Professional Documents
Culture Documents
Fracture Behavior Epoxy
Fracture Behavior Epoxy
Amar C. Garg
Department of Aeronautical Engineering, Indian Institute of Technology,
Powai, Bombay 400076 (India)
and
Yiu-Wing Mai
Department of Mechanical Engineering, University of Sydney,
Sydney, NSW 2006 (Australia)
SUMMARY
1 INTRODUCTION
Epoxies are among the best polymeric materials being used in fiber-
reinforced composites and in many structural parts. They are also used as
179
Composites Science and Technology 0266-3538/88/$03.50 © 1988 Elsevier Applied Science
Publishers Ltd, England. Printed in Great Britain
180 Arnar C. Garg, Yiu-Wing Mai
*James and Harvill is° have recently shown that for wing structures a further weight saving
of some 8-12% can be achieved if the allowable strain can be increased from 0.4% to 0.6%.
182 Amar C. Garg, Yiu-Wing Mai
2 EPOXIES
For example, when curing agents such as aromatic amines, acid anhydrides,
BF3MEA and dicyandiamide are used with the right epoxy resins and fully
cured, the epoxies attain high glass transition temperature, T~, and are thus
suitable for high-temperature applications.
For aerospace applications, only the resins that can withstand high
temperatures (~ 177°C) are acceptable. Such epoxies are polyglycidyl ethers
of the Novolacs (Union Carbide ERR 0100), triglycidyl p-amino phenol
(Union Carbide ERLA 0510) and tetraglycidyl diamino diphenyl methane
(TGDDM, Ciba-Geigy MY 720). The suitable curing agents are dicyan-
diamide (DICY), BFaMEA, diamino diphenyl sulfone (DDS, Ciba-Geigy
EPORAL), methyl dianiline (MDA), meta phenylene diamine (MPDA) and
piperidine. The chemical compositions of such epoxies and curing agents are
presented in Appendix A.
During the curing of the epoxide resins, the crosslinking between the
epoxy molecules and reactive groups on each end of the curing agent occurs.
The amount of crosslinking affects the resin properties significantly. Lower
crosslink density improves toughness by permitting greater elongation
before breakage and also results in reduced shrinkage during cure. On the
other hand, higher crosslink density yields an improved resistance to
chemical attack and raises Tg but lowers the failure strain.
Two major toughening methods have been considered for the epoxies. The
details of these are described below.
attempts have been made in using siloxane and fluorocarbon elastomers for
toughening. 6s.68
To manufacture a toughened epoxy resin, the rubber is first dissolved in
the resin at 50-60°C and the solution is subjected to vacuum to remove
trapped air bubbles. The hardener is then dissolved with stirring, and the
resin-rubber-hardener mixture is cast. A second application of vacuum
removes air bubbles entrapped at this stage, but care is necessary with a
volatile hardener such as piperidine. The resin is then cured following a
standard cure cycle.
In the modified epoxide resin, the size or diameter, d, and volume fraction,
Vp, of the dispersed rubber particles depend on the curing process, rubber
content and acrylonitrile content in the rubber. Slow-cure resin systems tend
to form large rubber particles between 1 and 10~tm in diameter. In some
cases, smaller particles less than 0.5/~m in diameter are also formed, s°
probably at the later stages of the cure reaction when molecular mobility is
reduced. Rapid-cure resins tend to form small rubber particles. An increase
in rubber content leads to an increase in d a n d Vp.s7"69 However, rubber as a
dispersed phase occurs only when the rubber content is less than 20 wt %.
Above this concentration rubber blends with epoxy 16 or it acts as a matrix
and the epoxy separates as spherical particles, s° Depending on the solubility
characteristics of the epoxy resin, increasing the acrylonitrile content over a
certain range reduces Vp and d. Also the fracture toughness of modified
epoxy is affected by acrylonitrile content. This is shown to be highest with
acrylonitrile contents between 12 and 1 8 w t % . 6a
3 P R O P E R T I E S OF T O U G H E N E D EPOXIES
reduces the yield strength and increases the linear thermal expansion
coefficient; but it also significantly increases the fracture resistance. Because
of such a gain in toughness, rubber-toughened epoxies are being used for
various engineering applications.
The changes in properties of a toughened epoxy resin depend upon
various parameters, such as the size and volume fraction of dispersed rubber,
the strain rate, temperature and other environmental factors. Though a
significant amount of information on the physical and mechanical behavior
of toughened thermoplastic materials like high impact polystyrene (HIPS)
and acrylonitrile-butadiene styrene (ABS) is available, 5° the information on
thermoset resins such as epoxies is rather limited. In particular, very little
information is available for high temperature epoxies toughened with
rubber.
T a b l e 115'16 shows the variations of initial tensile modulus, E, tensile
strength, a t, linear thermal coefficient of expansion, ~, and glass transition
temperature, Tg, with elastomer concentration for CTBN-toughened
D G E B A (DER-332 Dow Chemical Co.) epoxide resin. It is seen that E
decreases sharply at first, up to about 4.5% elastomer. Above this
concentration, it remains almost constant up to about 15% CTBN and then
falls sharply again. Similarly, a t and Tg both decrease with increase in
elastomer content (except Tg at 30% CTBN). However, the opposite trend is
true for cc These results are slightly different from the behavior of two similar
CTBN-toughened DGEBA (MY 750 and GY 250 Ciba-Geigy) epoxide
resins studied by Scott and Phillips 3~ and Lahiff, sv who noticed only a
marginal change in E and Gt with rubber content. These properties are also
affected by temperature and loading rates, o t and E are significantly reduced
with increasing temperature. 63,87 The effect of strain rate on E is seen to be
marginal but o-t increases with increasing strain r a t e . 62'63'88,89
TABLE 1
Mechanical Properties of a Rubber-modified Epoxy
(CTBN-toughened D G E B A ) ~6
35
30
25
z
20
1.5
w O.5
~. O0
0 5 10 15 20 25 30
PHR RUBBER
Fig. 1. Variation of specific fracture energy, G~¢, with rubber content. (+) Yee and
Pearson; 8s ( I ) Bascom e t al.; 16 ( 0 ) present work.
-- 3.0
~ 2.5
~ 2.o
g T
to
0.5
o.o
-60 -~0 -20 0 Z0 ~0 60
TEMPERATURE('C)
(a)
t,5
/+.0
-- 3.5
~" 3.0
~- 2.5
~ 20
1.5
05
O0 . . . . .
-60 -40 -20 0 20 /+0 60
TEHPERATURE ('C)
(b)
Fig. 2. (a) Variation of fracture toughness, K=c,with temperature. ( , ) Kinloch et al.; 62 ( 0 )
present work. (b) Variation of fracture energy, Glc, with temperature (GY 250/piperidine/
CTBN X- 13).
TABLE 2
Fracture Toughness for Bulk Resin, as AdhesiVe Film and as
Matrix in Carbon Fiber-reinforced Composites 31
Particulate fillers are often used to reinforce epoxies for various engineering
applications in order to improve several mechanical and physical properties.
The fillers reduce cost, degree of shrinkage, exothermic temperature rise and
coefficient of thermal expansion, and raise the thermal conductivity and T~.
Failure mechanisms in toughened epoxy resins--a review 189
The addition of these fillers to the resin also affects its mechanical properties.
The stiffness, strength and fracture toughness of particle-filled epoxy resins
are higher than those of pure epoxy resin. 77'78 The changes in properties
depend upon the type of filler, particle size, volume fraction, surface
treatment of the particles, strain rate and the test environment. Various
studies of the fracture toughness (in terms of the critical stress intensity
factor, K~¢) of the particle-filled epoxy as a function of particle volume
fraction, V~, and surface treatment during fast 7°'75- 77,81 and slow 71 crack
propagation show that the K~¢ value of the composite is increased. In
addition, increases in Young's modulus, E, and yield stress, ay, have also been
reported. (See Table 3.)
TABLE 3
Influenceof Fiber Particle Sizeand VolumeFraction on E, % and K~ of an
Epoxy-Silica Composite 77
40 -- 1 "90 -- --
(silane-treated)
The recent studies of Moloney et al. 77 have indicated that KI¢ of the
composite is little affected by the type of filler particle and its size (see Table
3). Their studies show that the matrix/particle adhesion, i.e. resulting from
treatment of particles with silane coupling agents, does not improve K~c
appreciably. Similarly, Broutman and Sahu 92 have shown that for a glass-
bead-filled epoxy resin, K~cfor treated and untreated glass beads coated with
silane coupling agent remained the same. By contrast, treatment with
silicone resin doubles the fracture toughness as a result of resin/filler
debonding and resin plasticization by the silicone. 92 In further work on
glass-filled epoxy resin, Spanoudakis and Young 76 show that although Kic is
not much affected by surface treatment the fracture energy, G~c, is. Poor
190 Amar C. Garg, Yiu-Wing Mai
16
50°C
lZ,
12
10 o
OC
08
-70°C
0,6
0~,
02
0 [ . _ ~ [ ~ I I t
01 02 03 0.~, 05 06
VOLUME FRACTION,Vf
(a)
5I 50Oc
4
J
2
x
I L 1 L I
01 02 03 04 05 06
VOLUME Ft~A£TIONVf
(b)
Fig. 3. Fracture energy, G,~, against volume fraction, Vr, of glass particles for (a) an
unmodified epoxy and (b) a rubber-modified epoxy. (After Kinloch et al. 84)
Failure mechanisms in toughened epoxy resins--a review 191
TABLE 4
Mechanical Properties of Hybrid Particulate Epoxies 95
a Weight percentage.
* For a review of the parameters affecting the strength and toughness of particlc-tilled epoxy
resins, see Moloney et al.l 5
192 Amar C. Garg, Yiu-Wing Mai
1 2 0 1 2
A (ram) A (ram) A (ram)
z
o~
coo
2O0
o°
/ ~
z
Q_
~00
2OO
o i
I - - 8 I~
A(mm) A(mm}
Fig. 4. Load (P) versus deflection (A) curves for toughened epoxies associated with the
different types of crack growth in the compact tension geometry. Type C through to type A
occur with increasing temperature and decreasing stress rate of test. (After Kinloch el a/. ('2)
Failure mechanisms in toughened epoxy resins--a review 193
for load control, where C is the specimen compliance defined as the load-
point displacement per unit load. Clearly, dGc/dA is a material property and
gsf is a function of the specimen geometry. The more negative the gsf, the
194 Amar C. Garg,Yiu-WingMai
more stable is the cracking. Characteristic values of gsf have already been
tabulated for a variety of testpiece geometries in Refs 96 and 106. So far two
explanations have been given for unstable 'stick-slip' cracking in epoxies.
Both require eqn (1) to be violated, i.e. dGc/dA is less than the gsf. W h y Gc
changes with A probably explains the mechanism of crack instability. Mai
and Atkins 19'96 first suggested that in terms of crack velocity, A, and crack
acceleration, ,4, eqn (1) can be rewritten as
1 dG~ ~I"
G~ dA A > gsf (3)
If G¢ decreases with crack velocity, A, sufficiently so that eqn (3) may not be
satisfied, then unstable cracking will follow. Indeed, Mai and Atkins 19 and
Andrews and Stevenson 113 have shown that for certain epoxies dG¢/dA is
negative. Kinloch and Young ~4 are uncertain whether this is the cause or
the consequence of 'stick-slip' crack propagation. Partly this is because
Gledhill and Kinloch ~ 5 could not p r o m o t e stable cracking by improving
the gsf and partly because they only obtained positive dGc/dA in the epoxies
they studied. It must be stressed, however, that cracking can only be made
stable in otherwise unstable crack systems if eqn (1) is satisfied. If dGJdA is
very negative the gsf may have to be very m u c h improved with external
stabilizers before any stable cracking can be obtained. 19'96 During 'stick-
slip' the crack velocity is not uniform and only a mean crack velocity
between upper and lower values can be estimated) 1° The corresponding
mean fracture toughness is given by G c = Q Gx/~G~, where G~ and GA are the
initiation and arrest toughnesses, and Q is a function of the specimen
geometry. 96 The positive dGJdA results obtained by Gledhill and Kinloch
only refer to the initiation,which must be higher for higher crosshead speeds
of the testing machine. There are no measurements for K~¢ and A during the
jump. Based u p o n these arguments it is felt that the negative dGJdA
explanation for unstable 'stick-slip' cracking in epoxies cannot be
completely ruled out. In fact this proposal has been recently supported by
Leevers 1lO and Maugis, ~~2 who both assume that if G~ decreasing with A is a
material property, 'stick-slip' occurs when a mean crack velocity is imposed
within this region. Both authors also discuss the effects of specimen
geometry (i.e. gsf) in controlling 'stick-slips'. Why G¢ should decrease with A
is not clear. Because of the low crack velocity at which stick-slip can be
observed in epoxies Maugis 112 suggested that the mechanism may be one of
internal friction ~16 and not one of isothermal-adiabatic transition*) t7
IZ
Eq (~)
%6 o
v.
(a) (b)
l q(~)
% =t+25 MPo
c :2l/* ~m
I
5 110
(c)
Fig. 5. Variation of Kic/Kic~ ratio with x/P for (a) unmodified epoxies, (b) rubber-modified
epoxies, and (c) hybrid glass particle-toughened epoxies. (After Kinloch and co-
workers. 84.~o)
particle-toughened epoxies. Any crack tip radius bigger than Pc gives types
A and B crack growth characteristics.
While the crack tip blunting model offers a simple mechanism for, and an
elegant analysis of, crack instability in unmodified and toughened epoxies it
must be mentioned that Cherry and Thomson 12 ~ have observed slow stable
crack growth prior to unstable fracture in a certain epoxy system and this
Failure mechanisms in toughened epoxy resins--a review 197
&
0%
0o
-%
y.
1 1
60 120 180
TRUE TENSILE YIELD STRESS, aty(MPa)
(a)
3 I
2 x
@v
A
@a
D
o
o
0 7'5 150
O~y (MPo)
(b)
Fig. 6. Relationship between KI~/KI~ ratio and true tensile yield stress, a,y, showing the
various types of crack growth. (a) Unmodified and rubber-modified epoxies and (b) hybrid
particle-toughened epoxies. (After Kinloch and co-workers. 8'~'119)
198 Amar C. Garg, Yiu-Wing Mai
result is in direct conflict with the crack tip blunting theory. It is also
pertinent to ask whether the blunting model in which dG~c/dA or dK~¢/dA is
negative is any different to the negative dG~¢/dA proposal. Figure 6 shows
that in the instability region of type B crack growth dK~c/da,y is negative and
in the stable type C crack growth region dK~¢/da,y ~0. It is c o m m o n
knowledge that in polymeric materials a,y increases with strain rate (g). When
crack tip blunting occurs the crack tip is stationary and the strain rate in the
crack tip zone (gs) is of the order e.rk/K,96 where e.y is the yield strain. When
crack propagation eventually occurs the strain rate for the moving crack is
~ m ( = anE
" 2 ~y/K~c
3 2 ) and is proportional to the crack tip velocity 6. Thus the aty
axis in Fig. 6 may be thought of as the 6 axis. Since e;m > ~ there is an effective
increase in aty due to d when the blunt crack extends. At the same time, K~
decreases from its blunt tip value to Kt~ for a sharp crack so that dKtJdA (or
dK~Jdaty ) and dKtJId,4 are both negative and unstable cracking ensues by
violation of eqn (1). It appears, therefore, that the crack tip blunting theory
and the negative dGJd,4 model are qualitatively equivalent. Type C crack
growth is stable because dKtJdA m0; however, the stable type A crack
growth cannot be explained in terms of eqn (1) for it only applies to linear
elastic solids. Crack stability analysis in the presence of large plastic flow
remote from the crack tip has been given by Atkins and Mai, 96 Turner ~22
and Paris et alJ 23 The remote plastic flow work accompanying crack growth
acts as a sink for excess energy absorption and this has the effect of
stabilizing otherwise unstable fractures. As mentioned before, it must be
questioned whether valid Ktc measurements can be made in this region.
It is shown in the previous section that the transitions between the three
different types of crack growth can be controlled by varying either the test
temperature or the applied crosshead/strain rate. This result is just a
reflection of the viscoelastic nature of epoxies. As such the time-temperature
superposition principle in viscoelasticity is expected to apply to the specific
work of fracture measurements, G~c. Hunston and co-workers have
vigorously pursued this line of approach. 9°'1~9'~ 24. By plotting the fracture
energy, G~¢, against time-to-failure, tf (taken as the time from initial load
application to onset of crack propagation), at various temperatures the
curves can be shifted along the tf axis to obtain the best overlap in order to
give a master curve for G~c. It is shown that this shift factor, a v, can be
described by the Arrhenius equation
Failure mechanisms in toughened epoxy resins--a review 199
Zt~
Stress retoxohon Dynamic mechanica[
~ Ophr 0 Ophr
20 " ~ 41, 5phr (bimodal • 5 phr
distribution) ~1 5 phr (bimodo[
• 17 5phr distribution)
16 ~ + 175phr
Yietd
12 15phr
o 13
2 t+
0 ÷
(T =Tg(lOO*g)}
-L,
-1/+0 -100 -60 -20 20
-(Tg-T)
Fig. 7. Shift factor, a x, as a function of temperature needed for time-temperature
superposition of G~c for rubber-modified epoxies. Rubber contents are indicated after each
test method. (After Hunston et a/ri °)
Hunston and co-workers 119'12+ have suggested that G~c+ measures the
limiting toughness at low temperatures and high stress rates, fl gives the
magnitude of the toughening and depends on the rubber content, 124 and m
assesses the time-dependence of the fracture energy and appears to be a
function o f the rubber particle size distribution. 124 Equation (7), therefore, is
not only useful for the prediction of Gnc given any tf and T, but it can also be
used to compare the fracture properties of epoxies of different composi-
tions.t19 However, much further work is required to establish the usefulness
of the parameters G,c~, fl and m for the characterization of the structure-
property relationship.
200 Amar C, Garg, Yiu-Wing Mai
,&
SLow rote tests: T
• -600C "
x -50 Rubber modified epoxy J
A -~.0 (15 phr odded rubber Vf=018)
o -JO
v -20
. 0
6 • 20
v 30
• 50 . /
i:pQct tests: /
3 To:Tg(I00°C) /
L-~
_ . x ~ O Unmod,fledepoxy /
, . :,~-#~ . .* 4. v. ? . r ~ ~ ,
-20 -15 -I0 -5
tog (tf/nT)(S)
Fig. 8. Fracture energy, G]c,versus reduced time-to-failure,tt/ax. (After Hunston et al.9°)
0 0 o
0 0
0
0
~ m m
Fig. 9. Crack toughening mechanisms in rubber-filled modified polymers: (1) shear band
formation near rubber particles; (2) fracture of rubber particles after cavitation; (3) stretching,
(4) debonding and (5) tearing of rubber particles; (6) transparticle fracture; (7) debonding of
hard particles; (8) crack deflection by hard particles; (9) voided/cavi tated rubber particles; (10)
crazing; (11) plastic zone at craze tip; (12) diffuse shear yielding; (13) shear band/craze
interaction.
Fig. 10. Crack pinning mechanism. The bowed crack front is at the verge of breaking away
from pinning. (After Lange. 7°)
202 Amar C. Garg, Yiu-Wing Mai
which is dissipated irreversibly (e.g. as heat) when the particle fails either by
debonding from the matrix or by tearing is given by 57
AG,¢ = 47 Vp[1 - 6/(22 + 2 + 4)3 (8)
where 2 is either the extension ratio at the time of debonding or rubber
tearing, Y is either the energy per unit area of interface required to d e b o n d
rubber from the matrix or the rubber tearing energy, and Vp is the volume
fraction of rubber particles.
2[22
7p = X ~/ ar (9)
where t/is the stress concentration factor at the particle equator and ar is the
fracture stress. The n u m b e r of particles, N, of mean radius, [, per unit area of
surface is 126
N = 3(Vp/n[2) (10)
Thus the increase in toughness due to the fracture of particles is
AG,~ = 3t/z[EVpe, 2 (11)
where q (= af/E) is the fracture strain of the particle. For a uniaxial stress
field t/= 2; hence
AGI~ = 12lEe 2 Vp (12)
(4) Crack-pinning
The obstacles (hard filler particles) create the obstruction to the propagation
of the crack front and cause an increase in toughness by bowing out the
crack front between the particles (Fig. 10). Lange 7°'v4 has given a relation for
the increase in fracture energy due to pinning as
assuming a Dugdale plastic line zone model it can be shown that the two
parameters are related by
~¢ = 8~yrye/~E (18)
if Poisson constraint effects are neglected. Since the fracture energy, G~c, is
given by the p r o d u c t of ay anf 6 c, and blunting causes 6¢ to increase, the
fracture resistance is hence enhanced. In the crack tip blunting theory l~s
referred to in the previous section 3c is taken as equal to the crack tip radius,
p, at onset of crack growth. Equation (4) gives the toughness enhancement,
K~¢, as a function of p with two adjustable parameters, o't¢ and c. Figure 5
shows the success of this model in predicting K~¢ values for several
unmodified and toughened epoxies.
A similar fracture criterion for blunt notches in epoxies has been provided
by Narisawa et al., ~33 who state that the craze and fracture are initiated at
the elastic-plastic boundary, r = dy (ahead of crack tip), where high stresses
exist owing to stress triaxiality in plane strain conditions. The fracture
initiates when the dilatational stress at this point, dy, becomes critical.
The hydrostatic stress, S, at a point, r, ahead of a crack tip of radius p can
be obtained from slip-line theory and Tresca's yield criterion. Thus it can be
shown for a rigid, perfectly-plastic material with a Poisson ratio, v, of 0.5
that 133
S = ~y{1 + 2In(1 + r/p)} (19)
where Zy is the shear yield stress.
Fracture occurs when at r -- dy, S attains a critical value, S¢. Narisawa et
al. have determined S~ and dy for several epoxies at various loading rates.
This criterion can be used to predict the crack growth behavior in a m a n n e r
similar to the Kinloch and Williams l~s analysis. In terms of the principal
stresses, a~, o 2 and ~3, ahead of the crack tip, it can be shown that for plane
strain
S = ~((~l "{- ~2 -~- ~3) = 2(1 + v)aax~/3[2zcdy(1 + p/2dy)] '/2 (20)
At the onset of fracture S = S¢, so that eqn (20) becomes
Equation (21) is similar to eqn (4) in that S~ and dy now become the two
adjustable fracture parameters. But, unlike ate and c, both S~ and dy are
strong functions of loading rate)33 In this respect eqn (21) is not as useful a
fracture criterion as eqn (4), which applies for a wide range of loading rates
and test temperatures.
Garg and Mai ~34 have also derived the following failure equation for
Failure mechanisms in toughened epoxy resins--a review 205
epoxies based on the blunt crack tip model by considering both the complete
stress state at the point of fracture initiation and the pressure dependence on
fracture, i.e.
K,¢/K,~ = [x//-~-+ B - A ] (22)
where
A= (k - 1)(1 + v)(1 + q)5/2 (23a)
(1 + 2q) 2 - 2q - 4v(1 - v)(1 + q)2
and
k(1 + q)3
B = (1 + 2q) 2 - 2q - 4v(1 - v)(1 + q)Z (23b)
Here k = tree/tyro (ratio of compressive to tensile strength) and q = p/2c.
Equation (22) is given for plane strain, but for plane stress v = 0 in eqn (23).
Comparisons of the predicted K~c results for two epoxy resins using eqn (22)
and eqn (4) obtained by Kinloch and Williams 118 are given in Ref. 134.
5 TOUGHENING MECHANISMS
i.e. rubber stretching and tearing. In their study of toughened epoxies, crack
propagation was followed with an optical microscope. As the crack opened
up, stretching and failure of rubber particles was seen to be the major source
of energy absorption in toughened epoxies. Also they did not notice any
significant shear yielding either in unmodified or modified epoxies and they
therefore ignored the contribution of shear yielding to toughness. Neither
did they obtain any evidence of crazing. Based upon such observations,
Kunz-Douglass et al. have proposed a simple model (eqn (8)) to compute the
fracture energy of rubber-toughened epoxies. However, as commented by
Yee and Pearson, 88'89 this quantitative theory can only account for an
increase in toughness by a factor of one or two, whereas the actual increase
in toughness due to rubber particles is seen to be an order of magnitude
greater (Table 1). This has been acknowledged by Sayre e t a / . 1 3 8 in a later
paper. However, they explained that the discrepancy was due to the rubber
tear energy, 7, used in eqn (8) being considerably underestimated in quasi-
static tests. In the fracture experiments the tearing rate for the rubber
particles would be three to four orders of magnitude larger so that 7 is at
least 10 times bigger. If so, better agreement would be obtained between
theory and experiment. In addition, they pointed out the importance of the
state of cure and mechanical properties of the rubber in toughening.
Kinloch et al. 62 have also observed a few discrepancies between the model
of Kunz-Douglass et al. and their own experimental observations. For
example, the model fails to explain the stress whitening phenomenon, the
high toughness at high temperatures and the transitions between various
types of crack growth. It also predicts the wrong trend for the time-
temperature dependence of G~c and it ignores the very important
contributions to toughness by rubber cavitation and shear yielding in the
matrix. It is difficult to resolve these differences because of the sensitivity of
the mechanical properties to curing hardeners and curing conditions.
However, it must be pointed out that the rubber stretching and tearing
model should not be dismissed completely. Bandyopadhyay and co-
workers 58.59 have clearly identified particles stretching and bridging behind
the advancing crack tip in rubber-toughened epoxies from direct scanning
electron microscopy. Toughening by crack interface bridging behind the
propagating tip has been observed in cementitious materials 139-14~ and in
ceramics. 142-144 This gives rise to the so-called R-curve behavior 11~ in
which the crack growth resistance as measured by either Gc or K c increases
with crack growth, Aa. It is to be expected that an R-curve would be obtained
if rubber particle bridging is present behind the tip of the crack. The
condition for a propagating crack is that at the crack tip the effective stress
intensity factor, Ke, is equal to that of the unmodified matrix, Km. Thus
K s = K~ + K , ( A a ) = K m (24)
208 Amar C. Garg, Yiu-Wing Mai
where K r is the stress intensity factor due to the rubber particles bridging
over a distance, Aa, and its magnitude is negative, K a is the applied stress
intensity factor and at fracture it measured the critical value of the fracture
resistance, K c. Therefore the crack growth resistance of the modified epoxy is
Kc = K m - - Kr(Aa ) (25)
K, depends on the properties of the rubber particles and K c reaches a
maximum value, Ko~, when full bridging is established over a saturated
distance, Aa s, and at this point the rubber particles at the original crack tip
position just begin to tear. From experimental evidence given in Ref. 138 it
seems that Aa~ is of the order of several millimeters and an R-curve should
have been obtained. Unfortunately, no such curves have been published in
the literature for ~rubber-modified epoxies. Kunz-Douglass et al. 5v have
shown that the increased fracture toughness in terms of G is given by
Glc = Gm(l - Vp) + AGIc (26)
Since AGj¢ is given by eqn (8) and corresponds to rubber tear, the G~¢ value
calculated is therefore related to the maximum value, K~, of the R-curve, i.e.
K 2 = EG~ = E[Gm(1 - Vo) + AGI¢] (27)
Scott and Philtips 31 have also studied rubber-toughened epoxies and have
discussed the possibilities of mechanisms such as tearing of rubber particles,
localized crazing and plastic zone effect by plasticization of the resin.
However, no conclusive toughening mechanisms have been identified. It is
only in more extensive recent studies by Bascom and co-workers, 15.18 Yee
and Pearson 88'89 and Kinloch et al. 62"63 have more definite toughening
mechanisms been verified. Bascom et al. have noticed in the fracture surface
of specimens of toughened epoxies a stress whitening zone in the slow crack
propagation region. Such stress whitening was observed neither for
unmodified epoxy nor for the modified epoxy in the region of fast crack
growth or during an impact test. Detailed microscopic examination of the
stress whitening region showed the presence of small closely spaced holes
{larger than the original particle size) in the matrix and yielding (seen as tear
markings in the epoxy) of the matrix resin around the particles. Such large
holes are interpreted as being caused by the dilatational deformation of the
particles and the matrix. Also, the dilatation of rubber particles nucleates
local shear yielding of the epoxy matrix, causing a significant crack tip
deformation. Thus volume dilatation of rubber particles and the
surrounding matrix as well as shear yielding of the epoxy are proposed as
two mechanisms responsible for the increase in fracture toughness of
elastomer-modified epoxies.
Yee and Pearson 88"89 and Kinloch e t a / . 6 2 A 1 9 have also proposed similar
Failure mechanisms in toughened epoxy resins--a review 209
5.0
~,.0
÷
K~c o o
Km
3.0
2.0
1.0 i
1.0 20 3.0
%
Fig. 1 I. Increase in K~c relative to unmodified epoxy, Km, as a function of particle size to
interparticle distance, d/2b. - - . , Theoretical curve; 0 , resin A + alumina; O, resin A + silica;
[~, resin B + alumina; +, resin B + silica. (After Moloney et al. 77'79)
(b)
Fig. 12. Fracture surfaces for (a) zirconia--epoxy showing crack pinning 'tails' and (b)
zirconia-rubber-epoxy showing crack initiations and arrests in stress whitened region. (After
Lowfl 5)
Failure mechanisms in toughened epoxy resins--a review 213
tip blunting model, and a two-parameter fracture criterion (i.e. a~c and c) is
required (see Figs 5(c) and 6(b)).
Recent work by Low et al. 95 on zirconia-epoxy and zirconia-rubber-
epoxy systems (see Table 4) has also confirmed that crack pinning is the
major toughening mechanism in zirconia-epoxies and that localized shear
flow in the matrix material at the crack tip and rubber cavitation becomes
the dominant mechanism of toughening in the hybrid particulate epoxies.
Experimental evidence to support these conclusions are given in Fig. 12(a)
and (b). In Fig. 12(a) the tails around the zirconia particles are clear
indications of crack pinning operating in the zirconia-epoxy system. In Fig.
12(b), taken in the stress whitening region for the hybrid particulate epoxy, it
appears that there are many sites of crack arrest and crack initiation. Since
initiation involves the debonding/tearing of rubber particles the toughness is
very much enhanced.
In conclusion it may be said that both crack pinning and crack tip
blunting are two mechanisms responsible for the enhancement of toughness
in particulate composites. It is probably difficult to separate the two effects,
although some approximate rules can be applied. 75'84
Although crack deflection processes by tilting and twisting of the crack
front are known to produce toughening in some ceramic materials 127'128
they have not been observed in particle-filled epoxies except when the filler is
a short fiber. Low et al. 95 have observed very large stress whitening zones in
an epoxy/rubber/short-alumina fiber hybrid composite resulting from a
range of failure mechanisms in addition to crack front tilting and twisting.
These are fiber bridging behind the advancing crack tip, rubber cavitation
and plastic shear deformation of the matrix. The development of this stress
K~
g_
tic
t..
._°2
I 2 3 4 5 6 ? 8 9 10
0omoge Zone Exfension Ao (ram)
Fig. 13. Crack growth resistance curve for an epoxy-rubber-alumina fiber hybrid
composite as given in Table 4.
214 Arnar C. Garg, Yiu-Wing Mai
whitening zone gives rise to stable crack growth and hence a rising crack
resistance curve as shown in Fig. 13.
6 CONCLUSIONS
The effect of toughening agents such as CTBN rubbers and particulate fillers
on epoxide resins has been reviewed. It is noticed that the effect of elastomer
addition may increase the toughness up to about 30 times the fracture energy
of unmodified epoxies. However, the gain in toughness for the toughened
epoxy as matrix in a fiber-reinforced composite or as adhesive is much
smaller than for the bulk epoxy. This reduced toughness is explained by the
dependence of toughness on epoxy film thickness. 31'9°
Several toughening mechanisms are reviewed. Although there exists some
controversy about the origin of the high toughness of elastomer-modified
epoxies, the evidence collected so f a r 62'63'88'89 suggests that the deform-
ation processes near the crack tip caused by (i) localized cavitation in the
rubber, or at the particle/matrix interface, and (ii) plastic shear yielding in the
epoxy matrix account for the high toughness in such materials.
As regards the toughening mechanisms in particle-filled epoxies, the
possibility of various mechanisms is discussed and it is suggested that both
plastic yielding of the matrix and pinning of the crack front are responsible
for increasing the toughness of such materials. The relative contribution of
the mechanisms depends on the test temperature. Various factors such as
size and type of particle, volume fraction, and bonding between particle and
matrix play a role in the operative toughening mechanisms. 75- 8o Effective
toughening also depends on the geometric arrangements of the inclusions,
whether a single inclusion or a cluster of inclusions, and whether they are
uniformly distributed or not. 149
ACKNOWLEDGEMENTS
REFERENCES
117. Y.-W. Mai and N. B. Leete, J. Mater. Sci., 14 (1979), pp. 2264-7,
118. A. J. Kinloch and J. G. Williams, J. Mater. Sci., 15 (1980), pp. 987-96.
119. A.J. Kinloch, Mechanics and mechanisms of fracture of thermosetting epoxy
polymers, in Advances in Polymer Science--Vol. 72 (ed. K. Dfisek), Springer-
Verlag, Berlin, 1986, pp. 45-67.
120. J. G. Williams, Fracture Mechanics of Polymers, Ellis Horwood/John Wiley,
Chichester, 1984.
121. B. W. Cherry and K. W. Thomson, J. Mater. Sci., 16 (1981), pp. 1913-24.
122. C. E. Turner, in ASTM STP 677 (1979), pp. 614-28.
123. P. C. Paris, H. Tada, A. Zahoor and H. Ernst, in ASTM STP 668 (1979),
pp. 5-36.
124. D. L. Hunston and G. W. Buliman, 'Viscoelastic fracture behaviour for
different rubber-modified epoxy adhesive formulations', to be published.
125. J. Gurland, Fracture of metal matrix particulate composites, in Composite
Materials, Vol. 5: Fracture and Fatigue (ed. L. J. Broutman), Academic Press,
New York, 1975, pp. 45-91.
126. E. E. Underwood, Particle size distribution in Quantitative Microscopy (eds
R. T. DeHoff and F. N. Rhines), McGraw-Hill, New York, 1968, pp. 149-200.
127. K. T. Faber and A. G. Evans, Acta Metall., 31 (1983), pp. 565-76.
128. Idem. ibid., 31 (1983), pp. 577-84.
129. F. F. Lange, Phil. Mag., 22 (1970), pp.983-92.
130. F. F. Lange, J. Amer. Ceram. Soc., 54 (1971), pp. 614-20.
131. D. J. Green, P. S. Nicholson and J. D. Embery, J. Mater. Sci., 14 (1979),
pp. 1657-61.
132. S. Hashemi and J. G. Williams, J. Mater. Sci., 20 (1985), pp. 922-8.
133. I. Narisawa, T. Muralyama and H. Ogawa, Polymer, 23 (1982), pp. 291-4.
134. A. C. Garg and Y.-W. Mai, Failure prediction in toughened epoxy resins,
Composites Sci. & Techn., 31 (1988), pp. 225-42.
135. A. M. Donald and E. J. Kramer, J. Mater. Sci., 17 (1982), pp. 1871-9.
136. J. Lilley and D. G. Holloway, Phil. Mag., 28 (1973), pp. 215-20.
137. S. H. Carr, What we should be doing, in Proc. of the CriticalReview, Techniques
for the Characterization of Composite Materials, Army Materials and
Mechanics Research Center, Mass., AMMRC, MS 82-3 (1983), pp. 493-9.
138. J.A. Sayre, S. C. Kunz and R. A. Assink, Effect of rubber cross-link density and
tear energy on the toughness of rubber-modified epoxies, in ACS Advances in
Chemisto, Series No. 208: Rubber Modified Thermoset Resins (eds C. K. Riew
and J. K. Gillham), Amer. Chem. Soc., 1984, pp. 215-34.
139. A. Hillerborg, M. Modeer and P. E. Petersson, Cement and Concrete Research,
6 (1976), pp. 773-82.
140. M. Wecharatana and S. P. Shah, Cement and Concrete Research, |0 (1980),
pp. 833-44.
141. B. Cotterell and Y.-W. Mai, J. Mater. Sci., 22 (1987), pp. 2734-8.
142. Y.-W. Mai and B. R. Lawn, J. Amer. Ceram. Soc., 70 (1987), pp. 289-94.
143. P. L. Swanson, C. J. Fairbanks, B. R. Lawn, Y.-W. Mai and B. J. Hockey, J.
Amer. Ceram. Soc., 70 (1987), pp. 279-89.
144. R. F. Cook, C. J. Fairbanks, B. R. Lawn and Y.-W. M ai, J. Mater. Res., 2 (1987),
pp. 345-56.
145. A. L. Gurson, J. Eng. Mater. Techn., Trans. ASME, 99 (1977), pp. 2 15.
146. A.G. Evans, Z. B. Ahmad, D. G. Gilbert and P. W. R. Beaumont, Acta Metall.,
34 (1986), pp. 79 87.
Failure mechanisms in toughened epoxy resins--a review 221
147. R.M. McMeeking and A. G. Evans, J. Amer. Ceram. Soc., 65 (1982), pp. 242-6.
148. A. G. Evans and K. T. Faber, J. Amer. Ceram. Soc., 67 (1984), pp. 255-60.
149. J. C. M. Li and S. C. Sanday, Acta Metall., 34 (1986), pp. 537-43.
150. A. M. James and W. E. Harvill Jr, Advanced Composites, Lockhead Horizons,
1986, pp. 31-43.
151. A.C. Moloney, H. H. Kausch, T. Kaiser and H. R. Beer, J. Mater. Sci., 22 (1987),
pp. 381-93.
(a) Epoxies
1. Polyglycidyl ethers of the Novolacs (Union Carbide ERR 0100)
/o /o\
O--CHz--CH--CH 2 O--CH2--CH--CH 2 O--CHz--CH--CH 2
/ \0
O--CH2--CH--CH 2
/% y\ /o
CH2--CH--CH 2 CH2--CH--CH 2
CH2--CH--CH 2
\ N ~ C H 2 ~ N /
/ L_Z_/ L~_/ \
C~HO2 -/- C H - - C H 2 CH2--CH - -~C H 2
,NO
222 Amar C. Garg, Yiu-Wing Mai
~-NH 2
6. Diethylene triamine (DETA)
H 2N--CH2--CH2--NH--CHi--CH2--NH 2
7. Piperidine
(CH2)sNH
Failure mechanisms in toughened epoxy resins--a review 223
A P P E N D I X B: C H E M I C A L F O R M U L A FOR CARBOXYL-
T E R M I N A T E D B U T A D I E N E - A C R Y L O N I T R I L E (CTBN) RUBBER
HOOC--~(CH 2 - - C H = C H - - C H 2)~--(CH 2 - - C H ) y ] - C O O H
[_ CN J~
where on average x = 5, y = 1 and m = 10, giving a molecular weight of
3320 g m o l - 1.