You are on page 1of 19

Journal of the Mechanics and Physics of Solids

52 (2004) 1507 – 1525


www.elsevier.com/locate/jmps

A stress-gradient based criterion for dislocation


nucleation in crystals
Ronald E. Millera , Amit Acharyab;∗
a Department of Mechanical and Aerospace Engineering, Carleton University, 1125 Colonel By Drive,
Ottawa, Canada ON K1S 5B6
b Department of Civil and Environmental Engineering, Carnegie Mellon University,

Pittsburgh, PA 15213, USA

Received 16 July 2003; received in revised form 19 January 2004; accepted 21 January 2004

Abstract
Dislocations are the main lattice defects responsible for the strength and ductility of crys-
talline solids. The generation of new dislocations is an essential aspect of crystal defect physics,
but a fundamental understanding of the mechanical conditions which lead to dislocation nucle-
ation has remained elusive. Here, we present a nucleation criterion motivated from continuum
thermomechanical considerations of a crystalline solid undergoing deformation, and demonstrate
the criterion’s ability to correctly predict dislocation nucleation via direct atomistic simulations.
We further demonstrate that the commonly held notion of a nucleation criterion based on the
magnitude of local stress components is incorrect.
? 2004 Elsevier Ltd. All rights reserved.

Keywords: Dislocation nucleation; Dislocation mechanics; Atomistic simulation

1. Introduction

It has long been understood that dislocations—line defects in the arrangement of


atoms in crystals—are the primary defects responsible for the strength and ductility
of crystalline solids. A material in which dislocations are free to move and multiply
will generally be soft and malleable, whereas a material in which existing defects
cannot move and new dislocations cannot easily be created will be hard and brittle.
Understanding and controlling these aspects of crystalline materials has long been the

∗ Corresponding author. Tel.: +1-412-268-4566; fax: +1-412-268-7813.


E-mail address: acharyaamit@cmu.edu (A. Acharya).

0022-5096/$ - see front matter ? 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jmps.2004.01.007
1508 R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525

goal of materials scientists, seeking always to improve the mechanical behaviour of


engineering materials.
A fundamental question in crystal plasticity is that of a criterion for defect nucleation.
This is important, for instance, in brittle versus ductile behaviour, where the ease of
dislocation nucleation from a crack tip and subsequent dislocation mobility determine
the macroscopic material response. The criterion for dislocation nucleation is also of
fundamental importance in certain special circumstances such as in the nano-indentation
of initially defect free crystalline regions. Nano-indentation simulations similar to those
used in this work commonly reveal homogeneous defect nucleation below the indented
surface (e.g. Li et al., 2002; Zimmerman et al., 2001; De la Fuente et al., 2002;
Kelchner et al., 1998). In this case, the conditions leading to homogeneous nucleation
are of interest. Other molecular dynamics simulations reveal the details of heteroge-
neous defect nucleation and interactions in relatively complex structures, for example
in nano-crystalline materials (Yamakov et al., 2001). Here as well, it is desirable to
know the mechanical conditions for nucleation. Finally, the complex processes of dis-
location multiplication and interactions with other defects often exhibit the generation
of new defects as plastic Gow continues (Wirth et al., 2001; De Koning et al., 2003),
and as such provide a rigorous proving ground for criteria aiming to predict defect
nucleation.
On a practical note, recent advances in computational techniques have led to a branch
of computational materials science dedicated to understanding the complex interactions
between individual dislocations that give rise to macroscopic material behaviour. Com-
puter simulation techniques (Kubin and Canova, 1992; Van der Giessen and Needle-
man, 1995; Zbib et al., 1998) model the deformation of crystalline solids by explicitly
treating each dislocation in the material. Such an approach requires a set of parameters
that describe the fundamental behaviour of individual dislocations. SpeciJcally, these
techniques require such things as the mechanical conditions (i.e. the internal stress
state in the deforming material) necessary for an existing dislocation to move, grow
and multiply (as in Frank-Read sources), and the conditions for the spontaneous gen-
eration of new dislocations. Without the guidance of a fundamental understanding of
these underlying atomic scale processes, such modelling techniques must rely on ad
hoc and possibly incorrect assumptions.
The motion of dislocations in FCC crystals is often observed to obey Schmid’s Law,
whereby the dislocation will move when the resolved stress acting on it reaches a criti-
cal, material speciJc value. Other crystal structures often reveal so-called “non-Schmid”
dislocation behaviour (Duesbery and Vitek, 1998). For example, the motion of disloca-
tions in BCC metals exhibits a dependence on stress components other than the resolved
shear stress, and on the sense (±) of the resolved shear. These non-Schmid eNects
can largely be explained based on the details of the BCC dislocation core structures
(Ito and Vitek, 1999).
However, an elusive point in the understanding of dislocation mechanics has been
a satisfactory description of the mechanical conditions that give rise to dislocation
nucleation. In the absence of any compelling evidence to do otherwise, some (e.g.,
Gouldstone et al., 2001) have adopted a natural extension of the Schmid law, whereby
it has been assumed that a critical, material-speciJc stress level must be reached at
R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525 1509

a point in a crystal for nucleation to occur. The work of Khantha et al. (1994a,b)
and Sun et al. (2002) on the cooperative nucleation of dislocation loops is based on
statistical mechanical considerations geared towards an understanding of the brittle to
ductile transition. These analyses do not develop a criterion for homogeneous nucleation
and its location. Rice’s (1992) analysis of dislocation nucleation from a crack tip led
to a nucleation condition based on the so-called unstable stacking fault energy of
crystalline slip. This criterion has served as a basis for several studies of dislocation
nucleation from a crack (Sun et al., 1993; Rice and Beltz, 1994; Gumbsch and Beltz,
1995; Xu et al., 1995). The Rice criterion has been extended by other authors to
nucleation from the corner of a knife-edged indenter (Shenoy et al., 2000), and also to
predict twinning near a crack tip (Tadmor and Hai, 2002), but it is not straightforward
to extend this criterion to a general state of loading or to instances of homogeneous
nucleation away from singular points like crack tips or indenter corners. More recently,
others (Li et al., 2002; Van Vliet et al., 2002), have developed a nucleation criterion
based on instability analysis and examined within the framework of a Cauchy–Born
approximation to crystal hyperelasticity.
Here, we present a criterion for dislocation nucleation developed from continuum
thermomechanical considerations that is naturally phrased in terms of stress gradients
(speciJcally the curl of the stress tensor). We then show, through atomistic simula-
tion of crystal deformation, that the criterion accurately predicts dislocation nucleation,
whereas a Schmid-like stress-based criterion clearly does not. Next we discuss, as far as
possible at this stage, similarities and diNerences between our criterion and that of Rice
(1992) and Li et al. (2002). Our results and discussion lead to, and quantify, a new
material property that appears to govern dislocation nucleation. This development sheds
light on a fundamental aspect of material behaviour: the mechanics of dislocation
nucleation.

2. “Schmid-like” criterion for nucleation

The Schmid law for dislocation motion states that a dislocation will become mobile
when the resolved shear stress, r , on a dislocation reaches a critical resolved shear
stress, crss , that is a property of the material
r ¿ crss ; (1)
where
r = s · T · n; (2)
where s is the slip direction, T is the stress tensor and n is the slip plane normal.
Thus, to extend this idea to a criterion for the nucleation of a dislocation, one can
assume that a critical resolved shear stress exists for defect nucleation. A Schmid-like
nucleation criterion will then predict nucleation of a speciJc dislocation (deJned by its
Burgers vector and slip plane normal) at the point in the crystal where r Jrst reaches
crss . In what follows, we will show that such a criterion is not an accurate predictor
of dislocation nucleation.
1510 R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525

3. Driving force for nucleation

Here, we outline the salient features of a mathematically well-set, nonequilibrium


dislocation Jeld theory that has recently been developed (Acharya, 2001, 2003, 2004),
building on the classical elastic theory of continuously distributed dislocations (KrRoner,
1981; Mura, 1963; Willis, 1967; Kosevich, 1979). The goal of this section is to show
the plausibility of a stress-gradient based nucleation criterion as predicated on this the-
ory. Complete details are provided in Acharya (2001, 2003, 2004). The Jeld equations
are
curl U p = −; (3)

Up = Ũ p ; (4)

T = T̂ (” − ”p ); ” := 12 (U + U T ); ”p := 12 (U p + U pT ); (5)

div T = 0; (6)

˙ = −curl ( × V ) + s; (7)

Ũ˙ p =  × V : (8)
In the above, U is the displacement gradient, Ũ p and U p are measures of ‘small’,
slip, and plastic distortion, respectively (second-order tensors). T̂ is a generally non-
linear function of the elastic strain describing the (symmetric) stress tensor T.  the
dislocation density tensor, and V is the dislocation velocity vector. s represents the
dislocation source (nucleation rate) term. A superposed dot represents a time deriva-
tive. In terms of the components, with respect to a rectangular Cartesian frame, of a
second-order tensor Jeld A, and a vector Jeld a,
(A × a)ij = ejrs Air as ;
(9)
{curl A}ir = erjk Aik; j ;
where erjk is a component of the third-order alternating tensor X .
The theoretical structure relies on the mathematical fact (akin to the Stokes–
Helmholtz resolution of a vector Jeld) that any square-integrable second-order ten-
sor Jeld A on the body R can be written uniquely as a sum, A = A⊥ + A , of A⊥ ,
a curl of a tensor Jeld (that may be thought of as a gradient of a vector Jeld for
practical purposes) that vanishes on the boundary, and A , another tensor Jeld that is
orthogonal to A⊥ , i.e. R A⊥ : A dV = 0. In general, the  and ⊥ Jelds corresponding
to two diNerent tensor Jelds are orthogonal.
The Jeld equations have the following meaning. The incompatible part of the plastic
distortion tensor is determined by the dislocation density tensor (3). The compatible
part of the plastic distortion is given by the compatible part of the slip distortion
(4). Eqs. (5) and (6) are the standard stress constitutive equation and equilibrium
equation for an elastic–plastic material. Eq. (7) is the statement of evolution for the
dislocation density tensor. It is developed by localizing an integral balance law for the
R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525 1511

burgers vector content of dislocations that thread an arbitrary material surface in the
body (Acharya, 2001). In this equation, the Jrst term on the right-hand side comes
from essentially kinematic considerations regarding the Gux of Burgers vector into the
surface carried by dislocation lines. The second term corresponds to generation of new
dislocations and is of primary importance here. The dislocation velocity vector at a
given point is understood as the instantaneous velocity of the inJnitesimal dislocation
type at that point. Finally, Eq. (8) represents the evolution of the slip distortion in the
body due to dislocation motion.
The constitutive inputs required by the theory are speciJcations of the crystal elas-
ticity, dislocation velocity and dislocation nucleation rate. Crystal elasticity is generally
well characterized; as an aid to the deJnition of the kinetic statements describing the
dislocation velocity vector and nucleation rate tensor, the mechanical dissipation im-
plied by the theory can be analysed to identify the corresponding ‘driving forces’.
These driving forces (not necessarily with physical units of force) are the energetic
conjugate Jelds to dislocation motion and nucleation within the theory that determine
the dissipated mechanical power per unit volume of material. Once identiJed, it is nat-
ural to pose the kinetics for motion and nucleation as functions of their corresponding
driving forces. This program is carried out by Acharya (2003) for the geometrically
linear theory and in Acharya (2004) for the theory of unrestricted nonlinearity. Here,
we outline the rationale behind our approach based on thermodynamics of irreversible
processes.
In any general irreversible process of a body, the Jrst law of thermodyamics asserts
that the rate of change of kinetic energy plus the rate of change of internal energy is
equal to the sum of the rate of working of the tractions applied to the body on its
exterior surface, the rate of working of the body forces acting on the body, and the
heat entering into the body through its external surface (we assume, for simplicity,
no volumetric heat supplies). When expressed mathematically, the rate of change of
kinetic energy gets eliminated in this balance statement with the familiar result
  
˙ dv = T : U̇ dv + q · n da; (10)
R R @R

where R is the region of space occupied by the body,  is the speciJc internal energy
per unit volume of the material, q is the heat Gux Jeld into the material through the
external surface of the body @R, and n is the outward unit normal to the external
surface of the body. The Jrst integral on the right-hand side of Eq. (10) represents the
‘stress-power’ term. If we now introduce the speciJc free energy per unit volume Jeld
through the relation
=  − ; (11)
where  is the absolute temperature and  is the speciJc entropy per unit volume
Jeld and consider isothermal conditions at a temperature 0 , then Eq. (10) may be
rewritten as
  
1 1 ˙ dv ;
Ṡ − Q̇ = T : U̇ dv −
0 0 R R
1512 R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525

S :=  dv
R
   (12)
Q := q · n da dt:
@R

The second law of thermodynamics states that the rate of entropy production in the
body characterized by the left-hand side of Eq. (12)1 is nonnegative in any process
of the body. With this in mind, we think of the right-hand side of Eq. (12)1 , i.e.
essentially the rate of working of the stresses minus the rate of change of the free
energy, as a measure of the dissipation in the body. Notice that Q̇ in Eq. (12) can
be negative (heat being emitted by the body) as is usual in plastic deformation, thus
suggesting that the mechanical dissipation is converted to a change in entropy of the
body and into heat produced by the body. Of course, for the discussion of dissipation
under isothermal conditions for the whole body, the heat Gux has to be treated as
a constitutively undetermined constraint reaction (Lagrange multiplier) Jeld with the
global heat input/output determined from balance of energy (12)1 .
In the theory being considered, the work of the stresses in the body can be stored as
free energy, and dissipated in plastic deformation due to slip produced by dislocation
motion (nonzero V ) and in nucleating dislocations (nonzero s). One mechanism of
dissipation of energy by the motion of a dislocation is understood as the energy lost
in setting oN lattice waves and vibrations induced due to generally nonuniform motion
of the spatial conJguration of atoms consisting the dislocation. A similar physical
interpretation may be associated with dissipation related to nucleation, as dislocation
nucleation also entails a sudden change in atomic conJguration at the nucleation site. It
seems reasonable to assume, from the sudden load drops that accompany a nucleation
event, that lattice waves would have to be set oN for at least a short instant during
a nucleation event, and it is the energy taken up by such waves that we think of as
dissipation due to nucleation along with an increase in entropy of the body due to
disorder produced by the creation of a dislocation in a lattice.
In order to analyse this partitioning of the stress power into stored and dissipated
parts, we examine the global (mechanical) dissipation in the theory,
 
D := T : U̇ dv − ˙ dv: (13)
R R
The free energy, like the stress, is assumed to depend only on the elastic strain, ”e =
” − ”p ,
= ˆ (” − ”p ): (14)
Consequently,
   
@ˆ @ˆ p
D= T − e : ”˙ dv + e
: ”˙ dv: (15)
R @” R @”

We assume hyperelasticity of the material and adopt



T̂ = e : (16)
@”
R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525 1513

The dissipation now takes the form


 
@ˆ p
D= e
: ”
˙ dv = T : ”˙p dv: (17)
R @” R

Then
 
p
D= T : ”˙ dv = (T + T⊥ ) : U̇ p dv
R R
 
= T : Ũ˙ p dv + curl WT : U̇ p dv; (18)
R R

where Eq. (4) and the orthogonality property of the  and ⊥ decompostion has been
used. In Eq. (18), WT is that Jeld with zero boundary conditions whose curl is T⊥ .
It is uniquely determined through the relations curl WT = T⊥ and div WT = 0.
Now, using Eqs. (3), (7), and (8)
  
p p
curl WT : U̇ dv = WT : curl U̇ dv = WT : (−)
˙ dv
R R R

= WT : (curl Ũ˙ p − s) dv
R
 
= T⊥ : Ũ˙ p⊥ dv − WT : s dv; (19)
R R

so that Eqs. (18) and (19) together imply


   
D= T : Ũ˙ p dv − WT : s dv = T :  × V dv − WT : s dv: (20)
R R R R

Thus, Eqs. (13) and (20) give the required partitioning of mechanical power as
   
T : U̇ dv = ˙ dv +  · V dv + −WT : s dv;
R R R R (21)
 := X (T); i = eijk Tjr rk :

As can be seen from Eq. (21), the work-conjugate Jeld to the dislocation motion is 
and that to the nucleation rate is −WT . We think of these Jelds as the ‘driving forces’
for V and s in the sense that, surely, in their absence these dissipative mechanisms
would not be able to dissipate any power. Indeed,  may be viewed as the Jeld analog
of the Peach–Koehler (P–K) force, as it can be shown to share many of the properties
of the P–K force (Acharya, 2003). Following the conventional wisdom of having the
dislocation velocity depend upon its driving force as in dislocation theory and discrete
dislocation plasticity, it would be appropriate to have V dependent on  in the present
Jeld-theoretic setting. Similarly, since −WT is the driving force for s, we will assume
that the nucleation criterion should be dependent on this quantity. It is to be noted here
that the concept of examining the form of the dissipation function in order to identify
1514 R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525

driving forces for dissipative mechanisms is standard in the thermomechanics of con-


tinua and has resulted in many meaningful conclusions in diNerent physical theories—
e.g. Rice (1971) shows that, granted the assumptions of single crystal plasticity theory
related to slip on speciJc slip systems and the free energy depending only on the local
value of the elastic distortion (Rice considers a more general situation with respect to
the latter assumption), the driving force for the slip rate on a slip system is the Schmid
stress on that system.
Based on the above line of reasoning, let a dislocation density of  Burgers vector
per unit area (perpendicular to a direction l) per unit time be nucleated at a point with
character m ⊗ l, where m is the unit vector in the direction of the Burgers vector and
l is the unit line direction. Then, s at that point may be written as
s =  m ⊗ l: (22)
Inserting this form in the last integral in Eq. (21) we obtain
 
−WT : s dv = (−m · WT · l) dv (23)
R R
and hence conclude that the driving force for nucleation of the density  is
− m · WT · l: (24)
Since the determination of WT from the stress Jeld requires the solution of a well-
deJned boundary value problem with T as data, it is clear that it is genuinely nonlocal
(integral nonlocality) in the stress Jeld. While WT can be determined from T with
relative ease, as a criterion for nucleation we felt it to be prudent to seek a simpler
construct, if possible. With some approximation this is obtained as follows:
The fact that the Jeld WT satisJes
curl WT = T⊥ (25)
by deJnition implies
curl(curl WT ) = curl T⊥ = curl T: (26)
This is so because curl T vanishes as T is a gradient of a vector Jeld. But
curl(curl WT ) = grad(div WT ) − ∇2 WT ;
(27)
eijk ekmn (WT )rn; mj = (WT )rj; ji − (WT )ri; jj
and since div WT = 0 by deJnition, we have from Eqs. (26) and (27)
∇2 WT = −curl T: (28)
We now deJne a tensor Jeld as

1 −curl T(x )
A(x) := − dVx ; (29)
4 R |x − x |
where |x − x | is the distance between the two points x; x . It is a well-known result
of potential theory (e.g. Carlson, 1967) that the Jeld A then satisJes
∇2 A = −curl T: (30)
R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525 1515

DeJning a tensor Jeld B that takes on the value of the Jeld −A on the boundary
of the body and satisJes ∇2 B = 0 in R, we note from Eqs. (28) and (30), and the
fact that WT vanishes on the boundary of the body by deJnition, that
∇2 (WT − {A + B}) = 0 on R
(31)
WT − {A + B} = 0 on @R:
Consequently, by the uniqueness of solutions to the Dirichlet problem for Laplace’s
equation we conclude that
WT = A + B (32)
Expanding the numerator of the integrand in the expression for A (29) about the point
x, we obtain as a Jrst approximation
WT (x) ≈ k(x)curl T(x) + · · · (33)
where k(x) depends on the geometry of the body. Based on Eqs. (24) and (33) we
assume that dislocation nucleation of a particular character depends upon the quantity,
Nm; l , deJned as
Nm; l = |m · curl T · l| = |mi erjk Tik ;j lr | (34)
and consider its justiJcation to be its ability to predict homogeneous nucleation events
in atomistic assemblies as we present in the following sections of this work.
Nm; l may be mechanically interpreted as essentially an areal intensity of force per
unit length corresponding to the sum of the force per unit length Jeld m · T · d x along
a circuit C encircling an inJnitesimal area A perpendicular to the line direction, i.e.
assuming m to be uniform in A
 
m·T ·dx= m · curl T · l dA: (35)
C A

Interestingly, for a dislocation with Burgers vector in the slip direction (i.e., the
close-packed direction), the stress-state dependence of the driving force  for dis-
location velocity is given exactly by the Schmid stress. As such, our theory seems
to naturally diNerentiate between the physical causes for dislocation motion and nu-
cleation. Given that dislocation nucleation refers to the breaking and reformation of
several atomic bonds along a slip plane that may be idealized as slipping of a sizeable
portion of a block of atoms over another, while dislocation motion refers to the ex-
tension of a slipped block through the advance of its boundary requiring the breaking
and reformation of, ideally speaking, a single bond, it is not surprising that the driving
force for nucleation should be diNerent from that of motion. Granted the above phys-
ical picture, it also seems natural that a criterion for dislocation nucleation should be
nonlocal and that for motion should be local.
The foregoing analysis does not strictly apply to a model whose elastic response
requires a consideration of Jnite deformation eNects, as in the case of an atomistic
assembly under the Cauchy–Born hypothesis operating with the macroscopic elastic
distortion F e (formally, nonlinearity in elastic response is considered above). However,
the essential nature of the thermodynamic argument does not change in an analysis of
1516 R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525

the theory under unrestricted nonlinearity (Acharya, 2004). Considering the above result
as motivation, the nucleation criterion that we propose for a rate-independent setting
then, is as follows. We assume that a critical value, Ncrit , exists for the speciJc material.
The criterion will predict nucleation of a speciJc dislocation (deJned by its Burgers
vector and line direction) at the point in the crystal where Nm; l for that dislocation
type reaches Ncrit , i.e.,
Nm; l ¿ Ncrit : (36)
For example, in FCC crystals, the nucleation of a Shockley partial dislocation on the
{1 1 1} family of slip planes would seem to be associated with the value of Ncrit based
on our observations of the dominant FCC slip behaviour.
To test whether this criterion accurately predicts dislocation nucleation it is necessary
to evaluate Nm; l at each point in the crystal for each possible combination of line
direction and Burgers vector. The largest value of Nm; l at each point in the crystal must
be compared to the corresponding value at all other points. Ultimately, the location and
magnitude of the largest of all possible Nm; l values will identify the predicted nucleation
site and dislocation type. As well, the value of Nm; l just prior to nucleation will quantify
the material property Ncrit for this material. Note that, in theory, the criterion should be
maximized over all possible combinations of Burgers vectors and line directions, i.e.
over the inJnite possible directions in space irrespective of the known underlying crystal
lattice. In practice however, it is computationally more tractable to consider only the
known likely candidates, for example all possible Shockley partials in an FCC lattice.
We emphasize that the theory leading to the criterion is not restricted in the way, only
the eWcient evaluation of the criterion is. Even within this restricted evaluation of the
criterion, we will show in what follows that the method is still predictive, as it can
accurately determine which of all the dislocations explicitly considered will in fact be
nucleated.
In the simulations which follow, we consider a special case in which the dislocation
line direction is conJned to be out of the plane of the simulation, which is 2D. This
simpliJes the calculation of Nm; l because the line direction is known, but we emphasize
that extension to 3D will introduce no signiJcant diWculties. The only added complexity
in 3D is that, because the dislocation line direction can lie anywhere in the slip plane,
maximization of the criterion over this one additional degree of freedom must be
performed at each point.

4. Atomistic simulation technique

To test whether the criterion of Eq. (36) is an accurate predictor of dislocation


nucleation, we have performed a number of atomistic simulations of the deformation
of perfect crystals. Simulations were performed using a recently developed technique
(Shilkrot et al., 2002) called coupled atomistics and discrete dislocations (CADD)
whereby a fully atomistic region, modelled by empirical interatomic potentials based
on the embedded atom method (EAM) (Daw and Baskes, 1984), is coupled to a
larger linear elastic region. However, because we are interested only in the initial
nucleation of dislocations, these calculations could have also been carried out using a
R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525 1517

straightforward and well-established lattice statics approach. The eNect of the CADD
formalism is only to facilitate more realistic boundary conditions and the automatic
detection and identiJcation of nucleated dislocations.
All simulations are two dimensional, but with a fully three-dimensional underlying
crystal lattice. This was achieved by employing periodic boundary conditions in the
z (out-of-plane) direction and a minimal z-dimension of the periodic cell to correctly
produce the crystal structure. This technique allows displacements in all three directions,
but variation of the displacement only in the x–y plane, and as a result permits the
nucleation of dislocations of general character (mixed edge and screw), but conJnes the
dislocation line to be in the z direction. Two crystal structures were considered: First,
a model hexagonal Al lattice with z = [0 0 0 1], modelled using the EAM potentials
due to Ercolessi and Adams (1994). This crystal, while not the equilibrium crystal
structure of Al, can be kept stable by artiJcially restricting the deformation to the x–
y plane. The hexagonal Al in this orientation is therefore convenient because its only
active slip systems are pure edge dislocations. Second, an FCC crystal of Ni oriented
with x = [1 1 1], y = [1Z 1 0] and z = [1Z 1Z 2] and modelled using the EAM potentials of
Foiles et al. (1986) was considered.
The modelling was carried out by building an initially perfect single crystal, and
then incrementally applying loads through Jxed displacement boundary conditions. The
atomic degrees of freedom in the crystal were then relaxed to their minimum potential
energy conJguration using the conjugate gradient (CG) minimization technique. After
some number of applied load steps, dislocations nucleated during the CG minimization
and the simulation was stopped, several intermediate atomic conJgurations between the
last (defect free) converged load step and the nucleation of the Jrst dislocation having
been stored during the process. These intermediate conJgurations were then analysed
in further detail. Because the CG minimization algorithm can, in some cases, take
a trajectory to the equilibrium conJguration that deviates from the physical one, we
have also performed one of the simulations (the indentation of hexagonal Al) using a
steepest descent (SD) algorithm. The diNerences between the CG and SD results were
not substantial and do not aNect the conclusions of the work.
Each intermediate conJguration leading up to the nucleation of a defect was used
to compute the atomic level stress components acting at each atomic position in the
region of interest. We make use of the atomic level stress formulation based on the
virial theorem, as Jrst described by Born and Huang (1954). For the embedded atom
method interatomic potentials, these stresses can be written in closed form on an atom-
by-atom basis, as described in Vitek and Egami (1987). Next, a Jnite element mesh
was deJned between all of the atoms, to allow for eWcient and reasonably accurate
calculation of the stress gradients between Jrst-neighbour atoms in the crystal. Linear,
three-noded triangular elements and their shape functions were used to interpolate the
stresses between atoms and provide stress-gradient information. Finally, the nucleation
criterion, Nm; l , was computed at every point in the crystal for every possible dislocation
(deJned by a Burgers vector and slip plane). The point at which the largest magnitude
of Nm; l occurred was identiJed as the predicted location of defect nucleation, while the
corresponding Burgers vector and slip plane for that maximum value were identiJed
as the predicted type of dislocation.
1518 R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525

At the same time, the atomic level stresses that were computed for each intermediate
conJguration were also used to compute the maximum resolved shear stress for the
Schmid-like nucleation criterion.
Three diNerent deformations were studied to observe dislocation nucleation. First,
indentation by a cylindrical indenter into a hexagonal lattice of Al atoms. Second,
the same indentation into an FCC lattice of Ni atoms was simulated. Finally, we
considered the uniaxial stretching of a hexagonal lattice of Al atoms into which we
initially introduced a small central void to serve as a nucleation site.
The indenter was modelled as a Jctitious atom with a radius of 100 A [ (compared
to the near-neighbour atomic spacing of 2:83 A [ in the hexagonal Al and 2:49 A[ in the
FCC Ni). This was achieved by introducing a repulsive interatomic potential between
the indenter and the atoms of the form
% = C(r − R)2 ; r 6 R;
(37)
% = 0; r ¿ R;

where r is the distance between the atom and the indenter, R is the indenter radius
[ and C = 100 eV A
(100 A) [ −2 . Thus, the indenter eNectively repelled atoms that were
[
less than 100 A from its centre, behaving like a rigid, frictionless sphere (Fig. 1).

5. Results

Fig. 2 shows the three simulations just after the nucleation of the Jrst sets of dislo-
cations. In the left-hand column, an overall view of the region of interest is presented
in order to orient the reader and for comparison to later Jgures which will show the
same regions. It should be emphasized that the actual size of the simulated crystal is
much larger than what is shown, but these Jgures highlight only the region near the
nucleation points. The frames in the right-hand column of Fig. 2 show a close-up of
the nucleated defects. The approximate locations of the dislocation cores (which nu-
cleate as dipole pairs) are denoted by the inverted “T”s, while the circles highlight the
locations where the defects nucleated. In order to make it easier to see the nucleation,
deformed Jnite elements are shown in the right-hand column of Jgures to highlight
atomic planes on which slip has taken place. Note that the nucleation of dipole pairs
is an artefact of the 2D calculation. In 3D, it is expected that dislocation loops would
nucleate. We mention, in passing, that the nucleation events in the indentation sim-
ulations coincide with a notable drop in the applied indenter load. This is consistent
with the observations of other atomistic simulations of indentation, as well as with the
energy dissipation premise in our thermomechanical formulation.
In Fig. 3, the same regions as those shown in the left-hand column of Fig. 2 appear
again. In this Jgure, the left-hand column shows contours of the largest critical resolved
shear stress, while the right-hand column shows contours of the largest value of Nm; l
from Eq. (34). In both columns, the contours have been obtained by computing the
relevant quantities for all possible dislocations that the crystal can support, but only
the largest value is plotted at each point. Thus, the red contours in each frame identify
R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525 1519

Fig. 1. Dislocation nucleation. Snapshots of the nucleated defects in the three atomistic simulations presented
here. The boxes in the Jgures of the left column are presented in close-up in the Jgures on the right. Locations
of the nucleated defects are approximately indicated in the right-hand Jgures by the inverted “T” symbols,
which were originally nucleated at the positions of the circles.

the predicted location of dislocation nucleation. These can be directly contrasted with
the black circles, which identify the actual locations at which the nucleation occurs.
It is clear from the left-hand column that the Schmid-like criterion does not accu-
rately predict defect nucleation. In both of the indentation simulations, the location
of maximum resolved shear stress is not in good agreement with the location of de-
fect nucleation. In the case of the void simulation, it does appear to agree with the
nucleation site, but it would seem that this agreement is coincidental.
In the right-hand column, we can clearly see that the gradient-based criterion is
largest exactly at the location of defect nucleation.
In the indentation of the hexagonal Al lattice shown in Fig. 3(b), we must note that
there is an artefact of the data analysis on the Jrst layer of atoms directly under the
indenter. The apparently large values along this layer are to be discounted as they do
1520 R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525

Fig. 2. Dislocation nucleation predictions. Comparison of the maximum Schmid stress (in the left column,
(a), (c) and (e)) with the maximum value of the stress-gradient based criterion (in the right column, (b),
(d) and (f)) for the three examples simulated. The circles indicate the actual locations of defect nucleation,
demonstrating that the Schmid-like criterion is not an accurate predictor of nucleation. The stress-gradient
based criterion, on the other hand, accurately predicts the location and the type of nucleated defect.

not correctly take into account the externally applied stresses from the indenter itself
—in essence the stress and stress gradients are not correctly computed for the topmost
row of atoms. The only other signiJcantly high Nm; l values occur at exactly the location
of defect nucleation. The maximum value is Nm; l = 0:0336 eV A [ −4 , and predicts the
nucleation of a dislocation with b = [a; 0; 0]. Close inspection of the nucleation process
(captured best through animated sequences of the simulation) reveals that this is in
fact the initial dislocation dipole pair which nucleates, but immediately thereafter the
dipole dissociates into the four dislocations shown in Fig. 2.
In the indentation of the FCC Ni lattice shown in Fig. 3(d), the artefact at the surface
is again present, and must be disregarded as in the Jrst example. There are several other
locations with relatively high Nm; l values, but the largest is found at the right-hand
nucleation site, where Nm; l = 0:0163 eV A [ −4 . In the sequence of conJgurations that
follow, a dislocation dipole nucleates at this point, followed closely by a dislocation
R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525 1521

0.016
0.014
0.014
0.012
0.012
0.01

Nm,l(eV/Å )
0.01

4
γ (eV/Å )
2

0.008
0.008
0.006
0.006
0.004 0.004

0.002 0.002

0 0
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
(a) δ /b (b) δ /b

Fig. 3. (a) Section of the (1 1 1) '-surface along the Shockley partial direction for FCC Ni, b = 1:437 A [
and (b) the value of Nm; l in the vicinity of the (1 1 1) slip plane during the sliding of two semi-inJnite
blocks of FCC Ni along the Shockley partial direction. The maximum value of Nm; l is Ncrit for this type of
dislocation.

dipole at the circle on the left, which in the conJguration shown here has a slightly
lower Nm; l = 0:0137 eV A [ −4 . The other regions near the surface where red contours
appear are unable to form dislocations, presumably because the nucleation events which
subsequently occur reduce the driving force on these regions. Finally, it is interesting to
note that the types of dislocations which nucleate on the left and right in this simulation
are diNerent. The one on the left is a 16 [1 2Z 1] Shockley partial whereas the one on the
right is a 16 [2Z 1 1] type. This is correctly borne out by the ranking of the magnitudes
of the Nm; l values in the two regions.
The void simulation provides a rigorous test of the criterion in the presence of a
free surface. Note that unlike the surface under the indenter, the void surface is not
loaded externally and thus the analysis routine which computes the atomic level stresses
is accurate in this case. Near a free surface, the atomic level stresses are strongly
inGuenced by the surface stresses which arise around the void, and as such may call
into question the applicability of the nucleation criterion. In fact, this would appear
not to be the case, as the location and type of dislocation nucleation is again correctly
predicted as shown in Fig. 3(f). This suggests the intriguing possibility of a very general
utility of the criterion. For example, it may be applicable in the heterogeneous regions
around grain boundaries and other microstructural features. Finally, the maximum value
of Nm; l = 0:0368 eV A [ −4 in this example is only 10% higher than the value computed
for the same material under a diNerent loading (the indentation of the Jrst example).
This is compelling evidence that there is in fact a material constant Ncrit that can be
tied to dislocation nucleation.
An important point needs to be made here. In our development of Nm; l , WT was
approximated by curl T. By deJnition, WT vanishes on the boundary of the body;
however, curl T does not, and this fact is useful in the prediction of nucleation from
a free surface. This suggests that in the theoretical development of driving forces,
the mathematical deJnition of the orthogonal decomposition used to deJne  and ⊥
components of a second-order tensor Jeld ought to be changed slightly so that the
1522 R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525

potential Jeld whose curl produces the ⊥ part is not required to vanish at the boundary.
Such a change in deJnition would result in an extra term reGecting dissipation at the
boundary involving WT , U̇ p , and Ũ˙ p whose consequences remain to be explored.

6. Comparison with other proposed nucleation criteria

6.1. Relation to Rice (1992) criterion

In an eNort to see the relationship between this criterion and the Rice criterion based
on the Peierls dislocation model, we have performed a simple atomistic simulation
whereby two semi-inJnite blocks of FCC Ni atoms are slid over each other on the
(1 1 1) plane along a Shockley partial slip direction. The change in energy of such
a system as a function of the amount of slip, ), is commonly used to compute the
so-called '-surface for the crystal, and the maximum excess energy per unit area along
a given slip direction is referred to as the unstable stacking fault energy (see, for
example, Sun et al., 1993). Fig. 3(a) shows the result of this calculation. We have
adopted the common approach of allowing relaxation of the atoms only in the direction
perpendicular to the (1 1 1) slip plane. In addition to the excess energy created in this
process, atomic level stress gradients are produced in the vicinity of the entire slip
plane. Thus, we have computed the maximum value of Nm; l at each ) and plotted
the results in Fig. 3(b). There are two things worthy of note in this result. First, the
maximum value of Nm; l occurs at the same degree of slip as the unstable stacking
fault energy. Second, the largest value of Nm; l is Nm; l = 0:0152 eV= A [ 4 . Interestingly,
this compares well with the critical value for the nucleation of Shockley partials in the
FCC Ni indentation described previously.

6.2. Relation to + criterion of Li et al. (2002)

The essential ingredient of the dislocation nucleation criterion of Li et al. (2002) and
Van Vliet et al. (2002) arises in continuum mechanics as the strong ellipticity condition
which Jnds use in considerations of uniqueness of incremental solutions, admissibility
of acceleration waves, and pointwise stability of local constitutive equations (Hill,
1962). It can also be interpreted as a threshold for linear stability of plane waves
in the theory of ‘small deformation superposed on large’ for a homogeneous base
state, and it is this interpretation that is perhaps most intimately connected to the issue
of dislocation nucleation in a continuum setting. While there are close connections,
in practical settings, between the typical occurrence of loss of strong ellipticity and
loss of ellipticity, the governing criterion for shear banding, these notions are not
equivalent, as stated in Van Vliet et al. (2002). It should also be noted that if dislocation
nucleation in a continuum setting is to be interpreted as a discontinuity in displacement,
as Van Vliet et al. (2002) interpret Rice’s (1976) work, then the criterion for shear
banding cannot serve as a useful one since displacement/velocity continuity and traction
continuity at a surface of discontinuity are essential ingredients of the analysis of shear
banding/stationary acceleration wave (Rice, 1976; Hill, 1962).
R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525 1523

Our nucleation criterion, Ncrit , arises from a nonlocal inelastic theory that diNers
signiJcantly in its mechanical structure from local elastic or elasto–plastic constitutive
theories. It arises from essentially thermodynamic considerations without the use of
further notions of stability. The diNerence in the mechanical structure of conventional
theories of inelastic response and the present one is further demonstrated by the fact
that mere thermodynamic analysis of conventional theories, as opposed to studies of
stability and uniqueness, does not provide conditions remotely close to what might be
interpreted as ones required for dislocation nucleation. Thus, Ncrit and the + criterion
are distinct from the conceptual standpoint. This can also be seen from the actual
criteria-+ depends upon local, instantaneous elastic moduli (based on the second Piola
–KirchhoN stress and the Green strain) and the Cauchy stress; Ncrit depends upon stress
gradients, or, roughly speaking, on the elastic distortion (F e ), the instantaneous moduli,
and the spatial gradient of elastic distortion. For the idealized setting of a homogeneous
base state, the + criterion would predict dislocation nucleation everywhere for a suit-
able level of deformation, while Ncrit would predict nucleation nowhere regardless of
the deformation level, and the latter would seem to be closer to what is physically
expected.

7. Conclusions

We have presented a stress-gradient based criterion for the nucleation of dislocations


in a crystal under applied loads. The criterion was then tested by performing direct
atomistic simulations of dislocation nucleation, and analysing the atomic level stresses
and stress-gradients in the crystal just prior to nucleation. In all of the examples pre-
sented, the criterion is exactly correct in predicting the location and type of defect
which nucleated during the simulation. By way of contrast, a Schmid-like criterion
based on a critical shear stress did not accurately predict the nucleation of crystalline
defects. The simulation results provide compelling evidence that the nucleation crite-
rion presented here is an accurate physical criterion for predicting defect formation,
and represents a signiJcant advance in our understanding of the fundamental processes
driving the plastic Gow of crystalline solids.
Our goal here has been to test whether a criterion based on continuum-level thermo-
mechanical arguments could correctly predict dislocation nucleation in an atomistic
model. Therefore, there was the need for the fully atomistic treatment of the simulations
to rigorously test the criterion. The favourable results of these tests, as presented herein,
open the door for the development of fully continuum models of defect nucleation,
without the need for explicit treatment of every atom. Indeed, this is a subject of
current development. Within such a continuum framework, we now have reason to
trust the Nm; l criterion due to the validation presented in this paper.
There are two additional current directions for this work. The Jrst is to extend the
simulations used for testing to fully three-dimensional situations and the nucleation of
dislocation loops. Secondly, it is of interest to investigate the relationships between
this criterion and other criteria such as that of Rice (1992). It may be possible to
interpret quantities like the unstable stacking fault energy in terms of gradients in the
1524 R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525

atomic-level stress, thus reconciling the apparently diNerent approaches to dislocation


nucleation.

References

Acharya, A., 2001. A model of crystal plasticity based on the theory of continuously distributed dislocations.
J. Mech. Phys. Solids 49, 761–785.
Acharya, A., 2003. Driving force and boundary conditions in continuum dislocation mechanics. Proc. R.
Soc. London A 459, 1343–1363.
Acharya, A., 2004. Constitutive analysis of Jnite deformation Jeld dislocation mechanics. J. Mech. Phys.
Solids 52, 301–316.
Born, M., Huang, K., 1954. Dynamical Theory of Crystal Lattices. Clarendon Press, Oxford.
Carlson, D.E., 1967. On Gunther’s stress functions for couple stress. Q. Appl. Math. XXV (2), 139–146.
Daw, M.S., Baskes, M.I., 1984. Embedded-atom method: derivation and application to impurities, surfaces,
and other defects in metals. Phys. Rev. B 29, 6443–6453.
De Koning, M., Cai, Wei., Bulatov, V.V., 2003. Anomalous dislocation multiplication in FCC metals. Phys.
Rev. Lett. 91 (2), 025503/1–025503/4.
De la Fuente, O.R., Zimmerman, J.A., Gonzalez, M.A., de la Figuera, J., Hamilton, J.C., Pai, Woei Wu,
Rojo, J.M., 2002. Dislocation emission around nanoindentations on a (001) fcc metal surface studied by
scanning tunneling microscopy and atomistic simulations. Phys. Rev. Lett. 88 (3), 036101.
Duesbery, M.S., Vitek, V., 1998. Plastic anisotropy in b.c.c. transition metals. Acta Mater. 46, 1481–1492.
Ercolessi, F., Adams, J.B., 1994. Interatomic potentials from Jrst-principles calculations—the force-matching
method. Europhys. Lett. 26, 583–588.
Foiles, S.M., Baskes, M.I., Daw, M.S., 1986. Embedded-atom-method functions for the fcc metals Cu, Ag,
Au, Ni, Pd, Pt and their alloys. Phys. Rev. B 33, 7983–7991.
Gouldstone, A., Van Vliet, K.J., Suresh, S., 2001. Nanoindentation: simulation of defect nucleation in a
crystal. Nature 411, 656.
Gumbsch, P., Beltz, G.E., 1995. On the continuum versus atomistic descriptions of dislocation nucleation
and cleavage in Nickel. Model. Simul. Mater. Sci. Eng. 3 (5), 597–613.
Hill, R., 1962. Acceleration waves in solids. J. Mech. Phys. Solids 10, 1–16.
Ito, K., Vitek, V., 1999. ENect of non-glide components of the stress tensor on deformation behavior of
BCC transition metals. Materials Research Society Symposium Proceedings, Vol. 538. pp. 87–92.
Kelchner, Cynthia L., Plimpton, S.J., Hamilton, J.C., 1998. Dislocation nucleation and defect structure during
surface indentation. Phys. Rev. B 58 (17), 11085–11088.
Khantha, M., Pope, D.P., Vitek, V., 1994a. The brittle-to-ductile transition—I: a cooperative dislocation
generation instability. Scr. Metall. Mater. 31 (10), 1349–1354.
Khantha, M., Pope, D.P., Vitek, V., 1994b. Dislocation screening and the brittle-to-ductile transition: a
Kosterlitz–Thouless type instability. Phys. Rev. Lett. 73 (5), 684–687.
Kosevich, A.M., 1979. Crystal dislocations and the theory of elasticity. In: Nabarro, F.R.N. (Ed.), Dislocations
in Solids. North-Holland Publishing Company, Amsterdam, pp. 33–141.
KrRoner, E., 1981. Continuum theory of defects. In: Balian, R., et al. (Ed.), Physics of Defects. North-Holland
Publishing Company, Amsterdam, pp. 217–315.
Kubin, L., Canova, G., 1992. The modelling of dislocation patterns. Scr. Metall. 27, 957–962.
Li, Ju., Van Vliet, K.J., Zhu, T., Yip, S., Suresh, S., 2002. Atomistic mechanisms governing elastic limit
and incipient plasticity in crystals. Nature 418 (6895), 307–310.
Mura, T., 1963. Continuous distribution of moving dislocations. Philos. Mag. 89, 843–857.
Rice, J.R., 1971. Inelastic constitutive relations for solids: an internal-variable theory and its application to
metal plasticity. J. Mech. Phys. Solids 19, 433–455.
Rice, J.R., 1976. The localization of plastic deformation. In: Koiter, W.T. (Ed.), Theoretical and Applied
Mechanics. North-Holland Publishing Company, Amsterdam, pp. 207–220.
Rice, J.R., 1992. Dislocation nucleation from a crack tip: an analysis based on the Peierls concept. J. Mech.
Phys. Solids 40, 239–271.
R.E. Miller, A. Acharya / J. Mech. Phys. Solids 52 (2004) 1507 – 1525 1525

Rice, J.R., Beltz, G.E., 1994. The activation energy for dislocation nucleation at a crack. J. Mech. Phys.
Solids 42, 333–360.
Shenoy, V.B., Phillips, R., Tadmor, E.B., 2000. Nucleation of dislocations beneath a plane strain indenter.
J. Mech. Phys. Solids 48, 649–673.
Shilkrot, L.E., Miller, R.E., Curtin, W.A., 2002. Coupled atomistic and discrete dislocation plasticity. Phys.
Rev. Lett. 89 (2) article #025501.
Sun, Y., Beltz, G.E., Rice, J.R., 1993. Estimates from atomic models of tension-shear coupling in dislocation
nucleation from a crack tip. Mater. Sci. Eng. A 170, 67–85.
Sun, Y.Q., Hazzledine, P.M., Hirsch, P.B., 2002. Cooperative nucleation of shear dislocation loops. Phys.
Rev. Lett. 88 (6) 065503, 1–4.
Van der Giessen, E., Needleman, A., 1995. Discrete dislocation plasticity: a simple planar model. Model.
Simul. Mater. Sci. Eng. 3, 689–735.
Van Vliet, K.J., Li, Ju., Zhu, T., Yip, S., Suresh, S., 2002. Quantifying the early stages of plasticity through
nanoscale experiments and simulations. Phys. Rev. B 67. 104105, 1–15.
Vitek, V., Egami, T., 1987. Atomic level stresses in solids and liquids. Phys. Stat. Sol. B 144, 145–156.
Willis, J.R., 1967. Second-order eNects of dislocations in anisotropic crystals. Int. J. Eng. Sci. 5, 171–190.
Wirth, B.D., Bulatov, V.V., De La Rubia, T.D., 2001. Atomistic simulation of dislocation-defect interactions
in Cu. Mater. Res. Soc. Conf. Proc. 650, R3.27.1–R3.27.6.
Xu, X.-P., Argon, A.S., Ortiz, M., 1995. Nucleation of dislocations from crack tips under mixed modes of
loading: implications for brittle versus ductile behaviour of crystals. Philos. Mag. A 72 (2), 415–451.
Yamakov, V., Wolf, D., Salazar, M., Phillpot, S., Gleiter, H., 2001. Length-scale eNects in the nucleation
of extended dislocations in nanocrystalline Al by molecular-dynamics simulation. Acta Mater. 49,
2713–2722.
Zbib, H., Rhee, M., Hirth, J.P., 1998. On plastic deformation and the dynamics of 3D dislocations. Int. J.
Mech. Sci. 40 113–127.
Zimmerman, J.A., Kelchner, C.L., Klein, P.A., Hamilton, J.C., Foiles, S.M., 2001. Surface step eNects on
nanoindentation. Phys. Rev. Lett. 87 (16), 165507.

You might also like