Engineering Fracture Mechanics Vol. 34, No. 1, pp. 65-86, 1989 0013-7944/89$3.00+ 0.

00
Printedin Great Britain. 0 1989MaxwellPergamonMacmillanplc.

THE FRACTURE BEHAVIOUR OF PMMA IN

MIXED MODES I AND II
T. M. MACCAGNOt and J. F. KNOTT
Department of Materials Science and Metallurgy, University of Cambridge, Pembroke Street,
Cambridge CB2 342, U.K.

Abstract-The brittle fracture behaviour of PMMA under mixed mode I/II loading has been
investigated using pre-cracked bend bar specimens loaded in antisymmetric and symmetric four
point loading. The accuracy of the antisymmetric configuration was verified by a photoelastic
investigation, and by comparison with results obtained using other mixed mode load arrangements.
The results are used to test and discuss three mixed mode fracture criteria. It is concluded that mixed
mode I/II brittle fracture of PMMA at room temperature is described best by a maximum tangential
tensile stress criterion based on the linear elastic stress field. This is found to be consistent with the
present understanding of the micro-mechanisms involved in fracture of PMMA.

1. INTRODUCTION
FOR THE general case of a three dimensional body containing a crack subjected to arbitrary loading,
Irwin[l] observed that the overall stress field in the vicinity of the crack tip can be separated into
the sum of three basic components, or ‘modes’, of crack front loading: mode I (opening), mode II
(sliding), mode III (tearing). Moreover, Irwin noted that the stress distributions for each mode are
of a similar form, with the mode I stress field characterized by the mode I stress intensity factor, K, ,
the mode II field by K,, , and the mode III field by K,,, . In the special case of mode I crack front
loading, fracture toughness evaluation is well established. However, crack front loading in real
structures is more likely to be some combination of the three modes, and it is important to develop
an understanding of cracking behaviour in these ‘mixed mode’ situations. The present paper
addresses the problem of brittle fracture throughout the range from mode I to mode II in the
amorphous glassy polymer polymethylmethacrylate (PMMA), and describes the findings of a
recent effort[2] to characterize mixed mode I/II fracture behaviour of this material.
PMMA is a relatively homogeneous isotropic medium exhibiting brittle fracture behaviour at
room temperature, In addition, it is relatively easy to introduce a natural sharp crack into this
material by pressing a razor blade into the bottom of a notch. Therefore, as a first approximation,
it seems reasonable to suppose that the linear elastic field equations of Irwin[3] and of Williams[4]
provide a good description for the stresses and strains near the tip of a sharp crack in PMMA.
Using the linear elastic field to describe the situation simplifies as much as possible any analysis of
experimental results. This factor, coupled with the relative ease of working with this material, has
prompted a number of experimental studies on mixed mode I/II fracture behaviour of PMMA
using a variety of configurations to obtain mixed mode loading. For example, a large plate contain-
ing a central crack inclined at an angle fl to a remote uniaxial tension has been used by Erdogan
and Sih[S], Williams and Ewing[6], Ueda et al.[7], and by Theocaris[8]. Maiti and Smith[9] used
plates containing a central crack inclined at an angle to a tensile load applied at pins located near
the crack faces. Banks-Sills and Bortman[lO] used centrally cracked plates of complex geometry
with the entire plate inclined to the tensile load. A thin walled cylinder, containing an inclined crack
and subjected to torsion, has been used by Ewing and Williams[ 111. Ewing et al.[ 121 used plates
containing an edge crack inclined to a tensile load or a bending moment. Richard[ 131has used edge
cracked plates with the entire plate inclined to the tensile load.
However, despite PMMA ostensibly being a ‘model’ material to test, no consensus has
developed with respect to a suitable approach to describe mixed mode I/II fracture behaviour in
this material, and a number of theories based on continuum mechanics have been advanced.
These theories can be roughly classed into three categories. The first category maintains that the

tCurrent address: Structures and Materials Laboratory, National Aeronautical Establishment, National Research Council
of Canada, Ottawa, Canada, KlA OR6.

65
66 T. M. MACCAGNO and J. F. KNOTT

tangential tensile stress component, bgs, of the crack tip stress field controls the cracking behaviour.
This is the so-called ‘maximum 0 BO’hypothesis, originally proposed by Erdogan and Sih[S], with
modified versions advanced by, among others, Williams and Ewing[6], and Maiti and Smith[l4].
The second category attempts to extend to mixed mode situations the energy release rate (i.e. G)
concept of Griffith[ 151.Workers who have developed the ‘maximum G’ hypothesis include Hussain
et a1.[16], Palaniswamy and Knauss[l7], Ueda et al.[7], and Hyashi and Nemat-Nasser[l8]. The
third category of mixed mode I/II fracture theories proposes that a ‘strain energy density’ factor,
S, may be evaluated in the crack tip region, and it it this quantity which controls cracking
behaviour. The original ‘minimum S’ hypothesis was proposed by Sih[l9], and modified versions
have been proposed by numerous workers including Jayatilaka et a1.[20], and Theocaris[8]. A
number of these modified versions have been discussed[21-241.
An important point to note in all of these attempts to describe mixed mode I/II behaviour is
that all are developed from the point of view of the material behaving as an ‘ideal’ continuum. None
of these theories addresses the actual mechanisms by which separation of the real material takes
place. However, it seems reasonable to expect that the micro-mechanisms resulting in crack advance
and a criterion describing overall fracture behaviour must be consistent with each other, and an
understanding of the micro-scale processes leading to fracture in PMMA could lead to a suitable
mixed mode fracture criterion. With this in mind, a brief discussion on the mechanism of fracture
in PMMA is given in the next section. Section 3 describes recent experiments[2] carried out on
PMMA using a loading configuration which allows accurate characterization of crack tip condi-
tions to be carried out in a straightforward manner. An analysis is presented in Section 4 comparing
the experimental results with predictions from various mixed mode hypothesis, and with the results
of other workers. Section 5 discusses the findings, and a brief summary is given in Section 6.

2. MECHANISM OF BRITTLE FRACTURE

In real materials containing a sharp crack there is a zone surrounding the crack tip which has
undergone inelastic deformation, In metals, plastic deformation is produced by resolved shear
stresses. In amorphous polymers, however, inelastic deformation can be produced by shear stress
(shear yielding), or by normal stresses (crazing), depending upon the conditions. At the tip of a
crack in PMMA, it is the second mechanism which is operative. The resulting ‘crazed’ zone is a thin
structure which is crack-like in appearance, but which is observed actually to consist of cavities and
thin fibrils of polymer matter oriented parallel to the direction of the largest tensile stress. Figure 1
shows a schematic representation of a craze at the tip of a crack in an amorphous glassy polymer.
It is found that crazes are able to support some level of stress, but the combination of cavities and
fibrils leaves the craze weak compared with uncrazed material. Thus a craze provides a preferred
pathway for crack propagation (see Hertzberg[25]).
Studies also show that craze development is not only associated with a stationary crack. It
appears that, even for high crack velocities, crazing also occurs continuously just ahead of, and in
the same direction as, a propagating crack tip. Thus crazing appears to be an intermediate state of
crack propagation, as well as being present prior to crack propagation. Because of this intimate
connection between crazing and cracking, it is appropriate to discuss the details of both processes.
General reviews of crazing and cracking are provided by Kramer[26] and by Doll [27].

2.1. Crazing
Oxborough and Bowden[28] show that experimental results for the initiation of a craze in
complex stress fields can be predicted very well by a criterion which considers both the strain, L, and
the hydrostatic component of the stress tensor, cr,. The criterion states that crazing will occur on

Crack

Fig. I. Schematic diagram of the region in the vicinity of a crack tip in PMMA.
 Fracture behaviour of PMMA 61

the plane perpendicular to the direction of maximum tensile strain, when this tensile strain reaches
a critical value, 6,. This critical strain is in turn an inverse function of a,,,, and o,,, must be greater
than zero. Thus crazing is predicted to occur only in a region of hydrostatic tension, and the critical
tensile strain required to initiate a craze decreases as the hydrostatic tension increases. This seems
reasonable since the observed structure of a craze suggests that a crazing is in some sense a
cavitation process, and it would be expected that dilatation promotes cavitation. Note also that
since the direction of maximum strain coincides with the direction of maximum stress in an
isotropic medium, Oxborough and Bowden’s criterion agrees with the observation that the fibrils
in a craze are oriented in the direction of the largest tensile stress.
To a certain extent, the macro-scale criterion of Oxborough and Bowden can be correlated
with a reasonably detailed micro-scale model of craze formation presented by Donald and
Kramer[29]. In a material which flows, a gradient of hydrostatic tension (i.e. negative pressure)
provides a driving force for matter to be sucked from a region of lower hydrostatic tension to a
region of higher hydrostatic tension. If the gradient exists in the vicinity of a free surface, this
suction may distort the free surface into a meniscus. Furthermore, if the gradient of hydrostatic
tension increases with time, it is conceivable that, after a time, the initially smooth meniscus may
become unstable and take on the appearance of finger-like protrusions. Advancement of these
protrusions beyond a certain distance results in their eventual break up into a trail of matter. Such
a ‘meniscus instability’ argument has been implicated in a wide variety of cavitation phenomena,
and if the crack front in a polymer is taken as the meniscus, this model accounts very well for many
detailed high magnification observations of the structure of crazes in amorphous polymers. The
meniscus instability model also accounts for the importance of hydrostatic tension observed by
Oxborough and Bowden, and it can be reasoned that the their critical strain, cc, relates to the initial
instability of the meniscus.
It is now possible to explain why crazing occurs simultaneously with crack propagation. A
craze can be thought of as a region of localized plastic deformation which is produced by tensile
stress. For the situation of a sharp crack under load, the conditions of high triaxial tensile stresses
that exist near the tip are the very conditions which promote craze development. Therefore, it seems
reasonable that as the crack propagates, the craze will propagate along with it. Is is interesting to
note that this contrasts with the situation for metals, where plastic deformation is caused by shear
stresses, and is inhibited by triaxial tensile stresses.

2.2. Crack propagation

From the meniscus instability model it is clear that bulk polymer-which initially consists of
coiled, intertwined, molecular chains-at first deforms into groups of protrusions, and then further
deforms into roughly cylindrical fibrils extending across the width of the craze. The high degree of
deformation involved in going from bulk polymer to polymer fibrils can be viewed as a drawing
process, where the numerous polymer chains in each protrusion uncoil and straighten in response
to the local tensile stress (see Young[30]). Continued drawing results in aligned bundles of stretched
molecular chains, which span the width of the craze, and are the observed fibrils.
Since the fibrils are the only matter which extend across the width of the craze, they must
support the entire load acting across the plane of the craze. Furthermore, because of their reduced
cross section, the fibrils are stressed proportionately higher than the adjacent uncrazed polymer.
Thus, not only have the fibrils undergone severe deformation, they are under a high level of stress.
It can be imagined that in narrower fibrils, the already stretched polymer chains reach a point where
the capacity to deform in response to the stress is completely exhausted. Any further increase in
stress beyond this critical level will result in chain rupture. The applied load then re-distributes onto
neighbouring chains thereby extending them, and under continued stressing more chains will break
(see Andrews and Reed[31]). This process repeats until the entire fibril breaks, and in a manner
similar to individual chain rupture, other fibrils in the neighbourhood will subsequently break.
Continuous breakdown of craze fibrils is, in effect, propagation of a crack.

2.3. Summary
It appears that prior to propagation of the crack, PMMA must undergo a certain amount of
tensile strain in order for the material to be in a condition suitable for fracture. This strain is
68 T. M. MACCAGNOand J. F. KNOTT

associated with the development of a craze, and the craze provides a relatively weak path along
which the crack propagates. The actual critical event to initiate the fracture, however, is associated
with the breakdown of craze fibrils, and therefore involves the attainment of some critical stress.
Subsequent propagation of the crack is the simultaneous rapid formation and splitting of a craze
at the crack tip (see Kambour and Robertson[32]).
It is interesting to observe that the notion that nucleation of fracture involves inelastic strain,
but actual propagation is controlled by stress, is widely recognized as describing brittle fracture of
mild steel at cryogenic temperatures (see Lin et a1.[33]).

3. EXPERIMENTAL CONSIDERATIONS
This part of the presentation describes the methods that were used to test PMMA in mixed
mode I/II loading. Section 3.1 outlines the rationale for selecting the particular loading configur-
ation that was adopted, and Section 3.2 summarizes the details of how this configuration was
applied to the present study. Section 3.3 describes the photoelastic study carried out to provide
verification that the configuration is suitable. Details concerning specimen preparation are given in
Section 3.4, while details concerning actual testing procedure are given in Section 3.5. The results
of the testing are given in Section 3.6.

3.1. Selecting a loading conjiguration

In choosing a suitable loading configuration for the present study, it was desired that a number
of requirements be fulfilled. To begin with, the configuration ideally should be able to provide the
entire range from pure mode I to pure mode II loading. This stipulation is not met by angled central
crack in a uniaxial tensile field because both K, and K,, approach zero as the angle of inclination
approaches zero, and therefore loading at the mode II end of the range cannot be obtained. This
is also true for an angled edge crack subjected to tension or to bending.
Another requirement is that the first term of the series expansion for the Irwin/Williams linear
elastic crack tip field be able to describe adequately the near tip conditions over as large a region
as possible. In this way the crack tip region can be characterized solely by K, and K,, , and there is
no need to resort to including higher order terms of the series expansion. This simplifies the
analyses, and the interpretation of results is hampered less by extraneous uncertainties. Consider-
ation of this requirement calls for an inspection of the relative significance of higher order terms for
different mode I/II loading arrangements.
Using the Cartesian indices x, y for the stress components rrij, and the polar coordinates r, 0 to
denote the position ahead of the crack tip, the linear elastic field may be written to include second
order terms as (cJ Rice[34]):

higher order terms in r (1)

where K is the appropriate stress intensity factor, and fX,(0), f,(e), f,(e) are functions of 8. The
second term of the expansion, L, is independent of r, and depends solely upon the magnitude of the
resolved normal stresses acting parallel to the crack. Included in the value of L is the component
of any remote normal stress resolved parallel to the crack. Also included is the ‘T-stress’ revealed
by the analysis of Irwin[3] and of Williams[4] on cracked bodies. These analyses indicate that a
remote normal stress, 6,) applied in one direction, causes a uniaxial normal stress, T, to develop at
90” to that direction. T Does not manifest itself in the first term of the series expansion, but the
component of T acting parallel to the crack plane appears in the second term. For a crack inclined
to the remote b,, this component has a magnitude of T sin’ /I, and it is apparent that T plays an
important role in the magnitude of the second term. In general, the magnitude of T is found to
depend upon the geometry of the particular loading configuration in question. Therefore, if it is
desired that the second term be insignificant compared to the first term, a suitable loading configur-
ation will have a value of T near to zero. For a large plate, Eftis and Liebowitz[35] show that
T = - IJ~. For an edge cracked plate loaded in uniaxial tension, which is of dimensions similar to
 Fracture behaviourof PMMA 69

Richard’s[ 131mixed mode specimen, Kfouri[36] shows that T N - O.O%, . For a bend specimen of
conventional dimensions, Larsson and Carlsson[37] find that T N O.OSa,.
Practical considerations are also important in selecting a suitable loading configuration. The
test specimens themselves should be relatively easy to machine and prepare for testing. For this
reason, all of the configurations making use of embedded cracks were rejected because of the
difficulties associated with first introducing an embedded slit and then forming a sharp crack at
the slit ends. The specimen developed by Banks-Sills and Bortman[lO] was rejected because of its
complex geometry. Such a geometry would be particularly sensitive to imprecise machining, and
this could lead to inaccuracies in subsequent analysis which assumes the ideal shape. To a lesser
degree, this is also a problem with the configuration of Richard[l3].
One approach which meets all of the requirements outlined above, and which is particularly
straightforward to apply, uses anti-symmetric four point loading developed by Gao et a1.[38] to
obtain mode II and mixed mode I/II, and uses conventional symmetric four point loading to obtain
mode I. In this way the full range from mode II to mode I can be investigated using familiar edge
cracked bend bar specimens, which are simple to machine and to pre-crack, and for which the
effects of higher order terms in the stress distribution are small. This is the approach that has been
adopted for the present study and is described in the next section.

3.2. Anti-symmetric and symmetric four point loading

Figure 2 shows the loading arrangement for anti-symmetric four point (AS4P) and symmetric
four point (S4P) bend specimens, together with the corresponding shear force and bending moment
diagrams. Given that the bending moment, M, is associated with mode I, and that the shear force,

I.
c, (1 n
I
--is0 Specimen
u v u v

k2S-I4 I-S-t-S-l

2P/3 P/3 P/2 P/2

I
IIf Applied
toads

t t
1

P/3 2P/3 Pi2 Pi2

Shear
force
r-F=
-

M=SP/P
Bending
moment

La 1 Ant1 -symmetric four point bend 1b) Symmetric four point bend

Fig. 2. Loadingconfigurationsfor mixed mode I/II testing: (a) anti-symmetric four point bend; and
(b) symmetric four point bend.
 T. M. MACCAGNO and J. F. KNOTT

Table I, Calibration functions for four

point bend configuration (after Wang
et a1.[39])

0.20 4.97 0.496

0.25 5.67 0.667
0.30 6.45 0.857
0.35 7.32 I .080
0.40 8.35 I.317
0.45 9.60 1.557
0.50 11.12 1.838
0.55 13.09 2.125
0.60 15.66 2.441
0.65 19.17 2.794

Q, is associated with mode II, it is evident that for the AS4P case (Fig. 2a) the ratio of mode I to
mode II varies with the positioning of the crack relative to the load points. Along the centre axis
of the loading arrangement M = 0, yet there is a substantial Q. Therefore, a crack positioned
exactly on the central axis is loaded in pure mode II. If the crack is positioned away from the centre,
Q remains constant and M increases, and therefore the mode I to mode II ratio increases. Pure
mode I is obtained from the S4P loading arrangement by positioning the crack in the central region
where Q = 0, and therefore where there is no mode II (Fig. 2b).
According to Wang et a1.[39], K, and K,, for a straight edge crack in a plate can be determined
from M and Q through the expressions:
M
K, = -~~3’2 ‘1’
K,, = - Q
BJ..j7’/2
Y”

where W is the width of the plate, B is the plate thickness, and Yi and Y,, are the mode I and mode
II calibration functions, respectively. Note that the Y-calibrations are functions of the ratio of
crack length to specimen width (a/W). Because bend bar specimens are widely used in pure mode
I fracture testing, Y, has been determined with great accuracy by a number of independent
investigators including Brown and Srawley[40], Wilson[41], and Benthem and Koiter[42]. The
results of all these investigators agree to within I %. Wang e? at. have determined Y,, by a boundary
collocation method for 0.40 < (a/W) < 0.60, and by a method which relates Y,, to Y,. The two
methods are found to agree within about &5%. Table 1 reproduces the values for Y, and Y,, as a
function of (a/W) as given by Wang et al.
At this point it is appropriate to introduce a convention used when discussing mixed mode I/II
situations. For the case of a central crack in a large plate loaded by a remote uniaxial tensile stress,
(T,, oriented at an angle /3 with respect to the axis of the crack, Ki and Ki, are given by[5]:
K, = B, & sin’ /I, Kii=cr,& sin@cosjI (3)
where a is half the length of the crack. By re-arranging this equation, the ratio of Ki to & can be
related to /I by:
Kl
-= tan/? (4)
&I
and it is apparent that the factor tan /I provides a convenient measure of the mode I to mode II ratio
regardless of the testing configuration employed. Hence, for situations which do not actually use the
inclined crack configuration:
4

where & is the ‘equivalent crack angle’.

&=tan-’
0 -
k;,
(5)

Returning once again to the bend bar situation, M and Q in eq. (2) are determined from the
applied load P, and from the relevant dimensions for each of the loading arrangements, For the S4P
 Fracture lxhaviour of PMMA 71

arrangement only K, is determined, and in this case:

P
b&ST (6)

where S is the lateral distance between the upper load point and the lower load point (see Fig. 2b).
With the AS4P arrangement (Fig. 2a) both Q and M are required, and they are given by:

The position of the crack, S,, which gives the desired K,/K,, can be determined by combining
eqs (2) and (7):
& = ~tanB&L/&) (8)
where tan & is equal to K,/K,, , as described above. In the present study, tests were carried out at
j3q = 90” (mode I), 75”, 60”, 45”, 30”, 15”, 0” (mode II). The loading point positions were marked
onto the specimen, with all measurements made relative to the tip of the sharp crack.
Figure 2 also shows the details of how the load is actually applied to the specimen. In both the
AS4P and S4P cases, two pin rollers are placed on the lower load plate, into grooves which have
been machined carefully at precisely the required distance apart. The specimen is placed onto the
rollers and another load plate/roller combination is placed onto the specimen. The rollers are
carefully aligned with the load point positions marked on the specimen. One more roller is placed
on top of the upper load plate and is positioned exactly on the central axis of the entire arrangement
by another locator groove. The diameter of the rollers was chosen to be relatively small (approxi-
mately 6 mm), in order to minimize the area over which the load was applied to the specimen.
From the discussion above it is apparent that the precise location of the tip of the crack is
critical in determining the ratio of mode I to mode II, and therefore a great deal of care must be
taken in this procedure. Even so, it was felt that the ~sitioning of the crack tip relative to the
central axis of the loading configuration (i.e. S,) can only be accurate to within about 0.3 mm.
Inspection of Fig. 2(a) shows that imprecision in S,, leads to more (or less) A4 than accounted for,
but does not affect Q. The error in & due to this imprecision in S, can be estimated. For a function
Z =f(A) the error in Z, AZ, can be determined from the error in A, AA, by the expression:

AZ = AAF (9)
dA ’
Applying this to the present situation where peg= tan-‘(K,/&,), the error in &,, A&, is:

(10)

Rearranging eq. (8) gives A(&/&,) in terms of AS, as:

(111
For the values AS, = 0.3 mm, W = 20 mm, and (a/W) = 0.5, eq. (10) becomes:
0.09
“=I = (1 + tan2 &) ’ (12)
and it is apparent that the error depends upon the ratio of mode I to mode II chosen. For &, = 75”,
A& = &0.35”, while for &, = 15”, A/& = i-4.8”.
3.3. Validation of conJguration
In order to prove that the AS4P load arrangement gives various mixtures of mode I to mode II
by simply altering the crack position relative to the loading points, a stress birefringence investi-
gation was carried out on bend specimens made from the photoelastic material polycarbonate.
Certain transparent materials exhibit birefringence under polarized light and the birefringence
EFM 34,1-E
12 T. M, MACCAGNO and J. F. KNOTT

pattern is intimately connected with the stress/strain field existing in the material. This behaviour
has been used by a number of investigators to obtain the stress distributions in engineering
components of complex geometry, and also to obtain Ki and Kir for geometries containing cracks
(see Sanford and Dally[43]).
The dimensions of the photoelastic specimens were 20 mm wide by 100 mm long, and were cut
from 3-mm thick polycarbonate sheet produced by ERTA. Polycarbonate was chosen as the
photoelastic material because it produces many birefringence lines for a relatively low load (see
Dally and Riley[44]), and because it was available in uniform thickness. The relatively small thick-
ness value was chosen so that there would be minimal residual stress present from manufacturing.
Care was taken also to minimize the residual stress introduced during machining. The specimens
were first cut to the approximate size, and then milled to the desired dimensions, with only a small
amount of material (about 0.05 mm) removed at each pass.
Ideally the investigation was to be done for a sharp crack, but poly~arbonate tends to yield
rather than crack, and introducing a sharp crack was found to be difficult. Andrews[45] observes
that polycarbonate undergoes a ‘ductile to brittle’ transition near -200°C and it was hoped that,
in a manner similar to notched PMMA at room temperature, notched polycarbonate specimens
cooled to - 196°C would form natural cracks by pressing a razor blade into the notch. Even at
this very low temperature however, the notch deformed plastically under the razor blade rather
than cracked. Fatigue pre-cracking was also tried, but observation under polarized light of the
unstressed pre-cracked specimens revealed that considerable residual stresses developed along
the crack length. In some cases it also appeared that the fatigue crack was not open all the way to
the tip, and therefore when the specimen was loaded the birefringence lines did not even develop at
the crack tip. To circumvent these problems, it was decided instead to machine a very thin slit
(0.15 mm) into the specimen and assume that this would be sufficient approximation to a sharp
crack.
Figure 3 shows the results of the investigation. Figure 3(a) shows the birefringence pattern for
a slitted specimen in S4P bending which is loaded to obtain mode I (& = 90”), while Fig. 3(b) and
(c) shows specimens in AS4P bending which are loaded to obtain mixed mode I/II. The actual
mixed mode I/II combinations chosen were & = 63” and /.$s= 27” in order to provide a direct
comparison with results on the shape of the stress/strain fields near a crack tip presented by
Shih[47]. Figure 3(d) shows the pattern for AS4P bending which is loaded to obtain mode II
(& = 0’). It is evident that the changes in the stress/strain field near the tip of the crack in
polycarbonate, for different crack positions relative to the loading points, are consistent with the
changes expected in going from mode I to mode II. That is, the field in mode I is symmetrical about
the crack tip, but this field becomes more skewed as the mode II component is increased.
These observations coincide well with other similar biref~ngence studies on crack tip stress
fields. Sanford and Dally[43] provide photographs of birefringence patterns for mode I and mixed
mode I/II, while Jones and Chisholm[46] show patterns for mode II. Richard[13] also shows
patterns for mode I, mixed mode and mode II. The present photoelastic patterns can also be
compared with the shape of the elastic-plastic boundary (assuming von Mises criterion) as
determined by Shih[47], and which are shown in Fig. 4. There is good qualitative agreement
between the patterns in Fig. 3 and those in Fig. 4.

3.4. Specimen preparation

Given the fracture toughness, K,, and uniaxial yield stress, o,, of a material, Brown and
Srawley[40] determined experimentally that for a through-the-thickness crack under mode I
conditions, the crack length must exceed about 2-2.5 times (_K,/o,)2, and the crack tip must be
positioned at a distance greater than about (Kc,/ao)2 from the specimen boundaries. They also
found that the thickness of the specimen must exceed about 2-2.5 times (K_,/o~)~ for plane strain
conditions to predominate across the crack front. In order to be conservative and to simplify
matters, Brown and Srawley concluded that each of the relevant dimensions should exceed the
critical dimension 2.5 (K,,/cQ2. For PMMA at room temperature, (Tois typically found to be about
80 MPa, and the mode I fracture toughness is typically about 1.6 MPafi (see Williams[48]). Note
that crOin this case is the crazing stress. The critical dimension for mode I testing of PMMA is,
therefore, about 1.Omm.
 Fracture behaviour of PMMA 73

03 c
(bl

06-
(d)
N-m
“< $ 0 4 -

Fig. 4. Shapes of plastic zone (after Shih[47]): (a) mode I (J, = 90”); (b) mixed mode I/II Cj!&
= 67”);
(c) mixedmode I/II (8, = 27”);and (d) mode II (B, = 0”).

Unfortunately there is no similar guideline for mode II testing, although an estimate can be
made following the principle that the mode I guideline exists in order to ensure that the plastic zone
size is small relative to the specimen dimensiotis. A critical dimension for mode II can be
determined, using knowledge of the extent of the plastic zone size for mode II, as compared with
the mode I plastic zone size. The results of Shih[47] presented in Fig. 4 provide this information for
materials with a plastic zone produced by shear stress. From this figure it is apparent that if &
is approximately equal in magnitude to K,,,, the mode II plastic zone extends to a distance about
four times as a far as the mode I plastic zone. The plastic zone in PMMA, however, is produced
by normal stress, and it is not obvious that Shih’s findings can be rigourously applied to the present
problem.
Further consideration though, suggests that Shih’s results can be used at least qualitatively. It
is not surprising that the maximum extent of a shear stress plastic zone will increase as the mode
II (i.e. shearing) component is increased. It may be anticipated, however, that a normal stress
plastic zone will be affected comparatively less by increasing the mode II component. Hence, a
mode II plastic zone size estimated by Fig. 4 can be expected to be conservative for the present
problem. Using this reasoning, K validity will be ensured for mode II testing of PMMA if the
relevant specimen dimensions exceed about 4 mm.
For the present study, testing was carried out on 5-mm thick ‘Perspex’ PMMA sheet produced
by ICI. The specimens were 20 mm wide by about 100 mm long, and were all machined from the
same sheet. A l-mm wide slit was machined to a depth of 4 mm in one edge of the specimen. A razor
blade was then slowly pressed into the bottom of the slit, and a natural crack formed to a depth of
about 10 mm. The cracks formed in this manner generally propagated straight down from the slit
and maintained a relatively straight profile through the thickness. Excessively curved or bowed
cracks were not tested. Inspection of the crack tips under an optical microscope revealed that the
cracks were sharp, and that a single craze ran ahead of the tip. A typical example is shown in Fig. 5.
74 T. M. MACCAGNO and J. F. KNOTT

Note that in this figure a groove runs parallel to the direction of the crack/craze. Although at this
magnification the groove appears somewhat like another crack/craze, it is in fact produced simply
by lateral contraction at the specimen surface, which occurs in response to the elongation
perpendicular to the crack plane (see Kambour and Robertson[32]).

3.5. Testing procedure

All mixed mode I/II testing was carried out on the S4P and the AS4P loading systems
described in Section 3.2. The mode I/II combinations investigated were: /I,, = 90” (mode I), 75”,60”,
45, 30”, 15, 0” (mode II). Recall that accurate location of the loading points is essential to obtain
the desired combination of shear force and bending and therefore great care was taken in this
procedure. These locations were all measured relative to the tip of the crack.
The tests were carried out at 20°C on a screw-driven Schenck universal testing machine of
50 kN capacity, under displacement control. The mechanical properties of polymers are found to
be very much dependent on the rate of testing. Therefore, a constant crosshead displacement rate
of 5 mm/min was selected, in order to compare directly the present results with the bend test results
on PMMA of Ewing et a1.[12]. A trace of load vs time for each test was made in order to provide
a permanent record and to given an indication of the relative overall material behaviour.

3.6. Results of testing

The load vs time trace for each specimen was linear all the way to the instant of sudden
catastrophic propagation of the crack and no stable crack growth was observed prior to this
instant. This suggest that all specimens fractured in a brittle manner. The load at fracture, PI, was
used to calculate the mode I and mode II stress intensity factors at failure, K,, and Knr, according
to eq. (2). The calculations are summarized in Table 2. Note that the average value of K,, at
P,, = 90 ’ is 1.87 MPa&. This compares with a K,, at j.& = 90” result of 1.85 MPa& obtained by
Ewing et af.[12] for an edge cracked bend specimen made from the same material, and tested at a
similar loading rate and temperature.
Figure 6 shows some specimens which have been tested through the entire range from mode
I to mode II. It may be seen that the introduction of a mode II component to the loading causes
the crack to extend at an angle to the original crack plane. Furthermore, this angle increases with
increasing mode II. The angle between the original crack plane and the initial portion of the
subsequent crack path was measured for each specimen using a microscope fitted with a rotating
stage. This initial fracture angle, &,, could be measured to within about 2”. Each specimen provides
four such angles and the variation was usually less than about 5”. The average of these 0, values for
each specimen is also presented in Table 2. All measurements were made at a microscope
magnification of 25 x .

Table 2. Summary of mixed mode I/II testing on PMMA

W B a SO
No. (mm) (mm) (mm) alW (mm) $I) (MP:&) (MP:& (!$
P-l 20.0 4.9 7.2 0.36 - 0.352 1.912 o.ooo 90.0 0.0
P-19 20.0 4.9 8.8 0.44 - 0.259 1.750 0.000 90.0 0.0
P-20 20.0 4.9 8.0 0.40 - 0.306 1.845 0.000 90.0 0.0
P-21 20.0 5.0 9.3 0.46 - 0.281 1.969 0.000 90.0 0.0
P-24 20.0 5.1 10.8 0.54 12.0 0.506 1.782 0.484 74.8 30.8
P-40 20.0 5.1 7.4 0.37 11.5 0.767 1.576 0.417 75.2 27.4
P-25 20.0 5.1 10.4 0.52 6.5 0.874 1.564 0.789 63.2 44.2
P-41 19.3 4.9 9.9 0.51 5.8 0.852 1.443 0.791 61.3 40.0
P-37 20.0 5.0 10.2 0.51 3.4 1.142 1.053 1.020 45.9 62.5
P-42 19.3 4.9 10.1 0.52 3.3 1.063 1.060 1.017 46.2 60.0
P-38 20.0 4.8 10.3 0.51 1.9 1.439 0.773 1.339 30.0 61.5
P-43 19.9 5.1 9.9 0.50 1.9 1.547 0.761 1.317 30.0 67.5
P-39 20.0 4.8 9.7 0.49 1.0 1.446 0.384 1.265 16.9 70.2
P-44 19.4 5.0 10.0 0.52 1.0 1.429 0.420 1.336 17.4 70.0
P-45 19.8 4.9 12.3 0.62 0.3 1.226 0.153 1.530 5.7 70.8
P-46 19.9 4.8 9.5 0.48 0.0 2.082 0.000 1.769 0.0 69.2
 Fracture behaviour of PMMA 75

Fig. 3. Results from photoelastic study: (a) mode I (Bog= 90”); (b) mixed mode I/II (g, = 67O);(c) mixed
mode I/II (& = 27”); and (d) mode II (& = O”).
16 T.M. MACCAGNO and J.F. KNOTT

Mode 11
component
increasing

Fig. 5. Crack tip region in PMMA showing a single craze at the tip of the crack, and showing a groove
produced by lateral contraction at the specimen surface.
Fig. 6. Mixed mode I/II specimens of PMMA after testing.
 Fracture behaviour of PMMA 11

‘fable 3. Maximum G dim&ion, es, asa function of equivalent crack

angle, 8,; and coefficients C,, , Cl2, C, as a function of angular
position, fI (after Hyashi and Nemat-Nasser(l8])

90 0.0 0.0 1.000 0.000 1.000

81 17.36 7.2 0.992 0.250 1.011
72 31.72 14.4 0.969 0.490 1.042
63 42.48 21.6 0.931 0.710 1.092
54 50.44 28.8 0.880 0.904 1.156
45 56.54 36.0 0.819 1.064 1.230
36 61.45 43.2 0.749 1.185 1.316
27 65.54 50.4 0.672 1.264 1.378
18 69.20 57.6 0.593 1.301 1.441
9 72.48 64.8 0.513 1.297 1.488
0 75.74 72.0 0.435 1.256 1.513
79.2 0.360 1.182 1.514

4. ANALYSIS OF RESULTS

4.1. Fracture angle

The fracture angle, 0,) vs the relative amount of model I to mode II loading, &, for the present
work is plotted in Fig. 7. Also plotted in this figure are the predictions for $ vs & according to
the three main hypotheses for mixed mode I/II fracture. The prediction for the maximum rreee
hypothesis is that obtained by Erdogan and Sih[S]. The prediction for the maximum G hypothesis
is given by the numerical results of Hyashi and Nemat-Nasser[ 181,developed from an analysis for
a crack with an infinitesimal kink at one end. These results are reproduced in Table 3. The
prediction for the minimum S hypothesis is that obtained by Sih[l9] using a value of v = 0.4 for
PMMA. It is apparent that the three predictions lie quite close to each other and that the
experimental points appear to follow any of the hypotheses to within experimental error.
It is of interest to compare the experimental points of the present work with experimental
points given in the literature and obtained by using other loading configurations to produce mixed
mode I/II. Figure 8 plots experimental values of 0, vs & presented in a number of other studies on
PMMA. Erdogan and Sih[S] present in tabular form the results of nine tests using pins located near
the crack and loaded to produce mode II. Each test gave two measures of fracture angle, and B0for
that specimen was taken as the average. Of the nine results, only the specimens that gave the highest
8, and the lowest 8, are presented in Fig. 8. Interestingly enough, Erdogan and Sih do not explicitly
present the 8, results of their inclined central crack tests, but Sih[ 191does in a later presentation. Sih
notes that four tests were performed at each value of /3 investigated, and that each test yielded two

(Mode I) (Mode II)

Equivalent crack afigle & (de@

Fig. 7. Experimental values of $ vs & for PMMA, compared with predictions according to 3 criteria.
 T. M, MACCAGNO and J. F. KNOTT

0 Present v.ork - AMP bend

& Erdogon and Slh 119631 - centre - crocked plot@
A Erdogon and S\h (19631 - PI” - Loaded CrOCk
13 W~ll~oms and Ewing (1972) - Mtre - crocked plate
# Ewng and Wlllroms (1974) - torson tube
ffl Ewing et al. f19761 - edge - cmcked fiate
0 UC& et a1 il983F - centre -crock&d plate

Fig. 8. Experimental values of 6, vs fi, from PMMA testing carried out by numerous workers, compared
with predictions according to 3 criteria.

measures of 0,. Once again however, onfy the highest and lowest average values for 0, at each &
are presented in Fig. 8. Williams and Ewing@] present graphicaliy a large number of values of $
for various values of 8, and many of these data are reproduced in Fig. 8. The precise values for each
data point were obtained by first enlarging the originat Wilhams and Ewing plot, and then
determining the coordinates. A similar procedure was followed for the inclined central crack tests
of Ueda et a1.[7], and for the torsion tube tests of Ewing and William@ I]. Ewing et a1.[12] present
in tabular form 0, for their inclined edge crack tests. However, for presentation in Fig. 8, the value

(ModeI) (Mode II)

Equivalent crack angle &, (deg)

Fig. 9. Experimental values of (0Okat j? )/(cT~ at peq= 90”) vs &, for PMMA, compared with prediction
actor 2.mg to maximum uBBcraterion.
 Fracture behaviour of PMMA 79

for 8, was calculated from the value of Pf given by Ewing et al. and using the mode I and mode II
calibration functions given by Bowie[49] for edge cracked plates.

4.2. ILoad at fracture

All of the hypotheses for mixed mode I/II fracture maintain that fracture occurs when
the appropriate quantity attains a critical value which is constant for the material. Therefore, the
critical quantity measured under mode I conditions should be equal to the critical quantity
measured under any combination of mode I and mode II,
This can be checked for the maximum bggcriterion in a straightfo~ard manner, assuming that
the material can be described by the linear elastic stress field. At fracture, the criticai value of the
maximum tangential tensile stress, a,,, can be expressed as (cj Erdogan and Sih[S]):

(13)

where Krr and K,,, are the values at failure for K, and K,,. Given that tan-‘&/K,,,) expresses the
relative amounts of mode I loading to mode II loading as the parameter &, a useful ratio is
obtained if each value of aok calculated at & is divided by the value of cr,, at 8, = 90” (i.e. the
mode I value). If the subscripts m and f are introduced to identify quantities calculated at #& and
at & = 90”, respectively, such a ratio can be written as:

04)

Thus, if egg, is in fact a constant, and if one assumes that r remains constant throughout the range
from mode I to mode II, then the ratio of (eooCat &) to (cBBC at fies = 90’) will be equal to unity for
all combinations of mode I and mode II. Figure 9 shows the result of this exercise when carried out
on K,, and K,,, obtained from the present work, assuming that 0, follows the maximum treecriterion.
It is apparent that ae,, calculated from the data remains essentially constant throughout the range
from mode I to mode II.
The same exercise can be carried out for expe~mental points from other investigations on
PMMA. The result is presented in Fig. 10. Once again, it is seen that the experimental points of the

s grn .8*~
$
b- 1.0 D 0
;: B D 0
0 8 a o@
2 V %
z
,g ::

- 05- 0 Present wwk - AS4P bend

A Erdoqan and Slh (1963) - pin-Loodedcmck
0 WINlams and Ewing 11972) - centre-crocked plate
9 Ewing and Willloms (1974) - torsion tube
a EWlnQ et 01. (19761 - edge-cracked plate
V Ueda et al. (1983) - centre - crocked @ate
0* I I I 1 I
90 75 60 45 30 15
.. _.
(Model) &lode XI)
Equlvaknt cmck angle & MegI

Fig. IO. Experimental values of (owe at ~~)/(u~ at & = 90°) vs & from PMMA testing carried out by
numerous workers, compared with prediction according to maximum Q, criterion.
 T. M. MACCAGNO and J. F. KNOTT

75 60 45 30 I5
(Mode I) (Mode II)
Equivalent crack angle 8, (deg)

Fig. Il. Experimental values of (G, at &)/(G, at & = 90”) vs & for PMMA, compared with prediction
according to analysis of Hyashi and Nemat-Nasser[ 181.

present work lie comfortably within the scatter of the experimental points of previous work. All of
the data support the proposition that bgO,is a constant for mixed mode I/II.
The other fracture hypotheses can be tested in a similar manner. Many investigators, including
Palaniswamy and Knauss[ 171,and Hyashi and Nemat-Nasser[ 181,maintain that G can be expressed
in terms of K, and K,, as:
G = C,,K:+C,,K,K,,+Cz2K;, (15)
where C,, ,C,2, C,,are numerically determined coefficients. Table 3 gives these coefficients, and
gives the direction of maximum G as a function of &, according to Hyashi and Nemat-Nasser.
Thus G, can be calculated using K,,and K ,,1, and using Table 3. The result is shown in Fig. 11 for
all experimental points for PMMA. Note that in a manner similar to that used above for bgsc, G,
evaluated at & is normalized by G, evaluated at & = 90”. From this figure, it is apparent that there
is considerable scatter in G, calculated from eq. (15). It is difficult to say whether G, remains
constant in going from mode I to mode II.
Sih[l9] maintains that the critical value of the strain energy density factor, S,, can be related
to K,r, K,,,and f& through the expression:

Sc = ; @I,K:r+ 2% WG, + ~,,K:,,) (16)

where:

Q,, =&(I +cosB,)(lc -cose(J

= 1 sin 8,[2 cos 8, - (K - l)]

“’ 16~

- cos e,) + (I + cos e,)(3 cos e. - I)]

and where ICis (3 - 4v) for plane strain and (3 - v)/(l + v) for plane stress. Note that there is a
dependence on Poisson’s ratio v. For PMMA there appears to be a variation in v quoted in the
literature, and therefore for the results of the present work, the ratio of (S, at /?,,) to (S, at & = 90’)
is calculated for two values of Poisson’s ratio: v = 0.4 (Fig. 12) and v = 0.3 (Fig. 13). It is apparent
 Fracture behaviour of PMMA

I I I I I
01
90 75 60 45 30 15

(Mode I) (Mode II)

Equivalent crack angle 8, (de-g)

Fig. 12. Experimental values of (SCat &)/(S, at j?, = 90”) vs & for PMMA, compared with prediction
according to minimum S hypothesis for Y = 0.4.

that both figures show similar trends, and that SCdoes not remain constant in going from mode I
to mode II for either value of v.

5. DISCUSSION
5.1. Validation of anti-symmetric four point loading
The results of the present work on PMMA support the use of the AS4P load arrangement to
study mixed mode I/II brittle fracture. When compared with experimental points for PMMA which

90 75 60 45 30 15
(Mode I) (Made III
Equivalent crack angle & (dag)

Fig. 13. Experimental values of (SCat &)/(& at & = 90”) vs &, for PMMA, compared with prediction
according to minimum S hypothesis for v = 0.3.
82 T. M. MACCAGNOand J. F. KNOTT

have been obtained using numerous other loading configurations (Figs 8 and 10) it can be seen that
the present points lie comfortably within the scatter. This provides further support, in addition to
the findings of the photoelastic study described in Section 3.3, that eq. (2) correctly determines K,,
and K,,,.
It is also worthwhile to emphasize that this method of obtaining mixed mode I/II is very
straightforward to apply, and can be used in conjunction with conventional symmetric four point
bending to investigate the entire range from mode I and mode II.

5.2. Mixed mode Illlfracture of Ph4MA

Section 2 outlines the sequence of events which is thought to describe brittle fracture of
amorphous glassy polymers in general. This section discusses whether such a fracture mechanism
can be applied to the present findings concerning mixed mode brittle fracture of PMMA.
From Fig. 6 it is apparent that crack propagation under mixed mode I/II occurs at an angle
to the plane of the original crack. When these specimens are observed under high magnification
(Fig. 14), it is seen that the extension has occurred directly from the tip of the crack, rather than
from the tip of the craze region that was present prior to fracture. It is concluded that this craze
region does not play any role in the fracture process of cracks which propagate at an angle.
From the discussion in Section 2, however, some manner of craze formation is expected as an
intermediate step of crack propagation. In the present situation this implies that a craze forms at
an angle to the plane of the original crack and that it is this craze which provides the path along
which the subsequent fracture occurs. No craze formation immediately prior to fracture was
observed in the present experiments, but in their mixed mode I/II tests on PMMA, Williams and
Ewing[6] noted ‘stable crack propagation’ prior to the fracture event. It is possible that this stable
crack propagation is actually craze formation. In addition, craze formation immediately prior to,
and in the direction of, crack propagation is reported by Ishikawa et al.[50] in blunt notch
specimens of PMMA. These investigators note that the stress at which craze formation is observed
is only slightly less than the stress at which fracture occurs. This fact, and the fact that the tests of
Williams and Ewing were carried out at a slower cross head displacement rate, suggests that
perhaps in the present tests the craze formation and subsequent crack propagation occurred
virtually simultaneously, and therefore were not separable events.
In an effort to establish whether or not crazing can occur at an angle to the original crack
plane, another series of tests were performed on PMMA, but this time a ‘crazing agent’ (methanol)
was introduced at the crack tip to promote crazing at relatively low loads. Such ‘environmentally
assisted’ crazing is treated in detail by Williams[48]. The specimens were prepared exactly as before,
and also loading to the same mode I/mode II ratios as before, but stressed to only 50% of the
applied load which would cause crazing/cracking without a crazing agent. Upon loading, the load
vs time trace was allowed to stabilize for some minutes and then the methanol was squirted into the
crack. A ‘wedge’ of very fine crazes emanating from the crack tip was observed to form almost
immediately, without subsequent crack propagation. A typical specimen is shown in Fig. 15. This
specimen was loaded at a mode I/mode II ratio of /I, = 60”, and it is apparent that the wedge of
crazes occurs at an angle to the plane of the original crack. The reason that a wedge of crazes forms
in the first place is that at this level of applied loading the criterion for crazing in methanol is met
over a region ahead of the crack tip, rather than only in a single direction ahead of the crack tip.
It is expected that this wedge would be narrower if the applied loading was decreased to a lower
level.
Figure 15 establishes conclusively that crazing can occur at an angle to the original crack
plane. In fact, if the angles of the two crazes which bound the wedge are measured, the average of
these two angles coincides very closely with the plane of maximum gee. This is shown in Fig. 16,
together with the average angle of the wedge for all of the other specimens tested in methanol.
Recalling from Fig. 7 that the angle of crack propagation also coincides very closely with the plane
of maximum uee, it appears that not only does crazing occur at an angle to the plane of the original
crack, this craze angle also coincides with the angle along which crack propagation occurs.
It is concluded that the present results for mixed mode I/II brittle fracture of PMMA are
consistent with the fracture mechanism outlined in Section 2.
 Fracture behaviour of PMMA 83

‘.

# 50 pm
‘15 I

Fig. 14. Crack tip region of mixed mode specimen loaded to & = 45”.
Fig. 15. Crack tip region of specimen tested in methanol (& = 60”).
 Fracture behaviour of PMMA

(ModeI) WcdeIf)
Equivalent crack angle @, (deg)

Fig. 16. Experimental values of 0, vs & for PMMA tested in methanol, compared with plane of
maximum uo8.

5.3. Criterion for fracture

From Fig. 10, it is apparent that the present results, and the results obtained using numerous
other configurations, are described very well by the maximum tree criterion first proposed by
Erdogan and Sih. That is, mixed mode I/II crack propagation in PMMA occurs from the crack tip
in the radial direction along which the maximum resolved tangential tensile stress, crgg,is maximum.
This criterion for fracture is consistent with the fracture mechanism that was proposed in
Section 2, and which was shown to be applicable to the present results in Section 5.2. A craze forms
from the crack tip in the radial direction along which the tensile strain is maximum, and since
PMMA is reasonably homogeneous and isotropic, this is also the direction along which the tensile
stress is maximum. Crack propagation is the simultaneous rapid formation and breakdown of this
craze, which is initiated when the local tensile stress exceeds the level that can be supported by the
craze fibrils.

6. CONCLUSIONS
(1) The combination of using both anti-symmetric and symmetric four point bend arrange-
ments is a convenient and easily applied method to study brittle fracture throughout the entire
range from mode I to mode II.
(2) A maximum tangential stress criterion based on the linear elastic stress field provides a very
good description for mixed mode I/II brittle fracture of PMMA at room temperature.
(3) The criterion is consistent with the present understanding of the micro-mechanisms
involved in fracture of PMMA.

Acknowledgements-The support for one of the authors (TMM) by the Edmonton Churchill Scholarship Foundation is
gratefully acknowledged. The authors also wish to thank Professor D. Hull for provision of research facilities.

REFERENCES
[I] G. R. Irwin, Fracture, in Encyclopedia of Physics (Edited by S. Flugge), Vol. 6, 551-590, Springer, Ikrlin (1958).
[2] T. M. Maccagno, Fracture in mixed modes I and II. PhD thesis, Dept of Materials Science and Metallurgy, Univ. of
Cambridge (1987).
[3] G. R. Irwin, J. appl. Mech. 24, 361-364 (1957).
[4] M. L. Williams, J. appl. Mech. 24, 109-114 (1957).
[5] F. Erdogan and G. C. Sih, J. Basic Engng 85, 519-527 (1963).
[6] J. G. Williams and P. D. Ewing, Int. J. Fracture Mech. 8, 441-446 (1972).
[7] Y. Ueda, K. Ikeda, T. Yao and M. Aoki, Engng Fracture Mech. 18, 1131-1158 (1983).
[8] P. S. Theocaris, Engng Fracture Mech. 19, 975-991 (1984).
86 T. M. MACCAGNO and J. F. KNOTT

[9] S. K. Maiti and R. A. Smith, J. Mech. Phys. Solids 31, 389-403 (1983).
[IO] L. Banks-Sills and Y. Bortman, Inc. J. Fracture 30, N-201 (1986).
[I I] P. D. Ewing and J. G. Williams, Int. J. Fracture 10, 537-544 (1974).
[I21 P. D. Ewing, J. L. Swedlow and J. G. Williams, Inr. J. Fracture 12, 85-93 (1976).
[13] H. A. Richard, Some theoretical and experimental aspects of mixed mode fractures”. Proc. 6th Inr. Conf on Fracture,
New Delhi, Vol. 5, pp. 3337-3344. Pergamon Press, Oxford (1984).
[14] S. K. Maiti and R. A. Smith, Inf. J. Fracfure 23, 281-295 (1983).
[15] A. A. Griffith, Phil. Trans. R. Sot. 221, 163-198 (1920).
[16] M. A. Hussain, S. L. Pu and J. Underwood, Strain energy release rate for a crack under combined mode I and mode
II. ASTM STP 560,2-28 (1974).
[17] K. Palaniswamy and W. G. Knauss, On the problem of crack extension in brittle solids under general loading, in
Mechanics Today-Vol. 4 (Edited by S. Nemat-Nasser), pp. 87-148. Pergamon Press, Oxford (1978).
[18] K. Hyashi and S. Nemat-Nasser, J. uppl. Mech. 48, 520-524 (1981).
[19] G. C. Sih, Int. J. Fracture 10, 305-321 (1974).
[20] A. de S. Jayatilaka, I. J. Jenkins and S. V. Prasad, Determination of crack growth in mixed mode loading system.
Proc. 4th Inr. Conf Fracture, Waterloo, Vol. 3, pp. 15-23. Pergamon Press, New York (1977).
[21] G. C. Sih and E. T. Moyer, Jr, Engng Fracfure Mech. 18, 731-733 (1983).
[22] P. S. Theocaris and N. P. Andrianopoulos, Engng Fracture Mech. 20,691-694 (1984).
[23] G. C. Sih and D. Y. Tzou, Engng Fracture Mech. 21, 977-981 (1985).
[24] R. N. L. Smith, Engng Fracture Mech. 26, 4633469 (1987).
[25] R. W. Hertzberg, Deformation and Fracture Mechanics of Engineering Materials, 2nd edn, pp. 218-222. Wiley, New
York (1983).
[26] R. J. Kramer, Microscopic and molecular fundamentals of crazing, in Crazing in Polymers-Aduances in Polymer
Science (Edited by H. H. Kausch), Vol. 52/53, pp. l-56. Springer, Berlin (1983).
[27] W. Doll, Optical interference measurements and fracture mechanics analysis of crack tip craze zones, in Crazing in
Polymers-Advances in Polymer Science (Edited by H. H. Kausch), Vol. 52/53, pp. 105-168. Springer, Berlin (1983).
[28] R. J. Oxborough and P. B. Bowden, Phil. Mug. 28, 547-559 (1973).
[29] A. M. Donald and R. J. Kramer, Phil. Mag. 43A, 857-870 (1981).
[30] R. J. Young, Introduction to Polymers. p. 285. Chapman-Hall, London (1981).
[31] E. H. Andrews and P. E. Reed, Molecular fracture, in Deuelopmenrs in Polymer Frucrure--I (Edited by E. H.
Andrews). Appl. Sci. Pub., London (1979).
[32] R. P. Kambour and R. E. Robertson, The mechanical properties of plastics, in Polymer Science (Edited by A. D.
Jenkins), Vol. 1, p. 811. N. Holland, London (1972).
[33] T. Lin, A. G. Evans and R. 0. Ritchie, Metall. Trans. A 18A, 641-651 (1987).
[34] J. R. Rice, J. Mech. Phys. Solids 22, 17-26 (1974).
[35] J. Eftis and H. Liebowitz, Int. J. Fracture Mech. 8, 383-392 (1972).
[36] A. P. Kfouri, Inr. J. Fracture 30, 301-315 (1986).
[37] S. G. Larsson and A. J. Carlsson, J. Mech. Phys. Solids 21, 263-277 (1973).
[38] H. Gao, Z. Wang, C. Yang and A. Zhou, Acta Merall. Sinica IS, 380-391 (in Chinese) (1979).
[39] K. J. Wang, C. L. Hsu and H. Gao, Calculation of stress intensity factors for combined mode specimens. Proc. 4th Int.
Conf: Fracture, Waterloo, Vol. 4. pp. 123-133. Pergamon Press, N.Y. (1977).
[40] W. F. Brown and J. E. Srawley, Plane strain crack toughness testing of high strength metallic materials. ASTM STP
410, l-65 (1966).
[41] W. K. Wilson, Engng Fracture Mech. 2, 169-171 (1970).
[42] J. P. Benthem and W. T. Koiter, Asymptotic approximations to crack problems, in Methods of Analysis and Solutions
lo Crack Problems (Edited by G. C. Sih), pp. 131-178. NoordholT, Leyden (1973).
[43] R. J. Sanford and J. W. Dally, Engng Fracture Mech. 11, 621-633 (1979).
[44] J. W. Dally and W. F. Riley, Experimenkd Stress Analysis, p. 486. McGraw-Hill, Tokyo (1978).
[45] E. H. Andrews, Fracfure in Polymers, pp. 52-53. Oliver & Boyd, London (1968).
(461 D. L. Jones and D. B. Chisholm, Engng Fracture Mech. 7, 261-270 (1975).
[47] C. F. Shih, Small scale yielding analysis of mixed mode plane strain crack problems. ASTM STP 560, 187-210 (1974).
[48] J. G. Williams, Fracture Mechanics of Polymers, p. 127 and p. 129. Ellis-Horwood, Chichester, U.K. (1984).
[49] 0. L. Bowie, Solutions of plane crack problems by mapping technique, in Methods of Analysisand Solutkm to Crack
Problems (Edited by G. C. Sih), pp. l-55. Noordhoff, Leyden (1973).
[SO] M. Ishikawa, H. Ogawa and I. Narisawa, J. Mucromol. Sci.-Phys. 19B, 421443 (1981).

(Received 15 September 1988)