3868
J. Phys. Chem. 1981, 85,3868-3872
ground-state dye concentration. The second-order rate constant for the quenching of MB+(T1) by MB+, k , as determined from the linear dependence of kd on [Mb+l0 is 4.1 X lo7 M-ls-l in aqueous solution at pH 8.2 and 3.6 X lo8 M-ls-l in acetonitrile medium. Under these conditions, prompt production of W2+., monitored at 520 nm, is followed by its slower formation during the decay of triplet dye. That the slow process of formation of MB2+. is the result of net electron transfer in the ground-state quenching of MB+(T1) was confirmed by simultaneous formation of semireduced dye, MBH+.. The above value of k, for ground-state quenching of MB+(T1)in water is significantly less than k, for groundstate quenching of MBH2+(T1)which we have previously reported,l 11.3 X lo7 M-l s-l (pH 2, 0.01 M C1-). The two values of k, are not strictly comparable because of differences in charge type and ionic strength. The lower charge of MB+(T1)and the somewhat higher ionic strength of the solution used in the present work both render k, for quenching of MB+(T1)by MB+(So)larger than it would be under strictly comparable conditions and reduce the observed difference between k values for ground-state quenching of MB+(T1) and MhH2+(T1). The observed higher value of k, for ground-state quenching of MBH2+(T1)is consistent with quenching via reversible electron transfer,14 particularly because Eared for MBH2+(T1),$1.33 V, is greater14than Eored for MB+(T1), +1.21 V. Comparison of efficiencies of net electron transfer, F1,in ground-state quenching of the two states of protonation of the triplet was not possible because the (14)Ohno, T.; Lichtin, N. N. J. Am. Chem. SOC.1980,102,4636-43.
prompt formation of MB2+.from excited MB+ interferes with accurate evaluation of F1by our procedure. Scavenging by NzO of ea,- Produced by Electron Photoejection from Fully Excited Methylene Blue. Flashing of M solutions of MB+ in water a t pH 8.2 which had been saturated with NzO by bubbling for more than 20 min, [N20] 0.03 M,15 resulted in absorbance at 880 nm 100 ps after the flash which was about 25% less than was obtained under otherwise identical conditions in solutions which had been purged with argon. Absorption a t 520 nm did not differ significantly under the two conditions. In view of the high specific rate of reaction of e, with N20, 8.7 X lo9 M-l s-l,I6 incomplete suppression of formation of MBH+. appears to be anomalous. Competition by ground-state quenching of MB+(T1)is too slow to account for the observed result, and significant T1-T1reaction has not been observed. The result suggests that NzO-- persists sufficiently long before dissociating to Nzand OH. to react principally by electron donation to MB+(So) and/or MB+(T1).17 The fact that saturation with N 2 0 had no effect on the yield of MB2+.suggests that whatever OHis formed from NzO does not give MB2+-as the principal product of its reaction with MB+.
-
Acknowledgment. This work was sponsored by the U.S. Department of Energy under contract EY-76-S-02-2889, (15)“Handbook of Chemistry and Physics”, 48th ed.; Weast, R. C., Ed.; The Chemical Rubber Publishing Co.: Cleveland, OH, 1967;p B201. (16)A n w M.;Bambanek, M.; Ross, A. B. Natl. Stand. Ref. Data Ser. (US’., atl. Bur. Stand.) 1973,43, 14. (17) Asmus, K.-D.; Henglein, A.; Beck, G. Ber. Bunsenges. Phys. Chem. 1966,70,459-66.
Extended X-ray Absorption Fine Structure Study of Co-Mo Hydrodesulfurization Catalysts Bjerne S. Clausen, Henrik Tops~e,Roberto Candla, J ~ r g e nVllladsen, HaMor Tops 08 Research Laboratories, DK-2800 Lyngby, Denmark
Bruno Lengeler, IFF, KernforschungsanlageJijllch, D 5 170 Jullch 1, West Germany
Jens Als-Nlelsen, and Finn Chrlstensen Ris0 Natlonal Laboratory, DK-4000 RosklMe, Denmark (Received: Aprll9, 1981; In Final Form: July 21, 1981)
By analyzing the extended X-ray absorption f i e structure (EXAFS) of the Mo absorption edge, we have obtained structural information about both calcined and sulfided Mo/A1203and Co-Mo/A1203catalysts. The calcined catalysts show only one strong backscatterer peak in the radial distribution function, which indicates that molybdenum is present in highly disordered structures. When the catalysts are sulfided, an ordering of the resulting molybdenum-containing phase takes place, as evidenced by the observation of a contribution from a second coordination shell as well. From a comparison with EXAFS data for well-crystallized MoS2, it is concluded that the molybdenum atoms in the sulfided catalysts are present in MoSz-likestructures. Furthermore, from an analysis of the amplitudes of the peaks, it is concluded that the MoS2-likestructures are ordered in very small domains. The results show for both the calcined and the sulfided states that the surroundings of the molybdenum atoms are not greatly influenced by the presence of the cobalt promoter atoms. Introduction Hydrodesulfurization (HDS) catalysts, which are used for removal of sulfur from various fossil hydrocarbon fuels, usually consist of Mo and Co dispersed on a high surface area support such as q-dumina. In view of their industrial
importance, HDS catalysts have been extensively studied by a variety of experimental techniques, and a number of hypotheses have been presented to describe the structure of the catalyst in both its precursor (calcined)and ita active (sulfided) states. For example, for the active state of the
0022-3654/81/2085-3868$01.25/00 198 1 American Chemical Society
Structure of Co-Mo Hydrodesulfurization Catalysts
catalyst, it has been proposed that the molybdenum is present either as some type of molybdenum oxysulfide (see, e.g., ref 1-3) or as MoS2 with cobalt either intercalated in the MoSz4t6or present as a separate phase of C O & ~ Recent Mossbauer emission spectroscopy mea~urements’,~ on this catalyst system have given evidence for a structure different from those mentioned above, and it was proposed7 that this structure, which contains molybdenum, cobalt, and sulfur (in a “CWMWS” phase), mainly consists of single S-Mo-S slabs (the basic building units of MoS2) with cobalt most probably substituting for the molybdenum atoms in the structure. Although Mossbauer spectroscopy is ideal for studying the amorphous or microcrystalline nature of the active phases present in HDS catalysts, the technique only gives direct information on the location of the cobalt atoms. In order to obtain structural information on the molybdenum atoms in the catalysts, an investigation has been undertaken using extended X-ray absorption fine structure (EXAFS). This technique has recently attracted much attention in studies of catalysts since it allows a determination of local structural parameters like coordination numbers and interatomic distances, and it has therefore been used to arrive at a structural description of “X-ray amorphous” systems (see, e.g., ref 9-11). Experimental Section The preparation of the HDS catalysts has been reported previously,8and only a few details will be given here. The Mo/A1203(8.6% Mo) catalyst was prepared by impregnating q-A1203 (250 m2/g) with ammonium heptamolybdate followed by drying and calcining in air at 775 K for 2 h. The CwMo/A1203 (Co/Mo = 0.6 (atomic ratio)) catalyst was prepared by adding cobalt nitrate to the above Mo/A1203 catalyst, followed by drying and by calcining in air a t 775 K for 2 h. A reference material of well-crystallized MoS2 was obtained from Riedel-de Haen AG (surface area (BET) = 1.79 m2/g; crystallite dimension obtained from the X-ray diffraction (002) reflection: D(002) 650 A). The Mo/A1203and the Co-Mo/A1203 catalysts in the calcined state and the well-crystallizedMoSz were pressed into homogeneous self-supporting wafers, 1.125 in. in diameter, and mounted between two pieces of adhesive tape. The sulfided alumina-supported catalysts were studied in situ by placing wafers (16 mm in diameter) in a specially designed cell equipped with X-ray-transparent windows. Sulfiding of the samples was carried out in the cell in a gas mixture of H2/H2S(270 H2S) a t 675 K for 4 h after which treatment the cell was sealed off. The absorber
-
(1)G. 6. A. Schuit and B. C. Gates, AZChE J., 19, 417 (1973). An updated version of this article was given by B. C. Gates, J. R. Ratzer, and G. C. A. Schuit in “Chemistry of Catalytic Processes”, McGraw-Hill, New York, 1979, Chapter 5. (2) F. E. Massoth, J. Catal., 36, 164 (1975). (3) P. C. H. Mitchell and F. TrifirB, J. Catal., 33, 350 (1974). (4) R. J. H. Voorhoeve and J. C. M. Stuiver, J. Catal., 23,243 (1971). (5) A. L. Farragher and P. Cossee, Proc. Int. Congr. Catal., 5th, 1972, 1301 (1973). (6) B. Delmon, Am. Chem. SOC.,Diu. Pet. Chem., P r e p . , 22, 503 11 ~ 977) - -.,..
(7) H. Topme, B. S. Clausen, R. Candia, C. Wivel, and S. Msrup, J . Catal., 68, 433 (1981). (8) C. Wivel, R. Candia, B. S. Clausen, S. Merrup, and H. T o p s ~ eJ. , Catal., 68, 453 (1981). (9) J. H. Sinfelt, G. H. Via, and F. W. Lytle, J. Chem. Phys., 68, 2009 (1978). (IO) R. M. Friedman, J. J. Freeman, and F. W. Lytle, J. Catal., 55,lO (1978). (11) F. W. Lytle, G. H. Via, and J. H. Sinfelt in “Synchrotron Radiation Research”, H. Winick and S. Doniach, Eds., Plenum Press, New York, 1980, Chapter 12.
The Journal of Physical Chemistry, Vol. 85, No. 25, 198 1 3889
n
I
I
I
I
20000
202m
2OLOO
E(eV1 Flgure 1. K-edge absorption coefficient of Mo vs. X-ray photon energy for the calcined catalysts: (a) Mo/A1203 catalyst; (b) Co-Mo/Al2O3 catalyst. Edge structure shown in insert.
-
thickness, x, of the samples was chosen such that px 1 ( p is the linear absorption coefficient) on the high-absorption side of the edge. The EXAFS experiments were performed a t DESY in Hamburg using the synchrotron radiation from the DORIS storage ring and the EXAFS setup at EMBL.12 The X-rays, which were emitted by positrons in the storage ring with an energy of 4.6 GeV and a typical current of 10 mA, were monochromatized by two Si (220) single crystals. The beam intensity was measured, before (lo)and after (I) passing through the sample, by use of two ionization chambers fiiled with Ar. All spectra were obtained at room temperature at 2-eV steps extending up to 1000 eV above the K edge of Mo. Results The raw EXAFS data for the Mo/A1203 and Co-Mo/ A1203catalysts in the calcined state are shown in Figure 1, where px = In (&,/I) is plotted against photon energy. The EXAFS data analysis used here has been described in detail in ref 13. This procedure involved a background subtraction by means of cubic spline functions, multiplication of the EXAFS by a factor 12, and normalization by the jump height at the K edge. The structural environment of the Mo atoms in the two catalysts, as determined by Fourier inversion of the EXAFS data, is shown in Figure 2, where R is the observed radial distance from the absorbing Mo atom. The radial distribution functions, which are quite similar for both catalysts, show the presence of only one distinct backscatterer peak undoubtedly being due to oxygen. Applying the principle of transferability of phase shifts,14one can estimate a Mo-0 interatomic distance of -1.73 A by use of a scattering phase shift (12) The EXAFS spectrometer at EMBL was designed by J. Bordas and J. C. Phillips. (13) B. Lengeler and P. Eisenberger, Phys. Rev. B, 21,4507 (1980); P. Eisenberger and B. Lengeler, ibid., 22, 3551 (1980). (14) P. H. Citrin, P. Eisenberger, and B. M. Kincaid, Phys. Reu. Lett., 36, 1346 (1976).
3870
The Journal of Physical Chemistry, Vol. 85, No. 25, 1981 I
Clausen et al.
I
1
I
I
I
I
I
20750
21000
h
a
I
A
b
I
\
2
6
4
R
8
10
(A)
Figure 2. Relative magnitude of the Fourier transforms of the data in Figure 1 after multiplication by k and background subtraction: (a) Mo/AI,O,
catalyst; (b) Co-Mo/A120, catalyst.
of 0.44 for the Mo-0 atom pair. Figure 3a-c shows the EXAFS spectra for well-crystallized MoSz (a) and the sulfided catalysts, Mo/A1203(b) and Co-Mo/A1203 (c), the latter two being obtained in situ in the H2/HzS mixture. The corresponding Fourier inversions of the data are shown in Figure 4a-c. It is observed that the peak due to the first coordination shell is located at essentially the same R value (-1.90 A) and has the same amplitude for both the catalysts and the MoS2 compound. The location of the peak due to the second coordination shell ( R 2.86 A) is also found to be essentially identical in all three samples. However, the amplitude of this peak is substantially lower in the catalysts (Figure 4, b and c) than in MoS2 (Figure 4a). Another difference between the catalysts and MoS2 is that the radial distribution function for MoS2also seems to show peaks which may be identified as originating from coordination shells outside the second one. Furthermore, the results show that there is no essential difference between the data of the sulfided Mo/A1203catalyst and the sulfided Co-Mo/A1203 catalyst. In MoS2 molybdenum has S in the first coordination shell and Mo in the second shell, and thus the first peak in the Fourier transform of the MoSz sample is attributed to S and the second peak to Mo backscattering atoms. The smallest Mo-S and Mo-Mo interatomic distances in MoSz are 2.41 and 3.16 A, respectively. By use of these values and the peak positions from the Fourier transform of well-crystallized MoSz (Figure 4a), empirical phase shifts for Mo-S and Mo-Mo scattering can be determined. It is found that C ~ M & 0.51 A and aMo-Mo N 0.30 A. These values agree quite well with those published by Cramer et al.15
-
~~
(15)S.P.Cramer, K. 0.Hodgson, E. I. Stiefel, and W. E. Newton, J. Am. Chem. Soc., 100, 2748 (1978).
20000
20250
20500
E(eV)
Figure 3. K-edge absorption coefficient of Mo vs. X-ray photon energy for well-crystallized MoS, (a), sulfided Mo/AI,O, sulfided Co-Mo/AI2O3 catalyst (c).
catalysts (b), and
In order to separate the contributions from the different coordination shells, a region surrounding each of the two main peaks in R space was also transformed back into k space for the transforms shown in Figure 4. The results of this Fourier filtering procedure for the first and second coordination shells are shown in Figures 5 and 6, respectively. It is found that for both MoS2and the two sulfided catalysts the envelope of the EXAFS amplitudes due to backscatterer atoms in the first coordination shell (Fi re 5 ) has a maximum located at the same k value ( 4.5 The contribution to the EXAFS from the second shell (Figure 6) gives also a maximum which is located at essentially the same position (-9 A-1) for the sulfided catalysts and MoS2, but the amplitudes of the EXAFS oscillations for the catalysts are smaller compared to those for the MoSz sample, in accordance with the Fourier transforms. N
pl).
Discussion The present EXAFS data show that for the Mo/Al2O3 and Co-Mo/Alz03 catalysts in the calcined (oxidic) state their radial distribution functions have only a single strong backscatterer peak due to oxygen atoms surrounding the molybdenum. The absence of more distinct peaks suggests that a great deal of disorder exists outside the first coordination sphere of the molybdenum atom. Schuit and co-workers1 have suggested the presence of a ‘‘Moo3monolayer” structure on the surface of the alumina in calcined catalysts. However, such a structure would presumably give rise to some degree of order outside the first coordination shell which is not observed in the present study. This shows that a single well-defined molybdenum-containing compound does not exist. Rather, the
The Journal of Physical Chemistry, Vol. 85, No. 25, 198 1 3871
Structure of Co-Mo Hydrodesulfurization Catalysts I
I
I
w
l
I
1
03
r
a
X
W
-0 1
b
-0 3 L
I
I
5
7 5
I 10
k(A-') Flgure 5. Fourier filtered data of the first coordination shell of wellcrystallized MoS, (-. -), sulflded Mo/AI2O3catalysts (-- -), and sulfided Co-Mo/AI2O3 catalyst (-).
I
C
'
I
I
0.1
Y X
R (AI Figure 4. Fourier transforms of the data in Flgure 3 after multiplication by k and background subtractton: (a) well-crystalllzed MoS,; (b) sulfided Mo/A1,03 catalyst; (c) sulfided Co-Mo/AQ03 catalyst.
results suggest that in the calcined catalyst molybdenum is present as isolated ions, chains, or small disordered patches. The Mo-0 bond length is estimated to be -1.73 A, using the observed peak position and the empirical phase shift published by Cramer et al.15 This value is typical for Mo-0 distances observed for tetrahedral-coordinated Mo (e.g., Fe2(MoO4)Jand is also in accordance with the shorter ones of the Mo-0 distances in CoMo04 and Moo3, in which Mo is octahedrally coordinated. Evidence from other spectroscopic techniques (see, e.g., ref 16-18) indicates that Mo is present in both tetrahedrally and octahedrally coordinated forms in the calcined state of CoMo/A1203 catalysts. From the present results it is not possible to obtain direct information about the coordination number of the molybdenum atoms, but such information could be obtained from a more detailed EXAFS (16) N.Giordano, J. C. J. Bart, A. Vaghi, A. Castellan, and G. Martinotti, J. Catal., 36,81 (1975). (17) H.Knozinger and H. Jeziorowski, J.Phys. Chem.,82,2002(1978). (18) J. Medema, C.van Stam,V. H. J. de Beer, A. J. A. Konings, and D.C. Koningsberger, J. Catal., 53,386 (1978).
5
10
75
k(
12 5
A-1)
Flgure 6. Fourler filtered data of the second coordination shell of well-crystallized MoS, (-. -), sulfided Mo/AI2O3catalyst (sulfided Co-Mo/A1203 catalyst (-).
- -),
and
study involving studies of several model compounds, which would allow accurate phase shifts and amplitude functions for the species in question to be obtained. It is worthwhile mentioning the shape of the absorption edge itself. In the absorption spectra in Figure 1, it is clearly observed that a low-energy shoulder is superimposed on the rising curve of the main peak. This shoulder arises from a transition which we will designate a 1s 4d (atomic) transition although contributions from ligand orbitals are also expected.lg By studying a series of Mo complexes with different sulfur and oxygen ligands, Cramerlg has found that this transition is particularly
-
(19) S. P.Cramer, SSRL Report 7807, June 1978.
3872
The Journal of Physical Chemistry, Vol. 85, No. 25, 1981
intense when molybdenum is doubly bound to oxygen. In contrast to the calcined catalysts, the Mo K edge of MoSz and the sulfided catalysts does not show a low-energy shoulder due to the above transition. Although Mo with terminal sulfur can show this transition, it was that it is virtually nonexistent when molybdenum has all sulfur ligands in a nontetrahedral coordination. Since all of the present absorption spectra were obtained with the same spectrometer resolution and step size, the presence of the low-energy shoulder in the case of the calcined catalysts and the apparent absence of this shoulder in MoSz and the sulfided catalysts must be due to different types of ligands in the samples. It is therefore likely that the major fraction of the molybdenum atoms in the sulfided catalysts are present in a sulfided environment and not as an oxysulfide as suggested by several Further support for this view is given below. The EXAFS data for the Mo/A1203and Co-Mo/A1203 catalysts in the sulfided state show that some degree of order exists outside the fiist coordination shell because two distinct backscatterer peaks are observed in the Fourier transforms of these catalysts (Figure 4, b and c). It is interesting that the data for the MoS2 model compound show many similarities with the data for the catalysts. First of all, two peaks are observed at essentially the same positions as for the sulfided catalysts. The first peak is undoubtedly due to backscattering from S whereas the second peak is due to backscattering from Mo. This assignment is also supported by the Fourier filtered data in Figures 5 and 6 which show maxima of the EXAFS amplitudes at k values which are in agreement with calculated values using the amplitude functions for S and Mo, respectively.20 For the sulfided catalysts it is found that not only the position but also the magnitude of the first peak are essentially the same as for MoS2. The above results therefore show that the molybdenum in the catalysts has the same coordination number of sulfur and the same Mo-S distance as in bulk MoS2. The Fourier filtered data shown in Figure 5 also support the conclusion that molybdenum oxysulfides are not present to a significant extent since the presence of oxygen ligands instead of sulfur ligands in the sulfided catalysts would give rise to a displacement of the maximum of the envelope of the EXAFS amplitudes to a lower k value (5 3 A-1),16,20-22 Figure 6 shows that for both the sulfided catalysts and bulk MoS2the maximum of the EXAFS amplitudes due to the atoms in the second coordination shell is located at about the same k value (-9 A-1). This suggests that the second peak in the sulfided catalysts is due to the same scattering element as in MoS2, i.e., Mo. Moreover, it is found that the position of the second peak is the same for the three samples (Figure 4). These results as well as the results for the first peak lead us to the conclusion that a MoS2-likephase is present in both the Mo/Al2O3and the (20) ELK. Teo and P. A. Lee, J. Am. Chem. SOC.,101, 2815 (1979). (21) P. A. Lee and G. Beni, Phys. Rev. B , 16, 2862 (1977). (22) IbK. Teo, P. A. Lee, A. L. Simons, P. Eisenberger, and B. M. Kincaid, J. Am. Chem. SOC.,99, 3854 (1977).
Clausen et al.
Co-Mo/A1203 catalysts in the sulfided state. It is interesting that the ratio between the second and the first peak amplitudes is significantly lower in the catalysts than in the well-crystallized MoS2sample (Figure 4). This result indicates either that the MoS2-likestructure in the catalysts has some degree of disorder outside the first coordination shell or that this structure consists of very small crystallites with a large fraction of atoms located at the surface. Both situations may give rise to a lower average coordination number in the second shell. For very small crystallites the fact that the Debye-Waller factor is smaller for surface than for bulk atoms may also lead to a decrease in the peak amplitude of the second shell. The average number of Mo scatterers in the second coordination shell can be estimated (neglecting the difference in Debye-Waller factors) for the two catalysts to be around 3 by use of the effective phase shift and the observed peak height normalized to the number of scatterers and their distance determined from the second peak of the Fourier transform of bulk MoS2. This indicates that the size of the crystallites (or domains) in which an order similar to that found in MoSz exists is -10 A. Thus, the sulfided catalysts may be thought to contain very small, MoS2-like crystallites (or domains) on the surface of the alumina support. These results are in accordance with the recent Mossbauer results' which showed that the active CoMo/A1203 catalysts contain MoS2-like structures in a highly dispersed state. In the latter study it was furthermore proposed that cobalt may substitute for molybdenum at surface positions. The present results are consistent with this proposal but provide, of course, no direct confirmation. This must await EXAFS studies on the Co edge, which are now being undertaken.
Conclusion The present EXAFS study shows that the molybdenum atoms in the calcined state of alumina-supported Mo and Co-Mo HDS catalysts are present in structures where no significant order exists outside the first oxygen coordination shell. These structures may be viewed as Mo atoms present as, for example, isolated ions, chains, or very small patches. When the catalysts are sulfided, an ordering of the resulting molybdenum-containing phase takes place as evidenced by the observation of two backscattering peaks. The first one is due to backscattering from S, and the results show that the Mo-S bond length as well as the coordination number are the same as in well-crystallized MoS2. The second peak is identified as arising from backscattering from Mo with the same Mo-Mo distance as in MoS2. However, the average coordination number in this shell seems to be lower than that found in bulk MoS2. These results suggest that the MoS2-likedomains formed upon sulfiding are very small. Acknowledgment. We are grateful to HASYLAB for offering beam time on the synchrotron radiation facility and to EMBL for the access to the EXAFS spectrometer. The assistance of J. Bordas and J. C. Phillips in performing the EXAFS experiments is greatly acknowledged. We also thank J. W. Hansen for help in the preparation of the samples.
