Journal of Thermal Analysis and Calorimetry, Vol.

89 (2007) 1, 175–180

LABTERMO: METHODOLOGIES FOR THE CALCULATION OF THE

CORRECTED TEMPERATURE RISE IN ISOPERIBOL CALORIMETRY

L. M. N. B. F. Santos1*, M. T. Silva1, B. Schröder1 and L. Gomes2

1
Centro de Investigação em Química, Departamento de Química, Faculdade de Ciências da Universidade do Porto
Rua do Campo Alegre 687, 4169-007 Porto, Portugal
2
Faculdade de Ciências de Saúde, Universidade Fernando Pessoa, Rua Carlos da Maia 296, 4150-150 Porto, Portugal

In isoperibol temperature-rise calorimetry, different approaches can lead to the corrected temperature rise. The description of the
methodology, as well as an improvement in the approach to the accurate numerical calculation of the corrected temperature rise in
isoperibol calorimetry, based on the comparison between the exponential functional description of the initial and final periods and
an independent third order polynomial fitting, is presented. The latter is focused on the strategy of minimizing the result dependency
on the definition of the main period of the reaction. Furthermore, the dedicated software application Labtermo is introduced.

Keywords: calorimetry, corrected temperature rise, extrapolation method, heat exchange corrections, isoperibol calorimetry,
Labtermo, Regnault–Pfaundler method, true temperature rise

Introduction method of calculating the heat loss corrections for

small isoperibol-calorimeter reaction vessels.
Isoperibol temperature-rise calorimeters are those in The exact calculation of the corrected tempera-
which the temperature of the calorimeter varies during ture rise in isoperibol calorimetry is one of the key
the course of the experiment but the temperature of the factors for accuracy in high precision calorimetry. An
environment remains constant. Procedures for the cal- isoperibol temperature-rise calorimeter is designed to
culation of the corrections for energy exchange be- have a big and constant thermal resistance to the
tween an ideal isoperibol calorimeter and its surround- surroundings and an insignificant thermal gradient
ings through heat exchange, with a constant stirring inside the calorimetric vessel. The calorimetric vessel
power, a constant thermal modulus and a constant en- and its surroundings, at constant temperature, could
ergy equivalent, have been described by several au- be considered as a system where the physical heat
thors [1–12]. Cases in which stirring power changes exchange model is simplified to allow a more
due to rotations of the calorimetric bomb (or analogous accurate calculation of the corrected temperature rise
problems in solution-reaction calorimetry) have been (the real temperature rise in the calorimeter if the
analyzed by Wadsö [2] and Gunn [6]. Oetting [5] pro- process would occur under perfect adiabatic condi-
posed and described a procedure to calculate the true tions, without any extraneous thermal effects).
temperature rise of an isoperibol calorimeter using In this work, an ideal isoperibol temperature-rise
only the after-period of the obtained experimental val- calorimeter will be considered; it will be focused on
ues; this procedure was especially developed for drop the description of the different methodologies and on
calorimeters. A complementary treatment was pre- the improvements of the numerical calculations of the
sented by Vanderzee [9] where some additional pertur- corrected temperature rise, based on a fitting
bations for the calculation of the true energy exchange procedure, in order to decrease the dependency of the
(changes in the energy equivalent, steady or constant numerical calculation from an exact definition of the
parasitic processes or non-linear parasitic processes) interval for the main period.
were taken into account. King and Grover [10] as well
as West and Churney [11] described and analyzed the
application of a two-body model in isoperibol calorim- Calculation of the corrected temperature rise
etry in which the calorimeter is represented by two
thermally connected regions, one surrounding the In isoperibol temperature-rise calorimetry, a tempera-
other and exchanging heat with the constant tempera- ture (T)–time (t) curve is usually divided into three pe-
ture environment. Hansen et al. [12] presented a riods as shown in Fig. 1. In the main period (tb to tc),

* Author for correspondence: lbsantos@fc.up.pt

1388–6150/$20.00 Akadémiai Kiadó, Budapest, Hungary

© 2007 Akadémiai Kiadó, Budapest Springer, Dordrecht, The Netherlands
 SANTOS et al.

dT/dt=k(T¥–TJ)+k(TJ–T) (3)
or
dT/dt=k(T¥–T) (4)
Assuming u, k and TJ being constant, a condition
which is nearly maintained for calibrations and experi-
ments at similar conditions, the analytical integration of
Eq. (4) gives an exponential function that describes the
temperature in the calorimetric vessel, T(t), as a function
of the time, t, considering an initial point, (Ti, ti),
T(t)=T¥–(T¥–Ti)exp[–k(t–ti)] (5)
Considering both the rate of change of tempera-
Fig. 1 Schematic representation of a typical temperature–time ture in the mean time of the initial period, tmi, and
curve in isoperibol temperature-rise calorimetry final period, tmf,
the chemical or physical process under investigation (dT/dt)tmi=u+k(TJ–Ttmi) (6)
occurs; the initial (ta to tb) and final (tc to td) periods are
(dT/dt)tmf=u+k(TJ–Ttmf) (7)
ideally governed by the energy exchange between the
calorimetric system and its surroundings only. The rate the value of the thermal leakage modulus, the cooling
of change of temperature during the initial and final pe- constant k, could be derived from Eqs (6) and (7).
riods can be described by a single body-model having a (dT / dt ) t mi – (dT / dt ) t mf
geometrically uniform but time-dependent tempera- k= (8)
ture; it is exchanging heat with an environment at a Tt mf – Tt mi
uniform, constant temperature. A two-body model
The equilibrium temperature, T¥, is, after all,
should be considered in systems with significant inter-
derived from Eq. (4) using the derivative at any point
nal thermal gradients. Considering the simplest single
of the initial or final periods, tz.
body-model, the rate of change of temperature during
the initial and final periods can be assumed to be (dT / dt ) t z
T¥ = + Tt z (9)
described by Newton’s law of heat transfer, k
dT/dt=u+k(TJ–T) (1) In order to minimize the interpolation errors of
where u – rate of change of temperature due to all the derivative, (dT / dt ) t z , and temperature, Tt z , the
constant thermal effects, like stirring, evaporation, time, tz, is chosen as the mean time of the initial or
self-heating of the temperature sensor, etc., T – tem- final periods, tmi or tmf.
perature, t – time, TJ – effective jacket temperature The corrected temperature rise, DTad, occurring
and k – thermal leakage modulus (overall heat-trans- during the main period, can be calculated as:
fer coefficient to the surroundings divided by the total DTad = Tt c – Tt b – DTexc (10)
heat equivalent of the calorimeter).
In isoperibol calorimetric systems, the thermal where Tt c and Tt b are the temperatures at the end (tc)
leakage modulus is usually small (a high total heat and the beginning (tb) of the main period, respectively,
capacity and a small overall thermal conductivity to and DTexc corresponds to the temperature change (dur-
the surroundings). ing the main period) which occurs due to the heat ex-
Considering the Eq. (1) for the description of the change and due to the constant thermal power within
rate of temperature change, term u represents the rate the calorimeter. DTexc can be calculated by integration
of change of temperature due to all constant thermal ef- of Eq. (4) over the time interval tb to tc, provided that,
fects and can be derived from the condition of a pseudo according to the previous assumption, u, k and TJ re-
thermal equilibrium, in which the heat exchange flow main constant during the calorimetric experiment.
balance is reached, and where dT/dt=0: tc tc

DTexc = ò (dT / dt )dt = k ò (T¥ – T )dt (11)

u=k(T¥–TJ) (2) tb tb

where T¥ is the convergence temperature when the tc

heat flow from the calorimetric vessel to the sur- DTad = Tt c – Tt b – k ò (T¥ – T )dt (12)
roundings equals the heat flow due to all constant tb
thermal effects inside the vessel. The substitution of
u, in Eq. (1), yields

176 J. Therm. Anal. Cal., 89, 2007

 CORRECTED TEMPERATURE RISE IN ISOPERIBOL CALORIMETRY

Calculation methodologies Calculation procedures

Definition of the time intervals In isoperibol calorimetry, the corrected temperature
rise, DTad, is usually calculated following the proce-
The calculation of the corrected temperature rise is dures known as the Regnault–Pfaundler method
based on Eq. (12). To derive this equation, it was as- (integration method) [1] or by the extrapolation
sumed that tb and tc are congruent with the true main method in several variations [1, 2].
period interval. The previous assumption is in many The Regnault–Pfaundler method can essentially
cases the main source for errors in the calculation of be described as the calculation of DTad using a numeri-
DTad, due to the difficulty to define if the power effect cal solution of Eq. (12), including the numerical inte-
due to the process in the calorimeter has ceased, and, gration of the temperature change during the main pe-
as a consequence, when the calorimeter reaches a riod which occurs due to heat exchange, and the calcu-
pseudo steady state in the final period. lation of k and T¥, from Eqs (8) and (9), respectively.
The choice of time tb is not problematic, as it is The results obtained by the Regnault–Pfaundler
usually well known when the process starts; the prob- method are highly dependent on the k value. The cal-
lem arises with the fixing of the end time of the main culation of k from experimental data is very critical
period, tc. The time tc should be taken as soon as pos- when the change in dT/dt, between the initial and final
sible, with the guaranty, that the temperature–time period, is very small (experiments with a small DTad).
evolution is already following Eq. (1). A premature The definition of the main period is also very crucial
choice of the time tc is a known problem in high preci- as it defines the time interval for the numerical inte-
sion solution–reaction calorimetry; it becomes even gration of the temperature change due to the heat ex-
more critical in small calorimeters with small heat ca- change to the surroundings and due to the constant
pacities and large cooling constants [12, 13]. thermal power within the calorimeter.
If the time tc is headily chosen, the following The extrapolation method, as described by
consequences arise: Challoner et al. [15], is based on the extrapolation of
• wrong fitting of the exponential function which is the temperature rise from the fitted exponential func-
used to describe the initial and final periods; tion, Eq. (5), for the initial and final periods, Ti and Tf.
• systematical deviations in the initial and final tem- This extrapolation is done at the time tx, where the tem-
peratures of the main period; perature-time integration of the fitted equation, Ti and
• introduction of errors in the calculation of the T–t Tf from ta to tx and from tx to td, respectively, equals the
derivative, dT/dt, in the value of the thermal leakage overall temperature–time integration of the experimen-
modulus, k, and in the convergence temperature, T¥. tal curve as represented by the following equation:
td tx td

Fitting equations ò T (t )dt = ò Ti (t )dt + ò Tf (t )dt

ta ta tx
(13)

The exponential function (Eq. (5)) describes the temper-

ature in the calorimeter vessel, T(t), as a function of the This is equivalent to the schematic representa-
time t, assuming a simple one-body heat transfer model. tion on Fig. 2a, where the temperature–time lines
In most cases, the exponential description is closely ap- (A, B) are extrapolated to a common point in time,
proximated by quadratic equations, which are, in nu- (tx). The time (tx) is selected in a manner which ascer-
merical terms, easier to deal with. It is still common tains that the two shaded areas in the graph are equal.
practice to fit the initial and final periods to straight The corrected temperature rise, DTad, is calculated
lines, especially when the exponent, [–k(t–ti)], is small, from the final and initial temperatures at the time tx.
or in the case of low quality of the temperature–time DTad=Tf(tx)–Ti(tx) (14)
data. Due to the symmetry, the derivative obtained from
the fit of linear or quadratic equations at the mean of the
fitting period is very close to the derivative obtained at Numerical procedures
the same point from the Eq. (5). It can be analytically In order to minimize errors due to an unintentionally
proven [14] that, in a series of equally time spaced data, premature choice of the end time of the main period tc,
a least square fitting of a linear or quadratic function to the following procedure is proposed:
any series of data gives exactly the same derivative at
• Fitting of independent polynomial equations to the
the middle point of the interval. Linear or quadratic fit-
tings show an equal accuracy but the use of a cubic initial and final periods:
polynomial fitting results in a significant increase in the In the presented procedure we will consider the
accuracy of the interpolation of the temperature–time fitting to a third-order polynom, with the initial pe-
derivative, even when the mean time is considered. riod, ta to tb

J. Therm. Anal. Cal., 89, 2007 177

 SANTOS et al.

[T ( t )]i = A i + Bi t mi + C i t mi2 + D i t mi3

+(Bi + 2C i t mi + 3D i t mi2 ) / k (19)
–[( Bi + 2C i t mi + 3D i t mi2 ) / k ]exp[– k ( t – t mi )]
and at the mean time of the final period, T(tmf)
[T ( t )]f = A f + Bf t mf + C f t mf
2 3
+ D f t mf
2
+(Bf + 2C f t mf + 3D f t mf )/k (20)
2
–[(Bf + 2C f t mf + 3D f t mf ) / k ]exp[– k ( t – t mf )]

Implementation of the numerical procedure

The described methodology procedures were imple-
mented in the windows software application Labtermo
[16]. Labtermo performs the calculation of the corrected
temperature rise from results obtained in isoperibol cal-
orimeters using the following procedures.

Regnault–Pfaundler method (integration method) (RP)

Fig. 2 Schematic temperature-time curve in isoperibol calorime-
• step 1 – Polynomial or exponential fitting of the
try: a – the temperature–time lines (A, B) are extrapolated
to a common point in time, (tx). The time (tx) is selected in initial and final periods, Eqs (15) and (16);
a way which will make the two shaded areas in the graph • step 2 – Calculation of k, from Eq. (17);
equally sized; b – fitting extrapolations in vicinity of the • step 3 – Numerical integration of the temperature
main period region (exponential and polynomial) exchange correction, DT exc , Eq. (11);
• step 4 – Calculation of T(tb) and T(tc) from Eqs (19)
[T(t)]ci=Ai+Bi t+Ci t2+Di t3 (15) and (20), respectively;
• step 5 – Calculation of DTad, from Eq. (10), using
and the final period, tc to td the data obtained in steps 3 and 4.
[T(t)]cf=Af+Bft+Cft2+Dft3 (16) Optionally, the thermal leakage modulus k can
Alternatively, an exponential function, in the be pre-defined. This feature offers the possibility of
form of Eq. (5), could be iteratively fitted. using a k-value obtained from other experiments, or
an averaged k-value which is typical for the system.
• Utilizing the analytical derivatives of the Eqs (15)
and (16) at the mean times of the initial and final
periods, tmi and tmf, respectively, to derive the ana- Extrapolation method (EM)
lytical solution for the thermal leakage modulus, k, • step 1 – Polynomial or exponential fitting of the
based on Eq. (8); initial and final periods, Eqs (15) and (16);
Bi – Bf + 2Citmi – 2Cf tmf + 3Ditmi 2 2
– 3Df tmf • step 2 – Calculation of k, Eq. (17);
k= 2 2 3 3
(17) • step 3 – Numerical iterative calculation of tx, where
Af – Ai + Bf tmf – Bitmi + Cf tmf – Citmi + Df tmf – Ditmi
the temperature–time integration of the derived
• Calculation of the convergence temperature, T¥, Eqs (19) and (20), [T(t)]i and [T(t)]f, from ta to tx and
from Eq. (9), using the analytical derivative of the from tx to td, respectively, equals the overall tempera-
temperature in the mean time of the initial period, ture–time integration of the experimental curve data.
T(tmi), and the interpolated temperature from • step 4 – Calculation of DTad from Eq. (14), using
Eq. (15) and the value of k, from Eq. (17). [T(tx)]i and [T(tx)]f calculated from Eqs (19) and (20).
T¥ = A i + Bi t mi + C i t mi2 + D i t mi3 + (18)
In order to evaluate and compare the proposed
methodology with the standard methods, the described
+(Bi + 2C i t mi + 3D i t mi2 ) / k
procedures were applied to the calculation of a typical
• Derivation of the analytical solutions of the function exothermic experimental data set. The calculations
that describes the initial and final periods, Eq. (5), were done for the same experimental temperature vs.
using the previous values of k and T¥ and anchoring time data set, based on an independent fitting of the ini-
the exponential functions at the interpolated temper- tial and main period to a first, second and third order
atures at the mean time of the initial period T(tmi) polynom, using a fixed size time interval for the initial

178 J. Therm. Anal. Cal., 89, 2007

 CORRECTED TEMPERATURE RISE IN ISOPERIBOL CALORIMETRY

and final periods (2000 s) and changing the time inter- be considered. From the results acquired for the dif-
vals for the main period (tb to tc) A (2990–4240 s); ference between the temperature at the time tc, T(tc),
B (2990–4490 s) and C (2990–4740 s). The main pe- Tc (exponential)–Tc (poly), one can conclude that the
riod, A (tb=2990 to tc=4240 s), was intentionally cho- chosen main period A is too short as the significant
sen too short in order to exemplify how the difference decrease of the difference, Tc (exponential)–Tc (poly),
between the temperature at the initial time of the final reveals, while facing an increase of the time tc
period, tc, obtained from the polynomial fitted Eq. (16) (intervals B and C).
and the exponential Eq. (20), can be used as a criterion While considering both the Regnault–Pfaundler
to evaluate when the calorimeter reaches a pseudo and the extrapolation method, the use of the exponen-
steady state in the final period. tial functions [Eqs (19) and (20)] results in a higher
Table 1 lists a summary of the results for the nu- reliability in the calculation of DTad, as compared to
merical treatment of a typical exothermic experimen- the use of the polynomial fitting functions [Eqs (15)
tal data set (data obtained in a high-precision and (16)], which holds for both methods.
isoperibol combustion calorimetry experiment). The Besides the observed small differences in DTad
presented results indicate that in the case of the con- attained via second and third order fittings, the results
sidered experimental data set, a linear fit is clearly not obtained for both methods are in good agreement
adequate to be simultaneously used in high-precision (considering ±5×10–5 K as the experimental tempera-
isoperibol calorimetry when long time periods have to ture resolution). Third order fittings result in more ac-
Table 1 Summary of the results for the numerical treatment of a typical experimental data set (exothermic high-precision
isoperibol combustion calorimetry), obtained using different sizes of the ‘main period’ A, B and C. ‘initial period’:
(2000 s); ‘final period’: (2000 s)
A B C
Experimental ‘main period’ (tb:tc)/s 2990:4240 2990:4490 2990:4740
k/s–1 0.00010756 0.00010772 0.00010828
T¥/°C 25.27656 25.27582 25.27452
Tb(poly)/°C 25.00181 25.00181 25.00181
Tb(exponential)/°C 25.00009 25.00009 25.00009
Tc(poly)/°C 25.34671 25.34488 25.34320
Linear fit Tc(exponential)/°C 25.34709 25.34525 25.34357
Tc(exponential)–Tc(poly)/°C 0.00038 0.00037 0.00037
DTad/K RP (poly) 0.34766 0.34780 0.34826
RP (exponential) 0.34976 0.34990 0.35036
EM (poly) 0.34573 0.34548 0.34554
EM (exponential) 0.34983 0.35001 0.35059
k/s–1 0.00010776 0.00010800 0.00010803
T¥/°C 25.27659 25.27590 25.27582
Tb(poly)/°C 25.00061 25.00061 25.00061
Tb(exponential)/°C 25.00068 25.00068 25.00068
Tc(poly)/°C 25.34682 25.34507 25.34325
2nd order fit Tc(exponential)/°C 25.34700 25.34511 25.34328
Tc(exponential)–Tc(poly)/°C 0.00018 0.00004 0.00003
DTad/K RP (poly) 0.34897 0.34921 0.34925
RP (exponential) 0.34908 0.34918 0.34921
EM (poly) 0.34842 0.34902 0.34902
EM (exponential) 0.34913 0.34922 0.34925
k/s–1 0.00010814 0.00010799 0.00010800
T¥/°C 25.27533 25.27576 25.27572
Tb(poly)/°C 25.00059 25.00059 25.00059
Tb(exponential)/°C 25.00063 25.00063 25.00063
Tc(poly)/°C 25.34671 25.34510 25.34329
3rd order fit Tc(exponential)/°C 25.34716 25.34518 25.34333
Tc(exponential)–Tc(poly)/°C 0.00045 0.00008 0.00004
DTad/K RP (poly) 0.34906 0.34928 0.34933
RP (exponential) 0.34947 0.34932 0.34933
EM (poly) 0.34656 0.34873 0.34873
EM (exponential) 0.34965 0.34937 0.34938
In the table, the symbols have the same meaning as described in the text.
RP (poly) and RP (exponential) – Regnault–Pfaundler procedure applying T(tb) and T(tc) as derived from the polynomial fit
[Eqs (15) and (16)] or the exponential function [Eqs (19) and (20)], respectively.
EM (poly) and EM (exponential) – extrapolation method procedure applying T(tb) and T(tc) as derived from the polynomial fit
[Eqs (15) and (16)] or the exponential function [Eqs (19) and (20)], respectively.

J. Therm. Anal. Cal., 89, 2007 179

 SANTOS et al.

curate interpolations of temperature of both mean ternative to other methods for the accurate calculation
times, T(tmi) and T(tmf), and the corresponding tem- of the corrected temperature-rise in high precision,
perature–time derivatives, when compared with a nearly ideal isoperibol calorimetry.
second order fitting interpolation.
As in any established methodology for the calcu-
lation of DTad, it is mandatory to use the same proce- Acknowledgements
dure in the calibration and the experiment itself, in or- Thanks are due to the Funda¸±o para a CiÃncia e Tecnologia,
der to partially cancel out the particular inherent Lisboa, Portugal, for the financial support through Centro de
numerical deficiencies. Investiga¸±o em Química da Universidade do Porto (Linha 5)
and to AgÃncia de Inova¸±o S. A. for the financial support to the
R&D project ‘CMFORNOS’. Bernd Schröder is grateful to FCT
Final remarks and the European Social Fund (ESF) under the 3rd Community
Support Framework (CSF) for the award of a Post-Doc scholar-
A new methodology for the accurate calculation of ship (BPD 35131/2000).
the corrected temperature rise in isoperibol calorime- Labtermo is a freeware application for educational,
try is proposed, based on the polynomial fitting of the R&D, and academic use.
initial and final periods, followed by the use of the ex- The temperature–time experimental data set file used as
ponential function derived from the analytical deriva- test sample is also available by request.
Supplementary information and software contact:
tives of the fitted cubic polynomial equation, in order
Luís M. N. B. F. Santos; e-mail: lbsantos@fc.up.pt.
to decrease the dependency of the results on the defi-
nition of the main period. The advantages of the
proposed procedure are the following ones: References
• The fittings of the Eqs (15) and (16) are independ-
1 H. C. Dickinson, Bull. Natl. Bur. Stand. U.S., 11 (1914) 189.
ently done, with a highest number of fitting param-
2 I. Wadsö, Sci. Tools, 13 (1966) 33.
eters, and, as a consequence, with an increase in 3 J. Coops, R. S. Jessup and K. van Nes, in F. D. Rossini
flexibility of the fitting even in the incidental case (Ed.), Experimental Thermochemistry, Vol. 1: Chap. 3,
when the time tc was prematurely chosen; Interscience, New York 1956.
• At the mean time of the initial and final periods of 4 L. D. Hansen and R. M. Hart, Thermochim. Acta,
the fitted cubic Eqs (15) and (16), both the interpo- 417 (2004) 257.
lated temperature, T(tm), and the value of the deriv- 5 F. L. Oetting, J. Chem. Thermodyn., 2 (1970) 727.
ative, {dT/dt}tm, are expected to be very accurate; 6 S. R. Gunn, J. Chem. Thermodyn., 3 (1971) 19.
• As a consequence of the previous considerations, a 7 S. Sunner, in S. Sunner and M. Mansson (Eds),
Experimental Chemical Thermodynamics, Vol. 1: Chap. 2,
reasonably accurate calculation of k and T¥, from
Pergamon Press, Oxford 1979.
Eqs (17) and (18), is achieved; 8 C. Mosselman and K. L. Churney, in S. Sunner, M. Mansson
• The functions in Eqs (19) and (20) and the value of (Eds), Experimental Chemical Thermodynamics,
k are less sensitive to the time tc; Vol. 1: Chap. 3, Pergamon Press, Oxford 1979.
• A minimal difference between the temperature at 9 C. E. Vanderzee, J. Chem. Thermodyn., 13 (1981) 1139.
the initial time of the final period, tc, obtained from 10 A. King and H. Grover, J. Appl. Phys., 12 (1941) 557.
the polynomial fitted Eq. (16) and the exponential 11 E. D. West and K. I. Churney, J. Appl. Phys., 39 (1968) 4206.
Eq. (20), can be used as a criterion to evaluate 12 L. D. Hansen, T. E. Jensen, S. Mayne, D. J. Eatough,
when the calorimeter reaches a pseudo steady state R. M. Izatt and J. J. Christensen, J. Chem. Thermodyn.,
7 (1975) 919.
in the final period; as shown in Fig. 2b.
13 M. Sakiyama and T. Kiyobayashi, J. Chem. Thermodyn.,
• The deviations of the experimental temperature–time 32 (2000) 269.
data from the derived exponential Eqs (19) and (20) 14 E. Kryszig, Advanced Engineering Mathematics,
can be used to evaluate the quality of the numerical Chap. 18, 8th Ed. Wiley, New York 1999.
calculation and the existence of some anomalous per- 15 A. R. Challoner, H. A. Gundry and A. R. Meetham,
turbations in the experimental data. Philos. Trans. R. Soc. London, A247 (1955) 553.
A criterion to evaluate the moment, after which 16 L. M. N. B. F. Santos, Ph.D. Thesis, University of Porto,
Porto 1995.
the power effect due to the process under investiga-
tion can be considered as being completed, is pro-
posed. This is based on the analysis of the difference Received: January 16, 2006
between the temperatures at the initial time of the fi- Accepted: June 28, 2006
nal period, tc, obtained from the fitted polynomial OnlineFirst: February 13, 2007
equation, and the derived exponential functional de-
scription. The proposed methodology is a robust al- DOI: 10.1007/s10973-006-7509-2

180 J. Therm. Anal. Cal., 89, 2007