J. Phys. Chem. 1991, 95, 1195-7804 and OCH3-are weaker, 1.7 and 0.07 eV, respectively, for 5-fold coordinate AI3+. These ideas also apply to H migration across A1203surfaces. The surface capacity of alumina for hydrogen is about IO-' times the number of possible O H groups,42 which suggests that most of the chemisorbed hydrogen occupies defect sites and will diffuse rapidly until finding such sites, which is also what our calculated chemisorption energies would indicate. Decomposition of the Heterolytic Pair: Methane vs Methanol Formation. Since the heterolytic pair is so stable, methanol generation from it would be a high-temperature process, and indeed none has been observed experimentally. Related to this is the strongly held dissociative chemisorption phase of methanol which has been observed on activated (partially dehydroxylated) 6-A1203.43 From microcalorimetry, the heat of adsorption was estimated to be -2.3 eV/CH30H in the low-coverage limit. The adsorption was thought to be heterolytic, with CH,O- bound to AI3+of low coordination. The reverse reaction, which corresponds to the recombination of the heterolytic pairs and the desorption of C H 3 0 H , would then be endothermic by 2.3 eV. Our calculations yield 2.12 eV for this process when CH30- is bound to 3-fold coordinated AI3+. Recombination to form CH4 and an 0 atom on the A13+ is endothermic by only 0.65 eV according to our calculations, making this a likely process. The 0 atom + 02or the two 0-which are formed by the elimination of methane would react quickly with gas-phase or spilled-over hydrogen to form surface OH or water. If H2 migrates over the surface by a heterolytic H+-02- + H--AI3+ pair mechanism, the reaction with 0 would yield two OH- and a cus AI3+.These different types of OH, those associated with H- and those not, should be dis(43) Busca, G.; Rossi, P. F.; Lorenzelli, V.; Benaissa. M.;Travert, J.; Lavalley, J.-C. J. Phys. Chem. 1985, 89, 5433.
7795
tinguishable by vibrational measurements. Such measurements might be helpful in determining whether the hydrogen that spills over diffuses over the A1203surface in the heterolytic form. The energy barrier for CH4formation cannot be calculated with the present A1203cluster model. The reaction might be expected to be of the SN2type, with H- from OH- on the surface displacing 02-from the adsorbed methoxy anion. The activation energy for OH- displacement from ROH by H-in the gas phase has been calculated to be about 21 kcal/moL4 To model this process would require a large cluster with a "catalytic pocket" where O H would be directed toward CH30- in such a way that a four-centered O-H..CH,-O transition state could form. Such sites would not be characteristic of a smooth surface but would be associated with steps or defects. A mechanism wherein CH; is released, subsequently capturing H to form CH4, would require weakening of the methoxy C-0 bond. This would happen if the surface of the alumina is reduced by, say, reductively homolytically adsorbed hydrogen so that electrons occupy AI band gap surface dangling orbitals. The reasons were given in a recent p u b l i c a t i ~ nand ~ ~ also account for the weakness of the 02--methoxy bond discussed above: when a radical binds to 02-, the electron promotion energy in the u* orbital renders the bond weak. Acknowledgment. S.-F. Jen is grateful for a graduate Fellowship from the BF Goodrich Co. Registry No. Carbon monoxide, 630-08-0; methoxy, 2143-68-2; methoxide, 33 15-60-4; hydrogen ion, 12408-02-5; methane, 74-82-8; methanol, 67-56-1; alumina, 1344-28-1. (44) Shi, Z.; Boyd, R. J. J. Am. Chem. SOC.1990, 112, 6789. (45) Shiller, P.; Anderson, A. B. J . Phys. Chem. 1991, 95, 1396.
Primary and Secondary Reaction Pathways in Ruthenium-Catalyzed Hydrocarbon Synthesis Rostam J. Madon,+ Sebastian C. Reyes, and Enrique Iglesia* Corporate Research Laboratories, Exxon Research and Engineering Co., Route 22 East, Annandale, New Jersey 08801 (Received: January 18, 1991)
Residence time studies show that n-paraffins, a-olefins, and cis-2-olefins are primary products during hyL. ocarbon syn . A s on Ru catalysts. Their formation, as well as that of branched isomers, is consistent with previously proposed surface reactions of alkyl groups on metal surfaces. Secondary hydrogenation and hydrogenolysis of a-olefins are inhibited by the water product of the synthesis step. However, a-olefin readsorption and surface chain initiation and cis-to-trans isomerization take place as secondary reactions. The decrease in a-olefin selectivity with increasing CO conversion and molecular size reflects the greater extent of readsorption as bed and pore residence times of a-olefins increase. Readsorption of a-olefins and chain initiation increases with molecular size because the rate of removal of olefins from liquid-filled catalyst pores is decreased due to intraparticle diffusion limitations. This diffusion-enhanced olefin readsorption accounts for the observed deviations of carbon number distributions from those predicted by Flory polymerization kinetics. Chain growth probability increases with chain length until an asymptotic value is reached. Readsorption decreases the contribution of termination via hydrogen abstraction to chain growth kinetics and leads to a heavier, more paraffinic product. In effect, differences in selectivity between small and large hydrocarbons are due to the increasing influence of pore residence times and the decreasing influence of bed residence times on secondary reactions as olefin size increases.
Introduction It became a p arent during early studies of Fischer-Tropsch (IT)synthesis', that olefins, paraffins, and oxygenates were all part of the product spectrum. Herrington' first suggested that growing hydrocarbon chains on catalyst surfaces terminate as
P
'Current address: Engelhard Corporation, 101 Wood Ave., Iselin, NJ 08830.
*To whom correspondence should be addressed.
paraffins or olefins and that the latter readsorb and initiate new growing chains. Later, Friedel and A n d e r ~ o nusing , ~ thermodynamic arguments, concluded that a-olefins and oxygenates were the major primary products. Pichler et al.,5-6on the basis of bed (I) (2) (3) (4) (5)
Fischer, F.; Tropsch, H.Ges. Abh. Kennr. Kohle 1928, 10, 3 13. Pichler, H.Ado. Cafal. 1952. 4 , 271. Herrington, E. F. G.Chem. Ind. 1946. 347. Friedel, R. A.; Anderson, R. B. J . Am. Chem. SOC.1950, 72, 1212. Pichler, H.; Schulz, H.; Elstner, M. Brenstoff-Chem.1967, 48, 78.
0022-365419 112095-7795%02.50/0 0 1991 American Chemical Society
7796 The Journal of Physical Chemistry, Vol. 95, No. 20, 1991
residence time studies, proposed that a-olefins were primary products that readsorbed and isomerized to internal olefins or were hydrogenated to paraffins. They suggested that a-olefins could also react to form larger molecules or crack into smaller ones.s The readsorption of a-olefin products and their further participation in the chain growth process are crucial reactions in FT synthesis. A theoretical analysis by Novak et al.' showed that readsorption of a-olefins (Cl;) leads to a decrease in the net termination rate of growing C, chains, consequently increasing the average molecular weight of the product. They assumed, in their analysis, that a-olefin readsorption rates are independent of molecule size. In this paper, we describe residence time studies on Ru/Ti02 and Ru/Si02 catalysts; these studies allow us to distinguish primary and secondary reaction steps that control hydrocarbon chain growth and product distribution. Our study focuses on the three major products of the FT synthesis on Ru catalysts: nparaffins, a-olefins, and &olefins. Branched hydrocarbons, which in our study constitute a much smaller fraction of the products, and the effect of olefin addition to H2 and CO during synthesis, are discussed elsewhere, We describe here how selectivity trends for light hydrocarbons differ markedly from those for heavy hydrocarbons. We also propose a model that describes the key role of a-olefin readsorption in controlling hydrocarbon functionality and carbon number distribution in FT synthesis.
Experimental Section Experiments were carried out in a 122-cm-long stainless steel reactor tube with 0.77-cm internal diameter heated by a three-zone copper furnace.8 Temperature and gas flows were controlled by using Eurotherm and Brooks controllers, respectively. A thermocouple well, embedded in the catalyst bed, allowed a travelling thermocouple to measure axial temperature profiles along the catalyst bed. Carbon monoxide, dihydrogen, carbon dioxide, and light (C5J hydrocarbons were analyzed on-line by using a Carle (Model 31 1 AGC) gas chromatograph. Heavy products (C5+) were collected and analyzed separately in a Perkin-Elmer gas chromatograph using a 2% SP-2100 column (Supelco) and in a high-temperature gel permeation chromatograph (Waters model 150C) using 1Styragel columns. Detailed analyses of the C5+ fraction were performed in a Perkin-Elmer gas chromatograph using a 30 X 0.25 mm i.d. fused silica capillary column coated with SE-54 (J & W Scientific). Two catalysts were used in the present study. Both catalysts (1.1 % Ru/Ti02 and 10.8% Ru/Si02) were prepared by using a Ru(NOJ3 solution (Engelhard) and a "slurry" technique in which the support is stirred into excess acetone containing the Ru salt and the solution evaporated at room temperature. Degussa P25 Ti02and Davison grade 62 Si02were sieved to yield pellet sizes between 80 and 140 mesh. Before catalyst preparation, Ti02was reduced in flowing H 2 at 823 K for 4 h and Si02was calcined at 873 K for 24 h. After Ru impregnation, the catalysts were reduced for 4 h at 723 K in flowing H2, cooled, passivated, and stored. After loading into the reactors, the catalysts were reduced again in flowing H2 for 4 h at 673 K. The metal content of the samples was measured by X-ray fluorescence. The Ru dispersion in the Si02-supported sample, measured by static H2 chemisorption, was 20%. Ru dispersion is defined as the percentage of metal atoms residing at crystallite surfaces. The metal particle size for Ru/Ti02 was obtained by transmission electron microscopy; the fresh catalyst showed a number average value of IO A, whereas the used catalyst showed a number average value of 22 A. Residence time studies were carried out by varying the flow rate of the H 2 / C 0 gas mixture over a packed bed of catalyst. In our experiments, we used 32.5 g (40 cm3 catalyst bed volume) ( 6 ) Pichler, H.; Schulz, H.; Hojabri, F. Brensioff-Chem. 1964, 45, 21 5. (7) Novak, S.;Madon, R. J.; Suhl, H. J . Caial. 1982, 77, 141. (8) Madon, R. J.; Taylor. W. F.Ado. Chem. Ser. 1979, 178, 93.
Madon et al. TABLE 1: Hydrocarbon Synthesis Rates on 1.1% Ru/TiOz and 10.8% Ru/SiOZ
space velocity, CO rate, mmol site-time yield, mol of V/V.h conversion, % of CO/g.h CO/g-atom surface Ru-s
Ru/Ti02" 1500 1000 500 300
18 26 53 85
3000 1998 1008 705 53 I
14 21 47 61 80
0.145 0.141 0.144 0.138
2.28 2.22 2.26 2.18
X
x 10-2 X X
RU/S~O~~ 0.121 0.1 I4 0.130 0.117 0.116
1.81
X
1.70 X 1.94 x 10-2 1.76 X 1.74 X
1.1 w t % Ru, 2 2 4 average Ru particle diameter; 0.50 dispersion assuming hemispherical crystallites; 483 K reaction temperature. 10.8 wt % Ru, 0.20 dispersion by hydrogen chemisorption; 477 K reaction temperature.
of Ru/Ti02 and 8.7 g (20 cm3 catalyst bed volume) of Ru/Si02. The latter system was diluted with 20 cm3 clean quartz powder of the same size range (80-140 mesh) as the catalyst. The travelling thermocouple used in these experiments showed axial temperature gradients of less than f l . 5 K. Ru/Ti02 was studied at 483 K and four gas-hourly space velocities from 300 to 1500 V/V.h with corresponding CO conversions of 85 to 18%. Space velocity is defined as the volumetric (STP) flow rate of CO and H2 per unit bed volume. Ru/Si02 was studied at 477 K and five space velocities from 530 to 3000 V/V.h with corresponding CO conversions of 80-14%. The absolute pressure in both sets of experiments was 550 kPa, and the H 2 / C 0 ratio was 2.0 f 0.03. The conversion versus reciprocal space-velocity plot was linear up to about 80%conversion on both catalysts, suggesting that space-time yields are independent of conversion level. Sufficient time (>75 h) was allowed between space-velocity changes in order to ensure steady-state behavior. This is particularly important when C5+products are collected for analysis. The C4- gaseous products, however, reach steady-state within a few hours (C24 h) after changes in space velocity. The catalyst did not deactivate during the experimental runs (>500 h) after steady-state conversion was initially reached within 48 h. For example, Ru/Ti02 was initially run at 1500 V/V-h for 230 h with a constant 18 f 1% CO conversion. After 336 h and experiments at other space velocities, the CO conversion was 19 f 1% at 1500 V/V.h and the selectivity was unchanged. On Ru, only a minor amount of C 0 2 is formed by the water gas shift reaction. The exit H 2 / C 0 ratio equals the inlet H 2 / C 0 ratio (irrespective of conversion or bed axial position) because the inlet ratio (2.03) closely resembles the stoichiometric usage ratio (2.05-2.10). In summary, we conclude that our results are not masked by catalyst deactivation or by H 2 / C 0 axial gradients along the fixed-bed reactor. Also, as discussed previously? our experiments are not influenced by heat transfer or by diffusional limitations of H, and CO. Results The effect of bed residence time on product selectivity is described here by plotting the latter versus space velocity [GHSV, V/V.h; calculated at 298 K and 101 kPa], inverse space velocity (space-time), or conversion. Selectivities in Figures 1-8 are given as the percentage of the converted CO that appears as a particular product; it is a carbon selectivity. Table I describes hydrocarbon synthesis rates on both catalysts at several values of space velocity. Site-time yields are independent of space velocity and very similar on Ru/Ti02 and Ru/Si02. I n all discussions that follow, the synthesis rate of a given hydrocarbon is proportional to its se(9) Madon, R J.; Bucker. E. R.; Taylor, w. R. Department of Energy, Final Report. Contract No. E (46-1)-8008. July 1977.
Ru-Catalyzed Hydrocarbon Synthesis
The Journal of Physical Chemistry, Vol. 95, No. 20, I991 7797 I
o.6
t
I
1
ETHANE
1
Ru/TiO,
l o
0 0.0
1 Zoo0
I
loo0
$1
I
3wo
SPACE VELOCITY (V/V.h)
3~
Figure 1. Space velocity effects on ethylene and ethane selectivity ( 0 , m, Ru/TiO,; 0, 0 , Ru/SiOz) [550 kPa, 483 K (Ru/TiOz), 477 K (Ru/SiOl), Hz/CO = 2.031.
Ii
I
I
500
loo0
1500
SPACE VELOCITY (V/V.h)
Space velocity effects on methane and C5+ selectivity (Ru/Ti02) [550 kPa, 483 K, Hz/CO = 2.031.
Figure 4.
3
I
I
I
Ru/TiO,
1-Butene
h
1-Nonene
/"
.
\
>
->
- 2
Propylene
s
t,
>-I-
W
n-Nonane 0-0
s
w 2 fJY
-I
W
fJY1
\I
/ I
n-Octadecanp
Q, -
0
3.0
I
-
yo 1-Butene
1500
O-
-
/ O
-
1000
Figure 5. Space velocity effects on C9and CI8selectivity (Ru/TiO,) [550 kPa, 483 K, H$O = 2.031.
RulSiO, 2.5
500
SPACE VELOCITY (V/V.h)
I
I
2-Octadecene
1
6 Propylene
.------A I-
0 W
A'
'
7
ts:
1.0
-*-*-e
: : : ; ; ;n 0
-A-A-A
00
. I _ 1000
2000
3000
SPACE VELOCITY (VIV.h)
Fipn kPa, 4
500
1000
1500
SPACE VELOCITY (V/V.h)
Space velocity effects on C3 and C, selectivity (Ru/Si K, H2/CO = 2.031.
) [550
lectivity, because the total conversion rate is independent of space velocity or CO conversion. Figures 1-3 show the effect of space velocity on C2, C3,and C4selectivities. The trends are identical on Ru/TiO, and Ru/ Si02. Olefin selectivities decrease with decreasing space velocity (increasing bed residence time). The paraffin selectivity, however, remains constant with space velocity.
Figure 6. Space velocity effects on CJ5paraffin selectivity (Ru/Ti02) [550 kPa, 483 K, H2/C0 = 2.03).
Figure 4 shows that C5+hydrocarbon selectivity decreases and CH4 selectivity increases with increasing space velocity. Identical trends are observed on Ru/Si02. In Figure 5 and 6 , we use C9, CIS, and C35as examples of selectivity trends in higher hydrocarbons. Here, I -nonene decreases while n-nonane and 2-nonene increase with decreasing space velocity. The trend for C9paraffin differs markedly from that for ethane, propane, or butane. The selectivity trends for CI8also differ slightly. n-Octadecane selectivity increases with decreasing space velocity in the same manner as n-nonane but unlike ethane, propane, or butane. 1-
7798 The Journal of Physical Chemistry, Vol. 95, No. 20, 1991
--
Ip n s
i
Ru/TiO, c4
CIS + TRANS
Madon et al. 20
i
,/-•
15
w
-1
W
v)
0
0.4
2 a v)
z 4.
10
U
t SPACE VELOCITY (V1V.h)
Figure 7. Space velocity effects on cis- and rrum-tbutene selectivity (Ru/Ti02) (550 kPa, 483 K, H2/C0 = 2.031. 5,
5
I
I
c*o
0 01 o'20
I-l
500
SPACE VELOCITY ( V / V * h ) Figure 9. Space velocity effects on cis/trans 2-olefin ratio (Ru/Ti02)
[ S O kPa, 483 K, H2/C0 = 2.031.
o
oo++-+7
I
RulTiO,
SPACE VELOCITY ( V I V - h )
t
01500
1000
'.
CIS
1 8
9
10 11 12 13 14 15 16 17 18 I 9 20 21 2 2
CARBON NUMBER
Figure 10. Carbon number effects on cis/trans 2-olefin ratio (Ru/Ti02) (1500 V/V.h, 550 kPa, 483 K, H2/C0 = 2.031. > k
I:
01 0
5
I
10
15
1
20
25
I
30
35
1
40
CARBON NUMBER
Figure 11. Carbon number and space velocity effects on chain termination probability (8.) (Ru/Ti02) (550 kPa, 483 K, H2/CO = 2.031.
(10)
Golden, D. M.; EEger, K. W.; &nson, S.W.J . Am. Chem. SOC.1964,
86. 5416.
velocities for a light molecule (C,) and for a heavy molecule (Cm). The cis/trans 2-butene ratio decreases with decreasing GHSV and extrapolates to the thermodynamic equilibrium value of 0.55. In contrast, the Cm cis/trans 2-olefin ratio is independent of space velocity; it equals 0.7, the thermodynamic equilibrium value, at all bed residence times. Cis/trans 2-olefin ratios (Figure 10,
The Journal of Physical Chemistry, Vol. 95, No. 20, 1991 7799
Ru-Catalyzed Hydrocarbon Synthesis
Isomers (Internal Olefins, Branched Hydrocarbons) t 0.6
1
1
1 1
0
1
a-Olefin (aC;)
Ru/Ti02, 1500 V/V.h) steadily decrease as carbon number increases up to about C19;then, the ratio remains constant at the thermodynamic equilibrium value. Large internal olefins reach thermodynamic equilibrium cis/trans ratios even at high space velocity ( I 500 V/V*h), whereas light 2-olefins (C,) do only at very low space velocity (
