pubs.acs.

org/JPCB Article

Constitutive Model of Radiation Aging Effects in Filled Silicone

Elastomers under Strain
A. Maiti,* W. Small,§ M. P. Kroonblawd,§ J. P. Lewicki, N. Goldman, T. S. Wilson, and A. P. Saab

Cite This: J. Phys. Chem. B 2021, 125, 10047−10057 Read Online

ACCESS
Downloaded via BEIJING COMPUTATIONAL SCI RSRCH CTR on November 9, 2021 at 02:09:12 (UTC).

Metrics & More Article Recommendations

See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

ABSTRACT: Filled silicone elastomers, an essential component in many technological

applications, are often subjected to controlled or unintended radiation for a variety of reasons.
Radiation exposure can lead to permanent mechanical and structural changes in the material,
which is manifested as altered mechanical response, and in some cases, a permanent set. For
unfilled elastomers, network theories developed and refined over decades can explain these
effects in terms of chain-scission and cross-link formation and a hypothesis involving
independent networks formed at different strain levels of the material. Here, we expose a filled
silicone rubber to gamma radiation while being under finite elongational strain and show that
the observed mechanical and structural changes can be quantitatively modeled within the same
theoretical framework developed for unfilled elastomers as long as nuances associated with the
Mullins effect are accounted for in a consistent manner. In this work, we employ Ogden’s incompressible hyperelastic model within
the framework of Tobolsky’s two-network scheme to describe the observed permanent set and mechanical modulus changes as a
function of radiation dosage. In the process, we conclude that gamma radiation induces both direct cross-linking at chain crossings
(H-links) and main-chain-scission followed by cross-linking (Y-links). We provide an estimate of the ratio of chain-scission to cross-
linking rates, which is in reasonable agreement with previous experimental estimate from Charlesby−Pinner analysis. We use density
functional theory (DFT)-based quantum mechanical calculations to explore the stability of −Si and −SiO radicals that form upon a
radiation-induced chain-scission event, which sheds light on the relative rates of Y-linking and H-linking processes.

■ INTRODUCTION
Filled and cross-linked elastomers and foams are materials of
tremendous technological importance1−3 with applications
spanning support cushions and pads, coatings, adhesives,
interconnects, acoustic/seismic-isolation, thermal/electrical
insulation, and bioimplants.4−10 Among the diverse choice of
polymers, silicones constitute a special class, possessing extreme
thermal and electrical stability and a high degree of
biocompatibility, which makes them ideal for applications
such as artificial organs/implants and biomedical devices.11−14
Given that in most applications, silicone materials are
expected to possess long lifetimes (years to decades), it is
important to develop age-aware models that account for the
interaction of the rubber or foam component with various
environmental stresses like temperature, humidity, oxidative
degradation, and radiation. The latter, which could be in the
form of high-energy X-rays, gamma rays, or an electron beam,
often arises in intended exposure, e.g., to perform sterilization of
a bioimplant component prior to placement within the subject’s
body. Radiation, especially penetrating gamma rays, is known to
induce permanent chemical changes in rubber, filled or unfilled,
which can lead to the scission of main polymer chains and the
formation of new cross-links.15
It is particularly interesting to explore the effect of cross-
linking and main-chain-scission on the structural and mechan-
ical response properties of elastomeric networks. Such a study
not only guides the construction of a radiation-age-aware model
but also puts to test several important network theories and
concepts developed over decades and provides useful insights
into the chemical effects of radiation at the molecular level.
Most of the fundamental theories have been developed for
unfilled rubber networks. The earliest important result states
that the mechanical response of rubber is primarily entropic in
nature, with the increase in free energy arising out of loss in
configurational entropy16−18 when a networked rubber speci-
men is deformed from its equilibrium state. A landmark result of
this analysis is the fact that if the cross-links (i.e., polymer
segments between two junctions) are assumed to be Gaussian
chains, and each junction is assumed to move affinely under
deformation, then the shear modulus μ of the rubber is
proportional to its number-density ν of cross-links (per unit
volume) in the form μ = νkBT, where kB is the Boltzmann
constant and T is the absolute temperature.
Received: June 5, 2021
Revised: August 6, 2021
Published: August 27, 2021

© 2021 American Chemical Society https://doi.org/10.1021/acs.jpcb.1c04958

10047 J. Phys. Chem. B 2021, 125, 10047−10057
The Journal of Physical Chemistry B
There have been many refinements of the above picture,
including (but not limited to) accounting for the effects of
junction fluctuations,17,18 contribution of physical entangle-
ments,19,20 non-Gaussian behavior at a larger deformation,21−26
and the effect of fillers27,28 that are used for enhancing the
mechanical modulus. For lightly cross-linked (or un-cross-
linked but entangled) materials, there is a complex time-
dependent response known as viscoelasticity. Theories for this
are still undergoing refinement,26,29−32 although some basic
concepts involving reptation tube dynamics were established in
the 1970s33 and more fully developed in the 1980s.34
Given that incorporating all the above theories to account for
the observed structural and mechanical response of radiation-
aged, filled rubber is a tall order, a few simplifying
approximations are needed to generate a practical, working
model. The first (well-justified) approximation for our system is
that the rubber is cross-linked enough that viscoelasticity can be
neglected, and the resulting (energy-conserving) response is
hyperelastic. This means that any stress−strain response is
governed by a strain energy functional.35−37
Ideally, we would like to employ a hyperelastic model that
quantitatively describes the observed stress−strain behavior of
unaged and radiation-aged rubber. However, as we will see
below, the picture gets complicated for filled rubber because of
the significant presence of the Mullins effect. 38,39 To
consistently extract a history-independent hyperelastic stress−
strain response for all samples, we focus on the last loading cycle
only up to strain levels beyond which the curve bends upward
due to prior maximum-strain history. Once we focus on this part
of the curve, we show that the response (for both aged and
unaged rubber) can be described well by the commonly
employed hyperelastic model due to Ogden. In fact, for our
material of interest, we show that the stress−strain response can
be modeled by just a single parameter, which can be directly
interpreted as the shear modulus of the material.
We note that the present work is intended as a more careful
theoretical treatment compared to our previous attempts to
analyze similar data.40,41 The previous work, although a useful
first step, employed the simple neo-Hookean model,16 which is
justified only to describe very small-strain behavior. The present
work accounts for the Mullins effect in a consistent manner,
employs a more accurate response model (i.e., Ogden), carefully
considers structural relaxation effects due to main-chain-scission
(Fricker’s transfer function), provides a range of estimates of the
ratio of Y-link versus H-link formation rates, and delves deeper
into the radiation-induced chemistry, both at the network-
topology level and at the molecular level.
■
EXPERIMENTAL PROCEDURE
Rectangular samples of commercial rubber TR-55 manufactured
by Dow Corning were used for the radiation aging experiments.
The material is primarily polydimethylsiloxane (PDMS) and
incorporates roughly 30 wt % of silica filler. Thin samples (∼1
mm thick) were stretched and held at a fixed elongation strain
while exposed to γ radiation from a Co-60 source (1.2 MeV,
∼0.1 Mrad/h dose rate), which provided a controllable
degradation pathway. To avoid oxidative degradation, the
samples were irradiated under vacuum until the desired
cumulative dosage (D) was reached. In the following, we
express elongational strain in terms of the stretch ratio λ = 1 + ε,
where ε is the engineering strain, and we use λa to denote the
stretch ratio corresponding to the constant “aging strain” the
system is under when exposed to radiation.
10048
pubs.acs.org/JPCB Article
Following exposure to controlled duration (and therefore
controlled cumulative dosages) of radiation, each sample was
removed from the irradiation chamber, released from the λa
strain, and allowed to relax under ambient conditions for 24 h.
The relaxed samples were then subjected to measurements of
the new equilibrium length, called the recovered length λs. Two
sets of experiments were performed: (i) in the first, samples
under λa= 1.20, 1.47, 1.67, and 1.84 were subjected to
cumulative dosages of 1, 3, 7, and 17 Mrad of radiation and
(ii) in the second, samples under λa= 1.67, 2.00, 2.33, and 2.67
were subjected to cumulative dosages of 5, 10, and 17 Mrad of
radiation. After several weeks of further equilibration, tensile
stress−strain analysis was carried out on a few of the specimens,
including specimens that were not stretched (i.e., λa= 1.00)
when exposed to radiation. The stress−strain analysis was
performed on rectangular specimens (∼3 mm wide by ∼40 mm
long) cut from the irradiated samples, as well as a pristine unaged
specimen, using an Instron 5565 dual-column electromechanical
test system with an initial grip separation of ∼20 mm and a
stretching rate of 20 mm/min.
■ MULLINS EFFECT
One of the challenges in modeling stress−strain response in a
filled rubber with a hyperelastic strain function is the omni-
presence of the so-called Mullins effect, as illustrated in Figure
1(a). A few common characteristics of the Mullins effect
Figure 1. (a) Engineering stress vs strain for unaged TR-55 silicone
rubber displaying strong Mullins effect; (b) modeling the last loading
curve from the left figure along with the best fit using the N = 1 Ogden
incompressible hyperelastic model. Only the part of the curve
independent of strain history (i.e., prior to upward swing) is modeled.
include38,39,42,43 (1) significant softening upon the first
unloading cycle; (2) an increase in softening with an increase
in the maximum strain in the first cycle; (3) subsequent
unloading showing much less softening as long as the previous
maximum strain is not exceeded; and (4) a small but noticeable
permanent set at the end of the first unloading curve, which
remains stable through subsequent cycles. The permanent set
typically increases upon unloading from an increased maximum
strain, although in some cases, it can recover after a long resting
time.44 There is an extensive literature study that attempts to tie
the Mullins effect to various causes, primarily involving quasi-
irreversible structural changes in filler configuration, chain-
slippage, and network damage.42 In this work, we do not delve
into such details but rather attempt to focus on the last loading
cycle (see Figure 1(b)), which represents a quasi-steady-state
stress−strain response largely free of energy loss due to
hysteresis or viscoelasticity.
All radiation exposure experiments in this work were
performed after the material underwent four or five loading−
unloading cycles; the aim was to minimize any coupling of
Mullins cycles with radiation effects.
https://doi.org/10.1021/acs.jpcb.1c04958
J. Phys. Chem. B 2021, 125, 10047−10057
The Journal of Physical Chemistry B
■
OGDEN HYPERELASTIC AND TOBOLSKY
2-NETWORK MODELS
As mentioned in the foregoing, we focus on the hyperelastic
response of the material; that is, any hysteretic loss due to
viscoelasticity is neglected in our analysis. With this caveat, we
chose a commonly employed hyperelastic strain energy model
due to Ogden, given by eq 1:45,46
N 2μj
α α α
W (λ1 , λ 2 , λ3) = ∑ 2
[λ1 j + λ 2 j + λ3 j − 3]
j = 1 αj
− p(λ1λ 2λ3 − 1) (1)
where λ1, λ2, and λ3 are the eigenvalues of the deformation tensor
F, and p is a Lagrange multiplier that imposes the constraint that
the material is incompressible, i.e., detF = λ1λ2λ3 = 1. The
parameters {μj} are related to the shear modulus μ by the
N
relation μ = ∑ j = 1 μj , and the parameters {αj} are exponents
that can be positive or negative.
If we focus on the stress−strain curve of the last loading cycle,
we find that in all cases, it is described well by just a single term,
i.e., N = 1 Ogden model (eq 2):
W1net(λ1, λ 2 , λ3| μ , α)
2μ
= 2 [λ1α + λ 2α + λ3α − 3] − p(λ1λ 2λ3 − 1)
α (2)
where μ is the shear modulus, and α is a scalar exponent. The
true (Cauchy) stress in the principal direction i is given by σi =
λi∂W1net/∂λi. Under a uniaxial engineering stretch εa along
direction 1, we have λ1 = 1 + εa and λ2 = λ3 = λ−1/2
1 (assuming an
isotropic and incompressible material). Using the boundary
condition σ2 = σ3 = 0, we obtain the following expression for the
Cauchy stress along direction 1:
2μ α
σ1net = (λ1 − λ1−α /2)
α (3)
where index subscript “1” in σ (to indicate the principal stretch
direction) has been omitted for simplicity. When eq 3 is fitted to
the last loading cycle of Figure 1, we find that for all our samples,
the best fit occurs for α → 0 in which limit eq 3 simplifies to
σ1net = 3μlnλ1 (4)
The corresponding engineering stress P is obtained simply as
σ1net/λ1, i.e.,
P1net = 3μλ1−1ln λ1 (5)
In our experiments, the material is subjected to a cumulative
radiation dose while being under an elongation engineering
strain εa (i.e., stretch ratio λa = 1 + εa). Any radiation-induced
cross-link formation under this condition happens in equili-
brium with the strained state, while main-chain-scission can
happen both to the original cross-links formed under zero strain
and to the new cross-links formed in the strained state. To model
such effects, we need a hyperelastic strain energy expression for a
two-network system, i.e., the original network (formed under
zero strain) and the induced network (formed under stretch
ratio λa). For an unfilled network, Tobolsky argued that the net
effect of these two networks is linear and additive.47 Assuming
Tobolsky’s picture also holds true for a filled system, we express
the net effect of these two networks as the following strain
energy expression (eq 6):
10049
pubs.acs.org/JPCB Article
W2net(λ1, λ 2 , λ3 , λa|μ1 , α1 , μ2 , α2)
2μ 2μ
= 21 (λ1α1 + λ 2α1 + λ3α1 − 3) + 22
α1 α2
{(λ1λa−1)α2 + (λ 2λa1/2)α2 + (λ3λa1/2)α2 − 3}
− p(λ1λ 2λ3 − 1) (6)
By using the same argument that led to eqs 3 and 4, it is
straightforward to obtain the following expression for Cauchy
and Engineering stress (in direction 1) for the two-network
system (in the limit α1, α2 → 0):
σ2net = 3μ1ln λ1 + 3μ2 ln(λ1/λa); P2net = σ2net /λ1 (7)
A primary effect analyzed in this work is an irreversible change in
length, as defined by the residual strain εs, or equivalently the
permanent stretch ratio λs = 1 + εs. The quantity λs is simply the
value of λ1 at which σ2net = 0, which from eq 7 is obtained as
follows:
λs = 1 + εs = λa1 − p (8)
with
p = μ1 /(μ1 + μ2 ) (9)
Since the shear moduli μ1 and μ2 are functions of the dosage D
(as made clearer in the sections below), the residual strain εs is a
function of both the aging strain and the cumulative radiation
dosage. Occasionally, for brevity, we will make loose reference to
λs and εs as the “permanent set.”
■ RADIATION AGING UNDER ZERO STRAIN
Figure 2(a) displays the results for stress−strain response upon
radiation exposure under zero strain for four different
Figure 2. (a) Modeling stress−strain response (last loading cycle) of
TR-55 rubber aged under various radiation dosages at zero strain; (b)
corresponding shear modulus as a function of radiation dosage,
showing an approximately linear behavior. The vertical error bars in
figure (b) indicate standard error due to sample-to-sample variation.
cumulative dosages. Each curve represents the stress response
averaged over two samples. Figure 2(b) plots the corresponding
mean shear modulus μ by fitting eq 5 to the stress−strain
response. To eliminate dependence on strain history (in this
case, memory of the previously attained maximum strain), only
the part of the response curve prior to upward swing was
modeled. From Figure 2(b), we see that the shear modulus
increases linearly as a function of dosage D, i.e.,
μ(D) = μ0 + kD (10)
From the data in Figure 2(b), we obtained a mean value of 0.57
MPa for the intercept μ0, and a mean value of 0.074 MPa/Mrad
for the slope k. From uncertainty analysis, the 90% confidence
https://doi.org/10.1021/acs.jpcb.1c04958
J. Phys. Chem. B 2021, 125, 10047−10057
The Journal of Physical Chemistry B
interval for the marginal distribution of the intercept μ0 was
(0.49,0.65), while that for the slope k was (0.069, 0.079), with a
negative correlation (−0.70) between the two quantities.
We note that all the analysis in this work is based on
experiments performed with our specific Co-60 radiation source
that has a fixed dosage rate (see the Experimental Procedure
section). As long as the γ photon absorption events are spatially
and temporally separated, one expects the overall model and the
conclusions to be largely independent of the dosage rate and
depend only on the cumulative dosage D. However, since such
absorption events involve a complex cascade of chemical
processes involving multielectronic excitations, a more extensive
experimental study with several different radiation sources is
necessary to probe possible dose-rate dependence of our model
parameters.
■
RADIATION AGING UNDER FINITE STRAIN:
PERMANENT SET
Figure 3 (symbols) displays results for the observed residual
strain εs for various values of dosage D and aging strain εa (note:
Figure 3. Residual strain (variable εs in the text, expressed here in
percent), a measure of the permanent set, in radiation-aged TR-55
samples while being under different aging strain levels (εa, indicated in
the legend in percent). Points indicate sample-averaged measured
values. The average experimental standard error is 2.2% (not shown for
clarity). Solid curves are the best model fit under the consideration of
both H-linking and Y-linking processes.
aging stretch ratio λa = 1 + εa). To model such data with eqs 8
and 9, one needs to first determine the evolution of the shear
stresses of the original and the induced networks, i.e., μ1 and μ2,
as a function of radiation dosage D. To make progress on this
front, we borrow polymer network concepts developed for
unfilled rubber networks and make the following implicit
assumptions about our filled system:
(1) Despite all the structural complexities in a filled elastomer,
there is a linear relationship between the shear modulus
(μ) and the chemical cross-link density (ν) in the bulk
elastomer for the pristine cross-linked polymer (i.e., prior
to radiation exposure), i.e., μ = CkBTν, where C is a
material-dependent dimensionless constant that accounts
for effects like filler enhancement, departure from affine
behavior, entanglements, and so on.
(2) Cross-linking effects due to radiation is in general
different at the filler surface than in the bulk polymer,
and one expects a different proportionality constant for
the radiation-induced change in the modulus, i.e., Δμ =
C′kBTΔν, where Δμ is the radiation-induced change in
the shear modulus of the material and Δν is the radiation-
induced change in cross-link density in the bulk
elastomer. The ratio β = C/C′ is important in expressing
radiation-induced bulk cross-link density in terms of
10050
pubs.acs.org/JPCB Article
pristine cross-link density. It is used in eq 13 below, where
we develop the two-network constitutive model. In
particular, the case C′ > C (i.e., β < 1) can be interpreted
as an increase in the filler-enhancement factor induced by
radiation (relative to the filler-enhancement factor in
pristine cross-linked rubber), which could be attributed to
much stronger radical activation in silica fillers,48,49 as
compared to radiation-induced cross-linking within the
bulk elastomer.
(3) When the material is radiation-aged while under a state of
strain, the resulting material behaves like an additive
combination of two independent networks, following the
model developed by Tobolsky47 for unfilled networks. In
fact, we have already used this assumption to derive eqs 8
and 9.
(4) If radiation induces main-chain-scission of the original
cross-links, there is a feedback effect in which a fraction of
the network 2 cross-links generated prior to that scission
event is effectively transferred to the original network (i.e.,
network 1). Such analysis was first carried out by Flory50
for a Gaussian-chain network and by Fricker51 for a
phantom network of Gaussian chains. Assuming a
phantom network model, in the following analysis, we
use Fricker’s formula to model such behavior, although
alternative approaches have recently been suggested in
the literature.52
■ RADIATION-INDUCED SCISSION AND
CROSS-LINKING RATES
High-energy electromagnetic or electron-beam radiation is well
known to produce permanent chemical changes in silicone
polymers.15,53−59 Two of the chemical changes particularly
relevant to change in network topology and mechanical moduli
are scission and cross-linking processes. Scission processes could
involve either main-chain-scission (i.e., breaking of Si−O
bonds) or side-chain-scission (i.e., breaking of Si−C or C−H
bonds). Any scission process typically produces radicals, which,
in the absence of radical scavengers, could persist long enough
until another active site with a radical is found and a new
chemical bond is formed by radical-pair annihilation. Using this
picture as a good approximation to reality (i.e., ignoring other
more complex processes, some of which are briefly mentioned in
the Insights from QM Modeling Section), let us introduce rates
of two broad classes of processes: (i) direct cross-linking of
active side chains belonging to two different load-bearing chains,
with a rate of kxl processes per radiation dosage, and (ii) main-
chain-scission followed by the bonding of a fraction (say, δ) of
the resulting active −Si and/or −SiO radicals with an active side-
chain site of a different load-bearing chain, with a rate of ksci
scission processes per radiation dosage. Process (i) produces a
fourfold-coordinated junction, which we refer to below as an H-
link, while process (ii) yields a threefold-coordinated junction,
which we refer to below as a Y-link.
Mechanical changes due to only H-link formation are simpler
to analyze, and we do that first in the following section.
■
CONSIDERING ONLY H-LINK PROCESSES
In this section, we analyze how the formation of H-links only
(i.e., no main-chain-scission) affects the mechanical properties
of the system and the permanent set. In the analysis below, we
make use of our heuristic understanding that shear moduli of the
https://doi.org/10.1021/acs.jpcb.1c04958
J. Phys. Chem. B 2021, 125, 10047−10057
The Journal of Physical Chemistry B
two networks (μ1 and μ2) are proportional to the corresponding
cross-link densities, even for the filled rubber material.
Let us consider the material radiation-exposed under an aging
strain εa (i.e., stretch ratio λa = 1 + εa). Prior to radiation
exposure, i.e., at D = 0, we have μ1 = μ0 and μ2 = 0 (where μ0 is
the intercept in eq 10).
Let us now consider an H-linking process. In this process, two
cross-links become four cross-links. The question is the strained
state status of the four new links. To address this point, we use
the important result from bead-spring simulations,60,61 which
showed that such cross-linking processes under a state of strain
do not alter the cross-link density of the original network. This
simply translates to the result that if the two chains involved in
the H-linking process belong to the original network (i.e.,
network 1), then after cross-linking two of the resulting four
chains effectively remain in the strained state of the original
network, while the other two chains belong to the induced
network (i.e., network 2). The above argument can be easily
extended to the case where either one or both the chain
segments involved in H-linking belong to network 2. In all such
cases, we see that H-linking leads to a net increase of two chains
in network 2, without any change in the number of chains (and
hence the chain density) of network 1. Thus, we can express
radiation-induced changes in shear moduli μ1 and μ2 by the
following equations:
dμ1 /dD = 0; dμ2 /dD = 2kxl (11)
which can be easily solved as follows (eq 12):
μ1(D) = μ0 ; μ2 (D) = 2kxlD
(12)
If this model is correct, then to have consistency with Figure 2,
we need to have 2kxl = k, where k is the slope in eq 10. From the
90% CI values (0.069 and 0.079) for k, this implies that kxl
should be ∼0.037 MPa/Mrad with an error margin of 0.0025.
However, we find that the value of kxl that fits the observed
permanent set data (symbols in Figure 3) is ∼0.052 MPa/Mrad,
which is inconsistent with the confidence interval obtained from
the mechanical modulus results of Figure 2. Additionally, even if
one uses the higher value of kxl, the fit is not as accurate as the
solid curves in Figure 3, which is obtained from a model that
incorporates both Y-linking and H-linking processes, as
described in the section below.
■
CONSIDERING BOTH H-LINKING AND Y-LINKING
PROCESSES
In the presence of both main-chain-scission (followed by Y-link
formation) and H-link processes, eq 11 needs to be modified.
Let kxl and ksci be H-linking and main-chain-scission rates,
respectively. Let r be the fraction of cross-links belonging to
network 1, and φ(r) be the fraction of monomers belonging to
network 1. From our model assumptions, we can relate r to the
shear moduli μ1 and μ2 by the following equation:
r = {μ0 + β(μ1 − μ0 )}/{μ0 + β(μ1 + μ2 − μ0 )} (13)
The above equation follows from our model that the pristine
(i.e., preradiation) cross-link density is proportional to μ0/C,
and the radiation-induced change in network 1 cross-link density
is proportional to (μ1 − μ0)/C′, etc., and note that β = C/C′, by
definition. Let us also assume that following main-chain-scission,
only a fraction δ of the broken links forms active (i.e., load-
bearing) cross-links through the Y-linking process. Then,
without the consideration of any feedback effect due to main-
10051
pubs.acs.org/JPCB Article
chain-scission of network 1 cross-links, eq 11 gets modified to
the following:
dμ1 /dD = −ksciφ(r ); dμ2 /dD = 4δksci − ksci(1 − φ(r ))
+ 2kxl = ksciφ(r ) + k (14)
with
k = (4δ − 1)ksci + 2kxl (15)
Equation 14 is derived based on the assumption that each
monomer in the system has the same probability of being the site
of main-chain-scission.
However, as mentioned earlier, there is an additional effect
that needs to be accounted for. This involves the fact that when
cross-linking is followed by scission of some of the network 1
linkages, then a fraction f of network 2 cross-links effectively
becomes part of network 1.50,51,62 When this effective transfer of
cross-links is taken into account, eq 14 gets modified as follows:
dμ1 /dD = −ksciφ(r ) + fμ2 ; dμ2 /dD = ksciφ(r ) + k − fμ2
(16)
For a phantom network of Gaussian chains, Fricker showed51,62
that f is approximately given by the ratio of the number of
network 1 cross-links that get scissioned and the sum of two
quantities, i.e., (i) number of network 2 cross-links that exist
prior to the scission events and (ii) number of network 1 cross-
links of the pristine system (i.e., prior to radiation exposure). In
the present case, the number of network 1 cross-links that
undergo scission between cumulative dosages D and D + dD is
proportional to βksciφ(r)dD, while the denominator is propor-
tional to μ0 + βμ2. Thus, according to the Fricker formula, in eq
16, we identify the fraction f as follows:
f ≈ βk sci φ(r )/(μ0 + βμ2 )
(17)
The initial conditions to eq 16 are
μ1(D = 0) = μ0 ; μ2 (D = 0) = 0 (18)
where μ0 (in eq 18) and k (in eqs 14, 15, and 16) are the
intercept and slope, as previously defined in eq 10 with mean
values of 0.57 MPa and 0.074 MPa/Mrad, respectively.
Determining the monomer fraction of network φ(r) in eqs 14,
16, and 17 is difficult. To build a practical working model, we use
a simple approximation based on the following argument. At
small times (i.e., small cumulative dosage D), we have r ≈ 1 and
φ(r) ≈ 1; i.e., almost all scissioned chains belong to network 1.
However, since the average length of the induced chains
(following scission/cross-linking processes) is smaller than that
of the original cross-links, for any D > 0, one expects r < 1 and
φ(r) > r. There is no simple way to derive a closed-form
expression for φ(r) as it likely depends on the molecular weight
distribution of the material at D = 0. To avoid such complexities,
we assume a tractable approximation φ(r) ≈ rγ, where 0 < γ < 1;
this is a simple formula that satisfies the desired relations φ(1) =
1, and φ(r) > r for 0 < r < 1, and φ(r) → 0 for r → 0. Assuming
that initially (i.e., at low dosages D ≪ 1) the average lengths of
network 2 cross-links are roughly half of those of the network 1
cross-links, we expect γ to be around 0.5.
If we use φ(r) = rγ with γ = 0.5 in eq 16, numerically integrate
to solve for μ1 and μ2 as a function of D, and then use eqs 8 and 9
to compute the residual strain εs, the best agreement with the
experimental data in Figure 3 happens for ksci ≈ 0.016 MPa/
Mrad and β ≈ 0.50. The corresponding computed results for εs
https://doi.org/10.1021/acs.jpcb.1c04958
J. Phys. Chem. B 2021, 125, 10047−10057
The Journal of Physical Chemistry B
as a function of the cumulative dosage D (for various values of
aging strain levels) are indicated by the solid lines in Figure 3
showing good agreement between the observed and model
results.
We point out that the above value of ksci corresponds to a
scission-to-cross-linking ratio of ksci/k = 0.22. This ratio is in
good semi-quantitative agreement with experimental esti-
mates63 from Charlesby−Pinner analysis64 upon radiation
aging of the same filled rubber system (i.e., TR-55) as used in
the present study. The optimized value of β ≈ 0.50 indicates that
radiation-induced filler-enhancement effect is a factor of β−1 ≈ 2
larger than the filler-enhancement effect in pristine TR-55.
We have experimented with other values of γ, e.g., between 0.5
and 0.7, and found the optimized values of ksci and β and the
quality of agreement in Figure 3 (in terms of RMSE deviation)
to be largely unchanged. This provides confidence in the
robustness of the model and the optimized parameters.
Additionally, the model results for the permanent set are
independent of the parameter δ (the fraction of scissioned chain
ends that form load-bearing Y-links, as introduced just above eq
14) as long as it is chosen within reasonable physical bounds
(e.g., between 0.5 and 1.0), but it depends on the choice of k
(which we chose as 0.074 MPa/Mrad, following the fit to Figure
2(b), see discussion following eq 10). On the other hand, from
eq 15, the parameter kxl depends on the choice of δ. Additionally,
the ratio of the number of Y-linking to H-linking events (rYH),
which according to the present model is rYH = 2δksci/kxl, also has
a strong dependence on δ as well. As we argue in the Insights
from QM Modeling Section below, we expect δ to be between
0.5 and 1. For δ= 0.5, 0.75, and 1.0, the Y-linking/H-linking ratio
rYH is 0.55, 1.1, and 2.5, respectively (using ksci = 0.016 and k =
0.074 MPa/Mrad). Thus, without knowing the parameter δ, we
cannot reliably predict the Y-linking/H-linking ratio rYH. It is
likely that δ varies significantly depending on the polymer
system, filler distribution, cross-linking history prior to
irradiation, and many other factors. This could explain the
lack of agreement in the literature on the relative importance of
the Y-linking and H-linking rates.56,58
Finally, using the above model, we can compute the stress−
strain response of the radiation-aged system relative to the new
equilibrium. To this end, we define the longitudinal stretch ratio
(in direction 1) in the aged system as λ′1 = λ1/λs, where λ1 is the
stretch ratio in the unaged system and λs is the permanent
stretch ratio defined previously. Then, eq 7 can be rewritten as
eq 19:
′ = 3μ1ln(λ1′λs) + 3μ2 ln(λ1′λs /λa) = 3(μ1 + μ2 )ln λ1′
σ2net
(19)
where the last step was simplified by using the result that σ2net as
defined in eq 7 is 0 at λ1 = λs. In other words, the aged material is
predicted to exhibit exactly the same stress−strain behavior as
the unaged material with a modulus of μ1 + μ2. Adding the two
subequations of eq 16 and integrating, we obtain eq 20:
μ1 + μ2 = μ0 + kD
(20)
which is identical to eq 10. In other words, the stress−strain
behavior and shear modulus of the radiation-aged system is
predicted to be independent of the aging strain λa (with strain in
the aged sample being defined relative to its new equilibrium
length that incorporates the permanent set). Such predicted
behavior largely agrees with our measured stress−strain curves
of the aged samples, within the uncertainty of sample-to-sample
variations.
10052
■
pubs.acs.org/JPCB Article
INSIGHTS FROM QM MODELING
From the discussion in the previous section, it is clear that the
ratio rYH of Y-linking to H-linking rates depends strongly on the
fraction (δ) of scissioned ends that form stable, load-bearing Y-
links. To estimate δ, one needs to understand the stability of
radicals formed following a scission event. In this section, we
explore such radical stability by computing the energetics of
possible pathways using DFT-based first-principles QM
calculations65,66 with the CP2K code.67
It is understood that there are number of complex reaction
channels in irradiated PDMS58 and silica fillers,48,49 including
ones not directly tied to cross-linking. In this section, we explore
some representative pathways using clusters of atoms in the gas
phase. The DFT calculations reported below employed the
generalized gradient approximation functional due to Perdew−
Burke−Ernzerhof68 with Grimme D3 dispersion corrections69
and the core electrons represented by the Goedecker−Teter−
Hutter pseudopotentials.70 CP2K is a mixed Gaussian/plane-
wave code; we used a double-ζ plus polarization Gaussian basis
set (DZVP) and a 200 Ha plane wave energy cutoff for
expansion of the charge density. The electronic structure was
evaluated with spin polarization at the Γ-point only. Reactions
considered here involved migration of a single radical, so the
spin multiplicity was taken as a doublet. We used the
nonperiodic Martyna−Tuckerman Poisson electrostatic solv-
er71 and a cubic simulation cell with a side length of 30 Å, which
is more than twice the maximum cluster size.
Chemical reaction pathways were sampled using a con-
strained “bond scan” optimization approach in which the
separation distance between two chosen atoms was incremented
through set values. This was implemented in CP2K by applying
a large harmonic bias potential to a single collective variable
defined as the separation distance between the two chosen
atoms. Each step in the scan was optimized subject to the biased
collective variable using the BFGS algorithm with a force
tolerance of 10−3 Ha/Bohr. These scans were performed starting
from a configuration optimized without constraints and were
iterated on a uniformly spaced grid with a resolution of 0.1 Å.
The initial configuration for each subsequent iteration of the
scan was taken as the optimized configuration from the previous
iteration. It should be noted that while this approach effectively
explores the potential energy surface, it does not necessarily fully
optimize local minima or transition states that would be
obtained without constraints.
Radiation damage in PDMS is understood to result from the
formation of radicals (see Figure 4) that arise from either main-
chain-scission (species A and B) or excitation of the side groups
(species C and D).15 Recombination reactions involving
biradical pairs are energetically favorable and result in the
formation of H or Y cross-links. More specifically, recombina-
tion reactions involving main-chain-scission products A and B
can lead to either a healed chain (e.g., A + B) or the formation of
Y-links (e.g., A + C). In contrast, reactions involving side group
radical species C and D can result in both types of cross-links. A
Y-link is formed if C or D reacts with A or B, whereas an H-link
forms if C or D reacts with another radical of type C or D. If
biradical reactions are the only chemical pathways considered,
then the only cross-link-producing outcome of main-chain-
scission is Y-link formation. However, given that the timescale
for cross-linking is governed by chain dynamics in which two
radicals must migrate to reach each other and react, it is
important to address the stability and lifetime of −Si (species A)
https://doi.org/10.1021/acs.jpcb.1c04958
J. Phys. Chem. B 2021, 125, 10047−10057
The Journal of Physical Chemistry B
Figure 4. Initial radicals formed in irradiated PDMS and example Y-
and H-type cross-links formed from biradical reactions. Species A and B
are formed as a pair from scission of the main chain, and species C and
D are formed in separate events from excitation of the side groups.
Atom centers with an initial radical are rendered in color.
and/or −SiO radicals (species B) formed upon chain-scission,
especially given the high reactivity (and the resulting instability)
of most radicals.
To explore the stability of −Si and −SiO radicals, we used
DFT calculations to explore reaction energetics of such radicals
with neighboring side groups in bulk PDMS. Three cases were
considered, including (a) reaction between a radical silicon
terminus (species A) and a methyl side group forming a cross-
link; (b) reaction between a radical silicon terminus and the
hydrogen on a methyl side group leading to an H-transfer; and
(c) reaction between a radical oxygen terminus (species B) and a
methyl side group. Constrained bond scan optimizations were
used to sample these reactions and are shown in Figure 5, panels
(a), (b), and (c), respectively. Distinct chemical configurations
sampled in these reactions are shown to the right of each energy
scan plot in the figure.
In the first case, there is a significant barrier to reaction
between a silicon radical and a methyl side group to form a cross-
link. The constrained optimization predicts an energetic penalty
of ∼35 kcal/mol to induce a reaction from state 1 to 2. This
reaction produces a Y-link and a free hydrogen radical that are
energetically unfavorable relative to the starting point (ΔE ∼
+15 kcal/mol). The back-reaction barrier is also quite large
(∼20 kcal/mol), indicating that state 2 would be kinetically
stable once formed. Owing to both the large barrier and lack of
energetic drivers for reaction 1 → 2, it is reasonable to expect
that the radical silicon terminus will not directly form a cross-link
unless it interacts with a more reactive site, such as another
radical located on the polymer or an unsaturated bond found in a
vinyl or phenyl side group.
The second energy profile shows an alternative reaction
channel with a radical silicon terminus (1 → 3). Compared to
case (a), there is a more modest energy penalty (15 kcal/mol) to
radical/H transfer resulting in a stable Si−H terminus and
methyl radical side group. Similar to case (a), there is no energy
driver for this reaction (ΔE ∼ +12 kcal/mol), but the back-
reaction barrier is very small (3 kcal/mol), indicating that state 3
10053
pubs.acs.org/JPCB Article
is unstable even if formed. Based on the potential energy
landscape, this radical/H transfer reaction is plausible but not
likely to be a dominate pathway.
Thus, according to the above analysis, the Si terminus is
expected to be stable until it forms a Y-link with another side-
group monoradical.
The energy landscape explored in the third case is more
complicated than the previous two. There is a very small barrier
(5 kcal/mol) to proton transfer between a radical oxygen
terminus and methyl side group (4 → 5). This reaction produces
a silanol (Si−OH) terminus and methyl radical side group
identical to species C in Figure 4. Compared to state 4, there is
also a modest energetic driver (ΔE ∼ −5 kcal/mol) to reach
state 5. Thus, the radical oxygen terminus is substantially less
stable than the radical silicon terminus and can react to transfer
the radical to a nearby side group. The back-reaction barrier is
also comparatively modest (9 kcal/mol), indicating the
possibility for a non-negligible equilibrium concentration of
state 4. However, this equilibrium will be sensitive to entropic
contributions arising from chain motions that cause the silanol
and methyl radical side group to diffuse apart, which would favor
state 5. Following the formation of a silanol, constrained
optimization uncovers another local minimum that corresponds
to a conformational change (5 → 5′) before a substantial 30+
kcal/mol barrier to further forward the chemical reaction. While
state 6 is a local minimum, it is unlikely that reaction 5 → 6
would occur through thermally activated chemistry.
Large barriers to hydrogen abstraction by a radical silicon
terminus are consistent with radiolysis experiments coupled to
electron spin resonance spectroscopy, which found that silicon
radicals persist under cryogenic conditions.72 Recent cryogenic
radiolysis experiments sensitive to methyl radicals found an
increase in the concentration of radical side group species C that
was correlated with the decay of free methyl radical (·CH3)
concentration,73 although these changes in concentration were
not equivalent. One possible contributor to this trend is that free
methyl radicals may scavenge hydrogen from nearby side groups
producing species C.73 The buildup of species C concentration
is also consistent with the proposed monoradical reactions in
Figure 5(b,c). Formation of Si−H bonds is consistent with
infrared and NMR measurements of irradiated PDMS
materials.53,58
Silanols are known to form in irradiated PDMS in the
presence of large concentrations of strong hydrogen donors such
as alkylthiols.56 Small quantities of silanol products were found
in hexamethyldisiloxane irradiated under an inert N2 atmos-
phere.74 Water is a common environmental antagonist known to
be coupled to silanol production. Experimental NMR measure-
ments found that thermal and radiative aging of PDMS in the
presence of 17O-labled water led to 17O uptake in silanol groups
and the main chain.75 Recent DFT calculations indicate that a
radical oxygen terminus can readily abstract hydrogen from
water, while the barrier for proton transfer between water and a
radical silicon terminus is substantially higher.76 Water can
couple with radiation effects to produce silanols77 and is
predicted to destabilize highly strained chains leading to
hydrolytic scission.78 Many mechanisms generate silanol
functionalities under a range of conditions in PDMS, which
complicates its use as a signature to conclusively validate the
proposed mechanism through experiments.
To summarize, the principal consequence for PDMS
radiolysis cross-linking chemistry inferred from Figure 5 is that
main-chain-scission can promote both Y- and H-link formation.
https://doi.org/10.1021/acs.jpcb.1c04958
J. Phys. Chem. B 2021, 125, 10047−10057
The Journal of Physical Chemistry B pubs.acs.org/JPCB Article

Figure 5. Constrained optimization scans of the potential energy surface for monoradical reactions between (a) radical silicon terminus and methyl
side group, (b) radical silicon terminus and hydrogen on methyl side group, and (c) radical oxygen terminus and methyl side group. Distinct chemical
configurations sampled in the optimization scans are indicated in the figure panels and shown at the bottom. Biased atoms are rendered in color.

This effect arises because radical side groups such as species C
are necessary intermediates for both types of cross-linking.
While the radical silicon terminus produced from scission is
quite stable and would primarily engage in Y-linking through the
biradical reaction, the radical oxygen terminus supports reaction
pathways for both Y- and H-linking through radical migration to
side groups. The value of the fraction δ (see the beginning of this
section for definition) is governed primarily by the relative rates
of the two types of linking processes resulting from the −SiO
terminus. For instance, if these rates are nearly equal, and if we
assume almost all cross-links to be load-bearing, we obtain δ≈
10054
0.75. In general, we expect δ to be between 0.5 and 1.0. A
comparatively large δ is consistent with the NMR radiolysis
experiments by Hill et al.58 that found Y-linking to be the
dominant route to cross-linking in PDMS over more well-
established H-linking mechanisms.53,55,74 Additional environ-
mental factors, such as large gas concentrations, may have other
synergistic interactions with the radical silicon and oxygen
termini. High-throughput computational screening of reactivity
under many conditions using graph descriptors offers a
promising route to elucidate the complex and often inseparable
https://doi.org/10.1021/acs.jpcb.1c04958
J. Phys. Chem. B 2021, 125, 10047−10057
The Journal of Physical Chemistry B
couplings of different atomic-scale aging mechanisms in PDMS
and other polymer materials.79
Before leaving this section, we would like to note that the
present analysis is relevant to radiation-induced scission/cross-
linking reactions in the bulk of the elastomer only. This is in
spirit of the previous sections where fillers are not explicitly
considered but assumed to influence mechanical properties and
radiation-induced net cross-linking through multiplicative
constants C, C′. A more accurate analysis of radiation-induced
chemical changes in filled rubber should take account of radical
formation and cross-linking chemistry at the filler surface. Given
the significant presence of SiOH termini at the silica filler
surfaces, such analysis should be similar to Figure 5(c), although
effects of other molecules, e.g., H2O at the filler surface need to
be explicitly modeled. In future studies, we have plans to explore
such reactions in more detail.
■ SUMMARY AND DISCUSSION
In this work, we show that radiation-induced permanent set and
mechanical property changes of a filled silicone elastomer can be
explained in terms of known concepts from network theory
originally developed for unfilled rubber materials. The
applicability of network theory concepts to filled systems hinges
on several assumptions, the most important of which involves a
direct proportionality between the shear modulus of the filled
system and the cross-link density of the underlying network, and
a direct proportionality (with a different proportionality
constant) between the radiation-induced modulus increase
and the increase in the cross-link density in the bulk elastomeric
network. We use such assumptions in conjunction with
Tobolsky’s two-network model to quantitatively interpret the
observed permanent set when filled-rubber samples are exposed
to radiation while being under an applied tensile strain. For a
consistent treatment of stress−strain response under the
omnipresent Mullins effect, we used the last loading cycle and
fitted it using Ogden’s incompressible hyperelastic model. We
found that for all our samples, the Ogden N = 1 model with zero
α exponent led to the best fit, which led to a simple
interpretation of the single Ogden parameter as the shear
modulus of the material. When used in conjunction with the
Tobolsky two-network model, such a framework was able to
quantitatively explain the observed permanent set data. Finally,
through DFT calculations, we explored the stability of radicals at
the −Si and −SiO termini that are generated following a chain-
scission event and its effects on the relative rates of Y-linking
versus H-linking processes. We found that the -Si terminus
radicals are relatively stable and highly probable to create Y-links
(if not regenerating the prescissioned chain through recombi-
nation with the −SiO terminus), while the −SiO terminus
radicals can potentially form either Y-links or get passivated
through hydrogen abstraction (in which case, an H-link is likely
to form).
Our analysis led to several interesting results, including the
following: (1) the radiation-induced modulus enhancement and
the permanent set data cannot be explained by H-linking
processes alone, and incorporating contributions of Y-linking
processes, which follow main-chain-scission, becomes neces-
sary; (2) the feedback effect due to structural relaxation
following main-chain-scission of network 1 cross-links can be
effectively modeled by Fricker’s transfer function, originally
developed for an unfilled phantom network of Gaussian chains;
(3) the predicted scission to cross-linking ratio that best fits the
observed stress−strain and permanent set data (i.e., Figure 2(b)
10055
pubs.acs.org/JPCB Article
and Figure 3) is ∼22%, which is similar to earlier experimental
estimates using Charlseby−Pinner analysis on a similar
system;63 (4) it is difficult to make a definite statement about
the relative rates of Y-linking versus H-linking events: depending
upon processing history and material details, the ratio of these
two types of cross-linking can be more than or less than 1, as we
illustrated for various reasonable values of the parameter δ; (5)
the filler-enhancement factor of the mechanical modulus in the
presence of radiation is a factor of ∼2 stronger as compared to
that in the absence of radiation; and (6) the −Si termini of
scissioned chains should lead to primarily Y-links, while the
−SiO termini likely lead to significant fractions of both Y-links
and −SiOH passivated ends.
While the above results are interesting, we realize that some of
the assumptions made are perhaps too simplistic. For instance,
in the network theory for unfilled rubber, it is well known that
physical entanglements lead to a different stress−strain response
than chemical cross-links,18−20 while in this work, we lump such
effect (along with filler enhancement) within a contribution
proportional to the chemical cross-link density. In fact, to
address this, we have explored the possibility of splitting the μ0
term into contributions from entanglement and chemical cross-
links present prior to radiation exposure and found that the best
fits to the data occur when the contribution due to entanglement
is zero. This is likely not due to the absence of entanglements but
rather reflects inconsistency of the isolated contribution of
entanglements to the hyperelastic stress−strain response of the
filled system, at least for the data set analyzed here. Another
aspect of the network theory neglected in this work is
consideration of different functionalities of the cross-link
junctions. For instance, one could borrow ideas from the results
of a phantom network model and weight the contribution of
each cross-link segment by a factor ≈(1 − ϕ−1 −1
1 − ϕ2 ) (where ϕ1
and ϕ2 are functionalities of junctions at the two ends).
However, that significantly complicates the analysis and
warrants knowing the starting functionality distribution of the
pristine network prior to radiation exposure. In addition, it is not
clear how a weighting factor such as the above gets modified in
the presence of fillers. During our analysis, we also made the tacit
assumption that the ratio of non-load-bearing to load-bearing
cross-link formation does not change with cumulative dosage.
Given that the average end-to-end distance of a network chain
between its two end-junctions is proportional to the square root
of the number of monomers in the chain, a strain of 2.67 is not
enough to strain any chemical bond in the siloxane system and
thus does not alter the chemical reactivity of the system.
However, at much larger macroscopic strain levels, where a
significant number of network chains are stretched to near
maximum-stretchable limits, one needs to account for stretches
in the chemical bonds in the system, which can result in a higher
susceptibility to radiation-induced chemistry.
Finally, we would like to note that the rates of cross-linking
(both Y and H) are kinetically governed by chain-diffusion rates
as well as factors such as the density of vinyl side groups,
peroxide additives, and various radical scavengers present in the
system. A more complete model needs to account for all the
above factors at an appropriate phenomenological level.
■
AUTHOR INFORMATION
Corresponding Author
A. Maiti − Lawrence Livermore National Laboratory,
Livermore, California 94550, United States; orcid.org/
0000-0003-0831-0700; Email: amaiti@llnl.gov
https://doi.org/10.1021/acs.jpcb.1c04958
J. Phys. Chem. B 2021, 125, 10047−10057
The Journal of Physical Chemistry B
Authors
W. Small − Lawrence Livermore National Laboratory,
Livermore, California 94550, United States
M. P. Kroonblawd − Lawrence Livermore National Laboratory,
Livermore, California 94550, United States; orcid.org/
0000-0002-5009-5998
J. P. Lewicki − Lawrence Livermore National Laboratory,
Livermore, California 94550, United States; orcid.org/
0000-0002-2467-702X
N. Goldman − Lawrence Livermore National Laboratory,
Livermore, California 94550, United States; orcid.org/
0000-0003-3052-2128
T. S. Wilson − Lawrence Livermore National Laboratory,
Livermore, California 94550, United States
A. P. Saab − Lawrence Livermore National Laboratory,
Livermore, California 94550, United States
Complete contact information is available at:
https://pubs.acs.org/10.1021/acs.jpcb.1c04958
Author Contributions
§ strain time−dependent behavior of elastomers. J. Mech. Phys. Solids
W.S and M.P.K. contributed equally to this paper.
Notes
The authors declare no competing financial interest.
■ ACKNOWLEDGMENTS
This work was performed under the auspices of the U.S.
Department of Energy by Lawrence Livermore National
Laboratory under Contract DE-AC52-07NA27344. M.P.K.
thanks I-Feng W. Kuo for technical assistance with the CP2K
code.
■
REFERENCES
(1) Mark, J. E.; Erman, B.; Eirich, F. R. Science and Technology of
Rubber; Academic Press: New York, 2005.
(2) Harper, C. A. Handbook of Plastics, Elastomers, & Composites;
MacGraw-Hill: New York, 2002.
(3) Ciesielski, A. An Introduction to Rubber Technology; Rapra
Technology Ltd.: UK, 1999.
(4) Owen, M. J. Siloxane polymers, 1st ed. Englewood Cliffs, NJ:
Prentice Hall, 1993.
(5) Morent, R.; De Geyter, N.; Axisa, F.; De Smet, N.; Gengembre, L.;
De Leersnyder, E.; Leys, C.; Vanfleteren, J.; Rymarczyk-Machal, M.;
Schacht, E.; et al. Adhesion enhancement by a dielectric barrier
discharge of PDMS used for flexible and stretchable electronics. J. Phys.
D Appl. Phys. 2007, 40, 7392−7401.
(6) Andersson, L. H. U.; Hjertberg, T. Silicone Elastomers for
electronic applications. I. Analyses of the noncross-linked fractions. J.
Appl. Polym. Sci. 2003, 88, 2073−2081.
(7) Wu, K. H.; Chao, C. M.; Yeh, T. F.; Chang, T. C. Thermal Stability
and Corrosion Resistance of Polysiloxane Coatings on 2024-T3 and
6061-T6 Aluminum Alloy. Surf. Coat. Technol. 2007, 201, 5782−5788.
(8) Buyl, F. Silicone Sealants and Structural Adhesives. Int. J. Adhes.
2001, 21, 411−422.
(9) Ikada, Y. Surface modification of polymers for medical
applications. Biomaterials 1994, 15, 725−736.
(10) Abbasi, F.; Mirzadeh, H.; Katbab, A. Bulk and surface
modification of silicone rubber for biomedical applications. Polym.
Int. 2002, 51, 882−888.
(11) Colas, A.; Curtis, J. Silicone biomaterials: history, chemistry and
major applications of silicones, printed in Biomaterials Science: An
Introduction to Materials in Medicine; 2nd edn., Elsevier Publications,
1996 80−86.
(12) Arkles, B. Look what you can make out of silicones (Biomedical
applications of silicones). Chem 1983, 13, 542−555.
10056
pubs.acs.org/JPCB Article
(13) Brook, M.A. Silicon in a biological environment. In Silicon in
Organic, Organometallic and Polymer Chemistry; Brook, M. A. (ed), John
Wiley and Sons: New York, NY, 2000, 459−479.
(14) Van Dyke, M. A.; Clarson, S. J.; Arshady, R. Silicone biomaterials,
in Introduction to Biomaterials; Arshady, R. (ed), Citus Books, London,
UK, 2003, 109−135.
(15) Palsule, A. S.; Clarson, S. J.; Widenhouse, C. W. Gamma
Irradiation of Silicones. J. Inorg. Organomet. Polym. 2008, 18, 207−221.
(16) Treloar, L. R. G. The Physics of Rubber Elasticity; Clarendon
Press: Oxford, 1975.
(17) Boyd, R. H.; Phillips, P. J. The Science of Polymer Molecules;
Cambridge University Press, 1996.
(18) Rubinstein, M.; Colby, R. H. Polymer Physics; Oxford University
Press, 2003.
(19) Edwards, S. F.; Vilgis, T. The effect of entanglements in rubber
elasticity. Polymer 1986, 27, 483−492.
(20) Rubinstein, M.; Panyukov, S. Elasticity of Polymer Networks.
Macromolecules 2002, 35, 6670−6686.
(21) Arruda, E. M.; Boyce, M. C. A three-dimensional constitutive
model for the large stretch behavior of rubber elastic materials. J. Mech.
Phys. Solids 1993, 41, 389−412.
(22) Bergström, J.; Boyce, M. C. Constitutive modeling of the large
1998, 46, 931−954.
(23) Itskov, M.; Ehret, A. E.; Dargazany, R. A Full-Network Rubber
Elasticity Model based on Analytical Integration. Math. Mech. Solids
2010, 15, 655−671.
(24) Davidson, J. D.; Goulbourne, N. C. A nonaffine network model
for elastomers undergoing finite deformations. J. Mech. Phys. Solids
2013, 61, 1784−1797.
(25) Ehret, A. E. On a molecular statistical basis for Ogden’s model of
rubber elasticity. J. Mech. Phys. Solids 2015, 78, 249−268.
(26) Li, Y.; Tang, S.; Kröger, M.; Liu, W. K. Molecular simulation
guided constitutive modeling on finite strain viscoelasticity of
elastomers. J. Mech. Phys. Solids 2016, 88, 204−226.
(27) Wang, M.-J. Effect of Polymer-Filler and Filler-Filler Interactions
on Dynamic Properties of Filled Vulcanizates. Rubber Chem. Tech.
1998, 71, 520−589.
(28) Vilgis, T. A.; Heinrich, G.; Klüppel, M. Reinforcement of Polymer
Nano-composites; Cambridge University Press: Cambridge, UK, 2009).
(29) McLeish, T. Tube theory of entangled polymer dynamics. Adv.
Phys. 2002, 51, 1379−1527.
(30) Boyd, R. H.; Smith, G. D. Polymer Dynamics and Relaxation;
Cambridge University Press, 2007.
(31) Tang, S.; Greene, M. S.; Liu, W. K. Two-scale mechanism-based
theory of nonlinear viscoelasticity. J. Mech. Phys. Solids 2012, 60, 199−
226.
(32) Dealy, J. M.; Read, D. J.; Larson, R. G. Structure and Rheology of
Molten Polymers; Hanser Publishers: Munich, 2018.
(33) de Gennes, P. G. Reptation of a Polymer Chain in the Presence of
Fixed Obstacles. J. Chem. Phys. 1971, 55, 572−579.
(34) Doi, M.; Edwards, S. F. The Theory of Polymer Dynamics;
Clarendon Press: Oxford, 1986.
(35) Ogden, R. W. Nonlinear Elastic Deformations; Dover
Publications: New York, 1984.
(36) Marckmann, G.; Verron, E. Comparison of Hyperelastic Models
for Rubber-Like Materials. Rubber Chem. Tech. 2006, 79, 835−858.
(37) Holzapfel, G. A. Nonlinear Solid Mechanics; John Wiley and Sons:
Chichester, UK, 2000.
(38) Mullins, L. Effect of Stretching on the Properties of Rubber.
Rubber Chem. Technol. 1948, 21, 281−300.
(39) Mullins, L. Softening of Rubber by Deformation. Rubber Chem.
Technol. 1969, 42, 339−362.
(40) Maiti, A.; Weisgraber, T. H.; Dinh, L. N.; Gee, R. H.; Wilson, T.
S.; Chinn, S. C.; Maxwell, R. S. Controlled manipulation of elastomers
with radiation: Insights from multiquantum nuclear-magnetic-
resonance data and mechanical measurements. Phys. Rev. E 2011, 83,
No. 031802.
https://doi.org/10.1021/acs.jpcb.1c04958
J. Phys. Chem. B 2021, 125, 10047−10057
The Journal of Physical Chemistry B
(41) Maiti, A.; Weisgraber, T. H.; Gee, R. H.; Small, W.; Alviso, C. T.;
Chinn, S. C.; Maxwell, R. S. Radiation-induced mechanical property
changes in filled rubber. Phys. Rev. E 2011, 83, No. 062801.
(42) Diani, J.; Fayolle, B.; Gilormini, P. A review on the Mullins effect.
Eur. Polym. J. 2009, 45, 601−612.
(43) Maiti, A.; Small, W.; Gee, R. H.; Weisgraber, T. H.; Chinn, S. C.;
Wilson, R. S.; Maxwell, R. S. Mullins effect in a filled elastomer under
uniaxial tension. Phys. Rev. E 2014, 89, No. 012602.
(44) Itskov, M.; Haberstroh, E.; Ehret, A. E.; Voehringer, M. C.
Experimental observation of the deformation induced anisotropy of the
Mullins effect in rubber. KGK, Kautsch. Gummi Kunstst. 2006, 59, 93−
96.
(45) Ogden, R. W. Large deformation isotropic elasticity − on the
correlation of theory and experiment for incompressible rubberlike
solids. Proc. R. Soc. London A 1972, 326, 565−584.
(46) Wikipedia page: https://en.wikipedia.org/wiki/Ogden_
(hyperelastic_model).
(47) Tobolsky, A. V. Properties and Structure of Polymers; Wiley: New
York, 1960.
(48) Dondi, D.; Buttafava, A.; Zeffiro, A.; Conzatti, L.; Faucitano, A.
The role of silica in radiation induced grafting and crosslinking of silica/
elastomers blends. Polymer 2012, 53, 4579−4584.
(49) Dondi, D.; Buttafava, A.; Zeffiro, A.; Bracco, S.; Sozzani, P.;
Faucitano, A. Reaction Mechanisms in Irradiated, Precipitated, and
Mesoporous Silica. J. Phys. Chem. A 2013, 117, 3304−3318.
(50) Flory, P. J. Elasticity of polymer networks cross-linked in states of
strain. Trans. Faraday Soc. 1960, 56, 722−743.
(51) Fricker, H. S. The effects on rubber elasticity of the addition and
scission of cross-links under strain. Proc. R. Soc. London A 1973, 335,
267−287.
(52) Yu, Y.; Yan, S.; Fang, Y.; He, Q.; Wang, H.; Qiu, Y.; Wan, Q.
Theoretical Model of Polymer Network Chain Formation under Strain.
ACS Omega 2018, 3, 15615−15622.
(53) Charlesby, A. Changes in silicone polymeric fluids due to high-
energy radiation. Proc. Roy. Soc. A 1955, 230, 120−135.
(54) Bueche, A. M. An investigation of the theory of rubber elasticity
using irradiated polydimethylsiloxanes. J. Polym. Sci. A 1956, 19, 297−
306.
(55) Miller, A. A. Radiation chemistry of Polydimethylsiloxane: I.
Cross-linking and gas yields. J. Am. Chem. Soc. 1960, 82, 3519−3523.
(56) Miller, A. A. Radiation chemistry of Polydimethylsiloxane: II.
Effect of Additives. J. Am. Chem. Soc. 1961, 83, 31−36.
(57) Hill, D. J. T.; Preston, C. M. L.; Whittaker, A. K.; Hunt, S. M. The
radiation chemistry of poly(dimethyl-siloxane). Macromol. Symp. 2000,
195, 95−102.
(58) Hill, D. J. T.; Preston, C. M. L.; Whittaker, A. K. NMR study of
the gamma radiolysis of poly(dimethyl siloxane) under vacuum at 303
K. Polymer 2002, 43, 1051−1059.
(59) Hill, D. J. T.; Preston, C. M. L.; Salisbury, D. J.; Whittaker, A. K.
Molecular weight changes and scission and cross-linking in poly
(dimethyl siloxane) on gamma radiolysis. Rad. Phys. Chem. 2001, 62,
11−17.
(60) Rottach, D. R.; Curro, J. G.; Grest, G. S.; Thomson, A. P. Effect of
strain history on stress and permanent set in cross-linking networks: A
molecular dynamics study. Macromolecules 2004, 37, 5468−5473.
(61) Rottach, D. R.; Curro, J. G.; Budzien, J.; Grest, G. S.; Svaneborg,
C.; Everaers, R. Permanent set of cross-linking networks: Comparison
of theory with molecular dynamics simulations. Macromolecules 2006,
39, 5521−5530.
(62) Rottach, D. R.; Curro, J. G.; Budzien, J.; Grest, G. S.; Svaneborg,
C.; Everaers, R. Molecular dynamics simulations of polymer networks
undergoing sequential cross-linking and scission reactions. Macro-
molecules 2007, 40, 131−139.
(63) Maxwell, R. S.; Chinn, S. C.; Alviso, C. T.; Harvey, C. A.; Giuliani,
J. R.; Wilson, T. S.; Cohenour, R. Quantification of radiation induced
cross-linking in a commercial, toughened silicone rubber, TR55 by 1H
MQ-NMR. Polym. Degrad. Stab. 2009, 94, 456−464.
10057
pubs.acs.org/JPCB Article
(64) Charlesby, A.; Pinner, S. H. Analysis of the solubility behaviour of
irradiated polyethylene and other polymers. Proc. R. Soc. A 1959, 249,
367−386.
(65) Hohenberg, P.; Kohn, W. Inhomogeneous electron gas. Phys.
Ther. Rev. 1964, 136, B864−B871.
(66) Kohn, W.; Sham, L. Self-Consistent Equations Including
Exchange and Correlation Effects. Phys. Rev. 1965, 140, A1133−A1138.
(67) Kühne, T. D.; Iannuzzi, M.; Del Ben, M.; Rybkin, V. V.; Seewald,
P.; Stein, F.; Laino, T.; Khaliullin, R. Z.; Schütt, O.; Schiffmann, F.; et al.
CP2K: An electronic structure and molecular dynamics software
package - Quickstep: Efficient and accurate electronic structure
calculations. J. Chem. Phys. 2020, 152, 194103.
(68) Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient
Approximation Made Simple. Phys. Rev. Lett. 1996, 77, 3865−3868.
(69) Grimmea, S.; Antony, J.; Ehrlich, S.; Krieg, H. A consistent and
accurate ab initio parametrization of density functional dispersion
correction (DFT-D) for the 94 elements H-Pu. J. Chem. Phys. 2010,
132, 154104.
(70) Goedecker, S.; Teter, M.; Hutter, J. Separable dual-space
Gaussian pseudopotentials. Phys. Rev. B 1996, 54, 1703−1710.
(71) Martyna, G. J.; Tuckerman, M. E. A reciprocal space based
method for treating long range interactions in ab initio and force-field-
based calculations in clusters. J. Chem. Phys. 1999, 110, 2810−2821.
(72) Ormerod, M. G.; Charlesby, A. The radiation chemistry of some
polysiloxanes: An electron spin resonance study. Polymer 1963, 4, 459−
470.
(73) Pankratova, L. N.; Saenko, E. V.; Bugaenko, V. L.; Makhlyarchuk,
V. V. Kinetics of decay of methyl radicals in irradiated dimethylsila-
nediol and polymethylsiloxanes. Mendeleev Commun. 2011, 21, 157−
159.
(74) Dewhurst, H. A.; St. Pierre, L. E. Radiation chemistry of
hexamethyldisiloxane, a polydimethyl-siloxane model. J. Phys. Chem.
1960, 64, 1063−1065.
(75) Patel, M.; Morrell, P.; Cunningham, J.; Khan, N.; Maxwell, R. S.;
Chinn, S. C. Complexities associated with moisture in foamed
polysiloxane composites. Polym. Degrad. Stab. 2008, 93, 513−519.
(76) Wang, P.-C.; Yang, N.; Liu, D.; Qin, Z.-M.; An, Y.; Chen, H.-B.
Coupling effects of gamma irradiation and absorbed moisture on
silicone foam. Mater. Des. 2020, 195, No. 108998.
(77) Kroonblawd, M. P.; Goldman, N.; Lewicki, J. P. Chemical
Degradation Pathways in Siloxane Polymers Following Phenyl
Excitations. J. Phys. Chem. B 2018, 122, 12201−12210.
(78) Kroonblawd, M. P.; Goldman, N.; Lewicki, J. P. Anisotropic
Hydrolysis Susceptibility in Deformed Polydimethylsiloxanes. J. Phys.
Chem. B 2019, 123, 7926−7935.
(79) Kroonblawd, M. P.; Goldman, N.; Maiti, A.; Lewicki, J. P. A
Quantum-Based Approach to Predict Primary Radiation Damage in
Polymeric Networks. J. Chem. Theory Comput. 2021, 17, 463−473.
https://doi.org/10.1021/acs.jpcb.1c04958
J. Phys. Chem. B 2021, 125, 10047−10057