J. Phys. Chem. 1988,92, 1291-1295
1291
EPR Study of the Superoxide Ion in Molybdenum Zeolites Huang Min-Ming,+Jeffrey R. Johns, and Russell F. Howe* Chemistry Department, University of Auckland, Private Bag, Auckland, New Zealand (Received: July 22, 1987)
An EPR study is described of 02-radicals formed in Mo-mordenite and Mo-ZSM-5 zeolites. An 0,species similar to that formed on oxide-supported molybdena catalysts can be generated on the external surface of the zeolites; this species contains magnetically equivalent oxygen atoms on Mo-mordenite and inequivalent oxygens on Mo-ZSM-5. A second Ozspecies,not previously known can be formed within the zeolite pores by pretreatment with ammonia; it is proposed that this treatment generates low-coordination Mo5+sites not found on oxide surfaces which can interact in a partially covalent manner with Oz. The relative importance of geometric and electronic factors in determining the EPR characteristics of 02-on molybdenum-containing surfaces is discussed.
Introduction The superoxide ion 02-has been widely studied as a surface species.' Electron transfer to adsorbed O2produces paramagnetic 02-with a characteristic EPR signal. On most surfaces the two oxygen atoms are observed to be magnetically equivalent, which is consistent with a simple ionic model in which the added electron occupies a orbital of 02,the degeneracy of which is lifted by the surface crystal field.' There are, however, a number of systems known in which analysis of I7O hyperfine splitting reveals the oxygens to be inequivalent. A particularly intriguing case of inequivalence is found with 02-on supported molybdenum catalysts. On thermally reduced Mo-A1203 catalysts the two oxygens have almost identical 1 7 0 hyperfine coupling constants, whereas on Mc-Si02 the coupling constants were observed to differ by 13 f 2 G.2 The difference between the two supports was attributed by Che and Tench2 to differences in surface topology. Ben Taarit and Lunsford, on the other hand, attributed inequivalence in 02-on Mo-SiO, to a degree of partial covalent bonding to the ~ u r f a c ein ; ~fully covalent peroxy species ROz the I7Ocoupling constants differ by ca. 30-50 G. We have shown more recently that 02-formed on photoreduced Mo-SO2 surfaces contains completely equivalent oxygen^.^ The relative importance of geometric (surface topology) and electronic factors in the bonding of 0,to molybdenum-containing surfaces thus seems to be still an open question. In order to pursue this point further we have examined the formation of 02-in molybdenum-containing zeolites. Transition-metal cations exchanged into zeolite matrices can be useful models for the surface sites on less well-defined oxide catalysts., We describe here the detailed EPR characterization of 02-species formed in molybdenum-exchanged mordenite and ZSM-5, and compare the structure of these with that of 02-on conventional supported molybdenum catalysts. Experimental Section Mordenite in the hydrogen-exchanged form was provided by Strem Chemicals, and ZSM-5 in the sodium-exchanged form (3.8 A1 per unit cell) by Chemistry Division, DSIR. Both zeolites were characterized by X-ray powder diffraction prior to use. Two different methods were used to introduce molybdenum into the zeolites. The adsorption of MoCIS from the vapor phase into H-mordenite and subsequent decomposition to produce Momordenite containing ca. 2 Mo per unit cell has been described previously.6 The alternative method involved aqueous ion exchange, and follows a procedure in the patent literature for preparation of Mo-Ye7 A detailed description of this method and the characterization of the Mo-mordenite and Mo-ZSM-5 zeolites will be presented elsewhere;* Mo02C12 is added to a stirred suspension of the zeolite in hydrochloric acid solution, producing mordenite containing ca. 0.4% Mo by weight (0.14 Mo per unit Permanent address: Department of Modern Chemistry, University of Science and Technology of China, Hefei, Anhui, Peoples Republic of China.
0022-3654/88/2092-1291$01,50/0
cell) and ZSM-5 containing 0.2% Mo by weight (0.13 Mo per unit cell). X-ray diffraction measurements showed that the crystallinity of the Mo-mordenite and Mo-ZSM-5 zeolites was fully retained. The molybdenum zeolites were subjected to various treatments in a high-vacuum cell fitted with a quartz side arm for EPR measurements. The cell also served as a reactor for in situ preparations from MoCIS vapor. Spectra were recorded at ca. 9.5 GHz on a Varian E4 spectrometer at either room temperature or 77 K. A DPPH standard (g = 2.0036) was used for determination of g values and spin concentrations were estimated by double integration of the observed signals and comparison with a CuSO., standard. Simulated spectra were obtained with the 0 on an ~ IBM 4 4341 ~ computer. Oxygen enriched to program ~ 36.8% in I7O was provided by Prochem.
Results Molybdenum-Mordenite. Mo-mordenite prepared by vaporphase adsorption of MoCl, shows after activation in vacuo at 450 OC two overlapping Mo5+ EPR signals: a major signal having g, = 1.949, gil = 1.900, and a minor component with a more anisotropic g tensor, g , = 1.985 and gll = 1.831.6 The total MoS+ signal intensity corresponded to typically 1 X 1020g-' (0.6 per unit cell). A subsequent exposure to 10 Torr of O2 at room temperature gave the new signal shown in Figure l a , which has g-tensor components identical with those reported for 02-on oxide-supported molybdenum catalysts' (the g3 component IS partially obscured by a symmetric signal at g = 2.003 probably due to carbonaceous impurities). The intensity of this signal (hereafter referred to as signal A) never exceeded about 10% of the MoS+ intensity (1 X 1019 g-l), and it could be reduced in intensity (but not completely removed) by prolonged evacuation at room temperature. Signal A was also slowly removed on exposure to 2,6-dimethylpyridine vapor at room temperature. The spectrum shown in Figure 2 was obtained on exposure of the activated Mo-mordenite to 170-enriched Oz. This contains I7Ohyperfine structure associated with signal A, the analysis of which is discussed below. Exposure of the activated Mo-mordenite to ammonia followed by subsequent outgassing at 450 "C enhanced the relative intensity (1) Che, M.; Tench, A. J. Adu. Cural. 1985, 32, 1. (2) Che, M.; Tench, A. J.; Naccache, C. J . Chem. SOC.,Faraday Trans. 1 1914. 70. 263. (3) Ben'Taarit, Y.; Lunsford, J. H. J. Phys. Chem. 1973, 77, 780. (4) Seyedmonir, S . R.; Howe, R. F. J. Chem. Soc., Faraday Tram. 1 1984, 80, 2269. ( 5 ) Mortier, W. J.; Schoonbeydt, R. A. Prog. Solid Stare Chem. 1985, 16, 1.
( 6 ) Johns, J. R.; Howe, R. F. Zeolites 1985, 5, 251. (7) Moorehead, E. L. U S . Patent 4297243, 1981. (8) Huang, M. M.; Howe, R. F., to be published. (9) Nilges, M. Ph.D. Thesis, University of Illinois at Urbana-Champaign, 1979.
0 1988 American Chemical Society
Min-Ming et al.
1292 The Journal of Physical Chemistry, Vol, 92, No. 5, 1988 g,=
2016
A
Ol1
17 16
00,
20 gauss
bLk
v 2
d
160170
b-
-
100 gaus?
Figure 1. EPR spectra of 02-in Mo-mordenite (a) signal A, formed in
MoQ-derived sample activated in vacuo; (b) signal B, formed in MoCI5-derivedsample reduced in NH3.
Figure 3. EPR spectra of "0-enriched 0 , (signal B) in Mo-mordenite. (a) spectrum observed in ion-exchanged zeolite reduced in NH,; (b) 160170and 1702in the computer simulation for a mixture of I7Ol6O-, ratio 0.3:0.3:0.4.using the g- and hyperfine-tensor components in Tables I and 11, and Lorentzian line widths of W, = 5.0, W, = 2.5, and W , = 5.5 G .
;,,.
~
10 gauss
";"
170160 4
M
-
3
f
v7. 1
0
80 gauss
Figure 2. EPR spectrum of 170-enriched0, (signal A) in Mc-mordenite. Hyperfine components of singly and doubly labeled species are indicated.
of the more anisotropic Mo5+ signal. Addition of O2 to such ammonia pretreated samples gave the new signal shown in Figure 1 b (signal B) and caused a significant reduction in the intensity of the anisotropic Mo5+ signal. Signal B also has an orthorhombic g tensor. The maximum intensities observed for signal B were up to 4 times those of signal A (corresponding to ca. 0.25 per unit cell). Signal B could be removed only by outgassing above room temperature (e.g., at 100 "C), and it was unaffected by exposure to 2,6-dimethylpyridine at room temperature. Heating in oxygen to 450 "C destroyed both Mo5+signals. Subsequent reduction in hydrogen a t 450 OC restored only the less anisotropic Mo5+signal (gl = 1.949,g,,= 1.900),and no new signals were formed on addition of oxygen at room temperature. Reduction in ammonia a t 450 "C followed by outgassing at this temperature restored both Mo5+ signals, and addition of oxygen at room temperature then produced signal B as before. Mo-mordenite prepared by aqueous ion exchange gave no EPR signals on outgassing in vacuo up to 450 OC. Reduction in hydrogen at this temperature produced a single Mo5+ signal (gl = 1.956,g,, = 1.897,maximum spin concentration 10'' g-'), and subsequent exposure to oxygen at room temperature gave traces
Figure 4. EPR spectrum of 02(signals A and B) formed in 9sMomordenite.
only (ca. 10l6g-l) of the signal A described above. Pretreatment of ion-exchanged samples with ammonia (adsorption at room temperature followed by outgassing up to 450 "C) produced both of the MoS+ signals formed on ammonia reduction of MoC1,derived Mo-mordenite (total spin concentration 5 x IO'* g-l); and addition of oxygen at room temperature gave a more intense (spin concentration 4 X 1017g-l) signal B similar to that shown in Figure 1b. This signal was immediately destroyed on exposure to propene at room temperature, whereas 2-ethylpyridine had no effect. Figure 3 shows signal B formed from I70-enriched oxygen on an ammonia pretreated ion-exchanged Mo-mordenite (a similar spectrum was obtained from the ammonia-pretreated MoCISderived zeolite; I7O coupling constants are listed in Table 11). When the H-mordenite used to prepare the MoCI,-derived zeolites was acid washed (1 M HCl) prior to adsorption of MoCIS, the reactivity toward oxygen of the resulting Mo-mordenite was identical with that of the ion-exchanged zeolites; Le., an extremely weak signal A formed in vacuum-activated samples, and an intense signal B in ammonia-pretreated samples. Figure 4 shows a spectrum obtained on addition of oxygen to an activated 95Mo-mordenite sample prepared by adsorption of 95MoC15into H-mordenite. The major signal present is signal A, with clearly resolved 95Mo hyperfine splitting about the gl component, but the spectrum also contains a contribution from
Superoxide Ion in Molybdenum Zeolites 9
-
1-
The Journal of Physical Chemistry, Vol. 92, No. 5, 1988 1293
2,016
n
I
5
0
~I"o"o1-
,EO G
,
,s.o~o,. , 6 8 G ,
~
11
_ A
__
2
-29 9, ' - p 3
I
Figure 6. EPR spectrum of I7O-enriched02-(signal A) in Mo-ZSM-5.
Figure 5.
EPR spectra of 0, in Mo-EM-5 (ion exchanged): (a) after
H2 reduction; (b) after single NH, reduction; (c) after repeated NH, reduction.
signal B. The 9SMohyperfine splitting about the g, component of signal B is less well resolved than that of signal A, and no splitting could be resolved about the gzor g3components of either signal. Molybdenum-ZSM-5. Mo-ZSM-5 zeolites prepared by aqueous ion exchange or by adsorption of MoCl, gave generally similar EPR spectra to those described for Mo-mordenites. For example, hydrogen reduction of ion-exchanged Mo-ZSM-5 produced a single MoS+ signal (gl = 1.951, gll = 1.895, spin concentration ca. 2 X 10l8g-I corresponding to 0.02 per unit cell), and addition of oxygen at room temperature formed the weak (3 X 10I6g-l) signal A shown in Figure 5a. This signal was removed by exposure to 2,6-dimethylpyridine at room temperature. Adsorption of ammonia and subsequent outgassing at 450 OC produced a second more anisotropic Mo5+ signal (gl = 1.987, glI= 1.837), and addition of oxygen to this sample gave the spectrum shown in Figure 5b, which contains both signals A and B (1 X 10'' g-l). Repeated cycles of ammonia adsorption and outgassing increased the intensity of the second Mo5+ signal (to ca. 3 X 10l8 g-l) and the contribution of signal B to the spectrum obtained on addition of oxygen (Figure 5c). The formation of signal B was accompanied by a distinct reduction in the intensity of the anisotropic Mo5+ signal. As in Mo-mordenite, signal B was unaffected by exposure to 2,6-dimethylpyridine. Figure 6 shows the signal A obtained from "0-enriched O2 with Mo-exchanged ZSM-5 reduced in H2. The differences between this signal and that shown in Figure 2 for mordenite are discussed further below. The corresponding signal B obtained with I70-enriched O2 in ZSM-5 is shown in Figure 7. Molybdenum-Y. As reported previously,I0 Mo-Y zeolites prepared by adsorption and decomposition of Mo(CO), in H-Y gave no oxygen radicals on exposure to 02. Likewise, no 02species could be formed in Mo-Y prepared by aqueous ion exchange." Adsorption and decomposition of MoCl, in H-Y (10) Abdo, S.; Howe, R. F. J. Phys. Ckem. 1983, 87, 1713.
Figure 7. EPR spectrum of I'O-enriched 02-(signal B) in Mo-ZSM-5.
TABLE I: p-Tensor Components of 02-Species on Molybdenum gxx" (g3)
gyy (g2)
(gl)
2.004 2.004
2.011 2.010
2.016 2.021
2.0042 2.004
2.0098 2.010
2.0176 2.018
2.0034
2.010
2.0167
2.0045
2.0103
2.0170
2.0034
2.0092
2.0173
gzz
Mo-M and Mo-ZSM-5 (this work)b A
B Mo03-Si02, thermally reduced ref 2 ref 3 Mo03-Si02, photoreduced ref 4
Mo03-A1203,thermally reduced ref 2
Mo-Y, ultrastable ref 16
perpendicular to 0-0 bond and parallel to surface. along the 0-0 bond. b*O.OO1. ' x axis
z axis
produced an Mo-Y zeolite which showed substantial loss of crystallinity in its X-ray powder diffraction pattern. Exposure of this material to O2 gave a weak 0; signal similar to signal A (gl = 2.016).
Discussion Identification of 02-.Two different EPR signals have been observed here when O2is adsorbed in Mo-mordenite and MoZSM-5 pretreated in different ways. The gtensor components of the signals A and B are compared in Table I with those of 02radicals in other systems. Signal A corresponds closely to 02signals reported previously for supported molybdena catalysts. The I7O hyperfine pattern observed for signal A in Mo-mordenite (Figure 2) is identical with that of 02-formed on photoreduced molybdena-silica catalyst^.^ The set of 11 lines from 1 7 0 2 and 6 lines from I7Ol6Ocentered on g, arise from an 02-species in which the oxygens are magnetically equivalent;' splitting about g, and g2 is too small to be resolved. Furthermore, the 95Mo hyperfine splitting in signal A (1.9 G) in Mo-mordenite is closely (1 1) Huang, M. M.; Howe, R. F. J . Catal., in press
1294 The Journal of Physical Chemistry, Vol. 92, No. 5, 1988 TABLE 11: Hyperfine Tensor Components and Spin Densities for 02-
total spin A x x ( 1 7 0 ) , A,,(95Mo), a density G G on 02-b Mo-mordenite (this work) A
B
-80 -83, -74
19 2.2
107 105
Mo-ZSM-5 (this work) A B
Mo03-Si02,thermally reduced ref 2 ref 3
-80, -68 -85, -78
1.003 109
-85, -72 -82, -69
105 101
ref 18
2
Mo0,-Si02, photoreduced
ref 4 MoO3-Al2O3.thermally reduced ref 2
-80 -80, -77
2
107 105
u f l G bCalculated assuming A, = A,, = 0 and using A, = -1651 G , Bo = -51 38 G
similar to that for 02-on 95Mo-enriched molybdena-silica (Table 11). We thus assign signal A in Mo-mordenite to a "normal" ionic 0,- radical which is symmetrically bound to Mo6+ sites, as on photoreduced molybdena-silica. Signal A with identical g-tensor components formed in MoZSM-5 has a different I7O hyperfine splitting pattern. Figure 6 shows two different sets of six lines centered on g3. As first pointed out by Che et al.? such a pattern is consistent with a single 02-species containing slightly inequivalent oxygens. The 12-G difference in 170coupling constants between the oxygens is comparable to that reported for 0,- on thermally reduced molybdena-silica catalyst^.^^^ Signal A in Mo-ZSM-5 is thus attributed to a normal ionic 0,- radical asymmetrically adsorbed on an Mo6+ site, as in thermally reduced molybdena-silica. Signal B observed in both Mo-mordenite and Me-ZSM-5 has no counterpart on supported molybdena catalysts. The " 0 hyperfine patterns for signal B (Figures 3 and 7) also indicate that the species responsible for this signal contains two slightly inequivalent oxygens; in Mo-mordenite the "0 coupling constants for the two oxygens differ by 9 f 2 G and in Mo-ZSM-5 by I i 2 G. The total spin density on both oxygens is calculated from the observed coupling constants (assuming the components not observed to have values of zero) to be 1.05 for Mo-mordenite and 1.09 for Mo-ZSM-5. Figure 3b shows a calculated I7Ohyperfine pattern for signal B in Mo-mordenite using the coupling constants ratio given in Table 11; the simulation employed a higher 170:'60 than in the experimental spectrum in order to enhance the hyperfine features, but the peak positions and relative intensities agree satisfactorily with those observed. In the simple ionic description of adsorbed 02-the low field g-tensor component is identified as g,,, where z is defined as the 0-0 bond direction, and the magnitude of g,, is determined by the cation charge at the adsorption site.] In this description the higher g,, value of 2.021 for signal B implies a cation site of lower charge than the Mo6+ site on which the usual O,-.species (signal A) is located. The observed 95Mohyperfine splitting in signal B proves that the 0; is interacting with molybdenum and not some other site in the zeolites. The observed g,, value would be consistent with an adsorption site Mo4+ (i.e., electron transfer to 0, from Mo3+ to form Mo4+02-). There are, however, several reasons why this explanation is unlikely. Formation of an 02species similar to that responsible for signal B has never been observed on well-reduced molybdena-alumina or molybdenumsilica catalysts which do contain Mo3+sites. There is no evidence that Mo6+ or Mo5+ in mordenite and ZSM-5 can be reduced readily to Mo3+ under mild conditions (molybdenum cations ion-exchanged into zeolite Y have been shown to be more difficult to reduce than those in supported molybdena catalysts"). Furthermore, signal B appeared only in samples which showed the Mo5+signal characterized by g-tensor components g, = 1 985, gl, = 1.831 (mordenite) or gl = 1.837 (ZSM-5), and although overlapping of the signals prevented a quantitative correlation
Min-Ming et al. being established, the appearance of signal B in all cases caused a reduction in the intensity of the Mo5+ signal. The alternative explanation for the larger g,, value of signal B is that there is a degree of covalent bonding between 0,- and an Mo6+site in the zeolites. The effect of covalency can be seen by comparing the ionic species A13+O; formed on alumina surfaces (g,, variously reported to lie between 2.034 and 2.040') with the covalent adduct [ C O ( N H ~ ) ~ Oformed ~ ] ~ + in zeolite Y (which may be formally represented as Co3+02-;g,, = 2.084).12 A characteristic feature of the covalent dioxygen adducts of cobalt is that the unpaired spin densities on the two oxygens differ markedly; this is consistent with the bent Co-0-0 structure which has been determined in several cases.' For [ C O ( N H ~ ) ~ Oin~zeolite ] ~ + Y, for example, the observed I7O hyperfine coupling constants for the two oxygen atoms are 80 and 60 G, respectively.I2 The inequivalence of oxygens for 02-in Mo-mordenite and MoZSM-5 (signal B) is considerably less than this (ca. 8 G difference in the ''0 coupling constants), and the shift in g,, (from 2.016 to 2.021) is also less, which would indicate a lesser degree of covalency than in the cobalt case. According to the bonding model proposed by Drago et al.I3 for cobalt dioxygen adducts, the degree of covalency depends on the amount of overlap of the metal d,z orbital with the lower energy a* orbital of 02.In the ionic case, there is no overlap, and metal dg electron is completely transferred to the a* orbital to form Co3+02-. In the completely covalent case, the metal electron and one of the unpaired electrons from O2occupy a molecular orbital formed from equal contributions of metal dZ2and oxygen a* orbitals, giving Co2+O,. In both cases, as well as for all situations between these two extremes, the unpaired electron remains in the other oxygen a* orbital and is completely localized on oxygen. From the comparison with cobalt dioxygen adducts we attribute signal B to a partially covalent Mo6+0, species in Mo-mordenite and Mo-ZSM-5 formed by interaction of 0, with Mo5+sites. The crucial point in this interpretation is that a change in the nature of the 0,--molybdenum bond must result in a shift in g,, (since g,, is determined by the separation of the a* orbitals), whereas variations in the 170coupling constants may be caused by electronic or geometric effects. There evidently exist MoS+sites in mordenite and ZSM-5 which can interact with oxygen in a more covalent manner than those on oxide surfaces, as discussed further below. We note that the 95Mohyperfine splitting observed for signal B is significantly larger than that for signal A; in the covalent model metal hyperfine splitting arises largely from spin polarization of the metal-oxygen o bond.13 Adsorption Sites. The relative concentrations of the two 0,species A and B formed in Mo-mordenite and Mo-ZSM-5 depend crucially on the method used to introduce molybdenum into the zeolite, and on the pretreatment. These variations can be accounted for in terms of the adsorption sites available. Vapor-phase adsorption of MoCl, into mordenite produces a high molybdenum loading (up to 2 Mo per unit cell), and exposure to 0, after activation in vacuo gives a relatively intense signal A. The disappearance of this signal on exposure to 2,6-dimethylpyridine proves conclusively that this 02-species is not located within the zeolite channels but must be on the external surface (and is thus readily removed on outgassing). Ion-exchanged mordenite, on the other hand, gives a very weak signal A after reduction in H2 and exposure to 0, (3 orders of magnitude less intense than in the MoC1,-derived zeolite, although the Mo content is only a factor of 10-20 less). Clearly the zeolite prepared from MoCI, contains a much higher concentration of molybdenum sites on the external surface which can form the ionic 0, species similar to that formed on supported molybdenum oxides. The external molybdenum sites differ from those of thermally reduced molybdena catalysts however in that the 02-formed has equivalent oxygens; this appears to be a unique feature of the MoC1,-derived zeolites. Acid washing of the mordenite prior to adsorption of MoC1, reduces the concentration of external sites, possibly by (12) Vansant, E. F.; Lunsford, J. H. Ads. G e m . Ser. 1973, I Z I , 441. (13) Drago, R. S.; Corden, B. B. Arc. Cbem. Res. 1980, 13, 353.
Superoxide Ion in Molybdenum Zeolites removing amorphous material from the zeolite channels and thus facilitating penetration of MoCl, into the channels. The 02-species responsible for signal B is located within the mordenite channels, as shown by the fact that it is not affected by molecules such as 2-ethylpyridine and 2,6-dimethylpyridine which cannot penetrate the channels at room temperature, and that it is more resistant to outgassing than species A. Formation of this species requires the presence of a particular Mo5+site (gl = 1.985, gll= 1.831). Small amounts of this Mo5+ site are formed during activation in vacuo of MoC1,-derived zeolites (signal B can be seen together with signal A in the spectrum obtained from vacuum-activated 95MoC15-derivedmordenite in Figure 4), but ammonia pretreatment is required to produce the highest concentrations of the Mo5+ site and signal B. In ion-exchanged mordenites the Mo5+ site and signal B are formed only after ammonia pretreatment. The ion-exchanging cation is Mo6+ (probably as the dioxo cation M0022+7.11).Ammonia will coordinate to Mo6+cations, thereby weakening their interaction with the zeolite lattice (EPR spectra show direct evidence for coordination of ammonia to Mo5+;a similar interaction is presumed to occur with Mo6+). On subsequent outgassing ammonia desorption causes reduction of the Mo6+ to Mo5+. The g-tensor of the Mo5+signal formed by ammonia treatment is more anisotropic than that of the first Mo5+. The close similarity of the first Mo5+ signal to those formed on thermal reduction of supported molybdena catalysts suggests that this signal is due to Mo5+ in square-pyramidal or distorted octahedral c~ordination.'~The second, more anisotropic signal may then be Mo5+ in distorted tetrahedral coordination. Such a tetrahedral Mo5+ in a low-coordination site will interact with oxygen more strongly than Mo5+ in supported oxide catalysts. A similar description can be given for ion-exchanged MoZSM-5. The weak signal A is due to 0; on the external surface; this species is identical with that formed on thermally reduced molybdena-silica catalysts in that the oxygens are inequivalent. The 0; formed within the ZSM-5 channels (signal B) is similar to the internal 0, formed in mordenite, and its formation requires the presence of a second, more anisotropic Mo5+ species. This species has similar g-tensor components to the corresponding Mo5+ species in mordenite. The lower value of gl,also suggests a distortion of the tetrahedral coordination. Adsorption of ammonia at room temperature and/or heating in ammonia followed by outgassing at high temperature produces the distorted tetrahedral Mo5+ by reduction of Mo6+, as in mordenite, although there may also be some conversion of the first (distorted octahedral) Mo5+ to the second as a result of ammonia treatment. The maximum concentrations of the internal 02-species observed never corresponded to more than ca. 10% of the Mo5+ concentration; Le., only a small fraction of the distorted tetrahedral Mo5+ species (14) Dai, P.S. E.; Lunsford, J. H. J. Catal. 1980, 64, 173.
The Journal of Physical Chemistry, Vol. 92, No. 5, 1988 1295 interact with oxygen, for reasons which are not understood. In the case of zeolite Y, 02-formation was observed only in circumstances where the zeolite lattice had been partly destroyed,14 and the weak signal obtained resembles that of the external 02on mordenite and ZSM-5. In ion-exchanged Mo-Y ammonia treatment produces an anisotropic Mo5+ species (gl = 1.985, g,, = 1.826) which is located in the zeolite @-cages.]' A similar species ( g l = 1.99, gll = 1.80) is formed during activation of Mo-Y prepared from Mo(CO)~.IOThe inability of both zeolites to form an internal 0; (signal B) can be attributed to the location of the reactive Mo5+ site; O2 cannot enter the P-cages of zeolite Y at room t e m p e r a t ~ r e . ~ , Comparison with Oxide Catafysts. Our observation of 02species with identical g-tensor components but different degrees of inequivalence on the external surfaces of Mo-mordenite and Mo-ZSM-5 further supports the suggestion of Che et al.' that magnetic equivalence or otherwise of the "normal" 02-species on molybdenum-containing surfaces is determined by the topology of the surface, i.e., by geometric factors. For example, thermal reduction of supported molybdena catalysts produces highly defective and disrupted surfaces due to removal of both terminal and bridging oxide ligands.I6J7 Photoreduction at low temperatures, on the other hand, produces Mo5+in a much more uniform en~ir0nment.l~ Thermal reduction of Mo-ZSM-5 produces also a disrupted molybdena-like external surface, whereas Mo-mordenite prepared from MoC15 evidently contains a more uniform external surface (possibly isolated oxomolybdenum cations). The fact that the 02-species on all of these surfaces have the same g-tensor components further suggests that they should be described as a simple radical anion generated by electron transfer from molybdenum into a .rr* orbital. The internal 02-species formed in Mo-mordenite and MoZSM-5 has no counterpart on oxide surfaces. In this case the magnetic inequivalence of the oxygens is accompanied by an increase of the value of g,,, which we take as evidence for partial covalent bonding to a particular coordination state of Mo5+which is not found on oxide surfaces. Acknowledgment. This work was supported by the donors of the Petroleum Research Fund, administered by the American Chemical Society. We thank Dr. D. M. Bibby of Chemistry Division, DSIR, for providing the ZSM-5 zeolite. Registry No. 02-,11062-77-4; Mo, 7439-98-7; NH,, 7664-41-7; 02, 1782-44-7. (15) Breck, D.W. Zeolite Molecular Sieves; Wiley-Interscience: New York, 1974. (16) Seyedmonir, S. R.; Howe, R. F. J . Chem. Soc., Faraday Trans. 1 1984, 80, 87. (17) Seyedmonir, S. R.; Howe, R. F. J . Cural., in press. (18) Che, M.; McAteer, J. C.; Tench, A. J. Chem. Phys. Lett. 1975, 31, 145.