Pergamon Engineerhlg Fracture Mechanics Vol. 58, No. 4. pp.

327-354, 1997
"('~ 1997 ElsevierScience Ltd. All rights reserved
Printed in Great Britain
PII: S0013-7944(97)00102-1 0013-7944/97 $17.00+ 0.00

SHORT FATIGUE CRACK BEHAVIOUR AND

A N A L Y T I C A L MODELS: A R E V I E W
K. HUSSAINt
Metallurgy Division, Dr A. Q. Khan Research Laboratories, PO Box, 502 Rawalpindi, Pakistan

Abstract--The use of engineering materials in critical applications necessitates the accurate prediction
of component lifetime for inspection and renewal purposes. In a cyclic loading situations, it is very im-
portant to be able to predict the growth rates of cracks from initiation to final fracture. Most of the in-
vestigations have been focused on the long fatigue cracks from the notched specimens, and linear
elastic fracture mechanics is used to analyse the crack growth data. The behaviour of the short fatigue
cracks less than a millimeter in length (short cracks) cannot be analysed by linear elastic methods
because of large-scale plasticity effects. A number of modification to LEFM and new models have been
introduced to correlate the behaviour of long and short fatigue cracks. This paper addresses the critical
view of the modifications and some of the models proposed for the analysis of short fatigue cracks. ~
1997 Elsevier Science Ltd

Keywords--analytical models, fatigue, short cracks.

1. INTRODUCTION
THE PROBLEM of the short fatigue cracks received attention for the first time when Pearson [1]
observed that linear elastic fracture mechanics (LEFM) failed to correlate the crack growth rate
of very small cracks (0.006-0.5 ram) with that of long crack, as shown in Fig. 1. Since then
interest has heightened in the study of short/small fatigue crack behaviour [2-15].
It is now well recognized that small pre-existing defects are an inherent feature of engineer-
ing components and structures. They may be formed as a result of material forming and fabrica-
tion techniques, or may be an adventitious results of careless transportation or handling. The
size of such defects may range from the order of microns, for metallurgical inhomogeneities
such as nonmetallic inclusions, to several millimeters for the case of welding defects such as slag
inclusions or heat-affected zone cracks. Over the majority of this range of crack sizes the major
portion of the fatigue life is spent whilst the crack is smaller than the non-destructive inspection
(NDI) limit, as illustrated in Fig. 2. It is therefore convenient to define the upper bound of the
small crack problem to correspond to the general NDI limit of around 0.5-1.0 mm. The import-
ance of the short crack regime in fatigue is apparent in Fig. 2, and it is useful to apply the prin-
ciples of linear elastic fracture mechanics and in particular the defect tolerant design approach.
However, there are certain practical and fundamental difficulties in existing fatigue design codes
when including such small cracks.

1.1. Effect of closure and microstructure

Several authors have proposed that changes in crack closure are the cause of observed
short crack behaviour [19, 20, 22, 24]. The use of an effective stress intensity range (AKeer) is one
method of accounting for effect of closure on crack growth. Crack closure, microstructure and
plasticity on short crack growth may either act independently or in conjunction.
It is well understood that closure reduces the stress intensity range at the tip of a growing
crack through premature crack face contact. There are several mechanism that can cause this
premature crack face contact, and they can be classified as (i) plasticity-induced closure, (ii)
roughness-induced closure, (iii) oxide-induced closure, and (iv) transformation-induced closure.
The actual closure level of a particular crack may be a combination of more than one of these
mechanism.

tAuthor to whom correspondence should be addressed.

327
328 K. HUSSAIN

Long through section cracks

16' > 0.25/,m ~

./',,,
"t3 ~..~.'f /-- Short surlace cracks
-6
10 / 0.006-0.Smm deep
/
I
167 /
/
/
16s
' 16 ' 100
AK{MPa~/-m)
Fig. I. Short crack behaviour in aluminium alloy[l].

Plasticity-induced closure is considered to arise from the elastic constraint in the wake of
the crack tip of material elements permanently stretched within prior plastic zones, leading to
an interference between mating crack surfaces. Based on experimental compliance measurements
in 2024-T3 aluminium alloy, Elber [25] proposed an empirical relationship between the ratio of
the closure stress intensity to maximum stress intensity and load ratio. Lindley and Richards[26]

102
-7
Long Crock Regime
Crack Detection by NDI
E Starting from
E

Nonpropagating
Cracks
v

U
10-2 Short Crack Regime

1
Grain s~zes

I 1 I I

0 0.2 0,4 0,6 0,8 1,0

Fatigue Life Fraction

Fig. 2. Illustration of the importance of the small crack in fatigue. Where failure originates from micro-
defects, the majority of the fatigue life will be spent as cracks smaller than the NDI detection limit[16].
 Short fatigue crack behaviour 329

work indicated that the plasticity-induced closure plays a far less significant role in influencing
fatigue crack growth.
Roughness-induced crack closure arises due to the roughness of the fracture surface near
threshold stress intensity. Roughness-induced crack closure is promoted by small plastic zone
sizes at the crack tip which induce a single shear mechanisms, small crack tip opening displace-
ments, coarse grained materials and microstructure, crack deflection and low load ratios where
the minimum crack tip opening displacements may be substantially smaller than the size of the
surface asperities. Purushothaman and Tien [27] first modelled the roughness-induced closure.
However, the model does not incorporate the role of mode II displacements, which influence the
roughness-induced closure.
Oxide-induced crack closure concept is based on the phenomenon that at low load ratios,
near threshold growth rates are significantly reduced in moist environments, compared to dry
environment. Moist environment lead to the formation of oxide layers within the crack, which
are thickened at low load ratios by fretting oxidation. The formation of corrosion deposits and
the process of oxide-induced crack closure are promoted by (i) small crack tip opening displace-
ments (CTOD), (ii) highly oxidizing environment, (iii) low load ratios, (iv) rough fracture sur-
faces, and (v) low strength materials where the plasticity-induced closure mechanism is
dominant. A number of independent studies by Paris et al. [28], Tu and Seth [29], Skelton and
Haigh [30], Kitagawa et al. [31], Stewart [32] and Suresh and co-workers[17, 33-35] suggested the
possibilities of crack closure due to corrosion debris affecting near threshold growth rates and
AKth values.
Transformation-induced crack closure can result in the materials which undergo stress- or
strain-induced phase transformations under cyclic loading. Such transformation-induced closure
will be promoted by metallurgical phase changes which show a large increase in volume and by
conditions which enhance the transformation, i.e lower temperatures, higher strain rates, etc.
This closure effect arises from the constraint of surrounding elastic material on the transformed
regions. Hornbogen [36] and Mayer et al. [37] reported that the pronounced crack growth retar-
dation in the metastable alloy is due to the transformation of part of the austenitic phase into
martensite in the stress field of the crack tip, accompanied by a volume increase and conse-
quently, residual compressive stresses.
The understanding of the many sources of closure outlined above has enabled a much more
consistent explanation to be obtained for the effects of many mechanical, microstructural and
environmental factors which have been documented to influence near-threshold growth rates
and the values of the fatigue threshold. The direct experimental closure measurements have con-
firmed the role of crack closure at low load ratios with long cracks. At high load ratios or with
small cracks, the closure effect may well be severely diminished, non-existent or cannot be so
effective as in case of long cracks.
James and Morris [22] suggested that in mechanically and physically small cracks, large-
scale near-crack-tip plasticity or less extent of crack closure may be a main reason why small
cracks propagate appreciably faster than long cracks subjected to the same nominal value of the
stress intensity factor range. Jono and Sugeta [38] founded that fatigue crack longer than
0.2 mm were shown to have a residual tensile deformed region in the wake of the crack tip and
exhibited retardation after load reduction due to the difference in closure levels. However, for
crack smaller than 0.1 mm, closure is insufficient to control the crack growth behaviour and the
high level loading may give rise to an acceleration of crack growth.
In spite of different findings of different researchers, crack closure has been successfully
used to explain the anomalous behaviour of short fatigue cracks. The work of Dowson et a1.[39]
on the early stage of short crack growth suggests that after cracks outgrow the grain boundary,
the amount of crack closure increases significantly. It appears that roughness-induced closure
plays an important role in this case, due to crack deflection after crossing the grain boundary. It
has been noticed that a combination of plasticity- and roughness-induced closures dominates
the early stage of fatigue crack growth. According to Morris [40] stage I interaction with grain
boundaries is found to induce closure stress close to the maximum applied stress, explaining the
effective crack retardation of stage I cracks by grain boundaries observed for A1-7075. In these
cases the closure load was found to slowly fall as cracking continues by a stage II mode into a
second grain. For stage II cracking it is possible to estimate closure load from the cracking path
330 K. HUSSAIN

as observed from the surface. Tanaka et al. [41] suggested that the anomalous behaviour of
short cracks is due to reduction in the ratio Kop/Kmax as crack size decreases. Recently Gall et
al. [42] presented an elastic-plastic finite element analysis of mode I crack growth and plasticity
induced crack closure. The model shows that plasticity induced crack closure will occur when
the crack tip plastic zone is smaller than a grain size. Closure levels are found to vary as the
crystallographic orientation of the model with respect to the crack line is changed. For some
orientations the closure levels are found to be negligible, while in others the levels approach
0.35 of the maximum applied load. Closure levels also vary as the angle between the slip planes
is varied.
The variation in growth rates due to different crystallographic orientation is reported by
Tanaka and co-workers [43, 44] during the study of microstructurally small fatigue cracks. The
orientation of the individual grains in the direction of crack growth may increase, decrease or
arrest the crack growth. Since the growth behaviour of short cracks can influence the fatigue
predictions, it is very important to consider the microstructure to model short crack growth
behaviour.
Crack retardation at microstructural interfaces is one explanation for the initially decreas-
ing growth rates shown schematically for short cracks in Fig. 3. This decrease occurs despite the
fact that AK, for constant amplitude loading, is increasing steadily as the crack length increases.
This observation has been attributed to two factors [47]. The first is the difficulty in propagating
slip across an interface, which may give rise to an incubation period, depending on the type of
the interface, e.g. high-angle grain boundary or interface with a second phase. The second factor
is a possible increase in closure stress as the crack tip crosses an interface, caused by increased
non-continum crack tip plasticity due to slip extending to the next interface. In a closure domi-
nated situation the decrease in growth rates can be the result of an increase in closure stress i.e.
a reduce in AKeff. As the closure level approaches that of a long crack the increase in closure
stress reduces gradually and the growth rate increases.
The effect of heat treatment is clearly demonstrated in two studies made on 0.4% C
steel [2,48]. James and Smith [48] studied a quenched and tempered 0.4% C steel with ferrite
grain size of l0 pm. They observed no microstructural dependence of growth for small cracks.
Higher growth rates were found for cracks whose surface length was less than 100 pm. The
short crack effect was explained by change in closure. However, in a study of a normalized
0.4% C steel with prior austenite grain size of 140 #m, tested at 350-400 MPa shear stress, Rios

Notch root crocks

A
Increasing a' //
..l J/ Linear.growth

/'I
I \ \\/

AKth Long crocks

AK
Fig. 3. Schematics illustration of types of short crack behaviour (dashed lines) which may be observed
over the first millimeter or so of growth[16].
 Short fatigue crack behaviour 33

Fig. 4. Development of a surface crack. (a) Optical micrograph of a replica taken before the test com-
menced, showing the ferrite phase on the prior austinite grain boundary. (b) SEM micrograph of a
replica taken after 20 735 865 cycles. Two cracks, X and Y (arrows), have formed in the ferrite, and
were unable to propagate any further. Az = 350.51 MPa. (c) SEM micrograph, 46 763 cycles after rais-
ing the stress level to Az = 414.55 MPa. The cracks have now joined-up and branching has already
occurred at one end (X). (d) Optical micrograph, 649 907 cycles after raising the stress level. The crack
has now branched at both ends[2].
332 K. HUSSAIN

Fig. 5. Micrographs of crack initiation sites: (a) crack initiated along a persistent slip band; (b) crack
initiated along a twin boundary (acoustic microscope image); (c) crack initiated in a grain with no vis-
ible persistent slip bands[49].
 Short fatigue crack behaviour 333

Fig. 6. Sequence of crack development (plastic replica). Pearlite banding parallel to specimen axis (a)
before testing, (b) after 158 000 cycles, (c) after 235 000 cycles, (d) after 600 000 cycles[50].
334 K. H U S S A I N

Fig. 7. Crack arrest at pearlite region (plastic replica), parallel to specimen axis[50].

Fig. 8. Sequence of crack development in C - M n steel (plastic replica) under push-pull loading [74],
N~ = 27 000.
 Short fatiguecrack behaviour 335

et al. [2] observed microstructurally dependent growth with cracks arresting at ferrite-pearlite
interfaces. The cracks were initiated in the grain boundary ferrite and grew rapidly, at an almost
constant rate, until they reached the end of ferrite plates. The growth rate was dropped dramati-
cally and the crack arrested at the end of ferrite plate. Figure 4 shows the development of some
of these cracks. Although crack closure is an important factor it has not been measured for
0.4% C steel. Short crack behahiour was observed for cracks with surface lengths in the range
200-300/am, which corresponded to the lengths of ferrite phase in which the cracks were in-
itiated. In the study of nickel base super alloy, Wasplay, Yate et al. [49] observed three types of
crack initiation: (i) cracks developed directly from persistent slip bands; (ii) initiated along a
twin grain boundary; and (iii) initiated in grains which showed no evidence of slip bands, as
shown in Fig. 5. Although there were three types of cracks initiation, all cracks were initiated at
45 to the principal axis in a relatively large grain. The surface crack growth rates slowed down
and reached a minimum at the grain boundary. The grain boundary blocking effect was
observed up to a few grains. In 0.4% C steel[50] it was found that cracks were initiated and pro-
pagated along the slip band; these are shown in Fig. 6. The pearlite region was acting as barriers
to crack propagation, see Fig. 7. In a ferriti-bainitic microstructure, Hussain et al. [74] observed
the initiation of the crack in ferrite region situated at prior austenite grain boundaries. In this
microstructure cementite plates were found to be strong barriers to crack growth. The crack in-
itiation and propagation behaviour is shown in Fig. 8. de Lange [51] observed deceleration in
fatigue cracks when they become as large as four to five grain diameters. In a high strength
steel[52] and an aluminium alloy[21], retardation in growth was observed at crack lengths equal
to the grain size. Figure 9 shows short and long fatigue crack growth behaviour in different ma-
terials. Forrest and Tate[56] observed that cracks developed along slip bands and propagated in
shear mode, but slowed down at grain boundaries. The degree of retardation depends on the
grain orientation. It is well documented that microstructural size affects the mechanical proper-
ties of the material, especially the monotonic properties. The Hall-Petch [58, 59] equation is the
obvious example of the dependence of tensile strength on grain size. Similar approaches were
used to relate the fatigue strength to the grain size[45, 55-57, 60].

1.2. Limitations
The fundamental difficulties arise primarily from two factors: (a) application of LEFM to
high stress situations which usually characterize the growth of short cracks; and (b) the exper-
imental observation of growth of short cracks at values of the applied stress intensity range well
below the conventional "long crack" threshold values (AKth). Such growth may occur at rates
several orders of magnitude greater than might be expected, from an extrapolation of the Paris
law regime to include low stress intensity values, and can cause a component failures at fraction
of predicted life time. A basic condition of the use of LEFM is that applied stress should be
well below the yield stress and that the extent of crack tip plasticity should be small compared
with crack size and component dimensions. Associated with this is a possible noncontinum
aspect to crack tip plasticity when the crack size is small compared with some microstructural
parameter. In this situation several studies have indicated that the size of the crack tip plastic
zone may then be determined primarily by the microstructure rather than the applied stress
intensity factor[16].
The short fatigue crack problem in fact is one created by fracture mechanics through a
breakdown in the analysis of short cracks. Fracture mechanics analysis of fatigue crack growth
assumes microstructural independence. Kitagawa and Takahashi[3] and Taylor[ll] found that
cracks smaller than a critical length cannot be accurately defined by the LEFM and these are
termed as "short" cracks. Different researchers[9, 12, 16, 46] defined the short cracks in different
ways. In accordance with ASTM specification, if the plastic zone size (rp) > a/50, then the crack
(a) is short; conversely if rp is smaller than this ratio then the crack growth kinetics are predict-
able by LEFM.
According to James and Knott[16], when the growth of a fatigue crack from small initial
defect is analysed by LEFM, different types of behaviour may be observed over the first milli-
meter or so of growth, as shown in Fig. 3. In all cases, there is an initially high growth rate at
AK values well below the conventional long crack threshold. The growth rate may then decrease
to minimum as AK increases. Such anomalous growth behaviour has been observed in many
336 K. HUSSAIN

_~Io"
"D
>,, ~Crock stopped growing ,,
under the opphed /// 7
Ij
loading. /7/// Air __
E
E R=0.1 .///
z 100Hz\ /.0 3~t/~/// Z -9
16' ~o
O ,, f k z ~ L ~/3so p
"O \
,c; ~\ / i v t~ ,"'-[Scatter
\V ~. ljI bond) /ocuum
, ,lo), ~,ir (b)
16' 3 z. 5 6 7 8 9 1 0 15 20
16" 1 2 s ~6 2b
~ K (MPo '/'m ) ZkK (MPo ~r'm )

Oo
lx16 5

>,

lx10-6 o~ o 40/
E
E 0 0
0
"I0 lx~ 0

13 o 525 MPo
"I3
626 ..
lxll~ l 6~7 -
- Long crock growth~AK th
(c}
I l | I . . . . *

10 30 SO

AK, (MPo V-rn )

Fig. 9. Example of short and long crack behaviour in different materials: (a) aluminium bronze[53], (b)
7075-T651 aluminium alloy[53], (c) C - M n steel[54].

materials including steels [18-20, 49, 50, 67], aluminium alloys [21], titanium alloys [22, 23], nickel-
base super alloys[24, 45] and cast bronze[10]. The short crack regime can be defined as the crack
length which shows high growth rates as compared to the long cracks. The higher growth rates
may be observed when:
(a) Cracks are comparable in length to the scale of local plasticity. This is particularly the
case when the crack is growing in a region of bulk plasticity.
(b) Cracks are smaller in length to some microstructural parameter. This is essentially a
breakdown in the homogeneous continuum requirement for the application of fracture
mechanics[10].
(c) The crack is long in terms of microstructure and local plasticity, higher growth rates
have been observed for physically small cracks (0.5-1.0 mm). This has been proposed to be
due to a breakdown in similitude concept of LEFM.
 Short fatigue crack behaviour 337

Kitagawa and Takahashi [3] reported that at crack lengths greater than 0.5mm, the
threshold stress intensity factor for fatigue crack growth is constant. The intensity factor devi-
ates from values predicted by LEFM if the crack length is smaller than 0.5 mm.
Due to the limitations, discussed above, the LEFM cannot be used to analyse the short fatigue
crack growth behaviour. Several mathematical models have been developed along with the modifi-
cation of LEFM, to predict the anomalous behaviour of short cracks. This paper addresses the
critical view of the models and their application to analyse short fatigue crack behaviour. In some
cases a comparison is made between predicted and experimental fatigue lifetime.

2. ANALYTICAL MODELS
Several characteristics of short cracks distinguish them from long cracks and contribute to
differences in growth rate. One of these is that the plastic zone size at the crack tip of short
crack is large with respect to the length of the crack, which is the one violation of the conditions
prescribed for linear elastic fracture mechanics. The second violation, the overall applied load
levels, may be high with respect to the yield strength of the material, is the small scale yielding
appropriate to LEFM analysis. The third violation is that the level of crack closure is in a state
of transition as the crack changes from a short to long crack. Crack closure reduces the effective
range of the stress intensity factor, AKe~, and there will be a greater driving force for growth of
a short crack as compared to long crack at a given AK level[62].
Contrary to long cracks, short crack growth strongly depends on microstructure.
Investigators have attempted to modify LEFM to account for the behaviour of the short cracks.
Mechanisms based on crack closure stress and crack deflection, elastic-plastic approaches such as
the J-integral, or simple semi-empirical approaches have been introduced. The methods are suffi-
ciently successful when the crack length is only a few times that of the relevant microstructural par-
ameters. A number of crack growth models have been proposed to describe the short fatigue crack
propagation, taking into consideration the microstructural effect on the growth behaviour.

2.1. Model based on crack tip strain

Chan and Lankford [61] modified the LEFM equation to consider the variation on the grain
orientation and effects of the grain boundaries. The model was based on the assumptions that near
threshold stress intensity for a long crack, the crack-tip opening displacement (CTOD) is larger for
small/short crack than for a nominally equivalent long through-crack, i.e the plastic strain range
associated with a small crack is higher than that of a long crack. Considering the influence of the
crystallographic orientations of the neighbouring grains and the distance of the crack-tip from the
nearest grain boundary, the plastic strain range at the crack tip was defined as:

A~p CAK"'[1 -
J )]
- X "'

The crystallographic function K(~) in terms of resolved shear stress in the grain with small
crack (grain A) and its neighbouring grain (B) was defined as

K(~)= 1 ~B (2)
,t-A "

In this approach, the local plastic strain range at the crack-tip was used as a measure of
fatigue damage. Crack advance by the failure of a crack-ti~ element of size AX' occurs when the
accumulated local plastic strain exceeds a critical value ~p. The number of cycles AN required
for failure of crack-tip element was given by

p
AN = (3)
A~p

and the crack growth rate was defined as

da
d--N =
CIAK,,,[I_(I_~B~D-2Xy"I
rA J \ D .1 I (4)
338 K. HUSSAIN

AX'C
Ci - , (5)
/3

From a physics and mechanics point of view, Chan and Lankford's model is a gross ap-
proximation to reality. The model predicts little or no deceleration in growth rate in the case of
similar orientation of grains. A comparison between long crack (analysed by LEFM) and short
crack growth behaviour [analysed by eq. (4)] is shown in Fig. 10.

2.2. Model based on modification of L E F M

McEvily et al. [62] proposed a short crack analysis considering large-scale plasticity effects,
crack closure and the fatigue crack growth threshold. The proposed short fatigue crack analysis
method is based on modification of elastic analysis in the presence of large-scale plasticity. They
used Irwin's expression [63] to analyse the experimental data as

K = limp_0KTtrrtV/-
~ (6)

which relates the stress intensity factor (K), to the stress concentration factor (KT). In this ex-
pression ~r is the applied stress and p is the radius of the stress raiser. Irwin has shown that it is
possible to develop a variety of stress intensity factors appropriate to large cracks from a knowl-
edge of the corresponding stress concentration factor when the radius p is allowed to approach
zero. In a case of a central notch (2a) wide sheet specimen loaded in tension, KT was defined as
KT = (1 + 2x/{a/p}) and therefore eq. (6) can be written as

K = limp_0 (1 + 2 ~']a_ ~//-~ (7)

\ VP.] V 4

10-s

K(~) = 0

KI~] : 0.3

(,3 =0.B
U
m=2
E 10.9 CI 3 ,, 10-9
n =0.2
z" ,',o" = 3t.9 NIPo
D = ,',0 p rn

KI'}) = ~ . 0 [

J
crack a r r e s t at
grain boundary

10-1o
2 5 10
AK, MPo -m
'o /.'0 '
BO 100
i 2 0 , pm

Fig. 10. The crack-tip strain model is used to simulate the anomalously rapid growth of small crack.
The simulated behaviour also includes crack growth perturbation including arrest at a grain
boundary [61].
 Short fatigue crack behaviour 339

which leads to the following expression:

K = ~v~. (8)

In their model, instead of allowing p to approach zero, they allow it to approach a finite value
Pc. This modification was justified as follows:
1. the fatigue crack tip is not of zero radius even at the minimum load;
2. the large plastic zone to length ratio of a short crack requires modification of the LEFM;
3. the stress range required to propagate a crack must remain finite rather than go to infinity
when the crack length is extremely small.
The "'K" was defined as

K= t~/--~-+ Y~-a ~. (9)

In the propagation of short crack, the applied stress levels are high with respect to the yield
strength of the material. As a result the plastic zone will be large than predicted by LEFM, and
therefore the effective crack length, aefr, i.e. [a + (1)/(2)rp] was used to calculate stress intensity
factor as

K=~/~-~-I-Y77Z~ a(sec ( ~'max

2~y "~/ + l)cr. (I0)
The short fatigue crack growth rate was defined as:

da = A(AKeff - AKemk)2 (11)

dN
where A is material constant, AKefr=Kmax-Kop and /Cop (crack tip opening
K) = (1 -e-~l})Kopmax, 0 _< gop-<Kopmax provides a means for taking into account the develop-

(a)

10

-1 ,,, ~ l If /~s "

/ .i /.4 .
/ ." .,4.'/,<;
~~65 .'//,<7
, .--'--
=o,,i ,...--
"

/..""
/ . ' J~ T l"
.,,4/ i K ,

/~ i/
.lily

lo'
-...joo/i
"3W"-
"

460 " ~ Experimental results

--- Colculated
1(~r -T - - - Colculoted
10
i I | i J
I I I I i
0,01 0.05 0,I 0,5 1,0 0,05 0,1 0.2 0,5 1,0 1,5
a (rnm) o (ram)

Fig. 11. A comparison of calculated and experimentalcrack growth rates in 1045 steel: (a) plain speci-
men, (b) specimencontaining0.1 mm hole[62].
340 K. H U S S A I N

ment of closure during the transition period. Kop is the crack opening level transition region, I is
the length of the short crack, is the parameter of unit mm -I and Kopmax is the crack opening
level for a long crack and is a function of stress ratio R. Then eq. (11) becomes

da _ A[AK - (1 - e-~")Kopma - AKeffth]2. (12)

dN
By using eq. (10), the above equation becomes
da [" ~l~e /-~ [ rca ) ]2
dN-LV4+ Y ~ / ~ a ~ s e c ~ a r + l Aa-(1-e-~'l)Kopmax-AKeffth . (13)

This equation is applicable to both long and short fatigue cracks. The graphical presen-
tation of eq. (13) and the experimental data is shown in Fig. 11. The model predicts crack
growth behaviour in a continuum manner, but does not deal with the usual acceleration and
deceleration behaviour of short fatigue cracks. Since LEFM is unsuitable for application in
short cracks, the AK approach becomes questionable in the analysis of short crack growth.
Although McEvily's approach is interesting in its own right, it has not been adopted by other
researchers.

2.3. Model based on strength of a slip

de los Rios et al. [2] proposed that crack growth rate is proportional to the strength of the
slip band, assuming that from a smooth specimen crack will initiate from slip band. The shear
stress (r) at slip band of length (L) and strength (ndb) was assumed as
pndb
~= /~ (14)

or

~" = g'~app - - TO (15)

where ~ is orientation factor, "lapp the applied shear stress, nd the number of dislocation, b the
Burger vector and To internal friction stress. The growth rate was defined as
da rL
dN - f l p (16)

where # is the shear modulus, " a " the crack length, fl the fraction of dislocation on the slip
band which takes part in the process of crack extension, and N the number of cycles. After a
number of cycle, when the crack length is equal to "a", the length of slip band was shortened to
L - a, and the crack growth rate was defined as
da z(L - a)
- - (17)
dN -f~

A comparison of experimental crack growth and predicted growth rate by eq. (17) is shown in
Fig. 12.

2.4. Model based on energy release rate

Later, de los Rios et al. [50] developed another model based on load and bulk energy con-
siderations, assuming that the local energy, i.e. the energy at the slip band, should be equal to
the crack extension energy for crack propagation. A second consideration of the model was the
nature and strength of the barriers that should be overcome to propagate the crack into the
next grain. The microprocesses of crack growth need to be related to the mechanics of crack
extension. The rate of energy released in a cycle by the system, in order to extend the crack
(da), is the energy released rate (G) and was defined as
dU
a = -- (18)
dR
where U is the energy of the slip band and dU/dR is the rate of change of slip band energy with
 Short fatigue crack behaviour 341

Q, b,
@
i0-~

10-7

10- / i/
J 0,01 0,1 0,01 0,1
e,,,

dJ
o 16'-
0

~0-7I /
8"I"=399 MPa / AT=399 MPo !
|
16 8 , l 1080.0] 0.I I
0,01 0.1
Crock Length~ rnrn
Fig. 12. Application of the model to various short crack cases in a normalized medium carbon steel;
the solid line represents predicted growth behaviour[2].

slip band length. During one cycle the slip band associated with the crack was simulated to a
linear distribution of close packed dislocation and was defined as

U = 1 [R/j(ndb)2 dr (19)
2 J,.0 r
where ~ is the shear modulus, nd the number dislocation in the slip band, b the Burgers vector,
r0 the core radius of the first dislocation, R the slip band length, and r the variable length in
front of the crack and along the slip band. Then eq. (19) can be written as
/~(ndb) 2 r,
U -- -~ Lm R - In r0]. (20)

As the crack length " a " increases, R decreases, and therefore the change of the energy of
the system with " a " may be equated to the change of the slip band energy with R and following
the Griffith approach as

dU #(ndb)2 -- G = - T- 2r t a (21)
dR 2R /~

ndb = (2rtaR) 5-r. (22)

#

By making the crack extension per cycle proportional to the displacement along the slip
band, the crack growth rate was defined as
da
= fl (2~aR) 5 -~ (23)
dN /~
342 K. HUSSAIN

The value of R, the effective slip band length, depends on the distance to the next barrier
and also on the strength of the barrier which in turn depends on the ratio of the resolved shear
stresses along the slip band plane on the two grains separated by the barrier. The argument
used by Chan and Lankford[61] to define the change in crack tip plastic strain, was used to de-
fine R in eq. (23), i.e.

R=D[1- ~b(-D-~)] '' (24)

where D is the distance from barrier to barrier and 4) is a function related to the relative crystal-
lographic orientation of the two grains; the grain (A) where the crack develops and grain (B) in
which the crack will next propagate. Function 4, was defined, in terms of the threshold shear
stresses along the slip bands, as

4) = 1 Tr}. (25)
"cA

q5 = 1 when the slip orientation of the next grain is most unfavourable and the crack will arrest.
If ~b = 0, it corresponds to the case where the orientation of the two grains is similar, i.e.
rA = 'cB. The combination of eqs (23)-(25) gives the following:
da
~1 ~)(D-X))
0.5/
-d-~=fl(2rcaD) D m~" (26)

A comparison of the experimental and predicted growth behaviour by eq. (26) is depicted in
Fig. 13.

O, b,

E
10

16'

16
) 10

1(3 6
o o
o
= 0,75
E = 1.0 1E
O ...... .9 ....... J ...... i . . . . . . . I i ....... | -

n,, 0.01 0.1 1.0 0.01 0,1 1,0

t'- C, d,
_~ ' o)

10

0
o
b..
o
10-'
I( o
16 =0,85
I , ,l,,d . . . . . |11] I m - . ,,,,.J , , ...... !

0.01 0,1 1,0 0.01 0.1 1,0

Crock Length mm
Fig. 13. Application of the model to various short crack cases in banded ferrite-pearlite microstructure
of 0.4% C steel; solid line represents predicted growth behaviour[50].
 Short fatigue crack behaviour 343

de los Rios et al. in their both models [2,50], proposed that crack growth rate is pro-
portional to the strength of the slip band [2] and bulk energy considerations, assuming that the
local energy, i.e. the energy at the slip band, is equal to the crack extension energy for crack
propagation. A second consideration of the model is the nature and strength of the barrier that
should be overcome to propagate the crack into the next grain [50]. Both models are successful
in predicting growth rate for individual cracks studied from replicas.
In smooth samples there are usually more than one cracks, each growing at different
growth rate and facing grain or second phase barriers of different orientation. Failure occurs
when these cracks combine into one dominant crack.

2.5. Model based on grain boundary effect

Hobson [64] used a statistical approach to accommodate this factor and proposed two
equations, one for "short" crack growth and the other for "physically" small cracks. The short
crack growth equation is

da---2s= C2(d - as) (27)

dN
where "as" = surface crack length, C2 = material constant in short crack region and d represents
the distance to the first microstructural barrier affecting crack growth, which may be a grain
boundary or any other metallurgical obstacle. This equation assumes that crack arrest, tempor-
ary or permanent, will occur at the grain boundary, when a = d i.e. (da)/(dN) = 0, depending
upon the applied stress level. This equation is dimensionally correct.
The crack growth behaviour of the crack length greater than the length "d" was expressed
by the following equation:
das
= C3as - DI (28)
dN
where C3 is the material constant and is a function of stress/strain and D~ represents crack
growth threshold. To calculate the value of Dl it was assumed that at the fatigue limit (das)/
(dN) -- 0 and that this happens when crack length equals the microstructural barrier distance
"d". Taking as = d, at fatigue limit (aVl) eq. (28) becomes
0 : C 3(O'FI)'nd -- DI

DI = C 3(trvl)"'d. (29)

Application of eqs (27) and (28), on 0.4% C steel tested at low and high strain rates and
on C-Mn steel tested under uniaxial loading is shown in Figs 14, 15 and 16.
Hobson calculated the value of "at' for individual crack by plotting the values of (das)/
(dN) = 0 against average as on a linear scale. Then a least-square fit was performed on those
data points where the growth rate was decreasing in relation to crack length. The value of "at'
was taken to be that value of as, where the extrapolated least-square equation intersected the
crack length axis. For a physical small crack, crack growth data corresponding to crack lengths
greater than the parameter "d" was used to derive the physical small crack growth expression.
It was assumed that the growth of cracks longer than "ae' was not affected by microstructure,
which results in shorter lifetime prediction than the actual fatigue life. The method used to cal-
culate the value of "at' is a numerical procedure, which gives the value of "at' different to the
actual microstructural length, i.e the grain size.
The assumptions and method proposed by Hobson yields good agreement with experimen-
tal data if the microstructural effect on the short crack behaviour is considered in the first grain
only. It has been observed by the author and reported by other researchers that the crack
growth behaviour is affected by the microstructures even when the crack is longer than one
grain diameter[10, 67-69].
344 K. HUSSAIN

50 0 Hghsto [kb .soo

AEt = 0,0039
A Ct = 0,026 006

J
u
u
Eq.27
E

q)
20t \ I I \ Eq.(27+28)
O
0,02

l0
'th \ Eq.28 I I kt Eq.28
0 th

0 SO I00 10 0 50 I00 150

e s (pro) e$()am)

Fig. 14. Crack growth characteristics at (a) high, and (b) low strain range[70].

2.6. Model based on plastic zone &teraction

Navarro and de los Rios[71] proposed a model which can describe the short and physically
small crack growth behaviour by a single equation. The usual behaviour of short fatigue cracks,
acceleration and deceleration, is described by successive blocking of the plastic zone by the
grain boundaries and the subsequent initiation of the slip in the next grain. Crack growth was
defined by
da
=f2~b I. (30)

Factor f2 represents the degree of irreversibility of slip on each stress cycle and can be equated
to the fraction of dislocation which are drawn into the crack during each reversal of stress cycle.
~bl is the crack tip displacement and is expressed as

= aa (31)
n

where G is the shear modulus, x = 1 for screw dislocation and 1 - v for edge dislocations, "a"
is the crack length and a is the applied stress. As the crack grows, while the plastic zone is
blocked by the grain boundary, the value of "n" increases in magnitude up to a critical value
(no), where the stress concentration is sufficiently high to activate a dislocation source in the
next grain; this is defined as

( ~ a _ _- _at.i)
_ (32)
nc = cos 20"comp

O"FI
 Short fatigue crack behaviour 345

i0-i A t = 2 5 0 MPa.

/
//
Eq.27
Io-s
QP

Eq.28
U

E
E 10-6
z

10-7
. . . . Model
I Experimental
(Conttmnt amplitude tests)

10-8 I I
0.01 O.i 1.0 I0.0
Surface crack length (rnm)
Fig. 15. Experimental and predicted short crack growth rates. Different symbols represent different
cracks and solid line shows predicted growth behaviour[66].

where O'compis the comparison stress, which is equal to flow stress within the plastic zone, and
~rFL is the fatigue limit. As slip initiates in the next grain the plastic zone extends across to the
grain boundary. The stress concentration ahead of the new extended plastic zone is defined as
i
ns = nc. (33)
t+2
where i = 1, 3, 5,..., the number of the half grains. The transfer of slip is repeated grain after
grain; crack growth accelerate (after slip is produced in the next grain) and decelerates (as the
crack approaches the grain boundary) during the process until failure. However, as the crack
approaches failure, the acceleration and deceleration in growth rate decrease.
A schematic of the Navarro-Rios model is shown in Fig. 17, which represents the decelera-
tion and acceleration behaviour at grain boundary and soon after crossing the grain boundary,
respectively, along with upper and lower curves. Figures 18 [71] and 19 [72] show the compari-
son of the experimental and calculated growth behaviour by the Navarro-Rios model. The the-
ory of the Navarro-Rios model is based on the assumption that the microstructural features
will affect the growth rate until failure of the specimen. However, it has been observed by the
author and reported by other researchers that after a certain length, cracks are not influenced
by microstructure. Hussain et al. [67] observed that effect of microstructure is prominent up to
the crack length equal to 3-4 grains diameter. Taylor and Knott [10] observed that growth
characteristics of those cracks which are shorter than 10 times the grain diameter, depending on
the surrounding microstructure. Keiro et al. [68] reported that when the surface crack length is
longer than three grain diameters, crack growth rates are not influenced by the microstructure.
According to Daeubler and Thompson [69], when a crack changes from stage I to stage II it is
no longer affected by microstructure.
346 K. HUSSAIN

xlO-I

Eq.27 Eq.28
xl0-=

V 0

x I0 "~ ~-I. x

E I

x 10 "4
z II

tn
0 I,
IN
E
x 10 "j

x 10"*
I I 4 I llIPl
x 101 x 10~ x 10~

Surface crack length (IJrn)

Fig. 16. Experimental (solid curves) and predicted (dashed curve) short crack growth behaviour[65].

2.7. Two-stage micromechanics model

On the basis of the observation discussed above, Hussain et al. [67] proposed a two-stage
short crack growth model. The first part of the model deals with short crack growth behaviour
dependent on the surrounding microstructure. The stress concentration in this region was
defined as in eq. (32). The second part of the model represents physically small crack region
(where growth rate is not influenced by its surrounding microstructure). The stress concentration
ahead of the crack tip will always attain a value sufficiently high to overcome the barrier
strength and therefore the growth rate will be independent of the microstructure. The value of
nc =n, which can be calculated using Bilby et al.'s relationship[73], as

n = cos . (34)
0"comp

In the short crack region (first stage), where crack growth decreases as the crack approaches
the grain boundary and increases when slip is activated in the next grain, the value of n is
defined by the position of the crack tip relative to the grain boundary. This is expressed by
eqs (32) and (33).
The crack growth rate in terms of the surface crack length is expressed as
das
dN = f2 q~* (35)
 Short fatigue crack behaviour 347

1000
n=n s

_~ 100

..c

3:
o
o
10

1 I / i i i i i Ii II I I I ii

1 10 20 50

No. of Grains

Fig. 17. Schematic of oscillating crack tip plastic displacement and upper and lower limit curves[71].

where
_, ~/1 - n2
4' -- - - Art.as. (36)
G n

In the Hussain-Rios[67] two-stage model, the expected deceleration and acceleration in the
crack growth, as the crack approaches grain boundaries and where slip is transmitted to the
next grain, is assumed to occur solely for the first three to four grains. The model predicts not
only the author's experimental observations but also supports the findings of others that short
crack growth is affected by the grain boundary up to a certain number of grains. The model can
be used to predict short fatigue crack growth behaviour in uniaxial as well as in torsional load-
ing. The constant x = 1 for screw dislocation (uniaxial loading) and r -- l - v for edge dislo-
cation (torsional loading). The experimental crack growth data and predicted curve is shown in
Figs 20 and 21, both in torsional and uniaxial loading.
The Hussain-Rios [67] model was latter modified to define growth behaviour of individual
crack. The crack length with minimum growth rate was considered to be the length (d) between
the barriers to growth. This length was found different for different cracks and it varied within
the one crack depending upon the size of the grains through which it propagated. To accommo-
date this experimental observation, the parameter "n" for a particular length, in short crack
region, of an individual crack was defined as[74]:
348 K. HUSSAIN

os S J

" o=.c
...- .-,o

A &O

Q

(.3

I| o 0 , o . .|

40 !00 500 1000 2000

Surface crack length as(Wn)
Fig. 18. The experimentaldata of Hobson et al. [70], for differentcracks (distinguished by differentsym-
bols) together with his own prediction (dashed line) and the prediction derived by Navarro and de los
Rios[71].

7 S
n= k (37)

i=-k

Equation (37) was used in eq. (36). The factor '~r2" was calculated for each stress level. The
curves in Fig. 22 were calculated using eqs (35)-(37). The calculated growth rates match reason-
ably well with the experimental results and showed a systematic decrease in growth rate at grain
boundaries.

3. SUMMARY
The propagation behaviour of long fatigue cracks is uniquely determined by the range of
elastic stress intensity factor AK. However, when the crack length is small, the growth rate is no
longer predictable from conventional AK-based fracture mechanics. Short cracks with dimen-
sions on the order of the material microstructure (grain size) are called microstructurally short/
small cracks. They show anomalously high growth rates when compared with long cracks at the
same AK value. The assumption of macroscopic continuum in fracture mechanics breaks down
for these microstructurally short cracks.
For microstructurally short cracks, the resistance to crack growth is much influenced by the
material microstructure such as grain orientation and grain boundary. The crack growth rate
will be different even when the crack is subjected to the same loading parameter of continuum
mechanics, AK and AJ. Local parameters will be necessary to express the crack driving
force [44]. Tanaka e t al. [75] calculated ACTOD and derived irregular growth of microstructu-
rally small cracks by assuming the crack growth range as a power function of ACTOD. Cracks
decelerate as they approach the grain boundary. When the crack length is larger than several
 Short fatigue crack behaviour 349

0,1 --z-cZ-____'----f ,' 0.1

o
O,Ol . . . . _
0.01
y.,
o
0.001 ~!//

o.o~ ;~9
~, 0,0001
"' ~ '
,-, 10 1000 10 1130 i
1000
f

0.1
t'r.

0.01
,t-

O
,5 0,001

0,0001

ODO001
c;oo
Surface Crack Length, ~rn

Fig. 19. Crack growth rate of the failure cracks (ferrite-pearlite, ~ ferrite-bainite microstructures)
plotted against crack length together with theoretical predictions of crack growth rate for crack tip pos-
ition near barrier (lower curves) and after crossing the barriers (upper curve): (a) z/'Cyc=1.29, (b) r/
ryc= 1.16, (C) Z/'~yc = 1.11 [72].

times the grain size (mechanically short crack), the resistance of the materials is regarded as
homogeneous. For such cracks an appropriate choice of fracture mechanics parameters such as
AKefr or AJ will make crack growth prediction possible. All of the crack closure mechanisms in-
dicate that the crack wake made by fatigue is responsible for crack closure. Since a short crack
has a short wake, the amount of crack closure is expected to be smaller and therefore, the
growth rate is higher for short cracks. Premature crack closure plays a dominant role in anoma-
lous, fast growth behaviour of mechanically short cracks[44].
Morris et al.[76] proposed a model of roughness-induced closure of short cracks, by consid-
ering the measured value of surface roughness. The analysis of the plasticity-induced closure of
short cracks indicates that the opening stress is lower under higher amplitude of stress
cycling[77]. The acceleration of crack growth due to a small amount of macro-plasticity is pre-
dictable from an increase of AKefrvalue. When the amount of cyclic plasticity is large, crack clo-
sure adjustment is not enough; some elastic-plastic mechanics parameters are necessary to
evaluate the crack growth rate. Among several parameters proposed, the J-integral appears to
be most successful. The AJ can be estimated from measurement of the cyclic hysteresis loop and
crack closure [6,78]. The amount of crack closure is smaller for shorter cracks under larger plas-
tic strain. Hoshide et al. [78] attributed the growth acceleration of microstructurally short cracks
350 K. HUSSAIN

x10 -I

xlO_ 2 S

g , ,,:Z
"E.-.a x10.3 S +x " +A A ~

xlO -4 s E \32 " !O~ 141 ~

A O+
u 1
2
x10 "s 5 &

xl0 -6 . . . . . . . . . . . . . . . |

5 I 5 I
xl01 x 102 xl0 3

Surface crack length, IJrn

Fig. 20. Crack growth rate vs surface crack length in torsion; Ar = 319.5 MPa. The curve represents
predictions using two-stage model. Different symbols represent experimental behaviour of different
cracks[67].

to the difference in crack growth micromechanisms. A dominant mechanism is the intergranular

fracture for short cracks, while the transgranular fracture with striations for long cracks.
Any predictive method of the crack growth behaviour of microstructurally small cracks
must include the statistical nature of the interaction of cracks with the material microstructure.
Chan and Lankford[61] model is based upon the concept of greater plastic strain and the
grain orientation. The reduced crack tip strain at grain boundaries causes deceleration in micro-
crack growth rate. However, if the orientation of the adjacent grains are same, the model pre-
dicts little or no decrease in growth rate.
The model proposed by McEvily et al. predicts crack growth behaviour in a continuum
manner, but does not deal with the usual acceleration and deceleration behaviour of short fati-
gue cracks. Since LEFM is unsuitable for application in short cracks, the AK approach becomes
questionable in the analysis of short crack growth.
de los Rios et al. in their models proposed that crack growth rate is proportional to the
strength of the slip band [2] and bulk energy considerations, assuming that the local energy, i.e.
the energy at the slip band, should be equal to the crack extension energy for crack propa-
gation. A second consideration of the model is the nature and strength of the barrier that
should be overcome to propagate the crack into the next grain [50]. Both models are successful
in predicting growth rate for individual cracks.
The assumptions and method proposed by the Hobson model yields a good agreement with
experimental data if the microstructural effect on the short crack behaviour is considered in the
first grain only. Hobson calculated the value of "d" for individual crack by plotting the values
 Short fatigue crack behaviour 351

~I0 -~

A A

Eq.32
~_
u
xi0-2
e
E 1
~ o
+

o
e" xi0-3 A o

',,e
U
o 5
..)

xIO ,,a,

5 1 5 I 5
x101 x10 2 xlO 3

Surface crack length, ~Jm

Fig. 21. Crack growth rate vs surface crack length in push-pull; Aa = 626.8 MPa. The curve represents
predictions using two-stage model. Different symbols represent experimental behaviour of different
cracks [67].

of das/dN against average as on a linear scale. Then a least-square fit was performed on those
data points where the growth rate was decreasing in relation to crack length. The value of "d"
was taken to be that value of as, where extrapolated least-square equation intersected the crack
length axis. For physical small crack, crack growth data corresponding to crack lengths greater
than the parameter "d" was used to derive the physical small crack growth expression. It was
assumed that the growth of cracks longer than "d" was not affected by microstructure, which
results in shorter lifetime prediction than the actual fatigue life. The method used to calculate
the value of "d" is a numerical procedure, which gives the value of "d" different to the actual
microstructural length i.e the grain size.
In the Navarro-Rios model the microstructural influence in the short crack regime extends
into the physically small crack regimes. For longer crack lengths, the difference between the
upper and lower crack growth curves is negligible and the theory is similar to Hobson's
approach.
Hobson considers only one barrier in the short crack regime. It appears simpler than the
Navarro-Rios model, because it considers the average growth rate, ignoring the crack growth
variations near the grain boundaries. In some respects it is more difficult since two parameters
for each equation have to determined, including the separation of the data into two regimes, to
apply the model.
The Navarro-Rios model relies only on using the data from the physically small crack
regime to derive the expression '~f2" in the terms of applied stress, which in turn is used to
352 K. H U S S A I N

xlO &o" = 687 MP=

x161l
O 0o 0 0 o 0 0
OO ~
o

xlo
ZXO"= 626 M Po.
:2,.

0 0 0 0 0
=_

g
.
o

g
0

&(~= 559 MPa

xl04

xl~)3 8
00 o ,,e
oe
el lw~" o Exp.
-5 Cal.
X10 i

120 20 320 420 520 620 720

Surface creck tength { pml

Fig. 22. C o m p a r i s o n of e x p e r i m e n t a l a n d calculated g r o w t h b e h a v i o u r o f short crack at different stress
levels [74].

describe the crack growth in short and physically small crack regions. The theory behind the
Navarro-Rios model is based on assumption of the continuous effect of microstructures up to
failure. This assumption becomes questionable in the presence of the experimental observation
that microstructural effect on growth behaviour is up to a certain length of crack.
The Hussain-Rios two-stage model is in fact a combination of the Hobson and Navarro-
Rios methods. Both its theory and application are simpler. In this approach the expected decel-
eration and acceleration in the crack growth is assumed to occur solely for the first few grains.
The factor "f2" is calculated from second-stage crack data, where grain boundary effects on
growth are assumed to be zero, and is used to describe the short crack growth behaviour which
considers the crack tip grain boundaries interaction. The model is easier and gives fatigue life
prediction comparable with experimental fatigue life. See Table 1 for a summary of the models.
For a detailed comparison between the Hobson, Navarro-Rios and Hussain-Rios models see
Ref. [67].

T a b l e 1. C o m p a r i s o n o f fatigue lifetime u n d e r t o r s i o n a l loading predicted by different models[67] for a final c r a c k size of

4 mm. The c o n s t a n t used for the c a l c u l a t i o n are f2, C2 and C3

Hussain-Rios model H o b s o n model N a v a r r o - R i o s model

Ar A c t u a l life
(MPa) time Factor.f2 Lifetime C2 C3 Lifetime Factor.f,_ Lifetime

417.58 15450 1.35 10- t 11 976 1.67 10-2 9.11 10-5 29102 7.1 10-2 21 208
369.58 96988 2.12 x 10-2 95 357 7.58 10-5 2.35 10-5 121 198 1.1 x 10-2 124043
357.50 177000 1.28 10-2 167971 3.26 10-4 1.63 10-5 184201 6.6 10-3 260908
344.94 362 112 7.46 10-3 306 874 1.32 10-4 1.1 10-5 292 907 3.83 10-3 463 902
328.95 702 462 3.62 10-3 686 390 3.8 10-5 6.5 10-6 568 570 1.86 10-3 1 035 666
319.48 1220000 2.33 10-3 1 122774 1.9 10-5 4.7 10-6 894252 1.19x 10-3 1446246
306.82 2290000 1.26 10-3 2222889 6.8 10-6 3.0x 10~ 1 867741 6.42 10-4 2613924
 Short fatigue crack behaviour 353

REFERENCES
1. Pearson, S., Engineering Fracture Mechanics, 1975, 7, 235-247.
2. de los Rios, E. R., Tang, Z. and Miller, K. J., Fatigue Engineering and Material Structures, 1984, 7, 97-108.
3. Kitagawa, H. and Takahashi, S., In Proc. ICM2, Boston. U.S.A,, 1976, pp. 627-632.
4. Dowling, N. E. and Begley, J. A., A S T M STP, 1976, 590, 82-103.
5. Brown, M. W., Fatigue Engineering and Material Structures, 1986, 1, 93-103.
6. Dowling, N. E., A S T M STP, 1977, 637, 97-121.
7. Wells, A. A., In Proc. Crack Propagation Syrup., Vol. 1, Cranfield, 1962, pp. 210-230.
8. Ewalds, H. L. and Wanhill, R. J. H., Fracture Mechanics. Arnold, London, 1986.
9. Miller, K. J., Fatigue Engineering and Material Structures, 1982, 6, 223-232.
10. Taylor, D. and Knott, J. F., Fatigue Engineering and Material Structures, 1981, 4, 147-155.
11. Taylor, D., Fatigue Engineering and Material Structures, 1982, 5, 305-309.
12. Chauhan, P. and Roberts, B. W., Metallic Material Technology, 1979, 131-136.
13. Smith, R. A., A S T M STP, 1983, 811, 264-279.
14. Lankford, J., Fatigue Engineering and Material Structures, 1985, g, 161 175.
15. Lankford, J., International Journal qfFracture, 1980, 16, R7-9.
16. James, M. N. and Knott, J. F., In AGARD(NATO), (CP-393), 10.1-10.12, October 1985.
17. Suresh, S. and Ritchie, R. 0., International Metal Review, 1984, 29, 445-475.
18. Kitagawa, H., Takahashi, S., Suh, C. M. and Miyashita, S., A S T M STP, 1979, 675, 420-449.
19. Breat, J. L., Mudry, F. and Pineau, A., Fatigue Engineering and Material Structures, 1983, 6, 349-358.
20. James, M. N. and Smith, G. C., in Proc. 6th. Int. ConJbrence on Fracture, New Delhi, India, 3, 1984, pp. 2117-2124.
21. Lankford, J., Fatigue Engineering and Material Structures, 1982, 5, 233-248.
22. James, M. R. and L Morris, W., Metallargy Transactions, 1983, 14A, 153 155.
23. Brown, C. W. and Hicks, M. A., Fatigue Engineering and Metallic Structures, 1983, 6, 67-76.
24. McCarver, J. F. and Ritchie, R. O., Material Science and Engineering, 1982, 55, 63-67.
25. Elber, W., A S T M STP, 1971, 486, 280.
26. Lindley, T. C. and Richards, C. E., Material Science Engineering, 1974, 14, 281.
27. Purushothaman, S. and Tien, J. K., In Strength of Metals and Alloys, ICSMA5, Con. Proc., Vol. 2, ed. P. Haasen et
al. Pergaman, New York, 1979, p. 1267.
28. Paris, P. C., Bucci, R. J., Wessel, E. T., Clark, W. G. and Mager, T. R., ASTM STP, 1972, 513, 141.
29. Tu, L. K. L. and Seth, B. B., Journal of Testing and Evaluation, 1978, 6, 66.
30. Skelton, R. P. and Haigh, J. R., Material Science and Engineering, 1978, 36, 17.
31. Kitagawa, H. Toyohira, S. and Ikeda, K., In Fracture Mechanics in Engineering Applications, ed. G. C. Sih and S.
R. Valluri. Sijthoff and Noordhof, The Netherlands, 1981.
32. Stewart, A. T., Engineering Fracture Mechanics, 1980, 13, 463.
33. Ritchie, R. O., Suresh, S. and Moss, C. M., Journal of Engineering Material Teehnology, Transctions o! the ASME,
1980, H(102), 293.
34. Suresh, S., Zamiski, G. F. and Ritchie, R. O., Metallurgy Transactions, 1981, 12A, 1435.
35. Suresh, S., Parks, D. M. and Ritchie, R. O., In Fatigue Thresholds, ed. J. Backlund, A. F. Dolom and C. J. Beevers.
EMAS, Warly, U.K., 1982, p. 392.
36. Hornbogen, E., Acta Metallurgica, 1978, 26, 147.
37. Mayer, H. R., StanzI-Tschegg, S. E., Sawaki, Y., Huhner, M. and Hornbogen, E., Fatigue Fracture and Enghteering
Material Structures, 1995, 18(9), 935.
38. Jono, M. and Sugeta, A., Fatigue Fracture and Engineering Material Structures, 1996, 19(2/3), 115.
39. Dowson, A. L., Halliday, M. D. and Beevers, C. J., In Proc. IRC'92, Birmingham, U.K., 1992.
40. Morris, W. L., Metallurgy Transactions. A, 1977, 8, 1087.
41. Tanaka, K., Nakai, Y. and Yamashita, M., International Journal of Fracture, 1981, 17, 519.
42, Gall, K., Sehitoglu, H. and Kadioglu, Y., Acta Materiala, 1996, 44(10), 3955.
43. Tanaka, K., Hoshide, T. and Sakai, N., Engineering Fracture Mechanics, 1984, 19, 805.
44. Tanaka, K., A S T M STP, 1989, 1020, 151.
45, Hicks, M. A. and Brown, C. W., Proceedings of Fatigue, 1984, 84(3), 1337-1347.
46, Ritchie, R. O. and Suresh, S., Material Science and Engineering, 1983, 57, L27-30.
47. Morris, W. L., James, M. R. and Buck, 0., Metallurgy Transactions, 1981, 12A, 57-64.
48, James, M. N. and Smith. G. C., International Journal of Fatigue, 1983, 5, 75-78.
49. Yates, J. R., Zhang, W. and Miller, K. J., Fatigue Fracture and Engineer#Tg Material Structures, 1993, 16(3), 351-
362.
50. de los Rios, E. R., Mohamed, H. J. and Miller, K. J., Fatigue Engineering and Material Structures, 1985, 8, 49-64.
51. de Lange, R. G., Transactions of the Metallurgy SocieO,, AIME, 1964, 230, 644-648.
52. Lankford, J., Engineering Fracture Mechanics, 1977, 9, 617-624.
53. Taylor, D., Fatigue Thresholds. Butterworth, Oxford, 1989, p. 146.
54. Hussain, K. and de los Rios, E. R., Scripta Metallurgica et Materiala, 1994, 30, 53-58.
55. Lankford, J., Fatigue Engineering and Material Structures, 1983, 6, 15-31.
56. Forrest, P. G. and Tate, A. E. L., Journal o['the hlstitute c~fMetals, 1964, 93, 438-444.
57. Brown, C. W. and Hicks, M. A., Fatigue Engineering and Material Structures, 1983, 6, 67-75.
58. Hall, E. Q., Proceedings o[the Physics Society, London, 1951, B64, 747-753.
59. Perch, N. J., Journal of the Iron and Steel Institute, 1953, 174, 25-28.
60. Tanaka, K. and Nakai, Y., Fatigue Engineering and Material Structures, 1983, 6, 315-327.
61. Chan, K. S. and Lankford, J., Scripta Metallurgica, 1983, 17, 529-532.
62. McEvily, A. J., Eifler, D. and Macherauch, E., Engineering Fracture Mechanics, 1991, 40, 571-584.
63. Irwin, G. R., Fracture mechanics. In First Symposium on Naval Structural Mechanics, 557. Pergamon, Oxford, 1960.
64. Hobson, P. D., Fatigue Engineering and Material Structures, 1982, 5, 323-327.
65. Hussain, K., Ph.D, Thesis, University of Sheffield, U.K., 1990.
354 K. HUSSAIN

66. Miller, K. J., Mohamed, H. J. and de los Rios, E. R., In The Behaviour of Short Fatigue Cracks, 491, EGFI,
Mechanical Engineering Publication, London, UK, 1986.
67. Hussain, K., de los Rios, E. R. and Navarro, A., Engineering Fracture Mechanics, 1993, 44(3), 425436.
68. Keiro, T., Takeshi, O., Yukio, N. and Zenji, A., Fatigue Engineering and Material Structures, 1986, 9, 1-14.
69. Daeubler, M. A. and Thompson, A. W., In Small Fatigue Cracks, ed. R.O. Ritchie and J. Lankford. A1ME, 1986,
pp. 157-164.
70. Hobson, P. D., Brown, M. W. and de los Rios, E. R., In The Behaviour of Short Fatigue Cracks, ed. K. J. Miller
and E. R. de los Rios. EGFI, Mechanical Engineering Publication, London, UK, 1986, pp. 441-459.
71. Navarro, A. and de los Rios, E. R., Fatigue Engineering and Material Structures, 1988, 11, 383-396.
72. de los Rios, E. R., Navarro, A. and Hussain, K., In Short Fatigue Cracks, ESIS 13, ed. K. J. Miller and E. R. de los
Rios. The European Structural Integrity Society, Sheffield, UK. 1992, pp. 115-132.
73. Bilby, B. A., Cottrell, A. H. and Swinden, K. H., Proceedings of the Royal Society, London A, 1962, 272, 304-314.
74. Hussain, K., Tauqir, A., Hashmi, F. H. and Khan, A. Q., Metallurgy Transactions, 1994, 25A, 2421-2425.
75. Tanaka, K., Akiniwa, Y., Nakai, Y. and Wei, R. P., Engineering Fracture Mechanics, 1986, 24, 803.
76. Morris, W. L., James, M. R. and Buck, 0., Engineering Fracture Mechanics, 1983, 18, 871.
77. Tanaka, K. and Nakai, Y., In Fatigue Crack Growth ThreshoM Concept, ed. D. L. Davidson and S. Suresh. AIME,
1984, p. 497.
78. Hoshide, T., Yamada, T., Fujimura, S. and Hayashi, T., Engineering Fracture Mechanics, 1985, 21, 85.

(Received 29 October 1996, in [~naljbrm 31 July 1997, accepted 4 August 1997)