Relaxation time of a polymer glass stretched at very large strains

HAL Id: hal-03063708

https://hal.archives-ouvertes.fr/hal-03063708v2

Submitted on 5 Jan 2021

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

Relaxation time of a polymer glass stretched at very

large strains

R. Sahli, J. Hem, C. Crauste-Thibierge, Franck Clement, D. Long, S. Ciliberto

To cite this version:

R. Sahli, J. Hem, C. Crauste-Thibierge, Franck Clement, D. Long, et al.. Relaxation time of a polymer

glass stretched at very large strains. Physical Review Materials, American Physical Society, 2020, 4

(3), �10.1103/PhysRevMaterials.4.035601�. �hal-03063708v2�

R. Sahli1,∗_{, J. Hem}1_{, C. Crauste-Thibierge}1_{, F. Clément}2_{, D.R. Long}2_{, S. Ciliberto}1_, 1 Univ Lyon, Ens de Lyon, Univ Claude Bernard, CNRS,

Laboratoire de Physique, UMR 5672, F-69342 Lyon, France and 2 Laboratoire Polymères et Matériaux Avancés, CNRS/Solvay, UMR 5268,

87 avenue des Frères Perret, 69192 Saint Fons Cedex, France

The polymer relaxation dynamic of a sample, stretched up to the stress hardening regime, is measured, at room temperature, as a function of the strain λ for a wide range of the strain rate ˙γ, by an original dielectric spectroscopy set up. The mechanical stress modies the shape of the dielectric spectra mainly because it aects the dominant polymer relaxation time τ, which depends on λ and is a decreasing function of ˙γ. The fastest dynamics is not reached at yield but in the softening regime. The dynamics slows down during the hardening, with a progressive increase of τ. A small inuence of ˙γ and λ on the dielectric strength cannot be excluded.

Mechanical and dynamical properties of polymers are intensively studied, due to their fundamental and

tech-nological importance [1]. When strained at a given strain

rate, beyond the elastic regime in which the stress is pro-portional to strain, glassy polymers exhibit a maximum in the stress-strain curves (yield point) at a strain of a

few percents [2] and the deformation becomes irreversible

(see Fig.1). At larger strains, depending on the history of

the sample [3], the stress drops (strain-softening regime)

before reaching a plateau corresponding to plastic ow. Strain-hardening may then occur at even larger strains, depending on the molecular weight and on the

cross-linking of the polymer [4]. The key new insights obtained

either by numerical simulations [57] or experiments [8

11] are that strain hardening appears to be controlled by

the same mechanisms that control plastic ow [1215].

However the microscopic mechanisms leading to such a

mechanical behavior are not fully understood [16, 17].

For example it is unclear to what extent the relaxation dynamics in polymer glasses is modied when the sam-ple is stretched into the plastic region (see for examsam-ple

refs.[1821]).

The purpose of this letter is to bring new insight into this problem by presenting the results of experiments in which we performed dielectric spectroscopy of polymer samples stretched till the strain hardening regime. Di-electric spectroscopy, allows the investigation of the dy-namics of relaxation processes by means of the

polariza-tion of molecular dipoles. It is directly sensitive to

poly-mer mobility and probes directly the segmental motion.

It can be used to quantify the mobile fraction of

poly-mers. Measuring the dielectric response of polymers in

situ had been pioneered by Venkataswamy et al. [22].

It is complementary to other techniques, such as

Nu-clear Magnetic Resonance [23] and the diusion of probe

molecules [2426], used to study the molecular dynamics

of polymers under stress. The dielectric spectroscopy has already been used in combination with mechanical defor-mation to study the dynamics in the amorphous phase of

polymer under stress [27,28]. The results of these

experi-ments were limited to the yield pointwhereasthe

experi-mental studies of the microscopic behavior and processes

FIG. 1. Stress evolution over a wide range of λ at several constant strain rates : 2.5 × 10−3

s−1 (black), 2.5 × 10−4s−1 (red), 2.5 × 10−5

s−1 (green) and 2.5 × 10−6s−1 (blue). Sev-eral identical specimens have been measured for each strain rate. Inset: tensile machine for dielectric measurements un-der stress : motor (A), load cell (B), linear transducer (C), sample fastening cylinders with the stretched polymer lm (D) and electrodes (E) connected to the dielectric spectrom-eter (not sketched).

during strain hardening are more scarce.

In this letter we present the results obtained by our original experimental apparatus which can measure with high accuracy the evolution of the Dielectric Spec-trum(DS) of a sample stretched, till the strain hardening regime, at dierent strain rates. Thus we can precisely

compare the DSsobtained as a function of stress at room

temperature with those (named DST) obtained as a

func-tion of temperature in an unstressed sample. This com-parison allows us to extract useful informations on the relaxation dynamics under stress at various strain rates. Our experimental results bring new important informa-tions because they clearly show an acceleration of the dynamics which reaches a maximum in the softening re-gions. Instead the molecular mobility slows down again during the strain hardening regimes.

2 Our experimental setup is composed by a home-made

dielectric spectrometer coupled with a tensile machine

for uniaxial stretching of lms [28, 29]. The scheme of

the experimental setup is presented in the inset of Fig.1.

More details are given in AnnexI.1.

The dielectric measurements are performed by conn-ing the polymer lm between two disc-shaped electrodes of 10 cm diameter. The good contact between the elec-trodes and the sample, during the whole experiment, is assured by an aqueous gel, which has very low electrical resistivity compared to the sample. We checked that the gel does not perturb the sample response because water

absorption in our samples is only 0.2%and by comparing

the response in the presence of the aqueous gel and in the

presence of mineral oil [29].

Dielectric spectroscopy allows the investigation of the dielectric response of a material as a function of the angu-lar frequency ω. It is expressed by the complex dielectric

permittivity or dielectric constant: ε(ω) = ε0_(ω)−iε00_(ω),

with ε0_{(ω)the real component of the dielectric constant}

which is related to the electric energy stored by the

sample, and ε00_{(ω) the imaginary component which}

in-dicates the energy losses. The loss tangent is tnδ(ω) =

tan δ(ω) = ε00_(ω)/ε0_{(ω), where δ is the phase shift}

be-tween the electric eld applied for measuring ε and the measured dielectric polarization.

We investigated an extruded lm MAKROPOL® _DE

1-1 000000 (from BAYER) based on Makrolon®

poly-carbonate (PC) with a Tg of about 150 ◦C. The

sam-ple sheets had a thickness of 125 µm. The tensile

ex-periments are performed at T = 25◦_{C much below T}

g.

We xed four dierent strain rates: 2.5 × 10−3 _s−1_,

2.5 × 10−4 s−1, 2.5 × 10−5 s−1 and 2.5 × 10−6s−1.

Typ-ical strain curves at the dierent strain rates are

plot-ted in Fig.1. Initially, the stress increases almost

lin-early up to its maximum (yield stress) which is reached at λ ' 1.08. The stress decreases after the yield strain softening regime for 1.09 < λ < 1.2. For λ > 1.2, the material is inside the strain hardening regime, where the force increases up to the lm crack.

1 . 0 1 . 2 1 . 4 1 . 6 1 . 8 1 . 0 0 0 1 . 0 0 5 1 . 0 1 0 1 . 0 1 5 1 . 0 2 0 1 . 0 1 . 2 1 . 4 1 . 6 1 . 8 0 5 1 0 1 5 2 0 ε 'n λ ( a ) ( b ) 0 1 0 2 0 3 0 4 0 5 0 6 0 H z 1 H z 7 H z 0 . 2 5 H z σ ( M P a ) 0 . 1 2 5 H z 6 0 H z 7 H z 1 H z 0 . 2 5 H z 0 . 1 2 5 H z T a n ( δ ) x 1 0 -3 λ 0 1 0 2 0 3 0 4 0 5 0 σ ( M P a ) FIG. 2. Dependence on λ of ε0

n (a) and tnδ (b) measured

at several frequencies during the stretching experiment at ˙γ = 2.5×10−4s−1. In both panels the solid lines (left hand or-dinate) represent ε0

n(a) and tnδ(b) as a function of λ, whereas

the dashed line (right hand ordinate) show the corresponding evolution of the stress, which reveals that the low frequency components of DSshave a maximum in the softening regime

of the polymer lm.

In this article we focus on the measurement at λ > 1.09

being the results at λ ≤ 1.08 already discussed in [28].

For each value of the strain rate the measurement of the dielectric and mechanical properties have been repeated on at least 3 samples, in order to check the reproducibility of the results. The maximum uctuations of the results observed in dierent samples is at most 11%. In order to compensate for the change of the thickness of the sample

we study ε0

n(f ) = ε0(f, λ)/ε0(400Hz, λ) because above

400Hz the only eect of the applied stress on the dielec-tric measurement is related to the change of the sample thickness induced by the large applied stress. The

qual-ity of this compensation can be checked looking at Figs.2

where we plot ε0

n(f, λ)and tnδ(f, λ) measured at various

frequencies as a function of λ. We clearly see that the ef-fect of the stretching on the DS decreases a lot by increas-ing the measurincreas-ing frequency and it almost disappears at

60Hz. This means that ε0_n cancels the dependenceof the

sample thicknessd(λ)and correctly estimates the varia-tion of ε induced by the strain. This gure also shows

that both ε0

n and tnδ reach the maximum in the

soft-ening regime then they decrease and remain constant in

the hardening regimes. Fig.3shows that the eect of the

strain on ε increases with ˙γ as already shown in ref.[28].

1 . 0 1 . 2 1 . 4 1 . 6 1 . 8 1 . 0 0 0 1 . 0 0 2 1 . 0 0 4 1 . 0 0 6 1 . 0 0 8 1 . 0 1 . 2 1 . 4 1 . 6 1 . 8 0 2 4 6 8 ε 'n λ ( a ) 0 1 0 2 0 3 0 4 0 5 0 σ ( M P a ) T a n ( δ ) x 1 0 -3 λ 0 1 0 2 0 3 0 4 0 5 0 σ ( M P a ) ( b )

FIG. 3. Normalized real part ε0

n(a) and tnδ (b) as a function

of λ measured at 7Hz for dierent strain rates : 2.5×10−3

s−1 (black triangles), 2.5 × 10−4

s−1 (red circles), 2.5 × 10−5s−1 (green squares) and 2.5 × 10−6

s−1(blue diamonds).

It is useful to study the evolution of the whole

DSsmeasured at ˙γ = 2.5 × 10−4s−1 for which the eect

of the strain on ε is very pronounced (see Fig.2) and at

the same time the measure lasts enough time to have a

good low frequency resolution of the DS. Figures4show

ε0n(f ) and tnδ(f) as a function of frequency measured

at various λ at ˙γ = 2.5 × 10−4_s−1_{. The DS}

srecorded

during the tensile test are compared to the DST,

i.e. the DS measured as a function of tempera-ture in the unstressed sample. This gure sum-marizes one the most important ndings of this investigation, that we will explain. First we no-tice that the low frequency part of DSshas roughly

the same amplitude of DSTmeasured at

temper-atures close to Tg. However the tnδ(f) of DSsand

of DSThave a very dierent dependence on f

indi-cating that a simple relationship between ˙γ and an eective temperature ( presented in several arti-cles [26, 28, 30, 31]) cannot be easily established

1 0 - 1 _{1 0} 0 _{1 0} 1 _{1 0} 2 _{1 0} 3 1 . 0 0 0 1 . 0 0 5 1 . 0 1 0 1 . 0 1 5 1 . 0 2 0 1 0 - 1 _{1 0} 0 _{1 0} 1 _{1 0} 2 _{1 0} 3 0 5 1 0 1 5 2 0 f ( H z ) 1 3 7 ° C

λ

ε

'

FIG. 4. The normalized real part ε0

n (a) and tnδ (b) as

a function of frequency measured, at T = 25◦_{C, during a}

stretching experiment at 2.5 × 10−4

s−1at dierent values of the strain : λ = 1.00 (equilibrium), λ = 1.08 (plastic yield), λ = 1.17(softening regime) and λ = 1.70 (hardening regime). In the background, continuous orange lines represent ε0

n (a)

and tnδ (b) measured at λ = 1 in the temperature range be-tween 137◦

C, (bottom orange curve) and 152◦C (top orange curve)at 1◦

Cincrement [28].

as we discuss later on. However in Figs.4,2 we no-tice several other important results. We clearly see that

in the stressed sample the maximum amplitude of ε0

n(f ) and tnδ at low frequencies is not reached at the yield but in the softening regime at λ = 1.17 and most importantly

the amplitude of DSsdecreases in the hardening regime

at λ = 1.70. This is a new and totally unexpected result, i.e. the eect of the strain on the dielectric constant is not monotonous.

In order to have a more detailed description of these experimental observations, we study the be-havior of tnδ(f, λ, ˙γ), which is not aected by any geometrical eect. We plot, in Fig.5, tnδ(f) versus frequency measured at λ = 1.17 at various ˙γ. We no-tice that the low frequency parts decrease as a power law

which can be tted by a power lawtnδ(ω) = 1/(ωτef f)β

where τef f and β are tting parameters and ω = 2πf.

We nd that β ' 0.5 ± 0.05 is independent of ˙γ

(con-tinuous lines in Fig.5a). The measured time scale τef f,

which is a function of ˙γ and λ, is plotted in Fig.5b).

Look-ing at this gure we see that with a very good accuracy

1 0- 2 1 0- 1 1 00 1 01 1 02 1 03 1 0- 3 1 0- 2 1 0- 6 1 0- 5 1 0- 4 1 0- 3 1 02 1 03 1 04 1 05 1 06 1 . 2 1 . 6 1 02 1 1 0 T a n ( δ ) f ( H z ) ( a ) . τeff (s ) γ( s - 1 ) l = 1 . 7 0 l = 1 . 3 0 l = 1 . 1 7 ( b ) a (λ ) λ

FIG. 5. (a) Dependence on frequency of tnδ, measured in the softening regime λ = 1.17, at dierent ˙γ: 2.5×10−3

s−1(black triangles), 2.5 × 10−4

s−1 (red circles), 2.5 × 10−5s−1 (green squares) and 2.5 × 10−6

s−1 (blue diamonds). The low fre-quency parts of each curve are tted by a power law (straight lines) tnδ = 1/(ωτef f)β The time scale τef fdepends on ˙γ but

the scaling power β = 0.5±0.05 is found to be constant within experimental errors. (b) The tted time scale τef f as a

func-tion of the strain rate ˙γ measured at several strains: λ = 1.17 (softening regime), λ = 1.30 and λ = 1.70 (hardening regime). The relaxation time follows the law τef f = a(λ)/ ˙γ (dashed

lines) with the coecient a(λ) plotted in the inset.

τef f = a(λ)/ ˙γwhere a(λ) has a minimum at λ ' 1.17 in

the stress softening regime.

We can consider the power law behavior of tnδ as the high frequency part of the Cole-Cole (Havriliak-Negami) model [32] for the α peak of the dielectric constant. Indeed, calling τ the characteristic time of the α peak, at frequencies ω >> 1/τ the Cole-Cole model takes the form:

ε0 ε∞ − 1 ' (εo− ε∞) ε∞ cos(βπ/2) (ωτ )β (1) and tnδ ' (εo− ε∞) ε∞ sin(βπ/2) (ωτ )β (2)

where ε∞and εoare respectively the high and low

frequency dielectric constants. In our experiment β ' 0.5 thus the two equations become:

ε0 ε∞ − 1 ' (εo− ε∞) ε∞ 1 (2ωτ )1/2, (3) tnδ ' (εo− ε∞) ε∞ 1 (2ωτ )1/2. (4)

These two equations show that tnδ ' (ε0

ε∞ − 1)

which is rather well veried by the experimen-tal data for all λ. This indicates that they are a rather good model of the dielectric constant un-der stress. Using eq.4, we identify the tting pa-rameter as τef f = 2τ /∆2 where ∆ = (εo− ε∞)/ε∞.

At this level we cannot distinguish whether the increase of ε0 _{and tnδ is induced by a decreasing}

4

We suppose rst that the stress has no inuence on τ but only on ∆. Keeping for τ the value of the unstressed sample we can estimate the value of ∆ under stress from the previous expression ∆ = p2τ /τef f. To give a lower bound on the

value of τ at 22o_{C we rely upon the high}

tempera-ture measurements of the dielectric properties of PC and on the estimation based on the WLF law [28, 34]. For example at T = 130o_{C, which is the}

minimum temperature at which the WLF law for PC can be tested, one nds using WLF τ ' 107_s.

Thus, as τ increases by lowering temperature, we simply assume that at T = 22o_{C, τ > 10}7_{s, which,}

in addition, is coherent with the fact that our samples are more than one year old [35].

From the data of g.5b) we obtain τef f ' 1/ ˙γ at

λ = 1.17. Thus in the hypothesis of constant τ, we nd ∆ > 7 at ˙γ = 2.5 × 10−6 _{and ∆ > 223 at ˙γ =}

2.5 × 10−3s−1. This value is 1000 times larger than that of the unstressed sample for which ∆ ' 0.12. This is nonphysical because ∆ cannot change of such a large amount even at very large strain (see refs.[33,36,37] and appendix I.2). Therefore one has to conclude that, although a small increase of the dielectric strength ∆ cannot be excluded, the observations cannot be explained without a de-creasing of τ with the mechanical stress, i.e. an acceleration of the dynamics, which agrees with other experimental results based on other tech-niques and other polymers [24,25,38]. Using our data we can estimate τ ' 0.5∆2_{a(λ)/ ˙γ, assuming}

that ∆ = 0.2, i.e. keeping x the value of the un-stressed sample. At λ = 1.17 we see that a(λ) ' 1 and using ˙γ = 2.5 × 10−3 _{one nds τ = 8s, which}

implies that one would not observe a clear power law at the highest strain rate because the approxi-mation ω >> 1/τ will be not fully satised for the low frequencies of our frequency range. There-fore at the highest ˙γ, one concludes that τ >> 8 s and ∆ > 0.2. At this point we can safely say that the main eect of the strain is a reduction of τ of several orders of magnitude although a small inuence on ∆ is necessary for the consistency of the observations.

Finally we compare these results with those that one extracts from the values tnδ(f) measured in the unstressed sample at dierent tempera-tures. In the range 137o_{C < T < 152}o_{C, where}

we observe an overlap of DSsand DSTin g.2 the

tnδ(f) has a power law dependence on f but the exponent β is a function of temperature: specif-ically 0.25 < β < 0.35 for 137o_{C < T < 152}o_C.

The observation that the β exponent measured under stress is larger than the one measured in the unstressed sample at temperature close to Tg,

has important consequences. Indeed in the Cole-Cole expression the smallest is the value of the exponent β the broadest is the distribution of

re-laxation times. Thus we conclude that our mea-surements are compatible with a narrowing of the distribution of relaxation times under stress ob-served in refs.[25, 38] with other techniques in other polymers.

Let us point out that in the frequency window of our measurements there are no other relax-ations which may inuence the results. The beta relaxation is not aected by the stress because DSsdoes not change above 400Hz. The gamma

re-laxation time is at very high frequencies at room temperature (see ref[36]) Structural changes that might also inuence the results are absent. Indeed we checked by Dierential Scanning Calorimetry (DSC) that polycarbonate does not crystallize under strain in the range considered here. Microscopic damaging may also contribute to the dielectric response. It has been shown that damaging occurs indeed during the strain hardening regime, but the volume fraction corresponding to these

damages is very small (< 10−4 _{for cellulose acetate [39]}

and for polycarbonate [40]) and the corresponding

per-turbation regarding the interpretation of the results is negligible.

Summarizing, we have studied the DSsof a sample of

polycarbonate at room temperature submitted to an ap-plied stress at dierent strain rates covering three order

of magnitude range. In our frequency window we

observe that, with respect to the unstressed sam-ple, tnδ increases at low frequencies as a function of ˙γ, whereas at high frequencies it remains un-changed. The increase is not monotonous and reaches its maximum in the softening regime. From the data we extract an eective time scale τef f ∝ 1/ ˙γwhich depends on λ and reaches its

min-imum in the softening regime to increase again in the hardening. We have also shown that the dis-tribution of relaxation times is narrower under stress than in the unstressed sample at T close to Tg. The large variation of τef f cannot be explained

only by the increase of the dielectric strength ∆, but it implies that the stress induces a shift towards high frequencies of the α peak without inuencing the high frequency part of the DS. Thus we conrm that the stress accelerates the polymer dynamics as already observed in other experiments [2328], which were limited to the yield points. We have extended the analysis to high values of strain in a polymer which presents stress hardening, nding two very important and unexpected results. Firstly the smallest τef f is

not reached at yield but in the stress softening regime (see Figs.2,3). Secondly the stress harden-ing regime is associated to a progressive increase of the eective relaxation time τef f = 2τ /∆2. This

increase may again be due either to an increase of τ or to a decrease of ∆ of at least a factor of 3 (see g.5), which, on the basis of the previous ar-guments, is too large. Furthermore this decrease

of ∆ will imply a non monotonous dependence of ∆ as a function of λ, which contradicts previous measurements [37,41]. Specically in ref. [41] the authors measured by NMR a segmental orienta-tion for stretched polycarbonate at room temper-ature, observing a monotonous increase of the ori-entational order parameter, whereas we observe a non monotonous behaviour of τef f versus λ. Thus

it is conceivable to say that such a behavior is in-duced by an increase of τ in the hardening regime.

As a conclusion our experimental results impose strong constrains on the theoretical models on strain softening and hardening.

I. APPENDIX I.1. Experimental methods

The mechanical part is composed by two cylinders (21 cm long, 1.6 cm of diameter) used to fasten the sam-ple, a load cell of 2000 N capacity, a precision linear trans-ducer to measure the strain over a range of 28 cm, and a brushless servo motor with a coaxial reducer. This device allows the investigation of polymer lm samples of a max-imum width of 21 cm. We use dog-bone shaped sheets with the eective dimensions of the area under

defor-mation of 16X16 cm2_{, in order to focus the deformation}

of the lm between the electrodes used to measure the DS. The sample is stretched at xed relative strain rate

˙γ = _dtd(L/Lo)Where L(t) is the length of the stretched

sample and Lo the initial length at zero applied force

F. Measurements are performed at room temperature

at constant ˙γ, till a stretch ratio λ = L(t)/L0 of about

1.8 (80 % strain) is reached. The force F (λ) applied to the sample is measured during all the experiment by a load cell. The value of the applied stress is dened as

σ = F (λ)/(Wodo)where Woand doare the initial width

and thickness of the sample.

We use an innovative dielectric spectroscopy technique

(see ref. [29]), which allows the simultaneous

mea-surement of the dielectric properties in a four orders of magnitude frequency window chosen in the range of

10−2− 103 _{Hz. This multi-frequency experiment is very}

useful in the study of transient phenomena such as

poly-mer lms deformation (see [29] for more details), because

it gives the evolution of the DS on a wide frequency range instead of a single frequency. Specically the de-vice measures the complex impedance of the capacitance

C formed by the electrodes and the sample. In all the

frequency range the device has an accuracy better than

1%on the measure of C and it can detect values of tnδ

smaller than 10−4_{. The real part of the dielectric}

con-stant of the sample is ε0 _{= Cd(λ)/(Sε}

o) where S is the

electrode surface, εo the vacuum dielectric constant and

d(λ) the thickness of the sample whose dependence on

λmust be taken into account to correctly estimate the

value of ε0_{. This compensation is not necessary for tnδ(f)}

because it does not depend on the geometry being the

ra-tio of ε00_/ε0_.

I.2. Kirkwood factor

The low frequency dielectric constant εo is

re-lated to ∞ via the Kirkwood factor g = 1 + z <

cos θi,j > where < . > stands for mean value, θi,j

is the angle between dipole i and dipole j, and z is the number of relevant nearest dipoles, which is usually taken to be about 10 (see ref. [33]). Specically one nds that εo= ε∞+ A g where A

is a material dependent factor. Therefore using this relationship, the quantity ∆ dened in the text becomes: ∆ = εo/ε∞− 1 = Ag/ε∞.

We notice that, independently of the probabil-ity distribution of θi,j, the value of < cos θi,j >

is limited between 0 and 1 and as a consequence 1 < g < 11. This means that, being ∆ = 0.2 in the unstressed sample, the maximum value of ∆ un-der stress is ∆ = 2.2. This value is much smaller than the values of ∆ estimated in the main text in order to explain the observations, assuming that only ∆ is aected by the strain. Thus on the ba-sis of the experimental data, we have to conclude that τ decreases in the stressed sample, i.e. the dynamics accelerates under strain. However this acceleration is associated to a small increase of ∆ as we discuss in the main text.

∗_{Present address of Riad Sahli: INM-Leibniz Institute}

for New Materials, 66123, Saarbr¨ucken, Germany

[1] J. D. Ferry, Viscoelastic Properties of Polymers (John Wiley and Sons, Inc., 1980).

[2] L. E. Nielsen and R. F. Landel, Mechanical Properties of Polymers and Composites (Marcel Dekker, New York, 1994).

[3] H. E. H. Meijer and L. E. Govaert, Macromolecular Chemistry and Physics 204, 274 (2003).

[4] H. E. H. Meijer and L. E. Govaert, Progress in Polymer Science 30, 915 (2005).

[5] R. S. Hoy and M. O. Robbins, The American Physical Society Division of polymer Physics 44, 3487 (2006). [6] R. S. Hoy and C. S. O'Hern, Physical Review E 82

(2010).

[7] R. S. Hoy, Journal of polymer science. Part B. Polymer physics 49, 979 (2011).

[8] H. G. H. Van Melick, H. E. H. Meijer, and L. E. Govaert, Polymer 44, 2493 (2003).

6

Tervoort, and U. W. Suter, Journal of polymer science. Part B. Polymer physics 46, 2475 (2008).

[10] D. J. A. Senden, J. A. W. van Dommelen, and L. E. Govaert, Journal of Polymer Science Part B: Polymer Physics 48, 1483 (2010).

[11] Jatin, V. Sudarkodi, and S. Basu, International Journal of Plasticity 56, 139 (2014).

[12] C. Chui and M. C. Boyce, Macromolecules 32, 3795 (1999).

[13] O. A. Hasan and M. C. Boyce, Polymer 34, 5085 (1993). [14] R. S. Hoy and M. O. Robbins, Polymer Physics 47, 1406

(2009).

[15] T. Ge and M. O. Robbins, Journal of Polymer Science Part B: Polymer Physics 48, 1473 (2010).

[16] E. J. Kramer, Journal of polymer science. Part B. Poly-mer physics 43, 3369 (2005).

[17] R. S. Hoy, Polymer Glasses. Chapter: Modeling strain hardening in polymer glasses using molecular simula-tions. pp 425-450 (Taylor and Francis Group: Boca Ra-ton, Edited by Connie B. Roth, 2016).

[18] A. Nicolas, E. Ferrer, K. Martens, and J. L. Barrat, Rev. Mod. Phys. , 045006 (2018).

[19] L. Conca, A. Dequidt, P. Sotta, and D. R. Long, Macro-molecules 50, 9456 (2017).

[20] P. K. Jaiswal, I. Procaccia, C. Rainone, and M. Singh, Phys. Rev. Lett. , 085501 (2016).

[21] C. B. Roth, Polymer Glasses (Taylor and Francis Group: Boca Raton, 2016).

[22] K. Venkatasway, K. Ard, and C. L. Beatty, Polymer Engineering and Science 22, 961 (1982).

[23] L. S. Loo, R. E. Cohen, and K. K. Gleason, 288, 116119 (2000).

[24] M. D. Ediger, Annu. Rev. Chem. 51, 99 (2000).

[25] B. Bending, K. Christison, J. Ricci, and M. D. Ediger, Macromolecules 47, 800 (2014).

[26] K. Hebert, B. Bending, J. Ricci, and M. D. Ediger, Macromolecules 48, 6736 (2015).

[27] J. Kalfus, A. Detwiler, and A. J. Lesser, Macromolecules 45, 4839 (2012).

[28] R. Perez-Aparicio, D. Cottinet, C. Crauste-Thibierge, L. Vanel, P. Sotta, J.-Y. Delannoy, D. R. Long, and S. Ciliberto, Macromolecules 49, 3889 (2016).

[29] R. Pérez-Aparicio, C. Crauste-Thibierge, D. Cottinet, M. Tanase, P. Metz, L. Bellon, A. Naert, and S. Cilib-erto, Review of Scientic Instruments 86, 044702 (2015). [30] P. Guan, M. Chen, and T. Egami, Phys. Rev. Lett. 104,

205701 (2010).

[31] K. Chen and K. S. Schweizer, Macromolecules 44, 3988 (2011).

[32] S. Havriliak and S. Negami, Journal of Polymer Science Part C-Polymer Symposium , 99+ (1966).

[33] Frohlich, Theory of Dielectics, Dielectric Donstant and Dielectric Loss (Oxford University Press, London, 1958). [34] D. Huang and G. B. McKenna, The Journal of Chemical

Physics 114, 5621 (2001).

[35] L. C. E. Struik, Physical Aging in Amorphous Polymers and Other Materials (Elsevier, Amsterdam 1978). [36] A. F. Yee and S. A. Smith, Macromolecules 14, 54 (1981). [37] M. Hernández, A. Sanz, A. Nogales, T. A. Ezquerra, , and M. López-Manchado, Macromolecules , 3176 (2013). [38] H.-N. Lee, K. Paeng, S. F. Swallen, and M. D. Ediger,

Science 323, 231 (2009).

[39] A. Charvet, P. Sotta, C. Vergelati, and D. R. Long, Macromolecules 52, 6613 (2019).

[40] S. Djukic, PhD thesis (to be defended in Lyon, 2020). [41] V.-D. Vogt, M. Dettenmaier, H. Spiess, and M. Pietralla,