Differential Scanning Calorimetry

Differential Scanning Calorimetry
Springer-Verlag Berlin Heidelberg GmbH

G. W. H. H6hne . W. F. Hemminger
H.- J. Flammersheim
Differential Scanning
Ca lori metry
2nd revised and enlarged edition
With 130 Figures and 19 Tables
" Springer
Dr. G.W.H. H6hne
(Retired from Universitat Ulm)
Morikeweg 30
88471 Laupheim, Germany
e-mail: gwh.hoehne@t-online.de
Dr. W. F. Hemminger
Physikalisch-Technische Bundesanstalt
Bundesallee 100
38116 Braunschweig, Germany
e-mail: WolfgangHemminger@ptb.de
Dr. H.-J. Flammersheim

Universitat Jena
Institut fur Physikalische Chemie
Lessingstraf5e 10
07743 Jena, Germany
e-mail: C7flha@uni-jena.de
ISBN 978-3-642-05593-5
Library of Congress Cataloging-in-Publication Data

Hohne, G. (Glinther) Differential scanning calorimetry: an introduction for practitioners / G. W. H. Hohne,
W. Hemminger, H.-J. Flammersheim. -- 2nd rev. and en!. ed. p. cm. Includes bibliographical references and index.
ISBN 978-3-642-05593-5 ISBN 978-3-662-06710-9 (eBook)
DOI 10.1007/978-3-662-06710-9
1. Calorimetry. 1. Hemminger, W., 1941-II. Flammersheim. H.-J., 1942-II1. Title.
QC29I.H64 2003
536'.6--dc21 2003050472
This work is subject to copyright. AII rights are reserved, whether the whole or part of the material is concerned,
specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on
microfilrns or in any other ways, and storage in data banks. Duplication of this publication or parts thereof is
only permitted under the provisions of the German Copyright 1.aw of September 9, 1965, in its current version,
and permission for use must always be obtained from Springer-Verlag Berlin Heidelberg GmbH.
Violations are liable for prosecution under the German Copyright Law.
http://www.springer.de
Springer-Veriag Berlin Heidelberg 2003
Originally published by Springer-Verlag Berlin Heidelberg New York in 2003
Softcover reprint of the hardcover 2nd edition 2003
The use of general 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 applica-
tion contained in this book. In every individual case the user must check such information by consulting the
relevant literature.
The instructions given for the practical carrying-out of HPLC steps and preparatory investigations do not
absolve the reader from being responsible for safety precautions. Liability is not accepted by the author.
Cover Design: Klinkel & Lopka, Heidelberg
Typesetting: Data-conversion by Fotosatz-Service Kohler GmbH, Wlirzburg
Production Editor: Christiane Messerschmidt, Rheinau
Printed on acid-free paper 52/3020 - 5 4 3 2 1
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. Measure-
ment values obtained by DSC allow heat capacity, heat of transition, kinetic data,
purity and glass transition to be determined. DSC curves serve to identify sub-
stances, to set up phase diagrams and to determine degrees of crystallinity.
This book provides, for the first time, an overall description of the most
important DSC measuring systems and measuring programs, including the
modulated temperature DSC. Furthermore many examples of typical and wide-
ly used applications of Differential Scanning Calorimetry are presented. Pre-
requisites for reliable measurement results, optimum evaluation of the mea-
surement curves and estimation of the uncertainties of measurement are, how-
ever, 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, and for
the recently introduced modulated temperature mode of operation (TMDSC) as
well. Temperature calibration and caloric calibration are described on the basis
of thermodynamic principles. Desmearing of the measurement curve in differ-
ent ways is presented as a method for evaluating the curves of fast transitions.
The instrumental data which are most important for the characterization of
Differential Scanning Calorimeters are defined, and it is explained how they are
determined experimentally. This enables every potential instrument buyer to
ask the manufacturer for measured characteristic data which will allow him to
compare the different instruments available. To make measurement results com-
parable, in addition to their traceability (via reference materials) a standardized
evaluation of measurement uncertainty is indispensable. How this may be done
is explained in some detail.
We are indebted to S.M. Sarge for a critical examination of the manuscript
and valuable suggestions for improvement, to S. Rudtsch for contributions to the
chapter on calibration and to the expression concerning uncertainties and in
particular to M. J. Richardson for helping us with the translation.
Braunschweig, Jena and Ulm G. W. H. Hahne

March 2003 w. F. Hemminger
H.-f. Flammersheim
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 Types of Differential Scanning Calorimeters and Modes

of Operation . . . . . . . . . . . . . . . . . . . . . 9
2.1 Heat Flux DSC . . . . . . . . . . . . . . . . . . . . . 10
2.1.1 Heat Flux DSC with Disk-Type Measuring System . 10
2.1.2 Heat Flux DSC with Turret-Type Measuring System 13
2.1.3 Heat Flux DSC with Cylinder-Type Measuring System 14
2.2 Power Compensation DSC . . . . . 17
2.2.1 Function Principle . . . . . . . . . . . . . . . . . . . . 17
2.2.2 Special Power Compensating DSC . . . . . . . . . . . 22
2.3 DSC with Combined Heat Flux and Power Compensation
Measuring System .. 25
2.4 Modes of Operation . . . . . . . . . . . . . . . . . 25
2.4.1 Constant Heating Rate . . . . . . . . . . . . . . . . 26
2.4.2 Variable Heating Rate (Modulated Temperature) . 27
3 Theoretical Fundamentals of Differential Scanning Calorimeters . 31

3.1 Heat Flux DSC . . . . . . . . . . . . . . 31
3.2 Power Compensation DSC . . . . . . . . . 48
3.3 Temperature-Modulated DSC (TMDSC) 50
3.3.1 The Temperature-Modulated Method 51
3.3.2 Influences of the Sample . . . 52
3.3.3 Influences of Heat Transport 57
3.3.4 Conclusions . . . . . . . . . . 63
4 Calibration of Differential Scanning Calorimeters 65

4.1 Aspects of Quality Assurance 66
4.2 Basic Aspects of Calibration . . . . . 66
4.3 Temperature Calibration . . . . . . . 69
4.3.1 Temperature Calibration on Heating 69
4.3.2 Temperature Calibration on Cooling 84
4.4 Caloric Calibration . . . . . . 86
4.4.1 Heat Flow Rate Calibration . . . 87
4.4.2 Heat (Peak Area) Calibration .. 90
4.4.3 Examples of Caloric Calibration 92
4.4.4 Caloric Calibration in Cooling Mode 97
VIII Contents
4.5 Conclusions Regarding the Calibration of DSCs 97

4.6 Reference Materials for DSC Calibration . . . . . 98
4.6.1 Reference Materials for Temperature Calibration 99
4.6.2 Reference Materials for Heat Flow Rate Calibration 101
4.6.3 Reference Materials for Heat (Peak Area) Calibration 107
4.7 Additional Calibration in Temperature-Modulated Mode
of Operation . . . . . . . 107
4.7.1 Calibration of Magnitude 113
4.7.2 Calibration of Phase . . . 113
5 DSC Curves and Further Evaluation . 115

5.1 Characteristic Terms of DSC Curves . 116
5.2 Parameters Influencing the DSC Curve 118
5.3 Further Evaluation of DSC Curves ... 119
5.3.1 Determination of the Real Sample Heat Flow Rate 119
5.3.2 The Baseline and the Determination of Peak Areas
(Enthalpy Differences) . . . . . . . . . . . . . . . . 121
5.4 Desmearing of the DSC Curve . . . . . . . . . . . 126
5.4.1 Correction of the Temperature and Heat Flow Rate Indicated 128
5.4.2 Subtraction of the Zeroline . . . . . . . . . . . . . . . . 129
5.4.3 Calculation of the True Heat Flow Rate into the Sample 129
5.4.4 Advanced Desmearing 133
5.4.5 Further Calculations . . . . . . . . . . . . . . 140
5.5 TMDSC Curves . . . . . . . . . . . . . . . . . 140
5.5.1 Reversing and Non-Reversing Heat Capacity 141
5.5.2 Complex Heat Capacity . . . . . . . . . .. 143
5.5.3 Curves from Step-Scan Evaluation . . . . . 145
5.6 Interpretation and Presentation of Results . 145
6 Applications of Differential Scanning Calorimetry 147

6.1 Measurement of the Heat Capacity 147
6.1.1 "Classical" Three-Step Procedure . . . . . . . . . . . 148
6.1.2 The ''Absolute'' Dual Step Method . . . . . . . . . . 153
6.1.3 General Precautions for the Minimization of Errors
and their Estimation . . . . . . . . . . 154
6.1.4 Procedure of Small Temperature Steps . . . . . . . . 155
6.1.5 The Temperature-Modulated Method . . . . . . . . 159
6.1.6 Typical Applications of Heat Capacity Measurements 160
6.2 Determination of Heats of Reaction 162
6.3 Kinetic Investigations . . . . . . . . . . . . . . . . . . 168
6.3.1 Introduction and Definitions . . . . . . . . . . . . . . 168
6.3.2 Experimental Prerequisites for a Reliable Kinetic Analysis 173
6.3.3 Selection of the Measuring Conditions -
Isothermal or Non-Isothermal Reaction Mode 175
6.3.4 Activation of the Sample by UV Irradiation 180
6.3.5 Different Strategies of Kinetic Evaluation . 183
6.3.6 Selected Examples and Possible Predictions 189
Contents IX
6.4 The Glass Transition Process . . . . . . . . . . . . . . 200

6.4.1 The Phenomenology of the Glass Transition . . . . . 200
6.4.2 The Nature of the Glass Transition and Consequences
for DSC Measurements . . . . . . . . . . . . . . . . . 201
6.4.3 Definition and Determination of the Glass Transition
Temperature Tg 203
6.4.4 Applications of Glass Transition Measurements . . . 212
6.4.5 The Dynamic Glass Process, an Example . . . . . . 217
6.5 Characterization of Substances, the Phase Behavior 219
6.5.1 Applications in Biology and Food Science 219
6.5.2 Applications in Pharmacy 221
6.5.3 Other Applications . . . . . . . . . 224
6.5.4 Porosity Measurements . . . . . . 228
6.6 Determination of Phase Diagrams 230
6.7 Safety Aspects and Characterization of High-Energetic Materials 232
6.8 Characterization of Polymers . . . . . . . . . 233
6.8.1 Effects of Origin and Thermal History . . . . . . . . . 234
6.8.2 Determination of the Degree of Crystallinity . . . . . 236
6.8.3 Advanced Characterization with the TMDSC Method 238
6.9 Purity Determination of Chemicals . . . . . . . . . . 241
7 Evaluation of the Performance

of a Differential Scanning Calorimeter 245
7.1 Characterization of the Complete Instrument 245
7.2 Characterization of the Measuring System 245
7.3 Characterization of the Results of a Measurement:
Uncertainty Determination 251
7.3.1 Black Box Method . 253
7.3.2 GUM Method . . . . 253
7.4 Check List for DSCs 257
Appendix 1 259
Appendix 2 263
References . 281
Subject Index 291

List of Symbols
A area, pre-exponential factor

C heat capacity, electric capacity
D diffusivity
E energy
H enthalpy
K factor (calibration), coefficient
L thermal conductance
P electric power, transfer function
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
i electric current
k proportionality factor, calibration factor, rate constant, conductivity
1 length, distance
m mass
n reaction order
p pressure
r rate of reaction, radius
t time
w weight fraction
x mole fraction
a degree of reaction
f3 heating rate
6 phase angle
y expansivity coefficient
E emissivity
A thermal conductivity
v stoichiometric number
tP heat flow rate
p density
a standard deviation
XII List of Symbols
r time constant
w angular frequency
X compressibility
~ extent of reaction, composition, conversion
Subscripts
A activation, amplitude
a amorphous
bl baseline
c extrapolated offset, crystalline
cal calorimeter
e extropolated onset
el electric
eq equilibrium
exp experimental
F furnace
f final
fix fixpoint
fus fusion
g glass
h, i, k, n running numbers
initial
iso isothermal
I liquid
lin linear
lit literature
M measurement point
m measured
mix mixing
o onset
p peak, constant pressure
prop proportional
4> 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
u underlying
V constant volume
o 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 creates 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
generation or consumption of heat, calorimetry is a universal method for inves-
tigating 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 cen-
tury. Although modern 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 calorime-
ters will be described only briefly in Appendix 2 to give the reader a more com-
prehensive 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). In addition to the measurement of heat,
DSCs are used to measure heat flow rates (power) and characteristic tempera-
tures of a reaction or a transition as well. The precise measurement of heat
capacities, which is of an increasing importance, has distinctly been improved
with modern DSCs. The measurement of heat not only includes integral (total)
heats of reaction or transition but also the determination of "partial heats"
developed within a selected temperature interval. Such values are of impor-
tance for kinetic evaluations, determination of crystallinity and purity (see
Chapter 6).
An accurate definition of Differential Scanning Calorimetry (DSC) is as
follows:
Differential Scanning Calorimetry (DSC) means the measurement of

the change of the difference in the heat flow rate to the sample and to a
reference sample while they are subjected to a controlled temperature pro-
gram.
2 1 Introduction
It is important to understand that DSC measures the change of a property -

namely of a heat flow rate difference - which normally is released due to an al-
teration of the sample temperature. When there is no alteration of the sample
temperature, no change of a heat flow rate difference can be measured (except
for possible chemical reaction heat flow rates). That means, a distinct tempera-
ture program, or in general a "mode of operation", is always part of a DSC mea-
surement (cf. Chapter 2). In the case of thermally activated reactions/transitions,
e. g. by a controlled stepwise change from one constant temperature to another,
the sample's temperature undergoes an alteration by internal processes and
causes the DSC signal.
It should even be mentioned, that heat only flows if there is a temperature
difference present. In other words, a non-zero heat flow rate difference implies
always a temperature difference between the sample and the reference and their
surroundings and a change of the heat flow rate difference implies a change of
the temperature as well.
Differential Thermal Analysis (DTA)
To distinguish a heat flux DSC from an apparatus for Differential Thermal

Analysis (DTA), the latter will be characterized in the following.
Differential Thermal Analysis is applied to measure the temperature differ-
ence between the sample to be investigated and a reference sample as a function
of temperature (or time). This temperature difference indicates a heat exchange
qualitatively. DTA is of more recent date than classic calorimetry. Its advantages
as compared with conventional calorimetry are the dynamic mode of operation
("scanning") which allows reactions or processes to be investigated which can
be thermally activated, 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 construction of Differential Scanning Calorimeters (DSCs)
with disk-type measuring systems.
The widely used term Differential Thermal Analysis (DTA) means a thermo-
analytical method, which is more than a thermoanalytical (measurement) tech-
nique (see the following definition), because the method (analysis) includes in
principle the thermoanalytical investigation procedure, i. e. the evaluation and
interpretation of the measured values. An accurate definition for the thermo-
analytical technique is:
Differential Thermometry (DT) means the measurement of the change

of the difference in temperature between the sample and the refer-
ence sample while they are subjected to a controlled temperature pro-
gram.
The accurate term for the affiliated thermo analytical method is then Differen-
tial Thermal Analysis (DTA). For a short description of the principle of DTA
measuring systems in comparison with DSC, see Appendix 1.
1 Introduction 3
Calorimetry and OTA Today
Owing to the new materials and sensors used and the application of modern
mechanical manufacturing (micro and nano technologies), advanced data pro-
cessing by modern electronic systems, all calorimeters available today - includ-
ing the so-called "classic" calorimeters - are instruments which allow precise
measurements to be carried out with high sensitivity and repeatability (see Sect.
7.2), their operation being relatively simple or perhaps even automated. Caloric
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 assur-
ance (e.g., the standard ISO/IEC 17025, 1999: General requirements for the
competence of testing and calibration laboratories) 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 Chapter 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; Oscar-
son, Izatt, 1992). A suggestion how to classify calorimeters is given in Appendix 2.
- With bomb calorimeters, combustion heats (gross calorific values) are mea-
sured on a large scale in industry (costs of fossil fuels) (Sunner, Mansson,
1979).
- Gas calorimeters (flow calorimeters) are used for the continuous or discon-
tinuous 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 costs for the individual consumer (Hyde,
Jones, 1960; Hemminger, 1988; Sarge, 1997; Ulbig, Hoburg, 2002; Jaeschke,
2002; Dale et al., 2002; Alexandrov, 2002).
- Drop calorimeters (usually self-made) allow mean heat capacities or enthalpy
differences to be quickly measured. Drop sample temperatures of up to
2000C are realized (Chekhovskoi, 1984).
- Different types of isoperibol mixing calorimeters are used to investigate reac-
tions between two fluids or between a fluid and a solid (reaction heats, heats
of solution, 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, as all test parame-
ters (temperature, time, addition of substances etc.) are completely docu-
mented and can be varied automatically (Landau, 1996).
Questions of process technology which are also important for the dimen-
sioning 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 industrial production plants (large-scale calorimetry,
mega-calorimetry, see, e.g., von Stockar, Marison, 1991 in Lamprecht et aI.,
1991).
4 1 Introduction
Special calorimeters (commercially available) are also used to investigate

aspects of safety technique, for example, to determine characteristic temper-
atures of decomposition or the kinetics of reactions (decompositions, run-
away reactions) under certain boundary conditions (see, for example, Grewer,
1987; Schwanebeck, 1991; Singh, 1993; for a survey,cf. Grewer,Steinbach, 1993).
- In highly sensitive flow calorimeters (usually of the isoperibol type, i. e., with
surroundings at constant temperature), the heat generation of biological sys-
tems and their change at varying conditions of life are investigated (for ex-
ample, addition of pharmaceuticals to bacterial cultures). In different types of
biocalorimeters, the metabolism of organisms and their change due to exter-
nal influences (optical, acoustical, mechanical, thermal, chemical) are studied
(see, e.g., Spink, Wadso, 1976; Lamprecht, Schaarschmidt, 1977; Beezer, 1980;
Lamprecht et aI., 1991).
- Heat capacities and heats of transition are directly measured with high accu-
racy using adiabatic calorimeters. The characteristic data of materials deter-
mined with their aid are an indispensable basis for the calibration of the DSCs
(see, e.g., Gr0nvold, 1967; Kagan, 1984; Jakobi et aI., 1993; Gr0nvold, 1993).
Properly calibrated DSCs allow specific heat capacities to be measured with
an accuracy of 1 to 2 % (see Sect. 6.1; 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 mea-
sure in the manufacture and for the control of 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 quickly measured on 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 (in particular polymers),
- comparison (relative) measurements (quality control, identification of sub-
stances or mixtures),
- stability investigations,
- evaluation of phase diagrams,
- purity determinations,
- kinetic investigations,
- safety investigations,
- determination of heat capacity and complex heat capacity (with TMDSC).
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 2400C),
- they are highly sensitive,
- they are most flexible as regards the volume and form of the crucibles,
- their reasonably-priced measuring system can be easily exchanged.
1 Introduction 5
Characteristic temperatures of transitions or reactions can be very well deter-

mined 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 Ther-
moanalysis (EGA), Thermomicroscopy (TOA) or Thermosonimetry (TS).
The coupling of DSC or DTA with Thermogravimetry (TG) is of particular
importance. In addition to information on changes in the heat flow rate (due, for
example, to changes of Cp) and heats of transition, the TG signal provides infor-
mation on whether volatile components are involved and which changes in mass
are to be attributed to a transition.
The different methods of gas analysis used together with DSC or DTA (usual-
ly together with TG/DSC or TG/DTA systems) are of increasing importance.
Via a carrier gas stream, gaseous reaction products from the calorimeter or DTA
device are transported to an apparatus (e. g., IR or mass spectrometer) in which
these products are analyzed. This allows a correlation between the characteris-
tic temperature of a reaction, gravimetric and/or enthalpic information and
the composition of the volatile reaction product to be made (see examples in
Mathot,1994a).
It is difficult to transfer the volatile products without adulteration from the
hot sample to the analysis system (condensation must be avoided). Various sys-
tems for coupling quadrupole mass spectrometers are commercially available
(up to temperatures of 2400 0c). Examples of the investigation of volatile com-
ponents applying methods of gas analysis can be found in Ohrbach et al., 1987;
Matuschek et al., 1991; Matuschek et al., 1993; Kaisersberger, 1997.
The advantage of simultaneous measurements is that the same sample is
investigated under identical conditions and that diverse information is ob-
tained by one measurement run which is important for interpretation pur-
poses. The following potential disadvantages of simultaneous instruments
should be mentioned: lower sensitivity, higher susceptibility to failure, in-
creased time and effort required for preparation and operation, higher instru-
ment costs.
In Differential Scanning Calorimetry which is widely applied, there are still
some actual problems of which one should be aware:
- The theoretically well-founded complete understanding of DSC is not yet
totally accepted. The average user does not yet know the limits to this method
and the sources of the systematic errors by which it is affected. This is in par-
ticular true for the temperature-modulated DSC, its theoretical background
is still a matter of controversial discussion.
6 1 Introduction
- There are not yet any practicable and experimentally tested recommenda-
tions for temperature and heat calibration and for measurement procedures
which are internationally accepted (see Chapter 4; and some examples in
Della Gatta et al., 2000).
- There is no international agreement on a single set of substances for the tem-
perature and heat calibration of DSC which have been measured with suffi-
cient accuracy, including a metrologically sound traceability to national/in-
ternational standards; instead there is a confusing variety of "certified refer-
ence materials" whose characteristic data are in part contradictory. Sub-
stances which are certified on a metrologically basis (i. e., traceable to the SI
units) are offered by the German Metrology Institute (PTB) (cf. Sect. 4.6).
Due to the rapidly increasing use of DSCs in various fields of application, some
negligence has gained ground - favored 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 Con-
federation 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 (cf.
Sect. 7.3). The calibration capability (cf. Chapter 4), different influencing
quantities (cf. Sect. 5.2) and known theoretical considerations (cf. Chapter 3)
should be taken into account. In many cases, the repeatability of a DSC is, for
example, simply, but wrongly, indicated 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 has not been mentioned here as, in
comparison with DSC, they are used only in specific fields. To name a few ex-
amples:
- low-temperature calorimetry for measuring heat capacities (see, e. g., Gmelin,
1987),
- more recent techniques of cp measurement (see, e.g., Lakshmikumar, 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 al., 1991; Wadso,
1993 and Kemp, 1993; Kemp et aI., 1998; Hansen, 2000),
1 Introduction 7
- deformation calorimetry on polymers (see, e. g., Kilian, Hahne, 1983; Godovsky,

Hahne, 1994),
- high-temperature calorimetry in material science (see, e.g., Bruzzone, 1985;
Bros, 1989).
- (high)-pressure calorimetry (see, e. g., in Mathot, 1994 b; Hahne, 1999 b),
- AC calorimetry (Sullivan, Seidel, 1968) and 3w calorimetry (Rosenthal, 1961).
Summary
During the last decades the development of Differential Scanning Calorimetry

from a half-quantitative DTA method (producing "thermograms") toward a true
calorimetric method (yielding quantitative caloric quantities) forged ahead.
This becomes manifest from the following items which hopefully will be proved
within this book.
- The theory of (heat-flux) DSCs is well-known and the function principle un-
derstood.
- Systematic (unavoidable) error sources and methods of correction are
known.
- Modern data treatment and powerful computers make it possible for every-
one to make the necessary evaluations without problems.
- Metrological founded calibration and measurement procedures improved the
certainty of the results which nowadays are comparable with those obtained
with classical calorimeters.
- The difference in quality of the results from heat-flux and power-compensat-
ed DSCs has diminished.
- Precise heat capacity measurements are possible and open the door for other
well-defined thermodynamic quantities.
- Coming to terms with the problem of the thermal inertia (thermal lag) en-
ables kinetic investigations which otherwise need much more experimental
efforts.
- DSC is a fast and reliable method in a wide field of different applications.
2 Types of Differential Scanning Calorimeters
and Modes of Operation
Two basic types of Differential Scanning Calorimeters (DSCs) must be distin-

guished:
- The heat flux DSC,
- The power compensation DSC.
They differ in the design and measuring principle what will be dealt with in this
chapter. The theoretical bases are presented later in Chapter 3. Common to all
DSC's is a differential method of measurement which is defined as follows:
A method of measurement in which the measured quantity (the measurand)
is compared with a quantity of the same kind, of known value only slightly dif-
ferent from the value of the measurand, and in which the difference between the
two values is measured (International 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 difference princi-
ple 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 difference signal - which is the measurement signal actually of
interest - can be strongly amplified, as the high basic signal (signal of the indi-
vidual measuring system) is almost 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 of the difference principle.
The differential signal is the essential characteristic of each Differential Scan-
ning Calorimeter. Another characteristic - by which it is distinguished from
most classic calorimeters - is the dynamic mode of operation. In other words,
the DSC can be operated by applying various "modes of operation": The tem-
perature can not only be held constant but also raised or lowered at a preset rate
which might be superimposed with a temperature modulation (see below).
A characteristic common to both types of DSC is that the measured signal
is proportional to a heat flow rate iP and not to a heat as is the case with most
of the classic calorimeters. This allows time dependences of a transition to be
observed on the basis of the iP(t) curve. This fact - directly measured heat flow
rates - enables all DSCs to solve problems arising in many fields of application
(see Chapter 6).
10 2 Types of Differential Scanning Calorimeters and Modes of Operation
2.1
Heat Flux OSC
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 well-defined heat con-
duction path with given thermal resistance. The primary measurement signal
is a temperature difference; it determines the intensity of the exchange and the
resulting heat flow rate tP is proportional to it.
In commercial heat flux DSCs, the heat exchange path is realized in different
ways, but always with the measuring system being sufficiently dominating. The
most important fundamental types are:
- The disk-type measuring system, where the heat exchange takes place via a
disk which serves as solid sample support.
Features: Simple and easily realizable design with a high sensitivity, the
sample volume is small, but the heat exchange between furnace and sample is
limited which allows only medium heating and cooling rates.
- The turret-type measuring system, where the heat exchange takes place via
small hollow cylinders which serve as elevated sample support.
Features: More sophisticated design with high sensitivity and fast thermal
response which allows large heating and cooling rates, the sample volume is
small.
- The cylinder-type measuring system, where the heat exchange between the
(big) cylindrical sample cavities and the furnace takes place via a path with
low thermal conductivity (often a thermopile).
Features: Very sensitive with a large sample volume but with a large time
constant which allows only low heating rates, the sensitivity per unit volume
is, however, very high.
2.1.1
Heat Flux OSC with Oisk-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 medi-
um thermal conductivity (Fig. 2.1 a). The samples (or the sample containers) are
positioned on this disk symmetrical to the center. The temperature sensors are
integrated in the disk. Each temperature sensor covers more or less the area of
support of the respective container (crucible, pan) so that calibration can be car-
ried out independent 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 and to the support must always be the same (center
pin or the like).
Metals, quartz glass or ceramics are used as disk materials. Type (and number)
of the temperature sensors (e.g., thermocouples, resistance thermometers) dif-
fer. The use of modern sensors on the basis of semi -conducting material leads
to a significant increase in sensitivity.
2.1 Heat Flux DSC 11
3
rzzzzzzzzl~zzzhzzZl 5
S R
T( f) 1---7>'---,------.
a 6T
calibra ticn
KIT)
time, temperature ~
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, IPps heat flow rate from furnace to sample crucible,
IPpR heat flow rate from furnace to reference sample crucible, <Pm measured heat flow rate,
K calibration factor. b Measured heat flow rate <Pm (schematic curve) (according to Hem-
minger, 1994)
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 arrange-
ment is ideally symmetrical (with samples of the same kind), the same heat
flows into sample and reference sample. The differential temperature signal6.T
(normally in form of an electrical potential difference) is then zero. If this
steady-state equilibrium 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:
c;llFS- c;llFR - - AT with AT= Ts- TR
As neither ideal thermal symmetry of the measuring system at all operating

temperatures nor thermal identity of the samples can be attained in practical
application, not even outside the transition interval, there will always be a signal
ATwhich depends on the temperature and the sample properties (cf. definitions
of zero line in Sect. 5.1). Calculations in Chapter 3 are based on the assumption
that this portion of the total signal is zero or has already been subtracted from
the measurement signal proper.
In the DSC the measurement signal AT is always obtained as electric voltage.
The heat flow rate c;llm (m: measured) is internally (in the software) assigned to
this signal AT by factory-installed provisional calibration:
c;llm = -k'AT
The measurement signal output by the DSC (Fig. 2.1 b) and accessible to the user
is c;llm (in IIW or m W). The calibration of the DSC must be checked by the user:
to what extent does c;llm represent the true heat flow rate c;lltrue which is released
or consumed by the sample?
Such a test (verification of calibration) can be carried out
- by measuring the steady-state heat flow rate into a sample of known heat
capacity C by "charging" this heat capacity in the quasi-steady-state of scan-
ning with a constant heating rate f3 =dT/dt:
(2.1)
- by comparing the integral over a transition peak with the expected (known)
heat of transition Qr (energy balance):
(2.2)
(for definition of baseline 4\" see Sect. 5.1)

(When resistance thermometers are used as temperature sensors, the ther-
mometers can be used also as heaters which is advantageous in calibration pro-
cedures, see Wolf et 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 tempera-
tures between -190C and 1600C. The maximum heating rates are about 100 K
min-I. Typical time constants of the AT-sensor (empty systems, no samples) are
between 3 and 10 seconds. The noise (definition see Sect. 7.2) of the measure-
ment signal lies between 0.511W and 20 llW (it also depends on the temperature
and the heating rate). The total uncertainty of the heat measurement amounts to
about 5 %, and it is to be expected that it cannot be reduced to less than 2 % even

if more time and effort were spent (cf. Sect. 4.4).
Heat Flux DS( with Triple Measuring System
An extension of the disk-type measuring system described above was presented

by TA Instruments (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 cru-
cible, and the third for the calibration (reference) material with well known heat
capacity (Jin, Wunderlich, 1993; see even Sect. 6.1). The temperature range of
this Dual Sample DSC is from 125 to 1000 K with heating rates up to 100 K min-I.
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 DS(s
Special containers for heat flux DSCs with disk-type measuring system allow high-
er pressure, in general up to 7 MPa (70 bar), to be applied. This makes it possible
to determine, for example, vapour pressures and heats of evaporation by means of
DSC (Wiedemann, 1991; Perrenot et al., 1992, see also in Mathot, 1994b).
2.1.2
Heat Flux DSC with Turret-Type Measuring System
The characteristic feature of this recently developed measuring system is that

the essential heat flow passes from the bottom of the furnace through the jacket
of two thin-walled cylinders to the top of them which serve as sample and ref-
erence sample support (cf. Fig. 2.2). Because of the short time since introduction
of this system there are only few experiences yet, but it seems to offer some ad-
vantages against the common disk-type DSC which are:
- a small system size combined with low mass,
- a very short heat conducting path,
- a strictly direct connection of the sample and reference sample to the furnace,
- no interference (cross talk) between sample and reference sample events.
This leads to a smaller furnace as well and the thermal response time becomes
much lower, which allows faster heating and cooling rates and higher frequen-
cies in the temperature-modulated mode.
In addition to the temperatures of the sample and reference sample support
the temperature in the middle of the bottom plate To (see Fig. 2.2) is measured
in this measuring system. With these data it is possible to determine additional
quantities which are essential for the performance of the DSC and correct for
the influence of thermal inertia. The theoretical background of this so-called
R
~::::::::1
Fig.2.2. Heat flux DSC with turret-type measuring system (TA Instruments). 1 elevated con-
stantan platform for sample and reference sample, 2 chromel area thermocouple, 3 constan-
tan body, 4 chromel - constantan thermocouple, 5 silver furnace, S sample substance,
R reference sample substance,~T platform temperature difference, To body (furnace) temper-
ature
Tzero DSC technology will be outlined in Chapter 3, it is much more sophisti-

cated than the simple formulae presented in Sect. 2.1.1 for the simple disk-type
DSC. Modern computer techniques offer the possibility of expensive online cal-
culations and enable the output of the corrected heat flow rate and the calculat-
ed temperature of the sample pan even while the measurement runs.
The tiny design necessitates a very accurate production of the components
with very narrow tolerances to ensure a good symmetry which is indispensable
for the construction of differential calorimeters. However, modern technology
enables this new type of heat flux DSC which seems to be a large step in the right
direction to overcome the well-known draw-backs of heat flux DSC (see Chap-
ter 3) in near future.
2.1.3
Heat Flux DSC with Cylinder-Type Measuring System
A block-type cylindrical furnace is provided with two cylindrical cavities, each

containing a cylindrical, fixed sample container which is connected with the fur-
nace 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 (according to Calvet,1948,
Fig. 2.3), 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. 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 decou-
pled; heat exchange takes place only with parts of the massive furnace. The
3
'-----0 !1 T 0 - - - - - - - '
Fig.2.3. Heat flux DSC with cylinder-type measuring system (Calvet, 1948; thermally decou-
pled sample containers) (according to Hemminger, 1994).
1 containers to take up sample and reference sample, 2 thermopiles, 3 furnace (with program-
mable temperature controller), 4 lid, S sample substance, R reference sample substance,
1'1 T temperature difference between the containers
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:
By analogy with the disk-type DSC, the following is valid for the (steady-state)
heat flow rates exchanged between furnace and sample !PFS, between furnace
and reference sample !PFR, and for the measured heat (output) flow rate !Pm and
the true heat flow rate !Ptrue into the sample:
!PFS - !PFR - -I1T

!Pm = -k' I1T
!Ptrue = K~ !Pm
An electric voltage proportional to I1T, or a heat flow rate signal internally cal-
culated from I1T (as in the case of the 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, according to Petit
et al., 1961), only a small fraction of the heat usually flows from the furnace to the
samples via thermocouples (Fig. 2.4). The greatest part of the heat exchange be-
Fig.2.4. Heat flux DSC with modified cylinder-type measuring system (thermally coupled
sample containers) (according to Hemminger, 1994).
1 containers to take up sample and reference sample, 2 thermopile(s), 3 furnace (with pro-
grammable temperature controller), 4 lid, 5 support of containers, S sample substance, R ref-
erence sample substance, I1T temperature difference between the containers
tween furnace and samples takes place via the holders of the sample containers,
via leads and gas layers between furnace and sample containers. Thermocouples
are preferably used to directly measure the temperature difference 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 holders, leads and gas layers,
depends on the position of the heat source (sample) in the container. When the
sample comes close to the upper edge of the container, the calibration factor can
change by more than 50 %, depending on the calorimeter type (cf. Sect. 4.4).
Compared with disk-type measuring systems, the classical cylinder-type
measuring system according to Calvet has the advantage of a much larger useful
volume. In return, greater thermal inertia (time constants of up to 40 min) must
usually be put up with. The large cavity volume allows different special contain-
ers (for example, for electric calibration, gas flow, mixture, high pressure etc.) to
be used and makes intervention (electrical, mechanical, 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 rela-
tively low noise.
Heat flux DSCs with cylinder-type measuring systems with sample volumes
of about 10 mL are available for a temperature range between -196C and
1500 0c. Due to the relatively great time constants of the measuring systems, the
2.2 Power Compensation DSC 17
maximum heating rates - depending on the cavity volume - lie at about 30 K

min-I. Special instruments with large volumes (up to 100 cm3 ) are used, for ex-
ample, in biology for the investigation of small animals at constant temperature
(cf. Moratzky et al., 1993; Schmolz et al., 1993).
High temperature systems can also be used in isoperibol (isoperibol: sur-
roundings at constant temperature) mode of operation, as a drop calorimeter
(cf. Appendix 2) to measure, for example, enthalpies of mixture up to 1500C
(see, e.g., Fan et al., 1993).
Other systems with smaller sample volume (around 850 }lL) and semicon-
ductor temperature sensors are used for highly sensitive measurements (detec-
tion limit around 50 n W). They are applicable in a temperature range from -45
to 120C and can be equipped with special vessels according to various investi-
gations (batch, ampoule, fluid-circulation, mixing, electric calibration by means
of a "Joule-heater", etc.). These DSCs can usually be heated maximally with
1 K min-I, isothermal operation is possible as well (noise ca. 0.2 }lW).
For heat flux DSCs with cylinder-type measuring systems according to Calvet
(decoupled sample containers), a rather simple theory has been developed
which quantitatively describes the functional correlation between the instanta-
neous measurement signal IPm and the original event IPr in the calorimeter (cf.
Sect. 3.1). According to this theory, "desmearing" (cf. Sect. 5.4) of the measured
curve - to obtain the IPr(t) function - can be performed rather easily, even on-
line, for example, by means of a commercially available electronic devices or via
proper software on the computer.
For disk- and turret-type heat flux DSCs the desmearing procedure is not
that simple but can be performed as well. For the latter DSC type the advanced
Tzero technology makes it possible to do the necessary calculations during the
measurement run (see Sects. 2.1.2 and 5.4.3).
2.2
Power Compensation DSC
The power compensation DSC belongs to the class of heat-compensating

calorimeters (see Appendix 2). The heat to be measured is (almost totally) com-
pensated with electric energy, by increasing or decreasing an adjustable Joule's
heat.
2.2.1
Function Principle
The commercial power compensation DSC I most frequently used is an instru-

ment with an isoperibol operation. The measuring system (Fig. 2.5) consists of
two identical microfurnaces which are mounted inside a thermostated alu-
minum block. The furnaces are made of a platinum-iridium alloy, each of which
contains a temperature sensor (platinum resistance thermometer) and a heating
1 The description of a power compensation DSC is based on the widely used DSC of Perkin-
Elmer Instruments.
Fig.2.S. Power compensation DSC (Perkin-Elmer Instruments). Set-up of the measuring sys-
tem (according to Hemminger, 1994).
S sample measuring system with sample crucible, microfurnace and lid, R reference sample
system (analogous to S), 1 heating wire, 2 resistance thermometer. Both measuring systems -
separated from each other - are positioned in a surrounding (block) at constant temperature
,--___ @ Calibration ~ cf> m
tJp tJp
T (I)
) - - --'----- --+1 Re (0 r der <Pm( f)
Fig.2.6. Power compensation DSC (Perkin-Elmer Instruments). Block diagram showing the
function principle (according to Hemminger, 1994).
Ts temperature of the sample furnace, TR temperature of the reference sample furnace,
IlT = Ts - TR, Pav average heating power, IlP compensation heating power, cf>m measured heat
flow rate (measurement signal)
resistor (made of platinum wire). The micro furnace is about 9 mm in diameter,

approx. 6 mm in height and has a mass of approx. 2 g. The time constant of the
DSC without sample pans is 1.5 s, and the isothermal noise is about 2 II W. The
maximum heating power of a microfurnace is about 15 W, the maximum heat-
ing rate is 500 K min -I. Maximum cooling rate can reach 200 K min-I, it depends
on the temperature difference between block and sample (Perkin-Elmer Pyris
Diamond DSC). The measuring range extends from -175C (block cooled with
liquid nitrogen) to 725C,
During heating-up, the same heating power is supplied to both microfurnaces
via a control circuit (Fig. 2.6) in order to change their mean temperature in
accordance with the preset heating rate (see Watson et aI., 19642 ). If there is ideal
thermal symmetry, the temperature of both micro furnaces is always the same.
When an asymmetry occurs, for example, as a result of a sample reaction, a tem-
perature difference results between the microfurnace accommodating the sam-
ple and the micro furnace containing the reference sample. The temperature dif-
ference is both the measurement signal and the input signal of a second control
circuit. This second circuit compensates most of the reaction heat flow rate by
proportional control by increasing or decreasing an additional heating power of
the sample furnace. The compensating heating power I1P is proportional to the
remaining temperature difference I1T (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. 2.7).
Again, a heat flow rate <Pm is assigned to the real measurement signal I1T as a
result of a factory-installed calibration, and fed in. The relations between I1T, <Pm
and the compensating heating power I1P are as follows:
I1P=-k l I1T
<Pm =-k 2 I1T
The factor kl 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 kl - a given compensating heating power
always corresponds to a given 11 T, independent of the temperature; k2 can there-
fore in principle be determined by one calibration measurement at one temper-
ature only.
As regards the formal aspects (cf. Sect. 2.1 for the heat flux DSC), the output
signal is also given as a heat flow rate signal <Pm (e. g. in m W), and the relation be-
tween <Pm and the true heat flow rate exchanged with the sample, <Ptrue = K<[> <Pm,
must also be determined by caloric calibration (see Sect. 4.4).
At higher temperatures of the measuring system, the heat flows exchanged with
the (isoperibol) surroundings (conduction, radiation, convection) are relatively
2 In the basic paper by Watson et al. the term Differential Scanning Calorimeter (DSC) is
coined, 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 of DSC and classic DTA.
time--l>
t
tJp
tJT
t
t
Fig.2.7. Power compensation DSC. Diagrammatic view of the signals in question (according
to Hemminger, 1994).
tPr heat flow rate released in the sample (exothermic, therefore negative), qr heat released in
the sample, ~ T* temperature increase in the sample furnace which would build up due to the
exothermic effect unless it was compensated (as in heat flux DSC), ~p compensation heating
power (negative, to compensate the exothermic effect), ~T 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), tPm output sig-
nal (proportional to the (negative) amplified ~T signal)
large (several Watt) compared with the quantity to be measured which is three
orders of magnitude lower. High requirements must, therefore, be met as far as the
uniformity of the heat exchange between the two microfurnaces and the sur-
roundings is concerned in order to keep a good symmetry and the uncertainties
of measurement small. Moreover, the shares of the various heat exchange mecha-
nisms and their respective amounts must depend only on the temperature, and
strict repeatability must be ensured. This is a rather hard demand if one considers
the huge heat exchange of the furnaces with the thermostated surroundings via
radiation and convection. Consequently the conditions for this heat exchange
should be kept as constant as possible. Conclusion: The microfurnaces must be
covered with lids of the same kind in order to "cover up" possible inhomogeneity
of sample and reference sample; the thermophysical properties of the crucibles
and lids must depend only on the temperature and never on sample properties.
As a result an asymmetry of heat exchange with the surroundings of, say, 10-4
causes a differential signal of about 1 mW, the zero line of the power compen-
sated DSC is never zero and often strongly curved. The manufacturer made
allowance for these problems and took care of an electronically compensation of
thermal asymmetries of the measuring system, which become apparent as a cur-
vature of the zero line. In this way, it is also possible to "straighten" the zero line
and/or incline it as desired, and with it the measured curve outside the peaks.
Compared with heat flux DSCs, the power compensation DSC offers the fol-
lowing advantages:
- The short heat conduction path between samples and heater and the relative-
ly small masses of the microfurnaces 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).
- Reaction heat flow rates are rapidly and to a large extent compensated by elec-
trical heating power. As a result, only small temperature differences I1T re-
main between the micro furnaces of sample and reference sample (approx.
1110 of those occurring in heat flux DSCs using the same sample and refer-
ence sample). This means that the calibration factor Kcp is practically inde-
pendent of the intensity and kinetics of the sample reaction.
- The total compensating energy (J I1Pdt) 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 re-
peatable. It can be taken into account with the aid of special electronics or by
the software. A single caloric calibration is then - in principle - sufficient to
determine the correlation between CPm (or CPr) and I1P (cf. Sect. 3.2).
The user must, however, keep in mind that the temperature difference between
the two micro furnaces is not compensated totally in commercial power com-
pensation DSCs. During the peak there is still a temperature difference 11 T pro-
portional to the reaction heat flow rate. That is to say, the power compensation
DSC can be considered a kind of DTA instrument with a 11 T as the measurement
signal, but with a I1T arising from a specific thermal event which is much small-
er 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 par-
ticular 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 pronounced, they should be carefully tested and a thorough verification
(with the relevant parameters varied) should be performed if the demands on
the accuracy of the measurements are high (H6hne, G16ggler, 1989).
2.2.2
Special Power Compensating DSC
Because of the importance of this type of calorimeters several modifications

have been done for essential special purposes:
High Pressure DSC
High pressure measurements are of great importance from the thermodynamic

point of view. The change of pressure enables a better insight into the thermo-
dynamic behavior of materials. High pressure is often used during production
and processing of materials and the change of properties with pressure is
of great interest (H6hne, 1999b). While high pressure DTA already exists for
a longer time in several laboratories worldwide (see, e.g., Szabo et aI., 1969) at
pressures up to 1 GPa, the number of high pressure DSCs is very limited
(Schmidt et aI., 1994). To our knowledge there exists only one power compen-
sated DSC (based on the Perkin Elmer DSC-7), which works up to a pressure of
500 MPa. Design, specifications as well as applications have been described by
Blankenhorn, H6hne, 1991 and H6hne, Blankenhorn, 1994.
Photo-DSC
Another modification concerns the possibility to follow light induced reactions

in a 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, sufficiently rich in energy, leads to 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; Flammersheim, 1981. Since 1987, the DPA-7 of Perkin-Elmer is
commercially available as a supplementary device of the DSC-7. (Heat flux DSCs
are also available with suitable additional devices.) The principle is shown in
Fig. 2.8.
Sample and reference substance are contained in the calorimeter at constant
temperature. The sample crucibles are either open or covered by quartz disks.
The lid of the microfurnace is provided with openings sealed with quartz disks.
If light high 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
Fig. 2.S. Light transfer light source

in a Photo-DSC, schematic
representation
optics
quartz cuvette with

water as IR-absorber
monochroma tor
neutral density fil ters
LU W calorimeter
place only as long as activation energy is supplied by irradiation. High-pressure

mercury vapor 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 inter-
est is selected using monochromatic 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 applications, it is of secondary importance whether sample
and reference cells are irradiated or the sample alone. For precise measurements
electronic stabilization 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 usu-
ally very fast. The greatest part of the reaction takes place within a few seconds
(cf. Figs. 6.13 and 6.14). There are two consequences:
- There is a considerable change in the measurement signal in periods compa-
rable with that of the DSC time constant. The measured and the 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 con-
version versus time curve (see. Sect. 5.4).
- Heat flow rates are so large (see Fig. 6.13) that the sample temperature devi-
ates considerably from the temperature of the microfurnace, even in power-
compensated DSCs. If a thermal resistance of 40 K W- 1 between sample and
temperature 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 measure-
ments are no longer isothermal, not even as an approximation. Kinetic analy-
sis must allow for this.
Another Photo-DSC is described by Saruyama, 1999. In this case the light is

used to transfer energy to the sample by absorption. A commercial DSC (Rigaku
DSC 8230) has been modified to enable the light to reach the lids of the
crucibles, which were covered with carbon black for better absorption. The
light intensity was modulated in time by two polarizers, one fixed and one
rotating with a certain frequency. This way a periodical changing (sinusoidal)
heat flow rate is added to the sample which causes a temperature modulation.
The light-modulated DSC is used for similar investigations as the temperature-
modulated DSC (see Sects. 3.3, 5.5) but allows a wider range of frequencies up to
1 Hz.
Special DSC for Fluids
For precise biochemical investigations or for the investigation of reactions

in fluids, commercial power compensation DSCs are often used that are operat-
ed adiabatically or quasi-adiabatically (Privalov, Plotnikov, 1989). In a former
version of such a calorimeter the sample cell is heated with constant power
and the (heated) jacket follows the cell temperature as closely as possible (adia-
batic or quasi-adiabatic conditions). Consequences of this operation mode are
that the heating rate is not constant but depends on the behavior of the sample
(temperature change due to release of heat of reaction, change in heat capacity)
and that cooling requires a jacket temperature that is significantly below the cell
temperature to remove heat from the samples, which leads to non-adiabatic
conditions.
A power compensation calorimeter for liquids (MicroCal LLC, see Plotnikov
et al., 1997) uses twin coin-shaped measurement cells (non-removable, volume
0.5 mL) in an adiabatically (for heating) operated jacket. The cells are connect-
ed to the outside via capillary tubes. Thermoelectric sensors measure the tem-
perature difference between the cells (power compensation control loop ) and
between reference cell and jacket (adiabatic control loop ). The time constant is
approx.5 s. The temperature range is from -10 to 130C (-45 to 150C for an ex-
tended temperature range version) with scan rates up to 1.5 K min-I. Heating
(adiabatic) and cooling (nonadiabatic) are performed by means of Peltier ele-
ments. To avoid boiling of the liquids, the system can be pressurized up to 3 bar.
The repeatibility of the baseline is stated to be around 0.2 p.W, the short time
noise (RMS, s. Sect. 7.2) around 70 nW.
New calorimeters offered by Calorimetry Sciences Corp. (CSC) use a temper-
ature-leading thermostat which is heated or cooled by means of Peltier elements
according to the preset scanning rate. The cells are placed non-removable inside
a jacket and are thermally connected to the thermostat via capillary tubes which
also serve to fill, to clean etc. the cells. The temperature difference sensors be-
tween sample cell and reference sample cell are semiconductor (bismuth tel-
luride) thermopiles.
The calorimeter can be operated between -10 C to 130C (extended temper-
ature range version to 160C) with scan rates (heating or cooling) up to 2 K min- I
and with a pressure up to 6 bar. Small sample masses of 10 p.g or 50 p.g are in-
vestigated in capillary type or cylindrical cells.
2.4 Modes of Operation 25
Power compensation heaters serve to minimize the temperature difference

between sample and reference sample cells. The time constant is approx. 5 s, the
baseline noise is stated to be ~ 15 nW and the repeatibility of the baseline -
even after refIlling the cells - to be within 0.5 ]lW. These calorimeters have suc-
cessfully been used in biochemistry to investigate proteines, in particular their
phase behavior in solution.
2.3
DSC with Combined Heat Flux and Power Compensation Measuring System
The "pure" power compensation DSC measuring system uses two thermally
decoupled microfurnaces with an integrated temperature sensor and heater in
each furnace (cf. Sect. 2.2). So called "hybrid systems" show a pair of sensor-
heater combinations on a disk. The temperature difference between the posi-
tions of the sample and the reference sample is measured by the temperature
sensors and (almost) compensated for by means of controlling the integrated
heaters. A good thermal coupling between temperature sensor and related
heater is a prerequisite for short time constants and a negligible cross heat flow
between the two sensor-heater elements. This construction combines advan-
tages of both the pure heat flux and the pure power compensation measuring
systems: stable baseline, short time constants, high resolution, low noise and
small temperature differences between environment (furnace) and measuring
system.
Another version with a combined heat flux and power compensation system
uses the disk with integrated thermopile and compensation heaters as bottom
plate for two distinct cylindrical chambers which are machined in a small cylin-
drical metal block. The crucibles for sample and reference sample are put into
these chambers and are positioned on the detector disk (temperature range
-150 to 550C, detection limit approx. 0.5 ]lW, crucibles with 30 or 100 ]lL).
In principle the addition of heaters plus a power compensation control loop
is not restricted to disk-type DSCs. Whenever a heater is part of the measuring
system - e. g., a calibration heater inside a cylinder type system - this heater may
be used for compensation purposes.
2.4
Modes of Operation
DSCs are generally operated by a controlled program which changes the tem-
perature in time. We distinguish between those modes of operation which leave
the heating rate constant (i.e., the classical DSC operation mode) and those with
variable (periodical or non-periodical) change of the heating rate.
2.4.1
Constant Heating Rate
In this mode of operation the controlled program follows the time law:
(2.3)
where To is the starting temperature of the run and Po the heating or cooling (neg-
ative) rate. In other words the temperature changes linearly in time. With common
DSCs heating rates up to 150 K min- 1 (power compensated DSC: 500 K min-I) can
be reached. The real cooling rates are lower (< 150 K min-I), because the transport
of heat out of the sample needs time (a compensation is not possible).
Isothermal Mode
In this mode the heating rate [Po in Eq. (2.3)] is zero, i.e., the temperature To is
kept constant. Consequently, if no transitions or reactions take place in the
sample, there is no heat exchanged with the sample and the heat flow rate should
read zero. This is in practice not the case, as all DSCs are not perfectly symmet-
ric and the heat exchange of the sample and reference sample with the sur-
roundings is somewhat different. This results in a non-zero heat flow rate even
in the isothermal case. One reason to perform isothermal measurements in a
DSC is to measure and check this asymmetry. Isotherms are used for calculation
of heat flow rate corrections due to asymmetries, they are included before and
after scanning sections in the case of precise heat capacity measurements, for
example (see Sect. 6.1). Another reason to perform isothermal measurements is
to determine the latent heat of reactions or transitions taking place in the sample
at a certain temperature (examples can be found in Chapter 6). Of course, the
baseline (the isotherm obtained without processes of the sample) must be sub-
tracted from the measured curve to get the true reaction heat flow rate.
Scanning Mode
In this mode the temperature changes linearly in time. Every material needs a
certain heat to be warmed up and the heat flow rate is proportional to the heat-
ing rate:
dT
tPcp= Cp ' - (2.4)
dt
with Cp as the proportionality factor. In a DSC the differential heat flow rate de-
pends on the differential heat capacity and heating rate. Generally the measured
heat flow rate in scanning mode is never zero and is made up of three parts:
(2.5)
The first term on right hand side is caused by the (unavoidable) asymmetry of
the DSC, the second term is caused by the difference in heat capacity of sample
and reference sample and the third term is the heat flow contribution from a re-
action or transition (latent heat) occurring in the sample. The first two parts de-
fine the "baseline" and the third part the "peak" of the measured curve. In case
of total symmetry and absence of processes, only the sample heat capacity caus-
es a signal (and can be determined from it). Scanning is the mostly used mode
of operation with DSCs.
High-Speed Scanning Mode

Recently Pijpers et al., 2002, published a method to perform DSC measurements
at very high, controlled (including constant) cooling and heating rates of hun-
dreds of degrees per minute with sub-milligram amounts of substance. To over-
come the thermal inertia (thermal lag) of the heat transfer to the sample special
measures have to be taken: the mass of the sample has to be very low and the ther-
mal contact between DSC and sample as good as possible. This is needed to re-
duce the time constant of the respective RC-element (cf. Sect. 3.3.3) which other-
wise can be much larger than the time constant of the DSC and the controllers
causing a huge smearing (cf. Sect. 5.4) of the measured heat flow rate curve.
The method works, however, well and allows, for example, one to investigate
melting kinetics and to override recrystallization processes which are unavoid-
able when testing polymers and often indistinguishably superimposed at lower
heating rates.
2.4.2
Variable Heating Rate (Modulated Temperature)
In this case a certain modulation term is added to the linear part of the temper-
ature-time function. The simplest and most used modulation type is the peri-
odic (harmonic, i. e., sinusoidal) one and Eq. (2.3) reads:
T(t) = To + Po t + TA sin (wt) (2.6)
with TA the amplitude and w the (angular) frequency of the modulation. To-
gether with Po, the "underlying" heating rate, there are three parameters which
can be chosen freely within certain limits. This influences the heating rate, which
follows from Eq. (2.6) as:
dT
- = Po + TA W cos (wt) (2.7)
dt
Depending on the relation of Po to TAW, we have to distinguish four cases (Fig. 2.9).
Quasi-Isothermal Mode
In this mode the underlying heating rate is zero and the temperature varies
around a constant temperature (see Fig. 2.9 curve a). As the amplitude of the
modulation is low (0.01 to 0.5 K normally) the temperature is almost constant
Fig.2.9. Temperature-mod- S
I
ulated modes of operation:
(a) quasi-isothermal,
4
(b) heating-cooling, (c) heat-
ing-iso, (d) heating-only
3
ClJ
'--
::J
~
d
2
'--
ClJ
Cl.
E
ClJ
0
-1
0 10
time
(quasi-isothermal). The heating/cooling rate varies between TAW. The advan-
tage of using this mode is the possibility to determine heat capacities even in the
isothermal case (see Sect. 6.1.5) what otherwise is not possible.
Heating-Cooling Mode
In this mode the amplitude is so large, that the heating rate (Eq. 2.7) changes its
sign periodically (see Fig. 2.9 curve b),i.e., the sample is periodically heated and
cooled. This is the case if the product TAW is larger than the underlying heating
rate Po.
The advantage of this mode is a rather large heating rate and therefore good
signal-to-noise ratio which is in particular useful if we have small sample masses.
The disadvantage on the other hand is that the sample is heated and cooled
during the run, this may cause problems in some cases, e. g., for polymers where
different processes (with different time constants) could occur within the heating
and cooling period. This would complicate the evaluation of the processes.
Heating-Only Mode
In this mode the amplitude is so small that the heating rate (Eq. 2.7) is always
positive (see Fig. 2.9 curve d), i.e. the sample is only heated and never cooled dur-
ing a run. This is the case if the product TAW is smaller than the underlying heat-
ing rate Po.
The advantage of this mode is that all processes only occurring during cool-
ing of the sample are suppressed and the remaining processes are easier to eval-
uate. The disadvantage is clear from the rather low amplitude which gives a bad
signal-to-noise ratio for all quantities derived from it.
Heating-Iso mode
This is the limiting case between the two modes above where TAW = Po: the
heating rate varies periodically between TAW + Po and zero. This is the mode
Fig.2.1O. Special tempera- 5

ture-modulated modes
t
of operation: (a) step-scan,
(b) sawtooth 4
3
-
<lJ
L..
:::I
0
2
L..
<lJ
CI..
E
(lJ
10
time ..
with maximum of amplitude (and signal-to-noise ratio) but without cooling the
sample (see Fig. 2.9 curve c).
Another advantage is that, at heating rate zero, the (even periodical changing)
heat flow rate has a minimum with no contribution from the sample heat
capacity [which follows from Eq. (2.4) and (2.7)]. This way the lower contour of
the fluctuating heat flow rate curve can be used to detect endo- or exothermic
processes occurring in the sample. After subtracting of cPo there is only cPr re-
maining in the moment of minimum (i.e. zero) heat flow rate [ef. Eq. (2.5)].
Sawtooth Mode
In this mode of operation different linear scanning rates change periodically

(Fig. 2.10 curve b). Again we can distinguish between heating-cooling, heating
only, and heating-iso modes of operation, depending on the choice of the re-
spective rates of temperature change.
Every DSC can be operated in this mode, as programming of several different
linear scanning ramps in one run normally is possible without special addi-
tional electronics or software. This is one advantage of this mode of operation.
Another advantage follows from the fact that every (non-sinusoidal) periodical
function can be expanded as a Fourier series (see textbooks of mathematics).
In the case of the sawtooth mode, the periodical function, superimposed on the
linear temperature change, reads:
_ (2n 1 3 . 2n 1 5 . 2n )
T(t) =c sin--2"sin--+2"sin--- ...
tp 3 tp 5 tp
with c as a proportionality factor containing the amplitude and period tp of

the periodic modulation. This function contains not only the basic frequency
f = lItp but also (odd) higher harmonics. The same is true for the measured heat
flow rate. By Fourier analysis the modulated heat flow rate at different frequen-
cies can be determined from one measurement.
A disadvantage of this temperature modulation mode is the discontinuous

(step-like) change of the heat flow rate, which may cause problems with the fur-
nace controller (overshoot) and produce additional (non-linear) heat flow rate
fluctuations.
Step-Scan Mode
The step-scan mode, recently introduced by Perkin-Elmer Instruments for

power-compensated DSCs and by Mettler-Toledo for heat flux DSCs, is a special
mode of operation: After changing the sample temperature, say, by 0.5 K at
a high heating rate (step-like, see Fig. 2.10 curve a) the heat flow rate is allowed
to equilibrate again. The respective response is evaluated. This can be done many
times in a periodical or non-periodical manner. In the periodical case the step-
scan is a special sawtooth mode, which is similar to the heating-iso mode in the
sinusoidal case, but with an asymmetric triangle shaped form of the tempe-
rature-time curve. This results in more than only the odd harmonics which
one can separate from the modulated part of the heat flow rate signal. Another
advantage is that a series of temperature steps is equivalent to a series of heat-
ing rate pulses. The heat flow rate response on a pulse-like event enables one
to determine the apparatus (Green's) function of the DSC with and without a
sample at every moment of the run. This offers an easy possibility to come to
correct ("desmeared", cf. Sect. 5.4.3) reaction heat flow rates as well as to mea-
sure the complex heat capacity Cp(w) from one run (see Sects. 4.7, 5.5.3).
3 Theoretical Fundamentals
of Differential Scanning Calorimeters
In all DSCs, a temperature difference AT - given as a voltage - is the original

measurement signal. In almost all instruments a heat flow rate tl>m (differential
heat flow rate) is internally assigned to AT ( cf. Chapt. 2). Independent of whether
the user obtains AT or tl>m from the respective DSC, knowledge of the functional
relation between the measured signal (AT, tl>m) and the quantity searched for
(the real heat flow rate tl>r consumed/produced by the sample) is important for
- the time-related assignment of tl>r to AT or tl>m (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
properties of the measuring system with regard to this relation,
- the estimate of the overall uncertainty of measurement.
The relation between tl>r and AT or tl>m can be derived in varying degrees of
approximation to real DSCs. Analytical solutions of the differential equations
are possible only for simple boundary and initial conditions and for quasi-
steady states. Numerical procedures and solutions can approximate the actual
conditions more exactly, however, without the clarity of the functional relations
given by an analytical solution. Basic considerations in this field are given by
Gray, 1968.
To ensure better differentiation from tl>n in the following section, AT - instead
of tl>m - is assumed to be the measurement signal, i. e., we search for the relation
tl>r(AT). These two quantities, AT and tl>m, are under normal circumstances
strictly proportional, with the exception of the opposite sign.
3.1
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.
32 3 Theoretical Fundamentals of Differential Scanning Calorimeters
Fig.3.1. Heat flux DSC /

(disk-type), model for zeroth
/ tJ!
approximation (linear
model).
S sample, R reference
sample, F furnace, A cross
Tr
lS T,
section of the heat conductor
between furnace and Sand
R,M distance between tem-
perature measurement point
and furnace F F
Ts T. / /
Zeroth Approximation
First, the heat flux DSC is represented by a very simple linear model (Fig. 3.1).
The following simplifications have been made in this approximation:
- steady-state conditions (i.e., constant heat flow rates),
- only one thermal resistance, the apparent resistance between furnace and
sample, is taken into account with no interaction (cross-talk) between sample
and reference sample,
- only the heat capacities of the sample and reference sample (Cs , CR) are taken
into account, other heat capacities are neglected,
- sample temperature and measured temperature are assumed equal,
- no heat exchange with the surroundings (i. e., no heat leak).
Figure 3.2 shows the equivalent electric circuit diagram for the zeroth approxi-
mation. This diagram serves to better understand the interrelations. From the
physical point of view, electric charge transport and heat transport are equiva-
lent processes, and many people find it easier to read electric circuit diagrams
than to visualize heat flows in real equipment.
R,s
'fS
Fig. 3.2. Equivalent electric
circuit for the linear model
of the heat flux DSC fjU
(see Fig. 3.1).
C capacitance, R resistance,
if. R'R
i current, U voltage, Sub-
scripts: S sample, R refer- U,
ence, F furnace
[.
T T
[s
O----------------T---------~----~
/77J777
The Biot-Fourier equation of the (steady-state) conduction of heat, together

with the formulation in absolute values, reads as follows:
~ I~I
- = - A . grad T or - = - A 1 grad T 1
A A
The amount of the heat flux tP/A is proportional to the gradient of the tempera-
ture; the thermal conductivity .\ 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.l (with Tp> Ts,
TR in scanning mode):
In the case of thermal symmetry, tPps = tPPR is valid, with Ts = TR.

If a constant (exothermic) heat flow rate (tP r < 0) is produced in the sample,
Ts increases by t:.Ts, the temperature difference Tp - Ts and thus the heat flow
rate tPps decrease. When the steady-state (i.e., uniform heating rate) is reached
again, for reasons of balance, the change of tPps, (t:. tPps), must be equal to tP r :
Nothing has changed on the side of the reference sample, hence:
Consequently
(Here, Ts > TR,hence tP r is negative: exothermic effect; t:.T = t:.TSR: measurement

signal)
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 tPr and
the measurement signal t:. T. The conditions of constant heat consumption can
be achieved when in scanning operation sample and reference sample have dif-
ferent, temperature-independent "heat capacities". A greater amount of heat
will always flow into the sample whose heat capacity is higher, in order that the
steady-state heating rate is maintained. With Cs > CR the following is then valid
for the difference between the heat flow rates to sample and reference sample:
There is no steady state during sample transitions or reactions; the above

approximation does not apply in these cases. Furthermore, Cs and CR (and thus
I:! cPSR) 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 approxima-
tion:
(3.1)
This relation describes the shape of the baseline and is the basic equation to de-
termine the heat capacity Cs . For the empty reference crucible (CR = 0)
I:!T
Cs = - K ' -
P
is valid.
When the heat capacity CR is known (i. e., if CR = Cref "# 0, using a reference
material), we get
I:!T
Cs=CR-K' . -
P
In practice, first a zero line I:!To will be recorded with both crucibles empty (to
check the asymmetry of the apparatus) which is subtracted from the measured
curves (for the measurement of heat capacities, see Sect. 6.1).
First Approximation
In the first approximation, non-steady-state processes in the sample are also

permitted which manifest themselves as "peaks" of the measured curve. This
means that I:!T is not constant in time. For the rest, the simplified arrangement
as for the zeroth approximation is used here.
cPps is the heat flow rate from the furnace to the sample, cPr(t) the time
dependent heat flow rate produced inside the sample (reaction, transition).
The following balance for the heat flow rates is then valid for the sample of heat
capacity Cs:
dTs
Cs - = cPps - cPr
dt
(exothermic: cPr negative, endothermic: cPr positive)
Withl:!T= Ts - TR ,
dTR dl:!T
Cs - - + Cs - - = cPps - cPr
dt dt
results.
Accordingly, the following holds for the reference sample (cPr = 0 by defini-
tion):
dTR
CR - - = cPFR
dt
When the difference between the two balance equations is calculated, the fol-
lowing is obtained:
The following is valid for the heat flow rates cPFS and cPFR :
and
where R FS and R pR are global heat resistances between the furnace and the sam-
ples. In the case of thermal symmetry, R FS =RFR =R, thus:
flT dTR dfiT

cPr = - - - (Cs - CR) - - Cs - (3.2)
R dt dt
This equation links the reaction heat flow rate cPr searched with the measured
signal fl T. 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 signal flT(t) appears. In analogy to the charging or discharging of a
capacitor of capacity C, a time constant T can similarly be defined for the heat
flow rates:
T=CsR
When fl T is changed, R is the effective thermal resistance to the "charging or dis-

charging" 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):
flT(t) T dfl T(t)

cPr (t) =- -R- - (Cs - CR) /3 - R . ~ (3.3)
The measured signal flT is not (I) proportional to the heat flow rate cPr at a given
moment but delayed with time and thus also distorted ("smeared"). Further-
more, when Cs "# 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
t
DoT
" time --..

"
Fig.3.3. Measured curve of a heat flux DSC (schematic) with heat released inside the sample
(exothermic effect).
t1T measured signal, R thermal resistance between furnace and sample, f3 heating rate, t1C =
Cs - CR difference between the heat capacities of sample and reference sample (here negative,
because CR > Cs), Q) peak area (exothermic effect); it is a measure of the heat released between
t, and tz ,@ area below the baseline; it is a measure of the heat required to heat the sample be-
tween t, and tz
-R . p. (Cs - CR) which is the initial deviation after the quasi-steady state has been
reached in scanning operation [cf. Eq. (3.I)]. This contribution is the measured
curve before/behind a peak which is parallel to the abscissa if R, l1C = Cs - CR and
p are constant (Fig. 3.3).
In reality, the term R (Cs - CR) 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. [NB: Eq. (3.3) is in principle
the so called Tian equation, originally derived for cylinder-type calorimeters.]
Regarding Eq. (3.3), the conclusions are as follows:
1. When the signall1 T measured at a given moment is to be assigned to the heat
flow rate IPr by which it is caused, the third term in Eq. (3.3) must be taken into
account (cf. desmearing, Sect. S.4).
R must be determined by calibration (cf. Sect. 4.4); the time constant T can
also be obtained from calibration measurements (cf. below "Higher-order ap-
proximations" and Sect. 7.2).
For t> to, the solution of the differential Eq. (3.3) for a heat pulse (/Jr at the
moment to has the form (see textbooks of mathematics)
Fig. 3.4. Measured 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
~~r-~-----------~-~-----
t-T-j
t
IlT
t, time--....
with T = Cs ' R. Up to the moment t = to, the solution is the steady-state

curve
For t ~ 00, the function ~T(t) returns to this curve (cf. Fig. 3.4).
2. For the total heat of reaction or transition Qr developed/consumed in the
sample, the following balance equation is valid:
where t1 , t2 are the beginning and end, respectively, of the peak.

With Eq. (3.3) inserted, the following is obtained:
1 [t2 t2 ] t2 T d~T
Qr=-- f~T(t)dt-f(-R~C'f3)dt - f - - d t (3.4)
R tl tl tl R dt
The content of the square brackets corresponds to area CD, i. e., the so-called
peak area between the measured curve and the (interpolated) baseline (defi-
nition see Sect. S.1) in Fig. 3.3.
- When ~C and R are not temperature- (or time-)dependent, the measured

curve outside the peak is a parallel to the abscissa. In this case, d~T/dt is
zero before and behind the peak. The contribution of the 3rd term vanish-
es 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 be-
fore 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.
Fig. 3.5. Partial integration

of a peak (heat flux DSC,
exothermic effect).
CR > Cs , CD partial peak area
(between tl and t*),@ area
below baseline (between tl
and t*); symbols see Fig. 3.3
6.T
t
t, time - .
3. For the partial integration (Fig. 3.5) of the peak between t, and t*, the contri-
bution of the 3rd term must be taken into acount at the point t*:
Qr(t*} = -~ [Y ilT(t} dt- f (-R ilCf3) dt] - f ~. dilT dt

R tl tl tl R dt
\. y ) '-v-------l
partial peak area correction term
Partial integrations of the peak are, for instance, necessary for kinetic investiga-
tions and to determine purity (see Sects. 6.3 and 6.9).
Higher-Order Approximations
The temperatures of sample and reference sample (assumed here to be homo-

geneous) are not measured directly. There is a certain distance and a thermal
/
Fig. 3.6. Heat flux DSC,
(one half), model of the 2nd
/
approximation.
S sample, F furnace, TMS tem-
perature at sample measure-
ment point
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, ll.U corresponds to ll.T, UF corresponds to TF, i corresponds
to tPr RFM thermal resistance between furnace and measurement points Ms , MR , R MS 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, CR
heat capacity of the reference sample, CFM apparent heat capacity between furnace and mea-
surement points Ms, MR
resistance between the temperature measurement points and the respective

sample (Fig. 3.6). Depending on the design of the measuring system, the resis-
tance is made up of several parts differing in quantity and originating in the
transition layer between sample and bottom of the crucible on the one hand and
crucible and support on the other hand as well as further resistances between
support and temperature sensor. When the sample temperature changes, the
temperature measurement point reacts only after some delay. The analogue
electric circuit diagram of a DSC in this "2nd approximation" is shown in Fig. 3.7.
The following is valid for this so-called "thermometer problem":
dTMs
TMS = Ts - r2 - -
dt
TMS temperature of the measurement point (e. g., junction of the thermocouple)
for the sample
Ts sample temperature (assumed homogeneous)
r2 characteristic time constant for the temperature relaxation between sample
and measurement point.
In analogy, the following is valid for the reference side:
dTMR
TMR =TR - r 2 - -
dt
(r2 is assumed to be equal for both sides).
With these equations, the following results for the difference:
dTMS dTMR)
flTsR = Ts- TR= TMS - TMR + T2 ( - - - - -
dt dt
fl Tm = TMS - TMR is the measured temperature difference. In other words, with

known T2 it is possible to calculate the temperature difference between sample
and reference sample from the temperature difference measured at the position
of the temperature probe. This was made use of in the so-called advanced Tzero
technique for the turret-type DSCs recently introduced by TA-Instruments (see
Sect. 2.1.2). However, the following relation results in analogy to the mathemat-
ical procedure of the 1st approximation if we use fl TSR instead of fl T:
iPr (t)
1[
= - Ii fl Tm
dflTm
+ R . (Cs - CR) f3 + T 1 dt + TIT 2 ~
d2flTm J (3.5)
[Tl = T from the first approximation, see Eq. (3.3) 1

In addition to the 1st derivative (slope) of the measured curve flTm(t), the 2nd
derivative (curvature) must be used in the 2nd approximation to get the heat
flow rate converted in the sample. The two time constants must be determined
experimentally by proper calibration procedures. The first time constant Tl is de-
termined by the apparent thermal resistance and heat capacity between furnace
and temperature sensor. The second time constant is determined by the appar-
ent thermal resistance and the "effective" heat capacity between sample and
sample temperature measurement point. Exact symmetry between the sample
and the reference side is assumed in this calculation.
The above-described approximation can be refined as desired: Possible asym-
metries can be taken into account by introducing different thermal resistors and
capacities on the sample and reference side into the network. This results in more
complicated equations which, however, can be solved. This enables one to cor-
rect not only the temperature but even the measured heat flow rate for possible
influences of the asymmetries (e. g., a curved zero-line, which otherwise must be
subtracted afterwards) during the measurement. Certain heat transfer problems
and a temperature gradient inside the sample and its influence on the peak
shape can be taken into account by introducing further thermal resistances and
capacities in the network of Fig. 3.7. A thicker sample with poor conductivity
can be considered as having been split up into different layers which are linked
with one another by heat-conducting boundary layers. In the end an additional
differential quotient in the differential equation and another time constant re-
sult for each additional thermal resistance in connection with a heat capacity.
The solution of such a refined approximation is as follows:
1 [ dflT d2flT
iPr(t)=-- flT+k o +k 1 --+k 2- -
d 3 flT
2 +k3 - - 3 + ...
J
Reff dt dt dt
In reality, the constants ki are terms into which the thermal resistances and ca-
pacities (and thus the time constants) of the arrangement enter. Calculation of
the reaction heat flow rate r/Jr presupposes that all ki and the measurement signal
AT and its time derivatives are known. It can be shown (Loblich, 1985) that, for
practical application, the 2nd order differential equation is a sufficiently good
approximation to calculate the true desmeared reaction heat flow rate; only the
time constants f1 and f2 must be known for this purpose. For a detailed treatment
how to determine the time constants experimentally, see Loblich, 1994.
So far we have only considered the case where it is assumed that no heat ex-
change takes place between sample and reference sample, which is true for the
turret-type DSC as well as for some of the cylinder-type DSCs. This simplifica-
tion is certainly not permissible for the widespread disk-type measuring sys-
tems. 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 use the analogue electric circuit represented in Fig. 3.9
for the 2nd approximation (Rohne, 1983). In order to formulate the desired dif-
ferential 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 cir-
cuit. For the circuit of Fig. 3.9, five equations for both voltages and currents are
then obtained. On the basis of the laws of electricity, these 10 equations are sim-
plified and combined to form one differential equation which is retranslated
into the language of heat transport. The result for a symmetric twin design (i.e.,
R FMS = RFMR = R, CFMS = CFMR = C) is as follows:
(3.6)
/
T,/
'--~==:='--~~---'%
-- 'R
/JT
Fig.3.8. Disk-type measuring system of a heat flux DSC (to calculate the tIIr(~n dependence).
Ms temperature measurement point on sample side, MR temperature measurement point on
reference side, tIIr heat flow rate produced/consumed by the sample
UF
Fig.3.9. Equivalent electric circuit for the disk-type measuring system according to Fig. 3.S.
S sample, R reference sample, M temperature measurement point, F furnace, U voltage, i current,
R resistance, C capacitance (cf.legend to Fig. 3.7)
This equation is similar to that of the 2nd approximation [Eq. (3.5)]. However, as
there is a 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 a sufficiently great distance from peaks
is d TRid t equal to the heating rate f3, and the second derivative of the reference
sample temperature is equal to zero only there. For the steady-state (st) case,
dilT d 2ilT
with tPr = 0, - - = 0 and - - 2 - = 0 the following is then valid and characterizes
dt dt
the baseline:
CR - Cs
ilTst = . f3
1 2
-+--
R RMM
When Eq. (3.6) is integrated to get the area of a transition peak, the following is
valid with the approximation dTR/dt::::: f3:
f12 tPr (t)dt = Qr = - ( -1 + -

II R RMM
2) rf ilT(t)dt- f 12 CR - Cs . f3dt ]
12
II II 1 2
-+--
R RMM
= - (R1+ -RMM2)
- . f (ilT(t) - ilT.t) dt
12
II
This integral describes the peak area between the measured curve and the (in-
terpolated) 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 ::::: R MS )'
The factor (l/R + 2IR MM ) 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 T 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. De-
pending on where sample and reference sample are arranged on the disk, RMM
(thermal resistance between sample and reference sample) change, meaning
that high reproducibility for the location of sample and reference sample is of
great importance for the ability of a DSC to be calibrated.
In the approximation Cs ::::: CR and RMS ::::: RMR , Eq. (3.6) changes into a 2nd order
differential equation of the following form:
The solution of this equation, i. e., the curve fl.T (t) as measured by the calorime-
ter, for a pulse-like CPr at t = 0 is the sum of two exponential functions:
(3.7)
where there is a complex dependence of the time constants Tl, T2 on the coeffi-
cients K, K1 , K 2 Here, T2 is determined in approximation by C . R, and Tl by
Cs ' RMS (cf. Eq. 3.6). Such a peak, produced by a heat pulse, is shown in Fig. 3.10.
The time constant Tl essentially determines the ascending slope, T2 the descend-
ing slope.
Fig.3.10. Measured signal

/:"T of the heat flux DSC for
a heat pulse generated in the
sample at the time to (sum
of two exponential functions
t
with the time constants 11 and
12 in 2nd approximation)
fl.T
1---+- - - - - - - -=-::-=-=-,...,-~------
to time
Numerical Simulation
As the equations become increasingly complex and as it is impossible to solve
them analytically without introducing simplifications, it is no longer recom-
mended to set up differential equations for calculations. This is in particular
true if we want to cover heat exchange by convection and radiation and include
non-linearities due to temperature-dependent thermal resistances and heat ca-
pacities. In this case, numerical simulation by the finite-element method could
be applied. Appropriate software is commercially available and runs on modern
PCs without problems. When the required time and effort are spent, the tem-
perature and heat flow fields and the measured AT(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" with given material properties for which the heat
flow rates and temperatures are calculated. On the basis of considerations with
regard to the energy balance, a large system of equations is obtained which is
solved by conventional methods. To show the essential results, this method has
been applied to a very simple one-dimensional model of a disk-type measuring
system as well as a real commercial disk-type DSC in one-dimensional approx-
imation (for details see Hohne, 1983).
The results for a simulated melting peak are represented in Figs. 3.11 to 3.13.
As can be seen, the shape of the peak depends strongly on the given conditions
and parameters. The position of the peak maximum, for example, changes with
the heating rate, with the thermal conductance of the sample and with the mass
(or heat of transition Qr) of the sample. The slope of the descending part is de-
termined by the heating rate and the thermal conductance of the sample. Only
the extrapolated peak onset temperature is relatively independent of experi-
mental parameters. This is why, together with the area, this temperature (Te) is
preferred to characterize the peak. In contrast to this, the peak temperature
maximum (Tp) and the peak width are not values suited to characterize transi-
tions (for definition of the characteristic temperatures, see Sect. 5.1). Figure 3.14
Fig.3.11. Dependence of
the peak shape on the
heating rate /1, calculated
by numerical simulation for
a heat flux DSC
1
bJ
Fig.3.12. Dependence of
the peak shape on the heat of
transition Q" calculated by
numerical simulation for a
heat flux DSC (heating rate:
2 Kmin- I)
1
6.[
time ..
Fig.3.13. Dependence of the
peak shape on the thermal
conductance L of the sample,
calculated by numerical
simulation for a heat flux
1
DSC (heating rate: 2 K min-I)
6.T
L=O.OOS W/K
shows the numerical simulation of a thermal event which is caused by the melt-
ing of a small sample of indium in a real commercial disk-type DSC. The result
is very similar to the real measurement in such a DSC.
It has furthermore been possible to show that the calibration factors KQ =
I Qrlpeak areal and Kw = 14'rf.6.TI clearly depend on measurement and sample
parameters (e. g., Qr, A.ample, f3) when radiation and convection are included into
such model calculations (Hahne, 1983). Table 3.1 gives the results for different
parameters as an example. The dependence of the calibration factor on such
parameters is a fact which is of importance in practical scanning calorimetry
and which must by all means be taken into consideration during calibration
(cf. Sect. 4.4). The obvious temperature dependence of the calibration factor is
0,25
!::,JIK
0,20
0,15
0,10
0,05
~

..
25 50 75 100 125
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 K min-I)
Table 3.1. Numerical simulation of a commercial disk-type DSC: Changes of the two calibra-
tion factors K<[> and KQ depending on changes of the heat of reaction (Qr), emissivity (E), heat-
ing rate ({3), density (J) and thermal conductivity (A) of the sample, respectively".
The initial standard parameters (belonging to K<[> = 1 and KQ = 1) are given in parenthesis
Parameter which Numerical value of the K<[> KQ

is varied parameter
QrinJ (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
(10) 1 1
1 1.001 0.930
0.1 1.004 0.875
0.01 0.991 0.870
(7310) 1 1
73100 0.998 0.935
731 1.015 0.885
(157) 1 1
15.7 0.998 1.016
1570 1.003 0.999
a The parameters have been changed in (unrealistic) steps of 10 only, to give a better impres-
sion of their influence.
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 heat ex-
change with the surroundings via radiation and convection which is strongly
temperature-dependent (non-linear heat leak). The calibration factor is also
strongly influenced by the surface quality (emissivity c) of the crucibles con-
taining sample and reference sample, which therefore should be well defined
and must not change during the measurement.
As far as steady-state reaction heat flow rates are concerned, the above calcu-
lations have shown that the calibration factor Kcp = - ([Jr/l1 T st also depends on the
quantity ([Jr itself (which is proportional to the difference between the heat
capacities, Cs - CR) when radiation heat exchange is included.
In addition, the calibration factors used for peak integration (Le., for deter-
mination of Qr) and for heat flow rates (e. g., for determination of Cs) are gener-
ally not equal. The difference in the calibration factors Kq, and KQ can also be
concluded from the following considerations. In a DSC there are always temper-
ature differences between sample and reference sample during the measure-
ment. As a result, the heat exchange between sample and reference sample and
the respective environment varies. As the radiation exchange increases, non-lin-
early with temperature, the overall heat exchange is also non-linearly linked
with the temperature difference between sample and reference sample. As a con-
sequence, the measured temperature difference (and thus the measured differ-
ential heat flow rate ([Jm) is non-linearly linked with the true heat flow rate into
the sample ([Jtrue' This results in the calibration factor Kef> in the (linearly formu-
lated) Eq. (2.1) becoming a function of the differential heat flow rate ([Jm itself:
(3.8)
As the measured heat flow rate ([Jm depends on sample parameters (Cp, m) and
on the heating rate, Kq, implicitly depends on these quantities as well. Further-
more, 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 peak integration. If, how-
ever, Kq, depends on ([Jm, this factor is not constant and cannot be moved in front
of the integral, the integration of Eq. (3.8) then yields:
From a comparison with Eq. (2.2) it follows that KQ is an integral mean value
over the function Kq,( ([Jm) in the region of the peak. As ([Jm may considerably vary
during a peak, the difference between Kq, and KQ is normally significant (see
Sect. 4.4.3). In these cases, a quantitative peak evaluation may lead to results af-
fected 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 all types of heat flux DSCs:
- For steady-state conditions
is valid [ef. Eqs. (3.1) and (3.8)].

With this function, for a known heating rate fl, the unknown heat capacity Cs
can be determined from the measured L1Pst It must be borne in mind that
the calibration constant Kq, depends on the difference of the heat capacities,
Cs - CR.
- For non-steady-state conditions (i.e., within a peak):
- The shape of the peak changes as a function of heating rate, thermal con-
ductivity, heat capacity and shape of the sample, and of the amount of the
heat of transition. As a result, the peak parameters (width, maximum,
height) change as well; only the extrapolated peak onset temperature Te is
to a certain degree independent of sample parameters.
- The calibration factor Kq, (true heat flow rate/measured heat flow rate) is
not constant but a (weak) function of the heat flow rate itself.
- The calibration factor KQ (heat of transition/measured 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 (emissivity), and sample position.
- The calibration factor for peak areas (KQ) differs from the calibration fac-
tor for steady-state heat flow rates (Kq,).
It is basically recommended that in DSC measurements the best symmetry pos-
sible between sample and reference sample be ensured (CR :::: Cs , RMR :::: R MS ,
identical sample containers). The results of the model calculations are con-
firmed by practical measurements (Sarge et aI., 1994).
3.2
Power Compensation DSC
In ideal power compensation DSCs, each L1T 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 would be mea-
sured directly, this signal would directly stand for the heat flow rate into the sam-
ple searched for. The calibration factor would then be identical to unity and
(neglecting "smearing" of the measured signal) Pr = Pm would be obtained.
There exists 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 T between the controlled heater and the sample location must
be taken into consideration. This leads to a measurement signal Pm which is
"smeared" in comparison with the processes inside the sample, because of the
thermal lag, and a differential equation of at least 1st order is therefore obtained
to describe the behavior:
Without exception, commercial power compensation DSCs are instruments in

which a temperature difference always occurs between sample and reference
sample (as is the case in heat flux DSCs). This temperature difference serves,
firstly, as the primary measurement signal and, secondly, it is used to electrically
compensate the measurement effect by means of a proportional controller (cf.
Sect. 2.2). However, the proportional control can never completely compensate
the measurement effect so that, even in these calorimeters, a ATsR remains be-
tween the micro-furnaces of sample and reference sample, and this both in the
steady state and during a peak.
The following is then valid:
with kprop set by the proportional control circuit.

O'Neill, 1964, analyzed such a system with a proportional controller; the
theoretical analysis of two different systems of power compensation DSC,
one with P (proportional) and one with PID (proportional, integral and differ-
ential) temperature control of the sample micro-furnace has been presented by
Tanaka, 1992.
As different temperatures also result in a different exchange of radiation
and convection 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 differ-
ences are much smaller than in heat flux DSCs. For power compensation DSCs
at present commercially 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, leading to a better res-
olution in time,
- the calibration factor is not exactly equal to 1 (i. e., calibration is necessary),
- the calibration factors for steady-state (heat capacity measurement) 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 (Hahne, Glaggler,
1989). As had been expected, the effects are, however, substantially smaller than
with heat flux DSCs. When power compensation DSCs are calibrated, this de-
pendence on parameters must be taken into consideration.
When heat capacities are measured (see Sect. 6.1) the reference micro-
furnace 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 losses measured in isothermal mode before and
after the scan mode (for details, see PoeBnecker, 1993 and Sect. 6.1).
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 the case if the DSC 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. The transfer theory has in par-
ticular been used successfully to describe the behavior of DSCs in temperature-
modulated mode (see Sects. 3.3 and 4.7). In the case of power compensated DSC
the situation is more complicated because of the control electronics involved.
For the most frequently used equipment from Perkin-Elmer it has been shown
that this DSC can be considered as a linear apparatus in the first approxima-
tion (Tanaka, 1992; Hohne, Schawe, 1993; Schawe et al. 1993, 1994). However, any
asymmetry between sample and reference side disturbs the linear behavior, and
the theory of linear response is strictly speaking not valid. This has con-
sequences for evaluation procedures which require linear response, such as de-
convolution (Sect. 5.4) in particular in peak regions of a measured curve.
Note:
In contradiction to what has been presumed in this chapter, every real DSC is not
strictly symmetric in its functionality. Even in the empty state, the temperatures
of the sample cell and of the reference cell are not equal so that a residual tem-
perature difference will arise, which may change with temperature during a scan
run. 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 curved shape 4'0(T). This so-called zeroline, which will mostly be
measured with empty crucibles in the calorimeter, is a function, which is addi-
tive to the measurement signal which stems from the sample processes. In every
real case, this function should first be subtracted from the measured curve be-
fore evaluations following the theoretical considerations can be made. We there-
fore must insert the term 4'm - 4'0 (or AT - ATo) instead of 4'm (or AT) in all
formulas used so far, this has been omitted in this chapter for clarity.
3.3
Temperature-Modulated DSC (TMDSC)
Different temperature-modulated calorimetric techniques have been known

since the beginning of the last century [for details see (Gmelin, 1997)]. Gobrecht
et al., 1971, were the first who ran a DSC with periodical changing temperature
to measure complex heat capacities. It is remarkable that the authors already in
this early paper point out that this method offers the possibility of "quasi-iso-
thermal investigation of heat capacity changes during annealing procedures or
chemical reactions", which nowadays belong to the most successful applications
of TMDSC. They even mention: ''A combination of linear and periodic heating
offers the advantage of the good temperature resolution of slow scan speeds and
the higher output signal due to the faster oscillations".
Unfortunately this article was published in a physical journal and was not
noticed by the manufacturers and thermoanalytical community. That's why another
20 years passed before M. Reading and coworkers came out with a temperature-
modulated DSC (Reading et al., 1993; Sauerbrunn et al., 1992) and the method be-
came commercially available. During the past decade this method became wide-
spread and it has proved its worth with numerous successful applications.
We want to present the fundamental ideas and the theoretical background
of TMDSC in what follows, but we restrict ourselves to present only those
facts which are absolutely essential for the practitioner. The complete theory
of TMDSC would go far beyond the scope of this book, in particular as it is still
a matter of development and discussion. The interested reader is referred to
the original literature. Of course we are not able to give a complete list of the
numerous authors which have contributed to the theory of TMDSC during the
recent years. Many of them are named in review articles on this topic (see, e.g.,
Simon, 2001) and in special issues about temperature-modulated DSC of Ther-
mochimica Acta (1997, 1999, 2001) and Journal of Thermal Analysis and
Calorimetry (1998).
3.3.1
The Temperature-Modulated Method
At the beginning of TMDSC the common (linear) temperature program of the

DSC was superimposed with a sinusoidal temperature fluctuation:
T(t) = To + (Jot + TA sin(wt) (3.9)
with {Jo the underlying heating/cooling rate, TA the temperature fluctuation am-
plitude and w = 2 n f the angular frequency of modulation. But other periodical
temperature fluctuations (sawtooth-like, triangular, rectangular, step-like, etc.)
are possible and have been used as well. From mathematics it is known, that
every periodic function can be written as a Fourier series and Eq. (3.9) can easily
be generalized for the non-harmonic case e.g. for an odd function:
T(t) = To + (Jot + L TA,n sin(nwt) (3.10)

n~l
i. e., every periodic temperature change can be considered as a sum of sinusoidal

functions. However, there is not only one sinusoidal frequency but also higher
harmonics at frequencies which are integer multiples of the base frequency but
with (generally) decreasing amplitudes.
It is even possible to subject the sample to non-periodic temperature changes
like a step or a steep ramp. In such cases the Fourier series in Eq. (3.10) must
be replaced by a Fourier integral and as a result we get in the heat flow rate
function a continuous spectrum of frequencies instead of discrete harmonics.
The theoretical background comes from linear response theory and Fourier
transform mathematics. However, the measurement may yield more informa-
tion if we use non-sinusoidal temperature modulations, but the mathematics
and even the evaluation is more sophisticated although the background, as far as
the theory of temperature-modulated method is concerned, remains generally
the same. To simplify the formulae and focus the understanding to the essential
points, we restrict ourselves here to the simple sinusoidal case without reserva-
tion of generality. If needed the results can be generalized to other periodic or
non-periodic temperature fluctuations.
From Eq. (3.9) it follows that the heating rate is not constant as in the case of
conventional DSC but reads:
dT
- = /30 + TA W cos(wt) (3.11)
dt
The heating rate fluctuates between a maximum (/30 + TAW) and a minimum
(/30 - TAW) value and, depending on the magnitude of the three measuring
parameters /30, TAand w = 2 n f, we have to distinguish between different modes
of TMDSC operation (see Sect. 2.4.2)
- quasi-isothermal mode: /30 = 0
- heating only mode: TAW < /30
- heating-iso mode: TAW = /30
- heating cooling mode: TAW> /30
All these modes are used in practice, often without consideration about possible
consequences concerning the result of the measurements, which may be differ-
ent depending on the sample and processes in question. However, there are
different temperatures and different temperature fluctuations as well as differ-
ent heating rates during the TMDSC run, and such processes which react either
to temperature or to heating rate changes will give different signals.
As a consequence of the temperature modulation of the furnace in the DSC, the
measured heat flow rate fluctuates as well. The DSC signal is, of course, influenced
by the sample and possible processes occurring, but even by the apparatus and the
heat transfer process to the sample. Although this in reality takes place together,
we will look at the different influences separately to simplify matters.
3.3.2
Influences of the Sample
To see what the heat flow rate in a TMDSC looks like we start from the very gen-
eral heat flow rate caused by any sample in a DSC [ef. Eqs. (2.4), (2.5), (4.2)]
dT
11> (T, t) = Cp (T) . - + l1>ex. (T, t) (3.12)
dt
which tells us that the heat flow into a sample has two components, one comes
from the always non-zero heat capacity Cp and the other from additional endo-
or exothermic processes occurring at a certain temperature with a certain rate.
To make the things more transparent we distinguish in the following between
different cases. They may, in reality, often happen together in the sample, but this
doesn't matter, because of the superposition principle - which is valid for heat -
we then simply measure the sum of the different heat flow rates in the DSC.
Case 7: Heat Capacity Sample - No Processes
This is the simplest case, the sample has a certain (vibrational) heat capacity
CiT), with a generally very weak temperature dependence, considered as con-
stant during one period of the temperature fluctuation. In this case Eq. (3.12),
with ~ex. = 0, together with Eq. (3.11) yields:
~(T, t) = CpfJo + Cp ' T A W cos(wt) (3.13)
the measured heat flow rate is the sum of two components: the first one, propor-
tional to the underlying heating rate, is almost constant and the second one
fluctuates cosinusoidal (i. e., sinusoidal, but shifted in phase). These two parts are
referred to as the underlying ~u and the periodic ~ component of the heat flow
rate. As Eq. (3.13) tells us, Cp can be determined from both components and here
the question arises how to determine them from the total signal. The method is
simple, we have to determine the average within one period by integration:
t+tpl2
I
t-tp/2
~(T, t) dt = Cp Po = ~u
If we shift this integral through the measured heat flow rate signal ("gliding
integration") we get the underlying heat flow rate. This is almost that curve
we would get from the conventional DSC (without modulation: TAW = 0). Now it
becomes obvious why Cp must not change during one period, this would give
faulty values.
Subtracting the underlying part from the measured heat flow rate yields the
periodic part:
ci>(T, t) = ~(T, t) - ~u(T, t) = Cp ' TA W cos(wt) (3.14)
This is a function which fluctuates in time around zero with the same frequen-
cy as the temperature, but shifted n/2 in phase with the amplitude:
~A= Cpo T A w
From the amplitude of the periodic part, which is normally determined via
Fourier analysis or other suitable mathematical procedures, we get the heat
capacity of the sample:
~A
Cp=cpm=-- (3.15)
TAw
The specific quantity cp - obtained by dividing Cp with the mass m - is often

called "reversing heat capacity". Of course it should be the same as the specific
heat capacity determined from the underlying curve or via a conventional mea-
surement. If not, there may be a zeroline (see Sect. 5.3.1) or calibration problem
(see Chapter 4) with the particular TMDSC.
Case 2: Heat Capacity Sample with Additional Processes
Again we assume that the sample has a certain (only vibrational) heat capacity
CP(T) with a very weak temperature dependence, but additional processes with
endo- or exothermic latent heat exchange are taking place as well. In this case
Eq. (3.13) must be extended with the excess heat flow rate:
<f>(T, t) = Cp f30 + Cp TA W cos(wt) + <f>ex'(T, t) (3.16)
The latter, a continuous function of temperature and time, can be expanded as a

Taylor series around the mean temperature during the periodical fluctuations
Tu = To + f3ot,
with generally very small temperature fluctuations and a slow change of Tu in

time the excess heat flow rate can be substituted with sufficient accuracy by the
first approximation of the Taylor series:
Inserting this into Eq. (3.16) yields:

a<f>ex. (T t)
<f>(T, t) = Cp f30 + Cp TA W cos(wt) + <f>ex'(Tu, t) + u, (T - Tu)
aT
from Eq. (3.9) we see that T - Tu = TA sin(wt) and after substitution and re-
arrangement we get:
a<f>ex. (T t)
<f>(T, t) = Cp f30 + <f>ex'(T, t) + Cp TA W cos(wt) + aT u, TA . sin (wt)
(3.17)
Again the heat flow rate is the sum of a non-periodic and a periodic part. The
non-periodic part - got from the measured signal via gliding integration - now
contains contributions from the heat capacity as well as from the processes oc-
curring in the sample:
(3.18)
this is exactly the signal we would measure if we would switch the temperature
modulation off [set TA = 0 in Eq. (3.17) and compare with Eqs. (3.12) and (4.2)].
Subtracting the underlying (non-periodic) part from the total signal yields the
periodic part:
_ oq,eX.(Tu,t) .
cP(T, t) = Cp TA W cos(wt) + oT . TA sm(wt) (3.19)
This is again a harmonically fluctuating function with the same frequency as the
temperature modulation. There are two contributions, one from the heat capac-
ity of the sample (the same as in case 1) and one from the temperature depend-
ence (1 st derivative) of the process involved. Both contributions are shifted nl2
in phase, i. e., they have to be added like orthogonal vectors. [Formally the re-
sulting amplitude can be interpreted as a complex quantity with the real part
Cp TA wand the imaginary part TA . 0 cPex (Tu,t)loT]. In otherwords,Eq. (3.19)
can be rewritten in the following form:
ci>(T, t) = cPA cos(wt + 6)
This is a (co )sinusoidal heat flow rate with amplitude
(3.20)
and - compared to the pure Cp contribution - a phase shift of

ocPex.
ay-(Tu,t)
6 = arctan - - - - (3.21)
Depending on the sign of the temperature derivative of the reaction heat

o q,ex. (Tu,t)lo T, which is positive for endothermic and negative for exothermic
events, the phase shift is positive or negative, respectively, whereas the amplitude
becomes always greater than that of the pure Cp-sample, regardless of whether we
have an endo- or exothermic event. The increase depends on the magnitude of
the temperature derivative of the reaction heat flow (but not linearly). It should
be emphasized that every reaction is somehow temperature dependent and con-
tributes therefore to the measured heat flow rate amplitude. But the contribution
is often small because 0 F(Tu, t)loT is generally a small quantity compared to
the product Cp (T) . w. In addition the latter contribution can be made even more
dominant by choosing a higher frequency (lower period) of modulation.
From the amplitude - as in case 1 determined from the modulated heat flow
rate with proper evaluation methods - an "apparent" heat capacity can be cal-
culated [see Eq. (3.15)]:
Gppp. = m cppp = (Cp (T2 + (~ o~x. (Tu, t) J (3.22)

This tells us that, in addition to the common vibrational heat capacity, there is
always a contribution from processes occurring in the sample. This so-called ex-
cess heat capacity depends on the temperature derivative of the reaction heat
flow rate as well as on the frequency. It causes always an increase of the appar-
ent heat capacity regardless of whether the process is endo- or exothermic in
character. To separate the excess heat capacity from the vibrational part, the
latter must be known, or measurements at different frequencies, but with the
same underlying heating rate, should be performed.
Formally the apparent heat capacity may be understood as the absolute value
of a complex heat capacity which can be given either as real and imaginary parts
or as magnitude (absolute value) and phase angle and the following relations
apply (see textbooks of mathematics):
ct = C; + iC; = Ict I . e = Ict I . cos 6 + i . Ict I . sin 6

iD
(3.23)
C"
and tan6=L
C;
by means of Eqs. (3.20), (3.21) and (3.15) it is easy to show that in our case the
real and imaginary parts of the complex heat capacity read:
1 o~ex.
and C; = w
--::1-
uT
(Tu, t) (3.24)
With other words, the real part is the common vibrational heat capacity and the
imaginary part is marked by the change of the process heat flow rate with tem-
perature divided by frequency.
Depending on the manufacturer of the TMDSC and the software used, these
useful quantities are calculated from the modulated signal and can be further
evaluated.
Case 3: Sample with Time-Dependent Heat Capacity
In this case we are free to allow time dependent changes of the degrees of
freedom to occur in the sample, e. g., relaxation processes like vitrification or
devitrification, but we consider the temperature dependence of the heat capaci-
ty to be very weak again. However, time dependent heat capacity means a non-
equilibrium state of the system or certain subsystems of the sample. The time
scales of the relaxation processes are comparable with the time scale of the
experiment. Within the scope of linear response the total heat flow from the
sample at any moment is the superposition of the heat flows from all subsystems
a
at that moment. That is, the simple equation ~(T, t) = Cp(T) . TIO t [see Eq.
(3.12)] is not valid any more in this case and must be replaced with an integral
equation:
~(T, t) = -d f~ ( Cp(T, t - t') . -OT(t'))

- : : I dt' (3.25)
dt -~ ut
This defines the so-called convolution product of the time dependent heat
capacity and the heating rate. The convolution theorem means that the convo-
lution product in time domain transforms into a common product in the fre-
quency domain via Fourier transform ~:
~ (If>(t = ~ (Cp(T, t . ~ (T(t

this is a complex frequency dependent function, a product of two complex func-
tions in Fourier space:
If>*(w) =Gp(w) T*(w) (3.26)
In other words, a time-dependent heat capacity is equivalent to a complex fre-

quency dependent heat capacity and the heat flow rate is frequency dependent
and complex as well. All evaluation and calculation should be done in Fourier
space to simplify matters. The resulting function [e. g., the heat capacity Gp (w) 1
can be inverse Fourier transformed to get the result in time domain again. It
should be emphasized that an evaluation of the heat capacity with Eq. (3.15)
instead of Eq. (3.26) will lead to incorrect results in the time-dependent case.
The apparent heat capacity would depend on the measurement parameters like
heating rate and temperature amplitude and the true time dependence would
never be obtained. Fortunately the relaxation times of most of the degrees of
freedom contributing to the common heat capacity are much smaller than the
time window of TMDSC experiments and their contributions appear as time in-
dependent. One process, however, the glass transition, which plays an essential
role in polymer science (see Sect. 6.4), necessitates to take a time dependent heat
capacity into account and thus a careful evaluation.
The three cases discussed in this section are the most important ones from the
practical point of view. There exist other possible cases which may be described
mathematically in a similar way. In reality all these cases occur seldom as isolat-
ed events as presented here, but are combined jointly in the same sample under
investigation. We refrain from presenting the rather complicated mathematics of
such combined cases. The advanced reader may be able to set up the formulae
himself by means of the guidelines presented here. However, we recommend
to choose experimental conditions and the measurement parameters such that
only one of the above cases is dominant in the TMDSC run and determines the
measured results and the other influences can be neglected. Because of the com-
plicated matter any interpretation of the measured results should be done very
carefully and all possibilities should be thought of.
3.3.3
Influences of Heat Transport
Beside the sample and the processes occurring, there are at least two more phys-
ical effects which influence the measured heat flow signal: the transport of heat
needs time and the limited conductivity together with the heat capacity of the
DSC parts gives rise to a (damped) thermal wave with changing amplitude and
phase on moving through the DSC. As a result both the amplitude and the phase
of the measured modulated heat flow rate depend on the DSC used, and the
apparatus must be properly calibrated (see Sect. 4.7) to come to reliable values,
and on the thermal diffusivity of the sample.
The calorimeter itself is constructed of several parts which the heat has to
pass on its way from the furnace to the sample. All these parts have a certain
thermal conductivity (i.e., a certain thermal resistance) and a certain heat
capacity which influences the heat flow and the thermal wave which is propa-
gating through the DSC in the case of modulation. Any contact area between
different parts acts as an additional thermal resistance. In addition the DSC
includes often some sophisticated electronics that amplifies the voltages from
the sensors to the measured signal transferred to the computer. But every DSC,
no matter how complicated it is, can be dissected into a network of simple
mechanical elements and, hopefully, linear electronics. As mentioned in Sect. 3.1
the heat transport and the transport of electric charge are physically equivalent
processes and the DSC may be described as an electrical network as well. The
influence of such a network on AC signals is equivalent to the influence of a
calorimeter on the heat flow rate in case of TMDSC.
Formally the total network describing a DSC (here called "box") has one
"input" (the temperature-time program) and one "output" (the heat flow
rate into the sample). To evaluate the behavior of an apparatus within the
framework of linear response, it is often sufficient to look at the so-called
"transfer function" P(w) of the "box" in question. This is a complex function
in frequency domain (Hohne, Schawe, 1993; Schawe et al., 1993) defined as the
quotient of the "output function" Out(w) and the "input function" In(w). The
transfer function (in frequency domain) is mathematically connected with
the "step response" or "pulse response" functions (in time domain) via Fourier
transform. It will go beyond the scope of this book to derive all details of the
features of these functions; the interested reader is referred to textbooks of
transfer theory. However, the transfer theory has proved to be a useful tool
for the theory of TMDSC and for the description of the apparatus influence.
Actually the transfer function is closely linked to the "calibration function" need-
ed to correct the measured heat flow rate amplitude and phase regarding appa-
ratus influences.
From the transfer theory of linear systems it is known that the overall transfer
function of a network can be calculated from the transfer functions of the indi-
vidual components. In particular it holds that for transfer elements connected
in series the total transfer function is the product of those from the elements. In
other words, the magnitudes of the complex functions have to been multiplied,
whereas the phases have to be added up. On the other hand, for transfer elements
connected in parallel the total transfer function is the sum of those of the com-
ponents. These facts enable one to breakdown the rather complicated heat trans-
fer network of a DSC into simple components having rather simple transfer
functions. These transfer functions can then be assembled properly and yield
the transfer behavior of the total TMDSC (both with and without the sample in-
cluded). Knowing the transfer function enables the determination of the cor-
rection function and thus a proper calibration.
Transfer Function of a Simple RC-Element

The simplest component of heat conducting networks is an object having a cer-
tain thermal resistance Rth and a certain heat capacity Cpo The respective elec-
trical analogy element is the low pass fllter built of a resistor R and a capacitor C
in series (see Fig. 3.2). The complex transfer function, taking AC-voltage (i. e., the
respective temperatures) as input and output functions, can easily be calculated
for this case (see textbooks of physics):
(3.27)
From this the magnitude (modulus or absolute value) and the phase (argument)
of the complex function can easily be determined:
1
IP(w) I = VRe2 (p) + Im2 (p) =---;=========~
V 1 + (WRtbCp)2
(3.28)
1m
arg(P(w = tan- 1 ( - -
(P) = tan- 1 (WRthCp)
Re(P)
For a given Rtb and Cp these functions are plotted in Fig. 3.15 logarithmically
(Bode plot). In the quasi-static case - for low frequencies w - the magnitude and
phase are 1 and 0, respectively, i. e., the temperature on the output side is the
same as on the input side of the RC-element. But in the case of periodic temper-
ature changes like in Eq. (3.9) on the input side of a thermal path, the output
temperature is lower, namely by a factor of P (w).
In other words the amplitude of the modulated part decreases and the phase
shifts on the way from furnace to the sample. What do we learn from that fact?
5
Fig.3.15. Transfer function
(magnitude and phase) of rad
a simple RC-element in form p
,1
I
of a Bode plot
0.10
c
en
cs ". :::;-
......
------- <1J
U1
cs
E ;; -'=
",'"; '
CL
",'"
0.01 '""""'-=-==-:=-_-_--_-_--1-____-'-____----' 0
001 01 1.0 rad Sl 10
w--
The temperature amplitude at the sample site, compared to the set value, is too low,
resulting in a too low apparent heat capacity as well. The error is the larger the
larger are the frequency and the thermal resistance and the heat capacity
of the thermal path. In addition even the phase is shifted, in other words, the eval-
uation mathematics would result in a complex heat capacity with a non-zero imag-
inary part even in cases where the sample has no time dependent heat capacity at
all. These artifacts are caused by the propagation of the thermal wave through the
network of thermal resistors and capacities from the furnace to the sample, an
unavoidable effect for which every TMDSC must be calibrated (see Sect. 4.7).
The RC-element may even serve as a simple model for the sample itself, which
always has a certain heat capacity and of course even a thermal resistance. From
this it follows that the effective temperature amplitude inside the sample de-
pends strongly on frequency, sample size (heat capacity) and its thermal resis-
tance. For larger sample sizes this influence must be corrected as well. In cases
where the thermal resistance or the heat capacity changes during the measure-
ment, the correction function (the reciprocal transfer function) changes as well.
This is in particular the case during phase transitions (e. g., melting), where the
apparent heat capacity and the thermal resistance changes dramatically.
Thermal Waves
Another approach to describe the influence of apparatus as well as sample prop-
erties on the modulated signal starts from the (one dimensional) thermal diffu-
sion equation:
(3.29)
with Dtb the thermal diffusivity. Solving this equation with the boundary condi-
tion of an periodical temperature change on one side of an infinite rod yields:
T(x, t) = TA,o . e-~~. sin (X)

rot - ~ with ~ __ ~2~h
~ LV (3.30)
This is obviously a damped temperature wave which propagates through the rod
in time but with exponentially decreasing amplitude. From Eq. (3.30) follows:
the local amplitude at the position x is determined by the thermal diffusivity of
the heat path material as well as the frequency:
x
TA(x) = TA,o' e-~ = TA,o' e-vooso'poPoWlkthx (3.31)
(Cp the specific heat capacity, p the density, kth the thermal conductivity, TA,O the
temperature amplitude at the (input) boundary of the rod, the phase shift is the
argument of the exponential).
From the known properties of common materials used in DSCs and the nor-
mal operation frequencies, we find 5 values of about 10 to 80 mm. As the di-
mensions of the thermal pathways usually are in the region of a few millimeters,
the exponential in Eq. (3.31) can be approximated with a linear function which
results in a formula similar to that found for the simple RC-element. The situa-
tion is, however, totally different if we have materials with low thermal diffusiv-
ity like polymers. In this case the ~ drops to values of 1 to 4 mm. This means
that the temperature amplitude changes already within one millimeter, i. e., the
amplitude cannot be considered as constant or linearly decreasing in such a
sample anymore but decreases exponentially. From the practical point of view
we have to consider an apparent temperature amplitude inside the sample which
is lower than the amplitude at the input boundary. To calculate it for a sample of
thickness d in the linear case, we have to integrate Eq. (3.31):
1 d
(TA(d=d!TA,Oe~dx=-~- l-e~
_~ TA o' ~ ( _4)
(:r -2~ (:)3

(3.32)
= TA,o [1- ~ : + ~ + - ... J

This formula contains the case of a simple RC-element as the first approxima-
tion, which can easily be seen if we introduce ~ from Eq. (3.30) and (3.31), and
divide by TA,o to get the transfer function.
Network of Heat Conducting Elements

The basic elements presented so far can even be helpful to model the properties
of the total equipment, i. e., the apparatus (DSC) plus sample what concerns the
heat transport influence on the measurement. In a first approximation the heat
flows along a one-dimensional pathway through the DSC and into the sample.
This can be modeled by a series of RC-elements (or a series of "rods" from dif-
ferent material properties in the thermal wave approach). The simplest model
contains three elements: one, for instance, for the part of the pathway from the
oven to the thermometer, one from the thermometer to the sample (or reference)
support (or pan) and one from there to the inner of the sample (or reference). All
these RC-elements have a different (apparent) thermal resistance and different
(apparent) heat capacities, which, of course, are unknown for the user but almost
fixed concerning the apparatus part (but they differ for the sample part).
From transfer theory it is known that for transfer elements connected in se-
ries the total transfer function is the product of the transfer functions from the
elements. In other words, the magnitudes (amplitudes) have to be multiplied,
whereas the phases connect additively. This way the total transfer function can
be deduced from Eq. (3.28) with different 1) = Rth,j CP,j for the different RC-ele-
ments in question:
1
IP(w)l =IT and arg(P(w = ~ tan-l (WRth,jCp)
j VI + (WRth,jCpY J
The more RC-elements are included, the more precise is the approximation of
the real transfer function of the thermal pathway of the DSC, but the more (ini-
Fig.3.16. Transfer function 5

(Bode plot) of a simple DSC rad
model (solid lines) together
I
p 4
I
with that of a more detailed
model (dotted lines)
QJ
3
"0
:::J 0.10
0:: 2 :.c
Vl
""
'"
E
OJ
Vl
'"c.
.<::
0.01 0
0.01 0.1 10 rad 5- 1 10
w----
tially unknown) parameters are included in the respective correction formula

and have to be determined for proper correction.
To show one result, the transfer behavior of the simplest possible complete
DSC network is as follows: one RC-element on the reference side (collecting the
heat flow path and the pan together to one apparent Rth and Cp) and two RC-el-
ements in series on the sample side (the apparatus part as on the reference side,
and the sample separated from the latter). The transfer function for this model
(which corresponds to a DSC model in between the 1st and 2nd approximation of
Sect. 3.1) is again defined as the quotient of the output function (i.e., the ATsig-
nal of the DSC) over the input function (i.e., the temperature of the heater). It
can be calculated straight forward like the single RC-element, with the network
rules in mind (for details see Hohne, 1999a; Hohne et al., 2002). The respective
transfer function is plotted in Fig. 3.16. If we compare this with that from the sim-
ple RC-element (Fig 3.15) we see a qualitatively similar behavior: the amplitude
drops with increasing frequency and there is an increasing phase shift. However,
there are quantitative differences, the frequency dependence is stronger, i. e., the
influence is larger for this network than for a single RC-element. Of course even
this result is only an approximation, but it fits already rather well to the real DSC.
The network can be refined including more RC-elements, e. g., the crucibles and
the heat transfer resistance to them, this results in a modified transfer function
(see Fig. 3.16 dotted lines) which give an even better approximation to the real-
ity, but is not very different from the simpler model.
Several correction formulae have been derived this way in the literature
which describe the DSC influence more or less satisfactory. We refrain from pre-
senting the formulae here, they impart no more information to the practitioners
than the above derived simple models. Instead we recommend to determine the
transfer function experimentally and correct the measured quantities as shown
in Sect.4.7.
3.3.4
Conclusions
The TMDSC method starts from a temperature program which contains a non-
periodic and a periodic part. This causes a heat flow rate signal with two com-
ponents too, a non-periodic underlying part and a periodic part. The former can
be separated by gliding averaging. Subtraction of the underlying from the total
signal yields the periodic part which can be treated in different ways to get the
information it contains. One possibility is to evaluate the amplitude, this results
in the "reversing" curve and after subtraction from the underlying curve in the
"non-reversing" curve. Another approach evaluates amplitude and phase angle
and results in a complex heat capacity with magnitude and phase or real and
imaginary part. Different processes contribute to the different signals in a differ-
ent way. This offers the possibility to separate the processes from one another by
choosing the right measurement conditions and the proper evaluation method.
In every case the measured quantities as well as the properties of the sample (e. g.
Cp ) must not change during one period.
But the fundamentals presented in this section show even that the effective
temperature amplitude inside the sample is never that of the heater which we
program in our TMDSC and use for the calculation. It is always lower and we
have to correct it to get proper results from the measurements. The correction
depends on the frequency and the thermal diffusivity (i. e., the heat capacity and
the thermal resistance) of the materials of the heat conducting path in the
calorimeter but unfortunately even on the properties of the sample itself. This
necessitates careful calibration of the TMDSC as far as the influence of the
thermal wave is concerned. To reduce the influence of the sample itself and its
(changing) properties - which normally are unknown - we have to reduce the
thickness of the sample to the lowest possible size, or we have to determine this
influence by means of a special procedure (see Sect. 4.7).
In summary, it may be said that temperature-modulated DSC is a very power-
ful method to measure material properties and to get more insight into sample
behavior, in particular as far as time dependent processes are concerned, but the
theory is not simple and one has to be careful in choosing the right measure-
ment parameters and the proper evaluation. But it should be emphasized that
linearity and stationarity of the whole system are essential conditions for the
temperature-modulated method to reveal meaningful results.
4 Calibration of Differential Scanning Calorimeters
"Calibration means the set of operations that establish, under specified condi-
tions, the relationship between values of a quantity indicated by a measuring
instrument or measuring system ... and the corresponding values realised by
standards" (from International Vocabulary of Basic and General Terms in
Metrology, 1994). "Standards" stands for Certified Reference Materials, or for
otherwise calibrated measuring instruments like temperature sensors, standard
resistors and the like, which are connected to international fixed quantities.
Calibration needs we1l4efined and tested procedures suitable to be done by well
trained users in the laboratory.
A recommendation for DSC calibration, given by the International Confeder-
ation for Thermal Analysis and Calorimetry (ICTAC), exists which describes
procedures for temperature as well as caloric calibration. This recommendation
is founded on a metrological basis.
However, a metrologically flawless, basic calibration of a DSC should be done
very carefully and is very time-consuming. For many routine applications of
DSCs, e.g., for (relative) comparisons of material properties, other written
"standards" exist which recommend to use selected parameters for the mea-
surements, e. g., certain sample masses, specific heating rates etc. The standards
(the written ones and the recommended reference materials) follow a "consen-
sus philosophy" (Archer, Kirklin, 2000) which is based on round robin test pro-
cedures and average values.
Before using one of the "standard" procedures common in practice, it's
a good idea to understand the principles of a metrologically correct calibra-
tion. This would enable the user to adapt (simplify) the correct calibration
procedure to their specific purposes instead of using the "standards" uncriti-
cally without considering their limitations. A collection of papers on the cali-
bration of calorimeters (DSC and others) has been published by Della Gatta et
al.,2000.
To what extent DSCs are amenable to reliable calibration depends on the
quality of the DSCs, the measuring system, the measurement parameters
and on the availability of calibration substances with precisely measured
properties. In this point, the situation is still quite unsatisfactory. Moreover,
no complete theory of the DSCs exists, which takes full account of the influ-
ences of the instrument on the measurement and inherent systematic error
sources. It should be borne in mind that it is the overall uncertainty of the cal-
ibration which yields the smallest possible systematic uncertainty of the meas-
urements.
66 4 Calibration of Differential Scanning Calorimeters
4.1
Aspects of Quality Assurance
Thermoanalytical instruments, including DSCs, are extensively used by indus-

try in quality assurance systems to warrant trustworthy, reproducible properties
of the products, even products in the sense of measurement results by which
products are characterized. The ISO 9000 series and the ISO/lEC 17025,1999
are the basic standards in the field of quality management and quality assur-
ance. The standard ISO/lEC 17025, in particular, specifies general requirements
to be met by test laboratories. It will no longer be sufficient to have technical and
scientific competence, this competence must be documented and made appar-
ent to others. DSC methods are used for quality assurance in many ways: for ex-
ample, for the inspection of incoming materials, as an accompanying measure in
manufacture and for the control of the final 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
application of sensitive calibration procedures and the use of calibration sub-
stances which are - if possible - traceable 1 to national/international standards
and - in the end - to the SI-units. This means: to create confidence in the quali-
ty of the measured values and to avoid repeat measurements, metrologically
flawless calibration procedures and traceable calibration substances (Certified
Reference Materials) must be developed on the basis of which reliable uncer-
tainty values can finally be assigned to the measurement results.
In the following sections calibration procedures and materials will be pre-
sented which are mainly based on investigations carried out by the German
Society for Thermal Analysis (GEFTA), which are the basis of the ICTAC recom-
mendations mentioned above (temperature calibration: Hahne et aI., 1990;
Cammenga et aI.1993; caloric calibration: Sarge et aI., 1994; calibration general:
Gmelin, Sarge, 1995; Sarge et aI., 1997; Gmelin, Sarge, 2000).
4.2
Basic Aspects of Calibration
In the following, reference will be made to some important thermodynamic as-

pects 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 f...H of the
phase transition (with T = constant); it is the difference between two variables
of state and, consequently, well defined. In a calorimeter, it is the heat of transi-
1 Traceability: Property of the result of a measurement or the value of a standard whereby

it can be related to stated references, usually national or international standards, through
an unbroken chain of comparisons all having stated uncertainties (from International
Vocabulary of Basic and General Terms in Metrology, 1994).
4.2 Basic Aspects of Calibration 67
tion Q which is measured. The relation between heat and enthalpy is given by the
1st law of thermodynamics and the definition of the enthalpy function:
dU=dQ +dw+ IdE; (1 st law of thermodynamics)

;
dH=dU+pdV+ Vdp (fromH == U+ pV)
From this it follows (with dW= -pdV):
dQ = dH - V dp - I dE;
;
(U internal energy, W work, E; different energy forms).
The enthalpy H of a system is a function of the variables of state, namely pres-

sure p, temperature T and composition ~. The total differential then reads:
dH=(aH) dP+(aH) dT+(aaH ) d~

ap T,s aT P,s ~ T,p
Calorimetrically, the measured quantity, the heat in differential form is:
dQrn = [( ~H)
P
-vJ r.s dp + (~;) p,S
dT + (~H)
~ T,p
d~ -7 dE; (4.1)
Under the conditions normally prevailing during DSC measurements, the first
term can almost always be neglected, even if closed crucibles are used (maxi-
mum pressure changes: 1 to 2 bar, i.e., dp :::: 0). The second term is the heat ca-
pacityat constant pressure and constant composition of the (reacting) system:
The third term in Eq. (4.1) is the isothermal and isobaric enthalpy change due,
for example, to a phase transition (AH = AtrsH), a mixing effect (AH = ArnixH) or
a reaction (AH = ArH).
Only if p :::: const. and I dE; = 0 and if the Cp course is known (namely the cor-
;
rect baseline during a thermal event, d. Sect. 5.3) the following equivalence can
be assumed for the melting of a pure substance [obtained by integration of
Eq. (4.1)]:
Afus H = Qfus
It should be emphasized that 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 bigger bead of
the melting substance. As a result, surface heat is exchanged (dEsurface =j:. 0) and
Eq. (4.1) becomes
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 Qfus and the enthalpy difference fiR is small, these contributions must
not be basically ignored. It is furthermore pointed out that common thermody-
namics starts from infinite phases and the fiR definition is based on this fact. For
smaller phases (small crystals), fiR depends on the size of the sample grains be-
cause the surface energy of the crystallites must not be ignored. In other words,
the specific heat of melting of a fine powder differs from that of a large crystal
and, as a result, the temperature of melting changes as well.
For the heat flow rate under ideal conditions (p =: const., IdE i =: 0) the fol-
lowing is obtained by differentiation of Eq. (4.1)
6Qrn
- = cPrn= C ~(T)-+ -
dT (OR) d5
- (4.2)
dt p. dt 05 p.T dt
The heat flow rate has two parts, the first is caused by the heat capacity of the
sample and the second is the reaction heat flow rate. The former is the baseline,
which must be subtracted from the measured heat flow rate to get the latter
which forms the basis for the kinetic evaluation of DSC curves (cf. Sect. 6.3).
DSC measurements are normally performed both in the heating and cooling
mode. Generally the temperature distribution in the DSC measuring system is
asymmetrical with respect to heating and cooling. The reason for this is that the
heat transfer is not strictly linearly dependent on the temperature difference.
Therefore heat flow rates are basically different during heating and cooling of
the same sample with the same rate. This results in different calibration factors
for heat and heat flow rate. In practice, the effect is very small, i. e., normally suf-
ficient symmetry with respect to heating and cooling can be assumed for heat
and heat flow rate calibration factors. In other words, the uncertainty of heat and
heat flow rates is normally larger than the asymmetry effect. This must, however,
be verified and the procedures are part of a careful calibration.
It is even obvious that the temperature of the sample deviates from the
set value differently on heating and cooling because of the thermal lag. This
deviation can even be asymmetric if we heat or cool the sample at a certain
rate because of non-linearity effects. This necessitates a separate calibration
in cooling mode as well, which creates another problem, because the conven-
tionally used calibration substances show supercooling effects for the relevant
phase transition. Consequently, temperature calibration in heating and in
cooling mode has to follow different procedures with different substances (see
Sect. 4.3).
After the calibration experiments have been completed, the electronics pro-
vided for this purpose will either be adjusted until the values indicated corre-
4.3 Temperature Calibration 69
sponds to the true ones, or adaptation will be ensured via the control software
(in a way usually not apparent to the user).
Any calibration already carried out by the manufacturer must be checked
(verification of calibration) for a newly purchased equipment and again after
appropriate periods. Regular calibrations provide important information about
the repeatability error and any long-term systematic variations (drift) of the
measured quantities. (It should be noted that repeatability error is not the same
as uncertainty of measurement, cf. Sects. 7.2 and 7.3).
4.3
Temperature Calibration
Temperature calibration means the unambiguous assignment of the tempera-

ture "indicated" by the instrument to the "true" temperature.
The "true" temperature is defined by fixed points with the aid of calibration
substances. It is reasonable to choose as calibration substances, if possible, the
substances used to realize the fixed points of the International Temperature
Scale of 1990 (lTS-90, internationally legal by law). 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 approxi-
mately. As the point of temperature measurement is not the point where the
sample is located, a systematic error will always occur in scanning operation
which depends on instrument and experimental parameters. To establish a
graph or table, showing the relation between the indicated and the true temper-
atures in dependence on experimental parameters, is normally very helpful in
practice and for uncertainty estimation. In every case, a table or graph should be
drawn up which shows the variation of the indicated temperature at different
heating rates. The calibration procedures described in the following take these
special features into account in a general way, independent of the respective DSC
type (see Hahne et aI., 1990).
In the case of an endothermic event, the DSC (in heating mode) records the
heat flow rate signal schematically shown in Fig. 4.1. The section between the
initial peak temperature Ti and the final peak temperature Tf is defined as the
peak (see Sect. 5.1). The baseline needed to get the different temperatures can be
interpolated by various methods between Ti and Tf (cf. Sect. 5.3.2). The intersec-
tion between the auxiliary line and the baseline suffices to fix the characteristic
temperature Te (extrapolated peak onset temperature) which is important here.
4.3.1
Temperature Calibration on Heating
- Selection of at least 3 calibration substances (cf. Sect. 4.6.1) which cover the
desired temperature range as uniformly as possible (it should be at least 3 in
order to detect possible non-linear temperature dependence).
Fig.4.1. Heat flow rate signal

of a DSC during a transition
in heating mode (schematic
representation, according to
Hemminger, 1994).
CD baseline (interpolated),
@ auxiliary lines, Ti initial
peak temperature, Te extrap-
olated peak onset tempera-
r
o
ture, Tp peak maximum c:
Ol
temperature, Tc extrapolated V)
I
peak offset temperature, I \
Tf final peak temperature r---~Y~---~-----~-~-~------
1\ 'I I I\ ,I
I I I
I
I
I
I
I
I
CD
I I
T; Tf
tempera ture ...
- At least two calibration samples of each substance are prepared (for repeat
measurements). The sample mass should correspond to that commonly used
in routine measurements.
- The transition is to be measured with each calibration sample at a minimum
of 3 different heating rates in the range of interest, including the smallest pos-
sible one. The second calibration sample of the same substance is also mea-
sured at the different heating rates.
- It is to be checked whether there is a significant difference between the char-
acteristic temperatures (Te) obtained at identical heating rates for the first
and second calibration sample of the same substance. If necessary, it is to be
checked whether the temperatures depend on other parameters (mass, loca-
tion of the sample in the crucible etc.).
- If this is not the case, Te is represented as a function of the heating rate and
the extrapolated value Te(f3 ~ 0) determined for zero heating rate (Fig. 4.2).
- The difference Ll Teorr (13 = 0) between the value Te (13 ~ 0) obtained in this way
and the respective fixed-point value TflX or the value Tlit taken from the liter-
ature (cf., e.g., Marsh, 1987; Cammenga et aI., 1993; Gmelin, Sarge, 1995; Sarge
et aI., 1997) is either used to change the instrument calibration according to
the manufacturer's instructions or it enters into a calibration table or curve
(cf. Fig. 4.3).
- If Te 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.).
Temperature calibration has then been completed.
(3 ~
Fig. 4.2. The extrapolated peak onset temperature Te as a function of the heating rate f3, and
construction of Te(f3 ~ 0).
ATe/6.f3 variation of Te (f3) with f3; O),@ different calibration substances
t
Fig.4.3. Temperature correction ATcorr (f3 = 0) as a function of the extrapolated peak onset
temperature Te(f3 ~ 0) for three calibration substances (O),@,).
ATearr (f3= 0) is the difference between Te(f3 ~ 0) (cf. Fig. 4.2) and the "true" value of the tem-
perature of transition. The curve ATcorr (Te) obtained by means of (at least) three calibration
substances shows the corrections to be applied to the measured values Te(f3 ~ 0) at different
temperatures
How are the correct temperatures assigned in practice?

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 determined by extrapolation to 13 =0 (static case, quasi-equilibri-
um temperature). The correction ATeorr (13= 0) is applied using the proper
calibration table or curve:
Ttrue = Te(f3 -7 0) + ATeorr (13 = 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"
processes), the value Te{f3~O) may be calculated from the mean slope
ATe/Af3 of the Te{f3) curve{s) of the calibration substance{s) (cf. Fig. 4.2 and
example below):
ATeorr (f3 =0) is again taken from the calibration table or curve (Fig. 4.3), so that
Ttrue = Te (f3 ~ 0) + ATeorr (f3 = 0) . The best thing to do is to start by carrying out
two measurements at clearly differing heating rates. Then it is checked whether
ATe/Af3 corresponds with the (mean) slope of the Te{f3) curve{s) of the calibra-
tion substances. If so, extrapolation to Te{f3 ~ 0) can be carried out at once.
If not, first the heating rate dependence Te (f3) must generally be determined as
described above.
To simplify the described method, for each heating rate applied, the corre-
sponding overall correction ATeorr (f3) = ATeorr (f3 = 0) - f3 . ATel Af3 can be listed
(cf. example below) so that the "true" temperature can be determined directly:
Here, a distinction by classes of substances (metals, organic substances) must

possibly be made because of their very different thermal behavior.
It is to be expected that, for a certain instrument, ATeorr {f3 = 0) will vary with
time which is why a recheck should be made after some time (every week to
every other month, depending on the DSC in question) by verification or recal-
ibration. The dependence of the extrapolated peak onset temperature ATel Af3 on
the heating rate is only related to the heat transport properties of the DSC and
the sample substance; it remains unchanged for a certain DSC and substance
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 non-zero heating rate, an auxiliary line must
be drawn at the angle a to the extrapolated baseline. The angle a can be taken
from the calibration experiment in which the same heating rate has been applied
(Fig. 4.4 a, b). The scale division of the axes must be the same in both cases;
otherwise the slope of the auxiliary line must be converted properly. (For more
advanced methods of determining characteristic temperatures in the case of
endothermic and exothermic events see Sarge, 1991; Schawe, 1993).
Uncertainties
Depending on the type of DSC, the minimum repeatability error of the deter-
mination of Te 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 be-
tween 0.1 and 0.8 K. An overall uncertainty of Te measurement between 0.3 and
1.0 K must be reckoned with. The overall uncertainty of the temperature cali-
bration should in every case be carefully estimated (including uncertainty of
temperature sensors, uncertainty of the determination of Te etc.).
tempera ture
CI
c::
en
Vi ......_.-:;
b
T,1 TP2
temp ern ture

Fig. 4.4 a, b. Assignment of the characteristic temperature Tp in complex thermal events.
a Calibration measurement to determine the angie 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 operation, 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 mea-
sured 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 temper-
ature (see Tp3)
Notes:
1. It has to be checked whether Te depends on the location of the calibration
sample in the sample container (in particular at high temperatures). In par-
ticular, for power compensation DSCs, due to the isoperibol operation, a tem-
perature profile develops in the sample support (in bottom plate of the micro-
furnace, Fig. 4.5) 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 Te indicated (see Hahne, GlOggler, 1989).
Such effects are also possible in heat flux DSCs (cf. Sect. 3.1), 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 cylindrical 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 depends strongly on
the heat produced and on sample properties and cannot be precisely deter-
mined. A reliable assignment of a temperature is therefore only possible 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:
Fig.4.5. Temperature profile in the microfurnace of a power compensation DSC (schematic).

The isoperibol mode of operation (constant ambient temperature) results in a temperature
profile in the sample support as shown schematically. During melting of a sample, this tem-
perature profile leads to time-dependent zones of melting (Tfw is first reached in the centre)
and consequently to different extrapolated peak onset temperatures Te when the radial posi-
tion (r) of small samples differs (according to Hohne, GlOggler, 1989)
Fig. 4.6 a, b. Heat resistance

Rth between temperature
measurement point and
sample.
a Diagrammatic view of
the sample system.
b Curve measured during
melting of a pure substance,
q, heat flow rate into the
sample, tany ascending slope
(dq,/dt) of the q,(t) curve
during melting (to calculate
Rth)
time ..
If there is a thermal resistance Rth between sample (Ts) and sensor tempera-
ture (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 homoge-
neous. When a pure substance melts (as for calibration, for example), Rth can be
determined approximately from the slope of the respective heat flow rate curve
4>r(t). From Eq. (4.3) the following results after differentiation (with Ts = con-
stant during melting, dTM/dt:::: f3 and tan y= (d4>r/dt)melting):
f3
Rth ". - - (4.4)
tan y
(Experiments at various heating rates shows the validity of this relation.)
Arguments in Support of the Temperature Calibration Procedure

The existing recommendations for the temperature calibration of DSCs (e. g.,
ASTM E 967-97: Standard Practice for Temperature Calibration of Differential
Scanning Calorimeters and Differential Thermal Analyzers, 1997) and the spec-
m
L
1
Te
326,-_,,--_,,--_,,--_-,--_-,--_-,--_.J......._
o 10 20 30 K/min 40
{3 ~
Fig.4.7. Extrapolated peak onset temperature Te in DC as a function of the heating rate f3 (two
measurement series of a heat flux DSC with 58 mg of lead)
Q-{) measured values, .... curve of average values. The curves show a non-linear dependence Te (f3)
only towards the highest heating rate and a dispersion of the Te-values, which depends on f3
ifications of most manufacturers take a standard heating rate (e.g., 10 K min-I)

as a basis. This method cannot be recommended since
- different heating rates require different corrections; shifting of Te is not
always linear due to the heating rate (cf. Fig. 4.7);
- 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 3 different heating rates (starting with the
lowest heating rate),
- the temperature fixed-points of the International Temperature Scale are de-
fined for the reference substances in phase equilibrium (i.e., static case, zero
heating rate).
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:
- Ti cannot be determined with the required reliability because of the noise; the
same applies to Tf (the repeatability error is between 2 and 15K depending on
substance, kind of transition etc.)
- Tp and Tc strongly depend on the thermal conductivity, on mass and layer
thickness (volume) of the sample substance, on the heating rate and on the
heat transfer from sample to sample container (crucible), which may change
on melting (cf. Fig. 4.8).
- Te depends least on heating rate and sample parameters (substance, thermal
conductivity, mass, layer thickness); any possible effect of melting (heat
transfer) should be checked by carrying out various similar experiments with
the same calibration sample.
2
cJ>m
4
+
6
300 320 340 360 380 C 400
tempera ture ~
Fig.4.8. Measured curves showing the peak temperature maximum Tp changing with the
heating rate p (heat flux DSC, lead, 58 mg, heating rate pfrom 5 to 50 K min-I).
4'm heat flow rate (arbitrary units). In addition to the shifting of Tp with p, the great changes
of Tc and Trare obvious (for definitions see Fig. 4.1)
Examples of Temperature Calibration on Heating

The results of the temperature calibration of a power compensation DSC in
heating mode are given as an example. Two samples each (between 2 and 6 mg)
of highly pure indium, tin and lead were used (in hermetically closed aluminum
crucibles). The results shown in Tables 4.1 to 4.3 were obtained for the non-cal-
ibrated device (cf. Figs. 4.9 a to c):
The correction curve shown in Fig. 4.10 has been obtained for the calibration
of the device.
In practice, for the purpose of temperature calibration, the potentiometer ad-
justment would in this case be changed by the mean value of 0.83 K. If necessary,
the continuation of the correction curve towards higher temperatures may be
determined on the basis of another calibration (e.g., with Zn). Different correc-
tions must possibly be applied in different temperature ranges. Nowadays with
modern DSCs, the user normally enters the results of the calibration measure-
ments (for example, with In and Pb) into the computer which internally calcu-
lates the correction curve and converts the temperature scale. Unfortunately this
is in most cases only done in linear approximation (slope and position, 2-point
calibration) which is not sufficient for precise measurements.
After the DSC has been calibrated with the aid of !lTeorr (f3= 0), various pure
substances are measured at different heating rates and Te is determined respec-
tively. Table 4.4 is obtained and includes !lTeorr(f3= 0) = Tlit - Te(f3 ~ 0), the cor-
rections for the various heating rates calculated from !l Teorr (f3 = 0) and the slope
of the respective linear fit curve:
!lTe
!l Teorr (f3) =!l Teorr (f3 =0) - f3-
!lf3
(The values for !lTeorr (f3) have in general been rounded to two decimals.)
Table 4.1 Temperature calibration with indium (cf. Fig. 4.9a)
Sample No. Run No. Heating rate in K min- 1 Te in C
10.0 156.21
1 2 10.0 156.16
2 10.0 156.25
2 2 10.0 156.21
5.0 155.81
2 5.0 155.91
2.5 155.65
2 2.5 155.67
1 1.0 155.64
2 1.0 155.65
0.5 155.52
2 0.5 155.65
2 2 0.5 155.53
0.1 155.49
2 0.1 155.50
Results:
Te (f3 ~ 0) = 155.523 C
TflX = 156.5985 C (ITS-90)
f:.Tcorr (f3 = 0) = + 1.0755 K
f:.Te/f:.f3 = 0.0682 K/(Kmin- 1)
Table 4.2 Temperature calibration with tin (cf. Fig. 4.9 b)
Sample No. Run No. Heating rate in K min- 1 Te in C
10.0 232.35
1 2 10.0 232.18
2 1 10.0 232.06
5.0 231.81
2 5.0 231.69
1 2.5 231.54
2 2.5 231.41
1 1.0 231.33
2 1.0 231.22
0.5 231.28
2 0.5 231.12
0.1 231.26
2 0.1 231.12
Results:
Te (f3 ~ 0) = 231.184 C
TflX = 231.928 C (ITS-90)
f:.Tcorr (f3 = 0) = + 0.744 K
f:.Te/f:.f3 = 0.103 K/(Kmin- 1)
Fig.4.9a-c. Extrapolated 156.5

peak onset temperature Te 7;lp.O) ISS.5n t
i
(in 0c) as a function of the 7i.
heating rate f3 to determine
t
M .., IP= 0)= """4t~
.1.111 K
Te(f3 ~ 0). (Power compen-
156.0
sation DSC, two measure- 0
ment series each with two T.
~
samples of different mass, In
closed aluminum crucibles).
a Sample material: indium, 155.5 To: 155.523 + 0.0682 . P
b Sample material: tin, a
c Sample material: lead. 5 K/min 10
Te(f3 ~ 0) extrapolated peak (3-
onset temperature at zero
heating rate, TflX fixed-point
temperature, i. e., true tem- 232.5
T.lp.O) 231.1841"( 0
perature of melting (ITS-90, 0
in the case of lead converted '( T'i, 231.9681 '(
to the ITS-90), 6.Tcorr = LlTco"IP= 0)= .0.784 K
i
TflX - Te(f3 ~ 0) temperature m.o
correction
T.
231.5 Sn
J
0.9--'
T.: 231.1841 0.103 .p
-
0
b 231.0
5 K/min 10
(3
328.0
t 0
T,lp.O) : 326.793 '(
328.0 7ij, : 321.502 '( 0
LlT,o"IP= 0)= .0.709 K
!
0
321.5 9
T.
0
Pb
T.= 326.793 0.0997 . P
c 326.5
K/min 10
Table 4.3 Temperature calibration with lead (cf. Fig. 4.9c)
Sample No. Run No. Heating rate in K min~l Te in C
1 10.0 328.22
2 10.0 327.50
3 10.0 327.52
2 1 10.0 327.95
2 2 10.0 327.87
5.0 327.10
1 2 5.0 327.09
2 5.0 327.42
2.5 326.94
2 2.5 327.23
1 1.0 326.76
2 1.0 326.97
1 0.5 326.78
2 0.5 326.95
1 0.1 326.75
2 0.1 326.93
Results:
Te (fJ ~O) = 326.793C
Tnx = 327.462C (converted from the IPTS~68 to the ITS~90)
!:lTeorr (fJ = 0) = + 0.669 K
!:lTe/!:lfJ = 0.0997 K/(Kmin~l)
1.4
K In
1.0
t
f':.. Tcorr(P=O)
0.6 Sn
Pb
0.2
C
..
150 200 250 300 350
Te (P """0)
Fig.4.1O. Correction curve for the temperature calibration of a power compensation DSC.
The individual corrections !:lTeorr(fJ = 0) have been taken from the results of the calibration
measurements shown in Figs. 4.9 a to c.
!:lTeorr(fJ = 0) difference between true temperature (fixed~point temperature) and extrapolated
peak onset temperature Te(f3 ~ 0) for zero heating rate (for Pb see legend of Fig. 4.9),
---- mean value: +0.83 K
Table 4.4 Measurements with various substances to determine ~ Te/~{3
Diphenyl ether (phenoxybenzene)
Te in Heating rate in ~Ttot({3) in

C Kmin- 1 K
27.05 0.5 -0.19 Tlit = 26.86C

27.10 l.0 -0.24 Te({3 --+0) = 26.93C
27.20 2.5 -0.34 ~ Tcorr({3 = 0) = - 0.07 K
27.73 5.0 -0.87
Gallium
Te in Heating rate in ~ Ttot ((3) in

C Kmin- 1 K
29.92 0.5 -0.16 Tlit = 29.7646C = Tfix

29.97 l.0 -0.21 Te({3 --+ 0) = 29.851 C
30.11 2.5 -0.35 ~Tcorr({3 = 0) = -0.09 K
30.35 5.0 -0.59
30.94 lO.O -1.18
C33H68 (paraffin) 1st transition
Te in Heating rate in ~ Ttot ((3) in

C K min- 1 K
67.35 0.5 0.05 Tlit = 67.4C

67.36 l.0 0.04 Te({3 --+0) = 67.3C
67.47 2.5 -0.07 ~Tcorr({3 = 0) = +0.1 K
67.66 5.0 -0.26
68.06 10.0 -0.66
C33H68 (paraffin) 2nd transition
Te in Heating rate in
C Kmin- 1
70.98 0.5 0.12 Tlit = 7l.1 C

70.98 l.0 0.12 Te ({3 --+ 0) = 70.91 C
71.10 2.5 o ~Tcorr({3 =0) = +0.2K
7l.27 5.0 -0.17
7l.70 10.0 -0.60
Benzoic acid
Te in Heatin~ rate in
C Kmin-
122.28 0.5 0.06 Tlit = 122.34C

122.23 l.0 0.11 Te ({3 --+ 0) = 122.22C
122.34 2.5 o ~Tcorr({3 =0) = +0.12 K
122.46 5.0 -0.12
122.72 lO.O -0.38
Table 4.4 (continued)
Indium
Te in Heating rate in tJ. Ttot (f3) in

C Kmin- 1 K
156.63 0.5 - 0.03 Tlit = 156.5985 DC = Tfix

156.67 1.0 - 0.07 Te({3 ~O) = 156.53 DC
156.80 2.5 - 0.20 tJ. Teorr ({3 = 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
Te in Heating rate in tJ.Ttot ({3) in

DC Kmin- 1 K
232.18 0.5 -0.25 Tlit = 231.928 DC = Tfll{

232.21 1.0 -0.28 Te({3 ~ 0) = 232.l2 DC
232.34 2.5 -0.41 tJ. Teorr({3 = 0) = - 0.19 K
232.52 5.0 -0.59
233.Ql 10.0 -1.08
Caffeine
Te in Heating rate in tJ.Ttot ({3) in

DC Kmin- 1 K
236.30 0.5 -0.20 Tlit = 236.1 DC

236.30 1.0 -0.20 Te ({3 ~ 0) = 236.24 DC
236.36 2.5 -0.26 tJ.Teorr({3 = 0) = -0.1 K
236.67 5.0 -0.57
236.96 10.0 -0.86
Lead
Te in Heating rate in tJ. Ttot ({3) in

DC Kmin- 1 K
237.99 0.5 -0.53 Tlit = 327.462 DC

238.00 1.0 -0.54 Te({3 ~ 0) = 327.93 DC
238.17 2.5 -0.71 tJ.Teorr({3 = 0) = -0.47 K
238.32 5.0 -0.86
238.81 10.0 -1.35
Table 4.5 The dependence of Te on the heating rate f3 for various classes of materials (re-
sults from Table 4.4)
Substance
Gallium 0.11
Indium 0.07
Tin 0.09
Lead 0.09
Mean value metals 0.090.02 a
Diphenyl ether 0.15

Paraffin C33H68 0.08
Benzoic acid 0.05
Caffeine 0.07
Mean value organic substances 0.09 0.04 a
Overall mean value 0.090.03 a
a Standard deviation On-I.
ATtot = Tlit - Te (P) is defined as the difference between the "true" temperature
Tlit and the extrapolated peak onset temperature at a heating rate p.
Obviously, for this calorimeter there is a linear relation between the shift of Te
and the heating rate p. The gradient ATe/AP of the respective fit-line yields the
values given in Table 4.5.
From this it follows that for this DSC, at a heating rate of 10 K min-I, a mean
value of 0.9 0.3 K results for the difference between Te(P) and the true tem-
perature of the transition. But, depending on the heat transfer between sub-
stance and crucible bottom, the difference may vary between 0.5 and 1.5 K. This
means an uncertainty of 0.5 K for the temperatures at this heating rate. Sepa-
rate measurements at different heating rates must be carried out for a certain
substance in order to determine the slope and thus the transition temperatures
more accurately.
Note:
The values obtained for Te in the first measurement of a calibration sample
(In, Sn, Pb) are systematically higher than those of the second and all subsequent
measurements (at a given heating rate). The reason for this is the heat transfer
between the sample and the bottom of the crucible, which has strongly improved
after the first melting (larger and better contact). As a result, Te changes. If, for
reasons of irreversibility of the process to be investigated, only the first meas-
urement of a sample can be evaluated, this effect must also be taken into account
when the uncertainties of Te (P ~ 0) are estimated. Another striking feature of
the results of the calibration measurements is that the Te of individual 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. 4.9a to c). The
328.0
C
327.8
327.6
1
Te 327.4
o 2 4 6 8 K/mln 10
{3 ~
Fig.4.11. Extrapolated peak onset temperatures T. of lead (0.2 mg) as a function of heating
rate p with the samples in various positions (1 to 4) in the microfurnace of the power com-
pensation DSC (cf. Fig. 4.5, according to Hohne, GlOggler, 1989)
reason is that, firstly, the sample containers and thus the heat transfer path 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 Te of
the measured curve. This is obvious from the results shown in Fig. 4.11 (cf. also
Hahne, Glaggler, 1989).
4.3.2
Temperature Calibrati.on on Cooling
In measurements at negative heating rates (cooling), the sample temperature is

higher than the indicated temperature. As a result, the correction !1 T (f3) from
the calibration table or curve must be applied with the sign reversed as com-
pared with heating (see Fig. 4.12).
The procedures to be applied comprise a check whether the calibration in
heating mode, performed before, is also valid in the cooling mode (symmetry
check). If an asymmetry is found, a separate calibration for this mode has to be
done; suitable procedures and substances are given here which have been adopt-
ed from a paper of Sarge et al., 2000.
Generally, substances with two types of phase transitions are recommended.
These are substances with first order phase transitions (liquid/solid) with only
small as well as reproducible supercooling, and substances with phase transi-
D. T(P)
o
PCII' Pbf.at
. - - - (Doling rate heating rate - -
Fig.4.12. Schematic representation of the temperature corrections in the heating and cooling
mode
tions of higher order (e.g., between different liquid crystal states) with no or
negligible supercooling.
Symmetry Check
- From the list of substances recommended (see Table 4.12), one substance
is selected, whose phase transition lies within the temperature range of in-
terest.
- One sample is weighed in, the mass should be in the range normally used.
- With this sample the transition is to be measured at three different heating
and cooling rates, respectively. At least two runs each should be done.
- For each peak obtained the extrapolated peak onset temperature Te or the
peak maximum temperature Tp (in accordance with Table 4.12) is deter-
mined.
- It is to be checked whether the results from the two experiments differ signif-
icantly.
- If this is not the case, Te is plotted as a function of the heating rate and Tp
normally as a function of the square root of the product of heating rate and
sample mass (cooling rates count negative). Regression lines are determined
separately for heating and cooling sections and these are extrapolated to zero
rate. The two temperatures obtained this way are compared with the respec-
tive true temperatures Ttrue , taking a possible supercooling to be expected
into account.
- Significant discrepancies indicate an instrumental asymmetry and a separate
calibration for the cooling mode is to be performed.
One example of such a symmetry test of three power compensated DSC, where
good symmetry is established, is presented in Fig. 4.13.
-----:::=:
340.6
K
_0--- o_____
x--x----
340.2 0- _0
c _ _c
r.
0- ~
340.0
~ ---- c-c __- x
....::. 339.8 O~
c
c
_ _x---
-~
____ x
339.6
x
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
(3 ~
Fig.4.13. Test of the symmetry of three different power compensation DSCs with respect to
the heating and cooling mode using a liquid crystal (according to Hohne et al., 1993)
Calibration procedure
The calibration procedure is similar to that in heating mode (see Sect. 4.3.1)
except for those substances where the peak temperature is to be evaluated and
Tp should be plotted as a function of the square root of the product of heating
rate and mass of the sample. For details see Sarge et al., 2000, where also an ex-
ample of a calibration in cooling mode can be found.
In many cases results of a calibration in the cooling mode are different from
those obtained in heating mode, but, on the other hand, additional calibrations
are time consuming and therefore steered clear of. In addition, a calibration in
the cooling mode is not as accurate as a calibration in the heating mode. As a
consequence, the temperatures measured in cooling mode are generally more
uncertain than those determined in heating mode.
4.4
Caloric Calibration
By means of caloric calibration (for a review see Sarge et al., 1994), the propor-
tionality factor between the measured heat flow rate cJ>rn and the true heat flow
rate cJ>true on the one hand, and between the measured exchanged heat Qrn and
the heat Qtrue really transformed, on the other hand, is to be determined:
cJ>true =Kif>' cJ>rn and Qtrue = KQ Om
Strictly speaking cJ>rn in this equation should be considered the measured heat
flow rate with the instrument zero line already subtracted, but as all correction
calculations are usually done using the measured curve cJ>rn itself, only this is
used in the following.
4.4 Caloric Calibration 87
The 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
(cf. ASTM E 968-99: Standard Practice for Heat Flow Calibration of Differential
Scanning Calorimeters, 1999),
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
(cf.ASTM E 793-01: Standard Test Method for Enthalpies of Fusion and Crystal-
lization by Differential Scanning Calorimetry, 2001).
Since Qtrue = f IPtruedt and Qm =f (IPm - <!>t,l) dt, KcI> and KQ should be identical
which is not, however, the case because in practice - throughout the duration of
the peak - KcI> depends on the temperature T (and therefore also on the time t)
and is in addition a function of IP (cf. Sects. 3.1 and 4.4.3). As a result, the equa-
tion IPtrue = KcI>' IPm can indeed be integrated but KcI> must not, however, be
placed in front of the integral. As stated already, KQ is not equal to KcI>; KQ is
rather a kind of integral mean value of KcI> over the area of the peak. In practice,
the difference between the two calibration 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 of DSCs become fully effective only in
the case of perfect thermal symmetry of the measuring system. In this case,
however, the measured signal is zero. In real measurements there will always be
asymmetries in the temperature field. The effects of such asymmetries are dealt
with in Sect. 4.4.3. Another problem leading to the same effect is the non-linear
heat exchange via radiation and convection dealt with in Sect. 3.1. The conclu-
sion to be drawn for heat calibration is as follows:
The thermophysical behavior of calibration sample and sample to be meas-
ured must be as similar as possible. As this is possible only approximately, sys-
tematic errors exist which must be estimated and included in the overall uncer-
tainty of measurement (Sect. 7.3).
4.4.1
Heat Flow Rate Calibration
In almost all DSCs commercially available, a heat flow rate signal IPm is internally
assigned as a measurement signal for the actual measurement signal flT. [When
the measurement signal is put out as a voltage flU (for example in mY), the fol-
lowing applies analogously, IPm being replaced by flU.] The heat flow rate cali-
bration defines the functional relation between IPm and the true heat flow rate
IPtrue absorbed or emitted by the sample: IPtrue = KcI>' IPm (steady state, IPm with the
zero line - i. e., empty crucibles, definitions see Sect. 5.1 - already subtracted).
The proportionality factor (calibration factor, calibration function) K~ usual-

ly depends on parameters such as temperature and - what is most important -
heat flow rate. In some DSCs, K~ is made unity by electronic or software means.
In these cases, too, the relation between cPmand cPt rue 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 in-
side the sample. This method offers the following advantages:
- the electric power (heating power) can be measured easily and with high
accuracy,
- heat flow rates of differing intensity can be generated without the calibra-
tion set-up being modified,
- the steady state can be adjusted for any period of time desired; the result-
ing 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 (cPtrue = 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 "desmear-
ing" (see Sect. 5.4),
- the time constant of the measuring system can be easily determined (cf.
Sect. 7.2).
The disadvantages of the electric 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
measuring system. (Peak area calibration may also be carried out using the
peak areas furnished by these measurements; see below.) The resulting cali-
bration curve K~(T) = cPtrue(T)/( cPm(T) - cf>t,l (T is represented in Fig. 4.14b;
cf>t,l is the baseline heat flow rate (cPm-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:
K (T) _ Cp(T) f3
cPtrue = Cp f3 so that I/> - cPm(T) _ cPo (T)
(cPo - 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 and thus no additional heat leak.
Fig. 4.14a, b. Electrical heat

flow rate calibration of
a heat flux DSC (cylinder-
type measuring system) by
means of a built-in heater. m- bl
r
a Calibration peaks generat-
ed electrically in isothermal
operation or during heating
(exo up). b Calibration factor m
K,p (calculated with the data
of a), schematic) to deter-
mine the reaction (real)
heat flow rate: K,p(n = ~truel 4't,1
(~m<n - 4>J,1(D) (with ~true a
as electrical heating power
Pel), ~m measured heat flow time, tempera ture ~
rate, 4>J,1 baseline value of

the heat flow rate
tempera ture
steady state region

Fig.4.15. Heat flow rate cali-
bration by means of a known
heat capacity Cpo
~m measured heat flow rate,
~o zero line value calibration
r
factor to determine the
true heat flow rate:
K,p= Cp<nf3I(~m<n - ~o<n)
.,.~
temperature ~
The disadvantages are:

- The calibration cannot be switched off in between, i. e., checking of the base-
line (or zero line) is not possible during a run (this leads to uncertainties in
cI>m - cI>o),
- there is a temperature profIle inside the sample (this leads to a "mean CP").
The calibration substance most frequently used is synthetic sapphire (a-AI20 3 ,
corundum), cf. Table 4.9.
4.4.2
Heat (Peak Area) Calibration
For peak area calibration, a known heat Qtrue, dissipated or consumed, is com-
pared with the area of the resulting peak (Figs. 4.16 and 4.17) and the following
is valid:
tf
Qtrue = KQ (T)f [ cI>m (t) - cI>t,\ (t )] dt or
ti
tf
Qtrue =f [K~ . cI>m (t) - K~ . cI>t,\ (t)] dt
ti
In principle, K~"# K;' is valid. Since K~ depends on cI>m and T - which is propor-
tional to time in all heat flux DSCs and also in power compensation DSCs -, it
cannot be placed in front of the integral. This can be done in a first approxima-
tion only if it is assumed that cI>m and cI>t,\ are of the same order of magnitude (no
dependence on cI and the peak width is small (no dependence on T), which is
in general the case with phase transitions of organic or inorganic substances,
but not, however, with melting and crystallization of many polymers.
Integration must be carried out over the whole peak in order that contribu-
tions of the 1st derivative of the measurement signal are insignificant.
Fig.4.16. Electrical peak

area calibration.
tPm measured heat flow rate,
1
tf
tio tf beginning and end
of the calibration peak, f(m-"Idt= A
~l baseline value of the
Ii
heat flow rate (electric power
switched off)
~ ~
time ...
Fig. 4.17. Peak area calibra-

tion by means of a known
t
heat of fusion (schematic).
4im measured heat flow rate,
ti, tf beginning, end of the A
peak, 4it.l baseline value of
the heat flow rate
time
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 of the electrically generated heat
Qtrue = - f Pel. dt = - WeI. = - U i . 11 t ,
- peaks of differing size can be produced,
- calibration is possible at any temperature,
- measurement effects (peaks) can be "repeated" (simulated) byappropri-
ately presetting the development of the heating power with time,
- measurement effects (peaks) can be "encompassed" during the run by
similar calibration 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
temperature interval.
2. Peak area calibration of DSCs is usually carried out by means of well-known
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 (se-
lected) calibration materials.
The disadvantages are:

- calibration possible only at discrete temperatures,
- no adaptation to the measured peak possible,
- systematic uncertainties due to the special shape of the sample tempera-
ture 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, ef. Sect. 5.3).
Procedure
The peak area 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 Te),
- determination of KQ(Te), drawing of a calibration curve or (establishing of a
table) or inputting of 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 consid-
erations must be taken into account too),
- measurement of the repeatability errors of the calibration factors (or calibra-
tion curves). This repeatability error must be clearly smaller than the esti-
mated 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 may be sufficient (e.g., In,
ef. Sects. 2.2, 3.2 and Fig. 4.20), as the calibration factor depends on temperature
only to a smaller extent than the uncertainty of heat determination generally is.
4.4.3
Examples of Caloric Calibration
Calibration Curve
A DSC with cylinder-type measuring system was calibrated with the aid of elec-
tric 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.
t:.u
Itemperature), time ~
tempera ture ~
Fig.4.18. Electrical heat calibration of a heat flux DSC (cylinder-type measuring system with
built-in heater, schematic curves).
L'lU measured signal (voltage), KQ calibration factor for peak area evaluation (in J V-I S-I),
A peak area (in V s), Wei electrical heating energy (belonging to one peak: i . L'lU . M). In
contrast 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 K<[> from the electrical heating
power
At regular intervals, heating pulses of defined duration and power were auto-
matically generated over the whole temperature range, the measured signal
being a voltage in ]lV. The peak areas were integrated applying evaluation soft-
ware. Figure 4.18 shows schematically the measured curve and the resulting cali-
bration curve.
A heat flux DSC with disk-type measuring system was calibrated with the aid
of sapphire, the measured curves are shown in Fig. 4.19. From these curves the
cp (1') function can be calculated (cf. Sects. 5.3.1 and 6.1). Comparison with liter-
ature values (Table 4.9) enables to calculate the calibration curve.
Differences Between K<p and KQ
When DSCs are calibrated with the aid of the heat of fusion of, for example, met-
als (peak area calibration) and via the specific heat capacity (heat flow rate cali-
bration), significant 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 (see Chapter 3). This is
confirmed by measurements:
B
mW
4
1 -4
~
0
~
-B
3
~
-12
0
OJ
.c
-16
-20
00
...
4.0 B.O 120 16.0 20.0 24.0 2B.0 32.0 36.0 min 44.0
time
Fig.4.19. Heat flow rate calibration of a disk-type DSC (endo down).
Upper curve: Crucibles empty (zero line), lower curve: 129.6 mg of sapphire, 10 K min-I, (ac-
cording to Sarge et aI., 1994)
300 500 K 600

tempera ture ..
Fig.4.20. Systematic difference between K.p 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 Hohne, Gloggler, 1989).
KQ calibration factor for peak area evaluation, K.p calibration factor for heat flow rate mea-
surement, I error bar of the sapphire measurement (heat flow rate calibration), I error bar of
the heat-of-fusion calibration (peak area calibration)
10
!-LV
mw 9
ta
7
11K
6
100 200 O( 300

temperature ..
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) (according to Sarge,
Cammenga,1985).
K calibration factor, - calibration curve with sapphire (3 measurements at lO K min-I),
.... average value curve of the calibration with melting samples
For a heat flux DSC with disk-type measuring system, Fig. 4.21 shows the dif-
ferences between the average calibration curve from of a number of heat -of-melt-
ing (peak area) calibrations and the curves from three individual calibration runs
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
affected 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 measur-
ing system amounted to 5 % when the sample mass was 7 mg; but for a sample
mass of 27 mg, the scatter decreased to half this amount (Doelman et aI., 1977).
Figure 4.22 shows the (temperature-dependent) systematic difference be-
tween the calibration factors, which results when one calibration is carried out
with the calibration heater permanently installed (under the bottom of the sam-
ple cavity) and the other with the electric heater installed inside the sample (heat
flux DSC with cylinder-type measuring system).
The influences of certain parameters (sample mass, heating rate) on the
measurement results described in the following of course appear analogously
during calibration and must be taken into consideration as well.
For a heat flux DSC with disk-type measuring system a dependence on the
heating rate resulted for the heat of transition of CsCI (Fig. 4.23).
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.
30
mJ
mV s
25
t 20
KQ
15 .. '
0 100 200 300 "( 400

tempera ture ~
Fig.4.22. Differences in the electrical peak area calibration of a heat flux DSC with cylinder-
type measuring system (according to Hemminger, Schonborn, 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
sample container
3.2
kJ
i
mol
3.0
20 30 40 K/min SO
Fig. 4.23. Dependence of the molar heat of transition q of CsCl on the heating rate f3 for a
disk-type DSC (according to Breuer, Eysel, 1982)
% 10 mg
2 x x
x
r x x x x IS
x
0 o 1 mg
KQ 0 0 IS 0
0
0
0 0
-1
0.31 0.64 1.25 2.5 5 10 20 40 80
(i/(Kmin- 1 ) ~
Fig.4.24. Dependence of the calibration factor KQ change on the sample mass and the heating
rate f3 for a power compensation DSC.
KQ calibration factor for peak area evaluation (determined by means of the heat of fusion of
In; according to Hohne, Gloggler, 1989)
4.5 Conclusions Regarding the Calibration of DSCs 97
4.4.4
Caloric Calibration in Cooling Mode
The heat flow rate calibration in the cooling mode can be performed by com-
plete analogy to that in the heating mode. The substances recommended for heat
flow rate calibrations in the heating mode (see Tables 4.9 and 4.10) can also be
used for calibrations in the cooling mode. Heat capacities measured in the cool-
ing mode can differ systematically from the results determined in the heating
mode. This effect is more pronounced at measurements on low thermal diffu-
sivity materials. The reason for that temperature- and! or instrument-dependent
behavior is a non-symmetric heat flow in heating and cooling mode. It is rec-
ommended to examine this effect carefully. Unfortunately the recommended
reference materials have a rather high thermal diffusivity and do not show this
effect so much. Consequently the possible differences must be included into the
uncertainty budget. However, a correction method based on the thermal lag de-
termination is described briefly in Sect. 7.3.
The heat calibration (peak area calibration) in cooling mode can be performed
also by complete analogy with that in the heating mode. The same substances
(with first order phase transitions) can be used, but supercooling must be consid-
ered. That means the dependence of the measured enthalpy of freezing as a func-
tion of the freezing temperature must be included into the evaluation. In other
words, if a substance freezes at a temperature Tfrez, lower than the temperature
of fusion Tfus , and the specific heat capacity of the substance in the (supercooled)
liquid state (Is) differs from that in the solid state (ss) a correction according to
is necessary. In Sarge et al., 2000, differences between the specific heat capacities
of heat calibration substances in the supercooled and the solid state have been
determined from literature data and approximated by a linear function. The re-
sulting correction coefficients are given in Table 4.13 together with the sub-
stances recommended for the heat calibration in cooling mode.
4.5
Conclusions Regarding the Calibration of D5Cs
The following conclusions can be drawn from the explanations of Chapter 3

(Theoretical Fundamentals), 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 Te (cf. Fig. 4.1, definition in
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
Te should be used to characterize phase transitions.
Te should always be corrected to zero heating rate (13 =0) unless shifting of Te
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; this 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
transition 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 can be
checked by means of two samples of as different a heat capacity as possible
(which must, however, be well-known).
The uncertainty of heat capacity measurements lies generally between 5 % and
20% for disk-type and between 3 % and 5 % for power compensation DSCs.A higher
accuracy requires considerable time and effort to be spent on a proper calibration.
Precise heat capacity measurements with heat flux DSCs are possible only if
all potential systematic error sources are kept constant by ensuring as identical
conditions as possible during calibration run and measurement run (equal
mass, equal thermal conductivity, equal heating range, equal heat capacity).
Remarks: 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 K!I> of approx. 1 to 1.6% is stated in
the literature (see Jin, Wunderlich, 1993).
For the recently introduced turret-type DSC (see Sect. 2.1.2) only insufficient
experimental experience has so far been gained to state reliable figures for the
total uncertainty of heat capacity measurements.
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 reliabil-
ity of the 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 possi-
ble. The peak to be measured should be "repeated" (simulated) by electrical cal-
ibration 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. Prerequisite for a higher reliability is a very elabo-
rate calibration, for example by direct comparison with electric energy.
4.6
Reference Materials for DSC Calibration
The requirements to be met by substances to be used for the calibration of DSCs

are as follows:
4.6 Reference Materials for DSC Calibration 99
- high (defined) purity (at least 99.999 %),

- precisely known characteristic data of 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 vapor 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, informa-
tion 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 (at surface) in the case of metal samples,
- no weighing errors (calibrated balance).
A survey of reference materials, including certified reference materials, for
heat flow and heat calibration has been published by Sabbah, 1999. Several ISO
Guides are related to reference materials in general, and their certification and
usage:
- ISO Guide 30, 1992: Terms and definitions used in connection with reference
materials,
- ISO Guide 31,2000: Reference materials - Contents of certificates and labels,
- ISO Guide 32, 1997: Calibration in analytical chemistry and use of certified
reference materials,
- ISO Guide 33,2000: Uses of certified reference materials (under revision),
- ISO Guide 34, 2000: General requirements for the competition of reference
materials producers,
- ISO Guide 35, 1989: Certification of reference materials - General and statis-
tical principles (under revision).
4.6.1
Reference Materials 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 2
Not all substances meet the above requirements. Cadmium, for example, has
a high vapor pressure in the liquid state during melting, which may result in in-
juries to persons and damage to equipment.
2 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 secondary reference points
TinK tin C
Triple point of equilibrium hydrogen 13.8033 -259.3467

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 373.124 99.974
Solidification point of indium (ITS-90) 429.7485 156.5985
Solidification point of tin (ITS-90) 505.078 23l.928
Solidification point of bismuth 544.552 27l.402
Solidification point of lead 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 96l.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 1828.0 1554.8
Solidification point of platinum 204l.3 1768.2
Solidification point of rhodium 2236 1963
Solidification point of iridium 2719 2446
The substances used to realize the fixed points of the ITS-90 are listed in
Table 4.6 starting with the triple point of hydrogen. These substances have been
marked by "ITS-90" (cf. ITS-90, Preston-Thomas 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 been changed.
(The fixed-point temperatures indicated in the IPTS-68 were valid only until De-
cember 31,1989). In addition, a selected set of secondary reference points was in-
cluded (see Beford et al., 1996).
To calibrate DSCs in the heating mode, the melting point at zero heating rate
[Te(P= 0)] is set equal to the solidification point stated in Table 4.6.
Unfortunately, no suitable fixed-point materials are available for some tem-
perature ranges which is why other substances, too, must be used for the temper-
ature calibration of DSCs. Taking the above criteria into consideration, the
German Society for Thermal Analysis (GEFTA) recommends that the substances
listed in 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 customary DSCs (Cammenga et al., 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
uncertainty of the calibration when these substances are used. It comprises the
Table 4.7. Materials recommended for temperature calibration of DSCs (cf. Cammenga et ai.,
1993)
Substance Transition Uncertainty Transition a Remarks

temperature
TinK TinoC dTinmK
Cyclopentane 122.38 -150.77 50 sIs

Cyclopentane 138.06 -135.09 50 sIs
Cyclopentane 179.72 - 93.43 50 sn 1
Water 273.15 0.00 10 sll 2
Gallium 302.9146 29.7646 sn 3
Indium 429.7485 156.5985 <1 sn
Tin 505.078 231.928 <1 sn 4
Lead 600.61 327.46 10 sn 5
Zinc 692.677 419.527 sll 3
Lithium sulfate 851.43 578.28 250 sIs 6
Aluminum 933.473 660.323 sn 7
Silver 1234.93 961.78 sn 8
Gold 1337.33 1064.18 sn 8
a Solid/solid or solidlliquid transition.

1 Measure in a hermetically sealed crucible only.
2 Air-saturated, bidistilled water in hermetically closed crucible.
3 Ga and Zn very easily react with aluminum which is usually used as crucible material (also
when carefully oxidized). To avoid alloy formation with potential subsequent destruction of
calorimeter components, it is therefore necessary to proceed very carefully and with great
attention. It is advised to use only fresh samples, to examine the crucible bottom for cracks
in the aluminum oxide layer, and to immediately cut short a series of experiments, as soon
as during successive experiments an increase in the peak width or a decrease of the peak
area is observed. The authors do not accept any responsibility for damages!
Note high vapor pressure of Zn and the strong supercooling of Ga.
4 Melt reacts with Al and Pt.
5 Melt reacts with Pt.
6 Anhydrate is hygroscopic, dehydration takes place from 100C, thus movement of particles
and high water-vapor pressure (do not use hermetically closed crucible).
7 Melt reacts strongly with Pt.
8 Reacts with Pt, melt dissolves oxygen.
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.
Reference materials for the symmetry test and calibration on cooling are list-
ed in Table 4.12.
4.6.2
Reference Materials for Heat Flow Rate Calibration
The substances Ah03 in form of a single crystal (sapphire) disc and pure copper
(eu) are suitable for heat flow rate calibration purposes. They are chemically
Table 4.8 Compatibility between calibration substances and crucible materials (according to Cammenga et ai., 1993)
IS
Calibration Q)
Q)
....
substance .:: ..s
"3
1::'"
Q)
a::l
0- a::l a a'"::l .::
0
....Q) ....
.8 '"0 u 's '"0
::;;; '"0 .:: Q)
.:: "0
Crucible material ~ ~ .5 ~ '"
....:I N ~ ~ ~<Zl
Corundum, Aiz03 0 0
"+ + + + + + + +
"
+
Boron nitride, BN 0 0 + + + + + + +
Graphite,C 0 0 + + + + + + + +
Silicate glass + + + + + + + x x
Quartz glass, Si0 2 + + + + + + + + + +
Aluminum, Al + + + + x x x ....
Aluminum, oxidized + + + + x x x n~
+ + + +
Silver,Ag + + x x g;
....
Gold, Au + + + x o~::l
Nickel,Ni + + + 0
Iron, Fe + + + + .....
t;j
Stainless steel + + + + +
Platinum, Pt + + + ....~
t1>
Molybdenum, Mo + + a..
Tantalum, Ta + + + + + eo.
C/)
t"\
Tungsten,W 0 0 + + + ~
::l
::l
S
+ : No solubility and influence on melting temperature to be expected. C1CI
- : Melt dissolves crucible material, greater change of melting temperature. n
: Partial solution processes possible with negligible change of melting temperature.
eo.
....0
X : Crucible melts. S
t1>
? : Compatibility unknown.
0: Combination cannot be realized. ....en~
stable and the heat capacity is known in a wide temperature range with an un-
certainty of about 0.1 % between 100 K and 300 K for Cu and between 100 K and
900 K for Ah03 (Tables 4.9 and 4.10).
Probably the most accurate measurements on Cu were carried out by Martin,
1987 and on Ah03 by Ditmars et al., 1982. The published results are based on
the IPTS-68. There is an ongoing discussion about the correct method for the
temperature scale adjustment of thermodynamic data from the IPTS-68 to the
ITS-90 (cf. Goldberg, Weir, 1992, Weir, Goldberg, 1996, Archer, 1993, 1997).
Therefore, the use of the original IPTS-68 based data is recommended here.
Because of the practical interest the uncertainty of the specific heat capacity of
Cu and Ah03 and some results of the temperature scale adjustment are dis-
cussed here briefly.
The measurements of Martin on Cu have been carried out in the temperature
range between 20 K and 320 K and he claimed an uncertainty of 0.1 % between
30 K and 300 K. Martin did not publish his measured values, only a polynomial
representation of them (Table 4.lOa). The tabulated values (Table 4.lOb) have
been calculated by means of Martin's polynomial representation. For DSC cali-
bration purposes Cu is of interest in the temperature range between approx.
100 K and 300 K. Archer, 1997, has adjusted Martin's data to the ITS-90 and pub-
lished modified polynomial coefficients. The difference between Martin's data
Table 4.9 a. Specific heat capacity of a- Ah03 (synthetic sapphire, corundum, NIST SRM 720)
(according to Ditmars et al., 1982)
Molar mass 101.9613 g mol-I
Temperature range Fitted function cp in J mol-I K- I To
45K:s; T< 125 K 40K
125 K :s; T < 273.15 K 125K
4
273.15 K < T:s; 2250 K L CiTi
i=-3
ai bi Ci
-3 -1.32506 . 108
-2 4.54238.106
-1 -5.475599. 104
0 6.966.10- 1 2.1993.10 1 2.574076. 102
1 5.9387 . 10-2 3.8853 . 10-1 -1.715032. 10-1
2 4.0357 . 10-3 1.3955 . 10-3 1.2897189 . 10-4
3 9.5173.10- 5 - 8.3967 . 10-5 -4.60768. 10-8
4 -3.5910. 10-6 1.9133. 10-6 6.31755. 10- 12
5 - 6.498 . 10-7 -3.1778.10- 8
6 4.089.10-9 2.9562 . 10-10
Table 4.9 b. Specific heat capacity of a- Alz0 3 (synthetic sapphire, corundum, NIST SRM 720)
Molar mass 101.9613 g mol- 1 (according to Ditmars et al., 1982 )
TinK cp in Jg-l K- 1 cp in Jmol- 1 K- 1
100 0.1261 12.858

150 0.3134 31.95
200 0.5014 51.12
250 0.6579 67.08
300 0.7788 79.41
350 0.8713 88.84
400 0.9423 96.08
450 0.9975 101.71
500 1.0409 106.13
550 1.0756 109.67
600 1.1039 112.55
650 1.1271 114.92
700 1.1467 116.92
750 1.1636 118.64
800 1.1783 120.14
850 1.1913 121.47
900 1.2030 122.66
950 1.2138 123.76
1000 1.2237 124.77
1050 1.2330 125.72
1100 1.2417 126.61
1200 1.2578 128.25
1250 1.2653 129.01
1300 1.2724 129.74
1350 1.2792 130.43
1400 1.2856 131.08
1450 1.2917 131.70
1500 1.2975 132.29
1550 1.3028 132.84
1600 1.3079 133.36
1650 1.3128 133.85
1700 1.3173 134.31
1750 1.3214 134.73
1800 1.3253 135.13
1850 1.3289 135.50
1900 1.3324 135.85
1950 1.3356 136.18
2000 1.3387 136.50
2050 1.3417 136.80
2100 1.3446 137.10
2150 1.3477 137.41
2200 1.3508 137.73
2250 1.3540 138.06
Table 4.lOa. Specific heat capacity of copper (according to Martin, 1987)

Temperature range Fitted function cp in Jmol-I K- '
20K < T~ 320K
aj
0 - 0.8209550462989
0.1877774093791
2 -0.1572548380193. 10-1
3 0.5828318431167. 10-3
4 -0.1420296394933.10-5
5 -0.3466012703872 . 10-6
6 0.1030643882976.10-7
7 -0.1621415050746. 10-9
8 0.1678243825986. 10-11
9 -0.1225826347399.10- 13
10 0.6497335630403 . 10-16
11 -0.2516918128676. 10-18
12 0.7065147741950. 10-21
13 -0.1400307720276.10-23
14 0.1858984964834.10-26
15 -0.1483657580275.10-29
16 0.5382629833814. 10-33
Table 4.10 b. Specific heat capacity of copper (according to Martin, 1987)

TinK cp in Jg-I K- 1 cp in Jmol-I K- 1
100 0.2520 16.02

120 0.2871 18.25
140 0.3122 19.84
160 0.3306 21.01
180 0.3444 21.89
200 0.3550 22.56
220 0.3635 23.10
240 0.3703 23.53
260 0.3758 23.88
280 0.3806 24.19
300 0.3848 24.45
320 0.3884 24.68
0.008
-o. 008 L--~-'---~-'---'---'-~--'---'---'---'---'______L--.>--.J
250 750 1250 1750 K 2250

temperature - -
Fig.4.25. Differences between tabulated and calculated (polynomial presentation) specific
heat capacity data of sapphire (Ah03): p =I!1c cr -cr
yn. (cf. Ditmars et al., 1982) in the tem-
perature range between 250 K and 2250 K
and the adjusted data by Archer is between -0.022% at 100 K and +0.054% at
320 K. Sabbah et al., 1999, used the Goldberg and Weir, 1992, method and came
to the conclusion that the adjustment is smaller than 0.06 %. The adjustments
of both authors are smaller than Martin's claimed uncertainty in the tempera-
ture range between 100 K and 300 K.
The specific heat capacity of Al20 3 has been determined in the tempera-
ture range between 10 K and 2250 K (Ditmars et al., 1982, Table 4.9a and 4.9b).
Figure 4.25 shows the difference between the tabulated data and the data calcu-
lated by means of Ditmars polynomial representation.
Although Ditmars estimated an uncertainty of 0.1 % in the temperature range
between 100 K and 350 K only, it is generally accepted that the uncertainty of the
measurements is better than 0.1 % in the temperature range between 100 K and
900 K. At higher temperatures the uncertainty increases to values between 1 %
and 2 % at 2250 K. The influence of the IPTS-68 to ITS-90 adjustment and the
data treatment (fitting) has been determined by Archer, 1993, and amounts to
approx. 0.12% at 100 K, up to 0.09% between 300 K and 430 K, up to 0.31 %
between 900 K and 1700 K and 1.4% at 2250 K. The ITS-90 adjustment of Weir
and Goldberg, 1996, amounts up to 0.18 % between 900 K and l300 K.
NB: eu should not be used at higher temperatures and only in an inert purge
gas flow, to avoid oxidation which would produce an additional heat flow rate
and falsify the measurement.
4.7 Additional Calibration in Temperature-Modulated Mode of Operation 107
4.6.3
Reference Materials for Heat (Peak Area) Calibration
The heats of transition of pure substances for peak area calibration should have
been measured with precision calorimeters. Furthermore, the results should be
traceable to basic SI units and an uncertainty should be stated. Such results are
available for only a few substances consequently fulfilling the above require-
ments. When several independent precise measurements have been carried out
on one substance, the ranges of uncertainty stated in some cases do not overlap
so the estimate of the best value and of the overall uncertainty is problematic.
Recommended substances are listed in Table 4.11 a. Table 4.11 b contains a list
of further commercially available substances used for calibration. The latter are
not recommended here because of a lack of information about the used certifi-
cation procedure and the uncertainty of the results.
The analysis of the most reliable enthalpy of fusion measurements allows a
statement of an uncertainty less than 0.3 % for Ga and In and 0.5 % for Sn and Bi.
For peak area calibration in cooling mode most of the substances used in
heating mode are suitable as well, Table 4.13 list these substances together with
the correction of the enthalpy of crystallization which takes care of the temper-
ature dependence of the transition enthalpy. The correction must be applied in
cases of large supercooling.
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 enthalpies
of transition indicated in Table 4.11 a are therefore sufficient to allow these sub-
stances to be used as calibration substances.
In special cases relative uncertainties less than 1 % (e. g., Archer, Carter, 2000) can
be achieved. Calibration by means of electrically generated heat should be aimed
for precision measurements. In addition, reference is made to the discussion of
thermodynamic aspects with respect to the peak area calibration in Sect. 4.2.
The discussion about the suitability of various substances for caloric calibra-
tion is still going on at the international level (St01en, Gr0nwold, 1999). The fol-
lowing complexes of problems are concerned:
- backing-up and re-determination of characteristic data, including an esti-
mate 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 or other transitions), calibrations in the cooling mode, or substances
suitable for high temperatures.
4.7
Additional Calibration in Temperature-Modulated Mode of Operation
Quantities determined from the periodic part of the heat flow rate in temperature-
modulated mode of operation (cf. Sect. 2.4.2) depend usually on the frequency of
modulation. The reason is that heat transport needs time and the thermal inertia
......
o
00
Table 4.11 a. Reference materials for heat (peak area) calibration
Substance Ttrs (ITS-90) Type of Qtrs Qtrs Qtrs Qus

in C transition a (St01en et aI., 1999) (NIST) (PTB) (Sarge et ai. , 1997)
in J g-l in J g-l in J g-l in J g-l
Cyclopentane -150.77 sis 69.60 0.35

Cyclopentane -135.09 sis 4.91 0.05
Cyclopentane -93.43 sll 8.63 0.09
Gallium 29.76 sll 79.97 0.27 80.097 0.032 b 80.14 0.33 f 79.88 0.72
Biphenyl 69.26 sll 120.41 0.57 c ...
Indium 156.60 sll 28.58 0,07 28.51 0.19 d 28.64 0.11 f 28.62 0.11 (")
Tin 231.93 sll 60.42 0.17 60.22 0.1g e 60.24 0.27 f 60.40 0.36 ~
Bismuth 271.40 sll 53.33 0.26 53.14 0.22f 53.8 2.1 ....
~
Lithium sulfate 578.28 sis 228 10 g"
Aluminum 660.32 sll 399.9 1.3 398 9 o
.....,
Silver 916.78 sll 104.6 2.1 CI
Gold 1064.18 sll 64.6 1.5 ~
'"....
a s: solid; 1: liquid. '"g.
b Archer, 2002. e.
Vl
()
C NIST SRM 2222 (NIST: National Institute of Standards and Technology, USA).
d NIST SRM 2232.

e NIST SRM 2220. 5"
~
f PTB Certified Reference Materials (PTB: Physikalisch-Technische Bundesanstalt, Germany). (")
S-....
s"
~
....
'"en
Table 4.11 b. Further Certified Reference Materials for temperature and heat calibration
Substance Ttrs Qtrs

in C inkJ mol- 1
Phenyl salicylate 41.79 19.18

Biphenyl 68.93 18.60
Naphthalene 80.23 18.923
Benzil 94.85 23.26
Acetanilide 114.34 21.793
Benzoic acid 122.35 17.98
Diphenylacetic acid 147.19 31.16
Lead 327.47 4.765
Zinc 419.53 7.103
Aluminum 660.33 10.827
source: http://www.lgc.co.uk.
(LGC: Laboratory of the Government Chemist, UK).
Table 4.12. Substances recommended for temperature calibration in the cooling mode
Substance Transition Evaluate Phase transition Supercooling

temperature
Ttrs in C Ttr, in K 8TinK
Adamantane s~s Te -64.49 208.66 <1

M24 a SA~N Tp 66 339 <0.5
HP-53 b SA~N Tp 116 389
BCH-52c N~I Tp 162 435
Indium s~l Te 156.6 429.8 <2
NaN0 3 s~s Tp 276 549 <0.5
Zn d s~l Te 419.5 692.7
a 4-cyano-4' -octyloxybiphenyl.
b 4-( 4-pentyl-cyclohexyl)-benzoic acid 4-propyl-phenyl ester.
c 4' -ethyl-4-( 4-propyl-cyclohexyl)-biphenyl.
d Zn very easily reacts with aluminum which is usually used as crucible material (also when
carefully oxidized). To avoid alloy formation with potential subsequent destruction of
in the aluminum oxide layer, and to immediately cut short a series of experiments as soon
s solid.
SA smectic A (liquid crystal phase).
N nematic (liquid crystal phase).
1 liquid.
Table 4.13. Substances recommended for heat (peak area) calibration in the cooling mode
Substance Fusion Enthalpy of fusion Temperature dependence b

temperature 1'1fus H in Jg-1 d1'1 trs H/dT= C;(T) - Cp(T)
Tfus in C = 1'1trs Cp (T) = a + b (Ttrs - Tfus)
a in Jg -IK- 1 binJg- 1K- 2
Cydopentane -150.77 C 69.60 c +0.38 -5.8. 10-3

Cydopentane -135.09 c 4.91 c -0.058 0
Cydopentane -93.43 8.63 +0.16 +3.1 . 10-3
Gallium a 29.7646 see Table 4.11 a +0.031 -4.6. 10-4
Indium 156.5985 see Table 4.11 a -0.0026 -2.6. 10-4
Tin 231.928 see Table 4.11 a -O.oI8 -3.1 . 10-4
Zinc a 419.53 108.09 +0.012 -7.3 . 10-4
Lithium sulfate 578.28 c 228.1 c -0.15 -7.7. 10-3
Aluminium 660.323 398.1 -0.19 +3.2. 10-4
a Ga and Zn very easily react with aluminum which is usually used as crucible material (also
when carefully oxidized). To avoid alloy formation with potential subsequent destruction of
in the aluminum oxide layer, and to immediately cut short a series of experiments, as soon
b When the equation (Sarge et at, 2000)
1'1trsH(Ttrs) = -1'1fusH(Tfus) + (a - bTfus ) (Ttrs - Tfus ) + 1/2 b (nrs - Tfus)
is used for the determination of the enthalpy of solidification from the enthalpy of fusion,
the signs of the quantities used are strictly to be taken into consideration: fusion processes
are endothermic and, therefore, enthalpies of fusion are positive; solidification processes are
exothermic and, therefore, enthalpies of solidification are negative. The signs of the given
differences in heat capacity relate to the transition of the low-temperature phase to the high-
temperature phase. Thus, for example, supercooling (Ttrs - Tfus < 0) leads to an increase in
the enthalpy of solidification when the heat capacity during the solid/liquid transition in the
temperature interval decreases (1'1ep < 0). The enthalpy of solidification being however neg-
ative, this means a reduction of the amount of the enthalpy of solidification.
The procedure can be applied for solid/solid transitions in an analogous manner.
c Solid/solid transition.
causes a loss in amplitude and a shifting of the phase of the heat flow rate modu-
lation on the path from the furnace to and into the sample. Therefore the (appar-
ent) heat capacity, determined from the modulated part of the heat flow rate sig-
nal depends strongly on frequency (or period of modulation) used. That's why the
TMDSC must be calibrated relating to the frequency dependence of the heat ca-
pacity in addition to the normal calibration of temperature, heat flow rate and
peak area mentioned above. The relevant equation in this special case reads:
Cp, true = k (w) . Cp, meas. (4.5)
The normal (vibrational) heat capacity does not depend on frequency and this
fact is used to determine the calibration factor for the frequency of the temper-
ature modulation used, or, in the case of investigations at different frequencies,

as a function of frequency in the region of interest.
The temperature-modulated mode of operation enables to determine time-
dependent (complex) heat capacity (see Sects. 3.3.2 and 5.5.2) from the modu-
lated heat flow rate. From the theory of TMDSC (see Sect. 3.3) the following
equation is valid within the scope of linear response
T~eas.(W) = T~ue(w)' P*(w)
with pew) the complex transfer function of the device. Because of the propor-
tionality of temperature gradient and the heat flow rate a similar equation can
be formulated for the heat flow rate
q,~eas.(w) = q,~ue(w) P*(w) (4.6)
and even for the heat capacity which is linearly connected with the heat flow rate
(4.7)
Generally the functions are complex (characterized by the star) but, of course, the
equation is even valid for real functions. Comparing Eq. (4.7) with Eq. (4.5) yields:
_ 1
K(w)=-- (4.8)
P*(w)
the calibration function sought after is the reciprocal transfer function of the
DSC. The latter can be determined easily with the tools of transfer theory
(Hahne et al., 2002). One elegant possibility is to measure the response of the
heat flow rate on a sudden step-like temperature change of the DSC. This func-
tion, if normalized, is the apparatus (Green's) function of the DSC, which, after
Fourier transform, results in the complex transfer function.
The procedure is simple, Fig. 4.26 shows an example of such a measurement.
To correct for the unavoidable asymmetry influences on the signal, a measure-
ment with empty pans (zeroline) was subtracted. Fourier transform results in a
complex function, which after normalization is plotted as magnitude and phase
in logarithmic form (Bode plot, Fig. 4.27). The reciprocal function is the calibra-
tion factor [see Eq. (4.8)], a complex function of was well. With that the calibra-
tion, in principle, is completed, as the correction of magnitude and phase at any
frequency can be obtained according to the rules of complex numbers.
However, the normal evaluation software of temperature-modulated DSCs
doesn't contain this calibration procedure and the user has to perform the cali-
bration in a more usual way. Any complex quantity can be expressed as magni-
tude (absolute value) and phase angle, consequently calibration can be done for
both quantities separately. Normally only the magnitude (absolute value) of the
heat capacity is of interest and the phase information is disregarded. Therefore
there is a demand for separate procedures of magnitude and phase calibration
which are to be presented in what follows.
/\ 141
I \
I \
I
I \
;
,......II....... \\...........................tg.r!~p.~r.gJ~xg.::::f.:......... 0 C
: I \
aJ ; I \
d !
............ ,
I
I
\
\ 140
I \
:3
o ! ._ \,___ sample line d
~: ", -.. -----=---... ----;;;~;;--"'='=--~=-- <-
ill
D..
E
ill
zero line subtraded

OL---~---L __~==~====b===d 139
-10 0 10 20 30 s 50
time - - - -
Fig.4.26. Temperature-step (0.5 K, dotted) response signals (dashed) of a power compensa-
tion DSC containing a sapphire sample (25 mg) as well as empty pans (for zeroline). Subtrac-
tion results in the response function (solid line)
c 5
rad
Plw)
4
0100
I I
3
2
1i 0.010 ill
.,,,
Vl
d d
..c
D..
0.001 0
--
001 01 10 rad S-l 10
(;J
Fig.4.27. Transfer function (Bode plot, solid lines), the Fourier transformed step response
function from Fig. 4.26, together with the really measured values (symbols) at different dis-
crete frequencies (dashed line: magnitude of the transfer function from non -zero line corrected
step response)
Any complex quantity can be separated into magnitude and phase:
f* = 11* I exp(i6)
from this follows, that for a product of complex quantities as in Eq. (4.5) or (4.7)
the magnitudes have to be multiplied, whereas the phases add up. In other
words the magnitu~e of the. complex heat capacity has to be corrected with a
calibration factor IK (w) 1 = K(w), whereas the phase has to be shifted by a cer
tain 6K (w).
No reference substance with known complex (frequency dependent) heat
capacity, however, exist. The normal vibrational heat capacity is a real-valued
quantity, in other words the magnitude of Cp is independent of frequency, in the

range used with TMDSC, and the phase angle is zero, this makes it suitable for
our purpose and the respective calibration procedure easy.
4.7.1
Calibration of Magnitude
A simple procedure for magnitude calibration starts from Eq. (4.5) and deter-
mines the calibration factor by means of a substance with known heat capacity:
The recommended reference materials for heat flow rate calibration (Sect. 4.6.2)
can be used for this purpose as well, sapphire is inert and the vibration frequen-
cies are many orders of magnitude faster than the modulation frequencies of
TMDSC. As a result the heat capacity Cp,Ref (see Table 4.9) is the same for all fre-
quencies (or periods) used in the temperature-modulation experiment. This has
been done for several frequencies and the (reciprocal) calibration factors are
added to Fig. 4.27. As expected the results of this magnitude calibration coincide
with the magnitude of the step-response evaluation performed with the same
sample in the same DSC, but needs much more time. Of course the simple pro-
cedure is sufficient and faster, if only one frequency (period) is used for the
TMDSC experiments. In this case there is a certain calibration factor instead of
a calibration function. It should, however, be born in mind that this factor
changes if the frequency (period) is changed. For every DSC there exists a fre-
quency range where the TMDSC calibration factor is close to one (proper heat
flow rate calibration for granted).
4.7.2
Calibration of Phase
The phase signal is strongly influenced by heat transfer properties of the DSC as
well as the sample itself. From Eq. (3.28) (see Sect. 3.3.3) follows that the phase
shift of a simple RC-element depends not only on frequency but even on the
thermal resistance and heat capacity:
i. e., the phase signal changes not only with temperature (as Cp and Rtb depend on
temperature) but even when the heat transfer to the sample and its heat capacity
changes. This makes the determination of an exact phase angle not so easy.
A possible calibration procedure starts from the fact that a pure vibrational
heat capacity is real-valued, i. e., the phase shift is zero. The measured phase shift
must be assigned to the heat transfer from the furnace to the sample, this seems
to be the correction c5K at the frequency in question we were looking for. Unfor-
tunately this is not true, as the heat capacity of the sample is different from the
heat capacity of the reference material (sapphire), even the phase shift is differ-
ent. The correction depends strongly on the sample and its properties Rth and
Cp Consequently a phase calibration can only be done if the total arrangement
DSC plus pan plus sample is unchanged. The only way out is to perform the above
mentioned step response measurement and determine the phase shift from the
transfer function. A temperature step can be done easily at every temperature
with the experimental setup of the TMDSC. Of course, one should choose a tem-
perature well outside of any transition or reaction, to be sure that only the vi-
brational heat capacity contributes to the signal.
However, there is another problem arising from the fact that the heat capaci-
ty of a sample (normally) changes during a transition or reaction. This implies
that even the phase signal changes. In other words, the correction 15K changes
during transitions and reactions and is not a constant value. This is in particu-
lar true for the glass transition region, where the heat capacity becomes com-
plex, and the magnitude changes as well. Weyer et aI., 1997, suggested a special
correction for this case where the correction 15K is assumed to change in a sig-
moidal way proportional to the magnitude of the complex heat capacity. This
assumption is true for RC-elements for small phase shifts, where the tangens can
be approximated by the argument. The authors could show that this method is
successful for the glass transition and yields the true imaginary part of the com-
plex heat capacity.
To sum up one may state that the phase correction 15K has to be determined at
different temperatures and, in particular, on both sides of a transition or reac-
tion to take care of possible heat capacity changes. The arrangement of sample
and DSC must be the same as for the TMDSC measurement, i. e., the sample pan
should not be touched. It should be ensured that only vibrational heat capacity
contributes to the signal at the temperatures in question. A special reference
substance is not needed, as the vibrational heat capacity of any sample is real-
valued and therefore results in a zero phase shift. The phase shift in dependence
on frequency at a certain temperature can easily be determined by measuring
the temperature-step response and Fourier transform the respective (normal-
ized) heat flow rate function.
It should be emphasized that the phase correction 15K changes during transi-
tions, the problem to get the right correction function for this region is not
solved yet in a satisfactory manner. The knowledge of the correction function
is, however, an indispensable condition to determine the proper complex heat
capacity. This is not an easy task and the reader interested in these questions is
referred to special literature (see Merzlyakov et aI., 2002).
5 OSC Curves and Further Evaluations
A Scanning Calorimeter measures heat flow rates in dependence of temperature

or time. Modern DSCs are nowadays always connected with a data acquisition
system and a powerful computer (PC). This allows one to present the measured
data online on a monitor in form of a curve. Normally the heat flow rate versus
the program temperature (or time) is plotted, but it is also possible to calculate
other quantities from the originally measured values and draw the respective
graphs on the screen as well. Modern computer techniques make it possible to
do even complicated evaluations of the just measured values in the background
while the measurement runs.
It should be remembered that the original quantities measured in a DSC are
a temperature difference (in form of a voltage) and the set value of the temper-
ature of the furnace (or sample support) which is electronically controlled to
follow a certain temperature program (other quantities are in certain cases
T, T,
temperature ..
Fig.S.1. Curve measured by a DSC with step of the baseline (Cp change) and endothermic
peak (Ist order transition).
IlCp change of the sample's heat capacity, Ti> Tf initial peak temperature, final peak tempera-
ture, CD initial segment of the measured curve, @ step of the measured curve due to IlCp ,
@ measured curve, interpolated baseline (between Tj and Tf in the peak region), final
segment of the measured curve, peak (measured curve)
116 5 DSC Curves and Further Evaluations
measured additionally). The temperature difference is internally transformed

into a differential heat flow rate (see Chapter 3) which is gathered at regular
intervals together with the temperature (or time) in question and stored inter-
nally. This raw data set establishes the DSC curve, it forms the basis for all fur-
ther evaluation.
The DSC curve (the measured curve, see Fig. 5.1) offers quick information on
the total measuring process. In addition to the usual measuring effects (Cp
changes, transitions, reactions) it can be seen
- whether the predetermined temperature range has been completely covered,
- whether disturbances of the apparatus (mechanical, electrical) occurred,
- whether there were irregularities or unusual shapes of the baseline,
- whether the characteristic temperatures and peak areas lie within the expect-
ed range.
5.1
Characteristic Terms of DSC Curves
Some characteristic terms are used to describe a measured curve. The respective
definitions are given in the following (see Fig. 5.2).
- The zero line is the curve measured with the instrument empty (i. e., without
samples and without sample containers, crucibles), or with empty sample
containers (crucibles without samples). It shows the thermal behavior of the
peak (endothermic)
1 /
1
1
f
I 1 \
I .-!-\-Q-
---I
-- I '\
I 1
: baseline (interpolated)
1 I I
I 1
\
zero line
T; T. T,
temperature ~
Fig.5.2. Definition of zeroline, baseline, peak and the characteristic temperatures (defini-
tions see text).
Ti initial peak temperature, T. extrapolated peak onset temperature, Tp peak maximum tem-
perature, Tc extrapolated peak offset temperature, Tc final peak temperature
5.1 Characteristic Terms of DSC Curves 117
measuring system itself and the degree and influence of unavoidable asym-
metries. The smaller its range of variation with temperature or in time (re-
peatability, see Sect. 7.2), the better the instrument.
- The baseline is the part of the heat flow rate curve produced of the DSC dur-
ing steady state conditions (no reactions or transitions in the sample). The
baseline is the sum of the zeroline and the heat flow rate caused by the differ-
ence of heat capacities on the sample and reference side in scanning mode.
- A peak in the measured curve appears when the steady state is disturbed by
thermally activated heat production or consumption in the sample (e. g., from
transitions or reactions). Peaks in heat flow rate curves, which are assigned to
endothermic processes, are normally plotted "upwards" (positive direction),
as heat added to a system is defined as positive in thermodynamics by inter-
national convention. A peak begins at Ti (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 processes associated with a
heat (e.g., melting or crystallization) lead to peaks (except for changes in the
heat transfer between the sample and the AT-sensor). Other transitions (e.g.,
glass transition) only lead to changes in the shape of the measured curve, for
example step-like changes (see Sect. 6.4).
- The interpolated baseline is the line which in the range of a peak is con-
structed in such a way that it connects the measured curve before and behind
the peak as if no reaction heat had been exchanged, i. e., as if no heat (peak)
had developed (for interpolation see Sect. 5.3.2).
- The characteristic temperatures are defined as follows:
- Ti 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
baseline. [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 baseline (not necessarily the absolute
maximum of the curve of measured values).
- T, Extrapolated peak offset temperature
Here the auxiliary line (see above) through the descending peak slope in-
tersects the baseline.
- Tr Final peak temperature
Here the curve of measured values reaches again the baseline, the peak is
completed.
5.2
Parameters Influencing the DSC Curve
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 anal-
ogy to each measurement.
Note: For highly precise measurements, the zeroline must be determined
prior to and after every measurement. Temperature and heat calibration must
then also be checked (verification) at regular intervals (depending on the spe-
cific 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, infor-
mation 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 zeroline (without crucibles) is influenced by the heating rate,
the kind of purge gas and its flow rate, likewise by the temperature of the sur-
roundings and by surface properties of the measuring system. If the zeroline
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 the furnace and from differences between the emissivities
of the two crucibles (lids), and influences due to type and material of the cru-
cible used.
2. Point 1. is also applicable to the shape outside a peak of the measured curve
with sample and reference sample placed in the crucibles. In addition, the
properties of the sample and reference sample (heat capacity and its temper-
ature 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 course 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. If this occurs
abruptly this causes a step-like or peak-like change as well, but this usually
appears statistically and can thus be distinguished from real Cp-changes or
transitions of uniform samples (with the same thermal history) which always
occur 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. 3.11,4.8),
- the thermal conductivity of the sample (cf. Fig. 3.13),
- the mass and heat capacity of the 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 of the sample in the crucible or measuring system (cf. Fig 4.11),
- the kind of purge gas in the measuring system, which influences the sepa-
ration (resolution) of closely adjacent peaks (and the calibration).
5.3 Further Evaluation of DSC Curves 119
4. In addition, attention must be paid to:

- the sample purity,
- the thermal history of the sample
- 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. For direct comparison of measurements all parameters
should be as similar as possible. Use samples of defined state, shape and purity.
5.3
Further Evaluation of DSC Curves
The original curve, measured with a DSC, contains not only the heat flow rate of
the sample, but even other parts caused by unavoidable asymmetries. In addi-
tion the heat flow rate is falsified (smeared) by the thermal inertia of the meas-
uring system (thermal lag}. To get the real sample heat flow rate and - if possible
- sample temperature and eliminate the disturbances, conversion of the original
measured curve must be performed. This is done in two steps, firstly, the differ-
ential heat flow rate into the sample is calculated and, secondly, the influence of
thermal lag is eliminated (desmearing) if needed.
5.3.1
Determination of the Real Sample Heat Flow Rate
To eliminate all influences from asymmetries of the DSC in question, the zero-
line must be subtracted from the measured curve. For precision measurements
the proper zeroline should be determined prior to and after a sample run, to
check whether there are significant differences, and subtracted. Of course all
parameters of the measurement must be kept the same, only the sample pan is
exchanged for an empty pan of the same type and with the same mass. Depend-
ing on the DSC and the stability of the surrounding conditions the zeroline may
be rather stable during a day (should be tested). If that is the case, and for less
precise demands, it could be sufficient to measure the zeroline only once a day
(after sufficient equilibration) and use it for all measurements of the same kind.
However, the zeroline must be subtracted from the measured heat flow rate to
get the true sample heat flow rate
The advantages of this somewhat time-consuming procedure are, however,

striking:
- the influences from apparatus and reference sample are eliminated,
- the remaining heat flow rate curve is almost straight and horizontal (outside
thermal events), as the heat capacity of the sample only changes slowly with
temperature,
- determination of the baseline is easy (no curvature),

- step-like changes of the heat capacity are easily detected and cannot be mixed
up with a peak,
- weak exothermic events (e. g., chemical reactions) can easily be detected when
the remaining heat flow rate deviates from the straight line and curves down-
ward (exothermic heat flow rate counts negative),
- changes of the heat capacity during transitions (melting, crystallization) can
easily be detected,
- specific heat capacity and thermodynamic quantities can easily be calculated.
This is why we recommend to subtract the zeroline prior to all further evaluation.
Note: The recently introduced turret-type DSC (TA Instruments) takes the
possible asymmetry of heat capacity and thermal conductivity (i.e., the zero-
line) into account and corrects for it internally with the so-called Tzero tech-
nique. As a result the measured curve is automatically zeroline-corrected and
the DSC produces a measured curve, which is almost zero if an empty sample
pan is used.
Heat Capacity and Other Thermodynamic Functions
The specific heat capacity is easily calculated from the measured heat flow rate
with the zeroline subtracted. From the definition dQ = CpdT follows:
dQs
Cp dt cPs cPm - cPo
cp = - = - - - = - - =
m dT m . f3 m . f3
m-
dt
In other words the specific heat capacity in steady state (no transitions and
reactions) is the sample heat flow rate divided by sample mass and heating rate.
As the latter quantities are given and constant for a DSC run, the cp curve and the
measured curve with zeroline subtracted are identical beside a constant factor.
Therefore the evaluation programs of modern DSCs calculate the measured
curve (after zeroline subtraction) in units of specific heat capacity (J g-l K- 1). It
should be mentioned that this quantity equals the real (vibrational) heat capaci-
ty only in the absence of transitions or reactions (peaks). In the case of thermal
events in the sample (in the region of peaks) it is an "apparent heat capacity", dif-
ferent from the static (vibrational) heat capacity and, in addition, "smeared" (see
Sect. 5.4). However, there are great advantages of such a presentation; the advan-
tages mentioned above are still all valid and there are even more:
- the heat capacity curve (outside peaks) can be compared with heat capacity
values from the literature, and serve as an calibration check,
- the heat capacity is a thermodynamically well defined quantity and opens the
way to determine thermodynamic potential functions.
Because of these advantages we recommend to calculate the specific heat capac-
ity curves generally from the original measured curve.
Normally the DSC runs at constant pressure, this yields the heat capacity at
constant pressure Cp (T). The enthalpy follows per definition by integration:
T
H(T) - H(To) = I Cp(T)dT
To
This function offers the possibility to determine enthalpy changes connected

with transitions or reactions in a thermodynamically exact way (cf. Sect. 6.1.6).
Other thermodynamic functions can be determined similarly, the entropy reads:
S(T) - S(To) = JCp(T)

To T
dT
and we may even calculate the free (Gibbs) energy:

G(T) - G(To) = H(T) - H(To) - T(S(T) - S (To.
5.3.2
The Baseline and the Determination of Peak Areas (Enthalpy Differences)
The baseline is produced for steady state conditions (see Sect. 5.1 and Figs. 5.1,5.2).
Within a peak, i. e., during a transition or reaction, the baseline is defined as that
curve between Tj and Tf in the region of a peak, which would have been recorded if
all cp changes (and changes of heat transfer and other influences) had occurred but
no heat of transition had been released. This is a virtual line which never is record-
ed and therefore must be determined otherwise. It is, however, an essential curve,
because the area between the baseline and measured curve is a measure for the la-
tent heat of the respective process (heat of transition or reaction).
Construction of the Baseline

The baseline has to be constructed in the region of peaks of all kinds to be able
to determine peak areas. For 1st order transitions, with or without coupled Cp
change (and/or change of the heat transport mechanism), the baseline can be
constructed in different ways (cf. Hemminger, Sarge, 1991). A definite baseline
can be constructed only if a pure Cp change 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 kinetics), the measured curve and
the respective baseline must possibly be "desmeared" as well (cf. Sect. 5.4). The
common baseline construction methods are as follows.
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 (example: annealing of
lattice defects during recrystallization of a plastically deformed metal). In
these cases, the baseline is interpolated in the course of the measured curve
of the 2nd run. The baseline uncertainty corresponds to the repeatability
error of the DSC. If the heat capacity of the reacted product is different from
that of the original sample, this method cannot be used.
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 from the measured curve. Thus, the
Cp change coupled to it can be calculated.
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 Ti and Tf> the following is valid in
good approximation (see Fig. 5.3).
( d<P) = (1- a)(d<P) + a(d<P)

dT bl dT Ti dT Tr
That is to say, aCt) determines how quickly the slope of the baseline changes
from the slope of the measured curve at Ti (with a = 0 for T < Ti ) in the in-
terval Ti ~ T ~ Td with a = a (t) 1to the slope of the measured curve in T[ (with
a = 1 for T> T[). In the interval between Ti and Tf> the following results by
approximation for the functional values ~I:
~I = (1 - a) <Pi,ex + a <p[,ex
where <Pi, ex and <p[, ex are the segments for the measured curve extrapolated
into the peak range from the left-hand and right-hand side, respectively. That
..
T,
tempera ture
Fig.5.3. Construction of the baseline taking the degree of reaction a (t) into account (ac-
cording to van der Plaats, 1984).
Tj initial peak temperature, Tr final peak temperature, <1\ex> <Pr,ex measured curve extrapo-
lated from Tj , Tf into the peak region
is to say, q,i,ex and q,f,ex 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 at Ti and Tf is not
extremely great; otherwise, this could be taken into account as well.)
Integration of the peak area can then be carried out using the difference be-
tween the (desmeared) measured curve and the baseline q,(T) - ~1(T).
4. If the heat capacity of the sample changes suddenly by L'lCs at the transi-
tion temperature Ttrs with a constant sample temperature prevailing dur-
ing transition (first-order phase transitions), the change in the heat capa-
city is described by a step function at T = Ttrs By the "RC-elements" (cf.
Sect. 3.3.3) 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 trans-
formed 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 uncertainties of the measurements and must
be taken into account when the overall uncertainty of measurement is esti-
mated.
Fig. 5.4. Possible baselines

when the reasons for the
baseline change between
Ti and Tf are not known.
Ti initial peak temperature,
Tf final peak temperature,
r
- linear extrapolation into
the peak region and jump
somewhere between Ti and
Tf - - - constructed by means rp
of an "apparent" transfor-
mation function, .... straight
line between Ti and Tf
T, 0
temperature ..
The baseline always undergoes changes when Cp changes; additional changes

may result when:
- the thermal conductivity of the sample changes substantially during transi-
tion,
- the thermal resistance between sample and temperature sensor (heat ex-
change conditions) 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 reac-
tion and the sample is positioned in an open crucible, or when the crucible is
deformed.
When such changes coincide with Cp changes, the uncertainty how the baseline
should be determined generally increases. There is then no method for con-
structing a "true" baseline.
As emphasized before, all the theoretical discussions presume a zeroline of
the DSC which is at least an absolute straight line. If not, the zeroline must first
be subtracted from the measured curve before the described procedures can be
carried out correctly. It is highly recommended to do this for every measure-
ment. The advantage to transform the measured curve into heat capacity units
(see above) is obvious in this context; taking possibly known heat capacities of
the sample before and after transition as a reference the Cp change during tran-
sition (i. e., the baseline) can easily be calculated.
Determination of Latent Heat Connected With a Peak Area
Integration of the measured curve with the baseline subtracted yields the
heat of transition or reaction. This heat equals the enthalpy of reaction or tran-
sition, however, only, if the heat capacity is constant during the transformation.
Otherwise the enthalpy of transition is a function of temperature (cf. Sect. 4.2
and 6.1.6). As a consequence the peak area is something like an integral
mean value of the enthalpy change within the temperature regime of the peak.
To get the total H(T) curve and from that the enthalpy change, the Cp curve
should be calculated from the measured curve and then integrated (see Sects. 4.2
and 6.1.6).
As stated above, 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 spec-
ified, which allows the latent heat (enthalpy of a transition or reaction) to be
determined from a Cp curve without a baseline being constructed. The way of
proceeding will be demonstrated by the example of melting, which is a 1st order
thermodynamic transition (Richardson, 1993). Figure 5.5 shows schematically
the C p curve prior to, during and after transition. The heat capacities Cp,s (T) for
the solid prior to melting and Cp,\ (T) for the liquid after transition differ in ab-
solute values (at equal temperature) and also in their temperature dependence.
The kind of functional dependence of the heat capacity on the temperature (lin-
ear dependence, polynomials of higher order) is unimportant. Extrapolations
into temperature ranges that are not directly measurable must, however, be reli-
able. If this is guaranteed, the heat absorbed over a wide temperature interval
f
>.
~
'-'
0
0-
0
'-'
(P . (T)
~
0
Q)
.c
tempera ture ..
Fig. 5.5. The determination of thermodynamically valid values for the enthalpies of transi-
tion (according to Richardson, 1993). The limiting temperatures Tl and T2 must be chosen well
below or above the transition interval.
fJ.H(T) = Q - A - B with
Q: ~::::: :1 (horizontal dashed) A: ~ (right dashed) B: ~ (left dashed)
can be divided into the Cp portions and the isothermal enthalpy of fusion that is
of interest. The following is valid (see even Fig. 6.8).
11 H (T) = H[ (T) - Hs (T)
= (H[(T2 ) - Hs(Tj - (Hs(T) - Hs(Tj - (H[(T2 ) - H[(T)
T T2
= (H[(T2 ) - Hs (Td) - f Cp,s (T) dT - f Cp,[ (T) dT
Tj T
=Q-A-B
The three quantities Q, A and B are accessible by experiment, these are the areas
hatched differently in Fig. 5.5.
Enthalpies of reaction uniquely defined from the thermodynamic viewpoint
can be determined analogously; however, this determination presents some spe-
cial features (cf. Sect. 6.2) and Figs. 6.6 to 6.9.
From Cp curves, got at low heating and cooling rates to avoid irreversibility
(if necessary after previous desmearing of the measured curve), the entropy
change in temperature ranges in which no transitions take place can be deter-
mined with sufficient accuracy according to the following equation:
T C (T)
S(T) - S(To) = f - p - dT
To T
This is not allowed in the region of phase transitions because the peaks are al-
ways smeared and the plotted temperature is never the true equilibrium tem-
perature Teq of the sample. However if, in compliance with the above prescrip-
tion, Teq and t:.H(Teq ) are known, the reversibly exchanged entropy of transition
can easily be calculated from the quotient t:.H(Teq)/Teq. It is then possible to
calculate changes in the entropy and the free enthalpy for arbitrary tempera-
ture intervals. If a certain substance exists in several forms, information on its
thermodynamic stability as a function of temperature can be obtained.
5.4
Desmearing of the DSC Curve
The user of a DSC is usually interested in the heat flow rate into the sample to be
investigated or in the related thermodynamic potential function (depending
on the temperature), which is either the specific heat capacity cp(Tsample) or the
enthalpy change t:.H(Tsample)' The measured curve produced by the calorimeter
as a function of time, or a temperature proportional to it, is, even in the form
of a cp curve, 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 for, 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
function (t:.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 dis-
pensed with and what 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 or-
der endothermic phase transition. Accordingly, desmearing takes place in sever-
al steps:
a) Linear transformation of the abscissa (indicated temperature) in compliance
with the results of the calorimeter's temperature calibration, and transforma-
tion of the ordinate (indicated heat flow rate) in compliance with the results
of the calorimeter's heat flow rate calibration.
b) Determination and subtraction of the zeroline (sample crucibles empty).
c) Mathematical elimination (deconvolution) of the influence which heat trans-
port phenomena inside the DSC (heat flow rate relaxation) exert on the mea-
sured curve.
d) Proper conversion of the heat flow rate function into the real specific heat
capacity of the sample, i.e., (non-linear) variable transformation of the tem-
perature measured by the DSC into the true sample temperature.
e) Integration leading to the thermodynamic potential function (enthalpy).
The required steps will be described in closer detail in the sections below.
5.4 Desmearing of the DSC Curve 127
-
Fig.S.6a-e. 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 Zeroline
(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 of the
enthalpy function.
<Pm heat flow rate (measured curve), <Ps heat flow rate (sample run), <Po zeroline heat flow rate,
<Ptrue heat flow rate after calibration, 6. <Ptrue sample heat flow rate ( <Ps minus <Po), <Pr reaction
heat flow rate into the sample, Tm temperature (measured), Ttrue temperature after calibration,
Ttr , transition temperature, Ts sample temperature, Cp heat capacity of the sample, 6.H enthalpy
change
5.4.1
Correction of the Temperature and Heat Flow Rate Indicated
This procedure has already been described in Sect. 4.3 and 4.4, 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 {3,
the following is valid:
Ttrue = Tm + I1Teorr{fJ) (S.1)

I1Teorr {f3) = I1Teorr {{3= 0) -I1T{f3)
11 Teorr ({3 =0) results from the calibration measurements, and 11 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 tem-
perature axis scaling of the output of the measured function, a shift which is dif-
ferent 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. (S.l) takes only influences from the instrument into account. Samples of great
thickness possibly require an additional correction which also considers the tem-
perature proflle inside the sample; this correction is described below (Sect. 5.4.4).
Note: The recently introduced turret -type DSC (TA Instruments, see Sect. 2.1.2)
takes the heat capacity and thermal conductivity of the measuring system into
account and corrects for the thermal lag with the so-called Tzero technique. As
a result either the temperature of the sample sensor or the temperature of the
sample crucible is calculated and used as temperature axis for the measured
curve. These corrections are normally sufficient for common measurements
and additional corrections of the temperature scale are not necessary if careful
calibration has been done.
The following equations are valid for the correction of the displayed heat flow
rate (cf. Chapter 3 and Sect. 4.4.1):
IPtrue = K~ IPm or IPtrue = - K . 11 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 con-
sistently; it results from the differing definitions of the sign of I1Tm (measured
temperature difference) and IPm. A linear transformation of the IP-axis of the
graph of the measured curve, 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
IP-axis being retained:
5.4.2
Subtraction of the Zeroline
It is not possible to produce a perfectly symmetrical DSC in which the differen-

tial signal between empty sample system and empty reference sample system is
exactly zero over the entire temperature range. This applies also to the measured
curve with the sample crucibles empty (zeroline).
To obtain the true heat flow rate into the sample (for example, to calculate the
sample's heat capacity), the zeroline must be separately determined by a second
measurement and subtracted from the curve actually measured. For the correct
determination of the zeroline see Sects. 5.3.1 and 6.1.1.
The correction of the instrument asymmetry presents the problem that the
two measurements can be carried out only one after the other; it must therefore
be ensured that the DSC's symmetry properties have not changed in the mean-
time. This is a strict condition which means that high requirements must be met
by the calorimeter (cf. Sect. 7.2). In addition, it is indispensable that the DSC
is opened between the two measurements and the sample crucible changed or
moved. This alone can change the symmetry properties, which results in the
zeroline 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 presented below.
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 mea-
surement and the sample, the measurement signal is always a smeared represen-
tation of the sample reaction heat flow rate. The kind and degree of smearing
depends basically on the thermal inertia of the system, in other words it depends
only on the DSC construction.
Good DSCs retain linearity and allow the theory of linear response to be ap-
plied 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 DSC can be described by normal differential
equations. The mathematical interrelation has already been dealt with in Chap-
ter 3. The desired function cPr(t) can therefore be calculated from the measured
function cPm (or AT(t) which is linked with cPm by the equation cPm = -k'AT(t,
using Eq. (3.6) in the simplified form:
dcPm(t) d 2 cPm(t)
cPr(t)=cPm(t)+al +a2 2 (5.2)
dt dt
Accordingly, the desired function is a simple sum of terms which include the
measured function and its derivatives. The coefficients al and a2 include the
time constants Tl and T2 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 cPm(t). It is,
of course, also possible to calculate it from the (stored) function cPm (t) after the
measurement has been concluded.
Although the problem is easily solved (mathematically or by electronic
means), the determination of the proper coefficients al and a2 is not so easy in
practice. The time constants Tl and T2 [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 al and a2 may be more complex than the approximate calculations
in Sect. 3.1 possibly predict. There is no other way than to adjust these two co-
efficients (by trial and error or by suitable fit mathematics) until the results of
the desmearing procedure (online or calculated) coincide with the original heat
flow rate cPr inside the sample. The best method to simulate an event is to switch
on a constant heat flow rate (for a certain time) with the aid of a built-in electric
heater.
Note: The recently introduced turret-type DSC (TA Instruments, Sect. 2.1.2)
takes the heat capacity and thermal conductivity of the measuring system into
account and corrects for it internally with the so-called advanced Tzero tech-
nique. As a result the measured curve is automatically desmeared and the soft-
ware produces a curve of the heat flow rate into the sample crucible as a function
of sample crucible temperature. This correction includes the three steps of
desmearing described so far. This method takes a careful calibration of the DSC
and unchanged measuring conditions as granted to give reliable results.
Numerical Methods
If the behavior of the DSC cannot be described by simple differential equa-

tions within the framework of the required accuracy (e. g., in the case of
power compensation 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 func-
tions from each individual event (Fig. 5.7). Another condition is that all mea-
sured functions (curves) of various pulse-like events should have the same shape,
in other words all these measured functions divided by their peak area (nor-
malizing) must yield the same function, the so-called apparatus function a (t)
T, T2
temperature
r
b
tempera ture ~
Fig. 5.7 a, b. 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, tPr heat flow rate devel-
oped in the sample, tPm measured heat flow rate (measured signal), Q heat of the event, a(t)
apparatus function of the DSC
(or even called Green's function). If these conditions are fulfilled, the following
is valid:
tl>m (t) = c f tl>r (t') . a (t - t') dt' == tl>r (t) * a (t ) (5.3)
This defines the so-called convolution product of two functions, a integral equa-
tion which often is abbreviated as "convolution product"with a star as operator.
The equation is valid for all DSCs which work in the above-described linear
manner, irrespective of whether a certain approximate formula [such as Eq. (3.6)]
is explicitly 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 (e.g., Sn).
The seamy side of this desmearing method, also called deconvolution 1, is the
rather ambitious mathematics required to solve the integral equation (5.3) for
the function of interest 4>r(t). There are essentially two methods, the Fourier
transform and the recursion method. Both require numerical calculations with a
computer.
The Fourier transform represents an integral operation:
{S (f(x = nl2 . f fey) . exp(ixy)dy

Applied to the convolution integral [Eq. (5.3)], it results in the convolution
theorem (see textbooks of mathematics):
Thus the convolution product turns into an ordinary product which can be
solved for
The desired function is obtained by an inverse Fourier transform:
This method can be applied in all cases. Today, the Fourier transform is com-
monly included in many mathematical software packages and therefore easily
available. The drawbacks of this procedure lie in its laborious course and ab-
stract nature, since the calculations are performed in Fourier space. Those who
lack experience in numerical Fourier transforms are advised to study some "pit-
falls" 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 (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:
4>r,O(t) = 4>m(t)
4>r,n(t) = 4>r,(n-I)(t) + (4)m(t) - aCt) * 4>r,(n-I)(t (5.4)
The deviation between the "reconvoluted", still inaccurate synthetic function

4>m,n (t) = a (t) * 4>r,(n-I) (t) and the measured function 4>m (t) is used additively
for the recursive correction of the approximation.
1 Deconvolution means to solve the convolution integral equation, it is a special case of

desmearing.
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. However, in the case of the
"smooth" curves commonly encountered in calorimetry, the procedure con-
verges quickly and without problems.
It should be noted that every desmearing 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 whose width is in the same
order of magnitude as the time constant of the DSC and if the exact time course
of the heat flow rate is needed (e. g., for kinetic investigations).
Both numerical desmearing procedures can be applied only after the mea-
surements have been concluded (offline). For the case of a discrete (sampled)
function, the convolution integral [Eq. (5.3)] can, however, be represented as a
linear system of equations.
If>m(ti) = c M L (If>r(tk)' a(ti-k+l)) (i = 1 ... n)

k=l
which can be resolved for If>r (tj):
(5.5)
As can be seen, each value of the desired function can be calculated from the
measured points If>m (tk) (k ~ i) measured before this moment and the (known)
discrete apparatus function a (tJ (i = 1... n). Even with this numerical deconvo-
lution it is then principally possible to calculate the desired function online dur-
ing the measurement and to put it out.
The problem is that the measured values (and the apparatus function) are not
quite 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 transforma-
tion and the noise fraction is, therefore, relatively high; the initial values of the
apparatus function are usually very small which - according to Eq. (5.5) - leads
to a very noisy signallf>r (tj), which in turn results in numerical instabilities which
may lead to a "run-time error" of the computer. For these reasons, online decon-
volution according to Eq. (5.5) is usually not suitable without special precautions
concerning the initial peak region in question. Of course, such calculations can
be performed offline 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 dis-
cussed.
Peak Shape During Melting

For a 1st order endothermic phase transition of a pure sample, desmearing as it
has been carried out so far furnishes, for example, a deconvoluted measured
curve as shown in Fig. 5.8a. When recorded as a cp function, the 1st order phase
transition has, however, the shape according to Fig. 5.8 b. Both functions are ob-
viously not alike. To be able to calculate the real cp function from the measured
function, the reasons must be known quantitatively which have led to the par-
ticular function according to Fig. 5.8 a. In the case of an endothermic phase tran-
sition, the sample temperature is constant for the duration of the phase transi-
tion; as the sample crucible is heated linearly, the heat flow rate into the sample
also increases linearly:
tPr = L . fl T = L . (Tm - Ts)
dtPr = L . (dTm _ dTs)

dt dt '----,-'
'----,-'
dt
"" f3 = 0
dtPr dTm
-=L- (5.6)
dt dt
d tPr dt d tPr
--=-=L
dt dTm dTm
(L apparent thermal conductance to the sample, f3heating rate).
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 Tl to
T2 (Fig. 5.8) and dividing the ordinate by the mass and the heating rate. As a
result, the triangular peak transforms to a 6-peak with the weighting factor of
the phase transition heat fltrsH, and the temperature (abscissa) becomes the
sample temperature.
For endothermic transitions of, for instance, non-pure substances one may
start from the fact that the rather low heat flow rate at the beginning of the tran-
sition 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 (TJimit in Fig. 5.9a) it reaches the heat flow rate limit of the apparatus
and the sample temperature falls behind the measured temperature. The limit-
ing 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 of the tran-
sition of the sample (Fig. 5.9 b). There is a non-linear connection between these
two temperatures which generally cannot be specified in detail.
In the case of impure substances which can be described with the aid of ther-
modynamics of eutectic multi-component systems, the theoretical shape of the
t
Fig. 5.8 a, b. Advanced
desmearing in the case of the
melting of a pure substance.
a Deconvoluted measured
curve, b Corresponding
cp(T) function.
~r true heat flow rate into
the sample, Tm temperature cj> r
measured, Ts sample tem-
perature, Ttr transition
temperature, L apparent
thermal conductance be-
tween sample crucible and a
..
sample, cp specific heat Tl T2
capacity of the sample
Tm
l~
t
(p
function shown in Fig. 5.9b is known (see Sect. 6.9). This allows the true melting
curve to be determined from the measured one by calculating along the theo-
retical curve and comparing the areas which must be equal according to the law
of the conservation of energy. This calculation may be called desmearing as well,
but it is far away from deconvolution due to the theory of linear response, which
in the melting region, of course, is not valid any more.
Thermal Lag Inside the Sample

Another problem is the smearing of the measured curve due to the temperature
profile developing in the sample itself. It is evident that the peak assigned to a tran-
sition in a thin sample differs substantially from that in a thick sample (cf. Sect.
3.1}.1t can be shown by the method of Laplace transformation that the tempera-
ture profile in the sample has a parabolic shape and that the mean temperature (T)
of the sample can be calculated according to the following formula (Hoff, 1991):
d2
(T) = llb - Cp P . P. - (5.7)
3..\
Fig. 5.9a, b. Advanced

desmearing in the case
of a melting of an impure
substance.
a Deconvoluted measured
curve, dashed line: hypo-
tenuse of a right-angled
1
triangle with the slope L, /1
/ 1
b Corresponding cp(T) func- / 1
/ 1
tion (the temperature scale is r--------=------f-----+------~------
not correlated to that of a). I 1
1 1
q,r true heat flow rate into I
1
1
1
the sample, T m temperature a
measured, Ts sample temper-
ature, Ttrs transition temper-
ature, Te extrapolated peak
onset temperature, L appar-
ent thermal conductance,
cp specific heat capacity of
the sample
Tend 11,,1
Tlb being the temperature at the lower (heated) boundary of the sample (Cp heat
capacity, p density, A thermal conductivity, f3 heating rate, d sample thickness).
The second term on the right-hand side of Eq. (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. (5.3)] of both parts, as these
events are connected in series (Hahne, Schawe, 1993). This seems to complicate
the desmearing procedure, but fortunately linear response theory offers the pos-
sibility to determine the total apparatus function (including sample properties)
from the "step-response". This can be done by analyzing the switch-on and
switch-off behavior of the scanning mode.
Step Response Analysis

The starting and ending of a DSC run always implies switching the heating rate
from zero to a constant value and vice versa. As a result the real heat flow rate
fClJ
2~
C
L..
-
:J:
0
~
c
ClJ
.<=
time ~
Fig.5.10. Curve measured with the heating rate switched from zero to f3 at to.
1 measured curve, 2 theoretical heat flow rate into the sample
O.B
Sl
I
0.6
0.4
C;
0.2
0.0
0 10 20 30 40 s 60
time ~
Fig.5.11. Apparatus function a (t) from step response arising from

1 switch-on, 2 switch-off
into the sample should change in a step-like manner, whereas the measured heat
flow rate rises with a certain delay (Fig. 5.1O). This step response function covers
both the heat transport and the sample behavior. By differentiation the pulse re-
sponse function can be derived from this function (Fig. 5.11), which is the appa-
ratus function (Green's function) of the total system. Using this apparatus func-
tion 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
of samples with poor thermal conductivity. This may be used as a proof of the
influence which the thermal lag and the temperature profile inside the sample
2.2
J
glf
1.8
f 1.6
Co
LJ
1.4
1.2
350 360 370 380 390 K 410
temperature ..
Fig.5.12. Heat capacity measurement showing a glass transition.
1 measured curve, 2 desmeared curve (10 mg polystyrene, heating rate 20 K min-I)
exert on the measured curve. An advantage of this method is the easy measure-
ment 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. 6.1). Thus the apparatus function in question is always
on hand.
Note: The newly introduced step-scan mode of operation (cf. Sect. 2.4.2)
makes it possible to get the step-response function and thus the Green's function
of the total system at every moment of the measurement - even during transi-
tions occurring in the sample - this offers the possibility to determine pure sam-
ple properties with all apparatus influences corrected for. Such a correction is in
particular fruitful for determination of the complex heat capacity (see Sect. 5.5.2)
during transitions.
Simple Thermal Lag Correction
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 smeared, but also the baseline and the measured curve when steady state cp
measurements are performed. Figure 5.13 may serve to explain this"thermallag"
effect: There is in principle a difference between the measured temperature of the
sensor and the true temperature of the sample during a scanning run (Sect.
S.4.1).1t increases with increasing heat flow rate into the sample due to the pro-
portionality of heat flow rate and associated temperature gradient. The heat flow
rate for its part depends on the heat capacity of the sample and on the heating
heating rate {3
----~----~--O>
t
~/
/~
t
~/
CPtr" //
~/
j
,...:::-
-e. (cpm- CPtru.) = tl(3)
~/
............................................... :/.'< ......... .
cu
..2:!. ~
c -e.
......... . . .. . \
:::::::::.. ::::::::::::::::::::::.::::>Z<:::::.;: :.
'-
~
. "
i
0 /'
~~~~~...
.....c
~
OJ
.<=
( (T5 -Tm)~t({3)
: .. . 0>
I
~ag
T
Fig.5.13. Difference between true (dashed line) and measured (solid lines) heat flow rate (or
heat capacity, respectively) curves.
T program temperature, Tst starting temperature, Ts sample temperature, Tm measured tem-
perature, 4'>true true sample heat flow rate, 4'>m measured heat flow rate
rate. As a consequence, the measured curves shift more and more to the right side
in Fig. 5.13 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 meas-
ured 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
temperature displayed, the heat flow rate (and the heat capacity) displayed are
also too low - compared with the true one - due to the increase in the sample
heat capacity with temperature (see Fig. 5.13).
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 is, however, to be taken into account and, if nec-
essary, it should be corrected by applying the desmearing procedures referred to
above. The thermal lag 6T itself can be approximately determined (Richardson,
Savill, 1975b and Vallebona, 1979) from the heat 6Q, 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 func-
tions in Fig. 5.10, cf. also Fig. 6.2):
6T= 6Q
Cp,s
(6T: thermal lag in K; A: area in J; Cp,s: heat capacity of sample in J K- 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 true heat capacity-
temperature function of the sample, Cp (Ts). An integration must be carried out
to obtain the thermodynamic potential function change:
T2
t:.H(T) = f Cp(T) dT
TJ
Usually, the pressure p is constant in DSC experiments and the enthalpy change
t:.H(T) is obtained from Cp; (see Sect. 5.3.1).
If the measurement would be carried out at constant volume, in analogy to
Cp , firstly Cv ( Ts) is obtained and from this, by integration, the (internal) energy
t:.U(T) as the respective thermodynamic potential function. (This is not very
important regarding DSC measurements as the realization of a constant volume
is very difficult.)
The time and effort to be spent on desmearing seems to be large and the pro-
cedures complex. It should, however, be noted that the individual steps of the
desmearing procedure described in this section must be carried out only if it is
necessary because of the accuracy requirement. In other words, the increase of
accuracy after desmearing should be larger than the uncertainty of the mea-
surement and the increased uncertainty caused by the desmearing procedure it-
self. For most applications, both deconvolution of instrument influences and the
advanced desmearing procedures can be dispensed with; in peak area evalua-
tions they are absolutely unnecessary. They are, however, required when, for ex-
ample, kinetic evaluations or a precise phase transition analysis have to be done.
5.5
TMDSC Curves
The temperature-modulated mode of operation (cf. Sect. 2.4.2) offers the possi-
bility to separate different processes from one another. A special evaluation of the
(modulated) heat flow rate curve results in different additional functions (curves)
which, depending on the sample and the processes of investigation manifest
themselves in a different way. Of course, a proper calibration and, if needed,
zero line correction is taken for granted before further evaluation is performed.
With a sinusoidal temperature modulation [Eq. (3.9) 1the heat flow rate curve
has normally the following form [see Sect. 3.3.2 and Eq. (3.17)1:
5.5 TMDSC Curves 141
This is the sum of two parts: the first two terms on the right side describe the
normal DSC curve got at a heating rate f30 [ef. Eq. (3.12)], the third and fourth
term form the periodic part of the measured curve. The first part is obtained by
"gliding integration" (over a period tp = 2n/w, see Sect. 3.3.2) which lets the
periodic part get always zero and leaves the first two terms unchanged [apart
from that this function is averaged (smoothed) within the integration interval].
The curve obtained in this way is called the underlying heat flow rate lPu, it is
almost the same as the curve got in the DSC with the temperature modulation
switched off (common mode of operation).
Subtracting the underlying curve from the measured curve yields the peri-
odic part of the measured heat flow rate:
_ alPexo(Tu,t)
IP(T, t) = IP(Tu, t) - lPu(T, t) = Cp TA W cos (wt) + a T TA . sin (wt)
= IPA cos(wt + c5) (5.9)
This is in reality a periodical (sinusoidal) fluctuating function with an ampli-

tude IPA which on the one hand is influenced by the heat capacity of the sample
(lst term on the right side) and on the other hand by the reaction or transition,
if the respective latent heat flow rate depends on temperature (2nd term on the
right side). The latter causes even a phase shift 6. The periodic part of the heat
flow rate function forms the basis for further evaluations of TMDSC. By Fourier
analysis, which normally is included in the software of TMDSCs, the amplitude
and phase angle of the modulated part of the heat flow rate function can be de-
termined. Further evaluation results in the determination of an "apparent heat
capacity", which in the case of processes occuring in the sample can be different
from the common (vibrational) heat capacity. Depending on the evaluation (and
mathematical) method several heat capacity curves, containing different infor-
mation of the sample, can be obtained.
5.5.1
Reversing and Non-Reversing Heat Capacity
These curves are the result of the mostly used evaluation procedure. The basis is
the following: if there are no transitions or reactions in the sample, the ampli-
tude of the modulated heat flow rate is proportional to the heat capacity of the
sample (see Sect. 3.3.2, case 1). Formally, even in the case of processes occurring
in the sample, an apparent heat capacity can be defined:
IPA (T)
cp,revo (T) = m TAW (5.10)
This function (curve) is traditionally called the reversing heat capacity. Outside
the region of transitions or reactions it corresponds to the normal (vibrational)
heat capacity of the sample, within the region of processes it contains compo-
nents of the process in question (see Eq. 5.9).
From the reversing heat capacity the reversing heat flow rate can be calcu-
lated:
IPrev.(T) = Cp,rev.(T) m Po
Subtracting that from the underlying heat flow rate, got via gliding integration
of the measured heat flow rate function, yields the so-called non-reversing heat
flow rate:
If the temperature dependence of IPex. [the second term on the right side of
Eq. (5.9)] can be neglected we get [with Eq. (5.8)]:
(5.11)
the heat flow rate curve of the process (with a certain latent heat). But this is
only true in the regions where Cp from the underlying curve and the reversing
Cp really coincide. It must be emphasized that this is only the case in regions of
pure (real-valued) vibrational heat capacity, in other words without contribu-
tions from possible complex heat capacity components. Normally this is true for
chemical reactions and the crystallization of polymers, but not in the region of
glass transition or melting. Outside of these critical regions the non-reversing
heat flow rate curve reproduces the heat flow rate curve of a common DSC, but
with the baseline (see Sect. 5.3.2) already subtracted.
The heat flow rate curves can easily be normalized and presented as heat ca-
pacity curves if needed:
IPnon-rev. (T)
Cp,non-rev. ( T ) = -a- -
m /"0
and the following is valid:
Cp,non-rev. (T) = Cp,ll (T) - Cp,rev. (T) (5.12)
This equation is formally right, but it must be emphasized again, that these quan-
tities are apparent heat capacities which are not really defined in a thermody-
namic sense in the region of transitions and reactions. All three quantities differ
from the respective heat flow rates only by a constant factor. In Fig. 5.14 the un-
derlying (even called "total"), the reversing and the non-reversing heat capacity
curves calculated from a TMDSC measurement of polyethylene terephthalate
(PET) are presented as an example.
Within the scope of this evaluation approach the phase signal is totally ignored,
all these curves are real-valued and describe either the real vibrational heat ca-
pacity, or, for reactions or transitions, the magnitude (absolute value) of an appar-
ent complex heat capacity. To obtain the complex heat capacity we have to pursue
a different path. .
5.5 TMDSC Curves 143
f: 2
1
o
""
~ -1
~ -2
"C;
~
V)
-3
60 100 140 180 220 O( 260

temperature ~
Fig.5.14. Reversing (upper), underlying (medium) and non-reversing (lower) heat capacity
curves from a TMDSC measurement of PET (normalized to specific heat capacity units; m: 5 mg,
Po: 2 K min-1,f: 28 mHz., TA: 1 K; according to Schawe, 1995 a)
5.5.2
Complex Heat Capacity
Another approach, which is valid within the limits of linear response, calculates
an apparent complex heat capacity from the measured heat flow rate curve.
Again this evaluation is only valid if the DSC is properly calibrated, and if the
calibration of magnitude and phase signals is carefully done as well. The latter
is not easy but, however, necessary to come to reliable results. If this is success-
ful, the knowledge of the apparent complex heat capacity will be very helpful in
understanding of time-dependent processes taking place in the sample.
To get the complex heat capacity we start from the same evaluation procedure
as before: by gliding integration we determine the underlying heat flow rate
curve, subtract that from the measured heat flow rate and get the periodic part.
By Fourier analysis or another mathematical treatment (see, e.g., Hohne, 1997)
the amplitude and phase angle are determined. From the heat flow rate amplitude
the magnitude (absolute value) of the apparent complex heat capacity is deter-
mined [see Eqs. (3.15) and (5.1O)]:
Icp(T) I = 4iA (T} (5.13)

mTAw
This function is identical to the reversing heat capacity (see Sect. 5.5.1), it con-
tains possible contributions from reactions or transitions (see Sect 3.3.2). To-
gether with the properly corrected phase signal 8 (see Sect. 4.7.2), the apparent
complex heat capacity can be defined:
cp = IcpI . exp (i . 8) = IcpI . cos (8) + i . IcpI . sin (8) (5.14)

From this follows the real part and the imaginary part of the complex heat ca-
pacity (the phase shift is assumed to be negative!):
Re(cp) = p=
C ICpl' cos (6) and Im(cp) = cp = ICpl' sin (6) (5.15)
As magnitude and phase always change with temperature, even the complex
heat capacity and its real and imaginary part are functions of temperature. In
Fig. 5.15 the result from such an evaluation of a TMDSC measurement of PET is
presented. From comparing these curves with those of Fig. 5.14, it follows that
there are huge differences, in particular in the region where transitions occur.
Only the underlying curve coincides with the magnitude curve as expected, be-
cause they are calculated in a identical way.
In the region of transitions or reactions the complex heat capacity depends
on the modulation frequency too (see Fig. 6,43). To gain information on time de-
pendent processes, different measurements at different frequencies (periods)
must be done. Performing non-sinusoidal (e.g., sawtooth) modulation enables
one to get the apparent complex heat capacity not only at the frequency of the
modulation but also at the frequencies of the higher harmonics from the same
run (for details see Merzlyakov, Schick, 2001 a).
The most elegant method, however, is to analyze the heat flow rate response
on a temperature step. After Fourier transform and proper correction using the
transfer function (see Sect. 4.7), we obtain in this way the complex heat capacity
cft (w) as a continuous function of frequency at the temperature where the step
was performed. This function contains all information concerning the time de-
pendent processes, after an inverse Fourier transform one gets the time depend-
ent heat capacity function at the temperature in question. However, this elegant
method is not available with the commercial evaluation software and therefore
}/lg:1
>-
-=
w
'"
Cl..
8 2
.....
'"
OJ
.r;
w
~
w
0
OJ
Cl..
Vl
-2
50 100 150 200 250 C 300
temperature _
Fig.5.15. Real part (dashed), magnitude (solid) and imaginary part (dotted) of complex
specific heat capacity calculated from a TMDSC measurement of PET (m: 5 mg, /30: 2 K min-I,
f: 28 mHz., TA: 1 K; according to Schawe, Hohne 1996)
5.6 Interpretation and Presentation of Results 145
not accessible to the normal user and we refrain from presenting the details of
this special evaluation here. The advanced reader is referred to the original
literature (Merzlyakov, Schick, 2001 b).
5.5.3
Curves from Step-Scan Evaluation
The recently, for power-compensated DSCs, introduced step-scan mode of oper-

ation (see Sect. 2.4.2; Cassel, 2000) orginates originally two curves called the spe-
cific heat capacity and the isokinetic baseline. The former is simply the interpo-
lated curve through the discrete heat capacity values got from the areas A of the
peaks caused by the discrete temperature steps t:..T (see Sect. 6.1.4):
A
cp = - - -
mt:..T
It contains the vibrational heat capacity as well as contributions of reactions and

transitions (as with sinusoidal modulation, Sect. 5.5.1), in other words, it is an
apparent heat capacity like the reversing Cp Using the extrapolated isokinetic
line (see below) as baseline makes a subtraction of a zeroline normally unnec-
essary. A careful peak area calibration is, of course, needed anyway.
The isokinetic baseline, on the other hand, is the interpolated line through the
end point of the isotherms after each temperature step. This equals the (isother-
mal) zeroline of the DSC (caused by asymmetry) but contains all the reaction
and transition heat flow rates produced by the sample even in isothermal mode.
Within the region of processes, the shape of this contours curve is highly influ-
enced by the period of time chosen for the equilibration after the temperature
step. The longer the isothermal period, the lower is the end value of the heat flow
rate caused by reactions or transitions which usually follow a time law. This
curve must be evaluated with due care and attention to get reliable results con-
cerning the reaction or transition heat.
The advanced step-response analysis presented in Sect. 5.5.2 can be carried
out as well. As a result we get a set of complex heat capacities cp (w), one for
every temperature step, which contains all necessary information about time
dependent processes occurring in the sample during the run. An indisputable
evaluation and interpretation is, however, not an easy task and the investigations
on this topic are still not completed.
5.6
Interpretation and Presentation of Results
For certain tasks (e. g., quality control of goods and materials received) compar-
ison of the measured curve with reference curves enables an identification of the
sample substance on the basis of a yes/no decision. This generally concludes in-
terpretation in these cases.
In the normal case, however, interpretation is preceded by an 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 a reliable evaluation

are the following:
- desmearing, if necessary (cf. Sect. 5.4),
- determination of the peak area or partial areas after construction of the base-
line (cf. Sect. 5.3.2 and 6.3),
- determination of Cp changes (cf. Sect. 6.4),
- determination of the uncertainties (cf. Sect. 7.3).
The data obtained from the evaluation form the basis of the interpretation. The
presentation of non-interpreted measured curves is poorly informative for the
non-specialist. A reliable interpretation can possibly be made only after several
experiments with, for instance,
- 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 DSC experiments alone is not possible. Other methods must then be
applied in addition. Only if the results obtained by all test methods can be inter-
preted in the same way is it probable that the interpretation is correct. For exam-
ple, in the case of kinetic investigations in inhomogeneous phases, one should
always be aware of the fact that the calculated values are affected by considerable
uncertainties.
When DSC results are presented, the following should be stated:
- sample characteristics (mass, purity, structure, ... ),
- instrument characteristics (type of DSC),
- test conditions (heating rate, purge gas),
- the curve originally measured (and, if used, the desmeared curve),
- details of the calibration procedures (materials and their characteristic data),
- details of the evaluation of the measured curve (specification of characteris-
tic temperatures, construction of baseline, peak integration),
- data obtained from the measured curve and the uncertainties by which they
are affected,
- basic formulas and calculation procedures which are used (e. g., for glass
temperature, or kinetics),
- 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 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 follow-
ing text. Sometimes the required information can be obtained from only a qual-
itative evaluation of the DSC curve. But most of the examples described in this
section demand precise measurements and critical, very often special, evalua-
tion 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 mea-
sured zeroline (see Sect. 5.1) has to be subtracted from the measured curve
before evaluation. In every case, the relationship between uncertainties in
the measurements and the quantities to be determined must always be borne
in mind.
6.1
Measurement of the Heat Capacity
The heat capacity is one of the most important material properties. There is
no other method which supplies the temperature dependent heat capacity as
quickly and over such a large temperature range with - for most purposes -
sufficient accuracy. A knowledge of the heat capacity of a material as a function
of temperature is the basis for determination of any thermodynamic quantity
(cf. Sect. 5.3.1).
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 pressure. The situation is somewhat more complicated if
one uses hermetically sealed crucibles or the special crucibles which are avail-
able 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 com-
posed of air or of gaseous reaction products. Strictly speaking neither Cp nor Cv
are 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.
148 6 Applications of Differential Scanning Calorimetry
In the following of this 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 determi-
nation (under steady-state conditions)
is valid both for heat flux calorimeters and for power compensating DSCs.
As, normally, the true heating rates of the sample and the reference material
are not accessible by experiment, they must be replaced by the average heating
rate p. If the heat capacity CR is known, Cs can be determined easily and quick-
ly from the measured differential heat flow rate d 4>SR' Several variants of the
experimental procedure are known, four widely used techniques will be dis-
cussed.
6.1.1
uClassicaluThree-Step Procedure
The procedure is illustrated in Fig. 6.1. The temperature-time curve during an

experiment 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 zeroline 4>0(1'), using empty cru-
cibles (of equal weight) in the sample and the reference sides. The tempera-
ture program should only be started when the isothermal heat flow rate at the
starting temperature Tst has been constant for at least one minute. If the DSC
is computer controlled, this can easily be automated by checking the differ-
ences between the current average value of the heat flow rate and that one
minute before with allowance for a predetermined noise level. The scanning
region between Tst and Tend can be 50 to 150 K in modern calorimeters. At the
isothermal end temperature Tend the above computer check must be repeat-
ed. For the evaluation procedure all three regions of the curve are needed. The
zero line reflects the (inevitable) asymmetry of the DSC.
2. A calibration substance (Ref) of known heat capacity CRef is placed into the
sample crucible (or into a crucible of same type and mass), whereas nothing
is changed on the reference side. Using the same experimental procedure as
for the zeroline, the following is valid:
K<t>(T) is a temperature dependent calibration factor (cf. Sect. 4.4.1).

3. The calibration substance in crucible S is replaced by the sample. In analogy
to the equation above we get:
6.1 Measurements of the Heat Capacity 149
12
sample
iso-
Iher~ (ribralion subslance
mal
al isolhermal
Tst 01 Tend
scanning mode
:a
.e-'"
'"
.e- Ie;
"-
o
...co
.c
I
.e-
.~
~
emply pan
:1
a 0
0 1st 4 12 min 16
lime ..
1
~
::J
...~
C1.
T,t
E
2!
b
lsi
time ..
Fig.6.1. The conventional three-step technique for the determination of the heat capacity.
a schematic course of measurement, b the temperature change during the run.
Ts! start temperature at time tst> Tend end temperature at time tend, t1>s, t1>Ref, t1>o heat flow rates
into sample, calibration substance and empty crucible, respectively,~ t1>SR differential heat flow
rate between sample and reference crucible
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. 6.1:
The calibration factor K~(T) need not, therefore, be known explicitly. If the
condition mscs::::: 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 crucibles of different masses if allowance is made for the different thermal
responses according to:
The specific heat capacity of the crucible material (second term on right side)
serves as a correction only. The values for common crucible materials (e.g., AI}
are known with sufficient accuracy. From an error estimation (see Sect. 7.3) one
can find that omitting this pan correction would result in an error < 10/0, if
the masses of all (AI) crucibles are selected to differ by less than 0.03 mg (at a
sample mass> 10 mg and a specific heat capacity >0.5 Jg-l K- 1).
Sources of error:
Ideal and real conditions during the recording of the zeroline and measured
curve of the sample are compared in Fig. 6.2. Three differences are obvious:
1. 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.
2. The measured heat flow rate (with zeroline subtracted) may be smaller than
the ideal (theoretical) one.
3. The isothermal levels at tst and tend differ from each other and from run to run
(and may often have non-zero values).
These discrepancies result from the finite thermal conductivity of the path be-
tween temperature sensor and sample and from the limited thermal conductivi-
ty of the sample itself (cf. Sect. 5.4). The sample operates both as a heat capacity
and as a heat resistance with respect to the thermal surroundings (PoeBnecker,
1990). The signal is therefore a summation of the heat flow stored in the sample
and that which passes through it and to the surroundings (heat leak). To be pre-
cise, it always appears, of course, as the differential heat flow rate between sample
and reference sides.
In the following the causes of the three above-mentioned deviations from
ideal behavior are considered in detail and possibilities for their correction are
given.
1. The smearing (caused by thermal inertia) of the measured heat flow rate
curve during the beginning of the scanning region, before steady state is
reached, reduces the temperature range over which calculations are valid. The
initial unusable temperature range can be estimated by !1T= 5 to 10 times
f3. reff The effective time constant reff results from a coupling of the time con-
r---
f I
I
I
~
...
c I
~
0
I
;:;:: I
c
C1J
I
I
c
oJ::
B
I
I
L ______ _
time ..
Fig.6.2. Idealized (dashed line) and real conditions (solid line) during a heat capacity mea-
surement.
Curve section AC: delay function of the sample due to the restricted heat transfer between
sample and sensor, hatched area ABC: the product of thermal lag r5T and heat capacity of the
sample
stants for sample and apparatus and that of the heat transfer between sup-
port, crucible and sample. As a rule the influence of the apparatus is predom-
inant. The time constants of modern DSCs may vary from 2 to lOs. For thicker
samples with poor thermal conductivity (e. g., polymers) the influence of the
sample may dominate reff.
2. As discussed in Sect. 5.4.4, the sample temperature is always lower (higher)
than the program temperature during heating (cooling) and the measured
heat flow rate <Pm differs always from the true value <Plr Assuming the worst
conditions (large samples, high heating or cooling rates, large heat capacity,
poor thermal contact between crucible and sample holder), the difference be-
tween both temperatures may be more than 10 K. This temperature error cST
(the thermal lag) can be estimated (see. Sect. 5.4.4) from the (delayed) heat
cSQ, which is proportional to the area ABC in Fig. 6.2. 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 gradi-
ent 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 of Ni (sample mass ca. 250 mg) differed by 10 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 mea-
sured heat flow rates are influenced by the thermal conductivity of the sample
and by the heat transfer resistance between sample and sample holder. The
heat transfer resistance, and thus the thermal lag, 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 capacity and a thermal conductivity similar to that of the
sample. But thermal conductivities of common calibration substances fall in
the following order (values in W cm- 1 K- 1): Cu (4.01) > sapphire disk (0.34) >
organic materials (0.05), which limits the use. The best general method for the
correction of all effects due to finite thermal conductivities is to use the spe-
cial desmearing procedure described by Schawe, Schick, 1991 (cf. Sect. 5.4.4).
3. The isothermal levels at TSI and Tend (resp. tSI and tend, cf. Fig. 6.1b) for zero-
line, calibration run and measurement differ from each other by amounts
which depend on the type of calorimeter, TSI and Tend and the temperature
interval in between as well as laboratory conditions. The offset of the isother-
mal levels must be corrected to a common level (normally zero) before the
heat capacities are calculated. The correction is only meaningful, if almost
comparable conditions for the total heat conduction path can be assumed for
the three successive runs (zeroline, 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 differ-
ences in the offsets of the isothermals should always be rejected. For precise
measurements, as a rule, the heat flow rates of the isothermals at TSI 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 that the change of
the isothermal heat flow rates with temperature can be approximated by a
straight line cPiso (T) (but see Sect. 6.1.3 for the limits, temperature intervals
have to be sufficiently small), the offset correction is very simple. Figure 6.3
demonstrates the procedure. With cPiso,sl and cPiso, end the heat flow rates of the
initial isothermal and the final isothermal, the following is valid:
The corrected experimental heat flow rates cPcorr(t) are then obtained by sub-
tracting the above correction term:
which should be used for the calculation of the heat capacity then.
12
w
expo curve <Pup ( TJ
8
t
.......................... ............................
.
corr. curve <Peorr( TJ

6
.2!
...
0
3:
0
4
;;::
'" 2
.c-
O I
~
-2
0 2 3 4 5 min 8
t,t
time ...
Fig.6.3. Correction of the experimental heat flow rate curve 4'exp for differing isothermals
6.1.2
The "Absolute" Dual Step Method
The second step (the calibration run) in the above outlined Cp measurement
procedure is not necessary, if the temperature dependent calibration factor
K<t>(T) is truly constant in time and has been determined carefully (see Sect.
4.4.1). The heat capacity of the sample can then be calculated as follows:
K<t> (T) (I/>s - 1/>0)

Cs =
f3. ms
This method obviously saves time compared with the procedure in Sect. 6.1.1.
A disadvantage is that conditions of thermal symmetry between sample and
reference sample are lost and with them the partial compensation of related
errors, and that the conditions of the calibration may have changed for the mea-
surement (and even the calibration factor) which increases the uncertainty of
the results.
The dual step method is, however, always a good technique for the investiga-
tion of samples with high thermal conductivities. The comments of Sect. 6.1.1
regarding errors and their corrections remain valid also for this simpler proce-
dure.
An interesting possibility to get better symmetry is to attempt to set up ex-

perimental conditions for the sample run which are very similar to those for the
zeroline run. For this purpose the sample is put into the correct crucible Sand
a reference substance CRef is placed in the crucible R on the reference side. The
(differential) signal L1 tP will approach that of the zero line tPo when Cs ::::: CRef The
sample specific heat capacity is then given by:
mRef K</>(T) . (L1tP - tPo)

Cs = - - CRef +
ms f3. ms
At first sight this seems to be an advantageous method, because the uncertainty

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 con-
ditions of both substances (or runs). Unfortunately it is very difficult to recog-
nize these influencing factors and to correct them because:
- Conditions for heat transfer may be poor, although rather small offsets of the
isothermals may conceal this and even suggest high-quality measurements.
- Thermal lag cannot be corrected by the simple procedure outlined in Sect. 6.1.1.
In addition, desmearing has not so far been carried out under these conditions.
6.1.3
General Precautions for the Minimization of Errors and their Estimation
The determination of reliable heat capacities with the procedures mentioned in

Sects. 6.1.1 and 6.1.2 requires the knowledge of the true heat flow rate into the
sample (i.e., the difference between recorded signal and the zeroline). The zero-
line cannot be obtained simultaneously from the same run, it is even impossible
to get it unless the sample crucible has been replaced once (or several times).
This implies that the original setup of sample, crucible and temperature sensors
can never be reproduced perfectly. The unavoidable changes in the conditions
for the heat transfer, which normally are concealed, are responsible for the
limited certainty of the obtained heat capacities. For minimum uncertainties
during routine operation, the following precautions are advisable:
- very careful sample preparation taking care of optimized thermal contacts at
all interfaces,
- sample and reference substance should have similar heat capacities and, as
good as possible, similar thermal conductivities,
- temperature intervals must not exceed 100 to 200 K for a single run. If larger
temperature ranges are needed, they should be divided into proper sub-inter-
vals. For isoperibol DSCs the initial and end isotherms are in principle non-
linearly connected (due to asymmetries of radiation and convection losses)
and this tendency increases at higher temperatures. This must be borne in
mind when selecting appropriate temperature intervals. Overlapping temper-
ature intervals are helpful to detect possible differences in the observed heat
capacities reflecting the uncertainty of the experiments,
- heating and cooling runs should be performed and compared,

- the measurement procedure should include calibration and zeroline runs
before and after the sample runs at the same conditions (possibly repeated
several times),
- constant purge gas flow,
- constant line voltage,
- preheating of the calorimeter (including the sample, if possible) to at least
10 to 15 K above the maximum temperature of the measurement in question,
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 (ice)
from the atmosphere or of volatile components of the sample on the cold sur-
roundings of the measuring system,
- constant room temperature,
- in the case of isoperibol DSCs, a very good temperature control of the ther-
mostated block, which influences the long-time stability of the recorded sig-
nal to a great extent. If, for instance, the DSC has been heated to a high final
temperature during the zeroline run and the stabilization time before the
sample run was not long enough, the following sample run would result in
a different isotherm. The same happens if the scan is started before the iso-
thermal baseline has reached a steady state (cf. Sect. 6.1.1).
Detailed discussions can be found in papers of Richardson, 1989; Suzuki, Wun-
derlich, 1984; Poe:Bnecker, 1990; Flynn, 1993 and Rudtsch, 2002.
For cp measurements performed with the mentioned methods, typical uncer-
tainties of 1-5% are frequently reported in the literature. The comparison of
the results of different authors is difficult, because the complete evaluation pro-
cedure is rarely published. Using the "Guide to the Expression of Uncertainty
in Measurement" (GUM, see Sect. 7.3.2), Rudtsch, 2002, presents a detailed
estimation of error for cp measurements according to the "three-step procedure"
(Sect. 6.1.1) in a power compensated DSC. If the complete uncertainty budget
is considered, the author obtains temperature-independent uncertainties of
U (cpJ/cp = 1.5 % for samples of sapphire and glass ceramics between OC and
600C (at a level of confidence of approximately 95%). Rudtsch suggests the
use of two semi-quantitative criteria to check whether a single run fulfils the
repeatability requirements: the temperature-dependent function of isothermal
heat loss (Poe:Bnecker, 1990, 1993; cf. Sect. 6.1.1) and the temperature-dependent
function of thermal lag. At first, these functions must be determined using well
defined reference samples. The single runs of unknown samples (within the re-
spective temperature intervals) can be accepted if the deviations from the values
of these functions fall within a preset confidence interval (e.g., 10%).
6.1.4
Procedure of Small Temperature Steps
The techniques described so far allow a large temperature range to be covered in
one run. This gives a continuous heat capacity-temperature curve. An alterna-
tive procedure is to measure the exchanged heat (area) which is connected with
mW
t
cu 3
....
<:>
-
~
o
<:>
cu 01
.z::.
a o
o 2 3 4 5 6 7 8 9 10 11 min 13
time ..
12
t:..T--_
r:
Q1
r
f--:
b o
o 2 3 4 5 6 7 8 9 10 11 mm 13
time ..
Fig. 6.4a, b. Discontinuous method for the determination of the heat capacity (upper curves:
sample run, lower curves: zeroline). a Small temperature intervals f:.T, the heat flow rate does
not reach the steady-state in the heating phase, b larger temperature intervals f:. T, the heat flow
rate comes to steady-state
a small temperature step. Thus the total temperature range must be divided into
narrow intervals (Fig. 6.4) which are successively scanned with isothermal periods
in between (Flynn, 1974; 1993). Depending on the temperature interval the pro-
cedure corresponds either to Fig. 6.4 a or to Fig. 6.4 b. The same operation should
then be repeated with empty crucibles (of same mass) to see whether there are
remaining peak areas (caused by asymmetries) which have to be subtracted to
get the correct areas.
If Ti and I1T represent the initial temperature and the temperature interval
respectively, the average specific heat capacity for the j-th temperature interval
between Tj-l = Ti + (j - 1) 11 T and Tj = Ti + j 11T can be calculated from the heat
Qj, which is proportional to the area enclosed by the sample and zero runs:
(cs (Tj = KQ(Tj) . Qj (6.1)

I1T ms
It is very important that the isothermal periods after a temperature step are
long enough to attain steady-state conditions. These times differ according to the
type of calorimeter and should be determined experimentally. The older power-
compensated Perkin-Elmer DSC's (DSC 2, DSC 7), for example, need 1 to 3 min-
utes, longer times are needed for samples that are large and/or have poor ther-
mal conductivity.
By analogy with Sect. 6.1.2, the empty reference crucible may be replaced by
a crucible containing a calibration sample. The "11" in the symbol 11 (Qj) of the
following equation is used to label this. If all crucibles, both on the sample and
reference side, are assumed to have the same mass, we obtain:
KQ(Tj) 11 (Q;) mRef' CRef

(cs(Tj = +---
I1T 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 ad-
vantage 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 and, of course, performed for a heating rate of zero. If the
I1T-intervals are in the range of 1 to 2 K the calculated average value (cs (Tj re-
produces the searched value Cs (Tj ) sufficiently. The interval may be increased to
5 or 10 K when there is a low temperature dependence of the heat capacity.
Early applications of the discontinuous technique concern the purity deter-
mination of organic substances (Staub, Perron, 1974; Zynger, 1975), without
really convincing success compared to the usual slow linear heating of the sam-
ples. All in all, the discontinuous procedure has been little used so far. Possible
reasons for this are:
- Long measurement times are needed due to the longer response times, also
in the case of the older power compensated DSC's. To determine the heat ca-
pacityover 100 K with the continuous procedures of Sects. 6.1.1 and 6.1.2 re-
quires 30 to 60 min at a heating rate of 10 K min-I. In contrast the discontin-
uous technique takes 400 to 600 min for I1T-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

10 K 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.
- Commercial software is normally not available and a major effort must be put
into the proper evaluation of the raw data.
- The accuracy of heat capacities measured in this way was not higher than for
those determined using the continuous technique. For normal sample masses
(l0 to 20 mg), temperature intervals of 1 K and specific heat capacities of
0.5 J g-I K-l, 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 melting peaks.
Step-Scan Procedure
The discontinuous heating or cooling technique made a comeback with the re-
cently introduced method of StepScan-DSC (SSDSC) of Perkin-Elmer (Cassel,
2000). The method utilizes the very low masses of sample and reference furnaces
and the rapid response times of the power-compensated DSC to perform a
rather fast repetitive sequence of short heat-hold segments over a large temper-
ature range (step-scan mode of operation, see Sect. 2.4.2). A zero line run and! or
a run with a reference substance (sapphire) as needed in the three step method
(Sect. 6.1.1) is in generally not necessary due to certainty and repeatability of the
heat flow rate and peak area calibration valid over wide temperature ranges.
However, an additional zeroline run (cf. Sect. 6.1.1 and 6.1.2) and its subtraction
from the sample run before peak area evaluation is highly recommended for
precision measurements of cp If the DSC is well balanced (equal masses for
sample and reference crucibles) a zeroline correction is not necessary.
The step-scan mode of operation is explained in Sect. 2.4.2. Typical tempe-
rature intervals (heating rates 10 K min-I) for each step are 1-2 K, typical iso-
thermal increments are 20-30 s. The result is a series of peaks (cf. Fig. 6.4a).
A cp-curve is generated from the peak areas as explained in Sect. 5.5.3.
Due to the very short times needed for each step, the measurement of the
thermodynamic (vibrational) heat capacity with SSDSC is not sensitive to base
line drifts, even at the highest accessible temperatures. Measurement of the heat
capacity with the step-scan method has some advantages:
- High heating rates can be used for the temperature step and the equilibration
time is short.
- The measuring time is distinctly lower (by a factor of 2-3) compared to the
common short step method.
- The data treatment is straightforward, Fourier analysis of the raw data is not
needed.
Moreover, for advanced applications it was shown (Merzlyakov, Schick, 2001 b)
that this SSDSC technique can be considered as time domain TMDSC. If desired
the frequency dependent (complex) heat capacity Ceff(W) can be obtained from
single experiments for nearly three decades of W (cf. Sect. 5.5).
6.1.5
The Temperature-Modulated Method
With the temperature-modulated DSC (TMDSC, see Sect. 2.4.2 and 3.3) the heat
capacity of the sample is determined from the amplitude of the modulated part
of the heat flow rate [see Eqs. (3.15), (5.13)]:
CPA
Cp=cpm=--
TAw
with CPA the heat flow rate amplitude, TA the temperature modulation amplitude,
w = 2n/tp the angular frequency (tp the period), and m the mass of the sample.
If the TMDSC is carefully calibrated (see Sect. 4.7) the method offers the possi-
bility to measure the heat capacity with some percent uncertainty in only one
run, the subtraction of a zeroline is normally not required (but recommended
for precision measurements). This method has the one big advantage of being
practicable even in isothermal mode of operation, where the DSC methods pre-
sented so far fail.
Outside of the temperature region of transitions or reactions the TMDSC
method results in the static (vibrational) heat capacity. The temperature depend-
ent values are, of course, independent of the frequency or period used, provided
the calibration has been done properly (see Sect. 4.7). That's why one may take
one suitable value, say a period of 60 s, for all such measurements, which simpli-
fies the calibration procedure markedly. The same is true for the temperature am-
plitude, which again should not influence the measured result if we move within
the limits of linearity (see Sect. 3.3).A proper value for TA is 0.5 K at a heating rate
of 2 K min-I. Figure 6.5 shows an example of such a measurement, the correspon-
dence between the measured and literature values is conspicuous.
The situation is, however, quite different if transitions or reactions occur in the
sample. Within the temperature region of such events the evaluation of the heat
flow rate amplitude results in an "apparent heat capacity" which contains not only
the vibrational, but even an "excess heat capacity" which contains contributions
from the latent heat of reaction or transition as well as possibly time dependent
changes of the heat capacity from relaxation processes (e.g., the glass transition,
see Sect. 6.4). In such cases the measured apparent heat capacity may very well
depend on frequency and it is just the frequency dependence which enables one
to get detailed knowledge of the processes involved, one example is given in Sect.
6.8.3. The deconvolution of an apparent heat capacity made up of several compo-
nents into its different contributions is not an easy task and needs in every case
frequency dependent measurements and therefore an extensive calibration, in
particular if a complex heat capacity (i.e., a time dependent heat capacity) has to
be determined which necessitates a precise phase signal as well. Such a deconvo-
lution needs special experiences and is still a matter of scientific discussion and
normally outside the scope of common heat capacity measurements. However, for
Fig. 6.5. TMDSC heat capac- 3

ity measurement for poly-
styrene (PS); solid line: JK- 1 g-1
reversing cp , dashed line:
underlying cp , symbols: liter- 2
ature values for the solid and t
liquid state (ATHAS, 2002) I
(m:4mg,po: 2 Kmin-I,
tp: 60 s, TA : 0.5 K) [p
OL-----~----~ ______L -_ _ _ __
-50 o 50 100 O( 150
temperature ..
most cases it is true that the higher the frequency, the lower is the contribution of
the excess heat capacity to the total apparent heat capacity. This offers the possi-
bility to get the static vibrational heat capacity via extrapolation of the measured
apparent heat capacity to infinite frequency where time dependent processes
don't contribute anymore (Schick et al., 2003). To extend the normal frequency
range of a common TMDSC, the evaluation of higher harmonics is essential, this
requires a non-sinusoidal temperature modulation (see Sect. 5.5).
6.1.6
Typical Applications of Heat Capacity Measurements
The heat capacity is a basic quantity for determination of other thermodynamic

quantities as well as material properties. Typical applications are listed in the
following:
- Very important from the theoretical point of view is, on the one hand, the
comparison of theoretical and measured heat capacities and, on the other, the
calculation 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 et al.1990,
ATHAS, 2002). The majority of these are DSC measurements and generally re-
fer to relatively high temperatures (> 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 calcu-
late 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 et al. 1990; Roles, Wunderlich 1993) the spectra may be approxi-
mately separated into group and skeletal contributions. The former are ob-
tained 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 Cv , the molar heat
capacity at constant volume. To compare with the experimental Cp a relation
between both quantities is needed. The thermodynamically exact conversion
T. V.y2
Cp-C v = - - -
X
cannot often be used because neither the thermal expansivity coefficient y
nor the compressibility X are known. A good approximation is to use a mod-
ified Nernst-Lindemann equation (Roles, Wunderlich, 1993).
3R Ao' Ci- T
Cp - Cv = 0
Cv ' Tfus
nus equilibrium fusion temperature; Ao empirical constant (for many poly-

mers: Ao = 3.9 10-3 K mol rl; R universal gas constant).
When the heat capacity functions Cp(T) are known from as near as possi-
ble to a K on, the thermodynamic functions H (T), 5 (T), F (T) and G (T) can be
calculated as normal:
T T C (T)
H(T) = Ho + f Cp(T) dT 5 (T) = 50 + f -p- dT
o o T
F(T) = U(T) - T 5(T) G(T) = H(T) - T 5(T)
Low temperature DSC measurements of Cp are not possible and for this tech-
nique only changes in these functions can be calculated.
- Calculation of the reaction enthalpies from the measured heat capacities of
reactants and products using Kirchhoff's law:
T2
L1 rH(T 2 ) = L1rH(Tl) + f I (Vi' Cm,i(T)) dT
Tl i
L1rH(Td, L1rH(T2 ) molar reaction enthalpies at Tl and T2 ; Cm,i(T) molar heat

capacity 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. 6.4) requires heat
capacity functions for the non-equilibrium glassy amorphous and the
equilibrium 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 + al T + a2 T2 + ... or C(T) = ao + al T + a2 T- 2 , ao is not need-
ed. It is generally found that a linear approximation (all ai > a 1 are zero) is suf-
ficient.
6.2
Determination of Heats of Reaction
The aim is to determine a thermodynamically well defined (temperature de-

pendent) reaction enthalpy. If a subsequent kinetic analysis is planned, the cal-
culation of a similarly well defined conversion-time curve is of equal impor-
tance. The interpretation assumes that both the initial and the final state of the
reaction are unambiguous and well-defined. Is this not the case, the results will
be, at best, semi-quantitative. Because of the limited precision of DSCs minor
effects due, for example, to changes in pressure, stress or surface contributions
may be neglected. The following, simplified equation is then obtained for the
measured heat flow rates in a DSC:
Q
(d ) = <Pm = Cp,5(T). dT + (OH) . d~
dt p dt o~ T,p dt
Cp,{;(T) is the heat capacity of the system at constant pressure and at constant
extent of reaction. The partial molar reaction enthalpy flrH = (oHlo~h;p is nearly
always replaced by the average reaction enthalpy (flrH), which is independent of
the instantaneous composition of the system. The Kirchhoff equation
describes the relation between the change of the 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 program). Advantages and disadvan-
tages of the two methods are discussed in detail in Sect 6.3.
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' Simple extrapolation of this <Pend to the start time yields the base-
line of the reaction peak. As there is no heat capacity contribution to the heat
flow rate in the isothermal mode, the relation between the heat produced or con-
sumed and the average reaction enthalpy is very simple:
t (;
Qm(t) =f (<Pm - <Pend) dt = (flrH) f 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 ]lW. 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 (which is essential for a
kinetic analysis), the temperature dependence of the reaction heat is also ob-
tained. Alternatively, if the Cp (1') functions of reactants and products are
known, .1.r H (1') may be calculated by means of the Kirchhoff equation.
The main problem with isothermal measurements results from the ill-de-
fined behavior 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 isother-
mal heat flow rate curve of a second run with the same (now totally reacted)
sample and exactly the same conditions. Nevertheless, minor differences may re-
main because of the different thermal conductivities of products and reactants
and because heat transfer conditions may have changed during the reaction.
In order to check the completeness of the reaction, the sample is often heated
to higher temperatures. However this is only useful if side reactions or shifts
of chemical equilibrium 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. 5.3.2. However the situation is more difficult, because it cannot be assumed
that reactants and products have almost the same and temperature-independent
heat capacities owing to the large temperature interval of the reaction (100 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 tran-
sition within the reaction region. In addition, the glass temperature of the orig-
inal reactants may shift into this temperature range. As a typical example, the
curing of an epoxide resin is shown in Fig. 6.6 (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 zeroline has no curvature or
has been subtracted and that changes in heat transfer conditions can be neg-
lected. 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 1:
A linear baseline is drawn between the points A and B (Fig. 6.6). 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. 6.7b, this is not true for this system, even after allowance for the
glass transition.
Fig.6.6. Curing of an epox-

ide resin (according to 2.5
Richardson, 1989) and the J
procedure mostly used to K-g
determine the heat of reac-
tion. The linear baseline
between start (A) and end t 10
A
B
(B) of the reaction ignores
all possible changes in heat >-
capacity ~
c
8" 1.S
c
'"
.c
~ 1.0
'"
300 350 400 450 K 500
tempera ture ..
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 transitions),
then the baseline can be calculated by iterations similar to those of case 3 in
Sect. 5.3.2. Method 1 above gives the Oth approximation to the extent of reaction
function ~(T). The heat capacities of the products can easily be obtained from
the rerun on the fully reacted sample. If the first run of the reaction mixture can
be started at least 30 K before the reaction starts, the heat capacity of the reac-
tants may be determined from this interval. Extrapolation into the reaction tem-
perature 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 of the baseline constructed in this way from the measured
curve, an average heat of reaction is obtained (the term reaction enthalpy should
be avoided, the quantity has no thermodynamic significance) for the tempe-
rature range between T J and T2 Improvements relative to method 1 are only
found if the heat capacity functions are known with sufficient precision (uncer-
tainties < 1 %). Obvious improvements in the definition of the baseline should
result from the use of temperature-modulated DSC (cf. Sect 6.1.5). By this
method it is possible to get both the heat flow rate curve (which is proportional
to the actual reaction rate) from the non-reversing curve, and the heat capacity
function of the reacting mixture from the reversing curve, from the same run
(see Sect. 6.3.6)
The degree of reaction function a(T) can be obtained from the ratio be-
tween partial heats (up to a certain temperature) and the overall heat of the re-
action.
Method 3:
Many of the above problems and difficulties can be avoided by integration of the
Cp curves (Richardson, 1989; 1992b; Flynn, 1993). The specific reaction enthalpy
Arh {Td at the starting temperature TI (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 Fig. 6.7 it follows that
The reaction enthalpy at temperature TI (Arh (TI is obtained as difference be-

tween the areas X and Y. The baseline is not necessary in this case, because only
differences are needed:
d r h (TI ) = 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 Rand
the liquid-like or rubbery state of the products P. The difference hI (P, 490 K) -
hI (R, 350 K) corresponds to the experimental quantity, the area X, defined by
the lines at Tl> T2 , cp = 0 and the reaction curve. The horizontal hatched area Y
follows from the rerun on the reacted sample.
490K
Y = ht (P, 490 K) - ht (P, 350 K) =I Cp,t (P) dT
350K
If the glass transition of the products is above TI , as is the case in Fig. 6.7, 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 Tbetween TI and T2 (Richardson, 1992b; cf. also Sect. 5.3.2).
At first a possible glass transition is disregarded. One obtains:
Arh (T) = h (P, T) - h (R, T)

= {h (P, T2) - h (R, Td) - {h (P, 12) - h (P, T - {h (R, T) - h (R, TI
or:
T2 T
Arh{T)=X- ICp{P)dT- ICp{R)dT=X-Z
T TJ
This procedure is schematically shown in Fig. 6.8. The areas X and Z are hatched
horizontally and vertically respectively. A simple rearrangement of the last
equation gives:
T2 T
Arh (T) =X - I cp{P) dT+ I (cp{P) - cp{R dT
TJ TJ
T
= Arh (TI ) + I {cp (P) - Cp (R dT
TJ
Fig. 6.7 a, b. The thermody-

namically correct determi- 2.5
nation of the specific heat
of reaction (according to J/IKgl
Richardson, 1989); for details
see text. a 1st run curve,
2.0
b 1st and 2nd run curves
Hatched areas extended
to cp = 0; ~ h (350 K) =
r>-
B
X - Y = hproducts (350 K) - ~
hreactants (350 K) = - 44.7 J g-l '"c. 1.5

'"
~
'"
C1I
..r:::.
~
~
C1I
c.
VI
10
a
..
300 350 400 450 K 500
tempera ture
2.5
J/IKgl
r
2.0 B
>-
~
'"c. 15
'"
~
'"
OJ
-
..r:::.
~
~
C1I
c.
VI
10
b
300 350 400
tempera ture
450
.. K 500
It is clear from Fig. 6.8 that this procedure is a direct application of the Kirchhoff
equation. An enthalpy-temperature diagram (Fig. 6.9) is especially clear and in-
structive. Here, for the sake of simplicity, temperature independent heat capaci-
ties 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 HI (R) and HI (P)] state. A re-
action 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
Fig.6.S. The thermodynam-

ically correct determination ~rHITI = X-Z
of a reaction enthalpy
(according to Richardson,
1
1992 b); for details see text
'p (R) 'p(PI
--- ~r"' I--
r
!\
-...'" Z
-.......
.c
"
a..
II>
X
r, T
temperature ..
Fig. 6.9. The determination

of reaction enthalpies from
an enthalpy-temperature
diagram (according to Flynn
1993); A: integral of Cp(n
for the reactants, B: integral
of Cp(n for the reaction
mixture, C: integral of Cp(n
for the products
1
>-
C1.
CI
.c
CQ.I
I __ - _-K[
I II
...l- I I
I I I
Tg (R) T Tg (P)
temperature ..
products [Tg (P) 1is very often somewhere between Tl and T2 A formal (Flynn,
1993) and thermodynamically correct procedure yields four different reaction
enthalpies and four different a values at each temperature (see Fig. 6.9):
glassy reactants -7 glassy products ArH =bd, ~ =bc/bd
glassy reactants -7 liquid (rubbery) products ArH = be, ~ =be/be
liquid reactants -7 glassy products ArH = ad, ~ = ac/ad
liquid reactants -7 liquid (rubbery) products ArH = ae, ~= ac/ae
Only the final case has any meaning for kinetic studies.
In exceptional cases, e. g., when there is a possibility of side or decomposition

reactions, the heating run should be stopped at a temperature at which such dis-
turbing reactions don't contribute to the heat flow rate. 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 the reaction totally in the isothermal mode.
6.3
Kinetic Investigations
6.3.1
Introduction and Definitions
Every chemical reaction is associated with a certain heat of reaction. From

this it follows that the heat flow rate is proportional to the rate of reaction. The
assignment of the time dependent heat flow rates to defined reactions leads to
kinetic data. Every quantitative kinetic analysis starts with the determination of
a continuous sequence of concentration-time data. DSC methods are widely
used to solve kinetic problems because of the simple and fast sample prepara-
tion and the wide range of experimental conditions - much information is
produced in a short time. The technique, by contrast with many other methods,
immediately gives a series of "reaction rates" as function of the extent of reac-
tion 5, dependent on time and temperature. The aim of kinetic investigations is
then to find quantitatively this functional relation. The methodology is quite
general and it is immaterial whether the reactions investigated come from inor-
ganic, organic or macromolecular chemistry. Continuing advances in instru-
mentation and data treatment facilitate refined calculations.
The framework of the kinetic description of reactions was originally devel-
oped in physical chemistry for reactions in homogeneous phases. A generalized
reaction with the educts A, B, ... and the products C, D, ... is described by the
following stoichiometric equation:
aA + bB + ... ~ cC + dD + ...
According to the definition (ni = ni,O + Vi 5),5 can be related to the consump-
tion of an educt (Vi < 0) or to the formation of a product (Vi> 0). The rate of re-
action is then quite generally given by:
1 d5
r=--
V dt
Considering a reaction in homogeneous phase and at constant volume, the rate
law for the overall or the elementary reaction may then be formulated as usual:
1 d5 1 dCi
r=--=--=!(CA,CB, ... T,p, ... )
V dt Vi dt
6.3 Kinetic Investigations 169
or quite general
r= f(n f(e}
If other possible variables such as concentrations of catalysts or inhibitors are

kept constant, one obtains at constant temperature and constant pressure fre-
quently:
1 dS n n
r = - - = k(T) eAA eBB
V dt
k(n 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.
Of course, a DSC curve does not provide the concentrations of the respective
components themselves, but they may be obtained unambiguously from the
measured heat flow rate 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 analyt-
ical methods (e. g., from IR, UV, NMR, MS, GC, HPLC etc.).
Using thermo analytical methods for the investigation of kinetic problems,
the concentrations of the reactants are usually replaced by the degrees of reac-
tion a = S/Smax> in particular if reactions in heterogeneous systems are investi-
gated.As a rule, a is set as the ratio of partial area and total area of the DSC peak:
Qt da If>(t}
a(t} = - and ----
Q~ dt Q~
But this is only valid if the above restrictions are fulfilled! In every case
the measured reaction heat Q~ must be checked and, if necessary, corrected
with respect to the final degree of reaction actually reached. This correction
must be made whatever the reason for the cessation of reaction - perhaps
caused by a great increase in viscosity (vitrification) or coming to an equilibri-
um state. All subsequent kinetic analyses are incorrect if this modification is
omitted. This is because the calculated a(t} is not related to the thermodynam-
ic degree of reaction but to an apparent final state that is specific to the individ-
ual experiment.
A general rate law is written as
da
- = k(n f(a}
dt
The functionf(a) is very simple for the smallest steps (elementary reactions) of
reactions in homogeneous systems. In the case of equimolar mixtures one ob-
tains then:
da
- = k(T) . c8- 1 (1- a)n
dt
da/dt is measured in units of reciprocal time, a ranges from 0 to 1. Using the

DSC in the scanning mode at a constant rate (fJ = dT/dt = const), the term da/dt
will be replaced by the term fJ da/dT (minor self-heating of the sample during
the exothermic reaction is neglected).
Using thermoanalytical methods, c8- 1 in k(T) c8- 1 is usually neglected. Ele-
mentary reactions have always constant and integer reaction orders n within
the whole range of reaction degrees, broken values suggest a more complicated
mechanism. 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. To some extent the
above rate equation 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 in homogeneous systems cannot be described, at least within
the total range of conversion, by the above n-th order rate law. A frequently used
overall rate law, which in addition allows for catalytic and autocatalytic effects,
is (Sourour, Kamal, 1976):
r = k(T) . am. (1 - a)n
The most general equation was suggested by Sestak, Berggren, 1971:
r= k(T) am. (1- a)n. [-In(1- a)]P
The formal reaction orders n, m and p are only parameters of variation, to ob-
tain a better fit to the experimental curves, an interpretation in the physico-
chemical meaning is difficult or impossible.
The situation is much more complicated for reactions in heterogeneous sys-
tems. 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 compo-
nents. Further, the results may be influenced by the energetic heterogeneity of
the solid surface. It is typical of heterogeneous reactions that very different
events such as nucleation, growth and diffusion can occur simultaneously and
successively. A large number of special functions f(a) in the rate laws for these
cases were derived in the past. Above all, the Avrami-Erofeev equation is of con-
siderable importance for describing the kinetics of both decomposition reac-
tions of solids and the isothermal or non-isothermal crystallization. The kinetic
analysis is additionally complicated because we have numerous additional
sources of error connected with sample preparation and measuring conditions.
Heat flow rate curves are frequently difficult to analyze without help from other
analytical methods (especially temperature programmed X-ray diffraction, hot
stage microscopy, thermogravimetry and possibly mass spectroscopy, atomic
absorption spectrometry and gas chromatography}. A review and a critical judge-
ment of the most frequently used rate laws is given by Brown et aI., 1980, and by
Galwey and Brown, 1998.
The temperature dependence of the rate constant k (T) is usually described by
the empirical Arrhenius equation or rarer by the Eyring equation, which follows
from the activated complex theory:
Arrhenius: k(T} = A exp (- ;~ )
Eyring:
k(T} = -h- kT
B
exp (~S")
R .exp (AH"
- RT )
kB is the Boltzmann constant, h the Planck constant

Both equations were first developed for gas phase reactions, but they are valid
also for reactions in homogeneous liquid phase. The Arrhenius parameters
(A preexponential factor, EA activation energy) and the Eyring parameters (AS"
activation entropy, AH" activation enthalpy) are for condensed systems related
to one another by:
Ah
AS"=Rln---
e k B T
The Eyring activation parameters are more suitable for the understanding of re-
lations between the structure of the reactants and the reactivity (Heublein et aI.,
1984).
Although open to criticism (Flynn, 1990), as a rule the traditional concept of
activation is also used for reactions in heterogeneous systems. The main theo-
retical difficulty results from the fact that the energy distribution of the immo-
bilized constituents of a solid is not represented by the Maxwell-Boltzmann
equation, respectively, that we find a number of surface sites with different
energies and energy distributions. For instance, Hunger et al., 1999, have deter-
mined the activation energy distributions for the desorption of water from dif-
ferent zeolites and they have definitely proved the existence of several different
adsorption places. But meanwhile two reasonable approaches supply also for
this type of reaction a theoretical foundation for the exponential k- T relation-
ship. Galwey and Brown, 2002, start from the band structure of a solid and con-
sider discrete interface energy levels in the locally modified environment
between the band structures of reactant and product phases. The occupancy of
the interface levels is directed by energy distribution functions based on Fermi-
Dirac statistics for electrons and Bose-Einstein statistics for phonons. At higher
energy differences between the interface levels and the highest occupied elec-
tron bands (compared with kB T) both distributions approximate to the expo-
nential term of the Maxwell-Boltzmann statistics. The second and promising

physical approach to the interpretation of the observed exponential k- T relation-
ship was developed by L'vov, 2001, 2002, for some types of solid state decompo-
sition reactions as an alternative to the traditional chemical approach based on
the Arrhenius concept. The reaction scheme (using the Hertz-Langmuir predic-
tion of the dependence of reactant evaporation on the equilibrium vapor pres-
sure) is based on the congruent dissociative evaporation of the reactants with
simultaneous condensation of the low-volatile reaction products. In the interval
between evaporation and condensation, the gaseous species can diffuse for
some distance from the primary sites. L'vov has shown for different substance
classes that this model also accounts for such features as: the mechanism of
nucleation and growth, the autocatalytic development of some decomposition
reactions and the influence of gaseous products on the reaction rate.
To summarize we can formulate three main objectives of kinetic investigations:
1. determination of the rate equation or the rate law and the corresponding ac-
tivation parameters,
2. determination of the reaction mechanism,
3. prediction of the reaction behavior for any complex time-temperature pro-
files.
First of all, kinetic analysis is of course an efficient tool for data reduction. A
large body of experimental data from a series of measurements with many data
points has then been reduced to a model with few parameters. Further, reactions
may be judged using different criteria. For example, the influence of variations
of the activation parameters can be discussed together with changes in the com-
position of the reacting system (stoichiometry, catalysts, solvents and fillers) or
with modifications of one reactant (number and position of substituents,
homologous series).
But the true aim of a kinetic analysis is an evaluation 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. The inseparable combination
of rate law, activation energy and preexponential factor is sometimes named the
"kinetic triplet" (Maciejewski, 2000). Frequently, the form of the overall rate law
already permits conclusions according to a possible mechanism. The rate law
can be simple or complicated. A complicated rate law always signals a complex
mechanism with a number of steps. On the other hand, the opposite conclusion
is not correct. A simple rate law can also be caused by the existence of a very slow
reaction step - the rate-determining step - within a number of much faster steps.
The investigation of the reaction mechanism is an extremely demanding task.
Much time and effort is required to reach it and supplementary investigations
using other analytical techniques are essential. Even then, the result is often only
a most probable mechanism or the knowledge of the most important steps.
A third aspect of kinetic investigations becomes more and more important.
The practitioner in industry needs a useful and quickly accessible reaction mod-
el, which on the one hand adequately describes the course of the reaction and on
the other allows reliable predictions for any temperature-time conditions. The
form of this reaction model (overall rate law, combination of formal steps in
a formal-kinetic model, use of the "true" mechanism as sum of all important

elementary reactions) is unimportant.
6.3.2
Experimental Prerequisites for a Reliable Kinetic Analysis
Kinetic investigations 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.
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 vapor pressure must be loaded into special
sealed crucibles. Errors due to the effect of pressure on the measured reaction
rates 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 out-
side the reaction 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 of the 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
composition should be carried out in order to exclude accidental errors dur-
ing 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 behaviour in a
reactor) the very different conditions must be recognized and, in particular,
allowance made for heat transfer effects.
3. Data Acquisition and Processing

- Some varieties of one-point-evaluations, e.g., In(f3IT~ax) versus lITmax
(Kissinger, 1956,1957), lnf3 versus lITmax (Ozawa, 1965, 1970) or lnf3 versus
lITa=o.s (van Dooren, Muller, 1983), are fast and easy to realize but no longer
state-of-the-art. Recommendable kinetic evaluations need the complete

a(t, T) function and often also that of the rate of conversion daldt(t,T).
Therefore, the correct baseline must be determined first (according to the
methods 1-3 in Sect. 6.2) and then subtracted from the measured curve.
Commercial evaluation programs offer usually the following possibilities
after the reaction range was selected:
- a straight line between start- and endpoint,
- a sigmoidal baseline whose differences from the simple linear case are al-
ways proportional to the area of the peak,
- horizontal lines starting either from the left or from the right. The left
starting base line should be used if for any reasons it is not possible to
determine the end of reaction, whereas the reaction start is well defined.
Using the right -hand horizontal line from the assumed end of the reaction
is the standard baseline method for isothermal measurements.
The reliability of kinetic results strongly depends on the correctness of the
baseline! Therefore, it would be ideal if the baseline could measured simulta-
neously and independently. This is nowadays possible by using the modern
temperature-modulated or step-scan techniques (see Sect. 5.5 and Reading et
al., 1994; van Assche et aI., 1995, 1996, 1997; Hu, Wunderlich, 2001; Flammers-
heim, Opfer mann, 1999a, b, 2001).
- In general, however, several reactions contribute to the measured heat flow rate.
Apart from so-called model free evaluations (see Sect. 6.3.5), an evaluation is
then possible only with respect to a certain model. If the gross reaction is made
up of, say, n reactions, all contribute to the measured heat flow rate and their re-
action heats can be treated as parameters and determined as such by variation
procedures. It is obviously much better to determine 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., tem-
perature dependent spectroscopic investigations of equilibrium states etc.
- The experimental curve has to be desmeared (Sect. 5.4) before any evaluation
if there 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 ther-
mometer 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
calculated if the instantaneous heat generated by the reaction is related to the
heat exchanged with the calorimeter (this is limited by the overall thermal
relaxation 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 10 Keven
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 func-
tions of concentration or degree of reaction.
In much of the literature it is unfortunately not clear whether the above aspects
have been considered and the corresponding results must be treated with reser-
vation.
6.3.3
Selection of the Measuring Conditions - Isothermal or Non-Isothermal
Reaction Mode
Classical techniques of chemical kinetics normally operate in the (quasi-)

isothermal 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 par-
ticular advantages and disadvantages (Flammersheim, 1988, 1999a; b, 2000,
2001). They complement each other and should not be regarded as alternatives.
Modern, highly sensitive DSC devices, powerful computerization and sophisti-
cated evaluation methods together allow and favor the combined realization and
evaluation of isothermal and non-isothermal experiments. Further, it does not
matter whether model-free or model-based evaluations are made. Apparently
different conclusions are always obtained if non-comparable reaction condi-
tions were selected (Flammersheim, Opfermann, 2001). There is a permanent
risk to overlook consecutive and concurrent reaction steps with higher activa-
tion energies in isothermal measurements. When planning measurements for a
kinetic evaluation, the selection of the proper window in the global tempera-
ture-time-conversion (concentration) space is extremely important. This condi-
tion is easy to fulfil at dynamic experiments. They should be carried out using
very different heating rates. At least three or better five measurements should be
made, a practicable and reasonable range of heating rates is between 0.5 and
10 K min-I. This should be sufficient for scanning an adequate range within the
global reaction field (Opfermann, 2000). Single-heating rate scans are useless for
a kinetic analysis (Opfermann, 2000; Burnham, 2000). Burnham formulates in
his paper that judges the results of the [CrAC kinetics project 2000: " ... kinetic
analysis using single heating rate methods should no longer be considered
acceptable in the thermal analysis community." On the other hand is it often dif-
ficult or impossible to fulfil the above condition for isothermal measurements.
This is due to the necessary restriction to a relatively narrow temperature range,
in practice rarely more than so K. All comments made up to now are just as valid
if temperature-modulated or step-scan modes of operation are applied. As already
mentioned the main advantage of such techniques is a reliable determination of
the baseline, but at the expense of much longer times for the experiments.
In the following some typical advantages and disadvantages of the two main
operating modes (isothermal and scanning) are particularized. But it should be
mentioned that several other temperature control modes have been used. The
advantages and drawbacks were reviewed by Ozawa, 2001.
Isothermal mode:
- A simple and immediate interpretation of the measured curve is possible be-
cause of the complete decoupling of the two variables, temperature and time,
if the minor changes of the sample temperature during the reaction can be
neglected. This method is especially useful when searching for the most prob-
able reaction mechanism. Two examples will illustrate this:
During the polyaddition reaction of aniline and the diglycidyl ether of
bisphenol-A (DGEBA) the maximum conversion rate is found at non-zero de-
grees of reaction (Fig. 6.10). This can only be due to autocatalytic and/or con-
secutive reactions. Furthermore it is obvious that not only the ratio of the
concentrations of both reactants has substantial consequences but also that
equal excess concentrations of amine or epoxide differ greatly in their effects.
This is a very important pointer to the possible reaction mechanism.
The second example (Fig. 6.11) is the polyaddition reaction of a mixture of
the two isomers 2,2,4-trimethylhexamethylene diisocyanate and 2,4,4-tri-
methylhexamethylene diisocyanate (TMDI) with 3,6-dioxaoctane-l,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
100
Wig
1
2.25
-
T : 405 K
'"
...
c all sample masses about 15 mg
1.50
--
:3:
0 ra tio 0f OGEBA to aniline
c
1: 1:1
'"
.r:::. 2: 1:2
-'"
.~
' v 0.75
C1.
3:
4:
1:3
1:4
II> 5: 2:1
0.00
o 17 34 51 68 min 85
time ~
Fig.6.10. Isothermal reaction curves for the polyaddition of aniline and bisphenol-A digly-
cidyl ether (DGEBA) in mixtures with different stoichiometry
Fig.6.11. 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
increase in the slope of the leading edge of the reaction peak in direct pro-
portion to the concentration of catalyst is unusual. This can only be explained
if one assumes a direct coupling of catalysis and autocatalysis.
- The baseline follows unambiguously by extrapolating the measured heat flow
rate after the complete reaction, assuming that all parameters, which can
influence the measurement directly or indirectly (e. g., the temperature of
the surroundings) are sufficiently constant. Demands on the apparatus can be
considerable, 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 reaction can be allowed to take place at low temperatures such that de-
composition or side reactions can be avoided.
- Isothermal measurements need more time than scanning runs.
- The main disadvantage, already mentioned in Sect. 6.2, is that the initial phase
of the reaction (30-60 s) 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,
lime
100
------ lime
10
------ min 30
0
Wg-' Wg-'
-0005 -005
1 1
30 '( SO '(
-0.10
-0010
-0.15
0
Wg-' Wg-'
-05
1 -10
70 '(
-10
1

-20
-15
a 0 2 min 50 65 C 80
lime ------ temperature ------
0
os K min
-, Wg 1
1 -10
- 0.2
1

-1
10 K min - 04
0 5 10 15 min -50 0 50 C 100
-----
b
time ----- temperature
Fig.6.12. Calculated DSC curves for isothermal (left) and scanning (right) mode of opera-
tion. Reliable experimental values for isothermal mode are only accessible for t > 45 s (right
of the vertical broken lines). a autocatalytic 2nd order reaction: A + B -7 C + 2B with 19
(A/s- I L mol-I) = l3, EA = 100 kJ mol-I, Qr = -50 kJ mol-I, b simple second order reaction:
A + B -7 C with 19 (A/s- I L mol-I) = 5, EA = 50 kJ mol-I, Qr = - 35 kJ mol- I
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.
- The useable range of temperatures is, however, often strongly restricted and too
small for reliable kinetic evaluations. Two examples will demonstrate this. First,
let us assume an autocatalytic 2nd order reaction, as frequently found in epoxy
curing reactions (Sourour, Kamal, 1976; Klee et al., 1998; Flammersheim, 1998).
Such reactions are easy to investigate isothermally, as they often start with
a low reaction rate. Three isothermal and three non-isothermal DSC curves
(Fig. 6.12 a) were calculated for the assumed reaction path A + B ~ C + 2B, with
19 (Als- I L mol-I) = 13, EA = 100 kJ mol- I and Qr = - 50 kJ mol-I. The calculated
initial rates at 30 DC and 50 DC are so small that the distortion of measured
DSC heat flow rates can be neglected or reliably corrected (Sect. 5,4, Flam-
mersheim, 2000). But the time (> 6 h) taken to complete the reaction at 30 DC
already requires a device with very good long-term stability. On the other
hand the reaction at 70 DC is already so fast that a considerable part of this re-
action occurs within the first 45 s (on the left of the broken line in Fig 6.12a),
then each attempt to correct the initial heat flow rate must fail. In other words,
the experimentally accessible temperature window for absolutely reliable
isothermal measurements is relatively narrow (about 30-40 K).
The situation becomes dramatic if an elementary step with an nth order reac-
tion (n > 1) dominates. This is shown in Fig. 6.12b for second order kinetics
(A + B ~ C, 19 [AI(s-1 L mol-I) 1= 5, EA = 50 kJ mol-I, Qr = - 35 kJ mol-I). It is
no longer possible to find a temperature range in which error-free measure-
ments can be made. At temperatures higher than 25 DC the heat flow rate is
strongly disturbed and falsified at the beginning, while at temperatures low-
er than 25 DC, the end of the reaction cannot be detected because signal and
noise will have the same magnitude. Again, non-isothermal runs are clearly
favorable.
- If isothermal experiments are planned, a rule-of-thumb may be helpful.
Appropriate temperatures should cover the first half of the leading edge of a
temperature programmed scan with low heating rates (0.5-2 K min-I)
Scanning mode:
- Scanning measurements need less time than isothermal experiments but
construction of the proper baseline may be a problem. But as already men-
tioned this can now be bypassed using the temperature-modulated or the
step-scan mode of operation.
The measurement can be started at temperatures well below that of the
beginning of the reaction, so steady-state conditions of the DSC are ensured.
Unfortunately, reactions may go to completion, even at low heating rates, at
such high temperatures that secondary reactions cannot be neglected (this is
especially relevant for organic reactions).
It is not clear from the shape of the curve what type of reaction is involved
and this is a drawback when searching for the most probable reaction mech-
anism.
- The screening of the complete global reaction field, by carrying out non-
isothermal measurements with strongly different heating-rates, is needed to
avoid misinterpretations. This needs time but is always possible without
problems (compare Figs. 6.12a and 6.12b).
6.3.4
Activation of the Sample by UV Irradiation
Reactions studied in the DSC are normally thermally activated. The wide range
of experimental conditions (scanning rate, temperature range, even complicat-
ed procedures with intermediate annealing phases) is responsible for the exten-
sive use of this technique.
Many reactions, however, can also be started by irradiation with sufficiently
high energy. The investigation of such reactions has considerable value for the
optimization of process control parameters and this is especially true for those
polyaddition and polymerization reactions for materials with widespread use.
Slight changes in the design of a conventional DSC yield a so-called Photo-DSC
(see Sect. 2.2.2), normally used in the isothermal mode of operation. First appli-
cations of this type of DSC were described by Wight, Hicks, 1978; Tryson, Shultz,
1979; Flammersheim, 1981. The light (normally UV), having sufficient energy to
start a chemical reaction, is brought to the sample by use of classical lens and
mirror systems or by glass fiber optics. The portion of the spectrum which is
of interest is selected using monochromatic filters and the desired intensity is
adjusted by means of neutral density filters or metal sieves with different mesh
width. For precise measurements electronic stabilization of the radiation inten-
sity of the light source is extremely important. Light-activated reactions under
conditions closely related to practice are usually very fast. The greatest part of
300
Jase line of the reaction mixture with the absence of UV radia tion
mW 2nd run
.-.----.-~.-.---
base line of the reacted sample
with the presence of UV radiation
150
:.
o
c
cu
..c
15
o ~------~--------~--------~--------~--------~
o 0.7 1.4 2.1 2.8 min 3.5
time ~
Fig.6.13. 1st and 2nd run for the photopolymerization of a multifunctional acrylate at room
temperature. (photoinitiator: 1% 2,2-dimethoxy-2-phenylacetophenone (benzyl dimethyl
ketal), light intensity: 5 mW cm-2 at 365 nm)
the reaction takes place within a few seconds (cf. Figs. 6.13 and 6.14). There are
two consequences:
- There is a considerable change in the measured signal in periods comparable
with that of the DSC time constant. The measured and true heat flow rates dif-
fer substantially. Desmearing is necessary prior to every evaluation in which
the time variable is involved.
- Heat flow rates (cf. Fig. 6.13) 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- 1 is assumed, temperature
differences of 5 to 10 K would result even for samples with very good thermal
conductivity. For organic samples, these differences are even greater. This
means that the measurements are never isothermal, not even approximative-
ly! Kinetic analysis must allow for this.
A typical example of a light-activated measurement is shown in Fig. 6.13. 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 desmearing (deconvolution, see Sect. 5.4)
500
mW -- -----
375
r I
~
c 250 I
,,,
~
I
~
0
;:;::
c
QJ
,
..c
125
"
"
"
r
"
"
0
0 0.5 10 15 min 2.0
time
Fig.6.14. Baseline corrected experimental (solid line) and desmeared (dashed line) curves
for the photoinduced reaction of Fig. 6.13
of the very fast reaction, is obtained easily and is sufficiently precise if, after
the completed reaction, the response of the sample to a light flash is recorded
(Flammersheim et al., 1991). If it were not for the fact that the heat produced
during the chemical reaction is evolved inside the sample, 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 solid line in Fig 6.14 corre-
sponds to the desmeared curve. The dashed curve is the original measured curve
after subtraction of the baseline from Fig. 6.13.
The kinetic evaluation of light-activated reactions has so far only been par-
tially successful because the true reaction course is overshadowed by additional
influences:
- 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 reac-
tion rates within the sample. However, the calorimeter always records an over-
all heat flow rate. As result of this rate gradient there is a corresponding con-
centration gradient of the reactive component. An exact solution of the sub-
sequent system of differential equations is only possible for drastically sim-
plified boundary and initial conditions (Tryson, Shultz, 1979; Shultz, 1984). If
the measurements have been desmeared, the expected correlation between
average reaction rate on the one hand and sample thickness and light inten-
sity on the other hand is found (Flammersheim, Kunze, 1998).
- The situation may be made more difficult by the increasing viscosity of the
mixture during the reaction, which nearly always is carried out at room tem-
perature. As a result, the reaction may be incomplete (Klemm et al., 1985). For
example, at room temperature phenyl glycidyl ether (PGE) reacts completely
during cationic polymerization induced by irradiation of diaryliodonium
salts, whereas the reaction of DGEBA stops at about 50 to 55 % conversion
(Klemm et al., 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 in-
sufficiently reliable experimental 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 photo-initiator. If diffusion
limitations 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, photo-initiators and inhibitors,
- wavelength and intensity of the radiation used.
6.3.5
Different Strategies of Kinetic Evaluation
The Traditional Way - Search of a Rate Law and its Activation Parameters
for the Overall Reaction
In the past, many evaluations were directed to find a rate law and its activation pa-
rameters which adequately describes the overall reaction within a range of conver-
sion as well as possible (cf. Sect. 6.3.1). For technological purposes a description of
an even 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 at all of any value - it must be done very care-
fully. An overall rate law may give a good description over a certain range of reac-
tion, 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 ef-
ficiency for heterogeneous 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 simpli-
fications that may sometimes be made to the rather complicated differential equa-
tions that describe the true reaction mechanism. In spite of this, a determination of
the gross rate law may be a useful first step in finding the reaction mechanism.
The risk of using gross rate laws for the description of the reaction is, that this
possibly may 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 in other ranges of con-
version are then interpreted too narrowly within the framework of that model
only. A typical example is the Sourour-Kamal equation (cf. Sect. 6.3.1), which is
often used in the polymer chemistry literature to describe the epoxide amine
polyaddition reaction. For stoichiometric reactions it reads:
da
- = kJ . C (1 - a)n + k2 . am . (1 - a)n
dt
C isthe concentration of a catalyst, nand m are formal reaction orders and kJ and
k2 are the rate constants of the catalyzed or auto catalyzed reaction. The chem-
istry of the reaction is thus reduced to a simple rate law containing catalytic and
autocatalytic steps.
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 experimental results available at that time. However, solvent-free reactions can
only be described by this equation up to conversion degrees of about 0.6, which
is 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 completely lost (n and m as fitting parameters alone, sometimes
even as temperature dependent ones: Ryan, Dutta, 1979; Chung, 1984; Keenan,
1987; non-verified assumption of diffusion control at higher conversions: Barton,
1980, 1985; Huguenin, Klein, 1985). If the detailed chemistry of the complex
process is unimportant, the value of such 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 Sourour-Kamal equation is frequently
used to describe successfully quite different reactions in polymer chemistry.
Three general methods have been developed for the determination of kinetic
parameters, although nearly innumerable varieties were developed:
a. The direct method uses the procedure of Borchardt, Daniels, 1957, which
needs both the concentrations (in DSC, the degrees of reaction) and the cor-
responding reaction rates:
~; = k(T) f(a) = A exp (- :;) f(a)

To evaluate simple rate laws of the form
f(a) =(1- a)n or more exactly: f(a) = CS- 1 (1- a)n
it is especially useful to proceed from the logarithmic form:
In -(da) =In A + n . In (1 - a) - RT
dt
EA
- + (n - 1) . In Co
Today multiple non-linear regression software from different calorimeter manu-

facturers allows to estimate the best values for n, InA and EA by evaluation of
a single data set from one scanning experiment. The risk of misinterpretations
is minimized if several data sets with different measuring conditions are
simultaneously evaluated. Originally developed for n-th order reactions, at pres-
ent various choices are offered for f(a) (simple rate laws, rate laws including
autocatalytic steps, typical rate laws for the kinetics of reactions in heterogene-
ous systems). The evaluation presupposes isokinetic behavior (no changes of the
mechanism within the examined temperature range). Traditionally this was
proved using somehow created "reduced curves" for isothermal scans and
subsequent comparison with master curves. For instance, all experimental time
values for certain conversion degrees can be scaled by the reduced time factor,
tred = t/ ta (ref) , where ta (ref) is the time for a reference conversion degree, frequent -
ly a = 0.5. If we have isokinetic behavior, all experimental plots of a against tred
should fallon one single curve. The comparison with theoretical master curves
for all known rate laws enables conclusions concerning the probable rate law. But
even in the case of single step reactions the differentiation between various reac-
tion models is sometimes difficult on the basis of a graphical comparison, espe-
cially using erroneous real measurements. Therefore, using the present possibil-
ities of computerization and evaluation methods, the determination of the most
probable rate law should be supported by statistical methods. The F-test yields
the information which reaction model has the highest fit quality and whether the
best reaction model is significantly better than other models. Modern variations
of the Borchardt-Daniels procedure also allow the evaluation of nonisothermal
measurements and avoid therefore the main disadvantage of isothermal runs,
the not exactly known initial phase of the reaction (see Sect. 6.3.3).
b. If the known rate law is used in its integrated form (the integral method)
a 1 t
f - d a =g(a) = k(T) f dt
o f(a) 0
only a few conversion/time data points are needed to calculate k(T) for
isothermal measurements leading immediately to the activation parameters
for a single temperature scan. Evaluation of isothermal experiments gives no
problems when the rate law is known. For scanning experiments using
a 1 A
f - d a =g(a) =- f exp - -
T (EA ) dT
o f(a) f3 0 RT
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. A review of numerous varieties using integral methods can be found in
the literature (e.g., Hemminger, Cammenga, 1989; Galwey, Brown, 1998). The
importance of integral procedures has decreased considerably with develop-
ments in computer technology.
c. Likewise, many varieties of differential methods are known. For instance,
Freeman, Carroll, 1958, used the logarithmic equation for n-th order reac-
tions in a form based on differences between neighboring data points
(da)
~ In dT = n . ~ In (1 - a) -
EA ~ (T1 )
R'
A graph of ~ In (da/dt) versus ~ In (1- a) (for constant increments of liT)
yields the reaction order and activation energy. This analysis is very sensitive to
experimental errors and this is particularly true for the procedure of Ellerstein,
1968, in which the last equation is differentiated once more.
It must repeatedly be remembered that the analysis of a single curve is prac-
tically useless. Moreover, preconceived ideas for a certain reaction mecha-
nism can often be confirmed by data fitting of single heating-rate data over a
particular range of the degree of reaction.
Isoconversional Methods
As already outlined, activation parameters and rate law are inseparable coupled
as "kinetic triplet". Vyazovkin, 2001a, has shown for single step reactions that
the activation parameters dramatically depend on the chosen model. Otherwise,
the majority of chemical reactions proceed as multi-step reactions. If a series
of non-isothermal measurements at different heating rates is carried out, the
activation energy can be obtained without having to specify a certain model.
Therefore, these methods are also called as model-free or model-independent
methods, although this description is probably somewhat misleading. To obtain
the preexponential factor, the rate law must be known. The only requirement for
isoconversional methods is the independence of the model f (a) from the degree
of reaction a. The mathematics of isoconversional evaluations is derived from

the basic types of differential and integral methods.
Using the differential isoconversional method according to Friedman, 1964,
In ( -da) = In (A . f (ai)) - -EA . -1

dT a=ai R Ti
In (daldt)a= ai is plotted versus lITi at different heating rates fJi' If the investi-
gated reaction is a single-step reaction, the lines for the different degrees of re-
action are parallel and have slopes (-EAIR) and intercepts In [A f(a)]. Other-
wise, the slopes are different. The preexponential factor can only be obtained if
f(a) is known.
The second frequently used variety, the integral isoconversional method, was
independent from one another developed by Ozawa, 1965 and Flynn, Wall, 1966.
In the equation
A. EA)
lng(a) =In ( -R- -lnfJ+lnp(z)
is p (z) (with z = EAIRT) an approximation for the so-called exponential integral.

By rearranging and using an approximation for p (z) (Doyle, 1962) one obtains:
A. EA) EA 1
In fJi = In ( - - -lng(a) - 5.3305 + 1.052 - . -
R R ~
Again, the plot In fJi versus lITi for constant a/s yields parallel lines for a single-
step reaction. EA results from the slope, A from the intercept if f(a) is known.
The performance of modern computers allows to calculate much better approx-
imations for all values of EAIR then that used by Doyle.
If the evaluation shows a significant dependence of the activation parameters
on the degree of reaction, this is an unmistakable sign that the reaction mecha-
nism has at least two elementary steps. Such behavior is very frequently found
for condensed phase reactions. The relative contributions of these reactions
to the overall reaction rate vary with temperature and the effective activation
energy varies with the extent of conversion. Vyazovkin et aI., 1996 a to 2001 b, has
therefore developed a variety of model-free evaluations, which no longer pre-
supposes the independence of the modelf(a) from the degree of reaction a but
rather the opposite, a conversion-dependent overall activation energy. He calcu-
lates for a set of n experiments carried out at different heating rates the mini-
mum of the function
i i I(EA,a, Ti(ta))
i=1 jn I(EA,a, Tj(ta))
with
I(EA, Ti(ta)) == f exp (-~)

o RTi(t)
dt
Assuming the simple additive superposition of the individual reactions for a

possible multiple-step mechanism, he obtains conversion-dependent formal
activation energies. Then it is also possible to obtain predictions for isothermal
conditions without the knowledge of both reaction model and pre-exponential
factor, no doubt an important advantage for the chemical engineer. Vyazovkin,
2000 b, also tries to interpret the conversion -dependent activation energies from
a mechanistic point of view but he modifies himself this aspect when he formu-
lates (Vyazovkin, 2000a): "However, one should not forget that the mechanistic
clues are not yet the reaction mechanism, but rather a path to it that can further
be followed only by using species-specific experimental techniques". Sewry and
Brown, 2002, emphasize this aspect on formulating: "Model-free methods of
kinetic analysis postpone the problem of identifying a suitable kinetic model
until an estimate of the activation energy has been made". However, in the mean-
time a number of problems show that Vyazovkin's procedure has fundamental
limits in spite of all advantages of the model-free analysis (Opfermann et al.,
2002). Vyazovkin and Lesnikovich, 1990, have also suggested a scheme that, using
the function EA = f(a), allows some conclusions regarding complex processes.
Another model-free evaluation method, the so-called non-parametric kinetics
method was recently suggested by Serra et aI., 1998 a, b.
Kinetic Evaluation Using a Formal Reaction Model
The description of the reaction behavior by a formal-kinetic model (Opfermann

et aI., 1995; Flammersheim, 2000; Flammersheim, Opfermann, 2001, 2002; Opfer-
mann et aI. 2002) is, at present, the most flexible, simple and fastest solution
concerning the fitting of reaction curves. Using the vocabulary and basic ideas
of usual reaction kinetics, the overall reaction is described by the combination
of formal reaction steps (independent, parallel, competitive or consecutive) with
constant activation parameters. The single steps can be described by all known
rate laws for homogeneous and heterogeneous reactions. Up to 16 curves ob-
tained at different heating rates and/or temperatures are included simultane-
ously in the non-linear regression. Using statistical tests the formal model is
searched which describes the experiment as well as possible. An interpretation
of the single steps and their parameters should be made very cautiously, if at all.
As a rule, the formal steps of the model do not correspond to real elementary
reactions. In other words, a high value of the fitting quality for the best model
does not mean that the kinetic description (Roduit, 2000) is correct in the
physicochemical meaning! But the industrial practitioner is usually more inter-
ested in predictions of the reaction behavior for arbitrary temperature-time
profiles, than knowing or understanding the real mechanism. As with the model-
free analysis, the knowledge of the composition of the reacting system is not
necessary. The obtained results allow the prediction of reaction degrees, heat
flow rates and relative concentrations of the formal reactants versus tempera-
ture or time for any complicated temperature/time profiles.
Like all other evaluations discussed so far, disadvantages result from the fact
that the obtained process parameters and the derived conclusions are valid only
for the tested composition of the reaction mixture. To be more exact, there are
two disadvantages: technically important reaction mixtures could have different

amounts of catalysts, non-reactive fillers or solvents; and, in addition, if the
stoichiometric composition of the reaction mixture cannot be varied, then im-
portant additional information concerning the verification, acceptance or rejec-
tion of an assumed reaction model is not available. This is true for such cases, in
which one or several elementary steps of the mechanism are influenced only by
a specific reactant.
Search and Use of the Real Mechanism
A new evaluation program (Flammersheim, Opfermann, 2002) uses the same

basic strategies and mathematical routines as during the kinetic analysis with
a formal-kinetic model. The crucial difference is that the individual steps of
the assumed mechanism are now treated and formulated as real elementary
steps. Even if the reaction mechanism is not too complex, such evaluations can
be done only if the experimental information is adequate. Therefore, besides
having a sufficient reaction window (a sufficient number of isothermal reaction
temperatures and/or heating rates) the initial composition of the reaction mix-
ture should also be varied as widely as possible. The new program is somewhat
more time-consuming (it was not developed to replace faster formal-kinetic
evaluations) but it is a valuable supplementary tool for kinetic evaluations. As a
rule, results from other analytical methods must be considered in defining an ad-
equate kinetic model, even if only the relevant elementary steps of the assumed
mechanism are considered.
Important features are:
- Unrestricted combination of elementary steps (consecutive, parallel, equilib-
rium).
- Simultaneous evaluation of both isothermal and non-isothermal measure-
ments.
- Simultaneous evaluation of curves for mixtures with variable compositions
and differing amounts of reactive (catalysts) and non -reactive (fillers, solvents)
substances.
- Consideration of changes in volume during the reaction.
- Test of significance regarding the fitting quality of different models and re-
garding an assumed number of reaction steps.
- Predictions for signal, fractional reaction and reactants for any temperature-
time program and for any composition of the reaction mixture.
The important steps of the new program are:
- Input of up to 16 DSC curves, selection of the base line and optional desmear-
ing and smoothing.
- Input of the assumed mechanism with all (important) elementary steps
(maximum 12 steps). Using symbolic letters or short names for the reactants,
the reaction mechanism can be formulated as usual in chemistry. Reversible
reactions with very high rate constants for both forward and backward reac-
tions can be formulated as equilibrium reactions.
- Input of the molar masses, specific densities, molar concentrations or initial

weights of all reactants, including those of a possible solvent. Molar masses
and densities are valid for all curves of the chosen model, while the stoichio-
metric composition of the reaction mixture can be different for each curve.
The program checks the mass balance.
- Input of useful initial values for all variables of the kinetic scheme: for ele-
mentary reactions log A, EA and Qr, for equilibrium reactions L1r Hand L1r s.
- Use of the differential equation solver, which is a combination of embedded
Prince-Dormand procedure of 4/5 degree with a Gear4 procedure (see Engeln-
Mtillges, Reuter, 1996). In the Prince-Dormand procedure two checks regarding
stiffness have been integrated. The calculation always starts with the Prince-
Dormand procedure. If both checks regarding stiffness are positive, then the
calculation continues using the Gear4 procedure. There are three criteria to
terminate the iterations. The first criterion is an inherent part of the algorithm
and causes the iterations to finish automatically, if the relative change of all pa-
rameter values is lower than a defined precision. User-defined criteria are the
number of iterations or a maximum value of the correlation coefficient.
- Output of all optimized parameters and their errors.
- Predictions of the reaction behavior can be made for any complex tempera-
ture-time profile using the favored model and its parameters. These predic-
tions concern the concentration behavior of all reactants and products, the
DSC signal and the degree of reaction.
6.3.6
Selected Examples and Possible Predictions
1. Example: Simple Test Reactions
Photochemically produced molten cis-azobenzene converts into the trans form

above room temperature in a well-defined 1st order reaction. The activation
parameters have been found to be In (A/S-i) = 27.7 and EA = lO3.6 kJ mol- i
(Eligehausen et al., 1989). By contrast with the situation for temperature and
caloric calibration, there are so far no internationally recommended test reac-
tions for kinetic investigations. 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 (Eckardt et al., 1998).
The conversion of cis-azobenzene is only a 1st order reaction in the liquid
phase. Before each measurement the solid substance (Tfus = 71.6 0c) must be
rapidly melted and quenched to the starting temperature (35C). 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. However, this is not possible because the liquid cis-azoben-
zene cannot be supercooled 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 approximation because the differences between a calculat-
ed sigmoidal baseline (from literature cp-values) and the straight line are very
100
Wig
2.25
~
t
<11
1 K min- 1
2 K min- 1
0
~ 4 K min- 1
~ 1.50
~ 8 K min- 1
0
.c;
<11
15 K min- 1
....
....
<11
0.75
0..
11"1
o
a 50 71 92 113 134 O( 155
temperature
1.00
Wig
1 '.75
..::!.
c
~ 0.50
o
370 K
c
'"
.c::
~ 0.25
'"
0-
Il>
o
b o 30 60 90 min 120
time ~
Fig. 6.1Sa, b. Measured (solid) and calculated (dashed) 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
small. An approximate (because of different thermal conductivities) but ade-

quate desmearing is possible using the crystallization peak of molten trans-azo-
benzene as the apparatus function. The correction is small because the isomer-
ization is rate-determining even at the highest used heating rates. A precise tem-
perature correction to zero heating rate values for all heating runs is very im-
portant.
Figure 6.15 shows the excellent agreement between experiments and calcula-
tions 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 calorime-
ter 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 K min-I). Assuming a fixed reaction order
of n =1 the procedure gives activation parameters of In (Als- I ) = 27.6 0.2 and
EA = (l02.5 0.8) kJ mol-I, in good agreement with Eligehausen et al., 1989.
Another example is the dimerization of cyclopentadiene (Flammersheim,
Opfermann, 1999b). The monomer (CPD) is easy to obtain from the commercial
dimer (DCPD) by cracking distillation at 150-170 DC and repeated distillation at
40 DC in an argon stream. The mechanism according to the following scheme is
slightly more complicated (Turnbull, Hull, 1968):
2CPD ~ DCPD (l)
2 CPD + DCPD ~ 2 DCPD (2)
DCPD ~ 2CPD (3)
The second order Diels-Alder dimerization is auto catalyzed and the back reac-
tion cannot be neglected above 140 DC. But this reaction should be suitable to test
the efficiency of the used kinetic software. Using the program "Component
kinetics" and taking into account the volume contraction during the reaction
(~CPD = 0.7970 g cm- 3, (JDCPD = 0.9770 g cm- 3 ), the best activation parameters are
shown in the Table 6.l.
Table 6.1
Reaction Parameter Optimized value
1: 2 CPD ~ DCPD 19 (Ails-II mol-I) 5.74 0.15

EA.I/(kJ mol-I) 69.43 1.00
2: 2 CPD + DCPD ~ 2 DCPD 19 (A 2/s- 1 L2 mol- 2) S.SO 0.30
EA 2 /(kJ mol-I) 73.79 2.40
3: DCPD ~ 2 CPD Ig (A3/s-l) 13.04 0.25
EA 3 /(kJ mol-I) 142.0 2.2
correlation coefficient 0.9997
Fig.6.I6. Concentration 12
profile for the dimeriza-
tion of CPD (heating mol l-1 O[
)
rate: 10 K min-I). 1: CPD,
2 and 3: DCPD formed 8
according to the simple 120
2nd order reaction (1) and 6
-
c
the autocatalytic reaction (2) ~
80
of the reaction scheme
0
<- 4
(see text) -'C
<lJ
u 2 40
c
0
u
0 o
0 4 8 12 16 min 20
time - - -
Using these results, the heat flow rates, conversion degrees and concentra-
tions of the reactants can be predicted for any temperature-time reaction pro-
files. As an example, Fig. 6.16 shows the concentration profile for a heating rate
of 10 K min-I. The program allows to differentiate between CPD formed during
the two reaction paths (1) and (2). With increasing temperature, in the case of
f3 = 10 K min- I above 185C, dominates the back reaction (3) for the overall
reaction. But the two forward reactions show a different behavior. While the
amount of CPD formed as a result of reaction (1) already decreases above 160C,
the percentage of the autocatalytic reaction (2) is still increasing. Another result
is important. Every nth-order reaction starts with maximal rate. Using the pa-
rameters of the table we can calculate for instance the storage periods for which
a predefined degree of conversion is not exceeded. In our example and for
a =0.02 these times are nearly 4 d at - 20C, 9.5 h at OC and only 72.3 min at
20C. That means, the dimerization of CPD should always be investigated with
freshly distilled monomer immediately after its preparation.
2. Example: Radical Polymerization

of (2,2-Dimethyl-l,3-dioxolan-4-y/)methyl Methacrylate (DOMA)
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-radicals is independent of the chain length, approximately constant rad-
ical concentration during the whole reaction, bimolecular chain termination
reactions). Under these circumstances, the rate of concentration change is pro-
portional to the concentration of the monomer (CM) as follows:
dCM
---=kpro
(r.
--
kini)"2
CI
1/2
'CM
dt kter
The quantity r is the yield of radicals due to the thermal homolysis of the ini-
tiator (I) [e. g., 2,2'-azodiisobutyronitrile (AcBN) or dibenzoyl peroxide 1, k pro ,
kter and kini are the rate constants of the chain propagation, chain termination
and initiation respectively. If the thermal initiator is replaced by a photoinitia-

tor, like benzoin methyl ether or 2,2' -dimethoxy-2-phenylacetophenone (benzyl
dimethyl ketal), the analogous overall rate law reads:
dCM ( r labS) 112

---= kpro ' - - 'CM
dt kter
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. 6.13), monofunctional methacrylates (methyl,
ethyl, propyl, 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 polymerization). After this stage the so-called gel or Trommsdorff-
Norrish effect dominates, this is recognizable by increasing reaction rates
despite decreasing monomer concentrations (Malavasic et al., 1986). This is
also valid (Fig.6.17) for the polymerization of {2,2-dimethyl-l,3-dioxolan-
4-yl)methyl methacrylate (DOMA) (Flammersheim,Klemm, 1985). There are no
qualitative differences between curves 1 and 2. In both cases the change of reac-
tion rate with time is small enough during the entire reaction that desmearing
is not necessary. Spectroscopic investigations 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 re-
action of 0.3 the rate law is 1st order. Knowing the activation term {rk ini CI )1/2 or
{rlabs )112, the overall rate constant gives the ratio kpro / k~~;. During the gel effect
this ratio is formally dependent on the degree of reaction reached. In contrast to
the nearly unhindered chain propagation reactions of the small monomer mol-
ecules, the chain termination reactions of the growing macroradicals are in-
6.0
mW
4.5
t 3.0
~
0
.....
~
....
0
1.5
0
cu
.J:
0
0 10 20 30 min 40
time ~
Fig.6.17. Light-initiated (1) and thermally activated (2) polymerization of DOMA. (curve 1:
A = 365 nm, 10 = 0.35 mW cm- 2, T = 298 K, 1 % 2,2-dimethoxy-2-phenylacetophenone as pho-
toinitiator; curve 2: T = 343 K, 1% 2,2-azodiisobutyronitrile as initiator)
20
mW
15
illumination period dark reaction

o~------~--------~------~------~--------~
o 0.6 1.2 1.8 min
time ..
Fig.6.18. Photopolymerization of DOMA with interruption of the irradiation 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 origin of the time scale.
Solid lines: measured curves, dashed lines: desmeared curves (A= 365 nm,Io = 3.5 mW cm- 2,
T = 298 K, 1% 2,2-dimethoxy-2-phenylacetophenone as photoinitiator)
creasingly hindered in the more and more viscous matrix. The result is a pro-
nounced dark reaction after switching off the light during the polymerization
(Fig. 6.1B). 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 O.B (curve 2). To get a better comparison
the second curve is shifted to the origin of the time scale in Fig. 6.1B.As a result,
the relative extent of the dark reaction is proportional to the degree of reaction
at that moment at which the light was interrupted. The rate of the dark reaction
(missing activation by Flab.) is described by:
CM kter CMo
---=-t----
dcM/dt kpro dCM/dt
If the ratio kpro Ik~~; is determined from the measured curve immediately before
the light is interrupted and, following this, the ratio kterl kpro from the dark reaction
(last equation) the rate constants kter and kpro 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 meaningless 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 kterl kpro ratio). The picture shows
this effect very clearly. The linear relation between -cM/(dcMldt) and t (see the
last equation) is only found from the desmeared signal (kterlkpro 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: Formal-Kinetic Evaluation of a Polymerization Reaction

with Partial Diffusion Control
The curing reaction of a solid mixture of a commercial epoxy (RUETAPOX VE

3579 or 2,2',6,6' -tetrabromobisphenol-A diglycidyl ether, Tg: 25 DC, used without
purification) with 5 wt.-% of a latent accelerator [Zn(OCNh(iMeImidh, crys-
talline powder, self-prepared] is initiated by the release of the 1-methylimidazole
above 70 DC (Flammersheim, Opfermann, 1999a, 2001). Although the mecha-
nism is complex, the measured curves can be fitted by a relatively simple, formal
model. But the kinetic evaluation is complicated because, with increasing reac-
tion degree, a partial or complete vitrification of the reaction mixture - de-
pendent on the heating rate - is observed. The resultant, complex shape of the
baseline is directly available from TMDSC measurements. Figure 6.19 presents
the course of the heat capacity during the reaction for three heating rates; curve
1: 0.1 K min-I, curve 2: 0.2 Kmin- I and curve 3: 2.0 K min-I. Curve 4 corresponds
to the averaged heat capacity of the reaction product, calculated from all second
runs. The picture clearly shows the devitrification of the initial epoxide at 25 DC,
the heating rate dependent vitrification (decrease of cp ) and the second devitri-
fication (increase of cp) at temperatures higher than the Tg, max of this system
(163 DC). In other words, at low heating rates the glass temperature of the reac-
tion product increases faster than the program temperature. Comparable results
were obtained by van Assche et al., 1995, 1996, 1997, for other vitrifying systems.
After partial or complete vitrification of the reaction mixture the reaction is no
longer controlled by the kinetics of the chemical reaction, but by diffusion
processes. Finally, if the program temperature exceeds the maximum glass tran-
sition temperature, one again enters the realm of chemical reaction control.
Figure 6.20 shows the isothermal reaction at 95 DC. The range of the isothermal
vitrification is clearly displayed by the decrease of the heat capacity (between
Fig.6.19. Specific heat 1.8
t
capacity change during cur-
ing of RUETAPOX VE 3579
at a heating rate of 1: 0.1, JK-' g-'
2: 0.2 and 3: 2.0 K min-I; >..
4: heat capacity of the reac- --
a 1.4
'-'
tion product Cl-
au
~ 1.2
'"
OJ
.s:::
'-'
OJ
CI-
Vl
0 50 100 150 O( 200

temperature ----
Fig. 6.20. Specific heat

capacity and reaction heat
flow rate for isothermal cur-
ing of RUETAPOX VE 3579
at 95C; the vitrification
process is clearly visible CL
<:> 3
co
by the decrease of the heat
14
capacity between the two
-04 -;;
vertical lines QJ
..c:
1.3
'"en
CL Cp
QJ
CL
1 2 '----------'---'---------'----'-----'-------' - 0 6 en
o 50 100 150 min 200
tim e -------
the both vertical lines}. But this is not coupled to a noticeable sudden increase of
the reaction rate! On the other hand, the picture indicates definitely a feature of
the reaction mechanism. One or more reaction steps are autocatalyticaI.
The kinetic evaluation must be capable of taking into account the more or less
complete change from chemically- to diffusion-controlled reaction steps. The most
frequently used possibility goes back to a formulation given by Rabinowitch, 1937:
1 1 1
-=-+--
k kdiff kchem
As usual, the temperature dependence of kchem is described by the Arrhenius
equation. Several approaches for kdiff (Huguenin, Klein, 1985; Deng, Martin, 1994;
Karkanas et aI., 1996; Wise et aI., 1997; Liu et al., 2001) are discussed in the liter-
ature. Van Assche et aI., 1995, 1996, 1997,2001, were able to show that for epox-
ide curing reactions the molecular motions frozen out at vitrification corre-
spond to the molecular motions needed for the chemical reaction. They define
for each run a mobility factor (heating rate and temperature dependent)
between 1 and 0 which is directly accessible from the heat capacity signal of
TMDSC measurements. This factor parallels the diffusion factor, defined as ratio
between the theoretical reaction rate without diffusion hindrances and the
actual reaction rate. The calculation of theoretical rate and diffusion factor
presupposes the knowledge of the correct reaction model, in general surely a
non trivial problem. If the correspondence between main glass transformation
and mobility of the reactive species can be assumed (this is not always observed,
see van Assche et aI., 1997,2001), it should be favorable to use an equation which
describes the diffusion hindrances independent of the chemistry of the process.
This is also advantageous for another reason, predictions of the reaction behav-
ior for any conditions require a model for both kchem and k diff A usable approxi-
mation is - despite of all criticism regarding the region of validity - a so-called
modified WLF equation as proposed by Wise et aI., 1997:
kdiff (1j is related by this equation to Tg(a). A simple and convenient regression
function for a number of measured Tg and a values obtained from partially
cured samples was proposed by Hesekamp, 1998:
Tg(a)
g!.
=Tg(O) . exp ( - -
a)
g2- a
For the investigated example were found: Tg(O) = 296.3 K, g! = 4.48 and
g2 = 12.26.
The general limitation of applying any types of WLF equation is based on the
fact that they should be used only for the liquid state above Tg Below Tg , one
obtains values for kdiff(T) which are increasingly too low. But it is reasonable
(Flammersheim, Opfermann, 2001) to assume:
- for T ~ Tg the WLF-equation is valid in the given form
- for T < Tg an Arrhenius type equation is used:
Both equations fulill the boundary conditions that kdiff and its first derivative are
continuously at T = Tg The C-parameters of the modified WLF equation should
not be discussed, as long as the true reaction mechanism is not known. Using
the formal-kinetic evaluation it is not clear how certain reaction steps are influ-
enced by the changing segmental length, van Assche, 1995, 1997,2001, during
the freezing of the reactive species.
The most simple formal-kinetic model that produces a practically perfect fit
for all heating rates (correlation coefficients> 0.9995) corresponds to a process
with two consecutive partial steps:
A~B~C
The corresponding rate law for the formal component B is:
dt
dCB
= A! . exp ( - EA'!)
RT . CA - (EA'2) . CB2 (1 + K
A2 . exp - RT cat ,2 cd
The decrease of the formal component A and the increase of the formal compo-
nent C correspond to the negative first and second term of this equation respec-
tively. Two data sets were available for the kinetic evaluation, each with 8 differ-
ent heating rates. The fit to the measured curves is perfect. The second reaction
step is auto catalyzed by the formal reactant "c" and this is taken into account by
the parameter Kcat,2 in the above equation. Only this reaction step is influenced
by the diffusion. If, as a trial, the diffusion control is not used for this step, no ac-
ceptable fit is reached. The model parameters for both data sets are summarized
in Table 6.2.
Table 6.2
Parameter Data set 1 Data set 2
19 (A 2 /s- 1) 7.94 1.05 8.90 1.12

19 (A1/s- 1) 6.48 0.70 6.20 0.86
EA,l/(kJ mol-I) 70.0 5.2 67.5 6.6
EA ,2/(kJ mol-I) 79.0 8.2 86.7 8.7
19 Kcat,2 0.826 0.170 0.788 0.170
19 (Kdiffusion,2/mol-1 m 3 S-l) -4.77 1.05 -4.39 0.78
CdK 9.9 3.1 8.1 2.1
C2 /K 25.9 21.8 28.6 19.4
Contribution of consecutive reaction 0.25 0.06 0.21 0.06
Correlation coefficient 0.9997 0.9996
Using these values, the course of the isothermal curing was correctly predict-
ed. Further, at low temperatures, the conversion degrees never reach "1" after an
extended period due to partial freezing. The corresponding values are correctly
predicted for different temperatures (110 C: calculated 0.58, found 0.57 and
0.61; 140C: calculated 0.80, found 0.79 and 0.79; 170C: calculated 0.94, found
0.93 and 0.94). Finally, a convincing argument for the validity of the selected
model is given in Fig. 6.21 for an experiment with a very slow heating rate,
0.25 K min-I. The heat capacity function that is directly accessed by experiment
(cf. Fig. 6.19) runs completely parallel to the increase of the glass temperature
and the degree of reaction calculated by means of the model parameters. This is
valid both for the vitrification at about 100C and the slow devitrification start-
ing at about 130C. This temperature range becomes larger at even lower heat-
ing rates. At heating rates faster than 1 K min- I Tg permanently lags behind the
program temperature, as the diffusion hindrances due to the partial freezing be-
come more and more unimportant.
18 200 10
tJK- 1 g-l O[ 08 t
>--
~
d
t
OJ
06
c
0
CL
14 100
~
d d
:::>
:=':'
-;:;
OJ

12
/
Il 50
~
~
OJ
CL
E
0.4
"0
a2
OJ
,,' T'!,./i OJ <lJ
:.....---- .
~
~
OJ en
CL
._cx./
OJ
Vl 'D
----
200 400 600 min 800
lime
Fig.6.21. Heat capacity, glass transition temperature and degree of reaction as function of
temperature and time for the curing of RUETAPOX VE 3579, heating rate: 0.25 K min- 1
Fig. 6.22. Calculated tem- 100
t
perature profile for the rate
controlled reaction of
RUETAPOX VE 3579 at a
% DC
t
c:
constant conversion rate of !:!
0.1 % min- 1 (i.e., a is propor- '"<- 60
OJ 150
tional to time; Tg parallels >
c:
<U
'-
-
::>
the reaction temperature
0
40
ec..
~
'-'
within a large temperature 0

T9 100 OJ
range) OJ
OJ 20 E
OJ
<-
""
<U
50 -
"0
0
time - - -
Another kind of prediction is the temperature respectively the heating rate

profile for a so-called rate controlled reaction. This is very important if in cer-
tain temperature ranges the change of any physical properties coupled with the
increasing degree of reaction is of primary technological interest. This could
be for example the thermal shrinkage during the production of ceramics. In
Fig. 6.22 it is assumed that for the epoxy curing a constant rate of conversion of
0.1 % min- 1 is desired. The result of such calculations for the chemical engineer
is a table of temperature intervals and their incident heating rates, possibly in-
cluding isothermal reaction steps. An interesting but surely not surprising result
(cf. Fig. 6.21) is that over large temperature intervals the program temperature
parallels the glass transition temperature.
Finally, a useful framework for understanding the changes which occur dur-
ing the cure of a thermosetting system is a Time-Temperature-Transformation
(TTT) diagram. Some examples including possible varieties were presented by
Wisanrakkit, Gillham, 1990, van Hemelrijck, van Mele, 1997 and Theriault at al.,
1999. The diagram in Fig. 6.23 is calculated for the isothermal cure as function
of time and curing temperature. The vitrification curve (Tg = Teure , the solid,
thick line in Fig. 6.23) separates the liquid range without diffusion hindrances
(left) from the glassy range (right) with increasing diffusion hindrances for the
chemical reaction. Sometimes, the practitioner needs the reaction behavior for
a number of constant differences between Tg and Teure , as shown by the solid,
thin lines in Fig. 6.23 for Tg - Teure = 10 and 20 K. The dashed curves show for a
number of curing temperatures the evaluation of Tg against 19 (time). Eventually,
also isoconversional curves (the dotted curves in Fig. 6.23) can be calculated for
a number of conversion degrees, in the example for 0.3,0.6,0.8,0.9,0.95 and 0.99.
It is also possible to construct a so-called Continuous-Heating-Transformation
(CHT) diagram (van Hemelrijck, van Mele, 1997), at which 19 (t) on the x-axis is
replaced by 19 (P>. Both types of diagrams enable a fast and rough assessment of
the reaction behavior of the system within a wide range of experimental condi-
tions. The common disadvantage is sometimes the limited practicability for real
quantitative predictions, because true isothermal or true heating processes with
constant heating rate are difficult to realize in technical processes. In such cases,
180
O( 19 (100%)
140
w
'-
100
:::J
o
....
~ 60
E
w
10~
time
Fig.6.23. Time-Temperature-Transformation (TTT) diagram for the isothermal cure of
RUETAPOX VE 3579 showing the increase of Tg at different curing temperatures (dashed),
constant conversion lines (dotted) and two curves with Tg - Teure = constant (thin solid lines),
the thick solid line corresponds to the vitrification line Tg - Teure = 0
the calculation of the reaction behavior (degree of reaction, Tg as function of a,

heat production) for any concrete preset and technically feasible temperature-
time profiles (cf. Fig. 6.21) is the better and more reliable way.
6.4
The Glass Transition Process
6.4.1
The Phenomenology of the Glass Transition
Qualitative Cp measurements intended to provide information regarding the

glass temperature are much commoner than direct determinations of the tem-
perature dependent heat capacity (cf. Sect. 6.1). To find the best experimental
procedure and an unambiguous interpretation of the results is, however, 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.
Typical values for the temperature ranges of the transition are 10 K for low-
molar mass glasses, 20 to 50 K for most of the organic polymeric glasses and
100 K or more for silicate network glasses. The glass transition is characterized
by an appropriately defined glass transition temperature Tg Within the transi-
tion region many macroscopic properties, which may have great practical im-
portance, change their values (viscosity, dielectric and especially mechanical
6.4 The Glass Transition Process 201
3.0
( 0
B
mW hea ting run
~
r 1.5
"
.....
~
~
"
'"
.c
cooling run
o ~ ______ ______ ________L -_ _ _ _ _ _

~ ~ ~ ______ ~
40 55 70 85 'C 115
temperature ..
Fig.6.24. Typical DSC curves of amorphous polymers in the glass transition region.
A: the glassy non-equilibrium state, B: glass transition region, C: "enthalpy relaxation" peak,
D: the equilibrium liquid state (sample: linear epoxide-amine polyadduct, heating and cool-
ing rate: 5 K min-I)
properties). Each of these could form the basis of an experimental determina-

tion of Tg With DSCs the glass transition is detectable by a step change of the
heat capacity /}.cp on heating or cooling. Both the temperature and magnitude of
this event are important. Observed values of /}.cp range from 0.1 to 2 J g-l K- 1
Figure 6.24 shows a typical DSC curve for a linear, high-molar epoxide-amine
polyadduct which was first cooled at the rate shown and then immediately re-
heated.
For polymeric glasses the temperatures of the transition range from - 50 to
300C 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 as least as important as is the melting tem-
perature.
6.4.2
The Nature of the Glass Transition and Consequences for DSC Measurements
Problems result from the fact that the glassy (or vitreous) solid is, thermo-
dynamically, 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 de-
fined second order transition. Only the liquid (or rubbery) state at the high tem-
perature 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 sys-
tem 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. 6.1) 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-de-
pendent, and classical thermodynamics is no longer valid. The system passes
through a sequence of non-equilibrium states during heating or cooling. The typ-
ical 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 distri-
bution of the intrinsic variables which vitrify (or devitrify) over a wide range.
A special problem when investigating the glass transition (key word: relax-
ation 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 his-
tory. This is important not only for theoretical investigations but also for practi-
cal DSC measurements of the glass transition.
Conclusions:
- In contrast to the measurement of equilibrium transitions, it is not possible
to get "equilibrium" values of the characteristic quantities Tg and /:"cp byex-
trapolation 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 magni-
tude is generally 3 to 5 K, whereas it is 15 to 18 K for the considerably stiffer
borosilicate network glasses. By contrast /:"cp shows much less dependence on
thermal history.
- Characterization of glasses by DSC measurements (Fig. 6.24) is mostly car-
ried out in the heating mode because this is practicable even for those heat
flux DSCs having relatively sluggish furnaces. A heating run is also advanta-
geous when it is important to characterize the "as received" glassy state (e. g.,
resulting from particular cooling or annealing procedures or chemical reac-
tion). It must be remembered that, on heating, the original glassy state can
change at temperatures as much as 50 K below the transition region (for in-
stance, if the heating rate is slower than the previous cooling rate). Tg then de-
pends on the heating rate. To sum up, values for Tg (and to a lesser extent /:"cp )
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
coupling of vitrification and devitrification processes (occurring during heat-
ing) 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 lower than 5 K/min for a reliable determination of
the rather small changes of the heat capacity. This may sometimes be difficult
because of the limited cooling rate of many DSC furnaces.
- 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). Correc-
tion of these influences is important, because larger sample masses (10-20 mg)
and larger scanning rates (10 or 20 K min-I) are used in practice. For this pur-
pose 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 6.1.1 (Figs. 5.10 and 6.2).
2. Placing a small piece of indium both on the bottom and in the middle of
the sample and determining the difference between the extrapolated onset
temperatures from the In melting 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 al., 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 al., 1979).
A disadvantage of all of these corrections is that only an average lag is ob-
tained and thus can be corrected.
6.4.3
Definition and Determination of the Glass Transition Temperature 19
Conventional Glass Transition Temperature

The main parameters used to characterize the glass transition are shown in
Fig. 6.25. For many applications it is important to know the temperature range
Tg,i to Tg,f over which the substance vitrifies on cooling, or devitrifies on heat-
ing. Unfortunately, the practical determination of these temperatures is prob-
lematic, there are large errors in their definition and it is difficult to give clear
instructions for their measurement. The situation is improved if clearer, more
characteristic temperatures 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 Tg,1I2 related to the Cp change (the temperature at
which Cp is midway between the extrapolated heat capacity functions of the
glassy and liquid state). The use of the latter is more meaningful as this temper-
ature is better related to the second characteristic quantity of the glass transi-
tion, 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
cooled from the liquid at a fixed rate.
Fig. 6.25. Definition of the most frequently used conventional quantities for characterization
of the glass transition.
Tg, e: extrapolated onset temperature, Tg, 1/2: half-step temperature, !:J.Cp : Cp-change at the half-
step temperature, Tg,i and Tg,f: initial and final temperatures of the glass transition, Tg,f - Tg,i:
temperature interval of the glass transition
Tg,e and T g,1I2 can both easily be obtained from routine measurements, this is the
reason for the nearly exclusive use of these pragmatically defined glass transi-
tion temperatures up to now. A repeatability error of 1 K is acceptable in prac-
tice. For heating runs the following practical and frequently used procedure is
recommended:
- The sample is heated to a temperature at least 15 to 30 K above Tg
- 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 K min-I). These rates lead to rel-
atively 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 quantitatively
by comparison of the original run with the rerun under the same conditions.
Neither Tg,e nor T g,l12 make any allowance for the non-equilibrium nature of the
glass transition. This is especially striking (Fig. 6.26) if the glass transition is
accompanied by "enthalpy relaxation peaks". These appear on heating curves as
endothermic events at the high temperature end of the glass transition range.
mW
annealed glass
3 ~
~
t 2
c>
.....
~
0
c>
""
~
o
Tg lqul ~.l\lqU'Z: fg .la'
g.lIla'
60 70 80 90 100 O( 110
temperature ..
Fig.6.26. DSC heating curves for an annealed (or slowly cooled) and a quenched glass show-
ing the paradoxical result that the conventional glass temperature(s) of the quenched (qu)
glass seem to be lower than those of the annealed (a) glass, for details see text. Tg,e: extra-
polated onset temperature, Tg,lf2: half-step temperature
Their height may be comparable with those of the melting peaks of crystalline
materials (Petrie, 1972). For the example shown in Fig. 6.26 the glass annealed
70 h at 68e has Tg,e = 89.1 e, Tg,1/2 = 87.3e whereas the quenched glass has
Tg,e = 81.6e, Tg,1I2 = 85.1 e. This use of Tg,e and Tg,ll2 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 addi-
tion Tg,e and Tg,1I2 react to the thermal history in a different manner.
Tg,e and Tg,1I2 are not useful for theoretical treatments of the 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).
The Thermodynamically Defined Glass Transition Temperature
The "glass transition", although a kinetic phenomenon, can be unequivocally

defined thermodynamically. This uses the so-called fictive temperature Tg,fic=Tf
first introduced by Tool, 1946. Tf is a well defined quantity that reflects the cur-
rent structural state of the glass to be characterized.
The concept of the fictive temperature was considerably extended by
Narayanaswamy, 1971, Moynihan et al., 1976 and de Bolt et al., 1976. An analo-
H9 (-P d_______
1
>-
C1.
Hg(all~
c;
:E
c:
OJ H9 (-P21/
Hg(a21~
temperature ..
Fig.6.27. Schematic enthalpy-temperature curves for the glass transition of an amorphous
sample at different cooling rates (-Pi) and subsequent heating after isotherm annealing at Tai,
for details see text. Hg : enthalpy of the different non-equilibrium glassy states, HI: enthalpy
function of the equilibrium liquid state, Tg (- Pi): glass transition temperatures on different
cooling (11311 > 1132 D, Tg (ai) glass transition temperatures on heating after different annealing
schedules
gously defined "thermodynamic" or "enthalpic" Tg-temperature, based on the

specific information of a DSC, was introduced by Flynn, 1974 and Richardson,
Savill, 1975a and Richardson, 1976. In the following section this thermodynam-
ically defined temperature Tf is presented as Tg without the additional sufflx.
To understand the definition of Tg , the enthalpy versus temperature diagram
(Fig. 6.27) 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 rates (negative {3) during this process the lower are the vitrification tem-
peratures. The functions Hg (- {31) and Hg (- {32) characterize the enthalpy of the
respective 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 measurements (Gilmour, Hay, 1977) show slight differences]. If the glass
is annealed at temperatures down to, at most, 50 K below the vitrification tem-
perature, 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) HI-function (Petrie,
1972; Peyser, 1983; Cowie, Ferguson, 1986; Agrawal, 1989). The figure shows this
for two annealing temperatures Tal and Ta2 of a glass, which was obtained at
a cooling rate - Pl. At Tal the annealing time was sufficient to reach the equilib-
rium 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 de-
vitrify exactly when the enthalpy line of the glass crosses that of the liquid but
the transition from Hg to HI is not sharp. To determine the intersection of en-
thalpy 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 Tg Tool, 1946 called this tempera-
ture the "fictive temperature", because during heating nothing happens at that
point. Hence Tg cannot be located directly on the measured curve, instead, on
heating, 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 Tal). This is the
reason for the paradoxical values, mentioned earlier, for the pragmatically de-
fined Tg,e or Tg,1/2 temperatures, when comparing slowly cooled (or annealed)
and quenched glasses. Superheating ends only at temperatures well above Tg , the
return to the equilibrium curve is now rapid and produces the so-called relax-
ation 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; Richardson,
1976; Moynihan et al., 1976) is used, Tg can easily be calculated from DSC mea-
surements.
The procedure in question (Richardson, Savill, 1975a; Richardson, 1976) is
explained in Fig. 6.28. We start with the definition, the equality of hg(T) and
hI (T) at Tg. The enthalpies h (T) for the glass and liquid are obtained by integra-
tion of the corresponding cp-functions, which can always be approximated by
linear equations:
1
glass: Cp,g= a + b T and hg(T) = a T + - b T2 + P
2
1
liquid: Cp,l =A +B T and hl(T) = A . T + - B . T2 +Q
2
The integration constants P and Qare not known but Q-P maybe obtained from
the difference hl (T2 ) - hg (Tl ), the hatched area in the Fig. 6.28, a directly accessi-
ble experimental quantity:
1 2 1 2
hl (T2 )-hg(Tl ) =A T2 - a Tl +-B T2 --b Tl + (Q-P)
2 2
4.5
l/(Kgl
3.5
r>-
3.0
'u
CI
0-
CI
2.5
C P.l
-------------
~
-;;
cu
..c: 2.0
~
~
cu
'" 1.5
0-
1.0
60 90 100 T2 'C 11
temperature ..
Fig.6.28. Determination of the thermodynamically glass temperature from DSC heating
curves with "relaxation peak" (Richardson, Savill, 1975a; Richardson, 1976).
Cp,g and cp,l are the (extrapolated) specific heat capacity functions of the glass and the liquid,
respectively, the hatched area (extended to cp = O) corresponds to the difference hi (T2 ) -
hg (TI ), for details see text
TJ and T2 are convenient arbitrary temperatures that must be chosen to be be-

low and above the glass transition region (i.e., in the glassy and in the liquid
state, respectively). Using Q - P from the above equation the desired glass tem-
perature is obtained by solving the quadratic equation:
1
- (B - b) . T~ + (A - a) Tg + (Q - P) = 0
2
Figure 6.29 shows an equivalent, graphical procedure (Moynihan et aI., 1976) for
determining Tg From the enthalpy definition it follows that:
T2 T2
f (Cp,I(T) - cp,g(T dT = f (cp(T) - cp,g(T dT
Tg TJ
where cp(T) is the experimentally determined curve and Cp,g(T) and Cp,t(T) are
the (linearily extrapolated) specific heat capacities of the glass and liquid re-
spectively, the integration limits TJ and T2 have the same meaning as those in
the previous figure. It must be guaranteed, however, that TJ (on heating) and T2
(on cooling) are definitely in the steady state region of the DSC. The lower limit
of the left hand side integral, the thermodynamic glass temperature Tg , must be
determined so that the integrals on both sides are equal. In other words, the area
4.5
J/!Kgl
3.5
1>-
3.0
~
0
0.. 2.5 Cp.1
0
~
------ ~~ =---=- =--=--

0
'"
..c 2.0
~
'u /
'"
0..
II>
1.5 cpg
1.0
60 T, 70 90 100 Tz C 11
Iempera lure ..
Fig.6.29. Construction to determine the thermodynamic glass temperature Tg , which is de-
fined by the equality of the different hatched areas (for other quantities see Fig. 6.28)
between Tl and Tz and the (extrapolated) cp-curves of glass and liquid (the left
side integral) must be equal to that between the experimental curve and the (ex-
trapolated) cp-curve of the glass (the right side integral). The two areas are
hatched differently in Fig. 6.29.
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 temperature 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 Tg) is available for extrapolation.
An error estimation was done by Richardson, Savill, 1975a: for a typical /).cp of
0.3 J g-l K- 1 an uncertainty of 0.3 J g-l for the enthalpy change would result in a
temperature uncertainty of 1 K. For total enthalpy changes of about 100 J g-l and
caloric errors of 1%, the determined Tg would be uncertain to 3K and
this is not acceptable in practice. Fortunately, some errors tend to compensate
each other in both procedures. For instance, an incorrectly extrapolated cp, I-curve
(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' Tg cannot, therefore, be deter-
mined to better than 0.5 to 1 K with this method during routine measure-
ments. However, this is not the reason why the thermodynamic Tg is so rarely used
in practice - 1 K is adequate for many investigations of common Tg relations.
In fact, the scarcity of thermodynamic Tg data has mainly been due to a lack
of suitable programs in manufacturers' software. The situation is now changing
rapidly and "fictive temperatures" (Tg,fic == Tg) can be calculated for most in-
struments. This is fortunate because for theoretical investigations of the kinetics
of the glass processes the situation is clear: only the thermodynamically defined
glass transition temperature reflects unambiguously the thermal history and all
the other conditions during the formation of the investigated material. Tg is thus
the central quantity for all kinds of relaxation studies. The aim of such investi-
gations is to reproduce the behavior of the glass in the transition region, i. e., in
the case of DSC measurements to reproduce the course of the function cp = cp (T)
precisely. As the changes in Tg , caused by different thermal histories, may only
be of the order of a few tenths K, Tg has to be determined at least with that pre-
cision.
To minimize the errors in determining Tg ,
- all error sources, associated with the determination of cp (cf. Sect. 6.1) must
be borne in mind and carefully excluded,
- the sample should remain untouched in the apparatus during all experimen-
tal manipulations even for (often time consuming) annealing experiments.
Annealing the sample outside the apparatus almost always yields unsatisfac-
tory 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 measure-
ments need not necessarily agree with those of other methods. Discrepancies
are caused by the different influences of the particular technique on the re-
laxation of the intrinsic variables (Duncan et aI., 1991). A formal conversion,
taking into account the various experimental influences, can be made using
the (modified) equation of Williams, Landel and Ferry (WLF) (Williams et aI.,
1955). Principal differences must be attributed to different interactions between
the method in question and the relaxation time spectrum of the intrinsic vari-
ables.
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. 6.30 the Tg values (open symbols) of polystyrene (determined as ex-
plained in this section) for different cooling (circles) and subsequent heating
(triangles) runs at different rates are shown (Schawe, 1996).As can be seen, there
is a significant difference between the results of heating and cooling obtained at
a particular rate. In addition, this difference increases with the heating (or cool-
ing) rate in question. From the theoretical point of view, there should not be any
difference between the thermodynamic glass transition temperature measured
in the heating and cooling mode if the sample has been cooled with the same
rate before the heating run and if relaxation effects during the cooling run can
be neglected. The glass transition should only depend on the procedure accord-
ing to which the sample has been transformed 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 relative to the measured one. Thus the
experimental glass temperatures are not fictive (or thermodynamic) values,
Fig. 6.30. Thermodynamic

glass transition temperature
Tg of polystyrene from DSC
measurements as a function
of heating rate. Circles: cool- 1,5
ing mode; triangles: heating
mode; open symbols:
as measured, solid symbols:
desmeared; according to
1.0
Schawe, 1996
'-'
.9) 0,5
0,0
370 375 380

Tg In K
though determined as such. If the step response desmearing procedure described

in Sect. 5.4.4 is applied and the glass transition temperature determined accord-
ingly, the result is quite different (solid symbols in Fig. 6.30). Now the corrected
values from the cooling and heating experiments almost superimpose. Fur-
thermore, there is clearly a systematic change in Tg with the cooling rate (l.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,1996).
Otherwise, Fig. 6.30 also offers a simple and rapid procedure for the practi-
tioner to obtain transition temperatures from non-desmeared curves. The mean
values of the conventional midpoint temperatures obtained during subsequent
heating and cooling runs with the same rate are practically independent of
smearing effects and representative for the chosen rate. Analogous to Fig. 6.30
some measurements with different rates allow the extrapolation to zero heating
rate. The temperature differences of such values between 10 participants of
a GEFTA round robin test (Schick, 1999) were smaller than 1 K and hence
absolutely sufficient for all practical purposes. Therefore this method is highly
recommended to come to a well-defined and reproducible glass transition tem-
perature.
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 cross-linked polymers), the evaluation of the DSC curve is difficult and
uncertain. Dynamic mechanical or dielectric measurements are then more

suitable.
6.4.4
Applications of Glass Transition Measurements
Many important applications of DSC measurement in the glass transition region

are connected with polymer research, these are:
- Theoretical investigations concerning the thermokinetics of the glass transi-
tion. The quantitative description and modeling of relaxation phenomena
implies either the determination of enthalpy changes (Petrie, 1972; Cowie,
Ferguson, 1986; Agrawal, 1989; Montserrat, 1992; Hay, 1992) or of the thermo-
dynamic glass transition temperature (Moynihan et al., 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 al., 1991). Quantitative descriptions of the phenomena are often
based on scanning experiments alone (Kovacs, 1963; Kovacs, Hutchinson,
1979; Moynihan et al., 1976; Ramos et al., 1984; Hutchinson, Ruddy, 1988;
Hutchinson, 1990, 1992; Chang, 1988). However, some aspects of glassy be-
havior can be better studied following various isothermal annealing sched-
ules (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.
It is well known that the approach of the structure toward equilibrium when
held at a constant temperature (up to 50 K below the glass transition) is non
exponential. This is caused by a broad distribution of different relaxation
times. The non-exponentiality can be considered by assuming a "stretched"
exponential expression (Moynihan et al., 1976).
T is a characteristic time, the exponent f3 is inversely proportional to the width

of the corresponding distribution of the relaxation times (0 < f3 < 1).
This model treats the relaxation behavior as thermorheologically simple,
assuming that the relaxation function is independent of temperature.
In addition, the relaxation rate according to the last equation does not
depend linearly on the initial temperature jump. But linearity can be restored
by the concept of the reduced-time integral (Narayanaswamy, 1971) for a
temperature jump from To (at which the material was initially equilibrated)
to T:
T dt
~=f--
To T(T, T f )
Then, the dependence of the fictive temperature Tf can be expressed accord-

ing to the Boltzmann superposition principle:
(-sP dT
T
Tf(t) = To + f (1- exp
To
The most frequently used expression for T (T, Tf ) although purely empirical is
the Moynihan equation (Moynihan et aI., 1976):
T (T, Tf ) = A . exp ( x -i1h* + (1 - i1h*)

x) . -
RT RTf
x is the non-linearity parameter (O < x <1). A is the pre-exponential factor

and i1h* is an effective activation energy. In other words, T is dependent on
both temperature and structure. This model successfully describes practically
all experimentally observed relaxation phenomena of inorganic and organic
amorphous materials.
Another model, based on the Adam-Gibbs equation (Adam, Gibbs, 1965) and
advanced by Hodge, 1987, gives more meaningful model parameters but it is
more rarely used and it does not provide for a better fit of the relaxation data.
- Tg as one of the most important properties of amorphous materials. For
production of new materials with defined properties, the prediction of the
glass transition temperature is a natural necessity. Predictions can be based
on group contributions (Becker, 1976, 1977, 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 (i1cp of the glass transition) are still
only rarely found.
- Finding of relations between Tg and caloric quantities of the glass or the glass
process (Becker, 1976; Batzer, Kreibich, 1982; Wunderlich, 1990).
- Relations between molar mass and glass transition temperatures of polymers.
The most widely used equations are those of Fox, Flory, 1950:
and Ueberreiter, Kanig, 1952:
1 1 A'
-=--+-
Tg Tg,= Mn
where Tg,= is the glass transition temperature of a polymer with infinite num-
ber 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
Fig.6.31. DSC measure- lb 11

ments of an epoxide cured at 4 h
160C for the times shown: 3h
the glass transition tempera- 2h
ture increases with extent
of reaction (according to 15 h
Wisanrakkit, Gillham, 1990) 1h
min
1 min
'"
QJ
..c
o 100 200 300 O( 400

tempera ture
investigations, the glass transition temperature must be determined from

heating runs, the thermodynamic Tg should always be used. All other (empir-
ical) Tg values 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
progress of a polymerization than is the heat production. This is especially true
towards the end of the reaction where the heat production is low and barely
detectable, whereas a distinct change of Tg may be observed (Mijovic, Lee, 1989;
Wisanrakkit, Gillham, 1990; Hale et aI., 1991), for instance, if the cross-link den-
sity increases. As typical example, DSC curves for an epoxide-amine system,
which has reacted for different times at 160 DC, are shown in Fig. 6.31.
It is generally not possible, and in any case it would be far too time-consum-
ing, to determine the degree of reaction only from the changes of Tg during
polymerization. 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 pre-
cisely determinable.
A "calibration curve" must be determined connecting the degree of reaction
(obtained by another method such as an IR technique or, for soluble poly-
mers, determination of the molar mass) with the Tg values obtained by DSC
measurements, nearly identical experimental conditions should be chosen
for the different methods. The empirical function Tg = tea), obtained this
way, is not related to a specific reaction model but it can be calculated if some
assumptions are fulfilled: for instance, the Fox-Flory model can be used to
describe the linear (L) polyaddition reaction:
Tg,Q and (Tg,=h 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 cross-linked polymers. In this case Tg shows a more
pronounced rise, due to the increasing cross-link density, than is the case for
linear polyaddition:
The relation between L".Tg,v and the cross-link density v is complex and the
separation of L".Tg into L".Tg,L and L".Tg, v may not be possible but Min et aI., 1993
were successful. They found by NIR measurements that in the system bisphe-
nol-A diglycidyl ether (DGEBA) and 4,4' -diaminodiphenyl sulfone (DDS) the
reaction of the primary amine hydrogens only increases the molar mass of
the linear polymer whereas that of the secondary amine hydrogen is respon-
sible for cross-linking. Hence the total degree of reaction a can be separated
into one part aL for the linear polymerization and another part a v for the
cross-linking reaction. Both parts can be individually obtained from the ex-
periment. If Tg,= represents the glass transition temperature of the completely
reacted and cross-linked polymer, the following is valid:
- 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. 6.4.3) which cannot be reproduced by any subse-
quent thermal manipulation (Montserrat Ribas, 1992). The reason for this ex-
ceptional relaxation behavior 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 con-
taining plasticizers. The influence of plasticizers on Tg of polymers or the
glass transition behavior of blends have often been investigated (Fox, Flory,
1950; Couchman, Karasz, 1978; Chee, 1985; Kalachandra, Turner, 1987; Braun
et al., 1988; Brekner et aI., 1988; Tsitsilianis, Staikos, 1992; Pomposo et aI., 1993).
Homogeneous, one-phase blends (random copolymers and miscible poly-
mers) have only one Tg transition, whereas heterogeneous, two-phase blends
(block copolymers and incompatible polymers) show the separate glass tran-
sitions of the components with L".cp values correlated with their relative
masses. As a rule, the glass transition range of homogeneous copolymers is

distinctly enlarged compared with the pure components. The glass transition
temperature of an ideal homogeneous system of two components can be cal-
culated from their Tg values and mass fractions Wi by the Fox, Flory, 1950
equation:
1 WI Wz
-=--+--
Tg TgI Tg.z
or by the Karasz equation (Couchman, Karasz, 1978):
In Tg = WI ~cp I
'
In Tg , I + Wz ~cp,
z Tg z
.
WI ~Cp.I + Wz ~cp,z
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; Schneider, 1999).
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 "amplifier" 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 transition and thus imply a homogeneous product. As an
example Fig. 6.32 shows the DSC curves of one homogeneous and two hetero-
geneous glasses before and after annealing. The two heterogeneous blends dif-
fer in the amount of both copolymer components. After annealing the homo-
geneous sample shows only one non-structured relaxation peak, whereas
the heterogeneous ones possess two separated relaxation peaks, though the
DSC-curves of the unannealed glasses show no signs of heterogeneity.
- The glass transition of semi crystalline polymers. The inverse correlation be-
tween the glass transition and melting process has been investigated (Schick
et aI., 1985, 1988). It is often found that the sum of the amounts of amorphous
material (determined from ~cp) 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 transi-
tion temperature and other characteristic temperatures, e. g., the melting
temperature or the temperature of the maximum crystallization rate.
- Testing of pharmaceuticals. Nowadays amorphous forms of pharmaceuticals
are in widespread commercial usage. Substances in amorphous state show
often different physical properties than the corresponding crystalline forms,
in particular, an improved bioavailability and ease of processing due to
Fig. 6.32. DSC curves of a

homogeneous (A) and two
heterogeneous (B, C) blends
which have been quenched
from the melt or annealed
a few K below Tg for 14 days. A
The relaxation peak( s) of
the annealed samples clearly
show that blends Band C are
phase separated (according quenched
to Jorda, Wilkes, 1988)
f annealed
B
--
)0
o quenched
o
'"
J:
annealed
(
quenched
30 50 70C 90
temperature ..
changed mechanical properties. On the other hand, there is a risk for an un-
desirable crystallization during storage, often connected with a change of the
efficacy of the drug. Therefore, the knowledge of the molecular mobility in
amorphous pharmaceuticals is of decisive importance for their practical use.
The amorphous pharmaceuticals must be kinetically stable much longer than
the expected lifetime of a drug. Both a possible tendency and the rate of
crystallization can be estimated from the storage temperature in comparison
with Tg and from the evaluation of corresponding relaxation experiments
(Hancock, Shamblin, 2001; Crowley, Zografi, 2001).
6.4.5
The Dynamic Glass Process, an Example
As mentioned above the glass process is a relaxation (time dependent) process.

Below the glass transition the system is in a non-equilibrium state which -
strongly depending on the temperature - slowly moves toward the equilibrium

state. The vitrification or devitrification is coupled with a change of mobility
(degrees of freedom) manifested to a change of the heat capacity, which thus
becomes a function of time (at least in the range of, and below the glass temper-
ature). This implies, that both the glass transition temperature as well as the heat
capacity becomes a (complex) function of frequency (cf. Sect. 3.3 and 5.5.2).
Therefore the temperature-modulated DSC method suggests itself (among
other methods such as dielectric or dynamic mechanical measurements) to in-
vestigate the dynamics of the glass process.
As an example to show the usability of the TMDSC method, the "activation
diagram" of the glass transition of glucose is shown in Fig. 6.33. The step-like
change of the real part of the complex heat capacity cp (co) in the glass transition
region is always coupled with a maximum in the imaginary part COmax This max-
imum shifts in a characteristic way to another frequency if we change the tem-
perature. The same is true if we, in a TMDSC experiment, change the tempera-
ture at a given frequency, again the imaginary part has a maximum at the glass
transition temperature (see Fig. 5.15). It can be shown that both descriptions
are equivalent. However, in Fig. 6.33 the maximum temperatures of dielectric ex-
periments (DRS) - done with changing frequencies at fixed temperatures - are
plotted together with the glass temperature - measured with different tempera-
ture-modulated calorimetric methods at fixed frequencies (Schick et al., 2002).
The result is convincing, it is obvious that TMDSC is a suitable method to mea-
sure the frequency dependence (dynamics) of the glass transition. The experi-
mental findings (symbols) fit to the theoretical prediction (solid line) of the
Vogel Fulcher Tammann approach for the calorimetric as well as for the dielec-
tric results. The systematic shift is due to the different stimulation of the glass
process in the electric and thermal experiment.
It should be mentioned and emphasized, that the dynamic glass process in-
vestigated with TMDSC and the thermal glass process (vitrification or devitrifi-
Fig.6.33. Activation dia- 6

gram of glucose from tem- 5
perature-modulated calori-
metric measurements (solid 4
symbols) and dielectric
results (open symbols)
t 3
together with fit functions 'VJ 2
-.:::J
from the Vogel Fulcher d

'--- 1
Tammann approach (accord- "-
ing to Schick et al., 2002) "2l' 0
= -1
0
-2
-3
-4
29 3.0 3.1 3 2 K- 1 33
1000/ T - -
cation) visible in common DSC curves are different processes, which result in
different glass transition temperatures. The glass temperature from the thermal
process depends on the thermal history (cooling rate, annealing time etc)
whereas the glass temperature from the dynamic process depends on the fre-
quency of the measurements. In a TMDSC experiment the former manifests it-
self in the underlying, and the latter in the reversing cp curve. From this follows
that the non-reversing curve (see Sect. 5.5.1) contains the difference of two in
principle different cp curves. Every, however done, evaluation of the non-revers-
ing cp curve in this region gives only accidental results.
6.S
Characterization 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 of the 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 inves-
tigation 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.
6.S.1
Applications in Biology and Food Science
The investigation of polymorphism - defined as the ability of a chemical com-

pound to crystallize into different states (modifications) - is increasingly im-
portant in food science.
A well-known example is cacao butter. Chocolate products with cacao butter
have the desirable consumer properties only if the p-form with a melting point
at 35C is applied. DSC measurements are an effective tool for both optimizing
the production method and the quality control.
Another example from food science is common butter. The objectives of in-
vestigations are an improvement of the cold consistency (spreading property) of
butter and the concurrent reduction of softening when the temperature is in-
creased. From DSC curves the influences of additives (triglycerides) with low-
melting point, and differences in the cream-ripening method can be directly
concluded (Schaffer et al., 2001).
A frequently used sugar substitute is isomalt, a nearly equimolar mixture of
two isomers containing crystal water. The crystalline material melts around
142-150C, and after cooling results a glassy state with a glass transition tem-
perature at about 60C. Like other carbohydrates used in food industry, isomalt
is very often used in its amorphous form. The final quality and the storage
stability of isomalt containing products depends on the aging behavior of the
non-equilibrium glassy state. Therefore, the properties of both crystalline and
0.03
WIg
1 0.02
!:!
a
.....
0.01
~
=-
a
'"
.&:.
u
3
u
0
C1J
0-
III
-0.01 L...-_ _ _ _--"-_ _ _ _ _-'--_ _ _ _ _'---_ _ _ _--'
18 21 24 27 C 30
temperature .-
Fig.6.34. DSC heating curves of the gel-liquid crystalline phase transition of a liposome.
1: 1.75% of dimyristoylphosphatidylcholine (DMPC) in water, 2: after addition of 3%, and
3: after addition of 5 % of vitamin D3 (mass: 15 mg, heating rate: 2 K min-I)
glassy isomalt were intensively studied (Cammenga et al., 1996; Borde, Cesaro,
2001).
The denaturation temperature is a measure for the thermal stability of pro-
teins, essential in food science. The determination of the denaturation heat
allows conclusions regarding a possible already existing thermal impairment
of the proteins. Therefore Schubring, 1999, has studied the thermal behavior of
herring muscle, skin and pyloric caeca during salting and ripening.
As a rule, applications in biochemistry or food industry require highly sensi-
tive DSC instruments. As an example, Fig. 6.34 shows the transition from gel to
liquid crystalline state of an appropriately prepared liposome of 1.75% of
dimyristoyl phosphatidylcholine (DMPC) in water. The reversible transition
(curve 1, Q = 0.47 J per g total sample mass, Tp = 24.6C) 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 D3. With increasing concentration
of vitamin D3 in the liposomes, the peak becomes broader and smaller and
the peak maximum shifts to lower temperatures (curve 2, 3 % of vitamin D3,
Q = 0.15 J g-I, Tp = 22.7C). Addition of more than 5% of vitamin D3 to the
DMPC-water lipid (curve 3) causes the disappearance of this peak.
6.S Characterization of Substances, the Phase Behavior 221
Another point of view concerns the interaction of common environmental

contaminants like phthalic esters and biomembranes (Bonora et aI., 2000).
Dipalmitoyl phosphatidylcholine was used as a model of cellular membranes
because lecithins are the major components in mammalian membranes. The
authors have found noticeable effects on temperature, heat, shape and width of
the main transition even in the presence of small amounts of phthalates.
6.5.2
Applications in Pharmacy
Without doubt most investigations on polymorphism concern compounds used
as pharmaceuticals. The majority of drugs show one or the other form of poly-
morphism (Grunenberg et aI., 1996).
General Remarks
Phase transitions between different modifications may be reversible (enan-

tiotropic) or irreversible (monotropic). This behavior can be understood
best using energy-temperature diagrams (Fig. 6.35, Burger, Ramberger, 1979;
Kuhnert-Brandstatter, 1996; Grunenberg et aI., 1996). Two modifications A and
B are called enantiotropic or monotropic forms, if the free (Gibbs) energy curve
of the liquid G1 crosses the respective curves of the solid modifications A
and B above (Fig. 6.35 a) or below (Fig. 6.35b) the intersection point of GA and
GB Burger has formulated two important rules which immediately follow from
Fig. 6.35.
The Heat-of Transition rule: If an endothermic (exothermic) transition enthalpy

is found on heating, the two polymorphs are enantiotrops (monotrops). Because
solid-state transitions are more or less kinetically hindered, apparent deviations
from this rule are frequently observed and the actual behavior depends strong-
lyon the heating rates (compare the next two examples). Generally, on heating,
the measured temperature of the DSC peak T;rs (AB) is higher than the thermo-
dynamic transition point Ttrs(AB) (Fig. 6.35a).
The Heat-of Fusion rule: If the form with the higher melting point shows a lower
heat of fusion the two forms are usually enantiotropic, otherwise they are
monotropic. This rule is a good help if the melting-point difference of the two
forms is small and heat capacity differences between supercooled liquid and
crystalline B can be neglected.
In principle, the detailed thermodynamic analysis of DSC data enables the
estimation of stability domains for possible modifications. This is possible if the
cp functions, differing only very slightly, can be measured with the necessary
precision. Of course, the quantitative determination of the complete energy-
temperature diagram according Fig. 6.35 is difficult and very time-consuming,
often even impossible. Otherwise, to find the thermodynamically stable form
of two modifications A and B within a given temperature range, the absolute
values of their thermodynamic functions of state (G, Hand S) are not needed
Fig. 6.35 a, b. Energy-tem-

perature diagram of a H,
dimorphic system.
a enantiotropic system,
b monotropic system.
t
>--
G: Gibbs free energy, C"l
H: enthalpy, Ttrs (AB): ther- ~B~~~~-
At--
ill - __
modynamic transition tem-
perature from A to B, T;rs --~
,~-
(AB): measured transition . ~--
temperature, Ttrs (BA): mea- -G A
sured temperature for the GB
monotropic transition B a ~ __________~__~____~____~____~G,
to A (A, B: solid phases,
1: liquid phase)
o T'rs(AB) T;rs(AB) Tfus(A) Tfus(B)
temperature/K -
H,
>-. llf"HIAI
en
::;;
c:
I ~--==-~CHffi-:;;-
- _ _ _ lll"H<BAl HB
ill B -___ - HA
A --~
--- "'" ; ~-_GB
~-GA
; G,
b
o
temperature/K -----
but only the differences against a reference state. In this case the liquid state of that
modification with the highest melting temperature is used as reference state.
With (aH) =C p and as)

(aT =T
Cp
one obtains:
aT p
p
Tfus,B T T
HB(T)-HA(T)=t1fusHA-t1fusHB+ f Cp,ldT+ f Cp,BdT- f Cp,AdT
Tfus, A Tfus,B Tfus, A
and
t1f H t1f H Tfus,B C l T C T C
SB(T)-SA(T)=~-~+ f ~dT+ f ~dT- f ~dT
Tfus,A Tfus,B Tfus,A T Tfus,B T Tfus,A T
The difference of the Gibbs free energies is then obtained as usual:

This way, Sacchetti, 2001, could verify that the transition temperature of two
modifications of acetaminophen is clearly lower than previously thought
(- 30 DC). This result is consistent with the recent observation that form B spon-
taneously converts to form A even at - 70 DC. Furthermore, Sacchetti could show
that the greater stability of form A is not caused by a lower packing energy but
by its higher entropy.
PharmaceuilcalAspecB
Investigations regarding the polymorphism of pharmaceutical substances are
especially important, as different modifications have because of different efficacy,
different solubilities and dissolution rates (bioavailability), and frequently also
different processability (Burger, Ramberger, 1979; Kuhnert-Brandstatter, 1996;
Giron, 1999). In general, the modification used should be the thermodynamic
stable one. Metastable or glassy products showing better dissolution rates, can
be used, if the stability is guaranteed for storage.
DSCs are also very useful for investigations of stability and physical and chem-
ical interactions between drugs and excipients. The knowledge of the polymor-
phism of a excipient including its aging behavior is extremely important for the
development of preparation forms such as suppositories, micro emulsions and so-
lutions in soft gelatin capsules. Formulations with amorphous substances tend to
form crystalline modifications with reduced bioavailability upon storage, strong-
ly dependent on climatic conditions (temperature, moisture content). The deter-
mination of the amorphous part in semi-crystalline drugs is usually performed by
X-ray diffraction. The detection limit is 5-10 %. Measuring the crystallization en-
thalpy, Giron, 1999, was able to detect amorphous parts less than 5%.
Examples
The first example of polymorphism concerns the melting of acetamide (Fig. 6.36).
During the first heating the thermodynamically stable modification (the as re-
ceived state) melts at about 80C. On cooling from the melt and in the absence
of crystal nuclei of the stable modification the metastable modification crystal-
lizes first in accordance with Ostwald's step rule. The second run thus shows the
melting peak of this metastable modification at about 65 DC. In the absence of
crystal nuclei neither a possible monotropic transition below 65 DC nor the re-
crystallization of the supercooled melt between 65 and 80C is found.
Remark: The formation of such metastable phases is one of the main prob-
lems encountered in the determination of phase diagrams, and this 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 1 in Fig. 6.37a) melts at about 103C. Rapid cool-
ing of the melt yields a glassy phase. On reheating the glass (curve 2) this first
recrystallizes to a metastable form at about 35C. Subsequent behavior depends
on the heating rates, either melting of the metastable modification at about
93 C or of the stable structure at 103C will be observed. This is because the rate
of transformation from the metastable to the stable form is low in the solid
phase (i. e., < 93C), only at low heating rates 1 K min-I) will there be enough
50.0
mW
37.5
~
1 25.0 /
/
"\
::
=-
0
/
2
c
C1I 12.5 /
-'=
/
0
50
.-._"
58
- ,/
66 74
--.-.
82 O( 90
temperature ~
Fig.6.36. DSC melting curves of the thermodynamically stable (1) and of the metastable (2)
modification of acetamide
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 of the metastable modification. The corresponding DSC
curves are shown in Fig. 6.37b. For better comparison of both curves, curve 2
was also run with a heating rate of 40 K min- I following a preliminary heating
to 98C at 1 K min- I and recooling to room temperature. At moderate heating
rates (ca. 5 K min-I), however, the sample reaches some intermediate state. Any
remaining metastable phase that has not transformed during the slow solid
state reaction, melts at 93C, it then transforms to the stable modification as
an exothermic reaction and finally melts at 103 C (curve 2 in Fig. 6.37a). This
can be proved by quenching the sample immediately after completion of the
exothermic reaction (at 100C) and reheating it again, when only the melting
peak of the stable phase is seen (curve 3).
6.5.3
Other Applications
Liquid Crystals
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. 6.38). The
first peak (Q = 77.3 J g-I, T = 69.0C) corresponds to the formation of a smectic
20
mW
15
f 10
.!!
...
CI
-...
:;J:
0
CI
..c: 5 2
3
a 0
40 55 70 85 100 e 115
25
tempera ture ...
70.0
mW
52.5 2
t
.!!
"" 30.0
....
:;J:
....
0
""
....
.c;
17.5
b oL===~==~====~~~~~
o(
55 70 85 100 115 130
tempera ture
Fig.6.37. Polymorphism of phenylbutazone.
a 1: 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 100DC (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 previous heating to
98 DC with 1 K min-I (heating rate: 40 K min-I)
20
mW
15
f
~
10
E
=-

-=c
'"
..c:
5
o~------~------~------~------~------~~----~
60 65 70 75 80 85 C 90
tempera ture
Fig. 6.38. Polymorphism of cholesteryl myristate.
1: first heating, 2: cooling to 60C, 3: subsequent heating, 4: heating after cooling from the melt
to room temperature (mass: 4 mg, heating and cooling rates: 5 K min-I)
mesophase. The following two peaks (Q = 2.5 Jg-I, T= 77.7C and Q = 1.9 Jg-l,
T = 83.0C) correspond to the formation of the cholesteric mesophase or the
transition 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 Jg-l, T = 68.5 C), 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 tran-
sitions are observed during the subsequent reheating (curve 3). The heat of tran-
sition 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. This is due to the fact that mesophase
transitions are probably second order transitions. The onset temperatures of the
heating and cooling runs do not differ from each other after proper correction
of the extrapolated onset temperatures to a heating or cooling rate of zero (on-
set differences of AT = 0.2 K and AT = 0.0 K for the two transitions). The crys-
talline phase, however, shows considerable supercooling. It crystallizes at about
30C, equivalent to a AT of about 40 K.
Extreme Melting Ranges
Sometimes melting processes extend over very broad ranges of temperature.

For low molar mass substances this may be due to large amounts of impurities
(cf. Sect. 6.9). For macromolecules this may be caused by the broad distribution
of the molar masses or by a distribution of the thickness of the crystal-lamellae.
The determination of the appropriate baseline and the choice of physically rea-
sonable 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.
Processes with Gas Exchange

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 experimen-
tal conditions, the type of purge gas, its flow rate, the geometry of the sample
holder, sample preparation and so on. A typical example is the thermal decom-
position reaction of calcium oxalate monohydrate. The temperature of the first
dehydration peak and the peak width are radically influenced both by the heating
rate and the effectiveness of the water vapor 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
decomposition of calcium oxalate, 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
t nematic -+
.!!. 25.0 liquid pha se
\
....
Q
:J:
0
Q
C1.I smectic -+ nema tic
\
~
12.5
o L -_ _ _ _ _ _ ~ ______ ~ ______ ~ ________ ~ ____ ~
o 40 80 120 160 c 200

tempera ture ..
Fig.6.39. DSC curve of the phase behavior of calcium stearate monohydrate (mass: 3 mg,
heating rate: 10 K min-I)
the presence of O2 the CO is completely or partially oxidized. The simultaneous

coupling of the DSC with thermogravimetry or mass spectrometry is advanta-
geous in these cases.
Another example, the phase behavior of calcium stearate monohydrate is
shown in Fig. 6.39. 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 vapor 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 crys-
talline to the smectic phase, from smectic to nematic and finally from nematic
to isotropic liquid. Within the homologous series of fatty acids the phase behav-
ior 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.
6.5.4
Porosity Measurement
Porous materials have an increasing technological importance. Due to the sur-

face contribution to the free energy of liquids within the confined geometry of
a porous solid, their melting and freezing properties are different from those
of the bulk phase. Frequently used experimental techniques are NMR, N2 -ad-
sorption, small and wide angle X-ray scattering (SAXS and WAXS), mercury
porosimetry and thermoporosimetry. The latter method was developed by Brun
et al., 1977, and permits the determination of both the mean pore size and the
pore size distribution from the freezing and melting DSC-curves of a suitable
liquid (e. g., water, cyclohexane, alkanes) confined in the pores. The steps of a
typical DSC experiment are:
1. gasing-out of the porous solid by annealing under reduced pressure,
2. saturation of the pores with the desired liquid,
3. sealing of the sample in vapor-tight sample pans, together with a small excess
droplet of the liquid,
4. first cooling run, starting well above the freezing temperature of the bulk
liquid,
5. subsequent heating run, possibly followed by further thermal cycles,
6. determination of the total pore and bulk liquid by weighing before and after
piercing the pan and annealing the sample under reduced pressure.
Generally it is found that both the position and the shape of the peaks from
repeated cooling/heating cycles are extremely reproducible. An alternative but very
time-consuming technique replacing the steps 3 and 4 was suggested by Neffati et
al., 1998 ("fractionation" of the crystallization or melting peak). It is applicable in
all cases of a distinct hysteresis between freezing and melting temperatures.
A typical experiment is shown in Fig. 6.40 for a porous silica gel (f3 = -1 and
+ 1 K min-I) ftlled with water. The large peaks correspond to the excess bulk
water, the broader and less intense ones to the "confined fluid". Contrary to the
shape of the peaks for the bulk liquid, the width of the peaks for the pore liquid
I
1- (onfined \later
Wg- 1
OJ 01.----+---
cJ
-1 1 - - - bulk water
cJ
OJ
-L
-2~----~----~~------~----~
-30 -20 -10 o 10
temperQture ..
Fig.6.40. DSC curves on heating (upper) and cooling (lower) of water confined in a porous
silica gel (heating and cooling rate: 1 K min-I)
doesn't become smaller for lower heating rates but is determined by the pore size
distribution. The highest usable heating and cooling rates must be determined by
preliminary experiments, in order to maintain thermodynamic "equilibrium" con-
ditions andto avoid the "smearing" of the peak shape due to the thermal inertia
of sample and DSC. Typical values are 0.25 to 2 K min-I, something dependent on
the concrete system (Faivre et al., 1999; Neffati, Rault, 2001; Ishikiriyama et al.,
1996). The pronounced but more or less statistical supercooling of the freezing
peak for the bulk water due to the delay of heterogeneous nucleation is uninter-
esting but often very disturbing if overlapping with the freezing peak of the pore
water. Other liquids, e.g., cyclohexane, show much smaller supercooling. Proba-
bly, the freezing in the nano-sized pores occurs through the slow penetration
of the freezing front, dependent on the pore size aperture. The solid-liquid
meniscus formed at the pore aperture should be spherical both for cylindrical
and for spherical pores. Contrary to this, the melting of the confined material is
initiated at the pore wall surface, the interface has the form of the pore surface.
The result is the distinct hysteresis between freezing and melting.
The profile of the DSC curve and the hysteresis between freezing and melting
allow conclusions on the shape of the mesopores. The shift of the transition tem-
perature is given by a generalized form of the Gibbs-Thomson equation (Faivre
et al., 1999):
o Ysl V.
I1T= - a T fus ' - ' - -
r I1fusH
where 11 T = (T?us - T) is the melting point temperature depression of the con-

fined liquid, T?us is the melting temperature of a crystal of infinite dimension,
Ysl is the average interfacial tension of the crystal, V. is the molar volume of the
crystal, I1fusH is the molar enthalpy of melting and r is the radius of the pores.
The factor a depends on the shape of the pores (Faivre et al., 1999). Apart from
uncertain values for Ysl the reliability of the results is limited by using the correct
value for t:..fusH. If the total peak area is divided by the total amount of the used
liquid and subsequently the part of the bulk phase obtained from the bulk liquid
peak is subtracted, one always obtains values for t:..fusH in the pores which are
distinctly smaller than those for the bulk phase. A distinct temperature depend-
ence of t:..fusH suggested by Neffati and Raul, 2001, is rather improbable. The
most probable explanation is the existence of anon-freezing layer on the surface
of the pores. This can be verified if the above equation is modified and rewrit-
ten (Schmidt et aI., 1995) as
K
t:..T=--
(r - t)
with
o
K= -a' T f us VI'--
V.
's A H
ilfus
That means the effective pore radius is reduced from r to (r - t). If the experi-
mental data for defined and systematically varied pore sizes are fitted to this
equation, t is in the order of one to three monolayers. The accessible range of r
is limited for large pores because of overlapping and missing separation of the
peaks for bulk and pores, and for very small pores because of missing filling
with the test liquid (dependent on its required space).
The same technique can also be used to obtain the distances of the cross-link
points of swollen polymers after equilibration in a appropriate swelling agent,
e. g., cyclohexane.
6.6
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 Appen-
dix 1) at the beginning of the twentieth century.
As an example, the phase diagram of the binary system benzil!acetanilide is
shown schematically in Fig. 6.41 together with the DSC curves of the heating
runs of the pure components (curves 1 and 7) and of five mixtures including
the eutectic (curves 2 to 6). In accordance with the phase diagram benzil and
acetanilide are completely miscible in the liquid and immiscible in the solid
phase. Pure acetanilide melts at 115C, pure benzil at 95C. The eutectic, with
a mole fraction of about XBen = 0.58, melts at 78C. In Fig. 6.41 a two different
melting pathways (mole fractions Xl = 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. 6.41 a). 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
6.6 Determination of Phase Diagrams 231
115 6C
t 95C
78C
x __
WIg
6 1--_2=--_ _
1
4
7
b OL-____~~c=~====~====~~=_~
60 73 86 99 112 C 125
tempera lure
Fig. 6.41 a, b. 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 melt-
ing process for two mixtures with the mole fractions Xl and X2, b DSC curves of the pure com-
ponents and of five mixtures, 1: acetanilide (XBen = 0),2: XBen = 0.1,3: XBen = 0.4,4: XBen = 0.578
(eutectic), 5: XBen = 0.75,6: XBen = 0.9,7: benzil (XBen = 1)
melt at well defined temperatures. These can be determined from the extra-
polated peak onset temperatures in the usual way, whereas the determination of
the liquidus temperature is not easy. The peak maximum temperature (if neces-
sary 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 K min-I) and small sample
masses should be used, to ensure approximate thermodynamic equilibrium at
every moment during the experiment and to avoid "smearing" (see Sect. 5.4) of
the DSC curve and thus incorrect temperatures. The heating mode is more suit-
able for the determination of phase diagrams than the cooling mode, because
many organic substances have a tendency to supercool considerably.
If the heats of fusion of the pure components are known and the binary sys-
tem behaves ideally in the liquid phase, the theoretical melting course may be
predicted, using the Schroder-Van Laar equation (Brezesinski, Dorfler, 1983; see
also Sect. 6.9).
The molar heat of fusion of the eutectic ~fus Heu can be calculated for ideal
eutectic systems (for which the eutectic is a simple blend of microcrystals) using
the normal rule of mixtures:
XA,XB: mole fractions of the components A and B'~fusHA'~fusHB molar heats of

fusion of the pure components.
For precise calculations ~fus HA and ~fus HB should be converted to the eutectic
temperature using Kirchhoff's law. If this simple additivity holds, the exact com-
position of the eutectic can be determined by plotting the heats of fusion of the
eutectic peak of different mixtures against the composition (Ding et aI., 2000).
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 general-
ly 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 be-
tween components. From these differences, the excess functions of entropy,
enthalpy and Gibbs energy can be calculated (Rai, Shekhar, 1993, Rai, Rai, 1998).
For positive (negative) excess free energy, the attractive interaction between
molecules of the same kind is stronger (weaker) than that one between mole-
cules of different components in the mixture.
Using available data from the literature Sangster, 1999, has published com-
puter calculated and thermodynamically consistent phase diagrams for 60 binary
eutectic systems of drugs. Among the results are also the excess Gibbs energy of
liquid and solid solution phases and the thermodynamic properties of intermedi-
ate compounds. Due to the very long times for the diffusion-controlled establish-
ment of the solid-liquid equilibrium, the determination of phase diagrams with
mixed crystals is difficult and should be supplemented by adiabatic calorimetric
measurements. Some examples are the systems trans-stilbeneltrans-azobenzene
and p-dichlorobenzenelp-dibromobenzene (Stosch et al., 1996) and d-carvonel
l-carvone (Gallis et aI., 1996a-c, 1999).
6.7
Safety Aspects and Characterization of High-Energetic Materials
Specific devices for this type of investigation are the reaction or safety calorime-
ters mentioned in Chapter 1. They provide detailed information for the opti-
mization of process parameters, reducing production costs and avoiding safety
6.8 Characterization of Polymers 233
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 exothermic decomposition reactions, have to be stored for long
periods at temperatures, such that the heat evolved is greater than that trans-
ferred to the surroundings. Under worst case conditions this may lead to a run-
away and subsequent fire or explosion. The direct detection at ambient temper-
ature of the extremely small heat flow rates in question is, in general, not possi-
ble with DTA/DSC (exceptions are Calvet calorimeters, cf. Sect. 2.1.3).
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 ob-
tained for the decomposition kinetics of, for example, p-nitroaniline. Using this
procedure it is possible to determine decomposition heat flow rates down to
10 mW kg-' at storage temperatures.
Decomposition reactions often react sensitively to traces of foreign materials.
With the annealing technique described above, the first detectable decomposi-
tion of azidophenylacetic acid was determined. In addition the effects of many
kinds of "impurities" were investigated. In a similar way the autocatalytic influ-
ence of the decomposition products was definitely proved.
DSC measurements supply valuable results regarding the thermal decompo-
sition of newer high-energetic materials as ammonium dinitramide (ADN)
and 2,4,6,8,1O,12-hexanitrohexaazaisowurtzitane (HNIW). To avoid safety risks,
small sample weights 1 mg) should be investigated in open sample pans
or in special high-pressure capsules. The decomposition kinetics of ADN is
strongly influenced by the reaction conditions (Tompa,2000) or impurities and
stabilizers (LObbecke et al., 1997). HNIW exists in several polymorphic forms
with different thermodynamic stability and shelf life at room temperature (Foltz
et al., 1994).
6.8
Characterization of Polymers
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.
6.8.1
Effects of Origin and Thermal History
Plastics are either amorphous or partial crystalline materials, i. e., the glass
transition temperature and/or the melting region as well as the degree of crys-
tallinity are essential quantities for the assessment of material properties and
working conditions. They depend distinctly on manufacturing (synthesis) and
storage conditions (thermal history) and a reliable determination of these quan-
tities is indispensable in polymer technology. DSC offers the possibility to charac-
terize plastics in a fast and satisfactory way. In fact the rapid development of a pre-
cise equipment for quantitative thermal analysis (DSC) was bound up with the in-
creasing importance of plastics. Polymer science is still a major area of application
for DSC (Mathot, 1994 a, b) and a polymer laboratory without a DSC is unthinkable.
To determine the properties and the thermal history of a polymer requires at
least a cycle of three measurements: Heating the original sample to a tempera-
ture 20 to 50 K above the melting region, cooling the sample to get it into a
defined state and reheat it again. The three DSC-curves are normally analyzed
with respect to:
- The glass transition region: the temperature(s), the width(s) and step
height(s) of the glass transition(s) (see Sect. 6.4) characterize amorphous
polymers, blends of them, and even the amorphous part of partial crystalline
samples. The glass transition temperature reflects the composition of the
components, the 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 par-
tial oxidation). In heterogeneous or micro-heterogeneous amorphous blends
two or more glass transitions are found, which are characteristic of the indi-
vidual components. The relative amounts can be estimated by evaluating the
respective step heights. In partial crystalline polymers the degree of amor-
phous state can be determined from the step height as well, whereas the width
of the glass transition region in blends is influenced by the miscibility of the
components. Further important information can be taken from the possible
existence and intensity of a relaxation peak in the region of the glass transi-
tion on heating. Glass transition evaluation is generally a very important ap-
plication field in DSC and is discussed in detail in Sect. 6.4.
- The melting region: the melting peak(s) temperature(s), the temperature
region of melting (peak width) and the peak area(s) (heats of fusion or tran-
sition and crystallization) characterize the crystalline polymer. As a rule it
is not difficult to identify the components of a plastic mixture, using the cha-
racteristic temperatures of the melting peaks. Often even a quite accurate
quantitative analysis is possible. Overlapping peaks can be separated by using
appropriate evaluation methods (see Opfermann, 1992), since the necessary
condition for using such programs - the additive superposition of the re-
spective peaks because of immiscibility in the solid phase - is nearly always
fulfilled for plastic mixtures. Some polymers appear in different polymor-
phous forms, for such plastics solid-solid phase transitions can be found. For
semi-crystalline products the degree of crystallinity is an essential quantity
and the determination of crystallinity is an important application of DSC

therefore we address this topic a special section (see Sect. 6.8.2). A relative
(within the type of plastic) degree of crystallinity can be obtained very easily
(ignoring all precautions described in Sect. 6.8.2) if the area of melting peak
is compared with that of the pure material of the same type.
- The crystallization behavior: crystallization is visible in a DSC because of the
exothermic heat connected with it. The exothermic peak is easily detected in
scanning mode of operation, its temperature and width characterize the crys-
tallization. Normally crystallization occurs on cooling from the melt, the
crystallization rate depends strongly on temperature and can be followed in
the DSC in isothermal mode of operation. If crystallization takes place dur-
ing a heating run (prior to the melting peak), the sample was definitely in a
non-equilibrium state caused by a special treatment such as quenching the
sample from the melt. This way the thermal history of a polymer sample can
be followed with DSC. In extreme cases, it is found that for bulk thermoplastic
moulded parts the degree of crystallinity varies within the parts, due to un-
even cooling rates during moulding, this can be quantified by determination
of the crystallization heat from the peak area of the different samples.
- Chemical reactions and thermal and oxidative decomposition: the beginning
of chemical reactions, in this case a polymerization or decomposition, is nor-
mally characterized by the appearance of an exothermic (or rarely endother-
mic) signal in the DSC. To determine this, the plastic is either annealed
isothermally in the DSC until such a signal indicates the beginning of reac-
tion or decomposition, or the sample is heated in scanning mode up to this
event. Polymerization reactions can even be followed quantitatively and
kinetic investigations can be performed (see Sect. 6.3). The curing 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). Even decomposition can be followed quantitatively. By choosing
an appropriate gas atmosphere (inert or reactive, e.g., oxygen), information
is available concerning the thermal or oxidative stability of the material. For
these measurements there is always a risk of contaminating and destroying
the measurement system, so the run should be interrupted immediately after
detecting the start of decomposition. However, thermogravimetry is in prin-
ciple much more suitable for this type of investigation.
From such measurements and the evaluation procedures mentioned, valuable

information can be gained which is used to characterize the plastic material, to
identify it and describe the properties. Such information is even useful for qual-
ity control and, of course, in case of complaints and material failure.
For products containing recycled plastics, most of the above mentioned ther-
mal properties may have changed in a characteristic manner. The observed de-
viations from the original behavior provide reliable information whether the
plastic in question is suitable for the intended application or not.
6.8.2
Determination of the Degree of Crystallinity
DSC is widely used for determination of the degree of crystallinity We (weight

fraction crystallinity) of partially crystalline polymers. The most simple and ob-
vious possibility is to relate the estimated heat of fusion (~fusH) to that of the
same polymer but with known crystallinity, provided that this value was deter-
mined by an independent method such as X-ray diffraction, Raman scattering,
density measurements etc. The determination of the required peak area is prob-
lematic, 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 dif-
ficult 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. 6.8), if one
uses fixed integration limits and a fixed reference value for the heat of fusion of
the totally crystallized sample.
Mathot, Pijpers, 1983, 1989, suggest a convenient procedure (Fig. 6.42), which
avoids the difficulties mentioned. The heat flow rate after the melting peak, which
is proportional to the heat capacity function Cp,a of the amorphous (liquid)
polymer, is extrapolated to that temperature T, at which the degree of crys-
tallinity We (T) is to be calculated. If, in doing so, the experimental curve is not
intersected (T > T*, e.g. T2 in Fig. 6.42), only the area Al has to be evaluated. If
the extrapolated line intersects the measured curve (T < T*, e.g. Tl in Fig. 6.42),
the area A2 has to be subtracted from A l ' The heat, corresponding to the areas Al
(or to A 1-A 2 ), is then divided by the reference value ~fush (T) of a completely
crystalline sample at this temperature. For linear polyethylene Mathot, Pijpers,
1983, suggest the following equation for the latter function:
~fush(T) = 293 - 0.3092 ,10- 5 (414.6 - TIK)2. (414.6 + 2TIK) J g-1

The degree of crystallinity We (T) in dependence on temperature is then ob-
tained according to:
Meanwhile the functions ha (T) for the completely amorphous and he (T) for the
crystalline form are known for numerous linear macromolecules with sufficient
accuracy (Wunderlich et al., 1990).
If the specific heat capacity curve is evaluated from the DSC measurement (see
Sect. 6.1), 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
A,
f
~
~ A,
:a
0
;;::
~
c:J
GI
.c.
temperature ..
Fig. 6.42. The 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 tem-
perature of intersection of the extrapolated liquid phase line and the measured curve (for
details see text)
the enthalpy difference function h (T) [or respectively, ha (T) and he (T) ]. This
opens a new thermodynamic based calculation of We (T), 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 ha (1'), he (1') and We (1') 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). Unfortu-
nately the two-phase model fails with some types of polymers, e.g., polyethyl-
eneterephthalate. The baseline can be calculated in such cases by use of im-
proved models, which include a third phase called the "rigid amorphous phase"
{Suzuki et al., 1984, Schick et al., 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 either
to the thermal glass transition or to the melting heat of the crystalline part.
Nevertheless, the heat capacity of this non-crystallized material can be approx-
imated by that of the crystalline phase.
For some polymers the rigid amorphous phase is absent and good results for
We may be obtained for these {polydimethylsiloxane, polybutadiene}, if the mea-
sured step change of the heat capacity at the glass transition temperature Tg
{cf. Sect. 6.4} is related to that of the completely amorphous polymer yielding the
amorphous amount Wa = LiCp{Tg}/LiCp,a{Tg}. In this case the degree of crys-
tallinity is We = 1 - Wa' The completely amorphous state can often be obtained by
quenching or ultraquenching the molten polymer.
The temperature-modulated mode of operation {cf. Sect. 2.4.2} is a very suit-
able method to determine the {vibrational} heat capacity even in the regions of
crystallization {see Sect. 5.5}. This way it is possible to precisely determine the
heat capacity from below the glass transition up to the beginning of the melting
region and even further if higher frequencies are used where the contributions
from melting dynamics are suppressed {Schick et al., 2003}. Together with liter-
ature values of the liquid and solid heat capacities of the sample in question
{ATHAS, 2002}, the degree of crystallinity {solid} can be determined precisely,
even its change in dependence on temperature in the case of hidden crystalliza-
tion. This way it is even possible to determine the rigid amorphous part and its
change with temperature. Schick et al., 2001, showed with this method that the
rigid amorphous part is closely connected with the crystallites and disappears
together with their melting.
6.8.3
Advanced Characterization with the TMDSC Method
Recently TMDSC {see Sect. 2.4.2} has become relevant to polymer science. This
method offers some more advantages and possibilities because of the additional
modulated {reversing} signal, which contains information about the dynamics of
the investigated process. The glass transition {see Sect. 6.4} is one example of
a relaxation process which yields different signals depending on the frequency
of modulation. The glass transition temperature depends on frequency in a
characteristic manner and the resulting "activation diagram" (see Sect. 6.4.5)
makes it possible to get a better insight into the glass process of polymers. Of
course, this is mainly of interest for theoretical and model investigations and has
hardly any relevance for more practical applications.
However, the situation is quite different for other time dependent processes
which may happen in polymers. In particular melting and crystallization are
generally time dependent and TMDSC offers the possibility to follow the change
of, say, the apparent heat capacity in time. The crystallization is normally not
seen in the modulated heat flow rate signal, since the crystallization process is
not influenced by the rather small temperature changes of the modulation,
but with crystallization the heat capacity changes as well and the latter can pre-
cisely be calculated from the heat flow rate amplitude (see Sect. 6.1.5) which, in
turn, enables the calculation of the degree of crystallinity and its change in time.
With TMDSC it is even possible to measure the change of the heat capacity in a
quasi-isothermal mode and follow the crystallization this way which opens the
possibility to investigate crystallization kinetics (cf. Sect. 6.3).
Chemical reactions (e. g., polymerization or curing) are generally also not
visible in the modulated signal of TMDSC, whereas any change of the heat
capacity, such as the vitrification or devitrification (see Sect 6.3.6), can easily be
determined from that signal, even in a quasi-isothermal mode. This enables
valuable insight into the reaction conditions and possible changes from concen-
tration control to diffusion control and the setting up of a TTT diagram for the
reaction in question (see Sect. 6.3.6). The modulated (reversing) signal may even
serve as baseline (see Sect. 5.1) to determine the enthalpy of reaction from the
underlying (conventional) DSC curve (see Sect.6.2). This is in particular of great
importance if the heat capacity changes during the reaction.
In the melting region of polymers the situation is more complicated, as the
melting process contributes to the modulated signal in a complex manner. On the
one hand the temperature modulation, though small in general, influences the
melting and thus the heat flow rate quite well and on the other hand melting of
polymers needs time which makes the modulated amplitude strongly frequency
dependent (see Sect. 3.3.2). Both effects together with the reversible melting,
which is confirmed for a large number of polymers (Wunderlich, 2003), affect
the modulated signal and gives rise to the so-called excess heat capacity (see
Sect. 3.3.2). In other words, the heat capacity determined from the modulated
heat flow rate signal, the apparent heat capacity, is larger than the normal vibra-
tional cp in the melting region and depends on frequency (see Fig. 6.43). From the
20
JK" g"
, S ~H,
1 11.5 ~H,
) IS mH,
10 ~ .is mH7
Icpl OS( 10 Kim,"

5
0
90 100 110 120 130 O( 150
tempera t ure ~
Fig.6.43. The magnitude of the apparent heat capacity of nascent ultra-high-molar-mass

polyethylene (UHMMPE) from TMDSC runs at different frequencies ({30: 0.1 K min-I, m: ca. 4 mg,
WTA: 3.6 mK rad S-I). A common DSC run (dotted; fJo: 10 K min-I, m: 3 mg) and literature
values of the amorphous (upper dashed line) and crystalline (lower dashed line) polyethylene
(ATHAS, 2002) are added for comparison
Fig.6.44. Excess heat capaci-

10
ties (solid curves) calculated
with Eq. (3.22) from the
measured apparent heat ca-
pacity (dotted curves) from
quasi-isothermal TMDSC
measurements at different
temperatures (UHMMPE;
mass: 2-3 mg, frequency:
12.5 mHz, tp : 80 s, TA : 53 mK)
0
0 50 100
time
----
150 min 250
Fig.6.45. Activation dia- 10 00 0 1<>1

gram with the time constants 000 (>
of three different processes []
00
evaluated from the excess
heat capacity curves of Fig. 100
(>
6.44 via a multi-exponential
fit. From the slopes activation c:
E (>(>(> c
energies can be calculated .......
!-- 10
(><>
[]
c: []
rP
[][]
1
24 25 26 K-' 27
liT 10 J
frequency dependence of the excess heat capacity conclusions concerning the

time law of the delayed melting [causing the well known "superheating" of poly-
mers (Wunderlich, Czorny, 1977)], can be drawn (Toda, Saruyama, 2001).
If we perform quasi-isothermal temperature-modulated measurements in
the melting region it is possible to separate irreversible contributions from
reversible excess cp and evaluate the time dependence quantitatively. Figure 6.44
shows as an example the result of such measurements in the case of a certain
ultra-high-molar-mass polyethylene. The calculation of excess cp and further
evaluation with proper fit functions (e. g., exponential functions) make it possi-
ble to separate different irreversible processes with different time constants in
the melting region of this polymer (Fig. 6.45). Additional evaluation of the phase
angle signal yields information about the exo- or endothermic character of the
respective process (see Sect. 3.3.2 case 2).
In summary, it may be stated that DSC and in particular TMDSC offer very
powerful tools in polymer science and technology. The method is indispensable
6.9 Purity Determination of Chemicals 241
for characterization of polymers and for quality control. Modern equipment

yields reliable results in short time and is often superior to other thermoanalyt-
ical methods.
6.9
Purity Determination of Chemicals
The theory of purity determination using DSC is based on the thermodynamics

of two-component systems. The simplest and most widely used theory presup-
poses an eutectic mixture of ideal behavior. For this case we find in textbooks of
chemical thermodynamics that the mole fraction x of one component (defined
here as the pure one) can be calculated as follows:
1 T i1H(T)
lnx=- f --dT
R Ttrs T2
(R, gas constant: 8.31441 J mol- 1 K- 1; i1H(T): phase transition enthalpy at tem-
perature T of the pure component in question, Ttrs : phase transition temperature
of this completely pure component}.
In the case of a negligible temperature dependence of the phase transition

enthalpy, the integral can be solved and we get
lnx = i1H (_1__ ~)

R Ttrs T
or x = exp (i1H(_1 _~))
R Ttrs T
This equation serves often to calculate ideal eutectic phase diagrams and the
melting behavior of slightly impure materials. If a change is made to the mole
fraction of the (eutectic) impurity Ximp = 1 - x, and In (1- Ximp) is approximated
(for small values of Ximp) by -Ximp as well as the product Ttrs Tby Trrs> we obtain:
Ttrs - T
Ximp (T) = i1H - - 2 - (6.2)
R T trs
This is the well-known van't Hoff equation which relates the decrease in the
melting temperature of the impure component to the amount of impurity in-
volved. In principle, any melting of an eutectic mixture starts at the eutectic tem-
perature and ends at the temperature T which is related to the composition of
the mixture via Eq. (6.2) (see Fig. 6.46) and may serve to determine the amount
of impurities.
The theoretical heat flow rate curve of eutectic mixtures can be calculated
from the phase diagram: an example is shown in Fig. 6.46 for different concen-
trations. 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
real curves measured in a power compensation DSC (Fig. 6.47) differ consider-
mW
1 3
...c 2
oL-________~==~~~~~~~L_~
340 345 350 K 355
temperature ..
Fig.6.46. Calculated melting peaks of an ideal eutectic system, normalized to unit area
(1: 0.20%; 2: 0.55%; 3: 1.05%; 4: 2.03%; 5: 4.67%; 6: 6.65% impurity)
ably from the theoretical shape because of the smearing due to the finite ther-
mal conductivity of the DSC (cf. Sect. 5.4). As a result 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. 6.41 a) relative to the initial
composition is the same as the relation between the heat of fusion used up to
that temperature and the total heat:
Ximp (T) = Qr == F (T) (6.3)

Ximp Qtot
The right-hand side of this equation is the quotient of the 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. (6.3) into Eq. (6.2) furnishes an equation according to which the
impurity Ximp can be determined from the slope of a plot of the temperature
versus lIF(T) which should be a straight line. In practice this is not the case
because of the difficulty in deciding when melting actually starts and the limited
heat transfer to the sample. Commercial software generally corrects the mea-
sured 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
6.9 Purity Determination of Chemicals 243
0.6
mW
2
0.4
f 3
~
C1J
c
OJ
....
~
0
;;::
~
c
0.2
C1J
..r:::.
0.1
0.0
320 330 K 360
tempera ture ~
Fig.6.47. Measured melting peaks of an eutectic system, normalized to unit area (impurities
the same as in Fig. 6.46)
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. These problems
restrict the method and limit the certainty of the results to an extent which nor-
mally is not accepted in certification procedures which are an obligation in
pharmacy and food industries.
Attempts have been made to improve the method and to bypass the smearing
of the DSC results. Bader et al. (1993) presented a method which determines
the impurity from the shape of the measured curve at the start of the melting
process, just 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 a mixture of two components and can
be calculated (see Fig. 6.46). If the (normalized) heat flow rate equation in ques-
tion is solved for the purity x as a function of temperature, a straight horizontal
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. The
same type of plot can be calculated if we insert the really measured heat flow
rates (normalized and with the baseline subtracted, see Fig. 6.47) instead of the
theoretical heat flow rate curve. This should yield a horizontal line at the right
purity value as well, but the result does not come up to what had been expected
because of the difficulty of precisely determining the baseline and thus the be-
ginning of the peak. It has, however, been shown (Bader et al., 1993) that some
minor parameter variations (well below the uncertainty of the measurement)
correct this problem satisfactorily. However, the problem, that even this method
starts from an ideal eutectic mixture is not solved this way.
Another method to bypass both the thermodynamic and the thermal lag
problem of the DSC was suggested by Sarge et al., 1988 and Stosch et al., 1998.
They simulate the expected melting curve, using an improved thermodynamic
approach (which includes even non-ideal mixing behavior) and a certain mathe-
matical model for the influence for the unavoidable thermal lag of the DSC, and
fit this to the measured melting curve with the purity as running parameter.
However, even this simulation is not free from thermodynamic model assump-
tions and the obtained results are thus uncertain and not better than those from
the other methods.
In summary, it can be stated that the accuracy of all purity evaluations is
largely influenced by:
- The thermodynamic model used (ideal or real mixing, eutectic or non-eutectic
behavior). At least the activity coefficients of the impurities are never known
and cannot be used for exact calculation of the phase diagram and thus the
expected melting curve.
- The smearing of the measured curve, which falsifies the calculated curve in
one way or the other from the shape of the peak.
Bearing in mind that most theories of purity determination start from the as-
sumption of eutectic mixtures, which is actually only one special case, the accu-
racy of every determination of purity by DSC should be considered to be a
rather limited approximation to the truth. Therefore it can only be used for
relative quality control, e. g., to compare different charges of the same chemical
relative to its purity, but never for absolute measurements. The latter is nowa-
days much more precisely done with HPLC and other modern analytical meth-
ods. That's why the importance of the purity determination with DSC recently
dropped distinctly. Nevertheless the method has still some importance because
it is fast and easy and yields at least approximate values of possible impurities.
7 Evaluation of the Performance of a Differential
Scanning Calorimeter
DSC furnishes information on temperature and heat flow rates (or respectively,
heat). Whether it is suited to solve the respective problem depends on the effi-
ciency of the instrument. The characteristic data of the DSC which describe the
instrument unambiguously 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 comparison with other DSCs possible.
A distinction can be made between:
1. the characterization of the complete instrument,
2. the characterization of the measuring system,
3. the characterization of the results of a 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
numerical values are suggested. The characteristic terms in question are the fol-
lowing:
- noise,
- repeatability,
- linearity,
- time constant,
- sensitivity,
- resolution.
246 7 Evaluation of the Performance of a Differential Scanning Calorimeter
1W} ~WNfol: ~WMvlRMS

m
time
Fig.7.1. The various definitions of noise (according to Hemminger, 1994). pp peak-to-peak
noise, p peak noise, RMS root-mean-square noise, ([Jrn measured heat flow rate
- The noise of the measured signal (given, for example, in ]lW) is indicated in
different ways (see Fig. 7.1):
- as peak-to-peak noise (pp): maximum variation of the measured signal in
relation to the mean signal value,
- 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 in-
stantaneous deviations of the 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., purge gas). The signal-to-noise
ratio is decisive for the smallest heat flow rate detectable ("heat flow rate res-
olution" or "detection limit" of the 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 ]lW/cm3 or ]lW/g) which allows the smallest detectable
heat flow rate to be estimated for a given sample.
The noise ("short-time noise") can be measured as follows:
- In the desired operating mode (isothermal or scanning mode at a specified
heating rate) and at the temperature of interest, the measured signal is am-
plified to such a degree (most sensitive measuring range) that the noise is
clearly 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 ]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 zeroline). It gives an impression of the distur-
bances from the environment to which the sensors are subject. It deter-
mines the maximum possible signal-to-noise ratio.
As regards the signal to noise ratio (and the resolution of DSCs, see
below) a comparative test of 22 different models of DSCs (from 8 manu-
facturers) using 4,4'-azoxyanisole (which shows a reversible liquid crystal
to isotropic liquid transition at ca. 134C with a small heat of transition
of ca. 2 Jg- 1) was performed under defined experimental conditions (van
Ekeren et aI., 1997). The signal to noise (pp) ratio was determined, the re-
ported values showed a large spreading, even for one distinct DSC model.
A real comparison with an understanding of the observed differences
between classes of DSCs or even within one class (e. g., power compensat-
ed) was not possible. The repeatability for a particular DSC was reported
to be good.
- The repeatability indicates the closeness of the agreement between the results
of successive measurements of the same measurand carried out under the
same conditions of measurement (according to the International Vocabulary
of Basic and General Terms in Metrology, 1994; cf. Sect. 7.3).
DSCs can be characterized with the aid of repeatability by measuring a sig-
nificant DSC quality (e. g., extrapolated peak onset temperature, peak area,
course of the zeroline etc.) several times under the same conditions on sam-
ples of the same kind. Examples:
- The repeatability of the zeroline is determined by measuring 4 or 5 zero lines
over the whole temperature range at medium scanning rate and super-
imposing the curves. The temperature-dependent range of the deviation
from the mean" zeroline (absolute or in percent) 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 zeroline (for
example when cp is measured).
- The repeatability of the peak area and of the baseline is determined by meas-
uring for example, the peak area caused by the melting of a pure metal sever-
al times under the same conditions. (Either only repeat measurements with-
out 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 su-
perimposed. The range related to the mean area shows the repeatability of
the peak area (heat of fusion) determination. A separate baseline repeatabili-
ty can be determined (according to the procedure with the zeroline, 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 re-
sults of measurements carried out on a sample using different instruments in
---------
An
CD
'-J CD Q) C0
mn --P-t-+~
t m ---<>
temperature ..
Fig.7.2. Linearity (according to Hemminger, 1994). <Pm measured signal (heat flow rate),
Al to A4 peak areas, belonging to sample masses ml to m4 (ml < m2 < m3 < m4). In the par-
tial figure: ratio Anlmn with the range of uncertainty of the area determination, - - - -devi-
ation from ideal linearity (here only for very small (1) and very big (4) sample masses)
different laboratories (round robin, inter-laboratory comparison; cf. Sect. 7.3).

The absolute or percentage deviation of the 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 of measurement or bias (see be-
low) of certain instruments.
- The linearity of a DSC describes the relation between the measured heat flow
rate IPm (signal) and the true heat flow rate IPtrue or between the measured heat
Qm (peak area A) and the true heat Qtrue. Linear relations are: IPtrue = K</J . IPm
or Qtrue = K Q Qm.
To determine the linearity means to show the validity of these relations.
If the proportionality factor K depends on parameters, e. g., the temperature,
the mass of the sample, the heating rate etc the linearity is only given for the
parameters in question. In such cases the dependence of K on parameters
must be measured and represented graphically. Only with ideal linearity K is
a constant. Examples, as the temperature dependence K = K (1') or the de-
pendence of the factor KQon the mass of the sample (KQ= KQ (m)),have been
shown in the calibration chapter. Figure 7.2 presents such a linearity test.
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 r of the DSC measuring system is the time interval in which
- after a sudden (step-like) change of a constant heat flow rate produced at
the place of the sample - the measurement signal has reached the e-th part of
its final value (i. e., changed 63.2 %). In other words: if a constant heat flow
rate, that causes a signal 4im,o, is switched off at the moment to, one time in-
terval dater the signal has decreased to 4im ,o/e (i. e. to 26.8 % of its original
value or by 63.2 %) (Fig. 7.3). The time constant determines to what degree
two successive thermal events (peaks) are recorded clearly separated from
each other (time resolution, see "resolution" below). Knowledge of the time
constant is important for the desmearing of DSC curves (see Sect. 5.4.).
The time constant of a DSC can be determined as follows (cf. Ulbrich,
Cammenga, 1993 and L6blich, 1994):
- When an electric calibration heater can be installed in the sample crucible,
4im,o, a constant heat flow rate is switched on until a constant signal is ob-
tained; after switching-off of the heater, the descending curve is evaluated
graphically (ace. to Fig. 7.3) or analytically:
4i (t) = 4im,o . exp (- tlr)
d4im(t) 4im (t)
=---
dt
so for any time t*, r is determined by the quotient of the measured heat
flow rate and its derivative
r= -
d4im(t*)/dt
Fig.7.3. Determination of
the time constant T (accord- ., ., . ., .. ., ., . r - - - - - - ,
ing to Hemminger, 1994).
After a constant measure-
ment signal 4>m,o has been
reached, the constant heating
current is switched off at the
time to' After the time inter-
val T, the measurement
t
signal 4>m,o has dropped
to 4>m,o/e
hme
Fig.7.4. Determination of
the time constant rwith the
melting peak of a pure
substance (according to
Hemminger, 1994)
time
t* should not be too close to the switching-off time to because further,

smaller time constants ( T2, T3 etc.) may there be superimposed on the max-
imum time constant Tmax (see below).
- When no calibration heater can be installed in the crucible, the descending
section of a transition peak (when a pure substance melts or crystallizes)
is evaluated by the "tangent method" (cf. Fig. 7.4). Several tangents are
plotted to the descending section and the time constants T) Tmax are
graphically determined from the intersections with the linearly interpolat-
ed baseline. Towards the end of the descending curve, a constant value T
(= M between t* and the intersecting tangent on the baseline) results: this
is the (greatest) time constant Tmax which describes the thermal inertia of
the measuring system in good approximation.
The (usually sudden) solidification of the supercooled melt of a pure
substance (e. g., pure tin) causes a pulse-like heat production in the sample.
The resulting heat flow rate curve is the so-called apparatus (or Green's)
function (see Sects. 3.3.3 and 4.7), this is ideally suitable to determine the
apparent time constant of the system with the tangent method.
- The sensitivity of a measuring system is defined as the ratio between the
change of the measurement signal and the change of the measured quantity
that creates the signal. In DSC systems the measured quantity is the heat flow
rate, and the signal output is usually an electric voltage, thus the ratio of f,U
(for example in }lV) and f, rP (for example in m W) yields the sensitivity given
in }lV/mW.
- The resolution of a DSC measuring system describes its ability to clearly iden-
tify overlapping thermal events (peaks) as separate ones.
Van Ekeren et aI., 1997, suggested to use 4,4'-azoxyanisole to measure the reso-
lution of DSCs. This substance shows two transitions close to one another: a solid
7.3 Characterization of the Results of a Measurement: Uncertainty Determination 251
Fig.7.5. Definition of "reso-

lution" as ratio alb, for
details see text (according to
van Ekeren et ai., 1997)
1
\ I
\ I
\ I
\ I
temperQ ture ..
to liquid crystal transition at about 117 C (heat of transition approx. 120 Jg- 1)
and a liquid crystal to isotropic liquid transition at approx.134 C (heat of tran-
sition approx. 2 Jg- 1). The two transitions are only 17 K apart and differ by a
factor of 60 as regards their heats of transition. The degree of separation of the
two transition peaks was used to quantify the "resolution" defined as the ratio
between the distances baseline - shoulder (a) and baseline - peak maximum (b)
(Fig. 7.5). From 22 DSCs of 8 different manufacturers tested with this method,
most of the DSCs gave values between 0.10 and 0.35, i. e., a peak that was between
10 and approx. 3 times larger than the distance baseline-shoulder (a). But a few
DSCs gave - under the experimental conditions of the test runs - values from
0.70 to 0.90 which makes a separation - and even a detection - of such narrow
peaks almost impossible.
7.3
Characterization of the Results of a Measurement: Uncertainty Determination
The quality of measurement results must be evaluated prior to their being in-
terpreted and published or used for further calculations (crystallinity, kinetics
etc.). This evaluation is made on the basis of the data characterizing the effi-
ciency of the measuring system (cf. Sect. 7.2).
To make measurements comparable, stating the measurement uncertainty is
a must. This is in particular true when the laboratory in question is "certified"
internally or by an administrative body. To come to an uncertainty measure dif-
ferent possibilities exist. Methods which use statistical tools applied to inter-
laboratory comparisons with terms like "random" and "systematic" measure-
ment errors, which are used to describe deviations from the "true value" that is
never known. A more recent, metrologically based, access to the assessment of
the accuracy of a measurement is described by the "Guide to the Expression of
Uncertainty in Measurement" (referred to as GUM, 1995; see Kessel, 2002). GUM
specific terms are "standard measurement uncertainty", "uncertainty budget",
"probability distribution" and others (cf. Sect. 7.3.2).
The description of the variability of results of measurements is usually sepa-

rated into three parts. According to the "International Vocabulary of Basic and
General Terms in Metrology", 1994, these parts are:
Repeatability characterizes the closeness of the agreement between the
results of successive measurements of the same measurand carried out under
the same conditions of measurement (cf. Sect. 7.2). Usually this means a multiple
use of the same DSC, the same measurement procedure, and the same reference
material for calibration by the same observer at the same conditions over a short
period of time. Furthermore, it is known that the results of DSC measurements
can depend on contact resistances between sample and furnace and the posi-
tioning of sample and reference sample inside of the furnace. A realistic deter-
mination of the repeatability of DSC-measurements should therefore include
a suitable procedure to consider these effects. In most cases the specification of
the repeatability is intended as a measure of the variability of DSC measure-
ments within the specific laboratory.
Reproducibility characterizes the closeness of the agreement between the
results of measurements of the same measurand carried out under changed con-
ditions of measurement (cf. Sect. 7.2). For DSC measurements these changed
conditions may include the measuring instrument, the observer, the method of
measurement, the evaluation procedure or software, the calibration standard,
conditions of use, or the time. In most cases the specification of a reproducibil-
ity is intended as a measure of the variability of DSC measurements between dif-
ferent laboratories.
Accuracy is an often used qualitative term which describes the closeness of
the agreement between the result of a measurement and the true value of the
measurand. But in all cases the true value of the measurand is not known.
Error is the difference between the result of a measurement and the true value
of the measurand. A systematic error is a mean that would result from an infinite
number of measurements of the same measurand carried out under repeatabil-
ity conditions minus the true value of the measurand.
Bias means a systematic error of the indication (reading) of a measuring
instrument. This means in our case a systematic difference of the results of DSC
measurements on well known (and stable) materials in comparison to the cor-
responding reference (literature) values.
Correction is a value added to or subtracted from the uncorrected result of a
measurement to compensate for the systematic error. In some cases a correction
is carried out by multiplication or division of the uncorrected results of a mea-
surement with a correction factor to compensate for the systematic error.
A further source of confusion is the different meaning of "estimation"
in common sense and in statistics. From the statistical point of view the
word "estimation" means the application of a well defined mathematical pro-
cedure (e.g., the calculation of the standard deviation of the mean of a popu-
lation) and not the making of a rough guess. Although in many cases experi-
ence based uncertainty estimations give reasonable results, a fair comparison
of methods and instruments and a credible specification of the uncer-
tainty should be based on a comprehensible (experimental and mathematical)
basis.
Table 7.1. Results of an inter-laboratory study (according to ASTM EI269-01)
Material Repeatability Reproducibility Bias

(95% confidence limit) (95% confidence limit)
}~%
Diphenyl ether +0.95%
Linear polyethylene -1.1%
Indium +0.8% resp. + 1.8 %a
a Depending on the literature value used.
7.3.1
Black Box Method
The black box method starts from the assumption that the single apparatus (e. g.,
a DSC) is unknown, as far as the uncertainty of the measurements is concerned.
The experimental basis for the specification of uncertainty are inter-laboratory
studies on different materials. These inter-laboratory studies are evaluated by
means of (classical) statistics including the comparison with reliable literature
data. The uncertainty determination with the black box method yields mainly
three values: the repeatability, the reproducibility and the bias. The method is
rather easy but time consuming and needs a lot of cooperating laboratories to
give reliable results, it is applied, e.g., by ASTM.
As an example, the specific heat capacities of diphenyl ether, linear polyethylene
and indium at 67 DC were determined within an inter-laboratory study (ASTM
EI269-0l) resulting in the relative uncertainty contributions given in Table 7.1.
7.3.2
GUM Method
With the "Guide to the Expression of Uncertainty in Measurement" (GUM,
1995) internationally accepted guidelines for evaluating and expressing the
uncertainty of measurement results were published. This new approach for the
unification of different uncertainty concepts is based on the principle of maxi-
mum (information) entropy and the Bayesian-statistics. It is somewhat compli-
cated, but really helpful to determine the total uncertainty in a well defined and
comparable way. In times where international comparability and certification
of laboratories becomes more and more common, it is likely the method of the
future. The main difference to the above method, used by ASTM, is that the
instrument (DSC) is not considered as a black box. Therefore, a basic part of the
uncertainty determination according to GUM is the mathematical model used
(the evaluation equation with the corrections). It must include all relevant parts
which contribute to the uncertainty of the measurement. The procedure to
create an uncertainty budget consists of four steps:
1. Development of the models of evaluation for all relevant physical quantities
and their relation to each other in mathematical terms. All significant correc-
tions should be identified and applied.
2. Preparation of the input data for the evaluation, including a list of all sources
of uncertainty and their contributions.
3. Calculation of the measurement result by means of the model in question and
calculation of the uncertainty propagation.
4. Averaging of the results of the measurements and corresponding uncertainties.
As an example the procedure for the determination of the uncertainty of specific
heat capacity measurements of a glassy ceramic by means of DSC according
to GUM is outlined in the following (the detailed description can be found in
Rudtsch,2002).
Considering a typical DSC measurement consisting of the empty (O), calibra-
tion sample (Ref) and sample (S) measurement. The specific heat capacity of the
sample Cs (at a given temperature) can be calculated by the comparison of the
heat flow rates into the sample <Ps and into the calibration sample <PRef according
to (cf. Sect. 6.1):
(7.1)
According to this simple model the main sources of uncertainty are the masses
of sample ms and calibration sample mRef> the heat flow rates <P of the three
experiments and the specific heat capacity of the calibration material CRef.
In the next step additional quantities of influence and the corresponding cor-
rections or uncertainty contributions must be identified and applied. One item
is the consideration of the influence of the temperature measurement. Because
of the fact that the heat capacities of sample, reference sample and empty system
are functions of temperature and that their temperature dependencies differ
from each other, a corresponding uncertainty contribution must be considered.
One possibility for the inclusion of this effect is the introduction of different
heat flow rate corrections L\ <P i for empty, sample and calibration sample mea-
surement. In doing this, the slope of a measurement run and a temperature
difference (i. e., the uncertainty of temperature measurement) are used for the
determination of L\ <Pi.
In the following discussion it is assumed that the DSC was calibrated (accord-
ing to Chapter 4) by measuring the phase transition temperatures of different
calibration materials as a function of the scanning rate and subsequent extrap-
olation to zero scanning rate. The resulting uncertainty consists of three parts,
the repeatability of the calibration procedure L\Tcalib , the uncertainty of the
phase transition temperature of the calibration material L\ Tmat and the uncer-
tainty caused by the interpolation between the different phase transition tem-
peratures of the calibration materials L\ Iiin .
Even if a careful temperature calibration of the instrument has been carried
out, the reading of the temperature sensor doesn't agree with the sample tem-
perature except for the limit of zero scanning rate. Therefore it is necessary to
find a method to determine and correct for the corresponding bias in the tem-
perature measurement during the scanning operation of the instrument. This
temperature difference depends on the scanning rate, the thermal resistances
between sample and temperature sensor and the thickness and thermophysical
properties of the sample itself (see Sect. 3.1 and 3.3.3). For some types of in-
struments a determination of the temperature difference between the tempera-
ture sensor of the DSC and the mean sample temperature is possible by the
measurement of the thermal lag 61iag (Richardson, 1997). The results of the
thermal lag analysis have shown that for power compensation DSCs the cor-
responding temperature difference is typically between 0.5 K and 3 K. Thus the
resulting model equation for the sample temperature Ts reads
Ts = T + ~Tcalib + 6Tmat + ~Tlin - 6T1ag (7.2)
In the next step the standard uncertainties u of all input quantities Xi of Eq. (7.2)
(e. g., 6 Tmat or ~ Tcalib ) must be determined. As an example, the standard
uncertainty of the temperature calibration u (~Tcalib) can be determined by
repeated measurements of the melting temperature (extrapolated peak onset
temperature at zero scanning rate). If normal (Gaussian) distribution can be
assumed for the measurand the standard deviation of the experiments is a mea-
sure of the standard uncertainty u (~Tcalib) and corresponds to a coverage
probability of approximately 68 % (coverage factor k = 1). In most cases a stan-
dard coverage factor k = 2 should be used. The assigned expanded uncertainty
U of a measurement corresponds to a coverage probability of approximately
95%. Table 7.2 shows the resulting uncertainty budget of the temperature mea-
surement at 500C. As a result, the expanded uncertainty (k = 2) of the temper-
ature measurement is U (Ts) = 1.2 K.
A second additional uncertainty contribution comes from the heat capacity
of the calibration sample. If NIST (National Institute of Standards and Technol-
ogy, Gaithersburg, USA) SRM-720 sapphire as calibration material is used, the
uncertainty can be taken from the certification report. But in many cases (non-
certified) sapphire from other sources is used. In this case an additional contri-
bution to uncertainty 6cRef is included which takes care of a possible specific
Table 7.2. Uncertainty budget for the temperature determination example
Input quantity a Estimate a Standard Probability Sensitivity Uncertainty

Xi Xi uncertainty distribution coefficient contribution
U(Xi) in K oTs Ui(Y) in K
oXi
T 500C
6 Tma, O.OK 0.1 normal 1.0 0.1
ll.Tcalib 1.9 K 0.3 normal 1.0 0.3
ll.Tjin O.OK 0.5 normal 1.0 0.5
6Tjag 1.0 K 0.2 normal 1.0 -0.2
Ts 500.9C 0.6
a Input quantities Xi are measurable quantities like the temperature T or the thermal lag 6Tjag.
The corresponding quantity in statistics is the estimator. The value of the estimator is the esti-
mate or best estimate of the unknown quantity and corresponds to the result of a measurement
(e. g., arithmetic mean).
heat capacity difference between the used calibration material and the NIST
SRM -720 sapphire. The modified model equation for the calculation of the spe-
cific heat capacity of the sample is then given by
+ ~IPs) - (IPo + ~IPo)

(IPs
(7.3)
(IPRef + ~ IPRef) - (IPo + ~ IPo)
In the next step the standard uncertainties U (Xi) of all input quantities Xi (e. g.,
mRef> ms, IPs, etc) in Eq. (7.3) must be determined. The corresponding contribu-
tion to the standard uncertainty Ui (cs) associated with the output estimate Cs
can be calculated according to
oCs
Ui (cs) = oXi U (Xi) (7.4)
For uncorrelated input quantities the standard uncertainty of the specific heat
capacity (combined uncertainty) is given by
(7.5)
Special investigations have shown that the relative uncertainty of the specific
heat measurement between 0 C and 600C doesn't depend on temperature. The
uncertainty budget for the specific heat capacity measurement has therefore as
an example been calculated at a temperature of 250C. Table 7.3 gives a sum-
mary of the results of this analysis.
The results of Table 7.3 show that the major source of uncertainty of DSC
measurements are the heat flow rate measurements of sample and calibration
sample. As a result, the final relative expanded uncertainty (k = 2) of the specif-
ic heat capacity determination is U (cs)/ Cs = 1.5 % with a reported result at 250C
of Cs = (1.015 0.016) J. g-l. K- 1
Table 7.3. Uncertainty budget for the specific heat capacity determination example
Input Estimate Standard Probability Uncertainty

quantity Xj uncertainty distribution contribution
Xj u(Xj) Uj (y)/J . g-l . K- 1
mRef 55.41 mg 0.006mg rectangular 0.1. 10-3

CRef 1.060 J . g-l . K- 1 0.001 J . g-l . K- 1 rectangular 1.2 . 10-3
DcRef 0.0 J . g-l . K- 1 0.003 J . g-l . K- 1 rectangular 2.9.10- 3
4>Ref 11.09 mW 0.05mW rectangular 4.6. 10- 3
l1~ef -0.009mW 0.003mW rectangular 0.3 . 10-3
ms 52.29 mg 0.006mg rectangular 0.1. 10-3
4>s 1O.13mW 0.05mW rectangular 5.3. 10- 3
l14>s -0.01 mW 0.003 mW rectangular 0.3. 10- 3
4>0 1.2mW O.lmW rectangular 1.5. 10- 3
l14>0 -0.002mW 0.0004mW rectangular 0.004.10- 3
7.4 Check List for DSCs 257
7.4
Check List for OSCs
The following check list serves to collect the essential characteristic data of a
DSC.1t may be used to
- ask the manufacturers for the values,
- establish a guideline when characteristic data are measured in one's own lab-
oratory or in the manufacturer's laboratory,
- compare different types of DSCs,
- compare the DSC data with the values required to investigate a problem.
Manufacturer:
Type of measuring system: o heat flux disk-type
o heat flux turret-type
o heat flux cylinder-type
o power compensation
Special features:
Sample volume (standard crucible): ..... mm3
Atmosphere
(vacuum?, which gases? pressure?):
Temperature range: from ..... to ..... C or K
Scanning rates: from ... to ... K/min steps: ... K min-I
Zeroline repeatability: from ..... flW (at ..... C)
to ..... flW (at ..... 0c)
Peak area repeatability: ..... % (at ..... 0c)
Total uncertainty for heat: ..... % (at ..... oC)
Extrapolated peak onset
temperature repeatability: ....... K (at ..... 0c)
Total uncertainty for temperature: ..... K (at ..... 0c)
Scanning noise (pp) at ..... K min-I: from ..... flW (at ..... 0c)
to ..... flW (at ..... 0c)
Isothermal noise (pp): from ..... flW (at ..... 0c)
to ..... flW (at ..... 0c)
Time constant with sample: ..... s
Additional facilities:
Appendix 1
Comparison of Heat Flux Differential Scanning Calorimeters

and Differential Thermal Analysis Instruments
The measurement signal generated in DTA instruments is the temperature dif-

ference AT between sample and reference sample. The measuring system com-
prises temperature sensors and a suitable holder or support to mechanically
fix the temperature sensors and the sample containers. A defined, constant
and good thermal contact must be established between temperature sensors and
sample containers. Strict repeatability of the measuring system over the whole
temperature range must be ensured; the system's properties are reflected in the
shape of the zeroline (i. e., curve measured with the measuring system empty or
with empty crucibles).
Depending on the desired measurement temperature, sample volume and
sensitivity, a choice can be made from among various DTA measuring systems.
In the following, the two contrasting basic types - the block-type measuring sys-
tem and the measuring system with free-standing crucibles - will be described.
DTA: The block-type measuring system
The following is characteristic of this type of measuring system (cf. Fig. ALl):
- The temperature difference AT is measured directly in the sample substances.
As a result, the temperature sensors are subject to the attack of the sample and
reference sample substance.
- The duration (resolution with time) and the height (sensitivity) of the mea-
surement 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 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 quantita-
tive evaluation of the heat in question is therefore impossible.
The generally high sensitivity and the rapid, almost instantaneous "response" to
sample reactions are advantages of this measuring system.
260 Appendix 1
I1T
s R
4
a
r
I1T
b
time ...
Fig. A 1.1. a Instrument for Differential Thermal Analysis (DTA) with block-type measuring
system, b Schematic measured curve L'1 T (t) (exo up).
1 "block" with cavities to take up sample and reference sample substance, 2 furnace, 3 differ-
ential thermocouple, 4 programmer and controller, S sample substance, R reference sample
substance, Tp furnace temperature, t time
Comparison of Heat Flux Differential Scanning Calorimeters 261
OTA: Measuring system with free-standing crucibles
This measuring system (Fig. A1.2) 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 thermocou-
ples for the purpose of AT measurements. The thermocouple junction is in ther-
mal 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
time
t
6.T
b
Fig. A 1.2. a Instrument for Differential Thermal Analysis (DTA), measuring system with free-
standing crucibles, b Schematic measured curve ,iT(t) (endo down).
1 crucible,2 furnace, 3 differential thermocouple, 4 programmer and controller, S sample sub-
stance, R reference sample substance, Tp furnace temperature, t time
262 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
samples, a choice can be made from among a great number of container types
(crucibles), 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
(temperature ranges), the adjustment of different purge gases (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
unfavourable effects of the block material (resolution, sensitivity) or the (isolated)
individual crucibles (delayed, "smeared" temperature measurement).
Heat flux OSC
In contrast to the majority of the DTA instruments of rather simple design,

DSCs should be capable of being calibrated for heat measurement. An unam-
biguous and repeatable assignment of the measurement signal proper, ~ T, to the
measurand d P must be ensured (~ P = PFS - PFR , 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 measuring 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 approxima-
tion when the temperature field of the measuring system does not primarily de-
pend 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 sig-
nal itself, dT, will also be small. As a consequence, the noise of the whole mea-
suring 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 param-
eters. This leads to systematic error sources, and the uncertainty of the result is,
therefore, much larger than the repeatability error of the measurements. Very
careful calibration, 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 var-
ious fields of application. In the following, a classification system for calorime-
ters and a couple of examples of different types of calorimeters will be present-
ed. 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

particular characteristics. Various classification systems are reasonable and
practicable (cf. Hemminger, H6hne, 1984; Rouquerol, Zielenkiewicz, 1986). It
may sometimes be difficult to arrange calorimeters in proper order in a rela-
tively simple system, and such a classification may even be impossible. Classi-
fication systems covering the entire field of calorimetry tend to become very
sophisticated, end in itself and are rather useless for practical applications.
When thermodynamic 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 satisfac-
torily 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 devel-
oped which will suffice to discuss characteristic features and error sources by
groups.
Criteria for the classification are the following:
1. The Principle of Measurement
1.1 Measurement of the energy required for compensating the heat to be
measured (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).
264 Appendix 2
2. Mode of Operation
2.1 Isothermal
2.2 Isoperibol
2.3 Adiabatic } static modes
2.4 Scanning of surroundings

2.5
2.6
2.7
Isoperibol scanning
Adiabatic scanning
Temperature-modulated
} dynamic modes
3. Construction Principle
3.1 Single calorimeter
3.2 Twin or differential calorimeter
Most of the calorimeters can be classified by the above-mentioned criteria. All
combinations of 1,2 and 3 are, of course, not possible as some of the criteria are
incompatible.
Calorimeters which are used today will be presented in the following accord-
ing to this classification. As the classic calorimeters will be dealt with only
briefly, reference will be made to the literature for further information on con-
struction details, error sources, methods for the evaluation of the measured val-
ues, and ranges of application of these instruments. The DSCs are treated in
closer detail in several sections in this book (see Chapters 2 and 3).
Examples of Colorimeters
The aim of this chapter 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 of the heat to be measured is
"suppressed", 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 of the 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 heating or Peltier's cooling). The compensation by means of reversible
expansion or compression of an ideal gas was described by Ter Minassian,
Milliou, 1983.
It is an advantage of all compensating methods that the measurements are
carried out under quasi-isothermal conditions and that heat leaks do not, there-
fore, represent important error sources. Moreover, in the case of electric com-
pensation, no calibrated temperature sensor is required for the measurement
but only a sensitive thermometer which controls the compensation power of a
controller so that the temperature remains constant.
Appendix 2 265
1st example: "Ice calorimeter"
1. 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 OC
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 dif-
ference of ice and water). The measurements which Ginnings and Corruccini
carried out at the National Bureau of Standards (NBS) during the late forties of
the last century were counted among the important applications of an ice
calorimeter. They measured heat with an uncertainty of about 0.02 %, the tem-
perature of the sample at the moment of its being dropped into the calorimeter
lying between 100C and 600 0c,
The liquid phase to gaseous phase transition was also made use of ("boil-off"
calorimeter), 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 magni-
tude higher than in solid to liquid transitions.
Phase transition calorimeters are relatively simple to construct and allow pre-
cise 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 sub-
stance (ice, liquid); the heat of transition must be known. (In general, the
calorimeter is calibrated electrically.)
2nd example: "Dissolution calorimeter"
1. Compensation of the heat to be measured with the aid of electric energy

2. Isothermal
This calorimeter serves to measure heats of solution. A container fIlled with

water is equipped with a stirrer, a controllable electric heater and a sensitive
thermometer (Fig. A2.2). At constant temperature, an endothermically 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 ther-
mometer activates the compensation control circuit so that, if possible, no tem-
266 Appendix 2
1 - -.....[> X
o
o
o
o
o
o
x
t
time ...
Fig. A2.1. a Ice calorimeter (Bunsen, 1870), b Measured curve (according to Hemminger,
1994).1 sample, 2 sample container (receiver), 3 ice, 4 water,S ice-water mixture, 6 mercury,
7 capillary tube, x position of the mercury meniscus. The displacement Lix* of the meniscus
is proportional to the heat Q exchanged with the sample (positive for an endothermic effect).
t.x* is determined taking the "pre-period" and the "post-period" into account
Appendix 2 267
v
5
2
1
~l
time
Fig. A2.2. a "Compensation calorimeter" (Bronsted, 1906), b Measured curve (according to
Hemminger, 1994). 1 sample (salt), 2 water, 3 stirrer, 4 electric heater, 5 temperature sensor,
6 controller, Pel electric heating power. The area below the measured curve Pel (t) corresponds to
the compensation heat and (at constant temperature) also to the endothermic heat of solution
Q of the sample substance, which has been searched:
t2
Q= J Pel (t) dt
t,
268 Appendix 2
perature change due to reaction heat takes place. As the Peltier power for the
compensation of exothermic effects cannot be measured with sufficient accura-
cy, 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 (con-
trolled). 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 compensated, the heating power is decreased with the
Peltier power remaining unchanged (see, for example, Christensen et al., 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, tem-
perature control is delayed by the heat transfer processes, and it is very difficult
to obtain spatially homogeneous temperature fields by electric heating.
The advantages of these quasi-isothermal calorimeters consist in the simple
and very precise measurement of the electric compensation energy and the pos-
sibility 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"

2. Adiabatic scanning
In these calorimeters (Fig. A2.3), the temperature program is preset. So much

electrical 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 sur-
roundings as well as possible to the temperature of the sample (or sample con-
tainer) (adiabacy). Calorimeters of this type allow the heat capacity to be mea-
sured with high accuracy (uncertainty :S0.1 %) (cf. 4th example of heat-accu-
mulating calorimeters).
The following is valid for the heat capacity C(T):
C (T) = Pel, 1 - Pel,2

(dT/dt: preset)
dT
dt
or
C(T) = P" (( ~~f (~:)J

Appendix 2 269
5
T
6. T = 0 6
t
f!1
--- ~ ____---CD
time
Fig. A2.3. 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(t), Pel (t) is controlled so that (dT/dth = (dT/dth
(outside the peak). b Measured curve (according to Hemminger, 1994). CD curve measured
with the calorimeter empty, @ curve measured with the sample placed in the calorimeter,
sample transition between tJ and t2 (peak)
270 Appendix 2
(1: run with the calorimeter empty, 2: run with the sample placed in the
calorimeter, Pel = i U electrical heating power, dTldt heating rate at the respec-
tive 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 tem-
perature remains constant during transition in spite of the fact that heating pow-
er is continued to be supplied, the heat of transition is directly determined from
the integral of the heating power over the transition time (Qtrs = Ii (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 of this type are used for the accurate, absolute and direct mea-
surement of heat capacities and heats of transition (cf., for example, Nolting,
1985; Kagan, 1984; an extension to a low-temperature system has been described
by Rahm, Gmelin, 1992).
4th example: "Power compensation DSC"

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

calorimeter (see Figs. 2.5, 2.6) comprises two identical measuring systems (twin
principle), one containing the sample, the other the reference sample. The tem-
perature difference between the two systems is measured. In a 1st approxima-
tion, disturbances from the surroundings have the same effect on both measur-
ing systems and therefore cancel out with respect to the temperature difference.
The individual sample supports (microfurnaces) are heated separately so that
they comply with the given temperature-time program. When there is ideal ther-
mal symmetry between the two measuring systems, the same heating power is
required for sample and reference sample. When additional heat is released or
consumed during sample transition (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 exothermic or endothermic transition
process. The measured signal is the temperature difference I1T (deviation from
the set value) to which the compensation heating power I1P is proportional:
I1P = k . 11 T. Calorimeters of this type ("power compensation DSC" or "differen-
tial power compensation scanning calorimeter, DPSC" or "dynamic power
difference calorimeter") are widely used (cf., for example, Watson et al., 1964;
Hemminger, Hohne, 1984). They are discussed at full length in Sects. 2.2 and 3.2
of this book.
Appendix 2 271
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 substance and a "calorimeter substance" with which the heat to be de-
termined is exchanged. This temperature change is measured. When the change
is not too large, it is proportional to the amount of heat exchanged. The propor-
tionality factor must be determined by calibration with a known amount of heat.
5th example: "Drop calorimeter"
1. 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 thermo-
stat ("isoperibol": uniform surroundings). The heat Q to be measured is ex-
changed with the "calorimeter substance" and the temperature change !:.T(t) is
measured (Fig. A 2.4).
The following is valid: Q = Ccal !:.T* (cf. Fig. A2.4c).
The proportionality factor Ccal is the heat capacity of the calorimeter substance
(the liquid in Fig. A2.4a) 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 dif-
ference 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 ... ).
Otherwise, it cannot be determined by calibration and represents an error source.
In order to guarantee strict repeatability of the heat exchange,!:. T 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!:. T* used to calculate Qmust be determined 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 !:.T(t)
curve measured before and after the sample has been placed into the calorimeter
(cf. also 2nd 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 500C; sample mass between 10 and 100 g).
In "aneroid" drop calorimeters (Fig. A2.4b), 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 2000C)
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 sub-
stance; this results in heat of reaction being released.
272 Appendix 2
-3
4
T '~ T
.;------;-Ir----. ~ c::=J I'--_~
__ L _
a b
(I
I
1 L1 r" I
r I
c
time ~
Fig. A 2.4. "Drop calorimeter" in isoperibol mode of operation a with liquid, b with aneroid
calorimeter substance and c measured curve (according to Hemminger, 1994). 1 sample,
2 calorimeter substance: (a) liquid, (b) solid, 3 temperature sensor, 4 stirrer, 5 radiation
shields. The temperature change L'lT* of the calorimeter substance has to be determined from
the measured curve according to defined rules, taking the "pre-period" and the "post-period"
into account
Q = Cca1 . L'l T* is valid.
Ccal calibration factor ("heat capacity" of the calorimeter substance and of other calorimeter
components)
Appendix 2 273
1
T
b
time ~
Fig. A2.S. a Adiabatic bomb calorimeter, b Measured curve (according to Hemminger, 1994).
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,
S temperature sensor, 9 controller
274 Appendix 2
6th example: ''Adiabatic bomb calorimeter"

measured
2. Adiabatic
In the case of the adiabatic bomb calorimeter (Fig. A2.5), the "combustion
bomb"which is (usually) filled with oxygen at high pressure immerses in water
(calorimeter substance). The water temperature is continually measured before
and after the electric ignition. With the aid of a controller, the temperature of
the surroundings is always adapted to this temperature (adiabatic jacket). The
temperature change with time in the surroundings serves as measurement
signal; it is determined from the drift of the measured curve before the ignition
(pre-period) and after the ignition (post-period) (as with the "drop calorime-
ter", see 5th example).
Bomb calorimeters of this type (usually automated) are widely used to mea-
sure the calorific value of solids or liquids under standardized conditions. In
general, they are calibrated with benzoic acid. Their uncertainty of measure-
ment 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 mea-
sured; bomb calorimeters 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"

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. A2.6). The temperature
difference between the medium flowing in and the medium flowing out is pro-
portional to the heat transferred. Calorimeters of this type are used to deter-
mine the calorific value of fuel gases; they are calibrated with gases of known
calorific value (e. g., methane) so that the specific heat capacity of the heat -con-
veying 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 princi-
ple 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 solu-
tions, or they are used to measure reaction heats. Two reacting solutions are, for
example, mixed in a reaction tube (Fig. A2.7). The uniform temperature of the
Appendix 2 275
T = const.
2 "--~-4
3
5
5
T,
3 I
----~@J~------
Fig. A2.6. "Gas calorimeter". 1 burner in which the gas to be measured is burnt, 2 heat ex-
changer, 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 con-
veying medium whose temperature increase T2 - TJ is measured. The "calorific value" (the
combustion heat) is proportional to T2 - TJ
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 differ-
ence 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 in-
stalled in the reaction tube. A sophisticated example of a flow calorimeter with
stimulated cardiac muscle developing heat in the perfused tube has been
described by Daut et aI., 1991. [Flow calorimeters are also designed as twin
calorimeters to cancel out the influence of the isoperibol surroundings (heat
leaks) ].
8th example: "Adiabatic calorimeter"
1. Measurement of the temperature change of a substance due to the supply of

a known amount of heat
2. Adiabatic
This type of calorimeter (Fig. A2.8) is not designed to measure an unknown

heat; instead, a well-known, electrically generated heat Q = Wei serves to change
the sample temperature by ~ T. The temperature of the 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 determine the phenomenological coefficient of the heat supplied, Q,
and the temperature change, ~T, of a substance (at constant pressure): the heat
capacity Cp (n. The following is valid: Cp = Q/~ T.
276 Appendix 2
1
j
I
enter of
I1 ......- - -
readand
b
time
Fig.A2.7. a "Flow-mix calorimeter", b Measured curve. 1,2 reactants, 3 reaction product,
4 temperature sensor. As in the case of the "drop calorimeter", the temperature change llT*
required to determine the mixing or reaction heat is obtained from the pre- and post-period
(affected by a drift) of the measured curve (T2 - T1) (t)
Appendix 2 277
t
T
.--.---- switch on P"

b
time
Fig. A2.8. a Adiabatic calorimeter, b Measured curve (according to Hemminger, 1994). 1 sam-
pie, 2 heatable sample container, 3 adiabatic jacket, 4 controller,S programmer and controller,
Pol electric heating power. When the constant measurement temperature T is adjusted, the elec-
tric heating energy WeI is supplied to the sample, increasing the sample temperature by 6 T*
278 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;
Jakobi et al., 1993).
Knowledge of the heat capacity and its temperature dependence is of utmost
importance in solid state physics and thermodynamics. In practical application,
I1T is kept as small as possible in order to determine the temperature depend-
ence 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 calori-
meters), the heat exchange with the isoperibol surroundings is kept low to make
the measured signal 11 T 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 de-
scribes the intensity of the exchange is then proportional to a heat flow rate ~ and
not to a heat. This allows time dependences of a transition to be observed on the
basis of the ~ (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 be-
tween 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 (see Sect. 2. 7. 7)

A metal, ceramic or quartz glass disk with the sample and the reference sample
(or the pans) positioned on it symmetrical to the center, is placed into a furnace
(Fig. 2.1a). 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 measure-
ment signal itself must, of course, reflect the sample properties). The signal
I1T is measured on the solid heat conductor (disk) between the supports for
sample and reference sample (I1T = Ts - TR ).
Appendix 2 279
Note: The disk-type DSC is not really a differential (twin) calorimeter, as

there is an influence of sample temperature changes on the reference tempera-
ture via the center part of the solid disk (cross-talk).
DSC with Turret-Type Measuring System (see Sect. 2.1.2)
In this variant of the heat flux calorimeters the turret like sample and reference
supports are soldered on the bottom plate of a silver furnace. This way the dis-
advantage of thermal coupling of sample and reference sample is get rid off and
we have a real differential calorimeter (twin). The function principle is other-
wise similar to the disk-type DSC (see above).
DSC with Cylinder-Type Measuring System (see Sect. 2.1.3)
In another type of heat flux DSC, the two cylindrical containers for sample and
reference sample are connected with the furnace ("surroundings") by one ther-
mopile each (Fig. 2.3). 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 (ilT) is proportional to the difference between the heat flow rates from the
furnace to the sample (ilTps - <Pps) and from the furnace to the reference sam-
ple (ilTpR - <PPR).
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 calorimeters
temperature change of the adiabatic
calorimeter substance
measured (heat-
accumulating calorimeters) Dynamic: Twin (differential)
scanning of surroundings calorimeters
Measurement of the heat isoperibol scanning
flow rate between adiabatic scanning
calorimeter substance and temperature-modulated
surroundings due to the heat
to be measured (heat-
exchanging calorimeters)
References
Adam G, Gibbs JH (1965) J Chern Phys 43: l39-146

Agrawal AJ (1989) J Polym Sci: Part B, Polym Phys 27: 1449-1461
Alexandrov YI (2002) Thermochim Acta 382: 55-64
Aras L, Richardson MJ (1989) Polymer 30: 2246
Archer DG (1993) J Phys Chern Ref Data 22: 1441-1453
Archer DG (1997) J Chern Eng Data 42 :281-292
Archer DG,Kirklin DR (2000) Thermochim Acta 347:21-30
Archer DG,Carter RW (2000) J Phys Chern B 104:8563-8584
Archer DG (2002) J Chern Eng Data 47: 304-309
ASTM: American Society for Testing and Materials publishes Standard Test Methods and
Standard Practices (ASTM website: www.astm.org)
ATHAS (2002) Internet Databank http://web.utk.edu/-Athas/databank/Intro.Html
Bader RG, Schawe JEK, Hahne GWH (1993) Thermochim Acta 229: 85-96
Barton JM (1980) Polymer 21 :603-606
Barton JM (1985) In: Dusek K (ed) Advances in Polymer Sciences, Springer, Berlin, Heidelberg,
New York: 111-154
Batzer H, Kreibich UT (1982) Angew Makromol Chern 105: 113-l30
Becker R (1976) Z Phys Chern-Leipzig 257:667-677
Becker R (1977) Z Phys Chern-Leipzig 258:953-966
Becker R (1978) Faserforschung und Textiltechnik 29:361-385
Beezer AE (1980) (ed) Biological micro calorimetry. Academic Press, London
Bedford RE, Bonnier G, Maas H, Pavese F (1986) Metrologia 33: l33-154
Blankenhorn K, Hahne GWH (1991) Thermochim Acta 187:219-224
Bonora S, Fini G, Piccirilli B (2000) J Therm Anal Cal 61 :731-743
Borchardt HJ, Daniels F (1957) J Am Chern Soc 79:41-46
Borde B, Cesaro A (2001) J Therm Anal Cal 66: 179-195
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, Darfier HD (1983) Colloid & Polymer 261 :423-426
Bransted IN (1906) Z Phys Chern 56:663-665
Bros JP (1989) J Less Common Metals 154:9-30
Brown ME, Dollimore D, Galwey AK (1980) In: Bamford CH, Tipper CF (eds) Comprehensive
chemical kinetics, vol 2, Reactions in the solid state. Elsevier, Amsterdam
Brun M, Lallemand A, Quinson JF, Eyraud C (1977) Thermochim Acta 21 : 59-88
Bruzzone G (1985) Thermochim Acta 96:239-258
Bunsen RW (1870) Ann Phys 141: 1-14
Burger A, Ramberger R (1979) Mikrochim Acta 2:259-271 and 273-316
Burnham AK (2000) Thermochim Acta 355: 165-170
Calvet E (1948) CR hebd 226: 1702-1704
Cammenga HK, Eysel W, Gmelin E, Hemminger W, Hahne GWH, Sarge SM (1993) Thermochim
Acta 219:333-342
282 References
Cammenga HK, Figura LO, Zielasko B (1996) J Therm Anal 47 : 427-434

Cassel B (2000) Amer Lab, January 2000, 23-26
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 1. Plenum, New York: 555-589
n,
Christensen Johnston HD, Izatt RM (1968) Rev Sci Instrum 39: 1356-1359
Chung TS (1984) J Appl Polym Sci 29:4403-4406
Couchman PR, Karasz FE (1978) Macromolecules 11: 117-119
Cowie JMG, Ferguson R (1986) Polym Commun 27:258-260
Crowley KJ, Zografi G (2001) Thermochim Acta 380: 79-93
Daut J, GroB T, Elzinga G (1991) Thermochim Acta 193:269-280
de Bolt MA,EasteaIAJ,Macedo PB,Moynihan CT (1976) J Am Chern Soc 59: 16-21
DengY,Martin GC (1994) Macromolecules 27:5141-5146
Ding MS,Xu K,JowTR (2000) J ThermAnal Cal 62: 177-186
Dale A, Lythall C,Aucott J, Sayer C (2002) Thermochim Acta 382:47-54
Della Gatta G, Beezer AE, Richardson MJ, Schiraldi A (eds) (2000) Special Issue: Calibration of
Calorimeters; Thermochim Acta 347: 1-132
Ditmars DA, Ishihara S, Chang SS, Bernstein G, West ED (1982) J Res Nat Bur Stand 87:
159-163
Doelman A, Gregges AR, Barrall EM (1977) In: Porter RS, Johnson JF (eds) Analytical Calorime-
try, vol 4. Plenum Press, New York
Domen SR (1987) In: Kase KR, Bjarngard BE,Attix FH (eds) The dosimetry of ionizing radia-
tion, vol II. Academic Press, Orlando, Florida
Doyle CD (1961) J Appl Polym Sci 5 :285-292
Doyle CD (1962) J Appl Polym Sci 24:639-642
Duncan JC, van de Weide JG, Wetton RE (1991) Kautschuk und Gummi 44:768-770
Eckardt N,Flammersheim HJ,Cammenga HK (1998) J ThermAnal52: 177-185
Eligehausen S, Sarge SM, Ohlschliiger G, Cammenga HK (1989) J Therm Anal 35:515-526
Ellerstein SM (1968) In: Porter RS, Johnson JF (eds) Analytical Calorimetry, Vol I. Plenum,
New York
Engeln-Miillges G, Reuter F (1996) Numerik-Algorithmen; VOl-Verlag, Diisseldorf
Fan M, Miiller F, Wilsmann W (1993) Thermochim Acta 224: 19-32
Faivre C, Bellet D, Dolino G (1999) Eur Phys J B 7: 19-36
Flammersheim HJ (1981) Wiss Beitr, Friedrich Schiller Universitiit Jena, 160-164
Flammersheim HJ (1988) J Therm Anal 33 : 55-65
Flammersheim HJ (1998) Thermochim Acta 310: 153-159
Flammersheim HJ (2000) Thermochim Acta 361 :21-29
Flammersheim HJ, Eckardt N, Kunze W (1991) Thermochim Acta 187:269-274
Flammersheim HJ, Eckardt N, Opfermann J (1993) Thermochim Acta 229:281-287
Flammersheim HJ, Klemm E (1985) Acta Polymerica 36:443-446
Flammersheim HJ,Kunze W (1998) J Therm Anal Cal 52: 125-130
Flammersheim HJ, Opfermann J (1999a) Thermochim Acta 337: 141-148
Flammersheim HJ, Opfermann J (1999b) Thermochim Acta 337: 149-153
Flammersheim HJ, Opfermann J (2001) Macromol Mat Eng 286: 143-150
Flammersheim HJ, Opfermann J (2002) Thermochim Acta 388:389-400
Flynn JH (1974) Thermochim Acta 8:69-81
Flynn JH (1990) J Thermal Anal 36: 1579-1593
Flynn JH (1993) Thermochim Acta 217: 129-149
Flynn JH, Wall LA (1966) J Polym Sci, Pt B (Pol Letters) 4:323-328
Foltz MF, Coon CL, Garcia F, Nichols AL (1994) Propellants Explos Pyrotech 19: 19-25
Fox TG, Flory PJ (1950) J Appl Phys 21 :581-590
Freeman ES, Carroll B (1958) J Phys Chern 62:394-397
Friedman HL (1964) J Polym Sci, Polym Symp C 6: 183-195
References 283
Gallis HE, Bougrioua F, Oonk HAJ, van Ekeren PJ, van Miltenburg JC (1996a) Thermochim
Acta 274:231-242
Gallis HE, van Miltenburg JC (1996b) Thermochim Acta 274:223-230
Gallis HE, van Miltenburg JC, Oonk HAJ, van der Eerden JP (1996c) Thermochim Acta
286:307-319
Gallis HE, van Ekeren PJ, van Miltenburg JC, Oonk HAJ (1999) Thermochim Acta 326:
83-90
Galwey AK, Brown ME (1998) In Chapter 3: Handbook of thermal analysis and calorimetry,
vol 1, principles and practice, Elsevier, Amsterdam 147-224
Galwey AK, Brown ME (2002) Thermochim Acta 386: 91-98
GEFTA, Gesellschaft fUr Thermische Analyse (German Society for Thermal Analysis)
http://www.gefta.uni-freiburg.de
Gilmour IW,Hay IN (1977) Polymer 18:281-285
Ginnings DC, Corruccini RJ (1947) J Res Nat Bur Stand 38:538-591
Giron D (1999) J Therm Anal Cal 56: 1285-1304
Gmelin E (1987) Thermochim Acta 110: 183-208
Gmelin E, Rodhammer P (1981) J Phys E: Sci Instrum 14:223-228
Gmelin E, Sarge SM (1995) Pure Appl Chern 67: 1783-1800
Gmelin E (1997) Thermochim Acta 304/305: 1-26
Gmelin E, Sarge SM (2000) Thermochim Acta 347: 9-13
Gobrecht H, Hamann K, and Willers (1971) J Phys E: Sci Instrum 4:21-23
Godovsky YuK, Hohne GWH (eds) (1994) Special Issue: Deformation Calorimetry of Polymers;
Thermochim Acta 247,1: 1-128
Goldberg RN, Weir RD (1992) Pure Appl Chern 64: 1545-1562
Gray AP (1968) In: Porter RS, Johnson JF (eds) Analytical Calorimetry, Vol 1, Plenum Press
New York
Grewer T (1987) Thermochim Acta 119: 1-15
Grewer T, Steinbach J (eds) (1993) Special Issue: Safety Techniques; Thermochim Acta 225,
2:143-301
Gr0nvold F (1967) Acta Chern Scand 21: 1695-1715
Gr0nvold F (1993) J Chern Thermodyn 25: 1133-1144
GrunenbergA, Henck JO,Siesler HW (1996) Int J Pharm 129: 147-158
GUM: Guide to the Expression of Uncertainty in Measurement (1st ed. 1993; corrected and
reprinted 1995) International Organization for Standardization, Geneva
Gunn SR (1971) J Chern Thermodyn 3: 19-34
Hale A, Macosko C, Bair HE (1991) Macromolecules 24:2610-2621
Hancock BC, Shamblin SL (200!) Thermochim Acta 380:95-107
Hanitzsch E (1991) Thermochim Acta 187:275-281
Hansen LD (ed) (2000) Special Issue: Biothermodynamics: Molecular, Organismal and Ecolog-
ical; Thermochim Acta 349: 1-184
Hay IN (1992) Progr Colloid Polym Sci 87:74-77
Hemminger W (ed) (1988) PTB Bericht PTB-W -37, Wirtschaftsverlag, Bremerhaven
Hemminger W (1994) "Calorimetric Methods" in Calorimetry and Thermal Analysis of
Polymers (Mathot VBF, ed), Hanser, Munich
Hemminger W, Schenborn K-H (1982) In: Miller B (ed) Thermal Analysis, Proc 7th ICTA, Vol
1, Wiley, New York 156-162
Hemminger W, Hohne GWH (1984) Calorimetry-Fundamentals and Practice; VCH Verlag,
Weinheim
Hemminger WF, Cammenga HK (1989) Methoden der Thermischen Analyse. Springer, Berlin,
Heidelberg, New York
Hemminger W, Sarge SM (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
Hesekamp D (1998) Chern Eng Technol21: 149-153
284 References
Heublein G, Hallpap P, Hauptmann S,Mann G (1984) EinfUhrung in die Reaktionstheorie. VEB

Deutscher Verlag fUr Grundstoffindustrie, Leipzig
Hodge 1M (1987) Macromolecules 20:2897-2908
HoffH (1991) Thermochim Acta 187:293-307
Hahne GWH (1983) Thermochim Acta 69: 175-197
Hahne GWH (1997) Thermochim Acta 304/305:209-218
Hahne GWH (1999a) Thermochim Acta 330:45-54
Hahne GWH (1999b) Thermochim Acta 332: 115-123
Hahne GWH, Blankenhorn K (1994) Thermochim Acta 238:351-370
Hahne GWH, Cammenga HK, Eysel W, Gmelin E, Hemminger W (1990) Thermochim Acta
160: 1-12
Hahne GWH, Glaggler E (1989) Thermochim Acta 151 :295-304
Hahne GWH, Schawe JEK (1993) Thermochim Acta 229:27-36
Hahne GWH,Schawe JEK,Schick C (1993) ThermochimActa 221: 129-137
Hahne GWH, Merzlyakov M, Schick C (2002) Thermochim Acta 391 :51-67
Horie K,Hiura H,Sawada M,Mita I,Kambe H (1970) J Polym Sci,Al Polym Chern 8: 1357-1372
Hu W, Wunderlich B (2001) J Therm Anal Cal 66:677-697
Huguenin FGAE, Klein MT (1985) Ind Eng Chern, Prod Res Dev 24: 166-171
Hunger B, Klepel 0, Kirschhock C, Heuchel M, Toufar H, Fuess H (1999) Langmuir 15: 5937-5941
Hutchinson JM (1990) J Polym Sci, Part B, Polym Phys 28:2127-2161
Hutchinson JM (1992) Progr Colloid Polym Sci 87:69-73
Hutchinson JM, Ruddy M (1988) J Polym Sci, Part B, Polym Phys 26:2341-2366
Hutchinson JM, Ruddy M, Wilson MR (1988) Polymer 29: 152-159
Hyde CG, Jones MW (1960) Gas Calorimetry, 2nd ed; Ernest Benn, London
International Vocabulary of Basic and General Terms in Metrology (1'1 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
Ishikiriyama K, Todoki M, Min KH, Yonemori S, Noshiro M (1996) J Therm Anal 46: 1177 -1189
ISO Guides from International Organization for Standardization http://www.iso.ch
ISOIIEC 17025: General requirements for the competence of testing and calibration laborato-
ries (1999)
ITS-90: The International Temperature Scale of 1990 (1990) Metrologia 27:3-10, 107
Jaeschke M (2002) Thermochim Acta 382:37-45
Jakobi R, Gmelin E, Ripka K (1993) J Therm Anal 40: 871-876
Jin Y, Wunderlich B (1993) Thermochim Acta 226: 155-161
Jorda R, Wilkes GL (1988) Polym Bull 20:479-485
Journal of Thermal Analysis (1998) Special issue 54
Kagan DN (1984) In: Maglic KD, Cezairliyan A, PeletskyVE (eds) Compendium of Thermo-
physical Property Measurement Methods 1. Plenum, New York: 461-526
Kaisersberger E (ed) (1997) Special Issue: Coupling Thermal Analysis and Gas Analysis
Methods; Thermochim Acta 295: 1-186
Kalachandra S, Turner DT (1987) J Polym Sci, Part B, Polym Phys 25: 1971-1979
Karkanas PI, Partridge IK,Attwood D (1996) Polym Int 41: 183-191
Keenan MR (1987) J Appl Polym Sci 33: 1725-1734
Kemp RB (1993) Thermochim Acta 219: 17-41
Kemp RB, Hunsen LD, Lamprecht I, von Stockar U (eds) (1998) Special Issue: Biothermody-
namics: From Human Beings to Molecules; Thermochim Acta 309: 1-202
Kessel W (2002) ThermochimActa 382: 1-16
Kilian HG, Hahne GWH (1983) Thermochim Acta 69: 199-219
Kissinger HE (1956) J Res Nat Bur Stand 57:217-221
Kissinger HE (1957) Anal Chern 29: 1702-1706
Klee J, Grutzner RE, Harhold HH, Flammersheim HJ (1998) Macromol Chern Phys 199: 1603-
1611
References 285
Klee J, Harhold HH, Flammersheim HJ (1990) Angew Makromol Chern 178:63-83

Klemm E, Flammersheim HJ, Martin R, Harhold HH (1985) Angew Makromol Chern 135: 131-
138
Kovacs AJ (1963) Fortschr Hochpol Forsch 3: 394-507
Kovacs AJ,Aklonis n, Hutchinson JM, Ramos AR (1979) J Polym Sci, Polym Phys Edn 17: 1097-
1162
Kovacs AJ, Hutchinson JM (1979) J Polym Sci, Polym Phys Edn 17: 2031-2058
Kuhnert-Brandstatter M (1996) Pharmazie 51 :443-457
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, Hahne GWH (eds) (1991) Special Issue: Calorimetry in the
Biological Sciences: Thermochim Acta 193: 1-452
Landau RN (ed) (1996) Special Issue: Reaction Calorimetry; Thermochim Acta 289: 101-369
Lau Sv, Suzuki H, Wunderlich B (1984) J Polym Sci Polym Phys Edn 22 :379-405
Lavoisier AL, de Laplace PS (1784) Histoire de l'Academie Royale des Sciences, Annee 1780:
355
Liu G, Zhang L, Qu X, Li Y, Gao J, Yang L (2001) J Therm Anal Cal 65:837-846
LObbecke S, Keicher T, Krause H, Pfeil A (1997) Solid State Ionics 101 :945-951
LOblich K-R (1985) ThermochimActa 83:99-106
LOblich K-R (1994) Thermochim Acta 231: 7-20
L'vov BV (2001) ThermochimActa 373:97-124
L'vov BV (2002) Thermochim Acta 386: 1-16
Maciejewski M (2000) Thermochim Acta 355: 145-154
Maglic KD, Cezairliyan A, Peletsky VE (eds) (1984) Compendium of Thermophysical Property
Measurement Methods 1, Plenum, New York
Malavasic T, Osredkar U, Anzur I, Vizovisek I (1986) J Macromol Sci Chern A 23: 853-860
Marsh KN (ed) (1987) Recommended Reference Materials for the Realization of Physiochem-
ical Properties. Blackwell Scientific Publications, Oxford
Martin DL (1987) Rev Sci Instrum 58(4):639-646
MathotVBF (1984) Polymer 25:579-599
MathotVBF, Pijpers MFJ (1983) J Thermal Anal 28:349-358
Mathot VBF, Pijpers MFJ (1989) Thermochim Acta 151: 241-259
Mathot VBF (ed) (1994a) Calorimetry and Thermal Analysis of Polymers. Hanser, Munich
Mathot VBF (ed) (1994 b) 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, Kettrup A (1993) J Thermal Anal 39: 141-1155
Merzlyakov M, Schick C (2001a) Thermochim Acta 377: 193-204
Merzlyakov M, Schick C (20mb) Thermochim Acta 380:5-12
Merzlyakov M, Hahne GWH, Schick C (2002) Thermochim Acta 391 : 69-80
Mijovic J, Lee CH (1989) J Appl Polym Sci 37:889-900
Min BG, Stachurski ZH, Hodgkin JH (1993) Polymer 34:4908-4912
Montserrat S (1992) J Appl Polym 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, Easteal AJ, de Bolt MA, Th.cker J (1976) JAm Ceram Soc 59: 12-16 and 16-20
Narayanaswamy DS (1971) JAm Ceram Soc 54:491-498
Nefatti R,Apekis L, Rault J (1998) J Therm Anal 54:741-752
Neffati R, Rault J (2001) Eur Phys J B 21 :205-210
Nalting J (1985) Thermochim Acta 94: 1-15
O'Neill MJ (1964) Anal Chern 36: 1238
Oetting FL (1970) J Chern Thermodynamics 2:727-739
Ohrbach K-H, Matuschek G, Kettrup A (1987) Thermochim Acta 121 :87-96
286 References
Okui N (1990) Polymer 31 :92-94

Opfermann I (1992) Technical report concerning the peak separation. Netzsch-Geratebau-
GmbH, Selb, Germany
Opfermann I (2000) I Therm Anal Cal 60:641-658
Opfermann I, Giblin F, Mayer I, Kaisersberger E (1995) Am Lab 27:34-41
Opfermann I, Kaisersberger E, Flammersheim HI (2002) Thermochim Acta 391: 119-127
Opfermann I, Wilke G, lung I, Ludwig W, Hagen S, Gebhardt M, Kaisersberger E (1991) Therm
Analysenverf in Ind u Forschung. Wiss Zeitschr Friedrich Schiller Universitat lena 51-68
Oscarson IL, Izatt RM (1992) In: Rossiter BW, Baetzold RC (eds) Physical Methods of Chem-
istry, 2nd Ed. Wiley, New York
Ozawa T (1965) Bull Chern Soc Ipn 38: 1881-1886
Ozawa T (1970) I Therm Anal 2:301-324
Ozawa T (2001) I Therm Anal Cal 64: 109-126
Parrish WmR (1986) Fluid Phase Equilibria 29: 177-192
Perez I, Cavaille IY, Calleja RD, Ribbeles ILG, Pradas MM, Greus AR (1991) Makromol Chern
192:2141-2161
Perrenot R, de Valliere P, Widmann G (1992) I Therm Anal 38:381-390
Petit 1- L, Sicard L, Eyraud L (1961) CR hebd 252: 1740-1742
Petrie SEB (1972) I Polym Sci A2 Polym Phys 10: 1255-1272
Peyser P (1983) I Macromol Sci-Phys, B22: 185-196
Pijpers MFI et al. (2002) Macromolecules 35:3601-3613
Plotnikov VV, Brandts 1M, Liu LN, Brandts IF (1997) Analytical Biochemistry 250: 237 - 244
Podesva I, Prochazka 0 (1979) Makromol Chern 180:2525-2530
PoeBnecker W (1990) I Therm Anal 36: 1123-1139
PoeBnecker W (1993) Thermochim Acta 229:97-109
Pomposo lA, Eguiazabal I, Calahorra E, Cortazar M (1993) Polymer 34:95-102
Preston-Thomas H (1976) Metrologia 12:7-17
Preston-Thomas H (1990) Metrologia 27:3-10, 107
Privalov PL, Plotnikov VV (1989) Thermochim Acta 139 :257-277
Rabinowitch E (1937) Trans Faraday Soc 33: 1225-1233
Rahm V, Gmelin E (1992) I Therm Anal 38:335-344
Rai US, Rai RN (1998) I Therm Anal 53:883-893
Rai US, Shekhar H (1993) I Therm Anal 39:415-428
Ramos AR, Hutchinson 1M, Kovacs AI (1984) I Polym Sci, Polym Phys, Edn 22: 1655-1695
Reading M, Elliott D, Hill VL (1993) I Thermal Anal 40 : 949-955
Reading M, Luget A, Wilson R (1994) Thermochim Acta 238:295-307
Regenass W (1985) Thermochim Acta 95: 351-368
Reichelt I, Hemminger W (1983) Thermochim Acta 69: 59-70
Richardson MI (1976) Plastics and Rubber, Materials and Applications 1: 162-167
Richardson MI (1989) In: Booth C, Price C (eds) Comprehensive Polymer Science, Vol I, Perg-
amon, Oxford: 867-901
Richardson MI (1992a) In: Maglic KD, Cezairliyan A and PeletskyVE (eds) Compendium of
Thermophysical Property Measurement Methods 2. Plenum, New York: 519-545
Richardson MI (1992b) Pure Appl Chern 64: 1790-1800
Richardson MI (1993) Thermochim Acta 229: 1-14
Richardson MI (1997) Thermochim Acta 300: 15-28
Richardson MI, Savill NG (1975a) Polymer 16:753-757
Richardson MI, Savill NG (1975b) Thermochim Acta 12:213-220
Roduit B (2000) Thermochim Acta 355: 171-180
Roles KA, Wunderlich B (1993) I Polym Sci, Part B, Polym Phys 31 :279-285
Rosenthal LA (1961) Rev Sci Instrum 32: 1033
Rossini RD (ed) (1956) Experimental Themochemistry, Vol I. Interscience Publishers, New York
Rouquerol I, Zielenkiewicz W (1986) Thermochim Acta 109: 121-137
Rudtsch S (2002) Thermochim Acta 382: 17 -25
References 287
Ryan ME, Dutta A (1979) Polymer 20:203-206

Sabbah R (1999) Special Issue: Reference Materials for Calorimetry and Differential Thermal
Analysis; Thermochim Acta 331 :93-209
Sabbah R, Xu-wu A, Chickos JS, Planas Leitao ML, Roux MV, Torres LA (1999) Thermochim
Acta 331 :93-204
Sacchetti M (2001) J Therm Anal Cal 63:345-350
Sangster J (1999) J Phys Chem Ref Data 28:889-930
Sarge SM (1991) Thermochim Acta 198:323-334
Sarge SM, Cammenga HK (1985) Thermochim Acta 94: 17-31
Sarge SM, Bauerecker S, Cammenga HK (1988) Thermochim Acta 129: 309-324
Sarge SM, Gmelin E, Hohne GWH, Cammenga HK, Hemminger W, Eysel W (1994) Thermochim
Acta 247: 129-168
Sarge SM (ed) (1997) PTB-Bericht PTB-ThEx-1, Wirtschaftsverlag,Bremerhaven
Sarge SM, Hemminger W, Gmelin E, Hohne GWH, Cammenga HK, Eysel W (1997) J Therm
Anal 49: 1125-1134
Sarge SM, Hohne GWH, Cammenga HK, Eysel W, Gmelin E (2000) Thermochim Acta 361: 1-20
Saruyama Y (1999) Thermochim Acta 330: 101-107
Sauerbrunn SR,Crowe B, and ReadingM (1992) Am Lab (08):44-47
Schaffer B, SzakaIy S, LOrinczy D (2001) J Therm Anal Cal 64: 659-669
Schawe JEK (1993) Thermochim Acta 229:69-84
Schawe JEK (1995 a) Thermochim Acta 260: 1-16
Schawe JEK (1995b) Thermochim Acta 261: 183-194
Schawe JEK, Schick C (1991) Thermochim Acta 187:335-349
Schawe JEK, Schick C, Hohne GWH (1993) Thermochim Acta 229:37-52
Schawe JEK, Schick C, Hohne GWH (1994) Thermochim Acta 244:49-59
Schawe JEK, Hohne GWH (1996) J Thermal Anal 46:893-904
Schawe JEK (1996) personal communication
Schick C (1999) Laborpraxis 5:92-94
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
Schick C, Wurm A, Mohamed A (2001) J ColI Polym Sci 279:800-806
Schick C, Wurm A, Mohamed A (2003) Thermochim Acta 396: 119-132
Schick C et al. (2002) NATAS proceedings: 112-117
Schmidt C et al. (1994) Thermochim Acta 238:321-336
Schmidt R, Hansen EW, Stocker M, Akporriaye D, Ellestad OH (1995) J Am Chem Soc
117:4049- 4056
Schmolz E, Lamprecht I, Schicker B (1993) Thermochim Acta 229: 173-180
Schneider HA (1999) J Therm Anal Cal 56:983-989
Schubring R (1999) J Therm Anal Cal 57:283-291
Schwanebeck W (1991) ThermochimActa 187:201-210
Serra R, Nomen R, Sempere J (1998a) Thermochim Acta 316:37-45
Serra R, Nomen R, Sempere J (1998b) J Therm Anal Cal 52: 933-943
Sestak J, Berggren G (1971) Thermochim Acta 3: 1-12
Sewry JD, Brown ME (2002) Thermochim Acta 390: 217-225
Shultz AR (1984) J Polym Sci, Polym Phys Edn 22: 1753-1771
Simon SL (2001) Thermochim Acta 374:55-71
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 E (ed) Methods of Biochemical Analysis. Wiley, New York
Staub H, Perron W (1994) Anal Chem 46: 128-130
288 References
Stevens GC, Richardson MJ (1985) Polym Commun 26: 77-80

St0len S, Gr0nvold F (1999) Thermochim Acta 327: 1-32
Stosch R, Bauerecker S, Cammenga HK (1996) Z Phys Chern 194:231-241
Stosch R et al. (1998) Mol Cryst Liq Cryst 312:285-295
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) J Therm Anal 29: 1369-13 77
Szab6 J, Luft G, Steiner R (1969) Chern Ing Tech 18: 1007-1014
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
Theriault RP, Wolfrum J, Ehrenstein GW (1999) Kunststoffe 89: 112-116
Thermochim Acta (1997) Special issue 304/305
Thermochim Acta (1999) Special issue 330
Thermochim Acta (2001) Special issue 377
TodaA, Saruyama Y (2001) Polymer 42:4727-4730
TompaAS (2000) ThermochimActa 357:177-193
ToolAQ (1946) JAm Ceram Soc 29:240-253
Tryson GR, Shultz AR (1979) J Polym Sci: Polym Phys Ed 17:2059-2075
Tsitsilianis C, Staikos G (1992) Macromolecules 25: 910-916
Turnbull AG, Hull HS (1968) AustJ Chern 21 :1789-1797
Ueberreiter K, Kanig G (1952) J Coli Sci 7:569-583
Ulbig P, Hoburg D (2002) Thermochim Acta 382: 27 -35
Ulbrich U, Cammenga HK (1993) Thermochim Acta 229:53-67
Vallebona G (1979) J Therm Anal 16:49-58
van Assche G, van Hemelrijck A, Rahier H, van Mele B (1995) Thermochim Acta 268: 121-142
van Assche G, van Hemelrijck A, Rahier H, van Mele B (1996) Thermochim Acta 286: 209-224
van Assche G, van Hemelrijck A, Rahier H, van Mele B (1997) Thermochim Acta 304/305:
317-334
van Assche G, van Mele B, Saruyama Y (2001) ThermochimActa 377: 125-130
van der Plaats G (1984) Thermochim Acta 72: 77-82
van Dooren AA, Miiller BW (1983) Thermochim Acta 65: 257 -267
van Ekeren PJ, Hall CM, Witteveen AJ (1997) J Thermal Anal 49: 1105-1114
van Hemelrijk A, van Mele B (1997) J Therm Anal 49 : 437 -442
von Stockar U, Marison IW (1991) Thermochim Acta 193 :215-242
Vyazovkin S, Lesnikovich AI (1990) Thermochim Acta 165:273-280
Vyazovkin S, Dollimore D (1996a) J Chern Inf Comp Sci 36:42-45
Vyazovkin S (1996b) Int J Chern Kinet 28:95-101
Vyazovkin S, Sbirrazzuoli N (1996) Macromolecules 29: 1867-1873
Vyazovkin S (1997) J Comput Chern 18:393-402
Vyazovkin S, Wight CA (1998) Int Rev Phys Chern 17:407-433
Vyazovkin S, Wight CA (1999) Chern Mater 11 :3386-3393
Vyazovkin S (2000a) Thermochim Acta 355: 155-163
Vyazovkin S (2000b) New J Chern 24:913-917
Vyazovkin S (2001 a) J Thermal Anal Cal 64:829-835
Vyazovkin S (2001 b) J Comput Chern 22: 178-183
Vyazovkin S, Sbirrazzuoli N (2001) Polymeric Material: Science & Eng 84: 1021-1022
Wadso I (1993) Thermochim Acta 219: 1-15
Watson ES, O'Neill MJ, Justin J, Brenner N (1964) Anal Chern 36: 1233-1238
Weir RD, Goldberg RN (1996) J Chern Thermodyn 28:261-276
Weyer S, Hensel A, Schick C (1997) Thermochim Acta 304/305 :267 -275
References 289
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 JD (1955) J Am Chem Soc 77:3701-3707
Wisanrakkit G, Gillham JK (1990) J Coatings Technology 62:35-50
Wise CW, Cook WD, Goodwin AA (1997) Polymer 38:3251-3261
WolfG, Schmidt H-G, Bohmhammel K (1994) Thermochim Acta 235:23-29
Wunderlich B (1990) Thermal Analysis. Academic Press, Boston
Wunderlich B (2003) Prog Polym Sci 28:383-450
Wunderlich B, Czornyj G (1977) Macromolecules 10:906-913
Wunderlich B et al. (1990) ATHAS databank update 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
Zynger J (1975) Anal Chem 47: 1380-1384
Subject Index
A - reactions 162,174,177,179,181,189,195
AC calorimetry 7 - uncertainty 121,123
accuracy 6,252 benzil 230
acetamide 223 benzyl dimethyl ketal 180
acetaminophen 223 bias 252
acetanilide 230 bioavailability 223
acrylate 192 biochemistry 219,220
activation energy 171,184-187 biological systems 6
activation entropy 171 Biot-Fourier equation 33
activation by radiation 180,192-195 bisphenol-A-diglycidylether 176
adiabatic Bode plot 59,62
- bomb calorimeter 274 bomb calorimeter 3,274
- calorimeter 4, 268 Borchardt & Daniels method 184
- DSC 13,24 butter 219
- scanning calorimeter 268
amplitude (TMDSC) C
- temperature 51,60 cacao butter 219
- heat flow rate 53 calcium oxalate monohydrate 227
- thermal wave 58 calcium stearate monohydrate 228
- transfer function 59 calibrant 3
analogue electric circuit (see network) 32, calibration 12,45,65-114
39,42,58 - caloric 86-92,
analytical equipment 5 - cooling 84-86, 97
aneroid drop calorimeter 271 - electrical 88, 91
aniline 176 - factor 12,19,45-47,86,93-97
annealing 202,217,228,233,235 - function 58,76,80,89-96
apparent heat capacity (see heat capacity) - heat 90-92
141 - heat flow rate 87-90
apparatus function 131,132,181,191,250 - influences 74,95,96
applications, DSC 147-244 - magnitude 113
Arrhenius equation 171,196 - peak area 90-92,107,108
ASTM 75,87 - phase 113
autocatalysis, (autocatalytic) 170,172,176, - procedure 69,70,86,88,92
178,183,192,196 - substances 6, 98-107
Avrami (Erofeev) equation 170 - symmetry check 85
azobenzene 189 - temperature 69-75,84-86
- thermodynamic aspects 66
B - TMDSC mode 107,110-114
baseline 42,116,117,121-124,162,227, caloric calibration 86-92,
237 calorimeter 1,263
- changes 124 - classification 263
- construction 121-124 - examples 264-279
292 Subject Index
calorimeter cryostate 6
- heat accumulating 271 crystal size 227
- heat compensating 264 crystallinity, degree of 236
- heat exchanging 278 curve (see DSC curve)
- types 3,263,264 cyclopentadiene 191
calorimetry 1,3,263 cylinder-type DSC 14-17
- applications 3 - advantages 16
- classification 263 - noise 17
- synopsis 263-279
Calvet type DSC 14-17 D
calcium decomposition 5,168,227,235
- oxalate monohydrate 227 deconvolution 126,132,181
- stearate monohydrate 228 deformation calorimetry 7
catalyst (catalytic) 169,170,173,176,183, deformation energy 68
188 degree of reaction 122,164,169,175,
certified reference material 65 182,185,189,194,198,214
characteristic temperatures 70,76,117 degree of crystallinity 236-238
characteristic terms 115-117 dehydration 227,228
characterization desmearing 126-140
- complete instrument 245 - advanced 133-138
- materials 3,219-228 - cp-measurements 152,154
- measuring system 245 - heat flow rate 128
- results of a measurement 251 - measured curve 126
checklist DSC 257 - numerical methods 130-133
cholesteryl myristate 224 - reaction curve 181,188,194
classification of calorimeters 263 - temperature 127,128
combined DSC 25 - Tg-measurements 203,210
combustion calorimeter 274 detection limit 246
comparison devitrification 56,195,198,202,239
- manufacturers 257 differential measurement (method) 9
- measurements 245 Differential Scanning Calorimetry
- types of DSC 259 (see DSC) 1
compatibility of substance and crucible differential temperature 12
102 Differential Thermal Analysis (DTA)
compensating heating power 17-20 2,259
compensation calorimeter 264-270 Differential Thermometry (DT) 2
complex heat capacity (see heat capacity) diffusivity, thermal 60
56,60 2,2-dimethyl-l ,3-dioxolan-4-yl)methyl
conductance (conductivity), thermal 33, methacrylate 192
45,57,134,152 dimyristoylphosphatidylcholine 220
construction principle 264 3,6-dioxaoctane-l,8-dithiol 176
controller of power compensation 18 disk-type DSC 10-13,278
convection heat transfer 45 - noise 12
convolution integral 56, 132 - uncertainty 12, 13
convolution product 57,132 drop calorimeter 3,271
cooling mode 84, 97 DSC 1,7
copper 105 - actual problems 5,6
correction 71,252 - applications 4,147-244
- of errors 252 - calibration 65-114
coupled apparatus 5 - characterization 245
cp (see heat capacity) - checklist 257
criteria of classification 263 - combined types 25
cross-talk, thermal 42 - fluids 24
crucible 10, 38, 102, 150 - heat flux 10-13,262,278
Subject Index 293
- light-modulated 24 excess
- linearity 50, 130,248 - heat flow rate 54
- modulated (see TMDSC) 27 - quantities 232
- noise 12,246 exothermic events 20,36,55,74
- oscillating (see TMDSC) extent of reaction 162,164,168
- performance 245 extrapolated peak completion temperature
- power compensation 17-25,48-50, 70,117
270 extrapolated peak onset temperature
- pressure 13 70-72,117
- temperature-modulated 50-63,159,218, Eyring equation 171
238
- theory 31-63 F
- triple system 13 fictive temperature 205
- types of 9-25 final peak temperature 70, 117
- uncertainty 12,13 finite-element method 44
DSC curve 115-146 fixed point 69,100
- calculated 115 flow calorimeter 4,274
- check 116 flow-mix calorimeter 274
- measured 115-119 food 219
- characteristic terms 116, 117 formal kinetic 173,187,195
- desmearing 126-140 Fourier
- evaluation 119-121 - analysis 53
- influences 118, 119 - integral 51,132
- kinetic evaluation 168 - series 51
- shape 118 - space 57
DTA 2,259 Fourier transform 57,58,132
- applications 4, 5 - inverse 57,132
dual sample DSC 13 free energy (see Gibbs free energy)
dynamic mode (see scanning mode) 26 Freeman & Carrol method 185
frequency of modulation 51
E fundamentals of DSC 31
electric calibration 88, 91
electric network 32,39,42 G
Ellerstein method 185 gas calorimeter 3,274
endothermic event 55 Gibbs free energy 121
energy (see internal energy) glass transition 57,161,195,200-219,
enthalpy 66,67,121,162,160-162 234
- determination 124-126 - characterization of glasses 213
- reaction 124, 125, 162 - during polymerization 163,195,214
- transition 67, 124, 125,221,241 - nature 201
enthalpy relaxation 202,212,216 - modeling of relaxation phenomena
enthalpy-temperature diagram 166,206 212
entropy 121,125,223,232 - phenomenology 200
epoxide 163,175,176,183,195 - relaxation peaks 204, 216
equivalent electric circuit 32,39,42 glass transition temperature 200
errors 252 - applications 212
- systematic 252 - definition 203
eutectic system 134, 230,241 - influence of molar masses 213
events, puIs-like 130 - of copolymers 215,216
examples of calibration 77,77,86,92 - of crosslinked polymers 215
examples of calorimeters 263 - of semicrystalline polymers 216
- heat accumulating 271 - thermodynamic (fictive) 206,207
- heat compensating 264 Green's function 131,136,138,181,250
- heat exchanging 278 GUM method 253
294 Subject Index
H - turret-type measuring system 13,14

harmonics 51 - theory 31-48
heat 1 - triple measuring system 13
- latent 54 heating-cooling mode 28,52
- measured 86,91 heating-iso mode 28,52
- true 90 heating-only mode 28, 52
heat accumulating calorimeters 271 heating rate 12,52
heat capacity (Cp ) 34,53,67,68,88, 120, heat of crystallization 234
147-160 heat of fusion 234
- apparent 55,57,141,142,145 heat of reaction 162-168
- change 123,200,201 heat of dissolution 265
- complex 56,60,138,143-145 heat of transition 44,67,107,219
- copper 105 heat pulse 36,43,130,131
- imaginary part 56, 144 heat radiation 45
- magnitude 56, 143 heat transfer 57
- non-reversing 142 heterogeneous system 168, 170, 171, 215,
- phase 56,143 217
- real part 56, 144 high-energetic materials 233
- reversing 141 high-pressure calorimeter 7,22
- sapphire 103, 104 high-speed mode 27
- time-dependent 56,57 high-temperature calorimeter 7
- vibrational 53,56 homogeneous system 168,170,171,215,
heat capacity (Cp ) measurement 6, 147-160 217
- classic procedure 148 hybrid system 25
- dual step method 153
- small step method 155
- sources of error 150 ice calorimeter 265
- step scan method 158 ICTAC 6,65
- with TMDSC 159 impurity 241-244
heat compensating calorimeters 264 indium 77 -80
heat convection 45 initial peak temperature 117
heat-exchanging calorimeters 278 internal energy 67
heat flow 1 International Temperature Scale (ITS) 69,
heat flow rate 12-20,19,52,31,34 99,100
- amplitude 53 interpolated baseline 117
- calibration 87-90,87-90 interpretation of results 6, 57, 145
- excess 54 ISO 3,66,99
- measured 86-90 isoconversional method 185
- non-reversing 142 isokinetic baseline 145
- periodic part 53,55 isomalt 219
- phase 53,55 isoperibol (constant surrounding
- resolution (see noise) 246 temperature) 264
- reversing 142 isothermal
- true 12,15,19,47,87-90,119,129-133 - mode 26,159,162,175,180
- underlying part 53, 55 - reaction curves 175
heat flux DSC 9-17, 31-48, 278 isothermals 148
- cylinder-type measuring system 14-17 ITS-90 (see International Temperature
- disk-type measuring system 9,10-13, Scale)
31-48
- dual sample 13, K
- equivalent electric circuit 32,39,42 kinetic analysis 162, 168
- models 31-48 kinetic evaluation 168
- modulated heating mode 27-30 - requirements 173
- numerical simulation 44-48 kinetic investigations 168
Subject Index 295
- calculation methods 184-188 - scanning 26,

- data sampling 173 - step-scan 30
- examples 189-200 - temperature-modulated 27-30,50-63
- formal kinetic model 187,195 modification (see polymorphism) 219
- heterogeneous phase 169,170,171,184 modulated DSC (see temperature-
- homogeneous phase 168,170,171 modulated DSC)
- prediction 172, 183, 189 modulated temperature (see temperature-
- rate laws 168-173 modulated)
- reaction mechanism 172,176,183,186, modulation of temperature
188 - frequency 51
- sample preparation 173 - amplitude 51
- strategies 183-189 - period 53
kinetics 168 - phase 53,
kinetic triplet 172
Kirchhoff-equation 161,162,166,232 N
nematic phase 228
L network model 32,39,42,58,59,61
lag, thermal (see thermal lag) nitroaniline 233
lamellae crystal 227,238 noise 12,17,19,246
latent heat 124-126 non-equilibrium state 56
lead 77-80 non-reversing curves 142
light 23 numerical methods 44, 130
light -activated reactions 23, 180, 192 numerical simulation 44
light-modulated DSC 24
linear response 50, 56, 58, 130 o
linearity 50,63, 130, 248 3-omega (3w) calorimetry 7
liposome 220 operation mode 25-30,264
liquid-crystalline mesophases 224 order ofreaction 169
liquid crystal 86,109,224 oscillating DSC (see TMDSC)
- temperature calibration 109 overall rate laws 170,172,183,193
low-temperature calorimetry 6 Ozawa-Flynn-Wall method 186
M P
measured curve (see DSC curve) partial peak area 38
measurement 251 particle energy 6
- interpretation 57,145 peak 34,116,117
- principle 263 - area 42,90,124-126
measuring system 259 - definition 116
mega-calorimetry 3 - characteristic temperatures 70,117
mesophase transitions 224 - heating rate dependence 44
metallography 230 - maximum 44,117
metastable phases 219 - overlapping 234
methacrylates 193 - partial integration 38
metrology 9 - repeatability 247
mineralogy 230 - shape 44-46,48,134
mixing calorimeter 3 - temperature maximum 117
mode of operation 25, 264 peak area calibration 90-92, 108
- heating-cooling 28,52 performance of DSC 245
- heating-iso 28,52 period (time) 29,53
- heating-only 28,52 periodic
- high speed scanning 27 - part of signal 53,55
- isothermal 26,162,175,189 - temperature change 51
- quasi-isothermal 27,52 PET 142,144
- sawtooth 29 pharmaceutical applications 221
296 Subject Index
phase (TMDSC) 53,55 pulse response function 37, 43, 58, 131, 137
- thermal wave 58 purity determination 241-244
- transfer function 59
phase behavior 219,241 Q
- acetamide 223 quality assurance 4, 66
- calcium oxalate monohydrate 227 quality control 233
- calcium stearate monohydrate 228 - aging of plastic 235
- cholesteryl myristate 224 - degree of crystallinity 234,236
- dimyristoylphosphatidylcholine 220 - oxidative decomposition 235
- liquid crystalline mesophases 224 - plastic blends 234
- metastable phases 219 - reactive resins 235
phase diagrams 5,230-232,241 - semicrystalline polymers 234
- acetanilide/benzil 230 - thermal decomposition 235
- eutectic 230,232,241 quasi-isothermal mode 27,52
- ideal/real systems 232 quenching 206,217
phase transition 67,219-228
phenylbutazone 223 R
phenylglycidyl ether 182 radiation activation 180
photo-DSC 22-24,180 radiation heat transfer 44
- base line 180 radical polymerization 192
photo reactions 180-183,192 radioactivity 6
- dark reaction 194 random errors 251
plastics see polymers rate laws 168,182,183,189
polybutadiene 238 RC-element 59,61
polydimethylsiloxane 238 reaction 68, 162, 168
polyethylen 236,239,240 - autocatalytic 170,172,179,181,189,195
polyethylenterephtalate (see PET) 237 - consecutive 175,176,187,188,197
polymerisation 180,182,183,192,214,215 - curves 165,187
polymers 160,192,200,212,233-241 - heat 162-168
- amorphous 234,236 - isothermal 175
- crystalline 234,237 - mechanism 172,176,183,186,188
- crystallization 235 - non-isothermal 163,175
- glass transition 200-219 - order 169,183,185,189
- melting 234 reaction activation
- recycling 235 - by radiation 180,192
- semi -crystalline (partial crystalline) - thermal 180
234,236,237 reaction calorimeter 3
polymorphism 219-228 reaction heat flow rate 34
polystyrene (PS) 210 recycling 233
power compensation DSC 17-25,48-50, reference material
270 - certified 65
- advantage 21 - calibration 98-107
- controler 18,49 - recommended 101,101-105,
- fluids 24 108-110
- special types 22-25 reference sample 32
pre-exponential factor 17l, 172, 186, 187, relaxation process 56
213 repeatability 247,252
presentation of results 145 reproducibility 248,252
pressure DSC 13, 22 resistance (see thermal resistance)
principle of measurement 263 resolution 250
process 53,142 result 145, 251
- endothermic 55 reversing
- exothermic 55 - heat capacity 54,141
pulse 36,43,130,131 - heat flow rate 142
Subject Index 297
S temperatures, characteristic 70, 117,

safety aspects 5, 232 2,2',6,6' -tetrabromo-bisphenol-A
- substances 233 diglycidylether 195
safety calorimeter 3,232 theory
safety risk 233 - DSC 31-63
sample heat flow rate 119 - heat flux DSC 31-48
sample model 34,60 - power compensation DSC 48-50
sample preparation - TMDSC 50-63
- cpmeasurement 150,154 theory of linear response 50, 56, 58,
- reaction 173 130
- Tg-measurements 204 thermal activation 180
sample properties 53, 58 thermal conductivity 33,45,57,58,136,
sample temperature 38, 135, 136 150,152,174,242
sapphire 103, 104 thermal diffusion equation 60
sawtooth mode 29,51 thermal history 202,234
semi-crystalline polymers 234,236,237 thermal lag 135-140,
sensitivity 43,250 - cp-measurements 151,154,155,157
Sestak-Berggren equation 170 - Tg-measurements 203
scanning mode 26 thermal resistance 32,35,59,60,75, 123
SI units 66 thermal wave 57,60
side reaction 163, 177 thermocouples 10,14
signal-to-noise ratio 247 thermodynamic functions 67,120,121,
silicate glass 201 140,161
simulation, numerical 44 thermodynamics 67,121,161
simultaneous techniques 5 thermometer problem 39
smearing 35,178,211,229 thermopile 15
smectic phase 224, 228 Tian equation 36
Sourour-Kamal equation 170,183 time-temperature-transformation (TTT)
stability investigations 235 diagram 199,293
standard 65 TMDSC (see temperature-modulated mode)
stationarity 63 27-30,50-63,159,218,238
steady state 32,37,42,48 - calibration 107-114
step-scan mode 30,138,145,158 - curves 140-145
step response function 58, 136, 145 - influence of sample 52-57
subscripts (table) XII time constant 13, 16, 19,35,39,43,48, 130,
substance identification 219 150,151,163,174,181,195,249,250
superposition principle 53,130,131 time dependent
surface energy 68 - heat capacity 57
symbols (table) XI - processes 54,238
symmetry 33,41,48,50,68,87 tin 78-80
- check 85 transfer function 58,59,61,62
systematic errors 251 - DSC model 62
transfer theory 58
T transition enthalpy 67,125
temperature trimethylene diisocyanate 176
- calibration 69-75,99 triple measuring system 13
- correction 128, 135 true heat 47,90
- fixed points 100 true heat flow rate 47,87,88,129-140
- true 7l,72 true temperature 7l,72
- profIle in sample 136 turret-type DSC 13,14,
temperature scale 69, 100 types
temperature-modulated mode 27-30, - of calorimeters 3,263-279
50-63,159,195,238 - ofDSC 9-25
- calibration 107,110-114 Tzero technology 14
298 Subject Index
U vitrification 56,169,170,195,198,199,202,
uncertainty 6,65,68,72,98,251-256 239
- black box method 253 Vyazovkin method 186
- budget 253-256
- determination 251
W
- GUM method 253
wave (see thermal wave)
- measurement 72,251
WLF equation 196,197,210
- temperature measurement 69, 72
work 67
- total 251
underlying signal 53, 55
Z
V zeroline 50,116,119,120,124,126,129,
van't Hoff equation 241 147,163
variable heating rate 27 - repeatability 247
verification of calibration 69 zinc 77

Differential Scanning Calorimetry

Încărcat de

Informații document

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Differential Scanning Calorimetry

Încărcat de

Drepturi de autor:

Formate disponibile

Differential Scanning Calorimetry

Springer-Verlag Berlin Heidelberg GmbH

With 130 Figures and 19 Tables

Dr. H.-J. Flammersheim

Library of Congress Cataloging-in-Publication Data

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

Braunschweig, Jena and Ulm G. W. H. Hahne

2 Types of Differential Scanning Calorimeters and Modes

3 Theoretical Fundamentals of Differential Scanning Calorimeters . 31

4 Calibration of Differential Scanning Calorimeters 65

4.5 Conclusions Regarding the Calibration of DSCs 97

5 DSC Curves and Further Evaluation . 115

6 Applications of Differential Scanning Calorimetry 147

6.4 The Glass Transition Process . . . . . . . . . . . . . . 200

7 Evaluation of the Performance

Subject Index 291

A area, pre-exponential factor

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

Differential Scanning Calorimetry (DSC) means the measurement of

It is important to understand that DSC measures the change of a property -

Differential Thermal Analysis (DTA)

To distinguish a heat flux DSC from an apparatus for Differential Thermal

Differential Thermometry (DT) means the measurement of the change

Calorimetry and OTA Today

Special calorimeters (commercially available) are also used to investigate

Characteristic temperatures of transitions or reactions can be very well deter-

- deformation calorimetry on polymers (see, e. g., Kilian, Hahne, 1983; Godovsky,

During the last decades the development of Differential Scanning Calorimetry

Two basic types of Differential Scanning Calorimeters (DSCs) must be distin-

c;llFS- c;llFR - - AT with AT= Ts- TR

As neither ideal thermal symmetry of the measuring system at all operating

(for definition of baseline 4\" see Sect. 5.1)

about 5 %, and it is to be expected that it cannot be reduced to less than 2 % even

Heat Flux DS( with Triple Measuring System

An extension of the disk-type measuring system described above was presented

Pressure Heat Flux DS(s

The characteristic feature of this recently developed measuring system is that

Tzero DSC technology will be outlined in Chapter 3, it is much more sophisti-

A block-type cylindrical furnace is provided with two cylindrical cavities, each

measurement signal proper is the temperature difference I1T of both sample

!PFS - !PFR - -I1T

maximum heating rates - depending on the cavity volume - lie at about 30 K

The power compensation DSC belongs to the class of heat-compensating

The commercial power compensation DSC I most frequently used is an instru-

,--___ @ Calibration ~ cf> m

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

The factor kl is a factory-set fIxed quantity of the proportional controller, k2 can

Because of the importance of this type of calorimeters several modifications

High Pressure DSC

High pressure measurements are of great importance from the thermodynamic

Another modification concerns the possibility to follow light induced reactions

Fig. 2.S. Light transfer light source

quartz cuvette with

place only as long as activation energy is supplied by irradiation. High-pressure

Another Photo-DSC is described by Saruyama, 1999. In this case the light is

Special DSC for Fluids

For precise biochemical investigations or for the investigation of reactions

Power compensation heaters serve to minimize the temperature difference

High-Speed Scanning Mode

T(t) = To + Po t + TA sin (wt) (2.6)

Fig.2.1O. Special tempera- 5

In this mode of operation different linear scanning rates change periodically

with c as a proportionality factor containing the amplitude and period tp of

A disadvantage of this temperature modulation mode is the discontinuous

If>(w) =Gp(w) T(w) (3.26)