I n d . E n g . C h e m . R e s . 1989,28,931-940 Massaldi, H. A.; Maymo, J. A. Error In Handling Finite Conversion Reactor Data by the Differential Method. J. Catal. 1968, 14, 61-68. McMillan, D. Process for Reduction of SOz with Hydrocarbon Vapor. US Patent 3615221, 1971. Mulligan, D. J. Reduction of Sulphur Dioxide Over Transition Metal Sulphides. M. Eng. Thesis, McGill University, Montreal, Canada, 1988.
931
Rosenberg, H. L.; Engdahl, R. B.; Oxley, J. H.; Genco, J. H. The Status of SOz Control Systems. Chem. Eng. h o g . 1975, 71, 66. Sarlis, J.; Berk, D. Reduction of Sulfur Dioxide with Methane over Activated Alumina. Ind. Eng. Chem. Res. 1988, 27, 1951-54.
Received f o r review August 5, 1988 Revised manuscript received January 23, 1989 Accepted February 19, 1989
High-Temperature Sulfidation-Regeneration of Cu0-A1203 Sorbents V. Patrickt and G. R. Gavalas* Department o f Chemical Engineering, California Institute of Technology, Pasadena, California 91 125
M. Flytzani-Stephanopoulos and K. Jothimurugesan Department o f Chemical Engineering, Massachusetts Institute o f Technology, Cambridge, Massachusetts 02139
Thermogravimetric analysis (TGA) and flow-reactor experiments were used to study sulfidationregeneration of highly porous Cu0-A1203 sorbents. In TGA studies, sulfidation of reduced sorbents produced a high-temperature form of digenite ( C U ~ + ~asS the ~ ) major crystalline product. When a platinum pan was used in the TGA, significant sulfur chemisorption on alumina occurred. Sulfur chemisorption on alumina was eliminated by use of a quartz sample pan. Reaction of the mixed-oxide sorbents with a mixture of H2S, H2, H20, and N2 in a packed-bed microreactor, a t temperatures between 550 and 800 "C, yielded prebreakthrough outlet-H2S levels considerably lower than those predicted by the sulfidation equilibrium of metallic copper. Independent reduction experiments confirmed t h a t alumina stabilizes CuO against complete reduction t o Cu. With this mechanism a t work, low prebreakthrough H2Slevels are attributed to sulfidation reactions of copper a t oxidation states of +1 or +2. T h e sulfided sorbents were completely regenerable in air/N2 mixtures with no deterioration of subsequent sulfidation performance. Among the most promising new technologies of power generation from coal are gasification integrated with combined-cycle power generation (gas turbine in series with a steam turbine), and coal gasification followed by oxidation in a molten carbonate fuel cell. Both technologies require removal of H2S (and COS) from the fuel gas prior to combustion. While this removal can be carried out at ambient temperatures by established technology, removal at high temperatures offers considerable improvement in process economics (Marqueen et al., 1986). In the original research on hot-gas desulfurization, pure metal oxides were tested, including those of zinc and iron (Morgantown Energy Technology Center, 1978). Recent work has focused on mixed metal oxides, especially zinc ferrite (ZnFe,O,). Above about 600 "C, however, sorbents containing zinc oxide lose zinc by reduction to the volatile metal. Iron and copper oxides are not subject to this limitation but do not possess sufficiently large sulfidation equilibrium constants to provide the required level of sulfur removal. In the case of copper oxide, the equilibria 2CuO + H2S + Hz = CuzS + 2Hz0 (1) CU~O + H,S CUZS + HzO (2) are very favorable. In the fuel gas atmosphere, however, the reaction CUO + H2 = CU + H2O (3) progresses rapidly so that sulfidation proceeds by reaction with metallic copper. 2Cu + H2S = CuzS + Hz (4) Present address: CR&DS, Monsanto Company, Creve Coeur,
MO 63167. 0888-5885/89/2628-0931$01.50/0
Table I. Equilibrium Constants for Sulfidation Reactions 1,2. and 4 from Data b s Robie et al. (1984) tema. K 7 00 800 900 1000 1100
reaction 1 21.3 19.2 17.6 16.3 15.2
reaction 2 11.0 9.9 9.0 8.3 7.8
reaction 4 4.1 3.7 3.4 3.2 3.0
As shown in Table I, the equilibrium constant of reaction 4 is much lower than those of reactions 1 and 2. The constant of reaction 4 is such that sulfidation of metallic copper generally does not provide adequate H2S removal for the aforementioned power generation applications. We have recently found that when copper oxide is employed in association with aluminum oxide, or iron oxide, or both, the level of sulfur removal is much higher than that obtained with pure copper oxide (Tamhankar et al., 1986). Such behavior was attributed to the retardation of copper reduction to the metallic form engendered by the association with iron or aluminum oxides. Previous work by Flytzani-Stephanopoulos et al. (1985) on the copper-containing mixed-oxide sorbents involved measurements of breakthrough curves in a packed-bed reactor and X-ray diffraction (XRD) analysis of fresh, sulfided, and regenerated sorbents to determine the crystalline phases present. The purpose of the present study was to obtain detailed mechanistic and qualitative kinetic information about the sulfidation and regeneration of the binary Cu0-A1203 sorbent. To this end, we have used extensively thermogravimetric analysis (in isothermal or temperature-programmed mode) combined with XRD and packed-bed reactor experiments. Some thermo0 1989 American Chemical Society
932 Ind. Eng. Chem. Res., Vol. 28, No. 7, 1989
gravimetric sulfidation and regeneration experiments of a copper-aluminum oxide sorbent were presented by Gangwal et al. (1989).
Experimental Section 1 Sorbent Preparation. A well-known, complexation method developed by Marcilly et al. (1970) was used to synthesize the mixed-oxide sorbents in a highly dispersed state The complexing agent (citric acid), equimolar to total metal cations, is added to an aqueous solution of metal nitrates of the desired stoichiometric proportions. An amorphous citrate precursor is prepared by evaporation of this solution. This evaporation proceeds first rapidly in a rotary evaporator under vacuum a t 70 "C, until a marked increase in viscosity of the solution is observed, and then for several hours (3-24) in a vacuum oven at 70 "C. until an amorphous solid foam forms. The foam is carefully broken up and calcined a t the desired temperature (between 550 and 900 "C) in air either under static conditions or under flow, to produce the final mixed oxide. 2. Wet Chemical Analysis. Hot nitric acid extraction, coupled with atomic absorption spectroscopy (AAS), was used to determine quantitatively the amounts of CuA1204, CuO, and A1,03 present in a given sample. Paulsson and Ros6n (1973) have reported that copper in the form of CuA120, cannot be extracted by hot nitric acid; however, copper present as a pure oxide is easily extracted. This wet chemistry has been confirmed in our laboratories. The concentration of the copper ions in solution was easily and accurately determihned by a modified Varian Techtron Model AA5 atomic absorption spectrometer. 3. X-ray Diffraction. A Siemens D500 step-scanning diffractometer employing Ni-filtered Cu K a radiation (1.54056 was used for qualitative chemical analysis of the polycrystalline components present in a sample. The X-ray tube was operated at 40 kV and 30 mA. X-ray powder diffractograms in a 28 range of 25-60" (or 30-50") were obtained that could detect the presence of CuO, CuA1,04, Cu20, CuAlO,, Cu, crystalline transition aluminas, Cu2S, CuS, and various other copper sulfides. Samples were finely ground and spread out evenly, so as to avoid preferred orientations, over a piece of double-stick tape adhered to a glass slide. Diffractograms were scanned at 0.1" intervals (in 28) for 60 s per interval. 4. Scanning Electron Microscopy. The samples were examined by an ETEC Corporation scanning electron microscope operating a t 20 kV with a resolution of 70 A. The sample was carefully ground and sprinkled on a metal stub containing a light coat of silver paste. This metal stub was then coated with a gold-palladium film 100 A in thickness prior to observation. 5. Thermogravimetric Analysis. A Du Pont 951 thermogravimetric analyzer interfaced through an analog-to-digital converter to a microcomputer served to measure the sample weight continuously. The quartz housing and flow path of the TGA were modified according to the system used by Ruth et al. (1972) so that corrosive gases, such as H2S, could be accommodated. A temperature programmer enabled either isothermal operation or operation under a linear temperature profile. It was verified experimentally that the rate of reduction of ultrapure CuO (99.999% purity) decreased when the particle size exceeded 25 mesh (707 pm), when the gas flow rate was less than 60 cm3/min or greater than 120 cm3/ min, or when the sample size exceeded 50 mg. On the basis of these findings, a 30-mg sample of particles, 120-170 mesh (88-125 km), was typically employed to minimize internal mass-transfer effects (e.g., intra- and inter-particle diffusion), and the gas flow rate was fixed at 80 cm3/min
a)
to minimize external mass-transfer effects (e.g., gas-film diffusion and particle entrainment). A flow rate of 80 cm3/min was also used for the stream of N2 serving as the protective backflow gas. For temperature-programmed reduction, a reactant gas containing 5% H2 in N2 was used and the temperature was ramped from 233 to 912 "C a t a rate of 1.6 "C/min. For experiments other than temperature-programmed reduction, the reaction-gas mixtures were 5% H2 in N2 for the reduction runs, 4.2% H2S in N,for the sulfidation runs, and air for the regeneration runs. 6. Packed-Bed Microreactor Experiments. The experimental system will be described here very briefly, for it has been described in detail by Tamhankar et al. (1986). The reactor consists of a quartz tube of 1-cm i.d. and 41-cm length loaded to a bed height of 4-6 cm with a mixture of sorbent granules (-20 to +35 mesh) and inert alumina particles of low surface area (Alcoa T-64, -28 to +48 mesh). The sorbent bed was supported by a fritted quartz disk on one end and packed with quartz wool at the opposite end. The reactor tube was mounted vertically inside an electric furnace, and the bed temperature was monitored by a K-type thermocouple moving inside a quartz thermowell (0.3-cm i.d.) concentric to the reactor tube. Different gases from cylinders passed through purifiers and calibrated flowmeters into a common gas line. The desired gas mixture flowed either upward (sulfidation) or downward (regeneration) through the sorbent bed. The lines leading to the reactor tube were insulated and heated. Nitrogen bubbling through water maintained at a constant temperature in a three-neck flask assembly was used to introduce known amounts of water vapor into the gaseous feed stream. Temperatures at various locations in the reactor system were monitored by K-type thermocouples connected to a multichannel digital readout. The reactor pressure in all cases was slightly above atmospheric. For a sulfidation run, fresh or sulfur-free sorbent was exposed to a feed gas containing H 2 (15-20%), H 2 0 (7-25%), H2S (0.2-1%), and N2 balance a t a constant temperature in the range 550-800 "C. Sulfided sorbents were regenerated using a mixture of 90 mol % N2 and 10 mol 9% air at temperatures between 550 and 800 "C. Feed gas rates of 200 cm3/min (STP) were typically used, and the gas hourly space velocity was -2000 h-' (STP)in most tests. The product gas was analyzed for H2S in sulfidation runs and SO, in regeneration runs by a HP-5830A gas chromatograph equipped with a flame photometric detector. The TGA experiments and the packed-bed microreactor experiments each have distinct advantages and disadvantages and provide complementary information. The packed-bed reactor experiments are suitable for obtaining overall sorbent performance under conditions similar to those of an industrial reactor. These experiments can provide information about sulfur-removal efficiency and can generate large samples for solid analysis. However, they are not well suited to kinetic studies because the inherent gradients of gas and solid composition along the reactor make the calculation of reaction rates mathematically tedious. TGA runs, on the other hand, involve uniform gas and solid composition and are better suited for kinetic investigations. However, the small sample size that must be used to avoid mass-transfer limitations may be insufficient for certain analytical procedures.
-
Results and Discussion TGA results are discussed in terms of normalized weight (W,) versus time, while microreactor results are given in terms of the mole fraction of H,S in the product gas versus
Ind. Eng. Chem. Res., Vol. 28, No. 7, 1989 933 Table 11. Fresh Sorbents Used in TGA Studies
BET sample 72-1-CA 33-1-CA6 11-1-CA3
mole % CuO:CuA1,04:A19On 47.5547.5 2:96:2
1oo:o:o
surface area, m2/g 120 2.72 1.11
normalized time (t/t*). Normalized weight is the ratio of the instantaneous weight change to the maximum weight loss as follows:
where m, mo,and mf are the instantaneous, initial, and final mass (at complete reduction to Cu), respectively. For example, following complete reduction of pure CuO to Cu, the final normalized weight is WN = 0. Subsequent complete sulfidation of this Cu to Cu2S results in WN = 1. Normalized time is the ratio of the actual run time, t , over the time that would be required for complete conversion of the sorbent to the postulated sulfide product, t*. 1. Mechanistic and Kinetic Studies. a. Sorbent Characterization. Two amorphous citrate precursors were prepared and exposed to 4-h calcinations. One precursor was calcined a t 900 "C to yield predominantly compound oxide (CuA1204,sample 33-1-CA6), while the other was calcined a t 550 "C to yield a mixture of oxides (CuO and A1203, sample 72-1-CA). For comparison, a sample of pure CuO was prepared by the same procedure and calcined in air a t 550 "C for 4 h (sample 11-1-CA3). Components present in crystalline form were identified by XRD analysis, while quantitative chemical compositions (Table 11) were determined by atomic absorption spectroscopy (AAS). SEM analysis showed the markedly different microstructures possessed by these three materials (Figure 1). Note that the micrographs display both a primary porous structure and a superimposed secondary texture. The secondary texture is smooth in the case of CuA1204(sample 33-1-CA6),homogeneous and coarse for pure CuO (sample ll-l-CA3), and inhomogeneous with granules for a mixture of oxides (sample 72-1-CA). Consequently, the presence of free alumina creates the appearance of crystallites on the surface of fresh sorbents. Figure 2 shows the results of temperature-programmed reduction (TPR) of the three sorbents. The major observation is that the reduction rate of CuA1204(sample 33-1-CA6) is more than 1order of magnitude lower than that of pure CuO (sample 11-1-CA3)proceeding through CuA102and Cu20 as intermediates. Furthermore, independent TGA experiments revealed that, for a sorbent containing a mixture of CuA1204,CuO and inert A1203,the major fraction of the pure oxide, CuO, were reduced rapidly, while a smaller fraction of the CuO and the compound oxide, CuA1204,was reduced very slowly. For sample 721-CA, 37 mass ?& of the CuO was observed to reduce more slowly than pure CuO. The association of CuO with A1203 was believed to be responsible for retarding reduction of this species (Patrick and Gavalas, 1989). This phenomenon has been reported for NiO/7-A1203 (Puxley et al., 1983) and CuO/y-A1203 prepared by impregnation (Friedman and Freeman, 1978; Strohmeier et al., 1985). The type of transition alumina present in fresh samples could not be ascertained by XRD. Since no amorphous halo was evident in X-ray diffractograms, it was assumed that the alumina in fresh samples existed in a microcrystalline form. Furthermore, both the high BET surface areas measured for fresh samples (Table 11) and the association between CuO and alumina, which was compa-
(1,) 33-I-CAB
(a) 72-1-CA
(c) 1 l - l - C A 3
Figure 1. Scanning electron micrographs of fresh oxides prepared by the citrate process. All cases a t 2750X: (a) CuO, A120s (72-1-CA); (b) CuA1204 (33-1-CA6); (c) CuO (11-1-CA3); markers are 1 pm. 1
0.8
0.6
z
3
0.4
0.2
0. 0.
200
100
300
TIFIE (blIN)
Figure 2. TPR of samples (a) 11-1-CA3, (b) 72-1-CA, and (c) 33-1CA6. Temperature increased by 1.6 OC/min starting from 233 "C.
rable to that reported for materials containing CuO and 7-A1203,indicated that some fraction of the alumina in fresh samples was 7-A1203. b. Isothermal Experiments in the TGA. Desulfurization of coal gas by CuO or Cu0-A1203 involves simultaneous reduction and sulfidation of the sorbent (reactions
934 Ind. Eng. Chem. Res., Vol. 28, No. 7, 1989 a
2.5
2
z
3
1.5
1
0-5
0. 0.
25
50
75
TIME (MIN)
Figure 3. Reduction, sulfidation, and air regeneration in series for sample 72-1-CA at (a) 600 "C, (b) 700 "C, (c) 800 "C, and (d) 900 "C. In all cases, regeneration commenced at 43 min from the origin of time. ( c ) Sulfidcd
1-4). For the purpose of mapping out the reaction network and measuring reaction rates, however, it is useful to separate reduction and sulfidation. Therefore, in the TGA experiments, reduction and sulfidation were carried out consecutively. Isothermal reduction and sulfidation of sample 72-1-CA were performed in the TGA system for temperatures of 600,700,800, and 900 "C (Figure 3). Consistent data were obtained by careful reproduction of sorbent heating procedures. Reduction was carried out with 5% H2 in N2 for the first 5 min, followed by sulfidation with 4.2% H2S in N2 for the next 35 min. XRD analysis of a sample following sulfidation a t 700 "C identified CulbS (digenite) as the dominant crystalline phase, while SEM analysis of the same sample revealed large crystals (5-12.5 pm) composed of interlocking malformed octahedra. In the majority of the octahedral building blocks, two of the eight faces predominated, creating flat hexagonal plates. This same crystal morphology was observed by Donnay et al. (1958) in scanning electron micrographs of digenite crystals grown from mixtures containing copper-to-sulfur ratios between 9:5 and 2:l and grown a t temperatures between 500 and 775 "C. Since for both Donnay's study and this study SEM was performed on samples cooled to room temperature, the observed phase was a low-temperature form of digenite, low digenite. There exists a high-temperature form of digenite, high digenite, that forms from low digenite a t temperatures greater than approximately 83 "C (Craig, 1974). High and low digenite both have sulfur atoms in approximate cubic close packing and are solid solutions of formula Cug+,S5. On the other hand, the composition range of low digenite (-0.166 < x < 0.125) is narrower than that of high digenite (-0.35 < x < 1.00),and low digenite has been isolated in the form of various polymorphs or superstructures of high digenite (Roseboom, 1966). For example, high digenite has a cubic cell of lattice parameter 5.56 A, while low digenite has a pseudocubic cell with lattice parameter (5.56 &N, where N has been found to
(d) S I J I f i d C d
Figure 4. Scanning electron micrographs (all cases at 2750X) of sample 72-1-CA (a) fresh, (b) reduced at 700 "C, (c) sulfided at 700 "C in a platinum pan, and (d) sulfided at 700 "C in a quartz pan (markers are 1 pm).
take on both integral (i.e., N = 5,6) and nonintegral (i.e., N = 5.2,5.5,5.7,5.8) values (Morimoto and Gyobu, 1971; Morimoto and Koto, 1970; Morimoto and Kullerud, 1963). While the method of powder XRD enables the distinction between high and low digenite, the method of single-crystal XRD is required for identification of a particular superstructure of low digenite. The succession of scanning electron micrographs shown in Figure 4 exhibits the marked structural changes accompanying consecutive reduction and sulfidation a t 700 "C. The fresh 72-10! having inhomogeneous grainy texture is converted upon reduction to a dispersion of spherical copper particles on alumina. Subsequent sulfidation produces a bimodal dispersion of large (2-6-pm), octahedral crystals and small ( 1. However, sulfidation a t 900 "C exhibits unexpected behavior, suggesting decomposition of the nonstoichiometric sulfides that have formed. Regeneration of sulfided sorbents in air was investigated to test the sorbent performance in repeated sulfidationregeneration cycles. Regeneration was carried out in air for 40 min following sulfidation. The regeneration of the sorbent nominally proceeds by the reactions cu2s + 20, = 2CuO + so2 (6) 19 + x CUg+xS5+ -02 = (9 + X)CUO 5SO2 (7) 2 followed by the solid-state reaction
Ind. Eng. Chem. Res., Vol. 28, No. 7, 1989 935 3.5
Table 111. Normalized Weight, WN,Following Reduction, Sulfidation, and Regeneration in the TGA regeneration sample T," "C panb reduction sulfidation 1.27 72-1-CA 700 P 0.03 1.10 0.07 0.96 1.03 Q 1.47 0.06 1.11 800 P 0.03 0.99 0.99 Q 1.04 1.10 11-1-CA3 700 P 0.00 1.00 0.00 1.06 Q 0.12 1.11 1.02 800 P 1.02 0.00 1.05 Q 1.12 0.24 1.26 33-1-CA6 700 P 1.06 0.28 0.93 Q 1.07 0.03 1.51 800 P 1.00 1.02 0.03 Q
3
2.5
z
P
2
"Temperature measured about 1 cm above the sample pan. b P is platinum; Q is quartz.
1.5
1
0.5 0
10
20
30
40
TIbE ( M I N )
Figure 5. Reduction followed by sulfidation of pure A120, using a platinum pan.
the extent of which depends on the regeneration temperature. However, the production of SOz in creactions 6 and 7 can lead to the simultaneous formation of sulfate by the following side reaction: CUO
+ so2 + 1/02= c u s o ,
(9)
Isothermal regeneration of sulfided 72-1-CA at 600, 700, 800, and 900 "C is included in Figure 3. Regeneration at 600 "C is rapid and yields both CuO and a significant fraction of CuSO, (corresponding to about one-third of the copper). The slow weight loss during the remainder of regeneration suggests the presence of other sulfur-containing species since CuSO, does not decompose below 650 "C. It is possible that a portion of the weight gain during regeneration is due to the adsorption of SOz on alumina or to the formation of aluminum sulfate. The presence of CuS04 was confirmed by XRD, while the presence of other sulfur-containing species could not be detected directly. A t 700 "C, the sulfate formed (about 8% of the copper) and decomposed at appreciable rates. At 800 and 900 "C, there is no evidence of sulfate formation during regeneration. XRD analysis shows a higher content of CuA1204 in the regenerated sorbent as compared to the fresh sorbent, in keeping with the fact that the regeneration temperatures are greater than the calcination temperature (550 "C) used in sorbent preparation. The sample regenerated a t 900 "C contained some crystalline CuA102 in disagreement with the thermodynamic data of Jacob and Alcock (1975). The implication is that a fraction of CuAIOz remained unreacted during both sulfidation and regeneration at 900 "C. A possible explanation for this finding is the formation of a layer of sintered CuAlZ0, surrounding CuAIOz during reduction, causing mass-transfer resistance to the sulfidation and regeneration of CuA102. c . Sulfur Chemisorption on Alumina. The sulfidation curves of Figure 3 suggest a contradiction: Weight gain exceeds W , = 1.1,which corresponds to complete conversion of copper to Cuss, (digenite), the most sul-
fur-rich form of sulfide identified by XRD. One possible explanation is that the alumina component of the sorbent contributes directly to sulfur retention. To test this possibility, a sample of pure A1203(prepared by the citrate process using calcination a t 550 "C) was exposed to isothermal reduction and sulfidation at different temperatures (Figure 5 ) , as in the case of sample 72-1-CA (Figure 4). The final weight (W,) used in the expression for WN in Figure 5 was estimated as 0.05Wi based on a separate TPR experiment (Thomas et al., 1981). During sulfidation, the alumina showed significant weight gain, which increased with temperature. Reduction and sulfidation in succession of an empty platinum pan at 700 and 800 "C revealed no weight change; thus, chemisorption of sulfur on the platinum pan or formation of platinum sulfides was negligible. Since alumina is inert to sulfidation, these observations indicate the adsorption of sulfur-containing species on alumina. d. Experiments using a Quartz Pan. In an attempt to eliminate any role of 'platinum on sulfur species chemisorption on the alumina in the sorbent, reduction and sulfidation experiments a t 700 and 800 "C were repeated using a quartz boat in the TGA. While the basic trends of the weight cycle were similar to those observed with the platinum pan, the weight excursion above W , = 1 was drastically reduced (Table 111). Note that this weight excursion was significant only for the alumina-containing samples. With both the quartz and the platinum pans, the weight gain following sulfidation increased with temperature. The difference in the weight excursion above W , = 1between the experiments with the platinum and quartz pans suggests that the platinum pan catalyzed the decomposition of H2S to elemental sulfur, which, in turn, chemisorbed readily on alumina. This conclusion is supported by the well-known, catalytic activity of platinum in HzS decomposition (Fukuda et al., 1978; Wore11 and Kaplan, 1979; Bartholomew and Agrawal, 1982). The rates of reduction for samples 72-1-CA, 33-1-CA6, and 11-1-CA3 revealed that the maximum rates of both reactions were highest for sample 72-1-CA and lowest for sample 11-1-CA3. This difference is attributed to sintering of copper, which occurred in sample 11-1-CA3 but not in 72-1-CA, in agreement with independent experiments (Patrick and Gavalas, 1989). The intermediate reduction and sulfidation rates exhibited by sample 33-1-CA6 were due to the slower reduction and sulfidation kinetics of CuAlZO4as compared to CuO. These trends were observed to be even more drastic at 800 "C than at 700 "C. Low digenite (Cus+,S5) was again identified by XRD in cooled samples, suggesting that high digenite was the major sulfidation product at 700 and 800 "C (Table IV) for all
936 Ind. Eng. Chem. Res., Vol. 28, No. 7, 1989 Table IV. XRD Analysis of Sulfided Sorbents sorbent sulfidation temp, "C crystallinity$ % CAT" 800 50-60 CA;I" 72-1-CAb 72-1-CAE 72-1-CAC 11-1-CA3b 11-1-CA3' 11-1-CA3' 33-1-CA6b 33-1-CA6' 33-1-CA6'
800 700 700 800 700 700 800 700 700 800
" From microreactor experiments. available.
'
50-60 NA NA NA NA NA NA NA NA NA
CuA1204
none none trace trace trace none none none major major mediuim
T G A experiments using the platinum pan.
three samples. In addition, a large quantity of unreacted CuA1204was identified in sample 33-1-CA6 after sulfidation a t 700 "C. This observation was consistent with the extent of reduction to metallic copper, which at 700 "C was nearly 100% for 72-1-CA and 11-1-CA3 but only 70% for 33-1-CA6 (Table 111). Copper oxide, CuO, and products of reduction Cu20, CuA102,and Cu were not identified a t the end of the sulfidation. Evidently, these compounds reacted more rapidly than CuA1204to form the sulfide cu9+xs5. In a study by Chung and Massoth (1980), it was found that CoA1204could not be sulfided a t 400 "C, and consequently, it was proposed that tetrahedrally coordinated C O +was ~ very stable. Both CoA1204and CuA1204possess spinel structures; however, the distribution of cations in each compound is different. CoA1204is a normal spinel with Co2+occupying exclusively tetrahedral sites and A13+ occupying exclusively octahedral sites (Navrotsky and Kleppa, 1968), while CuA1204is a partially inverse spinel with Cu2+occupying both tetrahedral (tet)and octahedral (oct) sites such that x = 0.4 in the formula [Cul-,A1,],,[Cu,A12-,lod04. I t is conceivable that the 60% Cu2+that occupies tetrahedral sites in CuA1204was responsible for the slow sulfidation kinetics observed for sample 33-1-CA6 at 700 and 800 "C. The XRD patterns of sulfided 33-1-CA6 reveal that the peaks of CuA1204are about twice as intense as those of Cug+,S5. In view of the sharpness of these peaks, the intensities are approximately proportional to the content of the crystalline compounds. However, comparing the weight loss during reduction with the weight gain during sulfidation (Table 111)shows that the content of the sulfide is actually higher than the content of unreacted CuA1204. Thus, the XRD intensity relations can be explained only by assuming that the sulfide is predominantly in an amorphous state form. Furthermore, the XRD intensities for Cug+,S5 formed by sulfidation of 11-1-CA3 are 2-3 times greater than those for Cug+,S5formed by sulfidation of alumina-containing samples, suggesting that a high fraction of copper sulfided in the presence of alumina forms amorphous or microcrystalline sulfide. The scanning electron micrographs (Figures 4 and 6) of sulfided alumina-containing samples (sulfided 72-1-CA and 33-1-CA6) show large (2-6-pm), octahedral crystals and small (
