Polymer Degradation and Stability xxx (2013) 1e11

Contents lists available at SciVerse ScienceDirect

Polymer Degradation and Stability

journal homepage: www.elsevier.com/locate/polydegstab

Review article

Review of polymer oxidation and its relationship with materials

performance and lifetime prediction
Mathew C. Celina*
Sandia National Laboratories, Materials Characterization and Performance Dept. 1819, PO Box 5800, MS 1411, Albuquerque, NM 87185-1411, USA

a r t i c l e i n f o a b s t r a c t

Article history: All polymers are intrinsically susceptible to oxidation, which is the underlying process for thermally
Received 2 May 2013 driven materials degradation and of concern in various applications. There are many approaches for
Received in revised form predicting oxidative polymer degradation. Aging studies usually are meant to accelerate oxidation
12 June 2013
chemistry for predictive purposes. Kinetic models attempt to describe reaction mechanisms and derive
Accepted 14 June 2013
Available online xxx
rate constants, whereas rapid qualification tests should provide confidence for extended performance
during application, and similarly TGA tests are meant to provide rapid guidance for thermal degradation
features. What are the underlying commonalities or diverging trends and complications when we
Keywords:
Accelerated polymer aging
approach thermo-oxidative aging of polymers in such different ways? This review presents a brief status
Lifetime prediction report on the important aspects of polymer oxidation and focuses on the complexity of thermally
Thermal oxidation accelerated polymer aging phenomena. Thermal aging and lifetime prediction, the importance of DLO,
Polymer degradation property correlations, kinetic models, TGA approaches, and a framework for predictive aging models are
Arrhenius extrapolation briefly discussed. An overall perspective is provided showing the challenges associated with our un-
Diffusion limited oxidation derstanding of polymer oxidation as it relates to lifetime prediction requirements.
Published by Elsevier Ltd.

1. Introduction briefly highlight the evolving knowledge with regard to some

Long-term thermo-oxidative degradation is a concern for any
polymer under normal environmental conditions, unless sealed
systems or purposely introduced inert conditions apply. Therefore,
a focal point in the field of ‘Polymer Degradation and Stability’ is
thermally driven degradation. The other big field of course is UV
induced photo-oxidation or radiation degradation, where much
overlap with thermal aging exists when the effects of temperature
on degradation processes are considered. Understanding thermal
aging processes, oxidation mechanisms, thermal acceleration
behavior, activation energies, and the convoluted manner in which
degradation pathways may present themselves, is an integral
requirement of meaningful polymer aging studies.
The different methods that are used to evaluate thermo-
oxidation may often yield diverging perspectives. It is not always
clear if or to what degree extrapolations have predictive value or
what complications could be introduced by raising the tempera-
ture. Polymer oxidation depends on simultaneous chemical and
physical phenomena and accelerative conditions add complexity to
this relationship. This overview perspective is therefore intended to
* Tel.: þ1 505 8453551; fax: þ1 505 8449781.
E-mail address: mccelin@sandia.gov.
important aspects when we deal with thermo-oxidative degrada-
tion. This also warrants some questions about the assumptions and
expected behavior which feed into predictive aging approaches.
Some of these challenges are: Can degradation chemistry be simply
accelerated? Under which conditions do accelerated aging experi-
ments lose their predictive value? How important are diffusion
limited oxidation (DLO) complications? Can kinetic modeling pro-
vide predictive capabilities, and if so, where are the limits? What
are the ongoing issues that need to be addressed to enable this
research field to move forward?
Therefore, to address such questions we will briefly review how
thermo-oxidative processes relate to different aspects of polymer
chemistry and physics. Further, we will discuss how reaction
chemistry has been correlated with mechanical property or molec-
ular weight changes. It is also important to consider the limits of
kinetic models or the circumstances under which diffusion limited
oxidation (DLO) conditions affect accelerated aging experiments.
This opens up a discussion about non-correlations between accel-
erated and ambient aging conditions. A questionable approach, for
example, is the overexposure of polymers in fast thermo-gravimetric
analysis (TGA) degradation methods, which in the literature unfor-
tunately may be regarded as a quick substitute for more extended
polymer aging studies. We hope to provide a few comments about
other intriguing developments and a perspective on how evolving

0141-3910/$ e see front matter Published by Elsevier Ltd.

http://dx.doi.org/10.1016/j.polymdegradstab.2013.06.024

Please cite this article in press as: Celina MC, Review of polymer oxidation and its relationship with materials performance and lifetime
prediction, Polymer Degradation and Stability (2013), http://dx.doi.org/10.1016/j.polymdegradstab.2013.06.024
2 M.C. Celina / Polymer Degradation and Stability xxx (2013) 1e11

material characterization approaches could aid more sophisticated 1000

spatially resolved degradation models in the future. Slower bulk degradation at higher T

Relative shift factor aT

2. Discussion 260°C

2.1. Thermal polymer degradation and lifetime prediction 100 150°C

Mechanistic changes Linear behavior for
Temperature coupled with oxygen is the most commonly used DLO
some T range
accelerative condition for polymer degradation as it is easy to apply.
100°C
It is also an integral part of most other exposure conditions (i.e. 10
photo-oxidation), as is schematically shown in Fig. 1. Materials
which do not ‘age well’ are commonly exposed to hotter temper- 25°C
atures until they may ‘fall apart.’ The best example for this approach
Mechanistic changes
is given by ramped temperature TGA, since TGA will not ‘fail’ to
induce thermal decomposition of organic polymers. In more 1 Faster degradation at lower T
commonly used accelerated aging studies, temperature and time 1.6 1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4
are perhaps selected based on project constraints, funding avail- -3 -1
ability or even the need to have Ph.D. students publish within a 1/T [10 K ]
three year timeframe. Whatever the reason, temperature is used for
Fig. 2. Most accelerated aging studies use some convenient temperature range where
a common purpose, namely to accelerate the underlying ‘degra- linear Arrhenius behavior may be observed over a limited temperature range. How-
dation chemistry.’ But is it sensible to use accelerated chemistry ever, non-linearity may be seen when a larger temperature range is examined.
from a TGA experiment to provide guidance for long term perfor-
mance of polymers under ambient conditions? It is clear that limits
associated with mechanistic changes in dominant chemistry over a
in acceleration factors need to be considered if the resulting
large temperature range are at the core of accelerated aging and
accelerated degradation pathways are to be representative.
lifetime predictions [1,5,6,8]. It is therefore imperative that for any
There are many indications that chemical processes which
accelerated thermal aging study all relevant parallel processes be
govern polymer oxidation often follow the ‘holy grail’ of kinetics,
considered which could add complexity to the analysis of degra-
which is Arrhenius behavior, as long as moderate conditions and
dation rates and definition of mechanistic reaction schemes. Such
perhaps limited temperature intervals are applied. But significant
issues are evident when dealing with DLO behavior and surface
evidence has also been collected when larger temperature regimes
versus bulk degradation phenomena [9e14], heterogeneous versus
are examined, that many materials show curvature in an Arrhenius
homogenous aging processes [15e23], liquid state versus solid state
plot of oxidation rates [1]. Oxidation rates measured over a large
kinetic frameworks [19,24,25], or complex TGA signatures [26]; all
temperature range are an excellent indicator for the relative
which affect the interpretation of thermal polymer oxidation.
importance of oxidative processes with temperature and underlying
Thermo-oxidative aging and lifetime prediction are also intrin-
activation energies [1e7]. Textbook chemistry does not always apply
sically linked with the action mechanism and behavior of polymer
and reaction kinetics competes with other processes which induce
stabilizers or additives. Much emphasis of applied polymer testing
non-linearity. A schematic of deviations from linear Arrhenius
and aging studies relates to the effectiveness of additives or com-
behavior over a large temperature range due to mechanistic changes
binations thereof. It is not the intention of this review to go into
and DLO is shown in Fig. 2. It shows an Arrhenius plot of relative
details here, since sufficient knowledge and excellent overview
degradation rate data, which are often called shift factors when
literature in this area exists [27e31]. Further, the concerns associ-
obtained from a timeetemperature aging data superposition
ated with evaluating stabilizers when using short term high tem-
analysis and may be normalized to a particular temperature. Shift
perature or similar exposure have been discussed before [20,21,32e
factors are proportional to rate and inversely proportional to aging
34]. Correlations between high temperature and low temperature
time for a particular aging state. It is common to observe linear
action of stabilizers can be non-existent, because crossover effects
Arrhenius behavior over a limited temperature range, but non-
exist which project the opposite rating from highly accelerated
linearity may evolve when a larger temperature range is examined
tests than observed at slower aging conditions [32e34]. The chal-
due to mechanistic variations in the aging process. The challenges
lenges of truly understanding stabilizer performance and their
behavior under accelerated conditions are very similar to the ex-
• Polymers are often exposed to a multitude of environments amination of materials for lifetime prediction purposes. It is also
• Materials can display particular sensitivities important to remember that thermo-oxidative degradation will
• Temperature, oxygen, hydrolysis, UV may act in combination
proceed despite the presence of antioxidants. Obviously rates will
O2 be much lower for stabilized materials, but significant degradation
Radiation
H 2O γ, e-beam, particles can occur even when high levels of antioxidant are present [35], or
UV, vac-UV in space when oxidation induction time (OIT) data may suggest a suffi-
ciently stabilized material [36]. In such situations significant ma-
H+, OH-, ions
terial degradation at lower temperatures is not necessarily
Polymer material Temperature
inhibited until all antioxidant is consumed, which suggests the
Metal contaminants participation of localized non-inhibited reaction zones [17,24,35].
Mechanical load
Other sensitivities Stress, creep
Microbial, solvents, surfactants etc 2.2. Correlation of degradation chemistry and mechanical
properties
Major differences between oxidative and inert conditions

Fig. 1. Various exposure conditions result in polymer degradation where oxygen and Materials performance is usually associated with expected
temperature are important aging parameters. mechanical properties or similar desired attributes (elasticity,

Please cite this article in press as: Celina MC, Review of polymer oxidation and its relationship with materials performance and lifetime
prediction, Polymer Degradation and Stability (2013), http://dx.doi.org/10.1016/j.polymdegradstab.2013.06.024
 M.C. Celina / Polymer Degradation and Stability xxx (2013) 1e11 3

toughness, adhesion, impact fracture resistance etc.). Aging studies Oxidation [%]
are intended to accelerate the degradation chemistry and the
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
resulting degradation state is often monitored with supporting
400
chemical analyses, such as IR signatures, discoloration, crosslink
state, molecular weight analysis, melting behavior, or morpholog-
Superposed tensile elongation data at 80°C
ical characterization. Of course, aging trends are commonly also

Tensile elongation [%]

average oxidation rate 4*10-11 mols/g-s
monitored using tensile elongation, modulus analysis, fracture 300
50% elongation corresponds to ~ 1.8% oxidation
sensitivity or similar mechanical tests. However, in many acceler-
ated aging situations material characterization may be limited to
small specimens or it is often more convenient to consider sub- 200
gram samples for chemical diagnostics, since mechanical tests
may not always be feasible. This then requires appropriate corre-
lations between ‘oxidative chemical state’ and ‘mechanical or
physical state.’ Our understanding and description of oxidative 100
polymer degradation processes depend strongly on well-described
cross-correlations between properties measured with diverging
methodologies. 0
In addition to our discussion in Section 2.1., as part of predictive 0 50 100 150 200 250 300 350
oxidative aging studies, the value of using oxidation rate mea-
surements over large temperature regimes has been demonstrated Aging time [d]
repeatedly [1e7,37]. This is advantageous for two reasons, namely
Fig. 3. Correlation between oxidation level and tensile elongation properties for an
that steady oxidation rates can be easily extrapolated in time and HTPB elastomer aged at 80  C. Data from Ref. [2].
integrated to yield predicted oxidation levels, and that a super-
position analysis of oxidation rates and subsequent Arrhenius plot
of the underlying shift factors allows for an overview of whether conditions (i.e. very high temperatures), stress cracking and
linear or curved Arrhenius behavior may apply, and to what degree oxidative damage are severe and commonly observed [46e49].
the corresponding activation energies may differ between high and Similarly, under aggressive thermal or radiation-thermal oxidative
low temperatures. Since the use of oxidation rates allows for pre- conditions measurable chemical changes can be detected [50e52],
diction of oxidative state (degradation level) with time, and further yet numbers for exact percent-oxidation level are usually not
since DLO behavior describes and predicts oxidation levels with provided. Interestingly, under photo-oxidative conditions epoxies
depth, such approaches require knowledge of oxidation levels can show strongly heterogeneous oxidation behavior when
versus mechanical properties to be of broader value. examined by AFM [53]. Also, for photo-oxidation of a range of
Cross-correlations can be established for many polymers; thermo-set coatings, a good correlation between carbonyl growth
however, experimentally this is not always easy. In general, poly- and AFM micro-hardness was observed [54e56]. Not knowing
olefins are the most challenging materials, as they are most sensi- exact molar absorptivities and homogeneity of the degraded films,
tive to localized oxidative damage, because of their semi-crystalline it remains challenging to establish chemical oxidation levels. To
morphology and dependence on tie-molecules which are address this issue, we have recently succeeded to measure oxida-
compromised by oxidative chain scission. Little oxidation is tion rates for two epoxy systems over a large temperature range.
required to result in stress cracking of semi-crystalline polyolefins. Coupled with extended accelerated thermal oxidation experiments,
The work by Billingham et al. throughout the 1970se1980s enabled we concluded that epoxies are more reactive than anticipated [57],
much progress in the understanding of these phenomena but that they can also accommodate significant oxidation levels
[15,16,25,38e40], followed by ongoing work in the 1990s which (>10%) before any cracking may develop (so far unpublished).
dealt with the heterogeneity of oxidative aging [17,18,24,41e43]. In order to establish meaningful oxidation level to mechanical
Recent unpublished data combining chemiluminescence (CL) with property correlations, significantly aged specimens are normally
oxuptake for PP suggest on the order of 0.01% oxidation to result in
beginning embrittlement for PP; similar numbers are obtained
when comparing bulk carbonyl-index with the earliest observa-
tions of brittle features as discussed later [44].
In comparison, oxidation of elastomers nearly always affects
their elastic modulus, mostly via oxidative crosslinking. Most
elastomers harden as part of oxidative aging which facilitates stress
cracking and loss of desirable properties. It is well known that in-
tegrated oxidation rates (i.e. oxidation levels) coupled with tensile
elongation or modulus profiling data, allows for excellent correla-
tions to be established [2,3,37,45]. An example is shown in Fig. 3 for
thermal oxidation of a soft polyurethane rubber, and in Fig. 4 for
localized oxidation levels and modulus changes during aging of
neoprene. The predicted oxidation levels and localized modulus for
the neoprene rubber material are based on thermal DLO modeling
of relative localized oxidation rates for this specific geometry.
Thermo-set materials, such as highly crosslinked epoxies, bis-
maleimides, benzoxazines or phenolics, seem to accommodate
Fig. 4. Diffusion limited oxidation and its predicted correlation with localized modulus
significant oxidation levels. So far exact numbers for oxidation for a neoprene elastomer specimen under highly accelerated thermal aging conditions
levels in correlation with modulus or other property changes for of 15 days at 125  C. It shows the absence of oxidative damage and therefore soft
these materials do not exist. Under highly oxidative pyrolytic material in the center. Model input data taken from Ref. [3].

Please cite this article in press as: Celina MC, Review of polymer oxidation and its relationship with materials performance and lifetime
prediction, Polymer Degradation and Stability (2013), http://dx.doi.org/10.1016/j.polymdegradstab.2013.06.024
4 M.C. Celina / Polymer Degradation and Stability xxx (2013) 1e11
required. Oxidation rates can be determined with short term
thermal exposure [58,59]. However, if high oxidation levels are
required so that measurable moduli changes or significant loss of
tensile elongation develop, then extended exposure and strongly
accelerative aging environments are needed [2e4,37]. Therefore,
while oxidation rates can be measured for low oxidation levels at
ambient temperature, higher oxidation levels depend on acceler-
ated aging; hence, oxygen consumption experiments often run in
parallel with aging studies. Surface modulus and modulus profiling
analyses, when combined with local oxidation state from DLO ap-
proximations/predictions, can also help to extract the degree to
which chemical oxidation affects elastic properties [3,12,37].
2.3. Molecular weight changes as part of oxidative degradation
Another polymer oxidation characterization approach which
provides significant mechanistic insight is linking oxidative
degradation with molecular weight changes. It is well known that
degradation will proceed via scission or crosslinking reactions.
Unfortunately, it is not always possible to predict how specific
polymers will behave. For example, it is generally accepted that
polyolefins show predominantly scission behavior under oxidative
conditions as evidenced by reduced molecular weight, but more
crosslinking characteristics under inert (non-oxidative) degrada-
tion which develops during g-irradiation under N2 or under high
temperature DLO conditions. However, morphology, crystallinity
and the aging temperature are important parameters which affect
this process. For example, crosslinked polyolefins with low crys-
tallinity seem to crosslink further under thermo-oxidative condi-
tions at high temperatures where partial crystalline melting
becomes active [60], but related materials show strong scission
behavior under photo-oxidative low temperature conditions [61].
In comparison, most elastomeric materials tend to crosslink under
oxidative conditions, leading to increased modulus and hardening
with embrittlement [2,3,12,37,58,62e65]. Thermo-set materials
tend to also crosslink under thermo-oxidative conditions with the
modulus increasing in parallel [46,48,54e56,66].
A systematic review of how molecular weight changes are
coupled with brittle behavior and oxidation as measured by
carbonyl growth was conducted by Fayolle and co-workers for
polyolefins [44,67e70] and other polymers [71e76]. In particular,
the results summarized how a critical molecular weight exists
which corresponds with a ductile to brittle transition during
oxidative polymer degradation [44,69,77]. For example critical Mw
values (in kDa) were provided as 28 for PC, 40 for PPMA, 60 for PS,
70 for PE, 200 for PP, or 130 for PP fibers, 70 for POM, 200 for PTFE
[71], or 17 for PA 6,6 [76]. Mechanistically, chain scission, followed
by a decrease in molecular weight and chemi-crystallization, was
then discussed as affecting tie molecules and interlamellar spacing
and ultimately leading to brittle sensitivity. Interestingly, embrit-
tlement is seen as occurring before significant damage in the
entanglement network, at higher than expected Mw. Because of the
involvement of tie molecules and critical interlamellar domains,
chain scission damage is essentially amplified to result in brittle
features [77].
Considering the relatively high remaining Mw when brittle
failure occurs and the limited reduction in average Mw from the
original material, it is clear that from a chemistry point of view very
little oxidation is required. For example, to discuss a few basic
numbers, if PP with an initial MW of 400 kDa oxidatively cleaves
and the critical Mw is on the order of 200 kDa, then ideally only one
effective scission in the middle of the chain is required to transition
the material from ductile to brittle. In this case only 0.008%
oxidation would be required. Three perfect scission events in this
simplistic example would drop Mw to 100 kDa and only require
Please cite this article in press as: Celina MC, Review of polymer oxidation and its relationship with materials performance and lifetime
prediction, Polymer Degradation and Stability (2013), http://dx.doi.org/10.1016/j.polymdegradstab.2013.06.024
0.024% oxidation. Such approximate numbers are consistent with
the essentially non-detectable carbonyl absorption when this
transition occurs (see Fig. 1 in Ref. [44]). A detection limit of
0.005 mol/kg carbonyl groups equals w1.3 carbonyl groups for Mw
of 270 kDa, the equivalent of 0.016% oxidation. In comparison, a
Nylon material undergoing degradation from Mw of 50e20 kDa via
CeN scission chemistry will require a minimum of w0.16% oxida-
tion for the relevant scission to occur. The required chain scissions
depend on the initial molecular weight and Mw associated with
failure and do not require much oxidative ‘chemistry,’ yet can
significantly change toughness and tensile properties.
This situation is further complicated when heterogeneity as-
pects in thermo-oxidation are considered. More localized reactions
can create more volatiles and small molecular weight fragments,
perhaps increasing oxidation chemistry yields. One such contro-
versy arises when the chemical variety of volatile species produced
during oxidation of PP or Nylon is considered, as analyzed with
sensitive mass spectrometry methods (MS) [78e82]. Unfortunately,
MS analysis does not easily quantify total amounts of volatiles per
sample weight and exposure condition, and without a clear un-
derstanding of percent oxidation level associated with the type and
absolute amounts of chain scission products, the underlying reac-
tion dynamics cannot be established. However, it is clear that each
small fragment originates from two scission events; further, it is
unlikely that for example one Nylon molecule only produces one X
volatile product and another molecule only one Y product to
maintain limited chain scission, while at the same time producing a
large range of degradation species [79,83]. Hence, the balance be-
tween total volatile yield and at the same time maintaining critical
average molecular weights as evidenced by ongoing ductile prop-
erties holds the key to the discussion of average homogeneous ki-
netics versus localized ‘chemistry amplification,’ here resulting in
higher scission yields with many different volatiles. Mechanical
properties and therefore critical Mw values were maintained,
despite the identification of a large number of volatile products in
corresponding experiments [79,83]. This cannot be explained with
a model of a few localized average chain scission events. Instead, it
is clear evidence for the participation of heterogeneous oxidation
domains which are not represented by current homogeneous ki-
netic models [76].
2.4. Kinetic modeling
Kinetic modeling is a broad approach that is pursued by a
number of research groups to help in the interpretation of polymer
oxidation, the evaluation of the underlying reaction mechanisms
and for the development of kinetic data extrapolation strategies
[84e88]. Kinetic models of oxidation pathways are built on detailed
mechanistic degradation chemistry, yet they do not easily embrace
the physical constraints of localized solid state reactions and
mechanistic variations with temperature, perhaps based on
morphology and reactivity zones. Reactive transport models which
address DLO are often grouped together with kinetic modeling
approaches, but are actually quite separate approaches, as the more
physical competitive process of localized diffusive oxygen reagent
supply and consumption terms are not necessarily dependent on
the exact chemical reaction mechanism. It is not intended here to
present a critique of kinetic modeling, but nevertheless caution
should be taken for a number of reasons. Some of the outstanding
challenges associated with kinetic modeling of thermal polymer
degradation are:
 The determination of rate constants from accelerated degra-
dation depends on the kinetic scheme that is being applied.
Parameters are often extracted based on data fits with initial
 M.C. Celina / Polymer Degradation and Stability xxx (2013) 1e11
mechanistic assumptions, but may then be discussed to justify
the underlying mechanism. Suitable fits may evolve for a
particular kinetic pathway, but other mechanisms may provide
correlations of similar quality. Correlations between kinetic
models and experiments which are meant to provide rate
constants are encouraging, but the models should be regarded
with some caution as they may not be exclusive. Models can
provide guidance, but obviously can be no better than the as-
sumptions applied to derive the underlying model input pa-
rameters (rate constants and activation energies).
 Kinetic parameters (rate constants) are often only available
when sufficiently fast degradation processes can be experi-
mentally observed, i.e. when conducting classic accelerated
aging studies to derive model input parameters. Long-term
aging behaviors where complicating physical processes may
evolve over large temperature differences are not easily
accommodated (as with any limited accelerated aging study)
and the applied models can therefore not confidently predict
them.
 Extrapolation is associated with assumptions of the prevailing
activation energy and perhaps Arrhenius linearity. For many
materials it has been shown that linearity may not apply and
therefore combinations of rate constants do not necessarily
follow basic model assumptions with temperature. Most ki-
netic models do not establish Arrhenius behavior for extended
temperature regimes, but rather assume it to be valid for lower
temperatures as part of extrapolations [76].
 Many of the kinetic models which have been proposed rely on
initial hydroperoxide levels to trigger follow-up reactions [46e
48]. For high temperature polymer degradation, it is ques-
tionable whether such hydroperoxide based approaches are
always appropriate, since there are limits to chemical forma-
tion and thermal stability of hydroperoxides. Another issue is
associated with initial hydroperoxide level assumptions. For
example, it is difficult to comprehend how a kinetic model
describing the thermo-oxidation of Nylon accommodates an
oxidation level of only 0.0056 mol/l carbonyl scission products
(0.016% oxidation) for mechanical failure, but requires
0.05 mol/l of initial hydroperoxides (i.e. a significant pre-
oxidation level of 0.14%) [76,89]. These hydroperoxides
should decompose and result in scission and follow-up
chemistry; therefore additional oxidative scission events may
occur. Further, the correlation between oxidation level and
effective scission required for failure should also depend on the
morphology and initial Mw condition.
 Current kinetic models also do not accommodate heterogene-
ity aspects in solid polymer degradation. There are significant
challenges associated with the more localized reactions and
reactivity zones encountered in solid polymer degradation
than covered by the widely-used description based on solution
chemistry reaction kinetics.
One example where high temperature and highly accelerated
polymer degradation, and for that manner any kinetic model fitting
such data, fails to predict long-term aging, is in situations where
‘inverse temperature effects’ are observed. Such phenomena evolve
when effective degradation levels may change with temperature or
depend on morphology, hence where the underlying assumption of
any kinetic model of equal ‘chemistry’ for equal physical property
changes is no longer valid. For example, this was observed for
combined thermal radiation oxidative degradation of semi-
crystalline cable insulation materials [90,91]. Another example is
given when rapid accelerated qualification testing of polymers
excludes the participation of oxidative processes, but when
oxidation will be the dominant long-term degradation process [92].
Please cite this article in press as: Celina MC, Review of polymer oxidation and its relationship with materials performance and lifetime
prediction, Polymer Degradation and Stability (2013), http://dx.doi.org/10.1016/j.polymdegradstab.2013.06.024
5
As discussed before, kinetic models which rely on data obtained
under DLO conditions lack predictive value for non-DLO conditions.
2.5. Importance of diffusion limited oxidation (DLO)
Whenever dealing with highly accelerated thermal degradation
of polymers and larger sample specimens, the complicating effects
of diffusion limited oxidation (DLO) must be considered. This sit-
uation applies when degradative acceleration does not only result
in much faster chemistry (as intended for predictive studies), but
also leads to the introduction of competing physical phenomena as
part of the degradation process. Fig. 5 shows schematically that
DLO can occur in many different accelerated aging scenarios. Of
course for very large specimens DLO may already exist at ambient
temperatures [2]. Most polymer degradation pathways evolve
around competition between oxidative versus inert degradation
processes, i.e. when oxygen supply is diffusion limited because of
too much reactive consumption within the material and spatially
dependent degradation zones evolve. Under accelerated condi-
tions, depending on geometry and specific material properties, any
oxidative aging process has the potential to transition from ho-
mogeneous aging into a heterogeneous oxidation process, perhaps
better defined as bulk versus surface dominated degradation. If the
exact oxygen supply and consumption rates are known under
thermal conditions, it is then possible to develop a spatially
dependent degradation model which couples the O2 supply with
consumption chemistry, for example applied under thermal only
[2,3,11,12,37], or under thermal-g-radiative conditions [93,94].
Diffusion limited oxidation (DLO) predictions describe spatially
dependent degradation chemistry based on competition between
oxygen supply (physics) and consumption (chemistry). Such
models explain spatially dependent changes in modulus, i.e. sur-
face degradation versus bulk property changes, and generally can
predict the magnitude of chemical degradation as a function of
depth. For larger specimens the ingress of oxygen from the surface
becomes a self-limiting maximum flux, i.e. when the decreasing
partial reactivity is integrated with depth, and further depth does
not contribute additional chemical oxidation reactivity (i.e. when
more material is added beyond the reactive zone). DLO models
which can predict localized chemistry have shown excellent cor-
relation with spatial variations in mechanical properties of
degraded polymers [2,3,12,37].
DLO complications are one of the most important components
of accelerated polymer aging. They apply to thermally driven, UV
photo-oxidation and radiation oxidative processes (see Fig. 5), as
well as many severely accelerated polymer testing approaches
where oxygen is involved, such as oxidation induction time (OIT)
Stress variable DLO conditions
Temperature excessive T
(high temperature
O2 oven, TGA or similar)
UV exposure strong dose rate
(high dose rate
O2 Polymer sample
wheatherometer)
O2 Radiation high dose rate
(intense gamma,
e-beam, ion beam)
Mechanical severe load
stress surface cracking
Fig. 5. DLO situations may arise during severely accelerated polymer degradation, all
resulting in preferential surface versus bulk degradation phenomena.
6 M.C. Celina / Polymer Degradation and Stability xxx (2013) 1e11

or TGA pyrolysis. Fig. 6 emphasizes that a significant increase in
temperature not only accelerates chemistry, but eventually in-
troduces additional physical processes. High temperature
methods, such as severely accelerated thermal aging (high tem-
perature ovens), oxidation induction time (OIT) or thermo-
gravimetric analysis (TGA) measurements are often subject to
DLO, even when small samples are used. Such techniques may
offer avenues for stability comparison of nearly identical material
formulations, but generally lack extrapolative value as discussed
later [32e34,36,95e97]. In some studies OIT data have been an
attractive approach to resolve stabilizer diffusion and depletion
aspects in polyolefins [98,99] or stabilizer profiles associated with
heterogeneous aging in PE nano-composites [100].
If the complicating phenomena of DLO are poorly understood, it
is clear that a non-correlation between ambient oxidative aging
and predictive aging under partially or mostly inert conditions can
quickly develop. As long as only surface properties are of interest,
DLO is not important. In contrast, whenever accelerated aging is
focused on uniform sample degradation or bulk property assess-
ments, then DLO is an important component of accelerated expo-
sure and must be considered. This means that there are intrinsic
limits for maximum temperatures, as long as temperature accel-
erates the oxidation rate, coupled with a maximum thickness or
volume for specimens for which homogeneous aging will apply.
Similarly, under UV or g-radiation exposure there is a limit in
thickness and dose rate for homogenous oxidative processes to
occur. DLO is one reason that high dose rate exposure (mostly for
sterilization purposes) results in surface polymer degradation,
whereas the bulk of the material experiences non-oxidative
conditions.
Rapid qualification testing of polymers for performance pre-
dictive purposes in engineering applications, where highly
accelerated conditions are applied based on significant tempera-
tures coupled perhaps with high UV or gamma dose rates, may
give a sense of tremendous over-exposure and hence the
perception of performance margins. However, unless very thin
films are used under such conditions, DLO effects will induce inert
dominated degradation, and those will most likely not correlate
with the use application, where oxidative conditions are normally
expected to be much more important. The situation described
here will apply to the qualification testing of cable assemblies
[92], to O-rings and seals and related materials. Similarly, DLO is
one reason that the reciprocity law (equal damage for equal dose)
under highly accelerated UV exposure cannot apply when the
dose rate is sufficiently high, meaning that for a particular thick-
ness non-oxidative photochemistry processes commence and
total polymer degradation yields will no longer be proportional
with dose.
DLO is at the core of an unresolved discrepancy between the
complexity of polymer degradation and knowledge gained in the
field of ‘polymer aging’, and the more engineering-based testing
approaches incorporated into standardized methods for lifetime
prediction purposes of many polymeric materials. Limited highly
accelerated testing methods using a few data points and relying on
linear Arrhenius-type extrapolations do not always reproduce
homogeneous degradation processes and do not accommodate
mechanistic changes and non-linear temperature dependencies.
DLO behavior is a key contributor to divergence between acceler-
ated thermal aging and actual aging behavior under use condi-
tions. A comprehensive understanding and prediction of DLO
requires two terms, namely the O2 permeation [101] and oxidation
rate as a material property (O2 consumption) which has been
published on numerous occasions [1,3,4,37,58,59,102,103]. Com-
putations then allow for spatially dependent oxidation chemistry
to be described, as shown for example in Fig. 4, which shows inert
degradation in the center and oxidatively dominated thermal
degradation toward the exterior for a particular geometry. DLO
models can also help establish the amount of material in a spec-
imen that actively oxidizes. Under homogeneous conditions the
material oxidizes throughout, whereas under heterogeneous
conditions only a small fraction of the material oxidizes at all. Fig. 7
shows such transitions for an XLPO material of different thick-
nesses with increasing temperature. With integration of localized
rates over the predicted degradation profile it is possible to
determine the amount of material that oxidizes at the equivalent
surface oxidation rate.
2.6. High temperature TGA approaches
Based on the evolving literature and perhaps driven by needs
originating from the standards community for more rapid testing
approaches as part of material qualification, there appears to be a
desire to embrace ramped TGA as one technique for rapid thermal
aging and as a kinetic evaluation tool with the potential to deliver
activation energies. Obviously such approaches are appealing
because of speed and available mathematical data interpretation
tools. While ramped TGA can certainly deliver data for pyrolytic
volatilization processes, for lifetime prediction purposes or more

OIT-T (DSC temperature ramp) TGA (Temperature ramp) High temp ovens Physics
In some instances these techniques may provide relative comparisons
OIT (Oxidation Induction Time) DLO TGA (isothermal) DTA, DMA
Isothermal DSC (220-160°C)
Temperature

Moderate temp ovens

Moderately accelerated aging
Monitor chemistry and
mechanical/physical changes Moderate UV exposure

Non-DLO Ambient aging processes

Goal: Representative test with predictive

capability for specific application Chemistry

DLO is the most important complication in highly accelerated oxidative degradation

Similar issues could occur for moisture diffusion and hydrolysis when significant rates are involved
Goal: Conduct aging under non-DLO to understand material behavior and chemistry

Fig. 6. Extensive temperature exposure transitions polymer oxidation from accelerated chemistry to include physics based anomalies. Diffusion limited oxidation can affect any
rapid high temperature characterization and aging methods.

Please cite this article in press as: Celina MC, Review of polymer oxidation and its relationship with materials performance and lifetime
prediction, Polymer Degradation and Stability (2013), http://dx.doi.org/10.1016/j.polymdegradstab.2013.06.024
 M.C. Celina / Polymer Degradation and Stability xxx (2013) 1e11 7

Very thin specimens will oxidize homogenously

even at very high T (high rates)
0.1mm
Homogeneous 1.0

Material fraction
At very low temperatures
0.8 thick materials will oxidize
homogenously
0.6 1mm

XLPO insulation
0.4 10mm material
0.2 100mm
Thick materials do not fully
0.0 oxidize at mid temperatures
Heterogeneous Significant DLO effects
250°C 170°C 100°C 60°C 25°C

1.8 2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6

1000/T [1/K]
Fig. 7. Example of DLO modeling for a crosslinked polyolefin material which can predict the equivalent fraction of material oxidizing at the surface oxidation rate. Any thickness will
transition to heterogeneous aging with increasing temperature. Even small 1 mm sized sample specimens for TGA experiments can be subject to DLO effects.

moderate oxidative aging processes TGA is questionable and should
be approached with caution for a number of reasons:
 For many polymers thermo-oxidative degradation is associated
with a weight increase due to incorporation of oxygen. In
contrast, TGA forces volatilization of the material and as such
represents a process that is often not encountered during
polymer aging.
 TGA yields Ea values in a high temperature regime that is far
removed from use temperatures. TGA is outside the framework
to deliver meaningful Ea for long-term polymer aging processes
and often yields Ea data that are much higher than observed in
other studies. Many polymers show curvature in Arrhenius
plots toward lower Ea values [1e3,5,6].
 Even for small samples, TGA under oxidative conditions is
subject to DLO effects. This method can therefore not easily
provide information for overall sample representative oxida-
tion kinetics and is greatly affected by diffusion of oxygen into
the material, and likewise volatile diffusion and desorption
from the sample surface. TGA polymer behavior can be very
different depending on whether a single sample (i.e. 10 mg) or
finely cut sample pieces are analyzed. Even isothermal TGA
measurements can be convoluted by DLO effects, which pro-
duce variability in activation energies extracted from such data
(see Fig. 8 for an EPDM material under oxygen). The results
shown in Fig. 8 from 210  C to 170  C imply that timeetem-
perature superposition of the weight loss versus aging time
curves is impossible. The relatively faster drop-off in weight
loss at 170  C is also indicative of the pronounced induction-
time behavior found for many EPR/EPDM materials during
 Modern TGA instruments may include software assisted data
analysis options. There is no limit to what can be done with
non-isothermal weight loss curves in terms of fitting differ-
ential equations from various mathematical models that
describe degradation conditions at different scanning rates.
However, it does not mean that the computational power
provided itself validates a particular mathematical treatment
or the resulting Ea and data analysis employed. Many studies
seem to lack a critical perspective on what the computational
output justifies and the challenges in the meaning of such
numbers for the material properties.
In principle, ramped TGA can provide kinetic data (i.e. Ea) when
the decomposition associated with weight loss involves well-
defined steps, and chemical conversion can be established to then
be examined with any of the existing mathematical treatments
[104e108]. These interpretations have the identical goal of extract-
ing Ea from ramped temperature experiments under different
scanning rates, but resolve the rate equation with different ap-
proximations for the determination of the kinetic triplet (functional
form of conversion, Ea and frequency factor). The CoatseRedfern
approach uses defined reaction orders (i.e. first or second) [107],
whereas Friedman and Vyazovkin rely on iso-conversional
10000
4% loss
3% loss
2% loss
1% loss
Ind time
1000
Time [min]

4% 82 KJ/mol
oven aging. While an Ea of 160 kJ/mol is observed for the in-
3% 72 KJ/mol
duction time and 1% weight loss data over the 170e210  C 2% 71 KJ/mol 4% 124 KJ/mol
temperature range, this material has yielded an Ea of 113 kJ/mol 1% 74 KJ/mol 3% 132 KJ/mol
tind. 73 KJ/mol 2% 147 KJ/mol
for loss in tensile elongation between 170 and 110  C and even 100
1% 160 KJ/mol
lower values with oxidation rate measurements toward 52  C tind. 160 KJ/mol
[103]. This is an example of the non-correlation in Ea that can
exist between lower temperature extended thermal aging and
a rapid TGA approach. 10 240°C 230°C 220°C 210°C 200°C 190°C 180°C 170°C

 TGA curves need to be thoroughly deconvoluted to yield

kinetic insight and activation energies, This requires actual 1.9 2.0 2.1 2.2 2.3
chemical conversion with temperature/time for clearly 1/T [10-3K-1]
defined decomposition steps. If multi-step reactions are
Fig. 8. Time data for equivalent weight loss levels taken from primary TGA curves from
involved they should consist of clearly separated single-stage isothermal analysis of an EPDM material under oxygen. It shows the effect of DLO
steps, as mentioned in standards such as ISO 11358 and resulting in curved Arrhenius behavior and the dependence of Ea on temperature and
ISO 11358-2. extent of weight loss.

Please cite this article in press as: Celina MC, Review of polymer oxidation and its relationship with materials performance and lifetime
prediction, Polymer Degradation and Stability (2013), http://dx.doi.org/10.1016/j.polymdegradstab.2013.06.024
8 M.C. Celina / Polymer Degradation and Stability xxx (2013) 1e11
treatment of the rate equation [106,108]. The Ozawa and FlynneWall
iso-conversional approach utilizes a linear approximation in the
integrated rate constant with respect to inverse temperature in a
narrow range of Ea/RT [104,105]. Such models are mathematically
correct as long as exact conversion can be derived from weight loss
curves for a particular decomposition step; and this represents the
key challenge with ramped TGA.
Ramped TGA data interpretation is intrinsically complicated or
often impossible because polymer degradation rarely proceeds
with well resolved individual decomposition steps. Weight loss
curves often show multiple parallel degradation steps, and hence
the activation energies when ‘brute-force’ mathematics are applied
to weight loss data as perceived conversion, do not replicate a
particular mechanistic step that may govern long-term aging under
limited weight loss conditions. Ramped TGA experiments tend to
focus on degradation chemistry pathways that for most polymers
are too far removed from the mechanisms that are relevant for the
usual application conditions. Unless thermal pyrolytic decomposi-
tion behavior is the focus of a particular study, ramped TGA pre-
sents a high risk approach and introduces significant uncertainty
for any polymer lifetime prediction purposes.
2.7. Other current developments
This review also requires a few comments on some related
encouraging research trends relevant to heterogeneity and
controlled degradation. There has long been a need to better
visualize heterogeneous thermo-oxidative degradation processes
before they appear as visible cracks. Chemiluminescence (CL)
originally provided direct evidence for interaction between
degrading PP particles [18] and spreading phenomena in solid
film samples [109e112], with ‘infectious degradation’ even
observed between different polymers [113] and likely carried by
low molecular weight peroxidic species [114]. The usefulness of CL
has been greatly expanded with the advent of pro-fluorescent
nitroxide reagents which allow monitoring the in-situ formation
of free radicals during thermo-oxidative [115e117] or even during
processing conditions [118]. Such avenues will greatly improve
diagnostic tools and will enable more insight into heterogeneous
polymer degradation aspects. It is equally important that a better
understanding of the effects of nano-particles on polymer
degradation behavior be developed [119,120], since nano-
particulate fillers can also affect degradation pathways and
rates. This plays into the accelerated environmental degradation
behavior of polymers (mostly polyolefins) where UV degradation
needs to be balanced with catalytically enhanced thermo-
oxidative processes, which for example can be facilitated by the
incorporation of pro-degradants. It is clear that optimized mate-
rials with on demand stability under environmental conditions
require a solid understanding of the competing behaviors be-
tween stabilizers, pro-degradants, processing and thermal versus
Fig. 9. Visual example of competing oxidative processes (thermal and UV) in a mock PV section showing oxidative degradation below the glass cover (UV) and under the inert cell
(thermal) due to O2 diffusion trapping in a thin film, but strong DLO behavior above the cell. The encapsulant above the cell center experiences non-oxidative photo-chemistry.
Please cite this article in press as: Celina MC, Review of polymer oxidation and its relationship with materials performance and lifetime
prediction, Polymer Degradation and Stability (2013), http://dx.doi.org/10.1016/j.polymdegradstab.2013.06.024
UV exposure conditions [121,122]. All this has to be considered
with the goal of true biodegradation, which appears to be
dependent on the interaction of fine polymer particulate degra-
dation products with moisture based on relative surface area,
versus just ‘bio-fragmentation’ leaving undesirable polymer trash
in the environment.
2.8. Research strategy for better predictive models
There is a need for broad polymer reliability models which can
predict changes in polymer performance associated with specific
physical properties based on an understanding of time-dependent
and spatially resolved degradation chemistry. One such application,
for example, can be found in photovoltaic (PV) modules where a
number of polymers are used as critical components and multiple
parallel degradation pathways are encountered. Long-term reli-
ability is a key demand, as is an understanding of the kinetics and
mechanisms that reduce power outcome with time. If the pro-
gression of spatially resolved chemistry is known as a function of
exposure conditions (for PV systems these are temperature, UV
doses, ingress of oxygen and moisture), and cross-correlations with
physical property are available as model input, then very powerful
predictive models can be developed. For example, oxidatively
driven degradation could be modeled for both thermo and photo-
oxidative processes subject to diffusion limited oxidation (DLO)
conditions (see a visual example in Fig. 9). Input parameters would
be oxygen supply based on permeation through a number of
polymers, as well as thermal and photo-oxidation rates. For the
latter, rates would need to be available in consumed oxygen mol/g s
and not only in produced carbonyl absorption versus deposited
dose as is usually reported. Subsequently, these models could be
improved by coupling oxidative degradation with parallel hydro-
lytic degradation processes, as long as moisture ingress parameters,
the kinetics of hydrolysis with temperature and localized water
concentration, and cross-correlation with the most important
physical property changes can be established. In the future the
evolving knowledge of polymer degradation processes coupled
with more easily available computational tools will allow for the
development of complex finite element (FE) DLO models to predict
the progression of spatially resolved property changes in polymers.
A generic approach for such a potential target is given below:
 Characterization of oxygen permeation through relevant
polymers as a function of temperature. Measurement of
permeability and its activation energy as material properties.
Derive numbers for oxygen solubility for all polymers involved.
 Characterization of thermo-oxidative polymer degradation.
Measurements of intrinsic thermal oxidation rates and their
activation energies.
 Estimations of oxidation rates under UV exposure from IR
changes (carbonyl, hydroxyl, etc.) as a function of deposited
 M.C. Celina / Polymer Degradation and Stability xxx (2013) 1e11
dose or alternative avenues to measure oxidation rates as an
intrinsic material property under photo-oxidative polymer
degradation conditions.
 Define localized oxygen supply and consumption scenarios for
specific geometries. Use DLO model to predict spatially
dependent oxidation states with progressive temperature and
UV exposure under static and dynamic conditions.
 Establish cross-correlation of ‘degradation chemistry’ with
‘performance critical physical properties’ using classic polymer
degradation approaches, i.e. linking %-oxidation with light
transmissivity, discoloration, adhesion, embrittlement or
similar features.
 Predict progressive material degradation and property changes
under long-term exposure conditions.
 Apply similar reaction DLO models to hydrolytic processes as
long as water ingress, solubility, permeation parameters, and
consumption kinetics have been established. This would also
require cross-correlations between ‘hydrolysis state’ and rele-
vant physical properties.
3. Conclusion
This brief summary of the status and major trends in polymer
oxidation, its measurements, modeling, and effects shows, that all
the intricate details which affect the oxidation of polymers and
basic property changes are not easily addressed. Many researchers
have contributed to the comprehensive understanding of degra-
dation phenomena, which has evolved in parallel with polymer
synthesis and processing over the last 60 years. Tremendous
progress has been achieved, considering the knowledge docu-
mented in the literature and the many polymers and aging phe-
nomena that have been resolved to make better materials. At the
same time there is great excitement about the evolving opportu-
nities, particularly when linking experimental science with the
growing power of numerical simulations.
With an appropriate level of experimental planning and
patience in aging programs and with the support of oxidation rate
measurements, excellent predictions on relative polymer stability
can be made. This is valid for thermal aging as well as photo-
oxidation where great progress in mechanistic details, stabilizer
behavior, and exposure dose to performance correlations has been
achieved. Over the last 25 years, it has also been established that
DLO effects can seriously complicate predictive aging experiments,
as the spatial distribution of oxidative chemistry can greatly differ
between accelerated aging and ambient conditions. This is one of
the biggest challenges in predictive aging, as it documents the
intrinsic limits in accelerating ‘chemistry’; at some point physical
processes will invariably govern the degradation behavior. With
that in mind, rapid qualification testing where a few weeks of
exposure under DLO conditions are perhaps intended to predict
performance over decades of service where DLO does not apply are
not only highly questionable, they are scientifically flawed. The
established interpretation of the complexity of polymer aging
processes when oxidation is involved has revealed the limits of
what accelerated aging can achieve or do incorrectly.
When combining polymer chemistry and physics, in the
simplest approach, oxidative polymer aging has never been just
temporal, but always temporal and spatial. Much research has been
dedicated toward understanding these phenomena. The develop-
ment of spatial damage distributions needs to include a discussion
of surface versus bulk degradation, DLO effects occurring during
overly fast aging experiments or in TGA screening methods, or
address the challenges embedded in ‘bio-fragmentable’ versus ‘bio-
degradable’ polymers and carefully tuned degradation on demand.
Please cite this article in press as: Celina MC, Review of polymer oxidation and its relationship with materials performance and lifetime
prediction, Polymer Degradation and Stability (2013), http://dx.doi.org/10.1016/j.polymdegradstab.2013.06.024
9
Polymer oxidation will never cease, but analytical methods will
continue to evolve, as will model capabilities. Kinetic models that
perhaps embrace more heterogeneous pathways consistent with
solid state reactions, incorporation of non-linear Arrhenius
behavior and morphological contributions to the effectiveness of
oxidative damage, should be expected. Significant progress will
occur in applied DLO models that soon will combine thermal and
parallel photo-oxidative processes. With similar efforts dedicated
to acquiring critical material parameters as model input (i.e.
diffusion and permeation constants, plus oxidation rates under
thermal and UV- and g-radiation), spatially dependent visualiza-
tion and prediction of polymer oxidation in complex assemblies
and geometries under dynamic exposure conditions are within
reach. With suitable model calibration parameters (i.e. oxidation
state versus physical property correlations) made available we will
be a significant step closer to our goal of predicting localized failure
in components based on ‘Polymer Oxidation.’
Acknowledgments
This paper is an invited contribution based on a plenary address
given at the 2012 MODEST conference. All previous and current
collaborators, colleagues, mentors, technical staff and students are
gratefully acknowledged. Adam Quintana is recognized for creative
approaches to DLO modeling. Sandia National Laboratories is a
multi-program laboratory managed and operated by Sandia Cor-
poration, a wholly owned subsidiary of Lockheed Martin Corpora-
tion, for the U.S. Department of Energy’s National Nuclear Security
Administration under contract DE-AC04-94AL85000.
References
[1] Celina M, Gillen KT, Assink RA. Accelerated aging and lifetime prediction:
review of non-Arrhenius behaviour due to two competing processes. Polym
Degrad Stab 2005;90:395.
[2] Celina M, Graham AC, Gillen KT, Assink RA, Minier LM. Thermal degradation
studies of a polyurethane propellant binder. Rubber Chem Technol 2000;73:678.
[3] Celina M, Wise J, Ottesen DK, Gillen KT, Clough RL. Correlation of chemical
and mechanical property changes during oxidative degradation of neoprene.
Polym Degrad Stab 2000;68:171.
[4] Gillen KT, Celina M, Bernstein R. Review of the ultrasensitive oxygen con-
sumption method for making more reliable extrapolated predictions of
polymer lifetimes. Ann Tech Conf Soc Plast Eng 2004;62:2289.
[5] Gillen KT, Bernstein R, Celina M. Non-Arrhenius behavior for oxidative
degradation of chlorosulfonated polyethylene materials. Polym Degrad Stab
2005;87:335.
[6] Gillen KT, Bernstein R, Derzon DK. Evidence of non-Arrhenius behavior from
laboratory aging and 24-year field aging of polychloroprene rubber mate-
rials. Polym Degrad Stab 2005;87:57.
[7] Bernstein R, Derzon DK, Gillen KT. Nylon 6.6 accelerated aging studies:
thermal-oxidative degradation and its interaction with hydrolysis. Polym
Degrad Stab 2005;88:480.
[8] Gillen KT, Celina M, Bernstein R. Methods for predicting lifetimes and
remaining lifetimes of polymers. In: Martin JW, Ryntz RA, Dickie RA, editors.
Service life prediction, challenging the status quo. Federation of Societies for
Coatings Technology; 2005. p. 67.
[9] Gillen KT, Clough RL. Quantitative confirmation of simple theoretical models
for diffusion-limited oxidation. In: Clough RL, Shalaby SW, editors. Radiation
effects on polymers. ACS series, vol. 475. Washington, DC: ACS; 1991. p. 457.
[10] Gillen KT, Clough RL. Rigorous experimental confirmation of a theoretical
model for diffusion-limited oxidation. Polymer 1992;33:4358.
[11] Audouin L, Langlois V, Verdu J, de Bruijn JCM. Role of oxygen diffusion in
polymer aging: kinetic and mechanical aspects. J Mater Sci 1994;29:569.
[12] Wise J, Gillen KT, Clough RL. Quantitative model for the time development of
diffusion-limited oxidation profiles. Polymer 1997;38:1929.
[13] Rincon-Rubio LM, Colin X, Audouin L, Verdu J. A theoretical model for the
diffusion-limited thermal oxidation of elastomers at medium temperatures.
Rubber Chem Technol 2003;76:460.
[14] Celina M, Gillen KT. Advances in exploring mechanistic variations in thermal
aging of polymers. In: Martin JWR, Rose A, Chin Joannie, Dickie Ray A, edi-
tors. Service life prediction of polymeric materials global perspectives.
Springer; 2009. p. 45.
[15] Billingham NC. Localization of oxidation in polypropylene. Makromol Chem
Macromol Symp 1989;28:145.
10 M.C. Celina / Polymer Degradation and Stability xxx (2013) 1e11
[16] Billingham NC. Physical phenomena in the oxidation and stabilization of
polymers. In: Pospisil J, Klemchuk PP, editors. Oxidation inhibition in organic
materials. Boca Raton: CRC Press; 1990. p. 249 [chapter 6].
[17] Celina M, George GA. A heterogeneous model for the thermal oxidation of
solid polypropylene from chemiluminescence analysis. Polym Degrad Stab
1993;40:323.
[18] Celina M, George GA, Billingham NC. Physical spreading of oxidation in solid
polypropylene as studied by chemiluminescence. Polym Degrad Stab
1993;42:335.
[19] Celina M, George GA. Heterogeneous and homogeneous kinetic analyses of
the thermal oxidation of polypropylene. Polym Degrad Stab 1995;50:89.
[20] Pospisil J, Pilar J, Marek A, Billingham NC, Horak Z, Nespurek S. Environ-
mental and micro-environmental influences affecting the performance of
polymer stabilisers. Polym Polym Compos 2004;12:111.
[21] Pospisil J, Pilar J, Billingham NC, Marek A, Horak Z, Nespurek S. Factors
affecting accelerated testing of polymer photostability. Polym Degrad Stab
2006;91:417.
[22] Pilar J, Michalkova D, Sedenkova I, Pfleger J, Pospisil J. NOR and nitroxide-
based HAS in accelerated photooxidation of carbon-chain polymers; com-
parison with secondary HAS: an ESRI and ATR FTIR study. Polym Degrad Stab
2011;96:847.
[23] Marek A, Kapralkova L, Schmidt P, Pfleger J, Humlicek J, Pospisil J, et al.
Spatial resolution of degradation in stabilized polystyrene and poly-
propylene plaques exposed to accelerated photodegradation or heat aging.
Polym Degrad Stab 2006;91:444.
[24] George GA, Celina M. Homogeneous and heterogeneous oxidation of poly-
propylene. In: Hamid SH, editor. Handbook of polymer degradation. New
York: Marcel Dekker Inc.; 2000. p. 277.
[25] Billingham NC, George GA. Chemiluminescence from oxidation of poly-
propylene: some comments on a kinetic approach. J Polym Sci Part B: Polym
Phys 1990;28:257.
[26] Colin X, Marais C, Verdu J. Kinetic modelling and simulation of gravimetric
curves: application to the oxidation of bismaleimide and epoxy resins. Polym
Degrad Stab 2002;78:545.
[27] Pospisil J, Nespurek S. Highlights in chemistry and physics of polymer sta-
bilization. Macromol Symp 1997;115:143.
[28] Zweifel H. Stabilization of polymeric materials. Berlin: Springer-Verlag; 1998.
[29] Pospisil J, Klemchuk PP, editors. Oxidation inhibition in organic materials,
vol. 1. Boca Raton: CRC Press; 1990.
[30] Pospisil J, Klemchuk PP, editors. Oxidation inhibition in organic materials,
vol. 2. Boca Raton: CRC Press; 1990.
[31] Pospisil J. Highlights in the inherent chemical activity of polymer stabilizers.
Environ Sci Pollut Control Ser 2000;21:191.
[32] Billingham NC, Bott DC, Manke AS. Application of thermal analysis methods
to oxidation and stabilization of polymers. In: Grassie N, editor. De-
velopments in polymer degradation. London: Applied Science; 1981. p. 63.
[33] Pospisil J, Horak Z, Pilar J, Nespurek S, Billingham NC, Habicher WD. Effect of
testing conditions on performance and durability of stabilizers in plastics.
Polym Polym Compos 2003;11:81.
[34] Pospisil J, Horak Z, Pilar J, Billingham NC, Zweifel H, Nespurek S. Influence of
testing conditions on the performance and durability of polymer stabilizers
in thermal oxidation. Polym Degrad Stab 2003;82:145.
[35] Celina M, Skutnik Elliott JM, Winters ST, Assink RA, Minier LM. Correlation of
antioxidant depletion and mechanical performance during thermal degra-
dation of an HTPB elastomer. Polym Degrad Stab 2006;91:1870.
[36] Woo L, Ling MTK, Eu B, Sandford C. Application and limitations on thermal
and spectroscopic methods for shelf-life prediction. Thermochim Acta
2006;442:61.
[37] Celina M, Wise J, Ottesen DK, Gillen KT, Clough RL. Oxidation profiles of
thermally aged nitrile rubber. Polym Degrad Stab 1998;60:493.
[38] Billingham NC, Prentice P, Walker TJ. Some effects of morphology on oxidation
and stabilization of polyolefins. J Polym Sci Polym Symp 1977;57:287.
[39] Billingham NC, Calvert PD. Applications of ultra-violet microscopy to studies
of the oxidation and stabilization of polymers. Pure Appl Chem 1985;57:
1727.
[40] Billingham NC. The physical chemistry of polymer oxidation and stabiliza-
tion. In: Scott G, editor. Atmospheric oxidation and antioxidants. Amster-
dam: Elsevier; 1993. p. 219.
[41] Goss BGS, Barry MD, Birtwhistle D, George GA. Modeling of infectious
spreading in heterogeneous polymer oxidation I. Development of a sto-
chastic model. Polym Degrad Stab 2001;74:271.
[42] Goss BGS, Blakey I, Barry MD, George GA. Modeling of infectious spreading in
heterogeneous polymer oxidation II. Refinement of stochastic model and
calibration using chemiluminescence of polypropylene. Polym Degrad Stab
2001;74:523.
[43] George GA, Celina M, Lerf C, Cash G, Weddell D. A spreading model for the
oxidation of polypropylene. Macromol Symp 1997;115:69.
[44] Fayolle B, Audouin L, Verdu J. A critical molar mass separating the ductile and
brittle regimes as revealed by thermal oxidation in polypropylene. Polymer
2004;45:4323.
[45] Gillen KT, Assink R, Bernstein R, Celina M. Condition monitoring methods
applied to degradation of chlorosulfonated polyethylene cable jacketing
materials. Polym Degrad Stab 2006;91:1273.
[46] Pochiraju KV, Tandon GP, Schoeppner GA, editors. Long-term durability of
polymeric matrix composites. Springer Science þ Business Media; 2011.
Please cite this article in press as: Celina MC, Review of polymer oxidation and its relationship with materials performance and lifetime
prediction, Polymer Degradation and Stability (2013), http://dx.doi.org/10.1016/j.polymdegradstab.2013.06.024
[47] Pochiraju KV. Modeling thermo-oxidative aging and degradation of com-
posites. In: Pochiraju KV, Tandon GP, Schoeppner GA, editors. Long-term
durability of polymeric matrix composites. Springer Science þ Business
Media; 2011. p. 383.
[48] Colin X, Verdu J. Mechanisms and kinetics of organic matrix thermal
oxidation. In: Pochiraju KV, Tandon GP, Schoeppner GA, editors. Long-term
durability of polymeric matrix composites. Springer Science þ Business
Media; 2011. p. 311.
[49] Colin X, Verdu J. Aging of organic matrix composite materials. In: Nicolais L,
Borzacchiello A, Lee SM, editors. Wiley encyclopaedia of composites. 2nd ed.
2012. p. 35.
[50] Djouani F, Zahra Y, Fayolle B, Kuntz M, Verdu J. Degradation of epoxy coat-
ings under gamma irradiation. Radiat Phys Chem 2013;82:54.
[51] Galant C, Fayolle B, Kuntz M, Verdu J. Thermal and radio-oxidation of epoxy
coatings. Prog Org Coat 2010;69:322.
[52] Queiroz DPR, Fraisse F, Fayolle B, Kuntz M, Verdu J. Radiochemical ageing of
epoxy coating for nuclear plants. Radiat Phys Chem 2010;79:362.
[53] Nguyen T, Gu X, Vanlandingham M, Byrd E, Ryntz R, Martin JW. Degradation
modes of crosslinked coatings exposed to photolytic environment. J Coat
Technol Res 2013;10:1.
[54] Gardette J-L, Rivaton A, Therias S. Photodegradation processes in polymeric
materials. Photochem Photophys Polym Mater 2010:569.
[55] Larche JF, Bussiere PO, Gardette JL. Photo-oxidation of acrylic-urethane
thermoset networks. Relating materials properties to changes of chemical
structure. Polym Degrad Stab 2011;96:1438.
[56] Larche JF, Bussiere PO, Therias S, Gardette JL. Photooxidation of polymers:
relating material properties to chemical changes. Polym Degrad Stab
2012;97:25.
[57] Celina MC, Dayile AR, Quintana A. A perspective on the inherent oxidation
sensitivity of epoxy materials. Polymer 2013;54:3290.
[58] Wise J, Gillen KT, Clough RL. An ultrasensitive technique for testing the
Arrhenius extrapolation assumption for thermally aged elastomers. Polym
Degrad Stab 1995;49:403.
[59] Assink RA, Celina M, Skutnik JM, Harris DJ. Use of a respirometer to measure
oxidation rates of polymeric materials at ambient temperatures. Polymer
2005;46:11648.
[60] Gillen KT, Assink RA, Bernstein R. Nuclear energy plant optimization (NEPO):
final report on aging and condition monitoring of low-voltage cable mate-
rials. Albuquerque and Livermore: Sandia National Laboratories; 2005.
p. 286. SAND2005-7331.
[61] Celina M, George GA. Characterization and degradation studies of peroxide
and silane crosslinked polyethylene. Polym Degrad Stab 1995;48:297.
[62] Gillen KT, Celina M. The wear-out approach for predicting the remaining
lifetime of materials. Polym Degrad Stab 2000;71:15.
[63] Gillen KT, Celina M, Keenan MR. Methods for predicting more confident
lifetimes of seals in air environments. Rubber Chem Technol 2000;73:265.
[64] Gillen KT, Terrill ER, Winter RM. Modulus mapping of rubbers using micro-
and nano-indentation techniques. Rubber Chem Technol 2001;74:428.
[65] Gillen KT, Assink RA, Bernstein R. Condition monitoring approaches
applied to a polychloroprene cable jacketing material. Polym Degrad Stab
2004;84:419.
[66] Colin X, Verdu J. Thermal ageing and lifetime prediction for organic matrix
composites. Plast Rubber Compos 2003;32:349.
[67] Fayolle B, Audouin L, Verdu J. Initial steps and embrittlement in the thermal
oxidation of stabilized polypropylene films. Polym Degrad Stab 2001;75:123.
[68] Fayolle B, Colin X, Audouin L, Verdu J. Mechanism of degradation induced
embrittlement in polyethylene. Polym Degrad Stab 2007;92:231.
[69] Fayolle B, Richaud E, Verdu J, Farcas F. Embrittlement of polypropylene fiber
during thermal oxidation. J Mater Sci 2008;43:1026.
[70] Fayolle B, Tcharkhtchi A, Verdu J. Temperature and molecular weight
dependence of fracture behaviour of polypropylene films. Polym Test
2004;23:939.
[71] Fayolle B, Audouin L, Verdu J. Radiation induced embrittlement of PTFE.
Polymer 2003;44:2773.
[72] Fayolle B, Colin X, Audouin L, Verdu J. Relating kinetic models to embrit-
tlement in polymer oxidative aging. ACS Symp Ser 2009;1004:135.
[73] Fayolle B, Verdu J. Radiation aging and chemi-crystallization processes in
polyoxymethylene. Eur Polym J 2011;47:2145.
[74] Fayolle B, Verdu J, Bastard M, Piccoz D. Thermooxidative ageing of poly-
oxymethylene, part 1: chemical aspects. J Appl Polym Sci 2008;107:1783.
[75] Fayolle B, Verdu J, Piccoz D, Dahoun A, Hiver JM, G’sell C. Thermooxidative
aging of polyoxymethylene, part 2: embrittlement mechanisms. J Appl
Polym Sci 2009;111:469.
[76] El-Mazry C, Ben Hassine M, Correc O, Colin X. Thermal oxidation kinetics of
additive free polyamide 6-6. Polym Degrad Stab 2013;98:22.
[77] Fayolle B, Richaud E, Colin X, Verdu J. Review: degradation-induced
embrittlement in semicrystalline polymers having their amorphous phase
in rubbery state. J Mater Sci 2008;43:6999.
[78] Bernstein R, Thornberg SM, Assink RA, Irwin AN, Hochrein JM, Brown JR,
et al. The origins of volatile oxidation products in the thermal degradation of
polypropylene, identified by selective isotopic labeling. Polym Degrad Stab
2007;92:2076.
[79] White GV, Smith JN, Clough RL, Ohlhausen JA, Hochrein JM, Bernstein R. The
origins of CO2 and NH3 in the thermal-oxidative degradation of nylon 6.6.
Polym Degrad Stab 2012;97:1396.
 M.C. Celina / Polymer Degradation and Stability xxx (2013) 1e11
[80] Thornberg SM, Bernstein R, Mowery DM, Klamo SB, Hochrein JM, Brown JR,
et al. Insights into oxidation pathways, from gaseous products of poly-
propylene with selective isotopic labeling. Macromolecules 2006;39:5592.
[81] Thornberg SM, Bernstein R, Irwin AN, Derzon DK, Klamo SB, Clough RL. The
genesis of CO2 and CO in the thermooxidative degradation of polypropylene.
Polym Degrad Stab 2007;92:94.
[82] Bernstein R, Thornberg SM, Assink RA, Mowery DM, Kathleen Alam M,
Irwin AN, et al. Insights into oxidation mechanisms in gamma-irradiated
polypropylene, utilizing selective isotopic labeling with analysis by GC/MS,
NMR and FTIR. Nucl Instrum Methods Phys Res Sect B 2007;265:8.
[83] Smith Jonell N, White Gregory 2nd V, White Michael I, Bernstein R, Hochrein
James M. Characterization of volatile nylon 6.6 thermal-oxidative degrada-
tion products by selective isotopic labeling and cryo-GC/MS. J Am Soc Mass
Spectrom 2012;23:1579.
[84] Gillen KT, Wise J, Clough RL. General solution for the basic autoxidation
scheme. Polym Degrad Stab 1995;47:149.
[85] Audouin L, Gueguen V, Tcharkhtchi A, Verdu J. “Close Loop” mechanistic
schemes for hydrocarbon polymer oxidation. J Polym Sci Part A: Polym Chem
1995;33:921.
[86] Rincon-Rubio LM, Fayolle B, Audouin L, Verdu J. A general solution of the
closed-loop kinetic scheme for the thermal oxidation of polypropylene.
Polym Degrad Stab 2001;74:177.
[87] Verdu S, Verdu J. A new kinetic model for polypropylene thermal oxidation
at moderate temperatures. Macromolecules 1997;30:2262.
[88] Verdu J. Oxidative ageing of polymers. Hoboken: ISTE Ltd and John Wiley &
Sons Inc.; 2012.
[89] El-Mazry C, Correc O, Colin X. A new kinetic model for predicting polyamide
6-6 hydrolysis and its mechanical embrittlement. Polym Degrad Stab
2012;97:1049.
[90] Celina M, Gillen KT, Wise J, Clough RL. Anomalous aging phenomena in a
crosslinked polyolefin cable insulation. Radiat Phys Chem 1996;48:613.
[91] Celina M, Gillen KT, Clough RL. Inverse temperature and annealing phe-
nomena during degradation of crosslinked polyolefins. Polym Degrad Stab
1998;61:231.
[92] Celina MC, Gillen KT, Lindgren ER. Nuclear power plant cable materials:
review of qualification and currently available aging data for margin as-
sessments in cable performance. Sandia National Laboratories; 2013. p. 136.
SAND 2013-2388.
[93] Gillen KT, Clough RL. Predictive aging results in radiation environments.
Radiat Phys Chem 1993;41:803.
[94] Wise J, Gillen KT, Clough RL. Time development of diffusion-limited oxida-
tion profiles in a radiation environment. Radiat Phys Chem 1997;49:565.
[95] Ding SY, Ling MTK, Khare AR, Woo L. Durability of a reference material over
diverse conditions. Thermochim Acta 2000;357e358:147.
[96] Woo L, Khare AR, Sandford CL, Ling MTK, Ding SY. Relevance of high tem-
perature oxidative stability testing to long term polymer durability. J Therm
Anal Calorim 2001;64:539.
[97] Fearon PK, Bigger SW, Billingham NC. DSC combined with chem-
iluminescence for studying polymer oxidation. J Therm Anal Calorim
2004;76:75.
[98] Viebke J, Gedde UW. Antioxidant diffusion in polyethylene hot-water pipes.
Polym Eng Sci 1997;37:896.
[99] Geertz G, Bruell R, Wieser J, Maria R, Wenzel M, Engelsing K, et al. Stabiliser
diffusion in long-term pressure tested polypropylene pipes analysed by IR
microscopy. Polym Degrad Stab 2009;94:1092.
[100] Wong W-K, Cheng S, Li CY, Ahmad I, Cairncross R, Hsuan YG. Depletion
mechanism of antioxidants in MDPE-clay nanocomposites under thermal
aging. Polym Degrad Stab 2012;97:192.
Please cite this article in press as: Celina MC, Review of polymer oxidation and its relationship with materials performance and lifetime
prediction, Polymer Degradation and Stability (2013), http://dx.doi.org/10.1016/j.polymdegradstab.2013.06.024
11
[101] Celina M, Gillen KT. Oxygen permeability measurements on elastomers at
temperatures up to 225  C. Macromolecules 2005;38:2754.
[102] Scheirs J, Bigger SW, Billingham NC. A review of oxygen uptake techniques
for measuring polyolefin oxidation. Polym Test 1995;14:211.
[103] Gillen KT, Celina M, Clough RL, Wise J. Extrapolation of accelerated aging
data e Arrhenius or erroneous? Trends Polym Sci 1997;5:250.
[104] Ozawa T. A new method of analyzing thermogravimetric data. Bull Chem Soc
Jpn 1965;38:1881.
[105] Flynn JH, Wall LA. A quick, direct method for the determination of activa-
tion energy from thermogravimetric data. J Polym Sci Part B Polym Lett
1966;4:323.
[106] Friedman HL. Kinetics of thermal degradation of char-forming plastics from
thermogravimetry. Application to a phenolic resin. J Polym Sci 1964;6:183.
[107] Coats AW, Redfern JP. Kinetic parameters from thermogravimetric data. Nat
(London, UK) 1964;201:68.
[108] Vyazovkin S. Evaluation of activation energy of thermally stimulated solid-
state reactions under arbitrary variation of temperature. J Comput Chem
1997;18:393.
[109] Celina M, George GA, Lacey DJ, Billingham NC. Chemiluminescence imaging
of the oxidation of polypropylene. Polym Degrad Stab 1995;47:311.
[110] Lacey DJ, Dudler V. Chemiluminescence from polypropylene. Part 1. Imaging
thermal oxidation of unstabilized film. Polym Degrad Stab 1996;51:101.
[111] Blakey I, Goss B, George G. Chemiluminescence as a probe of polymer
oxidation. Aust J Chem 2006;59:485.
[112] Blakey I, Billingham N, George GA. Use of 9,10-diphenylanthracene as a
contrast agent in chemiluminescence imaging: the observation of spreading
of oxidative degradation in thin polypropylene films. Polym Degrad Stab
2007;92:2102.
[113] Celina M, Clough RL, Jones GD. Initiation of polymer degradation via transfer
of infectious species. Polym Degrad Stab 2006;91:1036.
[114] Butler CH, Whitmore PM. Measurement of peroxides in the volatile degra-
dation products of polypropylene photooxidation. Polym Degrad Stab
2013;98:471.
[115] Micallef AS, Blinco JP, George GA, Reid DA, Rizzardo E, Thang SH, et al. The
application of a novel profluorescent nitroxide to monitor thermo-oxidative
degradation of polypropylene. Polym Degrad Stab 2005;89:427.
[116] Blinco JP, Keddie DJ, Wade T, Barker PJ, George GA, Bottle SE. Profluorescent
nitroxides: sensors and stabilizers of radical-mediated oxidative damage.
Polym Degrad Stab 2008;93:1613.
[117] Colwell JM, Walker JR, Blinco JP, Micallef AS, George GA, Bottle SE. Pro-
fluorescent nitroxides: thermo-oxidation sensors for stabilised poly-
propylene. Polym Degrad Stab 2010;95:2101.
[118] Moghaddam L, Blinco JP, Colwell JM, Halley PJ, Bottle SE, Fredericks PM, et al.
Investigation of polypropylene degradation during melt processing using a
profluorescent nitroxide probe: a laboratory-scale study. Polym Degrad Stab
2011;96:455.
[119] Sumino K-I, Taniike T, Terano M, George GA. Thermal and photooxidative
degradation behaviors of poly(propylene)/SiO2 nanocomposites with various
polymer morphologies. Macromol React Eng 2008;2:135.
[120] Wang DL, Watson SS, Sung L-P, Tseng IH, Bouis CJ, Fernando R. Effect of TiO2
pigment type on the UV degradation of filled coatings. J Coat Technol Res
2011;8:19.
[121] George G, Johnstone P, Warner J, Halley P, Laycock B, Gardi S, et al.
Degradable polymeric films comprising polymer and prodegradant. Appli-
cation: WO. Australia: CRC for Polymer Ltd; 2009.
[122] Nikolic M, Gauthier E, George K, Cash G, de Jonge MD, Howard DL, et al.
Antagonism between transition metal pro-oxidants in polyethylene films.
Polym Degrad Stab 2012;97:1178.