Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Fuel 80 (2001) 327±334

www.elsevier.com/locate/fuel

Kinetics of the thermal decomposition of oil shale


from Puertollano (Spain)
M.C. Torrente, M.A. GalaÂn*
Department of Chemical Engineering, University of Salamanca, Plaza de Los CaõÂdos, 1-5 37008 Salamanca, Spain
Received 13 March 2000; revised 6 June 2000; accepted 9 June 2000

Abstract
The kinetics of the thermal decomposition of Spanish oil shale was studied using isothermal and non-isothermal thermogravimetric
analyses (TGA). The isothermal TG data were studied at different particle sizes (0.250 £ 10 23, 0.505 £ 10 23 and 0.925 £ 10 23 m) and three
temperature plateaux (703, 688 and 673 K). For the kinetic analysis, the Integral Method was used.
The non-isothermal weight-loss data was analysed at different heating rates (5, 10, 15, 20 and 50 K/min) up to 1173 K and for a particle
size of 0.250 £ 10 23 m. Three methods were applied for the determination of the kinetic parameters: the Direct Arrhenius Plot Method, the
Integral Method and the Differential Method.
From an engineering point of view, the rate of the thermal decomposition of Spanish oil shale can be suitably described by overall ®rst-
order kinetics. We also found that transport effect problems affected the observed rate when the heating rates were higher than 10 K/min;
additionally, no mass and heat transfer resistance was observed for the different particle sizes studied.
The results obtained from the isothermal method give an apparent activation energy of 150 kJ/mol and a frequency factor of
2.11 £ 10 8 s 21, while for the non-isothermal method these results were: E ˆ 167 kJ=mol and A ˆ 2:16 £ 109 s21 : The pyrolysis rates of
Spanish shale are compared with those of Anvil Points, Clear Creek and Moroccan shale. q 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Oil shale; Pyrolysis; Spain; TGA

1. Introduction two successive ®rst-order reactions.


kerogen…s† ! bitumen …1†
Spain has moderate deposits of oil shale [1], which can be
exploited by pyrolysis to decompose the insoluble organic
matter called kerogen into the desired oil products. bitumen ! oil 1 gas 1 char …2†
Kerogen is a complex substance consisting of large mole- where the bitumen product refers to lower molecular weight
cules of carbon and hydrogen with nitrogen, sulphur and components which are vaporisable.
oxygen atoms introduced into the molecule. The average Braun and Rothman [3] examined the Hubbard and
molecular weight is in the order of 3000 and an approximate Robinson data and showed that the oil-production kinetics
empirical formula would be C200H300SN5O11. When shale could be explained more accurately by including a thermal-
particles are heated to about 623 K, the kerogen starts to induction period in the data analysis. This thermal-induction
decompose and partially vaporise, leading to ®nal products period was required in order to account for the non-isother-
that, at room temperature, consist of gas, oil, and residual mal heating effects in the experiments performed by
carbon. For the design of equipment to obtain oil from shale Hubbard and Robinson. Later, GalaÂn and Smith [4] deter-
it is necessary to know the rate of decomposition. mined the intrinsic isothermal kinetics of Colorado oil
Numerous attempts addressing the mechanistic and shales by incorporating an induction period and determined
kinetic points of view have been made to understand the the extent to which transport processes affected the rate
processes occurring during oil shale pyrolysis. Working decomposition and, also, the conditions under which these
with isothermal conditions, Hubbard and Robinson [2] transport effects could be eliminated.
obtained rate data that could be interpreted in terms of However, non-isothermal thermogravimetric analysis
offers certain advantages over the classic isothermal
* Corresponding author. Tel.: 134-923-294479; fax: 134-923-294574. method. First, this kind of analysis eliminates the errors
E-mail address: magalan@gugu.usal.es (M.A. GalaÂn). introduced by the thermal induction period; second, it
0016-2361/01/$ - see front matter q 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0016-236 1(00)00101-0
328 M.C. Torrente, M.A. GalaÂn / Fuel 80 (2001) 327±334

Nomenclature
a fraction of kerogen pyrolysed
b heating rate (K/min)
A frequency factor (s 21)
E activation energy (J/mol)
R gas constant (J/mol K)
T temperature (K)
t time (s)
t0 time at the start of the constant-temperature period (s)
w1 weight of the sample after complete pyrolysis of kerogen (mg)
w0 initial weight of the sample (mg)
wt weight of the sample at ªtº seconds (mg)

permits a rapid scan of the whole temperature range of Table 1


interest. For these reasons, several researchers have Characteristics of Spanish oil shale
employed the non-isothermal technique for oil shale pyro- Density (g/ml) 1.8
lysis [5±11]. In addition, the non-isothermal technique Ash (%) 62.8
permits one to study the in¯uence of the heating rate on Organic carbon (%) 26.0
the kinetic and on the yield of the pyrolysis [12,13]. Ash analysis (%)
However, the most effective method for determining the SiO2 56.6
Al2O3 27.6
most probable kinetic mechanism of polymer decomposi- Fe2O3 9.1
tion is a combined kinetic analysis of isothermal and non- CaO 2.6
isothermal thermogravimetric measurements. Such a MgO 2.2
combined analysis was ®rst performed by Campbell et al. Other oxides 1.9
[6].
Thakur and Nuttall [14], also using a combined thermo-
gravimetric analysis, compared the kinetics of Moroccan oil samples consisting of about 4 £ 10 26 kg (4 mg) were
shale with those of Jordan and Colorado shale and found added to the sample holder and pyrolysed in a nitrogen
that the decomposition rate of Moroccan oil shale is differ- atmosphere. Prior to heating, the system was purged at
ent from that of Jordan and Colorado shale and that it room temperature with nitrogen (¯ow rate 1.33 £ 1026 m 3/s)
depended on the temperature. This is not surprising because for 15 min to eliminate oxygen. The working conditions
oil shales have different origins and are found in differing were chosen to eliminate external transport effects, follow-
geological environments. ing Galan and Smith [4], these conditions were the same as
Here we used isothermal and non-isothermal TG those described above (¯ow rate of nitrogen and weight
measurements to determine the probable mechanism and sample).
kinetics of Spanish oil shale. Weight vs. time data were taken under two sets of oper-
ating conditions: isothermal and non-isothermal. In the ®rst
set, the temperature was increased from 298 K to the desired
2. Experimental
constant value by heating at the maximum rate: 160 K/min.
2.1. Oil shales Isothermal oil shale decomposition was carried out at 673,
688 and 703 K with sample sizes of 0.250 mm (50±70
The oil shale used in this work was obtained from mesh), 0.505 mm (30±40 mesh) and 0.925 mm (10±20
Puertollano in Ciudad Real (Spain). The elemental mesh). In the second set of experiments, the 0.250 mm
analysis of the oil shale is given in Table 1. The origi- sample was heated from ambient temperature to 1173 K.
nal shale was ground and sieved to give three particle Heating rates of 5, 10, 15, 20 and 50 K/min were employed.
sizes: 0.250 £ 10 23 m (0.290 £ 10 23 ±0.210 £ 10 23 m), Temperature calibration was carried out in the equipment
0.505 £ 10 23 m (0.590 £ 10 23 ±0.420 £ 10 23 m) and Table 2
0.925 £ 10 m (1.000 £ 10 23 ±0.850 £ 10 23 m). The
23
Fischer assay data
Fischer assay data are given in Table 2.
Oil (l/ton) 177
2.2. Thermogravimetric analysis Oil (%) 17.6
Water (%) 1.8
Spent shale (%) 78.4
Intrinsic kinetic data were obtained in a Perkin±Elmer,
Gas and loss (%) 2.2
TGS-2 thermogravimetric analysis apparatus. Then,
M.C. Torrente, M.A. GalaÂn / Fuel 80 (2001) 327±334 329

under conditions close to the working conditions. To test the or:


reproducibility of the thermal characteristics of the furnace,    
da A E
preliminary runs were made without shale in the basket for ˆ ´exp 2 ´…1 2 a† …7†
dT b RT
the same nitrogen ¯ow rate and two temperature plateaux of,
673 and 700 K. Reproducibility was good (^2 K) and a where b gives dT/dt.
constant ®nal temperature could be maintained. These In the present study, three models were used to evaluate
runs were made at an approximate heating rate of 160 K/ the kinetic parameters: (1) the Direct Arrhenius Plot
min. Method; (2) the Integral Method; and (3) the Differential
To check reproducibility, three runs were made under the Method. We assumed a single ®rst-order equation to
same conditions. When the three sets of results did not describe the decomposition reaction of kerogen in all of
agree, perhaps due to small differences in sample composi- these methods.
tion, a fourth run was made. Direct Arrhenius Plot Method. In this method, Eq. (7)
was rearranged to give:
2.3. Analysis of weight loss data    
1 da A E
ln ´ ˆ ln 2 …8†
The extent of conversion, or fraction of material pyro- …1 2 a† dT b RT
lysed, a , was de®ned by the expression: The values of E and A for the individual runs were calcu-
w 2 wt lated based on Eq. (8) by plotting the experimental values of
aˆ 0 …3†
w0 2 w1 ln[(1/(1 2 a )) £ (da /dT )] against 1/T.
Integral Method. This method estimates the activation
where w0 is the initial weight (mg), wt the weight after ªtº energy and frequency factor values of a reaction from the
minutes (mg), and w1 is the weight after complete pyrolysis overall conversion vs. temperature curves. Rearranging and
(mg). integrating, one has:
Assuming that the reaction order is close to unity it would
Z a da A ZT
be advantageous to simplify the kinetics to ®rst-order. This ˆ ´ exp…2E=RT† dT …9†
would greatly simplify subsequent use of the rate expression 0 …1 2 a† b T0
in model oil shale retorting. It should also facilitate compar-
where T0 is the initial temperature.
ison of our experimental results with those of previous
Integration of Eq. (9), using the exponential integral func-
investigators, most of which a priori assumed the reaction
tion, affords:
order to be unity. This assumption should be con®rmed   
later. Thus, the rate of kerogen decomposition is given by: A E E
2ln…1 2 a† ˆ ´ T´e2E=RT 1 ´Ei 2 …10†
  b R RT
da 2E
ˆ A´exp ´…1 2 a† …4†
dt RT Here, we made use of the fact that T0 is chosen low and the
rate at this temperature is negligible; therefore, the lower
where A is the frequency factor (s 21); E the activation limit can be replaced by zero. The exponential integral
energy (J/mol), R the gas constant (J/mol K), and T is the Ei …2E=RT† can be approximated by:
temperature (K).
   
E e2E=RT 1! 2! ¼
Ei 2 ˆ2 ´ 12 1 2
2.3.1. Isothermal study RT E=RT E=RT …E=RT†2
For the isothermal decomposition of kerogen, Eq. (4) was …11†
integrated to give:
If the ®rst three terms of the approximation are used, Eq.
ln…1 2 a† ˆ 2k´…t 2 t0 † …5† (11) becomes:
 
where t0 is the time at the start of the constant-temperature ART 2 2RT
2ln…1 2 a† < ´ 12 ´e2E=RT …12†
period. bE E
Thus, for ®rst order kinetics a plot of ln(1 2 a ) against
(t 2 t0) will yield a straight line, with a slope equal to 2k. Dividing both sides of Eq. (12) by RT 2 …1 2 2RT=E†=b
Then, the Arrhenius plot was used to determine activation and taking the logarithm, one obtains:
     
energies and frequency factors. 2b´ln…1 2 a† 2RT A E
ln 2
2 ln 1 2 < ln 2
RT E E RT
2.3.2. Nonisothermal study …13†
For the non-isothermal decomposition of solids, Eq. (4)
The values of E and A can be obtained by repeated least-
can be modi®ed by introducing the heating rate as:
squares ®t of Eq. (13) to the experimental data. By ®rst
 
da dT E using an approximate E in the left-hand side of Eq. (13),
´ ˆ A´exp 2 ´…1 2 a† …6†
dT dt RT it is possible to plot the left-hand side of Eq. (13) as a linear
330 M.C. Torrente, M.A. GalaÂn / Fuel 80 (2001) 327±334

Fig. 2. Arrhenius plots for the different particle sizes.

independent variables, 1/T and T 2 e2E=RT …1 2 2RT=E†=b:


The second variable is calculated using an approximate E
in the ®rst step. Once an approximate E has been obtained
Fig. 1. Kinetic data for a particle size of 0.925 £ 10 23 m. using the least-squares treatment, T 2 e2E=RT …1 2 2RT=E†=b
is redetermined using the new value of E and, successively,
a more accurate value of E is obtained.
function of 1/T and obtain 2E/R from the slope and A/E
from the intercept. The value of E thus obtained is used in
the values of the left-hand side and, successively, a more 3. Results and discussion
accurate value of E is obtained until no improvement in the
value of E is achieved. 3.1. Isothermal method
From Eq. (12), a can be obtained as:
( "  #) Fig. 1 shows fractional weight vs. time, plotted as
2ART 2 2E=RT 2RT suggested by Eq. (5), for the particle size of
a ˆ 1 2 exp ´e ´ 12 …14†
bE E 0.925 mm in the 673±703 K temperature range; analo-
gous ®gures were obtained for the others particle sizes.
Differential Method. This method determines the values
The average relative error over the entire weight-loss
of E and A from the instantaneous rate vs. temperature
curve, determined from replicate pyrolysis experiments,
curves. Substituting Eq. (14) in Eq. (6) and taking loga-
was approximately ^6%.
rithm, we obtain:
As can be seen, nearly straight lines, indicating ®rst-order
   
da E ART 2 2E=RT 2RT kinetics, are obtained down to a fractional weight change
ln ˆ ln A 2 2 ´e ´ 12 …15† that depends upon the temperature. The extents of conver-
dt RT bE E
sion that can reasonably be represented by ®rst order
The values of E and A can be obtained by ®tting Eq. (15) kinetics ¯uctuate between a values of 0.89 and 0.95.
to the experimental data, using the repeated least-squares Since it is known that the pyrolysis of oil shale is a rather
treatment. Speci®cally, one assumes Eq. (15) to have two complex process, this deviation from ®rst-order kinetics is
not unexpected [15]. For example, coke-forming reactions,
Table 3 which would lead to higher values of the ordinate in Fig. 1,
Isothermal kinetic parameters could in part account for the concave shape of the curves at
Particle size (m) T (K) k £ 10 24 (s 21) E (kJ/mol) A £ 10 8 (s 21) low values of (1 2 a ).
Speci®c rate constants evaluated from the slopes of the
0.250 £ 10 23 673 4.73 ^ 0.12 155 ^ 16 5.31 straight-line sections of the curves in Fig. 1 were used.
688 9.62 ^ 0.17
Linear regression analysis afforded the results indicated in
703 15.41 ^ 0.67
673 4.23 ^ 0.11 Table 3. The activation energy and frequency factor were
0.505 £ 10 23 688 8.30 ^ 0.19 148 ^ 15 1.34 determined using an Arrhenius plot, Fig. 2, for the different
703 13.07 ^ 0.34 particle sizes. It is interesting to note that there are no signif-
673 4.38 ^ 0.06 icant differences between the activation energies obtained
0.925 £ 10 23 688 8.55 ^ 0.21 148 ^ 14 1.31
for the different particle sizes studied Ð less than 10% Ð
703 13.49 ^ 0.29
which lies within the kinetic equation errors. The overall
M.C. Torrente, M.A. GalaÂn / Fuel 80 (2001) 327±334 331

Table 4
Isothermal rate constants for different origins oil shale (s 21)

T (K) Spanish shale (present study) Anvil Points shale [15] Clear Creek shale [15] Moroccan shale [14]
24a 24 24
673 4.80 £ 10 5.54 £ 10 5.12 £ 10 2.50 £ 10 23
688 8.62 £ 10 24a 8.90 £ 10 24 8.70 £ 10 24 3.29 £ 10 23a
703 1.51 £ 10 23a 3.63 £ 10 23 2.84 £ 10 23 4.15 £ 10 23a
723 3.07 £ 10 23a 1.20 £ 10 22a 1.31 £ 10 22a 5.17 £ 10 23
Activation energy (kJ/mol) 150 247 254 62.3
A (s 21) 2.11 £ 10 8 8.40 £ 10 15 2.94 £ 10 16 176.7
a
k values calculated by using kinetics parameters in the Arrhenius equation.

value of the kinetic parameters are E ˆ 150 kJ=mol and A ˆ Figs. 4 and 5 show the plots for the Direct Arrhenius Plot
2:11 £ 108 : The overall value was obtained using the data Method and the Integral Method for the different heating
for the three particle sizes simultaneously in the kinetic rates. The closeness of ®t indicates the validity of assuming
analysis rather than simply averaging the values for each a ®rst-order decomposition mechanism for a weight loss
particle size. between 3 and 10% of the total weight from the oil shale.
The isothermal speci®c rate constants for the decomposi- The kinetic parameters are shown in Table 5; as can be
tion of oil shales of different origins are compared in Table seen, different procedures give somewhat different activa-
4. The activation energy of Spanish oil shale lies close to the tion energies and associated pre-exponential factors. We
middle of the range for Moroccan and American oil shales. believe that in terms of conversion the data are more reliable
While the decomposition rate of Spanish oil shale is similar than the instantaneous rates as measured in our experiments.
to those of Anvil Points Shale and Clear Creek Shale at Although we used small-time increments, the rates
temperatures of 673 and 688 K, it is lower than the decom- measured are based on the average of true instantaneous
position rate of Moroccan oil shale. rates over the time interval. Therefore, the kinetic para-
meters obtained using the Integral Method are recom-
3.2. Non-isothermal mended as the best values determined from our
experimental data, since the errors in the constants are smal-
Fig. 3 offers the weight loss data for Spanish oil shale as a ler. However, the Direct Arrhenius Plot Method is the
function of temperature at different heating rates. This ®gure simplest and involves the least amount of effort, at the
has certain important features: The total extractable kerogen same time affording results that are in close agreement
content in the oil shales represents about 14 wt% of the total with those obtained using the Integral Method.
shale weight. No carbonate decomposition is observed, The values of the activation energy calculated with all
suggesting that there are no carbonates in Spanish oil shale. three methods at the different heating rates are plotted in
In the present study, three methods were applied for the Fig. 6. The effect of the heating rate on the kinetic para-
determination of the kinetic parameters: the Direct Arrhenius meters can be seen, the activation energy decreasing when
Plot Method, the Integral Method and the Differential Method. the heating rate is above 15 K/min, in accordance with the

Fig. 3. Non-isothermal TG curves for Spanish oil shale at various heating rates.
332 M.C. Torrente, M.A. GalaÂn / Fuel 80 (2001) 327±334

Fig. 4. Direct Arrhenius Plot Method for the different heating rates.

data in the literature [11,12,14,16±18]. When the heating shale can be described by a pseudo-®rst-order reaction.
rate is higher than 20 K/min, there is no dependence of this These kinetics are ªeffectiveº values and as such can only
on the kinetic parameters. This difference is due to a shorter be used to describe the macroscopic kerogen decomposition
exposure time to a particular temperature at faster heating rates process rather than the microscopic chemistry involved.
and this causes transport limitations due to heat transfer. Effective kinetic data are very useful for modelling a
In Table 6, the isothermal speci®c rate constants are complex chemical process.
compared with the speci®c rate constants calculated under The decomposition rate of Spanish oil shale is similar to
non-isothermal conditions at differing heating rates. Despite that of American oil shale at temperatures of 673 and 688 K
the cumulative error in the rate constants, in the isothermal and lower than that of Moroccan oil shale in the temperature
method the kinetic parameters are more similar to the range studied.
kinetic parameters for a heating rate of 15 K/min. We Among the three models used in non-isothermal condi-
believe that the most plausible explanation (among others) tions, the Integral Method yields lower deviation and hence
for this similarity is due to the weight loss occurring during provides a better ®t of the data.
the induction period in the isothermal runs.

Acknowledgements
4. Conclusions
This research was supported with funds from the
The isothermal and non-isothermal TG measurements ªComisioÂn Interministerial de Ciencia y TecnologõÂaº
have shown that the thermal decomposition of Spanish oil (CICYT) QUI1999-1011.

Fig. 5. Integral Method for the different heating rates.


M.C. Torrente, M.A. GalaÂn / Fuel 80 (2001) 327±334 333

Fig. 6. Comparison of the methods.

Table 5
Kinetic parameters calculated by different methods

Heating rates (K/min) Arrhenius plot method Integral method Differential method

E (kJ/mol)
5 171 ^ 17 167 ^ 6 157 ^ 21
10 171 ^ 16 170 ^ 7 148 ^ 15
15 142 ^ 10 142 ^ 3 132 ^ 10
20 102 ^ 4 123 ^ 3 103 ^ 4
50 115 ^ 9 121 ^ 6 107 ^ 7
A (s 21)
5 1.53 £ 10 10 2.16 £ 10 9 7.50 £ 10 9
10 9.23 £ 10 9 3.87 £ 10 9 1.14 £ 10 9
15 4.65 £ 10 7 3.84 £ 10 7 4.73 £ 10 7
20 3.27 £ 10 4 1.35 £ 10 6 2.33 £ 10 5
50 2.65 £ 10 5 7.91 £ 10 5 4.18 £ 10 5

Table 6
Comparison between isothermal and non-isothermal rate constants (s 21)

T (K) Isothermal data Non-isothermal data (Integral Method)

Overall 5 K/min 15 K/min 50 K/min

673 4.80 £ 10 24 2.36 £ 10 24 3.65 £ 10 24 3.21 £ 10 24


688 8.62 £ 10 24 4.52 £ 10 24 6.35 £ 10 24 5.14 £ 10 24
703 1.51 £ 10 23 8.42 £ 10 24 1.08 £ 10 23 8.08 £ 10 24
723 3.07 £ 10 23 1.86 £ 10 23 2.11 £ 10 23 1.43 £ 10 23
Activation energy (kJ/mol) 150 167 142 121
A (s 21) 2.11 £ 10 8 2.16 £ 10 9 3.84 £ 10 7 7.91 £ 10 5

References [7] Chen WJ, Nuttall HE. TGA study of Colorado Shale and new model
of pyrolysis. AIChE National Meeting Houston, Texas; 1±5 April
1979.
[1] Russel PL. Oil shales of the world, their origin, occurrence and [8] Shih SM, Shon HY. Process Des Dev 1980;19:420.
exploitation. Oxford: Pergamon Press, 1990. [9] Li S, Qian J. Fuel 1991;70:1371.
[2] Hubbard AB, Robinson WE. Decomposition study of Colorado Oil [10] Dogan OM, Uysel BZ. Fuel 1996;75:1424.
Shale, US Bureau of Mines, Report of Investigation 4744, 1950. [11] Willians PT, Ahmad N. Fuel 1999;78:653.
[3] Braun RL, Rothman AJ. Fuel 1975;54:129. [12] Arnold CH. Effect of heating rate on the pyrolysis of oil shale.
[4] GalaÂn MA, Smith JM. AIChE J 1983;29:604. Proceedings on Symposium in Industrial and Laboratory Pyrolysis,
[5] Granoff B, Nuttall HE. Fuel 1977;56:234. Division of Petroleum Chemistry, 169th ACS Meeting, 1975, Paper
[6] Campbell JH, Koskinas GH, Stout ND. Fuel 1978;57:372. no. 28.
334 M.C. Torrente, M.A. GalaÂn / Fuel 80 (2001) 327±334

[13] Rajeshwar K. Thermochim Acta 1981;45:253. [17] Shala D, Kopsch H, Sokic M, Neumann HJ, Jovanovic JA. Fuel
[14] Thakur DS, Nuttall HE. Ind Eng Chem Res 1987;26:1351. 1987;66:1185.
[15] Pan Z, Feng HY, Smith JM. AIChE J 1985;31:721. [18] Shala D, Kopsch H, Sokic M, Neumann HJ, Jovanovic JA. Fuel
[16] Derrall AY, Arnold C. Thermochim Acta 1976;17:165. 1990;69:490.

You might also like