Beruflich Dokumente
Kultur Dokumente
G. W. H. H6hne . W. F. Hemminger
H.- J. Flammersheim
Differential Scanning
Ca lori metry
2nd revised and enlarged edition
With 130 Figures and 19 Tables
"
Springer
Dr. 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
52/3020 - 5 4 3 2 1
Preface
G. W. H. Hahne
w. F. Hemminger
H.-f. Flammersheim
Contents
Introduction . . . . . . . . . . . . . . . . . . . . . . . . .
2.4
2.4.1
2.4.2
3
3.1
3.2
3.3
3.3.1
3.3.2
3.3.3
3.3.4
4
4.1
4.2
4.3
4.3.1
4.3.2
4.4
4.4.1
4.4.2
4.4.3
4.4.4
2.1
2.1.1
2.1.2
2.1.3
2.2
2.2.1
2.2.2
2.3
9
10
10
13
14
17
17
22
25
25
26
27
31
31
48
50
51
52
57
63
65
66
66
69
69
84
86
87
90
92
97
Contents
VIII
4.5
4.6
4.6.1
4.6.2
4.6.3
4.7
5.1
5.2
5.3
5.3.1
5.3.2
5.4
5.4.1
5.4.2
5.4.3
5.4.4
5.4.5
5.5
5.5.1
5.5.2
5.5.3
5.6
6
6.1
6.1.1
6.1.2
6.1.3
6.1.4
6.1.5
6.1.6
6.2
6.3
6.3.1
6.3.2
6.3.3
6.3.4
6.3.5
6.3.6
97
98
99
101
107
107
113
113
115
116
118
119
119
121
126
128
129
129
133
140
140
141
143
145
145
147
147
148
153
154
155
159
160
162
168
168
173
175
180
183
189
Contents
IX
200
200
201
203
212
217
219
219
221
224
228
230
232
233
234
236
238
241
245
245
245
251
253
253
257
Appendix 1
259
Appendix 2
263
References .
281
Subject Index
291
List of Symbols
C
D
E
H
K
L
P
S
T
U
V
W
a
c
d
i
k
p
r
t
w
x
a
f3
6
A
v
tP
XII
r
w
X
~
List of Symbols
time constant
angular frequency
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
1 Introduction
namely of a heat flow rate difference - which normally is released due to an alteration of the sample temperature. When there is no alteration of the sample
temperature, no change of a heat flow rate difference can be measured (except
for possible chemical reaction heat flow rates). That means, a distinct temperature program, or in general a "mode of operation", is always part of a DSC measurement (cf. Chapter 2). In the case of thermally activated reactions/transitions,
e. g. by a controlled stepwise change from one constant temperature to another,
the sample's temperature undergoes an alteration by internal processes and
causes the DSC signal.
It should even be mentioned, that heat only flows if there is a temperature
difference present. In other words, a non-zero heat flow rate difference implies
always a temperature difference between the sample and the reference and their
surroundings and a change of the heat flow rate difference implies a change of
the temperature as well.
Differential Thermal Analysis (DTA)
1 Introduction
1 Introduction
1 Introduction
Characteristic temperatures of transitions or reactions can be very well determined by DTA. Heats can still be estimated with an uncertainty between 20 and
50 %. DTA is applied in the following fields:
- comparison (relative) measurements (identification, control, comparison),
- safety research (stability investigations, also long-time investigations), see,
e. g., Hentze, 1984; Hentze, Krien, 1986,
- investigation of transitions, decompositions, reactions with gases,
- evaluation of phase diagrams.
DSC and DTA are also used together with other methods of thermal analysis or
other analytical techniques (simultaneous thermal analysis), most frequently
in connection with thermogravimetry (TG), more rarely with Evolved Gas Thermoanalysis (EGA), Thermomicroscopy (TOA) or Thermosonimetry (TS).
The coupling of DSC or DTA with Thermogravimetry (TG) is of particular
importance. In addition to information on changes in the heat flow rate (due, for
example, to changes of Cp) and heats of transition, the TG signal provides information on whether volatile components are involved and which changes in mass
are to be attributed to a transition.
The different methods of gas analysis used together with DSC or DTA (usually together with TG/DSC or TG/DTA systems) are of increasing importance.
Via a carrier gas stream, gaseous reaction products from the calorimeter or DTA
device are transported to an apparatus (e. g., IR or mass spectrometer) in which
these products are analyzed. This allows a correlation between the characteristic temperature of a reaction, gravimetric and/or enthalpic information and
the composition of the volatile reaction product to be made (see examples in
Mathot,1994a).
It is difficult to transfer the volatile products without adulteration from the
hot sample to the analysis system (condensation must be avoided). Various systems for coupling quadrupole mass spectrometers are commercially available
(up to temperatures of 2400 0c). Examples of the investigation of volatile components applying methods of gas analysis can be found in Ohrbach et al., 1987;
Matuschek et al., 1991; Matuschek et al., 1993; Kaisersberger, 1997.
The advantage of simultaneous measurements is that the same sample is
investigated under identical conditions and that diverse information is obtained by one measurement run which is important for interpretation purposes. The following potential disadvantages of simultaneous instruments
should be mentioned: lower sensitivity, higher susceptibility to failure, increased time and effort required for preparation and operation, higher instrument costs.
In Differential Scanning Calorimetry which is widely applied, there are still
some actual problems of which one should be aware:
- The theoretically well-founded complete understanding of DSC is not yet
totally accepted. The average user does not yet know the limits to this method
and the sources of the systematic errors by which it is affected. This is in particular true for the temperature-modulated DSC, its theoretical background
is still a matter of controversial discussion.
1 Introduction
- There are not yet any practicable and experimentally tested recommendations 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 temperature and heat calibration of DSC which have been measured with sufficient accuracy, including a metrologically sound traceability to national/international standards; instead there is a confusing variety of "certified reference materials" whose characteristic data are in part contradictory. Substances 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 Confederation for Thermal Analysis and Calorimetry (ICTAC) should offer special
training courses, and precise specifications for instruments and programs
should be drawn up.
The following can frequently be observed:
- A realistic estimate of the uncertainty of measurement is rarely made (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 examples:
- 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
Summary
During the last decades the development of Differential Scanning Calorimetry
from a half-quantitative DTA method (producing "thermograms") toward a true
calorimetric method (yielding quantitative caloric quantities) forged ahead.
This becomes manifest from the following items which hopefully will be proved
within this book.
- The theory of (heat-flux) DSCs is well-known and the function principle understood.
- Systematic (unavoidable) error sources and methods of correction are
known.
- Modern data treatment and powerful computers make it possible for everyone to make the necessary evaluations without problems.
- Metrological founded calibration and measurement procedures improved the
certainty of the results which nowadays are comparable with those obtained
with classical calorimeters.
- The difference in quality of the results from heat-flux and power-compensated DSCs has diminished.
- Precise heat capacity measurements are possible and open the door for other
well-defined thermodynamic quantities.
- Coming to terms with the problem of the thermal inertia (thermal lag) enables kinetic investigations which otherwise need much more experimental
efforts.
- DSC is a fast and reliable method in a wide field of different applications.
10
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 conduction 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 medium 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 carried 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) differ. The use of modern sensors on the basis of semi -conducting material leads
to a significant increase in sensitivity.
11
rzzzzzzzzl~zzzhzzZl
S
R
T( f) 1---7>'---,------.
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 differential 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 Hemminger, 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 arrangement 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
is generated which is proportional to the difference between the heat flow rates
to the sample and to the reference sample:
c;llFS- c;llFR - -
AT
= -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 scanning with a constant heating rate f3 =dT/dt:
(2.1)
- by comparing the integral over a transition peak with the expected (known)
heat of transition Qr (energy balance):
(2.2)
(for definition of baseline 4\" see Sect. 5.1)
(When resistance thermometers are used as temperature sensors, the thermometers can be used also as heaters which is advantageous in calibration procedures, see Wolf et aI., 1994).
Questions of calibration capability and calibration are dealt with in Sect. 4.4;
problems of "desmearing" arising in connection with it are discussed in Sect. 5.4.
Heat flux DSCs with disk-type measuring systems are available for temperatures between -190C and 1600C. The maximum heating rates are about 100 K
min-I. Typical time constants of the AT-sensor (empty systems, no samples) are
between 3 and 10 seconds. The noise (definition see Sect. 7.2) of the measurement signal lies between 0.511W and 20 llW (it also depends on the temperature
and the heating rate). The total uncertainty of the heat measurement amounts to
13
2.1.2
This leads to a smaller furnace as well and the thermal response time becomes
much lower, which allows faster heating and cooling rates and higher frequencies in the temperature-modulated mode.
In addition to the temperatures of the sample and reference sample support
the temperature in the middle of the bottom plate To (see Fig. 2.2) is measured
in this measuring system. With these data it is possible to determine additional
quantities which are essential for the performance of the DSC and correct for
the influence of thermal inertia. The theoretical background of this so-called
14
R
~::::::::1
Fig.2.2. Heat flux DSC with turret-type measuring system (TA Instruments). 1 elevated constantan platform for sample and reference sample, 2 chromel area thermocouple, 3 constantan body, 4 chromel - constantan thermocouple, 5 silver furnace, S sample substance,
R reference sample substance,~T platform temperature difference, To body (furnace) temperature
Tzero DSC technology will be outlined in Chapter 3, it is much more sophisticated than the simple formulae presented in Sect. 2.1.1 for the simple disk-type
DSC. Modern computer techniques offer the possibility of expensive online calculations and enable the output of the corrected heat flow rate and the calculated temperature of the sample pan even while the measurement runs.
The tiny design necessitates a very accurate production of the components
with very narrow tolerances to ensure a good symmetry which is indispensable
for the construction of differential calorimeters. However, modern technology
enables this new type of heat flux DSC which seems to be a large step in the right
direction to overcome the well-known draw-backs of heat flux DSC (see Chapter 3) in near future.
2.1.3
Heat Flux DSC with Cylinder-Type Measuring System
15
2
2
3
'-----0
!1 T 0 - - - - - - - '
Fig.2.3. Heat flux DSC with cylinder-type measuring system (Calvet, 1948; thermally decoupled sample containers) (according to Hemminger, 1994).
1 containers to take up sample and reference sample, 2 thermopiles, 3 furnace (with programmable 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:
!PFS - !PFR !Pm
-I1T
= -k' I1T
!Ptrue
= K~
!Pm
An electric voltage proportional to I1T, or a heat flow rate signal internally calculated 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
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 programmable temperature controller), 4 lid, 5 support of containers, S sample substance, R reference 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 containers (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 relatively low noise.
Heat flux DSCs with cylinder-type measuring systems with sample volumes
of about 10 mL are available for a temperature range between -196C and
1500 0c. Due to the relatively great time constants of the measuring systems, the
17
2.2
Power Compensation DSC
The power compensation DSC belongs to the class of heat-compensating
calorimeters (see Appendix 2). The heat to be measured is (almost totally) compensated with electric energy, by increasing or decreasing an adjustable Joule's
heat.
2.2.1
Function Principle
The commercial power compensation DSC I most frequently used is an instrument with an isoperibol operation. The measuring system (Fig. 2.5) consists of
two identical microfurnaces which are mounted inside a thermostated aluminum block. The furnaces are made of a platinum-iridium alloy, each of which
contains a temperature sensor (platinum resistance thermometer) and a heating
1
The description of a power compensation DSC is based on the widely used DSC of PerkinElmer Instruments.
18
Fig.2.S. Power compensation DSC (Perkin-Elmer Instruments). Set-up of the measuring system (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
Calibration
cf> m
tJp
T (I)
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)
19
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 Differential Thermal Analysis, i. e., the inventors of this DSC recognized clearly the systematic
relationship between this type of DSC and classic DTA.
20
time--l>
t
t
tJp
tJT
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 signal (proportional to the (negative) amplified ~T signal)
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 surroundings is concerned in order to keep a good symmetry and the uncertainties
of measurement small. Moreover, the shares of the various heat exchange mechanisms and their respective amounts must depend only on the temperature, and
strict repeatability must be ensured. 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 compensated 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 curvature 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 following advantages:
- The short heat conduction path between samples and heater and the relatively 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 electrical heating power. As a result, only small temperature differences I1T remain between the micro furnaces of sample and reference sample (approx.
1110 of those occurring in heat flux DSCs using the same sample and reference sample). This means that the calibration factor Kcp is practically independent 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 repeatable. 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 compensation DSCs. During the peak there is still a temperature difference 11 T proportional to the reaction heat flow rate. That is to say, the power compensation
DSC can be considered a kind of DTA instrument with a 11 T as the measurement
signal, but with a I1T arising from a specific thermal event which is much smaller than that developed in a heat flux DSC measuring system.
22
In conclusion it may be said that all the attributes of heat flux DSC systems
which depend on this temperature difference can also be found in the "real"
power compensation DSC, but to a lesser degree. The calibration factor in particular is not a constant figure but depends in principle on temperature, heat
flow rate, heating rate and peak area (cf. Sects. 3.2, 4.4). Though these effects are
not very 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
23
light source
optics
quartz cuvette with
water as IR-absorber
monochroma tor
neutral density fil ters
LU
calorimeter
24
25
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 sensorheater combinations on a disk. The temperature difference between the positions 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 advantages 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 cylindrical metal block. The crucibles for sample and reference sample are put into
these chambers and are positioned on the detector disk (temperature range
-150 to 550C, detection limit approx. 0.5 ]lW, crucibles with 30 or 100 ]lL).
In principle the addition of heaters plus a power compensation control loop
is not restricted to disk-type DSCs. Whenever a heater is part of the measuring
system - e. g., a calibration heater inside a cylinder type system - this heater may
be used for compensation purposes.
2.4
Modes of Operation
DSCs are generally operated by a controlled program which changes the temperature 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
2.4.1
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 (negative) 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 symmetric and the heat exchange of the sample and reference sample with the surroundings 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 subtracted 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 heating rate:
dT
tPcp= Cp ' -
dt
(2.4)
with Cp as the proportionality factor. In a DSC the differential heat flow rate depends 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
27
and reference sample and the third term is the heat flow contribution from a reaction or transition (latent heat) occurring in the sample. The first two parts define 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 causes a signal (and can be determined from it). Scanning is the mostly used mode
of operation with DSCs.
In this case a certain modulation term is added to the linear part of the temperature-time function. The simplest and most used modulation type is the periodic (harmonic, i. e., sinusoidal) one and Eq. (2.3) reads:
T(t)
= To + Po t + TA sin (wt)
(2.6)
with TA the amplitude and w the (angular) frequency of the modulation. Together 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
dt
= Po + TA W cos (wt)
(2.7)
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
S
4
3
ClJ
'-::J
'-ClJ
Cl.
ClJ
0
-1
time
10
(quasi-isothermal). The heating/cooling rate varies between TAW. The advantage 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 during a run. This is the case if the product TAW is smaller than the underlying heating rate Po.
The advantage of this mode is that all processes only occurring during cooling of the sample are suppressed and the remaining processes are easier to evaluate. 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
29
<lJ
L..
:::I
L..
<lJ
CI..
(lJ
time
..
10
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 remaining in the moment of minimum (i.e. zero) heat flow rate [ef. Eq. (2.5)].
Sawtooth Mode
In this mode of operation different linear scanning rates change periodically
(Fig. 2.10 curve b). Again we can distinguish between heating-cooling, heating
only, and heating-iso modes of operation, depending on the choice of the respective rates of temperature change.
Every DSC can be operated in this mode, as programming of several different
linear scanning ramps in one run normally is possible without special additional electronics or software. This is one advantage of this mode of operation.
Another advantage follows from the fact that every (non-sinusoidal) periodical
function can be expanded as a Fourier series (see textbooks of mathematics).
In the case of the sawtooth mode, the periodical function, superimposed on the
linear temperature change, reads:
_
T(t)
=c
(2n 1
3 . 2n
1
5 . 2n
)
sin--2"sin--+2"sin--- ...
tp 3
tp
5
tp
30
3 Theoretical Fundamentals
of Differential Scanning Calorimeters
3.1
Heat Flux DSC
Three steps of an analytical description of the functional principle of a heat flux
DSC will be presented in the following. The results of a numerical calculation
will be discussed afterwards.
32
tJ!
T,
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 approximation. This diagram serves to better understand the interrelations. From the
physical point of view, electric charge transport and heat transport are equivalent processes, and many people find it easier to read electric circuit diagrams
than to visualize heat flows in real equipment.
'fS
R,s
fjU
if.
R'R
U,
[.
T T
O----------------T---------~----~
/77J777
[s
33
- = - A . grad T
A
I~I
or -
= - A 1 grad T 1
The amount of the heat flux tP/A is proportional to the gradient of the temperature; 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
34
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' . -
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
dt
dt
Cs - - + Cs - - = cPps - cPr
results.
35
= 0 by defini-
dTR
CR - - = cPFR
dt
When the difference between the two balance equations is calculated, the following 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 samples. In the case of thermal symmetry, R FS =RFR =R, thus:
flT
dTR
dfiT
(3.2)
This equation links the reaction heat flow rate cPr searched with the measured
signal fl T. The second term takes the asymmetry of the measuring system into
account as regards heat capacities of sample and reference sample. The third
term considers the contribution of the thermal inertia of the system when a
measured signal flT(t) appears. In analogy to the charging or discharging of a
capacitor of capacity C, a time constant T can similarly be defined for the heat
flow rates:
T=CsR
When fl T is changed, R is the effective thermal resistance to the "charging or discharging" of the "capacity" Cs. With this resistance and with dTR/dt = /3, the
heating rate; as the reference sample is always in a steady-state heating mode, the
following results from Eq. (3.2):
cPr (t)
flT(t)
=- -R-
- (Cs - CR)
dfl T(t)
/3 - R . ~
(3.3)
The measured signal flT is not (I) proportional to the heat flow rate cPr at a given
moment but delayed with time and thus also distorted ("smeared"). Furthermore, 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
DoT
"
time
--..
"
Fig.3.3. Measured curve of a heat flux DSC (schematic) with heat released inside the sample
(exothermic effect).
t1T measured signal, R thermal resistance between furnace and sample, f3 heating rate, t1C =
Cs - CR difference between the heat capacities of sample and reference sample (here negative,
because CR > Cs), Q) peak area (exothermic effect); it is a measure of the heat released between
t, and tz ,@ area below the baseline; it is a measure of the heat required to heat the sample between t, and tz
-R . p. (Cs - CR) which is the initial deviation after the quasi-steady state has been
reached in scanning operation [cf. Eq. (3.I)]. This contribution is the measured
curve before/behind a peak which is parallel to the abscissa if R, l1C = Cs - CR and
p are constant (Fig. 3.3).
In reality, the term R (Cs - CR) reflects the temperature dependence of the
thermal resistance R (in general: of the heat transfer conditions) and of the heat
capacities Cs and CR causing a temperature dependence of the measured curve
even without any thermal effect due to the sample. [NB: Eq. (3.3) is in principle
the so called Tian equation, originally derived for cylinder-type calorimeters.]
Regarding Eq. (3.3), the conclusions are as follows:
1. When the signall1 T measured at a given moment is to be assigned to the heat
flow rate IPr by which it is caused, the third term in Eq. (3.3) must be taken into
account (cf. desmearing, Sect. S.4).
R must be determined by calibration (cf. Sect. 4.4); the time constant T can
also be obtained from calibration measurements (cf. below "Higher-order approximations" 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)
37
~~r-~-----------~-~-----
t-T-j
IlT
t,
with T
curve
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:
tl
tl
tl
dt
(3.4)
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 (definition see Sect. S.1) in Fig. 3.3.
- When ~C and R are not temperature- (or time-)dependent, the measured
curve outside the peak is a parallel to the abscissa. In this case, d~T/dt is
zero before and behind the peak. The contribution of the 3rd term vanishes when integration is carried out over the whole peak.
- In the real case, the 2nd term of Eq. (3.4) is not constant, i. e., the curve before and behind the peak is not parallel to the abscissa. In this case, the 3rd
term will not vanish but represents a correction of the values obtained by
peak integration.
38
6.T
t,
time - .
3. For the partial integration (Fig. 3.5) of the peak between t, and t*, the contribution of the 3rd term must be taken into acount at the point t*:
Qr(t*} =
tl
tl
\.
tl
dt
'-v-------l
correction term
Partial integrations of the peak are, for instance, necessary for kinetic investigations and to determine purity (see Sects. 6.3 and 6.9).
Higher-Order Approximations
The temperatures of sample and reference sample (assumed here to be homogeneous) are not measured directly. There is a certain distance and a thermal
/
39
dTMs
r2 - -
dt
TMS temperature of the measurement point (e. g., junction of the thermocouple)
for the sample
Ts sample temperature (assumed homogeneous)
r2
characteristic time constant for the temperature relaxation between sample
and measurement point.
In analogy, the following is valid for the reference side:
dTMR
TMR =TR - r 2 - dt
(r2
40
dTMS dTMR)
TMR + T2 ( - - - - dt
dt
1[
= - Ii
fl Tm
dflTm
d2flTm
(3.5)
In addition to the 1st derivative (slope) of the measured curve flTm(t), the 2nd
derivative (curvature) must be used in the 2nd approximation to get the heat
flow rate converted in the sample. The two time constants must be determined
experimentally by proper calibration procedures. The first time constant Tl is determined by the apparent thermal resistance and heat capacity between furnace
and temperature sensor. The second time constant is determined by the apparent thermal resistance and the "effective" heat capacity between sample and
sample temperature measurement point. Exact symmetry between the sample
and the reference side is assumed in this calculation.
The above-described approximation can be refined as desired: Possible asymmetries can be taken into account by introducing different thermal resistors and
capacities on the sample and reference side into the network. This results in more
complicated equations which, however, can be solved. This enables one to correct not only the temperature but even the measured heat flow rate for possible
influences of the asymmetries (e. g., a curved zero-line, which otherwise must be
subtracted afterwards) during the measurement. Certain heat transfer problems
and a temperature gradient inside the sample and its influence on the peak
shape can be taken into account by introducing further thermal resistances and
capacities in the network of Fig. 3.7. A thicker sample with poor conductivity
can be considered as having been split up into different layers which are linked
with one another by heat-conducting boundary layers. In the end an additional
differential quotient in the differential equation and another time constant result for each additional thermal resistance in connection with a heat capacity.
The solution of such a refined approximation is as follows:
1 [
dflT
d2flT
d 3 flT
iPr(t)=-- flT+k o +k 1 --+k 2- 2 +k3 - 3 + ...
Reff
dt
dt
dt
41
In reality, the constants ki are terms into which the thermal resistances and capacities (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 exchange 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 simplification is certainly not permissible for the widespread disk-type measuring systems. Figure 3.8 is a more realistic representation of the measuring system of
such a calorimeter. When the differential equation is to be set up for this system,
it may be convenient to use the analogue electric circuit represented in Fig. 3.9
for the 2nd approximation (Rohne, 1983). In order to formulate the desired differential equation, according to Kirchhoff's laws, the voltage balance and current
balance are made up for each loop and each node of the analogue electric circuit. For the circuit of Fig. 3.9, five equations for both voltages and currents are
then obtained. On the basis of the laws of electricity, these 10 equations are simplified 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
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:
ilTst =
CR - Cs
1
RMM
. f3
-+-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
= -
2)
RMM
rf
ilT(t)dt- f 12 CR - Cs . f3dt ]
12
II
II
RMM
-+--
(R1+ -RMM2)
. f (ilT(t) - ilT.t) dt
-
12
II
This integral describes the peak area between the measured curve and the (interpolated) baseline. The approximation is the better the smaller the last two
summands in Eq. (3.6). From this follows the rule that, when heats of transition
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. Depending 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 calorimeter, 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 coefficients 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 descending slope.
Fig.3.10. Measured signal
/:"T of the heat flux DSC for
a heat pulse generated in the
sample at the time to (sum
of two exponential functions
with the time constants 11 and
12 in 2nd approximation)
t
fl.T
1---+- - - - - - -
to
-=-::-=-=-,...,-~------
time
44
Numerical Simulation
As the equations become increasingly complex and as it is impossible to solve
them analytically without introducing simplifications, it is no longer recommended to set up differential equations for calculations. 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 capacities. 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 temperature 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 approximation (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 determined by the heating rate and the thermal conductance of the sample. Only
the extrapolated peak onset temperature is relatively independent of experimental 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 transitions (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
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)
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
DSC (heating rate: 2 K min-I)
..
1
6.T
L=O.OOS W/K
shows the numerical simulation of a thermal event which is caused by the melting 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
0,25
!::,JIK
0,20
0,15
0,10
0,05
..
75
50
25
time
125
100
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 calibration factors K<[> and KQ depending on changes of the heat of reaction (Qr), emissivity (E), heating rate ({3), density (J) and thermal conductivity (A) of the sample, respectively".
The initial standard parameters (belonging to K<[> = 1 and KQ = 1) are given in parenthesis
Parameter which
is varied
QrinJ
(0.3965)
3.965
0.03965
0.5 (AI)
0.25 (Ag)
(10)
1
0.1
0.01
(7310)
73100
731
(157)
15.7
1570
a
K<[>
KQ
1
1.012
0.929
1.001
1.001
1
1.001
1.004
0.991
0.915
0.888
1
0.930
0.875
0.870
1
0.998
1.015
1
0.935
0.885
1
0.998
1.003
1
1.016
0.999
The parameters have been changed in (unrealistic) steps of 10 only, to give a better impression of their influence.
47
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 affected by systematic errors, as the calibration factor also undergoes substantial
changes. Summarizing, on the basis of the results of the numerical simulations,
the following can be stated for all types of heat flux DSCs:
48
identical sample containers). The results of the model calculations are confirmed by practical measurements (Sarge et aI., 1994).
3.2
Power Compensation DSC
In ideal power compensation DSCs, each L1T signal appearing between sample
and reference sample would be immediately compensated by a corresponding
change in the heating power. The differential heating power required for this
purpose would be equal to the differential heat flow rate as the complete electric
energy is converted into heat. If the differential heating power would be measured directly, this signal would directly stand for the heat flow rate into the sample 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
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:
50
When heat capacities are measured (see Sect. 6.1) the reference microfurnace 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 particular been used successfully to describe the behavior of DSCs in temperaturemodulated 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 approximation (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 consequences for evaluation procedures which require linear response, such as deconvolution (Sect. 5.4) in particular in peak regions of a measured curve.
Note:
In contradiction to what has been presumed in 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 temperature 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 apparatusdependent 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 additive to the measurement signal which stems from the sample processes. In every
real case, this function should first be subtracted from the measured curve before evaluations following the theoretical considerations can be made. We therefore must insert the term 4'm - 4'0 (or AT - ATo) instead of 4'm (or AT) in all
formulas used so far, this has been omitted in this chapter for clarity.
3.3
Temperature-Modulated DSC (TMDSC)
Different temperature-modulated calorimetric techniques have been known
since the beginning of the last century [for details see (Gmelin, 1997)]. Gobrecht
et al., 1971, were the first who ran a DSC with periodical changing temperature
to measure complex heat capacities. It is remarkable that the authors already in
51
this early paper point out that this method offers the possibility of "quasi-isothermal 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 temperaturemodulated DSC (Reading et al., 1993; Sauerbrunn et al., 1992) and the method became commercially available. During the past decade this method became widespread 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 Thermochimica Acta (1997, 1999, 2001) and Journal of Thermal Analysis and
Calorimetry (1998).
3.3.1
The Temperature-Modulated Method
At the beginning of TMDSC the common (linear) temperature program of the
DSC was superimposed with a sinusoidal temperature fluctuation:
T(t) = To + (Jot + TA sin(wt)
(3.9)
with {Jo the underlying heating/cooling rate, TA the temperature fluctuation amplitude and w = 2 n f the angular frequency of modulation. But other periodical
temperature fluctuations (sawtooth-like, triangular, rectangular, step-like, etc.)
are possible and have been used as well. From mathematics it is known, that
every periodic function can be written as a Fourier series and Eq. (3.9) can easily
be generalized for the non-harmonic case e.g. for an odd function:
T(t) = To + (Jot + L TA,n sin(nwt)
(3.10)
n~l
52
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 information 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 reservation 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
- =
dt
/30 + TA W cos(wt)
(3.11)
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)
-
All these modes are used in practice, often without consideration about possible
consequences concerning the result of the measurements, which may be different depending on the sample and processes in question. However, there are
different temperatures and different temperature fluctuations as well as different 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
To see what the heat flow rate in a TMDSC looks like we start from the very general heat flow rate caused by any sample in a DSC [ef. Eqs. (2.4), (2.5), (4.2)]
dT
dt
+ l1>ex. (T, t)
(3.12)
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 endoor exothermic processes occurring at a certain temperature with a certain rate.
To make the things more transparent we distinguish in the following between
different cases. They may, in reality, often happen together in the sample, but this
doesn't matter, because of the superposition principle - which is valid for heat we then simply measure the sum of the different heat flow rates in the DSC.
Case 7: Heat Capacity Sample - No Processes
This is the simplest case, the sample has a certain (vibrational) heat capacity
CiT), with a generally very weak temperature dependence, considered as constant during one period of the temperature fluctuation. In this case Eq. (3.12),
with ~ex. = 0, together with Eq. (3.11) yields:
~(T,
(3.13)
the measured heat flow rate is the sum of two components: the first one, proportional 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
t-tp/2
= Cp Po = ~u
~(T, t) dt
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 frequency as the temperature, but shifted n/2 in phase with the amplitude:
~A=
Cpo T A w
From the amplitude of the periodic part, which is normally determined via
Fourier analysis or other suitable mathematical procedures, we get the heat
capacity of the sample:
~A
Cp=cpm=--
TAw
(3.15)
54
Again we assume that the sample has a certain (only vibrational) heat capacity
CP(T) with a very weak temperature dependence, but additional processes with
endo- or exothermic latent heat exchange are taking place as well. In this case
Eq. (3.13) must be extended with the excess heat flow rate:
<f>(T, t)
(3.16)
a<f>ex. (T
aT
t)
u,
(T - Tu)
from Eq. (3.9) we see that T - Tu = TA sin(wt) and after substitution and rearrangement we get:
<f>(T, t)
a<f>ex. (T
aT u,
t)
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 occurring in the sample:
(3.18)
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:
_
cP(T, t) = Cp TA W cos(wt)
oq,eX.(Tu,t)
.
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 capacity of the sample (the same as in case 1) and one from the temperature dependence (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 resulting amplitude can be interpreted as a complex quantity with the real part
Cp TA wand the imaginary part TA . 0 cPex (Tu,t)loT]. In otherwords,Eq. (3.19)
can be rewritten in the following form:
ci>(T, t) = cPA cos(wt + 6)
6 = arctan - - - -
(3.21)
events, the phase shift is positive or negative, respectively, whereas the amplitude
becomes always greater than that of the pure Cp-sample, regardless of whether we
have an endo- or exothermic event. The increase depends on the magnitude of
the temperature derivative of the reaction heat flow (but not linearly). It should
be emphasized that every reaction is somehow temperature dependent and contributes therefore to the measured heat flow rate amplitude. But the contribution
is often small because 0 F(Tu, t)loT is generally a small quantity compared to
the product Cp (T) . w. In addition the latter contribution can be made even more
dominant by choosing a higher frequency (lower period) of modulation.
From the amplitude - as in case 1 determined from the modulated heat flow
rate with proper evaluation methods - an "apparent" heat capacity can be calculated [see Eq. (3.15)]:
Gppp. =
m cppp =
(Cp (T2
+ (~
o~x.
(Tu, t)
(3.22)
56
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 excess 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 apparent 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):
and
(3.23)
C"
tan6=L
C;
by means of Eqs. (3.20), (3.21) and (3.15) it is easy to show that in our case the
real and imaginary parts of the complex heat capacity read:
and
C; = w
o~ex.
--::1-
uT
(3.24)
(Tu, t)
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 temperature divided by frequency.
Depending on the manufacturer of the TMDSC and the software used, these
useful quantities are calculated from the modulated signal and can be further
evaluated.
Case 3: Sample with Time-Dependent Heat Capacity
In this case we are free to allow time dependent changes of the degrees of
freedom to occur in the sample, e. g., relaxation processes like vitrification or
devitrification, but we consider the temperature dependence of the heat capacity to be very weak again. However, time dependent heat capacity means a nonequilibrium 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
at that moment. That is, the simple equation ~(T, t) = Cp(T) . TIO t [see Eq.
(3.12)] is not valid any more in this case and must be replaced with an integral
equation:
~(T,
t)
= -d f~ ( Cp(T, t dt
-~
t') . -OT(t'))
- : : I dt'
ut
(3.25)
57
This defines the so-called convolution product of the time dependent heat
capacity and the heating rate. The convolution theorem means that the convolution product in time domain transforms into a common product in the frequency domain via Fourier transform ~:
~ (If>(t
= ~ (Cp(T, t . ~ (T(t
this is a complex frequency dependent function, a product of two complex functions in Fourier space:
If>*(w)
=Gp(w) T*(w)
(3.26)
In other words, a time-dependent heat capacity is equivalent to a complex frequency dependent heat capacity and the heat flow rate is frequency dependent
and complex as well. All evaluation and calculation should be done in Fourier
space to simplify matters. The resulting function [e. g., the heat capacity Gp (w) 1
can be inverse Fourier transformed to get the result in time domain again. It
should be emphasized that an evaluation of the heat capacity with Eq. (3.15)
instead of Eq. (3.26) will lead to incorrect results in the time-dependent case.
The apparent heat capacity would depend on the measurement parameters like
heating rate and temperature amplitude and the true time dependence would
never be obtained. Fortunately the relaxation times of most of the degrees of
freedom contributing to the common heat capacity are much smaller than the
time window of TMDSC experiments and their contributions appear as time independent. One process, however, the glass transition, which plays an essential
role in polymer science (see Sect. 6.4), necessitates to take a time dependent heat
capacity into account and thus a careful evaluation.
The three cases discussed in this section are the most important ones from the
practical point of view. There exist other possible cases which may be described
mathematically in a similar way. In reality all these cases occur seldom as isolated events as presented here, but are combined jointly in the same sample under
investigation. We refrain from presenting the rather complicated mathematics of
such combined cases. The advanced reader may be able to set up the formulae
himself by means of the guidelines presented here. However, we recommend
to choose experimental conditions and the measurement parameters such that
only one of the above cases is dominant in the TMDSC run and determines the
measured results and the other influences can be neglected. Because of the complicated matter any interpretation of the measured results should be done very
carefully and all possibilities should be thought of.
3.3.3
Beside the sample and the processes occurring, there are at least two more physical 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
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 propagating 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" needed to correct the measured heat flow rate amplitude and phase regarding apparatus 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 individual 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 components. These facts enable one to breakdown the rather complicated heat transfer 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 included). Knowing the transfer function enables the determination of the correction function and thus a proper calibration.
59
(P) = tan-
(3.28)
(WRthCp)
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 temperature 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?
Fig.3.15. Transfer function
(magnitude and phase) of
a simple RC-element in form
of a Bode plot
rad
,1
0.10
c
en
cs
;;
",'"
".
:::;-
......
",'"; '
-------
0.01 '""""'-=-==-:=-_-_--_-_--1-____-'-____----' 0
001
01
1.0 rad Sl
10
w--
<1J
U1
cs
-'=
CL
60
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 evaluation mathematics would result in a complex heat capacity with a non-zero imaginary 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 depends strongly on frequency, sample size (heat capacity) and its thermal resistance. For larger sample sizes this influence must be corrected as well. In cases
where the thermal resistance or the heat capacity changes during the measurement, 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 properties on the modulated signal starts from the (one dimensional) thermal diffusion equation:
(3.29)
with Dtb the thermal diffusivity. Solving this equation with the boundary condition of an periodical temperature change on one side of an infinite rod yields:
T(x, t)
with
~ __ ~2~h
LV
(3.30)
This is obviously a damped temperature wave which propagates through the rod
in time but with exponentially decreasing amplitude. From Eq. (3.30) follows:
the local amplitude at the position x is determined by the thermal diffusivity of
the heat path material as well as the frequency:
x
TA(x)
(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 normal operation frequencies, we find 5 values of about 10 to 80 mm. As the di-
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 situation is, however, totally different if we have materials with low thermal diffusivity 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):
_~
TA o' ~ (
_4)
(TA(d=d!TA,Oe~dx=-~- l-e~
= TA,o
[1- ~ :
(3.32)
+ - ...
This formula contains the case of a simple RC-element as the first approximation, which can easily be seen if we introduce ~ from Eq. (3.30) and (3.31), and
divide by TA,o to get the transfer function.
=IT
j
VI + (WRth,jCpY
and
arg(P(w
= ~ tan-l (WRth,jCp)
J
The more RC-elements are included, the more precise is the approximation of
the real transfer function of the thermal pathway of the DSC, but the more (ini-
62
QJ
"0
:::J
rad
4
0.10
2 :.c
Vl
0::
""
E
'"
OJ
Vl
'"c.
.<::
0.01
0.01
0.1
10
rad
5- 1
10
63
3.3.4
Conclusions
The TMDSC method starts from a temperature program which contains a nonperiodic and a periodic part. This causes a heat flow rate signal with two components 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 different 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 powerful 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 measurement 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.
"Calibration means the set of operations that establish, under specified conditions, 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 Confederation 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 measurements, e. g., certain sample masses, specific heating rates etc. The standards
(the written ones and the recommended reference materials) follow a "consensus philosophy" (Archer, Kirklin, 2000) which is based on round robin test procedures 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 calibration. This would enable the user to adapt (simplify) the correct calibration
procedure to their specific purposes instead of using the "standards" uncritically without considering their limitations. A collection of papers on the calibration 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 influences 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 calibration which yields the smallest possible systematic uncertainty of the measurements.
66
4.1
Aspects of Quality Assurance
Thermoanalytical instruments, including DSCs, are extensively used by industry in quality assurance systems to warrant trustworthy, reproducible properties
of the products, even products in the sense of measurement results by which
products are characterized. The ISO 9000 series and the ISO/lEC 17025,1999
are the basic standards in the field of quality management and quality assurance. The standard ISO/lEC 17025, in particular, specifies general requirements
to be met by test laboratories. It will no longer be sufficient to have technical and
scientific competence, this competence must be documented and made apparent to others. DSC methods are used for quality assurance in many ways: for example, for the inspection of incoming materials, as an accompanying measure in
manufacture and for the control of the final products.
To enable DSC methods to be useful for quality assurance, they, too, must be
subject to the requirements of the relevant standards. Calibration of the DSC is
an important precondition to fulfil these requirements. Calibration involves the
application of sensitive calibration procedures and the use of calibration substances which are - if possible - traceable 1 to national/international standards
and - in the end - to the SI-units. This means: to create confidence in the quality of the measured values and to avoid repeat measurements, metrologically
flawless calibration procedures and traceable calibration substances (Certified
Reference Materials) must be developed on the basis of which reliable uncertainty values can finally be assigned to the measurement results.
In the following sections calibration procedures and materials will be presented which are mainly based on investigations carried out by the German
Society for Thermal Analysis (GEFTA), which are the basis of the ICTAC recommendations mentioned above (temperature calibration: Hahne et aI., 1990;
Cammenga et aI.1993; caloric calibration: Sarge et aI., 1994; calibration general:
Gmelin, Sarge, 1995; Sarge et aI., 1997; Gmelin, Sarge, 2000).
4.2
Basic Aspects of Calibration
In the following, reference will be made to some important thermodynamic aspects of calibration. The same considerations are valid for the phase transitions
of a pure substance and for chemical reactions.
The quantity assigned to the substance is the enthalpy difference f...H of the
phase transition (with T = constant); it is the difference between two variables
of state and, consequently, well defined. In a calorimeter, it is the heat of transi-
67
tion Q which is measured. The relation between heat and enthalpy is given by the
1st law of thermodynamics and the definition of the enthalpy function:
(1 st law of thermodynamics)
dH=dU+pdV+ Vdp
(fromH == U+ pV)
The enthalpy H of a system is a function of the variables of state, namely pressure p, temperature T and composition ~. The total differential then reads:
d~
~H)
-vJ r.s dp + (~;)
P
p,S
dT +
(~H)
d~ -7 dE;
~ T,p
(4.1)
Under the conditions normally prevailing during DSC measurements, the first
term can almost always be neglected, even if closed crucibles are used (maximum pressure changes: 1 to 2 bar, i.e., dp :::: 0). The second term is the heat capacityat 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
(dEsurface =j:.
0) and
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 thermodynamics 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 because 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 following is obtained by differentiation of Eq. (4.1)
dT (OR)
6Qrn
- = cPrn= C ~(T)-+ dt
p.
dt
05
p.T
d5
dt
(4.2)
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 sufficient 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 conventionally 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 provided for this purpose will either be adjusted until the values indicated corre-
69
sponds to the true ones, or adaptation will be ensured via the control software
(in a way usually not apparent to the user).
Any calibration already carried out by the manufacturer must be checked
(verification of calibration) for a newly purchased equipment and again after
appropriate periods. Regular calibrations provide important information about
the repeatability error and any long-term systematic variations (drift) of the
measured quantities. (It should be noted that repeatability error is not the same
as uncertainty of measurement, cf. Sects. 7.2 and 7.3).
4.3
Temperature Calibration
Temperature calibration means the unambiguous assignment of the temperature "indicated" by the instrument to the "true" temperature.
The "true" temperature is defined by fixed points with the aid of calibration
substances. It is reasonable to choose as calibration substances, if possible, the
substances used to realize the fixed points of the International Temperature
Scale of 1990 (lTS-90, internationally legal by law). The temperature "indicated"
by the instrument must be derived from the measured curves, which usually
requires extrapolation to zero heating rate in order to eliminate/minimize the
influences of instrument and sample parameters, if possible.
Static methods (thermodynamic equilibrium) are applied to realize the fixed
points of the temperature scale. In a DSC, these can be achieved only approximately. As the point of temperature measurement is not the point where the
sample is located, a systematic error will always occur in scanning operation
which depends on instrument and experimental parameters. To establish a
graph or table, showing the relation between the indicated and the true temperatures in dependence on experimental parameters, is normally very helpful in
practice and for uncertainty estimation. In every case, a table or graph should be
drawn up which shows the variation of the indicated temperature at different
heating rates. The calibration procedures described in the following take these
special features into account in a general way, independent of the respective DSC
type (see Hahne et aI., 1990).
In the case of an endothermic event, the DSC (in heating mode) records the
heat flow rate signal schematically shown in Fig. 4.1. The section between the
initial peak temperature Ti and the final peak temperature Tf is defined as the
peak (see Sect. 5.1). The baseline needed to get the different temperatures can be
interpolated by various methods between Ti and Tf (cf. Sect. 5.3.2). The intersection between the auxiliary line and the baseline suffices to fix the characteristic
temperature Te (extrapolated peak onset temperature) which is important here.
4.3.1
70
r
o
c:
Ol
V)
I
I
r---~Y~---~-----~-~-~-----1\ 'I
I
I
I
I
I
I
I
I
I
I
I\
CD
T;
tempera ture
,I
Tf
...
- 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 possible one. The second calibration sample of the same substance is also measured at the different heating rates.
- It is to be checked whether there is a significant difference between the characteristic 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, location 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 literature (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.
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 temperature 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
72
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 calibration 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 corresponding 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 determination 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 between 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 calibration should in every case be carefully estimated (including uncertainty of
temperature sensors, uncertainty of the determination of Te etc.).
73
tempera ture
CI
c::
Vi ......_.-:;
en
T,1
TP2
74
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 particular, for power compensation DSCs, due to the isoperibol operation, a temperature profile develops in the sample support (in bottom plate of the microfurnace, 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 determined. 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:
75
time
..
If there is a thermal resistance Rth between sample (Ts) and sensor temperature (TM ) (Fig.4.6a, b),
(4.3)
is valid for the difference between the indicated temperature TM - calibrated as
described above - and the sample temperature Ts considered to be homogeneous. When a pure substance melts (as for calibration, for example), Rth can be
determined approximately from the slope of the respective heat flow rate curve
4>r(t). From Eq. (4.3) the following results after differentiation (with Ts = constant during melting, dTM/dt:::: f3 and tan y= (d4>r/dt)melting):
f3
Rth ". - -
tan y
(4.4)
76
m
L
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
- Ti cannot be determined with the required reliability because of the noise; the
same applies to Tf (the repeatability error is between 2 and 15K depending on
substance, kind of transition etc.)
- Tp and Tc strongly depend on the thermal conductivity, on mass and layer
thickness (volume) of the sample substance, on the heating rate and on the
heat transfer from sample to sample container (crucible), which may change
on melting (cf. Fig. 4.8).
- Te depends least on heating rate and sample parameters (substance, thermal
conductivity, mass, layer thickness); any possible effect of melting (heat
transfer) should be checked by carrying out various similar experiments with
the same calibration sample.
77
o
2
cJ>m
4
6
300
320
340
tempera ture
360
380
400
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)
!lTe
f3!lf3
(The values for !lTeorr (f3) have in general been rounded to two decimals.)
78
Run No.
2
2
2
2
1
2
2
2
Results:
Te (f3 ~ 0)
TflX
f:.Tcorr (f3
f:.Te/f:.f3
Te in C
10.0
10.0
10.0
10.0
5.0
5.0
2.5
2.5
1.0
1.0
0.5
0.5
0.5
0.1
0.1
156.21
156.16
156.25
156.21
155.81
155.91
155.65
155.67
155.64
155.65
155.52
155.65
155.53
155.49
155.50
= 155.523 C
= 156.5985 C (ITS-90)
= 0) = + 1.0755 K
= 0.0682 K/(Kmin- 1)
Run No.
2
1
2
2
1
2
1
2
2
2
Results:
Te (f3 ~ 0)
= 231.184 C
= 231.928 C (ITS-90)
f:.Tcorr (f3 = 0) = + 0.744 K
f:.Te/f:.f3
= 0.103 K/(Kmin- 1)
TflX
Te in C
10.0
10.0
10.0
5.0
5.0
2.5
2.5
1.0
1.0
0.5
0.5
0.1
0.1
232.35
232.18
232.06
231.81
231.69
231.54
231.41
231.33
231.22
231.28
231.12
231.26
231.12
79
156.5
7;lp.O)
ISS.5n t
7i.
M .., IP= 0)=
"""4t~
.1.111 K
156.0
T.
155.5
To: 155.523
In
+
0.0682 . P
K/min
10
(3232.5
'(
T.
m.o
231.1841"(
231.9681 '(
LlTco"IP= 0)= .0.784 K
T.lp.O)
T'i,
Sn
231.5
J
0.9--'
b 231.0
328.0
!
T.
.p
K/min
10
326.793 '(
:
7ij,
321.502 '(
LlT,o"IP= 0)= .0.709 K
T,lp.O)
328.0
(3
0
0
0
0
321.5
0
Pb
T.= 326.793 0.0997 . P
c 326.5
K/min
10
80
Run No.
Te in C
2
2
1
2
3
1
2
10.0
10.0
10.0
10.0
10.0
5.0
5.0
5.0
2.5
2.5
1.0
1.0
0.5
0.5
0.1
0.1
328.22
327.50
327.52
327.95
327.87
327.10
327.09
327.42
326.94
327.23
326.76
326.97
326.78
326.95
326.75
326.93
1
2
2
1
2
1
2
1
2
Results:
Te (fJ ~O)
Tnx
!:lTeorr (fJ = 0)
!:lTe/!:lfJ
= 326.793C
= 327.462C (converted from the IPTS~68 to the ITS~90)
= + 0.669 K
= 0.0997 K/(Kmin~l)
1.4
K
In
1.0
0.6
Sn
Pb
f':.. Tcorr(P=O)
0.2
150
200
250
Te
(P """0)
..
300
350
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
81
Te in
C
Heating rate in
Kmin- 1
~Ttot({3)
27.05
27.10
27.20
27.73
0.5
l.0
2.5
5.0
-0.19
-0.24
-0.34
-0.87
Te in
C
Heating rate in
Kmin- 1
~ Ttot ((3)
29.92
29.97
30.11
30.35
30.94
0.5
l.0
2.5
5.0
lO.O
-0.16
-0.21
-0.35
-0.59
-1.18
in
Tlit = 26.86C
Te({3 --+0) = 26.93C
~ Tcorr({3 = 0) = - 0.07 K
Gallium
in
Te in
C
Heating rate in
K min- 1
~ Ttot ((3)
67.35
67.36
67.47
67.66
68.06
0.5
l.0
2.5
5.0
10.0
0.05
0.04
-0.07
-0.26
-0.66
Tlit = 67.4C
Te({3 --+0) = 67.3C
~Tcorr({3 = 0) = +0.1 K
0.12
0.12
Tlit = 7l.1 C
Te ({3 --+ 0) = 70.91 C
~Tcorr({3 =0) = +0.2K
in
Te in
C
Heating rate in
Kmin- 1
70.98
70.98
71.10
7l.27
7l.70
0.5
l.0
2.5
5.0
10.0
-0.17
-0.60
Benzoic acid
Te in
C
Kmin-
122.28
122.23
122.34
122.46
122.72
0.5
l.0
2.5
5.0
lO.O
Heatin~
rate in
0.06
0.11
-0.12
-0.38
Tlit = 122.34C
Te ({3 --+ 0) = 122.22C
~Tcorr({3 =0) = +0.12 K
82
Te in
C
Heating rate in
Kmin- 1
156.63
156.67
156.80
156.96
157.32
158.01
159.34
162.17
168.60
0.5
1.0
2.5
5.0
10.0
20.0
40.0
80.0
160.0
0.03
0.07
0.20
0.36
0.72
- 1.41
- 2.74
- 5.57
-12.00
Te in
DC
Heating rate in
Kmin- 1
tJ.Ttot ({3) in
K
232.18
232.21
232.34
232.52
233.Ql
0.5
1.0
2.5
5.0
10.0
-0.25
-0.28
-0.41
-0.59
-1.08
Te in
DC
Heating rate in
Kmin- 1
tJ.Ttot ({3) in
K
236.30
236.30
236.36
236.67
236.96
0.5
1.0
2.5
5.0
10.0
-0.20
-0.20
-0.26
-0.57
-0.86
Te in
DC
Heating rate in
Kmin- 1
237.99
238.00
238.17
238.32
238.81
0.5
1.0
2.5
5.0
10.0
-0.53
-0.54
-0.71
-0.86
-1.35
Tin
Caffeine
Tlit = 236.1 DC
Te ({3 ~ 0) = 236.24 DC
tJ.Teorr({3 = 0) = -0.1 K
Lead
Tlit = 327.462 DC
Te({3 ~ 0) = 327.93 DC
tJ.Teorr({3 = 0) = -0.47 K
83
Table 4.5 The dependence of Te on the heating rate f3 for various classes of materials (results from Table 4.4)
Substance
Gallium
Indium
Tin
Lead
0.11
0.07
0.09
0.09
0.090.02 a
Diphenyl ether
Paraffin C33H68
Benzoic acid
Caffeine
0.15
0.08
0.05
0.07
0.09 0.04 a
0.090.03 a
ATtot = Tlit - Te (P) is defined as the difference between the "true" temperature
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 temperature of the transition. But, depending on the heat transfer between substance 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. Separate 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 measurement 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
328.0
C
327.8
327.6
Te 327.4
{3
8 K/mln 10
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 compensation 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
85
D. T(P)
CII'
. - - - (Doling rate
Pbf.at
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 interest.
- 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 determined.
- It is to be checked whether the results from the two experiments differ significantly.
- 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 respective 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
340.6
K
340.2
r.
....::.
_0
340.0
339.8
339.6
x--x----
c _ _c
00-
-----:::=:
_0--- o_____
~
c
O~
c
_ _x---
-~
____ x
339.4
339.2 -2.5
-2.0
-1.5
-1.0
-0.5
(3
0.0
0.5
1.0
1.5
K/min
2.5
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 example 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 proportionality factor between the measured heat flow rate cJ>rn and the true heat flow
rate cJ>true on the one hand, and between the measured exchanged heat Qrn and
the heat Qtrue really transformed, on the other hand, is to be determined:
cJ>true =Kif>' cJ>rn
and
Qtrue = KQ Om
Strictly speaking cJ>rn in this equation should be considered the measured heat
flow rate with the instrument zero line already subtracted, but as all correction
calculations are usually done using the measured curve cJ>rn itself, only this is
used in the following.
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 Crystallization 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 equation 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 conclusion to be drawn for heat calibration is as follows:
The thermophysical behavior of calibration sample and sample to be measured must be as similar as possible. As this is possible only approximately, systematic errors exist which must be estimated and included in the overall uncertainty of measurement (Sect. 7.3).
4.4.1
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 following applies analogously, IPm being replaced by flU.] The heat flow rate calibration 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
The proportionality factor (calibration factor, calibration function) K~ usually depends on parameters such as temperature and - what is most important heat flow rate. In some DSCs, K~ is made unity by electronic or software means.
In these cases, too, the relation between cPmand cPt rue must be carefully checked.
The heat flow rate calibration can be performed in two ways:
1. Installation of an electric calibration heater in the place of the sample or in-
so that
K (T) _
I/>
Cp(T) f3
89
m- bl
r
m
4't,1
a
tempera ture
r
.,.~
temperature
90
For peak area calibration, a known heat Qtrue, dissipated or consumed, is compared with the area of the resulting peak (Figs. 4.16 and 4.17) and the following
is valid:
tf
or
ti
Qtrue
tf
=f
ti
In principle, K~"# K;' is valid. Since K~ depends on cI>m and T - which is proportional 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 approximation 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 contributions of the 1st derivative of the measurement signal are insignificant.
Fig.4.16. Electrical peak
area calibration.
tPm measured heat flow rate,
tio tf beginning and end
of the calibration peak,
~l baseline value of the
heat flow rate (electric power
switched off)
tf
f(m-"Idt= A
Ii
time
...
91
time
Peak area calibration can be performed in two ways:
1. If electrical calibration is possible (Fig. 4.16), the advantages connected with
electrical heat flow rate calibration (see above) apply analogously to peak area
calibration:
- Easy and accurate measurement of the electrically generated heat
Qtrue = -
Pel. dt = - WeI. = - U i . 11 t ,
92
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 considerations must be taken into account too),
- measurement of the repeatability errors of the calibration factors (or calibration curves). This repeatability error must be clearly smaller than the estimated overall uncertainty of the calibration (see above). The repeatability error
is the smallest possible uncertainty of measurement of caloric measurements.
The same calibration procedure is applied to power compensation DSCs,however,
in this case, calibration with one calibration material 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
Calibration Curve
A DSC with cylinder-type measuring system was calibrated with the aid of electric heaters which could be inserted into the sample containers. Only the insert
in the sample container was electrically heated, the second insert of the same
type placed into the reference container served to establish thermal symmetry.
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 automatically generated over the whole temperature range, the measured signal
being a voltage in ]lV. The peak areas were integrated applying evaluation software. Figure 4.18 shows schematically the measured curve and the resulting calibration 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 literature values (Table 4.9) enables to calculate the calibration curve.
Differences Between K<p and KQ
When DSCs are calibrated with the aid of the heat of fusion of, for example, metals (peak area calibration) and via the specific heat capacity (heat flow rate calibration), 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
B
mW
4
0
1
~
0
3
~
-4
-B
-12
OJ
.c
-16
-20
00
4.0
B.O
120
16.0
20.0
...
24.0
time
32.0
2B.0
36.0
min
44.0
300
500
tempera ture
600
..
95
10
!-LV
mw 9
ta
7
11K
100
200
temperature
..
O(
300
Fig.4.21. Differences between the heat-of-fusion calibration (peak area calibration) and the
calibration by means of a known heat capacity (heat flow rate calibration) (according to Sarge,
Cammenga,1985).
K calibration factor, - calibration curve with sapphire (3 measurements at lO K min-I),
.... average value curve of the calibration with melting samples
For a heat flux DSC with disk-type measuring system, Fig. 4.21 shows the differences between the average calibration curve from of a number of heat -of-melting (peak area) calibrations and the curves from three individual calibration runs
carried out with the heat capacity of a sapphire sample (heat flow rate calibration).
Dependence on Other Parameters
The repeatability error by which the calibration with a known heat capacity is
affected can additionally depend on the sample mass. To give an example: the
repeatability error of the calibration of a heat flux DSC with disk-type measuring 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 between the calibration factors, which results when one calibration is carried out
with the calibration heater permanently installed (under the bottom of the sample 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
30
mJ
mV s
25
20
KQ
..
15
'
300
200
100
tempera ture
"(
400
Fig.4.22. Differences in the electrical peak area calibration of a heat flux DSC with cylindertype 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
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
disk-type DSC (according to Breuer, Eysel, 1982)
KQ
0
-1
10 mg
f3 for
IS
IS
(i/(Kmin- 1 )
10
o 1 mg
0
20
40
80
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)
97
4.4.4
Caloric Calibration in Cooling Mode
The heat flow rate calibration in the cooling mode can be performed by complete 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 cooling mode can differ systematically from the results determined in the heating
mode. This effect is more pronounced at measurements on low thermal diffusivity materials. The reason for that temperature- and! or instrument-dependent
behavior is a non-symmetric heat flow in heating and cooling mode. It is recommended 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 determination 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 considered. That means the dependence of the measured enthalpy of freezing as a function 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 resulting correction coefficients are given in Table 4.13 together with the substances recommended for the heat calibration in cooling mode.
4.5
98
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
The requirements to be met by substances to be used for the calibration of DSCs
are as follows:
99
When there are any doubts as to the usability of a substance, i. e., when the risk
of persons being injured or equipment damaged cannot be excluded, information should be obtained from experts prior to using such substances.
The following must be observed when calibration samples are prepared:
-
The fixed-point temperatures of metals are associated with the flat part of the temperatureversus-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
Table 4.6. Defining fixed points of the ITS-90 and some secondary reference points
TinK
Triple point of equilibrium hydrogen
Triple point of neon (ITS-90)
Triple point of oxygen (ITS-90)
Triple point of argon (ITS-90)
Triple point of mercury (ITS-90)
Water-ice point
Triple point of water (ITS-90)
Melting point of gallium (ITS-90)
Boiling point of water
Solidification point of indium (ITS-90)
Solidification point of tin (ITS-90)
Solidification point of bismuth
Solidification point of lead
Solidification point of zinc (ITS-90)
Solidification point of aluminium (ITS-90)
Solidification point of silver (ITS-90)
Solidification point of gold (ITS-90)
Solidification point of copper (ITS-90)
Solidification point of nickel
Solidification point of palladium
Solidification point of platinum
Solidification point of rhodium
Solidification point of iridium
13.8033
24.5561
54.3584
83.8058
234.3156
273.15
273.16
302.9146
373.124
429.7485
505.078
544.552
600.612
692.677
933.473
1234.93
1337.33
1357.77
1728
1828.0
204l.3
2236
2719
tin C
-259.3467
-248.5939
-218.7916
-189.3442
- 38.8344
0.00
0.01
29.7646
99.974
156.5985
23l.928
27l.402
327.462
419.527
660.323
96l.78
1064.18
1084.62
1455
1554.8
1768.2
1963
2446
The substances used to realize the fixed points of the ITS-90 are listed in
Table 4.6 starting with the triple point of hydrogen. These substances have been
marked by "ITS-90" (cf. ITS-90, Preston-Thomas 1990).
Substances which were used to realize the defining fixed points of the International 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 December 31,1989). In addition, a selected set of secondary reference points was included (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 temperature ranges which is why other substances, too, must be used for the temperature calibration of DSCs. Taking the above criteria into consideration, the
German Society for Thermal Analysis (GEFTA) recommends that the substances
listed in 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
101
Table 4.7. Materials recommended for temperature calibration of DSCs (cf. Cammenga et ai.,
1993)
Substance
Cyclopentane
Cyclopentane
Cyclopentane
Water
Gallium
Indium
Tin
Lead
Zinc
Lithium sulfate
Aluminum
Silver
Gold
Transition
temperature
Uncertainty
TinK
TinoC
dTinmK
122.38
138.06
179.72
273.15
302.9146
429.7485
505.078
600.61
692.677
851.43
933.473
1234.93
1337.33
-150.77
-135.09
- 93.43
0.00
29.7646
156.5985
231.928
327.46
419.527
578.28
660.323
961.78
1064.18
50
50
50
10
<1
<1
10
250
Transition a
sIs
sIs
sn
sll
sn
sn
sn
sn
sll
sIs
sn
sn
sn
Remarks
1
2
3
4
5
3
6
7
8
8
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
+
+
+
+
+
+
+
+
+
+
+
+
+
+
0
0
0
0
0
0
+
+
+
+
+
+
+
+
....Q)
0-
Q)
1::'"
.::
Q)
+
+
+
+
+
+
'"
+
+
+
+
+
+
+
+
+
....:I
Q)
'"0
+
+
+
+
+
+
+
+
+
+
+
+
"+
.::
'"0
.5
::;;;
a
.8
a::l
Corundum, Aiz03
Boron nitride, BN
Graphite,C
Silicate glass
Quartz glass, Si0 2
Aluminum, Al
Aluminum, oxidized
Silver,Ag
Gold, Au
Nickel,Ni
Iron, Fe
Stainless steel
Platinum, Pt
Molybdenum, Mo
Tantalum, Ta
Tungsten,W
Crucible material
Calibration
substance
+
+
+
.::
....
....~
t;j
.....
o~::l
....
g;
....
n
+
+
....en~
t1>
....0
eo.
C1CI
S
n
::l
::l
t"\
a..
eo.
C/)
x
x
x
x
+
+
+
x
x
x
"
+
'"0
"0
....
~<Zl
t1>
x
x
+
+
+
.::
's
a::l
IS
+
+
+
+
+
+
+
a'"::l
"3
..s
Q)
Table 4.8 Compatibility between calibration substances and crucible materials (according to Cammenga et ai., 1993)
103
stable and the heat capacity is known in a wide temperature range with an uncertainty 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 discussed 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 calibration 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 published modified polynomial coefficients. The difference between Martin's data
Table 4.9 a. Specific heat capacity of a- Ah03 (synthetic sapphire, corundum, NIST SRM 720)
(according to Ditmars et al., 1982)
Molar mass 101.9613 g mol-I
Temperature range
To
40K
125K
ai
-3
-2
-1
0
1
2
3
4
5
6
CiTi
i=-3
6.966.10- 1
5.9387 . 10-2
4.0357 . 10-3
9.5173.10- 5
-3.5910. 10-6
- 6.498 . 10-7
4.089.10-9
bi
2.1993.10 1
3.8853 . 10-1
1.3955 . 10-3
- 8.3967 . 10-5
1.9133. 10-6
-3.1778.10- 8
2.9562 . 10-10
Ci
-1.32506 . 108
4.54238.106
-5.475599. 104
2.574076. 102
-1.715032. 10-1
1.2897189 . 10-4
-4.60768. 10-8
6.31755. 10- 12
104
Table 4.9 b. Specific heat capacity of a- Alz0 3 (synthetic sapphire, corundum, NIST SRM 720)
Molar mass 101.9613 g mol- 1 (according to Ditmars et al., 1982 )
TinK
cp in Jg-l K- 1
cp in Jmol- 1 K- 1
100
150
200
250
300
350
400
450
500
550
600
650
700
750
800
850
900
950
1000
1050
1100
1200
1250
1300
1350
1400
1450
1500
1550
1600
1650
1700
1750
1800
1850
1900
1950
2000
2050
2100
2150
2200
2250
0.1261
0.3134
0.5014
0.6579
0.7788
0.8713
0.9423
0.9975
1.0409
1.0756
1.1039
1.1271
1.1467
1.1636
1.1783
1.1913
1.2030
1.2138
1.2237
1.2330
1.2417
1.2578
1.2653
1.2724
1.2792
1.2856
1.2917
1.2975
1.3028
1.3079
1.3128
1.3173
1.3214
1.3253
1.3289
1.3324
1.3356
1.3387
1.3417
1.3446
1.3477
1.3508
1.3540
12.858
31.95
51.12
67.08
79.41
88.84
96.08
101.71
106.13
109.67
112.55
114.92
116.92
118.64
120.14
121.47
122.66
123.76
124.77
125.72
126.61
128.25
129.01
129.74
130.43
131.08
131.70
132.29
132.84
133.36
133.85
134.31
134.73
135.13
135.50
135.85
136.18
136.50
136.80
137.10
137.41
137.73
138.06
105
aj
0
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
- 0.8209550462989
0.1877774093791
-0.1572548380193. 10-1
0.5828318431167. 10-3
-0.1420296394933.10-5
-0.3466012703872 . 10-6
0.1030643882976.10-7
-0.1621415050746. 10-9
0.1678243825986. 10-11
-0.1225826347399.10- 13
0.6497335630403 . 10-16
-0.2516918128676. 10-18
0.7065147741950. 10-21
-0.1400307720276.10-23
0.1858984964834.10-26
-0.1483657580275.10-29
0.5382629833814. 10-33
cp in Jg-I K- 1
cp in Jmol-I K- 1
100
120
140
160
180
200
220
240
260
280
300
320
0.2520
0.2871
0.3122
0.3306
0.3444
0.3550
0.3635
0.3703
0.3758
0.3806
0.3848
0.3884
16.02
18.25
19.84
21.01
21.89
22.56
23.10
23.53
23.88
24.19
24.45
24.68
106
0.008
-o. 008
L--~-'---~-'---'---'-~--'---'---'---'---'______L--.>--.J
250
750
1250
1750
2250
temperature - Fig.4.25. Differences between tabulated and calculated (polynomial presentation) specific
yn. (cf. Ditmars et al., 1982) in the temheat capacity data of sapphire (Ah03): p =
perature range between 250 K and 2250 K
I!1c cr -cr
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 temperature range between 100 K and 300 K.
The specific heat capacity of Al20 3 has been determined in the temperature 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 calculated 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.
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 requirements. 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 certification 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 temperature 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 substances 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 calibration is still going on at the international level (St01en, Gr0nwold, 1999). The following complexes of problems are concerned:
- backing-up and re-determination of characteristic data, including an estimate of the uncertainties,
- testing of new substances for their suitability for caloric calibration purposes,
- search for substances suitable for special calibrations, for example polymers
(glass 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 temperaturemodulated 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
sis
sis
sll
-150.77
-135.09
-93.43
29.76
69.26
156.60
231.93
271.40
578.28
660.32
916.78
1064.18
Cyclopentane
Cyclopentane
Cyclopentane
Gallium
Biphenyl
Indium
Tin
Bismuth
Lithium sulfate
Aluminum
Silver
Gold
399.9 1.3
104.6 2.1
64.6 1.5
28.58 0,07
60.42 0.17
53.33 0.26
79.97 0.27
Qtrs
(St01en et aI., 1999)
in J g-l
80.097 0.032 b
120.41 0.57 c
28.51 0.19 d
60.22 0.1g e
Qtrs
(NIST)
in J g-l
s: solid; 1: liquid.
Archer, 2002.
C NIST SRM 2222 (NIST: National Institute of Standards and Technology, USA).
d NIST SRM 2232.
e NIST SRM 2220.
f PTB Certified Reference Materials (PTB: Physikalisch-Technische Bundesanstalt, Germany).
sll
sll
sll
sis
sll
sll
sll
sll
sll
Type of
transition a
Ttrs (ITS-90)
in C
Substance
28.64 0.11 f
60.24 0.27 f
53.14 0.22f
80.14 0.33 f
Qtrs
(PTB)
in J g-l
0.35
0.05
0.09
0.72
28.62 0.11
60.40 0.36
53.8 2.1
228 10
398 9
69.60
4.91
8.63
79.88
Qus
(Sarge et ai. , 1997)
in J g-l
......
....
'"en
s"
S-....
(")
5"
()
Vl
'"....
'"g.
e.
CI
o
.....,
g"
~
....
(")
...
00
109
Table 4.11 b. Further Certified Reference Materials for temperature and heat calibration
Substance
Ttrs
Phenyl salicylate
Biphenyl
Naphthalene
Benzil
Acetanilide
Benzoic acid
Diphenylacetic acid
Lead
Zinc
Aluminum
Qtrs
in C
inkJ mol- 1
41.79
68.93
80.23
94.85
114.34
122.35
147.19
327.47
419.53
660.33
19.18
18.60
18.923
23.26
21.793
17.98
31.16
4.765
7.103
10.827
source: http://www.lgc.co.uk.
(LGC: Laboratory of the Government Chemist, UK).
Table 4.12. Substances recommended for temperature calibration in the cooling mode
Substance
Adamantane
M24 a
HP-53 b
BCH-52c
Indium
NaN0 3
Zn d
Transition
s~s
SA~N
SA~N
N~I
s~l
s~s
s~l
Evaluate
Te
Tp
Tp
Tp
Te
Tp
Te
Phase transition
temperature
Supercooling
Ttrs in C
Ttr, in K
8TinK
-64.49
66
116
162
156.6
276
419.5
208.66
339
389
435
429.8
549
692.7
<1
<0.5
<2
<0.5
4-cyano-4' -octyloxybiphenyl.
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.
a
110
Table 4.13. Substances recommended for heat (peak area) calibration in the cooling mode
Substance
Fusion
temperature
Tfus in C
Cydopentane
Cydopentane
Cydopentane
Gallium a
Indium
Tin
Zinc a
Lithium sulfate
Aluminium
a
-150.77 C
-135.09 c
-93.43
29.7646
156.5985
231.928
419.53
578.28 c
660.323
Enthalpy of fusion
1'1fus H in Jg-1
69.60 c
4.91 c
8.63
Temperature dependence b
d1'1 trs H/dT= C;(T) - Cp(T)
= 1'1trs Cp (T) = a + b (Ttrs - Tfus)
a in Jg -IK- 1
binJg- 1K- 2
+0.38
-0.058
+0.16
+0.031
-0.0026
-O.oI8
+0.012
-0.15
-0.19
-5.8. 10-3
0
+3.1 . 10-3
-4.6. 10-4
-2.6. 10-4
-3.1 . 10-4
-7.3 . 10-4
-7.7. 10-3
+3.2. 10-4
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!
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 hightemperature 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 negative, 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.
Solid/solid transition.
causes a loss in amplitude and a shifting of the phase of the heat flow rate modulation on the path from the furnace to and into the sample. Therefore the (apparent) heat capacity, determined from the modulated part of the heat flow rate signal 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 capacity in addition to the normal calibration of temperature, heat flow rate and
peak area mentioned above. The relevant equation in this special case reads:
Cp, true =
(4.5)
The normal (vibrational) heat capacity does not depend on frequency and this
fact is used to determine the calibration factor for the frequency of the temper-
111
with pew) the complex transfer function of the device. Because of the proportionality of temperature gradient and the heat flow rate a similar equation can
be formulated for the heat flow rate
q,~eas.(w)
= q,~ue(w) P*(w)
(4.6)
and even for the heat capacity which is linearly connected with the heat flow rate
(4.7)
Generally the functions are complex (characterized by the star) but, of course, the
equation is even valid for real functions. Comparing Eq. (4.7) with Eq. (4.5) yields:
_
1
K(w)=-P*(w)
(4.8)
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 function, 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 measurement 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 calibration factor [see Eq. (4.8)], a complex function of was well. With that the calibration, 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 calibration in a more usual way. Any complex quantity can be expressed as magnitude (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
/\
141
I \
I \
I
\
,......II....... \\...........................tg.r!~p.~r.gJ~xg.::::f.:.........
;
:
;
aJ
I
I
I
............ ,
\
\
\
! ._
:3
o
~:
",
140
\,___
sample line
<-
ill
D..
ill
OL---~---L
-10
10
__~==~====b===d 139
20
30
50
time - - - Fig.4.26. Temperature-step (0.5 K, dotted) response signals (dashed) of a power compensation DSC containing a sapphire sample (25 mg) as well as empty pans (for zeroline). Subtraction results in the response function (solid line)
rad
Plw)
0100
1i 0.010
.,,,
0.001
001
01
--
10
(;J
rad
I
ill
Vl
..c
D..
S-l
10
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 discrete 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
113
Calibration of Magnitude
A simple procedure for magnitude calibration starts from Eq. (4.5) and determines 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 frequencies 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 frequencies (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 procedure 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 frequency 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. Unfortunately this is not true, as the heat capacity of the sample is different from the
114
heat capacity of the reference material (sapphire), even the phase shift is different. 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 temperature well outside of any transition or reaction, to be sure that only the vibrational heat capacity contributes to the signal.
However, there is another problem arising from the fact that the heat capacity 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 particular true for the glass transition region, where the heat capacity becomes complex, 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 sigmoidal 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 complex 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 reaction 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 realvalued 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 (normalized) heat flow rate function.
It should be emphasized that the phase correction 15K changes during transitions, 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).
t
T,
temperature
T,
..
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 temperature, CD initial segment of the measured curve, @ step of the measured curve due to IlCp ,
@ measured curve, interpolated baseline (between Tj and Tf in the peak region), final
segment of the measured curve, peak (measured curve)
116
5.1
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
1
1
1
---I
I
--
peak (endothermic)
/
\
.-!-\-QI
'\
I
baseline (interpolated)
T;
T,
T.
temperature
zero line
Fig.5.2. Definition of zeroline, baseline, peak and the characteristic temperatures (definitions see text).
Ti initial peak temperature, T. extrapolated peak onset temperature, Tp peak maximum temperature, Tc extrapolated peak offset temperature, Tc final peak temperature
117
measuring system itself and the degree and influence of unavoidable asymmetries. The smaller its range of variation with temperature or in time (repeatability, see Sect. 7.2), the better the instrument.
The baseline is the part of the heat flow rate curve produced of the DSC during 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 difference 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 international 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 constructed 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 intersects the baseline.
- Tr Final peak temperature
Here the curve of measured values reaches again the baseline, the peak is
completed.
118
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 analogy 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 specific DSC) at least with one calibration substance, e. g., indium. If the DSC shows
a distinct tendency towards drifting, a daily test is important. In this way, information is obtained about drift processes or scatter which cannot be assigned to
an exactly known parameter. This information enters into the estimate of the
overall uncertainty.
The influencing parameters are listed in the following:
1. The shape of the zeroline (without crucibles) is influenced by the heating rate,
the kind of purge gas and its flow rate, likewise by the temperature of the surroundings 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 crucible 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 temperature dependence) are of importance (differential measurement). In the case
of pure Cp changes of the sample (e. g., glass point, Curie point), these changes
determine the 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 separation (resolution) of closely adjacent peaks (and the calibration).
119
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
To eliminate all influences from asymmetries of the DSC in question, the zeroline 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. Depending 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
120
dt
cPs
cPm
cPo
cp = - = - - - = - - =
m
dT m . f3
m . f3
mdt
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 capacity 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", different from the static (vibrational) heat capacity and, in addition, "smeared" (see
Sect. 5.4). However, there are great advantages of such a presentation; the advantages 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 capacity curves generally from the original measured curve.
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) =
Cp(T)dT
To
JCp(T)
dT
T
To
The baseline has to be constructed in the region of peaks of all kinds to be able
to determine peak areas. For 1st order transitions, with or without coupled Cp
change (and/or change of the heat transport mechanism), the baseline can be
constructed in different ways (cf. Hemminger, Sarge, 1991). A definite baseline
can be constructed only if a pure Cp change occurs; changes of heat transport
conditions in the range of a peak give rise to uncertainties in the shape of the
baseline. For certain evaluations (for example kinetics), the measured curve and
the respective baseline must possibly be "desmeared" as well (cf. Sect. 5.4). The
common baseline construction methods are as follows.
1. For irreversible transformations without measurable cp change, the baseline
can basically be determined by repeating the measurement with the same
sample whose reaction process has come to an end (example: annealing of
lattice defects during recrystallization of a plastically deformed metal). In
these cases, the baseline is interpolated in the course of the measured curve
of the 2nd run. The baseline uncertainty corresponds to the repeatability
error of the DSC. If the heat capacity of the reacted product is different from
that of the original sample, this method cannot be used.
122
Ti
dT
Tr
That is to say, aCt) determines how quickly the slope of the baseline changes
from the slope of the measured curve at Ti (with a = 0 for T < Ti ) in the interval 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
tempera ture
..
T,
Fig.5.3. Construction of the baseline taking the degree of reaction a (t) into account (according to van der Plaats, 1984).
Tj initial peak temperature, Tr final peak temperature, <1\ex> <Pr,ex measured curve extrapolated from Tj , Tf into the peak region
123
r
rp
T,
temperature
..
124
Integration of the measured curve with the baseline subtracted yields the
heat of transition or reaction. This heat equals the enthalpy of reaction or transition, 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 specified, 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 absolute values (at equal temperature) and also in their temperature dependence.
The kind of functional dependence of the heat capacity on the temperature (linear dependence, polynomials of higher order) is unimportant. Extrapolations
into temperature ranges that are not directly measurable must, however, be reliable. If this is guaranteed, the heat absorbed over a wide temperature interval
125
f
>.
'-'
0
0-
(P . (T)
'-'
~
Q)
.c
..
tempera ture
Fig. 5.5. The determination of thermodynamically valid values for the enthalpies of transition (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:
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 )
= (H[(T2 )
Hs (Td) -
H[(T)
T2
Tj
=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 special 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 determined with sufficient accuracy according to the following equation:
S(T) - S(To) =
To
C (T)
- p - dT
This is not allowed in the region of phase transitions because the peaks are always smeared and the plotted temperature is never the true equilibrium tem-
126
perature Teq of the sample. However if, in compliance with the above prescription, 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 temperature 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 representation of the functions searched for, and the desired evaluations can be directly made.
When the accuracy requirements are higher, or kinetic evaluation must be performed, 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 dispensed 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 order endothermic phase transition. Accordingly, desmearing takes place in several steps:
a) Linear transformation of the abscissa (indicated temperature) in compliance
with the results of the calorimeter's temperature calibration, and transformation 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 transport phenomena inside the DSC (heat flow rate relaxation) exert on the measured 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 temperature 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.
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
5.4.1
Correction of the Temperature and Heat Flow Rate Indicated
This procedure has already been described in Sect. 4.3 and 4.4, however, for the
sake of completeness, reference to it will be made again in the following.
For the correction of the temperature indicated at the heating or cooling rate {3,
the following is valid:
Ttrue = Tm + I1Teorr{fJ)
I1Teorr {f3) = I1Teorr {{3= 0) -I1T{f3)
(S.1)
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 temperature axis scaling of the output of the measured function, a shift which is different for each heating rate. A simple correction is concerned here which must of
course be carried out only if it is significant for the accuracy required for
the measurements. The correction of the displayed temperature according to
Eq. (S.l) takes only influences from the instrument into account. Samples of great
thickness possibly require an additional correction which also considers the temperature proflle inside the sample; this correction is described below (Sect. 5.4.4).
Note: The recently introduced turret -type DSC (TA Instruments, see Sect. 2.1.2)
takes the heat capacity and thermal conductivity of the measuring system into
account and corrects for the thermal lag with the so-called Tzero technique. As
a result either the temperature of the sample sensor or the temperature of the
sample crucible is calculated and used as temperature axis for the measured
curve. These corrections are normally sufficient for common measurements
and additional corrections of the temperature scale are not necessary if careful
calibration has been done.
The following equations are valid for the correction of the displayed heat flow
rate (cf. Chapter 3 and Sect. 4.4.1):
IPtrue = K~ IPm
or
IPtrue = - K . 11 Tm
129
5.4.2
It is not possible to produce a perfectly symmetrical DSC in which the differential signal between empty sample system and empty reference sample system is
exactly zero over the entire temperature range. This applies also to the measured
curve with the sample crucibles empty (zeroline).
To obtain the true heat flow rate into the sample (for example, to calculate the
sample's heat capacity), the zeroline must be separately determined by a second
measurement and subtracted from the curve actually measured. For the correct
determination of the zeroline see Sects. 5.3.1 and 6.1.1.
The correction of the instrument asymmetry presents the problem that the
two measurements can be carried out only one after the other; it must therefore
be ensured that the DSC's symmetry properties have not changed in the meantime. This is a strict condition which means that high requirements must be met
by the calorimeter (cf. Sect. 7.2). In addition, it is indispensable that the DSC
is opened between the two measurements and the sample crucible changed or
moved. This alone can change the symmetry properties, which results in the
zeroline being shifted.
The design of the evaluation software of the DSC computers is normally such
that the zero line (empty crucibles) is first subtracted and then the corrections
of the temperature and heat flow rate scales (see above) are made. This order is
justified only if the scale corrections for the measurement and the baseline are
the same. In all other cases, first the correction and then the subtraction must be
made. The same is true of all advanced desmearing procedures presented below.
5.4.3
130
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:
cPr(t)=cPm(t)+al
dcPm(t)
dt
+a2
d 2 cPm(t)
2
dt
(5.2)
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 coefficients (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 technique. As a result the measured curve is automatically desmeared and the software produces a curve of the heat flow rate into the sample crucible as a function
of sample crucible temperature. This correction includes the three steps of
desmearing described so far. This method takes a careful calibration of the DSC
and unchanged measuring conditions as granted to give reliable results.
Numerical Methods
If the behavior of the DSC cannot be described by simple differential equations within the framework of the required accuracy (e. g., in the case of
power compensation DSCs), another method must be applied which has its
roots in the theory of linear response. It is valid for all measuring systems
which work linearly, that is to say the measured signal for two distinct pulse-like
events in the sample must be the sum (superposition) of the two single functions from each individual event (Fig. 5.7). Another condition is that all measured functions (curves) of various pulse-like events should have the same shape,
in other words all these measured functions divided by their peak area (normalizing) must yield the same function, the so-called apparatus function a (t)
131
T,
T2
temperature
r
b
tempera ture
(or even called Green's function). If these conditions are fulfilled, the following
is valid:
tl>m (t) = c f tl>r (t') . a (t - t') dt'
==
tl>r (t)
* a (t )
(5.3)
This defines the so-called convolution product of two functions, a integral equation 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
The seamy side of this desmearing method, also called deconvolution 1, is the
rather ambitious mathematics required to solve the integral equation (5.3) for
the function of interest 4>r(t). There are essentially two methods, the Fourier
transform and the recursion method. Both require numerical calculations with a
computer.
The Fourier transform represents an integral operation:
{S (f(x
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 commonly included in many mathematical software packages and therefore easily
available. The drawbacks of this procedure lie in its laborious course and abstract nature, since the calculations are performed in Fourier space. Those who
lack experience in numerical Fourier transforms are advised to study some "pitfalls" such as the "break-off effect" and the "sampling theorem", both obtained
by numerical treatment. Under specific conditions this simulates periodicities
and fluctuations which do not reflect any actual processes in the sample (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)
133
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 converges 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 measurements have been concluded (offline). For the case of a discrete (sampled)
function, the convolution integral [Eq. (5.3)] can, however, be represented as a
linear system of equations.
If>m(ti)
= c M L
k=l
(If>r(tk)' a(ti-k+l))
(i = 1 ... n)
5.4.4
Advanced Desmearing
The desmearing procedures described so far have essentially taken into account
the influences from the DSC measuring system and the interaction between
sample and instrument. The events inside the sample have not yet been discussed.
134
"" f3 = 0
dtPr
dTm
dt
dt
-=L-
(5.6)
d tPr
dt
d tPr
dt
dTm
dTm
--=-=L
135
Tl
..
T2
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 theoretical 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. -
3..\
(5.7)
136
/1
1
1
r--------=------f-----+------~-----I
I
1
1
1
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 possibility 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.
137
2~
ClJ
~
L..
:J:
0
ClJ
.<=
time
Fig.5.10. Curve measured with the heating rate switched from zero to f3 at to.
1 measured curve, 2 theoretical heat flow rate into the sample
O.B
Sl
0.6
0.4
C;
0.2
0.0
0
10
30
20
time
40
60
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 response function can be derived from this function (Fig. 5.11), which is the apparatus function (Green's function) of the total system. Using this apparatus function desmearing can be carried out as described. Figure 5.12 shows that such
a desmearing gives rise to a distinct change in the measured curve in the case
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
2.2
J
glf
1.8
f
Co
1.6
LJ
1.4
1.2
350
360
370
380
390
temperature
..
410
exert on the measured curve. An advantage of this method is the easy measurement of the correct apparatus function by only switching from the isothermal to
scanning mode and vice versa which is always done in the case of heat capacity
measurements (cf. Sect. 6.1). 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 transitions occurring in the sample - this offers the possibility to determine pure sample 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.
139
heating rate
-e.
..2:!.
c
'~
.....c
{3
----~----~--O>
~/
/~
,...:::cu
-e.
j
.........
(cpm- CPtru.)
CPtr"
= tl(3)
~/
//
~/
~/
. . .. . \
:::::::::.. ::::::::::::::::::::::.::::>Z<:::::.;:
:.
.
~~~~~...
i
/'
t
"
OJ
.<=
(
:
(T5 -Tm)~t({3)
..
0>
~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 temperature, 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 measured apparent heat capacity (calculated from the measured heat flow rate) and
the true heat capacity of the sample. As the sample temperature is lower than the
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 necessary, 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 functions 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
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 capacitytemperature function of the sample, Cp (Ts). An integration must be carried out
to obtain the thermodynamic potential function change:
T2
t:.H(T) =
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).
Cp , firstly Cv ( Ts) is obtained and from this, by integration, the (internal) energy
The temperature-modulated mode of operation (cf. Sect. 2.4.2) offers the possibility 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:
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 periodic part of the measured heat flow rate:
_
IP(T, t)
alPexo(Tu,t)
a T TA . sin (wt)
(5.9)
This is in reality a periodical (sinusoidal) fluctuating function with an amplitude IPA which on the one hand is influenced by the heat capacity of the sample
(lst term on the right side) and on the other hand by the reaction or transition,
if the respective latent heat flow rate depends on temperature (2nd term on the
right side). The latter causes even a phase shift 6. The periodic part of the heat
flow rate function forms the basis for further evaluations of TMDSC. By Fourier
analysis, which normally is included in the software of TMDSCs, the amplitude
and phase angle of the modulated part of the heat flow rate function can be determined. Further evaluation results in the determination of an "apparent heat
capacity", which in the case of processes occuring in the sample can be different
from the common (vibrational) heat capacity. Depending on the evaluation (and
mathematical) method several heat capacity curves, containing different information of the sample, can be obtained.
5.5.1
Reversing and Non-Reversing Heat Capacity
These curves are the result of the mostly used evaluation procedure. The basis is
the following: if there are no transitions or reactions in the sample, the amplitude 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:
cp,revo (T)
IPA (T)
= m TAW
(5.10)
This function (curve) is traditionally called the reversing heat capacity. Outside
the region of transitions or reactions it corresponds to the normal (vibrational)
heat capacity of the sample, within the region of processes it contains components of the process in question (see Eq. 5.9).
142
From the reversing heat capacity the reversing heat flow rate can be calculated:
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 contributions 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 capacity curves if needed:
IPnon-rev. (T)
= Cp,ll (T) -
Cp,rev. (T)
(5.12)
This equation is formally right, but it must be emphasized again, that these quantities are apparent heat capacities which are not really defined in a thermodynamic 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 underlying (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 capacity, or, for reactions or transitions, the magnitude (absolute value) of an apparent complex heat capacity. To obtain the complex heat capacity we have to pursue
a different path.
.
143
f:
2
1
""
-1
-2
-3
"C;
~
V)
60
100
140
180
temperature
220
O(
260
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 successful, 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 determined [see Eqs. (3.15) and (5.1O)]:
Icp(T) I =
4iA (T}
mTAw
(5.13)
This function is identical to the reversing heat capacity (see Sect. 5.5.1), it contains possible contributions from reactions or transitions (see Sect 3.3.2). Together with the properly corrected phase signal 8 (see Sect. 4.7.2), the apparent
complex heat capacity can be defined:
(5.14)
144
From this follows the real part and the imaginary part of the complex heat capacity (the phase shift is assumed to be negative!):
Re(cp) =
p=
and
Im(cp) = cp =
(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, because 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 dependent 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 dependent processes, after an inverse Fourier transform one gets the time dependent heat capacity function at the temperature in question. However, this elegant
method is not available with the commercial evaluation software and therefore
6
}/lg:1
>-
-=
'"
8
w
Cl..
.....
'"
OJ
.r;
w
~
OJ
Cl..
Vl
-2
50
100
150
temperature
250
200
300
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)
145
not accessible to the normal user and we refrain from presenting the details of
this special evaluation here. The advanced reader is referred to the original
literature (Merzlyakov, Schick, 2001 b).
5.5.3
Curves from Step-Scan Evaluation
The recently, for power-compensated DSCs, introduced step-scan mode of operation (see Sect. 2.4.2; Cassel, 2000) orginates originally two curves called the specific heat capacity and the isokinetic baseline. The former is simply the interpolated curve through the discrete heat capacity values got from the areas A of the
peaks caused by the discrete temperature steps t:..T (see Sect. 6.1.4):
A
cp = - - mt:..T
apparent heat capacity like the reversing Cp Using the extrapolated isokinetic
line (see below) as baseline makes a subtraction of a zeroline normally unnecessary. A careful peak area calibration is, of course, needed anyway.
The isokinetic baseline, on the other hand, is the interpolated line through the
end point of the isotherms after each temperature step. This equals the (isothermal) zeroline of the DSC (caused by asymmetry) but contains all the reaction
and transition heat flow rates produced by the sample even in isothermal mode.
Within the region of processes, the shape of this contours curve is highly influenced by the period of time chosen for the equilibration after the temperature
step. The longer the isothermal period, the lower is the end value of the heat flow
rate caused by reactions or transitions which usually follow a time law. This
curve must be evaluated with due care and attention to get reliable results concerning the reaction or transition heat.
The advanced step-response analysis presented in Sect. 5.5.2 can be carried
out as well. As a result we get a set of complex heat capacities cp (w), one for
every temperature step, which contains all necessary information about time
dependent processes occurring in the sample during the run. An indisputable
evaluation and interpretation is, however, not an easy task and the investigations
on this topic are still not completed.
5.6
Interpretation and Presentation of Results
For certain tasks (e. g., quality control of goods and materials received) comparison 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 interpretation 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
The output signal from a DSC, the heat flow rate as a function of temperature,
and any derived quantity, such as the heat of transformation or reaction or
any change of the heat capacity of the sample, may be used to solve many
different problems. The work required to evaluate the measured curve may
differ greatly from one case to another. This will become clear from the following text. Sometimes the required information can be obtained from only a qualitative evaluation of the DSC curve. But most of the examples described in this
section demand precise measurements and critical, very often special, evaluation procedures of the measured curve. In every case the basis of reliable results
is a careful calibration of the DSC (see Chapter 4). As a rule the separately measured 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 available for pressures up to the order of a hundred bar. In addition to the condensed
phases, the heat capacity of which is required, sealed crucibles always contain
a gaseous phase. In this case it makes no difference whether this phase is composed of air or of gaseous reaction products. Strictly speaking neither Cp nor Cv
are obtained because the thermal expansion of the sample cannot be prevented
and the pressure of the gas changes. However, the pressure dependence of the
heat capacity of condensed phases is very small and as the change of pressure in
the sealed crucibles is generally small, the measured heat capacity is nearly the
same as that at normal pressure.
148
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 determination (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 quickly from the measured differential heat flow rate d 4>SR' Several variants of the
experimental procedure are known, four widely used techniques will be discussed.
6.1.1
uClassicaluThree-Step Procedure
The procedure is illustrated in Fig. 6.1. The temperature-time curve during an
experiment is outlined in the lower figure and the response of the calorimeter is
shown above. The three steps are:
1. Determination of the heat flow rate of the zeroline 4>0(1'), using empty crucibles (of equal weight) in the sample and the reference sides. The temperature program should only be started when the isothermal heat flow rate at the
starting temperature Tst has been constant for at least one minute. If the DSC
is computer controlled, this can easily be automated by checking the differences between the current average value of the heat flow rate and that one
minute before with allowance for a predetermined noise level. The scanning
region between Tst and Tend can be 50 to 150 K in modern calorimeters. At the
isothermal end temperature Tend the above computer check must be repeated. For the evaluation procedure all three regions of the curve are needed. The
zero line reflects the (inevitable) asymmetry of the DSC.
2. A calibration substance (Ref) of known heat capacity CRef is placed into the
sample crucible (or into a crucible of same type and mass), whereas nothing
is changed on the reference side. Using the same experimental procedure as
for the zeroline, the following is valid:
149
12
sample
iso-
(ribralion subslance
Iher~
mal
al
isolhermal
01 Tend
Tst
scanning mode
.e-'"
'"
.e- Ie;
:a
...co
.c
.~
.e-
emply pan
:1
0
0
1st
"12
min
16
..
lime
1
~
::J
...~
T,t
C1.
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
those of the second step. Many of the possible sources of error for DSC measurements 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 between temperature sensor and sample and from the limited thermal conductivity 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 precise, 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-
151
f
~
...
;:;::
C1J
oJ::
r--I
I
I
I
I
I
I
I
I
I
B
L
time
______ _
..
Fig.6.2. Idealized (dashed line) and real conditions (solid line) during a heat capacity measurement.
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 support, crucible and sample. As a rule the influence of the apparatus is predominant. 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 between 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 gradient in such samples, there is a marked reduction in the overall temperature
error after correction for thermal lag (Hanitzsch, 1991). As an example the
Curie temperatures 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
mass and heating (or cooling) rate the differences between true and measured 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 special 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 zeroline, 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 isothermal 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 differences 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:
cPcorr(t)
which should be used for the calculation of the heat capacity then.
153
12
w
expo curve
t
.2!
...
<Pup ( TJ
.......................... ............................
.
corr. curve
<Peorr(
TJ
3:
0
;;::
'" 2
.c-
-2
~
t,t
3
time
...
4'exp
min
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:
Cs =
f3. ms
This method obviously saves time compared with the procedure in Sect. 6.1.1.
A disadvantage is that conditions of thermal symmetry between sample and
reference sample are lost and with them the partial compensation of related
errors, and that the conditions of the calibration may have changed for the measurement (and even the calibration factor) which increases the uncertainty of
the results.
The dual step method is, however, always a good technique for the investigation of samples with high thermal conductivities. The comments of Sect. 6.1.1
regarding errors and their corrections remain valid also for this simpler procedure.
154
An interesting possibility to get better symmetry is to attempt to set up experimental conditions for the sample run which are very similar to those for the
zeroline run. For this purpose the sample is put into the correct crucible Sand
a reference substance CRef is placed in the crucible R on the reference side. The
(differential) signal L1 tP will approach that of the zero line tPo when Cs ::::: CRef The
sample specific heat capacity is then given by:
mRef
Cs
= - - CRef +
ms
f3.
ms
155
156
mW
t
cu
<:>
....
01
<:>
cu
.z::.
10
11
min
13
..
time
12
t:..T--_
r:
Q1
f--:
6
time
10
11
mm
13
..
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
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 procedure 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:
(6.1)
It is very important that the isothermal periods after a temperature step are
long enough to attain steady-state conditions. These times differ according to the
type of calorimeter and should be determined experimentally. The older powercompensated Perkin-Elmer DSC's (DSC 2, DSC 7), for example, need 1 to 3 minutes, longer times are needed for samples that are large and/or have poor thermal conductivity.
By analogy with Sect. 6.1.2, the empty reference crucible may be replaced by
a crucible containing a calibration sample. The "11" in the symbol 11 (Qj) of the
following equation is used to label this. If all crucibles, both on the sample and
reference side, are assumed to have the same mass, we obtain:
(cs(Tj =
KQ(Tj) 11 (Q;)
I1T ms
mRef' CRef
+--ms
158
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=cpm=--
TAw
with CPA the heat flow rate amplitude, TA the temperature modulation amplitude,
w = 2n/tp the angular frequency (tp the period), and m the mass of the sample.
If the TMDSC is carefully calibrated (see Sect. 4.7) the method offers the possibility 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 presented so far fail.
Outside of the temperature region of transitions or reactions the TMDSC
method results in the static (vibrational) heat capacity. The temperature dependent 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 simplifies the calibration procedure markedly. The same is true for the temperature amplitude, 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 correspondence 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 components 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 deconvolution needs special experiences and is still a matter of scientific discussion and
normally outside the scope of common heat capacity measurements. However, for
160
Fig. 6.5. TMDSC heat capacity measurement for polystyrene (PS); solid line:
reversing cp , dashed line:
underlying cp , symbols: literature values for the solid and
liquid state (ATHAS, 2002)
(m:4mg,po: 2 Kmin-I,
tp: 60 s, TA : 0.5 K)
3
JK- 1 g-1
[p
______L -_ _ _ __
100 O( 150
50
OL-----~----~
-50
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 possibility to get the static vibrational heat capacity via extrapolation of the measured
apparent heat capacity to infinite frequency where time dependent processes
don't contribute anymore (Schick et al., 2003). To extend the normal frequency
range of a common TMDSC, the evaluation of higher harmonics is essential, this
requires a non-sinusoidal temperature modulation (see Sect. 5.5).
6.1.6
Typical Applications of Heat Capacity Measurements
The heat capacity is a basic quantity for determination of other thermodynamic
quantities as well as material properties. Typical applications are listed in the
following:
- Very important from the theoretical point of view is, on the one hand, the
comparison of theoretical and measured heat capacities and, on the other, the
calculation of the thermodynamic potential functions S(T), H(T), F(T) and
G (T). For polymers numerous, reliable experimental heat capacity results are
available and these have been collected in a data bank (Wunderlich et al.1990,
ATHAS, 2002). The majority of these are DSC measurements and generally refer to relatively high temperatures (> 200 K). Values at lower temperatures (as
close as possible to 0 K) must be measured using high-performance adiabatic
calorimeters. The main microscopic cause of the measured, macroscopic
heat capacity of solids is the vibrational motion. It is thus possible to calculate the heat capacity of solid polymers if the vibrational spectra are known.
According to the procedure of Wunderlich and co-workers (Lau et aI., 1984;
Wunderlich et al. 1990; Roles, Wunderlich 1993) the spectra may be approximately separated into group and skeletal contributions. The former are obtained from infrared and Raman spectra. The latter can be approximated by
Tarasov functions, the characteristic parameters for which are obtained by
fitting the experimental heat capacities at low temperatures, when the group
161
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 = - - -
Cp - Cv =
3R Ao' Ci- T
0
Cv ' Tfus
nus
equilibrium fusion temperature; Ao empirical constant (for many polymers: Ao = 3.9 10-3 K mol rl; R universal gas constant).
When the heat capacity functions Cp(T) are known from as near as possible to a K on, the thermodynamic functions H (T), 5 (T), F (T) and G (T) can be
calculated as normal:
T
H(T)
= Ho + f
Cp(T) dT
C (T)
5 (T)
= 50 + f
G(T)
= H(T) -
-p-
dT
T 5(T)
Low temperature DSC measurements of Cp are not possible and for this technique 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
Tl
(Vi'
Cm,i(T)) dT
162
6.2
Determination of Heats of Reaction
The aim is to determine a thermodynamically well defined (temperature dependent) reaction enthalpy. If a subsequent kinetic analysis is planned, the calculation of a similarly well defined conversion-time curve is of equal importance. The interpretation assumes that both the initial and the final state of the
reaction are unambiguous and well-defined. Is this not the case, the results will
be, at best, semi-quantitative. Because of the limited precision of DSCs minor
effects due, for example, to changes in pressure, stress or surface contributions
may be neglected. The following, simplified equation is then obtained for the
measured heat flow rates in a DSC:
Q
(d )
dt
. d~
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 disadvantages of the two methods are discussed in detail in Sect 6.3.
Reactions in the Isothermal Mode
Here, the sample is heated as rapidly as possible from a temperature at which the
system is inert (often room temperature) to the reaction temperature. The "drop
in" technique (cf. Appendix 2) may also be used - the sample is dropped into the
preheated calorimeter. The end of the reaction is indicated by a constant heat
flow rate <Pend' Simple extrapolation of this <Pend to the start time yields the baseline of the reaction peak. As there is no heat capacity contribution to the heat
flow rate in the isothermal mode, the relation between the heat produced or consumed and the average reaction enthalpy is very simple:
(;
Qm(t)
=f
(<Pm -
<Pend)
dt = (flrH) f d~
0
The only uncertainty is due to possible drift in the signal. The caloric error for a
typical sample mass of 10 mg and a heat production of 360 J g-l is less than 1%,
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 obtained. 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-defined 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 isothermal heat flow rate curve of a second run with the same (now totally reacted)
sample and exactly the same conditions. Nevertheless, minor differences may remain 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
2.5
J
K-g
10
>~
8"
1.S
'"
.c
~
'"
1.0
300
350
400
tempera ture
450
500
..
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 reactants may be determined from this interval. Extrapolation into the reaction temperature range is generally accurate enough. Otherwise, the heat capacities of all
components of the reaction mixture must be measured separately and the Cp
function calculated using simple mixing rules.
After subtraction 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 temperature range between T J and T2 Improvements relative to method 1 are only
found if the heat capacity functions are known with sufficient precision (uncertainties < 1 %). Obvious improvements in the definition of the baseline should
result from the use of 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 between partial heats (up to a certain temperature) and the overall heat of the 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 temperature range of the reaction. From Fig. 6.7 it follows that
The reaction enthalpy at temperature TI (Arh (TI is obtained as difference between the areas X and Y. The baseline is not necessary in this case, because only
differences are needed:
d r h (TI ) = X - Y = {hI (P, 490 K) - hI (R, 350 K - (hI (P, 490 K) - hI (P, 350 K
The enthalpy subscript l represents the liquid state of the reactants Rand
the liquid-like or rubbery state of the products P. The difference hI (P, 490 K) hI (R, 350 K) corresponds to the experimental quantity, the area X, defined by
the lines at Tl> T2 , cp = 0 and the reaction curve. The horizontal hatched area Y
follows from the rerun on the reacted sample.
490K
=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:
or:
= h (P, T) -
h (R, T)
TJ
This procedure is schematically shown in Fig. 6.8. The areas X and Z are hatched
horizontally and vertically respectively. A simple rearrangement of the last
equation gives:
T2
TJ
T
166
Fig. 6.7 a, b. The thermodynamically correct determination of the specific heat
of reaction (according to
Richardson, 1989); for details
see text. a 1st run curve,
b 1st and 2nd run curves
Hatched areas extended
to cp = 0; ~ h (350 K) =
X - Y = hproducts (350 K) hreactants (350 K) = - 44.7 J g-l
2.5
J/IKgl
2.0
>-
'"c. 1.5
'"
~
'"
C1I
..r:::.
C1I
c.
VI
10
300
350
400
tempera ture
..
450
500
2.5
J/IKgl
2.0
>-
'"c. 15
'"
~
'"
OJ
..r:::.
C1I
c.
VI
10
300
350
400
tempera ture
..
450
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 instructive. Here, for the sake of simplicity, temperature independent heat capacities are assumed. The enthalpies are then linear functions of the temperature.
The diagram also recognizes that both reactants and products may be in the
glassy [curves Hg (R) and Hg (P)] or liquid [curves HI (R) and HI (P)] state. A reaction usually proceeds at a measurable rate only when the reactants are in the
liquid state [above glass transition Tg(R)]. Further, the glass transition of the
167
~rHITI
= X-Z
'p (R)
'p(PI
---
~r"' I--
!\
-...'"
-.......
.c
"
a..
II>
r,
T
..
temperature
Fig. 6.9. The determination
of reaction enthalpies from
an enthalpy-temperature
diagram (according to Flynn
1993); A: integral of Cp(n
for the reactants, B: integral
of Cp(n for the reaction
mixture, C: integral of Cp(n
for the products
1
>-
C1.
CI
.c
CQ.I
I __ -
...lI
Tg (R)
_-K[
I
I
I
II
I
I
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
glassy reactants
liquid reactants
liquid reactants
-7
-7
-7
-7
glassy products
liquid (rubbery) products
glassy products
liquid (rubbery) products
ArH =bd,
ArH = be,
ArH = ad,
ArH = ae,
Only the final case has any meaning for kinetic studies.
=bc/bd
=be/be
= ac/ad
~
~
~=
ac/ae
168
Kinetic Investigations
6.3.1
cC + dD + ...
According to the definition (ni = ni,O + Vi 5),5 can be related to the consumption of an educt (Vi < 0) or to the formation of a product (Vi> 0). The rate of reaction is then quite generally given by:
1 d5
r=--
V dt
169
or quite general
r= f(n f(e}
V dt
eAA
eBB
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 analytical 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 reaction a = S/Smax> in particular if reactions in heterogeneous systems are investigated.As a rule, a is set as the ratio of partial area and total area of the DSC peak:
Qt
a(t} = Q~
and
da If>(t}
---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 equilibrium state. All subsequent kinetic analyses are incorrect if this modification is
omitted. This is because the calculated a(t} is not related to the thermodynamic degree of reaction but to an apparent final state that is specific to the individual experiment.
A general rate law is written as
da
- = k(n f(a}
dt
170
The functionf(a) is very simple for the smallest steps (elementary reactions) of
reactions in homogeneous systems. In the case of equimolar mixtures one obtains then:
da
The formal reaction orders n, m and p are only parameters of variation, to obtain a better fit to the experimental curves, an interpretation in the physicochemical meaning is difficult or impossible.
The situation is much more complicated for reactions in heterogeneous systems. During a chemical reaction of this type the reactants may always exist
in different phases (liquid-solid, solid-solid) or the system may change due to
the development of new phases and/or the vitrification of polymerizing components. 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 considerable importance for describing the kinetics of both decomposition reactions 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.
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 judgement 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:
= A exp (- ;~ )
Arrhenius:
k(T}
Eyring:
k(T} = -hexp
kT
B
(~S")
R
.exp (AH"
- RT )
Ah
AS"=Rln--e k B T
The Eyring activation parameters are more suitable for the understanding of relations between the structure of the reactants and the reactivity (Heublein et aI.,
1984).
Although open to criticism (Flynn, 1990), as a rule the traditional concept of
activation is also used for reactions in heterogeneous systems. The main theoretical difficulty results from the fact that the energy distribution of the immobilized constituents of a solid is not represented by the Maxwell-Boltzmann
equation, respectively, that we find a number of surface sites with different
energies and energy distributions. For instance, Hunger et al., 1999, have determined the activation energy distributions for the desorption of water from different zeolites and they have definitely proved the existence of several different
adsorption places. But meanwhile two reasonable approaches supply also for
this type of reaction a theoretical foundation for the exponential k- T relationship. Galwey and Brown, 2002, start from the band structure of a solid and consider discrete interface energy levels in the locally modified environment
between the band structures of reactant and product phases. The occupancy of
the interface levels is directed by energy distribution functions based on FermiDirac statistics for electrons and Bose-Einstein statistics for phonons. At higher
energy differences between the interface levels and the highest occupied electron bands (compared with kB T) both distributions approximate to the expo-
172
tivation parameters,
2. determination of the reaction mechanism,
3. prediction of the reaction behavior for any complex time-temperature profiles.
First of all, kinetic analysis is of course an efficient tool for data reduction. A
large body of experimental data from a series of measurements with many data
points has then been reduced to a model with few parameters. Further, reactions
may be judged using different criteria. For example, the influence of variations
of the activation parameters can be discussed together with changes in the composition of the reacting system (stoichiometry, catalysts, solvents and fillers) or
with modifications of one reactant (number and position of substituents,
homologous series).
But the true aim of a kinetic analysis is an evaluation of the correct reaction
mechanism: this is fully defined, if the sequence of all elementary steps is known
as are the activation parameters for each of these. The inseparable combination
of rate law, activation energy and preexponential factor is sometimes named the
"kinetic triplet" (Maciejewski, 2000). Frequently, the form of the overall rate law
already permits conclusions according to a possible mechanism. The rate law
can be simple or complicated. A complicated rate law always signals a complex
mechanism with a number of steps. On the other hand, the opposite conclusion
is not correct. A simple rate law can also be caused by the existence of a very slow
reaction step - the rate-determining step - within a number of much faster steps.
The investigation of the reaction mechanism is an extremely demanding task.
Much time and effort is required to reach it and supplementary investigations
using other analytical techniques are essential. Even then, the result is often only
a most probable mechanism or the knowledge of the most important steps.
A third aspect of kinetic investigations becomes more and more important.
The practitioner in industry needs a useful and quickly accessible reaction model, which on the one hand adequately describes the course of the reaction and on
the other allows reliable predictions for any temperature-time conditions. The
form of this reaction model (overall rate law, combination of formal steps in
173
174
175
- 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 considered, irrespective of whether subsequent calculations are made as functions of concentration or degree of reaction.
In much of the literature it is unfortunately not clear whether the above aspects
have been considered and the corresponding results must be treated with reservation.
6.3.3
Selection of the Measuring Conditions - Isothermal or Non-Isothermal
Reaction Mode
176
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 because 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 probable 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 degrees of reaction (Fig. 6.10). This can only be due to autocatalytic and/or consecutive 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-trimethylhexamethylene 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
...
-'"
-'"
:3:
0
1.50
ra tio
2:
3:
4:
5:
.r:::.
.~
' v 0.75
C1.
II>
0.00
OGEBA to aniline
1: 1:1
0f
17
34
1:2
1:3
1:4
2:1
68
51
time
min
85
Fig.6.10. Isothermal reaction curves for the polyaddition of aniline and bisphenol-A diglycidyl ether (DGEBA) in mixtures with different stoichiometry
increase in the slope of the leading edge of the reaction peak in direct proportion to the concentration of catalyst is unusual. This can only be explained
if one assumes a direct coupling of catalysis and autocatalysis.
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 longtime 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 decomposition 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
lime
100
------
lime
10
------
min 30
0
Wg-'
-005
Wg-'
-0005
SO '(
30 '(
-0.10
-0010
-0.15
Wg-'
-05
Wg-'
-10
70 '(
-10
-15
-20
2 min
50
os
K min
0
Wg
-,
- 0.2
-10
10 K min
C 80
65
temperature ------
lime ------
10
15 min
time -----
-50
-1
temperature
- 04
50
C 100
-----
Fig.6.12. Calculated DSC curves for isothermal (left) and scanning (right) mode of operation. 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).
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, Flammersheim, 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 reaction 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 reaction (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 measurements can be made. At temperatures higher than 25 DC the heat flow rate is
strongly disturbed and falsified at the beginning, while at temperatures lower 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 mentioned 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 mechanism.
- The screening of the complete global reaction field, by carrying out nonisothermal 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
6.3.4
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 complicated procedures with intermediate annealing phases) is responsible for the extensive 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 applications 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 intensity 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
2nd run
mW
150
:.
c
cu
..c
15
o ~------~--------~--------~--------~--------~
0.7
2.1
1.4
time
2.8
min
3.5
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)
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 differ 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 compensated 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 approximatively! 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
I
I
250
,,,
,
~
0
;:;::
c
QJ
..c
125
"
"
"
"
"
10
0.5
time
15
min
2.0
Fig.6.14. Baseline corrected experimental (solid line) and desmeared (dashed line) curves
for the photoinduced reaction of Fig. 6.13
182
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 corresponds 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 partially 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 reaction rates within the sample. However, the calorimeter always records an overall heat flow rate. As result of this rate gradient there is a corresponding concentration gradient of the reactive component. An exact solution of the subsequent system of differential equations is only possible for drastically simplified boundary and initial conditions (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 intensity 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 temperature. 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 insufficiently 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 lightactivated 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.
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 parameters which adequately describes the overall reaction within a range of conversion 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 carefully. An overall rate law may give a good description over a certain range of reaction, because the rate reacts very sensitively to those parameters which are present
in the rate law. It may be insensitive with respect to those parameters which do not
appear explicitly (e. g., stoichiometry of the reaction, solvents, catalysts, packing efficiency for 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 simplifications that may sometimes be made to the rather complicated differential equations that describe the true reaction mechanism. In spite of this, a determination of
the 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 conversion 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
dt
= kJ . C (1 - a)n
+ k2 . am . (1 - a)n
the 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-
C is
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
=(1- a)n
In dt
Co
Today multiple non-linear regression software from different calorimeter manufacturers allows to estimate the best values for n, InA and EA by evaluation of
a single data set from one scanning experiment. The risk of misinterpretations
is minimized if several data sets with different measuring conditions are
simultaneously evaluated. Originally developed for n-th order reactions, at present various choices are offered for f(a) (simple rate laws, rate laws including
autocatalytic steps, typical rate laws for the kinetics of reactions in heterogeneous systems). The evaluation presupposes isokinetic behavior (no changes of the
mechanism within the examined temperature range). Traditionally this was
proved using somehow created "reduced curves" for isothermal scans and
subsequent comparison with master curves. For instance, all experimental time
values for certain conversion degrees can be scaled by the reduced time factor,
tred = t/ ta (ref) , where ta (ref) is the time for a reference conversion degree, frequent ly a = 0.5. If we have isokinetic behavior, all experimental plots of a against tred
should fallon one single curve. The comparison with theoretical master curves
for all known rate laws enables conclusions concerning the probable rate law. But
even in the case of single step reactions the differentiation between various reaction models is sometimes difficult on the basis of a graphical comparison, especially using erroneous real measurements. Therefore, using the present possibilities of computerization and evaluation methods, the determination of the most
probable rate law should be supported by statistical methods. The F-test yields
the information which reaction model has the highest fit quality and whether the
best reaction model is significantly better than other models. Modern variations
of the Borchardt-Daniels procedure also allow the evaluation of nonisothermal
measurements and avoid therefore the main disadvantage of isothermal runs,
the not exactly known initial phase of the reaction (see Sect. 6.3.3).
185
b. If the known rate law is used in its integrated form (the integral method)
a
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
(EA ) dT
f - d a =g(a) =- f exp - -
f3
o f(a)
RT
there are difficulties because an exponential integral cannot be solved analytically. A solution is to use integral tables as calculated, for example, by Doyle,
1961. A review of numerous varieties using integral methods can be found in
the literature (e.g., Hemminger, Cammenga, 1989; Galwey, Brown, 1998). The
importance of integral procedures has decreased considerably with developments in computer technology.
c. Likewise, many varieties of differential methods are known. For instance,
Freeman, Carroll, 1958, used the logarithmic equation for n-th order reactions in a form based on differences between neighboring data points
(da)
~ In dT
= n . ~ In (1 - a) -
EA ~ (T1 )
R'
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
Again, the plot In fJi versus lITi for constant a/s yields parallel lines for a singlestep 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 approximations 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 mechanism 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 presupposes the independence of the modelf(a) from the degree of reaction a but
rather the opposite, a conversion-dependent overall activation energy. He calculates for a set of n experiments carried out at different heating rates the minimum of the function
i i
i=1
I(EA,a, Ti(ta))
jn I(EA,a, Tj(ta))
with
I(EA, Ti(ta)) ==
f exp (-~)
dt
RTi(t)
187
188
189
190
100
Wig
2.25
1 K min- 1
2 K min- 1
<11
4 K min- 1
0
~
1.50
8 K min- 1
~
0
15 K min- 1
<11
.c;
....
....
<11
0..
11"1
0.75
71
50
134
113
92
O(
155
temperature
1.00
Wig
'.75
..::!.
c
~ 0.50
370 K
'"
.c::
'"
0.25
0-
Il>
90
60
30
time
min
120
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
191
small. An approximate (because of different thermal conductivities) but adequate desmearing is possible using the crystallization peak of molten trans-azobenzene as the apparatus function. The correction is small because the isomerization is rate-determining even at the highest used heating rates. A precise temperature correction to zero heating rate values for all heating runs is very important.
Figure 6.15 shows the excellent agreement between experiments and calculations using multiple-curve analysis by non-linear regression. All curves from runs
at five heating rates and at four isothermal temperatures have been included in the
calculations (that for the lowest temperature is not shown in the figure). If minor
self-heating of the sample because of the limited heat exchange with the calorimeter is allowed for, no systematic differences are found between the single-curve
and the multi-curve analysis. In addition, the same activation parameters are
found even at higher heating rates (40 K min-I). Assuming a fixed reaction order
of n =1 the procedure gives activation parameters of In (Als- I ) = 27.6 0.2 and
EA = (l02.5 0.8) kJ mol-I, in good agreement with Eligehausen et al., 1989.
Another example is the dimerization of cyclopentadiene (Flammersheim,
Opfermann, 1999b). The monomer (CPD) is easy to obtain from the commercial
dimer (DCPD) by cracking distillation at 150-170 DC and repeated distillation at
40 DC in an argon stream. The mechanism according to the following scheme is
slightly more complicated (Turnbull, Hull, 1968):
2CPD ~ DCPD
(l)
(2)
DCPD ~ 2CPD
(3)
The second order Diels-Alder dimerization is auto catalyzed and the back reaction cannot be neglected above 140 DC. But this reaction should be suitable to test
the efficiency of the used kinetic software. Using the program "Component
kinetics" and taking into account the volume contraction during the reaction
(~CPD = 0.7970 g cm- 3, (JDCPD = 0.9770 g cm- 3 ), the best activation parameters are
shown in the Table 6.l.
Table 6.1
Reaction
Parameter
Optimized value
1: 2 CPD ~ DCPD
19 (Ails-II mol-I)
EA.I/(kJ mol-I)
19 (A 2/s- 1 L2 mol- 2)
EA 2 /(kJ mol-I)
5.74 0.15
69.43 1.00
S.SO 0.30
73.79 2.40
Ig (A3/s-l)
EA 3 /(kJ mol-I)
13.04 0.25
142.0 2.2
correlation coefficient
0.9997
2: 2 CPD + DCPD
3: DCPD
2 CPD
2 DCPD
192
Fig.6.I6. Concentration
profile for the dimerization of CPD (heating
rate: 10 K min-I). 1: CPD,
2 and 3: DCPD formed
according to the simple
2nd order reaction (1) and
the autocatalytic reaction (2)
of the reaction scheme
(see text)
12
mol l-1
O[
120
80
<-
-'C
40
<lJ
c
0
0
0
8
12
time - - -
16 min 20
Using these results, the heat flow rates, conversion degrees and concentrations of the reactants can be predicted for any temperature-time reaction profiles. As an example, Fig. 6.16 shows the concentration profile for a heating rate
of 10 K min-I. The program allows to differentiate between CPD formed during
the two reaction paths (1) and (2). With increasing temperature, in the case of
f3 = 10 K min- I above 185C, dominates the back reaction (3) for the overall
reaction. But the two forward reactions show a different behavior. While the
amount of CPD formed as a result of reaction (1) already decreases above 160C,
the percentage of the autocatalytic reaction (2) is still increasing. Another result
is important. Every nth-order reaction starts with maximal rate. Using the parameters of the table we can calculate for instance the storage periods for which
a predefined degree of conversion is not exceeded. In our example and for
a =0.02 these times are nearly 4 d at - 20C, 9.5 h at OC and only 72.3 min at
20C. That means, the dimerization of CPD should always be investigated with
freshly distilled monomer immediately after its preparation.
2. Example: Radical Polymerization
of (2,2-Dimethyl-l,3-dioxolan-4-y/)methyl Methacrylate (DOMA)
The radical polymerization of a monofunctional monomer is a simple 1st order
reaction, if a number of ideal conditions can be assumed (reactivity of the
macro-radicals is independent of the chain length, approximately constant radical concentration during the whole reaction, bimolecular chain termination
reactions). Under these circumstances, the rate of concentration change is proportional to the concentration of the monomer (CM) as follows:
dCM
---=kpro
dt
(r.
kini)"2
-kter
1/2
CI
'CM
The quantity r is the yield of radicals due to the thermal homolysis of the initiator (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
193
and initiation respectively. If the thermal initiator is replaced by a photoinitiator, like benzoin methyl ether or 2,2' -dimethoxy-2-phenylacetophenone (benzyl
dimethyl ketal), the analogous overall rate law reads:
dCM
( r labS) 112
---= kpro ' - 'CM
dt
kter
Here r is the quantum yield of the initiator, labs is the radiant power absorbed
by the sample. In contrast to polyfunctional acrylates, for which cross-linking
reaction starts immediately (Fig. 6.13), monofunctional methacrylates (methyl,
ethyl, propyl, butyl) follow the simple 1st order law up to degrees of reaction of
0.3 to 0.6 (dependent on the chemical nature during both thermal and lightinduced polymerization). After this stage the so-called gel or TrommsdorffNorrish 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-dioxolan4-yl)methyl methacrylate (DOMA) (Flammersheim,Klemm, 1985). There are no
qualitative differences between curves 1 and 2. In both cases the change of reaction rate with time is small enough during the entire reaction that desmearing
is not necessary. Spectroscopic 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 reaction 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 molecules, the chain termination reactions of the growing macroradicals are in-
6.0
mW
t
~
4.5
3.0
.....
~
....
0
1.5
cu
.J:
30
20
10
time
min
40
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 photoinitiator; curve 2: T = 343 K, 1% 2,2-azodiisobutyronitrile as initiator)
194
20
mW
15
5
dark reaction
illumination period
o~------~--------~------~------~--------~
1.2
0.6
1.8
time
min
..
Fig.6.18. Photopolymerization of DOMA with interruption of the irradiation at reaction degrees of 0.35 and 0.80 (arrows). Both curves are consecutive runs of the same sample, but the
second curve is shifted to the origin of the time scale.
Solid lines: measured curves, dashed lines: desmeared curves (A= 365 nm,Io = 3.5 mW cm- 2,
T = 298 K, 1% 2,2-dimethoxy-2-phenylacetophenone as photoinitiator)
creasingly hindered in the more and more viscous matrix. The result is a pronounced 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
dcM/dt
kpro
dCM/dt
---=-t----
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
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.
1.8
JK-' g-'
>..
-'-'
a 1.4
Clau
'"
.s:::
1.2
OJ
'-'
OJ
CIVl
50
100
150
temperature ----
O(
200
196
CL
<:>
co
14
-04 -;;
QJ
'"en
..c:
1.3
Cp
CL
1 2 '----------'---'---------'----'-----'-------' - 0 6
o
50
100
150 min 200
QJ
CL
en
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
kdiff
kchem
-=-+--
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)
a)
g!.
=Tg(O) . exp ( -
g2- a
g2 = 12.26.
= 296.3
K, g!
= 4.48
and
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:
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 influenced by the changing segmental length, van Assche, 1995, 1997,2001, during
the freezing of the reactive species.
The most simple formal-kinetic model that produces a practically perfect fit
for all heating rates (correlation coefficients> 0.9995) corresponds to a process
with two consecutive partial steps:
A~B~C
The corresponding rate law for the formal component B is:
dCB
dt
= 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 component C correspond to the negative first and second term of this equation respectively. Two data sets were available for the kinetic evaluation, each with 8 different 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 acceptable fit is reached. The model parameters for both data sets are summarized
in Table 6.2.
198
Table 6.2
Parameter
Data set 1
Data set 2
19 (A 2 /s- 1)
19 (A1/s- 1)
7.94 1.05
6.48 0.70
70.0 5.2
79.0 8.2
0.826 0.170
-4.77 1.05
9.9 3.1
25.9 21.8
0.25 0.06
0.9997
8.90 1.12
6.20 0.86
67.5 6.6
86.7 8.7
0.788 0.170
-4.39 0.78
8.1 2.1
28.6 19.4
0.21 0.06
0.9996
EA,l/(kJ mol-I)
EA ,2/(kJ mol-I)
19 Kcat,2
19 (Kdiffusion,2/mol-1 m 3 S-l)
CdK
C2 /K
Using these values, the course of the isothermal curing was correctly predicted. Further, at low temperatures, the conversion degrees never reach "1" after an
extended period due to partial freezing. The corresponding values are correctly
predicted for different temperatures (110 C: calculated 0.58, found 0.57 and
0.61; 140C: calculated 0.80, found 0.79 and 0.79; 170C: calculated 0.94, found
0.93 and 0.94). Finally, a convincing argument for the validity of the selected
model is given in Fig. 6.21 for an experiment with a very slow heating rate,
0.25 K min-I. The heat capacity function that is directly accessed by experiment
(cf. Fig. 6.19) runs completely parallel to the increase of the glass temperature
and the degree of reaction calculated by means of the model parameters. This is
valid both for the vitrification at about 100C and the slow devitrification starting at about 130C. This temperature range becomes larger at even lower heating 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 become more and more unimportant.
tJK-
18
200
10
1 g-l
O[
08
>-~
CL
14
100
/
Il
-;:;
OJ
OJ
CL
Vl
12
50
200
0.4
OJ
CL
OJ
a2
:=':'
"0
OJ
<lJ
~
en
OJ
----
400
lime
06
:::>
,,'
T'!,./i
:.....---- .
._cx./
OJ
t
c
600
min
800
'D
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
199
100
%
DC
'"<- 60
150
c:
!:!
OJ
>
c:
0
'-'
::>
40
<-
<U
""
"0
ec..
T9
OJ
OJ
<U
'-
20
100
OJ
OJ
50 -
0
time - - -
200
180
O(
19 (100%)
140
w
'-
100
:::J
o
....
E
w
60
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
201
3.0
B
hea ting run
mW
r
~
"
.....
1.5
"
'"
.c
cooling run
40
55
70
temperature
85
'C
______
115
..
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 cooling rate: 5 K min-I)
properties). Each of these could form the basis of an experimental determination of Tg With DSCs the glass transition is detectable by a step change of the
heat capacity /}.cp on heating or cooling. Both the temperature and magnitude of
this event are important. Observed values of /}.cp range from 0.1 to 2 J g-l K- 1
Figure 6.24 shows a typical DSC curve for a linear, high-molar epoxide-amine
polyadduct which was first cooled at the rate shown and then immediately reheated.
For polymeric glasses the temperatures of the transition range from - 50 to
300C whereas the corresponding range for silicate glasses is from 500 to
1000 C. Polymeric glasses are thus more suitable for DSC measurements than
inorganic glasses. For polymer chemists and materials scientists a knowledge of
the glass transition temperature is as least as important as is the melting temperature.
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, thermodynamically, far from equilibrium. The formation and behavior of a glass are
exclusively kinetic events. There are only formal similarities between the
Cp-change at an "ideal" glass transition and at a thermodynamically well defined second order transition. Only the liquid (or rubbery) state at the high tem-
202
Conclusions:
- In contrast to the measurement of equilibrium transitions, it is not possible
to get "equilibrium" values of the characteristic quantities Tg and /:"cp byextrapolation to zero heating or cooling rates: these quantities are determined
by the thermal history (scanning rates and annealing times). If, for instance,
the cooling rate is changed by an order of magnitude Tg will change by 3 to
20 K depending on the material in question. For flexible polymers the magnitude is generally 3 to 5 K, whereas it is 15 to 18 K for the considerably stiffer
borosilicate network glasses. By contrast /:"cp shows much less dependence on
thermal history.
- Characterization of glasses by DSC measurements (Fig. 6.24) is mostly carried out in the heating mode because this is practicable even for those heat
flux DSCs having relatively sluggish furnaces. A heating run is also advantageous when it is important to characterize the "as received" glassy state (e. g.,
resulting from particular cooling or annealing procedures or chemical reaction). 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 instance, if the heating rate is slower than the previous cooling rate). Tg then depends on the heating rate. To sum up, values for Tg (and to a lesser extent /:"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 heating) are avoided. Cooling runs, therefore, immediately reflect - possibly after
203
Fig. 6.25. Definition of the most frequently used conventional quantities for characterization
of the glass transition.
Tg, e: extrapolated onset temperature, Tg, 1/2: half-step temperature, !:J.Cp : Cp-change at the halfstep temperature, Tg,i and Tg,f: initial and final temperatures of the glass transition, Tg,f - Tg,i:
temperature interval of the glass transition
Tg,e and T g,1I2 can both easily be obtained from routine measurements, this is the
reason for the nearly exclusive use of these pragmatically defined glass transition temperatures up to now. A repeatability error of 1 K is acceptable in practice. For heating runs the following practical and frequently used procedure is
recommended:
205
mW
annealed glass
t
~
c>
.....
~
c>
""
o
60
Tg lqul
70
80
~.l\lqU'Z: fg .la'
temperature
g.lIla'
90
..
100
O(
110
Fig.6.26. DSC heating curves for an annealed (or slowly cooled) and a quenched glass showing 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: extrapolated onset temperature, Tg,lf2: half-step temperature
Their height may be comparable with those of the melting peaks of crystalline
materials (Petrie, 1972). For the example shown in Fig. 6.26 the glass annealed
70 h at 68e has Tg,e = 89.1 e, Tg,1/2 = 87.3e whereas the quenched glass has
Tg,e = 81.6e, Tg,1I2 = 85.1 e. This use of Tg,e and Tg,ll2 to characterize the glass
process gives the paradoxical result that a slowly cooled or annealed glass seems
to have a higher glass transition temperature than a rapidly cooled one. In addition Tg,e and Tg,1I2 react to the thermal history in a different manner.
Tg,e and Tg,1I2 are not useful for theoretical treatments of the kinetics of the
glass process, this is also true for certain relationships between Tg and other
properties (e. g., Tg as function of the molar mass or as a function of the degree
of conversion in a reacting system).
The Thermodynamically Defined Glass Transition Temperature
206
H9 (-P d_______
Hg(all~
>-
C1.
c;
:E
c:
OJ
H9 (-P21/
Hg(a21~
temperature
..
207
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 equilibrium enthalpy value of the liquid at that temperature, whereas this was not the
case at Ta2 Reheating of the annealed glass then proceeds along the enthalpy
lines Hg(al) or Hg(a2). From the theoretical point of view, the glass should devitrify exactly when the enthalpy line of the glass crosses that of the liquid but
the transition from Hg to HI is not sharp. To determine the intersection of enthalpy curves for the glassy and liquid states therefore requires the extrapolation
of these curves from temperatures, which are clearly above or below that of the
transition region. The point of intersection, obtained in this way, defines the
thermodynamic glass transition temperature Tg Tool, 1946 called this temperature 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 defined 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 relaxation peak. The reason for the superheating effect is the drastically decreased
mobility in the glassy state, which parallels the slow enthalpy decrease during
annealing.
If the enthalpy definition (Flynn, 1974; Richardson, Savill, 1975a; Richardson,
1976; Moynihan et al., 1976) is used, Tg can easily be calculated from DSC measurements.
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 integration of the corresponding cp-functions, which can always be approximated by
linear equations:
1
hg(T) = a T + - b T2 + P
glass:
and
Cp,g= a + b T
2
liquid:
Cp,l
=A +B T
hl(T) = A . T + - B . T2
2
and
+Q
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 accessible experimental quantity:
1
+ (Q-P)
208
4.5
l/(Kgl
3.5
3.0
>-
'u
CI
0CI
2.5
C P.l
-------------
-;;
cu
2.0
..c:
cu
'" 1.5
0-
1.0
90
60
temperature
100
T2
'C
11
..
TJ and T2 are convenient arbitrary temperatures that must be chosen to be below and above the glass transition region (i.e., in the glassy and in the liquid
state, respectively). Using Q - P from the above equation the desired glass temperature is obtained by solving the quadratic equation:
1
- (B - b) . T~ + (A - a) Tg + (Q - P) = 0
2
Figure 6.29 shows an equivalent, graphical procedure (Moynihan et aI., 1976) for
determining Tg From the enthalpy definition it follows that:
T2
Tg
T2
(Cp,I(T) - cp,g(T dT
= f (cp(T) TJ
cp,g(T dT
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 respectively, 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
209
4.5
J/!Kgl
3.5
3.0
>-
0..
2.5
Cp.1
------
'"
..c
~~ =---=- =--=--
2.0
'u
'"
0..
II>
cpg
1.5
1.0
60
T,
90
70
Iempera lure
100 Tz
11
..
Fig.6.29. Construction to determine the thermodynamic glass temperature Tg , which is defined 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 (extrapolated) 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 determined to better than 0.5 to 1 K with this method during routine measurements. 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
rapidly and "fictive temperatures" (Tg,fic == Tg) can be calculated for most instruments. This is fortunate because for theoretical investigations of the kinetics
of the glass processes the situation is clear: only the thermodynamically defined
glass 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 investigations 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 precision.
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 experimental manipulations even for (often time consuming) annealing experiments.
Annealing the sample outside the apparatus almost always yields unsatisfactory results because heat transfer conditions are not exactly reproducible after
replacing the sample in the DSC.
The values for the glass transition temperatures obtained from DSC measurements need not necessarily agree with those of other methods. Discrepancies
are caused by the different influences of the particular technique on the relaxation of the 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 variables.
In studies of this kind it is indeed important to ensure that experimental
errors have really been minimized and that any direct influence of the apparatus
on the results is at least understood, if not avoidable. To illustrate this problem,
in Fig. 6.30 the Tg values (open symbols) of polystyrene (determined as explained 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 cooling) 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 according 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,
211
1,5
1.0
'-'
.9)
0,5
0,0
370
375
Tg In K
380
212
dt
~=f-To T(T, T f )
213
Then, the dependence of the fictive temperature Tf can be expressed according to the Boltzmann superposition principle:
T
Tf(t)
= To + f
To
(1- exp
(-sP dT
The most frequently used expression for T (T, Tf ) although purely empirical is
the Moynihan equation (Moynihan et aI., 1976):
T (T,
Tf )
= A . exp ( x -i1h* + (1 RT
i1h*)
x) . -
RTf
Tg
Tg,=
Mn
where Tg,= is the glass transition temperature of a polymer with infinite number average molar mass and A and A' are constants for certain broad classes
of materials. Both equations can be used for high-molar mass polymers but
that of Ueberreiter and Kanig is much better for oligomeric glasses. If, in such
214
lb 11
4 h
3h
2h
15 h
1h
min
min
'"
QJ
..c
100
200
300
O(
400
tempera ture
215
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 bisphenol-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 responsible 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 experiment. If Tg,= represents the glass transition temperature of the completely
reacted and cross-linked polymer, the following is valid:
216
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 heterogeneous glasses before and after annealing. The two heterogeneous blends differ in the amount of both copolymer components. After annealing the homogeneous 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 between 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 transition 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
217
quenched
annealed
B
-'"
)0
quenched
J:
annealed
quenched
30
50
temperature
70C
90
..
changed mechanical properties. On the other hand, there is a risk for an undesirable 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
218
5
4
3
2
'VJ
d
1
'--"0
"2l'
-1
=
0
-2
-3
-.:::J
-4
29
3.0
3.1
3 2 K- 1 33
1000/ T - -
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 frequency of the measurements. In a TMDSC experiment the former manifests itself 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-reversing 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 investigation of substances with metastable phases. On the other hand there is the
disadvantage that it is not possible to carry out experimental manipulations
(e. g., nucleation of a supercooled liquid) inside the small and practically always
closed crucibles.
6.S.1
Applications in Biology and Food Science
The investigation of polymorphism - defined as the ability of a chemical compound to crystallize into different states (modifications) - is increasingly important in food science.
A well-known example is cacao butter. Chocolate products with cacao butter
have the desirable consumer properties only if the p-form with a melting point
at 35C is applied. DSC measurements are an effective tool for both optimizing
the production method and the quality control.
Another example from food science is common butter. The objectives of investigations are an improvement of the cold consistency (spreading property) of
butter and the concurrent reduction of softening when the temperature is increased. From DSC curves the influences of additives (triglycerides) with lowmelting point, and differences in the cream-ripening method can be directly
concluded (Schaffer et al., 2001).
A frequently used sugar substitute is isomalt, a nearly equimolar mixture of
two isomers containing crystal water. The crystalline material melts around
142-150C, and after cooling results a glassy state with a glass transition temperature at about 60C. Like other carbohydrates used in food industry, isomalt
is very often used in its amorphous form. The final quality and the storage
stability of isomalt containing products depends on the aging behavior of the
non-equilibrium glassy state. Therefore, the properties of both crystalline and
220
0.03
WIg
0.02
!:!
a
.....
=-
0.01
'"
.&:.
u
u
C1J
0-
III
-0.01
18
21
27
24
temperature
30
.-
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 proteins, 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 sensitive DSC instruments. As an example, Fig. 6.34 shows the transition from gel to
liquid crystalline state of an appropriately prepared liposome of 1.75% of
dimyristoyl phosphatidylcholine (DMPC) in water. The reversible transition
(curve 1, Q = 0.47 J per g total sample mass, Tp = 24.6C) corresponds to the
change of the hydrocarbon chains from an ordered (crystal-like) to a liquid-like
structure. The structural transition of the lipid can be perturbed by molecules
which intercalate among the lipid chains and hence hinder the regular packing
of the phospholipid molecules into gel-phase structures. This can be realized for
instance by addition of the lipophilic vitamin D3. With increasing concentration
of vitamin D3 in the liposomes, the peak becomes broader and smaller and
the peak maximum shifts to lower temperatures (curve 2, 3 % of vitamin D3,
Q = 0.15 J g-I, Tp = 22.7C). Addition of more than 5% of vitamin D3 to the
DMPC-water lipid (curve 3) causes the disappearance of this peak.
221
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 polymorphism (Grunenberg et aI., 1996).
General Remarks
Phase transitions between different modifications may be reversible (enantiotropic) or irreversible (monotropic). This behavior can be understood
best using energy-temperature diagrams (Fig. 6.35, Burger, Ramberger, 1979;
Kuhnert-Brandstatter, 1996; Grunenberg et aI., 1996). Two modifications A and
B are called enantiotropic or monotropic forms, if the free (Gibbs) energy curve
of the liquid G1 crosses the respective curves of the solid modifications A
and B above (Fig. 6.35 a) or below (Fig. 6.35b) the intersection point of GA and
GB Burger has formulated two important rules which immediately follow from
Fig. 6.35.
222
H,
>--
C"l
~B~~~~At-- __
ill
--~
,~-
~--
-G A
GB
__________~__~____~____~____~G,
Tfus(B)
temperature/K -
H,
llf"HIAI
>-.
en
- _ _ _ lll"H<BAl
::;;
I ~--==-~CHffi-:;;c:
ill
B
-___
-
HB
HA
--- "'"
--~
; ~-_GB
~-GA
G,
o
temperature/K -----
but only the differences against a reference state. In this case the liquid state of that
modification with the highest melting temperature is used as reference state.
With
(aH)
=C
aT
p
and
as)
(aT
Cp
=T
one obtains:
Tfus,B
HB(T)-HA(T)=t1fusHA-t1fusHB+
Cp,ldT+
Tfus,B
Tfus, A
Cp,BdT-
Cp,AdT
Tfus, A
and
t1f H
t1f H
Tfus,B
Tfus,A
Tfus,B
Tfus,A
SB(T)-SA(T)=~-~+
Tfus,B
Tfus,A
223
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 spontaneously 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 chemical interactions between drugs and excipients. The knowledge of the polymorphism of a excipient including its aging behavior is extremely important for the
development of preparation forms such as suppositories, micro emulsions and solutions in soft gelatin capsules. Formulations with amorphous substances tend to
form crystalline modifications with reduced bioavailability upon storage, strongly dependent on climatic conditions (temperature, moisture content). The determination of the amorphous part in semi-crystalline drugs is usually performed by
X-ray diffraction. The detection limit is 5-10 %. Measuring the crystallization enthalpy, 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 received state) melts at about 80C. On cooling from the melt and in the absence
of crystal nuclei of the stable modification the metastable modification crystallizes 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 recrystallization of the supercooled melt between 65 and 80C is found.
Remark: The formation of such metastable phases is one of the main problems encountered in the determination of phase diagrams, and this must always
be borne in mind to avoid misinterpretation of the observed curves.
Phenylbutazone is another well-known material that shows polymorphism.
The stable modification (curve 1 in Fig. 6.37a) melts at about 103C. Rapid cooling of the melt yields a glassy phase. On reheating the glass (curve 2) this first
recrystallizes to a metastable form at about 35C. Subsequent behavior depends
on the heating rates, either melting of the metastable modification at about
93 C or of the stable structure at 103C will be observed. This is because the rate
of transformation from the metastable to the stable form is low in the solid
phase (i. e., < 93C), only at low heating rates 1 K min-I) will there be enough
224
50.0
mW
37.5
1
~
::
25.0
=-
C1I
-'=
12.5
"\
.-._"
50
58
/
,/
66
74
temperature
--.-.
82
O(
90
Fig.6.36. DSC melting curves of the thermodynamically stable (1) and of the metastable (2)
modification of acetamide
time during the experiment for this to occur. At high heating rates (> 20 K min -I)
the formation of the stable modification does not happen, not even in the liquid
phase after the melting of the metastable modification. The corresponding DSC
curves are shown in Fig. 6.37b. For better comparison of both curves, curve 2
was also run with a heating rate of 40 K min- I following a preliminary heating
to 98C at 1 K min- I and recooling to room temperature. At moderate heating
rates (ca. 5 K min-I), however, the sample reaches some intermediate state. Any
remaining metastable phase that has not transformed during the slow solid
state reaction, melts at 93C, it then transforms to the stable modification as
an exothermic reaction and finally melts at 103 C (curve 2 in Fig. 6.37a). This
can be proved by quenching the sample immediately after completion of the
exothermic reaction (at 100C) and reheating it again, when only the melting
peak of the stable phase is seen (curve 3).
6.5.3
Other Applications
Liquid Crystals
DSCs are used very frequently for the investigation of polymorphism of substances with liquid-crystalline mesophases. The reliable evaluation of mesophase
transitions with small or very small transition energies (sometimes <0.1 J g-I)
makes great requirements on apparatus and working conditions. An example is
the behavior of cholesteryl myristate during heating and cooling (Fig. 6.38). The
first peak (Q = 77.3 J g-I, T = 69.0C) corresponds to the formation of a smectic
225
20
mW
15
f
.!!
...
10
CI
-...
:;J:
0
CI
..c:
3
a
0
25
40
55
85
70
tempera ture
100
...
115
70.0
mW
52.5
.!!
....
"" 30.0
:;J:
0
....
....
""
~
.c;
17.5
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 reheating 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
20
mW
f
~
E
=-
15
10
-=c
'"
..c:
o~------~------~------~------~------~~----~
60
65
70
75
tempera ture
80
85
90
mesophase. The following two peaks (Q = 2.5 Jg-I, T= 77.7C and Q = 1.9 Jg-l,
T = 83.0C) correspond to the formation of the cholesteric mesophase or the
transition into the isotropic melt. All phase transitions are reversible (compare
curves 1 and 4). The reason for the narrower peak shape of the first peak of
curve 4 (Q = 76.9 Jg-l, T = 68.5 C), compared with that of curve 1, is the better
heat transfer to the sample after melting. If the cooling run (curve 2) is stopped
before the crystalline phase appears, only the two liquid-crystalline phase transitions are observed during the subsequent reheating (curve 3). The heat of transition from the crystalline to the liquid-crystalline mesophase is usually - as in
this example - much higher than those of the following mesophase transitions.
Further, it can be demonstrated that there are practically no supercooling
phenomena for mesophase transitions. 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 (onset differences of AT = 0.2 K and AT = 0.0 K for the two transitions). The crystalline phase, however, shows considerable supercooling. It crystallizes at about
30C, equivalent to a AT of about 40 K.
Extreme Melting Ranges
Sometimes melting processes extend over very broad ranges of temperature.
For low molar mass substances this may be due to large amounts of impurities
227
(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 reasonable integration limits for the enthalpy calculation is then very problematic.
Mathot, Pijpers, 1983; Mathot, 1984; Mathot, Pijpers, 1989, recommend the use of
the heat capacity (cp) function rather than the normal heat flow rate curve of the
sample for evaluation to overcome these difficulties.
mW
mesophase
37.5
nematic -+
liquid pha se
.!!. 25.0
Q
....
:J:
Q
C1.I
~
12.5
L -_ _ _ _ _ _
40
______
______
80
tempera ture
________
160
120
____
200
..
Fig.6.39. DSC curve of the phase behavior of calcium stearate monohydrate (mass: 3 mg,
heating rate: 10 K min-I)
228
Generally it is found that both the position and the shape of the peaks from
repeated cooling/heating cycles are extremely reproducible. An alternative but very
time-consuming technique replacing the steps 3 and 4 was suggested by Neffati et
al., 1998 ("fractionation" of the crystallization or melting peak). It is applicable in
all cases of a distinct hysteresis between freezing and melting temperatures.
A typical experiment is shown in Fig. 6.40 for a porous silica gel (f3 = -1 and
+ 1 K min-I) ftlled with water. The large peaks correspond to the excess bulk
water, the broader and less intense ones to the "confined fluid". Contrary to the
shape of the peaks for the bulk liquid, the width of the peaks for the pore liquid
229
I
OJ
1-
(onfined \later
Wg- 1
01.----+---
cJ
-1
1 - - - bulk water
cJ
OJ
-L
-2~----~----~~------~----~
-30
-20
-10
temperQture
10
..
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" conditions 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 uninteresting but often very disturbing if overlapping with the freezing peak of the pore
water. Other liquids, e.g., cyclohexane, show much smaller supercooling. Probably, 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 temperature is given by a generalized form of the Gibbs-Thomson equation (Faivre
et al., 1999):
o Ysl
V.
I1T= - a T fus ' - ' - r I1fusH
where 11 T = (T?us - T) is the melting point temperature depression of the confined liquid, T?us is the melting temperature of a crystal of infinite dimension,
Ysl is the average interfacial tension of the crystal, V. is the molar volume of the
crystal, I1fusH is the molar enthalpy of melting and r is the radius of the pores.
The factor a depends on the shape of the pores (Faivre et al., 1999). Apart from
230
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 dependence 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 rewritten (Schmidt et aI., 1995) as
K
t:..T=-(r - t)
with
o
K= -a' T f us
V.
VI'--
's
ilfus
That means the effective pore radius is reduced from r to (r - t). If the experimental data for defined and systematically varied pore sizes are fitted to this
equation, t is in the order of one to three monolayers. The accessible range of r
is limited for large pores because of overlapping and missing separation of the
peaks for bulk and pores, and for very small pores because of missing filling
with the test liquid (dependent on its required space).
The same technique can also be used to obtain the distances of the cross-link
points of swollen polymers after equilibration in a appropriate swelling agent,
e. g., cyclohexane.
6.6
Determination of Phase Diagrams
The determination of phase diagrams in the fields of metallurgy and mineralogy
was one of the earliest applications of thermal methods. It was one of the decisive
reasons for the introduction of Differential Thermal Analysis (DTA, see Appendix 1) at the beginning of the twentieth century.
As an example, the phase diagram of the binary system benzil!acetanilide is
shown schematically in Fig. 6.41 together with the DSC curves of the heating
runs of the pure components (curves 1 and 7) and of five mixtures including
the eutectic (curves 2 to 6). In accordance with the phase diagram benzil and
acetanilide are completely miscible in the liquid and immiscible in the solid
phase. Pure acetanilide melts at 115C, pure benzil at 95C. The eutectic, with
a mole fraction of about XBen = 0.58, melts at 78C. In Fig. 6.41 a two different
melting pathways (mole fractions Xl = 0.1 and X2 = 0.9 of benzil) are sketched.
The measured DSC curve of every mixture shows two events. The first is the
peak due to the melting of the eutectic, the subsequent broad endothermic effect
is caused by the solution of the remaining solid component in the equilibrium
melt (following the pathway along the liquidus curve in Fig. 6.41 a). The final
point of this peak corresponds to the liquidus temperature according to that
mole fraction in the phase diagram. Both the pure components and the eutectic
231
115 6C
95C
78C
a
x
__
WIg
6 1--_2=--_ _
1
4
5
6
b
OL-____~~c=~====~====~~=_~
60
73
86
tempera lure
99
112
125
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 melting process for two mixtures with the mole fractions Xl and X2, b DSC curves of the pure components 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 extrapolated peak onset temperatures in the usual way, whereas the determination of
the liquidus temperature is not easy. The peak maximum temperature (if necessary corrected for thermal lag, see Sect. 5.4.4) of the second broad peak is a good
approximation. In every case slow heating rates 2 K min-I) and small sample
masses should be used, to ensure approximate thermodynamic equilibrium at
232
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 suitable 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 system 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 calorimeters mentioned in Chapter 1. They provide detailed information for the optimization of process parameters, reducing production costs and avoiding safety
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 transferred to the surroundings. Under worst case conditions this may lead to a runaway and subsequent fire or explosion. The direct detection at ambient temperature of the extremely small heat flow rates in question is, in general, not possible 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 obtained 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 decomposition of azidophenylacetic acid was determined. In addition the effects of many
kinds of "impurities" were investigated. In a similar way the autocatalytic influence of the decomposition products was definitely proved.
DSC measurements supply valuable results regarding the thermal decomposition 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
6.8.1
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 crystallinity 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 quantities is indispensable in polymer technology. DSC offers the possibility to characterize plastics in a fast and satisfactory way. In fact the rapid development of a precise equipment for quantitative thermal analysis (DSC) was bound up with the increasing 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 temperature 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 partial oxidation). In heterogeneous or micro-heterogeneous amorphous blends
two or more glass transitions are found, which are characteristic of the individual components. The relative amounts can be estimated by evaluating the
respective step heights. In partial crystalline polymers the degree of amorphous 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 transition on heating. Glass transition evaluation is generally a very important application 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 transition and crystallization) characterize the crystalline polymer. As a rule it
is not difficult to identify the components of a plastic mixture, using the characteristic temperatures of the melting peaks. Often even a quite accurate
quantitative analysis is possible. 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 respective peaks because of immiscibility in the solid phase - is nearly always
fulfilled for plastic mixtures. Some polymers appear in different polymorphous forms, for such plastics solid-solid phase transitions can be found. For
semi-crystalline products the degree of crystallinity is an essential quantity
235
236
6.8.2
Determination of the Degree of Crystallinity
We (T)
J g-1
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
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 temperature 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 polymer 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 temperature dependent change of the degree of crystallinity (the peak). Unfortunately the two-phase model fails with some types of polymers, e.g., polyethyleneterephthalate. The baseline can be calculated in such cases by use of im-
238
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 approximated 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 measured 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 crystallinity 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 suitable 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 literature 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 crystallization. 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
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,
239
but with crystallization the heat capacity changes as well and the latter can precisely 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 concentration 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 vibrational cp in the melting region and depends on frequency (see Fig. 6.43). From the
20
JK" g"
10
Icpl
, S ~H,
1 11.5 ~H,
) IS mH,
~ .is mH7
OS( 10 Kim,"
0
90
100
110
120
tempera t ure
O(
130
150
240
Fig.6.44. Excess heat capacities (solid curves) calculated
with Eq. (3.22) from the
measured apparent heat capacity (dotted curves) from
quasi-isothermal TMDSC
measurements at different
temperatures (UHMMPE;
mass: 2-3 mg, frequency:
12.5 mHz, tp : 80 s, TA : 53 mK)
10
50
100
10 00
---150
time
250
min
1<>1
000
(>
[]
00
100
(>
c:
E
!--
c:
(>(>(>
.......
10
(><>
[]
[]
rP
[][]
1
24
25
26
liT 10 J
K-'
27
241
6.9
Purity Determination of Chemicals
The theory of purity determination using DSC is based on the thermodynamics
of two-component systems. The simplest and most widely used theory presupposes an eutectic mixture of ideal behavior. For this case we find in textbooks of
chemical thermodynamics that the mole fraction x of one component (defined
here as the pure one) can be calculated as follows:
1
lnx=R
i1H(T)
f --dT
T2
Ttrs
(R, gas constant: 8.31441 J mol- 1 K- 1; i1H(T): phase transition enthalpy at temperature T of the pure component in question, Ttrs : phase transition temperature
of this completely pure component}.
Ttrs
or
(_1 _~))
T
x = exp (i1H
R Ttrs
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
Ximp
(T) = i1H - - 2 -
R T trs
(6.2)
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 involved. In principle, any melting of an eutectic mixture starts at the eutectic temperature and ends at the temperature T which is related to the composition of
the mixture via Eq. (6.2) (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 concentrations. With increasing impurity, the melting peak becomes lower and lower
and less sharp. In reality, for small amounts of impurity, the eutectic melting
peak and the starting of the main peak are hardly visible on the DSC curve. The
real curves measured in a power compensation DSC (Fig. 6.47) differ consider-
242
mW
1
...c
oL-________~==~~~~~~~L_~
350
345
340
temperature
355
..
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 thermal conductivity of the DSC (cf. Sect. 5.4). As a result the end temperature of the
peak cannot be determined with the required accuracy and the concentration of
the impurity must therefore be determined in a different way.
The software in commercial DSCs usually starts from the assumption that,
during the melting process, the instantaneous mole fraction of the mixture at
the temperature T (along the eutectic curve, cf. Fig. 6.41 a) relative to the initial
composition is the same as the relation between the heat of fusion used up to
that temperature and the total heat:
Ximp
(T) = Qr == F (T)
Ximp
(6.3)
Qtot
The right-hand side of this equation is the quotient of the partial peak area up
to the temperature T and the total peak area if smearing effects are disregarded.
For this quotient, the abbreviated form F (T) (relative partial area) is often used.
Inserting Eq. (6.3) into Eq. (6.2) furnishes an equation according to which the
impurity Ximp can be determined from the slope of a plot of the temperature
versus lIF(T) which should be a straight line. In practice this is not the case
because of the difficulty in deciding when melting actually starts and the limited
heat transfer to the sample. Commercial software generally corrects the measured values in such a way that a straight line is obtained in the plot. This is
of course only an approximation to correct for deviations from experimental and
243
0.6
mW
2
0.4
f
C1J
~
c
....
OJ
~
0
;;::
~
0.2
C1J
..r:::.
0.1
0.0
320
330
tempera ture
360
Fig.6.47. Measured melting peaks of an eutectic system, normalized to unit area (impurities
the same as in Fig. 6.46)
244
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 beginning 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 mathematical 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 assumptions 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 assumption of eutectic mixtures, which is actually only one special case, the accuracy 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 nowadays much more precisely done with HPLC and other modern analytical methods. 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.
DSC furnishes information on temperature and heat flow rates (or respectively,
heat). Whether it is suited to solve the respective problem depends on the efficiency of the instrument. The characteristic data of the DSC which describe the
instrument unambiguously must therefore be known. They allow a decision to
be taken as to whether the DSC will be suitable for the intended use, and they
also make a comparison with other DSCs possible.
A distinction can be made between:
1. the characterization of the complete instrument,
2. the characterization of the measuring system,
3. the characterization of the results of a DSC measurement.
7.1
Characterization of the Complete Instrument
The following serves to characterize the DSC instrument as a whole:
-
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 following:
-
noise,
repeatability,
linearity,
time constant,
sensitivity,
resolution.
246
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 instantaneous 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 resolution" 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 amplified 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 disturbances from the environment to which the sensors are subject. It determines the maximum possible signal-to-noise ratio.
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 manufacturers) using 4,4'-azoxyanisole (which shows a reversible liquid crystal
to isotropic liquid transition at ca. 134C with a small heat of transition
of ca. 2 Jg- 1) was performed under defined experimental conditions (van
Ekeren et aI., 1997). The signal to noise (pp) ratio was determined, the reported 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 compensated) 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 significant DSC quality (e. g., extrapolated peak onset temperature, peak area,
course of the zeroline etc.) several times under the same conditions on samples 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 superimposing 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 measuring for example, the peak area caused by the melting of a pure metal several times under the same conditions. (Either only repeat measurements without moving the sample crucible or with replacing the crucible after each run).
The curves with the peaks and the baselines constructed for each peak are superimposed. The range related to the mean area shows the repeatability of
the peak area (heat of fusion) determination. A separate baseline repeatability can be determined (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 results of measurements carried out on a sample using different instruments in
248
An
mn
CD
'-J CD
Q)
C0
--P-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 partial figure: ratio Anlmn with the range of uncertainty of the area determination, - - - -deviation from ideal linearity (here only for very small (1) and very big (4) sample masses)
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.
249
4im (t)
=---
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 (according to Hemminger, 1994).
After a constant measurement signal 4>m,o has been
reached, the constant heating
current is switched off at the
time to' After the time interval T, the measurement
signal 4>m,o has dropped
to 4>m,o/e
., ., . ., .. ., ., . r - - - - - - ,
t
hme
250
Fig.7.4. Determination of
the time constant rwith the
melting peak of a pure
substance (according to
Hemminger, 1994)
time
t* should not be too close to the switching-off time to because further,
smaller time constants ( T2, T3 etc.) may there be superimposed on the maximum time constant Tmax (see below).
- When no calibration heater can be installed in the crucible, the descending
section of a transition peak (when a pure substance melts or crystallizes)
is evaluated by the "tangent method" (cf. Fig. 7.4). Several tangents are
plotted to the descending section and the time constants T) Tmax are
graphically determined from the intersections with the linearly interpolated baseline. Towards the end of the descending curve, a constant value T
(= M between t* and the intersecting tangent on the baseline) results: this
is the (greatest) time constant Tmax which describes the thermal inertia of
the measuring system in good approximation.
The (usually sudden) solidification of the supercooled melt of a pure
substance (e. g., pure tin) causes a pulse-like heat production in the sample.
The resulting heat flow rate curve is the so-called apparatus (or Green's)
function (see Sects. 3.3.3 and 4.7), this is ideally suitable to determine the
apparent time constant of the system with the tangent method.
251
1
\
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 transition 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 interpreted and published or used for further calculations (crystallinity, kinetics
etc.). This evaluation is made on the basis of the data characterizing the efficiency 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 different possibilities exist. Methods which use statistical tools applied to interlaboratory comparisons with terms like "random" and "systematic" measurement 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
The description of the variability of results of measurements is usually separated into three parts. According to the "International Vocabulary of Basic and
General Terms in Metrology", 1994, these parts are:
Repeatability characterizes the closeness of the agreement between the
results of successive measurements of the same measurand carried out under
the same conditions of measurement (cf. Sect. 7.2). Usually this means a multiple
use of the same DSC, the same measurement procedure, and the same reference
material for calibration by the same observer at the same conditions over a short
period of time. Furthermore, it is known that the results of DSC measurements
can depend on contact resistances between sample and furnace and the positioning of sample and reference sample inside of the furnace. A realistic determination of the repeatability of DSC-measurements should therefore include
a suitable procedure to consider these effects. In most cases the specification of
the repeatability is intended as a measure of the variability of DSC measurements within the specific laboratory.
Reproducibility characterizes the closeness of the agreement between the
results of measurements of the same measurand carried out under changed conditions of measurement (cf. Sect. 7.2). For DSC measurements these changed
conditions may include the measuring instrument, the observer, the method of
measurement, the evaluation procedure or software, the calibration standard,
conditions of use, or the time. In most cases the specification of a reproducibility is intended as a measure of the variability of DSC measurements between different laboratories.
Accuracy is an often used qualitative term which describes the closeness of
the agreement between the result of a measurement and the true value of the
measurand. But in all cases the true value of the measurand is not known.
Error is the difference between the result of a measurement and the true value
of the measurand. A systematic error is a mean that would result from an infinite
number of measurements of the same measurand carried out under repeatability conditions minus the true value of the measurand.
Bias means a systematic error of the indication (reading) of a measuring
instrument. This means in our case a systematic difference of the results of DSC
measurements on well known (and stable) materials in comparison to the corresponding reference (literature) values.
Correction is a value added to or subtracted from the uncorrected result of a
measurement to compensate for the systematic error. In some cases a correction
is carried out by multiplication or division of the uncorrected results of a measurement with a correction factor to compensate for the systematic error.
A further source of confusion is the different meaning of "estimation"
in common sense and in statistics. From the statistical point of view the
word "estimation" means the application of a well defined mathematical procedure (e.g., the calculation of the standard deviation of the mean of a population) and not the making of a rough guess. Although in many cases experience based uncertainty estimations give reasonable results, a fair comparison
of methods and instruments and a credible specification of the uncertainty should be based on a comprehensible (experimental and mathematical)
basis.
253
Repeatability
(95% confidence limit)
Diphenyl ether
Linear polyethylene
}~%
Indium
a
Reproducibility
Bias
(95% confidence limit)
+0.95%
-1.1%
7.3.1
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 maximum (information) entropy and the Bayesian-statistics. It is somewhat complicated, 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 corrections should be identified and applied.
254
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), calibration 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 corrections 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 measurement. 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 (according to Chapter 4) by measuring the phase transition temperatures of different
calibration materials as a function of the scanning rate and subsequent extrapolation 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 uncertainty caused by the interpolation between the different phase transition temperatures 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 temperature 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 temperature 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
255
properties of the sample itself (see Sect. 3.1 and 3.3.3). For some types of instruments a determination of the temperature difference between the temperature 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 corresponding temperature difference is typically between 0.5 K and 3 K. Thus the
resulting model equation for the sample temperature Ts reads
Ts = T + ~Tcalib + 6Tmat + ~Tlin - 6T1ag
(7.2)
In the next step the standard uncertainties u of all input quantities Xi of Eq. (7.2)
(e. g., 6 Tmat or ~ Tcalib ) must be determined. As an example, the standard
uncertainty of the temperature calibration u (~Tcalib) can be determined by
repeated measurements of the melting temperature (extrapolated peak onset
temperature at zero scanning rate). If normal (Gaussian) distribution can be
assumed for the measurand the standard deviation of the experiments is a measure of the standard uncertainty u (~Tcalib) and corresponds to a coverage
probability of approximately 68 % (coverage factor k = 1). In most cases a standard 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 measurement at 500C. As a result, the expanded uncertainty (k = 2) of the temperature 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 Technology, Gaithersburg, USA) SRM-720 sapphire as calibration material is used, the
uncertainty can be taken from the certification report. But in many cases (noncertified) sapphire from other sources is used. In this case an additional contribution to uncertainty 6cRef is included which takes care of a possible specific
Table 7.2. Uncertainty budget for the temperature determination example
Input quantity a
Estimate a
Xi
Xi
T
6 Tma,
ll.Tcalib
ll.Tjin
6Tjag
Ts
Standard
uncertainty
U(Xi) in K
Probability
distribution
Sensitivity
coefficient
oTs
oXi
Uncertainty
contribution
Ui(Y) in K
500C
O.OK
1.9 K
O.OK
1.0 K
0.1
0.3
0.5
0.2
normal
normal
normal
normal
1.0
1.0
1.0
1.0
0.1
0.3
0.5
-0.2
500.9C
0.6
a Input quantities Xi are measurable quantities like the temperature T or the thermal lag 6Tjag.
The corresponding quantity in statistics is the estimator. The value of the estimator is the estimate or best estimate of the unknown quantity and corresponds to the result of a measurement
(e. g., arithmetic mean).
256
heat capacity difference between the used calibration material and the NIST
SRM -720 sapphire. The modified model equation for the calculation of the specific heat capacity of the sample is then given by
(7.3)
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 contribution to the standard uncertainty Ui (cs) associated with the output estimate Cs
can be calculated according to
Ui (cs)
oCs
= oXi U (Xi)
(7.4)
For uncorrelated input quantities the standard uncertainty of the specific heat
capacity (combined uncertainty) is given by
(7.5)
Special investigations have shown that the relative uncertainty of the specific
heat measurement between 0 C and 600C doesn't depend on temperature. The
uncertainty budget for the specific heat capacity measurement has therefore as
an example been calculated at a temperature of 250C. Table 7.3 gives a summary 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 specific heat capacity determination is U (cs)/ Cs = 1.5 % with a reported result at 250C
of Cs = (1.015 0.016) J. g-l. K- 1
Table 7.3. Uncertainty budget for the specific heat capacity determination example
Input
quantity
Xj
mRef
CRef
DcRef
4>Ref
l1~ef
ms
4>s
l14>s
4>0
l14>0
Estimate
Xj
Standard
uncertainty
Probability
distribution
Uncertainty
contribution
Uj (y)/J . g-l . K- 1
rectangular
rectangular
rectangular
rectangular
rectangular
rectangular
rectangular
rectangular
rectangular
rectangular
0.1. 10-3
1.2 . 10-3
2.9.10- 3
4.6. 10- 3
0.3 . 10-3
0.1. 10-3
5.3. 10- 3
0.3. 10- 3
1.5. 10- 3
0.004.10- 3
u(Xj)
55.41 mg
1.060 J . g-l . K- 1
0.0 J . g-l . K- 1
11.09 mW
-0.009mW
52.29 mg
1O.13mW
-0.01 mW
1.2mW
-0.002mW
0.006mg
0.001 J . g-l . K- 1
0.003 J . g-l . K- 1
0.05mW
0.003mW
0.006mg
0.05mW
0.003 mW
O.lmW
0.0004mW
257
7.4
Check List for OSCs
The following check list serves to collect the essential characteristic data of a
DSC.1t may be used to
- ask the manufacturers for the values,
- establish a guideline when characteristic data are measured in one's own laboratory 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
o
o
o
Special features:
Sample volume (standard crucible):
..... mm3
Atmosphere
(vacuum?, which gases? pressure?):
Temperature range:
Scanning rates:
Zeroline repeatability:
Appendix 1
260
Appendix 1
I1T
4
a
I1T
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 differential thermocouple, 4 programmer and controller, S sample substance, R reference sample
substance, Tp furnace temperature, t time
261
time
6.T
b
Fig. A 1.2. a Instrument for Differential Thermal Analysis (DTA), measuring system with freestanding crucibles, b Schematic measured curve ,iT(t) (endo down).
1 crucible,2 furnace, 3 differential thermocouple, 4 programmer and controller, S sample substance, R reference sample substance, Tp furnace temperature, t time
262
Appendix 1
Appendix 2
Calorimetry - a Synopsis
A great variety of calorimeters serve to measure heats and heat capacities in various fields of application. In the following, a classification system for calorimeters and a couple of examples of different types of calorimeters will be presented. The aim is to give a structured survey of the whole field of calorimetry which
may help to better recognize and evaluate the advantages of and limitations to
the DSCs which result from their mode of operation.
Classification
In a classification system, calorimeters are arranged in groups according to
particular characteristics. Various classification systems are reasonable and
practicable (cf. Hemminger, H6hne, 1984; Rouquerol, Zielenkiewicz, 1986). It
may sometimes be difficult to arrange calorimeters in proper order in a relatively simple system, and such a classification may even be impossible. Classification systems covering the entire field of calorimetry tend to become very
sophisticated, end in itself and are rather useless for practical applications.
When thermodynamic principles of heat exchange or the aspect of how
the caloric signal is formed are taken as a basis for classification, the result
will always be that a certain number of calorimeters are characterized satisfactorily whereas additional explanations or auxiliary definitions are required
for others. Nevertheless, a classification system is useful to show basic principles
of calorimetry. It should be based on existing instruments and open to future
developments. In the following, a simple system for practical use will be developed which will suffice to discuss characteristic features and error sources by
groups.
Criteria for the classification are the following:
1. The Principle of Measurement
1.1 Measurement of the energy required for compensating the heat to be
measured (heat-compensating calorimeters).
1.2 Measurement of the temperature change of a substance due to the heat to
be measured (heat-accumulating calorimeters).
1.3 Measurement of the heat flow rate between sample and surroundings due
to the heat to be measured (heat-exchanging calorimeters).
Appendix 2
264
2. Mode of Operation
2.1 Isothermal
2.2 Isoperibol
2.3 Adiabatic
2.4
2.5
2.6
2.7
Scanning of surroundings
Isoperibol scanning
Adiabatic scanning
Temperature-modulated
static modes
dynamic modes
3. Construction Principle
3.1 Single calorimeter
3.2 Twin or differential calorimeter
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 potentialities 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, wellknown 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, therefore, represent important error sources. Moreover, in the case of electric compensation, 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
266
Appendix 2
1 - -.....[> X
o
o
o
o
o
o
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
267
Appendix 2
v
5
T
a
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 accuracy, the calorimeter substance (e.g., water) is generally cooled with constant
Peltier power and at the same time heated with equally high Joule's heat (controlled). To compensate an endothermic effect, the heating power is increased in
order to keep the temperature of the calorimeter substance constant. When an
exothermic effect is to be 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, temperature control is delayed by the heat transfer processes, and it is very difficult
to obtain spatially homogeneous temperature fields by electric heating.
The advantages of these quasi-isothermal calorimeters consist in the simple
and very precise measurement of the electric compensation energy and the possibility of using highly sensitive sensors to measure temperature changes; these
sensors must not, however, be calibrated.
This is done in that a sensitive thermometer activates the compensation control
circuit so that, if possible, no temperature change due to reaction heat takes plase.
3rd example: "Adiabatic scanning calorimeter"
1. Compensation of the heat to be measured with the aid of electric energy
2. Adiabatic scanning
3. Single calorimeter
= Pel, 1 -
Pel,2
dT
dt
or
C(T) = P" ((
~~f (~:)J
(dT/dt: preset)
Appendix 2
269
3
4
6. T = 0
f!1
---
____---CD
time
...
270
Appendix 2
(1: run with the calorimeter empty, 2: run with the sample placed in the
calorimeter, Pel = i U electrical heating power, dTldt heating rate at the respective moment, C(T) comprises the heat capacities of sample and container.
The heat capacity of the empty container is determined by separate measurement).
When in a 1st order transition (e.g., melting of a pure metal) the sample temperature remains constant during transition in spite of the fact that heating power is 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 measurement 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).
Appendix 2
271
Heat-Accumulating Calorimeters
In the case of this calorimetric method, the effect of the heat to be measured
is not "suppressed" by compensation but leads to a temperature change in the
sample substance and a "calorimeter substance" with which the heat to be determined is exchanged. This temperature change is measured. When the change
is not too large, it is proportional to the amount of heat exchanged. The proportionality factor must be determined by calibration with a known amount of heat.
5th example: "Drop calorimeter"
1. Measurement of the temperature change of a substance due to the heat to be
measured
2. Isoperibol
3. Single calorimeter
The temperature of the surroundings is kept constant with the aid of a thermostat ("isoperibol": uniform surroundings). The heat Q to be measured is exchanged with the "calorimeter substance" and the temperature change !:.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 difference between calorimeter substance and surroundings, heat is exchanged. This
exchange must depend only on this temperature difference and must be reduced
as far as possible by appropriate measures (Dewar vessel, radiation shields ... ).
Otherwise, it cannot be determined by calibration and represents an error source.
In order to guarantee strict repeatability of the heat exchange,!:. T must amount to
only a few Kelvin. Since the temperature change of the calorimeter substance and
the unavoidable heat exchange with the surroundings take place simultaneously,
the temperature difference!:. T* used to calculate Qmust be determined according
to defined rules [e.g., International Standard ISO 1928-1976 (E); Rossini, 1956;
Gunn, 1971; Oetting, 1970; Sunner, Mansson, 1979] from the shape of the !:.T(t)
curve measured before and after the sample has been placed into the calorimeter
(cf. also 2nd example). Simple drop calorimeters serve to measure mean heat
capacities (temperature of the sample at the moment of its being placed into the
calorimeter: up to about 500C; sample mass between 10 and 100 g).
In "aneroid" drop calorimeters (Fig. A2.4b), the calorimeter substance is a
solid body of good thermal conductivity. The advantage over calorimeters filled
with liquid consists in that samples at high temperature (up to about 2000C)
can be dropped without the risk of evaporation or splashing.
Other examples of calorimeters of this class are instruments in which the
sample placed into it (solid body, liquid, gas) reacts with the calorimeter substance; this results in heat of reaction being released.
Appendix 2
272
'~
.;------;-Ir----.
-3
4
T
~ c::=J I'--_~
__ L
(I
I
L1 r"
I
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)
273
Appendix 2
1
T
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
measured
2. Adiabatic
3. Single calorimeter
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 calorimeter", see 5th example).
Bomb calorimeters of this type (usually automated) are widely used to measure the calorific value of solids or liquids under standardized conditions. In
general, they are calibrated with benzoic acid. Their uncertainty of measurement lies in the per mil range. (There are also "dry" bomb calorimeters, in which
the temperature change of the combustion vessel itself (of the bomb) is measured; bomb calorimeters with isoperibol surroundings are also used.) For a
detailed representation of bomb calorimetry, see Rossini, 1956; Skinner, 1962;
Sunner, Mansson, 1979.
7th example: "Flow calorimeter"
1. Measurement of the temperature change of a substance due to the heat to be
measured
2. Isoperibol
3. Single calorimeter (also designed as "twin")
In the case of the so-called "gas calorimeters", the heat to be measured is transferred, if possible completely, to a flowing medium (Fig. A2.6). The temperature
difference between the medium flowing in and the medium flowing out is proportional to the heat transferred. Calorimeters of this type are used to determine 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 -conveying medium must not be known. If an electric heater is used instead of the
burner, the specific heat capacity of the heat-conveying medium can in principle be measured as well (cf. Hemminger, 1988).
In biology, biochemistry and chemistry, flow calorimeters with liquids serve
to measure the heat development of microorganisms in certain nutrient solutions, or they are used to measure reaction heats. Two reacting solutions are, for
example, mixed in a reaction tube (Fig. A2.7). The uniform temperature of the
275
Appendix 2
T = const.
2
"--~-4
T,
3
----~@J~------
Fig. A2.6. "Gas calorimeter". 1 burner in which the gas to be measured is burnt, 2 heat exchanger, 3 "heat conveying medium" (e.g., air or water), 4 combustion gases, 5 temperature
sensor. Inside the heat exchanger, the hot combustion gases convey their heat to the heat conveying medium whose temperature increase T2 - 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 difference is a measure of the reaction heat. The calorimeter must be calibrated, either
with the aid of known reaction heats of liquids or with an electric heater installed in the reaction tube. A sophisticated example of a flow calorimeter with
stimulated cardiac muscle developing heat in the perfused tube has been
described by Daut et aI., 1991. [Flow calorimeters are also designed as twin
calorimeters to cancel out the influence of the isoperibol surroundings (heat
leaks) ].
8th example: "Adiabatic calorimeter"
1. Measurement of the temperature change of a substance due to the supply of
a known amount of heat
2. Adiabatic
3. Single calorimeter
This type of calorimeter (Fig. A2.8) is not designed to measure an unknown
heat; instead, a well-known, electrically generated heat Q = Wei serves to change
the sample temperature by ~ T. The temperature of the surroundings (adiabatic
jacket) is adapted to the measurement temperature with high accuracy in order
to avoid any heat exchange with the surroundings. Calorimeters of this type are
used to determine the phenomenological coefficient of the heat supplied, Q,
and the temperature change, ~T, of a substance (at constant pressure): the heat
capacity Cp (n. The following is valid: Cp = Q/~ T.
276
Appendix 2
1
j
I
I1 ......- - -
enter of
readand
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
.--.----
switch on P"
time
Fig. A2.8. a Adiabatic calorimeter, b Measured curve (according to Hemminger, 1994). 1 sampie, 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 electric 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 dependence of Cp as precisely as possible and to avoid errors due to inhomogeneous
temperature fields. The limits are determined by the noise and the uncertainty
of measurement of the temperature sensors (cf. Kagan, 1984; Zhiying, 1986).
Heat-Exchanging Calorimeters
In calorimeters which measure the temperature change of the calorimeter substance, i. e., heat-accumulating calorimeters (e. g., drop calorimeters, bomb calorimeters), 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 describes 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 between the individual measurement signals into account (Differential Scanning
Calorimeter: DSC).
9th Example: "Heat flux differential scanning calorimeter"
1. Measurement of the exchange of the heat to be measured between sample and
surroundings via a heat flow rate
2. Scanning of surroundings
3. Twin calorimeter
DSC with Disk-Type Measuring System (see Sect. 2. 7. 7)
A metal, ceramic or quartz glass disk with the sample and the reference sample
(or the pans) positioned on it symmetrical to the center, is placed into a furnace
(Fig. 2.1a). Heat exchange between furnace ("surroundings") and samples takes
place by heat conduction, radiation and convection. Strict repeatability of this
heat exchange as a function of the temperature (with the atmosphere remaining
unchanged) must be ensured. This is why a solid heat-conducting disk is used
which guarantees that the properties of the measuring system dominate. As a
result, the different characteristics of the individual samples contribute less
strongly to the kind of heat exchange than in the case of the DTA (the measurement signal itself must, of course, reflect the sample properties). The signal
I1T is measured on the solid heat conductor (disk) between the supports for
sample and reference sample (I1T = Ts - TR ).
Appendix 2
279
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 disadvantage of thermal coupling of sample and reference sample is get rid off and
we have a real differential calorimeter (twin). The function principle is otherwise similar to the disk-type DSC (see above).
DSC with Cylinder-Type Measuring System (see Sect. 2.1.3)
In another type of heat flux DSC, the two cylindrical containers for sample and
reference sample are connected with the furnace ("surroundings") by one thermopile 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 sample (ilTpR - <PPR).
MODE OF
OPERATION
CONSTRUCTION
(heat-compensating
calorimeters)
Measurement of the
temperature change of the
calorimeter substance
measured (heat-
accumulating calorimeters)
exchanging calorimeters)
Static:
isothermal
isoperibol
adiabatic
Dynamic:
scanning of surroundings
isoperibol scanning
adiabatic scanning
temperature-modulated
Single calorimeters
Twin (differential)
calorimeters
References
References
282
n,
References
283
Gallis HE, Bougrioua F, Oonk HAJ, van Ekeren PJ, van Miltenburg JC (1996a) Thermochim
Acta 274:231-242
Gallis HE, van Miltenburg JC (1996b) Thermochim Acta 274:223-230
Gallis HE, van Miltenburg JC, Oonk HAJ, van der Eerden JP (1996c) Thermochim Acta
286:307-319
Gallis HE, van Ekeren PJ, van Miltenburg JC, Oonk HAJ (1999) Thermochim Acta 326:
83-90
Galwey AK, Brown ME (1998) In Chapter 3: Handbook of thermal analysis and calorimetry,
vol 1, principles and practice, Elsevier, Amsterdam 147-224
Galwey AK, Brown ME (2002) Thermochim Acta 386: 91-98
GEFTA, Gesellschaft fUr Thermische Analyse (German Society for Thermal Analysis)
http://www.gefta.uni-freiburg.de
Gilmour IW,Hay IN (1977) Polymer 18:281-285
Ginnings DC, Corruccini RJ (1947) J Res Nat Bur Stand 38:538-591
Giron D (1999) J Therm Anal Cal 56: 1285-1304
Gmelin E (1987) Thermochim Acta 110: 183-208
Gmelin E, Rodhammer P (1981) J Phys E: Sci Instrum 14:223-228
Gmelin E, Sarge SM (1995) Pure Appl Chern 67: 1783-1800
Gmelin E (1997) Thermochim Acta 304/305: 1-26
Gmelin E, Sarge SM (2000) Thermochim Acta 347: 9-13
Gobrecht H, Hamann K, and Willers (1971) J Phys E: Sci Instrum 4:21-23
Godovsky YuK, Hohne GWH (eds) (1994) Special Issue: Deformation Calorimetry of Polymers;
Thermochim Acta 247,1: 1-128
Goldberg RN, Weir RD (1992) Pure Appl Chern 64: 1545-1562
Gray AP (1968) In: Porter RS, Johnson JF (eds) Analytical Calorimetry, Vol 1, Plenum Press
New York
Grewer T (1987) Thermochim Acta 119: 1-15
Grewer T, Steinbach J (eds) (1993) Special Issue: Safety Techniques; Thermochim Acta 225,
2:143-301
Gr0nvold F (1967) Acta Chern Scand 21: 1695-1715
Gr0nvold F (1993) J Chern Thermodyn 25: 1133-1144
GrunenbergA, Henck JO,Siesler HW (1996) Int J Pharm 129: 147-158
GUM: Guide to the Expression of Uncertainty in Measurement (1st ed. 1993; corrected and
reprinted 1995) International Organization for Standardization, Geneva
Gunn SR (1971) J Chern Thermodyn 3: 19-34
Hale A, Macosko C, Bair HE (1991) Macromolecules 24:2610-2621
Hancock BC, Shamblin SL (200!) Thermochim Acta 380:95-107
Hanitzsch E (1991) Thermochim Acta 187:275-281
Hansen LD (ed) (2000) Special Issue: Biothermodynamics: Molecular, Organismal and Ecological; 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
References
285
286
References
References
287
288
References
References
289
Subject Index
- reactions 162,174,177,179,181,189,195
- uncertainty 121,123
benzil 230
benzyl dimethyl ketal 180
bias 252
bioavailability 223
biochemistry 219,220
biological systems 6
Biot-Fourier equation 33
bisphenol-A-diglycidylether 176
Bode plot 59,62
bomb calorimeter 3,274
Borchardt & Daniels method 184
butter 219
AC calorimetry 7
accuracy 6,252
acetamide 223
acetaminophen 223
acetanilide 230
acrylate 192
activation energy 171,184-187
activation entropy 171
activation by radiation 180,192-195
adiabatic
- bomb calorimeter 274
- calorimeter 4, 268
- DSC 13,24
- scanning calorimeter 268
amplitude (TMDSC)
- temperature 51,60
- heat flow rate 53
- thermal wave 58
- transfer function 59
analogue electric circuit (see network)
32,
39,42,58
analytical equipment 5
aneroid drop calorimeter 271
aniline 176
annealing 202,217,228,233,235
apparent heat capacity (see heat capacity)
141
170
baseline 42,116,117,121-124,162,227,
237
- changes 124
- construction 121-124
292
calorimeter
- heat accumulating 271
- heat compensating 264
- heat exchanging 278
- types 3,263,264
calorimetry 1,3,263
- applications 3
- classification 263
- synopsis 263-279
Calvet type DSC 14-17
calcium
- oxalate monohydrate 227
- stearate monohydrate 228
catalyst (catalytic) 169,170,173,176,183,
188
certified reference material 65
characteristic temperatures 70,76,117
characteristic terms 115-117
characterization
- complete instrument 245
- materials 3,219-228
- measuring system 245
- results of a measurement 251
checklist DSC 257
cholesteryl myristate 224
classification of calorimeters 263
combined DSC 25
combustion calorimeter 274
comparison
- manufacturers 257
- measurements 245
- types of DSC 259
compatibility of substance and crucible
102
compensating heating power 17-20
compensation calorimeter 264-270
complex heat capacity (see heat capacity)
56,60
conductance (conductivity), thermal 33,
45,57,134,152
construction principle 264
controller of power compensation 18
convection heat transfer 45
convolution integral 56, 132
convolution product 57,132
cooling mode 84, 97
copper 105
correction 71,252
- of errors 252
coupled apparatus 5
cp (see heat capacity)
criteria of classification 263
cross-talk, thermal 42
crucible 10, 38, 102, 150
Subject Index
cryostate 6
crystal size 227
crystallinity, degree of 236
curve (see DSC curve)
cyclopentadiene 191
cylinder-type DSC 14-17
- advantages 16
- noise 17
D
decomposition 5,168,227,235
deconvolution 126,132,181
deformation calorimetry 7
deformation energy 68
degree of reaction 122,164,169,175,
182,185,189,194,198,214
degree of crystallinity 236-238
dehydration 227,228
desmearing 126-140
- advanced 133-138
- cp-measurements 152,154
- heat flow rate 128
- measured curve 126
- numerical methods 130-133
- reaction curve 181,188,194
- temperature 127,128
- Tg-measurements 203,210
detection limit 246
devitrification 56,195,198,202,239
differential measurement (method) 9
Differential Scanning Calorimetry
(see DSC) 1
differential temperature 12
Differential Thermal Analysis (DTA)
2,259
Differential Thermometry (DT) 2
diffusivity, thermal 60
2,2-dimethyl-l ,3-dioxolan-4-yl)methyl
methacrylate 192
dimyristoylphosphatidylcholine 220
3,6-dioxaoctane-l,8-dithiol 176
disk-type DSC 10-13,278
- noise 12
- uncertainty 12, 13
drop calorimeter 3,271
DSC 1,7
- actual problems 5,6
- applications 4,147-244
- calibration 65-114
- characterization 245
- checklist 257
- combined types 25
- fluids 24
- heat flux 10-13,262,278
Subject Index
-
light-modulated 24
linearity 50, 130,248
modulated (see TMDSC) 27
noise 12,246
oscillating (see TMDSC)
performance 245
power compensation 17-25,48-50,
270
- pressure 13
- temperature-modulated 50-63,159,218,
238
- theory 31-63
- triple system 13
- types of 9-25
- uncertainty 12,13
DSC curve 115-146
- calculated 115
- check 116
- measured 115-119
- characteristic terms 116, 117
- desmearing 126-140
- evaluation 119-121
- influences 118, 119
- kinetic evaluation 168
- shape 118
DTA 2,259
- applications 4, 5
dual sample DSC 13
dynamic mode (see scanning mode) 26
293
excess
- heat flow rate 54
- quantities 232
exothermic events 20,36,55,74
extent of reaction 162,164,168
extrapolated peak completion temperature
70,117
extrapolated peak onset temperature
70-72,117
Eyring equation 171
F
G
gas calorimeter 3,274
Gibbs free energy 121
glass transition 57,161,195,200-219,
234
- characterization of glasses 213
- during polymerization 163,195,214
- nature 201
- modeling of relaxation phenomena
212
- phenomenology 200
- relaxation peaks 204, 216
glass transition temperature 200
- applications 212
- definition 203
- influence of molar masses 213
- of copolymers 215,216
- of crosslinked polymers 215
- of semicrystalline polymers 216
- thermodynamic (fictive) 206,207
Green's function 131,136,138,181,250
GUM method 253
Subject Index
294
H
harmonics 51
heat 1
- latent 54
- measured 86,91
- true 90
heat accumulating calorimeters 271
heat capacity (Cp ) 34,53,67,68,88, 120,
147-160
- apparent 55,57,141,142,145
- change 123,200,201
- complex 56,60,138,143-145
- copper 105
- imaginary part 56, 144
- magnitude 56, 143
- non-reversing 142
- phase 56,143
- real part 56, 144
- reversing 141
- sapphire 103, 104
- time-dependent 56,57
- vibrational 53,56
heat capacity (Cp ) measurement 6, 147-160
- classic procedure 148
- dual step method 153
- small step method 155
- sources of error 150
- step scan method 158
- with TMDSC 159
heat compensating calorimeters 264
heat convection 45
heat-exchanging calorimeters 278
heat flow 1
heat flow rate 12-20,19,52,31,34
- amplitude 53
- calibration 87-90,87-90
- excess 54
- measured 86-90
- non-reversing 142
- periodic part 53,55
- phase 53,55
- resolution (see noise) 246
- reversing 142
- true 12,15,19,47,87-90,119,129-133
- underlying part 53, 55
heat flux DSC 9-17, 31-48, 278
- cylinder-type measuring system 14-17
- disk-type measuring system 9,10-13,
31-48
- dual sample 13,
- equivalent electric circuit 32,39,42
- models 31-48
- modulated heating mode 27-30
- numerical simulation 44-48
Subject Index
-
295
- scanning 26,
- step-scan 30
- temperature-modulated 27-30,50-63
modification (see polymorphism) 219
modulated DSC (see temperaturemodulated DSC)
modulated temperature (see temperaturemodulated)
modulation of temperature
- frequency 51
- amplitude 51
- period 53
- phase 53,
N
o
3-omega (3w) calorimetry 7
operation mode 25-30,264
order ofreaction 169
oscillating DSC (see TMDSC)
overall rate laws 170,172,183,193
Ozawa-Flynn-Wall method 186
Subject Index
296
polymorphism 219-228
polystyrene (PS) 210
power compensation DSC 17-25,48-50,
270
- advantage 21
- controler 18,49
- fluids 24
- special types 22-25
pre-exponential factor 17l, 172, 186, 187,
213
Q
quality assurance 4, 66
quality control 233
- aging of plastic 235
- degree of crystallinity 234,236
- oxidative decomposition 235
- plastic blends 234
- reactive resins 235
- semicrystalline polymers 234
- thermal decomposition 235
quasi-isothermal mode 27,52
quenching 206,217
R
reference sample 32
relaxation process 56
repeatability 247,252
reproducibility 248,252
resistance (see thermal resistance)
resolution 250
result 145, 251
reversing
- heat capacity 54,141
- heat flow rate 142
Subject Index
297
S
safety aspects 5, 232
- substances 233
safety calorimeter 3,232
safety risk 233
sample heat flow rate 119
sample model 34,60
sample preparation
- cpmeasurement 150,154
- reaction 173
- Tg-measurements 204
sample properties 53, 58
sample temperature 38, 135, 136
sapphire 103, 104
sawtooth mode 29,51
semi-crystalline polymers 234,236,237
sensitivity 43,250
Sestak-Berggren equation 170
scanning mode 26
SI units 66
side reaction 163, 177
signal-to-noise ratio 247
silicate glass 201
simulation, numerical 44
simultaneous techniques 5
smearing 35,178,211,229
smectic phase 224, 228
Sourour-Kamal equation 170,183
stability investigations 235
standard 65
stationarity 63
steady state 32,37,42,48
step-scan mode 30,138,145,158
step response function 58, 136, 145
subscripts (table) XII
substance identification 219
superposition principle 53,130,131
surface energy 68
symbols (table) XI
symmetry 33,41,48,50,68,87
- check 85
systematic errors 251
T
temperature
- calibration 69-75,99
- correction 128, 135
- fixed points 100
- true 7l,72
- profIle in sample 136
temperature scale 69, 100
temperature-modulated mode
50-63,159,195,238
- calibration 107,110-114
27-30,
Subject Index
298
U
uncertainty 6,65,68,72,98,251-256
- black box method 253
- budget 253-256
- determination 251
- GUM method 253
- measurement 72,251
- temperature measurement 69, 72
- total 251
underlying signal 53, 55
vitrification 56,169,170,195,198,199,202,
239
Vyazovkin method 186
W
zeroline 50,116,119,120,124,126,129,
147,163
- repeatability 247
zinc 77