Sie sind auf Seite 1von 15

Journal of the Mechanics and Physics of Solids 88 (2016) 252–266

Contents lists available at ScienceDirect

Journal of the Mechanics and Physics of Solids


journal homepage: www.elsevier.com/locate/jmps

The inverse hall–petch relation in nanocrystalline


metals: A discrete dislocation dynamics analysis
Siu Sin Quek a,n, Zheng Hoe Chooi b, Zhaoxuan Wu a, Yong Wei Zhang a,
David J. Srolovitz c
a
Institute of High Performance Computing, 1 Fusionopolis Way, #16-16, Connexis, Singapore 138632, Singapore
b
School of Mechanical & Aerospace Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore
c
Departments of Materials Science and Engineering & Mechanical Engineering and Applied Mechanics, University of Pennsylvania,
Philadelphia, PA 19104, USA

a r t i c l e in f o abstract

Article history: When the grain size in polycrystalline materials is reduced to the nanometer length scale
Received 19 June 2015 (nanocrystallinity), observations from experiments and atomistic simulations suggest that
Received in revised form the yield strength decreases (softening) as the grain size is decreased. This is in contrast to
11 November 2015
the Hall–Petch relation observed in larger sized grains. We incorporated grain boundary
Accepted 20 December 2015
(GB) sliding and dislocation emission from GB junctions into the classical DDD framework,
Available online 31 December 2015
and recovered the smaller is weaker relationship observed in nanocrystalline materials.
Keywords: This current model shows that the inverse Hall–Petch behavior can be obtained through a
Discrete dislocation dynamics relief of stress buildup at GB junctions from GB sliding by emitting dislocations from the
Polycrystals
junctions. The yield stress is shown to vary with grain size, d, by a d1/2 relationship when
Grain boundary sliding
grain sizes are very small. However, pure GB sliding alone without further plastic ac-
Inverse Hall–Petch
comodation by dislocation emission is grain size independent.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction

When a polycrystalline material is deformed and undergoes plastic deformation, yield strength scales with average grain
size by d−1/2, which implies that the yield strength increases when we decrease the grain size. This is the classical Hall–Petch
relationship proposed by Hall (1951) and Petch (1953). The relationship is phenomenological, but one of several ways to
understand it is by considering dislocation pile-ups. If the grain boundaries in a polycrystalline material are assumed to
obstruct dislocation motion, which is often the case, dislocation multiplication within a grain will cause dislocations to pile
up against the grain boundary (GB). Ahead of the leading dislocation in the pile-up, the shear stress induced by the pile-up,
σ ∝ τ L/x (cf. Leibfried, 1951; Hirth and Lothe, 1992), where τ is the applied stress, L is the length of the pile-up, and x is the
distance ahead of the leading dislocation in the pile-up, which is obstructed from moving beyond the GB. This is akin to a
crack with the crack tip at the GB. If we define the yield stress of the material as the applied stress when the leading
dislocation of the pile-up can penetrate through the GB into the adjacent grain, the stress ahead of the pile-up s must
exceed a critical stress barrier, σ ⁎, of the GB. Alternatively, one can think of σ ⁎ as the stress required to operate a dislocation
source on the opposite side of the GB as the dislocation pile-up. Activation of this source due to the stress induced from the

n
Corresponding author.
E-mail address: quekss@ihpc.a-star.edu.sg (S.S. Quek).

http://dx.doi.org/10.1016/j.jmps.2015.12.012
0022-5096/& 2015 Elsevier Ltd. All rights reserved.
S.S. Quek et al. / J. Mech. Phys. Solids 88 (2016) 252–266 253

pile-up is similar to describing the transmission of the leading dislocation in the pile-up through the GB as discussed by
Quek et al. (2014). The yield stress of the material is therefore given by

σy ∝ σ ⁎ x/L (1)

Since the size of the pile-up, L is limited by the size of the grain, d, we can assume that L ∝ d , and Eq. (1) gives the
relationship that yield stress, sy scales with d−1/2. The above equation simply suggests that as the grain size decreases, a
larger applied stress is required for the dislocation to penetrate into the adjacent grain, hence, a higher yield stress.
Nevertheless, one quickly realises that there is a limitation to the pile-up theory as described in the last paragraph. As the
grain size becomes very small (typically less than tens of nanometer), the grain interior becomes starved of dislocations and
dislocation sources and there would be no dislocation pile-ups. We therefore expect to see a deviation from the Hall–Petch
relation when the grain size reaches this regime and this has indeed been observed in experiments (e.g., see Chokshi et al.,
1989; Nieman et al., 1989; Fougere et al., 1992). Chokshi et al. (1989) performed hardness tests on nanocrystalline copper
(Cu) and palladium (Pd) samples processed by inert gas condensation (IGC). Their results showed a clear decrease in
hardness (or yield stress) with a decrease in grain size, i.e., inverse Hall–Petch behavior. Nieman et al. (1989) performed
similar experiments and while they did not clearly show a ”smaller-is-softer” trend for their nanocrystalline Cu and Pd
samples, their hardness vs. d−1/2 plot showed a shallower positive slope as compared with coarse grain samples. Interest-
ingly, Fougere et al. (1992) tested the grain size dependence of hardness from samples that progressively increase in grain
size by annealing and found that hardness increases with increasing grain size (inverse Hall–Petch) to a critical size followed
by softening with further increase in grain size (Hall–Petch). Lu and Sui (1993), while discussing the experiments of Fougere
et al. (1992) and Chokshi et al. (1989), suggested that the annealing process relaxed the grain boundaries and changed their
interfacial properties. These observation suggest that when the grain size is very small, the contribution from lattice dis-
location plasticity is decreased and a different mechanism – one accommodated by grain boundaries – may dominate the
plastic deformation.
Chokshi et al. (1989) argued that the inverse Hall–Petch relation they observed is because of Coble creep (diffusional
creep along grain boundaries) at room temperature. At such small grain sizes, diffusional creep along the grain boundaries is
enhanced and has a similar effect as GB sliding in coarse-grain polycrystals at high temperature. Masumura et al. (1998)
generalized the Hall–Petch relation based on conventional Hall–Petch for large grains and Coble creep for very smaller
grains. They considered a log-normal distribution of grain sizes and proposed a model that integrates over the grain size
distribution such that for a given grain microstructure, there are contributions from both the Hall–Petch and Coble creep
mechanisms. Their model provided a reasonable fit to some experimental observations.
Asaro et al. (2003) proposed a model that assumes that for grain sizes of about hundreds of nanometers, dislocations/
stacking faults emission from GBs dominates deformation, and as grain sizes are further reduced to tens of nanometers, the
deformation mechanism changes to one dominated by GB sliding. A GB-bulk composite model was proposed to account for
the transition as grain size changes, and GB sliding was modeled with a flow law. In such a GB-bulk composite model,
deformation mechanisms in different size regimes are treated independently of each other and which grain size regime they
dominate depends on the volume fraction of the GBs (and bulk). Using this model together with the averaging approach by
Asaro and Needleman (1985) over an aggregate of grains, Zhu et al. (2005) proposed a constitutive model for nanocrsytalline
metals and studied the effect of grain size distribution on the deformation response. While an important advance, the
deformation mechanisms in the grain interior and in the GBs are not coupled in these models. In reality, we would expect a
strong coupling between the deformation mechanisms. Wei et al. (2008a) extended the model to include GB diffusion using
the finite element method, in which enforcement of compatibility and equilibrium conditions ensure that the deformation
mechanisms are coupled. They also showed that the transition of deformation mechanisms as one reduces the grain size
recovers the increase in strain rate sensitivity of nanocrystalline metals. While these models predicted the transition of
deformation mechanism from bulk plasticity to GB-based mechanisms as grain size is reduced, they did not explicitly
recover the inverse Hall–Petch relation first observed by Chokshi et al. (1989).
Argon and Yip (2006) considered the GB as an amorphous region of which a viscous flow process describes GB shear.
They also considered dislocation plasticity as thermally assisted release of dislocations from forest dislocation intersections
within the grain interior. The relative contribution to the plastic strain from the two processes is obtained from the volume
fraction of the GB region in the material, hence with grain size. Once again, the two deformation mechanisms are assumed
to be independent of one another. Nevertheless, they are able to obtain an optimal grain size that gives the highest flow
stress. However, the incompatibility arising from their GB shear is not accounted for in their model where they have as-
sumed the polycrystal simply as a set of parallel shearing grain boundaries. Several other models have been proposed to
describe the inverse Hall–Petch phenomenon. These models generally describe mechanisms involving lattice dislocations,
grain boundaries, and/or their interactions. To avoid including an incomplete list, we refer interested readers to the excellent
review papers by Meyers et al. (2006) and Pande and Cooper (2009).
The reality of what happens during the deformation of nanocrystalline materials may also be a synergistic combination
of various mechanisms. Molecular dynamics (MD) simulations provide an appealing way to directly observe the de-
formation in nanocrystalline materials with no a-priori assumption of what mechanisms actually occur during the de-
formation. Despite the limitation of drastically high strain rates in MD simulations, they can still provide insights on the
deformation behavior of nanocrystalline materials. Early MD simulations of the deformation of nanocrystalline metals
254 S.S. Quek et al. / J. Mech. Phys. Solids 88 (2016) 252–266

include studies by Schiøtz et al. (1998), Van Swygenhoven and Caro (1998), Van Swygenhoven and Derlet (2001) and Schiøtz
and Jacobsen (2003). MD simulations revealed that nanoscale grains readily exhibit GB sliding at room temperature. The
time scale for MD simulations are too short for these observed GB sliding to be a result of long-range diffusion; hence, the
observed GB sliding events are the result of stress-induced atomic shuffling. Li et al. (2009) showed that GB sliding and
dislocation emission from GBs are common occurrences in MD simulations of nanocrystalline aluminum, and these me-
chanisms contribute competitively to strain recovery during annealing. Wu et al. (2013) recently performed large scale MD
simulations of nanocrystalline Ni nanowires and bulk materials, and showed that GB sliding precedes plastic deformation by
lattice dislocations in these nanocrystalline materials regardless of sample size. In order to characterize the early states of
deformation, they unloaded their structure after a strain of 0.025 and observed a permanent residual strain attributed to GB
sliding, despite the complete absence of dislocation activity. At larger strains, GB sliding was accompanied by dislocation
plasticity, where partial and/or full dislocations were emitted from GBs and particularly GB junctions (with few or no
dislocations nucleated within the grain interior).
MD simulations suggest that GB sliding and other GB mediated plasticity mechanisms cause the deviation from the Hall–
Petch behavior in nanocrystalline materials. The fact that dislocations are nucleated from GBs and GB junctions also suggests
that dislocation plasticity is source limited in the nanocrystalline regime rather than limited by dislocation pileups, as in
large grain samples. While MD simulations provided us with insights on possible deformation mechanisms, they often do
not provide quantitative data on the plastic strain contributions of individual deformation mechanism, and it is therefore
not clear how these mechanisms would result in the inverse Hall–Petch trend. Nevertheless, we could garner insights from
MD to propose new models for the deformation of such materials. In the same spirit, the discrete dislocation dynamics
(DDD) methodology generally imposes constitutive relationships that describe dislocation interaction mechanisms in ad-
dition to the dynamics driven by long range stress fields. However, conventional DDD, when applied to polycrsytalline
materials, often only treat GBs as obstacles to dislocation motion (see Shishvan and Van der Giessen, 2010; Balint et al.,
2005, 2008; Nicola et al., 2005; Kumar et al., 2009; Ahmed and Hartmaier, 2010). This generally resulted in dislocation
pileup against the GBs and for large grain sizes, reproduces the Hall–Petch relationship as it rightfully should. Nevertheless,
when applied to nanocrystalline grains, two major issues exist: (1) typical dislocation or Frank-Read source densities imply
that the average spacing between dislocations or sources is much larger than the average grain size (i.e., there are few if any
dislocations/sources), which means that the nanocrystalline material will remain largely elastic; and (2) there is no alter-
native mechanism available to contribute to plastic slip.
In a previous paper by Quek et al. (2014), a DDD framework was proposed that models GB sliding by having GB dis-
locations gliding along the GB plane. These GB dislocations are consistent with the displacement shift complete (DSC) lattice
of GB structures (see Balluffi et al., 1982) and their Burger's vectors are typically smaller than lattice dislocations. As a result,
GB sliding appears to be more continuous than lattice dislocation slip. Furthermore, the implementation allows for these GB
dislocation sources to operate at relatively low stress resulting in GB sliding, even in the absence of any lattice dislocation
activity. Taking a cue from MD simulations, it would be of interest to study if we can reproduce the deformation behavior of
nanocrystalline materials using a DDD model that includes the above mentioned GB sliding mechanism as well as including
dislocation emission from GB junctions. As discussed by Wu et al. (2013) and Quek et al. (2014), dislocations are emitted
from GB junctions as a result of stress buildup from GB sliding. Such a model would enable us to study how GB sliding and
dislocation emission from GB junctions contributes to the observed grain size effects on the yield behavior of nanocrys-
talline materials.
In Section 2, we review the DDD simulation methodology and describe how we implement GB sliding and dislocation
emission from GB junctions in the present research. We also highlight the parameters used for the simulations performed
here. Section 3 shows the simulation results using the proposed model for different grain sizes. In Section 4, we discuss the
results obtained and propose a theoretical analysis of the mechanisms observed and how they contribute to the observed
grain size effect on yield strength. We then generalize the model to describe the transition from inverse Hall–Petch to Hall–
Petch behavior with increasing grain size.

2. Discrete dislocation dynamics framework

The two-dimensional discrete dislocation dynamics (2D DDD) approach is a popular method of explicitly tracking the
motion of dislocations and hence, the accumulation of plastic strain during deformation. Gulluoglu et al. (1989) used the
approach to track dislocation distributions in materials, and Van der Giessen and Needleman (1995) utilised the framework
to incorporate dislocation plasticity during mechanical deformation, which takes into account the long range elastic effects
between dislocations as well as short range dislocation interactions. We briefly describe the framework here and refer
interested readers elsewhere for more details (e.g., see Van der Giessen and Needleman, 1995; Shishvan and Van der
Giessen, 2010; Balint et al., 2005; Nicola et al., 2005; Balint et al., 2008; Kumar et al., 2009; Ahmed and Hartmaier, 2010;
Chakravarthy and Curtin, 2010; Quek et al., 2014).
To develop the ideas that are the focus of this work, we employ a simple 2D model – extension to 3D is conceptually
straightforward, but we acknowledge that “book-keeping” of 2D line segments in 3D space can be arduous. In addition,
extension to 3D would most certainly introduce more variable parameters in order to implement the concepts introduced
here. In the 2D DDD model, edge dislocations are modeled as point sources of deformation, characterized by a Burger's
S.S. Quek et al. / J. Mech. Phys. Solids 88 (2016) 252–266 255

Fig. 1. Simulation cell and its periodic images. Each dislocation in the simulation cell corresponds to an infinite array of dislocations in the x-direction (e.g.,
see the shaded region). The stress field associated with each dislocation and its images decays rapidly to a constant value. The stress field associated with
each dislocation is obtained by analytically summing the stress fields of the infinite array and adding three periodic images in the positive and negative
y-directions.

vector b. The edge dislocations are infinitely long in the direction normal to the simulation plane. Periodic boundary con-
ditions are assumed in both the horizontal x and vertical y directions. We assume isotropic linear elasticity and each dis-
location can only glide on a fixed (crystallographic) slip plane (i.e. no cross-slip and climb). We first consider the effects of
the long range elastic field induced by each dislocation. The stress state at the position of each dislocation can be computed
as the sum of the stresses induced by all other dislocations as well as any externally applied stress. For a 2D doubly periodic
system, the analytical expressions for the stress components of an infinite array of dislocations in one direction (say, in the
x-direction) can be used (since isotropic elasticity is assumed). The stress fields associated with such an array decay ex-
ponentially to a constant value at large distance from the line of such dislocations. Hence, the stress field of the infinite
periodic array (in the x -direction) can be summed in the y -direction with truncation after just a few images (cf. Hirth and
Lothe, 1992; Gulluoglu et al., 1989; Cai et al., 2003; Kuykendall and Cai, 2013). In this work, we used a total of six periodic
images – three in the negative y-direction and three in the positive y-direction, as shown schematically in Fig. 1. Following
Kuykendall and Cai (2013), the stress components at any point (x,y) induced by an array of dislocations in the x-direction
with Burger's vector, b , are given by:

μbx ⎡ πY (CY cX − 1) − SY (CY − cX ) ⎤ μbysX ⎡ CY − cX − 2πYSY ⎤


x∞
σxx = ⎢ ⎥+ ⎢ ⎥
(1 − ν )lx ⎣ (CY − cX )2 ⎦ 2(1 − ν )lx ⎣ (CY − cX )2 ⎦ (2)

−μbxπY ⎡ CY cX − 1 ⎤ μbysX ⎡ 2πYSY + CY − cX ⎤


x∞
σyy = ⎢ ⎥+ ⎢ ⎥
(1 − ν )lx ⎣ (CY − cX )2 ⎦ 2(1 − ν )lx ⎣ (CY − cX )2 ⎦ (3)

μbxsX ⎡ CY − cX − 2πYSY ⎤ μbyπY ⎡ CY cX − 1 ⎤


x∞
σxy = ⎢ 2
⎥− ⎢ ⎥
2(1 − ν )lx ⎣ (CY − cX ) ⎦ (1 − ν )lx ⎣ (CY − cX )2 ⎦ (4)

where X = x/lx , Y = y/lx , sX = sin 2πX , cX = cos 2πX , SY = sinh 2πY , CY = cosh 2πY , μ is the shear modulus, ν is the Poisson's
ratio, and bx and by are the x and y components of the Burger's vector. We then obtain the total stress tensor by summing up
the contributions from all periodic images in the y-direction and add any externally applied stress, σ a . There are some
subtleties associated with the truncation of this series, as discussed by Kuykendall and Cai (2013). Once the stress tensor at
the position of a dislocation is computed, the Peach–Koehler force acting on the dislocation can be obtained as (cf. Peach and
Koehler, 1950)
256 S.S. Quek et al. / J. Mech. Phys. Solids 88 (2016) 252–266

F = ( σ ·b) × ξ (5)

where ξ is the line direction of the dislocation. The velocity of the dislocation can then be obtained via a mobility law

v = M· F (6)

where M is a mobility tensor with the form

(
M = mg I − n ⊗ n ) (7)

where mg is the glide mobility constant, n is the normal vector of the slip plane and I is the identity matrix. (I − n ⊗ n) is the
orthogonal projection matrix that projects vectors onto the plane with normal n . Note that dislocation climb is not con-
sidered in this work. The position of the dislocation along its slip plane ri can then be updated through a first-order time
marching scheme:
ri(t + Δt ) = ri(t ) + vi(t )Δt, (8)

where vi is the projection of v onto the dislocation glide direction (plane). In this model, we impose two constitutive rules:
(1) dislocation nucleation and (2) dislocation annihilation. In 2D, nucleation is modeled by randomly placing point, dis-
location sources (mimicking Frank-Read sources in 3D). Upon operation, each source will nucleate a dislocation dipole pair
of opposite Burger's vectors placed ln apart and equidistance from the source in opposite directions. Nucleation occurs when
the resolved shear stress at the source exceeds a prescribed source strength, sn, for a sufficiently long period of time, tn. ln is
the (unstable) equilibrium distance between the two dislocations when subjected to a resolved shear stress of sn. Further
details can be found in the paper by Van der Giessen and Needleman (1995). For dislocation annihilation, we remove from
the simulation any pair of dislocations with opposite Burger's vectors that come within ϵd of each other. ϵd = 6b⪡ln is
prescribed for all simulations carried out in this work.

Fig. 2. Schematic of GB sliding and dislocation emission from a GB junction in a 2D polycrystal.


S.S. Quek et al. / J. Mech. Phys. Solids 88 (2016) 252–266 257

2.1. Additional slip mechanisms

In a recent paper (cf. Quek et al., 2014), we proposed a model to describe GB sliding within a DDD framework. The basic
idea is as follows. We model grain boundary sliding in a manner akin to dislocation glide along its slip plane. We focus
initially on the case of flat grain boundaries, where the analogy with dislocation glide is particularly apt. As on a slip plane
within a crystal, grain boundary dislocations may be nucleated and glide along the GB. However, the Burgers vector of GB
dislocations are typically much smaller than lattice Burgers vector, consistent with the DSC picture of GB structure (cf.
Balluffi et al., 1982). While this may or may not replicate the detailed mechanism by which real GBs slide, from a kinematic
point of view, it is a discrete sliding mechanism that reduces to classical sliding in the limit that the size of the Burgers
vector and the barrier for GB dislocation nucleation both tend to zero. Note that these GB dislocations do not represent the
inherent structure of the GB. Such a description of GB sliding is consistent with athermal GB sliding, which can occur when
the resolved shear stress acting on the plane of the GB exceeds a critical threshold. This approach also has the advantage
that it is consistent with the manner in which dislocation dynamics is simulated in classical 2D DDD.
To model GB dislocation nucleation, we place sources (2D equivalents of Frank-Read sources as is usual in 2D DDD ) along
each GB. Because the Burgers vectors of GB dislocations are much smaller than those of lattice dislocations, these sources
operate at lower stresses than do lattice dislocation sources; hence, GBs slide relatively easy until the GB dislocations begin
to pile up against some barrier. In real crystals, such barriers can be overcomed by a myriad of mechanisms, including climb
(not considered here) or emission of lattice dislocations (considered here). One particularly important type of barrier in
nanocrystalline materials are junctions between triplets of grain boundaries. These are intrinsic barriers that obviously play
roles of increasing importance with decreasing grain size. Note that while the transient stress field do not accurately re-
present the physical stress field of the GB during GB sliding, the steady state stress field from the eventual pileup against the
GB junction is equivalent to that of the stress concentration at the junction as a result of GB sliding.
As discussed in Introduction, a common observation in MD simulations is the emission of dislocations (or partial dis-
locations) from GB junctions (see Schiøtz et al., 1998; Schiøtz and Jacobsen, 2003; Van Swygenhoven and Caro, 1998; Van
Swygenhoven and Derlet, 2001; Wu et al., 2013). This is not surprising since sliding along one GB leads to the development
of a stress singularity at the GB junction that cannot be fully relaxed by sliding along the other two GBs meeting at the
junction. We treat GB junctions as sites for heterogeneous nucleation of dislocations. Consistent with the standard 2D DDD
nucleation model, we treat the emission of lattice dislocations from the GB junction as the operation of a dislocation source
within a grain, near the junction. To illustrate this idea, consider a dislocation source in a crystal near a GB junction, as
shown in Fig. 2. When the source operates, it emits a dislocation pair (dipole) on one of the slip systems of the crystal in
which it resides. Dislocation nucleation occurs when the resolved shear stress exceeds the source strength, just like in

Fig. 3. Schematic illustration of a 16 hexagonal grain periodic microstructure (i.e., the simulation cell is periodic in the x- and y-directions). The circle
around each GB junction is the area wherein lattice dislocations sources are randomly placed. The radius of the GB junction heterogeneous nucleation
regions are fixed at R¼ 3 nm regardless of d, Lx and Ly and the source density is ρj⊥ = 0.025 nm−2 .
258 S.S. Quek et al. / J. Mech. Phys. Solids 88 (2016) 252–266

standard DDD. However, a source near a grain boundary cannot have a critical (unstable equilibrium) size larger than the
j
distance from the source to the closest grain boundary along the slip plane, ld. The source strengths, sn for these junction
sources are therefore given by
μb 1
σnj = .
2π (1 − ν ) 2ld (9)

Once a junction source operates, a dislocation dipole is nucleated, with one dislocation being pulled into the GB, close to the
junction, and the other driven away from the GB junction into the grain interior. The dislocation at/near the GB junction will
be of the sign to cancel part of the net Burgers vector associated with GB dislocation pile-up at the GB junction. This changes
the character of the GB junction, which is similar to how lattice dislocation absorption at a GB occurs, as described in Quek
et al. (2014). Operationally, we describe the heterogeneous nucleation of dislocations near the GB junction by randomly
placing dislocation sources with a prescribed source density ρj⊥ within a circular region of radius R and on a randomly
chosen slip system within its host grain. This accounts for the heterogeneous nucleation of lattice dislocations at GB
junctions.
GB sliding and dislocation emission from GB junctions are synergistic GB mechanisms for plastic relaxation of stress that
are distinct from crystal plasticity. A schematic example of these mechanisms is shown in Fig. 2. In the figure, an applied
shear stress activates the dislocation source (  ) at the center of one GB, resulting in GB sliding. The GB dislocations pile up
at the GB junction leading to a stress concentration. Note that the stress field ahead of the pileup at the GB junction is
similar to that ahead of a mode II crack tip; the stress field associated with this pile-up creates a stress that scales with 1/ r ,
where r is the distance ahead of the GB junction. The dislocation source (°) near the GB junction feels the stress field from
the junction and if it is high enough to operate the source, a lattice dislocation dipole is emitted: one runs back to the GB
junction and the other into the grain. The random nature of the location of the sources and the randomness of the choice of
slip system give the heterogenous nucleation a stochastic character.

2.2. Simulation model

To study the effects of grain size on the yield stress of nanocrystalline materials, we consider a 2D microstructure
consisting of uniformly sized, hexagonal grains, as shown in Fig. 3. Note that since we assume isotropic elasticity, there are
no image forces from the GBs and the GBs are simply obstacles to dislocation motion. The doubly periodic simulation cell
always consists of 16 grains; grain size is varied by changing the simulation cell dimensions, Lx and Ly. A grain i has six

Fig. 4. Dislocation configurations at different applied stress levels for grain size (a) d¼ 40 nm, and (b) d ¼10 nm. The red symbols along the GBs are the GB
dislocations that account for GB sliding. The blue symbols are lattice dislocations emitted from GB junctions. Axes dimensions are in terms of lattice
Burger's vector, b. (For interpretation of the references to color in this figure caption, the reader is referred to the web version of this paper.)
S.S. Quek et al. / J. Mech. Phys. Solids 88 (2016) 252–266 259

possible slip systems ϕ = nπ /3, where n is any integer, relative to the orientation of the grain 0 ≤ θi < π /3 (relative to a fixed
laboratory frame). As we simulate different grain sizes, we keep the same grain orientation distribution in order to minimize
the effect of variation in microstructure. In this work, we vary the grain size in the range 10 nm ≤ d ≤ 40 nm and
|b| = 0.2 nm . Based on these dimensions and typical dislocation source densities, ≈1010 to 1012 m  2 (cf. Balint et al., 2005),
there would only be a couple of dislocation sources in the entire simulations cell for the case of the largest grain size
simulated (d¼ 40 nm) and none for the smaller grain size samples (i.e., the average spacing between dislocations sources is
larger than our simulation cell dimensions). Therefore, no dislocations sources within the grains are included in our si-
mulations (with the exception of those associated with the GB junctions). This is consistent with the lack of experimental
observations of dislocation nucleation within grain interiors in nano-meter size grains. We randomly place GB dislocation

sources with a density of ρGB = 0.35 nm−1 along the GBs to simulate GB sliding. The Burger's vector of the GB dislocations is

assumed to be bGB = 0.1b, where b is the magnitude of the lattice Burger's vector. ρGB is kept constant as we vary grain size.
The sources placed along the GBs are randomly prescribed a nucleation length, ln, from a Gaussian distribution with a mean
of ln = 2 nm (corresponding to σn = 2.375 × 10−3μ). In addition, heterogeneous nucleation of dislocations near the GB
junctions is modeled by prescribing a circular region of radius R¼3 nm around the GB junction (circles drawn around GB
junctions in Fig. 3), wherein lattice dislocation sources (with a density of ρj⊥ = 0.025 nm−2) are randomly positioned. These
sources can emit full lattice dislocations; the nucleation length is determined by twice the distance of the source from the
GB junction.

3. Effect of grain size on the yield stress

With the model described in the previous section, we study the effect of grain size on the yield stress. The applied
a a
uniaxial stress syy is increased quasi-statically in increments of Δσyy = 0.005μ . In other words, we increment the stress and
integrate the dynamical equations for dislocation nucleation and dislocation migration forward in time until all dislocation
activity ceases. At this point, the applied stress is incremented again and the same procedure is repeated until the desired
maximum applied stress is achieved. The plastic strain is calculated from the motion of all of the dislocations in the system.
We define the yield stress to be the applied stress required to achieve a plastic strain, ϵyy of 0.002 (i.e., 0.2% offset yield).
p
a
Fig. 4 shows the dislocation configurations at different applied stresses. The applied stress, syy, will result in a resolved
shear stress along the non-horizontal GBs that leads to GB sliding; note that the resolved shear stress along the horizontal
GBs due to the applied load is zero. In both grain sizes, d ¼40 nm and d ¼10 nm (Figs. 4(a) and (b), respectively), at small
a
applied stress levels (σyy < 0.015μ ) GB sliding is observed to occur via the nucleation and glide of the GB dislocations (in red)
a
along the GBs. For the d¼ 40 nm case, lattice dislocations (in blue) begin to emerge from the GB junctions when σyy ≥ 0.020μ ,
a
while in the d¼10 nm case, lattice dislocations emission occurs earlier, σyy ≥ 0.015μ . Note that the lattice dislocations
emitted from the GB junctions are a result of the stresses associated with GB sliding being obstructed by the GB junctions.
Also note that at large stresses, some sliding of the horizontal GBs occurs due to the elastic fields induced as a result of
a
sliding of the other GBs (e.g., see the σyy = 0.025μ for both d ¼40 nm and d ¼10 nm in Fig. 4).
We plot the accumulated plastic strain ϵyy as a function of the applied stress for grain sizes d¼ 10 nm, 12.5 nm, 15 nm,
p

17.5 nm, 20 nm, 30 nm, 35 nm, and 40 nm in Fig. 5. For each grain size, a minimum of 10 independent simulations were
performed to account for statistical variation in the location of the dislocation sources. The error bars indicated the standard
deviations obtained from the different simulations. Fig. 5 shows that the plastic strain is largely unaffected by grain size for

Fig. 5. Plastic strain as a function of the applied stress for several grain sizes.
260 S.S. Quek et al. / J. Mech. Phys. Solids 88 (2016) 252–266

Fig. 6. Average yield stress as a function of grain size, d.

Fig. 7. Dislocation density obtained by dividing the number of dislocations by the area of simulation cell of (a) lattice dislocations emitted from triple
junctions, and (b) GB dislocations, as a function of applied stress for different grain sizes. Each data point is the equilibrated dislocation density for the
corresponding applied stress.
S.S. Quek et al. / J. Mech. Phys. Solids 88 (2016) 252–266 261

a a
σyy ≤ 0.015μ . At these stresses, plastic strain is dominated by GB sliding (see Fig. 4). For applied stresses above σyy = 0.015μ ,
the plastic strain increases with decreasing grain size for the smaller grain size simulations (d < 20 nm ). For large grain sizes,
(d ≥ 20 nm ), however, no statistically significant grain size effects are observed. We define the yield stress to be the applied
stress for which ϵ pyy = 0.002 as indicated by the horizontal, dashed line in Fig. 5. The average yield stress and the standard
error (for each grain size) obtained is plotted in Fig. 6 and it shows a clear ”smaller is weaker” trend (inverse Hall–Petch) for
d ≤ 20 nm . In this small size regime, the yield stress obtained from our DDD simulations grows with increasing grain size
(i.e. σy ∝ d n where n > 0). However, the range of the simulation data is too narrow (10–20 nm) to accurately derive an
exponent in the absence of a model. We return to the question of what is the appropriate value of n and what mechanisms
control it in the next section of this paper. For larger grain size, d > 20 nm , the yield stress is independent of grain size
within the statistical accuracy of our results.
To understand the origin of these observations, we plot the dislocation density as a function of applied stress for both
lattice and GB dislocations in Fig. 7. These densities are obtained by counting the number of GB or lattice dislocations and
divided by the area of the simulation cell. The data points shown in Fig. 7 are the equilibrated densities at each applied stress
value (i.e., obtained when dislocation activity has ceased). We first note that the lattice dislocation densities are an order
magnitude smaller than the GB dislocation densities. This, by itself, does not mean that the lattice dislocation contribution
to the plastic strain is insignificant, since the magnitude of the Burgers vector of the lattice dislocations is ten times larger
than that of the GB dislocations in the current simulations. Fig. 7(a) shows a trend very similar to the plastic strain (Fig. 5).
The strong correlation suggests that the grain size effect on plastic strain (and therefore the yield stress) is dominated by the
emission of lattice dislocations from GB junctions. Fig. 7(b), on the other hand, shows that the GB dislocation density
increases monotonically with decreasing grain size for any applied stress. This is not surprising since the total GB length
increases with decreasing grain size as 1/d. However, the GB dislocation density is not a direct measure of GB sliding, as
discussed in the next section.

4. Analysis

4.1. GB sliding

In the model presented above, GB sliding is represented by the nucleation and glide of GB dislocations along the GB
plane. This results in dislocation pile ups at GB junctions, which produce a long range stress field that both shuts down the
GB dislocation sources and leads to a significant stress concentration in the vicinity of the junctions. To analyze this si-
tuation, we consider the simple case shown schematically in Fig. 2 in which a single GB of length l is terminated at its ends
by a pair of GB junctions. We assume that a single GB dislocation source is located in the center of the GB (x ¼0) and a shear
stress τ is applied. Dislocations dipoles are emitted from the source, leading to a double pileup of GB dislocations against the
GB junctions until the back stress on the source is sufficient to stop the source from operating. This can be analyzed in terms
of a continuous distribution of dislocations with a spatial distribution ρGB (x ) (cf. Leibfried, 1951; Hirth and Lothe, 1992):
2(1 − ν )τ x
ρGB (x) =
μb (l/2)2 − x2 (10)

Applying Orowan's formula to determine the total plastic strain associated with this dislocation density distribution, we find
l /2 l /2 2(1 − ν )τ x2 πτ (1 − ν )l2 πτ (1 − ν )d2
ϵ gb
p = 2 ∫0 ρGB (x)bx dx = 2 ∫0 dx = = ,
μ (l/2)2 − x2 4μ 12μ (11)

where in the last expression, we make use of the fact that the GB length l is proportional to d (l = d/ 3 for the hexagonal
grains considered in the current model).
In order to extend this result to the general case of an arbitrarily oriented GB subject to an arbitrary stress, we simply
replace τ in the above equation with the resolved shear stress, σ r = τm, where m = cos(λ )cos(ϕ) is the Schmid factor and λ
and ϕ are the angle between τ and the glide plane, and the angle between τ and the glide direction, respectively. Both λ and
ϕ are independent of d. Therefore, the total plastic strain from all GBs in the microstructure can be written as:
NGB
πσir (1 − ν )d2
ϵGB
p = ∑
i=1 12μ

πτ (1 − ν )d2
∝ NGB
12μ

πτ (1 − ν )
∝ ,
12μ (12)
262 S.S. Quek et al. / J. Mech. Phys. Solids 88 (2016) 252–266

Fig. 8. Plastic strain from pure GB sliding (i.e., with no dislocation nucleation at GB junctions) as a function of applied stress for several grain sizes.

where NGB ∝ 1/d2 is the number of GBs in the microstructure. Eq. (12) is independent of the grain size d. In the model
presented above, this expression describes the total plastic strain associated with GB sliding, assuming that each GB slides
independently of each other. This assumption implies that there is no continuous sliding path from one GB to the next.
In order to validate this analysis, we simulated the case where emission of lattice dislocations from the GB junctions is
prohibited and all of the plastic strain is associated with GB sliding. However, note that in the DDD simulations, GB sliding is
not entirely independent of adjacent GBs. This is because the stress buildup at a GB junction as a result of GB sliding can
cause adjacent GBs to slide (by operating GB dislocation sources) – continuous GB sliding. The results, plotted in Fig. 8, show
that even when continuous GB sliding is allowed to occur, there is no significant grain size effect on plastic deformation, in
agreement with Eq. (12). This shows that continuous GB sliding from one GB to another does not occur readily since they
have to satisfy compatibility of the GB orientation with the loading direction. The size independence is easily understood.
Even though the total number of GBs increases with decreasing grain size, the length of each GB and therefore the size of the
dislocation pileup on each GB also decreases. The net result is that the plastic strain is independent of grain size, as seen
from Eq. (12), Fig. 8 and also Fig. 5 at small applied stress where few or no lattice dislocations have been emitted from the
GB junctions.

4.2. Yield

The current DDD framework describes the emission of a lattice dislocation from a GB junction through heterogeneous
nucleation of dislocations from the region surrounding the junction. In order to analyze the effect of grain size on the yield
stress, we examine how grain boundary sliding gives rise to the nucleation of lattice dislocations. As above, we first consider
the case of a single GB of length l with GB dislocations piling up at the junction as shown schematically in Fig. 2. The stress
field around the junction is akin to the stress field near the tip of a mode II crack of length l loaded in pure shear. This stress
field can be analyzed in terms of a continuous distribution of dislocations by (cf. Leibfried, 1951; McClintock and Argon,
1966; Hirth and Lothe, 1992):

l
σ(r ) ≅ τ .
r (13)

Since l ∝ d , we can rewrite Eq. (13) as

d
σ(r ) ∝ τ
r (14)

If there is a dislocation source ahead of the GB junction, it will operate if the source strength σ is smaller than the stress at
the source location (a distance r from the GB junction), σ(r ).
Since we focus here on the yield stress, we are concerned about the limit where only some of the GB junction sources are
active. The total density of lattice dislocations emitted from the GB junctions is proportional to the density of such junctions
and the fraction of such sources that operate, α, that is
ρd ∝ αρj , (15)

where ρj is the density of GB junctions in the microstructure. The density of GB junctions scales as ρj ∝ 1/d . We postulate 2
S.S. Quek et al. / J. Mech. Phys. Solids 88 (2016) 252–266 263

that the fraction of sources that operate scales as the ratio of the stress induced by the pileup at the source location to the
source strength. Here, we note that the actual distribution of source strengths is statistically distributed and that σ(r ) scales
the entire GB source strength distribution. That is,

α=0 for σ < σmin
σ ⁎ ⁎
α∝ ⁎ for σmin ≤ σ ≤ σmax
σmax (16)

σmin is the minimum source strength (this implies that no dislocation is nucleated/emitted if the stress induced by the pileup

is less than the strength of the weakest source) and σmax is the maximum source strength.
Inserting the definition of ρj, α and σ(r ) into Eq. (15), we find

ρd = 0 for σ < σmin
⎛ 1 ⎞⎛ τ ⎞ τ
ρd ∝ ⎜⎜ 2 ⎟⎟⎜⎜ ⁎ d1/2⎟⎟ ∝ ⁎ 3/2 ⁎
for σmin ⁎
≤ σ ≤ σmax
⎝ d ⎠⎝ σmax ⎠ σmaxd (17)

This implies that in the early stages of lattice dislocation plasticity (σ > σmin ), the total dislocation density emitted from
GB junctions scales with the shear stress on the boundary and grain size as, ρd ∝ τd−3/2. The density of lattice dislocation
emitted shown in Fig. 7(a) from the simulations shows agreement with Eq. (17), where the lattice dislocation density having
a largely linear trend with applied stress when the applied stress exceeds a critical value. There might be some discrepancy
from linearity observed for the smaller grain sizes, but in these cases, there is also a larger scatter of simulated data.
Assuming that each dislocation emitted from a GB junction travels across the grain, the total plastic strain accumulated
can be expressed as:
ϵ p ∝ ρd bd. (18)

Therefore, substituting Eq. (17) into the above equation yields


τ τb
ϵp ∝ bd ∝ ,
d3/2 d1/2 (19)

which can be rearranged to give

ϵ pd1/2
τ∝ .
b (20)

If we define yield as the point where the material accumulates a total plastic strain of ϵ⁎p , then the yield stress is

ϵ⁎pd1/2
σy ∝
b (21)

The main conclusion of this analysis is that the yield stress increases with grain size as d1/2 in the regime where most of the

Fig. 9. Average yield stress as a function of d1/2 , as per Eq. (21). The straight line shows a best fit line to the yield stress data for small grain sizes,
d ≤ 17.5 nm or d1/2 ≤ 4.183 nm1/2 . The goodness of fit measure for the linear regression, R2 = 0.99213.
264 S.S. Quek et al. / J. Mech. Phys. Solids 88 (2016) 252–266

Fig. 10. Hardness versus d1/2 for the experimental data from Refs. Chokshi et al. (1989) and Fougere et al. (1992).

lattice dislocations nucleate heterogeneously at GB junctions as a result of GB sliding.


Fig. 9 shows the same simulation data as in Fig. 6, but now replotted in accordance with our theoretical scaling result in
Eq. (21), i.e., σy ∝ d1/2 . The straight line in Fig. 9 is a linear regression of the small grain size data, i.e., for d ≤ 17.5 nm or
d1/2 ≤ 4.183 nm1/2. The goodness of fit measure for this linear regression line is R2 = 0.99213. We note that while the data in
Fig. 6 may suggest that σy ∝ d for the same data range, a log–log plot of sy versus d shows that σy ∝ d1/2 does indeed provide a
much better fit to the simulation data. At larger grain size (d ≥ 20 nm ), the yield stress appears nearly grain size in-
dependent within the statistics of our simulations (see Figs. 5 and 9). This is exactly the grain size range where the density of
lattice dislocations becomes nearly grain size independent, as shown in Fig. 7(a), where dislocation emission from GB
junctions becomes less dominant when d ≥ 20 nm . The plateau in the yield stress versus grain size at d ≥ 20 nm can be
attributed to the fact that for large grain sizes, the contribution to dislocation emission from GB junctions is small compared
with grain boundary sliding. Of course, in real polycrystals, as the grain size becomes large, the dominant contribution to the
grain size effect in yielding is dislocation pile-up at grain boundaries from sources internal to the grains (Hall–Petch be-
havior) – an effect not included in our simulations here.
To further validate our analysis, we replotted several sets of experimental yield strength data from Chokshi et al. (1989)
and Fougere et al. (1992) showing the inverse Hall–Petch behavior at small grain sizes, in our predicted form (i.e., σy ∝ d1/2 );
see Fig. 10. We find that the linear regression fits are very good to reasonable in all cases.
From the analyses above, the inverse Hall–Petch behavior is attributed to the coupling of GB sliding with lattice dis-
location emission from the GB junction. The coupling is a natural relief of the stresses at the GB junction. The current DDD
model is therefore able to reproduce the inverse Hall–Petch relationship and the yield stress–grain size scaling relationship
is derived based on mechanisms observed during the deformation of nanocrystalline materials. We recall that several
previous models for the inverse Hall–Petch were developed (see Meyers et al., 2006; Pande and Cooper, 2009 for reviews).
For example, Gryaznov et al. (1993) used a composite model consisting of a “grain” matrix with embedded, interfacial (GB)
layers. Using their model, they proposed an expression for the yield stress–grain size relationship of the form:

σy = σ0 + k 0 log d + ∑ kid−i /2
i=1 (22)

The log d term corresponds to the contribution to the yield stress by disclinations, while in general, the constants, ki and
which value of the exponent i that dominates deformation depends on the mechanism(s). However, this model is built upon
the phenomenological Hall–Petch relation instead of being derived fundamentally from deformation mechanisms.
Chokshi et al. (1989) suggested that the inverse Hall–Petch behavior they observed is because of Coble creep; a concept
further developed by Masumura et al. (1998) into a unified yield stress–grain size relation. They proposed that the yield
strength scales with d3 (Coble creep) for very small grains, and with the usual d−1/2 (Hall–Petch) for very large grains. Their
model superimposes the two mechanisms (and accounts for a log-normal grain size distribution). Using this approach, they
proposed a unified yield stress–grain size relationship of the form:

σy = f (d−1/2) + f (d−1) + f (d3) (23)

The second term in Eq. (23) is an ad hoc threshold term that was later discussed by Pande and Cooper (2009) in terms of
defect characteristics of GBs. However, Eq. (23) is only strictly valid for a log-normal grain size distribution, while away from
normal grain growth (in the absence of solute), grain size distributions are often not of this form. Clearly, Eq. (23) is not
consistent with the present simulation results and differs significantly from the model proposed here (at small d). This is not
S.S. Quek et al. / J. Mech. Phys. Solids 88 (2016) 252–266 265

Fig. 11. Different regimes of grain size effect on yield stress. Regime I is the classical Hall–Petch regime dominated by dislocation pile-up against GBs.
Regime IIII is the inverse Hall–Petch regime dominated by lattice dislocation emission from triple junctions as a result of GB sliding. And an intermediate
regime II where pure GB sliding dominates. The symbols are simulation data points obtained in this work.

surprising, however, given the differences in assumptions between those used in our discrete dislocation dynamics simu-
lation and the Coble creep model. Similarly, Wei et al. (2008b) considered thermally activated GB sliding and GB diffusion
with the absence of dislocation mechanism, and showed the strong coupling between GB sliding and GB diffusion. In
contrast, the model presented here describes athermal GB sliding and no GB diffusion is considered. This results in a
coupling with lattice dislocation emission from the GB junctions, which gives the inverse Hall–Petch behavior. Different
mechanisms have been proposed to account for deformation in nanocrystalline metals in the scientific community, and in
our opinion, these mechanisms probably contribute in varying degrees to the deformation. Part of the goal of this work is
therefore to propose these additional mechanisms revealed through a simple model that could possibly play a significant
role in explaining the size effects observed in these nanocrystalline systems.
Many decades of experimental study (cf. Hall, 1951; Petch, 1953; Nieman et al., 1989, 1991; Fougere et al., 1992; Sanders
et al., 1997) have shown that the yield stress of polycrystalline metals scales as ∝d−1/2 for micrometer and larger grain sizes,
in accordance with the Hall–Petch relationship. The focus of the present simulations study and analysis was on the very
small grain size regime, where smaller is weaker has been observed (e.g., Chokshi et al., 1989; Fougere et al., 1992). To model
this regime, we presented 2D DDD simulation results for a model polycrystal that includes two key effects not commonly
incorporated in 2D DDD; namely, (1) GB sliding and (2) dislocation emission from GB junctions. We showed that this reliably
reproduces the inverse Hall–Petch relationship at very small grain sizes (d ≤ 20 nm ). The simulation data and analysis
suggest that σy ∝ d1/2, in agreement with experiment. Interestingly however, our model predicts that GB sliding alone makes
no contribution to the grain size dependence of the accumulated plastic strain. It is only when coupled with dislocation
nucleation at triple junction that grain boundary sliding produces a grain size effect. This indicates that there could possibly
be a regime, intermediate between the Hall–Petch and inverse Hall–Petch regimes, that is dominated by pure GB sliding and
is grain size independent. We illustrate this schematically in Fig. 11. In its entirety, the yield stress–grain size relationship
can be described by the convex (lower) hull formed by the predicted behavior in each of the three regimes. We note that
regime II may not be easily achievable depending on where the crossover between Hall–Petch and inverse Hall–Petch
behavior occurs and the stress levels where the grain size independent regime (II) pertains. This will, of course, depend on
several factors including material system, processing method, etc. This may explain the discrepancies found in experiments
with regards to how exactly the yield stress–grain size relationship deviates from the Hall–Petch relationship for nano-
crystalline materials; some studies show a yield stress plateau with decreasing grain size (Nieman et al., 1989; Sanders et al.,
1997), while others see an inverse Hall–Petch relationship (Chokshi et al., 1989; Fougere et al., 1992).

5. Conclusion

We have described a discrete dislocation dynamics model to study the grain size effects of nanocrystalline metals. GB
sliding and dislocation emission from GB junctions are incorporated into the DDD framework, and by doing so, reproduced
the inverse Hall–Petch relation observed for very small nanocrystalline grains. It was found using the model that for small
applied load, the yield stress is dominated by GB sliding and is grain size independent. As the loading increases, stress build-
up as a result of GB sliding causes the emission of lattice dislocations at the triple junctions. For small grains, this con-
tribution to plastic slip is significantly large and there is a clear grain size effect. The yield stress is shown to scale linearly
with d1/2 in this regime where grain size is small. As grain size increases, the contribution to plastic slip from emission at
triple junctions diminishes as the triple junction density diminishes. Without the contribution from lattice dislocation
266 S.S. Quek et al. / J. Mech. Phys. Solids 88 (2016) 252–266

multiplication within the grain interior (as assumed in this work), the plastic strain becomes dominated by GB sliding for
larger grain sizes. From our analyses and simulations, we also show that GB sliding on its own, is grain size independent – a
subtle contradiction to conventional understanding. It is only with the emission of lattice dislocations from GB junctions as a
result of GB sliding, that the inverse Hall–Petch is observed for very small grains.

Acknowledgments

This work was supported by the A*STAR Computational Resource Centre through the use of its high performance
computing facilities. ZHC would like to acknowledge the support given by the Industry-sponsored FYP programme from the
School of Mechanical and Aerospace Engineering, Nanyang Technological University.

References

Ahmed, N., Hartmaier, A., 2010. A two-dimensional dislocation dynamics model of the plastic deformation of polycrystalline metals. J. Mech. Phys. Solids
58, 2054–2064.
Argon, A.S., Yip, S., 2006. The strongest size. Philos. Mag. Lett. 86, 713–720.
Asaro, R.J., Krysl, P., Kad, B., 2003. Deformation mechanism transitions in nanoscale fcc metals. Philos. Mag. Lett. 83, 733–743.
Asaro, R.J., Needleman, A., 1985. Overview no. 42 texture development and strain hardening in rate dependent polycrystals. Acta Metall. 33, 923–953.
Balint, D., Deshpande, V., Needleman, A., Van der Giessen, E., 2005. A discrete dislocation plasticity analysis of grain-size strengthening. Mater. Sci. Eng. A—
Struct. Mater. Prop. 400, 186–190.
Balint, D., Deshpande, V., Needleman, A., Van der Giessen, E., 2008. Discrete dislocation plasticity analysis of the grain size dependence of the flow strength
of polycrystals. Int. J. Plast. 24, 2149–2172.
Balluffi, R.W., Brokman, A., King, A.H., 1982. CSL/DSC Lattice model for general crystalcrystal boundaries and their line defects. Acta Metall. 30, 1453–1470.
Cai, W., Bulatov, V.V., Chang, J., Li, J., Yip, S., 2003. Periodic image effects in dislocation modelling. Philos. Mag. 83, 539–567.
Chakravarthy, S.S., Curtin, W., 2010. Effect of source and obstacle strengths on yield stress: a discrete dislocation study. J. Mech. Phys. Solids 58, 625–635.
Chokshi, A., Rosen, A., Karch, J., Gleiter, H., 1989. On the validity of the hall–Petch relationship in nanocrystalline materials. Scripta Metall. 23, 1679–1683.
Fougere, G.E., Weertman, J.R., Siegel, R.W., Kim, S., 1992. Grain-size dependent hardening and softening of nanocrystalline Cu and Pd. Scripta Metall. Mater.
26, 1879–1883.
Van der Giessen, E., Needleman, A., 1995. Discrete dislocation plasticity: a simple planar model. Model. Simul. Mater. Sci. Eng. 3, 689.
Gryaznov, V.G., Gutkin, M.Y., Romanov, A.E., Trusov, L.I., 1993. On the yield stress of nanocrystals. J. Mater. Sci. 28, 4359–4365.
Gulluoglu, A., Srolovitz, D., LeSar, R., Lomdahl, P., 1989. Dislocation distributions in two dimensions. Scripta Metall. 23, 1347–1352.
Hall, E., 1951. The deformation and ageing of mild steel: Iii discussion of results. Proc. Phys. Soc. Sect. B 64, 747.
Hirth, J., Lothe, J., 1992. Theory of Dislocations. Krieger Publishing Company, Malabar, Florida.
Kumar, R., Nicola, L., Van der Giessen, E., 2009. Density of grain boundaries and plasticity size effects: a discrete dislocation dynamics study. Mater. Sci. Eng.
A 527, 7–15.
Kuykendall, W.P., Cai, W., 2013. Conditional convergence in two-dimensional dislocation dynamics. Model. Simul. Mater. Sci. Eng. 21, 055003.
Leibfried, G., 1951. Verteilung von versefzungen im statischen gleichgewicht. Z. Phys. 130, 214–226.
Li, X., Wei, Y., Yang, W., Gao, H., 2009. Competing grain-boundary and dislocation-mediated mechanisms in plastic strain recovery in nanocrystalline
aluminum. Proc. Natl. Acad. Sci. 106, 16108–16113.
Lu, K., Sui, M.L., 1993. An explanation to the abnormal Hall–Petch relation in nanocrystalline materials. Scripta Metall. Mater. 28, 1465–1470.
Masumura, R.A., Hazzledine, P.M., Pande, C.S., 1998. Yield stress of fine grained materials. Acta Mater. 46, 4527–4534.
McClintock, F., Argon, A. (Eds.), 1966. Mechanical Behavior of Materials. Addison-Wesley, Reading, Massachusetts.
Meyers, M.A., Mishra, A., Benson, D.J., 2006. Mechanical properties of nanocrystalline materials. Progr. Mater. Sci. 51, 427–556.
Nicola, L., Van der Giessen, E., Needleman, A., 2005. Size effects in polycrystalline thin films analyzed by discrete dislocation plasticity. Thin Solid Films 479,
329–338.
Nieman, G., Weertman, J.R., Siegel, R.W., 1991. Mechanical behavior of nanocrystalline cu and pd mechanical behavior of nanocrystalline cu and pd
mechanical behavior of nanocrystalline cu and pd mechanical behavior of nanocrystalline cu and pd. J. Mater. Res. 6, 1012–1027.
Nieman, G.W., Weertman, J.R., Siegel, R.W., 1989. Microhardness of nanocrystalline palladium and copper produced by inert-gas condensation. Scripta
Metall. 23, 2013–2018.
Pande, C.S., Cooper, K.P., 2009. Nanomechanics of Hall–Petch relationship in nanocrystalline materials. Progr. Mater. Sci. 54, 689–706.
Peach, M., Koehler, J., 1950. The forces exerted on dislocations and the stress fields produced by them. Phys. Rev. 80, 436.
Petch, N., 1953. The cleavage strength of polycrystals. J. Iron Steel Inst. Lond. 174, 25–28.
Quek, S.S., Wu, Z., Zhang, Y.W., Srolovitz, D.J., 2014. Polycrystal deformation in a discrete dislocation dynamics framework. Acta Mater. 75, 92–105.
Sanders, P., Eastman, J., Weertman, J., 1997. Elastic and tensile behavior of nanocrystalline copper and palladium. Acta Mater. 45, 4019–4025.
Schiøtz, J., Di Tolla, F.D., Jacobsen, K.W., 1998. Softening of nanocrystalline metals at very small grain sizes. Nature 391, 561–563.
Schiøtz, J., Jacobsen, K.W., 2003. A maximum in the strength of nanocrystalline copper. Science 301, 1357–1359.
Shishvan, S.S., Van der Giessen, E., 2010. Distribution of dislocation source length and the size dependent yield strength in freestanding thin films. J. Mech.
Phys. Solids 58, 678–695.
Van Swygenhoven, H., Caro, A., 1998. Plastic behavior of nanophase metals studied by molecular dynamics. Phys. Rev. B 58, 11246–11251.
Van Swygenhoven, H., Derlet, P.M., 2001. Grain-boundary sliding in nanocrystalline fcc metals. Phys. Rev. B 64, 224105.
Wei, Y., Bower, A.F., Gao, H., 2008a. Enhanced strain-rate sensitivity in fcc nanocrystals due to grain-boundary diffusion and sliding. Acta Mater. 56,
1741–1752.
Wei, Y., Bower, A.F., Gao, H., 2008b. Recoverable creep deformation and transient local stress concentration due to heterogeneous grain-boundary diffusion
and sliding in polycrystalline solids. J. Mech. Phys. Solids 56, 1460–1483.
Wu, Z.X., Zhang, Y.W., Jhon, M.H., Srolovitz, D.J., 2013. Anatomy of nanomaterial deformation: grain boundary sliding, plasticity and cavitation in nano-
crystalline Ni. Acta Mater. 61, 5807–5820.
Zhu, B., Asaro, R.J., Krysl, P., Bailey, R., 2005. Transition of deformation mechanisms and its connection to grain size distribution in nanocrystalline metals.
Acta Mater. 53, 4825–4838.

Das könnte Ihnen auch gefallen