Catalytic reduction of carbon monoxide with hydrogen sulfide. 3. Study

Michael A. Henderson, William S. Epling, Craig L. Perkins, and Charles H. F. Peden , Ulrike Diebold. The Journal of Physical Chemistry B 1999 103 (25)...
0 downloads 0 Views 608KB Size
J. Phys. Chem. 1986, 90, 3132-3136

3132

Catalytic Reductlon of CO with Hydrogen Sulfide. 3. Study of Adsorption of Op, CO, and CO Coadsorbed with H2Son Anatase and Rutile Using Auger Electron Spectroscopy and Temperature-Programmed Desorption D. D. Beck: J. M White,* Department of Chemistry, University of Texas, Austin, Texas 78712

and C . T. Ratcliffet Exxon Research and Engineering Co., Annandale, New Jersey 08801 (Received: June 28, 1985)

In O2and CO adsorption on anatase, only one weakly bound molecular desorption state was observed. For CO on rutile, there was a strongly bound and a weakly bound state. For O2on rutile, a weakly bound state and two strongly chemisorbed states were observed. These strongly bound states are associated with the surface lattice anion vacancies produced on rutile (1 10). The amount of chemisorption in one of the strongly bound oxygen desorption states is correlated with the initial rate of irreversible adsorption of H2S on rutile. Coadsorption of CO and H2S indicates that strongly chemisorbed CO interacts with strongly chemisorbed H$3 to yield intermediates which desorb as CH3SH and CH4at T 800 K. At higher temperatures, the surface sulfide concentration dominates the -SH concentration, explaining the dominance of COS in the product mixture. Implications for commercial hydrodesulfurization catalysts are discussed.

-

Introduction Catalysts active in hydrodesulfurization play an important role in the refining of fossil fuels. Common commercial hydrodesulfurization catalysts are usually based on sulfided Moo3 or W 0 3 and have been the subject of considerable study.'-3 Efforts to identify the active sites and the role of promoters have led to the proposal of numerous models, which have yet to be resolved. Sulfided Ti02also catalyzes the reduction of nitrobenzene with H2S4and the reaction of CO with H2Sto form CH3SHand CH4.5 Rutile is more active than anatase in the reaction of CO with H 3 to form mercaptans and CH4, and pretreatment of the catalyst in H2S at elevated temperatures (-673 K) is required for catalyst a ~ t i v a t i o n .A ~ combined TPD (temperature-programmed desorption) and AES (Auger electron spectroscopy) study points to surface oxygen vacancies in a particular orientation on the rutile (1 10) crystal face as active sites for HIS chemisorption and subsequent decomposition.6 The surfaces of T i 0 2powders' and single have been extensively studied by using adsorbates other than H2S. Rutile is more easily reduced than anatase: and lattice oxygen segregates to the surface of r ~ t i 1 e . l ~ In the present study, 0,and CO chemisorption on anatase and rutile were studied by TPD. CO/H,S coadsorption was studied by TPD to gain insight about the active intermediate in the reduction of CO with H,S to yield hydrocarbon products. Experimental Section Experiments were performed in a two-chamber ultrahighvacuum system described p r e v i ~ u s l y . ~All ~ ~TPD ~ ~ ' spectra ~ were obtained at a heating rate of 2 K/s unless otherwise stated. AES data were obtained with a primary electron beam energy of 3 keV and a beam current ranging from 1 to 4 PA. The beam was defocused on the sample and turned on only when recording a spectrum in order to minimize beam damage effects. Charging effects were reduced by grounding the wire mesh that supported the powdered sample.6 The TiO, catalysts used here were characterized in a previous report.6 The anatase sample was synthesized from titanium isopropoxide and was designated A2, while rutile was synthesized from TiCl, and designated R2. Both the rutile and anatase samples contain significant amounts of surface carbon (C/Ti 10.25 as detected by AES). TPD re-

'

Present address: Physical Chemistry Department, General Motors Research Laboratories, General Motors Technical Center, Warren, MI

4x090-905s. ... . . ... *Present address: UNOCAL, Science and Technology Division, P.O. Box 76, Brea, CA 92621.

0022-3654/86/2090-3132$01.50/0

moves some of this carbon as CO and C 0 2 , but removal of the remainder requires treatment in O2at relatively high temperatures (>800 K). Adsorption experiments were performed with the cleanest surface achieved, defined as a C/Ti AES intensity ratio below 0.05. Before adsorption, most of the surface water was removed by several heating-cooling cycles between 140 and 873 K. Such treatments did not reduce the surface (detected by AES).17318 Gases used in this study (H2S, CO, and 0,) were obtained from Linde in the CP grade form and were further purified in a vacuum manifold by using the freeze-pumpthaw technique. Oxygen18-labeled 0,(99%) was obtained from Stohler, and oxygen18-labeled C O (99.97%) was obtained from Alfa-Ventron.

Results O2 TPD. Figure 1 shows O2TPD spectra after dosing a rutile sample (R2) with O2at 1 X lo-* Torr and 140 K for various times. Two principal desorption peaks were observed at T , = 164 K (state I) and T , = 416 K (state 11). Both are first-order with desorption activation energies of 5 8 kcal/mol (state I) and 26 kcal/mol (state 11). The former value is considered an upper limit

(1) F. E.Massoth, Adu. Catal., 27, 265 (1978). (2) P. Grange, Catal. Rev.-Sci. Eng., 21, 135 (1980). (3) P. C. H. Mitchell, Spec. Period. Rep.: Catalysis, 4, 175 (1981). (4) C. T.Ratcliffe and G. Pap, U S . Patent 4 115 523, 1978. (5) C. T. Ratcliffe and P. J. Tromp, US.Patent 4517 171, 1985. (6) D. D.Beck, J. M. White, and C. T. Ratcliffe, submitted for publication to J. Phys. Chem. (7) G. D. Parfitt, Prog. Surf. Membr. Sci., 11, 181 (1976). (8) J. Woning and R. A. Van Santen, Chem. Phys. Lett., 101,541 (1983). (9) Y.W. Chung, W. J. Lo, and G. A. Somorjai, Surf. Sci., 64, 588 (1977). (10) Y.W. Chung, W. J. Lo, and G. A. Somorjai, Surf. Sci., 71, 199 (1978). (11) M.L. Knotek, Surf. Sci., 91,L17 (1980). (12) J. B. Bates, J. C. Wang, and R. A. Perkins, Phys. Reo. B: Condens. Matter, 19,4130 (1979). (13) V. E. Henrich and R. L. Kurtz, Phys. Reu. E Condens. Matter, 23, 6280 (1981). (14) M.A. Vannice, P. Odier, M. Bujor, and J. J. Fripiat, presented to the 188th National Meeting of the American Chemical Society, Philadelphia, Aug. 1984. (15) D. D. Beck and J. M. White J. Phys. Chem., 88, 2764 (1984). (16) D. D. Beck, Ph.D. Dissertation, University of Texas, 1985. (17) D. P. Griffis and R. W. Linton, Surf. Interface Anal., 6 , 15 (1984). (18) L. E.Davis, N. C. MacDonald, D. W. Palmberg, G. E. Riach, and R. E. Weber, Handbook of Auger Electron Spectroscopy, Perkin-Elmer Corp., Physical Electronics Division, Eden Prairie, MN, 1976, pp 59-61.

0 1986 American Chemical Society

The Journal of Physical Chemistry, Vol. 90, No. 14, 1986 3133

Catalytic Reduction of CO with H2S

II

-1

5

L

= 0 5 0 1 L

140

I

I

I

1

200

300

400

500

600

T (K)

Figure 1. TPD spectra of O2for various doses at 140 K on rutile. Doses are labeled in langmuirs (1 langmuir = 1 X lod Torr s).

r

I

I

I

I

I

I

I

I

300

500 T (K)

700

300

1 X 0.025

900

Figure 2. TPD spectra of O2dosed at 1 X 10” Torr for 300 s at 140 K on a sulfided rutile sample. The sample was sulfided by predosing with H2S at 1 X lo-’ Torr for 2000 s at 673 K.

as the state I desorption peak temperature would probably decrease if a lower adsorption temperature were used. When the rutile sample was heated under vacuum at T > 800 K, before dosing 0,at 140 K, the intensity of state I1 increased and was sensitive to the heating time. A third oxygen desorption state (state 111; T, 445 K) was also detected as a shoulder of state 11. The intensities of states I11 and I were far less dependent upon annealing pretreatments than state 11. When the same experiments were performed with anatase (A2), only a low-temperature desorption peak analogous to state I on rutile was observed ( T , = 180 K a t saturation). In another experiment, rutile (R2) was dosed with I8O2at 1 X 10” Torr and 140 K for 344 s (saturation amount of chemisorbed oxygen on the surface). Very little isotope exchange with was surface oxygen took place in the TPD spectrum. (1sO’60 detected with T, = 500 K, but its intensity was 0.07 that of the I8O2peak.) Oxygen desorption at T > 800 K contained no lSO. There was very little isotope exchange even when an I8O2dose was followed by an I6O2dose. Adsorption of H2S and 0,.In another experiment, rutile (R2) was sulfided6 by exposure to H2S a t 1 X Torr for 2000 s at 673 K and was then exposed to 0,at 1 X 10” Torr for 300 s at 140 K. The subsequent TPD (Figure 2) shows O2desorbing in state 111 and H2S and SO2desorbing at higher temperatures. No state I1 O2desorbs. A low-temperature reaction to form water

-

200 (ARBITRARY U N I T S )

Figure 3. Initial slope of sulfur uptake measured by AES as a function of O2 TPD peak area. O2was dosed at 1 X 10“ Torr for 300 s at 140 K on clean rutile. After TPD, the sample was exposed to H2S at 1 X lod Torr and 673 K and monitored with AES during the treatment. The four experiments shown correspond to evacuation times at 1000 K ranging from 60 to 3600 s.

3

100

100 AREA O F O 2 TPD

140

200

300

400

500

600

T (K)

Figure 4. TPD spectra of CO for various doses at 140 K on rutile. Doses are labeled in langmuirs.

is evident in the small H 2 0 TPD peak coincident with the 0, peak. There are also a number of high-temperature reaction products. Oxygen chemisorption has been used in other studies as a probe for active sites in hydrodesulf~rization’~*~~ and has suggested the following experiment. A rutile sample (R2) was annealed under vacuum at 1000 K for various lengths of time ranging from 60 to 3600 s, cooled to 140 K, and dosed with O2at 1 X 10” Torr for 300 s. After O2 TPD, the S Auger signal (152 eV) was measured as a function of H2Sexposure at 1 X 10” Torr and 673 K. This p r d u r e was repeated for various annealing times under vacuum, using a new clean rutile sample each time. The area of the high-temperature oxygen desorption state IT correlated with the initial growth in the sulfur AES (Figure 3). This sulfur is identified with irreversibly adsorbed H2S. CO TPD. The TPD results of various CO doses on clean rutile (R2) at 140 K are shown in Figure 4. There are two desorption peaks ( T , = 175 and 437 K). The former did not saturate with (19) S. J. Tauster, T. A. Pecoraro, and R. R. Chianelli, J . Caral., 63,515 (1980). (20) R. R. Chianelli and S. J. Tauster, J . Caral., 71, 228 (1981).

Beck et al. i,oo--o“

I

cI n

I

I

-

-

” 2 < ) _

1

I H2S

I

l

a

cos

a

:

0.2 0.4 0.6 0.6 1 0 FRACTION OF C O S T A N D A R D C O V E R A G E

0.0

/

( S T A T E I)

b cs2

1.0

+--a+

I

I

I

r n c

CH,SH

I 140

0 0

0 2 0 4 0 6 0 6 1 0 F R A C T I O N OF H2S S T A N D A R D C O V E R A G E

300

500

700

900

0 0

( S T A T E II)

Figure 5. Poisoning of adsorption sites in H2S/C0 sequential dosing experiments at 140 K: (a) HIS TPD area in state I1 as a function of fractional precoverage of CO; standard coverage is equal to area of CO TPD state I peak when dosed alone at 1 X 10” Torr and 300 s at 140 K. (bf CO TPD area in state I as a function of fractional precoverage of H2S; standard coverage is equal to area of H2S state I1 TPD peak when dosed alone a t 1 X 10” Torr and 300 s at 140 K.

increasing dose and has a desorption activation energy of 1 6 kcal/mol. The latter shows first-order desorption kinetics, with an activation energy of 28 kcal/mol. As with O2adsorption, the intensity of the higher temperature TPD peak increased with the extent of preannealing under vacuum at 1000 K. C i 8 0doses on clean rutile resulted in insignificant isotope exchange with surface oxygen. (Very little CI6O was detected by TPD.) For anatase (A2), only the lower temperature CO desorption state was detected. Using procedures described elsewhere,6 we could not detect CO adsorption by IR on either a clean or a sulfided rutile sample, but a weak band at 2190 cm-I was detected on anatase A2. On a CO-dosed rutile sample which had only been outgassed at 300 K under vacuum, a methane TPD peak was observed ( T , = 380 K) with an intensity as high as 5% of CO. Rehydrating a vacuum-annealed rutile surface by exposure to H 2 0 vapor, followed by CO adsorption, led to a similar amount of CH4. Hydrogenation of CO on “wet” Ti02 has been reported el~ewhere.~ H 2 S / C 0 Coadsorption. A series of experiments were carried out in which anatase (A2) was exposed first to H2S at 1 X Torr and 140 K for various times and then to CO at 1 X lo4 Torr at 140 K for 300 s. TPD showed that H2Sin state I1 inhibits the adsorption of CO (Figure 5b). More extensive blocking occurs for larger H2Sexposures. When the dose sequence was reversed, Le., CO followed by H2Sat 140 K, H2Sadsorption in state I1 was not poisoned by C O (Figure Sa). Rutile (R2) gave the same results (not shown). Coadsorption experiments were also carried out at 300 K on rutile (R2). H2S was dosed fLr 300 s at 1 X 10” Torr followed by the same dose of CO. TPD (Figure 6a) shows no change in the peak positions of H2S and C O compared to each adsorbed alone (described above). Reversing the dose sequende gave the same results (Figure 6b).

T (K)

Figure 6. TPD of H2S/C0 sequential dosing at 300 K: (a) H2S dose followed by CO, both at 1 X 10” Torr for 300 s; (b) CO dose followed by H2S, both at 1 X 10” Torr for 300 s. After dosing the sample was cooled to 140 K prior to TPD in both cases.

However, reaction products were observed in these coadsorption experiments (Figure 6 ) . When H2Swas dosed first, a significant H2 peak not attributable to H2S fragmentation in the mass spectrometer was observed between 300 and 400 K. At T N 800 K, SO2 and CH3SH peaked, while COS and CH, continued to increase with temperature. CH3SCH3was detected with an order of magnitude lower intensity than CH3SH in a broad peak centered at 350 K. These products and their trends are similar to those observed in the steady-state reaction of H2S over rutiles5 Reversing the dose sequence (CO followed by H2S) gave CH3SH ( T , 475 K) and CH4 ( T , 350 K) TPD peaks, each having an intensity about 2 orders of magnitude below that of the H,S peak (Figure 6b). At still lower intensity (3 orders of magnitude below the H2S intensity), C2H6 was detected in a broad peak centered at 350 K. In this case, no TPD state a t T , = 850 K was observed for any products, and there was no H2 desorption. The sequential adsorption experiment (H2S followed by CO) was repeated on a sulfided rutile (R2) sample at 300 K (S/Ti AES signal intensity ratio was 0.5). The TPD spectrum was like that of Figure 6a, except that CH,SH, COS, and CH4 components were significantly more intense.

-

-

Discussion 0, Adsorption and Removal. Oxygen is involved in many chemisorption and reaction studies on Ti0,.21-27 Our results, like those for photoinduced oxygen a d ~ o r p t i o n show , ~ ~ that the in(21) S. Sato, T. Kadowaki, and K. Yamaguti, J . Phys. Chem., 88, 2930 (1984). (22) J. Cunningham, B. Doyle, and N. Samman, J . Chem. SOC.,Faraday Trans. I , 72, 1495 (1976). (23) A. R. Gonzalez-Elipe, G. Munuera, and J. Soria, J . Chem. Soc., Faraday Trans. I , 75, 748 (1979). (24) R. I. BicMey and F. S. Stone, J. Cafaf.,31, 389 (1973). (25) A. H. Boonstra and C. A. H. A. Mutsaers, J . Phys. Chem., 79, 1654 (1975). (26) S. Sato and J. M . White, J . Catal., 85, 592 (1981). (27) S . Sato and J. M. White, J . Am. Chem. SOC.,102, 7206 (1980).

Catalytic Reduction of CO with H2S teraction of O2with the titania surface depends on the outgassing temperature. In particular, state I1 desorption increases on rutile after extensive outgassing at T > 800 K. As discussed elsewhere: 4- and 5-coordinate Ti4+ rows on rutile (1 10) surfaces are believed to play a dominant role in the distinctive surface chemistry of rutile compared to anatase. Rutile state I1 O2desorption is assigned to these sites. N o state I1 O2 is found on anatase. State I11 is less dependent on the sample outgassing pretreatments and may represent adsorption on defects or oxygen vacancies of another kind. In an earlier study,28the TPD of adsorbed O2on Degussa P25 (25% rutile) also gave the states I, 11, and 111. State I, which appears on both anatase and rutile samples, is assigned to a weakly bound molecular adsorption on Ti4+ sites. Dissociative adsorption at 140 K is not important since isotope scrambling was insignificant. The experiments employing isotopically labeled O2also show that chemisorbed species in state I1 do not interact with lattice oxygeri. This result is in agreement with Sato and co-workers,21who found no isotope exchange between lattice oxygen and l8Ozin the absence of UV irradiation a t 300 K. Neither is there a chemical connection between the desorption in states I, 11, and 111and that at higher temperatures (>800 K). On partially sulfided rutile, oxygen chemisorption leads to some water, so residual hydrogen (probably as -SH) is available and reactive. This water desorption is reaction-limited and occurs a t the state I1 O2desorption temperature, indicating that oxygen activation may be rate-limiting. Reaction between 02,an anion vacancy, and an -SH group has been suggested for sulfided molybdena supported on A1203.29 The correlation between O2chemisorption in state I1 and the initial rate of H2Sdecomposition in Figure 3 clearly identifies the active site for sulfiding on rutile as the chemisorption site for O2 (state 11) and for CO. This conclusion is consistent with an assignment of these sites to the 5- and 4-coordinate adjacent Ti4+ sites on the rutile (1 10) surface planee6 A similar argument was used by Tauster and ~ o - w o r k e r to s ~suggest ~ ~ ~ ~that edge sites in MoS2 which chemisorb O2 were also active for hydrodesulfurization. Davis and Carver30 found that anion vacancies created on the basal plane of MoS2 and ReSz are active for oxygen chemisorption. A relatively small degree of sulfiding activity may be present on rutile even when no O2chemisorbs in state 11. This was also observed in the case of anatase and is probably due to the presence of defect sites, perhaps a t the intersection between two crystal planes. It is also possible that low sulfiding activity is sustained on other crystal planes. CO Adsorption. The Weakly bound C O state on both anatase and rutile was expected in light of the IR l i t e r a t ~ r e . ~ l Surface -~~ densities of this C O state can become quite large when relatively high dose pressures are used (as much as one C O to every two surface oxygen anion sites). Two configurations have been suggested for weakly bound CO states on silica34 where, at low temperatures (83 K), CO bonds through association with a hydroxyl group (SiOH-.CO) and a second species behaves very much like liquid CO. As our samples were mostly dehydroxylated prior to exposure to CO, it is unlikely that the low-temperature TPD peak is associated with adsorption through TiOH. Moreover, since our TPD experiments were carried out in a vacuum well below the vapor pressure of CO at 120-140 K,3%” the proposed liquidlike C O state is unlikely. The observed weakly bound CO TPD state

The Journal of Physical Chemistry, vel. 90, No. 14, 1986 3135 can be assigned to adsorption on surface oxygen anions through the carbon atom. This assignment is consistent with our finding that H2S, adsorbing through hydrogen bonding to surface oxygen sites, poisons CO adsorption. The strongly chemisorbed CO observed only on rutile ( T , 425 K) has no documented infrared analogue. At saturation, the intensity of the TPD peak represents a surface density of less than 1 CO for every 100 surface oxygen anion sites. This population can be increased by flash-heating the sample under vacuum to lo00 K prior to adsorption. A similar population increase is found for water adsorption in state 1116*38 and for state I1 02. The proposed adsorption sites are the exposed adjacent 5- or 4-fold coordinated Ti4+ cations on the rutile (1 10) surface. We point out in other work that the (110) rutile surface is the only stable low-index surface plane allowing exposed adjacent 4-coordinate Ti4+ sitese6 Adsorbates such as - O H and -SH can be either terminally or bridge-bonded to these sites. It is likely that these sites function like the special chemisorption sites that Vannice and S ~ d h a k a suggested r~~ for interacting with the oxygen end of the CO molecule, thus facilitating dissociation. This is consistent with our observation that high-temperature (673 K) treatment with H2S poisons these sites. &SIC0 Sequential Adsorption. In the low-temperature region (140-300 K) H2S displaces CO and molecularly bound H2S poisons sites which weakly bind CO for both anatase and rutile. Both the site-poisoning experiments and the desorption activation energies (1 5 kcal/mol for H2Sas opposed to 6 kcal/mol for CO) make it clear that molecularly adsorbed H2Sin desorption state I1 is more strongly bound to the surface than C O is desorption state I. By analogy with HpO, we have discussed H2S adsorption in this state in terms of hydrogen bonding to surface oxygen anions.6 Thus, the assignment of weak molecular CO chemisorption to oxygen anion sites is supported by these poisoning experiments. The lack of poisoning in the 300 K dosing experiments is explained by the following argument. The intensities of the H2S and CO TPD peaks ( T , 350 K) are small compared to what is observed at low temperatures after a 140 K dose. Thus, at low pressures and at 300 K, only low surface coverages are achieved, and significant poisoning of weakly bound species is not detectable. The similarity in the product distribution (CH4, CH3SH, COS) with temperature in Figure 6a to that observed in the steady-state reaction of H2Swith CO on rutile5 indicates that the same reaction mechanism holds over a broad range of pressures. Sulfiding enhanced the rate of formation of these products in both the present coadsorption experiments and in the steady-state experi m e n t ~ . This ~ supports the proposal that H2S decomposition (sulfiding) activates the surface. Our results are consistent with the suggestion that CH3SHand perhaps CH3SCH3are intermediates in the production of CH4, and as expected, TPD of adsorbed CH3SHon rutile differs very little from H2S TPD.40 It has been suggested that CH3SH is produced from a thioformic acid intermediate5 on the surface at temperatures above 700 K:

-

-

0

II

H,S/C-H

I1 (28) D. D. Beck, A. 0. Bawagan, and J. M. White, J . Phys. Chem., 88, 2771 (1984). (29) J. Bachelier, M. J. Tilliette, J. C. Duchet, and D. Cornet, J. Carol., 76, 300 (1982). (30) S. M. Davis and J. C. Carver, Appl. Surf. Sci., 20, 193 (1984). (31) K. Tanaka and J. M. White, J . Phys. Chem., 86,4708 (1982). (32) D. J. C. Yates, J . Phys. Chem., 6 5 246 (1961). (33) R. A. Dalla Betta and M. Shelef, J. Catal., 48, 1 1 1 (1977). (34) J. T. Yates, Jr., in Proceedings of the 2nd Symposium of the Indusrry-Uniuersity Cooperatiue Chemistry Program, Texas ACM University Press, College Station, TX, 1984, p 340. (35) Handbook of Chemistry and Physics, 53rd ed.,R. C. Weast, Ed., Chemical Rubber Company Press, Cleveland, OH, 1972, p D151. (36) R. W. McCabe and L. D. Schmidt, Surf.Sci., 60,85 (1976). (37) L. D. Schmidt, Catal. Reo.-Sci. Eng., 9, 115 (1974).

TIOZ

Although the proposed thioformic acid intermediate was not detected by I R in this study, the results of the coadsorption experiments suggest tht strongly bound C O in TPD state I1 is involved in the reaction to form CH3SH and CH3SCH3. We believe that CO and H2S(or -SH) strongly chemisorbed in lattice (38) M. Egashira, S. Kawasumi, S. Kagawa, and T. Seiyama, Bull. Chem. SOC.Jpn., 51, 3144 (1978). (39) M. A. Vannice and C. Sudhakar, J . Phys. Chem., 88, 2429 (1984). (40) D. D. Beck, J. M. White, and C. T. Ratcliffe, submitted for publication to J . Phys. Chem.

3136 The Journal of Physical Chemistry, Vol. 90, No. 14, 1986

anion vacancies on rutile (1 10) interact to form thioformic acid, which leads to methanethiol qnd dimethyl sulfide. It is possible to create lattice anion vacancies with close proximity (- 3 A) to each other on rutile (1 lo), and this would assist in the interaction between CO and H2S molecules chemisorbed in neighboring sites. Decomposition of H2Sin these sites provides a supply of hydrogen for hydrogenation of intermediates. From this discussion, thioformic acid emerges as the preferred route for the activation of CO. It can be hydrogenated to produce adsorbed CH3SHand CH3qFH3. The production of CH4at high temperatures is then readilj attributed to S-H or S-CH3 bond cleavage and hydrogenation. This is demonstrated in an adsorption study of CH3SH on clean and sulfided rutile.40 Disregarding kinetic effects, the expected products in this mechanism are H2, CH4, and CH3CH3. Methane and H2 are favored over ethane owing to the relative abundance of hydrogen atoms provided in the sulfiding step. The involvement of other types of adsorbed H2Sin hydrocarbon formation is unlikely for the following reasons. Rutile is far more active than anatase in hydrocarbon formation from H2S and CO, yet some -SH groups bond to the surfaces of both to isolated 5-coordinate Ti4+ as observed in the IR results. Molecularly adsorbed H2S is probably not involved in hydrocarbon formation, as it is weakly bound to the surface through hydrogen bonds, desorbs at low temperatures, and will be present at very low densities at typical operating temperatures (>573 K). The fact that the activity is increased by sulfiding points to the strongly bound -SH groups as being involved in hydrocarbon synthesis on rutile. It is unlikely that -SH groups themselves are chemisorption sites for CO since bonding to such sites is expected to be extremely weak. Such a proposal has been made for CO interacting with adsorbed -OH.34 Likewise, it is doubtful that other weakly bound forms of adsorbed CQ (observed in TPD) lead to the formation of hydrocarbons. Our proposed model for H2S decomposition on the active chemisorption sites on rutile (1 10) resembles that for sulfided Mo02 catalysts. Several models of the active commercial catalyst have been suggested in the literature, many of them emphasizing surface anion lattice vacancies as the active It has been suggested that edge sites in MoS2, which are analogous to the anion vacancy sites on rutile (1 lo), account for hydrodesulfurization Our active site model is particularly

Beck et al. similar to that of Kilanowski and co-workers,42 who suggested adjacent pairs of anion vacancies as one kind of active site. Activity depends on the enhanced ability of HIS to chemisorb in these sites$"5 releasing its protons either to the surface (creating -OH or hydride-like bonds to Ti3+") or indeed to chemisorbed CO. From AES and XPS evidence -SH further decomposes at 2700 K, leaving sulfides in the anion vacancies, some of which diffuse into the bulk.6 Interaction between chemisorbed CO and surface sulfide leads to COS, consistent with its predominance in the product distribution at higher temperatures. Direct chemisorption of CO on surface S2-is possible, but it is unlikely to lead to COS owing to an expected weak CO chemisorption bond.46

Summary For both O2 and CO TPD, a single TPD state was observed on anatase while two CO and three O2states were observed on rutile. The additional states on rutile are assigned to strong chemisorption in the oxygen vacancies, Le., bridge-bonding to adjacent 4-or 5-coordinate Ti cations on rutile (1 10). A correlation between oxygen chemisorption in these sites and activity in the decomposition of H2S was demonstrated. In sequential HIS and CO dosing experiments, preadsorbed H2S blocks some CO sites while preadsorbed CO is displaced by desorption and transferred to other sites upon exposure to H2S. In both cases, reaction products, mainly CH4 and CH3SH, are observed in TPD. This reduction of CO with H2S is discussed in terms of adsorption of CO and H2S in adjacent anion vacancies of the type described above, producing a thioformic acid intermediate which is then hydrogenated to adsorbed CH3SH,-SCH3, and CH3SCH3. Further hydrogenation leads to CH4. Registry No. Ti02, 13463-67-7; HIS, 7783-06-4; CO, 630-08-0; 02, 7182-44-7.

(41) T. Edwards, A. T. Groszek, and G. C. Stevens, British Patent 1455 193, 1976. (42) D. R. Kilanowski, H. Teeuven, V. H. J. deBeer, B. C. Gates, G. C. A. Schuit, and H. Kwart, J . Catal., 55, 129 (1978). (43) S. N. Yang and C. N. Satterfield, J. Coral., 81, 168 (1983). (44) M. LoJacono and W. K. Hall, J. Colloid Interface Sci., 58,78 (1977). (45) J. Valyon, R. L. Schneider, and W. K. Hall, J. Coral.,85,277 (1984). (46) J. Clarke, Surf. Sci., 102, 331 (1981).