Fatigue Fracture of Polycarbonate

MICHAEL T. TAKEMORI
Polymer Physics and Engineering Branch
Corporate Research and Development
General Electric Company
Schenectady, New York 12301

The fatigue fracture process in polycarbonate can be differ-

ent for cracks grown from notches and from surface crazes. Dur-
ing the discontinuous crack growth process (which occurs dur-
ing the very slow crack growth regime), full-width cracks
grown from notches have a very different crack tip plastic zone
when compared to that observed in localized cracks grown
from surface crazes. These localized surface cracks have a pair
of sharply defined shear bands at the crack tip (forming the
“epsilon” plastic zone) that are absent from notched-grown
cracks. The presence ofthe shear band pair in this plastic zone
produces two major modifications. The discontinuous crack
growth is extended to higher stress intensity ranges due to the
(crack tip) craze stabilization by the shear band pair. Moreover,
a shear fatigue fracture mode, absent in the notched speci-
mens, occurs upon termination of the discontinuous growth
process. The implications of this comparison are that estimates
of fatigue lifetimes of smooth samples based on fatigue crack
growth data (on notched specimens) may be erroneous.

INTRODUCTION ment was observed in PMMA by Mukherjee, et al. (3).

In polycarbonate, however, given the observed differ-
ecently it was discovered that fatigue crack growth ences in the structure of the crack tip plastic zone dur-
Rbehavior in unnotched (smooth bar) specimens of ing discontinuous crack growth, it should not be sur-
polycarbonate can exhibit an unusual crack tip plastic prising that through-thickness cracks manifest
zone consisting of a single craze and a pair of shear markedly different growth behavior when compared to
bands (1, 2). Due to its visual appearance, this plastic part-through cracks in unnotched samples. The eluci-
zone was aptly labeled the “epsilon” plastic zone. This dation and clarification of these differences (and simi-
unique zone was observed during the discontinuous larities) will be addressed in this article.
crack growth period (where “discontinuity” implies Earlier fatigue studies on polycarbonate have gener-
multiple load cycles between crack jumps). Surpris- ally concentrated either on the growth behavior of
ingly, the epsilon plastic zone was not observed during through-thickness fatigue cracks (4-14) or on lifetime
discontinuous crack growth in notched samples (e.g., and fractography studies of smooth-bar polycarbonate
compact tension or single-edge notched specimens) of (15-18). In the former area, linear elastic fracture me-
polycarbonate despite the (nominally) similar stress chanics concepts have been applied to the study of
distribution near the crack tip. In some regions along through-thickness fatigue crack growth behavior. The
the crack front, the crack tip plastic zone consisted of a crack growth dependence on stress intensity factor (K,)
single craze (and no shear bands) whereas, in other re- has been examined as a function of temperature, fre-
gions, multiple layers of crazed material formed the quency, mean stress, molecular weight, sample envi-
plastic zone ahead of the crack (again, with no shear ronment, etc. In the latter area, fractography has been
bands). the primary investigative tool for post-mortem fracture
Crack growth rate data obtained from conventional analysis of unnotched specimens. Fatigue cracks initi-
notched specimens can be applied to crack growth ki- ate from localized cracks (1,2, 15)and eventually prop-
netics in unnotched specimens for situations where the agate as part-through cracks. A variety ofcrack propaga-
stress distribution at the crack tips is similar. In fracture tion modes, such as discontinuous growth, single-cycle
mechanics terms, this requires comparable stress inten- advance, and shear fatigue fracture, have been studied
sity distributions and plastic zones along the crack by fractographic analysis.
front. This type of correlation would be useful for poly-
mers, since cracks often initiate from surface crazes and EXPERIMENTAL PROCEDURE
become localized “part-through” surface cracks rather Unnotched tensile bar samples with rectangular
than cracks that span t h e w i d t h of t h e specimen cross sections (0.32 . 1.3 cm) were machined from
(“through thickness” cracks). Indeed, reasonable agree- 0.32-cm-thick BPA-polycarbonate extruded sheets.

POLYMER ENGINEERING AND SCIENCE, OCTOBER, 1982, Vol. 22, No. 15 937
 Michael T . Takemori

The samples were then sanded and polished with vary-

ing grades of silicon carbide paper and aluminum oxide
powder (down to 0.05p). Tests were performed on an
Instron Model 1350 servohydraulic material test sys-
tem at frequencies ranging from 1to 10 Hz using a sinus-
oidal waveform. The loading was tension-tension with a
stress amplitude ranging from 17 to 21 MPa and an R
value of 0.1 (minimum to maximum stress ratio). All
tests were performed at room temperature (23” to 25OC)
in room atmosphere. Additional tests were performed
on compression-molded and injection-molded tensile
bars, with similar results having been obtained. Due to
the high residual compressive stresses on the surfaces of
the injection-molded specimens, fatigue crack initia-
tion was retarded when compared to the compression-
molded bars, where residual stresses were minimized. Fig. 2 . Fracture surface microstructure in the mirror and mist
Smaller tensile bars (machined from 0.15-cm-thick region during low-speed, discontinuous crack growth.
BPA-polycarbonate extruded sheet) were used for the
interrupted testing study of the growth kinetics of the will be used to refer to the notched case, whereas “lo-
smooth bar cracks. Compact tension specimens were calized discontinuous crack g r o w t h (LDCG) will refer
prepared from various thickness BPA-polycarbonate to the unnotched sample behavior. “Localized” is de-
extruded sheets for the crack propagation testing. All scriptive of the small part-through cracks that initiate in
tests were performed at an R value of 0.1. surface crazes in unnotched samples of polycarbonate
The fracture surfaces were examined with a Toho In- during fatigue. It appears that the epsilon plastic zone
struments Super I1 scanning electron microscope. The will only develop in cracks that are highly localized, al-
fracture surface profiles and the subsurface regions though a satisfactory explanation has not yet been found
were studied by polarized transmission light micros-
(1, 2).
copy with a Zeiss light microscope. For the subsurface DCG in polymers was first reported by Elinck et
birefringence studies, thin sections (approximately 50p al. (20) in fatigue crack propagation tests on polyvinyl
thick) of the sample were produced by a grinding and chloride (PVC). The number of crack arrest bands on the
polishing procedure, described elsewhere (19). fatigue fracture surface could only be related to macro-
FATIGUE FRACTURE OF NOTCHED scopic crack growth observations by postulating many
POLYCARBONATE load cycles per arrest band; i.e., DCG. Since then, fa-
tigue crack propagation tests by Hertzberg, Manson, and
The fatigue fracture surface ofa typical notched poly- coworkers (7, 21, 22) have revealed discontinuity in a
carbonate compact tension specimen is shown in Fig. 1 . variety ofother polymers. Direct observations have also
The range in stress intensity factor AKI (see Appendix A) confirmed DCG in PVC (23), and video recordings
is shown at appropriate points along the fracture sur- have provided growth kinetics information (21). One
face. At low AKI values (0.3 to 0.7 M N - ~ L - ~a’ smooth
~), common feature of DCG is a fracture surface appear-
“mirror” region and a rough “mist” region coexist. In ance that reveals a distinct variation in the size of the
the mirror region, a regular set of crack arrest bands is granular structure between neighboring bands. The
observed (Figs 2a;2b),these bands having been formed granular structure is believed to be the remnants of the
discontinuously; i.e., many load cycles elapsed between plastic zone craze, which experienced fatigue loading
crack jumps. In order to distinguish the different dis- and subsequent breakdown (7, 17,21,22). The coarsest
continuous growth behavior in notched and unnotched granules, which lie about 1/3of the distance to the next
samples, the term “discontinuous crack growth” (DCG) crack arrest band (i.e., 1/3 of the crack jump length), are
the result of the most severe fatigue degradation,
whereas the finer granules farther out reflect the less-
damaged craze fibrils in the less-mature craze (also a
lower stress region).
Figure 2 shows closeups of t h e D C G region in
notched polycarbonate where the mirror and mist re-
gions appear side by side, hence indicating two modes
of crack advance along the crack front. The mirror re-
gion, which usually occurs near the free surface of the
sample, is due to the development of a single craze
layer ahead of the crack front. In the mist region, how-
ever, multiple craze layers (craze bundling), which lead
to crack propagation on different craze planes along the
Fig. 1 . Typicul notched compact tension specimen futigue fruc- crack front, produce an irregular fracture surface.
ture surface. Where the different crack layers meet, vertical steps,

938 POLYMER ENGINEERING AND SCIENCE, OCTOBER, 1982, VoI. 22, No. 15
 Fatigue Fracture of Polycarbonate

often observed as ductilely drawn material, appear, single-cycle crack advance were postulated for the mist
and give the "mist" appearance (Fig. 2c). DCG has pre- region, 100 to 500 A crack jumps would be required on
viously been thought to be associated (7)with the mirror each load cycle. A mechanism for crack advance on this
region of notched polycarbonate and single cycle crack scale is difficult to imagine.) For AK, > 0.7 MN-m-3",
advance with the mist region (based on single-cycle fa- the mirror region disappears, DCG terminates, and the
tigue striations being observed in the high AK, region). fatigue crack growth rate accelerates (indicated by a
However, since the crack front appears to propagate slope increase from two to five in Fig. 3).
uniformly across the width of the sample, the crack At higher A&, the crack growth rate decreases as the
most probably propagates discontinuously in the mist fracture behavior undergoes a mode transition. Near
region as well (when mirror and mist regions coexist). the specimen free surfaces, plane stress conditions en-
Furthermore, when crack arrest bands are visible in the hance shear flow whereas, in the interior portion, plane
mist region, they appear to have a fracture surface ap- strain conditions prevail and craze formation pre-
pearance (Fig. 2d) similar to that observed in the DCG dominates. A mixed mode crack propagates, with k45"
mirror region (Fig. 2b). This indicates highly localized shear zones forming at the free surfaces of the speci-
regions of DCG in one of the multiple craze layers. men. The size of this shear zone, which increases with
Macroscopic fracture surface appearances (mirror or A&, has often been related to the size of the crack tip
mist) may, therefore, not be sufficient to indicate the plastic zone (22), although some studies indicate that
nature of the crack advance mechanism. the shear zone width can be many times larger than the
Figure 3 shows fatigue crack propagation data for fully reversed plastic zone (24). The shear lip width has
polycarbonate. The low AKI data were obtained at 85 also been described in terms of the competition be-
Hz, while all other data presented were obtained at 4 tween craze formation and shear yielding (ll), the
Hz. Due to the relative frequency insensitivity of fa- greater strain rate dependence of the craze stress lead-
tigue crack propagation in polycarbonate at room tem- ing to greater shear lip size at higher AK,.
perature (6, 8, 14), the data join relatively smoothly. At The advance of the crack front through the shear
low AKI, there is evidence for a threshold value; i.e., a zones is slower than through the plane strain crazed
minimum AKI necessary to propagate the crack. Near zone; hence, the crack front develops a "tongue" shape,
the threshold value, the crack growth per cycle is of the with the crack lagging at the free surfaces. At high A&,
order of molecular dimensions (approximately 10 A). the transition to a plane stress mode across the whole
This crack growth is, of course, an average rate which is thickness of the sample is completed. The thickness de-
only physically meaningful since the crack grows dis- pendence of the transition can be examined simply in
continuously with crack jumps of the order of several p. terms of the fully reversed plastic zone size (see Appen-
For AKI between 0.3 and 0.7 MN-m-312, the fatigue dix A, E q A 4 ) . For specimens between 3.3 and 9.5 m,
crack in polycarbonate propagates discontinuously. (If
t / r * , = 18 k 1

where crv = 60 M P a , t is the thickness of the sample, and

. .. . 3..
L
r*yis the fully reversed plastic zone size at the onset of
n lo-'r I.

O f the mode transition. This result is in good agreement

with that found by Schijve in aluminum alloy sheet ma-
terial (24).

U l L
PUWE
sm€M 1 The nature of the mixed mode fatigue crack propaga-
tion and the transition to a plane stress fatigue fracture
12-31
mode is revealed in F i g . 4 . In this sample, mixed mode
crack growth occurred for AK, values ranging from 2.5
to 3.5 MN-mP3l2, whereupon a transition to a pure shear
mode was observed. The surface profiles and the
subsurface birefringence of four thinned and polished
cross-sections (plane normal to crack growth direction)
are also shown in F i g . 4 . The ~ 4 5 shear" zones are
clearly evident. In the lowest AKI cross-section, the
IDS 10' crack front followed one of the 245"shear zone bound-
nK (MN ,.-*,a>
aries, forming the well-known 45"shear lips on the frac-
Fig. 3. Fatigue crack propagation in polycarbonate. The 6.1 mm ture surface. For the remaining three cross-sections
and 3.3 mm data refer to compact tension specimen thicknessep,
with the low-speed data taken at 85 H z and the high-speed, 6.1 shown, however, the crack propagated through the
mm data, as well as the 3.3 mm data, taken at 4 H z . The 3.3 mm shear zones, thus forming shear lips with fracture sur-
data exhibits a plane-stress-to-plane-strain transition at 2.5 faces at angles less than 45" to the crack plane. In fact,
MN-m-3'2,whereas the 6.1 mm data entera mixed mode fatigue for the highest AK, cross-sections, the crack front, al-
fracture at 3.5 MN-m-3'2.The low-speed, smooth-bar data were though completely in aplane stress mode, is nearly hor-
takenfrom interrupted testing during LDCG (see text), whereas
the high-speed, smooth-bar data were taken from postmortem izontal, exhibiting no shear "lips. Thus, the formation
"

fracture surface analysis of the full-width, single-cycle advance of the characteristic 545" shear lips on the fracture sur-
region. face is not a necessary condition for plane stress fatigue

POLYMER ENGINEERING AND SCIENCE, OCTOBER, 1982, Vol. 22, No. 15 939
 Michael T . Takemori
FRACTURE CRE)ss SECTION SURFACE PROFILE
SURFACE AND SUBSURFACE BIREFRIWNCE

Fig. 5 . l'ypicul unnotched tensile fatigue fracture surface (A)

and a ciew of a thinned section (-50 p thick) through crossed
polaroidr ( B ) . The plane of the section contains the crack propa-
gation direction (left to right) and the loading direction.

shear band pair in each epsilon plastic zone. This se-

quence of gradually increasing epsilon plastic zones
(the load amplitude is kept constant) creates a fishbone
structure when the cracks are viewed before final fail-
ure, as shown in Figure 6. The LDCG fracture surface
(Fig. 7) shows similar crack arrest band structure as the
DCG fracture surface (Fig. 26). This is a strong indica-
tion that the mechanism of discontinuous crack propa-
gation in both notched and unnotched samples is
closely related.
The details of the development of the epsilon plastic .
zone during LDCG in polycarbonate have been de-
scribed elsewhere (2). For those studies, the fatigue
Fig. 4. Typical notched compact tension specimens during
loading was periodically interrupted and in-situ cracks
mixed-mode crack growth. Ecidence o f t 4 5 " s h e a r zones in the
subsurface region is clear. Fructure through the shear zones were observed and photographed. The craze and shear
does not, however, necessarily follow the 245" shear-zone band growth kinetics, as well as the overall crack
boundaries, (IS seen in the subsurfuce uiew of the cross-sections. growth rate, were recorded. One sequence of photo-
graphs depicting the discontinuous nature of crack
fracture. The shear zones, on the other hand, are growth and the growth of the shear bands and the lead-
roughly contained within a 245" region. ing craze is shown in Fig. 8. Between successive crack
FATIGUE FRACTURE OF UNNOTCHED
POLYCARBONATE
The fatigue fracture surface of a typical unnotched
polycarbonate tensile fatigue specimen is shown in Fig.
5a. This sample was fatigued at 5 Hz under sinusoidal,
tension-tension loading with a stress amplitude of 18
MPa and an R value of 0.1. Five distinct regions can be
seen: craze initiation, localized discontinuous crack
growth (LDCG), shear fatigue fracture, full-width
single-cycle fatigue striations, and final-quarter-cycle
tensile failure. The surface profile and subsurface bire-
fringence of a thinned section are shown in Fig. 5b, re-
vealing further evidence of the marked differences in
behavior in each of the five regions. In the LDCG re-
gion, there is a one-to-one correspondence between the Fig. 6. Partial fatigue crack showing LDCG .shear hand puirs.
shear bands in the subsurface and the surface crack ar- The spucings between the shenr band bases vary f r o m 50 to 60
rest bands, with each shear band being one-half of the P.

940 POLYMER ENGINEERING AND SClfNCf, OCTOBFR, 1982, Vol. 22, No. 75
 Fatigue Fracture of Polycurbonute

does indeed reside at the base of the foremost “epsi-

lon.” Were the crack tip to lie further forward, then
there would be no driving force for shear band growth,
since the base of the shear band pair would lie behind
the crack tip.
On the basis of those studies, it was concluded (for
the particular test conditions chosen) that, during
LDCG, the leading craze also jumped discontinuously.
When the crack jump occurred, some of the excess en-
ergy released was immediately spent in creating new
crazed material. This “first cycle” crazed material has
some distinct characteristics (e.g., a certain fibrillar
structure or elongation), which is impressed into the
craze, since the next crack jump (many cycles hence) es-
sentially runs through the entire first cycle craze and
no further. During the first few load cycles (1 to lo),
the shear bands grow rapidly to 80 or 90 percent of their
full length. During subsequent load cycling, craze fa-
tigue and breakdown commences, and slow craze and
shear-band lengthening occurs, presumably in re-
sponse to the gradual weakening of the fatigued craze.
The growing shear bands, leaning forward, shield the
craze and raise the stresses at the craze tip, thus
promoting craze growth. Finally, the weakened craze
fails, most probably at the location of maximum fatigue
damage. The crack then jumps through the weakened
craze. This mechanism of craze fatigue and breakdown
is believed to be similar to that theorized for DCG in
notched polycarbonate; hence, it is not too surprising
that the crack arrest band surface structure (Figs. 2h and
7) are remarkably alike. The shear bands do play an im-
portant role, however. Basically, they shield the craze
and lower the craze stress. Therefore, the effective AK1
(hence, stress field in which the crack tip craze resides)
is lower than the nominal AKI calculated straightfor-
wardly by linear elastic fracture mechanics models.
The consequences of this shielding are manifested in
the crack growth rate data and the transition to a shear
fatigue fracture mode at the termination of the LDCG
process.
The stress intensity factor, A&, for the localized, dis-
continuous crack can be approximated using several stir-
face crack models (which are discussed more fully in
Appendix B). Corner cracks can be treated as quarter-
elliptical or quarter-circular cracks, while edge cracks
can be treated as semi-elliptical or part-circular cracks.
The basic form for the stress intensity factor does not
change (see E9 B 3 , ) , but a numerical prefactor (which
accounts for the influence of the front, back, and sides
of the sample) has a different value for each approxima-
tion. The crack growth rate data for a corner crack (using
the quarter-circular approximation) in a sample fa-
tigued at 1Hz with a stress amplitude of 18.6 MPa with
R = 0.1 is shown in Fig. 3. Although the overall range of
AKI for this corner crack during LDCG spanned from
F i g . 8. Sequenca of optical micrographs showing LDCG pro- 0.52 to 1.28 MN-mP3”, crack propagation rates were
cess. 100 cycles elapsed between the test interruptions and pho- only measured for AK1 > 0.87 MN-m-3/2, as shown in
tographs. Crack j u m p s ure indicuted by the u r o u s . Fig. 3. The range of AKI for LDCG was measured in
several other edge cracks (using the part-circular edge-
j u m p s , t h e shear band pairs can be observed to crack model) and resulted in values ranging from 0.30
lengthen, which is convincing proof that the crack tip to 1.12 MN-m-312.(Use ofthe semi-elliptical edge crack

POLYMER ENGINEERING AND SCIENCE, OCTOBER, 1982, Vol. 22, No. 15 94 1

 Michael T . Takemori

model would result in values approximately 20 percent

larger than those calculated from the part-circular
model. As explained in Appendix B, the part-circular
model is probably a more accurate description of the
LDCG edge cracks in polycarbonate.) The range of AKI
for LDCG is therefore much greater than that for DCG
in notched specimens (0.3to 0 . 7 MN-m-3'2).The reason
for the apparently larger span of AKI values for LDCG
is thought to be due to the shielding or craze-stabilizing
effect of the shear band pair, as explained in the previ-
ous section. Despite the high nominal AKl, the shielded
craze experiences a smaller effective AK1, thereby pro-
longing the discontinuous growth process.
The presence of the shear band pair in the epsilon
plastic zone may also explain the sudden termination of
the LDCG process. Although the number of load cycles
between crack jumps decreases with successive jumps,
it does not asymptotically approach a single cycle, as
might be expected (2).A rather abrupt and unexpected
end to the LDCG process is observed instead. The criti-
cal factor may thus not be the fatigue breakdown of the
craze, which is still quite stable even in the last epsilon
plastic zone. The shear bands, however, may grow to a
critical size, which drops the craze stress below the
growth threshold level, thereby terminating further
craze growth. The crack can no longer propagate in the
same plane as the craze; thus, the crack tip blunts and the
nature of the plastic zone changes drastically.
The terminal shear band pair grows considerably
larger than the penultimate pair. Two partial shear
cracks then initiate (one in each half of the shear band
pair) and propagates along the shear bands at 245" to
the main crack plane (Fig. 9). The crack advances in
every cycle and shear fatigue striations are often ob-
served on the shear fracture surfaces (Fig. 10). At some
Fig. 1 1 . A t the end of the shear fatigue fracture, the crack re-
critical point, the ligament of material joining the main sumes along the main cruck direction. The 45" shear fructure
crack from one half of the symmetric pair of partial can be clearly seen.
shear cracks tears and the shear fracture process termi-
nates (Fig. 11). yielding in the subsurface is evident in Fig. Fib. Crack
The crack subsequently propagates along a plane par- growth rates were determined for this region of single-
allel to the original crack (i.e., perpendicular to the ten- cycle advance. These data are also plotted in Fig. 3 in
sile loading direction). Considerable evidence of shear the high AKI region (3 to 9 MN-mP3l2) and show excel-
lent agreement with the plane stress crack growth data
obtained from notched specimens.

DISCUSSION
T h e fatigue fracture behavior of notched and
unnotched polymers may behave very differently. In
the case of polycarbonate, this difference is evident in a
strikingly unusual crack tip plastic zone, the epsilon
plastic zone, which is seen only in the unnotched speci-
mens. The presence of the shear band pair in this plastic
zone produces two major modifications to the fatigue
fracture process. The discontinuous crack growth pro-
cess, which is a slower growth process than the conven-
tional single-cycle crack advance mechanism, is ex-
t e n d e d to higher A K I values d u e to t h e craze
stabilization by the shear band pair. Moreover, a shear
Fig. 9. The terminution of the LUCG process w i t h the growth of fatigue fracture mode, absent in the notched speci-
two partial shear cracks, one in each shear band. These cracks
grow at 245" to the main crack plane. Each partiul crack is mens, occurs upon termination of the localized discon-
about 50 p long. tinuous growth regime. The existence of different crack

942 POLYMER ENGINEERING AND SCIENCE, OCTOBER, 1982, Vol. 22, No. 15
 Fatigue Fracture of Polycarbonate

tip plastic zones is probably due to the different bound- singularity at the crack tip and completely describes
ary conditions for the stress distributions along the the stress state about the crack tip (elastically). It is the
crack front and the tendency of polycarbonate to yield only parameter in E q A - l that is a function of the ap-
under plane stress conditions at the free surface (as op- plied farfield load and the geometry (crack shape,
posed to the crazing behavior under plane strain in the length, proximity to specimen surfaces). The singularity
sample interior). In the localized part-through surface creates a plastic zone about the crack tip where the ma-
cracks, the presence of the back (free) surface affects the terial yield stress is exceeded and deformation occurs
stress distribution, whereas, in the fully notched speci- (Fig. l2b). Thus, E q A-1 only applies in the elastic re-
men, the side (free) surfaces form the boundaries. gion outside of the plastic zone (here we are assuming,
Thus, although the stress intensity at the tip of the for simplicity, an elastic-plastic model). Under “small-
growing crack may be nominally the same in both scale yielding” conditions (i.e., when the plastic zone
notched and unnotched samples, the crack tip plastic dimensions are small relative to crack length and speci-
zones may be different due to the effects of the free sur- men dimensions), two different tensile configurations
faces. The implications of this comparison are that esti- (i.e., different load or crack geometry) with identical K I
mates of fatigue lifetimes of smooth samples based on values should produce the same surrounding elastic
fatigue crack growth data (on notched specimens) may stress distribution. This would, in turn, lead to similar
be erroneous. plastic zone development in the same material and,
hence, similar crack growth behavior. For this reason,
APPENDIX A: LINEAR ELASTIC the stress intensity factor has been a useful parameter
FRACTURE MECHANICS for studying crack development and propagation.
Linear elastic fracture mechanics can be used to de- For cyclic loading conditons, Paris and Erdogan (26)
termine the stress distribution and the deformation at a showed that the variation in stress intensity factor, AK,,
crack tip in a homogeneous, isotropic material (25). Un- could be related to the rate of fatigue crack propagation
der tensile (Mode I) loading, the stresses at a crack tip with a relationship of the form:
assume the form (Fig. 12):
(A-2)

where dc/dN is the crack growth rate and A and n are

wherefij(8) is a function of 8 only. K I is the stress inten- material parameters.
sity factor, which describes the severity of the l l f i The size of the plastic zone can be determined from
E q A-1 by using a yield criterion (such as the von Mises)
to determine the loci of points where the yield stress
A. CRACK TIP (fly) is exceeded. Under plane stress conditions, the
GEOMETRY plastic zone length ( r p )in the crack plane is given by:

Under plane strain conditions, the plastic zone length

tP A in the crack plane is dependent on the Poisson’s ratio (v)

for the material:

4 ~ ( 1- 41 (A-4)
CRACK TIP which, for v ranging from 0.2 to 0.5, is considerably
shorter than the plane stress plastic zone length.
The deformation in the crack tip plastic zone in poly-
mers is often localized in acraze ahead of the crack. This
B. CRACK TIP
STRESSSTATE

0
lh-& SINGULARITY
is a special case of tensile yielding on a discrete surface
ahead of the crack tip, which has been treated by the
Dugdale-Barenblatt model of a line plastic zone (25).
The line plastic zone length under monotonic loading is
given by:

(A-5)

which corresponds to the length that makes the

PLASTIC ELASTIC stresses at the zone tip bounded. Under load cycling
ZONE ZONE (fatigue) conditions, the small-scale yielding solutions
F i g . 12. a) the cruck tip geometry and b) the elastic stress singu- can be obtained from monotonic loading solutions by
larity. The size und shape of the plastic zone are much more replacing KI by AKI and a, by 2ry (27). The cyclic plas-
complicated than indicated by the schematic diagram. tic zone is thus reduced to 114 the size of the monotonic

POLYMER ENGINEERING AND SCIENCE, OCTOBER, 1982, Vol. 22, No. 15 943
 Michael T. Takemori

plastic zone (for tensile fatigue with O minimum load): Semi-Elliptical Edge Crack
Newman (28) has written a comprehensive review
(A-6) and assessment of the many models for semi-elliptical
edge cracks (Fig. 13c). The general solution is of the
This does not, however, account for the effects of crack form:
closure during reversed loading.
KI = @-‘cT, k F(c, b, t , W; 4) (B-3)
APPENDIX B: STRESS INTENSITY FACTOR FOR
where
A SURFACE CRACK
The determination of the stress intensity factor for sur-
face cracks (also known as “part-through” cracks) has
@= I,TI2
(1 - kZ sin’ +)“‘d@ (B-4)

been the subject of intense activity in recent years (28). is the complete elliptical integral of the second kind
Surface cracks are usually modeled by part-elliptical and
cracks (part-circular or part-parabolic cracks being spe- c 2
cial cases). For unnotched polycarbonate, cracks usu- k 2 = 1 - (T)
ally initiate from highly localized crazes on the speci-
men sides or corners and grow inwardly; hence, the F (also known as the “boundary correction factor”) is a
surface crack approximations are applicable. We thus function of the crack shape (b, c), of the sample geom-
propose three models for calculating the stress intensity etry (t = thickness, w = width), and angle 4. A useful
factor for surface cracks in polycarbonate (as well as approximation for F at 4 = n-I2 has been given by Paris
other polymers). By way of comparison, the stress in- and Sih (31):
tensity factor for a full-edge crack in a semi-infinite
body (Fig. 13a) is given by: F = 1 + 0.12(1 - +) (-$ -)lrc
tan 2t
l“
(B-6)

K , = 1.12am= (B-1) for c 5 b. For b < c 5 2b, Raju and Newman (28)have
developed the approximation for a small surface crack;
Quarter-Elliptical Corner Crack i.e., for c << t , b << w :
Kobayashi and Enetanya (29) derived numerical so-
lutions for the quarter elliptical corner crack in a quar-
F = (1 + 0.03-)b a (B-7)
C
ter infinite solid (Fig. 13b) as a function of aspect ratio
and
(clb).They presented graphical solutions for the stress
intensity factor along the entire crack front for three dif-
ferent aspect ratios (0.2,0.4,and 0.98).The stress inten-
sity factor exhibits a minimum close to the diagonal; In Fig. 14, the F/ Q, prefactor has been plotted for c/ b
i.e., at 45”to either edge, and rises to a value 20 percent values from 0 to 2. (c/ b = 0 corresponds to the limiting
higher near either free surface. The 0.98 aspect ratio re- case of the semi-infinite edge crack given by Ey B - I ) .
sults compare well with the tabulated finite element
calculations of Tracey (30) for quarter-circular cracks. Part Circular Crack
Along the diagonal, the stress intensity factor for the Numerical solutions for part-circular cracks (Fig. 13d)
special case of a quarter circular corner crack (of diagonal have been reported by Rooke and Cartwright (30)along
length c) is given by: the entire crack front for various values of c/a. The re-
2
K I = 1.09-aa, (B-2)
n-
SEMI-ELLIPTICAL EDGE
CRACK
THREE-DIMENSIONAL STRESS INTENSITY
SURFACE CRACK MODELS PREFACTOR

CRACK
CRACK* PLANE

PARIS, SIH
APPROXIMATION

RAJU, NEWMAN
f
Zb
APPROXIMATION

I I I I 1 I
0 1 2
A) SPMHNFINITE 6 ) OUARTER- C) SEMI-ELLIPTICAL D) PARTCIRCULAR clb
EDQE CRACK ELLIPTICAL EDQE CRACK EDGE CRACK
CORNERCRACK Fig. 14. Stress intensity prefuctor Fl@(Eq B-3) using two differ-
Fig. 13. Three-dimensional surfuce cruck models used t o culcu- ent upproximutions for clb s 1 ( E q s B-4 and B-6) und clh > 1
lute the cruck tip stress intensity factors. ( E q s B-7 urid B-8).

944 POLYMER ENGINEERING AND SCIENCE, OCTOBER, 1982, Vol. 22, No. 75
 Fatigue Fracture of Polycarbonate

sults at 6 = O are of the form:

where a varies from 1.0 to 1.023 for c ranging from 2a

to a.

Polycarbonate Surface Cracks

Figure 15 shows a photograph of a polycarbonate fa-
tigue crack during its localized discontinuous crack
growth phase. A comparison of the cyclic stress inten-
sity factor, determined by the semi-elliptical edge
crack approximation for c > b as well as the part-circular
crack model, revealed that the semielliptical model
produced a 15 to 20 percent higher estimate of the stress
intensity factor at 8 = 4 = 0. Visually, the part-circular
surface crack model (for c > a) appears to approximate F i g . 15. Polycarbonate surface crack during the LDCG process.
the shape of the crack front more closely; i.e., wide in The total crack length (normal to free surface) is 1.2 mrn.
the interior, and narrower at the surface due to the for-
mation of shear lips on either side of the crack opening.
These shear lips at the back of the crack act as wedges 9. M. Kitagawa, Bull. J S M E , 18, 240 (1975).
10. G. C. Martin and W. W. Gerberich, J . Muter. Sci., 11, 231
and prevent complete crack closure during the unload (1976).
portion of the tension cycling. This wedging action of 11. G. Pitman and 1. M. Ward, J . Muter. Sci., 15, 635 (1980).
the surface shear lips reduces the compressive stress 12. D. H. Banasiak, J. Appl. Poly. Sci., 21, 1297 (1977).
acting at the crack tip during the unload cycle, thus de- 13. S. Arad, J. C. Radon, and L. E. Culver, Polym. Eng. Sci., 12,
193 (1972).
creasing A&. Therefore, the smaller A& obtained with 14. J. C. Radon and L. E. Culver, Polymer, 16, 539 (1975).
the part-oircular model more accurately accounts for 15. N. J. Mills and N. Walker, J. Muter. Sci., 15, 1832 (1980).
the restraining action of the sheared material at the back 16. A. J. McEvily, R. C. Boettner, a n d T . L. Johnston,
of the crack. For these reasons, the part-circular model “Fatigue-An Interdisciplinary Approach,” Sagamore
is the preferred model for polycarbonate surface Army Mat. Res. Conf., 10, Syracuse University Press, Syra-
cuse, New York (1964).
cracks. 17. M. E. McKay, T.-G. Teng, and J. M. Schultz,J. Muter. Sci.,
14, 221 (1979).
ACKNOWLEDGEMENTS 18. G. H. Jacoby, “Electron Microfractography,” ASTM STP
Many helpful discussions with Dean Matsumoto, 453, 147 (1969).
Roger Kambour, and Thomas Morelli are acknowl- 19. A. S. Holik, H. P. Kambour, D. G. Fink, and S. Y. Hobbs,
“Microstructural Science,” 7, Proc. 11th Ann. Tech. Mtg.
edged. The skillful assistance of Steve Gifford in ob- Int. Metall. Society, Elsevier, Amsterdam (1978).
taining the fatigue crack propagation data, and the art- 20. J. P. Elinck, J. C. Bauwens, and G. Homes, Znt. J. Fruc.
istry of Andy Holik, Salvatore Alessi, and Jim Grande Mech., 7, 277 (1971).
in preparing and photographing the thinned sections 21. R. W. Hertzberg, M. D. Skibo, and J. A. Manson, Fatigue
are gratefully acknowledged. Mechanisms, ASTM STP 675,471 (1979).
22. R. W. Hertzberg and J. A. Manson, “Fatigue of Engineering
Plastics,” Academic Press, New York (1980).
REFERENCES 23. R. W. Hertzberg and J. A. Manson, J . Muter. Sci., 8, 1554
1. hl. T. Takemori and R. P. Kambour,J. Muter. Sci., 16, 1108 (1973).
(1981). 24. J. Schijve, Eng. Frac. Mech., 14, 789 (1981).
2. M. T. Takemori and D. S. Matsumoto, to be published, J . 25. See, e.g., J. G. Williams, “Stress Analysis of Polymers,”2nd
Poly. Sci., Poly. Phys. Ed. Ed., John Wiley & Sons (1980).
3. B. Mukherjee, L. E. Culver, and D. J. Burns,Expt. Mech., 9, 26. P. C. Paris arid F. Erdogan,]. Bas. Eng. Trans. A S M E Ser. D,
90 (1969). 85,528 (1963),
4. R. W. Hertzberg, H. Nordberg, and J. A. Manson,]. Muter. 27. J. R. Rice, S y m p . on Fatigue Crack Growth, ASTM STP
Sci., 5, 521 (1970). 415, ASTM, 247 (1967),
5. H. W. Hertzberg, J . A. Manson, and W. C. Wu, Progress in 28. J. C. Newman, Jr., “Part-Through Crack Fatigue Life Pre-
Flaw Growth and Fracture Toughness Testing, ASTM STP dictions,” ASTM STP 687, ASTM, 16 (1979).
536, ASTM, 391 (1973). 29. A. S. Kobayashi and A. N. Enetanya, “Mechanics of Crack
6. R. W. Hertzberg, J. A. Manson, and M. Skibo, Polym. Eng. Growth,” ASTM STP 590, ASTM, 477 (1976).
Sci., 15, 252 (1975). 30. D. P. Rooke and D. J. Cartwright, “Compendium of Stress
7. M. D. Skibo, R. W. Hertzberg, J. A. Manson, and S. L. Kim,J. Intensity Factors,” HMSO, The Hillindon Press, Uxbridge,
Muter. Sci., 12, 531 (1977). 281 (1976).
8. J. A. Manson, R. W. Hertzberg, S. L. Kim, and W. C. Wu, 31. P. C. Paris and G. C. Sih, “Fracture Toughness and Its Ap-
Adv. Chem. Ser., 154, 146 (1976). plications,” ASTM STP 381, ASTM, 30 (1965).

POLYMER ENGINEERING AND SCIENCE, OCTOBER, 1982, Vof.22, No. 15 945