Mechanical properties of undoped GaAs.

II: The
brittle-to-ductile transition temperature
Shanling Wang, Pirouz Pirouz
*
Department of Materials Science and Engineering, Case Western Reserve University, Cleveland, OH 44106-7204, USA
Received 2 March 2007; received in revised form 12 June 2007; accepted 13 June 2007
Available online 13 August 2007
Abstract
In this second part of a series of papers on the mechanical properties of GaAs, direct determination of the brittle-to-ductile transition
temperature T
BDT
of the same crystal as that used for compression experiments (see part I) is reported. The experimental technique
employed for this purpose is four-point bend testing of pre-cracked samples at dierent temperatures and strain rates. It is found that,
as in other semiconductors, T
BDT
of GaAs is sharp, and is very sensitive to and increases with the strain rate from 300 to 380 C for the
strain rate ranging from _ e 1 10
6
s
1
to _ e 5 10
5
s
1
. From the variations of T
BDT
with the strain rate _ e, the activation enthalpy
DH
d
for dislocation glide in undoped GaAs was determined to be 1.36 0.02 eV, a value very close to that reported for the slow b dis-
locations in such a crystal.
2007 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: Semiconductor compound; Bending test; Dislocation structure; GaAs
1. Introduction
As described in part I of this series [1], compression tests
at various strain rates on undoped GaAs exhibit a critical
temperature T
c2
at which there is an abrupt change in the
slope of the linear plot of ln(s
y
) vs. 1/T. We had already
found a similar behavior for the wide bandgap semicon-
ductor 4HSiC, where the intriguing point emerged that
the transition temperature T
c2
in the plot of the yield stress
s
y
vs. temperature T was in the same range as the brittle-to-
ductile transition (BDT) temperature T
BDT
in that semi-
conductor [2,3]. This observation was later veried by
direct measurements of T
BDT
of the same 4HSiC crystals
at dierent strain rates [4]. Based on these results, a new
model for the occurrence of BDT in semiconductors, the
magnitude of T
BDT
and its dependence on the strain rate
_ e was proposed [5]. It would thus be interesting to investi-
gate whether the T
c2
in GaAs, measured in part I [1], also
corresponds to the T
BDT
of this material. To the authors
knowledge, except for the indirect measurements of Fujita
et al. [6], there are no reports on direct determination of the
brittle-to-ductile transition temperature and its strain rate
dependence for this important semiconductor. For this
purpose, the T
BDT
of the same undoped GaAs crystal as
that used in part I [1] has been measured at dierent strain
rates using the technique of four-point bend testing [4,7].
Combining the results of this paper with the microstruc-
tural changes that were observed by transmission electron
microscopy (TEM), an attempt is made to correlate the
transition from brittleness to ductility to the changes that
take place in the type and character of dislocations that
control plastic deformation of GaAs in dierent tempera-
ture and stress regimes.
GaAs, like most other cubic compound semiconductors,
has a zincblende structure based on a face-centered cubic
lattice with a 110{111} slip system and a {110} cleavage
plane. Because of the polar nature of GaAs, there are two
types of non-screw dislocations in the crystal, commonly
known as a and b dislocations [8]. These are distinguished
by the nature of the atoms constituting the terminating
1359-6454/$30.00 2007 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2007.06.026
*
Corresponding author.
E-mail address: pxp7@cwru.edu (P. Pirouz).
www.elsevier.com/locate/actamat
Acta Materialia 55 (2007) 55155525
edge of the extra half plane. Conventionally, for disloca-
tions on the shue plane, if the edge consists of Ga atoms,
the dislocation is labeled a [or Ga(s)], and if it consists of
As atoms, it is labeled b [or As(s)] [9]; conversely, for dislo-
cations on the glide plane, a and b dislocations would cor-
respond to As(g) and Ga(g), respectively. There are many
experimental results in the literature that indicate widely
dierent mobilities for a and b dislocations in GaAs [10
13]. Since most mechanical properties of a crystal are
strongly inuenced by dislocation mobility, it would be
expected that the polarity of GaAs would also aect its
brittle-to-ductile transition behavior.
2. Experimental
The four-point bend technique of pre-cracked bar-
shaped samples to measure the brittle-to-ductile transition
temperature T
BDT
was rst employed by Samuels in silicon
[7]. More recently, Zhang et al. [4] used this technique to
measure T
BDT
in 4HSiC. For our four-point bend exper-
iments on GaAs, two types of bar-shaped samples with dif-
ferent orientations were prepared, shown in Fig. 1a and b.
These two sample types were chosen because in the sample
orientation shown in Fig. 1a, b dislocations are expected to
activate on inclined (111) and

11 slip planes when a

1

10 Knoop indent is made on the (001) crystal face. Con-

versely, for the sample orientation in Fig. 1b, a dislocations
are expected to be activated on inclined 1

1 and

11

1
slip planes when a [110] Knoop indent is made on the
(001) crystal face. The samples were in the form of
35 3 1 mm
3
parallelepipeds and their orientation was
such that the tensile stress on the {110} cleavage plane is
maximized while maintaining a reasonable resolved shear
stress on the {111} primary slip plane. Thus in one set of
samples (Fig. 1a), the 35 3 mm
2
top and bottom faces
of the sample were cut parallel to the (001) plane, the
35 1 mm
2
side faces were parallel to the 1

10 plane
and the 3 1 mm
2
end faces were parallel to the (110)
plane, while in the other set (Fig. 1b), the side faces were
parallel to (110) and the end faces were parallel to
1

10. The four-point bend jig and the cylindrical rollers

were made from molybdenum (Fig. 2). The rollers were
polished to give a surface nish comparable to that of
the GaAs samples. In the tests, the inner and outer rollers
of the jig were placed on the opposite (001) faces of
the sample; the bending arm d given by half the dierence
between the outer and inner rollers
1
2
L l was 10 mm.
To convert the values of load and displacement into stress
and strain, the equation given by Bruneau and Pratt [14]
for an elastic beam in a four-point bending conguration
was used. According to this equation, the normal stress
r
app
applied to the (110) end faces of the sample is given by
r
app
3Pd=wh
2
1
where P is the applied load in Newtons, and w and h are the
width and thickness of the sample, respectively, in
millimeters.
The strain rate in the central portion of the beam is
_ e
6h
_
d
3lL l L l
2
4h
2
3mm
1mm
precracks
line of contact between
outer roller and sample
surface
35mm
[001]
[110]
[-110]
[1-10]
[110]
[001]
Fig. 1. Schematic of sample dimensions and orientations.
x
2
P/2 P/2
P/2 P/2
h x
1
d
l
L
Fig. 2. Schematic of the four-point bending jig and geometry.
5516 S. Wang, P. Pirouz / Acta Materialia 55 (2007) 55155525
where
_
d is the crosshead displacement rate (s
1
), and L and
l are the separation between the outer and inner rollers,
respectively, in millimeters.
Before testing, the samples were polished with dierent
grit Al
2
O
3
powders (9, 5, 1 and 0.03 lm) on polishing pads
until a mirror-like surface was achieved. Five pre-cracks
were introduced using a Knoop indenter with a load of
50 g at room temperature. The samples were subsequently
annealed for 1 h at 300 C to relax the residual stress; a typ-
ical Knoop indent on one of the four-point bend samples is
shown in Fig. 3a.
After loading the sample, the four-point bend jig was
connected to the crosshead of a 1361 Instron servohydrau-
lic machine where the jig was enclosed in a ceramic tube in
which high-purity argon was own to prevent oxidation of
the sample and jig during high temperature tests.
Following the tests, a few of the deformed samples were
selected for TEM examination. For this purpose, thin foils,
parallel to the (110) cleavage plane of the deformed sam-
ples, were cut, polished and ion sputtered to electron trans-
parency. The TEM examination was conducted in a Philips
CM20 at an operating voltage of 200 kV.
3. Results
3.1. Basic characterization of the brittle-to-ductile transition
in GaAs
Initial tests were carried out at a strain rate of
_ e 1 10
6
s
1
over a range of temperatures below and
above the brittle-to-ductile transition. Fig. 4 shows three
typical loaddisplacement curves for samples tested below,
near, and above the transition temperature T
BDT
.
3.1.1. Brittle behavior
A typical loaddisplacement curve at a temperature in
the brittle regime (T < T
BDT
) of GaAs is shown by curve
(a) in Fig. 4. At this temperature, the stress/strain behavior
is linear all the way up to the point where the sample failed
catastrophically by fracture. In this regime, the GaAs sam-
ples typically fractured under a load of about 12 N. This
sort of behavior was observed for all temperatures below
the transition temperature T
BDT
. It should be emphasized
that the critical value of load at which failure occurred
(12 N) was the same for all temperatures up to T
BDT
.
The range T < T
BDT
thus denes the brittle regime for
the undoped GaAs under consideration; the linearity
of the loaddisplacement curve indicates that the material
responds elastically to the applied stress until it fails. In
all cases the failure originated at one of the ve pre-cracks
introduced by Knoop indentation and the cleavage plane
was the {110} plane containing that pre-crack. Fig. 3b
shows a typical fracture surface in the brittle regime: a ser-
ies of radial markings begin at the crack front and extend
towards the sample boundaries. The radial markings are
shallow cleavage steps, which lie on surfaces parallel to
the {110} primary cleavage plane.
Examination by an optical microscope revealed no slip
lines emanating from the corners of the unbroken Knoop
indents (see Fig. 3a). On the other hand, after the sample
was etched in KOH, small etch pits parallel to the indent,
and small microcracks emanating from the sharp tips of
the Knoop indent appeared (Fig. 5). It appears that when
a pre-cracked sample undergoes bending, a {110} cleavage
crack from one of the ve indents develops and ultimately
causes the cleavage of the whole sample.
3.1.2. Transition behavior
In the four-point bend tests of Samuels on Si [7], the
loaddisplacement curve at the transition regime T T
BDT
Fig. 3. (a) Plan-view of an unbroken Knoop indent on the (001) face of a
GaAs bend sample after deformation; (b) cross-sectional view of the (110)
fracture plane after sample failure from one of the Knoop indents.
0.0 0.5 1.0
0
5
10
15
20
25

A
p
p
l
i
e
d

L
o
a
d

(
N
)
Displacement (mm)
260
o
C
a
b
c
320
o
C
309
o
C
Fig. 4. Load vs. displacement at dierent temperatures at a strain rate of
1 10
6
s
1
.
S. Wang, P. Pirouz / Acta Materialia 55 (2007) 55155525 5517
is still linear but fracture occurs at a higher stress by a
factor of 1.8 than that measured in the brittle regime.
In our work on GaAs, the transition temperature T
BDT
has the following features (curve (b) in Fig. 4):
(1) As in Si, the maximum load to fracture in the transi-
tion regime is higher than the load to fracture in the
brittle regime, 20 vs. 12 N (a ratio of 1.7). A similar
sharp increase in the critical stress intensity factor at
the brittle-to-ductile transition temperature of silicon
was observed by St. John [15]; in his measurements,
the critical stress intensity at the transition tempera-
ture was up to ve times larger than the value mea-
sured in the brittle regime.
(2) At a temperature just above T
BDT
, the load to yield
the crystal is lower than 20 N.
(3) At T
BDT
, the sample shatters into several small pieces
after reaching a load of 20 N.
(4) The scatter in measurement of T
BDT
from several
samples at the same strain rate by this technique
was about 3 C.
3.1.3. Ductile behavior
When the temperature is higher than the transition value
(T > T
BDT
), the loaddisplacement curve no longer
remains linear until the sample fractures (see curve (c) in
Fig. 4). Rather, the linear elastic region of the curve is fol-
lowed by the onset of plastic yielding at a critical value s
uy
(upper yield point). After yielding, the GaAs samples were
examined under an optical microscope. Fig. 6 shows the
tensile surface of a four-point bend sample which had been
deformed in the ductile regime T > T
BDT
. Dense slip lines,
corresponding to traces of {111} slip planes on the (001)
face of the sample, are clearly seen around all the Knoop
indents. Since the four-point bend samples were cut in an
orientation whereby the normal stress was maximized in
a direction perpendicular to the {110} cleavage plane of
the crystal, the maximum shear stress in this orientation
activates the two slip systems 1

10111 and 1

1011

1
to an equal extent. As a result, when the deformed sample
is etched, a high density of etch pits emanating from the
indentation tips appear along the 1

10 direction on the
(001) crystal face (see Fig. 7).
In order to present the response of GaAs to the
applied load r
app
at any temperature T, the applied load
that results in an irreversible change in the crystal i.e.
fracture at r
appl
= r
F
in the brittle regime, or plastic
yielding at r
appl
= r
y
in the ductile regime was plotted
vs. T (see Fig. 8 for a strain rate of _ e 1:0 10
6
s
1
). In
such a plot, for all temperatures below T
BDT
(corre-
sponding to the peak in r
appl
), the sample is brittle and
fractures on the cleavage plane at a nearly constant nor-
mal stress r
n
(=r
appl
= r
F
) of about 100 MPa, whereas at
a temperature just above the brittle-to-ductile transition,
T T
BDT
, the material becomes ductile, deforms plasti-
cally and bends at a yield stress r
y
(=r
appl
) of about
180 MPa.
The stress applied either to fracture or to plastically
deform GaAs is shown in Fig. 9 as a function of tempera-
ture for four dierent strain rates 1.0 10
6
, 2.0 10
6
,
5.0 10
6
, and 1.5 10
5
s
1
. At every strain rate tested,
the plot of r
appl
(T) exhibits the same shape as Fig. 8, with
the peak in r
appl
corresponding to the T
BDT
at that partic-
Fig. 6. Knoop indentation proles after sample deformation at
_ e 1 10
6
s
1
and T = 360 C.
Fig. 5. Knoop indentation prole after sample deformation at
_ e 1 10
6
s
1
and T = 100 C.
5518 S. Wang, P. Pirouz / Acta Materialia 55 (2007) 55155525
ular strain rate. The values of T
BDT
at dierent strain rates
are presented in Table 1; note that the transition tempera-
ture T
BDT
systematically increases with increasing strain
rate. Table 2 compares these with the critical temperatures
T
c2
on the same GaAs crystal, as determined in part I [1].
As described above, two types of samples with dierent
orientations were prepared and tested. According to the
geometry of the samples, b dislocations are expected to
be activated on inclined (111) and

11 slip planes when

a 1

10 Knoop indent is made on the (001) face (corre-

sponding to Fig. 1a) and a dislocations should be activated
on inclined 1

1 and

11

1 slip planes when a [110]

Knoop indent is made (corresponding to Fig. 1b). The
results of a series of tests on the two types of samples at
a strain rate of _ e 1:0 10
6
s
1
are shown in Fig. 10. Sur-
prisingly, no signicant dierence is seen between the T
BDT
obtained from the two types of samples with dierent
orientations.
Fig. 11 shows the resolved shear stress s (rather than the
applied stress r
appl
) as a function of temperature at dier-
ent strain rates. With the present orientation of the four-
point bend samples, four slip systems, 0

1111;
250 300 350 400 450
40
60
80
100
120
140
160
180
200
220
Temperature (C)
1E-6
2E-6
5E-5
A
p
p
l
i
e
d

s
t
r
e
s
s

(
M
P
a
)
1.5E-5
Fig. 9. Temperature dependence of the applied stress needed to deform
the sample at four dierent strain rates.
Table 1
The brittle-to-ductile transition temperature T
BDT
for undoped GaAs at
dierent strain rates
_ e (s
1
) 1.0 10
6
2.0 10
6
5.0 10
5
1.5 10
5
T
BDT
(C) 309 3 328 3 346 3 375 3
Fig. 7. Knoop indentation prole after sample deformation at
_ e 1 10
6
s
1
and T = 360 C.
250 300 350 400
40
60
80
100
120
140
160
180
200
220
A
p
p
l
i
e
d

s
t
r
e
s
s

(
M
P
a
)
Temperature (
o
C)
1E- 6
Fig. 8. Temperature dependence of the applied stress needed to deform
the sample at a strain rate of _ e 1 10
6
s
1
.
Table 2
Comparison of the brittle-to-ductile transition temperature T
BDT
and the
critical temperature T
c2
(from Table 1 in Ref. [1]) for undoped GaAs
_ e (s
1
) 2.5 10
5
5 10
4
1 10
4
2 10
4
T
c2
(C)-single glide 295 3 310 3 325 3 342 3
_ e (s
1
) 1.0 10
6
2.0 10
6
5.0 10
5
1.5 10
5
T
BDT
(C) 309 3 328 3 346 3 375 3
250 300 350 400 450
40
60
80
100
120
140
160
180
200
220
A
p
p
l
i
e
d

s
t
r
e
s
s

(
M
P
a
)
Temperature (C)
1E-6 [1-10]
1E-6 [110]
Fig. 10. Temperature dependence of the applied stress needed to
deform the crystal at _ e 1 10
6
s
1
for the two dierent sample
orientations (see Fig. 1).
S. Wang, P. Pirouz / Acta Materialia 55 (2007) 55155525 5519
10

1111;

10

11 and 0

11, each with a Sch-

mid factor S 1=

6
p
, are equally activated. Irrespective
of the strain rate, s is constant in the brittle regime with
a value of about 100 MPa. On the other hand, the yield
stress s
y
= sr
appl
in the ductile regime (T > T
BDT
) decreases
with increasing temperature, as expected. Here, the applied
stress r
appl
is the stress required to yield the crystal, i.e. the
stress needed to generate dislocations and move them.
For a semi-elliptical surface aw in bending, Keays [16]
gives the following equation for the stress intensity factor
K
Ic
:
K
Ic
rMpa=Q
1=2
3
where r is the maximum out-ber tensile stress, M is a
numerical factor related to the aw and sample geometry,
a is the aw depth and Q is the following expression:
Q /
2
0:212r=r
y

2
4
where r
y
is the tensile yield stress and 0.212 (r/r
y
) is the
plastic zone correction factor. At low temperatures
(<300 C), the correction factor can be neglected, and
Eq. (4) reduces to
Q /
2
5
where / is the elliptic integral
/
Z
p=2
0
sin
2
h a=c
2
cos
2
h
1=2
dh 6
with h as the angular position on the crack front, and val-
ues for / can be found in standard mathematical tables.
For a small semicircular aw, M is 1.03 [17].
In order to determine K
Ic
for GaAs, we took the results
of samples that fractured in the brittle regime. The fracture
stress was calculated using Eq. (1) and the crack depths
were measured directly from the prole visible on the frac-
ture surface. In our low-temperature experiments, the aw
depth was between 17 and 20lm. Using all the informa-
tion, the stress intensity factor for undoped GaAs was
found to be:
K
Ic
0:48 0:04 MPa m
1=2
This is within the range of values reported in the literature
(0.43 0.50 MPa m
1/2
) and obtained by dierent tech-
niques (see e.g. [18]).
3.2. Features around Knoop indentation after deformation in
the brittle and the ductile regime
Using an atomic force microscope (AFM), the regions
around Knoop indentation sites were compared for sam-
ples bent at dierent temperatures (Fig. 12). For samples
tested in the brittle regime, deformation is very localized
and dislocation pile-ups form around the indentation site
(Fig. 12a). In the plastic regime, deformation is spread
out and dislocation pile-ups are less pronounced
(Fig. 12b); instead, slip lines form along the long axis of
the Knoop indents. Under the optical microscope, these
appear as shallow trenches with a depth of around
90 nm. The shallow trenches originate from the indentation
tips.
In order to compare the dislocation arrangement on the
activated and inactivated slip planes, two types of thin foils
were prepared by the FIB technique for TEM analysis.
These foils were all from samples oriented as shown in
Fig. 1a that had been deformed in the ductile regime (at
temperatures above T
BDT
). The two foil types were respec-
tively cut parallel and perpendicular to the 1

10 plane, i.e.
perpendicular and parallel to the [110] bending axis.
Fig. 13 shows a low magnication TEM micrograph of a
thin foil cut perpendicular to the 1

10 bending axis where

the indentation impression can be clearly observed at the
top of the gure. In this view, dislocations on both the

111 and 1

11 sets of activated slip planes can be

observed. The observed slip bands consist of a high density
of dislocations that have emanated from the indentation
site and run deep into the crystal; these dislocations
presumably form when the Knoop indentation is rst
introduced at room temperature. Fig. 14 is a low magni-
cation TEM micrograph of a thin foil cut parallel to the
bending axis. In this view, the (111) and 11

1 sets of inac-
tivated slip planes project along the [110] beam direction.
The absence of any curved dislocation in this gure con-
rms this because any dislocation on the latter two slip
planes would have appeared as a curved line. Instead, in
Fig. 14, one can see a high density of straight lines parallel
to the bending axis; these are projections of dislocations on
the activated

111 and 1

11 slip planes.
In order to investigate the lines in more detail, conven-
tional TEM specimen preparation techniques were
employed to prepare cross-sectional thin foils parallel to
the (111) slip plane from the same samples. Fig. 15 shows
the dislocations lying on this plane under dierent reec-
tions. The majority of dislocations are straight and parallel
to one of the 110 directions in the (111) primary slip
plane. Using the g b = 0 criterion, the straight lines were
determined to be perfect
1
2
h011i type dislocations with dif-
220 240 260 280 300 320 340 360 380 400 420 140
0
20
40
60
80
100
120
140
160
1E-6
2E-6
5E-6
1.5E-5
R
e
s
o
l
v
e
d

s
t
r
e
s
s

(
M
P
a
)
Temperature (C)
Fig. 11. Temperature dependence of the resolved stress needed to deform
the sample at dierent strain rates.
5520 S. Wang, P. Pirouz / Acta Materialia 55 (2007) 55155525
ferent Burgers vectors gliding on the (111) and

11
planes; the observed Burgers vectors were parallel to the
[101], [011], 0

11 and 10

1 directions, which corre-

sponded to the four activated slip systems. Using weak-
beam imaging, these dislocations were always found to be
dissociated over most or all of their lengths; an example
of one of the dissociated dislocations is shown Fig. 16.
4. Discussion
It is known that, under a suciently large shear stress, a
crystal undergoes plastic deformation by the nucleation
and glide of dislocations on the activated slip planes.
Because of the strong covalent/ionic bonding in semicon-
ductors, dislocations in such crystals generally encounter
a large PeierlsNabarro lattice resistance which makes dis-
location movement dicult at low temperatures. Conse-
quently, semiconductors are generally brittle at low
temperatures and, under a tensile stress, the interatomic
bonds break along the easiest the so-called cleavage
plane and the crystal fractures. As the temperature
increases, dislocation glide becomes easier because kinks
nucleate and migrate more readily at higher tempera-
tures. At a suciently high temperature, it is easier for
Fig. 12. AFM images of samples deformed at dierent temperatures.
S. Wang, P. Pirouz / Acta Materialia 55 (2007) 55155525 5521
dislocations to form and move under the shear component
of the load than for the bonds to break under its tensile
component. Thus, the transition from brittleness to ductil-
ity is usually thought of in terms of a competition between
the ease of dislocation nucleation/glide and the ease with
which the interatomic bonds break across the cleavage
plane [1921].
This competition between dislocation activity and frac-
ture was formulated in a simple way in Ref. [22] by consid-
ering the temperature dependencies of the yield stress s
y
and the fracture stress r
F
. For semiconductors, s
y
can be
expressed by the following empirical equation [2325]:
s
y
T A_ e
1=n
exp
DH
s
kT

7
where A and n are constants, and DH
s
is an energy param-
eter such that nDH
s
(=DH
d
) is approximately the activation
energy for dislocation glide DH
d
. On the other hand, the
fracture stress r
F
for a crystal is weakly temperature depen-
dent and can be regarded as a constant, i.e.:
r
n
T r
F
8
where r
n
is the normal component of the applied stress
r
appl
. The two curves dening s
y
(T) and r
n
(T) intersect at
the brittle-to-ductile transition temperature T
BDT
, obtained
by the simultaneous solution of Eqs. (7) and (8) [5,22]:
T
BDT

DH
s
k
B
ln
Sr
F
A_ e
1=n
9
where S is the Schmid factor relating the shear stress s to
the applied stress r
appl
(s = Sr
appl
).
According to Eq. (9), a plot of ln _ e vs. 1/T
BDT
should be
a straight line with a slope
nDHs
k
B

DH
d
k
B

corresponding
to the activation energy for dislocation glide. This was rst
suggested by St. John [15] during his experiments on silicon
and has been veried experimentally by a number of later
researchers both in silicon [26,27] and in other semiconduc-
tors (see Refs. [28] for Ge and [4] for SiC).
Fig. 17 is a plot of ln _ e vs. 1/T
BDT
for our data in GaAs.
The plot is a straight line with a slope that gives an activa-
tion enthalpy of 1.36 0.02 eV. The velocities of a and b
dislocations in GaAs with dierent doping levels have been
directly measured by a number of investigators including
Osvenskii et al. [10,29], Steinhardt and Haasen [30] and,
more recently, Warren [13] and Yonenaga and Sumino
[11,31]. Although the results of these dierent investigators
are not all consistent, there is general agreement that a dis-
locations move faster than b dislocations in both semi-insu-
lating and n-type GaAs and slower in p-type GaAs.
Warren [13] measured the temperature dependence of dis-
location velocity in semi-insulating GaAs with a dopant
concentration of 5 10
15
cm
3
and found an activation
enthalpy of 1.23 0.04 eV for the glide of a dislocations
and 1.35 0.02 eV for b dislocations. In the present exper-
iments, a GaAs crystal with a resistivity of more than
10
7
X cm was used, corresponding to a dopant concentra-
tion of less than 10
8
cm
3
, which is closest to the doping
level used by Warren [13]. Our result, 1.36 0.02 eV
obtained by measuring the strain-rate dependence of T
BDT
is very close to Warrens result for b dislocations
obtained by direct measurement of the dislocation velocity.
This may be an indication that it is the slow b dislocations
rather than the fast a dislocations that control the tran-
sition from brittleness to ductility in undoped GaAs.
The most signicant features of T
BDT
in semiconductors
are the sharp brittle-to-ductile transition over a very nar-
Fig. 13. Low magnication TEM micrograph of the thin foil perpendic-
ular to the bending axis using the reection g = 220; the arrow is parallel
to the g-vector.
Fig. 14. Low magnication TEM micrograph of the thin foil parallel to
the bending axis using the reection g

220; the arrow is parallel to the g-

vector.
5522 S. Wang, P. Pirouz / Acta Materialia 55 (2007) 55155525
row temperature range and the sensitive dependence of the
BDT temperature on the strain rate. In our measurements
on GaAs, the brittle-to-ductile transition occurs to within
3 C and the T
BDT
increases by 65 C when the strain
rate is increased by an order of magnitude. The sharpness
of T
BDT
suggests that the transition is controlled by a
nucleation event [19,20]. On the other hand, the strain rate
dependence of T
BDT
and the fact that the activation
energy derived from a plot of ln_ e) vs. 1/T
BDT
is the same
as that determined from the direct measurements of dislo-
cation velocity - indicates that the BDT transition is con-
trolled by dislocation velocity [21,32].
Fig. 15. Bright-eld TEM micrographs from a sample deformed at 330 C and _ e 5 10
6
s
1
imaged with reection (a) g

220, (b) g 20

2,
(c) g 11

1, (d) g 02

2 and (e) g

11

1; the arrow is parallel to the g-vector.

S. Wang, P. Pirouz / Acta Materialia 55 (2007) 55155525 5523
The BDT results of Samuels et al. on Si [27] were inter-
preted by Hirsch et al. [21] in terms of a model in which
existing dislocations in the plastic zone of the indentation
move towards the crack tip, where they transform into dis-
location sources. A source formed in this way can operate
rapidly in the very high crack tip stress eld to generate a
high density of dislocations that shield the crack and stop
fracture (initiation of ductility). This model was based on
an examination of the dislocation etch pits near the crack
tip at dierent temperatures. However, it should be noted
that etch pits do not indicate the nature of dislocations
as to whether they are partial or perfect. The model pro-
posed in Ref. [5] was developed by measuring T
c
and T
BDT
on the same crystals of SiC at dierent strain rates and
investigating the microstructure of crystals deformed above
and below the critical temperature T
c
. The close values of
T
c
and T
BDT
at dierent strain rates, and the observation
that dislocations generated above T
c
were invariably per-
fect while those generated below T
c
were single partials
dragging a stacking fault, led to the conclusion that the
BDT transition is closely associated with the nucleation
of the slower trailing partial at the transition temperature.
Thus, in this model, it is suggested that at temperatures
below T
c
only the more mobile leading partials nucleate
and participate in crystal deformation, while above T
c
both
leading and trailing partials nucleate to form perfect dislo-
cations, albeit dissociated into leading/trailing partial
pairs. Since a leading partial, once nucleated, shuts o fur-
ther partial nucleation from that source, and since the Bur-
gers vector of partial dislocations is much smaller than that
of perfect dislocations, crystal deformation below T
c
is very
restricted. As a consequence, in order to accommodate the
applied stress, the material will fracture, i.e. the crystal is
brittle at T < T
c
. On the other hand, above T
c
, the perfect
dislocations that nucleate can readily multiply, e.g. from
Frank-Read sources. Formation of an avalanche of perfect
dislocations each with a relatively large Burgers vector is
able to shear the crystal to a signicant degree, blunt the
crack tip and fully relieve the applied stress, i.e. the crystal
becomes ductile at T > T
c
.
The results of the present experiments on GaAs are sim-
ilar to those reported for Si and SiC. From the etch pits
appearing in samples deformed at high temperatures and
the TEM analysis of these dislocations, it may be con-
cluded that the dislocations are mainly emitted from the
crack tip. In addition, the TEM analysis of the dislocations
shows that those generated above the transition tempera-
ture are invariably perfect (dissociated). Unfortunately,
because of experimental diculties, it has not been possible
to determine the nature of dislocations below the transition
temperature satisfactorily in our experiments.
In general, the dierent core natures of a and b disloca-
tions aect the mechanical properties of compound semi-
conductors in dierent ways, and it was expected that
samples with orientations that preferentially activate a or
b dislocations would give rise to dierent transition tem-
peratures. However, this was not found to be the case in
our bending tests. Sumino and Shimizu [33] investigated
the mechanical properties of InSb by bending tests and
found dierent ow stresses in samples with dierent orien-
tations. They suggested that there was no direct relation-
ship between the dierent ow stresses and the mobility
of a and b dislocations in the plastic regime of InSb. They
pointed out that the majority or minority dislocations
could not result from the motion of a and b dislocations
but resulted instead from the motion of screw dislocations
which should be rate controlling in bending of either type
of samples. In their experiments, they observed that, in
addition to the primary slip system, dislocations in the
Fig. 16. Weak-beam image of a sample deformed at 330 C and
_ e 5 10
6
s
1
that shows the dissociation of a perfect
1
2
1

10 type
dislocation on its (111) glide plane.
15 16 17 18
-15
-14
-13
-12
-11
-10
.
l
n

(


)
1/T
BDT
(10
-4
K
-1
)
Fig. 17. Plot of logarithm of the strain rate vs. the reciprocal of the brittle-
to-ductile temperature T
BDT
.
5524 S. Wang, P. Pirouz / Acta Materialia 55 (2007) 55155525
secondary slip system were activated. It is unlikely that this
situation applies to our bending experiments. Firstly,
according to the Schmid factor, no secondary slip systems
should be activated in either of the two orientation type
samples. Secondly, our TEM experiments did not show a
high density of screw dislocations in the bent samples.
Although, at temperatures above T
BDT
in Fig. 10, there
may be some dierence in the yield stress of the two dier-
ently oriented samples, this issue was not pursued any
further.
The fact that no dierence was observed in the brittle-to-
ductile transition temperatures of samples with dierent
orientations may be indicating that a and b dislocations
are activated simultaneously on the (111) Ga and

1
As slip planes. It should be noted that the dislocations that
are activated at T
BDT
are not only from the existing dislo-
cations sources (nucleated when the pre-crack is introduced
in the crystal by room temperature Knoop indentation) but
also from the dislocation sources near the indentation tip
which can be activated by the shear component of the
applied stress eld. In this case, a and b dislocations can
be activated at the same time, and the brittle-to-ductile
transition in both cases could be controlled by the slowest
moving b dislocations. It is noteworthy that both a and b
dislocations were also observed by Warren [13] during his
four-point bend tests performed for measuring the disloca-
tion velocity in GaAs.
5. Conclusion
1. The brittle-to-ductile transition temperature T
BDT
of undoped GaAs is sharp to within 3 C and varies
from 309 C at _ e 1 10
6
s
1
to 365 C at _ e
1:5 10
5
s
1
.
2. The plot of ln_ e vs. 1/T
BDT
follows an Arrhenius rela-
tionship, and its slope gives an activation energy of
1.36 0.02 eV. This is very close to the value of
1.35 0.02 eV for glide of the slow b dislocations
(rather than 1.23 0.02 eV for the fast a dislocations)
as determined by direct measurement of dislocation
velocity in undoped GaAs crystals.
3. At temperatures above the T
BDT
, perfect dislocations,
dissociated into leading/trailing partial pairs, are acti-
vated by the application of an applied shear stress. Pre-
sumably these dislocations blunt the crack tip and lead
to the ductility of the crystal.
4. As mentioned in the previous paper, the values of T
BDT
and T
c
at four dierent strain rates and measured by
completely dierent techniques agree reasonably well
with each other leading to the conclusion that lack of
shear deformation in the brittle regime (i.e. below T
c
or T
BDT
) is not because of a complete inactivation of
dislocations but rather because of the inactivation of
trailing partial dislocations. As soon as the combination
of tensile stress/thermal energy is sucient to nucleate
the slow trailing partials (at T > T
BDT
), and perfect dis-
locations nucleate and glide, the crystal becomes ductile.
Acknowledgements
This work was supported by Grant No. DMR-0108303
from the National Science Foundation. The authors thank
Dr. Ming Zhang for his assistance during the course of this
work.
References
[1] Wang S, Pirouz P. Acta Mater 2007;55(16):550014.
[2] Samant AV. Eect of test temperature and strain-rate on the critical
resolved shear stress of monocrystallineSiC, Ph.D. Thesis, Case
Western Reserve University; 1999.
[3] Demenet J-L, Hong MH, Pirouz P. Scripta Mater 2000;43(9):
86570.
[4] Zhang M, Hobgood HM, Demenet JL, Pirouz P. J Mater Res
2003;18(5):108795.
[5] Pirouz P, Zhang M, Demenet J-L, Hobgood HM. J Appl Phys
2003;93(6):327990.
[6] Fujita S, Maeda K, Hyodo S. Phil Mag A 1992;65(1):13147.
[7] Samuels J. The brittle to ductile transition in silicon, D. Phil. Thesis,
University of Oxford; 1987.
[8] Haasen P. Acta Met 1957;5:5989.
[9] Alexander H, Haasen P, Labusch R, Schro ter W. Foreword to. J Phys
(Paris) 1979; 40:Colloque C6.
[10] Osvenskii VB, Kholodnyi LP. Soviet Phys Solid State
1973;14(11):28225.
[11] Yonenaga I, Sumino K. J Appl Phys 1989;65(1):8592.
[12] Yonenaga I, Sumino K. J Appl Phys 1993;73(4):16815.
[13] Warren PD. The relation between electronic and mechanical prop-
erties of non-metals, Ph.D. Thesis, University of Oxford; 1987.
[14] Bruneau AA, Pratt PL. Phil Mag 1962;7:187185.
[15] St. John C. Phil Mag 1975;32:1193212.
[16] Keays RH. Structures and Materials Reports 343, Report No. (April
1973).
[17] Petrovic JJ, Jacobson LA, Talty PK, Vasudevan AK. J Am Ceram
Soc 1975;58:1136.
[18] Yasutake K, Konishi Y, Adachi K, Yoshii K, Umeno M, Kawabe H.
Jap J Appl Phys 1988;27(12):223846.
[19] Kelly A, Tyson WR, Cottrell AH. Phil Mag 1967;15:56786.
[20] Rice JR, Thomson R. Phil Mag 1974;29:7397.
[21] Hirsch PB, Roberts SG, Samuels J. Proc R Soc Lond A 1989;421:
2553.
[22] Pirouz P, Zhang M, Demenet J-L, Hobgood RM. J Phys: Condens
Matter 2002;14(48):1292945.
[23] Alexander H, Haasen P. Dislocations and plastic ow in the diamond
structure. In: Seitz F, Turnbull D, Ehrenreich H, editors. Solid state
physics, vol. 22. New York: Academic Press; 1968. p. 27158.
[24] Alexander H. Dislocations in covalent crystals. In: Nabarro FRN,
editor. Dislocations in solids, vol. 7. Amsterdam: Elsevier Science
Publishers B.V.; 1986. p. 114234.
[25] George A, Rabier J. Rev Phys Appl 1987;22:94166.
[26] Brede M, Haasen P. Acta Metall 1988;36(8):200318.
[27] Samuels J, Roberts SG. Proc R Soc Lond A 1989;421:123.
[28] Serbena FC, Roberts SG. Acta Metall Mater 1994;42(7):250510.
[29] Osvenskii VB, Kholodnyi LP, Milvidskii MG. Sov Phys Solid State
1973;15(3):6612.
[30] Steinhardt H, Haasen P. Phys Stat Sol (A) 1978;49:93101.
[31] Yonenaga I, Sumino K. J Appl Phys 1987;62(4):12129.
[32] Hirsch PB, Roberts SG. Phil Mag A 1991;64(1):5580.
[33] Sumino K, Shimizu H. Phil Mag 1975;32(1):14357.
S. Wang, P. Pirouz / Acta Materialia 55 (2007) 55155525 5525