Sunteți pe pagina 1din 14

E. C.

Aifantis
Laboratory of Mechanics and Materials, Aristotle University of Tliessalonii<i, Ttiessalonil(i 54006, Greece and Center for Mectianics of Materials, Micliigan Tectinological University, Houghton, Mi 49931

Gradient Deformation IVIodels at Nano, IVIicro, and IVIacro Scales


Various deformation models incorporating higher-order gradients are discussed and their implications are considered in a variety ofproblems ranging from the determination of the size of dislocation cores or elastic dislocation interactions to the determination of wavelengths of dislocation patterns or heterogeneous dislocation distributions and the determination of the structure of solid interfaces and of localized strain zones during adiabatic shear deformation. Different scales are involved in each one of these problems: the nanoscale for single dislocations, the microscale for dislocation patterning, and the macroscale for adiabatic shear banding. Accordingly, different gradient models apply for each case, different types of gradient terms are involved and different expressions of the gradient coefficients are assumed.

Introduction

Since the author's suggestion (see, for example, Aifantis, 1982,1983,1984a, 1987 and references quoted therein) to introduce higher-order gradients in the relevant constitutive variables for addressing the problems of dislocation patterns and shear band widths, numerous papers have appeared in the mechanics and the materials science literature (see, for example, Coleman and Hodgdon, 1985; Bazant, 1986; Schreyer and Chen, 1986; Kratochvil and coworkers, 1986, 1993; Kubin and coworkers, 1988, 1993; Belytschko and Lasry, 1988; de Borst and Muhlhaus, 1992; Vardoulakis and Frantziskonis, 1992; Triantafyllidis and Bardenhagen, 1993; Hahner, 1993; Sluys and de Borst, 1994; Tomita, 1994; Zbib, 1994; and references quoted therein) dealing with these and related issues. Among them, the two most popular directions of research developed as an outgrowth of the gradient approach to deformation, were the discrete dislocation simulation modeling of dislocation patterns (e.g. Kubin, 1993; see also Amodeo and Ghoniem, 1988; Groma and Pawley, 1993; Hirth et al., 1996) and the gradient-dependent finite element modeling of shear banding (e.g. Tomita, 1994; see also Sluys, 1992) dispensing with the difficulties of mesh-size dependence and non-convergence in the material softening regime. Detailed considerations of specific aspects of gradient dislocation dynamics and gradient plasticity can be found in various articles by the author and his co-workers (see, for example, Walgraef and Aifantis, 1985; Triantafyllidis and Aifantis, 1986; Zbib and Aifantis, 1988, 1989, 1992; Vardoulakis and Aifantis, 1989, 1991; Muhlhaus and Aifantis, 1991; Ning and Aifantis, 1996; Zhu et a l , 1997), as well as in the author's reviews (Aifantis, 1992, 1994, 1995a, 1996). In this connection, it is noteworthy pointing out the outgrowth of another strain gradient theory of plasticity recently advocated by Fleck et al. (1994, 1997; see also related works by Acharya and Bassani, 1995, 1996 and Steinmann, 1996) which incorporates an asymmetric stress tensor and is motivated by Ashby's description of geometrically necessary dislocations. A variety of problems including strain gradient hardening, size effects in torsion and indentation, as well as crack tips and particulate composites were proposed as suitable candidates to be considered within a gradient plasticity framework. Even though a detailed comparison between the Fleck et al. "asymmetric stress" and the author's original "symmetric stress" gradient plasticity theory is out of the scope

Contributed by the Materials Division for publication in the JOURNAL or ENGINEERING MATERIALS AND TECHNOLOGY. Manuscript received by the Materials Division September 5, 1998; revised manuscript received December 15, 1998. Guest Editors; H. M. Zbib, J. P. Hirth, T. Khraishi, and R. Thom,son.

of the present article, it is emphasized that the original theory is better suited for stability and deformation patterning problems. Nevertheless, apart from the feature of asymmetric stress and the associated difficulty of motivating flow rules, there are many cases where the two theories give similar results. Following the success of the gradient theory in dealing with plastic instabilities at the microscale (dislocation patterning) or macroscale (shear banding) and, in particular, with the determination of spacings and/or widths of dislocation bands and shear zones, the author and his co-workers employed a special theory of gradient elasticity for eliminating the strain singularity from crack tips and dislocation cores (Altan and Aifantis, 1992,1997; linger and Aifantis, 1995; Exadaktylos et al, 1996; Gutkin and Aifantis, 1996). The obtained results provided an alternative derivation of Barenblatt's smooth closure condition for the crack faces, as well as an estimate for dislocation core sizes. The purpose of the present article is to outline the simplest possible structure of gradient theory necessary for describing a variety of spatial deformation features that cannot be captured by classical theories of elasticity or plasticity on one hand, and dislocation dynamics or continuous distributions of dislocations theories on the other. Special attention is given to the form and particular expressions assumed or derived for the gradient coefficients on the basis of the underlying physical configuration and deformation mechanisms. First, in Section 2, the nanoscale is considered and a special form of gradient elasticity is adopted to discuss the strain field of a screw dislocation, the structure of its core, its elastic energy, as well as short-range dislocation interactions. The results are also used to describe the structure of the Peierls-Nabarro dislocation in a crystalline lattice and the crack opening displacement of the mode III crack in an elastic solid. Then, in Section 3, the microscale is considered and several gradient dislocation dynamics models for cyclic and monotonic deformation are discussed. The role and nature of the various phenomenological gradient coefficients, often appearing in the form of diffusion-like coefficients for the various dislocation populations involved, are discussed and specific expressions in terms of the underlying deformation mechanisms are proposed. A similar discussion is also given for the reactionlike terms modeling dislocation production and annihilation and analogous expressions for the related phenomenological coefficients are also provided. The wavelength estimates obtained for the corresponding dislocation patterns on the basis of the aforementioned reaction-diffusion (R-D) models for defects are consistent with those measured experimentally. Finally, in Section 4, the macroscale is considered and a strain gradient theory based on a direct gradient modification of the flow stress is adopted to discuss the localization of deformation for two current problems of mechanics of materials for which higher order APRIL 1999, Vol. 121 / 189

Journal of Engineering Materials and Technology Copyright 1999 by ASME

Downloaded 15 Jan 2008 to 128.125.38.186. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

strain gradients were thought of playing not an important role in previous treatments. The first problem is concerned with solid interfaces and the second with adiabatic shear bands. The significance of gradient terms in determining the structure of the interface and the shear band profile is illustrated. 2 Nanoscale: Gradient Elasticity in Dislocation Cores and CracJs Tips In this section we consider gradient effects at the nanoscale by applying a special theory of gradient elasticity theory to study the structure of dislocation cores and crack tips. In particular, a screw dislocation is considered and a non-singular strain field is derived. The elimination of strain singularity is due to the gradient term and the corresponding gradient coefficient is of the order of the lattice parameter. The derived nonsingular solutions are used to estimate the extent of dislocation core and the nature of short range dislocation interactions, thus providing information which cannot be obtained by using classical elasticity theory. The results are also used to calculate the elastic energy which is found to be finite without introducing an arbitrary cut-off radius. Finally, the derived nonsingular solutions are used to obtain the crack opening displacement for a mode III crack by representing it with a continuous distribution of virtual dislocations. It is shown that, in contrast to the classical elasticity prediction, the crack faces close smoothly with the strain being zero, instead of infinite, at the tip. The special form of the gradient elasticity theory used, reads
a,j = keu^ij + 2ney - cV^CXejAy + 2/ie,j). (2.1)

which contrasts the corresponding profile obtained for the Peierls-Nabarro dislocation on the basis of classical elasticity theory which reads w{x,y^Q) = s i g n (>')( | arctan a/2' The two displacement profiles are plotted in Fig. 1 (right sketch) which indicates that the gradient model gives a narrower dislocation core than the Peierls model. The strain components associated with the gradient elasticity solufion given by (2.2) read (see Gutkin and Aifantis, 1996) 4n _ b_ ^.-^^K, rye \vc . (2.3a)

- K(
r^ nfc

vVc

(2.4)

where (ay, ey) denote the stress and strain tensors respectively, (X., fi) are the Lame constants, V^ denotes the Laplacian and c is the gradient coefficient which, for the present case of crystalline lattice, is taken of the order of the interatomic distance a; specifically vc = a/4. A derivation of (2.1) on the basis of a mixture-type formalism for two superimposed elastic phases can be found in Aifantis (1994) (see also Allan and Aifantis, 1997) where the aforementioned estimate for the gradient coefficient c is obtained (see also Altan and Aifantis, 1992). The solution for the displacement field associated with a screw dislocation was obtained by Gutkin and Aifantis (1996) on the basis of (2.1) as w(x, y) = u(x,y) + b sign (y) C" s sin sx
2TT
-|j'lV(l/c) +

where r denotes the radial coordinate from the dislocation line, with the first term in the bracket representing the singular classical elasticity solution and the second term with the Bessel function Ki representing the gradient elasticity contribution. It is noted that Ki(r/yc) -> (Vc )/r as r - 0 and, thus, the gradient term cancels the elastic singularity as the dislocation line is approached. Plots of the strain components given by (2.4) are provided in the aforementioned publication by Gutkin and Aifantis (1996). These plots show explicitly that the strain singularity disappears at the dislocation line, that a dislocation core may be defined ai r = r^ = 1.25a, and that the strain achieves extreme values at a distance ~ 12% within this core. For the gradient elasticity model described by (2.1) it turns out that the stress components are identical to those singular ones predicted by classical elasticity theory, i.e.. lib sin
2iT r

fib cos 9 2iT r

(2.5)

where (r, B) denote polar coordinates, in terms of which (2.4) can be rewritten as b sin B
Airr

Vc

\Vc (2.6)

/:

ds.
(2.2)

+ s'

b cos 6
A-nr

Vc \Vc

where u{x, y)
2TT

The corresponding gradient elasticity total strain energy defined by the expression
arctan I sign x 2

iy)[l

sign ( x ) ] | .
Jv

(2.2a) is the classical elasticity solution for the same problem. The constant b denotes the Burgers vector and sign y = -I-1 for y can be calculated explicitly as > 0 while sign y = -\ fox y < Q. The situation is depicted in Fig. 1 (left sketch) where the smooth transition exhibited by K\^ (2.8) the gradient solution is contrasted to the discontinuous profile 47r of the classical solution. By utilizing the above formula to consider the structure of with (ro,R) denoting the limits of integration for the cylindrical dislocation core in the Peierls model (e.g., Cottrell, 1953; see domain at hand, and Ko denoting modified Bessel function of also Aifantis, 1985) we obtain the result zero order. It is noted that

'"' 9 " ^{|; - '-\Tc

w{x, y ^ 0) = sign {y)\ - [1 - sign {x)]{ 1 - e

|/V^)|

HI

-In

2Vc

<Z

^^i:-

0,

(2.9)

(2.3) 190 / Vol. 121, APRIL 1999

and, thus, the final expression for E in (2.8) becomes Transactions of the ASME

Downloaded 15 Jan 2008 to 128.125.38.186. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

u u afr/4

(b)

n/b 1 - Gradient model 2 Peierls model

xia

Fig. 1(a)

Fig. K b )

Fig. 1 (a) Sctiematic of screw dislocation and displacement fields u(x, y = 0*) and iv(x, y, 0*) for tlie classical (curves U) and the gradient (curves <l) elasticity, respectively, (b) Dislocation core and comparison with Peierls model.

- In j= 47r 2Vc

(2.10)

47r

X 2

X\ 2

r\

ric

\4c)

ri^lc

\vc/_

(2.11)

leading to the disappearance of the cut-off radius usually assumed for the dislocation energy expression of the classical elasticity solution. Additional results, mainly concerned with dislocation interactions, are included in a forthcoming report (Efremidis et al., 1999) where applications of gradient elasticity models to dislocation and disclination configurations in relation to grain boundary and crack problems are detailed. For the completeness of the present discussion, however, we list briefly below results pertaining to the strain field of two interacting screw dislocations of the same or opposite sign (dislocation dipole). Also, by representing a mode III crack with a continuous distribution of screw dislocations, we derive on the basis of (2.2) or (2.3) an analytic expression for the crack opening displacement which dispenses with the difficulties (strain singularities) of the classical elasticity solution and provides a further substantiation of Barenblatt's smooth closure condition for the crack faces. Thus, in reference to Fig. 2 for a dipole of screw dislocations (top sketch), it can be shown that the strain e is given by the expression

where (r, Vi) denote the radial coordinates corresponding to the two dislocations of opposite sign placed at distance d apart, and (x, X|) are the respective Cartesian coordinates with Xi = x d. Fig. 2 [bottom sketch to the left (a)] gives the distribution of the strain component e^^ for y = 0 in units of b/(4n\lc) for a dipole arm d = 2vc, while a corresponding plot for d = is given in the same figure [bottom sketch to the right (b)]. Sohd fines indicated by the index 1 refer to the gradient solution, while dotted lines indicated by the index 2 refer to the classical elasticity solution exhibiting the well-known singular behavior at the dislocation lines. The results show that the strain is finite at the dislocation lines, approaching a zero value as the dipole arm d increases to infinity (noninteracting dislocations). Moreover, the strain is finite at the central point between the two dislocations, becoming zero when the two dislocations annihilate (rf = 0). It is also interesting to observe from the sketch (b) of Fig. 2 that the gradient and classical solutions become identical at d ^ 5yc s 1.25a, thus providing another independent estimate for the size of the dislocation core. Similarly, for the two parallel screw dislocations of the same sign depicted in Fig. 3 (top sketch), it turns out that the strain is given by

4-IT

4 + i_4,;^, 4 .
njc \ Vc

(2.12)

Fig. 2 Scliematic of a screw dislocation dipole with arm d (top) and strain distribution (bottom) for (a) d = 2^/c and (b) d = 1 0 ^ . Solid lines (1) correspond to the gradient solution and dotted lines (2) correspond to the classical elasticity solution.

The corresponding plots at y = 0 for d = 4vc' and d = vc are also given in Fig. 3 [bottom sketches (a) and (b)]. Again, the strain is measured in units of b/(4TTVc) and the spatial coordinate is normalized by The plots (solid lines indicated by the index 1) show that the strain is finite at both dislocation lines, in contrast to the classical elasticity solution (dotted lines indicated by the index 2) which exhibits singularities there. The value of this strain tends to zero as the dipole arm d increases to infinity (noninteracting dislocations). The level of strain between the two dislocation lines decreases as they glide to each other, becoming zero when the two dislocations coincide (d = 0). These results, as well as the results of the preceding paragraph, show that the short-range elastic interaction between dislocations as derived by the gradient theory is much smaller APRIL 1999, Vol. 121 / 191

Journal of Engineering IVIaterlals and Technology

Downloaded 15 Jan 2008 to 128.125.38.186. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

provides a smooth closure for the crack faces and a vanishing strain at the tip, in contrast to the classical solution which produces abruptly closing crack faces with an infinite value of the strain at the tip. 3 Microscale: Gradient Dislocation Dynamics in Cyclic and Monotonic Deformation Diffusion-like terms in the evolution equations for dislocation densities were first introduced by Aifantis (1981, 1982, 1983, 1984a, 1985) on the basis of an effective momentum balance for dislocation species, for interpreting the occurrence of instabilities in dislocation ensembles and the associated patterning phenomena (see also Bammann and Aifantis, 1982). A specific model for cyclic deformation and the associated persistent slip bands (PSBs) was proposed and analyzed in analogy to the Brusselator model of chemical kinetics by Walgraef and Aifantis (1985), and subsequently adopted and further elaborated upon (among others) by Glazov et al. (1995). Another model was proposed later for monotonic deformation and the associated dislocation clusters by Romanov and Aifantis (1993). Finally, a reaction-diffusion model for dislocation populations was proposed recently by Liosatos et al. (1997) to interpret patterning phenomena of misfit dislocations in thin films. These reaction-diffusion type models for dislocation populations in cyclic and monotonic deformation are reviewed in a unified manner in this section and new results are provided especially in relation to the microscopic interpretation of the gradient and kinetic coefficients entering into the diffusion and reaction like terms. 3.1 Cyclic DeformationThe WA Model. The Walgraef-Aifantis (WA) model for the persistent slip bands (PSBs) occurring during cyclic deformation of copper single crystals distinguishes between immobile and mobile dislocations of densities p, and p, which are assumed to evolve according to the reaction-diffusion equations of the form (e.g., Walgraef and Aifantis, 1985, 1988) dpi d^Pi = A ^ + g(p,)-/(p,,p,), ~dt
-T

(a)

"yz
0.5

i - Ji
0.5.

x/Jc

-to

\
\-0.51

10

x/Jc

Fig. 3 Schematic of two screw dislocations of the same sign at a distance d apart (top) and strain distribution (bottom) for (a) d = A'lc and {b)d = ^. Solid lines (1) correspond to the gradient solution and dotted lines (2) correspond to the classical elasticity solution.

than that resulting from classical elasticity theory. Thus, the elementary processes of nucleation of dislocation dipoles or formation of dislocation pile-ups, which involve interdislocation spacings at the nanoscopic range, are not difficult to occur; this being again in contrast to the predictions of the classical theory of elasticity. Finally, we conclude this section by showing how the gradient solutions (2.2) or (2.3) can be used to obtain the crack opening displacement for a mode III crack extended from / to / as illustrated in Fig. 4 (top sketch to the left). In the same figure (top sketch to the right) it is also shown the representation of the crack in terms of a distribution of virtual dislocations n(x) given by the expression n(x) = (2.13)

^/P

This is, in fact, the same distribution assumed for treating the problem within a classical elasticity theory setting. It is also employed here as the stress fields in both the classical and gradient theories are the same for the model (2.1) and the assumed boundary conditions (see Aifantis, 1992; Altan and Aifantis, 1997). It follows that the crack opening displacement is given by u{x; y = 0) w(x s)n(s)ds '^'ds. (2.14)

'dpm _ 5VD. + dt dx'

(3.1)
f(Pi,p..,),

with (Di, D,) denoting the diffusion-like coefficients for the immobile and mobile dislocations,/(p,, p,) denoting the interaction between mobile and immobile dislocations, and gipt)

(a)

(b)

10

4u,vc J-I

The integral in (2.14) can be calculated explicitly for x\ > giving All Ay. h{-ll{^)e-"^'\ x>l, (2.15)
X < -/,

u{x; y = Q) = <

y^Z-fo

-7,(//^/^)e"^;

and the corresponding plot is also given in Fig. 4 (bottom sketch). In this figure the crack opening displacement is normalized by u and the spatial coordinate by The index 1 is assigned to the gradient solution, while the index 2 is assigned to the classical solution. It is shown that the gradient solution 192 / Vol. 121, APRIL 1999

Fig. 4 Schematic of a mode-Ill crack extended from - / to / [sketch (a) top] and its representation by means of an array of screw dislocations with distribution n{x) [sketch (b) top]. Bottom sketch shows the resulting crack opening displacement and the smooth closure of crack faces for the gradient elasticity solution (curve 1), in contrast to the parabolic profile (curve 2) of the classical elasticity solution and the associated strain singularity at the crack tip.

Transactions of the ASME

Downloaded 15 Jan 2008 to 128.125.38.186. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

denoting the production of immobile dislocations. The interaction t e r m / ( p , , p ) has the form fiPi, Pm) = bpi - yp,pj, (3.1a)

_L
. yd

_L
Yd

while the production term g(pi) g(pi) = ap" api.

in its simplest form reads ?'(p) = fl>0. (3.1b)

4
(a) (b) R^R, d,<p = )|| V i 4> + DsS% <

The coefficient b, which plays the role of the bifurcation parameter for the problem and depends (e.g. exponentially) on the resolved shear stress, represents the freeing rate of immobile dislocations for increasing stress, while the coefficient y represents the pinning rate of mobile dislocations by immobile dipoles. The coefficient a, which for stability purposes needs to be positive, is a constant modeling the linear creation of immobile dislocations, while pf is a constant denoting a reference homogeneous solution for the density of immobile dislocations. The linear stability analysis of (3.1) indicates that a Turing instability occurs at b = b,= (^

Fig. 5 Dipole exchange mechanism, dipole narrowing, and resulting diffusivity due to the shift of dipole's center of gravity.

(3.6)

(3.2)

with a critical wavelength \,. given by q.

(3.3)

where q^ is the critical wave number, c = yp"^, and p',' = blyp'!. Next, we adopt the following estimates for the parameters appearing in (3.3) (Aifantis, 1987; Walgraef and Aifantis, 1988) D, bp,u, iDila ~ li, (3.3a)

2ypf
where v denotes average dislocation velocity, ^ 2.3 x lu " m denotes Burgers vector magnitude, and /, sa 1.6 X 10~^m denotes a mean free path of trapped dislocations of the order of dipole annihilation length. These estimates give a dependence for the wavelength V of the form K = Al{pi (A 10 - 16) which, for typical levels of dislocation densities, gives values for the critical wavelengths very close to those observed experimentally. Another interesting feature of the model (3.1) is its nonlinear behavior beyond the bifurcation point describing transitions between different types of dislocation patterns as the bifurcation parameter (externally imposed stress or strain) increases. This behavior can be studied within a two- or three-dimensional setting (e.g., Walgraef and Aifantis, 1985,1988) by considering the two-time scales near bifurcation associated with the slow modes j , (corresponding to the eigenvalue Wj == 0 ) and the fast modes r, (corresponding to the eigenvalue OJ, < 0 ) . By adiabatically eliminating the fast modes {d,r,i f 0 ) , one obtains the following " s l o w mode d y n a m i c s " equation for the slow modes a in real (as opposed to Fourier) space

where D\\ and D^ are new constants. The phase dynamics modeled by (3.6)2 allows for " s p o n t a n e o u s " translations of the layered pattern to occur, as well as for symmetry breakings in the form of "superdislocations" to take place and persist (for d,4> = 0 ) as in the case of dislocations in an elastic lattice. [In this connection it is noted that, for steady states, (3.6) is of a form similar to that for the displacement field of a screw dislocation in classical elasticity.] One of the most critical issues in reaction-diffusion models for dislocations and other defects is the availability of appropriate expressions for the diffusion-like coefficients and the ratelike constants. In this connection, it is pointed out that the original W A model was proposed without reference to microscopic formulas or derivations pertaining to the various phenomenological coefficients involved. Special concern was expressed, in particular, for the diffusion-like coefficients D, and D as well as for the reaction-like term p,p}. In fact, these three terms are necessary for discussing dislocation patterning and constitute the new elements of the proposed W A model as contrasted, for example, to the existing Gilman-type dislocation kinetics models. [For a review of such type of models see, for example, Bammahn and Aifantis, 1 9 8 2 ] . A derivation of a microscopic expression for the diffusion coefficient D, has already been provided (Aifantis, 1986; see also Walgraef and Aifantis, 1988) along with a mechanical basis for the diffusionlike terms. This expression is listed in (3.3a)i above. For the D, term, the estimate listed in ( 3 . 3 0 ) , was obtained by neglecting the interaction t e r m / ( p / , p,) and, effectively, by disregarding the p,-equation. However, no definite microscopic mechanism was suggested for justifying either the diffusion term Did^pjdx^ or the reaction term yp,pj, an issue taken up in the remaining of this subsection.

First, we provide a justification for the diffusion term D, d^pi I dx^ by identifying p, with the density of immobile dipoles and assuming a "dipole e x c h a n g e " reaction mechanism as shown in Fig. 5 (see also Differt and Essmann, 1993). It is seen that this reaction, i.e., the interaction between a mobile dislocation (p,) and an immobile dipole ( p , ) with height y,, (left sketch of Fig. 5) may lead to a new mobile dislocation and a narrower immobile dipole with height y f (right sketch of Fig. 5 ) . In addition to this process of dipole narrowing under constant total density of mobile and immobile dislocations, each dipole exchange event leads to a random shift of the center of gravity of the dipole (y,i/2; with y,, denoting mean dipole height). A ua\ (3.4) va d,cj = [e -d,(ql + W^y + d,Vl](^ corresponding diffusion coefficient may then be defined for the where e ~ {b - bc)/b,., V^ = Vj^ -I- V^,, denotes the Laplacian, process as D, ~ yl/Stj with t,/ denoting the average time beand (d^, dy, v, u) are constants related explicitly to the model tween two successive events. An estimate of t,i can be obtained parameters previously defined. By considering perturbations by means of the relation 1/f,, i 2pivy,i (r,7' defines the rate of {R, 4>) of the form R = R + R, 4> = 4>c. '^ ^ around an the process) with p, denoting average mobile dislocation denequilibrium reference state (/?, <^) defined by sity and V average dislocation velocity. The final expression for the diffusion coefficient D, is a = 2Ro cos {q^x + 4>,>),

(3.5)

Ra = ve/Bw,

</>o = const..

D,

PmVy,!

(3.7)

with Ro denoting the pattern amplitude and 0 the pattern phase, it turns out that

The above argument leads to both a justification of the diffusion APRIL 1999, Vol. 121 / 193

Journal of Engineering IVIaterials and Technology

Downloaded 15 Jan 2008 to 128.125.38.186. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

term for the immobile dislocations and an expression for the associated diffusion coefficient. Next, we provide a microscopic argument for the justification of the cubic term ypmpl by again identifying pi with the density of the immobile dipoles and considering the coupling of the dynamics with point defect agglomerates of density p. A kinetic equation for the evolution of the point defect agglomerates may initially be assumed of the form
dt
OtoPiPm PoPoP.

(3.8)

with the first term denoting point defect formation by edge dipole disintegration and the second term denoting agglomerate "clean out" by moving dislocations. It may thus be assumed that a dipole stabilization occurs, as a result of the competition between the process of agglomerate formation by mutual annihilation of dipole and mobile dislocations and the process of agglomerate removal by their "sweeping" by moving dislocations. By further assuming that the point defect agglomerate density p may be adiabatically eUminating {d,po '^ 0) due to the short-life time of these defects as compared to time-scales over which the density of dislocations (p,, p,) evolves, we have p = (aj/3^)pi; (3.9)

corresponding parameters {A,B, C) and are generally assumed to obey the relations a < g l , O = < / 0 < l , O s 7 s 2 . The various reaction mechanisms assumed in (3.11) can schematically be represented in Figs. 6{a)-6{d), while Fig. 7 is a schematic representation of dislocation reactions which, however, do not change the overall defect densities but the defect mobilities described by D, and An In fact, one may consider a "hopping" dislocation mechanism, as depicted in Figure 8a for dislocation configurations (positive and negative) between n-th and ( + l)-th jumps. The symbol A denotes the magnitude of dislocation jump (dislocation free path length), while P = PfSx) denotes an associated probability density distribution depicted in Fig. 8(fo). Then, a diffusion equation may be assumed for the total dislocation density p = p^ + p~, with a diffusion coefficient given by the expression D = Ao{v), (3.12)

where AQ denotes a mean free path and {i,') an average dislocation velocity. If a "cross-slip" dislocation mechanism is assumed, as depicted in Fig. 9(a) for an initial dislocation of Surges vector b gliding in the jc-direction with velocity v before its cross-slip and continuing glide on a new plane lying at a distance h apart, the following expression may be assumed for the diffusivity Dy in the y-direction
Dy = Ayiv), A, = ?7 ;

a relation which may be used directly in conjunction with the term PoPiP,,, that now enters into the system of equations describing the (pi, p,)-dynamics, instead of the originally used term pmPf- This combination leads to the substitution
PoPiPm

r-

1+^ + i f ^
h 2\h

(3.13)

PmPt,

(3.10)

i.e., to the appearance again of the necessary for patterning cubic term in an indirect manner. The above discussion providing further justification for the diffusion and reaction terms assumed in the original WA model will be expanded upon in a forthcoming report by Kalaitzidou et al. (1999) where additional microscopic mechanisms are invoked to derive explicit relations for the phenomenological coefficients defining the respective reaction-diffusion terms (see also Zaiser and Aifantis, 1999). 3.2 Monotonic DeformationThe RA Model. The Romanov-Aifantis (RA) three-element kinetics model for monotonic deformation is also of a reaction-diffusion type for the densities of mobile dislocations p, immobile dislocation dipoles pi, and disclinations p^ (Romanov and Aifantis, 1993). For the present discussion we will neglect the coupling with the higher order disclination term p,, and consider the {pm, Pt)dynamics only, with emphasis on the physical interpretation and schematic representation of the various defect reactions involved. The starting point in such a consideration is the following system of reaction-diffusion equations ^ ^ = Ap - Bpl + yCp^Pi + -D. ^ . (3.11) = aAp - Kpi + PBpl - Cp^pi + D i ^ ,

where ho = p,bl2iT{l - V)(T - Tf) and h = J^ hP(h)dh. The constant p denotes as usual shear modulus and v Poisson's ratio, b denotes the Burgers vector magnitude. A., denotes an average distance between cross-slip events in the glide plane, ho is the distance of dislocation immobilization for dipole formation, and P(h) is the probability for the cross-slip height to be h, while (T, T/) stand for the resolved shear and friction stress respectively. The formulae in (3.13) hold for the mobile dislocations p,. For the immobile dislocations p, it turns out that the same formulae hold with rj replaced by (1 77). A different and simpler expression for the cross-slip controlled diffusion coefficient may be derived on the basis of Fig. 9(b). By considering the exchange of mobile dislocations between slip planes we can write
dp, dt

= Uv r

[p,(y) -

P.,iO)]dy

_ Ylvyl aVml
dt
j,=o 3 ay |j,o

(3.14)

leading to a diffusion coefficient for the dislocation density in the J-direction of the form (3.15)

where p is the density of mobile dislocations, p, is the density of immobile dislocations, D, denotes the diffusion coefficient of mobile dislocations, and D, denotes the diffusion coefficient of immobile dislocations. The coefficient A stands for the rate of mobile dislocation multiplication, the coefficient K measures the rate of immobile dislocations disappearance, while the coefficients ( 5 , C) represent the cross-sections of the respective defect reactions (i.e., the interactions between mobile-mobile dislocations and mobile dislocations-immobile dipoles, respectively) and they may be further determined from microscopic models. Finally, the constants (a, p, y) denote fractions of the 194 / Vol. 121, APRIL 1999

'

where 11 denotes the cross-slip probability per unit glide area, 1; denotes average dislocation velocity and yc denotes the maximum cross-slip distance. The calculation implied in (3.14) involves a Taylor expansion up to the second order (the integral with the first spatial derivative vanishes due to symmetry). More details on such derivations with appropriate references and related microscopic expressions for the reaction-diffusion constants will be included in the aforementioned report by Kalaitzidou et al. (1999). By considering, for example, an effecTransactions of the ASME

Downloaded 15 Jan 2008 to 128.125.38.186. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

tive mass balance equation (Aifantis, 1981,1987 and Bammann and Aifantis, 1982) for both positive and negative mobile dislocations

X-v
bypassing [dp,-=0 ; dp,=0]

(a)
V

dislocation multiplication [ dp=Apdt ; dpi=0 ]

->(il) recombination [dp=0 ; dpi = 0 ]

initial mobile dislocation

(")

dipole generation [dp=0 ; dp,=aApdt]

dipole-dislocation interaction initial configuration

(i) + (ii)

=>

dp = A p d t

dpi = a A p d t

X-V L'~*v
change of defect mobilities D ,D|

(b)

,0)
-V

dipole formation [ dp = -pBpJ dt; dpi =PBp'dt ]

[dp = 0 ; dpi = 0 ]

^...-1.
V

interaction between two mobile dislocations

MH) direct annihilation [dp=-(l-P)BpJdt ;dpi=0]

(1) + (ii) + (iii)=>dp = 0 ; dpi = 0


Fig. 7 Schematic representation of dislocation reaction mechanisms for the R-D model of (3.11) resulting into the same densities but different defect mobilities.

(i) + (ii) =>

dp =-BpMt

; dp, = pBpMt

Fig. 68, b Schematic representation of dislocation reaction mechanisms assumed in the R-D model of (3.11). (c)

dt

dpi
dx

= - g

1 , 2

cp

(3.16)

i.
dipole disappearance due to point defects indirect annihilation [dpi = -Kpidt]

(d)

with g accounting for dislocation multiplication, c ' denoting the mean lifetime of mobile dislocations and * = -v"' = v being an average constant dislocation velocity, we can derive corresponding evolution equations for the sum p, = p,t + p^, and the difference 6, = p^ - p,;,. Adiabatic elimination of S, (d,6, f 0) then, results into the expression S, = (v/c)p, which, in conjunction with (3.15), gives dt g - cp, D, (3.17)

dislocation immobilization [ dp 5*0 ; dpi =0 ]

dipole-dislocation interaction initial configuration

dipole dtsassociation [ dp ^0 ; dpi *0 ]

where D, D^/C = vl, (/, = = v/c is the mean free path of mobile dislocations); thus arriving again at the general expressions for the effective dislocation diffusivity given in (3.12) and (3.13)i above. Dislocation interactions of a more general form, leading to effective dislocation diftusivities, may be assumed as a result of long-range defect forces. Dislocation mobility is then determined not only by the applied stress but also by the longrange stress field of the ensemble of all other dislocations. This influence may be accounted for by the random effective stress

* A ; - H * <*i *
. , = &

PA(X)

.^'"..

4_^..>
*A-,*l*
(a)

dipole disappearance [ dp=0 ; dpi 5*0 )

(i) + (ii)

=> dp =YCpPidt

A;(b)

(ii) + (iii) => dpi = - C p p, dt Fig. 6c, d Schematic representation of dislocation reaction mechanisms assumed In the R-D model of (3.11).

Fig. 8 (a) Dislocation hopping mechanism and (b) probability for the associated dislocation jumps.

Journal of Engineering IVIaterials and Technology

APRIL 1999, Vol. 121 / 195

Downloaded 15 Jan 2008 to 128.125.38.186. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

4 Macroscale: Gradient Interfaces and Adiabatic Shear Bands


0 -y<

--

(a)

(b)

Fig. 9

Dislocation cross-slip mechanisms.

fluctuations ST^n which, in turn, lead to a random fluctuation of the dislocation velocity in the glide x-direction. Then, in a frame moving with the average dislocation velocity u, the dislocation may be considered as performing a random walk with a diffusion coefficient
D = - {OV )rcor, = eff) TJ

vL

(3.18)

where {6v^) stands for the amplitude of dislocation velocity fluctuations, (^con, hon = vrcorr) denotc the corresponding correlation time (dislocation velocity fluctuations) and correlation length (effective stress fluctuation) and S = (din v/dr^fs)'^ is the strain rate sensitivity. If dislocation ensembles extending over several grains are considered with positive and negative dislocations gliding with an average velocity v within each grain of an average size d and a random orientation, an effective diffusivity D = (1/4) af(tan^ ip)v may be deduced with v being an average glide velocity and (tan^ </>) denoting a numerical factor resulting from the averaging over all glide planes and grain orientations. We conclude this section with a brief discussion of the steadystate inhomogeneous solutions of (3.11) by assuming also that the coefficients (a, D,) are vanishingly small. The resulting governing nonlinear differential equation is d'{bp)

In this final section we consider the effect of higher-order strain gradients at the macroscale by elaborating on two specific problems: (i) the structure of one-dimensional shear interfaces and (ii) the structure and associated critical conditions of onedimensional adiabatic shear bands. For the first problem it is shown in Section 4.1 that inclusion of the gradient term leads to a smooth transition of the strain across the bimaterial interface, in contrast to the classical theory of elasticity or plasticity giving discontinuous profiles. A comparison between the predictions of the author's "symmetric stress" strain gradient and the Fleck-Hutchinson (1993) "asymmetric stress" strain gradient theory is provided for elastic bimaterials with appropriate interfacial conditions for each theory. On assigning to the interface its own stress and strain field, the two aforementioned solutions can be obtained from the same formula by assuming different expressions for the interfacial shear moduli. By considering a softening type interfacial stress-strain graph and utilizing a previously obtained solution by Aifantis and Serrin (1983) for fluid interfaces, it is further shown by employing the symmetric stress gradient theory that strain localization is possible to occur at the solidsolid interface. This strain localization is illustrated for typical examples of real interfaces with their associated interfacial softening type stress-strain relations determined by atomistic considerations. For the second problem of adiabatic shear banding it is shown that inclusion of the Laplacian term in the constitutive expression for the flow stress, allows for nonvanishing critical wave numbers and associated hardening moduli to be obtained within a linear stability analysis. This is in contrast to standard linear stability analyses for adiabatic shear banding problems which, in the absence of strain gradient effects, predict vanishing wave numbers and hardening moduli, as the length scale introduced by the thermal conductivity is not always sufficient to capture the heterogeneity of plastic deformation. Moreover, if strictly adiabatic conditions are enforced, then the thermal conductivity term should not enter into the governing equations and the only internal length scale remaining is the one introduced by the strain gradient term. Wavelength estimates for adiabatic shear band widths are also provided and the corresponding expressions allow for an explicit comparison between the thermal conductivity and the strain gradient effects. Some of these issues were dealt with previously (Aifantis, 1994, 1995b, Zhu et al., 1995) but a detailed and complete analysis of the coupling effects due to the thermal conductivity and the strain gradient coefficients will be included in a forthcoming paper by Huang etal. (1999). 4.1 Strain Gradients in Interfaces and Tliin Films. Consider a bimaterial interface subjected to a shear stress at infinity as shown in Fig. 11 (top sketch). The corresponding strain field e,j is then given by 0 y y 0 (4.1)

+ bp - ibp,f +

<bpy bp + u

= 0,

(3.19)

where the spatial coordinate x is normalized by y and the various coefficients are redefined by the relations b = Bl A, V = -y/3, u = KB/AC. It turns out that (3.19) possesses periodic solutions as depicted in Fig. 10. The values of the parameters u and D are taken as 15 and 1.5 respectively in these plots and the initial values of mobile dislocation density p, is taken such that p{0) = 5/b and p'm(Q) - 0. Moreover, the mobile dislocation density p, is measured in units of b, while the immobile dislocation density p, is measured in units of yC/ A. The straight lines I and II designate the uniform stationary solutions (D, = D, = 0 ) . The solid lines designated by (1) represent the periodic distribution of the mobile dislocation density p,, while the dotted lines (2) represent the periodic distribution of the immobile dislocation density p,. For typical values of the model parameters, the wavelength \ of the spatial periodicity is estimated as X . ! = 0.1 1 pim.

Pm-Pi

where y denotes the shear strain, while the equilibrium equations ayj = 0 imply
= 0 => T = T

oy

(4.2)

Fig. 10 Stationary mobile (1) and immobile (2) periodic dislocation distributions for the R-D model of (3.11).

where T is the shear stress, y is the space coordinate normal to the interface and r " stands for the applied shear stress at infinity. The assumed gradient-dependent constitutive equation is of the form Transactions of the ASME

196 / Vol. 121, APRIL 1999

Downloaded 15 Jan 2008 to 128.125.38.186. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

1 -

M2 - M l M2

VfJ-lC, V/U1C2 + VM2fl

g-W/iA',.

y Sz 0,

1 Film (Material 1)

A'l - ^ 2 Ml

'JfJ'tCl VM1C2 + VM2C

pyiiiQici'

y < 0, (4.7)

Substrate (Material 2)

Fig. 11(a) Case 1 fci=0 27N, c-)=0.49Nl Case 2 (/|=3nm, /2=35^m)


lE-5

Hl=30GPa, H2=40GPa

^_^^^^^>^

which gives the distribution of y across the bimaterial interface. This solution may be contrasted to the one provided by Fleck and Hutchinson (1993) for the same problem by using their asymmetric stress gradient theory which for the situation at hand is summarized by the relations r = T ' + 2(9m/dy), T" = ny, m = l^fj,(dy/dy) where r denotes the total asymmetric stress, T ' is its symmetric counterpart and m denotes the couple stress with / being an internal material length. The equilibrium equation (4.2) is still valid and, thus, the same form of governing differential equation (with c, in the symmetric stress theory being formally set equal to ^,/?/2 in the couple stress theory) holds for both theories. This is also true for the standard boundary and interfacial conditions (4.5) and (4.6)|. However, the extra interfacial condition (4.6)2 is replaced by the requirement of equal couples at the interface, i.e., mi = OT2 = > l]lJ,,(dy,/dy) = l\fj,2i.dy2ldy) which, in turn, leads to the following strain distribution for the Fleck-Hutchinson theory 1-^^"^'
M2

-IB-S

/^^^
MiA

. g-^^/'.;
g-.Vi//j.

y^O, (4.8)
y ^Q

Ml^l + ^2^2

^
-2E-S

y =
0.02S Fig. 11(b) 0.030

Ml - M2
/U|

/Lt|/i - I - ,1^2/2

heu- atraJn

Fig. 11 (a) Schematic of bimaterlal Interface and (b) strain distribution across the Interface.

T = Kiiy)

CjV^y,

(4.3)

where K is, in general, a nonlinear function of y and c denotes the gradient coefficient, while here and subsequently the index i = 1,2 designates material 1 and 2, respectively. For a linearly elastic bimaterial we have
Kiiy) = fJ-iy,

The strain distribution given by (4.7) and (4.8) are depicted in Fig. 11 (bottom sketch) for typical values of the parameters involved as shown in the diagram. It is noted that the values of c, = 0.27N and c'2 = 0.49^ for the author's symmetric stress theory are chosen such that the internal lengths Vci/jUi and VC2/M2 defined by them to coincide with the internal lengths /, = 3 fim and I2 = 3.5 ^m of the Fleck-Hutchinson asymmetric stress theory for the used values of /ni = 30 GPa and fj,2 = 40 GPa. Next, it is noted that the distributions (4.7) and (4.8) can both be obtained from the same general expression if the problem is reformulated by introducing an interfacial stress T,, an interfacial strain -y; and an interfacial shear modulus /i; defined by the relations
r , = r , = T2 = T , 7 , = y, = -yj,

(4.3a)

T, = M/T/-

(4-9)

where /Ui and /Li2 denote the respective shear moduli. Then, introduction of (4.3a) and (4.2)2 into (4.3) and solution of the resulting one-dimensional differential equation yields 7 = + C,e^^' + Ae"-"^, (4.4)

Then, (4.7) and (4.8) can be written as 7


T

M-

M'

(4.10)

where the constants C, and ZJ, are to be determined from the conditions assumed at infinity and at the interface. The conditions at infinity are assumed to be of the form dy \y\ y
0

with M/ = MiM2(V/V^ + VM2/C2)/(MIVM2/C2 + M2VM1/C1) f"'' the first case, and fi, = fi,ij,2(vjJ-iCi + VM2C2)/(MIVM2C2 + lj,2^lfJ,,Ci) with Ci = fj,ilf/2 for the second case.

(4.5)

while the conditions assumed to hold at the interface are 0 dy, (972 7i = 72, ^ 1 ^ = :ixi ^ dy dy (4.6)

In fact, (4.10) can be obtained as a special case of the general solution of the bimaterial interface problem defined by the nonlinear inelastic constitutive equation (4.3), the conditions (4.5) at infinity, and the general interfacial condition (4.9)2. The solution of this problem can be expressed in terms of the analytical solution obtained for fluid interfaces by Aifantis and Serrin (1983) as y

J'

dy Vf(-y)

F(y)

2 r

(y) -

Ki{yT)]dy, (4.11)

where (4.6)i is a standard interfacial condition, while (4.6)2 is an extra interfacial condition associated with the gradient term. With (4.5) and (4.6), (4.4) becomes Journal of Engineering Materials and Technology

where , (yT) = T " and the values of the strain at the interface y, and at the two outer boundaries of the bimaterial yT and APRIL 1999, Vol. 121 / 197

Downloaded 15 Jan 2008 to 128.125.38.186. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

c,=0.01N, C2=0.001N Hl=30GPa, nj=40GPa tm=10MPa, 0=0.002

Fig. 12(a)
4&-

> 7m for the case (c). These qualitative plots are given for different ratios of T^IT,,, and the values C\ = 0.01 N and c, = 0.001 N respectively, while the values of jii and ^2 are taken to be /^i = 30 GPa and //2 = 40 GPa as before. Finally, it should be pointed out that the solution (4.11) and condition (4.12) can be used, in particular, for discussing the structure and strain distribution of thin films on rigid substrates. Some preUminary results are listed below for this case by utilizing atomistic calculations of Rose et al. (1981, 1984) to motivate the Thorn = K{y) softening curve modeling the "homogeneous' ' portion of the gradient-dependent flow stress T = Thom cV^7 defining the constitutive response of the thin film. Typical graphs for the adhesive energy and the corresponding stress-strain relation are given in Fig. 13, where the corresponding strain profiles for an Al- and Mg-film on a rigid substrate are also provided. These profiles are based on the following softening type graph (4.15) X where the atomistic material parameter a (not to be confused
T|,c

0.8 T 0.9X

)
^ . ^

-4&<l.

Fig. 12(b)
8E-6 -

0.8 x
4B-tf -

i
^ " = ^ = ^ -

0.9 T

^^^"^ /'
4B-6 -

1 1 1 1

-IB-6

Fig. 12(c) Fig. 12 (a) Softening type of interfacial stress-strain graph and corresponding strain distribution for (d) y, < y and (c) y, > y for different T"/Tm ratios.

72 should satisfy the "equal area" or "Maxwell's rule" condition (Aifantis and Serrin, 1983)

7 = Ji

[Kiiy)

- T]dy

= 0,

(4.12)
y(nm) 14

for a nonconvex or softening type graph r = K, ( 7 ) . For a linear elastic bimaterial of the type (4.3o), the solution (4.11) reads 1 (4.13)

12 10

which for an elastic interface, i.e. T, = ij,,yi, is reduced to (4.10) discussed above. For an inelastic interface of the form, for example, r , = T, - a{y, yj^ (4.14)

8 fa 4 2

Alfilm Mgfilm

/
10 20 30 Fig, 13(c) 40 50

where T, denotes the maximum (taken as 10 MPa) of the r 7 graph depicted in Fig. 12 (top sketch) and a is a numerical coefficient (taken as 0.002), the smooth transition profiles depicted in Fig. 12(b) or the strain localization profiles depicted in Fig. 12(c) can be obtained by utilizing the solution (4.11) and the condition y, < y, for the case (fe) or the condition 7, 198 / Vol. 121, APRIL 1999

stralndO")

Fig. 13 (a) Schematic of binding (or adhesion) energies, (b) corresponding stress-strain diagram, and (c) strain distribution in a thin fiim rigid substrate system.

Transactions of the ASME

Downloaded 15 Jan 2008 to 128.125.38.186. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

with the symbol used in (4.14) above) and strain parameter y are defined by a = e\T,/Eo and y = (5 - 5o)/\. In these definitions, e denotes the usual natural logarithmic base and T^ the maximum value of Thom (which is simply denoted by T in Fig. 13(b)), 6 denotes the actual interface separation distance and 5o its equilibrium value, EQ is the equilibrium value of the adhesive energy E = Eo(l H a y ) exp( ay) [it is noted that typical values for Eo are 700 erg/cm^ for Al and 500 erg/cm^ for Mg], and \ denotes the range over which strong interatomic forces act (0.336 A for Al and 1.77 A for Mg). These, rather qualitative results, will be expanded upon for a variety of real bimaterial interfaces and thin film-substrate systems in a forthcoming report by Karagiannis et al. (1999). 4.2 Strain Gradients in Adiabatic Sliear Banding. The final topic to be considered here within a strain gradient framework of macroscopic plasticity is the evolution of adiabatic shear bands. The general view prevailing so far for this problem was that introduction of higher-order gradients is not necessary, since the heat conductivity offers a natural way for capturing spatial features of shear bands such as widths and spacings. An alternative view was expressed by the author (Aifantis, 1992) who argued that it is the heterogeneity of deformation that induces inhomogeneous heating which, in turn, softens the material inducing additional plastic deformation and giving rise to an autocatalytic process leading to instability. Within such a framework, heat conduction can be neglected due to rapid deformations (adiabatic heating) with the internal length scale again provided by the higher-order strain gradients. In this connection, it is pointed out that the energy equation should now be modified to include the extra work done by the higher-order gradients. Thus, the modified equations of continuum thermomechanics which are appropriate for describing adiabatic shear banding are T = K(y, y, 9) cV^y, e = aSI^B + bTy + e, where the thermal diffusivity a, the gradient coefficient c, and the plastic work coefficient b may generally depend on the temperature d, while the extra work term e depends, among other things, on the gradients of y in a way compatible with the second law of thermodynamics (Aifantis, 1984b). The "homogeneous" part of the flow stress Ti,om = '<(y, y, d) may assume various forms with the power-law relation K(y, y, 9) = ij,y"y"'9" being commonly used (/u denotes the thermoviscoplastic strength coefficient, n is the strain hardening exponent, m is the strain rate sensitivity exponent and v is the coefficient of thermal softening). Other types of temperature dependence, for example a linear relationship of the form K ~ (I af^), have also been used. The common feature of all these "temperature softening" relations is that, when introduced to the homogeneous (without the Laplacians) counterparts of (4.16)i and (4.16)2 to derive a stress-strain relation (by eliminating the temperature 9), a. descending portion or softening regime is obtained for advanced stages of straining. The above view has been taken-up more vigorously by Zhu et al. (1995) who argued that for strictly adiabatic conditions, heat conduction could be considered negligible; and, as a result, the length-scale for this instability problem should be provided by the higher-order deformation gradient term rather than the heat conduction term. They verified this claim by numerical analysis which indicated that, for the cases examined, the width of the shear band and the corresponding stress-strain relationship in the post-localization regime are not sensitive to changes of the heat conductivity but depend critically on the gradient coefficient. More recently, the "linearized" and "initially postcritical" stability analysis of the one-dimensional system of equations describing shear banding with the inclusion of higherorder strain gradients, has been considered by the author and Journal of Engineering Materials and Technology (4.16)

///////////////////////
Fig. 14(a)

Fig. 14 (a) Scliematic of simpie shearing and (b) the dispersion diagram (o vs q in normalized coordinates q = qL and Si = ot/y.

his students (Huang etal., 1999; see also Aifantis, 1994, 1995b) and additional results pertaining to critical conditions and preferred wavelength selection, as well as to the competition between the heat conduction and the strain gradient terms have been obtained. To gain further insight into the above problem and related discussion, we consider below the governing one-dimensional differential equations pertaining to the simple shearing of a block made of a thermoviscoplastic material shown in Fig. 14(a). These are
9T

dy

d^ = pi), pc^9 = k -^^ + firy,

dv dy

T = K(y, y, 9) c

. _

d^y dy'

(4.17)

where a superimposed dot stands for material time differentiation identified in the sequel with partial time differentiation, ( T , y) denote shear stress and strain, (p,v) denote material density and velocity, k is the heat conductivity, c is the specific heat, P is the plastic work into heat conversion coefficient, and the rest of the symbols have been explained before. Equations (4.17)i and (4.17)2 express the momentum balance and energy balance respectively, while (4.17)3 is a kinematic compatibility relation and (4.17)4 is the gradient-dependent constitutive equation for the flow stress. It is noted that the extra work term e is neglected and that two length scales, one related to the thermal conductivity coefficient k and another related to the strain gradient coefficient c, are determining the thermomechanical response of the material. Linear stability analysis for the system of equations (4.17) proceeds by assuming a solution of the standard exponential form {y,T,9,v} = {y" + y, T" + f, 9 + i^v" + v]\ (4.18)

{y, f, ~9,v] = {y*, r*, 9*, v*}e''"'*'",

APRIL 1999, Vol. 121 / 199

Downloaded 15 Jan 2008 to 128.125.38.186. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

ciated "convergence" of the stress-strain graph in the softening regime is not decisively influenced by reasonable variations of the thermal conductivity length 4- Both of these difficulties seem to be removed when the strain gradient internal length 4 is included, even in the case where thermal conductivity is neglected (4 = 0). These initial results will be revisited in a forthcoming article by Huang et al. (1999) where a detailed study for the competing effects of 4 and 4 and their influence LO^ + ALO^ + Boj + C = 0, (4.19) on the emergence and evolution of adiabatic shear bands in relation to a pre-existing deformation inhomogeneity will be where A = (l/pcMq\k + c,,s) - P^y], B = ( 1 / presented. P^c^)q\pc^iH + cq^) + q'^ks] and C = (k/p^c^)q'*ih + q^c) On returning to the instability condition (4.22), it may be with y and T evaluated at the uniform time-dependent state. The quantities in (4.19) not defined earlier are the strain harden- argued that a threshold wave number q,h may be defined by the ing modulus h = {dKldy) > 0, the strain rate sensitivity s = relation (dK/dy) > 0, and the thermal softening parameter $ = (OK/ ks y dO) <0, while //denotes the total (under constant y) hardening max \H/ -1(4.24) qth modulus H = h + {P^TIpcj) due to both strain hardening and y pcjj L / v thermal softening. It is also noted that (4.19) can be reached by assuming the form (4.18) only for the fields 6 and v, as the above which, strain fluctuations do not grow as they are supfields y and T are then subsequently determined. pressed by strain gradient and heat conduction effects. For strucThe Ruth-Hurwitz criterion suggests that an instability occurs tural steels with a homogeneous flow stress of the form [Re (w) > 0] when any of the following four conditions are (4,25) Thorn = K(y, t , 0) = ^ly"r'0", satisfied the prerequisite for the instability relation H < 0 implies the A < 0 or B < 0 or C < 0 or A5 - C s 0. (4.20) condition n + v < 0. Also, for typical values of the thermomeHowever, in our case, A and C are always positive. It also chanical parameters (p, = 3.58 X 10" SI, n = 0.015, m = 0.019, turns out that periodic spatio-temporal patterns corresponding V = - 0 . 3 8 ) , (4.19) gives the dispersion diagram shown in Fig. to condition (4.20)3 are possible for the following critical values 14(b) indicating that, independently of the level of strain y, of the total strain hardening coefficient H and wave number fluctuations with short wavelengths (q > qa,) will not grow. The diagram can also be used as a filter for determining the q effect of a pre-existing imperfection on the instability process. In fact, by comparing the Fourier spectrum of an initial strain , k[s(k + cs)qt + <!?(T sy)ql] H = -cq -\ imperfection with the dispersion diagram, we can determine pc^sq 4- pc^y which modes will grow and which will die out depending on (4.21) the magnitude of their wave number in relation to q,^. $ . 2 , ycArk sky - pc^y) y + The competition between the strain gradient coefficient and k + kc^s + pclc the thermal conductivity can clearly be seen by neglecting strain hardening (n = 0) in (4.25). In this case, K(y, y, 9) = In fact, the eigenvalue of (4.19) associated with conditions py"'9" and (4.24) then becomes (4.21) is complex with a non-negative real part but the numerical analysis shows that it becomes quickly real as the shear 1 / mdak pCyOoC I (4.26) strain y increases. It is also noted that the above estimate of positive H for the initiation of instability is in contrast to previous instability analyses which, in the absence of strain gradient effects, always predict vanishing values for H and where 6o and TQ ( = /ito^o) denote the initial values of the <?. Nevertheless, for high strain rates y and typical material temperature and stress entering into the homogeneous solution, parameters, it turns out that | AB | > C > 0 since h > 0 and i while 70 is the imposed strain rate which, for the problem > 0. It follows that the appropriate instability condition is B < considered, remains constant throughout the deformation process. A threshold wavelength 4h (4ii ~ 1/^th) and a correspond0, yielding ing stability condition may then be defined by the relation
( '

where (q, oj) denote the wave number of the fluctuation and its growth rate, a superimposed " o " denotes uniform (timedependent) solution and a superimposed " " denotes the departure of the respective quantity from the uniform state. Moreover, this departure is assumed to be infinitesimal when is normalized with respect to its uniform counterpart. The corresponding dispersion equation reads

// + I c + ) ^'
pc/

0.

(4.22)

4h_
L V

1 (j^mejL^
V\ Tn-Vn

^y,IL'
To

1,

(4.27)

This implies that instability occurs when / / < 0, i.e. in the thermal softening regime. Moreover, two length scales 4 (associated with the strain gradient coefficient c) and 4 (associated with the thermal conductivity k) can be explicitly defined as ks c (4.23) pcH H The preliminary investigations reported by Zhu et al. (1995) suggest that, in general, the observed sizes of adiabatic shear band widths scale with 4 rather than 4 for typical values of the material parameters involved. This was documented by numerical analysis indicating that thermal conductivity alone (in the absence of gradient effects, c = 0) does not account for variations of shear band widths with variations of the internal length 4 and predicts, in fact, shear band widths smaller than those observed. Moreover, the ' 'mesh size dependence'' and the asso/, 200 / Vol. 121, APRIL 1999

where the reference strain 70 ^ pCtfialro and the length of the specimen L were used. It follows from (4.27) that for large strain rates 7 the effect of the first term (A;) in the parenthesis decreases, while the effect of the second term (c) increases. Thus, the strain gradient term prevails and plays more important role than the heat conduction in stabilizing the deformation for high strain rates. Finally, we consider the effect of the strain gradient coefficient on the preferred wave number qpr defined by the conditions (duj/dq) I ,.,^^ = 0 and uj \ ^^^ > 0. In view of the dispersion equation (4.19), we then have 3ckq'^,r + 2[kh + (ks + pcc)u),.]qlr + pcji'-o,,,+ p(k + c,s)u}f,, = 0. This gives the following bounds for a;,,^ Transactions of the ASME (4.28)

Downloaded 15 Jan 2008 to 128.125.38.186. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

J
c=o
"^ 1.5 /

Acknowledgments
~--_^ This work was completed as a part of the US Air Force grant No. F49620-95-1-0208 under the Mechanics and Materials Program directed by Walter Jones, with partial support of the US AFOSR grant No. F4962G-96-1-0478. The support of the EU Commission of European Communities under contracts No. ERB FMRX-CT96-0062 (TMR Research Network) and No. FI4S-CT96-0024 (Revisa) is also acknowledged. Discussions with my co-workers M. Zaiser and M. Gutkin, my graduate students I. Mastorakos and J. Huang and my undergraduate students K. Kalaitzidou and M. Karagiannis contributed to the presentation.

c=1 c=2

as

0.5

2 3

References
A. Acharya and J. L. Bassani, 199.5, "Incompatible lattice deformation and crystal plasticity," N. Ghoniem, ed.. Plastic ami Fracture Imtabilities in Material, AMD-Vol. 200/MD-57, 75-80, ASME, New York. A. Acharya and J. L. Bassani, 1996, "On non-local flow theories that preserve the classical structure of incremental boundary value problems," A. Pineau and A. Zaoui, eds., lUTAM Symp. on Micromechanics of Plasticity and Damage of Multiphase Material.^, Paris, Aug. 29-Scpt. 1, 1995, 3 - 9 , Kluwer, The Netherlands. E. C, Aifantis, 1981, "Elementary physicochetnical degradation processes," A.P.S. Selvadurai, ed.. Mechanics of Structured Media, 301-317, Elsevier, Amsterdam-Oxford-New York. E.C. Aifantis, 1982, "Some thoughts on degrading materials," S.N. Atluri and J.E. Fitzerald (eds) NSF Workshop on Mechanics of Damage and Fracture, Georgia Tech., Atlanta, 1-12. E.C. Aifantis, 1983, "Dislocation kinetics and the formation of deformation bands," G. C. Sih and J. W. Provan (ed.s) Defects, Fracture and Fatigue, Proceedings of International Symposium, May 1982, Mont Gabriel, Canada, 75-84, Martinus-Nijhoff, The Hague. E. C. Aifantis and J. B. Serrin, 1983, "Equilibrium solutions in the mechanical theory of fluid raicrostructures," / Coll. Iiiteif. ScL, 96, 530-547. E. C. Aifantis and J. B. Serrin, 1983, "The mechanical theory of fluid interfaces and Maxwell's rule," J. Coll. Interf ScL. 96, 517-529. E. C. Aifantis, 1984a, "On the microstructural origin of certain inelastic model.s," ASME JOURNAL MATEKIALS AND ENOINEERING TECHNOLOGY, 106 326-

Fig. 15(a)

2.5

V 0:5

1.5'. /

---

k=0 k=54 k=100

0.5

Fig. 15(b) Fig. 15 The effect of gradient coefficient c and heat conductivity k on the preferred wavelength: (a) effect of c (k = 0) and (b) effect of k (c = 0). The coordinates are normalized again as q = qL and 6) = laly.

k + c-s

(4.29)

and the following estimate for the characteristic time of the process /,. mr s (4.30) Hy H which can be used to estimate, in turn, shear band spacings when the rate of the overall deformation process is known. For strictly adiabatic conditions (fc = 0), it can easily be shown from (4.28) that
tc

k + c^s

H +

Icql,

c(4pc - s^)

2pcH + sy-^V cX'^pc - s"-) ) H(pcH + .yy$)


cCiX^pc s^)

(4.31)

It follows that for c = 0, tOp^ = -H/s and qp,. -> TO. These conclusions can also be deduced from the numerical results depicted in the diagrams of Fig. 15. The top diagram shows that (for fe = 0) a preferred wave number is clearly defined for c * 0, in contrast to the case c = 0. The bottom diagram shows that (for c = 0) the effect of thermal conductivity is negligible in selecting a preferred wave number which is, thus, left undetermined. Journal of Engineering Materials and Technology

330. E. C. Aifantis, 1984b, "Remarks on media with microstructures," Int. J. Engng Sci., 22, 961-968. E. C. Aifantis, 1985, "Continuum models for dislocated states and media with microstructures," E.C. Aifantis and J, P. Hirth (eds) The Mechanics of Dislocations, 127-146, ASM, Metals Park. E. C. Aifantis, 1986, "On the dynamical origin of dislocation patterns," Mater. Sci. Engng.. 81, 563-574. E. C. Aifantis, 1987, "The physics of plastic deformation," Int. J. Plasticity. 3,211-247. E. C. Aifantis, 1992, "On the role of gradients on the localization of deformation and fracture," Int. J. Engng. Sci., 30, 1279-1299. E. C. Aifantis, 1994, "Gradient effects at macro, micro and nano scales," J. Mech. Behavior Mats 5, 355-375. E. C. Aifantis, 1995a, "Pattern formation in plasticity," Int. J. Engng. Sci., 33, 2161-2178. E. C. Aifantis, 1995b, "Adiabatic shear banding: Higher-order strain gradient effects," A.M. Rajendran and R. C. Batra (eds) Constitutive Laws: Theory, Experiments and Numerical Implementations, 139-146, CIMNE, Barchelona. E. C. Aifantis, 1996, "Non-linearity, periodicity and patterning in plasticity and fracture," Int. J. Non-Linear Mechanics, 31, 797-809. B. Allan and E. C. Aifantis, 1992, "On the structure of the mode 111 crack-tip in gradient elasticity," Scripta Met Mater., 26, 319-324. B. Allan and E. C. Aifantis, 1997, "On some aspects in the special theory of gradient elasticity," ,/. Mechanical Behavior Mats, 8, 231-282. R. J. Amodeo and N. M. Ghoniem, 1988, "A review of experimental observations and theoretical models of dislocation cells and subgrains," Res. Mechanica 23, 137-160. D. J. Bammann and E. C. Aifantis, 1982, "On a proposal for a contintnim with itiicrostructure," Acta Mechanica, 45, 91-121. Z. P. Bazant, 1986, "Mechanics of distributed cracking," Appl. Mech Rev 39, 675-705. T. Belytschko and D. Lasry, 1988, "Localization limiters in transient problems," Int J. Solids Struct.. 24, 581-597. B. D. Coleman and M. L. Hodgdon, 1985, "On shear bands in ductile materials," Arch. Rat. Mech. Anal., 90, 219-247. A. H. Cottrell, 1953, Dislocations and Plastic Flow in Crystals, Oxford University Press, London. R. de Borst and H.-B. Muhlhaus, 1992, "Gradient-dependent plasticity: Formulation and algorithmic aspects," Int. J. Num. Mech. Ettgng., 29, 1365-1397. K. Differt and U. Essmann, 1993, "Dynamical model of the wall structure in persistent slip bands of fatigued metals: 1. Dynamical model of edge dislocation walls," Mat. Sci. Engng., A164, 295-299. G. Efremidis, T. Akintayo, M. Yu Gutkin and E. C. Aifantis, 1999, "Boundary Value Problems in Gradient Elasticity: Dislocation, Disclination and Crack

APRIL 1999, Vol. 121 / 201

Downloaded 15 Jan 2008 to 128.125.38.186. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

fields," AUT Reports in Mechanics of Materials: AUT/MMl, Aristotle University of Thessaloniki, Thessaloniki, Greece. G. Exadaktylos, I. Vardoulakis and E. C. Aifantis, 1996, "Cracks in gradient elastic bodies with surface energy," Int. J. Fracture, 79, 107-119. N. A. Fleck and J. W. Hutchinson, 1993, "A phenomenological theory for strain gradient effects in plasticity," J. Mech. Phys. Solids, 41, 1825-1857. N. A. Fleck and J. W. Hutchinson, 1997, "Strain gradient plasticity," J. W. Hutchinson and T. W. Wu (eds), Advances in Applied Mechanics, 33, 295-361. N. A. Fleck, G. M. Muller, M. F. Ashby and J. W. Hutchinson, 1994, "Strain gradient plasticity: Theory and experiment," Acta Metall Mater., 42, 475-487. M. Glazov, L. M. Llanes and C. Laird, 1995, "Self-organized dislocation structures (SODS) in fatigued metals," Phys. Stat. Sol., A 149, 297-321. I. Groma and G. S. Pawley, 1993, "Computer simulations of plastic behavior of single crystals," Phil Mag., A67, 1459-1470. M. Yu. Gutkin and E. C. Aifantis, 1996, "Screw dislocation in gradient elasticity," Scripta Materialia, 35, 1353-1358. P. Hahner 1993, "Modelling the spatiotemporal aspects of the Portevin-Le Chatelier effect," Mater. Sci. Engng., A164, 23-34. J. P. Hirth, M. Rhee and H. M. Zbib, 1996, "Modeling of deformation by a 3D simulation of multiple curved dislocations," J. Computer-Aided Mat. Design, 3, 164-166. J. Huang, I. Damtsa and E. C. Aifantis, 1999, "On the competition between thermal conductivity and strain gradient coefficients in adiabatic shear banding," preprint. K. Kalaitzidou, M. Avlonitis, M. Zaiser and E. C. Aifantis, 1999, "ReactionDiffusion Models for Dislocation Dynamics and Plastic Flow," AUT Reports in Mechanics of Materials: AUT/MM2, Aristotie University of Thessaloniki, Thessaloniki, Greece, [see also: A. E. Romanov and E. C. Aifantis, preprint.] M. Karagiannis, I. Mastorakos, J. Ning and E. C. Aifantis, 1999, "Applications of Gradient Theory to Bimaterial Interfaces and Thin Films," AUT Reports in Mechanics of Materials: AUT/MM3, Aristotie University of Thessaloniki, Thessaloniki, Greece. J. Kratochvil, 1993, "On the dynamic origin of dislocation structures in deformed solids," Mater. Sci. Engng., A164, 15-22. J. Kratochvil and S. Libovicky, 1986, "Dipole drift mechanism of early stages of dislocation pattern formation in deformed metal single crystals," Scripta Met, 20, 1625-1630. L. P. Kubin, 1993, "Dislocation patterning," H. Mughrabi (ed) Materials Science and Technology, Eds. R. W. Cahn, P. Haasen, E. J. Kramer 6, Plastic Deformation and Fracture of Materials, 137-190, VCH, Weinheim-New YorkBasel-Cambridge. L. P. Kubin and J. Lepinoux, 1988, "The dynamic organization of dislocation structures," P. O. Kettunen et al., (eds) Strength of Metal and Alloys (ICSMA 8), 1, 35-59, Pergamon Press, Oxford. N. Liosatos, A. E. Romanov, M. Zaiser and E. C. Aifantis, 1998, "Non-local interactions and patterning of misfit dislocations in thin films," Scripta Materialia, 38, 819-826. H. B. Muhlhaus and E. C. Aifantis, 1991, "A variational principle for gradient plasticity," Int J. Solids Struct, 28, 845-857. J. Ning and E. C. Aifantis, 1996, "Anisotropic and inhomogenous deformation of polycrystalline solids," A. S. Krausz and K. Krausz (eds) Unified Constitutive Laws of Plastic Deformation, 319-341, Academic Press, New York.

A. E. Romanov and E. C. Aifantis, 1993, ' 'On the kinetic and diffusional nature of Hnear defects," Scripta Met Mater., 28, 617-622. J. Rose, J. Ferrante and J. Sinith, 1981, "Universal binding energy curves for metals and bimetallic interfaces," Phys. Rev. Lett., 47, 675-678. J. Rose, J. Smith, F. Guinea and J. Ferrante, 1984, "Universal features of the equation of state of metals," Phy.s. Rev., B29, 2963-2969. H. L. Schreyer and Z. Chen, 1986, "One-dimensional softening with localization," ASME Journal of Applied Mechanics, 53, 791-797. L. J. Sluys, 1992, "Wave propagation, localization and dispersion in softening solids," Ph.D. dissertation, TU Delft, Delft, The Netherlands. L. J. Sluys and R. de Borst, 1994, "Dispersive properties of gradient and ratedependent media," Mech. Mater., 18, 131-149. P. Steinmann, 1996, ' 'Views on multiplicative elastoplasticity and the continuum theory of dislocations," Int. J. Engng. Sci., 34, 1717-1735, Y. Tomita, 1994, "Simulations of plastic instabilities in solid mechanics," Appl. Mech. Rev., Al, 171-205. N. Triantafyllidis and E. C. Aifantis, 1986, "A gradient approach to localization of deformation1. Hyperelastic materials," J. Elasticity, 16, 225-238. N. Triantafyllidis and S. Bardenhagen, 1993, "On higher-order gradient continuum theories in 1-D nonlinear elasticity derivation from and comparison to the corresponding discrete models," J. Elasticity, 33, 259-293. D. J. Unger and E. C. Aifantis, 1995, "The asymptotic solution of gradient elasticity for mode III," Int J. Fracture, 71, R27-R32. I. Vardoulakis and E. C. Aifantis, 1989, "Gradient-dependent dilatancy and its implications in shear banding and liquefaction," Ingenieur-Archiv, 59, 197-208. I. Vardoulakis and E. C. Aifantis, 1991, "A gradient flow theory of plasticity for granular materials," Acta Mechanica, 87, 197-217. I. Vardoulakis and G. Frantziskonis, 1992, "Microstructure in kinematic-hardening plasticity," Eur. J. Mech. A/Solids, 11, 467-486. D. Walgraef and E. C. Aifantis, 1985, "Dislocation patterning in fatigued metals as a result of dynamical instabilities," /. Appl. Phys., 58, 688-691. D. Walgraef and E. C. Aifantis, 1988, "Plastic instabilities, dislocation patterns and nonequilibrium phenomena," Res Mechanica, 23, 161-195. M. Zaiser and E. C. Aifantis, 1999, "Materials instabilities and deformation patterning in plasticity," Recent Research Developments in Metallurgical and Materials Sciences, Research Signpost. H. M. Zbib, 1994, "Strain gradients and size effects in nonhomogeneous plastic deformation," Scripta Met Mater., 30, 1223-1226. H. M. Zbib and E. C. Aifantis, 1988, "On the localization and post-locahzation behavior of plastic deformation1, 11, 111," Res Mechanica, 23, 261-305. H. M. Zbib and B.C. Aifantis, 1989, "A gradient-dependent flow theory of plasticity: Application to metal and soil instabilities," Appl. Mech. Rev., 42, 2 9 5 304. H. M. Zbib and E. C. Aifantis, 1992, "On the gradient-dependent theory of plasticity and shear banding," Acta Mechanica, 92, 209-225. H. Zhu, H. M, Zbib and E. C. Aifantis, 1995, "On the role of strain gradients in adiabatic shear banding," Acta Mechanica, 111, 111-124. H. T. Zhu, H. M. Zbib, and E. C. Aifantis, 1997, "Strain gradients and continuum modeling of size effect in metal matrix composites," Acta Mechanica, 121, 165-176.

202 / Vol. 121, APRIL 1999

Transactions of the ASME

Downloaded 15 Jan 2008 to 128.125.38.186. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

S-ar putea să vă placă și