Kinetic analysis of complex decomposition reactions using evolved gas analysis

Article abstract of DOI:10.1007/s10973-009-0026-3

For complex decomposition reactions, traditional methods, such as TG and DSC cannot fully resolve all of the steps in the reaction. Evolved gas analysis (EGA) offers another tool to provide more information about the decomposition mechanism. The decomposi

Full text of DOI:10.1007/s10973-009-0026-3

J Therm Anal Calorim (2009) 96:805–811
DOI 10.1007/s10973-009-0026-3
Kinetic analysis of complex decomposition reactions using evolved
gas analysis
John P. Sanders Æ Patrick K. Gallagher
NATAS2008 Conference
´
´
ꢀ Akademiai Kiado, Budapest, Hungary 2009
Abstract For complex decomposition reactions, tradi-
tional methods, such as TG and DSC cannot fully resolve all
of the steps in the reaction. Evolved gas analysis (EGA)
offers another tool to provide more information about the
decomposition mechanism. The decomposition of sodium
bicarbonate was studied by TG, DSC and EGA using a
simultaneous thermal analysis unit coupled to a FTIR. The
decomposition of sodium bicarbonate involves two reaction
products H2O and CO2, which are not evident from either TG
or DSC measurements alone. A comparison of the reaction
kinetics from TG, DTG and EGA data were compared.
product. EGA data offers the potential to simplify the anal-
ysis of complex decomposition reactions since the evolution
of various reaction products can be followed separately.
When working with EGA, there is a finite transfer time
which is the time that it takes for the gas evolved from the
sample to reach the gas analyzer. The transfer time is
influenced by the flow rate of the gases, volume of the gas
cell, transfer line and furnace, and the flow profile through
the system [1]. Directly coupled thermal analyzers utilizing
either Fourier transform infrared spectroscopy (FTIR) or
mass spectrometry (MS) which have been designed to
minimize the effect of transfer time have been described by
Kaisersberger and Post [1].
Keywords Evolved gas analysis ꢀ Kinetic analysis ꢀ
Simultaneous TG/DSC ꢀ Transfer time
In addition to instrumental design, several operating
parameters can influence the transfer time. Higher flow rates
reduce the transfer time but increase dilution [2]. The
detection limit of the EGA to the gases involved in the
reaction must also be considered. For FTIR analysis,
excessive dilution may reduce the minimum detection limit
for weakly absorbing species to an unacceptable level. Also
when using an FTIR, the spectral resolution and number of
co-added scans determine the time resolution of the evolved
gas data [2]. It is also important to note that calibration of
EGA data is required to obtain quantitative results [3]. In
most cases, calibration of the EGA signal is not required
since it is used in conjunction with Thermogravimetric (TG)
or Differential Scanning Calorimetry (DSC) data.
Introduction
This work was undertaken to determine if data from evolved
gas analysis (EGA) could be used for the kinetic analysis of
complex decomposition reactions. Traditionally, TG and
DSC have been used to study the kinetics of simple decom-
position reactions. When multiple decomposition products
are involved, complicated deconvolution procedures were
required to separate the contribution of each reaction
J. P. Sanders (&)
Various authors have used the difference between the
DTA or DTG peak temperature and the corresponding
evolved gas peak to quantify the transfer time [4, 5]. This
differential can be used to select experimental conditions
that minimize the effect of transfer time.
The National Brick Research Center, Clemson University,
100 Clemson Research Blvd, Anderson, SC 29625, USA
e-mail: jpsand@clemson.edu
P. K. Gallagher
Clemson University, Anderson, SC, USA
The decomposition of NaHCO3 has been studied by
several investigators [3, 6–10]. The reaction proceeds
according to the following path (Eq. 1) with the
P. K. Gallagher
The Ohio State University, Columbus, OH, USA
123

806
J. P. Sanders, P. K. Gallagher
simultaneous evolution of CO2 and H2O [6]. The decom-
position rate is dependent on the partial pressure of both
CO2 and H2O [7, 8]. In controlled rate EGA studies,
Yamada and Koga found that increasing the partial pres-
sure of CO2 retards the rate of reaction [7]. Conversely,
increasing the partial pressure of H2O appears to enhance
the reaction rate [7]. The decomposition rate is also
affected by self cooling due to the endothermic nature of
this reaction. Given the simultaneous evolution of CO2 and
H2O, sodium bicarbonate was selected to study the possi-
bility of using EGA data for kinetic analysis.
16 ꢁC min-1. A flow rate of 100 mL min-1 of nitrogen
was used for all experiments. Open alumina crucibles were
used in these measurements. The sample size ranged from
20 0.5 mg. The DSC sensitivity curve was determined
for each heating rate using a sapphire disk.
In this study, a resolution of 4 cm-1 was used for all
measurements. The number of co-added scans was reduced
for the faster heating rates to improve the time resolution of
the evolved gas data from the FTIR. These conditions,
along with the nitrogen flow rate were selected to minimize
the transfer time between the thermal analyzer and the
FTIR. The CO2 and H2O emission traces were determined
by integrating the FTIR absorption spectra from 2,450 to
2,200 cm-1 for CO2 and 2,200 to 1,600 cm-1 for H2O.
Netzsch Proteus and Thermokinetic software packages
were used for data analysis [12–14].
2NaHCO3ðsÞ ! Na2CO3ðsÞ þ CO2ðgÞ þ H2OðgÞ
ð1Þ
Values for the activation energy of this reaction range
from approximately 90 to 120 kJ mol-1 depending on the
experimental conditions and the computational techniques
used to model the experimental data [9, 10]. Several studies
have found that the decomposition kinetics can be
modelled adequately with a simple first order reaction,
while other investigators have applied more complicated
models [9–11]. The experimental conditions and
computational methods used to interpret the experimental
data have a profound effect on the determination of the
kinetic parameters [11].
Results and discussion
A typical decomposition profile is shown in Fig. 1 for a
sample heated in nitrogen at the intermediate heating rate
of 8 ꢁC min-1 with a flow rate of 100 mL min-1. The
decomposition of NaHCO3 was characterized by the
simultaneous emission of CO2 and H2O as noted previ-
ously [6]. Additionally, the reaction was characterized by a
large endotherm.
Experimental
To evaluate the effect of the flow rate on the transfer
time to the FTIR, this experiment was repeated at flow
rates of 75 and 50 mL min-1. The results of this compar-
ison are summarized in Tables 1 and 2. Table 1 shows the
effect of flow rate on the extrapolated onset temperature at
a heating rate of 8 ꢁC min-1 in the various flow rates.
Table 2 shows the effect of the flow rate on the peak
temperatures under the same conditions. For the TG data,
the inflection point was used in place of the peak
The sodium bicarbonate used in this study was a com-
mercial grade produced by Church and Dwight. The
material was screened to remove particles greater than
74 lm. Simultaneous TG/DSC measurements were made
on a Netzsch STA 449C which was coupled to a Brucker
Vector 22 FTIR. The details of the coupling of the thermal
analyzer and the FTIR have been described by Kaiserber-
ger and Post [1]. Simultaneous TG, DSC and EGA mea-
surements were obtained at heating rates of 1, 2, 4, 8, and
H2O
DSC /(mW/mg)
Fig. 1 Decomposition of
NaHCO3 (8 ꢁC min-1
TG /%
CO2
30
exo
↓
)
[3]
100
95
90
85
80
75
70
65
2.0
1.5
1.0
0.5
0
2.5
2.0
1.5
1.0
0.5
0.0
25
20
15
10
5
[3] Bicarb 8 31Mar08.dsu
TG
DSC
H2O
CO2
DTG
[3]
[3]
0
50
100
150
200
250
300
Temperature /°C
123

Kinetic analysis of complex decomposition reactions
807
appears to provide a minimum transfer time and was used
in all subsequent measurements. In addition to the flow
rate, the instrumental design and coupling system has been
optimized to reduce the effect of transfer time [1]. As
would be expected, flow rate had less of an effect on the
DSC data than the EGA data although the differences
between all of the data sets was very small.
Table 1 Effect of flow rate on extrapolated onset temperature at
8 ꢁC min-1
Flow rate
50 mL min-1 75 mL min-1 100 mL min-1
TG data
123.4
112.8
110.8
123.6
113.3
111.8
112.0
113.5
123.4
112.6
110.7
111.6
113.2
DSC data
DTG data
CO2 (EGA) data 112.3
H2O (EGA) data 113.8
The extrapolated onset temperatures as a function of
heating rate for each data set are shown in Fig. 2. As noted
previously, the extrapolated onset temperatures for the TG
data were higher than the other data sets and different from
the DTG curve. The other data sets including the EGA data
sets for CO2 and H2O were very similar. At the highest
heating rate, 16 ꢁC min-1, the extrapolated onset temper-
atures for the DSC, CO2 and H2O data sets were very
similar, but the onset for the DTG data began to diverge.
Similarly, the peak temperatures as a function of heating
rate for each data set are shown in Fig. 3. Less scatter was
Table 2 Effect of flow rate on peak temperature at 8 ꢁC min-1
Flow rate
50 mL min-1 75 mL min-1 100 mL min-1
TG dataa
DSC data
DTG data
146.2
149.3
146.2
146.2
148.8
146.2
149.7
149.7
145.6
148.7
145.6
148.7
148.7
CO2 (EGA) data 150.8
H2O (EGA) data 150.8
a
For the TG data, the inflection point was used
140
130
120
110
100
90
temperature in Table 2. No correction factors were
applied to the onset and peak temperatures reported in
Tables 1 and 2.
The extrapolated onset temperature for the TG data, was
higher than for the other simultaneously collected data sets.
The determination of the extrapolated onset temperature
for the TG data gives a different result than the other data
sets due to the shape of the TG curve relative to the other
data sets as seen in Fig. 1. This effect is especially apparent
when the onset temperature for the TG data and the
derivative of the TG data (DTG) are compared. This dif-
ference is due to the mathematical amplification of sensi-
tivity and noise made by differentiation. In other words, the
delay between the TG and DTG data is the result of the
calculation of the DTG from the TG data and is indepen-
dent of the experimental conditions. For the other data sets,
the delay between the decomposition event and the
detection of the event by the appropriate sensor is due to
the experimental conditions such as flow rate. The DSC,
DTG and EGA data for CO2 and H2O emission all resulted
in similar onset temperatures. In general, lower onset
temperatures were observed with increasing flow rate, as
would be expected, but the difference was very small.
Flow rate has a pronounced effect on the peak temper-
ature as demonstrated in Fig. 3. For the EGA data, a wider
range of peak temperatures was observed than for the TG,
DSC or DTG data. At the highest flow rate of
100 mL min-1, the peak temperature was similar for the
DSC and EGA data. The TG and related DTG data dis-
played the lowest peak temperatures. It is again important
to note that the inflection point was substituted for the peak
temperature with the TG data. The highest flow rate
0
2
4
6
8
10
12
14
16
18
Heating Rate (oCmin-1)
TG
DSC
DTG
CO2
H2O
Fig. 2 Comparison of onset temperatures
170
160
150
140
130
120
110
0
2
4
6
8
10
12
14
16
18
Heating Rate (oC min-1
)
TG
DSC
DTG
CO2
H2O
Fig. 3 Comparison of peak temperatures
123

808
J. P. Sanders, P. K. Gallagher
120
110
100
90
13
12
11
10
9
independence between the two terms. This correlation or
´
compensation effect was also noted by Jankovic and
´
Adnadevic in their study of the thermal decomposition of
¯
NaHCO3 [10].
The reaction kinetics for the TG, DTG, CO2 and H2O
data sets were also compared by fitting each data set to a
generic nth order, Fn, reaction model. This fitting was
accomplished, using the Netzsch Thermokinetics software
package, to allow a direct comparison between the data
sets. The data was fit over an a range of 0.0005–0.9995. A
higher level of fitting could have been achieved for each
data set using the non-linear regression capabilities of the
Thermokinetics software package, but this was not done to
allow for the comparison of the resulting kinetic parame-
ters using a single reaction model [12]. Further, allowing
multiple reaction steps, as recommended by Maciejewski
would naturally have resulted in a higher level of correla-
tion between the kinetic model and the experimental
data [11]. The significance of these improved values
with increasing variables always requires thoughtful
consideration.
8
80
7
70
6
0.0
0.2
0.4
0.6
Alpha (%)
0.8
1.0
Ea - TG Data
Ea - CO2 Data
Ea - H2O Data
Log A - TG Data
Log A - CO2 Data
Log A - H2O Data
Fig. 4 Ozawa Flynn Wall model free kinetic analysis comparison
(Ea—solid lines, Log A—dashed lines)
observed in the peak temperature data, but the difference
between the TG and DTG data and the EGA data does
seem to become somewhat more pronounced at the higher
heating rates. It does not appear that the transfer time
increases significantly with higher heating rates as might be
expected, but the differences between the TG and DTG
data sets and the DSC and EGA data sets become more
significant as the heating rate increased.
The results of the kinetic modelling for the TG, CO2 and
H2O data sets using the nth order reaction model are
illustrated in Figs. 5, 6 and 7, respectively. The quality of
the fitting for each of these data sets is illustrated by the
difference between the experimental data (symbols) and
the prediction from the kinetic model (lines). It is clear that
the quality of data fit, but not necessarily the meaningful-
ness, could have been improved with further kinetic
modelling. Further, a summary of the kinetic model data
using the nth order model is shown in Table 3 and Fig. 8.
The coefficient of determination, R2, is also reported for
each of the models in Table 3. The best fit, using the nth
order reaction model, was achieved for the TG data set,
while the CO2 and H2O data sets achieved a lower level of
fit. For the TG data, shown in Fig. 5, the inclusion of a
second reaction step would have helped to improve the fit
for the initial part of the decomposition curve. The kinetic
modelling of the CO2 and H2O data sets, shown in Figs. 6
and 7, would have benefited from at least one more reac-
tion step, but multiple reaction steps would likely be
required to achieve a higher level of correlation with the
experimental data.
The kinetic parameters for the TG, CO2 and H2O data
sets were compared using the Ozawa Flynn Wall model
free approach using the Netzsch Thermokinetics software
package [12, 13]. The kinetics of the DSC data set were not
considered due to the additional complication of the
instrumental correction factor required by the software for
DSC data [12]. This correction factor is designed to com-
pensate for the thermal transport phenomena associated
with DSC measurements [14, 15]. It should be noted that
no correction factors have been applied to any of the data
sets. Further, the DTG data was not considered in the
model free kinetics comparison due to limitations with the
software.
A comparison of the activation energy, Ea, and the pre-
exponential term, log A, for the TG, CO2, and H2O data
sets as a function of the fraction conversion, a, is displayed
in Fig. 4. As can be seen, the TG data resulted in both a
higher activation energy and pre-exponential term than the
two EGA data sets. The TG data also exhibited a higher
onset temperature and lower inflection point temperature
than the EGA data sets as noted previously. The CO2 and
H2O data sets resulted in very similar kinetic parameters. It
is interesting to note the similarity between the shape of the
Ea and log A curves for each data set, suggesting an
inability to accurately resolve a distinct separation or
Similar to the model free kinetic analysis described in
Fig. 6, the kinetic modelling using the nth order model
yielded kinetic parameters which appear to show the
kinetic compensation effect [10]. The compensation effect
for the kinetic parameters is very evident in Fig. 8. The
compensation effect makes a direct comparison of the
results of the kinetic modelling for each data set difficult.
Previous work by Sanders and Gallagher attempting to
123

Kinetic analysis of complex decomposition reactions
809
Fig. 5 Kinetic modelling of TG
data (nth order, single step
reaction)
Fig. 6 Kinetic modelling of
CO2 data (nth order, single step
reaction)
Fig. 7 Kinetic modelling of
H2O data (nth order, single step
reaction)
123

810
J. P. Sanders, P. K. Gallagher
100
80
60
40
20
0
Table 3 Comparison of kinetic parameters
TG
DTG
CO2 trace
H2O trace
Reaction model
Log A (s-1
E (kJ mol-1
nth order
10.16
98.54
0.85
nth order
9.11
nth order
7.34
nth order
7.68
)
)
90.23
0.90
77.36
0.77
79.98
0.80
Reaction order
R2
0.998
0.988
0.970
0.974
TG
DTG
CO2
H2O
correlate kinetic modelling of simultaneous TG/DSC data
for the decomposition of calcium carbonate also encoun-
tered this compensation effect [14, 15]. Both Brown and
Galwey have also discussed similar problems with the
current state of kinetic analysis of solid state decomposi-
tion reactions [16, 17]. Despite the lack of correlation in
kinetic parameters between the various data sets due to the
compensation effect, it is clear that each data set can be
described by the nth order reaction model with a relatively
high degree of correlation.
0
5
10
15
20
25
30
Time (min)
Fig. 9 Prediction of reaction rate at 125 ꢁC
decomposition of NaHCO3 were suitable for modelling
the decomposition reaction. The transfer times required to
move the gases from the thermal analyzer to the evolved
gas analyzer do not seem to significantly impact the
kinetic model.
The kinetic parameters reported in Table 1, were used
to predict the reaction rate at 125 ꢁC as a means of
comparing the modelling of each data set. The predictions
determined by each of the kinetic models are shown in
Fig. 9. The predictions based on each set of kinetic
parameters were very similar. The calculated Arrhenius
parameters must be taken as a pair rather than indepen-
dently significant parameters which emphasizes the point
of the compensation effect. The EGA data derived from
the FTIR for the evolution of CO2 and H2O from the
For the decomposition of NaHCO3, where the evolution
of CO2 and H2O occur simultaneously, kinetic modelling
with evolved gas data is relatively simple. In more com-
plicated systems where multiple decomposition products
are involved, the use of evolved gas data for kinetic
modelling allows for the possibility of studying the reac-
tion in much more detail than the traditional methods of
thermal analysis, such as TG or DSC.
100
95
90
85
80
75
10.5
10.0
9.5
9.0
8.5
8.0
7.5
7.0
0.92
0.90
0.88
0.86
0.84
0.82
0.80
0.78
0.76
Fig. 8 Kinetic parameters (nth
order reaction model)
TG
DTG
CO2
H2O
Ea
Log A
n
123

Kinetic analysis of complex decomposition reactions
811
3. Eigenmann F, Maciejewski M, Baiker A. Quantitative calibration
of spectroscopic signals in combined TG-FTIR system. Ther-
mochim Acta. 2006;440:81–92.
4. Sills I, He S. Determination of evolved gas transfer time from a
thermoanalyser to a coupled gas detection system using pyrite
oxidation. Thermochim Acta. 1999;339:125–30.
Conclusions
•
The transfer time for the detection of EGA signals can
be minimized by using an adequate flow rate.
–
For weakly absorbing species, increasing the flow
rate decreases the detection limit.
5. Roduit B, Baldyga J, Maciejewski M, Baiker A. Influence of
mass transfer on interaction between thermoanalytical and mass
spectrometric curves measured in combined thermoanalyser—
mass spectrometer system. Thermochim Acta. 1997;295:59–71.
6. Dei L, Guarini G. The thermal decomposition of NaHCO3. J
Therm Anal. 1997;50:773–83.
7. Yamada S, Koga N. Kinetics of the thermal decomposition of
sodium hydrogencarbonate evaluated by controlled rate evolved
gas analysis coupled with thermogravimetry. Thermochim Acta.
2005;431:38–43.
•
•
The extrapolated onset temperatures for the EGA
signals were lower than the onsets determined for the
TG data and very similar to the onsets determined for
the DTG data.
Peak temperatures were general higher for the EGA
data than for the TG or DTG data.
8. Kenner T, Khang S. Kinetics of the sodium bicarbonate—sulfur
dioxide reaction. Chem Eng Sci. 1993;48:2859–65.
9. Heda P, Dollimore D, Alexander K, Chem D, Law E, Bicknell P.
A method or assessing solid state reactivity illustrated by thermal
decomposition experiments on sodium bicarbonate. Thermochim
Acta. 1995;255:255–72.
–
The delay in peak temperature for the EGA data
appears to increase with increasing heating rate.
•
Model Free Kinetic Analysis predicts a lower activation
energy and pre-exponential term for the evolved gas
data when compared to the TG data.
´
´
10. Jankovic B, Adnadevic B. The use of the IKP method for eval-
¯
uating the kinetic parameters and the conversion function of the
thermal decomposition of NaHCO3 for nonisothermal thermo-
gravimetric data. Int J Chem Kinet. 2007;39:462–71.
11. Maciejewski M. Computational aspects of kinetic analysis, Part
B: the ICTAC Kinetics Project—the decomposition kinetics of
calcium carbonate revisited, or some tips on survival in the
kinetics minefield. Thermochim Acta. 2000;355:145–54.
12. Opfermann J. Kinetic analysis using multivariate non-linear
regression. I. Basic concepts. J Therm Anal Calorim. 2000;60:
641–58.
–
The activation energy and the pre-exponential term
seem to closely follow each other, suggesting a lack
of independence.
•
•
Fitting the data to a single step nth order kinetic model
resulted in an adequate fit for the TG, DTG and EGA
data.
13. Opfermann J, Kaiserberger E, Flammersheim H. Model-free
analysis of thermoanalytical data—advantages and limitations.
Thermochim Acta. 2002;391:119–27.
14. Sanders J, Gallagher P. Kinetic analysis using simultaneous TG/
DSC measurements: Part I: decompostion of calcium carbonate in
argon. Thermochim Acta. 2002;388:115–28.
–
The fit of the EGA data resulted in lower activation
energies, pre-exponential and reaction orders.
The kinetic data for each data set predicted similar
reaction rates at 125 ꢁC.
15. Sanders J, Gallagher P. Kinetic analysis using simultaneous TG/
DSC measurements: Part II: decomposition of calcium carbonate
having different particle sizes. J Therm Anal Calorim. 2005;
82:659–64.
–
EGA data were adequately able to predict reaction
rates for this complex reaction.
16. Brown M. Stocktaking in the kinetics cupboard. J Therm Anal
Calorim. 2005;82:665–9.
17. Galwey A. What theoretical and/or chemical significance is to be
attached to the magnitude of an activation energy determined for
a solid-state decomposition? J Therm Anal Calorim. 2006;86:
267–86.
References
1. Kaisersberger E, Post E. Practical aspects for the coupling of gas
analytical methods with thermal-analysis instruments. Thermo-
chim Acta. 1997;295:73–93.
2. Eigenmann F, Maciejewski M, Baiker A. Influence of measuring
conditions on the quantification of spectroscopic signals in TA-
FTIR-MS systems. J Therm Anal Calorim. 2006;83:321–30.
123