Accepted Manuscript

Title: Analyzing the network formation and curing kinetics of

epoxy resins by in situ near-infrared measurements with
variable heating rates

Author: E. Duemichen M. Javdanitehran M. Erdmann V.

Trappe H. Sturm U. Braun G. Ziegmann

PII: S0040-6031(15)00316-0
DOI: http://dx.doi.org/doi:10.1016/j.tca.2015.08.008
Reference: TCA 77307

To appear in: Thermochimica Acta

Received date: 29-5-2015

Revised date: 3-8-2015
Accepted date: 5-8-2015

Please cite this article as: E. Duemichen, M. Javdanitehran, M. Erdmann, V. Trappe, H.

Sturm, U. Braun, G. Ziegmann, Analyzing the network formation and curing kinetics
of epoxy resins by in situ near-infrared measurements with variable heating rates,
Thermochimica Acta (2015), http://dx.doi.org/10.1016/j.tca.2015.08.008

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
*Highlights (for review)

Highlights

 Introduction of a new computer controlled NIR cell with temperatures up to 300 °C

 Evaluation of epoxy curing chemistry by NIR

 Fast and reliable kinetic evaluation with variable heating rates

t
ip
 Kinetic examination for isothermal and complex curing scenarios

cr
us
an
M
ed
pt
ce
Ac

Page 1 of 39
Graphical Abstract (for review)

i
cr
us
an
M
ed
pt
ce
Ac

Page 2 of 39
*Manuscript

Analyzing the network formation and curing kinetics of epoxy

resins by in situ near-infrared measurements with variable
heating rates

E. Duemichen,a M. Javdanitehran,b M. Erdmann,a V. Trappe,a H. Sturm,a,c U. Braun,a

G. Ziegmannd

t
ip
a - BAM Federal Institute of Material Research and Testing, Unter den Eichen 87, 12205

cr
Berlin, Germany, +49 30 8104 4317
b - Clausthal University of Technology, Institute of Polymer Materials and Plastics

us
Engineering, Agricolastr. 6, 38678 Clausthal-Zellerfeld, Germany
c - TU Berlin, Institute of Machine Tools and Factory Management (IWF), Pascalstr. 8-9,
10587 Berlin an
d - Clausthal University of Technology, Centre of Material Technology, Agricolastraße 2,
M
38678 Clausthal-Zellerfeld, Germany
ed

Keywords: epoxy resins, curing kinetics, near-infrared spectroscopy, DSC

pt

Corresponding author: ulrike.braun@bam.de

ce

Abstract
Near-infrared spectroscopy (NIR) turned out to be well suited for analyzing the degree of cure
Ac

for epoxy systems. In contrast to Dynamic Scanning Calorimetry (DSC), where the degree of
epoxy conversion is determined indirectly by the released heat of reaction, NIR spectroscopy
is able to determine the conversion directly by analyzing structural changes. Therefore a new
heatable NIR cell was equipped with an integrated thermocouple, which enables the real
sample temperature to be controlled and monitored in situ during epoxy curing. Dynamic
scans at different heating rates were used for kinetic modelling, to define kinetic parameters
and to predict real curing processes. The kinetic models and their parameters were validated
with an isothermal and a more complex multi-step curing scenario. Two available commercial
epoxy systems based on DGEBA were used with an anhydride and with an amine hardener.
NIR results were compared with DSC data. The simulated conversion predicted with a model
1
Page 3 of 39
fitted on the basis of NIR and DSC dynamic scans showed good agreement with the
conversion measured in the isothermal curing validation test.. Due to the proven reliability of
NIR in measuring the reaction progress of curing, it can be considered a versatile
measurement system for in situ monitoring of component production in the automotive,
aerospace and wind energy sectors.

t
1 Introduction

ip
Epoxy resins present amorphous thermosetting materials with excellent mechanical strength

cr
and toughness, outstanding chemical, moisture and corrosion resistance and good thermal,
adhesive and electrical properties [1]. They have been used as coatings and adhesives, and

us
especially as matrix material in composites combined with glass or carbon fibres [2]. The
most important commercial epoxy resin is the linear difunctionalized diglycidyl ether of
bisphenol A (DGEBA). Multifunctional hardeners are added to the resin to form a three-
an
dimensional network. Typical hardeners are primary amines [3-5] and anhydrides [6-9]. The
hardener type, the hardener concentration and the thermal curing procedure affect the curing
M
process of epoxy resins. Especially the degree of cure has a high impact on the physical,
mechanical and electrical properties of epoxy systems [10]. As the curing rate and extent of
polymerization are highly dependent on processing conditions [11], knowledge of curing
ed

kinetics is essential to adjust the material properties. During and after reaction, the degree of
curing can be determined using indirect and direct methods. Indirect methods are based on
pt

rheology [12, 13], the velocity of ultrasound [14, 15] or the release of reaction heat [5, 16];
the last, based on Differential Scanning Calorimetry (DSC) [5, 11, 16-21], turned out to be an
ce

especially reliable method for investigating curing kinetics. However, all of these methods
measure a cumulative parameter which can be affected by local side reactions, gelation,
Ac

vitrification or temperature changes.

The epoxy conversion can also be measured directly by structural changes using
spectroscopic methods like near-infrared (NIR) [4, 22-25], mid-infrared (MIR) [26-29] or
Raman [30-32] spectroscopy. In the NIR region bands are often difficult to evaluate due to
mode coupling and overtones, hence MIR spectroscopy is often preferred to analyse chemical
composition. However, particularly for investigating the curing of epoxy resins it turns out
that the sharpness and unambiguousness of the characteristic absorption peaks in the NIR are
more reliable [4, 22, 33]. In 1963 Dannenberg et al. pioneered the NIR region for the
quantitative analysis of epoxy resins [34]. A comprehensive comparison of NIR and MIR
spectroscopy evaluated with DSC and size exclusion chromatography (SEC) on epoxy curing
2
Page 4 of 39
was carried out by Poisson et al. in 1996 [4, 22], where they demonstrated the comparability
of NIR, DSC and SEC to determine the degree of cure, whereas the MIR systematically
underestimates the conversion. The comparability of NIR and DSC could be confirmed with
isothermal experiments by Pandita et al. [24]. Furthermore, measuring in the NIR rather than
the MIR region offers the possibility to measure samples with greater layer thicknesses (up to
several millimetres), using low-cost glass cuvettes and materials containing glass fibres,
because the glass fibres absorb the infrared irradiation above an excitation wavelength of five

t
ip
microns.
A study of kinetic model parameters for epoxy resins can be carried out by means of

cr
isothermal or dynamic experiments with variable heating rates. Isothermal experiments at
high temperatures are difficult to realize and inadquate due to a lack of information at the

us
beginning of measurement, caused by a fast start of curing and by sample handling.
Nevertheless, isothermal experiments are used in the literature almost exclusively to
an
determine kinetic parameters with NIR spectroscopy [35-37]. This is probably due to the
absence of a well-controlled heatable NIR cell with adjustable heating rates. Therefore, in this
work dynamic NIR measurements with temperatures from 30 °C up to about 250 °C with
M
defined variable heating rates were performed to determine kinetic models. Two
commercially available DGEBA epoxy resin systems were investigated in this study. The first
ed

system studied, cured with an anhydride curing agent, was Araldite® LY556 HY917 DY 070.
The second epoxy system was Epikote™ RIM 135/137, which is cured with an amine curing
agent. The experimental results were evaluated with the “Thermokinetics” software by
pt

Netzsch. Finally the kinetic models determined were validated and compared with isothermal
runs and multi-step curing scenarios.
ce

2 Experimental
Ac

2.1 NIR measurements

The experiments were performed with a new heated NIR Cell from Pike Technologies
(Madison, USA) which was modified by RESULTEC analytic equipment (Illerkirchberg,
Germany) and used in transmission mode. This cell was developed for heating samples from
room temperature up to 300 °C with a maximum heating rate of 10 K min-1. Cuvettes of
optical glass (Starna Scientific limited, Essex, England, type: 1/G/1), of 1 mm layer thickness
were used. The computer-controlled dynamic experiments were carried out using four
different heating rates of 1, 3, 5 and 10 K min-1 from 30 to 250 °C. NIR spectra were recorded
using a Nicolet 6700 FT-IR spectrometer (Nicolet Instruments, Offenbach, Germany)
3
Page 5 of 39
equipped with a white light source and a mercury cadmium telluride detector (MCT-A). The
measurements were performed within a wavelength range from 4000 to 7500 cm-1 and with a
resolution of 4 cm-1, averaging 16 scans. Spectra were recorded continuously with a time
interval of 15.6 s per spectrum by OMNIC software (Thermo Fischer Scientific, Karlsruhe,
Germany).
For sample preparation about 2 – 3 g of resin was weighed in a glass vial. Subsequently,
calculated amounts of hardener and accelerator (if required) were added and stirred for about

t
ip
60 s to achieve homogeneity. The mixture was injected with a syringe. A thermocouple was
plunged into the resin to enable precise temperature determination. The recording time

cr
interval of the spectrometer and the thermocouple were different. Therefore, the temperatures
were assigned to spectra by linear regression of the temperature time profiles.

us
For the validation experiments isothermal curing was performed at 80 °C. The thermocouple
was inserted after the desired temperature was reached. The NIR cell was placed in the
an
spectrometer and the experiment was started. A more complex multi-step curing scenario was
performed, beginning at 30 °C. Subsequently the sample was heated to 60 °C at a rate of 4 K
min-1, and this temperature maintained for 10 minutes. It was further heated to 230 °C at 6 K
M
min-1, followed by 10 minutes at 230 °C. For further evaluations only the sample temperatures
from the inserted thermocouple were used.
ed

2.2 DSC measurements

pt

A TA Q-2000 DSC from TA Instruments (New Castle, USA) is used to measure the
reaction’s enthalpy and the glass transition temperature (Tg) of the conducted isothermal and
ce

dynamic experiments. The attached cooling system RCS90 allows the samples to be cooled
down rapidly to -90 °C. Hermetic aluminium pans are used, and samples weighing 6 – 8 mg
are placed in the pans. Several samples are prepared in a single batch and stored at -18 °C in
Ac

order to interrupt the reaction. The samples are tested with 1, 3, 5 and 10 K min-1 heating
ramps, with the temperature rising from -25 °C (40 K below onset temperature) to 280 °C.
The TA Q-2000 was calibrated with indium and sapphire specimens. The heat flow,
temperature and time were recorded and used to calculate the degree of conversion and the
conversion rate.

2.3 Materials
Two kinds of commercial available epoxy systems were used, an amine-cured epoxy system
and an anhydride-cured epoxy system. An overview of the trade and IUPAC names, chemical

4
Page 6 of 39
structures and composition of epoxy mixtures is given in Table 1 (all information is taken
from material data sheets).
The first resin system investigated was Araldite® LY556, which consists of DGEBA. It was
cured with the curing agent HY917, a methyltetrahydrophthalic anhydride (MTHPA). The
reaction was accelerated by DY 070, a 1-methylimidazole. They were mixed at a weight ratio
of 100:90:2 respectively. The anhydride cured epoxy system was purchased from Ciba
Specialty Chemicals Inc..

t
ip
The other resin, EpikoteTM MGSTM RIMR 135 was a mix of DGEBA and the aliphatic epoxy
1,6-hexanediol diglycidyl ether. It was cured with the EpikureTM Curing Agent MGSTM RIM

cr
H 137, an aliphatic primary amine mixture of isophorondiamine and alkyletheramine. The
weight ratio of resin and hardener was 100:30. The amine cured epoxy system was provided

us
by Hexion Inc..

2.4 Cure kinetics an

The reaction rate r for a conversion reaction from an educt E to a product P can be described
in a general form by two separable functions k and f (Equation 1). Here p is the concentration
M
of the product, T the temperature and t the time. The function k is called reaction rate and
depends on the temperature. The function f depends on the concentration of the product as
ed

well as on the type of reaction (reversible or irreversible, first, second or nth order, catalyzed
or non-catalyzed, etc.).
pt

dp
 r (t , T , p)  k (T (t ))  f ( p) (1)
dt
The product concentration p can be replaced by the conversion degree α, which is defined as
ce

α = (A0 – A(t))/(A0 - A  ) with A(t) the actual product signal at time t, A0 for the product signal
Ac

at the beginning, and A  the product signal at the end of the conversion. Here product signals
are determined areas between a recorded peak or curve and a baseline. For the NIR
measurements an integrated area of an epoxy absorption peak, and for the DSC measurements
the generated heat of reaction were used as product signals. The temperature dependence of
the reaction rate can be expressed with the Arrhenius equation
k = k0 exp(-EA/RT), with activation energy EA, ideal gas constant R and pre-exponential factor
k0. In addition, the heating rate β = dT/dt can be inserted. This leads to Equation 2.
d 1
  k0  e(  EA / RT )  f ( ) (2)
dT 

5
Page 7 of 39
For model-free calculations the function f is set to 1 and leads to the Arrhenius model. Taking
the logarithm of both sides and solving equation 2 for log β yields the linear Equation 3.
dT EA 1
log   log(  k0 )   (3)
d 2,3026  R T
By plotting log β versus 1/T it is possible to determine the activation energy as the slope and
the pre-exponential factor as the intercept with the log β axis, independent of the type of
reaction.

t
ip
Using Friedmann analysis [38] and Ozawa-Flynn-Wall analysis [39, 40], which will not be
discussed in detail here, it is possible to determine iso-conversional lines. The Friedman

cr
method is well suited to evaluate whether the reaction is accelerated or not by comparing the
slope of the iso-conversion lines with the experimental data. A slope of the iso-conversional

us
lines that is steeper than the experimental data at the beginning of the reaction indicates an
accelerated reaction. With Ozawa-Flynn-Wall analysis it is possible to determine whether a
an
reaction is a one-step or multi-step reaction by comparing the slope of the iso-conversional
lines. Different slopes of the iso-conversional lines indicate a multi-step reaction. Using the
slopes and intercepts of the iso-conversional lines, the activation energies and pre-exponential
M
factors can be calculated depending on the conversion degree for both analysis methods.
Monitoring the activation energy over the conversion degree further helps to find a suitable
ed

reaction model.
For model-based examinations, the term f(α) in equation 2 is replaced by a complex reaction-
type dependent equation which can be solved numerically. For evaluation of kinetic
pt

parameters (model-based and model-free), NETZSCH Thermokinetics software (NETZSCH

Thermokinetics 3.1, NETZSCH GmbH & Co. Holding KG, Selb, Germany) was used. It has
ce

been described in detail by Opfermann and Hädrich [41] and used successfully in past studies
[42, 43]. The core of the software is a multivariate non-linear regression based on the hybrid
Ac

Levenberg-Marquardt algorithm for solving equations numerically [43].

3 Results and Discussion

3.1 General curing mechanisms
The curing mechanism for epoxy-anhydride formulations with tertiary amines (in our case
imines) is quite complex due to the multiplicity of possible competitive reactions (Figure 2). It
is known that the initiation step of the reaction is a nucleophilic attack on the epoxy or the
anhydride group by the tertitary amine, resulting in zwitter ions with an alkoxide anion (top of
Figure 2, reaction A) or in an carboxylate anion (top of Figure 2, reaction B), respectively [6-
6
Page 8 of 39
8]. Matejka et al. found evidence for the reaction of the tertiary amine to the epoxy group by
NMR spectroscopy [44]. Therefore, it is assumed here that this reaction is more likely. The
propagation starts with a nucleophilic attack on an anhydride by the alkoxide anion (middle of
Figure 2), resulting in the formation of a carboxyl anion, which is less reactive due to charge
stabilization by resonance [8]. Subsequently the carboxyl anion can launch a nucleophilic
attack on a further epoxy group, yielding an alkoxide anion (starting group), resulting in the
formation of an alternating epoxy-anhydride network. Deviating from the presented

t
ip
mechanism, a possible propagation can occur through a nucleophilic attack by an alkoxide
anion on a further epoxy group, which is less likely, unless there is an excess of epoxy groups

cr
[7]. In addition, the formation of a carboxyl anion is strongly favoured due to resonance
stabilization. The auto-catalyzed character of the reaction implied that the tertiary amine was

us
regenerated. A possible regeneration is given by an elimination reaction (bottom of Figure 2).
Intermediate formation of alkoxide or carboxyl anions (by initiation or propagation) can
an
abstract β-hydrogen, producing a double bond between the tertiary amine and an acid or
alcohol. The alcohol or acid can accelerate the reaction, as for curing with primary amines
(see Figure 3). Although regeneration leads to an acceleration of the reaction, it can also result
M
in the termination of propagation by protonating the anions. However, formed hydroxyl
groups are also able to induce a polymerisation of epoxy-anhydride formulations through
ed

polyesterification [6].
For amine-cured epoxy systems, the first consistent mechanism was published by Schechter et
al. [45]. He suggested a push-pull mechanism with a termolecular intermediate state (in the
pt

sense of an intermediate product consisting of three molecules) between the hydroxyl, the
epoxy and the amine groups. The sigmoid concentration curve progression they observed was
ce

probably caused by formed hydroxyl groups which auto-catalyze the reaction. This
mechanism was refined by Smith et al. [46], who propose a hydrogen bond between a donor
Ac

molecule and an epoxy group (Figure 3), leading to weaker H-X and C-O-bonds. Now a
bimolecular reaction (hydrogen-bonded epoxy group with amine) leads to a termolecular
transition state which is proposed to be the rate-controlling step. This mechanism shows
overall acceptance in the literature [35, 37, 47]. When X in Figure 3 is an oxygen atom the
reaction is called catalysed; if X is a nitrogen atom originating from the amine hardener the
reaction is called non-catalysed [36, 37]. The fact that proton donors, especially water,
accelerate the curing process was confirmed by many authors [35, 45]. Side reactions of the
epoxy groups, like homopolymerization or etherification, are disregarded because they would
require higher reaction temperatures or catalysts [37]. The reactivity of secondary amines is

7
Page 9 of 39
discussed controversially in the literature; however, it is important for the morphology and the
network formation. The reactivity of the secondary amines, called the substitution effect, is
determined by comparing the reaction-rate constants of an epoxy group reacting with a
secondary amine (ksec) and with a primary amine (kprim). A ratio (ksec/kprim) of 0.5 means that
both hydrogen atoms (in the primary and secondary amine) show the same reactivities.
Various values between 0.05 and 1.12 were given by Rozenberg and Mijovic et al. for several
amine-cured epoxy systems under different reaction conditions [47, 48]. Thus no general

t
ip
conclusion can be drawn and the curing mechanism must be examined case-by-case.

cr
3.2 Spectroscopic monitoring (NIR)
Figure 4 shows our NIR measurements during the curing process. The figure illustrates an

us
overlay of selected, representative NIR spectra in the region between 4450 – 7500 cm-1 for
DGEBA-based resins with the anhydride (top) and amine (below) hardener at different
degrees of curing with rising temperature. an
In both the amine-cured and the anhydride-cured systems high intensive absorption bands can
be observed between 4552 – 4497 cm-1, originating from combination vibrations of C-H
M
stretching and CH2 deformation vibrations of the epoxy groups (Epoxy 1) [4, 34]. They
decrease with increasing temperature until they disappear completely. For both systems a less
ed

intensive epoxy absorption band of C-H stretching vibrations was also found between 6149 –
6026 cm-1 (Epoxy 2) [22]. Both epoxy peaks show similar behaviour with increasing
pt

temperature. Aromatic absorption bands of conjugated C=C stretching with C-H fundamental
stretching vibrations [22] are present between 5090 – 4651 cm-1 in both systems, with a minor
ce

change during temperature rise.

The amine-cured system has two further absorption bands arising from the amine groups
between 5090 – 4863 cm-1 and 6638 – 6371 cm-1, respectively. Both absorption bands
Ac

correspond to overtone combination vibrations of primary and secondary amines (NH 1 and
NH 2) [49] and decrease with temperature. Finally, for the amine-cured system a broad
absorption band is formed between 7158 – 6638 cm-1 with increasing temperature and time,
corresponding to OH stretching vibrations [34, 49] that overlay the amine absorption band at
6638 – 6371 cm-1. A summary of the identified absorption bands is presented in Table 2.
The epoxy conversion can be determined by the disappearance of the epoxy absorption bands
(Epoxy 1/2), the disappearance of the amine absorption (NH 1/2) bands, and by the formation
of the hydroxyl (OH) absorption peaks. All characteristic absorptions peaks were normalized
and integrated by linear baselines with the same limits as the absorption regions in Table 2.

8
Page 10 of 39
Figure 5 presents the conversion curves calculated over time for the systems cured with
anhydride (left) and amine (right). All conversion curves show a sigmoid curve progression.
For the anhydride-cured system the epoxy conversion curves (Epoxy 1 and 2) are not
congruent. The Epoxy 2 conversion curve starts to decline earlier than that for Epoxy 1.
Regarding the overlay spectra (left) for the anhydride system (Figure 4), it can be seen that for
the integration of the Epoxy 2 band it is difficult to set accurate parameters for the integration
limits and for the baseline due to the overlay of the Epoxy 2 vibrations with several C-H

t
ip
combination vibrations. Hence a possible interference by C-H overtones may also be
responsible for the non-congruent progression. Therefore Epoxy 1 shows more reliable

cr
results. However, for the amine-cured system the Epoxy 1 and 2 conversion rates show
congruent progression. Theoretically the amine conversion rates (NH 1/2) and epoxy

us
conversion rates should show the same progression. Nevertheless, the conversion curve of NH
1 declines faster than that of NH 2. The formation of another absorption band next to the NH
an
1 absorption band leads to an overlay that causes a shift in the baseline, resulting in a faster-
shrinking area. In contrast, the formation of the hydroxyl absorption band next to the NH 2
absorption band leads to an addition of area, causing a delayed decline that rises with
M
increasing time/temperature.
The Epoxy 1 band turns out to be the best for the epoxy conversion evaluation and was used
ed

for the kinetic modelling of curing. For an accurate integration of the epoxy band the
integration parameters must be set carefully. With increasing degree of curing, a shoulder
(about 4580 – 4550 cm-1) occurs beside the Epoxy 1 band. This phenomenon appears for the
pt

system cured with amine as well as for the anhydride-cured system. Therefore, the Epoxy 1
band was evaluated in the region 4452 – 4497 cm-1. Due to the high number of spectra in each
ce

experiment, no deconvolution (e.g. Gauss, Voigt or Lorentz functions) yielding improved

integrations [24] was possible with the equipment used.
Ac

It is known that the signals of vibrational spectroscopic measurements are influenced by

temperature [50, 51]. In order to investigate the dependence of temperature on the absorption
bands, pure DGEBA was heated from 30 °C up to 90 °C at 10 K min-1 and cooled down to
30 °C. Pure DGEBA was chosen instead of resin mixture to minimise any effects of chemical
reaction. In Figure 6 the integrated areas of the absorption bands of Epoxy 1, Aromatic 1 and
Aromatic 2 are shown in comparison to the sample temperature profile. All absorption bands
show a reduced peak area with increasing temperature, which recovered as the sample cooled
back down. Thus it is additionally proved that the rise in applied temperature did not cause
homopolymerization resulting in a disappearance of epoxy groups. The reduced area with

9
Page 11 of 39
increasing temperature can be attributed, on the one hand, to a reduced concentration due to
density and viscosity changes, and/or to intrinsic vibrational effects [52]. The synchronous
(proportional) progressions of the Epoxy 1 and Aromatic 1 signals with temperature enabled
the aromatic signals to be used as reference signals. Thus, the epoxy conversion (αepoxy,t) was
calculated according to Equation 4 by normalizing the areas of the Epoxy 1 (AEpoxy1) to the
Aromatic 1 (AAroamtic1) absorption bands. All determined conversion curves were baseline-
corrected.

t
ip
( AEpox1 / AAromatic1 )ti
 epoxy ,t  1  (4)
( AEpox1 / AAromatic1 )t 0

cr
For both epoxy systems, the normalized epoxy conversions of the different heating rates
depending on the temperature are shown in Figure 7. In both cases the epoxy conversions

us
proceed as a sigmoidal curve. Increasing heating rates shift the epoxy conversion to higher
temperatures. For the anhydride system the curve of 1 K min-1 shows a small deviation at
an
about 110 °C; a flattening of the curve could be observed. This is possibly caused by diffusion
effects (vitrification). Therefore, for the further kinetic calculation the 1 K min-1 conversion
curve was excluded for the anhydride-cured system.
M

3.3 Calorimetric monitoring (DSC)

ed

DSC is an established method to calculate the kinetic parameters of epoxy systems. In the
past, isothermal experiments [5, 11, 53] were run as well as dynamic experiments with
pt

variable heating rates [16-19, 54]. For direct comparison DSC measurements were thus
performed with dynamic heating rates approximately equal to those used in the NIR
ce

experiments. The linear baseline-corrected DSC curves of the anhydride-cured (left) and the
amine-cured (right) systems are presented in Figure 8. With decreasing heating rate, the peak
Ac

maximum shifted to lower temperatures, as did the peak area. This is a disadvantage of DSC
and led to problems in evaluating DSC data for low heating rates, where low sensitivity
makes it difficult to distinguish between noise and the enthalpy released. By comparison, the
sensitivity of NIR measurements is independent of the heating rate.

3.4 Kinetic evaluation

In the following section a model-free determination of curing kinetics parameters is carried
out for both epoxy systems using the NIR as well as the DSC experimental data. Subsequently
a proper chemical reaction model is added for f(α).

10
Page 12 of 39
First the activation energies and pre-exponential factors were calculated without a chemical
reaction model for the NIR (Figure 9) or the DSC (Figure 10) measurements for both epoxy
systems, by plotting log (β) over 1000 1/T. In accordance with ASTM E698, the temperatures
were chosen at the points of inflexion of epoxy conversion curves. The activation energies
and pre-exponential factors were calculated via the slope and intercept of linear regressions
according to Equation 3.
For the NIR measurements both linear regressions do not fit very well (Figure 9). A

t
ip
correlation coefficient of R = 0.9241 was determined for the anhydride-cured system and R =
0.9781 for the amine-cured system. For the anhydride-cured system an activation energy of

cr
EA = 75.1 ± 14.6 kJ mol-1 was calculated and a pre-exponential factor of log k0 = 7.26. The
activation energy for the amine-cured system was EA = 56.4 ± 4.7 kJ mol-1 and the pre-

us
exponential factor was log k0 = 4.90.
By measuring DSC, an activation energy of EA = 77.7 ± 2.6 kJ mol-1 and a pre-exponential
an
factor of log k0 = 7.71 were determined for the anhydride-cured system (Figure 10). The
linear regression coefficient was 0.9968. In comparison to the NIR results the values were
quite similar. The same could be observed for the amine-cured system with an activation
M
energy of EA = 57.1 ± 2.8 kJ mol-1 and a pre-exponential factor of log k0 = 5.11 (Figure 10,
right). The linear regression coefficient of R = 0.9929 shows a slightly higher deviation from
ed

perfect linearity.
The Friedman and Ozawa-Flynn-Wall analyses (not shown here) of both epoxy systems for
the NIR as well as for the DSC measurements suggested that it might be an auto-catalyzed
pt

multi-step reaction. A small variation in the activation energies, depending on the conversion
in both the Friedman and the Ozawa-Flynn-Wall analyses, also shows multi-step reactions for
ce

both epoxy systems and measurement techniques. Summarized for the kinetic calculations
without a reaction model, the linear regressions of the Arrhenius plots as well as the Friedman
Ac

and Ozawa-Flynn-Wall analyses indicate that a more complex model is needed to describe the
curing of the epoxy systems. Therefore a reaction-dependent equation was added for the term
f(α) in the kinetic equations. Through pre-tests we found out that choosing complex multi-step
reactions often leads to better progressions of the values measured using dynamic
measurements, but fail when one attempts to predict multi-step curing scenarios and
isothermal measurements. Complex multi-step reactions and the values of the parameters
often have no connection to the real chemical reactions. However, realistic equations and
values for varying reaction conditions are needed for good simulations. Therefore it was

11
Page 13 of 39
attempted to keep the reaction models as simple as possible and keep the parameters in
realistic dimensions.
An acceptable accordance for both epoxy systems was given by an auto-catalyzed one-step
reaction of the nth order. The thermokinetics software uses the following equation for such a
reaction (Equation 5).
 B with f ( )  (1   )n  (1  kcat   )
A  (5)

t
Where n is the reaction order and kcat the autocatalytic factor of the system.

ip
Equation 5 inserted in equation 2 yields:
d 1
  k0  e(  EA / RT )  (1   )n  (1  kcat   )

cr
. (6)
dT 
Therefore the four parameters EA, k0, n and kcat were determined with the software, which give

us
curve progressions that were in good accordance with the real data. Substituting dT with β · dt
in Equation 6 and summarizing the temperature-dependent reaction constant as k1 yields
Equation 7.
d
an
 k1  (1   )n  (1  kcat   ) (7)
dt
M
Assuming that k1·kcat gives a new constant, k2 yields Equation 8, which is very similar to the
Kamal-Sourour equation.
ed

d
 (k1  k2   ) (1   ) n (8)
dt
pt

The kinetic parameters determined by the software for the nth-order auto-catalyzed reaction
ce

model for both epoxy systems are summarized for the NIR as well as the DSC measurements
in Table 3. The fractional number for the reaction orders indicates complex reactions that
Ac

cannot be described simply.

For the NIR measurements the epoxy conversion curves (straight lines) determined by the nth-
order auto-catalyzed reaction model are plotted against measured values (symbols) for both
epoxy systems in Figure 11. For each system the simulated curves were in good accordance
with the progressions of the measured values. For both epoxy systems the determined
activation energies and pre-exponential factors (Table 3) deviate slightly from the ones
calculated without a reaction model. However, the addition of the reaction order as well as the
auto-catalytic factor leads to a better regression. This is shown by the good correlation
coefficient for the anhydride-cured and the amine-cured systems of 0.9998 and 0.9997,
respectively.
12
Page 14 of 39
In Figure 12 the DSC curves are presented for both epoxy systems of the simulated (straight
lines, determined by the autocatalytic model) as well as of the measured data values
(symbols). For both epoxy systems the simulations were in good accordance with the
measured data. This can be seen in the correlation coefficients for the anhydride-cured and the
amine-cured systems of 0.9989 and 0.9972, respectively, which are better than for the kinetic
calculations without a reaction model. Except for the slightly higher activation energy of the
amine-cured system in comparison to the model-free kinetic calculations, the activation

t
ip
energies and pre-exponential factors hardly differ. However, introducing the new parameters
(reaction order, auto-catalytic factor) to the basic model of kinetic calculations led to a much

cr
better accordance for both epoxy systems.

us
3.5 Verification of the kinetic model
An isothermal curing run was performed at 80 °C. The multi-step curing scenario started from
an
30 °C and was heated to 60 °C at 4 K min-1, where the temperature was held for 10 minutes.
Subsequently the temperature was set to 230 °C at a heating rate of 6 K min-1 and held there
for a further 10 minutes before passive cooling to room temperature. NIR conversion rates
M
were determined as described above. During the curing process the actual sample
temperatures were monitored by the internal thermocouple. These profiles were segmented
ed

into several linear and isothermal temperature steps for the simulations. With these developed
temperature profiles, simulations of the epoxy conversion of the NIR and DSC one-step auto-
pt

catalytic reaction models were calculated by the thermokinetics software. The simulated NIR
and DSC conversion rates were presented in comparison to the measured epoxy conversion
ce

rates determined by NIR. The sample temperature profiles were also plotted over to the
simulated temperature profiles.
The results for the anhydride-cured system are presented in Figure 13. At the beginning of
Ac

the isothermal experiment (left) the temperature rises very fast, with a maximum heating rate
of about 60 K min-1. This heating rate is far from the heating rates used to develop the model.
After 5 minutes a constant sample temperature of 79 °C is reached. The simulated NIR
conversion curve up to 40 minutes shows a slope smaller than the measured conversion curve.
Afterward the slope rises and the simulated NIR conversion curve is in good accordance with
the real conversion. In contrast, the simulated DSC conversion curve shows a steeper rise at
the beginning, leading to increased deviation over time.
For the multi-step scenario with moderate heating rates of about 3 and 6 K min-1, both
simulated models yielded excellent accordance with the real conversion. Only minor

13
Page 15 of 39
deviations can be seen between 0 and 30 minutes. The simulated conversion using DSC data
shows a slightly premature rise in conversion, whereas the simulated conversion using NIR
data is slightly delayed. However, after 40 minutes the deviations of both simulated curves
from the measured conversion rate is less than 1 %.
For the amine-cured system, the results to verify the auto-catalytic reaction model of the nth
order by an isothermal (left) and a multi-step curing scenario are presented in Figure 14.
In the isothermal experiment (left), during the first 20 minutes the DSC model is in good

t
ip
agreement with the measured conversion, followed by a delayed conversion with increasing
deviation of up to 10 %. In contrast, the simulated conversion using the NIR model differs

cr
slightly at the beginning and reduces with increasing time. At the end there is only a small
deviation of about 2 %. Therefore, the kinetic model determined by NIR leads to better

us
accordance for the isothermal curing scenario than does DSC.
In the multi-step curing scenario with moderate heating rates (right), the conversion
an
determined by the DSC model differs only at the beginning (up to 27 minutes) resulting in a
higher degree of conversion. Afterwards the model corresponds well to the real conversion. In
comparison, the conversion determined by the NIR model shows an almost perfect match,
M
with a deviation of less than 1 % at the end.
Summarizing the isothermal experiments, both simulated models deviate from the real
ed

conversion, possibly caused by high heating rates at the beginning which are far from the
heating rates used for modelling. However, deviations of the NIR model are acceptable and
lead to reliable predictions. For the multi-step curing scenario, moderate heating rates were
pt

used, yielding much better results for both systems. Although the DSC model shows a small
deviation at the beginning of the amine-cured system, both models lead to curve progressions
ce

which are in good accordance with the measured data.

It could be proved that the new NIR cell with precise temperature control is well suited for
Ac

monitoring the structural changes during curing by tracing the epoxy band in the NIR region.
In addition, reliable kinetic models could be developed by varying heating rates, with no lack
of information at the beginning of the measurements as in the isothermal experiments. These
models were comparable with the reaction models determined by DSC and suited to predict
isothermal and complex curing scenarios. The sensitivity of the NIR measurements is
independent of the heating rate. In contrast, in DSC slow heating rates are difficult to apply
due to poor sensitivity. Therefore, monitoring the network formation by NIR is advantageous,
especially for slow curing processes.

14
Page 16 of 39
4. Conclusion
In this work a new, precise heatable cell was introduced for the controlled curing of
thermosetting materials and monitoring the structural changes in the NIR region. Kinetic
calculations with and without a reaction model were developed by measuring the epoxy
conversion with different heating rates from room temperature up to 250 °C. As a reference
method DSC was used with the same experimental procedures. The experiments were carried
out with two different commercially available DGEBA-based epoxy systems, one with an

t
ip
amine and the other one with an anhydride hardener. The determined reaction models and
kinetic parameters were verified by isothermal and multi-step curing scenarios with the NIR

cr
cell for both epoxy systems. The most reliable results for the isothermal experiments were
given by the model based on the NIR experiments. For multi-step curing scenarios with

us
moderate heating rates both models lead to good results in accordance with the conversions
measured for both epoxy systems. In contrast to DSC, the sensitivity of the NIR
an
measurements is independent of the heating rate and therefore shows advantages.
It was demonstrated that NIR spectroscopy is well suited for the in situ analysis of the
network formation of epoxy systems by direct monitoring of structural changes. In addition,
M
kinetic parameters as well as reaction models can be determined by dynamic measurements
with variable heating rates, yielding excellent results in accordance with measured
ed

conversions. The NIR experiments were carried out in transmission here. Alternative
techniques using NIR to determine epoxy conversion on the surface of epoxy resins by
pt

reflexion are still under investigation. They might be a promising, fast and non-destructive
method to determine the degree of cure of industrial components.
ce

Acknowledgements
The anhydride-cured system was provided by the Virtual Institute “Nanotechnology in
Ac

Polymer Composites”. H.S, M.J. and G.Z. wish to thank the German Science foundation
(DFG) for supporting parts of this work within the project “Acting principles of nano-scaled
matrix additives for composite structures (FOR2021)”. Thanks to the DFG, the DSC
measurements could be carried out within the project “Development of a wireless
multifunctional sensor system for process control in fiber reinforced composite materials”. In
addition we want to thank Hexion Inc. for providing the amide-cured system.

15
Page 17 of 39
References
[1] H.Q. Pham, M.J. Marks, Epoxy Resins, in: Ullmann's Encyclopedia of Industrial
Chemistry, Wiley-VCH Verlag GmbH & Co. KGaA, 2000.
[2] H.-J. Arpe, Ullmann's encyclopedia of industrial chemistry, VCH Verlagsgesellschaft,
Weinheim, 1985.
[3] J.-P. Eloundou, J.-F. Gerard, D. Harran, J.P. Pascault, Temperature Dependence of the
Behavior of a Reactive Epoxy-Amine System by Means of Dynamic Rheology. 2. High-Tg
Epoxy-Amine System, Macromolecules, 29 (1996) 6917-6927.
[4] G. Lachenal, A. Pierre, N. Poisson, FT-NIR spectroscopy: Trends and application to the

t
kinetic study of epoxy/triamine system (comparison with DSC and SEC results), Micron, 27

ip
(1996) 329-334.
[5] K. Horie, H. Hiura, M. Sawada, I. Mita, H. Kambe, Calorimetric investigation of
polymerization reactions. III. Curing reaction of epoxides with amines, Journal of Polymer

cr
Science Part A-1: Polymer Chemistry, 8 (1970) 1357-1372.
[6] J. Rocks, L. Rintoul, F. Vohwinkel, G. George, The kinetics and mechanism of cure of an

us
amino-glycidyl epoxy resin by a co-anhydride as studied by FT-Raman spectroscopy,
Polymer, 45 (2004) 6799-6811.
[7] A.N. Mauri, N. Galego, C.C. Riccardi, R.J.J. Williams, Kinetic Model for Gelation in the
Diepoxide-Cyclic Anhydride Copolymerization Initiated by Tertiary Amines,
Macromolecules, 30 (1997) 1616-1620. an
[8] X. Fernàndez-Francos, X. Ramis, À. Serra, From curing kinetics to network structure: A
novel approach to the modeling of the network buildup of epoxy-anhydride thermosets,
Journal of Polymer Science Part A: Polymer Chemistry, 52 (2014) 61-75.
M
[9] R. Seifi, M. Hojjati, Heat of Reaction, Cure Kinetics, and Viscosity of Araldite LY-556
Resin, Journal of Composite Materials, 39 (2005) 1027-1039.
[10] M. Munz, H. Sturm, W. Stark, Mechanical gradient interphase by interdiffusion and
ed

antiplasticisation effect – study of an epoxy/thermoplastic system, Polymer, 46 (2005) 9097-

9112.
[11] S. Sourour, M.R. Kamal, Differential scanning calorimetry of epoxy cure: isothermal
cure kinetics, Thermochim. Acta, 14 (1976) 41-59.
pt

[12] C. Zhao, G. Zhang, L. Zhao, Effect of curing agent and temperature on the rheological
behavior of epoxy resin systems, Molecules, 17 (2012) 8587-8594.
[13] M. Mravljak, M. Sernek, The Influence of Curing Temperature on Rheological
ce

Properties of Epoxy Adhesives, Drvna Industrija, 62 (2011) 19-25.

[14] J. Doring, W. Stark, J. Bartusch, J. McHugh, W. Simon, Ultrasound process control
yields mechanical parameters of thermosetting plastics, Materialprufung, 49 (2007) 238-242.
Ac

[15] J. McHugh, W. Stark, J. Doring, Evaluation of the cure behaviour of epoxy resin using
rheometric and ultrasonic techniques, Springer-Verlag Berlin, Berlin, 2003.
1 D. o u C.N. Ca caval F. Mustat ǎ C. Ciobanu Cure kinetics of epoxy resins studied
by non-isothermal DSC data, Thermochim. Acta, 383 (2002) 119-127.
[17] R.A. Fava, Differential scanning calorimetry of epoxy resins, Polymer, 9 (1968) 137-
151.
[18] R.B. Prime, Differential scanning calorimetry of the epoxy cure reaction, Polym. Eng.
Sci., 13 (1973) 365-371.
[19] S. Vyazovkin, N. Sbirrazzuoli, Mechanism and Kinetics of Epoxy-Amine Cure Studied
by Differential Scanning Calorimetry, Macromolecules, 29 (1996) 1867-1873.
[20] S. Montserrat, Vitrification and further structural relaxation in the isothermal curing of an
epoxy resin, J. Appl. Polym. Sci., 44 (1992) 545-554.
[21] H. Liu, X. Wang, D. Wu, Preparation, isothermal kinetics, and performance of a novel
epoxy thermosetting system based on phosphazene-cyclomatrix network for halogen-free
flame retardancy and high thermal stability, Thermochim. Acta, 607 (2015) 60-73.
16
Page 18 of 39
[22] N. Poisson, G. Lachenal, H. Sautereau, Near- and mid-infrared spectroscopy studies of
an epoxy reactive system, Vib. Spectrosc., 12 (1996) 237-247.
[23] C.E. Miller, Near-Infrared Spectroscopy of Synthetic Polymers, Appl. Spectrosc. Rev.,
26 (1991) 277-339.
[24] S.D. Pandita, L. Wang, R.S. Mahendran, V.R. Machavaram, M.S. Irfan, D. Harris, G.F.
Fernando, Simultaneous DSC-FTIR spectroscopy: Comparison of cross-linking kinetics of an
epoxy/amine resin system, Thermochim. Acta, 543 (2012) 9-17.
[25] C. Dell'Olio, Q. Yuan, R.J. Varley, Epoxy/Poly(ethylene-co-methacrylic acid) Blends as
Thermally Activated Healing Agents in an Epoxy/Amine Network, Macromol. Mater. Eng.,
300 (2015) 70-79.

t
[26] G.A. George, G.A. Cash, L. Rintoul, Cure monitoring of aerospace epoxy resins and

ip
prepregs by Fourier transform infrared emission spectroscopy, Polym. Int., 41 (1996) 169-
182.

cr
[27] U. Braun, K. Brademann-Jock, W. Stark, Cure monitoring of epoxy films by heatable in
situ FTIR analysis: correlation to composite parts, J. Appl. Polym. Sci., 131 (2014).
[28] M.C. Celina, N.H. Giron, M.R. Rojo, An overview of high temperature micro-ATR IR

us
spectroscopy to monitor polymer reactions, Polymer, 53 (2012) 4461-4471.
[29] H. Wang, Y. Zhang, L. Zhu, Z. Du, B. Zhang, Y. Zhang, Curing behaviors and kinetics
of epoxy resins with a series of biphenyl curing agents having different methylene units,
Thermochim. Acta, 521 (2011) 18-25.
an
[30] L. Merad, M. Cochez, S. Margueron, F. Jauchem, M. Ferriol, B. Benyoucef, P. Bourson,
In-situ monitoring of the curing of epoxy resins by Raman spectroscopy, Polym. Test., 28
(2009) 42-45.
31 V.K. Hana Vašková aman spectroscopy of epoxy resin crosslinking in: Proceedings
M
of the 13th WSEAS international conference on automatic control, modelling & simulation,
World Scientific and Engineering Academy and Society (WSEAS), Canary Islands, Spain,
2011, pp. 357-361.
ed

[32] V. Janarthanan, G. Thyagarajan, Temperature and radiation effects on the Raman bands
of epoxy resin, J. Chem. Sci. (Bangalore, India), 102 (1990) 721-723.
[33] S.T. Cholake, M.R. Mada, R.K.S. Raman, Y. Bai, X.L. Zhao, S. Rizkalla, S.
Bandyopadhyay, Quantitative Analysis of Curing Mechanisms of Epoxy Resin by Mid- and
pt

Near-Fourier Transform Infra Red Spectroscopy, Defence Science Journal, 64 (2014) 314-
321.
[34] H. Dannenberg, Determination of functional groups in epoxy resins by near-infrared
ce

spectroscopy, Polym. Eng. Sci., 3 (1963) 78-88.

[35] S. Choi, A.P. Janisse, C. Liu, E.P. Douglas, Effect of water addition on the cure kinetics
of an epoxy-amine thermoset, Journal of Polymer Science Part A: Polymer Chemistry, 49
Ac

(2011) 4650-4659.
[36] L. Xu, J.H. Fu, J.R. Schlup, In Situ Near-Infrared Spectroscopic Investigation of Epoxy
Resin-Aromatic Amine Cure Mechanisms, J. Am. Chem. Soc., 116 (1994) 2821-2826.
[37] L. Xu, J.H. Fu, J.R. Schlup, In Situ Near-Infrared Spectroscopic Investigation of the
Kinetics and Mechanisms of Reactions between Phenyl Glycidyl Ether (PGE) and
Multifunctional Aromatic Amines, Ind. Eng. Chem. Res., 35 (1996) 963-972.
[38] H.L. Friedman, Kinetics of thermal degradation of char-forming plastics from
thermogravimetry. Application to a phenolic plastic, Journal of Polymer Science Part C:
Polymer Symposia, 6 (1964) 183-195.
[39] T. Ozawa, A New Method of Analyzing Thermogravimetric Data, Bull. Chem. Soc. Jpn.,
38 (1965) 1881-1886.
[40] J.H. Flynn, L.A. Wall, General treatment of the thermogravimetry of polymers, J Res Nat
Bur Stand, 70 (1966) 487-523.

17
Page 19 of 39
[41] J. Opfermann, W. Hädrich, Prediction of the thermal response of hazardous materials
during storage using an improved technique, Thermochim. Acta, 263 (1995) 29-50.
[42] J. Opfermann, J. Blumm, W.D. Emmerich, Simulation of the sintering behavior of a
ceramic green body using advanced thermokinetic analysis, Thermochim. Acta, 318 (1998)
213-220.
[43] M. Herrera, G. Matuschek, A. Kettrup, Main products and kinetics of the thermal
degradation of polyamides, Chemosphere, 42 (2001) 601-607.
44 L. Matějka J. Lövy S. Pokorný K. Bouchal K. Dušek Curing epoxy resins with
anhydrides. Model reactions and reaction mechanism, Journal of Polymer Science: Polymer
Chemistry Edition, 21 (1983) 2873-2885.

t
[45] L. Shechter, J. Wynstra, R.P. Kurkjy, Glycidyl Ether Reactions with Amines, Industrial

ip
& Engineering Chemistry, 48 (1956) 94-97.
[46] I.T. Smith, The mechanism of the crosslinking of epoxide resins by amines, Polymer, 2

cr
(1961) 95-108.
[47] B.A. Rozenberg, Kinetics, thermodynamics and mechanism of reactions of epoxy
oligomers with amines in: K. Dušek (Ed.) Epoxy esins and Composites II Springer Berlin

us
Heidelberg, 1986, pp. 113-165.
[48] J. Mijovic, A. Fishbain, J. Wijaya, Mechanistic modeling of epoxy-amine kinetics. 1.
Model compound study, Macromolecules, 25 (1992) 979-985.
[49] H. Yamasaki, S. Morita, Identification of the epoxy curing mechanism under isothermal
an
conditions by thermal analysis and infrared spectroscopy, J. Mol. Struct.
[50] G.W. King, R.M. Hainer, H.O. McMahon, Infra‐Red Absorption Spectra of Some
Polymers at Liquid‐Helium Temperatures, J. Appl. Phys., 20 (1949) 559-563.
M
[51] W.H. Avery, C.F. Ellis, Infra‐Red Spectra of Hydrocarbons I. Some Investigations of the
Temperature Dependence of Absorption Bands, The Journal of Chemical Physics, 10 (1942)
10-18.
[52] H. Hagemann, R.G. Snyder, A.J. Peacock, L. Mandelkern, Quantitative infrared methods
ed

for the measurement of crystallinity and its temperature dependence: polyethylene,

Macromolecules, 22 (1989) 3600-3606.
[53] E. Mounif, V. Bellenger, A. Tcharkhtchi, Time-Temperature-Transformation (TTT)
diagram of the isothermal crosslinking of an epoxy/amine system: Curing kinetics and
pt

chemorheology, J. Appl. Polym. Sci., 108 (2008) 2908-2916.

[54] N. Rabearison, C. Jochum, J.C. Grandidier, A cure kinetics, diffusion controlled and
ce

temperature dependent identification of the Araldite LY556 epoxy, J. Mater. Sci., 46 (2011)
787-796.
Ac

18
Page 20 of 39
Tables and Figures

Table 1. Summary of the ingredients of the anhydride-cured and amine-cured epoxy systems

Table 2. Summary of the identified absorption bands with references for the amine-cured and
anhydride-cured systems

Table 3. Summary of the kinetic parameters for the DSC and NIR measurements for the

t
amine-cured and anhydride-cured epoxy systems determined with the kinetics software

ip
cr
Figure 1. Computer-controlled heatable NIR Cell

us
Figure 2. Curing mechanism of epoxy-anhydride formulation with tertiary amines

Figure 3. Auto-catalyst reaction of curing for epoxies with amines according to Smith [46]

an
Figure 4. Overlay of selected, representative NIR spectra (4450 – 7500 cm-1) as a function of
temperature for the anhydride-cured (top) and amine-cured (below) systems (with a
temperature rise from 30 °C to 250 °C with 5 K min-1)
M
Figure 5. Normalized areas of the characteristic absorptions bands as a function of time,
measured from 30 °C to 250 °C with 5 K min-1 for the anhydride-cured (left) and the amine-
ed

cured (right) systems

Figure 6. Change in the integrated areas as a function of sample temperature for pure DGEBA
pt

Figure 7. Normalized epoxy conversion using NIR of different heating rates as a function of
ce

sample temperature for the anhydride-cured system (left) and the amine-cured system (right)

Figure 8. DSC curves of dynamic measurements of the anhydride-cured (left) and amine-
cured (right) systems at four different heating rates (baseline-corrected)
Ac

Figure 9. Arrhenius evaluation (log β over 1000 K/T) of the NI measurements to determine
the activation energies and pre-exponential factors according to ASTM E698 for the
anhydride-cured (left) and amine-cured systems (right) without a chemical reaction model

Figure 10. Arrhenius evaluation (log β over 1000 K/T ) of the DSC measurements to
determine the activation energies and pre-exponential factors according to ASTM E698 for
the anhydride-cured (left) and amine-cured systems (right)

19
Page 21 of 39
Figure 11. Simulated epoxy conversion curves (straight lines) in comparison to the selected
data values measured (symbols) for the anhydride-cured (left) and the amine-cured (right)
systems

Figure 12. Simulated DSC curves (straight lines) in comparison to the selected data values
measured (symbols) for the anhydride-cured (left) and the amine-cured (right) systems

Figure 13. Measured and simulated epoxy conversion (based on NIR and DSC models) as a

t
function of time for isothermal (left) and multi-step curing scenario (right) for the anhydride-

ip
cured system

cr
Figure 14. Measured and simulated epoxy conversion (based on NIR and DSC models) as a
function of time for isothermal (left) and multi-step scenario (right) for the amine-cured

us
system

an
M
ed
pt
ce
Ac

20
Page 22 of 39
Table 1

Table 1. Summary of the ingredients of the anhydride-cured and amine-cured epoxy systems

Anhydride-cured system (Araldite® LY556 HY917 DY 070)

Structure Name Parts

bisphenol A diglycidyl

t
Resin 100
ether (DGEBA)

ip
cr
methyltetrahydrophthalic
Hardener 100
anhydride (MTHPA)

us
Accelerator 1-methylimidazole 2
an
Amine-cured system (EpikoteTM MGSTM RIMR 135/RIMH 137)
M
ed

bisphenol A diglycidyl
> 50
ether (DGEBA)
Resin
pt

1,6-hexanediol diglycidyl
< 25
ether
ce

25 –
isophorondiamine
50
Ac

Hardener

alkyletheramine > 50

Page 23 of 39
Table 2

Table 2. Summary of the identified absorption bands with references for the amine-cured and anhydride-cured
systems
Absorption region Band Origin of absorption band Ref.
/ cm-1
Combination of C-H stretching with CH2 deformation of [22, 34,
4552 – 4497 Epoxy 1
oxirane group 37]
Aromatic
4651 – 4593
1 Combination bands of the aromatic conjugated C=C
[22, 37]
Aromatic stretching with C-H fundamental stretching
4722 – 4651
2

t
5090 – 4863 NH 1 Primary amine and NH2 combination bands [37, 49]

ip
First overtone of terminal fundamental C-H stretching of
6149 – 6026 Epoxy 2 [22, 37]
oxirane group
Overtones of primary and secondary amines N-H

cr
6638 – 6371 NH 2 [37, 49]
stretching
[34, 37,
7158 - 6638 OH First OH overtones
49]

us
an
M
ed
pt
ce
Ac

Page 24 of 39
Table 3

Table 3. Summary of the kinetic parameters for the DSC and NIR measurements for the amine-cured and
anhydride-cured epoxy systems determined with the kinetics software

DSC measurements NIR measurements

Anhydride- Amine-cured Anhydride- Amine-

cured system system cured system
cured system

Without a reaction model (ASTM E698)

t
EA / kJ mol-1 77.7 ± 2.6 57.1 ± 2.8 75.1 ± 14.6 56.4 ± 4.7

ip
log k0 7.71 5.11 7.26 4.90

cr
R 0.9968 0.9929 0.9241 0.9781

With an auto-catalyzed reaction model of the nth order

us
EA / kJ mol-1 77.6 62.3 78.8 56.0

Log k0 7.33 5.72 7.27 4.46

n 1.44 2.70
an 1.61 2.49

log kcat 0.80 0.68 1.09 1.20

M
R 0.9989 0.9972 0.9998 0.9997
ed
pt
ce
Ac

Page 25 of 39
Figure 1

i
cr
us
an
M
ed
pt
ce
Ac

Page 26 of 39
Figure 2

i
cr
us
an
M
ed
pt
ce
Ac

Page 27 of 39
Figure 3

i
cr
us
an
M
ed
pt
ce
Ac

Page 28 of 39
Figure 4

t
ip
cr
us
an
M
d
te
p
ce
Ac

Page 29 of 39
Figure 5

i
cr
us
an
M
ed
pt
ce
Ac

Page 30 of 39
Figure 6

i
cr
us
an
M
ed
pt
ce
Ac

Page 31 of 39
Figure 7

i
cr
us
an
M
ed
pt
ce
Ac

Page 32 of 39
Figure 8

i
cr
us
an
M
ed
pt
ce
Ac

Page 33 of 39
Figure 9

i
cr
us
an
M
ed
pt
ce
Ac

Page 34 of 39
Figure 10

i
cr
us
an
M
ed
pt
ce
Ac

Page 35 of 39
Figure 11

i
cr
us
an
M
ed
pt
ce
Ac

Page 36 of 39
Figure 12

i
cr
us
an
M
ed
pt
ce
Ac

Page 37 of 39
Figure 13

i
cr
us
an
M
ed
pt
ce
Ac

Page 38 of 39
Figure 14

i
cr
us
an
M
ed
pt
ce
Ac

Page 39 of 39