4810
J . Phys. Chem. 1986, 90, 4810-4819
to x-Fe,C (X324) had similar selectivities. This result was not unexpected, because the reduced catalyst is partly converted to Hagg carbide in the early part of the synthesis. Although reduction of the used carbided catalyst (X324C) removed some carbon and wax, the activity and selectivity were not changed. All three of these reduced and carbided samples produced a lower molecular weight product than obtained in tests X149 and X341. Reduced samples were also treated with N H , to product iron nitrides, Fe4N in X220 and Fe2N in X273A. We note that iron nitrides are hydrogenated to N H 3 in pure hydrogen at 230 OC in I or 2 h. In the presence of CO the nitrogen persists for many months, and the nitrogen removed is replaced by carbon to form an t-carbonitride. The Fe,N in X220 gained carbon rapidly early in the FTS to become t-carbonitride. The similarity of activity and selectivity may be expected. After 5 weeks of FTS the sample in X273A was treated with pure hydrogen at 300 "C, and the nitrogen was completely removed, but the sample showed the presence of t and x phases. The presence of nitrogen in the catalyst, probably as an t-nitride or carbonitride, caused a marked shift in selectivity toward lower molecular weight products and increased yields of oxygenates, particularly alcohols. The products reverted to those typical of reduced and/or carburized catalysts when the nitrogen was removed. These data suggest that the iron catalyst can exist in metastable states that persist for 6 or more months. Catalysts for FTS seem to be selectively poisoned by CO, particularly for reactions of hydrogen. For iron this includes the hydrogenation of olefins, hydrogenolysis of hydrocarbons, and hydrogenation of nitrides and carbides. Other reactions including deposition of carbidic
and elemental carbon, the water gas shift, and magnetite formation proceed at measurable rates. Concluding Remarks Thermodynamics is useful in many ways for studying syngas reactions. The energetics of producing a large variety of organic molecules are favorable under conditions of FTS or the alcohol syntheses. Introducing one or more hydroxyl or aldehyde groups and/or double or triple bonds decreases this reaction tendency. Thermodynamically the production of acetylene seems unlikely, and methanol and ethylene glycol require moderate and high operating pressure, respectively. The thermodynamics of producing carbohydrates as postulated by Fischer may be similar to those of ethylene glycol. The production of methane is so favorable that CH4 is virtually the only product possible if equilibria is permitted between all molecules considered. The products of many syngas reactions are often far from the equilibrium composition. Thus, the distribution of products has great diagnostic value in interpreting the mechanisms of the reactions. Thermodynamics also provides useful information on catalytic materials; will they reduce, carbide or nitride, or form sulfides, oxides, and volatile carbonyls? Synthesis tests on an iron catalyst pretreated in several ways suggest that many metastable steady states are possible. In addition, the catalyst is severely inhibited, presumably by carbon monoxide, for reactions with hydrogen. However, reactions forming carbides, carbonitrides, elemental carbon, and iron oxide proceed at a modest rate. Registry No. CO, 630-08-0; C 0 2 , 124-38-9.
CO Hydrogenation over Carbon-Supported Iron-Cobalt and Potassium-Iron-Cobalt Carbonyl Cluster-Derived Catalysts A. A. Chen,? M. Kaminsky,t G . L. Geoffroy,t and M. A. Vannice*+ Department of Chemical Engineering and Department of Chemistry, The Pennsylvania State University, University Park, Pennsylvania 16802 (Received: October 15, 1985: In Final Form: March 24, 1986)
FeCo ana K-promoted Fe-Co CO hydrogenation catalysts were prepared by dispersing iron, cobalt, and iron-cobalt carbonyl clusters on a clean. high-surface-area carbon support. High dispersions were achieved as indicated by chemisorption and XRD measurements. Because of their zero-valent state and the absence of oxygen functional groups on the carbon surface, these catalysts required no high-temperature reduction for activation. The CO,(CO)~/Ccatalyst was particularly active, produced only paraffins, and exhibited very good activity maintenance, whereas the C-supported Fe3(CO),, clusters had the low turnover frequencies and higher olefin/paraffin ratios associated with small Fe crystallites. The mixed-metal cluster catalysts showed intermediate behavior. Addition of one K atom to the clusters markedly decreased specific activity but greatly enhanced selectivity to olefins along with increasing the rate of CO disproportionation. Co, which is known to produce essentially all paraffins, was modified by the K SO that it produced only olefins and CH4. The K significantly increased activation energies for CO hydrogenation, increased the pressure dependency on H2,and made the CO pressure dependency more negative. The variation observed in specific activities indicated that some surface enrichment in Fe may have occurred in the mixed-metal catalysts.
Introduction The high activity and relatively low cost of traditional iron and cobalt Fischer-Tropsch catalysts have given them a preeminent position in catalysis research pertaining to CO hydrogenation. However, one limitation with these catalysts is that they are typically poorly dispersed. The preparation of well-dispersed metal and mixed-metal particles is a desirable goal since they may have improved activity, selectivity, and on-stream durability. Two approaches to this end are the employment of metal carbonyl clusters (MCCs) as catalyst precursors's2 and the use of very 'Department of Chemical Engineering. *Department of Chemistry.
0022-3654/86/2090-4810$01.50/0
high-surface-area supports, preferably with no reactive surface oxygen, that can entrap and stabilize small metal clusters or particles. In previous work we have successfully used an amorphous carbon (1400 m2/g) to prepare Fe and Fe-Ru catal y s t ~ ,and ~ , ~in this study we report the synthesis and characterization of a series of MCC-derived, carbon-supported Fe-Co catalysts along with a series of MCC-derived, K-promoted Fe-Co (1) Anderson, J. R.; Mainwaring, D. E. J. Catal. 1974, 35, 162. (2) Zwart, J.; Snel, R. J. Mol. Catal. 1985, 30, 305. (3) Jung, H. J.; Vannice, M. A.; Mulay, L. N.; Stanfield, R. M.; Delgass, W. N. J . Catal. 1982, 76, 208. (4) Kaminsky, M.; Yoon, K. J.; Geoffroy, G. L.; Vannice, M. A. J. Catal. 1985, 91, 338.
0 1986 American Chemical Society
The Journal of Physical Chemistry, Vol. 90, No. 20, 1986 4811
CO Hydrogenation over Carbon-Supported Catalysts
TABLE I: Metal Loadings of Carbon-Supported Carbonyl Cluster Catalysts Unpromoted Catalysts" catalvst
wt % Fe
wt % c o
pmol Fe/ ( a cat.)
pmol Co/ ( a cat.)
co2 Fe, Co,(high) Fe, Co2(low) Fe3Co FeCo, Fe3
0 0.6 0.4 0.4 0.7 4.4
4.1 2.6 1.6 1.3 1.3 0
0 108.9 73.2 73.2 119.6 786.0
700.8 440.7 274.5 215.5 211.9 0
+ +
mol fract Fe 0
FeICo
0 0.25 0.27 0.34 0.56
0.20 0.21 0.25 0.36 1 .o
m
Promoted Catalvsts catalyst KCo KFeCo, KFeCo KFe3Co KFe, KFeCo-(NO,)
wt % Fe
0 0.9 1.1 2.1 4.1 1.5
wt % Co
wt % K
pmol Fe/ (g cat.)
pmol Co/ (g cat.)
pmol K/ (g cat.)
mol ratio Fe/Cs
mol ratio K/Co
4.4 3.1 2.7 1.o
3 .O 0.6 1.6 0.8 1.1 1.2
0 160.7 194.6 372.4 137.7 266.1
745.7 51 8.6 459.3 176.5 0 255.9
767.2 153.4 409.2 202.0 282.0 306.9
0 0.3 0.4 2.1
1 .o 0.3 0.9 1.1
m
m
1 .o
1.2
0 1.5
K was not analyzed for in these samples.
catalysts, their chemisorption behavior toward C O and H2, and their catalytic activity and selectivity in the C O hydrogenation reaction. Recent studies of more conventional F d o catalysts have dealt with unsupported alloys5-' and supported bimetallic catalysts prepared by impregnation of dissolved metal In one case, carbon-supported bimetallic crystallites were created by CO disproportionation over unsupported alloy particles.I8 Although the catalytic behavior of supported homonuclear Fe and Co clusters has been examined in earlier ~ t u d i e s , ~only ~ ~three * ' ~ studies ~ ~ have
'
(5) Nakamura, M.; Wood, B. J.; Hou, P. Y.; Wise, H.Proc. 7th Int. Congr. Catal. 1981, 432.
(6) Moran-Lopez, J. L.; Wise, H.Appl. SurJ Sci. 1980, 4 , 93. (7) Artyukh, Y. N.; Rusov, M. T.; Boldyreva, N. A. Kinet. Kata. 1967, 8, 1319. (8) Tricker, M. J.; Vaishnava, P. P.; Whan, D. A. Appl. Catal. 1982, 3, 283. (9) Brown, R.; Cooper, M. E.; Whan, D. A. Appl. Catal. 1982, 3, 177. (IO) Smith, P. J.; Taylor, D.W.; Dowden, D.A.; Kemball, C.; Taylor, D. Appl. Carol. 1982, 3, 303. (1 1) Taylor, D. W.; Smith, P. J.; Dowden, D. A.; Kemball, C.; Whan, D. A. Appl. Catal. 1982, 3, 161. (12) Butt, J. B.; Schwartz, L. H.; Baerns, M.; Malessa, R. Ind. Eng. Chem. Prod. Res. Deu. 1984, 23, 51. (13) Arai, H.;Mitsuishi, K.; Seiyama, T. Chem. Lett. 1984, 1291. (14) Amelse, J. A.; Schwartz, L. H.; Butt, J. B. J . Catal. 1981, 72, 95. (15) Stanfield, R. M.; Delgass, W. N. J . Catal. 1981, 72, 37. (16) Arcuri, K. B.; Schwartz, L. H.; Piotrowski, R. D.; Butt, J. B. J . Catal. 1984, 85, 349. (17) Christensen, P. H.;Morup, S.; Clausen, B. S.; Topsoe, H.Proc. Int. Congr. Catal., 8th 1984, 2, 545. (18) Audier, M.; Bass, B.; Coulon, M. C l Mol. Chem. 1984, 1 , 33. (19) Iwasawa, Y.; Yamada, M.; Sato, Y.; Kuroda, H.J . Mol. Catal. 1984, 23, 95. (20) Iwasawa, Y.; Yamada, M.; Ogasawara, S.; Sato, Y.; Kuroda, H. Chem. Lett. 1983, 621. (21) Nakamura, R.; Oomura, A,; Okada, N.; Echigoya, E. Chem. Lett. 1982, 1463. (22) Phillips, J.; Clausen, B. S.; Dumesic, J. A. J . Phys. Chem. 1980.84, 1814. (23) Lisitsyn, A. S.;Golovin, A. V.; Kuznetsov, V. L.; Yermakov, Y. I. C1 Mol. Chem. 1984, 1 , 115. (24) Lisitsyn, A. S.; Golovin, A. V.; Kuznetsov, V. L.; Yermakov, Y. I. React. Kinet. Carol. Lett. 1982, 19, 187. (25) Lazar, K.; Matusek, K.; Mink, J.; Dobos, S.; Guczi, L.; Vizi-Orosz, A,; Marko, L.; Reiff, W. M. J . Catal. 1984, 87, 163. (26) Schay, Z.; Lazar, K.; Mink, J.; Guczi, L. J . Caral. 1984, 87, 179. (27) Hugues, F.; Bussiere, P.; Basset, J. M.; Bonneviot, L.; Olivier, D. Proc. Int. Congr. Catal. 7th 1981, 418. (28) Commereuc, D.; Chauvin, Y.; Hugues, F.; Basset, J. M.; Olivier, D. J . Chem. SOC.,Chem. Commun. 1980, 154.
involved stoichiometric Fe-Co metal cluster^.^^-^^ N o investigations have been found on K-promoted Co or Fe-Co clusters although K-Fe clusters on Al,03 and SiOzhave been previously ~tudied.~'
Experimental Section Catalyst Preparation. The amorphous carbon black (CSX-203 from Cabot Corp.) used in this study was desulfurized as previously described4 and had a final N 2 BET surface area of 1400 m2/g. The procedure also removed oxygen functional groups from the surface. Prior to supporting the clusters, the carbon was degassed for 8 h at 573 K. The carbonyl clusters-Fe,(CO)l,,38 HFe) ~ ]K[ FeCo,(CO) 12,39 Et,N [Fe3Co(CO)13],40 K [ F ~ C O ( C O ,41 C O , ( C O ) , ~ ]K , ~[~F ~ , C O ( C O ) ~K~[ ]C,O ~ (~C O ) , ] , ~and ~ K[HFe3(CO), 1]42-were synthesized according to literature procedures. The carbon support was impregnated to the point of incipient wetness under anaerobic conditions using standard Schlenk techniques with dry, degassed T H F or acetone as the solvents.43 Coimpregnated samples were prepared by sequential impregnation of CO,(CO)~and Fe3(C0)12. The metal loadings of all catalysts are shown in Table I. In this paper clusters will be denoted by their stoichiometric metal ratios, and, for brevity, carbonyl ligands will be omitted. The 3.2 and 2.0 wt % coimpregnated catalysts will be denoted Fe, Co2 (high) and Fe3 + Co2 (low), respectively. A KFeCo control sample was prepared from a solution of the mixed nitrates impregnated on CSX-203, and this sample will be referred to as KFeCe(N0,). A control sample of K-dopal carbon containing no Fe or Co was prepared by impregnating CSX-203 with a K N 0 3 solution. Catalyst Pretreatment. The fresh catalysts were loaded into
+
(29) Ferkul, H. E.; Berlie, J. M.; Stanton, D. J.; McCowan, J. D.; Baird, M. C. Can. J . Chem. 1983, 61, 1306. (30) Blanchard, M.; Bonnet, R. Bull. SOC.Chim. Fr. 1977, 7. (31) Meyers, G. F.; Hall, M. B. Inorg. Chem. 1984, 23, 124. (32) Schneider, R. L.; Howe, R. F.; Watters, K. L. Inorg. Chem. 1984, 23, 4593. (33) Ballivet-Tkatchenko, D.; Tkatchenko, I. J . Mol. Chem. 1981, 13, 1. (34) Roper, M.; Hemmerich, R.; Keim, W. Chem. Ing. Techn. 1984, 56, 152. (35) Hemmerich, R.; Keim, W.; Roper, M. J . Chem. SOC.,Chem. Commun. 1983, 428. (36) Lisitsyn, A. S.;Kuznetsov, V. L.; Yermakov, Y. I. Kinet. Katal. 1983, 24, 764. (37) McVicker, G. B.; Vannice, M. A. J . Catal. 1980, 63, 25. (38) McFarlane, W.; Wilkinson, G. Inorg. Synth. 1969, 8, 181. (39) Chini, P.; Colli, L.; Peraldo, M. Gazz. Chim. Ital. 1960, 1005. (40) Steinhardt, P. C.; Gladfelter, W. L.; Harley, D.; Fox, J.; Geoffroy, G. L. Inorg. Chem. 1980, 19, 332. (41) Ruff, J. K. Inorg. Chem. 1968, 7 , 1818. (42) Hodali, H. A.; Arcus, C.; Shriver, D. F. Inorg. Synth. 1980, 20, 218. (43) Shriver, D. F. The Manipularion of Air Sensitive Compounds; McGraw-Hill: New York, 1969.
4812
The Journal of Physical Chemistry, Vol. 90, No. 20, 1986
Chen et al.
TABLE 11: Chemisorption on Carbon-Supported Fe-Co Clusters after HTR fresh sample used catalyst"
catalyst
CO,
+
Fe, Co2(high) Fe, + Co,(low) Fe,Co FeCo, Fe, KCo KFeCo, KFeCo KFe,Co KFe, KFeCo-(NO,)
co
H2
@mol/ (g cat.)
Nmol H/ (g cat.)
297 325 13 130 53 540 156 61 164 318 297 135
74 89 18 28 86 105
106 58 139 60 66 57
co
H2
@mol/
Kmol H / (g cat.)
(g cat.) 86 84 IO 22 50 333 101 55 94 30 30 73
38 107 16
24 41
I I8 74 39 70 49 51 71
COad/Hadb fresh used 4.0 3.7 0.7 4.6 0.6 5.1 1.5 1.1
1.2 5.3 4.5 2.4
2.3 0.8 0.6 0.9 1.2 2.8 1.4 1.4 1.3 0.6 0.6 1.0
fresh sample H/MT CO/M,
HIM,
0.1 1 0.16 0.05 0.10 0.26 0.13
0.05 0.19 0.05 0.08 0.12 0.15
0.42 0.59 0.04 0.45 0.16 0.69 0.23 0.09 0.25 0.58 0.40 0.26
used catalyst CO/M, 0.12 0.15 0.03 0.08 0.15 0.42 0.14 0.08 0.14 0.05 0.04 0.10
dco,b nm fresh
used
1.8 1.3 18.7 1.7 4.7 1.1 3.3 8.3 3.0 1.3 1.9 2.9
6.1 5.0 28.0 9.7 5.0 1.8 5.2 9.7 5.3 15.0 20.6 1.2
"Used catalyst chemisorptions are corrected for weight change due to solvent loss. A fresh sample basis is used in this table. b C O uptake at 300 K is used for particle size and COad/Hadcalculations. M, is the total metal content in pnol/(g cat.).
chemisorption and reactor cells in an N2-filled glovebox, and all subsequent manipulations were also conducted under anaerobic conditions. The low-temperature reduction (LTR) pretreatment consisted of heating fresh samples in 40 cm3/min He flow to 473 K, substituting a 40 cm3/min flow of H2 for He, and holding at this condition for 2-5 h until no solvent peak could be detected by gas chromatography. A more severe high-temperature reduction (HTR) consisted of heating the sample to 673 K under a 40 cm3/min flow of hydrogen and holding at that temperature for 16 h. Chemisorption Measurements. Chemisorption measurements were performed in a stainless steel adsorption system equipped with a Balzers TSU-171 turbomolecular pump and high-vacuum stainless steel bellows valves (MDC Mfg.). The system provided 5 X IO4 Torr ultimate vacuum while typical vacuums in the region of the sample holder approached 1 X lo-' torr, as measured by a Granville-Phillips Model 260 Bayard-Alpert ionization gauge. Catalyst samples (-0.15 g) were loaded into Pyrex cells equipped with greaseless high-vacuum stopcocks and attached to the system via a high-vacuum O-ring joint. An MKS Baratron Model 310 high-precision capacitance manometer was used to monitor pressures during chemisorption measurements. CO chemisorption isotherms were obtained at both 195 and 300 K according to a procedure described e l ~ e w h e r e . ~ , ~ Emmett and Harkness had first shown that H, adsorption on bulk iron surfaces was activated,44and recent studies have substantiated this on small, supported Fe crystallite^^^^^^^^ and Co4' crystallites. Consequently, H, chemisorption was measured with a desorption method similar to that described by Amelse et al.,I4 which consisted of cooling from the pretreatment temperature to 273 K under flowing HI, evacuating to a pressure below 2 X IO4 Torr to remove weakly adsorbed hydrogen, isolating the sample, and heating quickly to the original pretreatment temperature. The amount of chemisorbed hydrogen was calculated from the final pressure obtained. Kinetic Measurements. The kinetic data were typically obtained under differential conditions at 1 atm using a glass, plug-flow microreactor described el~ewhere.~ Catalyst charges to the reactor were approximately 0.4-0.5 g and C O conversions were usually maintained below 10%; however, for the most active catalysts, conversions occasionally were allowed to reach 20% or more. For these particular runs, no significant deviations were noted on the extension of the Arrhenius plot from the lower conversion points. The catalytic data in all cases except Figures 4-6 were obtained after 20 min on stream with a bracketing technique involving a 20-min exposure to pure H2 between each set.48 This procedure (44) Emmett, P. H.; Harkness, R. W. J . Am. Chem. SOC.1935, 57, 1631. (45) Topsoe, H.; Topsoe, N.; Bohlbro, H. Proc. Congr. Catul., 7th 1981, 247. (46) Weatherbee, G. D.; Rankin, J. L.; Bartholomew, C. H. Appl. Cutul 1984, 11, 73 (47) Zowtiak, J. M.; Bartholomew, C. H. J Cutul. 1983, 83, 107
was successful in preventing deactivation in the nonpromoted catalysts, although the promoted catalysts sometimes showed strong deactivation over long periods of time. Partial pressure analyses were performed at 1 atm total pressure after the HTR step, with He as a diluent in the feed stream. All gases used were 99.999% purity and additional purification traps were used.3 Metal loadings were analyzed with both neutron activation analysis and atomic absorption spectroscopy. X-ray line broadening measurements were performed with a Rigaku Model 401 1B3 diffractometer with Cu Ka radiation. Particle sizes were calculated with the Scherrer equation with corrections for instrumental broadening.
Results Chemisorption Measurements. The results of C O adsorption at 300 K and H, desorption measurements after a HTR pretreatment are presented in Table I1 for both the fresh catalysts and those used in the kinetic studies. Neither pure nor K-doped CSX-203 exhibited any irreversible CO uptake. The data in Table I1 have been corrected for a small uptake of hydrogen (3.8 kmol H/(g cat)) found on both the blank and the K-doped support. Estimated crystallite sizes based on CO uptakes at 300 K were calculated by assuming spherical geometry, an adsorption stoichiometry of CO/M, = 1.0, and a cross-sectional area for both surface cobalt and iron atoms of 9.4 A*, a value obtained by assuming equal fractions of the three low-index crystal planes of Fe.3-4 The final relationship was d (nm) = 0.75/0, with the dispersion, D,defined as M,/M,, where M , represents surface atoms and M , is the total number of Fe Co atoms.49 Actiuity Measurements. Most of the Fe-Co catalysts in this study displayed substantial activity for C O hydrogenation at temperatures between 473 and 573 K, as shown in Figures 1 and 2. At any given temperature, the cobalt-only sample was always at least 100 times as active as its K-promoted analogue, which was the least active catalyst. In general, the family of unpromoted catalysts was much more active than the K-promoted clusters. To avoid extrapolation errors within a family, the activities of the unpromoted catalysts are compared at 498 K (Table 111), while those of the promoted catalysts are reported at 548 K (Table IV). All of the catalysts were quite active after the LTR step, indicating that the described anaerobic procedure for handling the samples was successful. In some cases small amounts of methane were observed after LTR prior to the introduction of CO, consistent with observations by Hucul and Brenner of methane formation below 473 K during the decomposition of various MCCs
+
(48) Vannice, M . A. J . Catal. 1975, 37, 449. (49) Emmett, P. H.; Takezawa, J. J . Res. Inst. Catal., Hokkaido Uniu. 1978, 26, 3 1 . (50) Hucul, D. A,; Brenner, A. J . Am. Chem. SOC.1981, 103, 217. (51) Reuel, R. C . ; Bartholomew, C. H. J . Cutal. 1984, 85, 63. (52) Anderson, R. B. The Fischer-Tropsch Synthesis; Academic: Orlando. FL, 1984.
The Journal of Physical Chemistry, Vol. 90, No. 20, 1986 4813
C O Hydrogenation over Carbon-Supported Catalysts
TABLE 111: Activities of Unpromoted Carbon-Supported Clusters at 498 K and 101 kPa with HI/CO = 3
%
co
% co
catalyst
conv. total
Fe3 + Co,(high) Fe, + Coz(low) Fe3Co FeCo, Fe3
14.5 4.4 1.6 1.6 3.0 2.0
13.2 4.1 1.4 1.5 2.7 1.2
c02
specific activity, activity,' pmol CO/(s pmol Mt) x io4 wmol CO/ (s g cat.) CO" CH4 co2 Pretreatment: LTR
conv"
1.25 0.41 0.12 0.16 0.21 0.1 1
17.83 7.47 3.46 5.40 6.50 1.48
turnover freq, s-' x io3 CO"
CH,
14.4 1 4.89 3.16 3.46 4.58 0.3 1
1.87 0.88 0.3 1 0.55 1.01 0.92
4.216 1 .26b 9.18* 1.216 4.066 0.216
3.406 0.83b 8.4Cjb 0.7g6 2.86b 0.04*
19.47 4.74 2.10 1.75 3.41 0.43
2.07 0.80 0.16 0.41 0.70 1.10
18.80' 5.18' 9.16' 3.77C 4.14C 0.52'
1 5.84c 3.10C 7.35' 2.33' 2.27c 0.10c
Pretreatment: HTR 19.2 co2 Fe3 + Co,(high) 4.8 1.2 Fe, + Co2(low) 1 .o Fe3Co 2.2 FeCo, 3.0 Fe3 "CO converted to hydrocarbons (HC) only. the used sample. 10.0
5 E
I
I
I
I
18.0 4.4 1.1 0.9 2.0 2.0
1.62 0.44 0.09 0.09 0.20 0.17
23.1 1 7.91 2.61 2.82 6.18 2.20
bTOF based on CO uptake at 300 K on the fresh sample. 'TOF based on CO uptake at 300 K on I
I
10.0
I
I
I
I
1.00-
:
-
\
80
E
=
0.10-
L
>-Iu a
0.01 -
i 0.001 I .6
\ 1.8
I IT
2.0
2.2
( ~ " xIO')
0001I 16
1
1 18 I/T
1 ( K-'X
I 20
L A 22
lo3)
Figure 1. Arrhenius plots for CO conversion to hydrocarbons over Fe-Co and K-FeCo catalysts after LTR, H2/C0 = 3.0 P = 1.O atm: (A)Co2; ( 0 )Fe, + Co2(high):(0)FeCo3; (0)Fe3Co;(0) Fe,; (@)KFeCo,;( 0 ) KFe,Co; (A) KCo.
Figure 2. Arrhenius plots for CO conversion to hydrocarbons over Fe-Co and K-Fe-Co catalysts after HTR, H2/C0 = 3.0, P = 1.O atm: (A)Co,; ( 0 )Fe3 + Co2(high);(0)FeCo,; (0)Fe,Co; (0)Fe,; (@)KFeCo,;( 0 ) KFe,Co; (A) KCo; (M) KFe,.
under H2;50however, the decomposition of T H F used in the preparation of our catalysts may have also contributed to the production of the observed methane. The activities of the Fe3 and Co, catalysts increased after the HTR step. In contrast, the activities of the unpromoted bimetallic catalysts either remained about constant or actually decreased slightly after HTR, indicating that activation of the bimetallic catalysts was apparently achieved after LTR. This observation is consistent with earlier reports stating that the presence of Co facilitates the reduction of Fe in bimetallic Fe-Co ~ y s t e m s ? , ' ~ * ' ~ although only residual amounts of oxides would be expected in these anaerobically handled catalysts. The unpromoted catalysts follow a general trend of decreasing activity with increasing Fe content if the activities are based on the turnover frequencies (TOFs) after HTR, as illustrated in Figure 3. These TOFs were calculated with the CO adsorption values at 300 K on the used
samples to count surface sites. Although the exact stoichiometry has not been verified, the assumption of C O / M , near unity on the used catalysts, which have larger metal particles, should allow an estimation of surface atoms within a factor of 2 and reflect relative changes within a catalyst s e r i e ~ . ~ The . ~ ' dashed lines in Figure 3 indicate the predicted TOFs of a metal particle whose surface and bulk compositions are equal to each other. No particle size effects were taken into account in this simplified linear approximation. Because of the similarity in uptakes after HTR, similar values are obtained with the H2 chemisorption measurements, but T O F values based on chemisorbed CO are more conservative. This trend does not exist after LTR, where the TOFs are calculated on the basis of chemisorption on the fresh samples. The catalyst activity showed no maximum-only a monotonic decline with increasing Fe, but the possibility of an optimal Fe-Co composition is not ruled out because most of the unpromoted bimetallic catalysts were cobalt-rich. (The Fe3Co catalyst contained inseparable cobaltate impurities and was therefore more cobalt-rich than its stoichiometric formula indicates.)
(53) Storch, H. H.; Golumbic, N.; Anderson, R.B. The Fischet-Tropsch and Related Syntheses; Wiley: New York, 1951.
The Journal of Physical Chemistry, Vol. 90, No. 20, 1986
4814
Chen et al.
TABLE I V Activities of Promoted Carbon-Supported Clusters at 548 K and 101 kPa with HJCO = 3
catalyst
% co conv total
%CO conv"
KCo KFeCo, KFeCo KFe3Co KFeld
9.7 4.7 12.9 3.4 7.4
3.5 2.0 6.7
1.9
0.21 0.55 0.22 0.15 0.17
KCo KFeCo, KFeCo KFe C KFetd KFeCo-(NO3)
5.5 4.0 10.3 2.8 8.1 4.5
2.0 1.6 5.3 0.9 2.8 1.4
0.07 0.50 0.24 0.10 0.26 0.16
1
.o
specific activity, pmol CO/ (s pmol Mt) x IO4 CH,
CO,
CO t o H C
coz/
turnover, freq, s-' x IO3 COG CH,
Pretreatment: LTR 2.83 0.82 8.11 3.66 3.38 1.01 2.70 0.32 2.34 0.44
4.70 6.25 5.23 5.57 6.83
1.66 0.77 1.52 2.4 2.9
1.4* 9.0* 1.4' 0.5* 0.2*
0.4' 4.1' 0.4b 0.1' 0.04*
Pretreatment: HTR 1.oo 0.19 7.32 2.21 3.59 0.74 1.89 0.37 3.49 0.72 3.06 0.47
1.69 6.85 4.54 3.82 6.66 6.62
1.7 0.93 1.26 2.20 1.92 2.15
0.7' 9.Ic 2.5' 3.Y 8.6c 2.2'
0. I C 2.7c 0.5' 0.7' 1.8' 0.3'
activity," pmolCO/ (s g cat.)
COG
" C O converted to hydrocarbons (HC) only. * T O F based on C O uptake at 300 K on the fresh sample. 'TOF based on CO uptake at 300 K on the used sample. dData reported at 300 'C. TABLE V: Selectivity of Unpromoted Catalysts after LTR and
HTR Pretreatment
catalyst
temp, K
% co conv to H C
CHI
C2=
co2 Fe, + Co2(high) Fe, Co2(low) Fe,Co FeCo, Fe3
485 498 522 523 523 540
6.5 6.6 5.0 5.4 7.2 5.2
91.6 86.1 92.1 73.5 84.4 44.2
0 0 0 1.5 0.5 10.5
mol C2 C3= LTR Pretreatment 4.2 0 0 6.3 5.0 0 3.8 11.0 7.0 1.7 10.9 12.7
cot Fe, Co,(high) Fe3 Co2(low) Fe3Co FeCo, Fe,
487 513 523 523 523 540
8.5 7.0 3.8 4.1 6.3 6.2
93.3 81.0 82.1 67.2 75.3 42.2
0 1.0 3.0 2.3 1.6 11.8
HTR Pretreatment 3.3 0 1.6 8.7 7.9 3.7 11.9 5.0 4.4 7.9 9.0 13.6
+
+ +
"(C,'
+ C3=)/(C2+ C3). 'nd
C3
C4=
C4
Cs
C6
C7
olefin/ paraffin ration
2.2 3.7 2.2 5.7 3.3 5.3
0 0 0 0 0.3 2.3
1.3 2.2 0.5 2.8 17 6.3
0.6 1.3 0.2 1.2 0.8 4.6
0.1 0.3 nd 0.6 0.3 2.1
nd' nd nd nd nd 1.2
0 0 0 0.3 0.2 1.4
2.2 3.6 1.9 5.7 4.4 4.6
0 0 0 0 0.6 2.8
0.9 2.4 1.3 4.7 3.6 6.6
0.3 1.1 0.1 2.3 1.5 4.6
nd 0.6 nd 1.1 0.6 3.6
nd nd nd nd nd 1.2
0 0.2 0.7 0.4 0.5 1.9
% HC product
'
indicates not detected.
As with the unpromoted catalysts, the promoted catalysts became active after only a LTR step, as shown in Table IV. A rough estimate of the rate of carbon deposition on these catalysts was made on the basis of the relative amounts of COz formed and of C O reacted to hydrocarbons (HC). A COZ/HCratio exceeding unity indicates the presence of the Boudouard reaction, assuming that the only sources of COz are carbon deposition and the secondary water-gas shift reaction. With regard to the promoted catalysts, carbon deposition was noted for almost all samples, and accordingly some deactivation was observed. For the KFeCo and KFeCo, catalysts, most of the catalytic activity could be restored after a H T R treatment, indicating that for these samples deactivation was caused by carbon deposition and not metal sintering. Although the activity of the KFe3 catalyst increased slightly, the KFe,Co and KCo catalysts displayed sharply decreased activities after the H T R step, indicating for the latter two catalysts the possibility of metal sintering. This was confirmed for the KFe3Co sample, which exhibited detectable XRD peaks indicating a particle size of approximately 15 nm, consistent with the crystallite size of 15 nm indicated by CO chemisorption. XRD analyses on the remaining promoted and unpromoted samples showed no peaks, indicating that the metal particles were probably smaller than 5 nm. While the XRD results do not agree well with many of the chemisorption measurements, the uncertainties in adsorption stoichiometry and the influence of K on adsorption could account for most of the discrepancies. Selectivity, Activation Energies, and Partial Pressure Dependencies. The hydrocarbon selectivities of the unpromoted catalysts at similar conversions are reported in Table V. No olefin formation was detected with the Co, catalyst, whereas the highest olefin-paraffin (o/p) ratio was found for the Fe3 catalyst. This
20,
I
I
I
I
1
LL
0 L
Fe M O L E FRACTION
Figure 3. Activity vs. metal particle composition for the unpromoted catalysts after HTR, T = 498 K,H,/CO = 3.0, P = 1.0 atm: (0)C O CH, formation. Dotted lines represent converted to hydrocarbons; (0) predicted values assuming a homogeneous particle composition.
is consistent with traditional observations regarding these bulk catalyst^,^^^^^ although some olefin formation over Co has been observed at low conversion^.^^*^^ For all of the unpromoted (54) Reuel, R. C.; Bartholomew, C. H. J . Coral. 1984, 85, 78. (55) Bartholomew, C. H.; Reuel, R. C. Ind. Eng. Chem. Prod. Res. Deu. 1985, 24, 56.
The Journal of Physical Chemistry, Vol. 90, No. 20, 1986 4815
C O Hydrogenation over Carbon-Supported Catalysts TABLE VI: Selectivities of Promoted Catalvsts after LTR and
catalvst ~, ~~
temp, K
% conv
HTR Pretreatment
mol % HC product C,= C,’ C, C,= C, LTR Pretreatment
CO to HC
CH,
28.0 27.5 20.9 19.9 19.6 27.0 24.7 21.9
KCo KFeCo KFeCo, KFe,Co KFe,
548 548 524 562 553
3.5 3.5 1.3 2.2
54.6 49.1 74.3 32.1 37.3
KCo KFeCo KFeCo, KFe,Co KFe, KFeCo-(NO3)
551 548 538 564 558 557
3.8 2.9 1.9 2.5 2.5 3.3
44.5 42.8 59.5 35.9 39.7 28.5
1.8
0 0 0 2.2 0
16.8 20.9 11.6
C,
C,
C,
C,
0 0 2.0 0 0
5.3 4.3 1.0 4.8 5.9
ndb
nd nd
0 0
8.3 9.2 3.4 11.1 9.2
0 0 0 0 0 0
9.3 12.3 5.5 9.6 7.1 4.8
0 0 0 0 0 7.0
4.1 3.9 0 3.1 4.9 6.9
0 0 0
15.0 14.4 6.7 18.6 20.1
olefin/ paraffin ratio‘
0.2
high high high
nd
nd nd
high
nd
nd
21.0
2.3 1.8 3.3
2.2
high high
2.1 0.6 3.1
21.2
HTR Pretreatment
“(C,’
+ C3=)/(C2+ C,).
1.9 0 0’ 3.4 3.1 4.0
19.1 16.6 13.6 17.6 20.5 23.2
nd nd nd nd
nd 3.7
13.1 14.6 11.3
bnd indicates not detected.
TABLE VII: Activation Energies and Reaction Orders for CO Hydrogenation over Fe-Co/C Catalysts
catalyst co,
+ Fe,(high) Fe3 + Co,(low)
Co;
Fe3Co FeCo, Fe3 Fe3 (from ref 4) KCo KFeCo, KFeCo KFe3Co KFe, KFeCo-(NO,)
LTR/HC formation 121 95 109 121 112 71 84 295 143 218 68
act. energy after pretreat.” E , kJ/mol HTR/HC LTR/CH4 formation HTR/CH4 123 89 104 97 100 63 63 232 120 180 81
112 93 104 145 102 69 100 187 146 167 121 89 189
partial pressure dependencies after HTR temp, K xb Yb X’c Y‘c
113 90 87 101 98 64 86 190 143 165 99 83 150
488 508 523 513 523 533 548 560 548 556 563 573 558
“From 473 to 523 K for unpromoted catalysts and 523 to 573 K for promoted catalysts. ~XP(ECH,/RT)PH/PCO” catalysts, the product distributions were consistent with the Anderson-Schulz-Flory chain-growth model, although the usual deviations were seen in the C, and C, hydrocarbon fraction^.^^ The bimetallic catalysts yielded no synergistic increase of C2 or C3 olefins-Co always yielded the smallest amounts of these products and Fe the largest. The incorporation of K into the cluster precursors resulted in a marked increase in olefin production and reduced methane formation, as illustrated in Table VI. At the low conversions listed, C2 and C3 paraffins were almost never observed, although these products were detectable in small amounts at conversions above 6-7%. The apparent activation energies for methanation and C O reacted to hydrocarbons, derived from data such as that in Figures 1 and 2, are listed in Table VII. The bimetallic catalysts exhibited activation energies between those for the single-metal catalysts. The promoted catalysts exhibited sharply increased activation energies compared to the unpromoted catalysts, with the exception of the Fe-rich KFe3Co and KFe3 samples. Although the KFe3Co sample had a low activation energy after LTR, this value increased after HTR and became more similar to the other catalysts. The KFe, sample deactivated strongly in the initial runs following the LTR, and its activation energy was not determined; however, after the HTR step, the catalyst stabilized sufficiently to allow an activation energy to be obtained. The reaction orders for both methanation and total hydrocarbon formation are also shown in Table VII. The addition of potassium to a given cluster usually made the CO partial pressure dependence more negative whereas the H 2 pressure dependence became more positive. The activity maintenance behavior of selected cluster catalysts is illustrated in Figure 4. The unusual shape for the KCo catalyst (56)
Ponec, V. Coal Sci. 1984, 3, 35.
0.6 1.1 1.1 1.O 0.6 1.1 1.1 1.1 1.1 1.o 1.8 1.4
-0.5 -0.5 -0.4 -0.1 -0.3 0.4 -0.5 -1.4 -0.9 -0.7 0.2 0.2 -0.4
1.5
0.7 1.1 1.3 1.0 0.8 1.0 1.0 1.2 1.3 1.8 1.8
2.0 1.9
b r H =~ A exp(EHC/RT)PH,”Pco”. ‘rCH4 = A’
1 I
0
-0.6 -0.5 -0.6 -0.4 -0.7 -0.2 -0.7 -1.8 -1.6 -1.0 -0.6 -0.3 -0.6
IO
20
30
40
I
1
50
60
-
T I M E O N STREAM ( h )
Figure 4. Activity maintenance behavior of Fe-Co and K-Co catalysts, normalized activity for CH, formation vs. time on stream, H,/Co = 3.0, P = 1.0 atm: (0)Co2, T = 498 K; (a)Fe3Co, T = 513 K; (A)KCo, T = 553 K; ( 0 ) 10% Fe/AI,O, from Fe(NO,), (ref 77).
is discussed in the next section, but it is worth noting that the high olefin selectivity initially observed with this sample (Table VI) was maintained throughout the run. Almost all of the deactivation of the unpromoted catalysts took place within the first few hours on stream. As shown, the cluster-derived catalysts perform fa-
4816
The Journal of Physical Chemistry, Vol. 90, No. 20, 1986
vorably compared to a conventionally prepared Fe catalyst
Discussion
Chemisorption. The stoichiometries of H2 and CO adsorption on supported Fe and Co are not as well-defined as on other Group VI11 metals, and no universally acceptable standard has been established for the measurement of Fe or Co surface areas by c h e m i ~ o r p t i o n . ~Ever ~ since the early work of Emmett and Harkness, hydrogen adsorption on iron catalysts has been known to be an activated process.44 Additional work by Brunauer and Emmett demonstrated that the adsorption of hydrogen was dissociative on iron and also that the adsorption was confined to the iron itself in promoted catalysts.58 More recent hydrogen adsorption measurements near 300 K on iron have found very low uptakes, consistent with the existence of a substantial activation energy barrier that appears to be greater on very small Fe crystallite^.^*^,^^^^,^^*^^ The adsorption-desorption technique initially used by Amelse et al. attempts to establish monolayer coverages by cooling from a high temperature to 273 K under hydrogen,14an approach supported by recent TPD results.46 The presence of a K promoter is known to increase the strength of hydrogen adsorption on iron.60*6' Studies of H2 adsorption on Co have been much more limited. Yates and Sinfelt found that the amount of hydrogen adsorption near 300 K was greatly influenced by the support used.62 More recently, TPD studies established the presence of a variety of adsorption sites for hydrogen on cobalt and confirmed that H2adsorption on Co is also an activated process.47 The most recent work used isotherms at 323-423 K to measure monolayer coverages of hydrogen on Co and found an adsorption stoichiometry of H/Co, near unity when the total H, uptake was chosen.51 The behavior of C O chemisorption on Fe and Co has been previously On cobalt it has typically been found to be nondissociative near 300 K,64-69and IR studies have led to the proposal that subcarbonyl species exist on Si02-supported CO?' whereas only bridged and linearly bonded CO species were found on low-surface-area Co films.7' A variation in CO chemisorption stoichiometries, when compared to H2 uptakes, was reported by Reuel and Bartholomew, who examined cobalt catalysts with different dispersions, metal loadings, and support material^.^' TWO different types of carbon were used in this last study, and the COad/Hadratio was found to vary between 0.4 and 2.3, depending on the support. The early work of Brunauer and Emmett showed that C O chemisorbs nondissociatively on both promoted and unpromoted catalysts at 195 K,58with C O adsorbing exclusively on the iron component of promoted catalysts.72 More recent ultrahighvacuum work on the (1 10) and (100) crystal faces of Fe indicates that some dissociation of CO may occur at 300 K,73974and Textor and Gay found that substantial dissociation takes place at 300 (57) Farrauto, R. J. AIChE Symp. Ser. 1975, 70, 9. (58) Brunauer, S.; Emmett, P. H . J . Am. Chem. SOC.1940, 62, 1732. (59) Sinfelt, J. H.; Yates, D. J. C. J . Catal. 1968, 10, 362. (60) Benziger, J.; Madix, R. J. Surf.Sci. 1980, 94, 119. (61) Ertl, G.; Lee, S. B.; Weiss, M. Surf: Sci. 1981, 111, L711. (62) Yates, D. J. C.; Sinfelt, J. H.; Taylor, W. F. Trans. Faraday SOC. 1965, 61, 2044. (63) Vannice, M. A. Cut. Rev.-Sei.Eng. 1976, 14, 153. (64) Vannice, M. A. In Catalysis: Science & Technology; Anderson, J . R., Boudart, M., Eds.; Springer-Verlag KH: Heidelberg, 1982; Vol. 111. p 139. (65) Moyes, R. B.; Roberts, M. W. J . Catal. 1977, 49, 216. (66) Bridge, M. E.; Comrie, C. M.; Lambert, R. M. Surf:Sci. 1977, 67,
__
79-4 -
(67) Prior, K. A.; Schwaha, K.; Lambert, R. M. Surf: Sci. 1978, 77. 193. (68) Papp, H. Surf. Sci. 1983, 129, 205. (69) Papp, H. Surf. Sci. 1985, 149, 460. (70) Heal, M. J.; Leisegang, E. C.; Torrington, R. G. J . Catal. 1978, 51, 314. (71) Gopalakrishnan, R.; Viswanathan, B. J . Colloid Interface Sci. 1984, 102, 2. (72) Emmett, P. H.; Brunauer, S. J . Am. Chem. SOC.1937, 59, 310. (73) Broden, G.; Gafner, G.; Bonzel, H. P. Appl. Phys. 1977, 13, 333. (74) Rhodin, T. N.; Brucker, C. F. SolidState Commun. 1977, 23, 275.
Chen et al. K at low CO coverages on Fe( 111).75 C O adsorption at 195 K on carbon-supported Fe was consistent with a stoichiometry of CO/Fe, = Jung et al. found that the CO uptake at 300 K could exceed that at 195 K by a factor of 3 or more on 1-3-nm Fe crystallites, implying that subcarbonyl formation could occur at 300 K on very small particle^.^^^^ Analysis of the adsorption results in Table I1 is complicated by three factors: the unestablished adsorption stoichiometries on either metal surface, the unknown influence of potassium on adsorption stoichiometry, and certain assumptions about the composition of the metal particles due to the large excesses of Co in some samples. Regardless, some general conclusions can be drawn. First, the COad/Hadratio varies between 0.5 and 5 on the fresh reduced samples, but after rereduction following the kinetic studies the values were much closer to unity except for the two pure metal cluster catalysts. This range of values is similar to that reported by Reuel and Bartholomew for supported Co catalyst^.^' Second, the samples containing high loadings of one pure metal cluster, either by design or due to residual impurities, such as Co,, Fe, Co2(high), Fe3Co, and Fe,, have very high COad/Hadratios near 4-5. Even if H coverages are not complete on these samples, this allows for the possibility that subcarbonyl formation was occurring on these small Co and Fe crystallites. Adsorption on Fe, Co, and mixed Fe-Co cluster-derived catalysts appears to be similar, consistent with the previous studies on Fe and Co surfaces. Finally, despite all the uncertainties, the ratios of adsorbed C O and H to the total number of metal atoms are high, which implies the presence of small metal particles. Thus, the use of MCC precursors to produce well-dispersed Fe-Co appears to have been verified. The absence of any detectable XRD peaks, except for the KFe3Co sample, further supports this conclusion. Because the Co and Fe were initally present in the zero-valent state, we have assumed that all the metal is reduced when calculating dispersions and crystallite sizes. Catalytic Activity. The activities for the unpromoted and promoted cluster-derived catalysts listed in Tables I11 and IV demonstrate some clear conclusions. Note that the specific activities also represent TOF values assuming 100% dispersion and thus represent minimum values. First, all of these catalysts are quite active after only a LTR step and in all probability do not require this pretreatment in H2 to become active syngas catalysts. The H T R step did not have a marked effect on the unpromoted samples although it increased the specific activity of the Co-only and Fe-only samples by 30-40% and it produced small decreases in the mixed-metal catalysts. The decrease in activity noted for some of the K-promoted catalysts after HTR may be a consequence of K covering the metal surface as the crystallites grow. Second, the presence of K produces a large decrease in activity, in agreement with earlier results.37 It also causes a large increase in CO, formation that is due in part to C O disproportionation, as indicated in Table IV by the ratios of CO, formed to C O reacted to hydrocarbons which exceed unity. Finally, some of these catalysts are very active, especially C-supported C O ~ ( C O )whether ~, judged by specific activity or on a TOF basis. Previous studies of nitrate-derived catalysts have shown that cobalt-only catalysts are typically 3-10 times more active then Fe-only catalysts regardless of support.5,13,'4-78,79 In this study, the Co, catalyst was also found to have a much higher activity than the Fe3 catalyst on either a TOF or a gram catalyst basis. The specific activities of these MCC-derived catalysts are compared to previous values from the literature in Table VIII. With regard to these comparisons, two items must be considered: the method of chemisorption used to count surface sites and crystallite size effects. The study of Reuel and Bartholomew, for
+
(75) Textor, M.; Gay, I. D.; Mason, R. Proc. R. SOC.London, A . 1977, 356, 37. (76) Boudart, M.; Delbouille, A,; Dumesic, J. A,; Khammouma, S.; Topsoe, H. J . Catal. 1975, 37, 486. (77) Jung, H. J.; Walker, P. L.; Vannice, M. A. J . Catal. 1982, 75, 416. (78) Vannice, M. A. J . Catal. 1977, 50, 228. (79) Vannice, M . A . J . Catal. 1982, 74, 199.
The Journal of Physical Chemistry, Vol. 90, No. 20, 1986 4817
C O Hydrogenation over Carbon-Supported Catalysts
TABLE VIII: Comparison of CO Hydrogenation Activities of Fe-Co Cluster-Derived Catalysts TOF - CO converted to H C s-lx 103
catalvst
temp, K
a
4.1% Co/C 4.1% Co/Si02 4.1 % c o / c 3.0% Co/C (type uu) 3.0% Co/C (Spheron) 3.0% Co/Si02 4.1% Co/C 4.5% c o / c 4.6% Co/Si02 4.1% Co/C 4.0% Co/Si02 0.67% Fe, 1.25% Co/C 3.85% Fe, 1.02% Co/Si02 4.4% Fe/C 4.4% Fe/C 4.94% Fe/Si02 5.0% Fe/Si02 4.12% Fe, 1.14% K / C 4.94% Fe, 0.065% K/Si02
453 453 498 498 498 498 523 523 523 548 548 523 523 523 548 523 548 523 523
2.9c 13 (est) 42.4 0.07c 0.64c 5.5c 156 36 506( 100‘ 16.4 4.0 3.2 6.4 15.0 77 0.8 15.0
catalyst
temp, K
4.1 % c o p 4.1% Co/SiO/ 4.1% C o / A I 2 0 2 4.1% Co/Cf 7.0% Co/SiO/ 5.0% Co/A1204 4.12% Fe, 1.1% K / C 3.9% Fe, 5.5% K/A1203 1.6% Co, 0.5% Fe,g 1.1% Ti/Si02 0.67%’ Fe, 1.25% Co/C
453 453 453 473 473 473 543 544 453 453
b 1.3‘
T O F - CH4 s-1 x 103
a 2.4c
18.8
35.8 0.005c 0.3c 2.6c 69 131 37d 20 224( 432e 13OOc 7 2‘ 13.5 8.6 1.6 1.1 0.6 2.3 1.2 6.5 770 16 1.4 0.2 5.5 total activity pmol CO/(s g metal) 2.67e 9.0 1.5 9.8 7.9h 5.4h 6.67 3.78 3.38 1.O‘
b
ref
1.oc
this 23 this 54 54 54 this 80 14 this 78 this 14 this this 14 78 this 14
15.8
59 24 191‘ 870‘ 7.1 0.2 0.4 160 0.3
study study
study study study study study study
ref this 23 23 this 30 30 this 37 36 this
study study study study
Based on H2 chemisorption on used sample after HTR. *Based on CO chemisorption at 298 K on used sample after HTR. ‘TOF includes CO, an integral formation. dBased on H 2 uptake at 373 K. (Extrapolated values. ’Derived from C O ~ ( C O ) ~gDerived . from H F ~ C O , ( C O ) , ~hUsing . reactor,
example, based the TOF of Co catalysts on the total uptake of Hz at the temperature of maximum adsorption,s’ whereas in this study, the methcjd used for Hz chemisorption included a short evacuation step at 273 K prior to the desorption step, which may have removed a portion of the more weakly adsorbed hydrogen; consequently, the number of Cos atoms may be somewhat underestimated, resulting in slightly higher TOF values compared to previous results for Co/C catalysts.51 However, our C O uptakes may overestimate Cos atoms; therefore the two bases used to calculate T O F values in Table VI11 are likely to establish upper and lower limits. Regardless, this uncertainty in stoichiometry is not significant enough to explain the much higher activity of the MCC-derived Co catalysts compared to most of those prepared from Co nitrate using carbon or Si0z.14,51-53,54,80 The TOFs obtained here are more directly comparable to those reported by Amelse et al.I4 because of the similarity of the Hz chemisorption techniques, and in this case the cluster-derived Co/C catalyst is 3 times more active than their Co/Si02 catalyst. Finally, this Co/C catalyst has a TOF, based on H2 chemisorption, that is about 6 times greater than a 4.0% Co/SiOz studied by V a n n i ~ e . ~ ~ Previous investigators have found that the T O F for both C O hydrogenation and methanation decreases as crystallite size decreases for both Fe and Co catalysts on a variety of supp o r t ~ ~ , and ~ ~ the , ~ TOF ~ . ~values ~ . ~for ~ the 4.4% Fe/C catalyst based on C O adsorption are very similar to previous results for very small Fe crystallite^.^.^^ Whether small Co crystallites also have intrinsically lower TOFs is not yet clear because of the variations between H2 and C O uptakes on larger Co crystallite^^^*^^ (see Table VIII). Although a lower TOF would be anticipated based on earlier s t ~ d i e s , ’ the ~ , ~small ~ * ~C-supported ~ Co crystallites
examined here exhibit quite high activity similar to that reported by Lisitsyn et N o effort was made in this study to determine the influence of large variations in crystallite size; however, most of the used samples appeared to have similar sizes ranging from 5 to 10 nm. As noted earlier, the activities of the potassium-promoted catalysts are much lower than the unpromoted catalysts in this study (see Table IV). It has been previously shown that at low levels of potassium promotion the activities of bulk Fe catalysts increase with increasing K concentration, but beyond a certain value, the activities start decreasing The levels of promoter in the catalysts of this study far exceed what are considered to be “optimal” values,83thus explaining the relatively low activity of the KFe, catalyst. The low activities of the K-promoted catalysts are similar to those reported by McVicker and Vannice3’ for complex-derived, K-promoted Fe supported on A1203. A comparison of the data in Table IV indicates that the rate of hydrocarbon formation over the KFeCo catalyst is higher than that over the nitrate-derived KFeCo-(NO,) catalyst, because most of the C O conversion over the nitrate catalyst can be attributed to the formation of C 0 2 . If the intrinsic activities of surface Co and Fe atoms are assumed to be unaffected by neighboring atoms, if the activities are considered to be additive, and if the crystallite compositions are homogeneous, then the average turnover frequencies of the MCC-derived samples are dependent only on the overall metal composition in a linear fashion, as shown in Figure 1. The 40-fold higher activity of the Co/C catalyst, compared t o the Fe/C catalyst, provides a wide variation in TOF values. The result that all the Fe-Co catalysts have average TOF values far below those
(80) Fernandez-Morales, I.; Geurrero-Ruiz, A.; Lopez-Garzon, F. J.; Rodriguez-Ramos, I.; Moreno-Castilla, C. Appl. Catal. 1985, 14, 159. (81) Fu, L.; Bartholomew, C. H. J . Catal. 1985, 92, 376.
(82) Dry, M. E. Catalysis, Science and Technology; Springer-Verlag: Berlin, Vol. 1, p 159. (83) Mross, W. D. Catal. Reu. Sci. Eng. 1983, 25, 591.
4818
The Journal of Physical Chemistry, Vol. 90, No. 20, 1986
expected based on metal analyses strongly implies that the metal surface is iron-rich. Such behavior has been found previously in Fe-Ru/C catalysts' and other Fe-Co system^.^*^^'^ It is consistent with surface segregation models that predict surface enrichment in iron,6~84~ss and the variations in selectivity also indicate that surface enrichment in iron has occurred.86 We do not think that phase separation has occurred to produce Fe-rich and Co-rich crystallites, each with identical surface and bulk concentrations, but that the Fe-Co clusters migrate to form crystallites with similar overall composition but higher surface than bulk concentrations of Fe. Further studies are needed to unequivocably verify this hypothesis. Selectivity. A direct comparison of the selectivity behavior of the Fe-Co cluster-derived catalysts with the available literature information is often complicated by the variety of operating conditions used in different laboratories. Under the reaction conditions here, the unpromoted Fe3 catalyst gave the lowest selectivity to methane (44 mol %), which was only about 2 / 3 of that found on comparable silica-supported Fe-only catalyst^^^^'^^^^ but the same as that found previously for small Fe crystallites on ~ a r b o n . ~ In . ? contrast, ~ the amounts of methane produced by the Co, catalyst are higher than values reported for silica-supported, Co-only ~ a t a l y s t s . Other ~ ~ ~ work ' ~ ~ has ~ ~indicated ~ ~ ~ a trend of increasing methane selectivity with decreasing particle size, regardless of the type of s ~ p p o r t . ' ~ , ~Consequently, ' the high methane selectivity found for our Co/C catalyst is consistent with the chemisorption and XRD data, which indicated the presence of small crystallites. For the bimetallic catalysts, the high methane selectivity found after the LTR step decreased after the HTR was performed, indicative of a surface enrichment of iron. Although the HTR step also enhanced Cz-C3 olefin/paraffin (o/p) ratios compared to the LTR step, these ratios were not as high as those found in other studies on Fe-Co alloy cataiysts.l6J8 The results here for the three Fe-Co catalysts within the narrow composition range of 0.20-0.25 mole fraction Fe imply the possibility that the o/p ratio increases with increasing particle size after HTR. An analogous result was reported earlier by Kellner and Bell for R U . ~ ' The FeCo3 catalyst does not follow this pattern, probably because of its higher Fe content. As seen in Tables V and VI, promotion of the Fe-Co catalysts by K caused a marked enhancement in olefin selectivity; for example, no olefins were obtained over the Co/C catalyst, while the addition of one K atom per Co produced only olefins, within detectable limits, plus CH4. Such high o/p ratios over Co are quite unusual as Co typically produces a highly paraffinic produ c t . ' * ~A ~ ~similar high selectivity to olefins over K-promoted Fe clusters has been observed, but greater amounts of higher molecular weight products were found in that An inspection of Table VI shows that most of the promoted catalysts produced hydrocarbons with a carbon number of 6 or lower, despite the high levels of promoter relative to the metal. Again, it is possible that a particle size effect is responsible for this behavior, although detectability limits due to low conversions are also likely to contribute. Previous workers have noted that the average hydrocarbon molecular weight decreases with decreasing particle size over Cos' and R u . The ~ ~ distribution toward lower molecular weights over these well-dispersed MCC catalysts would therefore be consistent with the small particle sizes implied by the lack of diffraction peaks for almost all of the promoted catalysts. Finally, the overall absence of C, or heavier hydrocarbons is worth noting as i t is somewhat unique for Co and Co-rich catalysts. Activation Energies and Reaction Orders. The activation energies of the unpromoted catalysts listed in Table VI1 are in reasonable agreement with most of the literature values for supported Fe and Co ~ a t a l y s t s . but ~ ~ Bartholomew ~ ~ ~ ~ ' ~ ~and ~ ~ ~ (84) Burton, J. J.; Machlin, E. S. Phys. Reu. Lett. 1976, 37, 1433 ( 8 5 ) Van Santen, R. A.; Sachtler, W. M. H. J . Catal. 1974, 33, 202. (86) Chen, A ; Kaminsky, M.; Geoffroy, G. L.: Vannice, M. A,, unpublished results. (87) Kellner, S. C.; Bell, A. T. J . Caral. 1982, 75, 251. (88) Agrawal. P. K : Katzer. J. R.; Manogue. W. H. J . Catal. 1981, 69, 312
Chen et al.
0
L
16
8
0
24
32
"
45
ON S T R E A M ( h l
TIME
Figure 5. Activity maintenance behavior of KCo: CO conversion to hydrocarbons and ratio of C02 to CO converted to hydrocarbons vs. time on stream, T = 553 K, H2/C0 = 3.0 P = 1.0atm: (0)CO conversion to hydrocarbons; (0)(TOF COz)/(TOF CO to HC).
oa
4
07-
O4
t
031 0
4
0
1
I
1
0
16
1
1
1 32
24
T I M E ON STREAM
( h )
Figure 6. Activity maintenance behavior of KCo: Anderson-SchulzFlory chain growth probability a vs. time on stream, T = 553 K, H2/C0 = 3.0, P = 1.0 atm.
Reuel frequently found significantly higher activation energies for Co on a number of supports, including c a r b ~ n Differences .~~~~~ in preparative technique and the supports are presumed to be responsible. Higher activation energies were observed for our promoted catalysts, a trend consistent with that reported for K-promoted, A1203-supportedcluster catalysts.37 The large increases in activation energies are a reflection of the high promoter content, since the work of Amelse et al.I4 and Campbell and Goodmang9showed that lightly promoted catalysts had similar activation energies to the unpromoted catalysts. In general, the hydrogen partial pressure dependencies of the unpromoted catalysts were near first order, in agreement with The addition values found in the literature for Fe and Co.63-77390 of K to the clusters increased the H2pressure dependence, which may reflect a change in the rate-determining step as no deactivation was observed with these catalysts during the measurement of their partial pressure dependencies. The CO partial pressure dependencies for the unpromoted catalysts were always negative and ranged from -0.2 to -0.7, similar to values for Fe and Co from other s t ~ d i e s The . ~ ~ promoted ~ ~ ~ ~catalysts ~ usually showed a more negative dependence than the unpromoted samples, an observation consistent with an increased CO heat of adsorption on these samaples. Such an increase has been predicted from theory and has been observed previously with promoted iron
catalyst^.^^^^^-^^ ~ Activity Maintenance. While the deactivation behavior of the unpromoted catalysts appeared quite typical, an unusual gain in both the methanation and overall activities following a steep initial (89) Campbell, C. T.; Goodman, D. W. Surf. Sci. 1982, 123, 413. (90) Dry, M. E.; Shingles, T.; Boshoff, L. J. J . Catal. 1972, 25, 99. (91) Dry, M. E.; Shingles, T.; Boshoff, L. J.; Oosthuizen, G. J. J . Catal. 1969, 15, 190. (92) Broden, G.; Gafner, G.; Bonze], H . P. Surf. Sci. 1979, 84, 295.
J . Phys. Chem. 1986,90, 4819-4824 deactivation was observed for the KCo catalyst after approximately 20 h on stream, as shown in Figure 4. Figure 5 shows that the hydrocarbon activity went through a minimum, while a monotonic decrease in the rate of C02 formation was noted throughout the run. At the end of the run, the COZ/HC ratio was near unity, indicating that the rate of carbon deposition had decreased while the hydrogenation increased. It is reasonable to expect that good initial contact between K and Co was achieved through the use of MCCs as catalyst precursors, and it may be that the strong effect of the promoter begins to diminish as carbon deposition proceeds. This is supported by the continuously declining chain growth probability for C3-Cs hydrocarbons in Figure 6. Some decrease in K-Co interaction could result from the physical removal of K from the catalyst surface; however, the continued high selectivity to olefins would not be expected had all the K been removed. Also, atomic absorption analyses on the fresh and used samples did not show any loss of metal or potassium for this catalyst. The work of Bonzel et al.93 showed that when carbon was deposited on K-promoted Fe foils the location of the K was always at the top of the carbon layer. The carbon buildup observed during the initial stages of the activity maintenance run on KCo could have, therefore, removed a fraction of the promoter from the cobalt surface. This possibility is consistent with the work of Audier et a1.I8 Some cleaning of the catalyst surface may be occurring (93) Bonzel, H. P.; Broden, G.; Krebs, H. J. Appl. SurJ Sci. 1983,16,373. (94) In this paper the periodic group notation in parentheses is in accord with recent actions by IUPAC and ACS nomenclature committees. A and B notation is eliminated because of wide confusion. Groups IA and IIA become groups 1 and 2. the d-transition elements comprise groups 3 through 12, and the p-block elements comprise groups 13 through 18. (Note that the former Roman number designation is preserved in the last digit of the new numbering: e.g., 111 3 and 13.)
4819
after 20 h on-stream because of the increased C O to hydrocarbon activity observed. Although water production was not explicitly quantified in this study, an inspection of the chromatograms taken near the end of the activity maintenance run showed a sharply reduced water peak. This implied that at these longer times, the fraction of the total COz formed by the water-gas shift reaction had increased, while the C 0 2 resulting from carbon deposition was reduced. The suppression of the carbon deposition rate after longer times on-stream indicated that no reassociation of the K with the partially cleaned Co surface took place. In summary, we have shown that well-dispersed, carbon-supported Fe, Co, Fe-Co, and K-promoted Fe-Co catalysts can be prepared from stoichiometric mixed-metal carbonyl clusters. These catalysts can be very active for C O hydrogenation and do not require a high-temperature reduction step because of the initial zero-valent state of the metal atoms and the absence of oxygen functional groups on the carbon surface. The Co/C catalyst derived from C O ~ ( C Owas ) ~ especially active, produced only paraffins, and exhibited very good activity maintenance. The Fe/C catalyst had much lower turnover frequencies but a much higher olefin/paraffin ratio. The mixed-metal clusters showed catalytic behavior intermediate between these two extremes, and the unpromoted samples displayed more Fe-like behavior after a hightemperature treatment. Addition of a potassium atom to the MCCs markedly decreased activity but greatly enhanced olefin selectivity. Further control of the K/metal ratio by coimpregnating MCCs and K-containing MCCs could provide a method to optimize olefin yields.
Acknowledgment. This study was supported by the National Science Foundation through Grant No. CPE-8205937. Registry No. CO, 630-08-0; C O ~ ( C O )10210-68-1; ~, Fe3(C0),*, 17685-52-8; Fe, 7439-89-6; Co, 7440-48-4; K, 7440-09-7.
A Kinetic Study of the Catalytic Oxidation of CO over Nafion-Supported Rhodium, Ruthenium, and Platinum Vincent D. Mattera, Jr.? Denise M. Barnes,*Sanwat Noor Chaudhuri,s William M. Risen, Jr.,* Department of Chemistry, Brown University, Providence, Rhode island 0291 2
and Richard D. Gonzalezl Department of Chemistry, University of Rhode island, Kingston, Rhode island 02881 (Received: December 20, 1985)
The Rh, Ru, and Pt ionomers of perfluorocarbonsulfonic acid (Nafion or PFSA) polymers have been formed and have been reduced to investigate the formation of metal particles within the ionic domains of the materials. The particle size distributions are peaked in the 25-40-A range postulated for these domains and are consistent with exertion by the ionomer of morphological control over the growth of transition-metal particles. The reduced ionomers catalyze the CO oxidation with the activity sequence Ru > Rh > Pt. Diffusion limitations occur over the Rh and Ru, but not the Pt, ionomer catalysts. Turnover frequencies for the formation of COz by oxidation of CO are reported as a function of temperature.
Functionalized organic polymers have been used as supports for heterogenized homogeneous catalytic processes.' Typically, transition-metal-containing complexes which are either active catalysts or potential catalytic precursors are bound to the support
by covalent attachmenk2 Although there have been reports of catalytic reactions involving transition-metal complexes incorporated into ion-exchange (polyelectrolyte) resins?" those materials are not thought to form ionic domains or to have morphological properties of ionomers. Ionomers are polymers that are func-
+AT&T Bell Laboratories, 600 Mountain Avenue, Murray Hill, N J 07974. 'AT&T Bell Laboratories, 5 5 5 Union Boulevard, Allentown, PA 18103. $Department of Fuels and Engineering, University of Utah, Salt Lake City, UT 84112. Department of Chemical Engineering, University of Illinois at Chicago, P.O.Box 4348, Chicago, IL 60680.
(1) Chauvin, Y.; Commereuc, D.; Dawans, F. Prog. Polym. Sci. 1977, 5 , 95. (2) Smith, T. W.; Wychick, D. J . Phys. Chem. 1980, 84, 1621. (3) Drago, R. S.;Nyberg, E. D.; A'amma, E. A.; Zombeck, A. Inorg. Chem. 1981, 26, 641. (4) Smith, R. T.;Ungar, R. K.; Baird, M. C. Tramition Met. Chem. ( N Y ) 1982, 7, 288.
Introduction
0022-3654/86/2090-4819$01.50/0
0 1986 American Chemical Society
