 1997; Schmitz et al., dicted (see Sec. 6.2.5). This classical set of
1994; Hono et al., 1992), Mn – Cu alloys experiments in our view definitely shows
(Gaulin et al., 1987), Ni – Si alloys (Chen et that the spinodal singularity does not play
al., 1988; Cho and Ardell, 1997), Fe – Cr al- any role in the unmixing kinetics of Al – Zn
loys (Katano and Iizumi, 1984), and vari- alloys. Note that it is clear that the effective
ous other ternary alloys, e.g., Cu – Ni – Fe potential in this alloy is predominantly
(Lyon and Simon, 1987; Lopez et al., 1993) short range; however, this alloy does have
or Ni – Al – Mo (Fährmann et al., 1995; a significant misfit of atomic radii between
452 6 Spinodal Decomposition
Al and Zn, and thus the “coherent phase di- t is curved and a well-defined exponent x
agram” (where elastic strain fields are not cannot be identified, to the final Lifshitz –
released) is depressed by about 30 K below Slyozov behavior, x = 1/3. Katano and Ii-
the thermodynamic (incoherent) miscibil- zumi (1984) interpret the result x = 1/6 in
ity gap, and the spinodal curves are then terms of the mechanism (Binder and Stauf-
similarly shifted, providing an example fer, 1974) that B-rich clusters in a solid
of the effects of elastic interactions (Sec. solution show a random diffusive motion
6.2.10). Despite these significant long- since B atoms evaporate randomly from the
range elastic forces, the data for Al – Zn cluster and re-impinge at another boundary
spinodal decomposition are surprisingly position, thus leading to a small shift in
similar to the Monte Carlo results on sim- the cluster center of gravity. The resulting
ple nearest-neighbor Ising models of al- cluster diffusivity decreases strongly with
loys, where all such elastic effects are not increasing cluster size. Assuming then that
included. In other cases, such as many bi- two clusters coalesce when they meet in
nary and ternary Ni alloys (see, for exam- their random motions, we arrive at km (t) ~
ple, Fig. 6-25) elastic misfit interactions t –1/(d +3) in d dimensions. Although such a
lead to highly anisotropic precipitate mor- mechanism certainly exists, it is not clear
phologies and to ordered arrangements of whether it ever dominates during a well-
precipitates (see Sec. 6.2.10). defined time interval, since the Lifshitz –
There is ample experimental evidence Slyozov – Wagner (1961) mechanism com-
for the scaling of the structure factor at late petes with it and should dominate, at least
stages (Eq. (6-67)), and the validity of for long times. Monte Carlo simulations
the Lifshitz – Slyozov exponent x = 1/3 (Eq. considering diffusion via vacanies indicate
(6-68)). Sometimes smaller exponents are that cluster – diffusion – coagulation could
found, e.g., km–1 (t) ~ t x with x ≈ 0.13 to be important at low temperatures (Fratzl
x ≈ 0.2 (see, for example, Furusaka et al., and Penrose, 1997). This cluster – diffu-
1986; Osamura, 1988), but typically data sion – coagulation mechanism was origi-
extending over only 1.5 decades in time are nally proposed to explain the correspond-
available. A general problem is that some ing small values of the exponent x seen in
misfit between matrix and precipitates is Monte Carlo simulations (Bortz et al.,
present in many real alloys. Moreover, va- 1974; Marro et al., 1975). However, more
cancy concentrations may be larger than accurate simulations (Huse, 1986; Amar et
the equilibrium value after the quench and al., 1988) are rather consistent with a grad-
then decrease gradually. Both effects could ual approach to the asymptotic Lifshitz –
affect the exponent in real alloys. Katano Slyozov – Wagner (1961) law without an
and Iizumi (1984), Furusaka et al. (1988) intermediate regime characterized by a
and Forouhi and De Fontaine (1987) found well-defined different exponent x. In fact,
a regime of 1.5 decades in time where the data can usually be fitted to an equation
x = 1/6 whereas after a rather sharp cross- derived by Huse (1986):
over x = 1/3 at later times. Other experi-
ments (e.g., Morii et al., 1988) where the km–1 (t) = A + B t 1/3 (6-84 a)
km–1 (t) vs. t relationship is also measured while the original treatment of Lifshitz and
over about three decades in time find a Slyozov (1961) yielded
much smoother crossover from the initial
stages, where the log – log plot of km–1 (t) vs. km–3 (t) = A¢ + B¢ t (6-84 b)
6.3 Survey of Experimental Results 453
and SiC – AlN ceramics (Kuo et al., 1987). diffusion in fluid mixtures proceeds much
The crystallization behavior of bulk metal- faster than in solids, we must consider
lic glasses, such as Zr – Ti – Cu – Ni – Be, has quenches in a narrow region just below Tc ,
been found to be strongly influenced by de- in order to take advantage of critical slow-
composition (e.g., Wang et al., 1998 b). ing down. This implies that we must al-
Of course, spinodal decomposition is ways work in the non-mean-field critical
also expected to occur in various nonmetal- region (in the phase diagram in Fig. 6-13),
lic mixed crystals, e.g., oxides such as the i.e., nonlinear effects are very strong, and
TiO2 – SnO2 system (Flevaris, 1987; Taka- so it is not expected that the linearized the-
hashi et al., 1988) and semiconducting ory of spinodal decomposition will account
GaInAsP epitaxial layers (Cherns et al., for these systems. In addition, very early
1988; Mcdevitt et al., 1992). Often these stages are not observable because the un-
situations are difficult because of lower mixing proceeds too fast. This is best seen
crystal symmetry (the SnO2 – TiO2 system when working with rescaled variables
is tetragonal) and strong elastic lattice mis- (Chou and Goldburg, 1979, 1981), defining
fit effects (Flevaris, 1987).
qm (t ) ∫ km x (t = 0)
and
6.3.3 Fluid Mixtures
t = t D (t = 0)/ x 2 (t = 0)
Binary fluids containing small mole-
cules such as lutidine – water (Goldburg et where x (t = 0) and D (t = 0) are the correla-
al., 1978; Goldburg, 1981, 1983; Chou and tion length and the interdiffusion constant
Goldburg, 1979, 1981) and isobutyric in the initial state, respectively. Early times
acid – water (Wong and Knobler, 1978, then mean a t of the order of unity. Fig. 6-
1979, 1981; Chou and Goldburg, 1979, 28 shows that only scaled times t Ê6 are
1981) are classic systems where nonlinear accessible, whereas earlier times are ac-
spinodal decomposition was observed via cessible for fluid polymer mixtures (see
light scattering techniques. Since the inter- Sec. 6.3.4). The growth rate exponent x
(qm–1
(t ) ~ t x ) crosses over from x near methanol and partially deuterated cyclo-
x ≈ 1/3 for t = 10 to x ≈ 1 at very late times, hexane, a perfect density matching at Tc is
as expected (Siggia, 1979) (see Eq. (6-70)). possible; see Houessou et al., (1985)) or
–1
The behavior qm (t ) ~ t –1/3, which for off- space experiments are performed, where
critical quenches where the structure corre- the gravitational effect can be reduced by a
sponds to well separated clusters is be- factor of 104. Beysens et al. (1988) showed
lieved to be the true asymptotic behavior that such experiments agree with the re-
for t Æ ∞ (Siggia, 1979), is attributed to a sults obtained from isodensity systems. By
cluster diffusion and coagulation mecha- such means not only can the accuracy of
nism (Binder and Stauffer, 1974; Siggia, data of the sort shown in Fig. 6-28 be sub-
1979). Note that for fluid droplets the stantially improved, but several decades
diffusion constants decrease in proportion can be added to a plot of qm vs. t (Fig. 6-
to their radii with increasing droplet size, 29). The full curve in this plot is the func-
due to Stokes’ law, and hence the cluster tion (Eq. (6-73)) proposed by Furukawa
diffusion – coagulation mechanism never (1984), which reduces to Eq. (6-70) for
becomes negligible in comparison with large t , i.e., the result of Siggia (1979).
the Lifshitz – Slyozov (1961) evaporation – More recently, a growth of the mean drop-
condensation mechanism. let size with the power-law exponent 1/3
Fig. 6-28 shows, however, that the cross- could be followed over more than seven
over to the asymptotic power law is grad- decades in a microgravity experiment (Per-
ual. This smooth behavior is approximately rot et al., 1994).
described by the nonlinear theory of spino- Interesting extensions involve spinodal
dal decomposition of Kawasaki and Ohta decomposition in fluid mixtures under
(1978). However, this theory can give only weak steady-state shear (Chan et al., 1988;
a poor account of the full structure func- Krall et al., 1992; Lauger et al., 1995;
tion. This is not surprising, since it is only Hashimoto et al., 1995; Hobbie et al.,
expected to be accurate for t Á10 (see Sec. 1996) or periodically applied shear (Bey-
6.2.6). Again, the problem of calculating sens and Perrot, 1984; Joshua et al., 1985),
the scaled structure function S̃ (ᒗ) (Eq. (6- as well as in strongly stirred mixtures
67)) arises and is as difficult as in the case where a turbulent suppression of spinodal
of solids (Furukawa, 1985 a). decomposition occurs (Pine et al., 1984;
A problem for binary fluids is the effect Chan et al., 1987; Easwar, 1992). Many of
of gravity. Since the two coexisting phases these observations can be explained on the
usually differ in density, one phase must go basis of the theoretical considerations of
to the top and the other to the bottom of the Onuki (1984, 1986 a, 1989 c).
container. This effect (for a more detailed Finally, we draw attention to phase sep-
discussion see Beysens et al., (1988)) im- aration phenomena in more exotic fluids
plies a smearing of the critical region for such as surfactant micellar solutions (Wil-
the unmixing critical point of a sample of coxon et al., 1988, 1995), polymer solu-
finite height, and also affects the late stages tions near their Q -point, e.g., polystyrene
of phase separation (where even hydrody- in methyl acetate (Chu, 1988), gel – gel
namic instabilities may set in; see Chan transitions (such a transition occurs, for ex-
and Goldburg (1987)). The effect of gravity, ample, in N-isopropylacrylamide gel with
however, can be strongly reduced if iso- water as a solvent, where this system phase
density systems are used (using mixtures of separates into two states with different sol-
6.3 Survey of Experimental Results 457
Figure 6-29. Behavior of the scaled wavevector qm (t ) versus scaled time t . Full curve is from Furukawa
(1985), data points refer to mixtures of methanol and cyclohexane in space-flight experiments (F) or to isoden-
sity mixtures of methanol and partially deuterated cyclohexane for various quench depths, as indicated. The
late-time behavior, where qm (t ) ~ t –1 is observed, is magnified in the inset. From Beysens et al. (1988).
vent concentrations and hence a different spinodal curve can be defined; (iii) chang-
swelling ratio of the gel; see Hirotsu and ing the molecular weights while other
Kaneki (1988)), and lipid membranes with parameters (in particular, intermolecular
dissolved protein. forces!) remain constant allows a more
stringent test of theories than for other
systems. For example, it has been possible
6.3.4 Polymer Mixtures
recently to obtain by laser scanning confo-
As pointed out in Secs. 6.2.4 and 6.2.5, cal microscopy (Jinnai et al., 1997) three-
mixtures of long flexible macromolecules dimensional images of phase-separated
are particularly well suited model systems polymer blends that look very similar to
for the study of phase separation kinetics: pictures obtained from theory (Fig. 6.30).
(i) owing to the mutual entanglement of the Therefore, it is gratifying that quantita-
random polymer coils, their interdiffusion tive agreement with the linear theory of
is very slow for high molecular weights, so spinodal decomposition was obtained in
the early stages can be studied; (ii) polymer careful experiments with various poly-
mixtures exhibit a well-defined mean-field mer blends, e.g., polystyrene – poly(vinyl
critical regime, where the linear theory of methyl ether) (Okada and Han, 1986; Han
spinodal decomposition should hold, and a et al., 1988; Sato and Han, 1988), polybu-
458 6 Spinodal Decomposition
6.4 Extensions
Basic aspects of the theory of spinodal
decomposition were treated in Sec. 6.2.
Only the generic phase diagram (Figs. 6-
3 b, 6-13, and 6-21) of a binary mixture
with a miscibility gap has been considered.
In this section, we briefly mention the re-
lated phenomenon of ordering kinetics
(Fig. 6-3 a) and also discuss cases where
formation of order and unmixing compete.
The effects of certain complications (im-
purities, effects of finite size and free sur-
faces) are briefly assessed.
complex variable), and also crystallo- from some reference value that fixes the
graphic order – disorder phase transitions in constant D 0 , and m 3 and m 4 are the chemi-
alloys and in adsorbed layers at surfaces cal potentials of 3He and 4He, respectively.
are known to exhibit such tricritical points. For fixed c, T, and D = m 3 – m 4 + D 0 the
In alloys, the first discussions of phase sep- static equilibrium follows from seeking the
aration in tricritical systems come from Al- minimum of
len and Cahn (1976, 1979 a, b) in the con- (6-86)
1 1 1 1
text of Fe – Al alloys. An elegant extension F = r˜ |y |2 + u˜ |y |4 + v |y |6 + c n D2
of the linearized theory of spinodal decom- 2 4 6 2
position to tricritical 3He – 4He mixtures where r̃ = r0 + 2 g Dc n , ũ = u – 2 c n g 2, and
was given by Hohenberg and Nelson c = c n [D – g | y |2 ] was eliminated from the
(1979). Nonlinear effects have mostly been equation. For ũ > 0 there is a second-order
studied for the model with simple relaxa- transition from normal fluid to superfluid
tional dynamics of the order parameter at r̃ = 0, whereas for ũ < 0 the transition is
(Dee et al., 1981; San Miguel et al., 1981). of first order. The critical point occurs for
Here we briefly mention the case of ũ = 0.
3
He – 4He mixtures, since the most detailed Whereas for the binary mixture a single
experiments exist for these systems (Hoffer Langevin equation for the concentration
et al., 1980; Sinha and Hoffer, 1981; Benda field c (x, t) and y (x, t) had to be derived
et al. 1981, 1982; Alpern et al., 1982). (Eq. (6-15)), here we need two equations.
3
He – 4He mixtures have no practical appli- In fact, the entropy density S (x, t) must
cations in materials science, but we men- also be added (Hohenberg and Nelson,
tion them here as a model system for the 1979), and thus the calculation becomes
study of phase separation near a tricriti- complicated. However, since y is not a con-
cal point. The concepts developed (and served quantity, it relaxes much more rap-
checked experimentally) for this model idly, hence the assumption is made that it al-
system can be carried over to more com- ways adjusts to the local equilibrium corre-
plex materials such as ternary mixtures, sponding to the local concentration c (x, t).
magnetic alloys, etc. Instead of Eqs. (6-4) Instead of the simple exponential function
and (6-8), the expression for the free en- in Eq. (6-23), additional oscillating terms
ergy functional is now due to the “second-sound” mode are found:
DᏲ
= ∫ dx ⎧⎨ r0 |y |2 + u |y |4
2
1 1 S (k, t) = a1 e2 R (k) t + a2 e[R (k) – D2 k ]t
(6-87)
kB T ⎩2 4 ¥ cos (u2 k t) + a3 e
2
– 2 D2 k t
cos (2 u2 k t) + a0
1 1
+ v |y |6 + r 2 | ∇y ( x )|2 where a0 , a1 , a2 , and a3 are constants, u2 is
6 2
1 −1 2 the second-sound velocity and D2 its damp-
+ c n c + g c |y |2 − (m 3 − m 4 + D0 ) c ing coefficient. In the spinodal region, R (k)
2
is positive, and not too close to the tricriti-
+ l02 | ∇c ( x )|2 ⎫⎬
1
(6-85) cal point for small k, R (k) – D2 k2 should
2 ⎭
also be positive. Therefore in addition to
where r0 , u, v, c n–1, l0 , and g are pheno- the first term on the right-hand side of Eq.
menological coefficients that should de- (6-87), which is the analog of Eq. (6-23),
pend only on temperature and not on the lo- there is another exponentially growing
cal 3He concentration c, which is measured term oscillating in time. This would imply
462 6 Spinodal Decomposition
Figure 6-34. “Snapshot pictures” of the computer simulation of the ordering process of a 120 ¥ 120 square lat-
tice model of an alloy with repulsive interactions between A – B pairs on nearest- and next-nearest-neighbor
sites, e nn = e nnn . A quench from a disordered configuration at infinite temperature to kB T/e nn = 1.33 is performed
for c– = 0.5 (a second-order transition to the (2 ¥ 1) structure occurs in equilibrium at kB Tc /e nn ≈ 2.1). Time evo-
lution occurs via random nearest-neighbor A – B exchanges (i.e., the model of Kawasaki (1972)). Times shown
refer to 10 MCS after the quench (upper part), 200 MCS (middle) and 1700 MCS (bottom). Only B atoms are
shown (using four different symbols to indicate whether a B atom belongs to a domain of type 1, 2, 3, or 4, see
Fig. 6-35 a), A atoms are not shown. From Sadiq and Binder (1984).
6.4 Extensions 465
Figure 6-35. (a) Four types of domains (1, 2, 3, and 4) in the (2 ¥ 1) structure of a binary alloy AB on the square
lattice (B atoms are indicated by black circles and A atoms by white circles). (b) Structure factor S (q, t) of
superlattice Bragg scattering, at times t after the quench. Since the lattice spacing is set to unity, the Bragg po-
sitions of the (2 ¥1) structure are (±p, 0) and (0, ± p). Only the variation with qx near qx = p is shown. Times are
measured in units of Monte Carlo steps per site. Data are for a computer simulation of a quench from T = • to
kB T/e nn = 1.33 as in Fig. 6-34. (c) Scaling plot of the data shown in (b), with the structure factor S (q, t) being
rescaled with the maximum intensity S (p, t) and the relative distance q/p – 1 from the Bragg position being re-
scaled with the half-widths s (t) of the peaks in (b). From Sadiq and Binder (1984).
466 6 Spinodal Decomposition
undercooling of the transition. Here we are find a law km–1 (t) ~ t 1/4 (Noda, 1988), for
more interested in the “spinodal ordering” Ni3Mn a crossover from km–1 (t) ~ t 1/4 at
as occurs in Cu3Pd (Takeda et al., 1987), short times to km–1 (t) ~ t 1/2 at later times is
Ni3Mn (Katano and Iizumi, 1988), Cu3Au found (Katano and Iizumi, 1988), and for
(Hashimoto et al., 1978; Nishihara et al., AuCu3 a law km–1 (t) ~ t 1/2 (i.e., Eq. (6-95))
1982; Noda et al., 1984), and K2Ba(NO2 )4 is established (Noda et al., 1984). Whereas
(Noda, 1988). the t 1/2 law is expected from various theo-
Fig. 6-36 shows examples of growing ries (Lifshitz, 1962; Allen and Cahn,
Bragg peaks, i.e. the experimental counter- 1979 b; Ohta et al., 1982), a t 1/4 law occur-
parts to simulation data such as shown ring over a transient period of time can per-
in Fig. 6-35. Whereas for K2Ba(NO2 )4 we haps be attributed to the “softness” of the
domain walls between the growing ordered
regions (Mouritsen, 1986; Mouritsen and
Praestgaard, 1988). An interesting obser-
vation of Lifshitz – Slyozov type coarsen-
ing of the structure factor is also reported
for colloid crystallization (Schätzel and
Ackerson, 1993).
In the two dimensional case, Wu et al.
(1989) measured a growth exponent x =
0.28 ± 0.05 for O on W (110). Since the
p (2 ¥1) phase is believed to have an eight-
fold ground-state degeneracy in this sys-
tem, whereas the theories mentioned above
refer to a two-fold degenerate ordering
only, the interpretation of this result is not
obvious. We note, however, that reasonable
scaling of the structure factor is observed
(Fig. 6-37). For the system silver on Ge
(111), a growth exponent x = 1/2 is found
(Henzler and Busch, 1990) and also good
scaling of the structure factor is seen. In
this system, on the other hand, a crossover
to slower growth occurs at later times; the
reason for this behavior is not clear – per-
haps it is due to pinning of domain walls at
defects (see Sec. 6.4.4). At this point, we
note that a similar slowing down of the
growth where km–1 (t) basically stops grow-
ing further, has also been seen in Ni3Mn
(Katano and Iizumi, 1988) and in the phase
Figure 6-36. (a) Time evolution of the 211 super-
separation of mixtures of flexible and
lattice peak of Ni3Mn annealed at 470 °C for times
up to 34 h. From Katano and Iizumi (1988). (b) Time semi-rigid polymers (Hasegawa et al.,
evolution of the (–12 , x , –52 ) superlattice peak of 1988). In neither of these systems is the
K2Ba(NO2 )4 annealed at 190 K. From Noda (1988). very slow growth at late stages understood.
6.4 Extensions 467
(6-95)) will hold until some characteristic A simulation of the early stages of spino-
length lc is reached, which depends on the dal decomposition in such an ABV model
concentration ci of the quenched impurities (Yaldram and Binder, 1991) shows that the
(lc ~ ci–1/d in d dimensions). At times t sig- general features of the structure factor
nificantly exceeding the time given by the S (k, t) are almost the same as those of the
condition km–1 (t) lc ≈ 1, a logarithmic growth direct-exchange AB model (Fig. 6-5), and
law (Villain, 1985) is expected: the two models can be approximately
mapped onto each other by adjusting the
km–1 (t) ~ ln t (6-97)
time scales. In real systems, however, the
A similar crossover in the relaxation from a behavior may be more complicated; the va-
fast to a slow growth law may also be cancy concentration does not need to re-
caused by interface pinning at extended de- main constant, and it may be that many va-
fects (e.g., dislocations, grain boundaries) cancies are created during the quench
rather than point defects. which later are annealed out by migration
to surfaces or recombination with intersti-
(ii) Vacancies tials. If this happens, the effective mobility
In solid mixtures, vacancies V are cru- M would itself depend on the time t after
cial for a microscopic description of inter- the quench. Also the vacancy concentra-
diffusion, which occurs via a vacancy tions in equilibrium may be different in A-
mechanism rather than by direct A – B ex- rich and B-rich domains, or may become
change (Flynn, 1972; Manning, 1986). Mi- preferentially enriched in domain boundar-
croscopically we therefore need concentra- ies. This did not occur in the simulation of
tions ciA (t), ciB (t), and ciV (t) for A atoms, B Yaldram and Binder (1991), since there the
atoms, and vacancies as dynamic variables, static properties were assumed with perfect
respectively, rather than the single concen- symmetry between A and B and there was
tration variable ci or concentration field no energy parameter associated with the
c (x, t) (Eq. (6-1)). Since ciV (t) 1, it may vacancies, but this model is still a gross
be possible to reduce the problem to Eq. over-simplification of reality. Fratzl and
(6-15), where the mobility M then needs to Penrose (1994, 1997) found that vacancies
be related to the jump rates GA and GB of may speed up the coarsening by changing
A and B atoms to vacant sites (and the va- the mechanism. Note that for very low va-
cancy concentration). However, owing to cancy content but very late stages a faster
correlation effects in the vacancy motion, growth law (x = 1/2) due to cluster – cluster
this is a difficult problem even in the non- aggregation has been suggested (Mukher-
interacting case (Kehr et al., 1989). There jee and Cooper, 1998). These problems
Eqs. (6-6) and (6-14) would yield an inter- may confuse the proper interpretation of
diffusion coefficient Dint = M/[c (1 – c)], real experiments.
i.e., M can be found if Dint can be simply
related to GA and GB and the average va- (iii) Electrical Resistivity of Metallic
cancy concentration c– V. Although M obvi- Alloys Undergoing Phase Changes
ously is proportional to c– V for small c– V, In metallic alloys the quasi-free elec-
computer simulations (Kehr et al., 1989) trons responsible for electrical conduction
show that the simple relationships pro- are scattered from the atomic disorder. In
posed in the literature to relate M to GA and the framework of the Born approximation,
GB are inaccurate. treating the scattering as elastic, the excess
470 6 Spinodal Decomposition
terminate the coarsening process and thus understand the crossover from these early
freeze in an inhomogeneous pattern. stages to the late stage of phase separation
where the structure factor develops to-
wards a scaling limit. It is not clear under
which conditions power-law behavior
6.5 Discussion km (t) ~ t – x can be observed at intermediate
stages, and what the appropriate interpreta-
The understanding of the dynamics of
tion of the exponent x is.
phase changes of materials has advanced
(iv) In the late stages the structure factor
considerably in recent years; the theoreti-
obeys the scaling behavior first suggested
cal results now present a fairly clear picture
by Binder and Stauffer (1974, 1976 b),
of at least some basic questions, and also
quantitatively reliable experiments are S (k, t) ~ [km (t)] –d S̃ {k/km (t)}
available for many systems. This chapter
While the scaling function S̃ (ᒗ) proposed
has emphasized the theoretical aspects; its
by Furukawa (1984) seems to account well
main conclusions can be summarized as
for the general shape of experimental data
follows:
on solid or polymer mixtures, several de-
(i) The linearized theory of spinodal de- tails are not reproduced correctly. Good ev-
composition holds only for mean-field type idence for Porod’s law, S̃ (ᒗ) ~ ᒗ– (d +1) for
systems (with long ranges of interactions) ᒗ 1, has been provided. The width of S̃ (ᒗ)
or in systems that are equivalent to them, increases when the volume fraction f of
such as polymer mixtures. precipitate phase decreases from that of the
(ii) Whereas a spinodal curve is well de- critical mixture (Fratzl and Lebowitz,
fined in the mean-field limit or for polymer 1989). There we find, as a function of tem-
mixtures in the limit of chain length N Æ∞, perature and f , both morphologies, inter-
in all other cases the spinodal is an ill-de- connected domains and well separated
fined concept. This implies that the transi- droplets. Such a transition can also be ob-
tion from nucleation to spinodal decompo- served as a function of time in computer
sition is gradual. For polymer mixtures, the simulations (Hayward et al., 1987; Lironis
width of this transition region may be nar- et al., 1989) and experiments on polymer
row, as estimations from Ginzburg criteria mixtures (Hasegawa et al., 1988) (see Fig.
show. Experiments are still needed to 6-38).
check the latter point, whereas the gradual (v) There is now agreement that in solid
transition from nucleation to spinodal de- mixtures the behavior at late times is given
composition in metallic alloys over a broad by the Lifshitz – Slyozov law, km–1 (t) ~ t 1/3,
temperature region has been established even in the percolative regime (Amar et al.,
experimentally. 1988). This has recently also been verified
(iii) Nonlinear spinodal decomposition in a solid – liquid system that satisfies all
during the early stages can be successfully assumptions of the LSW theory (Alkemper
described qualitatively by the Langer – et al., 1999). For critical quenches of fluid
Baron – Miller theory (for solid mixtures) mixtures at late times, Siggia’s (1979) re-
and the Kawasaki – Ohta theory (for fluid sult km–1 (t) ~ t holds, but further research is
mixtures). Recent experiments, however, needed to achieve an understanding of the
show that these theories are not quantita- exponents for the growth of ordered struc-
tively accurate. More work is required to tures in cases with a higher degeneracy of
472 6 Spinodal Decomposition
6.6 Acknowledgements
This work has profited considerably
from a longstanding and stimulating inter-
action of K. B. with D. Stauffer and D. W.
Heermann. This author thanks them and
also other co-workers, namely C. Billotet,
H.-O. Carmesin, H. L. Frisch, J. D. Gun-
ton, S. Hayward, J. Jäckle, M. H. Kalos, K.
Kaski, K. W. Kehr, J. L. Lebowitz, G. Liro-
nis, A. Milchev, P. Mirold, S. Reulein, A.
Sariban, and K. Yaldram, for fruitful col-
laboration.
The other author (P. F.) is particularly in-
debted to O. Penrose and J. L. Lebowitz for
a longstanding and fruitful collaboration.
Figure 6-40. “Snapshot pictures” of the two-dimen- Furthermore he is indebted to H. Gupta, C.
sional nearest-neighbor Ising model of a phase-sep- A. Laberge, F. Langmayr, P. Nielaba, O.
arating mixture evolving after a critical quench to Paris, G. Vogl, and R. Weinkamer for fruit-
T = 0.5 Tc at times t = 5000 MCS (a), t = 105 MCS (b) ful interactions.
and t = 9.8 ¥ 105 MCS (c). B atoms are shown
in black, A atoms are not shown. Data are for a
512 ¥ 512 lattice. From Amar et al. (1988).
474 6 Spinodal Decomposition
6.7 References Binder, K. (1987 a), Rep. Prog. Phys. 50, 783.
Binder, K. (1987 b), Colloid Polym. Sci. 265, 273.
Binder, K. (1989), in: Alloy Phase Stability: Gonis, A.,
Abinandanan T. A., Johnson W. C. (1993), Acta Stocks, L. M. (Eds.). Dordrecht: Kluwer, p. 232.
Metall. Mater. 41, 17; 27. Binder, K. (1998), J. Non-Equilib. Thermodyn. 23,
Abinandanan T. A., Johnson W. C. (1995), Mat. Sci. 114.
Eng. B 32, 169. Binder, K. (2001), in: Phase Transformations in Ma-
Acuña, R. J., Craievich, A. F. (1979), J. Non-Cryst. terials: Kostorz, G. (Ed.). Weinheim: Wiley-VCH,
Solids 34, 13. Chap. 4.
Akaiwa, N., Voorhees, P. W. (1994), Phys. Rev. E 49, Binder, K., Frisch, H. L. (1984), J. Chem. Phys. 81,
3860. 2126.
Akcasu, A. Z., Benmouna, M., Benoit, H (1986), Binder, K., Heermann, D. W. (1985), in: Scaling
Polymer 27, 1935. Phenomena in Disordered Systems: Pynn, R.,
Alexander, F. J., Chen, S., Grunau, D. W. (1993), Skjeltorp, T. (Eds.). New York: Plenum Press,
Phys. Rev. B 48, 634. p. 207.
Alkemper, J., Snyder, V. A., Akaiwa, N., Voorhees, Binder, K., Sillescu, H. (1989), in: Encyclopedia of
P. W. (1999), Phys. Rev. Lett. 82, 2725. Polymer Science, Vol. 2: Kroschwitz, J. (Ed.). New
Allen, S. M. (1977), Philos. Mag. 36, 181. York: Wiley, p. 297.
Allen, S. M, Cahn, J. W. (1976), Acta Metall. 24, 425. Binder, K., Stauffer, D. (1974), Phys. Rev. Lett. 33,
Allen, S. M., Cahn, J. W. (1979 a), Acta Metall. 27, 1006.
1017. Binder, K., Stauffer, D. (1976 a), Adv. Phys. 25, 343.
Allen, S. M., Cahn, J. W. (1979 b), Acta Metall. 27, Binder, K., Stauffer, D. (1976 b), Z. Phys. B 24, 407
1085. Binder, K., Billotet, C., Mirold, P. (1978), Z. Phys.
Alpern, P., Benda, Th., Leiderer, P. (1982), Phys. B 30, 183.
Rev. Lett. 49, 1267. Binder, K., Kalos, M. H., Lebowitz, J. L, Marro, J.
Als-Nielsen, J. (1976), in: Phase Transitions and (1979), Adv. Colloid Interface Sci. 10, 173.
Critical Phenomena, Vol. 5 a: Domb, C., Green, M. Binder, K., Frisch, H. L., Jäckle, J. (1986), J. Chem.
S. (Eds.). New York: Academic Press, p. 88. Phys. 85, 1505.
Amar, J. G., Sullivan, F. E., Mountain, R. D. (1988), Binder, K., Heermann, D. W., Milchev, A, Sadiq, A.
Phys. Rev. B 37, 196. (1987), in: Proceedings of Heidelberg Colloquium
Avrami, M. (1939), J. Chem. Phys. 7, 1103. on Glassy Dynamics: Van Hemmen, L., Morgen-
Bastea, S., Lebowitz, J.L. (1995), Phys. Rev. E52, 3821. stern, I. (Eds.). Berlin: Springer, p. 154.
Bates, F. S., Wiltzius, P. (1989), J. Chem. Phys. 91, Bortz, A. B., Kalos, M. H., Lebowitz, J. L., Zendejas,
3258. M. A. (1974), Phys. Rev. B 10, 535.
Benda, Th., Alpern, P., Leiderer, P. (1981), Physica Bray, A. J. (1990), Phys. Rev. B 41, 6274.
107 B, 157. Bray, A. J. (1994), Adv. Phys. 43, 357.
Benda, Th., Alpern, P., Leiderer, P. (1982), Phys. Bray, A. J. (1998), Phys. Rev. E 58, 1508.
Rev. B 26, 1450. Cahn, J. W. (1961), Acta Metall. 9, 795.
Beysens, D., Perrot, F. (1984), J. Phys. (Paris) Lett. Cahn, J. W. (1962), Acta Metall. 10, 179.
45, L-31. Cahn, J. W. (1965), J. Chem. Phys. 42, 93.
Beysens, D., Guenoun, P., Perrot, F. (1988), in: Dy- Cahn, J. W. (1966), Acta Metall. 14, 1685.
namic of Ordering Process in Condensed Matter: Cahn, J. W. (1968), Trans. Metall. Soc. AIME 242,
Komura, S., Furukawa, H. (Eds.). New York: 166.
Plenum Press, p. 373. Cahn, J. W., Hilliard, J. E. (1958), J. Chem. Phys. 28,
Billotet, C., Binder, K. (1979). Z. Phys. B 32, 195. 258.
Binder, K. (1977), Phys. Rev. B 15, 4425. Cahn, J. W., Hilliard, J. E. (1959), J. Chem. Phys. 31,
Binder, K. (1981), in Stochastic Nonlinear Systems 688.
in Physics, Chemistry and Biology: Arnold, L., Cahn, J. W., Larché, F. C. (1984), Acta Metall. 32,
Lefever, R. (Eds.). Berlin: Springer, p. 62. 1915.
Binder, K. (1983), J. Chem. Phys. 79, 6387. Carmesin, H.-O., Heermann, D. W., Binder, K.
Binder, K. (1984 a), in: Condensed Matter Research (1986), Z. Phys. B 65, 89.
Using Neutrons: Lovesey, S. W., Scherm, R. Chakrabarti, A., Toral, R., Gunton, J. D. (1993),
(Eds.). New York: Plenum Press, p. 1. Phys. Rev. E 47, 3025.
Binder, K. (1984 b), Phys. Rev., A 29, 341. Chan, S. K. (1977), J. Chem. Phys. 67, 5755.
Binder, K. (1984 c), in: Kinetics of Aggregation and Chan, C. K., Goldburg, W. I. (1987), Phys. Rev. Lett.
Gelation: Family, F., Landau, D. P. (Eds.). Amster- 48, 674.
dam: Elsevier, p. 209. Chan, C. K., Goldburg, W. I., Maher, J. V. (1987),
Binder, K. (1986), in: Festkörperprobleme – Ad- Phys. Rev. A 35, 1756.
vances in Solid State Physics, Vol. 26: Grosse, P. Chan, C. K., Perrot, F., Beysens, D. (1988), Phys.
(Ed.) Braunschweig: Vieweg, p. 133. Rev. Lett. 61, 412.
6.7 References 475
Che, D. Z., Spooner, S., Hoyt, J. J. (1997), Acta Falicov, A., Berker, A. N. (1995), Phys. Rev. Lett. 74,
Mater. 45, 1167. 426.
Chen, H., Rolat, S., Epperson, J. E. (1988), in: Dy- Flevaris, N. K. (1987). J. Am. Ceram. Soc. 70, 301.
namics of Ordering Processes in Condensed Mat- Flory, P. J. (1953), Principles of Polymer Chemistry.
ter: Komura, S., Furukawa, H. (Eds.). New York: Ithaka: Cornell University Press.
Plenum Press, p. 245. Flynn, C. P. (1972), Point Defects and Diffusion. Ox-
Cherns, D., Greene, P. D., Hainsworth, A., Preston, ford: Clarendon Press.
A. R. (1988), in: Microscopy of Semiconducting Forouhi, A. R., De Fontaine, D. (1987), Acta Metall.
Materials. Bristol: Institute of Physics, p. 83. 35, 1863.
Cho, J. H., Ardell, A. J. (1997), Acta Mater. 45, 1393. Fratzl, P. (1991), J. Appl. Cryst. 24, 593.
Chou, Y. C., Goldburg, W. I. (1979), Phys. Rev. A 20, Fratzl. P., Blaschko, O. (1988), in: Dynamics of Or-
2105. dering Processes in Condensed Matter: Komura,
Chou, Y. C., Goldburg, W. I. (1981), Phys. Rev. A 23, S., Furukawa, H. (Eds.). New York: Plenum Press,
858. p. 223.
Chu, B. (1988), in: Dynamics of Ordering Processes Fratzl, P., Lebowitz, J. L. (1989), Acta Metall. 12,
in Condensed Matter: Komura, S., Furukawa, H. 3245.
(Eds.). New York: Plenum Press, p. 463. Fratzl, P., Penrose, O. (1994). Phys. Rev. B 50, 3477.
Conley, J. G., Fine, M. E., Weertmann, J. R. (1989), Fratzl, P., Penrose, O. (1995). Acta Metall. Mater. 43,
Acta Metall. 37, 1251. 2921.
Cook, H. E. (1970), Acta Metall. 18, 297. Fratzl, P., Penrose, O. (1996). Acta Metall. Mater. 44,
Cook, H. E., De Fontaine, D., Hilliard, J. E. (1969), 3227.
Acta Metall. 17, 765. Fratzl, P., Penrose, O. (1997). Phys. Rev. B 55,
Craievich, A. F., Olivieri, J. R. (1981), J. Appl. Crys- R 6101.
tallogr. 14, 444. Fratzl, P., Lebowitz, J. L., Marro, J., Kalos, M. H.
Craievich, A. F., Sanchez, J. M. (1981), Phys. Rev. (1983), Acta. Metall. 31, 1849.
Lett. 47, 1308. Fratzl, P., Lebowitz, J. L., Penrose, O., Amar, J.
Dee, G., Gunton, J. D., Kawasaki, K. (1981), Prog. (1991), Phys. Rev. B 44, 4794.
Theor. Phys. 65, 365. Fratzl, P., Penrose, O., Lebowitz, J. L. (1999), J. Stat.
De Fontaine, D. (1979), in: Solid State Physics, Vol. Phys. 95, 1429.
34: Ehrenreich, H., Seitz, F., Thurnbull, D. (Eds.). Fredrickson, G. H., Binder, K. (1989), J. Chem.
London: Academic Press, p. 73. Phys. 91, 7265.
De Fontaine, D., Cook, H. E. (1971), in: Critical Furukawa, H. (1978), Prog. Theor. Phys. 59, 1072.
Phenomena in Alloys, Magnets and Super Con- Furukawa, H. (1984), Physica 123 A, 497.
ducters: Miller, R. E., Ascher, E., Jaffee, I. (Eds.). Furukawa, H. (1985 a), Adv. Phys. 34, 703.
New York: McGraw-Hill, p. 257. Furukawa, H. (1985 b), Prog. Theor. Phys. 74, 174.
De Fontaine, D., Kikuchi, R. (1978), NBS Spec. Publ. Furukawa, H. (1985 c), Phys. Rev. A31, 1103.
No. 496, 976. Furukawa, H. (1988), in: Dynamics of Ordering
De Gennes, P. G. (1979), Scaling Concepts in Poly- Processes in Condensed Matter: Komura, S., Fu-
mer Physics. Ithaca: Cornell University Press. rukawa, H. (Eds.). New York: Plenum Press,
De Gennes, P. G. (1980), J. Chem. Phys. 72, 4756. p. 35.
De Gennes, P. G. (1984), Phys. Chem. 88, 6469. Furusaka, M., Ishikawa, Y., Mera, M. (1985), Phys.
De Groot, S. R., Mazur, P. (1962), Non-Equilibrium Rev. Lett. 54, 2611.
Thermodynamics. Amsterdam: North-Holland. Furusaka, M., Ishikawa, Y., Yamaguchi, S., Fujino,
Descamps, M., Caucheteux, C. (1988), in: Dynamics Y. (1986), J. Phys. Soc. Jpn. 55, 2253.
of Ordering Processes in Condensed Matter: Ko- Furusaka, M., Fujikawa, S., Sakauchi, M, Hirano, K.
mura, S., Furukawa, H. (Eds.). New York: Plenum (1988), in: Dynamics of Ordering Processes in
Press, p. 257. Condensed Matter: Komura, S., Furukawa, H.
Dietrich, S. (1988), in: Phase Transitions and Criti- (Eds.). New York: Plenum Press, p. 281.
cal Phenomena, Vol. 12: Domb, C., Lebowitz, S. Gaulin, B. D., Spooner, S., Morii, Y. (1987), Phys.
L. (Eds.). New York: Academic Press, p. 1. Rev. Lett. 59, 668.
Doi, M., Edwards, S. F. (1986), The Theory of Poly- Gawlinksi, E. T., Gunton, J. D., Vinãls, J. (1989),
mer Dynamics. Oxford: Clarendon Press. Phys. Rev. B 39, 7266.
Easwar, N. (1992), Phys. Rev. Lett. 68, 186. Gerold, V., Kostorz, G. (1978), J. Appl. Crystallogr.
Edwards, J. T., Hillel, A. J. (1977), Philos. Mag. 35, 11, 376.
1231. Ginzburg, V. L. (1960), Sov. Phys. Solid State 2,
Eguchi, T., Oki, K., Matsumura, S. (1984), Mater. 1824.
Res. Soc. Symp. Proc. 21, 589. Glatter, O., Kratky, O. (1982), Small Angle X-Ray
Fährmann, M., Fratzl, P., Paris, O., Fährmann, E., Scattering. London: Academic Press.
Johnson, W. C. (1995), Acta Metall. Mater. 43, Glotzer, S. C., Di Marzio, E. A., Muthukumar, M.
1007. (1995), Phys. Rev. Lett. 74, 2034.
476 6 Spinodal Decomposition
Goldburg, W. I. (1981), in: Scattering Technique Hashimoto, T. (1993), in: Materials Science and
Applied to Supramolecular and Nonequilibrium Technology, Vol. 12: Cahn, R. W., Haasen, P.,
Systems: Chen, S. H., Chu, B., Nossal, R. (Eds.). Kramer, E. J. (Eds.). Weinheim: VCH, p. 251.
New York: Plenum Press, p. 383. Hashimoto, T., Izumitani, T. (1985), J. Chem. Phys.
Goldburg, W. I. (1983), in: Light Scattering Near the 83, 3694.
Phase Transition: Cummins, H. Z., Levanyuk, A. Hashimoto, T., Nishimura, K., Takeuchi, Y. (1978),
P. (Eds.). Amsterdam: North-Holland, p. 531. J. Phys. Soc. Jpn. 45, 1127.
Goldburg, W. I. (1988), Dynamics of Ordering Hashimoto, T., Itakuri, M., Shimidzu, N. (1986),
Processes in Condensed Matter: Komura, S., J. Chem. Phys. 85, 6773.
Furukawa, H. (Eds.). New York: Plenum Press, Hashimoto, T., Matsuzaka, K., Moses, E., Onuki, A.
p. 361. (1995). Phys. Rev. Lett. 74, 126.
Goldburg, W. I., Shaw, C. H., Huang, J. S., Pilant, M. Hayward, S., Heermann, D. W., Binder, K. (1987),
S. (1978), J. Chem. Phys. 68, 484. J. Stat. Phys. 49, 1053.
Goldburg, W. I., Aliev, F., Wu, X. L. (1995), Physica Heermann, D. W. (1984 a), Phys. Rev. Lett. 52, 1126.
A 213, 61. Heermann, D. W. (1984 b), Z. Phys. B 55, 309.
Gorentsveig, V. I., Fratzl, P., Lebowitz, J. L. (1997), Heermann, D. W. (1985), Z. Phys. B 61, 311.
Phys. Rev. B 55, 2912. Heermann, D. W., Klein, W. (1983a), Phys. Rev. Lett.
Grant, M., San Miguel, M., Vinals, S., Gunton, J. D. 50, 1962.
(1985), Phys. Rev. B 31, 302. Heermann, D. W., Klein, W. (1983 b), Phys. Rev.
Grest, G. S., Srolovitz, D. J. (1985), Phys. Rev. B 32, Lett. B 27, 1732.
3014. Heermann, D. W., Klein, W., Stauffer, D. (1982),
Griffiths, R. B. (1970), Phys. Rev. Lett. 24, 715. Phys. Rev. Lett. 49, 1262.
Guenoun, P., Gastand, R., Perrot, F., Beysens, D. Hennion, M., Ronzaud, D., Guyot, P. (1982), Acta
(1987), Phys. Rev. A 36, 4876. Metall. 30, 599.
Gunton, J. D., San Miguel, M., Sahni, P. S. (1983), Henzler, M., Busch, H. (1990), Phys. Rev. B 41, 4891.
in: Phase Transition and Critical Phenomena, Vol. Hikmet, R. M., Callister, S., Keller, A. (1988), Poly-
8: Domb, C., Lebowitz, J. L. (Eds.). London: Aca- mer 29, 1378.
demic Press, p. 267. Hillel, A. J., Edwards, J. T., Wilkes, P. (1975), Phi-
Gunton, J. D., Gawlinski, E. T., Kaski, K. (1988), in: los. Mag. 32, 189.
Dynamics of Ordering Processes in Condensed Hirotsu, S., Kaneki, A. (1988), in: Dynamics of Or-
Matter: Komura, S., Furukawa, H. (Eds.). New dering Processes in Condensed Matter: Komura,
York: Plenum Press, p. 101. S., Furukawa, H. (Eds.). New York: Plenum Press,
Gupta, H., Weinkamer, R., Fratzl, P., Lebowitz, J. L. p. 481.
(2001), Acta Mater. 49, 53. Hobbie, E. K., Kim, S. H., Han, C. C. (1996), Phys.
Guyot, P., Simon, J. P. (1982), in: Solid State Phase Rev. E 54, R 5909.
Transformations: Aaronsen, H. I., Laughlin, D. E., Hoffer, J. K., Campbell, L. J., Bartlett, R. J. (1980),
Schuka, R. F., Wayman, C. M. (Eds.). New York: Phys. Rev. Lett. 45, 912.
AIME, p. 325. Hohenberg, P. C., Halperin, B. I. (1977), Rev. Mod.
Guyot, P., Simon, J. P. (1988), in: Dynamics of Or- Phys. 49, 435.
dering Processes in Condensed Matter: Komura, Hohenberg, P. C., Nelson, D. (1979), Phys. Rev. B 20,
S., Furukawa, H. (Eds.). New York: Plenum Press, 2665.
p. 211. Hono, K., Babu, S. S., Hiraga, K., Okano, R., Saku-
Haasen, P., Gerold, V., Wagner, R., Ashby, M. F. rai, T. (1992), Acta Mater. 40, 3027.
(Eds.) (1984), Decomposition of Alloys: The Early Houessou, C., Guenoun, P., Gastaud, R., Perrot, F.,
Stages. Oxford: Pergamon Press. Beysens, D. (1985), Phys. Rev. A 32, 1818.
Han, C. C., Okada, M., Sato, T. (1988), in: Dynamics Houston, E. L., Cahn, J. W., Hilliard, J. E. (1966),
of Ordering Processes in Condensed Matter: Ko- Acta Metall. 14, 1685.
mura, S., Furukawa, H. (Eds.). New York: Plenum Huberman, B. A. (1976), J. Chem. Phys. 65, 2023.
Press, p. 433. Huse, D. A. (1986), Phys. Rev. B 34, 7845.
Hasegawa, H., Shiwaku, T., Nakai, A., Hashimoto, T. Huse, D. A. (1987), Phys. Rev. B 36, 5383.
(1988), in: Dynamics of Ordering Processes in Izumitani, T., Takanaka, M., Hashimoto, T. (1989),
Condensed Matter: Komura, S., Furukawa, H. J. Chem. Phys. 92, 3213.
(Eds.). New York: Plenum Press, p. 457. Jäckle, J., Pieroth, M. (1988), Z. Phys. B 72, 25.
Hashimoto, T. (1987), in: Current Topics in Polymer Jantzen, C. M. F., Herman, H. (1978), in: Phase Dia-
Science, Vol. 2: Ottenbrite, R. M., Itracki, L. A., grams: Material Science and Technology, Vol. 5:
Inoue, S. (Eds.). Munich: Hanser, p. 199. Alper, A. A. (Ed.). New York: Academic Press,
Hashimoto, T. (1988), in: Dynamics of Ordering p. 127.
Processes in Condensed Matter: Komura, S., Fu- Jinnai, H., Hashimoto, T., Lee, D., Chen, S.-H.
rukawa, H. (Eds.). New York: Plenum Press, (1997), Macromolecules 30, 130.
p. 421. Joanny, J. F. (1978), C. R. Acad. Sci. Ser. B 286, 89.
6.7 References 477
Johnson, W. A., Mehl, R. F. (1939), Trans. AIME Krall, A. H., Sengers, J. V., Hamano, K. (1992),
135, 416. Phys. Rev. Lett. 69, 1963.
Jones, R. A. L., Norton, L. J., Kramer, E. J., Bates, F. Kramer, E. J., Green, P., Palmstrom, C. (1984). Poly-
S., Wiltzius, P. (1991), Phys. Rev. Lett. 66, 1326. mer 25, 473.
Joshua, M., Goldberg, W. I., Onuki, A. (1985), Phys. Krausch, G. (1995). Mat. Sci. Eng. Rep. 14, 1.
Rev. Lett. 54, 1175. Kumaki, J., Hashimoto, T., Granick, S. (1996), Phys.
Jou, H. J., Leo, P. H., Lowengrub, J. S. (1997), J. Rev. Lett. 77, 1990
Comp. Phys. 131, 109. Kuo, S. Y., Viskar, A. V (1987), J. Am. Ceram. Soc.
Kadanoff, L. P. (1976), in: Phase Transitions and 70, C 125.
Critical Phenomena, Vol. 5 a: Domb, D., Green, Laberge, C., Fratzl, P., Lebowitz, J. L. (1995). Phys.
M. S. (Eds.). London: Academic Press, Ch. 1. Rev. Lett. 75, 4448.
Kampmann, R., Wagner, R. (1984), in: Decomposi- Laberge, C., Fratzl, P., Lebowitz, J. L. (1997). Acta
tion of Alloys: the Early Stages: Haasen, P., Ge- Mater. 45, 3946.
rald, V., Wagner, R., Ashby, M. F. (Eds.). Oxford: Lai, Z. W., Mazenko, G. F., Valls, O. T. (1988), in:
Pergamon Press, p. 91. Dynamics of Ordering Processes in Condensed
Kaski, K., Binder, K., Gunton, J. D. (1984), Phys. Matter: Komura, S., Furukawa, H. (Eds.). New
Rev. B 29, 3996. York: Plenum Press, p. 45.
Katano, S., Iizumi, M. (1984), Phys. Rev. Lett. 52, Langer, J. S., Schwartz, A. J. (1980), Phys. Rev. A 21,
835. 948.
Katano, S., Iizumi, M. (1988), in: Dynamics of Or- Langer, J. S., Baron, M., Miller, H. D. (1975), Phys.
dering Processes in Condensed Matter: Komura, Rev. A 11, 1417.
S., Furukawa, H. (Eds.). New York: Plenum Press, Langmayr, F., Fratzl, P., Vogl, G. (1992), Acta Met-
p. 321. all. Mater. 40, 3381.
Kawasaki, K. (1972), in: Phase Transitions and Crit- Larché, F. C., Cahn, J. W. (1982), Acta Metall. 30,
ical Phenomena. Vol. 2: Domb, C., Green, M. S. 1835.
(Eds.). New Academic Press, Ch. 11. Laslaz, G., Guyot, P., Kostorz, G. (1977), J. Phys.
Kawasaki, K., Ohta, T. (1978), Prog. Theor. Phys. (Paris) C 7, 406.
59, 362. Lauger, J., Laubner, C., Gronski, W. (1995). Phys.
Kawasaki, K, Imeada, T., Gunton, J. D. (1981), in: Rev. Lett. 75, 3576.
Perspectives in Statistical Physics: Raveché, H. J. Lawrie, I. D., Sarbach, S. (1984), in: Phase Transi-
(Ed.). Amsterdam: North -Holland, p. 203. tions and Critical Phenomena, Vol. 9: Domb, C.,
Kehr, K. W., Binder, K., Reulein, S. M. (1989), Phys. Lebowitz, J. L. (Eds.). New York: Academic Press,
Rev. B 39, 4891. p. 1.
Khachaturyan, A. G. (1966), Sov. Phys. Solid State 8, Lebowitz, J. L. Penrose, O. (1966), J. Math. Phys. 7,
2163. 98.
Khachaturyan, A. G. (1983), Theory of Structural Lebowitz, J. L. Marro, J., Kalos, M. H. (1982), Acta
Transformations in Solids. New York: Wiley. Metall. 30, 297.
Klein, W., Unger, C. (1983), Phys. Rev. B 28, 445. Lee, S. K. (1997), Mater. Sci. Eng. A 238, 1.
Koga, T., Kawasaki, K. (1991), Phys. Rev. A 44, R 817. Lee, S. K. (1998), Trans. JIM 39, 114.
Koga, T., Kawasaki, K. (1993), Physica A 196, 389. Lefever, R., Carati, D., Hassani, N. (1995), Phys.
Komura, S., Osamura, K., Fujii, H., Takeda, T. Rev. Lett. 75, 1674.
(1985), Phys. Rev. Lett. B 11, 1278. Leo, P. H., Lowengrub, J. S., Jou, H. J. (1998), Acta
Komura, S., Takeda, T., Osamura, K., Okuda, K. Mater. 46, 2113.
(1988), in: Dynamics of Ordering Processes in Li, D. Y., Chen, L. Q. (1998), Acta Mater. 46, 639.
Condensed Matter: Komura, S., Furukawa, H. Lifshitz, I. M. (1962), Sov. Phys. JETP 15, 939.
(Eds.). New York: Plenum Press, p. 251. Lifshitz, I. M., Slyozov, V. V. (1961), J. Phys. Chem.
Konigsveld, R. Kleintjens, L. A., Nies, E. (1987), Solids 19, 35.
Croat. Chem. Acta 60, 53. Lironis, G., Heermann, D. W., Binder, K. (1989),
Konishi, H., Noda, Y. (1988), in: Dynamics of Order- J. Phys. A: Math. Gen. 23, L329.
ing Processes in Condensed Matter: Komura, S., Liu, A. J., Durian, D. J., Herbolzheimer, E. Safran, S.
Furukawa, H. (Eds.). New York: Plenum Press, p. A. (1990), Phys. Rev. Lett. 65, 1897.
309. Livet, F., Bloch, D. (1985), Scr. Metall. 10, 19.
Kostorz, G. (1988), in: Dynamics of Ordering Pro- Lopez, V. M., Sano, N, Sakurai, T, Hirano, K. (1993).
cesses in Condensed Matter: Komura, S., Furu- Acta Metall, Mater. 41, 265.
kawa, H. (Eds.). New York: Plenum Press, p. 199. Luiggi, N., Simon, J. P., Guyot, P. (1980), Acta
Kostorz, G. (1991), J. Appl. Cryst. 24, 444. Metall. 28, 1115.
Kostorz, G. (1994), Solid – Solid Phase Transforma- Lyon, O., Simon, J. P. (1987), Phys. Rev. B 35,
tions, PTM’94: Johnson, W. C., Howe, J. M., 5164.
Laughlin, D. E., Soffa, W. A. (Eds.). The Metals, Maheswari, A., Ardell, A. J. (1993), Phys. Rev. Lett.
Minerals and Materials Society: Warrendale, p. 133. 70, 2305.
478 6 Spinodal Decomposition
Mainville, J., Yang, Y. S., Elder, K. R., Sutton, M., Ohta, T., Enomoto, Y., Kawasaki, K., Sato, A.
Ludwig Jr., K. F., Stephenson, G. B. (1997), Phys. (1988), in: Dynamics of Ordering Processes in
Rev. Lett. 78, 2787. Condensed Matter: Komura, S., Furukawa, H.
Manning, J. R. (1986), Diffusion Kinetics for Atoms (Eds.). New York: Plenum Press, p. 127.
in Crystals. Princeton: Van Nostrand. Ohta, T., Jasnow, D., Kawasaki, K. (1982), Phys.
Marro, J., Bortz, A. B., Kalos, M. H., Lebowitz, J. L Rev. Lett. 49, 1223.
(1975), Phys. Rev. B 12, 2000. Okada, M., Han, C. C. (1986), J. Chem. Phys. 85,
Marro, J., Lebowitz, J. L., Kalos, M. H. (1979), Phys. 5317.
Rev. Lett. 43, 282. Oki, K., Hasaka, M., Eguchi, T. (1973), Jpn. J. Appl.
Mashiyama, H. (1988), in: Dynamics of Ordering Phys. 12, 1522.
Processes in Condensed Matter: Komura, S., Fu- Okumura, H., Inoue, A., Masumoto, T. (1993), Acta
rukawa, H. (Eds.). New York: Plenum Press, p. Metall. Mater. 41, 915.
355. Onodera, Y., Hirano, K. (1986), J. Mater. Sci. Lett. 5,
Matsuura, M. (1988), in: Dynamics of Ordering Pro- 1048.
cesses in Condensed Matter: Komura, S., Furu- Onodera, Y., Hirano, K. (1988), in: Dynamics of Or-
kawa, H. (Eds.). New York: Plenum Press, p. 349. dering Processes in Condensed Matter: Komura,
Mazenko, G. F. (1994), Phys. Rev. E 50, 3485. S., Furukawa, H. (Eds.). New York: Plenum Press,
Mazenko, G. F. (1998), Phys. Rev. E 58, 1543. p. 293.
Mazenko, G. F., Wickham, E. (1995), Phys. Rev. Onuki, A. (1984), Phys. Lett. A 101, 286.
E 51, 2886. Onuki, A. (1986 a), Phys. Rev. A 34, 3528.
Mcdevitt, T. L., Mahajan, S., Laughlin, D. E., Bon- Onuki, A. (1986 b), J. Chem. Phys. 85, 1122.
ner, W. A., Keramidas, V. G. (1992), Phys. Rev. Onuki, A. (1989 a), J. Phys. Soc. Japan 58, 3065.
B 45, 6614. Onuki, A. (1989 b), J. Phys. Soc. Japan 58, 3069.
McPherson, R. (1987), J. Mater. Sci. Lett. 6, 795. Onuki, A. (1989 c), Int. J. Thermophys. 10, 293.
Meakin, P., Metiu, H., Petschek, R. G., Scalapino, D. Onuki, A. (1997), J. Phys.: Condens Matter 9, 6119.
J. (1983), J. Chem. Phys. 69, 1948. Onuki, A., Kawasaki, K. (1978), Prog. Theor. Phys.
Meier, H., Strobl, G. R. (1988), Macromolecules 20, Suppl. 64, 436.
649. Onuki, A., Kawasaki, K. (1979), Ann. Phys. 121,
Milchev, A., Heermann, D. W., Binder, K. (1988), 456.
Acta Metall. 36, 377. Onuki, A., Nishimori, H. (1991), Phys. Rev. B 43,
Mirold, P., Binder, K. (1977), Acta Metall. 25, 1435. 13649.
Monette, L., Liu, A. J., Grest, G. S. (1992), Phys. Onuki, A., Yamamoto, R., Taniguchi, T. (1997),
Rev. A 46, 7664. Phys. II (France) 7, 295.
Morii, Y., Gaulin, B. D., Spooner, S. (1988), in: Dy- Oono, Y., Puri, S. (1988), Phys. Rev. A 38, 434.
namics of Ordering Processes in Condensed Mat- Orlikowski, D., Sagui, C., Somoza, A., Roland, C.
ter: Komura, S., Furukawa, H. (Eds.). New York: (1999), Phys. Rev. B59, 8646.
Plenum Press, p. 239. Osamura, K. (1988), in: Dynamics of Ordering Pro-
Mouritsen, O. G. (1986), Phys. Rev. Lett. 56, 850. cesses in Condensed Matter: Komura, S., Furu-
Mouritsen, O. G., Praestgaard, E. (1988), Phys. Rev. kawa, H. (Eds.). New York: Plenum Press, p. 187.
B 38, 2703. Paris, O., Fährmann, M., Fratzl, P. (1995), Phys. Rev.
Mukherjee, S., Cooper, B. R. (1998), Mat. Sci. Eng. Lett. 75, 3458.
A 248, 101. Paris, O., Färmann, M., Fährmann, E., Pollock, T. E.,
Müller-Krumbhaar, H. (1974). Phys. Lett. 50 A, 27. Fratzl, P. (1997), Acta Mater. 45, 1085.
Nattermann, T. (1998), in: Spin Glasses and Random Penrose, O., Lebowitz, J. L. (1971), J. Stat. Phys. 3,
Fields: Young, A. P. (Ed.). Singapore: World Sci- 211.
entific, p. 277. Perrot, F., Guenoun, P., Baumberger, T., Beysens, D.,
Nielaba, P., Fratzl, P. Lebowitz, P. (1999), J. Stat. Garrabos, Y., Lenemdre, B. (1994), Phys. Rev.
Phys. 95, 23. Lett. 73, 688.
Nishihara, S., Noda, Y., Yamada, Y. (1982), Solid Pincus, P. (1981), J. Chem. Phys. 75, 1996.
State Commun. 44, 1487. Pine, D. J., Easwer, N., Maher, J. V., Goldburg, W. I.
Nishimori, H., Onuki, A. (1990), Phys. Rev. B 42, 980. (1984), Phys. Rev. A 29, 308.
Noda, Y. (1988), in: Dynamics of Ordering Processes Pistoor, N., Binder, K. (1988 a), Colloid Polym. Sci
in Condensed Matter: Komura, S., Furukawa, H. 266, 132.
(Eds.). New York: Plenum Press, p. 373. Pistoor, N., Binder, K. (1988 b), in: Polymer Motion
Noda, Y., Nishihara, S., Yamada, Y. (1984). J. Phys. in Dense Systems: Richter, D., Springer, T. (Eds.).
Soc. Jpn. 53, 4241. Berlin: Springer, p. 285.
Nose, T. (1987), Phase Transit. 8, 245. Porod, G. (1951), Kolloid Z. 124, 83; 125, 51.
Ohta, T. (1984), Prog. Theor. Phys. 71, 1409. Puri, S., Dünweg, B. (1992), Phys. Rev. A45, R 6977.
Ohta, T., Nozaki, H. (1989), preprint referred to by Puri, S., Frisch, H. L. (1997), J. Phys.: Cond. Matt. 9,
Bates and Wiltzius (1989). 2109.
6.7 References 479
Puri, S., Oono, Y. (1988), Phys. Rev. A 38, 1542. Sur, A., Marro, J., Lebowitz, J. L., Kalos, M. H.
Puri, S., Bray, A. J., Lebowitz, J. L. (1997), Phys. (1977), Phys. Rev. B 15, 535.
Rev. E 56, 758. Takahashi, J., Kuwayama, M., Kamiya, H., Takatsu,
Rao, M., Marro, J., Kalos, M. H., Lebowitz, J. L. M. (1988), Mater. Sci. 23, 337.
(1976), Phys. Rev. B 13, 2000. Takeda, S., Kulik, J., De Fontaine, D. (1987), Acta
Rikvold, P. A., Gunton, J. D. (1982), Phys. Rev. Lett. Metall. 35, 2243.
49, 226. Tanaka, H. (1993), Phys. Rev. Lett. 70, 53.
Rogers, T. M., Desai, R. C. (1989), Phys. Rev. B 39, Tanaka, H. (1994), J. Chem. Phys. 100, 5323.
11956. Tanaka, H. (1996), Phys. Rev. Lett. 76, 787.
Rogers, T. M., Elder, K. R., Desai, R. C. (1988), Taniguchi, T., Onuki, A. (1996), Phys. Rev. Lett. 77,
Phys. Rev. B 37, 9638. 4910.
Rouse, P. E. (1953), J. Chem. Phys. 21, 1273. Tokuyama, M., Enomoto, Y. (1993), Phys. Rev. E 47,
Sadiq, A., Binder, K. (1984), J. Stat. Phys. 35, 1156.
517. Tokuyama, M., Enomoto, Y., Kawasaki, K. (1987),
Sagui C., Somoza, A. M., Desai, R. (1994), Phys. Physica A 143, 187.
Rev. E 50, 4865. Tomita, H. (1984), Prog. Theor. Phys. 71, 1405.
Sahni, P. S., Gunton, J. D. (1980), Phys. Rev. Lett. 45, Tomokiyo, Y., Yahiro, K., Matsumura, S., Oki, K.,
369. Epudi, T. (1988), in: Dynamics of Ordering Pro-
Sahni, P. S., Gunton, J. D., Katz, S., Timpe, R. cesses in Condensed Matter: Komura, S., Furu-
(1982), Phys. Rev. B 25, 389. kawa, H. (Eds.). New York: Plenum Press, p. 233.
San Miguel, M., Gunton, J. D., Dee, G., Sahni, P. S. Tomozawa, M., MacCrone, R. K., Herman, H.
(1981), Phys. Rev. B 23, 2334. (1970), Phys. Chem. Glasses 11, 136.
San Miguel, M., Grant, M., Gunton, J. D. (1985), Toral, R., Chakrabarti, A., Gunton, J. D. (1988),
Phys. Rev. A 31, 1001. Phys. Rev. Lett. 60, 2311.
Sariban, A., Binder, K. (1989 a), Polym. Commun. Tringides, M. C., Wu, P. K., Moritz, W., Lagally,
30, 205. M. G. (1986), Ber. Bunsenges. Phys. Chem. 90, 277.
Sariban, A., Binder, K. (1989 b), Colloid Polym. Sci. Tringides, M. C., Wu, P. K., Lagally, M. G. (1987),
267, 469. Phys. Rev. Lett. 59, 315.
Sato, R., Han, C. C. (1988), J. Chem. Phys. 88, 2057. Unger, C., Klein, W. (1984), Phys. Rev. Lett. B 29,
Schätzel, K., Ackerson, B. J. (1993), Phys. Rev. E 48, 2698.
3766. Valls, O. T., Farrell, J. E. (1993), Phys. Rev. E 47,
Scher, M., Zallen, R. (1970), J. Chem. Phys. 53, 3759. R 36.
Schmitz, G., Hono, K., Haasen, P. (1994), Acta Villain, J. (1985), in: Scaling Phenomena in Disor-
Metall. Mater. 42, 201. dered Systems: Pynn, R., Skjeltorp, A. (Eds.). New
Selke, W. (1987), J. Phys. B 20, L 455. York: Plenum, p. 423.
Sequeira, A. D., Calderon, M. A., Kostorz, G., Peder- Villain, S. (1986), Europhys. Lett. 2, 531.
sen, J. S. (1995), Acta Metall. Mater. 43, 3427; 3441. Wagner, C. (1961), Z. Elektrochem. 65, 243.
Shinozaki, A., Oono, J. (1993), Phys. Rev. E 48, Wagner, R., Kampmann, R., Vorhees, P. W. (2001),
2622. in: Phase Transformations in Materials: Kostorz,
Sigga, E. (1979), Phys. Rev. A 20, 595. G. (Ed.). Weinheim: Wiley-VCH, Chap. 5.
Simon, J. P., Guyot, P., Ghilarducca de Silva (1984), Wang, Y., Khachaturyan, A. G. (1995), Acta Metall.
Philos. Mag. A 49, 151. Mater. 43, 1837.
Singhal, S. P., Herman, H., Kostorz, G. (1978), J. Wang, G. C., Lu, T. M. (1983), Phys. Rev. Lett. 50,
Appl. Crystallogr. 11, 572. 2014.
Sinha, D. N. Hoffer, J. K. (1981), Physica 107 B, 157. Wang, Y., Chen, L. A., Khachaturyan, A. G. (1992),
Snyder, H. L., Meakin, P. (1983), J. Chem. Phys. 79, Phys. Rev. 46, 11194.
5588. Wang, Y., Chen, L. A., Khachaturyan, A. G. (1993),
Srolovitz, D. J., Grest, G. S. (1985), Phys. Rev. B 32, Acta Metall. Mater. 41, 279.
3021. Wang, Y., Banarjee, D., Su, C. C., Khachaturyan, A.
Stanley, H. E. (1971), An Introduction to Phase Tran- G. (1998 a), Acta Mater. 46, 2983.
sitions and Critical Phenomena. Oxford: Oxford Wang, W. H., Wei, Q., Friedrich, S. (1998 b), Phys.
University Press. Rev. B 57, 8211.
Stauffer, D. (1985), An Introduction to Percolation Weinkamer, R., Gupta, H., Fratzl, P., Lebowitz, J. L.
Theory. London: Taylor and Francis. (2000), Europhys. Lett. 52, 224.
Stauffer, D., Ferer, M., Wortis, M. (1972), Phys. Rev. Wilcoxon, J. P., Schaefer, D. W., Kaler, E. W. (1988),
Lett. 29, 245. Phys. Rev. Lett. 60, 333.
Stephenson, G. B., Warburton, W. K., Haller, W., Wilcoxon, J. P., Martin, J. E., Odinek, J. (1995),
Bienenstock, A. (1991), Phys. Rev. B 43, 3417. Phys. Rev. Lett. 75, 1558.
Su, C. H., Voorhees, P. W. (1996), Acta Mater. 44, Wiltzius, P., Bates, F. S., Dierker, S. B., Wignall, G.
1987; 2001. D. (1987), Phys. Rev. A 36, 2991.
480 6 Spinodal Decomposition
Wong, N. C., Knobler, C. M. (1978), J. Chem. Phys. rukawa, H. (Eds.). New York: Plenum Press, p.
69, 725. 343.
Wong, N. C., Knobler, C. M. (1979), Phys. Rev. Lett. Yaldram, K., Binder, K. (1991), J. Stat. Phys. 63, 161.
43, 1733. Yeung, C. (1988), Phys. Rev. Lett. 61, 1135.
Wong, N. C., Knobler, C. M. (1981), Phys. Rev. A 24, Yokota, H. (1978), J. Phys. Soc. Jpn. 45, 29.
3205. Zarzycki, J., Naudin, F. (1969), J. Non-Cryst. Solids.
Wortis, M. (1985), in: Fundamental Problems in Sta- 1, 215.
tistical Physics, Vol. VI: Cohen, E. G. D. (Ed.). Zettlemoyer, A. C. (1969), Nucleation. New York:
Amsterdam: North-Holland, p. 87. Dekker.
Wu, P. K., Perepezko, J. H., McKinney, J. T., Lagally, Zhang, Z., Chakrabarti, A. (1994), Phys. Rev. E 50,
M. G. (1983), Phys. Rev. Lett. 51, 1577. R 4290.
Wu, P. K., Tringides, M. C., Lagally, M. G. (1989), Zhang, Z., Chakrabarti, A. (1995), Phys. Rev. E 52,
Phys. Rev. B 39, 7595. 2736.
Yagi, T., Lu, Z.-L. (1988), in: Dynamics of Ordering Zuo, J.-K., Wang, G.-C., Lu, T. M. (1989), Phys. Rev.
Processes in Condensed Matter: Komura, S., Fu- B 40, 524.
7 Transformations Involving Interfacial Diffusion
Gary R. Purdy
Yves J. M. Bréchet
a crystalline phase
a, b included, external phase
ga activity coefficient
GB surface excess concentration of solute B per unit area
d boundary thickness
e misfit
ea interaction parameter; a measure of the non-ideality of the a solid solution
h coefficient of lattice parameter change with composition
q angle of tilt
k gradient energy coefficient
ka bulk modulus of precipitate
l spacing of parallel edge dislocations
mA, mB solvent, solute chemical potentials
W alignment parameter
s specific interfacial Helmholtz energy
sA , s B specific grain boundary Helmholtz energy in pure A, B
s ab specific interfacial Helmholtz energy of a/b interface
j phase field parameter
y order parameter
(Monte Carlo), and dynamic aspects such Monte Carlo methods) still requires further
as diffusion (molecular dynamics). The development.
structural information stemming from A surprisingly large amount of our infor-
these methods is very much dependent on mation relates to a rather special kind of
the interatomic potential chosen, and espe- interface, the symmetric tilt grain bound-
cially on the anharmonic part, since the ary. As suggested by Fig. 7-1, low angle
structure of the interface is usually rather boundaries are made up of arrays of dislo-
distorted. The dynamic aspects of inter- cations, and the low angle symmetric tilt
faces include the interface diffusivity and boundary is composed of parallel edge dis-
the interface mobility. While the diffusivity locations, whose spacing l is given by
is accessible to these simulations (via mo-
l = b/ 2 sin (q / 2 ) (7-1)
lecular dynamics), the mobility is more of a
problem, because the system size is typi- where b is the magnitude of the Burgers
cally constrained by available computa- vector, and q the angle of tilt. Studies by
tional resources. This imposes, for exam- Krakow et al. (1986) suggest that this de-
ple, very large curvatures as driving forces, scription in terms of individual disloca-
and results in high interface velocities tions can be extrapolated, with some mod-
which might not be realistic reflections of ification, to high angles, e.g., to 26.5°.
experimental situations (Sutton, 1995). More generally, the grain boundary region
The chemical aspects accessible to is expected to possess the following char-
Monte Carlo simulation also possess two acteristics:
facets: the thermodynamic equilibrium to a) The high angle grain boundary is thin,
be attained and the kinetics required to perhaps two or three atomic diameters in
reach it. Provided that the interatomic po- width. To a first approximation, it may be
tential is reliable, the Monte Carlo method considered a high energy “phase”, con-
gives an accurate description of chemical strained to have constant volume.
segregation (Treglia and Legrand, 1998). b) The high angle grain boundary is peri-
However, the simulation of the kinetics odic in structure, like its low angle counter-
needed to reach this state encounters a new part. The geometric repeating units are not
problem of coupling the bulk diffusion to
the diffusion close to the interface. Monte
Carlo simulations with time residence al-
gorithms (Martin et al., 1998) have allowed
the inclusion of the role of vacancies in
bulk diffusion, but the treatment of vacan-
cies in solids with free surfaces or internal
interfaces remains a source of problems
(Delage, 1998).
The whole field of surface and interface
thermodynamics and kinetics of metallic
alloys with tendencies to ordering or un-
mixing has profited from Monte Carlo
simulations. The next step (to introduce Figure 7-1. A symmetric low angle tilt grain bound-
simultaneous structural relaxations and ary. The tilt angle q and the edge dislocation spacing
chemical aspects, e.g., via off-lattice l are related by Eq. (7-1).
486 7 Transformations Involving Interfacial Diffusion
the excess Helmholtz energy is computed, the interfaces is thus related to the coupling
and this excess energy is then taken as the term weighting the square gradients.
interfacial energy. In a single model we These continuum descriptions are in
therefore get both a “continuum structure” principle well suited for application to dif-
for the interface, and the specific interfa- fuse interfaces. Their facility of implemen-
cial Helmholtz energy. tation in computer simulations (since they
The time-dependent version of these do not involve front tracing methods) has
equations allows us to treat the motion of triggered much interest under the name of
the interface under a prescribed driving “phase field approaches” (Carter et al.,
force, provided that a relation is assumed 1997). They have recently been applied to
between the fluxes and the driving forces. various problems involving sharp inter-
For a conserved order parameter C (such faces using the following systematics: the
as concentration), the kinetic equation interface is artificially made diffuse with
takes the form: respect to a “phase field parameter” j . (For
instance, this parameter can be taken equal
∂C
= M [∇ 2 f ( C ) − k 2 ∇ 4 C ] (7-2) to 1 in a solid and to 0 in a liquid if we wish
∂t to address a solidification problem involv-
where k is a gradient energy coefficient, ing both solute and heat flow (Wheeler et
and for a non-conserved order parameter al., 1992). It can be taken to vary continu-
(such as a long-range order parameter, or a ously from +1 to –1 through a grain bound-
crystal orientation), we obtain: ary, mimicking the orientation change
across the boundary for application to a
∂y
= M [k 2 ∇ 2 y − f ′ (y )] (7-3) problem in grain growth (Chen et al.,
∂t 1998).) A Helmholtz energy equation is
The interfacial Helmholtz energy for a dif- proposed which involves both the concen-
fuse interface between two concentrations tration and the phase field parameter and
C1 and C2 is given by their spatial derivatives. The dynamic equa-
tions for the concentrations and the phase
C2
fields are solved using continuum methods
s = 2 Nv ∫ k 2 f (C ) dC (7-4)
C1
and the coupling term containing the phase
field gradient is set equal to zero so that the
Similarly, assuming a quadratic expan- sharp interface can be recovered.
sion of the Helmholtz energy with two min- These methods have been applied to
ima for order parameters y = 1 and y = – 1 solidification, phase transformations, grain
gives the following expression for the inter- growth, and chemically induced grain
facial Helmholtz energy of an antiphase boundary migration. They possess the ad-
boundary: vantage of relative ease of implementation
using classical numerical techniques, and a
8
s= Nv k 2 f ′′ (7-5) great flexibility in terms of the introduction
3 of such features as anisotropy. They share
In each case, the width of the interface scales the drawback of all Ginzburg – Landau type
as k /f 1/2: the stronger the gradient term com- phenomenological equations, i.e., that the
pared with the local chemical Helmholtz precise atomistic interpretation of both the
energy contribution, the more diffuse will mobility and the coupling terms is not al-
be the interface. The “structural width” of ways clear.
488 7 Transformations Involving Interfacial Diffusion
a) The activation enthalpy for grain Having summarized briefly the present
boundary self-diffusion is of the order of level of understanding of grain boundary
half that for volume diffusion. Indeed, a diffusion, it is perhaps worth emphasizing
good empirical correlation exists between that the correlation of structure with diffu-
grain boundary diffusivity and reduced sivity remains incomplete, and that much re-
temperature T /Tm for different metal struc- mains to be learned about the fundamental
tures. For f.c.c. metals, Gust et al. (1985) processes of interfacial diffusion in solids.
obtained (7-10) Interphase boundary diffusion is less
– well studied, to the extent that very few
d D b = 9.7 ¥ 10–15 exp (– 9.07 Tm /T ) m3/s
reliable data are available in the litera-
b) For the low angle symmetric tilt ture. Kaur and Gust (1988) cite only one
boundary, rates of diffusion are aniso- system for which extensive measurements
tropic, with faster diffusion occurring par- have been reported, the Sn – Ge/In system
allel to the dislocation lines. Undissociated (Straumal et al., 1981). One reason for the
dislocations have higher diffusivities than lack of direct information about this impor-
dissociated ones; the diffusivity increases tant aspect of interfacial diffusion may lie
linearly with dislocation density (and there- in the fact that, unlike grain boundaries, in-
fore with tilt angle q ) for angles up to terphase boundaries can support local equi-
about 10°. librium concentration gradients only in the
c) Anisotropic behavior is also obtained presence of gradients of curvature, temper-
for higher angle symmetric tilt boundaries. ature or stress. The isothermal planar inter-
d) Low angle twist boundaries, com- phase boundary is normally isoconcentrate.
posed of screw dislocations, are less effi- Mullins (1957) has considered a number
cient in transporting material than their tilt of important cases of interfacial diffusion
counterparts. in response to variations in curvature. For
492 7 Transformations Involving Interfacial Diffusion
the case of substitutional diffusion, and Helmholtz energy s A, where s is the spe-
under the simplifying assumptions of con- cific interfacial Helmholtz energy and A
stancy of composition of the two phases a the interfacial area. For constant s , the
and b and site conservation at the interface, force can be expressed as
we may write for the interfacial flux:
ps = K s (7-12)
A
Jab =− B
Jab =− Lab ∇ (m ab
A
− B
m ab ) which is the product of the surface Helm-
Lab Vm s ab holtz energy and the interface curvature
=− ∇Kab (7-11) K = dA/dV.
ca − cb
For smoothly curved interfaces the curva-
Here, L represents the mobility of atoms ture is the sum of the reciprocals of the two
along the interface, Vm is a molar volume, principal radii of curvature [(1/r1) + (1/r 2 )].
the m’s are chemical potentials within the For strongly anisotropic or faceted inter-
interface, s ab is the specific interfacial faces, an effective capillary force may still
Helmholtz energy and K ab is the interfacial be defined, through the derivative dA/dV, al-
curvature. though care must be taken in its formulation.
For the special case of an equilibrium
particle, the capillary force gives rise to
7.4 Forces for Interface Migration a variation of the specific Helmholtz
energy of the included phase: the resulting
It is useful to define and classify a set of relation between the zero-curvature equi-
generalized “forces” for interface migra- libria (C0 ) and the equilibrium composi-
tion. In most cases, the driving and retard- tions in the presence of curvature is sum-
ing forces are expressed in terms of a nor- marized in the Gibbs – Thompson relation-
mal force per unit area of interface, and ship. For two components we obtain the
therefore have the units of pressure. linearized expressions (Purdy, 1971)
It is conventional to separate the various
forces according to their origin (e.g., surface C0a (1 − C0a ) Vm ⎛ K s ⎞
DC a = (7-13 a)
Helmholtz energy, elastic energy, chemical ( C0a − C0b ) ⎜⎝ RT eb ⎟⎠
Helmholtz energy) and to some extent the
division is arbitrary. Its utility lies in the C0b (1 − C0b ) Vm′ ⎛ K s ⎞
DC b = (7-13 b)
ability to visualize the interplay of different ( C0a − C0b ) ⎜⎝ RT eb ⎟⎠
forces on a moving transformation interface.
It is also true that the definition of the where a is the included phase and b the
“force” for boundary migration is only a external phase, Vm is the molar volume of
part of the solution to the migration prob- the a phase, V m′ the molar volume of the
lem. The response function (or mobility) of a phase extrapolated to the composition of
the interface must also be known, and this ⎛ ∂ ln g b ⎞
the b phase and e b = ⎜ 1 + , where
∂ ln Cb ⎟⎠
function is expected to differ greatly from
one interface type to the next. ⎝
g b is an activity coefficient. This last term
is a measure of the departure of the parent
7.4.1 Capillary Forces
phase from ideal solution behavior.
These forces originate from the system’s An equivalent graphical construction
tendency to reduce its total interfacial due to Hillert (1975 a) is shown in Fig. 7-7.
7.4 Forces for Interface Migration 493
Figure 7-8. Helmholtz energy and interfacial com- 7.4.3.2 Plastic Response
position relationships for a composition-invariant
phase transformation. The available Helmholtz en- It is important to distinguish between
ergy for the transformation is used in part to drive the mechanically derived forces which origi-
interface, DF ¢, and the balance, DF 0, is dissipated in
nate from elastic strain energy, and those
the volume-diffusion field.
which result in the plastic relaxation of the
loaded specimen. In the latter case, the
additive term in the Helmholtz energy of a elastic energy may remain approximately
phase. Thus, for homophase variations, the constant during the course of the transfor-
local elastic energy density is given by mation, but the interfaces can experience
h 2 DC 2 Y as indicated in Eq. (5-12). virtual forces, in the sense that the Helm-
This term generally reduces the total holtz energy of the loading system is re-
driving force for a phase transformation. In duced by interface motion.
Eq. (5-12), DC is the local difference from As an example, consider the symmetric
the average concentration C0 . The equation tilt boundary of Fig. 7-9, which is capable
therefore has application to any case where of motion in response to the applied load.
a gradient of misfitting solute exists in an Studies of this type of boundary have
otherwise perfect crystal, and in particular shown that the dislocation boundary under-
to the case where a thin layer of altered goes normal migration, that is, synchro-
composition is in coherent contact with an nous glide, under an appropriate load. We
infinite bulk phase of composition C0 . can consider that there is a virtual mechan-
7.4 Forces for Interface Migration 495
event associated with local fluctuations in boundary migration are not fully under-
subgrain configuration; it is unlikely to be stood.
successfully modeled either by a standard When two or more components are in-
nucleation theory or by an averaging ap- volved, this picture must be extended.
proach for the stored energies. Information Superimposed on the intrinsic response of
on local misorientations (leading to differ- the interface, and frequently masking it, we
ences in local mobilities and local driving find the effects of solutes which are prefe-
forces) may be crucial, as has been sug- rentially attracted to or repelled from the
gested using vertex-based numerical simu- interface. The experiments of Aust and
lations (Weygand, 1998). Rutter (1959) first demonstrated conclu-
sively the powerful retarding effects of
trace amounts of solute on grain boundary
7.4.4 Frictional Forces
migration.
7.4.4.1 Solute Drag The idea of a “solute drag” perhaps orig-
inated with Cottrell (1953), who noted that
Each type of mobile interface will pos-
solute atoms would be attracted to disloca-
sess its own characteristic response func-
tions, and would therefore be required to
tion. For the simple tilt wall subject to a
diffuse along with the dislocations, or to be
mechanical force, we have noted that the
left behind by dislocations that had broken
force – velocity relationship is linear over a
free of their solute atmospheres. Similar
range of forces. If a considerable recon-
ideas were put forward by Lücke and his
struction of the crystal structure is required
co-workers (1957) for application to the
for the motion of an interface, as in the mo-
case where a solute is preferentially segre-
tion of a high-angle boundary in a single-
gated to grain boundaries. The most com-
component system, we expect that cross-
plete and informative theoretical treat-
boundary diffusive motion will play a sig-
ments of this effect are those of Cahn
nificant role. Turnbull’s (1951) expression
(1962) and Hillert and Sundman (1976).
for the intrinsic mobility, M, of a grain
The latter is based on a treatment put for-
boundary
ward earlier by Hillert (1969). The basic
d D b Vm reason for the solute drag force is simple: if
M= (7-16) solute is attracted to the boundary, it will
b 2 RT
tend to diffuse along with the moving
is based on a model in which atom jumps boundary. Depending on the relative rates
across the boundary are independent of one of boundary motion and solute diffusion,
another. This relation has not been con- the solute distribution may become asym-
firmed by experiment, and this is taken to metric, with more solute trailing the
suggest that more complex, cooperative boundary than leading it. Because of the
movements of atoms are generally in- mutual attraction between the boundary
volved in interface migration. Again refer- and the solute, the asymmetric solute dis-
ring to experiment, there exists a consider- tribution leads to a retarding force of the
able database drawn from measurements form
of grain boundary motion in high purity
pi = NV Ú (C – C0 ) (dE/dx) dx (7-17)
metals. Thus the intrinsic response of
high-angle grain boundaries is rather well where NV is the number of atoms per unit
documented, even if the mechanisms of volume, and dE/dx is a measure of the
7.4 Forces for Interface Migration 497
force between the solute atoms and the is not restricted to dilute solutions, nor is it
boundary. The effects of small amounts of restricted to grain boundaries. Figure 7-11
solute can be profound. The profiles in Fig. uses the simple square well to illustrate the
7-10 are due to Cahn (1962). They are effect of cross-boundary diffusion in set-
based on a triangular interaction potential ting the degree of asymmetry of the solute
between solute and boundary, and corre- profile. The drag term may be written in
spond (a) to a slowly moving boundary, (b) several ways. Among the most convenient,
to a boundary experiencing maximum so- Hillert and Sundman (1976) find:
lute drag, and (c) to a fast-moving bound-
ary. That is, the drag force goes through a RT v ( c − c0 )2
pi =
Vm
∫ Dc (1 − c ) dx (7-18)
maximum as the velocity is increased. If an
initial intrinsic force – velocity relationship In their approach, the ramps (zones 1 and 3
is assumed linear, and the solute drag term in Fig. 7-12), as well as the centre of the
is added, the total response function is boundary (zone 2) and the parent phase re-
quickly rendered nonlinear. Under more gion (zone 4), contribute to the drag. In-
severe conditions, the system can even be- deed, it is possible to separate out the con-
come unstable, displaying a region of ve-
∂p
locities where is negative.
∂v
Hillert’s formulation uses a different
interaction potential, initially square
(rather than triangular), and later a square-
topped potential with ramps at the edges.
The drag is evaluated as a rate of dissipa-
tion of Helmholtz energy due to diffusion
in and in front of the moving interface. The
model lends itself to numerical analysis; it
interface region (Eq. (7-18)). The solute Helmholtz energy defines the most prob-
field is schematized in Fig. 7-17 (a) for a able velocity. A second treatment of the so-
solute that is attracted to the interphase lute drag in massive growth (Bréchet and
boundary. The cross-boundary diffusion of Purdy, 1992) extended Cahn’s (1962) anal-
solute introduces a component of drag (re- ysis, and demonstrated that a finite drag
gions 1, 2 and 3), and diffusion ahead of term will always be present, however small
the boundary provides a further contribu- the velocity. This analysis, like that of Hil-
tion (region 4), which dominates the total lert and Sundman, therefore suggests that a
drag as the velocity approaches zero. The threshold driving force is a natural charac-
vd teristic of such transformations. However,
total drag is optimized for ≈ 1, and it
2D if the solute field in front of the interface is
may be that the optimum of dissipation of accompanied by a misfit, in the sense of
Eq. (5-12), we expect an additional term in
the force balance, a “pulling” force due to
the elastic energy contained in the coherent
composition gradient.
The above discussion is based on the
premise that the migration characteristics
of the massive front are those of a non-co-
herent interface. This is borne out by nu-
merous observations (Massalski, 1984)
which indicate, for example, that the trans-
formation interface is able to cross grain
boundaries in the parent phase without a
change in velocity or morphology. Dy-
namic observations indicate, however, that
the motion of the front is irregular, and of-
ten accomplished by a lateral process, in
which diffuse steps move parallel to the
interface plane (Perepezko, 1984). In the
view of Menon et al. (1988), lateral pro-
cesses are the rule in massive growth.
Hence models based solely on diffusion
normal to the boundary may need to be
modified. However, it is interesting that
Perepezko (1984) demonstrated a scaling
relationship for a wide range of alloy
systems, which yields a composite en-
Figure 7-17. a) Schematic concentration profile thalpy of activation for massive propaga-
through a massive transformation interface, i) at tion of 94 Tm J/mol. This value is similar to
equilibrium, and ii) in motion. b) Computed solute that for grain boundary diffusion, which
drag for the massive transformation in part a, for a suggests that diffusion within the interface
constant value of D. XA· and X B· represent the initial
composition of the parent phase expressed as mole is a common rate-determining feature for
fractions of the two components A and B. After Hil- all such transformations.
lert and Sundman (1976).
7.5 Examples 503
(1989), who systematically varied the mis- and unstrained product crystal. These con-
fit parameter in a ternary system, and siderations are contained in the phenomen-
showed that the phenomenon is suppressed ological treatment due to Bréchet and
when the misfit is brought to zero. An ap- Purdy (1992), in which the driving force is
proximate parabolic dependence of veloc- evaluated over a correlation length (l) on
ity on misfit was also found, consistent either side of the boundary. The steady
with Eq. (5-12). This suggests that elastic concentration profile is C (z). The force
strain energy also plays a role in propaga- then becomes:
tion of the grain boundary. (7-21)
l
In practice it is difficult to distinguish 1 ⎡1 ∂ F 2
⎤
∫ ( C ( z ) − C0 )2 + Y h 2 C ( z )2 ⎥ dz
the possible driving forces for the process. l 0 ⎢⎣ 2 ∂C 2 ⎦
At the highest temperatures, it is likely that
solute field stresses will be dominant. This where the second derivative is a measure of
appears to be the case for the study of Rhee the Helmholtz energy composition relation.
and Yoon (1989). In the early work of Hil- This approach has the virtue of including
lert and Purdy (1978), in which thin poly- both possible contributions to the driving
crystalline iron films were exposed to zinc force; it results in the conclusion that, as in
vapor, it was evident that the volume diffu- the massive transformation, a threshold ex-
sion penetration of the parent grains, in- ists, below which no motion is possible.
dexed by DV /v, was of atomic dimensions. An ambitious treatment of the problem
Hence, it was assumed that the driving has been put forward by Cahn et al. (1998).
force was entirely chemical. A comprehen- Their analysis is based on a phase-field
sive model should take into account all treatment of the grain boundary. They look
possible sources of the driving force. for travelling-wave solutions of the equa-
The moving grain boundary is subject to tions of motion, and treat their existence
a set of forces, which may in general in- as a requirement for forces capable of
clude chemical, elastic, frictional and cap- coupling with the boundary motion. They
illary forces. As in the massive transforma- find that coupling with the elastic field is
tion, there exists in principle a degree of indeed possible. However, their treatment
freedom corresponding to the concentra- gives no indication that a purely chemical
tion in the parent phase immediately adja- force is effective in moving the boundary.
cent to the boundary. In the conceptual
limit where no volume-diffusion penetra-
7.5.3 Discontinuous Precipitation
tion exists in advance of the boundary, the
concentration profile will be a step, and the Discontinuous precipitation, like chemi-
full chemical force pch will act across the cally-induced grain boundary migration,
interface. In a second limiting case, a con- involves the lateral diffusion of solute
centration gradient exists in front of the within a sweeping grain boundary. The dif-
boundary and an elastically-derived resid- ference lies in the nature and spacing of the
ual force is determined by the coherency solute sources/sinks. In the case of discon-
field. Intermediate cases would then corre- tinuous precipitation, these are members of
spond to a higher or lower degree of relax- a regular array of precipitates, whose spac-
ation of the concentration at the leading ing is a free variable, capable of internal
edge of the boundary towards a constrained adjustment. The reaction is found in a wide
equilibrium between strained parent crystal variety of precipitation systems, often at
7.5 Examples 505
lute-depleted volumes were left in the the steady state has attracted the attention
wake of bulging grain boundaries in super- of successive generations of theorists. In
saturated Al – 4.7 wt.% Cu. Fonda et al. this section, we review only the more ad-
(1998) and Mangan and Shiflet (1997) vanced theories, while acknowledging
studied the initiation of discontinuous pre- their geneology.
cipitation in Cu – 3% Ti alloys, and demon- Before proceeding, the question of the
strated that, at low supercooling, the initial steady state should be pursued in more de-
boundary motion is due to the presence of tail. It appears that the steady state can be
Widmanstätten precipitates, which grow achieved fairly generally, and that it can be
into one grain, and cause boundary dis- approached from higher or lower supersat-
placement into the other. Duly and Bréchet uration, such that the system has little or no
(1994) examined the initiation of discon- memory, and the steady state is characteris-
tinuous precipitation over a wide range of tic of the isothermal reaction conditions
temperatures, initial grain sizes and com- only. Nevertheless, the constancy of reac-
positions in Mg – Al alloys. They conclude tion front velocity has recently been ques-
that the Tu – Turnbull and Fournelle – Clark tioned by a number of authors, including
mechanisms dominate at low and high tem- Kaur and Gust (1988), Mangan and Shiflet
peratures respectively, and that the Purdy – (1997), and Fonda et al. (1998). This point
Lange mechanism may be important at is discussed in Sec. 7.5.3.3.
intermediate temperatures. It is convenient to divide the theoretical
It is now clear that a range of mecha- treatments into two classes depending on
nisms exist, each mechanism capable of in- whether the details of the transformation
itiating boundary motion. At lower temper- front are predicted as part of the develop-
atures, precipitation on the static boundary ment. In the “global” approach of Cahn
is a common precursor of discontinuous (1959), which builds on the ideas implicit
precipitation; the initial precipitates are ca- in Turnbull’s (1951) and Zener’s (1946)
pable of acting either directly, to pull the treatments, the reaction front is treated as a
boundary from its initial location, or indi- planar high-diffusivity path whose mobil-
rectly through CIGM to accomplish a simi- ity is rate determining. Thus the important
lar result. We conclude that, for most situa- input quantities to the theory include the
tions where the supersaturation is signifi- overall Helmholtz energy change accom-
cant, there will be no difficulty in initiating panying the passage of the front and the
grain boundary motion. interface mobility. Cahn (1959) first em-
phasized the importance of storage of
Helmholtz energy in the product phase, and
7.5.3.2 Theories of Steady Cooperative
showed how to estimate it from a knowl-
Growth
edge of the interphase boundary energy
The development of a steady or near- s a, b , the grain boundary diffusivity and
steady growth front, characterized by a distribution coefficient, s, and the interla-
regular spacing between precipitate lamel- mellar spacing S and velocity v. The Helm-
lae (or rods), S, and a constant velocity, v, holtz energy balance is written
is not immediate upon initiation, but is pre-
DF ≤ = DFch – 2 s ab Vm /S (7-22)
ceded by extended transient growth re-
gions. However, steady or near-steady con- Here DFch is the total chemical Helm-
ditions appear eventually to prevail, and holtz energy difference between the parent
508 7 Transformations Involving Interfacial Diffusion
and product phases, evaluated some dis- Certain aspects of the theory are amenable
tance from the interface. In determining to experimental verification, for example,
this quantity, the stored Helmholtz energy the concentration profiles left in the a
in the product a phase must be calculated product lamellae have been measured by
from a knowledge of the composition pro- Porter and Edington (1977) and by Solor-
file in the a lamellae. This is determined zano and Purdy (1984) using high resolu-
in theory from a knowledge of the grain tion elemental analysis, to give the trans-
boundary transport properties and the ve- port properties of the reaction front,
locity and spacing through {s D b d }, through Eq. (7-23). These same
measurements allow the evaluation of the
C0 − C ( z ) cosh ( z a / Sa )
= (7-23) amount of the total available Helmholtz
C0 − C3 cosh ( a / 2 ) energy for the transformation retained as
segregation in the product; this latter quan-
where C0 is the initial concentration in the
tity is also available, albeit in averaged
parent phase, C (z) is the local concentra-
form, through X-ray measurements of the
tion in the a phase behind the transforma-
lattice parameter of the product phase.
tion front, C3 is the concentration in the a
A heuristic description of discontinuous
phase in contact with the b lamellae and
precipitation, after Petermann and Horn-
v S2 bogen (1968), utilizes a rate equation of the
a= (7-24) type of Eq. (7-25) coupled with an approx-
s Db d
imate evaluation of the relaxation time for
Eq. (7-23) is a solution of the moving- grain boundary diffusion to the b lamellae,
boundary diffusion equation, for constant to give a velocity expression which is di-
v, S, and boundary concentration C3 ; the mensionally similar to that of Cahn.
solution reduces in approximation to a sine A second class of theoretical description
function when the centerline composition has been developed by Hillert (1969, 1972,
C (z = 0) is closer to C3 than to C0 . 1982). This approach has been termed “de-
e
Cahn (1959) next chose C3 = C ab , the tailed”, in the sense that a much deeper
equilibrium composition of a in contact knowledge of the transformation front is
with b at a planar interface. He was then implicit, even required, for its application.
able to evaluate the stored chemical Helm- The basis for this treatment is the applica-
holtz energy in the product phases from Eq. tion of a local force balance at every point
(7-23). To complete the description, he as- along the interface. Thus the possibility is
sumed a relationship between the total raised of determining the interface shape
Helmholtz energy change DF ≤ and the ve- for steady growth, as a consequence of the
locity of the form interplay of capillary, chemical, elastic,
and frictional forces, as defined in Sec. 7.4.
v = M≤ DF ≤ (7-25)
In simplified form, for application to the
Here, M≤ is a “global” mobility that dif- case where the composition profile (step),
fers from the intrinsic grain-boundary mo- DC, measured normal to the interface is not
bility, e.g., of Eq. (7-16). relaxed by volume diffusion, and where
The theory described above has been elastic and solute drag effects are negli-
modified by others, for example by Aaron- gible, the interfacial force balance becomes
son and Liu (1968), to take partial account
of capillary forces acting on the b lamellae. v/M = pch – ps (7-26)
7.5 Examples 509
which holds at every point. The solute pro- solute drag forces and severely nonplanar
file in the boundary will be given by Eq. shapes, thus allowing for consideration of
(7-23), and the concentration step (DC) at morphological stability of the transforma-
the interface will take a minimum value at tion front.
the centres of the lamellae, varying contin- Neither of the approaches described
uously to a maximum value at the a/a0 above is sufficient to predict the transfor-
interface immediately adjacent to the b la- mation state that will occur at given super-
mellae. Hillert has chosen the limiting saturation in a particular system. Each is
value of the solute concentration at this capable of generating a unique relationship
point, C3 in Eq. (7-23), as the capillary- between v and S, however.
modified solubility in a in equilibrium In considering the spacing problem for
with b under curvature 2/Sb . lamellar eutectoids, Zener (1964) set a
For a steady transformation, the right- thermodynamic limit, a minimum spacing
hand side of Eq. (7-26) is constant along for which all of the available Helmholtz
the interface, and the curvature must vary energy for the transformation is stored as
to match the chemical force, indexed by the interfacial Helmholtz energy in the prod-
composition step DC. uct. This minimum also exists for discon-
A further local force balance is required tinuous products:
at the a0 /a/b junction. This is depicted
by a vectorial balance of surface energies 2 s ab Vm
Smin = (7-27)
in Fig. 7-23. Thus the angles of intersec- DFch
tion and local curvatures are set at the junc-
tion, and the variation of the grain bound- At this (virtual) state, the system is in
ary curvature is required to match the vari- equilibrium and no growth is possible.
ation of chemical force along the boundary. Larger spacings corresponds to finite
The theory therefore allows the self-con- rates of growth. It has been suggested that
sistent calculation of the interface shape the steady spacing is one which maximizes
for given values of the spacing S, the veloc- the growth rate (Hillert, 1972) or the inte-
ity v, and the interfacial transport property gral rate of dissipation of Helmholtz
s D b d. energy (Cahn, 1959). Solorzano and Purdy
Sundquist (1973) extended Hillert’s (1984) put this latter hypothesis to the test,
treatment to include the possibility of and found a reasonable level of confirma-
tion for two well-characterized systems, out modification; however, the theories
Al – 22 at.% Zn and Mg – 9 at.% Al. need to take into account the possibility of
Building upon the local approach of Hil- lateral migration of the growth interface for
lert and co-workers, Klinger et al. (1996) lower supersaturation. A recent treatment
have proposed a solution which leads to a by Klinger et al. (1997 a) takes this into ac-
selection of the growth velocity without the count by describing the motion of the inter-
need for an optimization principle. This face as intermittent.
treatment relies on the cooperative nature
of the growth of the a and b lamellae, and
7.5.3.3 Experimental Observations
on the different characters of the interfaces
between the a and b phases and the parent The majority of experimental studies
a phase: one is a grain boundary and the have focused on microstructures developed
other is an interphase boundary. Assuming through steady isothermal growth. They
a definite composition for the b phase leads are therefore capable of description in
to a Mullins (1957) equation for transport terms of the theoretical treatments of the
along the a0/b interface determined by gra- previous section. The types of information
dients of curvature. Transport along the accessible to experiment are:
a0/a is considered to be driven by concen-
a) quantities that may be derived from
tration gradients, according to Cahn’s
conventional metallographic methods: av-
(1959) formulation. The condition at the
erage velocity v and spacing S;
triple junction is taken as a local equilib-
b) chemical information obtained either
rium, corrected for Gibbs – Thompson ef-
from averaging processes such as X-ray
fects. The requirement that the a and b
diffraction measurements of lattice param-
growth velocities match in order to have a
eters of product lamellae or from high res-
cooperative growth process is sufficient to
olution microanalyses of the product;
select the spacing and velocity of the front.
c) structural information relating to the
This approach has the advantage of avoid-
growth interface, as obtained for example
ing the introduction of optimization princi-
by transmission electron microscopy.
ples, and it has been shown to accurately
describe some experimental results for the In addition, for the test of theory we re-
Al – Zn system. It also predicts that steady- quire solution thermodynamic data, inter-
state solutions with invariant front shapes facial energies and interfacial diffusion co-
are possible within a limited range of driv- efficients. Many alloy systems have been
ing forces, suggesting the possibility of investigated. In this section we will focus
morphological instabilities outside this on the systems Al – Zn and Mg – Al, for
range. which much of the necessary data are avail-
All of these theories of steady growth as- able (e.g., Rundman and Hilliard, 1967;
sume that the motion of the interface is Cheetham and Sale, 1974; Hassner, 1974).
continuous. This assumption is probably Yang et al. (1988) collected velocity and
acceptable when the growth front is well- spacing data for Al – Zn alloys, and for a
rounded, but there is substantial evidence range of reaction temperatures and spac-
for faceted interfaces at low supersatura- ings, as summarized in Figs. 7-24 and 7-25.
tions, as discussed in detail in the next sec- Figure 7-26 gives the average composition
tion. In the high-supersaturation regime, of the a lamellae superimposed on the Al-
the continuum models seem to apply with- rich portion of the phase diagram. These
7.5 Examples 511
explanation based on the elastic interaction 7-29), and give the correct sign of the ef-
between the applied and coherency (solute) fect for the other five systems investigated,
stress fields. Hillert (1972) indicated how a for which no quantitative data were ob-
quantitative treatment of such an effect tained. We conclude that it is the plastic (or
could be formulated. Sulonen’s result and creep) response of the specimen which is
Hillert’s analysis have been widely quoted more plausibly coupled to the migration
in support of the existence of a solute (co- rates of differently oriented transformation
herency) field in front of the discontinuous interfaces.
transformation front, even at temperatures
where the calculated diffusion distance is
7.5.4 Interface Migration in Multilayers
negligible.
Dryden and Purdy (1990) reconsidered Multilayers have been utilized for criti-
the problem, and included the possible ef- cal experiments since the seminal work of
fect of a volume misfit in the transformed Hilliard and his co-workers (1954). They
region. It was found, in agreement with provide an ideal tool for the investiga-
Hillert, that the sign of the effect could be tion of the thermodynamics and kinetics of
predicted on the basis of an elastic solute heterogeneous systems far from equilib-
field interaction; however, the predicted ef- rium. Much of the work to date has been
fects on interface velocities are too small concerned with the approach to equilib-
by four orders of magnitude. The more rium via bulk diffusion. The recently de-
likely cause of the coupling is found in the veloped techniques for precisely controlled
plastic response of the dead-loaded speci- growth (e.g., molecular beam epitaxy)
men. The reduction in Helmholtz energy of prompt the exploration of the approach to
the loading device results in a virtual force equilibrium via interfacial diffusion and
on the transformation interface, as dis- migration. The theoretical investigations
cussed in Sec. 7.4.3.2. For the case of inter- summarized in this section deal with pos-
faces with normals parallel to the x, y, and sible effects, many of which are still to be
z (tensile) axes, these virtual forces are re- observed.
lated by Several classes of problems can be de-
(7-28)
⎛ 1 + 3W ⎞ Ze veloped, beginning with an A – B multi-
px = py = − p = (1 + 3W ) layer grown on a bicrystalline substrate
⎝ 2 ⎠ z 2
such that a grain boundary penetrates the
where px , py , and pz are forces per unit whole structure and provides a possible
area, Z is the applied tensile stress, e is the fast diffusion path. The simplest case oc-
stress-free strain in the transformed vol- curs when A and B are fully miscible.
ume, and W is an alignment parameter Homogenization can then occur by grain
which accounts for the possibility that the boundary motion, either via a cooperative
lamellae in transformed regions bounded mechanism, or by a “fingering solution” at
by x and y interfaces are aligned with re- the former A/B interfaces as illustrated in
spect to the z axis. Figure 7-30 (Klinger et al., 1997 a). In each
It is found that the virtual forces so de- case, the shape of the moving boundary, as
rived, in conjunction with independently well as its velocity and the concentration
derived grain-boundary mobilities, are ca- profile left in its wake, can be computed in
pable of explaining the sign and magnitude terms of driving forces, interface energies
of Sulonen’s data for Cu – Cd alloys (Fig. and diffusion coefficients.
7.5 Examples 515
Figure 7-30. Illustrating three different possibilities for the discontinuous homogenization of a monophase
multilayer containing a mobile grain boundary, assumed initially to bisect the multilayer. Situations correspond-
ing to (a) the steady state motion of the boundary; (b) an initially sinusoidal instability of the grain boundary,
and (c) a “fingering” instability of the moving boundary, after Klinger et al. (1997 b).
Cottrell, A. H. (1953), Dislocations and Plastic Flow Humphreys, F. J., Hatherly, M. (1995), Recrystalliza-
in Crystals, Oxford: Oxford University Press. tion and Related Annealing Phenomena. Oxford:
Dahmen, U. (1981), Scripta Metall. 15, 77. Pergamon, Elsevier Science Ltd., p. 212.
Delaey, L. (2001), in: Phase Transformations in Ma- Ju, C. P., Fournelle, R. A. (1985), Acta Metall. 33,
terials: Kostorz, G. (Ed.). Weinheim: Wiley-VCH, 71.
Chap. 9. Karlyn, D. A., Cahn, J. W., Cohen, M. (1969), Trans.
Delage, S. (1998), Ph.D. thesis, University of Paris V. TMS-AIME. 245, 197.
den Broeder, F. J. A. (1972), Acta Metall. 20, 319. Kaur, I., Gust, W. (1988), Fundamentals of Grain
Dryden, J. R., Purdy, G. R. (1990), Acta Metall. 38, and Interphase Boundary Diffusion. Stuttgart:
1255. Ziegler Press.
Duly, D., Bréchet, Y. (1994), Acta Metall. Mater. 42, King, A. H. (1987), Internat. Mater. Rev. 32, 173.
3035. Klinger, L., Bréchet, Y., Purdy, G. R. (1996), Acta
Duly, D., Cheynet, M. C., Bréchet, Y. (1994 a) Acta Mater. 45, 5005.
Metall. Mater. 42, 3843. Klinger, L., Bréchet, Y., Duly, D. (1997 a), Scripta
Duly, D., Cheynet, M. C., Bréchet, Y. (1994 b) Acta Mater. 37, 1237.
Metall. Mater. 42, 3855. Klinger, L., Bréchet, Y., Purdy, G. R. (1997 b), Acta
Emeric, E. (1998), Ph.D. thesis, Marseille. Mater. 45, 325.
Faridi, B. A., Crocker, A. G. (1980), Phil. Mag. A 41, Klinger, L., Bréchet, Y., Purdy, G. R. (1997 c), Acta
137. Mater. 45, 4667.
Ferrence, T. G., Balluffi, R. W. (1988), Scripta Klinger, L., Bréchet, Y., Purdy, G. R. (1998), Acta
Metall. 22, 1929. Mater. 46, 2617.
Fonda, R. W., Mangan, M. A., Shiflet, G. J. (1998), Krakow, W., Wetzel, J. T., Smith, D. A. (1986), Phil.
Metall. Mater. Trans. A 29, 2101. Mag. A 53, 739.
Fournelle, R. A., Clark, J. B. (1972), Metall. Trans. LeClaire, A. D. (1986), in: Solute-Defect Interac-
A 3, 2757. tions: Saimoto, S., Purdy, G. R., Kidson, G. V.
Gibbs, J. W. (1906), Scientific Papers, Vol. 1, Ther- (Eds.). Oxford: Pergamon, p. 251.
modynamic. New York: Dover Publications. Levine, H. S., MacCallum, C. J. (1960), J. Appl.
Gust, W. (1984), in: Decomposition of Alloys, The Phys. 31, 595.
Early Stages: Haasen, P., Gerold, V., Wagner, R., Lücke, K., Detert, K. (1957), Acta Metall. 5, 517.
Ashby, M. F. (Eds.). Oxford: Pergamon, p. 221. Mangan, M. A., Shiflet, G. J. (1997), Scripta Mater.
Gust, W., Hintz, B. B., Lucic, R., Predel, B. (1984), 37, 517.
in: Phase Transitions in Solids: Tsakalakos, T. Martin, G., Blackburn, D. A., Adda, Y. (1967), Phys.
(Ed.). New York: Elsevier, p. 513. Stat. Sol. 23, 223.
Gust, W., Mayer, S., Bogel, A., Predel, B. (1985), Martin, G., Athenes, M., Desgranges, C., Nasbur, M.,
J. Physique C 4 46, 537. Soissons, F. (1998), in: Diffusion Mechanisms in
Guttman, M. (1995), J. Phys. C7, 85. Crystalline Solids, Mishin, Y., Cowern, N. E. B.,
Hassner, A. (1974), Kristall. Tech. 9, 1371. Catlow, C. R. N., Farkas, D., Vogl, G. (Eds.). MRS
Hazzledine, P. M., Oldershaw, R. D. (1990), Phil. Symposium Proceedings, vol. 527, Warrendale,
Mag. A 61, 579. PA, USA: MRS, p. 35.
Hillert, M. (1969), in: The Mechanism of Phase Massalski, T. B. (1958), Acta Metall. 6, 243.
Transformations in Crystalline Solids. London: In- Massalski, T. B. (1984), Metall. Trans. 15 A, 421.
stitute of Metals, Monograph No. 33, p. 231. Menon, E. S. K., Plichta, M. R., Aaronson, H. I.
Hillert, M. (1972), Metall. Trans. 3 A, 2729. (1988), Acta Metall. 36, 321.
Hillert, M. (1975 a), in: Lectures on the Theory Meyrick, G. (1976), Scripta Metall. 10, 649.
of Phase Transformations: Aaronson, H. I. (Ed.), Michael, J. R., Williams, D. B. (1986), in: Interface
New York: AIME, p. 1. Migration and Control of Microstructure: Pande,
Hillert, M. (1975 b), Metall. Trans. 6 A, 5. C. S., Smith, D. A., King, A. H., Walter, J. (Eds.).
Hillert, M. (1982), Acta Metall. 30, 1689. Metals Park, OH, USA: ASM, p. 73.
Hillert, M., Lagneborg, R. (1971), J. Mater. Sci. 6, Morral, J. E., Purdy, G. R. (1995), J. Alloys and Com-
208. pounds 220, 132.
Hillert, M., Purdy, G. R. (1978), Acta Metall. 26, Mullins, W. W. (1957), J. Appl. Phys. 28, 333.
333. Pawlowski, A., Ziȩba, P. (1991), Phase Transforma-
Hillert, M., Sundman, B. (1976), Acta Metall. 24, tions Controlled by Diffusion at Moving Boundar-
731. ies, Krakow, Poland: Polish Academy of Sciences.
Hilliard, J. E., Averbach, B. L., Cohen, M. (1954), Perepezko, J. H. (1984), Metall. Trans. 15 A, 437.
Acta Metall. 2, 621. Perovic, A., Purdy, G. R. (1981), Acta Metall. 29,
Hoffman, D. W., Cahn, J. W. (1972), Surface Science 53.
31, 368. Petermann, J., Hornbogen, E. (1968), Z. Metallk. 59,
Hsieh, T. E., Balluffi, R. W. (1989), Acta Metall. 37, 814.
2133. Peterson, N. L. (1983), Internat. Met. Rev. 28, 65.
518 7 Transformations Involving Interfacial Diffusion
Plichta, M. R., Clark, W. A. T., Aaronson, H. I. Sutton, A., Balluffi, R. W. (1995), Interfaces in Crys-
(1984), Metall. Trans. A 15, 42. talline Metals, Oxford: Oxford University Press.
Porter, D. A., Edington, J. W. (1977), Proc. Roy. Soc. Suzuoka, T. (1964), J. Phys. Soc. (Japan) 19, 839.
A 358, 335. Tashiro, K., Purdy, G. R. (1983), Scripta Metall. 17,
Purdy, G. R. (1971), Met. Sci. J. 5, 81. 455.
Purdy, G. R., Lange, N. (1984), in: Decomposition of Tashiro, K., Purdy, G. R. (1987), Scripta Metall. 21,
Alloys, The Early Stages: Haasen, P., Gerold, V., 361.
Wagner, R., Ashby, M. F. (Eds.). Oxford: Per- Tashiro, K., Purdy, G. R. (1989), Metall. Trans. A 20,
gamon, p. 214. 1593.
Razik, N. A., Maksoud, S. A. (1979), Appl. Phys. 19, Treglia, G., Legrand, B. (1998), in: Tight Binding
331. Approach to Computational Materials Science:
Rhee, W.-H., Yoon, D. K. (1989), Acta Metall. 37, Turchi, P. E., Gonis, A., Columbo, L. (Eds.). MRS
221. Symposium Proceedings, vol. 491, Warrendale,
Rigsbee, J. M., Aaronson, H. I. (1979), Acta Metall. PA, USA: MRS, p. 275.
27, 351, 365. Tu, K. N. (1972), Metall. Trans. A 3, 2769.
Robinson, J. T., Peterson, N. L. (1973), Acta Metall. Tu, K. N., Turnbull, D. (1969), Acta Metall. 17,
21, 1181. 1263.
Rottman, C., Voorhees, D. W., Johnson, W. C. Turnbull, D. (1951), Trans. AIME 191, 661.
(1988), Scripta Metall. 22, 293. Verdier, M., Groma, I., Flandin, L., Lendvaii, J.,
Rundman, K. B., Hilliard, J. E. (1967), Acta Metall. Bréchet, Y., Guyot, P. (1997), Scripta Mater. 37,
15, 1025. 449.
Singh, J., Purdy, G. R., Weatherly, G. C. (1985), Vitek, V. (1984), in: Dislocations 1984: Veyssiere, P.,
Metall. Trans. A 16, 1363. Kubin, L., Castaing, J. (Eds.). Paris: Editions du
Smith, C. S. (1948), TMS-AIME 175, 15. CNRS, p. 435.
Smith, D. A. (1986), Proc. Int. Conf. on Solute-De- Wagner, R., Kampmann, R., Vorhees, P. W. (2001),
fect Interactions: Saimoto, S., Purdy, G. R., Kid- in: Phase Transformations in Materials: Kostorz,
son, G. V. (Eds.). Oxford: Pergamon, p. 128. G. (Ed.). Weinheim: Wiley-VCH, Chap. 5.
Solorzano, I. G., Purdy, G. R. (1984), Metall. Trans. Wanatabe, R., Koda, S. (1965), Trans. Natl. Res. Inst.
A 15, 1055. Metals (Japan) 7, 87.
Solorzano, I. G., Purdy, G. R., Weatherly, G. C. Weygand, D. (1998), Ph.D. thesis, Grenoble, INPG.
(1984), Acta Metall. 32, 1709. Weygand, D., Bréchet, Y., Lepinoux, J. (1999), Acta
Stephens, D. E., Purdy, G. R. (1975), Acta Metall. 23, Mater. 47, 961.
1343. Wheeler, A., Boettinger, W. J., McFadden, G. B.
Straumal, B. B., Bokshtein, B. S., Klinger, L. M., (1992), Phys. Rev. A 45, 7425.
Shvindlerman, L. S. (1981), Scripta Metall. 15, Whipple, R. (1954), Phil. Mag. 45, 1225.
1197. Yang, C. F., Sarker, G., Fournelle, R. A. (1988), Acta
Sulonen, M. S. (1964 a), Acta Polytechnica Scandi- Metall. 36, 1511.
navica 28, 5. Zener, C. (1946), Trans. AIME 167, 550.
Sulonen, M. S. (1964 b), Acta Metall. 12, 749. Zhang, W-Z., Purdy, G. R. (1993 a), Phil. Mag. A 68,
Sundquist, B. F. (1973), Metall. Trans. A 4, 1919. 279.
Suresh, V., Gupta, S. (1986), Z. Metallk. 77, 529. Zhang, W-Z., Purdy, G. R. (1993 b), Phil. Mag. A 68,
Sutton, A. (1995), in: Computer Simulation in Mate- 291.
rials Science, NATO ASI vol. 308, Pontikis, V., Zhang, W-Z., Perovic, V., Perovic, A., Weatherly, G.
Kirchner, H., Kubin, L. (Eds.), Dordrecht (Nether- C., Purdy, G. R. (1998), Acta Mater. 46, 3443.
lands): Kluwer, p. 163. Ziȩba, P., Gust, W. (1998), Internat. Mater. Rev. 43,
Sutton, A. P., Balluffi, R. W. (1987), Acta Metall. 35, 70.
2177.
8 Atomic Ordering
Gerhard Inden
atoms on the lattice sites defines the con- by s r = (s 1 , s 2 , …, s r ). The smallest clus-
figuration. ter is a point, the next one a pair, then a
In order to specify a configuration, we triplet, and so on. On an r-site cluster there
need an operator which identifies unequiv- are K r configurations, a much smaller num-
ocally the atomic species on an arbitrary ber than K N. The configurations of the N-
site n. A convenient means of identifica- point system can then be classified into
tion is to associate an integral number with groups with the same number of clusters
each constituent and to define a site opera- N (s r ) with the configuration s r . Instead of
tor s n , which takes these integral values N (s r ), it is preferable to work with the re-
corresponding to the constituent on site n. duced number r (s r ) = N (s r )/Nr , where Nr
Different choices are possible, e.g., is the number of equivalent r-site clusters
contained in the system. These fractions
⎧ K ⎛ K − 1⎞ are called cluster probabilities. They spec-
⎪ 2 ⎝ 2 ⎠ ify the configuration in the r-point cluster
⎧ 0 ⎧1 ⎪
⎪ 1 ⎪2 ⎪ approximation and constitute the most im-
⎪ ⎪ ⎪ 1 portant configurational variables. The ap-
s n = ⎨ 2 or ⎨ 3 or ⎨ (0)
−1 proximation depends on the size of the
⎪ ⎪ ⎪
⎪ K −1 ⎪K ⎪ largest cluster taken into account. For a
⎩ ⎩ ⎪ − K ⎛ − K + 1⎞ three-dimensional (3-dim.) lattice, it is
⎪ 2 ⎝ 2 ⎠
⎩ necessary to include at least one 3-dim.
cluster, otherwise the topological connec-
⎧ t 1 for i = A tion of the clusters for space filling cannot
⎪ t for i = B be taken into account correctly. For in-
or generally ⎨ 2
stance, if only pairs are considered, it is not
⎪
⎩ K for i = K
t possible to distinguish a 3-dim. configura-
tion from a 2-dim. Bethe lattice with the
Any configuration is then specified by the same coordination number.
vector s = (s 1 , s 2 , …, s N ). In total, there So far r defines a particular set of points
are K N different configurations. For binary 1, 2, …, r, and Nr is the number of clusters
alloys, the choice s n = ⎧⎨ B was first sug-
x having the same orientation in space, thus
−
⎩ xA differing from this particular set by a trans-
gested by Flinn (1956). The operator s n is lation in the lattice, or Nr = N. In this in-
sometimes called the spin variable because stance, we speak of oriented clusters and
of its correspondence with the Ising model oriented-cluster probabilities. The N-point
for binary alloys if we take s n = ± 1. Any system usually exhibits more symmetry
function of s n including s n itself is called a elements than translation, and clusters of
configurational variable. different orientation then become equiva-
Because N is a very large number, it is lent. Bringing these clusters together gives
not possible to handle such a large amount Nr > N.
of information. Therefore, we are forced to Suppose that an r-point cluster has been
work with a reduced amount of information selected for the description of the config-
by considering the configurations of much urations. Then the thermodynamic func-
smaller units called clusters. A cluster is tions derived, for example, with CVM, de-
defined by a set of lattice points 1, 2, …, r, pend on the probabilities of this cluster and
and a configuration on this cluster is given also on the probabilities of all subclusters.
8.2 Definition of Atomic Configurations 525
pn(i) is called the site-occupation operator Det M ≠ 0 and so we can invert the ma-
(Clapp and Moss, 1966) and is defined as trix M and define R = (R ik ) = M –1. From
follows: Eq. (8-2) we arrive at
⎧1 if an atom of type i occupies site n K
pn( i ) = ⎨ pn(1) = ∑ R1k s nk − 1
⎩0 otherwise k =1
K
Because i goes from 1 to K we have K op-
erators. Using this defintion we can imme-
pn( 2 ) = ∑ R2 k s nk − 1
k =1
diately write the following equations:
K
⎧ K
⎫ ⎛ ⎞ pn( K ) = ∑ RK k s nk − 1
⎪1 = ∑ pn
(i )
⎪ ⎜ 1 ⎟ k =1 (8-3)
⎪ i = 1K ⎪ ⎜ ⎟
⎪ ⎪ ⎜ ⎛ pn(1) ⎞ ⎛ 1 ⎞
⎪s n = ∑ t i pn
(i )
⎪ sn ⎟
i =1 ⎜ ⎟ ⎜ pn( 2 ) ⎟ ⎜ sn ⎟
⎪ ⎪ ⎜
⎨ 2 K
⎬ or ⎜ ⎟ or
⎜ pn( 3) ⎟⎟= R ⎜ s n2 ⎟
⎜ ⎟
s = ∑ t 2 (i )
p ⎜ s 2 ⎟
⎪ n i =1 i n ⎪ ⎜
n
⎟ ⎜⎜ ⎟⎟ ⎜ ⎟
⎪ ⎪ ⎜ ⎝ pn( K ) ⎠ ⎝ s nK − 1 ⎠
⎪ ⎪ ⎟
⎜ ⎟
⎪ K −1 K
K −1 (i ) ⎪ ⎜ K −1⎟
⎪s n = ∑ t i
This is the relation that connects the site-
pn ⎪ ⎜s n ⎟
⎩ i =1 ⎭ ⎝ ⎠ (8-2) occupation operators with the site opera-
tors. Because all the elements of the first
⎛ 1 1 1 … 1 ⎞ ⎛ pn ⎞ (1)
row of M are unity, the inverse matrix R
⎜ t1 t2 t 3 … t K ⎟ ⎜ pn( 2 ) ⎟
= ⎜ t12 t 32 … t K2 ⎟ ⎜ pn( 3) ⎟
exhibits the following properties: the ele-
t 22
⎜ ⎟ ⎜ ⎟ ments of the first column of R add up to
⎜⎜ K⋅− 1 K⋅− 1 K⋅− 1 … K⋅− 1 ⎟⎟ ⎜⎜ ⋅ ⎟⎟ unity and the elements of all other columns
⎝ t1 t2 t3 … t K ⎠ ⎝ pn( K ) ⎠ K
2 2 1 ⎛ 0 1 1⎞
ij
r nm = ∑ ∑ Rih Rjk 〈s nh −1 s mk−1 〉 R= ⎜ 2 0 − 2⎟
2 ⎜ ⎟
k =1 h =1 ⎝ 0 − 1 1⎠
or
According to Eqs. (8-2) and (8-5), we ob-
⎛ rnm
AA
⎞ ⎛ rnm
11
⎞ tain
AB ⎟
⎜ rnm 12 ⎟
⎜ rnm
⎜ BA ⎟ = ⎜ 21 ⎟ (8-7) 〈s n 〉 = rn(1) − rn( 3)
⎜ rnm ⎟ ⎜ rnm ⎟ 〈s n2 〉 = rn(1) + rn( 3)
⎜ BB ⎟ ⎜ 22 ⎟
⎝ rnm ⎠ ⎝ rnm ⎠
rnA = rn(1) = 12 ( 〈s n 〉 + 〈s n2 〉 ) (8-8)
⎛ 1 1 1 1⎞ ⎛ 1 ⎞
rnB = rn( 2 ) = (2 −
1
2 〈s n2 〉 )
1 ⎜ 1 1 − 1 − 1⎟ ⎜ 〈s n 〉 ⎟ 2
= ⎜ ⎟⎜ ⎟
4 ⎜ 1 − 1 1 − 1⎟ ⎜ 〈s m 〉 ⎟ rnC = rn( 3) = ( − 〈s n 〉 + 〈s n2 〉 )
1
2
⎝ 1 − 1 − 1 1⎠ ⎝ 〈s n s m 〉⎠
For the pair probabilities, we obtain, ac-
cording to Eq. (8-5),
⎛ rnm
AA
⎞ ⎛ rnm11
⎞ ⎛0 0 0 0 1 1 0 1 1⎞ ⎛ 1 ⎞
⎜ rnm ⎟ ⎜ rnm
AB 12 ⎟ ⎜0 2 2 0 0 0 0 − 2 − 2 ⎟ ⎜ 〈s n 〉 ⎟
⎜ AC ⎟ ⎜ 13 ⎟ ⎜ ⎟ ⎜ ⎟
1 ⎟ ⎜ 〈s n 〉 ⎟
2
⎜ rnm ⎟ ⎜ rnm ⎟ ⎜0 0 0 0 −1 −1 0 1
⎜ rnm
BA ⎟ ⎜ rnm
21 ⎟ ⎜0 0 0 2 0 −2 2 0 − 2 ⎟ ⎜ 〈s m 〉 ⎟
⎜ BB ⎟ ⎜ 22 ⎟ 1 ⎜ ⎟ ⎜ ⎟
⎜ rnm ⎟ = ⎜ rnm ⎟ = 4 ⎜ 4 0 −4 0 0 0 −4 0 4 ⎟ ⎜ 〈s n s m 〉 ⎟ (8-9)
⎜ rnm
BC ⎟ ⎜ rnm
23 ⎟ ⎜0 0 0 −2 0 2 2 0 − 2 ⎟ ⎜ 〈s n s m 〉⎟
2
⎜ CA ⎟ ⎜ 31 ⎟ ⎜ ⎟
1 ⎟ ⎜ 〈s m 〉 ⎟
2
⎜ rnm ⎟ ⎜ rnm ⎟ ⎜0 0 0 0 −1 1 0 −1
⎜ 2 ⎟
⎜ rnm
CB
⎟ ⎜ rnm32
⎟ ⎜0 −2 2 0 0 0 0 2 − 2 ⎟ ⎜ 〈s n s m 〉 ⎟
⎜ CC ⎟ ⎜ 33 ⎟ ⎜ ⎟ ⎜ 2 2 ⎟
⎝ rnm ⎠ ⎝ rnm ⎠ ⎝0 0 0 0 1 −1 0 −1 1 ⎠ ⎝ 〈s n s m 〉⎠
8.2 Definition of Atomic Configurations 529
Nc = ⎛ ⎞ ( K − 1) + ⎛ ⎞ ( K − 1)2 + …
r r
⎝ 1⎠ ⎝ 2⎠
+ ⎛ ⎞ ( K − 1)r = ( K − 1 + 1)r − 1 = K r − 1
r Figure 8-1. Subdivision of the face-centered cubic
⎝ r⎠ (f.c.c.) unit cell into four simple cubic sublattices for
a characterization of the superstructures L12 , L10 and
This is the dimension of the configura- P4/mmm. The occupations of the sublattices for
tional space, and the whole set of cluster these structures are given in Table 8-1.
530 8 Atomic Ordering
Table 8-1. F.c.c. and b.c.c. superstructures and values of their lro parameters.
difference between the point correlation In practice, the values of the pair correla-
functions or linear combinations between tion functions for the sro states are ob-
them (e.g., the same combinations as those tained from diffraction experiments. For
entering the structure factor for X-ray dif- this purpose, we usually introduce the
fraction). These linear combinations of the Cowley –Warren sro parameters a nij– m
point correlation functions are called lro (Cowley, 1950), which are directly related
parameters. In the case of the b.c.c. struc- to the measured intensities. They are de-
ture, we usually define the lro parameters fined as the deviation from the random
state:
x1 = ·s 1 Ò + ·s 2 Ò – ·s 3 Ò – ·s 4 Ò
x2 = ·s 1 Ò – ·s 2 Ò r nij− m = r nm
ij ji
+ r nm
x3 = ·s 3 Ò – ·s 4 Ò = [r n( i ) r m( j ) + r n( j ) r m( i ) ] (1 − anij− m )
Table 8-1 lists the b.c.c. superstructures, The index n – m stands for the neighbor dis-
together with the values of the parameters. tance between the positions n and m. Tak-
If there is no long-range order, then all ing again the t i values as before (i.e., ± 1
the positions are equivalent and the point for the binary case and 1, 0, – 1 for the ter-
probability takes one value r n(i) = x i . Taking nary case), we obtain the isotropic Cow-
ley –Warren sro parameters:
〈s n s m 〉 − 〈s n 〉 〈s m 〉
binary alloys: an − m =
1 − 〈s n 〉 〈s m 〉
〈s n s m2 〉 + 〈s n2 s m 〉 + 2 〈s n2 s m2 〉 − 4 xA (1 − xB )
ternary alloys: anAB
−m =
4 xA x B
− 〈s n2 s m2 〉 + 〈s n s m 〉 + 4 xA xC
anAC
−m =
4 xA x C
− 〈s n2 s m 〉 − 〈s n s m2 〉 + 2 〈s n2 s m2 〉 − 4 x C (1 − xB )
anBC
−m =
4 x B xC
range, the numerical values do not define Adopting again the values t i = ± 1, which
actually existing atomic configurations. were already used in the examples of Sec.
Therefore, this restricted range of values 8.2, the variables can both vary in the inter-
has been called the existence domain, or val [– 1, 1] (see Fig. 8-3): x1 takes the value
configuration polyhedron (Kudo and Kat- 1 for pure A and – 1 for pure B, x2 takes 1
sura, 1976). The dimension of the configu- for each pure component and – 1 for an al-
rational space depends either on the num- loy composed only of A– B pairs. These
ber of correlations which are taken into ac- ranges, however, are not fully accessible,
count or, equivalently, on the size of the for example because of the obvious con-
largest cluster. In the following sections the straints
existence domains will be investigated for
r1AA
2 7 0, r1BB
2 7 0, r1AB
2 70 (8-11)
some simple examples. The method ap-
plied closely follows the procedure devel- The constraints in Eq. (8-11) yield the fol-
oped by Finel (1987). For the sake of sim- lowing inequalities:
plicity, the analysis will be restricted to bi-
r1AA
2 7 0 fi 1 + 2 x1 + x2 7 0
nary alloys. For multicomponent systems,
AB
the procedure is analogous but has to be r1 2 7 0 fi 1 – x2 7 0 (8-12)
BB
carried out on a computer. r1 2 7 0 fi 1 – 2 x 1 + x 2 7 0
Existence domains are most useful for an
analysis of ground states. This will be shown
in Sec. 8.4. The ground-state analysis will be
limited to a finite range of interactions (e.g.
pair interactions between first and second
nearest neighbors). Therefore, we are partic-
ularly interested in the part of the existence
domain that is the subspace of correlation
functions required for a treatment with fi-
nite-range interactions. This existence do-
main in the configurational subspace will
be investigated in detail in this section.
If the relations in Eq. (8-12) are taken as new variable x3 = ·s 1 s 2 s 3 Ò must be intro-
equalities, they define a triangle in the duced, and the configurational space is
space (x1 , x2 ), as shown in Fig. 8-3. Any now three-dimensional. The constraints for
point inside the triangle abg fulfills these the triplet probabilities are
relations. However, these relations are (8-14)
AAA
not sufficient to define the existence do- r12 3 7 0 fi ⎛ 1 3 3 1⎞ ⎛ 1 ⎞
main. In fact, the vertex g, for instance,
AAB
r12 3 7 0 fi ⎜ 1 1 − 1 − 1⎟ ⎜ x1 ⎟
r12 3 7 0 fi ⎜⎜ 1 − 1 − 1 1⎟⎟ ⎜⎜ x2 ⎟⎟
70
corresponds to r1AA BB
2 = r1 2 = 0 (introduce
ABB
AAAA
r12 34 7 0 fi ⎛1 4 6 4 1⎞ ⎛ 1 ⎞
AAAB
r12 3 4 7 0 fi ⎜1 2 0 − 2 − 1⎟ ⎜ x1 ⎟
⎜ ⎟⎜ ⎟
⎜1 0 − 2
AABB
r12 34 7 0 fi 0 1⎟ ⎜ x2 ⎟ 7 0 (8-15)
⎜1 − 2 2 − 1⎟ ⎜ x3 ⎟
ABBB
r12 34 7 0 fi 0
⎜ ⎟⎜ ⎟
⎝1 − 4 6 − 4 1⎠ ⎝ x 4 ⎠
BBBB
r12 3 4 7 0 fi
534 8 Atomic Ordering
Table 8-2. Coordinates of the vertices shown in Fig. 8.3.2 F.C.C. Structure, First and
8-3. Second Neighbor Interactions
Vertex x1 x2 x3
First and second neighbor interactions
a 1 1 1 correspond more closely to the situation
h 1/3 – 1/3 –1 encountered in real alloys (Inden, 1977 a)
f – 1/3 – 1/3 1 than the previous case does. The internal
b –1 1 –1 energy contains one more term, the second-
nearest neighbor interaction. Consequently
we will finally be interested in the exis-
tence domain within the subspace of point
Table 8-3. Coordinates of the vertices of the config- and first and second neighbor pair correla-
uration polyhedron shown in Fig. 8-3. tion functions. The interesting part of the
existence domain will be a polyhedron in
Configuration Correlation functions
this three-dimensional space.
Vertex x1 x2 x3 x4 The analysis begins with the octahedron
r = 234567 in Fig. 8-4, which is defined as
Pure A a 1 1 1 1 a basic cluster because it includes both
L12 (A3B) e 1/2 0 – 1/2 –1 types of interaction. For this cluster, nine
L10 (AB) d 0 – 1/3 0 1 correlation functions have to be intro-
L12 (AB3 ) c – 1/2 0 1/2 –1 duced:
Pure B b –1 1 –1 1
x1 = ·s 2 Ò x6 = ·s 2 s 3 s 4 s 7 Ò
x2 = ·s 2 s 3 Ò x7 = ·s 3 s 4 s 5 s 6 Ò
x3 = ·s 2 s 7 Ò x8 = ·s 2 s 3 s 4 s 5 s 7 Ò
four-dimensional space (x1 , …, x 4 ), see x4 = ·s 2 s 3 s 4 Ò x9 = ·s 2 s 3 s 4 s 5 s 6 s 7 Ò
Table 8-3. The vertices abcde in Fig. 8-3 of x5 = ·s 2 s 3 s 7 Ò
the two-dimensional polygon in the sub-
space (x1 , x2 ) can immediately be taken x1 again corresponds to xA – x B . The exis-
from Table 8-3. tence domain will be a polyhedron in 9-
dim. space (x1 , …, x 9 ), which is defined by
the consistency relations for the various
configurations on the octahedron cluster:
i j… k
AAAAAA : ⎛ 1 6 12 3 8 12 12 3 6 1⎞ ⎛ 1⎞
BAAAAA : ⎜1 4 4 1 0 0 −4 −1 −4 − 1⎟ ⎜ x1 ⎟
BBAAAA : ⎜ 1 2 0 −1 0 − 4 0 −1 2 1⎟ ⎜ x2 ⎟
BAAAAB : ⎜1 2 − 4 3 − 8 4 −4 3 2 1⎟ ⎜x ⎟
⎜ ⎟ ⎜ 3⎟
BBBAAA : i j …k 1 ⎜ 1 0 0 − 3 0 0 0 3 0 − 1⎟ ⎜ x 4 ⎟ 70
r =
BBAAAB : 23…7 2 6 ⎜ 1 0 − 4 1 0 0 4 −1 0 − 1⎟ ⎜ x5 ⎟
BBBAAB : ⎜ 1 −2 −4 3 8 −4 −4 3 −2 1⎟ ⎜ x6 ⎟
ABBBBA : ⎜1 − 2 0 −1 0 4 0 −1 −2 1⎟ ⎜x ⎟
⎜ ⎟ ⎜ 7⎟
BBBBAB : ⎜ 1 − 4 4 1 0 0 −4 −1 4 − 1⎟ ⎜ x8 ⎟
BBBBBB : ⎝ 1 − 6 12 3 − 8 − 12 12 3 −6 1⎠ ⎝ x9 ⎠
536 8 Atomic Ordering
x1 x2 x3
Pure A (A1) 1 1 1
A5B, AB5 ±2/3 1/3 1/3
A3B, AB3 (L12 ) ±1/2 0 1
A3B, AB3 (D022 ) ±1/2 0 2/3
A2B, AB2 (Pt2Mo) ±1/3 – 1/9 1/9
A2B, AB2 ±1/3 0 – 1/3
AB (L10 ) 0 – 1/3 1
AB (L11) 0 0 –1
A 2B 2 0 – 1/3 1/3
Pure B (A1) –1 1 1
Figure 8-7. Intersection of the two configuration
polyhedra corresponding to the tetrahedron and the
octahedron, Fig. 8-4 and Fig. 8-5. All vertices of this
polyhedron correspond to existing configurations on statement made in the previous paragraph
the f.c.c. lattice. The corresponding superstructures that if we reduce the interaction range to
are indicated. Notice the symmetry with respect to
the stoichiometric composition AB. The vertical sec-
nearest neighbors and keep the two basic
tions represent the existence domains for a fixed al- clusters, the projected existence domain in
loy composition, here AB and A3B. the space (x1 , x2 ) does not change due to
the introduction of the octahedron, a larger
cluster than the tetrahedron.
hedron, we will have to intersect the two
polyhedra. The result of the intersection of
8.3.3 B.C.C. Structure, First and
Fig. 8-5 and Fig. 8-6 is shown in Fig. 8-7,
Second Neighbor Interactions
and the coordinates of the vertices are
given in Table 8-4. It is easy to show that The treatment of the b.c.c. structure is
all the vertices of this polyhedron corre- much simpler than that of the f.c.c. struc-
spond to existing atomic configurations. ture because an irregular tetrahedron as
Consequently, in the f.c.c. structure with a basic cluster already contains first and
first and second nearest-neighbor interac- second neighbor distances, as shown in
tions, we have to introduce at least the reg- Fig. 8-2. With this cluster, we must in-
ular tetrahedron and the octahedron as ba- troduce the variable x1 = ·s 1 Ò, x2 = ·s 1 s 3 Ò,
sic clusters. Fig. 8-7 also illustrates the x3 = ·s 1 s 2 Ò, x 4 = ·s 1 s 2 s 3 Ò, x5 = ·s 1 s 2 s 3 s 4 Ò.
This consistency relations are
ijkl
AAAA : ⎛ 1 4 4 2 4 1⎞ ⎛ 1 ⎞
BAAA : ⎜ 1 2 0 0 − 2 − 1⎟ ⎜ x1 ⎟
BABA : i j k l 1 ⎜ 1 0 0 − 2 0 1⎟ ⎜ x2 ⎟
r1 2 3 4 = 4 ⎜ ⎟ ⎜ ⎟ 70
BBAA : 2 ⎜ 1 0 − 4 2 0 1⎟ ⎜ x3 ⎟
BBBA : ⎜ 1 − 2 0 0 2 − 1⎟ ⎜ x 4 ⎟
BBBB : ⎝ 1 − 4 4 2 − 4 1⎠ ⎝ x5 ⎠
538 8 Atomic Ordering
Pure A (A2) 1 1 1 1 1
The concept of pair interactions has a
A3B, AB3 (D03) ±1/2 0 0 ⫿1/2 –1
long tradition and has been supported by
AB (B2) 0 –1 1 0 1
the theoretical work of Gautier (1984),
AB (B32) 0 0 –1 0 1
Bieber and Gautier (1984 a, b, 1986, 1987),
Pure B (A2) –1 1 1 –1 1
and Turchi et al. (1991 a, b), who came to
the conclusion that pair interactions play a
8.4 Ground States 539
dominant role in alloys, while higher-order from sro by energetic arguments. In fact,
cluster interactions are of minor impor- lro results from a topological constraint.
tance. This appears to be a good approxi- Beyond a critical number of unlike bonds,
mation for the class of alloys with negli- it is no longer possible to arrange these
gible relaxation effects. Furthermore, pair bonds on a lattice with the constraint to
correlations can be determined by diffrac- maintain the equivalence of all lattice
tion experiments (see, e.g., Schönfeld, points.
1999). In Eq. (8-17), no reference state is speci-
Grouping together all the energy contri- fied. We will refer the energy to the me-
butions from positions n m with the same chanical mixture of the pure components
distance k, we obtain the following results in the same crystal structure as the alloys
for the internal energy: and obtain the following expression from
L Eq. (8-17):
∑ ∑ ⎧⎨ z ( k ) N (n ) ∑ 〈 pn( i ) pn( kj ) 〉 Vij( k )
1
U˜ =
⎩2 1
n =1 k i, j
DU˜ = U˜ − N ∑ z(k ) (8-18)
⎫ 2
+N ∑ 〈 pn( i ) 〉 m i ⎬
(n ) k
i ⎭ × ( xA VAA +
(k ) (k )
xB VBB )− N ∑ mi0 r0( i )
L i = A, B
1
= ∑ ∑ Vij( k ) ∑ zn( k ) N (n ) rnijn k N N
= ( DmA + DmB ) + ( DmA − DmB ) 〈s 0 〉
2 k i, j n =1
2 8
L
+ ∑ mi N N
∑ N (n ) rn( i ) − ∑ z ( k ) W ( k ) + ∑ z ( k ) W ( k ) 〈s 0 s k 〉
i n =1 8 k 8 k
where n k defines a point in the k-th neigh-
where D m i = m i – m 0i and W (k) = – 2 VAB (k)
+
bor shell of an arbitrary position in sublat- (k) (k) (k)
VAA + VBB . The W are called exchange-
tice n taken as the origin, and z(k)
n is the cor-
energy parameters. They take positive val-
responding coordination number. We ob-
ues for an ordering tendency in the k-th
tain
shell and negative values for a separation
L
tendency. In order/disorder problems, we
∑ zn( k ) N (n ) rnijn k = z ( k ) N r ij0 k are not usually interested in the size of the
n =1
system, and the quantity DŨ/N can then be
and treated. If we fix temperature and pressure
L and consider only equilibrium states, the
∑ N (n ) rni = N r (0i ) number of independent chemical potentials
n =1
is reduced by one. For convenience we often
where position 0 stands for any position in take D mA + D m B = 0 and define an effective
the crystal taken as the origin. Therefore, chemical potential m * = –12 (D mA – D m B ).
the energy is Finding the ground states is equivalent
(8-17)
1 to finding the minimum of Ẽ, which is de-
U˜ = N ∑ ∑ z ( k ) Vij( k ) r ij0 k + N ∑ mi r (0i ) fined as the part of the internal energy that
2 k ij i
depends only on the configurational vari-
This expression contains only the isotropic ables:
pair probabilities because we assumed k +1
DU˜ 1
isotropy for the pair interactions. Conse- E˜ = + ∑ z ( k ) W ( k ) = ∑ xs Fs (8-19)
quently, it is not possible to distinguish lro N 8 k s =1
540 8 Atomic Ordering
Figure 8-9. Lines of constant grand canonical energy E˜ (Eq. (8-20)) for given sets of values F1 = m * =
–12 ( m A – m B ) and F2 = –34 W (1) in the case of the f.c.c. lattice and nearest-neighbor interactions. The configuration
polygon is that of Fig. 8-3. Extrema of the energy correspond to lines passing through the vertices of the poly-
gon. Depending on the sign of F1 and F2 the line through a vertex corresponds to a minimum or a maximum.
The vertices thus define the ground states. For particular F1 , F2 values the lines coincide with the sides of the
polygon. For these values, the ground state is degenerate.
These lines are shown in Fig. 8-9. If we ideal solution. The lines Ẽ = const. are par-
consider an alloy with ordering tendency allel to the x2-axis in this case, because n is
(W (1) > 0 ¤ F 2 > 0), the value Ẽ min is parallel to the x1-axis. Eq. (8-20) now re-
reached for the most negative x2 value, duces to
which is x2 = – 1/3. This point corresponds
F1 x 1 = Ẽ
to the L10 structure AB. If we consider an
alloy with separation tendency (W (1) < 0 For a positive value of F1 = m * =
¤ F 2 < 0), the value Ẽ min is reached for the –12 (D mA – D m B ), the most negative value
most positive x2 value, which is x2 = 1. In of Ẽ is obtained for x1 = (xA – x B ) = – 1, i.e.,
this case, the line Ẽ = Ẽ min coincides with a corresponding to pure B, and a negative
side of the configuration polygon. This line value of F1 corresponds to pure A.
joins the vertices corresponding to the pure The procedure outlined above can be re-
components A and B. The ground state is peated for any direction of the vector F ,
degenerate: it could be pure A and pure B see Fig. 8-9. As long as the lines of con-
in any proportion juxtaposed to each other, stant Ẽ pass through a vertex of the poly-
or any other arrangement of A and B gon, a unique ground state can be defined.
which, in the limit of N Æ ∞, does not If these lines coincide with the sides of the
form A– B bonds (e.g., constant number of polygon, the ground state is degenerate. It
2-dim. boundaries). could then be any mixture of the two states
Next, we may consider the case corresponding to the two vertices defining
12
F 2 = ––8
W (1) = 0, which corresponds to an the side of the polygon. It was mentioned
542 8 Atomic Ordering
before that the proportion of the two states When higher-order interactions are in-
differing in composition is not defined volved, as in the next paragraph, the deter-
because in these cases of degeneracy the mination of the ground-state diagram is not
overall composition of the system is not that simple, and it is useful to have a more
defined by a fixed value of the chemical analytical method for its determination.
potential. We can say, however, that the Such a method will now be introduced and,
state is two-phase. for the purpose of illustration, will be ap-
Finally, we can determine a so-called plied to the present very simple case.
ground-state diagram, which is shown in The ground-state diagram displays the
Fig. 8-10. It is worth noting that the single- ranges of F1 and F 2 values for which a
phase fields are sectors in the (F1, F 2 ) given state, e.g., L10 , is stable. We thus
plane limited by lines originating from the interpret Eq. (8-19) inversely to that of Eq.
origin. It is thus sufficient to know the di- (8-20): given a vertex in the configura-
rections of those lines. Furthermore, the di- tional space, e.g., vertex x v = (– 1 3, 0) cor-
agram is symmetric with respect to F1 = 0. responding to L10 , we define a unit vector
n* = x v /| x v |. Eq. (8-19) can then be rewrit-
ten as
(F |n*) = Ẽ /| x v | = d* (8-21)
In a method similar to that used with Eq.
(8-20), we interpret Eq. (8-21) as the equa-
tion of a line in the space (F1, F 2 ) with a
distance d* = Ẽ /| x v | from the origin. If we
fix the value of Ẽ, then all vectors F ful-
filling Eq. (8-21) must end in the line with
the normal n* and with distance d*. In Fig.
8-10, the result is shown for Ẽ = – 1 (energy
units). It shows that the vectors fulfilling
Eq. (8-21) describe a polygon. The coordi-
nates of the vertices of this polygon are
given in Table 8-6. This polygon is the dual
polygon to the configuration polygon
(Finel, 1987), and the space (F1, F 2 ) is
called the dual space. A face of the dual
Figure 8-10. Ground-state diagram and dual poly- polygon corresponds to each vertex of the
gon (to Fig. 8-9) for the case of an f.c.c. lattice with configuration polygon in the direct space
nearest-neighbor interactions. The axes are
(x1 , x2 ). All vectors F pointing to the same
F1 = m * = –12 ( m A – m B ) and F2 = –34 W (1). The solid lines
are radius vectors which subdivide the plane into re- face of the polygon define alloys that ex-
gions with a defined ground state. The directions of hibit the same ground state given by the as-
these vectors are defined by the vertices of the dual sociated vertex of the configuration poly-
polygon to Fig. 8-3. It corresponds to a fixed value of gon. If we vary Ẽ, then the polygon shrinks
the energy, here E˜ = – 1 (energy units), and is ob-
or blows up, respectively, and the vertices
tained from Eq. (8-21). The dual polygon already
contains complete information on the ground-state of the dual polygon vary along the lines
diagram. For F1 , F2 values corresponding to these that define the two-phase states in Fig. 8-
lines, the ground state is degenerate. 10. It is easy to verify that these lines are
8.4 Ground States 543
12
vertical to the faces of the configuration Let us first consider F 3 = –– 8
W (2) = 0.
polygon in Fig. 8-3. The normal vector n corresponding to the
The generalization of these results to planes of constant energy Ẽ is perpendicu-
higher dimensions is straightforward. A dual lar to the axis x3 in this case. If we consider
polyhedron defined by Eq. (8-21) is asso- an alloy with an ordering tendency in the
12
ciated with the (direct) configuration poly- first shell, F 2 = –– 8
W (1) > 0, then n is
hedron in such a way that each vertex of the oriented in the half space x2 > 0, and the
direct polyhedron corresponds to a face of the planes corresponding to Ẽ min are those
dual polyhedron, and vice versa. In order to passing through the vertices pure A, L12
get the vertices of the dual polyhedron, we with compositions A3B or AB3 , L10 with
have to solve Eq. (8-21) simultaneously for composition AB, and pure B, depending on
the number of vertices of the direct polyhe- the value of x1 = m * = –12 (D mA – D m B ). This
dron required to define the (hyper)plane. is in accordance with the results already
The application of this in three dimensions obtained in Sec. 8.4.3 (see Fig. 8-10).
will be shown in the next two paragraphs. However, an additional result can be de-
duced from Fig. 8-7: the energy planes pass
simultaneously through the vertices corre-
8.4.1.3 F.C.C. Structure, First and
sponding to the structures D022 and L12
Second Neighbor Interactions
(both either A3B or AB3 ), as well as L10
With first and second neighbor interac- and A2B2 . The ground states are thus de-
tions in the f.c.c. structure, the energy Ẽ generate in these instances.
depends on one more configurational pa- The complete ground-state diagram is
rameter, the second neighbor pair corre- obtained from the determination of the dual
AB
lation function x3 = ·s 0 s 2 Ò = 1 – 4 r 02 (see polyhedron associated with the existence
Eq. (8-7)). The relevant existence domain domain of Fig. 8-7. This dual polyhedron is
is now that shown in Fig. 8-7. We can again shown in Fig. 8-11. All F vectors pointing
proceed as previously. to (or piercing) a given face of the dual
Figure 8-11. Dual polyhedron (to Fig. 8-7) for the case of an f.c.c. lattice with first and second neighbor inter-
actions. The axes are F1 = m * = –12 ( m A – m B ), F2 = –34 W (1), and F3 = –34 W (2). The dual polyhedron defines the
ground-state diagram and corresponds to a constant energy E˜ = – 1 (energy units). All radius vectors pointing to
the same face correspond to the same ground state, which is labeled for each face, except the top, which is L1 1 .
Vectors coinciding with the boundaries of the faces correspond to degenerate ground states.
544 8 Atomic Ordering
polyhedron define alloys with the same coordinates of the dual vertices are given in
ground state, which is defined by the corre- Table 8-6. They have been derived by
sponding vertex of the configuration poly- inserting the vertices in Table 8-3 into Eq.
hedron. The ground-state diagram thus di- (8-21) for Ẽ min = – 1.
vides the dual space into (hyper)cones
formed by the vectors from the origin to the
8.4.1.4 B.C.C. Structure, First and
borderlines of the faces. Each cone defines
Second Neighbor Interactions
a ground state. Due to graphical limita-
tions, those cones are not shown in Fig. In the first and second neighbor interac-
8-11, but the ground states are labeled tions of the b.c.c. structure, the energy Ẽ
at each face of the dual polyhedron. The depends on the same variables (x1 , x2 , x3 )
as in the preceding example, and the rele-
vant existence domain is that shown in Fig.
Table 8-6. Coordinates of the dual polyhedra shown 8-8. We can now construct the dual polyhe-
in Figs. 8-10, 8-11 and 8-12. dron, solving Eq. (8-21) for Ẽ = – 1 and us-
Tetra- Tetrahedron – Tetrahedron
ing the vertex coodinates in Table 8-4. The
hedron Octahedron b.c.c. coordinates of the dual vertices are given in
f.c.c. f.c.c. Table 8-6, and the dual polyhedron is
shown in Fig. 8-12.
F1 F2 F1 F2 F3 F1 F2 F3
Figure 8-12. Dual polyhedron (to Fig. 8-8) for the case of an b.c.c. lattice with first and second neighbor inter-
actions. The axes are F1 = m * = –12 ( m A – m B ), F2 = W (1), and F3 = –34 W (2). The dual polyhedron defines the
ground-state diagram and corresponds to a constant energy E˜ = – 1 (energy units). All radius vectors pointing to
the same face of the polyhedron correspond to the same ground state, which is labeled for each face, except the
top one, which is B32.
8.4 Ground States 545
·s 1Ò 1 1 1 1
·s 2 Ò 1 1 1 1– 4 xB
·s 3 Ò 1– 4 xB 1 3 – 8 xB 1
·s 4 Ò 1– 4 xB 1– 8 xB –1 1– 4 xB
Figure 8-13. Atomic distribution on four subse- ·s 1 s 3 Ò 1– 4 xB 1 3 – 8 xB 1
quent (0 0 1) planes of the superstructures L11 (AB) ·s 1 s 4 Ò 1– 4 xB 1– 8 xB –1 1– 4 xB
and Amm2 (A3B). The structure Amm2 is composed ·s 2 s 3 Ò 1– 4 xB 1 3 – 8 xB 1– 4 xB
of alternating (0 0 1) planes of pure A and L11 . ·s 2 s 4 Ò 1– 4 xB 1– 8 xB –1 (1– 4 xB )2
·s 1 s 2 Ò 1 1 1 1– 4 xB
·s 3 s 4 Ò (1– 4 xB)2 1– 8 xB – (3 – 8 xB) 1– 4 xB
tions (Kudo and Katsura, 1976; Kanamori, It may be useful to spell out explicitly
1984). the pure configurational part of the Vn . In
the case of pair interactions the Vn with
8.4.2 Effective Cluster Interactions n ≥ 2 are equivalent to the pair exchange
(ECIs) z(n) (n)
energies, i.e., Vn = W . In this case
Up to this point only pair interactions 8
the specific lattice type is only introduced
have been considered with the exchange
in terms of the coordination numbers.
energies W (k) as dummy parameters. The
In the case of higher-order cluster inter-
internal energy DŨ, with reference to the
actions the lattice type is not only specified
pure components in the same crystal struc-
by the coordination numbers of the basic
ture, has been written in terms of an expan-
cluster and its subclusters, but most impor-
sion of these W (k) (Eq. (8-19)). This is a
tantly by the type of clusters they share
special case of the more general cluster ex-
with each other. Subclusters which are not
pansion (CE) of the internal energy DŨ in
shared define no new correlations. They are
terms of the correlation functions:
already taken into account by those of their
DŨ a parent cluster. We could thus expect that the
= ∑ Vn xn (8-24) Vn of such clusters do not show up in the
N 0
expansion, for example triplet correlations
where n refers to subclusters of the basic in the tetrahedron expansion of the f.c.c.
cluster a (n = 1 for point, n = 2 for pair lattice. This, however, is not true because
clusters, etc.) (Sanchez et al., 1984). This the Vk in the expansion, Eq. (8-24), not only
equation is exact if a represents the entire contain terms from the k-point clusters, but
system. It is expected, however, that Eq. also terms from higher clusters (n > k).
(8-24) converges rapidly such that the se- This becomes transparent if the Vn are
ries can be truncated. For a general discus- expressed in terms of the cluster exchange
sion of the method see Laks et al. (1992), energies, defined as
de Fontaine (1994) and Sanchez (1996).
l
The parameters Vn are supposed to em- Wij12……l r = − r Vij …l +
brace all possible energy sources that depend
∑ Vk k …k
k =i
on configuration, including all ground- 1234
(for example, WABBB = – 4VABBB + VAAAA +
state features like electronic energy, long-
3VBBBB). It should be noted that, unlike the
range elastic and Coulombic interactions.
pairs, these cluster-exchange energies do
But they might also contain configuration-
not represent the total energy of the clus-
dependent excitation energies such as those
ters, only successive corrections to the
arising from vibrational and electronic ex-
cluster energies. If total energies were con-
citations when finite temperatures are con-
sidered, the sharing of subclusters would
sidered, making the model interaction en-
have to be taken into account.
ergies temperature dependent. It should be
Considering the f.c.c. lattice and a trun-
appreciated that the parameters Vn repre-
cation at the regular tetrahedron cluster
sent energy contributions, not only from di-
(a = 4), Eq. (8-24) gives:
rect interactions within the cluster, but also
from the interactions of much longer range DU˜ (j )
outside the cluster range. They are thus = V0 + V1 ( xA − xB )
N
called effective cluster interactions (ECIs). + V2 x2(j ) + V3 x3(j ) + V4 x 4(j ) (8-25)
8.4 Ground States 549
lent to saying, in the pair approximation, molar volumes are not constant, and “local
that the W (k) can be used for all configura- relaxation” when atoms deviate from their
tions. Since the only source of error in Eq. ideal lattice positions. Ferreira et al. (1987,
(8-24) is the truncation, it is imperative to 1988) and Wei et al. (1990) have shown
carefully select the truncation. that, if the volumes depend on composi-
The ingredients for the determination of tion but only weakly upon configuration,
the ECIs are values of DŨ(j ), which may the energy of formation can be split into
be obtained from various sources (see two additive terms, the elastic energy re-
e.g. Zunger, 1994). The simplest approach quired to hydrostatically deform the pure
is to take experimental data of the corre- constituents from their equilibrium vol-
sponding quantities (e.g. Inden (1975 a, b, umes to the alloy volume, and the pure
1977 a, b; Oates et al., 1996)). This requires configurational or “chemical” term. The
a truncation at a level imposed by the avail- ECIs can be determined either from a set
ability of data and in most cases does not of as many ordered compounds as there are
provide satisfactory results. unknowns, or from a larger set, in which
The second approach is to derive DŨ(j ) case the ECIs are obtained by a fitting pro-
from first principles calculations, treating cedure (Lu et al., 1991, 1995). It must be
the random alloy by the Coherent Potential emphasized that the ECIs are “effective”
Approximation (CPA) and the electronic and their values depend on the level of
band structure with the Tight-Binding truncation.
method (Gautier et al., 1975; Ducastelle Cluster expansions for systems with
and Gautier, 1976; Treglia and Ducastelle, strong lattice relaxations converge slowly
1980; de Fontaine, 1984; Sigli and San- in real space. This has led to an alternative
chez, 1988; Sluiter and Turchi, 1989 a, b) route, treating the CE in the reciprocal
or with the Korringa, Kohn and Rostoker space (e.g. Laks et al., 1992).
method (Gyorffy and Stocks, 1983; John- Finally, it should be mentioned that pair
son et al., 1990). The approaches include interactions may also be obtained from the
the Generalized Perturbation Method and diffuse scattering of X-rays or neutrons
the Concentration Wave Method. The or- (for a recent review, see Schönfeld, 1999).
dering contributions are treated as pertur- The procedure is the inverse of the CVM
bations of the disordered states. The pertur- (see Sec. 8.5.1) or the MC technique (see
bation method is limited to alloys with sim- Sec. 8.5.3). Instead of calculating equilib-
ilar atomic species and provides essentially rium configurations using given energy pa-
pair interactions. rameters, experimental data of sro are taken
The third approach is the inversion as input in order to obtain the interaction
method of Connolly and Williams (1983) parameters. The methods have been pre-
(Ferreira et al., 1988, 1989; Wei et al., sented by Priem et al. (1989 a, b) for CVM,
1990, 1992; Terakura et al., 1987) . In this and Gerold and Kern (1987) and Livet
approach special atomic configurations (1987) for MC.
like in ordered compounds are selected and
their total energy is obtained from direct
electronic structure calculations.
It has been pointed out by Zunger (1994)
that it is necessary to account for two types
of relaxation, “volume relaxation” when
8.5 Phase Equilibria at Finite Temperatures 551
for the equilibrium at fixed composition, or where Na is the total number of a -clusters
by minimization of the grand potential in a system with N points, and Na (s ) is the
number of a -clusters with a given configu-
W (V, T, m i ) = U – T S – Â m i Ni ration specified by s ; the term {a}Na is a
i
short notation which will be used hereafter
for the equilibrium with exchange of atoms, for abbreviation. The ratio Na (s )/Na will
i.e., for given values of the chemical poten- be indicated by ra(s ).
tials. Eq. (8-29) overestimates the number of
The MC method simulates the configu- arrangements: two overlapping a -clusters
rations in a computer crystal. At a given cannot be permuted independently because
temperature and fixed chemical potentials, they must fit together with their over-
atoms are exchanged with a reservoir of at- lapping units. This becomes obvious if
oms with a probability defined in such a we look at the high temperature limit of Eq.
way that the equilibrium state is reached (8-29), which is W∞ = {1}Na , while the
after a sufficient number of atomic replace- exact expression of the limit of W is given
ments. In the case of fixed composition, at- by the number of configurations of N
oms are selected pairwise and interchanged points: W∞ = {1}N .Therefore, a correction is
with each other according to an equivalent needed in order to obtain the correct high-
probability. This method yields the equilib- temperature limit. Correcting for the points
rium configuration but not the thermody- would correspond to the generalized quasi-
namic functions. On the other hand, we ob- chemical approximation (Yang, 1945; Yang
tain complete and fine-scale information and Li, 1947, 1949). Based on geometrical
about the atomic configurations and the considerations, Kikuchi (1950, 1951) intro-
correlations at large distances, the limita- duced a correction in such a way that the
tion being imposed only by the size of the high temperature limit is not only obtained
computer crystal. Both methods will be ap- for the point cluster but also for all subclus-
plied in this section. ters of a. A short time later Barker (1953)
552 8 Atomic Ordering
arrived at the same result by a more mathe- The same reasoning now needs to be fol-
matical treatment, and the continued inter- lowed with the next subcluster a – 2. For
est in the derivation of the entropy formula this cluster, the number of arrangements of
is manifested in a number of papers, e.g., the basic cluster a, as well as the correction
Hijmanns and de Boer (1955), Burley (1972), term for a – 1, have already been counted.
Sanchez and de Fontaine (1978), Gratias et We can therefore write
al. (1982) and Finel (1987). For a dis- ma − 1 aa − 1
W = ({a}N ) ma aa ({a − 1}N )
cussion of the hierarchy of the cluster ap- m a
proximations, see Schlijper (1983, 1984, × ({a − 2}N ) a − 2 a − 2
1985). with aa – 2 determined from the equation
Following Barker, we define ma as the
number of a -clusters per point: Na = ma N. ma − 2 = ma naa − 2 aa + ma −1 naa −−12 aa −1
Then we can write (in Stirling’s approxi- + ma − 2 aa − 2
mation): Continuing this reasoning leads to
ma
( ma N )! ( N !) a
{a}Na = = W = ∏ ({n }N ) mn an (8-30)
∏ ra(s ) ma N )!
( ∏ ra(s ) N !)ma
((
n =1
s s
= ({a}N ) ma and to the following recursion scheme for
the exponent an , using the identities nnn = 1
We now want to correct for the overlap and aa = 1:
with the first subcluster a – 1, which has
one point less than a. Due to the indepen- (basic) cluster a : ma = ma naa aa
dent variation of the a -clusters, the over- a − 1: ma −1 = ma −1 naa −−11 aa −1
lapping cluster a – 1 has also been counted
a −1 + ma naa −1 aa
({a − 1}N ) ma na times, where naa −1 is the
number of times the a – 1 cluster is con- a − 2: ma − 2 = ma − 2 naa −−22 aa − 2
tained in a. The correct number of times + ma −1 naa −−12 aa −1 + ma naa −1 aa
the a – 1 cluster should be counted, pre-
a −n
suming it were an independent species, is
m
({a − 1}N ) a − 1. Therefore, we must correct
n : mn = ∑ ma − n nan − m aa − m with 1 Ù n Ù a
m =0
Eq. (8-29) as follows:
We thus obtain for the entropy in Stirling’s
ma − 1
({a − 1}N ) approximation
W = ({a}N ) ma a −1
({a − 1}N ) ma na a
S = − kB N ∑ mn an ∑ rn(s ) ln rn(s ) (8-31)
ma − 1 − ma naa − 1
= ({a}N ) ma ({a − 1}N ) n =1 s
Table 8-8. Numerical values of ma , naa ––nm , and the CVM exponents an , for the f.c.c. structure in the tetrahe-
dron – octahedron approximation and for the b.c.c. structure in the irregular tetrahedron approximation.
ma = 1 6 3 12 2 12 8 3 6 1
n a6 = 1
n a5 = 6 1
n a4a = 3 1 1
n a4b = 12 4 0 1
n a4c = 0 0 0 0 1
n a3a = 12 6 4 2 0 1
n a3b = 8 4 0 2 4 0 1
n a2a = 3 2 2 1 0 1 0 1
n a2b = 12 8 4 5 6 2 3 0 1
n a1 = 6 5 4 4 4 3 3 2 2 1
aa = 1 0 0 0 1 0 –1 0 1 –1
ma = 6 12 3 4 1
aa = 1 –1 1 1 –1
[
For the entropy, we thus obtain the fol-
lowing:
= − kB N 6 ∑ r1234
ij k l ij k l
ln r1234
{ij k l}
f.c.c. − 12 ∑ r123
ij k ij k
ln r123 (8-33)
{23 … 7}1N {1234}2N {12}6N {ij k}
S = kB ln
{123}8N {1}1N + 3 ∑ r12
ij ij
ln r12
{ij}
[
ij … k ij … k
= − kB N ∑
∑ r13i k ln r13i k − ∑ r1i ln r1i ]
r23 … 7 ln r 23… 7
{ij … k} +4
{i k} {i}
+2 ∑ ij k l
r1234 ij k l
ln r1234 (8-32)
{ij k l} The entropy equations (8-32) and (8-33)
−8 ∑ ij k
r123 ij k
ln r123 are valid for the sro state. In the lro state,
{ij k} the equivalence of certain clusters is
+ 6 r12
{ij}
∑ ij
ln r12ij
∑− r1i
{i}
ln r1i ] broken. For example, Eq. (8-33) has to be
written in the lro state with four sublattices
(Fig. 8-2):
554 8 Atomic Ordering
Figure 8-15. F.c.c. structure: calculated prototype phase diagrams for the case of nearest-neighbor interactions,
W (1) > 0, W (2) = 0. (a) CVM calculation in the tetrahedron – octahedron approximation (Finel and Ducastelle,
1986), (b) Monte Carlo simulation (Ackermann et al., 1986).
cide quantitatively, but their main features The CVM diagram in the tetrahedron –
agree qualitatively. In particular, they agree octahedron approximation, Fig. 8-15, dif-
as to the existence of a triple point at finite fers only slightly from that previously cal-
temperature, i.e. at culated in the tetrahedron approximation
(van Baal, 1973; Kikuchi, 1974), which
CVM
t tri CVM
= kB Ttri /(W (1)/4) ≈ 1.5
gave a triple point at t tri CVM
≈ 1.6. The
(Finel and Ducastelle, 1986) and at higher-order cluster approximation does
MC
t tri ≈ 0.8 (Ackermann et al., 1986) and 0.9 not lead to a strong shift of the triple point
(Diep et al., 1986). This point gave rise to to lower temperatures, contrary to an ear-
some controversy (discussed by Kikuchi, lier result by Sanchez et al. (1982) who
1986), caused by an earlier MC study
CVM
found t tri ≈ 1.2. The difference results
(Binder, 1980; Binder et al., 1981) which only from certain approximations in the
indicated that the phase boundaries extra- numerical treatment made by Sanchez et
polate to 0 K, and that a triple point does al., which were avoided by Finel (1987). In
not exist. The existence of this triple point Finel’s work, a more sophisticated CVM
is now well confirmed and has been further calculation was also made using the tetra-
corroborated by the studies of Lebowitz et hedron – octahedron for the ordered phases
al. (1985) and Finel (1994). The diagram in and the quadrupole tetrahedron for the sro
Fig. 8-15 replaces that of the earlier MC by states. In that approximation. Finel ob-
Binder. The MCs were not performed at tained complete agreement with the MC
sufficiently low temperatures to detect the work.
P4/mmm phase obtained in the CVM
(called L¢ by Finel, 1984).
8.5 Phase Equilibria at Finite Temperatures 557
8.5.4.2 F.C.C. Structure, First and agree fairly closely. Phase diagrams for
Second Neighbor Interactions other values of the ratio W (2)/W (1) are
given by Mohri et al. (1985) and Binder
In this section, the case of an ordering et al. (1983).
tendency in both shells will be treated first, Fig. 8-17 shows the results for an order-
namely, W (1) > 0 and W (2)/W (1) = 0.25. The ing tendency between first neighbors and a
results of the tetrahedron – octahedron separation tendency between second neigh-
CVM (Sanchez and de Fontaine, 1980) and bors, W (1) > 0, W (2) = – W (1). This situation
of MC simulations (Bond and Ross, 1982) corresponds to the calculation presented
are shown in Fig. 8-16. The two diagrams previously, in Fig. 8-14 c. As expected
from this diagram the low temperature tions on the shape of the miscibility gap is
states are two-phase states and the phase small. The shape obtained differs signifi-
boundaries extrapolate to the pure compo- cantly from the miscibility gap which is
nents and to the stoichiometric ordered usually calculated with the regular solution
phases. A complete series of prototype (i.e., point approx.) model.
diagrams calculated with the tetrahedron –
octahedron CVM is presented by Mohri et 8.5.4.3 B.C.C. Structure, First and
al. (1985). Second Neighbor Interactions
The case of a separation tendency be-
A series of prototype phase diagrams,
tween first and second neighbors has been
calculated with the tetrahedron CVM for
analyzed with the MC technique: Kutner
varying strengths of ordering tendency in
et al. (1982) analyzed the case W (1) < 0,
both neighbor shells, was first presented by
W (2) = 0 with the grand canonical simula-
Golosov and Tolstick (1974, 1975, 1976).
tion, while Gahn et al. (1984) analyzed the
Simultaneously, Kikuchi and van Baal
cases W (2) = 0 and W (2) ≠ 0 with a spe-
(1974) presented a diagram corresponding
cial canonical simulation. The results for
to the ratio W (2)/W (1) = 0.5, which is close
W (2) = 0 are identical in both treatments.
to the situation encountered in Fe – Si and
The resulting miscibility gaps are shown in
Fe – Al. Figs. 8-19 a, b and c display a series
Fig. 8-18 (on a reduced scale in order to
of diagrams calculated with the tetrahedron
show the variation in shape). It is found
CVM and with the MC method (Acker-
that the effect of second neighbor interac-
mann et al., 1989) for different strengths
and signs of the interchange energies. The
results of both methods are in good agree-
ment. In these diagrams, second-order tran-
sitions are indicated by a hachure. These
second-order transitions turn into first-or-
der transitions at tricritical points, and the
topology of the phase boundaries close to
these points exhibit the characteristics de-
rived by Allen and Cahn (1982).
Figure 8-19. B.c.c. structure: calculated prototype phase diagrams for ordering or separation tendencies
between first or second neighbors. The lines correspond to the CVM calculation in the tetrahedron approxima-
tion, the points correspond to the Monte Carlo simulation (Ackermann et al., 1989). The hachure indicates a
second-order transition. (a) W (1) = 2 W (2) > 0, (b) W (1) = – W (2) > 0, (c) W (1) = – W (2) < 0.
560 8 Atomic Ordering
even at T = 0 K. With the progress made in ond neighbor pair interactions and used a
first-principles calculations, it has become power series expansion in composition for
obvious that higher-order pair and multisite the disordered state. No ternary interac-
cluster interactions have to be used in the tions were included. The agreement of the
configurational part of enthalpy, and that it results for the limitrophe binary systems
is mandatory to take lattice and local relax- with experimental data is satisfying only in
ations into account. This will be illustrated a few of the cases studied, and very little is
using the Au – Ni system as an example. presented with respect to ternary isother-
When it comes to T > 0 K the cluster inter- mal sections. McCormack et al. (1996,
actions have to embrace all configuration- 1997) studied the systems Cd – Ag – Au
dependent excitation energies such as arise (1996) and Cu – Al – Mn (1997) using first
from thermal vibrations and electronic ex- and second neighbor pair interactions, but
citations. The inclusion of these effects failed to present phase equilibria for ter-
means that the cluster interactions become nary and limitrophe binary systems that
temperature dependent. There is still some could be checked by experiments. In
disagreement as to whether vibrational the following the two ternary systems
contributions can (Ozoliņš et al., 1998 b; Fe –Ti – Al and Fe – Co – Al will be dis-
van de Walle et al., 1998) or cannot (Craie- cussed. The Fe – Co – Al system will be
vich and Sanchez, 1997; Craievich et al., treated as a magnetic system with spin 1/2
1997 a, b) be included in the cluster expan- given to Fe, Co and Al. This leads in fact to
sion. a six-component system.
However, even the apparently most care- In multicomponent systems the number
ful first-principles calculations (Ozoliņš et of points within a chosen basic cluster may
al., 1998 a) remain insufficiently accurate not be large enough to accommodate all
for technological purposes. For the fore- components. Then the question arises as to
seeable future it seems clear that simpler what extent this is detrimental to the qual-
approaches will continue to be important, ity of the approximation. In order to give
e.g. for technological phase diagram calcu- some idea of this effect, a magnetic spin
lations and also for other applications re- 7/2 system with f.c.c. structure has been
lated to the thermodynamic properties. The analyzed by Schön and Inden (2001) using
systems Fe – Al and Ni – Al will be taken the tetrahedron CVM and Monte Carlo
here as examples for calculations of phase simulations. This spin problem is equiva-
diagrams and of the thermodynamic factor lent to an eight-component alloy problem.
for diffusion. The calculation of phase equilibria and
Surveys of the abundant literature de- thermodynamic properties of multicompo-
scribing applications to mainly binary al- nent systems is of paramount importance
loy systems can be found in the reviews by for technological design and materials de-
Inden and Pitsch (1991) and by de Fontaine velopment. The use of both a cluster ex-
(1994). pansion for the energy and the CVM for
At this stage multicomponent systems configurational entropy introduces severe
present apparently insurmountable prob- computational problems because of the
lems for first-principles calculations. This large number of cluster types that must be
even holds for ternary systems. Rubin and considered in both formalisms. It also has
Finel (1993) studied ternary Ti – Al – X to be emphasized that in practical applica-
(X = W, Nb, Mo) systems with first and sec- tions with coexisting phases of different
8.6 Application to Real Systems 563
8.6.2 The Thermodynamic Factor following parameters are obtained (in units
of Ordered Phases of kB K)
8.6.2.1 The B.C.C. Fe – Al System V0 = – 2190, V1 = 0, V2 = 1680,
Iron aluminides such as Fe3Al and FeAl V3 = 457, V4 = 0, V5 = 52.5
have received considerable attention as or equivalently
candidates for high temperature structural (8-38)
(1) (2)
materials due to their low cost, high W = 1680, W = 740, W 1234
FeAlFeAl =– 140
strength and good oxidation resistance. With these parameters the irregular tetrahe-
Schön and Inden (1998) assessed the sys- dron CVM yields the b.c.c. phase diagram
tem in order to derive ECIs from experi- shown in Fig. 8-25. The calculated critical
mental enthalpies of formation, corrected temperatures match the high temperature
for b.c.c reference states and converted into experimental data. However, at lower tem-
the paramagnetic state: peratures, the tricritical point and the two-
phase region A2 + B2 are not obtained and
j DH j [J/mol of atoms] the agreement with the experimental phase
boundaries A2 + D03 is poor. This is not
Fe3Al (D03) – 18 650 surprising in view of the large relaxation
FeAl (B2) – 27 940 effects, which are to be expected in this
FeAl(B32) – 21 570 system as seen from the variation of lattice
FeAl3(D03) – 18 650 parameter as a function of composition and
of state of order, Fig. 8-26. On the other
Introducing these enthalpies and the values hand, recent experimental data on chemical
of the correlation functions of the configu- potential measurements at 1000 K (Kley-
rations (see Table 8-5) into Eq. (8-24), the kamp and Glasbrenner, 1997) are very well
reproduced by the calculations, see Fig. in ordered alloys. Fig. 8-28 shows this vari-
8-27. ation at three different temperatures. Close
A very important quantity for the treat- to stoichiometric compositions of ordered
ment of diffusion is the thermodynamic phases, F varies dramatically. The amount
factor defined for a binary system A – B as of variation depends on the degree of lro,
reaching up to an order of magnitude at
x B dm B
F= least in almost fully lro states, as observed
dx B at T = 650 K for B2 at the composition
This quantity is smooth in a random solid xAl = 0.5. At second-order transition points
solution, but shows a complicated variation F changes discontinuously.
8.6 Application to Real Systems 567
Figure 8-28. Thermodynamic factor of diffusion in b.c.c. Fe – Al alloys, calculated as a function of composi-
tion at three temperatures, 650 K, 1000 K and 1400 K. The calculations were performed with the irregular
tetrahedron CVM using the parameters given in Eq. (8-38) (Schön and Inden, 1998). At second-order transi-
tions, F changes by a step. Close to the stoichiometric compositions xAl = 0.25 and 0.5 the value of F increases
as a function of the degree of lro. For example, the temperature of T = 650K is only slightly below the D03 /B2
transition temperature, see Fig. 8-25. Therefore, at xAl = 0.25 the increase is small compared to xAl = 0.5 where
F increases by more than an order of magnitude because the distance from the B2/A2 transition temperature is
large and the degree of lro almost maximum. With increasing temperature this effect becomes smaller.
8.6.2.2 The F.C.C. Ni – Al System Introducing these enthalpies and the values
The variation of the thermodynamic fac- of the correlation functions of the configu-
tor is even more pronounced in the Ni – Al rations (see Table 8-3) into Eq. (8-24), the
system because there the order – disorder following parameters are obtained (in units
transitions occur at much higher tempera- of kB K)
tures than in Fe – Al, as shown in the cal- V0 = – 4630, V1 = – 1563, V2 = 5051,
culated f.c.c. phase diagram in Fig. 8-29
(Schön and Inden, 1998). The ECIs were V3 = 1563, V4 = – 421
derived from the ground-state energies of or equivalently
L12 – Ni3Al, L12 – NiAl3 , and L10 – NiAl,
obtained by Pasturel et al. (1992) from W (1) = 3370, W 1234
NiAlAlAl = – 4810,
first-principles calculations: W 1234
NiNiNiAl = 1440 (8-39)
j DH j [J/mol of atoms] The calculated thermodynamic factor of
diffusion is shown in Fig. 8-30 for the two
Ni3Al (L12) – 48 000
temperatures indicated in Fig. 8-29. Due to
NiAl (L10) – 56 000
the high degree of lro, the variation of F in
NiAl3(L12) – 22 000
the range of stoichiometric composition is
568 8 Atomic Ordering
8.6.3.1 B.C.C. Fe – Ti – Al
Ohnuma et al. (1998 b) studied ordering
and phase separation in the b.c.c. phase of
the ternary system, with particular empha-
sis on ternary miscibility gaps between dis- Figure 8-32. Calculated isothermal section of the
ordered and ordered phases. The theoreti- b.c.c. Fe – Ti – Al system at 1173 K using the energy
parameters in Table 8-9. Second-order transitions are
cal analysis was based on the irregular tet- shown as broken lines. The enlarged part of the sec-
rahedron CVM. The analysis of the Fe – Ti tion shows the good agreement obtained between ex-
and Ti – Al binary systems led to the energy periments and calculation (Ohnuma et al., 1998 b).
570 8 Atomic Ordering
tween the results of Miyazaki et al. and hedron – octahedron CVM to calculate the
those from the CVM and MC pointed out phase diagram, which is shown in Fig. 8-
by Inden and Pitsch (1991) was removed 39 a. The topology of the diagram is correct
by the TEM – EDS analysis of Kozakai and but it does not have the accuracy necessary
Miyazaki (1994) confirming the calculated for technological purposes.
direction of the tie-lines. Zhang et al. (2000) tried the CSA in the
tetrahedron approximation. Because the
c/a ratio is not ideal (f.c.c. and h.c.p. lat-
8.6.4 H.C.P. Cd – Mg
tices become equivalent in the ideal case),
The Cd – Mg system is a prototype order- two irregular tetrahedra should be consid-
ing system for h.c.p. alloys in the same way ered (Onodera et al., 1994). Zhang et al.
as Cu – Au plays this role for f.c.c. alloys. (2000) tried the irregular tetrahedra, but
This system has been studied using first- found that the regular tetrahedron CSA
principles methods by Asta et al. (1993). gave an equally good description. The
The configurational energy was evaluated value of g was varied in such a way that the
by a cluster expansion up to seven-point two congruent maxima which appear at the
clusters, as required for a treatment with compositions A3B and AB3 for g = 1.42
the octahedron – tetrahedron CVM. Lattice (valid for the ideal h.c.p. and f.c.c., see Sec.
relaxation and vibrational energy were 8.5.5) moved towards the mid-composition
taken into account, but no local relaxations. until, at g = 1.8, the maxima disappeared,
The lattice relaxation cluster expansion just as observed in the Cd – Mg system.
was based on experimental results for dis- Starting with energies taken from Asta et
ordered alloys, and the other ECIs were ob- al. (1993), which proved to be excellent,
tained from total-energy calculations on or- only slight changes were applied in order
dered phases. In total 32 energy terms were to get an optimum description of all the
used. The ECIs were then used in the tetra- properties that can be checked with avail-
574 8 Atomic Ordering
Figure 8-39. Calculated phase diagrams of the hexagonal Cd – Mg system. a) Phase diagram calculated from
“first principles” (redrawn from Asta et al., 1993). b) Phase diagram calculated with the CSA using almost the
same parameters as in a) and a value g = 1.8. Experimental data from Frantz and Gantois (1971).
able experimental data: the phase diagram, At present, it can be concluded that the
the enthalpy of mixing of the disordered CVM and MC techniques supply a treat-
state, the enthalpies of formation of or- ment of sufficient sophistication to cor-
dered phases and the chemical potentials of rectly handle the statistical aspects of the
Cd and Mg. The cluster energies used were equilibria in solid solutions.
(the values of Asta et al. are given in brack- Much progress has been made in the
ets): field of first-principles calculations of total
energies, including lattice and local relaxa-
Cd3Mg: – 9.3 (–10.3); CdMg: –13.7 (–13.5);
tions, and sometimes including excitations.
CdMg3: – 9.95 (– 9.5) in kJ/(mol of atoms). From this the energy parameters of the sta-
tistical models can be calculated. The re-
The calculated phase diagram is shown
sults for binary systems are numerous, but
in Fig. 8-39 b. The agreement between cal-
the field of multicomponent systems is still
culation and experiment is very good, not
to be explored. With increasing numbers of
only for the phase diagram but also for all
components the cluster size also has to be
the other properties mentioned above.
increased.
For the solution of metallurgical prob-
8.6.5 Concluding Remarks lems, the phase equilibria between all
phases have to be considered. The order –
The examples treated above represent disorder equilibria within a given crystal
the large group of real systems that can be structure are only one part of this task.
analyzed using these techniques. Similar Other phases, such as the liquid phase or
findings were obtained for oxide systems, intermetallic compounds, have to be in-
e.g., by Burton (1984, 1985), Burton and cluded. These aspects have been discussed
Kikuchi (1984), Kikuchi and Burton (1988), at recent workshops on the thermodynamic
Burton and Cohen (1995), Tepesch et al. modeling of solutions and alloys (e.g. Cac-
(1995, 1996), Kohan and Ceder (1996). ciamani et al., 1997).
8.7 Appendix 575
8.7 Appendix
Table 8-11. Superstructures of the f.c.c. lattice. 1
Positions
1 Figures 8-40 to 8-45 represent the original f.c.c. unit cell and the unit cells of the superstructures; 2 “Struk-
turbericht” designation; 3 In terms of vectors of the f.c.c. structure.
576 8 Atomic Ordering
Figure 8-40. Original f.c.c. unit cell and unit cell of Figure 8-43. Original f.c.c. unit cell and unit cell of
the superstructure A5B (B2/m), indicated by bold the superstructure A2B (Pt2Mo), indicated by bold
lines. lines.
Figure 8-41. Original f.c.c. unit cell and unit cell of Figure 8-44. Original f.c.c. unit cell and unit cell of
the superstructure D1a (Ni4Mo), indicated by bold the superstructure L11 (CuPt), indicated by bold
lines. lines.
Figure 8-42. Original f.c.c. unit cell and unit cell of Figure 8-45. Original f.c.c. unit cell and unit cell of
the superstructure D022 (Ti3Al), indicated by bold the superstructure A2B2 (I4/amd), indicated by bold
lines. lines.
8.7 Appendix 577
Positions
Positions
A3 F63 /mmc a = a0 [1 0 0] ––
b = a0 [0 1 0] (1/3 2/3 3/4) 2 (A/B)
c = a0 [0 0 1]
D019 (A3B) P63/mmc a = a0 [2 0 0] ––
b = a0 [0 2 0] 6 h (5/6 2/3 1/4) 6A
c = c0 [1 0 0] 2 c (1/3 2/3 1/4) 2B
–
B19 Pmma a = c0 [0 0 1] ––
b = a0 [0 1 0] 2 f (1/4 1/2 5/6) 2A
c = a0 [2 1 0] 2 e (1/4 0 1/3) 2B
–
A 2B P21/m a = a0 [1 1 0] 2 e (1/2 1/6 1/4) A
b = a0 [1 2 0] 2 e (1/6 1/2 1/4) A
c = c0 [0 0 1] 2 e (5/6 5/6 1/4) B
–
A 2B p6m a = a0 [1 1] 2 b (1/3 2/3) A
(2-dim.) b = a0 [1 2] 1 a (0 0) B
Ducastelle, F., Gautier, F. (1976), J. Phys. F 6, 2039. Gyorffy, B. L., Stocks, G. M. (1983), Phys. Rev. Lett.
Eymery, J., Lançon, F., Billard, L. (1993), J. Phy- 50, 374.
sique I France 3, 787. Hansen, M., Anderko, K. (1958), Constitution of Bi-
Ferreira, L. G., Mbaye, A. A., Zunger, A. (1987), nary Alloys. New York: McGraw-Hill.
Phys. Rev. B 35, 6475. Hijmans, J., de Boer, J. (1955), Physica 21, 471.
Ferreira, L. G., Mbaye, A. A., Zunger, A. (1988), Houtappel, R. M. F. (1950), Physica 16, 425.
Phys. Rev. B 37, 10 547. Inden, G. (1975 a), Z. Metallkde. 66, 577.
Ferreira, L. G., Wei, S.-H., Zunger, A. (1989), Phys. Inden, G. (1975 b), Z. Metallkde. 66, 648.
Rev. B 40, 3197. Inden, G. (1977 a), Z. Metallkde. 68, 529.
Finel, A. (1987), Thesis, University Pierre et Marie Inden, G. (1977 b), J. de Physique 38, C7, 373.
Curie, Paris. Inden, G., Pepperhoff, W. (1990), Z. Metallkde. 81,
Finel, A. (1994), in: Statics and Dynamics of Alloy 770.
Phase Transformations: Turchi, P. E. A., Gonis, A. Inden, G., Pitsch, W. (1991), in: Materials Science
(Eds.). NATO ASI Series B, Physics, Vol. 319. and Technology, Vol. 5: Phase Transformation in
Dordrecht: Kluwer, p. 495. Materials: Haasen, P. (Ed.). New York: VCH
Finel, A., Ducastelle, F. (1984), in: Mat. Res. Soc. Press, p. 497.
Symp. Proc. 21: Phase Transformations in Solids, Jacobs, L., Rebbi, C. (1981), J. Comput. Physics 41,
Maleme (Crete) 1983. Amsterdam: North-Holland 203.
Publ., p. 293. Kainuma, R., Ikenoya, H., Ohnuma, I., Ishida, K.
Finel, A., Ducastelle, F. (1986), Europhys. Lett. 1, (1997), Defect and Diffusion Forum Vols. 143 –
135 and Erratum p. 543. 147, 425.
Finel, A., Gratias, D., Portier, R. (1984), in: L’ordre Kalle, C., Winkelmann, V. (1982), J. Stat. Phys. 28,
et le Desordre dans les Materiaux, Reynaud, F., 639.
Clément, N., Couderc, J. J. (Eds.). Ecole d’Hiver Kanamori, J. (1966), Prog. Theor. Phys. 35, 16.
Aussois 1984. Les Ulis: Editions de Physique, Kanamori, J. (1979), in: Int. Conf. on Modulated
p. 9. Structures: Cowley, J. M., Cohen, J. B., Salomon,
Flinn, P. A. (1956), Phys. Rev. 104, 350. M. B., Wuensch, B. J. (Eds.). AIP Conf. Proc. 53,
de Fontaine, D. (1979), in: Solid State Physics, Vol. p. 117.
34: Ehrenreich, H., Seitz, F., Turnbull, D. (Eds.). Kanamori, J. (1984), J. Phys. Soc. Jpn. 53, 250.
London: Academic Press, p. 73. Kanamori, J., Kakehashi, Y. (1977), J. de Physique
de Fontaine, D. (1984), in: High Temperature Or- 38, C-7, 274.
dered Intermetallic Alloys, Mat. Res. Soc. Proc., Kear, B. H., Sims, C. T., Stoloff, N. S., Westbrook, J.
Vol. 39: Koch, Liu, Stoloff (Eds.). Pittsburgh: Ma- H. (Eds.) (1970), Ordered Alloys. Proc. 3rd. Bol-
terials Res. Soc., p. 43. ton Landing Conf. Baton Rouge: Claitors Publ.
de Fontaine, D. (1994), Solid State Phys. 47, 33. Khachaturyan, A. G. (1978), in: Progress in Materi-
Fosdick, L. D. (1963), Methods Comp. Phys. 1, 245. als Science, Vol. 22. New York: Pergamon Press,
Frantz, C., Gantois, M. (1971), J. Appl. Cryst. 4, p. 1.
387. Kikuchi, R. (1950), Phys. Rev. 79, 718.
Gahn, U. (1982), J. Phys. Chem. Solids 43, 977. Kikuchi, R. (1951), Phys. Rev. 81, 988.
Gahn, U. (1986), J. Phys. Chem. Solids 47, 1153. Kikuchi, R. (1974), J. Chem. Phys. 60, 1071.
Gahn, U., Sauthoff, G., Pitsch, W. (1984), in: Decom- Kikuchi, R. (1977 a), Acta Metall. 25, 195.
position of Alloys, the Early Stages, Acta-Scripta Kikuchi, R. (1977 b), J. de Physique 38, C7-307.
Metallurgica Conf. Proc.: Haasen, P., Gerold, V., Kikuchi, R. (1986), Progr. Theor. Phys. Suppl. 87,
Wagner, R., Ashby, M. F. (Eds.). New York: Per- 69.
gamon Press, p. 16. Kikuchi, R. (1987), Physica 142 A, 321.
Gautier, F. (1984), in: High Temperature Alloys: The- Kikuchi, R., van Baal, C. M. (1974), Scripta Metall.
ory and Design: Stiegler, P. O. (Ed.). New York: 8, 425.
Am. Inst. Min. Engrs., p. 163. Kikuchi, R., Burton, B. (1988), Physica B 150, 132.
Gautier, F., Ducastelle, F., Giner, J. (1975), Philos. Kikuchi, R., Sato, H. (1974), Acta Metall. 22, 1099.
Mag. 31, 1373. Kikuchi, R., de Fontaine, D., Murakami, M., Naka-
Gerold, V., Kern, J. (1987), Acta Metall. 35, 393. mura, T. (1977), Acta Metall. 25, 207.
Golosov, N. S., Tolstick, A. M. (1974), J. Phys. Kikuchi, R., Sanchez, J. M., de Fontaine, D., Yama-
Chem. Solids 36, 899. uchi, H. (1980), Acta Metall. 28, 651.
Golosov, N. S., Tolstick, A. M. (1975), J. Phys. Kleykamp, H., Glasbrenner, H. (1997), Z. Metallkde.
Chem. Solids 36, 903. 88, 230.
Golosov, N. S., Tolstick, A. M. (1976), J. Phys. Koch, C. C., Liu, C. T., Stoloff, N. S. (Eds.) (1985),
Chem. Solids 37, 273. High Temperature Ordered Intermetallic Alloys I,
Gratias, D., Sanchez, J. M., de Fontaine, D. (1982), Mat. Res. Soc. Symp. Proc., Vol. 39. Pittsburgh:
Physica A 113, 315. Materials Res. Soc.
580 8 Atomic Ordering
Köster, W., Gödecke, T. (1980), Z. Metallkde. 81, Ozoliņš, V., Wolverton, C., Zunger, A. (1998 b),
770. Phys. Rev. B 57, 5897.
Kohan, A. F., Ceder, G. (1996), Phys. Rev. B 53, Pasturel, A., Colinet, C., Paxton, A. T., van Schilf-
8993. gaarde, M. (1992), J. Phys.: Condens. Matter 4,
Kozakai, T., Miyazaki, T. (1994), J. Mater. Sci. 29, 945.
652. Priem, T., Beuneu, B., Finel, A., Livet, F., de Novion,
Kudo, T., Katsura, S. (1976), Prog. Theor. Phys. 56, C. H. (1989 a). J. de Physique 50, 2217.
435. Priem, T., Beuneu, B., de Novion, C. H., Chevrier, J.,
Kutner, R., Binder, K., Kehr, K. W. (1982), Phys. Livet, F., Finel, A., Lefevbre, S. (1989 b). Physica
Rev. B 26, 2967. B 156 + 157, 47.
Laks, D. B., Ferreira, L. G., Froyen, S., Zunger, A. Renaud, G., Motta, N., Belakhovsky, M., Lançon, F.,
(1992), Phys. Rev. B 46, 12 587. Billard, L. (1987), Solid State Comm. 63, 569.
Lebowitz, J. L., Phani, M. K., Styer, D. F. (1985), J. Richards, M. J., Cahn, J. W. (1971), Acta Metall. 19,
Stat. Phys. 38, 413. 1263.
Li, Y. Y. (1949), Phys. Rev. 76, 972. de Ridder, R., van Dyck, D., Amelinckx, S. (1980),
Li, Y. Y. (1949), J. Chem. Phys. 17, 447. Phys. Stat. Solidi A 61, 231.
Lihl, F., Ebel, H. (1961), Arch. Eisenhüttenwes. 32, Rubin, G., Finel, A. (1993), J. Phys.: Condens. Mat-
483. ter 5, 9105.
Liu, C. T., Taub, A. I., Stoloff, N. S., Koch, C. C. Sanchez, J. M. (1996), in: Theory and Applications
(Eds.) (1989), High Temperature Ordered Inter- of the Cluster Variation and Path Probability
metallic Alloys III, Mat. Res. Soc. Symp. Proc., Methods: Moran-Lopez, J. L., Sanchez, J. M.
Vol. 133, Pittsburgh: Materials Res. Soc. (Eds.). New York: Plenum Press, p. 175.
Liu, C. T., Cahn, R. W., Sauthoff, G. (Eds.) (1992), Sanchez, J. M., de Fontaine, D. (1978), Phys. Rev.
Ordered Intermetallics – Physical Metallurgy and B 17, 2926.
Mechnical Behaviour, Nato ASI Series Vol. 213. Sanchez, J. M., de Fontaine, D. (1980), Phys. Rev.
Dordrecht: Kluwer, p. 525. B 21, 216.
Livet, F. (1987), Acta Metall. 35, 2915. Sanchez, J. M., de Fontaine, D. (1981), in: Structure
Lu, Z. W., Wei, S. H., Zunger, A., Frota-Pessoa, S., and Bonding in Crystals, Vol. II: O’Keefe, M.,
Ferreira, L. G. (1991), Phys. Rev. B 44, 512. Navrotsky, A. (Eds.). New York: Academic Press,
Lu, Z. W., Klein, B. M., Zunger, A. (1995), Model. p. 917.
Sim. Mater. Sci. Eng. 3, 753. Sanchez, J. M., de Fontaine, D. (1982), Phys. Rev.
McCormack, R., de Fontaine, D. (1996), Phys. Rev. B 25, 1759.
B 54, 9746. Sanchez, J. M., de Fontaine, D., Teitler, W. (1982),
McCormack, R., Asta, M., Hoyt, J. J., Chakoumakos, Phys. Rev. B 26, 1465.
B. C., Misture, S. T., Althoff, J. D., Johnsson, Sanchez, J. M., Ducastelle, F., Gratias, D. (1984),
D. D. (1997), Comp. Mater. Sci. 8, 39. Physica A 128, 334.
Miyazaki, T., Isobe, K., Kozakai, T., Doi, M. (1987), Schlijper, A. G. (1983), Phys. Rev. B 27, 6841.
Acta Metall. 35, 317. Schlijper, A. G. (1984), J. Stat. Phys. 35, 285.
Mohri, T., Sanchez, J. M., de Fontaine, D. (1985), Schlijper, A. G. (1985), Phys. Rev. B 31, 609.
Acta Metall. 33, 1171. Schön, C. G., Inden, G. (2001), Comp. Mater. Sci.,
Mouritsen, O. G. (1984), Computer Studies of Phase 20, 98.
Transitions and Critical Phenomena. Berlin: Schön, C. G., Inden, G. (1998), Acta Mater. 46, 4219.
Springer. Schönfeld, B. (1999), in: Progress in Materials Sci-
Newell, G. F. (1950), Phys. Rev. 79, 876. ence, Vol. 44. New York: Pergamon Press, p. 435.
Oates, W. A., Wenzl, H. (1996), Scripta Mater. 35, Sigli, C., Sanchez, J. M. (1988), Acta Metall. 36,
623. 367.
Oates, W. A., Spencer, P. J., Fries, S. G. (1996), Sluiter, M., Turchi, P. E. A. (1989 a), in: Alloy Phase
CALPHAD 20, 481. Stability: Stocks, G. M., Gonis, A. (Eds.). Dor-
Oates, W. A., Zhang, F., Chen, S-L., Chang, Y. A. drecht: Kluwer Acad. Publ., p. 521.
(1999), Phys. Rev. B 59, 11 221. Sluiter, M., Turchi, P. E. A. (1989 b), Phys. Rev. B 40,
Ohnuma, I., Ikeda, O., Kainuma, R., Sundman, B., 215.
Ishida, K. (1998 a), presented at the CALPHAD Sluiter, M., Turchi, P. E. A., Fu, Zezhong, de Fon-
Meeting, Beijing. taine, D. (1986), Phys. Rev. Lett. 60, 716.
Ohnuma, I., Schön, C. G., Kainuma, R., Inden, G., Stoloff, N. S., Koch, C. C., Liu, C. T., Izumi, O.
Ishida, K. (1998 b), Acta Mater. 46, 2083. (Eds.) (1987), High Temperature Ordered Inter-
Onodera, H., Abe, T., Yokokawa, T. (1994), Acta metallic Alloys III, Mat. Res. Soc. Symp. Proc.,
Metall. 42, 887. Vol. 81.
Ozoliņš, V., Wolverton, C., Zunger, A. (1998 a), Taggart, G. B. (1973), J. Phys. Chem. Solids 34,
Phys. Rev. B 57, 6427. 1917.
8.9 References 581
Tepesch, P. D., Garbulski, G. D., Ceder, G. (1995), Warlimont, H. (Ed.) (1974), Order – Disorder Trans-
Phys. Rev. Lett. 74, 2274. formations in Alloys. Berlin: Springer.
Tepesch, P. D. (1996), J. Am. Ceram. Soc. 79, 2033. Wei, S.-H., Ferreira, L. G., Zunger, A. (1990), Phys.
Terakura, K., Oguchi, T., Mohri, T., Watanabe, K. Rev. B 41, 8240.
(1987), Phys. Rev. B 35, 2169. Wei, S.-H., Ferreira, L. G., Zunger, A. (1992), Phys.
Treglia, G., Ducastelle, F. (1980), J. Phys. F 10, Rev. B 45, 2533.
2137. Whang, S. H., Liu, C. T., Pope, D. P., Stiegler, J. O.
Turchi, P. E. A., Gonis, A. (Eds.) (1994), Statics and (Eds.) (1990), High Temperature Aluminides and
Dynamics of Alloy Phase Transformations: NATO Intermetallics. Warrendale: The Minerals Metals
ASI Series B, Physics, Vol. 319. Dordrecht: and Materials Soc.
Kluwer, p. 361. Williams, R. O. (1980), Metall. Trans. A 11, 247.
Turchi, P. E. A., Sluiter, M., Pinski, F. J., Johnson, D. Williams, R. O. (1984), CALPHAD 8, 1.
D. (1991a), in: Mat. Res. Soc. Symp. Proc., Vol. Wolverton, C., Zunger, A. (1997), Comp. Mater. Sci.
186: Stocks, G. M., Giamei, A. F., Pope, D. P. 8, 107.
(Eds.). Pittsburgh: Materials Res. Soc., p. 59. Wolverton, C., Ozoliņš, V., Zunger, A. (1998), Phys.
Turchi, P. E. A., Sluiter, M., Pinski, F. J., Howell, R. Rev. B 57, 4332.
H., Wachs, A. L., Fluss, M. L., Johnson, D. D., Yang, C. N. (1945), J. Phys. Chem. 13, 66.
Stocks, G. M., Nicholson, D. M., Schweika, W. Yang, C. N., Li, Y. Y. (1947), Chin. J. Phys. 7, 59.
(1991b), in: Mat. Res. Soc. Symp. Proc., Vol. 166, Zhang, J., Oates, W. A., Zhang, F., Chen, S-L., Chou,
Shapiro, S. M., Moss, S. C., Jorgensen, J. D. K-C., Chang, Y. A. (2001), Intermetallics 9, in print.
(Eds.). Pittsburgh: Materials Res. Soc., p. 231. Zorn, R., Heermann, H. J., Rebbi, C. (1981), J.
Vaks, V. G., Samolyuk, G. D. (1999), J. Exp. Theor. Comp. Phys. 23, 337.
Phys. 88, 89. Zunger, A. (1994), in: Statics and Dynamics of Alloy
van de Walle, A., Ceder, G., Waghmare, U. V. Phase Transformations: Turchi, P. E. A., Gonis, A.
(1998), Phys. Rev. Lett. 80, 4911. (Eds.), NATO ASI Series B, Physics, Vol. 319.
Wannier, G. H. (1950), Phys. Rev. 79, 357. Dordrecht: Kluwer, p. 361.
9 Diffusionless Transformations
Luc Delaey
a name of phase
b name of phase
g name of phase
G interfacial energy
d0 shear strain
e name of phase
e0 strain associated with the transformation
j surface to volume ratio
s stress
sa applied stress
Figure 9-1. Classification scheme for the displacive/diffusionless phase transformations as proposed by Cohen
et al. (1979).
shiyama in 1934 and by Greninger and If a sample is polished into the marten-
Troiano in 1949 (orientation relationship site (= parent phase), a surface relief ap-
between austenite – the parent phase – and pears. An edge-on section of such a sample
martensite) represent breakthroughs in the is shown in Fig. 9-2 a (see Hsu, 1980). The
study of martensitic transformations. origin of the surface relief is indicated by
The most frequently used techniques the white arrows. Owing to the macro-
will now be briefly discussed and illus- scopic martensite shear (the two thinner ar-
trated. rows), a surface relief is obtained. This is
9.1 Introduction 589
Figure 9-2. Some examples of how to see or measure the presence and growth of martensite (see text for de-
tails).
590 9 Diffusionless Transformations
explained further in Sec. 9.8.1.1 (Fig. also the specific volume, which allows the
9-19). transformation temperatures to be deter-
The electrical resistance shows at the mined by dilatometry (Fig. 9-2 e, from
transformation temperatures a deviation Yang and Wayman, 1993).
from linearity versus temperature. This is Changes in mechanical properties are
shown in Fig. 9-2 b for the martensitic also measured while the sample is trans-
transformation in an Fe – Ni alloy and a forming. The Young’s modulus exhibits a
Au – Cd alloy. The difference in the resis- dip between the two transformation tem-
tance ratio for the two different materials is peratures Ms and Mf , as clearly visible in
obvious and remarkable (see Otsuka and Fig. 9-2 f for four different alloys (see Su-
Wayman, 1977). Measuring the electrical gimoto and Nakaniwa, 2000).
resistance while cooling or heating a sam-
ple is a very convenient and relatively easy
and accurate technique of determining the 9.2 Classification and
transformation temperatures Ms , Mf , As Definitions
and Af .
The heat exchanged with the surround- A structrual change in the solid state is
ings is becoming a more popular method of termed “displacive” if it occurs by coordi-
determining the transformation tempera- nated shifts of individual atoms or groups
tures. An example is shown in Fig. 9-2 c of atoms in organized ways relative to their
(Nakanishi et al., 1993). A DSC (differen- neighbors. In general, this type of transfor-
tial scanning calorimetry) curve allows any mation can be described as a combination
particular behavior of the martensitic sam- of “homogeneous lattice-distortive strain”
ple to be detected (for example, effects oc- and “shuffles”.
curring during heat treatments and/or def- A lattice-distortive deformation is a ho-
ormation steps). mogeneous strain that transforms one lat-
A number of martensitic transformations tice into another; examples are shown in
and materials are characterized by a so- Fig. 9-3. The homogeneous strain can be
called shape-memory effect (see Sec. represented by a matrix according to
9.11.2). Figure 9-2 d (courtesy of Memory
y = Sx (9-1)
Europe) shows the displacement of a
spring made of a NiTi shape-memory alloy. where the strain S deforms the lattice vec-
The spring controls a small valve in a cof- tor x into a lattice vector y. This strain
fee-making machine. At the temperature As is homogeneous because it transforms
the hot water starts to drop onto the coffee straight lines into other straight lines. A
powder. This valve is completely open as spherical body of the parent phase will thus
soon as the temperature Af is reached. The be transformed into another sphere or into
temperature range between As and Af seems an ellipsoidal body. The actual shape of the
to be most suitable for making the best cup ellipsoid depends on the deformation in the
of coffee. On cooling, the valve closes three principal directions. If a spherical
again. The “displacement – temperature” body is completely embedded inside the
curve measured on cooling does not coin- matrix phase and is undergoing the strain
cide with the heating curve. S, the volume and shape change associated
During a martensitic transformation, not with this deformation will cause elastic
only is the shape of the sample changed but and, sometimes, plastic strains in the parent
9.2 Classification and Definitions 591
of the remaining principal strains becomes lattice hase already been generated by the
positive (e2 > 0) and the other negative Bain strain, this additional strain should be
(e1> 0) (Bevis and Allan, 1974). a “lattice-invariant strain”. Slip and twin-
The strain ellipsoid for the above exam- ning in the product phase or in both phases
ple has a special shape. The cone of undis- are typical lattice-invariant strains; both
torted vectors of the product phase degen- deformation modes are shown schemati-
erates into a pair of planes, which rotate in cally in Fig. 9-6. The diffusionless phase
opposite directions in the pure strain. transformation can in this case be repre-
Hence either of them may be invariant if sented by an analog consisting of a pure
the total deformation of the lattice includes lattice strain (B), a rigid lattice rotation (R)
a suitable rotation. Because all the vectors and an inhomogeneous lattice-invariant
in this plane are undistorted, the transfor- deformation (P). The last factor is also as-
mation is said to be an “invariant plane sociated with a shape change, which mac-
strain (IPS)” type. The pure Bain strain is roscopically can be considered as homoge-
then equivalent to a simple shear on that in- neous. Such a combination is typical of
variant plane. Because this invariant plane martensite and is discussed in Sec. 9.8.1.4.
is also a matching plane between the ma- In cases where the lattice-invariant shear
trix and the product phase and both phases is twinning (as opposed to faulting or slip),
have to be present at the same time, a rigid type I twinning, where the twin plane orig-
body rotation (R) over an angle q is re- inates from a mirror plane in the parent
quired in order to bring the product and the phase, has been assumed. Otsuka (1986)
parent phases in contact along that plane, carefully analyzed for a number of systems
the habit plane. the possibility of a type II twinning as an
A similar situation is found in structu- alternative inhomogeneous shear. In type II
rally less sophisticated systems, namely the twinning, the shear direction stems from a
f.c.c. to h.c.p. transformation in cobalt. two-fold symmetry direction of the parent
Both phases are close packed and a simple phase. In a table, Otsuka (1986) compiled
shear on the basal plane transforms the cu- all the twinning modes observed in marten-
bic stacking into a hexagonal stacking. Be-
cause the atomic distances in the basal
plane do not change significantly during
the transformation, the plane of simple
shear is the plane of contact or the habit
plane (HP). This is the case, however, only
if there is zero volume change in the trans-
formation, i.e., in the case quoted above if
the h.c.p. phase has an ideal axial ratio of
1.633.
The situation becomes more complicated
if none of the principal strains is zero, but
of mixed sign. To achieve matching along
the plane of contact in cases where the two
Figure 9-6. Schematic representation of (a) the ho-
phases coexist, a deformation is needed ad- mogeneous lattice deformation, (b) the inhomogene-
ditional to the pure Bain strain in order to ous lattice-invariant deformation (slip and twinning),
have an invariant plane. Because the final and (c) the lattice rotation.
596 9 Diffusionless Transformations
site and found that most of them are type I et al. (1998) carried out computer simula-
or compound but that type II twinning had tions by using the molecular dynamics
only recently been observed. According to method. The result showed that the irra-
Nishida and Li (2000), five different twin- tional thin boundary did not show any
ning modes exist in TiNi and other shape steps. Thus, the above interpretation for an
memory alloys such as Cu-Al-Ni, Cu-Sn irrational twin boundary is justified. Ot-
etc., namely the {111} type I, the {011} suka and Ren (1999) have pointed out
type I, the 〈011〉 type II, the (100)-com- again the importance of type II twinning in
pound and the (001)-compound twins. the crystallographic aspects of martensite.
Type II twinning has recently received They also stress the role that martensite ag-
much attention as a mechanism for lattice ing has on the rubber-like behaviour of
invariant shear in some alloys. Since martensite, a point that has been a long-
type II twins have irrational twin boundar- standing unsolved problem. They showed
ies, the physical meaning of an irrational that the point defects play a primordial
boundary is still a controversial problem. It role. The deformation mechanisms of the
has proposed that an irrational boundary cold deformation of NiTi martensite have
consists of rational ledges and steps, the been thoroughly analyzed by Liu et al.
average being irrational. Thereafter, Hara (1999a, b). They also found an interplay
et al. (1998) carried out a careful study to between type I and type II twinning.
observe 〈111〉 type II twin boundaries in a The cubic to tetragonal transformations,
Cu-Al-Ni alloy by HRTEM, but they could which occur in a number of metallic and
not observe any ledges or steps. The boun- non-metallic systems, need some special
dary is always associated with dark strain attention. The volume change with these
contrast, and the lattice is continuous transformations is sometimes very small or
through the irrational boundary. Nishida even absent, and the c/a ratio does not
and Li (2000) also made extensive studies change abruptly but progressively (Fig.
on 〈011〉 type II twin boundaries in TiNi by 9-7); the transformation is then said to be
HRTEM, but they did not observe ledges or “continuous”. The c/a ratio can be smaller
steps either. Based on these experimental or larger than unity, depending on compo-
results, it is thus most likely that the type II sition and temperature. The shape change
thin boundary is irrational even on a micro- associated with the transformation is small
scopic scale, and the strains at the boun- enough in many systems, especially in
dary are elastically relaxed with wide twin those belonging to the quasi-martensites,
width. To confirm this interpretation, Hara for elastic accommodation alone to be suf-
Table 9-1. A schematic overview of some characteristics typical of the various types of diffusionless phase
transformations.
* C: continuous, D: discontinuous
** F: first order, S: second order
ficient for lattice matching. It is, however, phase transformations. A soft mode is,
possible for c and a to change abruptly with in simple terms, a vibrational mode, the
zero volume change. square of whose frequency tends toward
Based on the crystallographic aspects zero as the temperature approaches that of
discussed above, a list of the most typical the phase transition. The average static
characteristics of the diffusionless phase atom displacements resemble the frozen-in
transformations can be compiled (Table pattern of the vibrational displacements of
9-1). a certain vibrational mode. According to
Vallade (1982), “the crystal lattice vibra-
tions can within the harmonic approxima-
9.3.2 Pre-transformation State
tion be separated into independent plane
Diffusionless structural changes are waves (phonons) characterized by a set
achieved by atom displacements, such as of collective atomic displacements corre-
shuffles and shears. The new atomic con- sponding to a well defined frequency. The
figuration is already prepared in some ma- energy involved is a function of the squares
terial systems at temperatures above the of the momentum and of the eigenfrequen-
transition temperature. Atoms in the parent cies of the mode. The eigenfrequencies de-
phase then become displaced more easily pend only on the mass of the atoms and on
towards their positions in the new phase the force constants. It is clear that the van-
because the restoring force that is felt by ishing of one eigenfrequency corresponds
the displaced atoms diminishes on cooling. to the lack of restoring force for the mode:
In certain cases the restoring force even the amplitude can then grow without any
vanishes at the phase transition tempera- limit and the lattice is mechanically un-
ture. stable. Stability can be recovered only
Certain shuffle transformations result by changing atomic equilibrium positions
from a vibrational instability of the parent which, in turn, changes the force con-
phase and are therefore called “softmode” stants“.
598 9 Diffusionless Transformations
shear constant vanishes at Tc . The trans- the fully formed product. The exact mecha-
formation mechanism is therefore not a nisms for the various types of martensitic
combination of expanding and contracting transformations are still under debate.
atom movements, but a lock-in of long-
wavelength shear-type movements on
– 9.3.4 Microstructures
{110} planes in the ·110Ò directions in this
scheme. The microstructures that result from
As for the martensitic transformations, diffusionsless phase transformations show
the situation is not so straightforward. The typical features, which are related to the
Bain-type strains are concerned only with crystallograph of the transformation. The
the correspondence between initial and fi- transformation is associated with a reduc-
nal lattices; they do not give the actual ob- tion in symmetry; consequently, different
served crystal orientation relationships equivalent orientational states of the prod-
between them. Based on the experimen- uct phase are formed. A single crystal of
tally determined orientation relationships, the parent phase thus transforms to a col-
different transformation mechanisms have lection of the product-phase crystals, called
been proposed, such as shears on the planes variants, that are separated by interfaces.
and along the directions involved in the The higher the symmetry of the parent
orientation relationship. However, these phase and the lower the symmetry of the
shear mechanisms have been found to be product phase, the greater is the number of
too simple to be consistent with the experi- equivalent paths of transformation. The
mental facts. More recently, a transforma- number of equivalent orientations or vari-
tion mechanism has been proposed for ants is also determined by the symmetry
martensitic transformations of b.c.c. to elements that are maintained or broken due
close-packed structures; a condensing state to the Bain strain. The collection of vari-
of some soft phonon modes combined with ants constitutes the microstructure.
a homogeneous shear explains the variety The order of the transformation (whether
of structures that are found. For the same first or second order) also determines the
transformations, Ahlers (1974) proposed a microstructure: in the former case parent/
two-shear mechanism; the first shear product or heterophase interfaces in addi-
creates the close-packed planes, whereas tion to product/product or homophase inter-
the close-packed structure is obtained by faces are created, whereas in the latter only
the second shear. product/product interfaces are formed. In
Martensitic transformations are first- the former case the first plates formed can
order phase transformations that occur by grow to a larger extent than those formed
nucleation and growth. The growth stage later, which can lead, for example, to mi-
generally takes place by the motion of crostructures with fractal characteristics.
interfaces converting the parent phase to Fig. 9-10 shows a selection of character-
the fully formed product phase. Two types istic microstructures obtained through dif-
of paths have to be considered for the case fusionless phase transformations.
of nucleation, the “classical” and the “non-
classical” nucleation paths (Olson and Co-
9.3.5 Shape Changes
hen, 1982). The latter involves a continu-
ous change in structure whereas the former If we could transform a single crystal of
involves a nucleus of the same structure as the parent phase into a single crystal of the
9.3 General Aspects of the Transformation 601
Figure 9-12. A series of macrographs representing the shape change while mechanically straining a Cu–Al–Ni
martensite single crystal; (a) to (e) increasing with time (Ichinose et al., 1985).
If a single crystal of the product phase is temperature, similar shape changes are ob-
mechanically strained it can either be trans- served, provided that the reverse transfor-
formed to another or be deformed to a dif- mation is also diffusionless. The situation
ferently oriented single crystal of the prod- for second-order phase transformations is
uct phase (Fig. 9-12). Similar behaviour is straightforward; the sample whose shape is
also typical of a number of ferroelastic changed during the forward transformation
materials; the reorientation is there referred and possibly after deformation below Tc re-
to as “switching” (Wadhawan, 1982). The verts back to its original shape above Tc in
switching force in these materials is not a homogeneous and continuous way. For
only a mechanical stress but can also be an first-order transformations, the reverse
electric or magnetic field, the domains be- transformation is more complex and not
ing either electrically or magnetically po- yet well understood. Much depends on
larized. whether the forward transformation is
In first-order phase transformations, as completed or not, and whether the growth
shown above, the full transformation shape of the martensite plate occurs by bursts or
change is induced locally and is gradually under thermoelastic conditions (see be-
spread over the whole sample within a low). Occurrence of the reverse transfor-
small temperature interval, whereas in a mation does not necessarily imply that the
second-order phase transformation the original shape is restored. Depending on
sample changes its shape homogeneously the crystal symmetry of the product phase,
and continuously as soon as the critical more than one crystallographically equiva-
transition temperature Tc is reached. lent path can be followed for the reverse
Until now, shape changes have been dis- transformation. The shape changes that oc-
cussed that are induced by the forward cur during the reverse transformation are at
transformation. It is evident that if a sam- the origin of the shape-memory effect and
ple of the low-temperature phase, a single are discussed in Sec. 9.10.2.
crystal or a polyvariant, is heated to tem-
peratures above the reverse transformation
604 9 Diffusionless Transformations
which is higher than T0 . The difference tributions to the Gibbs energy change,
between the forward and the reverse trans- DGnon-chem, and Eq. (9-3) then becomes
formation temperatures is the transforma-
DGtot = DGchem + DGnon-chem (9-4)
tion hysteresis. The true first-order equilib-
rium temperature, T0 , which is calculated The transformation then proceeds until
from DG = 0, can thus only be bracketed DGtot becomes minimum or, if the phase
from experimental data for the forward and boundary is mobile, until the total force at
the reverse transformation temperatures, the parent-to-product interface is zero. If
and is not necessarily halfway between Ms the advancing direction of the interface is
and As. x, we can write
Strain energy resulting from the transfor-
[∂ (DGtot)/∂x] dx = 0 (9-5)
mational shape change and interfacial en-
ergy have been omitted from the free-en- or
ergy curves in Fig. 9-13. These two non- [∂DGchem /∂x] dx + [∂DGnon-chem /∂x] dx = 0
chemical-energy terms have to be consid-
ered, however, in the overall free-energy The sum of the non-chemical restoring
balance. The strain energy associated with forces is then identical with the chemical
the formation of a single domain of the driving forces. The difficult task now is
product phase is proportional to the volume to find expressions representing the non-
of that domain. The interfacial energy is chemical terms. Three thermodynamic ap-
not directly related to the volume of the proaches have been worked out, dealing es-
transformed domain but merely to its sur- sentially with the influence of the two non-
face-to-volume ratio, and, in the case of an chemical contributions on the transforma-
anisotropic interfacial energy, also to the tion behavior (Roitburd, 1988; Ball and
orientation of the interface. Both quantities James, 1988; Shibata and Ono, 1975, 1977).
are positive and thus consume part of the According to Roitburd (1988), the strain
chemical driving force for a forward trans- energy, which arises in crystals owing to a
formation. Both terms will, however, in- diffusionless phase transformation, can de-
crease the driving force for the reverse crease if the crystals are subdivided into
transformation, provided that the inter- domains arranged such that a maximum
facial coherence is not lost. The reverse compensation of the individual strain fields
transformation might start below T0 if a is achieved. In order to determine which ar-
negligible net driving force is required for rangements are energetically most favor-
nucleation. able, Roitburd calculates the strain energy
Considering the Gibbs energy per unit for arbitrary domain arrangements, and
molar volume, the total Gibbs energy then minimizes this energy. The formula-
change per unit molar volume for the for- tion of this problem is complex and can
mation of a single domain of the product hardly be solved in general, but he suc-
phase embedded in the matrix phase is ceeded for some specific cases.
given by Ball and James (1988) do not assume
any geometric restrictions on the shape or
DGtot = DGchem + (DGelast + j DGsurf) (9-3)
arrangements of the domains; they found
where j is the surface-to-volume ratio this necessary to determine microstructures
of the single domain. The two terms in occurring in complex stress fields, or to
parentheses are then the non-chemical con- explore new and unusual domain arrange-
606 9 Diffusionless Transformations
ments. The general aim of their work was Shibata and One (1975, 1977) use the
to develop mathematical models, using cal- Eshelby theory; the principle of their calcu-
culus of variations, capable of predicting lation is in a corrected version (Christian,
the microstructure, especially the micro- 1976) illustrated schematically in Fig.
structural details at the interface between 9-14. An embedded part of the parent
the parent and the product phases. At- phase is cut out (step a) and is allowed to
tempts have been made to predict the pos- transform stress-free into the product phase
sible interfaces between austenite and mar- (step b). A lattice-invariant deformation is
tensite from a minimization of a Gibbs en- applied (step c) and the transformed crystal
ergy function, which depends on the defor- is subjected to forces along its surface such
mation gradients of all possible domain that it is deformed to the original shape
variants and on temperature. A deforma- (step d). The thus deformed part of the
tion or domain is then termed stable if product phase is introduced in the empty
it minimizes the total energy. They show, space of the parent phase (step e), and the
among others, that a martensite–austenite forces are removed, creating internal
interface can exist as an energy-minimiz- stresses in both the product and the parent
ing sequence of very fine twins. A further phase. The total energy is then calculated
example of an intriguing application is as a function of all possible lattice orienta-
the formation of triangular Dauphiné twins tions, taking into account the actual elastic
in quartz, which become finer and finer in constants and the modes of lattice-invari-
the direction of increasing temperature. A ant deformation, twinning or slip.
Gibbs energy function accounting for this The total Gibbs energy of the system is
behaviour could be constructed. therefore a function not only of the intrin-
sic energies of the stress- and defect-free
parent and product phases, but also of the
arrangement of the domains. The non-
chemical component of the total Gibbs en-
ergy of the transforming system is lowered
by an appropriate rearrangement of the mi-
crostructure and/or by irreversible plastic
deformation.
If the structural change can be repre-
sented by an order parameter e, the Gibbs
energy of the system can then, according to
the theory of Landau–Devonshire, be rep-
resented by
G = G0 + a (T – T1) e2 – B e4 + Ce6 (9-6)
where a, B and C are constants and T1 > 0.
It can be shown that the high-temperature
phase becomes unstable with respect to
any fluctuation of e, as soon as the temper-
Figure 9-14. The necessary steps in calculating the ature reaches T1 on cooling, and hence is
elastic stresses induced by a transforming ellipsoid thermodynamically metastable between T0
(Christian, 1976). (For details see text.) and T1. Accordingly, the low-temperature
9.4 Shuffle Transformations 607
phase cannot exist at temperatures higher The most relevant thermodynamic data
than T2, which is the temperature above for the various diffusionless phase transfor-
which the low-temperature phase becomes mations are presented in Table 9-2.
unstable with respect to any fluctuation in e.
That additional undercooling is needed
for further transformation below Ms is due 9.4 Shuffle Transformations
(in part) to the non-chemical contributions,
which increase with increasing volume Shuffle transformations from a distinct
fraction of transformed product. class of diffusionless phase transitions. At
An interesting aspect of the diffusionless the unit-cell level the atom displacements
phase transformations that are accompa- are intercellular with little or no pure strain
nied by a volume and shape change is the of the lattice. The role of elastic strain en-
role played by external stresses, e.g., hy- ergy in shuffle-phase transformations is
drostatic or uniaxial. Both thermodynamics sufficiently small that the transformation
and experiments show that the transforma- can either occur continuously from the par-
tion temperatures are affected by the appli- ent to the product phase or that it is com-
cation of stresses. According to Wollants et pletely controlled by interfacial energy. In
al. (1979), the relationship between a uni- the former case the transformation is sec-
axially applied stress s and the transfor- ond order whereas in the latter it is a first-
mation temperature T depends on the order phase transformation.
transformation entropy and the transfor- Cohen et al. (1979) gave three examples
mational strain in the direction of the ap- which clearly illustrate the shuffle transfor-
plied stress. This relationship, the Clau- mations. The displacive transformation in
sius–Clapeyron equation for uniaxially strontium titanate is the prototype example
stressed diffusionless first-order phase of a pure shuffle transformation. The asso-
transformations, is ciated strain energy is so small that the
transformation occurs continuously. The
ds /dT = – D S/e = – DH*/[T0 (s ) e] (9-7)
b-to-w transformation in some Ti and Zr
where DH* = DH – FDl = DH – s e Vm = alloys shows, in addition to the shuffle
T0 (s ) D S is itself a function of the applied displacements, small homogeneous lattice
load, e = Dl/l, l is the total “molar length” distortions. These distortions are small
of the sample, and F is the applied load enough for the transformation mechanism
(s = F/A). This equation is similar to that and the resulting microstructure to be dom-
relating the equilibrium temperature to the inated only by shuffling. In ferroelectric
hydrostatic pressure, except for the nega- transformations, which are accomplished
tive sign on the right-hand side of Eq. (9-7). by shuffling, the interfacial energy is con-
This relationship between ds and dT is ex- stituted largely by electrostatic interaction
perimentally constant for most of the diffu- energies and is therefore dependent on the
sionless transformations, which means that orientation of the domain interfaces. The
the thermodynamic quantity DS is, within interfacial energy in those materials is
the experimental scatter, independent of strongly anisotropic and controls the poly-
temperature and stress. Knowing the trans- domain structure.
formation strain, uniaxial tensile tests are Phase transformations that can be en-
very useful for determining the transforma- tirely described by shuffle diplacements
tion entropy. are often found where the change in crystal
608 9 Diffusionless Transformations
Table 9-2. The elastic shear constant and some thermodynamic data characterizing the diffusionless phase
transformation (Delaey et al., 1982 b).
Ferrous g Æ a¢ 2–3 negative ≈ 10–1 ≈ 2000 – 300 5.8 150 – 450 200 – 400
gÆe 3–10 (positive 600 – 1800
for Ni > 30%)
structure is such that the point group to placement. For strontium titanate, the order
which the crystal structure of the product parameter would then simply be the rota-
phase belongs is a subgroup of that of the tion angle that describes the displacement
parent phase. In other words, some symme- of the oxygen atoms around the titanium
try elements of the high-temperature phase atoms (see Fig. 9-4). For convenience, the
are lost on cooling below the transition order parameter is taken as zero for the
temperature Tc . Because of this group/sub- high-temperature configuration and as non-
group relationship, the product phase pos- zero for the low-temperature phase. The
sesses two or more equally stable orienta- majority of such transitions are found in
tional states in the absence of any external chemical compounds (e.g., Rao and Rao,
field. The change in crystal structure can 1978). As soon as the critical temperature
easily be described by an order parameter Tc is reached on cooling, the order parame-
which itself is related to the shuffle dis- ter changes continuously. The thermody-
9.4 Shuffle Transformations 609
namics of such transformations are then in point-group symmetry are called ferroic
the temperature range close to Tc , which is phase transitions. We refer to a crystal as
dealt with by a Landau approach. being in a ferroic phase if that phase results
from a symmetry-lowering ferroic phase
transition”.
9.4.1 Ferroic Transformations
Not all ferroelastic phase transitions be-
Usually a phase transition that is domi- long to shuffle transformations as defined
nated by shuffling is associated with a in Fig. 9-1. Indeed, in addition to shuffle
change in some physical properties, such as displacements, as for example those in-
spontaneous electrical polarization, strain volved in the cubic to tetragonal transition
and magnetization. Because the crystal in barium titanate, the lattice may become
symmetry of the parent phase decreases homogeneously distorted. For the example
during the phase transition, two or more considered here, the lattice distortion oc-
equivalent configurations of the product curs discontinuously at the transition tem-
phase are formed. In the absence of any ex- perature; the lattice parameters change
ternal field, the average polarization of the abruptly (Fig. 9-15). Even below this tran-
product phase is zero. However, under a sition temperature, the lattice continues to
suitably chosen driving force, which may be homogeneously distorted. In cases
be an electrical field (E ), a mechanical where this lattice is tetragonal, the c/a ratio
stress (s), or a magnetic field (H), the do- steadily increases. For barium titanate the
main walls of the product phase move, change in c/a continues until the tempera-
switching the crystal from one domain or- ture for another first-order phase transition
ientation to the other. Owing to the applica- is reached. Many such phase transforma-
tion of a uniaxial stress, for example, one tions are encountered in chemical com-
orientation state can be trensformed repro- pounds. In some cases the amount of spon-
ducibly into the other, and the crystal is taneous strain is not large enough to con-
then said to be “ferroelastic”. The materials trol the microstructure. In others, the strain
exhibiting this property are called ferro- energy associated with the transformation
elastic materials. Similarly, we can define will be dominant. The transformation is
ferroelectric and ferromagnetic materials. then, according to Fig. 9-1, quasi-marten-
According to Wadhawan (1982), “phase sitic or martensitic. In ferroelectric materi-
transitions accompanied by a change of the als, the interfacial energy also has to be
610 9 Diffusionless Transformations
undistorted line can be found in the lattice- are very similar (large plates, occurrence of
distortive deformation. The f.c.c.-to-f.c.c.¢ variants and twins). A distinction between
transformation in cerium is considered as the two transformations cannot be made by
the prototype for a dilatational dominant simply judging only the product morphol-
transformation. Below 100 K cerium un- ogy, but rather a knowledge is required of
dergoes a pure volume contraction of about the morphological relationships between
16%; the ellipsoid of the f.c.c.¢ phase thus parent and product phases during the trans-
falls completely inside that of the high- formation itself. It may be adequate to say
temperature f.c.c. phase. The low-tempera- first what a quasi-martensitic transforma-
ture cubic to tetragonal transformation in tion is: a quasi-martensitic transformation
tin also appears to be dominated by dilata- is not a martensitic transformation, which
tion, although some deviatoric components itself is “a first-order phase transformation,
are present; the volume expansion is about that undergoes nucleation, passes through a
27%. The deviatoric component is not two-phase mixture of the parent and prod-
large enough to let the original sphere uct phases, and which product grows with a
intersect with the dilated ellipsoid. transformation front in a plate-like or lath-
The name “dilatational diffusionless like shape being indicative of a tendency
phase transformation” has been used by toward an invariant-plane interface” (Co-
Buerger (1951) but with a different mean- hen et al., 1979). If a deviatoric dominant
ing. In the systems he considers, for exam- transformation does not satisfy the above
ple the CsCl-to-NaCl transitions in many criterion, it should not be designated as
alkali metal halides, he defines the term martensitic but as quasi-martensitic.
dilatational as follows: “the transformation Three aspects are common to most of the
can be achieved by a differential dilatation materials that transform quasi-martensiti-
in which the structure expands along the cally: (1) the lattice distortion is small and
trigonal axis and contracts at right-angles deviatoric dominant and the change in lat-
to the axis”. Although the volume change tice distortion is continuous or nearly con-
in these and other related inorganic sys- tinuous; (2) a banded internally twinned
tems may be very large (up to 17%), the microstructure gradually builds up on cool-
transformation is, in the context of Fig. 9-1, ing below Tc ; and (3) a mechanical lattice
clearly not dilatation dominant but devi- softening is expressed by elastic shear
atoric dominant. See Kriven (1982) for a constants approaching zero as Tc is ap-
more detailed review of these dilatational proached. Because of the small lattice dis-
dominant transformations. tortion at the transformation, the ratio of
the strain energy to the driving energy for
transformation is small; this ratio has been
used by Cohen et al. (1979) as an alterna-
9.6 Quasi-Martensitic tive index to differentiate quasi-martensitic
Transformations transformations from martensitic.
The three aspects are now illustrated by
The quasi-martensitic and the marten- taking the manganese-based magnetostric-
sitic transformation are both deviatoric tive antiferromagnetic alloys as an example
dominant and are characterized by an un- (see Delaey et al., 1982a). One of the four
distorted line. The morphologies of the polymorphic states of manganese is the
product phases of the two transformations gamma f.c.c. phase which is stable only at
612 9 Diffusionless Transformations
high temperatures. Alloying with elements schematic representation of the phase dia-
such as Cu, Ni, Fe, Ge, Pd and Au stabi- gram of the Mn-based alloys is given in
lizes the f.c.c phase and the latter can be re- Fig. 9-17 a, showing the temperature–com-
tained by quenching. However, owing to position areas in which the various crystal
the antiferromagnetic ordering, the lattices structures and microstructures are ob-
become homogeneously distorted. This or- served. The accompanying variation in the
dering to the Mn atoms starts at a temper- lattice parameters as a function of tempera-
ature TN , which is the Néel temperature ture for the three classes of Mn-based al-
for the paramagnetic to antiferromagnetic loys is given in Fig. 9-17 b.
transition. The transformed product phase The changes in lattice parameters show
has a banded microstructure containing that the transformation is almost second or-
fine twins. The temperature at which this der, except for some alloys of class I and III
banded microstructure is formed does not where the transformation is weakly first
always coincide with the transformation order. A phase transformation is called
temperature TN. Vintaikin et al. (1979) di- “weakly first order” whenever the height of
vide these antiferromagnetic alloys into the discontinuous jump in the correspond-
three classes according to the relative posi- ing thermodynamic property is very small.
tions of the temperature TN and the temper- The formation of the twinned banded mi-
ature TTW. The latter is the temperature at crostructure extends over the entire volume
which the banded microstructure sets in. of the sample quasi-instantaneously and is
Depending on the type of lattice distortion, visible in polarized light because of the
the alloys are grouped into three classes, non-cubic structure of the product phase.
each class being characterized by the rela- Similar microstructures are observed in
tive positions of the two temperatures. A other quasi-martensitic product phases
such as V54 – x Ru46Osx (Oota and Müller,
1987). The microstructure, if properly
oriented with respect to the prepolished
surface, exhibits a surface relief effect that
is enhanced as the temperature decreases
below TTW. This surface relief proves that
the transformation is accompanied by a
shape change associated with each domain.
Because of the continuously changing lat-
tice parameters, accommodation stresses
are built up as the temperature decreases.
An appropriate arrangement of these do-
mains reduces the overall stored elastic en-
ergy; further changes in microstructure are
therefore expected to occur even below the
transition temperature.
Class II alloys do not exhibit the twinned
banded microstructure immediately below
Figure 9-17. Schematic representation of (a) the
phase diagram and (b) the variation of the lattice pa-
TN. In the temperature region between TN
rameters for the three classes of Mn-based alloys and TTW, broadening of some of the X-ray
(Delaey et al., 1982 a). diffraction peaks is observed, which is
9.7 Shear Transformations 613
attributable to a chaotic distribution of the gradually disappear and the product phase
a and c axes with small undercooling. At becomes a single domain maximizing the
TTW the banded structure becomes visible shape change. The shape change thus ob-
(point A in Fig. 9-17) and a tetragonal tained is gradually recovered on heating
structure can now be clearly detected by X- the sample and is completely recovered at
ray diffraction. If the sample is now heated, TN and not (as in the case of Mn-based
only the banded microstructure disappears, alloys of class II) at TTW, but at the point B
not at A but at a temperature B that coin- in Fig. 9-17. The quasi-martensitic alloys
cides with the Néel temperature. This thus also exhibit the shape-memory effect.
proves that on cooling, very small, submi- Some of the materials characterized by
croscopic tetragonal regions are first shuffle displacements during the phase
formed as soon as TN is reached. Hocke and transition may develop elastic strains as
Warlimont (1977) have shown that when transformation proceeds. As in ferroelec-
the distortion |c/a – 1| becomes greater trics, for example, in addition to the elastic
than 0.005, a critical value is obtained at strain energy, the dipole interaction energy
which the elastic strain is relaxed through also contributes to the polydomain forma-
coalescence of the small distorted regions tion. If the elastic strain energies are only
into large banded twinned regions. Thus, at slightly dominating, the transformation is
TTW there is not a phase transformation but quasi-martensitic; if, however, the elastic
a stress relaxation in the microstructure, strain energy is largely dominating, the
which results in a twinned microstructure. transformation can be martensitic.
The lattice distortive phase transformation Sometimes it becomes difficult to differ-
itself occurs at TN, followed immediately entiate between martensite and quasi-mar-
(class I) or after some undercooling (class tensite, as for example in In-based alloys.
II) by a domain rearrangement and macro- In particular, if quasi-martensitic samples
scopic twinning. are cooled in such a way that a temperature
Similar conclusions can be drawn for gradient is created across the sample, the
other quasi-martensitic transformations, as product phase and the parent phase then
for example in the iron – palladium alloys; coexist and apparently the transformation
the Pd-rich f.c.c. phase transforms on cool- goes through a two-phase region, the two
ing first to an f.c.t phase and at lower tem- regions being separated by a blurred or
peratures to a b.c.t. phase. The f.c.c.-to- planar interface. Such observations do not,
f.c.t. transformation, although sometimes of course, facilitate the distinction between
regarded as martensitic, shows all the char- quasi-martensite and true martensite.
acteristics of a quasi-martensitic transfor-
mation.
Because the formation of each single do- 9.7 Shear Transformations
main is associated with a shape change and
thus with accommodation stresses, the ap- In this section we discuss a special group
plication of an external stress to the trans- of phase transformations, the so-called
formed product will result in a macro- shear transformations or polytypic transi-
scopic shape change. As the domain boun- tions, which strictly belong to the marten-
daries, which for the Mn-based alloys coin- sitic transformations. According to Verma
cide with the antiferromagnetic boundar- and Krishna (1966), “polytypism may be
ies, are mobile, the banded structure will defined, in general, as the ability of a sub-
614 9 Diffusionless Transformations
Table 9-3. The crystallographic observables of the martensitic transformations in some metals and alloys
(courtesy G. Guénin et al. 1979*).
Alloy system Structural change Composition wt.% Orientation relationship Habit plane
* Gobin, P. F., Guénin, G., Morin, M., Robin, M. (1979), in Transformations de Phases à l’État Solide-Trans-
formations Martensitiques. Lyon: Dep. Gènie Phys. Mat., INSA
by shear on those planes and directions martensite would be a vector of the same
specified in the orientation relationships. size and the same orientation in the parent
However, it was found that the shear mech- phase before transformation. As indicated
anisms proposed by the K – S and N – W in Sec. 9.3.1, the macroscopic shape
relations are not consistent with these ex- change of an invariant-plane transforma-
perimental observations. The observations tion can be represented by a combination
made by Greninger and Troiano (1949) on of a pure lattice deformation (B), the so-
Fe–22% – Ni–0.8% C martensite were the called Bain strain, a rigid lattice rotation
key to the mathematical development of (R), and an inhomogeneous lattice-invari-
the crystallographic theory of martensite. ant deformation (P). The pure lattice defor-
They found that martensite plates exhibited mation either increases or decreases some
a surface relief that can be described by a vectors in length. According to Wayman
homogeneous shear along the habit plane, (1964), “the essence of the crystallographic
but this homogeneous shear could not theory of martensitic transformations is to
transform the f.c.c. lattice of the parent find a simple shear (of a unique amount, on
phase into the b.c.t. lattice of the marten- a certain plane, and in a certain direction)
site. If the f.c.c. lattice had undergone the such that vectors which are increased in
same homogeneous deformation, the struc- length due to the lattice deformation are
ture of the martensite would have been correspondingly decreased in length due to
trigonal. They therefore suggested that two the simple shear, and vice versa. Such vec-
types of shear are involved in the marten- tors which remain invariant in length to
sitic transformation: a “first” simple shear these operations define the potential habit
which is responsible for the macroscopic plane. Physically speaking, the ellipsoid
shape change, and a “second” shear which generated from the initial sphere by the lat-
needs to be added to obtain the structural tice deformation is distorted by the simple
change but which should produce no ob- shear into another ellipsoid which becomes
servable macroscopic change in shape. tangential to the initial sphere, the points of
Two years later, Bowles (1951) showed tangency being related along a diameter.”
that the shape deformation may be any in- This is clearly illustrated in Fig. 9-23,
variant plane strain. This opened the way to where the problem becomes two-dimen-
the formulation of the general theory of the sional, because one of the principal axes of
crystallography by Wechsler et al. (1953) the lattice deformation is taken as normal
and, independently, by Bowles and Mack- to the plane of shear.
enzie (1954). Almost equivalent theories Some complementary remarks concern-
were later developed by Bullough and ing the crystallographic theory should be
Bilby (1956) and Bilby and Frank (1960). made. The input data for the calculations
The reader may consult the following more are (i) the lattice parameters of the parent
elaborate reviews of these theories: Way- and martensite phases, (ii) the lattice corre-
man (1964), Christian (1965), Nishiyama spondence, and (iii) the lattice-invariant
(1978) and Ahlers (1982). shear. The output of the calculations is then
The basic assumption in the crystallo- the amount of inhomogeneous shear re-
graphic theories is that the interface be- quired to obtain the invariant plane condi-
tween the product and the parent phases is tion, the macroscopic shape change, and
undistorted, which means that any vector the orientation relationship. Because of the
that lies in this interface on the side of the lattice symmetries, differently oriented
622 9 Diffusionless Transformations
Cu – Al – Ni alloy, but they were unable to even at very low temperatures. This obser-
observe any ledges or steps. The boundary vation shows that the interface migration
is always associated with dark strain con- must be accomplished without appreciable
trast, and the lattice is continuous through thermal activation. The interface is thus
the irrational boundary. Nishida and Li “glissile”. In searching for models to ex-
(2000) also carried out extensive studies on plain the structure and mobility of the
·011Ò type II twin boundaries in TiNi using interface, we are concerned with the ideal
HRTEM, but they did not observe ledges or and the actual interface morphology. A
steps either. Based on these experimental careful experimental analysis of the inter-
results, it is thus most likely that the type II face structure is therefore required if we
thin boundary is irrational even on a micro- want to verify the various models that have
scopic scale, and the strains at the boun- been proposed. As the models treat the
dary are elastically relaxed with wide twin interface on an atomistic scale, the sub-
width. To confirm this interpretation. Hara structure of the interface should be studied
et al. (1998) carried out computer simula- by conventional and by high-resolution
tions by using the molecular dynamics transmission electron microscopy. The
method. The result showed that the irra- same holds for martensite-to-martensite
tional thin boundary did not show any interfaces, which in some alloys are also
steps. Thus, the above interpretation for an mobile. Recently, atomistic imaging of the
irrational twin boundary is justified. Ot- martensite/austenite and martensite/mar-
suka and Ren (1999) have pointed out tensite interfaces have been obtained. It is
again the importance of type II twinning in therefore not surprising that both aspects,
the crystallographic aspects of martensite. the observation of interface substructures
They also stress the role that martensite and the atomistic models, are treated
aging has on the rubber-like behavior of jointly in the literature. For further reading
martensite, a point that has been a long- concerning the interface structures and the
standing unsolved problem. They showed growth mechanism of martensite we refer
that the point defects play a fundamental to the review papers by Christian (1982),
role. The deformation mechanisms of the Christian and Knowles (1982), and Olson
cold deformation of NiTi martensite have and Cohen (1986). A summary of these pa-
been thoroughly analyzed by Liu et al. pers is given below.
(1999 a, b). They also found an interplay Let us first introduce the kinds of mar-
between type I and type II twinning. tensite interfaces concerned: glissile and
As already mentioned in Sec. 9.3.6, a non-glissile martensitic interfaces, with the
better and more complete agreement can be latter subdivided into the coherent and
achieved when the strain energy terms, semi-coherent interfaces. The two struc-
both bulk and interfacial, are included in tures, martensite and the parent phase, are
the calculation. said to be “fully coherent” if both lattices
have a matching plane parallel to the inter-
face. If a fully coherent interface is dis-
9.8.1.5 Structure of the Habit Plane
played, the crystal undergoes a shape de-
In a number of alloys, especially those in formation leaving all vectors in the inter-
which the so-called thermoelastic marten- face invariant. In general, the two phases
sites are formed, the interface between do not have a plane of atomic fit, so that
martensite and the parent phase is mobile, fully coherent martensite interfaces are ex-
624 9 Diffusionless Transformations
ceptional. A fully coherent martensitic in- electron micrographs show that the inter-
terface is, for example, that between f.c.c. face between martensite and the parent
and h.c.p. structures with lattice parame- phase and also the intervariant interfaces
–
ters such that a (f.c.c.) = ÷2 a (h.c.p.); the and the interfaces between the internal
atomic arrangement in the basal planes, twins in one martensite plate contain dis-
which constitute the interface between the continuities (“steps”) on an atomic scale,
two structures, are identical. Such transfor- the nature of which has not yet been com-
mations are found in Co and its alloys and pletely unravelled. These steps can be con-
in some Fe-based alloys. The situation at sidered as resulting from a small deviation
semi-coherent interfaces becomes more of the ideal habit plane, and would then be
complex. The models predict the presence comparable to those observed along the
of dislocations to correct the mismatch interfaces of tapered twins.
along the interface. If this coherent inter- An exact understanding of the structure
face moves, it is suggested that not all vec- of the interfaces involved in the martensitic
tors are left invariant and that the move- transformation (the parent – martensite, the
ment of dislocations causes shear in the intervariant, and the twin/twin interfaces)
product phase. Fig. 9-24 shows the slip as- is therefore essential in determining the
sociated with the interface dislocations. mechanism of transformation and the mo-
Internally twinned martensite has been bility of the interfaces.
reported to show a zig-zag parent – marten-
site interface, as observed by conventional
9.8.2 Thermodynamics and Kinetics
electron microscopy in, for example,
of the Martensitic Transformation
Ti – Mn and Cu – Al – Ni. Fine parallel stria-
tions have been observed in the interface
9.8.2.1 Critical Driving Force
between austenite and both the b¢-type and
and Transformation Temperatures
the g¢-type Cu – Al – Ni-martensite. These
striations have been accounted for in terms A quantitative thermodynamic treatment
of interfacial dislocations resulting from of the martensitic transformation requires a
random faulting on the basal plane of the precise knowledge of the thermodynamic
b¢-type martensite and the twinning planes equilibrium temperature T0 and of the
of the g¢-type martensite. High-resolution change in Gibbs energy at the transforma-
9.8 Martensitic Transformations 625
tion temperature Ms . Both can be in the g-phase. Because of the shape and
calculated and/or derived from measured volume changes associated with the trans-
data, as is shown here for two examples: formation, elastic strain energy also has to
the martensitic transformation in Fe – X be considered. The Gibbs energy is thus
(X = Ni, Ru, …) and in Cu – Zn – Al alloys. composed of chemical Gibbs energy, Gc ,
In the former example, both the parent and and strain energy, Ee , so that the Gibbs en-
the martensite phases have the same struc- ergy change accompanying the transforma-
ture as the equilibrium phases and hence tion may be written as
the data for the equilibrium phases can be
taken. In the latter example, both structures Gg Æ M = DGcg Æ a + DEea Æ M (9-10)
differ from that of the equilibrium phases, At temperatures below Ms , where both
which requires a more elaborate calculation. phases coexist and thus are in equilibrium,
For the Fe – X alloys, the Gibbs energy DG g Æ M | T = 0. DGcg Æ a | T is then exactly
per mole of the parent austenite phase g equal, but opposite in sign, to the sum of all
and of the martensite phase a are G g and non-chemical energies DGnc aÆM
| T . If the
G a, respectively. The change in Gibbs en- surface energies are neglected in compari-
ergy per mole, DG g Æ a, which for a mar- son with the high strain energies, the non-
tensitic transformation g Æ a is available chemical energy equals DEea Æ M | T , and
to the system at any temperature T, is then approaches zero at T = Ms . The strain en-
ergy stored in the material is the sum of
DG g Æ a | T = G a – G g (9-8) that produced by shearing and by volume
change. The former depends on the
This quantity is negative for temperatures
strength of the parent phase and thus also
at which the a-phase is the more stable and
on the grain size, hence Ms also depends on
positive for temperatures at which the g-
the grain size, as shown by Hsu and Xiao-
phase is the more stable. There is a charac-
wang (1989).
teristic temperature T0 corresponding to the
The necessary undercooling (T0 – Ms )
thermodynamic equilibrium between both
and superheating (As – T0 ) vary for differ-
phases, such that
ent alloy systems, and for certain materials
DG g Æ a | T = T0 = 0 (9-9) even with composition. A precise thermo-
dynamic definition of Ms and As cannot be
Because the transformation creates interfa- given, however, if the non-chemical ener-
aÆM
cial and elastic energies, the martensitic gies, DGnc , are not known. We can then
transformation g Æ a or a Æ g does not only say that Ms or As is the temperature at
start at T0 , but at a temperature below or which the quantity DG g Æ a (at T = Ms or
above T0 , respectively. It is therefore nec- T = As , respectively) is sufficiently nega-
essary to undercool or overheat, respec- tive or positive, respectively, to have a rea-
tively, until Ms or As is reached. At these sonable chance of nucleation.
temperatures the Gibbs energy change Two approaches are found in the litera-
DG g Æ a is sufficiently large to induce the ture for calculating the critical chemical
forward or reverse transformation, respec- driving force. The first is based on the ex-
tively. DG g Æ a (at Ms ) is then the critical perimentally determined Ms temperatures
chemical driving force. (Kaufman, 1965) and the other on a theo-
The martensite phase, represented by M, retical model for the non-chemical energies
is to be regarded as the a-phase embedded (Hsu, 1985).
626 9 Diffusionless Transformations
a a
The first approach has been used for fer- where GFe and GA are the Gibbs energies of
rous alloys, which can be classified into pure iron and pure element A as a b.c.c. a-
a
two systems: those with g-loops and those phase, respectively, and Gm is the Gibbs
with stabilized g-phases (see Fig. 9-25). energy of mixing of the martensite phase.
Figure 9-25 also gives the Ms temperatures The change in chemical Gibbs energy ac-
for the a- and e-martensite. If A is the al- companying the martensitic transformation
loying element for iron, the molar chemical g Æ a then becomes
Gibbs energy for the austenite phase (G g ) gÆa gÆa gÆa
DG = (1 – x) DGFe + x DGA
can be written as gÆa
g g g g + DGm (9-13)
G = (1 – x) GFe + x GA + Gm (9-11)
gÆa
The quantity DGFe represents the Gibbs
where x represents the atomic fraction of energy change for transformation g Æ a of
the element A in solid solution in the g- pure iron and can be assessed experimen-
austenite, (1 – x) the atomic fraction of tally from the measured heat of transforma-
g
iron, GFe the chemical Gibbs energy of tion and the specific heat of both phases.
g
pure iron as f.c.c. g-phase, GA the chemical gÆa
The quantity DGA cannot usually be ob-
Gibbs energy of element A as f.c.c. phase, tained from experiments because the ele-
g
and Gm the Gibbs energy of mixing of the ment A does not always exist in the two
g-phase. Similarly, the Gibbs energy of the modifications g and a; it must therefore be
a-phase can be given as estimated from thermodynamic models for
a a a a gÆa
G = (1 – x) GFe + x GA + Gm (9-12) solid solutions. The quantity DGm is the
9.8 Martensitic Transformations 627
Gibbs energy can, according to Hsu and It is known that martensite may also be
Zhou (1989), be described as induced by an external stress at tempera-
tures above Ms . The problem now is to cal-
DG b¢ – M = DG b¢ – b + DG b – a + DG a – a¢ culate the change in T0 due to changes in
+ DG a¢ – M (9-14) stress. As a first approximation, it is as-
PÆM
sumed that the driving force DGm | T = Ms
where b¢ – M represents the transformation
required for nucleation remains constant
from the the ordered b.c.c. phase to the or-
with temperature and thus independent of
dered 9 R-type of martensite, b¢ – b the or-
stress. Patel and Cohen (1953) calculated
der – disorder transition, b – a the transfor-
the work done on the stressed specimen;
mation from the disordered b.c.c. to the
their treatment provides a good under-
disordered f.c.c. phase having the same
standing of how an applied stress that is de-
composition, a – a¢ the disorder – order
composed into a shear stress along the
transformation in the f.c.c. phase, and
habit plane and a normal stress perpendicu-
a¢ – M the transition from the ordered f.c.c.
lar to it, affects the transformation temper-
phase to the ordered martensite phase. As-
ature. At Mss , which is the martensite start
suming a simplified relationship between
temperature when cooling under an applied
the degree of ordering and temperature,
stress s , the chemical Gibbs energy change
Hsu and Zhou found good agreement
equals the transformation work of the ex-
between the calculated and observed Ms .
ternal stress:
Their calculations show that ordering of
the parent phase, which cannot be sup- DGsP Æ M = 1/2 s a (9-15)
pressed even by severe quenching, strongly ¥ [d 0 sin 2 q ± e0 (1 + cos 2 q )] Vm
influences T0 .
where d 0 is the shear strain, s a the applied
stress, q the angle between the stress axis
and the normal to the operative shear plane,
e0 the corresponding strain associated with
the transformation, and Vm the molar vol-
ume. The quantity DH P Æ M can be meas-
ured by calorimetry and DS P Æ M can be
evaluated from stress-induced transfor-
mation experiments or calculated from Eq.
(9-16). The temperature T0 can be calcu-
lated thermodynamically or obtained more
or less accurately from the relationships
DG P Æ M = DH P Æ M – T DS P Æ M
(9-16)
DH P Æ M = T0 DS P Æ M
and
T0 = (As – Ms )/2 = (Af – Mf )/2 (9-17)
Figure 9-27. Gibbs energy G* versus temperature
and force for stressed samples: P and M represent However, it should be noted that the deter-
free energy surfaces for parent and martensite, re- mination of T0 does not always obey these
spectively (Wollants et al., 1979). simple relationships and that the calorimet-
9.8 Martensitic Transformations 629
Cooling
Figure 9-34. Schematic representation of some relevant features (volume-transformed product or transforma-
tion strain) experimentally observed in hysteresis curves corresponding to thermally induced and stress-induced
thermoelastic transformations: (a, e) single interface transformation in a single crystal; (b, f) multiple interface
transformation; (c, g) discontinuous jumps (bursts), (d, h) partial cycling behavior.
634 9 Diffusionless Transformations
tensite plates is then a direct function of the ing the control parameter, return to their
increase or decrease in stress. More elab- original values, a closed loop is formed
orate thermodynamic treatments of ther- (Fig. 9-34). The loop is always contoured in
moelasticity can be found in the papers by such a sense that it encloses a positive area,
Delaey et al. (1974), Salzbrenner and Co- representing the energy lost in the cyclic
hen (1979), Ling and Owen (1981) and Or- process. Therefore, hysteretic behavior is
tin and Planes (1989). always related to an energy-dissipative pro-
The best quantitative understanding of cess. The dissipated energy is much smaller
the kinetics of the martensitic transforma- in thermoelastic martensitic transforma-
tion is obtained from isothermal transfor- tions than in burst-type transformations.
mations, because they permit both the nu-
cleation and the transformation rates to be
determined. In those alloys exhibiting iso- 9.9 Materials
thermal martensite (Thadhani and Meyers,
1986), it is shown that at each temperature
9.9.1 Metallic Materials
the transformation starts in the austenite
and proceeds as a function of time. The A classification of the diffusionless dis-
transformation exhibits a C-curve behav- placive transformations in metallic materi-
ior. Isothermal martensitic transformation als is given in Table 9-4, where the alloy
kinetics consist of two effects: an initial in- systems are subdivided into three groups.
crease in the total volume fraction of mar- The origin of the martensitic transforma-
tensite, which is attributed to an autocata- tion in the first group lies in the allotropic
lytic nucleation of new martensite plates, transformation of the pure solvent. The
followed by a decrease due to the compart- parent phase of the alloys of this group thus
mentalization of the austenite into smaller does not show any remarkable mechanical
and smaller areas. instability. The second group consists of
the b b.c.c. Hume – Rothery alloys, which
are characterized by a moderate lattice in-
9.8.2.4 Transformation Hysteresis
stability in the temperature range above
Hysteresis behavior is one of the pecu- Ms . The third group is characterized by a
liar characteristics of both the thermal and drastic mechanical instability of the parent
stress-induced martensitic transformations. phase. Because the transformation is only
In several studies the origin of the fric- weakly first order (by this we mean a dis-
tional resistance opposing the interfacial continuous jump in the corresponding ther-
motion of martensite plates has been inves- modynamic property whose height, how-
tigated and described. From a practical ever, is very small) or even second order, it
point of view, the hysteresis phenomenon is in this group of alloy systems that we
is an important problem in the application find, in addition to the martensitic, the
of shape-memory alloys. In general hyster- quasi-martensitic transformations.
esis appears when, on passing through a lo- Traditionally, the ferrous and non-fer-
cal extreme value (maximum or minimum) rous martensites have been treated separ-
of any control parameter such as tempera- ately in the literature. Before going into de-
ture or stress, one or more state variables tail it will be an advantage to first compare
do not follow the original path in state and contrast ferrous and non-ferrous mar-
space. When all the state variables, includ- tensites and to do it in such a way that we
9.9 Materials 635
Table 9-4. Classification of metallic alloy systems showing diffusionless displacive transformations (Delaey
et al., (1982).
Table 9-5. A qualitative comparison between ferrous and non-ferrous martensites (Delaey et al., 1982b).
High values near the Ms Elastic constants of Low values near the Ms
the parent phase
Table 9-6. Summary of substructure, habit plane cant deviations from the above equation.
(H.P.) and orientation relationship (O.R.) for the four The tetragonality is abnormally lower for
types of a¢-martensite (Maki and Tamura, 1987).
X = Mn or Re and abnormally higher for
Morphology Substructure H.P. O.R.* Ms X = Al or Ni. Heating to room temperature
of the latter martensite results in a lowering
Lath (Tangled) (111)A K–S High of the tetragonality. The formation of do-
dislocations mains or microtwinning in the former al-
Butterfly (Straight) (225)A K–S
loys and ordering of the Al atoms in the lat-
dislocations ter have been put forward as the origin of
and the abnormal c/a ratio. This behavior has
twins moreover been related to the martensite
plate morphology (Kajiwara et al., 1986,
Lenticular (Straight) (259)A N
dislocations
1991). Kajiwara and Kikuchi found that in
and or or Fe – Ni – C alloys the tetragonality is abnor-
twins (3 10 15)A mally large and depends on the microstruc-
(Mid-rib) G–T ture. It is very large for a plate martensite,
while it is normal or not so large for a len-
Thin-plate Twins (2 10 15)A G–T Low
ticular martensite. They conclude that “the
* K–S: Kurdjumow–Sachs relationship, N: Nishiyama martensite tetragonality is dependent on
relationship, G–T: Greninger–Troiano relationship the mode of the lattice deformation in the
martensitic transformation. If the lattice
deformation is twinning, the resulting c : a
A distinct substructure, crystallographic is large, while in the case of slip it is small”
orientation of the habit plane and austenite (Kajiwara and Kikuchi, 1991).
to martensite orientation relationship are
associated with each morphology, as sum-
9.9.1.2 Non-Ferrous Alloys
marized in Table 9-6.
Because the carbon atom occupies octa- A classification base of the non-ferrous
hedral interstices in the austenite f.c.c. lat- alloy systems exhibiting martensite is
tice, special attention is drawn to the
Fe – X – C martensite. In the martensite lat-
tice, those interstitial positions are defined
by the Bain correspondence (Fig. 9-35).
Only those at the midpoints of cell edges
parallel to [001]B and at the centers of the
faces normal to [001]B are permitted. This
preferred occupancy affords an explanation
of the observed tetragonality c/a, the de-
gree of which is a function of the carbon
content:
c/a = 1 + 0.045 (wt.% C) (9-25) Figure 9-35. Schematic representation of the Bain
correspondence for the f.c.c. to b.c.t. transformation.
Careful X-ray diffraction of martensite, The square symbols represents the possible occupied
freshly quenched and maintained at liquid positions of the interstitial carbon atoms (Muddle,
nitrogen temperature, has shown signifi- 1982).
638 9 Diffusionless Transformations
the g¢1 – b≤1 , and finally the b≤1 – a¢1 marten- bohedral (the so-called R-phase) structural
site. change. At still lower temperatures the
The other typical example of the second rhombohedral R-phase transforms into a
group is the Ni – Ti-based alloy system; monoclinically distorted martensite. The
both prototype alloy systems constitute the R-phase also forms displacively and can be
shape-memory alloys (SMA). The occur- stress-induced, and shows all the character-
rence of a so-called “pre-martensitic” R- istics of a reversible transformation.
shape has long obscured the observations. Concerning the third group, in only a
The review by Wayman (1987), illustrates few cases, as in In – Tl, has definite proof
the complexity of the transformation be- been provided to justify the conclusion that
havior. During cooling, the high-tempera- the transformation is martensitic. Most of
ture ordered b.c.c. phase (P) transforms the transformations in these systems have
first to an incommensurate phase (I) and on to be classified as quasi-martensitic.
further cooling to a commensurate phase
(C), and finally to martensite. The P-to-I
9.9.2 Non-Metals
transformation is second order, whereas the
I-to-C transformation is a first-order phase Inorganic compounds exhibit a variety
transformation involving a cubic-to-rhom- of crystal structures owing to their diverse
Inorganic compounds
Alkali and ammonium halides MX, NH4X (NaCl-cubic ¤ CsCl-cubic)
Nitrates RbNO3 (NaCl-cubic ¤ rhombohedral ¤ CsCl-cubic)
KNO3, TlNO3, AgNO3 (Orthorhombic ¤ rhombohedral)
Sulfides MnS (Zinc-blende-type ¤ NaCl-cubic)
(Wurtzite-type ¤ NaCl-cubic)
ZnS (Zinc-blende-type ¤ wurtzite-type)
BaS (NaCl-type ¤ CsCl-type)
Minerals
Pyroxene chain silicates Enstatite (MgSiO3) (Orthorhombic ¤ monoclinic)
Wollastonite (CaSiO3) (Monoclinic ¤ triclinic)
Ferrosilite (FeSiO3) (Orthorhombic ¤ monoclinic)
Silica Quartz (Trigonal ¤ hexagonal)
Tridymite (Hexagonal, wurtzite-related)
Cristobalite (Cubic ¤ tetragonal, zinc blende-related)
Ceramics
Boron nitride BN (Wurtzite type ¤ graphite-type)
Carbon C (Wurtzite type ¤ graphite)
Zirconia ZrO2 (Tetragonal ¤ monoclinic)
Organics
Chain polymers Polyethylene (CH2–CH2)n (Orthorhombic ¤ monoclinic)
Cement
Belite 2 CaO · SiO2 (Trigonal ¤ orthorhombic ¤ monoclinic)
640 9 Diffusionless Transformations
Figure 9-40. Temperature-actuated switch designed so that it opens or closes above a particular temperature.
From “Shape Memory Alloys” by L. McDonald Shetky. Copyright (1979) by Scientific American, Inc. All
rights reserved.
ously to its “deformed” positions, thus in the third deformation step, because it
opening or closing the electrical contacts may not exceed the maximum strain that
on cooling. This repeated cycling, defor- can be recovered by the phase transforma-
mation in martensitic condition followed tion itself. Because these strains are asso-
by a heating – cooling, is called “training”. ciated with the martensitic transformation,
We can thus induce two-way memory by the maximum amount of recoverable strain
using a training procedure. is bound to the crystallography of the trans-
A further comment should be made here formation. Exceeding this amount of de-
concerning the shapes that can be remem- formation in the third fabrication step will
bered. We have to distinguish three shaping automatically result in unrecoverable de-
procedures: the fabrication step from raw formation.
material towards, for example, a coiled Many examples of shapes that can be re-
wire such as for the antenna, the fabrica- membered are possible. A flat SME speci-
tion of the “to be remembered position”, men can elongate or shorten during heat-
such as the additional shaping for the actu- ing, can twist clockwise or counter-clock-
ator, and the final deformation in the mar- wise, and can bend upwards or downwards.
tensitic condition, such as the bending of An SME spring can expand or contract dur-
the actuator. The first two fabrication steps ing heating. All this depends on the second
involve only classical plastic deformation and third fabrication steps.
and, therefore, the type and degree of de- What happens now if, for one reason or
formation are in principle not limited, pro- another, the specimen is restrained to ex-
vided that the material does not fail. The hibit the shape-memory effect? For exam-
degree of deformation, however, is limited ple, what happens if an expanded ring is
9.10 Special Properties and Applications 645
its deformed position. The elastic energy mainly because of economic factors and
that has been stored in the bias spring dur- the reliability of the material. The three al-
ing the heating cycle is now released, al- loy systems are Ni – Ti, Cu – Zn – Al and
lowing the plate to perform work also dur- Cu – Al – Ni. Generally, other elements are
ing the cooling cycle. To describe fully added in small amounts (of the order of a
such a working performing cycle, a ther- few weight %) in order to modify the trans-
modynamic treatment is needed (Wollants formation temperatures or to improve the
et al., 1979). The working performing mechanical properties or the phase stabil-
cycle can best be illustrated by taking a ity. In all three cases the martensite is
shape-memory spring that expands or con- thermoelastic. Maki and Tamura (1987)
tracts during heating or cooling and that reviewed the shape-memory effect in fer-
carries a load. The working performing cy- rous alloys, where a non-thermoelastic
cle can then be represented in a displace- Fe – Mn – Si alloy has also been found to
ment – temperature, a stress – temperature, show a shape-memory effect, and commer-
or an entropy – temperature diagram. cialization is being considered. The most
Although the shape-memory effect has important properties of shape-memory al-
been observed in many alloy systems, only loys are summarized in Fig. 9-43, in which
three systems are commercially available, the working temperatures, the width of the
Figure 9-43. Schematic representation of the most relevant shape-memory properties (courtesy Van Hum-
beeck, 1989).
9.10 Special Properties and Applications 647
then a result of plastic deformation of the some new approaches are also explained.
parent phase and of transformation plastic- Most of the information referred to in this
ity. Strain hardening and enhancement of section was presented at the most recent
nucleation of martensite also play an essen- ICOMAT international conference on mar-
tial role. When martensite is formed during tensitic transformations held in 1998 at
tensile deformation, the strain hardening Bariloche (Ahlers et al., 1999).
becomes large. Necking is then expected New directions in martensite theory are
to be suppressed, explaining the enhanced presented by Olson (1999). The nucleation
uniform elongation. Fig. 9-46 shows, as an of martensite, the growth of a single mar-
example, the amount of martensite, the tensite plate, the formation of, for example,
elongation and the ultimate strength meas- self-accommodating groups of martensite
ured after tensile tests of a TRIP steel as a plates, and this within single crystals of the
function of temperature, clearly illustrating parent phase as well as in polycrystalline
the enhanced elongation, especially in the material, and the constraints dictated by the
temperature range between Mss and Md . components where martensitic materials
Such a large elongation (sometimes over are only one (maybe the most important)
200%) can also be produced by subjecting functional element of the component, are
a TRIP steel specimen under constant load all influenced by different interactive lev-
to thermal cycles through the transforma- els of structures (ranging from solute atoms
tion temperature. to components). Nucleation is the first step
in martensite life, and a component whose
functional properties are attributed to those
9.11 Recent Progress in the of martensite, can be considered the final
Understanding of Martensitic step. Olson (1999) constructed a flow-
Transformations block diagram in which the martensitic
transformation is situated in a multilevel
We draw attention here to some recent dynamic system. This new system, shown
papers that illustrate recent progress in the in Fig. 9-47, and the one given in Fig. 9-1,
understanding of martensite, and in which offer powerful tools for a better under-
Kakeshita, T., Saburi, T., Shimizu, K. (1999), Mater. lin, D. E., Sekerka, D. F., Wayman, C. M. (Eds.).
Sci. Eng. A 273 – 275, 21. Warrendale: Met. Soc. AIME, p. 1145.
Kaufman, L. (1965), in: Physical Properties of Mar- Olson, G. B., Cohen, M. (1986), in: Dislocations in
tensite and Bainite: Special Report 93. London: Solids: Nabarro, F. R. N. (Ed.). New York: Else-
The Iron & Steel Inst., p. 48. vier Science Publishers, p. 295.
Kaufman, L., Cohen, M. (1958), Progr. Metal Phys. Olson, G. B., Hartman, H. (1982), Proc. ICOMAT
7, 165. ’82, J. de Physique 43, 855.
Krauss, G. (1992), in: Materials Science and Tech- Oota, A., Muller, J. (1987), Proc. ICOMAT ’86. Sen-
nology, Vol. 7: Pickering, F. B. (Ed.). Weinheim: dai: Japanese Institute of Metals, p. 1121.
VCH, p. 1. Ortin, J., Planes, A. (1989), Acta Metall. 37, 1433.
Krauss, G., Marder, A. R. (1971), Metall. Trans. 2, Osmond, M. F. (1895), Arts Chimiques 94, 480.
2343. Otsuka, K. (1986), Proc. ICOMAT ’86. Sendai: Japa-
Kriven, W. M. (1982), in: Proceedings of Solid to nese Institute of Metals, p. 35.
Solid Transformations. New York: AIME, p. 1502; Otsuka, K., Shimizu, K. (1986), Int. Metals. Rev. 31,
and in: International Summer Course on Marten- 93.
sitic Transformations. Leuven: Dep. Met. & Mat. Otsuka, K., Wayman, C. M. (1977), in: Reviews on
Eng., KV Leuven, p. 9.1. the Deformation Behavior of Materials, Feltham,
Kriven, W. M. (1988), J. Am. Ceram. Soc. 71, 1021. P. (Ed.). Intl. Quarterly Scient. Reviews Journal II,
Kurdjurnov, G. V. Sachs, G. (1930), Z. Phys. 64, 325. 81.
Lemons, R. A., Coldren, L. a. (1978), Rev. Sci. In- Otsuka, T., Ren, X. (1999), Mater. Sci. Eng. A 273 –
strum. 49, 1650. 275, 89.
Liao, K. C., Allen, C. W. (1982), in: Solid-State Patel, J. R., Cohen, M. (1953), Acta Metall. 1, 531.
Phase Transformations: Aaronson, H. I., Laugh- Pops, H. (1970), Metall. Trans. 1, 251.
lin, D. E., Sekerka, D. F., Wayman, C. M. (Eds.). Pelton, A. (2001) in: Phase Transformations in Ma-
Warrendale: Met. Soc. AIME, p. 1493. terials: Kostorz, G. (Ed.). Weinheim: Wiley-VCH,
Ling, H. C., Owen, W. s. (1981), Acta Metall. 29, Chap. 1.
1721. Purdy, G. R., Bréchet, Y. J. M. (2001), in: Phase
Liu, Y., Xia, Z., Van Humbeeck, J. (1999 a), Mater. Transformations in Materials: Kostorz, G. (Ed.).
Sci. Eng. A 273 – 275, 673. Weinheim: Wiley-VCH, Chap. 7.
Liu, Y., Xia, Z., Van Humbeeck, J., Delaey, L. Rao, C. N. R., Rao, K. J. (1978), Phase Transitions in
(1999 b), Mater. Sci. Eng. A 273 – 275, 679. Solids. New York: McGraw-Hill.
Maki, T., Tamura, I. (1987), Proc. ICOMAT ’86. Sen- Richman, R. H., Zimmerly, C. A., Inal, O. T., Hodg-
dai: Japanese Institute of Metals, p. 963. son, D. E., Rao, A. S. (1994), in: Proceedings
Massalski, T. B. (1984), Metall. Trans. 15 A, 421. International Conference on Shape Memory Alloys
Muddle, B. C. (1982), in: Proceedings of Solid to and Superplastic Technologies SMST ’94, Pelton,
Solid Transformations. New York: AIME, p. 1347. A., Hodgson, D., Duerig, T. (Eds.). Pacific Grove,
Mukherjee, K. (1982), in: International Summer CA, USA: SMTS Proceedings, p. 175.
Course on Martensitic Transformations. Leuven: Roitburd, A. L. (1988), in: Phase Transformations ’87:
Dep. Met. & Mat. Eng., KV Leuven, p. 4.1. Lorimer, G. W. (Ed.). Institute of Metals, p. 414.
Nakanishi, N. (1979), Progr. Mat. Sci. 24, 143. Rühle, M., Kriven, S. (1982), Proc. Phase Transf.
Nakanishi, N., Nagasawa, A., Murakami, Y. (1982), Confer. AIME, p. 1569.
in: ICOMAT ’82: Editions de Physique C 4, p. 35. Saburi, T., Nenno, S. (1987), Proc. ICOMAT ’86.
Nakanishi, N., Shigematsu, T., Machida, N., Ueda, Sendai: Japanese Institute of Metals, p. 671.
K., Tanaka, H., Inaba, T., Iwatsu, O. (1993), in: Salzbrenner, R. J., Cohen, M. (1979), Acta Metall.
International Conference on Martensitic Transfor- 27, 739.
mation ICOMAT ’92: Wayman, C. M., Perkins, J. Schetky McDonald, L. (1979), Sci. Am. 241, 74.
(Eds.). Carmel, CA, USA: Monterey Institute of Schroeder, T. A., Wayman, C. M. (1977), Acta Met-
Advanced Studies, p. 977. all. 25, 1375.
Nishida, M., Li, S. (2000), Mat. Sci. Forum 327/328, Shapiro, S. M. (1981), Metall. Trans. 12 A, 567.
103. Shibata, M., Ono, K. (1975), Acta Metall. 23, 587.
Nishiyama, Z. (1934), Sci. Rept. Tohoku Univ. 23, p. Shibata, M., Ono, K. (1977), Acta Metall. 25, 35.
325. Shimizu, K., Kakeshita, T. (1989), ISIJ Int. 2, 97.
Nishiyama, Z. (1978), Martensitic Transformations. Sikka, S. K., Vohra, Y. K., Chidambaram, R. (1982),
New York: Academic Press. Progr. Mat. Sci. 27, 245.
Olson, G. B. (1999), Mater. Sci. Eng. A 273 – 275, Sugimoto, K., Nakaniwa, M. (2000), Mat. Sci.
11. Forum 327/328, 363.
Olson, G. B., Cohen, M. (1982 a), Metall. Trans. A 6, Tamura, I., Maki, T., Hato, H., Aburai, K. (1969), J.
791. Jpn Inst. Met. 33, 1383.
Olson, G. B., Cohen, M. (1982 b), in: Solid-State Tas, H., Delaey, L., Deruyttere, A. (1973), Metall.
Phase Transformations: Aaronson, H. I., Laugh- Trans. 4, 2833.
654 9 Diffusionless Transformations
Thadhani, N. N., Meyers, M. A. (1986), Progr. Mat. “International Conference on Martensitic Transfor-
Sci. 30, 1. mations ICOMAT ’79” (1979), Cambridge (MA):
Vallade, M. (1982), in: International Summer Course Dep. of Mat. Science and Techn., M.I.T.
on Martensitic Transformations. Leuven: Dep. “International Conference on Martensitic Transfor-
Met. & Mat. Eng., KV Leuven, p. 3.1. mations ICOMAT ’82” (1982), Delaey, L., Chan-
Van Humbeeck, J. (1977), J. Phys. (Paris) IV, 7-C5, 3. drasekaran, M. (Eds.), Journal de Physique 43,
Van Humbeeck, J. (1989), Dep. Met. & Mat. Eng., Conf. C-4.
University of Leuven, Leuven, unpublished. “International Conference on Shape Memory Al-
Van Humbeeck, J. (1999), Mater. Sci. Eng. A 273 – loys” (1986), Youyi Chu, Hsu, T. Y., Ko. T. (Eds.).
275, 134. Beijing: China Academic Publishers.
Van Tendeloo, G. (1989), University of Antwerp, “International Conference on Martensitic Transfor-
Antwerp, unpublished. mations ICOMAT ’86” (1987), Tamura, I. (Ed.).
Verma, A. R., Krishna, P. (1966), Polymorphism and Sendai: The Japan Inst. of Metals.
Polytypism. New York: Wiley. “The Martensitic Transformation in Science and
Vintaikin, E. Z., Litvin, D. F., Udovenko, V. A., Technology” (1989), Hornbogen, E., Jost, N.
Schcherebinskij, G. V. (1979), in: Proc. of ICO- (Eds.). Oberursel: DGM-Informationsgesellschaft
MAT ’79. Cambridge: MIT-Press, p. 673. Verlag.
Wadhawan, V. K. (1982), Phase Transitions 3, 3. “Shape Memory Materials” (1989), Otsuka, K., Shi-
Warlimont, H. (1989), Vacuumschmelze GmbH, Ha- mizu, K. (Eds.), Proceedings of the MRS Interna-
nau. tional Meeting on Advanced Materials, vol. 9.
Warlimont, H., Delaey, L. (1974), Progr. Mat. Sci. Pittsburgh: Mat. Research Soc.
18, p. 1. “International Conference on Martensitic Transfor-
Wayman, C. M. (1964), Introduction to the Crystal- mations ICOMAT ’89” (1990), Morton, A. J.,
lography of Martensitic Transformations. New Dunne, B., Kelly, P. M., Kennon, N. (Eds.) to be
York: MacMillan Comp. published.
Wayman, C. M. (1983), in: Physical Metallurgy: “International Conference on Martensitic Transfor-
Cahn, R. W., Haasen, P. (Eds.). New York: Else- mations ICOMAT ’92” (1993), Wayman, C. M.,
vier Scientific Publishers, p. 1031. Perkins, J. (Eds.). Carmel, CA, USA: Monterey In-
Wayman, C. M. (1987), Proc. ICOMAT ’86. Sendai: stitute of Advanced Studies.
Japanese Institute of Metals, p. 645. “International Symposium and Exhibition on Shape
Wechsler, M. S., Lieberman, D. S., Renard, T. A. Memory Materials (SME ’99)” (2000). Zurich:
(1953), Trans. AIME 197, 1503. Trans. Tech. Publications.
Williams, J. C., Thompson, A. w. (1981), in: Metal- “International Conference on Shape Memory and
lurgical Treatises: Tien, J. K., Elliott, J. F. (Eds.). Superelastic Technologies (SMST ’97)” (1997),
Warrendale: Met. Soc. AIME, p. 487. Pelton, A., Hodgson, D., Russell, S., Duerig, T.
Wollants, P., De Bonte, M., Roos, J. R. (1979), Z. (Eds.). Santa Clara, CA, USA: SMST Proceedings.
Metallkd. 70, 113. “Displacive Phase Transformations and Their Appli-
Yang, J., Wayman, C. M. (1993), in: International cations in Materials Engineering” (1998), Inoue,
Conference on Martensitic Transformation ICO- K., Mukherjee, K., Otsuka, K., Chen, H. (Eds.).
MAT ’92, Wayman, C. M., Perkins, J. (Eds.). Car- Warrendale, OH, USA: TMS Publications.
mel, CA, USA: Monterey Institute of Advanced “European Symposium on Martensitic Transforma-
Studies, p. 1187. tions ESOMAT ’97” (1997), Beyer, J., Böttger, A.,
Mulder, J. H. (Eds.). EDP-Sciences Press.
“European Symposium on Martensitic Transforma-
tions ESOMAT ’94” (1995), Planes, A., Ortin, J.,
Proceedings Lluis Manosa. Les Ulis, France: Les Editions de
of International Conferences Physique.
“International Conference on Martensitic Transfor-
on Martensitic Transformations mations ICOMAT ’98” (1999), Ahlers, M., Kos-
torz, G., Sade, M. (Eds.). Amsterdam: Elsevier.
“Physical Properties of Martensite and Bainite” “Martensitic Transformation and Martensite” (1980),
(1965), Special Report 93. London: The Iron and Hsu, T. Y. (Xu Zuyao) (Ed.). Shanghai: Science
Steel Institute. Press (in Chinese).
“Shape Memory Effects in Alloys” (1975), Perkins, “Pseudoelasticity and Stress-Induced Martensitic
J. (Ed.). New York: Plenum Press. Transformations” (1977), Otsuka, K., Wayman, C.
“New Aspects of Martensitic Transformations” M. (Eds.), in: “Reviews on the Deformation Be-
(1976), Suppl. to Trans. Jap. Inst. Met. 17. havior of Materials”.
“Martensitic Transformations ICOMAT ’77” (1978),
Kiev: Academy of Science.
10 High Pressure Phase Transformations
Martin Kunz
HT high temperature
IR infrared
LDA low-density amorphous phase of H2O
NMR nuclear magnetic resonance
RT room temperature
XAS X-ray absorption spectroscopy
10.1 Introduction 659
Equation (10-7) states an intuitive trivial- tions (coesite can be quenched to room
ity, i.e., that in order to minimize the Gibbs pressure), both structures exhibit nearly
energy upon increasing pressure, a given ideal Si–O–Si angles of between 143° and
system will reduce its volume. Volume re- 144°. The phase transition is thus structu-
duction is thus the ultimate driving force of rally characterized by building the tetrahe-
structural change and phase transformation dral framework in coesite more densely
as a reaction on increasing pressure. How- than in a-quartz. This can be demonstrated
ever, the way a volume is reduced can be by looking at the oxygen surroundings in
different, depending on the initial structure both structures: In a-quartz, each oxygen
and configurational entropy of the system has three oxygen neighbors at ~ 2.6 Å,
and also on the amount of pressure applied. which represent the tetrahedral oxygen
In the following we will focus on the three neighbors. The closest oxygen atoms from
most important mechanisms responsible any other tetrahedron are found at a dis-
for structural changes at high pressure. tance of ~ 3.5 Å. In coesite on the other
These are framework flexion, increase in hand, there are also three oxygen neighbors
coordination number and pressure-induced at 2.6 Å around each O, indicating the stiff
ordering. behavior of the SiO4 tetrahedrals. The next
nearest oxygen neighbors in coesite, how-
ever, are observed at a distance of only
10.2.1 Framework flexion
3.0 – 3.2 Å (room pressure). Both structure
A number of technologically important types react to applied pressure with a very
oxide materials (e.g., zeolites, perovskites) large decrease in the Si–O–Si angles (Fig.
can be viewed as being built of relatively 10-1), while the Si–O bond lengths remain
rigid corner-linked polyhedra forming a more or less constant. Densification within
rather flexible framework. These materials the stability field of quartz is thus achieved
are known to react on changing tempera- by flexion of the framework through rigid
ture and/or pressure mainly by flexion at rotations of the SiO4 tetrahedra, resulting
the polyhedral joints rather than by polyhe- in a decrease of the Si–O–Si angle from
dral compression, i.e., bond length reduc- 144° to 125°. If this reaches a limit, denser
tion (e.g., Hazen and Finger, 1978; Velde packing of the polyhedra is achieved by
and Besson, 1982; Hemley et al., 1994). framework reconstruction. Further densifi-
Silica (SiO2 ) can be viewed as a typical ex- cation, however, eventually involves an in-
ample of the behavior of framework struc- crease in the coordination number (see Sec.
tures under pressure. It has a rich phase di- 10.2.2). Thus, a pure compression at R T or
agram in P – T space. The ambient condi- a rapid (shock) compression of a-quartz
tion phase is a-quartz, which consists of leads to an amorphization involving both
interconnected spirals of corner-linked framework collapse and an increase in co-
SiO4 tetrahedra. Silica has two high-pres- ordination number (Hemley et al., 1994).
sure polymorphs, of which only coesite can Pressure-induced amorphization seems to
be viewed as a tetrahedral framework be a common feature of many framework
structure. Its framework is characterized by structures, indicating a volume reduction
four-fold rings of tetrahedra connected to by a collapse of the framework. This de-
chains, which in turn are packed via shared stroys the long-range order while still
tetrahedral corners to the densest known largely maintaining the short-range order
tetrahedral framework. At ambient condi- of the primary building blocks. A similar
662 10 High Pressure Phase Transformations
Figure 10-1. Si–O–Si angles (solid symbols) and average Si–O bond lengths (open symbols) vs. pressure for
quartz (squares) and coesite (diamonds). Note the (within experimental errors) constant ·Si–OÒ distances with in-
creasing pressure, while the Si–O–Si angles show a significant negative correlation. This demonstrates that frame-
work structures (such as tetrahedral SiO2 polymorphs) adapt their volume through framework flexion. Data from
Levien et al. (1980) and Levien and Prewitt (1981).
effect is also thought to be responsible for of the tetrahedral framework into a glassy
the high-density form of amorphization of tetrahedral arrangement.
ice (Mishima et al., 1984; see also Sec. If the channels and holes in a framework
10.5.3.6). are large enough, we can even obtain the
The most open frameworks known in in- seemingly paradox result of a negative
organic chemistry are adopted by the zeolite compressibility. This happens if the molec-
structure family. These structures exhibit ular size of the pressure medium is small
some peculiar effects upon compression. enough to be squeezed into the channels.
Zeolite frameworks are characterized by Because the pressure medium seems to be
large cages and channels, which can be oc- packed more densely within the zeolite
cupied by extra-framework cations or mole- structure than in the fluid, the total volume
cules. As in silica, volume reduction on zeo- of the system (crystal plus pressure me-
lite can be most easily achieved by a reduc- dium) decreases. The volume of the crystal
tion of the framework channels, which is alone, however, increases with applied
achieved through rigid body rotations of the pressure. In a similar way, Hazen (1983)
framework tetrahedra and thus bending of observed compressional anomalies for zeo-
the Si–O–Si angles. A study of the natural lite 4a, which exhibits different phase tran-
zeolite natrolite for example showed that sitions depending on the pressure medium
compression leads to a continuous reduction used. All of the observed high-pressure
of the unit-cell volume without any structu- phases showed higher compressibilities at
ral phase transition up to ~ 7 GPa (Belitsky high pressure.
et al., 1992). Above 7 GPa, the investigated A more frequent but still unusual phe-
samples underwent amorphization, similar nomenon observed upon compressing
to the transition observed for silica. This framework structures is a negative linear or
amorphization again indicates the collapse areal compressibility. Materials with these
10.2 Pressure-Driven Phase Transitions 663
properties will expand in one or two di- (i.e., framework flexion) with increasing
mensions upon hydrostatic compression. pressure. An example of this is the increas-
As shown by Baughman et al. (1998), this ing orthorhombic distortion of MgSiO3
also implies a negative Poisson’s ratio, i.e., with increasing pressure (Fiquet et al.,
lateral contraction upon uniaxial compres- 2000).
sion. These authors also demonstrate that
properties such as a negative Poisson’s ra-
10.2.2 Increase in coordination number
tio are in most cases linked to geometric
constraints in hinged framework structures. Many crystal structures are best de-
Because such behavior can result in in- scribed by a closest packing of anions with
creasing surface area upon increasing hy- some of the interstitial sites (two tetrahedra
drostatic pressure, it is of potential interest and one octahedron per anion) occupied by
to material scientists. In principle it is con- a cation. In this simple, but in many cases
ceivable to manufacture these compounds very successful picture, the large and soft
into porous composite material with zero anions are in contact with each other and
or negative volume compressibilities. the small, more rigid cations are isolated
Another example of a structure family from each other and in contact only with
whose phase transitions are characterized their surrounding anions. The number of
by framework flexion are the perovskites. anions surrounding any given cation is de-
This material, with the general formula termined by the ratio of the “sizes” of the
ABX3, can be viewed as a stuffed deriva- cations and anions (for the problem of de-
tive of WO3-type structures. WO3 (BX3) is fining the “size” of an atom see e.g., Ross
built up of a three-dimensional network of and Price (1997)). In such a close-packed
corner-linked WO6 octahedra, the W (B) array of anions, rigid polyhedral rotation is
cations occupying the corners of the cube- usually not able to accommodate a volume
shaped unit cell. In perovskite the center of reduction imposed by increasing pressure.
this cube is occupied by the A cations. In The only way to optimize the volume upon
the ideal structure, the B–X–B angle at the increasing pressure is through more effi-
connecting octahedral corners is 180°. As cient packing of the anions involved, thus
shown by Glazer (1972), this octahedral reducing the anion–anion distances within
framework is susceptible to a number of the structure. This reduces the ‘size’ of the
distortions which are characterized by rigid anions without affecting the respective size
octahedral rotations, thus pure framework of the cations, which leads to a higher cat-
flexions. These distortions can even be ob- ion/anion size ratio. As a consequence, the
served at ambient conditions depending on coordination number of the cations in-
the nature of the A and B cations. At ambi- creases. Such a coordination increase is
ent conditions, the geometry is controlled generally accompanied by a lengthening of
by the relative size and thus the bonding re- the cation–anion distances, which at first
quirements of the cations involved. In a glance may appear a surprising effect for a
similar way the effect of compression de- high-pressure phase transition. However,
pends on the relative size at ambient condi- the lengthening of the first coordination
tions and the relative compressibilities of sphere is compensated by a shortening of
the A and B cations. In most cases the A the second shell. It is therefore worthwhile
cation is the more compressible unit, thus looking not only at the nearest neighbors of
leading to increasing polyhedral tilting a cation. When looking at both the first and
664 10 High Pressure Phase Transformations
second coordination shells of a cation, a ica (SiO2) from coesite (4-coordinated Si,
coordination increase can be visualized as 2-coordinated O) to stishovite (6-coordi-
making the bonding environment of a cat- nated Si, 3-coordinated O). While the co-
ion less distorted. Keeping this in mind, the esite structure is described by a very dense
increase in coordination number can be packing of corner-linked SiO4 tetrahedra,
quantitatively rationalized by the distortion stishovite adopts the rutile structure, char-
theorem (Brown, 1992). The compression acterized by chains of edge-sharing SiO6
of a close-packed array of anions leads first octahedra. Other prominent examples of
to a reduction in the anion–anion distance increasing coordination numbers at in-
without rearranging the actual packing. creasing pressure are the well-known phase
This reduces all cation–anion distances, transitions from the NaCl (6-fold) type
which leads to a strengthening of the indi- structure to the CsCl (8-fold) structure in
vidual bonds. In the framework of Brown’s alkali chlorides or the high-pressure transi-
bond valence approach, a general strength- tions from zincblende (4-fold) to NaCl or
ening of cation–anion bonds around a b-Sn (6-fold) in many AX semiconductors
given atom induces an ‘overbonding’ of (see Sec. 10.5.1).
the atoms, which in turn destabilizes the
structure. Because of the exponential rela-
10.2.3 Pressure-induced ordering
tionship between bond strength and bond
length (Fig. 10-2), a set of equal bond The potential effect of high pressure on
lengths will have a lower bond valence sum order–disorder phase transitions has only
than the same number of bonds with the recently been fully realized. The interested
same average value, but different individ- reader is referred to the excellent and
ual bond lengths (distorted arrangement). thorough review by Hazen and Navrotsky
An increase in coordination number thus (1996). In this section we only give a brief
reduces the overbonding of the atoms in summary of the main features of pressure-
the crystal by making the bonding environ- induced order–disorder phenomena.
ment around the atoms less distorted. A Atoms on a given crystallographic site
very instructive example of this effect is can order with respect to their chemical
the aforementioned phase transition in sil- species, exact position, magnetic moment
10.2 Pressure-Driven Phase Transitions 665
or electronic state. In the following, disor- vors an ordered arrangement (Hazen and
der with respect to chemical species is re- Navrotsky, 1996). This can be understood
ferred to as ‘substitutional disorder’, by the fact that an ordered arrangement of
whereas disorder on two neighboring sites two atomic species of different ‘sizes’
is called ‘structural disorder’. Changes in leads to an alternation of ‘small’ and
any of these order parameters across a ‘large’ layers or rods (Fig. 10-3 a). A disor-
phase transition are known to have an ef- dered arrangement, on the other hand,
fect on the molar volume of the material forces each of the disordered sites to have
(e.g., Owen and Liu, 1947). Depending on the apparent size of the largest atom shar-
whether the molar volume is smaller for an ing this site (Fig. 10-3 b). An ordering de-
ordered or disordered state, pressure can pendence of the molar volume (and there-
thus – in principle – promote ordering or fore a pressure dependence of the ordering)
disordering effects. Nevertheless, such ef- can also be observed in flexible framework
fects are very often hampered by reduced structures such as feldspars. At first glance,
diffusion rates at high pressure. It is there- the molar volume of such structures should
fore fair to say that at moderate tempera- not be critically correlated with the size of
tures (Ⰶ1500 K), high pressure does not the cations, because in these structures
necessarily induce order–disorder transi- most of the volume change induced by or-
tions, but certainly supports such reactions. dering or pressure can be accommodated
The fast kinetics above ~ 1500 K, however, by the intra-polyhedral angles (Sec.
allow rapid equilibration with respect to or- 10.2.1). The observed volume changes,
dering and even in many cases inhibit however, can be understood on the basis of
quenching of an ordering pattern stable at the variation of intra-polyhedral angles, de-
high pressure and high temperature (Hazen pending on the cation species occupying
and Navrotsky, 1996). the respective polyhedra (e.g., Geisinger et
Based on present data, the volume al., 1985). Simple geometric considera-
change DVdis = Vdisordered – Vordered of struc- tions show that varying distributions of a
tural or substitutional disorder tends to be given set of angles in space leads to differ-
positive, suggesting that high pressure fa- ent enclosed volumes (Fig. 10-4).
666 10 High Pressure Phase Transformations
The largest volume changes are ob- the charge-ordering transition to lower
served for substitutional ordering pro- pressures and eventually even induces an
cesses. A4+B22+O4 spinels, for example, iso-structural phase transition where the
show a difference in molar volume of up to mixed valence compound Sm32+Sm3+Bi33–
3.5% between the normal (fully ordered) changes to a purely 3-valent material
and the inverse (disordered on the octahe- Sm43+Bi34– (Ochiai et al., 1985).
dral site) modifications (Hazen and Yang, Another technically interesting phenom-
1999). The pressure behavior of the spinel enon that is connected to the pressure de-
family is especially interesting, because pendence of ordering and correlation of
they show both pressure-induced disorder- charge carriers is the well-documented
ing as well as pressure-induced ordering pressure dependence of Tc in certain ce-
for one and the same structure type de- ramic superconductors (e.g., Acha et al.,
pending on the chemical species involved 1997; Han et al., 1997). Although neither
(Wittlinger et al., 1998; Hazen and Yang, the superconductivity nor its relationship
1999). to high pressure is fully understood in ox-
The rather scarce data on pressure de- ide materials, it is justified to assume that
pendence of charge distribution seems to the observed strong shift in Tc with increas-
indicate that charge ordering tends to be ing pressure is connected to a subtle inter-
suppressed by high pressure. NaV2O5 for play between pressure-induced structural
example shows a charge-ordering transi- distortions, orbital overlaps and charge-
tion around 35 K at ambient pressure. This carrier distribution.
transition seems to be shifted to lower tem-
peratures at higher pressures and disap-
pears completely around 1 GPa (Ohwada et 10.3 Generating High Pressure
al., 1999). This suggests that in the case of
NaV2O5 , high pressure induces a charge- The technology of pressure cells com-
disorder phase transition. In a similar way, patible with in situ experiments for mate-
the application of pressure to Sm4Bi3 shifts rial characterization has experienced tre-
10.3 Generating High Pressure 667
mendous development since the ground- energy rises from E0 to EH. The Rankine–
breaking work of Bridgman. Consequently, Hugoniot relations combine these quanti-
there is a very comprehensive and vast lit- ties to (r0 = density at ambient conditions):
erature on this subject to which the more
VH = V0 (US – UP)/US (10-8)
committed reader is referred (Miletich et
al., 2000; Holzapfel, 1997; Eremets, 1996; P H = r0 U S U P (10-9)
Ahrens, 1987). Only a brief overview will
EH – E0 = (V0 – VH) PH/2 (10-10)
be given in this section.
There is a huge variety of different tech- The measurements of US and UP give the
niques for generating pressures. They can quantities on the left of Eqs. (10-8) to (10-
initially be divided into static methods and 10) for one experiment. Various experi-
dynamic techniques. ments at different strengths of explosion or
different velocities of the projectile pro-
duce different points, forming the Hugon-
10.3.1 Dynamic pressure generation
iot curve describing the Hugoniot equa-
The highest pressures (102 – 103 GPa) tions of state VH (PH). In order to reduce the
can be obtained using dynamic shock-wave Hugoniot equations of state into an isother-
generation. This is achieved by means of mal equation of state, careful thermody-
explosives or by a projectile that is acceler- namic corrections have to be applied
ated toward the target with a gas gun (Ah- (Poirier, 1991). A material passed by a
rens, 1980, 1987). The shock-wave tech- shock wave usually displays various
nique was originally developed in the mid- stages. Up to a pressure of the Hugoniot
1950s at Los Alamos, USA, in the course elastic limit (HEL) (0.2 to 20 GPa), the
of the development of atomic bombs sample behaves elastically, corresponding
(Walsh and Christian, 1955). In its simplest to the propagation of the longitudinal
case, the impact of a projectile on the target shock wave. Above the HEL, plastic defor-
produces a uniaxial shock wave. The shock mation of the material occurs, giving rise
wave passes through the sample at shock to the fluid-like behavior that creates the
velocity Us . The sample itself is acceler- Hugoniot curves. If a material undergoes a
ated to the sample velocity Up, and Us and phase transition, this is readily observed as
Up , together with the temperature, are the changes of slope in the Hugoniot curves,
quantities measured during a shock-wave separated by a mixed-phase regime (Fig.
experiment. The velocities are usually de- 10-5).
termined by measuring entrance and exit In this sense, in a shock-wave experi-
times of the shock wave. For samples of ment the material can only be investigated
a few millimeters in length, the time to by the difference in UP and US, which both
be measured is in the range of 10–1 to depend on the volume reduction of the
~ < 101 µs. Ignoring the yield strength of sample. The accuracy of these measure-
the solid (which is justified at the high ments becomes critical once the volume re-
shock pressures encountered during a ductions are small at very high pressures.
shock-wave experiment), the material be- Third-generation synchrotron radiation
haves as a fluid. In such an experiment, the sources or even free-electron lasers of the
volume decreases from V0 to VH, the tem- future may offer improved possibilities.
perature increases from T0 to TH and the The increase in X-ray flux of these X-ray
pressure from P0 to PH, while the internal sources may allow for stroboscopic X-ray
668 10 High Pressure Phase Transformations
Figure 10-6. Schematic drawing of (a) a girdle-type press and (b) a belt apparatus. The shaded area corre-
sponds to gasket material. S shows the sample position. Note how consequent cutting of stress amplifying cor-
ners on the girdle and decrease of length-to-width ratio on the piston lead from a girdle press to a belt design.
ders are closed on one side and the sample most straightforward way to reinforce a
is compressed with one piston from the cylinder is by increasing the ratio of its di-
other side. Open-end cylinders rely on two ameter to length. If this is pursued conse-
opposite pistons. The maximal pressure of quently, we arrive at two conical pistons
5 GPa attained by piston-cylinder devices compressing a sample contained in a girdle
is generally limited by the yield strength of (Fig. 10-6 a). A modification of the girdle
the cylinder. Various techniques exist to in- design is obtained by optimizing the coni-
crease this yield strength. In a technique cal shape of the anvils. A cycloid shape of
called frettaging the cylinder is prepared the pistons, as shown in Fig. 10-6 b, en-
such that in its default state, the outer part sures the optimal compromise ensuring
retains a residual tensile stress and the in- sufficient cylinder support at low pressure
ner part a compressive stress. Adding the while maintaining a reasonable stroke in
work stress leads to a partial cancellation the high-pressure regime. This ‘belt
of these stresses, thus allowing for the ap- design’ is by far the most widely used high-
plication of higher loads to the sample. pressure apparatus for materials synthesis
Pre-loading the cylinder is another way to and is frequently applied in industry for the
increase the yield strength of the cylinder synthesis of super-hard material. The max-
to higher values. Pre-load can be achieved imum pressure achieved by belt devices is
by using a soft but incompressible material around 10 GPa. Such devices have also
(lead) as the inner cylinder material and been optimized with respect to sample vol-
by initially overstraining it into its flow re- ume; a flat belt apparatus constructed by
gime. Another method relies on the appli- Fukunaga et al. (1987) was capable of
cation of an external load on the cylinder compressing a sample of 125 ml up to pres-
simultaneously with the application of the sures of 5 – 7 GPa.
load on the piston. Alternatively, cylinders The disadvantage of piston-cylinder as-
can be reinforced by winding a strong wire semblies is the intrinsic opacity of both pis-
under tension around the cylinder. The ton and cylinder, thus inhibiting most in
670 10 High Pressure Phase Transformations
situ observations. This problem is allevi- is also translated into an equatorial com-
ated by opposed anvil systems which oper- pression through wedges attached to the
ate with a gasketing system instead of a four equatorial anvils. Although initially
cylinder. The very first example of this designed for synthesis experiments, vari-
type was built by Bridgman (1952). A log- ous models of this type have been used
ical extension of the Bridgman design are in recent years at synchrotron radiation
profiled Bridgman anvils as first proposed sources for in situ X-ray diffraction studies
by Ivanov and coworkers in the 1960s in Japan (Photon factory and Spring-8),
(e.g., Ivanov et al., 1991) leading to tor- Brookhaven (NSLS) and Hamburg
oidal anvils (Fig. 10-7). This principle was (DESY). This is possible because the small
optimized most consequently in the gaps between the individual anvils allow
Paris–Edinburgh cell design (Besson et al., for entry and exit of X-rays, and the cell as-
1992). The advantage of X-ray and neutron sembly containing the sample is made of X-
transparent gasketing has been exploited to ray transparent material (BN, epoxy). With
the maximum in this cell. This is the reason tapered anvils made of sintered diamond,
why this cell design has advanced to the maximal pressures of 20 – 25 GPa at tem-
state-of-the-art model for in situ high-pres- peratures around 1000 °C can be attained.
sure neutron diffraction and is also em- A rather different geometry in multi-an-
ployed for X-ray experiments. A develop- vil technology is the split-sphere design
ment of the Los Alamos neutron group (Liebermann and Wang, 1992) pioneered at
pushes the limits of P and T attainable with Okayama University by Ito. It consists of a
toroidal anvils up to 50 GPa and 3000 K, uniaxial ram applying force on a split
thus enabling in situ neutron studies up to sphere (first stage) that contains a cube-
these conditions. shaped cavity with the body diagonal along
A quite different approach to compress- the axis of the ram. The cavity is occupied
ing large volumes up to pressures of by a cube built of eight WC cubes with
30 GPa was first developed by researchers truncated corners. The truncated corners in
in Japan using multi-anvil designs (Aki- turn form an octahedral cavity that hosts
moto, 1987). A very successful product is the cell assembly. The cell assembly is usu-
the DIA-type cubic anvil press. In this de- ally a MgO octahedron with a Pt capsule
vice, six tungsten carbide (WC) or sintered containing the sample in its center. A vari-
diamond anvils are arranged parallel to the ation of this design is the split-cylinder de-
faces of a cube that encloses the cell as- sign in which the sphere of the first stage is
sembly. A ram exerts an axial force, which replaced by a cylinder. Multi-anvil presses
10.3 Generating High Pressure 671
based on the split-sphere approach are also 10.3.2.2 Diamond anvil cells
used as in situ devices at the ‘Synchrotron
Radiation Source’ in Daresbury (UK) and Experiments at extreme pressures, which
the ‘Advanced Photon Source’ at Argonne are of interest not only to geophysicists and
National Laboratory (USA). planetologists, but also to physicists and
However, the determination of pressure chemists studying, for example, solidifica-
is not straightforward. In principle, pres- tion and metallization of ‘gases’, can be
sure can be calculated by dividing the force achieved using diamond anvil cells (DAC).
by the area on which it is acting. In practice Diamond anvil cells are in principle very
this is not applicable, mainly because of small opposed-anvil devices of Bridgman
the unpredictable friction losses and illde- type. The anvils are made of diamond sin-
fined compressibilities of the cell assem- gle crystals, shaped in the brilliant cut with
bly. Therefore, pressure has to be deter- the bottom tip truncated to form the anvil
mined through a calibration procedure. surface (culet). This simple design allows
When in situ diffraction techniques are im- the special properties of diamonds to be
possible, this can be done by determining used in two ways. First, the extreme hard-
phase transitions revealed through changes ness of diamonds allows very high pressure
in resistivity of metals and semiconductors to be generated. Pressures attained depend
(e.g., Bi, Ba, ZnS, GaAs, GaP). The re- of course on the size of the culet. Maximal
spective high-pressure phase transitions pressures of 500 GPa have been reported
are determined in hydraulic pressure de- (Xu et al., 1986) and pressures between
vices whose pressure can be directly meas- 100 and 200 GPa can be reliably repro-
ured by a pressure gauge. If in situ observa- duced. The second advantage of using
tion is possible, the equations of state of single crystalline diamonds as pressure
materials such as NaCl, Cu, Mo, Ag and anvils is the high transparency of diamond
Pd, which have been determined up to very for almost the entire electromagnetic spec-
high pressures through shock-wave experi- trum. This allows us not only to easily
ments, can be applied to calibrate the press. observe samples under high pressure, but
Because the reproducibility for a given also to probe them with spectroscopic
press and cell assembly is very high, pres- methods as well as X-ray diffraction
sure of subsequent experiments can be de- (Sec. 10.4). The strength of DACs is thus
termined from the force applied on the cell their huge range of pressure combined with
assembly. the ease of performing in situ experiments,
The strength of large-volume devices and this all with a device of the size of a
lies in their potential for material synthesis fist, which is also easy to operate by non-
at simultaneously high temperature and specialists.
high pressure. The relatively large size of The very first DAC was constructed us-
the compressed volume (~ 0.1 – 1 cm3) al- ing a big (8 ct) gem-quality diamond (taken
lows inclusion of a heater (usually cylindri- from smugglers by the US government and
cal graphite or LaCrO3 resistance heaters) donated to the US National Bureau of Stan-
as well as thermocouples, which enable dards) in which a hole was drilled (Jamie-
pressure and temperature to be combined son, 1957; Lawson and Tang, 1950). Pres-
in a very controlled way. Due to the large sure on the sample in the hole was applied
volume, these devices are limited to maxi- via a piston (piano string) pressed on the
mal pressures of 50 GPa at the very best. sample. The limited pressure range obtain-
672 10 High Pressure Phase Transformations
able with this approach was painfully real- research by Bassett (Merrill and Bassett,
ized when the 8 ct diamond was crushed 1974) and Mao (Mao and Bell, 1975). Fur-
during an experiment. As a consequence, ther details on the technology of DACs and
the opposed-anvil geometry was developed also on its various modifications and devel-
(Weir et al., 1959; Jamieson et al., 1959), opments are given in excellent reviews by
first by simply squeezing a powdered sam- Hazen and Finger (1982) and Miletich et
ple between the culets of two opposed an- al. (2000).
vils and later by introducing the gasketing As for the large volume devices, the ac-
technique (Van Valkenburg, 1964). Its prin- curate determination of pressure in a dia-
ciple has remained unchanged since its in- mond anvil cell is a difficult issue. The
vention and is as simple as it is efficient most accurate values are obtained by add-
(Fig. 10-8). A hole about 100 to 200 mm in ing an internal standard to the sample
diameter is drilled in a metal foil (Fe, W, whose equation of state is known with suf-
Re). This hole serves as sample chamber ficient precision to relate its diffraction
and is filled with the sample (powder or pattern to a pressure value (e.g., Angel et
single crystal) and a pressure medium (al- al., 1997). A very convenient and popular
cohol, liquid gas). The pressure medium is alternative, albeit not quite as accurate, is
compressed through an axial force exerted the exploitation of the pressure shift of flu-
by the diamonds on the gasket. The gasket orescence lines. The most frequently used
seals the sample chamber and at the same fluorescence is the R1 line of ruby (Mao et
time transforms part of the axial pressure al., 1978). This method again benefits from
into an equatorial pressure through the the transparency of the diamond high-pres-
plastic deformation of the gasket material. sure windows. A ruby, which is packed to-
In this way the pressure medium is com- gether with the sample into the gasket hole,
pressed isotropically and therefore trans- is illuminated with a green or blue laser
mits a hydrostatic pressure on the sample. that induces a red fluorescence line. The
This design was popularized for materials wavelength of this line depends on pres-
10.4 Probing Phase Transformations in Materials at High Pressure 673
sure (dl/dP = 0.37 nm GPa–1) and can thus is by measuring the black-body radiation
be used to determine the pressure within of the glowing sample and fitting it to
the sample chamber. More recent develop- Planck‘s spectral function (e.g., Bassett
ments of this approach use different fluo- and Weathers, 1987). When doing HP-HT
rescence lines with various pressure and experiments with laser heating we should
temperature dependence in order to simul- bear in mind the possibility of thermal
taneously determine pressure and tempera- pressure, i.e., the increase of pressure in
ture in the sample chamber (e.g., Datchi et the heated area, while the pressure deter-
al., 1997). mined by a ruby chip outside the hot-spot
Because of the very small volumes of the remains constant (Andrault et al., 1996).
sample chamber (~ 0.001 mm3), combining
pressure and temperature in a diamond an-
vil cell is a difficult task. In principle there 10.4 Probing Phase
are two different approaches, namely exter- Transformations in Materials
nal and internal heating. With external at High Pressure
heating, the whole sample chamber, includ-
ing diamonds, is enclosed in a resistance A very important aspect when doing ex-
heater (e.g., Hazen and Finger, 1982; Bas- periments on high-pressure phase transfor-
sett et al., 1993) that heats the entire assem- mations is the possibility of investigating
bly consisting of gasket, diamond anvils the phase transformation in situ at condi-
and sample. Temperature is measured via tions of high pressure and possibly simulta-
thermocouples attached to the outer dia- neously high temperature. As mentioned
mond facets, assuming that the high ther- above, this is much easier to do in static ex-
mal conductivity of the diamonds allows periments than with dynamic shock-wave
for only a very small temperature gradient techniques. This is mainly due to the very
between sample and outer diamond facets. short time that is available in a shock-wave
This technique has consequently been opti- experiment. We will therefore focus on
mized by using the gasket material itself as static experiments in the following. Among
a resistance heater, thus minimizing the the static experiments, the transparency of
heated volume and thermal gradients (Du- the diamond pressure windows in a DAC
brovinsky et al., 1997). An alternative ap- allows for much more versatile experimen-
proach is the use of an infrared laser beam, tal techniques compared with large-volume
which is focused through the (IR-laser presses. For large-volume experiments, di-
transparent) diamonds onto the sample, rect observation of the sample is limited to
where it is absorbed and thus transformed either transport properties (e.g., electric re-
into heat. This technique was again pio- sistance) or, if using some sort of electro-
neered by Bassett (e.g., Bassett and Ming, magnetic radiation as a probe, severe com-
1972) and then further developed and opti- promises in signal-to-background and ac-
mized by Boehler (Boehler and Chopelas, cessible space have to be accepted. Never-
1991) and Fiquet and Andrault (Fiquet et theless, the range of in situ techniques ap-
al., 1994). Measuring temperature with this plicable to both DACs and large-volume
technique is even more difficult because it devices has grown considerably and con-
explicitly assumes that only the sample is tinues to expand. There is an extensive spe-
heated up. The only way to obtain a quanti- cialized literature on each of the various
tative estimate of the sample temperature techniques. We will give here a brief over-
674 10 High Pressure Phase Transformations
rates, the actual decay event can be sepa- Brown et al. (1988) for a review). If an in-
rated from the primary synchrotron radia- cident X-ray photon hitting an atom in the
tion. This allows the individual decay sample has an energy equal to the energy
events and their interaction with the nu- difference between the ground state and the
clear and magnetic lattice to be investi- excited state of a core electron, it will be
gated directly. In such a ‘nuclear resonant absorbed by the atom while lifting the core
scattering’ experiment, the emitted g-rays electron to an excited state. Photons at this
of the individual nuclei in the material have energy have a high probability of being ab-
different wavelengths owing to the hyper- sorbed by the sample, which leads to char-
fine interaction of the magnetic 57Fe nu- acteristic absorption edges in the transmit-
cleus. The different wavelengths of radia- ted X-ray spectrum. Core electrons in their
tion emitted from different nuclei cause a excited state (photoelectrons) are delocal-
quantum-beat oscillation pattern, which in ized from their parent atom and can thus
turn yields information on the material in- interact with the intermediate surroundings
vestigated (e.g., Smirnov, 1999). The high of the atom as well as with other photoelec-
brilliance of synchrotron radiation sources, trons, leading to an oscillatory contribu-
which are required for such experiments, tion in the X-ray spectrum in the vicinity of
also makes nuclear forward scattering an absorption edge. It is this oscillatory
much easier to apply in combination with part that is extracted from an EXAFS ex-
diamond anvil cells. High-pressure experi- periment. It is dependent on the immediate
ments using nuclear forward scattering as a surroundings of an atom and therefore con-
probe to investigate structural and mag- tains information about the local neighbor-
netic phase transitions have therefore hood of an atom in a solid or a liquid.
quickly become very popular (e.g., Nasu, When photoelectrons are recaptured by an
1996; Lubbers et al., 1999). atom, they will emit characteristic fluores-
cence radiation which by itself can again
be used as a probe for characterizing a ma-
10.4.2.4 X-Ray absorption spectroscopy
terial. The advantage of XAS methods in
The high X-ray brilliance of modern syn- comparison with diffraction methods is
chrotron radiation facilities not only revo- their sensitivity to local and short-range ef-
lutionized high-pressure diffraction (see fects. This makes them very powerful for
Sec. 10.4.4), but also provided the opportu- the investigation of phase transformations,
nity to apply X-ray absorption spectros- not only in crystalline solids but also in
copy (XAS) methods in high-pressure amorphous solids and liquids (e.g., Buon-
research. In particular, developments on tempo et al., 1998).
large-volume presses (Paris–Edinburgh)
which were originally built for neutron dif-
10.4.2.5 NMR
fraction, proved to be very useful for in situ
XAS experiments. Most popular among the A final spectroscopic tool that can be ap-
high-pressure XAS techniques are the ‘ex- plied to samples under high pressure is
tended X-ray absorption fine structure’ NMR. An NMR experiment exploits the
(EXAFS) experiments (i.e., Katayama et interaction between the magnetic moment
al., 1997). The principle of EXAFS is well and the spin of the nucleus on the one hand
known and has been described since the and a static magnetic field disturbed by
early days of quantum mechanics (see pulses of a radio frequency field on the
10.4 Probing Phase Transformations in Materials at High Pressure 677
other. Introductory texts on NMR are avail- 10.4.3 Ultrasonic sound velocity
able, e.g., by Akitt (1983) or Kirkpatrick
(1988). The spin of a nucleus leads to a Measuring sound velocities at high pres-
precession motion of its magnetic moment sure is of great interest, mainly for seismol-
if placed in a static magnetic field. A mag- ogists who use this to correlate seismic
netic perturbation in the radio frequency models with mineralogical models of the
range of the static field causes a tilting of earth. The importance of this method for
the precession axis. The time the nucleus materials science is not quite so obvious al-
needs to revert its precession axis into its though in principle, discontinuities in the
static position depends on the nature of the change in sound velocity at high pressure
nucleus itself as well as its immediate sur- can be used to detect phase transformations
roundings in the crystalline or amorphous and to characterize them through the elastic
host. NMR, similar to XAS techniques, is tensor. A simple pulse-transmission or
therefore a useful probe for investigating pulse-echo method can be combined with a
short-range phenomena. Alternatively, the piston-cylinder apparatus in order to mea-
analysis of NMR spin echoes also allows sure velocities at pressures up to about
information on viscosity and self-diffusion 1 GPa. A more sophisticated approach is
rates in liquids to be extracted. ultrasonic interferometry, which can be
The application of NMR to high pressure combined with a split-sphere apparatus. In
is an experimental challenge because it has this technique, two phase-coherent pulses,
to deal with both the inherently small vol- separated in time by the approximate return
umes encountered in high-pressure experi- travel time in the sample, are applied to a
ments and the magnetic susceptibility of cell assembly under pressure, which is at-
most materials suitable for construction of tached to a buffer rod. When combining
high-pressure devices. Nevertheless, the this technique with a large-volume press,
first high-pressure NMR experiment was the anvils are simultaneously acting as buf-
performed as early as 1954, in combination fer rods. For two consecutive pulses, the
with a Bridgman-type press (Benedek and echo of the first pulse from the far end of
Purcell, 1954). The press was almost en- the sample overlaps with the echo of the
tirely made from a non-magnetic Be–Cu al- second pulse from the sample–buffer inter-
loy. Combining NMR and DAC is even face. The interference caused by this over-
more difficult because of the extremely lap leads to a modification of the amplitude
small sample volumes. However, Bertani et of the resultant signal. This amplitude in
al. (1992) developed an NMR–DAC, again turn can be modified by changing the car-
made of Be–Cu alloy. This DAC is compat- rier frequency, causing a beat pattern as a
ible with a cylindrical cryostat and is thus function of frequency. The difference on
able to combine high pressure and low tem- the frequency scale between two succes-
perature, allowing for the study of the pres- sive interference extremes yields the sound
sure and temperature dependence of the velocity, if the dimensions of the sample
Knight shift in solids (e.g., Kluthe et al., are known. This method has an advantage
1996). An NMR set-up for the investiga- over more traditional pulse-transmission
tion of gases and liquids at pressures up to techniques, in that it involves only one
20 bar has been developed by Woelk and interface, reducing the possible coupling
Bargon (1992). problems for the acoustic signals. Owing to
the applied high pressure, mechanical con-
678 10 High Pressure Phase Transformations
tact and thus coupling between the materi- acterize even subtle phase transformations
als forming the interface is further en- (Carpenter et al., 1998, Angel, 2000 b). In
hanced. A more thorough introduction to addition, measurements of single-crystal
this technique is given by Rigden et al. diffraction intensities serve to establish an
(1992) and Li et al. (1998). There have accurate structural model for the observed
been recent efforts to combine ultrasonic phase transformations and are therefore of
interferometry with diamond anvil cells. In great importance to the understanding of
this case, however, because of the very phase transformations at high pressure.
small sample volumes, the carrier frequen- X-ray powder diffraction is much more
cies have to be in the GHz range (Shen et difficult to perform at laboratory sources
al., 1998). (e.g., Haines et al., 1998), mainly because
of the inherently small sample volumes.
For this same reason even at synchrotron
10.4.4 Diffraction techniques
sources the dominating diffraction tech-
Probably the most versatile and at the nique for investigating powdered samples
same time powerful tools for characteriz- has been energy-dispersive diffraction (e.g.,
ing phase transformations at high pressure Holzapfel, 1997), making use of the much
are diffraction methods. There is a huge higher flux of a white beam compared
amount of literature on this field. Here we with monochromatic radiation. Energy-dis-
only briefly present the various domains of persive diffraction was initially the only
high-pressure diffraction giving appropriate method of choice for in situ diffraction
references for the more interested reader. studies in combination with large-volume
presses (Yagi et al., 1987, Weidner et al.,
1992, Mezouar et al., 1996). In the energy-
10.4.4.1 X-ray diffraction
dispersive mode, the diffraction signal is
A comprehensive introduction to the ba- recorded on the energy scale for a fixed
sic principles of X-ray diffraction is given Bragg angle using a Ge solid-state detec-
by, e.g., Stout and Jensen (1989). X-ray tor. Modern synchrotron radiation sources,
diffraction at high pressure has initially coupled with new two-dimensional detec-
only been possible in combination with di- tor technology, however, allow monochro-
amond anvil cells. Despite the high trans- matic powder experiments with both dia-
parency of diamonds to X-rays, diffraction mond anvil cells (e.g., Fiquet et al., 2000)
experiments suffer both from the small and large-volume presses (Parise et al.,
sample volume and from shielding and ab- 1998, Mezouar et al., 1999 a). In this tech-
sorption effects from the pressure cell com- nique, the diffraction signal of a monochro-
ponents. This is the reason why high-pres- matic X-ray beam is recorded as a function
sure experiments on sealed-tube sources of the Bragg angle. In contrast to energy-
are mostly restricted to single crystals, be- dispersive data, such angle-dispersive data
cause the diffraction signal of a crystal of a allow for a quantitative interpretation of
size of about 100 mm is sufficiently strong the diffracted intensities in order to estab-
to pass through the diamond anvil (Hazen lish and refine a structural model. Another
and Finger, 1982; Miletich et al., 2000, An- advantage of angle-dispersive diffraction
gel et al., 2000). By measuring accurate over energy-dispersive techniques is the
lattice parameters on single crystals (An- higher resolution achievable with mono-
gel, 2000 a), we are able to detect and char- chromatic radiation.
10.5 Examples 679
Drickamer, 1958; Paul, 1959; Goñi and NaCl (Cmcm, cinnabar, Imm2, Immm) or,
Syassen, 1998). b-tin (Imma, simple hexagonal) structures.
High-pressure driven phase transforma- An interesting aspect also is that a site-or-
tions have also been extensively studied in dered version of the diatomic b-tin struc-
the past 30 to 40 years using diffraction ture, which previously was believed to be
methods. In the following only an over- of great importance in semiconductor crys-
view of phase transformations in zinc- tal-chemistry, probably does not exist for
blende-type semiconductors will be given, any of the group III–V and II–VI semicon-
without claiming completeness even in this ductors. A further peculiarity, which has
limited segment of semiconductors. A most been found in Si, Ge and GaAs, is the oc-
thorough and comprehensive review of this currence of different phases at a given
subject is given by Nelmes and McMahon pressure, depending whether one is on a
(1998). compressional or decompressional branch
First structural diffraction studies on of the P-path. In a similar way, InSb, HgSe
these materials were performed on pow- and HgTe exhibit intermediate phases dur-
dered samples using energy-dispersive X- ing their sluggish phase transitions from 4-
ray diffraction (EDX) (see Sec. 10.4.4). fold to 6-fold coordination. These interme-
The picture emerging from these studies is diate phases (“hidden” phases) cannot so
a sequence of phase transitions where vol- far be isolated as single phases, but always
ume optimization is achieved by a stepwise occur together with the stable low- or high-
increase in coordination number (see Sec. pressure modification. Most of these pecu-
10.2.2). This increase in coordination num- liarities are believed to be linked to site-
ber was believed to happen through the fol- ordering problems (non-existence of dia-
lowing sequence of phase transformations: tomic b-tin) or to problems of achieving a
4-fold (diamond, zincblende, wurtzite) true equilibrium state when compressing at
Æ 6-fold (NaCl, b-tin) Æ 8-fold (simple room temperature („hidden“ phases).
hexagonal) Æ 8-fold with six close second Systematic experiments at simultaneously
nearest neighbors (bcc, CsCl) or 12-fold high temperature and high pressure are
(ccp, hcp). For a description of the struc- therefore the next experimental step to be
ture types relevant for semiconductor crys- taken to obtain a complete understanding
tal chemistry, see Nelmes and McMahon of semiconductor phase transformations.
(1998). The availability of in situ angle- In the following the phase transforma-
dispersive X-ray powder diffraction tions in Si, Ge, GaAs and InSb will be dis-
(ADX) (see Sec. 10.4.4) has changed this cussed in more detail, because they are in-
simple model dramatically, mainly because structive examples of most of the special
of the much higher resolution of the ADX features encountered in the high-pressure
technique relative to EDX. The overall pic- behavior of group IV, III – V, and II – VI,
ture emerging from these new experiments semiconductors.
is that at least the change from 4-fold to 6-
fold coordination is characterized by a de-
10.5.1.1 Si and Ge
crease in symmetry, in contrast to the gen-
eral crystal-chemical trend of increasing The phase transformations of Si and Ge
symmetry with increasing pressure. The are quite similar. The main differences are
new 6-coordinated phases can be structu- found in large discrepancies in the respec-
rally described as distorted derivatives of tive transformation pressures and in an ad-
10.5 Examples 681
mained a controversial topic for many Two more phases of GaAs have been
years. GaAs has the zincblende structure at found. The first occurs upon decompres-
ambient conditions. A high-pressure phase sion of the GaAs-II phase. It leads to a 4-
transformation to a GaAs-II phase was first coordinated cinnabar structure (McMahon
found through electrical measurements and Nelmes 1997). The other can be grown
(Minomura and Drickamer, 1962). In the by heating the GaAs-II phase to about
1980s, it was recognized that the structure 450 K at a pressure of ~ 14 GPa (McMahon
would most probably be of orthorhombic et al., 1998). Its structure (SC16) is the di-
symmetry. However, space group as well as atomic equivalent of the BC8 structure
structural topology remained ambiguous found in Ge and Si.
(Shimomura et al., 1980; Baublitz and Ru-
off, 1982; Weir et al., 1989) in that a pro-
10.5.1.3 InSb
posed model (space group Pm2m) implied
very close Ga–Ga and As–As contacts. InSb, which at ambient conditions also
ADX experiments again proved to be the adopts the zincblende structure, is probably
key to resolving the puzzle, although the the most impressive example of a phase di-
very similar X-ray scattering power of Ga agram that has been completely rewritten
and As created additional problems. The in the past 5 – 10 years. The problem of
high resolution of the ADX patterns were, very similar scattering powers is even more
however, able to reveal small reflections, severe in InSb than in GaAs. In order to un-
which allowed for an experimental distinc- ravel the crystal chemistry of its high-pres-
tion between Pm2m and Cmcm (which can sure phases, the ADX technique had to be
be described as a distorted NaCl structure). supported by anomalous scattering in order
Ga and As form slightly distorted NaCl- to distinguish between true symmetries and
type planes parallel (001). Neighboring pseudo-symmetries caused by the similar
planes are shifted parallel to [010] by about scattering powers (Nelmes and McMahon,
half a Ga–As distance (Nelmes and McMa- 1998). The peculiar feature of InSb is that
hon, 1998). Upon further compression, two different successions of high-pressure
GaAs has been reported to undergo another phases as a function pressure can be ob-
phase transition to a simple hexagonal served, albeit in a reproducible way. The
structure (Weir et al., 1989). However, a di- basic succession is from the zincblende
atomic compound such as GaAs, cannot type to a superstructure of Cmcm (InSb-
adopt the simple hexagonal structure, al- IV) around 3 GPa. At 9 GPa the structure
though the topology of the atoms does have transforms to an orthorhombic distorted
this symmetry. To adopt this structure, NaCl phase (InSb-III) with space group
GaAs would have to be completely site- Immm. The Immm phase appears to be at-
disordered. It can be speculated that the tained through an intermediate site-disor-
similar scattering power of Ga and As may dered orthorhombic phase with space
make a diffraction pattern (because of lim- group Imma. The transformation from
ited resolution and signal-to-noise ratio) Imma occurs with time (a few hours) or
appear to have simple hexagonal symme- upon slight heating. InSb-III, in turn, trans-
try, when in truth it remains of orthorhom- forms at 17 GPa to a new phase with yet
bic symmetry. This would imply that GaAs unknown structure, and upon further com-
remains in Cmcm structure up to the high- pression over 21 GPa, to a site-disordered
est pressure investigated, i.e., 108 GPa. body centered cubic structure. A new fea-
10.5 Examples 683
ture in this case is that the transition from 10.5.2 Materials in the B–C–N system
the zincblende-type phase to InSb-IV can
also be preceded by an additional phase Diamond (i.e., the high-pressure poly-
transition around 2.1 GPa, which produces morph of carbon) is not only a precious and
a mixture of a site-disordered b-tin phase popular gem stone, it also has great techno-
and the InSb-II phase with orthorhombic logical significance owing to its extraordi-
symmetry (Immm). This mixture recrystal- nary physical properties. The most promi-
lizes to InSb-IV when left for a few hours. nent of these is a Mohs hardness of 10 (di-
Even more peculiar is the fact that if, dur- amond is the hardest material known) mak-
ing the presence of the mixed phase the ing it a most efficient abrasive material.
pressure is increased to about 3 GPa within The reason for the extreme hardness of di-
a few hours, the formation of InSb-IV is amond lies in the combination of high co-
suppressed (Nelmes et al., 1993a). The valency together with the smallness of the
InSb-II phase persists up to 9 GPa where it core electron shell of carbon, distinguish-
does not show any phase transition to InSb- ing it from the isomorphic compounds Si
III, because InSb-III already has the Immm and Ge (see Sec. 10.5.1.1). Also unusual is
structure. the combination of a very high thermal
As outlined in this brief summary of the conductivity coupled with a low electrical
unusual high-pressure behavior of GaAs conductivity, which makes it an interesting
and InSb, there seems to be a number of material as a heat sink in microelectronic
unusual phenomena in the structural be- applications. Graphite (i.e., the low-pres-
havior of zincblende-type semiconductors sure polymorph of C) also has technologi-
at high pressure. These phenomena may cal importance thanks to its electrical (e.g.,
well be linked to kinetic problems as they graphite electrodes) as well as its mechani-
are often encountered during phase trans- cal (steel additive, graphite composite
formations at high pressure. As shown by material, nanotubes) and lubricating prop-
Leinenweber (1993) and Kunz et al. erties. Therefore, (pseudo-)isoelectronic
(1996), such kinetic problems can be alle- systems of C have attracted much attention
viated by adding high temperature to high in the hope of obtaining materials with
pressure. Very often heating a compressed graphite- and diamond-like properties.
sample may eventually lead to a stable There is reason to believe that ternary com-
phase. In a similar way, the nonexistence of pounds in the B–C–N system with graph-
a site-ordered b-tin phase could be linked ite-like structures might be semiconductors
to microstrain and/or kinetics. Mezouar et with high thermal stability or, alternatively,
al. (1999 b) showed that different phases diamond-type phases in the B–C–N field
can be obtained at a given pressure, de- may exhibit abrasive properties exceeding
pending on the microstrain and hydrosta- those of diamond combined again with a
ticity present in a pressure cell. It will thus much higher thermal stability (Kurdyumov
be a challenge in future high-pressure semi- and Solozhenko, 1999).
conductor research to combine high pres- Crystallographic investigations of super-
sure and high temperature while simultane- hard materials in situ at high pressure pose
ously probing the sample. an additional challenge: the extreme hard-
ness of the material causes strong devi-
atoric stresses when compressing a pow-
dered sample (Weidner et al., 1994 a, b).
684 10 High Pressure Phase Transformations
The deviatoric strain caused by the devi- a pressure of about 8 GPa (5 GPa if a C-so-
atoric stress leads to anisotropic peak lution in a metal melt is used as precursor)
broadening, which in turn makes the accu- usually has to be combined with a tempera-
rate interpretation of powder patterns more ture of ~ 1000 K in order to obtain cubic di-
difficult. amond. The crystal structure of cubic dia-
mond is the monatomic equivalent of the
zincblende structure [see Sec. 10.5.1 and
10.5.2.1 C
Nelmes and McMahon (1998) for detailed
Carbon crystallizes in three different descriptions of these structure types]. Al-
polymorphs (Fig. 10-9 a). Graphite, the though the phase boundary has a positive
stable phase at ambient conditions, is char- slope and thus would extend to much lower
acterized by a stacking of hexagonal pressures at room temperature, graphite
layers, where the individual layers are compressed at room temperature does not
formed by a honeycomb net of 6-mem- readily adopt the diamond structure. In-
bered rings formed by C. The layered stead, pure compression of graphite at
structure is responsible for the very low room temperature up to 12 GPa and subse-
hardness (Mohs 1 – 2) as well as the aniso- quent annealing at about 1000 K leads to
tropic electrical properties of graphite. the formation of a hexagonal form of dia-
Graphite can exist in a hexagonal (2H) and mond (londsdaleite). This transformation
rhombohedral (3R) modification, where is reversible in the temperature range
these polytypes differ in the stacking of the between 1000 and 1300 K. If heated above
layers: ABABAB for 2H and ABCAB- 1400 K, however, londsdaleite converts
CABC for 3R. If graphite is pressurized, it to cubic diamond. Subsequent further
transforms to diamond. For kinetic reasons, changes in the P – T field do not induce any
Figure 10-9. P – T phase diagrams for (a) carbon and (b) BN. Note that the two topologies are identical. (b) is
different from (a) in that it is shifted to lower pressures. This causes the phase boundary between h-BN (graph-
ite-type) and c-BN (diamond-equivalent) to cross the zero pressure line at ~ 1600 K, thus making c-BN the
stable phase at ambient conditions. After Bundy (1989) and Solozhenko (1994, 1999).
10.5 Examples 685
further phase transformations as long as we type cubic phase (c-BN) crosses the 1 bar
remain above the graphite–diamond phase line at a temperature of about 1600 K. This
boundary. This implies that londsdaleite, makes the zincblende-type c-BN the stable
which crystallizes in the wurtzite structure, phase at ambient conditions (Solozhenko,
is metastable with respect to diamond. The 1994; Solozhenko et al., 1999). However,
stability of diamond under high pressure the growth of c-BN at atmospheric pres-
seems to be much higher than that of sili- sure is only possible in the presence of
con or germanium (see Sec. 10.5.1.1). No supercritical fluids and seed crystals. Spon-
phase changes have been observed up to taneous crystallization of c-BN has not
pressures of 550 GPa (Xu et al., 1986). been observed at pressures below 2 GPa
This is attributed to the fact that C, unlike (Solozhenko, 1994). This fact has, of course,
Si and Ge, does not have any core p-elec- important technological consequences, as
trons, which brings the valence electrons in c-BN is the phase with the desired abrasive
C much closer to the nucleus and thus sta- properties. The graphite-like phase (h-BN)
bilizes the cubic high symmetry structure is the stable phase at low pressure and high
(Bundy, 1989). temperatures. The graphite-like phase can
In the B–C–N phase field, there are four occur as hexagonal (2H) or rhombohedral
important ternary phases known, namely (3R) polytype (e.g., Britun et al., 1999).
BCN, BC2N, BC3N and BC4N. In addition Like most binary and ternary phases in the
to these, there are numerous binary phases B–C–N system, h-BN has a turbostratic
such as BN and several B-rich B–N solid structure. This structure can be described
solutions, B4C and BC3 along the B–C line as a graphite structure with good in-plane
and C3N4 as the only binary carbon–ni- ordering but random orientation of the in-
tride. Many of these phases can be synthe- dividual planes around the layer normal.
sized as graphite-like compounds through This leads to the absence of hkl or h0l
sputter deposition or chemical vapor depo- reflections, while hk0 and 00l reflections
sition (CVD), yielding thin films suitable exhibit very irregular peak shapes (e.g.,
for abrasive protection. Because in this Andreev and Lundström, 1994). The phase
chapter we are mainly interested in high- boundary between c-BN and h-BN has a
pressure phase transformations, we will fo- positive slope and meets the liquidus in a
cus on structural transformations relevant triple point at 3480 K and 5.9 GPa (Solo-
to the synthesis of bulk samples. zhenko et al., 1999). Similar to the carbon
phase diagram, the BN also forms a meta-
stable phase (w-BN) at high pressure,
10.5.2.2 B–N
which crystallizes in the wurtzite structure.
The III-V analogue to carbon is BN. Ow- The transitions between the various poly-
ing to its industrial importance as a dia- types are either martensitic or diffusional,
mond substitute in drilling and polishing depending on the synthesis and formation
applications, a large amount of research conditions (shock compression vs catalytic
has been done in this field. The topology of synthesis) as well as on the ordering degree
the phase diagram is very similar to that of of the precursor material (Kurdyumov,
carbon (Fig. 10-9). However, it is shifted 1995). A curious but potentially very inter-
towards lower temperatures so that the esting BN phase has recently been synthe-
phase boundary between the graphite-type sized by laser heating a c-BN sample up to
hexagonal phase (h-BN) and a zincblende- 5000 K at pressures between 5 – 15 GPa
686 10 High Pressure Phase Transformations
(Golberg et al., 1996). This procedure was pressurized to about 7 GPa and tem-
yielded multi-walled nanotubes, which did peratures between 600 and 1200 K. This
not contain any other inclusions as they are procedure yielded faceted crystals of a-
frequently obtained when producing nano- and b-C3N4.
tubes via a plasma-arc discharge method.
10.5.2.4 B–C
10.5.2.3 C–N
The crystal chemistry along the B–C
A lot of effort has been put into the syn- axis has been well studied (Thévenot,
thesis of sp3-bonded carbon nitrides, again 1990). The most important binary is B4C.
in the hope of creating super-hard materi- Its structure is a stuffed derivative of a-
als. This hope has been supported by theo- boron, where in B4C the individual boron
retical calculations predicting the stability icosahedral chains are linked through a
of C3N4 phases whose bulk moduli were three-atom C–B–C chain. So far no high-
calculated to exceed the value of diamond pressure phases of B4C are known. How-
(Sung and Sung, 1996; Liu and Cohen, ever, owing to the electron-deficient nature
1989, 1990). These theoretical studies pre- of the intra-icosahedral bonding, boron-
dict five different phases of C3N4. Besides rich solids such as B4C are expected to be
a graphite-like phase, an a- and a b-phase inverted-molecular solids, where the com-
are expected. These would be structurally pressibility of the molecules is higher than
identical to the corresponding phases of that of the intermolecular links. This is un-
Si3N4. Their expected bulk moduli are 425 usual because in most cases the compress-
GPa and 451 GPa, respectively. A more re- ibility of a molecular crystal is controlled
cent theoretical study (Teter and Hemley, by the stiffness of the weak van der Waals
1996) predicts two more high-pressure bonding between the molecules, while the
phases. A cubic phase should be structu- bonds within the molecules are rather rigid.
rally identical to the high-pressure Zn2SiO4 The inverted compressibility was experi-
polymorph. The bulk modulus for this mentally confirmed for B4C (Nelmes et al.,
phase is calculated to be higher than that of 1995). Such an inverted behavior is re-
diamond (496 GPa vs. 444 GPa). Nguyen markable for a compound like B4C in view
et al. (1998) managed to synthesize a bulk of the very strong and covalent B bonds
C3N4 phase by laser heating a C60 film with within the B12 molecules.
liquid nitrogen in a diamond anvil cell at
2000 K and 18 GPa. The diffraction pattern
10.5.2.5 B–C–N
of the retrieved phase is very close to the
one predicted theoretically for the cubic Ternary phases in the B–C–N system are
structure by Teter and Hemley (1996). The rather difficult to grow, mainly because of
measured compressibility, however, is the high stability of BN–C mixtures rela-
much higher than predicted (249 GPa vs. tive to true B–C–N phases. Nevertheless,
449 GPa). The phase was found to decom- several ternary phases have been synthe-
pose upon pressure release below 14 GPa. sized (see Solozhenko (1997) for a review).
Successful attempts at bulk synthesis of Graphite-type phases with ternary compo-
C3N4 have also been reported by He et al. sitions are mostly grown using CVD meth-
(1998) and Dymont et al. (1999). In both ods or nitration of B- and N-containing
cases a precursor containing C, N and H compounds in the condensed state. Investi-
10.5 Examples 687
gations of possible diamond-type struc- commonly called ice. Presently, there are
tures in this system are difficult owing to 12 crystalline and two amorphous phases
the difficulty of distinguishing true ternary of ice accepted. Crystalline ice phases are
phases from a mechanical mixture of C and labeled with Roman numerals from I to XII
BN with diffraction methods. In order to in the approximate sequence of their dis-
create diamond-type ternary phases (exhib- covery. The high structural variability can
iting zincblende- or wurtzite-related struc- be explained by a subtle interplay between
tures), the diffusion during the high-pres- enthalpic and entropic contributions. The
sure/high-temperature experiment has to be flexible O–H…O hydrogen bond with a
suppressed, otherwise the ternary phases strong O–H bond and at least one weak
decompose into c-BN and diamond. This O…H hydrogen bond around each hydro-
can be understood on the basis of the sig- gen enables the formation of a large num-
nificant positive deviation of Cx(BN)1–x ber of energetically similar topologies. In
solid solutions from Vegard’s law. It im- addition, the total energy of the system can
plies that ternary high-pressure phases will be significantly altered through the entropy
have a higher volume than a linear combi- of a disordered distribution of the O–H…O
nation of the components involved and will asymmetry in the framework.
thus, provided sufficient atomic diffusion The flexibility of the hydrogen bonds is
occurs, be unstable at high pressure. The intimately linked to their asymmetric
most efficient way of preventing diffusion geometry. With simple bond valence and
and thus segregation is shock compres- ionic radius arguments, it is possible to
sion (e.g., Komatsu et al., 1996). Another show that this asymmetry is a direct conse-
possibility of transforming graphite-type quence of the fact that the closest un-
C–B–N material into a diamond-type phase strained O–O distance is longer than the
is by laser heating a sample in a diamond sum of two relaxed O–H bonds. This would
anvil cell (Knittle et al., 1995). The sam- lead to an underbonding of the hydrogen
ples were grown at pressures between 30 atom. As outlined above (see Sec. 10.2.2),
and 50 GPa combined with temperatures such an underbonding can be compensated
between 2000 and 2500 K. This approach for by an asymmetric coordination geome-
appears so far to be the only published ex- try. This model implies straight O–H..H
ample where a ternary phase could actually geometries for strong hydrogen bonds, but
be grown from both ternary graphitic pre- bent (because under-constrained) O–H…O
cursors and mechanical mixtures. It should configurations for weaker hydrogen bonds
be mentioned, however, that the analytical (Brown, 1976, 1995). Therefore the model
data of Knittel et al., (1995) could equally correctly predicts the counter-intuitive fact
well be interpreted as resulting from me- that high-pressure phases of ice will tend to
chanical mixtures of dispersed diamond have weaker hydrogen bonds. This is due
and c-BN (Solozhenko, 1997). to the requirement for a denser packing of
water molecules in high-pressure phases
(see Sec. 10.2.1), leading to bent and thus
10.5.3 H2O
weakened hydrogen bonds.
Water not only occurs as the life-sustain- The hydrogen bond is not only critical
ing liquid phase, but also has an exception- for H2O with respect to its structural as-
ally rich phase diagram (Fig. 10-10) in its pects, but also with respect to its electronic
solid state. All solid phases of water are properties. The disordering of the hydro-
688 10 High Pressure Phase Transformations
Figure 10-10. (a) P – T phase diagram for H2O, (b) enlargement of the central section of (a). Note that in (a)
the pressure scale is logarithmic. (After Petrenko and Whitworth, 1999.)
gen bonds strongly increases the polariz- tions, lack of kinetics favors a chaotic
ability and electrical conductivity of H2O. framework collapse leading to an amor-
There exists a vast literature on the struc- phous phase (Sec. 10.5.3.6).
ture, physics and chemistry of H2O, includ-
ing a very comprehensive and thorough re-
10.5.3.1 Ice Ih, XI and Ic
view by Petrenko and Whitworth (1999).
A common feature of all ice phases is The ice phase occurring at ambient pres-
that the H2O molecules form a three-di- sure (ice Ih) has a hexagonal structure,
mensional tetrahedral framework. This, where the oxygen atoms occupy the posi-
combined with the versatility of the hydro- tions of the monatomic version of the wurt-
gen bonds, makes H2O an interesting study zite structure (see Sec. 10.5.1.1). The hy-
case for understanding the possible mecha- drogen atoms are placed asymmetrically
nisms of a framework to reduce its volume. between neighboring oxygen atoms. Each
At low pressures, H2O adopts rather open oxygen thus forms two O–H bonds and two
frameworks (see Sec. 10.5.3.1). At pres- O…H hydrogen bonds. The distribution of
sures above ca. 0.6 GPa, however, it is able the O–H bonds and O…H hydrogen bonds
to reduce the molar volume significantly is disordered, resulting in space group
by forming two interpenetrating, but un- P63/mmc (hcp space group). The local dis-
connected frameworks (see Sec. 10.5.3.5). tribution of the H atoms, however, has to
In the intermediate range between 0.2 and follow the two ice rules (Bernal and
0.6 GPa, it adjusts to the imposed volume Fowler, 1933), which are valid for all other
constraint mainly by straining the individ- ice phases; (1) there are always two O–H
ual bonds (Sec. 10.5.3.3 and Sec. 10.5.3.4), bonds for each oxygen, (2) there is only
which can lead to a weakening of O…H one H per O–O contact. These rules follow
hydrogen bonds with increasing pressure, directly out of bond-valence rules and cat-
as mentioned above. Under certain condi- ion–cation repulsion considerations.
10.5 Examples 689
resulting in a positive slope of the liquidus within the stability field of ice V. The
in P – T space. new phase can reproducibly be grown at
0.55 GPa and 260 K with slow cooling
rates. Its structure consists of two symmet-
10.5.3.4 Ice V, Ice IV and Ice XII
rically independent sets of oxygen atoms.
Further compression of ice III or ice II The first one forms zig-zag chains parallel
leads to ice V. The structure of ice V is a c. Neighboring chains are rotated by 90°.
complex framework built by 4- and 5- These chains are connected by the sec-
membered oxygen rings (Kamb et al., ond type of oxygen atoms thus forming
1967). Each oxygen atom is tetrahedrally a tetrahedral framework, whose density
surrounded by four neighboring oxygen at- (1.437 g cm–3) is very close to that of ice
oms. The symmetry is monoclinic with IV. While in ice IV, the increase in density
space group A2/a. The hydrogen atoms are relative to ice V is achieved by self-pene-
distributed in a disordered arrangement tration, ice XII compresses by additional
between the oxygen links. In fact, the two- bending. In both metastable phases, the hy-
fold axis of this space group does not allow drogen positions are disordered between
for any proton ordering. Proton ordering, if the O–O links.
it occurs, would thus imply a noncentro-
symmetric space group Aa.
10.5.3.5 Ice VI, Ice VII, Ice VIII, Ice X
A special feature of ice V is that in its
stability field, at least two ice phases with- Compressing ice V beyond 0.6 GPa in-
out their own stability fields have been de- duces a phase transformation to ice VI.
scribed. Ice IV has been observed within This is the first example of an ice phase,
the stability field of ice II, ice V and ice VI. where the structure is characterized by two
Even a liquidus for ice IV, which lies interpenetrating frameworks which are not
within the field of these solid phases has linked together (‘self-clathrates’, Kamb,
been established (Chou and Haselton, (1965)). In ice VI, the topologies of two
1998). The structure of ice IV is quite re- frameworks are identical. They are sym-
markable. Only slightly puckered 6-mem- metrically related by an inversion and
bered rings are linked together within a shifted relative to each other by a/2 + b/2 +
plane by strongly elongated 6-membered c/2. The individual frameworks can be de-
rings forming a puckered plane with scribed as chains of tetrahedra, extended
trigonal symmetry. The planes are stacked along c. All tetrahedra of one framework
in an ABABAB sequence where A- and B- have the same orientation. The chains of
planes are shifted relative to each other by each framework are connected parallel a
the value of the hexagonal a-axis. Like- and b with each other, but unconnected
type planes (A–A and B–B, respectively) with the chains of the other sublattice.
are linked together by hydrogen bonds Ice VII has, together with ice VIII and
through the regular 6-membered rings of ice X, the densest H2O topology. Its pack-
the neighboring planes. This results in a ing is as efficient as it is simple and can be
single, self-penetrating framework with a described by two interpenetrating frame-
higher density (1.436 g cm–3) than ice V works of the cubic ice Ic. The hydrogen po-
(1.402 g cm–3). sitions in ice VII are completely disordered
Only recently, another new phase of ice (Walrafen et al., 1982; Kuhs et al., 1984).
has been described by Lobban et al. (1998) The interesting aspect of this structure is
10.5 Examples 691
that the oxygen atoms form a body cen- In ice X, the O–O contacts are squeezed
tered cubic lattice. This implies that each sufficiently close together that the hydro-
oxygen atom has eight nearest neighbors. gen can sit unstrained in the center of the
However, it forms hydrogen bonds only to O–O-link. Such a symmetric O–H–O ar-
four of them in tetrahedral orientation and rangement thus has no potential for disor-
in this way forms two interpenetrating but der in the first place. Such a phase is differ-
independent frameworks. This also implies ent from a double-well phase with a dy-
that the O–O distances to the hydrogen- namically disordered proton, which from a
bonded oxygen neighbors are equivalent to physical point of view should still be called
the non-bonded O–O contacts. ice VII. Ice X has been predicted by Kamb
The disorder–order transition from ice and Davis (1964). Although experimental
VII to ice VIII was first observed through evidence for a symmetric phase was ob-
dielectric measurements (Whalley et al., served from IR spectroscopic studies by
1966). Although the structure crystallizes the appearance of a new mode characteris-
in the tetragonal space group I41/amd, the tic of a symmetric O–H–O configuration
positions of the oxygen atoms deviate only (Goncharov et al., 1996; Aoki et al., 1996,
little from a body centered cubic lattice, Pruzan et al., 1997), it should be pointed
confirming the order–disorder nature of the out that up to now there is no direct experi-
phase transformation of ice VII to ice VIII mental evidence for a truly proton-ordered
(Kuhs et al., 1984). The ordering pattern of single-well phase of ice X. This lack of ex-
the O–H bonds is ferroelectric within one perimental evidence is due to the difficulty
sub-lattice, but antiferroelectric between the of performing experiments that are able to
two frameworks. In ice VIII, the O–D dis- localize hydrogen or deuterium positions in
tances remain almost constant with increas- the pressure range of 60 GPa and more.
ing pressure (Nelmes et al., 1993 b), al-
though the O–O distances decrease signifi-
10.5.3.6 Amorphous ice at high
cantly. This leads to a decrease in the dis-
pressure
tance between the two potential wells of the
hydrogen position. Therefore, the entropy As mentioned above, there are two
difference between the ordered and disor- amorphous high-pressure phases of H2O in
dered state decreases gradually with in- the solid state. A first phase is formed by
creasing pressure, leading to a progressively compressing ice Ih at 77 K (where slow ki-
steeper phase boundary VII–VIII, which netics inhibit the recrystallization of ice II)
eventually ends as a vertical slope at 62 GPa to the extrapolated liquidus of Ih. Its den-
(Pruzan et al., 1993). At this point, the en- sity in the quenched state is 1.17 g cm–3.
ergy barrier between the two proton posi- Upon heating, this high-density amorphous
tions is sufficiently small to be crossed with (HDA) phase transforms to a second amor-
the thermal energy of the protons, thus lead- phous phase with lower density (0.94 g
ing to a dynamic disorder. This phase thus cm–3) at 125 K (LDA). LDA, in turn can be
has the average position of the hydrogen at- transformed into HDA at 77 K by compres-
oms in the middle between the two oxygen sion to 0.6 GPa.
atoms. An identical structural configuration The research on the high-pressure phases
can be expected for a structure with a single of ice is far from completed. The under-
potential minimum as is predicted for ice X standing of the various amorphous phases
at pressures exceeding 62 GPa. and the theoretical and experimental quest
692 10 High Pressure Phase Transformations
for ice X and beyond are just the most Angel, R. J., Downs, R. T., Finger, L. W. (2000), in:
Comparative Crystal Chemistry, Reviews in Mine-
prominent examples of a lively topic with ralogy, Vol. 40: Hazen, R. M. (Ed.). Washington
relevance to physical chemistry and mate- DC: Mineralogical Society of America, Chapter 5,
rials science. in press.
Aoki, K., Yamawaki, H., Sakashita, M., Fujihisa, H.
(1996), Phys. Rev. B 54, 15673.
10.6 Acknowledgements Arlt, T., Kunz, M., Stolz, J., Armbruster, T., Angel,
R. J. (2000), Contrib. Mineral. Petr. 138, 35.
Bacon, G. E. (1975), Neutron Diffraction, 3rd edn.
The author wishes to thank Vladimir Oxford: Clarendon Press.
Solozhenko, Richard Nelmes, Mohammed Bassett, J. M., Ming, L.-C. (1972), Phys. Earth
Planet. Mater. 6, 154.
Mezouar, Guillaume Fiquet, and Denis An- Bassett, W. A., Weathers, M. S. (1987), in: AGU
drault for valuable discussions. This work Geophysical Monograph, Vol. 39, Mineral Phys-
received support from the ‘Schweizerische ics 2, Manghnanni, M. H., Syono, Y. (Eds.).
Washington DC: American Geophysical Union,
Nationalfonds’ (Swiss National Science p. 129.
Foundation) through grant # 21-052682.97. Bassett, W. A., Shen, M. J., Bucknum, M. J., Ming
Chou, I. (1993), Rev. Sci. Instrum. 64, 2340.
Baublitz, M., Ruoff, A. L. (1982), J. Appl. Phys. 53,
6179.
10.7 References Baughman, R. H., Stafström, S., Cui, C., Dantas, S.
O. (1998), Science 279, 1522.
Abd-Elmeguid, M. M., Micklitz, H., Buschow, K. J. Belitsky, I. A., Fursenko, B. A., Gabuda, S. B., Khol-
H. (1980), J. Phys. IV, Colloque, 41, C1. dev, O. V., Seryotkin, Y. V. (1992), Phys. Chem.
Acha, C., Loureiro, S. M., Chaillout, C., Tholence, J. Minerals 18, 497.
L., Capponi, J. J., Marezio, M., Nunez-Regueiro, Benedek, G. B., Purcell, E. M. (1954), J. Chem.
M. (1997), Physica C 282–287, 1167. Phys. 22, 2003.
Ahrens, T. J. (1980), Science 207, 1035. Bernal, J. D., Fowler, R. H. (1933), J. Chem. Phys. 1,
Ahrens, T. J. (1987), in: Methods of Experimental 515.
Physics, Vol. 24A: Sammis, C. G. Henyey, T. L. Bertani, R., Mali, M., Roos, J., Brinkmann, D.
(Eds.). New York: Academic Press, p. 185. (1992), Rev. Sci. Instrum. 63, 3303.
Akimoto, S. (1987), in: AGU Geophysical Mono- Besson, J. M., Mokhtari, E. H., Gonzales, J., Weill,
graph, Vol. 39, Mineral Physics 2: Manghnanni, G. (1987), Phys. Rev. Lett. 59, 473.
M. H., Syono, Y. (Eds.). Washington DC: Ameri- Besson, J. M., Hamel, G., Grima, T., Nelmes, R. J.,
can Geophysical Union, p. 1. Loveday, J. S., Hull, S., Häusermann, D. (1992),
Akitt, J. W. (1983), NMR and Chemistry. New York: High Pressure Res. 8, 625.
Chapman and Hall. Binder, K. (2001), in: Phase Transformations in Ma-
Anderson, M. S., Swenson C. A. (1984), J. Appl. terials: Kostorz, G. (Ed.). Weinheim: Wiley-VCH,
Phys. 56, 2697. Chap. 4.
Andrault, D., Fiquet, G., Peyronneau, J., Itié, J. P., Boehler, R., Chopelas, A. (1991), Geophys. Res. Lett.
Richet, P., Gillet, Ph. (1996), in: Proceedings of 18, 1147.
the 3rd NIRIM International Symposium on Ad- Bridgman, P. W. (1935), Proc. Am. Acad. Arts Sci.
vanced Materials (ISAM ’96), National Institute 70, 285.
for Research in Inorganic Materials, Tsukuba, Ja- Bridgman, P. W. (1940), Proc. Am. Acad. Arts Sci.
pan, p. 297. 74, 11.
Andreev, Y. G., Lundström, T. (1994), J. Appl. Crys- Bridgman, P. W. (1946), Collected Experimental Pa-
tallogr. 27, 767. pers of P. W. Bridgman, Vol. VI, p. 164.
Angel, R. J. (2000a), in: Comparative Crystal Chem- Bridgman, P. W. (1952), Proc. Am. Acad. Arts Sci.
istry, MSA Reviews in Mineralogy, Vol. 40: Hazen, 81, 167.
R. M. (Ed.). Washington DC: Mineralogical So- Britun, V. F., Kurdyumov, A. V., Petrusha, I. A.
ciety of America, Chapter 8, in press. (1999), Mater. Lett. 41, 83.
Angel, R. J. (2000b), in: Transformation Processes in Brown, G. E. Jr, Calas, G., Waychunas, G. A., Petiau,
Minerals, MSA Reviews in Mineralogy, Vol. 39: J. (1988), in: Spectroscopy Methods, MSA Reviews
Redfern, S., Carpenter, M., Salje E., Seifert F. in Mineralogy, Vol. 18: Hawthorne F. C. (Ed.).
(Eds.). Washington DC: Mineralogical Society of Washington DC: Mineralogical Society of Amer-
America, p. 38. ica, p. 431.
Angel, R. J., Allan, D. R., Miletich, R., Finger, L. W. Brown, I. D. (1976), Acta Crystallogr. A 32, 24.
(1997), J. Appl. Crystallogr. 30, 461. Brown, I. D. (1992), Acta Crystallogr. B 48, 553.
10.7 References 693
Brown, I. D. (1995), Can. J. Phys. 73, 676. Hanfland, M., Schwarz, U., Syassen, K., Takemura,
Bundy, F. P. (1989), Physica A 156, 169. K. (1999), Phys. Rev. Lett. 82, 1197.
Buontempo, U., Filipponi, A., Farcia-Martinez, D., Hazen, R. M. (1983), Science 219, 1065.
Postorino, P., Mezouar, M., Itié, J. P. (1998), Phys. Hazen, R. M, Finger, L. W. (1978), Geological Soci-
Rev. Lett. 80, 1912. ety of America, Annual Meeting 1978, Toronto;
Carpenter, M. A., Salje, E. K. H. (1998), Eur. J. Min- Abstracts, 10, 418.
eral. 10, 693. Hazen, R. M., Finger, L. W. (1981), Rev. Sci. In-
Carpenter, M. A, Salje, E. K. H, Graeme-Barber, A. strum. 52, 75.
(1998), Eur. J. Mineral. 10, 621. Hazen, R. M., Finger L. W. (1982), Comparative
Chou, J. M., Haselton, H. T. (1998), Rev. High Pres- Crystal Chemistry. New York: John Wiley & Sons.
sure Sci. Technol. 7, 1132. Hazen, R. M., Navrotsky, A. (1996), Am. Mineral.
Crain, J., Ackland, G. J., Maclean, J. R., Piltz, R. O., 81, 1021.
Hatton, P. D., Pawley, G. S. (1994), Phys. Rev. B Hazen, R. M., Yang, H. (1999), Am. Mineral. 84,
50, 13043. 1956.
Datchi, F., LeToullec, R., Loubeyre, P. (1997), J. He, D. W., Zhang, F. X., Zhang, X. Y., Qin, Z. C.,
Appl. Phys. 81, 3333. Zhang, M., Liu, R. P., Xu, Y. F., Wang, W. K.
Dubrovinsky, L. S., Saxena, S. K., Lazor, P. (1997), (1998), J. Mater. Res. 13(12), 3458.
Geophys. Res. Lett. 24, 1835. Hemley, R. J., Prewitt, C. T., Kingma, K. J. (1994) in:
Duclos, S. J., Vohra, Y. K., Ruoff, A. L. (1987), Phys. Silica, MSA Reviews in Mineralogy, Vol. 29:
Rev. Lett. 58, 775. Heany, P. J., Prewitt, C. T. Gibbs, G. V. (Eds.).
Duclos, S. J., Vohra, Y. K., Ruoff, A. L. (1990), Phys. Washington DC: Mineralogical Society of Amer-
Rev. B 41, 12021. ica, p. 41.
Dymont, V. P., Nekrashevich, E. M., Starchenko, I. Holzapfel, W. B. (1997), in: High-Pressure Tech-
M. (1999), Solid State Commun. 111, 443. niques in Chemistry and Physics: Holzapfel, W.
Fiquet, G., Andrault, D., Itié J. P., Gillet, P., Richet, B., Isaacs, N. S. (Eds.). Oxford: Oxford University
P. (1994), EGS XIX General Assembly, Annales Press, p. 64.
Geophysicae, C6, Supplement I, Vol. 12, 24. Ivanov, A. N., Nikolaev, N. A., Pashkin, N. V., Sa-
Fiquet, G., Dewaele, A., Andrault, D., Kunz, M., venko, B. N., Smirnov, L. S., Taran, Y. V. (1991),
LeBihan, T. (2000), Geophys. Res. Lett. 27, 1; 21. Institute of Theoretical and Experimental Physics.
Fukunaga, O., Yamaoka, S., Akaishi, M., Kanda, H., Preprint No. 91 - 76. Dubna.
Osawa, T., Shimomura, Nagashima, T., Yoshi- Jackson, S. M., Nield, V. M., Whitworth, R. W.,
kawa, M. (1987), in: AGU Geophysical Mono- Oguro, M., Wilson, C. C. (1997), J. Phys. Chem. B
graph, Vol. 39, Mineral Physics 2: Manghnanni, 101, 6142.
M. H., Syono, Y. (Eds.). Washington DC: Ameri- Jamieson, J. (1957), J. Geol. 65, 334.
can Geophysical Union, p. 17. Jamieson, J. C., Lawson, A. W., Nachtrieb N. D.
Geisinger, K. L., Gibbs, G. V., Navrotsky, A. (1985), (1959), Rev. Sci. Instrum. 30, 1016.
Phys. Chem. Minerals 11, 266. Kamb, B., Davis, B. L. (1964), Proc. Natl Acad. Sci.
Gillet, P., Hemley, R. J., McMillan, P. F. (1998), in: USA 52, 1433.
Ultra-High Pressure Mineralogy, MSA Reviews in Kamb, B. (1965), Science 150, 205.
Mineralogy, Vol. 37: Hemley, R. J. (Ed.). Wash- Kamb, B., Prakash, A., Knobler, C. (1967), Acta
ington DC: Mineralogical Society of America, Crystallogr. 22, 706.
p. 525. Kamb, B., Prakash, A. (1968), Acta Crystallogr. B
Glazer, A. M. (1972), Acta Crystallogr. B 28, 2284. 24, 1317.
Golberg, D., Bando, Y., Eremets, M., Takemura, K., Katayama, Y., Mezouar, M., Itié, J. P., Besson, J. M.,
Kurashime, K., Yusa, H. (1996), Appl. Phys. Lett. Syfosse, G., LeFevre, P., DiCicco, A. (1997), J.
69, 2045. Phys. IV, Colloque 7, C2.
Goncharenko, I. N., Mirebeau, I. (1998), Rev. High Kawada, S. (1972), J. Phys. Soc. Jpn. 32, 1442.
Pressure Sci. Technol. 7, 475. Kirkpatrick, R. J. (1988), in: Spectroscopy Methods,
Goncharov, A. F., Struzhkin, V. V., Somayazulu, M. MSA Reviews in Mineralogy, Vol. 18: Hawthorne
S., Hemley, R. J., Mao, H. K. (1996), Science 273, F. C. (Ed.). Washington DC: Mineralogical Society
218. of America, p. 341.
Goñi, A. R., Syassen, K. (1998), Semiconductors and Klotz, S., Besson, J. M., Hamel, G., Nelmes, R. J.,
Semimetals 54. New York: Academic Press, p. 247. Loveday, J. S., Marshall, W. G. (1996), High Pres-
Haines, J., Leger, J. M., Chateau, C., Bini, R., Ulivi, sure Res. 14, 294.
L. (1998), Phys. Rev. B 58, R2909. Klotz, S., Besson, J. M., Braden, M., Karch, K., Pa-
Han, C. Y., Lin, W., Wu, Y. S., Yin, B., Tang, D. S. vone, P., Strauch, D., Marshall, W. G. (1997),
(1999), Physica C 282–287, 1189. Phys. Rev. Lett. 79, 1313.
Handa, Y. P., Klug, D. D., Whalley, E. (1988), Can. J. Kluthe, S., Markendorf, R., Mali, M., Roos, J.,
Chem. 66, 919. Brinkmann, D. (1996), Phys. Rev. B 53, 11369.
694 10 High Pressure Phase Transformations
Knittle, E., Kaner, R. B., Jeanloz, R., Cohen, M. L. McMahon, M. I., Nelmes, R. J. (1993), Phys. Rev. B
(1995), Phys. Rev. B 51, 12149. 47, 8337.
Komatsu, T., Nomura, M., Kakudate, Y., Fujiwara, S. McMahon, M. I., Nelmes, R. J. (1997), Phys. Rev.
(1996), J. Mater. Chem. 6, 1799. Lett. 78, 3697.
Krisch, M., Mermet, A., San Miguel, A., Sette, F., McMahon, M. I., Nelmes, R. J., Allan, D. R., Bel-
Masciovecchio, C., Ruocco, G., Verbeni, R. monte, S. A., Bovornratanaraks, T. (1998), Phys.
(1997), Phys. Rev. B 56, 8691. Rev. Lett. 80, 5564.
Kuhs, W. F., Finney, J. L., Vettier, C., Bliss, D. V. Mermelstein, N. H. (1999), Food Technol. 53, 86.
(1984), J. Chem. Phys. 81, 3612. Merrill, L., Bassett, W. A. (1974), Rev. Sci. Instrum.
Kuhs, W. F., Lobban, C., Finney, J. L. (1998), Rev. 45, 290.
High Pressure Sci. Technol. 7, 1141. Mezouar, M., Besson, J. M., Syfosse, G., Itié, J. P.,
Kunz, M., Leinenweber, K., Parise, J. B., Wu, T. C., Häusermann, D., Hanfland, M. (1996), Phys.
Bassett, W. A., Brister, K., Weidner, D. J., Vau- Status Solidi B 198, 403.
ghan, M. T., Wang, Y. (1996), High Pressure Res. Mezouar, M., LeBihan, T., Libotte, H., LeGodec, Y.,
14, 311. Häusermann, D. (1999a), J. Synchrotron Rad. 6,
Kurdyumov, A. V. (1995), Powder Metall. Metal 1115.
Ceram. 34, 409. Mezouar, M., Libotte, H., Députier, S., LeBihan, T.,
Kurdyumov, A. V., Solozhenko, V. L. (1999), Sverkh- Häusermann, D. (1999b), Phys. Status Solidi B
terdye Materialy (J. Superhard Mater.) 21, 3. 211, 395.
Lawson, A. W., Tang, T. Y. (1950), Rev. Sci. Instrum. Miletich, R., Allen, D. R., Kuhs, W. F. (2000), in:
21, 815. Comparative Crystal Chemistry, MSA Reviews in
Leinenweber, K. (1993), National Synchrotron Light Mineralogy, Vol. 39: Hazen, R. M. (Ed.). Wash-
Source, Annual Report. ington DC: Mineralogical Society of America,
Levien, L., Prewitt, C. T., Weidner, D. J. (1980), Am. Chap. 2, in press.
Mineral. 65, 920. Miller, P. H., Taylor, J. H. (1949), Phys. Rev. 76, 179.
Levien, L., Prewitt, C. T. (1981), Am. Mineral. 66, Minomura, S., Drickamer, H. G. (1962), J. Phys.
324. Chem. Solids 23, 451.
Li, B., Ganglin, C., Gwanmesia, G. D., Liebermann, Mishima, O. H., Calvert, L. D., Whalley, E. (1984),
R. C. (1998), in: Properties of Earth and Planetary Nature 310, 393.
Materials at High Pressure and Temperature. Nasu, S. (1996), High Pressure Res. 14, 405.
AGU Geophysical Monograph, Vol. 101: Mangh- Nelmes, R. J., McMahon, M. I., Hatton, P. D., Piltz,
nani, M. H., Yagi, T. (Eds.). Washington DC: R. O., Crain, J. (1993a), Phys. Rev. B 47, 35; B 48,
American Geophysical Union, p. 41. 9949 (Corrigendum).
Liebermann, R. C., Wang, Y. (1992), in: High Pres- Nelmes, R. J., Loveday, J. S., Wilson, R. M., Besson,
sure Research: Application to Earth and Planetary J. M., Pruzan, Ph., Klotz, S., Hamel, G., Hull, S.
Sciences. AGU Geophysical Monograph, Vol. 67: (1993b), Phys. Rev. Lett. 71, 1192.
Syono, Y., Manghnani, M. H. (Eds.). Washington Nelmes, R. J., Loveday, J. S., Wilson, R. M., Mar-
DC, American Geophysical Union, p. 19. shall, W. G., Besson, J. M., Klotz, S., Hamel, G.,
Line, C. M. B., Whitworth, R. W. (1996), J. Chem. Aselage, T. L., Hull, S. (1995), Phys. Rev. Lett. 74,
Phys. 104, 10008. 2268.
Liu, A. Y., Cohen, M. L. (1989), Science 245, 841. Nelmes, R. J., Liu, H., Belmonte, S. A., Loveday, J.
Liu, A. Y., Cohen, M. L. (1990), Phys. Rev. 41, S., McMahon, M. I., Allan, D. R., Häusermann,
10727. D., Hanfland, M. (1996), Phys. Rev. B 53, R2907.
Lobban, C., Finney, J. L., Kuhs, W. F. (1998), Nature Nelmes, R. J., McMahon, M. I. (1998), Semiconduc-
391, 268. tors and Semimetals 54. New York: Academic
Londono, J. D., Kuhs, W. F., Finney, J. L. (1993), J. Press, p. 145.
Chem. Phys. 98, 4878. Nguyen, J. H., Caldwell, W. A., Benedetti, L. R.,
Lubbers, R., Wortmann, G., Grünsteudel, H. F. Kruger, M. B., Jeanloz, R. (1998), Materials Re-
(1999), Hyperfine Interact. 123, 529. search Society Symposium Proceedings, Vol. 499,
Mao, H. K., Bell, P. M. (1975) Carnegie Institution p. 303.
Washington Year Book, 74, p. 402. Ochiai, A., Suzuki, T., Kasuya, T. (1985), J. Magn.
Mao, H. K., Bell, P. M., Shaner, J. W., Steinberg, D. Magn. Mater. 52, 13.
J. (1978), J. Appl. Phys. 49(6), 3276. Ohwada, K., Nakao, N., Takesu, N., Isobe, M., Ueda,
McCammon, C. (2000), in: Transformation Pro- Y., Fujii, Y. (1999), IUCr XVII Glasgow ’99. Con-
cesses in Minerals, Vol. 40: Redfern, S., Carpenter, gress and General Assembly. Abstract Book,
M., Salje E., Seifert F. (Eds.). Washington DC: P09.OC.002, 114.
Mineralogical Society of America, p. 241. Ondrey, G., Kamiya, T. (2000), Chem. Eng. 107,
McCammon, C., Peyronneau, J., Poirier, J. P. (1998), 26.
Geophys. Res. Lett. 25, 1589. Owen, E. A., Liu Y. H. (1947), Philos. Mag. 38, 354.
10.7 References 695
Parise, J. B., Weidner, D. J., Chen, J., Liebermann, Solozhenko, V. L. (1994), Diamond Related Mater. 4, 1.
R. C., Chen G. (1998), Annu. Rev. Mat. Sci. 28, Solozhenko, V. L. (1997), Eur. J. Solid State Inorg.
349. Chem. 34, 797.
Pasternak, M. P., Taylor, R. D. (1990), Hyperfine Solozhenko, V. L., Turkevich, V. Z., Holzapfel, W. B.
Interact. 62, 89. (1999), J. Phys. Chem. B 103, 2903.
Pasternak, M. P., Taylor, R. D. (1996), in: Mössbauer Solozhenko, V. L. (1999), J. Phys. Chem. B 103,
Spectroscopy applied to Magnetism and Materials 2903.
Science, Vol. 2: Long, G. J., Grandjean F. (Eds.). Stout, G. H., Jensen, L. H. (1989), X-ray Structure
New York: Plenum Press, p. 167. Determination. New York, John Wiley & Sons.
Pasternak, M. P., Farrell, J. N., Taylor, R. D. (1986), Sung, C. M., Sung, M. (1996), Mater. Chem. Phys.
Hyperfine Interact. 28, 837. 43, 1.
Paul, W. (1959), J. Phys. Chem. Solids 8, 196. Tammann, G. (1900), Annalen der Physik, Series 4,
Pedersen, K. (1997), FEMS Microbiol. Rev. 20, 399. 2, 1.
Pelton, A. D. (2001), in: Phase Transformations in Teter, D. M., Hemley, R. J. (1996), Science 271, 53.
Materials: Kostorz, G. (Ed.). Weinheim-VCH, Thévenot, F. (1980), J. Europ. Ceramic Soc. 6, 205.
Chap. 1. Van Valkenburg, A. (1964), Diamond Res. 17.
Petrenko, V. F., Whitworth, R. W. (1999), The Phys- Velde B., Besson J. M. (1982), in: High Pressure in
ics of Ice, Oxford: Oxford University Press. Research and Industry, 8th AIRAPT Conference
Poirier, J. P. (1991), Introduction to the Physics of the Proceedings, Vol. 2, p. 704.
Earth’s Interior, Cambridge: Cambridge Univer- Vohra, Y. K. Brister, K. E., Desgreniers, S., Ruoff, A.
sity Press, Chap. 4.7. L., Chang, K. J., Cohen, M. L. (1986), Phys. Rev.
Pruzan, Ph., Chervin, J. C., Canny, B. (1993), J. Lett. 56, 1944.
Chem. Phys. 99, 9842. Walrafen G. E., Abebe, M., Mauer, F. A., Block, S.,
Pruzan, Ph., Wolanin, E., Gauthier, M., Chervin, J. Piermarini, G. J., Munro, R. (1982), J. Chem.
C., Canny, B., Häusermann, D., Hanfland, M. Phys. 77, 2166.
(1997), J. Phys. Chem. B 101, 6230. Walsh, J. M., Christian, R. H. (1955), Phys. Rev. 97,
Rigden, S. M., Gwanmesia, G., Jackson, I., Lieber- 1544.
mann, R. C. (1992), in: High Pressure Research: Weidner, D. J., Vaughan, M. T., Ko, J., Wang, Y., Liu,
Application to Earth and Planetary Sciences. X., Yeganeh-Haeri, A., Pacalo, R. E., Zhao, Y.
AGU Geophysical Monograph, Vol. 67: Mangh- (1992), in: High Pressure Research: Application to
nani, M., Syono, Y. (Eds.). Washington DC, Amer- Earth and Planetary Sciences. AGU Geophysical
ican Geophysical Union, p. 167. Monograph, Vol. 67: Syono, Y., Manghnani, M. H.
Ross, N. L., Price G. D. (1992) (Eds.), The Stability (Eds.). Washington DC, American Geophysical
of Minerals, London: Chapman & Hall, Chap. 1. Union, p. 13.
Ruoff, A. L., Li, T. (1995), Annu. Rev. Mater. Sci. 25, Weidner, D. J., Wang, Y., Vaughan, M. T. (1994a),
249. Geophys. Res. Lett. 21, 753.
Schwarz, U., Takemura, K., Hanfland, M., Syassen, Weidner, D. J., Wang, Y., Vaughan, M. T. (1994b),
K. (1998), Phys. Rev. Lett. 81, 2711. Science 266, 419.
Shen, A. H., Reichmann, H. J., Chen, G., Angel, R. Weir, C. E., Lippincott, E. R., Van Valkenburg, A.,
J., Bassett, W. A., Spetzler, H. (1998), in: Prop- Bunting, E. N. (1959), J. Res. Natl Bur. Standards
erties of Earth and Planetary Materials at High USA 63A, 55.
Pressure and Temperature. AGU Geophysical Weir, S. T., Vohra, Y. K., Vanderborgh, C. A., Ruoff,
Monograph, Vol. 101: Manghnani, M. H., Yagi, T. A. L. (1989), Phys. Rev. B 39, 1280.
(Eds.). Washington DC, American Geophysical Whalley, E., Davidson, D. W., Heath, J. B. R. (1966),
Union, p. 71. J. Chem. Phys. 45, 3976.
Shimomura, O., Kawamura, T., Fukamachi, T., Ho- Whalley, E., Heath, J. B. R., Davidson, D. W. (1968),
soya, S., Hunter, S., Bienestock, A. (1980), in: J. Chem. Phys. 45, 3976.
High Pressure Science and Technology, Vol. 1: Wittlinger, J., Werner, S., Schulz, H. (1998), Acta
Vodar, B., Marteau, P. (Eds.). Oxford: Pergamon Crystallogr. B 54, 714.
Press, p. 534. Woelk, K., Bargon, J. (1992), Rev. Sci. Instrum. 63,
Shimomura, O., Takemura, K., Fujihisa, H., Fuji, Y., 3307.
Ohishi, Y., Kikegawa, T., Amemiya, Y., Matsu- Xu, J. A, Mao, H. K, Bell, P. M (1986), Science 232,
shita, T. (1992), Rev. Sci. Instrum. 63, 967. 1404.
Sinogekin, S. V., Bass, J. D. (1999), Phys. Rev. B 59, Yagi, T., Akaogi, M., Shimomura, O., Tamai, H., Ak-
R14141. imoto, S. (1987), in: High Pressure Research in
Slykhouse, T. E., Drickamer, H. G. (1958), J. Phys. Mineral Physics. AGU Geophysical Monograph,
Chem. Solids 7, 207. Vol. 39: Manghnani, M. H., Syono, Y. (Eds.).
Smirnov, G. V. (1999), Hyperfine Interact. 123/124, Washington DC, American Geophysical Union,
31. p. 141.
Index