See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/309717379

Comprehensive Structural Integrity

Chapter · January 2003

CITATIONS READS

5 1,164

4 authors, including:

David L. Davidson Kwai S Chan

Southwest Research Institute (Retired 1999) Southwest Research Institute
228 PUBLICATIONS   4,104 CITATIONS    303 PUBLICATIONS   6,285 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Electron channeling View project

(Cu 1-x Co x ) 2 (OH)PO 4 Solid Solution View project

All content following this page was uploaded by David L. Davidson on 25 April 2018.

The user has requested enhancement of the downloaded file.

4.05
Small Fatigue Cracks
D. DAVIDSON, K. CHAN, R. McCLUNG, and S. HUDAK
Southwest Research Institute, San Antonio,TX, USA
4.05.1 INTRODUCTION 129

4.05.2 DEFINITIONS 130

4.05.3 METALLURGICAL CHARACTERIZATION 134

4.05.4 MECHANICS 136

4.05.4.1 Experimental Micromechanics 136
4.05.4.1.1 Stress intensity factor 136
4.05.4.1.2 Plasticity effects 136
4.05.4.1.3 Closure 136
4.05.4.1.4 Constraint 139
4.05.4.1.5 Microstructural effects 139
4.05.4.1.6 Similitude 139
4.05.4.1.7 Summary 140
4.05.4.2 Analytical Mechanics 140
4.05.4.2.1 Stress intensity factors 141
4.05.4.2.2 Plasticity effects 141
4.05.4.2.3 Closure 143
4.05.4.2.4 Constraint 145
4.05.4.2.5 Notch effects 145
4.05.4.2.6 Microstructural effects 146
4.05.4.2.7 Internal stress field 149
4.05.4.2.8 Summary 149
4.05.5 ENGINEERING LIFETIME PREDICTION 150
4.05.5.1 When is Small-crack Behavior Significant? 150
4.05.5.2 When is a Crack Small? 151
4.05.5.3 Practical Life Prediction Methods 151
4.05.5.3.1 Crack growth threshold effects 151
4.05.5.3.2 Accelerated crack growth above threshold 152
4.05.5.4 Statistical Issues 153
4.05.5.5 Small Cracks in Advanced Materials 154
4.05.6 ENVIRONMENTAL EFFECTS 154
4.05.6.1 Gaseous Environments 155
4.05.6.2 Liquid Environments 155
4.05.7 EXPERIMENTAL TECHNIQUES 158

4.05.8 CONCLUSIONS 160

4.05.9 REFERENCES 161

4.05.1 INTRODUCTION
nents and structures. The growth behavior of
Fatigue cracks are often ‘‘small’’ for a large these small cracks is sometimes significantly
fraction of the total life of engineering compo- different from what would be expected based

129
130 Small Fatigue Cracks
on conventional (i.e., large-crack) fatigue crack The fundamental reason for this disagree-
growth (FCG) rate test data and standard ment between measured small- and large-crack
FCG design and analysis techniques. In growth rate data is a lack of similitude.
particular, small fatigue cracks are sometimes Although nominally calculated DK values for
observed to grow faster than corresponding large and small cracks may be the same, the
large cracks at the same nominal value of the local driving force or the local resistance for
stress intensity factor range, DK. Small cracks crack growth may be different due to the
have also been observed to grow at significant effects of localized plasticity, crack closure and
rates when the nominal applied DK is less than other shielding, microstructural influences on
the threshold value, DKth, determined from crack-tip strain, localized crack-tip chemistry,
traditional large-crack test methods. Therefore, or other phenomena. In some cases, the basic
a structural life assessment based on large- continuum mechanics assumptions of material
crack FCG analysis methods can be noncon- homogeneity and small-scale yielding may be
servative if the initial crack size is sufficiently violated for small-crack analysis.
small that the life is dominated by small-crack Small-crack behavior is a complex subject,
growth. due to the variety of factors that may affect
The fatigue life of many structures and small cracks and the variety of microstructures
components is estimated using so-called ‘‘in- used in engineering structures. Many diffe-
itiation’’ or ‘‘total life’’ methods such as stress rent researchers have published small-crack
life or strain life. These empirical methods are data and offered various explanations and
generally based on laboratory data in which models to rationalize these data, and apparent
smooth specimens without initial (known) disagreements are not uncommon in the
cracks are cycled to failure. Although these literature. This short chapter is a general
methods describe the fatigue life associated introduction to the subject of small cracks that
with ‘‘initiation’’ of an engineering-sized crack, provides an organizational framework for
in reality small cracks may be present and published data and summarizes the most
growing during much of this life. Therefore, an current, as of early 2000s, understandings of
understanding of the mechanics of small-crack the phenomena, with special attention to
behavior may be useful in describing fatigue implications for engineering life prediction.
damage development, especially in complex The serious student should consult more
problems for which the simple empirical extensive review articles (Hudak, 1981; Suresh
models perform poorly. and Ritchie, 1984; Miller, 1993) and collections
In contrast to large-crack growth rates, of small-crack papers (Ritchie and Lankford,
which generally increase with increasing DK, 1986; Miller and de los Rios, 1986, 1992;
small-crack growth rates are sometimes ob- Larsen and Allison, 1992; Ravichandran et al.,
served to increase, decrease, or remain constant 1999) for further details and references.
with increasing DK. A variety of typical small- This chapter begins with definitions of the
crack growth rate behaviors are illustrated different types of small cracks, and material
schematically in Figure 1. characteristics related to small cracks are then
reviewed. A detailed look at small cracks from
a mechanics perspective follows, including
insights from experimental micromechanics
and a variety of micromechanical models.
Practical suggestions are given for treating
small cracks in engineering life prediction. The
special problem of environmental effects on
small cracks is briefly discussed, brief guide-
lines are given for experimental methods to
measure the growth rates of small cracks, and
general conclusions close this chapter.

4.05.2 DEFINITIONS
All small cracks are not the same. Different
mechanisms are responsible for different types
of ‘‘small crack’’ effects under different condi-
tions. Criteria that properly characterize small-
crack behavior in one situation may be entirely
Figure 1 Typical small-crack growth rate beha- inappropriate in another. It is critical, there-
viors, in comparison to typical large-crack behavior. fore, to understand the different types of small
 Definitions 131
Table 1 Small fatigue crack classification.
Description Approximate dimension Reason for small-crack behavior
Microstructurally small a/rp 4–10 Excessive plasticity
R-ratio shift
Mechanically small a/Do 2–10 Crack closure
Excessive crack-tip plasticity
Small-crack size approximation aoao Derived from Kitagawa diagram
Chemically small Ao5 mma Liquid environment
2a: surface crack length; rp: plastic zone size at crack-tip or notch plastic field size; D: a relevant microstructural dimension.
a
Critical size is a function of frequency and reaction kinetics.

cracks before selecting suitable analytical

treatments for life prediction. The classifica-
tions used in Table 1 help separate the various
types of small crack.
A crack is generally considered to be
‘‘microstructurally small’’ when all crack di-
mensions are small in comparison to charac-
teristic microstructural dimensions. The
relevant microstructural feature that defines
this scaling may change from material to
material, but the most common microstructur-
al scale is the grain size. The small crack and its
crack-tip plastic zone may be embedded
completely within a single grain, or the crack
size may be of the order of a few grain
diameters.
Typical crack growth data for microstructu-
rally small cracks are shown for a 7075
aluminum alloy in Figure 2, along with
traditional large-crack data for the same
material (Lankford, 1982). Note that small-
crack growth can occur at nominal DK values
below the large-crack threshold. Small-crack
growth rates are often faster than would be
predicted by the extrapolated large-crack Paris
equation (the dashed line in Figure 2), and the
apparent linear or Paris slope for the small- Figure 2 Typical FCG data for microstructurally
small cracks and large cracks (source Lankford,
crack data can be smaller than for the large-
1982).
crack data. Crack arrest (temporary or perma-
nent) can occur at these low DK values, and this
arrest is sometimes observed to occur when the
crack size is of the order of the microstructural
unit dimension. However, not all small cracks
arrest or even slow down at these microstruc-
tural barriers. As the crack continues to grow,
the small-crack da/dN crack growth rate data
often merge with large-crack data.
A crack is generally considered to be
‘‘mechanically small’’ when all crack dimen-
Figure 3 Schematic of relationship between me-
sions are small compared to characteristic
chanically small cracks and plastic zones.
mechanical dimensions. The relevant mechan-
ical feature is typically a zone of plastic
deformation, such as the crack-tip plastic zone Many microstructurally small cracks are also
or a region of plasticity at some mechanical mechanically small, but our focus in this
discontinuity (e.g., a notch). The crack may be section is on mechanically small cracks that
fully embedded in the plastic zone, or the are microstructurally large.
plastic zone size may simply be a large fraction Typical crack growth data for mechanically
of the crack size, as illustrated by Figure 3. small cracks in unnotched configurations are
132 Small Fatigue Cracks

Figure 4 Typical FCG data for mechanically small cracks and large cracks (source Davidson et al., 1996).

Figure 5 Typical FCG data for short cracks at

notches (source McClung and Sehitoglu, 1992).

shown in Figure 4 for an HSLA steel (David-

son et al., 1996). Note again that small-crack
growth can occur below the large-crack thresh- Figure 6 Comparison of crack size and grain
old. The slope of the linear region often diameter in fracture specimens containing: (a) small
appears to be roughly the same for small- crack and (b) a short or large crack (source Chan
and large-crack data, but the small-crack data and Lankford, 1988).
sometimes fall above the large-crack trend line
when expressed in terms of nominal DK. (early 2000s) accepted definition for a ‘‘small’’
Cracks growing in notch fields can exhibit a crack requires that all physical dimensions (in
characteristic ‘‘fish-hook’’ growth behavior particular, both the length and depth of a
(Pearson, 1975), as illustrated in Figure 5 surface crack) are small in comparison to the
(McClung and Sehitoglu, 1992). Here, crack relevant length scale. The relevant length scale,
growth rates are much faster than for compar- and hence the specific physical dimensions,
able large cracks when the cracks are extremely vary with the particular material, geometry,
small in comparison to the notch radius. These and loading of interest. In contrast, a crack is
small-crack growth rates can actually decrease defined as being ‘‘short’’ when only one
with increasing crack growth and then even- physical dimension (typically, the length of a
tually merge with large-crack data. through-crack) is small in comparison to the
One warning about nomenclature should be length scale. These definitions are illustrated in
noted. The terms ‘‘small crack’’ and ‘‘short Figure 6. Because the crack front of a short
crack’’ both appear in the literature, and crack interrogates many different grains, it is
sometimes the two appear to be used inter- not usually subject to strong microstructural
changeably. However, the two terms have effects. However, it should be noted that this
distinct meanings among many researchers. In nomenclature distinction has not always been
the US research community, the currently observed in the literature, and some authors
 Definitions 133
(especially in Europe) employ the terms with
nearly reverse meanings. Whatever the usage,
the reader should carefully observe which type
of ‘‘little’’ or ‘‘physically small’’ crack is the
subject of a given application.
Experiments on a variety of ferritic and
martensitic steels in aqueous chloride environ-
ments have shown that under corrosion-fatigue
conditions, small cracks can also grow sig-
nificantly faster than large cracks at compar-
able DK values (Gangloff, 1981). This
phenomenon is believed to result from the
influence of crack size on the occluded chem-
istry that develops at the tip of fatigue cracks.
The specific mechanism responsible for this
‘‘chemically small’’ crack effect is believed to
be the enhanced production of embrittling
hydrogen near the crack tip resulting from a
crack-size dependence of one or more factors
that control the evolution of the crack-tip
environment—specifically, convective mixing,
ionic diffusion, or surface electrochemical
reactions (Gangloff, 1984). This mechanism is
distinctly different from that responsible for
the enhanced rate of crack growth in micro- Figure 7 FCG rate vs. stress intensity factor range
structurally or mechanically small fatigue (DK) in 4130 steel for various crack sizes, crack
cracks in vacuum or gaseous environments. geometries, and environments (Gangloff, 1985).
However, the enhanced crack-tip plasticity Enhanced growth rates of small surface and edge
associated with microstructurally or mechani- cracks compared to long cracks in aerated 3% NaCl
cally small cracks could further stimulate the environment illustrate chemical crack size effects.
Small surface crack data also depend on applied
electrochemical reactions through the creation
stress levels, as indicated. No such crack size or
of additional fresh and highly reactive surfaces stress level dependencies occur in the reference air
at the crack tip. environment. (Line marked ‘‘absolute’’ is the theo-
The regimes of mechanically and micro- retical corrosion FCG rate assuming a maximum
structurally small cracks can overlap (Hudak hydrogen concentration, i.e., without inhibition by
and Chan, 1986). Many microstructurally oxygen—see Gangloff, 1985.)
small cracks are also mechanically small,
although a microstructurally small crack in a
large grain or single-crystal material may be
mechanically small. Many mechanically small
cracks of practical importance are microstruc-
turally large.
The chemical crack-size effect is clearly
illustrated for 4130 steel in an aqueous NaCl
environment in Figure 7. Note that corrosion-
FCG rates from small surface cracks (0.1–
1 mm deep), as well as short through-thickness
edge cracks (0.1–3 mm), are appreciably faster
than corrosion-FCG rates from large through- Figure 8 Kitagawa diagram for estimating ao.
thickness cracks (25–40 mm) in standard com-
pact tension specimens. It is also interesting to
note that the corrosion-FCG rates for small large cracks. Thus, in relation to the fatigue
surface cracks decrease with increasing applied small-crack effect, the chemical small-crack
stress (at a given DK), and this trend is opposite effect is of potentially greater importance since
to the dependence of applied stress on crack it can occur over a much larger range of crack
growth rates in small fatigue cracks. Moreover, sizes (up to 5 mm).
all of the corrosion-fatigue crack growth rates Different definitions of crack length can also
in NaCl are enhanced compared to those in a lead to confusion. In some treatments, surface
moist laboratory air environment, even though crack length is 2a (Figure 1), or it may also be
the latter were generated with both small and 2c (Figure 5), or a (Figure 8). Verify the
134 Small Fatigue Cracks
definition of crack length when it is used in crack a crack?’’ At one extreme, Miller (1997)
modeling calculations. has provocatively proposed that crack propa-
The relationship between small cracks, large gation begins on the very first cycle of fatigue
cracks, and crack initiation is illustrated by the loading, with the immediate formation of local
Kitagawa diagram (Figure 8) (Kitagawa and surface or subsurface features that will pro-
Takahashi, 1967). This diagram integrates the gressively grow into a crack of engineering size.
traditional large-crack fracture mechanics At the other extreme, it is the common practice
threshold, DKth, with the traditional smooth in many industries to regard crack ‘‘initiation’’
specimen endurance limit, DSe. The region in as the formation of this engineering-sized
the Kitagawa diagram both above the hor- crack. Depending on the industry, the standard
izontal endurance limit line and above the ‘‘initial crack size’’ for a fracture mechanics
sloping threshold line is ‘‘unsafe’’ in that analysis commonly ranges from 125 mm
fatigue failure is predicted to occur. The region (0.0127 cm) to 1.25 mm (0.127 cm). The specific
below both lines is ‘‘safe’’ in that fatigue failure size chosen may also be driven by the
is predicted not to occur. The intersection of capability of available nondestructive evalua-
the two lines is commonly denoted as ao. tion (NDE) methods; the ‘‘initial crack size’’ is
The Kitagawa diagram demonstrates that the smallest size that can be detected reliably in
cracks smaller than ao must be able to grow at practice.
nominal stress intensity factors that are less The classic question is perhaps more practi-
than the large-crack threshold, since smooth cally posed: ‘‘When do I begin to treat the
specimens fail by the initiation and growth of crack as a crack in life prediction analysis?’’ If
microcracks to failure at stresses greater than the life prediction method is based entirely on
the endurance limit. As discussed later in this large-crack fracture mechanics, then the cor-
chapter, El Haddad and co-workers (El Had- rect answer may be of the order of ao, because
dad et al., 1979; Tanaka et al., 1981) have cracks larger than that size generally obey
proposed that the safe/unsafe boundary is large-crack trends. However, the introduction
properly represented not by the two intersect- of simple small-crack models may permit
ing straight lines, but by the curving dashed line accurate treatment of crack growth down to
in Figure 8 that provides a smooth transition crack sizes of the order of the relevant
between the endurance limit and threshold microstructural dimension. The introduction
lines. This construction implies that cracks of more sophisticated micromechanical models
smaller or slightly larger than ao can also grow may permit accurate treatment of crack growth
both below the large-crack threshold and below down to even smaller sizes. Selection of the
the endurance limit stress. The transition curve most appropriate approach will depend on
has been explained either as an increase in the many practical factors. What is important is
crack driving force or as a decrease in the that small-crack behavior represents the bridge
threshold as crack size decreases. between fatigue crack nucleation and large
As noted earlier, cracks that are smaller than FCG, and small-crack models are necessary to
some characteristic microstructural dimension understand and predict fatigue damage devel-
(e.g., the grain size) are known to exhibit faster opment in these size scales.
growth rates and lower thresholds under
certain conditions, although crack deceleration 4.05.3 METALLURGICAL
and/or arrest may occur when the crack tip CHARACTERIZATION
reaches a microstructural discontinuity. The
dashed scallop just below the horizontal ‘‘Fatigue crack initiation’’ is a term used
endurance limit line at the far left-hand side differently by scientists studying fatigue in the
of the Kitagawa diagram in Figure 8 represents laboratory and by engineers designing and
this sort of behavior—cracks that initiate and maintaining structures. To the scientist, initia-
begin to grow at stresses below the endurance tion is the number of cycles required to
limit, but then arrest as they grow to slightly generate the smallest crack that can be detected
longer lengths unless the stress is raised above by any means. To the engineer, initiation is
the endurance limit. likely to mean the smallest crack that can be
Because the terms used to define small cracks determined by a reliable NDE technique,
overlap, and may be confusing, some ‘‘rules of which would, to the scientist, probably mean
thumb’’ are given in Table 1 that may help that growth of a small crack is included in this
clarify the descriptions of size effects as well as definition.
help in approximating the conditions under To the scientist, initiation sets the limit on
which cracks may be considered as small. the minimum size of a small fatigue crack
The topic of small cracks provides important (Miller, 1987), assuming that the component
perspectives to a classic question, ‘‘When is a does not have any existing cracks. Initiation is
 Metallurgical Characterization 135
a complex process that can lead to a crack that When characterizing small cracks with a
is extremely small (2 mm) or one that ‘‘pops in’’ microscope, a grain-sized crack usually indi-
(initiates and rapidly grows) at up to 100 mm cates that the crack either formed at that size
long. Fatigue cracks in commercial materials or grew within a few cycles to the size seen
start at metallurgical stress concentrations, (Chan, 2002). If an inclusion or pore is located
such as inclusions and pores, as well as within along the crack length, then determining the
regions where a number of adjacent grains of minimum size is more difficult, but it is usual
nearly the same orientation have the slip practice to assume the minimum size as that of
characteristics of a single large grain. Surface the pore or inclusion. At high cyclic stresses,
finish has a strong influence on fatigue crack crack initiation sites are usually the same in
initiation (Taylor and Clancy, 1991) and many samples of a given material, but for low
residual stresses due to surface preparation cyclic stresses, multiple locations may occur
can have considerable influence on the number which may be the effect of environment, or
of cycles to initiation. The anisotropy and the sampling of other defects (Ueno and
texture of a material also have strong influ- Kishimoto, 2001).
ences. Cracks can initiate also in grain bound- Fractography from small cracks should
aries, with or without the influence of reflect the complexity of the microstructure
environment, or at grain boundary brittle and the planarity of slip. The nature of the slip
intermetallics, or due to the chemical segrega- characteristics of the microstructure will dic-
tion of impurities to the grain boundary. tate the fracture surface topography near the
Suresh (1998) has a summary of the fatigue origin of the crack. In general, highly planar
crack initiation process in both pure (labora- materials should have very rough surfaces near
tory) and commercial metals. the crack origin, becoming smoother as the
Planar slip materials may concentrate slip crack grows, while multiple slip materials will
and form persistent slip bands that eventually evidence less of this transition. Intergranular
become cracks. Larger grained materials are fracture must be carefully differentiated from
more prone to slip band cracks than small planar slip fractures.
grained materials. In some alloys, the apparent Stage I is a term used to describe a small
grain size, as seen in an optical microscope, crack in the region near the origin in pure
may not be an accurate view of the ‘‘slip grain metals and in some single-phase polycrystal-
size.’’ An example of this is a ferritic–pearlitic line alloys. Cracks exhibiting stage I beha-
steel, where the underlying alpha grain has vior are usually initiated by the intrusion–
only one crystalographic orientation (Daeubler extrusion mechanism from persistent slip
and Thompson, 1986), or a grain existing in the bands, and they grow at an angle to the stress
gamma phase before a martensitic transforma- axis. As these cracks grow larger, they trans-
tion, which is larger than any dimension in the form to stage II cracks that grow more
transformed structure. perpendicular to the stress axis. Fatigue crack
Texture, intentional or not, may enhance formation in complex materials, as used in
fatigue crack initiation if groups of grains are structural applications, usually cannot be
oriented for easy slip within the applied stress characterized by these terms since the fatigue
field. Surface preparation has an enormous cracks generally initiate at defects as stage II
influence on fatigue crack initiation that is cracks.
likely to override most metallurgical considera- Specimens or components loaded under
tions. Machining operations and surface treat- comparatively low stress may, or may not,
ments like shot peening cause deformation that have small fatigue cracks that grow discon-
can extend up to millimeters below the surface. tinuously. Grain boundaries and phase bound-
Manufacturing operations may create surface aries have been observed to stop fatigue cracks
roughness, fissures, laps, and pits that start in some materials. Numerous examples of the
fatigue cracks. These defects may be due to periodic arrest of small fatigue cracks at
machining, cleaning, or handling. microstructural features have been given (see
Because of all the factors that can influence Table 2).
fatigue crack initiation, it is very difficult to However, there are many studies where
predict the minimum size and location of a crack arrest by microstructure was not ob-
naturally initiated fatigue crack. Miller (1987) served. Several examples of this behavior are
has suggested that cracks of 2–4 mm exist in all also included in Table 2. It is often difficult to
structures and should be considered as a determine in published research results whether
universal minimum crack size, but this concept or not growth was sometimes arrested because
is not applicable universally, especially in crack growth is usually given on a logarithmic
commercial materials used in many engineering scale where zero growth cannot be directly
structures. included.
136 Small Fatigue Cracks
Table 2 Examples of intermittent FCG as related to microstructural features.
Material Authors Microstructural feature
a/b Ti alloys Bolingbroke and King (1986) a/transformed b boundaries
Ti–8.6Al Wagner et al. (1986) Grain boundaries of a-Ti
Five types of steel Tokaji et al. (1988) Pearlite/ferrite and grain boundaries
Steels De los Rios et al. (1985) Ferrite grain boundaries and ferrite/pearlite
boundaries
HSLA steel Chan et al. (1997) No influence of microstructure
a/b Ti alloys Brown and Taylor (1984) Grain and phase boundaries did not
significantly alter crack growth
Ni-base superalloy Boyd-Lee and King (1994) High angle grain boundaries

4.05.4 MECHANICS tip) is greater for small cracks than for large
cracks at the same value of computed DK, as
4.05.4.1 Experimental Micromechanics shown in Figure 10 (Lankford and Davidson,
The micromechanics of small fatigue cracks 1984). Similar to CTOD, the effective strain
has been compared to large fatigue cracks at (approximately the same magnitude as the
the same value of computed DK. Direct maximum shear strain) at the crack tip is
measurements of crack opening load, crack greater for small cracks than for large cracks,
opening displacement (COD), and displace- as shown in Figure 11 (Lankford and David-
ments within the crack-tip deformation zone son, 1984). Both the larger crack-tip strain and
have been made (Lankford and Davidson, CTOD are consistent with small fatigue cracks
1984) using the stereoimaging technique. From growing faster than large cracks at the same
the measured displacements, strains in the computed DK.
crack-tip zone were computed (Williams et al., Plastic zone size and shape measurements
1980), and from this information, values of show that crack-tip plasticity for small fatigue
equivalent DK (Davidson and Hudak, 1995) cracks is different than for large cracks. For
and constraint (Davidson, 1999) were com- large fatigue cracks, crack-tip strains are more
puted also. or less equally distributed around the crack tip,
Detailed comparisons between small and while for small cracks, strains are concentrated
large fatigue cracks are summarized in the more intensely B301 from axis of loading. The
following sections. equally distributed zone of strain found for
large cracks is indicative of mixed-mode
loading, while the ‘‘butterfly’’ shape of the
4.05.4.1.1 Stress intensity factor zone of strain for small cracks indicates a stress
Small- and large-crack mode I COD are state that is mainly plane stress.
proportional to the square root of distance
behind the crack tip (see also Chapter 2.03).
4.05.4.1.3 Closure
Mode I COD for the entire length of a 26 mm
crack is shown in Figure 9(a), and for each end When load is applied, small fatigue cracks
in Figures 9(b) and (c), where it is seen that open to the crack tip as shown in Figure 12.
COD ¼ CoOd, here d is the distance from the Compared to the way large cracks ‘‘peel open’’
crack tip. This is the same function that nearly to the tip at relatively small loads, small
describes COD for an elastic crack and implies cracks open to the tip with much more
that fracture mechanics provides an applicable difficulty, indicating that the forces that
framework by which to describe small (as well oppose crack opening (probably residual
as large) fatigue cracks. However, if values of stresses in the plastic zone), which are con-
DK are computed using Co, magnitudes differ- centrated close to the tip of the crack, have
ent from those computed using stress and crack much more influence on the opening of small
length are obtained; thus linear elastic fracture cracks than on large cracks (James and
mechanics (LEFM) must be modified if it is to Sharpe, 1989). This is an indication that the
be used to provide an adequate description of size of the plastic zone relative to the crack
small-crack behavior. length influences the behavior of small cracks
more than large cracks. Also, the large size of
4.05.4.1.2 Plasticity effects the plastic zone relative to the crack length
violates the LEFM assumption of small-scale
Crack-tip opening displacement (CTOD) yielding. The closure of small and large cracks
(defined as the COD 1 mm behind the crack differs, as shown in Figure 13, which plots the
 Mechanics 137

Figure 9 COD in the loading direction for: (a) the whole crack, (b) left end, and (c) right end, showing that
COD is proportional to Od (d ¼ distance from crack tip). This behavior indicates K-controlled behavior.

Figure 10 CTOD vs. DK. Line shown is a least-squares fit through large crack data.
138 Small Fatigue Cracks

Figure 11 Effective strain range at the crack tip as a function of DK. The line shown is a least-squares fit
through the large crack data.

Figure 12 Opening of small cracks with applied load in an aluminum alloy as compared to a large crack.
Small cracks ‘‘peel open’’ more rapidly with increasing load than does the large crack.

ratio DKeff/DK vs. 1/Kmax, where DKeff ¼ DKeff/DK until they are large enough to be
Kmax
Kopen, for R ¼ 0.1. It has been shown ‘‘large cracks.’’ The constant ratio for small
for large cracks that using 1/Kmax for the cracks corresponds to Kopen
Kmin ¼ 0.56Kmax,
abscissa results in a collapse of data taken at or that Kopen increases linearly with Kmax
0.1oRo0.8 into a single line (Hudak and (Davidson and Hudak, 1995). The variation
Davidson, 1988). For large cracks, DKeff ¼ 0 at in values measured at one level of DK occurs
DKth and rises to equal DK as K increases. because cracks do not grow on each cycle;
Small cracks maintain a constant value of thus, the measurements of Kopen will vary
 Mechanics 139

Figure 13 Differences in crack closure between small and large cracks. U ¼ (Kmax–Kop)/DK is approximately
constant until crack size increases sufficiently to be considered a large crack. The large crack data cluster
around a line that extrapolates to DKth when DKeff ¼ 0.

depending on when the data were taken during condition for internal cracks. Thus, their
the growth process. growth rates may not be affected as much by
changing constraint as for surface small cracks.
4.05.4.1.4 Constraint
4.05.4.1.5 Microstructural effects
Constraint is different for small surface
cracks than it is for large cracks that grow at Most cracks, small and large, exhibit both
the same computed DK (Davidson, 1999), mode I (perpendicular to the crack plane) and
provided that the DK is near or below the mode II (sliding parallel to the crack plane)
large-crack DKth. A small surface crack usually CODs. Although some of the mode II may be
exists in an approximately plane-stress condi- due to the cracks not being exactly perpendi-
tion, because it is a small purtubation in an cular to the loading axis, much of it is due to
otherwise practically one-dimensional (1D) anisotropic deformation within the plastic zone
stress field, while a large through-crack, loaded (Davidson, 2000).
at a near threshold DK, is approximately in a Small fatigue cracks typically do not grow
state of plane strain because the crack essen- continuously, especially for lower cyclic stress
tially acts as a very sharp notch that has a 3D levels (see Table 2). Large fatigue cracks loaded
stress field. This difference in constraint leads at a DK near threshold also do not grow
to a larger crack-tip strain and COD for small continuously (Vaidya, 1992). This intermittent
cracks relative to a large crack growing at the growth characteristic of fatigue cracks in many
same DK. The lower level of constraint materials results from the cycle-by-cycle accu-
exhibited by small fatigue cracks may also be mulation of crystal damage in the crack-tip
a major contributing factor to the lower levels process zone that is a precursor to further
of crack opening load, relative a large crack crack elongation.
near-DKth, although the fact that the length of
the crack wake is smaller also contributes to 4.05.4.1.6 Similitude
this effect. Lower crack opening load (closure)
means that the effective DK (the crack ‘‘driving The sum of all these differences in crack-tip
force’’) is larger for small fatigue cracks than mechanics indicates that the similitude present
for large cracks, which results in a higher crack for large FCG rate as DK changes in the linear
growth rate. (Paris) region is lost when crack size is reduced
Although constraint is lower for small sur- into the small-crack region. The larger crack-
face fatigue cracks, this may not be the same tip strains for small cracks are consistent with a
140 Small Fatigue Cracks

Figure 14 DK computed from experimentally measured CTOD values compared to the DK computed from
crack length and applied stress. For small cracks, the driving force for crack growth (DKexp.) is larger than
expected from the computed value, and for the large crack it is smaller than expected due to crack closure.

larger CTOD. The larger CTOD and lower independent line integral have shown larger DJ
levels of crack opening load for small cracks values for small cracks than for large cracks
are consistent with a lower level of constraint, under equivalent applied DK, further confirm-
and all these factors are consistent with the ing that the crack driving force is higher for
differences in plastic zone sizes and shapes. small cracks (Chan et al., 1986).

4.05.4.1.7 Summary 4.05.4.2 Analytical Mechanics

The differences between small and large Extensive efforts have been made to model
fatigue cracks can be summarized by com- the crack driving force and the growth
paring the crack driving forces as shown in characteristics of short and small cracks (e.g.,
Figure 14. Experimental values of DK for small see Ritchie and Lankford, 1986; Miller and de
cracks were computed from DJ, using mea- los Rios; Suresh and Ritchie, 1984; Ravichan-
sured CTOD and crack-tip strains, while dran et al., 1999; McDowell, 1996; Miller,
applied DK values were computed from crack 1993). These investigations may be broadly
length and applied stress using categorized into four groups according to the
pffiffiffiffiffiffi phenomena that were modeled, which include:
DK ¼ F Ds pa ð1Þ modified stress intensity factor; large-scale
plasticity effects (Newman, 1994; Dowling,
where F is the boundary correction factor. For 1977; El Haddad et al., 1980; Chan, 1986;
the large crack, computed values are reduced Hudak and Chan, 1986; Chan et al., 1986);
by measured closure, which for R ¼ 0.1 is crack closure effects (Hudak, 1981; Newman,
approximately the measured DKth. Small-crack 1994; Ritchie and Suresh, 1982; Suresh and
DK experimental values exceed DK applied Ritchie, 1982; Suresh, 1985; Tanaka et al.,
because of differences in crack-tip plasticity, 1981; Tanaka, 1985); plastic constraint; con-
etc. The trend line through the small-crack tinuum limit (Kitagawa and Takahashi, 1967;
data meets that of the large crack at about the El Haddad et al., 1979; Topper and El
point where the large-crack growth rates are Haddad, 1981), and microstructural size effects
becoming proportional to DK (beginning of the (Tanaka et al., 1981; Tanaka, 1985; Chan,
Paris region). Computations based on a path- 1986; Chan and Lankford, 1988); and internal
 Mechanics 141
stress field (Barenblatt, 1987; Deshpande et al.,
2003). Note that phenomenon being modeled is
listed in the same sequence used previously in
experimental micromechanics, although not all
experimentally derived measurements have
been modeled and vice versa.
Both crack closure and large-scale plasticity
contribute to the small-crack phenomenon by
causing a breakdown in mechanical similitude
at the crack tip. In contrast, microstructural
size effects arise when the crack size is smaller
than the microstructural size (e.g., grain size)
so that properties of an individual microstruc-
tural unit or grain affect the response of the
Figure 15 Crack propagation rate vs. crack length
small crack. Microstructural size effects, which for microstructurally small to mechanically small
involve mechanisms such as slip band blocking cracks (source Miller, 1993).
by grain boundaries, slip anisotropy, texture,
grain shape, crack shape (Ravichandran and
pore, particle, or an optimally orientated
Larsen, 1992), and others, contribute to small-
surface grain. Under these circumstances, the
crack behavior by virtue of a breakdown in the
driving force for FCG is not accurately treated
microstructural similitude at the crack tip.
by LEFM and the large-scale yielding experi-
Heterogeneous microstructural deformation
enced at the crack tip must be used to correct
has also been purported to cause an addi-
the crack driving force. Elastic–plastic fracture
tional driving force for small cracks (Baren-
mechanics (EPFM) has been used to address
blatt, 1987; Deshpande et al., 2003), including
this breakdown of mechanical similitude.
internal microstresses resulting from hetero-
When the crack size is of the order of the
geneous deformation in individual grains
microstructural unit or grain size, microstruc-
(Barenblatt, 1987) or image stresses acting
tural fracture mechanics (MFM) is required to
on near-tip dislocations (Deshpande et al.,
address this breakdown of similitude. Figure
2003).
15 illustrates the crack-size regimes where
MFM, EPFM, and LEFM are applicable for
4.05.4.2.1 Stress intensity factors fatigue cracks (McDowell, 1996; Miller, 1993).
Summaries of these modeling efforts and
The growth rates of small fatigue cracks important findings are highlighted in the
have been correlated to crack driving force following sections.
parameters that are defined as the product of
the stress range and the crack length raised
to a power. Instead of a power of 1/2 as 4.05.4.2.2 Plasticity effects
dictated by LEFM, the power to the crack
length can be obtained by fitting to experi- Several relations have been developed to
mental crack growth data of small cracks. compute the local crack driving force for short
Such an empirical approach was first proposed and small cracks subjected to large-scale
in the 1950s for treating long cracks (Frost yielding (Newman, 1994; Dowling, 1977; El
et al., 1974), but it also has subsequently been Haddad et al., 1980; Chan, 1986; Hudak and
used for treating the growth of small cracks. Chan, 1986). These relations are formulated
Since it has been shown that fracture me- either expressly or implicitly on the basis of DJ,
chanics is applicable to small fatigue cracks the cyclic J-integral. The DJ-integral was first
(e.g., small- and large-crack da/dN data merge applied to elastic–plastic fatigue under gross
at large-crack lengths), the use of a power yielding conditions (Dowling, 1977; El Haddad
other than 1/2 for small and large cracks is not et al., 1980). Subsequent work has demon-
justified. Thus, the empirical approaches strated that the DJ approach is applicable to
(Wang, 1996; Nisitani et al., 1994; Hamm fatigue loading where gross yielding is absent
and Johnson, 1999) will not be included in this but the stresses are high in relation to the yield
review. strength of the material (Newman, 1994; Chan,
The LEFM assumption of small-scale yield- 1986; Hudak and Chan, 1986; Chan et al.,
ing is violated in small cracks under several 1986). A nonlinear crack-tip parameter de-
fatigue conditions, which include fatigue load- scribing both short and small cracks is the
ing under high stresses in relation to the yield equivalent stress intensity factor range, DKeq,
strength, and for crack propagation within a which is defined in terms of the DJ-integral as
local plastic strain field, such as that of a notch, (Newman, 1994; Chan, 1986; Hudak and
142 Small Fatigue Cracks
Chan, 1986; Chan et al., 1986)
 1=2
Esy d
DKeq ¼ ð2Þ
1
Z2

where E is Young’s modulus, sy is the yield

stress, and d is the CTOD; Z ¼ 0 for plane
stress and Z ¼ n for plane strain. In a series of
studies, the CTODs of small fatigue cracks
were measured. The DKeq parameter computed
based on the experimental CTOD via Equation
(2) was successful in collapsing small- and
large-crack data, as shown in Figure 16(a) for
astrology. In addition, the validity of Equation
(2) for short and small cracks has been
demonstrated for other structural alloys (New-
man, 1994; Chan, 1986; Hudak and Chan,
1986; Chan et al., 1986).
Instead of using experimental values of
CTOD, Equation (2) can be used in conjunc-
tion with theoretical values that are computed
based on the analytical or numerical methods.
A common practice in EPFM is to add a
portion of the plastic zone to the crack length
to compute the stress intensity factor. This
approach was adopted to compute the crack
driving force for short fatigue cracks (New-
man, 1994) as shown in Figure 16(b) for 7075-
T6 aluminum. In this approach, an effective
stress intensity factor range has the form given
by (Newman, 1994)
  pffiffiffiffiffiffiffi
DKp eff ¼ ðSmax
So Þ pa0 F ða0 =wÞ ð3Þ

where Smax is the maximum stress, So is the

crack opening stress, F is the boundary
correction factor, w is the specimen width,
and the effective crack length, a0 , is given by
a0 ¼ a þ fnz rp ð4Þ

where rp is the cyclic plastic zone size and fnz ¼

0:25 for several cracked bodies (Newman,
1994).
For smooth specimens containing small
surface cracks subjected to large-scale cyclic Figure 16 Comparison of experimental and pre-
plasticity, Dowling suggested that the DJ- dicted small-crack growth rates based on Equation
parameter may be obtained from the J-integral (3): (a) astroloy (source Davidson and Hudak, 1995)
and (b) 7075-T6 aluminum alloy SENT specimens
solution of Shih and Hutchinson for mono- (source Newman, 1994).
tonic loading as follows (Dowling, 1977; El
Haddad et al., 1980):
 
others (Tanaka, 1983; Lamba, 1975) success-
DJ ¼ 2pb0 DWe þ f ðnÞDWp a ð5Þ fully correlated FCG rates of small surface
cracks and large cracks in steels (Figure 18) for
where b0 is the surface flaw shape correction cracks larger than 90 mm or five times the grain
factor; f(n) is a function of the cyclic hardening size (18 mm). The lack of correlation for small
exponent, n, as derived in the work of Shih and cracks below this size was attributed to
Hutchinson (1976); DWe and DWp are the inhomogeneous plasticity in the microstructur-
elastic and plastic work density, respectively, as al units. A practical engineering approach to
shown in Figure 17. Based on this DJ- estimate DJ has been developed by McClung
parameter, Dowling (1977) and subsequently et al. (1997).
 Mechanics 143

Figure 17 Schematic showing: (a) a line contour G encircling the tip of a crack and along the upper and lower
crack surfaces, and (b) comparison of the assumed Ds–De behavior in the DJ-integral with the actual behavior
at the crack tip.

oxide wedge and roughness-induced crack

closure are microstructure sensitive; as a result,
they have been modeled at the microstructural
scale levels via the micromechanical approach.
Plasticity-induced crack closure in short
cracks was modeled by Newman (1994), who
used the Dugdale model (Dugdale, 1960) to
represent the crack-tip plastic zone and com-
pute the residual plastic strain on the surfaces
of the crack wake. Contact of the fracture
surfaces during loading leads to delayed crack
opening, thereby reducing the stress intensity
range between opening and maximum loads.
The amount of plasticity-induced crack closure
depends on the stress state at the crack tip, the
crack-tip strain, the amount of plastic strain in
the crack wake, and the residual stress in the
plastic zone. For through-thickness cracks in
several Al alloys, Newman and co-workers
found that the plasticity-induced crack closure
Figure 18 Comparison of FCG rates vs. DJ for was lower in short cracks, but it increased with
small cracks with a large-crack scatter band (source increasing crack length. The higher crack
Dowling, 1977). growth rates exhibited by short cracks can be
correlated with an effective stress intensity
4.05.4.2.3 Closure range, DKeff, as given by
Several crack closure mechanisms have been DKeff ¼ Kmax
Kopen ð6Þ
identified in metals and structural alloys, which
include contact of the crack surfaces resulting The stress intensity factor at crack closure is
from plasticity (Hudak, 1981; Newman, 1994), calculated on the basis of an analytical
oxide wedging (Ritchie and Suresh, 1982), and formulation based on the Dugdale model
fracture-surface asperities (Suresh and Ritchie, (Dugdale, 1960).
1982). The plastic-wake phenomenon has been Elastic–plastic finite element methods
modeled using the continuum mechanic ap- (FEMs) have also been used extensively to
proach. Crack closure by fracture surface simulate the crack closure behavior of fatigue
asperities is often referred as roughness-in- cracks in structural alloys. Prior reviews
duced closure (Suresh and Ritchie, 1982). Both (McClung, 1996; Newman, 2002) of the FEM
144 Small Fatigue Cracks
analyses indicated that the mesh size of the compressive axial stresses that serve to clamp
elements must be sufficiently fine in length scale the crack closed even at some nonzero remote
and large in element scale in order to capture loads. The compressive syy stresses extend for
the entire cyclic crack-tip plastic zone and the an appreciative distance behind the crack tip,
residual stretch in the crack wake. Most, if not gradually increasing in intensity near the crack
all, FEM simulations performed prior to 1985 tip. As the crack is loaded remotely, these
did not meet these two requirements, and these residual stresses are gradually overcome closer
results are suspect (McClung, 1996). With an and closer to the crack tip, causing the crack to
appropriate mesh size reduction, FEM results gradually peel open. This sequence of the crack
provide direct insight into the mechanics opening process has been confirmed by direct
phenomena that cause plasticity-induced clo- experimental measurements (McClung and
sure. Figure 19 shows the cumulative axial Davidson, 1991).
plastic strains, epyy ; at zero load for a plane- Figure 19 illustrates the closure process in
stress short fatigue crack that has grown under plane stress. Some investigators have suggested
R ¼ 0 loading from left to right into the current that closure does not occur in plane strain,
field of view (McClung, 1996). The plastic noting that plastic contraction in the thickness
strains in front of the crack tip were induced by direction is not available to feed axial plastic
the most recent load cycle, as the crack-tip stretch. Finite element studies (Sehitoglu and
singularity elevated the stresses ahead of the Sun, 1991; McClung et al., 1991; McClung and
crack well into the plastic regime. Since these Davidson, 1999) have shown that closure can,
plastic strains are not fully reversed upon in fact, occur in plane strain, and that the
unloading to remote zero load, they remain material required to feed axial plastic stretch
as permanently stretched material while the comes from in-plane transverse contraction
crack tip moves ahead, leaving behind the (i.e., negative epxx strains). A plot of cumulative
plastic wake clearly shown in the contour plot. plastic strains for a plane-strain fatigue crack
The material required to feed this plastic can be obtained from Figure 19, by switching
stretch comes from plastic contraction in the epxx and epzz ; and the magnitude of all plastic
out-of-plane (zz) direction; the plastic strains, strains is scaled down by nearly a factor of 5.
epzz ; are nearly mirror images of the epyy strains. The in-plane residual stress field for the plane-
The material remote from the crack tip, which strain fatigue crack is considerably more
remains elastic, constrains the plastically de- complex due to the in-plane plastic contrac-
formed material, both in the wake and ahead tion. Color contour plots providing further
of the crack tip, upon unloading, and induces information about the residual stress and

Figure 19 FEM-calculated cumulative plastic strains at zero load near a plane-stress fatigue crack (source
McClung, 1996).
 Mechanics 145
plastic strain fields in both plane stress and understanding, as of early 2000s, of the effects
plain strain are available elsewhere (McClung of plastic constraint on plasticity-induced crack
et al., 1991). closure and FCG in a review (Newman, 2002).
In addition to plasticity-induced crack clo- 2D FEM analyses (Fleck and Newman, 1988;
sure, other closure mechanisms are important McClung et al., 1991; Newman, 2002) indicated
in influencing the near-threshold behavior of that plasticity-induced crack closure occurs
large cracks. These include oxide-induced under the plane-strain condition also, but the
crack closure (Ritchie and Suresh, 1982), amount of crack closure is smaller and the
roughness-induced crack closure (Suresh and opening load is lower compared to the plane-
Ritchie, 1982), and crack closure resulting stress condition. Furthermore, 3D analyses
from crack branching and deflection (Suresh, (Chermahini et al., 1988; Chermahini and
1985). These closure processes are often Blom, 1991; Riddell et al., 1999) showed that
referred to as crack-tip shielding mechanisms the crack opening stresses are higher in the
that reduce the local driving force for crack exterior (free surface) region and lower in the
growth by decreasing the local DK for the crack interior region because of higher plastic con-
tip during the fatigue cycle (Ritchie and Yu, straint in the interior region.
1986). Suresh and Ritchie have modeled Newman et al. (1986) have analyzed the
roughness-induced crack closure by consider- effects of plastic constraint on the growth of
ing the fracture surface asperity height in small cracks using a tensile strip model that
relation to the mode I and mode II CODs. incorporates a plastic constraint factor for
Their model gives the stress intensity at describing constrained deformation at the tip
closure, Kopen, as (Suresh and Ritchie, 1982) of a fatigue crack. This factor, which is used as
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi a measure of the increase in the flow stress
2gx during constrained plastic deformation, has a
Kopen ¼ Kmax ð7Þ value ranging from 1 for plane stress and 3 for
1 þ 2gx
the plane strain. By adjusting the value of the
constraint factor, the growth of a half-penny
where g is a nondimensional fracture surface crack from a notch to final failure was
roughness factor, equal to the fracture surface simulated successfully. The simulation, how-
asperity height divided by the length of the base ever, required a set of prescribed values of the
of the asperity, and x ¼ uII/uI, the mode II to plastic constraint factor at various crack
mode I displacements at the crack tip. Accord- lengths. An independent prediction of the
ing to this model (Suresh and Ritchie, 1982), plastic constraint and the growth rates of a
roughness-induced crack closure requires a small crack as it extends does not appear to
residual mode II displacement at the tip of the exist in the literature.
unloaded crack. Experimental evidence indi-
cates the presence of a mode II displacement at
the tip of loaded large, short, and small cracks 4.05.4.2.5 Notch effects
subjected to nominal mode I loading. The A small crack growing from the root of a
crack-tip mode II displacement arises from two hole, notch, or other stress concentration can
possible sources, which are: (1) crack deflection exhibit accelerated growth rates (Pearson,
and (2) noncontinuum slip behavior in the 1975; McClung and Sehitoglu, 1992). Some-
crack-tip plastic zone. Suresh (1985) has times the small crack grows very rapidly at
modeled crack closure due to a combination first, then slows down with increasing crack
of crack deflection and roughness-induced length, and then gradually accelerates again,
crack closure. These closure mechanisms, which finally merging with conventional large-crack
are operative in the crack wake, are considered trends. This ‘‘fish-hook’’ behavior was illu-
not to be operative in small and short cracks strated earlier in Figure 5.
when the crack length is short. The stress The small-crack effect at notches appears to
intensity factor due to closure, however, in- be caused by several different factors, including
creases with increasing crack length when one local plasticity and crack closure, and hence this
or more mechanisms become operative. effect may be a special case of the effects
discussed in the two preceding sections. The
4.05.4.2.4 Constraint effect is most pronounced when the fatigue
loading causes plastic deformation at the notch
Plasticity-induced crack closure is more pre- root (due to more severe applied loading and/or
valent in fatigue cracks under the plane-stress a more severe stress concentration factor) and
condition than those under plane-strain condi- the small crack is growing within this plastic
tion due to differences in the crack-tip plastic field. However, accelerated growth rates can be
constraint. Newman summarized the current observed beyond the notch plastic zone. The
146 Small Fatigue Cracks

Figure 21 A schematic showing the dependence of

yield stress within the plastic zone on the a/D ratio
for small, short, and large cracks (source Chan and
Lankford, 1988).

large cracks because of the different number of

units encountered at their crack tips (Chan and
Lankford, 1988). As illustrated in Figure 6, the
number of grains encountered by a small crack
of initial length, 2ai, embedded within a single
grain is unity. However, as the half-length, a,
of the small crack increases, the number of
grains along the crack front increases propor-
tionately, and ultimately approaches that of a
through-thickness large crack. Because of this
sampling effect, the mechanical properties
(e.g., yield stress, number of slip planes, etc.)
Figure 20 Stress distributions at maximum load for of the crack-tip plastic zone of a small crack
an uncracked body (top) and changes in crack vary with crack length and approach that of a
opening stresses for a crack growing from the notch
(bottom). E ¼ Young’s modulus, H ¼ plastic mod-
large crack (average properties) as the crack
ulus, a relates to FEM mesh size (nondimensional). length increases.
The variation of yield stress of the crack-tip
plastic zone with crack size for small cracks is
acceleration effect is also most pronounced when illustrated in Figure 21. Fatigue cracks usually
the fatigue loading has a negative stress ratio. initiate in an optimally oriented surface grain
Numerous researchers (cited in McClung with a low Taylor factor, e.g., M ¼ 2, where M
and Sehitoglu, 1992) have experimentally is the Taylor factor. As the crack length
measured crack opening levels for cracks increases, the value M, and therefore the yield
growing from holes or notches and found that stress, increases due to increasing numbers of
changes in growth rates corresponded to grains encountered by the crack tip. The ratio
gradual changes in crack closure with increas- of the plastic zone size to crack size is of the
ing crack length. Finite element studies of order of unity when the small crack is
closure (McClung and Sehitoglu, 1992) have embedded within a grain; thus, the small crack
confirmed this phenomenon. Figure 20 com- is subjected to a large-scale plastic field. As the
pares a series of finite element simulations of crack length increases, the plastic zone size to
crack closure with the corresponding stress crack size decreases due to increasing yield
fields in the uncracked body. The closure stress and it approaches that of a large crack.
arguments also explain the dependence of the Several microstructural length scale para-
small-crack effect on both the ratio and the meters have been suggested for delineating the
magnitude of the applied fatigue stresses. transition from small-crack behavior to the
polycrystalline macroscopic behavior experi-
4.05.4.2.6 Microstructural effects enced by a large crack. The ao parameter
proposed by El Haddad et al. (1979) is defined
Small surface cracks are more susceptible to on the basis of the Kitagawa diagram (Kita-
microstructural effects than through-thickness gawa and Takahashi, 1967) and it is given by
 Mechanics 147
(El Haddad et al., 1979) and the sloping portion of the line to the left in
 2 Figure 21 corresponds to the horizontal solid
1 DKth line to the left in Figure 8.) This threshold
ao ¼ ð8Þ
p F Dse dependence on crack size can be presented in
terms of ao as follows:
where DKth is the large-crack FCG threshold, F rffiffiffiffiffiffiffiffiffiffiffiffiffi
is the boundary correction factor, and Dse is sc lc a
DKth ðaÞ ¼ DKth ð10Þ
the fatigue limit or endurance limit. This a þ ao
parameter represents the minimum crack
length or the continuum limit beyond which which correlates well with both the threshold
both mechanical and microstructural simili- data of various crack sizes for several alloys.
tudes of large cracks are satisfied and LEFM is Equation (9) has been derived by several
valid (see again Figure 8). means.
El Haddad et al. (1979, 1981) suggested that The meaning of ao in terms of microstruc-
ao may be added to the actual crack length, a, ture has not been established. El Haddad et al.
to obtain an effective crack length, which is (1979) considered ao as an intrinsic crack
then used to compute the stress intensity factor length inherent in a material. Using a different
range, leading to approach, Tanaka and co-workers (1981, 1985)
treated small cracks in terms of a blocked slip
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
DK ¼ F Ds pðao þ aÞ ð9Þ model in which ao is taken to be the length of
the slip band blocked by a grain boundary. The
for the small crack. El Haddad et al. (1979, blocked slip band model gives the small-crack
1981) successfully correlated crack growth rate threshold as (Tanaka et al., 1981; Tanaka,
data for short and large cracks using this 1985)
approach. rffiffiffiffiffiffiffiffiffiffiffiffiffi rffiffiffi  
a a a
Alternatively, the El Haddad formulation sc
DKth lc
¼ DKth þ 2sfr arccos
and the ao parameter have proven useful for a þ ao p a þ ao
modeling the dependence of small-crack ð11Þ
threshold on crack size. In other words, the
small-crack effect is treated as a change to the where sfr is the friction stress for dislocation
crack growth resistance, rather than the crack motion along the slip band. Equation (11) is
driving force. As shown in Figure 22 and also reduced to Equation (10) when sfr ¼ 0: Instead
by Tanaka et al. (1981), extensive experimental of a blocked slip band, a FCG threshold can
evidence indicates that the fatigue crack thresh- also arise from dislocation shielding and the
old, DKth sc
; for small cracks increases with presence of a dislocation-free zone at the crack
increasing crack length and approaches the tip (Wilkinson et al., 1998).
large-crack threshold, DKth lc
: (Note that Figure A limitation in predicting ao as defined
22 is effectively just the Kitagawa diagram according to Equation (8) is that it does not
tilted on its side: the nearly horizontal portion relate explicitly to any microstructural size
of the line to the right in Figure 22 corresponds parameter. This limitation has been resolved
to the sloping solid line on the right in Figure 8, by considering the slip process within a single
grain. Utilizing Tanaka’s concept of a slip
band blocked by a grain boundary, an equa-
tion relating ao to the grain size has been
derived (Chan, 1999):
 
D DsH ðf
1Þ 2
ao ¼ ð12Þ
2 aMDse

with
Ds
f¼ ð13Þ
Dsmy

and
DKII;th
a¼ ð14Þ
Figure 22 FCG thresholds of small cracks in mill- DKI;th
annealed Ti–6Al–4 V (source Brown and Taylor,
1984) and in a lamellar TiAl alloy (source Chan, where D is grain size, H is the Heaviside
1999). function (H(x) ¼ 1 when x41; H(x) ¼ 0 when
148 Small Fatigue Cracks
xr0), M is the Taylor factor, and Dsmy is the of the tip of small cracks. Experimental
microscopic yield stress of the surface grain. evidence for large plastic strains associated
Combining Equations (10) and (12) leads to with small fatigue cracks has already been
rffiffiffiffiffiffiffiffiffiffiffiffiffiffi presented. For these small cracks, DK requires
sc lc a corrections for large-scale plasticity due to a
DKth ¼ DKth ð15Þ
a þ bd high applied stress, a low microscopic yield
stress, and the lack of constraint and micro-
with structural similitude at the crack tip. Under
 this circumstance, the crack driving force can
1 DsH ðf
1Þ 2 be expressed as (Chan and Lankford, 1988)
b¼ ð16Þ
2 aMDse
DKeq ¼ qp qm DK ð17Þ
b goes to zero when the applied stress range is
less than microscopic yield stress. Figure 23 where qp is the large-scale plasticity correction
illustrates the dependence of the threshold factor and qm is the microstructural similitude
stress on crack size computed for a lamellar correction factor. Expressions for these two
TiAl alloy at four different colony sizes using correction factors have been derived (Chan and
this model. Lankford, 1988):
Besides the small-crack threshold, blocked pffiffiffi    
1=2
8 sy psmax
slip bands have been used to explain crack qp ¼ ln sec ð18Þ
deceleration and arrest of small cracks by p smax 2sy
microstructural boundaries. Several treatments
and
of grain boundary interaction with a small
crack are available in the literature (Chan and pffiffiffi rffiffiffi
4 2sy b smy
Lankford, 1983; de Los Rios et al., 1985; qm ¼ 1 þ 1
ð19Þ
pDs a sy
Navarro and de Los Rios, 1988; Wang and
Hutchinson, 2001). In these formulations, the where b is the cyclic plastic zone size, and smax
driving force for FCG is reduced when a small is the maximum stress.
fatigue crack is blocked by a grain boundary. A Microstructure tends to influence the growth
wide range of crack growth behaviors, includ- of small cracks similarly. As a result, an effort
ing crack acceleration, deceleration, and arrest, has been made to derive the growth kinetics of
can be simulated depending on the slip activity small cracks from a microstructure-based
within the first grain and the ease of slip model that explicitly relates crack growth to
transmission into the second and subsequent stress range (Ds), cycles-to-failure, crack size,
grains. In general, the threshold-like behavior, microstructural unit size (D), and other mate-
sometimes termed a microstructural FCG rial parameters. This approach leads to (Chan,
threshold, occurs at the boundary neighboring 2002)
a grain or phase that is not oriented optimally
( 
for easy slip. In contrast, easy slip in favorably da lpð1
nÞ Ds
2Mk 2
oriented grains allows the development of a ¼ 2a
dN 8M m
large plastic zone and high plastic strains ahead
 3
1=2a
D
 2 a2a
1 ð20Þ
h

where Ds is the stress range, a is the fatigue life

exponent, k is the slip plane friction stress, n is
Poisson’s ratio, m is shear modulus, h is the slip
band width, and l ( ¼ 0.005) is a universal
constant (Chan, 2002). The expression de-
scribes the growth of a small crack by repeated
renucleation at the crack tip. Since crack
growth does not occur at stress range below
the fatigue limit, this small-crack growth
mechanism operates only at stress ranges
above the fatigue limit. In addition, the crack
growth rate of small cracks is driven mostly by
the stress range (Ds) and may or may not
Figure 23 Calculated threshold stress as a function depend on the crack length or stress intensity
of half crack length for four colony sizes, D, in a range (DK), depending on the value of the
Kitagawa diagram for R ¼ 0.1 (source Chan, 1999). fatigue life exponent, a. According to this
 Mechanics 149
model, the crack growth rate of small cracks is tions indicate that under cyclic loading, FCG is
independent of crack length when a ¼ 1/2. In DK controlled for sufficiently long cracks.
contrast, da/dN increases with crack length Below a critical crack size, the fatigue thresh-
when a41/2, but da/dN decreases with crack old decreases with crack size and cyclic crack
length when ao1/2. The value of the exponent growth continues even at stress intensity ranges
a, where 0oar1, depends on the degree of slip below the large-crack threshold. Despande
irreversibility and the stacking fault energy. et al. (2003) attributed the reduction in the
This model is a relatively new formulation fatigue threshold with crack size to a progres-
whose usefulness and limitation are still being sive increase in internal stress under cyclic
explored. loading. It was argued that the internal stress
arose from the irreversibility of dislocation
4.05.4.2.7 Internal stress field motion that allowed the density of anti-
shielding dislocations to accumulate near the
The notion of an internal microstress field crack tip as cyclic loading proceeded. This
was first suggested by Barenblatt (1987). The internal stress decreases with increasing crack
basic idea is that inhomogeniety in the micro- length as the crack tip moves away from the
structure creates a field of self-balanced mi- free surface. The result is that the threshold for
crostresses whose length scales are of the order the short crack is predicted to increase with
of the microstructural length scale. The ampli- increasing crack length, as shown in Figure 24,
tude of the internal microstress field is gov- which shows a plot of threshold stress normal-
erned by the grain size distribution, ized by the yield stress as a function of the
manufacturing methods, and processing con- crack length, a. While Figure 24 shows a
ditions. Barenblatt contended that the accel- similar trend experimental data (Figure 22), it
eration, deceleration, and crack arrest is uncertain that the predicted behavior arises
behaviors observed in small cracks when they from internal stresses and not from plasticity
interact with grain boundaries could be ratio- effects since both were present in the model
nalized by combining the internal microstress simulation. Furthermore, small cracks often
field with the external stresses in the stress maintain half-penny shapes as they grow
intensity range factor calculation. This hypoth- rapidly both in depth and in length. For these
esis has not been rigorously tested. A major small cracks, internal stresses resulting from a
difficulty is that small-crack growth studies are free surface do not appear to play an important
generally performed on fully annealed materi- role in their growth response.
als that are unlikely to contain any internal
microstresses that are of sufficient magnitude 4.05.4.2.8 Summary
to enhance the crack-tip strains and the crack
growth rates significantly. In addition, the A wide variety of micromechanical and
internal microstress field is comprised of self- analytical methods have been developed and
balancing static residual stresses whose main are available in the literature for treating the
effect is to change the mean stress and the R- growth of small cracks in various structural
ratio. The net effect on FCG should be zero alloys subjected to fatigue loading. These
when local da/dN variations are averaged models cover different crack size regimes and
along the crack path. Furthermore, the varia- differ in the underlying fatigue mechanisms
tions of the da/dN response of small crack controlling the crack growth process. Invari-
seldom show a pattern of periodicity or an ably, these small-crack models all share the
oscillatory pattern that could be linked to a common feature that there exists a character-
self-balancing residual stress field. istic length scale below which the behavior
Deshpande et al. (2003) presented an analy- expected from large crack breaks down and a
sis of the mode I crack growth behavior of new treatment is needed in order to describe
geometrically similar edge-cracked single-crys- the growth behavior of cracks smaller than the
tal specimens of varying length subjected to characteristic crack size. Experimental evi-
monotonic and cyclic loading. Plastic deforma- dence indicates that small-crack behavior arises
tion was modeled through the motion of edge from several origins, sometimes exhibiting
dislocations generated at and near the crack tip opposite and contrasting effects on the fatigue
in an elastic solid with the lattice resistance to threshold and the growth kinetics. A single
dislocation nucleation, motion, interaction model that encompasses all of the small-crack
with obstacles, and annihilation incorporated fatigue mechanisms does not exist. As a result,
through a set of constitutive equations. The a practitioner predicting fatigue life based on
fracture properties were specified through an the growth of small cracks might wish to select
irreversible cohesive relation applied to an from the various models the one that is most
embedded fracture process zone. These simula- pertinent to their problems of interest. There is
150 Small Fatigue Cracks

Figure 24 Dependence of short-crack FCG threshold on crack length predicted by the dislocation-induced
internal stress model (source Deshpande et al., 2003).

sufficient overlap among the various models on bulk damage strain-life or stress-life ana-
that, in many cases, deciding which short or lyses. However, as noted earlier, the total life in
small-crack model to select might not be so many strain-life applications is often domi-
important if the crack growth characteristics nated by the growth of small cracks, especially
that are different from the large crack can be in the low-cycle fatigue (LCF) regime where
specifically addressed. crack formation occurs very early in life and
final crack sizes are still relatively small.
Therefore, the damage accumulation process
4.05.5 ENGINEERING LIFETIME during LCF, which is often treated as an
PREDICTION ‘‘initiation’’ problem, is often actually a small-
4.05.5.1 When is Small-crack Behavior crack growth process. Small-crack analysis
Significant? techniques may provide valuable new insights
into some difficult LCF lifing problems, such
Small-crack behavior will not be an impor- as load interaction effects.
tant issue for applications in which initial Durability assessments of aircraft structures
defects are large or fatigue cracks are large at based on fracture mechanics typically employ
initiation. Small cracks are not a consideration much smaller initial flaw sizes than for
for many traditional mechanical engineering standard damage tolerance assessments (e.g.,
design/analysis applications employing damage 0.0127 cm (0.127 mm) rather than 0.127 cm
tolerance concepts (such as aircraft structures), (1.27 mm)). The initial flaw size for a durability
because the initial flaw size (which may be assessment may also be based on an equivalent
based on conventional NDE inspection limits) initial flaw size (EIFS) that is back-calculated
is usually much larger than the small-crack from some economic total life. In either case, it
regime. Furthermore, small-crack behavior will is possible that the initial flaw may exhibit
not be practically significant for some high- small-crack growth characteristics.
cycle fatigue (HCF) problems in which a very Small-crack behavior may also be very
large fraction (perhaps 490%) of the total life important for HCF applications when various
is consumed by the formation of even a damage states (e.g., foreign object damage
microscopic crack, which then grows quickly (FOD), fretting, or superimposed LCF) are
to failure under the high-frequency loading. also operative (Hudak et al., 2002). Although
However, small-crack behavior must be pure HCF (constant amplitude loading with no
considered in numerous other fracture me- superimposed damage) may induce cracking
chanics applications. When damage tolerance relatively late in life, these damage states can
methods are applied to more highly stressed produce a small crack relatively early in life.
safety-critical structures, tolerable flaw sizes The original damage source itself may not be
are generally smaller and NDE requirements able to grow the crack to failure, but the
are stricter. These applications have histori- critical question is whether the crack will
cally been treated with safe-life methods based propagate as a result of the HCF loading.
 Engineering Lifetime Prediction 151
Subsequent crack growth will probably depend often calls for simpler, easier-to-implement
on the magnitude of DKTH(a), because any approaches. Although these engineering pro-
crack growth will quickly result in failure due cedures may omit some details, they generally
to the high frequency of loading; thus, the capture the basic mechanisms adequately to
potential crack size effect on threshold may be accurately predict fatigue lifetime.
very significant. Engineering approaches to life prediction are
typically restricted to cracks that are at least of
the order of the microstructural dimension.
4.05.5.2 When is a Crack Small? Cracks smaller than the grain size introduce a
host of other analytical issues, including
Beyond these general qualitative guidelines,
mixed-mode cracks and highly localized plas-
several specific quantitative criteria are avail-
ticity effects. Cracks that are larger than a few
able to help determine when a crack is ‘‘small.’’
grains are often successfully modeled (albeit
Just because a crack is physically smaller in
approximately) using conventional mode I
comparison to traditional fracture mechanics
fracture mechanics, including regularly shaped
applications does not mean that its growth
crack models (e.g., semi-ellipses) and stress
rates will exhibit small-crack anomalies.
intensity factor solutions.
One criterion compares the physical crack
As noted in the introduction, small-crack
size to the Kitagawa diagram parameter, ao.
behavior may differ from conventional large-
Cracks that are distinctly larger than ao are less
crack behavior in two ways: non-negligible
likely to exhibit small-crack behavior, whereas
growth below the large-crack threshold, or
cracks that are of the order of ao or smaller are
growth above the large-crack threshold at
likely to show small-crack effects. A second
faster rates than would be predicted by
and potentially related criterion compares the
conventional large-crack fracture mechanics.
crack size directly to a dimension with micro-
Therefore, it is appropriate to consider two
mechanical relevance, often the grain size.
types of engineering life prediction models,
Cracks that are larger than about five times
addressing first the small-crack threshold and
this characteristic dimension are less likely to
then accelerated small-crack growth above
exhibit small-crack effects; cracks that are of
threshold.
the order of the microstructural dimension or
smaller are highly likely to show some micro-
structural influence that will cause crack 4.05.5.3.1 Crack growth threshold effects
growth to differ from large-crack behavior.
Another set of diagnostic criteria focuses on The simplest engineering treatment of small
a mechanical characterization of the global or cracks is to adopt a smaller value of the crack
local loading. If the applied stress range is growth threshold than would ordinarily be
distinctly lower than the endurance limit, then used for large cracks. One useful engineering
small-crack growth is unlikely. If the applied approach to predicting the effect of crack size
stress range is of the order of the endurance on threshold is the El Haddad expression cited
limit or higher, then the Kitagawa diagram previously (see comments accompanying Equa-
suggests that the crack will not observe the tion (10)):
large-crack threshold, and this may imply a rffiffiffiffiffiffiffiffiffiffiffiffiffi
sc lc a
small-crack effect. This criterion may be DKth ðaÞ ¼ DKth
a þ ao
ultimately based on a comparison between
the crack size and the size of the plastic zone at lc
where DKth is the measured large-crack thresh-
the crack tip (see Table 1). Therefore, other
old. This approach has been employed, e.g., in
factors that elevate the size of the local plastic
the widely used NASGROt (2002) software
zone—such as the presence of a stress concen-
for FCG analysis. The parameter ao can be
tration from which the crack is growing—may
easily estimated from the endurance limit and
also influence small-crack behavior. When the
large-crack threshold values, which are com-
crack size is not significantly larger than the
monly available. However, this expression
crack-tip cyclic plastic zone size, small-crack
should not be used for crack sizes smaller than
behavior should be expected.
the microstructural unit size.
Another simple, generally conservative ap-
4.05.5.3 Practical Life Prediction Methods proach to small-crack thresholds is to employ a
‘‘closure-free’’ large-crack threshold as a lower-
As noted earlier under analytical mechanics, bound representation of the small-crack
a variety of detailed analytical and numerical threshold (Hertzberg et al., 1992). This ‘‘clo-
models are available to address small-crack sure-free’’ threshold is either obtained from
behavior. However, engineering life prediction constant-R threshold tests at RZ0.7 or from
152 Small Fatigue Cracks
constant-Kmax, increasing-R threshold tests the use of multiple analytical corrections,
(Doker et al., 1982), which can be performed including adjustments to both the threshold
more quickly and which typically terminate at and the crack driving force above threshold.
R of the order of 0.9. The resulting threshold Because the problem is so multidimensional, it
has been shown to be a conservative represen- is not practical to illustrate the solution with a
tation of small-crack data for many engineer- simple demonstration problem. Instead, re-
ing and advanced materials (Ritchie and commendations are given for practical engi-
Peters, 2001). This approach may be over- neering approaches to various aspects of the
conservative as the crack grows to larger sizes, small-crack problem. The significance of these
because it does not provide a defined transition approaches for life prediction will depend on
to large-crack behavior. the specific problem. In some cases, corrections
It is important to note that some reported for the small-crack behavior may have a trivial
large-crack threshold values are themselves impact on the life calculation, while in other
anomalously high due to the threshold test cases, the difference between including or not
procedure itself, in which load shedding causes including small-crack behavior may be the
artificial increase in crack closure (McClung, difference between infinite survival and rapid
2000). In fact, some investigators have sug- failure.
gested that all large-crack thresholds obtained
from conventional load-shedding methods are 4.05.5.3.2 Accelerated crack growth above
anomalous, and that a large-crack threshold threshold
properly measured will in fact agree with the
small-crack threshold (Forth et al., 2002). In some cases, the growth rates of the small
These issues are unresolved as of early 2000s, cracks will be faster than exhibited by large
but highlight the need for caution in the cracks at the same nominal DK values, and
characterization and use of crack growth these differences must be predicted to perform
thresholds. For example, an anomalously high accurate life prediction. El Haddad et al. (1979)
large-crack threshold can lead to a substantial suggested that accelerated growth rates could
overprediction of the Kitagawa parameter, be properly predicted by adding ao to the
ao, as well as the tolerable levels of crack size physical crack size. However, some of his
and stress. demonstration problems included other effects
If the threshold is adjusted to a smaller such as mechanical notches, and the generality
value, the question of how to characterize of this recommendation has not been verified.
crack growth rates between the original and Some of the differences in growth rates arise
adjusted threshold values emerges. It is com- from differences in crack closure. This is
mon, and often adequate, to extrapolate the especially true when the small cracks are
trends of the original large-crack data above growing under substantially stresses that are a
threshold downward to the new threshold. large fraction of yield stress. A wide range of
Hudak et al. (2002) have employed the El numerical and experimental studies has shown
Haddad small-crack threshold equation in that crack opening behavior is significantly
practical engineering methods to predict the influenced by the magnitude of the applied
threshold stresses for HCF in the presence of stresses (McClung, 1991a, 1991b). Simple ana-
supplemental damage such as FOD or fretting. lytical treatments that describe this effect, such
Both FOD (which creates a small tear) and as the closure equations derived by Newman
fretting involve steeply decreasing stress gra- (1984) in the FASTRAN modified Dugdale
dients near the specimen surface where small closure model, are available to the practicing
cracks might grow. The question of whether a engineer. The Newman equations predict
small crack will arrest or continue to grow is changes in the effective range of the stress
answered by comparing the small-crack thresh- intensity factor as a function of maximum
old (which increases with crack size) with the stress, stress ratio, and constraint.
crack driving force (which may increase or Newman (1983) has also employed FAS-
decrease with crack size, depending on the TRAN to address small-crack behavior by
nature of the local stress field). modeling the initiation site of small cracks. As
Because of the wide range of phenomena noted earlier, small cracks growing from large
that cause small-crack effects, and the wide mechanical notches can exhibit accelerated
range of modeling approaches required to growth rates, and Newman has argued that
address these phenomena, the practicing en- this is also an explanation for the accelerated
gineer must be careful to select the appropriate growth of small cracks growing from small
model. All small cracks are not the same, and microstructural features such as inclusions.
all small-crack models do not address the same While changes in crack closure do appear to
behavior. Some practical problems may require be one of the factors in the accelerated growth
 Engineering Lifetime Prediction 153
rates of microstructurally small flaws (David- crack closure model has been proposed
son and Hudak, 1995), it is not clear that the (McClung, 1991a, 1991b) for FCG near stress
Newman approach captures the correct me- concentrations that estimates the local crack
chanisms. Nevertheless, the Newman approach opening stresses from: (a) the local stresses at
has exhibited some success and may be a useful maximum load in the uncracked, notched body,
engineering tool. and (b) the crack opening stresses at the same
Crack closure corrections change the effec- applied stress in a cracked, unnotched body.
tive crack driving force and hence the predicted This notch effect on closure is also addressed
growth rate. However, as stresses increase explicitly by the Newman FASTRAN model
further, closure corrections alone may be (Newman, 1983).
insufficient to explain and compensate for It was noted earlier that a substantial
changes in growth rates. In particular, as portion of the total life of a smooth specimen
nominal and local stresses increase, the con- may actually comprise crack growth, albeit the
ventional linear elastic fracture mechanics growth of very small flaws. That observation
description of the crack driving force in terms has led a number of researchers to attempt to
of DK may be inadequate. In this case, the construct S–N curves entirely from fracture
elastic–plastic description of the crack driving mechanics arguments, i.e., by integrating crack
force in terms of the closure-corrected range growth equations. Newman and co-workers
of the J-integral, DJeff, is a practical substi- (Newman, 1995) have explored this area
tute (McClung et al., 1997); see analytical extensively, employing the FASTRAN model
modeling. This engineering analysis treats (with the micronotch crack initiation site
not only the closure effects but also includes described earlier) to generate S–N curves for
plastic effects (enhancements in crack-tip several common aerospace alloys.
plasticity due to crack size) as well as the fully
plastic effects for severe reversed loading. The 4.05.5.4 Statistical Issues
J-integral itself can often be estimated with
acceptable engineering accuracy using the No discussion of life prediction that includes
reference stress method, which requires only a small-crack growth phenomena would be
K solution, a limit load solution, and an complete without some mention of the varia-
elastic–plastic constitutive model. bility in growth rates. From an engineering
Note that these stress-driven changes in perspective, FCG exhibits both macroscopic
crack closure and in the elastic–plastic crack variability (from crack to crack) and micro-
driving force generally lead to a uniform scopic variability (along the path of a single
translation of the da/dN vs. DK relationship; crack). The variability arises, at least in part,
in other words, the new crack growth curve has because all cracks are growing through vari-
the same slope but is shifted upwards in the able microstructures. Small cracks have fre-
linear Paris regime from the original baseline quently been observed to exhibit greater
large-crack curve. These acceleration effects growth rate variability than large cracks. When
are not diminished as the crack grows. the crack is small compared to the microstruc-
Other small-crack effects are more explicitly ture, and therefore interrogates only a small
dependent on crack size, and hence result in number of microstructural features, the effects
changes in the slope of the da/dN vs. DK of the microstructural variability on crack
relationship as the crack grows. These phe- advance may become more pronounced, and
nomena include microstructural effects and this behavior becomes even more pronounced
notch effects, both of which may involve for cracks growing near a threshold condition.
closure transients as well as local changes in However, some of the variability in small-
plasticity effects. crack growth rates is an artifact of crack length
Cracks growing from notches may exhibit a measurement errors and the small increments
small-crack effect that must be addressed by a in crack growth between measurements. When
modification in the life prediction methods. the crack size interval between successive
The effect appears to be significant only when measurements is small, the relative contribu-
the fatigue loading causes substantial plasticity tion of measurement error to the calculated
at the notch root and only for crack lengths value of da/dN increases. Furthermore, when
less than half of the notch root radius. It the crack measurement interval is small,
should be noted that notch root plasticity can successive measurements may capture micro-
also cause significant stress redistribution and scopic variations in crack growth rates that
changes in the local stress ratio, both of which would be averaged out naturally by a larger-
will impact the crack driving force, and these crack measurement interval (as for large
may also need to be addressed by an appro- cracks). Large cracks also show greater varia-
priate elastic–plastic stress analysis. A simple bility when crack length measurements are
154 Small Fatigue Cracks
taken at the same intervals as for small cracks structural integrity, and the key question is
(Vaidya, 1992). whether small cracks grow at applied DK
Unless very careful measurements are made, values below the large-crack threshold.
variability in crack growth rates appears as In ceramics that exhibit phase-transforma-
though it is scatter that is based on sampling a tion toughening (Ritchie and Peters, 2001),
random process. In engineering analysis, it small cracks appear to grow below the large-
may be useful to characterize variability using crack threshold, but at decreasing rates that
the methods of statistics (Kendall and King, often lead to arrest. However, large cracks
1988; Boyd-Lee and King, 1993). experience considerable crack-tip shielding due
The proper treatment of small-crack growth to the local phase transformations, and in fact
rate variability depends at least in part on the the development of stable zones of transformed
overall analysis goals. If safety is the para- material appears to explain the deceleration of
mount concern, then it may be more important small cracks. Large-crack data that are prop-
to characterize the ‘‘worst-case’’ upper-bound erly corrected for this transformation shielding
growth rate in some manner (Ritchie and appear to provide a reasonable lower bound to
Peters, 2001). Alternatively, if accuracy is the the small-crack behavior. In ceramics that do
paramount concern, then it may be necessary not exhibit phase transformations (Ritchie and
to capture the full scatter in growth rate Peters, 2001), small-crack data appear to agree
behavior, including periods when cracks are closely with large-crack data if proper correc-
nonpropagating, using a statistical method tions are introduced for any residual stress
(Boyd-Lee and King, 1993). It is essential to artifacts introduced by the small-crack experi-
ensure that any statistical treatment of small- mental procedures.
crack behavior leads to an accurate description Of the various intermetallic alloys considered
of scatter in total crack growth life, not merely for structural application, only gamma-based
scatter in the instantaneous crack growth rate. TiAl has been evaluated in depth for small-
A common mistake in a statistical treatment of crack effects (Ritchie and Peters, 2001; Chan
FCG is to focus on capturing the scatter in and Shih, 1997, 1998). Although lamellar
instantaneous crack growth rate. Since crack microstructures clearly exhibit superior fracture
growth rates are often self-averaging (cracks toughness and large-crack fatigue behavior due
speed up and slow down as they grow), to ligament bridging, cracks smaller than the
engineering treatments focusing on variation bridging zones can propagate at stress intensity
in da/dN can overestimate the observed varia- factor ranges well below the large-crack thresh-
tion in N. At this writing, no standard, widely olds. In the duplex microstructures, small
accepted methods for statistical treatment of cracks and large cracks exhibit similar growth
small-crack behavior have emerged. rates, and so a lower-bound large-crack thresh-
old is a practical approach to modeling small-
4.05.5.5 Small Cracks in Advanced Materials crack behavior. Such a design safeguard is
more difficult to achieve for lamellar micro-
Most of the preceding discussion was fo- structures because of a larger grain size coupled
cused on the behavior of fatigue cracks in with a wider variation in grain size.
traditional engineering metals, which exhibit
some ductility. In contrast, many advanced 4.05.6 ENVIRONMENTAL EFFECTS
materials (such as ceramics and intermetallics)
typically exhibit little or no ductility, and so Chemically active environments can enhance
critical crack sizes in these materials are often the rate of FCG in metals by reacting with the
relatively small. As a result, the behavior of fresh and highly reactive surface produced by
small fatigue cracks may be especially signifi- active straining at the crack tip. The enhance-
cant. The question is whether these advanced ment in crack growth rate is the result of either
materials exhibit small-crack effects similar to the individual or combined effects of film
ductile engineering metals. rupture/anodic dissolution or hydrogen em-
As a general rule, advanced materials exhibit brittlement at the crack tip. These processes
very high Paris exponents (10–50); i.e., crack tend to accentuate the influence of loading and
growth rate, da/dN, increases sharply with microstructural variables on FCG rates. The
small increase in the stress intensity factor focus here is to examine the influence of
range. In fact, crack growth rates are actually environment on the growth kinetics of small
much more sensitive to Kmax than to DK. cracks. In doing so, it is convenient to consider
Threshold values are often relatively close to two classes of environments—gaseous and
fracture toughness values, and crack growth liquid—since the influence of these environ-
typically leads to rapid failure. Therefore, the ments on small crack behavior can differ
threshold is the critical parameter in design for significantly.
 Environmental Effects 155
4.05.6.1 Gaseous Environments cracks caused the effects of oxidation enhanced
crack closure to be minimized, while for large
Most of the small-crack data discussed cracks there can be a large influence of crack
previously have been generated in laboratory wake, as has been shown especially for
air. Although this environment is often as- aluminum alloys and steels. In ferritic steels,
sumed to be benign, this is often not the case transgranular crack growth can be changed to
because of the presence of low partial pressures intergranular cracking when cracks are propa-
of water vapor. This gaseous environment can gated in humid air, and this has been shown to
cause enhancements in crack growth rates of be caused by the transport of H þ to grain
an order of magnitude or more (vs. vacuum), boundaries by crack tip plasticity. Hydrogen
particularly for high-strength materials. Un- also alters the growth of cracks in titanium
fortunately, most of these studies have not alloys by a similar mechanism.
explicitly considered the influence of water The effects of air, dry nitrogen (a few ppm of
vapor on small-crack behavior. This is because water vapor), and vacuum were compared for
it is difficult to generate small-crack data in small cracks growing in 7075 T651 using the
vacuum, which serves as a baseline for asses- data of Petit (1999) and Lankford (1982).
sing the extent of environmentally assisted Vacuum and dry nitrogen gave similar results.
crack growth. Nevertheless, a limited number From the crack-tip micromechanics and com-
of systematic studies of the influence of water parisons of crack growth rates, it is concluded
vapor and other gases such as hydrogen have that the effect of water vapor on crack growth
been made and it is worthwhile to briefly is much greater than the effect of crack length.
summarize them here before proceeding to the The next section examines the effects on FCG
more complex case of liquid environments. of liquid environments which are much differ-
A review of the influence of gaseous envir- ent from those just described for gaseous
onments on the growth of both short (through) environments.
and small (surface) fatigue cracks in several
materials has been given by Petit (1999). 4.05.6.2 Liquid Environments
Although there was some influence of crack
length on crack growth in vacuum for 7075 A chemical crack size effect can occur in
T651 aluminum alloy, Ti–6Al–4 V, and E460 high-strength materials, particularly steels, that
steel, the predominant effect was that of air, are exposed to aerated aqueous environments.
which increased the growth rates of both This phenomenon is due to a mechanism
large and short cracks relative to vacuum, different from that of the fatigue small-crack
and for all three materials, the growth rate was phenomenon. The chemical crack size effect,
higher for short cracks, giving similar results first reported by Gangloff (1981), was illu-
for small cracks. These results were obtained at strated by the data in Figure 7 for the case of
35 Hz. Most of the differences observed were 4130 steel in a deaerated aqueous 3.5% NaCl
attributed to the effect of air on fatigue crack environment (Gangloff, 1985). As can be seen
closure levels, which were measured from corrosion-FCG rates from small surface cracks
compliance using a back face strain gauge. (0.1–1 mm deep), as well as short through-
Short cracks were artificially created by remov- thickness edge cracks (0.1–0.3 mm), are appre-
ing the wake of large cracks using spark ciably faster than corrosion-FCG rates from
erosion cutting. long through-thickness cracks (25–40 mm) in
Enhanced crack growth in air has been compact tension specimens. Moreover, all of
attributed to the influence of hydrogen. The the corrosion-FCG rates in NaCl solution are
water molecule dissociates into H þ and OH
enhanced compared to those in a moist
at the crack tip where unoxidized metal is laboratory air environment even though the
exposed by crack growth. The H þ atom is then latter were generated for both small and large
carried into the zone just ahead of the crack tip cracks. It is also interesting to note that the
by dislocations where it affects the cohesion of corrosion-FCG rates for small surface cracks,
the metal, which results in a lower strain to at a given DK value, decrease with increasing
fracture (crack tip strain) and CTOD (David- applied stress such that at 1,260 MPa the
son and Lankford, 1981; Davidson and Camp- results approach those for the long cracks in
bell, 1996). The crack growth increment (Da) is the compact tension specimen. Interestingly,
not as much affected as is the number of cycles this stress level dependence of the crack growth
(DN). There is an additional effect of oxidation kinetics of chemically small cracks is opposite
which causes closure to increase (Al alloys and to that for small fatigue cracks in gaseous
steels in water and water vapor and crack environments, which suggests mechanistic dif-
growth at high temperature) by filling the crack ferences between chemical and other small
wake. The lack of a crack wake for short fatigue cracks.
156 Small Fatigue Cracks
The chemical small-crack effect is believed to Crack-tip creation of H:
be governed by the availability of embrittling 2H þ þ 2e
¼ 2H
hydrogen at the crack tip, which in turn is Oxygen consumption of H:
1

controlled by the mass transport of ionic 2O2 þ 2H þ 2e ¼ H2O
species to the crack tip—a process that can
depend on crack-size and applied-stress level. For alloy steels, dissolution of chromium also
This mass-transport-controlled mechanism is occurs, and chromium hydrolysis contributes
very different from the mechanical mechanisms to additional crack-tip acidification, while the
of crack closure and large-scale yielding that presence of ionic species (such as Cl
) serves to
are believed to be responsible for small-crack maintain overall charge balance during the
effects in inert or mildly aggressive environ- above reaction sequence. The last reaction
ments (e.g., water vapor in air). above is critical since it serves to limit the
Gangloff (1984) proposed a specific mechan- availability of embrittling H needed for en-
ism for the chemical small-crack effect in steels hanced crack growth rates, and in turn depends
exposed to aerated aqueous environments that on the transport of oxygen to the crack tip.
is controlled by oxygen transport to the crack Mass transport modeling results in Figure 25
tip. According to this mechanism, the supply of shows that oxygen transport to the crack tip
embrittling hydrogen [H] at the crack tip occurs by diffusion for very small cracks
depends on the overall electrochemical reaction (o0.5 mm), and by convection (pumping) for
sequence involving: (1) anodic dissolution of larger cracks (45 mm), with suppressed oxy-
metal at the crack tip, (2) metal ion hydrolysis gen transport for intermediate crack sizes
that produces local crack-tip acidification, and (Gangloff, 1984; Turnbull, 1982). Note that
(3) hydrogen ion reduction that produces in this figure the crack-tip oxygen concentra-
embrittling atomic hydrogen [H], and hydrogen tion relative to the bulk oxygen concentration
ion consumption due to a cathodic reduction (Ctip/Cbulk) is represented on a negative loga-
reaction as oxygen is transported to the crack rithmic scale, thus large negative numbers
tip. For steels, the reactions would be as follows: indicate low oxygen concentrations. As shown
in Figure 25, oxygen depletion occurs ranging
Crack-tip dissolution: from 0.5 mm to 5 mm—the crack- size regime
Fe ¼ Fe þ þ þ 2e
where most crack growth data have been
Acidification by hydrolysis: generated and enhanced crack growth rates
Fe þ þ þ H2O ¼ (FeOH)
þ H þ have been observed. Consequently, as shown in

Low
Oxygen

Experiment
20
r = 10 sec (0.1 Hz)
∆K = 20 MPa.m1/2
R = 0.10
16 Convection
−log (Ctip /Cbulk)O2

12

Diffusion

High
Oxygen
0
10− 2 10− 1 1 10
Crack Depth (mm)
Figure 25 Analytical mass transport predictions of relative oxygen concentration at the tip of a growing
corrosion fatigue crack in steel exposed to aerated NaCl at a DK value of 20 MPa m1/2 illustrating the crack
size dependence for diffusion and convection (crack pumping) processes (sources Gangloff, 1984; Turnbull,
1982).
 Environmental Effects 157
1.0
4130 Steel 3% NaCl
∆K = 22 MPa.m1/2 10−2

(da/dNCF-Net)/(da/dNCF-Abs)

da/dNCF-Net (mm/cycle)
0.10

10−3

0.01 SEN
Surface
CT
10−4
1 10
Crack Length (mm)
Figure 26 Corrosion-FCG rate vs. crack length in 4130 steel exposed to aerated 3% NaCl at a DK of
22 MPa m1/2 (Gangloff, 1984). Note similarity of crack length dependence of rates to crack length dependence
of oxygen transport in Figure 24. Solid symbols indicate transgranular cracking, open symbols indicate
intergranular cracking, and half-solid symbols indicate mixed transgranular–intergranular cracking. (da/
dNCF-Net is measured corrosion fatigue crack growth rate, and da/dNCF-Abs is theoretical maximum rate
shown in Figure 7.)

Figure 26, crack growth rates at a given DK results in Table 3 also show that the limiting
exhibit a crack-size dependence with a max- crack size below which chemical crack-size
imum at intermediate crack sizes that is effects are observed and above which they are
strikingly similar to that for minimum oxygen not observed ranged from 0.8 mm to 5 mm.
(maximum H) concentration in Figure 25. Thus, chemical crack size effects occur over a
The point of minimum oxygen concentration significantly larger-crack size range than is
in Figure 25 is also influenced by applied stress found for small cracks growing in air. Conse-
since increased applied stress causes increased quently, chemical crack size effects could be of
(mean) CODs and consequently less efficient potentially greater practical significance in life
crack pumping. Thus, crack growth rates for assessment and prediction than are fatigue
chemical small-crack data tend to decrease crack size effects. The critical unresolved
with increasing COD, as shown in Figure 27 question is: How widespread are chemical
for a range of crack geometries, stress levels, crack size effects?
and loading conditions (Gangloff, 1984). As indicated in the last two columns of
Most of the small-crack data in liquid Table 3, chemical crack size effects have been
environments have been generated for steels reported to be absent in aluminum alloys 2024
exposed to aqueous environments containing and 7075 exposed to an aqueous 1% NaCl
either sodium chloride (NaCl) or synthetic environment (Piascik and Willard, 1994; Pias-
seawater. These results are summarized in cik, 1999). However, it should be noted that
Table 3, updated from a previous review these data were generated at a relatively high
(Gangloff and Wei, 1986) to include new data cyclic frequency of 5 Hz, where environment
on aluminum alloys. As indicated, steels enhanced crack growth rates are not likely to
exhibit small-crack growth rates that are have occurred at any crack size. A similar lack
accelerated by factors of 2–500, compared to of a chemical crack size effect at high frequency
long-crack data at comparable DK values. (10 Hz) was also observed in ASTM A289-B
Furthermore, the degree of crack growth rate (see Table 3). This emphasizes the point that
enhancement due to the chemical crack size chemical crack size effects should only be
effect depends primarily on material strength expected to occur under conditions that result
level—with increasing strength resulting in in environment-enhanced fatigue crack growth
enhanced chemical crack size effects. The rates for long cracks. The above aluminum
158 Small Fatigue Cracks

4130 Steel 3% NaCl

SEN
Surface Crack
CT
SC−High R
(da/dNCF-Net)/(da/dNRef)

100

10

1 10 100 1000
Maximum Crack Opening @ Notch (
m)

Figure 27 The dependence of corrosion fatigue crack growth rates of small cracks on the maximum crack
mouth opening displacement for a variety of specimen geometries (Gangloff, 1984). (Specimen types:
SEN ¼ single edge notch, CT ¼ compact tension, SC ¼ surface crack. da/dNCF-Net is measured corrosion
fatigue crack growth rate, and da/dNRef is reference air rate shown in Figure 7.)

alloy data were also generated in a deaerated preferred and most widely used specimens
environment; thus, even if the test frequency promote the initiation of naturally small sur-
had been lower, this environment may not have face or corner cracks in rectangular or cylind-
produced chemical crack size effects via the rical specimens, rather than growing a large
mechanism described above for steel for which crack and then machining away material in the
the presence of oxygen is required. Thus, crack wake to leave a small or short crack.
additional experiments on aluminum alloys, However, it is inherently difficult to find a
as well as other nonferrous alloys, are war- microscopically small crack in the test speci-
ranted to further assess whether or not the men after it has initiated, but while it is still
chemical crack size effect occurs in materials small, in order to monitor its growth. It is
other than steel. easier to find the crack if its probable location
can be somehow restricted. For example, a
4.05.7 EXPERIMENTAL TECHNIQUES three-point bend specimen or a tensile speci-
men with a mild hourglass shape restricts the
Fatigue life prediction that includes small- initiation site to a narrow band along the
crack effects often uses baseline data collected region of highest stress. A sharper notch is
from conventional large-crack and smooth more successful in localizing the initiation site,
specimens, which are combined using the but the notch itself can significantly influence
Kitagawa diagram. In some applications, the resulting crack growth behavior.
however, it may be necessary to obtain direct For materials with few natural defects, some
experimental evidence for small-crack beha- investigators have put microdefects (such as
vior. Unfortunately, small-crack growth rates extremely small electro-discharge machined
can be measured with the standard test notches, environmental pits induced by a drop
procedures developed for large cracks. Small- of a corrosive fluid, or ion beam milled holes)
crack tests usually require different specimen into the specimen to serve as crack starters.
geometries and different specimen preparation However, it can be difficult to manufacture a
techniques, different crack length measurement microdefect that is small enough not to
techniques and equipment, and different data influence the growth behavior of the resulting
analysis techniques. small fatigue crack in the size range of interest,
The study of small cracks requires detection while at the same being big enough to cause the
of crack initiation and growth while physical fatigue crack to form at the defect before
crack sizes are extremely small, and this another crack initiates at some other location
requirement influences specimen design. The in the specimen. Furthermore, any biasing of
 Experimental Techniques 159
Table 3 Literature results demonstrating the effect of crack size on corrosion fatigue in materials exposed to
aqueous environments.
Limiting
chemical small
Yield stress Crack size da=dNSMALL crack size
Material (MPa) Environment effects? da=dNLONG (mm)

AISI 4130 (0.3C– 1,330 3% NaCl No—air, 1.2–500 42.5

0.9Cr–0.2Ni) vacuum
Yes—NaCl
ASTM A289-B 1,120 655 kPa H2 No—air 1.0–100 0.8
(18.2Mn–4.8Cr– 231C
0.6C)
Sat. H2O No—10 Hz
H2
801C Yes—
0.02 Hz
H2
HY130 (0.1C– 930 3% NaCl No—air 4 41.0
0.5Cr–5Ni)
Yes—NaCl
972 3% NaCl No—air 2 1.1
Yes—NaCl
972 3% NaCl No—air 1.0–1.8 free 0.9–1.4
corrosion
Yes—NaCl 0.8–1.3 cathodic 5
950 Seawater No 1.0 o0.5
13Cr (0.03C– 770 Water Yes 4 ?
12.8Cr–5Ni)
Q1N (0.17C– 625 Seawater No—air 4 2.0
1.2Cr–2.4Ni)
Yes—NaCl
BS4360: 50D 370 Seawater No—air 3 1.0–3.0
(0.18C–1.2Mn)
Yes—NaCl
EN5 (0.3C– 300 Seawater Yes—NaCl 2 2.0
0.7Mn)
2024Al 166 1% NaCl, No 1
deaerated
7075Al 517 1% NaCl, No 1
deaerated
After Gangloff and Wei (1986).

the crack origin (by either means) to an manner that will minimize residual stresses
‘‘unnatural’’ location could have some influ- (Larsen et al., 1992). Preparation methods that
ence on the resulting growth behavior for the terminate with low-stress mechanical and then
smallest flaw sizes if the local microstructure nonmechanical operations (e.g., electropolish-
differs from locations where natural initiation ing or chemical polishing) may be preferable.
would have occurred. This limitation can be However, since the growth rates of small
circumvented by studying a sufficiently large surface cracks in engineering components can
number of cracks that many microstructural be influenced by residual stress fields arising
variations are sampled. from fabrication of the component, the resi-
Specimen fabrication (e.g., machining and dual stress similitude—or lack thereof—be-
polishing the surface) introduces near-surface tween test coupon and component residual
residual stresses of some magnitude. Residual stresses should be considered when the labora-
stresses that are small enough to have only a tory data are applied to component life
minor effect on the behavior of large cracks or prediction.
the total lifetime of a smooth specimen may Specimen preparation methods may also
still be large enough to have a pronounced influence the visibility of the small crack on
effect on the growth behavior of a small crack. the specimen surface, which can be critical for
Therefore, the specimen preparation methods any visual method of crack length measure-
should be carefully chosen and carried out in a ment. It is easier to find small cracks visually
160 Small Fatigue Cracks
using polished surfaces, rather than etched, be applied during data analysis. However,
and the etch may influence the initiation variations in the crack shape can be a source
process. The relationship between microstruc- of inaccurate DK calculation, especially for
ture and the crack is then determined by microstructurally small cracks, and some con-
etching after crack growth. firmation of crack shape is desirable. Fracto-
Several well-established experimental tech- graphy can often be used to obtain crack shape
niques are available for measuring the size of as a function of size. Interactions between
small fatigue cracks, and hence deducing their closely spaced multiple cracks that affect
growth rates. These techniques include replica- growth rates are more likely to occur in the
tion, optical microscopy, scanning electron small-crack regime (especially under higher
microscopy, and potential drop. Each techni- cyclic stresses) and must be addressed. Special
que has unique strengths and limitations, and attention must be given to the minimum
different techniques are optimum for different interval between successive crack length mea-
circumstances. All are useful for measuring the surements, Da. Closely spaced measurements
growth of fatigue cracks with the size of the are often needed to capture key crack-micro-
order of 50 mm and greater, and some are structure interactions, but measurement error
applicable to even smaller cracks. Detailed can significantly influence apparent variations
descriptions of these and other techniques are in da/dN for extremely small Da values (Larsen
available in Larsen and Allison (1992). et al., 1992). Some of the alleged evidence for
Replication (Swain, 1992) is one of the most large scatter in small-crack growth rate data is
commonly used techniques. While fatigue actually just the result of measurement inter-
cycling is interrupted and a static force is vals that are too small relative to the measure-
applied to the specimen, a small piece of thin ment error. Variation in crack growth rates
cellulose acetate sheet is softened with acetone, also occurs when cracks arrest during part of
gently applied to the specimen surface, and the measurement interval, which can be com-
allowed to dry for a few minutes. The acetate mon for small fatigue cracks.
replica forms a permanent record of the surface Further information is available in appendix
topography, and is subsequently viewed in an X3 to ASTM Test Method E 647 (2002)
optical or scanning electron microscope to (‘‘Guidelines for measuring the growth rates
measure surface crack length. Replication is a of small fatigue cracks’’). This appendix does
useful means to record the growth of small, not prescribe complete, detailed test proce-
naturally initiated flaws by working backwards dures. Instead, it provides general guidance on
from a larger crack in a chronological series of the selection of appropriate experimental and
replicas. However, replication has been shown analytical techniques and identifies aspects of
to influence crack growth rates artificially in the testing process that are of particular
some materials due to chemical interactions, importance when fatigue cracks are small.
and replication can also be very labor intensive
and time consuming.
Direct microscopic methods, either optical 4.05.8 CONCLUSIONS
(Larsen et al., 1992) or SEM (Davidson, 1992),
permit direct inspection of the crack and may 1. In general, small fatigue cracks exhibit
facilitate the permanent capture of a photo- different growth behaviors than large
graphic record, but require advance knowledge fatigue cracks. This is particularly true
of the crack origin and may be limited in for structural metals and alloys that have
resolution by the available microscopic equip- high levels of fracture toughness, but the
ment. Nonvisual techniques such as potential differences may be smaller for brittle
drop (Gangloff et al., 1992) offer automation materials.
advantages but again require prior knowledge 2. A fatigue crack is considered to exhibit
of where the crack will form (since probe small-crack behavior when it grows below
location is critical) and may also be limited in the large-crack DK threshold or when the
resolution by the available instrumentation. It growth rates are above an extrapolation of
may be possible to use replication to locate a the linear (Paris) region of da/dN vs. DK
crack while it is still small, and then use a high- for large cracks.
resolution method such as the SEM to study 3. The similitude exhibited by large cracks in
ensuing growth in more detail, but the resolu- the linear (Paris) region is lost for small
tion of the replica itself is limited, and usually fatigue cracks because the crack driving
less than the SEM. force cannot simply be described by
Many small surface cracks develop shapes LEFM.
that are approximately semi-elliptical, and the 4. COD measurements indicate that cyclic
standard K solutions for these geometries can stress intensity factor (DK) is a valid
 References 161
description of the driving force for fatigue use of the EPFM parameter DJ. Engineer-
cracks, small and large; however, compu- ing methods have been developed for both
tation of DK from crack length and applied approaches.
stress may seriously underestimate the 12. Cracks growing from stress concentrations
local, crack-tip, driving force for small such as notches can exhibit small-crack
cracks. effects including acceleration, deceleration,
5. Defining the size range for small-crack and arrest. These phenomena can be
behavior is complex because all physically explained in terms of changes in crack
small cracks do not exhibit anomalous closure and local crack driving force which
behavior. If experimental data from small include the effects of notch stress and
cracks are not available, a crack size below strain fields.
which small-crack behavior may be ex- 13. Considerable variation in growth rates is
pected can be estimated from microstruc- often observed for small fatigue cracks,
ture and by using the Kitagawa diagram. and these cracks often grow discontinu-
6. Determining the microstructural unit re- ously. Some growth rate variations may be
levant to the growth of small fatigue cracks an artifact of the techniques used to
may be complicated by allotropic trans- measure crack length. Variations in the
formations and texture effects, but can be growth rates can be included in calcula-
as simple as grain size. Below a crack size tions of lifetime through the use of
of 2–10 microstructural units, small-crack statistics.
behavior may be expected. There may be 14. Experimental techniques for measuring
conditions, however, when plastic zone size small FCG rates are substantially different
also needs to be included in this estimate from conventional techniques used for
(see Table 1). large cracks. A knowledge of small-crack
7. The Kitagawa diagram provides a method measurement techniques is essential for
for estimating a critical crack size, ao, obtaining valid growth rate data.
below which small-crack behavior may be
expected. This estimate is independent of The environment in which a small crack grows
direct microstructural and mechanics con- usually greatly influences its growth rate.
siderations. Cracks grow more slowly in vacuum than in
8. The most significant differences between humid air, and cracks growing in liquids often
small and large fatigue cracks are in crack exhibit a ‘‘chemical small crack effect.’’ Rea-
closure, crack-tip plasticity, constraint, sons for these effects are known.
and interactions with local microstructure.
Detailed micromechanical models have
4.05.9 REFERENCES
proven to be useful for characterizing and
explaining, and in some cases predicting, ASTM E647-00, 2002, Standard test method for measure-
these differences. ment of fatigue crack growth rates. Annual Book of
9. It may be necessary to include small ASTM Standards, vol. 03.01.
G. I. Barenblatt, 1987, On a model of small fatigue cracks.
fatigue crack behavior in engineering life- Eng. Fract. Mech., 28, 623–626.
time predictions. Selections of the appro- R. K. Bolingbroke and J. E. King, 1986, The growth of
priate models requires first determining the short fatigue cracks in titanium alloys IM1550 and
nature of the relevant small-crack effect. A IM1318. In: ‘‘Small Fatigue Cracks,’’ eds. R. O. Ritchie
number of engineering methods, simpler and J. Lankford, The Metallurgical Society, Warrendale,
PA, pp. 129–144.
than the detailed micromechanical models, A. Boyd-Lee and J. E. King, 1993, In: ‘‘Fatigue Design,’’
have been developed to include small-crack ESIS 16, eds. J. Solin et al., Mech. Engineering
behavior. Publications, London, UK, pp. 283–296.
10. Use of the El Haddad et al. model together A. Boyd-Lee and J. E. King, 1994, Short fatigue crack path
determinants in polycrystalline Ni-base superalloys.
with the Kitagawa diagram provides an Fatigue Fract. Eng. Mater. Struct., 17, 1–14.
explicit expression for the decrease in K C. W. Brown and D. Taylor, 1984, The effects of texture
threshold as crack size diminishes. Thresh- and grain size on the short fatigue crack growth rates in
olds for small cracks can also be approxi- Ti–6Al–4V. In: ‘‘Proceedings of the International Sym-
mated using crack growth data obtained posium on Fatigue Crack Growth Threshold Concepts,’’
eds. D. L. Davidson and S. Suresh, The Metallurgical
from large cracks grown at high stress Society of AIME, Philadelphia, PA, pp. 433–446.
ratios (o0.7), which might include con- K. S. Chan, 1986, Local crack-tip field parameters for large
stant Kmax tests. and small cracks theory and experiment. In: ‘‘Small
11. The accelerated growth rates of small Fatigue Cracks,’’ eds. R. O. Ritchie and J. Lankford,
The Metallurgical Society, Warrendale, PA, pp.
cracks sometimes observed above the 407–425.
large-crack threshold can be addressed K. S. Chan, 1999, Fatigue crack growth thresholds of TiAl
through crack closure corrections and the alloys. In: ‘‘Proceedings of the 2nd International
162 Small Fatigue Cracks
Symposium on Gamma Titanium Aluminides,’’ eds. Strain and Plastic Deformation Aspects of Fatigue
Y.-W. Kim, D. M. Dimiduk and M. H. Loretto, TMS, Crack Growth, ASTM STP 637,’’ ASTM, Philadelphia,
Warrendale, PA, pp. 517–525. PA, pp. 97–121.
K. S. Chan, 2002, A microstructure-based fatigue crack D. S. Dugdale, 1960, Yielding of steel sheets containing
initiation model for structural alloys. In: ‘‘Proceedings of slits. J. Mech. Phys. Solids, 8, 100–104.
Fatigue 2002,’’ ed. A. F. Blom, EMAS, West Midlands, M. H. El Haddad, N. E. Dowling, T. H. Topper and K. N.
UK, vol. 1, pp. 347–354. Smith, 1980, J-integral application for short fatigue
K. S. Chan and J. Lankford, 1983, A crack tip strain model cracks at notches. Int. J. Fract., 16, 15–30.
for the growth of small fatigue cracks. Scr. Metall., 17, M. H. El Haddad, K. N. Smith and T. H. Topper, 1979,
529–532. Fatigue crack propagation of short cracks. ASME J.
K. S. Chan and J. Lankford, 1988, The role of Eng. Mater. Technol., 101, 42–46.
microstructural dissimilitude in fatigue and fracture of N. A. Fleck and J. C. Newman, Jr., 1988, Analysis of crack
small cracks. Acta Metall., 36, 193–206. closure under plane strain conditions. In: ‘‘Mechanics of
K. S. Chan, J. Lankford and D. L. Davidson, 1986, A Fatigue Crack Closure, ASTM STP 982,’’ pp. 319–341.
comparison of crack-tip field parameters for large and N. E. Frost, K. I. Marsh and L. P. Pook, 1974, ‘‘Metal
small fatigue cracks. ASME J. Eng. Mater. Technol., Fatigue in Engineering,’’ Wiley, New York.
108, 206–213. S. C. Forth, J. C. Newman, Jr. and R. G. Forman, 2002,
K. S. Chan, Y.-M. Pan, D. Davidson and R. C. McClung, Generating fatigue crack growth thresholds with constant
1997, Fatigue crack growth mechanisms in HSLA-80 amplitude loads. In: ‘‘Fatigue 2002,’’ ed. A. F. Blom,
steels. Mater. Sci. Eng. A, A222, 1–8. EMAS, West Midlands, UK, Vol. 4, pp. 2337–2344.
K. S. Chan and D. S. Shih, 1997, Fatigue and fracture R. P. Gangloff, 1981, The criticality of crack size in
behavior of a fine-grained lamellar TiAl alloy. Metall. aqueous corrosion fatigue. Res. Mech. Lett., 1, 299–306.
Mater. Trans. A, 28A, 79–90. R. P. Gangloff, 1984, Oxygen inhibition model of the
K. S. Chan and D. S. Shih, 1998, Fundamental aspects of chemical crack size effect in corrosion fatigue. In:
fatigue and fracture in a TiAl sheet alloy. Metall. Mater. ‘‘Embrittlement by the Localized Crack Environment,’’
Trans. A, 29A, 73–87. ed. R. P. Gangloff, TMS-AIME, Warrendale, PA, pp.
R. G. Chermahini and A. F. Blom, 1991, Variation of 265–290.
crack-opening stresses in three-dimensions: finite thick- R. P. Gangloff, 1985, Crack size effects on the chemical
ness plate. Theor. Appl. Fract. Mech., 15, 267–276. driving force for aqueous corrosion fatigue. Metall.
R. G. Chermahini, K. N. Shivakumar and J. C. Newman, Trans., 16A, 953–969.
Jr., 1988, Three-dimensional finite-element simulation of R. P. Gangloff, D. C. Slavik, R. S. Piascik and R. H. Van
fatigue crack growth and closure. In: ‘‘Mechanics of Stone, 1992, Direct current electrical potential measure-
Fatigue Crack Closure, ASTM STP 982,’’ eds. J. C. ment of the growth of small cracks. In: ‘‘Small-crack
Newman, Jr. and W. Elber, ASTM, Philadelphia, PA, Test Methods, ASTM STP 1149,’’ eds. J. M. Larsen and
pp. 398–413. J. E. Allison, ASTM, pp. 116–168.
M. A. Daeubler and A. W. Thompson, 1986, Microstruc- R. P. Gangloff and R. P. Wei, 1986, Small crack-
tural effects on short fatigue cracks in eutectoid steels. environment interactions—the hydrogen embrittlement
In: ‘‘Small Fatigue Cracks,’’ eds. R. O. Ritchie and J. perspective. In: ‘‘Small Fatigue Cracks,’’ eds. R. O.
Lankford, The Metallurgical Society, Warrendale, PA, Ritchie and J. Lankford, The Metallurgical Society,
pp. 157–164. Warrendale, PA, pp. 239–264.
D. L. Davidson, 1992, The experimental mechanics of R. W. Hamm and W. S. Johnson, 1999, A unification of
microcracks. In: ‘‘Small-Crack Test Methods, ASTM small- and large-crack growth laws. Fatigue Fract. Eng.
STP 1149,’’ ASTM, pp. 81–91. Mater. Struct., 22, 711–722.
D. L. Davidson, 1999, Fatigue crack tip constraint and R. Hertzberg, W. A. Herman, T. Clark and R. Jaccard,
closure as a function of crack length. In: ‘‘Small Fatigue 1992, Simulation of short crack and other low closure
Cracks,’’ eds. K. S. Ravichandran, R. O. Ritchie and Y. loading conditions utilizing constant Kmax DK-decreas-
Murakami, Elsevier, Oxford, pp. 109–118. ing fatigue crack growth procedures. In: ‘‘Small-crack
D. L. Davidson, 2000, Mixed mode crack opening Test Methods, ASTM STP 1149,’’ ASTM, pp. 197–220.
displacement measurements for small fatigue cracks S. J. Hudak, Jr., 1981, Small crack behavior and the
growing in Astroloy at 201C. Fatigue Fract. Eng. Mater. prediction of fatigue life. ASME J. Eng. Mater. Technol.,
Struct., 23, 445–452. 103, 26–35.
D. L. Davidson and J. B. Campbell, 1996, ‘‘New S. J. Hudak, Jr. and K. S. Chan, 1986, In search of a
Techniques for Characterizing Corrosion and Stress driving force to characterize the kinetics of small crack
Corrosion,’’ eds. R. H. Jones and D. R. Baer, The growth. In: ‘‘Small Fatigue Cracks,’’ eds. R. O. Ritchie
Metallurgical Society, Warrendale, PA, pp. 163–174. and J. Lankford, The Metallurgical Society, Warrendale,
D. L. Davidson, K. S. Chan and R. C. McClung, 1996, Cu- PA, pp. 379–405.
bearing high-strength low-alloy steels: The influence of S. J. Hudak, Jr., K. S. Chan, G. G. Chell, Y.-D. Lee and R.
microstructure on the initiation and growth of small C. McClung, 2002, A damage tolerance approach for
fatigue cracks. Metall. Trans. A, 27, 2540–2556. predicting the threshold stresses for high cycle fatigue in
D. L. Davidson and S. J. Hudak, Jr., 1995, The crack-tip the presence of supplemental damage. In: ‘‘Fatigue—
mechanics and growth rates of small fatigue cracks in David L. Davidson Symposium,’’ TMS, pp. 107–120.
ASTROLOY. Metall. Mater. Trans. A, 26, 2247–2257. S. J. Hudak, Jr. and D. L. Davidson, 1988, The dependence
E. R. de los Rios, H. J. Mohamed and K. J. Miller, 1985, A of crack closure on fatigue loading variables. In:
micro-mechanics analysis for short fatigue crack growth. ‘‘Mechanics of Fatigue Crack Closure, ASTM STP
Fatigue Fract. Eng. Mater. Struct., 8, 49–63. 982,’’ eds. J. C. Newman and W. Elber, ASTM,
V. S. Deshpande, A. Needleman and E. Van der Giessen, Philadelphia, PA, pp. 121–138.
2003, Discrete dislocation plasticity modeling of short M. N. James and W. N. Sharpe, 1989, Closure develop-
cracks in single crystals. Acta Mater., 51, 1–15. ment and crack opening displacement in the short crack
H. Doker, V. Bachmann and G. Marci, 1982, ‘‘Fatigue regime for fine and coarse grained A533B steel. Fatigue
Thresholds,’’ eds. J. Backlund, A. Blom and C. Beevers, Fract. Eng. Mater. Struct., 12, 347–361.
EMAS, Warley, UK, vol. 1, pp. 45–57. J. M. Kendall and J. E. King, 1988, Short fatigue crack
N. E. Dowling, 1977, Crack growth during low-cycle growth behavior: data analysis effects. Int. J. Fatigue, 10,
fatigue of smooth axial specimens. In: ‘‘Cyclic Stress– 163–170.
 References 163
H. Kitagawa and S. Takahashi, 1967, In: ‘‘Proceedings of NASGROt, 2002, Reference Manual, Version 4.01,
the 2nd International Conference on Mechanical Beha- Southwest Research Institute.
vior of Materials, Boston, Massachusetts,’’ pp. 627–631. A. Navarro and E. R. de los Rios, 1988, An alternative
H. S. Lamba, 1975, The J-integral applied to cyclic loading. model of the blocking of dislocations at grain bound-
Eng. Fract. Mech., 7, 693–703. aries. Phil. Mag., 57, 37–42.
J. Lankford, 1982, The growth of small fatigue cracks in J. C. Newman, Jr., 1983, Nonlinear fracture mechanics
7075-T6 aluminum. Fatigue Eng. Mater. Struct., 5, approach to the growth of small cracks. In: ‘‘Behaviour
233–248. of Short Cracks in Airframe Components, AGARD CP-
J. Lankford and D. L. Davidson, 1984, Near-threshold 328,’’ ed. H. Zocher, pp. 6.1–6.26.
crack tip strain and crack opening for large and small J. C. Newman, Jr., 1984, A crack opening stress equation
fatigue cracks. In: ‘‘Fatigue Crack Growth Threshold for fatigue crack growth. Int. J. Fract., 24, R131–R135.
Concepts,’’ The Metallurgical Society, Warrendale, PA, J. C. Newman, Jr., 1994, A review of modeling small-crack
pp. 447–463. behavior and fatigue-life predictions for aluminum
J. M. Larsen and J. E. Allison (eds.), 1992, In: ‘‘Small- alloys. Fatigue Fract. Eng. Mater. Struct., 17, 429–439.
crack Test Methods, ASTM STP 1149,’’ American J. C. Newman, Jr., 1995, Fatigue-life prediction methodo-
Society for Testing and Materials. logy using a crack-closure model. J. Eng. Mater.
J. M. Larsen, J. R. Jira and K. S. Ravichandran, 1992, Technol., 117, 433–439.
Measurement of small cracks by photomicroscopy— J. C. Newman, Jr., 2002, Advances in finite-element
experiments and analysis. In: ‘‘Small-crack Test Meth- modeling of fatigue-crack growth and fracture. In:
ods, ASTM STP 1149,’’ ASTM, pp. 57–80. ‘‘Proceedings of Fatigue 2002,’’ ed. A. F. Blom, West
R. C. McClung, 1991a, The influence of applied stress, Midlands, UK, vol. 1, pp. 55–70.
crack length, and stress intensity factor on crack closure. J. C. Newman, Jr., M. H. Swain and E. P. Phillips, 1986,
Metall. Trans. A, 22, 1559–1571. An assessment of the small-crack effect for 2024-T3
R. C. McClung, 1991b, A simple model for fatigue crack aluminum alloy. In: ‘‘Small Fatigue Cracks,’’ eds. R. O.
growth near stress concentrations. J. Pres. Ves. Technol., Ritchie and J. Lankford, The Metallurgical Society,
113, 542–548. Warrendale, PA, pp. 427–452.
R. C. McClung, 1996, In: ‘‘Proceedings of Fatigue 96: The H. Nisitani, N. Kawagoishi and M. Goto, 1994, ‘‘Hand-
6th International Fatigue Congress, Berlin, Germany,’’ book of Fatigue Crack Propagation in Metallic Struc-
vol. 1, pp. 113–130. tures,’’ ed. A. Carpinteri, Elsevier, Amsterdam, The
R. C. McClung, 2000, Analysis of fatigue crack closure Netherlands, pp. 733–778.
during simulated threshold testing. In: ‘‘Fatigue Crack S. Pearson, 1975, Initiation of fatigue cracks in commercial
Growth Thresholds, Endurance Limits, and Design, aluminum alloys and the subsequent propagation of very
ASTM STP 1372,’’ eds. J. C. Newman and R. S. Piascik, short cracks. Eng. Fract. Mech., 7, 235–247.
ASTM, pp. 209–226. J. Petit, 1999, Influence of environment on small fatigue
R. C. McClung, G. G. Chell, D. A. Russell and G. E. crack growth. In: ‘‘Small Fatigue Cracks: Mechanics,
Orient, 1997, A practical methodology for elastic–plastic Mechanisms, and Applications,’’ eds. K. S. Ravichan-
fatigue crack growth. In: ‘‘Fatigue and Fracture dran, R. O. Ritchie and Y. Murakami, Elsevier, pp.
Mechanics: 27th Volume, ASTM STP 1296,’’ eds. R. S. 167–178.
Piascik, J. C. Newman and N. E. Dowling, ASTM, R. S. Piascik, 1999, The growth of small corrosion fatigue
pp. 317–337. cracks in alloy 7050. In: ‘‘Proceedings of FATIGUE ’99
R. C. McClung and D. L. Davidson, 1991, High resolution Conference Volume IV,’’ eds. X.-R. Wu and Z.-G.
numerical and experimental studies of fatigue cracks. Wang, pp. 2237–2242.
Eng. Fract. Mech., 39, 113–130. R. S. Piascik and S. A. Willard, 1994, The growth of small
R. C. McClung and D. L. Davidson, 1999, Near-tip and corrosion fatigue cracks in alloy 2024. Fatigue Fract.
remote characterization of plasticity-induced fatigue Eng. Mater. Struct., 17, 1247–1259.
crack closure. In: ‘‘Advances in Fatigue Crack Closure K. S. Ravichandran and J. M. Larsen, 1992, Behavior of
Measurement and Analysis: Second Volume, ASTM small and large fatigue cracks in Ti-24Al-11Nb—effects
STP 1343,’’ eds. R. C. McClung and J. C. Newman, Jr., of crack shape, microstructure and closure. In: ‘‘Pro-
ASTM, West Conshohocken, PA, pp. 106–127. ceedings of Fracture Mechanics: 22nd Symposium,
R. C. McClung and H. Sehitoglu, 1992, Closure and ASTM STP 1131,’’ eds. H. A. Ernst, A. Saxena and
growth of fatigue cracks at notches. J. Eng. Mater. D. L. McDowell, ASTM, Philadelphia, PA, vol. I,
Technol., 114, 1–7. pp. 727–748.
R. C. McClung, B. H. Thacker and S. Roy, 1991, Finite K. S. Ravichandran, R. O. Ritchie and Y. Murakami
element visualization of fatigue crack closure in plane (eds.), 1999, ‘‘Small Fatigue Cracks: mechanics, Me-
stress and plane strain. Int. J. Fract., 50, 27–49. chanisms, and Applications,’’ Elsevier, Oxford, UK.
D. L. McDowell, 1996, Basic issues in the mechanics of W. T. Riddell, R. S. Piascik, M. A. Sutton, W. Zhao, S. R.
high cycle metal fatigue. Int. J. Fract., 80, 103–145. McNeill and J. D. Helm, 1999, Determining fatigue
K. J. Miller, 1987, The behaviour of short fatigue cracks crack opening loads from near-crack-tip displacement
and their initiation: Part II. General summary. Fatigue measurements. In: ‘‘Advances in Fatigue Crack Closure
Fract. Eng. Mater. Struct., 10, 93–113. Measurement and Analysis, ASTM STP 1343,’’ 157–174.
K. J. Miller, 1993, Materials science perspective of metal R. O. Ritchie and J. Lankford (eds.), 1986, ‘‘Small Fatigue
fatigue resistance. Mater. Sci. Technol., 9, 453–462. Cracks,’’ Metallurgical Society, Warrendale, PA.
K. J. Miller, 1997, The three thresholds for fatigue crack R. O. Ritchie and J. O. Peters, 2001, Small fatigue cracks:
propagation. In: ‘‘Fatigue and Fracture Mechanics: 27th Mechanics, mechanisms, and engineering applications.
Volume, ASTM STP 1296,’’ eds. R. S. Piascik, J. C. Mater. Trans., 42, 58–67.
Newman and N. E. Dowling, ASTM, pp. 267–286. R. O. Ritchie and S. Suresh, 1982, Some considerations on
K. J. Miller and E. R. de los Rios (eds.), 1986, ‘‘The fatigue crack closure at near-threshold stress intensities
Behavior of Short Fatigue Cracks,’’ European Group on due to fracture surface morphology. Metall. Trans. A,
Fracture, Publication No. 1. 13, 937–940.
K. J. Miller and E. R. de los Rios (eds.), 1992, ‘‘Short R. O. Ritchie and W. Yu, 1986, Short crack effects in
Fatigue Cracks,’’ European Structural Integrity Society, fatigue—a consequence of crack tip shielding. In: ‘‘Small
Publication 13, Mechanical Engineering Publications, Fatigue Cracks,’’ eds. R. O. Ritchie and J. Lankford,
London. TMS, Warrendale, PA, pp. 167–189.
164 Small Fatigue Cracks
H. Sehitoglu and W. Sun, 1991, Modeling of plane strain pearlitic steel. Fatigue Fract. Eng. Mater. Struct., 11,
fatigue crack closure. ASME, J. Eng. Mater. Technol., 331–342.
113, 31–40. T. H. Topper and M. H. El Haddad, 1981, Fatigue strength
C. F. Shih and J. W. Hutchinson, 1976, Fully plastic prediction of notches based on fracture mechanics. In:
solutions and large scale yielding estimates for plane ‘‘Fatigue Thresholds: Fundamentals and Engineering
stress crack problems. AMSE Trans., J. Eng. Mater. Applications,’’ eds. J. Bäcklund, A. F. Blom and C. J.
Technol., 98, 289–295. Beevers, Engineering Materials Advisory Services, War-
S. Suresh, 1985, Fatigue crack deflection and fracture ley, UK, pp. 777–798.
surface contact: micromechanical models. Metall. Trans. A. Turnbull, 1982, A theoretical evaluation of the influence
A, 16A, 249–260. of mechanical variables on the concentration of oxygen
S. Suresh, 1998, ‘‘Fatigue of Materials,’’ 2nd ed., Cam- in a corrosion fatigue crack. Corr. Sci, 22, 877–893.
bridge University Press, Cambridge. A. Ueno and H. Kishimoto, 2001, ‘‘Proceedings of the
S. Suresh and R. O. Ritchie, 1982, A geometric model for International Conference on Fatigue in the Very High
fatigue crack closure induced by fracture surface rough- Cycle Regime,’’ eds. S. Stanzl-Tschegg and H. Mayer,
ness. Metall. Trans. A, 13A, 1627–1631. University of Agricultural Sciences, Vienna, Austria, pp.
S. Suresh and R. O. Ritchie, 1984, The propaga- 111–117.
tion of short fatigue cracks. Int. Metals. Rev., 29, W. V. Vaidya, 1992, Fatigue threshold regime of a low
445–476. alloy ferritic steel under closure-free testing conditions:
M. H. Swain, 1992, Monitoring small-crack growth by the Part II. Hysteresis in near-threshold fatigue crack
replication method. In: ‘‘Small-Crack Test Methods, propagation: An experimental assessment. J. Test. Eval.,
ASTM STP 1149,’’ ASTM, pp. 34–56. 20, 168–179.
K. Tanaka, 1983, The cyclic J-integral as a criterion for L. Wagner, J. K. Gregory, A. Gysler and G. Luetjering,
fatigue crack growth. Int. J. Fract., 22, 91–104. 1986, Propagation behavior of short cracks in a Ti-8.6Al
K. Tanaka, 1985, Short-crack fracture mechanics in fatigue alloy. In: ‘‘Small Fatigue Cracks,’’ eds. R. O. Ritchie and
conditions. In: ‘‘Current Research on Fatigue Cracks,’’ J. Lankford, The Metallurgical Society, pp. 117–128.
Materials Research Series, eds. T. Tanaka, M. Jono and C. H. Wang, 1996, Effect of stress ratio on short fatigue
K. Komai, The Society of Materials Science, Japan, vol. crack growth. J. Eng. Mater. Technol., 118, 362–366.
1, pp. 79–100. C. H. Wang and J. W. Hutchinson, 2001, Interactions of
K. Tanaka, Y. Nakai and M. Yamashita, 1981, Fatigue fatigue cracks with elastic obstacles. Int. J. Fract., 109,
growth threshold of small cracks. Int. J. Fract., 17, 263–283.
519–532. A. J. Wilkinson, S. G. Roberts and P. B. Hirsch, 1998,
D. Taylor and O. M. Clancy, 1991, The fatigue perfor- Modelling the threshold conditions for propagation of
mance of machined surfaces. Fatigue Fract. Eng. Mater. stage 1 fatigue cracks. Acta Mater., 46, 379–390.
Struct., 14, 329–336. D. R. Williams, D. L. Davidson and J. Lankford, 1980,
K. Tokaji, T. Ogawa and S. Osako, 1988, The growth of Fatigue crack tip plastic strains by the stereoimaging
microstructurally small fatigue cracks in a ferritic- technique. Exp. Mech., 20, 134–139.

Copyright 2003, Elsevier Ltd. All Rights Reserved. Comprehensive Structural Integrity
No part of this publication may be reproduced, stored in any retrieval system or ISBN (set): 0-08-043749-4
transmitted in any form or by any means: electronic, electrostatic, magnetic tape,
Volume 4; (ISBN: 0-08-044155-6); pp. 129–164
mechanical, photocopying, recording or otherwise, without permission in writing
from the publishers.

View publication stats