Professional Documents
Culture Documents
Differential Scanning Calorimetry (1st Ed.) - Dr. G. W. H. Höhne, Dr. W. F. Hemminger, Dr. H.-J. Flammersheim PDF
Differential Scanning Calorimetry (1st Ed.) - Dr. G. W. H. Höhne, Dr. W. F. Hemminger, Dr. H.-J. Flammersheim PDF
Differential Scanning
Ca lori metry
2nd revised and enlarged edition
" 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
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
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 1
Appendix 1 259
Appendix 2 263
References . 281
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 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
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
2000°C 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
- 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
Summary
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
rzzzzzzzz£l~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
12 2 Types of Differential Scanning Calorimeters and Modes of Operation
is generated which is proportional to the difference between the heat flow rates
to the sample and to the reference sample:
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)
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
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
2.1.3
Heat Flux DSC with Cylinder-Type Measuring System
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
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:
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-
16 2 Types of Differential Scanning Calorimeters and Modes of Operation
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 -196°C and
1500 0c. Due to the relatively great time constants of the measuring systems, the
2.2 Power Compensation DSC 17
2.2
Power Compensation DSC
2.2.1
Function Principle
1 The description of a power compensation DSC is based on the widely used DSC of Perkin-
Elmer Instruments.
18 2 Types of Differential Scanning Calorimeters and Modes of Operation
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
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)
2.2 Power Compensation DSC 19
I1P=-k l ·I1T
<Pm =-k 2 ·I1T
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.
20 2 Types of Differential Scanning Calorimeters and Modes of Operation
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)
2.2 Power Compensation DSC 21
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.
22 2 Types of Differential Scanning Calorimeters and Modes of Operation
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
Photo-DSC
monochroma tor
neutral density fil ters
LU W calorimeter
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 550°C, 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.
26 2 Types of Differential Scanning Calorimeters and Modes of Operation
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
2.4 Modes of Operation 27
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.
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:
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
28 2 Types of Differential Scanning Calorimeters and Modes of Operation
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
2.4 Modes of Operation 29
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
_ (2n 1 3 . 2n 1 5 . 2n )
T(t) =c sin--2"sin--+2"sin--- ...
tp 3 tp 5 tp
Step-Scan Mode
3.1
Heat Flux DSC
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
3.1 Heat Flux DSC 33
~ 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):
Consequently
(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
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.
3.1 Heat Flux DSC 35
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:
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=Cs·R
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
36 3 Theoretical Fundamentals of Differential Scanning Calorimeters
t
DoT
-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)
3.1 Heat Flux DSC 37
t, time--....
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:
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.
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*:
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
/
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
3.1 Heat Flux DSC 39
dTMS dTMR)
flTsR = Ts- TR= TMS - TMR + T2 ( - - - - -
dt dt
iPr (t)
1[
= - Ii fl Tm
dflTm
+ R . (Cs - CR) • f3 + T 1 dt + TIT 2 ~
d2flTm J (3.5)
1 [ dflT d2flT
iPr(t)=-- flT+k o +k 1 --+k 2- -
d 3 flT
2 +k3 - - 3 + ...
J
Reff dt dt dt
3.1 Heat Flux DSC 41
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
42 3 Theoretical Fundamentals of Differential Scanning Calorimeters
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:
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
3.1 Heat Flux DSC 43
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.
1---+- - - - - - - -=-::-=-=-,...,-~------
to time
44 3 Theoretical Fundamentals of Differential Scanning Calorimeters
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
3.1 Heat Flux DSC 45
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
46 3 Theoretical Fundamentals of Differential Scanning Calorimeters
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
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.
3.1 Heat Flux DSC 47
(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:
48 3 Theoretical Fundamentals of Differential Scanning Calorimeters
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
3.2 Power Compensation DSC 49
"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:
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)
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
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:
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
3.3 Temperature-Modulated DSC (TMDSC) 53
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.
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:
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=cp·m=-- (3.15)
TA·w
54 3 Theoretical Fundamentals of Differential Scanning Calorimeters
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:
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)
3.3 Temperature-Modulated DSC (TMDSC) 55
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:
(3.20)
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):
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.
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:
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 ~:
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
58 3 Theoretical Fundamentals of Differential Scanning Calorimeters
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.
3.3 Temperature-Modulated DSC (TMDSC) 59
(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 S·l 10
w--
60 3 Theoretical Fundamentals of Differential Scanning Calorimeters
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:
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'poPoWlkth°x (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-
3.3 Temperature-Modulated DSC (TMDSC) 61
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,O·e~dx=-~- l-e~
_~ TA o' ~ ( _4)
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-
62 3 Theoretical Fundamentals of Differential Scanning Calorimeters
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----
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.
4.1
Aspects of Quality Assurance
4.2
Basic Aspects of Calibration
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:
dQ = dH - V dp - I dE;
;
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
68 4 Calibration of Differential Scanning Calorimeters
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
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).
70 4 Calibration of Differential Scanning Calorimeters
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.
4.3 Temperature Calibration 71
(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
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:
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.).
4.3 Temperature Calibration 73
tempera ture
CI
c::
en
·Vi ......_.-:;
b
T,1 TP2
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:
time ..
If there is a thermal resistance Rth between sample (Ts) and sensor tempera-
ture (TM ) (Fig.4.6a, b),
(4.3)
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
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)
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)
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)
4.3 Temperature Calibration 79
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
80 4 Calibration of Differential Scanning Calorimeters
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) = 326.793°C
Tnx = 327.462°C (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
4.3 Temperature Calibration 81
Gallium
Te in Heating rate in
°C Kmin- 1
Benzoic acid
Te in Heatin~ rate in
°C Kmin-
Indium
Tin
Caffeine
Lead
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
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
84 4 Calibration of Differential Scanning Calorimeters
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
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.
86 4 Calibration of Differential Scanning Calorimeters
-----:::=:
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:
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).
88 4 Calibration of Differential Scanning Calorimeters
K (T) _ Cp(T)· f3
cPtrue = Cp • f3 so that I/> - cPm(T) _ cPo (T)
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 ~
tempera ture
r
factor to determine the
true heat flow rate:
K,p= Cp<nf3I(~m<n - ~o<n)
.,.~
temperature ~
90 4 Calibration of Differential Scanning Calorimeters
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.
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 ...
4.4 Caloric Calibration 91
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
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.
4.4 Caloric Calibration 93
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.
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:
94 4 Calibration of Differential Scanning Calorimeters
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)
10
!-LV
mw 9
ta
7
11K
6
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).
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.
96 4 Calibration of Differential Scanning Calorimeters
30
mJ
mV· s
25
t 20
KQ
15 .. '
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
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.
4.6
Reference Materials for DSC Calibration
4.6.1
Reference Materials for Temperature Calibration
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.)
100 4 Calibration of Differential Scanning Calorimeters
Table 4.6. Defining fixed points of the ITS-90 and some secondary reference points
TinK tin °C
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
4.6 Reference Materials for DSC Calibration 101
Table 4.7. Materials recommended for temperature calibration of DSCs (cf. Cammenga et ai.,
1993)
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~
4.6 Reference Materials for DSC Calibration 103
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
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
104 4 Calibration of Differential Scanning Calorimeters
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 )
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
0.008
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
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).
S-....
s"
~
....
'"en
4.7 Additional Calibration in Temperature-Modulated Mode of Operation 109
Table 4.11 b. Further Certified Reference Materials for temperature and heat calibration
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
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
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!
s solid.
SA smectic A (liquid crystal phase).
N nematic (liquid crystal phase).
1 liquid.
110 4 Calibration of Differential Scanning Calorimeters
Table 4.13. Substances recommended for heat (peak area) calibration in the cooling mode
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
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!
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:
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-
4.7 Additional Calibration in Temperature-Modulated Mode of Operation 111
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
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.
112 4 Calibration of Differential Scanning Calorimeters
/\ 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
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)
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
4.7 Additional Calibration in Temperature-Modulated Mode of Operation 113
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
114 4 Calibration of Differential Scanning Calorimeters
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
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)
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.
118 5 DSC Curves and Further Evaluations
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
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 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.
5.3 Further Evaluation of DSC Curves 121
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
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).
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
5.3 Further Evaluation of DSC Curves 123
T, 0
temperature ..
124 5 DSC Curves and Further Evaluations
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
5.3 Further Evaluation of DSC Curves 125
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-
126 5 DSC Curves and Further Evaluations
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
128 5 DSC Curves and Further Evaluations
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:
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):
5.4.2
Subtraction of the Zeroline
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.
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-
130 5 DSC Curves and Further Evaluations
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
T, T2
temperature
r
b
tempera ture ~
(or even called Green's function). If these conditions are fulfilled, the following
is valid:
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).
132 5 DSC Curves and Further Evaluations
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:
Thus the convolution product turns into an ordinary product which can be
solved for
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 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.
(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.
134 5 DSC Curves and Further Evaluations
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.
d2
(T) = llb - Cp • P . P. - (5.7)
3..\
136 5 DSC Curves and Further Evaluations
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.
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
S·l
I
0.6
0.4
C;
0.2
0.0
0 10 20 30 40 s 60
time ~
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
138 5 DSC Curves and Further Evaluations
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.
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
5.4 Desmearing of the DSC Curve 139
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)
140 5 DSC Curves and Further Evaluations
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)
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· TA·W (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).
142 5 DSC Curves and Further Evaluations
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
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
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)]:
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
A
cp = - - -
m·t:..T
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
146 5 DSC Curves and Further Evaluations
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.
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
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 calibration factor K~(T) need not, therefore, be known explicitly. If the
condition mscs::::: mRefCRef holds, the experimental conditions are very similar to
150 6 Applications of Differential Scanning Calorimetry
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-
6.1 Measurements of the Heat Capacity 151
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
152 6 Applications of Differential Scanning Calorimetry
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.
6.1 Measurements of the Heat Capacity 153
12
w
expo curve <Pup ( TJ
8
t
.......................... ............................
.
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:
6.1.3
General Precautions for the Minimization of Errors and their Estimation
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
156 6 Applications of Differential Scanning Calorimetry
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
6.1 Measurements of the Heat Capacity 157
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:
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
6.1 Measurements of the Heat Capacity 159
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=cp·m=--
TA·w
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
160 6 Applications of Differential Scanning Calorimetry
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
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).
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
6.2
Determination of Heats of Reaction
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.
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%,
6.2 Determination of Heats of Reaction 163
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.
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.
164 6 Applications of Differential Scanning Calorimetry
~ 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.
6.2 Determination of Heats of Reaction 165
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
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:
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
166 6 Applications of Differential Scanning Calorimetry
X - Y = hproducts (350 K) - ~
'"
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
6.2 Determination of Heats of Reaction 167
1
1992 b); for details see text
'p (R) 'p(PI
--- ~r"' I--
r
!\
-...'" Z
-.......
.c
"
a..
II>
X
r, T
temperature ..
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.
168 6 Applications of Differential Scanning Calorimetry
6.3
Kinetic Investigations
6.3.1
Introduction and Definitions
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}
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
170 6 Applications of Differential Scanning Calorimetry
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
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.
6.3 Kinetic Investigations 171
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:
Eyring: •
k(T} = -h- kT
B
exp (~S")
R .exp (AH"
- RT )
6.3.2
Experimental Prerequisites for a Reliable Kinetic Analysis
1. Sample Requirements
- After the introduction of the sample into the DSC the amount of reaction
should be negligible prior to the attainment of a stable steady state. If this is
not the case it must be determined separately and allowance made for this
effect.
- The reaction mixture must not react with the material of the crucibles nor
should there be any catalytic influence.
- Samples with an appreciable 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.
- 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
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,
178 6 Applications of Differential Scanning Calorimetry
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
-----
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).
6.1 Measurements of the Heat Capacity 179
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).
180 6 Applications of Differential Scanning Calorimetry
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)
6.3 Kinetic Investigations 181
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
182 6 Applications of Differential Scanning Calorimetry
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 Kinetic Investigations 183
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
184 6 Applications of Differential Scanning Calorimetry
In -(da) =In A + n . In (1 - a) - RT
dt
EA
- + (n - 1) . In Co
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
(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
186 6 Applications of Differential Scanning Calorimetry
A. EA)
lng(a) =In ( -R- -lnfJ+lnp(z)
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
6.3.6
Selected Examples and Possible Predictions
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
6.3 Kinetic Investigations 191
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
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 185°C, 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 160°C,
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 - 20°C, 9.5 h at O°C and only 72.3 min at
20°C. That means, the dimerization of CPD should always be investigated with
freshly distilled monomer immediately after its preparation.
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
6.3 Kinetic Investigations 193
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)
194 6 Applications of Differential Scanning Calorimetry
20
mW
15
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
6.3 Kinetic Investigations 195
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.
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
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:
6.3 Kinetic Investigations 197
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
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.
198 6 Applications of Differential Scanning Calorimetry
Table 6.2
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; 140°C: calculated 0.80, found 0.79 and 0.79; 170°C: 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 100°C and the slow devitrification start-
ing at about 130°C. 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
:.....---- .
~
~
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
6.3 Kinetic Investigations 199
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..
~
'-'
time - - -
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
6.4
The Glass Transition Process
6.4.1
The Phenomenology of the Glass Transition
3.0
( 0
B
mW hea ting run
~
r 1.5
"
.....
~
~
"
'"
.c
cooling run
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)
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-
202 6 Applications of Differential Scanning Calorimetry
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
6.4 The Glass Transition Process 203
6.4.3
Definition and Determination of the Glass Transition Temperature 19
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.
6.4 The Glass Transition Process 205
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 68°e has Tg,e = 89.1 °e, Tg,1/2 = 87.3°e whereas the quenched glass has
Tg,e = 81.6°e, 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).
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
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
208 6 Applications of Differential Scanning Calorimetry
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
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
6.4 The Glass Transition Process 209
4.5
J/!Kgl
3.5
1>-
3.0
~
0
0.. 2.5 Cp.1
0
~
'u /
'"
0..
II>
1.5 cp•g
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
210 6 Applications of Differential Scanning Calorimetry
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,
6.4 The Glass Transition Process 211
'-'
.9) 0,5
0,0
6.4.4
Applications of Glass Transition Measurements
T dt
~=f--
To T(T, T f )
6.4 The Glass Transition Process 213
(-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):
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
214 6 Applications of Differential Scanning Calorimetry
1 min
'"
QJ
..c
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:
1 WI Wz
-=--+--
Tg Tg•I Tg.z
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
6.4 The Glass Transition Process 217
f annealed
B
--
)0
o quenched
o
'"
J:
annealed
(
quenched
30 50 70°C 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
-2
-3
-4
29 3.0 3.1 3 2 K- 1 33
1000/ T - -
6.5 Characterization of Substances, the Phase Behavior 219
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
0.03
WIg
1 0.02
!:!
a
.....
0.01
~
=-
a
'"
.&:.
u
3
u
0
C1J
0-
III
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 = 24.6°C) 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 = 22.7°C). 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
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
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
222 6 Applications of Differential Scanning Calorimetry
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.
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
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 80°C. 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 80°C 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 103°C. 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 35°C. Subsequent behavior depends
on the heating rates, either melting of the metastable modification at about
93 °C or of the stable structure at 103°C 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., < 93°C), only at low heating rates « 1 K min-I) will there be enough
224 6 Applications of Differential Scanning Calorimetry
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 98°C 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 93°C, 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 100°C) 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 = 69.0°C) corresponds to the formation of a smectic
6.5 Characterization of Substances, the Phase Behavior 225
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)
226 6 Applications of Differential Scanning Calorimetry
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 60°C, 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= 77.7°C and Q = 1.9 Jg-l,
T = 83.0°C) 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
30°C, equivalent to a AT of about 40 K.
(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.
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
6.5.4
Porosity Measurement
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
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
115 6C
t 95°C
78°C
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
232 6 Applications of Differential Scanning Calorimetry
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:
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.
234 6 Applications of Differential Scanning Calorimetry
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
6.8 Characterization of Polymers 235
6.8.2
Determination of the Degree of Crystallinity
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
6.8 Characterization of Polymers 237
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-
238 6 Applications of Differential Scanning Calorimetry
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,
6.8 Characterization of Polymers 239
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
0
90 100 110 120 130 O( 150
tempera t ure ~
0
0 50 100
time
----
150 min 250
rP
[][]
1
24 25 26 K-' 27
liT 10 J
6.9
Purity Determination of Chemicals
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}.
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-
242 6 Applications of Differential Scanning Calorimetry
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:
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 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
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.
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.
7.2 Characterization of the Measuring System 247
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. 134°C 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
248 7 Evaluation of the Performance of a Differential Scanning Calorimeter
---------
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)
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.
7.2 Characterization of the Measuring System 249
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
250 7 Evaluation of the Performance of a Differential Scanning Calorimeter
Fig.7.4. Determination of
the time constant rwith the
melting peak of a pure
substance (according to
Hemminger, 1994)
time
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).
252 7 Evaluation of the Performance of a Differential Scanning Calorimeter
}~%
Diphenyl ether +0.95%
Linear polyethylene -1.1%
Indium +0.8% resp. + 1.8 %a
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.
254 7 Evaluation of the Performance of a Differential Scanning Calorimeter
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
7.3 Characterization of the Results of a Measurement: Uncertainty Determination 255
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
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 500°C. 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
oXi
T 500°C
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 500.9°C 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).
256 7 Evaluation of the Performance of a Differential Scanning Calorimeter
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
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 600°C 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 250°C. 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 250°C
of Cs = (1.015 ± 0.016) J. g-l. K- 1 •
Table 7.3. Uncertainty budget for the specific heat capacity determination example
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
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
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
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
2. Mode of Operation
2.1 Isothermal
2.2 Isoperibol
2.3 Adiabatic } static 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
A warm sample placed in an ice calorimeter (Fig. A 2.1) transfers its heat to a O°C
ice jacket. As a result, a certain mass of ice (to be determined) melts. In the case
of the ice calorimeter according to Lavoisier, de Laplace, 1784, the melted water
was weighed, whereas Bunsen, 1870, determined the change of the ice/water
ratio on the basis of the change in the volume of the whole mixture (density 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 100°C 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.)
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.
5
T
6. T = 0 6
t
f!1
--- ~ ____---CD
time
Fig. A2.3. a"Adiabatic scanning calorimeter".
...
(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).
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.
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 500°C; 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 2000°C)
can be dropped without the risk of evaporation or splashing.
Other examples of calorimeters of this class are instruments in which the
sample placed into it (solid body, liquid, gas) reacts with the calorimeter sub-
stance; this results in heat of reaction being released.
272 Appendix 2
-3
4
T '~ T
__ 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
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.
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) ].
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
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).
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).
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).
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
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
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
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