J . Phys. Chem. 1991, 95, 9966-9974
9966
Studles of the Hydrogen Held by Solids. 27. H,MoS, Takayuki Komatsu and W. Keith Hall* Department of Chemistry, Chevron Science Center, University of Pittsburgh, Pittsburgh, Pennsylvania 15260 (Received: February 26, 1991)
The sorption of hydrogen measured as H,MoS2 has been studied at 0.72 IPH2I 65 atm on both single-crystalline MoSz and microcrystalline material prepared by decomposition of (NH4)2MoS4in 14% HzS/H2. The surface area was varied from about 60 to 7 mz g-l by hydrogen treatments at various temperatures. Studies were also made with both sulfided and reduced molybdena-alumina catalysts. The (550-Torr) isobars obtained with all these materials were remarkably similar. They revealed a molecular adsorption at low temperatures in amounts smaller than, but commensurate with V,,the nitrogen monolayer capacity, and an activated sorption to an equilibrium state above 400 K. Uptakes on the supported catalysts x 50.045 were were about IO-fold larger than on the microcrystalline material. For MoS,, uptakes in the range 0.016 I recorded at 1 atm and these data were cross-checked by several methods. They were measured at P H 1 atm. Sampson et a1.4 studied the sorption at elevated pressures using a differential scanning calorimeter. They observed an increase in heat evolution with hydrogen pressure up to 65 atm. The data suggested that saturation was reached at about 50 atm. Their inelastic neutron spectra showed that at higher pressures a new peak appeared at about 400 cm-' in addition to the peak at 662 cm-' found at 1 atm. The latter peak was the same as that previously assigned to M0-S-H.j Thus, it was suggested4 that the 400-cm-' peak corresponded to the hydrogen sorbed into the bulk. These findings were confirmed by Jones et al.," who also studied the high-pressure sorption by inelastic neutron scattering spectroscopy. They proposed that at higher pressures (10-50 atm) hydrogen is sorbed as molecules at 473 K rather than as Mo-S-H groups as at lower pressures. Polz et al.12obtained a weak band at 529 cm-' from the Raman spectrum of MoS2 before hydrogen sorption. They assigned this band to (S-S)2- structure and suggested that it might react with H2 to form two SH-. However, their experiments could not show whether all or just part of the sorbed hydrogen was held in this way. The molecular orbital calculations of Anderson et aI.l4 indicated that the most stable chemisorption form of hydrogen is heterolytically held at edges of the MoS, crystal layers, resulting in the formation of S-H and Mo-H bonds. It was also shown (13) Jones, P.N.; Knozinger, H.; Langel, W.; Moyes, R. B.;Tomkinson,
J. SurJ Sei. 1988, 207, 159.
(14) Anderson. A. B.: AI-Saigh, Z. Y.;Hall, w.K. J . phys. them. 1988, 92,803.
that the energy barrier is low for the conversion of the latter (reductively) into more of the former. Ratnasamy and Fripiat'j measured the IR spectra of MoS2 obtained by the reduction of MoS3. They observed strong bands around 2500 cm-I, which were assigned to S-H stretching bands. These bands persisted even after evacuation at 673 K. This work has never been confirmed although several attempts have been made. Dianisj carried out TPD of sorbed hydrogen using MoS2 and potassium-doped MoS2. MoS2 gave a broad band at 500-600 K, while K/MoS2 gave a relatively sharp peak at about 473 K. They proposed that the broad peak obtained with MoS2 contains two peaks: One is the same as appeared in K/MoS2, and the other is centered at about 558 K, indicating the presence of a second form of bound hydrogen. Matemova9 adopted the reaction between SH groups and Ag+ in pyridine solution to determine the amount of S H present on MoS2. This reaction is an accepted way to measure SH groups in organic compounds. MoS2 prepared by decomposing ATTM was mixed with a pyridine solution of AgN03 a t 295 K. From the amount of Ag' lost in the solution and from the generation of HN03, the number of S H groups on MoS2 was estimated. This corresponded to x = 0.014. Although this value is low relative to most of the other reported x values, it does provide chemical evidence that at least a portion of H, is in the form of SH groups. Jalowiecki et aI.I1 reported that hydrogen in H,MoS2 reacted with isoprene at 423 K to form pentenes. When MoS2 was pretreated with H2 at 723 K, the amount of this active hydrogen reached a maximum of x = 0.012. These workers also found that the hydrogenation of pentadiene responded similarly and suggested that their data correlated with the number of edge sites having multiple coordinative unsaturation. They found, as did we,16that MoS2 after sulfiding was inactive for both isomerization and hydrogenation. Significant activity appeared only for S/Mo < 2. Interestingly, this is where p-type MoS2 changes into an n-type semiconductor.8 Blake et al. also carried out reactions in the presence of gas-phase hydrogen over MoS2 and Ho.oaMoS2.They found that the latter had a much higher activity than the former P.;Fripiat, J. J . Trans. Faraday SOC.1970,66, 2897. (16)Hall, W.K.;Schneider, R. L.;Goldwasser, J. Proc. 8th Int. Congr. Carol. (Berlin) 1984, IV-273. (15) Ratnasamy,
9968 The Journal of Physical Chemistry, Vol. 95, No. 24, 19
for butadiene hydrogenation6 and for hydrodesulfurization of tetrahydrothiophene;” earlier: they reported thiophene hydrodesulfurization was not affected, perhaps because a “ m o n value of H, was achieved under their reaction conditions. Wises reported that MoS2 (a slice of a single crystal) was more active for butene hydrogenation and for HDS of butyl mercaptan as a p-type (S/Mo > 2) than as an n-type semiconductor, in apparent contradiction to the findings of Jalowiecki et al.” The composition of the crystal was controlled by equilibration in varying H2S/H2 ratios. The foregoing summary of the literature reveals a variety of opinion concerning the nature of the hydrogen comprising H,MoS2, although most workers favor the view that at least a portion of H, is present as S-H groups. If this is formed by a reductive process, it could profoundly affect the catalytic behavior of the material. Variation exists both in the recorded experimental results and in their interpretation, suggesting that uncontrolled variables exist. Since microcrystalline MoS2 is a key ingredient of HDS catalysts, the need for a better understanding of these systems is obvious. The results of some experiments designed for this purpose are presented herein. A similarity in behavior is exhibited with several kinds of MoS,, with supported MoS2 (sulfided Mo/A1,03) and even with reduced Mo/AI2O3catalysts. Experimental Section MoS2 Preparations Studied. Microcrystalline MoS2 was prepared from ammonium tetrathiomolybdate (ATTM) by the method of Vasudevan and Weller.’* Ammonium paramolybdate (24 g) was dissolved in 30 mL of hot water. After 160 mL of aqueous ammonium hydroxide (NH, = 30 wt %) was added, pure hydrogen sulfide was bubbled through the solution. Its temperature gradually increased, and dark red crystals of ATTM precipitated at about 333 K. After cooling, the crystals were removed by filtration, washed with ethanol, and dried in vacuo at room temperature. 1R spectra of the ATTM obtained in this way showed no significant absorption between 800 and 1000 cm-I, where a Moo3 impurity would be evident. ATTM was then heated abruptly from 295 to 623 K (within 5 min) in a flowing mixture of 14% H2S in H2 to obtain microcrystalline MoS2. This flash heating has been reported to be an effective way to obtain high surface area MoS2.19 The MoS2 was then passivated in a flowing mixture of 10% O2 in He, first at 77 K for 30 min and then at 295 K for 30 min. It was stored under a dry argon atmosphere. Before each measurement, the passivated MoS2 was weighed in air and then resulfided in situ with 14% H2S in H2 (673 K, 1 h), and this was followed by reduction in dry H2 to 773 K (1 h). The MoS2 regenerated in this way was flushed with purified H e or N 2 (723 K, 1 .5 h). The sample lost about 5% of its original weight (after passivation) during these treatments. We describe such samples as “fresh MoS,”. The a values were obtained gravimetrically on a microbalance (Cahn Instruments Inc. C-1100). MoS, (-0.1 5 g) was oxidized in flowing 0, at 773 K to form M00,.I9 The a value was calculated from the weight loss of the sample. Two kinds of single crystals of MoS2 were supplied to us by Dr. R. R. Chianelli. One was a hexagonal-shaped natural single crystal of about IO-” diameter. This was cut into small pieces with a razor blade before pretreatments. The other was a powdered single crystal synthesized to be rich in edge planes. These materials were pretreated in the same way used with microcrystalline MoS2 to obtain fresh MoS,. The specific surface areas of the microcrystalline and the powdered single-crystalline MoS2 were determined by the BET method. When the area was too small to measure with N2at 77 K, an upper limit of 1 m2 g1was assigned to the material. An alumina-supported molybdenum catalyst, Mo (8 wt %)/ A1203,was prepared by the incipient wetness method using ammonium paramolybdate and y-A1203(Ketjen, 000-1, 5E). After (17) Blake, M. R.; Eyre, M.; Moyes, R. B.; Wells, P. B. Bull. SOC.Chim. Belg. 1981, 90, 1293. (18) Vasudevan, P. T.; Weller, S. W.J . Carol. 1986, 99, 235. (19) Kalthod, D. G.; Weller, S. W.J . Cutul. 1985, 95, 455.
Komatsu and Hall calcination in flowing oxygen at 773 K overnight, the catalyst was reduced with hydrogen at 773 K for 2 h and at 873 K for 2 h to e/Mo H 1.8. The reduced Mo/A1203 was sulfided in flowing H2S ( I4%)/H2 at 623 K for 3 h to obtain sulfided Mo/AI2O3. The surface areas of the parent alumina and the oxidized form of the prepared catalyst were 239 and 226 m2 g-I, respectively. Cases Used and Purification procedures. H2 (Linde, prepurified grade, >99.99%) and He (Linde, high-purity grade, >99.995%) were passed through a Matheson gas purifier (Model 452) to remove possible residual H 2 0 . N 2 (Linde, prepurified grade, >99.998%), H2S (Matheson, CP grade, >99.5%), and the mixture of H2Sand H2 (Linde, custom grade, H2S14%)were used without further purification. For isotope dilution experiments, D2 (Linde, C P grade, >99.5%) was diffused through a Pd alloy tube. Measurements of H Sorption. ‘Fresh MoS,” was used for this purpose at 453 K unless otherwise stated. The amount of sorbed hydrogen is expressed by x for H,MoS,. The x values were determined by five methods. (1) For isobars (550 Torr), the sorptions were measured volumetrically in an all-glass BET system. The pressure was kept constant by changing the system volume. These measurements were kinetically limited so a constant amount of time ( 1 h) was given at each temperature for ‘equilibration”. With this choice, the rate of change, dx/df, became insignificant at each temperature although at the lower temperatures equilibrium was not achieved. (2) A high-pressure microbalance (Cahn Instruments, Inc., C-1100) was used to determine x for the hydrogen sorption at superatmospheric pressures (1-35 atm). MoS, (-4 g) was put in a Pyrex sample bucket, which was suspended by a Pyrex hangdown wire into a stainless steel tube (id. = 25 mm). A Pyrex tube (0.d. = 24 mm) was inserted into the stainless steel tube to minimize corrosion of its inner surface by H2S. MoS2 was pretreated in situ at 1 atm to obtain fresh MoS,. The sample weight was recorded under static conditions to eliminate any effects of gas flow. Note that both methods 1 and 2 measure the gas sorbed (uptakes) from the initial fresh catalyst state at the pressure used in the experiment. Buoyancy corrections were made as necessary, e.g., when changing gas and/or temperature. They were calculated as follows. The weight of the sample including the bucket and hangdown wire was measured at 295 K under several pressures (1-15 atm) of He. The perfect gas law was used in the form AP = (RT/MV)AW,where AP and AWwere differences in pressure and weight from those at 1 atm. Now AW was plotted vs AP, giving a straight line passing through the origin of slope MV/RT. All factors of the slope are precisely defined numerically except V, which is the displacement volume of the MoS2 and its containment vessel. Once this was determined, corrections could be accurately calculated for changes in temperature or surrounding gas (molecular weight) at any pressure of interest here. Using quartz powder (having essentially no surface area) instead of MoS,, we measured the weight changes under conditions similar to those used in our experiments ( T = 295-673 K, P = 1-35 atm). The buoyancy changes obtained from these weight changes agreed well with those calculated from the displacement volume. The maximum possible error obtained was less than 30 pg. The weight of the uptake of hydrogen corresponding to x = 0.01 on the 4-g sample amounted to 250 pg. (3) Temperature-programmed desorption (TPD) of the sorbed hydrogen was also used for the measurement of the sorption under pressure (1-65 atm). MoS, (1.6 g) in a stainless steel tube (i.d. = 10 mm) was pretreated to obtain fresh MoS,. Hydrogen was sorbed at elevated pressure and at 453 K; the sample was then cooled in H2 at the same elevated pressure to 295 K. At this temperature the sorbed H was stabilized. The pressure was then let down, and the sample was purged with N2 until the concentration of hydrogen at the outlet (monitored by TCD) fell below the detectable level. TPD was carried out in flowing N2 from 295 to 673 K a t the heating rate of 10 K min-I. The x value was determined by integrating the TPD peak. The method was modified for samples with low x values. After hydrogen sorption, the sample was cooled in H2 as before. After
The Journal of Physical Chemistry, Vol. 95, No. 24, 1991 9969
The Hydrogen Held by Solids 0.12
0.10
4
1: x
0.2
Temperature, K
.;
Figure 1. Isobars of hydrogen sorbed by MoS, at 550 Torr: first, measured with increasing temperatures, 0;second, decreasing temperatures, @; third, increasing temperatures, 0; fourth, increasing tempera-
tures after resulfiding MoS, at 673 K, fifth, decreasing temperatures, 0; sixth, increasing temperatures, 0. One hour was allowed at each temperature for "equilibration". the gas phase was removed by evacuation at 295 K for 30 min, TPD was carried out in a closed system with a cold trap (77 K ) filled with 5A molecular sieves. The evolved Hz was collected in the trap from which it was recovered on warming, and it was measured using N, carrying gas as it passed through the TCD. (4) The isotope dilution method was used to determine whether the hydrogen sorbed and desorbed represented all or only part of that held by the MoS2. (Some bound hydrogen could have been inadvertently introduced during preparation and pretreatment.) Hydrogen sorption was carried out with MoSz (1.6 g) in a Pyrex tube, which was part of an all-glass circulation system also containing a TCD cell. This device contained two gold-covered tungsten filaments in the measuring cell and two tungsten filaments in the reference (D,) cell, which formed two legs of a Wheatstone bridge. The cells were kept at 324 K in a thermostated sand bath. The method of analysis was copied from Cheselske et aLZ0 A measured amount of Dz was introduced onto the sample at 573 K. The circulation of gas was carried out first with a cold trap (77 K) upstream of TCD for 2 h, second by bypassing the TCD with the trap warmed to 295 K, and third through TCD again with the cold trap. This procedure was intended to equilibrate all the hydrogen species in both the gas and solid phases. Raising the sample temperature did not effect further reaction. From the voltage change in the TCD output before and after the exchange was complete, the atom fraction H/(H + D)in the gas phase could be determined from a calibration chart and the x value of the initial H,MoS, was derived from this.20 (5) ' H N M R spectra were obtained on a Bruker MSL-300 spectrometer using a high-power proton probe. H,MoS, prepared in a Pyrex reaction vessel was transferred in vacuo into an attached side arm made from an NMR sample tube (0.d. = 5 mm). After evacuation at 295 K, the sample tube was sealed and put into the NMR probe. Wide-line IH NMR spectra were measured at 295 K at a frequency of 300 MHz with 90' pulses of 6 ps. The acquisition of 2000 (or 1000) scans was carried out with a recycle time of 4 s between scans. Integrated intensities were obtained, and quantitative analysis of the amount of H "seen" by the method was derived with use of solid hexamethylbenzene as a standard. The absolute accuracy of the method was estimated to be about 15% of the value obtained. Results Isobars of Hydrogen (550 Torr) sorbed by fresh MoS, were determined volumetrically, and the results are presented in Figure (20) Cheselske, F. J.; Wallace, W. E.; Hall, W. K. J . Phys. Chcm. 1959, 63, 505.
Temperature, K Figure 2. Isobar of hydrogen sorbed by sulfided Mo/A1203 (solid line) and by reduced Mo/Al20, (dashed line) at 550 Torr. As in Figure 1, 0 and A represent first isobars: 0 and A,second isobars; and 0 and A, third isobars.
I . The BET surface area and the a value were 55 mz g-l and 1.983, respectively. The system was allowed 1 h at each temperature to equilibrate. The sorption temperature was first increased from 77 to 598 K in a stepwise manner (first isobar) and then was decreased to 295 K (second isobar) and increased again (third isobar). The data suggest that the second and third isobars are near equilibrium while most of the points below 450 K on the first are kinetically limited. At 77 K in the first isobar (e), the hydrogen uptake corresponded to an x value of 0.085 calculated as H,McS,. This value was equivalent to a surface concentration of 2.9 X lOI4 H2 molecules cm-2, which is smaller than the N2 monolayer capacity, VU (6.2 X lOI4N, molecules cm-,). It may be supposed that this is a site-selective form of molecular adsorption of HZ.The uptake decreased to almost zero at 195 K, but again increased sharply above 300 K; the uptake began to decrease at about 480 K as expected from thermodynamic considerations. The third isobar reproduced the second. This sorption behavior is in qualitative agreement with previous After MoS, was resulfided at 673 K, a fourth isobar was obtained (dashed curve, Figure 1). It reproduced the first isobar, except for a slight decrease in the activated chemisorption above 400 K. When the sample was cooled to 77 K (fifth isobar), the H uptake exceeded the initial value (x = 0.084), increasing to 0.1 17. This low-temperature sorption was reversible as can be seen in the sixth isobar. The difference (0.1 17-0.084 = 0.033) is in fair agreement with the activated chemisorption. Isobars for both sulfided and reduced Mo/AI,03 catalysts are shown in Figure 2. These two isobars were remarkably similar to each other. Hence,it may be concluded that whatever processes are responsible for the sorption on the sulfided catalyst must also be occurring on the reduced (oxidic) form. Figures 1 and 2 are qualitatively similar. However, the activated sorption on both Mo/AI,OS catalysts occurred at higher temperatures than required for MoS2. The equilibrium region was reached above 700 K for the sulfided and about 800 K for the reduced catalysts, respectively; these findings may be compared to about 450 K with MIS,. When the sample temperature is lowered to 295 K (second isobars), the hydrogen uptakes increased to their maximum values of H/Mo = 0.47 (sulfided) and 0.73 (reduced), respectively. These values are more than 1 order of magnitude higher than that of MoS2 (X = H/Mo = 0.045). In order to determine the hydrogen sorption at elevated pressures, a microbalance was used. A typical experiment is outlined in Figure 3. The weight of fresh MoS, at 673 K in He atmosphere was defined as an initial weight ( Wo). When the sample was exposed to a high pressure of H2 (P> 1 atm) at 453 K, the weight increased to reach a stable value ( W , ) within 30 min.
9970 The Journal of Physical Chemistry, Vol. 95, No. 24, 1991
Komatsu and Hall
w2
Time
-
1
I
t
1
.c
0
Figure 3. Measurement of hydrogen uptake by microbalance. All weights are corrected for buoyancy: Wo, in He ( I atm) at 673 K (fresh in H2 (P> 1 atm) at 453 K; W2,in H2 (P> 1 atm) after MoS,); W,, cooling to 295 K; W,,in H2(1 atm) at 295 K; W,,in He (1 atm) at 295 K; W,,in He ( 1 atm) at 673 K.
X
0.02
wO
TABLE 11: Comparison of x Valws for H,MoS, Obtained by Different Methods at Hydrogen Pressure 01 1 atm and 453 K surface x of H,MoS, H2 pretreatment area CY of temp(K) (m2g-') MoS, MB@ TPD IDb 623 673 723 773 623 (fresh)c
55 43 32 17 55
1.983 1.960 1.956 1.953 1.983
0.044 0.038 0.030 0.019 0
0.045 0.040 0.029 0.018 0
0.041 0.028 0.016 2)" 0.021 (>6)"
"Surface coverage of H atoms (see text). bThe surface areas of these samples were immeasurably small by the BET (N2) method. I
20
I
I
40
I
I
J
60
Surface a r e a , m2g-' Figure 6. Relation between the amount of sorbed hydrogen and the surface area of MoS,. The equilibrium data (0)were taken from Figure 5 . The data (0 and 4) are those for 6.7 m2 g-I MoS, and the synthetic single crystal presented in Figure 7. The two sets of data do not match exactly for reasons shown in Figure 4.
TABLE I V Determination of x bv 'H NMR
H2 pretreatment
temp (K) 573 623
x of H,MoS, by NMR 0.056 (0.050)" 0.035 (0.040jo 0.044 (0.040)"
673
" From the data collected using
by TPD
0.067 0.058
0.052
x(NMR)/ x(TPD) 0.84 (0.75)" 0.60 (o.69j" 0.85 (0.77)"
1000 scans.
2 ppm
c e -*
32
20 40 Pressure, a t m Figure 7. Effect of hydrogen pressure on the amount of hydrogen sorbed by microcrystalline MoS, and by synthetic (4) and natural (6) single crystals of MoS2. The sorption temperatures were 453 K ( 0 )and 473 K (Qand 6). "0-
other hand, could determine values of x down to 0.001 with an error of no more than &lo%. In order to determine the sorption capabilities of MoS2 of very low surfaces areas, such measurements were made. The results for three low-area samples are compared in Figure 7 with the corresponding data for higher surface area MoS, (obtained by microbalance after quenching and flushing with He, ( W,- Wo)).These latter data should be compared with those shown in Figure 5 (the W, - Wodata). Note that the curves in Figure 7 tend to level at high pressure whereas those of Figure 5 do not. The microcrystalline MoS, having a surface area of 6.7 m2 g-' was prepared by pretreating MoS2 with H2 at 773 K for 13 h. The H, values for the two single crystals of MoS2, the one shaved from a natural single crystal of MoS2 ( 0 )and the other from a synthetic single crystal ( 0 )with a high ratio of edge/basal plane surface area, were also determined and found to be significant, the latter being nearly equivalent to the 6.7 m2 g-I sample. Both single crystals sorbed hydrogen though neither had a measurable surface area by the BET (N2) method. To handle the data, an upper limit of 1 m2 g-l was assigned to them. These data are shown as the bottom three curves of Figure 7 and are presented numerically in Table 111. It should be noticed that the synthetic crystal sorbed more hydrogen than the natural crystal and as much as the sample having an area of 6.7 m2 g-l. The x values at 35 atm (Table 111) for these single crystals were 0.005 (natural) and 0.021 (synthetic), respectively. The surface coverages with H atoms (shown in parentheses in Table 111 as 0 = H/S,) were calculated by dividing the amount of sorbed H atoms by the surface concentration of S atoms based on the basal plane concentration: 1.2 X atoms cm-2.21 They were found to be >2 (21) Hulliger, F. Srrucrural Chemisrry of Layer-type Phases; Levy, F., Ed.; D. Reidel: Holland, 1976.
80
40
0 PPm
-40
-80
Figure 8. IH NMR spectra of H,MoS, prepared by H2 pretreatment at several temperatures: (a) 573 K; (b) 623 K; (c) 673 K. See Table IV.
(natural) and >6 (synthetic) based on a surface area of 1 m2 g-l. These results, which are lower limits calculated for an upper limit of 1 m2 g-' for the surface areas, demonstrate that some sorption occurs internally in the crystals. The amount of sorbed hydrogen can also be determined by 'H NMR. All of the samples gave one broad peak centered a t -2 ppm. The spectra obtained with 2000 scans are shown in Figure 8 for the samples listed in Table IV where the x values derived from the peak areas may be compared with those obtained by TPD method. Two sets of N M R measurements were carried out to estimate the experimental error; one used 2000 scans for data acquisitions and the other 1000 scans. The data from the latter are indicated in parentheses. Three H,MoS, preparations were made by H2 pretreatments at different temperatures followed by hydrogen sorption at 1 atm. In all samples, 70-8096 of sorbed hydrogen measured by TPD was detected by NMR. The signals were nearly Gaussian in shape and were fairly narrow for solids (fwhm N 25 ppm). Hence, the chemical shift could be deduced with fair accuracy. It was -2 ppm. This value is consistent with H bound to S,Le., S H groups, or with nonacidic (alcoholic) OH groups. Discussion
A great variability exists in the values of x reported by various workers on preparations of various origins (Table I). These may be divided roughly into two groups: those with H uptakes too large to be accommodated on the BET surface area and those that could be, if every surface sulfur atom is able to hold an H atom. Our data for microcrystalline MoS2 formed from ATTM fell in the latter class whereas the data obtained from the shaved single crystals, in spite of their small H uptakes, fell in the former because of their very low surface areas.
9912 The Journal of Physical Chemistry, Vol. 95, No. 24, 1991 The assumption that nearly every surface S atom can hold an H atom at T > 300 K would seem unlikely. Such surfaces would tend to readily evolve H2S,but of course this occurs to a limited extent at slightly higher temperatures. The activated chemisorption must be atomic; it is unrealistic to suppose that H2 molecules could be held by physical forces at T > 300 K. We are therefore driven to conclude (as have others) that at least a portion of the uptake is internal, perhaps intercalated between the planes of the chalcogenide wafers at sites inaccessible to N2 molecules at 77 K. (The c-axis is known4 to expand as H2 is sorbed.) This hypothesis is supported by the data in Figure 6, where reasonable extrapolations of the activated chemisorptions to zero surface area yield nonzero intercepts. These data suggest that as much as one-third of the total uptakes are at locations unavailable to N2at 77 K. Data for the single crystals, while not quantitative, support this picture. In Figure 6, the data from Figure 7 for the synthetic single crystal at I , 15, and 35 atm are also plotted (b),assuming a surface area of 1 m2 g-l. These TPD values were nearly equal to the extrapolated values ( W , - Wo),suggesting that most of the sorbed hydrogen of this low-area materials is located in the bulk. The results for the 6.7 m2 g-' MoS, ( 0 )did not fit exactly with the rest of the data, but according to Jalowiecki and co-workers," this material has been recrystallized; possibly much of the sorption is internal. The data in Figure 7 and Table 111 showed significant differences between the x values of the two single crystals. The synthetic crystal, which has a much higher edge/basal plane surface area ratio, sorbed much more hydrogen at all pressures. This suggests an important role for the edge planes in the hydrogen sorption. The edge/basal plane area ratio in microcrystalline M a 2 would depend on preparation and pretreatment. Therefore, the great variability in the literature data in Table I may result partly from differences in these ratios, but also from the extent of the sulfur deficiency (a)as already demonstrated." Another factor is the pretreatment. In our microbalance experiments, a significant amount of sulfur desorbed as H2S during H2 treatment which followed sulfiding. If hydrogen sorption is carried out without this H, pretreatment or with Hz pretreatment at low temperatures, hydrogen can react with this sulfur to form H2S. As indicated by D i a n i ~ MoS, , ~ sorbed H2S in amounts comparable to H2; this would result in an overestimation of the x value when it is measured volumetrically from the consumption of H2. This may explain some of the variability shown in Table 1. The isobars shown in Figure 1 have some interesting features. Two distinctly different kinds of adsorption are occurring at high and low temperatures, respectively. At 77 K the adsorption was rapid, reversible, and surprisingly large, vis., about 50% of V ,(N,) on MoS,. It had most of the attributes of physical adsorption; desorption was essentially complete at 195 K. Yet the adsorption of the H, was much too large relative to V, to be simple physical adsorption. Note that the sulfided and reduced Mo/AI2O3catalysts (Figure 2) behaved identically as the MoS2 at 77 C T C 300 K except that the adsorptions were about 1 order of magnitude larger. The supported catalysts have been intensively investigated at this low t e m p e r a t ~ r e . ~From ~ - ~ ~these works it is known that at 77 K a small portion is chemisorbed on catalytically active sites that can be poisoned by the selective chemisorption of NO or 02.23-24 The remainder is thought to be held in a molecular precursor state where it experiences a substantial barrier to rotation.2s-26 This results in the higher enthalpy of adsorption required to explain the abnormally large adsorption reported (22) Millman, W. S.; Van Cauwelaert, F. H.; Hall, W. K. J . Phys. Chem. 1979.83. 2764.
(23) Cifillo, A. C.; Dereppe, J. M.; Hall, W. K. J . Caral. 1980,61, 170. (24) Millman, W. S.; Hall, W. K . Proc. 7th Inr. Congr. Catal. (Tokyo)
1980. 8-1304.
(2s) Van Cauwelaert, F. H.; Hall, W. K. Trans. Faraday SOC.1970,66, 454. (26) Van Cauwelaert, F. H.; Hall, W. K. J . Colloid Inferface Sci. 1972, 38, 138.
Komatsu and Hall herein. Rapid exchange occurs on the N M R time scale between these two species.,) The fresh MoS, may be expected to follow the same pattern. Above about 300 K the heterolytically chemisorbed hydrogen atoms diffuse onto remote sites where they are energetically stablei4as SH. Above about 400 K (MoS,, Figure 1) the rates of adsorption and desorption became sufficiently large to bring the system onto the equilibrium isobar. Now, when the system was returned to 77 K, it was discovered that an additional uptake (0.117-0.084 = 0.033) had occurred about equal to the chemisorption (0.034) above 300 K. In other words, the dissociatively chemisorbed hydrogen is held at locations remote from the molecular chemisorption sites that are occupied at 77 K. Thefirst isobars were taken with increasing temperature, the second on decreasing the temperature, and the third on raising the temperature again. The fact that the second and third yielded essentially reproducible results may be taken as evidence that the system is in equilibrium above about 350 K. As the temperature was lowered to room temperature, however, the rate of change, dxldt, became very slow. When the sample was cooled to 295 K (second isobar), the uptake reached the value of x = 0.045. The surface coverage of H atoms was then calculated to be 0.26 H/S,. Evidently there is sufficient surface capacity to hold a sorption of this size provided a substantial fraction of the sulfur atoms on the basal planes is covered with SH. It is, however, not easy to conceive that so large a fraction of the surface S atoms can be converted to S-H, especially if the uptake is homolytic and, as suggested by Polz and coworkers,', corresponds to the reaction of s;- into 2 SH-. On the other hand, if the uptake is reductive, each SH-formed would correspond to the reduction of one Mo4+to Mo3+(or, if a collective election picture is appropriate, the injection of one electron per H, into the conduction band formed by the ordered array of Mo4+ ions in the central planes of the chalcogenide wafers). Reductive uptake would thus be expected to produce easily detectable physical effects and changes in catalytic behavior. Such effects have been reported.2,8J1+24,27 Sulfided Mo/AI,O, catalysts may be regarded as submicroscopic MoS2 bound to the alumina ~ u r f a c e . ~The ~ ? structure ~~ of the reduced molybdena clusters held by alumina is less certain, but clearly a recrystallization has occurred and such systems have been sometimes characterized as "supported MOO," bound to the alumina The hydrogen isobars for an 8% Mo on alumina (sulfided or reduced) are shown in Figure 2; they are remarkably similar to each other. Moreover, they also resemble closely those of Figure 1, but with a couple of significant differences. The relatively large adsorption (Figure 2) at 77 K fell to near zero between 195 and 300 K. On the other hand, the uptakes were 1 order of magnitude larger, both at 77 K and on the equilibrium isobars. In addition, activated adsorption set in at a somewhat higher temperature than with fresh MoS,. For the reduced Mo/A1203, the sorption of H2at 77 K and 550 Torr (1 3 cm3 (NTP) of H, g-') was now only about 25% of V, (52 cm3(NTP)of N2 g-l), but the larger portion of the latter may have been on the uncovered portion of the alumina surface. Millman et al.,, studied the adsorption of H,and of D2 on Ketjen CK-300 alumina and on a molybdenaalumina catalyst (8% Mo) made from it. After reduction to Mo4+, the H2 adsorption was 1.1 X lOI4 Hz cm-, while that on the alumina base was 7.8 X lor3 H, cm-,. The data of Figure 2 were in fair agreement with these data; viz., about 1.5 X l o i 4 H2cm-, were adsorbed at 77 K. Additionally, the N M R study of Cirillo et al.23revealed that at (27) Valyon, J.; Engelhardt, H.; Kallo, D. Acra Chim. Hung. 1987, 124, 83. (28) Knhzinger, H. Proc. 9rh fnr. Congr. Caral. (Calgary) 1988, 5, 20. (29) Hayden, T. F.; Dumesic, J. A. J . Catal. 1987, 103, 336. Hayden, T. F.; Dumesic, J. A.; Sherwood, R. D.; Baker, R. T. K. fbid. 1987, 105, 299. (30) Parekh, B. S.; Weller, S. W. J . Catal. 1977,47, 100, Ibid. 1978,55, 58. (31) Liu, H.-C.; Huan, L.; Weller, S. W. J . Caral. 1980, 61, 282; fbid. 1980, 66, 65. (32) Hall, W. K. In Chemistry and Physics of Solid Surfaces VI; Vanselow, R., Howe, R., Eds.; Springer Verlag: New York, 1986; p 73 ff.
The Hydrogen Held by Solids least three types of adsorption occur at 77 K on reduced Mo/AI2O3 catalysts: a strongly and a weakly held molecular species and a dissociated species. The high-temperature "equilibrium" sorption must occur mainly on the molybdena portion of the surface as the adsorption of H2 on alumina is very small under these conditions. When Pt is added to alumina and spillover occurs, the amount of the sorption is nearly 1 order of magnitude lower33than on the equilibrium isobar. Nevertheless, spillover onto sites remote from the dissociation centers on the MoS, is suggested by the data because the dissociatively chemisorbed hydrogen does not interface with the lowtemperature adsorption. The equilibrium values of x found for the supported catalysts were from IO to 20 times larger than for the fresh MoS,, viz., about one H for every four S in sulfided catalyst and for every three 0 (of the postulated M a 2 ) in reduced catalyst. The specific surface areas of the bound molybdena or MoS, species would necessarily be very high (essentially molecular clusters) to accommodate all this as surface S H or OH. It is likely that hydrogen adsorbed at the crystal growth edges diffuses into the lattice at least for several layers. The zero surface area intercepts of Figure 6 would be consistent with this idea. In the case of the related chalcogenides NbS, and TaS,, definite crystal lattices have been established by neutron diffraction for H,NbS2 ( x = 0.76) and D,TaS2 (0.1 < x < 0.4).34 The high-temperature equilibrium sorption of hydrogen on reduced Mo/AI2O3has been known for many years as HR. It was first discovered as hydrogen missing from the mass balance35 in reduction-reoxidation cycles and then by direct observation of the reversible adsorption and desorption of H2.36 The first isobar of the reduced Mo/AI20, (Figure 2) above 295 K closely resembles the desorption profile of HR obtained by evacuating the catalyst at successive increasing temperatures after cooling in H2 from 781 to 295 K.36 HR was shown3' to poison olefin metathesis and to promote hydrogenation, but no one has yet been able to satisfactorily explain this phenomena. Now, the data of Figures 1 and 2 suggest that HR is part of a common phenomena that may occur in a variety of oxide and sulfide systems where it may play an important role both in catalysis and in the control or modification of the physical properties. In previous NMR36and IR3* studies of HR on reduced Mo/A1203 catalysts, only about half of the hydrogen in this species appeared in the respective spectra and this was identified as OH. Bonnelle and c o - w ~ r k e r have s ~ ~ described similar phenomena for the copper-chromium oxide system. In addition, they have studied the H,MoS, system" and this work deserves special comment. They reported an abrupt drop in surface area (from 47 to 7 m2 g-l) and a change in "texture" above 573 K where S was removed by reduction with H,. (Evidently S H groups were formed as intermediates, as H2S was evolved.) H,MoS, was formed in this process, and x and a were determined separately, the former by removal by reaction with isoprene at 423 K. Values of 0.002 Ix I0.012 were obtained, but usually x I0.009. Most interestingly, x passed through a maximum (0.012) at a reduction temperature of 723 K where a = 1.94, Le., just at the point where recrystallization with concomitant decrease in surface area occurred. These experiments afford a partial explanation of the great variability in x displayed in Table I. The authors note, however, "the quantities (x values) obtained ... are much lower than those reported by other authors." Comparison with our data in Tables I and I1 confirms this point but also shows that their values make (33) Conner, W. C. Private communication. (34) Riekel. C.; Reznik, H. G.: Schbllhorn, R.: Wright, C. J, J . Chem. Phvs. 1979. 70.. 5203. ._.. 135j Hail, W. K.; Massoth, F. E. J . Caral. 1974, 34, 41. (36) Cirillo, A. C., Jr.; Dollish, F. R.; Hall. W. K. J . Coral. 1980.62. 379. (37) Engelhardt, J.; Valyon, J. Hydrogen Eflects in Catalysis; Paal, Z., Menon, P. G., Eds.; Dekker: New York, 1988; p 565. (38) Millman, W. S.;Crespin, M.; Cirillo, A. C., Jr.; AMo, S.; Hall, W. K. J . Caral. 1979, 60. 404. (39) Aissi, C . F.; Daage, M.;,Wrobel, G.; Guelton, M.; Bonnelle, J. P. Appl. Carol. 1982. 3, 187. Jalowiecki, L.; Daage, M.; Bonnelle, J. P. Appl. Caral. 1985, 16. I .
The Journal of Physical Chemistry, Vol. 95, No. 24, 1991 9973 up an appreciable fraction of ours. Note also that our a values fell in the range 1.95 I a I1.98, whereas their range was 1.70 Ix I2.01. Therefore, as near as we can tell, our preparations were essentially identical with theirs, but (with the exception of the one preparation with area 6.7 m2 g-' given in Table 111) were obtained in the transition region where the "change in texture"" was just beginning, but not yet complete. The data suggest that their generally low x values may be the result of incomplete removal of H, by hydrogenation of isoprene. If so, then it may be that this is not a single species or that removal is limited by kinetic considerations. Marcq et a1.40reported a similar result for the molybdenum bronze, H,Mo03, formed a t 333 K. The similarity of the present systems with the hydrogen bronzes is unmistakable. The amount of hydrogen sorbed on the surface of MoS, (Hsurf) can be obtained from Figure 6 by subtracting the extrapolated intercepts (H-uptake in the bulk) from the total uptake. On this basis, the percentages of the total held by the surface at 1 atm were 86% (55 m2 g-l), 79% (32 m2 g-l), and 66% (17 m2 g-l), respectively, for the catalysts shown in Figure 5. These values are similar to the ratios of the amounts determined by N M R to the amounts determined by TPD (Table IV (0.600.85)). Possibly, the internal hydrogen is N M R inactive. The chemical shift of the NMR peak was about 2 ppm (Figure 8), which is in a range of 'H chemical shifts of SH in organic compounds: 1 to 4 p p ~ n . ~ This ' suggests that the hydrogen sorbed on the surface formed S H groups. The missing H (sorbed in the bulk) may be bound to paramagnetic Mo centers where the corresponding 'H N M R signal would be broadened beyond detection. This raises another interesting question: viz., if H, is sorbed reductively forming SH groups, Mo4+should be reduced to paramagnetic Mo3+. Why, then, is the 'H N M R line (from these S H ) not similarly affected? A tentative explanation is that the freed electron is not localized on an adjacent Mo3+ and/or the proton is free to move from sulfur to sulfur. According to Wise,8 the MoS, wafers are semiconductors having a filled 4d,1 band 1.75 eV below a conduction band made up from 4d, and 4 d , ~ ~levels. z The semiconductor becomes n-type on reduction as electrons begin to populate the conduction band. Thus, Mo3+ do not act as isolated centers containing unpaired electrons and consequently may lower, but not effectively eliminate, the 'H NMR signal. Of course, another possibility is that no reduction actually occurs. If, as suggested by Polz et al.,12 S2,- is reacted with H, to form 2 SH-, no electrons would be injected into the lattice but release of H2Sshould be observed as the temperature is raised. Further careful experimental work is needed to verify or invalidate this possibility. The present data suggest that x is very strongly surface area dependent but that a portion is held in the bulk, Le., that the solid crystals act as a reservoir for sorbed hydrogen. Moreover, the data from the single crystals (Figure 7) indicate that the amount so held is pressure dependent and those of Table I1 that it becomes reversible below T = 673 K (TPD). The data suggest that, with the preparations made from ATTM, H, is dissociated heterolytically at sites on the edges of the chalcogenide wafers but diffuse away to other nearby locations at temperatures above 350 K. Anion (sulfur) vacancies are necessary for this adsorption-desorption process; effectively, they are an important part of the sites where H2 is dissociated before it spills into the bulk reservoir. The way that H, is partitioned between surface and bulk sites is an important unanswered question that will depend, among other things, on the ratio of edge to basal plane area, i.e., whether the crystal habit is needle or platelike. The much larger x values for the supported sulfided (or reduced) catalysts reflect these points. In these instances, the surface area of the MoS, on A1203may be estimated to be -500 m2 g-I of MoS, or about 1 order of magnitude larger than the microcrystalline MoS2 Hence, we may
-
'
(40) Marq, J. P.; Wispenninck, X.; Poncelet, G.; Keravis, D.; Fripiat, J. J. J . Cafal. 1982, 73, 309. (41) Bobey, F. A . N M R Data Tables for Organic Compounds; Interscience: New York,1967.
9974
J. Phys. Chem. 1991, 95, 9974-9979
has associated high catalytic activity with ptype semiconductivity suppose that the observed IO-fold larger uptake of hydrogen is and found that this is much diminished by reduction or loss of mainly a surface phenomenon. sulfur which converts the system to n-type. This emphasizes a Jalowiecki et al.” have stated, ‘the H* liberated in this work common problem in the literature; Le., it is difficult to correlate is different in nature from the SH group as no H* is found on all of the work since many materials identified simply as MoS2 a completely sulfided solid.” We are dubious about this because are not identical; they differ in texture and overall composition coordinatively unsaturated centers are necessary for H2 dissoci(anion vs cation vacancies) depending upon source and preparation ation. If they do not exist when CY = 2.01, no sorption would be method. Impurities, particularly bound oxygen, may play an possible. They also studied hydrogenation of the pentadienes and important role as will the ratio of edge to basal plane area. interpreted their results on the basis of the Siegel as Coordinatively unsaturated centers present on the edges are bound occurring by reaction of hydrogen oxidatively adsorbed on moto be important in adsorption and catalysis. Some or all of these lybdenum sites having three coordinatively unsaturated positions. factors may be required to explain the variability in the data P r e v i o ~ s l y , ~we ~ *presented ~ ~ * ~ ~ our reasons for thinking that compiled in Table I. In the present work we have adopted the heterolytically adsorbed hydrogen functions in the hydrogenation view that, on the submicroscopic scale, all MoSz exists as simple of butadiene. Of course we do agree that multiple coordinative chalcogenide wafers but that variations in texture and habit may unsaturation is a necessary prerequisite for hydrogenation activity, occur on crystal growth. Complete elimination of oxygen is but we have not seen any conclusive evidence indicating that the difficult if not impossible, but we have used treatment with Siegel model is operative. In this connection, an interesting set H2S/H2, which is the most effective technique available for this of conflicting results should be noted. Jalowiecki et al.,” like purpose. Finally, we have deliberately introduced CUS by ourselves?2 have noted that reductive removal of sulfur from MoS2 treatment with pure H2 at elevated temperature and specified the is essential to turn on the catalytic activity for hydrogenation sulfur deficiency as MoS,. Still, more research is obviously needed and/or isomerization. Moreover, the activity has been shown to increase more than linearly with increasing extent of r e d u c t i ~ n ~ ~ * ’ ~to understand such experimental contradictions. by formation of coordinatively unsaturated sites (CUS), which Acknowledgment. Thanks are due to the National Science selectively chemisorb H2, NO, and 02.Wise,’a on the other hand, Foundation for support of this work under Grant DMR-9045228. Helpful discussions with Dr. R. R. Chianelli, R. B. Moyes, and (42) Sicgel, S. J. Caral. 1973, 30, 139. Siegel, S.; Outlaw, J.; Garti, N. H. Knozinger are gratefully acknowledged. Ibid. 1978, 52, 102.
The Nature of Rhenium in Silica-Supported Pt-Re Clusters: Synchrotron X-ray Diffraction Studies K. S. Liang,* F. Z. Chien: G. J. Hugbes, G. D. Meitzner, and J. H. Sinfelt Corporate Research Laboratory, Exxon Research and Engineering Company, Annandale, New Jersey 08801 (Received: April I , 1991; In Final Form: August 12, 1991)
Using synchrotron anomalous X-ray diffraction, we studied the structure of bimetallic Pt-Re catalysts with Si02as a support. Our results reveal the presence of Re with a face-centered cubic structure in these catalysts. We propose that this form of Re is grown on Pt by an epitaxial effect.
I. Introduction The study of matter in small particle forms is not only a fascinating area of research but also has important technological implications in catalysis.’ To directly observe the structure of small particles of the order of 10 A in size is found to be difficult. The main structural tools developed recently for such studies include high-resolution electron micr0scopy,2~scanning tunneling microscopy$ anomalous X-ray diffraction using synchrotron rad i a t i ~ nand , ~ extended X-ray absorption fine s t r u ~ t u r e . ~The *~ primary goal of this study is to extend our earlier synchrotron X-rav studies of Pt catalvstsS to a model bimetallic catalvst svstem: Pt-Re supported on SfOz. Bimetallic cluster catalysts have had a major impact in industrial catalysis, in particular in the production of aromatic hydrocarbons for automotive fuels.6 Extensive work has been directed to the understanding of the structural and electronic properties of these cluster^.^.^ Frequently the clusters are compositionally inhomogeneous due to segregation of one constituent to the surface. Therefore, obtaining a precise description of the cluster structure is often a formidable task. For such studies, X-ray techniques based on atomic core-level absorption are commonly employed (e.& EXAFS). To a lesser extent, the application of ‘Present address: Tamkang University, Taipei, Taiwan.
the anomalous X-ray diffraction techniaue for supported metal crystallites has also k e n t l y been demonskated on heterogeneous metal catalyst^.^*^*^ Although the materials used in this work differ from the Pt-Re reforming catalysts used commercially, they serve the purpose of exploring the interactions between Re and Pt at the atomic level. For the completeness of our discussions, X-ray absorption edge analysis was also employed in the characterization of some of the samples. A brief account of the main results of this study has already been reported in a previous paper.8 One of the main structural features revealed by this study is the presence of a ( 1 ) See,for example: Physics and Chemistry of Small Clusters; Jena, P., Rae, B. K,, Khanna, s. N., Eds.; Plenum: New York, 1987, ( 2 ) Marks, L. D.; Smith, D. J. Narure 1983, 303, 316. ( 3 ) Smith, D. J.; Petford-Long, A. K.;Wallenberg, L. R.;Bovin, J.-0. Science 1% 233,872. ( 4 ) Gam, E.; Sattler, K.; Clarke, J. fhys. Rev. Lett. 1988, 60, 1856. Humbert, A.; Dayez, M.; Sangay, S.; Chapon, C.; Henry, C. R.J. Vac. Sci. Techno/, 1990, A8.311, ( 5 ) Liang, K.S.;Laderman, S.S.;Sinfelt, J. H. J. Chem. Phys. 1987,86,
2352. ( 6 ) See, for example: Sinfelt J H Bimetallic Caralysts: Discweries, and Applicutiom; York, 1983. (7) Sinfelt, J. H.; Via, G. H.; Lytle, F. W. Card. Rev. 1984, 26, 81. (E) Liang, K. S.; Hughes, G. J.; Sinfelt, J. H. fhysica B 1989, 158, 135. ( 9 ) Samant, M. G.; Bergeret, G.; Meitzner, G.; Gallezot, P.; Boudart, M. J . Phys. Chem. 1988, 92, 3547.
0022-3654/9 1/2095-9974%02.50/0 0 199 1 American Chemical Society
