Ind. Eng.
Chem. Process Des. Dev., Vol. 18, No. 1, 1979 163
Kinetics of the Fischer-Tropsch Reaction over Iron Harvey E. Atwood and Carroll 0. Bennett' Department of Chemical Engineering, University of Connecticut, Storrs, Connecticut 06268
Product distributions and reaction rates were measured for the reaction of a mixture of 66.7% H2-33.3% CO over a commercial promoted, fused iron catalyst. Data are presented for a pressure of 2.0 MPa, over the temperature range 523-588 K, for a range of conversion up to 70 % . The results are used for t h e preliminary design of an energy storage system based on the Fischer-Tropsch synthesis. Product distributions over a cobalt-silica catalyst are also presented.
Introduction The Fischer-Tropsch synthesis may provide an attractive method of storing energy in connection with the operation of a combined-cycle power plant using gasified coal as a fuel. The gasifier could be sized according to the average load rather than the peak load if its product could be stored. The direct storage of large volumes of the synthesis gas is cumbersome, but if part of the gas is converted to a liquid, its volume becomes manageable. The product from the gasifier is cleaned and scrubbed to acceptable concentrations of sulfur compounds. The resulting low-Btu fuel is ideal as feed to a gas turbine. The exhaust gas from the turbine is used in the combined cycle to raise steam for an auxiliary turbogenerator. By this means the integrated clean system becomes economically competitive with a direct coal-fired plant with stack-gas scrubbing (Kydd, 1975). For the past 30 years commercial experience with the Fischer-Tropsch process has been limited essentially to SASOL, the South African coal conversion facility. Some engineering know-how for this plant is also available in the U S . from M. W. Kellogg and from Fluor. The literature has recently been reviewed by Vannice (1976) and by Mills and Steffgen (19731, supplementing the comprehensive earlier works of Anderson (1956) and Storch et al. (1951). However, direct experience with readily available commercial catalysts is lacking. The present study is a preliminary step toward providing an independent basis for the design of a Fischer-Tropsch plant. Commercially available catalysts have been compared, and the results were given by Borghard and Bennett (1977). Promoted, nitrided fused iron ammonia synthesis catalysts showed good activity and stability. A cobalt-silica catalyst was the most active. These results were based on tests lasting from 200 to 300 h, and no tests for sulfur resistance were made. The latter subject is reviewed in a recent publication by Madon and Shaw (1977). T h e present work is aimed at the preliminary design (simulation) of a commercial reactor to handle 250 million standard cubic feet per day of synthesis gas. For this design we needed the intrinsic chemical kinetics of the reaction, and these data were obtained in a gradientless reactor. In small-scale tubular reactors (Bennett and Borghard, 1979) we have found that a 1:1 mixture of HZ/CO led to plugging by wax and carbon formation. A 2 1 H,/CO mixture did not exhibit this problem: this stoichiometry corresponds also to the usage ratio for the cobalt catalysts. Studies were made over two catalysts-a CCI nitrided fused iron catalyst and a Harshaw precipitated cobalt silica catalyst. The pressure was 2.00 MPa. and the temperature range was 250 to 315 "C. Space velocities were varied to give as large a range of conversions as possible. The pressure was chosen to be compatible with 0019-7882/79/1118-0163$0 1.OO/O
Table I. Calculated Response Factors compound
factor
H*
0.0 1.00 0.64 1.34 0.57 0.66 0.52 0.51 0.51 0.49
co CH, CO,
c,
H*O c 3
c4 c5
C6
c, c* c,, c 9
0.42 0.40
0.37 0.36
compound Cll Cl, c,3
c,, c,,
c "3
36
MeOH EtOH C30H C40H C,OH
factor
0.34 0.31 0.31 0.30 0.30 0.30 0.64 0.57 0.82 0.94 1.57 1.49
a gasifier and the downstream gas turbine; the temperatures correspond to those expected in a fixed bed based on information in the literature already cited. In the consideration of the energy-storage application, all the synthesis gas is sent to the Fischer-Tropsch reactor during an off-peak period, and the conversion is chosen to give an exit gas, after condensation of C3+hydrocarbons, which has 75% of the heating value of the synthesis gas. It is thus important to measure the product distribution as well as the kinetics for the two catalysts at the various conditions already given. On the basis of the experimental results, various commercial reactors have been simulated to show the effects of some of the parameters in their design. A one-dimensional, heterogeneous, plug-flow model was used for the calculations. Experimental Methods The flow diagram for the apparatus (see Figure 1) is essentially the same as that used by Borghard and Bennett (1979). In the latter work four tubular reactors were operated in parallel in the reactor oven. In the present work the gradientless reactor is placed in parallel with the oven, and only one tubular reactor is used. No kinetics were measured in the tubular reactor, but its product distribution at various conversions has been included in the data on that subject. In all the work reported here, curves of mole fraction of product vs. conversion of CO have been found to be the same for either reactor and for all the temperatures studied. The intrinsic reaction rates on which the reactor simulation will be based were obtained in the gradientless reactor described by Brown and Bennett (1970). All lines were electrically heated where necessary to prevent condensat ion. For these quantitative experiments it was important to have accurate chemical analyses of the product. Gas chromatography was used, and response factors were C 1978 American Chemical Society
164
Ind. Eng. Chem. Process Des. Dev., Vol. 18, No. 1, 1979 APPARATUS
o-Hp-Iron Hp-Cobalt
0-
He
z
Y LL
010
010
F
030
050
0.70
090
CONVERSION OF CARBON MONOXIDE
Figure 2. Mole fraction of hydrogen and carbon monoxide in the product stream. Figure 1. Flow diagram of apparatus: TCD, thermal conductivity detector; CO, chromatographic oven; REF, reference columns; S, switching valve; G, sampling valve; VO, valve oven; RO, reactor oven; C , flow control valve; P, pressure regulator; F, feed tanks; GR, gradientless reactor; T, traps; R, rotameter; V, volumeter; Id,liquid sample post; GC, gas column; AC, aqueous column; HC, hydrocarbon column.
e
o 0
Y
- Cobalt
- Iron
Table 11. Chromatographic Conditionsa phase gas
LHC ~
aqueous ~~~~
Chromosorb W none (80/100) active phase Porapak Q 5% Carbowax Chromosorb 1 0 2 (80/100) 20M (80/100) in. x IO f t in. x 10 f t dimensionsb in. x 1 0 ft Heflow 104 23 26.5 (mL/min) sample size 1 mL 1 PL 1 CtL temperature 40-250 "C 40-260 "C 75-260 "C ramp 30 "C/min 30 "C/min 30 "C/min compounds 30 70 10 reference Porapak Q 5% Carbowax Porapak Q column 20M Detector temperature = 300 "C; injector temperature = 325 "C; thermal conductivity detector: WX tungsten All columns filaments; filament current = 100 mA. were stainless steel. support
none
determined for all the important components; these are given in Table I. A negative hydrogen peak was obtained, but this could not be used as an accurate measure of the hydrogen concentration. The response factor of H2 is infinite at 17% H 2 in He (Minter, 1968), and these conditions were approached in some of the peaks passing the detector. It was possible, however, to calculate the hydrogen in the product from a material balance. The only unknown in the product is the H2 mole fraction. If a molecular weight of the product is assumed, the flow of reacting gas mixture can be calculated; it was not measured. The known composition and flows of the three product phases (gas, liquid hydrocarbons, and aqueous phase) are used to calculate the concentration of H2 in the product gas which is necessary to satisfy the material balance. The resulting product molecular weight is then computed and compared with the assumed value. This molecular weight is then altered by a simple convergence scheme, and the calculations are iterated until a constant product molecular weight is attained, within a reasonable tolerance. A computer program was written for these calculations. Details can be found in Atwood's thesis (1977). The conditions used in the columns are given in Table 11. The columns were arranged so that the reference
010
I
I
I
I
030
050
070
090
CONVERSION OF CARBON MONOXIDE
Figure 3. Hydrogen/carbon monoxide usage ratios.
helium flow could be switched to the various reference columns as the different analytical columns were used. This procedure was important, for it assured the regular base lines shown in the chromatograms. If the same reference column is used for the three analyses, the temperature programming causes excessive base-line drift for at least one of the columns. The detector output was fed into a Varian digital integrator, which then computed the areas under the various peaks. These areas were punched on cards and put into the computer program which calculated the hydrogen in the product by the iterative procedure already described. It also produced the conversions, rates, and concentrations of the products. The conversions and rates have been calculated in three ways: on the basis of the appearance of carbon compounds in the product, the difference between CO fed and CO in the product, and the appearance of oxygen in the products. The degree of concordance of the three values is a measure of the adequacy of the measured quantities as tested against a mass balance. In all but a few cases these quantities agreed to within about 10%. The average of the three is the basis of the figures and tables given here. The rate, based on CO disappearance, for example, is calculated by the simple equation r = (CO),X(SV) (1) The 2:l H2-CO mixtures were obtained from Matheson; pretreatment was limited to a charcoal filter to remove carbonyls. The two catalysts used are described in Table 111, where their conditions of pretreatment are also given. Product Distribution Figure 2 shows how the H2and CO are consumed during reaction over the two catalysts. Over iron, the usage ratio of H2to CO is only about 0.75, as shown in Figure 3. The system with a 2:l feed ratio behaves as if there were a large excess of hydrogen. Over the cobalt, the usage ratio corresponds more to the ratio of H 2 / C 0 fed, although a t low conversion H2 is used up even faster than this 2:l ratio.
Ivd. Eng. Chern. Process Des Dev , Vol. 18, No. 1, 1979
0 90
I
Table 111. Description of Catalysts
0.70
W
0
$ 050 0
z P
fn
a
030
z V
0IO 010
030
050
070
I .. 0.90
CONVERSION OF CARBON MONOXIDE
Figure 4. Conversion of hydrogen versus conversion of carbon monoxide. 5
b z
0
165
/
035r
CCI Fused Iron Catalyst Iron is fused in the presence of oxygen to form magnetite. Promoters are added t o give the following nominal composition FeO 30-37% FeP, 65-58% Fe < l-'/,% AW, 2-3% 0.4-0.8% KlO SiO,
