Physica B 406 (2011) 30893097

Contents lists available at ScienceDirect

Physica B
journal homepage: www.elsevier.com/locate/physb

Atomic structure and diffusivity in liquid Al80Ni20 by ab initio molecular

dynamics simulations
W.Y. Wang a,b,n, H.Z. Fang a, S.L. Shang a, H. Zhang a, Y. Wang a, X. Hui b, S. Mathaudhu c, Z.K. Liu a
a

Department of Materials Science and Engineering, The Pennsylvania State University, University Park, PA 16802, USA
The State Key Laboratory of Advanced Metals and Materials, University of Science and Technology Beijing, 100083 Beijing, China
c
Materials Science Division, US Army Research Ofce, Research Triangle Park, NC 27709, USA
b

a r t i c l e i n f o

abstract

Article history:
Received 15 February 2011
Received in revised form
12 April 2011
Accepted 4 May 2011
Available online 7 May 2011

The atomic structure and diffusivity in liquid Al80Ni20 are studied by ab initio molecular dynamics
simulations. The local structures are analyzed by the pair correlation function, structure factor,
coordinate number, HonneycuttAnderson bond pair, and Voronoi tessellation methods. It is observed
that the amount of icosahedral clusters increases, and the liquid becomes more ordered as the
temperature decreases. The predicted self-diffusion coefcients of Al and Ni via the mean square
displacements are very close to each other and agree well with the quasi-elastic neutron scattering
measurements in the literature. The observation of equal self-diffusivity of Al and Ni is attributed to the
formation of local solute-centered polyhedra, coupling the migration of Al and Ni. The Manning
dynamic correlation factor is evaluated and found to be close to unity. The predicted interdiffusion
coefcients using the Darken equation agree well with experimental data in the literature.
& 2011 Elsevier B.V. All rights reserved.

Keywords:
Diffusivity
Liquid Al80Ni20
AIMD
Short range order

1. Introduction
Metallic glasses are an important group of materials with
properties signicantly different from their crystal counterparts
[14]. Metallic glasses form once the nucleation of crystalline
phases from liquid is prohibited during quenching. One way to
understand the tardiness of crystallization is to study how atomic
structures of liquid evolve as the temperature decreases along
with the atomic mobility in the liquid. They have been typically
obtained through classic molecular dynamics (MD) simulations
[59]. In recent years, with the development of more efcient
algorithms and higher computing powers, the ab initio molecular
dynamics simulations (AIMD) introduced by Car and Parrinello
[10] are becoming more widely used [1114]. The advantage of
the AIMD approach over the classic MD approach is that the
atomic forces in AMID simulations are calculated on the y based
on the density functional theory [15]. In a series of studies, we
have used the AIMD approach to derive the detailed atomic
structures in a wide range of simple and complex alloys with
tendency to form bulk metallic glasses (BMG) [1620], demonstrating the great potential of the AIMD approach.
The chemical and structural short-range ordering and diffusion
coefcients in AlNi alloys have been studied via neutron
n
Corresponding author at: Department of Materials Science and Engineering,
The Pennsylvania State University, University Park, PA 16802, USA.
Tel.: 1 814 863 9957.
E-mail address: yuw129@psu.edu (W.Y. Wang).

0921-4526/$ - see front matter & 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.physb.2011.05.013

scattering and MD simulations [2129]. Maret et al. [21] and

Gruner et al. [23] investigated the structure factors through the
neutron scattering, demonstrating the existence of chemical
short-range ordering. Jakse et al. [2729] observed the strong
short-range ordering tendency of the Al80Ni20 liquid by using
AIMD simulations. Furthermore, Horbach et al. [22] and Griesche
et al. [26] explored the interdiffusion coefcient from the measured tracer diffusion coefcients via Darken equation with the
Manning factor. However, the inuence of short-range ordering
and the thermodynamic factor on chemical diffusion in liquid has
not been well studied.
In the present work, we carry out AIMD simulations for liquid
Al80Ni20 to understand its atomic structures and atomic diffusion
coefcients in our efforts of developing low density Al-based
metallic glasses. The results from the current in-depth analysis of
AIMD simulations are compared with both experimental data and
MD simulations in the literatures. Furthermore, both dynamic
correlation factor and thermodynamic factor are evaluated from
the AIMD simulations and used in calculating the interdiffusion
coefcients. It is demonstrated that the AIMD simulations, even
though with shorter simulation time and smaller supercell sizes
than classic MD simulations, generate reliable structural and
diffusion data in comparison with experimental results. The paper
is organized as follows. In Section 2 the simulation approach
and the characterization techniques are discussed in details. The
results are listed in the Section 3 in terms of structure analysis
and diffusivity evaluations. Finally, a summary is presented in
Section 4.

3090

W.Y. Wang et al. / Physica B 406 (2011) 30893097

2. Simulation and evaluation methodology

The AIMD simulations in the present work are conducted by
employing the Vienna ab initio simulation package (VASP) [30,31]
with the generalized gradient approximation [32] for the exchange
correlation functional and the projector augmented wave [33]
method for the electronion interaction. The Newtons equation of
motion is solved via the Verlets algorithm with a time step of 5 fs,
and the simulations were performed at the G point only with a low
precision as commonly used in AIMD simulations [11,12,33,34]. All
the calculations are carried out in canonical ensemble, i.e. constant
volume and temperature, with atomic conguration relaxation and
temperature controlled by a Nose thermostat [35]. The wave
functions are sampled on 1  1  1 k-point mesh in terms of the
MonkhorstPack scheme [36]. The plane wave cutoff energy is
269.6 eV, and the energy convergence criterion of electronic selfconsistency is chosen as 1  10  4 meV/atom for all the calculations
(both default values). At each temperature, the supercell volume is
systematically varied through the relaxation setting ISIF7 in
VASP, and the equilibrium volume is obtained when the pressure
equals zero.
The cubic supercell of the Al80Ni20 alloy is initially constructed
with 200 atoms randomly distributed. The initial conguration is
equilibrated at a temperature well above the melting point to
ensure an equilibrium liquid state. AIMD simulations are carried
out at 2000, 1900, 1800 and 1330 K, respectively, with an equilibrated time of 10 ps. AIMD calculations with a supercell of 150
atoms are also carried out, showing almost identical results as the
supercell of 200 atoms. All results reported in this paper are from
the supercell with 200 atoms except specied otherwise. 2000
congurations are collected at each temperature for diffusion
analysis, and the last hundred congurations are used for structure
analysis with the pair correlation function, structure factor, coordination number, bond pair analysis, and Voronoi tessellation methods. The mean square displacements are used for calculating the
diffusion coefcients according to the Einsteins relation.
In the structure analysis, the partial pair correlation function,
gij(r), is dened as follows [18,20,37,38]:
*
+
Ni
X
nij r, Dr
V
gij r
1
Ni Nj i 1 4pr 2 Dr
where V denotes the volume of the supercell, Ni and Nj are the
numbers of i and j atoms in the supercell, nij(r,Dr) is the number
of j atoms in the sphere shell from r to r Dr of the i atom, and the
bracket / S represents the time average of different congurations obtained from molecular dynamics simulations. The total
pair correlation function, gtotal(r) is the weighted sum of the
partial pair correlation functions [18,39]
n X
n
n X
n
X
X
ci cj fi qfj q
oij gij r
2

2 gij r
f q
i1j1
i1j1

2
P
where f q ci fi q2 with fi q being the atomic scattering
factor, ci the mole fraction of element i, and q(4p/l)sin y the
transferred momentum at the diffraction angle y and the X-ray
wavelength l. Neutron scattering measurements give the atomic
scattering factor of Al and Ni being 3.45 and 10.31, respectively [40].
The weight factor oij is typically treated as a constant independent
of r, but composition dependent, equal in reciprocal and real space
as in the case of neutron scattering experiments [23]. The reduced
pair correlation function, Gij(r), and reduced generalized correlation
function, G(r), are given as [39,41]

gtotal r

Gij r 4pr0 rgij r1

Gr


1
g r4pr0 r 2
r total

3
4

where r0 is the average atomic density in the supercell, and r the

atomic density in the spherical shell with a radius of r and the total
structure factor of Stotal(q). For binary systems, Faber and Ziman [42]
showed that Stotal(q) can be obtained from the scattered intensity
per atom through the following equation:
hD
E 
2 i
n X
n
Iq f q2  f q
X
Stotal q

oij Sij q
5

2
f q
i1j1
where Sij(q) is the partial structure factor between atoms i and
j obtained from the Fourier transform of gij(r), as follows
[20,28,43,44]:
Z 1
sinqr
dr
6
Sij q 1
4pr0 r 2 gij r1
qr
0
On the other hand, Bhatia and Thornton [45] represented the
scattering function of a binary alloy in terms of three structure
factors: the numbernumber structure factor, SNN(q), the
numberconcentration structure factor, SNC(q), and the
concentrationconcentration structure factor SCC(q). At temperatures above the Debye temperature and in the long-wavelength
limit (q-0), SNN(0) and SCC(0) denote the mean square thermal
uctuations in the atom number and concentration, respectively,
and SNC(0) represents the correlation between these two uctuations [44,45]. They can be calculated from the FaberZiman
structure factors from Eq. (6)
SNN q ci2 Sii q cj2 Sjj q 2ci cj Sij q

SNC q ci cj ci Sii qSij qcj Sjj qSij q

SCC q ci cj 1 ci cj Sii q Sjj q2Sij q

SCC(0) is related to the second derivative of the molar Gibbs

energy g [22,45] through the thermodynamic factor, F

ci cj @2 g
ci cj

kB T @ci @cj
SCC 0

10

Further detailed geometrical analyses include the average

coordination number, the bond pair analysis, and the Voronoi
tessellation to understand the atomic short- and medium-range
ordering in the liquid. The partial coordination number, Zij, is
dened as the number of j atoms in the rst neighboring sphere
centered at i atom with the radius rmin and can be calculated from
the partial pair correlation functions using the following equation
[18]:
Z rmin
Zij
4pr 2 rj gij r dr
11
0

where rmin is the cutoff distance dened by the rst-minimum

position of gij r, and rj the average atomic density of element j.
The average coordination number of element i can be obtained as
P
Zi j Zij . The average coordination number of a solution can be
estimated by Z ci Zii Zji cj Zjj Zij [46], which can also be
calculated in terms of Voronoi index n3 ,n4 ,n5 ,n6 ,. . ., where nk
denotes the number of k-edged face of a Voronoi polyhedron,
P
with Z
nk being the average coordination number [12,20].
For the bond pair analysis, the method proposed by Honeycutt
and Anderson [47] is used with the local environment represented by the indexes i, j, k and l. When the distance between root
pair (atoms A and B) is less than their bonding distance
measured by the rst minimum position of gij(r), the two atoms
are called a bond pair, with i1, otherwise i2. The index j
represents the number of common nearest-neighbor atoms forming bonds with the root pair, the index k denotes the number of
bonds among the neighboring atoms; and the index l is a
parameter used to distinguish local structures when i, j, k are
the same. In this notation, 1551, 1541 and 1431 represent the

W.Y. Wang et al. / Physica B 406 (2011) 30893097

icosahedral type of clusters, 1441 and 1661 the BCC-type bond

pairs, 1421 and 1422 the FCC and HCP-type bond pairs, and 1311
and 1321 the disordered atomic arrangements [5,13,20,4749].
In MD simulations, the self-diffusion coefcient of atom i can
be determined as shown below [22,48,5052]:
DP
E
2
Ni
9Rj t t0 Rj t0 9
/Ri 2 tS
j1
Di lim
lim
12
t-1
t-1
6t
6Ni t
2

where /Ri tS is the mean-square displacement of component i,

Ni the number of atom i, Rj the coordinates of the jth atom i, t0 the
P i
2
origin of time, and N
j 1 9Rj t t0 Rj t0 9 the ensemble average
of the square displacement of atom i. In a binary AB system,
Darken [53] derived the relation between the interdiffusion
coefcient and self-diffusion coefcients as
~ AB cB DA cA DB F
D

13

where F is the thermodynamic factor. Bardeen [54] and Manning

[55] modied the Darken equation by adding a dynamic correlation factor to take into account the ux difference of different
atoms. The dynamic correlation factor, l, can be calculated from
the velocity correlation functions, shown as follows [22]:

Lkl t

Nk X
Nl
1 X
!k
!l
/ m i tU m j 0S,
3Nck cl i 1 j 1

ia j when k l

14

3091

Z 1
cA cB
LAA t LBB t2LAB t dt
15
cA DA cB DB 0
!k
where m i t denotes the velocity vector of ith atom of the k
species at time t, and /  S the time average. Eq. (13) is thus
modied as

l 1

DAB cB DA cA DB Fl

16

The positions of individual atoms and their atomic environments are analyzed after each AIMD simulation step, and the
resulting pair correlation functions, structure factors, coordination numbers, bond pairs, mean-square displacements, dynamic
correlation factor, and diffusion coefcients are presented in the
next section.

3. Results and discussion

3.1. Pair correlation functions and structure factors
Fig. 1 shows the pair correlation functions as a function of
temperature with Fig. 1(a) for gtotal(r), and Fig. 1(b)(d) for GAlAl(r),
GAlNi(r) and GNiNi(r), respectively, along with available experimental
neutron scattering data at 1330 K [21] and classic MD simulations at
1320 K [28] in the literature. Remarkable agreement between the
current AIMD simulations and experimental data can be observed

Fig. 1. Pair correlation function of the liquid Al80Ni20 alloy (a) gtotal(r), with each curve shifted by 0.5, and the reduced partial pair correlation functions, (b) GAlAl, (c) GAlNi,
and (d) GNiNi with the experimental [21] and the MD [28] data superimposed.

3092

W.Y. Wang et al. / Physica B 406 (2011) 30893097

except the height of the rst peak in GNiNi, which is probably due to
the supercell size. The classic MD simulation results seem different
from the experimental data because the atomic density was not
included in evaluating the reduced pair correlation function (see
Eq. (3)). The total structure factors, STotal q, and the partial structure
factors, SAlAl(q), SAlNi q, and SNiNi q, are shown in Fig. 2, again
compared excellently with data from neutron scattering measurements with slight differences in SNiNi(q) and the third peak of
SAlNi q.
Fig. 3 shows the BhatiaThornton structure factors derived
from the FaberZiman structure factors (Eqs. (7)(9)) in comparison with the experimental data [21] and the MD simulation
results [22] in the literature. At high temperatures, the atoms are
randomly distributed, and the concentrationconcentration
BhatiaThornton structure factor, SCC q, varies slightly with
respect to temperature. At low temperatures, the number of
clusters or short-range ordering increases, and the positions of
the rst peaks of both SNC q and the SCC q move to smaller q, and
the widths of their splitting second peaks decrease, but their

intensities increase, e.g. SNC(q) 0.15 at 1800 K and

1
SNC q 0:19 at 1330 K at the same q 2:46 A , as shown in
Fig. 3(c) and (d). The differences in the peak positions and
intensities of SNC between MD data and our results at 2000 K
shown in Fig. 3(c) is due to the different formalism of Bhatia
Thorntons structure factor used in the Ref. [22] and in the present
work. It should be noted that the BhatiaThornton formalism [45]
is usually used in analysis of neutron scattering experiments
[21,46] as did in the present work, enabling a fair comparison
with experimental data. The splitting peak in the structure factor
describes repeating structural units involving neighboring clusters [22]. The uctuation of the concentration could be extracted
from the limit of SCC q-0, and the small value of SCC(q0)
reveals a strong ordering tendency.
3.2. Local structure analysis
Fig. 4 plots the distribution of average coordination number as
a function of temperature. It is observed that the coordination

Fig. 2. FaberZiman structure factor of the liquid Al80Ni20 alloy at various temperatures shifted by 0.5 between neighboring curves: (a) total structure factor, STotal(q); and
partial structure factor (b) AlAl, SAlAl(q); (c) AlNi, SAlNi(q); and (d) NiNi, SNiNi(q). The experimental results (red stars in the gures) are determined at 1330 about 70 K
higher than the liquidus temperature [21]. (For interpretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

W.Y. Wang et al. / Physica B 406 (2011) 30893097

3093

Fig. 3. BhatiaThornton structure factor of liquid Al80Ni20 with the neutron scattering data [21] and the MD data [22]: (a) various structure factors at 1330 K; and structure
factors at various temperatures (b) numbernumber, SNN(q); (c) numberconcentration, SNC(q); (d) concentrationconcentration, SCC(q). Note that the multiplication by
1/cicj is used to increase the amplitude of SCC q, leading the asymptotic value SCC(q) 1 for q-N. The plots at different temperatures are shifted from each other in steps of
0.5 in (b) and 0.1 in (c) and (d).

numbers of majority atoms are between 11 and 14. The average

coordination number changes from 10.0(1) with rmin 4.02 A at
2200 K to 10.1(4) with rmin 3.43 A at 1330 K, indicating the
distance between the rst nearest neighbors and the centered
atom is decreasing with the temperature in order to keep the
closed packed structure.
The percentages of various bound pairs and their summations
into icosahedral, FCC and HCP, and BCC groups are shown in
Fig. 5. It depicts that the icosahedral bond pairs are dominant at
temperatures above the melting temperature, followed by the
FCC and HCP-type bond pairs and the BCC-type bond pairs. The
icosahedral clusters prevent the formation of a long-range period
lattice and contribute to the formation of many disordered
icosahedra and their associated inter-cluster voids [56]. Our AIMD
results are in agreement with the AIMD simulations by Jakse and
Pasturel [27], who reported that the sum of the 1551 and 1541
bond pairs increases to more than 46% at 1220 K, and the sum of

1441, 1431, 1421 and 1422 bond pairs decreases to 40%, which
are 26% and 41% at 1330 K in our simulations, respectively.
The local atomic packing in liquid Al80Ni20 is further explored
by the Voronoi tessellation method, shown in Fig. 6. In binary
systems, it has been found [12,19,20] that the preferred polyhedron type depends on the ratio of solute to solvent (Rn) atomic
sizes, i.e. the FrankKasper type with Rn 41.2, the icosahedral
type with Rn o0.902, the bi-capped square Archimedean antiprism type with Rn o 0.835, and the tricapped trigonal prism
packing type with Rn o0.732. In liquid Al80Ni20, Rn equals to
rNi 1.62 A),
indicating the tendency to form
0.89 (rAl 1.82 A,
icosahedral type clusters. It is observed in Fig. 6 that Voronoi
polyhedra with the index of /0, 2, 8, 1S, /0, 2, 8, 2S, /0, 3, 6, 1S,
/0, 3, 6, 2S, /0, 3, 6, 3S and /0, 3, 6, 4S are the major polyhedra,
among which the percentages of /0, 2, 8, 1S, /0, 2, 8, 2S, and
/0, 3, 6, 3S polyhedra are the highest. Furthermore, the amount
of the perfect icosahedron with the index of /0, 0, 12, 0S

3094

W.Y. Wang et al. / Physica B 406 (2011) 30893097

Fig. 4. Percentages of average coordination number (CN) at various temperatures: (a) coordination numbers between 7 and 12 and (b) coordination numbers between 13
and 17.

Fig. 5. Percentages of bond pairs as a function of temperature: (a) various bond pairs, (b) the variation of icosahedral type (1551, 1541 and 1431), FCC and HCP type
(1422 and 1421), BCC type (1441 and 1661), and random type (1311, 1321 and others). The other bond pairs less than 1% are not shown.

increases dramatically at low temperatures. We can thus conclude that the icosahedral structure, including the defective ones
/0, 2, 8, 2S, /0, 2, 8, 1S and /0, 3, 6, 3S, is the dominant
structure in liquid Al80Ni20, in agreement with the bond pair
analysis discussed above.
The aforementioned clusters can form local ordering structures through vertex-, edge-, face- and intercross-sharing of
neighboring clusters. These sharing schemes lead to correlation
peaks in the partial pair correlation functions [57,58]. Fig. 7 shows
an icosahedral medium range ordering at 2000 K, consisting of
two /0, 3, 6, 3S and two /0, 0, 12, 0S. Comparing this
icosahedral medium-range ordering with the partial correlation
functions, Al atoms can be found from an Al- or Ni-center nearest
neighbor shell and the extended nearest neighbor shell, so both
GAlAl(r) and GNiAl(r) have the rst and second peaks (see Fig. 1).
Furthermore, at 1500 and 1330 K, there is clearly third peak in the
gtotal(r) in Fig. 1(a), indicating the icosahedra medium-range
ordering at low temperatures. Since the Ni atoms are far from
each other, the height of the rst peak of GNiNi(r) is smaller than

the extended peaks (see Fig. 1(d)). Additionally, the shoulder

peaks in the structure factor around the second peak in STotal q in
Fig. 2(a) indicate the existence of chemical short-range ordering
[23,24].
3.3. Diffusion coefcients
Fig. 8 shows the mean-square displacements of Al and Ni as a
function of time at various temperatures obtained in the present
work. For time below 0.1 ps, the mean-square displacement is
proportional to t2, as expected for ballistic motion [59]. For longer
times the mean-square displacement increases linearly with time,
which is the indication of long-range diffusion [59]. The meansquare displacements starting at 2 A 2 or later than 0.5 ps and
ending at larger than 4 A 2 are used for diffusion analysis. The
ln D vs: 1=T curve is shown in Fig. 9 and agrees very well with the
result from the quasi-elastic neutron scattering experiments [22].
Both our simulations and the experiments [60] show that the DNi
in binary alloys is higher than the self-diffusion coefcient of pure

W.Y. Wang et al. / Physica B 406 (2011) 30893097

3095

Fig. 6. Percentages of various Voronoi polyhedra grouped in terms of coordination number as a function of temperature.

Fig. 7. Icosahedra medium-range order with the Ni atoms in the center of

icosahedra (golden, darker) and other positions (red, lighter), and Al atoms in
the center (yellow, darker) and other positions (blue, lighter). The icosahedra are
linked by vertex-shared (VS), face-shared (FS), and intercross-shared (IS) atoms.
(For interpretation of the references to colour in this gure legend, the reader is
referred to the web version of this article.)

liquid Ni [61,62] and is close to DAl. This can be understood

through cluster ordering discussed in the previous section.
As shown in Fig. 7, many Ni atoms are located in polyhedron
centers, and the displacements of polyhedra are thus largely
determined by the bonding between polyhedral, i.e. Al atoms.
Consequently, the migrations of Ni and Al are collaborative,
resulting in similar self-diffusion coefcients close to that of Al.
To calculate interdiffusion coefcients, the thermodynamic
factor and dynamic correlation factor need to be evaluated. The
thermodynamic factor can be obtained from SCC(q) in MD and
AIMD simulations [22] or from the thermodynamic database [63].
The results thus obtained are compared in Fig. 10 as a function of
temperature. It can be seen that the thermodynamic factor
obtained from the thermodynamic database is about double of
those obtained from MD simulations in the literature, while the
values from the present AIMD simulations are between. The
dynamic correlation factor evaluated from the current AIMD
simulations is plotted in Fig. 11 with the initial step considered
!
as the reference state, m i 0, showing that the Manning dynamic

3096

W.Y. Wang et al. / Physica B 406 (2011) 30893097

Fig. 8. Mean-squared displacements of Al and Ni from AIMD simulations at 2000,

1900, 1800 and 1330 K from the supercell with 200 atoms and 2200, 1500 K from
the supercell with 150 atoms.

Fig. 10. Thermodynamic factors in liquid Al80Ni20 as a function of temperature

calculated from the thermodynamic database by Dupin et al. [63], MD simulations
by Horbach et al. [22], and current AIMD simulations.

Fig. 9. Self-diffusion coefcients of Al and Ni in liquid Al80Ni20 compared with the

data of pure Ni by Meyer et al. [61] using quasielastic neutron scattering (QNS)
and Zhang et al. [62] by AIMD and the experimental and MD data in liquid Al80Ni20
by Maret et al. [21] and Horbach et al. [22].

Fig. 11. Manning dynamic correlation factor as a function of simulation time step
!
with the initial step considered as the reference state, m i 0.

correlation factor l in Eq. (15) approaches unity with increasing

simulation time. The diffusion mechanism in liquid Al80Ni20 thus
appears to be the same as for viscous ow [64] and strongly
affected by local chemical ordering [65], similar to the conclusion
by Tang et al. [66,67] when examining the characteristics of Be
diffusion in Zr-based BMG alloys where the diffusion process of
Be involves a group of atoms.
The interdiffusion coefcients calculated using the thermodynamic factors obtained from three sources are plotted in Fig. 12
with the experimental data and MD simulation results in the
literature superimposed [22]. It is observed that the interdiffusion
coefcients calculated by the Darken relation using the thermodynamic factors from the thermodynamic database and current
AIMD simulations are both within the experimental uncertainty
range, while the thermodynamic factor from the MD simulations
in the literature gives much lower interdiffusion coefcients in
comparison with the experimental data.

4. Conclusions
In summary, the atomic congurations and diffusion coefcients in liquid Al80Ni20 are obtained by means of the AIMD
simulations with following conclusions:

 The generalized pair correlation functions and FaberZiman



total and partial structure factors show a good agreement with

the experimental data.
The bond pair analysis shows that the icosahedral bond pairs
are dominant, followed by the FCC- and HCP-type bond pairs
and the BCC-type bond pairs. The percentage of the perfect
icosahedral clusters increases as the temperature decreases.
The predicted self-diffusion coefcients of Al and Ni via the
mean square displacements are almost equal, in agreement with the quasielastic neutron scattering data in the
literature.

W.Y. Wang et al. / Physica B 406 (2011) 30893097

Fig. 12. Interdiffusion coefcient in liquid Al80Ni20 calculated by the Darken

equation with thermodynamic factors shown in Fig. 10, compared with the MD
and experimental results by Horbach et al. [22].

 The Manning dynamic correlation factor is found to approach

unity with increasing simulation time. The interdiffusion
coefcients obtained from Darken equation show good agreement with experimental data.

Acknowledgements
This work was nancially supported by the National Science
Foundation (Grant no. DMR-1006557) and the Army Research
Laboratory (Contract no. W911NF-08-2-0064) in the Unites
States, National Natural Science Foundation of China (Grant nos.
50431030 and 50871013), and National Basic Research Program
of China (Grant no. 2007CB613901). W.Y. Wang acknowledges the
support from the Project Based Personnel Exchange Program with
China Scholarship Council and American Academic Exchange
Service ([2008] 3072). First-principles calculations were carried
out on the LION clusters at the Pennsylvania State University
supported by the Materials Simulation Center and the Research
Computing and Cyberinfrastructure unit at the Pennsylvania State
University. Calculations were also carried out on the INTI clusters
from the Computer Science Department at Pennsylvania State
University supported by NSF under Grant no. CISE-0202007 and
CyberStar cluster funded by NSF through grant OCI-0821527.

References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]

J. Schroers, Advanced Materials 22 (2010) 1566.

C.A. Angell, MRS Bulletin 33 (2008) 544.
A.L. Greer, Materials Today 12 (2009) 14.
P.G. Debenedetti, F.H. Stillinger, Nature 410 (2001) 259.
K.Y. Chen, H.B. Liu, X.P. Li, Q.Y. Han, Z.Q. Hu, Journal of Physics-Condensed
Matter 7 (1995) 2379.
Y.N. Zhang, L. Wang, W.M. Wang, Physics Letters A 372 (2008) 690.
S. Sastry, C.A. Angell, Nature Materials 2 (2003) 739.
A. Samanta, S.M. Ali, S.K. Ghosh, Physical Review Letters 87 (2001) 245901.
T. Tomida, T. Egami, Physical Review B 52 (1995) 3290.
R. Car, M. Parrinello, Physical Review Letters 55 (1985) 2471.
C. Woodward, M. Asta, D.R. Trinkle, J. Lill, S. Angioletti-Uberti, Journal of
Applied Physics 107 (2010) 113522.
H.W. Sheng, W.K. Luo, F.M. Alamgir, J.M. Bai, E. Ma, Nature 439 (2006) 419.
P. Ganesh, M. Widom, Physical Review B 74 (2006) 134205.

3097

[14] M.N. Asta, V. Ozolins, J.J. Hoyt, M. van Schilfgaarde, Physical Review B 64
(2001) 020201.
[15] R. Iftimie, P. Minary, M.E. Tuckerman, Proceedings of the National Academy
of Sciences of the United States of America 102 (2005) 6654.
[16] Z.K. Liu, Journal of Phase Equilibria and Diffusion 30 (2009) 517.
[17] X. Hui, H.Z. Fang, G.L. Chen, S.L. Shang, Y. Wang, Z.K. Liu, Applied Physics
Letters 92 (2008) 201913.
[18] R. Gao, X. Hui, H.Z. Fang, X.J. Liu, G.L. Chen, Z.K. Liu, Computational Materials
Science 44 (2008) 802.
[19] H.Z. Fang, X. Hui, G.L. Chen, Z.K. Liu, Applied Physics Letters 94 (2009)
091904.
[20] X. Hui, H.Z. Fang, G.L. Chen, S.L. Shang, Y. Wang, J.Y. Qin, Z.K. Liu, Acta
Materialia 57 (2009) 376.
[21] M. Maret, T. Pomme, A. Pasturel, P. Chieux, Physical Review B 42 (1990) 1598.
[22] J. Horbach, S.K. Das, A. Griesche, M.P. Macht, G. Frohberg, A. Meyer, Physical
Review B 75 (2007) 174304.
[23] S. Gruner, J. Marczinke, L. Hennet, W. Hoyer, G.J. Cuello, International Journal
of Materials Research 101 (2010) 741.
[24] I. Egry, L. Hennet, M. Kehr, G. Mathiak, S. De Panlis, I. Pozdnyakova,
D. Zanghi, Journal of Chemical Physics 129 (2008) 064508.
[25] S.K. Das, J. Horbach, T. Voigtmann, Physical Review B 78 (2008) 064208.
[26] A. Griesche, M.P. Macht, G. Frohberg, Defect and Diffusion Forum 266 (2007)
101.
[27] N. Jakse, A. Pasturel, Modern Physics Letters B 20 (2006) 655.
[28] N. Jakse, O. Le Bacq, A. Pasturel, Journal of Chemical Physics 123 (2005)
104508.
[29] N. Jakse, O. Lebacq, A. Pasturel, Physical Review Letters 93 (2004) 207801.
[30] G. Kresse, J. Furthmuller, Physical Review B 54 (1996) 11169.
[31] G. Kresse, J. Furthmuller, Computational Materials Science 6 (1996) 15.
[32] Y. Wang, J.P. Perdew, Physical Review B 44 (1991) 13298.
[33] G. Kresse, D. Joubert, Physical Review B 59 (1999) 1758.
[34] P.E. Blochl, Physical Review B 50 (1994) 17953.
[35] S. Nose, Journal of Chemical Physics 81 (1984) 511.
[36] H.J. Monkhorst, J.D. Pack, Physical Review B 13 (1976) 5188.
[37] H.Z. Fang, X. Hui, G.L. Chen, R. Ottking, Y.H. Liu, J.A. Schaefer, Z.K. Liu,
Computational Materials Science 43 (2008) 1123.
[38] Y.N. Zhang, L. Wang, W.M. Wang, J.K. Zhou, Physics Letters A 355 (2006) 142.
[39] K.D. Machado, J.C. de Lima, T.A. Grandi, Solid State Communications 143
(2007) 153.
[40] V.F. Sears, S. Kurt, L.P. David, Methods in Experimental Physics, vol. 23,
Academic Press, 1986, p. 521.
[41] R. Kaplow, S.L. Strong, B.L. Averbach, Physical Review A 138 (1965) 1336.
[42] T.E. Faber, J.M. Ziman, Philosophical Magazine 11 (1965) 153.
[43] M. Maret, P. Chieux, J.M. Dubois, A. Pasturel, Journal of Physics-Condensed
Matter 3 (1991) 2801.
[44] L. Wang, H. Cong, Y. Zhang, X. Bian, Physica B: Condensed Matter 355 (2005)
140.
[45] A.B. Bhatia, D.E. Thornton, Physical Review B 2 (1970) 3004.

[46] D. Holland-Moritz, S. Stuber,

H. Hartmann, T. Unruh, T. Hansen, A. Meyer,
Physical Review B 79 (2009) 064204.
[47] J.D. Honeycutt, H.C. Andersen, Journal of Physical Chemistry 91 (1987) 4950.
[48] H.Z. Fang, X. Hui, G.L. Chen, Z.K. Liu, Physics Letters A 372 (2008) 5831.
[49] H.W. Sheng, J.H. He, E. Ma, Physical Review B 65 (2002) 184203.
[50] S.Y. Wang, M.J. Kramer, M. Xu, S. Wu, S.G. Hao, D.J. Sordelet, K.M. Ho,
C.Z. Wang, Physical Review B 79 (2009) 144205.
[51] N.P. Lazarev, A.S. Bakai, C. Abromeit, Journal of Non-Crystalline Solids 353
(2007) 3332.
[52] T.K. Gu, J.Y. Qin, X.F. Bian, Applied Physics Letters 91 (2007) 081907.
[53] L.S. Darken, Transactions of the American Institute of Mining and Metallurgical Engineers 175 (1948) 184.
[54] J. Bardeen, Physical Review 76 (1949) 1403.
[55] J.R. Manning, Physical Review 124 (1961) 470.
[56] J.L. Finney, Nature 266 (1977) 309.
[57] H.W. Sheng, Y.Q. Cheng, P.L. Lee, S.D. Shastri, E. Ma, Acta Materialia 56 (2008)
6264.
[58] W.K. Luo, H.W. Sheng, E. Ma, Applied Physics Letters 89 (2006) 131927.

[59] F. Faupel, W. Frank, M.-P. Macht, H. Mehrer, V. Naundorf, K. Ratzke,

H.R. Schober, S.K. Sharma, H. Teichler, Reviews of Modern Physics 75
(2003) 237.
[60] S. Stuber, D. Holland-Moritz, T. Unruh, A. Meyer, Physical Review B 81 (2010)
024204.
[61] A. Meyer, S. Stuber, D. Holland-Moritz, O. Heinen, T. Unruh, Physical Review B
77 (2008) 092201.
[62] H. Zhang, W.Y. Wang, S.L. Shang, Y. Wang, X. Hui, L.Q. Chen, Z.K. Liu,
unpublished.
[63] N. Dupin, I. Ansara, B. Sundman, Calphad-Computer Coupling of Phase
Diagrams and Thermochemistry 25 (2001) 279.
[64] L. Wang, X.F. Bian, J.T. Liu, Physics Letters A 326 (2004) 429.
[65] S.K. Das, J. Horbach, M.M. Koza, S.M. Chatoth, A. Meyer, Applied Physics
Letters 86 (2005) 011918.
[66] X.P. Tang, R. Busch, W.L. Johnson, Y. Wu, Physical Review Letters 81 (1998)
5358.
[67] X.P. Tang, U. Geyer, R. Busch, W.L. Johnson, Y. Wu, Nature 402 (1999) 160.