Materials Science and Engineering, A176 (1994) 71-78 71

High temperaturefracture
Horst Vehoff
MPI fiir Eisenforschung, Max Planck Strafle 1, W-4000 Diisseldorf 1 (Germany)

Abstract

Deformation and fracture of materials at elevated temperatures are time-dependent processes. For different loading
modes, cyclic or monotonous, cracks can nucleate on the free surface by localized corrosion at particles or interfaces or
by cavity nucleation, growth and linkage in the interior of the material. In this short keynote paper corrosion effects are
only discussed briefly. The main part is focused on cavity nucleation and growth. On the basis of recent measurements of
cavity size distribution functions as a function of temperature, load and strain in bicrystals, bimaterial interfaces and
directionally solidified oxygen-dispersion strengthened alloys, cavity growth by diffusion (quasi-equilibrium and crack-
like cavities) and creep is discussed. In addition, experimental results on the effects of cavity nucleation on growth and the
influence of cavity linkage on residual life are reported. These results are discussed from the viewpoint of predicting
rupture life.

1. Introduction which rule out cavitation in order to understand the

Deformation and fracture of materials at elevated
temperatures are time-dependent processes. For dif-
ferent loading modes, cyclic or monotonous, cracks
can nucleate on the free surface by localized corrosion
at particles or interfaces or by cavity nucleation,
growth and linkage in the interior of the material.
Depending on the service conditions, either corrosion
and creep processes operate in parallel or one of them
dominates. In thie short review paper a complete
survey of the various corrosion and creep mechanisms
cannot be given. Rather the theoretical models are dis-
cussed in the light of new experimental results. The
main part of the paper is focused on cavity nucleation
and growth, since quantitative data on the time, tem-
perature and load dependences of cavity distribution
functions are now available in the literature. On the
basis of these data, the predictions of various cavity
nucleation and growth models are compared.
2. Crack growth by high temperature corrosion
fatigue
At high temperatures cracks can nucleate either by
cavity nucleation and growth or by local oxidation [1,
2]. Both mechanisms can operate in the same material
in parallel and it is difficult to determine which
mechanism dominates or if synergistic effects lead to a
drastic reduction in lifetime. Therefore the nucleation
and growth of cracks must be studied under conditions
role of oxygen in enhancing or delaying the growth rate
of cracks [3]. The simplest model for crack growth pro-
posed by Saxena and Bassani [4] assumes that the
growth rate is controlled by the diffusion of oxygen
along the grain boundary, which gives
da
- Z Aai = nAaid + n(Dtp) 1/2 ~ C(AK~)q + Cj v - i/2
dN
(1)
where at, D, K~ and v are local crack events, the dif-
fusion constant of oxygen, the stress intensity factor
and the loading frequency respectively. However, the
corrosion processes in commercial materials are com-
plicated. An example of a section through a microcrack
in the single-crystal superalloy CMSX-6 is given in Fig.
1. It is obvious that the corrosion processes at the crack
tip depend on the details of the alloying composition
and cannot be described easily. Therefore only some
simple experiments related to oxide scale growth will
be discussed below.
Depending on the loading rate, either transgranular
or intergranular cracks are observed in commercial
materials. In tests with FeSi and FeCrNi bicrystals at
very low frequencies, specimens which had sulphide
particles on the boundary showed corrosion and cavity
nucleation and growth [5]. Cavity nucleation can be
suppressed when tests are done on bicrystals with
boundaries which are free of particles [6]. Therefore
only clean material was used in ref. 6 for the high tem-
perature tests, in which no cavitation was observed for

0921-5093/94/$7.00 1994 - Elsevier Sequoia. All rights reserved

SSD10921-5093( 93 )02549-I
72 H. Vehoff High ternperaturefracture

Fig. 1. Metallographic section through a crack tip in a CMSX-6

single crystal after fatigue cracking with a hold time of 60 s.
Fig. 3. Intergranular fracture surface of an FeSi bicrystal with
transgranular markings due to superimposed hold times.

intercrysta~ne 41" ~ "t- n 8~ 8s

lOO
10
+ + ~ 1 ~
aE hold time
Z

FeSi-bicrystal X
+ Air 873K ~..~ ~o,
transcryst alline
, , i i J ~ , L I L j . . . . . . i . . . . . . . . i
1 O. 1
0.00 0.01 0. 10 1.00 0.01 o.1 1 lo loo looo
elongabon rate [#m/s] t [s]
Fig. 2. Cyclic crack growth rate da/dN as a function of elonga- Fig. 4. Plastic crack tip opening and scale thickness vs. loading
tion rate. time.

all test frequencies and the effect of corrosion on crack oxide scale grows laterally. This assumption was tested
growth can be examined separately. by Lombard et al. [6]. Their result is plotted in Fig. 4.
The dependence of the cyclic crack growth rate The thin lines give the local crack tip opening
da/dN on the elongation rate is shown in Fig. 2. At A6 = d t as a function of the loading time for several
high elongation rates the crack propagated inter- tests. In the same diagram (thick line) the growth of the
granularly. The crack started to grow along the grain oxide scale thickness d assuming a parabolic growth
boundary when the elongation rate was reduced, since l a w d 2 = kpt is plotted. The parabolic growth constant
the crack tip was kept on the boundary by local oxida- kp for ~-Fe was taken from the work of Takada and
tion [6]. Surprisingly, a further reduction in elongation Adachi [7]. For the test temperature of 600 C a value
rate resulted again in transgranular crack growth. An of 4.72 x 10-17 m 2 s-1 was used. The upper rate was
example of a fracture surface after a test with two the highest rate at which intergranular crack growth
elongation rates and alternate inter- and transgranular was observed, while the lower rate corresponded to the
crack growth is given in Fig. 3. The intergranular lowest test frequency. In tests in which the crack open-
cracks always left the boundary during a hold time in ing increased faster than the scale thickness, oxygen
air as indicated in Fig. 3 by the steps perpendicular to could reach the crack tip freely during loading and
the crack growth direction. This transition from inter- intergranular crack growth was observed. However,
to transgranular fracture can be explained if it is when hold times were introduced, the crack-tip-
assumed that the crack opens at a lower rate as the opening rate was much lower. In this case no preferen-
 H. Vehoff / ttigh temperaturefracture 73

tial attack occurred at the grain boundary, since the
crack was completely closed by oxidation products. In
these cases the crack path changed from intergranular
to transgranular (Fig. 3) and the growth rate was
reduced by nearly a factor of 10 (Fig. 2). However,
when the boundary contained particles at which
cavities could nucleate easily, the crack stayed on the
boundary, but now the crack grew by cavity coales-
cence as shown in Fig. 5. The influence of cavity
nucleation and growth on rupture life will be discussed
below neglecting corrosion effects.
3. Cavity growth
Measurements of cavity distribution functions were
recently published by Kim et al. [8] for Ni3A1 and by
Hiraga and Vehoff [9] for recrystallized Inconel MA
754. Both materials contained pores on grain bound-
aries and hence were suited to measure cavity growth
directly without the kinetics being obscured by cavity
nucleation processes. Figure 6 shows typical distribu-
tion functions of round cavities in Inconel MA 754
corrected by the Curz-Orive method [10]. The dis-
tributions for round and crack-like cavities (Fig. 7)
were fitted with a modified log-normal distribution
function
observed (Fig. 6) the method is straightforward. How-
ever, when continuous nucleation is found (Fig. 7), this
procedure can only be applied for the largest cavities,
with special care being taken not to count cavities
which are linked together. Figure 8 gives measured
growth rates for round cavities obtained by this
method. The plot clearly demonstrates that small
cavities grow very fast. A minimum in the growth rate
was found for cavity sizes of about 0.4 ~tm. For larger
cavity sizes the growth rate increased again. This type
of behaviour was predicted by several authors on the
basis of various models. The measurements of Hiraga
and Vehoff [9] yield for the first time a complete set of
cavity growth data including the asymptotic growth
behaviour of small and large cavities. On the basis of
these data the various models of cavity nucleation and
growth are discussed in the following sections. For this
purpose the results for 190 MPa are replotted in Fig. 9
(filled circles) together with fits according to various
growth models (open symbols).
At low stresses and for small cavity sizes the grain
boundary diffusivity should be lower than the surface
800 ~ MA 7 5 4
f/X 2 4 0 MPo

600 P ]
~'~ ~,/ ~ "~% o 1.0
N(R,t)=C~{(2x)12RczeXp[-I(In(~(~[~))2]+C41 E

(2) z 200

Cavity growth rates were evaluated from these curves

on the basis of the method proposed by Schneibel and 0 -- I I I - ~ ~
0 I 2 3 4 5
Martinez [11]. For cases in which no nucleation is
Cevity Diameter [/~rn]

Fig. 6. Size distribution functions for round cavities on quasi-

boundaries in In-Conel MA 754.

7500
CRACK-LIKE CAVITIES

I v 0.5
5000 o 1.0
o 1.4
o 1.9

g 2500

f o o

1 2 3
Cavity Diameter [/.~m]
Fig. 5. Scanningelectron microscopy(SEM) image of a crack tip
in an FeSi bicrystal which propagates by cavity growth and Fig. 7. Cavity distributions for crack-like cavities in Inconel
coalescence. MA 754.
74 H. Vehoff / High temperaturefracture
i i i ,
6 10 - 7
5 10 - 7 MA 7 5 4
M A 754 o
7 190 MPa
'~ =\ 260 MPa 4- 10 - 7
E O0
& 3 10 - 7 a v
0-7 " . * * * o o 0 ~gg
&e

g
2 10 - 7
o
c~
o 240 M P a o
8
'~ 2 10 -7 o o v
8 a
10 -7 I I o I I
I 1 I I 0.0 0.4 0.8 1.2 1.6 2.0
0.0 0.4 0.8 1.2 1.6 2.0
Cavity Radius [/~m]
Cavity Radius [/zrn]
Fig. 9. Cavity growth rate for round cavities (filled circles) and
Fig. 8. Cavity growth rates for round cavities on quasi-bound- fits according to unconstrained diffusional growth (open circles
aries as a function of cavity radius. [13]), diffusional growth of crack-like cavities (open triangles
[16]), power-law creep (open squares [18]) and diffusional
growth and power-law creep (open-diamonds).
diffusivity and quasi-equilibrium cavity growth should
be the dominant growth mechanism [12]. The model
given in ref. 12 was modified by Chen and Argon [13].
0
They introduced a diffusional length A in which matrix
creep serves to shorten the distance of grain boundary
diffusion necessary to accommodate the matter trans-
T T
ported from the cavity walls:

( Db(~bff'dOoo] 1/3
A = t / (3) I-" /i
DaB b, '
where Db, ~b and g~o are the grain boundary diffusivity,
<, A >
the grain boundary thickness and the applied strain
rate respectively. The geometry is depicted schemati-
cally in Fig. 10. For the growth rate, ignoring the capil-
larity terms [14], this yields [13] Fig. 10. Cavities on a grain boundary; all relevant geometric
parameters are given.
4 ~ h ( ~ ) d a _ 2~ In
gooa dt a ~--

+ a +A] L -4 '(a)21 43-} - (4) g~a \a! [ \ a i

where h(~p) is a geometric function. The open circles in + 1 a 2 -

Fig. 9 are fitted with eqn. (4). A value for ~p of 70 ,
which gives h(~p)=0.61, and a strain rate of
1 1 0 - 8 s- ~ are used, yielding the single fitting param- where
eter A = 1.47 btm. The SEM image of round cavities in
Fig. 11 shows that the distance of small cavities
exceeds 1.5 p.m; hence the obtained value of A is a =
47~h(~p)
[4 s i n ( l p / 2 ) ]
(DbrbOooAIm
3/2 \ O ~ , ~--~ ]
(6)

reasonable. However, with typical, data for Ni in eqn.

(3) a higher value of 13 ~tm is obtained. According to prescribes a coupling parameter, the ratio of effective
the fit at 973 K, Do6 b must be two orders of magnitude boundary diffusional conductance to surface diffu-
lower than in pure Ni when a value of 2.29 x 10 -21 is sional conductance [16]. Ds, 6 , Z~ and aoo are the
used for Db6 b of Ni [15]. surface diffusivity, the surface layer thickness, the sur-
The open triangles in Fig. 9 represent the result of a face energy and the applied load respectively. The fit
similar fit for crack-like cavities according to refs. 16 yields a value of 2.7 for a. However, taking 142 for
and 13: oo~A/x~ from the measurements, a value of 0.05 was
 H. Vehoff /
A
I cavities underestimated the growth rate as well. Best
Fig. 11. SEM image of round cavities after 1.4% strain at 973 K
in Inconel MA 754.
obtained for the ratio of the diffusional conductances;
hence the fit yields D~6~ ">Dbd b in contradiction to the
assumption of crack-like cavities. Comparing the
experimental data with the model given by Martinez
and Nix [14] showed that even for such low ratios but
at high enough stresses (large a) the crack-like solution
could remain valid owing to capillarity effects. How-
ever, even the large cavities shown in Fig. 11 main-
tained their equilibrium shape.
The assumption that the larger cavities grow by
creep describes the observed behaviour better than a
transformation from round to crack-like cavities. For
cavity growth by creep the growth law can be written as
[17]
(7)
3h(~p)
The shape factor 2', which depends on the ratio
between the stress components, can be approximated
as 1.1 for uniaxial loading. A fit according to eqn. (7) is
High temperaturefracntre 75
shown in Fig. 9 (open squares). However, a factor of 32
for 2' was obtained from the data when the applied
strain rate g ~ was used in eqn. (7). The total growth
history can be described if it is assumed that the small
cavities (less than 0.6 gin) grow by diffusion and the
large cavities by creep. Adding the solutions (4) and (7)
yields the open diamonds in Fig. 9 with A = 1.4 btm
and 2 '= 20. All models underestimate the growth rate
of the large cavities for the chosen experimental condi-
tions. However, small changes in the local stress dis-
tribution can change the local strain rate and hence the
growth rate by creep drastically. The findings discussed
above are consistent with the data of Kim et al. [8] for
Ni3A1. They found that growth models for crack-like
fits were obtained when unconstrained cavity growth
was assumed with a five-times-smaller diffusion coef-
ficient as for clean Ni3AI boundaries. The uncon-
strained and independent growth of small cavities can
be predicted correctly by the models presented above.
However, when cavity interaction and cavity linkage
occur, models based on the independent growth of
cavities yield erroneous results.
4. Microcrack growth
Larger cavities often grow together and form micro-
cracks. The stress distribution near the tip of a micro-
crack can lead to accelerated growth of neighbouring
cavities. In this case the nature of the creep damage
changes from being dominated by individual quasi-
equilibrium cavity growth to being dominated by
microcrack growth in the final stages of creep life. In
Ni3AI cavity coalescence was observed throughout the
lifetime and becomes predominant as early as after
65% of the life of the specimens [8]. In an oxygen-dis-
persion-strengthened (ODS) superalloy [9] the micro-
cracks dominate the life after 30% of the lifetime.
Figure 7 shows distributions of crack-like cavities and
microcracks (emerging from the linkage of small
cavities) as a function of strain. The number of micro-
cracks increased with time. From these distributions
the growth rate of the largest microcracks was mea-
sured by the method given in ref. 11, neglecting further
coalescence. The result is shown in Fig. 12. With
increasing strain the measured R dependence changes,
approaching an R dependence which is similar to the
dependence of microcracks on crack length [18]. The
measured rates are an order of magnitude higher than
the growth rate of equilibrium-shaped cavities under
otherwise identical conditions (Figs. 8 and 12). Hence,
at least in this material, the lifetime is controlled by the
early coalescence of cavities and the subsequent
growth of microcracks. In the example given above the
76 H. Vehoff / Hightemperaturefracture

10 - 5 , ~ ,
25
CRACK-LIKE CAVITIES l -~- 1082 h. 0-=21 MPa
Ire
,~ 20 /
A ~ j.I A 2 8 0 h. 0-=24 M P a

izo 6 1 O-S 15 / ~ [] 46 h. o-=39,1 MPa

0,4 - 1.3 % o
o 4 10 -e ,',R 1 . / 2 , ~ 7 f 1 - 1.9
>, aR 3/2
to

g,
"~ 2 10 -e
0U
I I l
2 4 6 8
0
Cavity Radius [/m] 0 50 100 150 200 250 300

Fig. 12. Cavity growth rates for crack-like cavities as a function Cavity radius [am]
of cavity radius.
Fig. 14. Cavity distributionsfor FeSi bicrystals.

1500
35

1000 3O 1= 1 2 5 h
0
>. 25
~/ cy= 13 1MPa
"o # / t=55h
20
>. 0

t) + N(t)=C2tO.35[m
-2] E
2
15

o + 10

1O0 , , ~ h , 5
0 200 400 600 800 1000 1200
0
Creep time [hi 0 50 100 150 200 250 300
Fig. 13. Total cavity density measured on a cleaved grain bound- Cavity radius [am]
ary in an FeSi bicrystal as a function of creep time.
Fig. 15. Cavity distributionsfor X8 CrNiNb 1613 bicrystals.

number of microcracks increased with time which

resulted in a reduction in life. However, often continu-
ous cavity nucleation occurs during creep which is pos-
tulated to delay the growth rate of cavities.
5. Cavity nucleation and growth
In many steels cavities nucleate during creep.
Detailed examinations of NiCr steels [19], austenitic
and ferritic bicrystals [20] and metal-ceramic inter-
faces [21] showed that cavities nucleated early within
the first 10% of life. Later in life cavity nucleation
either ceased completely or progressed slowly (Fig.
13). Figures 14 and 15 show typical distributions
obtained for FeSi and FeCrNi bicrystals respectively
after creep. The maximum of the distributions shifted
towards lower cavity sizes with increasing creep time, a
clear indication of continuous cavity nucleation. A
surprising observation was that, within the statistical
scatter, even after a pronounced creep time of 1082 h
no clear evidence for further growth of larger cavities
was observed (Fig. 14). A similar behaviour was
observed for copper-sapphire interfaces by Kennfick
and Raj [21] as shown schematically in Fig. 16. The
data points represent the measurements and the curve
gives the prediction of the model when unconstrained
cavity growth is assumed, which is reasonable for a
bimaterial interface.
Cavity growth laws cannot be measured directly
from distribution functions when continuous nuclea-
tion occurs. A model is needed for nucleation and
growth to calculate a distribution function, which in
turn is used to evaluate the constants of the growth and
nucleation laws. Riedel [22], assuming power laws for
nucleation and growth, has evaluated a cavity size dis-
tribution function by solving the continuity equation
for cavities in size space:
1Q+ O(Nt~)= 0 (8)
OR
 H. Vehoff / Hightemperature,fracture 77

1.0 300 ........ , .........

Interrupted Tests
g CAVITY A R E A v s TIME
0.8
o 700C
o heryx, X to= l s

i
C_ O'oo = 2 . 6 4 MPo
200 /
o 0.6
t_
< Cu < 1 1 1 > /
>, SAPPHIRE
0.4
o
o
100
0.2 , t
to= 1 h
/
0.0 I
101 10 2 10 3 10 4 10 5 10 6 0 ........ k ........ i
time [s] 10 10' 10 ~ 10 3 10 4 10 5 10 6

Creep time t [s]

Fig. 16. Growth in the area fraction of cavities at a Cu(11 1)-
sapphire interface and prediction from unconstrained growth Fig. 1 7. Measured growth rates in FeSi bicrystals with the growth
[21]. law/~ = A i R 21-1.35"

with If combined with continuous nucleation to give

I{ = A I R -~t -~, J* =A2U (9)
A2 1
The exponents a, fl and 7 depend on the cavity growth f N(R,t)dR=f J*dt= y + 1 t~+' =2~ (13)
model. For diffusion-controlled cavity growth a value
of/3 = 2 must be used. The solution yields a class of this yields for the growth rate
size distribution functions
ag t a
N(R,t)=A~ R~t,+r(1 1 - a R ~+' / ~a+>/':l-~) RoC4nh(~p
~ R2 with a=7+l (14)
A, l + /3 Alfl-"] (10)
Fitting the data gave values for 7 of - 1 and - 0.65 for
where J* is the nucleation rate and A] and A 2 a r e con- stainless steel and FeSi respectively and a was found to
stants which depend on temperature and stress be related to the exponent of the nucleation rate J* by
respectively. Equation (10) was used to fit the distribu-
tions in Figs. 14 and 15. With increasing creep time the a=7+2 (15)
maximum of the distributions shifted towards lower The growth rate is lowered by a higher degree as pre-
cavity sizes. Differentiating eqn. (10) with fl = 2 (cavity dicted by eqn. (14). The resulting growth law is plotted
growth by diffusion) yields for the maximum of the dis- in Fig. 17. It can be seen that 1 h of creep after the
tribution cavities grow only very slowly.
In most cases in which the growth of cavities is in-
R3(Nm,x, t)=A3t' " (11)
hibited, cavity growth is not the life-controlling factor.
Hence the growth law (9) must be explicitly time The reasons why on many interfaces larger cavities
dependent with a > 1. In other words, with increasing stop growing are not understood. It is speculated [21]
creep time the growth rate of the larger cavities should that the nature of the boundary changes with cavita-
decrease drastically as was observed in the experi- tion, which can change the grain boundary diffusivity,
ments. but measurements of grain boundary diffusivities on
Inhibited cavity growth was first proposed by Ishida cavitated and particle-rich boundaries are not avail-
and McLean [23]. They suggested that the vacancies able.
needed for diffusional growth are produced by disloca- Predictions of rupture lifetimes can be obtained by
tions which enter the grain boundary during creep. The integrating the distribution N(R, t) for the various
number of vacancies thus generated per unit time is growth laws according to
assumed to be proportional to the strain rate g. The
vacancies can be collected only from a grain boundary
area of 22 , which yields for the growth rate [24]
to= f :zR2N(R,t)dR (16)

g 22 where to is the cavitated area fraction at the grain

/~ 4zrh(~p)R 2 (12) boundary. The time to fracture, t~, is reached when a
 78
10000
~
" " ~ ' ~ - 7 ~ ~
constrained
continuous
1000 unc strained
/ instantaneous~
100 " ~'--- ~.n~clLt_i ? . _ ~ _ _ _
2 1/6-1 Mo steel 565"C
10 ' ' ' ' '
50 1O0
o" [MPo]
Fig. 18. Theoretical predictions of rupture lifetime for con-
strained and unconstrained cavity growth with and without con-
tinuous nucleation. (After ref. 25.)
critical grain boundary area fraction (.ofis cavitated:
(h(lp)kTwfl2/5 1
t f = 0 " 2 5 [ ~ D b O ] (j,)3/5
Predictions for various growth and nucleation laws are
shown schematically in Fig. 18 [25]. T h e chain curves
are the predictions for unconstrained growth and the
full curves for constrained growth. Continuous nuclea-
tion can enhance the rupture life by several orders of
magnitude. For the same material the various models
give predictions which deviate by orders of magnitude.
Experiments are needed which incorporate the statis-
tical aspects of void nucleation and growth in order to
understand the role of cavitation in rupture life. With
m o d e r n image-processing software measurements with
large ensembles are possible to give data which are
statistically significant.
6. Conclusions
Damage processes at high temperature are
extremely complicated, especially when corrosion pro-
cesses are involved. However, experimental methods
based on quantitative metallography combined with
m o d e r n image processing allow us to gather and
handle large amounts of data quickly. Distributions of
cavity and particle sizes as well as of nucleation sites
can now be measured as a function of the parameters
of interest with reasonable effort. T h e application of
H. Vehoff / High temperaturefracture
the theory to such data shows that many of the
observed dependences can be predicted correctly.
However, quantitative predictions were found to be
way off in most cases.
Measurements of this kind should be done on
various materials in order to obtain a database for the
development of predictive theories for the rupture life
of components under service conditions.
References
200
1 R. Raj, Acta Metall., 26 (1978) 995.
2 J. S. Wang, J. J. Stephens and W. D. Nix, Acta Metall., 33
(1985) 1009.
3 T. Ericsson, Can. Metall. Q., 18(1979) 177.
4 A. Saxena and J. L. Bassani, in J. M. Wells and J. D. Landes
(eds.), Fracture: Interactions of Microstructure, Mechanisms
and Mechanics, Metallurgical Society of AIME, Warrendale,
PA, 1984, p. 357.
5 R. Lombard and H. Vehoff, Scr. Metall., 24 (1990) 581.
6 R. Lombard, H. Vehoff and E Neumann, Z. Metallkd., 83
(17) (1992)463.
7 J. Takada and M. Adachi, J. Mater. Sci., 21 (1986) 2133.
8 H. K. Kim, E. J. Lavernia and J. C. Earthman, Acta Metall.
Mater., 40(1992) 1933.
9 K. Hiraga and H. Vehoff, JIMIS-7, Aspects of High Tempera-
ture Deformation and Fracture in Crystalline Materials,
Nagoya, July 1993, Japan Institute of Metals, Sendai, 1993,
p. 229.
10 L. Curz and M. Orive, Z Microsc., 107(1976) 235.
11 J. H. Schneibel and L. Martinez, Philos. Mag. A, 54 (1986)
489.
12 W. Beere and M. V. Speight, MetalL Sci., 12 (1978) 172.
13 I.W. Chen and A. S. Argon, A cta Metall., 29(1981) 1759.
14 L. Martinez and W. D. Nix, Scr. Metall., 15 ( 1981 ) 757.
15 H.J. Frost and M. E Ashby, Deformation Mechanism Maps,
The Plasticity and Creep of Metals and Ceramics, Pergamon,
Oxford, 1982.
16 T. Z. Chuang, K. I. Kagawa, J. R. Rice and L. B. Sills, Acta
Metall., 27(1979) 265.
17 B. Budiansky, J. W. Hutchinson and S. Slutsky, in H. G.
Hopkins and M. J. Sewell (eds.), Mechanics of Solids,
Pergamon, Oxford, 1982, p. 257.
18 H. Vehoff and E Neumann, Steel. Res., 63 (1992) 371.
19 H. C. Cho, J. Yu and I. S. Park, Metall. Trans. A, 23 (1992)
201.
20 R. Lombard and H. Vehoff, Mater. Sci. Forum, 126-128
(1993) 347.
21 C. M. Kennefick and R. Raj, Acta Metall. Mater., 40 (1992)
615.
22 H. Riedel, Z. Metallk., 76 (1985) 669.
23 Y. Ishida and D. McLean, Met. Sci., 1 (1967) 171.
24 W. Beere, J. Mater. Sci., 15(1980)657.
25 H. Riedel, Fracture at High Temperatures, Materials
Research and Engineering, Springer, Berlin, 1987.