Ind. Eng. Chem. Prod. Res. Dev. 1985, 2 4 , 10-15
10
the surfaces of P t / P d particles may become enriched in P d in oxidizing environments because of the greater stability of Pd oxides (Chen and Schmidt, 1979). Conclusion Our results indicate that sintering and loss of metal surface area was primarily responsible for the decreased CO conversion activity of the aged catalyst examined here. In addition, our results suggest that Pd may have been poorly dispersed during the preparation of the catalyst. We hope that this report will stimulate further application of high-resolution analytical electron microscopy to the diagnosis of automotive catalyst deactivation under “real-world” conditions and to the assurance of proper catalyst preparation. Acknowledgment
James C. Schlatter, Robert M. Sinkevitch, and Patricia J. Mitchell peformed the CO conversion and BET measurements as part of an earlier project. Members of the
Analytical Chemistry Department of the General Motors Research Laboratories performed the elemental analyses and X-ray diffraction mewurements. Lawrence F. Allard, Jr., of the Department of Materials and Metallurgy at the University of Michigan provided assistance with the operation of the microscope. Registry No. CO, 630-08-0; Pt, 7440-06-4; Pd, 7440-05-3; Mn, 7439-96-5;P, 7723-14-0; Pb, 7439-92-1; S, 7704-34-9; Zn,7440-66-6. Literature Cited Chen, M.; Schmidt, L. D. J. Catal. 1979, 56, 198. D a b Betta, R. A,; McCune, R. C.; Sprys, J. Ind. Eng. Chem. Prod. Res. Dev. 1976, 15, 169. Furey, R. L.; Monroe, D. R. Society of Automotive Engineering Paper No. 811228, 1983. Pennycook. S. K.; Howie, A,; Shannon, M. D.; Whyman, R. J. Mol. Catal. 1983, 20, 345. Summers. J. C.; Chou, T. S. General Motors Research Memorandum No. 14-75, NOV 22, 1978. Taylor, K. C. Ind. Eng. Chem. Prod. Res. Dev. 1876, 15, 264.
Received for review June 8, 1984 Accepted October 25, 1984
Cracking Catalyst Deactivation Models Wllllam J. Hatcher, Jr. Department of Chemical Engineering, The Universl~of Alabama, University, Alabama 35486
This study uses the kinetics developed for cumene cracking to simulate cracking and coking in a fixed bed. The Froment-Bischoff method for modeling catalyst deactivation was used. The simulations suggest significant coke profiles along the reactor length. This result is in disagreement with the ‘‘timean-stream’’ method of modeling catalyst deactivation. Time-on-stream would predict a uniform deactivation in a tubular isothermal reactor since its deactivation equation does not include the local concentration of reacting species. Comparisons are made with experimental data for catalytic cracking cumene and hexadecane in fixed beds.
Introduction The reactions of hydrocarbons on zeolite catalysts are accompanied by the deposit of coke on the surface of the catalyst, which decreases the reaction rate. Several methods have been proposed to account for the loss of catalytic activity due to coke deposition. Butt (1972,1978) has thoroughly reviewed the literature on coking and catalyst deactivation. Since coke formation in catalytic cracking represents a loss of desired products and the subsequent catalyst regeneration supplies a major part of the energy requirements of the process, factors leading to coke formation are of great commercial importance. In his pioneering work, Voorhies (1945) demonstrated that the coke formed on cracking catalysts depended on ”time-on-stream” or process time. Since that time several authors such as Wojciechowski (1968) and Jacob et al. (1976) have proposed to relate relative catalyst activity to time-on-stream. Froment and Bischoff (1961,1962) have argued that a more fundamental approach would require the consideration of mechanisms leading to a set of simultaneous equations expressing the rates of formation of both the main product as well as coke. Also, a relationship between the amount of coke on the catalyst and the decrease in catalytic activity would be needed. DePauw and Froment (1975) and Dumez and Froment (1976) experimentally observed an exponential relation between the relative catalyst activity and coke content in n-pentane 0196-432 118511224-001 0$01.50/0
isomerization on a platinum reforming catalyst and in 1-butene dehydrogenation on a chromia-alumina catalyst. Lin et al. (1982, 1983) also observed an exponential relationship for cumene cracking on a La-Y zeolite catalyst. A model in which coke or coke precursors were formed from both cumene as well as propylene product gave the best fit of the experimental data. Apparently, at low temperatures most of the coke forms from cumene feed, while at higher temperatures most of the coke forms from propylene. The same relative catalyst activity-coke relationship was found for the coking reaction as for the main reaction, suggesting that the coking reaction occurs on the same sites as the cracking reaction. Beeckman and Froment (1979,1980,1982) have investigated mechanisms that could explain the functional relationship between coke formation and the loss of catalyst activity. Their basic theory relates deactivation to site coverage and pore blockage by coke growing from the active site where it was deposited. When the rate of growth of coke and the rate of site coverage are of the same order of magnitude, the relative activities of the main and coking reactions are no longer identical. Cumene Cracking Kinetics. Previous studies on LaY zeolite and rare earth exchanged Y catalysts in both integral and differential reactors determined that cumene cracking was best represented by a single site, surface reaction controlling mechanism (Hatcher et al., 1979) 1985 American Chemical Society
Ind. Eng. Chem. Prod. Res. Dev., Vol. 24, No. 1, 1985
k IOKALC (PA- PRPS /K)
r=
1
(1)
+ KAPA + KRPR
This study was made at very low cumene partial pressures and low cracking temperatures so that coking and catalyst deactivation were negligible. Best and Wojciechowski (1973,1977) and Campbell and Wojciechowski (1971) have reported similar expressions for cumene cracking at higher temperatures and partial pressures. Studies of cumene cracking in a thermobalance reactor in our laboratory (Lin et al., 1982,1983) accounted for the coking reactions as well as the main reaction. The kinetics of cumene cracking and coking were represented by combined parallel and series coking reactions A
e
R
+
11
Table I. Kinetic Parameters Cumene Cracking 200 OC 500 O C klKA, kmol/kg-s 3.25 X 1.08 x 10-3 KA, bar-' 162 0.9 KR, bar-' 83.8 0.45 K 5.48 x 10-4 6.74 6.43 x 10-4 kC3l kglkg-s 3.54 x 10-3 7.17 x 10-3 kC41 kg/kg-s 1.0 x 10-9 u 20.8 7.2
S
\/ C
The initial rate expressions in the absence of coke for cracking and coking, respectively, were r0
=
~ K A ( ~- 'PRPs/K) A 1 + KAPA + KRPR
r: = r,: r: =
kc3KAPA
1
(2)
+ rC2O
0
(3b)
+ KAPA + KRPR
6
4
8
10
1, cm
Figure 1. Cumene cracking. Conversion vs. bed length: 200 " C , tf = 7680 s; curve 1 space time = 3277 s; 2, 820 s; 3, 205 s.
(34
+ I2dKRPR
1
I
I
,
0
2
4
I
.05
To account for deactivation of the main and coking reactions, the rates were determined as
r = rO@ rc = r,O@
(0 I@ I1)
(4)
(0 5 9 I1)
(5)
The deactivation function, or relative catalyst activity, was found to be
CP = exp(-aC,) At differential reactor conditions, the coking pression can be integrated to give
' [
C,=-ln ff
1+
ff(&KApA + k&RPR)t 1 + KAPA + KRPR
If eq 7 is substituted into eq 6, then
(~&APA + kc.&RPR)t 1 + KAPA
+ KRPR
I
This form of .the deactivation function is very similar to forms used in the time-on-stream approach to cracking catalyst activity decay. It should be noted, however, that the relationship in eq 8 applies only to differential reactor conditions in which the partial pressures of the reacting species are essentially constant. In the thermobalance studies most of the coke appeared to form from cumene a t the lower reaction temperatures and from propylene a t the higher reaction temperatures. Therefore, the Froment-Bischoff catalyst deactivation model would predict an ascending coke profile in a tubular reador at low temperatures and a descending one at higher temperatures. The time-on-stream model would predict a uniform profile at all conditions. Simulation of Cumene Cracking in a Fixed Bed The Froment-Bischoff deactivation model and the cumene cracking and coking kinetics were used in a fixed
6
8
IO
Lcm
Figure 2. Cumene cracking. Coke content vs. bed length: 200 "C, tf = 7680 s; curve 1 space time = 3277 s; 2, 820 s; 3, 205 s.
bed reactor simulation. The simulation uses a basic onedimensional model with no interfacial gradients. When the deactivation rate is relatively slow compared to the cracking rate, the reactor may be considered to operate in the quasisteady state and the accumulation terms may be dropped from the continuity equations. This leads to the following continuity equations for cumene, propylene, and benzene, respectively
#A/dz
= PBfl(-r - rCi/MA)
(9)
The set of differential equations (9-11) were solved numerically using the Runge-Kutta fourth-order method. The coke formation was found by finite differences. The computation procedure was to solve for gas composition a t each axial position a t time zero through the solution to eq 9-11. Then time was incremented one step, and the incremental coke formed on the catalyst at each axial position was computed by eq 12 and the gas compositions by eq 9-11. The incremental coke formed at each axial position was then added to the previous coke on the catalyst. Thereafter the process of incrementing time and
12
Ind. Eng. Chem. Prod. Res. Dev., Vol. 24, No. 1, 1985
,
1
2
0
0
2
4
6
8
I
4
Figure 3. Cumene cracking. Catalyst activity: 200 OC, tf = 7680 s; curve 1 space time = 3277 s; 2, 820 s; 3, 205 s.
6
10
Z,cm
IO
1,cm
6
Figure 6. Cumene cracking. Catalyst activity: 200 "C, T = 820 s; curve 1 time on stream = 1920 s; 2, 3840 s; 3, 5760 s; 4, 7680 s. 1
/A/ /
"lC
0.2 0
2
6
4
8
10
Lcm
Figure 4. Cumene cracking. Conversion vs. bed length: 200 OC, T = 820 s; curve 1time on stream = 1920 s; 2,3840 s; 3,5760 s; 4,7680 9.
;/
0.1
0
/ , 2
6
4
8
18
Z,Cm
Figure 7. Cumene cracking. Conversion vs. bed length: 500 "C, tf = 240 s; curve 1 space time = 4.28 s; 2, 2.14 s; 3, 1.07 s; 4, 0.54 s.
0
2
6
4
8
10
1,cm
Figure 5. Cumene cracking. Coke content vs. bed length: 200 "C, T = 820 8; curve 1 time on stream = 1920 s; 2, 3840 s; 3, 5760 s; 4, 7680 s.
0
2
6
4
8
10
1,cm
marching across the axial positions was repeated until the total time was completed. Values of the kinetic parameters used in the simulations are reported in Table I. The results of simulations a t a very low reaction temperature, 473 K, are shown in Figures 1-6. These figures illustrate the effects of space time and process time on conversion, coke formation, and relative catalyst activity. At a cumene feed partial pressure of 1 bar, equilibrium conversion at 473 K is approximately 2%. The simulations shown in Figures 1-6 were calculated for a cumene feed partial pressure of 0.005 bar. At these conditions equilibrium conversion is approximately 30%. As can be seen in Figure 1, the simulation a t the lowest feed rate ( 7 = 3277 s) resulted in a close approach to equilibrium and a uniform composition profile in the last half of the reactor. Figure 2 shows that the coke profile is also uniform in the same region. As expected, since the parallel coke formation is the predominant one at 473 K, the coke profiles descend along the length of the reactor.
Figure 8. Cumene cracking. Coke content vs. bed length 500 "C, tf = 240 s; curve 1 space time = 4.28 s; 2, 2.14 s; 3, 1.07 s; 4, 0.54 s.
The rate of coke formation is relatively slow at 473 K and a cumene feed partial pressure of 0.005 bar. The maximum coke on catalyst was less than 0.05 after over 2 h on stream. Even so, the effect of the coke on the relative catalyst activity was pronounced as illustrated by Figure 3. In Figure 4 the effect of process time on conversion is shown. At very early times the final conversion approached equilibrium, but it diminished due to catalyst activity loss with coke formation. Figure 5 illustrates the buildup of coke profiles with process time, and Figure 6 shows the change in relative catalyst activity profiles with time. The simulations at 773 K shown in Figures 7-12 are based on a cumene feed partial pressure of 1bar. Equilibrium conversion is approximately 93%. Figures 7-9 are for a relatively short process time of 240 s. The coking rate
--
Ind. Eng. Chem. Prod. Res. Dev., Vol. 24, No. 1, 1985 13
I
I
4
8.d
I
et O
I
4
2
8
1
I
8
10
0 .I
1,cm
Figure 9. Cumene cracking. Catalyst activity: 500 OC, tf = 240 s; curve 1 space time = 4.28 s; 2, 2.14 E; 3, 1.07 E; 4, 0.54 s.
2,cm
Figure 10. Cumene cracking. Conversion vs. bed length 500 OC, 7 = 2.14 E; curve 1 time on stream = 60 E; 2, 120 s; 3, 180 E, 4,240 E.
I
01
I
I
,
2
I
I
1
I
I
I
4
6
8
LO
(
1,cm
Figure 11. Cumene cracking. Coke content w. bed length 500 OC, 7 = 2.14 E; curve 1 time on stream = 60 s; 2, 120 s; 3, 180 E; 4, 240 E.
is extremely fast at 773 K and 1bar and coke levels up to 0.23 were predicted. At 773 K, the coking kinetics had been best fit by a model in which most of the coke formed from propylene. These kinetics result in the ascending coke profiles shown in Figure 8 and the descending relative catalyst activity profiles in Figure 9. It is interesting to note that mean reactor catalyst activity was only 25 to 45% of fresh catalyst activity after 4 min. T h e effects of process time are shown in Figures 1&12. The final conversion in Figure 10 was 0.54 after 60 s and dropped to 0.37 after 240 s. Coke profiies increasing with process time and relative
10
1 ,cm
Figure 12. Cumene cracking. Catalyst activity: 500 O C , 7 = 2.14 s; curve 1 time on stream = 60 E; 2, 120 E; 3, 180 s; 4, 240 s.
Figure 13. Cumulative conversion as a function of time on stream. Cumulative cracking at 430 OC: cat./Reagent: (v)0.016; (0) 0.008; (0) 0.004. Table 11. Kinetic Parameters in Figure 13 Simulation 430 "C klKA,kmol/kg-bar-s 0.5 x 10-3 KA,bar-' 1 KR,bar-' 0.5 K 1.538 kc3, kg/kg-s 1.5 X lo4 k a t kg/kg-s 4.5 x 10-3 a 50
catalyst activity profiles decreasing with process time are shown in Figures 11 and 12. Wojciechowski and co-workers, for example, Best and Wojciechowski (1977), have made extensive experimental studies of cumene cracking in fixed bed reactors. These studies use a reaction rate expression for the main reaction somewhat similar to eq 2 and a time-on-stream mechanism to account for catalyst activity decay. The experimental results are compared to the kinetics model by plotting the time averaged conversion R = Jr d t / J dt, vs. total time on stream. The plots have parametric curves for constant catalyst-to-reagent weight ratios. The product of the catalyst-to-reagent ratio times the total time on stream is the space time based on weight ( 7 ) . Using the cumene cracking and coking kinetics developed in our laboratory, simulations were made at conditions similar to those used experimentally by Best and Wojciechowski. Simulation results are illustrated by the curves shown in Figure 13, and the data points are from the experimental study of cumene cracking in a fixed bed by Best and Wojciechowski (1977). Kinetic parameters used in this simulation are shown in Table 11.
14
Ind. Eng. Chem. Prod. Res. Dev., Vol. 24, No. 1, 1985
0.06
1
1
I
0 '
0
04
02
0 6
08
10
Z I L
Figure 14. Weight fraction coke on catalyst as a function of bed length: n-hexadecane cracking a t 500 "C; tf = 3600 s. Numbers indicative v/v/h.
Simulation of Hexadecane Cracking Eberly et al. (1966) determined the coke on catalyst as a function of time-on-stream and space velocity a t 773 K €or cracking n-hexadecane. The data covered a range of 13.2 to 20 v/v/h (volume of feed per volume of catalyst per hour) liquid hourly space velocity and 1 to 60 min on stream. Some dependence of coke on space velocity was clearly illustrated. The curves of coke vs. space velocity at a fixed time on stream exhibit maxima. Voorhies (1945) observed no dependence of coke on space velocity, but his study was over a very narrow range of 0.6 to 1.2 v/v/h. At a given space velocity the increase in coke as a function of time-on-stream followed the relationship first reported by Voorhies (1945)
c, = at"
~~
(13)
where a and n are constants. Along the accumulated length of a fixed bed, the effective space velocity decreases from a very high value near the bed inlet to that of the total reactor at the outlet. Using this principle, Eberly and co-workers (1966) calculated the coke on catalyst as a function of bed length. At a given space velocity, a maximum occurred in the coke curve, and with increasing space velocity the position at which the maximum occurred moved toward the outlet of the bed. This fact was confirmed by conducting separate experiments in which various fractions of the bed were analyzed for coke content. A similar relationship of coke formation vs. space velocity was observed by cracking East Texas light gas oil a t 823 K. In the present study, the coke and conversion data reported by Eberly were used to fit the Froment-Bischoff deactivation model and the following reaction network ti
- x
P
4
Conclusions Coking and catalyst deactivation mechanisms in catalytic cracking are phenomena that are not well understood. Experimental studies in integral tubular reactors to measure actual coke profiles and conversion profiles could yield fundamental coking parameters. This information would be valuable in developing a better understanding of catalytic cracking kinetics. Nomenclature A = cumene C = cpke C, = coke content of the catalyst, kg/kg FA = flow rate of cumene, kmol/s F R = flow rate of propylene, kmol/s Fs = flow rate of benzene, kmol/s H = n-hexadecane K = equilibrium coefficient for cumene dealyklation reaction K , = reaction rate coefficient in hexadecane cracking, kmol/ kg-s K , = reaction rate coefficient for coke formation from hexadecane, kg/ kg-s K3 = reaction rate coefficient for coke formation from coke precursor in hexadecane cracking, kg/ kg-s K 4 = reaction rate coefficient for formation of cracked products, hexadecane cracking, kmol/kg-s k , = reaction rate coefficient for cumene dealyklation, kmol/ kg-s k,O = surface rate coefficient at zero coke content, kmol/active site-s KA = adsorption equilibrium coefficient for A, atm-' kc3 = rate coefficient of coke formation from cumene, kg/kg-s kr4 = rate coefficient of coke formation from propylene, .~ kg/kg-s KR = adsorption equilibrium coefficient for R, atm-' L = fixed bed reactor length, m L, = total concentration of active sites, sites/kg MA = molecular weight cumene MR = molecular weight propylene P = coke precursor in hexadecane cracking PA = partial pressure of A, atm PR = partial pressure of R, atm Ps = partial pressure of S, atm r = rate of cumene dealkylation, kmol/kg-s R = propylene rO = initial rate of cumene dealkylation at zero coke, kmol/kg-s r: = initial rate of coke formation, kg/kg-s r,: = initial rate of coke formation from cumene, kg/kg-s rcZo= initial rate of coke formation from propylene, kg/kg-s rcl = rate of coke formation from cumene, kg/kg-s rc2 = rate of coke formation from propylene, kg/kg-s S = benzene t = time, s X = fractional conversion of cumene 2 = axial distance, m N = deactivation coefficient pB = bulk density of catalyst bed, kg/m3 Q, = deactivation function = relative catalyst activity T = space time, on mass basis = kg of cat./kg of feed/s CI = reactor cross sectional area, m2 Registry No. Cumene, 98-82-8; hexadecane, 544-76-3;carbon,
cracked products
where H is hexadecane feed, P is coke precursor, and C is coke. Several authors, Agudo et al. (1981), for example, have proposed a dimer inteTmediate in cracking alkane molecules to account for product molecules heavier than the feed. Their proposal was used to develop the reaction network for hexadecane cracking. Figure 14 illustrates the model of hexadecane cracking at 773 K and 60 min. As can be seen, the model predicts a maximum coke content near the reactor inlet at low space velocity, and the location of the maximum moves toward the outlet as space velocity is increased. In this simulation, the ratio of K3 to K2 was 9 and the ratio of K1 to K2 was 2. There was good agreement between the simulations represented by the curves and Eberly's data indicated by the data points.
7440-44-0.
Literature Cited Agudo, A. L.; Asensio, A,; Corma, A. J . Catal. 1981, 69, 274. Beeckman, J. W.; Froment. G.F. Ind. €ng. Chem. Fundam. 1979, 78, 245. Beeckman, J. W.; Froment, 0.F. Chem. Eng. Sci. 1980, 35, 805. Beeckman, J. W.; Froment, G. F. Ind. Eng. Chem. Fundam. 1982, 21, 243. Best, D. A.; Wojciechowskl, B. W. J . Cab/. 1973, 37,74. Best, D. A,; Wojciechowski, B. W. J . Catal. 1977, 4 7 , 343. Butt, J. B. A&. Chem. Ser. 1972, No. 109, 259. Butt, J. B. ACS Symp. Ser. 1978, No. 72, 288. Campbell, D. R.; WoJckchowski,B. W. J . Catal. 1971, 23, 307. DePauw, R.; Froment, G. F. Chem. Eng. Sci. 1975, 3 0 , 789. Dumez, F.; Froment, G. F. Ind. Eng. Chem. Process Des. Dev. 1976, 15, 291. Eberiy, P. E.; Klmberlin, C. N. Miller, W. H.; Drushel, H. V. Ind. Eng. Chem. Process Des. Dev. 1986, 5, 193.
15
Ind. Eng. Chem. Prod. Res. Dev. 1985, 2 4 , 15-19 Froment, 0. F.; Bischoff, K. B. Chem. Eng. Sci. 1061, 16, 189. Froment, G. F.; Blschoff, K. B. Chem. Eng. Sci. 1062, 77, 105. Hatcher, W. J.; Park, S. W.; Lin, C. C. 88th Natlonal Meeting of AIChE, Houston TX, April 1979, Paper 72a. Jacob, S. M.; Gross, B.; Voltz, S. E. and Weekman. V. W. AIChE J. 1076,
Voorhies, A. Ind. Eng. Chem. 1045, 3 7 , 318. Wojciechowski, B. W. Can. J. Chem. Eng. 1068, 4 6 , 48.
Received for review June 19, 1984 Accepted November 19, 1984
-- , 71-11 - .. 93
I
Lin, C . C.; Hatcher, W. J. ACS Symp. Ser. 1082, 796, 249. Lin, c. c.; Park, s. w.; Hatcher, w. J. Ind. Eng. Chem. process Des. DW. 1083, 22, 609.
Presented at the National Meeting of the AIChE, Anaheim, CA, May 1984.
Hydrous Metal Oxide Ion Exchangers for Preparation of Catalysts for Direct Coal Liquefactiont Howard P. Stephens,* Robert G. Dosch, and Frances V. Stohl Sandia National Laboratories, Albuquerque, New Mexico 8 7 785
A group of hydrous oxide ion-exchange compounds which can be used to prepare hydrogenation catalysts by a novel synthesis route has been identified. These materials offer several advantages for catalyst formulation, including high surface area, dispersion of any metal or mixture of metals and promoters over a wide concentration range, and adjustment of substrate acidity as well a s active metal oxidation state. Catalysts prepared by exchanging the sodium ion of sodium hydrous titanate for Ni, Mo, and Pd were tested for use in direct coal liquefaction. Results of hydrogenation experiments performed with the addition of the catalysts to a slurry of a bituminous coal in a coalderived solvent demonstrated that hydrous titanate catalysts with active metal loadings as low a s 1% were equally effective for liquefying coal as a commercial Nl-Molalumina catalyst containing 15 % by weight active metals.
Introduction Interest in efficient conversion of coal to liquid fuels has encouraged exploration of catalytic materials for direct coal liquefaction. We have identified a group of hydrous oxide ion-exchange compounds of Ti, Zr, Nb, and Ta which can be used to prepare hydrogenation catalysts by a novel synthesis route involving exchange of active metals into these compounds. Hydrous oxide ion-exchange compounds have previously been investigated a t Sandia National Laboratories for use in decontamination of aqueous nuclear waste (Dosch, 1978, 1981) and as precursors for ceramic materials (Hankey et al., 1981). A number of properties of the compounds suggested their use as substrates for catalyst preparation: (1)any metal or mixture of metals can be uniformly dispersed in the materials over a wide concentration range by a simple process; (2) the materials have high surface areas; (3) they exhibit good chemical stability; (4) solution chemistry or high-temperature reactions can be used to provide active metal oxidation state control; ( 5 ) acidity and basicity of the substrate can be modified by ion exchange; and (6) the catalysts can be prepared on transition metal supports (Ti, Nb, and Ta) known to undergo strong metal support interactions (Tauster et al., 1981). Although these properties suggested versatility of hydrous oxide ion exchangers for catalyst preparation, it was not known if Catalysts synthesized by exchanging ions of active metals into the material would exhibit hydrogenation activity. The purpose of the experiments reported here was to explore the hydrogenation activity of these materials for slurry phase catalysis of direct coal liquefaction. It is also possible that hydrous metal oxide ion This work supported by the U.S.Department of Energy at Sandia National Laboratories under Contract DEAC04-DP00789. 0196-4321/85/1224-0015$01.50/0
exchangers may be used to prepare multifunctional hydroprocessing catalysts as well as catalysts for other processes such as oxidation, dehydrogenation, methanation, alkylation, and isomerization. The hydrous oxides investigated as active metal catalyst supports belong to a group of inorganic ion-exchange compounds represented by the empirical formula M(M’,O HJn where M is an exchangeable cation and M’ is Ti, N i , Zr, or Ta. Compounds in which M is a quaternary ammonium ion, for example the tetramethylammonium ion, are very soluble in water. Vapor pressure osmometry measurements on benzene extracts of these solutions give average molecular weights of 5000-15000 for the associated species. Thus, the solution chemistry, stoichiometry, and molecular weights of the compounds suggest empirical formulas such as [ (CH,),NTi,O,H], where n varies from 20 to 60. Upon dissolution in water, the tetramethylammonium compounds dissociate into ionic moieties, (CHJ4N+and Ti205H-,as indicated by the high electrical conductivities of the solutions. The anions can react with a solution of aqueous metal cations to form amorphous precipitates with combining ratios of one Ticontaining anion per unit charge on the cation. Although the structures of these precipitates have not been elucidated, they are thought to consist of a network formed by metal-oxygen and metal-hydroxide bridges which incorporate cations as illustrated in Figure 1. Details of the preparation of these compounds are described in the following section. Experimental Section Catalyst Preparation. The titanate system, the best characterized and least expensive of the hydrous oxides, was chosen for exploratory testing. Hydrous titanate catalysts were prepared by a technique which consists of synthesis of sodium hydrous titanate ion-exchange material 0 1985 American Chemical Society