J. MATER. CHEM., 1994, 4(11), 1723-1730 X-Ray Absorption Spectroscopic Study of the AIPO,-5 :Ferrocene Inclusion Compound and its Thermally Decomposed Products Astrid Lund,*a David G. Nicholson," Geraldine Lambleb and Brian Beagley" a Department of Chemistry, University of Trondheim, AVH, N-7055 Dragvoll, Norway Brookhaven National Laboratory, Upton, NY 11973, USA Department of Chemistry, University of Manchester Institute of Science and Technology, P.O. Box 88, Manchester, UK M60 7QD The synthesis and decomposition of an inclusion compound, AIPO, :Fe(C,H& is reported. This, and the decomposition products have been characterised by extended X-ray absorption fine structure spectroscopy. The clustering of iron atoms either through Fe-Fe or Fe-0-Fe interactions is shown to be absent.Instead, Fe"' is sited within the double 6-rings of the 12-ring system. A model for distortions of the local lattice about Fe"' is proposed. There is intense interest in the microporous materials collec- tively known as zeotypes, a class of materials that include aluminium phosphorus oxygen frameworks AlPOs (Al, P), silicon aluminium phosphorus oxygen frameworks, SAPOs (Si, P), metal-substituted aluminium phosphorus oxygen frameworks, MeAPOs (Me, Al, P) and metal-substituted aluminium phosphorus silicon oxygen frameworks, MeAPSOs (Me, Si, P). Interest in these materials has been increased by their structural relationships to the zeolites, which enjoy wide industrial applications as ion exchangers, molecular sieves and A1P04 frameworks are built up from strictly alternating A1-0 and P-0 tetrahedra whose formal charges are mutu- ally compensated.Thus, these materials have neutral frame- works without any ion-exchange properties or strong acid sites (with the possible exception of some metal- and silicon- substituted A~POS~,~). In the specific case of A1P04-5 there appears to be some weak acidity, which has been attributed to lattice This property is of consequence in the present study. Materials of this type are attractive hosts for catalytic species because of their excellent thermal and hydrothermal stabilities (e.g. A1PO4-5 withstands calcination at 1000"C9). The thermal stability of A1P04-5 makes the material a promising candidate for some industrial appli- cations and is one of the reasons that the material was chosen in the present study.In addition, A1PO4-5 also has features that commend it as a host in the study of channel inclusion compounds, namely channel or cavity size. The structure (Fig. 1) is well defined, has a large internal surface area, and eP 00 oAl OFe Fig. 1 View of the A1Po4-5 12-ring system that consists of alternating 4-rings and 6-rings with an Fe"' atom sited in the 6-ring. The A1(0), and P(O), tetrahedra that build up the framework are shown schematically. has the potential to impose size and shape selectivity on the product distribution arising from the molecular sieving effect.lO," To date, a number of organometallic guest molecules, including ferrocene, have been adsorbed on several zeolites and some of the properties of the resulting materials have been studied.12-18 Loading an organometallic complex such as ferrocene into dehydrated AlPOs may lead to materials in which the chemical and physical properties of the guest and host are modified.Interest in encapsulating organoinetallic complexes into AlPOs stems from work carried out on zeolites. The zeolite inclusion materials contain guests ranging from atoms, ions, metal clusters and salts to coordination complexes and organometallics. Extended X-ray absorption fine structure (EXAFS) spec- troscopy using a synchrotron radiation source is ir useful method for probing the local environment of a selected atom as the EXAFS contains information about the near neighbours surrounding the target atom.The parameters extracted include interatomic distances, coordination numbers and the degree of disorder (static and dynamic). We report here the results of an EXAFS study on the A1PO4-5:ferrocene inclusion compound and the fate of the iron atoms in the thermally decomposed products. Experimental Gel Preparation A1P04-5 was prepared according to procedures previously described." The synthesis gel was prepared using pseudo- boehmite (BA Chemicals Ltd., a subsidiary of British Alcan Aluminium plc), orthophosphoric acid (Merck, 85 wt.% solu- tion in H,O), triethylamine, (TEA; Merck, <0.2% H,O) and water. The molar gel composition was 1.0 TEA: 1.0 A120,: 1.0 P2O5:4OH20.Pseudoboehmite was added to a solution of orthophosphoric acid and water simultaneously with vigorous stirring until the gel was homogeneous (typically after 2 h). Upon combining the alumina slurry with the acid solution, the pH of the precursor mixture rose with ageing as the orthophosphoric acid slowly reacted with alumina. The pH stabilised to 1.4-1.5 after approximately 3.5 h. After ,iddition of TEA to the precursor mixture (still under stirring) the pH immediately increased to over 3 and gradually climbed to its final value of 3.5. The ageing of the gel is important because of the rise in pH during this period. Homogeneity was attained after ageing for 3 h when the pH stabilised at 3.5. Crystallisat ion The gel was transferred to a stainless-steel bomb lined with a Teflon cup with cover, pre-treated at 90 "C for 24 h and heated to 200°C at autogeneous pressure for 24 hZo after which time the crystallisation was completed, and the pH increased to 7.The colourless crystalline product was recovered by filtration, washed with distilled water and dried at room temperature. Routine X-ray powder diffraction (XRD) was used to check the crystallinity and the identity of the product. Template Removal The template in the as-synthesized AlPO, was removed by calcination at 600°C for 24 h. The final product consisted of colourless crystals. X-Ray diffraction confirmed that this material was A1PO4-5. Furthermore, AlPO,-5 retained its structure after calcination at 800 "C for 24 h.Activation ofA1P04-5 It is known17 that water molecules block the pores in the non- or partially-dehydrated zeolite NaY thereby excluding ferrocene from the non-or partially-dehydrated zeolite. Therefore dehydrating the A1PO4-5 molecular sieve is a crucial step. Activation was carried out using a similar procedure as described by Ozin and Godber.'* This involved three 2h cycles of evacuation and heating to 300 "C in a sealed system. After 6 h the sample was allowed to cool to room temperature. Adsorption of Ferrocene on A1PO4-5 Ferrocene, Fe( Cp), was diffused into the microporous AlP04- 5 framework by contact with a saturated solution of ferrocene in pentane, and allowed to react for 2 h. According to the literature" the rates of diffusion of the metallocenes into the pores of zeolite Y (faujasite-like structure) are fast, approxi- mately 90% being taken up from solution within 3 min.The saturated ferrocene solution was prepared by dissolving ferro- cene (Aldrich-Chemie, 98%, 1g) in pentane (Fluka Chemica, >99.5%, 50 ml). The green A1P04-5 :ferrocene product was filtered and washed with pentane until the excess ferrocene was washed out and the washings were colourless. The preparation was carried out in a glovebox under a nitrogen atmosphere, the flow of nitrogen being dried by means of liquid nitrogen. Heated In Vacuo at 200 "C Heating A1P04-5:ferrocene in uucuo at 200°C for 24 hours produces a change in colour from green to grey. Bearing in mind that ferrocene itself decomposes at a much higher temperature (470 "C), the structural integrity of the guest molecules in A1PO4-5 is of interest.Thermal Decomposition The thermal decomposition of the A1Po4-5:ferrocene inclusion compound was carried out by heating samples in a sealed evacuated quartz ampoule and also by heating in open air. The in vacuo decomposition was carried out by placing the ampoule in a stainless-steel vessel and heating to 600°C for 15 min. The product was black; this can be attributed to high-carbon content products (this material is discussed later). The second sample was decomposed at 600°C in air for 24 h in order to remove the high-carbon materials by oxidation, and to examine the effect of oxidising the organic residues at elevated temperatures on the iron environments. This product was orange-brown.The samples were analysed by atomic absorption spec- troscopy (AAS) using a Perkin-Elmer, Zeeman 5100 instru- ment. The analysis yielded ca. 1-2 wt.% iron in the AlP04- 5 :ferrocene compound and both its decomposed A1P04-5 :Fe products. This was confirmed by X-ray fluorescence analysis. J. MATER. CHEM., 1994, VOL. 4 The AlPO, samples were structurally characterised by XRD in order to confirm the microporous AlPO,-5 structures. Model Compounds Ferrocene, Fe(Cp), and iron (111j phosphate, FePO, were chosen as model (reference) compounds. The structures ar? knownZ1q2' and the Fe-C bond values are 2.064 and 2.045 A in the gas and solid phase, respectively (Table 1 ).The structure of FePO,, which is isotopic with the condensed AlPO, phase berlir~ite,'~is thus also related to that of a-quartz but with a nearly doubled c-axis. The Fe and P atoms in FePO, are tetrahedrally coordinated to oxygen atoms. X-Ray Powder Diffraction Data Collection XRD data were collected at the University of Trondheim, NTH using a Philips X-ray generator, a diffractometer employing a Cu tube as X-ray source and a graphite crystal monochromator or at the University of Manchester Institute of Science and Technology using a SCINTAG powder diffractometer. Extended X-ray Absorption Fine Structure Preparation of Samples and Data ColEection EXAFS data collection was carried out at the Daresbury Synchrotron Radiation Source (SRS), and the National Synchrotron Light Source (NSLS), Brookhaven.Data were obtained on stations 9.2 (Daresbury) and X11A (Brookhaven) using the iron K-edge (A=1.74334 A; energy =71 12 eV). Double-crystal silicon (220)and order-sorting silicon crystal (111) monochromators were used to scan the X-ray spectra on stations 9.2 (Daresbury) and X11A (Brookhaven), respect- ively. At Daresbury, ion chambers were used for detecting the intensities of the incident and transmitted X-rays. The first ion chamber was filled with 20% Ar (24.1 Torrj and the second ion chamber with 80% Ar (174.2Torr) and both chambers were filled to atmospheric pressure with He. At Brookhaven fluorescence data were collected using a Lytle dete~tor.'~ The experiments were performed with electron beam ener- gies of 2.0 and 2.5 GeV, and maximum stored currents of 205 and 21 1mA for Daresbury and Brookhaven, respectively.The harmonic rejection for the data collections was set to 70%. The materials that had been decomposed in vacuo were prepared in the form of sealed samples for EXAFS experiments immediately after opening the evacuated ampoules. One of the ampoules was opened under a dry nitrogen atmosphere and the EXAFS sample sealed in that atmosphere in order to compare its EXAFS spectrum with those that had been exposed to the atmosphere. The amount of material used to prepare the samples for transmission XAS was calculated from element mass fractions and the absorption coefficients of the elements25 just above the absorption edge to give log(Io/I) of almost unity.The well powdered samples were mixed with boron nitride so as to obtain a sample thickness of ca. 1 mm and placed in aluminium sample holders and Table 1 Fe-C bonds and root mean square vibrational amplitudes of ferrocene according to the literature sample method ref. rF€& %& solid X-ray 21 2.045 (k0.0l)a -gas phase GEDb 22 2.064 ( f0.003p 0.062 (k0.001)" 'The standard deviations in parentheses include the effect of errors in the electron wavelength. bGas-phase electron diffraction. J. MATER. CHEM., 1994, VOL. 4 held in place by Sellotape. Spectra of the inclusion compound and its thermally decomposed products (i.e. unknowns) were collected in the fluorescence mode, because they are relatively dilute in Fe content (1-2%). Samples (100 mg) were placed on a surface area of 1.2 cm2.Results X-Ray Powder Diffraction The X-ray powder diffractogram confirms that the product of r,! 2the synthesis is hFxagonal A1Po4-5 with the cell dimensions: -O.4 4 6 8 10 12 14 a =b = 13.697(1) A, c =8.375(1) A and y = 120" which are clo5e to the paiameters found in the literature26: a =b =13.726 A, c =8.484 A and y =120". The A1P04-5:ferrocene material gives an XRD pattern that closely matches the A1PO4-5 diffractogdam. The cell parameters a= b =13.753(2) A, c= 8.410(3)A and y= 120" for the latter are consistent with the A1P04-5 framework being essentially unmodified by the guest molecules, and the small changes in cell parameters reflect only minor adjustments of the framework. The decomposed black sample matched the AlPO,-5 XRD.A decrease in weight was not found upon further heating to 900°C indicating that the AlPO heated to 600°C is C-free. XRD patterns of the brown A1P04-5 :Fe sample also match the A1P04-5 diffractogram. Extended X-ray Absorption Fine Structure The EXAFS data were analysed using standard procedures within the EXCURV90 program based on a curved-wave formalism as described in detail in the literat~re.~~.~~ Data were measured, summed, corrected for dark currents and converted into k-space by using the EXCALTB29 routine. The EXBACK29 program was used to normalise the data and extract the EXAFS functions Xybs (E-&) by background subtraction.In the preliminary stages of refinement, Fourier filtering (FF) using various ranges was carried out in order to isolate the contributions to the EXAFS from the different distances. A Gaussian window function was used. Fig. 2 shows the unweighted raw spectra of AlPO4-5:Fe(Cp), and its two decomposed products. The k3 weighted experimental spectra of the three synthesised samples and their Fourier transforms are given in Figs. 3-5. Model Compounds The model (reference) compounds chosen were Fe(Cp), (Daresbury and Brookhaven) and FePO, (Brookhaven). EXCURV90 calculates phases for their individual scatterers, Fe and the C, P and 0 atoms by ab initio methods. The reference compounds were used to calibrate the parameter sets, which were used in calculating the EXAFS of unknown samples.In Fe(Cp), and FePO,, the actual number of each kind of atom (C, 0 or P) within the shells was kept constant at the crystal!ographic values. The Debye-Waller-type factor, A of 0.007 A2 obtained for the Fe(Cp), sample is comparable to the vibrational amplitude, (u =0.0077 A2) obtained from gas- phase electron diffraction (Table 1).22 A1PO4-5:Fe(Cp), (Daresbury) The parameters obtained from the Fe(Cp), model compound were transferred to the A1PO4-5 :Fe(Cp), sample. The refined structural parameters for the latter confirm that the structural integrity of the ferrocene molecule in the host framework is maintained (Table 2). Fig.3 shows the calculated data fitted to the experimental EXAFS. 412t I11 2 4 6 8 10 12 14 0.2 0.0 -0.2 -0.4 Is2 4 6 8 10 12 14 k IA-' Fig. 2 Raw spectra (unweighted) for (a)AlPO,-5 :Fe(Cp),, (b AlP0,-1 5 :Fe (black) product and (c) AlPO,-5 :Fe (brown) product The present work describes two decomposition routes for the A1PO4-5:Fe(Cp), inclusion compounds. The EXAFS results of these decompositions are discussed below. Material (1) A1P04-5:Fe-decomposed in vacuo (Brookhaven). The EXAFS spectra of two samples of this material were collected; one being the EXAFS sample pre- pared in a dry nitrogen atmosphere. The two spectra were similar thereby showing that the iron atoms are locked into the structure and therefore not readily available to atmos- pheric reactions under the conditions prevailing here.Clearly the possibility of the existence of iron-carbon containing materials in this sample has to be considered. As noted above, we find that the iron atoms are dispersed within the lattice of the AlPO. We suggest that the blackening of this material is due to tar-like decomposition products of cyclopen tadiene that adhere to the external surfaces of the AlPO. Accordingly, we have not used a model that incorporates coordination to carbon and the EXAFS of the A1P04-5 :Fe (black) sample was fitted using the transferable parameters of the model compound FePO,. The model parameter set is given in Table 2. A region of 1-4A was isolated from the EXAFS spectrum by means of ff.The short-distance region includes the nearest neigh!ours, 0, P and A1 (Fig. 4). Coordination shells beyond 4A contribute only weakly to the EXAFS. Refined parameters are given in Table 2. Two different Fe-0 J. MATER. CHEM., 1994, VOL. 4 -1o.o+ I I I I I I I I 4.0 6.0 8.0 10.0 1; kfA-' 2-o/ (b) 2.0 3.0 4.0 5.0 6.0 RIA Fig.3 (a) Background subtracttd k3 EXAFS spectrum and (b) its Fourier transform (window 1-5 A) for the A1P04-5 :Fe(Cp), material. Experimental (-) and theoretical (---). -8.0 6.0 8.0 10.0 12.0 RIA Fig.4 Background subtrFcted k3 EXAFS spectrum and its Fourier transform (window 1-4 A) for the A1P04-5 :Fe (black) product (1). Experimental (-) and theoretical (---). I I I I -+-8.01 ! 6.0 8.0 10.0! k IA-' m0 (b) o) 0.8 D3 c.-c 0.6 E EZ 0.4 c v)c -2 0.2 0.0 R/A Fig.5 (a) Background subtracted ).' EXAFS spectrum and (b) its Fourier transform (window 1-4 A) for the A1P04-5:Fe (brown) product (2).Experimental (-) and theoretical (---). distances (1.95 and 2.59 A) within the host are discernible with multiplicities of 4 and 2 for the first and second oxygen shell, respectively. There are no clusters present, because Fe-Fe flistances are absent. The composite Fe-0 distance of 1.95 A indicates bonding of iron to oxygen atoms of the framework in the host channel. Statistical tests3' were used to determine whether addition of the shell made a significant improvement to the fit. All the shells added were significant at the 99% probability level, and therefore included. Material (2) AlPO,-5 :Fe-decomposed in uir (Brookhaven).The EXAFS of the AlPO,-5 :Fe (brown) sample was Fourier filtered and the data analysis carried out in the same manner as for the black AlPO,-5 :Fe unknown as described above. The results of the refinement are presented in Table 2. Fig. 5 shows the calculated data fitted to EXAFS data. The first peak in the Fourjer transform (Fig. 5) was fitted to an Fe-0 distance of 1.91A with a multiplicity 4. The second and Jhird peak both include 2 Fe-0 distances of 2.46 and 2.64A. As we see this is a very narrow range and the seFond peak consists more or less of a mean distance of 2.6 A. The last peak included in the Fpurier transform corresponds to an Fe-P distance of 3.22 A with a multiplicity of 3.Statistical tests3' were carried out and all the shells added were significant at the 99% probability level and therefore included. Discussion The ferrocene guest molecules of the inclusion compound A1P04-5:ferrocene occupy positions within the host channels. It is known that the A1P04-5 structure is stable up to 1000 0C,9and the XRD of the present materials confirm that the structure is maintained after decomposing the inclusion material at 600 "C.Ferrocene itself is a very stable molecule, J. MATER. CHEM., 1994, VOL. 4 Table2 Refined structural parameters for Fe K-edge EXAFS data for the Fe(Cp), and FePO, model compounds and A1P04-5:Fe(Cp), and its decomposed product samples back Re sample scatterer EdeV NU rb/A A'lA2 FId ("/I Fe(CP)2f Fe (CP)zg FePO,g A1PO4-5:Fe(Cp), A1P04-5:Feh A1P04-5:Fe' C C 0 P 0 C 0 0 P A1 0 0 P 0 17.30 12.88 22.76 2 1.23 26.83 26.85 10 10 4 4 4 10 4.2(1) 2.96( 10) 2.86(22) 2.81 (30) 2.44( 15) 2.33( 14) 3.22(20) 4.09(9) 2.047( 1) 2.043( 1) 1.850( 1) 3.118(5) 3.498(9) 2.058(1) 1.952( 1) 2.594( 2) 3.052(5) 3.285(8) 1.9 14( 2) 2.465(5) 2.640( 5) 3.098(5) 0.0067( 2) 0.0062( 2) 0.007 1 (2) 0.0213( 12) 0.0147( 21) 0.009 1 ( 3) 0.01 74 ( 2) 0.0155( 6) 0.2089(17) 0.0330( 3 1 ) 0.0154( 4) 0.0097( 11) 0.0063(10) 0.0219( 12) 1.60 1.94 1.33 1.40 0.16 0.74 15.33 19.29 19.59 14.90 11.98 22.45 a Multiplicity. Distances, r, estimated standard deviations in parentheses.Note that systematic errors in bond distances arising from data collection and analysis are ca. ?0.02-0.03 A for well defined shells. Debye-Waller-type factor. dFit index = NPT [Resi kWTI2, where Res, i=l =Residual =j(obs -xcalc, NPT =number of data points and WT =integral weighting. Residual =R-factor, R= {c[(slobs -j(,,kc)k~]2}/ {I[(j(obs)k3]2}. Fe(Cp), Daresbury data. Fe(Cp), and FePO, Brookhaven data. 'A1P04-5: Fe after decomposition of ke(Cp), at 600T in V~CUUM. A1PO4-5:Fe after decomposition of Fe(Cp), at 600 "C in air. decomposing at ca. 3470 "C3' The decomposition of ferrocene sandwich. We note that heating this material to 200°C results guest molecules within the AlPO raises the question as to the in its colour changing from green to grey (see belob).This manner in which iron is incorporated into the AlPO. This is also distinguishes A1PO4-5:ferrocene from the zeolite one of the questions addressed below. inclusion material. Also of particular interest, and central to Of the AlPO/SAPO :ferrocene systems in~estigated~~,~~ this work, is whether decomposition leads to clustering ofthe AlP04-5:ferrocene material stands out because it is green iron either through direct Fe -Fe bonding or through whilst the others exhibit the yellow-brown colour of ferrocene. Fe-0-Fe bridging. This question is addressed below. Clearly, something in the behaviour of the A1P04-5 framework towards ferrocene is different from the others. It has already been remarked (see above) that A1PO4-5 exhibits weak acidity, EXAFS Data possibly due to lattice imperfections.It is also known that under acidic conditions ferrocene is oxidised to the ferricenium The parameters extracted from the EXAFS data lor the which has an intense blue colour for dilute A1PO4-5:ferrocene and its thermally decomposed PIoductscation F~(CP)~+, samples, while more concentrated samples are red.31 Our are listed in Table 2. Before discussing the details of the supposition is that the green of A1P04-5 :ferrocene arises from decomposed material we turn first to the precursor, the a mixture of ferrocene and ferricenium. Only minor amounts inclusion compound itself. of ferricenium are needed to produce a green material because the colour of ferricenium is much more intense than the colour A1PO4-5:Fe(Cp), of ferrocene. Support for the view that ferricenium is present Moller et d3' have studied the intracavity chemistry of is given by Mossbauer studies34 on the materials (see below) organometallic fragments including ferrocene in different acid and also from literature studies on zeolite reactions with forms zeolite Y by EXAFS.They obtained Fe-C distances ferro~ene.'~~'~~'~~~~ of 2.01 A for a half-sandwich fragment attached to the zeolite. From this it would seem that the environment around These workers did not collect EXAFS data for the zeolite :fer-ferrocene in the A1PO4-5 channels significantly modifies the rocene inclusion compound but instead took up data on a chemical stability of that molecule and chemical reaction can sample that had been heated to 200°C.In the present study take place. Mossbauer spectroscopy indicates that ca. 25YO EXAFS data pertaining to the A1P04-5:ferrocene and the ferricenium is present.34 This would mean that certain pos- A1P04-5:ferrocene heated at 200 "C in vucuo were collected itions within the channels constitute different environments and it was found that complete ferrocen: molecules are present than those occupied by the majority of the ferrocene molecules. with Fe-C distances of 2.06 and 2.05 A for the two samples, If the supposition that A1Po4-5 does actually contain respectively. The small amount of ferricenium was not Bronsted acid sites is correct it does not seem unreasonable detected. Related studies36 on A1PO4-5 :cobaltocene channel to couple the ferricenium sites with Bronsted acidity.It is inclusion compounds also show that the metallocene is present known from zeolite chemistry35 that the inner surface of the as such. acid forms are highly reactive. Moller et have studied the The Debye-Waller-type factor, A, from the EXAFS data chemistry of ferrocene in zeolite Y cavities. They conclude analysis (EXCURV90) can be compared with the root mean that thermally decomposing ferrocene at 200 "C in partially square vibrational amplitude, u, obtained from gas-phase acidic zeolite Y results in the formation of zeolite-attached electron diffraction (where u,= u of EXAFS). This calculation half-sandwich fragments, Fe+(Cp)(OZ),- . These workers sug-gives an A value of 0.0077 A2 for the gas phase.FOFAlP04-gest that the half-sandwich FeCp is bonded to two zeolite 5:Fe(Cp), the Debye-Waller-type factor is 0.0091 A2. These oxygen atoms thereby producing FeCp(OZ), . Although this values are similar within experimental error, and they show material is also green, our EXAFS data clearly show that that on going from the gas phase to the solid state the more A1PO4-5:ferrocene contains ferrocene and not a bound half- distant atoms of the lattice play only a minor role. Moreover, J. MATER. CHEM., 1994, VOL. 4 the gas phase value would seem to indicate that this effect is indeed negligible. The room temperature Mossbauer results3, reveal that molecular axes of the ferrocene guest molecules reorientate ca. > lop8 s-') rapidly within the channel of the A1PO4-5 host.This is reminiscent of the thiourea :ferrocene ~ystem.~~,~~ The same experiment is also consistent with the presence of ferricenium cations at ca. 25% of the ferrocene concentration. This lends support to the suggestion (see above) that the intense green colour indicates the presence of ferricenium in the sample. One of the aims in preparing A1P04-5:ferrocene was to study the decomposition of ferrocene within the A1PO4-5 molecular sieve. This is discussed in some detail below. Calcined A1P04-5:Fe(Cp), Fe-Fe bond distances in a material depend on the nature qf the metal environments. Thus, in a-Fe the distance is 2.48 A but in !morphous Fe metallic glasses a typical Fe-Fe distance is 1.8 A.39 This large decrease in bond length is consistent with increased covalent character of the bonds between Fe and the nonmetallic impurities in the glass.Clearly, we have to examine the possibility that formation of small metal clusters have been formed within the channels, and that interactions between them and the channel atoms also could produce short Fe-Fe distances. Hence, we have to examine the EXAFS for evidence of short and long Fe-Fe interactions. The first shell contribution to the EXAFS can only be fitted to oxygen-backscatters. From the parameters for the two decomposed materials (Table 2) it is clear that iron clusters are not formed on decomposing the inclusion compounds since Fe-Fe bond distances are absent. The absence of Fe---Fe interactions shows that the metal atoms are well dispersed in the structure and accordingly significant aggre- gation through oxygen-bridged Fe ---Fe clusters [for example in the form of FeO(0H)-based species] is excluded. We therefore conclude that the majority of the iron atoms are distributed within the structures at considerable distances from each other.Material (1 ) A1P04-5:Fe-decomposed in vacuo. The EXAFS analysis shows that iron is bonded to four oxygen atoms with longer range interactions to two additional oxygen atoms. Although, EXAFS (neglecting multiple scattering) does not contain information on the spatial distribution of the oxygen atoms about the central atom, there is information contained on this aspect in the pre-edge region.,' For some iron environ- ments there is a pre-edge feature that is assigned to Fe ls+3d transition^.^^ Although these transitions are forbidden by dipole selection rules they become allowed when the Fe 3d and 0 2p states mix.It has been shown that the position of the pre-edge peak42.43 is sensitive to the iron valence state and that its intensity is sensitive to the number of ligands and to the symmetry of the metal environment. For example, the intensity of the pre-edge feature is several times larger for tetrahedral Fe"' compounds than for octahedral Fe"' com-.~~pounds. Roe et ~ 1 have measured the areas of the pre-edge peak for 28 Fe"' complexes of varying coordination. The largest areas are shown consistently for tetrahedral environ- ments and the smallest for six and seven coordination.The decrease in area in going from four coordination to six coordination reflects the ligand field of the various symmetries. The pre-edge peak can therefore be used as a diagnostic test for tetrahedral Fe"' environments. A key example (FePO,, berlinite, a-quartz structure) is shown in Fig. 6 for a tetrahedral Fe"' environment. In berlinite the average 0-Fe -0 angle is tetrahedral within experimental error, although the individual I I I II II I I I 7100 7110 . 7120 7130 7140 energyleV Fig. 6 Normalised pre-edge intensity of (a) FePO, and (b) AIPO,: Fe (black) product and (c) AlPO, :Fe (brown) product angles range from 105.5 to 114.8'. The actual point symmetry is C2 (e.g.mixing of Fe 3d and 0 2p states, see above), but is close to Td.The distortion from the ideal angle allows mixing of orbitals that now have the same irreducible representation under the operations of the point group.Also shown is the pre-edge region for the A1P04-5 :Fe material ( 1). The edge energy (7124 eV) is similar to that for berlinite (7123 eV) and therefore defines the valence state as Fe"'. Comparison of the pre-edge regions of Fig. 6 shows that the Fe(O), environment in A1P04-5 is not tetrahedrally coordinated Fe"' and is consistent with the 4Fe-0 and 2Fe-0 distances derived from the EXAFS. The high temperature decomposition of ferrocene in the A1PO4-5:ferrocene takes place in an environment made chemically aggressive by the elevated temperature. From the EXAFS and from chemical analysis it is clear that ferrocene decomposed within the channels and was not sublimed out of the AlPO,.The availability of both Bronsted (see above) and Lewis sites makes possible the attachment of Fe"' to the channels. By using molecular modelling (INSIGHTII/ DISCOVER software package45) it is possible to visualise iron atoms coordinated to four framework oxygen atoms at Fe-0 distances that ace consistent with the observed com- posite distance of 1.95 A. These distances, which are longer than those in berlinite (Table 2),23 are characteristic of higher coordination numbers as indicated by the pre-edge data (Fig. 6). The distances derived from EXAFS show that iron is positioned within the double six-ring of the A1P04-5 structure.This is illustrated in Fig. 7 which shows details of the iron coordination. The iron atom is displaced from the centre of the ring thereby distorting the local lattice. The interaction between iron and A1P04-5 can be understood in terms of coordination from oxygen atoms of the P-0-A1 bridges to Fe3+, the formal charge being compensated by the anionic framework charges caused by the lattice imperfections (see above). A similar picture has been suggested by Lee et al. for CUSAPO,-~.~~ The two longer Fe---0 distances of 2.59 A reflect weak interactions within Fe -0-P and/or Fe -0-A1 bridging. Consistent with this interpretation are the Fe ---A1 and Fe ---P distances listed in Table 2.Interactions of this type have been described as secondary bonds by Alco~k.,~ These bonds are defined as those interatomic distances, often regarded as non- J. MATER. CHEM., 1994, VOL. 4 Fig. 7 Schematic model illustrating the displacement of the Fe"' atom from the symmetry axis of the six-ring as the ring system locally distorts from the symmetrical arrangement (left) to the unsymmetrical environment (right). Bonds between the atoms are not shown. Note that only every other P---A1 distance is bridged by an 0 atom. bonded distances, which are longer than normal bonds but shorter than van der Waals distances. Examples of other long distances in zeotypes include the Co-0- (Si,AI) bridges ~ [2.36( 3) A] found in cobalt(1r) exchanged zeolite A.48 Material (2) A1P04-5:Fe-decomposed in air.The AlP04- 5 :ferrocene decomposed jn air also yields four composite Fe-0 distances at 1.91 A that are similar to those in the above material (1). However, the iron environments differ somewhat because sample (2) contains four secondary Fe--- 0 interactions (Table 2). The pre-edge feature has also been investigated for the AIP04-5:Fe (2) material and Fig. 6 shows the pre-edge region. The edge energy (7127eV) is close to that for berlinite (7123 eV) and consistent with Fe"'. Furthermore, the pre-edge regions of Fig. 6 shows that the Fe(0)4 environment in the AlPO is not close to tetrahedral Fe"' but has a higher coordination. This lends support to the 4Fe-0 and 3Fe-0 interactions derived from the EXAFS (see above).The differ- ences between the two samples are small and can be attributed to adjustments of the local lattice about iron during the much longer period of annealing. Conclusions It is clear from this study that decomposition of ferrocene included in an A1P04-5 lattice both in the absence and presence of oxygen results in Fe3+ moving into the double six-rings of the framework. Bridging to six or more oxygens of the surroundings leads to irregular coordinations about Fe3+. The formal charge is balanced by anionic framework charges introduced via lattice defects. It is also clear that iron clusters are not formed within the molecular sieve when the guest molecule contains a single metal atom. A major factor that probably prevents clustering is the spacing of the ferrocene molecules along the channel axes.T>ese spacings are dictated by the periodic variation (8.410 A, c-direction) of the potential. Hence, if clustering is required, the most likely route for producing small clusters of metal atoms within the AlPO would seem to be to incorporate polymetal compounds into the host channels prior to calcination. Support of this work by the Norwegian Research Council for Humanities and Science (NAVF), VISTA (STATOIL) and the Nansen Foundation is gratefully acknowledged. We also appreciate the constructive comments of the referees. Part of this work was performed at X-11A Beamline at the National Synchrotron Light Source and is supported by the US Department of Energy under Contract numbers DE-AC02- 76CH00016 and DE-FG05-89ER45384.References 1 R. M. Barrer, Zeolites and Clay Minerals as Sorbents and Molecular Sieves, Academic Press, London, 1978. 2 Y. Murakami, A. Lijima and J. W. Ward, New Developments in Zeolite Science and Technology, Kodansha, Tokyo, 1986. 3 Studies in Surface Science and Catalysis, ed. P. J. Grobet, W. J. Mortier, E. F. Vansant and G. Schulz-Ekloff, Elsevier, Amsterdam, 1988, p. 37. 4 E. M. Flanigen, B. M. Lok, R. L. Patton and S. T. Wilson, Pure Appl. Chem., 1986,58,1351. 5 C. W. Lee, X. Chen and L. Kevan, Catal. Lett., 1992,15,75. 6 G. Dworezkov, G. Rumplmayr, H. Mayer and J. A. Lercher, Adsorption and Catalysis on Oxide Surfaces, Proceedings of a Symposium, Uxbridge, June 28-29, 1984, ed.M. Che and G. C. Bond, Elsevier, Amsterdam, 1985, p.163. 7 A. Bhattacharya, J. Das, S. Mitra and S. K. Roy, J. Chem. Tech. Biotechnol. 1992,54, 399. 8 V. R. Choudhary and D. B. Akolekar, J. Catal., 1987,103,115. 9 S. T. Wilson, B. M. Lok, C. A. Messina, T. R. Cannan and E. M. Flanigen, J. Am. Chem. Soc., 1982,104, 1146. 10 D. W. Breck, Zeolite Molecular Sieves, Wiley, New York, 1974. 11 R. Szostak, Molecular Sieves Principles of Synthesis and Zdentijcation, Van Nostrand Reinhold, New York, 1989. 12 P. A. Jacobs, Stud. Surf. Sci. Catal., 1982, 12, 71. 13 T. Bein and P. A. Jacobs, J. Chem. Soc. Faraday Trans. I, 1983, 79, 1819. 14 T. Bein and P. A. Jacobs, J. Chem. Soc. Faraday Trans.1, 1994, 80, 1391. 15 G. A. Ozin and J. Godber, ACS Symp. Ser., 1986,307,212. 16 G. A. Ozin, A. Kuperman and A. Stein, Angew. Chem., Int. Ed. Engl., 1989,28, 359. 17 G. A. Ozin and C. Gil, Chem. Rev., 1989,89,1749. 18 G. A. Ozin and J. Godber, J. Phys. Chem., 1989,93,878. 19 A. F. Ojo, J. Dwyer, J. Dewing, P. J. O'Malley and A. Nabhan, J. Chem. SOC., Faraday Trans, 1992,88,105. 20 N. J. Tapp, N. B. Milestone and D. M. Bibby, Zeolites, 1988, 8, 183. 21 J. D. Dunitz, L. E. Orgel and A. Rich, Acta Crystallogr., 1956, 9, 373. 22 A. Haaland and J. E. Nilsson, Acta Chem. Scand., 1968,22,2653. 23 H. N. Ng and C. Calvo, Can J. Chem., 1975,53,2064. 24 F. W. Lytle, Nucl. Instr. Methods Phys. Res., 1984,226, 542. 25 International Tables for X-Ray Crystallography, The Kynoch Press, Birmingham, UK, 1962, p.175.1730 J. MATER. CHEM., 1994, VOL. 4 26 27 28 29 30 31 32 33 34 35 J. M. Bennett, J. P. Cohen, E. M. Flanigen, J. J. Pluth and J. V. Smith, ACS Symp. Ser., 1983,218, 109. S. J. Gurman, N. Binsted and I. Ross, J.Phys. C: Solid State Phys., 1984, 17, 143. S. J. Gurman, N. Binsted and I. Ross, J. Phys. C: Solid State Phys., 1986,19,1845. N. Binsted, S. J. Gurman and P. C. Stephenson, SERC Daresbury Laboratory Programs: EXCALIB, EXBACK and EXCURV90. R. W. Joyner, K. J. Martin and P. Meehan, J. Phys. C: Solid State Phys., 1987, 20, 4005. M. Rosenblum, in The Chemistry of Organometallic Compounds, A Series of Monographs, ed. D. Seyferth, John Wiley, New York, 1965. A. Moen, Ph.D. Thesis, The University of Trondheim, AVH, to be submitted.A. Lund, Ph.D. Thesis, The University of Trondheim, AVH, sub-mitted February 1994. A. Lund, D. G. Nicholson, R. V. Parish and J. P. Wright, Acta Chem. Scund., in press. K. Moller, A. Borvornwattananont and T. Bein, J. Phys. Chem., 1989,93,4562. 37 38 39 40 41 42 43 44 45 46 47 48 M. D. Lowery, R. J. Witterbort, M. Sorai and D. N. Hedrickson, J. Am. Chem. SOC., 1990,112,4214. S. J. Heyes, N. J. Clayden and C. M. Dobson. J. Phys. Chem., 1991,95,1547. S. G. Saxena and K. B. Garg, Phys. Status Solid1 A, 1985,87, K25. D. Sunil, J. Sokolov, M. H. Rafailovich, R. Kotyuzhanskii, H. D. Gafney, B. J. Wilkens and A. L. Hanson, J. Appl. Phys., 1993,74,3768. A. Bianconi, in EXAFS for Inorganic Systems, ed. C. D. Garner and S. S. Hasman, Science and Engineering Research Council of Daresbury Laboratory, Daresbury, 1981, 13. G. A. Waychunas, M. J. Apted and G. E. Brown, Phys. Chem. Miner., 1983, 10, 1. N. Motta, M. De Crescenzi and A. Balzarotti, Phys. Rev: B, 1983, 27,4712. A. L. Roe, D. J. Schneider, R. J. Mayer, J. W. Pjrz, J. Widom and L. Que, Jr., J. Am. Chem. SOC., 1984,106,1676. INSIGHT11 version 2.1.O, DISCOVER version 2.8, INSIGHTIIi DISCOVER, Biosym Technologies, San Diego, CA, USA, 1992. C. W. Lee, X. Chen and L. Kevan, Catal. Lett., 1992,15,75. N. W. Alcock, Adv. Inorg. Chem. Radiochem., 1972,15,2. P. E. Riley and K. Seff, J. Phys. Chem., 1975,79. 1594. 36 M. Endregard, Ph.D. Thesis, The University of Trondheim, AVH, to be submitted. Paper 4/01 105H;Received 23rd February,1994