1186
I n d . Eng. Chem. Res. 1988,27, 1186-1193
A Pilot Plant Designed for the Process Study of Hydrodenitrogenation and Hydrogen Consumption Dejan U. Skala,* Marko D. Saban, and Jovan A. Jovanovid Faculty of Technology and Metallurgy, University of Belgrade, P.O.Box 494, 11001 Beograd, Karnegijeva 4 , Yugoslavia
Volker W. Meyn and Iradj G.-H.Rahimian German Institute for Petroleum Research, 3392 Clausthal-Zellerfeld, Federal Republic of Germany
A pilot plant with a trickle-bed reactor was designed and built with the purpose of investigation of reaction networks, determination of catalyst activity, and on-line measurement of hydrogen consumption during steady-state operation. Preliminary experiments were performed using a test mixture which consisted of quinoline, indole, and lauronitrile dissolved in decalin or hexadecane. Three different types of catalyst loadings were used (a mixture of Co-Mo, Ni-Mo, and pure Ni-Mo) in the following range of operating conditions: 250-370 OC,80 bar, and LHSV from 0.35 to 1.33 h-l. T h e obtained results showed that the catalyst activity changed with time with an increasing decline effect after 60 h of operation. The overall HDN process of nitrogen compounds could be expressed as a first-order reaction with an apparent activation energy of 103.4 kJ/mol. The determined differences between on-line measurement of the hydrogen consumption and the theoretical one predicted by using GC-MS product analysis were in the range of 25%. Most of the information on trickle-bed reactors (TBRs) applies to their use in large-scale processes. The results of laboratory-scale studies of specific chemical reactions in TBR have appeared in recent literature. This is directly correlated with the recent emphasis on the catalytic upgrading of heavy crudes, petroleum residues, coal liquids, and oils derived from tar sands and oil shales, i.e., potential energy sources containing large quantities of different types of sulfur, nitrogen, and oxygen compounds. These hetero compounds are removed by hydrogenation, giving H2S, NH,, and HzO. For determining the reaction mechanisms and kinetics of hydrodesulfurization (HDS), hydrodenitrogenation (HDN), and hydrodeoxygenation (HDO), model substances are often used. Table I presents a short review of some investigations in laboratory T B R s with different model substances. Laboratory Trickle-Bed Reactor (TBR) In order to obtain detailed information on the influence of pressure, temperature, LHSV, and type of catalyst on the kinetics of catalytic processes in laboratory TBRs, it is necessary to fulfil the conditions for neglecting the effects of internal and external mass transfer and temperature gradients. For application in direct scale-up, it is desirable to perform experiments using commercial catalysts in their original size and to investigate catalytic processes in laboratory TBR’s under the same superficial liquid velocity (GL,kg/(m2-s))as those in an industrial reador. In fact, these conditions require laboratory reador design of the same length as a commercial one to simultaneously match the mass velocity and contact time of a commercial reactor. Unfortunately, the design of laboratory T B R s is in many cases limited by some demand, e.g., relatively short laboratory reactors and very often only small volumes of catalyst to minimize the quantity of feed. In the literature, many different designs of small tricklebed reactors may be found. Table I1 sumarizes the information about several experimental setups, together with important remarks on the applied working conditions and the determination of hydrogen consumption directly in the pilot plant. Although all the cited laboratory units were designed as gas-liquid-solid reactors, the applied feed, 0888-5885/88/2627-1186$01.50/0
Table I. Literature Review for HDS, HDN, and HDO Reaction Studies in Trickle-Bed Reactor with Model Substances reference model substances Doelman and Vlugter, quinoline, isoquinoline, aniline, 1963 o-toluidine, benzylamine, propylamine Rollman, 1977 dibenzothiophene, dibutyl sulfide, quinoline, dibenzofuran, naphthalene, 2-methylnaphthalene, 2,3-dimethylnaphthalene Eliezer et al., 1977 dibenzothiophene, quinoline Badilla-Ohlbaum et al., phenanthrene, dibenzothiophene, 1979 dibutyl sulfide, quinoline, dibenzofuran Krishnamurthy et al., dibenzofuran 1981 Chu and Wang, 1982 dibenzothiophene, aniline, naphthalene, cumene Satterfield and Yang, quinoline 1984; Yang and Satterfield, 1983, 1984; Satterfield et al., 1985 Kumar et al., 1984 benzothiophene benzothiophene Hook and Akgerman, 1985 Furimsky et al., 1986 phenols and ortho-substituted derivatives
pressure, temperature, and hydrogen-to-oil ratio permit them to be considered, in some cases, as gas-solid, fixedbed reactors (Moore and Akgerman, 1985). Therefore, taking into account the above-mentioned reality, the pilot plant with a TBR used in this study can and did provide valuable information on the qualitative relationships in reaction networks, catalyst activity decline, and hydrogen consumption, but it could not yield intrinsic kinetics and the mass-transfer coefficient necessary for reactor design and scale-up. Unit Design The objective of the design was to build a small laboratory unit for hydroprocessing shale oil, heavy crude oil, and petroleum residues in an isothermal TBR. Good contacting needs to be achieved between the gas (H,) and 0 1988 American Chemical Society
Ind. Eng. Chem. Res., Vol. 27, No. 7, 1988 1187 Table 11. Small Laboratory T B R Literature Review reactor appld reaction conditions 1. d, cm 1. t, o c 2. L, cm catalyst type 2. P,bar reference 3. L,, cm and loading 3. LHSV, h-l Wolfson et al., 1951 1. 1.9 Cr-Mo-Zn 1. 500 2. 120 Fuller earth 2. 660 1000 cm3 3. 3. 1.3 1. 2.5 Cottingham et al., 1957 CO-MO 1. 540 2.5-3.2 mm 2. 20.0 2. 200 3. 75.0 3. 1.5 1. 350) is substantially lower due to the higher values of the Peclet number (L/d,> -2O/Pe In (1- XN)). Determination of the axial dispersion in the TBR was not performed, and further analysis of the HDN data obtained with loadings B and C was performed with the plug flow assumption, keeping in mind all the mentioned remarks and discussion in concordance with the specific reaction conditions and flashing of the reaction mixture. T h e Performance of the TBR In the case of neglecting the influence of liquid holdup or wetting effects, i.e., for an ideal trickle-bed reactor, the mole balance for a first-order HDN reaction may be expressed as -[ln (1- XN)] = k/LHSV (1) The validity of this assumption was tested in a log-log plot of In (1/(1-XN)) versus 1/LHSV in which a straight line must appear with a slope equal to unity. The above analysis is given in Figure 5, for both catalyst loadings, and the results show that the slopes of these straight-line plots varied from 0.25 to 1.03. The changes in the slopes are dependent on temperature for both catalyst loadings. These results show that the conversion of nitrogen compounds in the TBR depends not only on the liquid flow rate but also on contact between the liquid phase and catalyst particles. That means that more realistic models for a TBR must be used in the interpretation of kinetic data, models such as those based on liquid holdup or effective wetting effects with the mole balance expressed in the form -[ln (1- X,)] = k'(l/LHSV)" (2) with theoretical values of the exponent CY equal to 0.66 or 0.68 given by Henry and Gilbert (1973) or Mears (1974)
1190 Ind. Eng. Chem. Res., Vol. 27, No. 7, 1988 ' ~
6.0 -
a-hexadecane b-decalin
--- lmding " B ' i
- loading "C" I ~
/37nc-a
2oT \ 1
12c
;r 0'-
I-
-''-
o decalin
- 1 2-
L ___-1 L 6
-J-d-LLLLLL
15 lILHSV, h
8 10
B
-+
17
16
lo3
ia
K
Figure 6. Arrhenius plot of apparent HDN rate constant
20
Figure 5. Analysis of experimental data using effective wetting model (eq 2).
based on data and a correlation derived by Puranik and Vogelpohl (1974). Only for experiments performed at higher temperatures (above 340 "C) could the models of an ideal trickle bed be used for the kinetic interpretation of HDN. At lower temperatures, the stronger effect of catalyst wetting could be expected, together with the influence of the catalyst particle diameter as predicted by the corresponding model. The influence of temperature on constant CY in eq 2 could be explained using the results of a flash calculation obtained with the Redlich-Kwong equation of state (Table IV) These results confirmed the influence of the catalyst wetting effect in the case when the reaction, to a large extent, takes place in the liquid phase.
Kinetics of HDN Determination of the apparent rate constant for the HDN of the test mixture containing only quinoline, indole, and lauronitrile was based upon the following statements: at smaller nitrogen conversion, the influence of backmixing could be neglected; for higher total conversion obtained at temperatures above 340 "C, the influence of backmixing could be also neglected, although considerable nitrogen conversion was obtained due to the effect of flashing and reactions in the vapor and liquid phases; the rates of HDN in the vapor and liquid phases are comparable and almost identical as shown by Satterfield and Yang (1984) in the case of quinoline HDN; at lower temperatures, the conversions in the pilot plant depend on the liquid flow rate usually expressed by the effective catalyst wetting model. The calculated values of the apparent rate constant for the simultaneous HDN reaction of quinoline, indole, and lauronitrile are presented in the Arrhenius plot (Figure 6). The effect of temperature on the HDN rate appears to obey the Arrhenius equation with an apparent activation energy of E = 103.4 kJ/mol. This value is in good agreement with other literature data for HDN reactions, e.g., those given by Shah and Paraskos (1975) of 60-85 kJ/mol for the HDN of shale oil and Satterfield and Yang (1984) for the vapor- and liquid-phase HDN of quinoline with independent values for different stages in the complex reaction mechanisms of 18-187 kJ/mol. Comparison between the experimental values of nitrogen conversion and those calculated by using eq 2, a = 0.68, and the corresponding reaction rate constant (Figure 6) for temperatures below 340 "C is shown in Figure 7. For higher tempera-
K,,exp
Figure 7. Comparison of experimental and calculated values of nitrogen conversion for HDN of test mixture (quinoline, indole, and lauronitrile).
tures, eq 1was used for the calculation of nitrogen conversion with the same temperature dependency of the rate constant, k . The average difference between the experimental and calculated values of nitrogen conversion for loadings B and C are 12.3% (28 data) or 5.2% for loading C and 19.3% for B. The greatest difference between the calculated and experimental values of XNwas obtained for 280 "C. A possible explanation for such poor prediction with the effective catalyst wetting model at 280 "C could be in the large difference in the reaction rate of different steps in the HDN of quinoline, indole, and lauronitrile. Actually, at moderate temperatures only the reaction of lauronitrile HDN proceeds to the final steps, giving ammonia and CIzas the final product. That means that the measured nitrogen conversion is more indicative of the specific overall reaction of lauronitrile HDN than the overall HDN of the nitrogen compounds in the test mixture. A similar observation and analysis was given by Chu and Wang (1982) for the interpretation of the kinetics of kerosene HDS. A higher k', apparent rate constant, for HDN was determined using data obtained with catalyst loading C as compared to loading B. It differs in the range of 22% at 340 "C and 28% at 310 "C. Considerably higher values of k', determined with the Ni-Mo catalyst rather than with the mixture of Co-Mo + Ni-Mo, agree with the results of some other authors on the efficiency of Co-Mo and Ni-Mo in the HDN process (Satterfield and Cocchetto, 1975; Stern, 1979; Harvey et al., 1986). However, additional correction of k is necessary because different sizes of the particles in loadings B and C were used in the present study. Very simple calculations were made in order to give information about the influence of d, on nitrogen conversion. Equivalent particle diameters of the order of 3.73
Ind. Eng. Chem. Res., Vol. 27, No. 7, 1988 1191 and 2.4 mm were calculated for catalyst loadings B and C, respectively. This calculation was based on data presented in Table I11 (pellets 5 X 5 mm, extrudate 1.5 X 3 mm, mass fraction of the catalyst in TBR). However, different models for trickle-bed reactors show a somewhat different dependence on catalyst size. In the case of the liquid holdup model, the correction factor for the particle diameter is (3.73/2.4)4.66 = 0.75 and for the effective catalyst wetting model (3.73/2.4)0.18= 1.08. By use of the above-derived correction factor, approximately equal values of k were obtained for the liquid hold-up model and an even larger percent difference than those calculated (28%) when the effective catalyst wetting model was tested. But considering also that a considerable difference in k'was also determined for loadings C and B (22%) at 340 "C, where ideal plug flow in the trickle bed was assumed and proved, it seems that the earlier conclusion about the greater activity of the Ni-Mo catalyst for HDN could still be supported.
Theoretical Prediction and On-Line Measured Values of Hydrogen Consumption The main goal in designing the microreactor laboratory unit was to obtain information on the actual hydrogen consumption in a steady-state operation process. The test mixture of model nitrogen compounds used in the HDN study enabled the checking of on-line hydrogen consumption measurements with calculated values on the basis of known HDN mechanisms for quinoline, indole, and lauronitrile. It is well-known that the HDN of quinoline and indole goes via the saturation of the hetero ring as the first step in the HDN reaction network (Stern, 1979; Satterfield and Yang, 1984). The HDN of quinoline proceeds mostly via decahydroquinoline (DHQ), which is the preferred reaction pathway (Satterfield and Cocchetto, 1981) and via propylaniline (PA) without previous benzene ring saturation. The principal hydrocarbon product of the HDN of quinoline is propylcyclohexane (PCH), but propylbenzene (PB) and propylcyclohexene were also detected in smaller quantities. The HDN of indole proceeds in an analogous way with ethylcyclohexane (ECH) as the main product (Rollman, 1977; Stern, 1979). The liquid produds of the HDN of the test mixture were analyzed by a GC-MS combination (Perkin-Elmer SIGMA 1 GC with a 30-m capillary column coupled with an H P 5970 B mass selective detector operated by a computer). For quantitative evaluation of the HDN reaction products, a Varian Vista Series GC was used with a flame ionization detector (FID). A 30-m-longfused silica megabore column provided full and reproducible resolution of all the peaks. The retention times and response factors for each component were determined experimentally with pure compounds. The product distributions in the HDN of quinoline and indole in decalin at 80 bar and 0.75-h-l LHSV versus temperature are shown in Figures 8 and 9. Figure 8 shows almost complete conversion of quinoline to 1,2,3,4-tetrahydroquinoline(Py-THQ) at 250 "C which is the optimal temperature for the formation of this type of intermediate. At temperatures above 250 "C, the concentration of the second intermediate, DHQ, starts to increase and reaches its maximum at temperatures of about 300 "C; at higher temperatures the concentration of PCH and PB gradually increases. In the case of indole HDN, Figure 9, similar effects were observed. The somewhat lower reactivity of indole as compared to quinoline was manifested at temperatures below 370 "C. The first intermediate product, indoline
0
80 1'97 0 '
/-\ ,\
A
Py-THO DHO PCH PB
r' /
O
\
J
I
/
TEMPERATURE,
OC
Figure 8. HDN of quinoline (Q).Product distribution versus temperature. Catalyst loading A. i IND ECH
70 -1
\
60- \
-8: L50'o
/
),,\ \
TEMPERATURE, O C
Figure 9. HDN of indole (I). Product distribution versus temperature. Catalyst loading A.
(IND), has a maximal concentration at 250 "C with a decreasing tendency at higher temperatures. The final product is mostly ECH, although ethylbenzene (EB) was also detected in low concentrations. In the case of lauronitrile, only n-dodecane was detected with complete conversion already at lower temperatures. The stoichiometry for quinoline, indole, and lauronitrile HDN could be written in simple form with the assigned hydrogen consumption in each step as HDN QUINOLINE
lPyTHQi
I DHO I
IPCHAI
IPCHI
HON INDOLE
UND!
[EA!
ECHAI
(ECHI
HDlv LAURONITRILE
+2H
CllH23C3N 2 C ~ ~ H ~ ~ C H Z3- N C H~ ~H 2 +6 NH3
The reaction pathways in Figures 8 and 9 show that at temperatures above 370 "C 7 mol of hydrogen is consumed per mole of quinoline and 6 and 3 mol of H2per each mole of indole and lauronitrile, respectively. However, below 310 "C, hydrogen is mainly consumed in hydrogenation reactions which precede the hydrogenolysis of the C-N bond. The product distributions in the HDN of quinoline and indole in decalin at 310 "C and 80 bar for various LHSV's for loadings B and C are presented in Figures 10-13. Comparison of the product distribution (Figures 10 and 12) confirms the earlier discussion about the influence of different catalysts on the HDN reaction rate. The con-
1192 Ind. Eng. Chem. Res., Vol. 27, No. 7, 1988 @-THO
ej
89
0
DHO
0
pcH+PB &-THO
\
90-
\
\
A I IND ECHA
0
, '
, C
1 / LHSV, h
Figure 10. HDN of quinoline (Q).Product distribution versus l/LHSV. Catalyst loading B.
Figure 13. HDN of indole (I). Product distribution versus 1/ LHSV. Catalyst loading C.
A I IND o ETHKCYCLOHEXWINE ECH+ES (ECHAI
N 701
I
,
IiLHSV, h
Figure 11. HDN of indole (I). Product distribution versus 1/ LHSV. Catalyst loading B.
;i
0 Py-THO o DHO
mtm
80
0 Sz-THQ
x, , % Figure 14. Hydrogen consumption versus total conversion of niComparison of on-line measurement trogen in test mixture (XN), with calculated values.
l/LIiSV h
Figure 12. HDN of quinoline (Q).Product distribution versus l/LHSV. Catalyst loading C.
version of the test mixture at 310 "C on catalyst loading C is much higher than in the case of loading B, probably due to the higher hydrogenation ability of the Ni-Mo catalyst as compared to the Co-Mo catalyst. The results of on-line hydrogen consumption measurementa and hydrogen consumption calculated from GC data (Figures 8-13) are presented in Figure 14. The plot of H, consumed during HDN of the test mixture versus nitrogen conversion, XN,shows an average difference of 25% which is higher at lower nitrogen conversion or lower absolute values of hydrogen consumption. That is in good agreement with the discussion on the accuracy of the hydrogen consumption determination on the basis of experimental data in the first stage of the sulfidization procedure. In addition, the results provide a good starting point for further experimental investigations of hydrogen consumption during the hydrogenation of other raw materials (shale oil, heavy crude, coal liquids), in which more than
200 m3/m3 of hydrogen consumption could be expected (Harvey et al., 1986; Sullivan and Stangeland, 1979; Beuther et al., 1959). It is also important to point out that during the HDN process some side reactions take place which give as products lighter and heavier compounds than these identified and reported here. These products were detected in a total amount of less than 3% but were not identified. More detailed information on the chemical structure of the heavier products in the HDN of quinoline was given by Satterfield and Yang (1984). Because of the possible importance of the heavier HDN products in coke formation (Yang and Satterfield, 1984), further analytical research on the identification of these compounds needs to be done.
Conclusions 1. The presented results demonstrate the ability of the hydrogenation unit to determine the kinetics and reaction networks of complex hydrogenation reactions such as the HDN of quinoline, indole, and lauronitrile. It also enables the rapid checking of catalyst activity and the testing of different catalyst types. The main task appointed in unit design, expressed as on-line hydrogen consumption measurement, was successfully fulfilled. The latter statement is very important for further investigations in the domain of synthetic oil production from different sources (shale oil, tar sands, coal liquids).
Ind. Eng. Chem. Res., Vol. 27, No. 7, 1988 1193 2. The calculated hydrogen consumption is about 25% smaller than the one experimentally obtained on-line. The above difference is related to the accuracy of hydrogen flow and gas composition determination but also indicates that hydrogen was consumed not only in the main HDN reactions outlined in the text but also in some unidentified side reactions. 3. The hydrogenation unit has shown in practice the possibility of automatic operation and easy removal of the catalyst bed for the purpose of testing different catalysts. In preliminary experiments, three different catalyst loadings were used: a mixture of Co-Mo, Ni-Mo, and pure Ni-Mo catalyst. A higher HDN reaction rate was obtained with Ni-Mo catalyst (about 20-30%), which is in agreement with literature data (Satterfield and Cocchetto, 1975; Harvey et al., 1986). 4. The overall HDN of the model nitrogen compounds (quinoline, indole, and lauronitrile) was analyzed using a simplified model for first-order kinetics. This model is frequently presented in the literature for HDN kinetics of different nitrogen compounds, although for specific reaction conditions the reaction rate of quinoline HDN is defined as nearly zero order (Satterfield and Yang, 1984). 5. The apparent activation energy for the simultaneous HDN of quinoline, indole, and lauronitrile was found to be E = 103.4 kJ/mol. Acknowledgment This work was done under the Agreement of Scientific Educational Cultural and Technical Cooperation between the Socialist Federal Republic of Yugoslavia and the Federal Republic of Germany. Financial support by the KFA International Bureau D 5170 Julich on the German side as well as the Research Fund of the Region Belgrade, Business Association for Exploration, Production and Exploitation of Oil Shales Belgrade, and SOUR Jugopetrol Belgrade on the Yugoslavian side is gratefully acknowledged. Nomenclature d = diameter of reactor, cm dp = diameter of catalyst particle, cm DL = axial dispersion coefficient, cm2/s E = activation energy, J/mol G L = superficial liquid velocity, kg/ (m2-s) k = apparent reaction rate constant, eq 1,h-’ k’ = apparent reaction rate constant, eq 2, h-’ LHSV = liquid hourly space velocity, h-l L = length of catalyst bed, cm L, = length of reactor, cm P = pressure, kPa Pe = Peclet number based on particle diameter (Pe = Udp/DL) t = temperature, O C T = temperature, K U = interstitial liquid velocity, cm/s XN = nitrogen conversion Greek Symbol a = exponent (eq 2) Abbreviations TBR = trickle-bed reactor HDS = hydrodesulfurization HDN = hydrodenitrogenation HDO = hydrodeoxygenation
HPS = high-pressure separator LPS = low-pressure separator Q = quinoline
DHQ = decahydroquinoline PyTHQ = 1,2,3,4-tetrahydroquinoline Bz-THQ = 7,8,9,10-tetrahydroquinoline P A = propylaniline PCH = propylcyclohexane PB = propylbenzene I = indole IND = indoline ECH = ethylcyclohexane ECHA = ethylcyclohexylamine EB = ethylbenzene LN = lauronitrile STP = standard temperature and pressure Registry No. Co, 7440-48-4;Mo, 7439-98-7; Ni, 7440-02-0; quinoline, 91-22-5;indole, 120-72-9; lauronitrile, 2437-25-4. Literature Cited Ahmed, M. M.; Crynes, B. L. “Refining of Syn-Crudes”. ACS Symp. Ser. 1979,179, 175. Badilla-Ohlbaum, R.;Pratt, K. C.; Trimm, D. L. Fuel 1979,58,309. Beuther, H.; Flinn, R. A.; McKinlel, J. B. Ind. Eng. Chem. 1959,51, 1349. Chu, C. I.; Wang, I. Ind. Eng. Chem. Process Des. Deu. 1982,21,338. Cottingham, P. L.; White, E. R.; Frost, C. M. Ind. Eng. Chem. 1957, 49,679. de Bruijn, A. Proceedings of the Sixth International Congress on Catalysis, The Chemical Society, of London, 1977,p 951. Doelman, J.; Vlugter, J. C. Proceedings of the Sixth World Petroleum Congress, Frankfurt/Main, Germany, June 1963,p 247. Eliszer, K. F.; Bhinde, M.; Houalla, M.; Broderick, D.; Gates, B. C.; Katzer, J. R.; Olson, J. H. Ind. Eng. Chem. Fundam. 1977,16,380. Furimsky, E.;Mikhlin, J. A.; Jones, D. Q.; Adley, T.; Baikowitz, H. Can. J. Chem. Eng. 1986,64,982. Harvey, T. G.; Matheson, T. W.; Pratt, K. C.; Stanborough, M. S. Ind. Eng. Chem. Process Des. Deu. 1986,25 521. Henry, H. C.; Gilbert, J. B. Ind. Eng. Chem. Process Des. Deu. 1973, 12, 328. Hook, B. D.; Akgerman, A. Ind. Eng. Chem. Process Des. Deu. 1985, 24,507. Kang, C. C. “Refining of Syn-Crudes”. ACS Symp. Ser. 1979,179, 193. Krishnamurthy, S.;Panvelker, S.; Shah, Y. T. AIChE J. 1981,27, 994. Kumar, M.; Akgerman, A.; Anthony, R. G. Ind. Eng. Chem. Process Des. Dev. 1984,23,88. Mears, D. E.Adv. Chem. Ser. 1974,133,218. Moore, P. K.; Akgerman, A. Fuel 1985,64,722. Puranik, S. S.;Vogelpohl, A. Chem. Eng. Sci. 1974,29,501. Rollman, L. D. J. Catal. 1977,46,243. Regtop, R. A.; Crisp, P. T.; Ellis, J. Fuel 1982,61,185. Saban, M. MS. Thesis, TMF Beograd, 1986. Satterfield, C. N.Heterogeneous Catalysis in Practice; McGraw-Hill: New York, 1980; pp 259-268. Satterfield, C. N.; Cocchetto, J. F. AIChE J. 1975,21,1107. Satterfield, C. N.; Cocchetto, J. F. Ind. Eng. Chem. Process Des. Dev. 1981,20,53. Satterfield, C. N.; Yang, S. H. Ind. Eng. Chem. Process Des. Deu. 1984,23,11. Satterfield, C. N.; Smith, C. M.; Ingalls, M. Ind. Eng. Chem. Process Des. Deu. 1985,24,1000. Shah, Y. T.; Paraskos, J. A. Chem. Eng. Sei. 1975,30, 1169. Sivasubramanian, R.; Crynes, B. L. Znd. Eng. Chem. Prod. Res. Deu. 1979,18,179. Stern. E. W. J. Catal. 1979. 57. 390. Sulliv’an, R. F.; Stangeland,’ B. E. “Refining of Syn-Crudes”. ACS Symp. Ser. 1979,179,25. Wilson, M. F.; Kriz, J. F. Fuel 1984,63,190. Wolfson, M. L.; Pelipetz, M. G.; Damick, A. D.; Clark, E. L. Ind. Eng. Chem. 1951,43,536. Yang, S. H.; Satterfield, C . N. J . Catal. 1983,81,168. Yang, S. H.; Satterfield, C. N. Ind. Eng. Chem. Process Des. Deu. 1984,23,20. Received for review July 8, 1987 Revised manuscript received January 26, 1988 Accepted February 10,1988