5096
Ind. Eng. Chem. Res. 1997, 36, 5096-5102
Pd/δ-Al2O3 Catalysts for Isoprene Selective Hydrogenation: Regeneration of Water-Poisoned Catalysts Jen-Ray Chang,*,† Tzong-Bin Lin,‡ and Chil-Hung Cheng† Department of Chemical Engineering, National Chung Cheng University, Chia-Yi, Taiwan, ROC, and Chinese Petroleum Corporation, R. M. R. C., Chia-Yi, Taiwan, ROC
Water-containing isoprene was partial hydrogenated over δ-alumina-supported eggshell Pd catalysts with a weight-hourly space velocity (WHSV) of 18 h-1 (grams of feed per hour × grams of catalyst) at 40 °C and 30 atm (30.39 kg/cm2) total pressure. The isoprene conversion decreased with time on stream and became almost time invariant at about 50 h, with a value of 20%. The activity of the water-poisoned catalysts was only slightly recovered after the water-free isoprene was introduced; by this, the conversion was increased only to 24%. In contrast, by hydrogen treatment prior to the introduction of water-free feed, the water-poisoned catalyst was greatly reactivated. The improvement in activity was also influenced by temperature. The number of Pd sites available for CO chemisorption was increased and, thus, the catalyst activity was improved by a raise in temperature to 200 °C, then declined. At nearly steady state, the isoprene conversion by catalysts reactivated at 200 °C is about 93% of that by catalysts without waterpoisoning. The loss of the catalyst activity was thought to be caused by the Pd agglomeration during hydrogen reactivation. EXAFS (extended X-ray absorption fine structure) spectroscopy results indicated that the average Pd-Pd coordination number increased from 6.4 to 8.3 after reactivation of the catalyst with hydrogen at 400 °C. Introduction Pyrolysis gasoline, a byproduct of naphtha cracker, is unstable because of its highly active contents such as isoprene, cyclopentadienes, alkenyl aromatics, and other conjugated diolefins. Normally, δ-alumina-supported eggshell Pd catalysts are used in a hydrogenation process to stabilize pyrolysis gasoline by partially hydrogenating these active compounds (Griffiths et al., 1968; Lepage et al., 1987; Cheng et al., 1986). Besides bringing about the partial hydrogenation of the diolefins into high-octane gasoline blending components, olefins, the process also prevents such active contents from polymerizing into gums which may cause excessive pressure drop or catalyst decay downstream in the reactor or heat exchanger (Derrien, 1986; Lepage et al., 1987). If the pressure drop is too high, the whole unit must be shut down. In the previous paper (Chang and Cheng, 1997), we reported that water, introduced in the pyrolysis gasoline by foaming of the oil/water separator before the hydrogenation process, influences the catalytic properties of the Pd catalysts. Experimental results indicated that water-poisoning suppresses the hydrogenation of isoprene and inhibits the isomerization of 2-methyl-1butene and 3-methyl-1-butene to 2-methyl-2-butene, the main feedstock of the tert-amyl methyl ether (TAME) process. In this research, an attempt was made to regenerate the water-poisoned catalysts. Isoprene was chosen as the representative compound for the study of the selective hydrogenation of pyrolysis gasoline. The Pd catalysts were poisoned by carrying out the reaction with water-containing feed in a fixed bed reaction system with operation conditions of 40 °C and 30 atm (30.39 kg/cm2), similar to the conditions of commercial * Author to whom correspondence should be addressed. Phone: 886-5-272-0411 ext. 6239. Fax: 886-5-272-1206. Email:
[email protected]. † National Chung Cheng University. ‡ Chinese Petroleum Corporation. S0888-5885(97)00371-0 CCC: $14.00
plants (Jong et al., 1984). The water-poisoned catalysts were reactivated by hydrogen. The effects of hydrogen reactivation were examined by comparing the catalytic properties of the fresh, the water-poisoned, and the hydrogen-reactivated catalysts. The structural change induced by the hydrogen reactivation was characterized by extended X-ray absorption fine structure (EXAFS), which has been found to be useful for the characterization of supported metal catalysts, including sulfurpoisoned catalysts (Vaarkamp et al., 1992; Chang et al., 1997). Experimental Section Material and Catalyst Preparation. The δ-Al2O3 support was made by calcining γ-Al2O3 (A2U, Osaka Yogyo, particle size ∼2 mm) at 1000 °C for 6 h. The bulk density of the resulting δ-Al2O3 is 0.68 g/cm3. The BET surface area, mean pore diameter, and pore volume are 82.4 m2/g, 104.5 Å (1.045 × 10-8 cm), and 0.570 cm3/ g, respectively, determined with an Omnisop 360 analyzer. Preparation of the eggshell Pd catalysts has been described previously (Lin and Chou, 1994 and 1995). The support, impregnated with Pd(CH3COO)2, was filtered to remove solvent and then was calcined at 350 °C for 6 h. The catalyst contained 0.2 wt % Pd, measured by inductive plasma optical emission spectroscopy using a Jarnell-Ash 1100 instrument. Two different kinds of feed were used in this study. The isoprene feed was prepared by mixing 10 wt % isoprene (Merck) in toluene (99.22% purity, No. 4 reforming plant, Chinese Petroleum Corp., Taiwan, ROC). The pyrolysis gasoline feed was obtained from No. 4 naphtha cracking plant, and its physical and chemical properties are shown in Table 1. Catalytic Performance. As reported in the previous paper (Cheng et al., 1997), the catalytic performance tests were carried out in a continuous downflow fixedbed reactor with an inside diameter of 2.2 cm and an inside volume of 94.0 mL. The reactor was heated by a © 1997 American Chemical Society
Ind. Eng. Chem. Res., Vol. 36, No. 12, 1997 5097 Table 1. Physical and Chemical Properties of Pyrolysis Gasoline property
value
specific gravity diene value (by UOP 326/82) bromine no. total sulfur (ppm) existing gums (mg/100mL)
0.83 22 64 360 4-10
water-bath circulator, and the reaction temperature was monitored with a sensor in the center of the catalyst bed. The reactor was packed with 1.5 g of catalysts mixed with inert SiO2 (Merck) in a ratio of 1:20 by volume. A gradient packing method was used to minimize bypassing effects (Satterfield, 1975). The catalyst samples were reduced with flowing hydrogen at 12 L(NTP)/h hydrogen at 100 °C and 30 atm (30.39 kg/cm2) for 10 h, and then the hydrogenation reaction was carried out with a weight-hourly space velocity (WHSV) of 18 h-1 (grams of feed per hour × grams of catalyst), at 40 °C, 30 atm (30.39 kg/cm2), and a H2/oil molar ratio ) 2.262. For the water-poisoning tests, feed and water were pumped separately into the reaction system and mixed in a tube mixer (Koflo Corp., USA), and then the mixture was preheated to 40 °C before entering the catalyst bed. The reaction products were trapped by a condenser at -5 °C and analyzed by gas chromatography (Hewlett-Packard model 5890 A, FID model, equipped with a Petrocol DH 150, 150 m × 0.25 mm i.d., 1.0 µm phase film capillary column, and a SP4270 data processor) for the isoprene feed and by UOP 326/82 analytical method for the pyrolysis gasoline feed. To determine if the activity loss by water-poisoning can be compensated by a change of reaction temperature, after the conversion became time invariant (pseudo steady state), the reaction temperature was elevated from 40 to 60 °C and finally to 80 °C. For comparison, two regeneration methods were usedsone by the introduction of water-free feed, and another by treating the catalysts with flowing hydrogen at 12 L(NTP)/h hydrogen, 30 atm (30.39 kg/cm2), and temperatures ranging from 100 to 400 °C for 2 h prior to the introduction of water-free feed. Those waterpoisoned catalysts regenerated by hydrogen treatment are noted as hydrogen-reactivated catalysts. The effects of both regeneration methods were investigated by examining the activity and selectivity of isoprene hydrogenation. CO Chemisorption. A quartz tube was packed with about 1.0 g of catalyst sample. The catalyst sample was then reduced under the same operation conditions as those used in the catalytic performance test, except at 1 atm (1.013 kg/cm2). After the system became steady (20 mL/min He flow rate and 35 °C), consecutive 0.1 mL pulses of CO were injected into the catalyst bed with He carrier gas until none of the pulse was chemisorbed. The amount of chemisorption was then calculated by summing up the proportions of all pulses consumed. For the water-poisoned catalyst sample, before CO titration, the reduced catalyst sample was treated by flowing water-saturated N2 gas at 1 atm (1.013 kg/cm2) and 40 °C for 0.5 h. To investigate the effects of hydrogen-reactivation temperature on Pd dispersion, the water-poisoned catalyst was treated with flowing hydrogen at temperatures ranging from 100 to 400 °C for 2 h. X-ray Absorption Spectroscopy. The X-ray absorption measurements were performed on X-ray beam-
Figure 1. Effects of water-poisoning on the diene value of the C5 fraction of pyrolysis gasoline catalyzed by eggshell Pd/δ-Al2O3 catalysts at T ) 40, 60, 80 °C, P ) 30 atm (30.39 kg/cm2), WHSV ) 18 h-1, and H2/pyrolysis gasoline (mol) ) 2.262 (O: without water-poisoning; 4: with 5% water-poisoning).
line X-11A of the National Synchrotron Light Source (NSLS) at Brookhaven National Laboratory with a storage ring energy of 2.5 GeV and a beam current between 150 and 250 mA. A Si(111) double-crystal monochromator was used for energy selection, and it was detuned 20% at Eo + 50 eV to suppress higher harmonic radiation; resolution, ∆E/E, was estimated to be 2.0 × 10-4. The monochromator was scanned in energy from 200 eV below the palladium K absorption edge (24 350 eV) to 1200 eV above the edge. The waterpoisoned catalyst sample was prepared by chemisorption of water on the reduced catalysts, as described in the previous section. The hydrogen-reactivated catalyst sample was prepared by treating the water-poisoned catalysts with flowing hydrogen at 400 °C for 2 h. Samples were prepared and handled with exclusion of air on a double manifold Schlenk vacuum line and measured in a transmission mode at liquid nitrogen temperature. Results and Discussion The effects of water in the feed on the catalytic performance of the Pd catalysts were investigated by comparing the diene value (number of grams of iodine equivalent to the amount of malic anhydride that react with 100 g of sample; analytical method UOP 362/82) of the reaction products with the diene values of the feed of pure and water-containing pyrolysis gasoline. As shown in Figure 1, at the start of the run, significant catalyst deactivation was observed for both tests, while, in the near steady state, the diene value for the watercontaining feed was about 6.7, about 2 times that observed for the test with pure feed. Increased reaction temperature leads to an increase in conversion; however, increased reaction temperature also promotes the formation of gums. Since gums formed from the polymerization of highly reactive diolefins contained in pyrolysis gasoline not only cause difficulty in the separation of C5 and C9+ from pyrolysis gasoline but also increase the pressure drop of the reactor (see Table 2), the inlet temperature of the reactor for the selective hydrogenation of pyrolysis gasoline can not be higher than 120 °C. For the water-containing feed, the decrease of diene value of the reaction products with the elevation of reaction temperature is much lower than that for pure feed (Figure 1). Thus, compensating for the activity loss of the water-poisoned catalyst by
5098 Ind. Eng. Chem. Res., Vol. 36, No. 12, 1997 Table 2. Operation Conditions of Selective Hydrogenation of Pyrolysis Gasoline values
parameters reactor temperature (°C) start of run (SOR) end of run (EOR) reactor pressure (kg/cm2) LHSV (h-1) product specifications C5 fraction existing gums (mg/100 mL) diene value C9 204.4 °C fraction final boiling point (°C) existing gums (mg/100 mL)
inlet 60 120
outlet 140 180 30
3.5 4 max 3 max 204.4 max 4 max
Figure 2. Effects of water-poisoning on (A) total isoprene conversion, (B) the tendency to form isopentenes during isoprene conversion, and (C) the tendency to form 2-methyl-2-butene during isoprene conversion; conditions are T ) 40 °C, P ) 30 atm (30.39 kg/cm2), WHSV ) 18 h-1, and H2/isoprene (mol) ) 2.262 catalyzed by eggshell Pd/δ-Al2O3 catalysts (]: water-free feed; 4: 5% watercontaining feed; O: 20% water-containing feed; 3: water-free feed).
temperature elevation is impractical, and regeneration of the catalyst is necessary. For the process of selective hydrogenation of pyrolysis gasoline, normally, the catalysts are deactivated by gum formation and are reactivated by off-site or in situ oxidative regeneration. However, for water-poisoned catalyst deactivation, reductive regeneration is considered to be effective, since it can be performed in situ just by gradually decreasing flow of the feed, maintaining the flow of hydrogen, and adjusting the temperature. Regeneration of Water-Poisoned Catalysts. The total conversion of isoprene for pure feed and watercontaining feed are shown as a function of time on stream (Figure 2). As expected, the test with watercontaining feed exhibited lower initial conversion and underwent deactivation much more rapidly than the
other. At time invariant, the isoprene conversion was about 20% for the test with water-containing feed, in contrast to 80% for pure feed. As water concentration was increased from 5% to 20% by volume, no further catalyst deactivation was observed. Water-poisoning is not the only cause of the catalyst deactivation for the test with water-containing feed. However, as shown in Figure 2A, the conversion decreases only about 7% for water-free feed while it decreases about 70% for water-containing feed. The results indicate that, with the presence of water, the catalyst deactivation is caused mainly by water-poisoning. In the absence of water, gum formation is the main reason for catalyst deactivation. As shown in Figure 7 of our previous paper (Chang and Cheng, 1997), for the supported-palladium catalyst, two CO adsorption bands (located at 2066 and 1911 cm-1) were observed and were assigned as the terminal and the bridging CO ligand, respectively (Little, 1966). For the water-poisoned Pd catalyst, the νCO absorption bands for both the terminal and bridging ligands shift to higher frequency. This result suggests that water molecules are adsorbed on the Pd clusters. The adsorbed water removes electrons from Pd clusters, thereby decreasing the electron density of the Pd clusters. A decrease of the electron density on the Pd clusters lessens backbonding from Pd to the CO π* orbital, resulting in a shift of νCO to higher frequency (Cotton and Wilkinson, 1988). Inferred from FFT-IR results, the water-poisoning catalyst deactivation is considered to be caused mainly by the adsorption of water on the Pd clusters. Thus, we might expect that the water-poisoned catalysts could be reactivated by the desorption of the adsorbed water after feed was freed from water contamination. Contrary to our expectation, perhaps because the temperature (40 °C) was too low to desorb the adsorption water, the catalytic performance was only slightly improved (Figure 2A-C). Water-poisoning of the catalysts is caused by the irreversible adsorption of water molecules on active sites resulting from the foaming of the oil/water separator. Thus, these results indicate that, without further treatment of the water-poisoned catalysts, the catalyst activity cannot be recovered, even when the foaming problem has been solved. Hydrogen reactivation, a thermal treatment carried out in a reducing atmosphere to remove the adsorption water from the catalysts and to reduce oxidized palladium, leads to the reactivation of the catalysts. However, high-temperature hydrogen reactivation may also lead to palladium agglomeration. As shown in Figure 3A, the activity of the water-poisoned catalysts is evidently increased with hydrogen-reactivation temperature to the maximum of 200 °C. At nearly steady state, the isoprene conversion of the 200 °C hydrogenreactivated catalyst is about 93% of that of the catalyst without water-poisoning. The performance of the waterpoisoned catalysts can not be improved further by elevating hydrogen-reactivation temperature; experimental results indicated that the Pd dispersion and the isoprene conversion are inversely proportional to the hydrogen-reactivation temperature at temperatures higher than 200 °C (Figure 3A and Table 3). In the previous paper (Chang and Cheng, 1997), we reported that water-poisoning suppresses the hydrogenation of isoprene and the double-bond migration of
Ind. Eng. Chem. Res., Vol. 36, No. 12, 1997 5099
Figure 4. Raw EXAFS data for the water-poisoned catalyst (solid line) and the 400 °C hydrogen-reactivated catalyst (dotted line).
Figure 3. Effects hydrogen-reactivation temperature on (A) the isoprene total conversion. (B) the tendency to form isopentenes during isoprene conversion, and (C) the tendency to form 2-methyl2-butene during isoprene conversion; conditions are T ) 40 °C, P ) 30 atm (30.39 kg/cm2), WHSV ) 18 h-1, and H2/isoprene (mol) ) 2.262 catalyzed by eggshell Pd/δ-Al2O3 catalysts (]: fresh catalyst with water-free feed, 4: fresh catalyst with 5% watercontaining feed, O: 100 °C hydrogen-reactivated catalyst with water-free feed, 3: 200 °C hydrogen-reactivated catalyst with water-free feed, 0: 300 °C hydrogen-reactivated catalyst with water-free feed, ×: 400 °C hydrogen-reactivated catalyst with water-free feed). Table 3. Palladium Dispersion of the Fresh, Water-Poisoned, and Hydrogen-Reactivated Catalysts type of catalysts
palladium dispersion, (CO/Pd)
fresh catalyst water-poisoned 100 °C hydrogen-reactivated 200 °C hydrogen-reactivated 300 °C hydrogen-reactivated 400 °C hydrogen-reactivated
0.460 0.370 0.384 0.414 0.415 0.383
2-methyl-1-butene and 3-methyl-1-butene to 2-methyl2-butene. However, it slightly facilitates the formation of isopentenes. The increased tendency to form isopentenes was thought to be caused by the increasing affinity between Pd clusters and isoprene, and the decreased tendency to form 2-methyl-2-butene was caused by the lack of β-hydride. As shown in Figure 3B,C, the tendency toward the formation of isopentenes and 2-methyl-2-butene is only slight different (within experimental errors) for the fresh and the 300 °C higher hydrogen-reactivated catalysts, suggesting that the adsorpbed water was almost completely removed. Morphology of Pd Clusters on the Water-Poisoned and the Hydrogen-Reactivated Catalysts. X-ray absorption data from three scans of each sample were averaged, and then the preedge and background
Figure 5. Imaginary part and magnitude of Fourier transform (k3-weighted, ∆k ) 4.3-14.0 Å-1, Pd-Pd phase- and amplitudecorrected) of the raw EXAFS data for the water-poisoned catalyst (solid line) and the 400 °C hydrogen-reactivated catalyst (dotted line).
were subtracted. Each resulting spectrum was divided by the edge height to obtain EXAFS functions (van Zon, 1988). The comparison of EXAFS spectra for the waterpoisoned and 400 °C hydrogen-reactivated catalysts is shown in Figure 4. The raw EXAFS data for the samples have a signal to noise ratio >40 (The noise amplitude was determined at k ) 14 Å-1, and signal amplitude was determined at k ) 4 Å-1). Prior to the detailed EXAFS data analysis, k3weighted Pd-Pd phase- and amplitude-corrected Fourier transforms were determined for the EXAFS functions (4.3 < k < 14.0). The Fourier transforms provide the qualitative information about the influence of hydrogen reactivation on the structure of the supported Pt clusters. As shown in Figure 5, the Pd-Pd phaseand amplitude-corrected Fourier-transformed EXAFS functions of the water-poisoned and hydrogen-reactivated catalyst sample show that the peaks corresponding to the first metal-metal shell (at about 2.8 Å) and higher shells are at the same positions, while the peaks for the hydrogen-reactivated catalyst sample are higher than those of the water-poisoned one. Since the intensity of the Fourier-transformed EXAFS functions is proportional to the coordination number of Pd (Kamper, 1988), these results suggest that a remarkable agglomeration of Pd clusters occurred during hydrogen reactivation, while there is no significant difference in Pd-Pd bond distance between these two samples. Detailed EXAFS Analysis. A k2-weighted Fourier transformation without correction was performed on the EXAFS function over the range 3.0 < k < 14.0 Å-1 for the water-poisoned catalyst sample and over the range 3.1 < k < 15.4 Å-1 for the hydrogen-reactivated one. The major contributions were isolated by inverse Fourier transformation of the data in the range 1.14 < r