Available online at www.sciencedirect.

com
EUROPEAN
POLYMER
European Polymer Journal 43 (2007) 5107–5120
JOURNAL
www.elsevier.com/locate/europolj

Chemico-diffusion kinetics and TTT cure diagrams of DGEBA–

DGEBF/amine resins cured with phenol catalysts
F.X. Perrin *, Thi Minh Hanh Nguyen, J.L. Vernet
Laboratoire Matériaux à Finalités Spécifiques (UPRES no 1356), Université de Toulon et du Var, BP 132, 83957 La Garde Cedex, France

Received 12 October 2006; received in revised form 13 September 2007; accepted 27 September 2007
Available online 5 October 2007

Abstract

Curing of epoxy–amine resins with bisphenol A (BPA) as an external catalyst was studied from differential scanning
calorimetry analyses in isothermal and dynamic modes. Both phenomenological and mechanistic models have been tested.
The mechanistic model where epoxy cure is postulated to only occur through hydroxyl-catalyzed reactions, and assuming a
different reactivity of both types of hydroxyl groups (from BPA and epoxy chains) provided a reasonable fitting of the
whole set of experimental data. In particular, the latter model provides good predictive behavior for changes in the mixture
composition (BPA content varying in the range from 3 to 10 wt.%, relative to the weight of hardener), contrary to the
model based on the same reactivity of both types of hydroxyl groups.
The isothermal time–temperature–transformation (TTT) diagram including the time to vitrification and iso-Tg curves of
the complex epoxy system was also established.
 2007 Elsevier Ltd. All rights reserved.

Keywords: Epoxy; DSC; Time–temperature–transformation diagram; Mechanistic model; Kinetics

1. Introduction partial volatilisation; (iii) their low amine equivalent

weight that does not lend itself to slight errors in
Aliphatic amines have been used for ambient- component quantities nor to easy mixing ratios;
cure-temperature epoxy resin systems in applica- (iv) their high reactivity with atmospheric carbon
tions such as civil engineering, adhesives and archi- dioxide and moisture to form carbamates that
tectural and maintenance coatings. However, a wide may provoke loss of gloss, yellowing, or poor reco-
use of the low molecular weight polyamines (ethyl- atability and (v) the little impact resistance and poor
ene diamine, diethylene triamine, etc.) suffered from flexibility of the cured epoxy.
(i) their toxicity (skin sensitizer and respiratory irri- Interblending these amines with polyoxypropyl-
tant); (ii) their high vapor pressure that results in ene amine as a co-curing agent allows to circumvent
a great part of the above drawbacks. Polyoxyprop-
ylene diamine have a lower reactivity compared to
*
Corresponding author. that of low molecular weight aliphatic amines such
E-mail address: [email protected] (F.X. Perrin). as ethylene diamine or diethylene triamine. In case

0014-3057/$ - see front matter  2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.eurpolymj.2007.09.020
5108 F.X. Perrin et al. / European Polymer Journal 43 (2007) 5107–5120

mine D230 and diethylene triamine (DETA). The

H O
cure kinetics from DSC experimental data will be
O analysed through two different approaches: phe-
nomenological and mechanistic. Finally, the TTT
NHRR' diagram of the complex commercial epoxy–amine
Scheme 1. epoxy–amine addition with phenol catalyst. system will be calculated as it is a useful tool for
analysing and designing curing processes.
a relatively low curing time is desired (to reduce
production costs, for example), an external catalyst 2. Experimental procedure
can be added to the polypropylene diamine based
hardener. Amongst the potential candidates, 2.1. Materials
phenolic compounds are particularly attractive.
Indeed, their high hydrogen donor ability should The epoxy resin was a DGEBA/DGEBF
favour the opening of the epoxide ring by amine (65 wt.%) mixture kindly supplied by Rea Industrie
addition through the formation of an intermediate (France) with an average polymerization degree of
epoxy–hydroxyl complex as shown in Scheme 1. n = 0.08 and an epoxy equivalent weight of 174 g/
The same type of push–pull mechanism accounts mol as measured by titration. DGEBF is a low vis-
for the autocatalytic nature of the curing process cosity resin which is particularly useful for solvent-
observed with epoxy/amine systems, which is less or high solids/low VOC coatings or composites
related to the accumulation of hydroxyl groups dur- (where the finest viscosity is required to achieve a
ing curing. good fibre wetting). Blending DGEBF with DGEBA
Since the pioneering work by Schechter et al.[1], a prevents both resins from crystallization, making
great deal of effort has been made to study the kinet- them easier to handle at low temperature.
ics of epoxy–amine reactions [2–12]. However, to The hardener was a commercial mixture of
our knowledge, there are no previous studies related poly(oxypropylene diamine), Jeffamine D230, dieth-
to the cure modeling of epoxy–amine with phenol ylenetriamine (DETA) and bisphenol A (BPA). The
compound as external catalyst. chemical structures of these components and the 13C
The main objective of this work was to obtain a NMR chemical shifts are presented in Fig. 1. The
simple and reliable kinetic model that accounts for molecular weight of the poly(oxypropylene diamine)
the presence of an external catalyst in a complex (MJEFF), the weight percentage of each component
commercial formulation consisting of a diglycidyl and the amine equivalent weight (AEW) of the
ether of bisphenol F (DGEBF); diglycidyl ether of hardener were determined by quantitative 13C
bisphenol A (DGEBA); polypropylene oxide dia- NMR spectroscopy using the following equations:

18.9
77-78 73.5-75.0
45.4-45.9
75-76
46.0-46.3
H2N O
O x NH2

a b
77-78
18.9 16.5

40.6
51.6
154.6 140.7
H2N c NH2 HO
d OH
NH
30.5
41.0
114.3 126.9
13
Fig. 1. Chemical structures of the components present in the hardener used and C chemical shifts.
 F.X. Perrin et al. / European Polymer Journal 43 (2007) 5107–5120 5109

M JEFF ¼ 114  I b =I a þ 132 ð1Þ extraction with 2 · 20 mL of methanol under vigor-

I a M JEFF ous shaking (12 h). The combined extracts (40 mL)
%JEFF ¼ a  100% ð2Þ are concentrated prior to NMR analysis.
I M JEFF þ I c M DETA þ I d M BPA
I c M DETA
%DETA ¼ a  100% ð3Þ 2.3. Techniques
I M JEFF þ I c M DETA þ I d M BPA
%BPA ¼ 100  %JEFF  %DETA ð4Þ 1
H and13C NMR spectra were measured at room
100 temperature on a Brüker AV400 spectrometer oper-
AEW ¼ 4 5
ð5Þ
M JEFF
 % JEFF þ M DETA  %DETA ating at 400 MHz, using deuterated dimethylsulfox-
ide (DMSO-d6) as a solvent. Quantitative 13C
where the integrals I a, I b, I c and I d are the integrals
spectrum of the hardener was recorded using the
of the C a, C b, C c and C d carbons, respectively.
INVGATE Brüker sequence, with 90 orientation
MJEFF, MDETA and MBPA are the molecular weight
pulse length (9 ls), 15,000 Hz spectral width,
of JEFF, DETA and BPA. The calculated molecu-
16,000 data points, 1.36 s acquisition time, and a
lar weight of the polyoxypropylene diamine was
relaxation delay of 60 s: 600 scans were accumulated.
234 ± 3 g/mol which fairly agrees with the expected
The conversion of BPA after epoxy cure was
molecular weight for Jeffamine D230. The values of
determined from 1H NMR spectra of the residues
the weight percentage of each component was
obtained from the previously detailed extraction
%JEFF = 93.4%, %DETA = 3.5% and %BPA = 3.1%.
procedure adding naphthalene as an internal stan-
The latter value was in excellent agreement with
dard. The conversion of BPA, xBPA,was calculated
the BPA content determined from a liquid extrac-
by comparing the integral of the four aromatic pro-
tion procedure described elsewhere [13].
tons of BPA at 6.98 ppm (Ia) with the integral of the
The calculated AEW of the hardener was 57
four aromatic protons of naphthalene at 7.9 ppm
g/mol. From the EEW and AEW values, the stoichi-
(Ib):
ometric composition for this system was found to
correspond to a ratio of 33 parts of amine for 100 W N ðtÞ=W N ðt ¼ 0Þ I a ðtÞ=I a ðt ¼ 0Þ
xBPA ðtÞ ¼ 1  
parts of epoxy. These values are very close to the W R ðtÞ=W R ðt ¼ 0Þ I b ðtÞ=I b ðt ¼ 0Þ
values given in the technical data sheets for this ð6Þ
epoxy–amine system (34 parts of amine for 100
parts of epoxy). where WN is the weight of naphthalene and WR is
Bisphenol A (BPA) supplied by Aldrich was used the weight of the aliquot.
as received. Differential scanning calorimeter (TA Q-10 DSC
Instrument) was used to monitor the cure kinetics.
2.2. Sample preparation Samples of approximately 5–10 mg were placed in
hermetically sealed aluminium pans to prevent
Stoichiometric amounts of DGEBA/DGEBF evaporation of volatile reagents (DETA).
prepolymer and amine hardener were mixed at To determine the Tg as a function of fractional
room temperature for 1 min. Blends with different conversion, DSC samples were isothermally cured
catalyst content were also prepared adding a suit- in a oven at a specified cure temperature, Tcure
able amount of BPA in the DGEBA/DGEBF pre- (between 40 and 80 C) for various periods of time.
polymer at 60 C. After dissolution of BPA (in The cured samples were quenched in liquid nitrogen
few minutes) and cooling at room temperature, and then scanned at a rate of 10 C/min from 20 to
the hardener was then added with vigorous stirring 250 C to determine the Tg (taken here as the mid-
for 1 min. The as-prepared samples were immedi- point of the step transition) and the residual heat
ately used for analysis. of reaction DHr. In order to erase any physical age-
In this paper, the amount of BPA in the mixture ing effects that appear when Tg > Tcure, vitrified sam-
will be expressed as the weight percentage of BPA ples were first heated at a temperature just beyond
relative to the weight of hardener. the ageing peak, then quenched at liquid nitrogen
A simple extraction procedure was applied to temperature and finally re-scanned from 20 to
determine the extent of BPA conversion in cured 250 C at 10 C/min to determine Tg and DHr [14].
samples: it consists of a cryogenic grinding of about DSC in the temperature ramp conditions as
2 g of the gelled material, followed by a solid–liquid above was performed on the as-prepared blends to
5110 F.X. Perrin et al. / European Polymer Journal 43 (2007) 5107–5120

define the enthalpy for complete reaction, DHT (first The phenomenological Kamal model [15] is the
scanning) and the glass transition of the fully cured most popular for epoxy systems. The rate of conver-
system, Tg1 (second scanning). sion dx/dt and the extent of reaction x are related to
The conversion x of a partially cured sample was one another as follows:
calculated as
dx
x ¼ 1  DH r =DH T ð7Þ ¼ ðk 1 þ k 2 xm Þð1  xÞn ð10Þ
dt
In another series of experiments, isothermal runs
were conducted in the 40–80 C temperature range where k1 and k2 are specific rate constants with
with 10 C increments. Samples were inserted into Arrhenius temperature dependency and m and n
the DSC furnace, previously equilibrated at the de- are adjustable power exponents which have definite
sired temperature. The scan was stopped when the physical meaning in very limiting cases.
heat flow signal levelled off to the baseline. The reac- Eq. (10) accounts for an autocatalytic reaction in
tion rate, dx/dt, as a function of time has been cal- which the initial reaction rate is not zero. It is note-
culated through the equation worthy that despite the presence of BPA catalyst,
dx=dt ¼ ðdH =dtÞt =DH T ð8Þ the dx/dt versus x plots still show autocatalytic
kinetic behavior from the non-zero maximal reac-
where(dH/dt)t is the measured rate of heat flow. tion rate, thereby validating the use of Kamal model
The extent of reaction was calculated by to fit experimental data. However, it is noteworthy
that an autocatalyzed thermoset has typically its
x ¼ DH t =DH T ð9Þ
maximum heat evolution between 20% and 40%
where DHt is the heat generated at time t. conversion. The maximum rate for the present sys-
After the end of isothermal run at a temperature tem is observed at conversions around 10% which
Ti, the sample was re-scanned at a rate of 10 C/min accounts for the initial catalytic activity of BPA.
to determine the residual heat of reaction DHR(Ti) The graphic-analytical method presented by
and the attained glass transition temperature, Kenny [16] was applied to determine the kinetic
Tgmax(Ti). parameters of Eq. (10) from isothermal kinetic data
Gelation times were measured with a viscosity (see supporting information). Values of the four
follower which records the force that is required to kinetic parameters are listed in Table 1. The overall
move a solid body immersed in the changing liquid reaction order m + n is around 2 with the average
environment (Trombomat apparatus from Prode- value m = 0.25 and n = 1.62. The values of m are
mat). The solid body (100 mm long rod) has an lower than values found for epoxy systems that typ-
alternative horizontal motion with a defined magni- ically range above 0.3 [17–20]. This can be ascribed
tude and frequency. A force transducer linked to the to the specificity of the present system which con-
reciprocating rod provided a continuous measure of tains BPA as an external catalyst. Contradictory
the resistance to its’ movement. This force is the results are shown in the literature concerning the
image of the viscosity of the product. Here, the mea- temperature dependency of m and n parameters:
surement was performed under isothermal condi- for example, m was reported to remain constant
tions in a hemolyse-tube in order to lower the [21], increase [22,23] or decrease [24] when tempera-
effect of exothermy generated by the epoxy–amine ture increases depending on the nature of the epoxy
reaction. On the curve of apparent viscosity versus system. Here, m and n values are found to increase
time, the gel time was determined by the point of somewhat with increasing isothermal temperatures.
intersection of the tangent at the point of inflexion The change of m and n with T suggests that the
and the x-axis.
Table 1
3. Results and discussion Kinetic parameters of the Kamal equation
Temperature (C) k1 (105 s1) k2 (105 s1) m n
3.1. Phenomenological kinetic model 40 1.1 6.7 0.185 1.23
50 1.7 15 0.201 1.53
Both phenomenological and mechanistic models 60 3.4 34 0.288 1.67
have been tested to study the kinetics of the cure 70 6.5 69 0.292 1.75
reactions. 80 12.0 133 0.300 1.91
 F.X. Perrin et al. / European Polymer Journal 43 (2007) 5107–5120 5111

physical (or chemical) meaning of these parameters deed, the apparent rate constant k1 can be related to
have to be considered with care. the reaction catalyzed by BPA rather than to the
The relative contribution of autocatalytic and non-catalyzed or impurity-catalyzed reaction. As
non-autocatalytic reactions to the conversion rate seen later on, the development of relevant mechanis-
is given by tic models will confirm and clarify the latter state-
ment. Fig. 4 presents comparisons between
ðdx=dtÞautocat: k 2 m experimental data and predictions of the autocata-
¼ x ð11Þ
ðdx=dtÞnth k1 lytic model (Eq. (10)), with values of model param-
eters determined above. The Kamal model shows a
The plot of k2xm/k1 versus x at a given cure temper- good fit with experimental data up to about 80%
ature shows, as expected, that the contribution of conversion. The deviations observed in the latter
autocatalytic reaction to the epoxy conversion in- stage of the reaction can be associated to the onset
creases with the extent of reaction (Fig. 2). More of diffusion control.
interesting is the fact that the relative contribution To consider the diffusion effect, a semi-empirical
of autocatalytic reaction increases with isothermal relationship, based on free volume considerations
temperature at a given conversion, which is a direct was proposed by Chern and Poehlein [25]:
consequence of the respective activation energies 1
associated to rate constants k1 and k2 as seen below. fd ðxÞ ¼ ð12Þ
1 þ exp½Cðx  xc Þ
Fig. 3 shows that the reaction rate constants k1 and
k2 follow the Arrhenius relationship. The values of where C and xc are curve-fitted variables.
associated activation energies E1 and E2 are 55.9 For x  xc, the diffusion factor fd(x) approxi-
and 68.7 kJ/mol, respectively. The activation ener- mates unity and the effect of diffusion is negligible,
gies of the present system are in the same range than so that the reaction is kinetically controlled. As x
those reported in the literature. However, it is worth approaches xc, fd(x) begins to decrease and
mentioning that the activation energy associated to approaches zero as the reaction effectively stops.
the autocatalytic reaction, E2, is usually found to be A modification of the Chern and Poehlein
lower than E1 [17,18]. Here again, the opposite expression was proposed by Fournier et al. [26],
trend observed for the present system (E2 > E1) is introducing as a cutoff value the limiting conver-
associated to the presence of BPA in the blend. In- sion, xf:

15
80ºC
13 70ºC
60ºC
50ºC
11 40ºC

9
k2/lk1xm

-1 0 0.2 0.4 0.6 0.8 1

Fig. 2. Plots of k2/k1xm versus conversion x at different curing temperature.

5112 F.X. Perrin et al. / European Polymer Journal 43 (2007) 5107–5120

-4

-5
k1
-6 k2

-7
ln(k1),ln(k2)

y = -8267.7x+ 16.813
-8 R2= 0.9998

-9 y = -6722.2x+ 9.9489
R2= 0.9914
-10

-11

-12
0.0027 0.0029 0.0031 0.0033
1/T (K-1)

Fig. 3. Arrhenius plots of rate constants k1 and k2 from the Kamal model.

0.7
0.02

0.6
dx/dt.10 (s )
-1
3

0.5 0.01
dx/dt.103 (s-1)

0.4
0
0.9 0.95 1
0.3 experimental data x

Kamal model Eq.(10)

0.2
Kamal model with diffusion Eq.(12)
0.1
Kamal model with diffusion Eq.(13)

0
0 0.2 0.4 0.6 0.8 1
x

Fig. 4. Curve of reaction rate versus time with phenomenological model predictions for isothermal DSC runs at T = 80 C.

h x  x i1
f (12) and (13) are listed in Table 2. As it could be
fd ðxÞ ¼ 2 1 þ exp 1 ð13Þ
b anticipated, values of xc and xf are found to
increase with increasing curing temperature. A rea-
where b is an empirical parameter. sonable agreement is obtained between the curve-
In this study, fd(x) was determined as the ratio of fitted parameter xf and the final conversions deter-
the experimental reaction rate to the reaction rate mined from isothermal DSC runs. Consequently,
predicted on the basis of Kamal model in Eq. Eq. (13) can be treated as a single curve-fitted var-
(10). Values of xc, C, b, and xf obtained by applying iable, b, equation, as suggested by Fournier et al.
non-linear regression to fd(x) versus x data to Eqs. [26].
 F.X. Perrin et al. / European Polymer Journal 43 (2007) 5107–5120 5113

Table 2 epoxy groups (ie no substitution effect). This

Values of the critical conversion xc, parameter C, final conversion hypothesis is corroborated by the reactivity ratio
xf and parameter b at different temperatures
close to unity reported elsewhere for various ali-
Temperature (C) xc C xfa b phatic amines [29] and, notably for polyoxypropyl-
40 0.764 13.3 0.820 (0.79) 0.063 ene diamine D400 [30] whose structure is similar
50 0.796 23.8 0.840 (0.83) 0.066 to the major component (93.4 wt.%) of the hardener
60 0.824 45.3 0.867 (0.86) 0.037
70 0.894 89.7 0.917 (0.91) 0.019
used in this work.
80 0.901 79.8 0.928 (0.92) 0.022 Since the pioneering work by Smith [31], many
a researcher proposed kinetic models based on the
The data in brackets represent the final conversion determined
experimentally from isothermal DSC runs. activation of oxirane ring by hydrogen bonding
between the oxygen of the oxirane ring and any pro-
ton donor through a fast-pre-equilibrium step
Considering the diffusion effect, Eq. (10) can be [27,29,32–34]. Amine-hydroxyl complexes were also
generalized as follows: inferred to participate to the reaction [27]. However,
dx amine is certainly the proton acceptor in amine-
n
¼ ðk 1 þ k 2 xm Þð1  xÞ fd ðxÞ ð14Þ hydroxyl complexes, due to the higher acidity of
dt
hydroxyl groups. Consequently, the nucleophilic
Fig. 4 compares the experimental values with the character of amine in such a complex is lost and
data calculated by the Kamal model with diffusion its reaction with oxirane ring is prevented [29].
factor, according to Eqs. (12) and (13). Good agree- Amine–amine and hydroxyl–hydroxyl can also be
ment between experimental data and the autocata- neglected owing to a sort of a compensation effect,
lytic model with diffusion factor from Eq. (13) was that was already fully explained by Rozenberg [29].
found over the whole curing temperature range, According to the above discussion, two simple
compared to the model with diffusion factor from kinetic models referred as the single-equilibrium
Eq. (12). Contrary to the final conversion xf, the crit- model (SE model) and the multiple-equilibrium
ical conversion xc is not an observable quantity since model (ME model) have been tested.
the transition to the diffusion regime is gradual. The In the single-equilibrium model, the epoxy–
difficulties in determining xc with accuracy may ex- amine reaction is assumed to proceed either through
plain the better agreement of the model from Eq. the interaction between an epoxy–hydroxyl complex
(13) as compared to the model from Eq. (12) [26]. and an amino group (reaction (17)) or through an
impurity-catalyzed reaction (reaction (16)) as shown
3.2. Mechanistic kinetic models below
K
Due to the complexity of epoxy cure reaction, E þ OH E-OH ð15Þ
phenomenological models are preferred in studying K1
E þ A ! product þ OH ð16Þ
the epoxy cure processes. However, these models are
K2
limited to conditions (initial formulation, tempera- E-OH þ A ! product þ 2OH ð17Þ
ture, conversion range) that are similar to those
from which they are derived. In this respect, mech- where E, A, OH and E-OH denote epoxy group,
anistic models offer more flexibility and allow a bet- amine hydrogen, hydroxyl group (from BPA and
ter understanding of the network formation process. polyadduct) and complex between epoxy and hy-
Firstly, it is important to point out that in view of droxyl groups, respectively.
the low cure temperatures and stoichiometric condi- The same reactivity of both types of hydroxyl
tions, etherification and homopolymerization are groups, from BPA and polymer chains, are inferred
not significant[27,28] and can be neglected simplify- from the above reaction scheme. This mechanistic
ing the kinetic analysis. To confirm the lack of scheme has been already mathematically described
etherification reactions, unreacted BPA was quanti- [34] so we’ll give only the final set of equations here:
fied using 1H NMR spectroscopy (Eq. (6)) as
dx
detailed in the experimental part. BPA was found ¼ ð1  xÞ½k 01 ð1  x  yÞ þ k 02 y ð18Þ
to be totally recovered after epoxy curing. dt
Furthermore, primary and secondary amino A  ðA2  4½c0 þ xð1  c0 Þ  x2 Þ0:5
y¼ ð19Þ
hydrogens are assumed to have equal reactivity with 2
5114 F.X. Perrin et al. / European Polymer Journal 43 (2007) 5107–5120

with OH1 þ e-OH1 ¼ ðOH1 Þ0 ð30Þ

1 OH2 þ e-OH2 ¼ ðOH2 Þ0 þ eo x ð31Þ
A ¼ 1 þ c0 þ ð20Þ
K0
A new set of equations can be derived from Eqs.
where x is the conversion and c0 = (OH)0/e0,
(25)–(27) taking into account Eqs. (28)–(31)
K 0 = Ke0, k 01 = k1e0, k 02 = k2e0 and y = e-OH/e0.
The second proposed model is extended to explic-  0 0 
dx k 1 K 1 ð1  x  yÞ 0
itly include the BPA catalyst. In this ME model, the ¼ ð1  xÞ þ k2y ð32Þ
dt 1 þ K 01 ð1  x  yÞ
impurity-catalyzed reaction is neglected and epoxy y
cure is postulated to only occur through hydroxyl- K 02 ¼ ð33Þ
ð1  x  yÞðc0 þ x  yÞ
catalyzed reactions (reactions (23) and (24)), assum-
ing a different reactivity of both types of hydroxyl where c0 = (OH2)0/e0, y = e-OH2/e0, K 01 = K1e0,
groups, K 02 = K2e0, k 01 = k1(OH1)0 and k 02 = k2e0.
K1 The dependence of y versus x can be readily
E þ OH1 E-OH1 ð21Þ
derived from Eq. (33) and the expression is the same
K2
E þ OH2 E-OH2 ð22Þ as the one developed for the SE model except that
the equilibrium constant K 0 is replaced with K 02 :
K1
E þ OH1 þ A ! product þ OH þ OH1 þ OH2 ð23Þ
A  ðA2  4½c0 þ xð1  c0 Þ  x2 Þ0:5
K2 y¼ ð34Þ
E-OH2 þ A ! product þ 2OH2 ð24Þ 2
with
where the subscripts 1 and 2 denote the BPA-cata-
lyzed reaction route and the autocatalyzed reaction 1
A ¼ 1 þ c0 þ ð35Þ
route, respectively. K 02
This mechanistic scheme is mathematically For each conversion data from isothermal run, the
defined by the following set of equations: predicted dx/dt values were calculated for each
de model from Eqs (18)–(20) (SE model) and from
 ¼ k 1 ðe-OH1 Þa þ k 2 ðe-OH2 Þa ð25Þ
dt Eqs. (32), (34) and (35) (ME model) and compared
e-OH1 to the experimental dx/dt values.
K1 ¼ ð26Þ Fitting was realized by minimizing the sum of the
ðeÞðOH1 Þ
square error values between the experimental and
e-OH2 calculated values of the degree of conversion x.
K2 ¼ ð27Þ
ðeÞðOH2 Þ Only the experimental data preceding vitrification
The conversion in epoxy groups is defined by time were considered.
For the ME model, the determination of k 01 , k 02 ,
x ¼ ðe0  ðe þ e-OH1 þ e-OH2 Þ=e0 ð28Þ K 1 , K 02 from a four parameter fitting process without
0

From paths (24)–(27) and at a stoichiometric constraints on these parameters is a somewhat illu-
composition: sive task. That’s why a three parameter fitting pro-
cess with fixed parameter K 02 was preferred. The
a ¼ e þ e-OH1 þ e-OH2 ð29Þ
choice of K 02 as a fixed parameter was based on
e-OH1 in Eq. (29) can be neglected compared to two reasons: first, it appeared that the quality of
(e + e-OH2) during the whole course of the reaction: the fit was not very sensitive to K 02 compared to
at the initial time, this assumption is based on the the other three parameters. Second, the equilibrium
value of (OH1)0 (= 0.08 Eq l1) far lower than constant and its dependence with temperature for
e0(= 5.019 Eq l1); at more advanced stage of the complexation of isopropanol with phenylglycidyle-
reaction, this assumption is still valid due to the ther given by Rozenberg [29] can be taken as a fairly
accumulation of hydroxyl groups in epoxy chains reliable value of K2 in the present epoxy–amine sys-
that shift the equilibrium (22) toward the formation tem (Table 3). This infers from the fact that the
of e-OH2. structure and mean energies of the hydrogen bonds
The mass balance equations of the different spe- of epoxy–amine real polymer systems were shown to
cies, namely hydroxyl groups of BPA and epoxy be very much the same than in low molecular weight
chains, respectively give model compounds [29].
 F.X. Perrin et al. / European Polymer Journal 43 (2007) 5107–5120 5115

Table 3
Estimated parameters of mechanistic models as determined by isothermal DSC runs
SE model k1 k2 K
A1 E1 A2 E2 DH DS
3 5
3.9 · 10 51.9 7.04 · 10 62.3 25.9 74.9
ME model k1 k2 K1 K2 a
A1 E1 A2 E2 DH1 DS1 DH2 DS2
1.79 · 107 62.5 8.91 · 105 64.3 28.8 90.0 22.1 60.2
Preexponential factors Ai are in L/equiv s. Activation energies Eai are in kJ mol1. Enthalpy of the complexing reactions DHi are in
kJ mol1. Entropy of the complexing reactions DSi are in J mol1 K1.
a
Fixed parameter (from Ref. [29]).

As shown in Fig. 5, good agreement between isobutane-phenol [35] and DH = 22.1 kJ mol1
experimental data and the two mechanistic models and DS = 60.2 J mol1 K1 for isopropanol-phen-
was found, but some deviation attributed to vitrifi- ylglycidyl ether [29]). In the ME model, k1 values
cation was observed at high conversions. The anal- are far higher than k2 values (3–4 fold) in the whole
ysis of the diffusion controlled regime using the temperature range. Then, the more acidic hydroxyl
diffusion factor Eqs. (12) and (13) is omitted as it groups of BPA would be more efficient to activate
would provide no new insight compared to the the epoxy ring carbon atom than hydroxyl groups
results in Fig. 4. of epoxy chains.
It is also noteworthy that the ME model provides The values of activation energy obtained in this
a better agreement with data than the SE model in study agree reasonably well with those reported in
the early stages of the curing progress. Arrhenius- the literature[36]. For the SE model, the activation
type relationships were demonstrated for all param- energy of the hydroxyl-catalyzed reaction is higher
eters. The kinetic and thermodynamic parameters than that of the impurity-catalyzed reaction. The
are summarized in Table 3 for both models. The cal- same trend was obtained by Riccardi et al. [34] using
culated enthalpy and entropy of complexation are in the same model with a stoichiometric aliphatic
the same range as those reported in literature for amine based epoxy ((DGEBA/ethylene-diamine).
other epoxy–hydroxyl complexes (DH = 23.4 kJ For the ME model, reactions (23) and (24) are char-
mol1 and DS = 56.8 J mol1 K1 for 1,2-epoxy- acterized by nearly the same Ea values which could
be predicted as they both proceed by a push–pull
mechanism through the same type of epoxy–hydro-
0.7 80° C xyl–amine transition complex.
ME Model As noted before, an appropriate mechanistic
0.6 SE Model
model should provide good predictive behavior for
changes in the mixture composition. Then, the abil-
0.5 ity of the two models to describe the behavior of
stoichiometric epoxy–amine system made up with
dx/dt (103.s-1)

0.4 different proportions of BPA was tested. A fourth

order Runga–Kutta method was applied to Eqs.
0.3 (18) and (32) using the functions ki(T) and Ki(T)
determined from the previous fitting procedure, to
0.2 predict conversion during dynamic runs. Firstly,
Fig. 6 shows that the experimental curves shift
0.1 toward lower temperature when BPA concentration
increases. This confirms that BPA effectively cata-
0 lyzes epoxy–amine reaction. It is also clear from
0 1000 2000 3000 4000
Fig. 6 that the SE model is a bad model for purposes
Time/s
of prediction. In other words, the mechanism
Fig. 5. Curve of reaction rate versus time with mechanistic model depicted in reactions (15)–(17) that is based on the
predictions for isothermal DSC runs at T = 80 C. same reactivity of both types of hydroxyl groups,
5116 F.X. Perrin et al. / European Polymer Journal 43 (2007) 5107–5120

1
3 wt.-% BPA
0.9
7 wt.-% BPA

0.8 SE-model

0.7

0.6

0.5
x

0.4

0.3

0.2

0.1

0
0 50 100 150 200 250
T(°C)

1
3 wt.-% BPA
0.9 7 wt.-% BPA
ME model
0.8

0.7

0.6

0.5
x

0.4

0.3

0.2

0.1

0
0 50 100 150 200 250
T(°C)

Fig. 6. Dynamic conversion versus temperature profiles for stoichiometric DGEBA/DGEBF–amine mixtures with different BPA content.
Solid line represent the theoretical prediction of the single-equilibrium model and multiple-equilibrium model.

is not realistic. Conversely, the ME model predicts of adduct produced via the impurity-catalyzed reac-
successfully curves of curing for the whole range tion (16) and via the hydroxyl-catalyzed reaction
of BPA concentration employed, namely between (17) remain in the same order of magnitude for
3% and 10%. the 3–10 wt.% BPA range. These values can be com-
The relative amounts of reaction products pared with the literature data about stoichiometric
formed via the different routes of the two mechanis- DGEBA–ethylene diamine (EDA)[34] where the
tic models have been calculated for different compo- same mechanism is inferred: under isothermal cure
sitions of the reaction mixture (Table 4). Results conditions at 60 C, about 16% is formed via the
clearly show that the SE model is poorly sensitive impurity-catalyzed route whereas 84% is formed
to changes in BPA content: the relative amounts via the autocatalytic route. ME model is character-
 F.X. Perrin et al. / European Polymer Journal 43 (2007) 5107–5120 5117

Table 4
Comparison of the amount of products formed via the different routes of the single-equilibrium model and the multiple-equilibrium model
for isothermal cure at 60 C
BPA contenta 3% 5% 7% 10%
1 2 1 2 1 2 1 2
SE model, % 20.7 78.5 20.0 79.1 19.4 79.7 18.6 80.6
ME model, % 39.3 59.6 49.2 49.8 56.0 43.2 63.0 36.3
a
1 refers to reactions (16) of the SE model or (23) of the ME model. 2 refers to reactions (17) of the SE model or (24) of the ME model.

ized by totally different trends as shown by the dra- conversion, x. The one-to-one relationship between
matic impact of BPA concentration toward the rel- Tg and x evidences that there is no change in the
ative weight of each route of the model: the relative nature of the chemical reactions taking place in
weight of the reaction route including the epoxy– the temperature and BPA concentration ranges
BPA complex dramatically increases with BPA con- explored. Di Benedetto’s equation was used to
tent in the initial blend. The high catalytic efficiency model the master curve by treating k as an adjust-
of BPA is highlighted by the fact that comparable able structure-dependent parameter [38]:
amounts of adduct are produced via reactions (23)
ðT g1  T g0 Þkx
and (24) at BPA concentrations as low as 5%. T g ¼ T g0 þ ð36Þ
1  ð1  kÞx

3.3. Tg advancement and TTT diagram where Tg0, the Tg of the uncured material
(= 40 C) and Tg1, the maximum Tg (= 79 C)
The isothermal TTT cure diagram described by are both experimentally determined by DSC.
Gilham et al. is a very useful tool for describing The better fit was obtained with k = 0.44. The
the cure processes of epoxy systems [37]. It displays comparison of k and Tg0/Tg1 (= 0.66) indicates
in a synthetic way the various changes of material that, in this epoxy system, the ratio of lattice ener-
states occuring in the material during isothermal gies for a certain extent of reaction x with respect
cure such as vitrification and macroscopic gelation. to the mixture of monomers (x = 0), ex/em, is far
The glass transition temperature versus cure from unity [39].
advancement plot shown in Fig. 7 evidences the The gel times were determined under different iso-
one-to-one relationship between Tg and degree of thermal cure temperature (between 40 and 80 C)

90
experimental data at 3.wt% BPA
Dibenedetto model
70
10.wt% BPA
7.wt% BPA
50 5.wt% BPA

30
Tg (°C)

10

-10

-30

-50
0 0.2 0.4 0.6 0.8 1
x

Fig. 7. Relationship between Tg and the corresponding conversion degree, as calculated from the isothermal runs.
5118 F.X. Perrin et al. / European Polymer Journal 43 (2007) 5107–5120

from the apparent viscosity versus time curves. The A single apparent activation for the epoxy cure was
values for conversions at gel time were determined calculated from the gel time dependance with cure
by combining viscosity measurements and isother- temperature. This method is based on the assump-
mal DSC data. The values of conversions at gel point tion that the conversion at gel time is independent
were very close to each other giving a mean value of of the cure temperature. Under this assumption,
0.61 ± 0.02. Applying the gelation theory described the gel time measured at a given temperature is inver-
by Flory [40], the extent of reaction at the gel point, sely related to the apparent rate constant at this tem-
xgel, is defined as follows: perature. The gel times were found to follow the
 1=2 Arrhenius relationship as shown in Fig. 8. The
r apparent activation energy thus obtained is 51.3 kJ/
xgel ¼ ð37Þ
ðfe  1Þðfa  1 mol. This latter value is close to the values calculated
where fe and fa are the epoxy and amine functional- from the different kinetic models. The resulting mas-
ities and r is the epoxy/amine stoichiometric ratio. ter curve with Tref = 60 C and E = 51.3 kJ/mol is
For the present stoichiometric epoxy system, xgel shown in Fig. 9. A satisfactory superposition is ob-
is around 0.57. Application of the DiBenedetto tained up to the vitrification point, because near
Eq. (36), using the value x = 0.57 give the tempera- and above that point, diffusion limitations begin to
ture at which vitrification and gelation occur simul- influence the overall rate of the reaction.
taneously, Tggel = 4 C. Isoconversion (iso-Tg) contours were calculated
Time–temperature superposition principle of from the following relationship:
 
DSC data was used to calculate isoconversion con- E 1 1
tours and vitrification curve. Such an approach lntT c ¼ lntT ref þ  ð42Þ
R T c T ref
assumes that the reaction is only chemically con-
trolled with a single apparent activation energy where tT ref is the time, provided by the master curve,
[41,42]. Then, the reaction rate can be described to reach a particular Tg at cure temperature Tref and
by the usual kinetic rate equation: tTc is the time to reach the same Tg at a different
dx cure temperature Tc. The vitrification times tvit of
¼ kðT Þxf ðxÞ ð38Þ the vitrification curve were calculated using Eq.
dt
(42) with Tc = Tg. It must be noted that isoconver-
where k(T) is the reaction rate constant which de- sion (iso-Tg) contours and the vitrification curve
pends only on temperature and f(x) is some function were not determined by numerical integration of
of conversion. Integration of Eq. (38) at constant the kinetic model accounting for diffusion Eq. (14)
temperature and taking the natural logarithm will since the reaction orders m and n depend in an
give: undefined manner on cure temperature. The isocon-
Z x version curve x = 0.93 corresponds to the maximum
dx
ln ¼ ln kðT Þ þ lnt ð39Þ achievable experimental extent of reaction. The val-
0 f ðxÞ
ues measured for conversions at gel times are very
The left-hand side of Eq. (39) is only a function of
conversion and, thus, also a function of Tg: 10
F ðT g Þ ¼ ln kðT Þ þ lnt ð40Þ
9 y = 6166.7x - 10.633
Eq. (40) shows that for a chemically controlled reac- 2
tion, isothermal Tg versus log(time) curves at differ- R = 0.9952
ln(tg)

ent cure temperatures Tc could be superimposed by 8

simply shifting all curves horizontally by a constant
amount A(Tc), relative to a curve at fixed reference 7
temperature Tref:
AðT c Þ ¼ lntT ref  lntT c 6
  0.0028 0.0029 0.003 0.0031 0.0032 0.0033
E 1 1 1/T(K-1)
¼ ln kðT c Þ  ln kðT ref Þ ¼  
R T c T ref
Fig. 8. Natural log of the gel times versus 1/T used to calculate
ð41Þ the apparent activation energy of epoxy cure.
 F.X. Perrin et al. / European Polymer Journal 43 (2007) 5107–5120 5119

80

80ºC
60
60ºC
40ºC
40

20
Tg (°C)

-20

-40

-60
0 1 2 3 4 5 6 7 8
A(Tc) + ln(t)

Fig. 9. Superposition of the Tg versus ln time by horizontal shifting of the original curves for a reference temperature Tref = 60 C.

iso-conversional curves

100
0.93

80
0.91
Tg
8

0.9
Gelation 60
0.82

40

0.57
T(ºC)

20

Tggel 0.43
0
0.24

-20 0.13
Vitrification 0.05
Tg0
-40

-60
0 2 4 6 8 10
ln(t)

Fig. 10. Calculated TTT isothermal cure diagram for the epoxy system with 3% BPA. The dotted lines and solid lines correspond to
calculated curves while the individual data points represent experimental data.

closed to the theoretical value (0.57). For this rea- sonable agreement between the calculated gelation
son, the gelation curve was calculated as the isocon- curve and the experimentally determined gelation
version contour corresponding to x = 0.57 (the ideal times. The vitrification curve is S-shaped, as gener-
molecular gelation contour). Fig. 10 shows the rea- ally observed with thermosets [43].
5120 F.X. Perrin et al. / European Polymer Journal 43 (2007) 5107–5120

4. Conclusion [8] Wise CW, Cook WD, Goodwin AA. Polymer 1997;38:3251.
[9] Gonzalez M, Kindelan M, Cabanelas JC, Baselga J. Mac-
romol Symp 2003;200:111.
The catalytic effect of bisphenol A added in a [10] Vyazovkin S, Sbirrazzuoli N. Macromolecules 1996;29:1867.
complex epoxy–aliphatic amine mixture is clearly [11] Sbirrazzuoli N, Vyazovkin S, Mititelu A, Sladic C, Vincent
demonstrated from DSC data. In particular, L. Macromol Chem Phys 2003;204:1815.
although the autocatalytic character of epoxy– [12] Vyazovkin S, Sbirrazzuoli N. Macromol Rapid Commun
amine reaction is preserved, the maximum rate for 2006;27:1515.
[13] Perrin FX, Nguyen MH, Tran ML, Vernet JL. Polym Test
the present system is observed at a lower conversion 2006;25:912.
(around 10%), compared to other typical epoxy sys- [14] Wisanrakkit G, Gillham JK. J Appl Polym Sci 1990;41:2885.
tems without external catalysts (between 20% and [15] Kamal MR, Sourour S. Polym Eng Sci 1973;13:59.
40%). Moreover, the values of the kinetic parame- [16] Kenny JM. J Appl Polym Sci 1994;51:761.
ters extracted from the fitting of experimental data [17] Lopez J, Lopez-Bueno I, Nogueira P, Ramirez C, Abad MJ,
Barral L, Cano J. Polymer 2001;42:1669.
with the semi-empirical Kamal model agree with [18] Xie H, Liu B, Sun Q, Yuan Z, Shen J, Cheng R. J Appl
the presence of BPA in the epoxy system. Apart Polym Sci 2005;96:329.
from the Kamal phenomenological model, two sim- [19] Seung H, Kim WG, Yoon HG, Moon TJ. J Polym Sci Part
ple mechanistic kinetic models have been employed A 1998;36:773.
to fit experimental results in both isothermal and [20] Francis B, Lakshmana Rao V, Vanden Poel G, Posada F,
Groeninckx G, Ramaswamy R, Thomas S. Polymer
dynamic cure conditions. The kinetic model where 2006;47:5411.
oxirane rings are activated by two types of hydroxyl [21] Sourour S, Kamal MR. Thermochim Acta 1976;14:41.
groups with different reactivity, namely hydroxyl [22] Mijovic J, Kim J, Slaby J. J Appl Polym Sci 1984;29:1449.
from BPA and from polyadduct, through a fast- [23] Moroni A, Mijovic J, Pearce F, Foun CC. J Appl Polym Sci
pre-equilibrium appears to be the more realistic. 1986;32:3761.
[24] Boey MC, Qiang W. Polymer 2000;41:2081.
An excellent fitting of the kinetic controlled region [25] Chern CS, Poehlein GW. Polym Eng Sci 1987;27:782.
was obtained for all mixtures with different propor- [26] Fournier J, William SG, Duch C, Aldrige GA. Macromol-
tions of BPA in the range 3–10 wt.%, with a lonely ecules 1996;29:7097.
set of parameters in isothermal and dynamic cure [27] Mijovic J, Fishbain A, Wijaya A. Macromolecules
conditions. The TTT isothermal diagram of this 1992;25:979.
[28] Dusek K, Bleha M, Lunak S. J Polym Sci, Polym Chem Ed
DGEBF/DGEBA/jeffamine D230/DETA system 1977;15:2393.
with BPA catalyst has been constructed. This TTT [29] Rozenberg BA. Adv Polym Sci 1986;75:113.
diagram will allow the selection of optimal curing [30] Masood Siddiqi H, Dumon M, Eloundou JP, Pascault JP.
conditions for this epoxy system. Polymer 1996;37:4795.
[31] Smith IT. Polymer 1961;2:95.
[32] Xu L, Fu JH, Schlup JR. J Am Chem Soc 1994;116:2821.
Appendix A. Supplementary data [33] Flammersheim HJ. Thermochim Acta 1998;310:153.
[34] Riccardi CC, Fraga F, Dupuy J, Williams RJJ. J Appl
Supplementary data associated with this article Polym Sci 2001;82:2319.
can be found, in the online version, at doi:10. [35] West R, Powell DL, Lee MKT, Whatley LS. J Am Chem
1016/j.eurpolymj.2007.09.020. Soc 1964;86:3227.
[36] Horie K, Hiura H, Sawada M, Mita I, Kambe H. J Polym
Sci Part A 1970;8:1357.
References [37] Gillham JK. Polym Eng Sci 1986;26:1429.
[38] Pascault JP, Williams RJJ. J Polym Sci Polym Phys
[1] Schechter L, Wynstra J, Kurkjy RP. Ind Eng Chem 1990;28:85.
1956;48:94. [39] Nielsen LE. J Macromol Sci Revs Macromol Chem
[2] Barton JM. Adv Polym Sci 1985;72:111. 1969;C3:69.
[3] Zvetkov VL. Thermochim Acta 2005;435:71. [40] Flory PJ. Principles of polymer chemistry. NY: Ithaca;
[4] Swier S, Van Mele B. Macromolecules 2003;36:4424. 1953.
[5] Titier C, Pascault JP, Taha M, Rozenberg B. J Polym Sci [41] Gillham JK. Polym Eng Sci 1986;26:1429.
Part A 1995;33:175. [42] Simon SL, Gillham JK. J Appl Polym Sci 1993;47:461.
[6] Blanco M, Corcuera MA, Riccardi C, Mondragon I. [43] Chan LC, Naé HN, Gillham JK. J Appl Polym Sci
Polymer 2005;46:7989. 1984;29:3307.
[7] Zvetkov VL. Polymer 2002;43:1069.