J. MATER. CHEM., 1991, 1(6), 1071-1077 Alkali-metal Fluorides supported on y-Alumina Surface Reactions involving 18F- and 35S-labelled Sulphur Tetrafluoride and Thionyl Fluoride, 35S-labelled Sulphur Dioxide, l8F=and 14C-labelled Carbonyl Fluoride and 14C-labelled Carbon Dioxide Thomas Baird, Abdallah Bendada, Geoffrey Webb* and John M. Winfield* Department of Chemistry, University of Glasgow, Glasgow G12 8QQ, UK Reactions between the Lewis acids, SF,, F,SO, SO2, F2C0 and CO,, radiolabelled with 14C, 18F and 35Sas appropriate, and CsF or KF supported on y-alumina have been studied under heterogeneous conditions at room temperature. Fluorine-containing Lewis acids react in two ways, with surface F- anions to form Lewis acid- base complexes and to fluorinate surface hydroxyl groups.Maximum surface F-anion activity occurs at a metal fluoride loading of 5.5 mmol g-'. The reaction between CsF or KF and y-alumina to give hexafluoroalumin- ate salts, which occurs during the impregnation process, becomes more important as the metal fluoride loading increases over the range 0.6-20.0 mmol g-'. This reaction is accompanied by dehydroxylation of the surface of the support, but there is no evidence for the presence of free hydroxide anions. Keywords: Alkali-metal fluoride ; y-Alumina ; Radiotracer; Transmission electron microscopy; Surface species The widespread use of anhydrous ionic fluorides as bases in organic synthesis' has generated a large number of appli- cations in which ionic fluorides are supported on high-surface- area inorganic oxides in order to carry out reactions more efficiently under heterogeneous conditions.Several materials have been suggested as supports,2 but potassium fluoride supported on chromatographic alumina appears to be the reagent of choice in most situation^.^ Potassium fluoride supported on alumina is a complex material whose surface properties are composition dependent. Impregnation of alumina with KF in aqueous solution leads to the formation of K3AlF6 to some extent, possibly according to4 12KF+3H20+A1203 +6KOH +2K3AlFs (1) This has resulted in the suggestion that the basic properties of KF/alumina are due wholly' or in art^,^ to the presence of 'free' hydroxide anions on the surface. "F MAS NMR spectro~copy~~'and fast-atom bombardment mass spec-trometry (FABMS)' have provided good evidence for the presence of well dispersed F-ions on alumina at low loadings of KF, but the evidence for KOH is inconclusive.There was no evidence in the FABMS study for OH-except in very heavily loaded samples.' The carbonate anion has been ident- ified by FTIR spectroscopy in alkali-metal fluorides supported on alumina after exposure to the atmosphere;' this is indirect evidence for free OH- anion on the surface. In an attempt to clarify the surface properties of alkali- metal fluorides supported on alumina, particularly their vari- ation with the composition of the materials, we have studied the behaviour of CsF and KF supported on y-alumina towards volatile Lewis acids.The radiotracer approach used is similar to that employed to quantify the process of activating unsup- ported CsF for use as a heterogeneous base or catalyst." Our results are in agreement with many of those from previous physicochemical st~dies,~,',' notably the predominance of MF, M=Cs or K, at low loading and the formation of M3AlF6 as a result of the impregnation process. However, it is considered that eqn. (1) is unrealistic for describing the process, and an alternative explanation is offered based on a consideration of the y-alumina surface, an aspect that has been largely ignored in previous work. Some parts of this work have been reported in a preliminary communication.' Experimental Except where noted below, the vacuum, dry-box and radio- tracer techniques used in this work have been described previously.",' Instrumentation For the XRD measurements, a Philips diffractometer was used with Cu-Ka radiation and the powdered samples were mounted on adhesive tape in a glove box with the tape rolled into a cylinder; for TEM (JEOL JEM 1200 Ex) specimens were prepared by dipping an adhesive-coated carbon-filmed grid in the powder and were then inserted immediately into the microscope to minimise hydrolysis during transfer; for 27Al MAS NMR spectroscopy [Varian UXR-300/89 spec- trometer (IRL, University of Durham) at 76.1 52 MHz], samples were contained in a 0.3 cm3 zirconia tube; for vibrational spectroscopy (PE 983 and Spex Ramalog spec- trometers) solids were sampled in sealed Pyrex capillaries or as Nujol mulls, and gases were contained in a Pyrex cell with AgCl windows; B.E.T.areas were determined using a Pyrex system with N2 as adsorbate. The compositions of volatile product mixtures resulting from reactions of volatile Lewis acids with supported metal fluorides were determined by IR spectroscopy using a calibrated Pyrex manifold equipped with a constant volume manometer and IR gas cell, and experimen- tally determined pressure us. peak area calibration relationships. Materials The Lewis acids used were either commercial samples, C02 and SO2which were dried by multiple distillations over P205, or were prepared by standard methods, SF4,13 F2S0,14 F2C0,15 and stored in metal pressure vessels over activated NaF.Immediately before use, SF4 was purified further via its BF3 adduct.16 The labelled compounds SF3"F, 3'SF4, 18FFC0 and I4CO2 were prepared as previously described," F23'S0 and 3'S02 by hydrolysis of 3'SF4 over calcined y-alumina at 295 and 373 K, respectively, and "FFSO by hydrolysis of SF318F over calcined y-alumina at 295 K. 14C- J. MATER. CHEM., 1991, VOL. 1 Labelled F2C0 was prepared from 14C02 by the sequence The contents were shaken vigorously for 2 h, MeCN was removed in vacua at room temperature, and the vessel was 14C02 Zn CIF NaF F2 14C0 then heated under vacuum at 398 K for 16 h to decompose 14C0 C1F14C0 the Cs[OCF(CF,),]. Finally the material was calcined at (2) 523 K for 5 h and stored in the glove box.Bands in the IR The radiochemical purity of all 18F-labelled compounds was spectrum of the material due to the [(CF3),CFO]- anion checked by half-life and y-ray spectrum determinations and were absent. linear relationships between count rate and pressure were Preparations in which various quantities of dried metal verified experimentally for all 14C- and 35S-labelled com- fluorides and calcined y-alumina were ground together in the pounds. glove box, followed by calcination at 523 K, led to irreproduc- ible B.E.T. results from different preparations; this method of preparation was therefore discontinued. Preparation of Caesium and Potassium Fluorides supported on y-Alumina ResultsSupported metal fluorides, composition range 0.6-20.0 mmol g-', were prepared following the literature method4 from CsF y-Alumina from two sources (Degussa and Pural) was used or KF (B.D.H.Ltd. Optran grade) and y-alumina (Degussa in this work with essentially identical results. For consistency C or Pural Sb90). In a typical preparation, KF (44.0mmol) the data reported in Tables 1-4 refer to metal fluorides in distilled water (130 cm3) and y-alumina (10.0 g) were mag- supported on Degussa C y-alumina. netically stirred at room temperature for 1 h, the bulk of the water was removed by rotary evaporation at ca. 333 K, and the solid calcined under dynamic vacuum at 523 K for 5 h. Examination of y-Alumina-supported Caesium and Potassium All materials were stored under dry N2 in the glove box.Fluorides by Physical Methods Similar materials were prepared under non-aqueous con- B.E.T. areas of CsF and KF supported on y-alumina are ditions using alkali-metal heptafluoroisopropoxides,'7 dry given in Table 1. Impregnation of y-alumina by MF, M=Cs acetonitrile," and y-alumina previously calcined at 523 K. or K, under aqueous conditions resulted in marked reductions Typically, y-alumina (5.0g) was added in the glove box to a in B.E.T. area. At least seven samples, normally from different solution of CS[OCF(CF,)~] (22.0 mmol) in MeCN (20 cm3), preparations were examined for each metal fluoride loading. the mixture being contained in a Monel metal pressure vessel. Both preparative methods were satisfactory in the degree of Table 1 B.E.T.areas" of KF and CsF supported on y-alumina after calcination at 523 K (CsF/y-alumina)/m2 g- (KF/y-alumina)/m2g-' metal fluoride loading/ mmol g- ' prepared from H,O prepared from MeCN prepared from H20 prepared from MeCN 1.1 124-135 140-156 2.0 76-88 62-73 104-1 16 80-93 4.4 55-65 40-50 91-101 68-8 1 5.5 51-61 3 1-47 76-88 5 1-69 6.0 42-56 24-32 69-8 1 47-54 8.8 33-45 19-26 51-65 32-4 1 15.0 15-27 10-19 29-39 17-25 20.0 11-18 19-29 a 95% Confidence limits. B.E.T. area of y-alumina (Degussa C) calcined at 523 K = 155-165 m2 g-'. Table 2 Summary of examination of KF supported on y-alumina by XRD and TEM identified by XRD identified by TEM loading/mmol g- KF y-alumina K,AlF, KF particle size/A y-alumina 2.0 35 4.4 33 5.5 40 8.8 20.0 " An unidentified 'Al-F' phase was also present.Table 3 "A1 NMR chemical shifts" of metal fluoridely-alumina CsF/y -alumina KF/ y-alumina 4.4 54.3 5.4, -1.6 59.3 4.9, -0.7 8.8 61.6 5.2, -1.6 59.9 4.8, -1.8 10.0 60.5 4.8, -1.6 58.7 4.0, -1.3 20.0 55.2 -1.8 71.2 -1.3 =W.r.t. aluminium(III) chloride (aq). For y-alumina calcined at 523 K 6(27Al,,t)=73.1 and 6(27A10c,) 6.9. J. MATER. CHEM., 1991, VOL. 1 Table 4 18FExchange occurring on exposure of CsF or KF supported on y-alumina to OCF18F or SF318F at room temperature" ~~ volatile fluoride 18F specific count rate/ "F count rate from solid/ count min-' (mg atom F)-' count min- metal fluoride and loading/ mmol g- initialb finalb expt.' calc.' (a) Reactions with OCF18F CsF 4.4 14 598 12 624 23 889 24 096 CsF 4.4 14 028 12 134 24 045 24 682 KF 4.4 14 598 13 030 24 996 25 756 KF 4.4 14 028 12 521 23 499 23 373 CsF 8.8 13 408 12 234 15 375 15 732 CsF 8.8 14 904 13 531 18 417 18 723 ICF 8.8 13 408 12 431 19 161 19 623 KF 8.8 14 904 13 892 20 769 21 314 (b)Reactions with SF318F CsF 8.8 35 955 30 089 84 903 84 221 CsF 8.8 39 731 33 761 94 312 95 217 KF 8.8 35 955 32 161 90 41 1 91 007 KF 8.8 39 731 35 920 99 841 99 576 ~~~ Reaction conditions 1 h using 0.5 g supported fluoride and 1 mmol (300 Torr initial pressure) volatile fluoride.fraction of 18F exchange (%) 34 34 27 27 18 19 15 14 44 47 31 30 Error < f2%.'Determined as described in the text. Error 51%. reproducibility obtained. The data were broadly in agreement with a previous, more limited, e~amination.~ B.E.T. areas decreased as the metal fluoride loading increased and areas of KF/y-alumina were slightly greater than those of CsF/y- alumina. However, the effect of increasing the MF, M =Cs or K, loading was far more marked than the effect of a change in cation. Samples prepared by impregnation of y-alumina by M[OCF(CF,),] in MeCN, followed by thermal decompo- sition of the alkoxide and calcination of 523 K, resulted in slightly smaller B.E.T. areas for each loading and cation but in other respects showed identical behaviour.Samples that were calcined at 773 rather than 523 K had smaller surface areas. Representative results for samples prepared by the non- aqueous route were: loading 4.4mmol g-', 40-50 (Cs) and 68-81 (K) m2 g-'; loading 8.8 mmol g-', 19-26 (Cs) and 32-41 (K) m2 g-'. Structural studies of CsF/y-alumina and KF/y-alumina by powder XRD and TEM indicated, in agreement with previous XRD re~ults,~,'~F MAS NMR4,7,8 and FABMS9 studies, that metal fluoride particles were present at low loadings of metal fluoride. TEM examination of KF/y-alumina samples indi- cated that KF was well dispersed at loadings G4.4 mmol g-' (Table 2), however, the particles were mobile on the surface of the samples under the influence of the electron beam; consequently, quite large, well formed KF particles were detected (Fig.1). As expected y-alumina was present at all compositions examined (Table 2); the only other aluminium- containing species positively identified were M3A1F6, M =Cs or K, which were detected by XRD at loadings 28.8 mmol g-(Table 2). The latter finding is consistent with the previous XRD study of KFly-al~mina.~ Strong bands in the IR spectra of MF/y-alumina samples, loading >5.5 mmol g-I, at v,,, 580 and 395 cm-' indicated the presence of [A1F613-by comparison with the spectra of unsupported [A1F6I3-salts;'* these bands were absent in samples of lower MF loading whose spectra were almost identical to that of calcined y-alumina. Raman spectra con- tained a very broad band, vmaX 596cm-', at MF loadings >6.0 mmol g- ', which was not observed at lower loadings.The band could be due to v1 of [A1F6I3- although it occurred at a higher wavenumber than v1 in unsupported [A1F6I3- salts.l9 27Al MAS NMR spectrum of MF/y-alumina samples were similar at all MF loadings (4.4-20.0 mmol g-I) examined and Fig. 1 Electron micrograph and diffraction pattern for KF/y-alumina, loading 4.4 mmol g-consisted of two broad signals due to 27Al nuclei in tetrahedral and octahedral environments2' (Table 3). The signals at lower applied frequency resembled a combination of those from calcined y-alumina and K3A1F6, prepared by a literature method,21 and thus the component ca. Sppm was assigned to 27Aloc, in y-alumina and that at ca.-1.6 ppm to [A1F613-. The latter component appeared to grow in intensity with increasing MF loading and the higher-frequency component was observed only as a shoulder in the 20.0 mmol g -samples. Although the identification of the [A1FJ3-anion by 27Al MAS NMR spectroscopy must be regarded as tentative, the line shapes were distinctly different from those obtained from fluorinated y-alumina samples and there was no evidence for AlF3 nor for any other simple fluoroaluminate(Ir1) species. Interaction of y-Alumina-supported Caesium and Potassium Fluorides with Volatile Lewis Acids Exposure of y-alumina-supported metal fluorides, prepared from aqueous solution, to sulphur tetrafluoride at room temperature resulted in very exothermic reactions, particularly using samples in which the MF loading was low; for example, the temperature of the outer surface of the Pyrex reaction vessel rose to >400 K when the MF loading was 4.4 mmol g- l.Thionyl fluoride and sulphur dioxide were identified as products in all cases; unchanged SF, was identified only at MF loadings >5 mmol g- '. Similar behaviour was observed using F2S0 or carbonyl fluoride, although the reactions were less exothermic. The changes in composition of the volatile products with variation in the loading of CsF or KF (Fig. 2) indicated that although hydrolysis occurred in every case its extent was greater when M =K. In both materials the extent of hydrolysis decreased as the loading of MF increased.There was no evidence for SiF, in the vapour phase, suggesting that HF, which must have been the other product from the hydrolysis reactions, was retained completely by the surface. The IR spectra of the solids after removal of volatile material and pumping at room temperature, showed that the anions [SF,]-(ref. 22) and [FS02]-(ref.23), [FSOJ, and [F3CO] -(ref. 24) had been formed from reactions that involved SF,, F2S0 and F,CO, respectively. There was no evidence for the formation of the anions [HF2]-, [F3SO]- or [FC02]-. Admission of 35SF4, F235S0, or 35S02 to MF/y-alumina samples at room temperature led to rapid growth in 35S surface count rates indicating that adsorption and/or reaction of the Lewis acid had occurred on the surface.When material from the vapour phase was removed, the 35Ssurface count rates fell to <30% of their maximum values, with the magnitude of the decrease depending on the adsorbate used, the identity of MF, and its loading. Multiple treatment of CsF/y-alumina (4.4mmol g- ') with 35SF4 (initial pressure 3 10 Torrt) indicated (Fig. 3) that although small increases in 35S activity occurred on the surface, the 35S surface count rate due to permanently retained species, presumably [SF,] -and [FS02]-, remained constant. Similar behaviour was observed when F214C0 or 14C02 were admitted to MF/y-alumina samples except that the surface interactions were complete within the time of the first 14C surface count-rate determination. Removal of volatile material after F2I4CO treatment resulted in 14C surface count rates that were <45% of the original values depending on the MF loading. On removal of 14C02, however, surface count rates decreased to the background level irrespective of loading and identity of M.In all cases the maximum 35Sor 14C surface count rates were pressure-dependent, over the range 10-350 Torr, and in contrast to unsupported CsF," saturation coverages of the surface were not observed. More importantly, the surface count rates varied with the identity of MF and its loading in a characteristic way, as illustrated for F214C0 in Fig. 4. The maximum surface count rate occurred at a loading of 5.5 mmol g-'; for a given loading CsF had a slightly larger value, and at 20.0 mmol g-' the count rate decreased to background levels when the volatile material was removed.Admitting SF318F, 18FFS0, or 18FFC0 to MF/y-alumina samples at room temperature led to immediate detection of 18Factivity from the solids. Growth in the 18F count rate t 1 Torrzl33.322 Pa. Fig. 2 Variation in the volatile product composition (mol%) after exposure of (a)SF,, (b) OSF, and (c) F,CO (1 mmol in each case) to CsF/y-alumina and KF/y-alumina samples (0.5g), loadings in the range 1.1-15.0 mmol g-'. Reaction conditions, 1 h at room tempera- ture, initial pressure 300 Torr. ., OSF,; & SO,; *, F,CO;SF,; 0, 0,co2 84 -h s-E 72-Y !2 601 a 48 > 36-.-0 c..-rn g 24-E 8 l2 t 0 961 884 h s-8 72.Y rn c 60 L P Q) -48, 0 r-O 36 .-0 Y.-24 E 8 l2 I1 0 96-cc 1 84 -s-72-Y rn Y 02 60-2 P ," 48-c -0 5 36-C .-Y.-24-E8 12 -I 0 J. MATER. CHEM., 1991, VOL. 1 2.5 5.0 7.5 10.0 12.5 15.0 (MF/y-alumina loading)/rnmol g-' I I I I I I 2.5 5.0 7.5 10.0 12.5 15.0 (MF/y-alumina loading)/rnrnol g-' I I I I I I 2.5 5.0 7.5 10.0 12.5 15.0 (M Fly-al urn ina load ing)/rnmol g-' J. MATER. CHEM., 1991, VOL. 1 desorption time/min 60 50 40 30 20 10 0 0 10 20 30 40 50 60 adsorption tirne/min Fig. 3 Adsorption-desorption behaviour observed on exposure of CsF/y-alumina, loading 4.4 mmol g-', to "SF, at room temperature. Initial pressure of SF, =310 Torr in each case r .-C E 90r 1 1 I I I 1 zo 5 10 15 20 (MF/y-alumina loading)/rnmol g-' Fig. 4 Variation of 14C surface count rate with composition.Upper trace CsF/y-alumina, lower trace KFly-alumina. Initial pressure of F2C0=300 Torr was rapid over the first 0.5 h and slower thereafter. Removal of volatile material after 1 h led to small decreases in solid count rates but most of the activity was retained even after pumping. Specific ''F count-rate determinations for SF318F, '*FFSO, and "FFCO before and after reaction indicated that partial "F exchange had occurred between MF/y-alum- ina (8.8 mmol g-') and SF3"F or ''FFCO but not with I8FFSO. The extent of "F exchange was marginally greater when M=Cs, and in the "FFCO, MF/y-alumina systems, where measurements were possible for the loading 4.4 mmol g-', it was greater at the lower loading (Table 4).In separate experiments it was established that there was no observable "F exchange at room temperature between SF318F or "FFCO and M3AlF6, M =Cs or K, compounds under hetero- geneous conditions, therefore the ''F exchange reactions with supported metal fluorides were ascribed to the presence of labile C1'FF2C0]- or [SF418F] -anions as the intermediates. "F Exchange between unsupported CsF and "FFCO has been observed at room temperature but there was no observ- able exchange between CsF and SF3"F under identical conditions." The exchange between SF318F and supported CsF observed here may have been the result of local heating from the hydrolysis reaction leading to the reversible decomposition of the [SF,"F] -anion.The assumption that "F activity produced in MF/y-alum- ina was the result of hydrolysis of SF3"F or "FFCO on the surface, formation of [SF,"F-] and [18FS021-or ["FF2C0]-, and "F exchange was tested by calculating the solid count rates that would result from the combination of the three routes (Table4). The extent of hydrolysis was determined from vapour-phase IR and manometric measure- ments (Fig. 2) and anion formation from manometric studies and vapour phase 35Sor 14C data. Combining these with the "F data led to calculated "F solid count rates that were in satisfactory agreement with those determined experimentally (Table 4).The behaviour of the "F count rate from the solid, deter- mined after 1 h exposure to SF3"F, "FFSO, or "FFCO at room temperature, with change in the loading of MF was identical, irrespective of the Lewis acid or metal fluoride used and is illustrated for "FFCO vs. KF/y-alumina and CsF/y- alumina in Fig. 5. The largest count rates were observed at the lowest loading (0.6 mmol g-') used. They decreased sharply over the range 0.6-10.0 mmol- ', there being relatively little change thereafter. Count rates from KF/y-alumina were slightly larger than those from their CsF counterparts. Although the behaviour of MF/y-alumina samples, pre- pared from the MeCN route, towards volatile Lewis acids was not investigated in detail, it appeared to be similar in most respects to that described above.The most noticeable difference was that hydrolysis occurred to a smaller extent. Discussion The behaviour of the fluorine-containing, volatile Lewis acids towards CsF and KF supported on y-alumina indicates that the main reactions occurring at the surface are the hydrolyses SF4-+F2SO+S02 and F2CO+C02 and formation of Lewis acid-base complexes. Radiotracer experiments using the weak p-emitters 14C and 35Sprobe only the surface, since self-absorption prevents the detection of radiation from the bulk solid. Weakly adsorbed surface species are observed in all cases and perma- nently retained species in all cases except for C02. The results obtained are very similar to those of an earlier radiotracer m i 1 1 1 1 eo 5 10 15 20 (MF/y-alumina loading)/mmol g-' Fig.5 Variation of '*F solid count rate with composition. Upper trace KF/y-alumina, lower trace CsF/y-alumina. Initial pressure of F2C0=300 Torr; count rate determined after 1 h in each case investigation using unsupported CsF. lo This finding, together with the spectroscopic detection of the fluorine-containing anions that would be expected from the addition of F-ions to SF4, SO2 and F2C0, suggest strongly that F-ions are directly involved in the Lewis base adsorption sites. Although C02 and SO2 are both weakly adsorbed on the surface of calcined y-alumina, it does not appear that there is a major contribution from this source to the observed variation in 14C and 35S surface count rates with change in loading of MF/y-alumina, M =Cs or K (Fig.4). If this were the case a closer correlation with B.E.T. areas (Table 1) would have been expected. It is considered, therefore, that the surface count rate us. MF loading relationships reflect the variation in F-ion surface concentrations. The maximum surface concentration of F-ion is observed at 5.5mmol g-I for both supported fluorides (Fig.4). A loading of 5 mmol g-I of KF on alumina has been shown to correspond to the optimum reagent activity for methylation of phenol, while the greatest catalytic activity in Michael addition reactions was observed at a loading of 0.6mmol g-1.4 Our TEM study shows the existence of KF particles on the surface at loadings <5.5 mmol g-' (Table 2).Taken overall, the structural and spectroscopic evidence indicates, in agreement with previous w~rk,~~~~~ that <5.5 mmol g- MF is the major fluorine-containing species on the surface. There is no evidence, however, for monolayer coverage of the surface by MF, as has been assumed implicitly by some previous worker^.^,^ Increasing the loading of MF above 5.5 mmol g-' leads to a sharp decrease in 14C and 35Ssurface count rates (Fig.4), indicating that the conversion of MF to M3A1F6 during the impregnation process becomes more important as the loading of MF increases. The only fluoroaluminate positively ident- ified, in agreement with previous was [A1F6]3- but, rather surprisingly, TEM did not reveal its presence. Because of this, we suggest tentatively that M3AlF6 is located in the bulk material rather than on the surface.Information concerning the nature of the support can be inferred from the composition of the volatile products after exposure of SF4, F2S0, and F2C0 to the supported metal fluorides and from the variation of 18F count rates from the solids with change in MF loading. The rapid and exothermic hydrolysis reactions that occur, are consistent with the use of y-alumina to catalyse hydrolysis and methanolysis of acyl and phosphoryl fluorides under heterogeneous condition^.^' NMR spectroscopic investigation of alumina impregnated with aque- ous NH4F and subsequently calcined at high temperature, has been interpreted on the basis of the replacement of Al-OH surface groups by Al-F groups,26 and related behaviour is shown by sili~a.~~,~~ In the present work the hydrolysis reactions become less dominant as the MF loading, and thus the proportion of M3AlF6, increases (Fig.2). This is to be expected since the B.E.T. areas decrease (Table l), although if free MOH were to be formed according to eqn. (l), the reverse effect would be expected. A decrease in the 18Fsolid count rate is observed also as the MF loading increases, irrespective of the fluorine- containing Lewis acid used (Fig. 5). The reactions of '*FFSO with the supported metal fluorides do not involve 18F exchange nor formation of [F,SO]-, hence the 18F count rates for the solids must arise solely from 18Fdeposited on the surface as a result of hydrolysis.This source appears to be the major contribution in the other reactions, where 18F exchange and complexation with F-ion both occur (Table 4). Rather surprisingly, there is no evidence that HF, which is a product expected from the hydrolysis reactions, is released to the vapour phase. The nature of the surface-adsorbed HF is speculative but it is possible that it is dissociatively adsorbed J. MATER. CHEM., 1991, VOL. 1 in a manner similar to that suggested for anhydrous HCl on calcined y-alumina.28 It is noteworthy that 18Fcount rates from supported KF are generally greater at a given loading than those from supported CsF (Fig. 5 and Table 4).This is consistent with the volatile product compositions after hydrolysis (Fig. 2) and the greater B.E.T. areas of supported KF materials (Table 1). In contrast, 14C and 35S surface count rates for which the major contributions are from interactions that involve F-ions, are in the reverse order, supported CsF >supported KF (Fig. 4). The composition of the defect spinel29 y-alumina is [A12 .,(vacancy), in which the hydroxyl groups are exclusively on the ~urface.~'.~~ A recent TEM is in agreement with the earlier suggestion29 that the predominant surface plane is (1 10). Because of its amphoteric nature, a y-alumina surface can be modified by exposure to dilute solutions of aqueous electrolytes, leading to the specific adsorption of Group 1133or Group 134cations or F-anion.34c We suggest that under the impregnation conditions used to prepare supported alkali-metal fluorides, surface adsorption of F-ions is followed by further reaction leading to [A1F6I3-. Were metal hydroxides to be formed in this reaction, according to eqn. (l), the extent of hydrolysis observed on exposure of the supported metal fluorides to SF4, F2S0 or F2C0 would have been expected to increase with increased metal fluoride loading, which is contrary to the behaviour observed.In order to preserve charge neutrality, we suggest that the reaction of MF, M=Cs or K, with y-alumina to give M3A1F6 is accompanied by dehydroxylation of the y-alumina surface to give H20 and surface [Al-0-1 groups which act as specific sites for M ad~orption,~~.~~+ and which are also potentially basic sites.Calcination of y-alumina also leads to a partial dehydroxyl- ation of the surface, the extent of which depends on the calcination temperature, although complete dehydroxylation cannot be achieved without loss of the y-alumina struc-t~re.~'.~~Weakly bound surface water can be removed more easily, for example the results of a recent study of the chlorination of y-alumina, calcined at various temperatures, by anhydrous HC1 are consistent with the removal of weakly bound surface water below 373 K.28 Some of the surface water on the supported metal fluoride, however, is likely to be strongly bound, for example via hydrogen bonding to the MF particles, and will not be removed completely by calcination at 523 K.The importance of 'residual water' in determining the activity and the selectivity of oxide-supported metal fluorides3v4 and other supported reagents3' is well docu- mented. The behaviour of SF4, F2S0, and F2C0 towards MF/y-alumina materials of varying composition does not enable the effects due to the hydroxylated support surface and strongly bound water to be differentiated. Both are likely to be important. We thank SERC for support of this work, the staff of the SERC solid-state NMR service (University of Durham) for the 27Al NMR spectra, and a referee for helpful comments. References 1 J. H. Clark, Chem. Rev., 1980, 80, 429; G.G. Yakobson and N. E. Akhmetova, Synthesis, 1983, 169.2 e.g. J. Yamawaki and T. Ando, Chem. Lett., 1979, 755; J. H. Clark, A. J. Hyde and D. K. Smith, J. Chem. SOC., Chem. Com- mun., 1986,791; J. Ichihara, T. Matsuo, T. Hanafusa and T. Ando, J. Chem. SOC., Chem. Commun., 1986, 793; R.G. Sutherland, A. S. Abd. El Aziz, F. Piorko and C. C. Lee, Synth. Commun., 1987, 17, 393; L.M. Harwood, G.C. Loftus, A.Oxford and C. Thomson, Synth. Commun., 1990,30, 649. J. MATER. CHEM., 1991, VOL. 1 1077 3 e.g. T. Ando, J. Yamawaki, T. Kawate, S. Sumi and T. Hanafusa, Bull. Chem. SOC. Jpn., 1982, 55, 2504; J. H. Clark, D. G. Cork and M. S. Robertson, Chem. Lett., 1983, 1145; J. Yamawaki, T. Kawate, T. Ando and T. Hanafusa, Bull. Chem. SOC. Jpn., 1983, 56, 1885; J. H. Clark, D.G.Cork and H. W. Gibbs, J. Chem. SOC., Perkin Trans. I, 1983, 2253; D. Villemin and M. Ricard, Tetrahedron Lett., 1984, 25, 1059; F. Texier-Boullet, 19 20 21 1982,495,200; U. Bentrup, R. Stodolski and L. Kolditz, Z. Chem., 1986, 26, 187. K. Wieghardt and H. H. Eysel, 2. Naturforsch., Teil B, 1970, 25, 105; M. J. Reisfeld, Spectrochim. Acta, Part A, 1973, 29, 1923. D. Muller, W. Gessner, H-J. Behrens and G. Scheler, Chem. Phys. Lett., 1981, 79, 59. H. M. Haendler, F. A. Johnson and D. S. Crocket, J. Am. Chem. D. Villermin, M. Ricard, H. Moison and A. Foucaud, Tetra- SOC., 1958, 80, 2662. 4 5 hedron, 1985, 41, 1259; J. M. Melot, F. Texier-Boullet and A. Foucaud, Synthesis, 1987,364;D. E. Bergbreiter and J. J. Lalonde, J. Org. Chem., 1987, 52, 1601. T. Ando, S.J. Brown, J. H. Clark, D. G. Cork, T. Hanafusa, J. Ichihara, J. M. Miller and M. S. Robertson, J. Chem. SOC., Perkin Trans. 2, 1986, 1133. L. M. Weinstock, J. M. Stevenson, S.A. Tomellini, S. H. Pan, T. Utne, R. B. Jobson and D. F. Reinhold, Tetrahedron Lett., 1986, 27, 3845. 22 23 24 25 L. F. Drullinger and J. E. Griffiths, Spectrochim. Acta, Part A, 1971, 27, 1793; K. 0. Christe, E. C. Curtis, C. J. Schack and D. Pilipovich, Znorg. Chem., 1972, 11, 1679. E. A. Robinson, D. S. Lavery and S. Weller, Spectrochim. Acta, Part A, 1969,25, 151; D. F. Burow, Inorg. Chem., 1972, 11, 573; K. Garber and B. S. Ault, Znorg. Chem., 1983, 22, 2509. K. 0. Christe, E. C. Curtis and C. J. Schack, Spectrochim. Acta, Part A, 1975, 31, 1035. G. H. Posner, J. W. Ellis and J. Ponton, J.Fluorine Chem., 1981, 6 T. Ando, J. H. Clark, D. G. Cork, T. Hanafusa, J. Ichihara and 19, 191. 7 T. Kimura, Tetrahedron Lett., 1987, 28, 1421. J. H. Clark, E. M. Goodman, D. K. Smith, S. J. Brown and J. M. 26 27 J. R. Schlup and R. W. Vaughan, J. Catal., 1986, 99, 304. C. V. A. Duke, J. M. Miller, J. H. Clark and A. P. Kybett, Spectro- 8 9 10 Miller, J. Chem. SOC., Chem. Commun., 1986, 657. C. V. A. Duke, J. M. Miller, J. H. Clark and A. P. Kybett, J. Mol. Catal., 1990, 62, 233. S. J. Brown and J. M. Miller, Can. J. Spectrosc., 1989, 34, 33. K. W. Dixon and J. M. Winfield, J. Chem. SOC., Dalton Trans., 28 29 chim. Acta, Part A, 1990, 46, 1381. J. Thomson, G. Webb and J. M. Winfield, J. Mol. Catal., 1991, 67, 117. B. C. Lippens, Ph.D. Thesis, Delft University of Technology, 1961. 11 12 13 14 15 16 1989,937. T. Baird, A. Bendada, G. Webb and J. M. Winfield, J. Fluorine Chem., 1991,52,245. K. W. Dixon, M. F. Ghorab and J. M. Winfield, J. Fluorine Chem., 1987, 37, 357. F. S. Fawcett and C. W. Tullock, Znorg. Synth., 1963, 7, 119. C. W. Tullock and D. D. Coffman, J. Org. Chem., 1960,25,2016. F. S. Fawcett, C. W. Tullock and D. D. Coffman, J. Am. Chem. SOC., 1962, 84, 4275. D. D. Gibler, C. J. Adams, M. Fischer, A. Zalkin and N. Bartlett, 30 31 32 33 34 35 S. Soled, J. Catal., 1983, 81, 252. J. B. Peri, J. Phys. Chem., 1965, 69, 220. A. Reller and D. L. Cock, Catal. Lett., 1989, 2, 91. C-P. Huang and W. Stumm, J. Colloid Interface Sci., 1973, 43, 409. (a) L. Vordonis, P. G. Koutsoukos and A. Lycourghiotis, J. Catal., 1986, 98, 296; (b) Langmuir, 1986, 2, 281; (c) J. Catal., 1986, 101, 186. e.g. M. Onaka, K. Sugita and Y. Izumi, J. Org. Chem., 1989,54, 11 16. 17 18 Inorg. Chem., 1972, 11, 2325. M. E. Redwood and C. J. Willis, Can. J. Chem., 1967, 45, 389. M. Epple, W. Riidorff and W. Massa, Z. Anorg. Allg. Chem., Paper 1/03696C;Received 19th July, 1991