THERMOSETS
1. Introduction
IUPAC defines a thermosetting polymer as a “prepolymer in a soft solid or viscous
state that changes irreversibly into an infusible, insoluble polymer network by
curing” with two notes: (A) Curing can be induced by the action of heat or suitable
radiation, or both and (B) a cured thermosetting polymer is called a thermoset (1).
Unlike thermoplastic polymers, chemical reactions are involved when producing
or fabricating parts. As a result of these reactions, the materials cross-link and become “set,” that is they can no longer flow or be dissolved. Cure most often is thermally activated, hence the term thermoset, but network-forming materials whose
cure is light or radiation activated are also considered to be thermosets. Some
thermosetting adhesives cross-link by a dual cure mechanism, that is, by either
heat or light activation. In contrast to cross-linked elastomers or rubbers, the glass
transition temperature (Tg ) of thermosets is generally above room temperature.
In this article, we describe the distinguishing characteristics of thermosetting materials; the more common thermosetting resins including their cure
chemistry; thermoset applications; materials and process characterization; cure,
including cure kinetics and cure monitoring; and conclude with a description of
selected thermoset processes.
Uncured thermosets are mixtures of small reactive molecules, often
monomers. They may contain additives such as catalysts to promote or accelerate
cure. Most thermosets are used in filled or reinforced form to reduce cost, modify
physical properties, act as a binder for particles or fibers, reduce shrinkage during
cure, or to provide or enhance fire resistance. In general, thermosets possess good
dimensional stability, thermal stability, chemical resistance, and electrical properties. Because of these attributes, they find widespread use in several applications
such as adhesives; primary and secondary structural members in aerospace; countertops and floors for manufacturing facilities and homes; printed circuit boards,
conductive polymer elements, and encapsulation materials for electronic applications; biomedical and dental materials, especially adhesives; and recreational
products such as tennis racquets, bicycle frames, golf clubs, and fishing rods.
Epoxy resins are probably the best known of the thermoset family whose members
also include Phenolic resins, unsaturated polyester, polyurethanes, dicyanate,
bismaleimide, acrylate, and many others (see POLYESTERS, UNSATURATED;
ISOCYANATE-DERIVED POLYMERS). Good thermoset references include May (2),
Prime (3), Pascault and co-workers (4,6), Van Mele and co-workers (5), and Guo (7).
1
Encyclopedia of Polymer Science and Technology. Copyright c⃝ 2017 John Wiley & Sons, Inc. All rights reserved.
DOI: 10.1002/0471440264.pst519.pub2
2
THERMOSETS
Fig. 1. Schematic, two-dimensional representation of thermoset cure. For simplicity difunctional and trifunctional co-reactants are depicted. Reprinted with permission from
Ref. 3. Copyright 1981 Elsevier.
2. Cure
Thermosets are distinct from thermoplastic polymers in one major respect: Their
processing includes the chemical reactions of cure. Cure is the process that transforms small reactive molecules into a cross-linked three-dimensional network
with good structural and physical properties. It includes chemical reactions, the
development of specific properties, and physical transformations and transitions
such as gelation and vitrification. Because it is the most studied and best understood cure reaction, we use the epoxy–amine system as a model for general
thermoset behavior.
Cure starts with A-stage monomers (or oligomers) (a); proceeds via simultaneous linear growth and branching to a B-stage material below the gel point (b);
continues with formation of a gelled but incompletely cross-linked network (c);
and ends with the fully cured, C-stage thermoset (d). This is illustrated schematically in Figure 1 for difunctional and trifunctional coreactive monomers. Reaction
in the early stages of cure (a to b in Fig. 1) produces larger and branched molecules
THERMOSETS
3
Fig. 2. Macroscopic development of rheological properties (eg, 𝜂 0 = zero-shear viscosity)
and mechanical properties (eg, Ge = equilibrium shear modulus) during network formation,
illustrating the approach to infinite viscosity and the first appearance of an equilibrium
modulus at the gel point. Reprinted with permission from Ref. 10. Copyright 1998 Springer.
and reduces the total number of molecules. Macroscopically, the thermoset can be
characterized by an increase in its viscosity 𝜂 (see Fig. 2). As the reaction proceeds
(b to c in Fig. 1), the increase in molecular weight accelerates and all the chains become linked together at the gel point into a network of infinite molecular weight.
The gel point coincides with the first appearance of an equilibrium modulus Ge , as
shown in Figure 2. At this point, the thermoset can support a load without flowing. Unless impeded by vitrification reaction continues beyond the gel point (c to d
in Fig. 1) to complete the network formation. Macroscopically, physical properties
such as modulus and glass transition temperature build to levels characteristic of
a fully developed network.
Thermosets are amorphous and exhibit a glass transition, the reversible
change upon heating from a hard glassy state to a viscous or rubbery one. As
the cure reactions proceed, the material will change from a viscous liquid to a
rubber and to a glass if the cure temperature is below its maximally attainable
Tg . Figure 1 schematically depicts these changes from a liquid to a cross-linked
solid. On cooling from above Tg , the thermoset at all stages of cure will transition from viscous or rubbery to hard and glassy. Subsequent sections will cover
the changes in Tg with cross-linking. Tg is the approximate midpoint of the of the
temperature range over which the glass transition takes place; for more details
on the glass transition, see Seyler (8).
2.1. Gelation. Gelation is the incipient formation of a cross-linked network, and it is the most distinguishing characteristic of a thermoset. As illustrated
in Figure 1, cure begins by the growth and branching of chains. As the reaction proceeds, the increase in molecular weight accelerates, and eventually several chains
become linked together into a network of infinite molecular weight. The abrupt
and irreversible transformation from a viscous liquid to an elastic gel or rubber
is called the gel point. The gel point of a chemically cross-linking system can be
defined as the instant at which the weight-average molecular weight diverges to
infinity (9).
4
THERMOSETS
Fig. 3. Idealized TTT cure diagram. A plot of the times to gelation and vitrification during
isothermal cure versus temperature delineates the regions of four distinct states of matter: liquid, gelled rubber, gelled glass, and ungelled glass. Reprinted with permission from
Ref. 17. Copyright 1983 John Wiley & Sons, Inc.
A thermoset loses its ability to flow and is no longer able to be processed
above the gel point, and therefore gelation defines the upper limit of the work
life. An example is Five-Minute Epoxy, which can be found in any hardware store
and where gelation occurs in about 5 minutes at room temperature. After the
two parts are mixed, the user must form an adhesive joint within 5 minutes
while the adhesive can still flow before it becomes rubbery, and then keep the
repaired part fixtured until cure is sufficiently complete, typically for one to several hours. Full cure may take up to 24 hours. A distinction may be drawn between the phenomenon of molecular gelation and its consequence, macroscopic
gelation. Molecular gelation occurs at a well-defined stage of the chemical reaction, provided the reaction mechanism is independent of temperature and free
of non-cross-linking side reactions (11–13). It is dependent on the functionality,
reactivity, and stoichiometry of the reactants. Macroscopic consequences of gelation include a rapid approach toward infinite viscosity and development of elastic
properties not present in the pregel resin. Molecular gelation may be detected as
the point at which the reacting resin just becomes insoluble, or as the point where
the mechanical loss tangent becomes frequency independent (9–12,14,15). Macroscopic means to approximate gelation include the time to reach a specific viscosity and the G′ = G′′ crossover in a dynamic rheology measurement. Frequencyindependent damping peaks accompanied by a small increase in storage modulus,
commonly observed in thermosetting resins on glass fiber or wire mesh substrates,
have been attributed to gelation (16,17; see Fig. 35 as an example). Such measurements are used to construct time–temperature-transformation (TTT) diagrams
such as Figure 3 (16,17). Beyond the gel point, the reaction proceeds toward the
formation of one infinite network with substantial increases in cross-link density,
glass transition temperature, and ultimate physical properties.
THERMOSETS
5
Gelation in step-growth systems such as epoxy–amine typically occurs between 40% and 80% conversion (degree of cure 𝛼 = 0.4–0.8). The gel point may
be calculated if the chemistry is known (11,12,14). Gelation in ionic chain-growth
systems such as cationic polymerization of epoxies typically occurs in the 30%–
40% conversion range. For high functionality and radical chain-growth systems,
gelation may occur at much lower conversions. The degree of conversion at the gel
point 𝛼 gel is constant for a given thermoset, independent of cure temperature, provided the cure mechanism does not change with temperature. Gelation is considered to be isoconversional, and for this reason the time to gel versus temperature
can be used to measure the activation energy for cure (see eq. 8, Section 4.2). There
is no discontinuous change in the rate of the reaction on going through gelation
and gelation does not inhibit cure. As a result, gelation cannot be detected by techniques sensitive only to the chemical reaction rate, such as differential scanning
calorimetry (DSC). Gelation may be measured rheologically (see Fig. 31).
2.2. Vitrification. Vitrification is glass formation due to Tg increasing
from below Tcure to above Tcure as a result of the cure reaction and is defined by
Gillham as the point where Tg = Tcure (16). This is a practical rather than precise
definition since Tg depends on the method of measurement. It is a completely distinct phenomenon from gelation and may or may not occur during cure depending
on the cure temperature relative to Tg∞ , the glass transition temperature for the
fully cured network. Vitrification may also occur at slow heating rates, usually
below 2◦ C/min, when the reaction rate becomes faster than the heating rate. Vitrification can occur anywhere during the reaction to form either an ungelled glass
or a gelled glass. It can be avoided by curing at or above Tg∞ . In the glassy state,
the reaction rate will usually undergo a significant decrease as the cross-linking
reaction becomes controlled by the diffusion of reactants. It is common for complete vitrification to result in a decrease in the reaction rate by two to three orders
of magnitude.
Unlike gelation, vitrification is reversible by heating, and chemical control of
cure may be re-established by heating to devitrify the partially cured thermoset.
Vitrification may be detected and the time/temperature to vitrify measured by
a step decrease in heat capacity by modulated-temperature DSC (MTDSC; see
Figs. 27 and 28, and temperature-modulated differential scanning calorimetry
[TMDSC]), as a frequency-dependent transition in the loss modulus and tan 𝛿
peaks by dynamic mechanical analysis (DMA), and as a frequency-dependent
transition in the dielectric loss factor during cure by dielectric analysis (see
Fig. 36).
Generally, the shift from chemical control to diffusion control of the reaction
may be observed by a slowing of the reaction rate, which can be readily observed
for thermosets that cure by step-growth and ionic chain-growth polymerization
when Tg reaches about 10–20◦ C above Tcure (see Figs. 11 and 15). Though not
common in these systems, it is possible for diffusion to control the cure kinetics
prior to vitrification for very fast reactions; it is also possible for sluggish reactions to remain under chemical control well into the glassy state (see Refs. 3–5).
Radical chain-growth polymerizations are diffusion controlled from the start of
the reaction but also exhibit a similar slowing in their reaction rate. In this case,
the slowing is due to additional diffusion limitations for most steps as well as a
significant decrease in the efficiency of the initiation step (18).
6
THERMOSETS
Table 1. Glossary of Characteristic Cure Parameters
𝛼
𝛼 gel
tgel
tvit
Tcure
Tg
Tg0
gel Tg
Tg∞
Chemical conversion (eg, of epoxide or isocyanate groups), degree of cure
𝛼 at the gel point
Time to gelation, gel time
Time to vitrification
Cure temperature, a process parameter
Glass transition temperature, a material property
Tg for uncured thermoset with degree of conversion 𝛼 = 0
Tg for thermoset with degree of conversion 𝛼 gel
Tg for fully cured thermoset with degree of conversion 𝛼 = 1
In most cases, it is advantageous to avoid vitrification to achieve complete
cure in a reasonable time, which necessitates cure at temperatures close to or
greater than Tg∞ . But there can be advantages to curing in the glassy state, including the ability to complete the cure process without expensive fixturing while
still maintaining dimensional stability; see Section 6.2 and (121).
Isothermal TTT cure diagrams, as illustrated in Figure 3, are a useful tool
for illustrating the phenomenological changes that take place during cure, such
as gelation, vitrification, complete cure, and degradation (16,17). Three critical
temperatures are marked on the temperature axis of the TTT cure diagram: Tg0 ,
the glass transition temperature of the completely unreacted thermoset; gel Tg , the
temperature at which gelation and vitrification coincide; and Tg∞ , the glass transition temperature of the fully cured network (see Table 1). In this generalized
diagram, the times to gelation and the times to vitrification are plotted as functions of the cure temperature. At temperatures below Tg0 , the uncured thermoset
must react in the glassy state and any reaction is therefore slow to occur. It is recommended that unreacted systems be stored well below Tg0 ; 20–50◦ C below Tg0 ,
that is deep into the glassy state to minimize reaction during storage (see Section 7.9 on Adhesive Processes for comments about storing premixed and frozen
adhesives). Between Tg0 and gel Tg , the liquid resin will react without gelation until its continuously rising glass transition temperature becomes coincidental with
the cure temperature, at which stage vitrification begins and the reaction in most
cases becomes diffusion controlled. Remember that gel Tg is the temperature at
which gelation and vitrification occur simultaneously.
At temperatures between gel Tg and Tg∞ , the viscous liquid changes to a
viscoelastic fluid, then to a rubber, and finally to a glass. Gelation precedes vitrification, and a cross-linked rubbery network forms and grows until its glass
transition temperature coincides with the cure temperature, where the reaction
likely becomes diffusion controlled. At temperatures above Tg∞ , the network remains in the rubbery state after gelation unless other reactions occur, such as
thermal degradation or oxidative cross-linking. Note that in the manufacture
of carbon–carbon composites, network degradation is part of the process (see
Phenolic Resins).
The handling and processing of thermosets are very much dependent on gelation and vitrification. For example, thermosets are often identified at three stages
of cure, A, B, and C:
THERMOSETS
•
•
•
7
A-stage refers to an unreacted resin or resin mixture;
B-stage refers to a partially reacted and usually vitrified system, below the
gel point, which, upon heating will devitrify, flow and be processed, and subsequently cure toward completion;
C-stage refers to the completely cured network.
Thus, to B-stage a thermoset requires vitrification prior to gelation, which
can be accomplished by maintaining the reaction temperature below gel Tg . Bstaging often provides systems that are optimized for handling and processing.
In general, thermosets that need to be solid during the precure stage of processing, for example, powder coatings, will be B-staged to Tg ≥ Tprocess , whereas thermosets that need to have some flow or tack during precure processing, for example,
prepregs, will be B-staged to Tg ≈ Tprocess . Prepregs are sheets of oriented fibers
or fabric that are impregnated with resin and B-staged; typically several layers
are laminated together under heat and pressure to fabricate a part.
2.3. Degree of Cure, Conversion. Thermoset properties develop during the cure process. Because so many other properties are dependent on degree
of cure or conversion, this is arguably the most important property of a thermoset.
Since conversion and the glass transition temperature Tg are related (see Section
2.4), this structure–property relationship is unique to thermosets and is a key
aspect of both processing and characterizing these materials.
The most common analytical tools for measuring degree of cure are DSC from
the cure exotherm; thermogravimetric analysis (TGA) where a volatile component
such as water or formaldehyde is a product of the cure reaction; and Fourier transform infrared (FT-IR) and Raman spectroscopies from the concentration of functional reactive groups. Time-temperature superposition relationships are considered in Section 4.3 on cure kinetics.
DSC measures heat flow into a material (endothermic) or out of a material
(exothermic). Thermoset cure reactions are exothermic, that is heat is produced
as a result of the chemical reactions. Figure 4 illustrates isothermal cure of an
epoxy–amine in the DSC at three temperatures. The maximum in the isothermal
DSC at t > 0, with tmax decreasing with increasing temperature, is characteristic of
autocatalytic cure. The maximum is due to competing effects of the rate accelerating from the increasing concentration of the alcohol produced in the epoxy–amine
reaction and the rate decreasing due to consumption of the epoxide and amine reactants (Schemes 5 and 6 in Section 3.3). Note that the ordinate of the DSC curve
is the heat flow dH/dt, which is proportional to the reaction rate d𝛼/dt = (dH/dt)
÷ ΔHrxn , the heat of reaction (see eq. 1). The reaction rate is sometimes expressed
as dx/dt.
Figure 5 illustrates an alternative DSC method where individual samples
are partially cured at various time/temperature combinations and then scanned in
the DSC. ΔHrxn is the total heat liberated from an uncured thermosetting system
and can be measured in a DSC heating experiment. ΔH is expressed as heat per
mole of reacting groups (kJ/mol or kcal/mol) or per mass of material (J/g or cal/g).
The basic assumption underlying the application of DSC to thermoset cure is that
the measured heat flow dH/dt [the ordinate of a DSC measurement, measured in
8
THERMOSETS
6
DGEBPA – EDA (B = 1)
5
70o
Reaction
rate
4
3
60o
2
50o
1
0
0
20
40
60
80
100
Time (min)
Fig. 4. Isothermal DSC of epoxy resin at three temperatures. Reprinted with permission
from Ref. 19. Copyright 1970 John Wiley & Sons, Inc.
Fig. 5. A series of DSC temperature scans at 10◦ C/min of an epoxy–amine cured isothermally at 160◦ C for different times, showing Tg increasing and the residual exotherm decreasing with increasing cure times. Reprinted with permission from Ref. 20. Copyright
1990 John Wiley & Sons, Inc.
THERMOSETS
9
J/s (W) or cal/s] is proportional to the rate of the chemical cure reaction, d𝛼/dt, as
shown in equation (1).
d𝛼∕dt =
(dH∕dt)
ΔHrxn
(1)
In cases such as the epoxy–amine reaction, it is further assumed that the heat of
reaction for epoxy with primary amine is the same as that for the reaction of epoxy
with secondary amine, and that the activation energies are the same. For systems
with a symmetrical DSC exotherm, these parameters are usually within a few
percent of each other (3–5). As pointed out in Section 3.3 on epoxy resins, there is
usually a substitution effect where the secondary amine is somewhat less reactive
than the primary amine, but the effect on the overall kinetics measured by DSC
is usually not significant. In the small number of systems where steric hindrance
is significant, the substitution effect is much larger, the DSC exotherm exhibits a
high temperature shoulder or even two distinct exotherms, and an overall reaction
cannot be assumed. In such cases equations 1 and 2 may not apply. This is one
reason why a cure study should always begin with a survey DSC, for example, at
10◦ C/min.
In terms of a DSC measurement, the conversion or degree of cure at time t,
𝛼 t ,DSC , is defined as
𝛼t,DSC =
)
(
ΔHrxn − ΔHres
ΔHrxn
(2)
where ΔHres is the residual heat of reaction at time t, also illustrated in
Figure 5. The cure of some thermosets, for example, those with higher functionality reactants, may be topologically restricted and unable to reach complete conversion. In those cases ΔHrxn is replaced by ΔHult , in which case 𝛼 t,DSC becomes 𝛼 ult ,
the maximum achievable conversion (see pp. 1390–1410 in Ref. 3 for a detailed
discussion). Sometimes the symbol x is used instead of 𝛼 to denote conversion or
degree of cure.
Figure 6 shows conversion versus ln(time) curves for the same epoxy for cure
temperatures from 100 to 180◦ C. Conversion was calculated from equation 2. Note
that the curves are parallel during the first part of cure and, at the lower cure
temperatures, less than 100% conversion is achieved.
TGA measures mass flow out of a sample, that is, mass or weight loss. Mass
loss in thermoset systems includes the loss of solvents, volatile products associated with cure, and degradation products associated with thermal and thermal–
oxidative degradation. For several thermosets, such as phenolic and amino resins,
an integral part of cure is the formation of condensation products, such as water or
formaldehyde (see Scheme 3). For blocked isocyante systems, unblocking involves
volatile loss and the unblocking temperature can be measured by DSC (21) and
TGA (22).
When the loss of mass is an integral part of the cure chemistry, TGA can
be applied to thermoset cure. The most common example is phenolics, where
each bond formation is accompanied by the loss of a water and/or formaldehyde
10
THERMOSETS
Fig. 6. Conversion versus time data for the same epoxy–amine system shown in
Figure 5: ⊡ 100◦ C, ◦ 120◦ C, ▴ 140◦ c, ⬧ 150◦ C, ■ 160◦ C, □ 180◦ C. Reprinted with permission
from Ref. 20. Copyright 1990 John Wiley & Sons, Inc.
molecule. Assuming a one-to-one relationship between mass loss and conversion
leads to the TGA equation for conversion
𝛼t,TGA =
mi − mt
Δmt
=
mi − mf
Δmrxn
(3)
where 𝛼 t,TGA is the TGA conversion, mi is the initial mass at the beginning of the
mass loss process, mf is the final mass at the end of the mass loss process, and mt
is the instantaneous mass at time t. Note that equation 3 is analogous to the DSC
conversion given in equation 2.
Figure 7 shows the TGA mass % and derivative DTG in percent/min curves
for a thermoplastic-thermoset adhesive as a function of temperature. The first
heat shows ∼10% mass loss attributed to the cure of the phenolic component. The
dashed line and the DTG signal show the thermal decomposition of the fully cured
adhesive.
Figure 8 shows the mass % versus time curves for the curing of the phenolic portion at several isothermal temperatures. For the adhesive in this system,
equation 3 reduces to 𝛼 t,TGA = Δmt /10. As an example for mt = 96%, we calculate
𝛼 t,TGA = Δmt /10 = 4/10, that is the degree of cure is 40%. Note that this degree of
cure occurs at all temperatures but at different times in Figure 8.
In this manner, conversion versus time curves at all temperatures may be
constructed, fit to different rate equations and analyzed for rate constants (see
Section 4 on cure kinetics). Note that Tg∞ for this phenolic is 235◦ C, which means
that vitrification will occur at all temperatures, slow the reaction and complicate
the kinetic analyses. The effects of vitrification are especially evident at lower
temperatures, and close to full cure is only achieved at the highest cure
Mass (%)
THERMOSETS
100
4
80
0
60
–4
40
–8
20
–12
11
–16
0
300
500
Temperature (ºC)
100
700
Fig. 7. TGA of thermoplastic-thermoset adhesive. Reprinted with permission from
Ref. 23. Copyight 2000 Springer.
100
98
Mass (%)
100°C
96
125
150
175
94
200 (2 samples)
92
225
90
0
100
200
Time (min)
300
400
Fig. 8. Isothermal TGA of thermoplastic-thermoset adhesive. Reprinted with permission
from Ref. 23. Copyright 2000 Springer.
temperatures. Vitrification will occur at higher conversion with increasing cure
temperature.
FT-IR and Raman spectroscopies are able to analyze organic and inorganic
materials, including thermosets. Results are obtained by spectral interpretation,
for example, band allocation to functional groups and comparison with reference spectra, or imaging such as two-dimensional mapping of functional groups
(see VIBRATIONAL SPECTROSCOPY). They are capable of characterizing the cure
12
THERMOSETS
reaction from the liquid state through gelation to the completion of conversion,
and even into the glassy state with good precision, sensitivity, and temporal resolution. Determination of chemical conversion versus time and temperature can
be made from measurements of functional group intensity with progress of the reaction. For the epoxy–amine reactions shown in Schemes 5 and 6 (in Section 3.3),
one can measure consumption of epoxide and primary amine, production, and subsequent consumption of secondary amine, and production of tertiary amine and
hydroxyl, thus allowing simultaneous examination of several kinetic processes associated with cure. FT-IR and Raman spectroscopies have advantages over other
techniques such as DSC for their ability to distinguish multiple simultaneous reactions which may have different heats of reaction and activation energies.
Normal vibrations related to a change in dipole moment are infrared active. Functional reactive groups in thermosetting systems, such as epoxide, amine,
and hydroxyl, typically have large dipole moments and strong infrared absorptions. Near infrared (FT-NIR) spectroscopy (12,000–4,000 cm−1 ) has several advantages over the traditional mid-IR spectroscopy (4000–400 cm−1 ), including
less complex absorption spectra and the transparency of glass in this region,
which allows it to be used as a sample support and for the possibility of in
situ, real-time cure monitoring, for example, by fiber optic NIR spectroscopy
(24). Epoxide, primary amine, and secondary amine concentrations can be measured directly. When these concentrations are known, the hydroxyl and tertiary
amine concentrations and the extent of etherification can be deduced from epoxy–
amine chemistry and kinetic considerations (25–27). The ratio of rate constants
R (= k2 /k1 ) for the epoxy–cheme (Scheme 6) to the epoxy–primary amine reaction (Scheme 5) can be measured. R = 0.5 for equal reactivity, which is often assumed, for example, for kinetic analyses from DSC measurements. PazAbuin and co-workers (28) report a simple method and R values around 0.4
for three systems but <0.2 for an epoxy–cycloaliphatic amine system. St. John
and George (29) reported R = 0.22 for a TGDDM/DDS system (tetraglycidyl4,4′ -diaminodiphenylmethane/diaminodiphenylsulfone). Kinetic calculation allows one to determine other functional group concentrations not easily measured.
The cure of epoxy model systems has been characterized by FT-NIR (28–35) and
Raman (36) spectroscopies. Quantitative measurements are in good agreement
with those obtained from other techniques, such as DSC and size-exclusion chromatography (also known as gel-permeation chromatography) (28). The epoxy–
amine reaction has been studied by Raman spectroscopy (30,36,37).
NIR conversion is typically measured from the concentration of epoxide obtained from the epoxide band at 4528 cm−1 , a combination of the stretching fundamental at 3050 cm−1 with the CH2 deformation fundamental at 1460 cm−1 , as
defined in equation 4,
𝛼t,NIR =
A0 − At
A0 − A∞
(4)
where A0 represents the normalized absorbance initially, At is the absorbance at
time t, and A∞ is the absorbance at complete cure. Figure 9 shows the normalized
epoxy conversion of an anhydride-cured system at four different heating rates,
measured by a heatable NIR cell with an integrated thermocouple that enables
THERMOSETS
1.0
13
Anhydride-cured system
0.8
1 K min-1
3 K min-1
5 K min-1
10 K min-1
0.6
0.4
0.2
0.0
50
75
100
125
150
175
200
225
Fig. 9. Normalized epoxy conversion using NIR at different heating rates as a function
of sample temperature for an epoxy–anhydride system. Reprinted with permission from
Ref. 35. Copyright 2015 Elsevier.
actual sample temperature to be controlled and measured in situ during cure (35).
Kinetic analyses were performed by isoconversional methods. Note that the shape
of the conversion curve in Figure 9 from infrared measurements is very similar
to the conversion–time data presented in Figure 6 and essentially identical to the
conversion–time data in Figure 13, both from DSC data. The cure profiles are for
two types of epoxies (epoxy–amine in Figs. 6 and 13 and epoxy-anhydride in Fig. 9),
but clearly demonstrate the characteristic conversion profile during thermoset
curing. Also note the similar deviations in the 1◦ C/min curves at high conversion
in Figures 9 and 13, in both cases attributed to vitrification (see Section 4.2)
2.4. Glass Transition Temperature–Conversion Relationship. Another very useful tool is the Tg –conversion relationship. DSC provides the most
common measure of conversion from the exotherm associated with the cure reaction as well as Tg . There are many instances where the exotherm is obscured
by other thermal events, such as the loss of water or formaldehyde during cure
of phenolics, or the exotherm becomes too small to measure accurately near the
final stages of cure (eg, see longer time curves in Fig. 5). The well-established
relationship between Tg and conversion validates Tg as an equal measure of
degree of cure, extending cure characterization to many more thermosetting
systems.
As can be observed in Figure 5, there appears to be a relationship between
the increasing glass transition temperature and decreasing residual exotherm.
Several workers have shown that for most thermoset systems there is a unique
relationship between the chemical conversion of a thermoset and its glass transition temperature, independent of the cure temperature and thermal history (see,
eg, Refs. 38–40). While a few exceptions have been noted, this appears to be the
14
THERMOSETS
Fig. 10. Relationship for Tg and fractional conversion for the same epoxy–amine system
shown in Figure 6. Different symbols represent material cured at different temperatures:
■ 100.5◦ C, • 117.2◦ C, ▴ 139.1◦ C, ◦ 157.1◦ C, □ 180.7◦ C, ▵ 200.4◦ C. The solid curve represents the best-fit calculation from the modified DiBenedetto equation (41,42). Reprinted
with permission from Ref. 20. Copyright 1990 John Wiley & Sons, Inc.
general case (see Ref. 3). Thus, measurement of Tg may be assumed equivalent to
direct measurement of conversion.
Figure 10 illustrates the Tg –conversion relationship for the same epoxy–
amine system shown in Figures 5 and 6 cured at several temperatures. All the
data can be seen to collapse to a single curve and are fitted to the empirical
DiBenedetto equation (41,42). Excellent theoretical treatises of the Tg –conversion
relationship can be found in Refs. 38–40 and 43. As can be seen from the increasing curvature of the data in Figure 10, Tg is a more sensitive measure of cure than
is conversion, as measured by ΔHres , in the latter stages of cure, which are often
the most critical. From the slope of the curve for the final 10% of cure for this
system [5.4◦ C/(% conversion)] and an ability to measure Tg to ±2◦ C (8), control of
degree of cure to 0.5% from measurement of Tg is readily feasible. For thermosetting systems in which weight loss is associated with cure or for which there is
no simple or unambiguous measure of conversion, Tg is often the only practical
means to monitor extent of cure.
The similarity of the Tg –time data in Figure 11 with the conversion–time
data of Figure 6 is a consequence of the Tg –conversion relationship and illustrates
the ability to monitor cure through measurement of Tg . Figure 11 also directly illustrates vitrification, defined as the point at which Tg becomes equal to Tcure (16)
and designated at each cure temperature by an arrow. Note that the progress of
cure is significantly impeded shortly after vitrification, which marks the shift from
chemical control to diffusion control of the reaction. From a practical perspective,
the data show that to get full cure in a reasonable time, cure temperatures in
THERMOSETS
15
Fig. 11. Tg versus ln(time) for data at different cure temperatures: ▴ 100◦ C, ⬧ 120◦ C, ×
140◦ C, ◊ 150◦ C, ■ 160◦ C, □ 180◦ C. Isothermal vitrification (Tg =Tcure ) at each cure temperature is designated by an arrow. Same epoxy–amine system as in Figures 6–8. Reprinted
with permission from Ref. 20. Copyright 1990 John Wiley & Sons, Inc.
the range of 180◦ C would give rapid reaction rates and avoid vitrification (Tg∞ =
178◦ C).
3. Thermosetting Resin Systems
In the uncured state, thermosetting materials are generally mixtures of small
reactive molecules that form networks; base resins, catalysts, initiators, and/or
accelerators; along with fillers that can be particles, fiber-based (either woven
or nonwoven chopped fibers), or nanosize fillers. This section covers the various
types of thermosetting base resins. The most widely used thermoset base resins
are unsaturated polyesters, phenolics such as phenol–formaldehydes, and epoxy
resins. There are also specialty thermoset base resins utilized in advanced engineering applications such as electronics, composites, and aerospace. Bio-based
monomers from starting materials such as vegetable oils and agricultural byproducts show promise to match properties from petroleum-based thermosets and
improve biodegradability (44,45).
3.1. Unsaturated Polyesters. Unsaturated polyester resins consist of a
linear unsaturated polyester polymer, a cross-linking monomer, activators or accelerators, and inhibitors to retard cross-linking until the resin is to be used (see
POLYESTERS, UNSATURATED). Cross-linked polyesters are mainly used in glass
fiber reinforced form for large structural parts. The linear unsaturated polyester is
typically the condensation product of an unsaturated dibasic acid, such as maleic
anhydride, a glycol, such as ethylene glycol, and a saturated dibasic acid, such as
phthalic anhydride, to modify the degree of unsaturation as shown in scheme 1.
16
THERMOSETS
O
C
COOH + HO
HOOC
Diol
or
O
C
R OH + O
C
C
O
Maleic anhydride
O
O
Saturated anhydride
O
H O C
O
O
O
C O R O C CH CH C O R OH
n
Scheme 1. R and R′ may be aliphatic or aromatic species
The most common cross-linking monomer is styrene; methyl methacrylate is
added for resistance to sunlight and weathering.
Cross-linking starts with a free-radical initiator, usually an organic peroxide. Free radicals may be formed by thermal decomposition of an initiator such
as tert-butyl perbenzoate (TBP) or benzoyl peroxide, by ultraviolet (UV) or visible
light decomposition, or by chemical decomposition in ambient temperature applications utilizing activators such as dimethylaniline. Typical inhibitors such as
hydroquinone, p-benzoquinone, and phenothiazine are added in very small quantities (<500 ppm) and are used to increase the shelf life. During curing, the inhibitors initially consume the newly formed free radicals and reaction does not
proceed until the inhibitor system is completely depleted. After an induction period, the free radicals initiate exothermic cross-linking reactions involving the
unsaturation in the polyester chains and the styrene monomer with gelation occurring at low levels of conversion. Reaction proceeds to form a highly cross-linked,
three-dimensional network. In Scheme 2, the typical cure pathway is shown where
a linear unsaturated polyester is cured using styrene with a peroxide initiator and
suitable catalyst such as cobalt octoate/naphthenate:
3.2. Phenolic Resins. Phenol–formaldehyde resins are employed in a
wide range of applications, including commodity construction materials such as
plywood and oriented strand board to high technology applications such as honeycomb and carbon–carbon composites (see PHENOLIC RESINS). They are composed
of a wide variety of structures based on the reaction products of phenols with
formaldehyde. Either resole or novolac resins are formed, depending on the mole
ratio of formaldehyde to phenol (F/P) and catalyst.
Resole resins are produced from an excess of formaldehyde (F/P > 1) and
contain reactive methylol groups that can condense to a network structure. The
cross-linking reactions in these one-part systems are exothermic and produce water and formaldehyde as volatile products, as illustrated in Scheme 3.
Novolac resins are produced from an excess of phenol (F/P < 1) to form
linear polymers, as illustrated in Structure 1, with molecular weights of 200–
5000 and glass transition temperatures of 45–70◦ C. In two-stage novolac systems,
THERMOSETS
17
O
R
O C CH CH PE
CH CH C
Unsaturated polyester
+
O
Styrene
Catalyst
CH
O
R
CH2
O C CH CH PE
CH CH C
CH2
O
CH
Crosslinked network
Scheme 2. PE = unsaturated polyester resin from Scheme 1
Scheme 3.
cross-linking is accomplished by reaction with hexamethylene tetramine (HMTA),
which is prepared from formaldehyde and ammonia and may be considered a latent source of formaldehyde.
Structure 1.
The typical cure pathway for a novolac cured with HMTA is shown in
Scheme 4.
3.3. Epoxy Resins. Probably the best-known thermoset is epoxy (2) (see
EPOXY RESINS). The largest use of epoxies is in protective coatings, with other
applications including printed circuit board (PCB) laminates, electronic materials, structural composites, flooring, and adhesives. The higher cost of epoxy resins
vis-à-vis commodity thermosets and thermoplastics is justified by their superior
properties and longer service life. Cured epoxies provide excellent mechanical
strength and toughness; outstanding chemical, moisture, and corrosion resistance;
good thermal, adhesive, and electrical properties; absence of volatiles and low
18
THERMOSETS
OH
p,p
o,o
o,p
OH
HO
OH
Novolac
HMTA
OH
p,p
OH
o,o
p,p
o,p
HO
o,p
o,o
OH
o,p
OH
OH
OH
OH
Cross-linked network
Scheme 4. o = ortho, p = para.
shrinkage on cure; and good dimensional stability. Epoxy resins are characterized by a structure containing the epoxide or oxirane group (see Structure 2). The
most widely used epoxy resins are the diglycidyl ethers of bisphenol A.
Structure 2.
Others include brominated bisphenol A resins, which impart fire resistance,
epoxy phenol novolac resins, bisphenol F epoxy resins, epoxy cresol novolac resins,
cycloaliphatic epoxy resins, and TGDDM. Liquid crystalline epoxy resins combine
the properties of epoxy resins and liquid crystals with diverse applications such
as microelectronics packaging materials, optical waveguides, adhesives, and color
filters (46).
The number of repeat units, n, in Structure 2 can be as high as 35 but is
usually 0–6, where 0 denotes the monomer. Epoxy resins can be cured by a variety of cross-linking agents known as hardeners (reactive comonomers) or by
catalysts that promote homopolymerization. Common examples of homopolymerization include the anionic polymerization promoted by Lewis bases, such as
tertiary amines and imidizoles, and cationic polymerization promoted by Lewis
acid BF3 complexes, such as boron trifluoride monoethylamine (BF3 ⋅NH2 C2 H5 ).
Photoinitiated cationic cure of epoxies, especially cycloaliphatic epoxies, with
photoinitators such as aryldiazonium and diaryliodonium salts provide coatings
and adhesives with long shelf life and near-instantaneous cure. But it is more
THERMOSETS
19
Scheme 5.
Scheme 6.
common to cross-link epoxies with a coreactant such as a diamine, polyamide,
or acid anhydride. The low exotherm of epoxy–anhydride systems make them
suitable for uses in large mass epoxy cures. Amines account for close to 50% of
all epoxy curing agents. For cross-linking to occur, at least one of the reactants
must be trifunctional or higher. Epoxy resins are typically difunctional, reacting
through the oxirane group, although in some cases, such as homopolymerization
or when epoxy is in stoichiometric excess, reaction can occur through the –OH
groups.
Diamines, a well-studied coreactant with epoxy, are tetrafunctional where
each amine hydrogen can react. The reactions of primary amine and secondary
amine with epoxide are illustrated in Schemes 5 and 6, respectively. The heat of
reaction ΔHrxn for epoxy–amine is ∼106 kJ/mol (∼25.5 kcal/mol) for reaction with
both primary amine and secondary amine (3–5). The activation energy for cure
can vary from 40 to 125 kJ/mol (10–30 kcal/mol) (3–5)
For equal reactivity of primary amine and secondary amine with epoxy, k2 /k1
= 0.5. But there is usually a substitution effect where the secondary amine is less
reactive and k2 /k1 < 0.5, often in the range of 0.40–0.45. When steric hindrance is
significant, k2 /k1 can be < 0.2.
In general, aliphatic amines are used with bisphenol A resins where roomtemperature cures are desired and heat-deflection temperatures (HDT) or glass
transition temperatures (Tg ) below 100◦ C can be tolerated. Aromatic amines can
provide HDTs and Tg s up to 170◦ C but require curing at elevated temperatures.
Anhydride-cured bisphenol A systems offer long pot life, low exotherm, excellent
adhesion and electrical properties, and HDT/Tg values between 125 and 170◦ C.
A small amount of tertiary amine is frequently used to accelerate the cure reaction, which proceeds by addition esterification and addition etherification. Cycloaliphatic epoxides respond well to acidic hardeners and can have HDT/Tg values approaching 200◦ C. TGDDM is a tetrafunctional epoxide which is often cured
with diaminodiphenylsulfone (DDS) to form high temperature, high performance
epoxy systems for aerospace applications. Commercial TGDDM/DDS systems,
which are typically epoxy rich, represent materials in which secondary amine hydrogens are significantly less reactive than primary amine hydrogens, and the
etherification reaction of epoxide with hydroxyl is significant.
3.4. Polyurethanes. Polyurethanes are produced by the reaction of a
polyfunctional isocyanate with a polyol or other reactant containing two or more
20
THERMOSETS
Scheme 7.
groups reactive with isocyanate, usually hydroxyls (see ISOCYANATE-DERIVED
POLYMERS, POLYURETHANES). Thermoset polyurethanes require that one reactant
be trifunctional or higher and that the Tg of the cross-linked network be above
room temperature. Products include rigid foams, coatings, adhesives, and binders.
The reaction of an isocyanate with an alcohol, illustrated in Scheme 7, is exothermic and forms a urethane or carbamate linkage without the production of volatile
products. Blocked isocyanates provide room temperature stable, one-component
systems, such as wire and coil coatings that can be cured at high temperatures.
Typically, temperatures above 120◦ C are required to release the blocking group
and regenerate the isocyanate group. Polyurethane powder coatings, which are Bstaged solid systems applied to a substrate as a dry powder, often employ blocked
isocyanates. Coalescence and cure at high temperatures result in tough, durable
coatings.
3.5. Acrylates. Acrylate-containing monomers can be used to rapidly
form high cross-link density networks using free-radical initiators, radiation initiators (UV, UV–visible, and electron beam), and Michael addition initiators. Very
fast cure speed can be obtained over a wide range of formulations, making this
particularly useful for high speed curing of coatings and adhesives.
Typical acrylic monomers contain either the acrylate functionality Structure
3) or the methacrylate functionality (Structure 4), where R can be aromatic or
aliphatic, and also contain linear, branched, or cyclic alkanes, ethers, or esters
(see ACRYLIC ESTER POLYMERS).
Structure 3.
Structure 4.
Free-radical curing coatings and adhesives can be formulated using
temperature-sensitive free-radical initiators such as peroxide and azonitrile compounds. Typical peroxide initiators are benzoyl peroxide and cumyl hydroperoxides. The most common azonitrile compound is 2,2′ -azobisisobutyronitrile. With
these initiators, free radicals are generated by thermal decomposition leading to
an active free radical. Once a free radical is formed, it propagates to form a polymer by the following reaction pathway, where Scheme 5 illustrates initiation and
Scheme 6 propagation.
THERMOSETS
21
Scheme 8.
Scheme 9.
Selection of the functionality and stoichiometry of the monomers or
oligomers can yield cross-linked networks with various physical and mechanical
properties. For example, monomers are commercially available with mono-, di-,
tri-, and multifuntionality. Choosing various ratios of di-, tri-, and multifunctional
acrylate-containing monomers, the curing speed, Tg , hardness, solvent resistance,
and flexibility can be tailored for a specific application. Choice of the peroxide
is critical since there are a wide range of initiation temperatures. When using
free-radical initiators with low cure onset temperatures, one needs to be careful
with the shelf life of coatings or work life times for paste adhesives. Additionally,
the free-radical curing mechanism is sensitive to oxygen inhibition. Oxygen is an
effective free-radical scavenger and to achieve uniform curing in the thickness
of the coating, oxygen must be eliminated at the surface by using an inert
atmosphere (typically nitrogen) or by the addition of chemical additives that are
good oxygen scavengers.
Another common curing mechanism for acrylate-containing polymers is to
cure using UV light. Typical UV-cured coatings contain a free-radical-generating
photoinitiator. Radicals are generated upon exposure to UV light and quickly
initiate free-radical polymerization. Common photoinitators include benzophenone and the diaryl ketone derivatives including zanthone, thiozanthone, and 2chlorothioxanthone, along with Michler’s ketone, benzyl, and anthraquinone (47).
Under UV irradiation, solventless coatings and adhesives can be cured extremely
fast at high line speeds.
To expand the formulation toolbox, acrylates can be used as coreactants and
reactive diluents in epoxy systems cured using amines. Acrylate esters react with
amine curing agents through a Michael addition reaction, resulting in a secondary
amine–acrylate adduct that can further react with epoxy resins. In this manner,
highly cross-linked networks with controlled properties can be formulated.
Fast-curing coatings and adhesives using acrylate-containing monomers are
used in high performance automotive primers and topcoats, concrete bridge deck
coatings, industrial floorings, and specialty coatings for wood and paper. In the microelectronics industry, many photoresists use acrylate-containing monomers and
oligomers. High performance die attach adhesives use a wide variety of acrylatecontaining monomers, allowing solvent-free formulations, very fast curing, and
control of key physical properties such as Tg and modulus.
UV curing is used in high performance coating applications, such as topcoats for flooring applications (polymer and wood flooring), coatings for metal and
wood furniture, along with plastic and metal container coatings. UV curing is used
extensively in the processing of photoresists in the microelectronics industry.
22
THERMOSETS
Photodefinable solder masks used to protect the outside surfaces of circuit boards
and chip substrates make considerable use of acrylate-containing polymers.
3.6. Bismaleimide Resins. Polyimide resins are a class of high performance resins that have gained wide acceptance in a variety of applications (see
POLYIMIDES). Bismaleimides are thermosetting polyimides that cure by addition
reactions that avoid formation of volatiles. Unlike condensation polyimides that
evolve volatiles during curing and are typically used as thin films, bismaleimide
resins contain unsaturated double bonds that can be cured via several chemical
pathways (free-radical, Michael addition, or cycloaddition, such as Diels–Alder).
Thermal polymerization without the evolution of volatile by-products enables bismaleimide resins to be used in a large number of applications from electronic
materials (such as die attach adhesives), coatings, films, structural adhesives, encapsulants, and high performance fiber-reinforced composites.
Bismaleimide resins generic structure is shown in Structure 5.
Structure 5.
Depending on the desired properties, the R group can be either aliphatic or aromatic. For high Tg networks, aromatic bismaleimides are used. The most common
component in bismaleimide resins is 4,4′ bismaleimidodiphenylmethane (BMI)
(Structure 6):
Structure 6.
The BMI resin is a high melting point crystalline solid and when thermally cured, the resulting networks are very brittle. Bismaleimides are rarely
used alone in the solid form. In most applications, reactive comonomers (eg, vinyl
and allyl compounds, allyl phenols, aromatic amines) are used to achieve the desired final properties and processing characteristics. Commercially available resin
systems are typically two-component systems. For example, the Matrimid 5292
system from Huntsman Chemical (Matrimid 5292 originally developed by Ciba
Geigy) is a blend of BMI and O,O′ -diallyl bisphenol A (DBA) (Structure 7).
Structure 7.
THERMOSETS
23
DBA has low viscosity and is used in place of solvents to enhance the processability and increase the toughness of BMI composites. DBA copolymerizes with
BMI causing a decrease in the cross-link density, leading to a tougher network.
BMI resins for composites are cured at temperatures in the range of 200–250◦ C.
Liquid BMIs with long-chain aliphatic structures between the maleimide
groups are used for low modulus, low Tg , high flexibility applications. For electronic applications, BMIs have made a significant impact in die attach adhesives.
Liquid BMIs can be used to formulate high performance die attach adhesives
because of the ability to achieve very fast curing rates via free-radical polymerization with no void formation. Typical formulations are cured with peroxides and
can have reaction speeds allowing cure in less than 1 min (snap cure) at curing
temperatures in the range of 175–220◦ C. For these systems, a significant amount
of the network formation is completed within seconds allowing semiconductor
chips to be rapidly attached to leadframes or substrates and maintain a very high
production rate at the die bonder. Additionally, the die attach formulations can be
tailored to have low Tg , low modulus (to control stress and warpage during processing), low moisture absorption, and high adhesion at elevated temperatures. A
potential drawback of the peroxide curing is short work life and the need to store
the formulated adhesive at low temperatures to minimize reaction during storage. Typical work life (working time after thawing) is between 12 and 36 h and
premixed and frozen adhesives are typically stored at −40◦ C.
3.7. Dicyanate Resins. The term dicyanate resin, alternatively known
as cyanate ester resin, describes both a family of monomers and oligomers with
reactive cyanate end groups on an aromatic ring and the resulting cured resin networks (see review by Kessler (48)). Bisphenol A dicyanates were commercialized in
the late 1970s. The initial use was for high circuit density and high speed circuit
boards. During initial testing, exposure to high relative humidity caused moisture stability problems. Moisture resistance was improved by blending bisphenol A dicyanates with epoxy resins to reduce the amount of ester linkages in
the resulting copolymer. During the 1980s and 1990s, many more new dicyanate
ester monomers were developed (49), including a bicyclopentadiene containing
dicyanate (Dow 71787.02) (50). The bicyclopentadiene-modified cyanate exhibits
excellent electrical and mechanical properties and enhanced processability.
Bisphenol A dicyanates had high Tg s but were brittle and had low fracture toughness. Several approaches were developed to toughen cyanate esters including
the use of core/shell rubbers (51) and thermoplastic toughening agents such as
polyethersulfone, polysulfone, and polyetherimides (52).
The predominant curing pathway for cyanate ester is via cyclotrimerization
using metal coordination catalysts. Three types of metal coordination catalysts are
used: metal naphthenates, metal acetylacetonates (AcAc), and metal octoates. Of
these, Cu AcAc and Mn octoate are the most popular. Nonylphenol is also required
as a cocatalyst. Scheme 10 shows the cyclotrimerization reaction of bisphenol A
dicyanate to form the triazine network structure:
For very high Tg applications, dicyanates are cured without addition of epoxy
modifiers leading to a very tight network via cyclotrimerization. The advantage
of the cyanate ester family of monomers is the ability to formulate with epoxy
resins. Typical resins in electronic applications are combinations of tailored epoxy
monomers and oligomers along with cyanate esters.
24
THERMOSETS
Scheme 10.
Mitsubishi Gas Chemical pioneered the development of circuit substrates
by blending bisphenol A dicyanates with epoxies and a small amount of bismaleimide. The resulting bismaleimide/triazine/epoxy (BT epoxy) laminates are
the industry standard for low dielectric, high Tg chip packaging substrates.
Cyanate ester networks have very high Tg s, high modulus, and low dielectric
constant and have an attractive dielectric loss factor. These properties made the
cyanate ester family of resins attractive for circuit packaging applications. Liquid cyanate ester monomers (bisphenol F dicyanates) have low viscosity at 25◦ C
and exhibit good compatibility with other cyanate ester resins, epoxy resins, and
bismaleimide resins (52). Blending liquid cyanate ester resins with other cyanate
ester and epoxy resins provided a pathway to control the rheological properties
without sacrificing the final cured performance. Liquid cyanate esters enabled
the development of underfill resins for chip packaging applications, where low
viscosity is required during processing. After curing, cyanate ester resins provide
high Tg , low coefficient of thermal expansion, high toughness, and low moisture
absorption required for underfill applications.
Dicyanate resins are versatile thermosets and have been used in a wide
variety of applications. For example, RTX-366 (Hunstman Chemical), shown in
Structure 8, is a high molecular weight cyanate ester oligomer with excellent dielectric properties (low dielectric constant and loss factor) along with low moisture absorption (53). RTX 366 has been used in low outgassing and dimensionally
stable, high tolerance composites for communication satellites, space antennas,
microwave transparent radomes, and aircraft film adhesives.
N
C
O
CH3
CH3
C
C
CH3
CH3
O
C
N
Structure 8.
3.8. Allyl Resins. Allyl resins are diallyl monomers and B-staged prepolymers that can be cross-linked by free-radical polymerization through their
double bonds. The allyls in widest use are diallyl phthalate (DAP) resins and diallyl isophthalate resins. The prepolymers can be used as the sole resin, whereas the
monomers are commonly used as cross-linking agents for unsaturated polyesters;
use of triallyl isocyanurate as a cross-linking monomer results in excellent high
THERMOSETS
25
temperature properties. These resins are exceptional in their ability to maintain
electrical properties under high temperature and high humidity conditions. They
are also characterized by excellent dimensional stability, chemical resistance, and
a chemical purity that minimizes ionic contamination. Specialty allyls have been
traditionally used in eyewear applications.
4. Cure Kinetics
This topic is important in two regards: characterization of the thermosetting material itself and the design and optimization of cure processes. Cure kinetics generally fall into two categories. The first is aging or storage kinetics after reactants have been mixed, but prior to their use. Reaction of prepregs, premixed
and frozen adhesives, and powder coatings during storage fall in this category.
The objective in this case is to minimize reaction. The second and generally the
more important is the cure of the thermoset after the prepreg has been formed
into a structure, the resin has infused into a preform in a vacuum-assisted resintransfer molding (VARTM) process, or the adhesive joint has been formed. Included here are the characterization, design, and optimization of cure processes,
where issues such as the optimum time to apply pressure in an autoclave process or the time/temperature to reach full cure are considered. Development of a
kinetic equation is often needed to model the thermal cure process. When combined with rheological models, the flow behavior or chemorheology of the reacting
system can be projected (54). For some applications, for example, to estimate the
decrease in cure time resulting from a 10◦ C increase in temperature, all that is
needed is the activation energy for cure to make the calculation (eg, by solving
eq. 10 for t2 , where t1 , T1 , T2 , and E are known).
4.1. Kinetic Equations and Analysis. Cure kinetics is the mathematical relationship between time, temperature, and conversion. Two aspects of cure
kinetics are considered: traditional kinetic analyses and time–temperature superposition or TTS kinetics. Traditional kinetics involves models such as nth order and autocatalytic, fits conversion–time data taken at several temperatures
to these models to determine reaction orders and measure rate constants, and
measures the activation energy E from the Arrhenius equation. An assumption of
TTS kinetics is that cure can be described by a single activation energy, independent of conversion and temperature, which is a reasonable assumption for most
thermosetting systems. TTS kinetics creates master cure curves by superimposing conversion–time, and Tg –time curves through the Tg –conversion relationship,
utilizing the Arrhenius equation. It may be applied to the characterization of cure
as well as to the monitoring and control of systems which cure according to complex time–temperature profiles.
Isothermal conversion–time data can be measured continuously from the
beginning to the end of the reaction as in Figure 4, or individually by scanning
a series of partially cured samples as in Figure 5. Reliable techniques to obtain
conversion–time data are isothermal DSC, FT-IR, and Raman spectroscopies, and
isothermal TGA for thermosets such as phenolics that evolve a water or formaldehyde molecule for each bond formed. Continuous isothermal DSC measurements
such as Figure 4 have the advantage of simultaneously measuring conversion (𝛼)
26
THERMOSETS
and rate of conversion (d𝛼/dt), which allows use of derivative forms of the rate
equation such as equations 5 and 6. A disadvantage of this method is the danger
of not recording the initial reaction that may take place during heating to temperature (3). Individual measurements on partially cured samples can yield very
accurate data and, in the case of DSC, give both Tg and conversion (see Fig. 5 as
an example). However, they are time consuming and since the reaction rate is not
measured, integrated forms of the rate equation, such as equation (11) must be
used.
Equation 5 shows the general nth-order rate equation
d𝛼
= k(1 − 𝛼 )n
dt
(5)
where k is the rate constant, n the reaction order, and (1 – 𝛼) is the concentration
of reactant(s). One example is the second-order reaction of an isocyanate with an
alcohol to produce a polyurethane, as illustrated in Scheme 7. Integrated forms of
this equation exist for first-order, second-order, and general nth-order reactions;
equation (11) is the integrated form of the second-order rate equation. Since there
are only two unknowns, simple linear regression analysis may be used to determine values for k and n.
Equation 6 shows a common autocatalytic rate equation proposed by Sourour
and Kamal (55)
)
d𝛼 (
= k1 + k2 𝛼 m (1 − 𝛼 )n
dt
(6)
where k1 is the catalyzed rate constant, attributed to added catalyst, impurities and/or fiber or particle surfaces that have a catalytic effect; k2 is the autocatalyzed rate constant, due to catalytic species produced by the reaction; and
m and n are reaction orders. An example of autocatalysis is the reaction of an
epoxy resin with a diamine to produce an epoxy thermoset, as illustrated in
Schemes 5 and 6. In the epoxy–amine reaction, the alcohol, which is produced
in the reaction, catalyzes further reaction, resulting in autocatalysis. The chemistry for a stoichiometrically balanced reaction suggests that m = 1 and n = 2 in
equation 6. For real systems, values are often close but not identical to these. The
alcohol concentration is represented by 𝛼. Since there are four unknowns, nonlinear regression analysis must be employed, although k1 can be evaluated independently as the extrapolated reaction rate at t = 0 (cf Fig. 4 at t = 0), reducing the
number of unknowns to three. Autocatalytic kinetics is usually evaluated by the
derivative form of the autocatalytic rate equation (eg, eq. 6) with 𝛼 and d𝛼/dt data
collected by continuous isothermal DSC measurements. These equations describe
the cure reaction under chemical control. References 56–58 discuss incorporating the shift from chemical control to diffusion control upon vitrification into an
overall kinetic analysis.
27
dH/dt
THERMOSETS
2 °C/min
5 °C/min
10 °C/min
50
70
90
110
130
150
170
190
210
Temperature (°C)
Fig. 12. DSC of an epoxy at three heating rates. Reprinted with permission from Ref. 61.
Copyright 1977, 2002 Springer.
4.2. Activation Energy. Activation energy E and preexponential factor
A are measured from the Arrhenius equation:
(
k = A exp
−E
RT
)
(7)
where k is the rate constant, R is the gas constant (1.987 cal/K⋅mol ≡ 8.314
J/K⋅mol), and T is absolute temperature. Also note that two activation energies
are measured for autocatalytic cure, E1 and E2 , corresponding to the two rate
constants k1 and k2 .
The measurement of activation energy is often the first step in a kinetic
analysis utilizing time-temperature superposition kinetics. Additionally, activation energy by itself can often be useful, for example, by using equation 10 to
estimate how simple changes in time or temperature will affect the cure process.
The measurement of activation energy is based on the concept of isoconversion. In an isothermal measurement, the time to reach a constant conversion
level is measured as a function of temperature. A common example is gel time.
Equation 8 shows the relationship between time to gel and temperature where
the activation energy is measured from the slope of ln tgel versus T−1 , where T is
the absolute temperature.
(
tgel ∝ Aexp
E
RT
)
(8)
In a multiple heating rate study, the temperature to constant conversion versus heating rate is measured, as illustrated in Figure 12. Lyon presents a comprehensive review of this topic (59). Equation 9 shows the relationship between the
28
THERMOSETS
%
kJmol ^ –1
55
80
10 K/min
50
5 K/min
60
50
100
%
0
2 K/min
40
1 K/min
20
0
0
100
50
150
Fig. 13. Conversion in % versus temperature for epoxy–amine from DSC multiple heating rate measurements. In the insert, apparent activation energy E𝛼 versus conversion.
Reprinted with permission from Ref. 62. Copyright 2002 Elsevier.
temperature to reach constant conversion T𝛼 and the heating rate q where the
activation energy is measured from the slope of ln q versus T𝛼 −1 (K−1 ). A more
accurate value of E may be obtained by recalculating the constant 1.052 from tables in [60] and then recalculating E. Usually this iteration is only required to be
performed once (see ASTM E698-11). The peak temperature Tp is generally considered to be an isoconversional temperature and is sometimes used by itself to
estimate activation energy. Software to perform these analyses is provided by all
DSC manufacturers.
E𝛼 ≅
−R Δ ln q
1.052 ΔT𝛼−1
(9)
The vertical lines in Figure 12 represent constant conversion as measured
from the exotherm areas, for example, 10% for the solid lines and 20% for the
dashed lines. By continuing this process to include the entire cure exotherm, for
example, up to 90 or 95%, E𝛼 can be measured over the entire range of conversion,
as shown in the insert in Figure 13. E𝛼 , calculated from DSC curves between 2
and 20 K/min (≡ ◦ C/min), yielded an average value of 52 kJ/mol (12.4 kcal/mol).
While there was a small decrease in E𝛼 with conversion, the reaction mechanism
was considered to be constant. In such cases where E𝛼 is approximately constant
with conversion, the cure reaction is considered to be relatively uncomplicated and
the kinetic analysis is usually straightforward. On the other hand, a significant
change of E𝛼 with conversion is indicative of complex cure chemistry which must
be taken into account in the kinetic analysis. Often, but not always, complexity
can be observed in a survey run at 10◦ C/min as a shoulder on the DSC exotherm
or even a second peak.
The main portion of Figure 13 shows conversion versus temperature calculated from DSC curves at heating rates similar to those employed in the FT-IR
study shown in Figure 9. In both the DSC and FT-IR studies, there are similar
THERMOSETS
Conversion
(Tg)
T1
>
29
Master curve
T2
aT
In time
In reduced time = time at reference temperature, e.g. T2
Note log time scale, curves parallel
Fig. 14. Schematic of time-temperature superposition to create master curve for cure. R.
B. Prime, unpublished.
deviations in the 1 K/min curves vis-à-vis those measured at faster heating rates.
In both cases, the deviations were attributed to vitrification, where Tg is increasing faster than the sample temperature and the reaction becomes diffusion controlled. For this reason, the 1 K/min data were excluded from the kinetic analyses
in both studies (35,62).
4.3. Time-Temperature Superposition Kinetics. For applications of
TTS kinetics, we consider cure reactions that can be described with reasonable accuracy by a single activation energy, independent of conversion and temperature.
The majority of cure reactions fall in this category, and they typically exhibit a
single, symmetrical DSC exotherm. This includes epoxy–amine cure, where there
may be small differences in the activation energies for the primary amine–epoxy
reaction and the secondary amine–epoxy reaction. They are usually within a few
percent of each other, and an average or overall activation energy gives a good description of the cure (see DSC curves in Fig. 12 as an example). In cases where the
activation energies are significantly different, the DSC exotherm will be skewed
or there will be two distinct exotherms, in which case a more complex analysis of
cure must be undertaken (3). Activation energies for cure are typically between
40 and 125 kJ/mol (10 and 30 kcal/mol). More detailed discussions of kinetics, not
only related to cure but also to storage life before cure and aging after cure, can
be found in Refs. 3–5.
Figure 14 shows a simple schematic plot of two conversion–time, or Tg –
time, curves at temperatures T1 and T2 , where T1 > T2 . When plotted on a ln
time basis, the result is two curves separated by a shift factor aT defined by
equation 10. aT is the ratio of times to reach the same conversion or Tg at two
different temperatures, and it is described by a form of the Arrhenius equation,
where t is the time to constant conversion or constant Tg , E is the activation energy, R is the gas constant, and T is the absolute temperature. When E is constant
with conversion, aT , is constant and the curves are parallel over their entirety allowing the construction of a master curve. The assumption of a single or overall
activation energy means that the only effect of temperature is to speed up or slow
down the reaction.
t
aT = 2 =
t1
[
)]
(
E T1 − T2
RT1 T2
(10)
30
THERMOSETS
Fig. 15. Superposition of the Tg versus ln(time) data to form a master curve at 140◦ . Note
that vitrification points at all cure temperatures lie on the master curve, that is vitrification
occurs during chemical control of the reaction: ⊡ 100◦ C, ⬧ 120◦ C, □ 140◦ C, ▴ 150◦ C, ■
160◦ C, □ 180◦ C. Reprinted with permission from Ref. 20. Copyright 1990 John Wiley &
Sons, Inc.
Master curves are useful for summarizing kinetic data taken at different
time/temperature conditions and for predicting behavior at times and temperatures that differ from those that were measured. Note the parallel nature of the
curves prior to vitrification in Figures 6 and 12. Figure 15 shows the master curve
from shifting the data of Figure 11 along the ln time axis using the measured overall activation energy of 63.5 kJ/mol (15.2 kcal/mol) (20). The data were shifted to
a reference temperature of 140◦ C by shifting each curve by a constant factor [ln
(aT ) = ln(t140◦ C ) − ln(tT )] (see eq. 10) along the ln(time) axis so that its beginning section (Tg < 90◦ C) coincides with the curve for Tcure = 140◦ C. Figure 15
clearly shows the reaction under chemical control (master curve, solid line). Only
the chemically controlled reaction, which follows Arrhenius kinetics, is subject
to time–temperature superposition. Following vitrification, the individual curves
deviate from the master curve as the reaction slows and becomes controlled by
diffusion of reactants in the glassy state.
Note that in many thermoset cure processes, temperatures are high enough
relative to Tg∞ to avoid vitrification. A good example is a fast curing, twocomponent polyurethane used in a pultrusion process (63). Parts are made by
mixing the components in-line and rapidly processing and curing. In this study,
conversion–time data were collected from isothermal DSC measurements. The
master cure curve shown in Figure 16 was obtained by shifting these data to times
at a reference temperature of 80◦ C by means of equation 10, using the activation
energy for cure measured from multiple heating rate DSC measurements (3,64).
To obtain additional data at high conversion but without the complication of vitrification, heat-hold-cool DSC measurements were made that simulated process
cure profiles. Equivalent cure times at the same reference temperature of 80◦ C
THERMOSETS
31
100
Conversion (%)
80
60
45
60
40
80
20
Kinetic Eqn
0
0
50
100
150
Fig. 16. Master cure curve for polyurethane pultrusion resin system. The solid line is
calculated from the second-order kinetic equation found to fit the master curve. Reprinted
with permission from Ref. 63. Copyright 2005 Elsevier.
35
30
25
20
15
10
5
0
0
50
100
150
Fig. 17. Plot of conversion – time data from master curve according to second-order kinetics equation. Reprinted with permission from Ref. 63. Copyright 2005 Elsevier.
were calculated from the profiles by shifting individual points along the profile to
times at 80◦ C by means of equation 10 and summing those points (3,63,64).
As shown in Figure 17, the master curve was found to fit equation 11, the
integrated form of the second-order kinetic equation (eq. 11 with n = 2) and
a complete mathematical description of cure was provided. From this kinetic
equation, the development of conversion along the profile shown in Figure 18 was
determined. A conversion of 96.5% for the polyurethane resin cured according to
this profile was estimated. The kinetic equation also allowed the computation of
the master curve shown as a solid line in Figure 16.
1
= 1 + kt
1−𝛼
(11)
4.4. Effect of Catalysts, Fillers, and the Environment on Cure.
Since catalysts and fillers are generally incorporated into thermoset formulations,
it is important to realize they can have a profound influence on the rate of cure.
THERMOSETS
200
100
160
80
120
60
80
40
20
40
Temperature
0
Conversion (%)
32
0
1
2
Time (min)
Conversion
3
4
0
Fig. 18. Time–temperature profile for pultrusion process and development of conversion
along the process. Temperature (◦ C, ◦), conversion (%, ■). Reprinted with permission from
Ref. 63. Copyright 2005 Elsevier.
Comparative DSCs, for example, with and without the additive, contain a wealth
of information. If the two curves overlap, it may be concluded that the additive
has no effect on cure. If the area under the curve decreases, this may indicate a
lower extent of cure. If the curve shape stays the same but the curve shifts to lower
temperatures, the additive accelerates the reaction. A decrease in activation energy, for example, from multiple heating rate measurements, indicates catalytic
behavior of the additive. A change in shape of the DSC curve indicates a change in
the reaction chemistry. Isothermal DSC can be employed where more detail such
as reaction rates are desired.
Willard (65,66) evaluated the effects of peroxide initiator on the heats
of reaction and rate constants of a single DAP prepolymer. Both parameters were found to be sensitive to catalyst type and level. For commercial
applications, the combination of high heat of reaction and fast curing rate
is desirable. TBP and dicumyl peroxide satisfy these criteria and are in
general use at the 3-phr level. Widmann (67) found that diminishing levels
of a tertiary amine catalyst led to reduced reaction rates and higher activation energies. Miller and Oebser (68) showed that accelerators used in
Dicyandiamide(DICY)-cured epoxy systems could be sorted into two categories:
cocuring agents that shorten cure times but do not affect the activation energies
and catalysts that not only shorten the cure time but substantially lower the
activation energy. DICY, dicyandiamide, is a fine particulate crystalline amine
hardener that is dispersed in the epoxy resin. Because cure requires melting and
dispersion of the particles DICY affords one-part epoxy systems with good shelf
life.
The surfaces of fibers and fillers will often participate in the cure reaction.
Dutta and Ryan (69) studied the effects of a carbon black filler and a silica filler
on the cure kinetics of an amine–epoxy system by isothermal DSC. The carbon
black filler was found to have no effect on the heat of reaction or activation energy and only a small effect on reaction rates. The silica filler, on the other hand,
causes an increase in the heat of reaction and reduction in the activation energy.
THERMOSETS
33
Willard (65,66) showed that three common DAP fillers had little effect to a slight
enhancement on the heat of reaction. In stark contrast, a kaolinite clay virtually
eliminated curing, attributed to decomposition of the initiator by the acidic clay
surface.
Multiple authors have found that the presence of glass or carbon fibers could
affect the reaction rate but not change the reaction mechanism or final extent
of cure (70,71). The cure kinetics and final cross-linked state of an anhydridecured epoxy were shown to be strongly affected by absorption of the tertiary amine
catalyst at oxidation sites on carbon surfaces (72). Oxidized carbon surfaces initially accelerated epoxy–amine reactions, but inhibited later stages of the reaction so that the final state of cure was reduced (72). Oxidized carbon surfaces also
preferentially absorbed the amine hardener, resulting in a stoichiometric imbalance at the interface.
DMA studies of an epoxy–phenolic magnetic coating with and without iron
oxide and cured in air or nitrogen atmospheres showed that Fe2 O3 and O2 had
marked effects on the cure (73–75). A decrease in gel time when Fe2 O3 was incorporated in the coating was attributed to catalysis of the reaction by the acidic
iron oxide surface. The very high cross-link density, which requires the presence of
both iron oxide and air, demonstrated the active role of the oxide in the oxidative
cross-linking process.
Water is a variable component of the atmosphere with potential to significantly influence cure. It is a good idea, especially in a humid environment, to dry
all components prior to thermoset processing. As illustrated by the example above,
oxygen also has the ability to alter the cure, but it is usually not a significant factor. Free radical cure processes, for example, UV cure adhesives, is an exception
and may require an oxygen scavenger, an excess of the free radical initiator, or a
nitrogen environment for an effective cure (see Section 3.5).
Bair (76) found that small amounts of moisture naturally present in epoxy–
Novolac molding compounds could not escape during the transfer molding process
and retarded the cure, irreversibly lowering Tg by as much as 40◦ C. When dried
materials were used for transfer molding, high conversion levels and high Tg s
were obtained.
Brand (77) showed that curing a B-staged DICY-epoxy in a saturated water environment at ∼150◦ C caused an irreversible lowering of the ultimate Tg
by about 10◦ C relative to dry samples cured in ambient air. Samples cured in a
trichloroethylene atmosphere exhibited an initially very low Tg , but removal of
the solvent raised Tg to the level of samples cured in ambient. It was concluded
that absorbed water altered the cure mechanism. Noll (78) postulated two possible reactions of DICY with water, both of which resulted in an increase in the
number of amine hydrogens available to react, thus offering an explanation for
the decrease in Tg by altering the stoichiometry.
Garton and co-workers (79–81) investigated the cross-linking of epoxy resins
at interfaces. Absorbed moisture on the surface of Kevlar fibers had a significant
effect on the cure of an epoxy–anhydride system in the vicinity of the aramid
surface, causing an increase in the rate of anhydride consumption, a decrease
in the yield of ester products and a decrease in Tg (79). It was proposed that
the absorbed water reacts with the anhydride to produce a diacid that catalyzes
etherification.
34
THERMOSETS
80
Silica
Alumina
Diamond
BN
70
60
50
40
30
20
0
10
20
30
40
50
Filler content (vol%)
Fig. 19. Coefficient of linear thermal expansion versus filler content for an epoxy anhydride resin system. Reprinted with permission from Ref. 82 Copyright 2005 Elsevier.
5. Additives and Fillers
In this section, we consider components of a thermosetting system other than the
resins themselves. This includes additives such as catalysts to promote or accelerate cure. Most thermosets are used in filled or reinforced form to reduce cost,
modify physical properties, act as a binder for particles or fibers, reduce shrinkage
during cure, or to provide or enhance fire resistance. In the following sections, we
describe some of the more common uses of fillers in thermosets.
5.1. Particulate Fillers for Thermal Expansion. Particulate fillers are
frequently added to a thermosetting system to match the thermal expansion of
the cured thermoset to the surface to which it is bonded, or to reduce the difference in their thermal expansions to reduce warpage and thermal stresses (see
Section 6.7). This requires a reduction in thermal expansion coefficient 𝛼 (Coefficient of thermal expansion (CTE) in ppm/◦ C). Common particulate fillers used to
reduce thermal expansion include alumina (𝛼 = 8 × 10−6 K−1 ), borosilicate glass
(𝛼 = 3 × 10−6 K−1 ), fused silica (𝛼 = 0.55 × 10−6 K−1 ), boron nitride (BN; 𝛼 = 0.5
× 10−6 K−1 ), and diamond (𝛼 = 0.8 × 10−6 K−1 ). In Figure 19, typical reductions in
the coefficient of linear thermal expansion are shown for various filler types and
filler loadings (82).
5.2. Particulate Fillers for Electrical Conductivity. Electrically conductive adhesives (ECAs) are an important class of materials widely used in electronics. A typical ECA has a thermoset matrix and electrically conductive filler
particles. The thermoset polymer matrix is electrically insulating, and to achieve
a high degree of electrical conductivity metal fillers are added. Silver (Ag) is the
most common filler. Gold, nickel, and copper have also been used but to a much
lesser extent.
To develop electrical conductivity, the silver particles must form a conductive
path from one side of the adhesive joint to the other. If the filler loading is insufficient to develop a connected path through the resin matrix, electrical conductivity
will not develop. In Figure 20, a schematic representation of the filler loading is
THERMOSETS
35
Fig. 20. Schematic of percolation threshold, important for electrical and thermal conductivity in thermosets.
shown. The insulating resin is denoted by the yellow color, and the electrically
conductive (or as we shall see in the next section, thermally conductive fillers) is
denoted in blue.
Below the percolation threshold, particle–particle interactions do not form
a continuous path through the adhesive. At the percolation threshold, a continuous path develops and electrical conductivity dramatically increases. Achieving
a filler loading above the percolation threshold is the first requirement. In addition, the particle–particle contact is governed by the shape of the filler particle. Silver flakes are the most common filler geometry for ECAs owing to their
plate-like morphology. A good way to think about this is to visualize stacking poker
chips versus marbles in a thermoset matrix. The flake geometry provides a larger
contact area and thus promotes maximum particle–particle interaction. A third
requirement for achieving good electrical conductivity is to treat the filler surface,
allowing the silver flakes to come in very close proximity. The reason silver is an
excellent filler for ECAs is that it has a high affinity for silver–silver interaction.
When silver flakes are produced, a lubricant is applied to prevent the silver from
massive agglomeration. Typical lubricants are fatty acids such as stearic, oleic,
linoleic, and palmitic acids (83). The lubricants keep the silver flakes flowable,
so they can be compounded into the resin matrix. During curing, the lubricants
must be removed in order for the silver flakes to come in close proximity to one
another. Conductivity promoters such as short chain acids (like adipic acid) are
used in small amounts in the resin. Studies have shown that the lubricants are
36
THERMOSETS
1.2
Diamond
BN
Alumina
Silica
1.0
0.8
0.6
0.4
0.2
0.0
0
10
20
30
40
50
Filler content (vol%)
Fig. 21. Thermal conductivity versus filler content for an epoxy anhydride resins system.
Reprinted with permission from Ref. 82. Copyright 2005 Elsevier.
replaced at the silver surface by various types of acids with the result of increased
electrical conductivity (83). Optimization of the silver flake geometry, the lubricant, and conductivity promoters have resulted in adhesives with electrical resistivity less than 2 × 10−5 ohm-cm. Note that most ECA data sheets report resistivity in ohm-centimeters.
5.3. Particulate Fillers for Thermal Conductivity. The thermal conductivity (k) of an unfilled epoxy is about 0.2 W/m⋅K. Thermal conductivity of
thermosets that are also electrically conducting was discussed in the preceding
section. This property of a thermoset can be increased to 1–2.5 W/m⋅K while
maintaining its electrical resistance by the incorporation of fillers like quartz (caxis; k = 12 W/m⋅K), alumina (k = ∼30 W/m⋅K), BN (k = 300 W/m⋅K), or diamond
(k = 2000 W/m⋅K). Analogous to lowering the CTE as discussed above, adding
thermally conductive fillers can increase the composite thermal conductivity as
seen in Figure 21 (82).
As discussed in the preceding section on electrical conductivity, the filler aspect ratio or geometry also plays an important role in the development of thermal
conductivity. In Figure 21, BN was the most effective filler for increasing thermal
conductivity. The morphology of BN is plate like (84). In the work of Lee and Yu
(82), they used scanning electron microscopy to investigate the filler morphology
used in a thermoset matrix. In Figure 23, the filler geometries clearly show the
spherical nature of the silica and alumina fillers compared with the plate-like
morphology of the BN fillers.
Owing to the efficacy of the BN filler interactions, the thermal conductivity
increased nearly fivefold at 15 vol% loading. At this loading, as seen in Figure 21,
the thermal conductivity of the BN-filled composite is approximately twice that of
the diamond filled composite even though the thermal conductivity of diamond is
nearly an order of magnitude larger than BN. The data clearly demonstrate the
role of the filler aspect ratio and geometry on the thermal conductivity of the filled
composite.
THERMOSETS
37
5.4. Fibers. Glass fibers are commonly added to a thermoset, either as
chopped fibers dispersed in the liquid resin system or in fabric form. To achieve
high performance properties, it may be necessary to resort to carbon fiber reinforced composites. Less common is the use of organic fibers such as cellulose or
Kevlar, a high performance aramid fiber.
Glass fibers are used as a reinforcing agent for many polymer products to
form a strong and relatively lightweight fiber-reinforced polymer (FRP) composite
material called glass-reinforced plastic, also popularly known as “fiberglass.” The
most common type of glass fiber used is E-glass, which is aluminoborosilicate glass
with less than 1% w/w alkali oxides. E-glass has excellent mechanical properties
with modulus up to 50 GPa, tensile strength of 2–3 Gpa, and failure strains in
the range of 2.5%. In contrast to carbon fiber, glass can undergo more elongation
before it breaks. Thermal expansion is about 5 × 10−6 K−1 .
The interfacial strength between glass and thermosetting polymer matrices
has been significantly enhanced by silane-coupling treatments of glass fibers (85).
The coupling agents are generally organosilanes, X3 SiR, where R is a group that
can react with the polymer matrix and X is a group that can hydrolyze to form a
silanol group that can react with a hydroxyl group on the glass surface.
Most carbon fibers used in high-performance polymer matrix structural composites are made from polyacrylonitrile (PAN) fibers whereas those for less demanding applications can be manufactured using pitch or rayon as the precursor.
They possess excellent mechanical properties with modulus up to 900 GPa (high
modulus fibers), tensile strengths up to 7 GPa (high tensile fibers), and failure
strains in the range of 2.5%.
PAN fibers are initially oxidized at 200–400◦ C to maintain molecular orientation. The cross-linked, oxidized PAN fibers are then carbonized at 800–1000◦ C,
eliminating nitrogen and hydrogen from the fiber. The mean transverse CTE
varies from 5 × 10−6 to 10 × 10−6 K−1 and the longitudinal CTE from 1.6 × 10−6 to
2.1 × 10−6 K−1 depending on the fiber structural properties (86). During the carbonization process, the fibers are drawn to produce fibers 5–10 𝜇m in diameter and
ellipsoidal microvoids oriented with their principal axes in the fiber direction. The
microvoid geometry and void characteristics directly control fiber strength. The
fiber surfaces are passed through a dry oxidative environment such as air at elevated temperature or a wet oxidative environment, such as HNO3 , to produce
hydroxyl, carboxylic acid, or nitrate surface groups conducive to bonding to polymer matrices (87). At a later stage of the process, a proprietary coating or sizing
is deposited on the fiber surface, generally an epoxide. Such sizings can improve
fiber wettability, protect the fiber surface reactivity, and prevent surface damage
that could occur during composite fabrication.
5.5. Nanoscale Reinforcement. “Nano” refers to a size scale measured
in nanometers (nm), or 10−9 m. A working definition for the nanoscale is molecules
and structures with at least one dimension between 1 and 100 nm (88,89). In
addition to size, shape is also a component of nanostructures (90,91). The benefits of using nanoscale constituents, such as clay platelets, carbon nanotubes
and nanofibers, and silica-based organic hybrids, have been widely studied since
the early 1990s (see NANOCOMPOSITES, POLYMER–CLAY; NANOCOMPOSITES, LAYER-BY-LAYER ASSEMBLY). Azeez and colleagues (92) give an excellent review of the
38
THERMOSETS
processing, properties, and applications of epoxy clay nanocomposites. They note
that these materials reinforced with glass or carbon fiber have great potential for
application in the automotive and aircraft industries due to reduced component
weight and enhanced mechanical properties (93).
The unique combination of large interfacial area and small interparticle distances is responsible for material responses unexpected from that observed in
conventional microcomposites. Nanocomposite filler loading is typically <10% and
often <5% by weight as compared to 20–40% for conventional materials. The addition of nanoscale constituents provides not only reinforcement at low volume fractions but also synergism with the matrix molecules with which it shares the same
size scale. Significant enhancements have been observed in modulus, strength,
and toughness; dimensional stability at elevated temperature; decreased swellability and barrier resistance to small molecules; thermal stability and fire retardancy; and electrical and thermal conductivity. Both increases and decreases in
Tg have been observed in thermoset nanocomposites.
A number of studies have reported the processing and properties of thermoset/clay nanocomposites (94). Nanoclay materials, such as montmorillonite, are
layered silicates, that is they have a lamellar structure. Elementary clay platelets
consist of a 1-nm thick layer. Stacking of layers leads to a regular gap called a
gallery or interlayer whose d-spacing can be measured by X-ray diffraction. The
interlayer space can be penetrated by organic cations or polar organic liquids. Exchange reactions with organic cations (eg, amino acids or alkylammonium ions)
render the silicate surface organophilic. Such materials are referred to as organically modified layered silicates. To gain the property enhancements described
above requires dispersion of the filler at the nanometer level, that is down to the elementary clay platelet. Two morphologies are possible: An intercalated nanocomposite in which the unreacted resin penetrates the interlayer and is subsequently
cured and an exfoliated or delaminated nanocomposite where clay layers are
individually dispersed in the thermoset matrix. Intercalation leads to increased
d-spacing and can eventually result in exfoliation. Many nanocomposite systems
are a blend of intercalated and exfoliated structures. The ability to process thermoset nanocomposites depends on several compositional factors such as resin viscosity and reactivity, catalysts, and the effect of the nanoparticle itself. The filler
can play a significant role in the rheology of the uncured system as well as catalyze the chemical cross-linking reactions. Organically modified layered silicates
frequently catalyze the reaction so that the rate of intragallery cure exceeds extragallery cure and that in the bulk matrix. The extent and rate of cure determine
gelation and ultimately the final morphology, that is intercalated or exfoliated.
Composites with single walled carbon nanotube, multiwalled carbon nanotube, and carbon nanofiber reinforcements are referred to as one-dimensional
(1D) nanocarbon composites (95). They show promise for new lightweight materials with incredible mechanical, electrical, and thermal properties. 1D nanocarbon
materials are envisioned as multifunctional materials, for example, single materials used for structures as well as electrical and/or thermal conductors. One example is electronics in a space satellite that need to be lightweight and mechanically
supported, have the excess heat dissipated, and be protected from electromagnetic
interference. Other examples are structures that are also batteries and structures
that store hydrogen for fuel cells. 1D nanocarbon can also be added at low volume
THERMOSETS
39
fractions to the matrix for conventional continuous fiber-reinforced composites,
offering the unique capability of three-dimensional reinforcement (93).
Silica nanoparticle reinforcement has received some attention (96). A significant increase in fracture energy was observed for an epoxy–anhydride system
with silica nanoparticles ∼20 nm in diameter (96). Various toughening mechanisms were explored. Microscopic evidence supported a model of plastic void
growth that was most likely responsible for the increased toughness due to the
presence of the silica nanoparticles.
Polyhedral oligomeric silsesquioxane (POSS) reagents offer an opportunity
for preparing molecularly dispersed nanocomposites (98). They combine a hybrid inorganic–organic composition with nanosized cage structures having dimensions of 1–3 nm, comparable to those of most polymeric segments or coils.
POSS compounds contain an inner inorganic framework made up of silicon
and oxygen, which is externally covered by organic substituents that can embody a range of polar structures and functional groups. For example, monofunctional and multifunctional epoxy–POSS compounds react with epoxy systems and
methacryl–POSS compounds react with acrylic and vinyl ester resin systems.
Pittman and co-workers (99) have also conducted investigations on several additional POSS/thermosetting systems, including phenolic, dicyanate, and vinyl ester.
5.6. Toughening Agents. Thermosets typically have high modulus but
tend to be brittle and fracture at very low strains. Unnikrishnan and Thachil
(100) and Thomas and colleagues (101) present nice reviews of toughening of
epoxy resins. Toughening agents are often employed to increase flexibility, fracture toughness, and impact resistance so that these materials can be considered
for more demanding applications. One example is carbon fiber composites for
primary aircraft structures. Toughening agents typically fall into two categories
resulting in low Tg or rubber-modified and high Tg or thermoplastic-modified systems. In both cases, a homogeneous multicomponent system, a typical example of
which is an epoxy resin, amine hardener, and organic/polymeric modifier, phase
separates during cure into a heterogeneous morphology as a result of the growing
epoxy chain becoming less compatible with the toughening component (102). This
process is known as reaction-induced phase separation or RIPS.
To avoid deterioration of mechanical properties such as elastic modulus and
yield stress, low Tg modifiers can be substituted with high Tg engineering thermoplastics like poly(ether sulfone) or poly(ether imide) (103,104). However, the
improvements in impact resistance have proven to be only marginal and only the
rubber-toughened approach is in widespread use today.
To improve the impact resistance of inherently brittle epoxy thermosets,
toughening with rubber particles is commonly used (105). Such RIPS is the result of a decrease in the entropic contribution to the free energy of mixing during
polymerization (103). As an example, it has been observed that at low levels rubber modification can improve the toughness and impact properties of cured epoxy
resins (106,107). In situ phase separation occurs during cure of the epoxy matrix,
and the dispersed rubbery phase can introduce energy dissipation mechanisms.
An epoxy–amine system with excess epoxy and low levels of a carboxyl-terminated
reactive liquid rubber copolymer of butadiende and acrylonitrile (CTBN) is a typical example (108,109). The system is homogeneous at the start of cure, provided
40
THERMOSETS
the cure temperature is above some critical solubility temperature. Initially, in
the presence of piperadine, a secondary amine that is effective in promoting the
homopolymerization of epoxy, reaction of the carboxyl end groups of the CTBN
with epoxide produces a rubber-containing diepoxide that reacts with the remaining epoxy as cure proceeds. The rubber and epoxy become less compatible with
increasing cure, and a phase separation point is reached where rubber-rich domains precipitate in the epoxy-rich matrix. Typical domain sizes range from 0.1
to 10 𝜇m with volume factions from 0.01 to 0.20, both dependent on formulation and the cure process. Once gelation has occurred the morphology is fixed
(106,110–114). Long gel times promote complete phase separation. If the material gels quickly, the rubber domains may not have sufficient time to develop and
fewer, smaller domains form (111). If gelation occurs prior to phase separation,
the rubber is trapped in the network structure and no domains form.
Both the material and the process influence the phase separation. The compatibility of the rubber and the epoxy can be controlled by the acrylonitrile content of the rubber modifier as well as the cure conditions. CTBN modifiers of
higher acrylonitrile content are more compatible with epoxy in terms of solubility parameters, and they precipitate at a later stage of cure. Similarly, variation
in the cure process up to the gel point will produce different morphologies, and
therefore different material properties, from a single rubber-modified epoxy formulation. The volume fraction, domain size, and number of particles of phaseseparated rubber are determined by the competing effects of incompatibility, rate
of nucleation and domain growth, and the quenching of morphological development by gelation. While most modifiers yield morphologies in the micrometer
range, more recent interest in nanostructured thermosets has initiated research
in amphiphilic block copolymers as well (6,114,115, Chapt. 5 in Ref. 4, Chapt. 5
in Ref. 7, Chapt. 3 in Ref. 101). Modification of thermoset resins such as epoxies by these block copolymers can lead to the formation of nanosized inclusions
through self-assembly. The main advantage of block copolymers is that feature
sizes are determined by the lengths of the blocks, typically on the scale of nanometers, compared with micrometer size features formed by homopolymers or random
copolymers. Liquid crystalline thermosets, discussed in Section 3.3 (46), also undergo self-organization and may be described as nanostructured thermosets, for
example, they exhibit enhanced impact properties. Other properties can benefit as
well.
5.5. Effect on Rheology. In the previous sections, the role of fillers has
been discussed as playing a key role in the property modification of thermosets.
The advantage of using thermosets comes from the ability to process thermoset
formulations in the liquid (uncured) state to achieve the desired shape or function
and the curing reactions allow achievement of the final fully cured properties.
In this section, the impact that fillers have on the rheological properties will be
presented.
Thermoset monomers and oligomers are low viscosity liquids. In rheological terms, they are Newtonian liquids in that the viscosity is not a function of
the shear rate (or rate of deformation during processing). However, the addition
of fillers significantly changes the rheological properties. During the processing
of highly filled thermoset resins, the initial processing typically involves depositing or applying the filled resin in processes such as adhesive dispensing or bulk
THERMOSETS
41
Fig. 22. SEM micrographs of fillers: (a) silica, (b) diamond, (c) BN, all at 5,000×, (d) alumina at 20,000×. Reprinted with permission from Ref. 82. Copyright 2005 Elsevier.
molding compound (BMC) or sheet molding compound (SMC) processes. Three
factors impact the rheology in uncured systems: filler loading, filler aspect ratio
or geometry, and filler/matrix interfacial interaction.
In the case of highly filled adhesives such as die attach (for bonding semiconductor chips to a substrate in electronics), the viscosity increases significantly
with increased filler loading. This is illustrated in Figure 23 where viscosity is
plotted as a function of shear rate for an epoxy resin loaded at various levels with
spherical silica fillers.
The yellow-green symbols at the bottom of the graph represent the viscosity
profile for the base thermoset epoxy resin only. The viscosity is not a function of
the shear rate resulting in Newtonian behavior. With increasing filler loading,
the viscosity steadily increases, most dramatically at the lower shear rates. The
decreasing viscosity with increasing shear rate is called shear thinning (or also
may be termed shear rate thinning). Many filled adhesives take advantage of this
rheological property during dispensing. The shear rate is high in the needle or
nozzle resulting in a lower viscosity at all filler loadings. The formulator can use
the filler loading to adjust the flow properties depending on the application. In
the case of electrically and thermally conductive adhesives, high filler loadings
are typically required to reach the percolation threshold, thus the filler loading
and viscosity must be optimized to achieve the desired final properties along with
having acceptable processing (rheology).
The viscosity shear rate relationship is also impacted by the filler geometry
or aspect ratio. In Figure 24, the relative viscosity is plotted as a function of filler
volume fraction to demonstrate the role of filler geometry.
42
THERMOSETS
10000
0%
10%
20%
30%
40%
50%
60%
1000
100
10
0.001
0.01
0.1
1
10
100
Fig. 23. Viscosity versus Shear Rate. Reprinted with permission from Ref. 116. Copyright
2007 ICI Corporation.
20
spheres
14
grains
rods
16
plates
18
12
10
8
Increased aspect ratio
6
4
2
0
0
0.2
0.4
0.6
Fig. 24. Viscosity versus Filler Volume Fraction. Reprinted with permission from
Ref. 116. Copyright 2007 ICI Corporation.
For a given volume fraction loading, for example 0.2 in Figure 24, the viscosity increases by a factor of approximately five times in going from spheres to
rods. In a shear field at low filler loadings, spherical particles can roll past one another and have a small impact on the viscosity. As the aspect ratio increases, even
at low filler loading, the particle–particle interactions have a large effect on the
viscosity. BN was an effective filler at increasing the thermal conductivity of the
thermoset composite, but has a plate-like morphology and will make dispensing or
THERMOSETS
43
1.5
1000
8000
6000
0.5
Tan Delta
Storage modulus (MPa)
1.0
4000
0.0
2000
0
0
10
20
30
40
50
60
70
Temperature (ºC)
80
90
100
110
120
Fig. 25. Multifrequency DMA of typical thermoset. Storage modulus versus temperature
during heating at 2◦ C/min. Copyright TA Instruments.
processing more difficult. Formulators must optimize the filler loading and filler
geometry to achieve the desired final cured properties with acceptable processing.
6. Materials and Process Characterization
In this section, important thermoset material properties and characterization
methods will be discussed.
6.1. Glass Transition Temperature Tg . Tg is an important property in
several respects. For example, a material’s HDT, a common measure of upper use
temperature, will be determined by its Tg . Depending on the application, it might
be wise to limit a thermoset’s temperature exposure to 30–50◦ C below its Tg . As
a material approaches its Tg , starting approximately 20–30◦ C below Tg , physical
aging will occur at a noticeable rate and increase with the approach to Tg . Residual
stresses will also begin to relax over this temperature range, which may result in
dimensional changes.
As described in Section 2.4, the Tg –conversion relationship established Tg
as a valid measure of degree of cure. In many cases, for example, where the DSC
exotherm is small or contains overlapping endothermic events, it is the preferred
measure. Tg can be measured with good accuracy by several techniques, including
DSC (Figs. 5 and 26), thermomechanical analysis (TMA) (Fig. 46), DMA (Fig. 25),
and dielectric analysis.
Because it is a kinetic property, Tg is not a single point like a melting temperature but its value will change depending on method of measurement, where
in the transition Tg is taken and experimental parameters such as frequency. This
44
THERMOSETS
0.15
2.3
0.10
1
2 3
Tg∞
1.8
0.05
1
2 3
1.3
0.00
exo
50
100
150
Temperature (oC)
200
0.8
Fig. 26. MTDSC of Epoxy Anhydride after Partial Cure at 85 ◦ C. 2.5C◦ /min (±1 ◦ C/60 s).
Reprinted with permission from Ref. 120. Copyright 2006 Springer.
is illustrated in Figure 25, which shows the storage modulus and tan 𝛿 at several
frequencies for a cured thermoset. For example, Tg taken as the peak in tan 𝛿 will
increase by 5–7◦ C per decade increase in frequency, for example, from 1 to 10 Hz.
Although not shown the peak in the loss modulus is also a valid measure of Tg .
Since the value reported for Tg is dependent on how it is measured, it is necessary to include the method and conditions of measurement, for example, the DSC
midpoint at 10◦ C/min or the DMA peak in tan 𝛿 at 1 Hz and 2◦ C/min.
6.2. Modulated Temperature DSC. Modulated temperature differential scanning calorimetry (MTDSC), also known as TMDSC, is a family of techniques where a temperature modulation is overlaid on a linear heating or cooling
ramp resulting in a modulated heat flow. Good references include Riga and Judovits (117), Menczel and colleagues (118), and Reading and Hourston (119). This
technique is especially suited to characterizing thermoset cure because it can separate the DSC signal into reversing effects such as the heat capacity and nonreversing effects such as the cure exotherm. Since the glass transition is observed
as a step change in the heat capacity, MTDSC is an excellent way to measure
Tg free from nonreversing events such as enthalpy relaxation, residual cure, and
evaporation of volatile components. Below are examples of MTDSC of thermosets.
Figure 26 shows the nonreversing heat flow and heat capacity (Cpr ) curves
for an epoxy–anhydride system cured to different degrees at Tcure = 85◦ C. After full cure Tg∞ for this system was observed to be 135◦ C. Curve 1, after 165
min at 85◦ C, shows a Tg that is below Tcure and well separated from the residual exotherm, indicating that vitrification had not occurred during the isothermal
cure. After 230 min, we see in Curve 2 Tg in the vicinity of Tcure accompanied by
a small endothermic enthalpy relaxation peak. Enthalpy relaxation only occurs
below Tg , so its presence indicates that vitrification has occurred, but only to a
small extent. After 800 min, we can observe in Curve 3 that Tg has increased
15–20◦ C beyond Tcure and is accompanied by a large enthalpy relaxation peak,
THERMOSETS
45
Fig. 27. Non-isothermal cure of an epoxy–anhydride. Reprinted with permission from
Ref. 120. Copyright 2006 Springer.
indicating that extensive vitrification has occurred. Note the overlap of the enthalpy endotherm with the residual cure exotherm in the nonreversing heat flow.
Curing at heating rates that are slow enough to maintain the thermoset in
the glassy state but still fast enough to complete the cure in a reasonable amount
of time has practical applications. For example, it can be employed to control the
exotherm of highly reactive systems and prevent thermal runaways, especially
dangerous when large masses are involved. And it can be used to complete cure
without expensive fixturing while maintaining dimensional stability. For example,
a novel processing scheme, referred to as reactive thermoset processing (RTP),
proposes to eliminate expensive fixturing by maintaining the curing thermoset in
the “just glassy” state, where the curing thermoset is dimensionally stable but still
sufficiently reactive (121). RTP may also be effective in reducing residual stresses
induced by the process (122). An application of this approach is the amplifier used
in the Bell System undersea cable (123).The optical fiber is held in alignment and
partially cured to Tg ≈ 60◦ C, where Tg∞ for this epoxy is 105◦ C. Then the device is
removed from the fixture, and the cure of several devices is completed by heating
them in an oven at about 0.5◦ C/min until cure is complete. MTDSC can help in
establishing an appropriate heating rate for RTP.
Figure 27 shows the nonreversing heat flow and heat capacity curves for an
epoxy system at three heating rates. Curve 4 is the heat capacity after full cure,
which shows the glassy heat capacity and Tg∞ at ∼140◦ C.
•
•
At 0.2◦ C/min ramp rate (1), we see the liquid/rubbery heat capacity begin
to decrease upon vitrification just below 100◦ C and remain between the liquid/rubbery and glassy heat capacities as cure proceeds in the partially vitrified state. When cure is complete, devitrification occurs and the heat capacity
returns to the liquid/rubbery line.
At 0.4◦ C/min (2), vitrification occurs at a higher temperature and to a lesser
extent.
THERMOSETS
0.20
(a)
0.15
0.10
exo
0.05
0.1
reaction heat capacity
(b)
–1
–1
Heat capacity change (Jg K )
–1
NR heat flow (-Wg )
46
vitrification
0.0
–0.1
–0.2
–0.3
0
200
Time (min)
400
Fig. 28. QiDSC of diglycidal ether of bisphenol A and methylenedianiline at equal stoichiometry (Tg∞ ≈ 170 ◦ C). ± 1◦ C/60 s. The blue stars mark the vitrification point at ½ΔCp .
Reprinted with permission from Ref. 124. Copyright 2006 John Wiley & Sons, Inc.
•
At 0.7◦ C/min (3), vitrification is not observed and cure is accomplished entirely
under chemical control.
Thus MTDSC tells us that heating a vitrified sample at 0.2–0.4◦ C/min would
advance the cure but maintain the sample in the vitrified state. Measurements
such as these can establish an appropriate heating rate for RTP and save valuable
time compared to the alternative trial and error methods. They are also useful
for determining the minimum heating rate for multiple heating rate experiments
where cure needs to proceed to completion without vitrification (see Section 4.2).
Figure 28 illustrates MTDSC in the isothermal mode, referred to as quasiisothermal or Quasi-isothermal DSC (QiDSC) due to the temperature modulation.
The nonreversing heat flow (NHF) curves in a are similar to standard DSC curves
and contain the same information. For example, the cure of this epoxy system is
clearly autocatalytic as evidenced by the occurrence of the maximum reaction rate
well into the reaction and not at t = 0 characteristic of nth-order reactions. The
data are of sufficient quality to lend itself to a kinetic analysis. However, what is
not readily apparent in these curves is the occurrence of vitrification that causes
the reaction to slow down considerably before a full cure is achieved. The unique
ability of MTDSC to separate the reversing and nonreversing components of the
DSC curve allows vitrification to be clearly observed in the reversing heat capacity
curves, separate from the cure exotherm in the NHF curves. With that information
and upon close inspection the reaction can be seen to slow down following that
transition. These data can now be used to characterize the cure reaction under
chemical control plus the transition to diffusion control following vitrification.
THERMOSETS
47
6.3. Rheology, DMA, and Dielectric Analysis. Thermosetting resins
allow a wide range of processing options. When low molecular weight monomers
or oligomers are used as the starting materials, the viscosity can be very low,
enabling a wide variety of application methods. Typically, with the application of
heat, the low molecular weight species begin to react and increase their molecular
weight. As the network builds, the viscosity increases.
Steady shearing rheological methods are good for precisely measuring the
viscosity early in the curing and up to the gel point. For low viscosity monomers
and oligomers, cone and plate and parallel plate geometries are typically
used. Additionally, most rheometers allow precise temperature control, enabling
isothermal curing studies to be conducted. If viscosity development is required
over the entire conversion range, then steady shearing rheometry is not the recommended choice, since large torques will build up after the gel point and could
potentially damage the torque transducers (see RHEOLOGICAL MEASUREMENTS).
Dynamic mechanical methods (typically oscillatory parallel plate rheometry)
are commonly used to measure the dynamic mechanical properties from the liquid state to the solid state. By using small-amplitude oscillatory deformations (to
remain in the linear viscoelastic regime), the dynamic storage and loss moduli can
be obtained. From these quantities, the viscosity and modulus can be calculated
(125) (see DYNAMIC MECHANICAL ANALYSIS).
Dynamic mechanical measurements are useful to probe the curing of thermosets. The dynamic moduli can be measured at various frequencies, along with
using isothermal or nonisothermal temperature profiles. The dynamic storage
modulus G′ (𝜔) or E′ (𝜔) is the energy stored or the elastic component, where 𝜔
is the frequency. The dynamic loss modulus, G′′ (𝜔) or E′′ (𝜔) is the energy lost or
the viscous component. tan 𝛿 = E′′ /E′ = G′′ /G′ is the phase lag between stress and
strain and a common measure of damping. In rotational rheometry, the dynamic
moduli, G′ (𝜔) and G′′ (𝜔), can be used to calculate the complex viscosity 𝜂 * (𝜔) (125):
√
𝜂 ∗ (𝜔) =
[
]2 [
]2
G′ (𝜔) + G′′ (𝜔)
𝜔
(12)
where G′ (𝜔) is the dynamic storage modulus and G′′ (𝜔) is the dynamic loss modulus. From the dynamic moduli, G′ (𝜔) and G′′ (𝜔), the complex viscosity 𝜂 * (𝜔) can
be calculated.
In Figure 29, the complex viscosity is plotted as a function of time for isothermal cure of an epoxy-based thermoset. The initial viscosity is low and increases
rapidly at various temperatures.
In many composite applications, B-staged resins and fiber-reinforced
prepregs are used. For B-staged thermoset resins, measuring the viscosity profile during nonisothermal curing provides valuable insight into the flow window,
or the time when the resin is able to flow during processing. In Figure 30, the
complex viscosity and the percent conversion are plotted as a function of time for
a heating rate of 9.8◦ C/min to 175◦ C. In this experiment, the rheometer run was
aborted at specific time intervals, and the sample rapidly quenched and analyzed
for Tg and conversion using DSC (43). The initial B-stage conversion was 23%.
48
THERMOSETS
Fig. 29. Complex viscosity versus curing time for isothermal curing of epoxy resin. (1
Pa⋅s = 10 P). ◦ 123◦ C, □ 135◦ C, ▵ 145◦ C, • 157◦ C, ■ 162◦ C. Reprinted with permission
from Ref. 43. Copyright 1990 American Chemical Society.
The viscosity rapidly decreases as the resin softens at Tg . With further heating,
the viscosity reaches a minimum and then begins to increase with the onset of
the cure reaction. During the initial softening of the resin, the conversion did not
change, indicating that the temperature dependence of the viscosity is the controlling factor. When the viscosity approaches its minimum, the conversion begins to
increase dramatically. The continuation of the reaction causes the viscosity to increase and ultimately leads through gelation toward full cure.
Whereas steady shear rheometry can measure the initial stages of cure up to
the gel point (as illustrated in the left-hand side of Fig. 2), DMA can measure the
cure process of unsupported thermosets from just after gelation to vitrification or
the completion of cure, as well as the properties of cured thermosets (see the righthand side of Fig. 2). Oscillatory rheometry can measure viscoelastic properties
during cure from the liquid state to the solid state, as illustrated in Figures. 29
and 30.
Recently, Block and co-workers developed an innovative technique called
RheoDSC (126) to investigate the curing behavior of an epoxy amine thermoset.
This method combines rheological and calorimetric measurements inside a DSC
THERMOSETS
49
Fig. 30. Complex viscosity (■) and epoxy conversion (◦) as a function of curing time for a
sample heating rate of 9.8◦ C/min. (1 Pa⋅s = 10 P). Gelation occurs in the range of 40–50%
conversion. Reprinted with permission from Ref. 43. Copyright 1990 American Chemical
Society.
Fig. 31. Dynamic moduli and tan 𝛿 plotted as a function of time for isothermal curing of a
DGEBA epoxy and MDA (4,4′ -methylenedianiline) at 90◦ C and multiple frequencies. Left:
0.25, 0.44 and 0.79 Hz; right: 0.079, 0.14. 0.25, 0.44 and 0.79 Hz. Reprinted with permission
from Ref. 126. Copyright 2013 Elsevier.
cell. Figure 31 shows the dynamic moduli at multiple frequencies as a function of
temperature during isothermal curing at 90◦ C (126).
The uncured state (1) consists of difunctional epoxy (diglycidal ether of
bisphenol A [DGEBA]) and tetrafunctional amine (methylenedianiline [MDA]).
During isothermal curing, the first chemical reactions lead to chain extension up
50
THERMOSETS
160
140
120
100
Gelation
Vitrification
80
60
40
20
0
–20
10
100
1000
Time (min)
Fig. 32. TTT diagram for the DGEBA-MDA system constructed with the RheoDSC
method. Gelation line from point of frequency independence in tan 𝛿. Vitrification line from
the maximum in G′′ at 0.079 Hz. Reprinted with permission from Ref. 126. Copyright 2013
Elsevier.
to ∼35 min, as seen in the cartoon image (2). In the corresponding plot of G′ and
G′′ , one observes very little change in the loss modulus, G′′ , during the initial chain
extension. With further reaction of the epoxy with the multifunctional amine, the
network structure builds leading to a rapid increase in loss modulus, G′′ , and
the appearance of the storage modulus, G′ . Recall that the storage modulus is
the elastic component of the dynamic moduli and as expected, the building network exhibits elastic behavior. Examination of tan 𝛿 in the right side of Figure 31
shows that it becomes frequency independent between 75 and 80 min, satisfying
the classic definition of gelation (9–12,14). Using tan 𝛿 in DMA proves a good way
to measure gelation, a critical event during processing of thermosets. As discussed
previously, after gelation macroscopic flow ceases and this marks the end of flow
during processing.
For this epoxy amine system, Tg∞ (150◦ C) is greater than 90◦ C and therefore
one would expect vitrification to occur during isothermal curing this temperature.
In Figure 31, the cure progresses to the stage where Tg = Tcure which is the definition of vitrification. At (4), the series of frequency dependent tan 𝛿 peaks in the
right-hand figure show the occurrence of vitrification. Taking the time to vitrify
tvit from the peak in tan 𝛿, there is an approximately 10 min difference from the
highest frequency (shortest tvit ) to the lowest frequency (highest tvit ), illustrating
the frequency dependence of this transition.
By measuring the dynamic mechanical properties at several isothermal cure
temperatures, a TTT diagram can be constructed. In Figure 32, the TTT diagram
for the DGEBA-MDA system above was constructed using a series of isothermal
curing temperatures.
DMA is also useful to characterize the modulus temperature relationship using either three-point bending or double or single cantilever beam methods. Figure 33 compares the dynamic mechanical properties of hypothetical lightly crosslinked and highly cross-linked thermosets. The glassy moduli are indistinguishable, which is characteristic of all polymers. The storage modulus of the lightly
THERMOSETS
51
Fig. 33. Storage and loss moduli of hypothetical thermosets illustrating the glassy modulus, the glass transition, and the rubbery plateau. Copyright TA Instruments.
cross-linked thermoset shows a large step decrease starting at a low temperature
and ending at a low rubbery modulus. The loss modulus is intense with a maximum (one measure of Tg ) at a low temperature. The storage modulus of the highly
cross-linked thermoset shows a small step decrease at a higher temperature and
ending at a higher rubbery modulus. The loss modulus is less intense with its
maximum at a higher temperature.
When uncured thermosets in liquid or paste form are applied to a substrate
such as a metal shim (3,127), wire mesh (3,64,75,128,129), or glass fiber or filter
media (3,16,74), cure can be measured from the uncured state to the completion
of the reaction. Isothermal gelation and vitrification can be measured directly at
several temperatures to generate TTT diagrams such as Figures 3 and 32 (17), or
cure can be followed by measuring Tg versus time at several temperatures such as
in Figure 11 (19,74,128,129). Figure 34 illustrates the wire-mesh technique which
has been used in the characterization of the cure of magnetic disk coating (75),
powder coatings (64), and dielectric and conducting polymer thick films (129).
Figure 35 shows the multifrequency DMA during heating of a B-staged epoxy
from below Tg through the completion of cure (130). The frequency-dependent
transition starting near room temperature is the glass transition of the B-staged
epoxy, which gels near 140◦ C during heating at 2◦ C/min., as evidenced by the
small increase in storage modulus and frequency-independent tan 𝛿 peaks.
Analogous to dynamic mechanical measurements, dynamic dielectric measurements can also be used effectively to measure the changes in thermosetting
polymers during curing. Typical dielectric measurements use two polarized electrodes. Early methods used parallel plate electrodes. More recently, interdigitated
comb electrodes have proven to be more versatile for cure monitoring. All dielectric measurements are made by applying a sinusoidal voltage across the material between the electrodes and measuring the in-phase and out-of-phase current.
From these measurements, the conductance and capacitance are obtained. The
52
THERMOSETS
Fig. 34. Illustration of wire-mesh technique for DMA of liquid or paste materials. Tensile
mode illustrated with wire-mesh cut at 45◦ bias to minimize shear resistance of mesh. R.
B. Prime, unpublished.
Fig. 35. Multifrequency DMA of a B-staged, fiber glass-reinforced adhesive during
2◦ C/min scan. 0.5, 1, 2, 5, 10 and 20 Hz. Reprinted with permission from Ref. 130. Copyright
1991 North American Thermal Analysis Society.
THERMOSETS
53
Fig. 36. Dielectric loss factor as a function of time for isothermal curing of an epoxy resin.
Reprinted with permission from Ref. 131. Copyright 1986 Springer.
conductance is a measurement of the dissipation of energy (like the dynamic loss
modulus in rheology and DMA), and the capacitance is an indication of the energy stored (analogous to the dynamic storage modulus). In the case of the dielectric measurement, the conductivity consists of two components, one arising from
dipole orientation and the second arising from ionic conductivity (131).
The dielectric loss factor 𝜀′′ can be expressed as
𝜀′′ =
𝜎
+ 𝜀′′
d
2𝜋f 𝜀0
(13)
where 𝜎 is the ionic conductivity, f is the frequency, 𝜀0 is the permittivity of free
space, and 𝜀′′ d is the lumped contribution from all dipoles. The ionic conductivity
for a given material is controlled by the number and charge of the ionic species
and, most importantly, the mobility of the ion (see DIELECTRIC RELAXATION).
For thermoset curing, there is typically a large change in segmental mobility
during curing because of the stiffening of the developing cross-linked network.
Measuring the dielectric loss factor as a function of time and temperature is an
effective monitor of cure (131).
In Figure 36, the dielectric loss factor is plotted as a function of time for
isothermal curing of an epoxy resin. Early in the curing, the viscosity is low, and
thus the ionic mobility is high, leading to a large value in the loss factor. As curing
proceeds, the increase in molecular weight and cross-linking cause the ionic mobility to decrease due to increasing viscosity and the tightening of the network resulting in lower segmental mobility. The series of frequency-dependent loss peaks
have been correlated with vitrification (132), analogous to those observed by DMA
or rheology, for example, in Figure 31. An excellent review of dielectric measurements and cure monitoring for thermosetting resins may be found in a review by
Senturia and Sheppard (131).
Several important industrial applications involve the processing of B-staged
prepregs in composite fabrication. Embedded dielectric cure monitors can be used
to investigate the changes in the network structure during curing.
54
THERMOSETS
Fig. 37. Dielectric loss factor (right axis) and complex viscosity (left axis) plotted as a
function of time for a nonisothermal ramp to the cure temperature. Reprinted with permission from Ref. 133. Copyright 1989 John Wiley and Sons, Inc.
In Figure 37, the dielectric loss factor at several frequencies is plotted as
a function of time for a nonisothermal ramp and hold at the cure temperature.
Also in Figure 37, the complex viscosity (at 1 Hz) is plotted as a function of cure
temperature. In this case, the viscosity and dielectric properties were measured
simultaneously in a parallel plate rheometer (133). The figure clearly indicates increasing conductivity as the temperature exceeds the Tg of the B-staged resin and
the viscosity decreases. With further curing, the viscosity goes through a minimum
and then increases because of the rapid tightening of the cross-linked network.
The dielectric loss factor goes through a maximum and then decreases rapidly
as the cross-linking reaction proceeds. From the simultaneous measurement, the
correlation of the loss factor and the complex viscosity can be established (133). Interestingly, because of the limitation of the torque sensor, the viscosity data could
not be collected through the completion of the cure. The dielectric sensor is capable
of measuring the continued segmental mobility change far into the reaction, giving
a more complete analysis of the curing profile. In Figure 37, the final cure temperature was chosen to be lower than the Tg of the fully cured network. During cure,
the Tg of the network will reach the cure temperature causing vitrification. At this
point, the reaction rate will be dramatically slowed and the resin will transition
from rubbery to glassy. At vitrification, the torque limit was exceeded because
of the large increase in modulus at vitrification, but the temperature was held
constant and dielectric data acquisition continued. One may note the continued
THERMOSETS
55
Fig. 38. TGA of phenolic resole resin, showing mass losses due to cure (30–300◦ C) and
decomposition (400–650◦ C). Bottom curve is the derivative of the mass versus temperature curve. 10◦ C/min in air. Reprinted with permission from Ref. 134. Copyright American
Chemical Society.
but very slow curing reaction that continues at elevated temperatures even in the
gelled glass region.
6.4. Composition. Discussed in this section are the chemical composition of components prior to mixing and cure, the physical composition or relative
amounts of resin (aka matrix in a fiber-reinforced composite), fillers (particulate
or fiber; see FILLERS), and volatile components such as a high-boiling solvent in a
thermosetting system prior to or after cure. Chemical composition includes reactive components, catalysts, solvents, and fillers. During compounding (eg, mixing
the components and additives to make the final formulation), both the chemical
identity and equivalent weights of reactive components are important. For example, when compounding an epoxy–amine system the choice of epoxy resin and
amine hardener determine such things as the ultimate Tg , and the equivalent
weights allow calculation of the appropriate mix ratio.
TGA, which measures mass flow out of a material, provides a common means
to measure physical composition. Mass loss in thermoset systems includes the loss
of solvents, volatile products associated with cure, and degradation products associated with thermal and thermal-oxidative stability. Composition, for example,
percent solids and filler content, can be measured from evaporation of solvents or
complete decomposition of the resin system, respectively. For several thermosets,
such as phenolic and amino resins, an integral part of cure is the formation of
condensation products, such as water or formaldehyde (see Scheme 3). In such
cases TGA can provide insight into shrinkage resulting from mass loss as well as
completeness of cure. TGA can also measure the amount of water absorbed by a
thermoset as well as sorption/desorption characteristics as shown in Figure 40.
Figure 38 shows the TGA of a phenolic resole resin in air (134), where
30–40% weight loss is attributed to cure between 30 and 300◦ C and complete
56
THERMOSETS
Fig. 39. TGA measurement of variation of glass fiber content in molded part. Reprinted
with permission from Ref. 136. Copyright 1977 Perkin-Elmer.
decomposition occurs between 400 and 650◦ C. Pyrolysis in nitrogen leaves a significant residue or char. Pyrolysis of a carbon fiber phenolic composite in an inert
atmosphere leads to a carbon–carbon composite where the char is the matrix
(135). Analysis of the volatile products of Figure 38 by TGA coupled with mass
spectroscopy showed that, in addition to water, products associated with cure included volatile starting materials such as phenol from 30 to 150◦ C; evolved gasses
associated with the major portion of cure from 100 to 250◦ C, including loss of
the allyl group from the allyl phenyl ether resin; and products associated with
the latter portion of cure and/or the first stages of decomposition (200–300◦ C),
such as methanol. It was speculated that the allyl functions as a thermally labile
protecting group and its loss plays a key role in cure. Decomposition was observed
to occur with two major mass loss steps between 400 and 650◦ C (134).
Figure 39 demonstrates the utility of TGA in measuring the uniformity of
filler content within a part (136). A large nonuniformity of filler content was observed for a molded motor housing. A 9% difference was measured between the
bulk (“interior”) and a narrow inside ridge (“exterior”). The outer surface, which
contained little or no filler, was presumed to be the gel coat (136). Thermally inert
fillers like the glass fiber in this example can easily be measured by TGA. TGA
analysis showed the composition of a silver-filled thermosetting ink, referred to
as polymer thick film, to be 20% solvent, 10% polymer, and 70% silver (129). Measurement of the fiber content in carbon-fiber composites is usually not done by
TGA because of the difficulty of separating the degradation reactions of the matrix from those of the fiber.
6.5. Effect of Water. Absorption of water can affect the cure process
(see Section 4.4) as well as the properties of a cured thermoset. Figure 40 illustrates the ability of TGA to measure moisture sorption and desorption on an epoxy
THERMOSETS
100.4
Mass (%)
Still not at
equilibrium
with 64% RH
after 1350 min
64% RH
Sample equilibrated
with ambient RH
100.2
57
0% RH
100.0
Mass gain = 0.86%
99.8
Mass loss = 0.53%
Start 64% RH purge
99.6
99.4
0
500
1000
1500
Time (min)
2000
2500
Equilibrium with 0% RH after~600 min
Fig. 40. Water desorption/resorption measurements by isothermal TGA at room temperature. Reprinted with permission from Ref. 137. Copyright 2009 John Wiley & Sons, Inc.
0.25 cm diam. HEMI SPHERICAL
PENETRATION PROBE
0.7 cm
COMPOSITE SAMPLE
(1 cm × 0.3 cm)
0.7 cm
Fig. 41. Schematic of TMA sample support fixture to measure heat distortion temperature. Reprinted with permission from Ref. 138. Copyright 1977 Society for the Advancement
of Material and Process Engineering.
thermoset. The measurement is made isothermally at lab ambient temperature.
The measurement starts with a sample that has been equilibrated at lab ambient
(typically 23–25◦ C and 50% RH). A purge of 0% RH nitrogen is started, and the
sample loses mass for about 600 min before stabilizing with a 0.53% mass loss.
After 1250 min, the purge gas is switched to nitrogen with 64% RH and the sample immediately begins to gain mass. A mass gain of 0.86% is observed after 1350
min, but absorption is still occurring at an appreciable rate. Epoxies typically absorb between 1 and 3% moisture. Even this small amount can significantly reduce
Tg and mechanical properties.
Figure 41 illustrates a small three-point bend fixture that will fit in a TMA
(138). This allows for a miniature heat distortion (HDT) test (cf ASTM D648).
The samples were two-ply unidirectional graphite–epoxy composite. Figure 42
presents the results of a temperature scan on dry sample and samples immersed
in distilled water for the times shown. Note that after 14 days (not shown, ∼1%
58
THERMOSETS
TMA PROBE
DISPLACEMENT
323
373
423
409 K (277oF)
DRY
394 K (250oF)
1 DAY SOAK
384 K (232oF)
2
379 K (223oF)
3
371 K (208oF)
6
K
Fig. 42. TMA curves and extrapolated HDT values for two-ply unidirectional graphite
epoxy composite. Reprinted with permission from Ref. 138. Copyright 1977 Society for the
Advancement of Material and Process Engineering.
water sorption) the HDT decreased by a whopping 49◦ C (83 F) from 136 to
87◦ C. This example illustrates the ability of TMA to measure small samples and
shows that small amounts of absorbed water can have large effects on mechanical
properties.
6.6. Photocure. UV curing is used in high performance coating applications, such as topcoats for flooring applications (polymer and wood flooring), coatings for metal and wood furniture, along with plastic and metal container coatings.
UV curing is used extensively in the processing of photoresists in the microelectronics industry. Photodefinable soldermasks used to protect the outside surfaces
of circuit boards and chip substrates make considerable use of acrylate-containing
polymers. Dual cure adhesives can be tacked by UV curing of the edges or fillets
and then thermally curing the adhesive that cannot be reached by the light.
Typical UV-cured coatings contain a free-radical-generating photoinitiator.
Radicals are generated upon exposure to UV light and quickly initiate polymerization. Commercial applications utilize high intensity light sources to effect cure in
seconds or less. UV light sources have been coupled with several analytical techniques such as DSC (known as differential photocalorimetry or DPC), TMA, DMA,
and rheology. These characterization techniques necessarily use lower intensities
in line with the instrument response times. Often a series of long time/low intensity results can reasonably be extrapolated to short time/high intensities typical of commercial processes, giving insight into optimal processing conditions.
These techniques are also useful for comparing different material systems, different light source characteristics such as wavelength and intensity, and parameters
such as time, temperature and oxygen concentration.
Typically an isothermal baseline is established, and then a shutter is opened
to expose the sample to the light source for a fixed time before the shutter is
closed. Variable light sources and filters may be used to control the wavelength
and intensity. Figure 43 illustrates a typical DPC experiment. Run #1 on the
Exothermic heat flow (mW)
THERMOSETS
59
36
Run #2
Shutter
open
32
Shutter
closed
28
Run #1
24
20
0
0.5
1
1.5
2
2.5
3
Fig. 43. Perkin-Elmer DPC accessory with IR water filter to minimize IR component. Run
1 = uncured sample, Run 2 = same sample immediately afterwards, Run 1 minus Run 2
= true heat flow curve. Reprinted with permission from Ref. 139. Copyright 1985 North
American Thermal Analysis Society.
Fig. 44. Photorheology experiment utilizing high frequency/fast data sampling. Copyright TA Instruments.
uncured sample shows a fast exothermic reaction. Run #2 provides the baseline for
calculating the heat of reaction. The absence of any exotherm on the second run
confirms that the reaction is complete in less than 30 s.
Figure 44 shows a photorheology measurement of the storage and loss moduli of a photosensitive material. Very rapid cure is observed when the shutter is
THERMOSETS
Probe position (sample thickness)
60
Expansion
mode
Tg
Temperature
Fig. 45. Schematic of TMA in the expansion mode demonstrating linear expansion below
and above Tg . Reprinted with permission from Ref. 140. Copyright 2009 John Wiley & Sons,
Inc.
opened at 30 s. Gelation occurs almost immediately as observed by the G’ = G"
crossover, a common approximation of gelation (see Section 2.1).
6.7. Thermal Expansion. The coefficient of expansion or CTE of a material is often an important property, for example, when trying to match the expansion coefficient of one material bonded to another material such as copper to
minimize thermal stresses. CTE can also be a valuable tool to characterize
anisotropy. For isotropic materials 𝛼 x = 𝛼 y = 𝛼 z , but for filled or composite materials CTEs are often nonisotropic with unique values in the x, y, and z-directions.
The coefficient of linear expansion (𝛼), abbreviated as CTE or CLTE, is
defined as
𝛼=
dL∕dT
L0
(14)
where L is sample dimension, L0 the initial sample dimension, usually at ambient temperature, and T is temperature in degree centrigrades. 𝛼 has the units of
ppm/◦ C or m/m/◦ C. TMA is a simple yet very useful technique that can measure
coefficient of thermal expansion, Tg , and the dimensional change accompanying
the relaxation of stress stored in a material (140).
As shown in Figure 45, TMA in the expansion mode is capable of measuring
CTE and Tg . The extrapolated intersection of the glassy and rubbery expansion
curves is considered to be a precise measurement of the glass transition temperature. Note that CTE in the rubbery phase is about 3× that in the glassy phase.
As illustrated in Figure 46, a common TMA experiment consists of a first
heat to above Tg to relieve any internal stresses. To relieve stress but avoid additional cure in a partially cured thermoset the first heat should to but not beyond Tg . The sample is then cooled and the CTE and Tg determined in the second
heat.
THERMOSETS
61
Fig. 46. ΔL plotted against temperature for the gate end of a 16-pin dual-in-line package (DIP). Solid line first scan, dashed line second scan. Reprinted with permission from
Ref. 141. Copyright 1987 Society of Plastics Engineers.
In the first run, the expansion rate (the slope of the expansion curve) exceeds
the glassy CTE due to the slow relaxation of molded-in stress. This relaxation,
reflected in the dimensional change ΔL, increases significantly about 30◦ C below
Tg , showing that dimensional stability can be an issue well below the actual glass
transition temperature. Stress relaxation is essentially complete by the time the
temperature reaches Tg .
Many electronic components are small enough to be measured directly by
TMA. Figure 47 shows the results of one such study (142,143), which illustrate differences in both directional dependence and molding compound used.
TMA measurements can be valuable in the selection of molding compounds and
optimization of molding conditions. Note that the Phenolic A sample is the most
isotropic and as a result would be expected to have less residual stress than the
nonisotropic DAP A sample.
TMA in the penetration mode is useful for measuring the softening or penetration temperature Tp as an approximation to Tg . The penetration TMA measurement is less rigorous, can be performed more quickly, and is useful when only
the transition temperature is of interest. A high force is used, and the probe diameter is small to promote penetration into the sample. A similar measure of Tp can
be made on a very small piece of material, for example, an adhesive, by placing it
under the expansion probe.
Figure 48 shows penetration measurements on two thermoset samples: one
that passed an abrasion test and one that failed (144). Sample A passed the abrasion test and exhibits less penetration and a higher Tp . The Tg –conversion relationship suggests that Sample A has achieved a higher degree of cure than Sample
B. Sample B failed the abrasion test and exhibits significantly greater penetration
and lower Tp . TMA penetration measurements can quickly distinguish differences
in degree of cure that can affect functional performance.
62
THERMOSETS
Expansion
Fig. 47. Directional dependence of linear expansion so semiconductor device molded from
four encapsulants. X-axis is the direction of flow. Reprinted with permission from Ref. (142,
143). Copyright 1972 Society of Plastics Engineers.
Probe
Molded polyester
gel coat
Perkin-Elmer TMS-2
r = 0.48 mm
(0.019 in)
Sample A
Sample B
Penetration
10º /m
No load
50g load
Sample A passed abrasion test
Sample B failed abrasion test
50
100
150
200
Temperature(ºC)
Fig. 48. TMA penetration test to detect relative states of cure. The more highly crosslinked sample softens at a higher temperature, undergoes less penetration, and is more
abrasion resistant. Reprinted with permission from Ref. 144. Copyright 1977 Perkin-Elmer.
THERMOSETS
63
Fig. 49. Schematic of matched metal die molding.
7. Thermoset Processes
Thermosets are reactive systems that undergo transformation from liquid to rubbery gel to rigid glassy solid, and these transformations are unavoidably part of
the process. For thermal processes, variables are time, temperature, and pressure.
For light- or radiation-activated processes, variables are time, temperature, and
intensity. Rheology is important because a thermoset system first needs to flow.
For example, an adhesive or coating needs to flow to wet a surface and form an
interface, the matrix polymer needs to flow to consolidate the prepreg layers and
form a laminate, and the resin needs to flow to wet fibers and fill a mold in a pultrusion process. Speed and efficiency of these flow-related steps, which affect both
cost and performance, are dependent on the rheology of the system.
Selection of the manufacturing process is influenced by several variables, including quantity, size, thickness, design, and cost. Thermosets are fabricated by
processes using a wide range of pressure and temperatures, from room temperature cure to high temperatures exceeding 200◦ C for high Tg thermosets. Ambient
cure, especially of larger parts, often takes advantage of the exothermic heat of
reaction to raise the part temperature and accelerate curing. When parts are very
large, overheating can be a problem and the heat of reaction may need to be controlled. Many thermoset systems incorporate fibers to enhance properties; most
others incorporate fillers to modify properties or reduce cost.
Described below are several thermoset processes. This is not meant to be
comprehensive, but rather they illustrate the variety of processing options that
exist for thermosets (see also COMPOSITES, FABRICATION, for descriptions of additional thermoset processes).
7.1. Matched Metal Die Molding. A common, high volume, cost-effective
process for making a wide variety of parts is matched metal die molding. The process entails using two precision machined matched metal molds (an upper and
lower mold) as seen in Figure 49 and placed in a compression molding frame capable of applying heat and pressure.
Compression molding is a high volume, high pressure process suitable for
molding complex, high strength, and fairly large articles that need excellent surface finish. The compounded resin–fiber charge consists of either BMC or SMC.
64
THERMOSETS
The BMC charges are produced in a separate compounding operation where
resins, initiators, catalysts, chopped glass fibers, mold release agents, and other
additives are mixed in a high speed, high shear mixer such as a sigma blade mixer.
In the case of BMC, the result is a dough-like material that can be cut into a
charge, weighed to provide enough material to fully fill the molds, and then placed
in the compression molds.
For SMC, a continuous process is used to apply the resin to a carrier, add the
chopped fibers on top of the resin, add additional resin to the top surface, finally
using a compaction zone to mix and wet-out the fibers. The resultant SMC may
be cut to the desired size and weight. Heat and pressure are applied to allow the
BMC or SMC to soften, flow, and completely fill the mold cavity. The temperature
is set to get rapid curing to form a highly cross-linked network in a short process
time. For polyesters, temperatures are typically 100–160◦ C and pressures 1–13
MPa (150–2000 psi). Depending on thickness, size, and shape of the article, curing
cycles range from <1 to about 5 min.
7.2. Manual Processes. A good example of manual processing is fiber
glass-reinforced plastic (FRP) applications utilizing unsaturated polyester resins
(see POLYESTERS, UNSATURATED). A random-chopped strand fiber glass mat is
precut to fit the open-mold contour and impregnated by hand or spraying in
successive plies with catalyzed polyester resin. The application of glass mat
and resin at ambient temperatures is now universally employed and called the
hand-lay-up or contact-molding process. The quality of the product depends on
the skill of the operator in removing air and voids. Fabricated products include
chemical equipment, sanitary fixtures, architectural structures, and fiber glass
boats. In the spray-up technique, used for bathtub and shower applications, for
example, the catalyzed resin and chopped glass fiber are laid down simultaneously on the mold surface with specialized spray equipment. The appearance of
FRP products is greatly enhanced by incorporating a polyester gel coat, which
may contain pigments and is usually sprayed directly onto the treated mold surface and allowed to cure. The thixotropic characteristics of gel coats need to be
controlled, and attention must be paid to gel-time drift caused by absorption and
deactivation of the accelerator by the pigments and by thixotropic agents such as
fumed silica. Resin and glass-fiber reinforcement is applied directly over the gel
coat by hand-lay-up or spray-up techniques to produce a product, such as boats
and bathtubs, in which the surface of the composite has a glossy appearance free
of glass fibers.
7.3. Prepreg Processing. The composites industry makes extensive use
of thermoset prepregs. As discussed previously, thermosets offer the unique ability to tailor the curing chemistry to allow partial reaction or B-staging. Typical
prepreg processes impregnate a reinforcement such as carbon fiber or woven glass
cloth with a carefully formulated resin consisting of base resins, curing agents,
accelerators, tougheners, and other property enhancing additives. The process begins with the compounding process where the ingredients in the liquid resin mixture are properly blended in the right proportions. There are two main types of
prepreg process: Solvent treating and Solventless (hot melt)
In the case of the solvent-based prepreg process, the resin mixture is
formulated in a suitable solvent that permits easy ingredient dissolution and
THERMOSETS
65
Fig. 50. Schematic of prepreg process. Reprinted with permission from Hexcel Corporation. Copyright 2013 Hexcel Corporation.
controlled evaporation in the treater tower. A typical solvent treater tower is
shown in Figure 50.
Solvent-based treaters are typically used to make woven fabric (most commonly woven glass) prepregs. The woven fabric enters the resin dip pan in the
left side of Figure 50 (in red just under the blue rollers). By controlling the nip
rollers, the amount of resin applied to the moving glass cloth can be precisely controlled. In the first two zones in the treater, the solvent is evaporated. In the
fourth and fifth zones, heat is adjusted (or at constant zone temperatures the web
speed is changed) to control the amount of B-stage advancement. Thermosets have
the unique ability to be partially cured. The heating profile can be tailored to react the resin to a specific degree of conversion, with the B-stage conversion below
the gel point. As the moving web exits the treater tower, it is cooled to below Tg .
Prepregs are a good example of an ungelled glass (vitrification occurs as the moving web is cooled). During subsequent processing, for example, using a lamination
process, the prepreg softens, flows, additional chemical reaction causes gelation
along the path to full cure of the prepreg into the final composite.
A hot melt process is used to make nonsolvent-based prepregs. The hot melt
process consists of two steps: filming the hot melt resin and incorporating and
consolidating fibers into the hot melt films.
A schematic of the film process is shown in Figure 51, where the release
film is unwound and moves from left to right where the solventless resin matrix
is applied to the release film using a coating head. The resin content is carefully
controlled by the web speed and the resin flow at the coating head. Another release
film is applied to the top of the moving web to protect the resin. In the hot melt
process, the resin is heated to above Tg to facilitate good flow and metering onto
the carrier.
66
THERMOSETS
Release film
Step 1
Matrix
Coating head
Matrix film
Release film
Fig. 51. Schematic of the hot melt process for nonsolvated resins. Reprinted with permission from Hexcel Corporation. Copyright 2013 Hexcel Corporation.
Step 2
Film recovery
Release film
Matrix film
Reinforcement
Consolidation
Heating
Prepreg
Release film
Fig. 52. Schematic of the second stage of the hot melt prepreg process. Reprinted with
permission from Hexcel Corporation. Copyright 2013 Hexcel Corporation.
In the next step, the final prepreg is fabricated by using the film from step
1 and adding various types of fiber tows. This process is most commonly used to
produce unidirectional tape prepreg by feeding a multitude of fibers (eg, glass,
carbon, Kevlar) into the prepreg line. A schematic is shown in Figure 52.
The matrix film (blue film in this figure) was produced in step 1. The fibers
are aligned and fed into the nip between a release film and the matrix film. The
resin is heated to lower the viscosity and facilitate the wet-out of the fibers. The
heated resin with the fibers then pass through a consolidation zone where rollers
apply pressure to fully impregnate and wet-out the fibers. A release film is applied
on the top of the moving web to protect the prepreg surface. The hot melt process
produces a flexible prepreg with controlled tack and drape that will allow complex
composite shapes to be molded in subsequent autoclave processing.
7.4. Vacuum Bag and Autoclave. Articles made by hand layup or
spray-up processes are allowed to cure without external pressure. While this is
sufficient for many applications, maximum consolidation is not achieved and some
porosity remains. The application of even moderate pressure can frequently overcome these drawbacks and results in better mechanical properties, higher quality parts, and lower weight than achieved by contact molding. One way to apply
moderate pressure is to enclose the wet article in the mold inside a flexible membrane or bag and apply a vacuum. Atmospheric pressure presses the bag uniformly
THERMOSETS
67
Vacuum bagging
Bagging film
Vacuum hose
Bleeder
Breather
Release film
Peel ply
Vacuum valve
Pressure
sensitive tape
Sealant
Laminate
Tool
Release film
Flash tape
Dam
Fig. 53. Schematic of vacuum bagging process.
RTM
4
1
3
2
5
6
A
B
7
8
Fig. 54. Schematic of RTM process.
against the wet article, generating pressures between 70 and 100 kPa (10–14 psi).
Vacuum inside the bag also tends to degas the article, reducing its porosity. In
Figure 53, a schematic of a typical cross-section layup is shown.
Still higher pressures can be obtained by placing the vacuum assembly in
an autoclave at pressures in the range of 700–1400 kPa (∼100–200 psi). The bag
must be well sealed to prevent infiltration of high pressure air, steam, or water
into the molded part. An initial vacuum may be used in the autoclave process to
aid in gas and void removal, but the bag is usually vented to atmospheric pressure
during the high temperature, high pressure portion of the cure cycle so as not to
induce additional void growth from reduced pressure.
7.5. Resin Transfer Molding. Resin transfer molding (RTM) is a low
cost process used to mass-produce small and midsize parts of various shapes with
two finished sides. A schematic of the RTM process is shown in Figure 54. The
process requires two matched dies (1 and 2) that are clamped together. The fiber
68
THERMOSETS
preform (5) is first placed in the open mold and then the mold is closed and clamped
(3) to maintain dimensional tolerances during processing. To achieve very rapid
curing, a two part resin system is used which allows much faster gel and cure
times than open molding processes. The liquid resin A and catalyst B (7 and 8)
are mixed in-line using a mixing head (4) and subsequently injected at ambient
temperature into the center of the cavity at low pressure (6). The radial flow outward facilitates the fiber preform wetting and good venting. RTM leverages many
of the advantages of using thermosets. There is a large resin “tool kit” available
to achieve the desired final properties. The liquid resin matrix has low viscosity
during infusion, multiple cure chemistries are available to tailor the catalyst and
accelerator package for very fast curing, and there is flexibility to use a wide variety of fiber reinforcements in multiple part geometries.
Gel times for polyester cure are in the 3–5 min range, and molding cycles can
be as short as 20 min (see POLYESTERS, UNSATURATED). The size of articles that
can be produced by RTM is limited by the costs of constructing molds and clamping equipment. Products can be pigmented, gel-coated, and finished. The resin
systems may be modified by various types and amounts of fillers. RTM articles
can include encapsulated core materials, rib formers, and structural inserts.
7.6. Vacuum-Assisted Resin-Transfer Molding. VARTM or vacuum
bag resin infusion is a single-sided molding process that offers numerous cost advantages over traditional RTM because of lower tooling costs, room temperature
processing, and scalability to large structures. Because it is a single-sided process,
dimensional tolerance is less precise than with RTM. VARTM is a resin infusion
process in which dry reinforcement plies are placed onto a metal tool, usually aluminum, and coated with a release agent. The preform is covered with a release
film and multiple layers of a distribution medium. A vacuum bag is placed on top
of this layered construction, and a resin inlet and an outlet are positioned at opposite ends of the preform inside the vacuum bag. Vacuum is then drawn providing
a driving force that promotes resin infusion first through the distribution medium
and then through the fiber preform. When the laminate is completely saturated
with resin, the inlet and the outlet are shut and the panel is allowed to cure either
at room temperature or at higher temperatures in an oven. An additional advantage of the single-sided process is the potential to use photocuring to aid the cure
process. The VARTM process is poised for exponential growth as this manufacturing technology is inserted into major defense and commercial programs.
7.7. Reactive Injection Molding. Reactive injection molding (RIM), also
known as liquid injection molding, is a process in which a mixture of two or more
reactive materials is injected into a mold where the materials react to form a
thermoset structure. Advantages include low cost, low pressure requirements,
and flexibility in mold configuration. Conventional systems, such as isocyanate
with polyol to produce a polyurethane, release little or no volatiles. The generation of substantial volatiles in the mold is obviously undesirable and has presented a significant obstacle to phenolic-based systems, although some have been
reported (see PHENOLIC RESINS). Rigid, cross-linked structural polyurethane foam
systems produced by the RIM process contain a sandwich structure of solid skin
and a lower density microcellular core in a single operation (see POLYURETHANES).
Polyurethane structural foams utilize the smooth, hard skins as a source of impact
strength, stability to heat distortion, and combustion resistance. The development
THERMOSETS
69
Fig. 55. Schematic of pultrusion process.
Fig. 56. Schematic of filament winding process.
of internal mold release–based foam systems has further improved the productivity of the RIM process by reducing cleaning time and application of mold release
between cycles. The metered component streams are mixed and injected into a
closed mold. The chemical cure reaction begins immediately. The exothermic heat
evolved vaporizes a low boiling solvent contained in the polyol component, causing the liquid mass to foam. The mixture becomes progressively more viscous,
gels, and at the completion of the reaction is transformed into a rigid thermoset
network. Cycle times of 1 min or less can be achieved.
7.8. Pultrusion and Filament Winding. In pultrusion, a combination of
liquid resin and continuous fiber is pulled continuously through a heated die of the
required shape for continuous profiles, such as structural channel, angles, rods,
and sheets. Figure 55 is a schematic of a typical pultrusion process. The plastic
sides and top of stepladders are common examples of parts made using a pultrusion process. Common resins utilized are polyesters, epoxies, and polyurethanes.
To keep the cost low glass fibers are used. A typical process might be 3–5 min in
duration where resin temperatures ramp rapidly from ambient to temperatures
in excess of 150◦ C, hold for a short time, and ramp back down (see Figs. 16–18 and
accompanying text for an example of the curing of a two-component polyurethane
used in a pultrusion process).
The filament winding process is shown in Figure 56 and is similar to the pultrusion process. Here multiple fiber tows are impregnated with a liquid thermoset
resin to form a filament that is applied to a rotating mandrel.
Glass fiber tows are most commonly used with carbon and aramid fibers used
for high performance applications. Polyester resins, epoxy, or polyurethanes are
70
THERMOSETS
common resins systems. By controlling the fiber impregnation process, high fiber
volume fraction can be obtained in the final part leading to very high stiffness and
good mechanical properties. Filament winding is optimal for producing various
sizes of circular or oval cross section parts such as tanks, pipes, and containers.
7.9. Adhesive Processes. Most commercially available adhesive systems consist of either thermoplastic hot melt or reactive thermosetting resins.
Typical thermoset adhesive formulations contain base resin monomers and
oligomers, curing agents (hardeners), catalysts, adhesion promoters, and various
fillers. Adhesive properties depend on the complete mixing or adequate dispersion of the various ingredients. It is critical that solids be well dispersed. Typical
solid components of adhesives are fillers, but can include catalysts or hardeners.
Three-roll mills or multishaft mixers with high speed dispersing blades are used
to disperse solids into low viscosity components. The high shearing action helps
in breaking up solid agglomerates and ensures that the fillers are completely wet
out and uniformly distributed in the resin. Once the fillers have been completely
dispersed, additional low viscosity components can be incorporated into the formulation using double planetary or multishaft mixers to produce a homogeneous
mixture. The high shearing action during filler wet-out and the mixing process
introduces air into the formulation. The final mixing operation is typically done
under vacuum to completely outgas the resin mixture. Most commercially available double planetary and multishaft mixers have the capability to apply vacuum
and control temperature during the mixing operation.
For reactive chemistries, the resultant mixtures are placed in containers or
syringes and frozen to well below Tg0 to minimize cure reactions from occurring
during storage. These types of adhesives are termed premixed and frozen. Depending on the catalyst and chemistry, the work life (the time after thawing the
adhesive will retain the required processing parameters) can range from hours to
days. Some adhesives are packaged in two parts. Typical home-use epoxies come
packaged in a dual syringe with parts A and B. The hardener/catalyst is in one
part, and the resin components are in the other part. This allows a long storage
life at room temperature. Once dispensed, the two-part adhesive must be thoroughly mixed and used within the required time. For most industrial adhesives,
a Five-Minute Epoxy type of formulation is not acceptable, since the work life is
too short. At a minimum, the work life should be at least 8 h, allowing use over
a single shift operation. Most premixed and frozen adhesives have about a 24-h
work life.
When a premixed and frozen adhesive is used in production, care must be
exercised during thawing to minimize voiding and separation of the adhesive from
the wall of the syringe. Additionally, the environment should be controlled to minimize condensation from being incorporated into the adhesive. Rapid thawing increases the likelihood of voiding. Once the adhesive containing syringe is properly
thawed, it is placed on a dispensing device. Typical dispensers include needle dispensers that deliver a prescribed amount of adhesive onto the desired surface. The
adhesive is then dispensed into line, star, glob, or other patterns depending on the
application.
After the adhesive is dispensed, the material is cured. Most adhesives have
thermal curing reactions. The most common application is to cure the adhesive in an environmentally controlled oven (typically under nitrogen purge to
THERMOSETS
71
minimize oxidation). Adhesives with very rapid curing are termed snap cure
(about 1–2 min at high temperatures) or spot cure (approximately 10 s at high
temperatures). Specifically in electronic applications, the required high manufacturing throughput requires adhesives with very fast curing speeds. Some adhesives are also B-staged to remove tackiness and give long shelf life, but under
additional heating can soften, flow, and cure into a high performance network.
Very fast curing reactions can proceed after exposure to UV radiation or low
wavelength visible light (blue light). Typical applications of UV cure include wood
coatings (particleboard and simulated wood finishes), flexible coating for vinyl
flooring, and coatings for metal surfaces (47). UV exposure is also extensively used
in photoresist technology. However, adhesive applications are often limited by the
inability of light to reach the interior of the adhesive joint. Dual-cure adhesives
overcome this problem, allowing the exposed adhesive at the joint edges to be
cured by UV or blue light, rapidly tacking the parts to be bonded. Subsequently,
the bond interior can be thermally cured without the need for expensive fixturing.
BIBLIOGRAPHY
1. IUPAC. Compendium of Chemical Terminology, 2nd ed. (the “Gold Book”). Compiled
by A. D. McNaught and A. Wilkinson, Blackwell Scientific Publications, Oxford, UK
(1997). XML on-line corrected version: http://goldbook.iupac.org (2006-) created by
M. Nic, J. Jirat, B. Kosata; updates compiled by A. Jenkins. ISBN 0-9678550-9-8.
doi:10.1351/goldbook.
2. C. A. May, Epoxy Resins: Chemistry and Technology, 2nd ed., Marcel Dekker, New
York, 1987.
3. R. B. Prime, in E. A. Turi, ed., Thermal Characterization of Polymeric Materials, Academic Press, San Diego, CA, 1981, Chapt. 5; 1997, Chapt. 6.
4. J. P. Pascault, H. Sautereau, J. Verdu, and R. J. J. Williams, Thermosetting Polymers,
Marcel Dekker, New York, 2001.
5. B. Van Mele, H. Rahier, G. Van Assche, and S. Swier, in M. Reading and D. J.
Hourston, eds., Modulated Temperature Differential Scanning Calorimetry: Theoretical and Practical Applications in Polymer Characterisation, Springer, Berlin, 2010,
Chapt. 2.
6. J. P. Pascault and R. J. J. Wiliams, Epoxy Polymers: New Materials and Innovations,
Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, Germany, 2010.
7. Q. Guo, ed., Thermosets: Structures, Properties and Applications, Woodland Publishing
Limited, Cambridge, UK, 2012.
8. R. J. Seyler, ed., Assignment of the Glass Transition, STP 1249, ASTM, Philadelphia,
Pa., 1994.
9. H. H. Winter, Polym. Eng. Sci. 27, 1698 (1987).
10. H. H. Winter, M. Mours, M. Baumgartel, and P. R. Soskey, in A. A. Collyer and D. W.
Clegg, eds., Rheological Measurement, Chapman and Hall, London, 1998, Chapt 2, pp.
47–95
11. P. J. Flory, J. Am. Chem. Soc. 63, 3083 (1941).
12. P. J. Flory, Principles of Polymer Chemistry, Cornell University Press, Ithaca, N.Y.,
1953.
13. D. R. Miller, E. M. Valles, and C. W. Macosko, Polym. Eng. Sci. 1 9, 272 (1979).
14. D. R. Miller and C. W. Macosko, Macromolecules 9, 206 (1976).
15. M. Feve, Makromol. Chem., Macromol. Symp. 30, 95 (1989).
16. J. K. Gillham, Polym. Eng. Sci. 26, 1429 (1986).
72
THERMOSETS
17. J. B. Enns and J. K. Gillham, J. Appl. Polym. Sci. 28, 2567 (1983).
18. E. R. Soulé, J. Borrajo, and R. J. J. Williams, Polym. Eng. Sci. 46, 1641 (2006).
19. K. Horie, H. Hiura, M. Souvada, I. Mita, H. Kambe, and J. Polym. Sci. Polym. Chem.
Ed. 8, 1357 (1970).
20. G. Wisanrakkit and J. K. Gillham, 41, 2885 (1990).
21. T. Anagnostou and E. Jaul, J. Coat. Technol. 53, 35 (1981).
22. P. I. Kordomenos, A. H. Dervan, and J. Kresta, J. Coat. Technol. 54(687), 43 (1982).
23. P. L. Tabaddor, C. J. Aloisio, H. E. Bair, C. H. Plagianis, and C. R. Taylor, J. Therm.
Anal. Calorim. 59, 559 (2000).
24. J. Mijovic and S. Andjelic, Polymer 36, 3783 (1995).
25. C. Billaud, M. Vandeuren, R. Legras, and V. Carlier, Appl. Spectrosc. 56, 1413 (2002).
26. J. H. Fu and J. R. Schlup, J. Appl. Polym. Sci. 49, 219 (1993).
27. L. Xu, J. H. Fu, and J. R. Schlup, Ind. Eng. Chem. Res. 35, 963 (1996).
28. S. Paz-Abuin, A. Lopez-Quintela, M. Varela, M. P. Pellin, and P. Prendes, Polymer 38,
3117 (1997).
29. N. A. St. John and G. A. George, Polymer 33, 2679 (1992).
30. K. E. Chike, M. L. Myrick, R. E. Lyon, and S. M. Angel, Appl. Spectrosc. 47, 1631 (1993).
31. J. Mijovic and S. Andjelic, Macromolecules 28, 2787 (1995).
32. J. Mijovic, S. Andjelic, C. W. Yee, F. Belluci, and L. Nicolais, Macromolecules 28, 2797
(1995).
33. G. Lachenal, A. Pierre, and N. Poisson, Micron 27, 329 (1996).
34. L. Xu and J. R. Schlup, J. Appl. Polym. Sci. 67, 895 (1998).
35. E. Duemichen, M. Javdanitehran, M. Erdmann, V. Trappe, H. Sturm, U. Braun, and
G. Ziegmann, Thermochim. Acta 616, 49 (2015).
36. S. Swier, G. Van Assche, and B. Van Mele, J. Appl. Polym. Sci. 91, 2798 (2004).
37. C. J. de Bakker, G. A. George, N. A. St. John, and P. M. Fredericks, Spectrochim. Acta,
Part A 49, 739 (1993).
38. J. P. Pascault and R. J. J. Williams, J. Polym. Sci., Part B: Polym. Phys. 28, 85 (1990).
39. A. Hale, C. W. Macosko, and H. E. Bair, Macromolecules 24, 2610 (1991).
40. R. A. Venditti and J. K. Gillham, J. Appl. Polym. Sci. 64, 3 (1997).
41. L. E. Nielson, J. Macromol. Sci. C3, 69 (1969).
42. A. T. DiBenedetto, J. Polym. Sci., Part B: Polym. Phys. 25, 1949 (1987).
43. B. Fuller, J. Gotro, and G. Martin, in C. Craver and T. Provder, eds., Polymer Characterization: Physical Property, Spectroscopic, and Chromatographic Methods, Advances
in Chemistry Series 227, American Chemical Society, Washington, D.C., 1990, pp. 215.
44. R. L. Quirina and M. R. Kessler, in A. N. Netravali and C. M. Pastore, eds. Sustainable
Composites: Fibers, Resins and Applications, Lancaster, PA, DEStech Publications,
2014, Chapt. 3, pp. 51–86.
45. J. Bergman and M. R. Kessler, in J. H. Clark and F. E. I. Deswarte, eds., Introduction
to Chemicals from Biomass, 2nd ed., John Wiley & Sons, Inc, Chichester, West Sussex,
UK, 2015, pp. 157–204.
46. Y. Li and M. R. Kessler, in V. Kumar and M. R. Kessler, eds., Liquid Crystalline Polymers: Structure and Chemistry, Springer, Berlin 2016, pp. 1–18.
47. S. Peter Peppas, UV Curing: Science and Technology, Technology Marketing Corporation, Norwalk, C.T., 1980.
48. M. R. Kessler, in L. Nicolais, ed., in Wiley Encyclopedia of Composites, John Wiley &
Sons, Inc., Hoboken, N.J., 2012, pp. 658–672.
49. D. A. Shimp, in I. Hamerton, ed., Chemistry and Technology of Cyanate Ester Resins,
282, Chapman and Hall, Glasgow, U.K., 1994, Chapt. 10.
50. Dow Chemical Company, Product Bulletin, XU-71787.02 and XU-71787.07, Developmental Polycyanate Resins for Advanced Composites and Adhesives (1992).
51. P. C. Yang, D. Pickleman, and E. Woo, 35th Int. SAMPE Symp. 1131 (1990).
THERMOSETS
73
52. D. Shimp and W. Craig, 34th Int. SAMPE Symp. 222 (1989).
53. D. Shimp and S. Ising, 34th Int. SAMPE Symp. 1045, (1989).
54. C. A. May, ed., Chemorheology of Thermosetting Polymers, ACS Symposium Series
227, American Chemical Society, Washington, D.C., 1983.
55. S. Sourour and M. Kamal, Thermochim. Acta 14, 41 (1976).
56. C. S. Chern and G. W. Poehlein, Polym. Eng. Sci. 27, 788 (1987).
57. S. Matsuoka, X. Quan, H. E. Bair, and D. J. Doyle, Macromolecules 22, 4093 (1989).
58. G. Wisanrakkit and J. K. Gillham, J. Appl. Polym. Sci. 42, 2453 (1991)
59. R. E. Lyon, J. Test. Eval. 42(6), 1387 (2014).
60. C. D. Doyle, Anal. Chem. 33, 77 (1961).
61. S. J. Swarin and A. M. Wims, Anal. Calorim. 4, 155 (1977).
62. J. E. K. Schawe, Thermochim. Acta 388, 299 (2002).
63. R. B. Prime, C. Michalski, and C. M. Neag, Thermochim. Acta 429(2), 213 (2005).
64. C. M. Neag and R. B. Prime, J. Coat. Technol. 63(797), 37 (1991).
65. P. E. Willard, J. Macromol. Sci., Chem. A8(1), 33 (1974).
66. P. E. Willard, in J. Chiu, ed., Polymer Characterization by Thermal Methods of Analysis, Vol. 33, Marcel Dekker, NY, 1974, pp. 33–41.
67. G. Widmann, in I. Buzas, ed., Thermal Analysis, Vol. 3, Heydon, London, 1975, p. 359.
68. R. L. Miller and M. A. Oebser, Thermochim. Acta 36, 121 (1980).
69. A. Dutta and M. E. Ryan, J. Appl. Polym. Sci. 24, 635 (1979).
70. J. Mijovic and H. T. Wang, J. Appl. Polym. Sci. 37, 2661 (1989).
71. M.-F. Grenier-Loustalot and P. Grenier, Polymer 33, 187 (1992).
72. A. Garton, W. T. K. Stephenson, and S. P. Wang, J. Polym. Sci., Part A: Polym. Chem.
26, 1377 (1988).
73. R. B. Prime, Proc. of 13th North Am. Therm. Anal. Soc. Conf. 362 (1984).
74. R. B. Prime, J. Therm. Anal. 31, 1091 (1986).
75. J. M. Burns, R. B. Prime, E. M. Barrall, and M. E. Oxsen, in K. L. Mittal, ed., Polymers
in Information Storage Technology, Plenum Publishing Corp., New York, 1989, pp.
237–256.
76. H. E. Bair (1992), Proc. North Am. Therm. Anal. Conf. 21, 60 (1992).
77. J. Brand, Tech. Pap. IPC-TP (Inst. Printed Circuits) ICP-TP-152 (1977).
78. T. E. Noll, Tech. Pap. IPC-TP (Inst. Printed Circuits) ICP-TP-140 (1977).
79. A. Garton and J. H. Daly, J. Polym. Sci., Part A Chem. Ed. 23, 1031 (1985).
80. A. Garton, J. H. Daly, W. T. K. Stephenson, and D. M. Wiles, J. Polym. Sci., Part A:
Polym. Chem. 24, 541 (1986).
81. A. Garton and W. T. K. Stephenson, J. Polym. Sci., Part A: Polym. Chem. 26, 1031
(1988).
82. W. Sun Lee and J. Yu, Diamond Relat. Mater. 14, 1647 (2005).
83. D. Lu and C.P. Wong, Materials For Advanced Packaging, Springer, Berlin, 2009,
Chapt. 11, pp. 365
84. K. Sato, H Horibe, T. Shirai, Y. Hotta, H. Nakano, H. Nagai, K Nitsuishi, and K. Watan,
J. Mater. Chem. 20, 2749 (2010).
85. E. P. Plueddeman, Silane Coupling Agents, 2nd ed., Springer, Berlin, 1991.
86. C. Pradere and C. Sauder, Carbon 46(14), 1874 2008.
87. D. M. Riggs, R. J. Shuford, and R. W. Lewis, in G. Lubin, ed., Handbook of Composites,
Van Nostrand Reinhold Co., New York, 1982, pp. 196–271.
88. M. A. Ratner and D. Ratner, Nanotechnology: A Gentle Introduction to the Next Big
Idea, Prentice-Hall, Englewood Cliffs, N.J., 2003.
89. R. A. Vaia, T. Benson Tolle, G. F. Schmitt, D. Imeson, and R. J. Jones, SAMPE J. 37(6),
24 (2001).
90. E. R. Zubarev, M. U. Pralle, E. D. Sone, and S. I. Stupp, J. Am. Chem. Soc. 123, 4105
(2001).
74
91.
92.
93.
94.
95.
96.
97.
98.
99.
100.
101.
102.
103.
104.
105.
106.
107.
108.
109.
110.
111.
112.
113.
114.
115.
116.
117.
118.
119.
120.
121.
122.
THERMOSETS
J. D. Hartgerink, E. Beniash, and S. I. Stupp, Science 294, 1684 (2001).
A. A. Azeez, K. Y. Rhee, S. J. Park, D. Hui, Composites, Part B 45, 308 (2013).
J. Njuguna, K. Pielichowski, and J. R. Alcock, Adv. Eng. Mater. 9, 853 (2007).
T. J. Pinnavaia and G. W. Beall, eds., Polymer-Clay Nanocomposites, John Wiley &
Sons, Inc., New York, 2000.
B. Maruyama and K. Alam, SAMPE J. 38(3), 59 (2002).
H. Miyagawa, M. J. Rich, and L. T. Drzal, Proc. North Am. Therm. Anal. Conf. 30, 144
(2002).
B. B. Johnson, A. J. Kinloch, R. D. Mohammed, A. C. Taylor, and S. Sprenger, Polymer
48, 530 (2007).
G. Z. Li, L. Wang, H. Ni, and C. U. Pittman Jr., Inorg. Organomet. Polym. 11(3), 123
(2002).
C. U. Pittman Jr., G. Z. Li, and H. Ni, Macromol. Symp. 196, 301 (2003).
K. P. Unnikrishnan and E. T. Thachil, Designed Monom. Polym. 9(2), 129 (2006).
S. Thomas, C. Sinturel, and R. Thomas, Micro- and Nanostructured Epoxy/Rubber
Blends, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, Germany, 2014.
T. Inoue, Prog. Polym. Sci. 20, 119 (1995).
R. J. J. Williams, B. A. Rosenberg, and J. P. Pascault, Adv. Polym. Sci. 95, 128 (1997).
J. H. Hodgkin, G. P. Simon, and R. J. Varley, Polym. Adv. Technol. 9, 3 (1998).
A. J. Kinloch, Adv. Polym. Sci. 72, 45 (1985).
E. H. Rowe, A. R. Siebert, and R. S. Drake, Mod. Plast., 110 (1979).
J. N. Sultan and F. J. McGarry, Polym. Eng. Sci. 13, 29 (1973).
L. T. Manzione, J. K. Gillham, and C. A. McPherson, J. Appl. Polym. Sci. 26, 889 (1981).
L. T. Manzione, J. K. Gillham, and C. A. McPherson, J. Appl. Polym. Sci. 26, 907 (1981).
C. K. Riew, E. H. Rowe, and A. R. Siebert, in R. D. Deanin and A. M. Crugnola, eds.,
Advances in Chemistry, Series No. 154, American Chemical Society, Washington D.C.,
1976, p. 326.
J. K. Gillham, C. A. Glandt, and C. A. McPherson, in S. S. Labana, ed., Chemistry and
Properties of Crosslinked Polymers, Academic Press, New York, 1977, pp. 491–520.
C. B. Bucknall, Toughened Plastics, Applied Sciences Publishers, London, 1977, pp.
36–37.
B. S. Kim, T. Chiba, and T. Inoue, Polymer 34, 2809 (1993).
L. Ruiz-Perez, G. J. Royston, J. P. A. Fairclough, and A. J. Ryan, Polymer 49, 4475
(2008).
M. Martin-Gallego, R. Verdejo, A. Gestos, M. A. Lopez-Manchado, and Q. Guo, Composites, Part A 71, 136 (2015).
Private communication, Gerald Smyth, ICI Corporation, 2007.
A. Riga and L. Judovits, eds., Material Characterization by Dynamic and Modulated
Thermal Analytical Techniques, ASTM STP 1402, American Society for Testing Materials, West Conshohocken, Pa., 2001.
J. D. Menczel, L. Judovits, R. B. Prime, H. E. Bair, M. Reading, and S. Swier, in J. D.
Menczel and R. B. Prime, eds., Thermal Analysis of Polymers, John Wiley & Sons, Inc.,
Hoboken, N.J., 2009, Chapt. 2.
M. Reading and D. J. Hourston, eds., Modulated-Temperature Differential Scanning
Calorimetry, Springer, Berlin, 2006.
B. Van Mele, H. Rahier, G. Van Assche, and S. Swier, in M. Reading and D. J.
Hourston, eds., Modulated-Temperature Differential Scanning Calorimetry: Theoretical and Practical Applications in Polymer Characterisation, Hot Topics in Thermal
Analysis and Calorimetry Series, Springer, Vol. 6, 2006, Chapt. 2.
W. V. Breitigam, R. S. Bauer, and C. A. May, Polymer 34, 767 (1993).
D. Djokic, A. Johnson, A. Rogers, P. Lee-Sullivan, and N. Mrad, Composites, Part A 33,
277 (2002).
THERMOSETS
75
123. U.S. Pat. 4, 978,712 (1990), H. E. Bair and S. Matsuoka AT&T Bell Laboratories (Murray Hill, NJ).
124. S. Swier, G. Van Assche, and B. Van Mele, J. Appl. Polym. Sci. 91, 2814 (2004).
125. C. Macosko, Rheology: Principles, Measurements, and Applications, VCH Publishers,
New York, 1994.
126. C. Block, B. Van Mele, P. Van Puyvelde, G. Van Assche, React. Funct. Polym. 73, 332
(2013).
127. R. E. Wetton, in J. V. Dawkins, ed., Developments in Polymer Characterization, Elsevier, London, 1986, pp. 179–221.
128. S. H. Dillman, J. C. Seferis, and R. B. Prime, Proc. North Am. Therm. Anal. Conf. 16,
429 (1987).
129. R. B. Prime, Polym. Eng. Sci. 32, 1286 (1992).
130. W. J. Sichina and B. Matsumori, Proc. North Am. Therm. Anal. Conf. 20, 41 (1991).
131. S. D. Senturia and N. F. Sheppard, Epoxy Resins and Composites IV, Advances in
Polymer Science 80, Springer-Verlag, Berlin, 1986, pp. 1–47.
132. N. F. Sheppard Jr., M. C. W. Coln, and S. D. Senturia, 29th Natl. SAMPE Symp. Exhib.
1243 (1984).
133. J. Gotro and M. Yandrasits, Polym. Eng. Sci. 29, 278 (1989).
134. R. B. Prime and B. Shushan, Anal. Chem. 61, 1195 (1989).
135. J. Delmonte, Technology of Carbon and Graphite Fiber Composites, Van Nostrand
Reinhold Co., Inc., New York, 1981.
136. B. Cassel, Perkin-Elmer Therm. Anal. Appl. Study 19 (1977); see also Ref. 1.
137. H. E. Bair, in J. D. Menczel and R. B. Prime, eds., Thermal Analysis of Polymers, John
Wiley & Sons, Hoboken, N.J., 2009, Chapt. 3, pp. 269–270.
138. G. F. Sykes, H. D. Burkes, and J. B. Nelson, Proc. 22nd Natl. SAMPE Symp. Exhib.,
1977, pp. 350–364.
139. H. E. Bair and L. L. Blyer, Proc. North Am. Therm. Anal. Conf. 14, 392 (1985).
140. H. E. Bair, A. E. Akinay, J. D. Menczel, R. B. Prime, and M. Jaffe, in J. D. Menczel and
R. B. Prime, eds., Thermal Analysis of Polymers, John Wiley & Sons, Hoboken, N.J.,
2009, Chapt. 4, pp. 319–385.
141. H. E. Bair, D. J. Boyle, J. T. Ryan, C. R. Taylor, and S. C. Tighe, Soc. Plast. Eng. (Proc.
Annu. Techn. Conf.) 33, 362 (1987).
142. K. Ritchie, W. L. Hunter, and C. Malkiewicz, in H. G. Wiedemann, ed., Thermal Analysis: Volume 3: Organic and Macromolecular Chemistry, Ceramics, Earth Sciences,
Birkhäuser, Basel, Switzerland, 1972, Vol. 3, pp. 179–185.
143. K. Ritchie, W. L. Hunter, and C. Malkiewicz, Soc. Plast. Eng. [Proc. Annu. Tech. Conf.]
18, 114 (1972).
144. B. Cassel, Perkin-Elmer Therm. Anal. Appl. Study 19 (1977).
JEFFREY GOTRO
InnoCentrix, LLC, Rancho Santa Margarita, CA,
USA
R. BRUCE PRIME
IBM (Retired)/PrimeThermosets.com San Jose,
CA, USA