thermochimica acta

ELSEVIER
Thermochimica Acta 253 ( 1995) 183- 188

.'.
ii

Linearity and non-linearityin DSC: A critique on modulatedDSC *

Takeo Ozawa u'*. Katsuhiko Kanari b

Received8 August 1994;accepted23 August 1994

Abstract Recently a new interesting technique, modulated differential scanning calorimetry (MDSC), was published.This techniqueis basedon Fourier transformation,and its sound applicability depends linearity in thermal processes, on because Fourier transformationis the based on linearity or the superposition principle. If a process under observation is a non,linear process,the physical meaning of the Fourier transformation should be closely examined.The DSC output signalsof some processes, such as melting and reaction, have been revealedto be non-linear. Therefore,differentiation of linearity from nonJinearity in DSC thermal responses very important for modulatedDSC, and they are discussed this is in paper. In this discussion, methods to analyzeoscillating output signalsare also proposed. Keywords: Deconvolution; DSC; Dynamic DSC; Linearity;MDSC

Introduction Recently a new technique of heat-flux differential scanning calorimetry, hf-DSC, was set forth [,2]. In this new technique, temperature oscillation is added to linear heating, and oscillating thermal response in the form of the temperature difference is recorded. The thermal responseis analyzed by Fourier transformation, and the
* Corresponding author. * Presented at the Intemational and III Sino-Japanese Symposium on Thermal Measurements, Xi'an, 4-6 June 1994.

0040-6031/95/$09.50 Science B.V. All rights reserved O 1995- Elsevier ( 9 4 )0 2 0 8 8 - 4 s , s D /0 0 4 0 - 6 0 3 1

184

T. Ozawa. K. KanarilThermochimica Acta 253 (199t

183-188

response phase with the temperatureoscillation is thought to be causedby in reversible changeor equilibriumchangein the sample,while the response out of phasewith the temperature oscillationis postulatedto be non-reversible change the is it [,2]. Because Fourier transformation appliedto the response, is implicitly principle,holds in the response. assumed that lirtearity,or the superposition However,this is not alwaysclear,and non-linearresponses have beenobserved in DSC results [3]. In order to elucidate the applicability of this very interesting technique, is essential differentiate it to linear from non-linearresponse hf-DSC. in What can be done and what cannotbe done by this technique? answerto this The questionis strictly connectedwith the linearity and non-linearity of the process.In order to obtain,the correct answer,thermal responses, such as the baseline,peak area and peak shape,are considered theoreticallyand their linearitiesare examined by experimentalfacts. It is elucidatedthat linearity holds for the baselineand peak area, but thermal response sampletransformation is a non-linear process, the so by physical meaning of observationof the transformation behavior by this technique is not clear. However, the thermal responseby electrical heating, instead of the sample transformation, is a linear process.These theoretical considerationsare described in detail, and methods to analyze oscillating thermal response are proposedin this paper.

2. Linearity First of all, the mathematicaland physicalmeaningof linearity and nonJinearity should be made clear in relation to DSC. Linearity can be expressedby the equations f(x +y):f\x) +f(y) f(ax) : q71*1 (1) Q)

For DSC, x and y express quantitiesto be measuredand the function expresses its output signal. Eq. (l) is the basis of the superposition principle and Eq. (2) indicatesproportionality for quantitative measurement. Hitherto, two contradictorypostulations have beenmade.Tateno [4] postulated the applicabilityof a linear theory of measurement, DTA and he deconvoluted curvesto reactionrate curvesby using a transferfunction of the DTA apparatus. The transfer function corresponds f(u) for a pulsive u in Eq. (l), and the to deconvolutionis a procedurefor obtaining x + y +. . ., which occurs consecutively, from the observed f(x + y + ) by using the transfer function. This procedure deconvolution of wasmadeon the basisof linearityor the superposition principle.However,O'Neill [5] analyzed heat transferin power compensation the DSC (pc-DSC)and postulated that transformation the sample, of suchas melting, proceeds a heat-transfer-controlling by mechanism. Thesetwo contradictory postulates expressed are schematically Fig. l. Accordin ing to the linear theory of measurement, the output signal is deformed by

T. Ozawa, K. Kanarif Thermochimica Acta 253 (jgg5)

lE3_lgg

(al Quantity to be measured

(b)

o o

time
tri.irrl

,t

time Fig._ l. Schematic expressions of (a) linear theory of measurement and (b) heat-transfer-controlling mechanism. Reproduced lrom Ref. t3l with permission.

i&i 't
the transfer function as shown in Fig. la, and the peak height is proportional to the quantity to be measured. For "^u-pl., the DSC peak hefiht oimetting of a pure sample should be proportional to the sample mass, if [r*?ritv-t "rds in a melting process. However, o'Neill's postulate [5] can be expressed as in Fig. lb. The heat is transferred by temperature gradient and the sampre temperature is kept constant during the transformation, while the temperature of the heat source is controlled to increaselinearly. Therefore, the temperature gradient arso increaseslinearly and the rate of transformation is controlled by this linearly increasing temperature gradient, through which the heat is suppried. Thus the uioitio.rut "i.rgy'input due to the transformation is increased Iinearry in pc-DSC, and the aoaltonat temperature difference is similarly increasedlinearly in nr-nsc [3]. The ,"r* p."r. shape can be obtained by both types of DSC, because the heat transfer mechanism is the same [3]' Non-linearity is implicitly postulated by this heat transfer mechanrsmdue to the constant sample temperature during the transformation.

186

T. Ozawa, K. Kanarif ThermochimicaActa 253 (1995) 183,188

3. Experimental facts and theoretical considerations Experimental verification was made by one of the present authors [3]. With a pc-DSC, indium melting peaks were observed by changing the sample mass, and the results are reproduced in Fig. 2. Similar results were also obtained using DTA of a micro-sample. From these results we can conclude that the above heat-transfercontrolled melting proceedsin both types of DSC, and it can be concluded that the sample temperature is kept approximately constant during the melting. Thus, the temperature distribution within the sample can be neglected during the melting. However, a linear process was also observed with an hf-DSC [3]. Pulsive electrical heating was made by inserting an electrical heater into the sample cell, and some peaks were obtained by changing the current and duration. They were compared in relation with the current, and the peak height was proportional to the current. A peak of long duration was compared with a peak synthesized by superposition of successivepeaks of short duration, and excellent agreement was obtained between them. In other words, by adding f(x) and "f(y) we can get f(x + y), and this synthesized "f(x + y) is in excellent agreement with the f(x + y) observed for x * y. Thus it was concluded that linearity holds in the case of electrical heating. Similar results were obtained by using peaks obtained by infrared irradiation of the sample pan of a pc-DSC [3]. Thus linearity holds in both types of DSC, when

:'.

;{
..i-

*) >\

30s

t)

L< # li

-n

157.00c

time
Fig. 2. DSC curves of melting for high-purity indium lrom Ref. [3] with permission.

The broken line is for remelting.Reproduced

T. Ozav,a, K. KanarilTherntochintica Acta 253 (1995) 183,188

187

heat evolves or absorbs independently of the sample temperature. In this case the evolved or absorbed heat is consumed by the sample temperature change. For the baseline,it is clear that linearity holds, becauselinearity has been used in heat capacity measurementsby using the baseline shift. From a theoretical viewpoint, the temperature distribution in an hf-DSC apparatus was analyzed in detail for sample cells of cylindrical symmetry [6] and for sample cells of general geometry [7]. In the dynamic steady state at a constant heating rate, the temperature of every part around the sample and the reference material is increased linearly, so that the temperature gradient at any point is proportional to the heating rate and it is also proportional to the heat capacity of the part to which the heat is supplied from the particular point. Similarly, linearity holds in the peak area, because proportionality has been used in measurements of transformation heats. The linearity in this case was also theoretically considered for sample cells of cylindrical symmetry [6] and for sample cells of general geometry [7]. The following fundamental equation for heat conduction holds for the additional temperature, Z', due to a sample transformation in an hf-DSC apparatus

(3) * a, a, whereV, tr, c, p, t andLH are respectively Laplacianoperator,the thermal a

)"Y : rp I'

d7''

dLH

t.

)r'

a t: t n

(4)

where the thermophysical properties of the sample, i.e., c and p, are assumed to be constant during the transformation, and the integral on the left side corresponds to the peak area. Thus the integral equation for the peak area is mathematically equivalent to the equation for temperature distribution in the steady state, and linearity holds for the peak area, as it holds for the baseline.

4. Concluding remarks As seen in the above discussion, linearity holds for a baseline and peak area. Therefore, modulated DSC can be applied to heat capacity measurements, but applicability of this technique to melting behavior observation is not certainly proved. For this application, further considerations using a theoretical approach are necessary. For theoretical consideration of the baseline, an analytical approach can be applied. In this approach, it can be assumed that the temperature distribution within the sample is negligibly small, though linearity holds for the case of detectable temperature distribution in the sample. Therefore, a fundamental equa-

T. Ozawa. K. KanariiThermochimica Acta 253 (199,

183-188

tion for DTA derived by Vold [8] can be used and the temperature oscillation is introduced into the equation. Thus, the relation of a baseline with experimental parameters,such as heat transfer coefficients,heating rate, sample mass,etc., can be derived as an analytical solution. For examination of modulated DSC results on sample transformation, the above analytical approach cannot be applied, but computer simulation seems to be a powerful tool. For this approach another assumption can be introduced, i.e., constant temperature of the sample during the transformation. These approaches will give us valuable suggestions,and we will report them in the near future.

References (1993) D. andV.L. Hill,J. Therm. Anal..40 949. I M. Reading, Elliott and J. Anal.,40(1993) 931. [2] P.S.Gill, S.R.Sauerbrunn M. Reading, Therm. NetsuSokutei, (19^17) . 4 45. [3] T. Ozawa, Trans. Faraday Soc., (1966) 62 1885. [4] J. Tateno, Anal.Chem., (1964) 36 1238. t5l M.J. O'Neill, Bull.Chem. Soc. Jpn.,39 (1966) 2071. [6] T. Ozawa, and K. Kanari,Thermochim Acta,to be submitted. [7] T. Ozawa 21 683. [8] M.J. Vold,Anal.Chem., (1949)