1238
Ind. Eng. Chem. Res. 2008, 47, 1238-1244
SEPARATIONS Superior Sorbent for Natural Gas Desulfurization Dennis Crespo, Gongshin Qi, Yuhe Wang, Frances H. Yang, and Ralph T. Yang* Department of Chemical Engineering, UniVersity of Michigan, Ann Arbor, Michigan 48109
New sorbents were investigated for desulfurization of natural gas by adsorption at ambient temperature. The new sorbents included Cu(I)Y zeolite, CuCl/MCM-41, and CuCl/SBA-15 and were compared with the best commercial sorbents, AgY and activated carbon. H2S and dimethyl sulfide (DMS) were used as the model sulfur compounds. Ab initio molecular orbital calculations on adsorption energies showed that Cu(I)Y > AgY > Cu(II)Y for both H2S and DMS and H2S adsorbs more strongly than DMS, both in full agreement with experimental results. Moreover, Cu(I)Y, CuCl/MCM-41, and CuCl/SBA-15 were fully regenerable, while AgY was not. Fixed-bed adsorber breakthrough results indicated severe diffusion limitation in sulfur capture; thus, the mesoporous sorbents are advantageous for very high throughput applications. 1. Introduction Fuel cells such as the proton exchange membrane (PEM) fuel cell and solid oxide fuel cell (SOFC) are the most energy efficient and clean energy generation systems, including distributed power generation for future “hydrogen economy”. These fuel cells require hydrogen as the fuel. The most economical way to produce hydrogen is by catalytic reforming of natural gas, with LPG (i.e., commercial propane and butane) used to a lesser extent. These fuels are the fuels of choice for hydrogen because of their relative abundance, availability of supply infrastructure, and safety in handling. These fuels must be dosed with sulfur odorants such as thiols and sulfides for safe handling during transportation and utilization. There is also residual hydrogen sulfide (H2S) in these fuels (∼5-10 ppm). These sulfur compounds must be removed before being fed to the catalytic reformers because they would otherwise poison both the catalysts in the reformers and the catalysts in the fuel cells. The concentrations of sulfur in the pipeline natural gas and LPG are typically around 10 ppm or higher. The acceptable sulfur levels for reformers and fuel cells are well below 1 ppm and are preferably Cu(II)Y > Cu/AC, but the H2S adsorption capacities on Cu(I)Y and AgY are very similar to each other. That means both Cu(I)Y and AgY are excellent sorbents for H2S adsorption removal (while the cost of Ag is much higher than that of Cu).
Figure 2. Isotherms for DMS adsorption on different sorbents at 25 °C (DMS in He, at 1 atm): Cu(I)Y (b), AgY (4), Cu(II)Y (O), CuCl/SBA-15 (9), Cu/AC (0), fitting data (line). Table 2. Composition of Different Sorbents adsorbent
composition
preparation
Cu(I)Y
Cu54H3Al57Si135O384
Cu(II)Y
(CuOH)54H3Al57Si135O384
AgY Cu/AC
Ag63Al57Si135O384 5 wt % Cu/active carbon
VPIE at 700 °C and auto reduction at 450 °C VPIE at 700 °C and pretreatment at 350 °C LPIE and pretreatment at 350 °C from Su¨d-Chemie
Table 3. Langmuir-Freundlich Parameters for Adsorption of H2S and DMS at 25 °C adsorbent
adsorbate
AgY
H2S DMS H2S DMS H2S DMS H2S DMS
Cu(I)Y Cu(II)Y Cu/AC
qm (mmol/g)
B (ppm-(1/n))
n
6.35 39.21 7.15 4.70 5.08 6.50 0.97 1.59
0.16 0.015 0.24 0.14 0.14 0.04 0.11 0.084
1.40 2.59 2.03 1.44 1.97 1.83 0.81 1.93
Figure 2 shows the equilibrium isotherms of DMS on different sorbents at 25 °C. The data were fitted by the LangmuirFreundlich isotherm, also shown by the solid lines in the figure. It can be seen that all sorbents have very high capacities for DMS, even at very low concentration at the 0.5 ppm level. The capacity of the adsorption decreased in the order Cu(I)Y > AgY > Cu(II)Y > Cu/AC in the whole DMS concentration range investigated in this work. From Figures 1 and 2, it is clearly shown that the adsorption capacity of H2S is higher than that of DMS for each sorbent we studied, which was predicted from the molecular orbital calculations that showed stronger bonds with H2S than DMS. The parameters for the measured adsorption isotherms of H2S and DMS are summarized in Table 3. The temperature dependence of the Langmuir-Frendlich isotherm parameters (qm, B, and n) can be estimated from Table 3. The theoretical temperature dependence of the Langmuir constant, B, is approximately (by neglecting the effect of T in the pre-exponential factor)
B ∝ eE/RT
(3)
where E is the heat of adsorption (a positive value). The temperature dependence of qm can be estimated using the empirical “Gurvitsch rule”; i.e., qm is equal to the pore volume divided by the molar volume of the adsorbate (as a liquid). Thus, the temperature dependence of qm follows that of the liquid density. Since n is only an empirical parameter,
Ind. Eng. Chem. Res., Vol. 47, No. 4, 2008 1241
Figure 3. Isotherms for H2S adsorption on different sorbents at 60 °C (H2S in He, at 1 atm): Cu(I)Y (O), AgY (b), Cu(II)Y (0), Cu/AC (9), fitting data (line).
Figure 4. Isotherms for DMS adsorption on different sorbents at 60 °C (DMS in He, at 1 atm): CuY (9), AgY (0), Cu(II)Y (b), Cu/AC (O), fitting data (line).
there is no theoretical basis for the temperature dependence of n. A linear correlation between n and T would be a good approximation. 4.3. H2S/DMS Adsorption Isotherms at 60 °C. Figure 3 shows the isotherm of H2S on different sorbents at 60 °C. The data were also fitted by the Langmuir-Freundlich isotherm shown as solid lines in Figure 3. A similar trend was observed for the isotherms compared to the isotherms at 25 °C. The adsorption capacity decreases in the following order: Cu(I)Y > AgY > Cu(II)Y > Cu/AC. Figure 4 shows the isotherm of DMS on different sorbents at 60 °C. The data fitted by the Langmuir-Freundlich isotherm are also shown as solid lines in Figure 4. The results were also very similar to those at 25 °C. The parameters for the measured adsorption isotherms of H2S and DMS are summarized in Table 4. 4.4. CH4 Adsorption Isotherms. To get an idea about the performance of the adsorbents when working with natural gas, methane isotherms were measured because it is the major component of natural gas. The adsorption isotherms of CH4 on Cu(I)Y and AgY were measured at room temperature (25 °C) and 60 °C. It was found (Figure 5) that a very small amount of CH4 was adsorbed compared to that of the sulfur compounds, even at much higher pressures, which means these sorbents have high selectivity on sulfur compounds and still maintain a high sulfur capacity. 4.5. Adsorption Energies. Adsorption energies were calculated from experimental data for the different sulfur species (H2S and DMS) on the different sorbents. Adsorption isotherm data
Figure 5. Isotherms of CH4 on Cu(I)Y at different temperatures: ([) 25 °C, (0) 60 °C.
Figure 6. Breakthrough of DMS on fresh and regenerated Cu(I)Y diluted in 3A zeolite 20/50 (w/w) with GHSV ) 27 000 h-1 at 50 °C. Ci is the feed concentration, and Ct is the effluent concentration.
Figure 7. Breakthrough of DMS 10 (ppmw) on Cu(I)Y diluted in 3A zeolite 20/50 (w/w) at different GHSVs and 50 °C. Ci is the feed concentration, and Ct is the effluent concentration.
were used along with the Clausius-Clapeyron equation to obtain the heats of adsorption. The Clausius-Clapeyron equation is given by
d(ln P) ∆H )R d(1/T)
(4)
The obtained results are summarized in Table 5. As can be seen from the results, the adsorption energies follow the order Cu(I)Y > AgY > Cu(II)Y for both H2S and DMS. Moreover, the heats of adsorption for H2S are higher than that of DMS for all adsorbents. This is in fair agreement with the results from the molecular orbital calculations, shown in Table 6. 4.6. Regeneration of Cu(I)Y by Heating in an Inert Gas. The regeneration of the spent Cu(I)Y was investigated by
1242
Ind. Eng. Chem. Res., Vol. 47, No. 4, 2008
Table 4. Langmuir-Freundlich Parameters for Adsorption of H2S and DMS at 60 °C adsorbent
adsorbate
CuY
H2S DMS H2S DMS H2S DMS H2S DMS
Cu(II)Y AgY Cu/AC
qm (mmol/g)
B (ppm-(1/n))
n
2.36 1.85 1.44 1.38 2.32 1.78 1.61 0.42
0.43 0.11 0.19 0.10 0.13 0.06 0.10 0.06
1.33 0.73 1.05 1.21 0.87 0.83 1.40 0.96
Table 5. Energies of Adsorption (kcal/mol) for Sulfur Adsorbates (Obtained from Experimental Isotherm Data) adsorbate
∆E on Cu(I)Y
∆E on AgY
∆E on Cu(II)Y
H 2S DMS
-14.0 -11.3
-11.4 -7.9
-9.5 -6.6
Figure 8. Breakthrough of DMS (10 ppmw) on AgY diluted in 3A zeolite 20/50 (w/w) at GHSV ) 27 000 h-1 and 50 °C. Ci is the feed concentration, and Ct is the effluent concentration.
Table 6. Energies of Adsorption (kcal/mol) for Sulfur Adsorbates (from Ab Initio Molecular Orbital Calculations) adsorbate
∆E on Cu(I)Z
∆E on AgZ
∆E on Cu(II)Z
H2S DMS
-17.5 -13.2
-16.1 -7.8
nil -1.3
Table 7. Desorption Amount of DMS and H2S on Cu(I)Y regeneration conditions
[DMS] (mmol/g)
[H2S] (mmol/g)
fresh heat at 200 °C in He gas heat at 300 °C in He gas heat at 450 °C in He gas
1.7 1.01 1.30 1.48
3.23 1.47 1.76 2.23
heating in an inert gas at different temperatures. Table 7 shows the desorption amounts of H2S and DMS on Cu(I)Y at different heating temperatures. When the sorbent was heated at 200 °C, the amounts desorbed were 1.01 and 1.47 mmol/g for DMS and H2S, respectively, which corresponded to 59% and 45% of the original amount for the fresh sample. When the used sample was treated at 450 °C, 87% of DMS and 69% of H2S were desorbed. On the basis of these results, it can be concluded that, on the Cu(I)Y sorbent, the adsorption energy of H2S is higher than that of DMS, so the adsorbed DMS was relatively easily removed. This is in agreement with the adsorption energy results. CuY could be regenerated completely (100% regeneration) by reacting with air at 350 °C followed by autoreduction as we showed previously.8-10 4.7. Breakthrough Experiments. Fixed-bed breakthrough experiments were carried out to investigate the adsorption capacity of the different sorbents and compare their effectiveness in removing sulfur compounds at different flow conditions. DMS diluted in He was used as the sulfur-containing compound since it is the most difficult odorant to remove from the studied compounds.4-7 High adsorption capacity samples were diluted with 3A zeolite to provide a longer fixed bed and allow the use of a smaller amount of adsorbent sample to make the experiment length more manageable. DMS is not adsorbed in 3A due to pore size exclusion, so it is an excellent choice for this purpose (as a diluent). A reference run was performed on plain 3A zeolite, and the results show very little DMS adsorption capacity (0.00019 mmol/g), which is less than 1% of the breakthrough capacity for the tested materials. Furthermore, the experiments were carried out at an elevated temperature (50 °C) to decrease the amount of time per run since the samples have a lower capacity at higher temperature. The sample (typically 70-100 mg) was then placed on a vertical glass adsorber (6 mm i.d.)
Figure 9. Breakthrough of DMS (98.4 ppmw) on Cu(I)Y and AgY. GHSV ) 39 000 h-1 and 25 °C. Ci is the feed concentration, and Ct is the effluent concentration.
with a quartz frit. The reactor was wrapped with a heating tape and insulated with ceramic wool wrapping. A thermocouple was placed at the reactor surface to monitor the temperature. The temperature was controlled with a PID temperature controller. The outlet of the reactor is connected to a GC-FID, with a sensitivity tested to 2 ppm DMS, to continually monitor the gas as it exits the column. The breakthrough of DMS on VPIE Cu(I)Y is shown in Figure 6. The run was repeated after regeneration of the sample in air at 350 °C for 8 h followed by autoreduction.8 As can be seen, the adsorption capacity is almost identical to the fresh capacity, which further proves the sorbent is fully regenerable. The effect of different gas flow rates (gas hourly space velocity, GHSV) is shown in Figure 7; as expected, under the same conditions, a higher flow rate led to a lower adsorption capacity. To compare our sorbents with AgY zeolite,6,26 the effect of regeneration is investigated for DMS on AgY zeolite. The results are shown in Figure 8. As can be seen, the capacity is roughly halved after each regeneration cycle. This is consistent with the results of Satokawa et al.,6 which reported roughly half of the adsorption capacity for their Y with the highest Ag exchange. The effect of different flow rates was not investigated due to the severe adsorption capacity reduction seen after regeneration. A direct comparison between fresh samples of VPIE Cu(I)Y and AgY can be seen in Figure 9. The breakthrough capacity is higher for VPIE Cu(I)Y; however, the total adsorption capacity is slightly higher for AgY. This agrees with the isotherm data at higher concentrations where AgY has a higher capacity than Cu(I)Y. The mesoporous sorbents (based on MCM-41 and SBA-15) were studied next. These sorbents have very large and uniform pore sizes, as shown in Table 1. Figures 10 and 11 show the
Ind. Eng. Chem. Res., Vol. 47, No. 4, 2008 1243
Figure 10. Breakthrough for DMS (92.8 ppmw) on active carbon at 50 °C and different GHSVs. Ci is the feed concentration, and Ct is the effluent concentration. Figure 13. Isotherms for H2S adsorption (H2S in He, at 1 atm) on VPIE Cu(I)Y and CuCl/SBA-15 at 25 °C: fresh Cu(I)Y (9), regenerated Cu(I)Y (0), fresh CuCl/SBA-15 (b), regenerated CuCl/SBA-15 (O), fitting data (line).
Figure 11. Breakthrough of DMS on CuCl/SBA-15 with different GHSVs and at 25 °C. Ci is the feed concentration, and Ct is the effluent concentration.
Figure 12. Breakthrough of DMS on fresh CuCl/MCM-41, fresh CuCl/SBA-15, and regenerated CuCl/SBA-15 with GHSV ) 60 000 h-1 at 25 °C. Ci is the feed concentration, and Ct is the effluent concentration.
flow rate effect on the adsorption of DMS on activated carbon and CuCl/SBA-15. Again, the adsorption capacity decreases with increasing flow rate. Given the relatively high adsorption capacity of CuCl/SBA-15, second and third adsorption cycles were performed to verify the capacity of the regenerated material; this can be seen in Figure 12. As can be seen, the sample is also fully regenerable. Figure 12 also shows the comparison of two different materials with large pore sizes: CuCl/MCM-41 and CuCl/SBA-15. It is clear that CuCl/SBA-
15 outperforms CuCl/MCM-41 in the breakthrough capacity (at the same space velocity), due to the much larger pores of CuCl/ SBA-15. It is important to notice that in all cases the breakthrough occurs well before the equilibrium amount is reached. This can be attributed to a slow rate of pore diffusion. There are two main factors affecting these results. First, the gas flow rate used in the experiments is rather high, and thus, there is not enough time to reach equilibrium and early breakthrough occurs. Israelson performed similar experiments with lower flow rates (GHSV from 150 to ∼1200 h-1), which took months to complete, and the adsorption capacity was nowhere near the equilibrium capacity we obtained for our samples.4 The time required to achieve full adsorption capacity using such low flow rates will be well over 300 days and thus not practical for our study. Second, it was observed that the samples with smaller pores (zeolites) adsorb a lower fraction of the equilibrium amount than the larger pore samples (CuCl/SBA-15 and active carbon). King and Li proposed a redox mechanism for the adsorption on Cu(I)Y.27 It is possible that adsorbed molecules or some of the redox products (RSCu) may accumulate at the pore entrance and partially hinder the diffusion of other gas molecules, effectively reducing the rate of diffusion. This would explain the shape of the obtained breakthrough curves, which are very sharp, as opposed to the expected breakthrough curves, which should be slanted and slowly increase to the inlet concentration. A similar case can be made for AgY zeolite. This phenomenon does not seem so dramatic in larger pore samples. The fact that both small- and large-pore samples are affected by this reduced capacity leads to diffusion limitation. Further work including modeling and diffusion constant measurements are ongoing. 4.8. Second Adsorption Cycle Isotherms. Second adsorption cycle isotherms were measured for the materials that showed both relatively high adsorption capacity from the fresh sample adsorption isotherms and full regeneration in the breakthrough experiments, namely, VPIE Cu(I)Y and CuCl/SBA-15. Figure 13 shows these isotherms, and the results confirm the breakthrough experiment results: both samples are fully regenerable. In this case, VPIE Cu(I)Y was regenerated in air at 350 °C for 8 h and CuCl/SBA-15 was regenerated in He at 350 °C. These results are particularly important since a fully regenerable material is highly desired.
1244
Ind. Eng. Chem. Res., Vol. 47, No. 4, 2008
5. Conclusion The sulfur adsorption capacity was measured for various sorbent materials. It was found from ab initio molecular orbital calculations that the energy of adsorption of H2S was greater than that of DMS on both Cu(I)Y and AgY; furthermore, the energy of adsorption decreased in the following order: Cu(I)Y > AgY > Cu(II)Y. Energies of adsorption calculated by molecular orbital theory and those calculated by the ClausiusClapeyron equation using experimental data showed good agreement and the same trends. Breakthrough curves show early breakthrough (i.e., below equilibrium), likely due to diffusion limitation. Lower flow rates increased the breakthrough capacity. Cu(I)Y and AgY show the highest adsorption capacities from among the tested samples. However, when the regeneration capacity was tested, Cu(I)Y displayed superior performance by being fully regenerable. Furthermore, the Cu(I)Y shows little affinity for CH4, which is desirable for natural gas desulfurization. CuCl supported on mesoporous supports (MCM-41 and SBA-15) is also a superior sorbent because it is also fully regenerable and has the largest pore size, allowing less diffusion resistance. The results show that Cu(I)Y, CuCl/SBA-15, and CuCl/MCM-41 are the most promising sorbents for natural gas desulfurization.10 Acknowledgment We thank Dr. Luis Amestica of the International Copper Association for helpful discussions. This work was supported by the International Copper Association. Literature Cited (1) Natural Gas Supply Association Web site. http://www.naturalgas.org/ overview/background.asp. (2) Speiget, J. G. Fuel Science and Technology Handbook; Marcel Dekker: New York, 1990. (3) Reisenfield, F. C.; Kohl, A. L. Gas Purification, 2nd ed.; Gulf Publishing: Houston, 1974. (4) Israelson, G. Results of Testing Various Natural Gas Desulfurization Adsorbents. Mater. Eng. Perform. 2004, 13, 282. (5) King, D. L.; Birnbaum, J. C.; Singh, P. Sulfur Removal from Pipeline Natural Gas Fuel: Application to Fuel Cell Power Generation Systems. Pacific Northwest National Laboratory. Fuel Cell Seminar, Palm Springs, CA, Nov 18-21, 2002. (6) Satokawa, S.; Kobayashi, Y.; Fujiki, H. Adsorptive Removal of Dimethylsulfide and t-butylmercaptan from Pipeline Natural Gas Fuel on Ag Zeolites Under Ambient Conditions. Appl. Catal., B 2005, 56, 51. (7) Alptekin, G. O. Sorbents for Desulfurization of Natural Gas, LPG and Transportation Fuels. Sixth Annual SECA Workshop, Pacific Grove, CA, April 21, 2004. (8) Yang, R. T.; Herna´ndez-Maldonado, A. J.; Yang, F. H. Desulfurization of Transportation Fuels with Zeolites under Ambient Conditions. Science 2003, 301, 79.
(9) Bauman, S. Chemistry Highlights 2003. Chemical & Engineering News; American Chemical Society: Washington, DC, Dec 23, 2003. (10) Yang, R. T.; Wang, Y. H.; Amestica, L. Selective Sorbents for Natutal Gas Desulfurization. U.S. Patent Application 60/984,602, 2007. (11) Herna´ndez-Maldonado, A. J.; Yang, R. T. Desulfurization of Diesel Fuels by Adsorption via π-Complexation with Vapor-Phase Exchanged Cu(I)-Y Zeolites. J. Am. Chem. Soc. 2004, 126, 992. (12) Baes, C. F.; Mesmer, R. E. The Hydrolysis of Cations; Wiley: New York, 1976. (13) Hutson, N. D.; Reisner, B. A.; Yang, R. T.; Toby, B. H. Silver Ion-Exchanged Zeolites Y, X, and Low-Silica X: Observations of Thermally Induced Cation/Cluster Migration and the Resulting Effects on the Equilibrium Adsorption of Nitrogen. Chem. Mater. 2000, 12, 3020. (14) Xie, Y. C.; Tang. Y. Q. Spontaneous Monolayer Dispersion of Oxides and Salts onto Surfaces of Supports: Applications to Heterogeneous Catalysis. AdV. Catal. 1990, 37, 1. (15) Cai, Q.; Lin, W. Y.; Xiao, F. S.; Pang, W. Q.; Chen, X. H.; Zou, B. S. The Preparation of Highly Ordered MCM-41 with Extremely Low Surfactant Concentration. Microporous Mesoporous Mater. 1999, 32, 1. (16) Zhao, D. Y.; Huo, Q. S.; Feng, J. L.; Chmelka, B. F.; Stucky, G. D. Nonionic Triblock and Star Diblock Copolymer and Oligomeric Surfactant Syntheses of Highly Ordered, Hydrothermally Stable, Mesoporous Silica Structures. J. Am. Chem. Soc. 1998, 120, 6024. (17) Yang, F. H.; Herna´ndez-Maldonado, A. J.; Yang, R. T. Selective Adsorption of Organosulfur Compounds from Transportation Fuels by π-Complexation. Sep. Sci. Technol. 2004, 39, 1717. (18) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; et. al. Gaussian 03, revision C.02; Gaussian, Inc.: Wallingford, CT, 2004. (19) Cerius2, version 4.6; Accelrys: San Diego, CA. (20) Hay, P. J.; Wadt, W. R. Ab initio Effective Core Potentials for Molecular Calculations. Potentials for the Transition Metal Atoms Sc to Hg. J. Chem. Phys. 1985, 82, 270. (21) Hay, P. J.; Wadt, W. R. Ab initio Effective Core Potentials for Molecular Calculations. Potentials for Main Group Elements Na to Bi. J. Chem. Phys. 1985, 82, 284. (22) Hay, P. J.; Wadt, W. R. Ab initio Effective Core Potentials for Molecular Calculations. Potentials for K to Au Including the Outermost Core Orbitals. J. Chem. Phys. 1985, 82, 299. (23) Gordon, M. S.; Cundari, T. R. Effective Core Potential Studies of Transition Metal Bonding, Structure, and Reactivity. Coord. Chem. ReV. 1996, 147, 87. (24) Takahashi, A. New Sorbents for Purification and Bulk Separation by π-Complexation. Ph.D. Dissertation, University of Michigan, Ann Arbor, MI, 2002. (25) Herna´ndez-Maldonado, A.J. Desulfurization and Denitrogenation of Liquid Fuels via π-Complexation. Ph.D. Dissertation, University of Michigan, Ann Arbor, MI, 2004. (26) Satokawa, S.; Kobayashi, Y. Adsorbent for Removing Sulfur Compounds from Fuel gases and Removal Methods. U.S. Patent 6,875,410, 2005. (27) King, D. L.; Li, L. Removal of Sulfur Components from Low Sulfur Gasoline Using Copper Exchanged Zeolite Y at Ambient Temperature. Catal. Today 2006, 116, 526.
ReceiVed for reView August 21, 2007 ReVised manuscript receiVed October 20, 2007 Accepted October 23, 2007 IE071145I
