Environmental Effects on Crack Growth in Polymers

JC Arnold, University of Wales, Swansea, UK

r 2003 Elsevier Ltd All rights reserved.

This is a reproduction of J.C. Arnold, 6.06 - Environmental Effects on Crack Growth in Polymers, In Comprehensive Structural Integrity, edited by I.
Milne, R.O. Ritchie, B. Karihaloo, Pergamon, 2003, https://doi.org/10.1016/B0-08-043749-4/06123-1

Introduction 242
Chemical Degradation 243
Environmental Stress Cracking 244
Failure in Polymers 244
Overview 244
Craze Formation 246
Craze Propagation 248
Fluid/Polymer Interactions 249
Solubility 249
Diffusion in Polymers 252
Surface Energy 254
Accelerated Crack Growth Due to Chemical Degradation 256
Overview 256
The Effects of Material Structure 256
The Effects of Temperature and Stress 257
Degradation in Specific Environments 257
Hydrolysis, acidolysis, and alkolysis 258
Degradation in oxidizing liquids 259
Chlorine attack 259
Nitrogen dioxide and sulfur dioxide 260
Environmental Stress Cracking 260
The Mechanisms of ESC 260
Plasticization 260
Surface energy effects 260
ESC Test Methods 263
Bend methods 263
Creep tests 265
Constant-strain-rate and constant-stress-rate tests 265
Fracture toughness tests 266
Biaxial tests 267
ESC Craze Initiation 1: Stress/Strain/Time Criteria 267
ESC Craze Initiation 2: Polymer/Fluid Interaction Criteria 270
ESC Craze Growth 271
The Effects of Liquid Absorption 273
Environmental Effects of Crack Growth In Rubbers 274
References 274

Introduction

One of the major roles of designers and users of materials is to attempt to prevent unexpected failure in service. Unfortunately,
during their relatively short history of engineering use, this is a problem that has plagued polymers and has hindered their
expansion into structurally critical areas. One of the main reasons for this is the potential complexity of interacting failure
mechanisms. As the understanding of these failure mechanisms has improved, the occurrence of these failures has decreased, but it
remains a continuing concern.
Failure in service is most commonly characterized by a relatively sudden brittle failure, which is unexpected and can be
catastrophic in nature. Once the causes have been identified, they are often attributed to combinations of minor effects, such as

242 Comprehensive Structural Integrity, 2nd Edition, Volume 6 doi:10.1016/B978-0-12-822944-6.00154-7

 Environmental Effects on Crack Growth in Polymers 243

residual stresses, brief contact with minor components of mixed environments or temperature fluctuations. These combinational
causes of failure are the ones that are hardest to predict and account for many of the failures seen in service.
A recently published survey of in-service failures with industrial and commercial plastic and rubber products in the UK has
attempted to categorize the cause of failure in each case (Wright, 2001). This is not always an easy task as in many cases, the exact
underlying cause of failure is debatable, with several interacting mechanisms. For example, if a polycarbonate component were to
fail by brittle crack propagation due to fluctuating loading conditions, the cause would be attributed to fatigue. However, if there
were also occasional contact with an organic cleaning fluid, then the possibility of environmental stress cracking (ESC) would also
appear. If in addition, the material had been incorrectly dried during processing, then hydrolysis during processing would have
significantly reduced the polymer molecular weight, promoting the other failure mechanisms.
The proportions of in-service failure attributable to specific failure mechanisms are given in Table 1 below.
From this study, it is clear that environmental effects play a significant role in failure processes. ESC, the most common type of
failure, is caused by combinations of stress and an active environment. Of the other failure types, chemical attack most commonly
involves environmental effects combined with eventual crack growth. Many of the failures in the “other” category will also have
some environmental influence. It would be safe to assume that environmental contact plays a role in at least a third of all in-service
failures.
Specific examples of environmentally assisted failure are given below, some of which are highlighted in detail in the study
mentioned above (Wright, 2001).

(i) An acetal piping and valve system in a brewery failed due to contact with low concentrations of acetic acid.
(ii) Solvent-welded polystyrene piping systems have periodically failed due to the combination of residual stress at the solvent
weld and small amounts of solvent remaining in the polymer.
(iii) A fire-extinguisher head made from ABS was found to have cracked after two years in service. This was eventually tracked
down to a combination of the stresses present due to the interference fit and brief contact with a cleaning fluid. The
important aspect to this was that in order to reproduce failure, it was necessary to apply the stress for a considerable time
before the brief contact of the environment.
(iv) Polycarbonate food mixer bowls were found to have developed fine crazes on the inner surface. Although these did not
cause catastrophic failure, they were deemed to be a health hazard. They were traced to occasional contact with essential
oils present in certain foodstuffs.
(v) A nylon hose valve used at a swimming pool failed catastrophically. This was found to be due to attack by zinc chloride,
formed between chlorine in the atmosphere and contact with a brass rod.
(vi) Polyethylene bottles containing organic liquid fertilizer failed when stacked in a warehouse. Although the failure appeared
to be macroscopically ductile, it was actually due to microcracks formed by a combination of the stresses from the applied
top load, environmental contact with the agrochemicals and a reduced-molecular-weight grade of polyethylene.
(vii) The rapid crack growth seen in rubber seals during pressure transients in hydrocarbon gas environments in the oil industry.
Although significant stresses are developed by the change in pressure and consequent evaporation of gas dissolved in the
rubber, the dissolved gas also significantly reduces crack-growth resistance. Similar effects have been seen with high-
pressure carbon dioxide in the nuclear-power-generating industry.
(viii) The failure of many plastics in low-temperature gaseous environments, such as nitrogen, which can be a significant design
concern for cryogenic systems.
(ix) Low-stress fatigue crack growth rate in many rubbers is found to be independent of the stress and is actually caused by
ozone attack at the crack front. This leads to a nonzero crack growth rate, even under negligible stress.
It is clear from these examples and many experimental studies that environmentally assisted failure can be especially difficult to
predict due to the following features:
(i) The active environment is often a very minor component in a mixture. In some cases, it occurs as a reaction product
between other environments.
(ii) The stresses causing failure can be very low, often due to residual stresses from moulding or interference stresses on
assembly.
(iii) The contact period can be very short and yet still cause serious problems. There is one particularly dangerous type of failure
that involves low stress application for a long time period followed by brief environmental contact.

In order to study environmentally assisted failures, it is essential to distinguish between two quite distinct failure processes.
These can overlap in some cases and produce many similarities in the appearance of failure, but involve quite different
mechanisms.

Chemical Degradation
The first of these is a true chemical process, often accelerated by stress. Chemical degradation occurs with many polymer/
environment combinations and involves covalent-bond breakage within the polymer chain. This normally leads to chain scission
(molecular chain fracture) and reduction in molecular weight. This will reduce the material’s crack-growth resistance, but in many
244 Environmental Effects on Crack Growth in Polymers

Table 1 Causes of in-service failure.

Type of failure Percentage

occurrence

Environmental stress 25
cracking
Dynamic fatigue 15
Notched static rupture 14
Creep/relaxation 8
Chemical attack 7
UV attack 6
Thermal degradation 4
Other 21

After Wright (2001).

cases may also be localized at crack tips due to fluid stagnation and increased stress. It is clear that applied stresses can dramatically
accelerate the chemical degradation, either due to easier bond breakage of strained chains or by increased solubility of the
environment. When this mechanism operates, the term “stress corrosion cracking” is sometimes used, although it is clearly
different from the mechanism operating in metals.

Environmental Stress Cracking

The second is an acceleration of normal failure processes by the presence of an active environment. In this case, diffusion of the
environment into a stressed polymer, especially at or near the surface allows easier craze formation and growth due to a
combination of local plasticization and surface energy reduction. As such, it is a purely physical process with no chemical
reactions. There will obviously be cases where an environment can act in either of these ways and so the two mechanisms can
overlap.
These two mechanisms form the main focus of this section, but in addition, the environmental effects on rubbery materials are
treated separately. It could be argued that crack growth due to thermo-oxidative degradation or to UV exposure should be included
within this study; however, as these subjects have been treated extensively elsewhere (e.g., White and Turnbull, 1994), we have
omitted them.
In order to understand the interacting mechanisms involved in environmentally assisted fracture, it will be necessary to
consider all the underlying processes, including:
(a) solubility of fluids in polymers and the effects of stress on solubility,
(b) diffusion and the effects of stress,
(c) chemical degradation and the effects of stress,
(d) failure mechanisms with polymers, and
(e) surface energy effects.

Failure in Polymers

Overview
Before describing the effects that external environments can have on crack growth in polymers, it is important to review the
underlying failure mechanisms that occur within polymers in inert environments. One of the most important features associated
with polymer failure processes is the combined effect of testing rate and temperature (Fig. 1). This shows that as the rate of loading
increases, the type of failure changes from an essentially brittle failure at fast rates, through ductile failure at intermediate rates and
then to another seemingly brittle region at very slow rates. The effects of temperature can be seen in a broadly similar light: at very
low temperatures, truly brittle failure occurs; at intermediate temperatures, failure is often ductile; and at higher temperatures,
failure often appears to be brittle in nature. This complete range of behaviors is not always observed, depending on the particular
polymer and loading condition, but virtually all polymers will show this double transition behavior under certain conditions.
In the rapid/low-temperature regime, failure is truly brittle, with crack formation and growth with very little ductility at the
crack tip. In this region, crazes are sometimes seen at the front of cracks, but these will be discussed more fully later. With this type
of failure mode, there is little time for chain motion and so the polymer chains ahead of a stress concentration will rapidly become
significantly strained. This will then lead to either chain scission of the most strained chain segments, or some disentanglement if
there is any scope for molecular motion.
Studies of this failure mode using electron spin resonance (ESR) or infrared spectroscopy (IR) have shown that appreciable
numbers of free radicals are formed due to chain scission (Kausch, 1987; Andrews and Reed, 1973; Sohma and Sakaguchi, 1976).
 Environmental Effects on Crack Growth in Polymers 245

Fig. 1 The effect of testing rate on the failure of plastics.

The number of these free radicals increases as the temperature decreases (Kausch, 1986) and also (to a lesser extent) as the
molecular weight increases. Some chain disentanglement will also occur; however, it is widely thought that chain scission is the
dominant rate-controlling mechanism at low temperatures and fast rates.
As the loading rate decreases, or the temperature increases, more main-chain motion becomes possible. This, therefore, leads to
a situation where ductile shear yielding becomes the dominant failure mode. Shear yielding is a molecular process whereby
polymer chains can move cooperatively. Shear yielded zones are formed, generally with little or no volume change, normally at
451 to the principal stress. They can form at the front of a growing crack, thereby blunting it and providing toughness, or through
the bulk of the material to give uniform yielding.
At the slowest loading rates, and higher temperatures (but below the glass transition temperature or melting point), a third
regime is seen where failure is characterized by a flat fracture surface with low values of strain-to-failure. In this case, failure is
primarily due to the formation, growth and breakdown of crazes. Crazes are planar, highly voided and fibrillated regions that form
in a direction perpendicular to the principal tensile stress. The unusual feature of craze failure is that, although the failure is
macroscopically brittle in nature, it proceeds by a yielding process, albeit a highly localized one. For crazes to form, chain motion
and disentanglement must occur at the tip of a craze, and therefore higher temperatures and longer timescales favor this type of
failure.
There is therefore a competition between the types of failure that is controlled by the degree of molecular mobility. This leads
to the well-known phenomenon of a ductile-to-brittle transition. In many polymer systems, two transitions can be seen; one at
low temperatures, especially with impact loading, and a second at higher temperatures and more especially with long-term
loading. The double transition can lead to some confusion when considering the effects of external variables. For instance, there is
considerable evidence that an increase in temperature produces shear yielding in preference to crazing; however, this is only true
for the short-time transition. With lower applied stresses, crazing is again favored, albeit at much longer times.
This is shown schematically in Fig. 2, where the creep lifetime is plotted against applied stress for different temperatures. As the
stress decreases, the lifetime increases, until brittle failure due to crazing is seen. Obviously, extrapolations of lifetime from the
ductile region can give significant overestimates if the transition is not taken into account. By conducting tests at elevated
temperatures, it is possible to predict this transition with shorter tests.
Different types of polymers will display these regions to greater or lesser extents, e.g.:

(i) Elastomers above their glass transition temperature have very mobile chains which are held together with cross-links.
Therefore, appreciable chain disentanglement is prevented and so the failure is controlled by chain scission, as seen in the
long-time region of Fig. 1 above.
(ii) Elastomers below their glass transition temperature are somewhat similar, except that there is much less molecular mobility
and so scission-controlled fracture occurs at lower strains. Thermosets generally behave in a similar manner, except that
small amounts of yielding and crazing can be seen (Henkee and Kramer, 1984).
(iii) Brittle thermoplastics such as polystyrene tend to show mainly scission-dominated fracture at low temperatures/fast rates
and crazing at higher temperatures/slower rates. Unless the applied stress has zero or negative hydrostatic component
(shear or compression), yielding is not seen very often.
(iv) Tougher thermoplastics such as polycarbonate, polyethylene and nylons show the full range of behavior, although very low
temperatures are sometimes needed to observe scission-dominated brittle fracture.

The main feature of a polymer that controls the type of failure is therefore the degree of chain mobility. This is characterized
most commonly by the glass transition temperature (Tg), at which the material changes from a stiff and glassy material to a more
rubbery one. Generally, the further below the glass transition temperature a polymer is, the lower the amount of main-chain
mobility is possible. The time/temperature equivalence is also seen here; as the testing rate increases, there is less time for
cooperative chain motion and so the effective glass transition temperature increases somewhat. There is an additional
246 Environmental Effects on Crack Growth in Polymers

Fig. 2 The ductile-to-brittle transition seen with long-term testing of polymers.

complication when considering semi-crystalline polymers, some of which have values of Tg below room temperature (e.g., PE),
while others have much higher values (e.g., PEEK). For the polymers with low values of Tg, the behavior is more similar to that of
rubbery materials, while those with higher values of Tg have more similarity with glassy thermoplastics. In either case, the mode of
failure is also dependent on the degree of pullout of chains from crystalline regions and on the overall levels of crystallinity (Fisher
and Brown, 1973).
One other significant influence on the mode of failure is the degree of physical aging. For glassy polymers at temperatures
below Tg, gradual densification and loss of chain mobility occurs with time due to slow compaction of the glass towards its
equilibrium density (Struik, 1978). As this happens, it would be expected that the reduced molecular mobility would favor/
scission-dominated brittle fracture and crazing. It has been shown that the yield stress does increase with physical aging (Hill et al.,
1990) and this tends to lead to a greater tendency for brittle failure (Lee and McGarry, 1991; Bauwens-Crowet and Bauwens, 1982;
Broutman and Krishnakumar, 1976; Arnold, 1993) and for the formation of simple crazes rather than shear yielded zones (Donald
and Kramer, 1982). The picture is unfortunately not quite as simple as this as localized yielding or softening at the front of a craze
can remove any effects of prior aging, which may explain why aging seems to have little effect on crazing during ESC (Arnold,
1995a).

Craze Formation
Crazes are planar crack-like features with the key difference that they contain considerable amounts of oriented fibrils bridging the
two surfaces of the craze. Therefore, they are load-bearing and it is possible to have a craze spanning the entire section of a sample
with no failure. A typical crazed sample is shown in Fig. 3. In most cases, crazes are seen to initiate from the surface, or from an
observable internal flaw. This obviously suggests the requirement for a significant stress-concentrating feature; however, when
near-perfect samples have been tested, crazes have still been observed, albeit at higher stresses (Argon and Bessonov, 1977).
The initial stages of craze formation have been a subject of considerable academic study, and one of the most significant papers
on this subject was produced by Wellinghof and Baer (1975). They studied the deformation processes in thin films of solution-cast
polystyrene using transmission electron microscopy (TEM) of gold-decorated samples. They found that, during deformation, very
small (B30 nm) voids were formed under all conditions studied. The subsequent behavior of these voids then depended on the
temperature and molecular weight. At temperatures close to Tg, these voids formed uniformly throughout the sample, leading to
shear yielding. At lower temperatures, these voids were seen to coalesce into planar zones perpendicular to the applied stress. These
zones then collapsed further and formed recognizable crazes. The crazes were seen to be more stable with higher molecular
weights; with the lowest-molecular-weight material studied, the planar zone rapidly formed a crack.
This study presents a picture of craze initiation as follows, shown schematically in Fig. 4:
(i) Under the action of a stress, very small voided regions form within the material due to main-polymer-chain motion. These
will form more readily at areas of stress concentration.
(ii) Under certain conditions, these will form more easily in a zone running perpendicular to the principal stress.
(iii) The voids then coalesce to form a fibrillated planar structure.
(iv) This fibrillated structure then develops into a recognizable craze with highly oriented polymer fibrils bridging the two
surfaces.

The reasons for the formation of these microscopic voids have been considered by Gent (1970), who proposed that they form
under the action of any negative hydrostatic stress. The important proposal in this work is the idea that the stress acts to lower the
glass transition temperature of the polymer at a stress concentration, due to an increase in free volume for molecular motion,
which then allows cavitation. This approach has been developed to include the effects of liquid environments, which will be
discussed later.
 Environmental Effects on Crack Growth in Polymers 247

Fig. 3 A typically crazed sample (of polystyrene after creep testing in ethanol).

Fig. 4 The stages of craze development.

The important question with this proposed mechanism is why the microscopic voids form a planar zone from which crazes
develop, rather than occurring in a uniform distribution leading to shear yielding. It has been found that higher molecular weights,
higher temperatures and some cross-linking all tend to favor shear yielding over crazing (Berger and Kramer, 1988; Henkee and
Kramer, 1984; Donald and Kramer, 1982). It is important to note that the applied stresses used in these studies were all reasonably
large, leading to craze formation in relatively short times (as is experimentally convenient). Therefore, it is almost certain that the
effects seen were at the high-stress/short-time transition from brittle crazing to ductile shear yielding. The cause of the low-stress/
long-time transition back to brittle craze formation, which is favored by higher temperatures is less clear. Although no theoretical
explanation of these two transitions has yet been given in terms of the formation and coalescence of microvoids, it is possible to
formulate one as follows:
248 Environmental Effects on Crack Growth in Polymers

(i) Microvoids can be formed by either a scission-dominated mechanism or by a disentanglement-dominated mechanism. The
scission-dominated mechanism will occur faster, but require higher stress levels. The disentanglement-dominated
mechanism will take place more slowly but will occur at lower stresses.
(ii) Adjacent to a microvoid (in a direction perpendicular to the maximum principal stress) will be a region of increased
dilatational stress. Above and below the microvoid will be a region of decreased dilatational stress.
(iii) Under conditions of fast loading (high stresses) and low temperatures, the first microvoid to form is likely to be scission
dominated. This will increase the likelihood of another scission-dominated microvoid adjacent to it. The disentanglement-
dominated microvoids, which will occur more uniformly, will not have sufficient time to form before a zone of scission-
dominated voids have formed, leading to a craze.
(iv) As the temperature increases, or stress decreases (leading to longer lifetimes), or as the entanglement density increases
(which increases resistance to chain scission), there will be more scope for disentanglement-dominated microvoids to
form. These will form more diffusely, thereby blunting any stress concentration and so lead to shear yielding.
(v) At even lower stresses, and high temperatures, although disentanglement-dominated microvoids are the only ones formed,
they form much more slowly, so that the stress concentration adjacent to one void leads to subsequent void formation in a
narrow zone, again leading to crazing.
The size and geometry of crazes have also been extensively studied (Brown and Kramer, 1981; Paredes and Fischer, 1979; Doll,
1983; Kramer, 1983), and modeled (Leonov and Brown, 1991; Kotoul, 1996). From these studies, the following features have
been found:

(i) The craze fibrils are typically between 10 nm and 50 nm in diameter, depending on the polymer.
(ii) The thickness of the craze ranges from zero at the tip to B 0.5 mm for most polymers.
(iii) The oriented fibrils exhibit an extension ratio close to 4 (higher near the craze tip).
(iv) The stress at the surface of a craze is fairly constant along its length, showing that the fibril drawing stress is also reasonably
constant.

The typical craze geometry is shown in Fig. 5.

Craze Propagation
Once a craze has established, a period of growth follows, where it propagates through the material, drawing more material into the
craze tip and breaking down to form a crack. It has generally been found that the overall geometry of a growing craze remains fairly
constant for much of its lifetime, and so there must be a steady-state process of craze growth and breakdown.
Early models of craze growth suggested that there was a continuous formation of microvoids ahead of the growing craze which
were then incorporated into the craze. One difficulty with this idea is that a fully developed craze structure has been seen at small
craze tips where the thickness is less than 10nm (Lauterwasser and Kramer, 1979). The typical void size is of the order of 30 nm,
and so it would not be possible for larger voids to become incorporated into a smaller crazed zone. Additionally, the growth rates
predicted by modeling such a mechanism did not fit experimental evidence (Argon, 1973).
The most widely accepted model is now the Taylor meniscus instability model (Argon and Salama, 1976). This model is shown
schematically in Fig. 6, and is based on the idea that a growing craze front develops a series of irregularities, much as a liquid
meniscus does between two solid plates. This instability develops void fingers, which grow and then merge sideways, leaving
behind the craze fibrils. In Fig. 6(a), the wedge of the craze tip is shown, with plastically deformed material ahead of the craze tip.
In Figs. 6(b) and (c), the stages of void finger progression are shown, giving an (idealized) fibril network in its wake. The meniscus
model has received good qualitative experimental support (Donald and Kramer, 1982) as well as quantitative agreement between
theoretically modeled and experimentally observed craze fibril spacings (Kotoul, 1996; Argon and Salama, 1976; Brown and
Kramer, 1981).
For the craze to grow, fibrils will obviously need to lengthen before they finally break. There are two possible mechanisms by
which this can happen. One is that the craze fibrils creep without any new material being drawn into them from the bulk. The
second is that the craze fibrils grow by drawing material from the surface of the craze. The important difference between these two
mechanisms is that the fibril creep mechanism would predict the craze extension ratio to increase as the fibrils become longer, and
also that the craze volume fraction should decrease. The surface drawing mechanism predicts that the fibrils have a constant
extension ratio and that the craze volume fraction remains constant (Lauterwasser and Kramer, 1979).
Experimental evidence almost exclusively points to the surface drawing mechanism as the dominant one. The measured fibril
extension ratios are found to be constant apart from at the craze tip, where they are higher (Leonov and Brown, 1991; Donald and
Kramer, 1982). The craze volume fraction has also been found to be virtually constant in most cases (Trent et al., 1981). The only
cases where fibril creep has been observed have been with crazes formed in the presence of an active environment (Williams and
Marshall, 1975). With environments that have a significant plasticization effect, but a low rate of diffusion, the environment
diffuses preferentially into the craze fibrils. This leads to the fibrils having a lower yield strength than the bulk of the material and
so fibril creep occurs rather than surface drawing. If the diffusion rate is slightly higher, plasticization of the craze surface can now
take place and the surface drawing mechanism again dominates.
 Environmental Effects on Crack Growth in Polymers 249

Fig. 5 The cross-sectional geometry of a typical craze.

Fig. 6 The meniscus instability model of craze growth.

In most cases, craze fibrils at the back of the craze will eventually break to form a crack. The mechanism by which this occurs
has been studied extensively. Obviously, if the surface drawing mechanism were to continue unhindered, the craze fibrils would
simply lengthen continuously and the craze would grow in a stable manner across the entire sample. Although this is seen in a few
cases, in the vast majority of situations, craze fibrils break. This implies that a certain degree of fibril creep must occur once the
craze fibrils grow beyond a certain size. The experimental evidence suggests that this tends to occur in a thinned section in the
middle of a craze fibril, often termed the “mid-rib.” An analysis of craze fibril breakdown by Kramer (1979) has shown that this
effect is due to unstable creep within each fibril, caused by a combination of disentanglement and scission and that it is strongly
dependent on molecular weight. This agrees with the experimental findings that low-molecular-weight polymers do not form
stable crazes.
The final area to consider with respect to crazes formed in inert environments is the stress distribution along and ahead of a
craze. Detailed stress analyses stem from work by Bilby and Eshelby (1972). It predicts a constant craze stress, slightly below the
remote stress over the majority of the craze, with an increase to a maximum at the craze tip.

Fluid/Polymer Interactions

Solubility
The solubility of polymers in solvents and of fluids into polymers is governed by internal energy interactions. The most widely
used theories for such interactions use cohesive energy terms and solubility parameters. A good review of these theories is given by
Van Krevelen (1997) and in more detail by Barton (1983). The solubility of polymers in solvents and vice versa is determined by
the chemical structures, with the simplification that similarity of chemical structure leads to solubility. The physical nature of the
polymer will also have an effect, especially for liquids that cause appreciable swelling. In order to be absorbed into a polymer,
there must be sufficient space or chain flexibility to accommodate a liquid molecule. Therefore, amorphous polymers above their
Tg will swell more easily than those below their Tg, and crystalline blocks generally show very little absorption.
250 Environmental Effects on Crack Growth in Polymers

For a liquid molecule to be stable when dissolved in a polymer, the overall interaction (free) energy change must be negative.
This is made up from an entropy term, which always favors mixing and increases as the temperature increases, and an enthalpy
term, which depends on the chemical interactions. Consideration of these terms leads to the formulation of solubility parameters
(δ) that govern miscibility. The solubility parameter is the square root of the cohesive energy density, with units of (stress)1/2. The
first treatment of this was due to Hildebrand (Hildebrand and Scott, 1950) which resulted in the prediction that if the term
(δp–δs)2 was small, then mixing was favorable. The subscripts p and s refer to polymer and solvent, respectively.
This has then been taken further by Hansen (1969), who realized that different types of molecular interactions will act
somewhat independently. He used three terms, δd, δp, and δh to account for dispersive, polar, and hydrogen bonding interactions,
respectively. The total solubility parameter was then given by the square-root of the sum of the squares of these terms and
solubility was maximized when the following term was minimized:
h
2
2
2 i1=2
δdp
δds þ δpp
δps þ δhp
δhs

This approach generally gives good correlations with solubility concentrated in a region where the solubility parameters are
close, but it is certainly not universal.
Similar terms have been developed that define an interaction parameter χ (Mai, 1986). Such models have been shown to
predict solubility of polymers in solvents reasonably well with total dissolution occurring when the term above was less than 5
MPa1/2.
One of the main difficulties with this approach is the determination of the solubility parameters. Although they can be
measured and derived from evaporation enthalpies for liquids, they are much harder to determine for polymers. They can be
derived indirectly by determining the degree of swelling/solubility of a polymer in a range of solvents for which the solubility
parameters are known. A more direct method of determining the one-dimensional (1D) solubility parameter for polymers,
including semi-crystalline polymers has been proposed by Tanaka (1992), which shows quite good agreement with values
obtained indirectly from solubility measurements.
The difficulty of determining polymer solubility parameters is especially true for the partial solubility parameters. One of the
most consistent methods to derive them for polymers is to use the method of group contributions. This uses the idea that each
molecular group has its own contribution to each of the parameters. They can then be added, in an appropriate manner to generate
the overall values. The problem with polymers having large and variable molecular weights is overcome as the total molecular
volume is a key term in the calculations. For polymers, the methods are generally applied to individual repeat units.
Two methods of deriving these solubility parameters are generally used; the first due to Hoftyzer and Van Krevelen (1976) and
the second to Hoy (1989). Both are described in detail, with tabulated values in Van Krevelen (1997). Values of solubility
parameters for common polymers and for selected liquids are given in Tables 2 and 3.
The effects of temperature are relatively straightforward, in that the solubility generally increases as the temperature increases.
These temperature effects are encompassed within the entropy terms and also in expressions for the interaction energy.
The situation for liquids dissolving into a solid polymer is more complex than a polymer dissolving in a solvent as the
mechanical and structural details of the solid material become important. The best treatment of such cases, for both amorphous
polymers and semicrystalline polymers is given by Flory (1978). These treatments are based on interaction parameters as men-
tioned above, but these are not reproduced here as they are quite involved. They can be used to predict the solubility of a liquid in
a solid polymer at a given temperature. The theories assume that the amorphous polymer is relatively mobile and swelling stresses
do not occur. These theories predict the equilibrium volume fraction of a solvent in a polymer as follows:

lnfs þ ð1
fs Þ þ χ ð1
fs Þ2 ¼ 0
where fs is the solvent volume fraction and χ the interaction parameter.
Examples of the free energy of mixing for a liquid into a polymer for various values of χ are shown in Fig. 7. From this it can be
seen that with a low interaction parameter, no free energy minimum is seen and complete dissolution occurs. As the interaction

Table 2 Solubility parameters for various polymers (MPa1/2).

Polymer δ (total) δd δp δh V

PTFE 12.7 49.5

Polyethylene 16.5 16.5 0 0 32.9
Polypropylene 17.8 17.6 0.6 0 49.1
Polyisoprene 18.0 17.4 3.1 3.1 75.7
Polybutadiene 18.8 18.0 5.1 2.5 60.7
Polystyrene 20.1 17.6 6.1 4.1 98.0
PET 21.3 143.2
PVC 22.5 19.2 9.2 7.2 45.2
PMMA 23.5 18.8 10.2 8.6 86.5
Polyacrylonitrile 25.6 44.8
 Environmental Effects on Crack Growth in Polymers 251

Table 3 Solubility parameters for some commonorganic liquids (MPa1/2).

Liquid δ (total) δd δp δh V

Butane 14.1 14.1 0 0 101.4

Hexane 14.9 14.9 0 0 131.6
Decane 15.8 15.8 0 0 195.9
Diethyl ether 15.8 14.5 2.9 5.1 104.8
Oleic acid 15.8 14.3 3.1 5.5 320
Cyclohexane 16.8 16.8 0 0.2 108.7
Amyl acetate 17.2 15.3 3.1 7 148.8
Trichloroethane 17.6 17 4.3 2 100.4
Stearic acid 17.6 16.4 3.3 5.5 326
Carbon 17.8 17.8 0 0.6 97.1
tetrachloride
Ethyl acetate 18.2 15.8 5.3 7.2 98.5
Dioctyl phthalate 18.2 16.6 7 3.1 377
Toluene 18.2 18.0 1.4 2 106.8
Benzene 18.6 18.4 0 2 89.4
Dekalin 18.8 18.8 0 0 156.9
Methyl acetate 18.8 15.5 7.2 7.6 79.7
Methyl ethyl 19 16 9 5.1 90.1
ketone
THF 19.4 16.8 5.7 8 81.7
Acetone 20.1 15.5 10.4 7 74
Acetaldehyde 20.3 14.7 8 11.3 57.1
Acetic acid 21.3 14.5 8 13.5 57.1
Aniline 22.5 19.4 5.1 10.2 91.5
Butanol 23.1 16 5.7 15.8 91.5
Acetonitrile 24.6 15.3 18 6.1 52.6
Ethanol 26.6 15.8 8.8 19.4 26.6
Methanol 29.7 15.1 12.3 22.3 40.7
Ethylene glycol 32.9 17 11 2 55.8
Glycerol 36.2 17.4 12.1 29.3 73.8
Water 47.9 15.5 16 42.4 18

parameter increases, a minimum does occur, indicating a finite degree of swelling at equilibrium. As the interaction parameter
increases further, the minimum shifts to lower and lower volume fractions (never reaching zero) showing that solubility decreases
significantly.
The theory can then be taken further to incorporate the effects of the elastic resistance to swelling from a solid. It is fairly clear
that to incorporate a fluid into a polymer, the expansion due to swelling will be resisted by the polymer network, especially with
networks that are bound at cross-links or crystalline regions. Indeed, one commonly used method of determining the cross-link
density in rubbers is to measure the degree of swelling. These effects have been incorporated into the Flory–Rehmer expression
(Flory, 1978), which has the following form:
" #
lnfs þ ð1
fs Þ þ χ ð1
fs Þ2 G
1=3
¼

Vs ð1
fs Þ RT

where Vs is the liquid molecular volume, and G is the polymer shear modulus. This expression predicts that, as the shear modulus
increases and as the liquid molecular size increases, the degree of solubility decreases, as is also seen experimentally. This
expression can also be used in reverse to determine the interaction parameter once the modulus and solubility are measured
(Malone et al., 1993).
The theories are also extendable to multicomponent systems; however, it should be noted that as the number of components
increases, the complexity of the theoretical treatment increases dramatically. Simplifications can be made if it is assumed that the
polymer and one or more of the liquid components are essentially incompatible. In this case, the system can be approximated to a
binary one with a dilute solvent.
Most studies of ternary mixtures have considered complete solution of the polymer. For instance, Cowie and McEwen (1987)
studied the effects of solvent composition on polystyrene dissolved in diethyl ether/nitro-methane mixtures. They found that
although the theoretical models of Flory and others were generally close, some nonideality due to solvent synergism was seen.
Synergistic effects between solvents have also been seen with binary liquids dissolved in polymers. Vahdat (1993) found that
polymethyl glutamate immersed in mixtures of water/methanol and benzene/cyclohexane showed some nonideality in terms of
the amount of each solvent absorbed at equilibrium. Further theoretical work showed that, with two highly soluble liquids,
significant enhancement occurs, with the solubility of each increasing above the (single liquid) theoretical prediction. When one
252 Environmental Effects on Crack Growth in Polymers

Fig. 7 The free energy of mixing for polymer/solvent mixtures with various values of interaction parameter, χ, increasing from curve 1 to
curve 5.

liquid is highly soluble and the other has only limited solubility, the highly soluble one enhances the solubility of the other, but
its own solubility is decreased below the theoretical value. Where both liquids have limited solubility, an ideal situation exists,
with predictions matching the theory. Such work has been extended to consider copolymers and cross-linking by Shiomi et al.
(1995).
The effects of stress on solubility are also of some importance. An applied dilatational stress will provide an additional energy
promoting swelling. This has been analyzed by Mai (1986) following work by Gent (1970). The energy term arising from the stress
can be added to the equation for the energy of mixing, giving an overall formula as follows:
    
1
DGs ¼ RT ln 1
fp þ 1
f þ χf2p
DVs
m p
where DGs is the free energy of mixing; fp is the polymer volume fraction; m is the ratio of molecular volumes of polymer segment
and solvent; χ is the interaction parameter; D is the negative hydrostatic pressure; and Vs is the solvent molecular volume. This
equation can be used, along with a suitable equation for the interaction parameter, to predict the solubility of a solvent in a
polymer under a particular stress.
The theory above breaks down when the structure of the material becomes inhomogeneous. For instance, under stress,
fibrillated and voided regions such as crazes can form, which will show preferential liquid absorption, almost by a capillary action.

Diffusion in Polymers
The diffusivity or diffusion coefficient D is the amount of matter passing per second through a unit area under the influence of a
unit concentration gradient (dc/dx), as is expressed by Fick’s law:
 
dm dc
¼ DA
dt dx
where m is the amount of matter and A the area. If the diffusion coefficient is independent of concentration and time, then the
process is described as Fickian. Most small penetrant molecules obey Fickian diffusion, which is characterized by a weight increase
that is proportional to (time)1/2 for a simple immersion test.
In general, diffusion into polymers may be regarded as a thermally activated process, involving energy jumps of the diffusing
species from one site to another. As such, it can be characterized by an Arrhenius equation:
 

ED
D ¼ D0 exp
RT
where D0 and ED are constants for a particular polymer and diffusing species, respectively.
 Environmental Effects on Crack Growth in Polymers 253

The activation energy for diffusion, ED, is the most dominant factor in this and describes the energy needed for a diffusing
molecule to jump from one position to another. This will in turn depend on several other factors including the molecular size, the
amount and distribution of free volume within the polymer and the flexibility of the polymer chains.
Although it is often most accurate to measure the diffusion terms experimentally, they can be estimated from the molecular size
of the diffusing species and Tg of the polymer. Van Krevelen (1997) has derived formulas for these values as follows:
For rubbers:
 2 h
ED sx
2 i
10
3 ¼ 7:5
2:5  10
4 298
Tg
R sN2
and
ED
logD0 ¼ 10
3
4
R
where sx and s2N are the molecular diameters of the diffusing molecule and nitrogen (as a standard), respectively. The value for
nitrogenis 38 nm. R is the gas constant and Tg is the glass-transition temperature.
For amorphous polymers:
 2 h
ED sx
3=2 i
10
3 ¼ 7:5
2:5  10
4 Tg
298
R sN2
and
ED
logD0 ¼ 10
3
5
R
For semi-crystalline polymers, it is assumed that diffusion occurs only through the amorphous regions (Gugumus, 1995;
Hedenqvist, 1996).
Molecular dynamics simulations have been performed for studying small-molecule diffusion (Gee, 1995; Theodorou, 1996).
The basis behind these is to use powerful computational approaches to model the detailed cooperative motions of polymer chain
segments and diffusing species. Although they have the advantage of being able to account for molecular shapes as well as sizes,
the level of detail required (including complex viscoelastic models) and the processing time involved make them relatively
unwieldy in many cases.
There are two further limiting cases that should be discussed. The first is when the diffusion proceeds to a level that approaches
the equilibrium solubility. In this case, the standard Fickian case of weight gain increasing with square root of time no longer holds
as diffusion fronts from both sides of a sample have met.
The second case is when the assumption that the diffusion coefficient is independent of concentration is not valid. In such
cases, the diffusing molecules cause the chains to be pushed further apart, generating more free volume and molecular mobility,
thereby increasing the diffusion coefficient. In such cases, termed case II diffusion, the weight gain is linear with time and a sharp
diffusion front develops. The reason for this is that, as the diffusing fluid enters the polymer, it causes plasticization and a
consequent increase in molecular mobility. The diffusion rate is therefore increased in regions with higher concentration of the
diffusing species, which leads to the observed sharp distinction between swollen and nonswollen regions. In such cases, the value
of D will be significantly higher than would be expected from the equations given above. One way of dealing with this would
possibly be to assume that, as the concentration increases, Tg decreases for the polymer and so the value of ED will decrease. This,
however would give the value for self-diffusion in the swollen layer and this is not the ratelimiting step. Rather, the kinetics of
progression of the swollen/nonswollen front is the key step, and this is harder to predict.
A good illustration of case II diffusion was highlighted by Thomas and Windle (1981) for methanol in PMMA. They observed
case II diffusion at room temperature, with a sharp diffusion front and linear weight and thickness increasing with time. When the
diffusion fronts from either side of their samples met, relaxation of constraints led to sudden increases in sample length. They
compared their results to Flory–Huggins theory and concluded that it is possible to determine the interaction parameter χ from the
degree of swelling when the fronts have just met. They also found that, at higher temperatures, the situation changed to one
between case II and Fickian, presumably due to the increased molecular mobility in the unswollen region. Although they did not
exceed the glass transition temperature of the PMMA, it would be expected that, above Tg, the extra molecular mobility from
plasticization by the diffusant would have little effect. Indeed, no true examples of case II diffusion have been shown for PE or PP
at room temperature, mainly because the amorphous regions (where diffusion occurs) are above Tg and so exhibit relatively high
mobility before any swelling has occurred.
Generally, case II occurs only for amorphous polymers below their Tg and with highly soluble diffusing species. If it does occur,
however, significant stresses can be generated from the sharp diffusion fronts. This was also studied by Thomas and Windle
(1981), who found that during swelling, compressive stresses were generated at the surface. With desorption in air after swelling,
significant tensile stresses were developed, which were large enough to lead to crazing. Kawagoe and Nunomoto (1991) also found
that compressive swelling stresses in the PMMA/methanol system were high enough to considerably hinder fatigue failure.
As the occurrence of case II diffusion is due to plasticization, it is expected that diffusants that have higher solubilities will be
more likely to cause case II diffusion, as is indeed generally seen. In addition, there can be effects of synergism between
components of mixed environments, in a manner similar to the solubility effects. Suzuki et al. (1993) found that diffusion rates in
254 Environmental Effects on Crack Growth in Polymers

mixtures of benzene and cyclohexane were higher than the rates in either liquid alone. A similar effect was found by Valdakar et al.
(1993) for water/dioxane and water/propanol mixtures.
The final issue, which will be seen to be of prime importance later is the effect of stress on diffusion. Theories predicting that a
dilatational stress will increase the solubility of liquids in polymers have been mentioned above, but it has also been assumed that
a dilatational stress will increase the diffusion rate. Despite this, direct experimental evidence for this point is somewhat limited
and no formal theories have been developed. Certainly, if a dilatational stress increases the free volume, especially in localized
regions such as voids and crazes, then it would be expected that the diffusion rate would increase.
More et al. (1992) using laterally constrained PMMA samples in methanol found that, as the degree of lateral compression
increased, the rate of diffusion decreased. They found that a compressive stress of about 0.3MPa was sufficient to suppress
diffusion almost completely.
A more extensive study of PEEK in toluene has been published by Wolf and Fu (1996). They investigated the effects of applied
tensile stress, temperature and degree of crystallinity on the weight gain. Their results are very dramatic, showing that an applied
tensile stress can have a large effect on the weight change. They found that, for totally amorphous material, stresses of 7MPa and
10MPa produced an increase in both the rate of diffusion and the equilibrium absorption. They also found that the induction
time, attributed to the establishment of a surface-swollen layer, decreased with increasing stress. For higher levels of crystallinity,
the stress required to see this effect increased, where a critical stress between 25MPa and 30MPa was needed to accelerate the
diffusion process. Their results show several key features that are worth considering further.
The applied stress does have a significant effect on the diffusion rate, an effect that seems to be more linked to the amorphous
regions.

(i) With higher levels of crystallinity, the increase is only seen with higher stresses, suggesting that, when the amorphous
regions are more constrained, the effect is dramatically reduced.
(ii) There is a critical stress for this effect, which suggests that it is linked to the formation of sufficient free volume in the
amorphous region (maybe the microvoids discussed previously).
(iii) Higher temperatures increase the effect, again suggesting that critical molecular mobility in the amorphous regions is
important.

Surface Energy
The surface energy (or surface tension) will play a part in any surface interaction between a polymer and a fluid. It will determine
the degree of wetting, the ease with which fluids can penetrate into fibrillated regions, and, to some extent, the energetics of
fluid–polymer interactions. The importance of surface energy to environmental failure is somewhat debatable, and will be
discussed later.
The surface energy, g, expressed as energy per unit area, is another internal energy term, and has some similarities to solubility
parameters discussed above. It acts to minimize the free surface of a material, and controls many interfacial properties. The most
common expression of this for a liquid on a solid surface is via the contact angle. If the surface energies of liquid and surface are
similar, the contact angle will be low and wetting will occur. If the surface energies are dissimilar, the contact angle will be large
and beading will occur. An expression for the contact angle y, is given by
g
gsl
cosy ¼ s
g1
where the subscripts s, l, and sl refer to solid, liquid, and the solid/liquid interface, respectively. The difficulty with this is that it
does not allow direct evaluation of the interfacial energy, gsl, but this can be obtained from the following relationship (Girifralco
and Good, 1960):

gsl ¼ gs þ g1
2Fðgs g1 Þ1=2

4ðVs V1 Þ1=3
where F ¼  2
1=3 1=3
Vs þ V1

with Vl and Vs being the molar volumes of liquid and solid, respectively. This equation therefore allows the calculation of the
interfacial energy between a polymer and a liquid if the individual surface energies are known.
The surface energy has been found to have the following relationship with the solubility parameter (Hildebrand and Scott,
1950):
g ¼ 7:55  105 δ2:33 V 0:33
An alternative form of this relationship, using the cohesive energy density Eco exists in the following empirical formula:
g ¼ 0:75E0:66
co

Measurements of surface energies can relatively easily be achieved via contact angle measurements in the solid phase, or from
extrapolations from the melt phase, or by a theoretical method based on an additive quantity termed a parachor (Van Krevelen,
1997). Table 4 shows surface energies for a range of polymers, along with the values predicted from the cohesive energy density.
 Environmental Effects on Crack Growth in Polymers 255

Table 4 Surface energies of some common polymers (mJ m–2).

Polymer g observed From Eco gd gp

PTFE 20 20 18.4 1.6

Polypropylene 29.6 33 29.6 0
Polyethylene 35.7 30 35.7 0
Polystyrene 40.7 38 35.4 6.3
PMMA 41.1 38 29.6 11.5
PVC 41.5 40 39.5 2
PET 44.6 42 35.6 9.0
Nylon 6 46.5 64 32.5 14

Fig. 8 The effect of molecular weight of a polymer on its surface energy.

Table 4 shows a reasonable agreement between the predicted and the measured values, except where significant hydrogen
bonding occurs. In these cases, it is often better to split the surface energy into component forms, as is done with solubility
parameters. Although full 3D surface energy terms have rarely been used, the surface energy is often conveniently split into a
dispersive component and a polar component (including hydrogen bonding):

g ¼ gd þ gp

Values for these terms are also given in Table 4, from which it can be seen that there is some similarity with solubility parameter
values.
The temperature will obviously have some effect on the surface energy, as the internal bonding will vary with temperature. It
has been found that the major influence on this is via molar volume changes, and so the surface energy varies in a manner
proportional to the density raised to the fourth power (Van Krevelen, 1997). This means that the surface energy will decrease as the
temperature increases, but the exact nature will depend on the density variation. The interfacial tension between a polymer and a
liquid will therefore be complex. The density of the solid polymer will vary less than that of the liquid, and so the surface energy of
the liquid will decrease more rapidly with temperature. Therefore, if the surface energy of the polymer is greater than that of the
liquid, an increase in temperature will increase the contact angle, and vice versa.
The molecular structure of the polymer will also have some effect on the surface energy, with crystalline regions having a
significantly higher surface energy than amorphous regions. The molecular weight also has some effect, with the surface energy
increasing as the molecular weight increases, but reaching a saturation value at high molecular weights. This is shown in Fig. 8,
using the values for polystyrene (Wu, 1982).
256 Environmental Effects on Crack Growth in Polymers

Accelerated Crack Growth Due to Chemical Degradation

Overview
As discussed before, a distinction needs to be made between the chemical effects that a fluid environment will have on crack
growth in a polymer and the physical effects. Although a single environment may produce both of these types of effects and their
consequences will overlap, they do proceed by distinctly different mechanisms. In this section, the mechanisms considered all
involve main-chain bond breakage due to chemical reactions. In many cases involving chemical degradation, chemical attack
occurs uniformly, leading to general weakening of the material or loss of material from the surface. In order for a chemical
degradation process to have an influence on craze and crack formation and growth, one or more of the following factors have to be
operating:

(i) The degradation reactions produce a reduction in molecular weight so as to cause an overall reduction in fracture toughness
that then leads to easier crack formation.
(ii) The degradation occurs preferentially at a craze/crack tip due to the enhancement of the degradation rate by the local stress
concentration.

(i) The degradation occurs preferentially at a craze/crack tip due to the greater availability of surface in the highly fibrillated
craze region.
(ii) The degradation occurs preferentially at a craze/crack tip due to stagnation of the environment or buildup of degradation
products.

Not surprisingly, with the large variety of polymer structures in use, in a wide range of chemical environments, the number of
separate reaction mechanisms is large, especially as each degradation pathway is normally very complex, involving many inter-
mediate chemical species. Despite this, some generalizations can be drawn.
The vast majority of degradation reactions involve the scission of the main polymer backbone chain. In some cases, this may
not be the initial step, but it quickly follows if it is not. These scission events normally yield free-radical ends, which are highly
reactive unpaired electrons. These will rapidly react with neighboring molecules, which may be other polymer chains, degradation
products, oxygen, water or whatever environment is present. The degradation reactions in the presence of oxygen have been most
widely studied, and so it is appropriate to give a brief overview of these here (Grassie, 1979).
For simple hydrocarbon polymers (RH) the initiation step generally involves the formation of hydrocarbon free radicals R  ,
which then react with oxygen to form peroxide radicals RO2 . All polymeric materials will, via their polymerization and processing
history, have developed a certain concentration of peroxide radicals. The propagation of further oxidation then depends on slower
reactions and in particular the abstraction of hydrogen from adjacent polymer molecules by the peroxide:
RO2 þ RH-ROOH þ R 
then immediately
R  þ O2 -RO2
which reacts with RH again to produce another free radical R  (which immediately reacts with oxygen to replenish the peroxide
concentration) and an unstable hydroperoxide, which via decomposition or other reactions generate a further two reactive species.
The conversion of peroxides to hydroperoxides is the rate-determining step for the chain reaction.
Main-chain scission involves the breaking of C—C or other covalent bonds in the polymer backbone. It can occur from direct
chemical attack or as a secondary process, b scission, which results in a fragmentation of the polymer chain forming carbonyls and
alkyl radicals:
Chain scission can also occur from decomposition of polymeric hydroperoxides:

The Effects of Material Structure

The effects of chemical degradation on semicrystalline polymers have almost exclusively shown that degradation is concentrated in
the amorphous regions. This is not too surprising as the amorphous regions will have far less interchain bonding, greater mobility
as well as exhibiting higher solubilities and diffusion rates. In many cases, the difference is such as to give complete etching away
of the amorphous regions in preference to the crystalline regions, an effect which can lead to intercrystalline fracture. Such an
example of chemical attack is shown in Fig. 9 for polypropylene exposed to an oxidizing hypochlorite environment.
A comprehensive treatment of the rates of thermal oxidation of polyethylene and polypropylene has been conducted by
Gugumus (1998a, 1998b), who found that the ratedetermining step is usually diffusion of initiation species into the material. A
full theoretical model was then developed which takes into account the semi-crystalline nature and proposes that oxidation
spreads from one amorphous domain to another. This model shows good correlation with experimental results. The model was
initially developed for LDPE, but was then shown to be applicable to HDPE, LLDPE, and PP.
The effects of molecular weight are also important, although this is an area that has not received much study. The main effect of
a higher molecular weight is to provide some protection against the effects of chain scission. In most polymers, there is a critical
level of molecular weight, below which the properties, especially the toughness properties decrease significantly. This molecular-
 Environmental Effects on Crack Growth in Polymers 257

Fig. 9 The chemically degraded surface of a sample of polypropylene after exposure to hypochlorite, showing preferential attack of the
amorphous regions.

weight level generally corresponds to that required for entanglements. If a material starts with a higher molecular weight, it
requires more scission events to lead to a critical reduction in toughness.

The Effects of Temperature and Stress

As with all chemical reactions, an increase in temperature leads to an increase in degradation rate. This is normally expressed via an
Arrhenius equation, with an activation energy Eact. As the reaction pathways that occur during degradation are often extremely
complex, it is also likely that some intermediate reactions will become more or less thermodynamically stable. As the key
degradation reaction will be main-chain scission in most cases, the activation energies for degradation in a certain environment are
strongly dependent on the bond strength of the weakest bond in a polymer. Activation energies for thermal degradation show this
correlation well (Van Krevelen, 1997) with simple carbon-chain polymers having fairly high activation energies. Polymers with
oxygen or nitrogen bridges in the main chain (nylons, polyesters, PPO) have lower activation energies.
The effects of stress on chemical degradation form probably the most significant factor to stress-enhanced degradation.
Generally, an increased tensile stress on polymer chains will increase the rate of degradation. This will occur either because the
stressed chains are more susceptible to scission reactions or that, once broken, chains are less likely to simply recombine if they are
under stress. The exact relative importance of these two mechanisms has yet to be fully established.
The degree to which an applied stress increases the reaction rate can be modeled through a modified Arrhenius equation, as
given below:

 
Eact
as
kb ¼ k0 exp

RT

where kb is the rate of main chain scission, k0 is the natural vibration frequency, and a is the stress-enhancement factor, which has
generally been found to be between 0.003 and 0.007 for most unoriented polymers. From published data on fiber and film
degradation, the value of a seems to be lower for highly oriented material.
Using this formula and the value of activation energy for hydrolysis of polycarbonate of 80 kJ mol–1 (Bair et al., 1981), the
expected effects of stress and temperature on bond fracture can be seen in Fig. 10. This shows that as the stress is increased, the rate
of bond fracture increases dramatically, but the significance of the stress decreases as the temperature increases. This tends to fit
experimental and industrial evidence, with uniform chemical degradation more likely at higher temperatures and localized attack,
often at crack fronts, more likely at lower temperatures.

Degradation in Specific Environments

For information on the specific effects of individual environments on individual materials, chemical resistance databases are
available (see, e.g., Morris, 1994), with similar information available from material suppliers. The information given in such
sources is, however, somewhat limited and may simply address compatibility. In the sections below, some generalizations are
given on the main types of degradation reactions occurring with polymers.
258 Environmental Effects on Crack Growth in Polymers

Fig. 10 The theoretical rate of bond breakage due to hydrolysis in polycarbonate as a function of stress and temperature.

Hydrolysis, acidolysis, and alkolysis

Hydrolysis is one of the most common forms of chemical attack with polymers and it refers specifically to degradation due to the
hydrogen ions (H þ ) or hydroxyl ions (OH–) present in water. The concentrations of these ions are increased in acid or alkaline
conditions, and so a polymer that is degraded by H þ ions in water is said to be very prone to acidolysis.
The polymers that are most susceptible to hydrolysis reactions are those that have strong hydrogen-bonded groups (such as
ester, amide, imide, and carbonate) within the main chain. Such polymers include polyesters, nylons, polyimides, and poly-
carbonate. Polymers, such as the acrylic polymers which have such hydrogen-bonded groups within side groups and branches will
be less prone to hydrolysis, but will certainly not be immune.
The hydrolysis of polycarbonate has been extensively studied, and a schematic of the degradation route is given below:
Polycarbonate is known to suffer particularly from hydrolysis at elevated temperatures. Although the reaction rates are not
especially high, a relatively small reduction in molecular weight can have a serious effect on the toughness properties. Commercial
polycarbonates, especially moulding grades have fairly low molecular weights (Mw) of around 4  104 whereas the critical
molecular weight for chain entanglements is B 2.5  104. Therefore, an average of a single scission event per chain will bring
about almost complete loss of toughness. The hydrolysis reaction has been found to break the polymer back to bisphenol A (Bair
et al., 1981), a reaction that is accelerated by higher temperatures and levels of humidity (Gardner and Martin, 1979; Ghorbel et al.,
1995). The effects of stress have also been noted (Ghorbel et al., 1995), with the rate of degradation increasing tenfold under the
action of a tensile stress of a few MPa. The scale of this effect agrees reasonably with the expected variation derived from activation
energies shown in Fig. 10. In addition to hydrolysis in service, processing of polycarbonate with insufficient drying can lead to
hydrolysis and a reduction in molecular weight before entering service (Wright, 2001).
Nylons are also known to suffer from hydrolytic attack. Due to the significance of hydrogen bonding within the structure of
nylons, water absorption can be significant, often leading to hydrolysis, especially in acidic conditions. The mechanism of attack is
attributed to hydrolytic cleavage of the amide link, the mechanism that has also been found to be responsible for water damage in
polyaramid fibers such as Kevlar.
PET fibers and films have also been found to be susceptible to hydrolysis, via attack of the ester linkages. This has been found to
be particularly significant under alkaline conditions, and where significant stresses are present (Ezrin and Lavigne, 1994). Acidic
conditions can also increase the degradation rate (St. Clair et al., 1993).
High-temperature engineering polymers tend to have bonds other than the simple C—C ones within the main chain and so
often suffer from hydrolysis. This has been found with polyimides and polyetherimides, especially in alkaline environments at
high temperatures. Another slightly different form of attack with these materials has been documented as the formation of
microcavities from absorbed water, assisted by hydrolysis reactions (Robeson and Crisafulli, 1982). This study found that the
presence of sodium chloride (0.1% dissolved in the polymer) promoted this type of failure by providing extra nucleation sites.
Acetals (polyoxymethylene) are well known to suffer from hydrolysis in acidic environments. The presence of even dilute acids
causes a dramatic loss in crack resistance. For instance, Zuchowska et al. (1999) showed that fatigue lifetimes in 1% HCl were
reduced by a factor of 100. A further reduction by a factor of 20 was seen when the HCl concentration increased to 6%. This rapid
degradation in the presence of a stress and mild acidic conditions has led to numerous in-service failures with acetal components
(Wright, 2001).
 Environmental Effects on Crack Growth in Polymers 259

One further factor that can lead to accelerated hydrolysis and oxidation is the presence of transition metal ions, especially
copper. These can significantly increase the rate of degradation, partly due to an increase in the H þ or OH– ion concentration, but
also by acting as electron sinks/sources to catalyze many of the degradation mechanisms.

Degradation in oxidizing liquids

It has been found that for many polymers, exposure to HO2 and especially HO radicals from peroxide environments accelerate
the rate of oxidation. The mechanisms of this oxidation are generally similar to those seen with thermal and photo-oxidation and
have been studied for polyethers (Zuchowska et al., 1999), PP (Pan et al., 1998; Girois and Delprat, 1999), UHMWPE (Haken,
1996; Seidmann and Dulog, 1989; Pruitt and Bailey, 1998), PPO (Kaczmarek et al., 1995), and PEO (Shlyapnikov and Tyuleneva,
1997). These have generally shown similar features, in that attack occurs preferentially (or exclusively) within the amorphous
regions (especially tie molecules) and that oxygen-containing polymers are more severely attacked. Work with UHMWPE has
shown that it is quite resistant to the HO2 radical and oxidation only really proceeds if the HO radical is formed. With peroxide
solutions, this requires high concentrations, higher temperatures, and UV light or gamma irradiation.
Work on polypropylene has found that the oxidation reaction quickly becomes limited by the diffusion of oxygen through the
material (Pan et al., 1998). Other studies (Girois and Delprat, 1999) examining peroxide oxidation during processing have studied
the kinetics occurring under these conditions (noting that mixing processes give an even distribution of species through the
material). It may be possible to combine these approaches to give a treatment for peroxide-initiated oxidation in the bulk
solid phase.
Sodium hypochlorite (NaOCl) is often used as a bleaching chemical for textile fibers. The effect of hypochlorite on cellulose
fibers is therefore studied extensively (e.g., Abdul-Hafis, 1997). Hypochlorite is a nonspecific oxidizing agent and its activity
strongly depends on the pH of the reaction medium. Hypochlorite can attack ether bonds and it is also capable of cleaving C—H
bonds. The oxidizing reactions are fastest at pH’s between 2 and 7.5, when the chlorite is in the protonated form HOCl.
The general oxidizing reaction is given by

HOCl þ Polymer-Polymer
O þ HCl

This reaction will cause a lowering of the pH of the solution. Free radicals play an active role in the oxidation with hypochlorite.
A free radical chain is set up as follows:

HOCl þ OCl
- ClO þ Cl
þ  OH


OH þ OCl
- ClO þ OH


ClO þ OCl
-2Cl
þ O2 þ  OH

According to this scheme, the radical  OH will react with the chain. However, in alkaline solutions, hypochlorite can be given
as:

HOCl þ OH
-OCl
þ H2 O

and OCl– is the oxidant.

Chlorine attack
Cracking due to a chlorinated atmosphere (gaseous chlorine or HCl) has been observed with several polymers. With polyethylene,
polypropylene, and polybutylene, it has been found that the presence of chlorine will accelerate oxidation reactions, but degra-
dation is only severe when elevated temperatures or UV light are also present. In these cases, the chlorine can disassociate into
chlorine radicals, which will very easily abstract a hydrogen atom from a polymer chain, which is then likely to undergo a scission
reaction. It has been seen that polypropylene and especially polybutylene are more susceptible than polyethylene, due to the easier
abstraction of hydrogen from side groups (Fettes, 1964).
Chlorine gas has also been found to cause cracking in PVDF (Wright, 2001), which is normally regarded as a material with
good chemical resistance. In this case, the cause of cracking was the substitution of a chlorine atom in place of one of the hydrogen
atoms. This causes significant swelling stresses which can lead to cracking.
PVC can be treated in a slightly different manner as the chlorine changes the degradation mechanisms to some extent. With this
polymer, the first degradation step is often abstraction of HCl, which causes two accelerating processes.
First, the adjacent chlorine becomes more susceptible to removal, which leads to sequences of conjugated polyenes. These cause
discoloration and are likely sites for further oxidative attack. Second, HCl acts as a catalyst for many of the subsequent degradation
reactions. This may be very important in low pH hypochlorite environments, where HCl is also present, although little direct
evidence is available for this.
Due to dehydrochlorination, PVC degradation is often less dependent on initiators and is more thermally-activated. This was
shown to be the case during oxidation with OH– ions (Abbas and Sorvik, 1973), which led to zeroorder rate kinetics (purely time
and temperature dependent). This is probably due to the large numbers of initiation points remaining in most PVC grades after
processing. The rate of degradation does, however, increase significantly with an available oxidant to form hydroperoxides.
One final mode of attack that can lead to cracking in PVC has been seen with plasticized PVC where certain solvents can extract
the plasticizer, leaving a much more brittle material, which can undergo cracking due to contraction stresses (Wright, 2001).
260 Environmental Effects on Crack Growth in Polymers

Nitrogen dioxide and sulfur dioxide

Nitrogen oxides, especially the dioxide and sulfur dioxide cause degradation to various polymers by similar mechanisms. Nitrogen
dioxide is another hydrogen abstracter, which then allows nitration of the polymer chain, leading then to chain scission, as shown
below:

RH þ NO2 -R  þ HNO2
R  þ NO2 -RNO2 -Scission

The degradation of polystyrene in nitrogen dioxide follows this mechanism with an activation energy of 67 kJ mol–1 (Jellinek
and Toyoshima, 1967). The rate of attack is dependent on the partial pressure of NO2, and also on the temperature, but also quite
dramatically dependent on the presence of air. In the presence of air, the rate of attack was increased by a factor of five with the
same partial pressure of NO2. As with chlorine attack, the presence of side groups increases the rate of degradation due to easier
hydrogen abstraction, shown by the increased resistance of polyethylene compared to other polymers (Igarashi and DeVries,
1983b).
The effects of stress on degradation in NOx has also been studied with nylon fibers (Igarashi and DeVries, 1983a), where it was
found that stress did cause an increase in attack rate, due to increased susceptibility of tie molecules between crystalline regions.
The effects of stress were only really significant at high stresses. For instance, a stress of 80% of the UTS caused a twofold increase in
degradation rate, whereas a stress of 90% of UTS increased the degradation rate to 20 times that under no stress.
A particularly strong form of degradation was seen with unsaturated polymers (natural rubber, polybutadiene), where nitro-
addition across the double bond leads to significant chain scission, or in some cases cross-linking (Jellinek and Flajsman, 1970).
Sulfur dioxide causes degradation by similar mechanisms, with a greater influence of oxygen and UV light, due to SO3 radical
formation (Jellinek et al., 1969). Degradation has also been shown to be affected by high stress levels in nylon in a manner similar
to NOx (Igarashi and DeVries, 1983a). With SO2, however, some recovery is possible at low stresses, showing that the chain
scission step is not as rapid as with NO2.

Environmental Stress Cracking

The Mechanisms of ESC

ESC was first observed and studied experimentally around the 1960s and the understanding of the complex mechanisms of failure
involved has gradually developed since then. As a phenomenon, it was first observed with both glassy thermoplastics exposed to
certain organic solvents and with polyethylene in contact with detergents and other organic fluids. It was fairly rapidly established
that the process was a purely physical acceleration of the brittle failure mechanisms occurring in the absence of an active
environment, involving craze formation and breakdown. The complete understanding and prediction of the process remains some
way off even as of early 2000s, but significant progress has been made since the phenomenon was first observed and now most
aspects are at least qualitatively understood.
One of the main difficulties preventing a complete understanding of ESC is the complexity of interacting processes that are
thought to play a part. The main mechanisms are listed below.

Plasticization
Considerable evidence exists which shows that the degree to which an environment can plasticize a polymer determines the
propensity to ESC. However, in order for plasticization to lead to brittle failure along a craze, the plasticization must be localized,
which could be due to:

(i) increased plasticization at the highly stressed region at the tip of a growing craze or stress concentration, due to stress-
enhanced solubility;
(ii) increased plasticization in the highly voided and fibrillated region at the tip of a growing flaw, due to a higher surface-
energy-to-volume ratio;
(iii) increased diffusion due to the high local stresses at the tip of a flaw, leading to greater plasticization; and
(iv) increased diffusion along the voided and fibrillated zone of a craze leading to greater plasticization.

Surface energy effects

Evidence also exists that the effects of surface energy are important, via two mechanisms:

(i) Decreased surface tension in the liquid environment reducing the energy required to produce new surface, thereby making
void and craze formation easier.
(ii) Capilliary filling of crazes and voids by the environment due to lower surface energy leading to hydrostatic pressures within
the voids, aiding craze formation.

The following describes the most important experimental and theoretical work that has been conducted in an attempt to
establish the relative importance of these mechanisms.
 Environmental Effects on Crack Growth in Polymers 261

Fig. 11 The critical strain for crazing, plotted against solubility parameter for PPO in a range of organic fluids. (reproduced with permission from
Bernier and Kambour, 1968).

Some of the best early evidence for the plasticization mechanism came from Bernier and Kambour (1968), who studied poly
(phenylene oxide) in a range of environments. They found that the critical strain to cause crazing in a simple bend test varied in a
manner consistent with the solubility parameter of the environment. This is shown in Fig. 11, from which it can be seen that there
is a distinct minimum where crazing is most severe. It is assumed that the solubility parameter of the polymer lies at this minimum
and so the degree of ESC depends on the difference between the solubility parameter of polymer and environment. This would
certainly fit with the plasticization mechanism: the closer the solubility parameters, the greater the solubility and plasticization.
This was investigated further by measuring the equilibrium solubility of the environments in the polymer and the Tg of the
swollen polymer. The correlation between strain to craze and solubility is shown in Fig. 12, from which it can be seen that the
proposed mechanism is supported by the evidence. It should be noted that due to experimental difficulties in obtaining fully
swollen samples, Tg values were only obtained for five fully equilibrated samples. The correlation between solubility and Tg of
swollen polymer seems reasonable in this case. As the Tg of a two-component material is dependent on the volume fraction and Tg
of both components, it is expected that the Tg of each liquid environment would also play a role; but, as these contributions are all
very much less than that of the polymer, the liquid volume fraction, is the main contribution.
This work was further extended to cover polysulfone, polycarbonate, and polystyrene (Kambour et al., 1974). Similar rela-
tionships between strain to crazing and Tg of the solvent-equilibrated samples were found, albeit with a few exceptions. The same
was true for the correlation with solubility parameter.
If a local reduction in Tg of the polymer accelerates the crazing process, this should be related to the test temperature. Such a
relationship has indeed been observed by several workers (Bernier and Kambour, 1968; Kam-bour et al., 1974; Kambour, 1977;
Kramer, 1983; Andrews and Bevan, 1972) and is shown in Fig. 13. As the value of Tg–Ttest decreases to zero, full plasticization will
occur, leading to a dramatic reduction in craze resistance.
Further evidence for the effects of plasticization has come from the generally good correlations between ESC resistance and the
difference in solubility parameters between polymer and environment. These studies will be discussed later, from the viewpoint of
ESC prediction.
The evidence for the influence of surface-energy effects is rather less conclusive. The main reason for surface-energy effects being
an attractive mechanism is the observation that ESC can be very pronounced in material/environment combinations that show no
measurable weight changes during immersion tests. One such example that has been widely studied is that of polyethylene in
Igepal(TM) detergents. Obviously, with such a surface-active environment that shows no measurable weight gain, the reduction in
surface energy is an obvious mechanism to suggest for the ESC observed. Certainly, many environments that cause severe ESC
show little bulk diffusion, but may diffuse significantly into the highly voided and stressed region at the front of a craze.
Surface energy considerations are probably of secondary importance, and certainly have been incorporated into several the-
oretical models of ESC, with some success (Andrews and Bevan, 1972; Kefalas, 1995; Kotoul, 1996). Of these studies, only Kefalas
(1995) proposes that surface energy reduction is the primary cause of ESC.
As experimental evidence of ESC phenomena has increased, the mechanism of plasticization has gained support. The studies
on ESC have covered a wide range of polymers and environments, including:
262 Environmental Effects on Crack Growth in Polymers

Fig. 12 The critical strain for crazing, plotted against degree of solubility for PPO in a range of organic fluids. (reproduced with permission from
Bernier and Kambour, 1968).

(i) glassy polymers in a wide range of organic solvents,

(ii) semi-crystalline polymers in organic solvents and surfactants,
(iii) PVC and polyethylene in hydrocarbon gas mixtures,
(iv) nylons in salt solutions, and
(v) polystyrene and PMMA in gases at low temperatures.

With all of these combinations, the same basic mechanism of failure was seen to occur, albeit with slight differences depending
on temperature, test method, timescale of testing, whether the samples had pre-existing cracks and so on.
A transition is often seen between localized plasticization (where ESC occurs) and bulk plasticization, where ductile failure
occurs. This transition can occur due to longer timescales of testing (Arnold, 1995c) or with more soluble environments (Arnold
et al., 1996; Breen, 1994). Breen (1994), studying the effects of gaseous environments on PVC, noted a change from ductile failure
due to bulk plasticization in benzene and toluene vapors, to brittle ESC in hexane, decane, ethanol, and methane vapors. Such a
gradual transition certainly indicates that plasticization effects dominate.
The effects of local plasticization on the behavior of crazes has been studied extensively. It has been well established that the
environment enhances craze growth and breakdown due to local plasticization. In the majority of cases, this is due to enhanced
surface drawing of craze fibrils (Breen and Van Dijk, 1991; Brown and Njoku, 1986; Ward et al., 1991). In a few cases, e.g.,
polystyrene in methanol (Krenz et al., 1976), the diffusion rate of the environment into the polymer is slow and occurs only within
the fibrils, rather than the surface of the craze. Under these conditions, fibril creep becomes the dominant mechanism. Work by
Miller and Kramer (1990) found intriguing evidence that the plastic flow processes occurring within a craze fibril greatly enhance
the diffusion rate. This is not too surprising as it is well known that plastic flow is a method of regenerating free volume within a
polymer (Struik, 1978).
The final stages of craze breakdown in an environment have also been studied, and the evidence is that final failure is almost
exclusively due to chain disentanglement, with very little evidence for any chain scission (Brown and Njoku, 1986). Even with
lightly cross-linked polymers, disentanglement seems to be the dominant mechanism, with chain scission occurring only in the
very last stages of fibril failure (Miller and Kramer, 1990).
The effects of the environment on the growing crazes would be expected to have an effect on the craze microstructure, and this
has indeed been studied extensively. The general findings from these studies are that plasticization tends to increase the craze fibril
diameter (Brown and Njoku, 1986) and decrease the craze fibril volume fraction. This has been found to be approximately
constant at B 0.5 for a wide range of polymer crazes in air, but in active environments it decreases with increased plasticization,
down to values as low as 0.1 in some cases (Yaffe and Kramer, 1981). These two findings together show that as plasticization
increases, the craze structure tends to become coarser.
The influence of the crystalline microstructure on environmental craze growth has also been studied (Berger, 1989; Lustiger,
1986). Not surprisingly, the amorphous regions tend to be the most susceptible to craze formation, due to increased fluid
solubility and diffusion as well as being regions of easier chain motion. Crazes have been seen to deviate substantially (by up to
301 from their normal path) to avoid crystalline blocks. The formation of a craze through a semi-crystalline material is shown
schematically in Fig. 14. In order to provide good ESC resistance in a semi-crystalline material, it is therefore important to have a
 Environmental Effects on Crack Growth in Polymers 263

Fig. 13 The critical strain for crazing for various polymer/environment combinations, plotted against the difference between test temperature and
Tg of the swollen polymer. (reproduced with permission from Kramer, 1979).

fine-scale microstructure with large numbers of tie molecules between the crystalline blocks. In polyethylenes, this is achieved
through a well-controlled bi-modal molecular-weight distribution.
The effects of gaseous environments have been found to have many similarities, although low temperatures are often needed to
see any effect. The crazing of polystyrene (Wu and Brown, 1982) and PMMA (Imai and Brown, 1976) observed in gases at low
temperatures has been shown to be due to ESC with plasticization as the dominant mode. The degree of plasticization was found
to be dependent on the square root of the gas concentration. Similar behavior was seen for polystyrene in CO2 at higher
temperatures (Wang and Kramer, 1982).

ESC Test Methods

Due to the complexity of ESC mechanisms, the effects of test method can be quite important. Over the years, a wide variety of test
methods have been used to study ESC (Turnbull et al., 2000), and these are discussed below, along with their respective
advantages, disadvantages and limitations.

Bend methods
The most commonly used and simplest type of test method involves flexural deformation. The principle behind all of these
methods is that a fixed bending deformation is imposed on a strip sample, followed by either immersion in the environment or
application of the environment to the surface. They have the common features of involving simple and cheap equipment, giving
rapid results and allowing relatively easy interpretation.
One of the simplest types of tests is the cantilever bend test, where a strip is clamped at one end, a fixed load is applied to the
other, and the environment of interest is applied to the surface of the sample. The maximum stress, which occurs on the top
surface adjacent to the clamp is given by
264 Environmental Effects on Crack Growth in Polymers

Fig. 14 The development of a craze in a semi-crystalline polymer. (reproduced with permission from Lustiger, 1986).

6FL
s¼
wt 2
where F is the applied force, L is the sample length, w the sample width, and t the sample thickness.
This has the obvious advantage of simplicity; however, a sensible choice of applied load is needed to achieve crazing in a
reasonable time. It can also be practically difficult to apply the liquid to the top surface of the sample, due to high volatility or low
viscosity. It should also be noted that this is a constant-stress method, in contrast to most of the bend methods that apply constant
deformation and, therefore, the constant-stress method tends to give more severe conditions.
The Bell telephone test was one of the earliest to be developed and is still popular, especially with more flexible materials. The
method involves bending pre-notched samples into a holder which is then fully immersed in the test fluid. Due to the high strains
involved in the notch region, the method is only really of any use as a comparative method for qualitycontrol purposes.
The elliptical bend jig has been used widely in experimental studies (e.g., Wyzgoski and Jacques, 1977), and offers a simple
method of determining a critical strain to crazing. With this method, a strip sample is fixed to an elliptical jig, with semi-major and
minor axes of a and b. The strain at any point, X, on the surface is given by
   2
bt 1 b2 2
e¼ 1

X
2a2 a2 a4

The environment can then either be applied to the surface or the whole jig can be immersed. The fact that the strain varies along
the jig gives the advantage that crazing will be observed up to a certain position, from which the critical strain can be determined.
One of the disadvantages of this method is that the points of clamping of the sample to the jig can act as craze initiation sites.
A more widely used bend method is the three-point bend jig, as described by Vincent and Raha (1972), and shown in Fig. 15.
This uses a simple jig in which a strip sample is bent between three circular rods. The whole sample is then normally immersed in
the fluid of interest. The maximum strain occurs on the outer surface of the strip opposite the central rod and is given by

6δt
e¼
L2
where δ is the deflection of the center of the sample and L is the span length of the sample. As with the elliptical bend jig, the strain
will vary along the surface, decreasing linearly to zero at the outer rods, and so a band of crazing of length b is normally seen, going
from the center to a distance corresponding to the limiting strain for craze formation. This critical strain can be given by
 
L
b
e ¼ emax
L

where emax is the strain at the center. This method has been used very widely, as it easily gives a value for critical strain for craze
formation. The method has also occasionally been modified so that a constant bending force is applied, with the bending strain
measured by displacement transducers or strain gauges.
One important consideration with the bend methods is the time required for craze formation. Generally, samples are immersed
for times between 1 h and 24 h, in order to give convenient experimentation. This would seem to be rather arbitrary, given the fact
that ESC can take many months to appear in service. For this reason, the use of critical strain values derived from such short-term
bending tests can be open to criticism. With most of the bend tests, they involve a constant applied strain, and so stress relaxation
occurs during the test. Because of this, the driving force for craze formation drops, often quite rapidly, and so the likelihood of
craze formation decreases fairly rapidly. In this case, the use of a relatively short testing time is more justified.
A review of bend methods conducted by Arnold (1994) showed additional effects of stress relaxation. If crazes form, they have
the potential to open up and relax the strain still further. In this case, the potential for further crazing is greatly reduced. This has
implications for the determination of critical strains from bend tests. If the critical strain is determined from the length of the
crazed zone, this can give artificially high values as the strain has relaxed due to adjacent crazes. In the most severe case, a single
craze in the center of a bend test can open up to such an extent as to act as a hinge, reducing the strain throughout the sample to
zero. This throws doubt on the validity of critical strain values determined from elliptical jigs and from craze width measurements
with three point bend tests. It was found that a better (but longer-winded) method was to use a range of bend tests, with different
maximum strains, and determine the highest strain at which no crazes were seen.
 Environmental Effects on Crack Growth in Polymers 265

Fig. 15 A three-point bend tester commonly used for ESC studies.

Creep tests
The use of creep tests, where a constant (normally tensile) stress is applied to a sample involves more complex equipment, but can
give more detailed information. The principle is quite simple, with a sample (strip or dog-bone) held under a constant tensile force
whilst immersed in the environment. With some versions of this test method, the whole sample and the grips are immersed; with
others, a liquid-containing chamber is fixed around part of the sample. In either case, there are complications of fluid sealing to be
overcome.
The tests can be run simply to give a time-to-failure under a certain stress, or with the use of extensometers or displacement
transducers, the strain can be monitored continuously. The measurement of just time-to-failure only gives information on the total
time for both craze initiation and growth, whereas the use of strain measurement devices can give useful information on craze
initiation. Samples can be observed continuously for visible signs of crazing, which can be done remotely by the use of optical
systems which detect reflected light from craze surfaces. A more convenient method can be to use the creep strain and the concept
of departure points. The rationale behind these is that, when crazes form, they accommodate extra strain and so the strain– time
curve rapidly increases upwards compared to one measured in air. Such a departure point is shown in Fig. 16. It has been found
that the departure points correspond fairly well with the development of optically visible crazes (Wright and Gotham, 1983).
The main advantage with this method is that the stress and strain conditions are uniform throughout the material and once
crazes have formed, they tend to accelerate. The main disadvantage, in addition to more complex equipment, is that the time-to-
crazing is strongly dependent on the applied stress, and so considerable testing needs to be undertaken initially to determine the
optimum stress conditions to give sensible testing times.

Constant-strain-rate and constant-stress-rate tests

One way to overcome the issue of initially determining the required stress level in a creep test is to use a continuously increasing
strain or stress. Constant- (slow-) strain-rate tests are basically conventional tensile tests performed with the sample immersed
(either fully or partly) in the liquid environment. Again, sealing of the fluid is an issue, and measurement of the strain can be
complicated by the presence of the environment. The main advantage is that the test ensures failure within a reasonable time,
depending on the strain rate used. This test has been used widely during the 1990s (Arnold, 1995c, 1998a, 1998b).
A modification to this type of test is to use a constantly increasing stress, which is performed in a similar manner, except that
the load is controlled. The increasing stress can either be achieved by use of controlled servohydraulic test equipment or by the use
of an increasing load on a creep machine using a pumped-water method (Wright, 1996). Although such an equipment is normally
less available than the standard tensile test equipment, the main advantage is that the constant-stress-rate method gives an
accelerating craze once it has formed and so the formation of crazes can be seen more dramatically.
Departure points can be seen for both of the above tests, in addition to constant-stress (and constant-strain) tests. A com-
parison of these methods was performed by Arnold (1998a), with the departure point behavior shown in Fig. 16 for constant-
stress and Fig. 17 for constant-strain rates. From these it can be seen that, with constant stress, the environment curve departs
upwards, whereas with the constant-strain rate (and constant-stress rate) tests, departure is downwards. It was found that the stress
rate and strain rate give departures in a more controlled time, and the departure point is most obvious for the stressrate test, and
least obvious for the constant-strain test, due to stress relaxation.
266 Environmental Effects on Crack Growth in Polymers

Fig. 16 The creep strain against time, for a creep test of polycarbonate in air and propanol, showing a departure point due to craze formation.

Fig. 17 The stress against strain for a constant-strain-rate test of polycarbonate in air and propanol, showing a departure point due to craze
formation.

Fracture toughness tests

All of the methods described above relate to un-notched samples where craze initiation was investigated. Many other studies have
concentrated on craze growth using pre-notched samples (Williams and Marshall, 1975; Tonyali and Brown, 1986; Brostow et al.,
1991; Truong et al., 1987; Chan and Williams, 1983; Qian et al., 1993; Mai, 1975). For most of these, standard sample geometries
have been used including single-edge, center-edge, and doubleedge notched samples, compact tension samples and notched bend
samples. The stress application methods have included constantload and constant-displacement rate, often coupled with crack-
growth and crack-opening displacement (COD) measurements. The use of fracture mechanics in the study of ESC is discussed
later, but the main complexities of testing include measurement of cracks when samples are immersed in fluids and ensuring the
initial notch/crack is representative.
 Environmental Effects on Crack Growth in Polymers 267

Fig. 18 A schematic of the miniature-disk bend test equipment used for ESC studies.

Biaxial tests
Although the vast majority of ESC testing has taken place with tensile loading, some biaxial testing has taken place using a variety
of methods including pressurized plates and torsion of cylinders. The results of these studies will be discussed later, but one
additional method is worth discussing here. This is the miniature-disk bend test, which involves clamping a small circular disk in a
jig and indenting it with a semicircular punch at a constant displacement rate, as shown in Fig. 18. Although it is difficult to
convert the force– displacement behavior directly to stress and strain due to the complex loading geometry, it does give a good
correlation to standard tests. It is, therefore, good as a comparative test method with the distinct advantages that it uses small
samples and very small amounts of the liquid environment. It is a very easy test to conduct in a liquid environment and it tends to
accentuate any ductile-to-brittle transition (Arnold and Keeble, 1998).

ESC Craze Initiation 1: Stress/Strain/Time Criteria

In order to be able to predict the ESC process, there are three main hurdles to overcome. The first involves predicting under what
conditions of applied stress and strain crazes will initiate, and how long this will take. The second task, discussed in the next
section, is to determine under what conditions crazes will initiate in different polymer/environment combinations, and the final
task is to predict the subsequent growth and breakdown of the crazes, via a fracture mechanics approach.
It is well known that the formation of ESC crazes does not follow any simple pattern. There is normally an incubation time
before crazes are observed, which increases as the applied stress or strain decreases. The strain at which crazes form decreases as the
applied stress increases, and there is some evidence that this strain reaches an asymptotic minimum value. Much of the evidence
for a minimum critical value of strain to crazing comes from the use of constant-strain bend tests. Under these conditions, stress
relaxation (especially if crazes form) causes the driving force for crazing to decrease and so craze formation rapidly decelerates. This
gives the appearance of the crazing process having stopped below a certain strain. It is now evident from other test methods that
crazing development continues at a decreasing rate, and the existence of a minimum strain is questionable.
In an attempt to determine the controlling parameter for air crazing, Matsuo (1966) formed crazes around a steel ball
embedded in polystyrene. He found that crazes formed at an angle of 37.21, which was closest to the position expected if the
tensile strain was the most important. It has subsequently been found that the dilatational component certainly plays a part, with
high dilatational strains promoting crazing and hydrostatic pressures suppressing them (Moet et al., 1979). Sternstein and Ongchin
(1969) proposed that crazes form when
B
s1
s2 ¼ A þ
s1 þ s2
where s1 and s2 are the principal stresses, and A and B are time and temperature-dependent constants, respectively. The two main
difficulties with this criterion are that the “constants” A and B vary with both temperature and time, and that the “stress bias,” s1– s2 is
rather ill defined for different stress states.
Various investigations into the effects of multiaxial stress states have been carried out by Kawagoe and co-workers. Using
combinations of tension, compression, and torsion on cylindrical samples, they mapped out the craze locus for PMMA in both air
and in kerosene (Kawagoe and Kitagawa, 1981). The locus obtained in air is shown in Fig. 19, which indicates that crazing occurs
only when there is a dilatational component and only shear yielding is observed when the stress state is pure shear (torsion).
Interestingly, when similar tests were done in a kerosene environment, they found that crazing was observed in pure shear, and
even when the dilatational stress was slightly negative, as shown in Fig. 20. Further studies of this data indicated that the localized
stress at surface scratches was important (Kawagoe and Kitagawa, 1988). In further studies, they found that crazes were also
observed in PMMA in pure torsion in a range of alkanes (Kawagoe and Kitagawa, 1987) and alcohols (Kawagoe and Morita,
268 Environmental Effects on Crack Growth in Polymers

Fig. 19 The type of failure plotted against stress state for polystyrene tested in air. (reproduced with permission from Kawagoe and Kitagawa,
1981).

Fig. 20 The type of failure plotted against stress state for polystyrene tested in kerosene. (reproduced with permission from Kawagoe and
Kitagawa, 1987).

1994). The other interesting feature is that, in these cases, the normal correlation with the solubility parameter seemed to be
reversed. It was suggested that the reason for the crazing in pure shear was due to local dilatation caused by liquid diffusion at
stress concentrators. If this were the case, a correlation with diffusion rate, rather than solubility, may well be expected.
The vast majority of experimental studies have focused on tensile stresses, as they are the most severe for ESC formation. There
is still a need for a workable criterion to determine the stress/strain/time conditions for crazing. Various criteria have been
suggested, including a critical total strain, critical relaxed strain, and a critical relaxed strain energy density. A comparison of these
criteria was performed by Wright and Gotham (1983), who presented conclusive evidence that the relaxed strain was the critical
parameter for crazing. The concept of relaxed (or inelastic) strain is given as the total strain minus the short-term strain and can be
seen to be the strain due to slower (main-chain) relaxations. This criterion has received further experimental support for poly-
carbonate in ethanol (Arnold, 1995b) and in propanol (Arnold, 1998a). Fig. 21 shows the results in propanol, where the stress and
strain at departure for a range of test methods are plotted. Also plotted is the short-term modulus for the material, from which it
can be seen that departure occurs when the inelastic strain equals B0.25%. The effects of temperature were also studied (Arnold,
1998a), from which it was found that the criterion still holds, except that the value of critical relaxed strain decreases with
increasing temperature. Similar results were obtained for PC in methanol, PVC in dekalin and ethylene glycol, and polystyrene in
propanol and ethylene glycol. The criterion did break down either when test speeds were sufficiently high to prevent environ-
mental effects or were so slow that diffusional effects took over.
The concept of a critical relaxed strain can be envisaged with recourse to the accepted mechanism of craze development. After
the short-term “elastic” response to a stress, main-chain motion will occur, leading to microvoid formation. When these voids have
developed to a certain degree (with associated strain accumulation), crazing will become possible. More severe ESC agents will
enable craze formation at earlier stages during this process and so will yield a lower value of critical strain. An increase in
temperature can accelerate the process in manner similar to an increase in severity of ESC agent. Experimental evidence to support
 Environmental Effects on Crack Growth in Polymers 269

Fig. 21 The stress and strain at departure for polycarbonate in propanol, plotted alongside the short-term modulus.

Fig. 22 The departure time plotted against the post-immersion time for polycarbonate in ethanol.

this idea has come in the form of postimmersion tests (Arnold, 1995b, 1998a), where liquid contact commenced at various times
after application of stress in creep tests. For many of the combinations of materials studied, it was found that departure occurred,
as discussed above, and that it occurred at the same time, regardless of when the environment was introduced. For example, if a
departure time of 105 s was observed with application of stress and environment together, the departure time (after stress
application) was the same if the environment was added after 103 s, 104 s, or 8  104 s, as seen in Fig. 22. When the environment
was added after the normal departure time, departure occurred almost immediately.
270 Environmental Effects on Crack Growth in Polymers

Fig. 23 The crazing stress plotted against the interaction parameter χ, for polycarbonate in a range of fluids. (reproduced with permission from
Mai, 1986).

This effect has considerable importance. First, it supports the critical relaxed strain criterion, but with the important fact that the
environment seems to have no effect on the crazing process until the critical relaxed strain has been reached, at which point the
environment has a rapid effect. This suggests that a void network has to be established to a certain level before the environment
can play a part. Second, it has important practical implications. For instance, if a combination of material, environment and stress
gave a departure time of two years, a component under this stress without environmental contact for two years would then rapidly
craze if a brief environmental contact occurred. Testing involving immersion at the start of a test would not produce such crazing
prior to two years. Isolated cases of this type have been documented (Wright, 2001) and many others may well have occurred.

ESC Craze Initiation 2: Polymer/Fluid Interaction Criteria

The second task in predicting ESC failure is to determine the severity of particular environment/polymer combinations. Given the
enormous number of possible combinations, comprehensive testing is impossible and so predictive capability is needed. Much of
the work in this area follows the correlation with the solubility parameter noted by Bernier and Kambour (1968). Further
experimental work has shown that this correlation is not perfect by any means, and this led Vincent and Raha (1972) and Henry
(1974) to incorporate a hydrogen bonding parameter. This gave an improved correlation for a range of combinations, and led to
the inclusion of the full 3D Hansen parameters discussed above by Jacques and Wyzgoski (1979). This approach was used to
develop 3D maps of ESC severity against dispersive, polar, and hydrogen bonding parameters. This gave improved correlations,
but, due to the uncertainty in the partial parameters for polymers, it does not represent a true predictive capability.
The solubility-parameter approach was taken further by Mai (1986), who analyzed a series of experimental results in terms of
Flory– Huggins interaction theory. The analysis used by Mai incorporated the effects of molar volume and the stress on solubility.
He calculated the interaction parameter χ and plotted this against the stress required to cause crazing. The correlations that were
obtained were very good, as shown for polycarbonate in Fig. 23. The major difficulty with this analysis is that the interaction
parameter depends on the local stress at the tip of a craze, which depends on the stress required to cause crazing, which is the value
to be predicted. Mai’s analysis used the experimentally measured values of craze stress to show that the correlation between χ and
crazing stress is very good; however, the prediction was left with an intractable circular methodology. Further modeling by Breen
(1993) used a combination of interaction parameters and surface-energy effects to predict crazing in a series of hydrocarbon
liquids and vapors, but the effects of stress on solubility were not included. In order to develop the method of Mai, some way of
predicting the local stress at the tip of a flaw is needed and recent fracture mechanics models have been developed to address this,
which will be discussed later.
The work described above generally applies to single-liquid environments, whereas in practice many ESC agents are present as
mixtures. There is, therefore, a need to be able to extend the predictive capabilities to mixed environments. Work by Qian et al.
(1993) showed that the degree of ESC attack in a detergent/water mix varied nonlinearly with concentration. They found that a
very small concentration of detergent promoted quite severe ESC, whereas, at higher concentrations, the effects of concentration
were less marked. Surprisingly, the 100% detergent environment did not give the shortest lifetimes, an effect attributed to
softening of the whole sample due to greater plasticization.
 Environmental Effects on Crack Growth in Polymers 271

Fig. 24 The strain required for crazing plotted against the enthalpy contribution to the interaction parameter for polycarbonate in a range of
mixed environments.

A more detailed study of the effects of mixtures was performed by Arnold and Taylor (1999), using polycarbonate in mixtures
of alcohols. An initial correlation against the law-of-mixtures average solubility parameter for the environment showed (not
surprisingly) very poor correlation. An improved correlation was obtained when the enthalpy interaction parameter,
f v2 h
2 i
χH ¼ 2 ðδd1
δd2 Þ2 þ δp1
δp2 þ ðδh1
δh2 Þ2
RT
f3 v 3 h
2 i
þ ðδd1
δd3 Þ2 þ δp1
δp3 þ ðδh1
δh3 Þ2
RT
was used, where f is the volume fraction, u the molar volume, δd the dispersive solubility parameter, δp the polar parameter, and
δh the hydrogen bonding parameter. The subscripts 1, 2, and 3 refer to the polymer and the two liquids, respectively. This
correlation is shown in Fig. 24, which, considering that it neglects liquid–liquid interactions, is quite promising.

ESC Craze Growth

Numerous studies of ESC craze growth have been undertaken with a fracture mechanics basis. Many of these have been concerned
with slow, stable crack growth in polyethylenes. Indeed, the use of an ESC agent has been one way to accelerate the brittle fracture
of PEs to provide comparative predictions of long-term behavior in air. All of the craze growth studies use pre-notched samples,
where there is the obvious concern regarding the initial craze structure. Once a certain amount of environmental craze growth has
occurred, the craze structure seems to stabilize.
Early studies into ESC growth used a variety of fracture mechanics tools. Andrews and Bevan (1972) found that the surface
work to fracture correlated quite well with solubility parameters. A more extensive study by Williams and Marshall (1975) used
measurements of crack speed, da/dt vs. applied stress intensity factor, KI. They found that there were two distinct regimes of
behavior for a range of materials. At low applied-stress intensities, the process was relaxation controlled and the crack speed varied
with stress intensity according to
 
da
pKI10
dt
At higher crack speeds, the relationship changed to one where the craze growth was hindered and controlled by the speed of
fluid flow to the craze tip. Under these conditions, the relationship changed to
 
da
pKI2
dt
They also proposed that there was a critical COD that decreased with increasing severity of environment. This proposal received
support from Truong et al. (1987), who found that the critical COD of PMMA in methanol increased as the molecular weight
increased, i.e., as the craze resistance increased.
Many subsequent fracture mechanics studies of ESC growth have used the stress-intensity factor approach, with some success.
Mai (1975) studied the craze growth in PMMA in various liquid environments. He found that the crack speed varied with applied
stress intensity in a manner similar to that described above. At low testing rates, the liquid was able to keep up with the growing
craze and fast craze growth (and low toughness) was obtained. At higher testing rates, the environment lagged behind somewhat,
272 Environmental Effects on Crack Growth in Polymers

Fig. 25 The crack growth rate, da/dt, plotted against the stress intensity factor, KI, for polyethylenes in an Igepal(TM) detergent solution.
(reproduced with permission from Tonyali and Brown, 1986).

giving higher values of fracture toughness. The surprising feature of this work was that, at the highest speeds used, the fracture
toughness increased above that seen in air, i.e., the environment provided some toughening. This was attributed to the extra
surface work involved in developing a craze in an environment. The exact reasons for this effect are rather unclear, but it seems
that, at low testing rates, the environment not only keeps up with the crazing but also diffuses ahead of the craze, decreasing the
toughness by a classic ESC mechanism. At higher testing rates, although the environment can still keep up with the crazing, it
cannot diffuse as much into the surrounding material. The environment then provides additional resistance to craze growth,
possibly due to very localized swelling compression. It would be expected that at even higher testing rates, the toughening effect of
the environment would be lost due to the environment lagging well behind the craze front.
Chan and Williams (1983) also found a two-stage behavior with polyethylenes in detergents. They found that, at low testing
rates, with low KI, the process was determined by diffusion-controlled void growth. At higher testing rates, toughening occurred,
attributed to craze blunting. This can be understood in terms of environment availability at the craze tip. At low testing rates, the
environment is always present. At higher rates, the environment lags behind and the toughness observed in the absence of an
active environment begins to recover. This phenomenon was further investigated by Tonyali and Brown (1986), who found a
three-stage process in craze growth in PE/detergent tests, as shown in Fig. 25. At low applied-stress intensities, the craze growth was
controlled by relaxation processes as the environment was fully present at the craze tip. This gave low vales of craze/crack speed,
which increased with KI. At higher KI, the crack growth rate increased, eventually becoming constant indicating that the crack speed
was independent of KI. This was attributed to a crack growth controlled by fluid flow into the crack. This flow occurs independently
of the stress intensity, giving the constant crack speed observed. With the highest applied KI, corresponding to the longest cracks,
the crack speed decreased dramatically, which was attributed to crack blunting by normal plastic yielding, with no environmental
effect. The mechanism for this is presumably one where the environment loses contact with the rapidly growing crack (at lower KI/
crack lengths) which allows yielding and blunting, which then prevents craze re-formation when the crack has slowed.
The effects of stress intensity clearly vary with material and environment, depending on whether partial environmental contact
leads to higher or lower toughness. For instance with PMMA in water (Josserand et al., 1995), more extensive environmental
contact leads to increased toughness due to craze blunting from the rapidly diffusing water. As the crack speed increases, the effects
of blunting decrease and so the fracture toughness decreases.
In order to develop a full fracture mechanics model of ESC growth, a model of the stress distribution along the craze is needed.
The simplest one is the Dugdale model, which assumes a constant stress along the length of a craze. Although this has been shown
to be a simplification, it is used frequently (Breen 1995). Krenz et al. (1976) showed that the actual stress profile differs quite
markedly in different environments, especially with severe ESC agents such as methanol. A complex model of the craze stress
profile based on the meniscus instability model was developed by Brown (1989). Although this is quite comprehensive, it is
perhaps defeated by its own complexity, as there are many craze-geometry factors that need to be established in order to provide a
prediction of stress profile.
A slightly more usable predictive model of craze growth was developed by Brostow et al. (1991). They used relaxation-
controlled craze growth and a slightly simplified craze geometry to provide the following predictive formula between crack speed,
da/dt, and stress intensity, KI:
  
1
 1 1 da
logKI ¼ log a2 2GE þ log 1 þ
2 2 bacr dt
where a is the geometric factor appropriate to the crack geometry, G is the surface energy (between polymer and fluid), E is the
Young modulus, b is a time-independent constant, acr is the critical crack length given by
 Environmental Effects on Crack Growth in Polymers 273

Fig. 26 The stress/strain behavior of PMMA tested in air and in methanol after various pre-immersion times.

2GE
acr ¼
ps2

This model showed good agreement with experimental evidence for polyethylenes, but has yet to be validated with other
materials.
Further modeling, attempting to include the thermodynamics of fluid/polymer interactions have been undertaken by Okamoto
and Ohde (1986) and then extended by Kotoul (1996). Kotoul’s model is certainly the most ambitious one published as of early
2000s, involving a combination of irreversible thermodynamics and fracture mechanics. However, the model involves several
factors that can only be determined iteratively, and it has yet to be verified experimentally.

The Effects of Liquid Absorption

The limiting effects of liquid absorption on craze growth mechanisms have been described above, but the effects of diffusion and
absorption also seem to play a part in craze initiation. These effects have been studied for a range of combinations using slow
strain rate, creep and bend test methods (Arnold, 1995c, 1998a, 1998b; Arnold et al., 1996). The amount of liquid diffusion was
varied by using different test rates and by immersion of the samples in the environments prior to testing. A variety of effects were
observed, depending on the material/environment combination and on the time for diffusion.
At the fastest test rates, or with slow-diffusing environments, it might be expected that the environment has little effect as it
cannot keep up with the growing craze. Such behavior, common with notched samples, was not observed with un-notched
samples. Instead, conventional ESC was observed under these conditions, with a reduction in stress and strain to crazing.
As the amount of diffusion prior to crazing increased (slower testing or pre-immersion), it is common for the stress required for
craze initiation to increase. This is due to one of two effects, which are hard to separate: Very localized swelling at the surface or at
stress concentrations can lead to compressive stresses which hinder craze formation. This seems to be the dominant mechanism
with large but chemically compatible environments with high solubility but very low diffusion rates. Alternatively, the environ-
ment can diffuse slightly ahead of a stress concentration and produce flaw blunting. This seems to be more dominant with smaller,
faster diffusing species. In some cases, these effects can provide so much toughening that the failure stress increases not only above
that observed in the environment at faster rates, but also above that in air. This overall toughening has been observed with very
brittle cross-linked acrylics (Arnold et al., 1996) and phenolics (Arnold, 1995a, 1995b, 1995c).
As the amount of diffusion increases, the environment is able to further plasticize the material and bulk softening takes place.
Then a transition to failure occurs, more dominated by shear yielding (which can occur in a highly localized band similar to a
craze). This causes a reduction in the failure stress, but a dramatic increase in failure strain. All of these effects are seen in Fig. 26,
where slow-strain-rate tests on PMMA in methanol were conducted at the same strain rate but with different pre-immersion times.
A dramatic reduction in strength, compared to air, is observed for the shortest pre-immersion. As the pre-immersion time
increases, toughening due to compressive stresses and flaw blunting takes place, leading to an increase in the observed strength.
This reaches a peak with pre-immersion of between 30 min and 90 min, after which softening starts to become important. Similar
effects are observed over longer timescales with slower diffusing environments.
274 Environmental Effects on Crack Growth in Polymers

Fig. 27 The stress/strain behavior of a silicone rubber tested in air and high-pressure CO2.

Environmental Effects of Crack Growth In Rubbers

The effects of environment on crack (or tear) growth in rubbers is treated separately here, although many of the mechanisms of chemical
degradation and ESC can be seen to some extent with rubbers. It is a well-known fact that chemical attack of the polymer chains can
significantly weaken rubber, and the rate of attack is often increased at very high strains. For instance, oxidative attack increases the fatigue
crack growth rate of natural rubber, an effect which is eliminated with high-frequency testing or at rapid crack growth rates (Lake, 1972).
One of the most severe chemical environments is ozone, especially for unsaturated rubbers. The ozone (O3) molecule rapidly
attacks the double bond, and then leads to rapid chain scission, as shown schematically below:
Although levels of ozone are normally low, ozone cracking is observed in fatigue behavior of many rubbers. At very low tearing
energies (equivalent to stress intensities), the fatigue crack growth rate reaches a constant limiting low value of B3  10–9 cm/
cycle, which is due to ozone cracking (Lake and Lindley, 1965). Obviously, in these situations, the degradation due to ozone is not
influenced by the low stresses involved.
An equivalent to ESC failure has also been reported widely in rubbers due to purely physical interactions with swelling fluids.
As rubber has never been shown to form crazes, the mechanisms must be slightly different, but the basic principle is the same. The
absorbed fluid reduces the interchain bonding, thereby increasing molecular mobility and allowing easier fracture. The exact
mechanism is thought to be due to two reasons.

(i) The plasticizing fluid allows easier chain motion, which results in the shortest chain segments becoming strained to their limit
sooner than would otherwise happen. Essentially, this implies that more of the stress is carried by these fully strained segments and
less by other entangled segments. This leads to earlier fracture of the strained segments and consequently earlier fracture.
(ii) The plasticizing fluid reduces the bonding between rubber chains and filler particles, thereby again transferring more of the
stress to the fully strained chain segments.

Evidence for this type of failure has come from studies of various rubbers in a range of hydrocarbon liquids (Chiu and Mark,
1977; Thomas, 1981; Abu-Isa, 1983; Prygiel et al., 1991) and also in high-pressure CO2 (Davies et al., 1999). Tensile tests in high-
pressure CO2 showed considerable reduction in strength compared to tests in ambient air and in high-pressure N2. Typical results
are shown in Fig. 27. These results not only showed a dramatic reduction in strength, but also a lower initial modulus, thought to
be due to the reduction in polymer–filler interactions. The fracture surfaces observed in CO2 were also considerably different from
those in air, with a much smoother surface, indicative of a plasticization-enhanced failure.

References

Abbas, K.B., Sorvik, E.M., 1973. On the thermal deregulation of poly(vinyl chloride): II. The effect of atmosphere. J. Appl. Polym. Sci 17, 3577–3594.
Abdul-Hafis, S.A., 1997. Synthesis and characterisation of hypochlorite-oxidised starch composites. Polym. Deg. Stab 55, 9–17.
Abu-Isa, I.A., 1983. Elastomer gasoline blends interactions: 1. Effects of methanol-gasoline mixtures on elastomers. Rubb. Chem. Technol 56, 135–168.
Andrews, E.H., Bevan, L., 1972. Mechanics and mechanism of environmental crazing in a polymer glass. Polymer 13, 337–345.
Andrews, E.H., Reed, P.E., 1973. Molecular fracture. In: Kausch, H.H., Hassell, J.A., Jaffe, R.I. (Eds.), “Deformation and Fracture of High Polymers”. New York: Plenum, p. 259.
 Environmental Effects on Crack Growth in Polymers 275

Argon, A.S., 1973. Physical basis of distortional and dilatational plastic flow in glassy polymers. J. Macromol. Sci B8, 573–596.
Argon, A.S., Bessonov, M.I., 1977. Plastic flow in glassy polymers. Polym. Eng. Sci 17, 174–182.
Argon, A.S., Salama, M., 1976. The mechanism of fracture in glassy materials capable of some inelastic deformation. Mater. Sci. Eng 23, 219–230.
Arnold, J.C., 1993. The influence of physical aging on the creep rupture behavior of polystyrene. J. Polym. Sci. B: Polym. Phys 31, 1451–1458.
Arnold, J.C., 1994. The use of flexural tests in the study of ESC of polymers. Polym. Eng. Sci 34, 665–670.
Arnold, J.C., 1995a. The effects of physical aging on the brittle fracture behavior of polymers. Polym. Eng. Sci 35, 165–169.
Arnold, J.C., 1995b. Craze initiation during the environmental stress cracking of polymers. J. Mater. Sci 30, 655–660.
Arnold, J.C., 1995c. The influence of liquid uptake on ESC of glassy polymers. Mater. Sci. Eng A197, 119–124.
Arnold, J.C., 1998a. Craze initiation criteria during ESC of amorphous polymers. Plast. Rubb. Comp. Proc. Appl 27, 139–147.
Arnold, J.C., 1998b. The effects of diffusion on ESC initiation in PMMA. J. Mater. Sci 33, 5193–5204.
Arnold, J.C., Keeble, J.M., 1998. The use of miniature disk bend tests with plastic materials. Polym. Test 17, 597–611.
Arnold, J.C., Li, J., Isaac, D.H., 1996. The effects of pre-immersion in hostile environments on the ESC behavior of urethane-acrylic polymers. J. Mat. Proc. Technol 56,
126–135.
Arnold, J.C., Taylor, J.E., 1999. Improved thermodynamic approach for predicting the ESC behavior of PC in binary liquid mixtures. J. Appl. Polym. Sci 71, 2155–2161.
Bair, H.E., Falcowe, D.R., Johnson, G.E., Hellman, M.Y., 1981. Hydrolysis of PC to yield BPA. J. Appl. Polym. Sci 26, 1777–1786.
Barton, A.F.M., 1983. “CRC handbook of Solubility Parameters and other Cohesion Parameters,”. FL: CRC Press.
Bauwens-Crowet, C., Bauwens, J.C., 1982. Annealing of polycarbonate below the glass transition. Polymer 23, 1599–1604.
Berger, L.L., 1989. Solvent induced embrittlement in crystallizable polyarylate. J. Polym. Sci. B: Polym. Phys 27, 1629–1647.
Berger, L.L., Kramer, E.J., 1988. The effect of temperature on the transition from crazing to shear deformation in cross-linked PS. J. Mater. Sci 23, 2526–2543.
Bernier, G.A., Kambour, R.P., 1968. The role of organic agents in the stress cracking and crazing of PPO. Macromolecules 5, 393–400.
Bilby, B.A., Eshelby, J.D., 1972. Dislocations and the theory of fracture. In: Liebowitz, H. (Ed.), “Fracture”. vol. 1. New York: Academic Press, p. 111.
Breen, J., 1993. ESC of PVC and PVC-CPE: Part I. Crazing. J. Mater. Sci 25, 3769–3775.
Breen, J., 1994. ESC of PVC and PVC-CPE: Part II. Failure mechanisms. J. Mater. Sci 29, 3536–3543.
Breen, J., 1995. ESC of PVC and PVC-CPE: Part III. Crack growth. J. Mater. Sci 30, 5833–5840.
Breen, J., Van Dijk, D.J., 1991. Environmental stress cracking of PVC: effects of natural gas with different amounts of benzene. J. Mater. Sci 26, 5212–5220.
Brostow, W., Fleissner, M., Muller, W.F., 1991. Slow crack propagation in polyethylene: determination and prediction. Polymer 32, 419–425.
Broutman, L.J., Krishnakumar, S.M., 1976. Impact strength of polymers: 1. The effect of thermal treatment and residual stress. Polym. Eng. Sci 16, 74–81.
Brown, H.R., 1989. A model of environmental craze growth in polymers. J. Polym. Sci. B: Polym. Phys 27, 1273–1288.
Brown, H.R., Kramer, E.J., 1981. Craze microstructure from small angle X-ray scattering. J. Macromol. Sci. Phys B19, 487–522.
Brown, H.R., Njoku, N.G., 1986. The effect of plasticisation on craze microstructure. J. Polym. Sci. B: Polym. Phys 24, 11–18.
Chan, M.K.V., Williams, J.G., 1983. Slow stable crack growth in high density PE. Polymer 24, 234–244.
Chiu, D.S., Mark, J.E., 1977. Effect of swelling on the stress strain isotherms and ultimate properties of PDMS networks in elongation. Polym. Preprints 18, 618–622.
Cowie, J.M.G., McEwen, I.J., 1987. Polymer cosolvent systems XI. J. Polym. Sci. B: Phys 25, 1501–1509.
Davies, O.M., Arnold, J.C., Sulley, S., 1999. The mechanical properties of elastomers in high pressure CO2. J. Mater. Sci 34, 417–422.
Doll, W., 1983. Optical interference measurements and fracture mechanics analysis of crack tip craze zones. In: Kausch, H.H. (Ed.), “Advances in Polymer Science 52/53”.
Berlin: Springer, pp. 105–168.
Donald, A.M., Kramer, E.J., 1982. The competition between shear deformation and crazing in glassy polymers. J. Mater. Sci 17, 1871–1879.
Ezrin, M., Lavigne, G., 1994. Failures caused by additives and contaminants during processing. SPEANTEC 3, 3302–3305.
Fettes, E.M., 1964. “Chemical Reactions of Polymers”. New York: Interscience.
Fisher, S., Brown, N., 1973. Deformation of PTFE from 78 to 298K and the effects of environmental crazing. J. Appl. Phys 44, 4322–4327.
Flory, P.J., 1978. “Principals of Polymer Chemistry”. Ithaca, USA: Cornell University Press, pp. 541–594.
Gardner, R.J., Martin, J.R., 1979. Humid aging of plastics. J. Appl. Polym. Sci 24, 1269–1280.
Gee, R.H., 1995. Small penetrant diffusion in PB. Polymer 36, 1435–1440.
Gent, A.N., 1970. Hypothetical mechanism of crazing in glassy plastics. J. Mater. Sci 5, 925–932.
Ghorbel, I., Akele, N., Thominette, F., Spiteri, P., Verdu, J.J., 1995. Hydrolytic aging of polycarbonate: 2. Hydrolysis kinetics. J. Appl. Polym. Sci 55, 173–179.
Girifralco, L.A., Good, R.J., 1960. J. Phys. Chem 64, 561.
Girois, S., Delprat, P., 1999. Kinetic study of the photostabilisation of polypropylene films by an hydroxyphenylbenzotriazole. Polym. Deg. Stab 64, 107–114.
Grassie, N., 1979. “Developments in Polymer Degradation”. Applied Science. London.
Gugumus, F., 1995. Re-examination of the role of hydroperoxides in polyethylene and polypropylene. Polym. Deg. Stab 49, 29–50.
Gugumus, F., 1998a. Thermooxidative degradation of polyolefins in the solid state 6. Polym. Deg. Stab 62, 235–244.
Gugumus, F., 1998b. Thermooxidative degradation of polyolefins in the solid state 7. Polym. Deg. Stab 62, 245–257.
Haken, J.K., 1996. Degradative polymer analysis by chromatography. J. Chromatography A 756, 1–9.
Hansen, C.M., 1969. The universality of the solubility parameter. Ind. Eng. Chem. Prod. Res. Dev 8, 2–11.
Hedenqvist, M., 1996. Diffusion of small molecule penetrants in PE. Polymer 37, 2887–2902.
Henkee, C.S., Kramer, E.J., 1984. Crazing and shear deformation in cross-linked polystyrene. J. Polym. Sci 22, 721–737.
Henry, L.F., 1974. Prediction and evaluation of the susceptibility of glassy thermoplastics to ESC. Polym. Eng. Sci 14, 167–173.
Hildebrand, J.H., Scott, R.L., 1950. “The Solubility of Non-electrolytes”. Princeton, NJ: Van-Nostrand.
Hill, A.J., Heater, K.J., Agrawal, C.M., 1990. The effects of physical aging in polycarbonate. J. Polym. Sci. B: Polym. Phys 28, 387–405.
Hoftyzer, P.J., Van Krevelen, D.W., 1976. “Properties of Polymers,”, 2nd edn. Amsterdam: Elsevier, pp. 152–154.
Hoy, K.L., 1989. Tables of solubility parameters. J. Coated Fabrics 19, 53–71.
Igarashi, M., DeVries, K.L., 1983a. An electron spin resonance investigation of environmental effects on polymeric materials: 1. Nylon. Polymer 24, 769–782.
Igarashi, M., DeVries, K.L., 1983b. An electron spin resonance investigation of environmental effects on polymeric materials: 2. Polyethylene. Polymer 24, 1035–1041.
Imai, Y., Brown, N., 1976. Environmental crazing and intrinsic tensile deformation in PMMA. J. Mater. Sci 11, 425–433.
Jacques, C.H.M., Wyzgoski, M.G., 1979. Prediction of ESC from solubility considerations. J. Appl. Polym. Sci 23, 1153–1166.
Jellinek, H.H.G., Flajsman, F.F., 1970. Chain scission of butyl rubber by nitrogen dioxide in the absence and presence of air. J. Polym. Sci. A-1 8 (3), 711–726.
Jellinek, H.H.G., Flajsman, F.F., Kryman, F.J., 1969. Reaction of sulfur dioxide and nitrogen dioxide with polymers. J. Appl. Polym. Sci 13, 107–116.
Jellinek, H.H.G., Toyoshima, Y., 1967. Reaction of nitrogen dioxide with polystyrene films. J. Polym. Sci. A-1 5, 3214–3218.
Josserand, L., Schirrer, R., Davies, P., 1995. Influence of water on crack propagation in PMMA: craze stress and craze fibril lifetime. J. Mater. Sci 30, 1772–1780.
Kaczmarek, H., Linden, L.A., Rabek, J.F., 1995. Photoxidative degradation of PPO. Polym. Deg. Stab 47, 175–188.
Kambour, R.P., 1977. Mechanistic aspects of crazing and cracking of polymers in aggressive environments. In: Swann, P.R., Ford, F.P., Westwood, A.R.C. (Eds.), “Mechanisms
of Environment Sensitive Cracking of Materials”. London: Metal Society, pp. 213–225.
Kambour, R.P., Gruner, G.L., Romagosa, C.E., 1974. Bisphenol-A polycarbonate immersed in organic media: swelling and response to stress. Macromolecules 7, 248–253.
Kausch, H.H., 1986. Intersegmental interactions and chain scission. In: Brostow, W., Corneliussen, R.D. (Eds.), “Failure of Plastics”. Munich: Hanser, pp. 84–97.
276 Environmental Effects on Crack Growth in Polymers

Kausch, H.H., 1987. “Polymer Fracture”. New York: Springer, pp. 305–402.
Kawagoe, M., Kitagawa, M., 1981. Craze initiation in PMMA under biaxial stress. J. Polym. Sci 19, 1423–1433.
Kawagoe, M., Kitagawa, M., 1987. Craze initiation in PMMA exposed to n-alkanes. J. Mater. Sci 22, 3000–3004.
Kawagoe, M., Kitagawa, M., 1988. On criteria for craze initiation in glassy polymers. J. Mater. Sci 23, 3927–3932.
Kawagoe, M., Morita, M., 1994. Fracture mechanisms of PMMA under static torsion in alcohol environments. J. Mater. Sci 29, 6041–6046.
Kawagoe, M., Nunomoto, S., 1991. Crazing and fatigue of PMMA due to case II diffusion of methanol. Polymer 32, 3130–3135.
Kefalas, V.A., 1995. Solvent crazing as a stress-induced surface absorption and bulk plasticisation effect. J. Appl. Polym. Sci 58, 711–717.
Kotoul, M., 1996. Micromechanical and thermodynamical aspects of environmental crazing. J. Mater. Sci 31, 3333–3349.
Kramer, E.J., 1979. Environmental Cracking of Polymers. In: Andrews, E.H. (Ed.), “Developments in Polymer Fracture”. London: Applied Science, pp. 55–120.
Kramer, E.J., 1983. Microscopic and molecular fundamentals of crazing. In: Kausch, H.H. (Ed.), “Advances in Polymer Science 52/53”. Berlin: Springer, pp. 1–56.
Krenz, H.G., Kramer, E.J., Ast, D.G., 1976. The structure of solvent crazes in polystyrene. J. Mater. Sci 11, 2211–2221.
Lake, G.J., 1972. Mechanical fatigue of rubber. Rubber Chem. Technol 45, 309–321.
Lake, G.J., Lindley, P.B., 1965. The mechanical fatigue limit for rubber. J. Appl. Polym. Sci 9, 1233–1241.
Lauterwasser, B.D., Kramer, E.J., 1979. Microstructure, mechanics and fracture of crazes in polystyrene. Phil. Mag A39/4, 469–495.
Lee, A., McGarry, J., 1991. A creep apparatus to explore the quenching and aging phenomena of PVC films. J. Mater. Sci 26, 1–5.
Leonov, A.I., Brown, H.R., 1991. A model of fibril deformation in crazes. J. Polym. Sci. B: Polym. Phys 29, 197–209.
Lustiger, A., 1986. Environmental stress cracking: the phenomenon and its utility. In: Brostow, W., Corneliussen, R.D. (Eds.), “Failure of Plastics”. Munich: Hanser,
pp. 305–329.
Mai, Y.W., 1975. On the environmental fracture of PMMA. J. Mater. Sci 10, 943–954.
Mai, Y.W., 1986. ESC of glassy polymers and solubility parameters. J. Mater. Sci 21, 904–916.
Malone, S.P., Vosburgh, C., Cohen, C., 1993. Validity of the swelling method for the determination of the interaction parameter. Polymer 34, 5149–5153.
Matsuo, M., 1966. Observation of crazes in ABS-polymer and high impact polystyrene under the electron microscope. Polymer 7, 421–429.
Miller, P., Kramer, E.J., 1990. Environmental shear deformation zones and crazes in crosslinked PS and PPMS. J. Mater. Sci 25, 1751–1761.
Moet, A., Palley, I., Baer, E., 1979. Craze initiation criterion in glassy polymers under high pressure. Am. Chem. Soc. Organic Coat. Plastics 41, 424–429.
More, A.P., Donald, A.M., Henderson, A., 1992. Stress modified case II diffusion in PMMA. Polymer 33, 3759–3761.
Morris, N.Y., 1994. “Chemical Resistance of Polymers”. London: Plastics Design Library.
Okamoto, H., Ohde, Y., 1986. Environmental Stress Cracking: An Irreversible Thermodynamics Approach. In: Brostow, W., Corneliussen, R.D. (Eds.), “Failure of Plastics”.
Munich: Hanser, pp. 330–344.
Pan, J.Q., Lui, N.C., Lau, W.W.Y., 1998. Preparation and properties of new antioxidants with higher MW. Polym. Deg. Stab 62, 165–170.
Paredes, E., Fischer, E.W., 1979. Small angle x-ray investigations of craze structure in polycarbonate and PMMA. Macromol. Chem 180, 2707–2722.
Pruitt, L., Bailey, L., 1998. Factors affecting near threshold crack propagation in orthopedic UHMWPE. Polymer 39, 1545–1553.
Prygiel, A., Jelczewska, J., Slusarski, L., 1991. Interaction in elastomer solvent systems and its effect on the mechanical properties of polymers. Int. Polym. Sci. Technol 18, 84–92.
Qian, R., Lu, X., Brown, N., 1993. The effect of concentration of an ESC agent on slow crack growth in PE. Polymer 34, 4727–4731.
Robeson, L.M., Crisafulli, S.T., 1982. Microcavity formation in engineering polymers exposed to hot water. J. Appl. Polym. Sci 27, 2809–2814.
Seidmann, R., Dulog, L., 1989. Hydrogen peroxide generation from reduced polymer nitroxyls and oxygen. Macromol. Chem 190, 975–983.
Shiomi, T., Kuroki, K., Kobayashi, A., 1995. Dependence of swelling degree on solvent composition of two component copolymer network in mixed solvents. Polymer 36, 2443–2451.
Shlyapnikov, Y.A., Tyuleneva, N.K., 1997. Inhibited oxidation of polyethylene. Polym. Deg. Stab 56, 311–315.
Sohma, J., Sakaguchi, M., 1976. ESR studies on polymer radical produced by mechanical destruction and their reactivity. Adv. Polym. Sci 20, 109–158.
Clair St., T.L., Dezern, J.F., Croall, C.I., 1993. Mechanical property structure relationships of linear aromatic polyimide filmes. SPE ANTEC 1, 650–652.
Sternstein, S.S., Ongchin, L., 1969. Yield criteria for plastic deformation of glassy high polymers in general stress fields. ACS Polym. Preprint 10, 1117–1124.
Struik, L.C.E., 1978. “Physical Aging of Amorphous Polymers and Other Materials”. Amsterdam: Elsevier.
Suzuki, F., Onozato, K., Takahashi, N., 1993. Pervaporation of athermal mixtures of benzene/toluene by PET membrane and synergistic effect on the concentration dependence
of diffusion rate. J. Appl. Polym. Sci 50, 1833–1841.
Tanaka, N., 1992. Prediction of solubility parameter from the thermal transition behaviour in polymers. Polymer 33, 623–631.
Theodorou, D.N., 1996. Molecular simulations of sorption and diffusion in amorphous polymers. In: Neogi, P. (Ed.), “Diffusion in Polymers”. New York: Marcel Dekker, pp. 67–142.
Thomas, E., 1981. Will gasohol affect the rubber in your car. Mater. Eng 93, 46–50.
Thomas, N.L., Windle, A.H., 1981. Diffusion mechanics of the system PMMA—methanol. Polymer 22, 627–638.
Tonyali, K., Brown, H.R., 1986. On the applicability of linear elastic fracture mechanics to ESC of LDPE. J. Mater. Sci 21, 3116–3124.
Trent, J.S., Palley, I., Baer, E., 1981. Relationships between mechanical behavior and craze morphology in thin films of polystyrene. J. Mater. Sci 16, 331–340.
Truong, V.T., Allen, P.E.M., Williams, D.R.G., 1987. The effect of molecular properties on the mechanical behavior of PMMA III. Euro. Polym. J 23, 181–189.
Turnbull, A., Maxwell, A.S., Pillai, S., 2000. Comparative assessment of slow strain rate, 4-point bend and constant load test methods for measuring environment stress
cracking of polymers. Polym. Test 19, 117–129.
Vahdat, N., 1993. Solubility of binary liquid mixtures in polymeric materials. J. Appl. Polym. Sci 50, 1833–1841.
Valdakar, V.S., Kulkarni, M.G., Bhagwat, S.S., 1993. Anomolous sorption of binary solvents in glassy polymers. Polymer 24, 4300–4307.
Van Krevelen, D.W., 1997. “Properties of Polymers”. Amsterdam: Elsevier, pp. 129–188.
Vincent, P.I., Raha, S., 1972. Influence of hydrogen bonding on crazing and cracking of amorphous thermoplastics. Polymer 13, 283–287.
Wang, W.V., Kramer, E.J., 1982. The micromechanics and microstructure of CO2 crazes in PS. Polymer 23, 1667–1674.
Ward, A.L., Lu, X., Huang, Y., Brown, N., 1991. The mechanism of slow crack growth in polyethylene by an ESC agent. Polymer 32, 2172–2179.
Wellinghof, S., Baer, E., 1975. The mechanism of crazing in PS. J. Macromol. Sci. Phys B11 (3), 367–387.
White, J.R., Turnbull, A., 1994. Weathering of polymers. J. Mater. Sci 29, 584–613.
Williams, J.G., Marshall, G.P., 1975. Environmental crack and craze growth phenomena in polymers. Proc. Roy. Soc. London A342, 55–77.
Wolf, C.J., Fu, H., 1996. Stress-enhanced transport of toluene in PEEK. J. Polym. Sci. B: Polym. Phys 34, 75–82.
Wright, D.C., 1996. “Environmental Stress Cracking of Plastics”. Shrewsbury: Rapra Technology.
Wright, D.C., 2001. “Failure of Plastics and Rubber Products” Rapra. Technology. Shrewsbury.
Wright, D.C., Gotham, K.V., 1983. Solvent crazing criteria. Polym. Eng. Sci 23, 135–139.
Wu, S., 1982. “Polymer Interface and Adhesion”. New York: Marcel Dekker.
Wu, J.C.B., Brown, N., 1982. The tensile stress/strain behaviour of PS in inert and gaseous environments. J. Mater. Sci 17, 1311–1319.
Wyzgoski, M.G., Jacques, C.H.M., 1977. Stress cracking of plastics by gasoline and gasoline components. Polym. Eng. Sci 17, 854–860.
Yaffe, M.B., Kramer, E.J., 1981. Plasticisation effects on environmental craze microstructure. J. Mater. Sci 16, 2130–2136.
Zuchowska, D., Hlavata, D., Steller, R., Adamiak, W., Meissner, W., 1999. Physical structure of pololefin starch blends after aging. Polym. Deg. Stab 64, 339–346.