J. Phys. Chem. Solidv Vol. 48, No. II, pp. 103~-1074, 1987 OOZZ-3697187 13.00 + 0.

00
Printed in Great Britain. 0 1987 Pergamon Journals Ltd.

DUCTILE FRACTURE
W. M. GARRISON JR
Department of Metallurgical Engineering and Materials Science, Carnegie-Mellon University,
Pittsburgh, PA 15213, U.S.A.

and
N. R. MOODY
Sandia National Laboratories, Livermore, CA 94550, U.S.A.

Abstract-The purpose of this paper is to review ductile fracture of metallic systems. The goal has been
to present both mechanics and microstructural aspects of ductile fracture and to emphasize the interplay
between the approaches. The paper considers first experimental studies and efforts to model the three
stages of ductile fracture: void nucleation, void growth and void coalescence. To make explicit the
necessity of incorporating both mechanics and microstructure in studies of ductile fracture, the fracture
behavior of three different mechanical tests are discussed. These are axisymmetric tensile specimens with
and without notches, plane strain tensile specimens and specimens containing sharp notches used to
measure fracture initiation toughness.
Keywords: Ductile fracture, void nucleation, void growth, void coalescence, tensile ductility, plane strain
tensile, notched tensile, fracture toughness, modeling fracture initiation.

1. INTRODUCTION Puttick found that void initiation occurred by frac-

ture of non-metallic inclusions and by the decohesion
Ductile fracture is a mode of material failure of the inclusion-matrix interface. Crussard ef al.
in which voids, either already existing within the observed numerous cup-like indentations in electron
material or nucleated during deformation, grow until micrographs of carbon replicas of fracture surfaces
they link together, or coalesce, to form a continuous [A. They speculated that voids had produced the
fracture path. Other terms used for this fracture mode indentations which were found on the fracture sur-
are fibrous fracture, dimpled rupture or microvoid faces. The connection between second phase par-
coalescence, As Fig. 1 shows, the fracture surfaces of ticles, void nucleation and the fracture surface has
materials failing by ductile fracture are covered by been confirmed repeatedly in subsequent experiments
segments of the voids which coalesced to produce the employing optical metallography of sectioned speci-
fracture path. These void segments were termed mens, transmission electron microscopy of thin sec-
dimples by Plateau et al. [I]. The dimples in Fig. 1 tions from the vicinity of fracture surfaces, scanning
vary widely in size and some have obviously been fractography and fracture surface replication. An
nucleated at second-phase particles while the nuclei emphasis on the role of inclusions and particles in
for others are not as readily apparent. initiating ductile fracture is justified when one consid-
The most significant step toward understanding the ers the range of deformation states and the variety of
microstructural origins of ductile fracture was the materials in which it has been demonstrated: intemal-
discovery that voids nucleate at inclusions and ly oxidized copper alloys [8], various maraging steels
second-phase particles. This was not recognized until [9, lo], quenched and tempered high-strength steels
more than a century after the early work of Henry, [9], low-strength steels [ 1l] and aluminum alloys [ 121.
who in 1855 suggested that ductile metals fracture by However, while second-phase particles act as void
internal necking when stretched [2]. In 1927, Ludwik nuclei in many systems, there are systems where voids
sectioned a necked but unbroken tensile specimen of can be nucleated by other mechanisms. These include
aluminum and found that fracture had initiated as an formation of voids at interfaces as in a-b titanium
internal crack which had apparently started in the alloys [13-151, and at slipband intersections [l&20].
center of the neck [3]. Tipper suggested in 1948 that In addition, void nucleation by vacancy condensation
the tensile fracture of ductile metals proceeded by the has been suggested [21]. While there are other frac-
growth of voids, initiated by matrix decohesion from ture types which have a ductile character such as
nonmetallic inclusions [4]. Unequivocal supporting rupture [6,22], our interest here is in ductile fracture
evidence for this hypothesis was presented in 1959 by arising from the nucleation, growth and coalescence
Puttick [S] and was later corroborated by the work of of voids. Furthermore, the focus will be on metallic
Rogers [6]. On examining sectioned tensile specimens systems deformed at temperatures less than about
of copper and iron under the optical microscope one-third their melting temperature; therefore void
1035
1036 W. M. GARRISONJR and N. R. MOODY

l.A . scanning electron micrograph of the fracture surface of a fracture toughness specimen illustra ting
coallescence. The material is a low alloy steel and the inclusions visible on the fracture su are
manganese sulfides.

formation during creep and irradiation will not be and coalescence-combine to determine these prop-
considered. erties. For ductile fracture, the engineering properties
The locally high levels of plastic deformation asso- are determined by the interaction of stress and strain
ciated with the voids seen in Fig. 1 suggest that fields with the microstructure of the material. Because
ductile fracture requires high levels of energy absorp- ductile fracture involves both mechanics and micro-
tion. However, this is not always the case; various structure a diversity of approaches have been taken
properties, such as tensile ductility or toughness, to its study. One extreme is to use a continuum
associated with energy absorbed during the fracture mechanics approach to model ductile fracture. The
process, can be quite low. For example, even low other is the classic metallurgical approach of varying
strength materials exhibit very low tensile ductilities microstructural features and correlating the changes
if the volume fraction of a void nucleating second in properties with those changes. However, it is clear
phase is sufficiently high. In addition, the fracture that any solution to the problem of ductile fracture
initiation toughness of higher strength materials con- involves both mechanics and microstructural aspects.
taining reasonably low volume fractions of void It is hoped that the interplay between the mechanics
nucleating particles can be quite low even though and microstructural oriented approaches will be ap-
their tensile ductilities might be reasonably high. parent throughout this paper. To reinforce this theme
Moreover, materials which exhibit similar tensile the fracture behavior associated with three different
ductilities may have very different fracture initiation types of mechanical tests will be discussed. These will
toughnesses. Thus, the energy associated with ductile be axisymmetric tensile specimens of different geom-
fracture can vary enormously depending on the mate- etries, plane strain tensile specimens and specimens
rial and the nature of the mechanical test. containing sharp notches used to measure fracture
Over the past 20 yr there has been a large effort to initiation toughness.
improve the properties of engineering alloys failing This review coupled with others will provide the
by ductile fracture and to understand how the three interested reader with a good overview of our present
stages of ductile fracture-void nucleation, growth, understanding of ductile fracture and of current
 Ductile fracture 1037

research interests. Review articles on ductile fracture follows reference will be made to a nucleation strain.
include those by Van Stone et al. [23], Rosenfield [24] However, experimental evidence suggests that even
and Hahn et al. [25]. A more mechanics oriented for a fixed matrix material a nucleation strain is
review is presented by Howard and Willoughby [26]. hardly unique and will apparently vary within a
Schwalbe’s article is an excellent source for micro- given distribution of particles and with the stress
structural effects [27]. state.
Most experimental evidence suggests that void
nucleation occurs first at the larger particles within
II. STAGES OF THE FRACTURE PROCESS
the particle distribution. This has been observed in
For purposes of discussion the process of ductile copper alloys containing SiO, [8], for constituent
fracture can be divided into three stages: (1) void particles in aluminum alloys [28], for various types
nucleation, (2) void growth, and (3) void co- of inclusions in steels [9,29] in T.D. nickel [30] and
alescence. However, it is difficult to define where one for spheroidized carbides in steels [31]. Thus for
stage ends and the next begins during fracture of given test conditions the larger particles in a given
materials. Even if the material contains only one type distribution are expected to nucleate at lower strains.
of second-phase particle, void initiation will not In addition, replacing one particle distribution
occur simultaneously at all of the particles. Typically with another characterized by a smaller average
voids nucleate at the larger particles first. As the size should increase the strain required for void
fracture process continues, voids nucleated at the nucleation.
larger particles grow while voids are nucleated at the The strain at which void nucleation occurs is
smaller particles. The process becomes even more apparently very dependent on the stress state. Cox
complicated for materials that contain several types and Low have investigated the influence of stress
of second-phase particles. In these materials, voids state on void nucleation at MnS particles in 4340
will often nucleate first at a particular particle type, steel using smooth axisymmetric tensile specimens
and later in the fracture process at another set of and notched tensile specimens [9]. Figure 2 shows
particles. It is often the case that some of the particles the fraction of MnS inclusions with voids as a
may never initiate voids at all, but may affect the function of true plastic strain for the 4340 steel. In
fracture process only indirectly by influencing the the notched specimens, with a higher triaxial stress,
flow characteristics of the matrix. An example of such voids nucleated at lower strains than in the smooth
a complex system is the ultra high strength steel 4340. tensile specimens. Void nucleation in the high purity
This steel contains inclusions, often small undissolved steel, with its smaller average inclusion size, occurred
carbides, and, after tempering at 2OO”C,fine carbides at higher strains than in the commercial purity 4340.
precipitated on tempering. Voids initiate first at the Other evidence for the importance of stress state
inclusions, later in the fracture process at the un- on void nucleation come from studies of the effects
dissolved carbides and perhaps not at all at the of hydrostatic pressure. French and Weinrich [32]
precipitated carbides. While a sharp division between studied void nucleation at cementite particles in a
initiation, growth and coalescence might be some- spheroidized 1050 carbon steel using smooth tensile
what misleading, the following discussion of the specimens uniaxially deformed under superimposed
fracture process will consider the three stages hydrostatic pressure. They found that increasing the
separately. pressure from 0.1 to 300 MPa significantly retarded
void nucleation, but there was little change in the
1I.A Void initiation
II.A.1 Experimental observations. Void initiation F _F
100 -- , I
can occur by decohesion of the particle-matrix 2 . 0 SMOOTH SPECIMENS
interface and by fracture of the particle. For a given ? F A A NOTCHED SPECIMENS

type of particle, particle fracture is favored as the size SOLID SYMBOLS REPRESENT_
= 75

VJ
COMMERCIAL PURITY
of the particle increases although this is influenced by 5 OPEN SYMBOLS REPRESENT
the shape of the particle, its orientation (if not 4 HIGH PURITY

equiaxed), and by the strength of the particle-matrix g 50

3 F
bond.
Particle-matrix decohesion does not occur simul- z
I- 25
taneously at all points along the particle-matrix
5
interface even for spherical particles. In tensile speci- t?
mens, voids initiate at the particle-matrix interface k! I I I I I
00 0.10 0.20 0.30 0.40 050 060
along the axis of applied stress at the position of
TRUE STRAIN
maximum tensile stress. A number of factors
influence the conditions required for this nucleation Fig. 2. A plot of per cent inclusions with voids as a function
of the true strain for commercial and high purity heats of
event to occur. These include particle size, stress 4340 steel. The sulfur levels and inclusion volume fractions
state, interfacial cohesion, strength of the matrix associated with the commercial and high purity heats were
material, and particle volume fraction. In what 0.013 wt % and 0.004 wt % and 0.0014 and 0.0006 [9].
1038 W. M. GARRISONJR and N. R. MCQDY

1 I I I I I I -
IOO- ’ 18 Ni.300 GRADE MARAGING STEEL
YIELD STRENGTH (KS11
0 (I 04). Solution Treat.
lOoF- 50 hrs.
80 I- A (225),800’F- 3 hrs.
0 (2891, 900°F- 3 hrs.
A (2481, lOOO*F-3hrs.
n (3001, 800°F- 100 hrs.

F indicates a fractured specimen

I I I I I I I
0.4 0.8 0.8 I.0 1.2 I.4 1.8
PLASTIC STRAIN

Fig. 3. The per cent of Ti,S inclusions associated with voids as a function of strain for a maraging steel
aged to several different strength levels [29].

strain required for void nucleation as the hydrostatic Al,09 particles. They employed an expression devel-
pressure was increased from 300 to 600 MPa. This oped by Ashby [37] which relates the work of
result suggests that when large stresses are required adhesion (bond strength) and the strain at which
for void nucleation, a simple normal stress criterion decohesion occurs:
may not be adequate to describe the onset of void
nucleation, but rather a criterion based on both
stress and strain may be required [23]. Additional
support for this conclusion comes indirectly from the
work by Coffin and Rogers on the influence of where G is shear modulus, b is Burgers vector, d is
hydrostatic pressure on strip and wire drawing [33]. particle diameter, t is critical strain for separation,
On the other hand, Chen has presented evidence and k is mean particle spacing.
which suggests hydrostatic tension will influence only Specimens prepared by powder metallurgy tech-
void growth, not void formation [34]. niques were cold rolled to 75% reduction, then held
Psioda’s work on a 300 grade maraging steel 30min at 1100°C to equilibrate cavities. Contact
suggests that matrix strength, or perhaps more angles at voids were measured in a high-voltage
correctly the flow characteristics of the matrix, can electron microscope. The work of adhesion was
influence void nucleation [29]. By varying the aging determined from
conditions, the yield strength of the steel could be
varied from 104 to 300 ksi. The percent of T&S w,, = yu(l -cos O),
inclusions in the steel which were associated with
voids is plotted in Fig. 3 as a function of strain for where yu is the surface energy of the matrix (as-
the different strength levels. These data suggest that sumed) and 0 is the contact angle between matrix
void nucleation occurs at lower strains in higher and inclusion. In this and a subsequent study [38] it
strength materials. was found that MO or Cr additions to the iron matrix
It is also believed that the nucleation strain can be produced large improvements in ductility. Additions
influenced by altering the strength of the cohesive of Cr to an F&O% Ni base increased W,, to such
bond between the particle and the matrix. The an extent that no voids were found at A&O3 particles
addition of zinc to copper containing particles of after 75% cold work. A similar effect of Cr in
SiOz was found to reduce the nucleation strain and increasing adhesion between ThO, and Ni was noted
this was attributed to a reduction in the strength of by Franklin et al. [39]. Chromium also increases
the bond between the copper and the SiO, particles significantly the cohesion between ferrite and MnS
[35]. There have been only a few attempts to increase [40]. In a third investigation Easterling et al. strained
the bond strength between metal matrices and inclu- foils in the high-voltage microscope and related flow
sions. Fischmeister et al. [36] examined the effects of stress and ductility to 0 [41]. When 0 was less than
alloying on the bond strength between iron and 10” the material was brittle; high values of 0
 Ductile fracture 1039

produced ductile fractures, Sundstrom came to the interface between plate-like carbides and ferrite will
same conclusion on the basis of calculations per- result in decohesion whereas in the absence of such
formed on a plane strain model of Al,O, inclusions impurity segregation particle cracking was observed.
in y Fe [42]. For a large work of adhesion, 2 J rnm2, Void nucleation sites other than second phase
no decohesion occurred. For intermediate values, particles have been observed in many alloy systems.
decohesion was followed by void growth and for a Thompson and Williams [54] observed that relatively
very small work of adhesion, 0.02-0.2 Jmm2, brittle featureless dimples formed in single-phase a Ti-Al
fracture occurred. alloys. There were no void nucleating particles evi-
There is evidence that segregation of impurity dent which led to the suggestion that twin bound-
atoms to the particle-matrix interface can reduce the aries, grain boundaries, vacancy clusters or slip band
strains required for void nucleation. Rellick and intersections may have acted as nucleation sites.
McMahon [43] and Smith and Low [44] have shown Subsequently, Van Stone et al. [55,56] showed that
that trace elements such as phosphorous, antimony void nucleation occurred at twin boundaries in a
and tin segregate to ferrite-carbide interfaces during Ti-5Al-2.5 Sn alloy tested at room temperature and
temper embrittlement of steels. King and Knott have at intense slip band intersections when tested at
shown that temper embrittlement heat-treatment of cryogenic temperatures.
a low alloy steel results in a large reduction in In OFHC copper, Nieh and Nix [57] observed that
toughness even when failure is by dimpled rupture failure occurred by grain boundary void formation
[45]. Hippseley and Druce have investigated the during high-temperature creep tests. Both the grain
influence of temper embrittlement heat treatments on boundary void size and slip band spacing varied
the ductile fracture of the low alloy steel 300M inversely with applied stress. The one-to-one corre-
initially tempered at 650°C for 10 hr [46]. They found spondence between grain boundary void size and slip
that the temper embrittlement treatment significantly band spacing suggested that void nucleation resulted
reduced the strain required for initiation of voids at from the intersection of slip bands with the grain
the carbides. Using Auger techniques they observed boundaries.
strong segregation of phosphorus to the A diversity of void nucleation sites has been ob-
ferrite-carbide interface in the temper embrittled served in a + /I titanium alloys. In aged a + /? alloys,
material. failure often nucleates at cctitanium particles [13-l 51.
Voids can also be initiated by the fracture of The voids which form at these particles result from
second-phase particles. Systems in which fracture of the intersection of planar slip bands in the a particle
second-phase particles have been observed include with the particle-matrix boundary. These voids sub-
constituent particles in aluminium alloys [47], sequently grow along the a boundaries until they
cementite in steels [31], as well as sulfides, and impinge.
Ti(CN) and other inclusions in steels [9,29,48-501. The propensity for voids to nucleate at and grow
Gurland demonstrated that void nucleation occurred along grain boundaries permits a wide range of
by cracking of spheroidized cementite in a 1.05% fracture toughness values to be obtained from ti-
carbon steel under tensile, compressive, and torsional tanium alloys by changing the morphology and dis-
loading [31]. Cracking occurred perpendicular to the tribution of the void nucleating a phase through
direction of the maximum principal strain under all heat-treatment. This is illustrated by the study of
three loading conditions which coincided with the Galda et al. [58], of grain size effects on fracture
direction of the highest tensile stress and the largest toughness of Ti-6A1-4V. They observed one dimple
particles fractured at the lowest strains. Particle along each grain boundary in small grain size samples
fractures are often constrained to the second-phase and more than one dimple along a grain boundary in
particles, but in some cases, cracks are observed to large grain size samples. As a consequence, fracture
run continuously through the particle some distance toughness increased significantly with increased grain
into the matrix. Factors which influence the tendency size in small grain size samples, then exhibited a weak
for void formation by particle cracking include dependence on grain size in large grain size samples.
particle size [31,47], strength of the matrix [29], The transition between a strong and weak grain size
particle shape [51,52], and the strength of the effect appears to correlate with the onset of multiple
matrix-particle bond, which can be altered by segre- void formation along a grain boundary. The rela-
gation of impurity elements. Gangulee and Gurland tively high fracture toughness of titanium alloys
[47] found for a given particle volume fraction and heat-treated to obtain a Widmanstatten structure is
level of applied stress that the fraction of particles derived from fracture along the elongated a platelets
which fractured scaled with the square root of the [59]. Fracture along elongated a platelets results in
average particle diameter. In general, they found that pronounced crack plane deviations (crack path tor-
the fraction of particles which fractured scaled with tuosity). These crack plane deviations require the
~~di’~/f’~ where CTis the true applied stress, d is the crack to travel over a greater distance and require
average particle diameter andfis the particle volume more energy for propagation than compared to crack
fraction (Fig. 4). Restaino and McMahon (531 have growth during flat fracture [14, 15, 591. Void nu-
pointed out that segregation of impurities to the cleation at phase boundaries has been reported in

P.C.S.
4*/1
I--F
1040 W. M. GARRISONJR and N. R. MOODY

SYMBOL 0 0 a 0 0 0 0 0 e l 0
% Si 2.6 6.1 6.1 6.1 6.1 6.1 6.1 6.1 6.1 9.613.2

d 9.4 2.6 3.6 5.1 6.1 6.6 9.6 II.2 12.6 6.9 6.9

STRESS - STRUCTURE PARAMETER (X io3 psi - MICRONS I’*)

Fig. 4. A plot of the fraction of particles which have fractured as function of ~d”‘/f”~, where D is the
true applied stress, d is the particle diameter, and f is the particle volume fraction. The material is an
alum&m alloy containing various amounts of silicon and undissolved silicon particles [47]. (Reprinted
with permission from Trans. metall. Sot. 239, 271 (1967), a publication of the Metallurgical Society,
Warrendale, PA 15086, U.S.A.)

other two-phase alloys such as, for example, dual- that voids formed at slip band intersections or dis-
phase steels [198]. location cell walls during straining.
Voids have also been found to nucleate at sites In this section, experimental studies of void nu-
other than second-phase particles in age hardened cleation at second-phase particles have been sum-
metastable /l titanium alloys. These alloys deform marized. These studies suggest that void nucleation at
inhomogeneously by coplanar slip. As shown by low strains will be favored by large particle size, high
Gysler et al. [16], voids nucleate at slip band inter- interfacial stresses, high matrix strength, weak bond-
sections with grain boundaries and other slip bands ing between the matrix and particle, and high particle
resulting in ductile intergranular and faceted failures. volume fraction. In addition, it has been pointed out
Unless studied in detail, the role of void nucleation that void nucleation does not always occur at second-
at slip band intersections may be overlooked. As phase particles. However, it is emphasized that non-
demonstrated by Williams et al. [17] in a study of particle related void nucleation is not as prevalent
metastable jl titanium alloys, failure appeared to and important as particle related void nucleation.
occur by cleavage at low magnification. However, With the exception of titanium alloys, ductile fracture
examination of TEM replicas at high magnification of engineering alloys usually begins with the initiation
revealed that the “cleavage” facets were covered by of voids at second-phase particles.
parallel rows of small, equiaxed dimples. A similar II.A.2 Modelling void nucleation at particles. The
approach applied to hydrogen-induced faceted frac- experimental results presented in the previous
ture of IN903 [60-62] showed that microvoids also section show that the initiation of voids at second-
formed at slip band intersections. Single crystal phase particles can occur by two mechanisms:
studies [18, 191 found that failure can initiate by (1) decohesion of the particle-matrix interface which
microvoid nucleation at the intersection of intense begins at particular points on the interface or (2) by
shear bands. The use of high voltage electron micro- fracture of the particle. While void nucleation is often
scopy during deformation of 304 stainless steel [20] quantified by a nucleation strain, the nucleation
revealed that failure in this polycrystalline alloy can strain does not appear to be an absolute quantity for
also initiate at slip band intersections. Similar studies a given type of second-phase particle. Factors which
by Robertson and Birnbaum (631 of nickel showed influence the nucleation strain appear to be the
 Ductile fracture 1041

strength of the particle-matrix interface, the stress be regarded as the spacing of active slip bands. If void
state, and the flow strength of the matrix. For nucleation is reached at a critical stress u, then this
nucleation by particle-matrix decohesion, there is stress will be achieved at lower strains for larger
substantial evidence which suggests nucleation is particles. Rosenfield [24] has noted that the model
easier at larger particles, although this point is de- does not incorporate other possible effects such as
bated [23,64,65] and Rosenfield [24] has suggested stress state or the stress intensification at the interface
the evidence for this size effect is not unequivocal. In discussed by Gurland and Plateau [66]. Rosentield
addition, there appears to be a second size effect-a [24] suggested the tensile stress at the interface could
transition from void nucleation by particle-matrix result from summing the contributions of several
decohesion to void nucleation by particle fracture. effects.
Both continuum and dislocation approaches have
been taken to modelling void nucleation. Some of a,=a,+qa +a,,
these modelling efforts will be now summarized in a
chronological order. In all of these models the where bT is the stress derived by Ashby, qcr is the
second-phase particle is assumed to be much harder intensification of the stress suggested by Gurland and
than the matrix and to deform only elastically. Plateau and C~ would be the hydrostatic pressure or
Gurland and Plateau appear to be the first to tension. We note here that this list of contributions
model void initiation and their interest was in deter- to the interfacial stress has not been exhausted; other
mining the conditions required for void nucleation by contributions are possible. For example, there could
particle fracture [66]. They suggested a necessary be stresses at the interface generated during heat-
condition for particle fracture was that the strain treatment if the particle and the matrix have different
energy released by particle fracture be greater than coefficients of expansion [67-761. To the best of our
the surface energy of the crack surfaces. Assuming knowledge, such stresses have not been studied with
the stress field of the particle and of the crack had respect to void nucleation.
dimensions of the same magnitude as the particle, Tanaka et al. determined the critical strain for void
they gave an approximate expression of the applied nucleation by decohesion of the particle-matrix inter-
stress, u, required for particle fracture, face for a system consisting of an elastically defor-
ming spherical inclusion and a plastically deforming
1 Ey I’* matrix for the case of uniaxial tension [77]. Their
fJ=-- ,
q0a calculations suggest that for particle diameters
greater than about 500 A, the energy criterion is
where q is a stress concentration factor, a is the satisfied at very low strains. However, it is not clear
particle diameter, E is a weighted average of Young’s whether a purely continuum model can be used at
moduli of the particle and the matrix and y is the such small particle sizes. Tanaka et al. also con-
specific surface energy of the crack within the par- sidered the criterion that void nucleation will occur
ticle. This approach predicts that the stress required when the stress at the interface exceeds a critical level,
for particle fracture will increase with decreasing assumed to be the theoretical strength of the particle
particle size and with the bulk fracture toughness of or matrix, whichever is lower. These calculations
the material constituting the second-phase particle. suggest a critical matrix strain for decohesion which,
Ashby appears to have been the first to propose a unlike the results of Ashby, is independent of particle
model for void nucleation by particle-matrix deco- size.
hesion [37]. Ashby suggested that the most plausible Brown and Stobbs were the next to model void
approach was that particle matrix decohesion would nucleation by decohesion of the particle-matrix inter-
occur when a critical stress was achieved at the face [78]. Like Ashby, their model was derived in the
interface. Ashby, considering a particle engulfed by a context of a study of work hardening. Brown and
slip band, suggested that the loops encircling the Stobbs suggested that void nucleation would not
particle could result, via secondary slip, in an array occur unless the elastic energy stored within the
of prismatic loops. The vacancy loops would result in particle, and presumably lost when void nucleation
tensile stresses on the particle-matrix interface given occurred, was greater than the surface energy associ-
by ated with the formation of the void. They took the
elastic energy stored in the particle as
CR
a,-jg,

where c is the plastic strain, R is the radius of the where up*is a measure of the incompatibility between
particle, b is the matrix Burger’s vector and 1 the matrix and particle deformation, p* is the shear
the length of the array of prismatic loops. Ashby modulus of the particle and r,, is the particle radius.
suggested that in considering an array of particles, They took the increase in energy of the system due to
1 could be equated with half the particle planar the formation of the new surfaces of the void and the
spacing, or for widely separated particles, that I could elimination of the matrix-particle interface as 4ny&
1042 W. M. GARRISXN
JR and N. R. MOODY

where y is yM+ yp - ypM.Thus for void nucleation to negative pressure or hydrostatic component of the
occur stress. As the flow stress in tension is very close to the
mean value of the limits, they suggest

For the copper-silica system they investigated, and

making rather specific assumptions as to nature of where Y(E,) is the flow stress in the region of the
the operating dislocation mechanisms, they found second-phase particle at the average local plastic
strain within the region if the particle were not
+(X->‘p(!!>‘; present. There is no effect of particle size in this
approach and it must be corrected if the second-phase
particles are sufficiently closely spaced that their
and estimated the minimum strain required for void stress fields overlap. They further argue that a distinct
size effect appears only when the particle volume
nucleation as
fraction is sufficiently high for the particle to interact
and that local variations of particle volume fraction
371~ u2r0 ‘I4 exist. This makes it possible for some larger than
6:
>K&EZ ( > ’ average particles to be nearest neighbors at a spacing
equal to or smaller than the average spacing. They
where a is a constant governing the strength of forest construct a formalism by which one can correct for
hardening. This form predicts that the nucleation this effect of particle-particle interaction.
strain will increase with the particle size. Argon and co-workers then considered three types
Sundstrom’s contribution to modeling of void nu- of void initiating particles: spheroidized cementite in
cleation, although restricted solely to elastic defor- a 1045 steel, chromium particles in a copper matrix
mation within both the matrix and the inclusion, and Ti(C,N) in what apparently was an overaged
introduced two new modeling concepts [42]. He was maraging steel. They studied void nucleation in these
the first to incorporate the idea that decohesion systems using notched as well as axisymmetric tensile
would begin at the poles of the particle parallel to the specimens. For each specimen they determined the
tensile direction and then propagate along the inter- distance along the tensile axis below the fracture
face. In addition, he postulated that the hemi- surface where no void nucleation was observed. From
spherical cap thus formed must extend to some their calculations of strain and stress distributions in
critical size before it would be stable. Both of these such tensile specimens they were able to assign to this
concepts are incorporated in the later and much more point a nucleation strain, a flow stress Y(cg) and a
general theory of Fisher and Gurland 183,841. value for cr. From this they were able to calculate the
For the purely elastic deformation considered, critical stress required for decohesion of the
Sundstrom’s results are expressed in terms of a particle-matrix interface. They found that voids were
critical stress required for a stable crack to form at more easily nucleated at larger cementite particles
the particle-matrix interface. This critical stress in- than at smaller ones; this was explained in terms of
creases with decreasing particle size and as the work the particle-particle interactions. Their data sug-
of adhesion increases. While this work of adhesion is gested no such influence of particle size in the Cu-Ct
not formally defined it is assumed to be a constant for system where the volume fraction of chromium par-
a particular system and to be independent of particle ticles was small (S- 0.0059).
size. It has units of energy/unit area and presumably Goods and Brown [82] present in their overview of
it is equal to yw + yp - yup used in previous models. void initiation a new model based on the previous
Perhaps the most influential work done on void work of Brown and Stobbs. They assume that the
initiation was that reported by Argon et al. condition for void nucleation is
[64,65,79,80]. In their modeling they suggest for
larger (diameter>> 100 A) particles that voids will
initiate when the radial stress at the particle-matrix
interface exceeds some critical level. They then argue,
based on an hypothesis of Rhee and McClintock [81], where aH is the hydrostatic pressure, or is the flow
that the radial interfacial stress (r~,~)will be bounded stress (Y(Z,) in the terminology of Argon and co-
as workers), g1 is the interfacial stress required for
particle-matrix decohesion and 6, is a local stress
;k so,,-a,s2k, at the interface resulting from dislocation storage.
Writing the local stress as

where the lower bound is for a non-hardening ma-

terial and the upper bound is for a linear hardening
material. The flow stress in shear is k and uT is the
 Ductile fracture 1043

and assuming that the void initiating particles control the models in how they treat the energy criterion.
the flow stress according to There is no agreement between the models as to how
nucleation strain will depend on size.

1I.B. Void growth

After nucleation the voids will expand to a volume
where c,, is Orowan stress, they derive the following and shape determined by material properties and test
expression for the nucleation strain: conditions. Plastic deformation is apparently re-
quired for void growth. Indeed, void walls sometimes
exhibit wavy markings, which are the result of wavy
or serpentine glide on slip planes which intersect the
growing void. While dislocation motion is believed to
This expression predicts that the nucleation strain be responsible for void growth, few dislocation
will increase with particle size. While data are models for the growth of voids have been developed.
presented to support this approach many of the data Ashby, in his article on the work hardening of
are from studies of spheroidized cementite in steels. dispersion-hardened crystals [371, has proposed a
While it is true that in the materials studied the dislocation based void growth model. According to
reported nucleation strains appear to increase with this model, once the void is nucleated, the ratio of
increasing carbide size, the yield strengths also de- AV, the increase in volume, to the volume V of the
creased. Decreasing yield strength is believed to result void nucleating particle will scale as
in higher nucleation strains.
The most recent attempt to model void nucleation AV a
-w-
is due to Fisher and Gurland [83,84]. While the v 2’
model was developed to analyze results of experi-
mental studies in spheroidized steels it appears to be where a is the shear strain on the primary slip band.
very general. They adopted two criteria for void As noted by Rosentield [24], this model predicts
initiation. The first is that the radial stress at the growth rates which are much too small. In addition,
particle-matrix interface must exceed some critical the model is based only on the dislocation mech-
stress and the second that the energy change associ- anisms which cause high localized stresses around
ated with void nucleation must be sufficient to pro- particles and does not suggest how these dislocations
vide a stable configuration. While many of the results provide for void growth. McLean has suggested a
are obtained through numerical calculations they find dislocation model for void growth in which edge
that void nucleation is favored by a large particle size, dislocations, moving towards the nucleated void,
high flow stress, high stress triaxiality and a large merge with the void resulting in its enlargement;
plastic strain. This size effect emerges independent however, quantitative dislocation based descriptions
of particle interactions, Furthermore, unlike earlier of void growth have not been developed [85].
energy balance requirements for void nucleation, Modelling void enlargement has proven, and
their energy criterion is not generally satisfied even continues to be, an active area within continuum
for large deformations and reasonably large particles. mechanics. The pioneer in the continuum approach
II.A.3 Void nucleation: summary. Void nucleation is McClintock [86,87]. He developed equations de-
can occur at particles by decohesion of the scribing growth of voids of different geometries
particle-matrix interface or particle fracture. Experi- under different conditions.
mental studies suggest that in the case of McClintock’s basic approach was to obtain solu-
particle-matrix decohesion the nucleation strain will tions or approximations to the behavior of linear-
decrease as the strength of the particle-matrix bond hardening materials and of nonhardening plastic
decreases, as the particle size increases, as the stress materials. He then postulated that the behavior of
triaxiality decreases, as the flow strength of the moderately strain hardening materials could be ob-
matrix decreases and as the particle volume fraction tained by interpolating the solutions for linear hard-
increases. Particle fracture will be favored over inter- ening and non-hardening plastic materials. McClin-
facial decohesion by increasing the particle size, tack began by obtaining equations describing the
increasing the strength of the interfacial bonding and growth of cylindrical voids in both a linear hardening
by non-equiaxed particle shapes. In addition, there is material under axisymmetric stress and for a rigid-
evidence that the nature of the plastic flow of the nonhardening material under axisymmetric defor-
matrix, as well as matrix strength, is important in mation. The two equations are given in Table 1. He
void nucleation. then provided a means of obtaining the behavior of
There have been many attempts to model void moderately work hardening materials by inserting n
nucleation for the case of particle-matrix decohesion. as shown in Table 1 using the relation
All accept the premise that the stress at the interface
must exceed some critical level before void nucleation sin fia
lim - =a.
can occur. There are, however, differences between B-0 ( /9 >
1044 W. M. GARRISONJR and N. R. MOODY

Following earlier work by Berg [88], McClintock then ” = 0.15

obtained an equation describing the growth of a 2w = 180’

cyclindrical void of elliptical cross-section in a linear

hardening matrix under arbitrary biaxial loading.
This solution is shown in Table 1. By analogy with
the results for circular cylinders, McClintock argued
that the equation given in Table 1 would be a
reasonable approximation for the growth of a cylin-
drical void of elliptical cross-section in a rigid non-
hardening material subject to arbitrary biaxial load-
ing. The same method as used earlier was used to
estimate the void growth equation for moderately
hardening materials.
McClintock, Kaplan and Berg extended the Mc-
Clintock analysis for non-rotational hole growth in
shear bands [89]. The mean radius (R) of a void in
shear bands was found to increase as

[ 1,
(a) (b) (cl
--1 dR =-----sinh
1
(1-n): In we/W = 0.07 In “b/Ly ‘0.25 In WC/Iv = 0.37
R dy 2(1-n)
T, = 4wo T, = 2W, T,=4W,

where 7 is the shear stress in the band, u is the normal Fig. 5. Traces of observed void dimensions and the predic-
stress across it and n is the strain hardening coefficient tions of McClintock’s model. FV, is the initial specimen
thickness and W is the current thickness. r, is the specimen
in 7 = 7g”. In 1971, McClintock [90] published an
width 1911. (Reprinted with permission of Mark Perra,
order of magnitude estimate for the growth of ellip- author.)
soidal holes by taking the averages of the three
principal stresses
the McClintock equations greatly underestimate the
1 dR extent of void growth (Fig. 5), but confirm the
-- =&sinh[(l -n)$],
R dz importance of high stress-triaxiality in void growth
(Fig. 6).
where R is the mean of three semiaxes of the void, CI Another continuum description of void enlarge-
is a constant on the order of unity, 6 is the equivalent ment has been presented by Rice and Tracey [92] and
plastic strain, and a, is the mean stress. further developed by Tracey [93]. Rice. and Tracey
Perra [91] evaluated the McClintock equations considered a spherical void in a rigid-perfectly plastic
with cylinders of graphic embedded in blocks of high material. They found that for the case where the
purity copper. Notched and unnotched plane strain remote strain field consisted of a tensile extension at
tensile specimens were machined from these blocks rate i, that the following was a good approximation
such that the graphite cylinders were aligned perpen- for a wide range of stress states.
dicular to the tensile direction in the centers of the
specimens. The growth of the cylindrical voids was R
- =0.2*3exp[ ++:I,
determined metallographically. Perra’s work suggests iR

Table 1. McClintock’s void growth equations

Nonhardening Intermediate Linear hardening
n=O O<nz%l n=l

1 ,n;_J;r)[d(3
Elliptical cavity

1
hr A = &36 ) sinh J(3) (u, + ub) (1 -n),/(3)(a,+ ub)
& --[ 2 26 26
 Ductile fracture 1045

.o- I I I I I Mudry [95], their results suggest that at low volume

fractions of inclusions a will extrapolate to 0.5. These
results are summarized in Figs 7 and 8.
The common feature of the foregoing continuum
I
studies of hole growth is that relative growth rates are
.o - amplified over imposed strain rates by a factor rising
exponentially as a function of the ratio of the mean
normal stress to the effective or equivalent stress. The
second basic result is that work-hardening decreases
void growth rates for a given imposed strain rate.
However, these analyses are based on the growth of
.o -
isolated holes in an infinite media with the assump-
tion of macroscopic incompressibility, so their use
can be justified only for the case in which holes are
very small compared to the hole spacing. The studies

O_ 1
0 0.1 0.2 0.3 0.4 0.5 0.6
EOUIVALENT STRAIN AT IO% DAMAGE, &

XEL762-6378
3i’
2.5
1 /
Fig. 6. The influence of stress state, triaxility (~,,,,/a) on the E / / j
strain required to extend the internal void over 10% of the
neck of the specimen [91].

where R is the average void radius, a, is the mean

normal stress and z0 is yield stress in shear. Tracey
extended this approach to cylindrical voids in work
hardening materials and introduced some measure of
void-void interaction by considering each void in its
own domain of transverse dimension equal to the
void spacing. Void-void interaction was incorporated
by using finite body analysis to approximate the
acceleration in void growth due to particle-particle
STRESS TRIAXIALITY RATIO (urn/ a)
interactions. He found, as McClintock’s equations
suggest, that for a constant stress state void growth Fig. 7. A plot of ln{[ln(R/R,,)]/E} as a function of stress state
for three materials representing three different inclusion
decelerates as the work hardening coefficient is in- volume fractions. The results can be compared to the
creased. He also found for a tixed stress state and theoretical Rice and Tracey equation [94]. (Reprinted with
work hardening coefficient that the rate of void permission from Engng. Fract. Me& 22, 989 (1985). Per-
growth increased with the amount of strain. gamon Journals Ltd.)
There has been no test of the work of Rice and
Tracey which is as elegant as that of Perra’s test of 2, , T
I
McClintock’s equations for the growth of cylindrical
voids. However, Marini et al. have examined the rate
of growth of voids nucleated at large
(diameter = 60 pm) A&O, particles in powder-
processed low alloy steels tempered to a low strength
level [94]. They considered materials with the volume
fractions of A&O, of 0.02 and 0.005. For their
analysis of void growth they used an integrated form
of the Rice and Tracey equation

ln(R/&) = ac exp(/?a,/a)
t,~__--’
E 1
o -0.28 RICE AND TRACEY

replacing J(3) a,,,/2z, by 3/2(a,,,/u,) to account for

work-hardening and determined CLand /3 experi-
mentally. R,, and R are the average void radii before
o? 10-4 2 5
VOLUME
l0-s
FRACTION
2
OF
5
INCLUSIONS
10.’ 2

deformation and at the current value of the equiv-

alent plastic strain (c) respectively. They found Fig. 8. The pre-exponential factor a used by Marini et al.
plotted as a function of initial inclusion volume fraction
/I = 3/2, but that CIvaried with the volume fraction of [94]. (Reprinted with permission from Engng. Frucr. Med.
void nucleating particles. Including the results of 22, 989 (1985). Pergamon Journals Ltd.)
1046 W. M. GARRW)N JR and N. R. MOODY

by McClinrock and by Rice and Tracey have neg-

lected void interactions altogether. The work by
Traccy only very approximately accounts for void-
void interaction effects. The analysis of void growth
should address the crucial matters of finite shape
changes, interactions between neighboring voids, and
the unstable coalescence of neighboring voids or void
arrays. This necessarily involves formulations for
large deformations. Some early work of this type has
been reported by Needleman [96].
Investigations to date suggest that triaxiality is
important in determining both the rate of void
growth and the void shape. This strong effect of
triaxiality is included in both the continuum models
01
discussed. There does not appear to be any test as to 0 2 .4 6 .B IO I i
the role of work hardening capacity on the rate of TRUE STRAIN

void growth. However, there is some evidence that Fig. 9. Comparison of void lengths and widths as a function
strength level does influence the rate of void growth. of strain for maraging steel in low strength (annealed) and
high strength (aged) conditions [97]. (Reprinted with
Cox and Low have studied the void growth rates
permission from Scripfa Metall. 4, 87 (1970). Pergamon
for voids initiated at large manganese sulfides in 4340 Journals Ltd.)
and at Ti(C, N) particles in a 250 grade maraging
steel [9]. For both steels the rate of void growth
observed in notched tensile specimens were much fracture is void coalescence. In this stage the growing
larger than those observed in axisymmetric tensile voids link together, and actual fracture can proceed,
specimens. Moreover, Cox and Low’s work suggests either stably or unstably. It is useful to consider two
that the rate of growth of the void width is much processes by which growing voids coalesce. The first
lower than the rate of void depth, when stress tri- would occur in a material containing only one
axiality is low. In addition, the voids nucleated at the population of void nucleating particles activated
larger inclusions tended to grow faster. during the fracture process (there may be other
The data of Floreen and Hayden [97] and of Psioda particles but they are bonded to the matrix so
[29] suggest that the void growth rate increases with strongly that they do not nucleate voids). Co-
matrix strength. Floreen and Hayden measured from alescence occurs when the voids nucleated at these
cross-sections of tensile specimens the length (parallel particles impinge. In the second, voids are first
to tensile direction) and width of the 10 largest voids nucleated at particles in a particular particle popu-
as a function of strain for two microstructures. (The lation (the one with particles most weakly bound to
voids presumably nucleated at fractured Ti(C, N) the matrix). These voids grow as the material is
inclusions.) The first microstructure was an annealed plastically deformed. At some point, voids are nu-
condition which had a yield strength of about 135 ksi. cleated at a second particle distribution. These par-
Their paper suggests this was as-received material ticles are generally much smaller, more finely dis-
which is normally heavily overaged. The second persed and more strongly bonded to the matrix, so
microstructure was that obtained by aging to obtain that the strains required to nucleate voids at these
a yield strength of 279 ksi. As shown in Fig. 9, the particles are higher. Thus the voids which form first
void growth is more rapid for the high strength coalesce by the growth and coalescence of the voids
microstructure. Psioda also examined the influence of nucleated at the second particle population. This
yield strength on the rate of growth of voids in a process is often referred to as void sheet coalescence
maraging steel and found the rate increased with [6,22]. Experiments suggest that the coalescence
increasing matrix strength. event is, in general, rapid, occurring over a small
In summary, the continuum predictions for the interval of macroscopic strain [98].
growth of an isolated void predict that the rate of Two mechanisms have been proposed for co-
void growth will increase with void size and that the alescence by void impingement. In the mechanism
growth rates are amplified over imposed strain rates suggested by Cottrell [99], the ligaments between
by a factor rising exponentially as a function of the growing voids would simply neck down to a “point”.
ratio of the mean normal stress to the effective stress. Thus a fracture surface of such a failure would be
These equations are in qualitative agreement with voids of a rather uniform size and spacing with the
experimental observations but underpredict the ex- ridges around the dimples representing material
tent of void growth. drawn to a knife-edge fracture of zero cross-section.
A second mechanism which would result in direct
1I.C Void coalescence impingement of the growing voids is the slipping off
II.C.1 Mechanisms of void coalescence. The final mechanism observed in the tensile fractures of single
step in the idealized three stage process of ductile crystals [loo]. One should be able to distinguish
 Ductile fracture 1047

between the slipping off and necking down processes by a void sheet mechanism in which the voids within
by cross-sectioning fracture surfaces. While the au- the sheet were nucleated at carbides precipitated on
thors are unaware of studies of the mechanisms of tempering.
direct void impingement, there exist many examples While it is commonly stated that void coalescence
of direct void impingement. For example, it has been in maraging steels is by direct impingement, Roesch
reported that for certain heat-treatment conditions and Henry [104] have found that once the inter-
the fracture surfaces of maraging steels consist of metallics, precipitated on aging of these steels, reach
equiaxed dimples nucleated at titanium carbides [9]. a critical size they too can nucleate voids. Therefore,
In addition, fractures of spheroidized steels have one would expect in some aging conditions for the
been claimed to be through direct impingement of large voids nucleated at the Ti(C, N) inclusions to
voids nucleated at the carbides [ 101, 1021. An ex- coalesce by a void sheet process where the sheets are
ample of direct void impingement from work by formed by voids nucleated at the intermetallics.
L. Thompson [103], is shown in Fig. 10(a). Here, Low observed in 2000 and 7000 series aluminium
voids nucleated at large inclusions are seen to alloys that the large voids nucleated at constituent
coalesce by direct impingement for a Fe-Mn-Cr particles often coalesced by the formation of void
stainless steel. In Fig. 10(b), the voids nucleated at sheets [28,105, 1061. The void sheets were formed by
large inclusions do not always impinge directly, but the growth and coalescence of voids nucleated at
are often separated by regions of very small dimples. dispersoids, submicron-sized particles introduced in
The classic example of void sheet coalescence by the alloy to control grain growth.
the nucleation of voids at a second particle popu- In addition, void sheet coalescence can connect
lation in a region of shear localization between larger larger voids by nucleating microvoids in processes
particles is of course that of Cox and LOW 191.They which do not require a second particle population.
observed, as shown in Fig. 11, that large voids In a study of hydrogen effects on the y’ strengthened
nucleated at sulfide inclusions in 4340 steel coalesced superalloy IN903, Moody and Greulich observed
that voids nucleated at fractured matrix carbides
(NbC) and coalesced by a void sheet mechanism [60].
Void sheet formation occurred at grain boundary
carbides in uncharged samples and along coplanar
slip bands in hydrogen charged samples. High
magnification fractography and TEM of strained
samples suggested that the voids along the coplanar
slip bands formed at slip band intersections.
Ordinarily void sheet coalescence brings to mind
the classic micrograph in the paper by Cox and Low
[9]. However, Rogers defined the term void sheet
coalescence in rather more general terms [6,22] to
include notch and strain localization effects. Puttick
[5] found in smooth axisymmetric tensile specimens
of OFHC copper that voids form first in the center
of the tensile specimen and coalesce to form an
internal notch (Fig. 12). Rogers observed that the
notch tends to concentrate or localize the defor-
mation at its tip in narrow bands of high shear strain
at an angle of 50”-60” to the transverse plane. Under
the combined action of the applied tensile stress and
the resulting shear strain, “sheets” of voids are
nucleated in these bands, growing until coasescence
(by impingement) occurs, producing local fracture of
the “void” sheet. Thus, the requirement for void
sheet coalescence is that voids nucleate and coalesce
on a nominally 2D surface defined by a region of
strain localization. In general, void sheet coalescence
does not require a second population of void nucle-
Fig. 10. Scanning electron microscope fractographs from ating particles. Void sheet coalescence defined in
Charpy impact specimens of (a) 20Mn-13Cr-O.lN-1 SSi general terms is observed in many different situ-
steel fractured at room temperature, and (b) 16Mn-13Cr ations.
alloy fractured at - 196°C (1031. (Reprinted with permission Examples of void sheet coalescence in which a
from Mechanical Properiies and Phase Transformations in
second particle population is not required are various
Engineering Materials (edited by Antolovich, S. D., Ritchie,
R. 0. and Gerberich, W. W.). The Metallurgical Society, types of grain boundary fracture. Ductile inter-
420 Commonwealth Drive, Warrendale, PA 15086, (1985).) granular fracture can occur in steels when they are
1048 W. M. GARFCFQN JR and N. R. MOODY

Fig. 11. Coalescence of two voids by a void sheet mechanism in 4340 steel [9].

overheated [107-1091. Overheating occurs when a it is not clear if shear localization requires microvoid
steel is austenitized at a temperature sufficiently high formation in the region between the larger inclusions.
to dissolve some or all of the sulfide inclusions; on Edelson investigated strain concentration between
quenching, sulfides precipitate at the austenite grain voids by examining copper sheets containing arrays
boundaries. Depending on the relative amounts of of holes which were coated with a strain sensitive
inclusions within the matrix and on the grain bound- birefringent film [ 1141. He examined holes positioned
aries, ductile intergranular fracture can take place along 35” diagonals and holes positioned along 60”
[1lo]. In such situations the toughness is apparently diagonals. He found that in both arrays the local
controlled by the spacing of the particles on the grain plastic strains at the hole edges were much greater
boundary [l 111. Intergranular fractures are also than the macroscopic strain; furthermore, the holes
observed in high strength aluminium alloys [112]. in the 35” diagonal arrays were joined by a path of
Intergranular void sheet coalescence can also re- high strain, almost as high as the highest local strains
sult if there is a precipitate-free zone. Kirman at the edge of the holes. Similar experiments and
investigated the influence of aging on the micro- observations have been made by Bourcier and Koss
mechanisms of ductile fracture of an Al-Mg-Zn [115].
aluminum alloy (7075) [ 1131. He found that when the Various numerical computations tend to corrobo-
alloy was sufficiently overaged, fracture occurred rate these experimental observations. Needleman
intergranularly by the coalescence of voids nucleated [96], Nemat-nasser and Taya [116], and Tvergaard
at grain boundary MgZn, precipitates. The voids [117] have shown in finite element calculations based
nucleated preferentially at the grain boundary pre- on an elastic-plastic matrix with a smooth yield
cipitates due to strain localization in the relatively surface, that neighboring voids become linked by a
soft precipitate-free zone. highly strained region. However, no shear band
II.C.2 Modeling void coalescence by void sheet instability would occur in these materials at realistic
formation. The mechanism of void sheet coalescence strain levels [118]. Tvergaard considered a similar
suggests strong localization of strain between the initial void distribution, but permitted microvoid
larger voids. It is unclear if the void sheets are the nucleation. Using Gurson constitutive equations
product simply of strain localization between the [ 1191for a porous medium, he predicted strain local-
larger voids or microvoids strongly contribute to the ization and microvoid nucleation in a narrow band
localization of flow between the larger voids. Experi- that begins at a large void, and grows into the matrix
ments and continuum calculations suggest that strain in the direction of another large void (selected to be
concentration can occur before microvoids form, but at 40.8” from the initial void [118].
 Ductile fracture I1049

Fig. 12. Section through neck of a tensile specimen of copper showing profusion of voids and central crack
in the necked region [5]. (Reprinted with permission from Phil. Mag., Taylor &I Francis, Ltd. (1959).)

III. MECHANICAL PROPERTIES properties (strain to fracture for the tensile specimens
and the critical crack tip opening displacement at
In this section we will discuss the process of ductile fracture initiation in toughness type specimens) can
fracture in the context of specimens often used to be predicated in terms of microstructure and/or
assess fracture behavior. The specimen geometries to continuum treatments of the three stages of the
be discussed are: smooth axisymmetric tensile speci- ductile fracture process.
mens, notched axisymmetric tensile specimens, plane
strain tensile specimens and fatigue precracked speci- IILA Ductility of smooth axisymmetric tensile
mens such as compact tension specimens or three specimens
point bend specimens used to measure fracture initi- The most commonly reported engineering prop-
ation toughness. The topics to be addressed are the erty is the strain to fracture for a smooth axisym-
stress states and strain distributions in the particular metric tensile specimen. The strain to fracture is
geometries, the micromechanisms of the fracture equal to ln(A,/A,) where A, is the initial cross-
process, and finally the extent to which mechanical sectional area and A/ is the smallest cross-sectional
1050 JR and N. R.
W. M. GARRISON MOODY

area after fracture. Initially the cross-sectional area garded the fracture event to be the growth of this
remains a constant along the specimen length or central flaw [1271.
gauge length. At some strain, the uniform strain, the An important factor in determining the tensile
specimen begins to neck. Beyond the onset of neck- ductility is the volume fraction of particles. Edelson
ing, strain will become concentrated in the necked and Baldwin investigated the effects of independent
region, although plastic deformation occurs through variations of particle volume fraction and size on the
the length of the necked region until late in the tensile properties of copper [128]. The particles
fracture process. The problem of necking of an included voids and particles of molybdenum, lead,
axisymmetric tensile specimen has been assessed by alumina, graphite and iron. They found that tensile
Bridgman [ 1211and subsequently treated numerically ductility depended primarily on the total volume
by Needleman [122], Chen [123], and Norris et al. fraction of particles and voids. Their alloys were
[124]. The analyses suggest that the effective plastic prepared by powder methods and all materials to
strain (cP) is essentially uniform across the neck and which inclusions were deliberately added contained
may thus be calculated from the diameter change of 2% voids. Their plot of tensile ductility as a function
the specimen. The second conclusion is that the ratio of total volume fraction of inclusions and voids in
of the mean normal stress (a,) to the equivalent Fig. 13 largely supports their conclusion. However,
stress (5) increases as the necking progresses. In fact, the data are not totally conclusive as, for example,
at the center of the specimen a,,,/5 can increase from the ductility varies from about 0.15 to 0.5 at an
0.33 to about 0.6 at an equivalent plastic strain of 0.9 inclusion volume fraction of about 0.7. The general
[125]. trend observed by Edelson and Baldwin has been
Studies of the fracture process in smooth axisym- confirmed in investigations of steels [129, 1301 and
metric tensile specimens suggest that voids can ap- for other tensile geometries [130, 1601. However,
pear well before fracture, and their volume fraction while it might be argued that particle spacing does
is highest in the center of the tensile neck. This region not influence tensile ductility, there is little doubt that
is the volume combining both the highest plastic particle shape can be important. For example, in
strain and the highest ratio of 0,/d. The area fraction steels containing inclusions elongated in the rolling
of voids in cross-sectioned specimens decreases as the direction, the tensile ductility depends on how the
distance from the plane of the fracture surface tensile specimen is oriented with respect to the rolling
increases. direction [131, 1321.
There have been a number of studies of the extent The work of Edelson and Baldwin and others
of void growth prior to fracture and of the failure show the smooth axisymmetric tensile ductility will
mechanisms in axisymmetric tensile specimens. decrease with increasing particle volume fraction.
Rogers examined the fracture process in OFHC Continuum treatments of void growth can qual-
copper and observed that voids formed well before itatively explain this trend but they are insufficient to
fracture, with the largest voids at the center of the predict fracture strains. For example, the Rice and
neck [6,22]. His observations suggested that fracture Tracey equation for the growth of an initially
began at the center of the neck by the consolidation spherical void is
of several voids. Complete specimen fracture then
occurred by the propagation of the central crack via dR
either void sheet formation or by slip (ductile cut- - = 0.32 exp(l.5 ~,,,/a) dr,
R
ting). In a similar manner, Hancock and Mackenzie
examined the fracture behavior of two steels [ 1261
where R is the average radius of the void. On
but related their observations to the true stresses and
integrating with the assumption that stress state
the effective plastic strain (6,). They observed the
remains constant and there is no void nucleation
stress increased monotonically with KPuntil it reached
strain, the fracture strain is given by
a maximum value, then dropped sharply. Metal-
lographic examination of specimens showed that
before this maximum stress was achieved discrete Z, = ln(R,/R,)/O.32 exp(l.5 ~,,,/a),
holes had formed. Beyond the maximum stress, they
observed large central flaws formed either by im- where R, and Riare the final and initial void radii
pingement of some large voids or by the linking respectively and for a smooth axisymmetric tensile
through void sheet coalescence of several voids. specimen a,,,/6 is initially 0.33. In applying this
Therefore, Hancock and Mackenzie took the strain equation we will take 0.32 exp(l.5 a,/e) to be about
at the maximum stress to be the fracture strain. They 0.50. Then, in absence of R//R, measurements from
did not observe how the central flaw propagated the center of the fracture surface, one might take Rf
through the material. However, their measurement to be X,/2, where X0 is some measure of spacing of
of the central flaw size, at what they have defined as the voids or particles. A variety of possible choices
the fracture strain, and their use of these data in of X0 exist [133]. These include the mean free path,
estimating the fracture initiation toughness of the the average nearest neighbor distance on a plane and
steels investigated leaves little doubt that they re- the average nearest neighbor distance in a volume.
 Ductile fracture 1051

PRESENT WORK
I
0 COPPER- IRON-MOLY
0 COPPER HOLES
x COPPER Cl#?OMlUM
0 COPPER AWMINA
A COPFER IRON
P COPPER MOLYBDENUM
ZWILSKY 8 GRANT
( COPPER ALUMINA
1.0 . 3 COPPER SILICA

VOLUME FRACTI0N.f

Fig. 13. A plot of Ekielson and Baldwin’s observations of the influence of void and inclusion volume
fraction on the ductility obtained in smooth axisymmetric tensile specimens [128].(From Bdelson, B. I.
and Baldwin, W. M., Jr., The Effect of Second Phases on the Mechanical Properties of Alloys, Trans.
ASM, 55, 238 (1962). American Society for Metals, Metals Park, OH 44073, U.S.A.)

For spherical particles these spacings can be written voids initiated by fracture of matrix carbides in both
as hydrogen charged and uncharged samples. However,
final failure did not occur by impingement of these
i = (4/3) %(l -f)lf voids but by void sheet formation at grain boundary
mean free path, carbides in uncharged samples and at slip band
AZ= 0.72 &f -‘I2 intersections in hydrogen charged samples. The
average nearest neighbor distance on a amount of void sheet formation increased signifi-
plane, cantly with a corresponding decrease in the amount
A3 = 0.89 &f-l” of void growth around the matrix carbides as the
average nearest neighbor distance in a hydrogen concentration increased. To determine the
volume, applicability of void growth equations to calculation
of the fracture strain, the extent of void growth at
where f is the particle volume fraction and & is the failure corresponding to the average-sized matrix
.
average particle radius [1331. carbide was determined from fracture surface meas-
Calculations of the fracture strain using the Rice urements. The average matrix carbide diameter
and Tracey equation for void growth for these three (6.3 pm) and volume fraction (0.003) were measured
choices of X,, are compared with the Edelson and in a previous study [62].
Baldwin data in Fig. 14. All three spacings give the As shown in Fig. 15(a), a linear relationship exists
observed trend of ductility decreasing with increasing between the measured fracture strain and the loga-
particle volume fraction in the case of voids or rithm of the measured ratio of final void radius to
particles very weakly bound to the matrix. Using the void nucleating carbide radius as predicted by the
average 3D nearest neighbor spacing actually gives a models of McClintock [86], Rice and Tracey [92,93]
reasonable fit to the Edelson and Baldwin data over and Marini et al. [94]. However, the fracture strains
a wide range of volume fractions. predicted from these models differ significantly as
Using actual measurements of R//R,, however, shown in Fig. 15(b). (In addition, the predicted values
it appears that the Rice and Tracey equations for do not include the nucleation strain of 0.08 for the
void growth overestimate the fracture strain. Recent average sized carbide.) Only predictions from the
studies [60-62] of hydrogen effects on the fracture of model of McClintock [86] are in reasonable agree-
the y ’ strengthened superalloy IN903, which failed by ment with measured values. However, this model is
microvoid coalescence illustrates this point. The for cylindrical voids, albeit oriented such that growth
1052 W. M. GARREONJR and N. R. MCXDY

0 COPPER PlJMINA
A COPPER IRON
0 COPPER MOLYBDENUM
ZWILSKY EL GRANT

Q 0.1 0.2
VOLUME FRACTION,f

Fig, 14. The results of Edelson and Baldwin compared to plots of predicted tensile ductilities using the
Rice and Tracey equation for void growth assuming inclusion spacings of (a) mean free path; (b) average
nearest neighbor distance on a plane; and (c) average nearest-neighbor spacing in a volume.

) (a)
occurs in the transverse direction parallel with the
fracture strain measurements, whereas the voids that
formed around the carbides were spherical. The
spherical void model of Rice and Tracey [92,93] and
the empirical modification of this model by Marini et
al. [94] to account for void interactions overpredict
the fracture strains. These results suggest that
void-void interactions and deformation character
(coplanar slip) may exert a strong influence on void
growth.
In (RJR,) Another approach to predicting tensile ductility is
that of Brown and Embury [134]. According to their
approach, they assume that fracture occurs by rup-
(b) ture of the ligaments between voids, which occurs on
1.6 - a single plane when the void length becomes equal to
the spacing between the void nucleating particles
Rice a Trocey (Fig. 16). They assume that the strains associated
with the coalescence process will be small, and ignore
this contribution to the tensile ductility. Beginning
with these assumptions they derive an expression for
the fracture strain c~, as follows:

,=ln($)=ln(l +tg+c.)
0.2 0.3
Ef, measured

Fig. 15. (a) A plot of the correlation between tensile fracture

strain and the quantity ln(R,/R,) for IN903 charged to the
hydrogen levels shown in Table 2. (b) The measured frac-
ture strains are compared to fracture calculated using the This equation provides reasonable agreement with
various void growth equations summarized in Table 2. the experimental data they use for comparison, with
 Ductile fracture 1053

criterion based on the concepts of fracture mechanics.

This interesting work has not received the attention
it deserves although Stout and Gerberich have used
it in modeling fracture initiation toughness [136].
Hahn and Rosenfield used McClintock’s equations
for the growth of cylindrical voids. They integrated
this equation from zero strain to the fracture strain,
using the Bridgman analysis to obtain the current
values of the stress in terms of 5 and n to obtain an
equation In@/&) = F(i, n). Do is the initial particle
diameter and D is the maximum stable hole diameter.
They defined fracture as the condition where one
large void (or a small group of neighboring voids that
0 - have coalesced) becomes unstable and grows rapidly
as a crack. Regarding this crack as a critical flaw,
Fig. 16. The coalescence criteria of Brown and Embury.
Coalescence will occur when the vertical void height equals they suggest that its size is stress sensitive and take
the spacing between inclusions [134]. (Reprinted with per-
mission from The Microsiructure and Design of Alloys. The D (a/)* = const ,
Iron and Steel Institute (1975).)
where a,is the equivalent stress at fracture. Using this
the extent of agreement depending on the value used equation it is possible to estimate the fracture strain
for t,. The derivation of their equations begins with if F and -0, are known for one reference condition
the assumption that the rate of void growth, dRJd6, from the relationship.
is given by

Finally, while the work by LeRoy et al. [ 1371offers

where R,, is the initial void radius. At the strain at no new approaches as to how one might model tensile
which void growth ends (t,), the void length is ductility, it is probably the most comprehensive at-
tempt to calculate the tensile fracture strains through
2R, = 2R,(l + Q, continuum models and measured parameters charac-
terizing the microstructure and their influence on the
which is then equated to the mean planar spacing of fracture process.
spherical particles, where S is the inclusion volume
fraction. 1II.B Ductility of notched axisymmetric tensile
specimens
The fact that re-entrant notches can influence
fracture behavior has been known for some time
[138] and led to standardized impact tests for
To obtain the fracture strain [e, = h&4,/A,)] they notched specimens [139, 1401. However, the earliest
then essentially take reference to the use of notched axisymmetric tensile
specimens appears to be that of Ludwik and Schau
A0 [141]. They conducted smooth and notched tensile
-=1+e,
Ar tests of a low strength mild steel; the smooth tensile
specimens had diameters of 1Omm and the notched
and, modifying this to incorporate nucleation strain tensile specimens were 18 mm in diameter, and
the following equation is obtained: 1Omm in diameter at the base of the notch. They
reported that incorporating a notch, 0.4mm wide,
A increased the ultimate tensile strength by almost a
a=1 +e&.+c,.
A/ factor of two and lowered the reduction in area from
69% to only 22.5%. Their report contains the most
These last two steps would seem to be valid only for important consequences of notched tensile testing.
very small strains. In addition, the approach does not These are the elevation of the flow stress (notch
incorporate the important role of stress state in void strengthening) and a marked decrease in the strain to
growth. fracture.
Another approach to estimating tensile ductility is Notch strengthening is due to the elevation of the
that due to Rosenfield and Hahn [135]. This ap flow curve by what is termed plastic constraint. To
preach combines observation of the fracture process, maintain equilibrium, no stresses can act normal to
continuum equations for void growth, and a fracture the notch surface; all of the tensile load must be
1054 W. M. GARRISON
JR and N. R. MOODY

taken by the metal in the cross-section at the notch. analysis by Earl and Brown [143] to show that the
Therefore a mass of unstressed material exists above stress state (o,#) varied across the specimen rising
and below the notch. This material in the cross- from l/3 at the notch surface to
section of the notch wants to contract laterally
because of the Poisson effect, but it is constrained by (a,/~?) = l/3 + ln(d/2R + 1)
the material above and below. This constraint causes
both radial and transverse stresses in addition to the at the axis of the specimens, where d is the diameter
stresses along the tensile axis and leads to a triaxial of the specimen at the notch, R is the notch radius
as opposed to uniaxial stress state. This in turn leads and a,,,/5 at the center is used as a measure of stress
to an elevation of the stress-strain curve because triaxiality for the specimen. Thus a,,,/5 ranged from
higher applied stresses in the tensile direction are 0.33 to 1.99. Hancock and MacKenzie used three
needed to reach yielding. It appears that the reduced heats of steel: a steel designated Ql. a heat of
fracture strains are the direct result in the change in HY130, and a heat of electroslag remelted HY130
stress state. As shown in Fig. 17 the reduced tensile (HY 130/ESR). The first two heats contained inclu-
ductility correlates reasonably well with the increased sion stringers while the HYl30/ESR heat did not.
stress triaxiality 11421. They observed that fracture began with the nucle-
While notched tests have been used to provide ation of voids at inclusions which coalesced to form
qualitative information about “notch sensitivity of a large cavity at the center of the neck or notch. Also,
materials”, they have also been used extensively in the strain to fracture decreased rapidly as the stress
studies of ductile fracture and in modelling fracture state parameter a,,,/6 increased as shown in Fig. 18.
initiation ahead of sharp cracks. Our interest here is They also found, as shown in Fig. 18, that in the
what these tests tell us about the mechanisms of presence of inclusion stringers, the ductilities were
ductile fracture. Their importance is that these tensile much lower in the short transverse direction than in
specimens can provide information about the role of the long transverse direction. They concluded that
stress state in the various stages of ductile fracture, void nucleation occurred at very small strains in
nucleation, growth, coalescence and the total strain these materials, with void growth the dominant event
to fracture. leading to failure. Quantitatively, they found that the
Work in this area is most strongly associated with Rice and Tracey equation for void growth
a series of studies by investigators in the Department
of Mechanical Engineering at the University of dR
Glasgow. In the first paper of this series Hancock - = 0.28 exp(lSa,/C) dE,
R
and MacKenzie [126] utilized U-notched tensile
specimens. They used an extension of the Bridgman
provides an approximate failure criterion. For a
given steel tested in a given orientation, the fracture
strain could be expressed as
50, n

A Ql (aHY-80) i; = a exp(- 1.5 0,/e),

v ESRtaHY-130)

where c( varied with material. Thus the Rice and

IRRADIATED (ROOM Tracey equation for void growth seems to provide a
TEMP TO llOO°F)
30 - 0 24 ST ALUM.
qualitative explanation for both the effects of inclu-
sion volume fraction and stress state on ductility for
axisymmetric specimens.
Hancock and MacKenzie noted that the failure
initiation event represented a transition between
homogeneous void growth and the formation of a
large central flaw due to the coalescence of several
voids. The general interpretation of these results by
the Glasgow school is that fracture occurs by shear
localization over some small volume containing a
certain number of large inclusions which becomes the
central crack that propagates to the extremity of the
specimen. Hancock and MacKenzie suggest that the
Fig. 17. The reduction in area of notched tensile specimens coalescence is due to shear localization between
vs the reduction in area of the corresponding smooth adjacent voids and suggest further that the shear
tension specimens divided by 3a,/d where u,,,,/@is the localization, initiates at a critical void volume frac-
triaxiality of stress state for the notched specimen [142].
tion. The idea that fracture initiates when the void
(Reprinted with permission from J. eng. Mats. Tech. The
American Society of Mechanical Engineers, 345 East 47 volume fraction reaches a critical level is used in
Street, NY 10017 (1978).) many recent continuum studies of ductile fracture.
 Ductile fracture 1055

- EXPERIMENTAL
--- E’= a expt-3o n /2R

\/al 1 LTI
t
/

(a)
A
I
I
0 I
I I
2
i\
3
1
4
1
5
1
.6
1
7 8
1”
9 IO
EFFECTIVE PLASTIC STRAIN TO FAILURE INITIATION (&I

- EXPERIMENTAL
--- 5’; a exp(-30.,25)
Fig. 19. A drawing of an unnotched plane strain tensile
specimen, with dimensions defined by Corrigan et al. [151].
(Reprinted with permission of the American Welding
Society.)

shear localization will occur in volumes which con-

tain locally higher volume fractions of inclusions.
.4 -
(b)
Thus shear localization and fracture initiation are
0 ,I 2 3 4 S 6 7 .8 9 IO II 12 most likely to be associated with regions containing
EFFECTIVE PLASTIC STRAIN TO FAILURE INITIATION (i, 1
a locally higher inclusion volume fraction. Because of
the higher inclusion volume fraction, they are the
- EXPERIMENTAL
first regions to achieve the void volume fractions
--- h’= D apt-so m/2e, required for shear localization.
Subsequent papers [144,145] by the Glasgow
group confirmed that for steels containing inclusions
weakly bound to the matrix, and heat-treated to a
bainitic or martensitic structure, the tensile fracture
strains for a given material could be fitted reasonably
well to a curve of the form 4 = a exp( - lSu,,#).
However, they noted that the tensile ductilities of
(c)
high strength aluminum alloys and of a steel contain-
0 I 2 3 4 5 6 7 .8 .9 IO II 12 ing 50% pearlite were much less sensitive to stress
EFFECTIVE PLASTIC STRAIN TO FAILURE INITIATION ( Ef I
triaxiality than were the tensile ductilities of mar-
Fig. 18. (a) The effective plastic strain to initiate failure, 4, tensitic and bainitic steels. We mention this only to
as a function of the stress state for Ql tested in the long caution the reader that the dependence of tensile
transverse and short transverse directions. (b) The effective ductility on stress state observed for the martensitic
plastic strain required to initiate failure, E,, as a function of
the stress state for ESR tested in the long and short
and bainitic steels is by no means universal, and
transverse directions. (Note the lack of directionality in this appears to be associated with a particular fracture
steel compared with those prepared by more conventional mechanism.
techniques.) (c) The effective plastic strain required to ini- Concurrent with the work of the Glasgow group
tiate failure 4, as a function of the stress state for HY130 was an explosion of activity by the continuum
tested in the long and short transverse directions. (Note that
in the highly triaxial states, HY130S.T. fails at strains of the mechanics community, based on the work of Gurson
order of 0.02; whereas in uniaxial tension, failure occurs [119]. The important feature of Gurson’s work is that
after a strain of the order of 0.65.) [126]. (Reprinted with the deforming body is no longer incompressible, but
permission from J. Me&. Phys. Soli& 24, 147 (1976). dilatation of the body is permitted by void growth.
Pergamon Journals Ltd.)
He developed constitutive relations for such a mate-
rial which permit yield criteria and flow rules for
Interpretations of Hancock and MacKenzie’s data porous (voided) materials. The Gurson model, with
in later papers of the Glasgow group [144, 1451 tend some modifications, has been used in continuum
to reaffirm the importance of achieving a critical void analyses of ductile fracture for smooth and notched
volume fraction, but there is an emphasis that the axisymmetric tensile specimens, plane strain tensile
event is localized, occurring on a size scale on the specimens and in the highly localized region of
order of the spacing between the large inclusions. An damage accumulation associated with fracture ini-
important feature of these interpretations is that tiation at blunting crack tips.

PCS. 4,11-o
1056 W. M. GARRISONJR and N. R. MOODY

In applying the Gurson approach to ductile frac- methods have apparently been used to measure
ture in tensile specimens, there have been essentially fracture strains with this test and not all are com-
two approaches. The first is to introduce an in- pletely explained.
homogeneity in the material by introducing a region It appears that the plane strain, strain state is
containing initially a higher void density than the rest susceptible to strain localization. Anand and Spitzig
of the body. Then conditions leading to shear [ 1551 observed, using an age-hardened martensitic
localization within this region are determined. The steel and a specimen similar to that shown in Fig. 19,
second approach is to assume an initially uniform that slip bands formed at about 38” to the load axis
distribution of inclusions or voids and then deter- at strains about a fifth of the eventual fracture strain.
mine the void volume fractions and strains as a With increasing strain the number of slip bands
function of position within the body. Failure then became more numerous, wavy, longer and more
becomes a competition between the development of pronounced (Fig. 20). The fracture path (Fig. 20)
a critical void volume fraction within some region was quite flat and was also about 38” + a few degrees
and strain localization. to the load axis. It is not known if fracture occurred
Yamamoto [146] and Saje et al. [147] have deter- within the fine slip bands but it does appear that a
mined conditions under which planar void in- slant fracture is normally observed for specimens of
homogeneities, which extend through the body, lead the type shown in Fig. 19.
to shear localization. While they observe that planar Clausing [120] determined the strains to fracture
inhomogeneities can become regions of shear local- for a variety of materials using both plane strain
ization in axisymmetric tensile specimens such behav- tensile specimens and smooth axisymmetric tensile
ior is rarely observed. Using a criterion for failure specimens. Comparison of his results, shown in
suggested by Tvergaard [148] that some volume Table 3, show that the ductility was less for the plane
achieve a critical void volume fraction, Tvergaard strain specimens than for the smooth axisymmetric
and Needleman [149, 1501 have shown that void specimens. These data and the fact that the ratio of
coalescence in the center of the neck will precede plane strain and axisymmetric ductilities decreases
macroscopic strain localization in axisymmetric ten- with increasing strength and presumably, therefore,
sile specimens. decreasing work hardening exponent, suggest that
strain state and work hardening capacity could be
1II.C Plane strain tensiles important factors in determining ductility when the
Plane strain tensile samples have no strain along mechanism of fracture is ductile. Needleman and
the width direction. Such specimens can utilize the Rice [199] discuss the results of Clausing and, using
configuration shown in Fig. 19 or they can be yield surfaces with vertices, suggest the ratio of
achieved using U notches of various dimensions. The effective plane strain ductility to the ductility of
ratio of a,,,/~? for the specimen in Fig. 19 is 0.58, but smooth axisymmetric specimens will scale as
this ratio can be raised dramatically by using notched I(3 n)/(l + 3 n)l”*, where n is the work hardening
plane strain tensile specimens, Our interest in plane exponent.
strain tensile specimens is twofold: (1) these samples Numerical studies of the necking of plane strain
have been used in fundamental studies of the fracture tensile specimens by McMeeking and Rice [1561 and
process, and (2) they have been used to provide a Burke and Nix [157], which employed the simplest
measure of the critical fracture strain when modeling smooth yield surface theory of plasticity to character-
fracture initiation ahead of initially sharp but blunt- ize the material behavior, found only continued
ing crack tips. growth of the diffuse neck with no tendency for the
The specimen shown in Fig. 19 was essentially formation of localized shearing (shear bands). How-
developed by Corrigan et al. [151] to investigate the ever, other continuum analyses, adopting a variety of
biaxial strength of welds in heat-treated steel. The assumptions as to material homogeneity and often
plane strain stress state should result in a higher more complex yield surfaces, have shown that the
ultimate tensile strength than that observed for plane strain, strain state is unstable with respect to
smooth axisymmetric tensile specimens. To ensure the localization of plastic flow into shear bands.
that the elevation in ultimate tensile strength reached These analyses include those of Yamamoto [146],
the value expected from continuum analyses and was Saje et al. [147] and Ohno and Hutchinson [158]
independent of specimen thickness, Corrigan et al. which utilized the existence of material in-
required that W > 30 t, B I 3 t and t I 0.6T, where homogeneities (i.e. regions of higher initial void
W, B, t, T are defined in Fig. 19. This specimen, in volume fraction). As summarized by Tvergaard et al.
various forms, has since been used for a number of [159], other work has shown that yield surface vertex
investigations: determination of the yield locus for effects, departures from normality, and dilatational
titanium alloy sheet [152], studies in which the plane plastic flow can all lead to localization of plastic flow
strain tensile ductility was used to provide a measure under conditions of plane strain.
of the fracture strain ahead of crack tips A study by Hancock and Brown [125] of ductile
[120, 153, 1541, and for studies of ductile fracture fracture used both unnotched and notched axisym-
[144, 1451. We note in passing that a variety of metric tensile specimens and unnotched and notched
 Ductile fracture

Fig. 20. Plane strain tensile specimen of an aged maraging steel: specimen shown at a strain of 0.15 in
(a) and (b) and at the fracture strain of 0.17 in (c) and (d) [155]. (Reprinted with permission from J. Mech.
Phys. Solids 28, 113 (1980). Pergamon Journals Ltd.)

Table 2. Ductility and void growth in hydrogen charged plane strain tensile specimens. They used finite ele-
IN903 ment calculations to determine both the stress state
Calculated 6, (a,,&) and the strain at the position of fracture
4 initiation. When ~,,,/a at fracture is plotted as a
(meas.) RJR, q,,,/ci A B C function of effective strain at fracture, the data for
0 0.467 2.61 0.489 0.49 1.63 0.69 a particular steel tested in a given orientation fall in
770 0.192 1.63 0.376 0.25 0.98 0.42 a curve as shown in Fig. 21(a). This suggests that
1876 0.173 1.42 0.368 0.18 0.71 0.30
stress state and not the strain state is the important
2900 0.151 1.57 0.358 0.23 0.93 0.39
500 0.142 1.36 0.353 0.16 0.64 0.27 factor in determining the strain to fracture. indeed,
Hancock and Brown state:
A: McClintock: c,=

1.
“A recurrent theme of localization theories is the effect
J(3) 1 of strain state for quite a range of constitutive equations
WWR,) Imsinh&(l -n)+i
and physical mechanisms. However, the experimental
data of Figs 1518 do not bear this out as the data are
B: Rice and Tracey: c,= ln(R,/R,)O.283 exp(l.5 ~,,,/a). well correlated by the stress state parameter (a,/#) for
C: Marim et al.: L,= ln(R,/R,)/0.67 exp(l.5 u,,,/B) (using both axisymmetric and plane states of strain. The
a = 0.67 with volume fraction carbides = 0.003). ductility in plane strain tension is indeed less than that
1058 W. M. GARIUON JR and N. R. M~C~DY

Table 3. Comparison of smooth axisymmetric and plane strain tensile

ductilities
Yield strength Fracture strain
Steel (ksi) Axisymmetric Plane strain
ABS-C 39.1 1.04 0.75
A302-B 53.8 0.98 0.72
HY-80 85 1.22 0.86
HY-130(T) 136 1.06 0.52
18Ni(180) 178 1.00 0.42
IONi-Cr-MoCr 182 1.16 0.36
18Ni(250) 248 0.89 0.15

in axisymmetric tension as in Clausing’s (1970) work but portant role in this regard. . . . We find that in plane
this is consistent with the reduction in ductility due to strain bluntly notched specimens, the non-
the increase in triaxiality and its effect on hole growth
proportional loading arising from shear localization
rate, the presence of holes being an essential feature of
all the materials examined in the flow field before failure does lead to significantly earlier fracture strains.” Our
initiation.” conclusion, after reviewing experimental and con-
tinuum studies of plane strain behavior, is that the
Needleman and Tvergaard took up this challenge plane strain tensile specimens of the type shown in
in their 1984 paper [150]. Considering voided ma- Fig. 19 are prone to shear localization. In fact,
terial, a modified Gurson theory and a fracture specimens of this type fail by a nominally flat fracture
criterion that the local void volume fraction exceed a at angle N 35-45” to the load direction even for steels
critical level, they investigated both notched axisym- having yield strengths as low as 250 MPa (36 ksi)
metric and notched plane strain tensile specimens of [160].
geometries similar to those used by Hancock and While smooth and notched axisymmetric tensile
Brown. They were able to duplicate to some degree specimens fail by the formation of an interval cavity
the results of Hancock and Brown (Fig. 21b). They which propagates to the surface, plane strain tensile
concluded that “as long as deviations from constant specimens of the type shown in Fig. 19 tend to fail
triaxiality history are not too great, the onset of first at the surface. These surface initiated cracks
failure, when represented in a plot of stress triaxiality propagate towards the center of the specimen. The
vs effective strain is approximately represented by a apparent signature of this shear localization and
single curve. However, significant deviations from shear fracture is a fracture surface consisting of fine,
this behavior can occur for nonproportional histor- closely spaced, submicron dimples whose widths are
ies, with plastic flow localization playing an im- much larger than their depths. The last region to

---I $i~W.lETRIC DEFORMATION

_ FIA~o~~TRAIN DEFORMATION

- FAILURE LOCUS
STRESS STATE RANGE IN
? 7 DEFORMATION HISTORY

Fig. 21. (a) Failure locus for axisymmetric and plane strain tensile specimens reported by Hancock and
Brown for the steel 50D tested in the longitudinal transverse orientation [125]. (Reprinted with permission
from J. Mech. Phys. Solids 31, 1 (1983). Pergamon Journals Ltd.) (b) Failure locus found by Needleman
and Tvergaard assuming failure occurs when the void volume fraction (f) reaches a critical value (f,).
This is for a material in which voids nucleate at strain of 0.3 [150]. (Reprinted with permission from
J. Mech. Phys. Solids 32, 461 (1984). Pergamon Journals Ltd.)
 Ductile fracture 1059

Fig. 22. The fracture surface of a plane strain tensile specimen of an ultra-high strength steel (a). The
fracture surfaces at the edge (b) and the center (c) are compared at higher magnification (199).

0.40 -
fracture, the center, exhibits a fracture surface consis-
n HP9-4-20+Ti,Al640AO ting of larger dimples (Fig. 22). As shown in Fig. 23,
: - 0 HP9-4-20 640AO
?
the area of the fine dimpled region decreases as the
8 HP9-4-20 640-565
plane strain ductility increases. For those more inter-
2
E 0.30 -
ested in the micromechanisms of fracture of these
specimens, reports by Hahn et al. [161] and by
5
Clausing [ 1621 are useful.
$ Before ending this section, we would like to men-
2 0.20
tion two papers which provide information as to the
?
effect of inclusion volume fraction on plane strain
Z
0 tensile ductility. Speich and Spitzig investigated the
L : -\\--
0.10 influence of inclusion volume fraction on the smooth
2 axisymmetric and plane strain tensile ductilities for
;3 4340 steel tempered to several strength levels [ 1301. As
% shown in Fig. 24, neither measure of ductility is
I I I I markedly influenced by inclusion volume fraction at
0.00 0.10 0.20 0.30 040
the highest strength level. A dependence on volume
EFFECTIVE PLANE STRAIN DUCTILITY fraction appears only as the strength is reduced.
Fig. 23. A plot of the area of fine dimples as a function of Spitzig subsequently investigated the influence of
plane strain ductility for three ultra-high strength steels. inclusion volume fraction and specimen orientation
1060 W. M. GARRISONJR and N. R. MOODY

(a) I2

r \.
0 ESR (O.O02%S)
0 u~0.003-0.022%.51
1

0
0
ESR (0002XS)
U (0.003-0.022968)

A RE 10.003-0.0248s. r0 RE (0.003-0.024%5,
RARE- EARTH 1 RARE- EARTH 1

9x) MPo (135 ksi)

1210 MPo (175 ksi)

1410 MPo (205 kai)

‘;i%%i-,
14lOMPo(205kri)
0
A * 0
0 ” / n

1960MPo (265ksi) 1960 hiPo (265 ksi)

I I 1

0’ 0.04 0.08 U.IZ 0.04 0.08 0.12

VOLUME PERCENT INCLUSIONS VOLUME PERCENT INCLUSIONS

Fig. 24. (a) Axisymmetric and (b) plane strain tensile ductilities plotted as a function of inclusion volume
fraction for 4340 steel tempered to the indicated strength levels [130].

on the smooth axisymmetric and plane strain tensile If one begins to deform a specimen containing a
ductilitiesfor a low strength (250 MPa yield strength) sharp crack, linear elastic analysis predicts there will
steel [1601. Figure 25 shows that by using an inclusion be an infinite stress at the crack tip. However, the
projected area per unit volume or a mean local stress at the crack tip does not go to infinity because
inclusion area fraction both the transverse and short plastic deformation occurs, resulting in finite stresses
transverse properties can be placed on the same at the crack tip. The plastic deformation at the crack
curve. tip also has the effect of blunting the crack tip, as
shown in Fig. 26. As we will see crack tip blunting
1II.D Fracture initiation toughness is critical to our understanding fracture initiation
The plane strain fracture initiation toughness, K,=, toughness. The blunting process can be quantified by
is important because it is regarded as a material the crack tip opening displacement, 6, shown in Fig.
property and can be used to predict when fracture 26. The crack tip opening displacement can also be
will occur when a material contains a flaw of a given used as a measure of fracture initiation toughness. In
size (a) and is subjected to known nominal stresses this approach, fracture will initiate when crack tip
(a). Fracture will occur when K,c = K(a, a), where K blunting stops and 6 reaches a critical value (~5,~).
is the stress intensity factor for a given part or 6,c, K, and J,= are related by
specimen containing a crack.
Measurements of K,c are made using specimens
containing very sharp fatigue pre-cracks [163, 1641.
In addition, there are constraints on specimen dimen- J,c = K,,-( 1 - v *)/E (plane strain),
sions and the crack length. These limitations make
it often difficult to obtain or test specimens of the size where a,, is the flow stress and d, is a function of the
required to obtain valid K,= values for materials of ratio a,/E and the work hardening exponent n. It
high toughness. This can be overcome by introducing should be noted that blunting does not necessarily
the J-integral approach [164-1661 in which fracture result in the semicircular crack tip shown in Fig. 26.
initiation occurs when the J integral achieves a Unique solutions for blunting behavior are not
critical value, Jlc. As long as the J,c test is done possible and the blunting profiles shown in Fig. 27
under conditions of small-scale yielding, one can use are among those suggested [168] and observed. The
JIGto calculate the fracture toughness (K,c) one shape of the blunting crack tip is important because
would obtain if sufficiently large samples had been the strain distribution and rate of void growth ahead
tested. These quantities are related by [167] of the crack tip apparently change as the shape is
changed.
The levels of stress, strain and the stress state
Jlc = !$ (1 - v*) (plane strain).
(triaxiality) are important factors if fracture ini-
 Ductile fracture 1061

tiation occurs by ductile processes. The plastic strain (4

1.6
and the stress in the plane ahead of the crack are
0 . (0.004% Sl
plotted as a function of X,/6 in Fig. 28, where X0 is
0 . (0.021% Sl
the distance from the initially sharp crack tip. These A 4 ~O.O26%SI
0 0 A Transverse specimens
results were first derived by Rice and Johnson [169]. l . A Through-thickness specimens
Large scale deformation (blunting) at the crack tip

(4
1.5 -

1.4 -

1.3 -

1.2 -

1.1 - 02 -

1.0 - -I
.c_
0 1 2 3 4 5 6 7
0
I_ 09 - Local area fraction (%I
‘,
F 0.6 -
Fig. 25. The transverse and through thickness tensile duc-
tilities of a mild steel containing three sulfur levels plotted
z:: 0.7-
as a function of (a) inclusion volume fraction; (b) inclusion
; projected area per unit volume, A,, on a plane perpendic-
0.6 -
ular to the tensile axis; and (c) mean local area fraction of
sulfide inclusions on a plane perpendicular to the tensile axis
05-
[160]. (Reprinted with permission from Acta Metall. 33, 175
0.4 -
(1985). Pergamon Journals Ltd.)

0.3 -

0 . LO.004 96 S)
results in large plastic strains ahead of the crack tip.
0.2 - II n (0.021% S) The strain decreases to a low value at a distance of
A A (0.0269. 91
0 0 A Tronrverse
about 2 6. The stress increases with X,/6 until X,, is
0.1 -
l ’ h Through--thickness about 2 6; the stress remains high until the end of the
I I I I I I I I I plastic zone and then decreases. The plastic zone
0 0.02 0.040.060.06 0.10 0.12 0.14 0.16 0.18

Volume fraction of inclusions (%)

6 IC

.6 , , , , , , , , , , ,

14_

1.2 -

CRACK TIP
1.0 -
.G BLUNTS
e
z

8
2 o.e-
1
t
F
LJ_ 06-

0.4

0.2

0
_

-
.

0.4
0.
q .
A.
o q A Transverse
. A

0.6
(0.026
(0.021
(0.004

Through-

1.2 1.6
% S)
% S)
% S)

thickness

2.02.4 2.6 3.2 3.64.04.4 4.6

3 IC
ECTURE
IC

Fig. 26. A schematic diagram of the blunting process ahead

of an initially sharp crack. Fracture will occur when S
LI, (mm-‘) reaches a crItIca value a,=.
1062 JR and N. R. MWXIY
W. M. GARRISON

(4 +
SMOOTHLY 6-
BLUNTING
.N=.2

N=.I

b _)I 1 J
FLAT-NOSED
OR SQUARELY
BLUNTING
--- co/E = .0050
N=O

- - co/E = .0075
- ALL CASES, N =0
0 I I I 0 I
SHARP NOSE 0.5 1.0 1.5 2.0 2.5
BLUNTING x/ 8,

INITIAL BLUNTED CRACK

Fig. 27. Possible modes of nonsmooth crack blunting sug- SHARP CRACK
gested by McMeeking [168]. (Reprinted with permission
from J. eng. Mars. Tech. The American Society of Mechan-
ical Engineers, 345 East 47 Street, NY 10017 (1977).)

extends much further than 2 6. For a non hardening c:’ 1.0

material, the ratio (a,,/~,,) is equal to the ratio of a,,,/~? FULLY PLASTIC
plus l/,/3. Thus there are not only large gradients in SMALL SCALE YIELDING
0.5
both stress and plastic strain but in the stress state
ahead of the crack tip as shown in Fig. 29. All
modeling which hopes to relate 6,c, K,c or J,c to
0.5 1.0 1.5 2.0 2.5
mechanical properties which can be directly meas-
ured or to parameters which quantify microstructure X/6+

must account for the very steep gradients, especially Fig. 28. The stress in the vertical direction (u,,) normalized
the gradient in strain. Such models are numerous, by the yield strength (uO)and the strain ahead of a smoothly
and those by Krafft [170], Hahn and Rosenfield blunting notch in the plane of the crack as a function of X/6
where X is the distance from the initial crack tip and 6 is the
[171], Rice and Johnson [169] and the so-called crack tip opening displacement [168, 1691.
characteristic distance model will be discussed.
When Krafft [170] developed his model, only the
elastic stress distribution ahead of a sharp crack was The strain at which the tensile instability occurred
available. From this he argued that the strain should was taken equal to the work hardening exponent n
decay rapidly ahead of the crack tip. In his approach, and the position of the instability was taken as dT.
he assumed fracture initiation would occur when a The region between the crack tip and dT was termed
tensile instability was reached ahead of the crack tip. the process zone. Krafft calculated d, by requiring

Fig. 29. The equivalent plastic strain and level of triaxiality (~,,,/a) as a function of X/6 [180].
 Ductile fracture 1063

t----XC------l
--I &2Ro
> h
”

-n_ VOID SITE ENVELOPED BY

LARGE STRAIN REGION

FURTHER GROWTH,
LOCALIZED FLOW BEGINS

FINAL COALESCENCE

SIC”X,

Fig. 30. Fracture initiation as postulated by Rice and Johnson. An inclusion or pre-existing void of radius
&is positioned a distance X0 ahead of the crack tip. As the inclusion is enveloped by the strain field ahead
of the blunting crack the void grows. Fracture is taken to occur when the ligament between the blunted
crack tip equals the vertical void radius [169].

agreement between measured values of K, and his This model for fracture initiation toughness requires
model. He observed that dr could be much larger some critical fracture strain and some measure of
than the spacing of void nucleating particles, thus he distance. It suggests, as did the model of Krafft, that
did not suggest dT would be the same as inclusion K,c should increase with n, the work hardening
spacing. exponent.
Hahn and Rosenfield, in a 1968 paper, before the In 1972 Rice and Johnson [169] made the first
work of Rice and Johnson, assumed a linear strain explicit connection between 6,c and microstructure.
distribution in the plastic zone [ 1711. This approach Moreover, they obtained their results in context of
predicts the continuum equations for void growth [92,93].
They considered a pre-existing void, or particle
K, = (2Ea,$ 1)“2, weakly bound to the matrix, of radius &,, a distance
X0 ahead of an initially sharp crack, as shown in Fig.
where a,, is the yield strength, 1 is the size of the 30. They calculated numerically the void size as the
plastic zone and ZF is the strain at the crack tip at crack tip blunted. They then assumed fracture to
initiation of fracture. They assumed E/’ could be occur when the ligament between the blunted crack
taken as one-third the tensile strain as measured tip and the void equalled the vertical void radius.
from smooth axisymmetric tensile specimens. They Their numerical results are shown in Fig. 31, where
also found that I, determined by requiring agreement 6,c/X,, is plotted as function of X,/Z&. This allows 6,c
with theory and measured K,c could be expressed as to be written as
a function of n:

I = 25.4 x 10-+z2 + 0.0005)(m).

where F is a weakly varying function of (X0/&). In
Thus their semi-empirical expression for K,c is writ- applying their model they suggested regarding X0 as
ten as the average 3D nearest neighbor distance between
inclusions; thus X0 = 0.89&,f-“3, where &, is the
._ = 512Eu..Pf(n2+0.0005)
K,,- _ ,,. x 10-31”2(MNm -3’2). average inclusion radius and f is the inclusion volume
1064 W. M. GARRISON JR and N. R. MOODY

t 8,/X, AT FRACTURE 3 I I 1 I ’
smallscaleyleldlng
thcoretlcal
C”T”C,

alloy steel castings

6500C. I
effect of
tcmpumg
En25 -
550VA

I 1 I
Xo’Ro high-strength weldmctok
t”;:$: ,
0 I 1 1 I,
IO 20 30
0 100 150 200
NON METAL:: INCLUSION SPACING/INCLUSION RADIUS
Fig. 31. The quantity 6,c/X,, as determined plotted as a
function X0/R, by Rice and Johnson [169]. *,k

Fig. 33. Critical crack tip opening displacement divided by

inclusion spacing X,, plotted as a function of X,/E0 for
fraction. Therefore F(X,/&) in their numerical re- several steels and the theoretical curves of Rice and Johnson
sults would be F(f-I/‘). Hahn et al. [25] used the [194]. (Reprinted with permission from Micromechanirms
of Fibrous Crack Extension in Engineering Alloys. Metal
results of Rice and Johnson to show that one would Science (1980).)
expect K,C to scale as

As Fig. 33 shows, the Rice and Johnson model is

not obeyed by most systems, partly because it as-
Plotting ln[K,c/,/(u,E)] as a function of In(f) for sumes a specific failure criteria which is independent
several steels and aluminum alloys, they found that of strength level and fine-scale microstructure.
K,C did scale as f-‘/6 for materials of the same value The plot by Hahn et al. [25] of the data of Birkle
of a,/E (Fig. 32). They noted that this equation et al. [172] suggests the Rice and Johnson approach
predicts that the fracture toughness will increase with is correct but incomplete (Fig. 34). These data show
increasing yield strength which must be incorrect, as that for a given tempering treatment 6,c does scale
toughness normally decreases with increasing with X,,, but the rate of increase of 6,c with 1,
strength. Hahn et al. suggested that could be because increases as the yield strength is decreased by tem-
at “higher strength levels, the attendant lower rates pering. What this graph does suggest is that for steels
of work hardening favor strain localization and pro- of otherwise identical microstructures the 6,c will
mote plastic instability within the ‘intensely strained scale as X,,. There is a large body of data by Knott
region’, thereby promoting the coalescence of voids” and others which reinforce this conclusion
1251. [27, 172-1771.

uy /E x IO-’
L

13 4340 steel
~2OOM(l8N1)steel
Kf04SC-NI-Cr-MO sfeel 0045 C-NI-Cr-MO steel
0 300 M steel U 2014-T6 aluminum
n 2024-T8Sl aluminum

011 I I I L I I I 1 I

001 01 IO IO
f, , percent

Fig. 32. Intluence of inclusion volume fraction on the fracture toughness of ultra-high strength steels of
comparable yield strengths [25]. (Reproduced with permission from the Annual Reuiew of Maferials
Science, Vol. 2. 0 1972 by Annual Reviews Inc.)
 Ductile fracture 1065

Tempering Yield Stress

Temp. (“C) (Mn/m’)
1 205 1610-1635 / A
\ 315 I585
I 370 1505-1540
1 425 1455-1500
L 480 1380-1395

5=x&C ot fracture

Fig. 35. Schematic diagram of the critical strain controlled

or characteristic distance model for microvoid coalescence.
The equivalent plastic strain, 5, has been plotted as a
function of X/6, where X is the distance from the initial
crack tip and 6 is the crack tip opening displacement. l/’ is
0
the critical strain and 1: the characteristic fracture distance
[196].
Sulphide Spacing,pm

Fig. 34. The critical crack tip opening displacement as a

function of inclusion spacing in a Ni-Cr-Mo steel. (Data notched tensile specimens, the plastic strain ahead of
from [172] as plotted in [25].) the crack tip is plotted as function of triaxiality (i.e.
eliminating S/X in Fig. 29); on this same graph the
fracture strains of the notched tensile specimens are
The Rice and Johnson model has been criticized plotted as a function of stress state. The intersection
because it predicts the fracture toughness increases of these curves determines the actual fracture strain
with inclusion size. This results because an increase in which then determines I,*. Lee et al. in their studies
inclusion size at fixed volume fraction increases X0, of ultra high strength steels took lo+ to be a fixed
and hence 6,c if the fine-scale microstructure is not multiple of a particle spacing and found good agree-
changed. The data of Cox and Low [9] do not support ment with calculated and experimental 6,c using for
the notion that 6,c will increase with X0. However, it E,* the effective plane strain tensile ductility [153].
is possible that, in the case of the maraging steel this Ritchie et al. in their study of low strength pressure
was because reducing the size of the Ti(C,N) particles vessel steels used the approach outlined above for
made them more resistant to fracture and that reduc- notched tensile specimens to determine c/* and found
ing sulfur content in 4340 reduced not only the 1: was a multiple, a, of X0, the inclusion spacing, but
average inclusion size but the tendency to form that a was dependent on the material [180].
stringers. As the toughness specimens used by Cox The conceptual success of this model has led to
and Low were oriented with the crack direction a general conviction that any model for fracture
parallel to the rolling direction, a large improvement initiation toughness must incorporate some charac-
in toughness would be possible simply by removing teristic distance and some measure of ductility. How-
the stringers. Two studies [176,199] have shown that ever, there are numerous drawbacks to this model.
increasing X0 by increasing the average inclusion size First, it is difficult to interpret on a microstructural
at a fixed inclusion volume fraction does increase basis; in fact as it is presently constituted the model
toughness. is unable to predict how 6,c will scale with particle
A generalization of the above three approaches to spacing. Second, it does not incorporate the micro-
modeling fracture initiation toughness is the modified mechanisms of the fracture process. Third, there are
strain criterion approach or characteristic distance difficulties with using tensile specimens to estimate a
model [127, 178-1831. In this approach the equivalent critical fracture strain: (1) a tensile specimen does not
plastic strain (5) must exceed a critical strain (CT ) incorporate the very sharp strain gradient ahead of
over some characteristic distance Z$ as shown sche- the crack tip, and as a result widely spaced micro-
matically in Fig. 35. This approach implies structural features could influence tensile ductility but
have little effect on 6,c; (2) neither plane strain nor
notched tensile specimens reflect the changing stress
state experienced by a particular point ahead of the
In applying this model, Z/* has been taken as the plane blunting crack tip.
strain tensile ductility or it has been obtained using In an effort to maintain the importance of inclusion
tensile specimens of varying notch acuity. In using the spacing in the fracture initiation process suggested by
1066 W. M. GARRS~NJR and N. R. MOODY

Rice and Johnson’s model, Garrison proposed an

approach to initiation toughness for a material con-
taining a unimodal dispersion of spherical particles
weakly bound to the matrix [184]. These are termed
the primary particles. Such particles can be regarded
UNDEFORMED CONFIG. DEFORMED CONFIG.
as pre-existing voids which will grow when the par-
ticles enter the highly localized strain field ahead of
an initially sharp crack [169]. It is possible that void
sheet coalescence [6,9] might result in the coalescence
of the voids nucleated at the primary particles and
their linking to the crack tip prior to their linking by
a necking down of the ligament between them. Sup-
pose the system contains, in addition to the primary
2
particles, a dispersion of particles strongly bound to ._
o4 -
the matrix (secondary particles). By the process of
void sheet coalescence, the growing voids nucleated <
a* 3-
at the primary particles would link up via the voids
nucleated at the secondary particles. Therefore in this
model system one would expect the toughness to
depend on the extent of growth of the voids nucleated
at the primary particles and probably on the spacing
of these particles as suggested by both the Rice and
Johnson model [169] for ductile fracture and some
interpretations of experimental studies of ductile frac- Fig. 36. McMeeking’s calculations of R,/& and Rx/& as
ture [172-1771. This approach is consistent with a function of S/X,, where 6 is the crack tip opening
McMeeking’s discussion of the Rice and Johnson displacement and X0is the position of a void of initial radius
model [185]. &. McMeeking’s R, is R, in our terminology. R, should be
given approximately by the average of R, and Rx,where R,
McMeeking’s numerical calculations of void is the average of R, and R, [168].
growth provide a basis for introducing void growth
into a model of ductile fracture. Using the Rice and
Tracey model for void growth [92], McMeeking wherefis sulfide volume fraction and &, the average
obtained a graphical relationship between void size sulfide radius. The sulfide size distribution and
and S/X, [168, 1851, where 6 is the crack tip opening (RY/R,)are plotted as a function of R, in Fig. 37 for
displacement and X0 is the distance of a primary one of the steels studied. The value of RJR,
particle of radius RI from the initial crack tip. His evaluated at & [(R,/&)(,Jhas been used to charac-
results (Fig. 36) indicate that for a smoothly blunting terize the extent of void growth. ~5,~is plotted as a
crack tip, the average of R,/R,and Rx/R,, where R, function of [&(R,/R,)1,J in Fig. 38 for several steels
and R, are regarded as the maximum and minimum and the results are consistent with the proposed
radii of the void in the plane of fracture, should be scaling. It should be noted that this approach is
approximately linear in 6/X,. In what follows, the similar to that proposed by Bates [ 1791 and almost
average of R, and Rx will be called R,, the void identical to that of Thompson [200].
radius; this is reasonable for most ductile fractures in In addition, it appears this approach can, in some
which microvoids are reasonably equiaxed. Follow- situations, be used to assess the effects of two particle
ing Rice and Johnson [169] and identifying X0 as the distributions on initiation toughness. Two particle
average 3D nearest neighbor distance between pri- distributions, a narrow size distribution of small
mary particles, one would expect the critical crack tip chloride inclusions and a broad size distribution of
opening displacement to scale as large pores, are common in blended elemental, pow-
der processed titanium alloys. Three billets of powder
6,c - &@,I&). processed Ti lOV-2Fe3Al were studied to determine
This approach predicts the same dependence on X0, the effect of these two types of defects on fracture
toughness [188]. One billet had both pores and inclu-
the primary particle spacing, as the Rice and Johnson
sions, a second billet was hot isostatically pressed to
model, but (RJR,) introduces a measure of ductility
lacking in that model. leave just the inclusions, and a third billet, produced
This model has been tested using several ultra high from very low chloride powder, had just pores. The
strength steels [176, 177, 184, 186, 187 for which the effect of each type of defect on toughness was pro-
primary particles are spherical sulfides. The average nounced as shown in Table 4, where 6,= increased
3D nearest neighbor distance between sulfides, X0, from 5 pm in the billet with pores and inclusions to
has been calculated from 25 pm in the billet with just pores.
The role of each type of defect on a,, was examined
through the effect of particle spacing and void growth
 Ductile fracture 1067

at failure with the following relation between 6,c, the

average particle spacing X,,, and the amount of void
growth (R”/R,).

where R, is void radius at fracture and R, is the initial

void radius.? The amount of void growth to final
failure corresponding to the average inclusion, aver-
age pore and average combined inclusion and pore
sizes was determined from fracture surface mea-
surements. The average particle sizes and 3D spacings
1
!;
INCLUSION RADIUS. /tm
were determined by computer-controlled SEM digital
analysis of metallographically prepared samples.
The effect of particle spacing and corresponding
void growth at failure on 6,c is shown in Fig. 39. The
data in Table 4 and Fig. 39 indicate that both
parameters influence 6,c in these samples. It is also
apparent that in single particle distribution samples
those particles control 6,c. It is not directly apparent
from the data for the billet containing both pores and
inclusions if inclusions, pores, or a combination of
both defects control ~5,~.However, examination of
the fracture surface shows void growth occurs around
each type of particle and suggests that both inclusions
and pores control fracture toughness of samples -6 02 04 0.; 0.8 i 12 la lb 1.; 2 i
containing both types of defects. INCLUSION RADIUS, pm
However, this approach has definite limitations. As
shown by the data in Fig. 38, all of the materials Fig. 37. The size distribution of manganese sulfides found
examined to that point obeyed the scaling on the fracture surface of a K,= specimen and the ratio of
6K - XoUW&)l,. On investigating the fracture void to manganese sulfide size plotted as a function of
sulfide size. The extent of void growth associated with a
particular sulfide of radius R, is RJR,. The void radius, R,,
t Note that here the proposed scaling is 6,c- X0 is taken to be one-half the average or the maximum and
ln(R,/R,) rather than 6,c - &(R,/R,). minimum void diameters [176].

60- *
AFl410
1200v (AGED so-3 -
0
50-

40-
6
$r% 1050~C
30- 0

OAF 1410 (AGED 425-C)

20- l
S4OOC o BASE + Ni + Si

0
BASE+ Ni.
IO- BASE + Ni + Si +MoV
BASE *
BASE + Si
1 I I I
OO 25 50 75 100
X,(R,/R,)Ia, &ml

Fig. 38. The critical crack tip opening displacement, a,,-, plotted as a function of XO(RY/R,)Jh, where
X0 is the sulfide spacing and (RJR,) 1R.is the ratio of void to sulfide size evaluated at R.. The dark circles
(a) refer to steels of yield strength greater than 1500 MN/mZ. The open circles (0) refer to HP9-4-20
in the as-quenched condition for three different austenitizing temperatures [176,177,184,186,187].
1068 W. M. GARRISONJR and N. R. MCKIDY

Table 4. Powder processed blended elemental Ti-IOV-2Fe-3Al material properties

Yield
strength KK
Condition (MN rne2) 4 (MN m-3’*)
As-sintered
(pores and inclusions 883 0.021 29.1 5
Hot isostatically pressed
(inclusions only 928 0.24 48.3 12.6
As-sintered
(pores only) 786 0.058 62.9 25

Modeling data
Pores + inclusions Pores only Inclusions only

X0
Condition Olm) (&/&)I, 2) (&/&)I, 2) (&I&) Ifi
As-sintered
(pores and inclusions) 2.66 2.51 31 1.21 6.78 3.22
Hot isostatically pressed
(inclusions only) - - 6 3.61
As-sintered
(pores only) - - 31 1.99 -

I
30
I
In generalizing the approach at least two factors
8=X01n(R,IR,1
should be considered. First, the approach presently
assumes that the rate of void growth with respect to
/ 6 is a constant. It is possible this is not the case. For
20 I /’ example, it may be that the rate of void growth
changes when the blunting behavior changes. In the
as-quenched condition when n = 0.18, the profile of
2
IO
/
i’ .
0 PORES

INCLUSIONS
the blunted crack is smooth. When the HP9-4-20 is
tempered at 565”C, n = 0.08 and the crack tip profile
is square. It is possible the rate of void growth ahead
/ 0 PORES AN0 INCLUSIONS
o/ of a smoothly blunting crack tip is greater than the
/ rates of void growth around a crack tip blunting
! I I
OO IO 20 30 40 to a square profile. Second, it is possible that in
X,ln(R,/R,), pm materials of high ductility the voids which coalesce
with the crack tip are not particle nucleated voids, but
Fig. 39. A plot of 6,c as a function of X,ln(R,/R,) for the are super-voids resulting from the coalescence of a
three powder processed blended elemental titanium alloys
described in Table 4 [188]. number of particle nucleated voids.
The foregoing models all address fracture initiation
toughness in terms of a critical distance (which may
properties of a second heat of HP9-4-20 steel it was be in some systems linearly related to a particle
found that in the as-quenched condition its behavior spacing) and some measure of ductility. However,
was consistent with previous data. However, on none of these models describe with any completeness
tempering this heat at 200°C and 565°C S,c increases the fracture process. Fracture initiation can appar-
substantially but there is no increase in (R,/R,)I%, as ently occur via several processes, three of which are
shown in Table 5. Thus, there may be limits in using shown in Fig. 40. First is the internal necking used in
(RY/R,)lRO as a measure of the critical fracture the Rice and Johnson model [169]. This has been
strain; on the other hand, it is possible the approach observed by Ritchie et al. in studies of a low carbon
needs only to be modified. ferritic machining steel [193]. Aravas and McMeeking

Table 5. Mechanical properties and inclusion and void growth data for HP9-4-20
as a function of tempering
Tempering Flow Work
temperatures stress hardening 6,c
(“C) (MN me2) exponent b m) (R,IR,)l,
As-quenched 1398 0.18 15.5 2.50 10.3
200 1394 0.10 41.5 2.50 9.3
565 1313 0.08 91.1 2.50 10.1
 Ductile fracture 1069

tip. McMeeking [185] considered the same geometry

as Rice and Johnson and computed void growth as
a a function of 6/X, for work hardening exponents of
n-0 and n = 0.10 at positions directly in front of the
crack tip and at 45” to the crack tip. He found the
extent of void growth was not greatly influenced by
a modest degree of work hardening nor did the
angular position of the particle significantly influence
the extent of void growth for a given S/X,. Aravas
and McMeeking [ 1891later showed, examining cylin-
drical voids rather than spherical voids, that there is
a pronounced interaction between voids and the
blunted crack tip. They compared these results to
those obtained by using McClintock’s [86] equations
for growth of cylindrical voids and found that the
strong crack tipvoid interaction provided for in their
calculations result in void growth rates 6 times faster.
It is reasonable to expect that incorporating void-
Fig. 40. Idealizations of three possible fracture modes crack tip interactions will also result in much faster
during fracture initiation: (a) internal necking, (b) lccal- growth rates than McMeeking found for spherical
ization of shear along slip line spiral and (c) localization voids [185]. In addition, Aravas and McMeeking
of shear between crack tip and void not directly ahead of [ 1891found, on providing for the interaction between
crack tip [194]. (Reprinted with permission from Micro- the crack tip and the growing void, that the rate of
mechanisms of Fibrous Crack Extension in Engineering
Alloys. Metal Science (1980)) void growth decreased substantially with an increase
in work hardening capacity. Thus, on the basis of
void growth, high toughness should be favored by
in their numerical analyses have apparently achieved high work hardening capacity.
this process of internal necking [189]. Second, it has While void growth has been the subject of several
been observed that initiation can occur by the for- continuum investigations, there have been no experi-
mation of a “narrow” shear crack forming at the mental tests of their results. In addition, two aspects
blunted crack tip. This shear crack follows curved of void growth have received little or no theoretical
paths corresponding to the slip lines around a circular or experimental attention; these are the influence of
notch. These shear cracks link up with a growing void crack tip blunting geometry on void growth and the
located somewhere along the slipline curve. This has formation and subsequent growth of super-voids,
been observed in both aluminum alloys, quenched voids formed by the coalescence of several particle
and tempered steels and prestrained mild steels [ 1941. nucleated voids.
In addition, it is possible that fracture can occur by In obtaining the strains just ahead of a blunting
strain localization between the crack tip and a void crack, Rice and Johnson [169] assumed a smoothly
growing above or below the plane of the crack. Such blunted crack tip. Such profiles are often observed,
fractures would produce a zig-zag fracture. Such and are normally assumed in modeling crack tip
fractures have been predicted by the continuum behavior as in void growth calculations of void
calculations of Needleman and Tvergaard [195] in growth by McMeeking [168,185] and Aravas and
which they considered an array of large inclusions McMeeking [189]. However, non-smoothly blunted
and an additional array of smaller more closely crack tip profiles have been observed. McMeeking
spaced particles strongly bonded to the matrix. In this [ 1681considered two cases of non-smoothly blunting
case the strain localization was induced solely by the notches, shown in Fig. 27. He calculated both the
formation of voids at the small particles. These strains ahead of such crack tips and the rates of void
observations, and experimental results discussed growth directly ahead of the crack tip. As shown in
earlier, indicate that void nucleation, growth and Fig. 41, changes in crack tip blunting behavior
coalescence, particle spacing and shear localization greatly alter the strain and void growth rate just
are involved in the fracture process during fracture ahead of the crack tip. The rate of void growth goes
initiation. However, the relative importance of to zero directly ahead of the non-smoothly blunting
these various factors and the extent to which they cracks. It appears that low work hardening [190] and
are inter-related are not known. Of these factors low plastic constraint (triaxiality) [191] will favor
void growth and particle spacing have received the non-smooth blunting. Moreover, it is not known if
greatest attention. blunting behavior changes as 6 increases [192]. That
It is clear from the continuum descriptions of void the rate of void growth directly ahead of non-
growth that the high triaxiality at the crack tip will smoothly blunting crack tips may go to zero could
result in rapid void growth as the particle is en- have important implications in modeling fracture
veloped into regions of high strain close to the crack initiation. What is needed are calculations to deter-
1070 W. M. GARRISON JR and N. R. MOODY

- SMOOTHLY BLUNTED

I
tr ------ FLAT NOSED CASE
SMOOTHLY BLUNTED RY%
EY __---_ --- SHARP NOSED CASE
2.0 FLAT NOSED CASE Rx/R, 6
--_ SHARP NOSED CASE RY
1 ---- ----R
I Y
5
/ /

4-

3-

OO 0.5 1.0 1.5 2.0 25 I 0.2

’ 0.6
I I IOI , 14
1 , 16
I I

x18, 81%7

Fig. 41, A comparison of the strains and the extents of void growth ahead of a blunting crack for the
three indicated blunting geometries [168].

mine whether the rate of void growth above and continuum expressions for void growth provide a
below the plane of the non-smoothly blunting cracks qualitative explanation for experimental results, no-
are smaller or greater than the rate of void growth tably the effects of stress state and particle volume
directly ahead of a smoothly blunting crack. fraction on ductility.
While the importance of particle spacing has been Plane strain tensile specimens are, unlike axisym-
established for certain systems (primarily ultra high metric tensile specimens, prone to shear localization.
strength steels) this important clue to the nature of Cracks initiate at the surface at shear offsets and
the fracture process has yet to be exploited in alloy propagate inward by void sheet formation. Like
design or to be examined in detail. In terms of alloy axisymmetric tensile specimens, the ductility of plane
design it is clear that substantial improvements in strain tensile specimens can be sensitive to particle
toughness can be realized, even at constant inclusion volume fraction. Hancock and Brown [125] suggest
volume fraction, by maximizing the inclusion spacing that the ductility of plane strain tensile specimens is
by making the inclusions as large as possible. For primarily determined by stress-state and thus can be
example, the toughness of the steel AF1410 aged at rationalized in terms of the effect of stress state on the
510°C can be increased from 28 to 66pm by in- void growth rate. Given the tendency of these speci-
creasing the inclusion spacing from 2.3 to 7.6 pm by mens to exhibit flow localization, this conclusion
increasing the average inclusion radius from 0.18 to should be regarded as tentative [150].
0.64pm [199]. If the radius and spacing could be Fracture initiation toughness, unlike tensile duc-
increased to 1.92 and 23 pm respectively, 6,c should tility, is significantly influenced by particle spacing.
increase to almost 200 pm. While the Rice and John- This dependence on particle spacing arises from the
son calculations and the results of Birkle et al. suggest steep gradients in strain, stress, and stress state ahead
6,: will scale as X0 as long as the fine-scale micro- of the crack tip; such gradients are not observed in
structure is constant, there is really no fundamental tensile specimens. Because of this dependence on
understanding of why 6,c (at least over some range) paerticle spacing it is possible to manipulate defects
should be linear in X,. in such a way that two otherwise identical materials
will rank very differently in terms of tensile ductility
1II.E Mechanical properties: summary and fracture initiation toughness.
Fracture of smooth and notched axisymmetric
tensile specimens begins by the coalescence of several
IV. CONCLUSIONS
voids in the center of the specimen. Fracture then
occurs by the propagation of this central crack. At While the literature on ductile fracture is exhaus-
this point it is impossible to predict tensile ductilities. tive it is clear that many fundamental issues remain
However, in the absence of nucleation strains, the unresolved. Such issues include basic understanding
 Ductile fracture 1071

of the three stages of ductile fracture and our ability spherical particles of a reasonably low volume frac-
to predict measures of ductility and toughness. tion. Experience suggests that increasing the nucle-
Much additional work remains before the nu- ation strain by reducing particle size can also improve
cleation event is understood. For example, while toughness. Therefore, if indeed void nucleation be-
stress state, strength of the interfacial bond between comes more difficult as particle size decreases there
the particle and matrix, matrix flow properties, and will be, at a fixed particle volume fraction, a trade-off
particle volume fraction all appear to be important in between size and spacing. There have been no
determining the nucleation strain for nucleation by attempts to incorporate the nucleation event in
interfacial decohesion, there is no agreement, at least modeling fracture initiation toughness. Other funda-
in terms of modeling, as to the role of particle size on mental problems include explaining the effect of
the nucleation event. Separating nucleation from void particle spacing and how fine-scale microstructure
growth appears to be a major experimental difficulty. influences toughness. Our belief is that the extent of
Our current qualitative understanding of ductile void growth is a convenient method of assessing
fracture is primarily based on continuum descriptions microstructural effects, but if this concept is to be
of the growth of isolated voids. Experimental tests extended the rates of void growth for various blunt-
suggest the McClintock equations for the growth of ing geometries and the possibility of super-void for-
cylindrical voids greatly underestimate the extent of mation at crack tips need to be investigated.
void growth. The Rice and Tracey equation for the Finally, the mechanics community has been ex-
growth of a spherical void appears to hold reasonably tremely active in modeling ductile fracture. Some of
well at small strains in materials containing a small these calculations can be tested. Others are presented
volume fraction of inclusions. In any case, it is in such a way that tests of the predictions and results
unreasonable to apply these continuum models to are difficult. It would be useful to the metallurgist if
large deformations of materials containing even small a greater effort was made to summarize these results
particle volume fractions. in approximate closed form solutions.
Current modeling is inadequate to predict tensile
ductilities. The Rice and Tracey equations for void
Acknowledgements-The authors appreciate the useful dis-
growth using actual void and particle size mea- cussions with R. Asaro, A. W. Thompson and R. Stoltz.
surements over-predict tensile ductility. Attention is Nor would the manuscript have been completed without the
currently being given to how non-uniform particle skill of Marge Cain in preparing the manuscript, or the
distributions influence ductility. But even if this is patience of the editors. The work was supported in part by
the Army Research Office (WMG) and the Department of
successful in the absence of nucleation strains, there
Energy (NRM).
remains the problem of incorporating the nucleation
event in modeling tensile ductility. Here there are
several problems. How does one select a measure of
nucleation which might reflect the influence of whole
REFERENCES
distribution of particles? Second, care must be taken
to incorporate in modeling only those particles which 1. Plateau J., Henry G. and Crussard C., Rev. Met. 54,
participate in the fracture process. By increasing 200 (1957).
particle-matrix cohesion, for example, one might 2. Henry J.,.Proc. Ass. Ado. Sci. 9, 102 (1855).
3. Ludwick P. Z., 2. Ver. dt. Ing. 1532 (1927).
reduce the effective volume fraction of void nucle- 4. Tipper C. F., Metallurgiu 39. 133 (1948).
ating particles. Finally, in terms of modeling, the void 5. Puttick K. E., Phil. Msg. 4, $64 (i959).’
growth equations, integrated from a nucleation strain 6. Rogers H. C.. Trans. AZME 218. 498 (1960).
to a fracture strain, suggest nucleation strains will 7. Crissard C., Plateau J. Tamhanbar R.,‘Henry G. and
Lajeunesse D., Fracture (Edited by Auerbach B. L.,
improve ductility. However, it is not clear if the strain
Felbeck D. R., Hahn G. T. and Thomas D. A.), p. 524.
to fructase will increase by the nucleation strain as the Technology Press of MIT and John Wiley, New York
rate of void growth after nucleation might be faster (1959).
than in the absence of nucleation strains. 8. Palmer I. G. and Smith G. C., Proc. AZME Co&
There are even greater difficulties in predicting Oxide Dispersion Strengthening, Bolton Landing, NY,
June 1966. Gordon & Breach, New York (1967).
fracture initiation toughness. In the absence of nu-
9. Cox T. B. and Low J. R., Metall. Trans. 5, 1457 (1974).
cleation strains it appears that 6,c will, for a tixed 10. Argon A. S. and Im J., Metall. Trans., 6A (4) 839
fine-scale microstructure, scale as the appropriate (1975).
particle spacing at least over some range of X0. 11. Liu C. T. and Gurland J., Trans. ASM, 61, 156 (1968).
Unfortunately, the selection of the appropriate par- 12. Broek D., Engng Fruct. Mech. 5, 55 (1973).
13. Greenfield M. A. and Margolin H., Metall. Trans., 3,
ticle spacing can only be determined on the basis of 2649 (1972).
experience and detailed experiments. However, this 14. Margolin H., Greenfield M. A. and Greenhut I..
does suggest that improvement in toughness can be Tita&m Science and Technology (Edited by Jaffee
achieved by reducing the particle volume fraction or R. I. and Burke H. M.). Vol. 3. D. 1709. Plenum Press.
New York (1973). ” ‘_
by making the inclusions as large as possible; it is
15. Rogers D. H., Titanium Science and Technology (Edi-
emphasized that this is our judgement and pertains ted by Jaffe R. I. and Burte W. M.), Vol. 3, p. 1719.
only to systems containing a unimodal distribution of Plenum Press, New York (1973).

PCS. 48,11--H
1072 W. M. GARRISON JR and N. R. MCODY

16. Gysler A., Lutjering G. and Gerold V., Acta metall. 22, (Edited by Taplin D. M. R.), Vol. 2, p. 343.
901 (1974). ICF4, Waterloo, Canada. Pergamon Press, New York
17. Williams J. C., Boyer R. R. and Blackburn M. J., (1978).
Electron Microfractography (Edited by Brothers A. J.) 55. Van Stone R. H. and Low J. R., Jr, NASA Tech. Rep.
ASTM STP 453, p. 215. ASTM, Philadelphia (1969). I-Ti. Carnegie Mellon Universitv. Pittsbureh (1973).
18. Asaro R. J., Acta metall. 27, 445 (1979). 56. Van Stone R H., Low J. R., Jr-and Shannon J. L:,
19. Pegel B., Schaubitz K., Burck P., Luft A., Brenner B. NASA Rep. CR-135120, Carnegie Mellon University,
and Reitzenstein W., Scripta metall. 13, 47 (1980). Pittsburgh (1977).
20. Baver R. W., Ph.D. Dissertation, University of Vir- 57. Nieh T. G. and Nix W. D., Scripta metall. 14, 365
ginia, Charlottesville, VA (1975). (1980).
21. Wilsdorf H. G. F.. Mater. Sci. Engng 59, 1 (1983). 58. Galda K. H., Munz D. and Schwalbe K.-H., Proc. 3rd
22. Rogers H. C., Ductility, p. 31. American Society for Int. Conf Titanium, May 1976, Moscow, published as
Metals, Metals Park, OH (1968). Titanium and Titanium Alloys (Edited by T. C. Will-
23. Van Stone R. H., Cox T. B., Low J. R. and Psioda iams and A. F. Belov), Vol. 1, p. 649. Plenum Press,
J. A., ht. Metals Reo. 30 (4) 157 (1985). New York (1982).
24. Rosenfield A. R., Metall. Rev. 13, 29 (1968). 59. Chesnutt J. C., Rhodes C. G. and Williams J. C.,
25. Hahn G. T., Kanninen M. F. and Rosenfield A. R., A. Fractography-Microscopic Cracking Processes (Edi-
Rev. Mater. Sci. 2, 381 (1972). ted by Beachemond C. D. and Warke W. R.), ASTM
26. Howard I. C. and Willoughby A. A., Advances in STP 600, p. 99. ASTM, Philadelphia (1976).
Fracture Mechanics (Edited by Chell G. G.), Vol. 2, p. 60. Moody N. R. and Greulich F. A., Scripta metall. 19,
39. Aonlied Science Publishers, London (1984). 1107 (1985).
27. Schwaibe K.-H., Engng Fract. Mech. 9, 795 (1977). 61. Moody N. R., Stoltz R. E. and Perra M. W., Corrosion
28. Tanaka J. P., Pampillo C. A. and Low J. R., Jr, Review Cracking (Edited bv Horl B. D.)., D. 1 54. ASM. Metals
of Developments in Plane Strain Fracture Toughness Park, cH‘(l986). .
Testing (Edited by Brown W. F., Jr), ASTM STP 463, 62. Moody N. R., Stoltz R. E. and Perra M. W., Metall.
p. 191. ASTM, Philadelphia (1970). Trans ISA, 1469 (1987).
29. Psioda J. A., Ph.D. thesis, Carnegie Mellon University, 63. Robertson I. M. and Birnbaum H. K., Acta metall. 34,
Pittsburgh (1977). 353 (1986).
30. Wilcox B. A. and Clauer A. H., Trans. AIME, 212,750 64. Argon A. S., Im J. and Safoglu R., Metall. Trans. 6A,
(1958). 825 (1975).
31. Gurland J., Acta metall. 20, 735 (1972). 65. Argon A. S. and Im J., Metall. Trans. 6A, 839 (1975)
32. French I. E. and Weinrich D. F., Scripta metall. 8, 87 66. Gurland J. and Plateau J., Trans. ASM 56,442 (1963).
(1974). 67. Laszlo F., J. Iron Steel Inst. 147, 173 (1943).
33. Coffin L. F. and Rogers H. C., Trans. ASM 60, 672 68. Laszlo F., J. Iron Steel Inst. 148. 137 (1943).
(1967). 69. Laszlo F.: J. Iron Steel Inst. 150; 183 (1944).
34. Chen’C. W., Acta metall. 9, 68 (1961). 70. Laszlo F., J. Iron Steel Inst. 152, 207 (1945).
35. Humphreys F. J. and Stewart A. T., Surface Sci. 31, 71. Brooksbank D. and Andrews K. W., J. Iron Steel Inst.
389 (1972). 206, 595 (1968).
36. Fischmeister H. F., Navarra E. and Easterling K. E., 72. Brooksbank D. and Andrews K. W., J. Iron Steel Inst.
Metal Sci. J. 6, 211 (1972). 207, 474 (1969).
37. Ashby M., Phil. Mag. 14, 1157 (1966). 73. Brooksbank D. and Andrews K. W., J. Iron Steel Inst.
38. Gessinaer G. H.. Fischmeister H. F. and Lukas H. L., 208, 495 (1970).
Powde; Metall. i6 (31) 119 (1973). 74. Brooksbank D. and Andrews K. W., J. Iron Steel Inst.
39. Franklin J. E., Judd G. and Ansell G. S., Proc. 3rd 208, 582 (1970).
ICSMA, Vol. 1, p. 345. Iron and Steel Institute, 75. Brooksbank D. and Andrews K. W., J. Iron Steel Inst.
London (1973). 210, 246 (1972).
40. Cane B. J. and Middletown C. J., Metal Sci 15, 302 76. Brooksbank D. and Andrews K. W., J. Iron Steel Inst.
(1981). 210, 765 (1972).
41. Easterling K. E., Fischmeister H. F. and Navarra E., 77. Tanaka K.. Mori T. and Nakamura T.. Phil. Man.
Y 21.,
Powder Metall. 16 (31), 128 (1973). 267 (1970):
42. Sundstrom B., Engng Fract. Mech. 6, 483 (1974). 78. Brown L. M. and Stobbs W. M., Phil. Mag. 34, 351
43. Rellick J. R. and McMahon C. J., Metall. Trans. 5, (1976).
2439 (1974). 79. Argon A. S., J. Engng Mater. Tech., Trans. ASME, 98,
44. Smith C. L. and Low J. R., Jr, Metall. Trans 5, 279 60 (1976).
(1974). 80. Argon A. S., Im J. and Needleman A., Metall. Trans.
45. King J. E. and Knott J. F., Metall. Sci. 15 (1), 1 (1981). 6A, 815 (1975).
46. Hippsley C. A. and Druce S. G.,Acta metall. 31 (II), 81. Rhee S. S. and McClintock F. A., Proc. 4th U.S. Nat1
1861 (1983). Cong. Appl. Mech., Vol. 2, p. 1007. ASME, New York
47. Gangulee A. and Gurland J., Trans. AIME, 239, 269 (1962).
(1967). 82. Goods S. H. and Brown L. M., Acta metall. 27, 1
48. Butler B. R. and Allen P. L., Metal Sci. 11,462 (1977). (1979).
49. Kozasu I., Shimizu T. and Kubota H., Trans. Iron 83. Fisher J. R. and Gurland J., Metal. Sci. 15, 185 (1981).
Steel Inst. Jon 13. 20 (1973). 84. Fisher J. R. and Gurland J., Metal. Sci. 15, 193 (1981).
50. Beremin F. M., tietall. Trans. 12A, 723 (1981). 85. McLean D., Mechanical Properties of Metals, p. 224,
51. Lindlev T. C.. Oates G. and Richards C. E., Acta John Wiley, New York (1962).
metall.* 18, 1127 (1970). 86. McClintock F. A., J. appl. Mech., Trans. ASME 90,
52. Cochrane R. C., JISI, Effects of Second Phase Particles 363 (1968).
on Mechanical Properties of Steel, p. 101. The Iron and 87. McClintock F. A., Ductility, p. 255. ASM, Metals
Steel Institute, London (1971). Park, OH (1968).
53. Restaino P. A. and McMahon C. J., Jr, Trans. A.S.M. 88. Berg C. A., in Proc. 4th U.S. Nat1 Cong. Appl. Mech.,
60, 699 (1967). Vol. 2, p. 855. ASME, New York (1962).
54. Thompson A. W. and Williams J. C., Advances in 89. McClintock F. A., Kaplan S. M. and Berg C. A., Int.
Research on the Strength and Fracture of Materials J. Fract. Mech. 2, 614 (1966).
 Ductile fracture 1073

90. McClintock F. A., Fracture (Edited by H. Leibowitz), 123. Chen W., Inc. J. Solids Struct., 7, 685 (1971).
Vol. 3, 47. Academic Press. New York (1971). 124. Norris D. M. Jr, Moran B., Scudder J. K. and
91. Perra M. W, Ph.D. thesis; University of California, Quinones R. F., J. Mech. Phys. Solids 26, 1 (1978).
Berkeley, CA (1976). 125. Hancock J. W. and Brown D. K., J. Mech. Phys.
92. Rice J. R. and Tracey D. M., J. Mech. Phys. Solids 17, Solids 31, 1 (1983).
201 (1969). 126. Hancock J. W. and MacKenzie A. C., J. Mech. Phys.
93. Tracey D. M., Engng Fract. Mech. 3, 301 (1971). Solids 24, 147 (1976).
94. Marini B., Mudry F. and Pineau A., Engng Fract. 127. MacKenzie A. C., Hancock J. W. and Brown D. K.,
Mech. 22, 989 (985). Engng Fract. Mech. 9, 167 (1977).
95. Mudry F., These d’Etat, Universitt de Technologie de 128. Edelson B. I. and Baldwin W. M., Trans. ASM, 55,230
Compiegne (1982). (1962).
96. Needleman A., J. appl. Mech. Trans. ASME, 94, 964 129. Lentz J. S., J. Pressure Vessel Technology, Trans.
(1972). ASME 100, 77 (1978).
97. Floreen S. and Hayden H. W., Scripta metall. 4, 87 130. Speich G. R. and Spitzig W. A., Metall. Trans. 13A,
(1970). 2239 (1982).
98. Bluhm J. I. and Morrissey R. J., Proc. 1st Inc. Conf. 131. Bernard G., Grumbach M. and Moliexe F., Metals
Fracture (Edited bv Yokobori T. et al.).,. Vol. III. _v. Tech 2, 512 (1975).
1739. Sendai, Japan (1966). 132. Wilson A. D., J. Engng Mater. Tech., Trans. ASME,
99. Cottrell A. H., Fracture (Edited by Averbach B. L., 101, 265 (1975).
Felbeck D. R., Hahn G. T. and Thomas D. A.), p. 20. 133. Corti C. W., Cotterill P. and Fitzpatric A.. Inc. Metals
Technology Press of MIT and John Wiley, New York Rev. 19, 77 (1979).
(1959). 134. Brown L. M. and Embury J. D., Proc. 3rd Inc. Conf.
100. Beevers C. J. and Honeycombe R. W. K., Phil. Mug. Strength of Metals and Alloys, p. 161. Iron and Steel
7, 763 (1962). Institute, London (1975).
101. Garber R., Bernstein I. M. and Thompson A. W., 135. Rosenfield A. R. and Hahn G. T., Trans. ASM 59,962
Scripta metall. 10, 341 (1976). (1966).
102. Garber R., Bernstein I. M. and Thompson A. W., 136. Stout M. G. and Gerberich W. W., Metall. Trans. 9A,
Metall. Trans. 12A, 225 (1981). 649 (1978).
103. Thompson L. D., Proc. Earl R, Parker Symp. Struc- 137. LeRoy G., Emburv J. D.. Edward G. and Ashbv
ture Propery Relationships (Edited by Antolovich M. F.: Actn metall.-29, 1509 (1981).
S. D., Ritchie R. 0. and Gerberich W. W.), n. 391. 138. Kirkcaldv D.. Proc. Scot. Shiobuild. Ass.. , Glaseow.
WI
TME-AIME, Warrendale, PA (1986). _ Appendix, p. 74 (186&61). *
104. Roesch L. and Henrv G.. Electron Microfractoaravhv 139. Izod E. G., Engineering 76, 431 (1903).
(Edited by Brothers- A. J.), ASTM STP 453, p. 3. 140. Charpy G., Proc. Inc. Ass. for Testing Materials Vol.
ASTM, Philadelphia (1969). 1, Report III (1909).
105. Low J. R., Jr, Prospects ofFracture Mechanics (Edited 141. Ludwik P. and Scheu R., Stal Eisen 43, 999 (1923).
by Sih G. C., VanElstrand H. C. and Broek D:). Vol. Summarized in J. Iron Steel Inst. 108, 457 (1923).
35. Noordhoff, Leyden (1974). 142. Agogino A. M., J. Engng Mater. Tech., Trans ASME
106. Van Stone R. H., Merchant R. H. and Low J. R., Jr, 100, 348 (1978).
Fatigue and Fracture Toughness-Cryogenic Behavior 143. Earl J. C. and Brown D. K., Engng Fract. Mech. 8,599
(Edited by Hickev C. F., Jr and Broadwell R. G.). (1976).
ASTM STP 556, p. 93 ASTM Philadelphia (1973). ‘. 144. Thomson R. D. and Hancock J. W., Inc. J. Fract. 26,
107. Jay G. D. and Nutting J., Effects of Second Phase 99 (1984).
Particles on Mechanical Properties of Steel, v. 95. The 145. Hancock J. W. and Thomson R. D., Muter. Sci. Tech.
Iron and Steel Institute, London (1971). _ 1, 684 (1985).
108. Baker T. J. and Johnson R.. J. Iron Steel Inst. 211.783 146. Yamamoto H., Inc. J. Fract. 14, 347 (1978).
(1973). 147. Saje M., Pan J. and Needleman A., Inc. J. Fract. 19,
109. Ritchie R. 0. and Knott J. F., Metall. Trans. 5, 782 163 (1982).
(1974). 148. Tvergaard V., Int. J. Solid Struct. 18, 659 (1982).
110. O’Brien R. N., Jack D. H. and Nutting J., Heat 149. Tvergaard V. and Needleman A., Acta metall. 32, 157
Treatment ‘76, Proc. 16th Inc. Heat Treatment Conf, (1984).
p. 161. The Metals Society, London (1977). 150. Needleman A. and Tvergaard, J. Mech. Phys. Solidr
111. Baker T. J.. Sulfide Inclusions in Steel. _ v. 135. ASM. 32, 461 (1984).
Metals Park, OH (1975). 151. Corrigan D. A., Travis R. E., Ardito V. P. and Adams
112. Ludtka G. M. and Laughlin D. E., Metall. Trans. 13A, C. M., Jr., Welding J. Res. Suppl. 41, 123-S (1962).
411 (1982). 152. Lee D. and Backofen W. A.., Trans. AIME, 236, 1077
113. Kirman I., Metall. Trans. 2, 1761 (1971). (1966).
114. Edelson B. I., Trans. ASM 56, 82 (1963). 153. Lee S., Majno L. and Asaro R. J., Metall. Trans. 16A,
115. Bourcier R. J. and Koss D. A., Proc. 5th Inc. Co& 1633 (1984).
Fracture (Edited by Francois D.), Vol. 1, p. 187. 154. Slatcher S. and Knott J. F., Adv. in Fracture Research
Pergamon Press, New York (1980). (Fracture 84) Vol. 2, p. 1473. Pergamon Press, Oxford
116. Nemat-nasser S. and Taya M., Continuum Models of (1984).
Discrete Systems (Edited by Provan J. W.), p. 387. 155. Anand L. and Spitzig W. A., J. Mech. Phys. Solids 28,
Universitv of Waterloo Press, Waterloo (1980). 113 (1980).
117. Tvergaard V., Inc. J. Fract. 17, 389 (198i). 156. McMeeking R. M. and Rice J. R., Inc. J. Solids Struct
118. Tveraaard V.. J. Mech. Phvs. Solids 30. 265 (1982). 11, 601 (1975).
119. Gurson A. L.; J. Engng Miter. Tech., Trans. ASME, 157. Burke M. A. and Nix W. D., Inc. J. Solids Struct. 15,
99, 2 (1977). 379 (1979).
120. Clausing D. P., Inc. J. Fract. Mech. 6, 71 (1970). 158. Ohno N. and Hutchinson J. N., J. Mech. Phys. Solids
121. Bridgman P. W., Studies in Large Plastic Flow and 32, 63 (984).
Fracture. Harvard University Press, Cambridge, MA 159. Tvergaard V., Needleman A. and Lo K. K., J. Mech.
(1964). Phys. Solids 29, 115 (1981).
122. Needleman A., J. Mech. Phys. Solids 20, 111 (1972). 160. Spitzig W. A., Acta metall. 33, 175 (1985).
1074 W. M. GARRISON JR and N. R. MOODY

161. Hahn G. T., Barnes C. R. and Rosenfield A. R., 180. Ritche R. O., Server W. L. and Wullaert R. A., Metall.
Aerospace Research Lab. Rep. ARLTR75-0194, Trans. lOA, 1557 (1979).
Wright-Patterson Airforce Base, OH (1975). 181. Osborne D. E. and Embury J. D., Metall. Trans. 4,
162. Clausing D. P., U.S. Stel Res. Lab. Rep. 35.066 (1972). 2051 (1973).
163. ASTM E399-83. Standard Test Method for Plane 182. Pandey R. K. and Banerjee S., Engng Fract. Mech. 10,
Strain Fracture Toughness of Metallic Materials, 1983 817 (1978).
Annual Book of ASTM Standards, Section 3, p. 170. 183. Schwalbe K. H., Engng Fract. Mech. 6, 415 (1974).
American Society for Testing and Materials, Philadel- 184. Garrison W. M. Jr, Scripta metall, 18, 583 (1984).
phia (1983). _ 185. McMeeking R. M., J. Mech. Phvs. Solids 25. 357
164. Knott J. F., Fundamentals of Fracture Mechanics. (1977). -
Butterworths, London (1973). 186. Garrison W. M. Jr., Proc. Earl R. Parker Symp.
165. ASTM E813-81, Standard Test Method for J,=, a Structure Property Relationships (Edited by
Measure of Fracture Toughness, 1983 Annual Book of Antolovich S. D., Ritchie R. 0. and Gerberich W. W.),
ASTM St&a&is, Section 3, p. 762. American Society p. 187. TMS-AIME (1985).
for Testing and Materials. Philadeluhia (1983). 187. Garrison W. M. Jr.,‘Scrt&a metall. 20, 633 (1986).
166. Begley J. A. and Landes’J. D., Fiacturk Toughness, 188. Moodv N. R.. Garrison W. M. Jr. Smueereskv J. E.
ASTM STP 514, p. 201. American Society for Testing and Costa J. E. (to be published).’ I -
and Materials, Philadelphia (1974). 189. Aravas N. and McMeeking R. M., J. Mech. Phys.
167. Shih C. F., J. Mech. Phys. Solids 29, 305 (1981). Solids, 33, 25 (1985).
168. McMeeking R. M., J. Engng Mater. Technol. Trans. 190. Begley J. A., private communication.
ASME 99, 290 (1977). 191. Hancock J. W. and Cowling M. J., Metal Sci. 14, 293
169. Rice J. R. and Johnson M. A., Inelastic Behavior of (1980).
Solids (Edited bv Kanninen M. F.. Adler W. G.. 192. McMeeking R. M., Presented at the Adv. Res.
Rosenfield A. k. and Jaffee R. I.), p. 641: Workshop Chemistry and Physics of Fracture, Bad
McGraw-Hill, New York (1970). Reichenhall, FDR, 23 June-l July (1986).
170. Krafft J. M., Appl. Mater. Res. 3, 88 (1964). 193. Ritchie R. O., Smith R. F. and Knott J. F., Metal Sci.
171. Hahn G. T. and Rosenfield A. R., Applhations Related 9, 485 (1975).
Phenomena in Titanium Allovs, ASTM SIP 432. 194. Knott J. F., Conf. Proc.: Micromechanisms of Crack
pp. 5-32. American Society for-Testing and Materials, Extension Mechanics and Phvsics of Fracture II. Cam-
Philadelphia (1968). bridge (1980), in Metal Sci.. 14, 327 (1980).
172. Birkle A. J., Wei R. P. and Pellissier G. E., Trans. 195. Needleman A. and Tvergaard V., An Analysis of
ASM, 59, 981 (1966). Ductile Rupture Modes at a Crack Tip. Brown Univer-
173. Spitzig W. A., Trans. ASM, 61, 344 (1968). sity, Division of Engineering, Providence, RI (1986).
174. Lautridou J. C. and Pineau A., Engng Fract. Mech. 15, 196. Ritchie R. 0. and Thomuson A. W.. Metall. Trans.
55 (1981). 16A, 233 (1985). -
175. Green G. and Knott J. F., J. Engng Mater. Technol., 197. Marder A. R., Metall. Trans. 13A, 85 (1982).
Trans. ASME 98, 37 (1976). 198. Needleman A. and Rice J. R., Mechanics of Sheet
176. Garrison W. M., Jr., Metall. Trans. 17A, 669 (1986). Metal Forming: Material Behavior and Deformation
177. Garrison W. M., Jr. and Moody N. R., Metall. Trans. Analysts (Edited by Koistinen D. P. and Wang Neng-
18A, 1257 (1987). Ming), p. 237. Plenum Press, New York (1978).
178. McKlintock F. A., J. appl. Mech. Trans. ASME 25, 199. Handerhan K. J. Ph.d Thesis, Carnegie-Mellon Uni-
363 (1958). versity, Pittsburgh September (1987)
179. Bates R. C., Metallurgical Treatises (Edited by Tien 200. Thompson A. W. and Ashby M. F., Scripta Metall. 18,
J. K. and Elliott J. F.), p. 551. TMS-AIME, War- 127 (1984).
rendale, PA (1982).