You are on page 1of 13

Philosophical Magazine

ISSN: 0031-8086 (Print) (Online) Journal homepage: http://www.tandfonline.com/loi/tphm19

Grain boundary sliding as a deformation


mechanism during creep

Terence G. Langdon

To cite this article: Terence G. Langdon (1970) Grain boundary sliding as a


deformation mechanism during creep, Philosophical Magazine, 22:178, 689-700, DOI:
10.1080/14786437008220939

To link to this article: https://doi.org/10.1080/14786437008220939

Published online: 20 Aug 2006.

Submit your article to this journal

Article views: 407

Citing articles: 211 View citing articles

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=tphm20
Grain Boundary Sliding as a Deformation Mechanism during Creep

By TERENCE G. LANGDON
Department of Metallurgy, The University of British Columbia,
Vancouver, Canada
[Received 16 February 1970 and in final form 1 June 19701

ABSTRACT
A model for grain boundary sliding is developed in which sliding occurs
by the movement of dislocations along, or adjacent to, the boundary by a
combination of climb and glide. Under these conditions the strain rate
due to sliding is proportional to u g / d , where u is the applied stress and d is
the average grain diameter. It is shown that reports in the literature of
enhanced creep rates a t low stresses and/or small grain sizes may be
oxplained by assuming that the various deformation mechanisms, including
sliding, operate independently.

$ 1. INTRODUCTION
ALTHOUGHmany investigations have been carried out to determine the
dependence of the strain rate observed during steady-state creep on stress
and temperature, only limited attention has been devoted to the effect of
grain size. When the average grain diameter is large, typically greater
-
than 0.1 mm, similar steady-state creep rates are obtained under a given
set of experimental conditions for grain sizes differing by more than an
order of magnitude (Cropper and Langdon 1968) ; but a t smaller grain
sizes the creep rate may increase rapidly, due, it has been suggested, to
the increasing importance of grain boundary sliding (Barrett, Lytton and
Sherby 1967). The purpose of this paper, therefore, is t o examine creep
behaviour when more than one deformation process is taking place and, in
particular, t o develop a mechanism for sliding and to compare the pre-
dictions arising from this model with experimental data appearing in the
literature.

$ 2 . THEEQUATION FORSTEADY-STATE CREEP


Creep is a thermally activated process in which the creep rate for the ith
mechanism can be expressed as
s) exp ( - UI(u,T,s ) / k T ) ,. . . . . . ( 1 )
ki=fi(u, T ,
where k is Boltzmann’s constant, and the frequency factorfi and activation
energy U i are both dependent on the applied stress u, the absolute tempera-
ture T, and a ‘ structure term ’ s.
If two or more mechanisms operate under any given experimental
conditions, they may act either sequentially, so that the second only occurs
after the first has taken place and t h e total (or observed) steady-state
P.M. 2Y
690 T. G. Langdon on
creep rate, it,is then determined by the slowest process, or non-sequentially,
so that each occurs simultaneously and kt represents the sum of all of the
individual rates. I n the former case, if two mechanisms have different
activation energies, there is a change in the observed activation energy
for creep, U,, from a high to a low value with increase in temperature ; in
the latter case, the change is from a low to a high value. I n practice, the
values of U , determined experimentally usually remain constant or increase
with increasing temperature, although there are some isolated instances,
ascribed to rapid pipe diffusion, in which there is a decrease. The inference
is therefore strong that mechanisms operate non -sequentially over the
range of strain rates usually observed experimentally, a t least for those
having different activation energies, so that eqn. ( 1 ) can be replaced by
it= zf;(u, T ,s)exp ( - Ui(cr,T,s ) / k T ) ,. . . . . (2)
i

As a first approximation, the summation may be expressed in the form


it=E'g+ ig,,+ i d f f ( l ) +Cdiff(gb), a . * * . (3)
where the individual strain rates are igdue to dislocation processes occurring
within the grains, igbsdue to grain boundary sliding, and idiff(,, and
due to the stress-directed diffusion of vacancies through the lattice
idiff(g,r)
(Nabarro 1948, Herring 1960) and along the grain boundaries (Coble 1963),
respectively. Although other processes may contribute t o it under
certain experimental conditions, the mechanisms given by eqn. (3) appear
to make the major contribution during the normal high-temperature
creep of pure metallic and ionic polycrystals and many solid solution
alloys.
I n order to make use of this summation, an equation is needed for iBbs;
this is formulated in $ 3 , and the implications of eqn. (3) are considered
in 5 4 .

SLIDING
$ 3 . A MODELFORGRAIN BOUNDARY
3.1. The Mechanism of Sliding
Lifshitz (1963) has correctly pointed out that movement a t the grain
boundaries must occur as a natural consequence of diffusional creep, but
it is important to recognize that this merely accommodates the grain
elongation arising from vacancy diffusion, and it makes no direct contribution
to the total strain. Reported examples of a linear relationship between
igbsand u are restricted to very low stress conditions (for example,
u < 6 g/mm2 in zinc bicrystals, when no slip lines intersected the boundary
(Turner 1965)), so that models which treat sliding as a Newtonian viscous
phenomenon appear incapable of accounting for data obtained under the
more normal conditions of high-temperature creep.
Using transmission electron microscopy, it has been shown that there
are only a few isolated dislocations in, or immediately adjacent to, the
grain boundaries of an annealed but unstrained polycrystal, but that the
Grain Boundary Sliding as a Deformation Mechanism during Creep 691
dislocation content increases rapidly with increasing strain (Hale, Ishida,
Lin and McLean 1966, Ishida and Henderson Brown 1967). It is not
immediately clear whether these dislocations come into the boundaries
from the lattice or are formed in the boundary and thence move into the
grains ; whilst both processes probably occur, present evidence suggests
that they primarily originate within the grains and enter the boundary
region by slip (Lin and McLean 1968).
Such dislocations appear to retain their crystallographic Burgers
vectorst, and sliding is therefore possible if they move along the boundary
by alternate climb and glide (Ishida and Henderson Brown 1967) so that
the shear displacement arises from the component of the Burgers vector
parallel to the boundary. Furthermore, since the dislocations are supplied
by deformation within the grains, this leads naturally t o the constant
relationship observed experimentally between sliding and strain (Bell and
Langdon 1969).
Experiments on polycrystals suggest that the activation energy for grain
boundary sliding, Ugbs,is probably close to that for lattice self-diffusion,
UHd(e.g. Ciha, MiliEka and Cadek 1968), and a similar result was reported
from differential temperature tests on aluminium bicrystals (Horton 1969).
Some indirect evidence for the equivalence of Ug,,sand U , , is also obtained
by noting that the strain contribution from sliding usually remains constant
during the early stages of creep. It follows therefore that if U g b S <U , ,
then, under conditions of constant stress, sliding will become less
important with increasing temperature for those materials in which U, is
approximately independent of temperature. Tests on aluminium, in
which U , 2: ZT,, over a wide temperature range, have shown instead that
the sliding contribution remains invariant a t constant stress for tempera-
tures in the range 250-350"c (Davies, Stevens and Wilshire 1966) and
337-474"~ (Fazan, Sherby and Dorn 1954), thereby suggesting that
U g , , sl J~C = Us(,. Although further work is needed t o unequivocally
establish the relationship between Ug,,sand URd,such an equality may
arise if the dislocation movement takes place in a finite zone adjacent to
the boundary.

3.2. The Strain Rate due to Sliding, igbs


If grain boundary sliding occurs by the climb and glide of dislocations,
the rate of sliding is governed by the rate of climb since this is the slower
process. The frequency of climb, v ~ , is
, given by (Friedel 1964) :

v C = p j v ( z - 1)exp ( S J k )exp ( - Cr,/kT){exp (~,!22/kT)- 13, . . (4)


~

t Gleiter, Hornbogen and Baro (1968) and Baro, Gleiter and Hornbogen
( 1968j69) considered that ' grain boundary dislocations ' observed by electron
microscopy had non-crystallographic Burgers vectors lying in the plane of the
boundary : but there are reasons for not accepting this conclusion (Bell and
Langdon 1969).
2YZ
692 T. G. Langdon on

where p , is the probability of finding a jog on a dislocation, v is the vibra-


tional frequency of an atom, z is the coordination number, 8, is the entropy
of activation for diffusion, U , is the energy of activation for diffision, r Z
is the stress required t o induce climb along the boundary, and R is the
activation volume.
The rate of climb along the boundary, s, is
Nbv,,
S= . . . . . . . . . (5)
where N is the number of grain boundary dislocations per unit length, and
b is the Burgers vector. Assuming a uniform distribution of dislocations
along the boundary with a spacing determined by the stress field of a
single edge dislocation gives N N 2 4 1 - p)T/Gb, where p is Poisson's ratio,
T is the shear stress and G is the shear modulus.
If the dislocations move in a zone adjacent to the boundary, so that
U g b s ? u,d, eqn. (4) may be simplified by using the lattice self-diffusion
coefficient, D,, given by
D,-b2vexp(Sd/k)exp(- Ud/kT), . . . . . (6)
where ud =_ Usd.
The shear strain rate in a climb-glide process, j g b s ,controlled by the
rate of climb but producing strain through glide, is
j g b s = MAbS, . . . . . . . . . (7)
where M is the number of boundaries per unit volume and A is the total
area swept out by dislocations moving along a boundary. Since disloca-
tions come into the boundary region from slip on either side of the boundary,
IIf and A can be expressed as 6/.rrd3 and nd2, respectively, where d is the
average grain diameter.
-
Substituting these values into eqn. (7) for T&? < k T , and putting R b3,
T = r, = a/2,and j g b s
= 3 i g b B / 4 , gives :
4rpjb2(1 - p ) ( -~ 1 ) ~ '
E'gbuN D,, . . . . . (8)
dGkT
I n order to make use of eqn. (8)) an assumption is necessary concerning
the value of pj. Electron microscopy has indicated that dislocations in
the boundary region do not dissociate into partials (Ishida and Henderson

,
-
pj=exp ( - U j / E T ) where U , is the jog energy. Taking pi -
Brown 1967)) so that p , is probably close to the equilibrium value given by
(Ardell,
Reiss and Nix 1965), p 0 . 3 , and putting z = 12 for f.c.c. materials, eqn. (8)
reduces to
/3b2a2
igbe= -D,, . . . . . . . . (9)
dGkT
where fi is a constant close to unity.

0 4. STEADY-STATE
CREEPDUE TO NON-SEQUENTIAL
MECHANISMS
Several dislocation mechanisms have been proposed to account for strain
within the grains, but the bulk of present evidence suggests that, whilst
Grain Boundary Sliding as a Deformation Mechanism during Creep 693
existing theories may need to be modified to incorporate such features as
the stacking fault energy of the material, high-temperature creep is best
described by some form of dislocation climb process (Mukherjee, Bird and
Dorn 1969). Taking i, from the theory of dislocation glide controlled by
the rate of climb (Weertman 1957) and substituting into eqn. (3), together
with igbsdefined by eqn. (9) and the exact equations derived for <diff(z)
(Herring 1950) and idiE(g,,)
(Coble 1963), gives
37r2u2 (3)O.5,2.5bl.5
it = 2 (2p 5 ~ 2 b sinh
2 { & ' J ~ ~ 5 ~ 0 Dl~ 5 k T }

where I' is the number of Frank-Read sources per em3, w is the width of
the grain boundary, D,,, is the grain boundary self-diffusion coefficient,
8- 1, and B - 10 for equi-axed polycrystals.
Fig. 1

Diffusion compensated creep rate ( & / D )plotted as a function of stress for


three different grain sizes, for the situation where grain boundary sliding
occurs as an independent mechanism.

The predictions arising from eqn. (10) are shown schematically in fig. 1,
with the total creep rate normalized by dividing by the diffusivity, D ; the
situation is represented for T > 0*6T,, where T , is the melting point in
degrees Kelvin, so that, due to the difference in activation energies,
694 T. G. Langdon on
i,iff(gb)is not significant. The solid lines indicate that one deformation
process is dominant a t any given stress level, and the stress dependence is
divisible into four distinct regions. Considering firstly the line for
d = 0.1 mm, diffusional creep is important a t very low stresses with a stress
exponent of n = 1 (region 1). At slightly higher stress levels, grain boundary

range in which the creep rate is determined almost solely by i, with n 4.5 -
sliding is the dominant mechanism with n = 2 (II), and this leads to a stress

(111). At very high stresses, typically for it/D > lo9cm-2 (see. for example,
Cropper and Langdon 1968),dislocation climb takes place under conditions
where the number of vacancies exceeds that for thermal equilibrium, and
eqn. (10) breaks down ; in region IV, it is proportional to exp ( K u ) ,where
K is a constant.
Figure 1 shows also the effect of decreasing grain size for d = 0.01 mm and
0.005mm, respectively. Since sliding is less dependent on grain size than
diffusional creep, igbsis dominant over smaller ranges of stress with
decreasing d, but these ranges occur at higher stress levels. Furthermore,
since it is represented by the summation of individual strain rates, there
is in practice a gradual transition from one region to the next, as indicated
by the broken lines above the linear portions in fig. 1. These show clearly
the difficulty of revealing region 11 experimentally, especially for very
small grain sizes, except in the unlikely situation where sliding is the
major mechanism over values of i t / D covering several orders of magnitude.
This experimental difficulty is further proven by the broken line lying
below the line for d =0*1mm, which shows the transition from regions I
to 111 when sliding is absent.
Whilst fig. 1 refers specifically to the stress-directed diffusion of vacancies
through the lattice, a similar plot is also possible for diffusion via the grain
boundaries but with an inverse dependence on d3 in region I.

3 5.COMPARISON WITH EXPERIMENTAL DATA


No single investigation has covered a sufficiently large range of stress to
reveal the four regions indicated schematically in fig. 1 ; and the majority
of experiments has been confined solely to region 111. Several studies have
led to the suggestion that the stress exponent changes fairly abruptly from
a high value a t large u to a value close to unity a t small u, but the various
pieces of evidence put forward to support the n = 1 behaviour are not
entirely acceptable (Woodford 1969) and some of these results may cover
a transition regime in which n = 2.
To compare the predictions arising from eqns. (9) and (10) with experi-
mental results appearing in the literature, three setsof datawill beexamined.

5.1. itas a Function of u and d


The creep behaviour of two types of austenitic stainless steel was
investigated over a wide stress range at 7 0 0 ' ~by Lagneborg and Attermo
(1989); their results are given in fig. 2, in which both steels are of the
Grain Boundury Sliding as a Deformation Mechanism during Creep 695
20wt. yo Cr/35 wt. yo Ni type, but steel B has an additional 0 - 5 wt. yo Ti
and Al. For d = 25 pm in steel A, Ct is proportional t o 04.7over a range of
creep rates from 2 x
N %/hr to 2%/hr, corresponding to region I11
in fig. 1, but this linearity breaks down at faster rates (CJD 2 5 x 109cm-2;
region IV). At low rates ( 5 %/hr), gt is proportional to 03 in steel A
and to 02 in steel B (region 11). Measurements of grain boundary sliding
on steel A in region 111 with d = 120pm led to the suggestion that the
behaviour a t low stress levels was due to enhanced sliding ; but no measure-
ments of sliding were taken on specimens of the smaller grain size in region
11.

Fig. 2

I I 1 I I 1 I

Steady-state creep rate versus stress for specimens of two differentsteels tested
at 700"c. The broken lines show predictions arising from diffusional
creep via the grain boundaries (&~(~k,)) and grain boundary sliding
(&), respectively. (Experimental data from Lagneborg and Attermo
(1969).)

To permit a direct comparison with the predicted value of igbsarising


from eqn. (S), two broken lines are shown in fig. 2 for d = 25pm and 100pm,
respectively, where p , = D, is taken from the diffusion data for iron
in an austenitic stainless steel (Smith and Gibbs 1968) and G is calculated
from the value obtained a t 2 9 5 " (Salmutter
~ and Stangler 1960) corrected
for temperature by putting G ( 1 0 0 0 " ~ ) / G ( 0 "=~0.7.
) These lines show a
stress and grain size dependence which is in good agreement with the
696 T. G. Langdon on
experimental results for steel B, and also confirm that the values of cgbs
calculated from eqn. (8) are of a reasonable magnitude.
An anomaly arises, however, when the predicted strain rate is calculated
for grain boundary diffusional creep, as shown by the line for iClirte,,,at
d = 25 pm, where wD,,is taken from data reported for boundary diffusion
(Smith and Gibbs 1968). It is possible that diffusional creep is inhibited
in these materials because the grain boundaries are not perfect sinks and
sources for vacancies (e.g. Ashby 1969).

5.2. it as a Function of d at Constant u

-
The dependence of it on grain size is shown in fig. 3 for specimens of
polycrystalline copper tested at 496"c ( 0*57T,) under a constant stress
of 2.07 x 10*dyne/cm2(Barrett et al. 1967). The broken line shows the

Fig. 3

0 01 02 0 3 04
d (mm)
Steady-state creep rate versus grain size for specimens of polycrystalline copper
tested at 496"c under a constant stress of 2.07 x 108 dyne/cm2. The
broken line shows the authors' division of the data into two distinct
regions. The solid line shows the predicted trend if it is assumed that
& =2.00 x min-l at d =0.4 mm and the contribution of sliding to
the total strain is then 8%. (Experimental data from Barrett et al.
(1967).)

authors' division of the data into two distinct regions, so that a second
mechanism, postulated as grain boundary sliding, is only significant for
d s 0 . 1 mm.
The solid line shows the predicted dependence of tt on d calculated by
assuming that it = 2.00 x min-1 and i,,,/i, = 0.08 a t d = 0.4 mm, and
neglecting any contribution from diffusional creep. This calculated trend
Grain Boundary Sliding as a Deformation Mechanism during Creep 697
is in excellent agreement with the experimental points, and, furthermore,
the predicted values for igbs/it of 0.15, 0.26 and 0.46 a t d - 0 . 2 , 0.1 and
0.04 mm, respectively, are similar to those obtained experimentally for
from offset measurements (0.13, 0.17 and 0.34, respectively, where E
is the strain for a given process, and E , , , ~ / E ~ = i,,,/i, if the individual
mechanisms contributing to the deformation process operate non-
sequentially).

5.3. i,,,,/i, as a Function of a and d


Detailed measurements were carried out on a magnesium alloy
(Mg-0.78wt. yo Al) to determine the importance of grain boundary sliding
a t 2 0 0 " (~ 0*51T,,)(Bell and Langdon 1967, Langdon and Bell 1968).
N

The experimental points obtained a t et = 0.01 are shown in fig. 4 for grain
sizes of 3.8 x
N 1.8 x and 9 x cm, respectively, plotted as
cgbs/ety0 versus a.
A direct comparison with the predictions of eqn. (10) is made possible by
assuming that one of the experimental points is correct, and then using
this point to calculate the resultant dependence on stress and grain size
over the range covered experimentally. The lines shown in fig. 4 were
calculated by applying the first two terms of eqn. (10) a t the point
) d = 3.8 x
egbs/et = 30% ( = i g b 8 / i tfor cm a t a = 2.76 x lo8dyne/cm2.
The predicted curves are in excellent agreement with the experimental

Fig. 4

The contribution of sliding to the total strain (cgbs/'t) versus stress for a
magnesium alloy testfedat 2 0 0 " ~ . The curves are obtained by assuming
Egbs/q = 30% for d= 3.8 x cm at = 2-76 x 10s dyne/cm2, and
(I

neglecting the contribution from diffusional creep. (Experimental data


from Bell and Langdon (1967).)
698 T. G. Langdon on

levels where the theoretical prediction overestimates Egb8/Et by 15%.


To avoid the implication that E g h S / E t + l O O %
-
points, with the exception of the smallest grain size a t the lower stress

as u+O, it is necessary to
also include the contribution from id,E(o and ic,iK(gl,).
Two experimental
observations suggest a reasonable assumption for the importance of
diffusional creep a t d = 9 x cm and u = 1-38 x lo*dyne/cm2. Firstly,
Fig. 5

0 2 4 6 8
o- (10' dynelcm')

The.contribution of sliding to the total strain (Egbs/ct) versus stress for the
experimental conditions of fig. 4, corrected by including the contribution
from diffusional creep. The predicted trends for two additional grain
sizes are also shown.
under these conditions the calculated value for idif(g,,, was slightly less
than an order of magnitude slower than the observed creep rate. Secondly,
tests on pure magnesium a t 270"c have shown that, with d = 5 x lO-3cm,
diffusional creep via the grain boundaries is the dominant mechanism for
u 5 2.8 x 10' dyne/cm2 (Jones 1965); compensating for the difference in
temperature, this suggests that is dominant for d = 9 x cm a t

-20, it was assumed that iditf(gb,/it


i,liqg,,)/i~ar(l) -
0 5 1.9 x 10' dyne/cm2. Since the theoretical equations also predict
0.15 and i,itf(o/it0.01.
The theoretical curves arising from this assumption are given in fig. 5 for
-
the three grain sizes of fig. 4, and for d = 5 x 10-3 and 3 x 10-3cm,
respectively ; for simplicity, the experimental points are not included.
The curves of fig. 5 show that Egbs/Et increases with decreasing grain
size under conditions of constant stress a t the high stress levels, but this
trend is reversed a t low stresses. Furthermore, the stress a t which
Egbs/ct reaches a maximum decreases with increasing grain size, although
the maximum value attainable is then increased due to the limited
importance of diffusional creep in large-grained material. These points
Grain Boundary Sliding as a Deformation Mechanism during Creep 690
also follow from fig. 1 when it is noted that sliding is the dominant
mechanism over a wider range of stress for the larger grain sizes.

0 6. DISCUSSION
An essential feature of this analysis is that it assumes the various
deformation mechanisms t o operate non -sequentially. This requires that
each takes place independently of the other, and that, under limiting
conditions, each is capable of accounting for the total strain when there is no
contribution from any other process. Clearly, this condition is usually
satisfied for dislocation movement within the grains, and for diffusional
creep when grain movement takes place to maintain coherency across the
grain boundaries (Lifshitz 1963).
The situation for sliding due to dislocation movement is less obvious, but
it is apparent that, on a purely geometric basis, sliding cannot take place in
the absence of any other process, since stress concentrations will arise
both a t triple points and a t irregularities along the boundary ; further,
it may be argued that egbR can never exceed eg, so that there is an upper
limit on egbs/ct of -0.50. This may not be strictly correct, however,
away from since sliding involves movement of material both towards and
any triple point, and it is possible that stress concentrations are relieved,
a t least in part, by some localized re-adjustment through diffusion. I n
addition, the larger values of egbs/ct reported in the literature appear to
occur preferentially in those materials, and under experimental conditions,
where grain boundary cavitation is extensive (for example, the magnesium
alloy of 5 5.3), so that complete coherency a t the boundaries is not a rigid
requirement. Unfortunately, the experimental data of fig. 4 are not
sufficient to reveal a possible upper limit for egbs/ct, although the highest
value of 0.75 may include an erroneous contribution of up to 0.16 due to
the accommodation by grain movement of diffusional creep, and this
gives some credence to a levelling off near Egbs/ct 0.60. At values of
N

cgbs/et 5 0.50: the close agreement between experimental data and


theoretical predictions (figs. 3 and 4) suggests that it is then reasonable t o
consider the deformation mechanisms to operate non -sequentially ; but
more results are needed to decide whether this situation applies under all
conditions.
9 7. CONCLUSIONS
1 . If grain boundary sliding takes place by the movement of dislocations
along, or adjacent to, the boundary, by a combination of climb and glide,
the strain rate due to sliding is proportional to u2/d,where u is the applied
stress and d is the average grain diameter.
2. By assuming that sliding occurs as an independent process, good agree-
ment is obtained with results in the literature showing enhanced creep rates
a t low stresses and/or small grain sizes.
3. If the various deformation mechanisms operate non-sequentially, it
is shown that (i) a t constant stress, the contribution of grain boundary
700 O n Grain Boundary Sliding as a Deformation Mechanism during Creep

sliding to the total strain (cgbs/ct) increases with decreasing grain size a t
high stresses, but this trend is reversed a t low stress levels, and (ii) the
maximum value attainable by E g b s / q increases with increasing grain size,
although the stress a t which this occurs is then decreased.
4. More experimental data are needed to determine whether very high
values of c g b s / q ( ~ 0 * 6 0occur
) in practice, or whether the deformation
processes operate sequentially, under some conditions a t least, to give an
upper limit t o egbs/et.

ACKNOWLEDGMENTS
This work was carried out in part a t the Cavendish Laboratory,
Cambridge, with support from the Science Research Council, and in part
a t the University of British Columbia with support from Atomic Energy
of Canada, Limited.

REFERENCES
ARDELL,A. J., REISS,H., and NIX, W. D., 1965, J . appl. Phys., 36, 1727.
ASHBY,M. F., 1969, Scripta metall., 3, 837.
BARo, G., GLEITER,H., and HORNBOGEN, E., 1968j69, Muter. Sci. Engng, 3,92.
BARRETT, C. R., LYTTON, J. L., and SHERBY, 0. D., 1967, Trans. metall. SOC.
A . I . M . E . , 239, 170.
BELL,R. L., and LANGDON, T. G., 1967, J . Muter. Sci., 2, 313 ; 1969, Interfaces
Conference, edited by R. C. Gifkins (Sydney : Butterworths), p. 115.
, MILI~KA,K., and CADEK,J., 1968, Kovove' materialy, 6, 313.
C ~ H AK.,
COBLE,R. L., 1963, J . appl. Phys., 34, 1679.
CROPPER,D. R., and LANGDON, T. G., 1968, Phil. Mag., 18, 1181.
DAVIES,P. W., STEVENS, R. N., and WILSHIRE, B., 1966, J . Inst. Metals, 94,49.
FAZAN, B., SHERBY, 0. D., and DORN,J. E., 1954, Trans. metall. SOC.A . I . M . E . ,
200, 919.
FRIEDEL, J., 1964, Dislocations (Oxford : Pergamon Press), p. 104.
GLEITER,H., HORNBOQEN, E., and B&o, G., 1968, Acta metall., 16, 1053.
HALE,K. F., ISHIDA, Y., LIN, T. L., and MCLEAN,D., 1966, Sixth International
Congress for Electron Microscopy, Kyoto, edited by R. Uyeda (Tokyo :
Maruzen), p. 295.
HERRING, C., 1950, J. appl. Phys., 21,437.
HORTON, C. A. P., 1969, Scripta metall., 3, 253.
ISHIDA, Y., and HENDERSON BROWN, M., 1967, Acta metall., 15, 857.
JONES, R. B., 1965, Nature, Lord., 207, 70.
LAGNEBORG, R., and ATTERMO,R., 1969, J. Mater. Sci., 4, 195.
LANQDON, T. G., and BELL,R. L., 1968, Trans. metall. SOC.A . I . M . E . ,242,2479.
LIFSHITZ,I. M., 1963, Soviet Phys. J E T P , 17, 909.
LIN, T. L., and MCLEAN,D., 1968, Metal Sci. J . , 2, 108.
MUKHERJEE, A. K., BIRD, J. E., and DORN,J. E., 1969, Trans. A m . SOC.
Metals, 62, 155.
NABARRO, F. R. N., 1948, Report of a Conference on the Strength of Solids
(London : The Physical Society), p. 75.
SALMUTTER, K., and STANGLER, F., 1960, A. Metallk., 51, 544.
SMITH,A. F., and GIBBS,G. B., 1968, Metal Sci. J . , 2,47.
TURNER, P. A., 1965, Ph.D. Thesis, University of London.
WEERTMAN, J., 1957, J . appl. Phys., 28, 362.
WOODFORD, D. A., 1969, Mater. Sci. Engng, 4, 146.

You might also like