Sie sind auf Seite 1von 4

Scripta Materialia 55 (2006) 739742

www.actamat-journals.com

A model for the grain size dependent work hardening of copper


C.W. Sinclair,a,* W.J. Poolea and Y. Brechetb
a
Department of Materials Engineering, The University of British Columbia, 309-6350 Stores Road,
Vancouver BC, Canada V6T 1ZA
b
LTPCM, INP de Grenoble, BP75, 38402 St. Martin dHeres, France
Received 7 March 2006; revised 8 May 2006; accepted 15 May 2006
Available online 2 August 2006

A model is proposed for the low temperature tensile response of copper polycrystals with grain sizes in the range of 250 lm. The
initial work hardening behaviour is strongly grain size dependent and is considered to arise from a combination of kinematic and
isotropic hardening due to dislocationgrain boundary interactions. At larger strains, the inuence of grain size on work hardening
disappears owing to processes associated with dynamic recovery at boundaries.
 2006 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Copper; Work hardening; Grain boundaries; Internal stresses

In recent work on ne grained copper with D > considering the evolution of mobile and forest disloca-
1 lm [1] it was shown that the grain size dependence tion densities. No physical argument was given for the
of the tensile response could be dierentiated into scale dependence of the adjustable parameters control-
strongly grain size dependent and weakly grain size ling dislocation trapping.
dependent regimes in agreement with previous studies In this paper we propose a physically based frame-
[2]. Strong grain size dependence was found to exist over work for investigating these grain size eects on yield
the initial 5% tensile strain. Beyond this, the work and work hardening, in the spirit of the internal
hardening rates, when plotted versus the ow stress, variable models. The aim of our work is to explain
converged to that for the coarsest grain size. Within both the dependence of the initial work hardening
the grain size dependent regime, samples with grain sizes behaviour with the grain size, and the lack of depen-
D < 5 lm exhibited a transient low work hardening rate dence of the large strain work hardening behaviour with
(viewed as an inection in the stressstrain curve). This respect to the grain size.
feature is also evident in recent work on copper [3,4] and We consider that, in the scale independent limit, the
silver [5] with grain sizes of 15 lm. The paradox to be ow stress and work hardening rate of a material
understood in this case is the strong grain size eect at may be described based on the evolution of a single
yield and its progressive disappearance with increasing parameter; the dislocation density. In this regard, we
deformation. follow Kocks and Mecking [7,8] and write the net rate
It has been proposed that the inection in the ow of dislocation storage with imposed shear strain as,
stressstrain curve of the copper is attributable to a lack oq  p 
of forest dislocation storage within ne grains. Disloca- M k1 q  k2q 1
tions are envisioned to pass through the grain interiors oe
and become incorporated/trapped at grain boundaries where M is the Taylor factor. The terms k1 and k2 are
[4,6] with little statistical trapping within grains. This parameters directly linked to the athermal work harden-
argument does not readily explain, however, the conver- ing limit and dynamic recovery respectfully, and can be
gence towards a grain size independent work hardening determined from coarse grained samples or single crys-
rate at large strain. An alternative explanation applied tals. When Eq. (1) is coupled to the Taylor relation,
to silver polycrystals [5] argued that the inection could p
be described by a two parameter state variable model r1 r0 aMlb q 2
with a a constant, l the shear modulus and b the magni-
* Corresponding author. E-mail: chad.sinclair@ubc.ca tude of the Burgers vector, the grain size independent

1359-6462/$ - see front matter  2006 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.scriptamat.2006.05.018
740 C. W. Sinclair et al. / Scripta Materialia 55 (2006) 739742

work hardening rateow stress relationship is uniquely of dislocations which can be stored in the vicinity of the
dened. It has been shown conclusively that such models grain boundary, so that the grain boundary eectively
are suitable to describe accurately the stress strain becomes transparent at larger strains. In principal, this
behaviour of coarse grained copper in the absence of could be viewed as a separate process controlled by a
stage IV [7,8]. critical number of sites per slip band at the boundary,
We consider that grain size enters the problem in two nA . Thus the grain boundary eciency for generating
ways. The dislocations stored in the vicinity of the back stresses is controlled by the parameters nS (screen-
boundary will contribute both to forest hardening giving ing) and the grain boundary eciency for dislocation
an isotropic hardening contribution already included in storage is controlled by nA (recovery) . Given the level
later versions of the KocksMeckingEstrin models of detail considered in this model and our uncertainty
(e.g., [8,9]) and to the building up of back stresses regarding the specic values of these parameters, we
(giving a kinematic hardening contribution). At small take a simplied approach and consider that both
strains, grain boundaries act as perfectly opaque barri- screening and recovery are dictated by the same critical
ers to dislocations. Dislocations are stopped at the grain number of dislocations at the boundary, i.e.,
boundary causing a backstress to develop impeding the
progress of similar dislocations. In its simplest form this nS nA n 6
backstress may be written as,
Finally, it is necessary to consider an evolution law for
lb the strain dependence of the number of dislocations
r2 M n 3
D stopped at a grain boundary. The ux of dislocations
where n is the number of dislocations that have been arriving at a grain boundary per slip band can be calcu-
stopped at the boundary on a given slip band. In this lated as,
formulation, the grain is considered as isolated, which
is of course not the case in reality. We modify Eq. (3) dn k  n
1  7
to account for the fact that the net long range stresses de b n
predicted by (3) will be reduced as dislocations of oppo-
site sign arrive at the boundary in adjacent grains or on where k is the mean spacing between slip lines at the
other slip systems thereby screening the stress eld from grain boundaries. The ratio k/b gives the number of
the dislocations already at the boundary. Since we as- dislocations per slip band geometrically necessary to
sume for the sake of simplicity that dislocations arrive provide the deformation, and the corrective term
randomly at a grain boundary, even when the average (1  n/n*) accounts for the nite number of sites avail-
number is the same on both sides, only a proportion able for dislocations at the boundary. The magnitude
of them (the ones facing each other) will be eciently of n* will be controlled by the dominant mechanism of
screened. A simple physical argument suggests that once relaxation at the boundary. This process is of increasing
the number of dislocations at the boundary on a given importance as the number of dislocations at the bound-
slip band reaches a critical value (nS ) the long range ary approaches n*. When n ! n* the eective storage
stress eld will be completely screened. Consider single eciency at the boundaries becomes zero.
slip with nS sites per slip band available for dislocations Of course, the process outlined above will not be per-
at the grain boundary. For a dislocation occupying a site fectly ecient. In order to account for the higher rate of
on one side of the grain boundary, the probability that it dislocation storage in ne grained materials, we consider
is faced by a dislocation that will screen it is, that only a portion of any dislocation line is likely to
n reach the grain boundary. The other portions of the line
P  4 may extend away from the grain boundary where they
nS will act as forest dislocations as described by Saada
The dislocations stored at the grain boundary will have [10]. This leads to a modication of the overall evolution
a backstress eect with a probability of 1-P. Combin- equation for dislocation density (Eq. (1)) by adding an
ing this with Eq. (3) gives the net polarized stress arising additional grain size dependent storage term due to
from long range back stresses, trapping of dislocations at the grain boundary,
 

  oq p 1 n
lb n M k1 q  k2q k3 1  8
r2 M n 1   5 oe bD nA
D nS
where k3 accounts for the eciency of the dislocations at
When n reaches the critical value nS , the eciency of the grain boundary with respect to forest hardening. The
screening is complete and the long range back stress grain size dependent storage term in Eq. (7) is multiplied
disappears. by a corrective term (1  n/n*) in a parallel fashion to
The storage of dislocations of dierent sign on both Eqs. (5) and (6). The physical interpretation is the same
sides of the boundary leads to a high probability of as that above, i.e., that dynamic recovery at the bound-
annihilation (or a form of dynamic recovery) at the ary limits the storage of additional forest dislocations
boundary. Another possibility is that the stresses gener- within the grain. The mechanism we envision is that
ated could force the dislocations to be absorbed in the annihilation of a part of a loop in a grain boundary will
grain boundary and emitted from the grain boundary lead to reorganization and disappearance of the con-
when a critical stress is achieved. One could view both necting parts within the grains. In principle only a pro-
processes as a steady state which would limit the number portion will disappear in this reorganization but to limit
C. W. Sinclair et al. / Scripta Materialia 55 (2006) 739742 741

the number of adjustable parameters, we have assumed


a complete removal hence the last term in Eq. (8).
Eqs. (2), (5), (7) and (8) can now be combined to give
an overall expression for the ow stress of the material,
which includes both isotropic and kinematic contribu-
tions to the hardening,

lb  n p
r r0 M n 1   alb q 9
D n
It can be shown that the kinematic contribution to the
ow stress described by Eq. (5) reaches a maximum
(r*) at a critical strain (e*) when n = n*/2, which when
substituted into Eq. (5) and the integrated form of Eq.
(7) gives,
lb n
r M 10a
D 4 (a)

nb
e ln2 10b
k
Beyond this maximum the two contributions to the ow
stress evolve in opposite directions. The net shape of the
stress strain curve then depends on the sum of the hard-
ening given by the dislocation storage term and the soft-
ening coming from the loss of the backstress. As the
grain size decreases the later becomes more important
thus causing the stress strain curve undergo an inec-
tion. As n tends to n* both the backstress term and the
grain size dependent dislocation storage term disappear
and the ow stress becomes independent of grain size
and thus controlled by Eqs. (1) and (2).
The tensile response for copper at 298 K predicted by
Eq. (9) is compared in Figure 1 with experiment. The
values of k1 and k2 were xed by tting Eqs. (1) and (b)
(2) to the work hardening response of material with
D = 50 lm. The parameters n*, k3 and k were adjusted Figure 1. Comparison between model predictions (using parameters in
to provide a best t to experimental data for copper with Table 1) and experimental data on copper from [1] with grain sizes
grain sizes between 2 and 50 lm tested at room temper- between 2 and 50 lm. (a) Comparison between ow curves for three
ature (details about the material can be found elsewhere grain sizes. (b) Comparison with respect to work hardening rate as a
function of ow stress.
[1]). Values of n* and k which determine the shape of the
backstress were selected rst. The value of k3 is then dic-
tated by the stress at which the work hardening rate-ow
stress (Hr) curves coincide. At this point, the grain size Table 1. Values of parameters used to compare with experimental data
dependence disappears and the dislocation density ac- for copper
crued to that point must be equivalent to that for a Parameter Value
coarse grained sample deformed to attain the same ow Physical constants
stress. The best t values of n*, k3 and k were subse- k1 125 106 m1
quently found by iteratively modifying the three param- k2 4.1
eters in this sequence and are listed in Table 1. Figure 2 a 0.3
shows the separate contributions (kinematic and isotro- r0 20 MPa
pic) to the overall ow stress for the three grain sizes M 3.06
used to construct Figure 1. l 42 GPa
b 0.3 nm
The behaviour predicted by Eq. (9) captures the basic
features of the tensile response including the inection in Adjustable parameters (tted to experimental data)
the work hardening response and the grain size depen- k3 4.4
dence of the stress and work hardening rate at the point k 413 nm
of inection. Figure 1a illustrates the t to three grain n* 6.7
sizes (D = 50 lm, 12 lm and 2.1 lm). At suciently
large values of stress, the work hardening rate becomes to coincide with the point at which the Hr plot be-
grain size independent. A sensitive measure of the accu- comes grain size independent as illustrated by the inset
racy of the model is to compare against the ow stress in Figure 3.
(r0) and work hardening rate (H0) at the point of inec- Implicit in the above description is the assumption
tion as shown in Figure 3. For large grain sizes where no that n* and k are constant, i.e., independent of strain
inection is observed, the values of r0 and H0 were taken and grain size. The constant value for k is taken in place
742 C. W. Sinclair et al. / Scripta Materialia 55 (2006) 739742

of hardening is justiable if plastic relaxation is imper-


fect in the vicinity of the boundary as has been described
with reference to dispersion hardened materials [11,12].
Evidence for a grain size dependent ow stress contribu-
tion in copper polycrystals consistent with the form
given by Eq. (5) can be seen in the work of Schankula
et al. [13]. In this work the ow stress was separated into
grain size dependent and independent components via
stress relaxation tests. The results of these experiments
indicate that the grain size dependent component of
the ow stress decreases to a very low value at a strain
of approximately 0.05 for grain sizes of 29 and 55 lm.
This is qualitatively consistent with the current results
that predict the disappearance of the grain size depen-
dence of the ow stress at plastic strains of several
percent.
Figure 2. Illustration of the proposed contributions to the total ow A model has been introduced that predicts a strong
stress for two dierent grain sizes (D = 2.1 lm and D = 12 lm). Not grain size dependence over the initial portion of the
shown is the constant intrinsic lattice resistance, r0. For grain sizes tensile response of copper. The grain size dependence
much larger than 12 lm the backstress contribution becomes negligible
is assumed to arise due to the storage of dislocations
compared to the forest hardening.
at grain boundaries. The eect of grain boundaries is as-
sumed to manifest in two eects; (i) kinematic hardening
due to the development of long range internal stresses
and (ii) the storage of additional forest dislocations that
contribute to an enhanced level of isotropic hardening.
The inuence of the structure developed due to grain
boundaries is assumed to disappear at larger strains
owing to dislocation screening and dynamic recovery ef-
fects at grain boundaries. Applying the model to exper-
imental data for copper with grain sizes down to 2 lm
reveals very good agreement. Further information on
the behaviour of materials with dierent stacking fault
energies and with grain sizes between 0.1 and 10 lm,
Figure 3. (a) Comparison of experimental and predicted position of
the inection in the initial portion of the ow curve as described by the
as well as Baushinger tests allowing for a decoupling be-
parameters H0 and r0 (dened in the inset gure). (b) Predicted and tween kinematic and isotropic contributions to the work
observed dependence of the ow stress at inection and grain size. hardening, would allow for a critical evaluation of the
model.

of a more complex description of the poorly understood [1] D.J. Lloyd, Met. Sci. 14 (1980) 193.
variation of slip line spacing with stress or strain. Given [2] X. Haung, N. Hansen, Mater. Sci. Eng. A387389 (2004)
the small range of plastic stain (<0.05) over which the 186.
grain boundaries contribute to the ow stress in this [3] J.E. Flinn, D.P. Field, G.E. Korth, T.M. Lillo, J.
Macheret, Acta Mater. 49 (2001) 2065.
model, this is not an unreasonable approximation. The [4] C.W. Sinclair, W.J. Poole, Work hardening of ne grained
constant value of n* is consistent with pile-ups of materials, Ultrane Grained Materials III, Charlotte,
dislocations on closely spaced slip bands at grain bound- NC., United States, Minerals, Metals and Materials
aries. The fact that a critical number of dislocations (and, Society, Warrendale, United States, 2004. p. 59.
for example, not a critical density) describes the experi- [5] R.P. Carreker, W.R. Hibbard, Acta Metall. 1 (1953) 654.
mentally observed behaviour suggests that dynamic [6] K.J. Al-Fadhalah, Ph.D. Thesis, University of Illinois at
recovery at the grain boundaries is controlled by a criti- Urbana-Champaign, 2004.
cal stress coming from the accumulation of dislocations [7] U.F. Kocks, H. Mecking, Prog. Mater. Sci. 48 (2003)
that could trigger some structural re-shuing within 171.
the boundary rather than by a critical distance between [8] H. Mecking, U.F. Kocks, Acta Metall. 29 (1981) 1865.
[9] H.S. Kim, Y. Estrin, M.B. Bush, Mater. Sci. Eng. A316
dislocations below which they would annihilate. (2001) 195.
The proposed model requires some combination of [10] G. Saada, Phil. Mag., in Press.
grain size dependent hardening coming from the back- [11] L.M. Brown, W.M. Stobbs, Phil. Mag. 23 (1971) 1185.
stress and forest hardening contributions. Though most [12] L.M. Brown, W.M. Stobbs, Phil. Mag. 23 (1971) 1201.
models for grain size dependent work hardening con- [13] M.H. Schankula, D.J. Lloyd, J.D. Embury, Acta Metall.
sider only the latter mechanism, inclusion of both forms Mater. 18 (1970) 1293.

Das könnte Ihnen auch gefallen