Enthalpy recovery in semicrystalline polymers

Aleksey D. Drozdov
Institute for Industrial Mathematics
arXiv:cond-mat/0004029 v1 4 Apr 2000

4 Hanachtom Street, Beersheba, 84311 Israel

Abstract
Constitutive equations are derived for enthalpy recovery in polymeric glasses after
thermal jumps. The model is based on the theory of cooperative relaxation in a version
of the trapping concept. It is demonstrated that some critical temperature Tcr and some
critical degree of crystallinity fcr exist in a semicrystalline polymer above which the
energy landscape becomes homogeneous and structural relaxation ceases.

1 Introduction
This note is concerned with the kinetics of enthalpy recovery in polymeric glasses after
thermal jumps. Physical aging of polymers has been the focus of attention in the past three
decades, see monographs [1, 2, 3, 4, 5] and review articles [6, 7, 8]. Despite significant
successes in the study of out-of-equilibrium dynamics in disordered materials achieved in
the past decade (see reviews [9, 10, 11, 12, 13, 14]), it is difficult to mention a model that
adequately predicts structural relaxation in polymers. Even the mode-coupling theory [15]
(the most advanced among molecular models) fails to describe slowing down in the response
of amorphous polymers below the glass transition temperature Tg [13, 16, 17]. As a reason for
this conclusion, the neglect of cooperativity in the molecular reorientation is conventionally
mentioned [18, 19].
Enthalpy relaxation provides one of the most convenient ways to study aging of glassy
polymers [7, 20, 21, 22, 23]. The kinetics of enthalpy recovery is in close connections with
time-varying changes in static and dynamic moduli [24, 25, 26], specific volume [27, 28]
and yield stress [29, 30, 31, 32], as well as with transition from ductile to brittle fracture
[33, 34]. For some polymers (e.g., polyesters), DSC measurements demonstrate a dramatic
effect of the waiting time [35], whereas the free volume fraction determined by PALS shows
no pronounced decrease induced by physical aging [36].
Constitutive models for structural relaxation in glassy polymers are conventionally con-
fined to amorphous media (as exceptions, we would mention Refs. [37, 38]). This may be
explained by a belief that crystalline lamellas do not affect the process of structural relax-
ation occurring in amorphous regions [39]. This hypothesis may be accepted provided that
only one kind of amorphous domains exists in a semicrystalline polymer. A number of ob-
servations demonstrate, however, a more complicated picture, where two different kinds of
amorphous regions co-exist in a partially crystallized polymer: amorphous layers between

1
lamellas and amorphous domains between spherulites [40, 41, 42, 43]. This results in a
question whether the rate of physical aging is a linear function of the degree of crystallinity
which vanishes when the degree of crystallization f reaches 100% only (as it is postulated in
Fig. 6 of Ref. [39]), or the rates of structural relaxation substantially differ in interlamellar
and intralamellar zones which implies that physical aging may decline even in a partially
crystallized material.
The objective of this study is to derive a constitutive model for enthalpy recovery which
is based on the theory of cooperative relaxation in a version of the trapping concept [16, 44,
45, 46, 47]. To simplify the analysis, we do not distinguish explicitly between interlamellar
and intralamellar regions, but presume that the “averaged” distribution of energies of traps
may depend on the degree of crystallinity. An amorphous region is treated as an ensemble
of mutually independent cooperatively rearranged regions (CRR). A CRR is thought of as a
globule consisting of scores of strands of long chains [46]. The characteristic length of a CRR
in the vicinity of Tg amounts to several nanometers [48]. In the phase space, a CRR is treated
as a point trapped in its potential well (cage). At random times, the unit hops to higher
energy levels as it is thermally agitated. When the energy of thermal fluctuation exceeds the
height of a barrier between cages, a CRR can change its trap. Introducing several hypotheses
regarding the kinetics of hops from one potential well to another, we develop a nonlinear
parabolic equation for the probability density of traps with various potential energies. This
equation is applied to the analysis of enthalpy recovery in amorphous and semicrystalline
poly(ethylene terephthalate) (PET). The following conclusions are drawn:

1. the degree of crystallinity of a semicrystalline polymer affects the model’s param-

eters in a similar way as temperature influences those for an amorphous polymer (a
crystallinity–temperature principle analogous to the time–temperature principle in lin-
ear viscoelasticity [49]);

2. a critical level of crystallinity fcr < 100% exists at which the energy landscape becomes
homogeneous. This level may be treated as an analog to the critical temperature
for structural relaxation in amorphous polymers [50] predicted by the mode-coupling
theory.

The exposition is organized as follows. Constitutive equations for the kinetics of enthalpy
relaxation are developed in Section 2. These relations are verified in Section 3 by comparison
with experimental data. Some concluding remarks are formulated in Section 4.

2 A model for enthalpy relaxation

Denote by w the energy of a potential well with respect to some reference state [51]. It is
assumed that w > 0 for any trap and w = 0 for the reference state. At random times, a
CRR hops to higher energy levels in its potential well as it is thermally activated. Denote
by q(ω)dω the probability to reach (in a hop) the energy level that exceeds the bottom level
of its potential well by a value located in the interval [ω, ω + dω]. Referring to [10], we set
q(ω) = A exp(−Aω), where A is a material constant. The probability for a CRR in a trap

2
with potential energy w to reach the reference state in an arbitrary hop is given by
Z ∞
Q(w) = q(ω)dω = exp(−Aw).
w

The average rate of hops in a cage γ is determined by the current temperature T only,
γ = γ(T ). The rate of rearrangement P equals the product of the rate of hops γ by the
probability Q to reach the reference state in a hop,

P (w) = γ exp(−Aw). (1)

Denote by X the (time-uniform) concentration of traps per unit mass, and by p(t, w) the
current probability density of traps with potential energy w. The number of relaxing regions
(per unit mass) trapped in cages with the energy belonging to the interval [w, w + dw] and
rearranged during the interval of time [t, t + dt] is XP (w)p(t, w)dwdt. Unlike Refs. [52, 53],
we assume that not all flow units change their traps when they reach the reference energy
level and denote by F (t, w) the ratio of the number of relaxing regions returning to their
traps to the number of those reaching the reference state. The number of relaxing regions
leaving their cages (with the energy located within the interval [w, w + dw]) per unit mass
and unit time is given by X(1 − F )P pdw. The exchange of flow units is assumed to occur
only between the nearest neighbors on the energy landscape, that is between a trap with
the energy [w, w + dw] and traps with the energies [w − dw, w] and [w + dw, w + 2dw]. The
balance law for the number of flow units trapped in cages with the energy belonging to the
interval [w, w + dw] reads

∂p 1h i 1h i
= −(1 − F )P p + (1 − F )P p + (1 − F )P p ,
∂t 2 + 2 −

where the subscript indices “−” and “+” refer to appropriate quantities for the intervals
[w − dw, w] and [w + dw, w + 2dw]. Expanding the right-hand side of this equality into
the Taylor series, using Eq. (1) and introducing the notation Γ = 21 γdw 2, we arrive at the
differential equation for diffusion over the energy landscape

∂p ∂2 h i
= Γ 2 (1 − F ) exp(−Aw)p . (2)
∂t ∂w
We adopt the Metropolis transition rates [54],
(
1, p(t, w) ≤ p∞ (w),
F (t, w) = (3)
exp[−ǫ(p − p∞ )], p(t, w) > p∞ (w),

where p∞ (w) is the equilibrium density of traps and ǫ > 0 is a material parameter. An
important advantage of Eq. (2) compared to relationships suggested in Refs. [55, 56] is that
under condition (3) it does not impose restrictions on the equilibrium density of trap p∞ (w).
Referring to [16], we suppose that the inequality
Z 0
p(t, w)dw ≪ 1 (4)
−∞

3
is satisfied for any t ≥ 0 and describe the initial distribution, p0 (w), and the equilibrium
distribution, p∞ (w), by the Gaussian formulas
1 (w − W )2 1 (w − W )2
   
p0 (w) = √ exp − , p∞ (w) = √ exp − , (5)
2πΣ0 2Σ20 2πΣ∞ 2Σ2∞
where W , Σ0 , Σ∞ are adjustable parameters. Equations (5) imply that the average equi-
librium energies of traps are temperature-independent (the same value W is employed for
the initial and equilibrium distribution functions), whereas their variances strongly depend
on T . The first assertion is fairly well confirmed by experimental data in mechanical tests
[57, 58], whereas the other hypothesis is in agreement with the conventional scenario for the
growth in the ruggedness of the energy landscape with a decrease in temperature [10, 47, 59].
The level of disorder in an ensemble of CRRs is characterized by the configurational
entropy per rearranging region [60],
Z ∞
s(t) = −kB p(t, w) ln p(t, w)dw,
0

where kB is Boltzmann’s constant. The configurational enthalpy per unit cage h(t) is ex-
pressed in terms of the configurational entropy s(t) by means of the conventional formula
∂h
= T.
∂s
This equality is integrated for a standard one-step thermal test,
T (t) = T0 (T < 0), T (t) = T (T > 0), (6)
and an explicit formula is found for the enthalpy per unit mass H = Xh. The relaxing
enthalpy per unit mass ∆H(t) = H(t) − H(0) is given by
Z ∞h i
∆H(t) = Λ p0 (w) ln p0 (w) − p(t, w) ln p(t, w) dw (7)
0

with Λ = kB T X. Introducing the dimensionless variables w̄ = Aw and t̄ = t/t0 , where t0

is the characteristic time of aging, and setting Γ̄ = A2 Γt0 , W̄ = AW and Σ̄k = AΣk , we
arrive at the constitutive model, Eqs. (2), (3), (5) and (7), with six adjustable parameters
W̄ , Σ̄0 , Σ̄∞ , Γ̄, ǫ and Λ. These kinetic equations substantially differ from conventional
relations for enthalpy recovery in polymers [21, 61, 62, 63], because they do not refer to
a (purely phenomenological) concept of internal clock. As an analog of the material time
τ in Eqs. (2), (3), (5) and (7), we may mention the parameter ǫ that characterizes the
rate of changes in the energy landscape. An important difference between ǫ and τ is that
ǫ is independent of the current energy landscape (but, in general, temperature-dependent),
whereas the parameter τ is conventionally expressed in terms of the current enthalpy H by
means of the Narayanaswamy or the Adam–Gibbs equations.

3 Comparison with experiments

It is easy to check that the quantities W̄ and Γ̄ are interrelated: when one of them is
fixed, the other may be chosen to characterize the time scale. For convenience of numerical

4
simulation, we fix W̄ and determine Γ̄ by fitting observations. The value W̄ = 2.5 ensures
that inequality (4) is satisfied with a high level of accuracy.
We begin with experimental data for two amorphous PETs exposed in Ref. [33]. First,
observations are fitted for a homopolymer to determine the parameters Σ̄0 , Σ̄∞ , Γ̄, ǫ and Λ
that ensure the best approximation of observations. Afterwards, the amounts Σ̄0 , Σ̄∞ and
Γ̄ are fixed, and the quantities ǫ and Λ are determined by matching experimental data for
a copolymer. Figure 1A demonstrates fair agreement between observations and results of
numerical simulation. Using the specific gravity g = 1.34 g/cm3 at T = 65 ◦ C [33], we find
the volume concentration of relaxing regions in PET, Ξ = 6.39 · 1026 m−3 , which is rather
close to Ξ = 4.1·1026 m−3 for polycarbonate [64] and Ξ = 7.7·1026 m−3 for poly(vinyl acetate)
[65] found by PALS. To ensure that adjustable parameters are rather robust with respect to
changes in material properties and conditions of the test (a decrease in Tg by 5 K and an
increase in T0 by 15 K), we repeat this procedure using experimental data presented in Ref.
[66]. Figure 1B demonstrates fair agreement between observations and results of numerical
simulation for similar values of Σ̄0 , ǫ and Λ. An increase in Γ̄ depicted in Figure 1B compared
to Figure 1A seems quite natural, because this parameter is determined by the difference
between the glass transition temperature and the annealing temperature, ∆T = Tg − T . A
decrease in Σ̄∞ may be associated with a drop in Tg , because it is conventionally accepted
that the glass transition temperature grows when molecular mobility becomes more restricted
(e.g., because of an increase in the number of crosslinks [30] or in the molecular weight [67]),
which may be associated with the growth of inhomogeneity of the energy landscape.
To study the effect of the annealing temperature T on the kinetics of structural recovery,
we approximate experimental data for amorphous PET obtained in Ref. [35]. Figure 2A
demonstrates fair agreement between observations and results of numerical simulation in
the temperature range from Tg − 25 to Tg − 5 ◦ C. The standard deviation of energies of
traps in thermodynamic equilibrium Σ̄∞ is plotted versus the degree of undercooling ∆T in
Figure 2B. This figure demonstrate that the dependence Σ̄∞ (T ) is fairly well approximated
by the linear function
Σ̄∞ = a0 + a1 ∆T, (8)
where ak are adjustable parameters. In follows from Eq. (8) that some critical temperature
Tcr exists at which Σ̄∞ vanishes and the energy landscape becomes homogeneous. Results of
simulation imply that Tcr = Tg + 34.11 ◦ C which is in accord with the values of Tcr found by
fitting date in mechanical tests [50]. The quantities Γ̄ and ǫ are depicted in Figure 3A. This
figure show that the dependences Γ̄(T ) and ǫ(T ) are correctly approximated by the “linear”
functions
log Γ̄ = b0 − b1 ∆T, log ǫ = c0 + c1 ∆T (9)
with adjustable parameters bk and ck . In the vicinity of the glass transition temperature,
the apparent activation energy ∆E is calculated as [68]

d ln Γ̄ 

∆E = −R ,
d(1/T ) T =Tg

where R is the gas constant. This equality together with Eq. (9) results in the formula
∆E = RTg2 b1 . According to Figure 3A, ∆E = 156.1 kJ/mol, which is quite comparable
with ∆E = 206 kJ/mol determined for PET in mechanical tests [69]. An increase in the

5
parameter Γ̄ with temperature T is in agreement with the theory of thermally activated
processes. A surprising result is a decrease in ǫ with temperature. It follows from Eq. (3)
that this decrease provides an additional source for slowing down of the aging process in the
close vicinity of Tg which has not been accounted for in previous studies.
To assess the effect of crystallinity on the rate of structural relaxation, we approximate
experimental data for enthalpy recovery in semicrystalline PET with various degrees of
crystallinity f . For a detailed description of specimens and the experimental procedure,
see Ref. [43]. First, we approximate observations for a sample with the smallest value of
f and determine adjustable parameters of the model using the steepest-descent procedure.
Afterwards, we fix the quantities Σ̄0 and Λ and repeat matching experimental data with
three adjustable parameters, Σ̄∞ , Γ̄ and ǫ. Figure 4A demonstrates fair agreement between
observations and predictions of the model. The equilibrium standard deviation of energies
of traps Σ̄∞ is depicted versus the degree of crystallinity f in Figure 4B. The dependence
Σ̄∞ (f ) is correctly approximated by the linear function

Σ̄∞ = a0 − a1 f (10)

with adjustable parameters ak . Comparing Figures 2B and 4B, we conclude that an increase
in f for a semicrystalline polymer affects the equilibrium distribution of cages in a way
similar to that in which the annealing temperature T influences the energy landscape for an
amorphous material. By analogy with an amorphous polymer, one may define the critical
degree of crystallinity fcr for a semicrystalline medium as the percentage of crystallites at
which the parameter Σ̄∞ vanishes and the energy landscape becomes homogeneous. In
contrast to the conclusions of Ref. [39], Figure 4B demonstrates that fcr is essentially
less than 100%. The parameters Γ̄ and ǫ are plotted versus the degree of crystallinity in
Figure 3B. This figure shows that the dependences Γ̄(f ) and ǫ(f ) are correctly approximated
by the “linear” functions

log Γ̄ = b0 + b1 f, log ǫ = c0 − c1 f (11)

with adjustable parameters bk and ck .

Two kinds of crystallization of polymers are conventionally studied: thermal crystalliza-
tion, when a specimen is annealed at a fixed temperature in the rubbery region for a given
time (this procedure was employed in Ref. [43]) and mechanically induced crystallization,
when a sample is stretched slightly above Tg to a given extension ratio λ. To demonstrate
that the effect of the degree of crystallinity on the kinetics of physical aging weakly depends
on the crystallization procedure, we approximate experimental data for semicrystalline PET
samples where the level of crystallization is established by uniaxial stretching at an elevated
temperature. A detailed description of specimens and the experimental procedure can be
found in Ref. [34]. The extension ratio λ is transformed into the level of crystallinity using
Figure 1 of Ref. [70]. Unlike previous sets of experimental data, where ǫ is treated as a
function of the level of crystallinity f , we fit observations at various elongations λ with the
same value of ǫ found for an undeformed specimen. Figure 5A demonstrates good agreement
between experimental data and predictions of the model. The functions Σ̄∞ (f ) and Γ̄(f ) are
plotted in Figure 5B. This figure demonstrates that Eqs. (10) and (11) adequately describe
observations. The parameter b1 (that characterizes the influence of crystallinity on the rate

6
of rearrangement) acquires similar values for thermally and mechanically crystallized sam-
ples. The same assertion is true for the critical degree of crystallinity fcr which adopts the
value 33.9% for the stretched PET versus 42.7% for the PET annealed above Tg . On the
contrary, the parameter Σ̄∞ for the PET obtained by thermal crystallization exceeds that
for the hot-drawn PET by twice. This may be explain by the fact that stretching a rubbery
polymer establishes an additional order caused by partial destruction of a polymeric network
and alignment of long chains along the axis of loading [32].

4 Concludung remarks
Constitutive equations have been derived for enthalpy relaxation in glassy polymers after
thermal jumps. The model is based on the trapping concept which treats a disordered
medium as an ensemble of flow units rearranged at random times as they are thermally
activated. Adjustable parameters are found by fitting experimental data for amorphous and
semicrystalline poly(ethylene terephthalate). The following conclusions are drawn:

1. the constitutive equations correctly describe the kinetics of structural relaxation in the
sub–Tg region.

2. the model predicts the existence of some critical temperature Tcr at which the energy
landscape becomes homogeneous.

3. for semicrystalline polymers, the model implies the existence of some critical degree of
crystallinity fcr at which the energy landscape becomes homogeneous. The inequality
fcr < 100% may serve as an indirect confirmation that the processes of structural
relaxation in interlamellar and intralamellar regions differ from one another.

4. a correspondence may be established between the effects of temperature and the level
of crystallinity on the kinetics of enthalpy recovery.

Acknowledgments
The author is grateful to A.L. Svistkov for fruitful discussions. The work is supported by
the Israeli Ministry of Science through grant 1202–1–98.

References
[1] L.C.E. Struik, “Physical Ageing in Amorphous Polymers and Other Materials”, Elsevier,
Amsterdam 1978

[2] S.A. Brawer, “Relaxation in Viscous Liquids and Glasses”, Am. Ceram. Soc., Columbus
1985

[3] E.-J. Donth, “Relaxation and Thermodynamics in Polymers: Glass Transition”,

Academie-Verlag, Berlin 1992

7
 [4] S. Matsuoka, “Relaxation Phenomena in Polymers”, Hanser-Verlag, Munich 1992

[5] P.G. Debenedetti, “Metastable Liquids: Concepts and Principles”, Princeton Univ.
Press, Princeton 1996

[6] G.B. McKenna, in: “Comprehensive Polymer Science”, C. Booth, C. Price, Eds., Perg-
amon Press, Oxford 1989, Vol. 2, p. 311

[7] C.T. Moynihan, S.N. Crichton, S.M. Opalka, J. Non-Cryst. Solids 131–133, 420 (1991)

[8] M.D. Ediger, C.A. Angell, S.R. Nagel, J. Phys. Chem. 100, 13200 (1996)

[9] W. Kob, in: “Experimental and Theoretical Approaches to Supercooled Liquids: Ad-
vances and Novel Applications”, J. Fourcas, D. Kivelson, U. Mohanty, K. Nelson, Eds.,
ACS Books, Washington 1997, p. 28

[10] J.-P. Bouchaud, L.F. Cugliandolo, J. Kurchan, M. Mezard, in: “Spin Glasses and Ran-
dom Fields”, A.P. Young, Eds., World Scientific, Singapore 1998, p. 161

[11] J.-P. Bouchaud, Preprint cond-mat/9910387

[12] J. Hammann, E. Vincent, V. Dupuis, M. Alba, M. Ocio, J.-P. Bouchaud, Preprint

cond-mat/9911269

[13] W. Kob, C. Brangian, T. Stühn, R. Yamamoto, Preprint cond-mat/0003282.

[14] G. Tarjus, D. Kivelson, Preprint cond-mat/0003368.

[15] W. Götze, L. Sjögren, Rep. Prog. Phys. 55 241 (1992)

[16] J.C. Dyre, Phys. Rev. B 51, 12276 (1995)

[17] H.Z. Cummins, G. Li, Y.H. Hwang, G.Q. Shen, W.M. Du, J. Hernandez, N.J. Tao, Z.
Phys. B 103, 501 (1997)

[18] G. Adam, J.H. Gibbs, J. Chem. Phys. 43, 139 (1965)

[19] S. Matsuoka, A. Hale, J. Appl. Polym. Sci. 64, 77 (1997)

[20] I.M. Hodge, J. Non-Cryst. Solids 169, 211 (1994)

[21] J.L. Gomez Ribelles, M. Monleon Pradas, Macromolecules 28, 5867 (1995)

[22] D.J. Hourston, M. Song, A.Hammiche, H.M. Pollock, M. Reading, Polymer 37, 243
(1996)

[23] A. Brunacci, J.M.G. Cowie, R. Ferguson, I.J. McEwen, Polymer 38, 865, 3263 (1997)

[24] A.B. Brennan, F. Feller, J. Rheol. 39, 453 (1995)

[25] H. Yoshida, Thermochim. Acta 266, 119 (1995)

8
[26] J.M.G. Cowie, R. Ferguson, S. Harris, I.J. McEwen, Polymer 39, 4393 (1998)

[27] J.M.G. Cowie, S. Harris, I.J. McEwen, Macromolecules 31, 2611 (1998)

[28] A.D. Drozdov, Phys. Lett. A 258, 158 (1999)

[29] C. Bauwens-Crowet, J.-C. Bauwens, Polymer 23, 1599 (1982)

[30] W.D. Cook, M. Mehrabi, G.H. Edward, Polymer 40, 1209 (1999)

[31] J.M. Hutchinson, S. Smith, B. Horne, G.M. Gourlay, Macromolecules 32, 5046 (1999)

[32] A. Pegoretti, A. Guardini, C. Migliaresi, T. Ricco, Polymer 41, 1857 (2000)

[33] S.A. Jabarin, E.A. Lofgren, Polym. Eng. Sci. 32, 146 (1992)

[34] S. Mukherjee, S.A. Jabarin, Polym. Eng. Sci. 35, 1145 (1995)

[35] A. Aref-Azar, J.N. Hay, Polymer 23, 1129 (1982)

[36] E.-A. McGonigle, J.H. Daly, S.D. Jenkins, J.J. Liggat, R.A. Pethrick, Macromolecules
33, 480 (2000)

[37] C.K. Chai, N.G. McCrum, Polymer 21, 706 (1980)

[38] L.C.E. Struik, Polymer 28, 1521, 1534 (1987)

[39] M.R. Tant, G.L. Wilkes, J. Appl. Polym. Sci. 26, 2813 (1981)

[40] K.H. Illers, H. Breuer, J. Coll. Sci. 18, 1 (1963)

[41] G. Groenickx, H. Berghmans, G. Smets, J. Polym. Sci.: Polym. Phys. Ed. 14, 459
(1976)

[42] S.X. Lu, P. Cebe, Polymer 37, 4857 (1996)

[43] J.A. Diego, J.C. Canadas, M. Mudarra, J. Belana, Polymer 40, 5355 (1999)

[44] J.P. Bouchaud, J. Phys. I France 2, 1705 (1992)

[45] C. Monthus, J.-P. Bouchaud, J. Phys. A 29, 3847 (1996)

[46] P. Sollich, Phys. Rev. E 58, 738 (1998)

[47] G. Diezemann, U. Mohanty, I. Oppenheim, Phys. Rev. E 59, 2067 (1999)

[48] A.K. Rizos, K.L. Ngai, Phys. Rev. E 59, 612 (1999)

[49] J.D. Ferry, “Viscoelastic Properties of Polymers”, J. Wiley & Sons, New York 1980

[50] A.D. Drozdov, Europhys. Lett. 49, 569 (2000)

[51] M. Goldstein, J. Chem. Phys. 51, 3728 (1969)

9
[52] R.E. Robertson, J. Appl. Phys. 49, 5048 (1978)

[53] R.E. Robertson, R. Simha, J.G. Curro, Macromolecules 17, 911 (1984)

[54] H. Böttger, V.V. Bryksin, “Hopping Conduction in Solids”, VCH Publishers, Weinheim,
1985

[55] P. Feltham, Phys. Stat. Sol. 30, 135 (1968)

[56] F. Dobes, Acta Metall. 28, 377 (1980)

[57] A.D. Drozdov, Comput. Mater. Sci. 15, 422 (1999)

[58] A.D. Drozdov, Modelling Simul. Mater. Sci. Eng. 7, 1045 (1999)

[59] C.M. Roland, P.G. Santangelo, K.L. Ngai, J. Chem. Phys. 111, 5593 (1999)

[60] I. Avramov, A. Milchev, J. Non-Cryst. Solids 104, 253 (1988)

[61] C.T. Moynihan, A.J. Easteal, J. Wilder, J. Tucker, J. Phys. Chem. 78, 2673 (1974)

[62] A.J. Kovacs, J.J. Aklonis, J.M. Hutchinson, A.R. Ramos, J. Polym. Sci.: Polym. Phys.
Ed. 17, 1097 (1979)

[63] I.M. Hodge, Macromolecules 20, 2879 (1987)

[64] J. Bohlen, J. Wolff, R. Kirchheim, Macromolecules 32, 3766 (1999)

[65] G. Dlubek, Th. Lüpke, J. Stejny, M.A. Alam, M. Arnold, Macromolecules 33, 990 (2000)

[66] S.E.B. Petrie, Polym. Preprints 15(2), 336 (1974)

[67] G.A. Adam, J.N. Hay, I.W. Parson, R.N. Haward, Polymer 17, 51 (1976)

[68] L.C.E. Struik, Polymer 38, 733 (1997)

[69] L.P. Chen, A.F. Yee, E.J. Moskala, Macromolecules 32, 5944 (1999)

[70] P. Varma, E.A. Lofgren, S.A. Jabarin, Polym. Eng. Sci. 38, 237 (1998)

10
 pppppppppppbpppppppppp 2
6.0 pp pp p pp pp 4.0
prp p p p p p
ppp ppppp pppp pppppbpp
p p p 1 b
p p p pbppppbpp pbpp
pp pp ppp pp pp
bpppbbbb
r p p p p p pppppppppppppppppppppbpp ppp b
bp ppppp p p p
p p p
p p pp p
p p p p p p p p pp p pp p p p bpppp ppp
∆H ∆H p p p
p p p pppp ppppppprpbp p pppp pp pp b bp p p p
p
p
pp prppp pppp p p ppp p pp pp
p p r
p p p
p p
p ppppppp pppppppp pbp pp
p pp p pp
p p p
p b
ppp pppppppp p pppp bp ppp pp
rp ppppppppp pp pp pppppp p pp b pp
pb p pp
ppbpp ppp
pp p
p pp
ppp pbp p
p ppp
p pp p
pppbp pp p
0.0 A 1.0
p pp B
0.0 log t 1.5 0.0 log t 2.0
Figure 1: A—The relaxation enthalpy ∆H J/g versus time t days for amorphous PET
(Tg = 80 ◦ C) quenched from T0 = 80 to T = 65 ◦ C. Circles: experimental data [33].
Solid lines: prediction of the model with Σ̄0 = 0.19, Σ̄∞ = 0.95 and Γ̄ = 8.0 days−1 .
Curve 1: homopolymer, ǫ = 0.2, Λ = 4.0 J/g; curve 2: copolymer (3.5 mole % of cyclohexane
dimethanol), ǫ = 0.17, Λ = 4.4 J/g. B—The relaxation enthalpy ∆H J/g versus time t h for
amorphous PET (Tg = 75 ◦ C) quenched from T0 = 95 to T = 65 ◦ C. Circles: experimental
data [66]. Solid line: prediction of the model with Σ̄0 = 0.15, Σ̄∞ = 0.38, Γ̄ = 1.24 h−1 ,
ǫ = 0.2 and Λ = 4.2 J/g

3.0 3 1.0
pppppppprppppppppppppp ppppppppppppppppppp 2
pp
pppppppppppppp ppprppppppppppppppppp ppppp2 pp
p p pp p pp p p p p pp pp pp p 4
p p p p p p p p p p p p p ppppppp pppp
pppppppppr
pp p p p p p pppppp pppppp
∆H pppppp 2 ppp pp∗ Σ̄∞ p p p p p pb
p p
r p p p p p p pp p p p p pppp pppppp pp
p p p p p
p p p pp
pppp p p p
p pp p
ppp p p p
p pp p p
p pp p p p p p pppbpppp pp ppp
ppppppppppppppppppbppppppppppppppppppppp 1
p p
pppp ppppppppppp pppppppppppb∗
p p p bp ppp ppp pppppp p p p p
pp2
p p p p p p p p p p p
pp pp prpp ppppppp ppppp
pp pp pp pppppppppbppppppppp pppp pppp ppp p
p p
p pp bp p p p p p p p p p
p p p p p pppppp pppppp
pppppppppppp2 pppp pppppp
bpp p pppp p p p pp
p
pp p p p p p p
p p
p p pp p
∗
ppbppppppppppp p pp p
pppppppppppppp p2 ∗p p p p p p p pp pp pp
pp pp
p p p pp pp
p p p p p ∗
p p p p p p p pp p p p p
p pp p
0.0 A 0.0 B
0.0 log t 2.0 0.0 ∆T 30.0
Figure 2: A—The relaxation enthalpy ∆H J/g versus time t h for amorphous PET (Tg =
76 ◦ C) quenched from T0 = 277 to the test temperature T ◦ C. Symbols: experimental data
[35]. Solid lines: prediction of the model with Σ̄0 = 0.25 and Λ(Tg ) = 2.9 J/g. Curve 1:
T = 71; curve 2: T = 67; curve 3: T = 60; curve 4: T = 52. B—The parameter Σ̄∞ versus
∆T ◦ C. Circles: treatment of observations. Solid line: approximation of the experimental
data by Eq. (8) with a0 = 0.433 and a1 = 0.013

11
 1ppb 2pp 1
3.0 2ppppp pppp pp 0.0 2.0 0.0
p pp ppppp p
pppp p ppp ppp
p pp ppp p pp pp
p
pppp
ppp p
p p p pppp
pp pp ppp
ppppb
ppp p p p ppppppp ppppr p pppb
p r
log Γ̄ log ǫ log Γ̄ ppp log ǫ
p p ppp
pppp pppppp pppp pb pp
ppp
pppp p p p p p pppppp p pppp
pp pp p p
r p p p p p pppp ppppp
pp p ppppp pppp pr ppp p
pp pp p
p p p p p pp p pppppp p pp r
ppp pp ppp
rp ppppp pppp ppp pp pppp ppp
p p p p pppppp p pp
p ppb pppp
p p p p p p p p b pppp pp pp pppp
rppp pppp
p p p p p p pp pr pppp ppp pp
ppppp p pp ppp p pp pp
p pppppp pppp p pppp
p ppp
p pp
pppp pppbpp pppp
pppp A ppp p ppppp
p ppp B
0.0 pppb −3.0 1.0 ppp pppp −3.0
0.0 ∆T 30.0 0.0 f 50.0
Figure 3: The parameters Γ̄ h−1 (curve 1) and ǫ (curve 2) versus ∆T ◦ C for amorphous PET
(A) and versus f % for semicrystalline PET (B). Symbols: treatment of observations [35, 43].
Solid lines: approximation of the experimental data by Eqs. (9) and (11). A: b0 = 3.686,
b1 = 0.154, c0 = −2.424, c1 = 0.073; B: b0 = 0.618, b1 = 0.038, c0 = 0.247, c1 = 0.071

pppp
3.0 1.2 p pp
pppp
pppp
Σ̄∞ bppp
pppp
∆H pp p 1 ppp b
ppppp ppbppp pppp
ppppppbppp ppp
pp p p p p p p pp 2 pppp
p pp pp pp ppppprpppp pp p
p p p p p pp pp p p p p p p p p p ppprppppp pppp
pp p
p p pp p pp p b pp p p pp pp pppp
ppppp pp ppppp ppppp pp ppp ppppp2 p ppp pp 3
prp ppppppppppppppppppp pppppppppp2 b pp pppp b
b p p p p p
b
p p p p
pp
pp pp p pp pppp ppppppppppp2 p p p p p p p p p p p ∗
p p p p p p p p p ∗
p ppp pp 4 pp p
pppp
p p p p p p p ppp p p p p p p ppppppppppppppppp p p p p p p p p p p pp pp pp p pp p
ppp bppp p pp ppppppppppppp pppppppppprpppppp pppp pppp∗ p pp p pp pp p p
p pppp
p p p p pppp
p p p ppp pp p ppp ppp pp p ppp p ppppppppp2
p ppp p p p p p pp p
p pp p p p p p p p p pppp
p p p pppppppppppppppppppppp pppppp ppp p pppp p pp∗ pppppppppppp pp p
0.0 p ppp A 0.0
pppp B
pp
−1.0 log t 1.5 0.0 f 50.0
Figure 4: A—The relaxation enthalpy ∆H J/g versus time t h for semicrystalline PET
(Tg = 80 ◦ C) quenched from T0 = 100 to T = 65 ◦ C. Symbols: experimental data [43]. Solid
lines: prediction of the model with Σ̄0 = 0.2 and Λ = 1.5. Curve 1: f = 14%; curve 2:
f = 21%; curve 3: f = 27%; curve 4: f = 32%. B—The parameter Σ̄∞ versus f %. Circles:
treatment of observations. Solid line: approximation of the experimental data by Eq. (10)
with a0 = 1.519 and a1 = 0.036

12
 5.0 A pp 1 1.0 B 2p pppp pppp ppp 0.0
p
p p p pppppppppppp 2 p p p
pr
pp p ppp p p pp pp pp p pp
p pbppppppppppp pp p p ppp ppp
∆H Σ̄∞ log Γ̄
p
p pp p p p
pbpppp prpppprp pppp
p pppppppp ppp p p rpppp ppp p
p
p p p p
pp ppp pp pp pp pp ppp pp 3
p p bppppp p
pp pppbprpp pprppppppppppp2 pp ppppppp pp pppp
p p p
ppppbpppppppp2 pp pp pp pp pp pp pp pp ppppppp pp pp pp pp
p p 2 p
rp p p2
p p p b pppp pppppp p
ppp pppp2
p p p p
ppp p p p p p ppppppp p p pp pppp p p p
p p p p p ppppppppppppppppppppp pp ppp ppp
pppp pp ppppppp ppppppp ppp
pp p rp p p
pppp ppp pp pbpppp ppp pp pppp
p p ppp pppppp
p p p p ppppp pp pppppppppppppppppp pp prppp pp
p ppppppbp p
ppppp pppp
ppp ppp ppp p p p
p pppp p pp pp ppppp p p ppppppp p1
p p p p p ppp pppppp
p
pp p pp ppp p pp pppp bpppppppprpppppp pppp prpppp
p p p
pppppbpppp
pppp pp ppppppppp p pp p ppp pppp ppp ppppppppp pppppppppppppppppppppppppppppppp∗
pp p p p ppp p p p p ppppppppppp∗ pppppppppppp∗ ppppp∗ppppppp∗ pp pp pp pp pp pp ppp pp pp pp pp 4 ppp pppppp
0.0 ∗ 0.0 −1.0
0.0 log t 2.5 0.0 f 30.0
Figure 5: A—The relaxation enthalpy ∆H J/g versus time t days for semicrystalline PET
(Tg = 77 ◦ C) uniaxially stretched at T0 = 100 ◦ C and quenched to T = 40 ◦ C. Symbols:
experimental data [34]. Solid lines: prediction of the model with Σ̄0 = 0.12, ǫ = 0.2 and
Λ = 3.38. Curve 1: λ = 1.0; curve 2: λ = 2.0; curve 3: λ = 3.0; curve 4: λ = 4.0. B—The
dimensionless parameter Σ̄∞ (curve 1) and the parameter Γ̄ days−1 (curve 2) versus f %.
Circles: treatment of observations. Solid lines: approximation of the experimental data by
Eqs. (10) and (11) with a0 = 0.603, a1 = 0.018 and b0 = −0.868, b1 = 0.027

13