Polymer Testing 20 (2001) 199–204

www.elsevier.nl/locate/polytest
Data Interpretation
The glass transition temperature Tg of polymers—
Comparison of the values from differential thermal analysis
(DTA, DSC) and dynamic mechanical measurements
(torsion pendulum)
*
Jens Rieger
BASF Aktiengesellschaft, Polymer Physics/Solid State Physics, 67056 Ludwigshafen, Germany

Received 6 January 2000; accepted 3 March 2000

Abstract

It is often maintained that the glass transition temperatures determined by dynamic mechanical thermal analysis
(DMTA) are higher than the corresponding glass transition temperatures measured by differential thermal analysis
(DTA, DSC). Examination of a large number of polymeric systems revealed that this statement is not generally true
when a torsion pendulum with a frequency of measurement of 1 Hz is used as the DMTA instrument. On the contrary,
it is found on average that the relationship Tg,TP⬇Tg,DTA holds.  2000 Elsevier Science Ltd. All rights reserved.

Keywords: Glass transition temperature; Thermal analysis; Dynamic mechanical analysis

1. Introduction approaching the glass transition temperature from high

temperatures the molecules or segments of molecules
The glass transition temperature Tg is an important increasingly impede one another’s movement and the
physical characteristic of plastics. Amorphous polymers viscosity rises rapidly until the material finally solidifies.
such as polystyrene, poly(methyl methacrylate), polybut- The consequence of this is that the measured glass tran-
adiene, etc. are hard and brittle at temperatures below sition temperature depends on the thermal prehistory of
Tg. On heating to temperatures above Tg they become the sample and on the physical method by which it has
elastic like rubber or liquefy depending on their molecu- been determined. The fact that these two factors are often
lar weight and/or degree of cross-linking. In the case of not adequately clarified when Tg values are quoted
semicrystalline materials (polyethylene, polypropylene, occasionally results in uncertainties which can have far-
polyamide, etc.) Tg is well below the melting point and reaching consequences, especially in the case of patent
marks the temperature range in which on cooling the proceedings.
brittleness increases markedly. The present understand- The most common methods for determining the glass
ing of the glass transition has been discussed at length transition temperature are differential thermal analysis
in the literature [1–9]. (DTA, DSC) and dynamic mechanical thermal analysis
In contrast to fusion and crystallization phenomena, (DMTA). In what follows glass transition temperatures
the glass transition is not a phase transition as defined that were determined by both methods are compared.
in thermodynamics. It is rather a kinetic effect. When
2. Glass transition temperature in torsion
pendulum and DTA-experiments
* Fax: +49-621-92281.
The torsion pendulum belongs to the group of
E-mail address: [email protected] (J. Rieger).
dynamic mechanical test methods. A disk with a known

0142-9418/01/$ - see front matter  2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 2 - 9 4 1 8 ( 0 0 ) 0 0 0 2 3 - 4
200 J. Rieger / Polymer Testing 20 (2001) 199–204

moment of inertia connected to the test specimen via a the temperature at which tan d attains its maximum value
rod is deflected by about 1° at the start of a measurement is very common in the literature. There are probably two
and subsequently performs an oscillation which is reasons for this. First, at the time when measurements
damped by the sample. In the case of the torsion pendu- were still evaluated manually it was simpler to calculate
lum, the moment of inertia is usually chosen in such a tan d⬇⌳/p directly from the amplitudes of oscillation.
way that the frequency of measurement is 1 Hz. The Secondly, maxima in ⌳ can sometimes be more easily
damping of the oscillation depends on the shear modulus analyzed than the corresponding effects in G⬘ and G⬙.
and loss modulus (G⬘, G⬙) of the sample. A pair of values There are, however, some arguments in favor of the
for G⬘ and G⬙ are obtained by analyzing the oscillation definition of the glass transition temperature by the
curve at a particular temperature. This procedure is car- maximum value of G⬙ rather than by the maximum value
ried out repeatedly over a defined temperature range. The of tan d.
result produced is a thermogram of G⬘ and G⬙ as a func-
tion of the temperature. The shear modulus G⬘ is a meas-
1. G⬙ is a measure of the dissipated energy. The tem-
ure of the rigidity of a material. The glass transition tem-
perature of maximum loss (maximum of G⬙) is the
perature can be read off from the position of the peak
transition temperature in question. The temperature at
of the loss modulus G⬙, as is discussed below. By com-
which the ratio of stored and dissipated energy exhib-
parison with DTA the torsion pendulum has the advan-
its a maximum is from the physical point of view
tage that in multiphase systems even very small amounts
only a derived variable.
(of a few percent) of a component can still be detected.
2. The temperature at which G⬙ for a substance attains
Review articles on the measurement of relaxation and
its maximum value does not depend on whether the
glass transitions using DMTA methods have been
material is pure or is present in a mixture (polymer
presented by Illers [10] and Boyer [7]. The measurement
blend). The maximum value of tan d is different. In
parameters for the torsion pendulum method are laid
pure form the corresponding temperature, dependent
down in the German norm DIN 53445. The data
on the decrease in G⬘, is at a higher temperature than
presented here have been recorded in accordance with
when the corresponding component has been mixed
this standard using an instrument from the Myrenne
into a system which exhibits further glass or fusion
company.
transitions at higher temperatures.
Figure 1(a) and Fig. 2(a) show by way of example the
3. In a more recent study Peleg demonstrates that the
temperature dependencies of the loss modulus G⬙ and of
temperature at which tan d is maximum, especially in
the logarithmic decrement ⌳ of a latex film based on
the case of smeared glass transitions, does not
acrylate and of a mixture of polyaryl ether ketone and
coincide with the softening temperature at which the
polyether sulfone, respectively, recorded with the torsion
shear or tensile modulus (G⬘, E⬘) fall to low values
pendulum. ⌳ is defined by the logarithm of the quotient
[11].
of the amplitudes of two successive oscillations. If damp-
ing of the material is not too strong the relationship
⌳=pG⬙/G⬘ holds. The phase angle d is defined by tan From what has been said it follows that it is practical to
d=G⬙/G⬘ and so one obtains the relationship tan d⬀⌳. use the temperature at which G⬙ exhibits a local
For the torsion pendulum experiment the glass tran- maximum as the definition of the glass transition tem-
sition temperature can be defined in principle in three perature. Illers and Baur came to this conclusion as early
ways: as the 1970s but they were unable to influence the prac-
tice of defining Tg by way of tan d. Boyer considered
1. The temperature at which G⬘ has fallen to a certain that the two former BASF research scientists were right.
value; In his review article, however, he used Tg(tan d=max.)
2. The temperature at which tan d has its maximum in order to allow comparison of as many literature refer-
value; ences as possible [7].
3. The temperature at which G⬙ has its maximum value. In the DTA thermogram the glass transition appears
as a smeared transition (see Fig. 1(b) and Fig. 2(b)).
The first definition is sometimes used for some classes Here, in accordance with the German norm DIN 53765,
of product since it takes the softening of the material the glass temperature is defined as the temperature at half
into account in a direct manner. In general, however, the the height of the step (midpoint temperature, Tg,midpoint).
definition is not suitable since it cannot be used for To determine this temperature straight lines are extended
example in the case of polymer blends (the decrease in along the left-hand and right-hand branches of the heat
G⬘ for the component having the lower glass transition flow curve. Tg is obtained from the point of intersection
temperature is only weakly exhibited when it is present of the bisecting line of the angle with the measured
in small amounts). curve. An extensive discussion of the glass transition in
The definition of the glass transition temperature as DTA experiments is provided in [12].
 J. Rieger / Polymer Testing 20 (2001) 199–204 201

Fig. 1. Thermograms recorded for an acrylate-based dispersion film using the torsion pendulum (a) and DTA (b).

3. Comparison of data from DTA and torsion uncertainty in the specification of Tg values according to
pendulum DTA typically lies between 1 and 5 K depending on the
width of the specific glass transition. To calibrate the
In order to examine how the glass temperatures which torsion pendulum a thermocouple was fused into a test
were determined by DTA or torsion pendulum relate to specimen. The temperature values from the thermo-
one another examples were selected from the numerous couple were compared with the simultaneous tempera-
samples studied in our laboratory as typical representa- ture display on the control unit of the torsion pendulum
tives of all classes of polymers. These involve for at the usual heating rate of 1 K/min. The uncertainty
example impact-resistant modified polystyrenes, poly- in the specification of Tg values from torsion pendulum
esters, HT blends, dispersion films, etc. measurements amounts to about 1–2 K. For the DTA
The DTA/DSC instruments employed were calibrated and torsion pendulum experiments samples having a
in the usual manner by means of the melting points of comparable thermal prehistory were prepared.
mercury (⫺38.9°C) and tin (231.8°C) using a heating For reasons of space, not all data records are presented
rate of 20 K/min. The sample size was about 15 mg. The here. Instead of this Fig. 3 shows the frequency with
202 J. Rieger / Polymer Testing 20 (2001) 199–204

Fig. 2. Thermograms recorded for a blend of polyaryletherketone and polyethersulfone using the torsion pendulum (a) and DTA (b).

which a certain temperature difference was measured the temperature of the maximum value of G⬙ and the
between the glass transition temperatures from DTA and temperature of the maximum for tan d=G⬙/G⬘ is related
from the maximum value of G⬙. It can be seen that on to the width of the glass transition determined from the
average Tg,DTA and Tg,G⬙ agree to within a few Kelvins. DTA data must also be answered in the negative. This,
In individual cases, however, differences of up to ±10 K however, is not in the least surprising. As was made clear
and more are observed. The correlation between Tg,DTA above, the temperature of the position of the peak in tan
and Tg,tan d is distinctly less marked (see Fig. 4). It may d depends on whether the corresponding component is
already be concluded at this point that the statement that incorporated into a blend in which still higher transition
Tg,tan d and Tg,G⬙ are always greater than Tg,DTA is not true. temperatures are observed. If that is the case the tem-
Further analysis of the data yielded that the gap perature of the maximum in G⬙ and the temperature of
between the glass transition temperature determined by the maximum value of tan d differ only slightly. If on
DTA and the value defined by the maximum value of the other hand we are dealing with a single-component
G⬙ is not correlated to the width of the glass transition material or with the highest transition temperature occur-
⌬Tg. The question as to whether the interval between ring in the system the two temperatures generally dif-
fer significantly.
 J. Rieger / Polymer Testing 20 (2001) 199–204 203

Fig. 3. Histogram indicating the frequency with which a certain temperature difference was found between the glass transition
temperatures from DTA and from the maximum value of G⬙ in the course of examining 83 polymer systems.

Fig. 4. Histogram indicating the frequency with which a certain difference was found between the glass transition temperatures
from DTA and from the maximum in tan d=G⬙/G⬘ in the course of examining 46 polymer systems.
204 J. Rieger / Polymer Testing 20 (2001) 199–204

4. Comparison with the literature References

Wetton states in a textbook on the characterization of [1] W. Burchard, M. Fischer, U.T. Kreibrich, R. Schmid, G.
Wegner, in: In H. Batzer (Ed.), Polymere Werkstoffe, vol.
polymers that Tg,G⬙ from DMTA experiments is always
I, Thieme Verlag, Stuttgart, 1985.
higher than Tg,DTA (see [6], p. 219; Wetton takes DMTA [2] M.J. Richardson, in: V.B.F. Mathot (Ed.), Calorimetry and
as meaning dynamic mechanical analysis in the fre- Thermal Analysis of Polymers. Hanser Verlag, Munich,
quency range of 0.01 Hz to several hundred Hz). The 1994.
evaluation of our data in Fig. 3 demonstrates that this [3] M.J. Richardson, in: C. Booth, C. Price (Eds.), Compre-
statement is not true. hensive Polymer Science, vol. I, Pergamon Press, Oxford,
Aklonis and MacKnight express themselves more cau- 1989, p. 867.
tiously when they say that Tg,tan d and Tg,G⬙ are generally [4] B. Wunderlich, Thermal Analysis, Academic Press, Bos-
greater than Tg,DTA [13]. This statement too is misleading ton, 1990.
at a measurement frequency of 1 Hz. [5] L.H. Sperling, Introduction to Physical Polymer Science,
Wiley, New York, 1992.
Achorn and Ferrillo maintained on the basis of their
[6] R.E. Wetton, in: B.J. Hunt, M.I. James (Eds.), Polymer
own measurements on some polymers that Tg,DTA Characterisation. Blackie, London, 1993, p. 178.
coincides with the average value of Tg,tan d and Tg,E⬙ to [7] R.F. Boyer, in: In H.F. Mark, N.M. Bikales (Eds.), Encyc-
within 4 K [14,15]. On the basis of the results discussed lopedia of Polymer Science and Technology, Suppl. Vol.
we are unable to confirm this statement. II, Wiley, New York, 1977, p. 745.
Our finding that Tg,DTA⬇Tg,G⬙ is supported by sup- [8] C.A. Angell, Science 267 (1995) 1924.
plementary measurements of the tensile modulus in [9] B. Frick, D. Richter, Science 267 (1995) 1939.
which similar agreement between Tg,DTA and Tg,E⬙ was [10] K.H. Illers, Ber. Bunsenges. 70 (1966) 353.
found [16]. [11] M. Peleg, Rheol. Acta 34 (1995) 215.
In conclusion it must be emphasized that what is said [12] J. Rieger, J. Therm. Anal. 46 (1996) 965.
[13] J.J. Aklonis, W.J. MacKnight, in: Introduction to Polymer
here applies to the case of DMTA investigations carried
Viscoelasticity, 2nd ed. Wiley, New York, 1983, p. 60.
out at a frequency of 1 Hz. At higher frequencies it can [14] P.J. Achorn, R.G. Ferrillo, J. Appl. Polym. Sci. 54
generally be assumed that above measurement fre- (1994) 2033.
quencies of about 100 Hz the glass transition temperature [15] R.G. Ferrillo, P.J. Achorn, J. Appl. Polym. Sci. 64
determined by DMTA is greater than the value measured (1997) 191.
by DTA under normal conditions. [16] A. Zosel, BASF AG, internal report.