Chemical Engineering Science 58 (2003) 157 162

www.elsevier.com/locate/ces

A study of thermal-cracking behavior of asphaltenes

Jinsheng Wang, Edward J. Anthony
CANMET Energy Technology Centre, Natural Resources Canada, 1 Hannel Drive, Ottawa, Ont., Canada K1A 1M1 Received 26 September 2001; received in revised form 17 April 2002; accepted 23 August 2002

Abstract Asphaltenes are problematic substances for heavy-oil upgrading processes. Recently interesting ndings on thermal-cracking kinetics of an asphaltenic residue were reported, but a proposed model which considered parallel reactions for oil + gas and coke formation could not describe the behavior at higher temperatures. It was suggested that in such cases oils participated in secondary coke-forming reactions. Here we reexamine the data and give the expression for the oil + gas yield as a function of asphaltene conversion or residence time, which describes the data well. Further, we show that an empirical relation for coke formation and asphaltene conversion gives a reasonable description of the kinetics and prediction of the cracking behavior at high conversion level or long residence time, and that this method is much simpler. The maximum yield of oil + gas and the conversion level corresponding to the maximum yield can also be predicted easily. Further, our proposed approach is not dependent on assumed reaction orders of cracking. Crown Copyright ? 2002 Published by Elsevier Science Ltd. All rights reserved.
Keywords: Fuel; Kinetics; Mathematical modelling; Reaction engineering; Asphaltene; Thermal cracking

1. Introduction In the petroleum industry the demand for light products is increasing. To meet this demand, reneries convert a portion of their residuals into light fractions. Such residues contain large amounts of asphaltenes, which are high molecular weight and non-volatile aromatic compounds. Asphaltenes cause many problems in oil production, residual processing and heavy oil combustion (Callejas & Martnez, 2000; Seki & Kumata, 2000; Rogel, 2000; Martnez, Benito, & Callejas, 1997; Benito, Martnez, Fern andez, & Miranda, 1995; Wilt, Welch, & Rankin, 1998; Artok et al., 1999; Victorov & Smirnova, 1998; Permsukarome, Chang, & Fogler, 1997; Speight & Long, 1996), such as deactivation of catalysts and formation of sludge or sediment, which clog fuel lters, separators, and nozzles. They are precursors of coke formation in cracking reactions and cause undesirable properties in the products. Understanding cracking behavior of asphaltenes is of particular importance to heavy-oil processing, as asphaltenes comprise the heaviest fraction. Martnez et al. (1997) presented interesting experimental results on the thermal cracking of asphaltene. They used a
Corresponding author. Tel.: +1-613-996-2868; fax: +1-613-992-9335. E-mail address: banthony@nrcan.gc.ca (E. J. Anthony).

batch reactor to study the cracking of an asphaltenic residue from a synthetic crude obtained by coal liquefaction. The amount of asphaltene and yields of two classes of products oil+gas and coke were reported as a function of residence time and a three-lump model was used to t the data.
k1
Asphaltene Oil + Gas

k3 k2
Coke

where k1 , k2 and k3 are the relevant reaction rate constants. It was reported that the data could be tted well with the assumption of second-order reaction for asphaltene cracking at temperatures 425 C, 435 C and 450 C, without considering coke formation via secondary cracking of oils. However, the data no longer tted well at 475 C. Therefore, it was suggested that the secondary formation of coke due to cracking of oil became increasingly important with increasing temperature. Since the integration of the corresponding rate equations involving secondary cracking is not straightforward, the concentrationtime predictions for the products were not developed for 475 C. In this paper, we reexamine the asphaltene cracking data, and derive the concentration and conversion or residence time relation for the yields involving the secondary cracking

0009-2509/02/$ - see front matter Crown Copyright ? 2002 Published by Elsevier Science Ltd. All rights reserved. PII: S 0 0 0 9 - 2 5 0 9 ( 0 2 ) 0 0 4 3 0 - X

158

J. Wang, E. J. Anthony / Chemical Engineering Science 58 (2003) 157 162

of oil, by direct integration of the rate equations. Further, we have developed a relationship between the coke yield and conversion, which would make the predictions much easier.

and asphaltene content from Eqs. (1a) and (1c) CC CC 0 = r2 (CA0 CA ) or yC yC 0 = r2 (yA0 yA ); (6a) (6b)

2. Model and analyses By using the three-lump scheme of Martinez et al. for a batch reactor the following relations can be obtained: d CA 2 ; = (k1 + k2 )CA dt d CO 2 k 3 CO ; = k1 C A dt d CC 2 = k 2 CA + k 3 CO ; dt (1a) (1b) (1c)

where CC and CC 0 denote the amount of coke and initial coke, respectively. r2 = k2 = (k1 + k2 ). Furthermore, by substituting Eq. (2) for CA in Eqs. (5a) and (6a), the residence time dependence of CO and CC (or yO and yC ) can be obtained. 3. Results and discussion From Eqs. (5b) and (6b) it is clear that a plot of (yO yO0 ) or (yC yC 0 ) against (yA0 yA ) would yield a linear relationship with zero intercept, if the model holds. The reported data are summarized in Table 1. The plots in terms of Eqs. (5b) and (6b) are shown in Fig. 1, where the lines represent linear regression results. Indeed, for 425 C and 435 C, good linear relationships are shown. Moreover, the sums of the slopes for oil+gas and coke, as shown in the gure, are very close to unity. This conrms the prediction from the model, i.e., r1 + r2 = k1 = (k1 + k2 ) + k2 = (k1 + k2 ) = 1 when the secondary cracking is negligible. The plots for 450 C appear to deviate from Eqs. (5b) and (6b). Nevertheless, the sum of the slopes from linear regression is close to unity. However, the plots for 475 C clearly show that Eqs. (5b) and (6b) no longer hold, and Eq. (4) which involves the secondary cracking may be applicable. Eq. (4) contains two parameters r1 and r3 , which re ect the formation and cracking of oil + gas, respectively. The values can be determined by tting Eq. (4) to the data. On the other hand, for the sake of consistence to the model, the value of r1 may be estimated from the temperature dependence of r1 . The value of r1 for 475 C was estimated to be 0.52 using the Arrhenius equation k = k0 exp E RT : (7)

where C denotes concentration and t is residence time. The subscripts A, O and C denote asphaltene, oil + gas and coke, respectively. In Eqs. (1b) and (1c) the secondary cracking of oil is assumed to be an apparent rst-order reaction. The solution of Eq. (1a) is CA = CA0 ; 1 + CA0 (k1 + k2 )t (2)

where CA0 is the initial concentration of asphaltene. For the oil + gas lump, the following relation is obtained from Eqs. (1a) and (1b) d CO k1 k 3 CO = + : 2 d CA k1 + k 2 k1 + k 2 C A (3)

With the initial condition CO = CO0 at CA = CA0 , we obtain from Eq. (3) CO = exp(r3 =CA )[(r1 CA0 + CO0 ) exp(r3 =CA0 ) r1 CA exp(r3 =CA ) r1 r3 Ein(r3 =CA0 ) + r1 r3 Ein(r3 =CA )]; (4)

where r1 = k1 = (k1 + k2 ) and r3 = k3 = (k1 + k2 ). Ein(x) = x ex =x d x. Eq. (4) is similar to the one given by Week man and Nace for gasoline yield in plug ow reactors in uid catalytic cracking (Weekman & Nace, 1970). The difference of the present system is that CO0 = 0, and the reactor is a batch one instead of a continuous- ow one. When the secondary cracking is negligible, r3 = 0 and Eq. (4) reduces to CO CO0 = r1 (CA0 CA ) or yO yO0 = r1 (yA0 yA ); (5b) (5a)

where y denotes weight fraction in the mixture. On the other hand, with r3 = 0 we get a similar relation for coke yield

By presetting this value, we tted Eq. (4) to the oil + gas data, and as can be seen from Fig. 2, a reasonable t is achieved. It can also be seen that the oil + gas fraction is predicted to decrease as the conversion of asphaltene increases further and reduce to zero at complete conversion of asphaltene. This is not likely, however, since light gas would form along with coke formation due to the secondary cracking of oil. Consequently, the nal products would be coke and light gas. For a better description of the system near complete conversion of asphaltene, the current three lumps may not be adequate. Unfortunately, the gas yield was not separately reported giving the required information to modify the model. For the present discussion we have to keep the three-lump framework and note that the relation may not be valid in the region of very high conversion of asphaltenes.

J. Wang, E. J. Anthony / Chemical Engineering Science 58 (2003) 157 162 Table 1 Reported product distribution (wt%) for thermal cracking of an asphaltene residue (Martnez et al., 1997)

159

Residence time (min) Before cracking Asphaltenes Coke Oil + gas 425 C Asphaltenes Coke Oil + gas 435 C Asphaltenes Coke Oil + gas 450 C Asphaltenes Coke Oil + gas 475 C Asphaltenes Coke Oil + gas 65.5 15.0 19.6 5 10 20 30 40

58.5 17.0 24.5 56.7 17.9 25.4 54.5 19.2 26.4 51.8 19.8 28.4

53.8 19.3 26.9 51.9 19.6 28.5 46.3 21.0 32.7 44.5 21.2 34.3

44.4 23.6 32.0 41.5 25.1 33.5 38.1 25.7 36.2 31.5 31.2 37.3

42.5 24.1 33.4 34.9 28.6 36.5 32.3 30.6 37.1 29.6 37.4 33.0

39.7 25.1 35.2 30.9 30.9 38.2 29.6 32.0 38.4 28.9 38.6 32.5

30 20 10 0 30 20 10
435oC

425 C

y = 0.6032x

oil+gas coke
y = 0.392x

The residence time dependence of the oil + gas fraction can be obtained by combining Eq. (4) with Eq. (2). However, this would make the expression for CO quite complicated. The expression for CC , the coke yield would also be complicated, since it should be obtained from an overall material balance yO y O 0 + y C y C 0 = y A 0 y A : (8)

y = 0.5607x

oil+gas coke
y = 0.4363x

yield, %

0
30 20 10
y = 0.4359x
450oC

What is desirable is a simple relation for oil+gas and asphaltene contents. A simple relation between coke yield and conversion was given for catalytic cracking of gas oils (Wilson, 1997) yC = k 1 + C0 ; (9)

oil+gas coke

y = 0.561x

0 30 20 10 0 0 10 20 30 40

475oC

oil+gas coke

where yC is the weight fraction of coke. k is a coe cient. = 1 y1 and y1 is the weight fraction of unconverted gas oil. C0 is related to the content of carbon residue in gas oil. By analogy with catalytic cracking of gas oils, we assume the following relation for the asphaltene system: yC = k yA 0 y A + yC 0 : yA (10)

conversion (yA0 yA), %

Fig. 1. Plots in terms of Eqs. (5b) and (6b) for the dependence of oil+gas and coke yields on asphaltene conversion. The equations are results of linear regression.

The di erence between the two systems is that the weight fraction of gas oil in the feed is unity, whereas the fraction of asphaltene in the feed is yA0 . If this relation of Eq. (10) holds, i.e., there is a linear relationship between yC and (yA0 yA )=yA , the oil+gas content as a function of the asphaltene content can be determined from material balance. As shown in Fig. 3, the plots in terms of Eq. (10) can be

160

J. Wang, E. J. Anthony / Chemical Engineering Science 58 (2003) 157 162

60 50 measured predicted by eq.(11) predicted by eq.(4)

yoil + gas, %

40 30 20 10 0 0 10 20 30 40

50

60

70

yA0 yA, %
Fig. 2. Oil+gas yield as a function of asphaltene conversion at 475 C. The symbols represent measured yield and the curves represent predicted yield.

45 40 35 y = 17.371x + 15 2 R = 0.9522

coke yield, %

30 25 20 15 10 5 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4

425C 435C 450C 475C

(yA0 yA)/yA
Fig. 3. Correlation of coke yield with asphaltene conversion in terms of Eq. (10). The equation is the result of linear regression for the data of 475 C and R2 is the square of correlation coe cient.

taken as linear, and converging to yC =yC 0 =0:15 at yA =yA0 . A linear regression of the plotted data for 475 C yielded k = 0:17 for the relation given by Eq. (10). To obtain the oil+gas fraction as a function of the asphaltene fraction, we use the relation of the material balance Eq. (8). Substituting Eq. (8) into Eq. (10), we obtain yO = yO0 + (yA0 yA ) 1 k yA : (11)

Predicted yO based on Eq. (11) is shown in Fig. 2, in comparison with the tting by Eq. (4). It can be seen that Eq. (11) also gives a reasonable prediction, although it does not contain any term related to the secondary cracking of oil. Near complete conversion of asphaltene the two equations predict di erent rates for the decrease of oil + gas fraction. It should be noted here that Eq. (11) is not supposed to predict the conversion level at which the oil + gas fraction reduces to zero, since it is based on the expression given by

Eq. (10), which becomes innite at the complete conversion (yA = 0). However, as has been discussed earlier, Eq. (4) may not be valid at this conversion level either. At least in the medium range of conversion, Eq. (11) would be useful. In addition to this simplicity, the prediction for the oil+gas fraction with Eq. (11) requires only the data of asphaltene and coke fractions. Moreover, this prediction involves no assumptions of reaction orders for the cracking of asphaltene and the cracking of oil + gas. Eq. (11) also predicts a conversion level at which maximum oil+gas yield is obtained. With d yO = d yA = 0 Eq. (11) yields
yA =

kyA0 :

(12)

Using the values k = 0:17 from the coke data in Fig. 3 and yA0 =0:655 from Table 1, we obtain yA =0:33. Accordingly, at conversion level yA0 yA = 0:325 the maximum oil+gas yield is expected. This yield is calculated to be 0.354 by

J. Wang, E. J. Anthony / Chemical Engineering Science 58 (2003) 157 162

161

0.6 0.5 0.4 yoil+gas 0.3 0.2 0.1 0 0 20 40 60 t, min

Fig. 4. Oil+gas yield as a function of residence time. The symbols represent measured yield for 475 C and the curve represents predicted yield.

80

100

120

substituting yA into Eq. (11), and this is indeed close to the observed maximum yield 0.373. With the three-lump model the maximum yield can also be predicted. By letting d yO = d yA = 0 in Eq. (3) we obtain

yO max =

k 1 CF 2 yA ; k3

(13)

where CF is the concentration of the feed. The value of yA can be obtained by combining Eq. (13) with Eq. (4). However, numerical computation is required since an analytical expression for yA can not be derived. The oil+gas fraction can also be predicted as a function of residence time. For this purpose a relation between asphaltene content and residence time is needed. Eq. (2) gives this relation, but it was obtained with the assumption of second-order reaction for asphaltene conversion. In reality, the data at 475 C deviate signicantly from the assumption, as was observed from a plot in terms of Eq. (2) (not shown). Although a higher reaction order might give a better t of the asphaltene conversion data, for convenience of treatment we just take the second-order approximation. With the value of the rate constant in terms of Eq. (2) determined from a linear regression, the predicted oil+gas yield as a function of residence time is shown in Fig. 4. In comparison with the reported data, the prediction appears to be reasonable. The time at which the maximum oil+gas yield is reached can also be predicted, from Eqs. (2) and (12), as

the approach is the same. Of course, the prediction of the time dependence can be improved if a good correlation of asphaltene cracking with the residence time is established. Although we are not able to give a theoretical foundation to Eq. (10), the equation enables a reasonable description and prediction for the thermal cracking of asphaltene, and the expression for the yields and the manipulation of data are much simpler than those obtained for the three-lump model. Since the relationship given by Eq. (10) is linear, it can be determined with a microreactor, and then used for extrapolation and prediction of the yield in large plants. An implication from the above results is that, the oil+gas yield at low temperatures may also decrease with increasing conversion, but at higher conversion levels than those studied by Martnez et al. So long as the relation in Eq. (10) holds, the decrease is expected according to Eq. (11). This is also consistent with the three-lump model. As the fraction of the oil+gas lump increases to higher level, the secondary cracking, whose rate depends on the concentration of oil+gas, would become signicant and result in the decrease. However, this prediction should be veried experimentally. 4. Conclusions Thermal cracking of asphaltenes occurs in important heavy-oil upgrading processes such as coking and visbreaking. Our analyses conrmed that at lower temperatures the three-lump model which considered parallel reactions of oil+gas and coke formation described the cracking behavior well, whereas at higher temperature the secondary cracking of oil may be considered. Unfortunately, the resulting expression for the oil+gas yield is quite complicated. With an empirical relation for the coke yield and the asphaltene conversion, a much simpler expression for the oil+gas yield can be obtained and a reasonable description has been achieved. Prediction of the maximum yield is also

t =

yA0 =k 1 : CA0 (k1 + k2 )

(14)

As has been seen earlier, Eq. (10) is an analogy of Eq. (9), which is an empirical relation in catalytic cracking. The prediction for the oil+gas selectivity based on the resultant Eq. (11) is not related to the three-lump model, except for the prediction of the residence time dependence where the assumption of the second-order reaction for asphaltene cracking is used. It should be noted that for other reaction orders,

162

J. Wang, E. J. Anthony / Chemical Engineering Science 58 (2003) 157 162

much simpler. This approach is not dependent on the reaction orders of cracking of asphaltene and oil. Moreover, since the empirical relation is linear, it can be determined with a microreactor, and then used for extrapolation and prediction of the yield in large reactors. The relations we developed here have the potential to be useful in describing thermal-cracking processes for heavy oils. Notation C C0 Cj0 E k k1 k2 k3 R ri T t t y yA yj0 concentration, kg= m3 a constant in Eq. (9) related to the content of carbon residue in gas oil initial concentration of component j , kg= m3 activation energy in Arrhenius equation x Ein(x) = ex x d x a coe cient in Eq. (9) relating coke fraction with gas oil fraction rate constant oil + gas formation rate constant for coke formation from asphaltene cracking rate constant for coke formation from oil + gas cracking gas constant in Arrhenius equation =ki = (k1 + k2 ) (i = 1; 2; 3) temperature, K residence time, s time corresponding to maximum oil + gas yield weight fraction weight fraction of asphaltene corresponding to maximum weight fraction of oil + gas initial weight fraction of component j (=1 y1 ), where y1 is the weight fraction of unconverted gas oil

Subscripts A C F O asphaltene coke feed oil + gas

References
Artok, L., Su, Y., Hirose, Y., Hosokawa, M., Murata, S., & Nomura, M. (1999). Structure and reactivity of petroleum-derived asphaltene. Energy and Fuels, 13, 287296. Benito, A. M., Martnez, M. T., Fern andez, I., & Miranda, J. L. (1995). Visbreaking of an asphaltenic coal residue. Fuel, 74, 922927. Callejas, M. A., & Martnez, M. T. (2000). Hydroprocessing of a Maya residue. 1. Intrinsic kinetics of asphaltene removal reactions. Energy and Fuels, 14, 13041308. Mart inez, M. T., Benito, A. M., & Callejas, M. A. (1997). Thermal cracking of coal residues: Kinetics of asphaltene decomposition. Fuel, 76, 871877. Permsukarome, P., Chang, C., & Fogler, H. S. (1997). Kinetic study of asphaltene dissolution in amphiphile/alkane solutions. Industrial Engineering Chemistry and Research, 36, 39603967. Rogel, E. (2000). Simulation of interactions in asphaltene aggregates. Energy and Fuels, 14, 566574. Seki, H., & Kumata, F. (2000). Structural change of petroleum asphaltenes and resins by hydrodemetallization. Energy and Fuels, 14, 980985. Speight, J. G., & Long, R. B. (1996). The concept of asphaltenes revisited. Fuel Science Technology International, 14, 112. Victorov, A. I., & Smirnova, N. A. (1998). Thermodynamic model of petroleum uids containing polydisperse asphaltene aggregates. Industrial Engineering Chemistry and Research, 37, 32423251. Weekman, V. W., & Nace, D. M. (1970). Kinetics of catalytic cracking selectivity in xed, moving and uid bed reactors. A.I.Ch.E. Journal, 16, 397404. Wilson, J. W. (1997). Fluid catalytic cracking technology and operations. Tulsa, Oklahoma: Penn Well Publishing Company. Wilt, B. K., Welch, W. T., & Rankin, J. G. (1998). Determination of asphaltenes in petroleum crude oils by Fourier transform infrared spectroscopy. Energy and Fuels, 12, 10081012.