2800
Ind. Eng. Chem. Res. 1993, 32, 2800-2811
Development of a Zero-Emissions Sulfur-Recovery Process. 1. Thermochemistry and Reaction Kinetics of Mixtures of H2S and CO2 at High Temperature Gavin P. Towlert and Scott Lynn' Department of Chemical Engineering, University of California at Berkeley, Berkeley, California 94720 When hydrogen sulfide is heated above 600 O C in the presence of carbon dioxide, the conversion of H2S to elemental sulfur is greater than when hydrogen sulfide is heated alone. Formation of elemental sulfur is favored by high temperature, low pressure, and low water content in the gas. The rate-limiting step is the thermal dissociation of H2S. The hydrogen then equilibrates rapidly with C02, forming CO and H2O via the water-gas-shift reaction. The equilibrium of H2S dissociation is therefore shifted t o favor the formation of elemental sulfur. The main byproduct is COS, which is formed by a reaction between C02 and H2S that is analogous to the water-gas-shift reaction. A quench rate of 1000 "C/s or greater is sufficient to prevent loss of elemental sulfur by back-reaction or reaction to COS during cooling. Formation of small amounts of SO2 and CS2 is thermodynamically feasible but has not been observed. Molybdenum disulfide is the best catalyst for H2S dissociation of those discussed in the literature. A process based on this chemistry has significant advantages over the Claus process in that it need not produce any tail gas, it allows recovery of the chemical (or fuel) value of the hydrogen from the H2S, and it requires much less stringent process control.
Introduction The recovery of sulfur from fossil fuels is of major importance in the chemical and energy industries. Sulfur must be removed from fuels in order to comply with legislation designed to prevent environmental damage due to acid precipitation. Sulfur must also be removed from petrochemical feedstocks to prevent degradation of catalysts used in downstream processing. If this sulfur can be recovered in a usable form then its sale will partially offset the cost of its removal from the fossil fuel. The amount of sulfur recovered from fossil fuels accounted for roughly 62 % of all sulfur consumed in the United States in 1991 (Chem. Eng. News, 1992). Most industrial sulfur recovery is carried out using variants of the Claus process, which is based on the partial oxidation of H2S by S02: 2H2S+ SO2 s 3 s + 2H20
(1)
To provide the SO2 for this reaction, a part of the H2S is burned with added air, which leads to the problems described below. The gas mixture then passes through several stages of catalytic conversion, with condensation and removal of product sulfur between stages. Because of the introduction of air to the process, there is a considerable amount of nitrogen flowing through the Claus plant (>60% of the gas stream at any point downstream of the combustor). This inert material must be removed as tail gas; however, after three or four conversion stages the tail gas still contains some sulfur-containing species, typically 2000-3000 ppm of H S plus S02. The tail gas must therefore be sent to a treatment plant to remove these contaminants to an acceptable level. A number of tail-gas processes are reviewed by West (1984). Of these, the process most widely used industrially is the SCOT process; however, a SCOT unit may cost as much as the Claw plant itself (West, 1984). If a sulfur-removal process is run in conjunction witha Claus plant and tail-gas cleanup process, then overall recovery of sulfur may be as high as + Current address: Centre for ProcessIntegration, Department of Chemical Engineering, U.M.I.S.T., P.O.Box 88, Manchester M60 lQD, United Kingdom.
0888-5885/93/2632-2800$04.00/0
99.8% (West, 1984). This technology allows economic recovery of sulfur; however, it is very costly because of the large number of processing steps needed to prevent sulfur emissions, and it does not recover the chemical or fuel value of the hydrogen from the H2S. An alternative route for sulfur recovery suggested by Raymont (1975) was to decompose H2S either thermally or catalytically, and hence recover the hydrogen as well as the sulfur: H2S e H2 + (1/2)S2
(2)
This route received much attention in the 1970's and early 1980's, but proved uneconomic due to the high temperatures required to achieve significant conversion. Attempts to improve the conversion by continuous separation of the products were attempted, but these also proved uneconomic due to the high costs associated with separating H2 from H2S at high temperatures (Fukuda et al., 1978). Several processes were designed using reaction 2 (Banderman and Harder, 1982; Fukuda et al., 1978), but none was developed commercially. The chemistry which led to the present work was observed experimentally while experiments on the desulfurization of coal gas using limestone were carried out. These experiments involved heating limestone in an atmosphere containing 96 mol% C02, 4% H2, and 1% H2S. The reaction-gas mixture entered the outer tube of a reactor that extended into a furnace and was withdrawn through a smaller, concentric inner tube. The difference in diameters between these tubes was such that the residence time in the inner tube was only one-thirtieth the residence time in the outer tube. The gases exiting the reactor were consequently quenched very rapidly by countercurrent heat exchange with the cooler incoming gases. More details of the experimental design are given in Towler (1992). When these experiments were carried out at a temperature of 800 OC, it was observed that a yellow deposit, which was found to be elemental sulfur, formed inside the inner tube in the region where the tubes exited the furnace. The reaction by which sulfur formation occurs was initially postulated as 0 1993 American Chemical Society
Ind. Eng. Chem. Res., Vol. 32, No. 11, 1993 2801 C02 + H2S CO + H20 + (1/2)S2 (3) which can be seen to be the sum of reaction 2, Le., H2S decompoeition, and the reverse of the water-gas-shift reaction: CO + H20 F=? CO, + H, (4) Reaction 3 is similar in form to the Claus reaction in that it achieves partial oxidation of H2S and could therefore form the basis for a potential sulfur-recovery process. Furthermore, the byproduct of reaction 3 is carbon monoxide, which can be reacted with steam to generate hydrogen, thereby effectively recovering the H2 from the H2S. Initial stages in the development of such a process are described below, and a full description of the process is given in the second part of this paper (Towler and Lynn, 1993). However, we cannot claim to be the first to have reported this reaction, as the same chemistry was patented by Bowman (1991) and used as the basis for a process which will be discussed below. We were unaware of Bowman's patent until the work reported here was nearly completed.
Previous Work A great deal has been written on the oxidation of H2S by SO2 as required in the Claus process. There has also been much research into the thermal decomposition of H2S as a source of hydrogen, including some efforts that tried to react the H2S with CO to form COS and H2 (see below). Very little research has been reported addressing the partial oxidation of H2S with C02. As was noted above, North American interest in the thermal decomposition of H2S began with a paper by Raymont (1975), although it had received some attention in Japan prior to that date (Kotera, 1976). The main focus of Raymont's work (and much of that which followed)was on H2S decomposition as a source of hydrogen, rather than as a means of S recovery. Raymont found the decomposition of H2S (reaction 2) to be thermodynamically unfavorable at temperatures below 1800 K. His kinetic studies showed that the reaction proceeds rapidly to equilibrium without requiring catalyst at temperatures above 1250 K, though he did not state the residence time of the reactor in which these experiments were carried out. Below this temperature catalysis is required to achieve a satisfactorily rapid rate; however, Raymont did not identify a particular catalyst. Raymont also correctly observed that the equilibrium conversion of H2S could be enhanced by combining the decomposition reaction with a more favorable reaction. Because of his emphasis on H2 recovery, he chose the reaction 2CO + s2
* 2cos
(5)
giving the overall reaction
+
H2S CO * H2 + COS
(6)
gation of sulfur conversion in flames and found the primary mechanism to be second order. It should be noted, however, that their experimental temperatures were significantly higher than those of interest to this study. Kaloidas and Papayannakos (1989) investigated the noncatalytic thermal decomposition of H2S in the temperature range 60+860 OC and pressure range 1.3-3.0 atm. They developed a kinetic model based upon a free-radical mechanism with the splitting of H2S into free-radical intermediates as the rate-limiting step. The model gave good agreement with their experimental results, and they found the rate of decomposition to be given by = kIPH2S
where k l = 784.1 exp (-23600/T) (mol/(cm3.s-atm))and PH~S is the partial pressure of H2S in atm. They demonstrated clearly that the rate of thermal decomposition was first-order in PH~S under the conditions observed and cited numerous references confirming the presence of H', HS', and S' radicals in gaseous systems containing H2S at high temperatures (Bradley and Dobson, 1967a,b;Levy and Merryman, 1965; Merryman and Levy, 1967, 1972; Norrish and Zeelenberg,1957). They also determined that alumina (A12031 does not catalyze the decomposition reaction. Several catalysts for H2S decomposition have been described; in particular, molybdenum disulfide, MoS2, was identified at an early stage (Katsumoto et al., 1973). Fukuda et al. (1978) described the kinetics of H2S decomposition over molybdenum disulfide and also determined that tungsten disulfide is a slightly less effective catalyst than molybdenum disulfide, whereas NiS is much less effective due to the formation of NiS2. Chivers et al. (1980) performed similar testa and found that Cr2S3 gave similar performance to WS2, while FeS, COS,and a range of copper sulfides (CU~S, CugS5, and CuS) were not effective as catalysts. Chivers et al. also found that MoS2 was the most effective catalyst above 600 "C, but WS2 and Cr2S3 were more effective below 600 "C. Possible mechanisms for H2S decomposition over MoS2 are described by Sugioka and Aomura (1984) and Katsumoto et al. (1973). Chivers and Lau (1985) reported catalytic activity for Li2S but found that sodium and potassium sulfides (Na2S, K2S) and polysulfides (Na2S2, K2S2, Nan&, K2S3, Na&, and K2S4) were not catalytically active; however, these sulfides react with the gas, forming amorphous mixed polysulfideproducts. Chivers and Lau (1987)investigated vanadium sulfide, v2s3, and mixed sulfide systems (v2s3/ FeS, V2S3/CugS5 and V2SdZnS) in both flow and thermaldiffusion-column reactors a t temperatures between 400 and 800 OC. They suggested that v2s3 and V2S3/CugS6 both perform better than MoS2 in flow reactors, with v2s3/ CugS5 being particularly effective at high temperatures. Kotera (1976) investigated the decomposition route via reaction 6, but found this pathway to be complicated by the reaction
The COS could then be converted to S2 via
*
2 c 0 s + so2 2Co2 + (3/2)s2 (7) the SO2 being produced in a manner similar to the Claus process. He suggested that reaction 6 be carried out under conditions that allow continuous separation of Ha from COS, i e . , in a reactor with metal-alloy-membrane walls. Since this technology was not available, attention shifted to continuous removal of sulfur instead. Roth et al. (1982)investigated the mechanism of thermal decomposition of H2S a t low concentrations (1965 K) as part of an investi-
(8)
2cos
* CO, + CS,
(9)
He therefore switched his attention to the thermal decomposition route. Interest in the COS route has recently been revived by Gangwal et al. (1991), who proposed using coal gas to reduce SO2 to elemental sulfur via 2co
+ so2F? 2CO2+ (l/n)S,
(10)
and suggested that coal gas would also react with COS and H2S to form elemental sulfur. The maximum tem-
2802 Ind. Eng. Chem. Res., Vol. 32, No. 11, 1993
perature they investigated was 650"C, and they suggested a process pressure of 20 atm. Their experiments showed that a high conversion of HzS to elemental sulfur could be obtained in a system containing coal gas and SOz;however, they were not aware that the H2S must be undergoing oxidation by COZ as well as SO2 in this system. The pressure selected was too high for formation of elemental sulfur to be strongly favored. Fellmuth et al. (1987) examined the reaction of H2S and C02. They were, however, chiefly interested in zeolite deactivation and did not extend their studies above 300 "C. The main reaction they observed was therefore H2S + CO,
* COS + H,O
(11)
which they found to be catalyzed by basic zeolites. This reaction can be seen to be analogous to the water-gas-shift reaction and is important as a mechanism for COS formation. The equilibriumconstant, KW,for this reaction is fitted over the temperature range 450-1100 "C by the expression
KsBB = 0.4347 exp(-2917/T)
(12)
The endothermic nature of the reaction and weak temperature dependence of the equilibrium constant give further evidence to support the comparison with the watergas-shift reaction, and we might therefore expect this reaction to have equally rapid kinetics. Unfortunately most of the research on this reaction has been related to the development of catalytic processes for COS hydrolysis prior to gas scrubbing and was thus carried out at low temperatures, typically below 300 "C. Although there is much discussion of the competitive absorption of different species and the nature of the catalytic sites, only one paper was found to give an activation energy for the reaction. George (1974) found the activation energy to be 12 kcaU mol for COS hydrolysis on cobalt molybdate at 230 "C. The Arrhenius plot he presented showed, however, that this energy was based on a line drawn through only four data points. The line fitted the three points at lower temperatures reasonably well, but substantially underpredicted the point at the highest temperature. There is therefore good reason to believe that George may have overestimated the activation energy, especially when we consider that we are operating several hundred degrees above the temperatures used in his study. In any case, the value 12 kcal/mol is low for an activation energy, confirming the similarity between this and the water-gasshift reaction and giving us good reason to believe this to be an important step in COS formation. An alternative route by which COS may be formed in the H2S/C02 system is reaction between CO and S2 via reaction 5, which is the reverse of the thermal dissociation of COS. The kinetics of COS decomposition was studied by Schecker and Wagner (1969) in the temperature range 1500-3100 K. They found the rate to be second order and given by dCcos dt --lo
14.2
61000 exp -- RT )'btdccOs
(
(13)
where CCOSis the concentration of COS (mol/cm3),Cbd is the total concentration (ie., COS + inerts) (mol/cm3), and R is the ideal gas constant (cal/(mol K)). The results of Schecker and Wagner (1969) were subsequently conf i i e d by Chenery et al. (1983). Dokiyaet al. (1978)looked at the reverse of COS decomposition, i.e., the formation of COS from CO and elemental sulfur. They did not report a kinetic expression; however, their data showed the
conversion attained to be strongly influenced by flow rate, indicating that this reaction can be quenched. It should be clear that for the HzS/COZsystem at high temperatures we are concerned with the interplay between a large number of possible gas-phasereactions. For process design purposes we wish to be able to predict the reactor size required to achieve a given conversion of H2S. We must also understand the mechanism by which the main byproduct, COS, is formed if we are to design a satisfactory quench. Although most of the reactions in the H2S/C02 system have been analyzed in isolation there is little information in the temperature range of interest, and no report of the kinetics of H2S decomposition in the presence of C02 was found in the literature.
Thermochemical Analysis
A series of calculations was performed to evaluate the equilibrium properties of mixtures of carbon dioxide and hydrogen sulfide at high temperatures. These calculations were used to confirm the experimental findings described above and to determine suitable operating conditions for a sulfur-recovery process based on this chemistry. The calculations were carried out using a modified version of a computer simulation developed by Whitney et al. (1987),whichcalculates the equilibrium composition of a reaction mixture, given a feed stream cornposition, together with a temperature and pressure. The calculation is performed by first selecting a set of independent chemical reactions, then determining material balances for all elements present, and finally determining the composition of the mixture having the minimum Gibbs free energy by using a robust, multidimensional NewtonRaphson iteration procedure. For the high temperatures (>600 "C) and low pressures ( 4 0 bar) of interest to this study, it was reasonableto assume that all gas-phase species behave ideally; therefore fugacity corrections were not made. Thermochemicaldata were taken from the JANAF Thermochemical Tables (Chase et al., 1985). Following the findings of Kaloidas and Papayannakos (19871, S2 was assumed to be the only elemental sulfur speciespresent. In actuality, small amounts of other sulfur allotropes will also exist; therefore the calculations may slightly underpredict the fraction of elemental sulfur in the gas at equilibrium. This is a safe-side approximation that considerably shortens calculation time and obviates the selection of a thermodynamic database for the other elemental sulfur species. The sensitivity of the results to this assumption is discussed below. In analyzing the results of these calculations the most important informationis the distribution of sulfur between the different sulfur-containing species. The results are therefore presented as graphs showing the fractional distribution of sulfur, i.e., the fraction of the total sulfur present as each species on a molar basis. The sulfurcontaining species included in the calculations were H2S, S2,SO2, CS2, COS, and SOS. Results for CS2 and SO8 are not reported, however, as neither ever accounted for more than 0.002 of the total sulfur present. In discussing the results of the survey it is useful to note that since we consider eight chemical compounds (Cop, CO, Ha, H20, H2S, SZ,, 9 0 2 , and COS) containing four elements (C, H, 0, and S), we require four reaction equations to specify the equilibrium composition of the gas phase. For the purposes of this discussion these will be taken as being the water-gas-shift reaction, CO + H20 Z= C02 + H, the H2S decomposition reaction,
(4)
Ind. Eng. Chem. Res., Vol. 32, No. 11, 1993 2803 1.0,
0.10
0.8
I
0.08
B VI
5
0.06
0.6 -
Q VI
c 0.4
In
e
1
I.
2 004
v)
e Q
002
". , nn I 900
"."
d"
0
2
4
6
8
10
89
0.00 12
I
1000
1100
1300
1200
1400
Temperature (K)
Figure 1. Equilibrium distribution of sulfur-containing species aa a function of temperature for H2S only (1atm).
-
0.0 900
/
0 00
1000
1100
1200
1300
1400
Temperature (K)
Figure 2. Effect of temperature on sulfur distribution for a feed containing 50% HzS, 50% COZ at 1 atm.
H2S ~i H2 + (1/2)S,
(2)
the Claus reaction, 2H2S + SO, e (3/2)S2+ 2H20
(1)
and the formation of COS from CO and S2,
2co + s, Fi 2 c o s
(5) It should be noted that these reactions merely form an independent set for equilibrium calculations and are not necessarilyindicative of the reaction mechanism. Reaction mechanism and kinetics are discussed below. Effect of Temperature. Figure 1shows the equilibrium sulfur distribution given by the decomposition of H2S as a function of temperature at 1atm in the absence of C02. Figure 2 shows the distribution for an initial feed of 50% H2S, 50% C02. Comparing the figures, it can be seen that the fraction of S present as elemental sulfur at equilibrium has been enhanced by a factor of roughly 2 a t all temperatures. This significant increase in conversion suggests that a process based on this chemistry would be more successfulthan a process based on the decomposition of H2S in the absence of COa and demonstrates the effectiveness of the water-gas-shift reaction for shifting the equilibrium of H2S decomposition.
The equilibrium yield of S2 increases with increasing temperature because of the endothermic nature of HzS decomposition (reaction 2). Increasing H2S decomposition causes an increase in the amount of HzO present in the gas, which helps drive the reverse of the Claus reaction (1);therefore the equilibrium fraction of SO2also increases with temperature. This imposes an important limit on the operating conditions for the sulfur-recovery process as formation of SO2 is detrimental to process operation and should be avoided whenever possible. The formation of COS is favored by low temperatures, since reaction 5 is exothermic as written. An important consequence of this thermal behavior is that if a mixture of C02 and H2S is brought to equilibrium at high temperature and then cooled slowly, the elemental sulfur is able to back-react to H2S and COS, thereby losing part or all of the reaction yield. This can be prevented by quenching, Le., cooling the gas rapidly to a temperature a t which the sulfurconsuming reactions are very slow compared to processing timescales. Quenchingto about 600"C a t a rate of roughly lo00 K/s was found experimentally to be sufficient. Effect of Pressure. The water-gas-shift reaction (4) is equimolar, and therefore its equilibrium composition is independent of pressure. All of the elemental-sulfurforming reactions lead to a net increase in the number of moles; consequently the amount of S present as elemental sulfur decreases with increasing pressure as can be seen in Figure 3. An interesting, and perhaps unexpected, result is that the equilibrium fraction of SO2 also decreases. If we consider the ratio (number of moles of products)/ (number of moles of reagents) for the S2-formingreactions, we find that for reactions 2 and 5 the ratio is 1.5, whereas for the Claus reaction it is only 1.17. COS formation and H2S decomposition are therefore much more sensitive to pressure than the Claus reaction, and consequently when the pressure is increased S2 is more likely to be converted to H2S or COS via these routes than to SO2 via the Claw reaction. Obviously this is somewhat dependent on the initial composition of the gas (particularly the amount of HzO present) since the water-gas-shift equilibrium is pressure-independent. Effect of the COdH2S Ratio in the Feed. Figure 4 shows the effect of increasing the initial mole fraction of H B in C02 on the equilibrium distribution of sulfur compounds. At very low H2S/C02 ratios, SO2 is the most favored product; however, the fraction of the S present as SO2 falls rapidly as the initial HzS concentration is
2804 Ind. Eng. Chem. Res., Vol. 32, No. 11, 1993
,
0.20
Feed sat w i t h H,O
a t 4OoC
i\\
0.15
Yield of S2 per mole of H,S in Reactor Feed
0 5 i i
04
0.10
0.3
0.05 0.2
0.0
n 50
""
0.1
"."I
0
10
20
30
40
Inltid X H,S In COz
Figure 4. Effect of the initial H&C02 ratio on the equilibrium distribution of sulfur-containing species at 1 atm, 900 OC.
0.0
1
o
io
I
,
,
1
I
/
,
I
l
20
30
40
so
60
70
a0
90
loo
Initial Z H,S in CO,
,E 0
E c
0.6
(900°C, 1 bar)
Figure6. Yield of elemental sulfur as a function of feed concentration (900 OC,1 bar).
I
10.06
~
Table I. Effect of Including Sulfur Allotropes fraction of Sulfur-Containing Species
v)
8
P
species
S2
ss
s 4 s5
S6
increased and the mixture becomes more reducing in nature. The fraction of sulfur present as S2 passes through a maximum at around 6% H2S in C02, since increasing the amount of H2S present not only makes the mixture less oxidizing in nature, but also lowers the S2 fraction by dilution with H2S. A similar effect is observed for COS. It should be noted that the results for COS are plotted on an expanded scale (4X magnification) and that the COS fraction is therefore much less sensitive to the feed composition than the S2 fraction. Effect of H20. For processing reasons, it is likely that any gas stream containing H2S and CO2 would also contain a small amount of water vapor. (This arises from the use of absorber/stripper loops for acid-gas cleaning-the acid gas stream usually leaves the stripper saturated with water at condenser temperature.) It is therefore important to understand the effect of the initial HzO concentration on the equilibrium distribution of sulfur-containing species. Figure 5 shows this variation at 900 OC and 1atm, for a feed containing 1:l C02 and HzS, as the water fraction is increased from 0 to 109%. At l-atm pressure, 10% initial water concentration corresponds to the feed gas being saturated with water at 46 OC. Perhaps the most salient feature of Figure 5 is that none of the species is strongly affected by increasing the H2O fraction over this range.
SI
S2only 0.324 0 0 0 0 0
S2, ...,Se 0.318 0.00649 0.000303
species Se H2S SO2
O.ooOo5
COS
0.000123 0.000107
CS2
fraction of Sulfur-Containing Species S2only S2, ...,SS 0 O.ooOo24 0.614 0.613 0.000811 0.000798 0.0607 0.0604 0.0000087 0.0000087
In particular, the equilibrium fraction of S2 decreases only from 0.32 to 0.29, a change of under 10%. Increasing the water content helps drive the reverse Clam reaction; thus the SO2 fraction increases, though not strongly. The effect of increasing the feed H20 content on the water-gas-shift reaction is to drive the reaction toward C02 and H2, thereby tending to reduce H2S decomposition. This also reduces the CO concentration and hence favors the reverse of reaction 5, leading to COS decomposition and causing the COS fraction to decrease as the H2O fraction increases. The increased COS decomposition to some extent counteracts the reduced H2S decomposition, giving rise to the observed weak dependence of S2 fraction on H20 content. Allowance for Sulfur Allotropes. As was stated above, elemental sulfur can exist as a number of allotropes S2, ...,SSunder the conditions of this study. To confirm the results of Kaloidas and Papayannakos (1987) cited above, the base calculation (1 atm, 900 "C, 5050 H2S: CO2) was repeated with the sulfur speciesS3,...,Sa included. The results are shown in Table I. The fraction of S present as elemental sulfur was 0.3240 in the case where only S2 was considered and 0.3256 in the case where all S allotropes were allowed. This confirms that the effect of including the compounds S3,..., Se is negligible. Reactor Yield. In considering the design of a sulfurrecovery process the optimum reactor feed is determined by several factors, which are somewhat dependent upon the process configuration, and will therefore be discussed in greater detail in the second part of this paper (Towler and Lynn, 1993). To a first approximation, however, we can consider the best feed composition to be that which maximizes the number of moles of elemental sulfur
Ind. Eng. Chem. Res., Vol. 32,No. 11,1993 2806 Vent
T
U U I
hookout TUlk
-H
I
I
I
I
I
AbiorpUon ScNbbclr
N.08
1 I
L
Pipellno
I
v
11
I
u u u u u
C02 NP E2S E2
CO
8
SRI 8610 Gal Chromatoirsph
Sulfur
Condemer
Figure 7. Experimental apparatus. Inner tube o.d, 0 3 m m td 3.9 mm
I
C u Inlet
II II
I
h
. I
cu
outlet
I
QwtrT b r m o w l l a.d 7 mm
Outer Tuba 1.d. 21.8 mm
T
Cu .
Sampling/
smep G u Port
.
kction
Figure 8. Reactor dimensions.
produced per mole fed to the reactor, i.e., the yield per reactor pass. We can determine the yield at chemical equilibrium quite easily, by multiplying the fraction of S present as elemental sulfur by the fraction of HzS in the feed. Figure 6 shows the reactor yield at 900 "C, 1 bar, as a function of the concentration of HzS in the feed (drybasis). The solid lines in Figure 6 are for a dry feed, in which case the maximum yield is 0.167 at an initial concentration of 67% HzS in C02. This represents an improvement of 36% over the yield in the absence of C02. This could raise the question of whether the presence of C02 is really necessary to the process; however, it must be remembered that the water-gas-shift reaction (41, by significantly lowering the hydrogen concentration, also minimizes back-reaction during the quench. It is therefore much easier to maintain a high yield in the presence of C02 than in its absence. As noted above, the feed will normally contain some H2O; therefore the calculations were repeated for a feed saturated with water at 40 "C, 1 atm. The results are shown as the dashed lines in Figure 6. In this case the maximum yield of elemental sulfur per mole of feed gas is 0.1452 a t a feed fraction 71% H2S in C02. This is somewhat lower than the dry case, mainly due to dilution: if we correct the results to a dry basis, the yield is 0.157, Le., an improvementof 27.7% over the yield in the absence
of C02. As noted above, the optimum reactor feed composition may depend upon other factors as well as reactor yield, and this will be discussed in the second part of this paper (Towler and Lynn, 1993). It is, however, important to note that the maximum in the yield curve is in fact rather broad and that high yield can therefore be achieved over a wide range of feed compositions. This has some important implications (discussed below) when a sulfurrecovery process based on this chemistryis compared with a process based on the Claus reaction.
Experimental Section Experiments were carried out in the continuous-flow apparatus shown in Figure 7. The C02 and H2S (industrial grade, supplied by Mattheson Gas Products, Newark, CA) were mixed at the desired flowrates using high-precision, low-flow-rate rotameters (Omega Engineering Inc., Stamford, CT), before being sent to the reactor. The reactor consisted of two concentric quartz tubes mounted in a high-temperaturefurnace. The gas entered the outer tube of the reactor, where it was preheated by contact with the exiting gas and by heating tape around the outside of the reactor. The gas was then further heated in the furnace section of the reactor before exiting through the inner tube. The difference in the tube diameters caused the
2806 Ind. Eng. Chem. Res., Vol. 32, No. 11, 1993
h
900
1000
800
900
700
800
U 0-
0-
u
2
0
3e
e
600
i
c
i
located here
F
500
600
0 Thermowell temperature V Inner tube temperature 0 Thermowell temperature (repeat)
400
300 -5
700
D
Inner tube entrance
,
I
f
I
,
I
,
0
5
10
15
20
25
30
500
/ 400
35
0
1
Distanoe Along Reactor
OC.
Knowledge of the temperature-distance profile and the gas flow rate allows calculation of the reactor temperature time profile, i.e., the temperature history experienced by the gas as it passed through the reador. We can also calculate the quench rate from the highest temperature attained to any lower temperature. Such calculations were incorporated into the kinetic models described below. A typical temperature-time profile is shown in Figure 10. To investigate the effect of quench rate, some experiments were run with a 3.06-mm quartz rod inserted into the inner tube. This reduced the flow area and increased the gas velocity by a factor of roughly 2.6, allowing the quench rate to be increased by roughly the same factor.
3
4
Time
Figure 9. Reactor temperature profile (T-t = 800 "C, flow rate = 2.935 mL/s).
residence time in the inner tube to be roughly 4% of that in the outer tube, which, together with the temperature profile obtained in the reactor and judicious selection of flow rates, ensured that the gas spent long enough at high temperature to guarantee significant conversion of H2S, before being cooled rapidly as it left the furnace. The physical dimensions of the reactor are given in Figure 8. These dimensions are of great importance in the modeling of the reactor; each dimension is the average of several measurements (taken at different points of the reactor where possible). The gas leaving the reactor passed through a condenser that collected more than 99% of the elemental sulfur formed during the reaction. The gas was then sampled using an SRI 8610 Gas Chromatograph (SRI Instruments, Torrance, CA), which was used to detect for C02, CO, COS, H2S, and S02. This, together with knowledge of the feed rates, allowed the composition of the gas leaving the reactor to be determined. The temperature profile of the reactor was measured for different heater set-points and gas flow rates. The profile was found to vary slightly with flow rate; therefore three standard flow rates were selected and the temperature profile was measured for each. A typical temperature profile for the reactor is shown in Figure 9. The temperature profile in the inner tube was found by removing the condenser and inserting a thermocoupleinto the tube. The temperature in the outer tube was measured by a thermocouple in the reactor thermowell and a slight correction was made to allow for radiative effects. The uncertainty in the temperatures measured was roughly 3
2
5
6
7
0
(8)
Figure 10. Typical Reactor Temperature-Time Ronie (T- = 900 T,F = 2.935 mlis)
Figure 10. %icd reactor temperaturetime profiie (T,t = 900 "c, flow rate = 2.935 mL/s). Table 11. Exwrimental Conditions and Results temp (OC) 950 950 950 950 950 950 950 950 900
900 900 900
flowrate (d/s) 2.935 2.935 2.935 2.935 1.468 1.468 1.468 1.468 2.935 2.935 2.935 1.468
feedconcn ( % HzS in COz) 5 5 10 10 5 5 10 10 5 10 10 5
conv (%)
34.4 35.2 34.0 34.4 54.3 53.1 43.8 44.3 21.5 20.0 22.2 36.1
outlet COSmole fraction x 103 4.05 4.43 2.82 3.99 3.57 4.15 4.79 5.16 3.86 3.36 2.99 3.32
Rasults Because we are concerned with t L e interp.dy between several reactions over a wide range of temperature and composition,where substantial changes in the equilibrium behavior of the system can occur, the results are best presented in tabular, rather than graphical, form. The conversion of HzS achieved for various experimental conditions is given in Table 11. Comparingthe experiments at high flow rate (where we are furthest from equilibrium conversion),we can see that the fractional conversion obtained with a feed of 10% H2S in C02 is roughly the same as that obtained with 5% H2S in C02. This behavior would be expected if the ratelimiting step were the first-order thermal decomposition of HzS. At low H2S feed concentrations the water-gasshift reaction lowers the hydrogen concentration to such an extent that back-reactionby recombination of hydrogen and elemental sulfur is negligible. The overall reaction would thus behave as a first-order irreversible reaction, for which the conversion in a plug-flow reactor is independent of the feed concentration. We see that this limit is obtained at both set-point temperatures (but note, however, that the gas does not experience a single temperature, but a range or profile, as indicated in Figure 9). Obviously as the conversion approaches its equilibrium value this approximation can no longer be true, and we see that the experimentsrun at high temperature and low flow rate do not satisfy this condition as equilibrium effects are more
Ind. Eng. Chem. Res., Vol. 32,No. 11,1993 2807 Table 111. Outlet Compositions Measured for Quenched and Nonquenched Experiments. feed H2S quench fraction rate (%) (K/d 5 990 5 2560 10 1040 10 2700
C02 0.928 0.930 0.873 0.872
outlet composition Hz and H2S CO H2O COS 0.0218 0.00973 0.0260-000357 0.0218 0.00884 0.0248 0.00415 0.0524 0.0108 0.0408 0.00479 0.0519 0.0113 0.0412 0.00516
If this is the mechanism of H2S conversion, then the rate of change of the H2S partial pressure is given by d@H2S)/dt = -klPH# + k&-@S:'2
S2
0.0112 0.0103 0.0180 0.0180
In all cases the flow rate was 1.468 mL/s and the maximum temperature in the profile was 950 "C.
where kl should be the first-order rate constant as found by Kaloidas and Papayannakos (1989) and other symbols are as defined in the Nomenclature section. If we use suitable average values for the quantities in the second term on the right-hand side of eq 14 and integrate, we obtain a "partially integrated" algorithm:
(I
important for these runs. The error in the reported conversions can be assessed from the difference between repeated runs and is typically of the order of 3% ,though it may be as high as 11%for the experiments with the lowest conversion. The outlet compositions found for the high-quench experiments (in which a quartz rod was inserted into the reactor exit tube) and experiments run under the same conditions with lower quench rates are shown in Table 111. Table I11 shows that the outlet compositions for the high-quench experiments are practically identical to those for the corresponding low-quench experiments; i.e., the increase in quench rate gives no added conversion. (The apparent slight increase in COS formation for both quenched experiments is negligible compared with the error in the COS measurements, but see below.) This confirms that the quench rates used in the bulk of the experiments (>lo00 "C/s) were high enough to quench completely both the H2S decomposition back-reaction and the COS-forming reactions. These results are also useful in analyzing the formation of COS, as discussed below.
Kinetic Model In developing a model for the kinetic behavior of the H2S/C02 system at high temperature, our main goal is to be able to predict the conversion of H2S to elemental sulfur, i.e.,to identify the rate-limiting step(s) in H2S conversion. A secondary goal is to predict the formation of carbonyl sulfide, so that reactor conditions can be arranged to minimize formation of this byproduct. Several models were tried; however, in the interest of brevity we shall concentrate upon the model that gave the best agreement with the data. The reactor is assumed to be in plug flow and is divided into sections corresponding to the interkls at which the temperature was measured. For each interval the average temperature can be found, and hence, knowing the feed flow rate and pressure, the time the gas spends in the interval may be calculated (the flow rate can also be corrected to allow for the increase in molar flow rate that occurs on reaction). Given the residence time and temperature for each interval we can use a suitable algorithm to find the concentration of each species a t the end of the interval. By treating the intervals as a set of plug-flow reactors in series, we eventually obtain a prediction of the final outlet concentration. The pseudo-first-order-irreversible behavior observed a t high flow rates suggested that the rate-limiting step of H2S conversion is the thermal decomposition of H2S, with the hydrogen thereby produced reacting with C02 via the water-gas-shift reaction. The water-gas-shift reaction is known to equilibrate very rapidly at temperatures above 600 "C and may therefore be assumed to be always a t equilibrium.
(14)
PH2S,n
-- PH,S,n-l
exp(-kl,nAtn) +
which may be used to update the H2S partial pressure from one interval to the next. The first term in this equation allows for decomposition with no reverse reaction, and the second term is a correction to allow for the reverse reaction. It can be seen that in the limit as n and At,, tend to infinity at a constant temperature, the first term goes to zero while the second term approaches the equilibrium concentration. Having found the number of moles of H2 formed from the H2S decomposition and knowing the number of moles of CO2 fed, the concentrations of Con, CO, H2, and H2O are then found by solving the water-gas-shift reaction equilibrium equation and three mass balances (C, H, and 01,and the concentration of elemental sulfur is found from a sulfur mass balance. This basic model gave good agreement with the experimental data using the published kinetic parameters of Kaloidas and Papayannakos (1989);however, it is unsatisfactory in that it does not describe the formation of the byproducts COS and SO2. Sulfur dioxide formation was accounted for by assuming that the Claus reaction was always at equilibrium. This is a reasonable approximation above 700 "C, but in practice this did not significantly alter the predictions of the model as SO2 was formed at very low levels under the conditions studied. Sulfur dioxide was not detected in any of the noncatalytic experiments; therefore the validity of this assumption could not be tested and remains a subject for future investigation. The formation of COS is not so simple to describe. There are two likely mechanisms for COS formation, namely reactions 5 and 11. The kinetics of these reactions was described above, though it should be noted that the work of George (1974) was carried out under conditions somewhat removed from those of this study. Since most of the formation of CO and S2 occurs a t the highest temperatures in the profile, we would expect reaction 5 to be significant as a COS-forming mechanism only at the highest temperatures or during cooling, whereas reaction 11 would be a reasonable mechanism at lower temperatures when substantial amounts of CO and S2 have not yet formed. (Indeed, reaction 11plus the reverse of reaction 5 would be a second-order alternative mechanism for H2S conversion.) In practice, both reactions are important, as will be shown by considering the experimental results. The COS concentrations detected were greater than the concentrations that would have been found if COS had been in equilibrium with the other species at the maximum temperature in the profile. As was shown above, COS formation is favored by lower temperatures. These high COS concentrations must therefore have been formed either as the gas entered the reactor or else as it cooled
2808 Ind. Eng. Chem. Res., Vol. 32, No. 11, 1993 Table IV. Comparison of Experiments and Model Predictions of the Conversion of HsS
exptl conv 34.4 35.2 34.0 34.4 54.3 53.1 43.8 44.3 21.5 20.0 22.2 36.1
CO and S2 in Equilibrium
Dredictions of model with various assumptions 5% variation E,& from 5% variation of E.& for of E& for published reactions 1 and 8 reaction 1 data 34.2 34.6 35.1 34.5 34.8 35.4 33.8 33.5 34.3 33.8 33.5 34.3 57.9 57.5 58.6 57.7 58.1 58.8 54.3 54.6 55.2 54.3 54.6 55.2 20.0 19.8 19.6 19.2 19.2 19.0 19.3 19.3 19.0 37.4 36.9 37.3
upon leaving. If COS were formed during cooling (via reaction 5), then the outlet concentration would show a dependence upon the quench rate, i.e.,the residence time for cooling. It can be seen from Table I11that this is clearly not the case; in fact, the COS outlet concentrations may even be higher at higher quench rates. Alternatively, COS may be formed via reaction (11)when the gas first enters the furnace. As the gas flows to regions of higher temperature, a point would be reached a t which the COS concentration is greater than its equilibrium value, at which point COS would begin to dissociate via the reverse of reaction 5 (thoughnot necessarilyvia the reverse of reaction 11 since COZ and HzS could still be present at greaterthan-equilibrium concentrations). By this hypothesis the outlet COS fraction would not be greatly affected by the quench rate (as long as this is rapid enough to prevent significant reaction when the gas reaches lower temperatures) and might even increase at higher quench rates as the time for COS decomposition a t high temperature is reduced. This mechanism not only gives a better explanation of the data, but is also more physically reasonable. We know that reaction 11 is somewhat similar to the water-gasshift reaction, (4), (see above), and, given that it has such alow activation energy,it would be unreasonable to neglect its effect at temperatures above 800 "C. Both reactions 5 and 11were therefore incorporated into the model, again using published kinetic parameters, and using an algorithm similar to eq 15 to upgrade the COS concentration from one interval to the next. Equation 15 was also modified to account for the presence of reaction 11. Mathematical details of the algorithm used are given in Towler (1992). Table IV shows the experimental results and model predictions obtained using the published values of the activation energies (Kaloidas and Papayannakos (1989) for HzS decomposition, and George (1974) for reaction 11). The preexponential constants used were adjusted to fit the data over the temperature range observed and were 6.0 X lo6 s-1 for HzS decomposition and 2.4 X lo6 cm3 mole-1 s-l for reaction 11. The root-mean-square error between the experimental conversions and those predicted by the model is about 12%, and the error between the predicted and observed COS concentrations is roughly the same as the error in the COS measurements. The predicted COS concentration for a typical experiment is plotted against temperature in Figure 11, from which we see that the model gives the right qualitative behavior. Figure 11also shows the concentration of COS that would be found if COS and SZwere in equilibrium and the concentration that would be present if COS were only formed from CO and SZas predicted by the kinetics
-2
-
-4
-
-6
-
-10 -a
--
-12
-
-14
-
4 Scheckcr and Wagner
Entrance t o Quench
Tube
0
Measured Concentration
-16 1
0
20
10
30
40
50
Distance From Reactor Entry (cm)
Figure 11. Variation of COS concentration withd i c e into reactor (maximum temperature 950 "C, flow rate 1.468 mL/s, 5% H2S in
Cod.
of Schecker and Wagner (1969). Clearly neither of these gives an adequate description of COS formation. We can improve the model predictions somewhat if we allow constrained variation of the activation energies for reactions 2 and 11. For HzS decomposition this is not unreasonable, since the experiments of Kaloidas and Papayannakos (1989) were performed outside our temperature range and must necessarily contain some experimental error; furthermore, Fukuda et al. (1978) found an activation energy of 42 kcal/mol, which is somewhat lower than the 43.7 kcal/mol found by Kaloidas and Papayannakos. If we use an activation energy 5 % lower than that of Kaloidas and Papayannakos (41.5 kcal/mol), we obtain the third column of Table IV. The preexponential constant in this case is 2.4 X 10 s-l. The root-mean-square error is reduced to 11.3%,which is of the order of experimental error. The COS concentration predictions are not affected by this change, as COS formation occurs in the lowtemperature region of the entry tube, Le., before the HzS conversion is significant. We can, however, improve the prediction of COS concentration somewhat by allowing similar variation of the activation energy of reaction 11. Again this is not unreasonable, due to the rather dubious nature of the graph presented by George (1974). We should also note that although George was studying the reaction in the presence of catalyst, his experiments were carried out 600 "C below ours, so extrapolation may not be justified. If the activation energy for COS hydrolysis is taken as 8 kcal/mol, we obtain the results of the fourth column of Table IV, which predict both conversion and COS formation to within experimental error (for details of the COS predictions see Towler (1992)). Obviously the model predictions could be further improved by including more reactions in the model; however, this merely increases the number of adjustable parameters available and is therefore rejected. The model thus obtained gives an adequate description of the experimental data over the range of conditions studied and can be used for reactor design. Some limitations of this model will be discussed below.
Effect of Catalysis Having identified the rate-limiting reaction for the HzSl COZsystem to be the thermal dissociationof HzS, it follows
Ind. Eng. Chem. Res., Vol. 32, No. 11,1993 2809 that one would improve the overall reactor performance by supplying a catalyst for this reaction. This is of great importance in the development of a process based on this chemistry, as use of catalysis considerably reduces the reactor volume required. Even with catalysis the reactor is likely to be the most expensivecapital item in the process designed (Towler, 1992). Catalysts for the thermal decomposition of H2S were discussed above; in particular, molybdenum disulfide (MoS2) is well-known as a catalyst for H2S decompositionat high temperatures. A short series of experiments was therefore performed to demonstrate that MoSz would act as a catalyst for the H2S/C02 system. It should be stressed that this was only a preliminary investigation to demonstrate feasibility and that a full analysis of the catalysis of this system would be a major piece of work in its own right and was beyond the scope of this study. The apparatus described above was modified by packing the reactor with MoS2 catalyst supported on quartz wool. Initially, experiments were performed with the reactor packed with quartz wool in the absence of catalyst; the quartz wool alone was determined to have a negligible effect on reaction kinetics. The experiments were then repeated with the same mass of quartz wool onto which MoS2 powder had been loaded. The conversion increased significantly at all temperatures studied (up to 900 "C), and no evidence of catalyst deactivation was observed over 52 h of operation. This does not eliminate the possibility that catalyst deactivation is occurring;however, it suggests that if it occurs at all then it is a rather slow process. In several cases the distribution of sulfur-containing species was close to that expected for chemical equilibrium at the highest temperature in the profile, with the exception that the fraction of SO2 was lower than expected. It is not clear whether SO2 was not formed at high temperature or whether it was lost by back-reaction during cooling. The experimental conditions were not sufficiently well defied to allow detailed modeling of the experiments with catalyst (Towler, 1992). It would be possible to extend the model described above by using the kinetic expression developed for H2S thermal decomposition over MoS2 by Fukuda et al. (19781,but this would not account for the effect of catalysison reaction 11 and hence COS formation. Since the catalytic mechanism is known to involve formation of H*, HS*, and S* free radicals (Katsumoto et al., 1973),it is unreasonable to expect the catalyst to affect only the H2S-decomposition reaction. We may however, use the kinetics of Fukuda et al. (1978)as a firstorder approximation for process design if we assume the gas reaches the equilibrium concentration (a reasonable assumption in the light of the experimental results) and the catalyst loading is the same as that used experimentally by Fukuda et al. For a fundamental understanding of the catalytic behavior a more complete analysis is necessary, which should include not only the effect of the catalyst on the other reactions in the system, but also the preparation of catalysts and optimization of properties such as surface loading, etc. This was beyond the scope of this work; however, the beneficial effects of catalysis on the H2S/ COz system were clearly demonstrated. More details of this work are given in Towler (1992). Discussion
The goal of this study was to develop an understanding of the kinetics that could be used to design a reactor for the Zero-Emissions Sulfur Process discussed in part 2 of this series (Towler and Lynn, 1993). For this purpose the model developed above is adequate, since it describes both
H2S conversion and the formation of the main reaction byproduct, COS. On a more fundamental level, however, the model is somewhat less satisfactory. Firstly, undoubtedly the greatest source of error in the model is also, unfortunately, the hardest to quantify. The model assumed the reactor to consist of two plug-flow reactors in series, each with a different residence time, corresponding to the inner and outer flow volumes. In practice the Reynolds numbers used experimentally were 2 or 4 for the outer tube and 48 or 96 for the inner tube, so the flow was laminar in all cases. For a first-order isothermal reaction the conversion in a laminar-flow tubular reactor is somewhat less than that in a plug-flow reactor, mainly due to bypassing by material that flows through the reactor at velocities faster than the average velocity. If the velocity profile is known (e.g., for a simple tube), the conversion can be found mathematically (this involves use of the exponential integral function for a firstorder isothermal reaction). In our case, however, the outer tube contains two nonconcentric,nonparallelsmaller tubes (the inner tube and thermowell) and has a marked longitudinal (and possibly also a slight radial) temperature gradient,all of which make evaluation of the velocity profile a complex computational problem in fluid dynamics. Furthermore, the high molecular diffusivities at the high temperatures obtaining in the reactor cause some radial mixing and tend to offset the dispersion of residence times caused by the velocity profile. Exact characterization of the laminar-flowreactor is therefore difficult and the plugflow assumption was thus necessary, but this may impose a large error on the model predictions. Experiments that were run with the outer tube entirely packed with quartz wool (as preparation for the catalysis experiments) gave somewhat higher conversionsthan experiments performed in the absence of this packing. It was not clear, however, whether the quartz wool was promoting radial mixing and thus bringing the reactor closer to plug flow or whether channelingeffects (the quartz wool could not be distributed evenly) were causing changes in the temperature profile experienced by the gas. Secondly, at the high temperatures in question it is not reasonable to suppose that the net rate of reaction is affected only by the collisions of molecular species. In practice there are likely to be many short-lived free radical species present at low concentrations (for example, H*, HS*, S*,OH*, SO*, etc.) which probably play an integral part in the reaction mechanism(s)via a series of initiation, chain propagation, and termination reactions, much the same as is found in the analysis of combustion systems. The presence of these species was referred to by Kaloidas and Papayannakos (1989)and earlier workers, as noted above. The success of the experiments with a catalyst that is known to promote the formation of HS* radicals is also good evidence of free-radical initiation as the ratelimiting step; however, HO* radicals are also known to be of great importance in the water-gas-shiftreaction kinetics and we did not feel confident enough in the experimental resulta (for reasons discussed above) to extend our analysis to proposing a free-radical mechanism. Our analysis was thus limited to molecular species. A more complete analysis of the kinetics would take account of free-radical effects, and also of other minor reactions and byproducts. This would require rather more accurate data than are presented here, in particular, the flow pattern in the reactor would have to be more accurately described. Such an analysis must therefore be the subject of future work, and a suitable concluding note is that the engineering model
2810 Ind. Eng. Chem. Res., Vol. 32, No. 11, 1993
developed above will be useful not only in process design, but also in designingan apparatus for more detailed kinetic studies.
Process Synthesis One can easily envisiona processbased on this chemistry, in which a stream containing H2S and C02 in the correct proportions is produced by absorbing these species from a suitable sour gas. This mixture is sent to a furnace, where it reacts forming S2, CO, HzO, H2, and small amounts of COS, SOZ,and CS2. The sulfur is condensed and the remaining acid gases are absorbed in a second absorber, leaving a product gas consisting chiefly of Hz and CO. The acid gases can than be stripped from the solvent in a stripper and recycled to the reactor. A detailed description of the process, operating conditions, and flowsheet will be presented in part 2 (Towler and Lynn, 1993). There are, however, some advantages to a process based on this chemistry that were apparent from an early stage, including some that arise from the thermochemical considerations discussedabove. The most important advantage is that by using CO2 instead of air as oxidant the tail gas problem is eliminated. The concentration of sulfur-containing species in the gas streams leaving the process (the sweetened gas and the product gas) is therefore determined by the operating conditions of the two absorber/stripper loops, and may be controlled to a very low value (a few ppm) depending on how lean the solvent is stripped in each stripper. Note that the process is strictly "zero-emissions" in the sense that there is no tail gas. Effluent streams of course have some sulfur content, but this may be controlled to as low a level as can be achieved using absorber/stripper technology. Secondly, the product gas leavingthe process is a mixture of CO and HZsaturated with water vapor a t the absorber temperature. This gas may be used as fuel or synthesis gas; in the latter case it may be mixed with steam and sent to a shift reactor for conversion to hydrogen. The choice between these uses depends on economic considerations that may be expected to vary from site to site; however, the value of the hydrogen from the H2S is recovered in all cases. Further advantages arise from the breadth of the maximum in the yield curve (Figure 6). The Claus process requiresvery tight control of the air-to-HzS ratio to achieve an exact stoichiometric ratio of 2 H2S to 1 SO2 in the reactor sequence. If this ratio is upset slightly, then the unconverted HzS or SO2 is passed on to the tail-gas unit, creating considerable difficulties. For a process based on the interaction between COZ and H2S, the COz-to-HaS ratio is not critical owing to the rather broad maximum in the reactor yield. The Zero-Emissions Sulfur Process is therefore much more robust and requires much less sophisticated control. A similar problem is caused for the Claus process if the incoming H2S stream is contaminated by hydrocarbon material. This is often the case if lowtemperature, high-pressure absorption is used, and is caused by condensation of organicsin the absorber solution and subsequent vaporization in the stripper. If hydrocarbon material is present in the Claus process feed, it consumes some of the oxygen from the air feed, making control of the H2S-to402 ratio more difficult. Fluctuations in the amount of hydrocarbon material present obviously exacerbate this problem. For the Zero-Emissions Sulfur Process any organic material sent to the furnace will react with COz, forming carbon monoxide and hydrogen, and will end up as part of the product gas. This
does not significantly affect the Zero-Emissions Sulfur Process (except to slightly improve the quality of the product gas). Other advantages of the new technology compared to the Claus process will be discussed in part 2 (Towler and Lynn, 1993).
Acknowledgment
This research was funded by the Morgantown Energy TechnologyCenter through the U.S.Department of Energy under Contract DE-AC03-76SF00098. Nomenclature
Ci = concentration of species i (mol/cm3) Cbd = total concentration of the gas phase (mol/cm3) Eact= activation energy (cal/(mol-K)) kl = first-order, noncatalytic, H2S thermal decomposition rate constant (mol/(cm3.s.atm)) kz = rate constant for noncatalytic thermal-recombination . ~atm-1) reaction between Hz and S2 (moF5 ~ m - ls-1 KWs = equilibrium constant for the "sulfur-gas-shift"reaction, reaction 8 n = (subscript) property evaluated under the conditions of the nth interval P = total pressure (atm) Pi = partial pressure of species i (atm) r = rate of H2S thermal decomposition (mol/(cm3-s)) R = ideal gas constant (cal/(mol.K)) t = time (a) At = residence time (s) T = temperature (K) Tset= furnace set-point temperature ("C) AT = temperature difference (K) * = denotes that the species is a free radical
Literature Cited Bandermann, F.; Harder, K. B. Production of Hz via Thermal Decomposition of H2S and Separationof Hz and H2S by PressureSwing Adsorption.Znt. J. Hydrogen Energy 1982,7 (6), 471-475. Bowman, M. G. Thermochemical Cycle for Splitting Hydrogen Sulfide. U.S. Patent 4,999,178, March 12, 1991. Bradley,J. N.;Dobson, D. C. Oxidationof Hydrogen Sulfidein Shock Waves, I. Absorption Studies of OH and SO2 in H&S-02-Ar Mixtures. J . Chem. Phys. 1967a, 46 (8), 2865. Bradley, J. N.;Dobson, D. C. Oxidationof Hydrogen Sulfidein Shock Waves, 11. The Effect of Added Hydrogen on the Absorption of OH and SOz. J. Chem. Phys. 1967b, 46 (8), 2872. Chem. Eng. News 1992, June 29, 37. Chase, M. W., Jr.; Davies, C. A.; Downey, J. R., Jr.; Frurip, D. J.; McDonald,R. A,;Syvemd,A. N. JAIVAF Thermochemical Tables, 3rd Ed.;National Bureau of Standards: Washington, DC, 1985; Vol. 14. Chenery,J. A.;Fakhr, A.; Wood,M. I.; Simpson,C. J. S. M. Vibrational Analysisof the Products fromthe ThermalDecompositionof OCS and COz. Chem. Phys. Lett. 1983, 96(2), 143-147. Chivers,T.;Lau, C. TheThermalDecompositionof HydrogenSulfide Over Alkali Metal Sulfides and Polysulfides. Znt. J . Hydrogen Energy 1986, 10 (l),21-25. Chivers,T.;Lau, C. The Thermal Decompositionof HydrogenSulfide Over Vanadium and Molybdenum Sulfides and Mixed-Sulfide Catalyst in Quartz and Thermal-DiffusionColumn Reactors.Znt. J. Hydrogen Energy 1987,12 (4), 235-243. Chivers, T.; Hyne, J. B.; Lau, C. The Thermal Decomposition of HydrogenSulfide Over TransitionMetal Sulfides.Int. J. Hydrogen Energy 1980,5 (5), 499-506. Dokiya, M.; Fukuda, K.; Yokokawa, H.; Kameyama, T. The Study of Thermochemical Hydrogen Preparation. VI. A Hydrogenevolving Step through the H2S-CO Cycle. Bull. Chem. SOC.Jpn. 1978,51 (l), 150-153. Fellmuth,P.;Lutz, W.;Biilow, M. Influenceof Weakly Co-ordinated Cations and Basic Sites Upon the Reaction of H2S and COZon Zeolites. Zeofites 1987, 7 (4), 367-371.
Ind. Eng. Chem. Res., Vol. 32, No.11,1993 2811 Fukuda, K.; Dokiya, M.; Kameyama, T.; Kotera, Y. Catalytic Decomposition of Hydrogen Sulfide. Znd. Eng. Chem. Fundam. 1978,17(4),243-248. Gangwal, S.K.;McMichael, W. J.;Dorchak, T. P. The Direct SulfurRecovery Process. Environ. h o g . 1991,lO (3,186-191. George, Z.M. Kinetics of Cobalt-Molybdate-Catalyzed Reactions of SO2 with H2S and COS and the Hydrolysis of COS. J . Catal. 1974, 32,261-271. Kaloidas, V.; Papayannakos, N. Hydrogen Production from the Decomposition of Hydrogen Sulfide. Equilibrium Studies on the System HzS/H2/Si (i = 1,...,8)in the Gas Phase. Znt. J. Hydrogen Energy 1987,12(6),403-409. Kaloidas, V.;Papayannakos, N. Kinetics of Thermal, Non-catalytic Decomposition of Hydrogen Sulfide. Chem. Eng. Sci. 1989, 44 (ll), 2493-2500. Katsumoto, M.; Fueki, K.; Mukaibo, T. An Investigation of the GasSolid Interface Reaction. Bull. Chem. SOC.Jpn. 1973,46,36413644. Kotera, Y. The Thermochemical Hydrogen Program at N.C.L.I. Znt. J. Hydrogen Energy 1976,1,219-220. Levy, A.; Merryman, E. L. The Microstructure of Hydrogen Sulfide Flames. Combust. Flame 1965,9,229. Merryman, E. L.; Levy, A. Kinetics of Sulfur Oxide Formation in Flames: I1Low Pressure HzS Flames. J. Air Pollut. Control 1967, 17,800. Merryman, E. L.; Levy, A. Disulfur and the Lower Oxides of Sulfur in Hydrogen Sulfide Flames. J. Phys. Chem. 1972,76(14),1925. Norrish, R. G. W.; Zeelenberg, A. P. The Combustion of Hydrogen Sulfide Studied by Flash Photolysis and Kinetic Spectroscopy. R o c . R. SOC.Ser. A 1957,240,293.
Raymont, M. E. D. Make Hydrogen from Hydrogen Sulfide. Hydrocarbon Process. 1975,54 (7), 139-142. Roth,P.; L&r, R.; Barner, V. Thermal Decomposition of Hydrogen Sulfide at Low Concentrations. Combust. Flame 1982,45(31,273286. Schecker, H. G.; Wagner, H. G. On the Thermal Decomposition of COS. Int. J. Chem. Kinet. 1969,1, 541-549. Sugioka, M.; Aomura, K. A Possible Mechanism for Catalytic Decomposition of Hydrogen Sulfide over Molybdenum DisuKde. Znt. J . Hydrogen Energy 1984,9(111, 891-894. Towler, G. P. Synthesis and Development of Processes for the Recoveryof Sulfur From Acid Gases.Ph.D. Dissertation, University of California, Berkeley, 1992. Towler, G. P.; Lynn, S. Development of a Zero-Emissions SulfurRecovery Process. 2. A Sulfur-Recovery Process Based on the Reactions of HzS and COz at High Temperature. Znd. Eng. Chem. Res. 1993,following paper in this issue. West, J. R. Sulfur Recovery. In Kirk-Othmer Encyclopedia of Chemical Technology; Grayson, M., Ed.; Wiley: New York, 19W, Vol. 22,pp 267-297. Whitney, G. M.; Bang, Y.; Denn, M. M.; Petersen, E. E. Sulfur Capture During Partial Coal Combustion. Chem. Eng. Commun. 1987,55(1-6),83-93.
Received for review January 11, 1993 Revised manuscript received June 8, 1993 Accepted June 24, 1993. Abstract published in Advance ACS Abstracts, September 1, 1993.
