Journul of Molecular Catalysis, 59 (1990) 221-231 221

EFFECTS OF THE ACTIVE PHASE-SUPPORT INTERACTION IN VANADIUM OXIDE ON TiOa CATALYSTS FOR o-XYLENE OXIDATION

GARRIELE CENTI, DAVIDE PINELLI and FERRUCCIO ‘IRIFIRO

Department of Industrial Chemistry and Materials, V.le Risorgimento 4, 40136Bobgna (Italy)

Catalytic performance in o-xylene oxidation of two samples of vanadium oxide on rutile or anatase TiOz supports, prepared by mechanical mixing, is analyzed with reference to their characterization by XRD, F”I’-IR and chemical analyses. Results show that the active phase, characterized by a Vv:Viv ratio of 2:l and an IR band at 995 cm-‘, does not interact directly with the titania surface, but rather with a Viv-modified surface. It is suggested that these Vn’ surface sites act as the ‘clinging’ sites for stabilization of an upper active phase. The same active phase is found on both anatase- and r-utile-based samples. Competing migration towards the bulk of these Vrv sites on the x-utile sample causes a slight decrease in catalyst performance. In particular enhanced formation of the intermediate product, phthalide, is observed.

Introduction

V,OS-TiOz(anatase) is a well-known catalytic system [l-51 showing a specific support-active phase interaction. It is generally agreed that a bidimensional amorphous layer of Vv forms on the surface of the support [6-g]; this so-called ‘monolayer’ is considered to be the active phase. It is suggested that the excess V-oxide deposited remains as crystalline VzO,, which is thought to be inactive. The formation of a paracrystalline VzO, upper layer, however, also has been shown [71 and the presence of Viv on the surface recognized [lo-131, but its importance in the catalytic cycle is usually neglected. The structure of TiO, is important in determining the characteristics of the active phase 1143. Only TiO,(anatase) is thought to create the monolayer giving rise to a highly active and selective catalyst in o-xylene oxidation. However, there is no generally accepted theory about how TiOz influences VzO,. According to Vejux and Courtine 1141, the fit between the (010) planes of V,O, and the structure of anatase results in a preferential exposure of Vv=O groups on the surface 115,161. Since these groups are considered to be the active sites, this explains why the catalytic performance of V/Ti/O catalysts is better than that of V205. Other authors,

0304-5102/90/$3.50 0 Elsevier Sequoia/Printed in The Netherlands

222

however, believe that the active phase has a different structure as compared with crystalline V,O,. They postulate a monolayer consisting of (i) mono- dispersed VO1 units located at the cationic sites with two terminal oxygens (V=O) and two bridging oxygens (V-O-Ti) [17,18]; (ii) an amorphous disordered layer of V-O polyhedra [19-211 that forms in a ‘wetting’ process of V,O, by TiOz, the difference in surface free energy being the driving force of the migration 1221; (iii) a strongly disordered vanadium oxide species preferentially deposited at the edges of the titania [23]; and (iv) patches of vanadium pentoxide on the surface of the support [241.

In previous studies 125-281 we have shown the highly complex nature of surface vanadium oxide species on titania, evidencing the presence of several types of species and, in particular, of non-negligible amounts of superficial strongly interacting VN species. The similar catalytic behavior of impregnated catalysts utilizing rutile and anatase has also been demon- strated [251. Therefore, questions still remain about the formation of the active phase and its real nature.

Reported in this paper are the results of research carried out to characterize vanadium oxide on anatase or rutile TiOz prepared using a mechanical-mixing method. Although this preparation method is seldom used in studies reported in the literature, we choose to employ it for this study because, in spite of its intrinsic simplicity, it allows a better understanding of the solid state reactions of the spreading of V,O, on the support surface and its reduction to VN, without interference of any other chemical phenomena. It also is a preparation method used for the synthesis of industrial catalysts for o-xylene oxidation and allows very active/selective catalysts to be prepared.

Experimental

Catalyst preparation The catalysts were prepared by mechanical mixing and grinding ( 1 min)

of TiOZ and VzO, (Carlo Erba reagent grade). Generally, 7.7 wt.% V,O, was used, an amount typical for industrial

preparations. Anatase (9.6 m2 g-l> and rutile (8.9 m2 g-l) prepared by TiCl, hydrolysis were used in order to obtain highly pure Ti02 supports and to exclude interferences from doping. After mixing and grinding, the samples were calcined in air at 500 “C for 16 h or longer. Hereinafter, the following notation will be utilized: A indicates catalysts prepared using an anatase Ti02 support and R indicates catalysts prepared using a rutile Ti02 support.

Catalytic tests The catalysts were tested in a conventional laboratory apparatus with a

tubular fixed bed reactor working at atmospheric pressure, and on-line gas chromatographic analysis of reagent and product compositions 125,261. The standard reactant composition was 1.5% o-xylene, 20.5% O2 and 78% N2. The

catalyst (0.52 g) was loaded as grains (0.250-0.420 mm). A thermocouple, placed in the middle of the catalyst bed, was used to verify that the axial temperature profile was within 5 “C. The catalytic results refer to the stable catalytic behavior obtained after at least 60 h on stream under typical reaction conditions for o -xylene oxidation.

Chemical analysis The samples (about 0.5 g) were moistened with 50ml of a room

temperature dilute (4 Ml H&SO, solution for 15 min under stirring and then filtered. The amount of vanadium was determined separately in the filtered solution and in the residual sample dissolved in boiling concentrated H2S04 (16 M). The total amounts and the valence state of vanadium were determ- ined by a titrimetric method as previously reported [25,261. The vanadium species extracted by the room temperature dilute sulphuric acid solution will be, hereinafter, called ‘soluble’ or ‘weakly-interacting’ species, whereas the remaining species determined after dissolving the residue are called ‘in- soluble’ or ‘strongly-interacting’ species. The dissolution and chemical analysis procedures were chosen after preliminary tests with reference samples [251.

Characterization X-ray diffraction patterns (XRD) (powder technique) were obtained

using N&filtered Cu K, radiation and a Philips computer-controlled in- strument. Surface areas were determined using the BET method and a Carlo Erba Sorptomatic instrument. A Perkin Elmer 7200 Fourier transform infrared (FYI-IR) spectrometer was used to record the IR spectra using the KBr disc technique. Calibrated amounts of V-Ti-C samples (around 0.1% w/w with respect to KBr) were used and electronically calibrated subtraction of the TiOZ contribution to the spectrum was done.

Resldts

The results of XRD analysis of A and R samples after grinding (a), alter calcination (air, 500 “C, 16 h) (b), after the catalytic tests (c) and after the catalytic tests and consecutive oxidation (air, 400 “C, 4 h) (d) are summarized in Table 1. Only a small decrease in the relative intensity of crystalline V,O, diffraction lines after calcination was observed, while after the catalytic test and after consecutive oxidation no crystalline vanadium oxide could be detected. The apparent higher relative intensity of V,O, XRD reflections in the R sample in comparison to that in the A sample derives from the different intensities of the mean peaks of anatase TiO, and rutile TiOZ. The intensity ratio (101) anatase: (100) i-utile is approximately 4:l [29]. In all cases the diffraction lines relative to rutile in A samples and anatase in R samples were negligible and did not change after the catalytic tests. For the rutile sample, a decrease in the cell volume was also observed afk catalytic tests, indicating the formation of a Vn’ solid solution in the r-utile matrix, in agreement with literature data 1301.

224

Z@ativa intensities 96 (Z/I,) of the main X-ray diffraction lines of A and R samples afk grinding (a), after calcination fair, 500 “C, 16 h) fb), after cata&tie tests (~1 and afkr the catalytic tests and consacutive oxidation (air, 400 “c, 4 h) Id)

Phase

sample

TiO, TiO, v2ol5

anatafre rutile

hkl= 101 200 004 110 211 101 001 110 400 d, (ihf = 3.51 1.89 2.38 3.24 1.69 2.49 4.38 3.40 2.88

untreated A (a)

(b) (c) (d)

R, (a) (b) Cc) Cd)

z/r,” = 100 33 22 100 60 41 100 90 65 qwwA 100 28 25 1 - - 7 6 4 ~&KmA 100 27 24 1 - - 5 4 3 ~II,wwA 100 28 25 2 - - - - - qlmA 100 27 24 1 - - - - -

I/Z~~~~~ 8 2 - 100 44 47 29 13 10 ~lfoflOOBt 8 2 - 100 43 46 25 10 8 f&DJO,R 6 1 - 10038 44 - - - qmo, 6 1 - 100 37 43 - - -

*Reference values taken from the relative ASTM tables (4-0477 for TiO, anatase, 4-0551 for TiOg rutile, 9-387 for V,O,).

The chemical analyses of the samples after calcination (Table 2) indicate the presence of insoluble VN. The amounts for anatase and rutile samples were slightly smaller than those necessary to achieve a complete coverage of the TiOz surface (0.9-l.Owt.%), assuming a value of about 0.1 wt.% of V,O, per m2 of surface for monolayer coverage. The evolution of the amount of insoluble VW as a function of the time of calcination for an anatase sample (Fig. 1) shows that the amount of Vrv does not further increase significantly after the first 24 h calcination. This indicates that the

TABLE 2

Chemical analyses of A and R samples &er grinding (a), after cakination (air, 500 “C, 16 h) (b), a&w catalytic tests (c) and after the catalytic tests and mnsecutive oxidation (air, 400 “C, 4 h) (d)

Sample Chemical analyses” Mean valence state of soluble fractionb

v’v insol. Vn’ soluble Vv soluble

A (a) 0.0 0.0 7.7 5.00 (b) 0.8 0.0 6.9 5.00 (c) 0.9 2.2 4.6 4.68 (d) 0.9 0.6 6.2 4.91

R (a) 0.0 0.0 7.7 5.00 (b) 0.8 0.0 6.9 5.00 (c) 4.1 0.9 2.6 4.74 (d) 4.1 0.2 3.6 4.95

“All amount6 are expressed as % by weight of corresponding V,O, per g of sample. b The mean valence state is determined as follows: 4 + (mol Vv)/(mol Vn’ + mol V”).

225

1

0.75

0.5

0.25

0

V(IV)insduble(asV+,vff%) I

value of monolayer

0’ @

7 0

9 , I I 0 25 50 75 100

Time of calcination, hours

Fig. 1. Effect of time of calcination on the amount of insoluble VN for catalysts prepared using an anatase-TiO, support.

value of about 0.1% wt. per m2 of surface area corresponds to the limiting value for these samples.

The catalytic performances in o-xylene oxidation of A and R catalysts are shown in Figs. 2 and 3, respectively. The behavior of the two catalysts is similar. The total C&-selectivity of A is analogous to that of R, but the

280 290 300

Temperature, C

Fig. 2. Catalytic behavior iu o-xylene oxidation using vanadium-oxide on anatase. Symbols: (X) conversion, (0) yield to phthalic anhydride, (A) yield to phthalide, ( X) yield to carbon oxides. Experimental conditions: 0.52 g of catalyst; WF = 36496; 1.5% o-xylene, 20.5% Oz. Note: steady-state catalytic behavior after at least 60 h on stream.

25

0 270 280 290 300

Temperature, C

310

Fig. 3. Catalytic behavior in o-xylene oxidation using V-oxide on rutile. Symbols, experimental conditions and note as for Fig. 2.

phthalic anhydride selectivity of the R sample is slightly smaller because of the higher yields of phthalide. Phthalide is an intermediate in the pathway of phthalic anhydride synthesis (Scheme 1). When the soluble fraction of vanadium is removed by washing, the catalysts become less active, but are still more active than pure anatase or r-utile TiOz (Fig. 4). The activity is roughly proportional to the amount of surface Vn’ sites 1251. The washed A samples after calcination are slightly less active than A samples (washed) after catalytic tests (Fig. 4), indicating a slight increase in surface VW sites after this treatment, in agreement with chemical analysis data (Table 2). On the contrary, a considerable decrease in activity was found for the cor- responding R samples (Fig. 41, indicating that surface Vrv sites migrate towards the bulk, forming a solid solution in the rutile structure, in agreement with XRD data. EPR data 125,311 also confirm these indications.

In the soluble part of the discharged catalysts an analogous mean valence state for both anatase and rutile samples (Table 2) is noted. This value (around 4.7) corresponds to a V”/Vrv atomic ratio of 2:1, taking into account a small fraction of V,O, present in the R sample as shown by l?I’-IR data (Fig. 5 A). A clear difference between the two supports is that the total

O-Xylene

Scheme 1.

phthalide .a phthalic anhydrl e

227

Conv. (300 C), %

'Wl

A(l) A(2) A(3) R(l)

Sample

Fig. 4. Comparison of the conversion at 300 “C in o-xylene oxidation for the A and R samples as such (l), after consecutive removal of the upper soluble layer (2) and when the upper soluble layer is removed before the catalytic tests (3). Experimental conditions as for Fig. 2.

amount of the soluble part is lower in the case of the rutile catalyst because of the larger insoluble VW fraction.

The FT-IR spectra in the v(V=O) region of the discharged catalyst (A, R) are reported in Fig. 5 (the TiOa contribution has been subtracted from the spectra). Before the catalytic tests only the spectrum of crystalline V,O, could be detected, whereas after the catalytic tests three bands, namely at

.r- rut i le

Fig. 5. FT-IR spectra in the v(V=O) region of anatase and rutile samples after cakination (a), after catalytic tests (b) and after consecutive oxidation (air, 4OO”C, 4 h) CC).

1020, 1010 and 995cm-’ can be identified. The first, corresponding to the symmetrical stretching of V”=O in crystalline VzO,, is almost negligible, while the 995 cm-l band is the most intense in spectra of both R and A catalysts. The intensity of the band at 995 cm-l is approximately propor- tional to the total amount of the soluble fraction of vanadium with a Vv:VN ratio of 2:l. A rough correlation was, in fact, observed in several samples [31] between the intensity of the FT-IR band at 995 cm-‘, the amount of soluble vanadium fraction which could be attributed to the non-crystalline active species with a Vv:Vn’ ratio of 2:l and the selectivity to phthalic anhydride in o-xylene oxidation.

Chemical analyses (Table 2) and FT-IR data (Fig. 5c) of the discharged catalysts after consecutive calcination (air, 400 “C for 4 h) showed that they are relatively stable to oxidation. The mean valence state in the soluble fraction remains below 5.0, the value corresponding to complete re-oxidation, and in the FT-IR spectra the band centered at 995 cm-’ is still present, but the relative intensity of the bands at 1020 cm-l (v,(V=O) in crystalline VzO,) and at 1010 cm-’ (reasonably attributed to an intermediate phase in going from the active phase to crystalline V,O,> increases. The phases which form by oxidation are not XRD detectable (Table 1). The effect is much more drastic in R samples than in A samples, suggesting in the latter an increase in the stabilization of the reduced phase characterized by the IR band at 995 cm-‘. No oxidation of insoluble VW occurs in either the R or A samples, indicating the very strong stabilization as a consequence of the interaction with the TiOz surface.

Discussion

Modification of the TiO, surface by VW ions Chemical analyses show the formation of relevant amounts of insoluble

VN species by interaction of V” ions with the surface TiO, sites. The formation of these VN sites occurs during calcination in air and in the absence of any reducing agent. The redox and chemical (solubility) properties of the VN sites are dramatically altered by the interaction with the TiOz surface, in comparison with VN-oxide; for example, VzO, easily oxidizes to V”-oxide even at 100 “C, and is soluble in dilute H,SO,. XRD (Table 1) and EPR [31] data clearly exclude the formation of a solid solution in A samples, whereas this effect is present in R samples. The formation of these VN surface sites, at least for A samples, occurs by a specific reaction between TiOz surface sites and V” sites. This chemical reaction is the driving force for the spontaneous reduction of V” in oxidizing conditions and for its surface migration. In fact, the reaction does not occur only in the regions of contact between TiOz and V,O, particles. The grain size of the two oxides does not have a significant influence on the amount of VN determined after long term calcination (100 h). Furthermore, a change of the V,0s:Ti02 w/w ratio from 7:93 (as used in this work) to 93:7 does not result in a significant

229

change in the amount of Viv per m2 of Ti02 surface. These indications suggest that Viv does not form only at the interface between Ti02 and V,O, particles, but may ‘walk’ on the surface to specific TiO, sites where they react, forming Vw. Furthermore, these data suggest that the value found for anatase samples (about 0.1% wt. per m2 of surface area) corresponds to the maximum amount of Viv on the surface of Ti02. This surface coverage,

w assuming a good distribution of V ions, indicates a mean vanadium to vanadium distance of about 4 A. This distance corresponds to that expected on the basis of a reaction model (assuming the (010) plane of Ti02 as the predominant one on the surface of a low surface area sample) 1321 of vanadium ions with two neighboring Ti-OH groups and formation of a distorted octahedral vanadyl group (as shown by EPR analysis [25,27,28]) which interacts with underlying Ti4’ ions. The mean distance of 4A indicates that some bridging T&O-Ti oxygens and Ti4’ ions still remain available on the surface modified by Vrv for possible secondary interaction with the upper layer (soluble fraction), but the main interaction of the upper layer occurs with surface vanadyl groups.

In R samples after calcination, the data for surface reactivity (Fig. 4) indicate that surface Viv ions are present in an amount quite comparable to that in A samples, but these VN ions are relatively unstable and during the catalytic tests tend to migrate towards the bulk of r-utile Ti02, forming a solid solution, as confirmed by the change in the cell parameters (XRD analysis). These surface VN ions (insoluble fraction) also may be formed during the catalytic tests 1261. In the R sample it is thus probable that a competition exists during the catalytic tests between formation of these VN sites and their tendency to migrate towards the bulk of rutile Ti02. When the steady -state catalytic behavior is reached, ‘clusters’ of VN are present in the R sample as clearly shown by ESR characterization 1311, in contrast to well dispersed vanadyl ions present on the surface of sample A after similar treatment [31]. The competition with bulk migration thus limits the possibility of surface diffusion of vanadium ions, leaving probably a large fraction of surface uncovered on the rutile sample, even after attainment of steady-state catalytic behavior.

Support effect on active phase behavior We have shown in the previous part that during the calcination step

some Vv ions react with Ti02, forming VN sites. The rest of the vanadium oxide remains as such (crystalline V20,) on the surface, as clearly shown by IR and XRD measurements. This part of the initial vanadium, however, transforms during the catalytic tests, giving rise to the active phase for o-xylene oxidation to phthalic anhydride, with a parallel increase in both activity and selectivity. This active phase is XRD amorphous; it is probably characterized by a Vv/VN ratio near to 2:l and an IR band at 995 cm-‘. This IR band may be attributed to stretching of the Vv=O bond in a distorted octahedral environment. The lowering in frequency, as compared with

v,V=O in V,O, (1020 cm-l), can be tentatively explained by the presence of Vrv in the neighbouring octahedral Vv.

Our data suggest the presence of the same active phase on both rutile and anatase-based catalysts. The catalytic performance of the two samples is comparable, as well as the IR, XRD and chemical analysis data. The principal difference in the catalytic behavior is a slightly reduced oxidizing power of the R samples, giving rise to the formation of an higher amount of phthalide, clearly a product intermediate in the synthesis of phthalic anhydride:

However, Table 2 and Fig. 5 suggest some differences between A and R catalysts, which may be summarized as follows:

(a) A smaller amount of active phase (soluble fraction of Table 2) is present in the rutile sample.

(b) The active phase on the rutile support is slightly less stable towards reoxidation, as shown by the tests of consecutive calcination after catalytic tests (Fig. 5).

However, it should be noted that even increasing the initial amount of VzO, in the catalyst R in order to compensate the loss of surface vanadium oxide caused by migration toward the bulk of Vrv ions, no improvements in the catalytic performances can be observed [311. This indicates that the difference observed in the catalytic behavior between the A and R samples cannot be attributed to a smaller amount of active phase on the surface of the R sample due to the partial migration of vanadium toward the bulk of the r-utile matrix.

These data thus suggest the probable presence of the same active phase on rutile- and anatase-based catalysts, but a lower effect of support-induced stabilization on the rutile-based catalysts. These catalysts show a slight decrease in catalytic performance in o-xylene oxidation and, in particular, an enhanced formation of the intermediate product, phthalide.

As previously discussed, this active phase does not interact directly with the TiO, surface, but rather with the surface modified by Viv ions. It was also shown that the mean amount of Viv sites on the rutile surface is probably lower than on the anatase surface, due to the tendency of Viv to migrate toward the bulk of the rutile structure. On the other hand, on the rutile surface ‘clusters’ of Vrv ions are present, in contrast to isolated vanadyl ions on the anatase surface. It seems thus reasonable to indicate a different type of interaction with the upper active phase between the A and R surfaces modified in a different manner by the Vn’ ions.

It is thus also reasonable to suggest tentatively that the role of the crystalline structure of TiOz is not related to crystallographic fitting or wetting effects, as indicated in the literature [8,9,14,17,22], but rather to the ability to stabilize specific Viv surface sites which act as the ‘clinging’ sites for the upper active layer. We may also remark that no stable Viv sites could be detected on VzO,-Al,O, or -SiOa catalysts whose only treatment was calcination 1331. It is well known that these supports exhibit poorer performances in o-xylene conversion to phthalic anhydride [ 181.

231

References

1 M. S. Wainwright and N. R. Foster, Cutal. Rev. Sci. Techrwl., 19 (1979) 211. 2 G. C. Bond and P. Konig, J. Cutal., 77 (1982) 309. 3 W. E. Slinkard and P. B. Degroot, J. Catal., 68 (1981) 423. 4 G. C. Bond, A. J. Sarkany and G. D. Pa&t, J. Cc&Z., 57 (1979) 476. 5 A. J. van Hengstum, J. C. van Ommen, H. Bosh and P. J. Gellings, Appl. Catal., 8 (1983)

369. 6 G. C. Bond and K. Briickman, J. Chem. Sot., Faraday Discuss., 72 (1981) 235. 7 G. C. Bond, J. P. Zurita, S. Flamerz, P. J. Gellings, H. Bosh, J. C. van Ommen and B. J. Kip,

AppZ. Catal., 22 (1986) 361. 8 G. C. Bond, J. P. Zurita and S. Flamerz, Appl. Catal., 27 (1986) 353. 9 I. E. Wachs, R. Y. Saieh, S. S. Chan and C. C. Chersich, AppZ. Catal., 15 (1985) 339.

10 A. Andersson and S. L. T. Andersson, in R. K. Grasselli and J. F. Brazdil (eds.), Solid State Chemistry in Catalysis, ACS Symp. Ser. 279; American Chemical Society, Washington D.C., 1985, p. 121.

11 M. Gasior, I. Gasior and B. Grzyboska, AppZ. Catal., 10 (1984) 87. 12 M. Rusiecka, B. Grzybowska and M. Gasior, AppZ. Cutal., 10 (1984) 101. 13 P. Meriaudeau and J. C. Vedrine, Nouv. J. Chem., 2 (1978) 133. 14 A. Vejuz and P. Courtine, J. Solid State Chem., 23 (1978) 93. 15 M. Inomata, K. Mori, A. Miyamoto, T. Ui and Y. Murakami, J. Phys.Chem., 87 (1983) 754. 16 M. Inomata, A. Miyamoto and Y. Murakami, J. Phys. Chem., 85 (1981) 2372. 17 R. Kozlowski, R. F. Pettifer and J. M. Thomas, J. Phys. Chem., 87 (1983) 5176. 18 J. Haber, A. Kozlowska and R. Kozloski, J. Catal., 102 (1986) 52. 19 Y. Nakagawa, T. Ono, H. Miyata and K. Kubokawa, J. Chem. Sot. Faraday Trans. 1, 79

( 1983) 2929. 20 H. Miyata, K. Fujii and T. Ono, J. Chem. Sot., Fumduy Trans. 1, 84 (1988) 3121. 21 J. L. G. Fierro, L. A. Arrua, J. M. Lopez Nieto and G. Kremenic, Appl. Cutal., 37 (1988) 323. 22 J. Haber, T. Machej and T. C. Zeppe, Surf. Sci., 151 (1985) 301. 23 A. Baiker, P. Dollemneier, M. Ghnski and A. Reller, AppZ. Cutal., 35 (1987) 351. 24 M. Del Arco, M. J. Holgado, C. Martin and V. Rives, J. Cutal., 99 (1986) 19. 25 F. Cavani, G. Centi, E. Foresti, F. Trifiro’ and G. Busca, J. Chem. Sot., Faraday Trans. 1, 84

(1988) 237. 26 F. Cavani, G. Centi, F. Parrinello and F. Trifuo’, in B. Delmon, P. Grange, P. A. Jacobs, G.

Poncelet (eds), Preparation of Catalysts ZV, Elsevier, Amsterdam, 1987, p. 227. 27 G. Busca, G. Centi, L. Marchetti and F. Trifiro’, Langmuir, 2 (1986) 568. 28 G. Busca, G. Centi, L. Marchetti and F. Trifiro’, J. Chem. Sot., Faraday Trans. 1, 81 (1985)

1663. 29 R. Spurr and H. Myers, Anal. Chem., 31 (1966) 2481. 30 I. Gasior, M. Gasior, B. Grzybowska, R. Kozlowski and J. Sloczynski, Bull. Acad. PoZ. Sci., 11

(1979) 829. 31 G. Centi, E. Giamello, D. Pinelli and F. Trifiro’, in preparation. 32 K. Morishige, F. Kanno, S. Ogawara and S. Sasaki, J. Phys. Chem., 89 (1985) 4404. 33 T. Tanaka, H. Yanashita, R. Tsuchitani, T. Funabiki and S. Yoshida, J. Chem. SOC., Faraday

Trans. 1, 84 (1988) 2987.