308
Energy & Fuels 1992,6, 308-314
Hydrocarbon Selectivity from Cobalt Fischer-Tropsch Catalysts Ian C. Yates and Charles N. Satterfield* Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139 Received July 12, 1991. Revised Manuscript Received March 3, 1992
A cobalt Fischer-Tropsch catalyst was studied in a continuous-flow,well-stirred slurry reactor at 220-240 "C, 0.5-1.5 MPa, H2/C0 feed ratios between 1.5 and 3.5, H2 conversions between 6 and 68%, and CO conversions between 11 and 73%. Increasing space velocity (decreasing conversion) or decreasing reador H2/C0 ratio decreased the yield of (undesired) C1 products and increased the yield of (desired)Clo+ products. Reactor temperature and pressure had little effect on the carbon number distribution. The extent of the readsorption of l-alkenes, formed by the primary synthesis, into growing chains on the catalyst surface in contrast to hydrogenation to the corresponding alkane or isomerization to the 2-alkene reflects an intricate interplay among many variables.
Introduction The chemistry of the Fischer-Tropsch synthesis can be described as the polymerization of single-carbon units on The distribution of products may be a catalytic characterized by a parameter a,the Schulz-Flory chain growth probability; in ita simplest formulation a is the likelihood that a C, species on the catalyst surface will add another carbon and become a C,,, species, rather than desorb as a product. More than one chain growth probability may exist, and in some interpretations of experimental data the chain growth probability is treated as increasing continuously with molecular size. The existence of product distributions with two distinct chain growth probabilities, a, for low carbon numbers and a2for high carbon numbers, has been observed experimentallyon iron catalysts by many investigator^^-^ and has been reported also for cobalt in a study by Schulz.s On iron-based Fischer-Tropsch catalysts, a sharp 'break" from a,,with a value usually near 0.6, to a2,having a higher value, usually near 0.9, occurs on a semilogarithmic plot of mole fraction versus carbon number, typically at about Clo. The double-a and a mathematical development for its interpretation are discussed in detail by Donnelly et Much less information is available on cobalt catalysts than on iron catalysts, especially over a substantial range of molecular weight of products. The secondary reactions of l-alkenes can have a significant effect on the observed product distributions on cobalt. l-Alkenes can incorporate into growing chains, the extent of which is affected by competitive reactions, largely l-alkene hydrogenation and isomerization to the 2-alkene. Literature Review Reports on hydrocarbon distributions from cobalt catalysts date back to the early days of the Fischer-Tropsch (1)Herington, E.F. G. Chem. Ind. 1946,65,346. (2) Anderson, R. B. The Fischer-Tropsch Synthesis; Harcourt, Brace & Jovanovich: New York, 1984.
(3)Donnelly, T. J.; Yates, I. C.; Satterfield, C. N. Energy Fuels 1988, 2,734. (4)KBnig, L.;Gaube, J. Chem.-Ing. Tech. 1983,55, 14. (5)Huff, G. A., Jr.; Satterfield, C. N. J. Catal. 1984,85,370. (6)Dictor, R. A.; Bell, A. T. J. Catal. 1986,97,121. (7)Donnelly, T. J.; Satterfield, C. N. Appl. Cotal. 1989,52,93. (8) Schulz, H. Report to the Bundesministerium fur Forschung und Technologie 'Katalysatoren und Selektivitiitalenkung bei der FischerTropsch-Synthese", BMFT-FB-T 80-124,Nov., 1980.
synthesis in Germany, in which cobalt was used in industrial plants during World War 11. Results from this early German work and other studies immediately following World War 11are summarized by Storch et al.9 On the Co/Th02/MgO/kieselguhr catalysts studied, the selectivity to heavy products increased with decreasing H2/C0 ratios, and reportedly went through a maximum with respect to pressure between 0.5 and 0.85 MPa and remained unaffected by conversion. From a considerable number of more recent studies, we comment only on those most closely related to our work. Product distributions can be affected by many variables such as operating pressure, temperature, degree of conversion, H2/C0 ratio, catalyst composition, and nature of reador, such as integral fixed bed versus a well-mixed flow reactor. Thus some reported results may seem to be contradictory. Borghard and Rennett'O studied a 34 wt % Co/Si02 catalyst at 2.03 MPa and 250 "C with a H2/C0 feed ratio of 2 in a fixed bed reactor and compared the product distribution with that from several iron catalysts. In differential reactor studies performed at 250 "C, atmospheric pressure, and conversions below 2% on a Co/M2O3 catalyst, Rautavuoma and van der Baanl' report data on the C1-C7 hydrocarbon distribution. Values of CY varied from 0.49 to 0.82. Beuther et a1.12 describe FischerTropsch synthesis on 100 co:18 Th02:200A 1 2 0 3 and 21.9% Co: 0.5% Ru:2.2% ThO2:74.5% A 1 2 0 3 catalysts. The data are somewhat difficult to interpret, but the selectivity to higher hydrocarbons can be inferred to decrease with increasing temperature. The most extensive results are from a variety of studies by Schulz and co-workers. Some of these appear in rather inaccessible sources and frequently work with cobalt is discussed in conjunction with more substantial studies on iron. The most extensive studies were apparently obtained by Schulza on five different cobalt compositions, all containing Thozand a silica gel support. Three also contained MgO. Reaction conditions were varied from 0.1 to 3.3 (9)Storch, H. H.; Golumbic, N.; Anderson, R. B. The Fischer-Tropsch and Related Syntheses; John Wiley and Sons: New York, 1951. (10)Borghard, W. G.; Bennett, C. 0. Ind. Eng. Chem. Process Res. Deu. 1979,18, 18. (11)Rautavuoma, A. 0.I.; van der Baan, H. S. Appl. Catal. 1981,1, 247. (12)Beuther, H.; Kibby, C. L.; Kobylinski, T. P.; Pannell, R. B. United States Patent 4,399,234,August 16, 1983.
0887-0624/92/2506-0308$03.00/00 1992 American Chemical Society
Hydrocarbon Selectivity from Cobalt Catalysts MPa, 175 to 210 "C, and H2/C0 ratios usually of 1.8-2.0, utilizing a fixed bed reactor. Molar distributions were reported up to about CWfor three of the five catalysts. A clean double-a product distribution was reported, breaking at about Clo, but the effect was much more pronounced for the catalysts containing MgO, and it varied significantly with temperature and conversion. The effect was attributed to incorporation of 1-alkenes formed as primary products. From this study and a later report summarizing much of his work, Schulz et al.13 drew a number of generalizations. The ratio of alkenes to alkanes decreased with increasing carbon number, increased with increasing pressure, and decreased with increasing space velocity (lower conversion). These are basically trends in alkene hydrogenation. Higher molecular weight alkenes were postulated to be more strongly adsorbed; the reduction in hydrogenation at higher pressures was attributed to inhibition of alkene adsorption by increased adsorption of CO. At high space velocities (low conversion) increased hydrogenation was attributed to higher hydrogen pressure. Hydrogenation increased with increasing temperature, which can also be interpreted in terms of competitive adsorption of CO that would be expected to decrease at higher temperatures. Methane selectivity was reported to be in the range of 7-14 carbon % (% of total C in hydrocarbon products, qualitatively very similar to wt % ) and increased with increasing temperature. The formation of 2-alkenes was lower at low conversions and decreased with increasing pressure and temperature. The isomerization of 1-alkenes to 2-alkenes increased with increasing carbon numbers, possibly as a result of longer residence times of higher molecular weight 1-alkenes. Operating with a Berty reactor at 250 OC, Sarup and Wojciech~wski'~ examined the effect of H2/C0 ratio on product distribution over a Co/Si02 catalyst. Products from C1 to C30 were identified as predominantly linear alkanes, monomethyl isomers and some 1-and 2-alkenes. The C3+ products were reported to follow a single-a Schulz-Flory distribution. However, graphs of the data taken after a significant time on stream show a possible second-a beginning in the Clo range. The chain growth probability decreased with increasing H2/C0 ratio. Fu et al.15 studied 10 and 15 w t % Co on A1203at 0.1 MPa in a differential reactor. Temperature was varied from 200 to 235 "C, H2/C0 ratios from 0.5 to 3, and a Cfold range of space velocities was examined. Selectivities are reported in terms of C-number ranges, the highest being C12+. The resulting a1values varied from about 0.6 to 0.8. Chain growth probability decreased with increasing inlet H2/C0 ratio and increasing temperature.
Energy & Fuels, Vol. 6, No. 3, 1992 309
l
0.001
r
' 0
I
5
10
I5
20
Carbon Number. Cn
Figure 1. Representative Schulz-Flory diagram showing a double-a-type distribution (220OC, 1.48 m a , feed rate = 0.032 L(STP)/min/g of cat.). (H2/CO)h = 1.64. (H2/CO) in reactor = 1.65. al = 0.54,a2 = 0.91,and s2 = 5.4.
and intermediates leaving the porous catalyst (see later). The fact that in a slurry reactor data can be obtained at uniform temperature and composition facilitates interpretation, in contrast to difficulties that may be encountered with data from a fixed bed reactor. The reactor was initially charged with 400 g of n-octacosane that had been previouslyrecrystalked in tetrahydrofuran (HPLC grade, >99.9% purity) to remove a bromine impurity.le The catalyst was prepared by an outaide facility and is of the approximate composition of the cobalt catalysts used at Ruhrchemicas The nominal composition, as reported to us by ita manufacturer, is 21.4 wt % Co (as Co), 3.9 wt % Mg (as Mg), and the remainder diatomaceous earth. Supplied as an extrudate, the catalyst was ground and sieved to 52-92 wm (170-270ASTM mesh). For data acquisition reactor temperature was varied between 220 and 240 "C. Pressure was varied from 0.5 to 1.5 MPa and H2/C0 feed ratios from 1.5 to 3.5. By changing space velocity between 0.085 and 0.008 L(STP)/min/g of cat. (unreduced basis) (cat. = catalyst) total synthesis gas conversions were varied from 11 to 70%. Data were obtained during the course of one run with the same catalyst that extended over nearly 2200 h. Rate data were obtained simultaneously and that portion of the study has been p~blished.'~That paper describes the method of operation, trapping, and analysis of products and other details. The activity of the catalyst was constant throughout the study.
Hydrocarbon Carbon Number Distributions Cobalt catalysts form mostly straight chain hydrocarbons. In the range of Clo to C20,such products are of value as diesel and jet fuels. Heavier waxy producta in the C20+range can be hydrocracked back to lower molecular Experimental Section weight fuels. Particular attention was paid here to the selectivity to the desired Clo+ fraction and the undesired The experimenta were performed in a continuous, mechaniC1 fraction. Some of the effects of operating conditions cally-stirred, 1-L autoclave. The slurry reactor and ancillary on the secondary hydrogenation or isomerization of 1 4 equipment are described in detail else~here.'~J'The reactor and kenes formed by the primary reaction were also estimated. ita contents are well-mixed and the reactor operated free of Representative Product Distributions. Figures 1 mass-transfer limitations of H2and CO into catalyst parti~les.'~J~ However,there may have been diffusionallimitations on products and 2 show representative Schulz-Flory diagrams of products volatilized from the reactor. The C3+ data are well described by a double-a Schulz-Flory model. The (13) Schulz, H.;Rosch, S.; Gokcebay, H. Proc. 64th CIC Coal Symp., solid line in Figures 1 and 2 is the best-fit nonlinear reOttawa 1982. gression of a double-a model as developed by Donnelly et (14) Sarup, B.; Wojciechowski, B. W. Can. J. Chem. Eng. 1984,62,249. aL3 Above about Cl5-CzO,the overhead product distribu(15) Fu, L.; Rankin, J. L.; Bartholomew, C. H.C l Mol. Chem. 1986, 1, 369. tion deviates increasingly from that actually formed be(16) Huff, G. A., Jr.; Sattertield, C. N. I d . Eng. Chem. Fundam. 1982, cause of retention of heavy products in the reactor. 21, 479.
(17) Donnelly, T. J.; Satterfield, C. N. Appl. Catal. 1989, 56, 231. (18) Huff, G. A., Jr.; Satterfield, C. N. Ind. Eng. Chem. Process Des. Deu. 1984, 23, 696.
(19) Yates, I. C.; Satterfield, C. N. Energy Fuels, 1991,5, 168.
Yates and Satterfield
310 Energy & Fuels, Vol. 6, No. 3, 1992 l t
0.6 1
-
I
C
0
0.4
3
0 8
i
0
0.0 5
10
I5
20
Carbon Number, Cn
0.1
0.01
0.00 1 35
40
45
50
Fraction o! C ,
1
~
’
0.020
0.050
0.080
Space Velocity Wminlgcatl
Figure 2. Representative Schulz-Flory diagram showing a double-a-typedistribution (230O C , 0.79MPa, feed rate = 0.020 L(STP)/min/g of cat.). (H2/CO) in reactor = 1.39. (H,/CO)i, = 1.55. CY^ = 0.54,a2 = 0.88,and fl = 5.
30
1
I
’
0.001
55
Carbon Number, Cn
Figure 3. Schulz-Flory diagram for reactor liquid at completion of run. a2 = 0.87. This model was used as a basis for comparison of the data from this study with data from previous studies on iron catalysts. For the two material balances shown, a1 is calculated to be about 0.54 and az near 0.89. The ‘break” carbon number, the carbon number at which the contributions of both a1and a2are equal, is near 5, which is lower than is typically observed on iron catalysts.2‘t22 For a description of this two-site interpretation of the double-a, see refs 4, 20, and 22. Figure 3 shows a Schulz-Flory diagram of a wax sample taken from the reactor slurry at the end of the run. The value of azestimated by linear regression of the data between Cm and (&was 0.87. Although this sample represents the s u m of all product distributions from the entire run, note that this value of a2is close to that calculated by the nonlinear regression from the overhead products. The asymptotic linear relationship holds over a wide range of carbon numbers, indicating that chain growth proba(20) Huff, C. A.,Jr. Fischer-Tropsch Synthesis in a Slurry Reactor. Sc.D. Thesis, Massachusetts Institute of Technology, Cambridge, MA, 1982. (21) Matsumoto, D. K. The Effects of Selected Process Variables on the Performance of an Iron Fischer-Tropsch Catalyst. Sc.D. Thesis, Massachusetts Institute of Technology, Cambridge, MA, 1987. (22) Donnelly, T. J. Product Distributions of the Fischer-Tropsch Synthesis. Ph.D. Thesis, Massachusetts Institute of Technology, Cambridge, MA, 1989.
Figure 4. Effect of space velocity on product yield. The Clo+ yield is greater at higher space velocities. Data at 240 O C , 0.79 MPa, and (H2/CO), = 2. bility reaches a constant value at high carbon numbers. The Effect of Operating Parameters. A useful method of reporting results on product distributions is to calculate al,az,and either s1, the breakpoint between the two distributions, or xl, the product fraction formed by the a1distribution. In addition, we determine respective 95% confidence intervals or standard deviation of parameter estimate. However, s1 or x1 is difficult to estimate precisely from regression of experimental data and, therefore, an additional method of reporting selectivity data is useful. Because of low volatility, products at higher carbon numbers tend to remain in the slurry liquid. Therefore, to calculate a complete distribution of the products being synthesized,the mole fractions above about C15 must be estimated. The procedure we used here is as follows. First, the data from C3 to CN were fitted and estimates of al, cyz, and s1 made. Second, the mole fractions from CN to Cleo were extrapolated and used to generate estimates of the ‘data” in terms of weight fractions. All products above CN .are assumed to have the molecular weight of alkanes. Third, the actual weight of products produced at C1 and Cz are included and the data are then expressed in terms of weight classes. Weight classes which are industrially relevant are C1, Cz-C4(light gas), C5-Cg (gasoline), and Clo + (diesel and wax). Finally, to check that the estimate of the weight classes is reasonable, the closure on carbon, including the extrapolated hydrocarbons, is estimated and required to fall between 95 and 105%. To simplify presentation of the results here, only C1 and Clo+ products are reported. These two product classes are representative of undesirable and desirable products, respectively. Space Velocity. Figure 4 shows the effect of space velocity on the yield of products at 0.79 MPa, 240 OC, and Hz/CO feed ratio of 2. This Hz/CO feed ratio is near the usage ratio on cobalt and therefore varying space velocity has little effect on reactor Hz/CO. Increasing space velocity (decreasing conversion) increases the fraction of Clo+, while decreasing the yield of C1 products. The effect may plausibly be related to the extent of chain incorporation of 1-alkenes relative to being hydrogenated or isomerized to 2-alkenes. At higher space velocity (lower residence time), 1-alkene are hydrogenated or isomerized less than at lower space velocities because of the decrease in residence time of the 1-alkenes. Pressure. Figure 5 shows that, at 220 “C and 0.017-0.018 L(STP)/min/g of cat. (unreduced basis) of
Energy & Fuels, Vol. 6,No. 3, 1992 311
Hydrocarbon Selectivity from Cobalt Catalysts
0.30
I A Data
o.6
E
I
2
0
1_1_
E
LL I
s M
."
5
0.1 0
L
0.00
0.0 1.00
0.50
0
1.50
1
ratio, 220 "C.
0.50
Icl
C
0 4
s M
3
E
0.4
I
4
Figure 7. Increase of weight fraction of C1with increasing H,/CO
t I
3
p"2/pco
Pressure [MPal
Figure 5. Plot showing that pressure has no noticeable effect on the product distribution. Data at 220 "C and feed rate of 0.017-0.018L(STP)/min/g of cat. (H2/CO)i, = 2.
.-0 0 E LL
2
.-0 4d
A
V
0.40
2
:\
I
I
A
\
s M
I
0.2
ldi Fraction ot
0.0
'
200
C,
.-
f
0.30
0.20
I
220
240
1 l-' 0
A
Data
Liwar
l A
\
\
ReQreSSlm
I,
1
2
3
4
260
Temperature [TI
Figure 6. Effect of temperature on product distribution. Tem-
perature has no effect on product distributions. Total synthesis gas conversions are between 31 and 33%, allowing comparison of similar product to reactant ratios. H2/C0 = 2 synthesis gas, the selectivity to C1 and CIO+ products remained constant over a range of total reactor pressure of 0.5-1.5 MPa. In a study at 175 "C, and pressures of 0.12-3.3 MPa, using a (H2/CO) inlet ratio of 1.8, Schulzs reported that methane selectivity decreased with increasing pressure. The Clo+ product distribution scattered considerably with pressure. The (H2/CO) exit ratio was nearly constant, in the range of 1.6-1.7. Temperature. Figure 6 shows the dependence of the weight fractions of C1 and Clo+ for total synthesis gas conversions between 31 and 33% at 220 and 240 "C. Data are presented at comparable conversion, rather than equivalent space velocities, because the ratio of product to reactant concentration appears to have a marked effect. No trend is observed over this limited temperature range. In a study at 1.7 MPa and 170 to 190 "C, using a (H2/CO) inlet ratio of 1.0, Sch@ reported that methane formation increased with temperature but there was little effect of temperature on the Clo+ distribution. However, in his work the (H2/CO) ratio varied substantially through the reactor. The exit value decreased from 0.65 at 170 OC to 0.24 at 190 "C. Reactor H2/C0 Ratio. Figure 7 shows the effect of reactor H2/C0 ratio on the relative yield of C1 products at 220 "C. Similar plota were generated showing the
Figure 8. Decrease of weight fraction of Clo+ with increasing
H2/C0 ratio, 220 "C.
analogous trends at 230 and 240 0C23but are not shown here. Increasing reactor H2/C0 ratios increases the relative weight fraction of C1 at all temperaturea, although the trend is more apparent at lower temperatures. Methane may be formed by a mechanism separate from chain growth, which may have a positive dependence on the PH /Pco ratio, as is examined below. fiigure 8 shows the effect of reactor H2/C0 ratio on the fractional yield of products in the Clo+ range at 220 OC. Again, similar plots were generated showing the analogous trends at 230 and 240 OCF3 but are not shown here. Increasing reactor H2/C0 ratio decreases the yield of highmolecular-weight products relative to total hydrocarbons synthesized. This decrease can be primarily attributed to the increase in rate of production of low-molecular-weight products, particularly methane. This general trend has also been reported by others.8J0J5
Selectivity to Various Product Classes Figures 9 and 10 show component Schulz-Flory diagrams including the distribution of three major classes, n-alkanes, 1-alkenes, and n-alcohols. Three other components were observed in much lower concentrations than these at each carbon number; in order of relative abun(23) Yates, IC., The Slurry-Phase Fischer-TropschSynthesis. Ph.D. Thesis, Massachusetts Institute of Technology, Cambridge, MA, 1990.
Yates and Satterfield
312 Energy &Fuels, Vol. 6, No. 3, 1992 1
.-
!
f
1
0.1
c
n-Alcokols
1
l
E E
/'
t
.a
0
rt
-
0.01
0
E"
0.00 1
,
1
,
,
5
10
0.0001
0
15
80
0 0 00
20
0.20
0.40
0 52
P,, [MPal
Figure 11. Rate of methane plus methanol formation, well fitted by eq 1.
-
1 Total
8
0 n-Alkanes
s-
EE
400 Y
240'C
A 1-Alkenes
C
.*
0 n-Alcohols
0.1 l o j :A**
:
.a
a 0
****
0.01 ;
P
E
200
-
i
08686
A
0 c
i
5a
A * OOe :Oooo~. A
s
A A
0.001 T
A
0
A 0
A A
A
0.000 1
0 ' 0.00
0.20
0.40
0.60
P,, [MPal C a r b o n N u m b e r , Cn
Figure 10. Component product distribution showing primary producta, n-alkanes, 1-alkenes,and n-alcohols (240 OC, 0.79 MPa, and feed rate of 0.035 L(STP)/min/g of cat.). (H2/CO), = 2.15. dance, they were 2-alkenes, branched alkanes, aid aldehydes (the last only at C2and C3and in very low concentrations). Methane lies above the line that would be predicted by a double-a Schulz-Flory correlation, while C2 products lie below, as is characteristic of most Schulz-Flory diagrams of hydrocarbon p r ~ d u c t s .The ~ observed Czconcentration on cobalt is generally less than on iron.8 Both the n-alkanes and n-alcohols exhibit a double-a type distribution; that is, at low carbon numbers, the mole fraction of products drops off quickly, while, at higher carbon numbers, the mole fraction drops off more slowly. Above C3,the fraction of synthesized products that are hydrogenated increases with increasing carbon number. In contrast to n-alkanes and n-alcohols, 1-alkenes appear to follow a single-a type distribution in Figures 9 and 10. The extent to which observed products represent the primary synthesis versus secondary reactions varies depending on reactor conditions. l-Alkenes, presumed to be the primary product of the synthesis by Schulz et al.,I3 may be hydrogenated to n-alkanes or isomerized to 2-alkenes. The ratios of 1-alkeneln-alkane and l-alkene/Balkene both decrease with increasing carbon number, as was also observed by Schulz8 and Rautavuoma and van der Baan." As shown below, a large fraction of n-alkanes appear to be produced by the hydrogenation of 1-alkenes.
Figure 12. Rate of formation of Cz+ compounds,well fitted by eq 1. Since the formation of C1products appeared markedly different from the formation of other components, we examined the fit of the rate expreasion developed elsewhere for the consumpton of synthesis gas to the formation of C1 and C,+ products as two separate groups. The expression id9
Figure 11shows that the rate of C1formation is well fit by the linearized form of equation 1. Figure 12 shows that the rate of Cz+formation is also well fit by eq 1. Thus, the decrease in yield of higher molecular weight products with increasing Hz/COdoes not seem to result from competing rate mechanisms for formation of C1versus C,+ products.
Secondary Reactions of 1-Alkenes In characterizing secondary reactions, the rates of formation of ethane, n-butane, and 2-butene were studied. Ethane represents hydrogenation only; the behavior of the C4compounds is taken as representative of that of the C,+ products. As discussed by Hanlon and S a t t e ~ f i e l dand ,~~ Donnelly and Satterfield,' C4 products are the highest molecular weight products which do not split between (24) Hanlon, R. T.;Satterfield, C. N. Energy Fuels 1988,2, 196.
Energy & Fuels, Vol. 6, No. 3, 1992 313
Hydrocarbon Selectivity from Cobalt Catalysts vapor and liquid phases in our traps and therefore are the least subject to experimental error. There are several publications on the effects on product distribution of adding ethene to synthesis gas. Most of these deal with iron catalysts. A recent paper by Adesima et al.= discusses product distribution from adding ethene, utilizing a cobalt/kieselguhr catalyst at 200 "C and 0.1 MPa, and gives guidance to the literature. The most clean-cut conclusionscome from studies with labeled compounds. In using nonlabeled compounds, generally relatively high concentrations of alkene were added, e.g., 5-10 mol % of the synthesis gas feed, to magnify reaction effects and make possible overall material balances to search for the fate of the added alkene. The results therefore may depart significantly from those encountered when the alkene concentrations are those produced by the primary reaction, which are much lower. The most complete study of the effects of adding selected labeled compounds during synthesis on cobalt is probably that of Schulz et alF6 They added ethene (14C), propene (l-14C and 2-I4C), butane (lJ4C), 2-methylpentadecane (15-14C),and 1-hexadecene (l-15C) at concentrations ranging from 0.10 to 0.78 vol % of total synthesis gas feed. They ran their fixed-bed reactor at 185-190 "C, 0.1 MPa, with H2/C0 = 2 synthesis gas at a space velocity of 75 L(STP)/min/L of cat. (unreduced basis). For the 1-alkenes studied, hydrogenation to alkanes was the predominant reaction, with 67 mol % of the added ethene hydrogenated,51 mol % of the propene, and 79 mol % of the 1-hexadecene. Of the added 1-alkene which was not hydrogenated, they found that 88 mol % of the ethene was incorporated into growing chains, 63 mol % of the propene, and 31 mol % of the 1-hexadecene, thus producing higher molecular weight products. Kibby et al.,27in a scouting study available in preprint form, co-fed ethene or propene at 10 vol % with (H2/CO)B of 1 or 2 to a reduced 30 wt 3' % Co, 5 wt % Tho2, on 65 wt % AlzO, catalyst in an internal recycle reactor operated at 195 OC and 0.79 MPa. Present analysis of their data at (HP/CO)in= 2 indicates that the weight percent of C5+ hydrocarbons increased to 62.5 wt % with ethene addition and 54.7 w t % with propene addition, as compared with 44.6 wt 5% with no added 1-alkenes. With etbpne addition the weight percent of C,+ hydrocarbons increased to 85.4 wt 7% at (H2/CO)B= 1from 62.5 wt % at (HZ/CO), = 2. This effect of decreasing incorporation with increasing H2/C0 ratio is consistent with a set of reaction pathways for alkenes in which hydrogenation and incorporation both consume 1-alkenes competitively; thus, at a higher H2/C0 ratio, more alkenes are hydrogenated and consequently less are incorporated into growing chains. Most alkene addition studies reported on cobalt were performed in conjunction with a more extensive study of feed additions to iron catalysts; as a result, cobalt catalysts were studied at only a few different process conditions. Although 1-alkenes appear to be incorporated into growing chains on cobalt, there is little agreement as to the extent to which incorporation affects the product distribution; further, no studies exist which cover a wide range of process conditions; for example, the only study at a pressure above 1 atm is apparently that of Kibby et al. and apparently only Adesima et al. varied the H2/C0 ratio. A systematic study of the combined effects of pressure, ~~~~
(25) Adesima, A. A.; Hudgins, R. R.; Silveston, P.L.Appl. Catal. 1990, 62, 295. (26) Schulz, H.; Rao, B. R.; Elstner, M. Erdol Kohle 1970, 23, 651. (27) Kibby,C. L.;Pannell, R.B.; Kobylinski, T.P.Prepr.-Am. Chem. Soc., Diu. Pet. Chem. 1984, 29(4), 1113.
1.00
I
0.80
.
I
,/A
A A
v
0.60
\
?I
\-
la
0.40
I
d Oa20
I
,I :z 1 Rewewon
0.00 0.0000
0.0005 P,,'P,,/P,
0.0010
WAI
Figure 13. Rate of ethane formation. Most ethane is produced from ethene, according to a simple hydrogenation model; data at 220 "C.
0.60
I
A
0.40 9 &I \
2
1.0s
E,
L
=s
0.20
d
0.00 0.000
l0.002 P,,,P",/P,,
Linear Reqession
0.004
0.006
IMP4
Figure 14. Rate of n-butane formation. Most n-butane is produced from 1-butene,according to a simple hydrogenation model; data at 220 OC. temperature, and H2/C0 ratio remains to be performed. Rate of Ethane Formation from Ethene. Figure 13 shows the dependence of the rate of formation of ethane on the ratio of P ~ # H , / P c O in the reactor; data at 220 "C are shown as representative. This assumes that ethane formation is a simple hydrogenation process inhibited by adsorbed CO. While the data scatter somewhat, much of the variation in the rate of ethane formation can be explained by such a simple model. The vertical axis intercept of this figure represents the amount of ethane produced as primary product. By this model, over mast of the range of operating variables studied, the amount of ethane produced by the primary synthesis would be less than half of the total amount synthesized. Rate of n -Butane Formation from 1-Butene. Figure 14 shows the dependence of the rate of formation of nbutane on the ratio of Pc,H&/Pco in the reactor; data at 220 "C are shown as representative. Again, a simple hydrogenation process inhibited by adsorbed CO is assumed. While there is considerable scatter in the data in Figure 14, the rate of n-butane formation clearly increases The with increasing values of the ratio PCH, &H' JPCO. vertical axis intercept of this figure is related to the amount of n-butane produced in the primary synthesis; over the range of operating variables studied, the amount of n-
Yat es and Sat t erfield
314 Energy & Fuels, Vol. 6, No. 3, 1992 50000 A
inhibited by carbon monoxide. Equation 2 can be linearized, yielding
Data
P1-butene - 1 KCSCO -+-
RZ-butene
30000
20000 d
'a
'p
10000
a0 0.00
0.40
0.20
0.60
P,, [MPal 20000
1-
Data Linear Repression
I
I
~
3
P
3
.-i E
10000
:
L
P
d J 0.00
0.20
0.40
0.60
P,, [MPal
Figure 15. (a, top) Rate of 2-butene formation. Most 2-butene is produced from 1-butene,according to a simple isomerization model (see eq 2), 220 "C. (b, bottom) Rate of 2-buteneformation. Most 2-butene is produced from 1-butene,according to a simple isomerization model (see eq 2), 240 O C .
butane produced by primary synthesis by this simple model would be generally less than half of the total amount synthesized. The behavior of these two simple hydrogenation models is consistent with the observations of Schulz8J3and Rautavuoma and van der Baan" on cobalt and Donnelly and Satterfield' on iron that increasing hydrogen to carbon monoxide ratio decreases the 1-alkeneln-alkane ratio. Rate of 2-ButeneFormation from 1-Butene. 2-A1kenes are not considered to be a primary product but are assumed to be produced solely by isomerization of l-alkenes.13 To represent the rate, a Langmuir-Hinshelwood-type relationship of the following form is proposed:
k
(3)
Following the relationship of eq 3, Figure 15, a and b, depicts the dependence of the ratio P1-hbne/RZhMe on Pco at 220 and 240 "C, respectively. Both figures show that increasing Pco will decrease the rate of 1-butene isomerization. The conclusion of Schulz et al.13that the ratio of 1-alkenes/total alkenes increases with increasing total pressure is consistent with eq 2 since the range of values of H2/C0 through their fixed bed reactor was within the range of 1.8-1.6 throughout his wide range of pressure. In very recent careful studies by Iglesia et al.28and Madon et al.% on ruthenium-catalyzed synthesis in a fixed-bed reactor, the observed deviations from Flory polymerization kinetics were interpreted as being caused by readsorption of primary 1-alkenes which was enhanced by slow diffusion out of catalyst pores. Alkene transport would logically decrease with increased molecular weight, causing a continuously increasing observed value of a and decreasing alkene content with increased carbon number. The effects of bed residence time were also interpreted by the same model. Contrary to the conclusions of most previous studies, they proposed that selective 1-alkene readsorption and chain initiation instead of secondary hydrogenation accounted for the observed effects of bed residence time on the 1-alkene/alkane ratio. These studies were performed on 80-140 mesh size catalysts, supported and unsupported, at 0.1-2.0 MPa, 203 OC and at a H2/C0 ratio of 2.1/1. This ratio was the stoichiometric one and hence it was essentially constant through the bed. These papers were not published until after the conclusion of our experimental work, and we did not consider the possibility of olefin pore diffusion limitations. The extent to which the above model may apply to cobalt catalysts is uncertain, but it does impose a new perspective that will need to be considered in future studies on cobalt catalysts. The extent of 1-alkene readsorption to form heavier products in contrast to hydrogenation to the corresponding h e or isomerization to the 2-alkene reflects an intricate interplay among many variables. These include catalyat composition, including the role of promoters and supports, particle size, pore size and residence time, and the effects of pressure and temperature. The partial pressure of CO, the H2/C0 ratio and their changes with degree of conversion plus a variety of competitive adsorption effects among reactants and producta add further complexities.
Acknowledgment. This study was supported by the
Office of Fossil Energy, U.S.Department of Energy, under contract No. DE-AC22-87PC79816. Registry No. Co, 7440-48-4;CO, 630-08-0.
Equation 2 assumes that the rate of 2-butene formation is simply proportional to the concentration of 1-butene and
(28) Iglesia, E.; Reyes, S. C.; Madon, R. J. J. Catol. 1991, 129, 238. (29) Madon, R. J.; Reyes, S. C.; Iglesia, E. J. Phys. Chem. 1991, 95, 7795.
