DYNAMIC MECHANICAL ANALYSIS IN THE ANALYSIS

OF POLYMERS AND RUBBERS

1. Introduction and History of DMA

Dynamic mechanical analysis (DMA) is the technique of applying a stress or

strain to a sample and analyzing the response to obtain phase angle and de-
formation data. These data allow the calculation of the damping or tan delta (δ)
as well as complex modulus and viscosity data. Two approaches are used: (a)
forced frequency, where the signal is applied at a set frequency and (b) free res-
onance, where the material is perturbed and allowed to exhibit free resonance
decay. Most DMAs are the former type, whereas the torsional braid analyzer is
the latter. In both approaches, the technique is very sensitive to the motions of the
polymer chains and it is a powerful tool for measuring transitions in polymers.
More specialized techniques such as the use of ultrasonics, laser pulses, and so
on do exist in specialized markets (1). It is estimated to be 100 times more sen-
sitive to the glass transition than differential scanning calorimetry (DSC), and
it resolves other more localized transitions such as side chain movements that
are not detected in the DSC. In addition, the technique allows the rapid scan-
ning of a material’s modulus and viscosity as a function of temperature, strain,
or frequency. DMA may also be referred to as dynamic mechanical thermal anal-
ysis (DMTA), dynamic mechanical spectroscopy, or dynamic thermomechanical
analysis.
DMA is a very important tool in the modern polymer laboratory. Despite
that, only a few books have concentrated on the technique (2). The first attempt
to do oscillatory experiments to measure the elasticity of a material was by Poynt-
ing (3) in 1909. Other early work gave methods to apply oscillatory deformations
by various means to study metals (4), and many early experimental techniques
were reviewed by te Nijenhuis (5) in 1978. Miller’s book on polymer properties (6)
referred to dynamic measurements in this early discussion of molecular structure
and stiffness. Early commercial instruments included the Weissenberg rheogo-
niometer (1950) and the Rheovibron (1958). The Weissenberg rheogoniometer,
which dominated cone-and-plate measurements for over 20 years following 1955,
was the commercial version of the first instrument to measure normal forces (7).
By the time, Ferry wrote the Viscoelastic Properties of Polymers (8) in 1961, dy-
namic measurements were an integral part of polymer science and he gives the
best development of the theory available. In 1967, McCrum and co-workers (9)
collected the current information on DMA and DEA (dielectric analysis) into
his landmark textbook. The technique remained fairly specialized until the late
1
Encyclopedia of Polymer Science and Technology. Copyright 
c 2015 John Wiley & Sons, Inc. All rights reserved.
DOI: 10.1002/0471440264.pst102.pub2
2 DYNAMIC MECHANICAL ANALYSIS

sixties when commercial instruments became more user-friendly. About 1966,

Gillham developed the TBA (10) and started the modern period of DMA. In 1971,
Starita and Macosko (11) built a DMA that measured normal forces and from this
came the Rheometrics Corporation. In 1976, Bohlin also developed a commercial
DMA and started Bohlin Rheologia. Both instruments used torsional geometry.
The early instruments were, regardless of manufacturer, difficult to use, slow, and
limited in their ability to process data. In the late seventies, Murayani (12) and
Read and Brown (13) wrote books on the uses of DMA for material characteriza-
tion. Several thermal and rheological instrument companies introduced DMA’s in
the same time period, and currently most thermal and rheological vendors offer
some type of DMA. Polymer Labs offered a dynamic mechanical thermal ana-
lyzer (DMTA using an axial geometry in the early 1980s. This was soon followed
by an instrument from Du Pont. Perkin-Elmer developed a controlled stress ana-
lyzer based on their thermomechanical analyzer (TMA) technology, which was de-
signed for increased low-end sensitivity. Triton Technologies, acquired by Mettler
Toledo in 2010, developed a very easy to use design that formalized both immer-
sion and humidity methods. On the high force end of applications, Mettler Toledo
also makes an instrument capable of 40 N loads and both Gabo and MetaVib
work exclusively in high load instruments. Small vendors include Hitachi, Bosch,
and others. The competition between vendors has led to easier to use, faster, and
less expensive instruments. The revolution in computer technology, which has so
affected the laboratory, changed the latter, and DMA of all types became more
user-friendly as computers and software evolved. This movement to easier to use
instruments has led to more use in quality control and the development of ASTM
standards (14).

2. Theory and Operating Principles

2.1. Forced Frequency Analyzers. If a constant load applied to a sam-

ple begins to oscillate sinusoidally (Fig. 1), the sample will deform sinusoidally.
This will be reproducible if the material is deformed within its linear viscoelas-
tic region. For any one point on the curve, the stress applied is described in
equation (1)

σ = σo sin ωt (1)

where σ is the stress at time t, σo is the maximum stress, ω is the frequency

of oscillation, and t is the time. The resulting strain wave shape will depend on
how much viscous behavior the sample has as well as how much elastic behavior.
In addition, the rate of stress can be determined by taking the derivative of the
above equation in terms of time:

dσ/dt = ωσo cos ωt (2)

The two extremes of the material’s behavior, elastic and viscous, provide the lim-
iting extremes that will sum to give the strain wave. The behavior can be under-
stood by evaluating each of the two extremes. The material at the spring-like or
 DYNAMIC MECHANICAL ANALYSIS 3

Fig. 1. Oscillating stress applied to a sample.

Hookean limit will respond elastically with the oscillating stress. The strain at
any time can be written as

ε (t) = Eso sin(ωt) (3)

where ε(t) is the strain at anytime t, E is the modulus, σo is the maximum stress
at the peak of the sine wave, and ω is the frequency. Since in the linear region σ
and ε are linearly related by E, the relationship is

ε (t) = εo sin(ωt) (4)

where εo is the strain at the maximum stress. This curve shown in Figure 2a
has no phase lag (or no time difference from the stress curve) and is called the
in-phase portion of the curve.
The viscous limit was expressed as the stress being proportional to the
strain rate, which is the first derivative of the strain. This is best modeled by
a dashpot and for that element, the term for the viscous response in terms of
strain rate is described as

ε (t) = ηdσo /dt = ηωσo cos(ωt) (5)

or

ε (t) = ηωσo sin(ωt + π/2) (6)

4 DYNAMIC MECHANICAL ANALYSIS

Fig. 2. Material responses in DMA.

where the terms are as above and η is the viscosity. Substituting terms as above
makes the equation

ε (t) = ωσo cos(ωt) = ωσo sin(ωt + π/2) (7)

This curve is shown in Figure 2b. Now, take the behavior of the material that lies
between these two limits. That curve is shown in Figure 2c and is intermediate
between the above cases. The difference between the applied stress and the re-
sultant strain is an angle, δ, and this must be added to equations. So the elastic
response at anytime can now be written as

ε (t) = εo sin(ωt + δ) (8)

Using trigonometry, this can be rewritten as

ε (t) = εo [sin(ωt) cos δ + cos(ωt) sin δ] (9)

This equation, corresponding to the curve in Figure 2c, can be separated into the
in-phase and out-of-phase strains that corresponds to curves like those in Figures
2a and 2b, respectively. These sum to the curve in Figure 2c and are

ε = εo sin(δ) (10)

ε = εo cos(δ) (11)

 DYNAMIC MECHANICAL ANALYSIS 5

Fig. 3. Decay curve.

and the vector sum of these two components gives the overall or complex strain
on the sample:

ε∗ = ε + iε (12)

2.2. Free Resonance Analyzers. If a suspended sample can swing

freely, it will oscillate like a harp or guitar string until the oscillations gradu-
ally come to a stop. The naturally occurring damping of the material controls the
decay of the oscillations. This produces a wave, shown in Figure 3, which is a
series of sine waves decreasing in amplitude and frequency. Several methods ex-
ist to analyze these waves and are covered in the review by Gillham (15). These
methods have also been successfully applied to the recovery portion of a creep-
recovery curve where the sample goes into free resonance on removal of the creep
force (16).
From the decay curve, the period, T, and the logarithmic decrement, , can
be calculated. Several methods exist for both manual and digital processing (15,
17). Fuller details of the following may be found in McCrum and colleagues (17)
and Gillham (15). Basically the decay of the amplitude is evaluated over as many
swings as possible to reduce error:
 
 = 1/j ln An/A(n+j ) (13)

where j is the number of swings and An is the amplitude of the nth swing. For one
swing, where j = 1, the equation becomes
 
 = ln An/A(n+j ) (14)

if for a low value of  where An /An+1 is approximately 1, the equation can be

rewritten as

1  2  
≈ An − An+1 2 /An2 (15)
2
6 DYNAMIC MECHANICAL ANALYSIS

from this, since the square of the amplitude is proportional to the stored energy,
W/Wst , and the stored energy can be expressed as 2π tan δ, the equation becomes

 ≈ 12 ( W/Wst ) = π tan δ (16)

which gives us the phase angle, δ. The time of the oscillations, the period, T, can
be found using the following equation:
 
M 1 + 2
T = 2π (17)
1 4π2

where 1 is the torque for one cycle and M is the moment of inertia around the
central axis. Alternatively, T can be calculated directly from the plotted decay
curve as

T = (2/n) (tn − t0 ) (18)

where n is the number of cycles and t is time. From this, the shear modulus, G,
can be calculated, which for a rod of length, L, and radius, r, is
   
4π2 ML 2 mgr 
G= 1+ 2 − (19)
NT2 4π 12N

where m is the mass of the sample, g the gravitational constant and N is a geo-
metric factor. In the same system, the storage modulus, G , can be calculated as

G = (1/T2 )(8πML/r4 ) (20)

Having the storage modulus and the tangent of the phase angle, the remaining
dynamic properties can be calculated.
Free resonance analyzers normally are limited to rod or rectangular samples
or materials that can be impregnated onto a braid. This last approach is how the
curing studies on epoxy and other resin systems were done in torsion and gives
these instruments the name of torsional braid analyzers (TBA).

3. Instrumentation

One of the biggest choices made in selecting a DMA is to decide whether to chose
stress (force) or strain (displacement) control for applying the deforming load to
the sample (Figs. 4a and 4b). Strain-controlled analyzers, whether for simple
static testing or for DMA, move the probe a set distance and use a force balance
transducer or load cell to measure the stress. These parts are typically located on
different shafts. The simplest version of this is a screw-driven tester, where the
sample is pulled one turn. This requires very large motors so the available force
 DYNAMIC MECHANICAL ANALYSIS 7

Fig. 4. Instrument types.

always exceeds what is needed. They normally have better short time response
for low viscosity materials and can normally perform stress relaxation experi-
ments easily. They also usually can measure normal forces if they are arranged
in torsion. A major disadvantage is that their transducers may drift at long times
or with low signals.
Stress-controlled analyzers are cheaper to make because there is only one
shaft, but somewhat trickier to use. Many of the difficulties have been allevi-
ated by software, and many strain-controlled analyzers on the market are really
stress-controlled instruments with feedback loops making them act as if they
were strain controlled. In stress control, a set force is applied to the sample. As
temperature, time, or frequency varies, the applied force remains the same. This
may or may not be the same stress: In extension, for example, the stretching and
necking of a sample will change the applied stress seen during the run. However,
this constant stress is a more natural situation in many cases and it may be more
sensitive to material changes. Good low force control means they are less likely to
destroy any structure in the sample. Long relaxation times or long creep studies
are more easily preformed on these instruments. Their biggest disadvantage is
that their short time responses are limited by inertia with low viscosity samples.
Most stress-controlled rheometers today use a feedback loop to mimic strain con-
trol and some, like the Mettler Toledo DMA/SDTA861e , will switch automatically
from one mode to the other depending the samples behavior,
Since most DMA experiments are run at very low strains (0.5% maximum)
to stay well within a polymers linear region, it has been reported that the both
analyzers give the same results. However, when one gets to the nonlinear region,
the difference becomes significant as stress and strain are no longer linearly re-
lated. Stress control can be said to duplicate real-life conditions more accurately
since most applications of polymers involve resisting a load. However, the choice
is best made on the specific application.
DMA analyzers are normally built to apply the stress or strain in two ways
(Figs. 4c and 4d). One can apply force in a twisting motion so one is testing the
sample in torsion. This type of instrument is the dynamic analog of the constant
shear spinning disk rheometers. While mainly used for liquids and melts, solid
8 DYNAMIC MECHANICAL ANALYSIS

samples may also tested by twisting a bar of the material. Torsional analyzers
normally also permit continuous shear and normal force measurements. Most of
these analyzers can also do creep-recovery, stress-relaxation, and stress–strain
experiments.
Axial analyzers are normally designed for solid and semisolid materials and
apply a linear force to the sample. These analyzers are usually associated with
flexure, tensile, and compression testing, but they can be adapted to do shear and
liquid specimens by proper choice of fixtures. Sometimes the instrument’s design
makes this inadvisable, however (eg, working with a very fluid material in a sys-
tem where the motor is underneath the sample has the potential for damage to
the instrument if the sample spills into the motor). These analyzers can normally
test higher modulus materials than torsional analyzers and can run TMA studies
in addition to creep-recovery, stress-relaxation, and stress–strain experiments.
Despite the traditional selection of torsional instruments for melts and liq-
uids and axial instruments for solids, there is really considerable overlap between
the types of instruments. With the proper choice of sample geometry and good fix-
tures, both types can handle similar samples, as shown by the use of both types
to study the curing of neat resins. Normally axial analyzers can not handle fluid
samples below about 2.5 Pa s, and torsional instruments will top out with the
harder samples (the exact modulus depending on the size of the motor and/or
load cell.)

4. Applications

4.1. Thermoplastic Solids and Cured Thermosets. The thermal

transitions in polymers can be described in terms of either free volume changes
(18) or relaxation times (19). While the latter tends to be preferred by engineers
and rheologists in contrast to chemist and polymer physicists who lean toward
the former, both descriptions are equivalent. Changes in free volume, vf , can be
monitored as a volumetric change in the polymer, by the absorption or release of
heat associated with that change, the loss of stiffness, increased flow, or by the a
change in relaxation time.
The free volume of a polymer, vf , is known to be related to viscoelastic-
ity (20), aging (21), penetration by solvents (22), and impact properties (23).
Defined as the space a molecule has for internal movement, it is schematically
shown in Figure 5a. A simple approach to looking at free volume is the crankshaft
mechanism (24), where the molecule is imagined as a series of jointed segments.
From this model, it is possible to simply describe the various transitions seen a
polymer. Other models exist that allow for more precision in describing behavior;
the best seems to be the Doi–Edwards model (25). Aklonis and Knight (26) give a
good summary of the available models, as does Rohn (27).
The crankshaft model treats the molecule as a collection of mobile segments
that have some degree of free movement. This is a very simplistic approach, yet
very useful for explaining behavior. As the free volume of the chain segment in-
creases, its ability to move in various directions also increases (Fig. 5b). This
increased mobility in either side chains or small groups of adjacent backbone
atoms results in a greater compliance (lower modulus) of the molecule. These
 DYNAMIC MECHANICAL ANALYSIS 9

Fig. 5. Free volume and molecular motions in polymers.

Fig. 6. Idealized DMA scan.

movements have been studied and Heijboer (28) classified β and γ transitions by
their type of motions. The specific temperature and frequency of this softening
help drive the end use of the material.
Moving from very low temperature, where the molecule is tightly com-
pressed, to higher temperatures the first changes are the solid-state transitions.
This process is shown in Figure 6. As the material warms and expands, the free
volume increases so that localized bond movements (bending and stretching) and
side chain movements can occur. This is the gamma transition, Tγ , which may
also involve associations with water (29). As the temperature and the free volume
10 DYNAMIC MECHANICAL ANALYSIS

continue to increase, the whole side chains and localized groups of four to eight
backbone atoms begin to have enough space to move and the material starts to
develop some toughness (30). This transition, called the beta transition Tβ , is not
as clearly defined as described here (Fig. 6). Often it is the Tg of a secondary com-
ponent in a blend or of a specific block in a block copolymer. However, a correlation
with toughness is seen empirically (31).
As heating continues, the Tg or glass transition appears when the chains
in the amorphous regions begin to coordinate large-scale motions (Fig. 6). One
classical description of this region is that the amorphous regions have begun to
melt. Since the Tg only occurs in amorphous material, in a 100% crystalline ma-
terial there would not be a Tg . Continued heating drives the material through the
Tα * and Tll (Fig. 6d). The former occurs in crystalline or semicrystalline polymer
and is a slippage of the crystallites past each other. The latter is a movement of
coordinated segments in the amorphous phase that relates to reduced viscosity.
These two transitions are not universally accepted. Finally, the melt is reached
(Fig. 6e) where large-scale chain slippage occurs and the material flows. This is
the melting temperature, Tm . For a cured thermoset, nothing happens after the
Tg until the sample begins to burn and degrade because the cross-links prevent
the chains from slipping past each other.
This quick overview provides an idea of how an idealized polymer responds.
Now a more detailed description of these transitions can be provided with some
examples of their applications. The best general collection of this information is
still McCrum’s 1967 text (24).

4.1.1. Sub-Tg Transitions. The area of sub-Tg or higher order transitions

has been heavily studied (32) as these transitions have been associated with me-
chanical properties. These transitions can sometimes be seen by DSC and TMA,
but they are normally too weak or too broad for determination by these meth-
ods. DMA, DEA, and similar techniques are usually required (33). Some authors
have also called these types of transitions (34) second-order transitions to dif-
ferentiate them from the primary transitions of Tm and Tg , which involve large
sections of the main chains. Boyer reviewed the Tβ in 1968 (35) and pointed out
that while a correlation often exists, the Tβ is not always an indicator of tough-
ness. Bershtein and Egorov (36) have reported that this transition can be consid-
ered the “activation barrier” for solid-phase reactions, deformation, flow or creep,
acoustic damping, physical aging changes, and gas diffusion into polymers as the
activation energies for the transition, and these processes are usually similar.
The strength of these transitions is related to how strongly a polymer responses
to those processes. These sub-Tg transitions are associated with the materials
properties in the glassy state. In paints, for example, peel strength (adhesion)
can be estimated from the strength and frequency dependence of the subam-
bient beta transition (37). Nylon 6,6 shows a decreasing toughness, measured
as impact resistance, with declining area under the Tβ peak in the tan δ curve.
Figure 7 shows the relative differences in the Tβ compared to the Tg for a high im-
pact and low impact nylon. It has been shown, particularly in cured thermosets,
that increased freedom of movement in side chains increases the strength of the
transition. Cheng and co-workers (38) report in rigid rod polyimides that the beta
transition is caused by the noncoordinated movement of the diamine groups al-
 DYNAMIC MECHANICAL ANALYSIS 11

Fig. 7. Beta transitions and toughness.

though the link to physical properties was not investigated. Johari and their as-
sociates have reported in both mechanical (39) and dielectric studies (40) that
both the β and γ transitions in bisphenol-A based thermosets depend on the
side chains and unreacted ends, and that both are affected by physical aging and
postcure. Nelson (41) has reported that these transitions can be related to vibra-
tion damping. This is also true for acoustical damping (42). In both of this cases,
the strength of the beta transition is taken as a measurement of how effectively
a polymer will absorb vibrations. There is some frequency dependence involved
in this, which will be discussed later in Section 5.7.
Boyer (43) and Heijober (28) showed that this information needs to be con-
sidered with care as not all beta transitions correlate with toughness or other
properties. This can be due to misidentification of the transition, or that the
transition does sufficiently disperse energy. A working rule of thumb (44) is that
the beta transition must be related to either localized movement in the main
chain or very large side chain movement to sufficiently absorb enough energy.
The relationship of large side chain movement and toughness has been exten-
sively studied in polycarbonate by Yee (45) as well as in many other tough glassy
polymers (46).
Less use is made of the Tγ transitions, and they are mainly studied to un-
derstand the movements occurring in polymers. Wendorff (47) reports that this
transition in polyarylates is limited to inter- and intramolecular motions within
the scale of a single repeat unit. Both McCrum and co-workers (24) and Boyd (48)
similarly limited the Tγ and Tδ to very small motions either within the molecule
or with bound water. The use of what is called 2D-IR, which couples at FTIR and
a DMA to study these motions, is a topic of current interest (49).

4.1.2. The Glass Transition (Tg or Tα ). As the free volume continues to

increase with increasing temperature, the glass transition, Tg , occurs where large
segments of the chain start moving. This transition is also called the alpha tran-
sition, Tα . The Tg is very dependent on the degree of polymerization up to a value
known as the critical Tg or the critical molecular weight. Above this value, the
Tg typically becomes independent of molecular weight (50). The Tg represents a
12 DYNAMIC MECHANICAL ANALYSIS

Fig. 8. Operating range by DMA.

major transition for many polymers, as physical properties changes drastically as

the material goes from a hard glassy to a rubbery state. It defines one end of the
temperature range over which the polymer can be used, often called the operat-
ing range of the polymer, and an example of this range is shown in Figure 8. For
where strength and stiffness are needed, it is normally the upper limit for use. In
rubbers and some semicrystalline materials such as polyethylene and polypropy-
lene, it is the lower operating temperature. Changes in the temperature of the
Tg are commonly used to monitor changes in the polymer such as plasticizing by
environmental solvents and increased cross-linking from thermal or UV aging.
The Tg of cured materials or thin coatings is often difficult to measure by
other methods, and more often than not the initial cost justification for a DMA is
measuring a hard to find Tg . While estimates of the relative sensitivity of DMA
to DSC or DTA vary, it appears that DMA is 10–100 times more sensitive to the
changes occurring at the Tg . The Tg in highly cross-linked materials can eas-
ily be seen long after the Tg has become too flat and diffuse to be seen in the
DSC (Fig. 9). A highly cross-linked molding resin used for chip encapsulation was
run by both methods, and the DMA is able to detect the transition after it is
undetectable in the DSC. This is also a known problem with certain materials
such as medical-grade urethanes and very highly crystalline polyethylenes.
The method of determining the Tg in the DMA can be a manner for disagree-
ment as at least five ways that are in current use (Fig. 10a). It should be noted
that DSC has multiple methods too (Fig. 10b). Depending on the industry stan-
dards or background of the operator, the peak or onset of the tan δ curve, the onset
of the E drop, or the onset or peak of the E curve may be used. The values ob-
tained from these methods can differ up to 25°C from each other on the same run.
In addition, a 10–20°C difference from the DSC is also seen in many materials.
In practice, it is important to specify exactly how the Tg should be determined.
For DMA, this means defining the heating rate, applied stresses (or strains), the
 DYNAMIC MECHANICAL ANALYSIS 13

Fig. 9. Electronic encapsulation material by DMA and DSC.

Peak tan δ = 140.5 °C

Onset E' = 133.1°C

Tf
Peak E" = 136.7°C
Heat flow/mW
Modulus/Pa

tan δ

ΔH/(J/g)
Onset E" = 127.3°C
Inflection point ΔCp

Onset
Onset tan δ = 130.0°C
Temperature/˚C

Temperature/˚C
(a) (b)

Fig. 10. Tg by DMA and DSC.

frequency used, and the method of determining the Tg . For example, the sample
will be run at 10°C min−1 under 0.05% strain at 1 Hz in nitrogen purge (20 cc
min−1 ) and the Tg determined from peak of the tan δ curve.
It is not unusual to see a peak or hump on the storage modulus directly
preceding the drop that corresponds to the Tg . This is shown in Figure 11. This
is also seen in the DSC and DTA and corresponds to a rearrangement in the
molecule to relieve stresses frozen in below the Tg by the processing method.
These stresses are trapped in the material until enough mobility is obtained at
the Tg to allow the chains to move to a lower energy state. Often a material will
be annealed by heating it above the Tg and slowly cooling it to remove this effect.
For similar reasons, some experimenters will run a material twice or use a heat—
cool–heat cycle to eliminate processing effects.
 14 DYNAMIC MECHANICAL ANALYSIS

1.12e+9

1.00e+9
The overshoot at the Tg
is caused by stress relief
9.00e+8

8.00e+8

7.00e+8
Modulus (Pa)

6.00e+8

5.00e+8

4.00e+8

3.00e+8

2.00e+8

1.00e+8

100
25.4 30 35 40 45 50 55 60 65 70 75 81.8
Temperature (˚C)

Fig. 11. Stress relief.

4.1.3. The Rubbery Plateau, Tα * and Tll . The area above the Tg and be-
low the melt is known as the rubbery plateau, and the length of it as well as its
viscosity is dependent on the molecular weight between entanglements (Me ) (51)
or cross-links. The molecular weight between entanglements is normally calcu-
lated during a stress-relaxation experiment, but similar behavior is observed in
the DMA. The modulus in the plateau region is proportional to either the number
of cross-links or the chain length between entanglements. This is often expressed
in shear as

G ∼
= (ρRT)/Me (21)

where G is the shear storage modulus of the plateau region at a specific tempera-
ture, ρ is the polymer density, and Me is the molecular weight between entangle-
ments. In practice, the relative modulus of the plateau region shows the relative
changes in Me or the number of cross-links compared to a standard material.
The rubbery plateau is also related to the degree of crystallinity in a ma-
terial, although DSC is a better method for characterizing crystallinity than
DMA (52). Also as in the DSC, there is evidence of cold crystallization in the tem-
perature range above the Tg (Fig. 12). That is one of several transitions that can
be seen in the rubbery plateau region. This crystallization occurs when the poly-
mer chains have been quenched (quickly cooled) into a highly disordered state.
On heating above the Tg , these chains gain enough mobility to rearrange into
crystallites, which causes sometimes a dramatic increase in modulus. DSC or
one of its temperature-modulated (MT-DSC) variants, such as Mettler’s ADSC
or TOPEM or PerkinElmer’s StepScan, or Water’s TMDSC, for example, can
be used to confirm this (53). The alpha star transition, Tα *, the liquid–liquid
transition, Tll , the heat set temperature, and the cold crystallization peak are
 DYNAMIC MECHANICAL ANALYSIS 15

Fig. 12. Cold crystallization in PET.

Fig. 13. Crystal-crystal slip.

all transitions that can appear on the rubbery plateau. In some crystalline and
semicrystalline polymer, a transition is seen here called the Tα * (54). Figure 13
shows this in a sample of polypropylene. The alpha star transition is associated
with the slippage between crystallites and helps extend the operating range of a
material above the Tg . This transition is very susceptible to processing-induced
changes and can be enlarged on decreased by the applied heat history, processing
conditions, and physical aging (55). The Tα * has been used by fiber manufacturers
to optimize properties in their materials.
In amorphous polymers, the Tll transition is seen instead, which is a liquid–
liquid transition associated with increased chain mobility and segment–segment
associations (56). This order is lost when the Tll is exceeded and regained on
cooling from the melt. Boyer (57) reports that, like the Tg , the appearance of
the Tll is affected by the heat history. The Tll is also dependent on the num-
ber average molecular weight, Mn , but not on the weight average molecular
weight, Mw . Bershtein and co-workers (58) suggest that this may be considered
as quasi-melting on heating or the formation of stable associates of segments
16 DYNAMIC MECHANICAL ANALYSIS

Fig. 14. Effect of MW on flow.

on cooling. While this transition is reversible, it is not always easy to see, and
Boyer (59) spent many years trying to prove it was real. Not everyone accepts
the existence of this transition. This transition may be similar to some of the
data from temperature-modulated DSC experiments showing a recrystallization
at the start of the melt (60). In both cases, some subtle changes in structure are
sometimes detected at the start of melting. Following this transition, a material
enters the terminal or melting region.
Depending on its strength, the heat set temperature can also be seen in the
DMA. While it is normally seen in either a TMA or a Constant Gauge Length
(CGL) experiment, it will sometimes appear as either a sharp drop in storage
modulus (E ) or an abrupt change in probe position. Heat set is the temperature
at which some strain or distortion is induced into polymeric fibers to change its
properties, such as to prevent a nylon rug from feeling like fishing line. Since
heating above this temperature will erase the texture, and polyesters must be
heated above the Tg to dye them, it is of critical importance to the fabric indus-
try. Many final properties of polymeric products depend on changes induced in
processing (61).

4.1.4. The Terminal Region. On continued heating, the melting point,

Tm , is reached. The melting point is where the free volume has increased so
the chains can slide past each other and the material flows. This is also called
the terminal region. In the molten state, this ability to flow is dependent on the
molecular weight of the polymer (Fig. 14). The melt of a polymer material will
often show changes in temperature of melting, width of the melting peak, and
enthalpy as the material changes (62), resulting from changes in the polymer
molecular weight and crystallinity.
Degradation, polymer structure, and environmental effects all influence
what changes occur. Polymers that degrade by cross-linking will look very dif-
ferent from those that exhibit chain scission. Very highly cross-linked polymers
will not melt as they are unable to flow.
The study of polymer melts and especially their elasticity was one of the
areas that drove the development of commercial DMAs. Although a decrease in
the melt viscosity is seen with temperature increases, the DMA is most commonly
 DYNAMIC MECHANICAL ANALYSIS 17

Table 1. ASTM Tests for the DMA

D3386 CTE of Electrical Insulating Materials by TMA
D4065 Determining DMA Properties Terminologya
D4092 Terminology for DMA Tests
D4440 Measurement of Polymer Melts
D4473 Cure of Thermosetting Resins
D5023 DMA in Three Point Bending Tests
D5024 DMA in Compression
D5026 DMA in Tension
D5279 DMA of Plastics in Torsion
D5418 DMA in Dual Cantilever
D7028 Glass Transition Temperature of Polymer Matrix Composites
D7750 Cure Behavior of Thermosetting Resins by DMA using an Encapsulated
Specimen Rheometer
E 228 CTE by TMA with Silica Dilatometer
E 473 Terminology for Thermal Analysis
E-756 Measuring Vibration-Damping Properties of Materials
E 831 CTE of Solids by TMA
E 1363 Temperature Calibration for TMA
E 1545 Tg by TMA
E 1640 Tg by DMA
E 1824 Tg by TMA in Tension
E 1867 Temperature Calibration for DMA
E-1953 Standard Practice for Description of Thermal Analysis and Rheology Apparatus
E-2254 Standard Test Method for Storage Modulus Calibration of DMA
E-2425 Standard Test Method for Loss Modulus Conformance of DMA
F-3131 Specification for Epoxy/Cotton Raw Materials for the Use in Bearing Cages
a This standard qualifies a DMA as acceptable for all ASTM DMA Standards.

used to measure the frequency dependence of the molten polymer as well as its
elasticity. The latter property, especially when expressed as the normal forces, is
very important in polymer processing.

4.1.5. Frequency Dependencies in Transition Studies. The choice of a

testing frequency or its effect on the resulting data must be addressed. A short
discussion of how frequencies are chosen and how they affect the measurement of
transitions is in order. Considering that higher frequencies induce more elastic-
like behavior, there is some concern a material will act stiffer than it really is if
the test frequency is chosen to be too high. Frequencies for testing are normally
chosen by one of three methods. The most scientific method would be to use the
frequency of the stress or strain that the material is exposed to in the real world.
However, this is often outside of the range of the available instrumentation. In
some cases, the test method or the industry standard sets a certain frequency and
this frequency is used. Ideally a standard method like this is chosen so that the
data collected on various commercial instruments can be shown to be compatible.
Some of the ASTM methods for DMA are listed in Table 1. Many industries have
their own standards so it is important to know whether the data are expected
to match a Mil-spec, an ASTM standard, or a specific industrial test. Finally,
18 DYNAMIC MECHANICAL ANALYSIS

Fig. 15. General trends in frequency scans.

one can arbitrarily pick a frequency. This is done more often than not, so that 1
Hz and 10 rad s–1 are often used. As long as the data are run under the proper
conditions, they can be compared to highlight material differences. This requires
that frequency, stresses, and the thermal program be the same for all samples in
the data set.
Lowering the frequency shifts the temperature of a transition to a lower
temperature (Fig. 15). At one time, it was suggested that multiple frequencies
could be used and the Tg should then be determined by extrapolation to 0 Hz. This
was never really accepted as it represented a fairly large increase in testing time
for a small improvement in accuracy. For most polymer systems, for very precise
measurements, one uses a DSC. Different types of transitions also have different
frequency dependencies; McCrum and co-workers (24) listed many of these. If one
looks at the slope of the temperature dependence of transitions against frequency,
one sees that in many cases the primary transitions such as Tm and Tg have a
different dependence on frequency than the lower temperature transitions. In
fact, the activation energies are different for α, β, and γ transitions because of the
different motions required and the transitions can be sorted by this approach (2).
4.2. Polymer Melts and Solutions. A fluid or polymer melt responds
to strain rate rather than to the amount of stress applied. The viscosity is one of
the main reason why people run frequency scans. As the stress–strain curves and
the creep recovery runs show, viscoelastic materials exhibit some degree of flow
or unrecoverable deformation. The effect is strongest in melts and liquids where
frequency versus viscosity plots are the major application of DMA.
Figure 16 shows a frequency scan on a viscoelastic material. In this exam-
ple, the sample is a rubber above the Tg in three-point bending, but the trends
and principles apply to both solids and melts. The storage modulus and complex
viscosity are plotted on log scales against the log of frequency. In analyzing the
frequency scans, trends in the data are more significant than specific peaks or
transitions.
Starting with the viscosity curve, η*, a fairly flat region appears at low fre-
quency, called the zero shear plateau (63). This is where the polymer exhibits
Newtonian behavior, and its viscosity is dependent on molecular weight, not the
 DYNAMIC MECHANICAL ANALYSIS 19

Fig. 16. Effect of frequency on transitions in filled polypropylene.

strain rate. The viscosity of this plateau has been shown to experimentally re-
lated to the molecular weight for Newtonian fluid:

η ∝ cMv1 (22)

for cases where the molecular weight, Mv , is less than the entanglement molecu-
lar weight, Me and for cases where Mv is greater than Me :

η ∝ cMv3.4 (23)

where ηo is the viscosity of the initial Newtonian plateau, c is a material constant,

and Mv the viscosity average molecular weight. This relationship can be writ-
ten in general terms, replacing the exponential term with the Mark–Houwink
constant, a. Equation 23 can be used as a method of approximating the molecular
weight of a polymer. The value obtained is closest to the viscosity average molec-
ular weight obtained by osmometry (64). In comparison with the weight average
data obtained by gel permeation chromatography, the viscosity average molecu-
lar weight would be between the number average and weight average molecular
weights, but closer to the latter (65). This was orginially developed for steady
shear viscosity but applies to complex viscosity as well. The relationship between
steady shear and complex viscosity is fairly well established. Cox and Merz (66)
found that an empirical relationship exists between complex viscosity and steady
20 DYNAMIC MECHANICAL ANALYSIS

shear viscosity when the shear rates are the same. The Cox–Merz rule is stated:

η(ω) = η(γ̇) γ̇=ω

(24)

where η is the constant shear viscosity, η* is the complex viscosity, ω is the fre-
quency of the dynamic test, and dγ/dt is the shear rate of the constant shear test.
This rule of thumb seems to hold for most materials to within about ±10%. An-
other approach is the Gleissile’s (67) mirror relationship that states the following:

ηγ̇ = η+ (t) t=1/γ̇

(25)

when η+ (t) is the limiting value of the viscosity as the shear rate, γ̇, approaches
zero.
The low frequency range is where viscous or liquid-like behavior predomi-
nates. If a material is stressed over long enough times, some flow occurs. As time
is the inverse of frequency, this means materials are expected to flow more at
low frequency. As the frequency increases, the material will act in a more and
more elastic fashion. Silly Puttyۛ, the children’s toy, shows this clearly. At low fre-
quency, Silly Puttyۛ flows like a liquid while at high frequency it bounces like a
rubber ball.
This behavior is also similar to what happens with temperature changes. A
polymer becomes softer and more fluid as it is heated, and it goes through transi-
tions that increase the available space for molecular motions. Over long enough
time periods, or small enough frequencies, similar changes occur. So one can move
a polymer across a transition by changing the frequency. This relationship is also
expressed as the idea of time–temperature equivalence (68). Often stated as low
temperature is equivalent to short times or high frequency, it is a fundamental
rule of thumb in understanding polymer behavior.
As the frequency is increased in a frequency scan, the Newtonian region
is exceeded and a new relationship develops between the rate of strain, or the
frequency, and the viscosity of the material. This region is often called the power
law zone and can be modeled by

η∗ ∼
= η(γ̇) = cγ̇ n−1 (26)

where η* is the complex viscosity, γ̇ is the shear rate, and the exponent term n is
determined by the fitting of the data. The can also be written as

σ∼
= η(γ̇) = cγ̇ n (27)

where σ is the stress and η is the viscosity. The exponential relationship is why
the viscosity versus frequency plot is traditionally plotted on a log scale. With
modern curve-fitting programs, the use of log–log plots has declined and is a bit
anachronistic. The power law region of polymers shows shear thickening or thin-
ning behavior. This is also the region in which the E –η* or the E –E crossover
 DYNAMIC MECHANICAL ANALYSIS 21

Fig. 17. Mw by crossover point.

point is found. As frequency increases and shear thinning occurs, the viscosity
(η*) decreases. At the same time, increasing the frequency increases the elastic-
ity (E ). This is shown in Figure 17. The E –η* crossover point is used an indicator
of the molecular weight and molecular weight distribution (69). Changes in its
position are used as a quick method of detecting changes in the molecular weight
and distribution of a material. After the power law region, another plateau is
seen, the infinite shear plateau.
This second Newtonian region corresponds to where the shear rate is so
high that the polymer no longer shows a response to increases in the shear rate.
At the very high shear rates associated with this region, the polymer chains are
no longer entangled. This region is seldom seen in DMA experiments and usually
avoided in use because of the damage done to the chains. It can be reached in
commercial extruders and causes degradation of the polymer, which causes the
poorer properties associated with regrind.
As the curve in Figure 17 shows, the modulus also varies as a function
of the frequency. A material exhibits more elastic-like behavior as the testing
frequency increases and the storage modulus tends to slope upward toward
higher frequency. The storage modulus’ change with frequency depends on the
transitions involved. Above the Tg , the storage modulus tends to be fairly flat
with a slight increase with increasing frequency as it is on the rubbery plateau.
The change in the region of a transition is greater. If one can generate a modulus
scan over a wide enough frequency range (Fig. 18), the plot of storage modu-
lus versus frequency appears like the reverse of a temperature scan. The same
time–temperature equivalence discussed above also applies to modulus, as well
as compliance, tan delta, and other properties.
The frequency scan is used for several purposes that will be discussed in
this section. One very important use, that is very straightforward, is to survey
the material’s response over various shear rates. This is important because many
22 DYNAMIC MECHANICAL ANALYSIS

Fig. 18. Temperature and frequency scans.

materials are used under different conditions. For example, adhesives, whether
tape, Band-AidsTM , or hot melts, are normally applied under conditions of low fre-
quency, and this property is referred to as tack. When they are removed, the re-
moval often occurs under conditions of high frequency called peel. Different prop-
erties are required at these regimes, and to optimize one property may require
chemical changes that harm the other. Similarly, changes in polymer structure
can show these kinds of differences in the frequency scan. For example, branch-
ing affects different frequencies (25).
For example, in a tape adhesive, sufficient flow under pressure at low fre-
quency is desired to fill the pores of the material to obtain a good mechanical
bond. When the laminate is later subjected to peel, the material needs to be very
elastic so it will not pull out of the pores (70). The frequency scan allows mea-
surement of these properties in one scan so ensuring that tuning one property
does not degrade another. This type of testing is not limited to adhesives as many
materials see multiple frequencies in the actual use. Viscosity versus frequency
plots are used extensively to study how changes in polymer structure or formu-
lations affect the behavior of the melt. Often changes in materials, especially in
uncured thermosetting resins and molten materials, affect a limited frequency
range, and testing at a specific frequency can miss the problem.
It should be noted that since the material is scanned across a frequency
range, there are some conditions where the material–instrument system acts
like a guitar string and begins to resonate when certain frequencies are
reached. These frequencies are either the natural resonance frequency of the
sample–instrument system or one of its harmonics. This is shown in Figure 19.
Under this set of experimental conditions, the sample–instrument system is os-
cillating like a guitar string and the desired information about the sample is ob-
scured. Since there is no way to change this resonance behavior as it is a function
of the system (in a free resonance analyzer, this effect is necessary), it is required
 DYNAMIC MECHANICAL ANALYSIS 23

RESONANCE
η*
1010 HARMONICS 1010

Complex viscosity
E′
Storage modulus
109 109

108 108

107 107

106 106
10–2 10–1 100 101 102
Frequency (Hz)

Fig. 19. Resonance effects.

Fig. 20. Analysis of the cure profiles for a beta staged carbon fiber prepreg.

to redesign the experiment by changing sample dimensions or geometry to escape

the problem. Often the design of the instrument will try and place the natural
resonance frequency of the material outside its operating range (71). Even so,
since it is changed by the sample, it does occur in real life. Using a sample with
much different dimensions, which changes the mass, or changing from extension
to three-point bending geometry changes the natural oscillation frequency of the
sample and hopefully solves this problem.

4.3. Thermoset Curing. The DMA’s ability to give viscosity and modu-
lus values for each point in a temperature scan allow estimation of kinetic behav-
ior as a function of viscosity. This has the advantage of describing how fluid the
material is at any given time, so as to determine the best time to apply pressure,
what design of tooling to use, and when the material can be removed from the
mold. The simplest way to analyze a resin system is to run a plain temperature
ramp from ambient to some elevated temperature (71). This “cure profile” allows
collection of several vital pieces of information as shown in Figure 20.
24 DYNAMIC MECHANICAL ANALYSIS

4.3.1. Cure Studies in the DMA. Before analyzing the cure in Figure 20
in more detail, it should be mentioned that, in curing studies, all three types of
commercial DMA’s are used. The shape of curve and the temperature of events
follow the same pattern. The values for viscosity and modulus often differ greatly.
Both types of forced resonance DMA’s also use samples impregnated into fabrics
in techniques that are referred to as “torsional braid.” There are some problems
with this technique as temperature increases will cause an apparent curing of
nondrying oils as thermal expansion increases friction. However, the “soaking of
resin into a shoelace,” as this technique has been called, allows one to handle
difficult specimens under conditions where the pure resin is impossible to run in
bulk (due to viscosity or evolved volatiles). Steel envelopes can also be used (73).
Composite materials like graphite–epoxy composites are sometimes studied in
industrial situations as the composite rather than the “neat” or pure resin be-
cause of the concern that the kinetics may be significantly different. In terms of
ease of handling and sample, the composite is often easier to work with. Some ax-
ial instruments have addressed this with the development of specialized clamps
for holding thin liquids in the shear mode. For example, the Mettler DMA/SDTA
861e low viscosity shear clamps can work with samples down to 2.5 Pa sec (ap-
proximately the viscosity of warm honey) when a very small gap is used.
Another special area of concern is paints and coatings (72) where the ma-
terial is used in a thin layer. This can be addressed experimentally by either a
braid as above or coating the material on a thin sheet of metal. The metal is often
run first, and its scan is subtracted from the coated sheet’s scan to leave only the
scan of the coating. Envelopes of stainless steel sheet or fine mesh can also be
used, and these containers also allow the running of powders and extremely fluid
materials in an axial designed DMA (73). This is also done with thin films and
adhesive coatings.
A sample cure profile for a commercial two-part epoxy resin is shown in Fig-
ure 20. From this scan, it is possible to determine the minimum viscosity (η*min ),
the time to η*min , and the length of time it stays there, the onset of cure, the
point of gelation where the material changes from a viscous liquid to a viscoelas-
tic solid, and the beginning of vitrification. The minimum viscosity is seen in the
complex viscosity curve and is where the resin viscosity is the lowest. A given
resin’s minimum viscosity is determined by the resin’s chemistry, the previous
heat history of the resin, the rate at which the temperature is increased, and the
amount for stress or stain applied. Increasing the rate of the temperature ramp
is known to decrease the η*min , the time to η*min , and the gel time. The resin gets
softer faster, but also cures faster. The degree of flow limits the type of mold de-
sign and when as well as how much pressure can be applied to the sample. The
time spent at the minimum viscosity plateau is the result of a competitive rela-
tionship between the material’s softening or melting as it heats and its rate of
curing. At some point, the material begins curing faster than it softens, and that
is where the viscosity starts to increase.
As the viscosity begins to climb, an inversion is seen of the E and E values
as the material becomes more solid-like. This crossover point also corresponds to
where the tan δ equals 1 (since E = E at the crossover). This is taken to be the gel
 DYNAMIC MECHANICAL ANALYSIS 25

point (74), where the cross-links have progressed to forming an “infinitely” long
network across the specimen. At this point, the sample will no longer dissolve
in solvent. While the gel point correlates fairly often with this crossover, it does
not always. For example, for low initiator levels in chain addition thermosets, the
gel point precedes the modulus crossover (75). A temperature dependence for the
presence of the crossover has also been reported (69). In some cases, where pow-
der compact and melts before curing, there may be several crossovers (76). Then,
the one following the η*min is usually the one of interest. Some researchers (77)
believe the true gel point is best detected by measuring the frequency depen-
dence of the crossover point. This is done by either by multiple runs at different
frequencies or by multiplexing frequencies during the cure. At the gel point, the
frequency dependence disappears (78). This value is usually only a few degrees
different from the one obtained in a normal scan and in most cases not worth the
additional time. During this rapid climb of viscosity in the cure, the slope for η*
increase can be used to calculate an estimated Ea (activation energy) (79). This
will be discussed below, but the fact that the slope of the curve here is a func-
tion of Ea is important. Above the gel temperature, some workers estimate the
molecular weight, Mc , between cross-links as:

G = RTρ/Mc (28)

where R is the gas constant, T is the temperature in kelvins, and ρ is the den-
sity. At some point, the curve begins to level off and this is often taken as the
vitrification point, Tvf .
The vitrification point is where the cure rate slows because the material has
become so viscous that the bulk reaction has stopped. At this point, the rate of
cure slows significantly. The apparent Tvf however is not always real: Any ana-
lyzer in the world has an upper force limit. When that force limit is reached, the
“topping out” of the analyzer can pass as the Tvf . The use of a combined tech-
nique like DMA-DEA (80) to see the higher viscosities or the removal of a sample
from parallel plate and sectioning it into a flexure beam is often necessary to see
the true vitrification point. A reaction can also completely cure without vitrifying
and will level off the same way. One should be aware that reaching vitrification or
complete cure too quickly could be as bad as too slowly. Often a overly aggressive
cure cycle will cause a weaker material as it does not allow for as much network
development, but gives a series of hard (highly cross-linked) areas among softer
(lightly cross-linked) areas.
On the way to vitrification, an important value is 106 Pa s. This is the vis-
cosity of bitumen (81) and is often used as a rule of thumb for where a material
is stiff enough to support its own weight. This is a rather arbitrary point, but
is chosen to allow the removal of materials from a mold and the cure is then
continued as a postcure step. As will be seen below, the postcure is often a vital
part of the curing process.
The cure profile is both a good predictor of performance as well as a sensi-
tive probe of processing conditions. A final note on cure profiles should be that a
volume change occurs during the cure (82). This shrinkage of the resin is impor-
26 DYNAMIC MECHANICAL ANALYSIS

tant and can be studied by monitoring the probe position of some DMAs as well
as by TMA and dilatometry.

4.3.2. Photocuring. A photocure in the DMA is run by applying a UV

light source to a sample that is held a specific temperature or subjected to a spe-
cific thermal cycle (83). Photocuring is done for dental resin, contact adhesives,
and contact lenses. UV exposure studies are also run on cured and thermoplastic
samples by the same techniques as photocuring to study UV degradation. The
cure profile of a photocure is very similar to that of a cake or epoxy cement. The
same analysis is used and the same types of kinetics developed as is done for
thermal curing studies.
The major practical difficulty in running photocures in the DMA is the
current lack of a commercially available photocuring accessory, comparable to
the photocalorimeters on the market. Several manufacturers now offer standard
accessories to do UV curing in the DMA, and these attachments can often be
adapted for Near Infrared (NIR) or Raman testing. The University of Colorado
Dental School has adapted a DMA to allow simultaneous exposure to UV as well
as monitoring by NIR (84), allowing the correlation of the chemical cure and me-
chanical properties.

4.3.3. Modeling Cure Cycles. The above discussions are based on using
a simple temperature ramp to see how a material responds to heating. In ac-
tual use, many thermosets are actually cured using more complex cure cycles to
optimize the trade-off between the processing time and the final product’s proper-
ties (85). The use of two stage cure cycles is known to develop stronger laminates
in the aerospace industry. Exceptionally thick laminates often also require mul-
tiple stage cycles to develop strength without porosity. As thermosets shrink on
curing, careful development of a proper cure cycle to prevent or minimize internal
voids is necessary.
One reason for the use of multistage cures is to drive reactions to comple-
tion. Another is to extend the minimum viscosity range to allow greater control in
forming or shaping of the material. The development of a cure cycle with multiple
ramps and holds would be very expensive if done with full sized parts in produc-
tion facilities. The use of the DMA gives a faster and cheaper way of optimizing
the cure cycle to generate the most efficient and tolerant processing conditions.

4.3.4. Isothermal Curing Studies. Often curing is done at a constant

temperature for a period of time. This is how the data needed for the kinetic
models discussed in the next section are normally collected. It is also how rubber
samples are cross-linked, how initiated reactions are run, and how bulk polymer-
izations are performed. Industrially, continuous processes, as opposed to batch,
often require an isothermal approach. UV light and other forms of nonthermal
initiation also use isothermal studies for examining the cure at a constant tem-
perature.

4.3.5. Curing Kinetics by DMA. Several approaches have been developed

to studying the chemorheology of thermosetting systems. MacKay and Halley (86)
recently reviewed chemorheology and the more common kinetic models. A funda-
 DYNAMIC MECHANICAL ANALYSIS 27

mental method is the Williams–Landel–Ferry model (87), which looks at the vari-
ation of Tg with degree of cure. This has been used and modified extensively (88).
A common empirical model for curing has been proposed by Roller (89). In the lat-
ter approach, samples of the thermoset are run isothermally as described above
and the viscosity versus time data collected. This is plotted as log η* versus time
in seconds, where a change in slope is apparent in the curve. This break in the
data indicates the sample is approaching the gel time. From these curves, the
initial viscosity, ηo , and the apparent kinetic factor, k, can be determined. By plot-
ting the log viscosity versus time for each isothermal run, the slope, k, and the
viscosity at t = 0 are apparent. The initial viscosity and k can be expressed as

ηo = η∞ e Eη/RT
(29)

k = k∞ e Ek/RT
(30)

Combining these allows to set up of the equation for viscosity under isothermal
conditions as

ln η (t) = ln η∞ + Eη/RT + tk∞ e Ek/RT

(31)

By replacing the last term with an expression that treats temperature as a func-
tion of time, the equation becomes
t
ln η (T, t) = ln η∞ + Eη/RT + k∞ e Ek/RT
dt (32)
0

This equation can be used to describe viscosity–time profiles for any run where
the temperature can be expressed as a function of time. The activation energies
can now be calculated. The plots of the natural log of the initial viscosity (de-
termined above) versus 1/T and the natural log of the apparent rate constant, k,
versus 1/T are used to give us the activation energies, Eη and Ek . Compari-
son of these values to the k and E to those calculated by DSC shows that this
model gives larger values (70). The DSC data are faster to obtain, but they do not
include the needed viscosity information.
Several corrections have been proposed, addressing different orders of reac-
tion (90) (the above assumes first order) and modifications to the equations (92).
Many of these adjustments are reported in Roller’s 1986 review (92) of curing
kinetics. It is noted that these equations do not work well above the gel tempera-
ture. This same equation has been used to predict the degradation of properties
in thermoplastics successfully (93).

4.3.6. Mapping Thermoset Behavior: The Gillham–Enns Diagram. An-

other approach to attempt to fully understand the behavior of a thermoset was
developed by Gillham and co-workers (94) and is analogous to the phase di-
agrams used by metallurgists. The time–temperature–transformation diagram
(TTT) or the Gillham–Enns diagram (after its creators) is used to track the ef-
fects of temperature and time on the physical state of a thermosetting material.
28 DYNAMIC MECHANICAL ANALYSIS

GEL

Fu
Tg∞

ll
RUBBER CHAR

cu
Vitr

re
ific
SOL–GEL atio
n
RUBBER

Temperature of curing
Devitrifi
cation
GEL GL
ASS
Ge Tg – T
la
tio V g∞
n itr
ifi
ca
tio
n SOL–GEL GLASS

LIQUID

Tg =
gel T
g

Tggel SOL GLASS

Tg Log time
0

Fig. 21. Gillham–Enns diagram.

Figure 21 shows an example. Running isothermal studies of a resin at various

temperatures and recording the changes as a function of time can do this. One
has to choose values for the various regions, and Gillham has done an excellent
job of detailing how one picks the Tg , the glass, the gel, the rubbery, and the
charring regions (95). These diagrams are also generated from DSC data (96),
and several variants (97), such as the continuous heating transformation and
conversion–temperature–property diagrams, have been reported. Surprisingly
easy to do, although a bit slow, they have not yet been accepted in industry despite
their obvious utility. A recent review (98) will hopefully increase the use of this
approach.

Symbols, Abbreviations, and Acronyms

Ea Activation energy
E* Complex Modulus
E Storage Modulus
E Loss Modulus
f Frequency
G Shear modulus
J Compliance
k Deformation
k Rate constant
Me Entanglement molecular weight
Mc Molecular weight between cross-links
Mw Molecular weight
T Period
 DYNAMIC MECHANICAL ANALYSIS 29

Tα,β,γ Transition (subscript type)

vf Free volume
 Logarithmic decrement
γ Shear strain
γ̇ Shear strain rate
δ Phase angle
tan δ Tangent of the phase angle, also called the damping
ε Tensile strains
ε̇ Strain rate
η Viscosity
η* Complex viscosity
η Storage viscosity
η Loss viscosity
Torque
ρ Density
σ Stress
ω Frequency, Hz
DEA Dielectric analysis or analyzer
DSC Differential scanning calorimeter
DMA Dynamical mechanical analysis or analyzer
DMTA Dynamic mechanical thermal analysis or analyzer
TBA Torsional braid analyzer
TGA Thermogravimetric analyzer

Acknowledgments

The authors wish to acknowledge the help and support of Mettler-Toledo NA,
especially B. Weddle, and the Laboratory for Advanced Materials and Optimized
Processing at the University of North Texas, especially Prof. W. Brostow.

BIBLIOGRAPHY

1. N. McCrum, B. Williams, and G. Read, Anelastic and Dielectric Properties of Polymeric

Solids, Dover Publications, N. Y., 1991, pp. 192–200.
2. (a) K. Menard, Dynamic Mechanical Analysis: A Practical Introduction, CRC Press,
Boca R aton, Fla., 1999; (b) M. Sepe, Dynamic Mechanical Analysis for Plastic Engi-
neering, Plastic Design Library, New York, 1998; (c) T. Murayama, Dynamic Mechani-
cal Analysis of Polymeric Materials, Elsevier Science Publishing Co., New York, 1977.
3. J. H. Poyntang, Proc. R. Soc., Series A 82, 546 (1909).
4. A. Kimball and D. Lovell, Trans. Am. Soc. Mech. Eng. 48, 479 (1926).
5. K. te Nijenhuis, in G. Astarita, G. Marruci, and L. Nicolais, ed., Rheology Volume 1:
Principles. Plenum Press, New York, 1980, pp. 263.
6. M. L. Miller, The Structure of Polymers, Reinhold Publishing, New York, 1966.
7. J. Dealy, Rheometers for Molten Plastics, Van Nostrand Reinhold, New York, 1992, pp.
136–137 and 234–236.
8. J. Ferry, Viscoelastic Properties of Polymers, 3rd ed., John Wiley & Sons, New York,
1980.
30 DYNAMIC MECHANICAL ANALYSIS

9. N. McCrum, B. Williams, and G. Read, Anelastic and Dielectric Effects in Polymeric

Solids, Dover Publications , New York, 1991.
10. J. Gilham and J. Enns, Trends Polym. Sci. 2, 406 (1994).
11. C. Macosko and J. Starita, SPE J. 27, 38 (1971).
12. T. Murayama, Dynamic Mechanical Analysis of Polymeric Materials, Elsevier Science
Publishing Co., New York, 1977.
13. B. E. Read and G. D. Brown, The Determination of the Dynamic Properties of Polymers
and Composites, John Wiley & Sons, New York, 1978.
14. American Standards and Test Methods: http://www.astm.org/
15. (a) J. Gillham, in J. Dworkins, ed., Developments in Polymer Characterizations: Volume
3, Applied Science Publisher, Princeton, N.J., 1982, pp. 159–227; (b) J. Gillham and J.
Enns, Trends Polym. Sci. 2(12), 406 (1994).
16. U. Zolzer and H-F. Eicke, Rheol. Acta 32, 104, (1993).
17. N. McCrum, B. Williams, and G. Read, Anelastic and Dielectric Properties of Polymeric
Solids, Dover Publications, New York, 1991, pp. 192–200.
18. P. Flory, Principles of Polymer Chemistry, Cornell University Press, I thaca, N.Y., 1953.
19. R. Bird, C. Curtis, R. Armstrong, and O. Hassenger, Dynamics of Polymer Fluids, 2nd
ed., John Wiley & Sons, New York, 1987.
20. (a) J. D. Ferry, in Ref. 8; (b) J. J. Aklonis and W. J. McKnight, Introduction to Polymer
Viscoelasticity, 2nd ed., John Wiley & Sons, New York, 1983.
21. (a) L. C. E. Struik, Physical Aging in Amorphous Polymers and Other Materials, El-
sevier Science Publishing Co., New York, 1978; (b) L. C. E. Struik, in W. Brostow and
R. D. Corneliussen, eds., Failure of Plastics, Hanser Publications, New York, 1986;
(c) S. Matsuoka, in W. Brostow and R. D. Corneliussen, eds. Failure of Plastics, Hanser
Publications, New York, 1986; (d) S. Matsuoka, Relaxation Phenomena in Polymers,
Hanser Publications, New York, 1992.
22. J. D. Vrentas, J. L. Duda, and J. W. Huang, Macromolecules 19, 1718 (1986).
23. W. Brostow and M. A. Macip, Macromolecules 22(6), 2761 (1989).
24. N. McCrum, G. Williams, and B. Read, Anelastic and Dielectric Effects in Polymeric
Solids, Dover Publications , New York, 1967.
25. M. Doi and S. Edwards, The Dynamics of Polymer Chains, Oxford University Press,
New York, 1986.
26. A. Aklonis and J. Knight, Introduction to Viscoelasticity, John Wiley & Sons, New York,
1983.
27. C. L. Rohn, Analytical Polymer Rheology, Hanser-Gardener Publications, New York,
1995.
28. J. Heijboer, Int. J. Polym. Mater. 6, 11 (1977).
29. N. McCrum, G. Williams, and B. Read, Anelastic and Dielectric Effects in Polymeric
Solids, Dover Publications, New York, 1967.
30. R. F. Boyer, Polym. Eng. Sci. 8 (3), 161 (1968).
31. C. L. Rohn, in Ref. 21, pp. 279–283.
32. (a) J. Heijboer, Int. J. Polym. Mater. 6, 11 (1977); (b) M. Mangion and G. Johari, J.
Polym. Sci, Part B: Polym. Phys. 29, 437 (1991); (c) G. Johari, G. Mikoljaczak, and J.
Cavaille, Polymer 28, 2023 (1987); (d) S. Cheng et al., Polym. Sci. Eng. 33, 21 (1993);
(e) G. Johari, Lect. Notes Phys. 277, 90 (1987); (f) R. Daiz-Calleja and E. Riande, Rheol.
Acta 34, 58 (1995); (g) R. Boyd, Polymer 26, 323 (1985); (h) V. Bershtien, V. Egorov, L.
Egorova, and V. Ryzhov, Thermochim. Acta 238, 41 (1994).
33. B. Twombly, Proc. NATAS 20, 63 (1991).
34. (a) C. L. Rohn, Analytical Polymer Rheology, Hanser: Cincinnati, (1995), 185–210;
(b) J. Heijboer, Int. J. Polym. Mater. 6, 11 (1977).
35. R. Boyer, Polym. Eng. Sci. 8(3), 161 (1968).
 DYNAMIC MECHANICAL ANALYSIS 31

36. V. Bershtien and V. Egorov, Differential Scanning Calorimetery in the Physical Chem-
istry of Polymers, Ellis Horwood, Chich ester, UK, 1993.
37. B. Coxton, personal communication , 1998.
38. S. Cheng et al., Polym. Sci. Eng. 33, 21 (1993).
39. G. Johari, G. Mikoljaczak, and J. Cavaille, Polymer 28, 2023 (1987).
40. M. Mangion and G. Johari, J. Polym. Sci, Part B: Polym. Phys. 29, 437 (1991).
41. (a) F. C. Nelson, Shock Vib. Dig. 26, 11 (1994); (b) F. C. Nelson, Shock Vib. Dig. 26, 24
(1994).
42. W. Brostow, personal communication, 2004.
43. R. Boyer, Polym. Eng. Sci. 8(3), 161 (1968).
44. (a) G. Johari, Lect. Notes Phys. 277, 90 (1987); (b) J. Heijboer, Int. J. Polym. Mater. 6, 11
(1977); (c) J. Heijboer, in J. Prins, ed., Physics of Non-Crystalline Solids, Interscience,
New York, 1965; (d) J. Heijboer, J. Polym. Sci. C 16, 3755 (1968); (e) L. Nielsen, J.
Macromol. Sci., Phys. 9, 239 (1974).
45. A. Yee and S. Smith, Macromolecules 14, 54 (1981).
46. G. Gordon, J. Polym. Sci. A2 9, 1693 (1984).
47. J. Wendorff and B. Schartel, Polymer 36, 899 (1995).
48. R. H. Boyd, Polymer 26, 323 (1985).
49. (a) I. Noda, Appl. Spectrosc. 44(4), 550 (1990); (b) V. Kien, Proc. 6th Sympos. Radiat.
Chem. 6, 463 (1987).
50. L. H. Sperling, Introduction to Physical Polymer Science, 2nd ed., John Wiley & Sons,
New York, 1992.
51. C. Macosko, Rheology, VCH Publishers, New York, 1994.
52. (a) F. Quinn and T. Hatakeyama, Thermal Analysis, John Wiley & Sons, New York,
1994; (b) B. Wunderlich, Thermal Analysis, Academic Press, New York, 1990.
53. (a) J. Schawe, Thermochim. Acta 261, 183 (1995); (b) J. Schawe, Thermochim. Acta 260,
1 (1995); (c) J. Schawe, Thermochim. Acta 271, 1 (1995); (d) B. Wunderlich, A. Boller,
and Y. Yin, J. Therm. Anal. 42, 949 (1994); (e) P. Cortes, I. Fraga, Y. Calventus, F.
Roman, J. Hutchinson, and F. Ferrando. Materials 7, 1830–1849 (2014); (f) A. Hammer,
ed., Applications of Thermal Analysis, Mettler Toledo Inc., Greifensee, Switzerland,
2012, pp. 12–14.
54. (a) R. H. Boyd, Polymer 26, 323 (1985); (b) R. H. Boyd, Polymer 26, 1123 (1985).
55. (a) S. Godber, personal communication, 2001; (b) M. Ahmed, Polypropylene Fiber- Sci-
ence and Technology, Elsevier Science, New York, 1982.
56. A. Lobanov and S. Frenkel, Polym. Sci. USSR 22, 1150 (1980).
57. (a) R. Boyer, J. Polym Sci., Part B: Polym. Phys. 30, 1177 (1992); (b) J. K. Gilham et al.,
J. Appl. Polym. Sci. 20, 1245 (1976); (c) J. B. Enns and R. Boyer, Encycl. Polym. Sci.
17, 23–47 (1989).
58. V. Bershtien, V. Egorov, L. Egorova, and V. Ryzhov, Thermochim. Acta. 238, 41 (1994).
59. C. M. Warner, MS thesis, Central Michigan State University, 1988.
60. (a) B. Cassel, PETAN #69 DDSC, Perkin-Elmer, Norw alk, Conn., 1995; (b) W. Sichina,
Proc. NATAS 23, 137 (1994); (c) B. Wunderlich, A. Boller, and Y. Jin, J. Therm. Anal
1994, 42, 307 (1994); (d) B. Wunderlich, Modulated DSC, Univesity of Tennessee.,
Knoxvil le, Tenn., 1994.
61. (a) J. Dealy, Melt Rheology and its Role in Plastic Processing, Van Nostrand Reinhold,
Toronto, Canada, 1990; (b) N. Chereminsinoff, An Introduction to Polymer Rheology
and Processing, CRC Press, Boca Rat on, Fla., 1993.
62. (a) E. Turi, ed., Thermal Characterization of Polymeric Materials, Academic Press,
Bosto n, Mass., 1981; (b) E. Turi, ed., Thermal Analysis in Polymer Characterization,
Heydon, London, 1981.
63. C. Rohn, in Analytical Polymer Rheology, Hanser: Cincinnati, (1995), p. 203.
64. M. Miller, The Structure of Polymers, Reinhold, New York, 1966, pp. 611–612.
32 DYNAMIC MECHANICAL ANALYSIS

65. S. Rosen, Fundamantal Principles of Polymeric Materials, Wiley Interscience, New

York, 1993, pp. 53–77 and 258–259.
66. (a) W. Cox and E. Merz, J. Polym. Sci. 28, 619 (1958); (b) P. Leblans, H. Booj, J. Palmer,
and G. Tiemersma-Thone, J. Polym. Sci. 21, 1703 (1983).
67. W. Gleissele, in G. Astarita, G. Marruci, and L. Nicolais, eds., Rheology, Volume 2,
Plenum Press, New York, 1980, pp. 457.
68. D. W. Van Krevelin, Properties of Polymers, Elsevier Science, New York, 1987, p. 289.
69. C. Macosko, Rheology, VCH Publishers, New York, 1996, pp. 120–127.
70. (a) L.-H. Lee, ed., Adhesive Bonding, Plenum Press, New York, 1991; (b) L.-H. Lee, ed.,
Fundamentals of Adhesion, Plenum Press, New York, 1991.
71. (a) G. Martin, A. Tungare, and J. Gotro, Polymer Characterization, ACS, Washington
D. C., 1990; (b) M. Ryan, ANTEC Proc. 31, 187 (1973); (c) C. Gramelt, Am. Lab. Jan-
uary, 26 (1984); (d) S. Etoh, SAMPE J. 3, 6 (1985); (f) F. Hurwitz, Polym. Compos. 4(2),
89 (1983).
72. (a) M. Roller, Polym. Eng. Sci. 19, 692 (1979); (b) M. Roller, and J. Gilham, J. Coating
Technol. 50, 57 (1978).
73. (a) F. Warren, P. Royall, P. Butterworth, and P. Ellis, Carbohydrate Polymers 90, 628
(2012); (b) P. Royall, C. Huang, S. Tang, J. Duncan, and M. Brown, Int. J. Pharm 301,
181 (2005); (c) R. Ou, Y. Xie, Q. Wang, S. Sui, and M. Wolcott, Holzforschung 69, 223
(2015).
74. (a) M. Heise, Polym. Eng. Sci. 30, 83 (1990); (b) K. O’Driscoll, J. Polym. Sci.: Polym.
Chem. 17, 1891 (1979); (c) O. Okay, Polymer 35, 2613 (1994).
75. M. Hiese and G, Martin, J. Gotro, Polymer Eng. Sci. 30 (2), 83 (1990).
76. K. Wissbrun, D. Rhun, and M. Hannon, J. Coating Technol. 48, 42 (1976).
77. (a) F. Champon, J. Rheol. 31, 683 (1987); (b) H. Winter, Polym. Eng. Sci. 27, 1698
(1987); (c) C. Michon, G. Culvelier, and B. Launay, Rheol. Acta 32, 94 (1993).
78. C. Michon, G. Culvelier, and B. Launay, Rheol. Acta 32, 94 (1993).
79. I. Kalnin and K. Hollands, Epoxy Resins, ACS, Washington D.C., 1970.
80. DEA is dielectric analysis, where an oscillating electrical signal is applied to a sample.
From this signal, the ion mobility can be calculated, which is then converted to a
viscosity (see McCrum for details). DEA will measure to significantly higher viscosities
than DMA.
81. H. Barnes, J. Hutton, and K. Walters, An Introduction to Rheology, Elsevier Science
Publishing, Co., New York, 1989.
82. A. W. Snow, and J. Armistead, J. Appl. Polym. Sci. 52, 401, 1994.
83. (a) T. Renault, NATAS Notes 25, 44 (1994); (b) H. L. Xuan and C. Decker, J. Polym.
Sci.: Part A 31, 769 (1993); (c) W. Shi and B. Ranby, J. Appl. Polym. Sci. 51, 1129
(1994).
84. (a) J. Stansbury, H. Lu, J. Nie, and C. Bowman, Dental Mater. 25, 33–38 (2009); (b) C.
Bowman, H. Kilambi, H. Lu, P. Shah, and J. Stansbury, Biomaterials 26, 1329–1336
(2005).
85. (a) R. Geimer, R. Follens, J. Kowtsky, A. Christensen, and G. Myers, J. Appl. Polym.
Sci. 47, 1481 (1993); (b) R. Roberts, SAMPE J. 5, 28 (1987).
86. P. J. Halley and M. E. MacKay, Polym. Eng. Sci. 36(5), 593 (1996).
87. J. Ferry, Viscoelastic Properties of Polymer, Wiley and Sons: New York, (1980) 266–
330.
88. (a) J. Mijovic, J. Comp. Met. 23, 163 (1989); (b) J. Mijovic, SAMPE J. 23, 51 (1990).
89. (a) M. Roller, Metal Finishing 78, 28 (1980); (b) M. Roller, ANTEC Proc. 24, 9 (1978);
(c) J. Gilham, ACS Symp. Series 78, 53 (1978); (d) M. Roller, ANTEC Proc. 21, 212
(1975); (e) M. Roller, Polym. Eng. Sci. 15, 406 (1975); (f) M. Roller, Polym. Eng. Sci. 26,
432 (1986).
90. C. Rohn, Problem Solving for Thermosetting Plastics, Rheometrics, Austin, Tex., 1989.
 DYNAMIC MECHANICAL ANALYSIS 33

91. (a) J. Seferis, M. Dusi, and C. May, Chemorheology of Thermosetting Polymers, ACS,
Washington, D.C. 1983, pp., 301; (b) R. Patel and S. Patel, J. Therm. Anal. 39, 229
(1993).
92. M. Roller, Polym. Eng. Sci. 26, 432 (1986).
93. M. Roller, personal communication 1998.
94. (a) J. Gillham and N. Schneider, Polym. Compos. 1, 97 (1980); (b) J. Enns, and
J. Gilham, J. Appl. Polym. Sci. 28, 2567 (1983); (c) L. C. Chan, N. Noe, and J. Gilham, J.
Appl. Polym. Sci. 29, 3307 (1984); (d) J. Gillham, Polym. Eng. Sci. 26, 1429 (1986); (e)
S. Simon, J. Gilham, J. Appl. Polym. Sci. 51, 1741 (1994); (f) G. Palmese, J. Gilham, J.
Appl. Polym. Sci. 34, 1925 (1987); (g) J. Enns, and J. Gilham, in C. Craver, ed., Polymer
Characterization, ACS, Washington, D.C., 1983.
95. (a) J. Gillham and S. Simon, J. Appl. Polym. Sci. 53, 709 (1994); (b) J. Enns and J.
Gillham, Trends Polym. Sci. 2, 406 (1994).
96. A. Otero, M. Clavaguera-Mora, S. Surinach, M. Baro, and N. Clavaguera, Thermochim.
Acta 203, 379 (1992).
97. (a) J. Gillham and G. Wisenrakkit, J. Appl. Polym. Sci. 42, 2453 (1991); (b) B. Osinski,
Polymers 34, 752 (1993).
98. J. Enns and J. Gillham, Trends Polym. Sci. 2, 406 (1994).

KEVIN P. MENARD
Mettler Toledo, Columbus, OH, USA

NOAH R. MENARD
Veritas Testing and Consulting LLC, Denton, TX,
USA