DSC Book

Differential Scanning Calorimetry
Springer-Verlag Berlin Heidelberg GmbH

G. W. H. Hahne, w. Hemminger,
H.- J. Flammersheim
Differential Scanning
Calorimetry
An Introduction for Practitioners
With 136 Figures and 13 Tables
i Springer
Dr. G. W. H. Hahne
Universitat Ulm
Sektion Kalorimetrie
Albert-Einstein-Allee 11
D - 89069 Ulm
Dr. W. Hemminger
Physikalisch-Technische Bundesanstalt
Bundesallee 100
D - 38116 Braunschweig
Dr. H.-J. Flammersheim

Universitat Jena
Institut fiir Physikalische Chemie
LessingstraBe 10
D - 07743 Jena
ISBN 978-3-662-03304-3 ISBN 978-3-662-03302-9 (eBook)

DOI 10.1 007/978-3-662-03302-9
Library of Congress Cataloging-in-Publication Data

Htlhne. G. W. H. (Giinther. Walther. Heinrich)
Differential scanning calorimetry: an introduction for practitioners / G. W. H. Hohne.
W. Hemminger. H.-J. Flammersheim.
Includes bibliographical references and index.
1. Calorimetry. 1. Hemminger, W. II. Flammersheim, H.-J., III. Title.

Q C291.H64 1995 95-20483
This work is subject to copyright. All rights are reserved. whether the whole or part of the
material is concerned. specifically the rights oftranslation. reprinting, reuse of illustrations.
recitation, broadcasting. reproduction on microfilm or in other ways. and storage in data
banks. Duplication of this publication or parts thereof is permitted only under the
provisions of the German Copyright Law of September 9. 1965. in its current version, and
permission for use must always be obtained from Springer-Verlag. Violations are liable for
prosecution act under German Copyright Law.
© Springer-Verlag Berlin Heidelberg 1996

Originally published by Springer-Verlag Berlin Heidelberg New York in 1996.
Softcover reprint of the hardcover 1st edition 1996
The use ofgeneral descriptive names. registered names. trademarks. etc. in this publication
does not imply. even in the absence of a specific statement. that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
Product liability: The publisher cannot guarantee the accuracy of any information about
dosage and application contained in this book. In every individual case the user must
check such information by consulting the rlevant literature.
Typesetting: Data-conversion by Fotosatz-Service Kohler OHG. Wiirzburg

SPIN:10077368 52/3020-543210 - Printed on acid-free paper
Preface
Differential Scanning Calorimetry (DSC) is a well established measuring method

which is used on a large scale in different areas of research, development, and
quality inspection and testing. Over a large temperature range, thermal effects can
be quickly identified and the relevant temperature and the characteristic caloric
values determined using substance quantities in the mg range. Measurement values
obtained by DSC allow heat capacity, heat of transition, kinetic data, purity and
glass transition to be determined. DSC curves serve to identify substances, to set up
phase diagrams and to determine degrees of crystallinity.
This book provides, for the first time, an overall description of the most impor-
tant applications of Differential Scanning Calorimetry. Prerequisites for reliable
measurement results, optimum evaluation of the measurement curves and esti-
mation of the uncertainties of measurement are, however, the knowledge of the
theoretical bases of DSC, a precise calibration of the calorimeter and the correct
analysis of the measurement curve.
The largest part of this book deals with these basic aspects: The theory of DSC
is discussed for both heat flux and power compensated instruments; temperature
calibration and caloric calibration are described on the basis of thermodynamic
principles. Desmearing of the measurement curve in different ways is presented as
a method for evaluating the curves of fast transitions.
The instrumental data which are most important for the characterization of Dif-
ferential Scanning Calorimeters are defined, and it is explained how they are
determined experimentally. This enables every potential buyer to ask the man-
ufacturer for measured characteristic data which will allow him to compare the dif-
ferent instruments available.
We are indebted to S.M. Sarge for a critical examination of the manuscript
and valuable suggestions for improvement and especially to M. J. Richardson for
helping us with the translation.
G. W. H. Hahne
w. Hemminger
H.-J Flammersheim
Contents
1 Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . .
2 Types of Differential Scanning Calorimeters 7
2.1 The Heat Flux DSC . . . . . . . . . . . . . . . 8

2.1.1 Heat Flux DSC with a Disk-Type Measuring System 8
2.1.2 Heat Flux DSC with a Cylinder-Type Measuring System 11
2.2 The Power Compensation DSC . . . . . . . . . . . . . . 14
3 Theoretical Fundamentals of Differential Scanning

Calorimeters . . . 21
3.1 The Heat Flux DSC 21

3.2 The Power Compensation DSC 38
4 Calibration of Differential Scanning Calorimeters 41
4.1 Aspects of Quality Assurance 41

4.2 Thermodynamic Aspects 42
4.3 Temperature Calibration 43
4.3.1 Calibration Procedure . . 45
4.3.2 Arguments in Support of the Temperature Calibration Procedure 52
4.3.3 Examples of Temperature Calibration 54
4.4 Caloric Calibration . . . . 61
4.4.1 Heat Flow Rate Calibration . . . 62
4.4.2 Peak Area Calibration . . . . . . 65
4.4.3 Examples of Caloric Calibrations 68
4.5 Conclusions Regarding the Calibration of DSCs 72
4.6 Calibration Substances . . . . . . . . . . . . . 73
4.6.1 Calibration Substances for Temperature Calibration 74
4~6.2 Calibration Substances for Heat Flow Rate Calibration 75
4.6.3 Calibration Substances for Peak Area Calibration . . . 76
VIII Contents
5 The DSC Curve 81
5.1 Characteristic Tenns 82

5.2 Influencing Parameters .. . . . . . . . . . . . . . . . . . 83
5.3 The Baseline and the Detennination of the Thennodynamic
Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.4 Desmearing ofthe DSC Curve . . . . . . . . . . . . . . . 89
5.4.1 Correction of the Temperature and Heat Flow Rate Indicated 91
5.4.2 Subtraction of the Baseline . . . . . . . . . . . . 92
5.4.3 Calculation of the True Flow Rate into the Sample 92
5.4.4 Advanced Desmearing 96
5.4.5 Further Calculations 102
5.5 Interpretation and Presentation 103
6 Applications of Differential Scanning Calorimetry . . . . . . . 105
6.1 General Applications Without Sophisticated Evaluation

of the Measured Curve . . . . . . . . . . . . . . 105
6.1.1 Identification of Substances, the Phase Behavior 105
6.1.2 Detennination of Phase Diagrams . . . . . . . . 111
6.1.3 Safety Aspects . . . . . . . . . . . . . . . . . . 113
6.1.4 Characterization of Polymers by Thennal Analysis;
Effects of Origin and Thennal History . . . . . . . . . . . . . 114
6.1.5 Detennination of the Degree of Crystallinity . . . . . . . . . . 115
6.2 Applications Requiring a Detailed Evaluation ofthe Measured
Curve . . . . . . . . . . . . . . . . . 118
6.2.1 Measurements ofthe Heat Capacity 118
6.2.1.1 The "Classical" Three Step Procedure 119
6.2.1.2 The "Absolute" Dual Step Method .. 124
6.2.1.3 A Variation of the Classical Technique 124
6.2.1.4 A Discontinuous Procedure . . . . . . 127
6.2.1.5 Detennination Using the Dual DSC 129
6.2.1.6 General Precautions for the Minimization of Erors 129
6.2.1.7 Typical Applications of Heat Capacity Measurements 130
6.2.1.8 Heat Capacity Measurements with Temperature Modulated DSC 132
6.2.2 The Investigation of Reactions .. 132
6.2.2.1 Detennination of Heats of rEaction 133
6.2.2.2 Kinetic Investigations . . . . . . . 139
6.2.2.2.1 Introduction and Definitions .. . 139
6.2.2.2.2 Experimental Preconditions for a Reliable Kinetic Analysis 141
6.2.2.2.3 Isothennal or Non-isothennal Reaction Mode? . . . . . . 143
6.2.2.2.4 Thennal Activation of the Sample or Activation by High Energy
Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
6.2.2.2.5 Remarks on the Influence of Different Phases . . . . . . . . . . 149
6.2.2.2.6 Evaluation of Overall Rate Laws and Fonnal Kinetic Parameters 149
Contents IX
6.2.2.2.7 Determination of the True Reaction Mechanism 152

6.2.2.2.8 Selected Examples . . . . . . . . . . . . . 153
6.2.3 The Glass Transition Process . . . . . . . . . . 160
6.2.3.1 The Phenomenology of the Glass Transition . . 160
6.2.3.2 The Nature ofthe Glass Transition and Consequences
for DSC-Measurements . . . . . . . . . . . . . . . . 161
6.2.3.3 Definition and Determination of the Glass Transition
Temperature 1'g . . . . . . . . . . . . . . . . . . . . 163
6.2.3.3.1 Conventional Glass Transition Temperatures . . . . . 163
6.2.3.3.2 The Thermodynamically Defined Glass Transition Temperature 165
6.2.3.4 Applications of Glass Transition Measurements . . . . . . . . 171
7 Evaluation of the Performance of a Differential Scanning

Calorimeter . . . . . . . . . . . . . . . . . 181
7.1 Characterization of the Complete Instrument 181

7.2 Characterization of the Measuring System .. 181
7.3 Characterization of the Results of a Measurement 186
7.4 Check List for DSCs 188
Appendix 1 189
Appendix 2 193
References 211
Subject Index 217

List of Symbols
A area, pre-exponential factor

C heat capacity, electric capacity
E energy
H enthalpy
K factor (calibration), coefficient
L thermal conductance
P electric power
Q heat
R resistance (thermal, electric), gas constant
S entropy
T temperature
U internal energy, voltage
V volume
W work, electric energy
a coefficient, apparatus function
c specific heat capacity
d distance
electric current
k proportionality factor, calibration factor, rate constant
I length, distance
m mass
n reaction order
p pressure
r rate of reaction
t time
w weight fraction
x mole fraction
a degree of reaction
f3 heating rate
r expansivity coefficient
E emissivity
A. thermal conductivity
v stoichiometric number
CP heat flow rate
p density
(1 standard deviation
'f time constant
X compressibility
~ extent of reaction, composition, conversion
XII List of Symbols
Subscripts
A activation
a amorphous
bl baseline
c extrapolated offset, crystalline
cal calorimeter
corr correction
e extrapolated onset
el electric
eq equilibrium
exp experimental
F furnace
f final
fix fixpoint
fus fusion
g glass
h. i. k. n running numbers
i initial
iso isothermal
I liquid
lit literature
M measurement point
m measured
mix mixing
0 onset
p peak, constant pressure
prop proportional
(fJ related to heat flow rate
Q related to heat
R reference sample
r reaction
Ref reference material used for calibration
(e. g. Certified Reference Material)
S sample
s solid
st steady state
th thermal
tot total
tr true
trs transition
V constant volume
0 zero, zero line
1 Introduction
The objective of calorimetry is the measurement of heat. To measure heat means to

exchange heat. The exchanged heat tends to effect a temperature change in a body
that can be used as a measure of the heat exchanged, or the process of heat exchange
is connected with a heat flow which leads to local temperature differences along its
path which again serve as a measure of the flowing heat.
As chemical reactions and many physical transitions are connected with the ge-
neration or consumption of heat, calorimetry is a universal method for investigating
such processes. Measuring devices in which an exactly known amount of heat is
input into a sample, or abstracted from it, and the temperature change in the
sample is measured (determination of the heat capacity, for example), are also
referred to as calorimeters.
Caloric measurements have been carried out since the middle of the 18th century.
Although modem Differential Scanning Calorimeters (DSC) are widely used today,
the "classic" calorimeters cannot be dispensed with in precision measurements and
for special applications. The most important classic calorimeters will be described
only briefly in Appendix 2 to give the reader a more comprehensive survey of the
field of calorimetry (for a more general presentation of calorimetry, cf. Hemminger,
Hohne, 1984; Oscarson, Izatt, 1992).
The topic of this book is Differential Scanning Calorimetry (DSC) using as a
measuring instrument the Differential Scanning Calorimeter (DSC) available in
various designs (see Chapter 2).
According to the International Confederation for Thermal Analysis and Calo-
rimetry (ICTAC) the definition of Differential Scanning Calorimetry is as follows:
A technique in which the heat-flow rate (power) to the sample is monitored
against time or temperature while the temperature of the sample, in a specified
atmosphere, is programmed.
Differential Thermal Analysis (DTA)

To distinguish a heat flux DSC from an apparatus for Differential Thermal Analy-
sis (DTA), the latter will be characterized in the following.
Differential Thermal Analysis is applied to measure the temperature difference
between the sample to be investigated and a reference sample as a function oftem-
perature (or time). This temperature difference indicates a heat exchange quali-
tatively. DTA is more recent than classic calorimetry. Its advantages compared with
conventional calorimetry are the dynamic mode of operation ("scanning") which
allows reactions or processes to be investigated which can be thermally activated,
2 1 Introduction
and the high sensitivity to anomalies of the temperature-time function. DTA

allows characteristic temperatures to be determined and qualitative statements
made on heat due to reaction. The further development of DTA has led to the con-
struction of Differential Scanning Calorimeters (DSCs) with disk-type measuring
systems.
The ICTAC definition of Differential Thermal Analysis is as follows:

A technique in which the difference in temperature between the sample and a refer-
ence material is monitored against time or temperature while the temperature of the
sample, in a specified atmosphere, is programmed.
For a short description of the principle of DTA measuring systems in compari-
son with DSC, see Appendix 1.
CalOrimetry and DTA Today

Owing to the new materials and sensors used and the application of modem elec-
tronic systems, all calorimeters available today - including the so-called "classic"
calorimeters - are instruments which allow precise measurements to be carried out,
their operation being relatively simple or perhaps even automated. Calorimetric
methods are used in many fields for quality assurance purposes (cf. series of
standards ISO 9000). The standards to be applied in the field of quality assurance
(e.g. the standard EN 45001) demand a large variety of measures to ensure reliable
measurement results, for example, the application of well-proven methods for the
preparation of samples, reliable measurement methods, calibration methods and
calibration materials traceable to standards (see also Sect. 4).
In the following, reference is made to some of the fields in which calorimeters
are widely used today (see Hemminger, Hohne, 1984; Wunderlich, 1990; Oscarson,
Izatt, 1992). A suggestion how to classify calorimeters in given in Appendix 2.
With bomb calorimeters, combustion heats (gross calorific values) are measured
on a large scale in industry (cost of fossil fuels) (Sunner, Mansson, 1979),
gas calorimeters (flow calorimeters) are used for the continuous or discontinu-
ous measurement of the calorific value of fuel gases (e.g. natural gas), both in
frontier-crossing commercial transactions (between supplier and buyer) and for
the calculation of the cost for the individual consumer (Hyde, Jones, 1960; Hem-
minger, 1988),
drop calorimeters (usually home-made) allow mean heat capacities or enthalpy
differences to be measured quickly. Drop sample temperatures of up to 2000°C
are realized (Chekhovskoi, 1984),
different types of isoperibol mixing calorimeters are used to investigate reactions
between two fluids or between a fluid and a solid (reaction heats, heats of solu-
tion, heats of mixing, adsorption heats) (see e.g. Parrish, 1986),
reaction calorimeters and safety calorimeters allow model tests of procedures
applied in industrial chemistry to be carried out. These instruments provide
valuable support in development and optimization tasks since all test parameters
(temperature, time, addition of substances etc.) are completely documented and
can be varied automatically.
I Introduction 3
Questions of process technology which are also important for the dimensioning
of production facilities, may be well in the fore here (Regenass, 1985). Reaction
heat released is connected with the degree of conversion and exploitation, for
example in biotechnology, where calorimetric studies are carried out on in-
dustrial production plants (large-scale calorimetry, mega-calorimetry, see e. g.
von Stockar, Marison, 1991 in Lamprecht et al., 1991).
Special calorimeters (commercially available) are also used to investigate as-
pects of safety technique, for example, to determine characteristic temperatures
of decomposition or the kinetics of reactions (decompositions, run-away reac-
tions) under certain boundary conditions (see, for example, Grewer, 1987;
Schwanebeck, 1991; Singh, 1993; for a survey, cf. Grewer, Steinbach, 1993),
in highly sensitiveflow calorimeters (usually of the isoperibol type, i. e. with sur-

roundings at constant temperature), the heat generation of biological systems and
their change under varying conditions oflife are investigated (for example, addi-
tion of pharmaceuticals to bacterial cultures). In different types of biocalorim-
eters, the metabolism of organisms and their change due to external influences
(optical, acoustical, mechanical, thermal, chemical) are studied (see e.g. Spink,
Wadso, 1976; Lamprecht, Schaarschmidt, 1977; Beezer, 1980; Lamprecht et al.,
1991; Schmolz et al., 1993; Moratzky et al., 1993),
heat capacities and heats of transition are measured directly with high accuracy
using adiabatic calorimeters. The characteristic data of materials determined
with their aid are an indispensable basis for the calibration of the DSCs (see e.g.
Gmnvold, 1967; Kagan, 1984; lakobi et al., 1993; Gmnvold, 1993). Properly
calibrated DSCs allow specific heat capacities to be measured with an accuracy
of 1 to 2% (see Sect. 6.2.1; cf. Richardson, 1992a).
New fields of application are constantly being opened up for the modern, highly
automated DSCs. DSC is· increasingly used in the field of quality assurance for
many purposes: for the inspection of raw materials, as an accompanying measure in
the manufacture and for checking the finished products. The basic limitations to,
and the problems of these instruments should not, however, be forgotten in view of
the ease of operation and evaluation.
DSCs allow reaction heats and heats of transition, or heat flow rates and their
changes at characteristic temperatures, to be measured quickly using small sample
masses (milligram range; in the case of classic calorimeters: gram range), in wide
temperature ranges and with an accuracy which is usually sufficiently high for the
respective purpose. DSCs are applied in the following fields (among others):
characterization of materials,
comparison (relative) measurements (quality control, identification
of substances or mixtures),
stability investigations,
evaluation of phase diagrams,
purity determinations,
kinetic investigations,
safety research.
4 1 Introduction
Now as before, instruments used for Differential Thermal Analysis (DTA) offer
particular advantages when special problems are to be investigated:
they can be used at very high temperatures (up to about 2400 oq,
they are highly sensitive,
they are very flexible as regards the volume and form of the crucibles,
their reasonably-priced measuring system can be easily exchanged.
Characteristic temperatures of transitions or reactions can be determined very well
by DTA. Heats can still be estimated with an uncertainty between 20 and 50%. DTA
is applied in the following fields:
comparison (relative) measurements (identification, control, comparison),
safety research (stability investigations, also long-time investigations), see e.g.
Hentze, 1984; Hentze, Krien, 1986,
investigation of transitions, decompositions, reactions with gases,
evaluation of phase diagrams.
DSC and DTA are also used together with other methods of thermal analysis or
other analytical techniques (simultaneous thermal analysis), most frequently in
connection with Thermogravimetry (TG), more rarely with Evolved Gas Analysis
(EGA), Thermomicroscopy or Thermosonimetry.
The coupling of DSC or DTA with Thermogravimetry (TG) is of particular im-
portance. In addition to information on changes in the heat flow rate (due, for ex-
ample, to changes of heat capacity) and heats of transition, the TG signal provides
information on whether volatile components are involved and which changes in
mass can be attributed to a transition.
The different methods of gas analysis used together with DSC or DTA (usually
together with TGIDSC orTGIDTA systems) are of increasing importance. Via a car-
rier gas stream, gaseous reaction products from the calorimeter or DTA device are
transported to an apparatus in which these products are analyzed. This allows a cor-
relation between the characteristic temperature of a reaction, gravimetric and/or
enthalpic information and the composition of the volatile reaction product to be
made (see examples in Mathot, 1994 a).
It is difficult to transfer the volatile products without alteration from the hot
sample to the analysis system (condensation must be avoided). Various systems for
coupling quadrupole mass spectrometers are commercially available (up to
temperatures of 2400oq. Examples of the investigation of volatile components
applying methods of gas analysis can be found in Ohrbach et aI., 1987; Matuschek
et aI., 1991; Matuschek et aI., 1993.
The advantage of simultaneous measurements is that the same sample is inves-
tigated under identical conditions and that diverse information is obtained by one
measurement run which is important for interpretation purposes. The following
potential disadvantages of simultaneous instruments should be mentioned: lower
sensitivity, higher susceptibility to failure, increased time and effort required for
preparation and operation, higher instrument costs.
In Differential Scanning Calorimetry which is widely applied, there are still
some actual problems of which one should be aware:
1 Introduction 5
there is not yet a theoretically well-founded complete understanding ofDSC. The

average user, therefore, does not yet know the limits to this method and the
sources of the systematic errors by which it is affected,
there are, as yet, no practicable and experimentally tested recommendations for
temperature and heat calibration and for measurement procedures which are
internationally accepted (see Sect. 4),
there is no international agreement on a single set of substances for the tempe-
rature and heat calibration of DSC which have been measured with sufficient
accuracy, including a metrologically sound traceability to national/international
standards; instead there is a confusing variety of "certified reference materials"
whose characteristic data are in part contradictory.
Due to the rapidly increasing use of DSCs in various fields of application, some
negligence has gained ground - favoured by the ease of operation and evaluation -
which would be inconceivable in "classic" calorimetry. Remedial measures should
be taken; for example, the national societies of the International Confederation for
Thermal Analysis and Calorimetry (ICTAC) should offer special training courses,
and precise specifications for instruments and programs should be drawn up.
The following can frequently be observed:
a realistic estimate of the uncertainty of measurement is rarely made. The cali-
bration capability (cf. Sect. 4), different influencing quantities (cf. Sect. 5.2) and
known theoretical considerations should be taken into account (cf. Chapter 3). In
many cases, the repeatability of a DSC is, for example, simply, but wrongly, in-
dicated as accuracy of the measured data (cf. Sect. 7.3),
interpretation of the DSC measurement results is often insufficient or faulty,
when
- uncertainties of measurement are not taken into consideration,
- systematic error sources are disregarded,
- the measured curve is not "desmeared" (cf. Sect. 5.4),
- the laws of thermodynamics, kinetics, are not taken into account,
- uncritical confidence is placed in the evaluation programs provided by the
manufacturer,
- the results are not confirmed by other measuring methods.
A great number of calorimetric methods have not been mentioned here as, in com-
parison with DSC, they are used only in specific fields. To name a few examples:
low-temperature calorimetry for measuring heat capacities (see e.g. Gmelin,

1987),
more recent techniques of heat capacity measurement (see e.g. Lakshmikurnar,
Gopal, 1981; Maglic et aI., 1984),
measurement of the energy of particle radiation (see e.g. Domen, 1987),
measurements on biological systems (see e.g. Lamprecht et aI., 1991; Wads6,
1993; Kemp, 1993),
deformation calorimetry on polymers (see e.g. Kilian, H6hne, 1983; Godovsky,
H6hne, 1994),
6 1 Introduction
- high-temperature calorimetry in material science (see e.g. Bruzzone, 1985;

Bros, 1989),
- (high-)pressure calorimetry (see e. g. in Mathot, 1994 b),
- "oscillating" ("temperature-modulated") DSC (see Wunderlich et aI., 1994;
Reading et ai., 1994; Mayorga et ai., 1994; Schawe, 1995) and AC calorimetry
(Sullivan, Seidel, 1968).
2 Types of Differential Scanning Calorimeters
Two types of Differential Scanning Calorimeters (DSCs) must be distinguished:

- the heat flux DSC,
- the power compensation DSC.
They will be dealt with in this chapter. The theoretical bases are presented in
Chapter 3.
Both types of DSC use a differential method ofmeasurement which is defined as
follows:
A method of measurement in which the measurand is compared with a quantity
of the same kind, of known value only slightly different from the value of the
measurand, and in which the difference between the two values is measured (Inter-
national Vocabulary of Basic and General Terms in Metrology, 1984).
The characteristic feature of all DSC measuring systems is the twin-type design
and the direct in-difference connection of the two measuring systems which are of
the same kind. It is the decisive advantage of the differential principle that, in first
approximation, disturbances such as temperature variations in the environment of
the measuring system and the like, affect the two measuring systems in the same
way and are compensated when the difference between the individual signals is
formed.
Moreover, the differential signal - which is the measurement signal actually of
interest- can be strongly amplified, as the high basic signal (signal of the individual
measuring system) is also compensated when the difference is formed.
An extension to form multiple measuring systems (three or four) connected back
to back does not mean a fundamental change in the differential principle.
The differential signal is the essential characteristic of each Differential Scanning
Calorimeter. Another characteristic - by which distinguishes it from most classic
calorimeters - is the dynamic mode of operation. The DSC can be heated or cooled
at a preset heating or cooling rate (isothermal mode is possible as well).
A characteristic common to both types ofDSC is that the measured signal is pro-
portional to a heat flow rate <1> (and not to a heat as is the case with most ofthe clas-
sic calorimeters). This allows time dependences of a transition to be observed on the
basis of the <1>(t) curve. This fact - directly measured heat flow rates - enables the
DSCs to solve problems arising in many fields of application (see Chapter 6).
8 2 Types of Differential Scanning Calorimeters
2.1 The Heat Flux DSC

The heat flux DSC belongs to the class of heat-exchanging calorimeters (for the
classification, see Appendix 2). In heat flux DSCs a defined exchange of the heat
to be measured with the environment takes place via a thermal resistance. The
measurement signal is the temperature difference; it describes the intensity of the
exchange and is proportional to the heat flow rate (/J.
In commercial heat flux DSCs, the well-defined heat conduction path is realized
in different ways, with the measuring system being sufficiently dominating. The
most important fundamental types are:
the disk-type measuring system with solid sample support (disk).
Features: It allows high heating rates, its time constants and the sample volume
are small, but it has a high sensitivity per unit volume,
the cylinder-type measuring system with integrated sample cavities.
Features: Provided with large cavities and sample containers, it allows only low
heating rates, its time constants and the sample volume are large, but it has a low
sensitivity per unit volume.
2.1.1 Heat Flux DSC with a Disk-Type Measuring System
The characteristic feature of this measuring system is that the main heat flow from
the furnace to the samples passes symmetrically through a disk of good thermal
conductivity (Fig. 2.la). The samples (or the sample containers) are positioned on
this disk symmetrical to the centre. The temperature sensors are integrated into the
disk or fixed on its surface. Each temperature sensor covers more or less the area
supporting the respective container (crucible, pan) so that calibration can be carried
out independently of the sample position inside the container (cf. Sect. 4.3). To
keep the uncertainties of measurement as small as possible, the arrangement of
sample and reference sample (or of the containers) and temperature sensor in
relation to one another must always be the same (centre pin or the like on the con-
tainer bottom).
Metals, quartz glass or ceramics are used as disk materials. Type and design of
the temperature sensors differ (thermocouples, resistance thermometers).
When the furnace is heated (in general linearly in time, more recently also in a
modulated way), heat flows through the disk to the samples. When the arrangement
is ideally symmetrical (samples of the same kind), equally high heat flow rates flow
into sample and reference sample. The differential temperature signal tlT (original-
ly a difference between electric potentials) is then zero. If this steady-state equi-
librium is disturbed by a sample transition, a differential signal is generated which
is proportional to the difference between the heat flow rates to the sample and to the
reference sample:
As neither ideal thermal symmetry of the measuring system at all operating tem-
peratures nor thermal identity of the samples can be attained in practical appli-
cation, not even outside the transition interval, there will always be a signal llT
which depends on the temperature and the sample properties (cf. definition of
zero line in Sect. 5.1). Chapter 3 is based on the assumption that this portion of
the total signal is zero or has already been subtracted from the measurement signal
proper.
3
rzzzzzzzz£l~zz6zZ21 5
S R
T(t) t--~--.----...,
'--- _ _6_T_ _- - l <<

calibra tion
a K(TJ
b time, tempera ture

Fig. 2.1. a Heat flux DSC with disk-type measuring system. 1 disk, 2 furnace, 3 lid, 4 differ-
ential thermocouple(s). 5 programmer and controller, S crucible with sample substance,
R crucible with reference sample substance, ~ heat flow rate from furnace to sample
crucible, ~R heat flow rate from furnace to reference sample crucible, C/lm measured heat
flow rate, K calibration factor.
b Measured heat flow rate C/lm (schematic curve) (acc. to Hemminger, 1994)
The measurement signal ~Tis always obtained as electrical voltage. In almost all
heat flux DSCs, a heat flow rate tPm (m: measured) is internally (in the computer)
assigned to this signal ~Tby factory-installed provisional calibration:
tPrn = -k'· ~T
The measurement signal output by the DSC (Fig. 2.1 b) and accessible to the user is
tPrn (in /-1W or m W). The calibration of the DSC must be checked by the user: to what
extent does tPrn represent the true heat flow rate tPtrue which is released or consumed
by the sample?
This test can be carried out:
by measuring the steady-state heat flow rate into a sample of known heat capa-
city C, "charging" of a "heat capacity" in the quasi-steady-state of scanning with
constant heating rate f3 = dTldt:
C . f3 = tPtrue = K", . tPrn (2.1)
by comparing the integral over a transition peak with the expected (known) heat
of transition Qr (energy balance):
Qr = Qtrue = KQ . f (tPrn - ~I) dt (2.2)
(for definition of baseline ~.. see Sect. 5.1)
(When resistance thermometers are used as temperature sensors, these thermom-
eters can be used also as heaters which is advantageous in calibration procedures,
see Wolfet aI., 1994).
Questions of calibration capability and calibration are dealt with in Sect. 4.4;
problems of "desmearing" arising in connection with it are discussed in Sect. 5.4.
Heat flux DSCs with disk-type measuring systems are available for temperatures
between - 190 and 1500°C. The maximum heating rates are about 100 K min -I.
Typical time constants (empty systems, no samples) are between 3 and 10 s. The
noise ofthe measurement signal (definition see Sect. 7.2) lies between 1 mW and
50/-1W (it also depends on the temperature and the heating rate). The total uncer-
tainty of the heat measurement amounts to about 5 %, and it is to be expected that
it could not be reduced to less than 2 % even if more time and effort were spent (cf.
Sect. 4.4).
Heat Flux DSC with Triple Measuring System

An extension of the disk-type measuring system described above has been pre-
sented by TA Instruments Inc. (formerly Du Pont). It is a "Dual Sample DSC" with
three locations on the metallic (constantan) disk to receive sample crucibles.
The three locations are equipped with temperature sensors (cf. Fig. 2.1 a for the
normal DSC) which can measure the temperature differences between them. This
configuration makes it possible to measure the heat capacity in a single run, using
one measuring position for the sample, the second for an empty crucible, and the
third for the calibration (reference) material with well known heat capacity; (Jin,
Wunderlich, 1993; see Sect. 6.2.1.5). The temperature range of this Dual Sample
DSC is from 125 to 1000 K with heating rates up to 100 K min-I.
2.1 The Heat Flux DSC II
The development of an adiabatic high temperature triple-cell DSC system (up to

1500 K) has been described by Takahashi, Asou, 1993.
Pressure Heat Flux DSCs

Special containers for heat flux DSCs with disk-type measuring system allow
higher pressure, in general up to 7 MPa (70 bar), to be applied. This makes it pos-
sible to determine, for example, vapour pressures and heats of evaporation by
means ofDSC (Wiedemann, 1991; Perrenot et aI., 1992, see also articles in Mathot,
1994 b).
Heat Flux DSC with Modulated Heating Mode

A modification ofthe purely linear heating mode of heat flux DSCs with disk-type
measuring system is realized if it is superimposed by a sine function (temperature
modulated DSC). In addition to the heating rate the amplitude and the frequency of
the modulating sine function can be changed. This operation mode is suitable for
disk-type DSCs with a time constant (see Sect. 7.2) which is small compared with
the modulation frequency. Otherwise a separation of the thermal effects due to the
temperature and due to the heating rate (see Sect. 6.2.1.8) would not be possible.
2.1.2 Heat Flux DSC with a Cylinder-Type Measuring System
A block-type cylindrical furnace is provided with two cylindrical cavities, each con-
taining a cylindrical, fixed sample container which is connected with the furnace or
directly with the other container by means of several thermocouples (thermopiles),
which are the characteristic features of this type of measuring system.
In the original cylinder-type measuring system (acc. to Calvet, 1948, Fig. 2.2),
the outer surfaces of each sample container are in contact with a great number of
thermocouples connected in series between the container and the furnace cavity.
The thermocouple bands or wires are the dominating heat conduction path from the
furnace to the samples. Heat conduction path and temperature difference sensors
are identical. Both sample containers are thermally decoupled; heat exchange takes
place only with parts ofthe massive furnace. The measurement signal proper is the
temperature difference I1T of both sample containers averaged over the surfaces; it
is generated by differential connection of both thermopiles:
I1T = I1TFR - I1TFF = I1TsR

The following is valid for the (steady-state) heat flow rates exchanged between fur-
nace and sample, ~ , between furnace and reference sample, ~, and for the meas-
ured heat (output) flow rate IPm and the true heat flow rate IPtrue into the sample:
IPm = -k' ·I1T

IPtrue = K • . IPm
12 2· Types of Differential Scanning Calorimeters
3
'-------0 !::, T 0 - - - - - - - - '
Fig. 2.2. Heat flux DSC with cylinder-type measuring system (Calvet, 1948; thermally
decoupled sample containers) (acc. to Hemminger, 1994).
1 containers to take up sample and reference sample, 2 thermopile, 3 furnace (with pro-
grammable temperature controller), 4 lid, S sample substance, R reference sample substance,
AT temperature difference between the containers
A voltage proportional to AT, or a heat flow rate signal internally calculated from
dT(as in the case ofthe DSCs with disk-type measuring system, cf. Sect. 2.1.1) is
put out as a measurement signal. It must be checked in both cases to what extent the
measurement signal corresponds to the (steady-state) heat flow rate actually
exchanged or to what extent the integral over the measured 'peak corresponds to the
known heat of transition (cf. Sects. 4.4 and 5.4).
In modified cylinder-type measuring systems (for example, acc. to Petit et aI.,
1961), no heat flows from the furnace to the samples via thermocouples (Fig. 2.3).
The heat exchange between furnace and samples takes place via the mechanical
connections to the sample containers, via leads and gas layers between furnace and
sample containers. Thermocouples are used to measure the temperature difference
directly between sample and reference sample containers. Both containers are no
longer thermally decoupled, however, the temperature difference measured between
them is again proportional to the differential heat flow rate from the furnace to the
samples, part of which is now also exchanged between the two containers.
In all cylinder-type measuring systems, the calibration factor may depend on
the position and height of the sample in the container, the reason being that
the efficiency of heat conduction paths, such as mechanical connections, leads
and gas layers, depends on the position of the heat source (sample) in the con-
tainer. When the sample comes close to the upper edge of the container, the
calibration factor can change considerably, depending on the calorimeter type (cf.
Sect. 4.4).
Fig.2.3. Heat flux DSC with modified cylinder-type measuring system (thermally coupled
sample containers) (acc. to Hemminger, 1994).
1 containers to take up sample and reference sample, 2 thermopile(s), 3 furnace (with
programmable temperature controller), 4 lid, 5 support of containers, S sample substance,
R reference sample substance, ~Ttemperature difference between the containers
Compared with disk-type measuring systems, cylinder-type measuring systems

have the advantage of a larger useful volume. In return, greater thermal inertia (time
constants of up to 40 min) usually have to be accepted. The large cavity volume
allows different special containers (for example, for electrical calibration, gas flow,
mixture, high pressure etc.) to be used and makes intervention (electrical, mecha-
nical, gas exchange, ... ) and direct observations (acoustical, optical) in the sample
cavity possible. The great number of thermocouples (up to 1000) generates a high
output signal at relatively low noise.
Heat flux DSCs with cylinder-type measuring systems are available for a tempe-
rature range between - 190 and 1500 °C. Due to the relatively great time constants
of the measuring systems, the maximum heating rates - depending on the cavity
volume - lie at about 30Kmin- l . The instruments with large volumes (up to
100 cm 3) are used, for example, in biology for the investigation of small animals at
constant temperature (cf. Moratzky et aI., 1993; Schmolz et aI., 1993). Other
systems are used for highly sensitive measurements (detection limit about 1 f.l,W).
They can usually be heated with maximally 1 K min-I; in most cases operation is
isoperibol (isoperibol: surroundings at constant temperature) within a restricted
temperature range.
High temperature systems can be used in an isoperibol mode of operation as drop
calorimeters (cf. Appendix 2), to measure, for example, enthalpies of mixing (see
e.g. Fan et aI., 1993).
For heat flux DSCs with cylinder-type measuring systems according to E. Calvet
(decoupled sample containers), a rather simple theory has been developed which
quantitatively describes the functional correlation between the instantaneous mea-

surement signal «Pm and the original event «Pr in the calorimeter (cf. Sect. 3.1).
According to this theory, "de smearing" of the measured curve - to obtain the «Pr (t)
function - can be performed rather easily "on-line", for example by means of an
electronic device.
For disk-type DSCs, the desmearing procedure is not so simple but can be per-
formed as well (see Sect. 5.4).
2.2 The Power Compensation DSC

The power compensation DSC belongs to the class of heat-compensating calori-
meters (see Appendix 2). The heat to be measured is (almost totally) compensated
with electric energy, by increasing or decreasing an adjustable Joule's heat.
The commercial power compensation DSC * most frequently used is an instru-
ment with isoperibol mode of operation. The measuring system (Fig. 2.4) consists
of two microfurnaces of the same type made of a platinum-iridium alloy, each of
which contains a temperature sensor (platinum resistance thermometer) and a heat-
Fig.2.4. Power compensation DSC (Perkin-Elmer Corp.). Set-up of the measuring system
(acc. to Hemminger, 1994).
S sample measuring system with sample crucible, microfurnace and lid, R reference sample
system (analogous to S), I heating wire, 2 resistance thermometer. Both measuring systems-
separated from each other - are positioned in a surrounding (block) at constant temperature
* The description of a power compensation DSC is based on the widely used DSC of the
Perkin-Elmer Corp.
ing resistor (made of platinum wire). The microfumace is about 9 mm in diameter,

approx. 6 mm in height and has a mass of approx. 2 g. The time constant is slightly
smaller than 2 s, and the isothermal noise is about 2 IlW The maximum heating
power of a microfumace is about 14 W, the maximum heating rate is 500 Kmin- I
(Perkin-Elmer DSC 7).
Both rnicrofumaces - separated from each other (thermally decoupled) - are
positioned in an aluminium block of constant temperature. The measuring range
extends from -175 (block cooled with liquid nitrogen) to 725°C.
During heating-up, the same heating power is supplied to both microfumaces via
a control circuit (Fig. 2.5) in order to change their mean temperature in accordance
with the preset heating rate (see Watson et aI., 1964 *). If there is ideal thermal
symmetry, the temperature of both microfumaces is always the same. When an
asymmetry occurs, for example as a result of a sample reaction, a temperature
difference results between the microfumace accommodating the sample and the
@ (alibra tion
~
¢m
jjp jjp
Ts TR
Sample- Reference-
furnace furnace
Ts TR
Pav Pav
T (f)
I--------~ Recorder ¢m( f)
'---=-----'
Fig.2.S. Power compensation DSC (Perkin-Elmer Corp.). Block diagram showing the func-
tion principle (acc. to Hemminger, 1994).
Ts temperature of the sample furnace, TR temperature of the reference sample furnace, IlT =
Ts - TR• Pay average heating power, IlP compensation heating power, <Pm measured heat flow
rate (measurement signal)
• In the basic paper by Watson et aI., the term Differential Scanning Calorimeter (DSC) is coin-
ed, but the title of the paper is: A Differential Scanning Calorimeter for Quantitative Dif-
ferential Thermal Analysis, i. e. the inventors of this DSC recognized clearly the systematic
relationship between this type ofDSC and classic DTA.
microfurnace containing the reference sample. The temperature difference is both

the measurement signal and the input signal of a second control circuit. This second
circuit tries to compensate the reaction heat flow rate by proportional control by
increasing or decreasing an additional heating power. The compensating heating
power provided for this purpose amounts to up to ± 700 mW. The compensating
heating power IlP is proportional to the remaining temperature difference ilT
(because of the proportional controller). The time integral over the compensating
heating power is proportional to the heat Qr which was consumed or released in the
sample (Fig. 2.6).
Again, a heat flow rate 4>m is assigned to the real measurement signal ilT as a
result of a factory-installed calibration, and fed in. The relations between ilT, 4>m and
the compensating heating power IlP are as follows:
The factor k) is a factory-set fixed quantity of the proportional controller, k2 can be

changed at the instrument with the aid of a potentiometer or it is adjusted via the
software (calibration). The factor k2 is almost independent of measurement
parameters (e.g. temperature), as - via k) - a given compensating heating power
always corresponds to a given ilT; k2 can therefore - in principle - be determined
by one calibration measurement.
As regards the formal aspects (cf. Sect. 2.1, heat flux DSC), the output signal is
also given as a heat flow rate signal 4>m (e.g. in mW), and the relation between 4>m
and the true heat flow rate exchanged with the sample, 4>true = K~· 4>m' must also be
determined by caloric calibration (see Sect. 4.4).
Thermal asymmetries of the measuring system, which become apparent as a
curvature of the zero line, can be electronically compensated. In this way, it is also
possible to "straighten" the measured curve outside the peaks and/or incline it as
desired.
At higher temperatures of the measuring system, the heat flows exchanged with
the (isoperibol) surroundings (by conduction, radiation, convection) are relatively
large compared with the quantity to be measured. High requirements must, there-
fore, be met as far as the uniformity of the heat exchange between the two micro-
furnaces and the surroundings is concerned in order to keep the uncertainties of
measurement small. Moreover, the shares of the various heat exchange mechanisms
and their respective amounts must depend only on the temperature, and strict
repeatability must be ensured. Conclusion: The microfurnaces must be covered
with lids of the same kind in order to "cover up" inhomogeneities of sample and
reference sample; the thermophysical properties of the crucibles and lids must
depend only on the temperature.
Compared with heat flux DSCs, the power compensation DSC offers the follow-
ing advantages:
the short heat conduction path between samples and heater and the relatively
small masses of the micro furnaces allow an almost instantaneous response to a
sample reaction. Due to the small time constant, desmearing is required only in
a few cases (cf. Sect. 5.4),
time~
t time~
fjP
fjtl~--~IC====·~~-t-im-e~--
Fig.2.6. Power compensation DSC. Diagrammatic view of the signals in question (ace. to
Hemminger, 1994).
<fir heat flow rate released in the sample (exothermic, therefore negative), qr heat released in
the sample, tiT· temperature increase in the sample furnace which would build up due to the
exothermic effect unless it was compensated (as in heat flux DSC), tiP compensation heating
power (negative, to compensate the exothermic effect), tiT temperature increase actually
occurring in the sample furnace (corresponds to the residual deviation from the theoretical
value, which cannot be completely compensated by the proportional control), <fIm output
signal (proportional to the (negative) amplified tiT signal)
reaction heat flow rates are rapidly and to a large extent compensated by electri-
cal heating power. As a result, only small temperature differences tlT occur be-
tween the microfurnaces of sample and reference sample (approx. 1110 of those
occurring in heat flux DSCs between sample and reference sample supports).
This means that the calibration factor Kq, is practically independent ofthe inten-
sity and kinetics of the sample reaction,
the total compensating energy (f tlPdt) is equal to the reaction heat or heat of
transition,
the temperature dependence of the control circuit properties (above all of the
temperature sensors and the heating elements) is known and strictly repeatable.
It can be taken into account with the aid of electronic circuits or by the software.
A single caloric calibration is then - in principle - sufficient to determine the
correlation between 4>m (or 4>r) and tlP (cf. Sect. 3.2).
The user must, however, keep in mind that the temperature difference between the
two microfurnaces is not totally compensated for in commercial power compensa-
tion DSCs. During the peak there is still a temperature difference IlT proportional
to the reaction heat flow rate. That is to say, the power compensation DSC can be
considered a kind of DTA instrument with a tlT as the measurement-signal, but
with a tlT (arising from a specific thermal event) which is much smaller than that
developed in a heat flux DSC measuring system.
In conclusion, it may be said that all the attributes of heat flux DSC systems
which depend on this temperature difference can also be found in the "real" power
compensation DSC, but to a lesser degree. The calibration factor in particular is not
a constant figure but depends in principle on temperature, heat flow rate, heating
rate and peak area (cf. Sects. 3.2,4.4). Though these effects are not very pronoun-
ced, they should be carefully tested and a thorough calibration (with the relevant
parameters varied) should be performed if the demands on the accuracy of the
measurements are high (Hohne, Gloggler, 1989).
The design, specifications and application of a power compensated DSC (based
on the Perkin-Elmer DSC-7) used under high pressure (up to 500 MPa) have been
described by Blankenhorn, Hohne, 1991.
Photo-DSC
Slight changes in the design of a DSC make it possible to irradiate the sample with
light. A DSC modified in this way is referred to as Photo-DSC.
Irradiation of the sample with light having enough energy leads to a reaction
in the Photo-DSC. The heat generated during these reactions is recorded. First
applications of this type of DSC were described by Wight, Hicks, 1978; Tryson,
Shultz, 1979; Flarnmersheim, 1981. The DPA-7 of Perkin-Elmer has been com-
mercially available as a supplementary device for the DSC-7 since 1987. (Heat flux
DSCs are also available with suitable additional devices.) The principle is shown in
Fig. 2.7.
Sample and reference substance are contained in the calorimeter at constant tem-
perature. The sample crucibles are either open or covered by quartz disks. The lid
of the microfurnace is provided with openings sealed with quartz disks. Iflight high
light source
optics
quartz cuvette with

water as IR-absorber
monochroma tor
neutral density filters
calorimeter
Fig. 2.7. Photo-DSC, schematic representation
in energy is incident on the sample, it is either absorbed directly or - as is usually

the case - with the aid of a photo-initiator. A reaction usually takes place only as
long as energy is supplied by irradiation. High-pressure mercury vapour lamps with
a large number of spectral lines are frequently used as a light source; however,
hydrogen, deuterium or xenon lamps are also suitable. The undesired infrared
portion of the spectrum can be quantitatively absorbed by quartz cells filled with
water. The portion of the spectrum which is of interest is selected using mono-
chromatic filters and the desired intensity is adjusted by means of neutral
density filters or metal sieves with different mesh width. It is a particular advantage
that the DSC is also suitable for the direct measurement of the incoming radiation
flow (light intensity). For this purpose, a graphite disk of known emissivity is
substituted for the sample. This disk absorbs the greater portion of the incident
light and transfers it to the calorimeter in the form of a heat flow. For most appli-
cations, it is of secondary importance whether sample and reference cells are
irradiated or the sample cell alone. For precise measurements electronic stabiliza-
tion of the light intensity of the lamp is important. Otherwise the total noise of the
observed curve will be substantially higher than the noise of the DSC measuring
system.
Light-activated reactions under conditions closely related to practice are usually
very fast. The greatest part of the reaction takes place within a few seconds
(cf. Figs. 6.19 and 6.20). There are two consequences:
there is a considerable change in the measurement signal in periods comparable
with that of the DSC time constant. The measured and true heat flow rates differ
substantially. Desmearing is necessary prior to every evaluation in which the
time variable is involved, for example, when calculating the conversion versus
time curve (cf. Sect. 5.4),
heat flow rates (cf. Fig. 6.19) are so great that the sample temperature deviates
considerably from the temperature of the microfurnace, even in power com-
pensated DSCs. If a thermal resistance of 40 K W-I between sample and tempe-
rature sensor is assumed, temperature differences of 5 to 10K would result even
for samples with very good thermal conductivity. For organic samples, these
differences are even greater. This means that the measurements are no longer iso-
thermal, not even as an approximation. Kinetic analysis must allow for this.
Adiabatic DSC
For precise biochemical investigations, commercial power compensation DSCs are
often used that are operated adiabatically. There are different types with a sample
volume of approx. 5 ml and a noise (cf. Sect. 7.2) of approx. 1 ~W (Privalov,
1989).
3 Theoretical Fundamentals of Differential Scanning
Calorimeters
In all DSCs, a temperature difference tlT - given as a voltage - is the original

measurement signal. In almost all instruments a heat flow rate 4Jm (differential heat
flow rate) is internally assigned to tl T. Independent of whether the user obtains tl T
or 4Jm from the respective DSC, knowledge of the functional relation between the
measured signal (tlT, 4Jm> and the quantity searched (the reaction heat flow rate 4Jr
consumed/produced by the sample) is important for:
- the time-related assignment of 4Jr to tlT or 4Jm (investigation into the kinetics of
a reaction),
- the determination of partial heats of reaction,
the evaluation and assessment of the influences of operating parameters and pro-
perties of the measuring system with regard to this relation,
the estimate of the overall uncertainty of measurement.
The relation between 4Jr and tlT or 4Jrn can be derived in varying degrees of ap-
proximation to real DSCs. Analytical solutions are possible only for simple boun-
dary and initial conditions and for quasi-steady-states. Numerical procedures and
solutions can approximate the actual conditions more exactly, however, without
the clarity ofthe functional relations given by an analytical solution.
Basic considerations in this field are given by Gray, 1968.
To ensure better differentiation from 4Jr , in the following section, tlT - instead of
4Jrn - is assumed to be the measurement signal, i.e. we search the relation 4Jr (tlT).
The two quantities tl T and 4Jrn , are strictly proportional, with the exception of the
opposite sign.
3.1 The Heat Flux DSC

Three steps of an analytical description of the functional principle of a heat flux
DSC will be presented in the following. The results of a numerical calculation will
be discussed afterwards.
The Zeroth Approximation

First, the heat flux DSC is represented by a simple linear model (Fig. 3.1).
The following simplifications have been made in the model of the zeroth ap-
proximation:
- steady-state (constant) heat flow rates,
22 3 Theoretical Fundamentals of Differential Scanning Calorimeters
/
/ tJl
Tf S
>£cfJ fS
Ts TR
/
"
Fig.3.1. Heat flux DSC (disk-type), model for zeroth approximation (linear model).
S sample, R reference sample, F furnace, A cross section of the heat conductor between
furnace and S and R, AI distance between temperature measurement point and furnace
only the thermal resistance between furnace and sample is taken into account,
i. e. no interaction between sample and reference sample,
with the exception of Cs , CR (heat capacities of the sample, reference sample),
no heat capacities are taken into account,
sample temperature equal to measured temperature,
no heat exchange with the surroundings (heat leak).
Figure 3.2 shows the equivalent electric circuit diagram for the zeroth approximati-
on. This diagram serves to understand the interrelations better. From the physical
point of view, electric charge transport and heat transport are equivalent processes,
and many people find it easier to read electric circuit diagrams than to visualize heat
flows in real equipment.
/),u
Fig.3.2. Equivalent electric circuit for the linear model of the heat flux DSC (see Fig. 3.1).
C capacitance, R resistance, i current, U voltage. Subscripts: S sample, R reference, F furnace
In absolute values, the Biot-Fourier equation of the (steady-state) conduction of

heat reads as follows:
41
- = - A . grad T or
1411
- =AlgradTI
A A
The amount of the heat flux lPIA is proportional to the gradient of the temperature;
the thermal conductivity A is the proportionality factor.
In the one-dimensional model referred to above, this equation is reduced as
follows for the left-hand and right-hand subsystem shown in Fig. 3.1 (with TF> Ts ,
TR in scanning mode):
In the case of thermal symmetry,

IPps = ern
is valid, with Ts = TR.
If a constant (exothermic) heat flow rate (lPr < O)is produced in the sample, Ts
increases by tlTs , the temperature difference TF - Ts and thus the heat flow rate IPps
decreases. When the steady-state (i.e. uniform heating rate) is reached again, for
reasons of balance, tllPps, the change of IPps, must be equal to lPr :
A'A
tllPps = lPr = - - - tlTs
tll
Nothing has changed on the side ofthe reference sample, hence:
tlTs = tlTSR = Ts - TR and lPr = L1lPsR = IPps - IPpR
Consequently
A'A A'A
<P. =- - - (Ts -TR) = - - - tlTSR =-K' tlT
r tll tll
(measurement signal: tlT = tlTs~
(Here, Ts> TR, and lPr is negative: exothermic effect.)

In this simple steady-state model, K is given completely by the properties of the heat
conduction path between the furnace and the samples. This means that in the steady-
state, there is direct proportionality between the measurand lPr and the measurement
signal tlT. The conditions of constant heat consumption can be achieved when in
scanning operation sample and reference sample have different, temperature-inde-
pendent "heat capacities". A greater amount of heat will always flow into the sam-
ple whose heat capacity is higher, in order that the steady-state heating rate is main-
tained. With Cs > CRthe following is then valid for the difference between the heat
flow rates to sample and reference sample:
L1lPSR = - K' . tlTSR and Ts < TR
There is no steady state during sample transitions or reactions; the above approxi-
mation does not apply in these cases. Furthermore, Cs and CR (and thus L1<PSR )
change with temperature, but these changes are in many cases rather slow and do
not affect the steady-state condition very much.
When there is such a quasi-steady state, the following is valid in approximation:
tl<PSR = P(Cs - CJ and thus P(Cs - CJ = -K' . tlT (3.1)
This relation describes the shape of the baseline and is the basic equation to deter-
mine the heat capacity Cs . For the empty reference crucible (CR = 0)
tlT
C = -K'·-
s P
is valid.
When the heat capacity CR is different from zero (i. e. if CR = CRef using a
reference material),
tlT
Cs =CR - K ' ·P-
In practice, first a zero line tlTo will be recorded with the crucibles empty (to check
the asymmetry of the apparatus) which is subtracted from the measured curves.
For the measurement of heat capacities, see Sect. 6.2.1.
The First Approximation

In the first approximation, non-steady-state processes in the sample are also per-
mitted which manifest themselves as "peaks" of the measured curve. This means
that AT is not constant in time. For the rest, the simplified arrangement as for the
zeroth approximation is used here.
~sis the heat flow rate from the furnace to the sample, <Pr(t) the heat flow rate
produced inside the sample (reaction, transition). The following balance equation
for the heat flow rates is then valid for the sample of heat capacity Cs :
dT.
Cs dt s = ~s + <Pr (exothermic: <Pr negative, endothermic: <Pr positive)
With tlT = Ts - TR,

dTR dtlT
Cs - + Cs -d = ~s + <Pr
dt t
results.
Accordingly, the following holds for the reference sample (<Pr = 0 by definition):
dTR
CRdt =~
When the difference between the two balance equations is formed, the following is
obtained:
dTR dtlT
~ - ~ = (CR - CJ dt + Cs dt
+ I<Pr I
The following is valid for the heat flow rates ~s and ~R:
and
where R FS and RFR are global heat resistances between the furnace and the samples.
In the case ofthermal symmetry, R FS = RFR = R, thus:
I:1T) dTR dI:1T

CPr=- (- -(Cs-CJ - - Cs - (3.2)
R dt dt
This equation links the reaction heat flow rate CPr searched with the measured sig-
nall:1T. The second term takes the asymmetry of the measuring system into account
as regards heat capacities of sample and reference sample. The third term considers
the contribution of the thermal inertia of the system when a measured signaII:1T(t)
appears. In analogy to the charging or discharging of a capacitor of capacity C, a
time constant 'f can similarly be defined for the heat flow rates:
'f = Cs . R
When I:1T is changed, R is the effective thermal resistance to the "charging or
discharging" of the "capacity" Cs . With this resistance and with dTR/dt = {3, the
heating rate; as the reference sample is always in a steady-state heating mode, the
following results from Eq. (3.2):
(3.3)
The measured signal I:1T is not proportional to the heat flow rate CPr at a given
moment but delayed with time and thus also distorted ("smeared"). Furthermore,
when Cs =F CR is valid, the measurement signal is not equal to zero - even if CPr is
equal to zero and steady-state conditions prevail- but has the value - R . {3 (Cs - CR )
which is the initial deviation after the quasi-steady state has been reached in scan-
ning operation (cf. Eq. (3.1». This contribution is the measured curve before/
behind a peak which is parallel to the abscissa if R. I:1C = Cs - CR and (3 are constant
(Fig. 3.3).
In reality, the term R (Cs - CJ reflects the temperature dependence of the
thermal resistance R (in general: of the heat transfer conditions) and of the heat
capacities Cs and CR causing a temperature dependence of the measured curve even
without any thermal effect due to the sample. (Equation (3.3) is in principle the so-
called Tian equation).
Regarding Eq. (3.3), the conclusions are as follows:
1. When the signall:1T measured at a given moment is to be assigned to the heat
flow rate CPr by which it is caused, the 3rd term in Eq. (3.3) must be taken into
account (cf. Desmearing, Sect. 5.4).
R must be determined by calibration (cf. Sect. 4.4); the time constant -rean also
be obtained from calibration measurements (cf. below "Higher-order approxi-
mations" and Sect. 7.2).
t
6.T
-R . (3.6. [ 2
time ---..
Fig.3.3. Measured curve of a heat flux DSC (schematic) with heat released inside the sample
(exothermic effect).
aT measured signal, R thermal resistance between furnace and sample, {:J heating rate, IlC =
Cs - CR difference between the heat capacities of sample and reference sample (here negative,
because CR > Cs , <D peak area (exothermic effect); it is a measure of the heat released between
t, and t2 , ~ area below the baseline; it is a measure of the heat required to heat the sample
between t, and t2
For t> to, the solution of the differential Eq. (3.3) for a heat pulse 4'r at the
moment to has the form
with T= Cs . R. Up to the moment t = to, the solution is the steady-state curve
For t -7 co, the function IlT(t) returns to this curve (cf. Fig. 3.4).
2. For the whole heat of reaction or transition Qr developed/consumed in the sam-

ple, the following balance equation is valid:
'2
Qr = f <P,(t) dt
"
where tl> t2 are the beginning and end, respectively, ofthe peak.
~--~--~------~----~-~------
t--
:
T----;:
f
~T
Fig.3.4. Measurement signal of a heat flux DSC for an exothermic heat pulse at the time to
(model of the 1st approximation).
'r time constant; symbols see Fig. 3.3
With Eq. (3.3) inserted, the following is obtained (see Fig. 3.3):
1 [ '2 '2 ] '2 1" ddT dt

Qr=- - ft.T(t) dt- f(-R)' t.C- f3dt - f _. - (3.4)
R '1'1 'I R dt
The content of the square brackets corresponds to area 1, i. e. the so-called peak
area between the measured curve and the (interpolated) baseline (definition see
Sect. 5.1).
When t.C and R are not temperature- (or time-)dependent, the measured
curve outside the peak is a parallel to the abscissa. In this case, ddTldt is zero
before and behind the peak. The contribution of the 3rd term vanishes when
integration is carried out over the whole peak,
in the real case, the 2nd term of Eq. (3.4) is not constant, i. e. the curve
before and behind the peak is not parallel to the abscissa. In this case, the 3rd
term will not vanish but represents a correction of the values obtained by peak
integration.
3. For the partial integration (Fig. 3.5) ofthe peak between t( and t*, the contribu-
tion of the 3rd term must be taken into acount at the point t*:
1 [ ,. ,. ] ,. 1" ddT
Qr(t*) =- - ft.T(t) dt- f (-R)' t.C· f3dt - f - ' - dt
R '1'1 'I R dt
'-----v----'
partial peak area correction term
Partial integrations of the peak are necessary for kinetic investigations and to
determine the purity (see Sects. 6.2.2 and 6.2.4).
t
b,T
-R . f3 .L1 [ 2
t* time ~
Fig. 3.5. Partial integration of a peak (heat flux DSC, exothermic effect).
CR > Cs , <D partial peak area (between t1 and t*), ~ area below baseline (between t1 and t*);
Symbols see Fig. 3.3
/
/
Fig. 3.6. Heat flux DSC (one halt), model of the 2nd approximation.
S sample, F furnace, TMS temperature at sample measurement point
Higher-order Approximations
The temperatures of sample and reference sample (assumed here to be homogene-
ous) are not measured directly. There is a thermal resistance between the tempera-
ture measurement points and the respective sample (Fig. 3.6). Depending on the
design of the measuring system, the resistance is made up of several parts differing
in amount and originating in the transition layers between sample, bottom of the
RFM R MS
S
b"U
RFM RMR
MR
4
[FM [R [FM [s
Fig. 3.7. Equivalent electric circuit for the 2nd approximation.

S sample side, R reference side, F furnace, Ms measurement point on sample side, MR mea-
surement point on reference side. aU corresponds to aT, Up corresponds to Tp, i corresponds
to <J)T' RPM thermal resistance between furnace and measurement points Ms, MR, RMS thermal
resistance between sample measurement point and sample, RMR thermal resistance between
reference sample measurement point and reference sample, Cs heat capacity of the sample, C R
heat capacity of the reference sample, CPM effective "heat capacity" between furnace and
measurement points Ms, MR' For correspondences between electrical and thermal quantities
cf. page 22
crucible and support; there are further resistances between support and temperature
sensor. When the sample temperature changes, the temperature measurement point
reacts after some delay. The analogue electric circuit diagram of the DSC for this
"2nd approximation" is shown in Fig. 3.7. The following is valid for this so-called
"thermometer problem":
TMS temperature of the measurement point (e. g. junction of the thermocouple) for
the sample
Ts (homogeneous) sample temperature
't"2 characteristic time constant for temperature balance between sample and
temperature measurement point.
In analogy, the following is valid for the reference side:
('t"2 is to be equal for both sides).

With these equations, the following results for the difference:
dTMS dTMR)
ATsR = Ts - TR = TMS - TMR + 1'2 (- - - - -
dt dt
(ATM = TMS - TMR: measured temperature difference)
The following relation results in analogy to the mathematical procedure of the 1st
approximation if we use the above ATsR instead of AT:
(3.5)
In addition to the 1st derivative (slope) of the measured curve ATM(t), the 2nd deri-
vative (curvature) must be used in the 2nd approximation to reconstruct the heat
flow rate converted in the sample; two time constants occur which must be deter-
mined. The first time constant 1'J is determined by the thermal resistance and heat
capacity between furnace and temperature sensor. The second time constant is
determined by the thermal resistance and the "effective heat capacity" between
sample and sample temperature measurement point.
The above-described approximation can be refined as desired: In order that
the temperature gradient inside the sample and its influence on the peak shape can
be determined, the sample can be considered as having been split up into different
layers which are linked with one another by heat-conducting boundary layers. An
additional differential quotient in the differential equation and another time
constant results for each additional thermal resistance in connection with a heat
capacity. The solution of the refined approximation is as follows:
1[ dAT d2AT d AT
- + ... ]
3
<lJ.{t)=-- AT+kj,+kJ - + k2 - - +kJ -
R dt dt 2 dt 3
In reality, the constants k; are terms into which the thermal resistances and capa-
cities (and thus the time constants) of the arrangement enter. Calculation of the
reaction heat flow rate <f>r presupposes that all k/ and the measurement signal AT and
its time derivatives are known. It can be shown (L6blich, 1985) that for practical
application, the 2nd order differential equation is a sufficiently good approximation
to calculate the true desmeared reaction heat flow rate; the time constants 1'J and 1'2
must be known for this purpose. For a detailed treatment how to determine the time
constants experimentally, see L6blich, 1994.
So far we have only considered the simplified case where it is assumed that no
heat exchange takes place between sample and reference sample. This simplifica-
tion is certainly not permissible for disk-type measuring systems. Figure 3.8 is a
more realistic representation of the measuring system of such a calorimeter. When
the differential equation is to be set up for this system, it may be convenient to
./;
/
s R T/
L...--_~'-----=:;==-_---' %
-- ¢FR
LJT
Fig.3.8. Disk-type measuring system of a heat flux DSC (to calculate the c1>r(~T) de-
pendence)
Ms temperature measurement point on sample side, MR temperature measurement point on
reference side, c1>r heat flow rate produced/consumed by the sample
Fig.3.9. Equivalent electric circuit for the disk-type measuring system according to Fig. 3.8.
S sample, R reference sample, M temperature measurement point, F furnace, U voltage, i cur-
rent, R resistance, C capacitance (cf.legend to Fig. 3.7)
use the analogue electric circuit represented in Fig. 3.9 for the 2nd approximation
(Hohne, 1983). In order to formulate the desired differential equation, according to
Kirchhoff's laws, the voltage balance and current balance are made up for each loop
and each node of the analogue electric circuit. For the circuit of Fig. 3.9, five equa-
tions for both voltages and currents are then obtained. On the basis of the laws of
the science of electricity, these 10 equations are combined to form a differential
equation and retranslated into the language of heat transport. The result for a sym-
metric twin design (R FMS = RFMR = R, CFMS = CFMR = C) is as follows:
tP. =-
r
(2 +~)tlT- [c- C(1 +
R RMM S
R
MS
(~+_2
R RMM
))] dtlT
dt
This equation is similar to that ofthe 2nd approximation (Eq. (3.5». However, as
there is the thermal resistance RMM between sample system and reference sample
system, a thermal effect in the sample will also affect the reference side and thus TR •
Only in the steady-state case and at sufficiently large distance from peaks is d TR/dt
equal to the heating rate 13, and the second derivative of the reference sample tem-
perature is equal to zero only there. For the steady-state case (st), with
dtlT
tJ)r = 0 -=0
dt
the following is then valid and characterizes the baseline:
CR-CS
tlT., = 13
1 2
-+--
R RMM
When Eq. (3.6) is integrated over the area of a transition peak, with the approxima-
tion d 1i/d t "" 13, the following is valid:
12
J4>,dt=Qr=
I,
( -1 +2)
- ['2JtlTdt-J12 13
R RMM I, I, (1- + -2)- 1
CR - Cs dt
R RMM
1 2)'2
=- ( - + - J(tlT-tlT.,)dt
R RMM It
This integral describes the peak area between measured curve and (interpolated)
baseline. The approximation is the better the smaller the last two summands in
Eq. (3.6). From this follows the rule that, when heats of transition are determined
with heat flux DSCs, the sample and the reference sample should be as similar as
possible (CR "" Cs , RMR "" RMS).
The factor (lIR + 2IRMM ) is decisive for the sensitivity of the calorimeter; the
greater the thermal resistance of the disk, the higher the peak for a given heat of
transition. However, as a result, the time constant increases as well, i. e. the system
becomes more inert. In addition, this factor allows the conclusion to be drawn that
the ratio of R to RMM plays an important role for the sensitivity. Depending on where
sample and reference sample are arranged on the disk, RMM (thermal resistance
t
b,T
1---+- - - - - - - -.:-:::-~-""'""'-"....-------
time
Fig.3.10. Measured signal11T of heat flux DSC for a heat pulse generated in the sample at
the time to (sum of two exponential functions with the time constants T\ and T2)
between sample and reference sample) change, meaning that high reproducibility
for the location of sample and reference sample is of great importance for the capa-
bility of a DSC to be calibrated.
In the approximation Cs "" CR and R MS "" RMR , Eq. (3.6) changes into the 2nd
order differential equation stated below:
d~T d 211T
<P.(t) =-K 'IlT-K\ - -K2 - -
r dt dt2
The solution of this equation, i.e. the curve llT(t) recorded by the calorimeter, for
a heat pulse cPr at t = 0 is the superposition of two exponential functions:
(3.7)
where there is a complex dependence of the time constants 'f\, 'f2 on the coefficients
K, K\, K 2. Here, 'f\ is determined in approximation by Cs . RMS , and 'f2 by C . R
(cf. Eq. 3.6). Such apeak, produced by a heat pulse, is shown in Fig. 3.10. The time
constant 'f\ essentially determines the ascending, 'f2 the descending part of the
curve.
Numerical Simulation
As the equations become increasingly complex and as it is impossible to solve them
analytically without introducing simplifications, it is no longer recommended to set
up differential equations for calculations, which also cover heat exchange by con-
vection and radiation and include non-linearities due to temperature-dependent
thermal resistances and heat capacities. In this case, numerical simulation by the
finite-element method (FEM) should be applied. Appropriate computer programs
are commercially available also for personal computers. When the required time
and effort are spent, the temperature and heat flow fields and the measured I1T(t)
curves can be simulated for any complex arrangement and any thermal process
inside the sample.
The finite-element method consists in splitting the whole arrangement up into
sufficiently small "cells" for which the heat flow rates and the material properties
are defined. On the basis of considerations with regard to the energy balance, a
system of equations is obtained which is solved by conventional methods. FEM has
been applied to a very simple model of a disk-type measuring system and to a com-
mercial disk-type DSC (see Hohne, 1983).
The results are represented in Figs. 3.11 to 3.13. As can be seen, the shape of
the peak depends strongly on the test parameters. The position of the peak maxi-
mum, for example, changes with the heating rate, the thermal conductance of
the sample and with the mass (or heat of transition Qr) of the sample. The slope of
the ascending part is determined by the heating rate and the thermal conductance of
the sample. Only the extrapolated peak onset temperature is relatively independent
of test parameters. This is why, in addition to the area, this temperature
(1;,) is preferred to characterize the temperature of a peak. In contrast to this, the
peak temperature maximum (1;') and the peak width are no values suited to
characterize transitions (for definition of the characteristic temperatures, see
Sect. 5.1).
Figure 3.14 shows another result of the numerical simulation of the thermal
events in a commercial disk-type DSC, namely the DSC signal which is caused by
the melting of a small sample of indium.
It has also been possible to show that the calibration factors KQ = I QJpeak area I
and KtIJ = I CPr/I1TI clearly depend on test parameters (Q" A.ample, 13) when radiation
time ~
t
/:::,T
Fig. 3.11. Dependence of the peak shape on the heating rate {J, calculated by numerical simu-
lation for a heat flux DSC
time ..
f
/j,T
Fig.3.12. Dependence of the peak shape on the heat of transition Qro calculated by numerical
simulation for a heat flux DSC (heating rate: 2 K min-I)
time ..
L=O.OOS W/K
Fig.3.13. Dependence of the peak shape on the thermal conductance L of the sample, calcu-
lated by numerical simulation for a heat flux DSC
0.25
fj, TIK
0.20
0.15
0.10
0.05
o
o 25 50 75 100 s 125
time ..
Fig. 3.14. Numerical simulation of the signal (absolute values) generated by the melting of an
indium sample in a disk-type DSC (1.38 mg In, 10 Kmin- I )
and convection are included (Hohne, 1983), for an example see Table 3.1. This is a
fact which is of importance in practical scanning calorimetry and which must be
taken into consideration during calibration (cf. Sect. 4.4). The clear temperature
dependence of the calibration factor is another conspicuous feature although the
temperature dependence of the heat capacities and of the heat resistance of the
substances has not been included in these model calculations. The reason is to be
found in the intensive radiative heat exchange with the surroundings (heat leak),
which is strongly temperature-dependent.
The calibration factor is also strongly influenced by the surface quality (emis-
sivity) of the containers of sample and reference sample.
As far as steady-state reaction heat flow rates are concerned, the above calcula-
tions have shown that the calibration factor K~ = - cI>/aTst also depends on the
amount of cI>r (or the difference between the heat capacities, Cs - CJ when radiative
exchange is included. In addition, the calibration factors for a peak integration
(determination of Qr) and for steady-state heat flow rates (determination of Cs) are
generally not equal.
The difference in the calibration factors K~ and KQ can also be concluded from
the following considerations. In a DSC there are always temperature differences
between sample and reference sample during the measurement. As a result, the heat
exchange between sample and reference sample and the respective environment
Table 3.1. Numerical simulation of a commercial disk-type DSC: Calibration factors K. and KQ
depending on the heat of reaction (aQ), emissivity (e), heating rate «(J), density (p) and thermal
conductivity (.t) of the sample *.
The initial standard parameters (belonging to K. = 1 and KQ = 1) are given in parentheses
Parameter which Numerical value of the K.

is varied parameter
aQinJ (0.3965) 1
3.965 1.012
0.03965 0.929
e 0.5 (AI) 1.001 0.915
0.25 (Ag) 1.001 0.888
(JinKs-I (10) 1 I
1 1.001 0.930
0.1 1.004 0.875
0.01 0.991 0.870
p in kg m- 3 (7310)
73100 0.998 0.935
731 1.015 0.885
.tinWK-1m-1 (157) I
15.7 0.998 1.016
1570 1.003 0.999
* The parameters have been changed in steps of a factor of 10 only, this is often not realistic but
gives an impression of their influence.
varies. As the radiation exchange increases, however, non-linearly with tempera-

ture, the overall heat exchange is also non-linearly linked with the temperature
difference between sample and reference sample. As a consequence, the measured
temperature difference (and thus the measured differential heat flow rate cPm ) is
non-linearly linked with the true heat flow rate into the sample cPtrue . This results in
the calibration factor Kq, in the (linearly formulated) Eq. (2.1) becoming a function
of the differential heat flow rate cPm :
(3.8)
As the measured heat flow rate cPm depends on sample parameters (Cp , mass m) and
on the heating rate, Kq, too, implicitly depends on these quantities. Furthermore, it
follows from Eqs. (3.3) and (3.6) that Kq, is determined by the thermal conduction
path and its properties. As these quantities are always a function of temperature, the
calibration factor is always also temperature-dependent.
On the other hand, for the determination of the true heat from the peak area,
Eq. (2.2) is valid, which is obtained from Eq. (2.1) by integration over the peak. If,
however, Kq, depends on cPm this factor is not constant and cannot be placed before
the integral, and the integration ofEq. (3.8) yields:
Qtrue = f( cPtrue - cI\I,true) dt = f(Kq, (cPm ) • cPm - Kq, (cI\I) . cI\1) dt
From a comparison with Eq. (2.2) it follows, that KQ is an integral mean value over
the function K,z. (<P.J in the region of the peak. As <Pm may considerably vary during
a peak, the difference between K,z. and KQ is normally significant (cf. Sect. 4.4.3).
In these cases, a quantitative peak evaluation may lead to results affected by
systematic errors, as the calibration factor also undergoes substantial changes.
Summarizing, on the basis of the results of the numerical simulations, the
following can be stated for heat flux DSCs:
For steady-state conditions:
is valid (cf. Eq. (3.l».

With this function for a known heating rate /3, the unknown heat capacity Cs can be
determined from the measured ilTst • It must be borne in mind that the calibration
constant K,z. depends on the difference of the heat capacities, Cs - CR •
- For non-steady-state conditions (peak):
- the shape of the peak changes as a function of heating rate, thermal conduc-
tivity, heat capacity and shape of the sample, and ofthe amount of the heat of
transition. As a result, the peak parameters (width, maximum, height) change
as well; only the extrapolated peak onset temperature 1'. is to a certain degree
independent of sample parameters,
- the calibration factor K,z. (true heat flow rate/measured heat flow rate) is not
constant but a (weak) function ofthe heat flow rate itself,
the calibration factor KQ (heat of transition/peak area) is not constant but
changes more or less with temperature, heating rate and thermal conductivity
of the sample as well as with the heat of transition, surface quality (emissivi-
ty) and sample location,
the calibration factor for peak areas (KQ) differs from the calibration factor for
steady-state heat flow rates.
It is basically recommended that in DSC measurements the best symmetry possible
between sample and reference sample be ensured (CR "" Cs , RMR "" R MS , identical
sample containers). The results ofthe model calculations are confirmed by practical
measurements (Sarge et aI., 1994).
3.2 The Power Compensation DSC
In ideal power compensation DSCs, each ilT signal appearing between sample and
reference sample would be immediately compensated by a corresponding change in
the heating power. The differential heating power required for this purpose would be
equal to the differential heat flow rate as the complete electric energy is converted
into heat. If the differential heating power was measured and put out, this signal
would directly stand for the heat flow rate into the sample which has been sought.
The calibration factor would then be identical to unity and - without "smearing" of
the measured signal - <Pr = <Pm would be obtained. There is, however, no such ideal
power compensation DSC. In real power compensation DSCs, the sample is always
put into a container and then placed into the heater; at least one heat conduction path
and, as a result, at least one time constant -r between the controlled heater and the
sample location must therefore be taken into consideration. This leads to a mea-
surement signal <Pm which is "smeared" in comparison with the processes inside the
sample, and a differential equation of at least the I st order is therefore obtained
d<Pm
<1>.=<1>.+-r--
r m dt
Without exception, commercial power compensation DSCs are instruments in

which a temperature difference occurs between sample and reference sample (as
is the case with heat flux DSCs). This temperature difference is, firstly, the mea-
surement signal and, secondly, it is used to electrically compensate the measure-
ment effect by means of a proportional controller (cf. Sect. 2.2). However, the pro-
portional control can never completely compensate the measurement effect so that,
even in these calorimeters, a ~TSR remains between the microfurnaces of sample and
reference sample, and this in both the steady state and during a peak.
The following is then valid:
<Pm = - ~rop • ~TSR
with kprop set by the proportional control circuit.

O'Neill, 1964, analyzed the system with a proportional controller; the theoreti-
cal analysis of two systems of power compensation DSC, one with P (proportional)
and one with PID (proportional, integral and differential) temperature control ofthe
sample microfurnace was presented by Tanaka, 1992.
As different temperatures also result in a different exchange of radiation and con-
vection with the surroundings, conclusions similar to those for heat flux DSCs
would have to be drawn. This means that, in principle, what has been stated in the
previous section is also valid for power compensation DSCs. However, it applies to
a lesser extent as, due to the compensation, the temperature differenGes are much
smaller than in heat flux DSCs. For power compensation DSCs at present commer-
cially available, it is therefore basically to be expected that
the shape of the peak depends on sample parameters, the time constant(s) being,
however, smaller than those of heat flux DSCs,
the calibration factor is not exactly equal to 1 (i.e. calibration is necessary),
the calibration factors for steady-state (determination of the heat capacity) and
for peak evaluation are not the same,
the calibration factor depends on the temperature, the heating rate, the thermal
conductivity of the sample, the amount of the heat of transition and on the
sample location.
Practical measurements have confirmed these assumptions (Hohne, Gloggler,
1989). As had been expected, the effects are, however, substantially smaller than
with heat flux DSCs. When power compensation DSCs are calibrated, this depen-
dence on parameters must be taken into consideration.
When heat capacities are measured (see Sect. 6.2.1), the reference microfumace
very often does not contain a sample. Due to the complex heat transfer conditions
in the non-symmetrical DSC measuring system, heat losses then arise which cannot
be compensated. These heat losses are different in static and dynamic operation. An
analysis shows that the dynamic measurement error can be determined by means of
the static losses measured before and after heating (for details, see PoeBnecker,
1993).
Another method to describe the behavior of an apparatus is deduced from the
transfer theory, namely the theory of linear response which can be applied if the
apparatus in question behaves linearly. This is true if it can be described with the aid
of linear differential equations of any order. We have shown in Sect. 3.1 that heat
flux DSCs can be described in this way if radiation and convection heat exchange
are disregarded.
In the case of power compensated DSC, the situation is more complicated be-
cause of the control electronics involved. For the equipment of the Perkin-Elmer
Corp. most frequently used, it has been shown recently that this DSC can be
considered as a linear apparatus in first approximation (Tanaka, 1992; Hohne and
Schawe, 1993; Schawe et at., 1993, 1994). However, any asymmetry between
sample and reference side disturbs the linear behavior, and the theory of linear
response is strictly speaking no longer valid. This has consequences for desmearing
procedures which require linear response, such as deconvolution (Sect. 5.4), in
particular in peak regions of a measured curve.
Note:
In contradiction to what has been presumed in Chapter 3, no real DSC is
strictly symmetric in its functionality. Even in the empty state, the temperatures
of the sample and of the reference cell are not equal so that a residual temperature
difference will arise, which may change with temperature during a scan. As
a consequence, the empty DSC will produce a measured curve, which is indeed
neither zero nor constant with temperature but has an apparatus-dependent shape
tPo(T). This zero line, which will mostly be measured with the crucibles in the calo-
rimeter cells empty, is a function, that is additive to the measurement signal which
stems from the sample processes. In every real case, this function should first be
subtracted from the measured curve before evaluations according to the theoretical
considerations can be made. We therefore must insert the term <Pm - <Po ( or AT - ATo)
instead of <Pm (or AT) in all formulas used so far, this has been omitted in this
chapter for clarity.
4 Calibration of Differential Scanning Calorimeters
There is not, as yet, an international agreement on procedures for the temperature

calibration and caloric calibration of DSCs which have been definitely backed up
both theoretically and experimentally. The results of investigations in this respect
will have to be taken into account in the relevant calibration regulations, and they
also show the basic limits set to the reliability of measurements. To what extent
DSCs are capable of being calibrated depends on the quality of the DSCs' measur-
ing system, on operational parameters and on the availability of precisely measured
calibration materials. Here, the situation is still quite unsatisfactory. Moreover, no
complete theory of the DSCs exists, which takes full account of the systematic
influences from the instrument and inherent error sources. It should be borne in
mind that it is the overall uncertainty of the calibration which yields the smallest
possible systematic uncertainty of the measurements.
4.1 Aspects of Quality Assurance
Thermoanalytical instruments, including DSCs, are extensively used by industry in

quality assurance systems to warrant trustworthy, reproducible properties of the
products, even when we mean products which are the measurement results by which
products are characterized. The ISO 9000 series and the EN 45 000 series are the
basic standards in the field of quality management and quality assurance. The
standard EN 45001 in particular specifies general requirements to be met by test
laboratories. It will no longer be sufficient to have technical and scientific com-
petence, this competence must be documented and made apparent to others. DSC
methods are used for quality assurance in many ways: for example for the inspec-
tion of raw materials, as an accompanying measure in manufacture and for the con-
trol of the finished products.
To enable DSC methods to be useful for quality assurance, they, too, must be
subject to the requirements of the relevant standards. Calibration of the DSC is an
important precondition to fulfil these requirements. Calibration involves the appli-
cation of reasonable calibration procedures and the use of calibration substances
which are - if possible - traceable * to national/international standards. This means:
* Traceability: Property of the result of a measurement or the value of a standard whereby it

can be related to stated references, usually international or national standards, through an
unbroken chain of comparisons all having stated uncertainties (International Vocabulary of
Basic and General Terms in Metrology, 2nd ed.,1994).
42 4 Calibration of Differential Scanning Calorimeters
to create confidence in the quality of the measured values and to avoid repeat mea-
surements, metrologically sound calibration procedures and traceable calibration
substances (Certified Reference Materials) must be developed on the basis of which
reliable uncertainty values can finally be assigned to the measurement results.
In the following sections, calibration procedures and materials will be presented
which are mainly based on investigations carried out by the German Society for
Thermal Analysis (GEFTA). Recommendations for calibrating DSCs and for cali-
bration materials suitable for temperature calibration and caloric calibration are
made. They are regarded as a first step in the development of internationally accept-
ed recommendations (temperature calibration: Hohne et aI., 1990; Cammenga et
aI., 1993; caloric calibration: Sarge et aI., 1994).
4.2 Thermodynamic Aspects

In the following, reference will be made to some important thermodynamic aspects
of calibration. The same considerations are valid for the phase transitions of a pure
substance and for chemical reactions.
The quantity assigned to the substance is the enthalpy difference !l.H of the
phase transition (with T = constant); it is the difference between two variables of
state and, consequently, well defined. Inside the calorimeter, it is the heat of transi-
tion Q which is measured. The relation between heat and enthalpy is given by the
first law of thermodynamics and the definition of the enthalpy function:
dU = dQ + dW + L dE; (first law of thermodynamics)
i
dH = dU + P . dV + V· dp (from enthalpy definition)
From this it follows (with dW = -p . dV):
dQ = dH - V . dp - L dE;
i
(U internal energy, W work, E; various forms of energy)
The enthalpy H of a system is a function of the following variables of state: pres-

sure p, temperature T and composition ~:
H=H(P,T,~)
The total differential is:
dH= (OH) dp + (OH) dT + ~H) d~

op T,~ aT p.~ \a~ p,T
Calorimetrically, the differential heat which is measured is:
dQm = [(OH) _V] dp + lOH) dT + ~H) d~ - L dE; (4.1)

op T,~ \oT p,~ \a~ p,T I
Under the conditions normally prevailing during DSC measurements, the first term
can almost always be neglected, even if closed crucibles are used (maximum pres-
sure changes: 1 to 2 bar, dp "" 0). The second term is the heat capacity at constant
pressure and constant composition ofthe (reacting) system:
Cp.~ (T) =oT

-(OH)
p.~
The third term in Eq. (4.1) is the isothermal and isobaric enthalpy change due, for
example, to a phase transition (1lH = ~trs H), a mixing effect (~H = ~mixH) or a
reaction (~H = ~rH).
Only if p "" const. and L dE; = 0 and if the Cp share is known (the correct base-
;
line in the course ofa thermal event, cf. Sect. 5.3) is the following (obtained by in-
tegration ofEq. (4.1)) valid for the melting ofa pure substance:
~H= Qfus
i. e. the heat of melting equals the enthalpy difference only if the above restrictions
are valid. A change in the surface already takes place if many small single crystals
fuse together and form a big bead of the melting substance. As a result, surface heat
is exchanged (dEsurface =1= 0) and Eq. (4.1) becomes ~Q = ~fusH - ~surfaceE.
Almost the same is valid when deformation energy is exchanged during melting
or when there is an interaction between the melted substance and the surface of the
sample pan (wetting energy). Though the difference between the heat of fusion
~fus Q and the enthalpy difference ~H is small, these contributions must not be basi-
cally ignored. It is furthermore pointed out that ~H is defined only for the in-
definitely great "phase". For smaller phases (small crystals), ~H depends on the
size of the sample grains (crystal size) which means that the specific heat of
melting of a fine powder differs from that of a large crystal. (As a result, the tem-
perature of melting changes as well.)
By differentiation of Eq. (4.1), the following is obtained for the heat flow rate
under ideal conditions (p "" const., LdE;"" 0).
;
(4.2)
This equation forms the basis for the kinetic evaluation of DSC curves (cf.
Sect. 6.2.2.2).
4.3 Temperature Calibration

Temperature calibration means the unambiguous assignment of the temperature
"indicated" by the instrument to the "true" temperature.
The ''true'' temperature is defined by fixed points with the aid of calibration sub-
stances. It is reasonable to choose as calibration substances, if possible, the sub-
stances used to realize the fixed points of the International Temperature Scale of
1990 (ITS-90). The temperature "indicated" by the instrument must be derived from
the measured curves, which usually requires extrapolation to zero heating rate in
order to eliminate/minimize the influences of instrument and sample parameters, if
possible.
Static methods (thermodynamic equilibrium) are applied to realize the fixed
points of the temperature scale. In a DSC, these can be achieved only approximate-
ly. As the point of temperature measurement is not the point where the sample is
located, a systematic error will always occur in a scanning operation which depends
on instrument and experimental parameters. The calibration procedure described in
the following takes these special features into account in a general way, independent
of the special DSC type (see Hohne et aI., 1990).
After calibration has been completed, the potentiometer provided for this pur-
pose will either be adjusted until the temperature indicated corresponds to the true
one, or adaptation will be ensured via the internal computer program (in a way
usually not apparent to the user), or a graph or table is established showing the
relation between the indicated and the true temperatures. In each case, a table
should be drawn up which shows the variation of the indicated temperature at
different heating rates.
A calibration already carried out by the manufacturer must be checked. Regular
calibrations provide important information about the repeatability error and any
\ I
(J)\,'(J)
~
t I
I
I
'"en
c
-Vi
I
I
I
I
I \
I---<Y'~ - - - ~ - - - - - ~ ---=--0-----
I, 'I I 1\ ,I
I I T I I
I
I
I
I CD I
I
I I I
I I I
T, Te Tp
tempera ture
Fig.4.1. Heat flow rate signal of a DSC during a transition (schematic representation, acc. to
Hemminger, 1994).
CD baseline (interpolated), <2l auxiliary lines, T; initial peak temperature, T. extrapolated peak
onset temperature, Tp peak maximum temperature, 1'c extrapolated peak completion tempera-
ture, Tr final peak temperature
long-term systematic variations (drift). (It should be noted that repeatability error is
not the same as uncertainty of measurement, cf. Sects. 7.2 and 7.3).
In the case of an endothermic event, the DSC records the heat flow rate signal
schematically shown in Fig. 4.1. The section between the initial peak temperature 1';
and the final peak temperature Tfis defined as peak (for definitions see Sect. 5.1).
The interpolated baseline is interpolated by various methods between 1'; and Tf
(cf. Sect. 5.3). The intersection between the auxiliary line and the baseline suffices
to fix the characteristic temperature T. (extrapolated peak onset temperature) which
is important here.
4.3.1 Calibration Procedure
Selection of at least 3 calibration substances (cf. Sect. 4.6.1) which cover the
desired temperature range as uniformly as possible (at least 3 in order that
nonlinearities can be detected),
at least two calibration samples of each substance are prepared (for repeat mea-
surements). The sample mass should correspond to that commonly used in
routine measurements,
- the transition has to be measured with each calibration sample at at least 5 differ-
ent heating rates in the range of interest, including the smallest possible one. The
second calibration sample of the same substance is also measured at different
heating rates (cf. the example below),
it has to be checked whether there is a significant difference between the charac-
teristic temperatures (especially 1'.) obtained at identical heating rates for the
first and second calibration sample ofthe same substance. If necessary, it should
be checked whether the temperatures depend on other parameters (mass, location
of the sample in the crucible etc.),
if this is not the case, T. is represented as a function of the heating rate and the
extrapolated value T. (f3 -+ 0) determined for zero heating rate (Fig. 4.2),
the difference t1Tcorr (f3 = 0) between the value 1'.(f3 -+ 0) obtained in this way
and the respective fixed-point value Tfix or the value lIit taken from the literature
(cf. e.g. Marsh, 1987; Cammenga et aI., 1993) is either used to change the instru-
ment calibration according to the manufacturer's instructions or it enters into a
calibration table or curve (cf. Fig. 4.3),
- if T. depends not only on the heating rate but also on other parameters, these
dependences should be represented accordingly (location of the sample in the
container, position of the sample container in the measuring system, open/closed
sample container, sample mass, sample shape (foil, bead), atmosphere, material
of sample container etc.).
In each case, a table (or graph) should be made which shows the variation of the
indicated temperature (or of that read from the measured curve) in relation to the
true temperature at different heating rates (cf. example below).
Temperature calibration has then been completed.
How are the correct temperatures assigned in practice?
{3 ..
Fig. 4.2. The extrapolated peak onset temperature T. as a function of the heating rate f3, and
construction of T. (f3 ~ 0).
b.T.lb.f3 variation of T. with f3; <D, ~ different calibration substances
t
Fig.4.3. Temperature correction b.Teorr (f3 = 0) as a function of the extrapolated peak onset
temperature 1". (f3 ~ 0) for three calibration substances (<D, ~, @).
b.Tcorr (f3 = 0) is the difference between T. (f3 ~ 0) (cf. Fig. 4.2) and the "true" value of the
temperature of transition. The curve obtained by means of (at least) three calibration sub-
stances shows the corrections to be applied to the measured values 1". (f3 ~ 0) at different
temperatures
1. When the accuracy requirements for temperature measurements are high (e. g.
thermodynamic investigations), the substance to be investigated is measured at
various heating rates. The desired characteristic temperature (e. g. Te) is deter-
mined by extrapolation to {3 = 0 (quasi-equilibrium temperature). The correction
I!1Tcorr ({3 = 0) is applied using the proper calibration table or curve (cf. above):
Ttrue = Te ({3 ~ 0) + I!1Tcorr ({3 = 0)
2. When the accuracy required in the determination of Te is not so high and/or the
process to be investigated depends strongly on the heating rate ("kinetic" pro-
cesses), the value 1'.({3 -+ 0) may be calculated from the mean slope tlT.ltl{3 of
the 1'.(fJ) curve(s) of the calibration substance(s) (cf. above and Fig. 4.2 and
example below):
tlTcorr ({3 = 0) is again taken from the calibration table or curve (Fig. 4.3), so that
The best thing to do is to start by carrying out two measurements at clearly differ-
ing heating rates. Then it is checked whether tlT.ltl{3 corresponds with the (mean)
slope of the T.({3) curve(s) of the calibration substances. If so, extrapolation to
T.({3 -+ 0) can be carried out at once. Ifnot, first the heating rate dependence T.({3)
must generally be determined as described above.
To simplify the described method, for each heating rate applied, the correspond-
ing overall correction tlTcorr (f3) = tlTcorr {f3 = 0) - {3 . tlT.ltl{3 can be listed (cf.
example below) so that the "true" temperature can be determined at once:
Ttrue = Tm + tlTcorr(f3)
Here, a distinction by classes of substances (metals, organic substances) must pos-

sibly be made.
It is to be expected that, for a certain instrument, tlTcorr ({3 = 0) will vary with
time which is why a recheck should be made about every three months by re-
calibrating the instrument. The dependence of the extrapolated peak onset tem-
perature tlT.ltl{3 on the heating rate is only related to the properties of the sample
substance; it remains unchanged for a certain DSC and need not, therefore, be
regularly checked.
In order to assign the correct temperature to the individual phases in the case of
complex thermal events, with {3 =1= 0, an auxiliary line must be drawn at the angle a
to the extrapolated baseline. The angle a can be taken from the calibration experi-
ment in which the same heating rate has been applied (Fig. 4.4). The scale division
ofthe axes must be the same in both cases; otherwise the slope of the auxiliary line
must be converted. (For more advanced methods of determining characteristic tem-
peratures in the case of endothermic and exothermic events, see Sarge, 1991 and
Schawe, 1993).
Uncertainties
Depending on the type ofDSC, the minimum repeatability error of the determina-
tion of 1'. on pure metals amounts to approx. ± 0.02 K (sample exactly in the same
place in the crucible or measuring system); in the other cases it varies between 0.1
and 0.8 K. An overall uncertainty of measurement of T. between 0.3 and 1.0 K must
be reckoned with. The overall uncertainty of the temperature calibration should in
every case be carefully estimated (uncertainty of temperature sensors, uncertainty
of the determination of T. etc.).
I
/
CI
c:
en
VI
a tempera ture
CI
=
C
en ~_ _ _
r" T,2
b tempera tUre
Fig.4.4. Assignment of the characteristic temperature Tp in complex thermal events.
a Calibration measurement to determine the angle a at a specific heating rate,
b Construction of characteristic peak maximum temperatures Tp of a complex thermal event
with the aid of angle a (measurement with the same heating rate as for a). In heating opera-
tion, due to the thermal lag, each endothermic thermal event in the sample is "indicated" too
late, i. e. at too high a temperature. To find the "true" temperature of a characteristic segment
of the measured curve in good approximation, a lower temperature must always be assigned
instead of the indicated temperature (read at an angle of 90°). This is done with the aid of
the angle a. In the case of exothermic events the true temperature can be higher than the
measured temperature (cf. Tpl)
Notes:
1. It has to be checked whether T. depends on the location of the calibration sam-
ple in the sample container (in particular at high temperatures). In the case of
power compensation DSCs, due to the isoperibol operation, a temperature pro-
file develops in the bottom plate of the microfurnace which - depending on the
sample position - results in an earlier (center) or later (boundary) melting of a
calibration sample - and thus in a lower or higher temperature T. being indicated
(Fig. 4.5) (see Hohne, Gloggler, 1989).
Such effects are also possible in heat flux DSCs, however, to a much lower
extent due to the non-isoperibol operation. In cylinder-type measuring systems,
above all the dependence on the (vertical) position of the sample in the cylindri-
cal container must be checked.
2. In the case of exothermic events, the sample temperature can be higher than the
measured temperature. The amount of this deviation cannot be precisely deter-
mined. The assignment of a temperature is therefore useful only at the beginning
of the exothermic event.
The increase in the sample temperature in the course of the reaction can be
estimated on the basis of the following considerations:
~i I
i ~
I
I
i
I I
I I
T k--L~ ~us"n
:.A ~t2
I I t,
r -<>
Fig. 4.5. Temperature profile in the micro furnace of a power compensation DSC (schematic).
The isoperibol mode of operation (constant ambient temperature) results in a temperature pro-
file in the sample support as shown schematically. During melting of a sample, this tempera-
ture profile leads to time-dependent zones of melting (Tfus is first reached in the center) and
consequently to different extrapolated peak onset temperatures T. when the radial position (r)
of small samples differs (acc. to Hohne, Gloggler, 1989)
r¢
b
Fig. 4.6. Heat resistance l4. between temperature measurement point and sample.
a Diagrammatic view of the sample system,
b Curve measured during melting of a pure substance, <1> heat flow rate into the sample, tan r
ascending slope (d<P/dt) of the <1>(t) curve during melting (to calculate R th )
If there is a thermal resistance RIb between sample (Ts) and temperature sen-
sor (TM ) (Fig. 4.6a, b),
(4.3)
is valid for the difference between the indicated temperature TM - calibrated as
described above - and the sample temperature Ts considered to be homogeneous.
When a pure substance melts (calibration, for example), RIb can be determined
approximately from the slope of the respective heat flow rate curve tP(t). From
Eq. (4.3), the following results after differentiation (with Ts = constant during
melting, dTM/dt "" Pand tan r = (dtP/dt)melting):
P
Rlb " , , - - (4.4)
tan r
(Experiments at various heating rates will show how well this relation is valid.)
3. In measurements at negative heating rate (cooling), the sample temperature is

higher than the indicated temperature. As a result, the correction IlT(f3) from
the calibration table or curve must be applied with the sign reversed as com-
pared with heating (see Fig. 4.7), whereas the correction at heating rate zero
ll1;,orr(f3 = 0) remains unchanged.
There are no calibration materials so far with well-defined transition tem-
peratures in the cooling mode as the process of undercooling of pure calibration
substances is not definitely known. Temperature calibration in the cooling mode
is therefore not possible. However, symmetry of the heat transfer phenomena
has been generally taken for granted, at least in heat flux DSCs. In the case of
power compensation DSCs, the situation is more complicated because of the
electronic and control equipment involved. This type ofDSC may exhibit asym-
metric behavior if the electronic system is not precisely adjusted. Whether or not
a DSC behaves symmetrically in the heating and cooling mode can be tested with
the aid of substances which show phase transitions without undercooling. The
temperature of the phase transition must not be known very precisely; it should
only be certain that it is the same in the heating and cooling mode, and that
undercooling is below 0.2 K. This is the case for the smectic/nematic transition
of certain liquid crystals (see, for example, Hohne et aI., 1993). Such a test of
three different power compensation DSCs is documented in Fig. 4.8. As can be
seen, symmetry due to heating and cooling ofthe instrument in question is given
within the limits of accuracy of the measurements, however, a small deviation
from exact symmetry is also apparent. When the electronic system is incorrect-
ly adjusted, the asymmetry of this type of calorimeter can indeed be important
-'"
QJ
'-
=>
'-
Q,I
-
CJ..
E
Q,I
___ _ ____ -.!eift=Ol

[~~~~//T,~
...-
./ I
I
I
.. {3COOI
cooling ra te
o {3heat
hea ting ra te ~
Fig. 4.7. Schematic representation of the temperature corrections in the heating and cooling
mode
340.6
340.2
f. 340.0
339.8
....!:
339.6
339.4
339.2 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 K/min 2.5
Fig. 4.8. Test of the symmetry of three different power compensation DSCs with respect to
the heating and cooling mode using a liquid crystal (acc. to Hohne et aI., 1993)
and may lead to errors in the temperature values determined in the cooling mode.
In the case of heat flux DSCs (in particular those of the disk-type), the asym-
metry need not be detected because of physical reasons. Nevertheless, it is not a
mistake to check it as well.
In general, the assignment of the characteristic temperatures of cooling curves
is affected by greater uncertainties than in the case of heating curves.
Investigations using liquid crystals as calibration substances are being carried
out at present.
4.3.2 Arguments in Support of the Temperature Calibration Procedure
The existing recommendations for the temperature calibration ofDSCs (e. g. ASTM
E 967-83, ASTM E 794-85) and the specifications of most manufacturers take a
standard heating rate (e.g. 10 Kmin- I ) as a basis. This method cannot be recom-
mended since
different heating rates require different corrections; shifting of Te is not always

linear due to the heating rate (cf. Fig. 4.9),
the sample temperature is equal to the measured temperature only at zero heating
rate; to ensure safe extrapolation to zero heating rate, measurements should be
carried out at at least 5 different heating rates (starting with the lowest heating
rate),
the temperature fixed points of the International Temperature Scale are defined
for the reference substances in phase eqUilibrium (i. e. static, zero heating rate).
334
"(
326~ __~__~__~__~__~__~__~___
o 10 20 30 K/min 40
{3 •
Fig.4.9. Extrapolated peak onset temperature T. in °C as a function of the heating rate P(two
measurement series of a heat flux DSC with 58 mg of lead).
0-0 measured values, ---- curve of average values. The curves show a non-linear depen-
dence T.({3) only towards the highest heating rate and a dispersion of the T.-values, which
depends on p
2
<Pm
~ 4
6
300 320 340
tempera ture
360
. 380 D( 400
Fig.4.10. Measured curves showing the peak temperature maximum Tp changing with the
heating rate p (heat flux DSC, lead, 58 mg, heating rate p from 5 to 50 K min-I).
tl>m heat flow rate (arbitrary units). In addition to the shifting of Tp with p, the great changes of
T. and Tr are obvious (for definitions see Fig. 4.1)
Only Te can be used as a characteristic temperature of a peak to define the DSC's

temperature scale (cf. Fig. 4.1), the reasons being the following:
Ii cannot be determined with the required reliability because of the noise; the
same applies to Tf (repeatability error between 2 and 15 K depending on sub-
stance, kind oftransition etc.),
Tp and Tc strongly depend on the thermal conductivity, mass and layer thickness
(volume) of the sample substance, on the heating rate and the heat transfer from
sample to sample container (crucible), which may change due to melting (cf.
Fig. 4.10),
1'. depends least on heating rate and sample parameters (substance, thermal con-
ductivity, mass, layer thickness); any possible effect of melting (heat transfer)
should be checked by carrying out various similar experiments with the same
calibration sample.
4.3.3 Examples of Temperature Calibration
The results of the temperature calibration of a power compensation DSC according

to Sect. 4.3.1 are given as an example. Two samples each (between 2 and 6 mg) of
highly pure indium, tin and lead were used (closed aluminium crucibles). The
results shown in Tables 4.1 to 4.3 were obtained for the non-calibrated device (cf.
Figs. 4.11a to c):
The correction curve shown in Fig. 4.12 has been obtained for the calibration of
the device.
Table 4.1 Temperature calibration with indium (cf. Fig. 4.11 a)
sample No. run No. heating rate in K min-I Te in °C
1 1 10.0 156.21
1 2 10.0 156.16
2 1 10.0 156.25
2 2 10.0 156.21
1 1 5.0 155.81
2 I 5.0 155.91
1 1 2.5 155.65
2 1 2.5 155.67
1 1 1.0 155.64
2 1 1.0 155.65
1 1 0.5 155.52
2 1 0.5 155.65
2 2 0.5 155.53
1 1 0.1 155.49
2 1 0.1 155.20
Results:
Te (13 -70) = 155.523 °C
Tfi• = 156.5985 °C (ITS-90)
t..Tcorr(f3 = 0) =+ 1.0755 K
t..T/t..f3 = 0.0682 K/(Kmin-I)
Table 4.2 Temperature calibration with tin (cf. Fig. 4.11 b)
sample No. run No. heating rate in K min-I T. in °C
1 1 10.0 232.35
1 2 10.0 232.18
2 1 10.0 232.06
1 1 5.0 231.81
2 1 5.0 231.69
1 1 2.5 231.54
2 1 2.5 231.41
1 1 1.0 231.33
2 1 1.0 231.22
1 1 0.5 231.28
2 1 0.5 231.12
1 1 0.1 231.26
2 1 0.1 231.12
Results:
1'. (fj-70) =231.184°C
Tfi • = 231.928 °C (ITS-90)
AToorr(fj = 0) = +0.744 K
AT.!Afj = 0.103 KI(Kmin- l )
Table 4.3 Temperature calibration with lead (cf. Fig.4.11c)
sample No. run No. heating rate in K min-I T. in °C
I 1 10.0 328.22
1 2 10.0 327.50
1 3 10.0 327.52
2 1 10.0 327.95
2 2 10.0 327.87
1 1 5.0 327.10
1 2 5.0 327.09
2 1 5.0 327.42
1 1 2.5 326.94
2 1 2.5 327.23
1 1 1.0 326.76
2 1 1.0 326.97
1 1 0.5 326.78
2 1 0.5 326.95
1 1 0.1 326.75
2 1 0.1 326.93
Results:
T.(fj-70) =326.793°C
Tfi • = 327.462 °C (converted from the IPTS-68 to the ITS-90)
AToorr(fj = 0) = +0.669 K
AT.!Afj = 0.0997 KI(Kmin- l )
156.5
~(p ..OI = 155.523 t
=156.614t~
M",IP=OI= +1.111 K
o
To ~ In
155.5 0 To: 155.523 + 0.0682 • P
5 K/min 10
a p-
232.5
T.lp ..OI = 231.1841 '(
8
'( Til. = 231.9681'(
t:..Tc."IP= 01= +0.784 K
I
232.0
To
m5
Sn
T.: 231.1841 + 0.103 • P
231.0
5 K/min 10
b p-
328.0
328.0
f 327.5
Te
Fig.4.11. Extrapolated peak onset temperature T. (in 0c) as a function of the heating rate {3
to determine T.({3 ~ 0). (Power compensation DSC, two measurement series each with two
samples of different mass, closed aluminium crucibles).
a Sample material: indium, b Sample material: tin, c Sample material: lead. T.({3 ~ 0) extra-
polated peak onset temperature at zero heating rate, Tfix fixed-point temperature, i.e. true tem-
perature of melting (ITS-90, in the case of lead converted from the IPTS-68 to the ITS-90),
ATeoc, = T fix - T.W ~ 0) temperature correction
1.4
K In
1.0
r
.6 Teorr (P=O)
0.6 Sn
Pb
0.2
150 200 250 300 O( 350
Fig.4.12. Correction curve for the temperature calibration of a power compensation DSC.
The individual corrections dTcocr (fJ = 0) have been taken from the results of the calibration
measurements shown in Figs. 4.11 a to c.
dTcorr(fJ = 0) difference between true temperature (fixed-point temperature) and extrapolated
peak onset temperature T.(fJ = 0) for zero heating rate (for Pb see legend of Fig. 4.11),
---mean value: +0.83 K
In practice, for the purpose of temperature calibration, the potentiometer adjust-

ment will in this case be changed by the mean value of 0.83 K and, if necessary, the
shape of the correction curve towards higher temperatures fixed on the basis Of
another calibration (e. g. with Zn). Different corrections must possibly be applied in
different temperature ranges. For many DSCs, the user feeds the results of the
calibration measurements (for example, with In and Pb) into the computer which
internally converts the temperature scale (slope and position).
After the DSC has been calibrated with the aid of tlTcorr (13 = 0), various pure sub-
stances are measured at different heating rates and T. is determined. Table 4.4 is
obtained and includes tlTcorr (f3 = 0) = ·Iiil - T. (13 ~ 0), the corrections for the
various heating rates calculated from tlTcorr (f3 = 0) and the slope ofthe respective
fitted line:
(The values for tlTcorr (f3) have in general been rounded to one decimal.)
tl Ttol is defined as Iiil - T. (13), i. e. the difference between the ''true'' temperature
Iiil and the extrapolated peak onset temperature at a heating rate 13.
For this calorimeter there is a linear relation between the shift of T. and the hea-
ting rate 13. The gradient tlT.ltlf3 ofthe respective fitted line yields the values given
in Table 4.5.
From this, it follows that for this DSC, at a heating rate of 10 Kmin- I , a mean
difference of 0.9 ± 0.3 K results between T. (13) and the true temperature of the tran-
sition. Depending on the heat transfer between substance and crucible bottom, the
shift may vary between 0.5 and 1.5 K.
Table 4.4 Measurements with various substances to determine 6.T/6.fJ
Diphenyl ether (phenoxybenzene)
To in heating rate in 6.T,ot (fJ) in

°C Kmin- 1 K
27.05 0.5 -0.19 IIit = 26.86 °C

27.10 1.0 -0.24 T.(f3 ~ 0) = 26.93 °C
27.20 2.5 -0.34 6.Tcorr {f3 = 0) = - 0.Q7 K
27.73 5.0 -0.87
Gallium
To in heating rate in 6.Ttot (fJ) in

°C Kmin- 1 K
29.92 0.5 -0.16 IIit = 29.7646 °C = TCix

29.97 1.0 -0.21 T.(fJ ~ 0) = 29.851 °C
30.11 2.5 -0.35 6.Tcorr (fJ = 0) = -0.09 K
30.35 5.0 -0.59
30.94 10.0 -1.18
C33H68 (paraffin) 1st transition
To in heating rate in 6.Ttot (fJ) in

°C Kmin- 1 K
67.35 0.5 0.05 IIit = 67.4 °C

67.36 1.0 0.04 T.(fJ ~ 0) = 67.3 °C
67.47 2.5 -0.07 6.Tcorr (fJ = 0) = +0.1 K
67.66 5.0 -0.26
68.06 10.0 -0.66
C33H68 (paraffin) 2nd transition
To in heating rate in 6.T,,,, (fJ) in

°C K min- 1 K
70.98 0.5 0.12 IIit = 71.1 °C

70.98 1.0 0.12 T.(fJ ~ 0) = 70.91 °C
71.10 2.5 0 6.Tcorr (fJ = 0) = +0.2 K
71.27 5.0 -0.17
71.70 10.0 -0.60
Benzoic acid
To in heating rate in 6.T,ot (/3) in

°C Kmin- 1 K
122.28 0.5 0.06 IIit = 122.34°C

122.23 1.0 0.11 T.(fJ ~ 0) = 122.22°C
122.34 2.5 0 6.Tcorr {f3 = 0) = +0.12 K
122.46 5.0 -0.12
122.72 10.0 -0.38
Table 4.4 (continued)
Indium
T.in heating rate in aT,ot (fJ) in

°C Kmin- 1 K
156.63 0.5 - 0.03 lIit = 156.5985 °C = Tfix

156.67 1.0 -0.D7 T.(fJ ~ 0) = 156.53 °C
156.80 2.5 - 0.20 aTcor,(fJ = 0) = +0.07 K
156.96 5.0 - 0.36
157.32 10.0 - 0.72
158.01 20.0 - 1.41
159.34 40.0 - 2.74
162.17 80.0 - 5.57
168.60 160.0 -12.00
Tin
T. in heating rate in aT,,,, (fJ) in

°C Kmin- 1 K
232.18 0.5 -0.25 lIit = 231.928 °C = Tfix

232.21 1.0 -0.28 T.(f3 ~ 0) = 232.12 °C
232.34 2.5 -0.41 aTcorr(fJ = 0) =-0.19K
232.52 5.0 -0.59
233.01 10.0 -1.08
Caffeine
T.in heating rate in aT,,,, (/3) in

°C Kmin- 1 K
236.30 0.5 -0.20 lIit= 236.1 °C

236.30 1.0 -0.20 T.(fJ ~ 0) = 236.24 °C
236.36 2.5 -0.26 aTeot,(fJ = 0) =-0.1 K
236.67 5.0 -0.57
236.96 10.0 -0.86
Lead
T. in heating rate in aT,,,, (fJ) in

°C Kmin- 1 K
237.99 0.5 -0.53 lIit = 327.462 °C

238.00 1.0 -0.54 T.(f3 ~ 0) = 327.93 °C
238.17 2.5 -0.71 aTcorr(fJ = 0) = -0.47 K
238.32 5.0 -0.86
238.81 10.0 -1.35
Table 4.5 The dependence of T. on the heating rate /3 for various classes of materials (results
from Table 4.4)
Substance ilT.lil/3 in KJ(K min-I)
Gallium 0.11
Indium 0.07
Tin 0.09
Lead 0.09
Mean value metals 0.09±0.02*
Diphenyl ether 0.15

Paraffin C33H68 0.08
Benzoic acid 0.05
Caffeine 0.07
Mean value organic substances 0.09±0.04*
Overall mean value 0.09±0.03 *
* standard deviation 0"._1
Separate measurements at different heating rates must be carried out for a sub-
stance in order to determine the transition temperatures more accurately.
Note:
The values obtained for Yo in the first measurement of a calibration sample (In, Sn,
Pb) are systematically higher than those of the second and all subsequent measure-
ments (at a given heating rate). The reason for this is the heat transfer between the
sample and the bottom of the crucible, which strongly improved after the first melt-
ing (larger contact area). As a result, Yo is lower. If, for reasons of irreversibility of
the process, only the first measurement of a sample can be evaluated, this effect
must also be taken into account when the uncertainties of T.(f3 ~ 0) are estimated.
Another striking feature of the results of the calibration measurements is that the
individual 1'. of various samples of the same substance are very well situated on a
straight line, but that some of the straight lines deviate strongly from one another
(e. g. Figs. 4.11 a to c). The reason is that, firstly, the sample containers and thus the
heat transfer to the microfurnace are not identical, i. e. each sample encounters
different contact points and heat flux conditions, and that, secondly, the location of
the sample in the microfurnace of the power compensation DSC clearly influences
the 1'. of the measured curve. This is obvious from the results shown in Fig. 4.13 (cf.
also Hohne, Gloggler, 1989).
4.4 Caloric Calibration 61
... (0
328.0 Pb 0
x ... Q)
K o"CV
327.8 ". . CD
327.6
r
Te
327.4
o 2 4 6 B K/min 10
(3 •
Fig.4.13 Extrapolated peak onset temperatures To of lead (0.2 mg) as a function of heating
rate f3 with the samples in various positions (l to 4) in the microfumace of the power com-
pensation DSC (cf. Fig. 4.5, acc. to Hohne, Gloggler, 1989)
4.4 Caloric Calibration

By means of caloric calibration (for a review see Sarge et aI., 1994), the propor-
tionality factor between the measured heat flow rate <Pm and the true heat flow
rate <Ptrue and between the measured exchanged heat Qm and the heat Qtrue really
transformed is to be determined:
<Ptrue = K.· <Pm and Qtrue = KQ . Qm
Strictly speaking, <Pm in this equation should be considered the measured heat flow
rate with the instrument zero line (which is constant) already subtracted, but as all
correction calculations are usually done using the measured curve <Pm itself, only <Pm
is used in the following.
This calibration is carried out either as "heat flow rate calibration" in the
(quasi-) steady-state
- by electrical heating applying the well-known power,
- by "charging" the known heat capacity of the calibration sample
or as "peak area calibration" by integration over a peak which represents a known
heat
- by electrical heating applying the well-known energy,

by applying the known heat resulting from a phase transition (melting) of a pure
substance.
Since Qtrue = f cJ>true . dt and Qm = f (cJ>m - <1\.) dt, K,p and KQ should be identical which
is not, however, the case because in practice - throughout the duration of the
peak - K,p depends on the temperature T (and therefore also on the time t) and is in
addition a function of cJ> (cf. Sect. 4.4.3). As a result, the equation cJ>true = K,p' cJ>m can
indeed be integrated but K,p must not, however, be placed in front of the integral.
As stated already, KQ is not equal to K,p; KQ is rather a kind of integral mean value
of K,p over the area of one peak. In practice, the difference between the two calibra-
tion factors is between 0.5 and several per cent. Both types of calibration must
therefore be carried out separately.
The advantages of the twin principle ofDSCs become fully effective only in the
case of perfect thermal symmetry of the measuring system. In this case, the mea-
sured signal is, however, zero. In real measurements there will always be asym-
metries in the temperature field. The effects of such asymmetries are dealt with in
Sect. 4.4.3. The conclusion to be drawn for heat calibration is as follows:
The thermophysical behavior of calibration sample and sample to be measured
must be as similar as possible. As this is possible only approximately, systematic
errors exist which must be estimated and included in the overall uncertainty of mea-
surement (cf. Sect. 7.3).
4.4.1 Heat Flow Rate Calibration
In almost all DSCs commercially available, a heat flow rate signal cJ>m is internally
assigned as measurement signal for the actual signal AT. (When the measurement
signal is put out as a voltage AU (for example in mY), the following applies analo-
gously, cJ>m being replaced by AU.) The heat flow rate calibration defines the func-
tional relation between cJ>m and the true heat flow rate cJ>true consumed or generated
by the sample:
(steady state, cJ>m with the zero line - empty crucibles - (definitions see Sect. 5.1)
already subtracted).
The proportionality factor (calibration factor, calibration function) K,p usually
depends on parameters such as temperature and - what is most important - heat
flow rate. In some DSCs, K,p is made unity by electronic or software means. In these
cases, too, the relation between cJ>m and cJ>true must be carefully checked.
The heat flow rate calibration can be performed in two ways:
1. Installation of an electric calibration heater in the place of the sample or inside

the sample. This method offers the following advantages:
the electric power (heating power) can be measured easily and with high accu-
racy,
heat flow rates of differing intensity can be generated without the calibration
setup being modified,
the steady state can be adjusted for any period of time desired; the resulting
conditions are most similar to those of a Cp measurement,
the heater can be switched on or off at any temperature desired so that the
position of the baseline (<Ptrue = 0) can also be checked in between (even with
a sample placed in the crucible),
by appropriately presetting the development of the heating power with time,
measurement effects (peaks) can be "repeated" (simulated) so that heat flow
rates leading to such a peak can be assigned without "desmearing" (see Sect.
5.4),
the time constant of the measuring system can be easily determined (cf. Sect.
7.2).
The disadvantages of the electrical heat flow rate calibration are the following:
the heater can hardly be installed in disk-type measuring systems,
heaters permanently installed in measuring systems are not situated at the
sample location; this leads to systematic errors,
heat fluxes in the wires lead to systematic uncertainties, "heat leaks". (Ensure
thermal symmetry of sample and reference side!)
Figure 4.14a shows the electrical calibration in a DSC with cylinder-type mea-
suring system. (Peak area calibration may also be carried out using the peak areas
furnished by these measurements; see below.) The resulting calibration curve
Kq,(T) = <Ptrue(T)/(<Pm(T) - 1"1\1 (T)) is represented in Fig. 4.14b; 1"1\1 is the base-
line heat flow rate (<Pm-curve without electric power).
2. Heat flow rate calibration can also be carried out with a sample of known heat
capacity (Fig. 4.15). The following is valid for the heat flow rate absorbed by the
sample (without reference sample) in (quasi-) steady state heating mode:
<Ptrue = Cp • {3 so that Kq, (T) = CiT) . (<Pm(T){3_ <Po(D)

(<Po: zero line value with empty crucibles.)
The advantages of this calibration method are the following:

applicable in all DSCs,
- calibration heat flux at the sample location,
no leads (wires) required.
The disadvantages:
the calibration cannot be switched off in between, i. e. checking of the base-
line (or zero line) is not possible during a run (leads to uncertainties in <Pm - <Po),
there is a temperature profile inside the sample (yields "mean C/').
The calibration substance most frequently used is synthetic sapphire (a-AI20 3 ,

corundum), cf. Table 4.9.
¢m- ¢bl
1
¢m
¢bl
a time, tempera ture ~
t
Kej>
b temperature
Fig.4.14. Electrical heat flow rate calibration of a heat flux DSC (cylinder-type measuring
system) by means of a built-in heater.
a Calibration peaks generated electrically in isothermal operation or during heating,
b Calibration factor K .. (calculated with the data of a, schematic) to determine the reaction
(real) heat flow rate: K.(T) = 4>true/( 4>m (T) - tJ\i (T»(with 4>true as electrical heating power p..),
4>m measured heat flow rate, cf>t,. baseline value of the heat flow rate
steady state region
temperature ..
Fig.4.15. Heat flow rate calibration by means of a known heat capacity Cp •
IPm measured heat flow rate, 1P0 zero line value; calibration factor to determine the true heat
flow rate: K~ = Cp(T) . fJ/( IPm (T) - 1P0(T»
4.4.2 Peak Area Calibration
For peak area calibration, a known heat Qtrue, dissipated or consumed, is compared
with the area of the resulting peak (Figs. 4.16 and 4.17):
If
Qtrue = KQ (T) f [<Pm (t) - «1\1(t)] dt or

'i
If
Qtrue =f [K~· <Pm (t) - K; . ~I(t)] dt

'/
'*'
In principle, K; K~ is valid. Since K. depends on <Pm and T - which is proportio-
nal to time (in all heat flux DSCs and also in power compensation DSCs) -, it can-
not be placed in front of the integral. This can be done in a first approximation only
if it is assumed that <Pm and «1\1 are of the same order of magnitude (no dependence
on <P) and the peak width is small (no dependence on T), which is in general the
case with phase transitions, not, however, with many polymer transformations.
Integration must be carried out over the whole peak in order that contributions of
the 1st derivative of the measurement signal are insignificant.
Peak area calibration can be performed in two ways:
1. If electrical calibration is possible (Fig. 4.16), the advantages connected with
electrical heat flow rate calibration (see above) apply analogously to peak area
calibration:
- easy and accurate measurement ofthe electrically generated heat
Qtrue = - f P.I . dt = - W.I = - U . i . M,
t,
f(¢-¢
m bl ldt =
A
1 t,
rpm
t, t,
time
Fig. 4.16. Electrical peak area calibration.
tPm measured heat flow rate, t j , tf beginning, end of the calibration peak, <flt,l baseline value
of the heat flow rate (electric power switched oft)
t,
f(¢m-¢bI1dt= A
1 t,
t, t,
time
Fig. 4.17. Peak area calibration by means of a known heat of fusion (schematic).
tPm measured heat flow rate, t j , tf beginning, end of the peak, 4\t baseline value of the heat
flow rate
peaks of differing size can be produced,

calibration is possible at any temperature,
measurement effects (peaks) can be "repeated" (simulated) by appropriately
presetting the development of the heating power with time,
measurement effects (peaks) can be "flanked" during the run by similar cali-
bration peaks.
The disadvantages are the same as those stated in the case of electrical heat flow
rate calibration:
the electric heater can hardly be installed in disk-type measuring systems,
- systematic errors result if heaters are permanently installed (different loca-
tions of heater and sample),
wires give rise to systematic uncertainties (heat leaks).
Care must be taken that electrical calibration is carried out only over a small tem-
perature interval.
2. Peak area calibration of DSCs is usually carried out by means of the heats of
melting of pure substances (Fig. 4.17). The advantages of this method are as
follows:
applicable in all DSCs,
calibration heat at the sample location,
- no wires required,
simultaneous temperature and heat calibration is possible with some (selec-
ted) calibration materials.
The disadvantages:
- calibration possible only at discrete temperatures,
no adaptation to the measured peak shape possible,
systematic uncertainties due to the special shape of the sample temperature
curve Ts(t) during the calibration procedure (melting),
uncertainties resulting from the determination of the area (definition of the
integration limits and shape of the baseline, cf. Sect. 5.3).
The calibration procedure for heat flux DSCs is thus as follows:
selection of calibration substances which cover the desired temperature range
and whose thermophysical characteristic data are similar to those of the
sample,
weighing-in of such masses which approximately generate a heat effect as is
found in normal measurements,
adjustment of customary heating rates. (Be careful in the case of calibration
substances which melt close to the start temperature: the quasi-steady state
must have been reached, the relaxation effects due to the start must have
faded; apply lower heating rate if necessary),
evaluation of peak (area and extrapolated peak onset temperature T.),
determination of KQ(T.), drawing ofa calibration curve (establishing a table)
or inputting measured data into the apparatus according to the manufacturer's
specifications (turning of potentiometer, or by software),
- estimate of the uncertainty of the calibration (uncertainty of the weighing, of

the baseline, of the integration limits, of the values for the heat of fusion taken
from the literature; the estimated uncertainty from theoretical considerations
must be taken into account),
measurement of the repeatability errors of the calibration factors (or calibra-
tion curve). This repeatability error must be clearly smaller than the estimated
overall uncertainty of the calibration (see above). The repeatability error is the
smallest possible uncertainty of measurement of caloric measurements.
The same calibration procedure is applied to power compensation DSCs, however,
in this case, calibration with one calibration material is in general sufficient (e. g.
In, cf. Sect. 2.2, 3.2 and Fig. 4.20), as the calibration factor depends on temperature
only to a small extent.
4.4.3 Examples of Caloric Calibrations
Calibration Curve
A DSC with cylinder-type measuring system was calibrated with the aid of electric
heaters which could be inserted into the sample containers. Only the insert in the
sample container was electrically heated; the second insert of the same type placed
into the reference container served to establish thermal symmetry.
At regular intervals, heating pulses of defined duration and power were auto-
matically generated over the whole temperature range, the measured signal being
put out as an voltage in f.1V. The peak areas were integrated applying evaluation soft-
ware. Fig. 4.18 shows schematically the measured curve and the resulting calibra-
tion curve.
A heat flux DSC with disk-type measuring system was calibrated with the aid of
sapphire (Fig. 4.19).
Differences Between K~ and KQ

When DSCs are calibrated with the aid ofthe heat of fusion of, for example, metals
(peak area calibration) and via the specific heat capacity (heat flow rate calibration),
systematic differences occur in practice. This is shown in Fig. 4.20 by the example
of a power compensation DSC.
The 1% difference in the case of a power compensation DSC suggests that the
difference will be substantially greater for heat flux DSCs. This is confirmed by
measurements:
For a heat flux DSC with disk-type measuring system, Fig.4.21 shows the
differences between the average value of a number of heat-of-melting (peak area)
calibrations and three (scattering) individual calibrations carried out with the heat
capacity of a sapphire sample (heat flow rate calibration).
Dependence on Other Parameters

The repeatability error by which the calibration with a known heat capacity is af-
fected can additionally depend on the sample mass. To give an example, the
repeatability error of the calibration of a heat flux DSC with disk-type measuring
f
/:,u
(temperature), time ~
temperature
Fig.4.18. Electrical heat calibration of a heat flux DSC (cylinder-type measuring system with
built-in heater, schematic curves).
I'1U measured signal (voltage), KQ calibration factor for peak area evaluation (in JV-I S-I), A
peak area (in V s), W.I electrical heating energy (belonging to one peak: i . I'1U' I'1t). In con-
trast to Fig. 4.14, the peak areas of the electrical calibration peaks were evaluated here to
obtain K Q • It is also possible to simultaneously determine Koj)from the electrical heating power
system amounted to 5 % when the sample mass was 7 mg; for a sample mass of
27 mg, the scatter decreased to half this amount (DoeIman et aI., 1977).
Figure 4.22 shows the temperature-dependent systematic difference between the
calibration factors, which results when one calibration is carried out with the cali-
bration heater permanently installed (under the bottom ofthe sample cavity) and the
other with the electric heater installed inside the sample (heat flux DSC with cylin-
der-type measuring system).
The influences of certain parameters (sample mass, heating rate) on the mea-
surement results described in the following, of course, appear analogously during
calibration and must be taken into consideration.
For a heat flux DSC with a disk-type measuring system a dependence on the
heating rate (Fig. 4.23) resulted for the heat of transition ofCsCI.
Figure 4.24 shows the relative dependence of the calibration factor on the
heating rate for two different sample masses for a power compensation DSC.
6
mW
4r::==:;-~
r -4
.....<'
OJ
.... -6
~
0
;;::
.....<' -12
OJ
..r:::.
-16
-20
0.0 4.0 8.0 12.0 16.0

time
20.0
Fig. 4.19. Heat flow rate calibration curve of a disk-type DSC.

..
24.0 28.0 32.0 36.0 min 44.0
Upper curve: Crucibles empty zero line, lower curve: 129.6 mg of sapphire, 10K min-I (acc.
to Sarge et al., 1994)
r
%
lin Isn lSi Pb
o
300 400 500 K 600

temperature
Fig. 4.20 Systematic difference between K. and KQ in a power compensation DSC. Heat flow
rate calibration was carried out with sapphire (solid line) and with copper (crosses); peak area
calibration with In, Sn, Bi, Pb (according to Hahne, Gloggler, 1989).
KQ calibration factor for peak area evaluation, K. calibration factor for heat flow rate mea-
surement, I range of uncertainty of the sapphire measurement (heat flow rate calibration),
I range of uncertainty of the heat-of-fusion calibration (peak area calibration)
10
f-lY
mw 9
11K
6
tempera ture
Fig.4.21. Differences between the heat-of-fusion calibration (peak area calibration) and the
calibration by means of a known heat capacity (heat flow rate calibration) (ace. to Sarge,
Cammenga, 1985).
K calibration factor, - calibration curve with sapphire (3 measurements at 10 K min-I),
.... average value curve of the calibration with melting samples
30
mJ
mY·s
25
t 20
15L-__'_"__~____~______~_______
o 100 200 300 400
temperature
Fig. 4.22. Differences in the electrical peak area calibration of a heat flux DSC with cylinder-
type measuring system (acc. to Hemminger, SchOnbom, 1982).
KQ calibration factor for peak area evaluation, - permanently installed calibration heater,
below the sample container, .... miniature heater installed in a copper sample, inside the sam-
ple container
3.2
kJ
mar
r u ~...o......--o-_:'----v
q
2.8 U-~ __ ~ __ ~ __L -____ ~ ____ ~~ ____ ~_
5 10 15 20 30 40 K/min 50
{3 ..
Fig. 4.23. Dependence of the molar heat of transition q of CsCI on the heating rate fJ for a
disk-type DSC (acc. to Breuer, Eysel, 1982)
3
% 10 mg
2 x x
r
x
x
x x ~ o 1 mg
KQ x x 0
0
0 0 ~ 0
0 0 0
-1~~--~~--~~--~~--~~--~
0.31 0.64
Fig. 4.24 Dependence of the calibration factor KQ on the sample mass and the heating rate fJ
for a power compensation DSC.
KQ calibration factor for peak area evaluation (determined by means of the heat of fusion of
indium, acc. to Hohne, Gloggler, 1989)
4.5 Conclusions Regarding the Calibration of DSCs
The following conclusions can be drawn from the explanations of Chapter 3 (Theo-
retical Fundamentals of DSCs), Sect. 4.3 (Temperature Calibration) and Sect. 4.4
(Caloric Calibration).
Temperature Calibration and Temperature Measurement

From among all possible characteristic temperatures which may be assigned to a
peak, only the extrapolated peak onset temperature 1'. (cf. Fig. 4.1, definition see
Sect. 5.1) is relatively independent of sample and test parameters (mass, layer
thickness, heat transfer, thermal conductivity and heating rate). This is why only T.
should be used to characterize phase transitions.
4.6 Calibration Substances 73
T. should always be corrected to zero heating rate (f3 = 0) unless shifting of 1'.
with the heating rate is the subject of the investigation (for example for kinetic
problems). The location of the sample inside the crucible and the calorimetric
measuring system is of importance; it should always be the same in precision
measurements.
Caloric Calibration and Heat Measurement

Different calibration factors are to be expected for measurements of heats of trans i-
tion and heat capacities; they must always be determined separately.
It is to be expected that the calibration factor for the heat flow rate depends on
the heat flow rate itself (i. e. on the heat capacity of the sample). This must be
checked by means of two samples of as different a heat capacity as possible (which
must, however, be known).
The reliability of heat capacity measurements by means of heat flux DSCs
lies between 5 and 20%; it is 3 to 5 % when power compensation DSCs are used.
A higher accuracy requires considerable time and effort be spent on the calibration.
When a dual-sample heat flux DSC is used (triple cell DSC, see Sect. 2.1.1), the
uncertainty of the heat capacity measurements is lower. A scatter (RMS, see Sect.
7.2) ofthe calibration factor KtlJofapprox. 1.0 to 1.6% is stated in the literature (see
Jin, Wunderlich, 1993).
As regards modulated DSCs, insufficient experimental experience has so far
been gained to state reliable figures for the total uncertainty of heat capacity mea-
surements.
Precise heat capacity measurements with heat flux DSCs are possible only if all
potential systematic error sources are kept constant by ensuring conditions as iden-
tical as possible during the calibration run and measurement run. (Equal mass,
equal thermal conductivity, equal heating range, equal heat capacity.)
In the case of peak area calibration, dependence of the calibration factor on the
mass of the sample, the heating rate, the peak size, the temperature and on sample
properties must be taken into consideration. The respective correlations have to be
clarified during calibration and taken into account when the reliability ofthe results
is estimated.
To ensure precise measurements, calibrations prior to the measurements should
in principle be as similar to the measurement proper as much as possible. The peak
to be measured should be "repeated" (simulated) by electrical calibration if there is
the possibility of doing this.
Due to the systematic error sources, a ± 5 % reliability can be attained when heats
of transition are measured with heat flux DSCs. This limit is at ± 1% for power
compensation DSCs. A prerequisite for a higher reliability is a very elaborate
calibration, for example by direct comparison with electric energy.
4.6 Calibration Substances

The requirements to be met by substances to be used for the calibration ofDSCs are
as follows:
high (defined) purity (at least 99.999%),

precisely known characteristic data of the substances,
non-hygroscopic, insensitive to light, non-toxic,
no decomposition, chemically stable,
no reaction with the material of the crucible (see Table 4.8) or with atmosphere,
negligible vapour pressure,
grains not too small.
When there are any doubts as to the usability of a substance, i. e. when the risk of
persons being injured or equipment damaged cannot be excluded, information
should be obtained from experts prior to using such substances.
The following must be observed when calibration samples are prepared:
clean preparation tools,
clean crucible,
no adsorption layers,
no oxide layers (surface) in the case of metal samples,
no weighing errors.
4.6.1 Calibration Substances for Temperature Calibration
For the temperature calibration of DSCs, those substances should preferably be

used with which the fixed points of the International Temperature Scale of 1990
(ITS-90) are realized *.
Not all substances meet the above requirements. Cadmium, for example, has a
high vapour pressure in the liquid state during melting, which may result in injuries
to persons and damage to equipment.
The substances used to realize the fixed points of the ITS-90 are listed in Table
4.6 starting with the triple point of neon. These substances have been marked by
"ITS-90" (cf. ITS-90, 1990).
Substances which were used to realize the defining fixed points of the Interna-
tional Practical Temperature Scale of 1968 (cf. IPTS-68, 1976) have also been
included. (The IPTS-68 was valid until December 31, 1989.) The fixed-point tem-
peratures indicated in the IPTS-68 have been converted to the ITS-90. In addition,
allist of secondary reference points was published (Bedford et aI., 1984).
To calibrate DSCs in the heating mode, the melting point at zero heating rate
(T.({3 = 0» is set equal to the solidification (resp. melting) point stated in Table 4.6.
Unfortunately, no suitable fixed-point materials are available for some tempera-
ture ranges which is why other substances, too, must be used for the temperature
calibration of DSCs. Taking the above criteria into consideration, the German
Society for Thermal Analysis (GEFTA) recommends that the substances listed in
* The fixed-point temperatures of metals are associated with the flat part of the temperature-
versus-time curve obtained during the slow solidification or melting (in the case of Ga) of
very pure metals at a pressure of 101.325 kPa. To calibrate DSCs, it is, however, appropriate
to apply the heating mode. (For other calibration substances, cf. Marsh, 1987.)
Table 4.6 Defining fixed points of the ITS-90 and some fixed points of the IPTS-68
(no longer valid) converted to the ITS-90
TinK tin °C
Triple point of neon (ITS-90) 24.5561 -248.5939

Triple point of oxygen (ITS-90) 54.3584 -218.7916
Triple point of argon (ITS-90) 83.8058 -189.3442
Triple point of mercury (ITS-90) 234.3156 - 38.8344
Water-ice point 273.15 0.00
Triple point of water (ITS-90) 273.16 0.01
Melting point of gallium (ITS-90) 302.9146 29.7646
Boiling point of water 373.12 99.97
Solidification point of indium (ITS-90) 429.7485 156.5985
Solidification point of tin (ITS-90) 505.078 231.928
Solidification point of bismuth 554.533 271.403
Solidification point oflead 600.612 327.462
Solidification point of zinc (ITS-90) 692.677 419.527
Solidification point of aluminium (ITS-90) 933.473 660.323
Solidification point of silver (ITS-90) 1234.93 961.78
Solidification point of gold (ITS-90) 1337.33 1064.18
Solidification point of copper (ITS-90) 1357.77 1084.62
Solidification point of nickel 1728 1455
Solidification point of palladium 1827 1554
Solidification point ofpJatinum 2041 1768
Solidification point of rhodium 2235 1962
Solidification point of iridium 2719 2446
Table 4.7 are used for temperature calibration. These materials have proved their
practical worth, they were carefully tested and cover the temperature range of custo-
mary DSCs (Cammenga et aI., 1993).
For the compatibility between calibration substances and crucible materials see
Table 4.8.
Only the uncertainty of the temperature measurement is decisive for the uncer-
tainty of the calibration when these substances are used. It comprises the pure
uncertainty of measurement of the sensors, repeatability errors of the heat transfers
in the measuring system and uncertainties in the processing of the analogue signal
until it is recorded and evaluated.
4.6.2 Calibration Substances for Heat Flow Rate Calibration
The substances Al20 3 and Cu are suitable for heat flow rate calibration purposes.
They are chemically stable and the heat capacity is known with an uncertainty of
about 0.1 % (Tables 4.9 and 4.10).
Table 4.7 Materials recommended for temperature calibration ofDSCs

(cf. Cammenga et aI., 1993)
substance transition uncertainty transi- ref.** remarks ***

temperature tion *
TinK tin °C L\TinmK
Cyclopentane 122.38 -150.77 50 sIs a

Cyclopentane 138.06 -135.09 50 sIs a
Cyclopentane 179.72 - 93.43 50 sn a
Water 273.15 0.00 10 sll b 2
Gallium 302.9146 29.7646 sll c 3
Indium 429.7485 156.5985 sll c
Tin 505.078 231.928 sll c 4
Lead 600.61 327.46 10 sll b 5
Zinc 692.677 419.527 sll c 6
Lithium sulphate 851.43 578.28 250 sIs d 7
Aluminium 933.473 660.323 sll c 8
Silver 1234.93 961.78 sll c 9
Gold 1337.33 1064.18 sn c 9
* s: solid; 1: liquid
** a: Aston et aI., 1943, b: Preston-Thomas, 1976 (IPTS-68), c: Preston-Thomas, 1990
(ITS-90), d: Cammenga et aI., 1993
*** 1: Only in hermetically sealed crucible. 2: Air-saturated, bidistilled water in hermetically
closed crucible. 3: Melt reacts with AI; note that strong undercooling occurs. 4: Melt
reacts with AI, Pt. 5: Melt reacts with Pt, oxidizes quickly (use protective gas). 6: Melt
(and vapour) react with AI, Pt; high vapour pressure at melting point (approx. 20 Pa).
7: Anhydrate is hygroscopic, thus weigh-in as Li2S04 . H20. Dehydration takes place
from 100°C, thus turbulent movement of particles in the crucible and high water-vapour
pressure (do not use hermetically sealed crucible). Must not melt (change of properties).
8: Melt reacts strongly with Pt. 9: Melt dissolves oxygen, reacts with Pt.
4.6.3 Calibration Substances for Peak Area Calibration
The heats of transition of pure substances for peak area calibration should have been
measured with adiabatic precision calorimeters (cf. Appendix 2). Such results are
available for only a few substances which fulfill the above requirements (Sect. 4.6).
When several independent precise measurements have been carried out on one sub-
stance, the ranges of uncertainty stated often do not overlap so that the estimate of
the best value and of the overall uncertainty is problematic. Recommended sub-
stances are listed in Table 4.11.
The analysis of the relatively large number of measured values of In and Sn
obtained by different calorimetric methods allows a minimum uncertainty of about
± 0.5 % of all values stated in the literature to be assumed (Hemminger, Raetz
1989). When only a few measured values can be used as a basis, these can be affect-
ed by substantially greater actual uncertainties as is shown, for example, by the
maximum deviations or standard deviations of the values available for Bi (Raetz,
1989).
Table 4.8 Compatibility between calibration substances and crucible materials (according to Cammenga et a!., 1993)
Q)
1ii
Q) ..=
,.e..
::s §
~
Q) '" 'E
0.. §
Q) § .§
.9 .. ;;
'1:1
os g .~ .
~ =a s: Q) ..s ~ :s!0
crucible material U $ ~ ..s ~ ~ N :.::l :;( tI) 0
Corundum, Al20 3 D D + + + + + + + + +
Boron nitride, BN D D + + + + + + + ? ?
Graphite, C D D + + + + + + + +
Silicate glass + + + + + + ? + x x
Quartz glass, Si02 + + + + + + + + + +
Aluminium, Al + + + + x x x
Aluminium, oxidized + + + + + + + + x x x
Silver, Ag + + ? x x
Gold, Au + + + x
Nickel, Ni + + ? +
Iron, Fe + + + ? + .j>.
Stainless steel + + + + ? + a-
Platinum, Pt + + + (")
+ ? e!.
Molybdenum, Mo + ? ? ? ? 6'
Tantalum, Ta + + ? + ? ? ? + + ~.
0
Tungsten, W D D ? ? + ? + + ::s
til
+ : No solubility and influence on melting temperature to be expected. <T
=
- : Melt dissolves crucible material, greater change of melting temperature. '"
: Partial solution processes possible with negligible change of melting temperature. ...~
x : Crucible melts. '"
? : Compatibility unknown.
D: Combination cannot be realized. -..l
-..l
Table 4.9 Heat capacity of a-aluminium oxide (according to Sarge et aI., 1994)
Alumina (highly pure a - AIP3' synthetic sapphire, corundum)
Molar mass 101.9613 g mol-I
Tin K (ITS-90) cp in J-I g K-I Cpo mol in J mol-I K-I
100 0.1260 12.85

150 0.3133 31.94
200 0.5014 5l.12
250 0.6579 67.08
300 0.7794 79.47
350 0.8706 88.77
400 0.9419 96.03
450 0.9976 101.72
500 1.0414 106.19
550 1.0762 109.73
600 1.1042 112.59
650 l.1272 114.93
700 l.1465 116.90
750 l.1630 118.58
800 l.1776 120.07
850 l.1906 121.40
900 1.2026 122.61
950 1.2135 123.73
1000 1.2238 124.78
1050 1.2333 125.75
1100 1.2423 126.66
1150 1.2507 127.52
1200 1.2585 128.32
1250 1.2658 129.07
1300 1.2727 129.77
1350 1.2792 130.43
1400 1.2854 131.06
1450 1.2912 131.66
1500 1.2968 132.23
1550 1.3022 132.77
1600 1.3074 133.30
1650 1.3124 133.81
1700 1.3172 134.30
1750 1.3217 134.77
1800 1.3260 135.20
1850 1.3300 135.61
1900 1.3335 135.97
1950 1.3367 136.29
2000 1.3395 136.58
2050 1.3420 136.83
2100 1.3444 137.08
2150 1.3471 137.35
2200 1.3506 137.71
2250 1.3558 138.24
Table 4.10 Heat capacity of copper (molar mass 63.546 g mol-I) (acc. to Sarge et aI., 1994)
Tin K (ITS-90) cp in J-I g K-I Cp • mol in J mol-I K-I
100 0.2520 16.01

120 0.2867 18.22
140 0.3122 19.84
160 0.3309 21.03
180 0.3447 21.90
200 0.3550 22.56
220 0.3631 23.08
240 0.3699 23.51
260 0.3758 23.88
280 0.3809 24.21
300 0.3850 24.47
320 0.3875 24.63
Table 4.11 Calibration substances for peak area calibration

(acc. to Sarge et aI., 1994)
substance Ttrs (ITS-90) type of transition * Qtrs uncertainty **

in °C in J g-I
Cyc10pentane -150.77 sis 69.60 (±0.5%)

Cyc10pentane -135.09 sis 4.91 (± 1.1 %)
Cyclopentane - 93.43 sll 8.63 (± 1.1 %)
Gallium 29.7646 sll 79.88 (±0.9%)
Indium 156.5985 sll 28.62 (±0.4%)
Tin 231.928 sll 60.40 (±0.6%)
Bismuth 271.403 sll 53.83 (±3.9%)
Lithium sulphate 578.28 sis 228.1 (±4.6%)
Aluminium 660.323 sll 398.1 (±2.3%)
* s: solid; 1: liquid.
** uncertainties: 2 times standard deviation Un_I'
Heat flux DSCs have a number of systematic error sources (cf. Sect. 3.1). In our
opinion, the reliability of the results obtained with them is therefore of the order of
approx. ± 5 % in routine operation. The uncertainties of the heats of transition indi-
cated in Table 4.11 are therefore sufficient to allow these substances to be used as
calibration substances.
Calibration by means of electrically generated heat should be aimed at for preci-
sion measurements. In addition, reference is made to the discussion of thermo-
dynamic aspects with respect to the peak area calibration in Sect. 4.2.
The discussion about the suitability of various substances for caloric calibration
is at present still going on at the international level. The following complexes of
problems are concerned:
backing-up and redetermination of characteristic data, including an estimate of

the uncertainties,
testing of new substances for their suitability for caloric calibration purposes,
search for substances suitable for special calibrations, for example polymers
(glass transition, heat capacity, special transitions).
5 The DSC Curve
The DSC curve (measured curve) offers quick information on the total measuring
process (Fig. 5.1). In addition to the usual measuring effects (Cp changes, transi-
tions, reactions) it can be seen
whether the predetermined temperature range has been completely covered,
whether disturbances ofthe apparatus (mechanical, electrical) occurred,
whether there were irregularities or unusual shapes of the baseline,
whether the characteristic temperatures and peak areas lie within the expected
range.
~
tempera ture
Fig. 5.1. Curve measured by a DSC with step of the baseline (Cp change) and endothermic
peak (I st order transition).
dCp change of the sample's heat capacity, 7;, Tfinitial peak temperature, final peak tempera-
ture, CD initial segment of the measured curve, ® step of the measured curve due to d Cp ,
@ measured curve, ® interpolated baseline (between 7; and Tf in the peak region), ® final
segment of the measured curve, @peak(measuredcurve)
82 5 The DSC Curve
5.1 Characteristic Terms

Some characteristic terms are used to describe a measured curve. They are defined
in the following (Fig. 5.2).
The zero line is the curve measured with the instrument empty, i.e. without sam-
ples and without sample containers (crucibles), or without samples and with the
sample containers (crucibles) empty. It shows the thermal behavior of the mea-
suring system without samples. The smaller the range of variation (repeatability,
see Sect. 7.2), the better the instrument,
the (interpolated) baseline is the line which in the range of a peak is constructed
in such a way (cf. Sect. 5.3) that it connects the measured curve before and
behind the peak as if no heat had been exchanged, i. e. as if no heat (peak) had
developed,
a peak in the measured curve appears when the steady state is disturbed by
thermally activated heat production or consumption in the sample. Peaks in heat
flow rate curves, which are assigned to endothermic processes, are plotted
"upwards" (positive direction), as heat added to a system is defined as positive
in thermodynamics. A peak begins at T; (first deviation from the baseline, see
below), ascends/descends to the peak maximum/minimum, Tp (see below), and
merges into the baseline again at Tf • Only those transitions associated with a heat
of transition (e. g. melting) lead to a peak (except for changes in the heat transfer
peak (endothermic)
I
I /
I
I
t
I I \
I I -Q_'::::-:--<;>---
-~\ 1\
\ I I
---
I baseline (interpolated)
I I
I
\
zero line
T. r.
temperature ..
Fig. 5.2. Definition of zero line, baseline, peak and the characteristic temperatures (defini-
tions see text).
1'; initial peak temperature, To extrapolated peak onset temllerature, Tp peak maximum tem-
perature, 7;, extrapolated peak completion temperature, Tf final peak temperature
5.2 Influencing Parameters 83
between the sample and the temperature sensor). Other transitions (e.g. glass tran-
sitions) only lead to changes in the shape of the measured curve, for example step
changes (see Sect. 6.2.3),
- the characteristic temperatures are defined as follows:
- Ii Initial peak temperature
Here the curve of measured values begins to deviate from the baseline, the
peak begins,
Te Extrapolated peak onset temperature
Here the auxiliary line through the ascending peak slope intersects the base-
line. (The auxiliary line is drawn through the (almost) linear section of the
ascending peak slope, either as inflectional tangent or as fitted line. The
distinction between the two methods is of no significance in practice, as the
resulting difference is smaller than the repeatability error of the measurement
results.),
- Tp Peak maximum temperature
This temperature designates the maximum value of the difference between the
curve of measured values and the interpolated baseline (not necessarily the
absolute maximum of the curve of measured values),
Tc Extrapolated peak completion temperature
Here the auxiliary line (see above) through the descending peak slope inter-
sects the baseline,
- Tr Final peak temperature
Here the curve of measured values reaches again the baseline, the peak is
completed.
5.2 Influencing Parameters

In Sects. 4.3 and 4.4, reference has already been made to the influence of some
parameters in connection with the calibration. These statements apply by analogy to
each measurement.
For highly precise measurements, the baseline must be determined prior to and
after the measurements. Temperature and heat calibration must then also be checked
at regular intervals (depending on the specific DSC) at least with one calibration
substance, e. g. indium. If the DSC shows a distinct tendency towards drifting, a
daily test is important.
In this way, information is obtained about drift processes or scatter which cannot
be assigned to an exactly known parameter. This information enters into the
estimate of the overall uncertainty.
The influencing parameters are listed in the following:
1. The shape of the zero line (without crucibles) is influenced by the heating rate,
the kind of purge gas and its flow velocity, likewise by the temperature of the
surroundings and by surface properties of the measuring system. If the zero line
is measured with the crucibles empty added to this are influences due to unequal
masses of the crucibles, differences in the heat transfer between the crucibles and
84 5 The DSC Curve
the furnace, differences between the emissivities of the two crucibles (lids), and
influences due to type and material of the crucibles used.
2. Point 1 is also applicable to the shape of the measured curve outside a peak with
sample and reference sample placed in the crucibles. In addition, the properties
of the reference sample (heat capacity and its temperature dependence) are of
importance (differential measurement). In the case of pure Cp changes of the
sample (e.g. glass point, Curie point), these changes determine the shape of the
measured curve which then contains the desired information. Unfortunately,
changes of the measured curve can also take place if the conditions of heat
transfer to the sample change abruptly. This type of step changes usually appears
statistically and can thus be distinguished from the glass transition of homo-
geneous samples which always occurs at the same temperature.
3. Point 2 is also applicable to the shape of the measured curve with peak. The peak
itself is additionally influenced by
the heating rate (cf. Figs. 3.11,4.10),
- the thermal conductivity of the sample (cf. Fig. 3.13),
the mass and heat capacity ofthe sample (cf. Fig. 4.24),
the structure of the sample (powder, granulates, foil, ... ),
the thermal resistance between sample and temperature sensor (cf. Reichelt,
Hemminger, 1983),
the location ofthe sample in the crucible or measuring system (cf. Fig. 4.13),
the kind of gas in the cavity of the measuring system influences the separation
(resolution) of closely adjacent peaks (and the calibration).
In addition, attention must be paid to:
the sample purity,
- the thermal history of the measuring system (cf. Suzuki, Wunderlich, 1984).
Conclusion
Ensure that all the parameters for the measurement are as similar as possible to
those for calibration. Use samples of defined state, shape and purity.
5.3 The Baseline and the Determination

of the Thermodynamic Functions
The baseline has been defined in Sect. 5.1 (cf. Figs. 5.1, 5.2):
In the region of a peak, the baseline is the curve between 7; and If which would have
been recorded if all Cp changes (and changes of heat transfer etc. that also influence
the baseline) had occurred but no heat of transition had been released. The area
between baseline and measured curve is a measure of the heat of transition.
Construction of the Baseline

For 1st order transitions with or without coupled Cp change (and/or change of the
heat transport mechanism), the baseline must be constructed (cf. Hemminger,
Sarge, 1991). A definite baseline can be constructed only if a pure Cp change
5.3 The Baseline and the Determination of the Thermodynamic Functions 85
occurs; changes of heat transport conditions in the range of a peak give rise to
uncertainties in the shape of the baseline. For certain evaluations (for example kine-
tics), the measured curve and the respective baseline must possibly be "desmeared"
(cf. Sec. 5.4).
1. For irreversible transformations without measurable Cp change, the baseline can
basically be determined by repeating the measurement with the same sample
whose reaction process has come to an end. In these cases, the baseline is inter-
polated with the help ofthe measured curve ofthe 2nd run (example: annealing
of lattice defects during recrystallization of a plastically deformed metal). The
baseline uncertainty corresponds to the repeatability error of the DSC.
2. For transformations with continuous Cp changes (without changes of the heat
transfer conditions), the transformed mass fraction is in principle known for any
particular time of the transformation (conservation of mass). Thus, the Cp
change coupled to it is also known.
3. Expressed more generally and in a purely formal manner (for example for a non-
horizontal measured curve outside the peak), the following is valid: When the
degree of reaction a(t) is known (possibly only by approximation), the baseline
can be constructed according to van der Plaats, 1984.
For the change of the slope of the baseline between 1; and Tr , the following is
valid in good approximation (see Fig. 5.3).
(-dcP) =(l-a) (dcP)

- +a (dcP)
-
dt bl dt Ti dt Tr
r,
Tf
temperature
Fig. 5.3. Construction of the baseline taking the degree of reaction a(t) into account (acc. to
van der Plaats, 1984).
T; initial peak temperature, Tr final peak temperature, «Pi"" «Pr", measured curve extrapolated
from 1j, Tr into the peak region
86 5 The DSC Curve
That is to say, a(t) determines how quickly the slope of the baseline changes
from the slope of the measured curve in 1'; (with a = 0 for T < 1';) in the interval
'Ii : :; T:::;; Tr with a = a (t)) to the slope of the measured curve in Tr (with a = 1 for
T> Tr). In the interval between 1'; and Tr, the following results by approximation
for the functional values <1\1:
<1\1 = (1 - a) <P. e, + a· cPrex
where <P. e, and cPre, are the segments for the measured curve extrapolated into the
peak range from the left-hand and right-hand side, respectively. That is to say, <P.ex
and cPrex are to be calculated as polynomials and extrapolated into the peak range
in order that the baseline can be calculated according to the above equation.
(When deriving this relation it has been assumed that the difference between the
slopes of the measured curves in 1'; and Tr is not extreme; otherwise, this should
be taken into account as well.)
Integration of the peak area can be carried out at once using the difference
between the (desmeared) measured curve and the baseline cP(t) - <1\1 (t).
4. If the heat capacity of the sample changes suddenly by ~ Cs at the transition tem-
perature Ttrs with a constant sample temperature prevailing during transition
(first-order phase transitions), the change in the heat capacity is described by a
step function at T= Ttrs • By the "RC-elements" of the measuring system (cf. Figs.
3.7,3.9), this step function is transformed into an exponential function (of time)
(see Hemminger, Sarge, 1991).
5. For transformations showing a jump of the baseline the cause of which is not
known and which is not necessarily coupled to the fraction transformed during
the reaction (for example spontaneous change of the heat transfer between sample
and container), there is no method for finding the "correct" baseline which can
be backed up theoretically. Figure 5.4 shows several possible constructions. The
peak area very much depends on the selected baseline. For this reason, several
possibilities should always be tried out (see Hemminger, Sarge, 1991). The
differences in the peak areas (or heat flow rates) appear as systematic uncertain-
ties of the measurements and must be taken into account when the overall un-
certainty of measurement is estimated.
The baseline undergoes changes when Cp changes; additional changes may result
when:
- the thermal conductivity of the sample changes substantially during transition,
the thermal resistance (heat exchange conditions) between sample and tempera-
ture sensor changes during transition (e. g. during melting),
the conditions for the heat transfer between sample and surroundings change, for
example when the emissivity of the sample changes due to a reaction and the
sample is positioned in an open crucible.
When such changes coincide with Cp changes, the uncertainty of how the baseline
should be determined generally increases. There is then no method for constructing
a "true" baseline.
As emphasized previously, all the theoretical discussions presume a zero line
which is at least an absolute straight line. If not, the zero line must first be sub-
5.3 The Baseline and the Determination of the Thermodynamic Functions 87
T. Tf
tempera ture
Fig. 5.4. Possible baselines when the reasons for the "baseline jump" between 1j and Tr are
not known.
1j initial peak temperature, Tr final peak temperature, - linear extrapolation into the peak
region and jump somewhere between 1j and Tr , - - - constructed by means of an "apparent"
transformation function, .... straight line between 1j and Tr
tracted from the measured curve before the described procedures can be carried out
correctly. It is highly recommended that this be done for every measurement. This
does not usually require more time and effort to be spent on measurements, as in
most cases the zero line does not change very much during a set of measurements
if the conditions which influence the zero line (see Sect. 5.2) are kept constant. It
is therefore possible to use the same zero line - measured in the morning - for
all measurements of the same kind during a working day. Only for highly precise
measurements is a separate zero line required in each case and they Ulerefore take
twice the time.
Except for a constant factor - namely the heating rate and the mass (see Sect.
2.1.1) - and, of course, the calibration factor, the measured curve with the zero line
subtracted is identical with the heat capacity curve. As a consequence, the baseline
for its part is proportional to the pure heat capacity of the sample in question.
Bearing this in mind, a method can be specified, which allows the latent heat (of a
transition or reaction) to be determined from a corrected zero line or a Cp curve
without a baseline being constructed.
The procedure will be demonstrated using melting as an example, which is a 1st
order thermodynamic transition (Richardson, 1993). Figure 5.5 shows schematical-
ly the Cp curve prior to, during and after the transition. The heat capacities Cpo s (T) for
the solid prior to melting and Cp •1 (T) for the liquid after transition differ both in their
absolute values (at the same temperature) and also in their temperature dependence.
88 5 The DSC Curve
r
.....>-
·u
'"
0..
'"
....
..... (P.s ITI
'"cu
..c:.
tempera ture ~
Fig. 5.5. The determination ofthennodynamically valid values for the enthalpies of transition
(ace. to Richardson, 1993). The limiting temperatures T. and T2 must be chosen well below or
above the transition interval.
IlH(T) 0= Q-A -B, Q: t::::::1 A: ~ B: ~
The kind of functional dependence of the heat capacity on the temperature (linear
dependence, polynomials of higher order) is unimportant. Extrapolations into
temperature ranges that are not directly measurable must, however, be reliable. If
this is guaranteed, the heat absorbed over a wide temperature interval can be
divided into Cp portions and the isothermal enthalpy of fusion that is of interest. The
following is valid (cf. Fig. 6.15):
IlH(T) = Hl(T)-H.(T)
llH(T) = (Hl (T2) - H.(T,» - H.(T) - (H.(Tl» - (H.(T2» - Hl(T»

T 72
llH(T) = (Hl(T2)-H.(T1» - !Cp •• (1) dT- !Cp,l(T)dT
T. T
llH(T) = Q-A-B
The three quantities Q, A and B can be determined from the measured curve
(hatched differently in Fig. 5.5).
Enthalpies of reaction uniquely defined from the thermodynamic viewpoint can
be determined analogously; however, this determination presents some special fea-
tures (cf. Sec. 6.2.2.1) and Figs. 6.13 to 6.16.
At low heating and cooling rates, if necessary after previous desmearing of the
measured curve, the entropy change over temperature ranges in which no transi-
5.4 Desmearing of the DSC Curve 89
tions take place can be determined with sufficient accuracy according to the follow-
ing equation:
7i dT
6.S = fCp(T)-
TI T
The situation is different if phase transitions are included, in principle these are
shifted on the measured curve relative to the true temperature of equilibrium Too. (cf.
Sect. 5.4.4). Nevertheless, if in compliance with the above prescription, Too. and
6.H(Too,) are known, the reversibly exchanged entropy of transition can be calculated
from the quotient 6.H(Too,)IToo,. It is then possible to calculate changes in the
entropy and the free enthalpy for arbitrary temperature intervals. If a certain sub-
stance exists in several forms, information on their thermodynamic stability as a
function of temperature can be obtained.
5.4 Desmearing of the DSC Curve

The user ofDSC is usually interested in the heat flow rate into the sample to be in-
vestigated or in the related thermodynamic potential function (depending on the
temperature), which is either the specific heat capacity cp(TsamPle) or the enthalpy
change 6.H(Tsample). The measured curve put out by the calorimeter usually shows
the temperature difference between sample and reference support (TMS - TMJJ as a
function of time, or a function proportional to it, which is only a more or less
"smeared" representation of the function searched. Degree and quality of this
"smearing" differ depending on the type of DSC, and they, moreover, depend on
measurement parameters (heating rate, temperature, sample size). The theoretical
relations which lead to this smearing have been quantitatively described in
Chapter 3 in various approximations.
In many cases the curve measured by the DSC is a sufficiently exact representa-
tion of the functions searched, and the desired evaluations can be directly made.
When the accuracy requirements are higher, or kinetic evaluation must be per-
formed, the measured curve (and the interpolated baseline) must be corrected and
converted prior to their being used to determine the thermodynamic potential func-
tion (6.H) of the sample or its derivative (Cp ). This procedure is referred to as
"desmearing" and will be described in detail in the following. It should therefore be
made quite clear, when individual corrections are required, when they can be
dispensed with and which will be the consequences for the uncertainty of the
results.
Figure 5.6 schematically shows the relation between the heat flow rate curve
recorded by the DSC and the desired potential function for the case of a 1st order
endothermic phase transition. Accordingly, desmearing takes place in several steps:
1. Linear transformation of the abscissa (indicated temperatures) in compliance
with the results of the calorimeter's temperature calibration, and transformation
of the ordinate (indicated heat flow rates) in compliance with the results of the
calorimeter's heat flow rate calibration.
90 5 The DSC Curve
~rue ~ b
~L.I / \ ~s -~o
}
Ttrs
~rue ~
1, I ./1
Ttr"s
~rue ~
)d
HJ Ttrs
Ts
-
LI~
e
Ttrs
Ts
-
Fig. 5.6. Curve measured by a DSC for a 1st order endothermic transition and its evaluation.
a Scaling of temperature and heat flow rate axes on the basis of the calibration, b Zero line
(crucibles empty) subtracted from the measured curve, c Desmearing due to influences from
the apparatus (thermal lag), d Conversion into the heat capacity function, e Calculation ofthe
enthalpy function.
tPm heat flow rate (measured curve), tPs heat flow rate into the sample, tPo zero line heat flow
rate, tPtrue heat flow rate after calibration, LitPtrue sample heat flow rate tPs minus zero line heat
flow rate tPo , tP, reaction heat flow rate into the sample, Tm temperature (measured), Ttrue tem-
perature after calibration, Ttrs transition temperature, Ts sample temperature, Cp heat capacity
of the sample, t;.H enthalpy change
2. Determination and subtraction of the zero line (sample crucibles empty).

3. Mathematical elimination (deconvolution) of the influence which heat transport
phenomena inside the DSC (heat flow rate relaxation) exert on the measured
curve.
4. Conversion of the heat flow rate function into the specific heat capacity of the
sample, i.e. (non-linear) variable transformation ofthe temperature measured by
the DSC into the true sample temperature.
5. Integration leading to the thermodynamic potential function (enthalpy).
The required steps will be described in closer detail in the sections below.
5.4.1 Correction of the Temperature and Heat Flow Rate Indicated

This procedure has already been described in Sect. 4.3, however, for the sake of
completeness, reference to it will be made again in the following.
For the correction of the temperature indicated at the heating or cooling rate f3,
the following is valid.
(5.1)
!:.Tcorr (f3) = !:.Tcorr (f3 = 0) - !:.T(f3)
!:.Tcorr (f3= 0) results from the calibration measurements, and !:.T(f3) must either be
determined experimentally as it may be different from one sample to the other, or it
is determined by approximation from the results of the calibration measurements
(cf. Sect. 4.3.1). The above equations describe a shift of the temperature axis
scaling of the output measured function, a shift which is different for each heating
rate. A simple correction is concerned here which must of course be carried out only
if it is significant for the accuracy required for the measurements. The correction of
the displayed temperature according to Eq. (5.1) takes only influences from the
instrument into account. Samples of great thickness possibly require an additional
correction which also considers the temperature profile inside the sample; this cor-
rection is described below (Sect. 5.4.4).
The following equations are valid for the correction of the displayed heat flow
rate (cf. chapter 3 and Sect. 4.4.1).
<Ptrue = K~· <Pm or <Ptrue = -K· !:.Tm
The factors K~ (or K) are determined by suitable calibration measurements (cf.

Sect. 4.4.1). The negative sign in the above equation is not always used consistent-
ly; it results from the differing definitions of the sign of !:.Tm (measured temperature
difference) and <Pm. A linear transformation ofthe c:P-axis ofthe measured curve put
out, with the form of the function maintained, is, however, only possible if K~ (or
K) is not temperature-dependent in the range of interest. This case is often given
only for relatively small temperature ranges (e. g. for a transition peak). In all other
cases (above all when the heat capacity is measured over temperature ranges of
100 K or more), the form of the function must be recalculated according to the
following equation, the scaling of the T-axis being retained:
<Ptrue(T) = K.(T) . <Pm (T) or <Ptrue(T) = - K(T) . !:.Tm (T)
92 5 The DSC Curve
5.4.2 Subtraction of the Baseline
It is not possible to manufacture a perfectly symmetrical DSC in which the differen-

tial signal between empty sample system and reference sample system is exactly
zero over the entire temperature range. This also applies to the measured curve with
the sample crucibles empty (zero line).
To obtain the true heat flow rate into the sample (for example, to calculate the
sample's heat capacity), the instrument's asymmetry must be separately determined
by a second measurement and the zero line (crucibles empty) subtracted from the
signal actually measured.
To correctly determine this zero line, all parameters of the zero line measurement
should correspond to those of the actual measurement (with the sample crucible
being, however, empty in this case).
The correction of the instrument asymmetry presents the problem that the two
measurements can be carried out only one after the other; one must therefore
ensure that the DSC's symmetry properties have not changed in the meantime. This
is a strict condition which means that high requirements must be met by the
calorimeter (cf. Sect. 7.2).
In addition, it is unavoidable that the DSC is opened between the two measure-
ments and the sample crucible changed or moved. This alone can change symmetry
properties, which results in the zero line being shifted.
The design of the evaluation software of the DSC computers is normally such
that the zero line (empty crucibles) is first subtracted and then the corrections of the
temperature and heat flow rate scales (see above) are made. This order is justified
only if the scale corrections for the measurement and the baseline are the same. In
all other cases, first the correction and then the subtraction must be made. The same
is true of all advanced desmearing procedures.
5.4.3 Calculation of the True Heat Flow Rate into the Sample
Owing to the design of the DSC's measuring system, the heat flow rate into the
sample cannot be measured directly; the measurement always yields only a heat
flow rate at a certain distance from the sample, outside the sample crucible. Due to
the finite thermal conductivity of the material between this point of measurement
and the sample, the measurement signal is always a smeared representation of the
reaction heat flow rate, the kind and degree of smearing depending basically only
on the DSC construction.
Good DSCs allow the theory of linear response to be applied to this problem.
Using this tool, ways have been found to calculate the true reaction heat flow rate
from the measured heat flow rate curve. These methods are generally referred to as
deconvolution or desmearing; they will be described in the following.
Solution of the Differential Equation

For simple cases, the behavior of a calorimeter can be described by normal dif-
ferential equations.
The mathematical interrelation has already been dealt with in Chapter 3. The
desired function tPr(t) can therefore be calculated from the measured function tPm
(or t:.T(t) which is linked with tPmby the equation tPm = - k' . t:.T(t )), using Eq. (3.6)
in the following form:
(5.2)
Accordingly, the desired function is a simple sum of terms which include the
measured function and its derivatives. The coefficients aJ and a 2 include the time
constants 'fJ and 'f2 of the instrument and thus thermal resistances and capacities of
the measuring system. Equation (5.2) can easily be realized by an electronic circuit
with operational amplifiers. The desmeared signal can then be determined "online"
(simultaneously) from the measured signal tPm(t). It is indeed also possible to
calculate it from the (stored) function tPm(t) after the measurement has been con-
cluded.
Although the problem is easily solved (mathematically or by electronic means),
the determination of the proper coefficients aJ and a 2 is not so easy in practice. The
times constants 'fJ and 'f2 (Eq. (3.7)) can be determined from the measured function
of a pulse- or step-like event (cf. Fig. 3.10), but the actual relation to aJ and a2 may
be more complex than the approximative calculations in Sect. 3.1 possibly predict.
There is no other way than to adjust these two coefficients until the results of the
desmearing procedure (on-line or calculated) coincide with the original heat flow
rate tPr inside the sample. The best method to test it is to switch on a constant heat
flow rate (for a certain time) with the aid ofa built-in electric heater.
Numerical Methods
If the behavior of the DSC cannot be described by simple differential equations
within the framework of the required accuracy (e. g. in the case of power compensa-
tion DSCs), another method must be applied which has its roots in the theory of
linear response. It is valid for all measuring systems which work linearly, that is to
say, the measured signal for two distinct pulse-like events in the sample must be the
sum (superposition) of the two single functions from each individual event (Fig.
5.7). Another condition is that all measured functions (curves) of various pulse-like
events should have the same shape, in other words all these measured functions
divided by their peak area (normalizing) must yield the same function, the so-cal-
led apparatus (or Green's) function a(t).
If these conditions are fulfilled, the following is valid:
tPm(t) = ftPr(t')·a(t-t')dt' =a(t) * tPr(t) (convolution) (5.3)
This equation is valid for all DSCs which work in the above-described linear man-
ner, irrespective of whether a certain approximative formula (such as Eq. (3.6)) is
known. Only the "apparatus function" a (t) must be known which can easily be
derived from a pulse-like event (usually produced with a built-in electric heater), or
by the sudden solidification of a strongly supercooled melt of a pure metal.
94 5 The DSC Curve
T2
a tempera ture
t
\/
,"-'-
,., /
/
/
.......
------
b tempera ture
Fig. 5.7. Linear response of a DSC.
a Pulse-like heat events inside the sample at temperatures TJ and T2 in the scanning mode,
b Measured DSC curve, - - - -: hypothetic (measured) curves of the individual events at TJ , T2 , .
- : recorded (superposed) curve measured for two successive events, «1>, heat flow rate
developed in the sample, «1>m measured heat flow rate (measured signal), Q heat of the event,
a(T) apparatus function of the DSC
The seamy side of this desmearing method, which is also called "deconvolu-
tion" * is the rather ambiti()us mathematics required to solve the integral equation
(5.3) for the function of interest 4>,(t). There are essentially two methods, the
Fourier transform and the recursion method. Both require numerical calculations
with a computer.
* Deconvolution is a particular case of "desmearing" carried out by solving the "convolution

integral".
The Fourier transform represents an integral operation:

j(f(x)) = 1I2n· If(y) exp(ixy)dy
Applied to the convolution integral (Eq. (5.3)), it yields (see textbooks of mathe-
matics):
j(tPm(t)) = j(tPr(t)) . j(a(t)) (convolution theorem)
Thus the convolution product turns into an ordinary product which can be solved
for
j(tPr(t)) = j(tPm(t))
j(a(t))
The desired function is obtained by an inverse Fourier transform:
This method can be applied in all cases. Today, the Fourier transform is commonly
included in the program library of computers. The drawbacks of this procedure lie
in its laborious course and abstract nature, since the calculations are performed
in Fourier space. Those who lack experience in numerical Fourier transforms are
advised to study some "pitfalls" such as the "break-off effect" and the "sampling
theorem", both obtained by numerical treatment. Under specific conditions this
simulates periodicities and fluctuations which do not reflect any actual processes in
the sample. Please refer to the literature for further details (e. g. Bracewell, 1965).
The recursion method for solving the convolution integral, Eq. (5.3), starts from
the following recursion formula:
tPro(t) = tPm(t)
tPrn(t) = tPr(n_I)(t) + (tPm(t) - a(t) * tPr(n_I)(t)) (5.4)
The deviation between the "reconvoluted", still inaccurate synthetic function

a(/) * tPr(n_I)(/) and the measured function tPm(t) is used additively for a simple cor-
rection of the approximation.
The recursion formula (5.4) does not converge for all event functions. Abrupt
changes and steps (on-off effects and similar phenomena) generate oscillations
of the approximation function which diverge rapidly. In the case of the "smooth"
curves commonly encountered in calorimetry, the procedure converges quickly and
without problems.
The desmeared curve tPr(t) is the basis of any further kinetic evaluation since the
result of these procedures strongly depend on the exact shape of the transformed
mass fraction, i. e. the reaction heat flow rate function, (see Sect. 6.2.2.2).
It should be noted that every deconvolution procedure increases the noise. The
better the resolution in time the higher the noise. Desmearing should therefore be
carried out only if it is necessary, i.e. for events (reactions) whose width is in the
same order of magnitude as the time constant of the DSC.
96 5 The DSC Curve
Both numerical desmearing procedures can be applied only after the measure-
ments have been concluded (off-line). For the case ofa discrete (sampled) function,
the convolution integral (Eq. (5.3» can, however, be represented as a linear system
of equations.
;
«Pm(t;)=M L«Pr(tk)'a(t;_HI) (i=I ... n)
k=1
which can be resolved for «Pr(t;):
(5.5)
As can be seen, each value of the desired function can be calculated from the mea-
sured points «Pm(tk)(k :s; i) measured before this moment and the stored discrete
apparatus function a (t;)(i = 1 ... n). Even with this numerical deconvolution it is
then basically possible to calculate and display the desired function online during
the measurement.
The problem is that the measured values (and the apparatus function) are not
quit exact but noisy, i. e. uncertain, and that the errors enter into the calculation
progressively.
In addition, the measured values are small at the beginning of a transformation
and the noise fraction is, therefore, relatively high; the initial values of the appara-
tus function are usually very small which - according to Eq. (5.5) -leads to a very
noisy signal «Pr(t;), which in turn results in numerical instabilities which may lead
to a "run-time error" of the computer. For these reasons, online deconvolution
according to Eq. (5.5) is usually not suitable without special precautions concerning
the initial peak region in question. Of course, the calculations can be performed off-
line as well, i. e. after the experiment has been finished.
5.4.4 Advanced Desmearing
The desmearing procedures described so far have essentially taken into account the
influences from the DSC measuring system and the interaction between sample and
instrument. The events inside the sample have not yet been discussed. For a 1st
order endothermic phase transition of a pure sample, desmearing as it has been car-
ried out so far furnishes, for example, a deconvoluted measured curve as shown in
Fig. 5.8 a. When recorded as a cp function, the 1st order phase transition has, how-
ever, the shape according to Fig. 5.8b. Both functions are obviously not alike. To be
able to calculate the cp function from the measured function, the reasons must be
known quantitatively which have led to the particular function according to
Fig. 5.8 a. In the case of an endothermic phase transition, the sample temperature is
constant for the duration of the phase transition; as the sample crucible is heated
linearly, the heat flow rate into the sample also increases linearly:
l/Jr = L . t::.T= L . (Tm - Ts)
dl/Jr = L. (dT
m _ dTs)
dt dt dt
'--' '--'
=={3 =0
dl/Jr dTm
-=L- (5.6)
dt dt
dl/Jr dt dl/Jr
-·-=-=L
dt dTm dTm
(L thermal conductance between sample crucible and sample, {3 heating rate)
T,
a
A
l
b
Fig. 5.S. Advanced desmearing in the case of the melting of a pure substance.
a Deconvoluted measured curve, b Corresponding cp(T) function.
<PI true heat flow rate into the sample, Tm temperature measured, Ts sample temperature,
TIrS transition temperature, L thermal conductance between sample crucible and sample,
cp specific heat capacity of the sample, pheating rate
98 5 The DSC Curve
Accordingly, the increase in the heat flow rate depends on the heat transport
conditions to the sample and on the heating rate.
In this case, advanced desmearing of the measured curve of a 1st order phase
transition consists in simply compressing the abscissa in the range TJ to T2 (Fig. 5.8)
and dividing the ordinate by the mass and the heating rate. As a result, the triangular
peak becomes a o-peak with the weight factor of the phase transition heat !!H, and
the measurement temperature becomes the sample temperature.
For endothermic transitions of, for instance, impure substances it is to be started
from the fact that the rather low heat flow rate at the beginning of the transition is
scarcely influenced by the limited thermal conductivity. But if the heat flow rate·
increases and approaches the magnitude of the phase transition peak, the shape of
the measured peak will be smeared more and more. At a certain moment (llimit in
Fig. 5.9a) it reaches the heat flow rate limit ofthe apparatus and the sample tem-
perature falls behind the measured temperature. The limiting factor is the same as
for the phase transition of a pure substance, namely the thermal conductance L of
the path from the temperature measuring point to the sample. The measured end
temperature is not the end temperature ofthe transition of the sample (Fig. 5.9b).
There is a nonlinear connection between these two temperatures which generally
cannot be specified.
In the case of impure substances which can be described with the aid of thermo-
dynamics of eutectic multi-component systems, the theoretical shape of the
function shown in Fig. 5.9b is known (see Sect. 6.2.4). This allows the true melting
curve to be determined from the measured one by calculating along the theoretical
curve and comparing the areas which must be equal according to the law of the
conservation of energy. Such calculation may be called desmearing as well, but it is
far removed from deconvolution due to the theory of linear response.
Another problem is the smearing of the measured curve due to the temperature
profile developing in the sample. It is evident that the peak assigned to a transition
in a thin sample differs substantially from that in a thick sample. It can be shown by
the method of Laplace transformation that the temperature profile in the sample has
a parabolic shape and that the mean temperature (T) of the sample can be calculat-
ed according to the following formula (Hoff, 1991):
(5.7)
lib being the temperature at the low "hot" boundary of the sample (Cp heat capacity,
p density, A. thermal conductivity, f3 heating rate, d thickness).
The second term on the right-hand side ofEq. (5.7) is an additional correction of
the temperature scale. In reality both the finite thermal conductivity of the heat path
to the sample and the temperature profile of the sample influence the shape of the
measured peak. In the case of linearity, the total apparatus function (often called
Green's function as well) is the convolution product (Eq. 5.7) of both parts, as
these events are connected in series (Hohne, Schawe, 1993). This seems to com-
plicate the desmearing procedure, but fortunately the total apparatus function can
be determined in one moment by analysing the switch-on and switch-off behavior
S.4 Desmearing of the DSC Curve 99
t /
/"/ :
I
/ I
~ ________ ______ f / _____ +I ______ L-______
~
I I
I I
I I
I I
Tend IIrsi
b
Fig. 5.9. Advanced desmearing in the case of a melting of an impure substance.
a Deconvoluted measured curve, dashed line hypotenuse of a right-angled triangle with the
slopeL,
b Corresponding cp(T) function (the temperature scale is not correlated to that of a).
4>, true heat flow rate into the sample, Tm temperature measured, Ts sample temperature,
T... transition temperature, T. extrapolated peak onset temperature, L thermal conductance, cp
specific heat capacity of the sample
of the scanning mode. The starting and ending of a DSC run always implies switch-
ing the heating rate from zero to a constant value and vice versa. As a result, the real
heat flow rate into the sample should change in a step-like manner, whereas the
measured heat flow rate rises with a certain delay (Fig. 5.10). This step response
function covers both the heat transport and the sample behavior. By differentiation
the impulse response function can be derived from this function (Fig. 5.11), which
is the apparatus function in which the investigator is interested. Using this appara-
tus function desmearing can be carried out as described. Figure 5.12 shows that
such a desmearing gives rise to a distinct change in the measured curve in the case
100 5 The DSC Curve
t 2~~----
time ..
Fig. S.10. Curve measured with the heating rate switched from zero to fJ at to.
1 measured curve, 2 theoretical heat flow rate into the sample
0.8
5- 1
0.6
t
.....
0.4
<:I
0.2
.
0.0
0 10 20 30 40 5 60
time
Fig. S.l1. Apparatus function a(t) from step response arising from
1 switch-on, 2 switch-off
of samples with poor thermal conductivity. This may be used as a proof of the in-
fluence which the thermal lag and the temperature profile inside the sample exert
on the measured curve. An advantage ofthis method is the easy measurement of the
correct apparatus function by only switching from the isothermal to scanning mode
and vice versa which is always done in the case of heat capacity measurements (cf.
Sect. 6.2.1). Thus the apparatus function in question is always on hand.
The lag between sample and measured temperature due to the thermal resistance
between sensor and sample causes an error not only because the shape of a peak is
2.2
J
glf
1.8
f 1.6
~
LJ
1.4
.
1.2
350 360 370 380 390 K 410
tempera ture
Fig. 5.12. Heat capacity measurement showing a glass transition.
1 measured curve, 2 desmeared curve (10 mg polystyrene, 20 K min-I)
heating rate {3
f
-e.
Q.J
~
~
'"
::::J
'-
f
-e.
~
c:I
c....
-
~
0 "
~
c:I
Q.J
..r::. .....
c:I
c..
.....
c:I
lag c:I
QJ
..r::.
T
Fig. 5.13. Difference between true and measured heat flow rate (resp. heat capacity) curves.
T program temperature, T" starting temperature, Ts sample temperature, Tm measured tem-
perature, <Ptruc true sample heat flow rate, <Pm measured heat flow rate
102 5 The DSC Curve
smeared, but also the baseline and the measured curve when pure Cp measurements
are performed. Figure 5.l3 may serve to explain this "thermal lag" effect: There is
in principle a difference between the measured temperature of the sensor and the
true temperature ofthe sample during a scanning run. It increases with the heat flow
rate into the sample due to the proportionality of heat flow rate and associated tem-
perature gradient. The heat flow rate for its part depends on the heat capacity of the
sample and on the heating rate. As a consequence, the measured curves shift more
and more to the right side in Fig. 5.l3 with increasing heating rate and heat
capacity of the sample.
Due to the proportional relation between heat flow rate and heat capacity of the
sample, an analogous relation is valid for the difference between the measured
apparent heat capacity (calculated from the measured heat flow rate) and the true
heat capacity of the sample. As the sample temperature is lower than the tempera-
ture displayed, the heat flow rates (and the heat capacities) displayed are also too
low - compared with the true ones - due to the increase in the sample's heat capa-
city with temperature (see Fig. 5.l3).
As the temperature dependence of the heat capacity is normally not very strong,
the effect discussed (the thermal lag) is not very important. For highly precise
measurements « 1%), it has, however, to be taken into account and, if necessary, it
should be corrected by applying the desmearing procedures referred to above. The
thermal lag oTitself can be approximately determined (Richardson, Savill, 1975b
and Vallebona, 1979) from the heat oQ, which is proportional to the area A between
the step function at varying scanning rate and the measured function which is the
step response function (i.e. the area between the two functions in Fig. 5.10):
oQ
3T=-
Cp,s
(oTthermallag in K; A area in J; Cp,s heat capacity of sample in JK- 1)
With this value, the temperature scale can be corrected for the lag of the sample
temperature caused of the finite thermal conductivity and the influences of the
sample.
5.4.5 Further Calculations
The desmearing procedures described above furnish the heat capacity-temperature

function of the sample, Cp(Ts). An integration must be carried out to obtain the
thermodynamic potential function:
T2
l1H (T) = f Cp(T) dT
TI
Usually, the pressure p is constant in calorimetric experiments and the enthalpy

difference l1H(T) is obtained from Cp; (see note on l1H in Sect. 5.3).
If the measurement is carried out at constant volume, in analogy to Cp , first
Cv(Ts) is obtained and from this, by integration, the energy difference l1U(T) as the
5.5 Interpretation and Presentation of Results 103
thennodynamic potential function. (This is not very important regarding DSC

measurements as the realization ofa constant volume is very difficult.)
The time and effort to be spent on desmearing seems to be great and the pro-
cedure complex. It should, however, be noted that the individual steps of the de-
smearing procedure described in this section must be carried out only ifthe accuracy
required for the measurements is higher than the correction to be expected as a
result of the respective procedure. For most applications, both deconvolution of
instrument influences and the advanced desmearing procedures can be dispensed
with; in peak area detennination they are absolutely superfluous. They are, however,
usually required when, for example, kinetic evaluations or a precise analysis ofthe
phase transition behavior of substances is concerned.
5.5 Interpretation and Presentation of Results

For certain tasks (for example, receiving inspection of materials), by comparison of
the measured curve with reference curves, an identification of the sample substance
is perhaps possible on the basis ofa yes/no decision. This generally concludes inter-
pretation in these cases.
In the nonnal case, however, interpretation is preceded by the evaluation of
the curve in order that data are obtained which are characteristic of the sample
substance and/or the transition investigated. The steps of the evaluation are the
following:
desmearing, if necessary (cf. Sect. 5.4),
detennination of the peak area or partial areas after construction of the baseline
(cf. Sect. 3.1 "The first approximation" and Sect. 6.2.2.2),
detennination of Cp changes (cf. Sect. 6.2.3),
- estimation of the uncertainties.
The data obtained from the evaluation fonn the basis ofthe interpretation. The pre-
sentation of uninterpreted measured curves is not very infonnative. An interpreta-
tion can possibly be made only in several steps, for instance by
- variation of the sample parameters (particle size, mass, sample shape, ... ),
- variation of operating parameters (heating rate, kind of atmosphere, ... ).
The interpretation must correctly describe the effects of the parameters changed.
Sometimes, for example with complex reactions, a final interpretation by means of
the DSCs is not possible. Other methods must then be applied in addition. Only if
the results obtained by all test methods can be interpreted in the same way is it prob-
able that the interpretation is correct. For example, in the case of kinetic investiga-
tions in inhomogeneous phases, one should always be aware of the fact that the cal-
culated values are affected by considerable uncertainties.
When DSC results are presented, the following should be stated:
- sample characteristics (mass, purity, structure, ... ),
- instrument characteristics (type ofDSC),
104 5 The DSC Curve
- test conditions (heating rate, atmosphere),

the curve originally measured (and a desmeared curve, should desmearing be
necessary),
details of the calibration procedures (materials and their characteristic data),
details of the evaluation of the measured curve (specification of characteristic
temperatures, construction of baseline, peak integration),
data obtained from the measured curve and the uncertainties by which they are
affected,
basic formula and calculation procedures which are used (e.g. for purity deter-
mination),
interpretation on the basis of the DSC results (taking the uncertainties into
account),
whether the interpretation could be confirmed by variation of the parameters or
by other measuring methods,
- comparable results from the literature.
6 Applications of Differential Scanning Calorimetry
The output signal from a DSC, the heat flow rate as a function of temperature, and
any derived quantity, such as the heat of transformation or reaction or any change
of the heat capacity of the sample, may be used to solve many different problems.
The work required to evaluate the measured curve may differ greatly from one case
to another. This will become clear from the following text. It will also become
obvious that an adequate evaluation may need more effort when the signal to noise
ratio is low.
6.1 General Applications Without Sophisticated Evaluation

of the Measured Curve
The required information can often be obtained from only a qualitative evaluation
of the DSC curve. A simple yeslno-decision is sometimes enough. Quantitative
information may not be needed.
6.1.1 Identification of Substances, the Phase Behavior
Such investigations are more meaningful only when they are coupled with other
structure-sensitive analytical methods such as hot stage microscopy, infrared
spectroscopy and X-ray diffraction. The main advantage ofthe DSC method lies in
the ease and simplicity of operation - in particular both the choice and the rapid
change of the required temperatures. This is very important for the investigation of
substances with metastable phases. On the other hand there is the disadvantage that
it is not possible to carry out experimental manipulations (e. g. nucleation of a
supercooled liquid) inside the small and practically always closed crucibles.
If the phase behavior of a sample is reversible, a good practice should always be
the evaluation of the second run rather than the first, because only then is there a
reproducible heat transfer pass from the sample to the apparatus. Investigations
regarding the polymorphism of pharmaceutical substances are especially impor-
tant, as different modifications or isomers have different rates of solution and thus
even different physiological effects on human organisms.
The first example (Fig. 6.1) concerns the melting of acetamide. During the first
heating the thermodynamically stable modification (the as received state) melts at
about 80°C. On cooling from the melt the metastable modification crystallizes first
106 6 Applications of Differential Scanning Calorimetry
50.0
mW
37.5
"\
f /
2
25.0 /
Cl
c... \
~ /
2 \
0
;:;::
~
Cl
OJ 12.5 /
..r:::;
o
50
-----
58
- /
66 74 82 O( 90
tempera ture ..
Fig.6.1. Melting of the thermodynamically stable (curve 1) and of the metastable (curve 2)
modification of acetamid
in accordance with Ostwald's step rule. The second run thus shows the melting peak
of this metastable modification at about 65 °e. The formation of metastable phases
is one of the main problems encountered in the determination of phase diagrams,
and this possibility must always be borne in mind to avoid misinterpretation of the
observed curves.
Phenylbutazone is another well-known material that shows polymorphism. The
stable modification (curve I in Fig. 6.2a) melts at about 103 °e. Rapid cooling of
the melt yields a glassy phase. On reheating the glass (curve 2) this first recrystal-
lizes to a metastable form at about 35°e. Subsequent behaviour depends on the
heating rates, either melting of the metastable modification at about 93 °e or ofthe
stable structure at 103 °e will be observed. Because the rate of transformation
from the metastable to the stable form is low in the solid phase (i. e. < 93°C), only
at low heating rates « I Kmin- I) will there be enough time during the experiment
for this to occur. At high heating rates (> 20 K min-I) the formation of the
stable modification does not happen, not even in the liquid phase after the melting
ofthe metastable modification. The corresponding Dse curves are shown in Fig.
6.2 b. For better comparison of both c,urves, curve 2 was also run with a heating rate
of 40 Kmin- I following a preliminary heating to 98 °e at I Kmin- I and recooling
to room temperature. At moderate heating rates (ca. 5 Kmin- I), however, the sam-
ple reaches some intermediate state. Any remaining metastable phase that has not
transformed during the slow solid state reaction, melts at 93 °e, it then transforms
6.1 General Applications Without Sophisticated Evaluation of the Measured Curve 107
20
mW
15
r 10
~
c
'-
~
0
c
C1J
..<:: 5 2
3
0
25 40 55 70 85 100 DC 115
a tempera ture ~
70.0
mW
52.5 2
1C1J
30.0
~
c
'-
~
~
~
C
<1J
..<::
17.5
o
55 70 85 100 115 DC 130
b tempera ture
Fig.6.2. Polymorphism of phenylbutazone.
a I: first heating run, 2: heating run after quenching from the melt, 3: heating run after reheat-
ing the quenched melt up to the end of the exothermic peak at 100°C (mass: 5.1 mg, heating
rate: 5 K min-I),
b two limiting cases in heating the quenched melt, 1: complete melting of the metastable phase
during fast heating, 2: melting of the stable phase after previously heating to 98°C with
I K min-I (heating rate: 40 K min-I)
to the stable modification as an exothermic reaction and finally melts at 103°C

(curve 2 in Fig. 6.2a). This can be proved by quenching the sample immediately
after completion of the exothermic reaction (at 100°C) and reheating it again, when
only the melting peak of the stable phase is seen (curve 3).
DSCs are used very frequently for the investigation of polymorphism of sub-
stances with liquid-crystalline mesophases. The reliable evaluation of mesophase
transitions with small or very small transition energies (sometimes <0.1 J g-I)
makes great requirements on apparatus and working conditions. An example is the
behavior of cholesteryl myristate during heating and cooling (Fig. 6.3). The first
peak (Q = 77.3 J g-I, T = 69.0 0c) corresponds to the formation of a smectic meso-
phase. The following two peaks (Q = 2.5 J g-I, T = 77.7 °C and Q = 1.9 J g-I,
T= 83.0 0c) correspond to the formation of the cholesteric mesophase or the tran-
sition into the isotropic melt. All phase transitions are reversible (compare curves 1
and 4). The reason for the narrower peak shape of the first peak of curve 4 (Q = 76.9
J g-I, T = 68.5 DC), compared with that of curve 1, is the better heat transfer to the
sample after melting. If the cooling run (curve 2) is stopped before the crystalline
phase appears, only the two liquid-crystalline phase transitions are observed during
the subsequent reheating (curve 3). The heat of transition from the crystalline to
the liquid-crystalline mesophase is usually - as in this example - much higher than
those of the following mesophase transitions. Further, it can be demonstrated that
there are practically no supercooling phenomena for mesophase transitions. The
onset temperatures of the heating and cooling runs do not differ from each other
20
mW
15
1
.::;
~
10
=
:Jo
-::;
OJ
..c:
..
65 70 75 80 85 O( 90
60
tempera ture
Fig. 6.3. Polymorphism of cholesteryl myristate.

1: first heating, 2: cooling to 60°C, 3: subsequent heating, 4: heating after cooling from the
melt to room temperature (mass: 4 mg, heating and cooling rates: 5 Kmin- I )
0.03
Wig
r 0.02
~
'"
~
0.01
3
~
'"
C1J
.&::
'-' 3
'-'
0
C1J
0..
III
-0.01
18 21 24 27 °C 30
tempera ture •
Fig. 6.4. DSC heating curves of the gel-liquid crystalline phase transition of a Iiposome.
I: 1.75 % of dimyristoylphosphatidylcholine (DMPC) in water, 2: after addition of 3 % and 3:
of5% of vitamin D3 (mass: 15 mg, heating rate: 2 Kmin- ' )
after proper correction of the extrapolated onset temperatures to a heating or cool-

ing rate of zero (onset differences of AT = 0.2 K and AT = 0.0 K for the two transi-
tions). The crystalline phase, however, shows considerable supercooling. It crystal-
lizes at about 30°C, equivalent to a AT of about 40 K.
Besides the pharmaceutical applications already mentioned, OSCs are also used
for investigations of stability and of physical and chemical interactions between the
active components of drugs. Recently OSC has become increasingly used in both
the food industry and biochemistry. As a rule these applications require very high
sensitive OSC measurements. One example (Fig. 6.4) shows the transition from gel
to liquid crystalline state of a appropriately prepared liposome of 1.75% of di-
myristoylphosphatidylcholine (OMPC) in H20. The reversible transition (curve 1,
Q = 0.47 J per g total sample mass, Tp = 24.6 0C) corresponds to the change of the
hydrocarbon chains from an ordered (crystal-like) to a liquid-like structure. The
structural transition of the lipid can be perturbed by molecules which intercalate
among the lipid chains and hence hinder the regular packing of the phospholipid
molecules into gel-phase structures. This can be realized for instance by addition of
the lipophilic vitamin 03. With increasing concentration of vitamin 03 in the lipo-
somes, the peak becomes broader and smaller and the peak maximum shifts to
lower temperatures (curve 2: 3% of vitamin 03, Q =0.15 J g-I, Tp = 22.7 0C). Addi-
tion of more then 5% of vitamin 03 to the OMPC-water lipid (curve 3) causes the
disappearance of this peak.
Sometimes melting processes extend over very broad ranges oftemperature. For
low molar mass substances this may be due to large amounts of impurities (cf. Sect.
6.2.4). For macromolecules this may be caused by the broad distribution of the
molar masses or by a distribution of the thicknesses of the crystal-lamellae. The
determination of the appropriate baseline and the choice of physically reasonable
integration limits for the enthalpy calculation is then very problematic. Mathot,
Pijpers, 1983; Mathot, 1984; Mathot, Pijpers, 1989 recommend the use of the heat
capacity (Cp ) function rather than the normal heat flow rate curve of the sample for
evaluation to overcome these difficulties.
Additional problems are found when DSC is used to investigate processes which
lead to the production or absorption of gases e. g. dehydrations or oxidations. Here
both qualitative and quantitative results are influenced by the experimental con-
ditions - the type and flow rate of the purge gas, geometry of the sample holder,
sample preparation and so on. A typical example is the thermal decomposition
reaction of calcium oxalate monohydrate. The temperature of the first dehydra-
tion peak and the peak width are radically influenced both by the heating rate
and the effectiveness of the water vapour transport by the purge gas. Thus every
attempt at a kinetic evaluation risks interpreting the transport conditions rather
than the kinetics of the decomposition process. The second step in the decomposi-
tion of calcium oxalate monohydrate, the elimination of carbon monoxide, to yield
CaC03 , is strongly influenced by traces of oxygen in the purge gas because in
50.0
crystalline -+ smectic
mW / mesophase
37.5
r nema tic -+
~ liquid pha se
25.0 dehydratation ~
\
~
:J<
0
0
C1J smectic -+ nema tic
..c.
12.5
\
o L -_ _ _ _ _ _ ~ ______ ~ ______ ~~ ______ ~ _ _ _ _ _ __ J
o 40 80 120 160 ·c 200

temperature ..
Fig. 6.5. DSC curve of the phase behavior of calcium stearate monohydrate (mass: 3 mg,
heating rate: 10 Kmin- 1)
6.1 General Applications Without Sophisticated Evaluation of the Measured Curve III
presence of O 2 the CO is completely or partially oxidized. The simultaneous

coupling of the DSC with thermogravimetry or mass spectrometry is advantageous
in these cases.
Another example, the phase behavior of calcium stearate monohydrate is shown
in Fig. 6.5. Shape and temperature of the first peak (dehydration) and the separation
of it from the second peak are strongly influenced by the effective removal of the
evolved water vapour and thus by the experimental conditions (open, partially or
completely closed sample crucibles; flow rate and type of purge gas). The three
other peaks correspond to the transitions from the crystalline to the smectic phase,
from smectic to nematic and finally from nematic to isotropic liquid. Within the
homologous series of fatty acids the phase behavior changes both quantitatively
(for rather small differences in the chain length) and qualitatively (for larger
differences). The DSC method may therefore also be used to solve analytical
problems.
6.1.2 Determination of Phase Diagrams
The determination of phase diagrams in the fields of metallurgy and mineralogy

was one of the earliest applications of thermal methods. It was one of the decisive
reasons for the introduction of Differential Thermal Analysis (DTA, see Appendix
1) at the beginning of the twentieth century.
As an example, the phase diagram of the binary system benzillacetanilide is
shown schematically in Fig. 6.6 together with the DSC curves of the heating runs
of the pure components (curves 1 and 7) and of five mixtures including the eutec-
tic (curves 2 to 6). In accordance with the phase diagram benziI and acetanilide are
completely miscible in the liquid and immiscible in the solid phase. Pure acetanili-
de melts at 115 °C, pure benzil at 95 °C. The eutectic, with a mole fraction of about
XBen = 0.58, melts at 78°C. In Fig. 6.6a two different melting pathways (mixtures
with XI = 0.1 and X2 = 0.9 of benzil) are sketched. The measured DSC curve of
every mixture shows two events: The first is the peak due to the melting of the
eutectic, the subsequent broad endothermic effect is caused by the solution of the
remaining solid component in the equilibrium melt (following the pathway along
the liquidus curve in Fig. 6.6a). The final point of this peak corresponds to the
liquidus temperature according to that mole fraction in the phase diagram. Both the
pure components and the eutectic melt at well defined temperatures. These can be
determined from the extrapolated peak onset temperatures in the usual way,
whereas the determination of the liquidus temperature is not easy. The peak maxi-
mum temperature (if necessary corrected for thermal lag, see Sect. 5.4.4) of the
second broad peak is a good approximation. In every case slow heating rates
«2 Kmin- I ) and small sample masses should be used, to ensure approximate
thermodynamic equilibrium at every moment during the experiment and to avoid
"smearing" (see Sects. 3 and 5.4) of the DSC curve and thus incorrect temperatu-
res. The heating mode is more suitable for the determination of phase diagrams than
the cooling mode, because many organic substances have a tendency to supercool
considerably.
115°C
t 95 °c
78 O(
Ac 0.25 0.5 0.75 Xz Ben

a Xee. --
WIg
61--'-----
g.z:: 5
u
.~ 6
Co
'"
oL-~--~--~~====~====~~~~~
..
60 73 86 99 112 °C 125
b tempera lure
Fig.6.6. The phase behavior of the system acetanilide (Ac) - benzil (Ben).
a Schematic phase diagram of the eutectic system, the arrows mark the course of the melting
process for two mixtures with the mole fractions XI and X2'
b DSC curves of the pure components and of five mixtures, I: acetanilide (x_= 0), 2: xBen =
0.1,3: x_= 0.4, 4: XBcn= 0.578 (eutectic), 5: x Ben = 0.75, 6: x Ben = 0.9,7: benzil (XBen= I)
If the heats of fusion of the pure components are known and the binary system
behaves ideally in the liquid phase, the theoretical melting course may be predicted,
using the SchrOder-Van Laar equation (Brezesinski, Dorfier, 1983; see also Sect.
6.2.4).
The molar heat of fusion of the eutectic dru.Heu can be calculated for ideal eutec-
tic systems (for which the eutectic is a simple blend of microcrystals) using the
normal rule of mixtures:
XA, X B mole fractions of the components A and B, tlrusHA' tlrusHB molar heats of
fusion of the pure components.
For precise calculations tlfus HA and tlfusHB should be converted to the eutectic
temperature using Kirchhoff's law.
If this simple additivity holds, the exact composition of the eutectic can be deter-
mined by plotting the heats of fusion of the eutectic of different mixtures against the
composition. This gives two different straight lines to the right and to the left of
the eutectic and the intersection gives the required composition. This procedure is
generally also a good approximation for non-ideal mixtures.
For real systems, discrepancies between calculated and experimental eutectic
heats of fusion can lead to conclusions regarding the nature of interactions between
components. From these differences, the excess functions of entropy, enthalpy and
Gibbs energy can be calculated (Rai, Shekhar, 1993). For positive (negative) excess
free energy, the attractive interaction between molecules of the same kind is
stronger (weaker) than that one between molecules of different components in the
mixture.
6.1.3 Safety Aspects
Specific devices for this type of investigation are the reaction or safety calorimeters
mentioned in chapter 1. They provide detailed information for the optimization of
process parameters, reducing production costs and avoiding safety risks. However,
valuable results may be obtained using conventional DTA or DSC (Hentze, 1984).
Safety risks in the chemical industry occur when materials, known for their exo-
thermic decomposition reactions, have to be stored for long periods at temperatu-
res, such that the heat evolved is greater than that transferred to the surroundings.
Under worst case conditions this may lead to a runaway and subsequent fire or
explosion. The direct detection at ambient temperature ofthe extremely small heat
flow rates in question is, in general, not possible with DTAIDSC (exceptions are
Cal vet calorimeters, cf. Sect. 2.1.2).
Hentze, 1984, avoided this problem by annealing the substance
- at constant temperature and various times
- constant times and various temperatures
and comparing the subsequently measured melting and decomposition peaks with
those of a fresh, unannealed sample. In this way quantitative results were obtained
for the decomposition kinetics of, for example, p-nitroaniline. Using this procedure
it is possible to determine decomposition heats down to 10 mWkg- 1 at storage tem-
peratures.
Decomposition reactions often react sensitively to traces of foreign materials.
With the annealing technique described above, the first detectable decomposition of
azidophenyl acetic acid was determined. In addition the effects of "impurities" of

many kinds were investigated. In a similar way the autocatalytic influence of the
decomposition products was definitely proved.
Another variant of investigations in the field of safety techniques, suggested by
Hentze, 1984, utilizes very slow (0.01 to 0.1 Kmin- 1) heating of the sample in
home-made devices of the DTA type. During the decomposition both heat and
pressure are recorded. For the identification of the decomposition products the
DTA apparatus is coupled to a mass spectrometer. By this technique the start of
the decomposition of p-nitroaniline was observed at about 200°C, in very good
agreement with the corresponding value from annealing experiments.
6.1.4 Characterisation of Polymers by Thermal Analysis; Effects of Origin

and Thermal History
Quality control of a material is extremely important both for inspection on delivery

and to insure the integrity of manufactured products, this is especially relevant in
the plastics industry. Until recently the prime reason for inspection was to ensure
continuity of properties if, for instance, the manufacturer, the batch number, or even
the manufacturing process, were changed. Now, with the need for recycling used
plastics many more questions must be answered, and thermoanalytical methods
(DSC, TG, TMA, DMA) have proved very powerful in this respect. The results are
easily comprehensible and readily interpreted. This is especially so when plastics of
the same chemical nature are recycled for the first time and either analyzed as such
or as a mixture with the original plastic. Another important application is the
investigation of service failures of plastics.
The DSC curve can be analyzed with respect to the following quantities:
temperature(s) and step height(s) of the glass transition(s).
The glass transition temperature reflects the composition of the components,
conditions during production, and sometimes also the aging of the plastic
product during its use (caused, for instance, by decreasing molar masses and a
change of their distribution due to cracking processes by partial oxidation). In
heterogeneous or microheterogeneous amorphous blends two or more glass tran-
sitions are found, which are characteristic of the individual components. The
relative amounts can be estimated by evaluating the respective step heights.
Further important information can be taken from the possible existence and
intensity ofa relaxation peak in the region of the glass transition (cf. Sect. 6.2.3).
- temperatures and heats of fusion, transition and crystallization effects.
As a rule it is not difficult to identify the components of a plastic mixture, using
the characteristic temperatures of the melting peaks. Often even a quite accurate
quantitative analysis is possible. Qverlapping peaks can be separated by using
appropriate evaluation methods (Opfermann, 1992), since the necessary conditi-
on for using such programs - the additive superposition of the respective peaks
because of immiscibility in the solid phase - is nearly always fulfilled for plastic
mixtures.
A relative (within the type of plastic) degree of crystallinity for semicrystalline

products can be obtained very easily (ignoring all precautions described in Sect.
6.1.5) if the area of the melting peak is compared with that of the completely
crystalline material of the same type. If a crystallization peak is observed prior
to the melting peak, then the crystallization process of the material during
production was incomplete. In extreme cases, it is found that for bulk thermo-
plastic moulded parts the degree of crystallinity varies within the parts, due to
uneven cooling rates during moulding.
onset of thermal and oxidative decomposition.
To determine this, the plastic is either annealed isothermally in the DSC until the
appearance of an exothermic or endothermic signal indicates the beginning of
decomposition, or the sample is heated in scanning mode up to this event. By
choosing an appropriate gas atmosphere (inert or reactive e.g. oxygen), infor-
mation is available concerning the thermal or oxidative stability of the material.
For these measurements there is always a risk of contaminating the measurement
system, so the run should be interrupted immediately after detecting the start of
decomposition. However, thermogravimetry is in principle much more suitable
for this type of investigation.
existence and intensity of exothermic reactions peaks.
The cure of resins and lacquers depends on both the chemistry (including
stoichiometry, filler and catalyst content) and on the specific heating procedures.
As a result the cured products may differ considerably in their thermal and
mechanical properties. This can be demonstrated either by repetition of the
experiment under different conditions or by coupling with other methods (IR,
GC, NMR, TMA, DMA).
For products containing recycled plastics, most of the above mentioned thermal
properties may have changed in a characteristic manner. The observed deviations
from the original behavior provide reliable information whether the plastic in que-
stion is suitable for the intended application or not.
6.1.5 Determination of the Degree of Crystallinity
DSC is widely used for determination ofthe degree of crystallinity We (weight frac-
tion crystallinity) of sernicrystalline polymers. The most simple and obvious possi-
bility is to relate the estimated heat of fusion (l1fusB) to that of the same polymer but
with known crystallinity, provided that this value was determined by an independent
method such as X-ray diffraction, Raman scattering, density measurements etc. The
determination of the required peak area is problematic, because it is not easy to fix
the lower integration limit. The melting of polymers often starts more than 50 K
below the maximum temperature of the peak and it is very difficult and rather
arbitrary to determine the beginning of this, at first, very slow process on the DSC
curve. The situation is made more difficult by the fact that both the degree of the
crystallinity and the heat of fusion are dependent on temperature. The use of the
heat of fusion measured at higher temperatures yields a value for the crystallinity
which is almost certainly too small compared with the value that is actually required
- that at room temperature. Nevertheless, this simple procedure is sufficient for

determining relative (within the same type of plastic) crystallinities (e.g. for
quality control, cf. Sect. 6.1.4), if one uses fixed integration limits and a fixed
reference value for the heat of fusion of the totally crystallized sample.
Mathot, Pijpers, 1983 and 1989, suggest a convenient procedure (Fig. 6.7), which
avoids the difficulties mentioned. The heat flow rate after the melting peak, which
is proportional to the heat capacity function Cp •• ofthe amorphous (liquid) polymer,
is extrapolated to that temperature T, at which the degree of crystallinity wc(T) is to
be calculated. If, in doing so, the experimental curve is not intersected (T> T*, e. g.
T2 in Fig. 6.7), only the areaAI has to be evaluated. If the extrapolated line intersects
the measured curve (T < T*, e.g. TI in Fig. 6.7), the area A2 has to be subtracted
from AI' The heat, corresponding to the areas Al (or A 1- A 2), is then divided by the
reference value liru.h (T) of a completely crystalline sample at this temperature. For
linear polyethylene Mathot, Pijpers, 1983, suggest the following equation for the
latter function:
liru.h (T) = 293 - 0.3092 x 10-5 • (414.6 - T)2 . (414.6 + 2T) J g-I
Al
f
QJ
~
c Al
....
C
QJ
.c
r,
temperature ..
Fig. 6.7. Determination of the degree of crystallinity according to the procedure of Mathot,
Pijpers, 1983, at two different temperatures T. < T* and T2 > T*, where T* is the temperatu-
re of intersection of the extrapolated liquid phase curve and the measured curve (for details
see text)
The degree of crystallinity we(D in dependence on temperature is then obtained

according to:
Meanwhile the functions h.(D for the completely amorphous and he(D for the
crystalline form are known for numerous linear macromolecules with sufficient
accuracy (Wunderlich et aI., 1990).
If the specific heat capacity curve is evaluated from the DSC measurement (see
Sect. 6.2.1), a simple integration from To to T of the cp-curve - and the respective
functions of the amorphous and crystalline cp-function (from literature) - yields
the enthalpy difference function h (D (resp. h.(D and he(D). This opens a new
thermodynamic based calculation ofwe(D, without the necessity of knowing the
course of the, often non-linear, baseline in the melting range:
Of course, the baseline in the melting interval can be calculated too, if a suitable
model and the functions h.(D, he(D and we(D are known for the semicrystalline
polymer. The simplest assumption is that the properties of a semicrystalline poly-
mer can be adequately described by a two phase model with only crystalline and
amorphous parts. The measuring curve is then calculated according to:
The first two terms correspond to the baseline, the third one describes the tem-
perature dependent change of the degree of crystallinity (the peak). Unfortunately
the two phase model fails with some types of polymers, e.g. polyethylene-
terephthalate. The baseline can be calculated in such cases by use of improved
models, which include a third phase called the 'rigid amorphous phase' (Suzuki
et aI., 1984, Schick et aI., 1985). This phase is thought to exist on the lateral surface
of the crystal lamellae and should have a very limited mobility. It follows from
this that the rigid amorphous phase does not contribute neither to the thermal glass
transition nor to the melting heat of the crystalline part. Nevertheless, the heat capa-
city of this non-crystallized material can be approximated by that of the crystalline
phase.
For some polymers (polydimethylsiloxane, polybutadiene) the rigid amorphous
phase is absent and good results for We may be obtained for these, if the measured
step change of the heat capacity at the glass transition temperature Tg (cf. Sect. 6.2.3)
is related to that of the completely amorphous polymer yielding the amorphous
amount w. = !:.Cp(Tg)/!:.Cp.• (Tg). In this case the degree of crystallinity is We = I - w•.
The completely amorphous state can often be obtained by quenching or ultra-
quenching the molten polymer.
6.2 Applications Requiring a Detailed Evaluation

of the Measured Curve
All the examples described in this section demand precise measurements and crit-
ical, very often special, evaluation procedures of the measured curve. In every case
the basis of reliable results is a careful calibration of the DSC (see chapter 4). As a
rule, the separately measured zero line (see Sect. 5.1) has to be subtracted from the
measured curve before evaluation. In every case, the relationship between uncer-
tainties in the measurements and the quantities to be determined must always be
borne in mind.
6.2.1 Measurements of the Heat Capacity
A knowledge of the heat capacity of a material as a function of temperature is the

basis for determination of any thermodynamic quantity. The use of normal, not
hermetically sealed, DSC crucibles (with a lid which rests on the sample and may
be lightly closed by crimping), always gives the heat capacity Cp at constant pres-
sure. The situation is somewhat more complicated if one uses hermetically sealed
crucibles or the special crucibles which are available for pressures up to the order
of a hundred bar. In addition to the condensed phases, the heat capacity of which is
required, sealed crucibles always contain a gaseous phase. In this case it makes no
difference whether this phase is composed of air or of gaseous reaction products.
Strictly speaking, neither Cp nor Cvare obtained because the thermal expansion of
the sample cannot be prevented and the pressure of the gas changes. However, the
pressure dependence of the heat capacity of condensed phases is very small and as
the change of pressure in the sealed crucibles is generally small, the measured heat
capacity is nearly the same as that at normal pressure.
In the following ofthis section the suffix ''p'' will be omitted so that C is the heat
capacity at constant pressure Cp and c the corresponding specific (per mass unit)
quantity cp for the sample (subscript S) or reference (subscript R).
According to Sects. 3.1 and 3.2 the basic equation for heat capacity determination
is valid both for heat flux calorimeters and for power compensating DSCs. As the
true heating rates of the sample and the reference material are not accessible by the
experiment, they must be replaced by the average heating rate {3. If the heat capacity
C R is known, Cs can be determined easily and quickly from the measured differen-
tial heat flow rate !l.c:I>gR. Several variants of the experimental procedure are known,
four widely used techniques will be discussed.
6.2 Applications Requiring a Detailed Evaluation of the Measured Curve 119
6.2.1.1 The "Classical" Three Step Procedure
The procedure is illustrated in Fig. 6.8. The temperature-time curve during an ex-
periment is outlined in the lower figure and the response of the calorimeter is
shown above. The three steps are:
1. Determination of the heat flow rate of the zero line <1>0 (T), using empty crucibles
(of equal weight) in the sample and the reference sides. The temperature pro-
gram should only be started when the isothermal heat flow rate at the starting
temperature T.t has been constant for at least one minute. If the DSC is compu-
ter controlled, this can easily be automated by checking the differences between
the current average value of the heat flow rate and that one minute ago with allo-
wance for a predetermined drift level. The scanning region between T.t and T.nd
can be 50 to 150 K in modem calorimeters. At the isothermal end temperature
Tend the above computer check must be repeated. For the evaluation procedure all
three regions of the curve are needed. The zero line reflects the (inevitable)
asymmetry ofthe DSC.
2. A calibration substance (Ref) of known heat capacity CRef is placed into the sam-
ple crucible (S), whereas nothing is changed on the reference side (R). Using the
same experimental procedure as for the zeroline, the following is valid:
CRef mRef f3 = Kt/J(T) (~ef - <1>0)
Kt/J(T) is a temperature dependent calibration factor (cf. Sect. 4.4.1).
3. The calibration substance (Ret) in crucible S is replaced by the sample (S). In
analogy to the equation above we get:
Cs ms f3 = Kt/J(T) (<1>s - <1>0)
The specific heat capacity Cs (at a given temperature) can. be calculated by a
simple comparison of the heat flow rates into the sample and into the calibration
substance as illustrated in Fig. 6.8:
The calibration factor Kt/J(I) need not, therefore, be known explicitely. Ifthe con-
dition ms Cs "" mRefCRef holds, the experimental conditions are very similar to
those of the second step. Many of the possible sources of error for DSC mea-
surements then tend to have at least partial compensation.
For the previous and the following considerations it is always assumed, that the
same crucible has been used on the sample side. If during the second and third step
different crucibles must be used, crucibles of the same kind with nearly the same
mass (mer) should be used. It is possible to make routine measurements using cru-
cibles of different masses if allowance is made for the different thermal responses
according to:
12
sample --...
iso- :
Iher~ ribratiOn subs lance
mal
al isolhermal
'" Tst aI Tend
'"
~ 6 scanning mode
0
-e-
"-
0
1-;;
-e-
1 ~
...'"
..c:
~
-e-
~
\. emply pan :)
o
o 1st 4 12 min 16
a lime ~
r
~
=>
T,I
::...
-...
0-
E
1st
b lime •
Fig. 6.8. The conventional three-step technique for the determination of heat capacity.
a schematic course of measurement,
b the temperature change during the run.
T.. start temperature at time I.. , T.... end temperature at time t.... , CZIs, 4>tt.r, "'0
heat flow rates
into sample, calibration substance and empty crucible respectively, ACZIsR differential heat flow
rate between sample and reference crucible
The specific heat capacity of the crucible material is needed only as a correc-
tion. Those for common crucible materials are known with sufficient accuracy.
Omitting the correction results in an error < 1%, if the masses of all crucibles (of
AI) differ by less than 0.03mg for a sample mass >10mg (specific heat capacity
>0.5 J g-I K-I).
Sources of error:
Ideal and real conditions during the recording of the zero line and measured curve
of the sample are compared in Fig. 6.9. Three differences are obvious:
The quasi steady-state conditions in the scanning and final isothermal regions
are not reached immediately after changes in the scanning program but with a
certain delay.
ii The measured heat flow rate (with zero line subtracted) may be smaller than the
ideal (theoretical) one.
iii The isothermal levels at tst and tend differ from each other (and may often have
non-zero values).
These discrepancies result from the finite thermal conductivity of the path between
temperature sensor and sample and from the limited thermal conductivity of the
sample itself(cf. Sects. 5.4.3 and 5.4.4). The sample operates both as a heat capa-
city and as a heat resistance with respect to the thermal surrounding (PoeBnecker,
1990). The signal is therefore a summation of the heat flow stored in the sample
and that which passes through it (heat leak). (Of course, it always appears as the
differential heat flow rate between sample and reference sides).
r---
I
I
I
'"
.... I
I
-
3:
o
I
I c
I B
I
I L ______ _
time
Fig. 6.9. Idealized (dashed line) and real (solid line) curves during a heat capacity measure-
ment.
Curve section AC: delay of the heat flow rate due to restricted heat transfer between
sample and sensor, hatched area ABC: the product of thermal lag 13T and heat capacity of
the sample
In the following the causes of the three above-mentioned deviations from

ideal behaviour are considered in detail and possibilities for their correction are
given.
The smearing of the measured heat flow rate curve during the beginning of
the scanning region reduces the steady state temperature range over which cal-
culations are valid. The initial unusable temperature range can be estimated by
tlT= 5 to 10 times f3. "relf' The effective time constant "relfresults from a coupling
of the time constants for the sample and apparatus. As a rule the influence of the
apparatus is predominant. The time constants of modem DSCs may vary from
2 to lOs. For thicker samples with poor thermal conductivity (e.g. polymers) the
influence ofthe sample may dominate "rolf.
11 As discussed in Sect. 5.4.4, the sample temperature is always lower (higher) than
the program temperature during the heating (cooling) mode. The measured heat
flow rate (fJm always differs from the true value (fJtr. Assuming the worst condi-
tions (large samples, high heating or cooling rates, large heat capacity, bad ther-
mal contact between crucible and sample holder), the difference between both
temperatures may be more than 10K. This temperature error 8T (the thermal
lag) can be estimated (see. Sect. 5.4.4) from the heat 8Q, which is proportional
to the area ABC in Fig. 6.9. This procedure gives a reasonable approximation
even for thick samples and/or those with poor thermal conductivity. Although
there is still a rather large temperature gradient in such samples, there is a
marked reduction in the overall temperature error after correction for thermal lag
(Hanitzsch, 1991). As an example the Curie temperatures ofNi (sample mass ca.
250 mg) differed bylO K for the original heating and cooling runs, whereas the
difference was reduced to 3.6 K after using this temperature correction method.
For a particular sample mass and heating (or cooling) rate the differences
between true and measured heat flow rates are influenced by the thermal con-
ductivity of the sample and by the heat transfer resistance between sample and
sample holder. The heat transfer resistance can be minimized by proper sample
preparation and by correct positioning of the sample in the DSC. It is essential
to ensure completely flat bases for the crucibles, uniform sample thickness, size
and position.
Thermal conductivity effects can be partially compensated if the calibration
substance has a heat capacitiy and a thermal conductivitiy similar to that of the
sample. Thermal conductivities of common calibration substances fall in the
following order (values in W cm- 1 K-l):
Cu (4.01) > Pt (0.72) > sapphire disk (0.34) > powdered a-A120 3
According to a GEFTA (German Society for Thermal Analysis) recommenda-

tion (Sarge et aI., 1994) Pt is only partly suitable, because heat capacities are not
known to the required accuracy (<; 0.5 %). If one uses copper, oxide layers on the
surface and oxygen in the gaseous phase must be excluded.
The best general method for the correction of all effects due to finite thermal
conductivities is to use the special desmearing procedure described by Schawe,
Schick, 1991 (cf. Sect. 5.4).
iii The isothermal levels at Tat and Tend (resp. tst and tend' cf. Fig. 6.8 b) for zero line,
calibration run and measurement differ from each other by amounts which
depend on the type of calorimeter, Tat and Tend and the temperature interval in
between. The offset of the isothermal levels must be corrected to zero before the
heat capacities are calculated. The correction is only meaningful if almost com-
parable conditions for all heat conduction pathes can be assumed for the three
successive runs (zero line, calibration substance, sample). However, PoeBnecker,
1990, has shown by a detailed theoretical treatment of the heat transfer in a power
compensated DSC that measurements with large differences in the offsets of the
isothermals should always be rejected. The heat flow rates of the isothermals at
Tat and Tend should not differ more than 5 % of the difference between the heat
flow rates in the isothermal and the scanning region.
If it is assumed (but cf. Sect, 6.1) that the change of the isothermal heat flow
rates with temperature can be approximated by a straight line <P.so(T) within
sufficiently small temperature intervals, the offset correction is very simple.
Fig. 6.10 demonstrates the procedure. If <P.so,st and <P.so.end are the heat flow rates
of the initial isothermal and the final isothermal, the following is valid:
The corrected experimental heat flow rates (,fJcorr(t) are then obtained from:
(,fJcorr(t) = (,fJexp(t) - <P.so(t)
12
W
expo curve ¢exp! n
8
1
....................................................
corr. curve ¢eo,,! n
6
2
...'"
:J:
0
4
'"
(1J
..c. 2
0 \
~
-2
0 2 3 4 5 min 8
tst
time
Fig.6.10. Correction of the experimental heat flow rate curve 4>..P for different isothermals
6.2.1.2 The "Absolute" Dual Step Method
The second step in the Cp measurement procedure (the calibration run, cf. Sect.
6.2.1.1) is not necessary, if the temperature dependent calibration factor K~(T») is
truly constant with time. Then the careful calibration of the heat flow rate within the
required temperature interval need only be done once. The heat capacity of the sam-
ple can then be calculated as follows:
This method obviously saves time compared with the procedure in Sect. 6.2.1.1.
A disadvantage is that conditions of thermal symmetry between sample and cali-
brant are lost and with them the partial compensation of errors.
The dual step method is always a good technique for the investigation of sam-
ples with high thermal conductivities. The comments of Sect. 6.2.1.1 regarding
errors and their corrections remain valid for this procedure.
An interesting possibility is to attempt to set up experimental conditions for the
sample run which are very similar to those for the zero line run. For this purpose the
sample is put into the correct crucible S and a reference substance CRef is placed in
the correct crucible R on the reference side. The (differential) signal 1l.(P will
approach that of the zero line (Po when Cs "" CRef • The sample specific heat capaci-
ty is then given by:
mRef K~(T) . (1l.(P- (Po)
Cs = -- CRef + -------
ms p. ms
At first sight this is a fascinating idea because the accuracy of the heat capacity
measurements should come close to that of the calibration substance. However, the
quality of the measurements is markedly influenced by possible differences in the
thermal conductivities and the heat transfer conditions of both substances (or runs).
Unfortunately it is very difficult or nearly impossible to recognize these influencing
factors and to correct them because:
Conditions for heat transfer may be poor, although rather small offsets of the iso-
thermals may conceal this and even suggest high-quality measurements.
Thermal lag cannot be corrected by the simple procedure outlined mSect.
6.2.1.1. In addition, de smearing has not so far been carried out under these con-
ditions.
6.2.1.3 A Variation of the Classical Technique
Precise measurements of the heat capacity using DSCs require rather narrow
temperature intervals - not exceeding 100 to 150 K - otherwise nonlinearities in
the temperature dependence of the isothermals cannot be neglected. However,
heat capacities are usually required over much larger ranges of temperature. The
region of interest must then be divided into smaller intervals. In this case the
two conventional procedures (cf. Sects. 6.2.1.1 and 6.2.1.2) are cumbersome,
because empty (for zero line) and sample crucibles must be exchanged several
times. The measuring time is increased and errors result from the inevitable lack
of reproducibility in conditions for heat transfer. A better procedure was suggested
by Hanitzsch, 1991 (see Fig. 6.11). The temperature range of interest is scanned
three times, first with empty crucibles, then with the calibration substance and
finally with the sample. After each run the subsequent cooling after a heating run
(or the subsequent heating after a cooling run) to the start temperature Tst is
interrupted every 50 to 100 K to get several isothermal heat flow rate levels.
The intermediate isothenrtal levels are recorded. Fig. 6.11 a explains this pro-
cedure for the case of a calibration with synthetic sapphire. The program tem-
perature and the heat flow rate into the sample are plotted as function of time.
Analogous curves are obtained for the empty pan and the sample run. From the
heat flow rates of the initial, end and intermediate isothermals a polynomial of
third or higher order can be calculated, which describes the generally existing
curvature of the temperature dependence of the isothermals to a very good appro-
ximation:
Fig. 6.11 b shows the isothermal reference line (here 4)0)' the experimental and the
corrected (due to the procedure of Sect. 6.2.1.1) heat flow rate curves. The
calculation of the heat capacities is then analogous to the three step or dual step
technique.
If the calibration factor K~(T) is independent of the temperature - this is a good
approximation for Perkin-Elmer power compensation DSCs - the calculation of the
heat capacities can be further simplified and the required time reduced (Hanitzsch,
1991). If deviations of the calibration factor from unity are known from a preceding
run and these deviations have been included in the hardware of the apparatus or the
software, the quantity K~(T) as third unknown is cancelled in the following two
equations:
calibration run:
measuring run:
The specific heat capacity Cs is then obtained without need for the zero line 4)0;
The third term of this equation can be neglected ifthe masses of the empty crucibles
mer.s, and mer.Ref differ by no more than ±0.03 mg. The comments of Sect. 6.2.1.1
regarding errors and their corrections remain valid for this procedure.
cI>"I."P (t) 3 600
°C
t 400
t
ZOO ~
-
~ :::>
o 7 <>
'-
QJ
0..
E
QJ
~
o 20 40 60 min 80
a time
t
<>
QJ
..c
o 200 400 °C 600

b
temperature
Fig. 6.11. Modified procedure for the determination of heat capacity for non-linear changes
in the isothermals between the start and end of a run (acc. to Hanitzsch, 1991).
a program temperature T (t) and heat flow rate <PeaI,OXJ> (t) during a calibrating run with synthetic
sapphire, I: initial isothermal at T.. , 2: scanning part of the heating run, 3: final isothermal at
Tend, 4 to 7: intermediate isothermals during subsequent cooling to T.. ,
b <PeaI,OXJ> measured calibration heat flow rate, <Po calculated polynomial fit of the isothermals
(3 -7 and I) measured on cooling, <PeaI,con with <Po corrected calibration heat flow rate
6.2.1.4 A Discontinuous Procedure
The techniques described so far allow a large temperature range to be covered in one
long run. This gives a continuous heat capacity-temperature curve. An alteqlative
procedure is to measure exchanged heats (areas). Thus the total temperature range
must be divided into narrow intervals (Fig. 6.12) which are successively scanned
with isothermal periods in between (Flynn, 1974; 1993). Depending on the tem-
perature interval the procedure corresponds to Fig 6.12a or 6.12b. The operation
must then be repeated with empty crucibles. If 1'; and AT represent the initial tem-
perature and the temperature interval respectively, the average specific heat capa-
city for the j-th temperature interval between Tj-i = 1'; + (j - I) AT and Tj = 1'; +j AT
can be calculated from the heat Qj' which is proportional to the area enclosed by the
sample and zero runs:
(c (T)) = KQ(Tj} . Qj
S J AT. ms
It is very important, that the isothermal periods are long enough to attain steady
state conditions. These times differ according to the type of calorimeter. The iso-
peribolic Perkin-Elmer DSC, for example, needs 1 to 3 minutes, the longer times
are needed for samples that are large and/or have poor thermal conductivity.
By analogy with Sect. 6.2.1.2, the empty reference crucible may be replaced by
a crucible containing a calibration sample. The "A" in the symbol A (Qj) of the
following equation is used to emphasize this. If all crucibles, both on the sample and
reference side, are assumed to have the same mass, we obtain:
K (T) . A(Q } m .C
(cs(Tj})= Q j j + Ref Ref
AT· ms ms
This discontinuous technique replaces the disadvantage of the determination of
absolute values of the heat flow rates from the continuous method with the advan-
tage of the integration method for heat determination. A further advantage is that no
correction for thermal lag is required but the temperature calibration must be very
precise. Ifthe AT-intervals are in the range of 1 to 2 K the calculated average value
(cs(Tj}) reproduces the searched value cs(Tj} sufficiently. The interval may be
increased to 5 or 10K when there is a low temperature dependence of the heat
capacity.
The discontinuous procedure has been little used so far. Possible reasons for
this are:
Long measurement times are needed. To determine the heat capacity over 100 K
with the continuous procedures of Sects. 6.2.1.1 to 6.2.1.3 requires 30 to 60 min
at a heating rate of 10K min-i. In contrast the discontinuous technique takes 400
to 600 min for AT-intervals of 1 K.
This time can, however, be reduced to 150 to 75 min if a 1 K interval is measured
every 5 or 10K only. This is perfectly acceptable when heat capacities are only
slowly changing functions of temperature. The measurement time then becomes
comparable with that of the other techniques.
mW
f
2 3
<:>
c....
Q1
I). T-- -- I). T:---
o
o 2 3 4 5 6 7 8 9 10 11 min 13
a time
12
_ - - 6.T - - - I).T--...;
c
'-
Q,
c
~ 4
(
o
o 2 3 4 5 6 7 8 9 10 11 min 13
b time
Fig. 6.12. Discontinuous method for the determination of the heat capacity.
a small temperature intervals I::r.T, the heat flow rate does not reach the steady-state,
b larger temperature intervals I::r.T, the heat flow rate comes to a steady-state
If adequate software is not available a major effort must be put into the evalua-
tion of data.
- The accuracy of heat capacities measured in this way is not higher than for those
determined using the continuous technique. For normal sample masses (10 to
20 mg), temperature intervals of 1 K and specific heat capacities of 0.5 J g-I K- 1,
the heats to be determined are only in the order of 5 to 10 mJ. This is comparable
with the area of the fusion peak of indium samples with masses of 0.15 to 0.3 mg.
Unfortunately, minor errors in determination of the isothermal heat flow rates,
which are unavoidable, affect the results of the cp determination much more than
is the case for In fusion peaks.
6.2.1.5 Determination Using the Dual Sample DSC
In a dual sample DSC (cf. Sect. 2.1.1) two samples can be analyzed simultaneously
in the same thermal environment. It becomes possible to measure the heat capacity
of a sample in a single run (Jin, Wunderlich, 1993). An empty crucible is placed
in the reference position and the two sample crucibles contain the sample and the
calibrant (Ref) respectively. The advantage of a single-run experiment, beside time
savings, results from reduction of systematic experimental errors (e. g. due to
thermal fluctuations of the signal), compared with the two or three successive runs
necessary with a normal DSC. Assuming ideal symmetry of the cells and exact
sample placement (within ±0.2 mm e.g. using an autosampler), the temperature-
dependent calibration factors K~(T) should be identical for both sample cells. The
Cs(T) function of the sample would then immediately result from the ratio of the
heat flow rates into the sample and into the calibration sample.
In reality there are asymmetries in the measuring system and it is essential to
determine the different K~(T) factors for both sample positions. Taking into account
all possible sources of error (asymmetry of the measuring system, temperature lag,
non-linear temperature calibration, heating rate differences between sample and
reference), the uncertainties (RMS, see Sect. 7.2) of all measurements (using
standard samples of sapphire, selenium, aluminum, quartz, polystyrene, and
sodium chloride with well known data from adiabatic calorimetry) were estimated
to be <1 % at temperatures above 300 K, <3% below 200K, and <2% between
200 and 300 K (Jin, Wunderlich, 1993).
6.2.1.6 General Precautions for the Minimization of Errors
Calculation of heat capacities following the procedures of Sects. 6.2.1.1 to 6.2.1.3

needs true heat flow rates into the sample (the difference between recorded signal
and the zero line). The zero line cannot be obtained simultaneously, but only after
the sample crucible has been replaced once (or several) times. In doing so the
original geometry between sample, crucible and temperature sensors can never be
reproduced perfectly. It follows that changes in the conditions for the heat transfer,
which cannot be immediately detected, are responsible for the great difficulties in
obtaining precise heat capacities. In order to limit the uncertainties to values

smaller than 3 to 5 % during routine operation, the following precautions are
advisable:
- very careful sample preparation to optimize thermal contacts at all interfaces,

sample and calibrant should have similar heat capacities and, if possible, similar
thermal conductivities,
- temperature intervals no longer than 100 to 200 K for a single run. Larger tem-
perature ranges should be divided into smaller intervals. For isoperibolic DSCs
the initial and end isothermals are nonlinearely connected (due to asymmetries
of radiation losses) and this tendency increases at higher temperatures. This must
be borne in mind when selecting a propriate temperature interval. Differences in
the observed heat capacities in overlapping temperature regions clearly reflect
the uncertainties of the experiments,
comparison of heating and cooling runs,
an effective program of operation. The essential zero line and calibration runs
should be carried out before and after the, perhaps multiple, sample runs under
the same conditions,
constant line voltage,
constant purge gas flow,
preheating of the calorimeter (including, if possible, the sample) to at least 10 to
15 K above the maximum temperature ofthe following measurement in order to
stabilize the system, optimize the heat transfer conditions, and remove adsorbed
moisture or retained solvent,
for low temperature operation, prevention of any condensation of water from the
atmosphere or of volatile components of the sample on the cold surroundings of
the measuring system,
constant room temperature (for precise measurements),
- in the case of isoperibolic DSC, very good temperature control ofthe isoperibo-
lic surroundings, which influences the long-time stability of the recorded signal
to a great extent. If, for instance, the DSC has been heated to a high final tem-
perature during the zero line run and the stabilization of the temperature of the
sample holder surroundings has not been sufficient, the following sample run
would give a different isothermal. The same happens if the experiment is started
before the isothermal baseline has reached a steady state (cf. Sect. 6.2.1.1).
Detailed considerations can be found in papers of Richardson, 1989; Suzuki,

Wunderlich, 1984; PoeBnecker, 1990, and Flynn, 1993.
6.2.1.7 Typical Applications of Heat Capacity Measurements
Calculation of the reaction enthalpies from the measured heat capacities of

reactants and products using Kirchhoff's law:
llrH (Tz) = llrH (T1) + JL (v;' em.; (T» dT

;
~H(TI)' ~(T2) molar reaction enthalpies at TI und T2; Cm.i(T) molar heat
capacitiy of the i-th component; Vi stoichiometric factor of the i-th component
(positive for products, negative for reactants).
Refined models for the description of the melting of polymers over broad
(> 100 K) ranges of temperatures require heat capacities of the amorphoUs and
crystalline fraction in this region.
Characterization of the glass transition process (cf. sect. 6.2.3) requires heat
capacity functions for the non-equilibrium glassy amorphous and the equi-
librium liquid state. Most such applications need only the change of the heat
capacity with temperature. Thus for commonly used polynomials of the type
C(T)=ao + a l T+ a2 T2 + ... or C(T)=ao + al T+ a2 T-2, ao is not
needed. It is generally found that a linear approximation (all a j > al are zero) is
sufficient.
Very important from the theoretical point of view is, on the one hand, the com-
parison of theoretical and measured heat capacities and, on the other, the calcu-
lation of the thermodynamic potential functions S(T), H(T), F(T) and G (T).
For polymers numerous reliable experimental heat capacity results are available
and these have been collected in a data bank (Wunderlich, 1990). The majority
ofthese are DSC measurements and generally refer to relatively high temperatu-
res (>200 K). Values at lower temperatures (as close as possible to 0 K) must be
measured using high-performance adiabatic calorimeters.
The main microscopic cause of the measured macroscopic heat capacity of
solids is the vibrational motion. It is thus possible to calculate the heat capacity
of solid polymers if the vibrational spectra are known. According to the
procedure of Wunderlich and co-workers (Lau et aI., 1984; Wunderlich, 1990;
Roles, Wunderlich, 1993) the spectra may be approximately separated into
group and skeletal contributions. The former are obtained from infrared and
Raman spectra. The latter can be approximated by Tarasov-functions, the
characteristic parameters for which are obtained by fitting the experimental
heat capacities at low temperatures, when the group vibrations are not yet
excited.
The result of the calculation is C y, the molar heat capacity at constant volume.
To compare with the experimental Cp a relation between both quantities is
needed. The thermodynamically exact conversion
cannot often be used because neither the thermal expansivity r nor the com-
pressibility X are known. A good approximation is to use a modified Nemst-
Lindemann equation (Roles, Wunderlich, 1993).
Cp - Cy= 3R· Ao· C;· T/(Cy· Tll..)

Tg" equilibrium fusion temperature; Ao empirical constant (for many polymers:
Ao = 3.910-3 KmolJ- I ).
When the heat capacity functions CiT) are known from as near as possible
to 0 K, the thermodynamic functions H(T), S(T), F(T) and G(T) can be cal-
culated as nonnal:
T
T C (T)
H(T) =Ho + I Cp(T) dT S(n=so+ I -p- dT
o o T
F(T) = U(T)- T· S(T) G(T) =H(T)- T· S(T)
Low temperature DSC measurements of Cp are not possible and for this tech-
nique only changes in these functions can be calculated.
6.2.1.8 Heat Capacity Measurements with Temperature Modulated DSC

Recently, "modulated" ("oscillating", "dynamic") DSCs has become available
(Reading et aI., 1994). In this type of calorimeter, the common DSC principle has
been modified to the extent that a (small) sinusoidal (or other periodic) tenn is
added to the linear temperature-time function:
T(t) = To + {3' t+ ATsinmt
It has been shown (Wunderlich et aI., 1994) that this type of DSC allows the heat
capacity of the sample to be detennined in one run. The easiest way is to deter-
mine it in the isothennal mode from the differential amplitude Ad (from sample and
reference sensors), the amplitude AT' the angular frequency m, the thennal con-
ductance L (between sample sensor and sample) and the heat capacity of the
empty crucible Co:
C s = Ad/AT v(L/m)2 + q
In principle it is also possible to detennine the heat capacity of the sample in the
same way from the non-isothermal, modulated scanning run, but in this case a
horizontal zero line must be guaranteed. Care must be taken that the proper cali-
bration is carried out, i.e. the detennination of the effective thennal conductance
L (in W K-I). With this proviso, heat capacities can be measured with this type
of calorimeter with an error of 1% and better. This is one of the reasons why
modulated (oscillating, dynamic) DSC is of considerable interest at present.
Another theoretical approach proceeds from the definition of a complex heat
capacity function, which exist in the case of time dependent processes (e. g. the
glass transition) and can be measured with the aid of temperature modulated DSC.
As this approach is rather new, and a brief explanation is difficult, the reader is
referred to the original literature (Jung et aI., 1992; Schawe, 1995).
6.2.2 The Investigation of Reactions
The results of such investigations are the (temperature dependent) heat of reaction
and the kinetics of the reaction. A definition of the true reaction mechanism (not
just a fonnal mathematical fitting), the so called kinetic analysis, is a far more
difficult problem. Any kinetic analysis demands a knowledge of the conversion
rates as function of the progress of reaction. This is expressed as the extent of reac-
tion .; or the degree of reaction a. Whatever the aim, interpretation assumes that
both the initial and the final state of the reaction are unambiguous and well-defined.
Without this information results will be, at best, semi-quantitative.
6.2.2.1 Determination of Heats of Reaction
The aim is to determine a thermodynamically well defined reaction enthalpy. If a

subsequent kinetic analysis is planned, the calculation of a similarly well defined
conversion-time curve is of equal importance. Because of the limited precision of
DSCs minor effects due, for example, to changes in pressure, stress or surface con-
tributions may be neglected. The following simplified equation is then obtained for
the measured heat flow rates in a DSC :
Q
(d ) = <Pm = C (n. dT + (OH)
dt p p,~ dt oJ!
~ ~p
dt
Cp , ~ (n is the heat capacity of the system at constant pressure and at constant extent
of reaction. The partial molar reaction enthalpy ~rH = (oHI O';)T,p is nearly always
replaced by the average reaction enthalpy (~rH), which is independent of the
instantaneous composition of the system. The Kirchhoff equation
describes the relation between the change ofthe heat capacity of the reacting system
and the temperature-dependent reaction enthalpy. Here the Vi are the stoichiometric
numbers and Cp,i the partial molar heat capacities of reactants and products.
Reactions may be carried out isothermally or non-isothermally (in scanning
mode or following a special temperature programm). Advantages and disadvan-
tages are discussed in detail in Sect 6.2.2.2.
Reactions in the Isothermal Mode

Here, the sample is heated as rapidly as possible from a temperature at which the
system is inert (often room temperature) to the reaction temperature. The "drop in"
technique (cf. Appendix 2) may also be used - the sample is dropped into the
preheated calorimeter. The end of the reaction is indicated by a constant heat flow
rate <Pend' Extrapolation of this <Pend to the start yields the baseline of the reaction
peak. As there is no heat capacity contribution in the isothermal mode, the relation
between the heat produced or consumed and the average reaction enthalpy is very
simple:
~
Qm (t) = J(<Pm - J
<Pend) dt = (~rH) d';
o 0
The only uncertainty is due to possible drift in the signal. The caloric error for a
typical sample mass of 10 mg and a heat production of 360 J g-l is less than 1%, if
drift during a one hour reaction is less than 10 ~W. For long-term reactions drift
must therefore be comparable with the short time noise (cf. sect. 7.2).
If a reaction is investigated at several temperatures (essential for a kinetic an-
alysis), the temperature dependence of the reaction heat is also obtained. Alter-
natively, if the Cp(T) functions of reactants and products are known, Il r H(T) may
be calculated by means of the Kirchhoff equation.
The main problem with isothermal measurements results from the ill-defined
behaviour of the signal following the initial introduction of the sample. Even for
DSCs with very short time constants this time amounts to at least 10 to 15 s. Most
of this error can, however, be eliminated by subtracting the measured curve of a
second run with the same (now reacted) sample and exactly the same conditions.
Nevertheless, minor differences may remain because of the different thermal con-
ductivities of products and reactants and because heat transfer conditions may
change during the reaction.
In order to check the completeness ofthe reaction, the sample is often heated to
higher temperatures. However, this is only useful if side reactions or shifts of
chemical equilibria at higher temperatures can be excluded. If additional reaction
has been detected, the clear assignment of a reaction heat to a definite temperature
may be lost.
Reactions in the Scanning Mode
In principle, the correct evaluation corresponds to that already discussed in Sect. 5.3.
However, the situation is more difficult, because it cannot be assumed that reactants
and products have almost constant heat capacities owing to the large temperature
interval of the reaction (l 00 K and more). In contrast to phase transitions, which
follow a zeroth order reaction law, chemical reactions do not terminate at, or
shortly beyond, the reaction peak maximum. To complicate matters, the reaction
products may have a glass transition within the reaction region. In addition, the
glass temperature of the original reactants may shift into this temperature range. As
a typical example, the curing of an epoxide resin is shown in Fig. 6.13 (Richardson,
1989). The reaction starts immediately after the glass transition of the reactants at
A and is finished at B. (In the following it is always assumed that the zero line
has no curvature or has been subtracted and that changes in heat transfer conditions
can be neglected). Reaction enthalpies may be determined in the following ways (in
the first case only the heat flow rate curve has to be measured, all other methods
require Cp measurements).
Method I: A linear baseline is drawn between the points A and B (Fig. 6.13). This
simple and widely used procedure is only an approximation. It is better the smaller
the heat capacity difference between reactants and products. The validity of this
assumption can be decided very easily by comparing the original run with a rerun
on the reacted material. Both curves should then fit at A and B. As can be seen from
Fig. 6.14 b, this is not true for this system, even after allowance for the glass
transition.
2.5
J
K·g
t 10
A
B
'-'
c
8" 1.5
~
c
OJ
..c::.
-'-'
' -'
~ 1.0
VI
300 350 400 450 K 500

temperature ..
Fig. 6.13. Curing of an epoxide resin (acc. to Richardson, 1989) and the procedure mostly
used to determine the heat of reaction. The linear baseline between start (A) and end (B) ofthe
reaction ignores all possible changes in heat capacity
Method 2: If the temperature dependent change of the heat capacity Cp.~(T) of

the reacting system cannot be neglected, but the heat capacities of reactants and
products are known over the relevant temperature range (and there are no glass tran-
sitions), then the baseline can be calculated by iterations similar to those of case 3
in Sect. 5.3. Method 1 above gives the zeroth approximation to the extent of reac-
tion function ~ (T). The heat capacities of the products can easily be obtained from
the rerun on the fully reacted sample. If the first run ofthe reaction mixture can be
started at least 30 K before the reaction starts, the heat capacity ofthe reactants may
be determined from this interval. Extrapolation into the reaction temperature range
is generally accurate enough. Otherwise, the heat capacities of all components of
the reaction mixture must be measured separately and the Cp function calculated
using simple mixing rules.
After subtraction ofthe baseline constructed in this way from the measured cur-
ve, an average heat of reaction is obtained (the term reaction enthalpy should be
avoided, the quantity has no thermodynamic significance) for the temperature ran-
ge between T\ and T2 • Improvements relative to method 1 are only found if the heat
capacity functions are known with sufficient precision (uncertainties < 1%).
Obvious improvements in the definition of the baseline should result from the use
of "modulated" ("oscillating", "dynamic") DSC (cf. Sect 6.2.1.8). By this method
it is possible to get both the heat flow rate curve (which is proportional to the actual
2.5
J/(Kg)
t
2.0 B
~
'"'
C
1.5
Q.
c
-'"'
c
cu
.r::.
-'"'
'"'
cu
Q. 1.0
VI
300 350 400 450 K 500

a tempera ture ~
2.5
J/(Kg)
r
2.0 B
->-
'"'
C
1.5
Q.
c
-'"'
c
cu
.r::.
~
'"'
cu
Q.
VI
1.0
300 350 400 450 K 500

b tempera ture ~
Fig.6.14. The thermodynamically correct determination of the specific heat of reaction (ace.
to Richardson, 1989); for details see text.
a 1st nm curve,
b I st and 2nd nm curves.
Hatched areas extended to cp = 0; Ilh (350 K) = X - Y = hproiuda (350 K) - II.- (350 K)
=-44.7Jg-I
reaction rate) and the heat capacity function of the reacting mixture from the same
run.
The extent of reaction function ~(T) can be obtained from the ratio between par-
tial heats (up to a certain temperature) and the overall heat of the reaction.
Method 3: Many of the above problems and difficulties can be avoided by integra-
tion of the Cp curves (Richardson, 1989; 1992b; Flynn, 1993). The specific reaction
enthalpy !l.rh (T) at the starting temperature T) (e. g. 350 K) may be determined in
a thermodynamically correct way even if a glass transition occurs within the tem-
perature range of the reaction. From the procedure of Fig. 6.14 it follows that
The reaction enthalpy at temperature T) (i.e. !l.rh(T)) is obtained as difference

between the areas X and Y. The baseline is not necessary in this case, because only
differences are needed:
!l.rh (T) = X - Y = «hI (P, 490 K) - hI (R, 350 K» - (hI (P, 490 K) - hI (P, 350 K»
The enthalpy subscript l represents the liquid state of the reactants R and the liquid-
like or rubbery state of the products P. The difference hl(p' 490 K) - hl(R, 350 K)
corresponds to the experimental quantity, the area X, defined by the lines at T), T2 ,
cp = 0 and the reaction curve. The horizontal hatched area Y follows from the rerun
on the reacted sample.
490K
Y = hI (p, 490 K) - hI (p, 350 K) = J Cp,l (P) dT

350K
If the glass transition ofthe products is above T), as is the case in Fig. 6.14, Cp,l
must be extrapolated as shown. Without this extrapolation another result would be
obtained:
The procedure can also be extended to the determination of reaction enthalpies

at any temperature T between T) and T2 (Richardson, 1992b; cf. also Sect. 5.3).
At first a possible glass transition is disregarded. One obtains:
!l.rh (T) = h (P, T) - h (R, T)

= (h (P, 12) - h (R, 7;» - (h (P, 12) - h (P, T»
- (h (R, T) - h (R, 7;»
or:
T2 T
!l.rh (T) = X - Jcp(P) dT - Jcp(R) dT= X - Z

T T)
This procedure is schematically shown in Fig. 6.15. The areas X and Z are hatched
horizontally and vertically respectively. A simple rearrangement of the last equation
gives:
T2 T
tlrh(D=X- f cp(P)dT+ f (cp(P)-cp(R»dT

TI TI
T
= tlrh e1;) + f (cp(P) - cp(R» dT

TI
It is clear from Fig. 6.15 that this procedure is a direct application of the Kirchhoff
equation. An enthalpy-temperature diagram (Fig. 6.16) is especially clear and
instructive. Here, for the sake of simplicity, temperature independent heat capacities
are assumed. The enthalpies are then linear functions of the temperature. The
diagram also recognizes that both reactants and products may be in the glassy
(curves Hg(R) and Hg(P» or liquid (curves H1(R) and H1(P» state. A reaction
usually proceeds at a measurable rate only when the reactants are in the liquid
state (above glass transition Tg(R». Further, the glass transition of the products
(Tg (P» is very often somewhere between 1) and 12. A formal (Flynn, 1993) and
thermodynamically correct procedure yields four different reaction enthalpies and
four different ~ values at each temperature (see Fig. 6.16):
glassy reactants ~ glassy products tlrH = bd, ~ = bclbd
glassy reactants ~ liquid (rubbery) products tlrH = be, ~ = bclbe
liquid reactants ~ glassy products tlrH = ad, ~ = ac/ad
liquid reactants ~ liquid (rubbery) products tlrH = ae, ~ = ac/ae
Only the final case has any meaning for kinetic studies.
l\, H(T) = X-Z
t- r-/
(p (RI (p (PI
-
....
'"
Cl.
"1\
-'"'"....
LJ
Z
.J::
-........
.~
Cl. ~ X
III
Fig.6.15. The thermodynamically

r, T correct determination of a reaction
enthalpy (acc. to Richardson, I 992b);
tempera ture for details see text
f I
I B
>-
c. I
c::I I
..r:::.
I
c
ClJ I d c
Tg (R) T Tg (P)
temperature ..
Fig.6.16. The determination of reaction enthalpies from an enthalpy-temperature diagram;
for details see text.
A: integral of Cp(T) for the reactants, B: integral of Cp(T) for the reaction mixture, C: inte-
gral of Cp(T) for the products
Rarely, e. g. where there is a possibility of side or decomposition reactions, the

heating should be stopped at a temperature at which such disturbing reactions can
still be excluded. The reaction should then be completed isothermally at this
temperature. Baselines are needed for both the scanning and isothermal parts of
the reaction (Richardson, 1989). However, it may be more convenient to carry out
all the reaction in the isothermal mode.
6.2.2.2 Kinetic Investigations
6.2.2.2.1 Introduction and Definitions

As mentioned in Sect. 4.4, the assignment of the time dependent heat flow rates to
defined reactions lead to kinetic data. The true aim of a kinetic analysis is an eva-
luation of the correct reaction mechanism: this is fully defined, if the sequence of
all elementary steps is known as are the activation parameters for each of these.
A large body of experimental data has then been reduced to a few equations which
describe the relevant process, so reliable predictions are possible. For example, the
influence of variations in the activation conditions can be described together with
the effects of changes in the composition of the reacting systems. Much time and
effort is required to reach this points and supplementary investigations are essen-
tial. If the evaluation of a given system is limited only to a mathematical descrip-
tion of the reaction by formal kinetic parameters (overall rate equation, overall
activation parameters), less effort is clearly required. However, the possibility of

predicting reaction rates for systems with another composition is then much
restricted or almost impossible.
Every quantitative kinetic analysis starts with the determination of an unbroken
sequence of concentration-time data. Using DSC gives a direct measurement of the
rate of conversion versus time. The rate of reaction law for the overall or the
elementary reaction may then be formulated as usual:
1 d~ 1 dCI
r=-·-=- ·_=f(CA,CB, ... ,p,T)
V dt Vi dt
At constant temperature and constant pressure and postulating a simple mechanism

one obtains:
1 d~ 1 de;
r = - . - = - . - = k (T) . C~A • C~B ....
V dt VI dt
k(T) is the rate constant; the exponents nA' nB' ... are the partial orders of the
respective reactants in the rate law. They are generally not identical with the
stoichiometric coefficients. Their sum is equal to the overall reaction order n.
As a rule the experiment does not provide the concentrations of the respective com-
ponents themselves, but rather quantities (or changes of these quantities with res-
pect to time) which are proportional to the concentrations or to the concentration
changes. With DSCs, the change in concentration of a component at time t may be
obtained from the measured heat flow rate, but only unambignously, if
the heat of reaction is independent of the extent of reaction,
the overall reaction is an elementary reaction, i. e. only one heat-producing
reaction exists,
the initial and final states of the reaction are known.
It is clear from these restrictions that, .in general, a kinetic evaluation of DSC
measurements is only meaningful if supplemented by results from other analytical
methods (e. g. from IR, uv, NMR, MS, GC, HPLC etc.).
Using thermoanalytical methods for the investigation of kinetic problems, the
concentration of the reactants are frequently replaced by the degrees of reaction
a = t;I ~max , in particular if reactions in heterogeneous systems are investigated. For
simple n-th order reactions in the case of equimolar mixtures one obtains:
da
- = k (T) . c~ . (1 - a)ft
dt
For any other mechanism a general conversion functionf(a), dependent on the
special type of the homogeneous or heterogeneous reaction, is introduced:
da
-=k(T)·f(a)
dt
diddt is measured in units of reciprocal time, a ranges from 0 to 1.
Using the DSC in the scanning mode at a constant rate (J3 = dTldt = const), the
term daJdt will be replaced by the term f3 daJdT (minor self-heating of the sample
during the exothermic reaction is neglected).
Usually the temperature dependence of the rate constant k (T) is describ~d by the
empirical Arrhenius equation or by the Eyring equation, which follows from the
activated complex theory:
Arrhenius: k (T) = A . e-EAIRT
Eyring: kB'T
k(T) = - - . eM IR. * *
e-i'JIIRT
h
kB is the Boltzmann constant, h the Planck constant.
The Arrhenius parameters (A preexponential factor, EA activation energy) and the
Eyring parameters (~S* activation entropy, ~H* activation enthalpy) are related to
one another by (for condensed systems):
A·h
~H* =EA -R' T ~S*=R ·In - - -
e' k B • T
The Eyring activation parameters are more suitable for the understanding of rela-
tions between the structure of the reactants and the reactivity (Heublein et aI., 1984).
DSC methods are widely used to solve kinetic problems because ofthe simple and
fast sample preparation and the wide range of experimental conditions - much
inform!ltion is produced in a short time. The technique, by contrast with many other
methods, immediately gives a series of "reaction rates" as a function of the degree
of reaction a. The methodology is quite general and it is immaterial whether the
reactions investigated come from inorganic, organic or macromolecular chemistry.
Continuing advances in instrumentation and data treatment facilitate refined calcu-
lations. However, these are only meaningful after consideration of specific aspects
due to sample preparation, some peculiarities of the DSC method, data sampling
and the processing of the raw data.
6.2.2.2.2 Experimental Preconditions for a Reliable Kinetic Analysis

1. Sample Requirements
After the introduction of the sample into the DSC the amount of reaction
should be negligible prior to the attainment of a stable steady state. If this is
not the case it must be determined separately and allowance made for this
effect.
The reaction mixture must not react with the material of the crucibles nor
should there be any catalytic influence.
Samples with an appreciable vapour pressure must be loaded into special seal-
able crucibles. Errors due to the effect of pressure on the measured C are
usually insignificant. Much larger errors can occur because of changes in the
concentration of a volatile component, for example a catalyst, used in very
low concentrations, could be partially in the gaseous phase outside the reac-
tion mixture. The change in the concentration in the reaction mixture can be
allowed for to a sufficient approximation, if the vapor pressure of the volatile
component and the volume ofthe gaseous phase are known.
Multiple measurements should always be made (with the same operation
parameters) to check on the experimental repeatability. In addition some spot
checks on other independently prepared reaction mixtures of the same com-
position should be carried out in order to exclude accidental errors during
sample preparation. If the mixture is reactive at room temperature, loaded
crucibles should be stored in liquid nitrogen.
2. Features Peculiar to DSC Methods
The reaction mixture in the small and sealed crucibles of a DSC cannot be
stirred. The results can, therefore, only be partially interpreted if concentra-
tion gradients occur during the reaction, or if there are transport processes to
the gaseous phase (e.g. the decomposition of hydrates).
It is not possible to add solid or liquid components to the reaction mixture
after closing the crucibles.
To apply these results to technical processes (e.g. to predict behavior in a
reactor) the very different conditions must be recognized and, in particular,
allowance made for heat transfer effects.
3. Data Aquisition and Processing
If one needs in addition to the a{t) function also that of the rate of conver-
sion da{t)/dt for a kinetic analysis, the baseline must be determined first
(according to the method 1 or 2 in Sect. 6.2.2.1) and then subtracted from the
measured curve. It is also possible to proceed from method 3 in that section.
In every case the measured reaction heat Qrn must be checked and, if neces-
sary, corrected with respect to the final degree of conversion actually reached.
This correction must be made whatever the reason for the cessation of reac-
tion - perhaps caused by a great increase in viscosity (vitrification) or coming
to an equilibrium state. All subsequent kinetic analyses are incorrect if this
modification is omitted. Otherwise the calculated a{t) is not related to the
thermodynamic degree of conversion but to an apparent final state that is
specific to the individual experiment. If the correct heat of reaction Q_ can be
related to only one heat-producing reaction, the resultant overall a{t) curve is
the true normalized degree of reaction function, independent of any model
considerations. Similarly the normalized rate of conversion follows from the
current heat flow rate:
Q, . da <P{t)
a(t)=- and a{t)=-=-
Q- dt Q_
In general, however, several r~actions contribute to the measured heat flow
rate. An evaluation is then possible only with respect to a certain model. Ifthe
overall reaction is made up of, say, n reactions, all contribute to the measured
heat flow rate and their reaction heats can be treated as parameters and deter-
mined as such by variation procedures. It is obviously much better to deter-
mine as much supplementary information as possible (for the optimum case

of n - 1) for the individual heats of reaction by changes in the DSC conditions
or other measurements, e. g. temperature dependent spectroscopic investiga-
tions of equilibrium states etc.
The experimental curve has to be desmeared (Sect. 5.4) before any evaluation
ifthere are significant changes in the heat flow rates at time periods which are
comparable with the time constants of the calorimeter.
The true sample temperature may differ from that measured by the thermo-
meter which is outside the sample, because of the limited thermal conduc-
tivity of the sample and its surroundings. This is especially true for larger
sample weights and higher reaction rates. Strictly speaking, even "isothermal"
reactions proceed non-isothermally. The true sample temperature can be cal-
culated ifthe instantaneous heat generated by the reaction is related to the heat
exchanged with the calorimeter (this is limited by the overall thermal relaxa-
tion time constant of sample and calorimeter). For example, with very fast
light-activated reactions there are sometimes heat generation rates of 100 mW
and more leading to temperature corrections greater than 10K even for
calorimeters with small time constants.
Assuming that absolute and random errors for the measured signal are inde-
pendent of time, the relative errors of the measured reaction rates increase at
the end of the reaction. Therefore, efficient software should weight the data
appropriately.
Correct calibration of the DSC is essential - the temperature scale, the heat
flow and the scanning rates.
Changes of the sample volume during the reaction (e.g. up to 8% decrease in
volume during epoxide amine curing reactions are possible) have to be con-
sidered, irrespective of whether subsequent calculations are made as functions
of concentration or degree of reaction.
In much of the literature it is unfortunately not clear whether the above aspects have
been considered. For example, a kinetic analysis of non-isothermal crystallization
may have serious flaws if published heat flow rate curves are without proper base-
lines. Again, if high heating or cooling rates (up to 40 Kmin- I ) are used, leading to
large changes in heat flow rates, it is rare to find that curves have been desmeared
prior to analysis.
6.2.2.2.3 Isothermal or Non-isothermal Reaction Mode?

Classical techniques of chemical kinetics normally operate in the (quasi-) iso-
thermal mode. By contrast, it is very convenient to make kinetic studies by DSC
using the scanning mode. Both modes are possible in DSC and have particular
advantages and disadvantages (Flammersheim, 1988). They complement each
other and should not be regarded as alternatives.
Advantages and disadvantages of the isothermal mode.
A simple and immediate interpretation of the measured curve is possible because
of the complete decoupling of the two variables, temperature and time, if the
lOO
Wig
2.25
--
f T = 405 K
~
CI all sample masses about 15 mg
'-
3:
0
1.50 ra tio of DGEBA to aniline
-
;;::
CI
1: 1: 1
QJ
.c::. 2: 1:2
~
<.J
3: 1:3
<.J
QJ
Cl.
0.75 4: 1:4
II>
5: 2:1
0.00 L -_ _ _....L...-_ _ _--l..._ _ _- ' ' - - - -_ _ _- L -_ _ _---l
o 17 34 51 68 min 85
time ..
Fig. 6.17. Isothermal reaction curves for the polyaddition of aniline and bisphenol-A-digly-
cidyl ether (DGEBA) in mixtures with different stoichiometry
minor changes of the sample temperature during the reaction can be neglected.
This method is especially useful when searching for the most probable reaction
mechanism. Two examples will illustrate this:
During the polyaddition reaction of aniline and the diglycidylether ofbisphenol-
A (DGEBA) the maximum conversion rate is found at non-zero degrees of reac-
tion (Fig. 6.17). This can only be due to autocatalytic and/or consecutive reac-
tions. Furthermore it is obvious that not only the ratio of the concentrations of
both reactants has substantial consequences but also that equal excess concen-
trations of amine or epoxide differ greatly in their effects. This is a very impor-
tant pointer to the possible reaction mechanism.
The second example (Fig. 6.18) is the polyaddition reaction of a mixture of the
two isomers 2,2,4-trirnethylhexamethylene diisocyanate and 2,4,4-trimethyl-
hexamethylene diisocyanate (TMDI) with 3,6-dioxaoctane-1,8-dithiol (TGDT)
in presence of varying concentrations of a catalyst. As expected, the initial
conversion rate is proportional to the catalyst. concentration but the increase in
the slope of the leading edge of the reaction peak in direct proportion to the
concentration of catalyst is unusual. This can only be explained if one assumes
a direct coupling of catalysis and autocatalysis (cf. Sect. 6.2.2.2.8).
- The baseline follows unambiguously by extrapolating the measured heat flow
rate after the complete reaction, assuming that all parameters, which can in-
fluence the measurement directly or indirectly (e. g. the temperature of the
OJ
/'
/
--- -
.....-
- - - -::..-:----:-=---=-
/ /'
/ /
Wig I /
1
/
I
I I
.2!. I I T : 373 K
....
<:>
\2 I / all sample masses about 10 mg
~
a 0.15 '-.V
J
<:> J catalyst concentra tion
'"
..r:: J 1: 1%
-....
....
C1J
Co
J
J
2: 2%
3: 4%
Vl
3J
J
I
\
\..1
0
0 40 80 120 min 260
time ...
Fig.6.1S. Isothermal reaction curves for a stoichiometric system of trimethylhexamethylene
diisocyanate and 3,6-dioxaoctane-l,8-dithiol with pyridine as a catalyst. The change in the
reaction rate on the leading edge of the peaks shows direct coupling of catalysis and auto-
catalysis
surroundings) are sufficiently constant. Demands on the apparatus can be con-

siderable, depending on the reaction time in question. Above all the long-
time drift of the signal should not be greater than the short-time noise. The
standard controls offered by the manufacturers of commercial DSCs are not
generally sufficient. For instance, for a Perkin-Elmer DSC it is necessary to con-
trol the temperature of the isoperibolic block to better than ± 0.01 K (by adding
of a controlled Peltier cooling device or a cryostat with circulation of the coolant)
and stabilize the room temperature to ± 0.2 K, for to come to optimum condi-
tions, i. e. to a caloric error below 1%.
The reaction can be allowed to take place at low temperatures such that decom-
position or side reactions can be avoided.
The main disadvantage, already mentioned in Sect. 6.2.2.1, is that the initial
phase ofthe reaction cannot be measured precisely, as the steady state conditions
are disturbed on introduction of the sample. The solution given, using the
difference between the reaction curve and the fully reacted rerun, for kinetic
analysis, is only partially successful. A better solution is to always choose
sufficiently low reaction temperatures so that unavoidable errors are minimized
to values which do not influence model calculations. In addition, smearing of
the heat flow rate data and deviations of the sample temperature from that of
the isothermal surroundings are so small that the experimental data need not be
corrected.
Isothermal measurements need more time than scanning runs, because reaction
isotherms must be recorded at a minimum of five different temperatures that are
as widely separated as possible.
Advantages and disadvantages of non-isothermal measurements.

Scanning measurements need less time than isothermal experiments but con-
struction of the proper baseline may be a problem.
The measurement can be started at temperatures well below that ofthe beginning
of the reaction, so steady state conditions of the DSC are ensured. Unfortunate-
ly, reactions may go to completion, even at low heating rates, at such high tem-
peratures that secondary reactions cannot be neglected (this is especially relevant
for organic reactions).
The kinetic parameters can, in principle, be obtained from a single measurement,
if the reaction mechanism is known. The experimental data cover such a wide
temperature/concentration/time regime, that reliable results concerning the
mechanism are possible. Unfortunately it is not clear from the shape of the cur-
ve what type of reaction is involved and this is a drawback when searching for
the most probable reaction mechanism.
6.2.2.2.4 Thermal Activation of the Sample or Activation

by High Energy Radiation
Reactions that are studied in the DSC are normally thermally activated. The wide
range of experimental conditions (scanning rates, temperature range, possibility of
more complicated procedures with intermediate annealing phases) is responsible
for the extensive use of this technique.
Many reactions, however, can also be started by radiation with sufficiently high
energy. The investigation of such reactions in a "Photo-DSC" (cf. Sect. 2.2) has con-
siderable value for the optimization of process control parameters and this is espe-
cially true for those polyaddition and polymerization reactions that give materials
with widespread uses.
A typical example of a light-activated measurement is shown in Fig. 6.19. It is
often found, even in the case of absolutely symmetric radiation conditions on both
the sample and the reference side, that the baselines change during the dark and the
illumination phase, as in the figure. Simple subtraction of a second curve measured
with the same procedure but with the reacted substance removes this effect. Light-
initiated reactions, using conditions in the DSC which come close to those used in
practice, are always extremely fast. The apparatus (or Green's) function, which is
needed for the subsequent desmearing (deconvolution, see Sect. 5.4), is obtained
easily and is sufficiently precise if, ~fter the completed reaction, the response of the
sample to a light flash is recorded (Flammersheim et aI., 1991). If it were not for the
fact that the heat produced during the chemical reaction is evolved inside the sam-
ple, but during the light flash it is mostly a surface effect, this response would be an
ideal apparatus function for the sample properties in question. The dashed line in
300
jase line of the reaction mixture with the absence of UV radiation
mW 2nd run
.------~.--- .
base line of the reacted sample

with the presence of UV radiation
150
'"
<1J
..c:
75
o ~------~--------~--------~--------~--------~
o 0.7 1.4 2.1 2.8 min 3.5
time ..
Fig. 6.19. 1st and 2nd run for the photopolymerization of a multifunctional acrylate at room
temperature.
(photoinitiator: 1 % 2,2-dimethoxy-2-phenylacetophenone (benzyldimethylketal), light inten-
sity: 5 mW cm-2 at 365 nm)
Fig. 6.20 corresponds to the deconvoluted curve. The solid curve is the original
measured curve after subtraction of the baseline (2nd run) from Fig. 6.19.
The kinetic evaluation of light-activated reactions has so far only been par-
tially succesful because the true reaction course is overshadowed by additional in-
fluences:
Irradiation of the initiator causes a time dependence of the concentration of the
initiating radicals but this effect can be easily corrected if the kinetics of the
decomposition are known.
Samples have finite thickness (0.1 to 0.4 mm). It follows from the Beer-Lambert
law that there is a varying light absorption and hence a gradient of the reaction
rates within the sample. However, the calorimeter always records an overall heat
flow rate. As result of this rate gradient there is a corresponding concentration
gradient of the reactive component. An exact solution of the subsequent system
of differential equations is only possible for drastically simplified boundary and
initial conditions (Shultz, 1984).
The situation may be made more difficult by the increasing viscosity of the mix-
ture during the reaction, which nearly always is carried out at room temperature.
As a result, the reaction may be incomplete (Klemm et aI., 1985). For example,
at room temperature phenyl glycidyl ether (pGE) reacts completely during catio-
nic polymerization induced by irradiation of diaryliodonium salts, whereas the
500
-----
mW
I
~
o
250 I
'- I
o
3: I
;:: I
I
o I
QJ
..r::::
125 I
"
"
"
"~
o
o 0.5 10 15 min 2.0
time .-
Fig.6.20. Baseline corrected experimental (solid line) and desmeared (dashed line) curves
for the photoinduced reaction of Fig. 6.19
reaction ofDGEBA stops at about 50 to 55% conversion (Klemm et aI., 1985).

In this case, a kinetic analysis is impossible and even the estimation of the final
degree of reaction is difficult because of missing or insufficiently reliable expe-
rimental information from other analytical methods. To solve the problem, it was
supposed in this special case, that the same heat is produced during the reaction
of the oxirane ring both for PGE and DGEBA.
The experimental situation can be improved if extremely thin « 0.1 mm) samples
are used with the lowest possible concentrations of photoinitiator. If diffusion limi-
tations are small during the whole reaction, as in the case of the reaction of PGE,
one then finds a simple first order rate law with respect to the monomer.
The advantages that lead to the application of DSCs for investigation of light-
activated reactions are essentially the fast, reliable and quantitative collection of
data that reflect the effects of those parameters which influence the conversion rate:
type of monomer,
- type and concentration of solvents, photoinitiators and inhibitors,
wavelength and intensity of the radiation used.
6.2.2.2.5 Remarks on the Influence of Different Phases

Reactions in Homogeneous Systems
In this case both reactants and products are in a homogeneous (generally liquid)
phase during the whole reaction. Solvent-free systems are normally investigated,
because of the higher reaction heat per unit volume of sample. The measured heat
flow rate curves are a reliable basis for subsequent kinetic analyses. The classic
approaches of kinetics for homogeneous systems may be used without restriction.
To some extent this also holds when reactants and products are completely or
partially immiscible, the reaction is virtually irreversible and the products have no
autocatalytic influences.
Many reactions can be described within a wide range of degrees of reaction by
simple n-th order rate laws
r = k (T) . (1 - an)
or by laws, which in addition allow for catalytic and autocatalytic effects
Non-integer exponents m and n suggest complex reaction mechanisms and the

converse, integer exponents, do not automatically imply a simple mechanism.
Reactions in Heterogeneous Systems

During a chemical reaction of this type the reactants may always exist in different
phases (liquid-solid, solid-solid) or the system may change due to the development
of new phases and/or the vitrification of polymerizing components. Although open
to criticism (Flynn, 1990) (and in the absence of any obviously better approach)
reactions ofthis type are normally treated using procedures that are formally simi-
lar to those for homogeneous systems. In addition to this very fundamental problem
there are numerous additional sources of error connected with sample preparation
and measuring conditions. In particular, if the gaseous phase is important in the
reaction, there is always the risk that the analysis interprets the kinetics of transport
through the sample rather than the reaction mechanism. Heat flow rate curves are
frequently difficult to analyze without help from other analytical methods (espe-
cially temperature programmed X-ray diffraction, hot stage microscopy, thermo-
gravimetry and possibly mass spectroscopy and gas chromatography).
It is typical of heterogeneous reactions that very different events such as nuclea-
tion, growth and diffusion can occur simultaneously and successively. A review and
a critical judgement of the most frequently used rate laws is given by Brown et aI.,
1980.
6.2.2.2.6 Evaluation of Overall Rate Laws and Formal Kinetic Parameters

The primary aim of kinetic investigations is the determination of adequate rate laws.
For technological purposes an adequate description of even a complex reaction by
a mathematical model is often sufficient. Of course, reliable predictions are then
only possible within the often very narrow limits of the data sets included into the
evaluation. The physical and chemical use of such rate laws is severely restricted
and - if it has any value whatever - it must be done very carefully. An overall rate
law may give a good description over a certain range of reaction, because the rate
reacts very sensitively to those parameters which are present in the rate law. It may
be insensitive with respect to those parameters, which do not appear explicitly (e. g.
stoichiometry of the reaction, solvents, catalysts, packing efficiency for heteroge-
neous samples, flow rate and type of the purge gas). A rate law can often be found
by systematic trial and error, this is a result of possible simplifications that may
sometimes be made to the rather complicated differential equations that describe
the true reaction mechanism. In spite of this, a determination of the overall rate law
may be a useful first step in finding the reaction mechanism.
The risk of using overall rate laws for the description of the reaction is that it
may possibly hinder further advances towards the true mechanism. For instance, the
validity within a certain range of conversion may be taken as confirmation of a
postulated, but often too simple, model. Deviations over other ranges of conversion
are then interpreted within the framework of that model only. A typical example is
the following equation (Sourour, Kamal, 1976), which is often used in the polymer
chemistry literature to describe the epoxide amine polyaddition reaction. For
stoichiometric reactions it reads:
da
dt = k\ . C • (1 - a)" + k2 . am . (1 - a)"
The chemistry of the reaction is thus reduced to a simple rate law containing cata-
lytic and autocatalytic steps. c is the concentration of a catalyst, n and m are formal
reaction orders and k\ and k2 are the rate constants of the catalyzed or autocatalyzed
reaction.
With n = 2 and m = 1 this equation was originally derived by Smith, 1961, as the
most probable reaction model for the epoxide amine system on the basis of all ex-
perimental results available at that time. However, solvent-free reactions can only
be described by this equation up to conversion degrees of about 0.6, equivalent to a
degree of polymerization of only 2. There have been numerous attempts to retain
this simple rate law but any connection with the chemistry of the process is com-
pletely lost. If the detailed chemistry of the complex process is unimportant, the
value of any further corrections to this equation simply reflects the better fit that
results from any increase in the number of parameters. It is hardly surprising,
therefore, that the rate law of Sourour, Kamal, 1976, has become independent of the
original reaction and that it is used to describe quite different reactions in polymer
chemistry.
To describe epoxide amine reactions to higher degrees of reaction, n and m are
taken as fitting parameters alone, sometimes even as temperature dependent ones
(Ryan, Dutta, 1979; Chung, 1984; Keenan, 1987). The term "reaction order" is then
meaningless. The data can still only be fitted, even after such corrections, up to
degrees of reaction of about 0.9 and an additional correction is introduced: it is sug-
gested that there is a change from kinetic control to diffusion control at higher con-
versions (Barton, 1980; Huguenin, Klein, 1985). At first sight this seems an obvious
correction and a review ofthe most important literature, to 1984, has been given by
Barton, 1985. Whether the assumption of a diffusion-controlled reaction is justi-
fied, can and should be experimentally checked by choosing the experimental con-
ditions in such a way that diffusion control is extremely unlikely. For example, the
reaction can be carried out in solvents or at temperatures well above Tg but the most
convincing approach is to use low-molar mass, monofunctional model systems. In
this way it has been shown (Klee et aI., 1990) that diffusion control is not the best
way to correct for the failure of the above equation.
Three general methods have been developed for the determination of kinetic
parameters:
- The starting point is the procedure of Borchardt, Daniels, 1957, which needs
both the concentrations (in DSC, the degrees of reaction) and the corresponding
conversion rates. To evaluate simple rate laws of the form
j(a) = (I - a)n
it is especially useful to proceed from the logarithmic form:
In - (da) =lnA+n·ln(1-a)--+(n-l)·lnc
dt
EA
RT
o
Today multiple nonlinear regression software from all calorimeter manufacturers

allows one to estimate the best values for n, InA and EA by evaluation of a
single data set from one scanning experiment. Various choices are offered for
j(a) (simple rate laws, rate laws including autocatalytic steps, typical rate laws
for the kinetics of reactions in heterogeneous systems). It must always be remem-
bered that the analysis of a single curve can give results that could be misunder-
stood - a reliable decision between possible reaction models is rarely possible
(Opfermann et aI., 1991). Moreover, preconceived ideas for a certain reaction
mechanism can often be confirmed by data fitting over a particular range of the
degree of reaction.
If an integrated form of a rate law is used
a 1 t
J- da=g(a)=k(T)·Jdt
of(a) 0
only a few conversion/time data points are needed to calculate the activation
parameters. Evaluation of isothermal experiments gives no problems when the
rate law is known. A review of important cases can be found in the literature, e.g.
Hemminger, Cammenga, 1989. For scanning experiments using
J- - da = g (a) = - Je-EAIRT dT
a 1 A T
o j(a) f3 0
there are difficulties because an exponential integral cannot be solved analyt-

ically. A solution is to use integral tables as calculated, for example, by Doyle,
1961. The importance of integral procedures has decreased considerably with
developments in computer technology.
Freeman, Carroll, 1958, used the logarithmic equation for n-th order reactions in
a form based on differences between neighboring data points
R.
dT = n . A In (1- a) - EA A T
A In ( da) (1)
A graph of AIn (dal dt) versus AIn (1- a ) (for constant increments of liT) yields
the reaction order and activation energy. This analysis is very sensitive to ex-
perimental errors and this is particularly true for the procedure of Ellerstein,
1968, in which the last equation is differentiated once more.
If the evaluation shows a significant dependence of the activation parameters on the
degree of reaction, this is an unmistakable sign that the reaction mechanism has at
least two elementary steps.
6.2.2.2.7 Determination of the True Reaction Mechanism

Including the results of other analytical methods, the system of differential equa-
tions for the suggested reaction mechanism is formulated and solved. Latest deve-
lopments in numerical analysis (Opfermann et al., 1991) allow one to determine the
optimized parameters even for complicated systems of differential equations. An
analytical solution is not necessary and there is no longer any need for simplifica-
tions. The result is the correct order of reaction and optimized activation parameters
for each elementary step in the reaction mechanism.
Satisfactory agreement between calculated and experimental curves, including
the use of statistical tests, is decisive in the acceptance or rejection of a specific
reaction model. The success of such decisions is strongly influenced by suitable
experimental conditions. For scanning measurements, at least three different heating
rates should be used (Opfermann et al., 1991). If analysis assumes the validity of
the usual rate laws for the kinetics of homogeneous systems (solid state or surface
reactions are negligible), maximum information follows from careful design of
experiments. It is essential to investigate the effects of different stoichiometries,
catalyst and solvent concentrations and possibly added intermediates or reaction
products. Of course, this is often difficult because of the small volume of sample
and the low heat flow rates involved. An evaluation from only one measured curve
has no value even if there is an excellent fit of the experimental data.
Current software packages allow the systematic inspection of up to 10 to 15 assu-
med rate laws both for homogeneous and solid state reactions. Combinations of
several steps (simultaneous or consecutive reactions) are possible. However, the
most powerful procedure is to dispense with trial and error and instead formulate
the most probable system of differential equations using the input routine of the
software (Flarnmersheim et al., 1993). If the rates of reversible reaction steps are
found to be much higher than those of subsequent reactions, the former can be
replaced by chemical equilibria.
Of course, a complicated reaction mechanism can often be simplified if certain
elementary reactions have very different rates and sometimes if equilibria are in-
volved, a much simpler rate law than describes the overall reaction. In contrast to
rate laws which result from systematic trial and error, the simpler rate law obtained
in this way is based on the true reaction mechanism and may thus be reliably extra-
polated to different reaction conditions. Even here, they must be tested by spot
check measurements within the framework of the model used.
6.2.2.2.8 Selected Examples

1. Example
Photochemically produced molten cis-azobenzene converts into the trans form
above room temperature in a well-defined 1st order reaction. The activation para-
meters have been found to be In (Als-I) =27.7 andEA = 103.6kJmol-1 (Eligehausen
et aI., 1989). By contrast with the situation for temperature and caloric calibration,
there are so far no internationally recommended test reactions for kinetic investiga-
tions. The use of this well-known reaction could be a first step in this direction, to
test individual calorimeters, the sample preparation technique and the evaluation
procedure.
The conversion of cis-azobenzene is only a 1st order reaction in the liquid
phase. Before each measurement the solid substance (Tfils = 71.6 Qq must be rapid-
ly melted and quenched to the starting temperature (35 Qq. Minor reaction at this
stage has no influence on the subsequent calculation of the kinetic parameters. The
true baseline of this non-isothermal measurement should be constructed using the
heat capacities of cis- and trans-azobenzene, according to method 2 of Sect. 6.2.2.1.
However, this is not possible because the liquid cis-azobenzene cannot be super-
cooled sufficiently to measure cp(T) down to temperatures below the start of the
isomerization reaction and a straight baseline must be used. This is a good approxi-
mation because the differences between a calculated sigmoidal baseline (from
literature cp-values) and the straight line are very small. An approximate (because
of different thermal conductivities) but adequate desmearing is possible using the
crystallization peak of molten trans-azobenzene as the apparatus function. The cor-
rection is small because the isomerization is rate determining even at the highest
used heating rates. A precise temperature correction to zero heating rate values for
all heating runs is very important.
Fig. 6.21 shows the excellent agreement between experiments and calculations
using multiple-curve analysis by non-linear regression. All curves from runs at
five heating rates and at four isothermal temperatures have been included in the
calculations (that for the lowest temperature is not shown in the figure). If minor
self-heating of the sample because of the limited heat exchange with the calorimeter
is allowed for, no systematic differences are found between the single-curve and
the multi-curve analysis. In addition, the same activation parameters are found'
even at higher heating rates (40 Kmin-I). Assuming a fixed reaction order
of n = I the procedure gives activation parameters of In (Als-I) = 27.8 ± 1 and
EA = 102.8±2.7kJmol-l , in good agreement with Eligehausen et aI., 1989.
2. Example
The radical polymerization of a monofunctional monomer is a simple 1st order
reaction, if a number of ideal conditions can be assumed (reactivity of the macro-
lOa
Wig
2.25
t
III
1 K minot
2 K minot
...c 4 K minot
~
3: 1.50
0
;::: 8 K minot
~
c
III
..c 15 K minot
'-'
'-'
III
0.75
a.
Vl
0
50 71 92 113 134 ·C 155
a tempera ture •
1.00
Wig ~
350 K
r
0.75
~
::
'"
;., 0.50
~
370 K
c
<1J
..c
'-'
'u 0.25
<1J
Cl..
'"
0
0 30 60 90 min 120
b time ~
Fig. 6.21. Measured (solid lines) and calculated (dashed lines) curves for the cis-trans
isomerization of azobenzene in the liquid phase.
a in scanning mode at different rates,
b in isothermal mode at different temperatures
radicals is independent of the chain length, approximately constant radical concen-

tration during the whole reaction, bimolecular chain termination reactions). Under
these circumstances, the rate of concentration change is proportional to the con-
centration of the monomer (cM ) as follows:
-
dc
dtM = ~ro •
(r.kw:
k . )1/2
m
• Cj'2. CM
The quantity r is the yield of radicals due to the thermal homolysis ofthe initiator
(I) (e. g. 2,2' -azodiisobutyronitril (AiBN) or dibenzoylperoxide), ~ro, kk:r and Aini are
the rate constants of the chain propagation, chain termination and initiation, res-
pectively. If the thermal initiator is replaced by a photoinitiator, like benzoin methyl
ether or 2,2' -dimethoxy-2-phenyl acetophenone (benzyldimethylketal), the an-
alogous overall rate law reads :
Here r is the quantum yield of the initiator, labs is the radiant power absorbed by the
sample. In contrast to polyfunctional acrylates, for which cross-linking reaction
starts immediately (Fig. 6.19), monofunctional methacrylates (methyl-, ethyl-, pro-
pyl-, butyl-) follow the simple 1st order law up to degrees of reaction of 0.3 to 0.6
(dependent on the chemical nature during both thermal and light-induced polyme-
rization). After this stage the so-called gel or Trommsdorff-Norrish effect domina-
tes, this is recognizable by increasing reaction rates despite decreasing monomer
concentrations (MalavaSic et aI., 1986). This is also valid (Fig. 6.22) for the
6.0
mW
I
~
...
c
4.5
3.0
-
~
0
1.5
c
OJ
.r:::
..
0
0 10 20 30 min 40
time
Fig.6.22. Light-initiated (1) and thermally activated (2) polymerization ofDOMA.
(curve 1: A = 365 run, 10 = 0.35 mW cm-2, T= 298 K, 1% 2,2-dimethoxy-2-phenylacetophe-
none as photoinitiator; curve 2: T= 343 K, 1% 2,2-azodiisobutyronitril as initiator)
polymerization of (2,2-dimethyl-l,3-dioxolan-4-yl)methylmethacrylate (DOMA)

(Flammersheim, Klemm, 1985). There are no qualitative differences between cur-
ves 1 and 2. In both cases the change of reaction rate with time is small enough
during the entire reaction that desmearing is not necessary. Spectroscopic invest-
igations on the soluble reaction product verify the complete reaction of the acrylate
double bonds, whereas the dioxolane ring remains intact under these experimental
conditions. Up to a degree of reaction of 0.3, the rate law is 1st order. Knowing the
activation term (Tkjnic, )112 or (T1abs)ll2, the overall rate constant gives the ratio
kpro / ~12. During the gel effect this ratio is formally dependent on the degree of reac-
tion reached. In contrast to the nearly unhindered chain propagation reactions of the
small monomer molecules, the chain termination reactions of the growing macro-
radicals are increasingly hindered in the more and more viscous matrix. The result
is a pronounced dark reaction after switching off the light during the polymeriza-
tion (Fig. 6.23). The first run was interrupted at a = 0.35. After the dark reaction
was complete the same sample was once more radiated and the light switched off
again at a overall degree of reaction of 0.8 (curve 2). To get a better comparison the
second curve is shifted to the origin of the time scale in Fig. 6.23. As a result, the
relative extent of the dark reaction is proportional to the degree of reaction at that
20
mW
15
illumination period
o L -_ _ _ _ _ _ ~ ________ ~ ________ ~ ________ ~ ______ ~
o 0.6 1.2 1.8 2.4 min 3

time •
Fig. 6.23. Photopolymerization of DOMA with interruption of the radiation at reaction de-
grees of 0.35 and 0.80 (arrows). Both curves are consecutive runs of the same sample, but the
second curve is shifted to the same origin of the time scale.
Solid lines: measured curves, dashed lines: desmeared curves (A. = 365 nm, 10 = 3.5 mW cm-2,
T= 298 K, 1% 2,2-dimethoxy-2-phenylacetophenone as photoinitiator)
moment at which the light was interrupted. The rate of the dark reaction (missing
.activation by r labs) is described by:
CM kwr CMo
---=-_·t---
deM ~ro deMo
dt dt
If the ratio ~ 1kJ:c12 is detennined from the measured curve immediately before
the light is interrupted and, following this, the ratio lGe/~ from the dark reac-
tion (last equation) the rate constants kwr and ~ro result from the combination
of these. Though desmearing is not necessary during the illumination period
with normal light intensities, the evaluation of the dark period signal is meaning-
less without it, especially if the dark reaction is performed at low conversions
(because of the very rapid decrease of the heat flow rate due to a large IG./~
ratio). The picture shows this effect very clearly. The linear relation between
-cM/(dcM/dt) and t (see the last equation) is only found from the desmeared
signal (kw)~ is about 310 at a = 0.35 and about 11 at a = 0.80). Measurement
of this reaction in DSCs is only possible if their time constants are lower than 2
to 4 s.
3. Example
The formation of polythiolcarbamates from diisocyanates and dithiols cannot be
described by a simple rate law even if autocatalytic effects are included. All discus-
sion of the most probable reaction mechanism must account for the following ex-
perimental observations (cf. also Fig. 6.18):
The initial reaction rate is proportional to the concentration of the catalyst

(pyridine, picoline, lutidine, collidine).
The reaction includes autocatalytic stePs.
Catalysis and autocatalysis are directly coupled.
Isocyanate and dithiol are not equivalent in the stoichiometry of the reaction
mixture.
A possible partial diffusion control can be excluded under the experimental con-
ditions used.
The existence of dithioVcatalyst, diisocyanate/catalyst and product/diisocyanatel
catalyst complexes could be proved using spectroscopic methods (NMR, uv,
IR). Further, the equilibrium constants of the first two complexes could be deter-
mined.
The apparent activation energy, which follows from the temperature dependence
of the overall reaction rate, is unusually low.
Following the general recommendations given above, reaction curves for a given
system (dithiol, diisocyanate, catalyst) including variations of stoichiometry,
temperature and catalyst concentration have been made for the non-linear regres-
sion analysis. The most probable reaction mechanism is (Flammersheim et aI.,

1993):
k,
S+K .=
k_,
SK (1)
1,
I+K .=
k_,
IK (2)
k,
IK+S -+ P+K (3)
k.
IK+P+S -+ 2P+K (4)
(S dithiol, K catalyst, I isocyanate, P product, IK and SK complexes of the catalyst

with I and S).
In this scheme SK (step 1) is a nonreactive and IK (step 2) a reactive complex. Step

3 corresponds to the normal catalytic reaction between IK and the dithiol, step 4
represents the much rarer combination of catalysis (via IK) and autocatalysis
(via P). An impression ofthe quality of this model is given in Fig. 6.24, which com-
0.20
Wig
r
0.15
1: 1 DDT I 1 TMDl
2: 1 DDT I 2 TMol
OJ
... 3: 2 DDT I 1 TMol

~
c::I
~
0.10 4: 3 DDT I 1 TMol
0
;:;::
c::I
OJ
..c::.
-....
....
OJ
C.
0.05
T = 363 K
'"
0.00
o 40 80 120 min 160
time ..
Fig.6.24. Comparison between measured (solid lines) and calculated (dashed lines) reaction
isotherms of the system TMDIIODT/3,5 lutidine. For exact compositions see text
pares measured and calculated isothenns for the system TMDIIoctane-I,8-dithiole

(ODT)/3,5-lutidine (concentrations are given in the table).
Component concentrations for the reactions of Fig. 6.24
Stoichiometric ratio curve number concentrations in moll- i

ODTffMDI
ODT TMDI 3,5-lutidin
1:1 4.881 4.870 0.477

1 :2 2 3.142 6.359 0.561
2: 1 3 6.834 3.225 0.361
3:1 4 7.709 2.343 0.439
The calculated activation parameters for this system are

EA = 43.5 ±2 kJ mol-I and In (All mol-I S-I) = 8.27 ±0.25 for step 3
EA = 44.7±3 kJmol- 1 and In (A1l2 mol-2 s- l ) = 7.94±0.45 for step 4.
The enthalpies and entropies of reaction are
d,H=-10.5 ± I kJmol- 1 and ~rS=-35 ±3 J K-I mol-I for SK
d,H=-32±3kJmol-1 andd,S=-94±1l JK-I mol-I forIK.
The already mentioned very low apparent activation energy of the overall reaction
follows from the coupling of the exothennic fonnation of IK with an reaction
energy of -32 kJ mol-I during the reversible step 2 and the consecutive reactions
which have normal activation energies of about 45 kJ mol-I.
The formulation of step 4 as triple impact reaction is extremely unlikely and it
should be replaced by the two following steps:
Ie,
IK + P ;:= PIK (5)
k_,
PIK+S '"
~ 2P+K (6)
The fit to the experimental curves is just as good as without this extension, cor-
relation coefficients are generally higher than 0.998. However, it is impossible to
calculate kinetic parameters for the coupled reactions (5) and (6).
This demonstrates a typical situation in calculating optimized model parameters
for more complicated reaction mechanisms. The number of model parameters is
very often so large that although an excellent fitting of the experimental curves
is possible, the parameters in question (rate constants, equilibrium constants) are
rather uncertain. In general the reasons for this are experimental (noise and
uncertainty of the measured quantities) rather than mathematical. Improvements
may be expected, not from additional DSC experiments, but from other analytical
methods. For the present, therefore, although the elementary reaction steps (5) and
(6) must be favoured, reliable kinetic parameters remain unavailable (the ratio of
both rate constants is also unknown). Their replacement by step (4) is a necessary
consequence. The situation is analogous for the determination of the equilibrium
constants for the reactions (1) and (2) in the above scheme. The calculation
provides only mean equilibrium constants. Formally, it should be possible to
optimize the products, concentration times activity coefficient, instead of the
concentrations to get the thermodynamic equilibrium constants (in the sense that
these are independent of the composition of the system). Of course, this fails be-
cause the parameter set is now twice as large and yields meaningless equilibrium
constants.
6.2.3 The Glass Transition Process
6.2.3.1 The Phenomenology ofthe Glass Transition
Indirect Cp measurements intended to provide information regarding the glass tran-

sition are much commoner than direct determinations of the temperature dependent
heat capacity (cf. Sect. 6.2.1). The best experimental procedure and an unambi-
guous interpretation of the results is not possible without basic information about
the nature of the glassy state and the transition process between glassy (amorphous)
and liquid states.
Most materials can be obtained in the glassy state by suitable treatment. In-
organic systems have been known since ancient times, they are responsible for the
name of this state of matter. However, from the DSC point of view, organic glasses
and, especially, macromolecular glasses are of special importance.
1YJ>ical values for the temperature ranges of the transition are 10K for low-molar
mass glasses, 20 to 50 K for most of the organic polymeric glasses and lOOK or
more for silicate network glasses. The glass transition is characterized by an appro-
priately defined glass transition temperature 'fg. Within the transition region many
macroscopic properties, which may have great practical importance, change their
values (viscosity, dielectric and especially mechanical properties). Each of these
could form the basis of an experimental determination of 'fg. With DSCs, the glass
transition is detectable by a step change of the heat capacity t::..cp on heating or
cooling. Both the temperature and magnitude of this event are important. Observed
values of t::..cprange from 0.1 to 2 J g-l K- 1• Fig. 6.25 shows a typical DSC curve for
a linear, high-molar mass epoxide-amine polyadduct which was first cooled at the
rate shown and then immediately reheated.
For polymeric glasses the temperatures of the transition range from -100 to
300°C whereas the corresponding range for silicate glasses is from 500 to
1000 °C. Polymeric glasses are thus more suitable for DSC measurements
than inorganic glasses. For polymer chemists and materials scientists a knowledge
of the glass transition temperature is at least as important as is the melting tem-
perature.
3.0
c o
hea ting run
-....
OJ
CI
1.5
~
0
CI
A
OJ
J:
cooling run
o
o
40 55 70 85 .( 115
temperature ..
Fig. 6.25. Typical DSC curves for an amorphous polymer in the glass transition region.
A: glassy non-equilibrium state, B: glass transition region, C: "relaxation peak", D: equilibri-
um liquid state (sample: linear epoxide-amine polyadduct, heating and cooling rate: 5 K min-I)
6.2.3.2 The Nature of the Glass Transition and Consequences

for DSC Measurements
Problems result from the fact that the glassy (or vitreous) solid is, thermodynam-
ically, far from equilibrium. The formation and behavior of a glass are exclusively
kinetic events. There are only formal similarities between the Cp-change at an
"ideal" glass transition and at a thermodynamically well defined second order
transition. Only the liquid (or rubbery) state at the high temperature end of the glass
transition is at equilibrium and thermodynamics is valid without any restrictions in
this region. Here any external (experimental) influence is slow compared with the
mobility of the internal degrees of freedom which thus always are in equilibrium.
Equilibrium thermodynamics can also be applied if the internal degrees of freedom
react very slowly, in this case the system is in a frozen (vitrified) state with respect
to external influences. This is true for temperatures well below the glass transition.
The heat capacity function can, therefore, be determined (see Sect. 6.2.1) for both
ranges without any restrictions. During the glass transition, changes of both the
intrinsic and the measurement variables occur on the same time scale, the measured
quantities become time-dependent, and classical thermodynamics is no longer
valid. The system passes through a sequence of non-equilibrium states during
heating or cooling. The typical asymmetric shape of the glass transition curve
(an extensive tail on the low temperature side and a fairly abrupt end at high
temperatures) is due to a distribution of the intrinsic variables which vitrify (or

devitrify) over a wide range.
A special problem when investigating the glass transition (key word: relaxation
phenomena) is caused by the effect of the previous history of the glass on the
thermal behavior. The system has, so to speak, a memory of its thermal history.
This is important not only for theoretical investigations but also for practical DSC
measurements of the glass transition.
Essential conclusions are:
- In contrast to the measurement of equilibrium transitions, it is not possible to get

'equilibrium' values of the characteristic quantities Tg and llcp by extrapolation
to zero heating or cooling rates: these quantities are determined by the thermal
history (scanning rates and annealing times). If, for instance, the cooling rate is
changed by an order of magnitude Tg will change by 3 to 20 K depending on the
material in question. For flexible polymers the magnitude is generally 3 to 5 K,
whereas it is 15 to 18 K for the considerably stiffer borosilicate network glasses.
By contrast llcp shows much less dependence on thermal history.
Characterization of glasses by DSC measurements (Fig. 6.25) is mostly carried
out in the heating mode because this is practicable even for those heat flux DSCs,
having relatively sluggish furnaces. A heating run is also advantageous when it
is important to characterize the "as received" glassy state (e.g resulting from
particular cooling or anealing procedures or chemical reaction). It must be
remembered that, on heating, the original glassy state can change at temperatu-
res as much as 50 K below the transition region (for instance, ifthe heating rate
is slower than the previous cooling rate). Tg then depends on the heating rate. To
sum up, values for Tg (and to a lesser extent llcp) are only meaningful with
respect to the chosen experimental conditions and - as will be pointed out later-
in the context of the particular definition of the "glass transition temperature".
If the glass sample to be investigated is formed only by cooling from the liquid,
a cooling run is indeed better for its characterization. Problems due to the cou-
pling of vitrification and devitrification processes (occuring during heating) are
avoided. Cooling runs, therefore, immediately reflect - possibly after desmearing
of the experimental curve - the kinetics of the vitrification. The cooling rates
must not be too low (~5 Kmin- I ) for a reliable determination of the rather small
changes of the heat capacity. This may sometimes be difficult because of the
large inertia of many DSC furnaces in the cooling mode.
In addition the measured curve of the glass process is falsified by the limited heat
transfer between sample and temperature sensor (cf. sect. 5.4.4). Correction of
these influences is important, because larger sample masses (l0 - 20 mg) and lar-
ger scanning rates (10 or 20 Kmin- I ) are used in practice. For this purpose the
advanced desmearing with the aid of the step response function (cf. Sect. 5.4.4)
is very advantageous.
Thermal lag can be determined using the following methods.
1. Determination as explained in the Sects. 5.4.4 and 6.2.1.1 (Figs. 5.10 and 6.9).
2. Placing a small piece of indium both on the bottom and in the middle of the
sample and noting the difference between the extrapolated onset temperatures
from the In fusion peaks which corresponds to the thermal lag. Attention
must be focused on a good thermal contact between the indium and the
sample.
3. A temperature correction suggested by Hutchinson et aI., 1988. This method
appears to be problematical, however, as it already presupposes the validity of
one of the models which describes the glass process (Kovacs et aI., 1979).
A disadvantage of all of these corrections is that only an average lag is obtained
and thus can be corrected.
6.2.3.3 Definition and Determination of the Glass Transition Temperature Tc
6.2.3.3.1 Conventional Glass Transition Temperatures
The main parameters used to characterize the glass transition are shown in Fig. 6.26.
For many applications it is important to know the temperature range T g•i to Tg,f over
which the substance vitrifies on cooling, or devitrifies on heating. Unfortunately,
the practical determination of these temperatures is problematic, there are large
errors in their definition and it is difficult to give clear instructions for their mea-
surement. The situation is improved if clearer, more chararacteristic temperatures
0.20
mW
~.f - 'g,;
-,
0.15
~
t 0.10
E
=-
0
c
'"
..r:: 0.05
o L -_ _ -L~____~____~~~--~----~--~--~----~
50 'g.; 60 70 'g.80 'g.\\ 90 100 'g.f ·c 120
tempera ture ~
Fig. 6.26. Definition of the most frequently used conventional quantities for characterization
of the glass transition.
Tg ,.: extrapolated onset temperature, ~.1/2: half-step temperature, ACp : Cp-change at the half-
step temperature, ~.; and ~.finitial and final temperatures of the glass transition, Tg,f - ~.i
temperature interval of the glass transition
from the transition region are used. These are the extrapolated onset-temperature
Tg.e (analogous to the peak onset temperature) and the half-step temperature 7;.112
related to the Cp change (the temperature at which Cp is midway between the extra-
polated heat capacity functions of the glassy and liquid state). The use of the latter
is more meaningful as this temperature is better related to the second characteristic
quantity of the glass transition, the Cp change. A third definition, the temperature of
the inflection point of the glass transition curve is seldom used.
Under certain circumstances glass transition temperatures defined in this way
may be used without restriction to compare experimental data:
1. All investigations are made in cooling mode at the same rate.
2. If evaluation must be done on heating runs, the sample must be previously coo-
led from the liquid at a fixed rate.
Tg.. and 7;,112 can both easily be obtained from routine measurements, this is the
reason for the nearly exclusive use of these pragmatically defined glass transition
temperatures up to now. A repeatability error of ± 1 K is acceptable in practice.
For heating runs the following practical and frequently used procedure is recom-
mended:
The sample is heated to a temperature at least 15 to 30 K above 7;.
Short (5-10 min) annealing at this temperature in order to establish thermo-
dynamic equilibrium and erase the "memory" (with respect to its thermal
history) of the system.
Rapid programmed cooling (or quenching) to a temperature at least 50 K below
the glass transition.
Immediate reheating at constant rate (10 or 20 Kmin-I). These rates lead to
relatively high temperature errors (3 to 10K) and, in addition, a, broadening
of the transition, on the other hand problems caused by relaxation processes
during the transition are avoided. The thermal history can be obtained quanti-
tatively by comparison of the original run with the rerun under the same con-
ditions.
Neither Tg.. nor Tg.112 make any allowance for the nonequilibrium nature of the glass
transition. This is especially striking (Fig. 6.27), if the glass transition is accompa-
nied by "enthalpy relaxation peaks". These appear on heating curves as endother-
mic events at the high temperature end of the glass transition range. Their height
may be comparable with those of the melting peaks of crystalline materials (Petrie,
1972). For the example shown in Fig. 6.27 the glass annealed 70 h at 68°e has
Ta.. = 89.1 °e, Ta. I12 = 87.3°e whereas the quenched glass has Tg.• = 81.6°e,
Tg.1/2 = 85.1 °e. This use of Tg.. and Ta. 112 to characterize the glass process gives the
paradoxical result that a slowly cooled or annealed glass seems to have a higher
glass transition temperature than a rapidly cooled one. In addition Tg,e and Tg.112
react to the thermal history in a different manner.
Til,. and Tg. 112 are not useful for theoretical treatments ofthe kinetics of the glass
process, this is also true for certain relationships between Tg and other properties
(e.g. Tg as function of the molar mass or as a function of the degree of conversion
in a reacting system).
mW
annealed glass
~
1
~
2
...
c
~
0
c
'"
.c
o
60 70 100 .( 110
temperature ..
Fig. 6.27. DSC heating curves for an annealed (or slowly cooled) (a) and a quenched (qu)
glass showing the paradoxical result that the conventional glass temperatures of the quenched
glass seem to be lower than those of the annealed glass, for details see text. Ta.. extrapolated
onset temperature, T,,112 half-step temperature
6.2.3.3.2 The Thermodynamically Defined Glass Transition Temperature
The "glass transition", although a kinetic phenomenon, can be unequivocally defi-

ned thennodynamically. This uses the so-called fictive temperature 1'g,fic first intro-
duced by Tool, 1946. 1'g,fic is a well defined quantity that reflects the current struc-
tural state of the glass to be characterized.
The concept of the fictive temperature was considerably extended by Naraya-
naswamy, 1971, Moynihan et aI., 1976, and de Bolt et aI., 1976. An analogously de-
fined "thennodynamic" or "enthalpic" 1'g-temperature, based on the specific infor-
mation of a DSC, was introduced by Flynn, 1974, Richardson, Savill, 1975a, and
Richardson, 1976. In the following section this thennodynamically defined tem-
perature Tg. fie is shown as 1'g without the additional suffix.
To understand the definition of 1'g, the enthalpy versus temperature diagram (Fig.
6.28) will be discussed. Heat capacity functions for the glass and liquid can be
described approximately by straight lines within a temperature range of 50 to
100 K. The enthalpy functions are then slightly parabolic curves. For simplicity,
curvature of the enthalpy functions is neglected in this figure, in other words Cp is
assumed to be temperature independent. Depending on the cooling rate, the sample
vitrifies (changes from HI to Hg) at different temperatures. The lower the cooling
H9 (-{3,) _________
1
>-
Cl..
Hg(a1)~
d
:E
c:
OJ Hg (-{3d/
Hg(a2)~
temperature ..
Fig. 6.28. Schematic enthalpy-temperature curves for the glass transition of an amorphous
sample at different cooling rates (- fJJ and subsequent heating after isothermal annealing at Tai ,
for details see text. H, enthalpy of the different non-equilibrium glassy states, HI enthalpy func-
tion of the equilibrium liquid state, T,(-fJi) glass transition temperatures on different cooling
(I PI 1> 1~I), ~ (a;) glass transition temperatures on heating after different annealing schedules
rates (-/3) during this process the lower are the vitrification temperatures (1'g(-/3I»
or 1'g(-/Jz». The enthalpy functions Hg(-/3I) and Hg(-fJz) characterize the two res-
pective glasses. For simplicity it is further assumed that all enthalpy functions are
parallel, in other words that the heat capacities of the glassy state are assumed to be
independent of the conditions during vitrification (although very precise measure-
ments (Gilmour, 1977) show slight differences). If the glass is annealed at tem-
peratures down to, at most, 50 K below the vitrification temperature, the mobility
of the frozen states is still so large that internal degrees of freedom are not totally
frozen and can relax towards equilibrium. During this process the Hg-function
approaches the (extrapolated) Hcfunction (Petrie, 1972; Peyser, 1983; Cowie,
Ferguson, 1986; Agrawal, 1989). The figure shows this for two annealing tempera-
tures Tal and 1'.2 of a glass, which was obtained at a cooling rate -/31' At 1'.1 the
annealing time was sufficient to reach the equilibrium enthalpy value of the liquid
at that temperature, whereas this was not the case at Ta2 • Reheating of the annealed
glass then proceeds along the enthalpy lines Hg(al) or Hg(a2). From the theoretical
point of view, the glass should devitrify exactly when the enthalpy line of the glass
crosses that ofthe liquid but the transition from Hg to Hl is not sharp. To determine
the intersection of enthalpy curves for the glassy and liquid states therefore requires
the extrapolation of these curves from temperatures, which are clearly above or
below that of the transition region. The point of intersection, obtained in this way,
defines the thermodynamic glass transition temperature I'g. Tool, 1946, called this
temperature the fictive temperature, because during heating nothing happens at that
point. Hence I'g cannot be located directly on the measured curve, instead, on heat-
ing, the system progresses further along the Hg-curve (superheating effect). This is
more pronounced the better the glass has been annealed (i.e. after annealing at Ta2 it
is far more intensive than after annealing at 1,;1)' This is the reason for the parado-
xical values, mentioned earlier, for the pragmatically defined I'g,e or I'g.1I2 tempera-
tures, when comparing slowly cooled (or annealed) and quenched glasses.
Superheating ends only at temperatures well above I'g, the return to the equilibrium
curve is now rapid and produces the so-called relaxation peak. The reason for the
superheating effect is the drastically decreased mobility in the glassy state, which
parallels the slow enthalpy decrease during annealing.
If the enthalpy definition (Flynn, 1974; Richardson, Savill, 1975a, and Richard-
son, 1976; Moynihan et aI., 1976) is used, I'g can easily be calculated from DSC
measurements.
The procedure in question (Richardson, Savill, 1975 a; Richardson, 1976) is
explained in Fig, 6.29, We start with the definition, the equality of hg(T) and hl(T)
4.5
J/(Kg)
3.5
t
>-
3.0
0
0..
0
2.5
~
( p,1
0
-------------
2.0
QJ
.<=
~
'u
QJ
'" 1.5
0..
1.0
60 Tl 70 90 11
tempera ture ~
Fig. 6.29. Determination of the thermodynamic glass temperature from DSC heating curves
with a "relaxation peak" (Richardson, Savill, 1975a; Richardson, 1976),
cp " and Cp,l are the (extrapolated) specific heat capacity functions of the glass and liquid, res-
pectively, T. and T2 are the integration limit temperatures in the glassy and liquid state, res-
pectively, the hatched area (extended to cp = 0) corresponds to the difference h t (T2 ) - hg(T.).
For details see text
at Ta. The enthalpies h(T) for the glass and liquid are obtained by integration of the
corresponding cp-fimctions, which can always be approximated by linear equations:
glass: cp.g=a+b·T and hg(T)=a'T+~b'T2+P
liquid: cp.l =A + B· T and hl(T) =A . T+~B' T2 + Q

The integration constants P and Q are not known but Q-P may be obtained from
the difference hl(TJ - hg(T1), the hatched area in Fig. 6.29, a directly accessible
experimental quantity:
hl (T2)-hg (1j) =A . T2 -a'1j +~B' n- ~b· 11+ (Q-P)
TI and T2 are convenient arbitrary temperatures that must be chosen to be below and
above the glass transition region (i. e. in the glassy and in the liquid state, res-
pectively). Using Q - P from the above equation the desired glass temperature is
obtained by solving the quadratic equation:
!(B- b)· T~+ (A -a) Tg + (Q-P) = 0
Fig. 6.30 shows an equivalent, graphical procedure (Moynihan et al., 1976) for
determining r..
From the enthalpy definition it follows that:
T2 T2
J(Cp.l (T) - cp.g(T» dT = J (cp (T) - cP,II(T» dT
T. T\
where cp(T) is the experimentally determined curve and Cp,g(T) and Cp,l(T) are the
(linearily extrapolated) specific heat capacities of the glass and liquid, respectively,
the integration limits TI and T2 have the same meaning as those in the last figure. It
must be guaranteed, however, that TI (on heating) and T2 (on cooling) are definite-
ly in the steady state region of the DSC. The lower limit of the left hand side inte-
gral, the thermodynamic glass temperature Tg, must be determined so that the inte-
grals on both sides are equal. In other words, the area between TI and T2 and the
(extrapolated) cp curves of glass and liquid (the left side integral) must be equal to
that between the experimental curve and the (extrapolated) cp curve of the glass
(the right side integral). The two areas are hatched differently in Fig. 6.30.
It is clear from the equation above that absolute values of heat capacities are not
required for calculating Tg • It is sufficient to know cp differences and this is also the
case for the Richardson and Savill procedure. Nevertheless, their changes with tem-
perature must be determined very precisely. This demand can only be fulfilled if the
repeatability of the experimental curve is very good (for the same thermal history)
and if a sufficiently large temperature range (more than 50 K on each side of Ta) is
available for extrapolation.
An error estimation was done by Richardson, Savill, 1975a: for a typical Acp of
0.3 J g-I K-I an uncertainty of 0.3 J g-I for the enthalpy change would result in a
temperature uncertainty of ± I K. For total enthalpy changes of about 100 J g-I and
caloric errors of ± I %, the determined Tg would be uncertain to ± 3 K and this is
not acceptable in practice. Fortunately, some errors tend to compensate each other
in both procedures. For instance, an incorrectly extrapolated Cp,l curve (Fig. 6.30)
4.5
J/IKgJ
3.5
t 3.0
€.,
C>. 2.5
8
~
------
cu 2.0
-......
..c:
cu
C>.
III
1.5 Cp•g
1.0
60 r, 70 90 100 Tz DC 11
tempera ture ..
Fig. 6.30. Construction to determine the thermodynamic glass temperature T., which is de-
fined by the equality of the different hatched areas (for other quantities see Fig. 6.29)
has the same influence on both hatched areas, but the same only holds to a limited
extend for the extrapolation of cp•g ' Ta cannot, therefore, be determined to better
than ± 0.5 to ± 1 K with this method during routine measurements. However, this is
not the reason why the thermodynamic Ta is so rarely used in practice - ± 1 K
is adequate for many investigations of common Ta relations.
In fact, the sparsity of thermodynamic Ta data has mainly been due to a lack of
suitable programs in manufacturers software. The situation is now changing rapid-
ly and "fictive temperatures" (Ta.fic = Ta> can be calculated for most instruments.
This is fortunate because for theoretical investigations of the kinetics of the glass
processes the situation is clear: only the thermodynamically defined glass transiti-
on temperature reflects unambiguously the thermal history and all the other condi-
tions during the formation of the investigated material. Tg is thus the central quan-
tity for all kinds of relaxation studies. The aim of such investigations is to reprodu-
ce the behavior of the glass in the transition region, i.e. in the case of DSC measu-
rements to reproduce the course of the function cp = cp(T) precisely. As the changes
in Ta, caused by different thermal histories, may only be of the order of a few tenths
K, Til has to be determined at least with that precision.
To minimize the errors in determining Ta,
- all error sources, associated with the determination of cp (cf. Sect. 6.2.1) must be
borne in mind and carefully excluded.
the sample should remain untouched in the apparatus during all experimental
manipulation even for (often time consuming) annealing experiments. Annealing
the sample outside the apparatus almost always yields unsatisfactory results
because heat transfer conditions are not exactly reproducible after replacing the
sample in the DSC.
The values for the glass transition temperatures obtained from DSC measurements
need not necessarily agree with those of other methods. Discrepancies are caused
by the different influences of the particular technique on the relaxation of the in-
trinsic variables (Duncan et aI., 1991). A formal conversion, taking into account the
various experimental influences, can be made using the (modified, if necessary)
equation of Williams, Landel, Ferry (WLF) (Williams, 1955). Principal differences
must be attributed to different interactions between the method in question and the
relaxation time spectrum of the intrinsic variables.
In studies of this kind it is indeed important to ensure that experimental errors
have really been minimized and that any direct influence of the apparatus on the
results is at least understood, if not avoidable. To illustrate this problem, in Fig. 6.31
the Tg values (open symbols) of polystyrene (determined as explained here) for
different cooling (circles) and subsequent heating (triangles) runs at different rates
are shown (Schawe et aI., 1996). As can be seen, there is a significant difference bet-
ween the results of heating and cooling obtained at a particular rate. In addition, this
difference increases with the heating (or cooling) rate in question. From the theore-
tical point of view, there should not be any difference between the thermodynamic
2.0
1.5
.("
c
'E
:::.:: 1.0
.......
co.
'-"
9) 0.5
0.0
370 375 380

Tgln K
Fig.6.31. Thermodynamic glass transition temperature T& of polystyrene from DSC measure-
ments as a function of heating or cooling rate. (Circles: cooling mode; triangles: heating mode;
open symbols: as measured, solid symbols: desmeared, acc. to Schawe et al., 1996)
glass transition temperature measured in the heating and cooling mode if the sam-
ple has been cooled with the same rate before the heating run. The glass transition
should only depend on the procedure according to which the sample has been trans-
formed from the liquid state to the glassy state. The differences measured result
from the smearing effect due to the heat transfer path and the temperature profile
inside the sample (cf. sect. 5.4) which causes a lag of the sample temperature rela-
tive to the measured one. Thus the experimental glass temperatures are no fictive or
thermodynamic ones, though determined as such. If the step response desmearing
procedure described in Sect. 5.4.3 is applied and the glass transition temperature
determined accordingly, the result is quite different (solid symbols in Fig. 6.31).
Now the corrected values from the cooling and heating experiments almost super-
impose. Furthermore, there is clearly a systematic change in Tg with the cooling rate
(1.9 K per decade). This figure is comparable with the corresponding value from
the activation plot of mechanical or dielectrical measurements carried out on the
same sample. Nevertheless, absolute values are shifted half a decade with respect
to dielectric results if one calculates an effective frequency from the cooling rate in
question and this agrees with modulated temperature DSC measurements (Schawe,
unpublished).
Although the DSC method is very convenient for the characterization of the glass
transition, it is not very sensitive. If the cp changes are small and take place over a
broad temperature interval (as, for instance, is the case for lightly crosslinked poly-
mers), the evaluation of the DSC curve is difficult and uncertain. Dynamic me-
chanical or dielectric measurements are then more suitable.
6.2.3.4 Applications of Glass Transition Measurements

Many important applications ofDSC measurement in the glass transition region are
connected with polymer research, these are:
theoretical investigations concerning the thermokinetics of the glass transition.
The quantitative description and modelling of relaxation phenomena implies
either the determination of enthalpy changes (Petrie, 1972; Cowie, Ferguson,
1986; Agrawal, 1989; Montserrat, 1992; Hay, 1992) or of the thermodynamic
glass transition temperature (Moynihan et aI., 1976; Stevens, Richardson, 1985).
An understanding of the relaxation processes is not only of theoretical interest
for physicists but also allows a better access to physical aging phenomena in
glassy, polymer materials (Struik, 1978; Cowie, Ferguson, 1986; Perez et aI.,
1991). Quantitative descriptions of the phenomena are often based on scanning
experiments alone (Kovacs, 1963; Kovacs, Hutchinson, 1979; Moynihan et aI.,
1976; Ramos et aI., 1984; Hutchinson, Ruddy, 1988; Hutchinson, 1990; 1992;
Chang, 1988). However, some aspects of glassy behavior can be better
studied following various isothermal annealing schedules (Petrie, 1972; Cowie,
Ferguson, 1986; Agrawal, 1989; Montserrat, 1992; Hay, 1992). In this method
the enthalpy difference between the annealed glass and the quenched glass is
determined and evaluated using the empirical Williams-Watts function.
Tg as one of the most important properties of amorphous materials. For produc-
tion of new materials with defined properties, the prediction of the glass tran-
sition temperature is a natural necessity. Predictions can be based on group con-

tributions (Becker 1976; 1977; 1978; Wunderlich, 1978). Possible discrepancies
between calculated and experimental Tg values may then contribute to a better
understanding of the relations between structure and properties of a specific
material. Similar investigations concerning the correlation between structure and
the caloric information (Mp of the glass transition) are still rare to find.
finding of relations between Tg and caloric quantities of the glass or the glass
process (Becker, 1976; Batzer, 1982; Wunderlich, 1990).
relations between molar mass and glass transition temperatures of polymers. The
most widely used equations are those of Fox, 1950:
A
1:=T
g g.-
-M
-
n
and Uberreiter, Kanig, 1952:

1 1 A'
-=-+-
Tg Tg._ Mn
where Tg._ is the glass transition temperature of a polymer with infinite number
average molar mass and A and A' are constants for certain broad classes of
materials. Both equations can be used for high-molar mass polymers but that
of Ueberreiter and Kanig is much better for oligomeric glasses. If, in such in-
vestigations, the glass transition temperature must be determined from heating
runs, the thermodynamic Tg should always be used. All other (empirical) Tg's are
unsuitable because of the inevitable relaxation processes on heating (Agrawal,
1989; Aras, Richardson, 1989).
investigation of polymerization reactions. Tg is often more sensitive to the pro-
gress of a polymerization than is the heat production. This is especially true
towards the end ofthe reaction where the heat production is low and barely detec-
table, whereas a distinct change of Tg may be observed (Mijovic, Lee, 1989;
Wisanrakkit, Gillham, 1990; Hale et aI., 1991), for instance, if the crosslink den-
sity increases. As typical example DSe curves for an epoxide-amine system,
which has reacted different times at 160 °e, are shown in Fig. 6.32.
It is generally not possible, and in any case it would be far too time-consuming,
to determine the degree of reaction only from the changes of Tg during poly-
merization. In addition the difficulties mentioned in determining exact Tg values
would result in very uncertain results for any kinetic analysis. The situation
is even more complicated because the glass transition for reacting systems
may often be overlapped by the reaction heat flow and thus not precisely deter-
minable.
A "calibration curve" must be determined connecting the degree of reaction
(obtained by another method such as an IR technique or, for soluble polymers,
determination of the molar mass) with the Tg values obtained by DSe measure-
ments, nearly identical experimental conditions should be chosen for the dif-
ferent methods. The empirical function Tg =f( a), obtained this way, is not
related to a specific reaction model but it can be calculated if some assumptions
lb 11
4h
3h
2h
15 h
1h
f
<IJ
~
....
0
-
::J:
0
0
<IJ
~
o 100 200 300 O( 400

tempera ture •
Fig. 6.32. DSC measurements of an epoxide cured at 160°C for the times shown: the glass
transition temperature increases with extent of reaction (acc. to Wisanrakkit, Gillham, 1990)
are fulfilled: for instance, the Fox-Flory model can be used to describe the
linear (L) polyaddition reaction:
Tg = Tg,o + a . «Tg,~)L - Tg,o) = Tg,o + ~Tg,L
Tg,o and (Tg,~)L are the limiting values of the glass transition temperatures for the
monomer mixture and the linear polymer with infinite molar mass, respectively.
As early as 1970 Horie et al. tried to extend these considerations to reactions
which give crosslinked polymers. In this case Tg shows a more pronounced rise,
due to the increasing crosslink density, than is the case for linear polyaddition:
Tg = Tg,o + ~Tg,o = Tg,o + ~Tg,L + ~Tg,v
The relation between ~Tg, v and the crosslink density v is complex and the sepa-
ration of ~Tg into ~Tg,L and ~Tg.v may not be possible but Min et aI., 1993, were
successful. They found by NIR measurements that in the system bisphenol-A-
diglycidylether (DGEBA) and 4,4' -diaminodiphenyl-sulfone (DDS) the reaction
of the primary amine hydrogens only increases the molar mass of the linear poly-
mer whereas that of the secondary amine hydrogen is responsible for cross-
linking. Hence the total degree of reaction a can be separated into one part a L
for the linear polymerization and another part a v for the crosslinking reaction.
Both parts can be individually obtained from the experiment. If Tg,_ represents
the glass transition temperature of the completely reacted and crosslinked poly-
mer, the following is valid:
Tg = Tg,o + M~,L + M~,v = Tg,o + aL((Tg,_)L - Tg,o) + av(Tg,_ - (Tg,_)d
Investigation of enthalpy relaxation as a function of the degree of cure. Polymers
heated for the first time after their production often show very large "relaxation
peaks" (Sect. 6.2.3.3.2) which cannot be reproduced by any subsequent thermal
manipulation (Montserrat-Ribas, 1992). The reason for this exceptional relaxa-
tion behaviour is not yet known. The shape and intensity of the relaxation peak
reflect somehow the specific conditions of the polymer forming process and so
far it has not been possible to decode this kind of "memory".
The glass transition temperature of copolymers, blends and polymers containing
plasticizers. The influence of plasticizers on Tg of polymers or the glass transiti-
on behavior of blends have often been investigated (Fox, 1950; Couchman,
Karasz, 1978; Chee, 1985; Kalachandra, Turner, 1987; Braun et aI., 1988; Bre-
kner et aI., 1988; Tsitsilianis, Staikos, 1992; Pomposo et aI., 1993). Homoge-
neous blends have only one Tg transition, whereas heterogeneous blends show the
separate glass transitions of the components with llcp values correlated with their
relative masses, The glass transition temperature of an ideal homogeneous
system of two components can be calculated from their Tg values and mass frac-
tions Wi by the Fox equation (Fox, 1950):
1 WI W2
-=-+-
Tg Tgl Tg2
or by the Karasz equation (Couchman, Karasz, 1978):
WI • llcpl • In Tgl + W2 • llcp2 • In Tg
In~= 2
WI' llcpl +W2' llCp 2
Possible interactions between the components may have very different effects
depending on the composition of the system in question. Several equations have
been suggested to describe the influence of molecular interactions on Tg with the
aid of up to three additional parameters (PodeSva, Prochazka, 1979; Braun et aI.,
1988).
Relaxation experiments are helpful for deciding the sometimes difficult problem
of whether a polymer blend is homogeneous or heterogeneous (Jorda, Wilkes,
1988; Tsitsilianis, Staikos, 1992). Following their proposals, a clearly detectable
relaxation peak formed during the annealing of the sample is used as an "ampli-
fier" for glass transitions, which otherwise only can be detected with difficulty.
In this way glass transitions can even be detected which are only a few degrees
apart from one another and which would normally overlap to give one broad tran-
sition and thus imply a homogeneous product. As an example Fig. 6.33 shows the
DSC curves of one homogeneous and two heterogeneous glasses before and after
annealing. The two heterogeneous blends differ in the amount of both co-
polymer components. After annealing the homogeneous sample shows only one
quenched
f annealed
-
QJ
c
r...
B
--
~
0
quenched
C
QJ
.t::
annealed
(
Fig.6.33. DSC curves of a homogeneous (A) and
quenched two heterogeneous (B,C) blends which have been
quenched from the melt or annealed a few K
below T. for 14 days. The relaxation peak:(s) of
30 50 70 "C 90 the annealed samples clearly show that blends B
and C are phase-separated (acc. to Jorda, Wilkes,
tempera ture 1988)
non-structured relaxation peak, whereas the heterogeneous ones possess two

separated relaxation peaks, though the DSC curves of the unannealed glasses
show no signs ofheterogenity.
- the glass transition of semicrystalline polymers. The inverse correlation between
the glass transition and melting process has been investigated (Schick et aI.,
1985; 1989). It is often found that the sum of the amounts of amorphous materi-
al (determined from Acp) and of crystalline material (determined from the heat
of fusion) is lower than expected for a two phase model. In these cases there must
be a third phase (called "rigid amorphous"), which shows neither amorphous
(glass transition) nor crystalline (melting) behavior. In addition, correlations
have been found (Okui, 1990) between the glass transition temperature and other
characteristic temperatures, e. g. the melting temperature or the temperature of
the maximum crystallization rate.
6.2.4 Purity Determination of Chemicals
The theory of purity determination using DSC is based on the thennodynamics of

two-component systems. The simplest and most widely used theory presupposes an
eutectic mixture of ideal behavior. For this case we find in textbooks of chemical
thennodynamics that the mole fraction x of one component (defined here as the
pure one) can be calculated as follows:
1 T !ili(T)
Inx=- J - - d T (6.1)
R Tin T2
(R, gas constant: 8.31441 J mol-I K-I; !ili(T): phase transition enthalpy at tempera-
ture T of the pure component in question, TIrs: phase transition temperature of this
completely pure component).
In the case of a negligible temperature dependence of the phase transition enthalpy,

eq. (6.1) is reduced to
(6.2)
This equation serves to calculate eutectic phase diagrams and the melting behavior
of slightly impure materials. If a change is made to the mole fraction of the (eu-
tectic) impurity Ximp = 1 - x and In (1 - Ximp) approximated (for small values of
Ximp) by - ximp and the product TIrs • T by T;;', we obtain:
Ttrs - T
x imp (T) -- !ili -
R.T.2- (6.3)
Irs
This is the well-known van't Hoff equation which relates the decrease in the melt-
ing temperature of the impure component to the amount of impurity involved. In
principle, any melting of an eutectic mixture starts at the eutectic temperature and
ends at the temperature T which is related to the composition of the mixture via
Eq. (6.3) (see Fig. 6.34) and may serve to determine the amount of impurities. The
theoretical heat flow rate curve of eutectic mixtures can be calculated: the result
is shown in Fig. 6.34 for different concentrations. With increasing impurity, the
melting peak becomes lower and lower and less sharp. In reality, for small amounts
of impurity, the eutectic melting peak and the starting of the main peak are hardly
visible on the DSC curve. The measured curve (Fig. 6.35) is also smeared due to the
finite thennal conductivity (cf. Sect. 5.4) so that it differs considerably from the
theoretical shape (Fig. 6.34). The end temperature of the peak cannot be determined
with the required accuracy and the concentration of the impurity must therefore be
determined in a different way.
The software in commercial DSCs usually starts from the assumption that,
during the melting process, the instantaneous mole fraction of the mixture at
the temperature T (along the eutectic curve, cf. Fig. 6.6 a) relative to the initial com-
mW
1 3
~
~
~
0
~
<11
.c.
oL-----------~~~~~~~~~~
340 345 350 K 355
tempera ture •
Fig. 6.34. Theoretical melting peaks, normalized to unit area (1: 0.20%; 2: 0.55%;
3: 1.05%; 4: 2.03%; 5: 4.67%; 6: 6.65% impurity)
position is the same as the relation between the heat of fusion used up to that tem-
perature and the total heat:
(6.4)
The right-hand side of this equation is the quotient ofthe partial peak area up to the
temperature T and the total peak area if smearing effects are disregarded. For this
quotient, the abbreviated form F(T) (relative partial area) is often used. Inserting
Eq. (6.4) into (6.3) furnishes an equation according to which the impurity ximp can
be determined from the slope of a plot of the temperature versus VF(T) which
should be a straight line. In practice this is not the case because ofthe difficulty in
deciding when melting actually starts and the limited heat transfer to the sample.
Commercial software generally corrects the measured values in such a way that a
straight line is obtained in the plot. This is of course only an approximation to
correct for deviations from experimental and theoretical shortcomings (e. g. the
smearing effects), nevertheless it is widely used. To avoid larger errors, small
sample masses and low heating rates should be used for purity determination.
However, the signal will then be noisy and the partial areas cannot therefore be
detected very precisely.
Recently Bader et a!., 1993, presented a new method which determines the impu-
rity from the shape of the measured curve at the start of the melting process, just
0.6
mW
2
0.4
f 3
ClJ
~ OJ
....
Cl
~
0
;;:
Cl
0.2
ClJ
..c
0.1
0.0
320 330 340 K 360
tempera ture ~
Fig. 6.35. Real melting peaks measured in a DSC and normalized to unit area (impurities as
in Fig. 6.34)
behind the eutectic point where the heat flow rates are still small and falsification
due to smearing low. In principle, the shape of the melting peak is determined from
the phase diagram of an ideal mixture of two components and can be calculated
from Eq. (6.1) or (6.2) (cf. Fig. 6.34). If the equation in question is solved for x as
a function of temperature and the (normalized) heat flow rate at the respective tem-
perature, a straight line should be obtained which intersects the ordinate at the value
of the purity of the sample and drops to 1.0 at the maximum temperature of the peak
(Fig. 6.36). The same type of plot can be calculated from the heat flow rates
(normalized and with the baseline subtracted) of the measured curves (Fig. 6.35), it is
shown in Fig. 6.37. The result does not come up to what had been expected because
of the difficulty of precisely determining the baseline and thus the beginning of the
peak. It has, however, been shown (Bader et aI., 1993) that some minor parameter
variations (well below the uncertainty of the measurement) correct the problem
satisfactorily (Fig. 6.38). Further details of this special method are not given here,
as it has not yet been accepted by the manufacturers and the user may be unable to
implement it in his computer himself. The interested reader is referred to the origi-
nal publication. The advantage of using the shape at the start of the melting peak
(rather than the end temperature or the partial areas procedure) for evaluation is the
1.00 III 1
112
mole
fraction 3
0.98 4
t 0.97
>-
.~
0.96
:::>
5
Q.
0.95
0.94
6
0.93
330 335 340 345 350 K 360
tempera ture ~
Fig.6.36. "Purity function" from theoretical melting peaks of Fig. 6.34
1.01
mole
fraction
0.99
0.98
t
>-
0.97
.~
:::>
Q. 0.96
0.95
0.94
0.93
330 335 340 345 350 K 360
temperature ~
Fig.6.37. "Purity function" from measured melting peaks ofFig. 6.35
1.00 ++1 11+1 I I I I '","PI I I I I I I I I 1

mole 11++1"1"1+1 I I I I I 1 1 1 1 1 I 2
fraction I I'" I "1"++ I I I 1 I 1 "'" I + I 1"1" 1 I +1" I 3
0.98 +++1++1 1 +","11+111+1"++1 I , 4
1 Q.96
++ I + I I • I + I I +* I I l..v+ I I I I I I 5
0.94
+++++++++ I I I I I I I air I I I +++'"1- I I 6

0.92 L -_ _ _--'-_ _ _ _..L-_ _ _- ' -_ _ _ _--'-_ _ _- - '
330 335 340 345 K 355

temperature ..
Fig.6.38. Corrected "purity function" of measured melting peaks from Fig. 6.35 (impurities
as in Fig. 6.34)
negligible falsification due to smearing of this part of the curve. This permits the
use of larger heating rates and sample masses which increases the signal-to-
noise ratio and the accuracy of the measurements.
From the example given it follows that the accuracy of all purity evaluations
is largely influenced by the smearing of the measured curve, as the results are
calculated in one way or the other from the shape of the peak. Bearing in mind that
most theories of purity determination start from the assumption of eutectic mix-
tures, which is actually only one special case, the accuracy of every determination
of purities by DSC should be considered to be a rather limited approximation to the
truth. Nevertheless, it is convenient and therefore often used in quality control.
7 Evaluation of the Performance of a Differential
Scanning Calorimeter
The DSC furnishes information on temperature and heat flow rates (resp. heat).
Whether it is suited to solve the respective problem depends on the efficiency of
the instrument. The characteristic data of the DSC which describe the instrument
unambigously must therefore be known. They allow a decision to be taken as to
whether the DSC will be suitable for the intended use, and they also make a com-
parison with other DSCs possible.
A distinction can be made between:
1. the characterization of the complete instrument,
2. the characterization ofthe measuring system,
3. the characterization of the results ofa DSC measurement.
7.1 Characterization of the Complete Instrument

The following serves to characterize the DSC instrument as a whole:
measuring principle (heat flux or power compensation DSC),
temperature range,
potential heating rates and temperature-time programs,
- usable sample volume,
- atmosphere (gases which may be used, vacuum, pressure).
7.2 Characterization of the Measuring System

In this section characteristic terms are presented which may be used to describe the
efficiency of a DSC measuring system. Instructions how to determine the numeri-
cal values are suggested. The characteristic terms in question are the following:
- the noise,
the repeatability,
the linearity,
the time constant.
The noise of the measured signal (given, for example, in ~W) is indicated in
different ways (see Fig. 7.1):
as peak-to-peak noise{pp): maximum variation ofthe measured signal in rela-
tion to the mean signal value,
182 7 Evaluation of the Performance of a Differential Scanning Calorimeter
time
Fig. 7.1. The various definitions of noise (ace. to Hemminger, 1994).
pp peak-to-peak noise, p peak noise, RMS root-mean-square noise, <Pm measured heat flow
rate
as peak noise (P): maximum deviation of the measured signal from the mean
signal value,
as root-mean-square noise (RMS): root of the mean value of the squared
instantaneous deviations ofthe measured signal from the mean signal value.
These three definitions of noise are statistically dependent on one another; in
general, the following is valid: p = 0.5 pp; for a sine-shaped measurement signal,
the RMS noise is equal to 0.35 pp noise; for a statistical random signal, the RMS
noise is about 0.25 pp noise.
The noise of the DSC and DTA instrument depends on the heating rate, the
temperature and on other parameters (e. g. atmosphere). The signal-to-noise ratio
is decisive for the smallest heat flow rate detectable ("heat flow rate resolution"or
"detection limit" ofthe DSC). This threshold for the heat flow rate determination
amounts to about 2 to 5 times the noise (cf. Wies et aI., 1992).
It is often expedient to indicate the sample volume- or sample mass-related
noise (for example in J.lW cm- 3 or J.lW g-l) which allows the smallest detectable
heat flow rate to be estimated for a given sample.
The noise (so called "short-time noise") can be measured:
In the desired operating mode (isothermal or scanning mode at a specified
heating rate) and at the temperature of interest, the measured signal is ampli-
fied to such a degree (most sensitive measuring range) that the noise is clear-
ly recognizable.
The mean variation (pp, p or RMS) of the signal over a period of about
1 min furnishes the respective (short-time) noise (for example in J.lW) (with
the amplification factor taken into account, if necessary).
The isothermal noise should be the smallest noise possible (compare the
scanning noise of the zero line). It gives an impression of the disturbances
from the environment to which the sensors are subject. It determines the maxi-
mum possible signal-to-noise ratio.
The repeatability indicates the closeness ofthe agreement between the results of
successive measurements ofthe same kind carried out with a specific instrument
7.2 Characterization of the Measuring System 183
(acc. to International Vocabulary of Basic and General Terms in Metrology,

1984).
DSCs can be characterized by the repeatability by measuring a significant DSC
quality (e. g. extrapolated peak onset temperature, peak area, course of the zero
line etc.) several times under the same conditions on samples of the same kind.
Examples:
The repeatability ofthe zero line is determined by measuring 4 or 5 zero lines
over the whole temperature range at medium scanning rate and superimposing
the curves. The temperature-dependent range of the deviation from the
"mean" zero line (absolute or in %) gives the repeatability.
This ± range of scatter gives an impression of the uncertainty of the heat
flow rate calibration and is the systematic part of the uncertainty when an
absolute heat flow rate must be determined that is related to the zero line (for
example when cp is measured).
The repeatability of the peak area and of the baseline is determined by mea-
suring for example the peak area caused by the melting of a pure metal sever-
al times under the same condition. (Either only repeat measurements without
moving the sample crucible or with replacing the crucible after each run). The
curves with the peaks and the baselines constructed for each peak are super-
imposed. The ± range related to the mean area shows the repeatability of
the peak area (heat of fusion) determination. A separate baseline repeatability
can be determined (acc. to the procedure with the zero line, see above). The
repeatability of the peak area (taking the baseline repeatability into account)
furnishes the calibration uncertainty of the peak area (heat) calibration which
is the smallest uncertainty for heat measurements.
When the repeatability is given as a ± range (scatter), it must be stated which
measure is used: the standard deviation, the maximum deviation or another
characteristic value.
Note:
The reproducibility describes the closeness of the agreement between the results
of measurements carried out on a sample using different instruments in different
laboratories (round robin, interlaboratory comparison). The absolute or percentage
deviation ofthe mean value of the results of an instrument from the total mean value
is a measure of the reproducibility. Low reproducibility may point to systematic
errors a/measurement (see below) of certain instruments.
The linearity of a DSC shows the relation between the measured heat flow rate
<Pm (signal) and the true heat flow rate <Ptrue or between the measured heat Qm (or
peak area A) and the true heat Qtrue: <Ptrue = K~· <Pm or Qtrue = KQ . Qm .
The respective proportionality factor K depends on parameters, e. g. the
temperature, the mass of the sample, the heating rate etc.
To determine the linearity means to measure the dependence of the respective
K on parameters and to represent this function graphically, i. e. to show
K~ = <Ptruel <Pm or KQ = Qtru.lQm as a funtion of a parameter. An ideal linearity
means K = const.
184 7 Evaluation of the Perfonnance of a Differential Scanning Calorimeter
Examples:
The temperature dependence K = K (T) is determined by caloric calibration
(cf. Sect. 4.4). In some DSCs, this dependence is linearized electronically or
by means of the pertinent software. This apparent ideal linearity is of course
affected by the full uncertainties of calibration and interpolation.
The dependence of the factor KQ on the mass of the sample (KQ = KQ(m»
is determined by measuring the peak areas related with the melting of, for
example, indium samples of different mass, with the other parameters
unchanged (Fig. 7.2).
Note:
With decreasing mass, the minimum heat of fusion can be assessed which can be
distinguished with reasonable accuracy from the noise and the baseline uncertainty.
The linearity can be tested with regard to, for example,

the scanning rate,
the sample position in the crucible,
the surface-to-volume ratio of the sample.
The time constant 'f of the DSC measuring system is the time interval in which
- after a stepwise change of a constant heat flow rate which is produced at the
place of the sample - the measurement signal reaches the new final value up to
the e-th part (i. e. up to 63.2 %). In other words, if a constant heat flow rate that
temperature ..
Fig.7.2. Linearity (ace. to Hemminger, 1994).
cJ)m measured signal (heat flow rate), AI to A4 peak areas, belonging to sample masses ml to m.
(ml < m2 < m3 < m4 )· In the partial figure: ratio An/mn with the range of uncertainty of the area
determination, ---- deviation from ideal linearity (here only found for very small (j) and
very big@ sample masses)
7.2 Characterization of the Measuring System 185
causes a signal t.Pm • O is switched off, the signal drops down to t.Prn,o/e (36.8 %; i. e.
it decreases by 63.2%) after the time interval 'f (Fig. 7.3). The time constant
determines to what degree two successive thermal events (peaks) are recorded
clearly separated from each other (resolution with time). Knowledge of.the time
constant is important for the desmearing ofDSC curves (see sect. 5.4.).
The time constant can be determined as follows (cf. Ulbrich, Cammenga,
1993; L6blich, 1994):
When an electric calibration heater can be is installed, a constant heat flow
rate, t.Prn,o, is switched on until a constant measurement signal is obtained;
after switching-off of the heater, the descending curve is evaluated according
to Fig. 7.3 or evaluated analytically:
t.P(t) = t.Pm• O • exp (-tl'f)
dt.P t.P
dt 'f
so for any time t* we get

t.P(t *)
'f=---
dt.P
dt
t* should not be too close to the switching-off time because other, smaller
time constants ('f2' 'f3 etc.) may then be superimposed on the maximum time
constant 'fmax (see below).
When no calibration heater can be installed, the descending section of a peak
is evaluated by the "tangent method" when a pure substance is melted (cf.
.............. r - - - - - ,
time
Fig. 7.3. Determination of the time constant T (acc. to Hemminger, 1994).
After a constant measurement signal <Pm,o has been reached, the constant heating current is
switched off at the time to' After the time interval T, the measurement signal <pm.O has dropped
to <Pm,oIe
186 7 Evaluation of the Perfonnance ofa Differential Scanning Calorimeter
1
I
I
I
I \
- -0- ~ -.:q. - --=---
L...y-J
T
time
Fig.7.4. Determination of the time constant 'I" from the melting peak of a pure substance (acc.
to Hemminger, 1994)
Fig. 7.4). Several tangents are plotted to the descending section and the time
constants 'Tt ... 'Tmax are graphically determined from the intersections with the
linearly interpolated baseline. Towards the end ofthe descending curve, a con-
stant value (At between t* and the intersecting tangent baseline) results: the
(greatest) time constant 'Tmax which describes the thermal inertia of the mea-
suring system in good approximation.
From the measured curve, at the moment of the (usually sudden) solidifica-
tion of the (supercooled) melt of a pure substance (procedure as above).
It is sometimes important to know the sensitivity of a DSC measuring system.
This quantity gives the relation between the change of the measurement signal
and the change ofthe measured quantity that creates the signal. In DSC systems
the measured quantity is a change in the heat flow rate LlcP, for example in f..LW,
the signal output is an electric voltage LlU, for example in f..Lv, so the sensitivity
is given in f..LV f..LW-t.
7.3 Characterization of the Results of a Measurement

The quality of measurement results must be evaluated prior to their being used in
calculations (purity, kinetics etc.) and interpreted. This evaluation is made on the
basis ofthe data characterizing efficiency of the measuring system (cf. Sect. 7.2).
Terms describing the quality of measurement results (cf. International Vocabu-
lary of Basic and General Terms in Metrology 1984, 1994):
the accuracy describes how well the result of a measurement approximates the
. "true" value. The accuracy is determined by random (unforeseeable, uncor-
related) and systematic errors of measurement. With the conditions of measure-
7.3 Characterization of the Results of a Measurement 187
ment unchanged, systematic errors are constant; they may comprise known and
unknown components.
Known systematic errors (deviations) are allowed for in the measurement
result as corrections. The uncertainty ofthese corrections, unknown systematic
errors and random errors must be estimated. The standard deviation is an esti-
mate of the random errors.
The total uncertainty of measurement indicates a range of values in which the
true value of the measurand lies with high probability. The total uncertainty is
often referred to as accuracy (more precisely: inaccuracy); it is not identical with
the repeatability or the reproducibility (cf. sect. 7.2).
All the relevant quantities should be carefully evaluated and taken into account
when the uncertainty of a measured value is estimated. Statement of the uncertainty
enables the reader or colleague to judge the rank of the investigations published.
Example (see Fig. 7.5):
uncertainty
of correction
~
Icorrection I
[-=-
~- -
~ random
estima ted
unknown devia tion
systematic
devia tion
corrected :!: to tal uncertaint y

resul t 1= measured (additive or due to statistics)
value
Fig. 7.5. Diagram to describe the relation between the terms "result of a measurement",
"measured value" and "total uncertainty" (cf. text)
188 7 Evaluation of the Performance of a Differential Scanning Calorimeter
Result of a measurement = measured value (e. g. mean value of a series

of measurements) (100 IlW)
+ known systematic error (correction) (-lOllW)
Overall
+ uncertainty of the correction, estimated (± 1 IlW)
{ + unknown systematic error, estimated (± 2 IlW)
uncertainty
+ random error (e.g. twice the standard deviation) (±4IlW)
= 90 IlW ±7 IlW (additive total uncertainty)
(taking no account of the error propagation or the like).
7.4 Check List for DSCs

The following check list serves to collect the essential characteristic data of a DSC.
It may be used to
ask the manufacturers for the values,
establish a guide-line when characteristic data are measured in one's own
laboratory or in the manufacturer's laboratory,
compare different types ofDSCs,
- compare the DSCs data with the values which required to investigate a problem.
Manufacturer:
Type of measuring system: o heat flux disk-type
o heat flux cylinder-type
o power compensation
Special features:
Sample volume (standard crucible): ..... mml
Atmosphere
(vacuum?, which gases? pressure?):
Temperature range: _ from ..... to ..... °C or K
Scanning rates: from ... to ... K min- 1 steps: ... K min- 1
Zero line repeatability: from ± ..... IlW (at ..... 0c)
to ± ..... IlW (at ..... 0c)
Peak area repeatability: ± ..... %(at ..... 0c)
Total uncertainty for heat: ± ..... % (at ..... 0c)
Extrapolated peak onset temperature
repeatability: ....... K(at ..... 0c)
Total uncertainty for temperature: ± ..... K(at. .... 0c)
Scanning noise (Pp) at ..... Klmin: from ±-..... IlW (at ..... 0C)
to ± ..... IlW (at ..... 0c)
Isothermal noise (Pp): from ± ..... IlW (at ..... 0C)
to ± ..... IlW (at ..... 0C)
Time constant with sample: ..... s
Additional facilities:
Appendix 1
Comparison of Heat Flux Differential Scanning Calorimeters

and Differential Thermal Analysis Instruments
The measured signal generated in DTA instruments is the temperature difference tlT
between sample and reference sample. The measuring system comprises tempera-
ture sensors and a suitable holder or support to fix the temperature sensors and the
sample containers mechanically. A defined, constant and good thermal contact must
be established between temperature sensors and sample containers. Strict repeat-
ability of the measuring system over the whole temperature range must be ensured;
the system's properties are reflected in the shape of the zero line (i. e. curve mea-
sured with the measuring system empty or with empty crucibles).
Depending on the desired measurement temperature, sample volume and sen-
sitivity, a choice can be made from among various DTA measuring systems. In the
following, the two contrasting basic types - the block-type measuring system and
the measuring system with free-standing crucibles - will be described.
DTA: The Block-type Measuring System
The following is characteristic ofthis type of measuring system (cf. Fig. A1.I):
The temperature difference tlT is measured directly in the sample substances.
As a result, the temperature sensors are subject to the attack by the sample and
reference sample substance.
The duration (resolution with time) and the height (sensitivity) of the measured
signal are determined by the thermal resistance between sample substance and
block, sample substance and temperature sensor and by the properties of the
block material.
The duration and height of the measurement signal and as its shape also depend
on the location of the temperature sensor in the sample substance and on the
(changing) thermophysical properties of the sample substance. A quantitative
evaluation of the heat in question is therefore impossible.
The generally high sensitivity and the rapid, almost instantaneous "response" to
sample reactions are advantages ofthis measuring system.
190 Appendix 1
6.T
3
s R
6.T
b time
Fig. A 1.1. a Instrument for Differential Thermal Analysis (DTA) with block-type measuring
system,
b Schematic measured curve t!.T(t) (exothermic transition).
I "block" with cavities to take up sample and reference sample substance, 2 furnace, 3 dif-
ferential thermocouple, 4 programmer and contro\ler, S sample substance, R reference
sample substance, TF furnace temperature .
Appendix 1 191
DTA: Measuring System with Free-standing Crucibles

This measuring system (Fig. A 1.2) is more frequently used in DTA instruments than
the block-type measuring system. A characteristic feature is that free-standing
sample crucibles are put over the protective tubes of the thermocouples for the
purpose of I:1T measurements. The thermocouple junction is in thermal contact with
the bottom of the crucible.
The following is characteristic of this type of measuring system:
- The thermophysical properties of the sample substance affect the height and
shape of the measured curve only slightly. The peak shape is, to a large extent,
2
4
~(t)
3
a '----0 I:::. T
time
r
I:::.T
b
Fig. A 1.2. a Instrument for Differential Thermal Analysis (DTA), measuring system with
free-standing crucibles,
b Schematic measured curve AT(t) (endothermic transition).
1 crucible, 2 furnace, 3 differential thermocouple, 4 programmer and controller, S sample
substance, R reference sample substance, TF furnace temperature
192 Appendix 1
determined by the containers, this allows a semi-quantitative evaluation of the

heat in question from the peak area.
- Depending on the problems to be solved and the mass and structure of the sam-
ples, a choice can be made from among a great number of container types (cru-
cibles), among them "poly-plate" crucibles, micro-crucibles worked out from the
thermocouple bead, crucibles for "self-generated atmosphere" and others.
- The high flexibility as far as the usability of different furnaces is concerned (tem-
perature ranges), the adjustment of different gas atmospheres (including vacuum)
and the usability in instruments for simultaneous thermal analysis.
In other types of DTA instruments, sleeves containing the sample substance are
placed on the thermocouple junctions which are either unprotected or covered by
thin tubes. With this the attempt is made to have the advantages offered by direct,
undelayed temperature measurement, without having to put up with the unfavour-
able effects of the block material (resolution, sensitivity) or the (isolated) individual
crucibles (delayed, "smeared" temperature measurement).
Heat Flux DSC
In contrast to the majority of the DTA instruments of rather simple design, DSCs
should be capable of being calibrated for heat measurement. An unambiguous and
repeatable assignment of the measurement signal proper, IlT, to the measurand IltP
must be ensured (IltP= C'l\-s - tPn, differential heat flow rate from the furnace to the
samples). The DSC should be suitable for the investigation of different substances
and transitions under different test conditions. This is why the properties of the mea-
suring system may depend as little as possible on the sample properties, the type of
transition and on test parameters (e. g. heating rate).
Within certain limits, these requirements can be fulfilled in good approximation
when the temperature field of the measuring system does not primarily depend on
sample properties but on the properties of the measuring system itself which are
given by its design.
This dominance of the measuring system, which is essential for DSCs, follows
from the fact that - compared with the sample - the measuring system is made up
of solid parts and has defined heat conduction paths. The sample reaction can then
be regarded as a small disturbance of the steady-state temperature field established
during heating. As this disturbance must be small, the measured signal itself, IlT,
will also be small. As a consequence, the noise ofthe whole measuring set-up must
be low to allow the sensitive measurement of heat flow rates.
In all DSCs the calibration factor depends on instrument and sample parameters.
This leads to systematic error sources, and the uncertainty ofthe result is, therefore,
much larger than the repeatability error of the measurements. Very careful calibra-
tion, with the sample parameters adjusted, is, therefore, necessary.
Appendix 2
Calorimetry - a Synopsis
A great variety of calorimeters serve to measure heats and heat capacities in various
fields of application.
In the following, a classification system for calorimeters and a couple of ex-
amples of different types of calorimeters will be presented.
The aim is to give a structured survey of the whole field of calorimetry which
may help to better recognize and evaluate the advantages of and limitations to the
DSCs which result from their mode of operation.
Classification
In a classification system, calorimeters are arranged in groups according to par-

ticular characteristics. Various classification systems are reasonable and practicable
(cf. Hemminger, Hohne, 1984; Rouquerol, Zielenkiewicz, 1986). It may sometimes
be difficult to arrange calorimeters in proper order in a relatively simple system, and
such a classification may even be impossible. Classification systems covering the
entire field of calorimetry tend to become very sophisticated, an end in itself and
rather useless for practical applications. When thermodynamical principles of heat
exchange or the aspect of how the caloric signal is formed are taken as a basis for
classification, the result will always be that a certain number of calorimeters are
characterized satisfactorily well whereas additional explanations or auxiliary
definitions are required for others. Nevertheless, a classification system is useful to
show basic principles of calorimetry. It should be based on existing instruments and
open to future developments. In the following, a simple system for practical use will
be developed which will suffice to discuss characteristic features and error sources
by groups.
Criteria for the classification are the following:
1. The principle ofmeasurement
1.1 Measurement of the energy required for compensating the heat to be mea-
sured (heat-compensating calorimeters).
1.2 Measurement of the temperature change of a substance due to the heat to be
measured (heat-accumulating calorimeters).
1.3 Measurement of the heat flow rate between sample and surroundings due to
the heat to be measured (heat-exchanging calorimeters).
194 Appendix 2
2. Mode ofoperation
2.1 isothernlal )
2.2 isoperibol static modes
2.3 adiabatic
2.4 scanning of surroundings )
2.5 isoperibol scanning dynamic modes
2.6 adiabatic scanning
3. Construction principle
3.1 Single calorimeter
3.2 Twin or differential calorimeter
Most ofthe calorimeters can be classified by the above-mentioned criteria. All com-
binations of the criteria 1, 2 and 3 are, of course, not possible as some ofthe criteria
are incompatible.
Calorimeters which are used today will be presented in the following according
to this classification. As the classic calorimeters will be dealt with only briefly,
reference will be made to the literature for further infornlation on construction
details, error sources, methods for the evaluation of the measured values, and
ranges of application of these instruments. The DSCs, however, are treated in
closer detail in several sections ofthis book (see chapters 2 and 3).
Examples of Calorimeters
The aim of this section is to give the reader an idea of the variety of calorimeters
offered in addition to DSCs. This will help to better classify the range and poten-
tialities of the DSC methods in comparison with different calorimetric methods.
Heat-compensating Calorimeters
In the case of this calorimetric method, the effect ofthe heat to be measured is "sup-
pressed", i. e. temperature changes of the sample or of the calorimeter's measuring
system, or temperature differences in the measuring system due to the caloric effect
are compensated. For this purpose, an equally high, well-known amount of energy,
with the sign reversed, is added.
Possibilities: Compensation ofthe heat to be measured with the aid of the "latent
heat" of a phase transition (e.g. ice calorimeter) or with electric energy (Joule's heat
or Peltier's cold). The compensation by means of reversible expansion or compres-
sion of an ideal gas was described by Ter Minassian, Milliou, 1983.
It is an advantage of all compensating methods that the measurements are car-
ried out under quasi-isothernlal conditions and that heat leaks do not, therefore,
represent important error sources. Moreover, in the case of electric compensation,
no calibrated temperature sensor is required for the measurement but only a sensi-
tive thernlometer which controls the compensation power of a controller so that the
temperature remains constant.
Appendix 2 195
1st example: "Ice calorimeter"
I. Compensation of the heat to be measured by latent heat

2. Isothermal
3. Single calorimeter
A warm sample placed in an ice calorimeter (Fig. A 2.1) transfers its heat to a O°C
ice jacket. As a result, a certain mass of ice (to be determined) melts. In the case of
the ice calorimeter according to Lavoisier, de Laplace, 1784, the melted water was
weighed, whereas Bunsen, 1870, determined the change of the ice/water ratio on the
basis of the change in the volume of the whole mixture (density difference of ice
and water). The measurements which Ginnings and Corruccini carried out at the
National Bureau of Standards (NBS) during the late forties were counted among the
important applications of an ice calorimeter. They measured heat with an uncer-
tainty of about 0.02 %, the temperature of the sample at the moment of its being
dropped into the calorimeter lying between 100°C and 600 °C.
The liquid to gaseous phase transition was also made use of ("boil-off" calo-
rimeter), in particular since in this case the difference between the density of both
phases - and thus the sensitivity - is by two or three orders of magnitude higher than
in solid to liquid transitions.
Phase transition calorimeters are relatively simple to construct and allow precise
measurements to be performed. A disadvantage is that measurements can be carried
out only at one temperature, i. e. at the temperature of transition of the respective
calorimeter substance.
The quantity to be determined is the transformed mass of calorimeter substance
(ice, liquid); the heat of transition must be known. (In general, the calorimeter is
calibrated electically.)
2nd example: "Measurement of heats of solution"
I. Compensation of the heat to be measured with the aid of electric energy

2. Isothermal
A container filled with water is equipped with a stirrer, a controllable electric

heater and a sensitive thermometer (Fig. A 2.2). At constant temperature, an endo-
thermically solving salt is added. The heater is adjusted so that the temperature
of the liquid remains constant. The supplied electric energy is then equal to the heat
of solution of the salt (Bronsted, 1906).
Expressed more generally, the following is valid:
In calorimeters of this type, the heat to be measured is compensated by Joule's heat
or with the aid of the Peltier effect. This is done in that a sensitive thermometer
activates the compensation control circuit so that, if possible, no temperature
change due to reaction heat takes place. As the Peltier power for the compensation
196 Appendix 2
f-----{> X
6 7
o
o
o
o
x
t
b time
Fig. A 2.1. a "Ice calorimeter" (Bunsen, 1870),
b Measured curve (acc. to Hemminger, 1994).
I sample, 2 sample container (receiver), 3 ice, 4 water,S ice-water mixture, 6 mercury, 7 capil-
lary tube, x position of the mercury meniscus.
The displacement Ax· of the meniscus is proportional to the heat Q exchanged with the
sample (positive for an endothermic effect). Ax· is determined taking the "preperiod" and the
''postperiod'' into account
Appendix 2 197
v
5
2
t
~I
b time
Fig. A 2.2. a "Compensation calorimeter" (Bronsted, 1906),
1 sample (salt), 2 water, 3 stirrer, 4 electric heater, 5 temperature sensor, 6 controller, P.l elec-
tric heating power.
The area below the measured curve P.l(t) corresponds to the compensation heat and (at con-
stant temperature) also to the endothermic heat of solution Q of the sample substance, which
has been searched:
198 Appendix 2
of exothermic effects cannot be measured with sufficient accuracy, the calorimeter

substance (e.g. water) is generally cooled with constant Peltier power and at the
same time heated with equally high Joule's heat (controlled). To compensate an
endothermic effect, the heating power is increased in order to keep the temperature
of the calorimeter substance constant. When an exothermic effect is to be compen-
sated, the heating power is decreased with the Peltier power remaining unchanged
(see for example Christensen et ai., 1968). These calorimeters do not attain the
strictly isothermal state of phase transition calorimeters, as the difference between
actual and set temperature value must be non-zero in order that the electrical power
control is activated. In addition, temperature control is delayed by the heat transfer
processes, and it is very difficult to obtain spatially homogeneous temperature
fields by electric heating.
The advantages ofthese quasi-isothermal calorimeters consist in the simple and
very precise measurement of the electric compensation energy and the possibility
of using highly sensitive sensors to measure temperature changes; these sensors
must not, however, be calibrated.
This is done in that a sensitive thermometer activates the compensation control
circuit so that, if possible, no temperature change due to reaction heat takes plase.
3rd example: "Adiabatic scanning calorimeter"
1. Compensation of the heat to be measured with the aid of electric energy

2. Adiabatic scanning
In these calorimeters (Fig. A 2.3), the temperature program is preset. So much elec-
trical heating power is supplied to the sample as is necessary to comply with the
given temperature program. (In practical application, the electrical heating power
required for sample heating is often preset and the resulting heating rate measured.)
Heat losses are minimized by adapting the temperature of the surroundings as well
as possible to the temperature of the sample (or sample container) (adiabacy). Calo-
rimeters of this type allow the heat capacity to be measured with high accuracy
(uncertainty SO.1 %) (cf. 4th example of heat-accumulating calorimeters).
The following is valid for the heat capacity C (T):
P. -p.
C (T) = 01,1 el.2
(dT/dt: preset)
dT
dt
or
Appendix 2 199
5
T
11 T = 0 6
a
t t,
CD
P.I
b
--- time ..
CD
Fig. A 2.3. a "Adiabatic scanning calorimeter".

1 sample, 2 sample furnace, 3 heatable (adiabatic) shield, 4 temperature difference sensor,
S temperature sensor, 6 programmer and controller, Pel electrical heating power. When there is
a variation of the temperature with time T(I), Pel(t) is controlled so that (dT/dt)1 = (dT/dl)z
(outside the peak),
b Measured curve (acc. to Hemminger, 1994). <D curve measured with the calorimeter empty,
<i) curve measured with the sample placed in the calorimeter, sample transition between 11 and
12 (peak)
200 Appendix 2
<D run with the calorimeter empty

@ run with the sample placed in the calorimeter
Pel=i' U electrical heating power

dTldt heating rate at the respective moment
C(T) comprises the heat capacities of sample and container. The heat
capacity of the empty container is determined by separate measure-
ment.
When in a 1st order transition (e.g. melting of a pure metal) the sample temperature
remains constant during transition in spite of the fact that heating power is con-
tinued to be supplied, the heat of transition is directly determined from the integral
of the heating power over the transition time (Qtrs = Ji (t) . U (t) dt).
The adiabatic shield guarantees quasi-isothermal conditions while the heat to be
determined is compensated with the aid of electric energy.
Calorimeters ofthis type are used for the accurate, absolute and direct measure-
ment of heat capacities and heats of transition (cf. for example Nolting, 1985;
Kagan, 1984; an extension to a low-temperature tum system has been described by
Rahm, Gmelin, 1992).
4th example: "Power compensation DSC"
1. Compensation of the heat to be measured with the aid of electric energy

2. Isoperibol scanning
3. Twin calorimeter
The temperature of the sample surroundings remains constant (isoperibol). The

calorimeter (Figs. 2.4,2.5) comprises two identical measuring systems (twin prin-
ciple), one containing the sample, the other the reference sample. The temperature
difference between the two systems is measured. In 1st approximation, disturban-
ces from the surroundings have the same effect on both measuring systems and
therefore cancel out with respect to the temperature difference. The individual sam-
ple supports (microfurnaces) are heated separately so that they comply with the
given temperature-time program. When there is ideal thermal symmetry between
the two measuring systems, the same heating power is required for sample and ref-
erence sample. When additional heat is released or consumed during sample tran-
sition (exothermic or endothermic process), the sample's heating power is regulated
by means of a proportional controller so that the electric heat supplied is decreased
or increased by just the amount as has been generated or consumed during the exo-
thermic or endothermic transition process. The measured signal is the temperature
difference ll.T (deviation from the set value) to. which the compensation heating
power ll.P is proportional: ll.P - ll.T. Calorimeters ofthis type ("power compensation
DSC" or "differential power compensation scanning calorimeter, DPSC" or "dy-
namic power difference calorimeter") are widely used (cf. for example Watson et
aI., 1964; Hemminger, Hohne, 1984). They are discussed at full length in Sects. 2.2
and 3.2 of this book.
Appendix 2 20 I
Heat-accumulating Calorimeters
In the case of this calorimetric method, the effect of the heat to be measured is not
"suppressed" by compensation but leads to a temperature change in the sample sub-
stance and a "calorimeter substance" with which the heat to be determined is
exchanged. This temperature change is measured. When the change is not too great,
it is proportional to the amount of heat exchanged. The proportionality factor must
be determined by calibration with a known amount of heat.
5th example: "Drop calorimeter"
I. Measurement of the temperature change of a substance due to the heat to be measured

2. Isoperibol
The temperature of the surroundings is kept constant with the aid of a thermostat
("isoperibol": uniform surroundings). The heat Q to be measured is exchanged
with the "calorimeter substance" and the temperature change I:1T(t) is measured
(Fig. A 2.4).
The following is valid: Q = Ceal • I:1T* (cfFig. A 2.4c).
The proportionality factor Ccal is the heat capacity of the calorimeter substance
(a liquid in Fig. A 2.4a) plus that of the other calorimeter components (stirrer,
thermometer, etc.), which cannot be exactly defined. This factor is determined by
calibration with electric energy (Joule's heat). As soon as there is a temperature
difference between calorimeter substance and surroundings, heat is exchanged.
This exchange must depend only on this temperature difference and must be
reduced as far as possible by appropriate measures (Dewar vessel, radiation shields
etc.). Otherwise, it cannot be determined by calibration and represents an error
source. In order to guarantee strict repeatability of the heat exchange, I:1T must
amount to only a few Kelvin. Since the temperature change of the calorimeter
substance and the unavoidable heat exchange with the surroundings take place
simultaneously, the temperature difference I:1T* used to calculate Q must be deter-
mined according to defined rules (e.g. International Standard ISO 1928-1976 (E);
Rossini, 1956; Gunn, 1971; Oetting, 1970; Sunner, Mansson, 1979) from the
shape of the I:1T (t) curve measured before and after the sample has been placed
into the calorimeter (cf. also 6th example). Simple drop calorimeters serve to
measure mean heat capacities (temperature of the sample at the moment of its
being placed into the calorimeter: up to about 500°C; sample mass between 10 and
100 g).
In "aneroid" drop calorimeters (Fig. A 2.4 b), the calorimeter substance is a solid
body of good thermal conductivity. The advantage over calorimeters filled with
liquid consists in that samples at high temperature (up to about 2000°C) can be
dropped without the risk of evaporation or splashing.
Other examples of calorimeters of this class are instruments in which the sample
placed into it (solid body, liquid, gas) reacts with the calorimeter substance; this
results in heat of reaction being released.
202 Appendix 2
3 3
4
T
1ijm
T
~c::J
~
_L_
a b
1
T
c
Fig. A 2.4. "Drop calorimeter" a with liquid, b with aneroid calorimeter substance (in iso-
peribol mode of operation) and c measured curve (ace. to Hemminger, 1994).
1 sample, 2 calorimeter substance: (a) liquid, (b) solid, 3 temperature sensor, 4 stirrer, 5 radia-
tion shields.
The temperature change AT· ofthe calorimeter substance has to be determined from the mea-
sured curve according to defined rules, taking the "preperiod" and the "postperiod" into
account
Appendix 2 203
6 8
T-
4 LC;5l..\:::::::J- - - - -
3
1
2
5
b time
Fig. A 2.S. a Adiabatic "bomb calorimeter",
1 sample in combustion pan, 2 vessel, 3 ignition device (electrodes with heating filament),
4 calorimeter substance (water), 5 adiabatic jacket, 6 stirrer, 7 temperature difference sensor,
8 temperature sensor, 9 controller
204 Appendix 2
6th example: "Adiabatic bomb calorimeter"
I. Measurement of the temperature change of a substance due to the heat to be measured

2. Adiabatic
In the case of the adiabatic bomb calorimeter (Fig. A 2.5), the "combustion bomb"
which is (usually) filled with oxygen at high pressure immerses in water (calo-
rimeter substance). The water temperature is continually measured before and after
the electric ignition. With the aid of a controller, the temperature of the surround-
ings is always adapted to this temperature (adiabatic jacket). The temperature
change with time in the surroundings serves as measurement signal; it is deter-
mined from the drift of the measured curve before the ignition (preperiod) and after
the ignition (postperiod) (as with the "drop calorimeter", see 5th example).
Bomb calorimeters of this type (usually automated) are widely used to measure
the calorific value of solids or liquids under standardized conditions. In general,
they are calibrated with benzoic acid. Their uncertainty of measurement lies in the
per mil range. (There are also "dry" bomb calorimeters, in which the temperature
change of the combustion vessel itself (of the bomb) is measured; bomb calorim-
eters with isoperibol surroundings are also used.) For a detailed representation of
bomb calorimetry, see Rossini, 1956; Skinner, 1962; Sunner, Mansson, 1979.
7th example: "Flow calorimeter"
1. Measurement of the temperature change of a substance due to the heat to be measured

2. Isoperibol
3. Single calorimeter (also designed as "twin")
In the case of the so-called "gas calorimeters", the heat to be measured is trans-
ferred, if possible completely, to a flowing medium (Fig. A 2.6). The temperature
difference between the medium flowing in and the medium flowing out is propor-
tional to the heat transferred. Calorimeters of this type are used to determine the
calorific value offuel gases; they are calibrated with gases of known calorific value
(e.g. methane) so that the specific heat capacity of the heat-conveying medium
must not be known. If an electric heater is used instead of the burner, the specific
heat capacity of the heat-conveying medium can in principle be measured as well
(cf. Hemminger, 1988).
In biology, biochemistry and chemistry, flow calorimeters with liquids serve to
measure the heat development of microorganisms in certain nutrient solutions, or
they are used to measure reaction heats. Two reacting solutions are, for example,
mixed in a reaction tube (Fig. A 2.7). The uniform temperature of the reacting
agents is measured before they are mixed and then at a point at which the reaction
in the flowing liquid has come to an end. The temperature difference is a measure of
the reaction heat. The calorimeter must be calibrated, either with the aid of known
reaction heats of liquids or with an electric heater installed in the reaction tube.
Appendix 2 205
Fig. A 2.6. "Gas calorimeter".

I burner in which the gas to be measured is burnt, 2 heat exchanger, 3 "heat conveying
medium" (e. g. air or water), 4 combustion gases, 5 temperature sensor.
Inside the heat exchanger, the hot combustion gases convey their heat to the heat conveying
medium whose temperature increase T2 - TI is measured. The "calorific value" (the com-
bustion heat) is proportional to T2 - TI
A sophisticated example of a flow calorimeter with stimulated cardiac muscle

developing heat in the perfused tube has been described by Daut et al., 1991. (Flow
calorimeters are also designed as twin calorimeters to cancel out the influence of
the isoperibol surroundings (heat leaks)).
8th example: "Adiabatic calorimeter"
I. Measurement of the temperature change ofa substance due to the supply ofa known
amount of heat
2. Adiabatic
This type of calorimeter (Fig. A 2.8) is not designed to measure an unknown heat;
instead, a well-known, electrically generated heat Q = J¥.I serves to change the
sample temperature by tJ.T. The temperature ofthe surroundings (adiabatic jacket)
is adapted to the measurement temperature with high accuracy in order to avoid any
heat exchange with the surroundings. Calorimeters of this type are used to deter-
mine the phenomenological coefficient of the heat supplied, Q, and the temperatu-
re change, tJ.T, ofa substance (at constant pressure): the heat capacity Cp(T). The
following is valid:
206 Appendix 2
t
j
I
enter of
: ----- react and
b time
Fig. A 2.7. a "Flow-mix calorimeter",
b Measured curve.
I, 2 reactants, 3 reaction product, 4 temperature sensor.
As in the case of the "drop calorimeter", the temperature change ilT'" required to determine
the mixing or reaction heat is obtained from the pre- and postperiod (affected by a drift) ofthe
measured curve (T2 - Tt)(t)
Appendix 2 207
2
3
(SWitch off P"
t
T
: _____ switch on P"

I
b time
Fig. A 2.8. a "Adiabatic calorimeter",
b Measured curve (ace. to Hemminger, 1994).
1 sample, 2 heatable sample container, 3 adiabatic jacket, 4 controller, 5 programmer and con-
troller, Pel electric heating power.
When the constant measurement temperature T is adjusted, the electric heating energy w.,is
supplied to the sample, increasing the sample temperature by aT*
208 Appendix 2
Calorimeters of this type allow the specific heat capacity and latent heats to be
measured with the greatest possible accuracy (Gmelin, ROdhammer, 1981; lakobi
et aI., 1993).
Knowledge of the heat capacity and its temperature dependence is of utmost
importance in solid state physics and thermodynamics. In practical application, llT
is kept as small as possible in order to determine the temperature dependence of Cp
as precisely as possible and to avoid errors due to inhomogeneous temperature
fields. The limits are determined by the noise and the uncertainty of measurement
of the temperature sensors (cf. Kagan, 1984; Zhiying, 1986).
Heat-exchanging Calorimeters
In calorimeters which measure the temperature change of the calorimeter sub-

stance, i. e. heat-accumulating calorimeters (e. g. drop calorimeters, bomb calorim-
eters), the heat exchange with the isoperibol surroundings is kept low to make the
measured signal llT as great as possible. In the calorimeters referred to in this
section, a defined exchange of the heat to be measured with the surroundings is
deliberately aimed at, the reason for this being that the measured signal which
describes the intensity of the exchange is then proportional to a heat flow rate cPand
not to a heat. This allows time dependences of a transition to be observed on the
basis ofthe cP(t) curve (the power compensation DSC also offers this possibility).
The twin design allows disturbances from the surroundings, which affect both
systems in the same way, to be eliminated by taking only the difference between the
individual measurement signals into account (Differential Scanning Calorimeter:
DSC).
9th Example: "Heat flux differential scanning calorimeter"
1. Measurement of the exchange of the heat to be measured between sample and surroundings
via a heat flow rate
2. Scanning of surroundings
3. Twin calorimeter
DSC with Disk-type Measuring System (ef Sect. 2.1.1)
A metal, ceramic or quartz glass disk with the sample and the reference sample
(or the pans) positioned on it symmetrical to the centre, is placed into a furnace
(Fig. 2.1 a). Heat exchange between furnace ("surroundings") and samples takes
place by heat conduction, radiation and convection. Strict repeatability of this
heat exchange as a function of the temperature (with the atmosphere remaining
unchanged) must be ensured. This is why a solid heat-conducting disk is used which
guarantees that the properties of the measuring system dominate. As a result, the
different characteristics of the individual samples contribute less strongly to the
kind of heat exchange than in the case of the DTA (the measurement signal itself
Appendix 2 209
must, of course, reflect the sample properties). The signal tlT is measured on the
solid heat conductor (disk) between the supports for sample and reference sample
(tlT= Ts - Tp).
DSCwith Cylinder-type Measuring System (ef. Sect. 2.1.2)
In another type of heat flux DSC, the two cylindrical containers for sample and
reference sample are connected with the furnace ("surroundings") by one thermo-
pile each (Fig. 2.2). The heat from the furnace to the samples preferably flows
through the thermocouple wires which are at the same time the dominant heat
conduction paths and the temperature difference sensors. When a differential
connection is provided between the outputs of both thermopiles, the measured
signal (tlT) is proportional to the difference between the heat flow rates from the
furnace to the sample (tlTFS - ~s) and from the furnace to the reference sample
(tlTFR - <PFR)·
Survey of the classification of calorimeters
PRINCIPLE Of MODE OF CONSTRUCTION

MEASUREMENT OPERATION
Measurement of the energy

required to compensate the
heat to be measured
(heat-compensating
calorimeters) Static:
isothermal
Measurement of the isoperibol Single
temperature change of the adiabatic calorimeters
calorimeter substance
measured (heat-
accumulating calorimeters) Dynamic: Twin
scanning of (differential)
Measurement of the heat surroundings, isoperibol calorimeters
flow rate between scanning adiabatic
calorimeter substance and scanning
surroundings due to the heat
to be measured (heat-
exchanging calorimeters)
References
Agrawal AJ (1989) Pol Sci: Part B, Pol Phys 27: 1449

Aras L, Richardson MJ (1989) Polymer 30:2246
ASTM E 967-83 (1984) Standard Practice for Temperature Calibration of Differential Scan-
ning Calorimeters and Differential Thermal Analyzers; Annual Book of ASTM Standards,
vol 14.02:815
ASTM E 794-85 (1986) Standard Test Method for Melting and Crystallization Temperatures
by Thermal Analysis; Annual Book of ASTM Standards, vol 14.02: 703 -707
Aston JG, Fink HL, Schumann SC (1943) J Amer Chern Soc 65 : 341-346
Bader RG, Schawe IEK, Hohne GWH (1993) Thermochim Acta 229: 85
Barton 1M (1980) Polymer 21 :603-606
Barton 1M (1985) In: Dusek K (ed) Advances in Polymer Sciences, Springer, Berlin, Heidel-
berg, New York: 111-154
Batzer H, Kreibich UT (1982) Angew Makromol Chern 105: 113 - 130
Becker R (1976) Z phys Chern 257:667-677
Becker R (1977) Z phys Chern 258:953-966
Becker R (1978) Faserforschung und Textiltechnik 29: 361- 385
Beezer AE (1980) (cd) Biological microcalorimetry. Academic Press, London
Beford RE, Bonnier G, Maas H, Pavese F (1984) Metrologia 20: 145 -155
Blankenhorn K, Hohne GWH (1991) Thermochim Acta 187:219-224
Borchardt HJ, Daniels F (1957) J Am Chern Soc 79:41-46
Bouwstra JA, de LeevvVV, van Miltenberg JC (1985) J Chern Thermodyn 17:685-695
Bracewell RM (1965) The Fourier transform and its applications. McGraw-Hill, New York
Braun D, Kohl PR, Hellmann GP (1988) Makromol Chern 189: 1671-1679
Brekner MJ, Schneider HA, Cantow HJ (1988) Polymer 29:78-85
Breuer K-H, Eysel W (1982) Thermochim Acta 57: 317 - 329
Brezesinski G, Dorfier HD (1983) Colloid & Polymer 261 :423-426
Bronsted IN (1906) Z Phys Chern 56:663-665
Bros JP (1989) J Less Common Metals 154:9-30
Brown WE, Dollimore D, Gallway AK (1980) In: Bamford CH, Tipper CF (eds) Comprehen-
sive chemical kinetics, vol 2, Reactions in the solid state. Elsevier, Amsterdam
Bruzzone G (1985) Thermochim Acta 96: 239 - 258
BunsenRW(1870)AnnPhys 141:1-14
Calvet E (1948) CR hebd 226: 1702-1704
Cammenga HK, Eysel W, Gmelin E, Hemminger W, Hohne GWH, Sarge SM (1993)
Thermochim Acta 219: 333 - 342
Chang SS (1988) Polym Commun 29:33-35
Chee KK (1985) Eur Polym J 21 : 29 - 31
Chekhovskoi VYa (1984) In: Maglic KD, Cezairliyan A, Peletsky VE (eds) Compendium of
Thermophysical Property Measurement Methods I. Plenum, New York
Christensen JJ, Johnston HD, Izatt RM (1968) Rev Sci Instrum 39: 1356-1359
212 References
ChungTS (1984) J Appl Pol Sci 29:4403-4406

Couchman PR, Karasz FE (1978) Macromolecules 11 : 117 - 119
Cowie JMG, Ferguson R (1986) Polym Commun 27:258-260
CRC Handbook of Chemistry and Physics (1990 - 1991) 71 8• edn, CRC Press
Daut J, GroBT, Elzinga G (1991) Thermochim Acta 193 :269-280
de Bolt MA, EastealAJ, Macedo PB, Moynihan CT (1976) JAm Chern Soc 59: 16-21
Doelman A, Gregges AR, Barrall EM (1977) In: Porter RS, Johnson JF (eds) Analytical Calo-
rimetry, vol 4. Plenum, New York
Domen SR (1987) In: Kase KR, Bjiirngard BE, Attix FH (eds) The dosimetry of ionizing
radiation, vol II. Academic Press, Orlando, Florida
Doyle CD (1961) J Appl Pol Sci 5 :285-292
Duncan JC, van de WeIde JG, Wetton RE (1991) Kautschuk und Gummi 44:768-770
Eligehausen S, Sarge SM, Ohlschliiger G, Cammenga HK (1989) JTherm Anal 35 : 515-526
Ellerstein SM (1968) In: Porter RS, Johnson JF (eds) Analytical Calorimetry, Vol 1. Plenum,
New York
EN 45000 Series (European Standards)
EN 45001 (European Standards) General criteria for the operation of testing laboratories
Fan M, Muller F, Wilsmann W (1993) ThermochimActa 224: 19-32
Flammersheim HJ (1981) Wiss Beitr, Friedrich Schiller Universitiit Jena, 160-164
Flammersheim HJ (1988) J Therm Anal 33 : 55-65
Flammersheim HJ, Klemm E (1985) Acta Polymerica 36:443-446
Flammersheim HJ, Eckardt N, Kunze W (1991) Thermochim Acta 187: 269 - 274
Flammersheim HJ, Eckardt N, Opfermann J (1993) Thermochim Acta 229: 281- 287
Flynn m (1974) Thermochim Acta 8:69-81
Flynn m (1990) Therm Anal 36: 1579-1593
Flynn m (1993) Thermochim Acta 217: 129-149
Fox TG, Flory PJ (1950) J Appl Phys 21:581-590
Freeman ES, Carroll B (1958) J Phys Chem 62:394-397
Gilmour IW, Hay IN (1977) Polymer 18:281-285
Ginnings DC, Corruccini RJ (1947) Res Natl Bur Stand 38:538-591
Gmelin E (1987) Thermochim Acta 110: 183-208
Gmelin E, Radhammer P (1981) J Phys E: Sci Instrum 14: 223 - 228
Godovsky YuK, Hahne GWH (eds) (1994) Special Issue: Deformation Calorimetry of
Polymers; Thermochim Acta 247, 1: 1- 128
Gray AP (1968) In: Porter RS, Johnson JF (eds) Analytical Calorimetry, Vol1 Plenum Press
Grewer T (1987) Thermochim Acta 119: 1- 15
GrewerT, Steinbach J (eds) (1993) Special Issue: Safety Techniques; Thermochim Acta 225,
2: 143-301
Gr"nvold F (1967) Acta Chem Scand 21: 1695-1715
Grenvold F (1993) J Chern Thermodynamics 25: 1133-1144
Gurm SR (1971) J Chern Thermodynamics 3: 19-34
Hale A, Macosko C, Bair HE (1991) Macromolecules 24: 2610- 2621
Hanitzsch E (1991) .Thermochim Acta 187: 275 - 281
Hay IN (1992) Progr Colloid Polym Sci 87: 74-77
Hemminger W (ed) (1988) PTB Bericht PTB-W-37, Wirtschaftsverlag, Bremerhaven
Hemminger W (1994) Chapter 2 "Calorimetric Methods" in Calorimetry and Thermal Anal-
ysis of Polymers (Mathot VBF, ed), Hanser, Munich
Hemminger W, SchOnbom K-H (1982) In: Miller B (ed) Thermal Analysis, Proc 7 th ICTA, Vol
1, Wiley, New York 156-162
Hemminger W, Hahne GWH (1984) Calorimetry-Fundamentals and Practice; VCH Verlag,
Weinheim
References 213
Hemminger WF, Cammenga HK (1989) Methoden der Thermischen Analyse. Springer,

Berlin, Heidelberg, New York
Hemminger W, Raetz K (1989) PTB-Mitteilungen 99: 83 - 88
HemmingerW, Sarge StM (1991) J Therm Anal 37: 1455-1477
Hentze G (1984) Thermochim Acta 72: 127-138
Hentze G, Krien G (1986) Thermochim Acta 107: 61-74
Heublein G, Hallpap P, Hauptmann S, Mann G (1984) Einfiihrung in die Reaktionstheorie.
VEB Deutscher Verlag fUr Grundstoffindustrie, Leipzig
HoffH (1991) Thermochim Acta 187:293-307
Hohne GWH (1983) Thermochim Acta 69: 175 -197
Hohne GWH, GlogglerE (1989) Thermochim Acta 151:295-304
Hohne GWH, Cammenga HK, Eysel W, Gmelin E, Hemminger W (1990) Thermochim Acta
160:1-12
Hohne GWH, Schawe JEK (1993) Thermochim Acta 229:27
Hohne GWH, Schawe JEK, Schick C (1993) Thermochim Acta 221: 129-137
Horie K, Hiura H, Sawada M, Mita I, Kambe H (1970) J Pol Sci Al 8: 1353-1372
Huguenin FGAE, Klein MT (1985) Ind Eng Chern, Prod Res Dev 24: 166- 171
Hutchinson JM (1990) J Pol Sci, Part B, Pol Phys 28: 2127 - 2161
Hutchinson JM (1992) Progr Colloid Polym Sci 87: 69-73
Hutchinson JM, Ruddy M (1988) J Polym Sci, Part B, Pol Phys 26, 2341- 2366
Hutchinson JM, Ruddy M, Wilson MR (1988) Polymer 29: 152 - 159
Hyde CG, Jones MW (1960) Gas Calorimetry, 2 nd ed; Ernest Benn, London
International Vocabulary of Basic and General Terms in Metrology (1 81 ed 1984, 2nd ed 1994)
Beuth, Berlin
IPTS-68 (1976) The International Practical Temperature Scale of 1968, Amended Edition of
1975; Metrologia 12:7-17
ITS-90: The International Temperature Scale of 1990 (1990) Metrologia 27: 3 - 10
Jakobi R, Gmelin E, Ripka K (1993) J Therm Anal 40; Proc 10 th ICTA (Morgan OJ, ed),
871-876, John Wiley, Chichester and Akademiai Kiad6, Budapest
Jin Y, Wunderling B (1993) Thermochim Acta 226: 155-161
Jorda R, Wilkes GL (1988) Polym Bull 20:479-485
Jung DH, Known TW, Bae OJ, Moon IK, JeongYH (1992) Sci Technol3 :475
Kagan ON (1984) In: Maglic KD, Cezairliyan A, Peletsky VE (eds) Compendium of
Thermophysical Property Measurement Methods I. Plenum, New York
Kalachandra S, Turner DT (1987) J Pol Sci, Part B, Pol Phys 25 : 1971- 1979
Keenan MR (1987) J Appl Pol Sci 33: 1725-1734
Kemp RB (1993) Thermochim Acta 219: 17-41
Kilian HG, Hohne GWH (1983) Thermochim Acta 69: 199-219
Klee J, Horhold HH, Flammersheim HJ (1990) Angew Makromol Chem 178: 63 -83
Klemm E, Flammersheim HJ, Martin R, Horhold HH (1985) Angew Makromol Chem
135:131-138
Kovacs AJ (1963) Fortschr Hochpol Forsch 3, 394-507
Kovacs AJ, Aklonis JJ, Hutchinson JM, Ramos AR (1979) J Polym Phys, Pol Phys Edn
17:1097-1162
Kovacs AJ, Hutchinson JM (1979) J Pol Sci, Pol Phys Edn 17: 2031- 2058
Lakshmikumar ST, Gopal ESR (1981) J Indian Inst Sci 63 (A):277-329
Lamprecht I, Schaarschmidt B (eds) (1977) Applications of Calorimetry in Life Sciences. de
Gruyter, Berlin
Lamprecht I, Hemminger W, Hohne GWH (eds) (1991) Special Issue: Calorimetry in the
Biological Sciences; Thermochim Acta 193: 1-452
Lau Sv, Suzuki H, Wunderlich B (1984) J Pol Sci Pol Phys Ed 22,379-405
214 References
Lavoisier AL, de Laplace PS (1784) Histoire de I:Academie Royale des Sciences, Annee
1780:355
Loblich K-R (1985 a) Thermochim Acta 83 :99-106
Loblich K-R (1986 b) Thermochim Acta 85 : 263 - 266
Loblich K-R (1994) Thermochim Acta 231 :7-20
Maglic KD, Cezairliyan A, Peletsky VE (eds) (1984) Compendiwn of Thermophysical
Property Measurement Methods 1. Plenwn, New York
MalavaSic T, Osredkar U, Anzur I, Vizovisek 1(1986) J Macromol Sci ChemA23:853-860
Marsh KN (ed) (1987) Recommended Reference Materials for the Realization of Physio-
chemical Properties. Blackwell Scientific Publications, Oxford
MathotVBF (1984) Polymer 25:579-599
Mathot VBF, Pijpers MFJ (1983) J Thermal Anal 28: 349 - 358
MathotVBF, Pijpers MFJ (1989) Thermochim Acta 151 :241-259
Mathot VBF (ed) (1994a) Calorimetry and Thermal Analysis of Polymers. Hanser, Munich
Mathot VBF (ed) (1994b) Special Issue: Thermal Analysis and Calorimetry in Polymer
Physics; Thermochim Acta 238: 321- 415
Matuschek G, Ohrbach K-H, Kettrup A (1991) Thermochim Acta 190: 111-123
Matuschek G, Ohrbach K-H, KettrupA (1993) JThermAnal39: 1141-1155
Mayorga OL, Navarro Rascon A, Freire E (1994) Thermochim Acta 238: 309 - 319
Mijovic J, Lee CH (1989) J Appl Pol Sci 37:889-900
Min BG, Stachurski ZH, Hodgkin JH (1993) Polymer 34:4908-4912
Montserrat S (1992) J Appl Pol Sci 44: 545 - 554
Montserrat-Ribas S (1992) Progr Colloid Polym Sci 87:78-82
Moratzky TH, Burkhardt G, Weyel W, Wegener G (1993) Thermochim Acta 229: 193 - 204
Moynihan CT, EastealAJ, de BoltMA, Tucker J (1976)Am Ceram Soc 59: 12-16 and 16-20
Narayanaswamy OS (1971) JAm Ceram Soc 54:491-498
Nolting J (1985) Thermochim Acta 94: 1-15
O'Neill MJ (1964) Anal Chern 36: 1238
Oetting FL (1970) Chern Thermodynamics 2: 727 -739
Ohrbach KH, Matuschek G, Kettrup A (1987) Thermochirn Acta 121: 87 - 96
Okui N (1990) Polymer 31 :92-94
Opfermann J, Wilke G, Jung J, Ludwig W, Hagen S, Gebhardt M, Kaisersberger E (1991)
Therm Analysenverf in Ind u Forschung. Wiss Z FSU Jena 51- 68
Opfermann J (1992) Technical report concerning the peak separation. Netzsch-Geriitebau-
GmbH, Selb, Germany
Oscarson JL, Izatt RM (1992) In: Rossiter BW, Baetzold RC (eds) Physical Methods of
Chemistry, 2 nd Eds. Wiley, New York
Parrish WmR (1986) Fluid Phase Equilibria 29: 177-192
Perez J, Cavaille N, Calleja RD, Ribbeles JLG, Pradas MM, Greus AR (1991) Makromol
Chern 192:2141-2161
Perrenot R, de Valliere P, Widmann G (1992) ThermAnaI38:381-390
Petit J-L, Sicard L, Eyraud L (1961) CR hebd 252: 1740-1742
Petrie SEB (1972) J Pol Sci PartA-2 10: 1255-1272
Peyser P (1983) J Macromol Sci-Phys, B22: 185-196
Podesva J, Prochazka 0 (1979) Makromol Chern 180: 2525 - 2530
PoeBnecker W (1990) J Therm Anal 36: 1123 -1139
PoeBnecker W (1993) Thermochim Acta 229: 97 - 109
Pomposo JA, Eguiazabal I, Calahorra E, CortazarM (1993) Polymer 34:94-102
Preston-Thomas H (1976) Metrologia 12:7-17
Preston-Thomas H (1990) Metrologia 27: 3 - 10
Privalov PL, Plotnikov VV (1989) Thermochim Acta 139: 257 - 277
References 215
Raetz K (1989) ThermochimActa 151 :323-331

Rahm V, Gmelin E (1992) JTherm Anal 38:335-344
Rai US, Shekhar H (1993) J Therm Anal 39:415-428
Ramos AR, Hutchinson JM, Kovacs AJ (1984) J Pol Sci, Pol Phys, Edn 22: 1655-1695
Reading M, Luget A, Wilson R (1994) Thermochim Acta 238: 295 - 307
Regenass W (1985) ThermochimActa 95:351-368
Reichelt J, Hemminger W (1983) Thermochim Acta 69: 59-70
Richardson MJ (1976) Plastics and Rubber, Materials and Applications 1: 162-167
Richardson MJ (1989) In: Booth C, Price C (eds) Comprehensive Polymer Science, Vol I,
Pergamon, Oxford: 867-901
Richardson MJ (1992 a) In: Maglic KD, Cezairliyan A and Peletsky VE (eds) Compendium of
Thermophysical Property Measurement Methods 2. Plenum, New York 529
Richardson MJ (I 992 b) Pure Appl Chern 64: 1790-1800
Richardson MJ (1993) ThermochimActa 229: 1-14
Richardson MJ, Savill NG (1975a) Polymer 16:753-757
Richardson MJ, Savill NG (I 975b) Thermochim Acta 12:213-220
Roles KA, Wunderlich B (1993) J Pol Sci, Part B, Pol Phys 31: 279- 285
Rossini RD (ed) (1956) Experimental Themochemistry, Vol I. Interscience Publishers, New
York
Rouquerol J, Zielenkiewicz W (1986) ThermochimActa 109: 121-137
Ryan ME, DuttaA (1979) Polymer 20:203-206
Sarge SM (1991) ThermochimActa 198:323-334
Sarge SM, Cammenga HK (1985) Thermochim Acta 94: 17 - 31
Sarge SM, Gmelin E, Hohne GWH, Cammenga HK, Hemminger W, Eysel W (1994)
Thermochim Acta 247 : 129 - 168
Schawe JEK (1993) Thermochim Acta 229: 69- 84
Schawe JEK (1995) Thermochim Acta 261 : 183 -194
Schawe JEK, Schick C (1991) ThermochimActa 187:335-349
Schawe JEK, Schick C, Hohne GWH (1993) Thermochim Acta 229: 37 - 52
Schawe JEK, Schick C, Hahne GWH (1994) Thermochim Acta 244: 49 - 59
Schawe JEK, Winter W, Hahne GWK (1996) Thermochim Acta, submitted
Schick C, Kramer L, Mischok W (1985) Acta Polymerica 36:47-53
Schick C, Fabry F, Schnell U, Stoll G, Deutschbein L, Mischok W (1988) Acta Polymerica
39:705-710
Schmolz E, Lamprecht I, Schicker B (1993) Thermochim Acta 229: 173-180
Schulze FW, Petrick HJ, Cammenga HK (1977) Z phys Chern NF 107: 1- 19
Schwanebeck W (1991) Thermochim Acta 187: 201- 21 0
ShuitzAR (1984) J Pol Sci, Pol Phys Edn 22: 1753-1771
Singh J (1993) Thermochim Acta 226 :211-220
Skinner HA (ed) (1962) Experimental Thermochemistry, Vol II. Interscience Publishers, New
York
Smith IT (1961) Polymer 2:95-108
Sourour S, Kamal MR (1976) Thermochim Acta 14:41-59
Spink H, Wadso I (1976) In: Glick D (ed) Methods of Biochemical Analysis. Wiley, New York
Stevens GC, Richardson MJ (1985) Polym Commun 26: 77 - 80
Struik LCE (1978) Physical Aging in Amorphous Polymers and Other Materials. Elsevier,
Amsterdam
Sullivan PF, Seidel G (1968) Phys Rev 173: 679
Sunner S, Mansson M (eds) (1979) Combustion Calorimetry. Pergamon Press, Oxford
Suzuki H, Grebowicz J, Wunderlich B (1984) Makromol Chern 186: 1109 -111
Suzuki H, Wunderlich B (1984) JTherm Anal 29: 1369-1377
216 References
Takahashi Y, Asou M (1993) Thermochim Acta 223 : 7 - 22

Tanaka S (1992) Thermochim Acta 210: 67 -76
Ter Minassian L, Milliou F (1983) J Phys E: Sci Instrum 16:450-455
Tool AQ (1946) JAm Ceram Soc 29: 240- 253
Tryson GR, Shultz AR (1979) J Pol Sci: Pol Phys Ed 17: 2059 - 2073
Tsitsilianis C, Staikos G (1992) Macromolecules 25:910-916
Ueberreiter K, Kanig G (1952) J CoIl Sci 7: 529
Ulbrich U, Cammenga HI( (1993) Thermochim Acta 229: 53 -67
Vallebona G (1979) JThermAnaI16:49-58
van der Plaats G (1984) Thermochim Acta 72: 77 - 82
Von Stockar U, Marison IW (1991) Thermochim Acta 193: 21 5 - 242
Wadso I (1993) Thermochim Acta 219: 1-15
Watson ES, O'Neill MJ, Justin J, BrennerN (1964) Anal Chem 36: 1233-1238
White GK, Collocott SJ (1984) J Phys Chem Ref Data 13: 1251-1257
Wiedemann HG (1991) Thermochim Acta 187:245-255
Wies S, Geyer W, Eysel W (1992) J Therm Anal 38 : 277 - 287
Wight FR, Hicks GW (1978) Polym Engn and Sci 18:378-381
Williams ML, Landel RF, Ferry 10 (1955) J Am Chern Soc 77:3701-3707
Wisanrakkit G, Gillham JK. (1990) Journal of Coatings Technology 62:35-50
Wolf G, Schmidt H-G, Bohmhammel K (1994) Thermochim Acta 235: 23 - 29
Wunderlich B (1990) Thermal Analysis. Academic Press, Boston
Wunderlich B at al.: ATHAS databank update (1990) University of Tennessee, Knoxville,
USA
Wunderlich B, Jin Y, Boller A (1994) Thermochim Acta 238:277-293
Wunderlich B, Jones LD (1969) J Macromol Sci 3: 67 -79
Zhiying Z (1986) Scienta Sinica B 24: 1239-1247
Subject Index
AC calorimetry 6 - caloric 61 ff, 75f

accuracy 5, 186ff - curve 68
acetamide 105 - electrical 63,65
acetanilide III f - examples 54, 68
acrylate 147 - factor 12,36, 38f, 61 ff
activation energy 141 - function 62
activation by radiation 146 - heat flow rate 62ff,75
adiabatic - on cooling 5lf
- bomb calorimeter 204 - peak area 65ff, 76
- calorimeter 3,76,208 - procedure 45
- DSC 20 - substances 73ff,79, 119
- jacket 204 - temperature 43 ff, 72, 74
- scanning calorimeter 189 caloric calibration 61 ff
analogue electric circuit diagram 29,31 calorimeter 1, 193
aneroid drop calorimeter 20 I - classification 8, 193
aniline 144 - examples 194 ff
annealing 174 - heat accumulating 201
apparatus function 93 ff, 146 - heat compensating 194
applications, DSC 105ff - heat exchanging 208
Arrhenius equation 141 - safety 113
assignment of temperature 48 - types 2f, 193ff
autocatalytic effect 157 calorimetry 1, 193 ff
azobenzene 153 - applications 2
- classification 193
baseline 32,45,82, 84ff, 92, 135 - synopsis 193
- .changes 86 catalyst 115, 144
- construction 84 ff characteristic temperatures 44, 83
- definition 82 characteristic terms 81, 82 ff
- in determining of crystallinity 117 characterization
- of reactions 133, 135, 138, 144, 146, 153 - complete instrument 181
- repeatability 183 - materials 3, 105, 114
benzil Illf - measuring system 181
bio calorimeter 3 - results of a measurement 186
Biot-Fourier equation 23 checklist DSC 188
bomb calorimeter 2, 204 cholesteryl myristate 108
Borchardt & Daniels method 151 classification of calorimeters 8, 193 f, 209
combustion calorimeter 204
calcium oxalate monohydrate 110 comparison
calcium stearate monohydrate III - manufacturers 188
calibration 10,36,41 ff - measurements 3,183
218 Subject Index
comparison disk-type DSe 8ff,208

- types ofDSe 188, 189f drop calorimeter 2,201
compatibility of substance and crucible 77 DSe
compensating heating power 14 ff - actual problems 4
compensation calorimeter 194 - applications 3, 105
composition 114 - calibration 41
conductance, thermal 35 - characterization 181
construction principle 194 - check list 188
controler of power compensation 15 - curve 81,83
convection 36 - dynamic 6, 11, 132
convolution integral 93 f - heat flux 8 ff, 192, 208
convolution product 93 f - linearity 183
cooling mode 51 - modulated 6,11,132
correction 46, 187 - noise 181
- of errors 187 - oscillating 6, 11, 132
cp see heat capacity - parameters 83
- change 84f, 161 f - performance 181
criteria of classification 193 - power compensation 7, 14 ff, 38 ff, 200
crucible 28,38,77, 120 - theory 21 ff
cryostate 5, 145 - types of 1,7, 189,208
crystal thickness 110 DSe curve 81 ff
crystallinity, degree of 115 - characteristic terms 82
curve, measured 81 ff - desmearing 89ff
cylinder-type DSe II f, 209 - evaluation 103
- kinetic evaluation 139ff
decomposition 110, liS, 139 DTA If, 189f
- oxydative liS - applications 2ff
- thermal 115 dual sample DSe 10, 129
deconvolution 93 ff, 146 dynamic mode 7, 146
deformation calorimetry 5
deformation energy 43 electric calibration 61 ff, 65 ff
degree of reaction 133,140 Ellerstein method 152
degrees of crystallinity 115 ff enthalpy 42, 130f
- rigid amorphous phase 117 - determination 87 f
- semicrystalline polymers liS - reaction 88, 132 ff
- two-phase model 115 f - transition 87 f, 116
dehydration 110 enthalpy relaxation 16lf, 165 f
desmearing 89ff enthalpy-temperature diagram 139
- advanced 96 ff entropy 132
- cp-measurements 121 epoxide 134f, 144, 150
- DSe curve 81,89ff errors 129
- heat flow rate 92 ff - random 186
- numerical methods 93 ff - systematic 184, 186
- reaction curve 143, 146, 153 eutectic 111, 176
- temperature 91 ff events, puis-like 93
- Tg-measurements 162,171 f examples of calibration 54, 68
detection limit 182 examples of calorimeters 194 ff
devitrification 162 - heat accumulating 201 ff
differential measurement (method) 7 f - heat compensating 194 ff
Differential Thermal Analysis 1f, 189 f - heat exchanging 208 f
dimyristoylphosphatidylcholine 109 exothermic events 49
Subject Index 219
extent of reaction 132f - - correct determination 134f

extrapolated peak completition temperature - - isothermal curves 133
44,83 - - temperature programmed mode 134
extrapolated peak onset temperature 34, 38, - of solution 195
44ft', 72, 83 - of transition 42,76, 114
Eyring equation 141 - true 61ft', 183
heat accumulating calorimeters 201
fictive temperature 165 f heat capacity 22, 24 f, 117 ft', 130
final peak temperature 44, 83 - applications 130
finite-element method 33 - curve 87
fixed point 43,54, 75f - measurement 73,118ft'
flow calorimeter 3, 204 - - classic procedure 119
flow-mix calorimeter 206 - - discontinuous procedure 127
formal kinetics 149 - - dual sample DSC 129
Fourier transform 95 - - dual step method 124
- inverse 95 - - errors 121 ft', 129
Freeman & Carrol method 152 - - modified classic procedure 124
- - with modulated DSC 132
gas calorimeter 2, 205 - of copper 79
glass transition 114,131, 145f, 160ft' - of sapphire 78
- nature 161 heat compensating calorimeters 194
- phenomenology 160 heat-exchanging calorimeters 208
- relaxation peaks 167 heat flow I
glass transition temperature 163 ft' heat flow rate 7f, 16,21,23
- applications 171 ft' - calibration 62ft',75f
- - characterization of glasses 171 - displayed 91
- - modeling of relaxation phenomena - exchanged 11,16
171,174 - measured 61 ff, 184
- - polymerization 134,172 - resolution 181 f
- - T, and molar masses 172 - transfer 86
- - T, of copolymers 174 - true 11,16,36,61 ft', 92ft', 183
- - T. of semicrystalline polymers 175 heat flux DSC 8 ft', 21 ff, 192, 208
- conventional 163ff - cylinder-type measuring system 11 ff, 209
- - inconsistencies 164 - disk-type measuring system 8ft', 21 ft', 30,
- - measurement 162,164 208
- definition 163 f - dual sample 10,129
- errors of 168 - equivalent electric circuit 22, 27, 31
- - magnitude 169 - modells 21 ft'
- - minimization 169 - modulated heating mode 11
- thermodynamic - numericalsimulation 33 ft'
- - calculation 167ft' - theory 21 ft'
- - definition 165ft' - triple measuring system 10, 129
- - measurement 167ft' heat pulse 26, 93 f
Green's function 93,98, 146 heat resistance 50
heterogeneous system 149
heat 1 high-pressure calorimeter 6
- measured 61 ft' high-temperature calorimeter 5
- of crystallization 114 homogeneous system 149
- offusion 114, 116
- of reaction 133ft' ice calorimeter 196
- - base line 134 identification 105
220 Subject Index
impurity 176ff lamellae crystal 110

inaccuracy 187 latent heat 87
indicated temperature 43 f light-activated reactions 19,146,155
initial peak temperature 44, 83 linear response 39,92
International Temperature Scale (1990) 43f, linearity 183 f
75 liposome 109
interpretation liquid-crystalline mesophases 108
- of results 5,103f liquid crystals 51, 108
- of thermal behavior 114 - temperature calibration 51
isoperibol 14, 15f, 194, 200f, 204, 209 low-temperature calorimetry 5, 145
isothermal
- calorimeter 194 f, 209 measured curve 81 ff
- mode 7, 143, 156 measured heat flow rate 61 ff
- reaction curves 133 measurement 186
isothermals 119 ff - principle 193
ITS-90 43 f, 75 measuring system 188, 189ff
mesophase transitions 108, III
kinetic analysis 132, 139 metallography III
kinetic evaluation 132 ff metastable phases 105 f
- requirements 141 mineralogy 111
kinetic investigations 3, 132, 139ff mixing calorimeter 2
- activation mixtures 111, 176
- - by radiation 146ff, 155 mode of operation 193
- - thermal 146,153, 155f - isoperibol 14
- base lines of reaction curves 134 ff, 144 f, - isothermal 7, 143, 153 f
153 - scanning 7,146, 153f
- calculation methods 149 ff modification 105 f
- - Borchardt-Daniels lSI modulated DSC 6, 11, 132
- - Ellerstein 152
- - Freeman-Caroll 151
naphthalene 176
- - non-linear regression 151,153,
nematic phase III
157
nitroaniline 113
- - Opfermann lSI
noise 10,95,181 f
- data sampling 141
numerical methods 93
- phase system
numerical simulation 33
- - heterogeneous 149
- - homogeneous 149
- reaction mechanism 139, 152ff operation mode 193
- - formal kinetic parameters ISO f order of reaction 140
- - rate laws 140, 149 oscillating DSC 6, 132
- sample preparation 141 overall rate laws 149 ff
- substances oxidation 11 0
- - aniline 144
- - azobenzene (cis, trans) 153 partial peak area 27
- - bisphenol-A-diglycidylether 144 peak
- - 3,6-dioxaoctane-I,8-dithiol 144,159 - area 32,65f
- - methacrylates 156 - definition 82 f
- - phenylglycidyl 147 - maximum 33
- - trimethylene diisocyanate 144, 159 - overlapping 114
kinetics 139 ff - partial integration 27
Kirchhoff-equation 130, 133, 137 - repeatability 182
Subject Index 221
- shape of 34 t, 38 principle of measurement 193

- temperature maximum 83 purity determination 3, 176 ff
peak area calibration 65ff,76 - calculations 176
Peltier cooling 145 - eutectic systems 176
performance of DSC 181 ff - new method 177
pharmacy 109, 176 - smearing effects 177 f
phase behavior 105 ff, 176 f - van't Hoff equation 176
- acetamid 105
- calcium oxalate monohydrate 110 quality assurance 2, 41
- calcium stearate monohydrate 110 quality control 114ff
- cholesteryl myristate 108 - aging of plastic 114
- dimyristoylphosphatidylcholine 109 - degree of crystallinity 115
- liquid crystalline mesophases 108 - oxidative decomposition 115
- metastable phases 105 f - plastic mixture 114
phase diagrams 3, III ff - reactive resins 115
- acetanilidelbenzil IlIff - semicrystalline polymers 115
- eutectic melt III - thermal decomposition 115
- eutectic mixture Ill, 176f quasi steady state 25, 39
- eutectic systems 113, 176 quenching 175
- phenanthrene/naphthalene 176
phase transition 105 ff radiation 35, 146f
phenanthrene 176 radicalic polymerisation 153 f
phenylbutazone 105 random errors 186
photo-DSC 18ff,146f rate laws 140 f, 149
- base line 146 reaction 132 ff
photo reactions 146ff - autocatalytic 144
- chain initiation 155 - consecutive 144
- chain propagation 155 - curves 115,133ff
- chain termination 155 - heat of 133
- dark reaction 155 - isothermal 133, 143
plastics see polymers - mechanism 152
polybutadiene 117 - non-isothermal 134, 143
polydimethylsiloxane 117 - order 140, 150
polyethylen 116 reaction calorimeter 2
polyethylenterephtalate 117 recursion method 95
polymerisation 146, 155 recycling 114
polymers 114,116,131,161 ff reference sample 23 ff
- amourphous 117,131 repeatability 5,44, 18If, 186f
- recycling 114 reproducibility 182, 186
- semicrystalline 115,130,175 result 186
polymorphism 105 f, 108 round robin 183
polystyrene 101,170
postperiod 202 safety aspects 3, 113
potential function 130f - p-nitroaniline 113
power compensation DSC 7, 14 ff, 38 ff, 200 safety calorimeter 2, 113
- control 15 safety techniques 113
- function 17 sample preparation
preexponential factor 141 - cp-measurements 122, 130
preperiod 202 sample preparation
presentation of results 103 - reactions 141
pressure DSC 11 - Tg-measurements 164
222 Subject Index
sample temperature 22, 43 thermal lag 99ff, 100, 102, 122

semicrystalline polymers 115, 130, 175 - cp-measurements 122
sensitivity 32, 186 - Tg-measurements 162
side reaction 139 thermal resistance 22 ff, 49 f, 86
signal-to-noise ratio 182 thermodynamic functions 42, 84ff,
silicatic glass 160 130ff
simultaneous techniques 4 thermodynamics 42, 130
smearing 25, 122, 177 thermometer problem 27
smectic phase 111 Tian equation 25
stability 105 time constant 10, 13, 16,25,29,30,38,93,
- investigations 3,113, 114f 122, 184ff
steady state 21,23,37 transition enthalpy 87
step response function 99 triple measuring system 10
subscripts XII (???) true heat 36
substance identification 105 true heat flow rate 11,16,36,61 ff, 92ff
surface energy 43 true temperature 43
symbols XI (???) types
symmetry 23 f, 40 - of calorimeters 8, 193 f, 209ff
systematic errors 186 - ofDSC 7
temperature uncertainty 5, 10,47

- calibration 43, 52, 72 - heat measurement 10
- correction 91, 98 - measurement 44, 187
- displayed 91 - overall 42
- profile in sample 98 - temperature calibration 47
temperature scale 43,74 - total 187f
temperature-modulated DSC 6, II, 132
temperatures, characteristic 44, 83 van't Hoff equation 176
theory of DSC 21 ff vitrification 162, 166
theory of linear response 39,92 ff
thermal activation 146 zero line 16,40,82,83,87,92, 118f,
thermal conductivity 23,35,86, 122 183
thermal history 114 - repeatability 182f
Springer-Verlag
and the Environment
We at Springer-Verlag firmly believe that an

international science publisher has a special
obligation to the environment, and our corpo-
rate policies consistently reflect this conviction.
We also expect our busi-

ness partners - paper mills, printers, packag-
ing manufacturers, etc. - to commit themselves
to using environmentally friendly materials and
production processes.
The paper in this book is made from

low- or no-chlorine pulp and is acid free, in
conformance with international standards for
paper permanency.

DSC Book

Uploaded by

Copyright:

Available Formats

DSC Book

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

DSC Book

Uploaded by

Copyright:

Available Formats

Differential Scanning Calorimetry

Springer-Verlag Berlin Heidelberg GmbH

With 136 Figures and 13 Tables

Dr. H.-J. Flammersheim

ISBN 978-3-662-03304-3 ISBN 978-3-662-03302-9 (eBook)

Library of Congress Cataloging-in-Publication Data

1. Calorimetry. 1. Hemminger, W. II. Flammersheim, H.-J., III. Title.

© Springer-Verlag Berlin Heidelberg 1996

Typesetting: Data-conversion by Fotosatz-Service Kohler OHG. Wiirzburg

Differential Scanning Calorimetry (DSC) is a well established measuring method

2 Types of Differential Scanning Calorimeters 7

2.1 The Heat Flux DSC . . . . . . . . . . . . . . . 8

3 Theoretical Fundamentals of Differential Scanning

3.1 The Heat Flux DSC 21

4 Calibration of Differential Scanning Calorimeters 41

4.1 Aspects of Quality Assurance 41

5 The DSC Curve 81

5.1 Characteristic Tenns 82

6 Applications of Differential Scanning Calorimetry . . . . . . . 105

6.1 General Applications Without Sophisticated Evaluation

6.2.2.2.7 Determination of the True Reaction Mechanism 152

7 Evaluation of the Performance of a Differential Scanning

7.1 Characterization of the Complete Instrument 181

Subject Index 217

A area, pre-exponential factor

The objective of calorimetry is the measurement of heat. To measure heat means to

Differential Thermal Analysis (DTA)

and the high sensitivity to anomalies of the temperature-time function. DTA

The ICTAC definition of Differential Thermal Analysis is as follows:

CalOrimetry and DTA Today

in highly sensitiveflow calorimeters (usually of the isoperibol type, i. e. with sur-

there is not yet a theoretically well-founded complete understanding ofDSC. The

low-temperature calorimetry for measuring heat capacities (see e.g. Gmelin,

- high-temperature calorimetry in material science (see e.g. Bruzzone, 1985;

Two types of Differential Scanning Calorimeters (DSCs) must be distinguished:

2.1 The Heat Flux DSC

2.1.1 Heat Flux DSC with a Disk-Type Measuring System

'--- _ _6_T_ _- - l <<

b time, tempera ture

Heat Flux DSC with Triple Measuring System

The development of an adiabatic high temperature triple-cell DSC system (up to

Pressure Heat Flux DSCs

Heat Flux DSC with Modulated Heating Mode

2.1.2 Heat Flux DSC with a Cylinder-Type Measuring System

I1T = I1TFR - I1TFF = I1TsR

IPm = -k' ·I1T

Compared with disk-type measuring systems, cylinder-type measuring systems

quantitatively describes the functional correlation between the instantaneous mea-

2.2 The Power Compensation DSC

ing resistor (made of platinum wire). The microfumace is about 9 mm in diameter,

microfurnace containing the reference sample. The temperature difference is both

The factor k) is a factory-set fixed quantity of the proportional controller, k2 can be

quartz cuvette with

in energy is incident on the sample, it is either absorbed directly or - as is usually

In all DSCs, a temperature difference tlT - given as a voltage - is the original

3.1 The Heat Flux DSC

The Zeroth Approximation

In absolute values, the Biot-Fourier equation of the (steady-state) conduction of

In the case of thermal symmetry,