See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/243957490

Kissinger kinetic analysis of data obtained under different heating schedules

Article  in  Journal of Thermal Analysis and Calorimetry · November 2008

DOI: 10.1007/s10973-008-9200-2

CITATIONS READS

71 269

3 authors:

Pedro E. Sánchez-Jiménez José M. Criado

Instituto de ciencia de Materiales de Sevilla Spanish National Research Council
99 PUBLICATIONS   2,234 CITATIONS    265 PUBLICATIONS   8,823 CITATIONS   

SEE PROFILE SEE PROFILE

Luis A Pérez-Maqueda
Spanish National Research Council
195 PUBLICATIONS   6,454 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Kinetics of Solid-State Reaction: A Continuous Challenge View project

Fondecyt Chile:CREEP OF DISPERSION-STRENGTHENED COPPER BASE ALLOYS View project

All content following this page was uploaded by Pedro E. Sánchez-Jiménez on 22 May 2014.

The user has requested enhancement of the downloaded file.

 Journal of Thermal Analysis and Calorimetry, Vol. 94 (2008) 2, 427–432

KISSINGER KINETIC ANALYSIS OF DATA OBTAINED UNDER

DIFFERENT HEATING SCHEDULES

P. E. Sánchez-Jiménez, J. M. Criado and L. A. Pérez-Maqueda*

Instituto de Ciencia de Materiales de Sevilla (CSIC-Universidad de Sevilla), C. Americo Vespucio 49, 41092 Sevilla, Spain

The dynamic heating rate method developed by TA Instruments (Hi-Res™) is a kind of sample controlled thermal analysis in which
a linear relationship between the logarithm of the heating rate and the rate of mass change is imposed. It is shown in this paper that
the reacted fraction at the maximum reaction rate strongly depends on the parameters selected for the Hi-Res heating algorithm,
what invalidates the use of the Kissinger method for analysing Hi-Res data unless that the reaction fits a first order kinetic law. Only
in this latter case, it has been demonstrated that it is not required that a constant value of the reacted fraction at the maximum reac-
tion rate is fulfilled for determining the activation energy from the Kissinger method. In such a case the Kissinger plot gives the real
activation energy, independently of both the heating schedule used and the value of the reacted fraction, am, at the maximum.

Keywords: Hi-Res, kinetics, Kissinger method

Introduction for overlapping processes as compared with conven-

tional linear heating rate TG [2]. As in dynamic
In sample controlled thermal analysis (SCTA) meth- heating rate, neither the evolution of the reaction rate
ods, the evolution of the temperature is determined by nor the evolution of temperature is known, it makes
the reaction progress through a feedback system useless the conventional kinetic analysis methods
[1–3]. Thus, unlike conventional methods were the proposed in literature for experimental data obtained
temperature follows a predefined function of time, under a constant heating rate. The Kissinger method
usually linear increase with time or isothermal, in the [15] is extensively used in literature for determining
SCTA method the evolution of temperature is not pre- the activation energy of experimental data obtained
determined because it depends on the evolution of the under linear heating rate conditions. Thus, more than
sample with the temperature. The SCTA methods re- 2000 citations can be found for the original publica-
quire monitoring the evolution of the sample. For tion, many of them in this journal [16–25]. Recently,
thermogravimetry, the mass loss or mass gain (TG the Kissinger method has been extended [26–30] to
signal) is a direct measure of the reacted fraction, the kinetic analysis of Hi-Res data. The scope of this
while the DTG signal gives a direct measure of the re- paper is to carry out a critical study of the generaliza-
action rate [4, 5]. The most conventional SCTA tion of the Kissinger method for determining the acti-
method is the constant rate thermal analysis (CRTA) vation energy from data obtained under heating
where the reaction rate is maintained constant at a schedules different than those assumed in the devel-
value previously selected. The kinetic analysis of opment of this method.
solid-state reactions by means of CRTA has been ex-
tensively described in literature [6–12]. A different
approach to SCTA is the High Resolution TG Theoretical
(Hi-Res) also known as dynamic heating rate method.
The Hi-Res is a combined rate and temperature con- For solid-state transformations, the reaction rate can
trolled thermal analysis method where neither the be written in the form:
temperature nor the reaction rate follows a predefined da
function, but the heating rate is influenced by the evo- = Ae –E / RT f (a ) (1)
lution of the reaction rate through an algorithm. Thus dt
the heating rate varies in a range from a maximum to being a the reacted fraction, A the preexponential fac-
minimum selected values depending on the evolution tor of Arrhenius, E the activation energy, T the abso-
of the reaction rate [2, 13, 14]. It has been claimed lute temperature, R is the gas constant and f(a) is the
that this method provides a better resolution power reaction kinetic model [31, 32]. Under a constant

* Author for correspondence: maqueda@cica.es

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

© 2008 Akadémiai Kiadó, Budapest Springer, Dordrecht, The Netherlands
 SÁNCHEZ-JIMÉNEZ et al.

heating rate, b, Eq. (1) would be written in the follow- The expressions of the function f(a) and f¢(a) for
ing form: the kinetic models used for describing solid-state re-
da A –E / RT actions are shown in Table 1. It is clear that in the case
= e f (a ) (2) of a first order reaction (F1 model), f¢(a)= –1 and
dT b Eq. (5) becomes
The Kissinger method [15] is based on the study æb ö AR E
of the rate equation at the maximum reaction rate. At lnçç m2 ÷ = ln - (6)
÷
this point d2a/dt2 is equal to zero and, thus, we get è Tm ø E RTm
from Eq. (1): Equation (6) allows to conclude that, provided
d 2a æ Eb m öæ da ö that a first order reaction is involved and bm>0, the plot
=çç ÷ç
dt 2 2
+ Af ¢ (a m )e –E / RTm ÷è dt ÷ø =0 (3) of the left hand side of this equation as a function of
è m
RT ø m
1/T would lead to a straight line whose slope gives the
where Tm, am and (da/dt)m are the temperature, re- activation energy, independently of both the heating
acted fraction, and reaction rate at the maximum schedule used for reaching the maximum reaction rate
while bm is the value of dT/dt at the point in which the and the value of the reacted fraction, am, at this point.
maximum reaction rate is reached. bm would be con- If the reaction does not follow a first order ki-
stant and equal to b only in the case that the experi- netic model, the slope of the plot of lnb m /Tm2 vs. 1/T,
ments were recorded at a constant heating rate as was according with Eq. (7), would lead to the activation
it formerly assumed by Kissinger. It follows from energy only in the case that am were independent of
Eq. (3) the heating schedule used. It has been shown in litera-
ture [33–41] that if the a–T plots are recorded under a
Eb m constant heating rate b (i.e.; bm=b), the value of am is
2
=- Af ¢(a m )e –E / RTm (4)
RT m nearly independent of the value of the heating rate.
Criado an Ortega [36, 37, 39] and Budrugeac and
Equation (4) can be rearranged after taking loga- Segal [41] have shown that exists a dependence be-
rithms into the Kissinger equation: tween the values of am of the different kinetic model
æb ö æ – AR ö E quoted in Table 1 resulting from linear heating exper-
lnçç m2 ÷ = lnç
÷ è E f ¢ (a m ) ÷ø - RT (5) iments and the actual value of E/RT, although it has
è Tm ø m
been clearly demonstrated that the error in the deter-
mination of the activation energy from the Kissinger
plot is lower than 5% for values of E/RT higher than
Table 1 f(a) kinetic functions and their first derivatives
10. It is noteworthy to remark that under hyperbolic
or logarithmic heating schedules [38] the value of am
Symbol f(a) f¢(a) is dependent on the kinetic law obeyed by the reaction
but independent of the E/RT value.
-1
R2 (1-a )1/ 2 Under dynamic heating rate (Hi-Res) conditions,
2(1- a )1/ 2 heating rate and reaction rate are correlated
-2
[13, 14, 42] by the following equation
R3 (1-a )2 / 3
3(1- a )1/ 3
S
æ 4R ö
lnb = lnbS -rateçç ÷÷ (7)
è 256 ø
F1 (1-a ) –1
being b the heating rate at the time t; bS the starting
heating rate selected by the user that is decreasing as
n(1- a )[ - ln(1- a )]1–1/ n nln(1- a )+ n-1 far as the rate (expressed in min–1) increases; R a pa-
An
[ - ln(1- a )]1/ n rameter selected by the user and usually called resolu-
tion and S a sensitivity parameter that generally is
1 -1 fixed at 1. Equation (7) can be also written in the
D2
[ - ln(1-a )] [ln(1- a )]2 (1- a ) form:
da
D3
3(1- a )2 / 3 1/ 2- (1- a )–1/ 3 lnb = lnbS -3910
. × –3 ×4 R (8)
2[1- (1- a )1/ 3 ] [1- (1- a )1/ 3 ]2 dt
or
3 - (1- a )–4 / 3
D4
2[(1- a )–1/ 3 -1] 2[(1- a )–1/ 3 -1]2

428 J. Therm. Anal. Cal., 94, 2008

 DATA OBTAINED UNDER DIFFERENT HEATING SCHEDULES

dT da model by assuming the same kinetic parameters used

ln = lnbS -3910
. × –3 ×4 R (9)
dt dt for simulating the curves included in Fig. 1. The val-
ues of am and Tm corresponding to these curves are
Salin and Seferis [30] have concluded that the shown in Table 2. It can be observed that in this case,
Kissinger method can be extended to the determina- contrarily to what occurs in the case of Hi-Res experi-
tion of the activation energy from experimental data ments, the value of am is nearly independent of the
obtained under the heating schedule expressed by heating rate as it is well established in literature. On
Eq. (7) [or the equivalent Eqs (8) or (9)]. This method the other hand, Fig. 3 shows that the plot of the data
has been often used for the kinetic analysis of experi- included in Table 2 according with Eq. (6) leads to a
mental data obtained under Hi-Res conditions by a straight line with a correlation coefficient equal to
number of authors [26–30]. However, it must be 1.000 and an activation energy E=200 kJ mol–1 in ex-
pointed out that Salin and Seferis [12] did not demon- cellent agreement with the theoretical value. These
strated if am is independent of the experimental pa- results confirm that the use of the Kissinger method
rameters selected in the dynamic heating rate algo- for determining the activation energy of first order re-
rithm as demanded by the Kissinger method. Thus, actions from rising temperature experiments does not
the validity of this method must be reanalysed. This is
one of the scopes of this work.

Results and discussion

It was mentioned above, that values of am should be

identical for the different curves to be analyzed by the
Kissinger method. We have tried to resolve the sys-
tem constituted by the differential Eqs (1), (3) and (9)
and we have not found an analytical solution for am,
what suggests that this parameter is not independent
of the bS and R values selected for recording the
Hi-Res experiments. In order to confirm if the values
of am are dependent on the experimental conditions
selected for the dynamic heating rate experiments, the
set of curves shown in Fig. 1 have been simulated by Fig. 1 Curves simulated by assuming Hi-Res heating schedule
assuming a F1 kinetic model, dynamic heating rate a F1 kinetic model and the following kinetic parame-
conditions (bS=50 K min–1 and R=2, 3 and 5, respec- ters: E=200 kJ mol–1; A=1014 min–1. A value of bS=
tively) and the following kinetic parameters: 50 K min–1 and values of the resolution parameter R,
A=1014 min–1, E=200 kJ mol–1. These curves have equal to a – 2, b – 3 and c – 5, respectively, have been
been simulated solving, by numerical integration with assumed
the Mathcad software, the system constituted by the
differential Eqs (1) and (9). The values of am resulting
at the different values of R selected for the simulation
are included in Table 2 together with the correspond-
ing ones of bm and Tm. These results clearly show that
values of am are very much dependent on the condi-
tions selected for the high resolution experiment.
However, we must bear in mind that it was concluded
in the theoretical section that in the case of first order
reactions, Eq. (6) would be accomplished indepen-
dently of the heating schedule used for reaching the
maximum reaction rate and whatever would be the
value of the reacted fraction at the maximum. In order
to check this assertion, it has been considered of inter-
est to carry out the combined analysis of the Hi-Res
results included in Table 2 together with those result- Fig. 2 Curves simulated by assuming constant heating rates of
ing for a set of da/dt–T curves obtained at different a – 5, b – 20 and c – 50 K min –1, respectively, and the
constant heating rates. Figure 2 shows a set of curves same kinetic model, i.e. F1, and kinetic parameters,
simulated at different heating rates for a F1 kinetic i.e. E=200 kJ mol–1 and A=1014 min–1, as in Fig. 1

J. Therm. Anal. Cal., 94, 2008 429

 SÁNCHEZ-JIMÉNEZ et al.

depend on the profile of the heating schedule used for and R controlling the heating schedule. Figure 5 shows
achieving the maximum reaction rate because a con- that the plot of the values of ln(bm/Tm2 ) taken from
stancy of the value of the reacted fraction at the maxi- Table 3 as a function of the corresponding values of
mum is not required. 1/Tm does not lead to a good straight line as pointed out
The accurate determination of the activation by the poor regression coefficient found (r=0.9575).
energy by the Kissinger method for kinetic models Moreover, the apparent activation energy E=
different from F1 requires, according to Eq. (5), an 1319.7 kJ mol–1 obtained from the slope of this linear
identical value for a at the maximum reaction rate, am, plot is considerably higher than the value of
for all the analyzed curves. It is well known [33–41] that 200 kJ mol–1 assumed for simulating the curves in
this requirement is fulfilled under linear heating Fig. 4. These results suggest that the Kissinger method
program but this requirement is not accomplished under cannot be extended for determining the activation
Hi-Res heating program as shown in Table 2. Thus, it energy from Hi-Res rising temperature experiments
would be interesting to analyse the influence of the lack unless that the reaction obeys a F1 kinetic law.
of constancy of the reacted fraction at the maximum Figure 6 shows the set of da/dt–T plots simu-
reaction rate on the activation energy supplied by the lated by assuming different linear heating rates and
Kissinger equation when applied to analysis of Hi-Res the same kinetic model and kinetic parameters used
results of reactions that follow a kinetic model different for simulating the Hi-Res curves in Fig. 4. The values
from F1. Figure 4 shows a set of diagrams calculated of Tm and am corresponding to these curves are in-
with the Mathcad software by assuming an A2 kinetic cluded in Table 3. It can be observed that the value of
model, an activation energy E=200 kJ mol–1, a pre-
exponential factor A=1014 min–1 and dynamic heating
rate conditions (bS=50 K min–1 and R=2, 3, 4 and 5,
respectively). The values of bm, am and Tm resulting
from these curves are included in Table 3. It is clear
again that the value of the reacted fraction, am, at the
maximum is strongly dependent on the parameters bs
Table 2 Values of am, bm and Tm for the sets of curves in
Figs 1 and 2

am bm/ Tm/
Conditions
K min–1 K
b=5 K min–1 0.612 5 714.2
b=20 K min –1
0.610 20 743.0
b=50 K min–1 0.609 50 763.4 Fig. 4 Curves simulated for an A2 kinetic model by assuming
bS=50 K min Res=2
–1
0.501 14.7 736.5 the same kinetic parameters of Fig. 1, i.e. E=
bS=50 K min–1 Res=3 0.433 5.5 716.2 200 kJ mol–1, A=1014 min–1, Hi-Res heating with bS=50
and resolution parameter R, equal to a – 2, b – 3, c – 4
bS=50 K min Res=5
–1
0.327 0.5 671.8
and d – 5, respectively

Fig. 3 Kissinger plot of the values of ln(b m /Tm2 ) as a function Fig. 5 Kissinger plots of the values of ln(b m /Tm2 ) as a function
of 103/Tm calculated from Table 2 of 103/Tm calculated from Table 3

430 J. Therm. Anal. Cal., 94, 2008

 DATA OBTAINED UNDER DIFFERENT HEATING SCHEDULES

Table 3 Values of am, bm and Tm for the sets of curves in

Figs 4 and 6 Acknowledgements
am bm/ Tm/ Financial support from projects TEP-03002 from Junta de
Conditions
K min–1 K Andalucía and MAT 2008-06619/MAT from the Spanish
b=5 K min–1 0.62 5 715.0 Ministerio de Ciencia e Innovación is acknowledged.
b=10 K min–1 0.62 10 729.2
b=20 K min –1
0.62 20 743.9
References
b=50 K min–1 0.62 50 764.3
bS=50 K min–1 Res=2 0.50 5.6 737.8 1 L. A. Perez-Maqueda, P. E. Sanchez-Jimenez and
bS=50 K min–1 Res=3 0.43 0.6 721.9 J. M. Criado, Current Topics Catal., 6 (2007) 1.
2 J. Rouquerol and T. O. Sorensen, ‘Sample Controlled
bS=50 K min–1 Res=4 0.39 1.1·10–2 706.5
Thermal Analysis’, T. O. Sorensen and J. Rouquerol, Eds,
bS=50 K min Res=5
–1
0.33 2.7·10–6 692.3 Kluwer, Dordrecht 2003, p. 1.
3 J. Rouquerol, J. Therm. Anal. Cal., 72 (2003) 1081.
4 J. M. Criado, L. A. Perez-Maqueda, M. J. Dianez and
P. E. Sanchez-Jimenez, J. Therm. Anal. Cal.,
87 (2007) 297.
5 M. J. Dianez, L. A. Perez-Maqueda and J. M. Criado, Rev.
Sci. Instrum., 75 (2004) 2620.
6 P. A. Barnes, M. J. Tiernan and G. M. B. Parkes,
J. Therm. Anal. Cal., 56 (1999) 733.
7 S. Bordere, F. Rouquerol, J. Rouquerol, J. Estienne and
A. Floreancig, J. Thermal Anal., 36 (1990) 1651.
8 F. J. Gotor, L. A. Perez-Maqueda, A. Ortega and
J. M. Criado, J. Thermal Anal., 53 (1998) 389.
9 L. A. Perez-Maqueda, J. M. Criado and F. I. Gotor, Int. J.
Chem. Kinet., 34 (2002) 184.
10 L. A. Perez-Maqueda, A. Ortega and J. M. Criado,
Thermochim. Acta, 277 (1996) 165.
11 M. H. Stacey, Anal. Proc., 22 (1985) 242.
Fig. 6 Curves simulated by assuming constant heating rates of 12 M. J. Tiernan, P. A. Barnes and G. M. B. Parkes, J. Phys.
a – 5, b – 10, c – 20 and d – 50 K min –1, respectively, Chem. B, 103 (1999) 338.
an A2 kinetic model and the same kinetic parameters 13 P. S. Gill, S. R. Sauerbrunn and B. S. Crowe, J. Thermal
assumed in Fig. 1, i.e. E=200 kJ mol–1 and Anal., 38 (1992) 255.
A=1014 min–1 14 S. Sauerbrunn and P. Gill, Am. Lab., 26 (1994) 29.
15 H. E. Kissinger, Anal. Chem., 29 (1957) 1702.
am is practically independent of the heating rate as 16 L. Abate, I. Blanco, F. A. Bottino, G. Di Pasquale,
E. Fabbri, A. Orestano and A. Pollicino, J. Therm. Anal.
would be expected from linear rising temperature ex-
Cal., 91 (2008) 681.
periments. Thus, f¢(a) is constant and, according with 17 S. M. Ahmed and M. I. Abd-Elrhaman, J. Therm. Anal.
Eq. (4), the plot of ln(b/Tm2 ) vs. 1/Tm shown in Fig. 3 Cal., 91 (2008) 195.
leads to a straight line with a regression coefficient 18 M. V. Borrachero, J. Paya, M. Bonilla and J. Monzo,
r=0.99999 whose slope gives an activation energy J. Therm. Anal. Cal., 91 (2008) 503.
E=199.7 kJ mol–1 that matches the value used for the 19 A. d’Almeida, D. W. Barreto, V. Calado and
simulations. J. R. M. d’Almeida, J. Therm. Anal. Cal., 91 (2008) 405.
20 M. D. Fernandez and M. J. Fernandez, J. Therm. Anal.
Cal., 91 (2008) 447.
Conclusions 21 A. Kropidlowska, A. Rotaru, M. Strankowski, B. Becker
and E. Segal, J. Therm. Anal. Cal., 91 (2008) 903.
22 A. Révész, J. Therm. Anal. Cal., 91 (2008) 879.
It can be concluded that, unless a first order kinetic 23 J. M. Salla, X. Fernandez-Francos, X. Ramis, C. Mas,
law were involved, the Kissinger method cannot be A. Mantecon and A. Serra, J. Therm. Anal. Cal.,
directly applied to the kinetic analysis of rising tem- 91 (2008) 385.
perature experiments obtained under any heating 24 S. Vecchio, L. Campanella, A. Nuccilli and
schedules without previously checking that the values M. Tomassetti, J. Therm. Anal. Cal., 91 (2008) 51.
of reacted fraction at the maxima remain unchanged 25 J. Zhang, Y. Y. Liu, J. L. Zeng, F. Xu, L. X. Sun,
for all analyzed curves. This condition is fulfilled if W. S. You and Y. Sawada, J. Therm. Anal. Cal.,
the a–T plots are recorded under a linear heating rate 91 (2008) 861.
but not if these curves are obtained with a Hi-Res ris-
ing temperature schedule.

J. Therm. Anal. Cal., 94, 2008 431

 SÁNCHEZ-JIMÉNEZ et al.

26 C. Gamlin, M. G. Markovic, N. K. Dutta, 35 G. Gyulay and E. J. Greenhow, Thermochim. Acta,

N. R. Choudhury and J. G. Matisons, J. Therm. Anal. Cal., 6 (1973) 254.
59 (2000) 319. 36 J. M. Criado, R. Garciarojas and J. Morales, Thermochim.
27 A. J. Gu and G. Z. Liang, J. Appl. Polym. Sci., Acta, 25 (1978) 257.
89 (2003) 3594. 37 J. M. Criado and A. Ortega, J. Non-Cryst. Solids,
28 B. A. Howell and J. A. Ray, J. Therm. Anal. Cal., 87 (1986) 302.
83 (2006) 63. 38 J. M. Criado and A. Ortega, Thermochim. Acta,
29 X. G. Li and M. R. Huang, J. Appl. Polym. Sci., 103 (1986) 317.
71 (1999) 565. 39 J. M. Criado and A. Ortega, Acta Metal., 35 (1987) 1715.
30 I. M. Salin and J. C. Seferis, J. Appl. Polym. Sci., 40 D. Chen, X. Gao and D. Dollimore, Thermochim. Acta,
47 (1993) 847. 215 (1993) 109.
31 J. M. Criado and L. A. Perez-Maqueda, J. Therm. Anal. 41 P. Budrugeac and E. Segal, J. Therm. Anal. Cal.,
Cal., 80 (2005) 27. 88 (2007) 703.
32 J. M. Criado and L. A. Perez Maqueda, ‘Sample 42 J. Rouquerol, T. O. Sorensen, P. A. Barnes,
Controlled Thermal Analysis: Origin, Goals, Multiple E. L. Charsley, E. Fesenko and M. Reading, ‘Sample
Forms, Applications and Future’, O. T. Sorensen and Controlled Thermal Analysis’, T. O. Sorensen and
J. Rouquerol, Eds, Kluwer, Dordrecht 2003, p. 55. J. Rouquerol, Eds, Kluwer Academic Publishers,
33 J. H. Flynn and L. A. Wall, J. Res. Nat. Bur. Stand. (A), Dordrecht, Netherlands 2003, p. 16.
70 (1960) 487.
34 H. H. Horowitz and G. Metzger, Anal. Chem.,
35 (1963) 1464. DOI: 10.1007/s10973-008-9200-2

432 J. Therm. Anal. Cal., 94, 2008

View publication stats