摘要:

J. CHEM. SOC.FARADAY TRANS.,1994, 90(18), 2775-2781 2775 EXAFS Studies of Molecular Geometries of Some Coil and Col" Porphyrins Monica Endregard" and David G. Nicholson Department of Chemistry, University of Trondheim,A VH,N-7055 Trondheim, Norway Raymond J. Abraham and Ian Marsden Robert Robinson Laboratories, University of Liverpool, P.O. Box 147, Liverpool, UK L69 3BX Brian Beagley Department of Chemistry, UMIST, P.O. Box 88,Manchester, UK M60 IQD ~~ Co K-edge EXAFS data of some (porphinato)cobalt(tii) complexes with amine ligands in chloroform solution are reported. Data for the following compounds were collected : chloro(tetraphenylporphinato)cobalt(iit) (TPPCoCI) complexed with pyridine (A), isoquinoline (B) and cyclohexylamine (C) and 4-methylpiperidine complexed with the sterically hindered chloro(tetra-o-dichlorophenylporphinato)cobalt(iii) (TCIPCoCI) (D) and chloro(tetra-o- difluorophenylporphinato)cobalt(iit) (TFPCoCI) (E).In addition to the four N atoms from the macrocyclic porphyrin ligand, the cobalt atom is coordinated to either one amine and a chlorine atom (A and B) or two axial amine ligands with the chloride as a counterion (C-E). An interesting feature is the longer Co-CI distances in A and B relative to those in theosolid state. The Co-N distances are represented by a single composite distance ranging from 1.91(1) to 1.95(1) A. The Co-N bond lengths for the ortho-substituted porphyrins were found to be similar to those in the less hindered TPP complexes. The EXAFS for the cobalt(ii) compounds TClPCo (F) and TFPCo (G) gave distances similar to those of the crystal structure of TPPCo.The equatorial Co-N distances depend more on the conformation of the porphyrin core than on the cobalt oxidation state, as demonstrated by the similarity between the Co"l-N and Co"-N distances. Metalloporphyrins are currently being studied intensively in and used to obtain geometric information on porphyrin-the search for molecular metals and selective oxidative cata- ligand'.' '*' and porphyrin-porphyrin complexes.7~9~'0~12 lysts. Their high chemical and thermal stability and their We have previously shown1' that the cobalt-nitrogen ability to form complexes with many different metals are distances in chloro(pyridine)(phthalocyaninato)cobalt(rn) unique assets.Also their propensity to form stacked com- obtained by EXAFS supported previous assumptions made plexes on the one hand, and the relative ease of synthesising in obtaining a ring-current model for the phthalocyanine ring sterically protected metalloporphyrins on the other, makes and lead to an improved model. Here we investigate the them of considerable potential in these important areas of steric effects of o-phenyl substituents on the porphyrin ring. A research.' The use of porphyrins as molecular metals depends detailed NMR and molecular mechanics investigation of critically on the geometry and interactions between the these compounds has been given," and we will show that the macrocycles;2 similarly their use as oxidative catalysts is EXAFS results complement the NMR studies.largely determined by the steric constraints and geometry of Steric hindrance appears to have a considerable impact on the axial ligand.3 complex formation between porphyrins and amines, as We report here an extended X-ray absorption fine struc- demonstrated by the fact that the sterically hindered N-ture (EXAFS) spectroscopic study of some methylpiperidine ligand quantitatively reduces TPPCo"'C1 (porphinato)cobalt(mI) complexes with amine ligands in chlo- to the Co" species.16 2-Methylpiperidine is also sterically hin- roform solution and two solid (porphinato)cobalt(rr) com- dered (albeit to a lesser extent) and reduction to the Co" pounds. EXAFS was chosen because of its unique ability to complex is only partial. Similar steric effects are absent in the provide precise structural information about a given metal case of 3-methyl- and 4-methyl-piperidine, and reduction does atom in solution as well as in the solid state.Strong EXAFS not take place. This illustrates that the environment around signals are observed from metalloporphyrins because of the the metal is crucial to the chemistry. local symmetry provided by the macrocyclic ring system. In addition to first-shell nitrogen atoms, distant carbon atoms in the ring also contribute to the EXAFS.4-6 The present investigation includes chloroform solutions of TPPCoCl complexed with pyridine (A), isoquinoline (B) and cyclohexylamine (C). The effect of ortho substituents on the phenyl rings was examined in chloroform solution by analys- ing the 4-methylpiperidine complexes of TClPCoCl and X = H : TPPCoCl TFPCoCl, D and E, respectively.Fig. 1 shows the porphyrins X = F : TFPCOCI referred to in this study. The results for the cobalt(I1) com- X = CI : TClPCoCl pounds TClPCo" (F) and TFPCO" (G) are included for pur- poses of comparison. These experiments are closely connected to NMR measure-ments on Co"' porphyrins and phthalocyanines complexed with axial amine ligand~.~-'~ The NMR spectra are domi- Fig.1 Structures of chloro(tetraphenylporphinato)cobalt(III)nated by the large ring-current shifts produced by the circu- (TPPCoCl), chloro(tetra-o-difluorophenylporphinato)cobalt(m)lating .n electrons of the porphyrin macrocycle.A refined (TFPCoCl)and chloro(tetra-o-dichlorophenylporphinato)cobalt(I~~) the chlorine atom is not shownmodel of the porphyrin ring current has been presented14 (TClPCoCl); The steric effect of o-chloro substituents has recently been revealed by NMR and X-ray crystallography of [(l-methyl- imidazole),TClPCo] +BF4-.' The chloro substituents cause large changes in the 59C0 chemical shifts and linewidths and increased ruming of the porphyrin core with respect to the unhindered analogue [(imidazole),TPPCo] +OAc-.' The chloro groups effectively block the space above and below the porphyrin plane which leads to a fixed ligand orientation. The geometries of hindered and unhindered cobalt(rrr) porphyrins complexed with pyridine, 1-methylimidazole, 4- methylpiperidine and isoquinoline have recently been investi- gated by NMR and molecular mechanics calculations.'3 The preferred orientation of the axial amine ligands with respect to the porphyrin ring was deduced.For the sterically hin- dered porphyrins TClPCoCl and chloro(tetramesity1-porphinato)cobalt(m) (TMPCoCl) complexed with the planar amine ligands, the eclipsed position (ligand over N) is pre- ferred as shown by, for example, the above crystal structure of [( 1-methylimidaz~le)~TClPCo]-tBF, -.l7 The staggered orientation is preferred in the case of the less hindered porphyrins TFPCoCl and TPPCoCl. The non-planar 4-methylpiperidine complexes are more similar, involving relax- ation of the Co-N bond from the vertical.It was suggested that this bond was lengthened so as to accommodate the increasing steric strain. Experimental The preparations of the Cot" porphyrin compounds have been described in more detail el~ewhere.'~*'~ TPP and the ortho-substituted TXP (X = Cl, F) were prepared by the methods of Alder et a1." and Lindsey and Wagner," respec-tively. Cobalt was incorporated into the porphyrins asco",21.22 and the Co"' complexes were prepared from the corresponding Co" complexes by warming the latter with HCl dissolved in a solution of chloroform and methanol fol- lowed by evaporation. Conversions from free porphyrin to the Co" complex and from Co" to Co"' porphyrin were mon- itored by UV-VIS spectroscopy. The Co"' solutions were prepared immediately prior to data collection by dissolving appropriate amounts of the Co"' porphyrin in chloroform (typically 40 mg ml-') and adding the amine.Two types of complexes are formed depending on the molar ratio between porphyrin and amine. A molar ratio of 1 : 1 gives an octahedral complex with the sixth coordi- nation site occupied by a chloride counter ion acting as a ligand (A and B). When the ratio is increased to greater than 1 :2, the chloride is displaced by an amine and is present as a counter ion (C-E). The solid porphyrin samples were crystallised by evapo- rating a chloroform solution of 4-hydroxypiperidine and TXPCo"'C1 (X = Cl, F). Compounds F and G were formed by the subsequent reduction of Co"' to Con. The reference compound tris(ethylenediamine)cobalt(rIr)chloride hydrate (Co(en),Cl, 3H20) was obtained from Aldrich Chemical Company.X-Ray absorption data were collected at the Daresbury Synchrotron Radiation Source, station 7.1, operating at an energy of 2.0 GeV and an average current of 200 mA. This station is equipped with an order-sorting Si( 1 11) mono- chromator that was offset to 50% of the rocking curve for harmonic rejection. Focussing optics were not used. Room-temperature cobalt K-edge data (E = 7710 eV, A = 1.60811 A) were registered over the energy range 7422- 8307 eV in the fluorescence mode because the samples were dilute in cobalt. An NaI scintillation counter was used. Two scans were collected for each sample. A spectrum of the refer- ence compound was also collected in the transmission mode, J. CHEM.SOC. FARADAY TRANS., 1994, VOL. 90 the sample being diluted with boron nitride in order to attain log(Io/I) in the range 1-2 just above the absorption edge. The fluorescence sample was also diluted with boron nitride so as to give a suitable fluoresence spectrum. The solid samples were finely ground and held between strips of sellotape. The porphyrin solutions were prepared immediately prior to data collection and injected into a specially designed sample holder with Kapton windows. Data were corrected for dark currents, converted to k-space, summed and background-subtracted using EXCALIB and EXBACK23 to yield the EXAFS function ZbS(k).Model fitting was carried out with EXCURV90,23 using curved- wave theory and ab initio phase shifts. The central atom (Co) and backscattering atom (C and N) phase shifts were calcu- lated within EXCURV90 using a 2 + 1 1s core hole correc- tion; they were not iterated in the least-squares fitting process.Fourier transforms were corrected for the phase shifts of the relevant atoms. The low-energy cut-off fsr all samples was 30 eV. The full energy range was applied for solution B and the solid samples (F and G), but the range for the remaining solution data was reduced to 9.4 (E) and 10.2 A-' (A, C and D) because of lower signal to noise at higher k. The data were k3 weighted in order to compensate for the diminishing amplitude at high k owing to the decay of the photoelectron wave.The solution data (A-E) were Fourier filtered using a Gaussian window function to include only the two, three or four first shells using a convenient cut-off which was not complicated by overlapping Fourier-transform shells. This has been shown not to introduce truncation errors.24 Data for the solid samples (F and G) were Fourier filtered in the range 1.0-25.0 A. This filter removes only the low- frequency contributions below 1 A, but does not remove the high-frequency oscillations, i.e. no noise removal. Only those shells significant at the 99% level2' were included in the final models. Details of the final models are listed in Tables 1 and 2 which give interatomic distances (rJ, Debye-Waller-like factors (20~) and the multiplicities (N),i.e.the number of atoms in a given shell n. The fits are shown in Fig. 2-4. The reference compound, tris(ethylenediamine)cobalt(rII) ~hloride,~~,~~was used to check the validity of the ab initio phase shifts and to establish the general parameters, AFAC (proportion of absorption causing EXAFS) and VPI (accounts for inelastic scattering of the photoelectron).28 These parameters were then transferred into the analyses for the compounds being investigated, thereby reducing any residual systematic error in the multiplicities. The estimated standard deviations (esd) for distances derived by EXAFS are 0.01 A at small r and 0.10 A at large r (ca. 4 A), with +20% accuracy for 2a2. Results and Discussion Solution Data The results are listed in Table 1, and the fits are shown in Fig.2 and 3. All coordinated nitrogen atoms are averaged to a single weighted distance, which varies from 1.91 (1) to 1.95 (1) A. Thus, the Co-N bonds to porphyrin and axial amine ligand could not be resolved by EXAFS. Samples A and B have one axial amine ligand with the sixth octahedral posi- tion being occupied by a chlorine atom. This corresponds to the formulae (pyridine)TPPCoCl and (isoquino1ine)-TPPCoCl. The other solutions (C-E) gave distances and coordination numbers consistent with complex salts with for- mulae [(cyc1ohexy1amine),TPPCo] 'C1 -(C) and C(4-methyl- piperidine),TXPCo]+Cl-, where X = C1 and F for D and E, respec ti vel y. As a result of the relatively rigid macrocyclic system, con- tributions from more distant atoms in the porphyrin ring are J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Table 1 Results of EXAFS curve fitting for A-E, all samples in chloroform solution (ca.40 g 1-') multiplicity r/A 2a'lA' E,/eV R(%) A Co-N 5 1.911 (2) 0.0072 (2) 19.1 (4) 13 co-Cl 1 2.569 (3) 0.005 (1) co. . -c 10 2.880 (7) 0.035 (2) B Co-N 5 1.913 (1) 0.0046 (1) 19.5 (3) 11 co-Cl 1 2.552 (5) 0.013 (1) co. . .c 8 2.948 (2) 0.0088 (5) co. . -c 6 3.266 (3) 0.007 (1) C Co-N 6 1.924 (2) 0.0071 (3) 17.7 (3) 14 CO. . -c 8 2.981 (4) 0.012 (1) D Co-N 6 1.930 (2) 0.0088 (2) 17.7 (3) 13 co. * *c 12 2.971 (3) 0.018 (1) co. ..c 4 3.323 (8) 0.012 (2) E Co-N 6 1.947 (2) 0.0099 (3) 16.0 (4) 16 CO. .*c 12 3.015 (4) 0.016 (1) CO...c 4 3.357 (7) 0.004 (2) Each distance (r, Co-X for bonding and Co...X for non-bonding) is associated with a multiplicity (or coordination number) and thermal vibration (Debye-Waller-like factor, 20'). E, is the refined correction to the threshold energy of the absorption edge. The multi- plicities are assigned on structural chemical grounds and the spectra Fourier filtered using a Gaussian window function and a convenient cut-off which includes two, three or four shells. The standard devi- ation in the least significant digit as calculated by EXCURV90 is given in parentheses. This is a measure of the precision (as opposed to accuracy) to which the parameters are determined within the con- straints of the model.Systematic errors will degrade the accuracy, with major contributions stemming from inexact treatment in the theory. Thus, the estimates of precision overestimate the accuracy, particularly in cases of high correlation between parameters. The estimated standard deviations for distances are 0.01 A at small r and 0.04 A at larger (ca. 3 A) r, with +20% accuracy for 20'. The residual index, R, was calculated as R= seen in the Fourier transforms. The multiplicities and approximate distances for these atoms were deduced from reported crystal structures. Table 3 compares the EXAFS and crystallographic distances. Distances extracted from EXAFS are often shorter than those obtained from crystallography. Table 2 Results of EXAFS curve fitting for the solid Co" com-pounds F and G; selected distances out to 4.5 A deduced from the crystal structure of TPPCo" are included for comparison with the EXAFS results multiplicity r/A 2a2/A2 E,/eV R(%) EXAFS" F Co-N 4 1.945 (2) 0.0060 (3) 17.1 (5) 22 co..*c 8 2.980 (4) 0.010 (1) co.. -c 4 3.323 (10) 0.011 (3) co-* .c 4 3.747 (15) 0.007 (3) co-. .c 8 3.946 (14) 0.014 (3) G Co-N 4 1.920 (3) 0.0071 (4) 17.7 (5) 26 co. . .c 8 2.966 (5) 0.011 (1) CO. * *c 4 3.295 (9) 0.010 (2) co*. -c 4 3.741 (15) 0.011 (3) co. . .c 8 4.363 (12) 0.012 (3) X-ray TPPCo" Co-N 1.949 (3) CO. * *c 2.998 (5) co-..c 3.403 (5) CO. .*c 3.779 (5) CO. .c 4.215 (8) " This study; the standard deviation in the least significant digit, as calculated by EXCURV90, is given in parentheses.The data were Fourier filtered over the range 1.0-25.0 A using a Gaussian window function (i.e.no noise removal). See also notes to Table 1. Ref. 36. This is largely due to the inexact treatment in the theory29 coupled with the lack of resolution of similar distances [i.e. the Co-(N), and Co-amine distances]. It is therefore most appropriate when considering the Co-N lengths in these molecules to compare the relative distances as determined by EXAFS rather than the absolute values. The radius of the central porphyrin hole can be altered by puckering or ruffling the rings and ranges between 1.929 and 2.098 (Porphinato)cobalt(rrI) crystal structures have average Co-N(porphyrin) bond lengths that range from 1.954 (4)3' to 1.985 (9) a.32The distances reported in the present study are short, although equal to the smallest porphyrin hole (1.929 A). The sterically hindered (porphinato)cobalt(IIr) complex [( 1-methylimidazole),-TClPCo]+BF,-has a similar axial Co-N distance; 1.942 (6) A.17 The phthalocyanines generally have shorter Co-N (macrocycle) distances than porphyrins, but similar distances to the axial ligand~.~~.,~ Thus, the EXAFS study of chloro(phthalocyaninato)cobalt(rn)(s) and chloro(pyridine)-(phthalocyaninato)cobalt(III)(s)15 shows that the correspond- ing distances are 1.91 (1) and 1.89 (1) A, respectively.The axial Co--N distance for TPPCo"' complexes with amine ligands varies from 1.93 (2)" to 2.060 (3) A.34 Hence, the dis- tances in the present study are equal to the shortest reported value within experimental error.Structures of Chluro(pyridine)(tetruphenylporphinato)cobu~t(ili) (A)and Chloro(isoquinoline)(tetruphenylporphinato)cobalt(IrI) (B) The Co-N(porphyrin) and Co-"(amine) distances were averaged to a single distance at 1.91 (1) I$ in both com-pounds. The chlorine atom is directly bonded to cobalt C2.57 (1) and 2.55 (1) A, for A and B, respectively]. The corre- sponding crystallographic distances for A35 are 1.976 (6) (TPP), 1.978 (8) (pyridine) and 2.251 (3) 8, (CI). Ruffled (porphinato)metal cores generally have shorter Co-N dis-tances than planar ones.36 This expresses the desire of the system to minimise the metal-N distances within the con- straints of the macrocycle.Thus, the distances reported here are consistent with a puckered ring system. An NMR study using the ring-current program DIPCALCi4 on the bispyridine complex of TPPCoCl gave much better agreement for the freely rotating than for a fixed- ligand orientation,' and it was suggested that the ligand has considerable rotational mobility. The even shorter distance obtained by EXAFS for the monopyridine complex C1.91 (1) A] supports this conclusion. The Co-Cl distances are significantly longer in the com- plexes dissolved in chloroform than in the crystal structure of A. In this connection, note that an unusually long Co-Cl distance has been reported for a paramagnetic five-coordinate Co"' compound C2.507 (2) Of particular interest is the fact that the five-coordinate porphyrin TPPCoCl is diamagnetic in the solid state, but changes partly to paramagnetic species in inert solvents such as chloroform when the temperature is raised.38 Certainly, the analogous non-planar OETPPCO"'C~ has been found to be paramagne- tic.Structure of Bis(cyclohexyfamineMtetruphenylporphinuto) cobalt(II1)Chloride (C) The corresponding crystal structure has not been reported. The short Co-N(porphyrin) distance c1.92 (1) 14) indicates a ruffled (N), grouping. The axial Co-N(amine) distance C1.92 (1) A] is also short, but similar within error limits to distances reported for other amine ligands."*" J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 1 1 1 J 4 6 8 10 1 2345 k/A-RIA 1.6I Q,U w.-c 1.2 0,2 0.8 'c C g 0.4 r m.---. . . . . 0 4 6 8 10 12 12345 k1A-I RIA 1.6 0.8 0* C2 0.4 0 4 6 0 10 12345 k/A-' RIA Fig. 2 Observed (-) and calculated (---) Fourier-filtered k3-weighted EXAFS and their Fourier transforms for (a) A, (b) B and (c) C; parameters in Table 1 Structuresof Bis(4-methyZpiperidine)(tetra-o-dichZorophenyl-that these sterically hindered complexes (D and E) exist as porphinato)cobaZt(III) Chloride (0)and Bis(4-methyZpiperidine) six-coordinate cations in chloroform solution with Co-N (tetra-o-diJiuorophenyZporphinato)cobuZt(IIr)Chloride (E) distances which are similar to those of the unhindered com- The fits are shown in Fig. 3, and the parameters listed in plexes A, B and C (Table 1).These results agree with the Table 1. Table 3 compares the obtained distances with rele- crystal structures of [( 1-methylimidazole),TClPCo]+BF4-'' vant crystal structures. The multiplicities and distances show and the unhindered analogue [(imidazole),-Table 3 Intramolecular distances (A) (Co-X for bonding and Coo .-X for non-bonding) from the solution study and from the crystal struc- tures of A"' and the cations [(piperidine),TPPCo'"] + ' and [( l-methylimidazole)2TC1PCo111]the multiplicities (N)are included for the + ;d crystal-structure data + +PyrTPPCoClb (piperidine),TPPCo (1-rnethylimidaz0le)~TClPCoA" B" C" D" E" r/A r/A N r/A r/A r/A r/A N rlA N rlA Co-N 1.91 (1) 1.91 (1) 4 1.976 (6) 1.92 (1) 1.93 (1) 1.95 (1) 4 1.978 (5) 4 1.977 (5) Co-N 1.91 (1) 1.91 (1) 1 1.978 (8) 1.92 (1) 1.93 (1) 1.95 (1) 2 2.060 (3) 2 1.942 (6) CO-Cl 2.57 (1) 2.55 (1) 1 2.251 (3) -CO-* .C 2.88 (2) 2.95 (2) 8 3.018' 2.98 (2) 2.97 (2) 3.02 (2) 8 3.028 (14) 8 3.018 (22) CO.*-C2.88 (2) 3.27 (4) 2 2.909 (9) -2.97 (2) 3.02 (2) 4 3.022 (5) 4 2.9468 CO*--C 2.88 (2) 3.27 (4) 4 3.402' -3.32 (4) 3.36 (4) 4 3.433 (6) 4 3.408 (14) ~ This study, with esds in parentheses.The corresponding multiplicities and Debye-Waller-like factors are given in Table 1. Ref. 35. Ref. 34. Ref. 17. 'Average of eight distances within the range 2.994-3.052 A. Average of four distances within the range 3.388-3.420 A. 9 Average of four distances within the range 2.907-2.978 A. J. CHEM.SOC. FARADAY TRANS., 1994, VOL. 90 L I t c 4 6 a 10 12345 k/A-' RIA Observed (-) and calculated (---) Fourier-filtered k3-weighted EXAFS and their Fourier transforms for (a)D and (b)E; parameters Fig. 3 in Table 1 TPPCo]+OAc-for which the distances are only slightly different. However, the steric requirements greatly affect the deviations of the core atoms from the mean plane of the porphyrin, the largest displacements in the hindered and unhindered complexes being 0.23 and 0.07 A, respectively. The o-chloro groups combined with crystal packing forces constrain the phenyl groups to be nearly perpendicular to the macrocyclic ring and also fix the axial ligand orientation. This effect has also been shown in a recent molecular mechanics and NMR study on various hindered and unhindered Co"' porphyrins complexed with amine ligands.l3 Planar ligands complex edge-on to the porphyrin with a fixed r 1 1 I 6 h s ,"o .y -6 4 6 8 10 12 1 2 3 4 5 k/A-' RIA Fig.4 Observed (-) and calculated (---) Fourier-filtered (1.0-25.0 A) k3-weighted EXAFS and their Fourier transforms for (a)F and (b)G; parameters in Table 2 orientation, staggered (ligand over N atoms) for the less steri- cally hindered complexes (o-fluoro substituents) and eclipsed ligand over rneso-carbon atoms) for the complexes with o-chloro and -methyl substituted porphyrins. Increased steric strain and a possible lengthening in the axial Co-N bond are suggested in the complexes with the non-planar 4-methylpiperidine complexes.No clear evidence of such a lengthening in the axial bond for D and E with respect to the unhindered complexes A-C could be found in the present study. A complicating factor in this context is that the equa- torial and axial Co-N bonds had to be averaged to one distance to avoid severe correlation between EXAFS param- eters. However, it shows that the axial Co-N bonds are shorter than ca. 2.1 A. Solid Samples The results are listed in Table 2 and the fits shown in Fig. 4. The EXAFS data show that there are only four N atoms in the first shell. The absence of axial amine ligands and/or a bonded chloride is consistent with reduction to the cobalt(I1) species TClPCo" (F) and TFPCo" (G).We note that Kastner and Scheidt,' found that (acety1)TPPCo"' and some other derivatives are unstable in chloroform solution, decomposing within a few hours.Certainly, our results are compatible with the observation that (porphinato)cobalt(ur) complexes with sterically hindered amines are reduced in chloroform solu- tion.16 On the other hand, freshly made chloroform solutions of the ortho-substituted porphyrins form six-coordinate Co"' cations with 4-methylpiperidine (D and E). On the basis of these observations, the requirement for Co"' complexes to crystallise as such would seem to be that the axial ligand must orientate in a sterically favourable manner with respect to the phenyl rings to minimise steric repulsions. Planar imid- azole ligands crystallise as octahedral complex salts with TClPFe"'C10, and TClPCo"'BF, , respectively.17e41 In both compounds the axial amine ligands have fixed orientations.The fact that piperidine is not planar combined with unfa- vourable crystal packing forces may explain the failure of complexes with ortho-substituted TPPCo"' to crystallise. An important factor in EXAFS is the phenomenon of multiple scattering which enhances contributions from distant atoms due to the electron wave being focussed by the first-shell atoms with a concomitant phase shift and, in turn, a displacement in distance. This leads to apparently shorter or longer distances as the program moves that particular wave in an attempt to correct the phase, and higher multi- plicities and/or low Debye-Waller factors.Multiple scat-tering effects are known to be important for ring ligands (such as the present compounds) and near collinear (150-180") atomic arrangment~.~' The Debye-Waller-like factors and the errors in the distances out from ca. 3.5 8, are to be regarded in this context (the multiplicities being fixed). That multiple scattering pathways do not significantly affect the results for the first three-shells was tested by repeating the calculations on data that were Fourier filtered to include only these shells. This cut-off can be made without causing large truncation effects, and is one that is not complicated by over- lapping Fourier-transform shells. It also eliminates com-plications in data analysis due to multiple scattering effects which could arise in more distant shells.Structure of( Tetra-o-dichZorophenylporphinato)co~aIt(II)(F) and (Tetra-o-dij?uorophenyZporphinato)cobaZt(II)(G) Table 3 shows that the distances and multiplicities obtained for samples F and G are in good agreement with the crystal structure of TPPCO".~~ C0'I-N bond distances are generally longer than those in the corresponding Co'" compounds.42 J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 However, the porphyrin core conformation has a greater impact on the equatorial Co-N distances than the oxidation state, as demonstrated by the distance in TPPCo" which is 1.949 (3) and the short distances derived by EXAFS, 1.95 (1) and 1.92 (1) 8, for F and G, respectively.These dis- tances are similar to the corresponding co-N composite distances observed in the solutions. The authors thank the Norwegian Science and Research Council (NAVF) and VISTA for financial support, the Royal Norwegian Council for Industrial and Scientific Research (NTNF) for a stipend to M.E. and the Science and Engineer- ing Research Council for an earmarked studentship to I.M. We especially thank Mr. Jostein Rise for designing and con- structing our EXAFS cells. References 1 B. M. Hoffman and J. A. Ibers, Acc. Chem. Res., 1983,16, 15. 2 J. P. Collman, J. T. McDevitt, C. R. Leidner, G. T.Yee, J. B. Torrance and W. A. Little, J. Am. Chem. SOC., 1987,109,4606. 3 P. Battioni, J-P. Renaud, J. F. Bartoli and D. Mansuy, J.Chem. SOC.,Chem. Commun., 1986,341. 4 S. P. Cramer, in X-Ray Absorption, ed. D. C. Koningsberger and R. Prins, Wiley, 1988, p. 271. 5 R. W. Joyner, J. A. R. van Veen and W. M. H. Sachtler, J. Chem. SOC., Faruday Trans. 1, 1982,78, 1021. 6 B. van Wingerden, J. A. R. van Veen and C. T. J. Mensch, J. Chem. SOC.,Faruday Trans. 1, 1988,84,65. 7 R. J. Abraham, S. C. M. Fell, H.Pearson and K. M. Smith, Tetrahedron, 1979,35,1759. 8 R. J. Abraham, G. R. Bedford and B. Wright, Org. Magn. Reson., 1982, 18,45. 9 R. J. Abraham, P. Leighton and J. K. M. Sanders, J. Am. Chem. SOC.,1985, 107, 3472. 10 K. M. Smith, F. W. Bobe, D. A. Goff and R. J. Abraham, J. Am. Chem. SOC., 1986,108,1111. 11 R. J. Abraham and C. J. Medforth, Magn. Reson. Chem., 1988, 26, 803.12 R. J. Abraham, A. E. Kowan, K. E. Mansfield and K. M. Smith, J. Chem. SOC., Perkin Trans., 1991,2, 515. 13 R. J. Abraham and I. Marsden, Tetrahedron, 1992,48,7489. 14 R. J. Abraham, S. C. M. Fell and K. M. Smith, Org. Mugn. Reson., 1977,9,367. 15 M. Endregard, D. G. Nicholson, R. J. Abraham, I. Marsden and B. Beagley, Acta Chem. Scund., 1994, in the press. 16 R. J. Abraham and C. J. Medforth, Mugn. Reson. Chem., 1988, 26, 334. 17 H. Bang, J. 0. Edwards, J. Kim, R. G. Lawler, K. Reynolds, W. J. Ryan and D. A. Sweigart, J. Am. Chem. SOC., 1992, 114, 2843. 18 J. W. Lauher and J. A. Ibers, J. Am. Chem. SOC.,1974,%, 4447. 19 A. D. Adler, F. R. Longo, J. D. Finarelli, J. Goldmacher, J. Assour and L. Korsakoff, J.Org. Chem., 1967,32,476. 20 J. S. Lindsey and R. W. Wagner, J. Org. Chem., 1989,54,828. 21 J. W. Buchler, in Porphyrins and Metulloporphyrins, ed. K. M. Smith, Elsevier, Amsterdam, 1975, p. 179. 22 A. D. Adler, F. R. Longo, F. R. Kampas and J. Kim, J. Znorg. Nucl. Chem., 1970,32,2443. 23 N. Binsted, J. W. Campbell, S. J. Gurman and P. C. Stephenson, EXCALIB, EXBACK and EXCURV90 programs, SERC Dares- bury Laboratory, 1990. 24 N. Binsted, S. L. Cook, J. Evans, G. N. Greaves and R. J. Price, J. Am. Chem. SOC.,1987,109,3669. 25 R. W. Joyner, K. J. Martin and P. Meehan, J. Phys. C: Solid State Phys., 1987,20,4005. 26 K. Nakatsu, Y. Saito and H. Kuroya, Bull. Chem. SOC. Jpn., 1956,29,428. 27 A. Whuler, C. Brouty, P. Spinat and P. Herpin, Acta Crystal-logr., Sect.B, 1975,31, 2069. 28 S. J. Gurman, N. Binsted and I. Ross, J. Phys. C: Solid State Phys., 1984, 17, 143. 29 N. Binsted, R. W. Strange and S. S. Hasnain, Biochemistry, 1992, 31,12117. 30 F. A. Cotton and G. Wilkinson, Adoanced Inorganic Chemistry, Chichester, 5th edn., 1988, p.355. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 278 1 31 J. A. Kaduk and W. R. Scheidt, Znorg. Chem., 1974,13,1875. 39 C. Medforth, personal communication. 32 T. Sakurai, K. Yamamoto, H. Naito and N. Nakamoto, Bull 40 M. E. Kastner and W. R. Scheidt, J. Organomet. Chem., 1978, Chem. Soc. Jpn., 1976,49,3042. 157,109. 33 B. Moubaraki, M. Ley, D. Benlian and J-P. Sorbier, Acta Crys- 41 K. Hatano, M. K. Safo, F. A. Walker and W. R. Scheidt, Znorg. tallogr., Sect. C, 1990,46,379. Chem., 1991,30,1643. 34 W. R. Scheidt, J. A. Cunningham and J. L. Hoard, J. Am. Chem. 42 N. E. Kime and J. A. Ibers, Acta Crystallogr., Sect. B, 1969, 25, Soc., 1973,95, 8289. 168. 35 T. Sakurai, K. Yamamoto, N. Seino and M. Katsuta, Acta Crys- 43 C. Hedtmann-Rein, M. Hanck, K. Peters, E-M. Peters and H. G. tallogr., Sect. B, 1975,31,2514. von Schering, Znorg. Chem., 1987,26,2647. 36 P.Madura and R. Scheidt, Znorg. Chem., 1976,15,3182. 37 N. A. Bailey, E. D. McKenzie and J. M. Worthington, J. Chem. SOC.,Dalton Trans., 1977,763. Paper 4/01742K; Received 23rd March, 1994 38 K. Yamamoto, J. Uzawa and T. Chijimatsu, Chem. Lett., 1979, 89.

ISSN:0956-5000    

DOI:10.1039/FT9949002775

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

J. CHEM. SOC. FARADAY TRANS., 1994, 90(18), 2783-2790 EXAFS Data Analysis for Lanthanide Sesquioxides P. Malet, M. J. Capitan, M. A. Centeno, J. A. Odriozola and 1. Carrizosa Departamento de Quimica lnorganica e lnstituto de Ciencia de Materiales de Sevilla, Centro Mixto Universidad de Sevilla-C.S.l.C., P.O. Box 553,41071 Sevilla, Spain The EXAFS spectra of lanthanide sesquioxides (LU,~,, Sm203, La203) at the lanthanide L,, ,-edge are analysed. The complex coordination polyhedron around the lanthanide cation is modelled with the minimum number of coordination shells that still allows the acquisition of coordination numbers and shell distances in agreement with the radial distribution functions from crystallographic data. Theoretical phase shifts and backscattering amplitudes with an amplitude reduction factor, S: = 0.73, are reliable for reproducing the experimental EXAFS data.A model with one Lu-0 and two Lu-Lu shells simulates the Lu coordination polyhedron in C-Lu,03, while for La,O, the model includes two oxygen shells for simulating the nearest neighbours and one longer distance that averages that of the La-La pairs. The coordination around Sm in Sm,O, is the most complex and two-shell models are needed to simulate the Sm-O and Sm-Sm absorber-backscatterer pairs. The models obtained are applied in the EXAFS analysis of dispersed Ln,O,/AI,O, samples, where X-ray diffraction fails to detect the structure adopted by the lanthanide phase. The results show that in an Sm,O,/AI,O, sample calcined at 800 "C, very small Sm,O, particles are formed.In an La,O,/AI,O, sample with low loading the analysis procedure allows the detection of the aluminium atoms that are present with the oxygens around the lanthanum ions, thus suggesting the incipient formation of a bidimensional LaAIO, phase. The local structure around the lanthanide atom is often of nation polyhedra around the lanthanide atoms in the Ln203 interest in non-crystalline and highly dispersed systems '-, phases with a limited number of shells, thus allowing the such as halide glasses (fibre optics and lasers), metallic glasses analysis of the EXAFS spectra of the unknown compound to (high strength, low-density materials) or, from a most general be made with a feasible number of parameters.point of view, disordered system^.^ However, examples of the In previous papers7-' we have reported EXAFS data application of the EXAFS technique to systems in which analyses for Sm,O,/Al,O, samples using phase and ampli- lanthanide atoms are involved are scarce in the literature. On tude functions calculated from McKale et uZ."*'~ for the the other hand, the structure of supported rare-earth-metal various contributions, which were shown to be reliable. oxides is a major concern in surface chemistry and cataly- However, their use has recently been discouraged12 and the sis.*-' Thus, the addition of small amounts of these oxides to use of the FEFF code', to calculate phase and amplitude y-Al,O, inhibits its sintering process at high temperatures. A functions is recommended. surface compound, usually amorphous in structure, has been In this paper we analyse the EXAFS spectra of Ln203 ses- suggested to be responsible for such behavio~r.'.~ EXAFS quioxides (LU,~,, Sm203, LU,~,) using theoretical phase spectroscopy is well suited to this structural problem.shifts and backscattering amplitudes calculated from FEFF Changes in the catalytic performance of supported rare-and propose a reasonable model that simulates the compli- earth-metal oxides in reactions such as the oxidative coupling cated structure of these compounds with the minimum of methane7** and hydrogenation over Ln,O,-promoted number of coordination shells, still yielding coordination rhodium catalysts' could be ascribed to changes in the struc- numbers and shell distances in agreement with the RDF ture of the highly dispersed or amorphous supported phase.obtained from crystallographic data. These models and the However, difficulties in the interpretation of the EXAFS parameters of the different coordination shells obtained to fit spectra of rare-earth-metal oxides have limited, to our know- the spectra could be used as inputs in the analysis of the ledge, the application of the technique to these systems. EXAFS spectra of materials with unknown local structure The analysis of the EXAFS spectra of crystalline com- around the lanthanide. pounds of lanthanide elements is not straightforward, even The physical meaning of the parameters obtained in the when the local structure around the lanthanide atom is analysis of the experimental spectra with a limited number of known.EXAFS spectra are measured at the lanthanide LI,,- shells and the errors introduced by approximating several edge, and the data range available is limited by the super- distances to only one shell, have been studied by analysing position of the L,,-edge, especially for elements with the theoretical EXAFS spectra including all the distances present lowest atomic numbers. The limited data range results in a in the crystalline structures. Alternative models, their sta- low resolution when using Fourier-transform (FT) techniques tistical significance and physical meaning, have also been to obtain the radial distribution function (RDF) around the considered. lanthanide.Moreover, RDFs deduced from crystallographic data for the oxides show that coordination polyhedra around Experimentalthe lanthanide atoms are complex, having short Ln-0 and Ln-Ln distances that extend over a rather wide range. Two X-Ray absorption spectra of La,O,, Sm,O, and Lu,O, main problems arise from the complex structures of the bulk from Sigma Chemical Co. calcined at 900 "C were recorded at compounds that are also expected in the dispersed systems. the L,,, absorption edge of the lanthanide (La, 5491 eV, Sm, Crystalline oxides cannot be used as EXAFS reference com- 6721 eV and Lu, 9250 eV). The Lu,O, spectrum and those of pounds to obtain experimental phase shifts and backscatter- the Ln,O,/Al,O, samples were measured at station EXAFS ing amplitudes for Ln-0 and Ln-Ln absorber-backscatterer I11 at LURE DCI, Orsay (France), while Sm203 and La203 pairs, and theoretical references have to be used.The second spectra were recorded at station 8.1 at the SRS, Daresbury problem is that it is necessary to model the complex coordi- Laboratory (UK). Monochromatization was obtained using double silicon crystals working at the (311) reflection at LURE and the (111) reflection at the Daresbury laboratory. The measurements were carried out in transmission mode using optimized ion chambers as detectors. Samples were placed on Kapton Sellotape with an absorbance of ca. 2.5 (Ap, < 1) just above the absorption edge, and measured at room temperature. Analysis and handling of the EXAFS spectra were carried out by using the program NEWEXAFS from the Eindhoven University of Technology.This program uses standard procedures to extract the EXAFS spectra from the measured absorption data. l4 Normalization was achieved by dividing by the height of the absorption edge. EXAFS data were analysed by multiple-shell fitting in k and R spaces using phase shifts and backscattering amplitudes calculated from FEFF.' Standard deviations of fitted parameters were calculated for a mean noise level of Estimated system- atic errors in R (* 1-3%) and N (f10-15%) are, in general, higher than the calculated standard deviations. In some cases, theoretical spectra including all the absorber-backscatterer distances in the crystallographic radial distribution functions were generated.Debye-Waller factors and AEo values were assumed to be zero in these calculations. Results and Discussion Radial distribution functions (RDF) around the lanthanide ions in the sesquioxides (Table 1) were calculated from crys- tallographic data."-" The simplest RDF corresponds to Lu203 which has a C-type structure under atmospheric pres- sure.15 In C-Ln203 structures six oxygen atoms appear around each Ln atom in an almost regular coordination J. CHEM.SOC. FARADAY TRANS., 1994, VOL. 90 sphere, while two different Ln-Ln distances, each of them with a coordination number of six, form the second coordi- nation shell. As shown in Table 1 there are two types of cationic sites with slightly different coordination environment and a relative occupation of 1 :3.In the rather complex Sm,O, structure16 three types of Sm atoms with equal rela- tive abundance are found. In the RDF around the Sm atoms oxygen neighbours appear at distances of up to 2.755 A, with an 0:Sm coordination number of seven, while Sm neigh- bours appear in the range 3.3-4.2 A with an Sm :Sm coordi- nation number of 12. In La203, only one type of La atom is found with seven oxygen neighbours at the nearest distances and 12 La atoms between 3.76 and 3.94 A.17 Fig. 1 shows the unfiltered oscillatory EXAFS functions, Ak), at the lanthanide L,,-edge for Lu203, Sm203 and La,O,, as well as the imaginary part and the absolute value of their associated k3-weighted FTs. Owing to the presence of the L,,-edge, the available k range decreases in the order Lu > Sm >La, leading to a parallel decrease in the intensity and resolution of the associated FTs.In agreement with the radial distribution functions calculated from crystallographic data, peaks in the uncorrected FT at ca. 2 and 3.5 A have been assigned to oxygen and lanthanide coordination shells. Oxygen neighbours at distances longer than 3.5 A scarcely contribute to the whole EXAFS spectrum, which is domi- nated by the heavy lanthanide atoms at long distances.'*" EXAFSData Analysis LU203 As shown in Fig. 1, in the Lu case maxima are well resolved in the FT and therefore the oxygen coordination shell can be Table 1 Radial distribution functions around lanthanide atoms as calculated from crystallographic data (distances in A) LU~O~' 0 6 x 2.242 Lu 6 x 3.433 6 x 3.931 1 x 2.249 2 x 2.286 0 1 x 2.481 2 x 2.554 1 x 2.703 1 x 3.593 1 x 3.342 2 x 3.633 Sm 2 x 3.713, 2 x 3.736 2 x 3.854, 1 x 3.858 2 x 4.178 1 x 2.372 2 x 2.373 0 1 x 2.452 3 x 2.726 1 x 3.678 3 x 3.768 La 3 x 3.866 6 x 3.940 3 x 4.599 0 6 x 4.641 3 x 4.791 a Ref.15. 'Ref. 16. Ref. 17. Smz03* 4 x 2.213 2 x 2.276 2 x 3.922 2 x 3.433, 4 x 3.451 2 x 3.931, 4 x 3.947 1 x 2.288 2 x 2.317 1 x 2.375 2 x 2.485 1 x 2.755 1 x 3.884 1 x 3.342 2 x 3.633 2 x 3.692, 2 x 3.743 2 x 3.782 1 x 3.869 2 x 4.178 1 x 2.260 2 x 2.277 1 x 2.306 2 x 2.564 1 x 3.111 1 x 3.797 -1 x 3.618, 2 x 3.633 2 x 3.736, 2 x 3.743 2 x 3.782 2 x 3.854, 1 x 3.869 La2OJC J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 0.5 0.0 h3 x 4 -0.5 -1 .o 5 10 15 2 4 6 k/A-' RIA Fig. 1 A, Experimental EXAFS spectra for Lu203 (a), Sm203(b) and La,O, (c) and B, associated k3-weighted FT (Ak ranges are 2.0-16, 2.2-12 and 2.2-9.9 A-' for Lu203, Sm,O, and La203, respectively) isolated by Fourier filtering. A direct transformation was per- formed in the 1.9-14.9 k' range using a k2 weighting scheme followed by an inverse Fourier transformation in the 0.20-2.3 A interval. The isolated oscillations are plotted in Fig. 2 and can be fitted by a single Lu-0 coordination shell at 2.22 A, in good agreement with the bond length calculated from crystallographic data (Table 1).The fit was forced to be good in k and R spaces, and working with k' and k3 weigh-ting schemes, since it has been shownlg that strongly coupled parameters such as coordination number (N) and Debye- Waller factor (Aa2) or shell distance and AEo are better decoupled with this procedure. Theoretical phase shift and backscattering amplitude functions were calculated with the FEFF pr~grarn,'~ with an input value of ACT' = 0 (absolute Debye-Waller factor). Theoretically it is expected that an amplitude reduction factor, Sg, is required to correct the cal- culations for the decrease in the overlap of the passive elec- trons between the initial and final states of the absorbing atom, its value lying in the 0.6-0.8 range at k > 7 A for most of the atoms.'8b.20 This factor was calibrated by setting Sg = 1 as a first value to calculate the backscattering amplitude, yielding an 0 :Lu coordination number of 4.39 & 0.02.Since the crystallographic value is six, Sg should be set to 0.73. This value is close to that previously determined by the same procedure' when using McKale's phase shifts and back- scattering amplitudes, and lies within the range reported for most of the elements and close to those reported for other lanthanide elements (Pm, 0.729; Yb, 0.758;18'). Therefore an Si value of 0.73 will be employed in further calculations. Once the Lu-0 shell had been fitted the difference between the raw spectrum and the fitted Lu-O oscillation was calcu- lated (dotted lines in Fig.2). This difference mainly includes Lu-Lu contributions at 3-4 A. Lu-Lu contributions were isolated by performing a k2-weighted FT in the 3.6-15.6 A-' range and an inverse FT between 2.2 and 4.2 A, and are plotted in F@. 3. An FT of these isolated oscillations cor- rected by the Lu-Lu phase shift and backscattering ampli- tude [Fig. 3(b)] shows two clear maxima at ca. 3.5 and 3.9 A, very close to those determined from the diffraction data (Table 1). The oscillations can be fitted by employing two shells at 3.44 and 3.93 8, (Fig. 3), leading to a total Lu :Lu coordination number of 13, in agreement (within the preci- sion of the EXAFS technique) with the crystallographic value (12.0).Best-fit parameters and associated standard deviations are given in Table 2. In summary, a one-shell model for the Lu-0 and a two-shell model for the Lu-Lu contributions have been found to be adequate to simulate the mean coordi- nation polyhedron around Lu atoms in the Lu20, structure. Sm203 Since the k range for the analysis is limited by the presence of the Lredge, which lies 600 eV beyond the Lnredge, some overlapping occurs between the two main contributions in the uncorrected FT. The best way to isolate them has been found by generating theoretical spectra that include all the 1 A 41 nSO x .y -1 k/A-' 5 I0 15 Fig. 2 Lu203: Isolated EXAFS oscillations for the Lu-0 first shell klA-' and associated k2-weighted FT. (-) Experimental data; (---) fit Fig.3 Lu203: Isolated EXAFS oscillations for Lu-Lu shells and with a one-shell model; (. * .) difference spectrum (total experimental associated k2-weighted FT (Ak = 4-15.5 A-') corrected by Lu-Lu phase shift. (-) Experimental data; (---)spectrum minus fitted Lu-O shell). Table 2 coordination N 0 6.0 f 0.02 Lu 7.2 f 0.09 Lu 5.8 & 0.13 fitted data. EXAFS parameter values for crystalline compounds: Lu203 Aa2/A2 RIA AE'IeV 0.0053 f O.OOO1 2.22 f0.001 -7.5 f0.1 0.0046fo.Ooo1 3.44 f0.001 -4.1 f0.1 0.0047 f O.OOO1 3.93 f0.001 -7.4 f 0.2 S: has been set to 0.73 (see text). 2786 I I 0.2 0-4 0.3 0.10.2 E z-s 0.1 + x 0.0* 2 c,0.0 .-ij L -0.1 -0.1 93 -0.2 -0.3 L-L-2-L -0.2 2 4 6, 8 1012 k/A-' Fig.4 Sm,O,: (a) EXAFS oscillations for Sm-0 nearest-shell con- tributions and (b) associated k'-weighted FT (Ak = 3-11.5 A-'). (-) Experimental data; (---) fit with a two-shell model. distances in Table 1, and performing FTs with different k weighting schemes with or without correction by phase shifts and backscattering amplitudes. The Sm-0 contributions due to oxygen neighbours at dis- tances up to 3.04 A, were isolated by performing a k'-weighted Sm-0 phase-corrected FT (Fig. 4, Ak = 2.1-12.3 A-', AR = 0.96-3.10 A). Two alternative models (Table 3) will fit these Sm-0 oscillations. The one-shell model yields an 0 :Sm coordination number (6.2) below that known from X-ray diffraction (XRD) data (7.0) at the shorter Sm-0 dis- tance expected.The alternative two-shell model (Table 3, Fig. 4) gives a total 0 :Sm coordination number of 7.0 and Sm-0 distances in agreement with the values expected from XRD data. The reliability of the two-shell model for Sm-0 dis-tances up to 3.04 A has been checked by generating a theo- retical EXAFS spectrum that includes all the Sm-0 distances shown in Table 1 and groups the shells step by step, avoiding distortions of total coordination numbers and mean dis-Table 3 Sm203: different models for Sm-0 shells (a) Experimental EXAFS oscillations N Ao2/A2 RIA AE'jeV one-shell model O(1) 6.2 0.00822 2.34 -4.2 two-shell model W)O(2) 4.3 2.7 0.00575 0.02307 2.33 2.53 -3.5 -13.3 (b) Theoretical EXAFS oscillations N Ao/A2 RIA AE'IeV 3.66 O.OO0 99 2.291 0.2 2.35 0.001 61 2.527 -0.1 0.67 O.OO0 72 2.728 0.0 0.33 O.OO0 00 3.111 0.0 3.66 O.OO0 99 2.29 1 0.2 2.83 0.005 75 2.532 1.2 0.33 O.oO0 00 3.111 0.0 4.04 0.001 95 2.285 1.o 2.5 1 0.00378 2.506 3.5 4.98 0.009 93 2.276 5.2 Si has been set to 0.73 (see text). J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 tances. Oxygen atoms between 2.249 and 2.375 A can be grouped in only one shell with an 0:Sm coordination number of 3.66 and a mean Sm-0 distance of 2.291 A, in agreement with XRD data. Sm-0 distances of 2.481-2.564 can also be grouped in one shell at 2.527 A with a total 0 :Sm coordination number of 2.35.This four-shell model (Table 3) reflects the deviations in Sm-0 distances around the mean values with increased Debye-Waller factors and slight deviations in AEo values. A three-shell model groups the central 0 shells at one distance leading to a slight under- estimation of the total coordination number, further increases in the Debye-Waller factors and deviations in the AEo value of the central shell. Errors in the mean Sm-0 distances determined with this model including only three shells are <0.04 A. Alternative models with one and two shells have also been considered. The third shell with an 0 : Sm coordi- nation number of only 0.333 is expected to be missed in the fitting of the experimental spectrum.A low 0 :Sm coordi- nation number is expected when fitting the whole spectrum with only one shell, as observed in the fitting of the experi- mental spectrum. Once the Sm-0 contributions had been fitted they were subtracted from the experimental spectrum. Sm-Sm contri-butions [Fig. 5(u)] were isolated from the difference spectrum by performing a k3-weighted Sm-Sm phase-corrected FT (Ak = 4.0-12.3 A-') and an inverse FT in the 2.7-4.5 A range. The analysis of the isolated oscillations yields two Sm shells, with a total Sm: Sm coordination number of 13.1 (expected value 12), and shell distances within the ranges observed by XRD. The analysis procedure is summarized in Table 4. A first shell is introduced at cu. 3.65 A. After refine- ment of this contribution, an Sm-Sm phase-corrected FT of the difference spectrum suggests the presence of, at least, one more Sm-Sm shell at cu.4.2 A. This new shell was introduced and the parameters of both shells refined, leading to the fit shown in Fig. 5(b). The reliability of this two-shell model has been checked by generating a theoretical EXAFS spectrum which includes all the Sm-Sm distances and groups the theoretical shells step by step (Table 4). In a five-shell model deviations in Sm-Sm distances around the mean values would be reflected only in increased Debye-Waller factors. The three-shell model aver- ages the central Sm distances leading to a slight underesti- mation of the total coordination number, a further increase in the Debye-Waller factor and deviations in AEo values.Errors in the mean Sm-Sm distances determined with this 1 ' LA-4-5 46 8 10122 4 6 k/A-' RIA Fig. 5 Sm,O,: (a) EXAFS oscillations for Sm-Sm contributions and (b), associated k3-weighted FT corrected by Sm-Sm phase shift (Ak = 4-11.5 A-'). (-) Experimental data. (---) Fitted data with a two-shell model. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Table 4 Sm20, :different models for Sm-Sm shells (a) Experimental EXAFS oscillations one-shell model Sm(1) 9.2 0.00897 3.64 -1.3 two-shell model Sm(1) 8.5 0.00779 3.64 -1.4 Sm(2) 4.6 0.00697 4.16 -8.4 (b) Theoretical EXAFS oscillations ~~ 0.67 0.000 00 3.342 0.0 2.33 0.00003 3.631 0.0 5.35 0.OOO 86 3.740 0.0 2.33 0.00004 3.859 0.0 1.33 0.000 00 4.178 0.0 0.72 0.00251 3.320 6.0 7.90 0.006 01 3.682 4.4 2.72 o.oO0 75 4.122 8.2 8.68 0.006 28 3.666 6.6 2.43 0.000 16 4.130 6.5 model including only three shells are <0.06 A.Since the first Sm-Sm shell, with the lowest coordination number, seems to be missed in the fit of the experimental spectrum, an alterna- tive two-shell fit of the theoretical spectrum was performed (Table 4), a slight underestimation of the total coordination number results, while errors in Sm-Sm distances are kept within kO.07 A. The two-shell model for Sm-Sm and Sm-0 absorber-backscatterer pairs was considered to be adequate. To avoid errors in the coordination parameters that may be introduced during the isolation procedures, the set of parameters obtained in the analysis of the isolated EXAFS oscillations was used as input data to be refined in the analysis of the complete experimental EXAFS spectrum (Fig.6) filtered by performing an uncorrected k3-weighted FT (Ak = 2.12-1 1.96 A-') and an inverse FT (AR = 1.2-4.3 A). Refined param- eters are included in Table 5, and, taking into account the complexity of the system, their agreement with the RDF obtained from crystallographic data is excellent, providing a good model for the analysis of this type of compound when using theoretical phase shifts and backscattering amplitudes calculated from FEFF. A comparison of this model with that previously reported' for the analysis of the EXAFS spectrum of Sm203 with McKale's phase shifts and backscattering amplitudes shows that fit parameters obtained with both models are in agree- ment within the precision of the EXAFS technique.In general, higher Debye-Waller factors are calculated when using McKale's tables. The main differences are found in the parameters determined for Sm-Sm shells. The fit achieved 2 4 6 81012 2 4 k/A-' RIA Fig. 6 Sm20,: Fit of the whole EXAFS spectrum. (a)EXAFS oscil- lations; (b) uncorrected k3-weighted FT (Ak = 3-11.5 A-'). (-) Experimental data; (---) theoretical data (EXAFS parameters in Table 5). with McKale's tables requires three Sm-Sm shells to repro- duce the experimental ~pectrum,~ and overestimates the coor- dination number at the shortest distance (1.7 Sm neighbours at 3.35 A) when compared with the crystallographic value (0.67 Sm neighbours at 3.342 A).On the other hand, this short Sm-Sm distance is omitted when fitting with the phase shifts and backscattering amplitudes calculated from FEFF, which may be justified by its low coordination number. In this case the isolation of both contributions was not attempted, since a theoretical EXAFS spectrum including all the distances in Table 1 shows that there is severe overlap- ping between both maxima in the uncorrected FT. The pres- ence of the L,,-edge at 410 eV beyond the L,-edge leads to a limited data range in k space and thus to poorly resolved FTs. Moreover, contributions to the FT due to oxygen neigh- bours between 4.5 and 4.8 A are in this case alsounresolved, appearing as a weak shoulder at long R values, with the maximum at ca.3.8 A. A detailed analysis of the theoretical EXAFS spectrum generated by including all the La-0 and La-La distances in Table 1 shows that nearest oxygen shells give rise to two dis- tinct maxima in the 0.5-1.5 8, range (uncorrected), and can be modelled by two different La-0 distances at 2.38 and 2.72 A, with coordination numbers of four and three, respectively (Table 6). Destructive interferences were found between the EXAFS oscillations generated by these two La-0 subshells. The analysis of the theoretical spectrum also shows that the positions of La-La absorber-backscatterer pairs between 3.768 and 3.940 A can be approximated to only one distance at 3.90 and those of oxygen neighbours at longer distances can be approximated to one shell at 4.62 A.As shown in Table 6, a model including two shells for the shortest La-0 distances, and two more shells for La-La and long La-0 absorber-backscatterer pairs leads to mean distances and coordination numbers in agreement with XRD data within the precision of the EXAFS technique. Deviations in Table 5 EXAFS parameter values for crystalline compounds: Sm,O, coordination N Ao2/A2 RIA AE'IeV 0 4.3 f0.3 0.0057 f0.o006 2.32 f0.006 -0.8 f0.5 0 2.7 f0.3 0.0231 & 0.0010 2.49 f0.01 -15.0 f0.8 Sm 7.5 f0.1 0.0069 k0.0001 3.64 +_ 0.001 -0.5 & 0.1 Sm 4.6 f 0.1 0.0064 f0.0003 4.15 +_ 0.003 -7.2 f0.2 J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Table 6 EXAFS parameter values for crystalline compounds: La,O, (a) Theoretical EXAFS oscillations four-shell model 0 4.0 0.001 56 2.38 0.4 0 3.1 O.OO0 90 2.72 1.o La 13.1 0.006 57 3.90 -1.4 0 11.9 0.00461 4.62 1.o three-shell model 0 4.0 0.001 56 2.38 0.4 0 3.1 O.OO0 90 2.72 1.o La 13.0 0.005 76 3.93 -4.3 (b) Experimental EXAFS oscillations three-shell model 0 4.1 f0.4 0.0112 fo.oO06 2.39 f0.01 2.0 f0.4 0 3.3 f0.5 0.0055 f0.0010 2.72 f0.01 3.2 f0.8 La 7.1 f0.1 0.0074 f O.OO01 3.87 f0.001 1.4 f0.1 four-shell model 0 4.1 f0.6 0.01 12 f0.0010 2.39 f0.01 2.0 0.6 0 3.3 f0.9 0.0055 f 0.0015 2.72 f 0.02 3.2 f1.5 La 7.2 f 0.1 0.0076 0.0002 3.86 f0.001 2.3 f0.1 0 4.8 f0.2 0.0096 kO.OO08 4.64 f 0.005 1.5 & 0.2 Sg has been set to 0.73 (seetext).absorber-backscatterer distances around the mean values are crystallographic distance, the low La :La coordination reflected in increased Debye-Waller factors and slight devi- number should be noted. An even lower La :La coordination ations in AEo values. Discarding the long La-0 distances in number was reported' when analysing the La203 EXAFS the three-shell model (Table 6) introduces larger errors, spectrum with McKale's phase shifts and backscattering although still tolerable, in the parameters determined for the amplitudes. Several possibilities have been considered to La-La absorber-backscatterer pairs. This analysis of a theo-explain this severe error.Interferences between the waves retical EXAFS spectrum leads to the conclusion that the originating from La-La shells and those originating from three-shell model should be adequate for the analysis of the La-0 shells at distances >3.5 A, not considered in the experimental data of La203. model, can be discarded. Long La-0 distances were not The analysis of the experimental EXAFS spectrum with included in the three-shell model used to analyse the theoreti- two La-0 and one La-La shell leads to the parameters col- cal spectrum, and the oxygen neighbour at 3.678 A was dis- lected in Table 6, which provide an excellent fit of the experi- carded in both the three- and four-shell models, leading in all mental data in k and R space (Fig.7). As expected from the cases to slightly overestimated La :La coordination numbers. analysis of the theoretical spectrum, the total 0 :La coordi- As expected, the inclusion of a fourth La-0 shell accounting nation number is 7.4 at the two closest distances (2.39 and for the 12 oxygen neighbours in the range 4.5-4.8 A (see 2.72 A), which agrees within experimental error with the crys- Table 6) does not improve significantly the La :La coordi- tallographic values. However, although the value for the nation number. In our view, the high reactivity towards H20 La-La distance (3.87 A) is also in agreement with the mean and C03 of the La203 when exposed to the atmosphere should not be forgotten since, although the La,O, sample was calcined immediately before its use, it is difficult to avoid contamination during handling and recording of the EXAFS 4 spectrum.Fig. 8(u) compares the EXAFS spectra for a just calcined La203 sample and an La203 sample that had been exposed 2 to the atmosphere for six months. Changes in the coordi- nation of lanthanum ions after exposure to the atmosphere n 30 are evident; thus, at long distances [see Fig. 8(b)], the m maximum at ca. 3.5 A in the uncorrected FT, ascribed to -Y La-La absorber-backscat terer pairs, decreases in intensity -2 and a new maximum appears at ca. 4 A. These changes may be ascribed to the formation of lanthanum hydroxy-carbonates. The crystalline structure of lanthanum hydroxy- -4 carbonate phases can be understood as a regular succession 246810 2 4 6 down the main crystallographic axis of a planar distribution k1A-l RIA of alternating [La(OH)2+], layers and C0,'-anions, Fig.7 La,O,: fit of the whole EXAFS spectrum. (a) EXAFS oscil-resulting in a separation between the [La(OH)2+], layers of lations; (b) uncorrected &'-weighted FT(A&= 3-9.5 A-1). (-) cu. 5 A.2' The hydration of the A-type Ln203 samples has Experimental data; (---) theoretical data (EXAFS parameters in been demonstrated to occur through the penetration of H20 Table 6). along the ternary axis of the oxide, which causes an increase J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 l*OI Fig. 8 La,03: changes in the EXAFS spectrum after exposure to the atmosphere; (-)just calcined sample; (---) sample exposed to the atmosphere for six months.(a)EXAFS oscillations; (b)associated k3-weightedFT at long distances (Ak = 3.0-9.9 A-'). in the cell volume.22 This topochemical mechanism is also present in the substitution of C1-, NO3-and/or SO4'-anions in lanthanide hydroxy Taking into account the above evidence, we propose that La,03 undergoes a trans- formation during the EXAFS experiment in which [Lao];' layers are separated by OH-and/or CO,'-anions, accord- ing to a model proposed earlier,24 resulting in a structure similar to that found in LaOCl (BiOCl type). If this is the case, and taking into account the fact that the layers are formed by [OLa,] tetrahedra sharing edges, we should expect that the La3+ ions are surrounded by eight oxygen ions at distances of around 2.40 A and four lanthanum ions at 3.90 A.The low La-La coordination number should then be ascribed to the incipient transformation of the La203 sample into an oxohydroxide during the recording of the EXAFS data. Application to Dispersed Systems The EXAFS parameters obtained in the analysis of the spectra of the bulk sesquioxides can be used to model the coordination polyhedra around the lanthanide ions in the analysis of the EXAFS spectra of unknown systems. The examples selected are two Ln,O,/Al,O, samples where X-ray diffraction fails to detect the structure adopted by the lantha- nide phases, since XRD patterns contain only lines that can be ascribed to y-Al,O, .Fig. 9(a) shows the EXAFS spectrum at the Sm L,-edge for a 5 mol% Sm,03/A1,03 sample, which was obtained by a -1.0 1 1" _-1 4 6 8 10 12 2 k/A-RIA Fig. 9 (a)Experimental EXAFS oscillations (-) and best-fit func- tion (---) for a 5 mol% Sm,03/Al,0, sample. (b)Uncorrected k3-weighted FTs for Sm2O3/Al2O3 (-) and bulk Sm203 (---, Ak = 2.1-12 A-' in all of the FTs). coprecipitation method' and calcined at 800 "C, and its associated k3-weighted FT [Fig. 9(b), solid line]. The uncor- rected k3-weighted FT for bulk Sm203 has also been included for comparison, showing that the local surroundings of the Sm atoms are similar in both compounds, although the maximum at cu. 3.5 A, ascribed to Sm-Sm absorber-backscatterer pairs, has almost vanished in the Sm,03/A1,0, sample, thus indicating the presence of small Sm,O, particles in the coprecipitated sample. The EXAFS spectrum of Sm,03/Al,03 was analysed by introducing, as a first guess, the set of parameters obtained for the bulk oxide.The ampli- tudes of the shells (Nand Aa') were first allowed to change to obtain the best fit of the experimental data, and then AEo and R were taken into the refinement. The set of parameters obtained and the best-fit function are included in Table 7 and Fig. 9(a) dashed line. Fig. 1qa) includes the EXAFS spectrum at the La Lmedge for a 10% w/w La,03/A1,0, sample, which was obtained by an impregnation method' and calcined at 900 "C. When com- paring its associated k3-weighted FT [Fig. lqb)] with that of bulk La203 [Fig.lqc), solid line], the main differences are found in the high4 region, where the maximum at ca. 3.5 A ascribed to La-La absorber-backscatterer pairs has almost disappeared in the supported sample. However, attempts to fit the EXAFS spectrum of La,O,/Al,O, with only two short La-0 distances cannot account for the peak at ca. 3 in the uncorrected FT, thus suggesting the presence of new neigh- bours in the lanthanum coordination sphere. These new neighbours have been identified as aluminium atoms by com- parison with the EXAFS spectrum of a 40% w/w Table 7 EXAFS parameters for Ln203/A1203 samples coordination N Aa2/A2 RIA AEo/eV ~ ~ ~~ ~~ 0 4.7 f 0.3 0.0106 f o.Oo04 2.38 f 0.004 -1.1 f 0.4 0 1.9 f 0.4 0.0267 f 0.0015 2.70 f 0.016 -2.6 f 1.5 Sm 2.7 f 0.1 0.0096 f 0.0003 3.72 0.002 2.3 f 0.3 10% La,O,/Al,O, 0 4.4 f 1.0 0.0094 f 0.0019 2.50 f 0.02 1.7 f 0.3 0 3.8 & 1.0 0.0055 f 0.0035 2.71 f 0.02 2.4 f 0.8 A1 3.9 & 0.1 0.0140 f 0.0002 3.34 f 0.002 -6.4 f 0.2 40% La,O,/Al,O, 0 5.0 & 1.4 0.0042 f 0.0019 2.48 & 0.02 4.4 f 0.5 0 7.4 f 1.9 0.0032 f 0.0030 2.70 f 0.02 1.5 f 0.6 A1 6.2 f 0.2 0.0015 f O.Oo04 3.30 & 0.004 -2.3 f 0.3 La 7.0 f 0.2 0.0068 f 0.0002 3.74 f 0.002 0.0 f 0.1 2790 1.a 1 00.5 E-1 5 h rc v)?5 0.0 w 55 U .-z L-0.5 JO9 -1 .o -5 Fig.10 (a) Experimental EXAFS oscillations (-) and best-fit function (---) for a 10% w/w La,O,/AI,O, sample.(b) Uncorrected k3-weighted FT for 10% w/w La,03/Al,03. (c) Uncorrected k3-weighted FT for bulk La,O, (-) and a 40% w/w La,O,/AI,O, sample (---, Ak = 2.6-9.8 A-' in all of the FTs). La,O,/Al,O, sample [Fig. lqc), dashed line] prepared by the same procedure, where the formation of an LaAlO, phase can be unambiguously stated by XRD. The analysis of the EXAFS spectra for the latter sample (Table 7) yields an 0 : La coordination number of 12.4 and the presence of La-A1 and La-La absorber-backscatterer pairs at 3.30 and 3.74 8, respectively, which are the main species responsible for the peaks at ca. 3 and 3.5 A in the uncorrected FT. These results are in agreement with XRD data and with the RDF, which can be calculated from crystallographic data for LaA10,,25 where A1 and La neighbours appear around the lanthanum atoms at mean distances of 3.28 and 3.79 A, respectively.Comparison of the uncorrected FTs for the 10% and 40% w/w La,O,/Al,O, supported samples allows us to ascribe the maximum at ca. 3 8 in the sample with low loading to A1 neighbours, thus suggesting the incipient for- mation of a bidimensional LaA10, phase around the lantha- num atoms in this sample. The best-fit function obtained by considering two La-0 and one La-A1 shell is included in Fig. 1qa) and corresponds to the EXAFS parameters given in Table 7. Conclusions We have shown that L,,,-edge EXAFS spectra of Ln,O, oxide systems, having a complex structure, can be modelled with a feasible number of shells even in the case of lantha- num, for which the energy difference between the L, and L,,, absorption edges is only ca.400 eV. Theoretical phase shifts and backscattering amplitudes calculated from FEFF, with an amplitude reduction factor of Sg = 0.73, are shown to be reliable in the modelling of the experimental spectra, thus overcoming the difficulty of finding EXAFS reference com- pounds with regular coordination around the lanthanide element. The obtained coordination polyhedra, which model the complex coordination around the lanthanide atoms in the oxides, are found to be useful as initial models for obtaining J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 information on the local order around rare-earth-metal ions in highly dispersed samples of unknown structure.In La,O,/AI,O, samples, the analysis procedure allows the detection of aluminium atoms, which are present around lanthanum ions alongside oxygen atoms. Finally, it has been shown that the analysis of systems con- taining lanthanum has to be undertaken carefully, since the hydration/carbonation undergone by La,O, when exposed to the atmosphere can result in changes in the coordination polyhedron around lanthanum. Financial support has been obtained from DGICYT (Projects PB88-0257 and PB92-0665). References 1 Dynamical Processes in Disordered Systems, ed. W. M. Yen, Materials Science Forum, Trans Tech Publications Ltd., Aeder- mannsdorf 1990, vol. 51. 2 Halide Glasses, ed.M. Yaname and C. T. Moynihan, Materials Science Forum, Trans Tech. Publications, Aedermannsdorf 1988, vol. 32,33. 3 S. H. Sohn and Y. Hamakawa, Jpn. J. Appl. Phys., 1992, 31, L963. 4 Y. He, S. J. Poon and G. J. Shiflet, Science, 1988,241,1640. H. Schaper, E. B. M. Doesburg and L. L. van Reyen, Appl. Catal., 1983, 7,21 1. 6 Y. Xie, M. Qian and Y.Tang, Sci. Sin. Ser. B, 1984,6, 549. 7 M. J. Capitan, P. Malet, M. A. Centeno, A. Muiioz-Paez, I. Car-rizosa and J. A. Odriozola, J. Phys. Chem., 1993,97,9233. 8 P. Malet, M. J. Capitan, M. A. Centeno, J. J. Benitez, I. Carri-zosa and J. A. Odriozola, Stud. Surf Sci. Catal., in the press. 9 P. Malet, J. J. Benitez, M. J. Capitan, M. A. Centeno, I. Carri-zosa and J. A. Odriozola, Catal.Lett., 1993, 18, 81. A. G. McKale, G. S. Knapp and S-K. Chan, Phys. Rev. B, 1986, 33, 841. 11 A. G. McKale, B. W. Veal, A. P. Paulikas, S-K. Chan and G. S. Knapp, J. Am. Chem. SOC., 1988,110,3763. 12 D. C. Koningsberger, Jpn. J. Appl. Phys., 1993,32,532; 877. 13 J. Mustre de Leon, J. J. Rehr, S. I. Zabinsky and R. C. Albers, Phys. Rev. B, 1991,44,4146. 14 D. E. Sayers and B. A. Bunker, in X-Ray Absorption: Principles, Applications and Techniques of EXAFS, SEXAFS and XANES, ed. D. C. Koningsberger and R. Prins, Wiley, New York, 1988. (a) R. S. Roth and S. J. Schneider, J. Res. NBS, A, Phys. Chem., 1960, 64, 309; (b) A. Fert, Bull. SOC. Fr. Mineral. Cristallogr., 1962,85, 267. 16 D. T. Cromer, J. Phys. Chem., 1957,61,753. 17 W. C. Koehler and E. 0.Wollan, Acta Crystallogr., 1953,6,741. 18 (a) B. K. Teo, EXAFS: Basic Principles and Data Analysis, Springer-Verlag, Berlin, 1986, pp. 165-170; (b)pp. 85-89. 19 (a) F. W. H. Kampers, Thesis, Eindhoven University of Tech- nology, 1988, p. 38; (b) F. W. H. Kampers and D. C. Kon- ingsberger, Faraday Discuss. Chem. SOC., 1990,89, 137. E. A. Stern, B. Bunker and S. M. Heald, in EXAFS Spectros- copy: Techniques and Applications, ed. B. K. Teo and D. J. Joy, Plenum Press, New York, 1981. 21 H. Dexpert, G. Sciffmacher and P. Caro, J. Solid State Chem., 1975,30,301. 22 R. Tueta and A. M. Lejus, Rev. Chim. Miner., 1973,10,105. 23 V. A. Msorin, V. V. Sakharov and L. M. Zaitsev, Russ. J. Znorg. Chem., 1974,19,804. 24 P. E. Caro, J. Less-Common Met., 1968, 16, 367. S. Geller and V. B. Bala, Acta Crystallogr., 1956,9, 1019. Paper 4/01402B; Received 9th March, 1994

ISSN:0956-5000    

DOI:10.1039/FT9949002783

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

J. CHEM. SOC. FARADAY TRANS., 1994, 90(18), 2791-2797 Single-crystal Structure of C,, at 300 K Evidence for the Presence of Oxygen in a Statically Disordered Model W. Bensch, H. Werner, H. Bartlt and R. Schlogl*$ lnstitut fur Anorganische Chemie University of Frankfurt, Marie-Curie Str. II,60439 Frankfurt, Germany The structure of solvent-free single crystals of sublimed c6, has been analysed at 300 K by X-ray diffraction. A unique solution of atom coordinates gave a satisfactory agreement with the intensity data. This indicates that the c60 balls exhibit a time-averaged preferred position in the crystals which are, however, heavily affected by rotational disorder. The structure reveals the truncated icosahedral molecular shape with two sets of carbon- carbon bond distances.Contour plots indicate the anisotropic distribution of electron density within the almost perfectly spherical shell of the molecule. The analytically pure crystals contained an impurity of molecular oxygen located statistically over some hexagons of each 'buckyball ' resulting in a limiting stoichiometry of c60 O2. Based on NMR,'.' X-ray diffraction studies3v4 and neutron diffraction data5 it is frequently assumed that within the molecular crystals of c60 the highly symmetric buckyballs rotate freely at 300 K. This would preclude structure analysis by diffraction methods at atomic resolution. The high molec- ular symmetry of the buckyball allows for such molecular dynamics, the detailed explanation of which, in terms of the nature of the intermolecular interaction potential, has led to considerable and controversial theoretical effort^.^.^ Very recently, the NMR interpretation of isotropic rotation with a high rotation frequency close to the value of free rotation was stated again in a study of the orientational ordering of c60.' One way to suppress the adverse influence of the molecular dynamics of the free molecules is to use low-symmetric deriv- atives of c60 either as metal-organic complexes9-" or cla- thrate c~mpounds.'~-'' A significant body of detailed structural information has now emerged from such studies confirming our general picture about the molecule but differ- ing in numerical details relevant for e.g.theoretical studies. This may be related to a certain distorting effect of the inter- acting ligands on the pristine structure. Another method to analyse the atomic structure of ful-lerenes is their analysis at very low temperatures where molecular dynamics is frozen out. Such studies carried out on powders5 and on single crystals4 yielded both structural details and information about the restricted molecular dynamics at intermediate temperatures.One study16 using precision intensity data measured as function of temperature found a first-order transition at 259 K followed by a complex evolution of orientational ordering with two anomalies at 165 K and 85 K. In addition, time-dependent effects of heating and cooling on diffracted X-ray intensities were noted. The numerous studies on the solid-state dynamics below the first- order phase transition with diffraction techniques, thermal methods and spectroscopy are summarised in a detailed review.17 In there it is pointed out that disorder is an intrin- sic property of a high-symmetry crystal made from molecules with non-crystallographic symmetry elements.The single-crystal structure of the OsO, derivative of c60 yielded a football made up from flat pentagons and hexagons in the expected topology with a radius of 351.2(3) pm and two types of bonds of length 138.8(9) and 143.2(5) pm. The t Institut fur Kristallographie der Universitat Frankfurt. $ Present address: Fritz Haber Institut der Max-Planck Gesell- schaft, Faradayweg 4, D-14195 Berlin, Germany. synchrotron powder diffraction study of pure c6, at 300 K was analysed with a model of freely rotating carbon atoms describing a sphere of homogeneous electron density with a radius of 352(1) pm.Numerous st~dies~-',~' have reported a phase transition at ca. 250 K associated with a change of the lattice type from centred F into simple cubic P.A phase tran- sition was also observed in NMR which con- cluded that in the high-temperature form the molecular carbon resonance is motionally narrowed by either free rota- tion or a rapid jump rotation. The transition temperature was, however, significantly lower as observed by diffraction and by thermal In a more detailed analysis21 the longitudinal relaxation time was found to exhibit, at 250 K, an anomaly which was interpreted as an indication of a transition from an almost free rotation to a jump rotation with a change in rotational barrier from the high-temperature value of 50 meV to a low-temperature value of 250 meV.Details of this analysis can be found in the review17 on the ordering transition. The recent DTA study" detected two high-temperature transitions at 250 and 310 K and con-cluded that structural and molecular-ordering transitions may be independent phenomena in the complex molecular dynamics of fullerenes. The transition temperature for the structural phase transition from fcc (face-centred cubic) to sc (simple cubic) varies in the literature between 249 and 260 K.22 This discrepancy in the detection of a first-order tran- sition which should exhibit a sharp change in the ordering parameter arises from different techniques of observation (diffracted intensities, thermal conductivity, electrical conductivity) and different definitions of the transition tem- perature (onset, inflection or completion of the anomaly) but implies further that details of the crystal quality may influ- ence the solid-state dynamic properties.The structural analysis of solvated varieties has clearly shown how important the sample purity is for an analysis of this solid, being affected by extraordinary rotational disorder effects arising from the highly symmetric shape of the molecu- lar units. It is often assumed that c60, after sublimation in high vacuum, is pure and air stable at ambient conditions.It could be shown, however, that purified c60 does take up molecules from its gaseous environment and intercalates them into its void ~ystem.~~-~~ The kinetics of gas uptake is severely affected by the stacking defect density of the solidz4 and may vary from crystal to crystal as has been recently illustrated by high-pressure studies27 which gave clear evi- dence of the dependence of the phase-transition temperature on the structural and chemical integrity of the c60 material. For this reason special attention is given to the possible pres- ence of molecular oxygen in the crystals. The importance of lattice defects for the solid-state proper- ties was also highlighted in a study2' of c60 by low-temperature high-resolution electron diffraction crystal-lography. Superlattice structures which were not detectable by integral X-ray diffraction methods,' were clearly apparent in the spatially high resolving electron diffraction patterns.The presence of such ordered defects in addition to stacking order defects will affect the structural dynamics directly and indirectly oia a modified rea~tivity~~,~,towards various gases. The purpose of the present single-crystal diffraction study using conventional crystallographic tools is to test the possi- bility of obtaining structural data at atomic resolution of nominally pure c60 at ambient temperature. First results were reported previously30 showing the phase transition to occur in our crystals at 255 K by the criterion of non-zero reflections forbidden in the fcc phase which is fully in agree- ment with later literature reports.,, Most of the diffraction studies published so far used either sub-ambient tem-peratures, derivatives of c60 or unconventional diffraction techniques and focused on the structural dynamics of the material.The present study is intended to serve as a reference for structural investigations of molecular adducts of fullerenes with single-crystal techniques. In addition, the effect of molec- ular oxygen, unavoidably present in fullerenes handled in air, on the crystal properties will be explored. Finally, as the importance of lattice defects as an essential property of highly symmetric fullerene solids was already stressed in the liter- ature,I7 it seems justified to address a possible discrimination between intrinsic structural properties occurring in all indi- viduals of solid C,, and extrinsic properties which may differ in each crystal of this molecular solid.Results The Material Home-made c60 was highly purified23 and sublimed. Single crystals of 0.1 mm length suitable for structural analysis were grown from the purified powder using the sublimation tech- nique. Several batches were investigated with growth times between 1 and 20 days. A total of 25 crystals were investi- gated with rotation photographs showing frequent twinning of equally large crystals. Only crystals without this twinning were further studied3' at 300 K, at 255 5 K and at 100 K.The present study focuses on the room-temperature structure of c60 and is complementary to structure determinations at low temperatures4v3 below the structural phase transition. FTIR measurements showed that the samples of purified c60 were clean to the standard discussed in a recent vibra- tional study of the structural phase tran~ition.~~ We note that in our spectra and in all literature data the intensity ratio of the spurious doublet at ca. 2327 cm-' did not change with the intensity ratios of the four main fullerene lines. This is taken as indication that a trace of CO, is present in the crys- tals. The densities of several single crystals were determined by the flotation method and found to be on average 1.765 g cm-over seven crystals of batch c with values ranging from 1.74 to 1.77 g ~m-~.Analysis of these crystals with energy- dispersive X-ray spectroscopy gave no detectable impurities of alkali metal or silicon. All handling of the material was done under purified Ar and structure determinations were carried out with the crystals sealed under Ar in Lindemann capillaries. A systematic investigation of the sample density with subli- mation parameters revealed a clear dependence of the density on the velocity of material deposition in the hot zone. The J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 20 15 10 5 r-:10 Q)CI.c 5 0 15 10 5- 0- I I a 10 12 28/degrees Fig. 1 Section of powder diffraction patterns around the (111) reflection (Cu-Ka transmission geometry, samples in Lindemann tubes) of three batches (a, b, c) of sublimed C6*.The asymmetric 'foot' superimposed on the normal Bragg reflection arises from stacking defects.averaged density varies from 1.765 g cm-3 (batch c) for 20 days to 1.720 g (batch b) after 2 days to 1.694 g (batch a) for rapid sublimation within 1 day. The calculated density of pure c6, is 1.689 g cm-3 with four molecules in the unit cell (lattice parameter 1415.2 pm). Note that the measured density is significantly above this value despite the diluting effect of lattice defects present in real crystals. High-resolution powder diffraction data of the three sample batches indeed revealed differences in stacking order.Fig. 1 shows the enlarged portion of the diffractograms around the (111) reflection revealing a varying degree of stacking disorder indicated by the wedge-shaped wide 'foot' of the main peak.33 These findings suggest strongly the presence of a light-element impurity in varying amounts. The impurity was iden- tified by several analytical methods including tempera-ture-programmed desorption (TPD), FTIR measurements, isotope labelling experiments and EPR spectroscopy. There is no doubt that the crystals contained extra oxygen present at 300 K as molecules. Heating above 450 K in inert atmo- sphere or vacuum removes most of the oxygen. Exposure to oxygen at elevated temperatures leads to the formation of epoxides and eventually to fullerene decomposition with evolution of CO and CO,.Details of the reaction analysis can be found el~ewhere.,~-,~ The identification of intercalated oxygen in single crystals of C,, showing no signs of structural disorder or defects in X-ray photographs but exhibiting a too high density is regarded as a significant finding and will be illustrated by an experiment carried out with batch c of the material used in the present study. Note that the crystals were optically smooth and exhibited a dark lustre in the microscope. Such material was heated in ultra-high vacuum to the sublimation temperature of ca. 700 K and exposed, after cooling to 300 K, to an atmosphere of 1802 and I6O2. After evacuation at 300 K the crystals were exposed to lo-, mbar I6O2 and heated at a rate of 0.25 K s-' to 650 K.The mass spectrometric response recorded with parallel detection is displayed in Fig. 2. The desorption of molecular oxygen of masses 32 (l6O,) J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Fig. 2 Temperature-programmed desorption of 180,-labelled oxygen from C,, . The desorption intensities (measured with a quad- rupole mass spectrometer equipped with a multichannel option from Hiden Analytical and processed for background contributions) were observed during a linear heating experiment of ca. 10 mg material of batch b under a vacuum of 1 x mbar. Dosing was done in situ at 450 mbar. The fragments chosen illustrate the (isotope labelled) evolution of CO (m/z 28, 29), 0, (m/z 32, 34, 36) and CO, (m/z 44,46, 48).The intensity scale is equal for all fragments and indicates the ion current in A. and 36 (1802) is clearly evident. No scrambling products at m/z 34 were observed indicating the molecular nature of the intercalate. Above 580 K oxidation of the fullerene caused the evolution of CO and CO, . The significant amounts of l80 incorporated indicate that the intercalated molecular oxygen acted as a precursor in the oxidation reaction. The small amount of CO being desorbed with the molecular oxygen was caused by incomplete reversibility of the intercalation : purification of c60 from oxygen in high vacuum always leads to a small amount of gasification products which are co- intercalated into the fullerene crystal.Single-crystal Data Collection and Data Reduction X-Ray studies were performed on a CAD-4 instrument at 300 K (Cu radiation) and on a STOE AED I1 instrument in the temperature range 100-300 K with Mo radiation. The known structural phase transition at ca. 250 K was found3' with all crystals at 256 & 3 K. The change of the lattice type was fol- lowed by the intensity changes of several reflections of the P system not allowed in F symmetry which occur irrespective of the heating/cooling rate (between 1 and 20 K h-') reversibly and in a narrow temperature range of a few K. An integral investigation of the transition with oriented oscil- lation photographs (orientation [ 1001, 30" oscillation angle, cooling rate 8 K h-I) showed that at 254 K the expected additional P-allowed reflections3' begin to occur.Further cooling to 100 K caused a significant gain in intensity of the sharp reflections besides partial disintegration of the crystal into micro-domains as seen from rings of diffracted intensity. Warming to 300 K for 24 h restored the initial crystal quality (reflection profiles). The volume of the unit cell changed from 2.8344 x lo3 pm3 at 300 K to 2.7976 x lo3 pm3 at 200 K and to 2.7759 x lo3pm3 at 100 K. The crystal structure at 300 K was determined from a crystal of batch c using a data set collected with Cu radiation in the 28 range from 2" to 148" yielding 1532 observations. The density of the crystals as well as the volume of the unit cell require four c60 molecules per translation unit.The point symmetry of the molecule is m35 of which m3 is a subgroup satisfying the crystallographic translational requirements. It is plausible to adopt this symmetry for the crystal structure model of a face-centred cubic closed-packed arrangement of the buckyballs. The resulting space group in accordance with the point symmetry is Fm-3. With this assumption the mea- sured intensities reduce into 281 unique data with an internal R value of 2.55%. The lattice constant was at 300 K 1415.2(1) pm, at 200 K 1409.1(2) pm and at 100 K 1405.4(3)pm. The alternative space group Fm-3m requires an additional four-fold symmetry axis which would in a non-twinned crystal only be possible with a special packing of two pairs of molecules rotated by 90".Merging the intensities in space group Fm-3m leads to 187 unique data with an internal R value of 3.22%. The difference in the internal R values between the two Laue groups amounts to 0.67% which is too small as a single indication for a clear distinction of the correct space group. In Laue group m3 the intensities are only cyclically permutable whereas in m3m the intensities are permutable. The sensitive intensities I(hkZ) and Z(khl) were thus extracted from the measured data set and averaged in the two Laue groups. This procedure yields an internal R value of 4.30% for m3 and 5.93% for m3m. These results together with the observation that the least-squares refine- ments in the space group Fm-3 lead consistently to a better model, led us to conclude that Fm-3 is the correct space group for the present crystal.Single-crystal diffraction data sets collected from various crystals grown from batches a and b as starting material yielded similar but non-identical lattice parameters and inten- sity distributions in the scattering law. Some examples are listed in Table 1. Using the described method for distinction of the space group the difference in R value between the two possibilities varied significantly with one example yielding even better R value for the space group Fm-3m. The scattering law for the crystal of batch c, i.e. the total scattered intensity as a function of the scattering angle is dis- played in Fig. 3. This unusual representation of a single-crystal data set shows that the total scattered intensity decays rapidly with the scattering angle indicating significant dis- order within the crystal.The intensity shows in addition an unusual periodic modulation which is caused by the spherical shape of the buckyball molecule. Its description with a homo- geneous sphere of electrons represented by a higher-order Bessel function3 (in the shape of the envelope of the intensity distribution) is, however, for small scattering angles a poor representation of the scattering law. Structural Model In the early stages of refinement attempts were made to fit the observations with three unique C atoms with isotropic Table 1 Structural parameters of several single crystals of C,, crystaldesignation (batch lattice parameter and number) symmetry IPm b5 cubic 1416.6 (30) b7 cubic 1405.4 (20)" 1417.9 (2) b 11 cubic 1406.9 (l0)ll 1415.8 (2) a1 cubic 1418.2 a2 cubic 1415.5 (1 1) a3 cubic 1417.7 a7 cubic 1419.9 (5) hexagonal 1003, 1003, 2456 " Determined at 100 K.2794 J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 'Oo0 1 0.1 0.2 0.3 0.4 0.5 0.6 sin Of2 Fig. 3 Scattering law of C6, at 300 K. The integrated intensity dis- tribution is shown for an average in space group Fm-3. thermal parameters. Only one set of parameters and a restrained variation of the bond lengths [C(l)-C(2): 142.0(5) pm, C(2)-C(3): 134.q5) pm, C(1)-C(3): 243.q5) pm] led uniquely to the truncated icosahedral shape of the molecule with a diameter of 348 pm.This model yielded R values between 30% and 40% depending on the constraints and the number of observations used. In the following steps starting values for the main axes and orientations of the probability ellipsoids were applied such as to simulate the spherical average surface of the buckyball. In this stage of refinement it is essential to assign the correct signs to the different U,, whereas the magnitude of the Uii and U, displacement parameters is of minor importance. No further constraints were applied to the anisotropic displace- ment parameters during the refinement. Different disorder models were considered with split sites for the three carbon atoms to account for possible static rota- tional disorder.The main problem of such models was, as expected, to find the correct site occupation factors (sof) for the splitted carbon atom sites as the sof are highly correlated with the displacement factors. Several disorder models gave better R values which may be due to the enlarged number of parameters. In all cases, however, the anisotropically refined carbon atoms were not defined positively. Therefore, we decided to refine the structural model with only three carbon atoms with the disadvantage of high values for the U, com-ponen ts. The final refinement of the three carbon atom model resulted in an R value of 14.3%using 270 observations and 24 parameters. The highest residual electron density of 0.375 x e pm3 was located above 8 of the 20 hexagons at a distance to carbon of 147 pm.This relatively low electron density must be compared with the electron density of a carbon atom found in a difference Fourier synthesis using only two carbon atoms in the refinement procedure. The refinement with C(2) and C(3) (80% total scattering power from the C atoms) exhibited a difference peak with a height of 0.59 x e pm3 for the missing C(l) atom. It needs to be considered that the residual electron density pointing to an additional atomic site above some of the six-membered rings may be an artefact from series termination effects. The only selective occurrence and the high overall temperature factor of the structure render this explanation highly unlikely.Keeping in mind the proven presence of molecular oxygen and noting that the (hOO) reflections were non-zero in intensity we concluded that the additional atomic site marks one atom of the oxygen molecules attached to the buckyballs. Fig. 4 Average structure of the C,, molecule with its oxygen adduct in the crystal at 300 K. The three unique carbon atoms are labelled. The oxygen molecules occur with detectable electron density only for the bonding atom, the free arom has too many inequivalent locations in the void of the crystal packing (see text). The position of each oxygen atom is only partly occupied. An oxygen atom Cjustified by non-crystallographic arguments) on a partially occupied special position was included in the refinement.The sof of the oxygen atom was refined to 0.029(4) [U,,= 0.05(1)]. This complies with eight atoms per unit cell (maximum stoichiometry C,,O,) resulting in a perfect match between calculated and measured densities. The R value reduced finally to 11.20% (270 observations, 29 parameters) with the weighted R value coming down to 5.18%. This value needs to be compared with the literature data from powder profile refinements of 8-10%17 and with the internal R value of our data set of 4.30%. The resulting molecular shape of the oxygen adduct with C,, is displayed in Fig. 4. The parameters of this model are given in Table 2. The dashed oxygen atom positions indicate one atom of an oxygen molecule chemisorbed to the surface of a buckyball molecule.This coordination is not to be mis-taken for a covalent bond, as an increase in temperature of less than 100 K is suficient to remove the interaction (see Fig. 2). The short distance of 30 pm between the molecule and the carbon surface is indicative of a significant dipolar intera~tion~~between the molecules giving rise to the EPR23 signal of the adduct. The low binding energy of the adduct despite the significant orbital interaction is in line with the electron-poor character of the C,, molecule. The other atom of the oxygen molecule has many locations owing to the lack of spatial constraints in the voids of the fullerene crystal so that it does not show up as a clear maximum in the difference Fourier map. For the buckball two sets of bond lengths of ca.134 and 142 pm were obtained. These values are artificially shortened by the rotational disorder as discussed by Bur@ et aL4 and no meaningful correction is possible in the present model without data about the details of the disorder. The model Table 2 Parameters of the unique structural elements for c6, at 300 K atom X Y z U C(1) C(2) C(3) 0(1) 0.2349 (5) 0.2015 (5) 0.1722 (4) 0.1379 (5) 0.0 0.0845 (4) 0.1599 (2) 0.1379 (5) 0.0517 (7) 0.0946 (4) 0.0469 (7) 0.1379 (5) 0.358 (4) 0.482 (4) 0.430 (4) 0.150 (4) J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 requires, however, two sets of bond lengths with a difference of ca. 5 pm in order to yield a buckyball as solution. It was noted that the other models which do not expand into even- shaped balls lead to bond lengths in better agreement with gas-phase data34 and lead also to better R values but are considered as chemically meaningless.The carbon atoms exhibit large anisotropic displacement parameters (ADP) indicating significant intermolecular dis- order (projection of different rotamers). The overall effect of such high ADP on the calculated structure factors is their faster decay with increasing scattering angle compared to a well ordered situation. Exactly this is observed experimen- tally, as illustrated in Fig. 3. A significant improvement of the R value towards the internal R value may thus be achieved artificially by omitting suitable observations or in a physi- cally meaningful manner by considering different rotamers in the model. This latter procedure was successfully applied to the low-temperature structure which was explained by the superposition of two unequally populated rotamer~.~By assigning the phase transition of 250 K to an order-disorder transition it may be seen as an increase in the number of rotamers from two to a finite larger number at high tem- perature.Their small difference in energy renders it, however, inadequate to model these high-temperature rotamers con- sidering the other types of influences on the orientational dis- tribution as there are stacking defects and impurities. It is, further, quite likely that their population is different from crystal to crystal. The finite number of rotamers differs from the picture of free rotation and requires an averaged uneven electron dis- tribution on the buckyball surface.This is demonstrated by electron-density maps within and perpendicular to the least- squares plane of one hexagon of the buckyball. Results at various temperatures are shown in Fig. 5. It is evident that the electron density in the high-temperature form is smeared out but not uniform. Lowering the temperature through the t t Fig. 5 Electron density distribution, s, calculated from the measured scattering law for one six-membered ring at various temperatures (top 300 K, centre 248 K, bottom 100 K). The electron density maximum (at the calculated atom positions denoted by dots) is 3.5 x lop3e ~m-~.The length scale is in units of 5 pm. Fig. 6 Electron contour plot of a section through the equator of two C,, molecules at 300 K. The dots represent the atom positions of the six-membered ring used for the electron contour plots in Fig. 5. The contours range from 0.5 x to 3.5 x e pm-’. phase transition reduces the number of rotamers and clearly improves the localisation of the electron density at the ring corners. This must not be mistaken for a change in the intra- molecular electron density distribution as it is only a conse- quence of the reduction in the number of projected rotamers. In addition, we found that the variation of the observed structure amplitudes with increasing scattering angle is aniso- tropic for different directions (the plot in Fig.3 is the spatial average). Such a modulation is not in agreement with the expected scattering factor for a freely rotating buckyball. The cross-section through the electron density distribution at the equator of one buckyball at 300 K in Fig. 6 shows the expected hollow structure of the buckyball and the empty void space between the molecules. A thin shell of electrons in an inhomogeneous internal distribution forms the wall of the molecules. It becomes apparent that the representation of a C6, molecule as a giant atom is a poor approximation of its spatial electron distribution. A group scattering factor for a c60 ball was calculated using the program NORMAL with the MULTAN package. In agreement with the observed electron-density distribution and the progress of the observed structure factors (see Fig.3) this calculated structure factor exhibits strong modulations in contrast to the shape of a scattering factor derived from a hollow sphere of electrons distributed on the surface of a c60 ball. Discussion The successful analysis of the diffracted X-ray intensity dis- tribution in terms of an atomically resolved molecular struc- ture with a clear minimum between observed and calculated intensities for one set of atomic parameters excludes the model for the molecular dynamics of free isotropic rotation of independent buckyballs in the crystal at 300 K. The present study confirms the shape of the c60 molecule in its crystal at 300 K and illustrates the dominating influence of the rotational disorder on the crystal structure.The time- and space-averaging X-ray diffraction experiment cannot dis- tinguish between a hopping rotation dynamic disorder at 300 K and a static rotational disorder. Time-resolved structural clearly reveal the dynamic nature of the dis- J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 order. It may be described as a tumbling motion of the mol- ecules between several orientationally inequivalent states. The 2D NMR observation^",^^ and the neutron scattering studies5 imply that, indeed, a complex motional system with more than two states is present. These observations let us hesitate to try to fit our observations to a two-state rotation- al disorder system as was done by Bur@ et d4Such models yield good agreements between calculated and experimental scattering laws but overestimate the physical content of the static X-ray diffraction experiment.The measured scattering law is, however, satisfactorily described with a conventional model of a perfect buckyball with the well known point sym- metry of the c6, molecule expanding from one set of three unique carbon atoms. There is no necessity to invoke a special des~ription~,~~*~~,~~ of the crystal structure in terms of a modulated spherical electron distribution which ends at a picture with the same physical content as the present analysis. If the aim of a structural study is the determination of atomically resolved dimensional parameters, then the concept of using sets of derivatives of c60 with large ligands symmetrically bonded to the balls may be more successful than building complicated models of the disorder of the free fullerene molecule which may fall short of physical justification’ within a diffraction experiment.These findings have implications for the solutions of struc- tures of fullerene derivatives in which the buckyballs retain their shape. There is no a priori reason to assume the absence of rotational disorder in compounds without the phase tran- sition at 250 K. In such cases, as represented by intercalation compounds, the fullerene molecules may either be statically disordered (frozen state of the dynamic structure of the pris- tine molecule, ‘glassy’) or slowed in their dynamics resulting in a different phase-transition temperature.Incorporation of neutral molecules such as pentane or noble gases in the void system of c60 does modify the molecular dynamics of such crystals as has been shown by high-pressure treatments fol- lowed by DTA analy~is.’~ Ordering of the c60 molecules with respect to each other would occur at finite temperatures if the undirected intermolecular interaction would be signifi- cantly amplified by spatially modulated charges on each mol- ecule. Such a situation will occur upon charge transfer from the guest species intercalated in the crystal and localisation of the extra charges within the c60 electronic structure. In this sense it may be possible that the space group of a given single crystal may depend on the abundance of polar- ising oxygen molecules which affect the intermolecular orien- tation (e.g.by forming pairs of buckyballs). The distribution of the rotamer population over a large but finite number of possibilities in the fcc phase varies, in the absence of a signifi-cant energy barrier,21 with the growth history of the crystal and seems thus not to be specific for the material c60. This was illustrated by the variation in structural properties of single crystals grown with varying rates of sublimation (see Fig. 1 and Table 1). The parameter variation reflects the dis- order of the molecular crystals. For this reason the choice of space groups which is only differentiated by the relative inter- molecular orientation is also affected by the rotational dis- order and may thus be either Fm-3 or Fm-3m depending on the rotamer population.The definitive statement about the correct choice of the space group for the material c60 which ‘does not adopt the Fm-3 space group at any temperat~re’~seems in the light of the present discussion overstated. The electron density contour plot of Fig. 6 shows clearly that the buckyballs naturally exhibit nodes in the potential distribution which may cause the energy barriers in the rota- tion process required for a jump mode. It is pointed out that the nodes appear even in the experimental electron distribu- J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 2797 tion which is the average over several rotamer orientations.6 K. H. Michel, J. R. D. Copley and D. A. Neumann, Phys. Rev. This result is fully in line with a synchrotron room-temperature single-crystal diffraction study3 arrived at after a different structure analysis with a modulated spherical elec- tron distribution and a spatial inhomogeniety of +lo% of the total electron density of the molecule. The density variation of many single crystals has shown 7 8 9 Lett., 1992,68, 2929. E. Burgos, E. Halac and H. Bonadeo, Phys. Rev. Lett., 1992, 68, 3598. R. Blinc, J. Seliger, J. Dolinsek and D. Arcon, Phys. Rev. B, 1994, 49,4993. J. M. Hawkins, A. Meyer, T. A. Lewis, S. Loren and F. J. Hol- lander, Science, 1991, 252, 312. the range of impurity doping. The identification of purified and cleanly handled c60 as oxygen adduct has implications beyond the possible effect of these additional molecules on the molecular dynamics of the crystals.The interaction between oxygen and c60 is much reduced as compared to 10 11 12 A. L. Balch, V. J. Catalan0 and J. W. Lee, Znorg. Chem., 1991, 30,3980. P. J. Fagan, J. C. Calabrese and B. Malone, Science, 1991, 252, 1160. B. Morosin, P. P. Newcomer, R. J. Baughman, E. L. Venturini, D. Loy and J. E. Schirber, Physica C, 1991,84,21. that in the epoxide C6,0. The epoxide exhibits, however, the same fcc structure as pristine c60 and undergoes a structural phase transition similar to that of pure c60 .38 This indicates that the presence of a single extra atom and thus the oxygen (and C02) impurities may exhibit only a limited influence on the rotational behaviour which is in contrast to the drastic 13 14 15 M.F. Meidine, P. B. Hitchcock, H. W. Kroto, R. Taylor and D. R. M. Walton, J. Chem. SOC.,Chem. Commun., 1992, 1534. S. M. Gorun, M. A. Greaney, V. W. Day, C. S. Day, R. M. Upton and C. E. Briant, in Fullerenes :Synthesis, Properties and Chemistry of Large Carbon Clusters, ed. G. S. Hammond and V. J. Kuck, American Chemical Society, Washington DC, 1992. P. R. Birkett, C. Christides, P. B. Hitchcock, H. W. Kroto, K. impurity effects on the transport properties3’ described recently. Nevertheless, care should be taken to consider puri- fied c60 even as single crystals after exposure to air as a clean material. The small effect of the attachment of an extra atom onto a hypothetically freely rotating buckyball is quite unex- pected as at least orientational preference of the free rotation 16 17 18 19 Prassides, R.Taylor and D. R. M. Walton, J. Chem. SOC., Perkin Trans. 2, 1993, 1407. R. Moret, Phys. Rev. B, 1993,48, 17619. P. A. Heiney, J. Phys. Chem. Solids, 1992,53, 1333. N. Yao, C F. Klein, S. K. Behal, M. M. Disko, R. D. Sherwood, K. M. Creegan and D. M. Cox, Phys. Rev. B, 1992,45, 11366. A. Dworkin, H. Szwarc, S. Leach, J. P. Hare, T. J. Dennis, H. W. should occur in the oxide adduct. Kroto, R. Taylor and D. R. M. Walton, C.R. Acad. Sci. Paris, The determination of the oxygen stoichiometry by crystal- lographic methods in one of our crystals is in good agree- ment with the independent determination of the maximum oxygen uptake from gas volumetric experiments26 arriving at a limiting composition of C6,02 .The molecular character of 20 21 22 1991,312,979. H. Yang, P. Zheng, Zh. Chen, P. He, Y. Xu, Ch. Yu and W. Li, Solid State Commun., 1993,89, 735. R. Tycko, G. Dabbagh, R. M. Fleming, R. C. Haddon, A. V. Makhija and S. M. Zahurak, Phys. Rev. Lett., 1991,67, 1886. H. Kasatani, H. Terauchi, Y. Hamanaka and S. Nahashima, the oxygen present at 300 K in the fullerene crystal has been shown unambiguously by isotope labelling desorption and by electron spectro~copy.~~The adsorption site within the crystal is most likely an octahedral site, as deduced from NMR experiments in the literat~re.~’ A localisation of the end-on bonded oxygen by diffraction is not possible due to 23 24 25 Phys.Rev. B, 1993,47,4022. H. Werner, J. Blocker, U. Gobel, B. Henschke, W. Bensch and R. Schlogl, Angew. Chem., 1992,31,868. H. Werner, M. Wohlers, D. Bublak, J. Blocker and R. Schlogl, Fullerene Sci. Technol., 1993, 1,457. H. Werner, Th. Schedel-Niedrig, M. Wohlers, D. Herein, B. Herzog, M. Keil, R. Schlogl, J. Kirscher and A. M. Bradshaw, J. the rotational disorder of the substrate and tumbling motion of the adsorbate at 300 K. The electronic interaction between 26 Chem. SOC., Faraday Trans., 1994,90,403. K. Kaneko, C. Ishii, T. Arai and H. Suematsu, J. Phys. Chem., oxygen and the fullerene mentioned above follows clearly from the ob~ervation~~ of an infrared-active 0-0 stretching frequency which is significantly lowered in energy relative to the Raman band in free molecular oxygen.The determination of the structure of chemisorbed molecu- 27 28 29 1993,97,6764. G. A. Samara, L. V. Hansen, R. A. Assink, B. Morosin, J. E. Schirber and D. Loy, Phys. Rev. B, 1993,47,4756. G. van Tendeloo, S. Amelinckx, M. A. Verheijen, P. H. M. van Loosdrecht and G. Meijer, Phys. Rev. Lett., 1992,69, 1065. A. B. Harris and R. Sachidananadam, Phys. Rev. Lett., 1993,70, lar oxygen in a bonding state prior to the formation of a covalent organoperoxide on the (inner) surface of a pure sp2 carbon material offers the chance to contribute to the under- standing of oxygen activation in the reaction sequence of carbon oxidation, a process which is of fundamental rele- 30 31 32 102. W. Bensch, H. Werner and R.Schlogl, Proc. Int. Carbon Conf., Essen, German Carbon Group, Frankfurt, 1992, p. 492. S. Liu, Y. J. Lu, M. M. Kappes and J. A. Ibers, Science, 1991, 254,408. K. Kamaras, L. Akselrod, S. Roth, A. Mittelbach, W. Honle and vance but yet only poorly understood in its mechanistic details.40 33 H. G. von Schnering, Chem. Phys. Lett., 1993,214,338. D. E. Luzzi, J. E. Fischer, X. Q. Wang, D. A. Ricketts-Foot, A. R. McGhie and W. J. Romanow, J. Muter. Res., 1992,7, 335. Crystallographic problems were discussed with E. Egert, J. Bats and E. Paulus. We thank E. Egert for access to the CAD 4 instrument. Financial support came from the Bundes- minister fur Forschung und Technologie, the Fonds der Che- mischen Industrie and from a special grant from Du Pont. 34 35 36 K.Hedberg, L. Hedberg, D. S. Bethune, C. A. Brown, H. C. Dorn, R. D. Johnson and M. de Vries, Science, 1991,254,410. R. Blinc, J. Seliger, J. Dolinsek and D. Arcon, Europhys. Lett., 1993, 23, 355. D. A. Neumann, J. R. D. Copley, R. L. Capelletti, W. A. Kamita- kahara, R. M. Lindstrom, K. M. Creegan, D. M. Cox, W. J. Romanow, N. Coustel, J. P. McCauley Jr., N. C. Maliszewskyj, References 1 C. S. Yannoni, R. D. Johnson, G. Meijer, D. S. Bethune and J. R. Salem, J. Phys. Chem., 1991,95,9. 2 R. Tycko, R. C. Haddon, G. Dabbagh, S. J. Glarum, D. C. Douglass and A. M. Mujsce, J. Phys. Chem., 1991,95,518. 3 P. A. Heiney, J. E. Fischer, A. R. McGhie, W. J. Romanow, A. M. Denenstein, J. P. McCauley Jr. and A. B. Smith 111, Phys. Rev. Lett., 1991,66, 291 1. 4 H. B. Burgi, E. Blanc, D. Schwarzenbach, S. Liu, Y-J. Lu, M. M. Kappes and J. A. Ibers, Angew. Chem., 1992,104,667. 37 38 39 40 J. E. Fischer and A. B. Smith 111, Phys. Rev. Lett., 1991,67, 3808. P. C. Chow, X. Jiang, G. Reiter, P. Wochner, S. C. Moss, J. D. Axe, J. C. Hanson, R.K. McMullan, R. L. Meng and C. W. Chu, Phys. Rev. Lett., 1992, 69, 2943. G. B. M. Vaughan, P. A. Heiney, D. E. Cox, A. R. McGhie, D. R. Jones, R. M. Strongin, M. A. Cichy and A. B. Smith 111, Chem. Phys., 1992,168, 185. R. K. Kremer, T. Rabenau, W. K. Maser, M. Kaiser, A. Simon, M. Haluska and H. Kuzmany, Appl. Phys. A, 1993,56,211. F. Atamny, J. Blocker, A. Dubotzky, H. Kurt, G. Loose, W. Mahdi and R. Schlogl, J. Mol. Phys., 1992,76, 851. 5 W. I. F. David, R. M. Ibberson, T. J. S. Dennis, J. P. Hare and K. Prassides, Europhys. Lett., 1992, 18, 219. Paper 41024426;Received 25th April, 1994

ISSN:0956-5000    

DOI:10.1039/FT9949002791

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

J. CHEM. SOC. FARADAY TRANS., 1994, 90(18), 2799-2802 High-resolution Electron Microscopy Studies of a Microporous Carbon produced by Arc-evaporation Peter J. F. Harris Chemical Crystallography Laboratory, University of Oxford, 9 Parks Road, Oxford, UK OX1 3PD Shik Chi Tsang, John B. Claridge and Malcolm L. H. Green Inorganic Chemistry Laboratory, University of Oxford, South Parks Road, Oxford, UK OX1 3QR The soot produced as a byproduct of fullerene synthesis by arc-evaporation consists of a microporous carbon with a surface area, after activation with carbon dioxide, of ca. 700 m2 g-’. Here, we investigate the structure of this material, and its appearance after electron irradiation and high-temperature heat treatment, using high- resolution electron microscopy.We show that the heat treatment transforms the new carbon into a structure containing large, tube-like pores, rather than into polycrystalline graphite. This suggests that the arc-evaporated carbon may have a novel, fullerene-related microstructure, and that it may be the precursor for nanotube forma- tion. The discovery that C,, and other fullerenes could be extracted from the carbon soot produced by arc-evaporation of graphite rods’ has stimulated enormous interest. Natu- rally, most of this interest has focused on the fullerenes them- selves rather than on the residual soot which remains following fullerene removal. Nevertheless, there are good reasons to believe that this residual material also has a novel microstructure which may be of considerable theoretical and practical interest.In a recent study2 we investigated the properties of the soot following removal of fullerenes and ‘activation’ in carbon dioxide. We found that the material had a very high internal surface area (ca. 700 m2 g-’), and that it displayed molecular sieving properties indicative of a microporous structure with most of the pores 65 8, in diam- eter. High-resolution electron microscopy (HREM) confirmed that the structure was highly microporous, with an appear- ance quite similar to that of some conventional high-surface- area carbons. Here, we describe a more detailed programme of studies of the arc-evaporated soot using HREM. We compare the properties of the soot with those of a high-surface-area carbon prepared by pyrolysis of Saran resin, a poly(viny1idene chloride)/poly(vinyl chloride) copolymer.In this way we aim to determine how the fullerene-related soot differs from conventional microporous carbon. Experimental Preparation of Arcevaporated Soot Arc-evaporation of graphite rods was carried out in the stan- dard way for C,, production. Thus, electrolytic grade rods were arced in 150 Torr helium using a dc voltage of 32 V and a current of 180-200 A. Under these conditions ca. 70% of the vaporised carbon deposited onto the walls of the evapo- ration vessel, while ca. 30% ‘distilled’ onto the cathodic rod, as shown in Scheme 1. The soot which formed on the walls of the vessel (Cl) was carefully collected and the soluble ful- lerenes were extracted with toluene for three days in a Soxhlet apparatus, after which the solvent was removed under reduced pressure.The residual insoluble soot, C3, was then dried and activated by treatment in a stream of carbon dioxide, with a flow rate of 20 ml min- at 850 “Cfor ca. 5 h. Under these conditions the reaction C + CO, -,2CO occurs and after the 5 h period a weight loss of ca. 15% was observed. The final product was a microporous carbon which we designate C4. Preparation of Saran Char The starting material was Saran 416 powder, a copolymer containing ca. 90% poly(viny1idene chloride) (PVDC) and 10% poly(viny1 chloride) (PVC), with a melting point of 445 K. 1 g of the powder was first heated to 438 K under nitrogen and held at this temperature for 68 h.The sample was then further heated to 933 K under nitrogen at a rate of 1 K min-’ and was held at this temperature for 12 h, before being cooled to room temperature under nitrogen. The apparent surface area of the resulting carbon measured by BET (N2) was 655 m2 g-’. High-temperature Heat Treatment The arc-evaporated carbon C4 and the Saran char were both subjected to high-temperature heat treatments under vacuum in a positive-hearth electron gun (4 kV, 0.4 A) for periods of ca. 4 h. Although an exact measure of temperature was not carbon vaporised from the graphite rod during the arcing processa /\ raw carbon soot deposited on the walls of the vessel (Cl) extracting the f u I I e renesb with toluene so Ive nt ..insoluble soot (C3)I activated with C02 at 850°C for5n 4microporous carbon (C4) \carbon deposited onto cathode rod containing fullerene-related structuresC(C2) a Arc provided by a dc supply at ca. 200 A and 32 V. *The extracted fullerenes (3-5%) are C,, , C,, and higher fractions. The fullerene-related structures include nanotubules (buckytubes) that are closed at both ends and some nanoparti- cles. Scheme 1 made, samples typically reach temperatures in the region of 2500-3500 K in this apparatus. High-resolution Electron Microscopy Specimens were prepared for electron microscopy by grinding them gently in a pestle and mortar and dusting the powder directly onto holey carbon films.The microscopes employed were a JEOL 2000FX instrument with a maximum acceler- ating voltage of 200 kV and a JEOL 4000FX instrument with a maximum accelerating voltage of 400 kV. Only thin regions which extended over a hole in the support film were imaged. In addition to imaging the samples, we carried out a series of electron irradiation experiments on both the C4 carbon and the Saran carbon. Ugarte has that elec- tron irradiation can transform nanotube-containing soot into quasi-spherical concentric fullerene particles (‘carbon onions’), and we have shown’ that C4 can also be trans- formed into onions in this way. The aim of the present experiments was to compare the behaviour of a conventional carbon during electron irradiation with that of the arc-evaporated carbon. The irradiation experiments involved removing the microscope’s condenser aperture and focusing the beam on a small area of sample for periods of up to 2 h.A range of accelerating voltages were used for the irradiation, from 200 to 400 kV. Results Examination of Freshly Prepared Carbons Plate 1 shows a typical high-resolution micrograph of the microporous carbon C4. It can be seen that the structure consists mainly of randomly curved layers enclosing micro- pores in the sub-nanometre range. In some areas, spiral arrangements of layers were present, as indicated in Plate 1, but the interlayer spacings in these spiral structures were ca. 5-6 A, i.e. considerably larger than the graphite c distance of 3.4 A.Small graphitic regions were sometimes seen, and occasionally nanoparticles similar to those observed in the cathodic soot were observed,’ but the great majority of the material displayed no obvious crystalline order. The carbon prepared from Saran was somewhat similar in appearance to the C4 carbon, as shown in Plate 2, although the carbon layers appeared to be somewhat more tightly packed, and no obvious spiral structures were seen. Electron Irradiation When the C4 carbon was irradiated with an intense electron beam in the 2000FX instrument, using an accelerating voltage of 200 kV, a transformation to quasi-spherical con- centric fullerene particles was observed after a period of 30-40 min. Surprisingly, irradiation at the same accelerating voltage in the 4000FX instrument did not produce a similar transformation even after periods of greater than 1 h.Irradia- tion in this instrument at higher voltages (300 and 400 kV) similarly failed to produce a complete transformation into onions, although a rounding of the overall morphology was observed. This difference between the behaviour of C4 under irradiation in the two microscopes is not yet understood, but two possible reasons suggest themselves. First, the beam current in the 4000FX was probably lower than in the 2000FX. A direct measure of beam current was not possible because the intensities employed were higher than the maximum that could be recorded by the microscopes. Also, the smallest spot size obtainable on the 4000FX appeared somewhat larger than that obtained on the other microscope.In either case, these experiments indicate that beam current J. CHEM.SOC. FARADAY TRANS., 1994, VOL. 90 may be more important than accelerating voltage in inducing onion formation. A high-resolution image of a region of C4 which has been completely converted into spheroidal onion particles is shown in Plate 3. It can be seen that the range of particle sizes is rather narrow: the smallest onion observed consisted of three concentric layers (approximate outer diameter 21 A) and the largest of 12 layers (approximate outer diameter 82 A). Ugarte observed a much broader range of sizes in irradi- ated nanotubes/nanoparticles. In general, the onions appeared to be approximately spheroidal, although some- times they were observed to pass through a faceted configu- ration.Plate 4(a) shows a nine-shell particle with a circular outline, while Plate 4(b)shows the same particle a short time later exhibiting slight faceting. This phenomenon of momen tary faceting was also reported by Ugarte? As pointed out by a number of workers, it is surprising that more faceting is not observed. If the onions are assumed to consist of concentric fullerenes, then the giant fullerenes which make up the outer shells would be expected to experience considerable strain when forced into an approximately spherical shape More work is needed to understand why the onions are so remark-ably spherical. Electron irradiation of the Saran carbon was also carried out.Evolution into carbon onions also occurred here, although considerably longer irradiation times were needed than for the C4 carbon (this probably explains why onion formation was not observed in our previous study’ when we irradiated a microporous carbon derived from coconut shell). Plate 5 shows a Saran sample following irradiation for ca. 100 min at 200 kV in the 2000FX instrument. Again, the irra- diated region appears to have been almost completely con- verted into spheroidal onion particles, and there is a slightly wider range of particle sizes than for the C4 carbon. Note that this is the first demonstration of onion formation by the irradiation of a conventional (as opposed to arc-evaporated) carbon.Heat-treated Carbons High-temperature heat treatment of the arc-evaporated carbon C4 produced a structure apparently made up of large pores which were often extended in shape, resembling single- layer nanotubes, as can be seen in the micrograph shown in Plate 6. Like nanotubes, the extended pores were almost invariably closed, and exhibited a variety of capping mor- phologies. In some cases features were observed which are thought to be indicative of the presence of seven-membered carbon rings;6*’ an example is arrowed in Plate 6. In most cases the extended pores were bounded by single carbon layers, although multilayer structures were also present, as shown in Plate 7. Other structures were also seen, which appeared to be random in shape rather than tube-like, often exhibiting quite sharp faceting.We note that de Heer and Ugarte’ have found that raw fullerene soot (Cl) can also be transformed into nanoparticles and nanotubes by high- temperature heat treatment. A similar heat treatment of the Saran char produced a quite different structure, as shown in Plate 8. Here, the micro- porosity of the original material has disappeared, and has been replaced by a polycrystalline structure made up of con- joined graphitic layer groups. This kind of structure is typical of ‘non-graphitising’ carbons following high-temperature heating.’ Discussion In this study we have attempted to distinguish between the microporous carbon produced by arc-evaporation, C4, and a J. CHEM.SOC. FARADAY TRANS., 1994, VOL. 90 Plate 1 High-resolution electron micrograph of the microporous carbon C4. Arrow shows area with apparently spiral structure. The scale bar in this and all other figures is 50 A. Plate 2 Micrograph of the Saran-derived microporous carbon Plate 3 Micrograph of C4 carbon following intense electron irradiation at 200 kV for ca. 30 min, showing region completely transformed into quasi-spherical concentric fullerene particles (carbon onions) P. J. F. Harris et al. (Facing p. 2800) J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 (a) (b1 Plate 4 Nine-shell carbon onion particle in irradiated C4 sample: (a)image showing circular profile, (b) image displaying slight faceting Plate 5 Carbon onions formed by electron irradiation of Saran-derived carbon Plate 6 Micrograph of C4 carbon heated to 2500-3000 K for 4 h using a positive-hearth electron gun, showing development of large tube-like pores.Arrow indicates morphological feature indicative of the presence of heptagonal rings.’ J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Plate 7 Multilayer tube-like structures in heat-treated C4 carbon Plate 8 Saran carbon following heat-treatment with electron gun, showing extensive development of graphitic structure J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 conventional microporous carbon derived from Saran polymer, by using HREM. We first examined both materials in their fresh state and found them to be very similar in appearance.We then carried out a series of electron irradia- tion experiments on both the C4 carbon and the Saran carbon and found that in both cases a transformation to con- centric fullerene particles was observed. This shows that the occurrence of such a transformation does not imply that the original carbon must have a fullerene-related microstructure, as we suggested in our earlier paper.2 However, note that the transformation to ‘onions’ was considerably easier for the C4 carbon than for the Saran carbon, suggesting that it has a more fluid structure. High-temperature heat treatments were then carried out on both materials, and here a very striking difference in behav- iour was observed. In the case of the Saran carbon, the heat treatment resulted in the formation of polygonised and bent layer groups similar to those observed many years ago in studies of ‘non-graphitising’ carbons.’ However, high-temperature heat treatment of the C4 carbon resulted in a quite different structure apparently made up of large pores, often with extended, tube-like shapes.These tube-like pores were invariably capped, indicating the presence of pentagonal rings, and, as noted above, often displayed morphological features which have been interpreted as resulting from hep- tagonal We therefore believe that the original struc- ture of C4 contains both pentagons and heptagons distributed randomly throughout a hexagonal network, producing continuous curvature, as shown in Fig. 1. This structure resembles the ‘random schwartzite’ structure pro- posed by Townsend et d.,”although with many fewer seven- membered rings.Also, we believe that C4 consists of relatively small fragments, rather than the extensive structure envisaged by Townsend et a/. Other workers have put forward similar models of ‘fullerene soot’ to the one we have proposed,’ 1*12 although they have tended to emphasise the importance of spiral structures in the soot. There is some evi- dence that spiral structures are present in C4 (see Plate l), but the majority of the material does not seem to have this form. We believe that it is quite feasible that a structure such as the one shown in Fig. 1 could form in a carbon arc. Accord- ing to the ‘pentagon road’ theory of fullerene a~sembly,’~ the formation of isolated pentagonal rings is energetically favoured during condensation as this leads to the mini- misation of dangling bonds.In C,,, 12 isolated five-membered rings are distributed symmetrically to produce the closed, icosahedral structure. However, if pentagons were to occur in the ‘wrong’ positions then closure would be much a Fig. 1 Illustration of our proposed structure for C4 carbon. The fragment shown has 45 hexagons, four pentagons and one heptagon. less likely, and a randomly curved structure might develop. Kroto and McKay considered a similar growth mecha- nism,l4 but assumed that pentagons would stack epitaxially one above the other, leading to spiral shell structures. As noted above, the structure of C4 does not appear to be domi- nated by spiral structures.The observation that C4 can be transformed into nanotube-like structures by high-temperature heat treatment suggests that it may be the precursor for the production of carbon nanotubes. One could therefore propose the following model for nanotube formation. In the initial stages of arc- evaporation, a C4-like material (plus fullerenes) would con- dense onto the cathode, and the condensed material would then experience extremely high temperatures as the arcing process continued, resulting in the formation first of single- layer, nanotube-like structures and then of multilayer nano- tubes. This contrasts with previous models of nanotube formation, which have envisaged a direct vapour-condensation process.One could also speculate that the spiral structures observed in C4 might be the precursors for the nanoparticles, which are also invariably present in cathodic soot.’ Recent work by Lin et a/.” has produced very interesting results which appear to confirm our model of nanotube growth. They carried out arc-evaporation employing a com- posite anode, the centre of which consisted of a copper rod. The central region of the resulting cathodic deposit was found to contain many single-layer structures similar to those reported here, while the outer region consisted largely of multilayer tubes. They speculate that this could be a result of a change in the temperature distribution on the electrode sur- faces due to the presence of copper in the anode.This would be consistent with the model of nanotube growth presented here, since in their experiment the central region of the cathodic deposit would presumably experience a lower tem- perature, so that the C4 material would be transformed only to single-layer structures, while the outer region would expe- rience temperatures similar to those produced in a normal arc, resulting in multilayer structures. A comment should be made here on the reactivity mea- surements described in our previous paper.2 We reported there that the carbon C4 was less reactive towards oxidation with carbon dioxide than conventional microporous carbon. We now believe that these measurements may be misleading owing to the presence of potassium in the conventional carbons, leading to enhanced reactivity.Experiments are in progress to establish the true reactivity of C4 in comparison with conventional microporous carbon. Finally we note that the new arc-evaporated microporous carbon may have many novel and useful characteristics in addition to those we have already described. Thus its elec- tronic and optical properties, for example, may be worthy of further study. We are grateful for the use of a JEOL 4000FX transmission electron microscope in the Department of Materials, Uni- versity of Oxford. We also thank C. Menhart for assistance with the high-temperature heat treatments, and R. Wence of the Dow Chemical Company for the supply of Saran resin. Funding was provided by the Gas Research Institute.References 1 W. Kratschmer, L. D. Lamb, K. Fostiropoulos and D. R. Huffman, Nature (London), 1990,347,354. 2 S. C. Tsang, P. J. F. Harris, J. B. Claridge and M. L. H. Green, J. Chem. SOC.,Chem. Commun., 1993,1519. 3 D. Ugarte, Nature (London), 1992,359,707. 2802 J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 4 5 6 7 8 9 10 D. Ugarte, Europhys. Lett., 1993, 22,45. P. J. F. Harris, M. L. H. Green and S. C. Tsang, J. Chem. Soc., Faraday Trans., 1993,89, 1189. S. Iijima, T. Ichihashi and Y. Ando, Nature (London), 1992, 356, 776. S. Iijima, Mater. Sci. Engng. B, 1993,19, 172. W. A. de Heer and D. Ugarte, Chem. Fhys. Lett., 1993,207,480. L. L. Ban, in Surj..e and Defect Properties of Solids, ed. M. W. Roberts and J. M. Thomas, Specialist Periodical Reports, Royal Society of Chemistry, London, 1972, vol. 1, p. 54. S. J. Townsend, T. J. Lenosky, D. A. Muller, C. S. Nichols and V. Elser, Phys. Rev. Lett., 1992,69, 921. 11 12 13 14 15 H. Werner, D. Herein, J. Blocker, B. Henschke, U. Tegtmeyer, Th.Schedel-Niedrig, M. Keil, A. M. Bradshaw and R. Schlogl, Chem. Phys. Lett., 1992, 194, 62. J. C. Scanlon and L. B. Ebert, J. Phys. Chem., 1993,97,7138. R.E. Smalley, Acc. Chem. Res., 1992,25,98. H. W. Kroto and K. McKay, Nature (London), 1988,331,328. X. Lin, X. K. Wang, V. P. Dravid, R. P. H. Chang and J. B. Ketterson, Appl. Phys. Lett., 1994,64, 181. Paper 4/02878C; Received 16th May, 1994

ISSN:0956-5000    

DOI:10.1039/FT9949002799

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

J. CHEM. SOC. FARADAY TRANS., 1994, 90(18), 2803-2807 Characterization of Carbon-supported Ruthenium-Tin Catalysts by High-resolution Electron Microscopy G. Neri and R. Pietropaolo Faculty of Engineering, University of Reggio Calabria, 89100 Reggio Calabria, Italy S. Galvagno and C. Milone Department of Industrial Chemistry, University of Messina, 98166 Messina, Italy J. Schwank Department of Chemical Engineering, University of Michigan, Ann Arbor, MI 48109, USA Ruthenium and bimetallic ruthenium-tin catalysts supported on high-surface-area carbon are characterized by high-resolution electron microscopy (HREM) and CO chemisorption. The characterization results give important insights into the reasons for the remarkable catalytic properties of Ru-Sn bimetallic catalysts in selective hydro- genation reactions of a,/.?-unsaturated aldehydes.In the monometallic Ru/C sample, ruthenium is present as small particles, showing no variation in the crystalline structure and lattice parameters with respect to the bulk state. At low Sn : Ru ratios, the average metal particle size is smaller than for monometallic Ru/C, whereas at higher Sn : Ru ratios Ru particles agglomerate on the surface of the support. According to HREM observations, most metal particles in the bimetallic catalysts possess a structure consistent with that of elemental ruthenium. CO chemisorption results indicate that the surface of the ruthenium particles is partially, covered by tin species. Only in the samples with high tin loading are a few Ru-Sn alloy particles found, in addition to a majority of particles having the structure of pure ruthenium.The high selectivity of Ru-Sn/C catalysts for selective hydro- genation of the C-0 group in a,/.?-unsaturated aldehydes can thus be attributed to the presence of small ruthe-nium particles whose surface is covered with tin species which are most likely in ionic form. RuSn alloy particles do not appear to make a significant contribution to the catalytic activity. Bimetallic catalysts are commonly used in many important processes in the petroleum industry. They also have potential applications in the synthesis of fine chemicals for the per- fumery and food industries. In our laboratory bimetallic systems containing platinum and one element of Group 14, such as Ge and Sn, have been investigated.These catalysts have remarkably high selectivity in the hydrogenation of a,/.?-unsaturated aldehydes to the corresponding unsaturated alcohols.’** Ruthenium is an attractive catalyst for the hydrogenation of carbonyl compounds. Moreover, the addition of tin to monometallic catalyst has been found to affect favourably the catalytic activity and/or selectivity towards the desired prod- ucts. Basset et aL3 reported that Ru-Sn supported on SiO, is a very active and selective system for the hydrogenation of ethyl acetate to ethanol under mild conditions. Ru-Sn cata-lysts, promoted with boron, have been shown to catalyse the selective hydrogenation of esters and unsaturated aldehydes to alcoh01.~ The effect of tin has been attributed to modifi- cations of the structural and/or electronic properties of ruth- enium either by dilution of the active surface atom ensembles or by formation of an alloy with the second metal com- ponent.Similar results obtained in our laboratory on Ru-Sn sup-ported on high-surface-area carbon confirm that addition of tin to ruthenium catalyst strongly modifies the performance of the noble meta1.5*6 In the selective hydrogenation of a,/.?-unsaturated aldehydes addition of tin to ruthenium changes the product distribution leading to a selectivity higher than 90% towards the formation of unsaturated alcohol.6 It also causes, at low tin loading, an increase of the catalytic activity. Higher tin loadings poison the active centres and the cata- lytic activity decreases.The chemical state and location of tin relative to ruthenium in bimetallic Ru-Sn catalysts is still an open question. Some authors have suggested that Sn is present as ionic tin; however, alloy formation has been detected by XRD in Ru-Sn catalysts promoted with b~ron.~,~,’ The objective of this work is to use HREM to obtain microstructural information on the Ru-Sn/carbon catalysts previously used for selective hydrogenationSv6 and to find a relationship between the structure of these bimetallic cata- lysts and the catalytic properties. Direct imaging of lattice planes by HREM has already proved to be a versatile tech- nique to obtain information on the crystallographic structure of small metal particles8 It should be possible to determine where the tin is located in these catalysts and to what extent Ru-Sn alloy particles are present.Experimental Catalyst samples were prepared by using the incipient wetness technique. Carbon (Chemviron SCXII, 80-100 mesh, surface area 900-1100 m2 g-’) was co-impregnated with aqueous solutions of RuCl, and Sndl, having the appropri- ate metal concentration. The catalysts were dried at 393 K in air for 1 h followed by reduction under flowing H, at 673 K for 2 h. They were then stored in air and reduced in situ under very mild conditions, 343 K and 0.1 MPa H, for 1 h, before catalytic tesk5v6 The ruthenium loading was held con- stant at 2 wt.% in all catalysts, whereas the tin loading was varied from 0 to 1.56 wt.%.The numerical value in the cata- lyst code, reported in Table 1, indicates the nominal atom% of Ru on the basis of the nominal metal loading; e.g. catalyst code Ru95/C refers to a sample with 95 atom% Ru and 5 atom% Sn. Chemisorption of CO was measured in a conventional pulse system operating at room temperature. The catalyst sample (0.1-0.5g) was reduced in flowing hydrogen (at 673 K for 2 h) followed by flushing for 3 h in a helium stream at 673 K with subsequent cooling in flowing helium to room tem- perature. Calibrated CO pulses were injected by means of a sample loop into the helium carrier gas and detected by a thermal conductivity detector. A mixture of 10 vol.% CO in He was used to inject a small amount of CO.By comparing the amount of CO reaching the detector and the amount of CO injected into the system, the quantity of CO adsorbed on Table 1 Composition and characterization of Ru/C and Ru-Sn/C samples catalyst code Ru (wt.%) Sn (wt."/,) CO :Ru" d/nmb Ru100/C 2.0 -0.261 5.9 (5.3)' Ru95/C 2.0 0.12 0.220 3.5 Ru90/C 2.0 0.26 0.115 3.5 Ru80/C 2.0 0.58 0.096 3.4 -dRu70/C 2.0 1.01 0.003 eRu60/C 2.0 1.56 --d aRatios given relative to one Ru atom. By TEM. 'By CO chemi- sorption. Not evaluated. Below the detection limits. the catalyst could be determined. Additional pulses were injected until no further CO uptake was noted. Blank experi- ments on the carbon support proved that there was no measurable uptake of CO on the support itself. Table 1 lists the various catalysts, their nominal loading, CO : Ru ratio, and average particle sizes obtained from transmission elec- tron microscopy. Under the experimental conditions used, Sn by itself does not adsorb measurable quantities of CO.Electron microscopy studies of the catalysts were per-formed on a JEOL 2000 FX instrument operating at 200 kV and a JEOL 4000 EX microscope operating at 400 kV. The latter instrument was equipped with a top entry state goni- ometer for high-resolution work and directly interfaced with a Gatan camera for real-time image processing. The catalyst specimens for electron microscopy were pre- pared by gently grinding the powder samples in an agate mortar, suspending and sonicating them in isopropyl alcohol, and placing a drop of the suspension on a holey carbon copper grid.After evaporation of the solvent, the specimen was introduced into the microscope column. During speci- men preparation, the samples were exposed to air. The top entry stage goniometer in the JEOL instrument is not well suited for interfacing with sample manipulators, and there is, at present, no commercial high-resolution electron' micro- scope available with in situ reduction capability. The lattice spacings of the metal particles were obtained by Fourier transformation of the lattice fringe image, utilizing a 40 (41 h S 30-W 0>.g 20-I-U 10. m0-. r d/nm 40 I 1 J. CHEM. SOC.FARADAY TRANS., 1994, VOL. 90 Table 2 Crystallographic data for Ru and Ru-Sn alloys lattice constant/A component crystal system a C ~~~~ ~ Ru Ru,Sn, hexagonalcubic 2.706 9.351 4.282 - Ru,Sn, RuSn, tetragonal tetragonal 6.172 6.389 9.915 5.693 solid-state camera (COHU) interfaced with a Macintosh computer for image calculation. The distances on the Fourier transform were calibrated using an Si( 111) specimen, with a (220) lattice spacing of 0.192 nm. Table 2 shows a listing of the crystal structure data for the metallic components which can exist in the Ru-Sn system. Results and Discussion Plate 1 shows a typical low-magnification transmission elec- tron micrograph of the monometallic Ru100/C catalyst. The ruthenium particles are clearly visible on the surface of the carbon support.The particle size distribution obtained by measuring the diameter of several hundred particles in micro- graphs of this sample is reported in Fig. 1. The surface mean diameter, d,, calculated from the equation: d, = 1,n,d:/1,n,d? was 5.9 nm, where n,represents the number of particles of a given diameter, d,. Chemisorption of CO has been used to obtain complementary average particle size data. The average Ru particle size, d,, was determined from the CO uptake by the expression d, = 6V/S, where V is the ruthenium metal volume and S the surface area of Ru. The metal surface area was measured assuming a ruthenium surface density of 1.63 x1019 atoms m-2 and a stoichiom- etry CO : Ru = 1 : 1.' The agreement between the average Ru particle size calculated from chemisorption and the Ru particle size determined by transmission electron microscopy (Table 1) is satisfactory within the accuracy of both methods. The same Ru100/C sample was also examined in a high- resolution electron microscope operating at 400 keV where the lattice structure of the metal can be resolved.An HREM image of a ruthenium particle of about 3-4 nm diameter is 50> 1 40-s v 5 30-C -12345678910 d/nm "1234567910 d/nrn d/nrn Fig. 1 Particle size distributions of Ru 1OO/C and Ru-Sn/C samples:(a)Ru100/C; (b)Ru95/C; (c) Ru90/C and (d)Ru80/C J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Plate 1 Micrograph of the monometallic Ru100/C Plate 2 HREM image of Ru100/C showing the [1213] orientation of a ruthenium particle of about 3-4 nm diameter 1 nm Plate 3 HREM image of Ru100/C showing a 2 nm ruthenium particle having a hexagonal shape oriented with the [OOOl] plane parallel to the electron beam G.Neri et al. (Facing p. 2804) J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 10 nm 10 nm Plate 4 TEMs from Ru-Sn bimetallic samples: (a)Ru90/C; (b)Ru70/C 20 rim Plate 5 Higher-magnification TEM of the Ru70/C sample. The agglomerate shown is composed of Ru particles of about 3 nm in diameter. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 10 nm Plate 6 Thin section of Ru60/C showing small particles -2nm Plate 7 High-resolution micrograph of a metal particle of the Ru60/C sample.The Fourier-transform patterns indicate an fcc structure with d spacings of 0.221 and 0.211 nm, respectively, corresponding to an Ru,Sn, alloy. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 shown in Plate 2. The Ru(10i0) and Ru(li01) planes with spacings of 0.234 and 0.205 nm, respectively, were imaged consistent with the lattice fringes of the hcp Ru structure. Very small metal particles in close contact with a support can show deviations of the lattice parameter from the bulk values for the metal. For example, small palladium particles supported on titania have been reported to show a lattice expansion in the (111) planes of palladium.” For small platinum particles supported on alumina the single-crystal structure was found to be preserved for particles as small as 1 nm,” while other authors” reported a decrease in the lattice parameter for platinum proportional to the reciprocal of the platinum particle size.The latter result was attributed to surface stress effects caused by the high surface : volume ratio of small particles.” On Pt/C and Pt/SiO, we have recently observed no deviation from the bulk fcc value in the lattice spacing of small particle^.'^*'^ On Pt/SiO, only a few par- ticles of about 1 nm diameter or less showed a change in the lattice parameter. Literature data on lattice parameters of supported ruth- enium are very scarce. Datye et a/.’’ have shown that Ru supported on MgO preferred low-index planes with the {OOOl}, { lOi0) and { 101l} planes of Ru being exposed.Their results show no significant variation from the hexagonal structure characteristic of the bulk ruthenium. Our observa- tions here are in agreement, as the analysis by Fourier trans- formation of lattice images of particles in Ru100/C showed no deviation from the lattice parameter of the bulk structure within the experimental error associated with the measure- ment of the d spacings,16 even for metal particles as small as about 1 nm diameter. This indicates that most of these par- ticles were single crystals. Numerous well faceted metal par- ticles of hexagonal shape are present on the external surface of the carbon support. We observed such a clear particle faceting only in the monometallic Ru sample.Plate 3 showed one Ru particle of about 2 nm oriented with the electron beam parallel to the [OOOl] direction. In the bimetallic cata- lysts containing Sn, the small metal particles were generally spherical, rather than faceted. Anno and Hoshino have pre- dicted that the equilibrium shape of small supported particles should be dependent on size.17 Below 2 nm the metal par- ticles would be spherical, whereas above this size they would be flat and faceted. In agreement with the above findings, Buglass et d.18 have reported faceted particles in ruthenium crystallites larger than 2 nm on their Ru/C samples. Transmission electron micrographs (TEM) of some bimetallic samples are presented in Plate 4. Catalysts with low tin content (< 20 atom%) exhibit metal particles smaller than those observed on the monometallic sample (Fig.1). When the tin loading increased (samples Ru70/C and Ru60/C) a massive agglomeration of smaller particles was observed. These agglomerates consisted of many small crys- tallites clustered together to form larger aggregates distrib- uted randomly on the carbon surface (Plates 4 and 5). These large agglomerates are predominantly located at the external surface of the carbon. In addition, there are numerous small crystallites clearly visible through thin sections of the carbon support (Plate 6). These smaller crystallites are probably located inside the small pores of the carbon support, and therefore, their growth appears to be limited by the diameter particles in the bimetallic catalysts with low tin content (<20 atom%) revealed on all particles examined a hexagonal struc- ture characteristic of pure, elemental ruthenium.This raises the question of where the tin is located. One possibility is that tin species, either in zero-valent or oxidized form, deco- rate the surface of these ruthenium particles. Such sub- monolayer or monolayer quantities of tin would not be visible in HREM images. Furthermore, it is very likely that the exposure of the specimen to air prior to examination in the microscope would have caused surface oxidation of the tin species, if present. An oxidized tin surface layer would not give rise to sufficient diffraction contrast to become vi$ble in the micrograph.Therefore, HREMs have to be interpreted in conjunction with surface-sensitive probes, such as chemisorp- tion and catalytic probe reactions. To test the hypothesis of Sn species covering the surface of Ru particles, the CO :Ru ratio was used to measure the Ru dispersion (Ru atoms on the surface :total Ru atoms). From the metal particle size dis- tribution histogram shown in Fig. 1 it is apparent that the average particle sizes of the bimetallic catalysts are smaller than that of the monometallic RulOO catalyst. If the metal particles in the bimetallic Ru-Sn/C catalysts consisted of pure ruthenium without any tin on the surface, there would be a larger number of ruthenium surface sites available for CO chemisorption. Therefore, one would expect an increased uptake of CO, compared with the Ru100/C sample.Experi- mentally, the opposite trend is observed, and the CO :Ru ratio on the Ru-Sn/C samples decreased with increasing tin content (Table 1). This suggests that tin, which does not adsorb CO under our experimental conditions, blocks some of the active ruthenium surface sites. Fig. 2 shows the value of Ru dispersion in the bimetallic catalysts normalized to the Ru dispersion of the Ru100/C sample (relative dispersion), and plotted as a function of the tin loading. Relative disper- sion values were calculated by dividing the CO :Ru uptake ratio determined from chemisorption experiments on the bimetallic samples by the CO : Ru uptake value obtained for the RulOO sample.The experimental points fit well the nor- malized Ru dispersion which would correspond to a mono- layer growth of tin on the surface of Ru, as indicated by the dashed line in this figure. This is in agreement with an inter- action between Ru and Sn possibly occurring through the formation of Ru-Sn bimetallic surface aggregates or surface alloys not extending beyond the first few atomic layers of the particles,” or it could be due to the presence of ionic tin species, most likely in the form of tin oxides. Even if the particle surface should be reconstructed as a consequence of air exposure of the reduced sample, HREM can give valuable information on the structure of the particle core. The finding that the metal particles in the Ru-Sn samples with low tin loading had a structure consistent with \ g 0.8 Q \ .-2 \ \0.6 .-\ \U p 0.4 o\ .-\nc \-of the pores.For these catalysts showing large agglomer- ?!0.2 ations, it is not possible to derive a reliable particle size dis- tribution. On Pt-Sn catalysts, dispersed on the same carbon support, it has recently been shown that addition of tin causes increases of the average particle size, and at the highest tin loading large particles with raft-like structures were 0bser~ed.l~ Fourier transformation of the image structure of individual Ob 10 20 \%I To 20 100 Sn/( Ru + Sn) Fig. 2 Normalized Ru dispersion as a function of the Sn : (Sn + Ru) ratio: (---) calculated Ru dispersion which would correspond to a monolayer growth of tin on the surface of ruthenium that of pure Ru is very meaningful.It rules out the formation of bulk alloys in the catalysts with low tin loading. A different picture emerged for the catalysts with high tin loading, Ru70/C and Ru60/C. Most metal particles in these catalysts still possessed the structure of pure ruthenium. However, a few particles were found with the structure of ruthenium-tin alloy. There was no evidence for the presence of separate, pure tin particles or tin oxide particles on the carbon support. Plate 7 shows one particle of about 20 nm on sample Ru60/C which does not show the hexagonal struc- ture of pure ruthenium. The d spacings, evaluated from the Fourier transfomation of the structure image, are 0.221 and 0.211 nm, respectively, very close to the values of the (330) and (420) planes of the fcc Ru3Sn7 alloy.20 The bulk phase diagram of the Ru-Sn system shows the existence of three different alloy phases.Crystallographic data of these alloys are given in Table 2. RuSn, and Ru,Sn3 have a tetragonal structure whereas Ru3Sn7 is fee.,' The presence of RuSn, tetragonal alloy cannot be ruled out because the similarity in the lattice spacing with Ru3Sn7 makes it very difficult to dif- ferentiate the structure image of these two alloys.22 Despite air exposure during specimen preparation for microscopy, it was possible to preserve Ru-Sn alloy structures with tin in its zero-valent form. This is in agreement with our prior HREM observation of tin alloys in Pt-Sn/C13 and Pt-Sn/SiO, cata-lyst~.~~Of course, this does not preclude that the surface layer of these alloy particles may be oxidized and/or recon- structed.Our findings are in agreement with those of Desh- pande et aL7 who identified by X-ray diffraction (XRD) the Ru3Sn7 alloy in Ru-Sn-B unsupported catalysts. On alumina-supported catalysts they were unable to identify any alloy formation, probably owing to the very low amount of alloy, and to limitations of the XRD technique. It is unlikely that the few Ru3Sn7 bulk alloy particles observed here are responsible for the change in activity and selectivity. The majority of the particles in the samples with high tin loading had a core of pure ruthenium, and CO chemisorption gave again evidence that the ruthenium surface is partially blocked by tin species.The catalytic behaviour of the Ru/C and Ru-Sn/C cata-lysts towards the hydrogenation of C-C and C-0 groups has previously been tested in the reduction of 8-methyls tyrene, hydrocinnamaldehyde and cinnam-aldeh~de.~.~It has been observed that on addition of tin, the activity (expressed per g of Ru) towards the hydrogenation of the C=C double bond decreased, whereas the rate of hydro- genation of the C=O group went through a maximum at intermediate Sn :Ru ratios. Addition of tin in the hydro- genation of cinnamaldehyde increased the selectivity towards the formation of the unsaturated alcohol. On the basis of the characterization results reported in this paper, the catalytic behaviour of the Ru-Sn catalysts for hydrogenation of unsaturated aldehydes can be explained.Hydrogenation of the C-C double bond depends on the availability of ruthenium surface sites which are responsible for H, activation. Addition of tin blocks the ruthenium surface, thereby decreasing the number of ruthenium active sites and lowering the C-C double-bond hydrogenation activity. However, tin species on the ruthenium surface, when present in ionic form, would enhance the reactivity of the C=O group. Ionic tin can increase the polarization of the C=O bond, making it more reactive towards the attack of a weak nucleophilic agent such as the hydrogen chemisorbed on the nearby Ru atoms.This model of ionic tin decorating the surface of ruthenium particles is consistent with the observed activity and selectivity trend^.^,^ However, we cannot entirely rule out that, after reduction, zero-valent tin species are present on the surface of ruthenium particles, J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 perhaps forming surface alloy or bimetallic Ru-Sn surface aggregates. Such submonolayer tin concentrations would go unnoticed in HREM images, and it would probably be impossible to maintain these tin species in a reduced state during specimen preparation for microscopy. The presence of tin surface species can alter the adsorption characteristics of the ruthenium surface either through an electronic effect (ligand effect) and/or through a dilution effect, which results in a modification of the geometry of the active sites needed for the reaction (ensemble effect).Conclusions The microstructural HREM characterization data of carbon- supported Ru and Ru-Sn catalysts, and the particle morph- ologies and size distributions measured by transmission electron microscopy, greatly facilitate the interpretation of activity and selectivity trends reported for selective hydro- genation of a,/?-unsaturated aldehyde^.'.^ HREM proves that in the monometallic Ru/C sample nanometre-scale ruthenium crystallites are present showing no deviation in crystal struc- ture and lattice parameters with respect to bulk ruthenium. In the bimetallic Ru-Sn/C catalysts, at low tin loading (<20 atom%), the majority of the metal particles have a structure consistent with that of pure ruthenium, without any evidence for bulk alloy formation.In these samples the metallic par- ticles have an average particle size smaller than that observed in the monometallic Ru/C sample. While electron microscopy gives information about the structure of the metal particles, it is not surface sensitive and cannot differentiate whether or not tin species decorate the surface of these ruthenium particles. CO chemisorption data complement the microscopy data and provide information about the number of Ru surface sites. From the decrease in CO uptake despite a decrease in particle size, we can con- clude that tin must be decorating the surface of these Ru par- ticles, partially blocking the Ru sites needed for CO chemisorption.This conclusion is consistent with the observed decrease in C=C bond hydrogenation activity with increased tin loading, as it is known that Ru sites are required for this reaction. From the enhanced selectivity on the Ru-Sn catalysts for hydrogenation of the C=O group of unsaturated aldehydes, we can infer that under reaction con- ditions a significant fraction of the tin on the surface must be in ionic form. In the catalysts with high tin loading (>20 atom%), HREM reveals that only a few metal particles are present in the form of ruthenium-tin alloys, while the structure of most of the metal particles still corresponds to that of pure ruth- enium.Similar to the case of the low-loading catalysts, CO chemisorption indicates that tin species partially block the ruthenium surface. These catalysts have very low activity, indicating that the presence of alloy particles is not a domi- nant feature controlling the catalytic behaviour. The high selectivity for C-0 bond hydrogenation supports the hypothesis that the excess of tin (i.e.that not involved in alloy particles) is present in ionic form. Support by CNR for G.N. is gratefully acknowledged. We thank Dr. John Mansfield and Mr. Bryan Demczyk of the University of Michigan Electron Microbeam Analysis Labor- atory for valuable assistance in the microscopy work. References 1 Z. Poltarzewski, S. Galvagno, R. Pietroapolo and P. Staiti, J.Catal., 1986, 102, 190. 2 S. Galvagno, Z. Poltarzewski, A. Donato, G. Neri and R. Pietropaolo,J. Chem. SOC.,Chem. Commun., 1986, 1729. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 3 P. Louessard, J. P. Candy, J. P.Bourneville and J. M.Basset, in Structure and Reactivity of Surfaces, ed. C. Morterra, A. Zecchina and G. Costa, Elsevier, Amsterdam, 1989,p. 591. 4 V. M. Deshapande, K.Ramnrayan and C. S. Narasimhan, J. Catal., 1990,121, 174. 5 S. Galvagno, A. Donato, G. Neri and R. Pietropaolo, Cutal. Lett., 1991,8,9. S. Galvagno, A. Donato, G. Neri, G. Capannelli and R. Pietropaolo, J. Mol. Catal., 1993,78,237.V. M. Deshpande, W. R. Patterson and C. S. Narasimhan, J. Catal., 1991, 121, 165. D.J. Smith, D. White, T. Baird and J.R. Fryer, J. Catal., 1983, 81,107. 9 J. R. Anderson, in Structure of Metallic Catalysts, Academic Press, New York, 1975. 10 J. W. M. Jacobs and D. J. Schryvers,J. Catal., 1987,103,436. 11 D. White, T. Baird, J. R. Fryer, L. A. Freeman, D. J. Smith and M. Day, J. Catal., 1983,81, 119. 12 C. Solliand and M. Flueli, Surf: Sci., 1985, 156,487. 13 G. Neri, S.Galvagno and J. Schwank, J. Chem. Sw., Faruduy Trans., submitted. 14 A. Sachdev and J. Schwank, J. Catd., 1989,lu), 353. 15 A. K. Datye, A. D. Logan and N. J. Long, J. Catal., 1988,109, 76. 16 R. Sinclair and G.Thomas, Metall. Trans. A, 1978,9, 373. 17 E.Anno and R. Hoshino, Sur$ Sci., 1984,144,567. 18 J. G. Buglass, S. R. Tennison and G. M.Parkinson, Catal. Today, 1990,7,209. 19 X. Wu, B. C.Gerstein and T. S. King, J. Catul., 1990,123,43. 20 ASTM, X-Ray Powder Data File 26-504. 21 M.Hansen, in Constitution of Binary Alloy, McGraw-Hill, New York, 1958,p. 3257. 22 ASTM, X-Ray Powder Data File 18-1 143. 23 J. Schwank, K. Balakrishnan and A. Sachdev, in New Frontiers in Catalysis, Proc. ZOth Znt. Congress on Catalysis, ed. L. Guczi, F. Solymosi and P. Tetenyi, Elsevier, Amsterdam, p. 905, 1993. Paper 3/07648B;Receioed 31st December, 1993

ISSN:0956-5000    

DOI:10.1039/FT9949002803

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

J. CHEM. SOC. FARADAY TRANS., 1994, 90(18), 2809-2813 Ru-Cu/SiO, Catalysts :Characterization by FTlR Spectroscopy Carmelo Crisafulli, Rosario Maggiore and Salvatore Scire Dipartimento di Scienze Chimiche, Universita di Catania , Wale A. Doria 6,/-95127Catania , Italy Signorino Galvagno Dipartimento di Chimica lndustriale, Universita di Messina, Cas. Post. 29, 1-98166 Sant'Agata di Messina , Italy Silica supported Ru-Cu bimetallic catalysts have been studied by FTlR spectroscopy. Experiments have been carried out on two series of samples prepared from RuCI, and Ru(NO)(NO,), . The systems studied contained a total amount of metal of ca. 2 wt.%. On the monometallic Ru/SiO, samples, adsorption of CO led to an asym- metric band centred at 2036 cm-' and two small bands at a higher frequency.The band intensity was greater for the samples prepared from Ru(NO)(NO,),. CO adsorption on Cu/SiO,, leads to a band at 2117 cm-' which easily disappears upon outgassing at room temperature. On the bimetallic samples, the IR spectra indicate the formation Ru-Cu particles having a surface covered mainly by Cu; there is more Cu on the surface of the samples prepared from RuCI,. In recent years supported bimetallic catalysts have received considerable attention because of their ability to control the activity, selectivity and stability of many reactions of indus- trial interest. Among the bimetallic catalysts, Ru-Cu samples have been found to be of particular interest as model systems because the two metals are practically immiscible in bulk.The large influence of the inert copper on the activity of ruthenium indicates, however, that the two metals form bimetallic aggregates.' To investigate the nature of the bimetallic particles, a large number of investigations have been carried out on powder samples'-6 and on single crys- tal~.~.~Recent papers on supported Ru-Cu catalysts have also shown that the surface composition, and therefore the chemisorption properties and the catalytic activity, are strongly influenced by the precursor salts used and by the nature of the ~upport.~.~ In this paper we report a detailed FTIR study of CO adsorbed on Ru-Cu/SiO, samples pre- pared from Ru(NO)(NO,), . The results are compared with similar samples prepared from RuC1,. The main purpose of this work was to determine the surface composition and to find a correlation between the catalytic properties and the degree of the Ru-Cu interaction.Experimental Ru-Cu samples were prepared by incipient wetness impreg- nation of the support with aqueous solutions of Ru(NO)(NO,), or RuCl, (Johnson Matthey) and Cu(NO,), (Carlo Erba) having an appropriate concentration of metals. The concentration of salts in the solution was adjusted to yield a total (Cu + Ru) metal content of ca. 2 wt.%. The support used (supplied as powder by GRACE) was a silica gel with a BET surface area of 306 m2 g-'. After impregna- tion catalyst samples were dried at 393 K for about 24 h and reduced in a flow reactor at 673 K for 1 h with a stream of pure H,.Chemisorption of CO was measured in a conventional pulse system operating at room temperature. Pulses of 10 vol.% CO in He were used. Under these conditions negligible amounts of CO were chemisorbed by the support and on Cu/SiO, . Details of the experimental conditions used are reported elsewhere.' Chemical composition and the CO :Ru ratio of the Ru-Cu/SiO, samples are reported in Table 1. Chemisorption of CO, instead of H, , was used to avoid spill- over phenomena. Spillover of hydrogen from Ru to Cu has been reported previously.' The sample code used has the following meaning: the first two letters indicate the support used (SD = silica type D) while the three digits indicate the atomic percentage of ruthe- nium in the metallic phase.The last letters indicate the pre- cursor used [Cl = RuCl, ,N = Ru(NO)(NO,),]. For IR studies the powdered samples were pressed into thin self-supporting discs of about 25 mg cm-' and 0.1 mm thick using a pressure of 15 x lo3 bar. Pellets were evacuated and reduced in pure H, raising the temperature slowly to 673 K over a period of 2 h and held there for another 1 h. The sample was then evacuated for 30 min at 673 K and cooled at Table 1 Chemical composition and CO uptakes of Ru-Cu/SiO, samples code RU (wt.96) cu (wt.%) [Ru : (Ru + Cu)] (at.%) CO uptake /cm3(STP) (g cat.)-' CO :Ru SDlOON 2.0 - 100 4.41 0.994 SDOSON 1.7 0.3 80 3.52 0.934 SD04ON 1 .o 1 .o 40 1.76 0.794 SD020N 0.6 1.4 20 1.33 1 SD1 00Cl 2.0 - 100 0.67 0.152 SD080Cl 1.7 0.3 80 0.61 0.163 SD020CI 0.6 1.4 20 0.04 0.032 SDOOO - 2.0 0 a - Not detectable.2810 room temperature. CO was passed over the reduced sample at a pressure of 20 mbar, unless otherwise specified. Sub- sequent evacuations were performed at room temperature (RT) or at higher temperatures. The IR spectra were collected on an FTIR spectrophotom- eter Perkin-Elmer System 2000 with a resolution of 2 cm- '. No spectrum of adsorbed carbonyl species was revealed on the pure SiO, support. Data are reported as difference spectra obtained by subtracting the spectrum on the sample recorded before the interaction with CO and are normalized to the same amount of catalyst per cm2 (25 mg ern-,). Results MonometalicRu and Cu Catalysts Fig.1A shows the IR spectra of CO adsorbed at room tem- perature on the monometallic Ru/SiO, sample prepared from Ru(NO)(NO,) (SD100N) and reduced within the IR cell at 673 K for 1 h. After admission of CO [Fig. lA(a)] a very intense asymmetric band centred at 2036 cm-' (LF) is observed. A shoulder at 2083 cm-' (MF) and a band of smaller intensity at 2142 cm-' (HF) are also found. The band at higher frequency (ca. 2165 cm-') is due to 2200 2000 1800 wavenumber/cm -' Fig. 1 A, FTIR spectra of co adsorbed on the 'reduced' SDlOON sample; (a)20 mbar of co;(b) after l5 min Outgassing at RT; (') after 15 min outgassing at 473 K; (d)after 15 min outgassing at 573 K.B, FTIR spectra of co on the 'reduced' SDlml sample: (a) 20 mbar of CO; (b) after 1 min outgassing at RT; (c) after 10min outgassing at RT. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 residual CO in the gas phase. Upon evacuation at 298 K [Fig. lA(b)] the bands at 2036 and 2142 cm-' are quite stable, whereas the shoulder at 2083 m-' disappears. Evac- uation at 473 K [Fig. lA(c)] removes a large fraction of the chemisorbed CO and causes a shift to lower frequency of the LF band. After evacuation at 573 K [Fig. lA(d)] almost all chemisorbed CO is removed. The band at 2036 cm- ',which shifts to lower frequency as the coverage decreases, can be assigned to CO linearly adsorbed on metallic Ru. The coverage dependence of the carbonyl frequency on Ru/SiO, is attributed to dipoleaipole interactions between neighbouring adsorbed CO oscil-lators." However, note that before evacuation the band at 2036 cm-' is already a broad band.Therefore, it may be possible that a band at a lower frequency already exists, but it could not be assigned to Ru,CO-bridged species since a band at a significantly lower frequency (1870-1910 cm-') has been reported for these species.' The medium-frequency (MF) and high-frequency (HF) bands have been previously assigned to a small fraction of carbonyl species adsorbed on partially reduced RU.'~,'~,'~ The three bands observed in Fig. 1A(a) have an intensity ratio similar to that previously reported for Ru/SiO, prepared from Ru(NO,), .lo The spectra of CO adsorbed on the Ru/SiO, sample pre- pared from RuCl, (SD100C1) are reported in Fig.lB.-AlsO in this case three bands at 2036, 2080 and 2145 cm-' are observed. However, differences in shape and intensity are evident from a comparison with the spectra of Fig. 1A. On SDlOOCl the peak at 2036 cm-' is sharper and its intensity is lower than that found on SD100N. Note also that in contrast to what is observed on SD100N, the band at 2036 cm-' is less stable to evacuation (Fig. 1B). The differences observed between the two samples are likely to be related to different metal particle sizes. Smaller particles lead to a larger amount of Ru surface atoms and most likely to greater heter-~geneity.'~Chemisorption data (Table 1) have shown that samples prepared from Ru(NO)(NO,), chemisorb a larger amount of CO and therefore they have a larger fraction of Ru surface atoms. Formation of smaller metal particles on samples prepared from Ru(NO)(NO,), ,with respect to those prepared from RuCl, ,has been previously reported on silica- supported samples4 Fig.2A shows the IR spectra of CO adsorbed on the SDlOON sample after interaction of the sample with 0, at room temperature ('oxidized' sample). After evacuation of the sample previously used to record the spectra of CO adsorbed on the 'reduced' catalyst, oxygen (20 mbar) was passed over the sample at RT. After equilibration, CO was admitted, in the presence of O,, into the cell. The same procedure was used for all the 'oxidized' samples.Two strong bands at 2073 and 2130 cm-' are observed. Minor features are the three shoulders at 2043, 2021 and 2009 cm-'. The shift upon evac- uation of the bands at 2043 and 2021 cm-' to lower fre- quency can be explained by assuming that these bands are related to CO adsorbed on reduced ruthenium slightly per- turbed by the presence of oxygen. Removal of oxygen (by evacuation) shifts the bands to lower wavenumbers. Fig. 2B shows three well resolved peaks for the 'oxidized' SDlOOCl sample at 2138, 2083 and 2028 cm-'. For both samples, the two bands at higher frequency are quite stable to evacuation whereas the band at 2028-2030 cm- decreases in intensity and shifts to a lower frequency. These results confirm the assignment of the MF and HF bands to CO adsorbed on oxidized species.10 The band at 2028 cm-1 is likely to be related to metallic Ru still remaining on the sample after oxidation at room temperature.This is con- firmed by the shift to lower frequency observed upon evac- uation. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 2073 A 2083 B 1 2200 2000 1800 wavenumber/cm -' Fig. 2 A, FTIR spectra of CO adsorbed on the 'oxidized' SDlOON sample: (a) 20 mbar of CO; (b) after 10 min outgassing at RT. B, FTIR spectra of CO adsorbed on the 'oxidized' SDlOOCl sample: (a) 20 mbar of CO; (b) after 1 min outgassing at RT; (c) after 16 rnin outgassing at RT. Fig. 3 shows the IR spectra of CO chemisorbed at room temperature on Cu/SiO, (SDOOO).On the 'reduced' sample (Fig. 3B) only a band at 2117 cm-'is observed. The band is weak and broad and disappears upon evacuation at room temperature. According to Millar et a1.,15 this band can be ascribed to linear CO adsorbed on a stepped surface of a high-index plane of copper. As shown in Fig. 3B the fre- quency of the band is not influenced by evacuation. This is in contrast with the results of ref. 15 which show that the v(C0) position is a function of coverage. In the presence of 0, (Fig. 3A) the CO band increases and shifts to 2123 cm-'. An analogous band, attributed to CO adsorbed on CuO,' 5-' has been described previously. However, note that Hoffmann and Paul18 have reported a band at 2123 cm-' on Cu/Ru(OOl) which has been assigned to CO adsorbed on small clusters of Cu on top of Ru(OO1).Also on the 'oxidized' sample the band due to CO adsorbed on copper disappears after evacuation at RT. The frequency of CO adsorbed on 'oxidized' Cu is relatively similar to that of the HF band on Ru. However, the differences in stability towards evacuation would allow the metal sites on which CO is chemisorbed to be identified. This is in agreement with our chemisorption measurements which have shown no uptake of CO on Cu/SiO,. In fact, by using a flow system only CO irreversibly chemisorbed can be detected. 2123 A B2117 I ln cu8 I I 2200 2000 waven u mber/cm -' Fig. 3 A, FTIR spectra of CO adsorbed on the 'oxidized' SDOOO sample: (a)50 mbar of CO; (b) after 15 s outgassing at RT; (c) after 15 min outgassing at RT.B, FTIR spectra of CO adsorbed on the 'reduced' SDOOO sample: (a)85 mbar of CO; (b) after 15 s outgassing at RT; (c) after 1 rnin outgassing at RT. Bimetallic Ru-Cu Catalysts Fig. 4 shows the results obtained from the bimetallic samples prepared from Ru(NOXNO,), . Note that by increasing the Cu : Ru ratio there is a progressive decrease of the broad band at 2036 cm-'. The band at higher frequency (2132-2142 cm-') increases with the Cu :Ru ratio, then decreases with increasing Cu content. A shift of this band to lower wave- numbers is also observed. The band at 2036 cm- ' is charac- teristic of CO chemisorbed on reduced Ru. The band at higher frequency is in the same spectral region as the band due to CO adsorbed on Ru'+ and of CO adsorbed on Cu sites.However, it cannot be assigned to an Ru'+-CO species because it has quite a low stability towards outgass- ing. Fig. 5 shows the behaviour of the CO bands to evac- uation on the monometallic SDlOON and on the bimetallic SD020N samples. On the sample SDlOON the band at 2036 cm -remains practically constant upon evacuation whereas the intensity of the band at 2142 cm-' decreases to ca. 70%. A much larger decrease of the band at 2132 cm-' is observed on the SD020N catalyst. Upon the same evacuation treat- ment the intensity of the band at 2132 cm-' is reduced to <25%. This suggests that this band is related to CO chemi- sorbed on Cu surface atoms, which agrees with previous results of Knozinger and co-workers.' Moreover, a fre-quency of 2138 cm-' has been reported for single Cu atoms adsorbed on Ru(OOl)." Note also that the intensity of the band at 2132 cm-' measured on the SD020N sample is higher than that observed on the SDOOO (monometallic Cu/SiO,) sample despite the lower amount of Cu present in the sample.J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Q) Cm e n8 2200 2000 1800 wavenumber/cm-' Fig. 4 FTIR spectra of CO adsorbed on Ru-Cu/SiO, prepared from Ru(NO)(NO,), .20 mbar of CO. The effect of addition of Cu to the Ru samples prepared from RuCl, is reported in Fig. 6. On the sample SDOgOCl, the band of CO adsorbed on metallic Ru (2036 cm-l) decreases and shifts slightly to higher frequency.A strong band at 2138 cm-' is also observed. This latter band can be easily eliminated by evacuation at room temperature and therefore is assigned to CO chemisorbed on Cu. On the sample with the highest Cu : Ru ratio, only a strong band at 2138 cm-' is visible. In the spectral region where the band of CO adsorbed on Ru is expected, only a very weak and broad peak is observed. It can be therefore concluded that the amount of Ru atoms on the surface of this sample is very small which agrees with the low CO : Ru ratio measured by chemisorption. This suggests that on this sample all Ru atoms are encapsulated by Cu. Q) m gn 2200 2000 1800 2200 2000 1800 wavenumber/cm-' wavenumber/cm-' Fig. 5 A, Influence of outgassing at RT on the SDIOON sample: (a) 1 mbar of CO; (b) after 5 min outgassing at RT; (c) after 35 min outgassing at RT.B, Influence of outgassing at RT on the SD020N sample: (a) 1 mbar of CO; (b)after 5 min outgassing at RT; (c)after 35 min outgassing at RT. Q) cm ze 0 n2 m 2200 2000 wavenumber/cm-' Fig. 6 FTIR spectra of CO adsorbed on Ru-Cu/SiO, prepared from RuCl, . 20 mbar of CO. Discussion Before discussing the IR results for the bimetallic Ru-Cu/SiO, samples let us review briefly the results of cata- lytic activity obtained in the hydrogenolysis of propane over the two series of Ru-Cu/SiO, samples.20 On addition of Cu the turnover frequency (TOF) of hydrogenolysis of propane measured at 200 "C (expressed as moles of propane converted per second and per Ru surface atom) was found to decrease for both series of catalysts.20 However, the addition of Cu to samples prepared from RuCl, was much more effective; samples having a Cu : Ru ratio greater than four showed a decrease of more than three orders of magnitude in TOF.For samples prepared from Ru(NO)(NO,), the decrease in cata- lytic activity for the same variation of the Cu :Ru ratio was only one order of magnitude.,' To explain the lower catalytic activity of the Ru-Cu cata-lysts it was suggested that the active sites for propane hydro- genolysis are made of ensembles of n adjacent Ru atoms present at the surface and that the catalytic activity is related to the probability of finding such ensembles.Therefore, it has been concluded that Ru and Cu, even though they are immis- cible in the bulk state, form bimetallic crystallites. The pres- ence of inert copper on the surface of these aggregates would decrease the fraction of exposed Ru atoms and, more impor- tant, the number of active ensembles, which varies as (1 -a)" (where a is the fraction of inert copper present on the surface).,l This explains the fact that upon addition of Cu, the catalytic activity decreases much more rapidly than the Ru surface atoms. On the basis of this hypothesis it was suggested that the amount of Cu interacting with Ru (which lowers the number of active ensembles) depends, in part on the support and on the precursor used for catalyst preparation.The formation of bimetallic Ru-Cu particles having a surface covered by Cu is favoured by using RuCl, as precursor. A similar conclusion was drawn by Damiani et aL3from similar Ru-Cu samples. The formation of bimetallic Ru-Cu aggregates is confirmed by the present IR investigation. The adsorption of CO on bimetallic Ru-Cu catalysts gives spectra which are not a J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 simple combination of the bands of Ru and Cu. Further con- firmation of the interaction between Ru and Cu comes from the shift in the frequency of CO chemisorbed on Cu which is 2117 cm-' on the monometallic sample and drives up to 2138 cm-' on the bimetallic catalysts. In agreement with a previous investigation on Ru-Cu/SiO, ,l9 this shift indicates, that the Cu surface atoms are modified by an interaction with Ru which leads to a positive polarization of the Cu sites. However, in this case we would expect v(C0) to shift towards lower wavenumbers for CO adsorbed on Ru in interaction with Cu.Since this shift is not observed, it is concluded that the electronic interaction is weak. The change in the position of the band of CO adsorbed on Cu can therefore be attrib- uted to the presence of small clusters of Cu. Isolated small clusters of Cu on Ru have been reported to give a v(C0) of 2138 cm-'.I8 Moreover, note that when Cu is added to Ru the intensity of the band attributed to CO chemisorbed on Cu is greater than that for the monometallic Cu sample. This indicates that in the bimetallic particles, a large fraction of copper atoms is located on the surface.This results either from the formation of smaller particles of copper and/or through the formation of bimetallic Ru-Cu particles having a surface mainly covered by Cu. This latter hypothesis is in agreement with the lower sublimation heat and surface tension of Cu with respect to Ru which would favour a segregation on the surface of the Group 11 metal. The possibility that Cu forms smaller aggregates in the bimetallic catalysts compared with the monometallic SDOOO sample is not easy to verify. A transmission electron micros- copy (TEM) analysis of the bimetallic samples does not allow discrimination between Ru and Cu. Moreover, even in the monometallic sample, the contrast of the Cu particles in the TEM micrographs is too low to detect the Cu particles easily.The use of chemisorption techniques, such as chemisorption of N20, cannot be employed in the bimetallic samples since the Ru atoms would also interact with the probe molecule. However, note that on the SD020C1 sample, it was not pos- sible to use FTIR to detect bands due to CO adsorbed on Ru. This is in agreement with the low CO :Ru ratio mea-sured on this sample (Table l). A previous investigation by HREM of similar Ru-Cu samples prepared from RuCl, has shown that the decrease of the amount of Ru on the surface cannot be ascribed to a sintering of the Ru particles. The bimetallic catalysts do not in fact show any large Ru par- ticles.22 It can therefore be concluded that Ru and Cu form bimetallic particles on which Cu is mainly located on the surface.The FTIR results reported in this paper have confirmed that the surface composition of the bimetallic Ru-Cu par-ticles is strongly dependent on the ruthenium precursor used. The use of RuCl, favours the formation of bimetallic particles with a surface enriched by Cu atoms. The preferential surface enrichment on the samples prepared from RuCl, cannot be attributed to their lower metal dispersion. In fact, previous investigations carried out on Ru-Cu samples prepared by using the same ruthenium precursor have shown that the highest degree of interaction between the two metals and the highest Cu surface coverage are obtained on the samples with higher di~persion.~, The results reported in this communica- tion do not allow us to draw any conclusions on the role of the chemical nature of the ruthenium precursor on the surface composition.However, previous TPR (temperature- programmed reduction) experiments indicate that the differ- ent surface composition could be related to the different reduction temperature of the precursor salts. TPR experi- ments carried out on RuCl,/SiO, , Ru(NO)(NO,),/SiO, and Cu(NO,),/SiO, have shown reduction peaks at 130, 211 and 240°C, re~pectively.~~.~~ It is proposed that, owing to the lower reduction temperature of RuCl, with respect to the Cu precursor, Ru nucleation centres are formed first, and Cu atoms are deposited on top of them when the Cu salt reduction begins.In the case of the samples prepared from Ru(NO)(NO,), , the reduction of the ruthenium precursor starts at a temperature closer to that observed for Cu(NO,),/SiO, . This suggests that for the samples prepared from Ru(NO)(NO,),, the nucleation centres of Ru and Cu are formed in a very narrow range of temperatures and there- fore the bimetallic particles which are formed are likely to be more homogeneous than those prepared from RuCl, . This work was partially supported by a financial contribution from MURST and the Progetto Finalizzato Chimica Fine 11. The authors also thank Prof. G. Ghiotti for helpful dis- cussions. References 1 J. H. Sinfelt, J. Catal., 1973, 29, 308.2 S. Y. Lai and J. C. Vickerman, J. Catal., 1984,90, 337. 3 D. E. Damiani, E. D. P. Millan and A. J. Rouco, J. Catal., 1986, 101, 162. 4 A. J. Hong, A. J. Rouco, D. E. Resasco and G. L. Haller, J. Phys. Chem., 1987,91,2665. 5 X. Wu, B. C. Gerstein and T. S. King, J. Catal., 1990, 121, 271. 6 G. Ghiotti, F. Boccuzzi, A. Chiorino, S. Galvagno and C. Crisa- fulli, J. Catal., 1993, 142,437. 7 J. E. Houston, C. H. F. Peden, D. S. Blair and D. W. Goodman, Surf. Sci., 1986, 167,427. 8 B. Sakakini, A. J. Swift, J. C. Vickerman, C. Harendt and K. Christman, J. Chem. SOC.,Faraday Trans. I, 1987,83,1975. 9 C. Crisafulli, R. Maggiore, G. Schembari, S. Scire and S. Gal-vagno, J. Mol. Catal., 1989,50, 67. 10 G. H. Yokomizo, C. Louis and A.T. Bell, J. Catal., 1989, 120, 1. 11 C. R. Guerra and J. H. Schulman, Surf. Sci., 1967,7,229. 12 E. Guglielminotti,Langmuir, 1986, 2, 812. 13 J. Schwank, G. Parravano and H. L. Gruber, J. Catal., 1980,61, 19. 14 R. A. Dalla Betta, J. Phys. Chem., 1975,79,2519. 15 G. J. Millar, C. H. Rochester and K. C. Waugh, J. Chem. SOC., Faraday Trans. I, 1991,87, 1467. 16 M. A. Kohler, N. W. Cant, M. S. Wainwright and D. L. Trimm, J. Catal., 1989, 117, 188. 17 K. P. de Jong, J. W. Geus and J. Joziasse, J. Catal., 1980, 65, 437. 18 F. M. Hoffmann and J. Paul, J. Chem. Phys., 1987,87, 1857. 19 R. Liu, B. Tesche and H. Knozinger, J. Catal., 1991,129,402. 20 R. Maggiore, C. Crisafulli, S. Scir6 and S. Galvagno, in Pro-ceedings of the 10th International Congress on Catalysis, ed. L. Guczi, F. Solymosi and P. Tetenyi, Elsevier, Amsterdam, 1993, vol. B, p. 1831. 21 J. A. Dalmon and G. A. Martin, J. Catal., 1980,66,214. 22 A. G. Shastri, J. Schwank and S. Galvagno, J. Catal., 1986, 100, 446. 23 C. Crisafulli, R. Maggiore, S. Scire, C. Milone and S. Galvagno, J. Mol. Catal., 1993,83, 237. 24 S. Galvagno, C. Crisafulli, R. Maggiore, G. R. Tauszik and A. Giannetto,J. Therm. Anal., 1985,30,611. Paper 4/00490F; Received 26th January, 1994

ISSN:0956-5000    

DOI:10.1039/FT9949002809

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

J. CHEM. SOC. FARADAY TRANS., 1994, 90(18), 2815-2820 FTIR Study of the Influence of Sulfate Species on the Adsorption of NO, CO and NH, on CuO/AI,O, Catalysts Mohamed Waqif and Mahjoub Lakhdar Universite Hassan 11, Faculte des Sciences I, Casablanca, 02 Maroc Odette Saur and Jean-Claude Lavalley* Laboratoire Catalyse et Spectrochimie, URA CNRS 4 14-ISMRA-Universite, 6,Boulevard du Marechal Juin-14050 Caen-Cedex , France CuO/AI,O, catalysts have been sulfated by oxidation of various amounts of SO,. The surface properties of the samples were studied using probe molecules. The IR spectra of CO species adsorbed on the non-sulfated sample show a strong band at 2125 cm-' with a shoulder at 2160 cm-', which shift to 2165 and 2180 cm-', respectively, with a concomitant decrease in intensity as the amount of sulfate increases.Adsorbed NO species on the non-sulfated catalyst give rise to two weak bands at 1754 and 1873 cm-'. On sulfated samples, the first band disappears whereas that at 1873 cm-' shifts towards higher wavenumber and reaches 1917 cm-' for large amounts of sulfate. These results show that the copper electronic state is affected by sulfation. The positive charge on the copper sites increases with the amount of sulfate. NH, adsorption is also influenced by sulfation. Sulfate species are adsorbed on basic sites and high sulfation prevents NH, dissociative adsorption while it favours coordinated and protonated species on acid-base pair sites. Elimination of NO, from flue gas is an important subject because the emission of these oxides is one of the causes of acid rain and air pollution.The main methods for removal of NO, have been reviewed recently.' Selective catalytic reduction by NH, seems to be the most efficient process. Many catalysts based on noble metals (Pd, Ru, Pt) or transition-metal oxides have been studied. ' Amongst the latter, vanadia on titania has received much attention.'-6 The presence of sulfur oxides in an oxidizing atmosphere is inevi- table and their emission also contributes to acid rain and air pollution. Moreover, the SO, chemisorbed species are oxi- dized and stable sulfate species may remain on catalysts and influence the reduction of nitric oxides.'** Methods for simul- taneous removal of NO, and SO, have been developed and for about 30 years copper-based catalysts were tested for such a reaction.'.' Supported or unsupported copper oxide is used in a wide variety of other processes in the chemical industry.For instance, it has been shown that copper is an active com- ponent in the oxidation of hydrocarbons" or alcohols.' '-13 Various physical methods (XPS, EPR, TPR, IR) have been used to characterize copper and its interaction with the support in these copper-based materials, especially on A1,0, .l4-I7 IR spectroscopic studies using probe molecules such as CO and NO have been used in particular to deter- mine the oxidation state of copper.' '-" When copper/alumina was used as a sorbent catalyst for SO, removal, sulfate species linked to A1-0 sites and CuO were observed by IR and XPS.30*31 It has been shown that the presence of anions, such as sulfate species, modifies the acid-base character of the o~ides~'-~~ and we have recently reported that the enhancement of acidity depends on the amount of sulfate.36 The aim of this work was to study the effect of sulfate species on the electronic properties of copper.Apart from CO, which is generally used as a probe for such IR spectros-copy characterization, we have also studied the adsorption of the reactants, NO and NH, . Experimental The CuO/Al,O, sample (112 m2 g-', 4.88% w/w CuO on AI,O,) was prepared by wet adsorption of copper acetate on industrial y-Al,03, followed by a thermal decomposition in air at 450°C.Sulfate ions were introduced on the sample evacuated at 450°C by heating known amounts of SO, with a large excess of 0, at 450°C for about 14 h. Then the sample was again evacuated at 450°C for 2 h. The sulfate content was determined by an elementary microanalysis using coulometry. The notation used, the amounts of SO, added and the result of S analysis are reported in Table 1. For IR studies, self-supported discs of about 15 mg cm-, were used. They were evacuated at 450°C for 2 h. The gases were introduced at room temperature. The time required for equilibrium was very short and the spectra were scanned after ca. 10 min. All the spectra were recorded at room tem- perature using a Nicolet MX-1 FTIR spectrometer. Spectra of adsorbed CO, NO and NH, species were obtained by sub- tracting the absorbance of the activated catalyst and that of the gas phase when applicable.Results Spectra of Evacuated Sample The spectra of the evacuated samples show broad bands in the 3900-3500 cm-' and 1500-1000 cm-' ranges due to OH and sulfate groups, respectively (Fig. 1 and 2). Analysis of the broad band in the 3900-3500 cm-' range by a curve-fitting program leads to the appearance of four bands at 3780, 3725, 3650 and 3580 cm-' for the S-0 sample. The intensity of the highest-wavenumber bands decreases while the amount of sulfate increases. Only a very broad band is observed on the spectra of the S-3 sample. In the 1500-1000 cm-' range, Table 1 Notation of the samples, amounts of SO, used for sulfation and S content notation amount of SO,/pmol g- s (wt.%) s-0 - - s-0.2 50 0.15 S-0.3 100 0.3 S-0.6 200 0.6 S-1.3 400 1.3 s-3 lo00 3.2 J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 I 3800 3700 3600 3500 wavenumber/crn-l Fig. 1 IR spectra in the 3850-3500 cm-' range of: (a) S-0, (b)S-0.3, (c)S-0.6, (6)S-3 spectra show the v(S=O) band near 1375 cm-' and v(S-0) ones near 1070-1030 cm-', as previously observed on alumina.35 Increasing the amount of sulfate leads to the width of the first band owing to the superposition of another one since different sulfate species were formed on alumina when the amount of sulfate is higher than 2 pmol m-*. Moreover, for the S-3 sample other bands appear near 1230 and 1170 cm-' which had been assigned to sulfate species on copper ions.30 NO Adsorption After introduction of successive doses of ca.15 pmol g-' of NO onto the unsulfated sample, two weak bands were observed. In the presence of gas (P, = 1.3 x lo3 Pa) their maxima are at 1873 and 1754 cm-'. Their intensity increases in unison (Fig. 3), but the amount of chemisorbed NO is too low to determine the integrated molar absorption coefficient of these bands by adding successive portions of NO. The adsorbed species are easily removed by evacuation (2 min at room temperature sufficed) showing that they are very loosely held on surface. The presence of only 50 pmol g-' of sulfate leads to the disappearance of the 1754 cm-' band, whereas the band at 1873 shifts to 1882 cm-'. Fig.4 shows that the shift of this band increases with the amount of sulfate; the band wave- number reaches 1917 cm-' for the S-3 sample. The intensity I , I I I 1900 1700 wavenurn ber/crn -Fig. 3 IR spectra of species adsorbed on S-0 after addition of (a)17 pmol g-* NO, (b)150 pmol g-' NO (c) 850 pmol g-' NO, and (d)in the presence of NO gas (P, = 1.3 x lo3Pa) of the band is enhanced in the presence of sulfate species (Fig. 4). In Fig. 5, we report the variation of the absorbance between 1950 and 1825 cm-' as a function of the amount of added NO on the various samples; a linear increase is h I I 1950 1850 1750 wavenurnber/cm-' Fig. 4 IR spectra of adsorbed NO in presence of gaseous NO (P, = 1.3 x lo3Pa) on (a)S-0, (b)S-0.2, (c)S-0.3, (6)S-1.3, (e) S-3 L i 1400 1100 0 100 200 300 wavenurnber/cm-' amount of added NO/pmol g-l Fig.2 IR spectra of the sulfate species obtained by oxidation of Fig. 5 v(N0) band area us. added amounts of NO on (a) S-0, (b) various amounts of SO, :(a) S-0.2,(b)S-0.3, (c)S-0.6, (6)S-1.3, (e)S-3 S-0.2, (c) S-0.3, (d)S-1.3, (e)S-3 J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 evident followed by saturation. The linear portions of the curves exhibit nearly the same slope for the four samples, and could allow evaluation of the integrated molar absorption coeficient (E = 14 f2 cm pmol-'). It is almost constant showing that it does not vary with v(N0). From E and the area of the v(N0) band at saturation, we can estimate the number of NO adsorption sites as 48 pmol g-' on S-0.3, 58 pmol g-' on S-0.6, 66 pmol g-' on S-1.3 and 95 pmol g-' on S-3 (Fig.5). These numbers can be compared with the amounts of sulfate and copper. The adsorbed NO species formed on sulfated samples are more stable than those on pure CuO/Al,O, and, for example, on S-3 they resist evac- uation at room temperature for more than 1 h. CO Adsorption Fig. 6 shows the spectra of adsorbed species on the different samples in the presence of CO gas (P, = 1.3 x lo3 Pa). A strong band is observed at 2125 cm-' accompanied by a shoulder at 2160 cm-', in the spectrum of the non-sulfated sample [Fig. 6(a)]. The spectrum of CO adsorbed on the sul- fated samples shows a band in this spectral region.Its wave- number has been progressively shifted from 2125 to 2165 cm-' as the amount of sulfate increases. The shoulder also shifts from 2160 to 2180 cm-'. Conversely, the intensity decreases when the amount of sulfate increases (Fig. 6). v(C0) is correlated to v(S-0) (Fig. 7) which increases with the amount of sulfate.30 The CO adsorption is partly reversible at room temperature (Fig. 8). The overall amount of CO adsorbed and the irreversible part decrease with increasing amount of sulfate. However, the fraction of CO remaining adsorbed after evacuation increases with sulfate content. In order to characterize the origin of the band at 2125 cm-' with a shoulder at 2160 cm-' in the spectrum of the unsulfated sample, treatment with H, has been carried out at various temperatures before introducing CO.v(C0) does not change even after H, treatment at 400°C. However, its inten- sity decreases from 45 to 10 in arbitrary units and reoxida- tion under 0, at 400°C does not fully restore it (20 arb. units). The integrated molar absorption coefficient of the 2 125 cm -'band can be evaluated as 23 cm pmol -'. CO and NO Coadsorption On the S-0.6 sulfated sample, 1.3 x lo3 Pa of NO were intro- duced. As previously observed, a strong band appeared at m cu F cu 1400 /1392 r I EY 1384 II s. 1376 1368 2 120 2130 2140 2150 2160 2170 v ( CO)/cm -' Fig. 7 v(C0) us. v(S-0) 1892 cm-'. Then 1.3 x lo3 Pa of CO were added and gave rise to one v(C0) band at 2135 cm-'.The absorption bands due to CO or NO adsorbed alone and those due to NO and CO coadsorption are compared in Fig. 9. Note that the shoulder due to v(C0) near 2180 cm-' is absent and that the intensity of v(C0) at 2135 cm-' is slightly decreased (35%) relative to that observed with CO alone. As for NO, the intensity of the 1892 cm-band is almost unaffected (< 10%) when NO is coadsorbed with CO. ujn--. ?? -0 10 n h 0 0, I I 0 25 50 75 100 evacuation ti me/m in Fig. 8 Variation of the v(C0) band area during evacuation at room temperature: (0)S-0, (+) S-0.2, (0)S-0.3, (V)S-0.6, (A)S-1.3, (+)s-3 I 2200 1900 2200 2050 wavenum ber/cm -' 'waven um ber/cm -Fig.9 IR spectra of adsorbed species on the S-0.6 sample: (a)after Fig. 6 IR spectra of adsorbed CO (P, % 1.3 x lo3 Pa) on (a)S-O,(b) CO addition in the cell; (b) after NO addition; (c) after NO then CO S-0.2, (c)S-0.3, (d) S-0.6, (e)S-1.3,O S-3 addition I* B 3700 3100 wavenumber/cm-' '' 3700 3100 1550 1250 wavenumber/cm-' wavenumber/cm-' Fig. 10 IR spectra of adsorbed NH,: A, B, on the S-0 sample after NH, addition (pol g-'): (a)55, (b) 133, (c) 1.3 x lo3 Pa NH,, fol- lowed by evacuation at (d)room temperature (e) 100 "C,(f) 200 "C. C, D, on the S-3 sample after NH, addition (pmol g-'): (a) 133, (b) 425. (c),(d),cf)as for parts A and B. NH, Adsorption The first doses adsorbed on S-0 gave rise to the spectra shown in Fig.10A and B with bands at 3370,3260cm-'and 3150 cm-', and 1620 and 1240 cm-'. Such bands were assigned to coordinated NH, species on alumina.37 Other bands appear at 1450, 1490 and 3530 cm-' (very broad) with increasing amount of adsorption. The NH,-treated sample was then heated to 100, 200 or 400°C under vacuum or in the presence of NH, gas phase. The intensity of the bands at 1620 and 1240 cm-' decreases, whereas the temperature increases. Conversely, the bands at 1450 and 1490 cm-' remained even after evacuation at 400°C. Moreover, new bands appeared at 2280, 2250 and 2060 cm-' when the sample had been evacuated at 200 or 400"C (Fig. 11). The adsorption of successive doses of NH, on the S-3 sample has been studied. The spectra of the adsorbed NH, species on that sample are reported in Fig.1OC and D. Bands at 1620 cm-', 3360, 3260 and 3150 cm-' are observed as in the case of the unsulfated sample, but their intensity is higher. The 1240 cm-' band is masked by absorption bands due to the sulfate species. The v(S=O) sulfate band normally near 1400 cm-' is highly perturbed and wide bands appear near 0a3I_ M 2250 2050 wavenumber/cm-' Fig. It IR spectra of adsorbed species after NH, addition on the S-0 sample (P, x 5 x lo3 Pa), followed by evacuation at (a)200, (b) 400 "C J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 1250 cm-'. No bands appear at 1450, 1490 and near 3530 cm-' in contrast with the spectra of NH, adsorbed on the non-sulfated sample. The intensity of the various bands decreases concomitantly when the sample was heated under vacuum at high temperature, but we did not observe new bands in the 2300-2000 cm-' range.The sulfate band at 1400 cm-' was progressively restored. Heating the sample under NH, pressure decreases the intensity of the sulfate band while a band develops at 1450 cm-'. This can be explained by the formation of ammonium sulfate. Discussion IR Study of OH Groups and Sulfate Species The spectra of the samples show bands due to OH groups in the 3800-3600 cm-range and bands in the 1400-1000cm-region assigned to sulfate species. The bands due to the type I, I1 and I11 OH groups3* of pure alumina could be observed on the S-0 sample spectrum although they are not well resolved.This poor resolution could be explained by a lack of crystallinity of the alumina.39 The interaction of Cu2+ ions with the alumina OH groups was studied recently and it was shown that it is not specific to one type.40 These authors concluded that for samples with <6-7 wt.% CuO, copper ions are well dispersed on the alumina surface. AI-0-Cu-0-A1 links have been formed by exchange of the hydrogen of the hydroxy groups and Cu atoms. Sulfation leads first to a decrease of the v(0H) absorbance between 3780 and 3730 cm-' when the amount of sulfate increases. A single broad band remains for the highly sulfated sample from 3750 to 3600 cm-' which could be assigned to internal OH groups4' or hydrogen-bonded OH groups linked to sulfate species.The same broad band was observed on the spectrum of highly sulfated alumina without copper. The sulfate species spectra observed in the 1400-1000 cm-' range (Fig. 2)on pure al~mina~~and on CuO/Al,O, 30 have previously been discussed. In the pure alumina spectra, a band at 1380 cm-' assigned to v(S=O) was observed for low sulfate coverage,35 while for higher amounts, another species was characterized by a band near 1405 cm-'. The latter was assigned to S20,2-42 or to SO, species on Al-0.30 On CuO/Al,O,, other species on Cu-0-A1 or CuO sites had been characterized by IR bands in the 1220-1080cm-' region.,' The spectra of the various samples (Fig. 2) show that the relative amount of these different sulfate species depends on the amount of sulfate.NO Adsorption NO has widely been used as a probe molecule to study properties of various catalysts, for instance V205/Ti02 and Fe/Si02,43 and especially the oxidation state of Cu on pure Cu02' or supported on silica or al~mina.'~*'~,~~ The energy diagram of the NO molecule shows that the 7c* orbital is higher in energy than that of the 0 and N atoms and this orbital is therefore unstable. During the interaction between NO and surface ions, electron transfer preferentially occurs from NO towards the metallic ions. As Cu2+ is an electron acceptor while Cu+ a donor (the d level is full), bonds are expected between NO and Cu2+ ions. Such an interaction had been demonstrated by Gandhi and Shelef'* on copper oxide by thermogravimetric methods.Recently, Lin and co- worker~~~used low-temperature IR spectroscopic methods to study adsorption of NO on CuO/y-Al,O,. They observed three bands at 1888, 1862 and 1800 cm-'.They attributed the first two bands to the vibration of NO molecules J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 adsorbed on two Cu2+ sites in different surroundings (CuO/Al,O, or CuAl,O,). The third band was attributed to coordination of NO on Cu' sites, although no such Cu+ sites were reported in their XPS study. In Fig. 3, we observe at room temperature only two bands at 1873 and 1754 cm-' on the non-sulfated sample which could be due to NO adsorbed in a dimeric form as on reduced oxide or zeolites.22 However, two bands were also observed on Fe/SiO, at 1805-1825 and 1755 cm-' and were, respectively, attributed to NO monomer on Fe2 + and FeO.,, On sulfated samples, the intensity of the band at 1873 cm-' increases with the amount of sulfate and shifts towards higher wavenumbers while that at 1754 cm- ' disappears. As the SO,' -group attracts electrons, its presence must increase the positive charge on copper ions.We consider that the 1880 cm-' band is due to NO adsorbed on Cu2+ sites and that at 1754 cm- 'to NO on Cu+ sites. On an oxidized A120, sample containing 10% w/w CuO, Knozinger and co-workers26 observed only one band at 1865 cm-', also assigned to NO-Cu2+. These authors also observed bands in the range 1500-1600 cm-' upon heating, which indicates the formation of nitrito and nitrato surface complexes.We did not observe bands in this range at room temperature, without heating the sample. CO Adsorption On pure CuO, Busca" observed a rather strong band near 2115 cm-' due to linearly coordinated CO on Cu+ ions exposed on the CuO surface. Bands due to carbonates in the 1700-1200 cm-l range were also noted. Lin and co-worker~~~observed three bands at 2170,2148 and 2120 cm-' by IR at low temperature. They assigned the first band to CO adsorbed on Cu'+ and the last to CO on a Cu+ site. Since this last band is less sensitive to the increase in temperature, they concluded that the Cu+ adsorption sites of CO were stronger than the CU" sites. Primet and co-~orkers~~ also observed a v(C0) band near 2120 cm-' when CO was adsorbed on CuO/Al,O, , but they assigned it to CO coordi- nated to a Cu2' site.The band assignments for CO interacting with copper oxides vary widely in the literature, but it is generally report- ed that CO adsorption leads to bands appearing in the 2105- 2130 cm-' range for CO on metallic copper, or in the 2115-2130 and 2120-2140 cm-' ranges for CO on Cu+ and Cu2+,respectively. On CuO/Al,O, without sulfate, we have noted a band at 2125 cm- ',which we assign to CO adsorption on Cu+ sites in agreement with ref. 20 and 29. This band shifts towards higher wavenumber and its intensity decreases with sulfate loading. Fig. 7 shows a correlation between v(C0) and v(S=O), showing the positive charge of the Cu sites increas- ing with sulfate loading.Our assignment of the 2125 cm-' band to CO adsorption on Cu+ sites accounts for the energy levels of CO. The CT orbital of CO is lower in energy than that of the 0 and C atoms and this orbital is therefore relatively stable. Conse- quently, it is difficult to transfer the CT electron of CO. Cu+ is a better electron donor than Cu2+ (3d"). Therefore, the transfer of a 3d electron to the n* orbital of CO is easier for Cu+ than Cu2+. The decrease in the intensity is due to the attractive effect of SO,,-groups which favours an increase of the positive charge on the Cu atoms. It could also be partly explained by the poisoning of the Cu2+ ions by sulfate groups, since the IR spectra of these SO,2- species show bands near 1150 cm-' which were assigned to in interaction with Cu or CU-A~.~' The experiment with H2-reduced CuO/A1,0, did not show any new band due to metallic copper but a decrease of the intensity of the v(C0) band assigned to the Cu' site.This may be due to a decrease in the dispersion of copper with H, reduction. We did not observe a band of significant intensity resulting from the adsorption of CO on Al,O, since the bands expected in the 2200-2240 cm-' range for CO coordination on activated alumina were not detected. Note that the sulfate species poison the strongest Lewis-acid sites, A13+, and emphasize the acidity of the least acidic sites.36 The spectrum obtained by CO adsorption after intro- duction of NO (Fig.9) shows simultaneously the band char- acteristic of NO coordinated to Cu2+ ions (band at 1892 cm-') and that due to CO coordinated to Cu+ ions (band at 2135 cm-'). The presence of the two different sites (well dis- tinguished by the use of the two probes) is confirmed. The shoulder observed at 2160 cm- 'upon CO adsorption (Fig. 6) and assigned to some Cu2+ ions does not appear when CO and NO are coadsorbed, which confirms that the adsorption of NO on Cu2+ ions is stronger than that of CO. NH, Adsorption IR studies of NH, adsorbed on activated A1,0, have been the subject of many Apart from species coordi- nated to Lewis-acid sites, characterized by IR absorption bands at 1620 and 1240 cm-' and hydrogen-bonded species giving rise to broad bands near 3300 cm-', some authors observed other bands in the 1550-1450 cm-' range whose assignment was not easy.Hall and co-~orkers~~ thought that these bands could result from dismutation of NH, on acid- base pair sites leading to NH,+and NH,-. On sulfated alumina, activated up to 400 "C, a strong band at 1620 cm-due to coordinated species was observed, but there were no bands in the 1550-1450 cm-I range. This is in agreement with Hall's interpretation, since sulfate species poison the basic sites and the dismutation could not occur. Moreover, NH, coordination shifts the v(S=O) band towards lower wavenumber as observed for pyridine ad~orption.~~ On CuO/Al,O,, we observe the same bands as on pure alumina [Fig. 10A, B]. They are also assigned to the forma- tion of coordinated species (bands at 1620 and 1240 cm-') and to NH,' and NH,- ions (bands at 1450 and 1490 cm-').The NH,-treated S-0sample has been heated under vacuum and this treatment leads to an increase in the inten- sity of the NH4+ and NH2- absorption bands (1450, 1490, ca. 3500 cm-') and to the appearance of new bands in the 2200-2050 cm-' range (Fig. 11). The last bands could be assigned to species resulting from NH, dissociation and espe- cially to Cu-NSN species, in agreement with Rochester and co-~orkers~~ or to surface-bound a~ide.~' On spectra of sulfated CuO/Al,O, samples (Fig. lOC, D) only one band (1620 cm-') due to NH, coordinated species is observed, since the bands due to sulfate species overlap the 1240 cm-' band.NH, adsorption sites are close to these species since the sulfate absorption bands are highly per- turbed. Heating in the presence of NH, gas leads to the for- mation of NH,' species (band at 1450 cm-') and ammonium sulfate, as previously observed.46 Some hydrogen-bonded species are also formed (negative absorb- ance near 3700 cm- ') but dissociation of NH, does not occur. This could be explained by the poisoning of basic sites and copper sites by sulfate. Conclusion This IR study of copper on alumina catalysts containing various amounts of sulfate ions shows that the presence of 2820 J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 23 A. A. Davydov, in Infrared Spectroscopy of Adsorbed Species on such anions increases the positive charge of the copper sites. the Surface of Transition Metal Oxides, Wiley, Chichester, 1990.Moreover, owing to sulfate adsorption on basic sites, high 24 M.C. Marion, E. Garbowski and M. Primet, J. Chem. SOC.,sulfation of the catalysts prevents NH, dissociative adsorp- Faraday Trans., 1990,86,3027.tion. Only coordinated and protonated species are then 25 R. Hierl, H. Knozinger and H. P. Hurbach, J. Catal., 1981, 69, formed. This shows that not only the acid-base properties 475. but also the copper electronic state of the catalysts are 26 R. Hierl, H. P. Hurbach and H. Knozinger, J. Chem. SOC., affected by sulfation. The reduction of NO, by NH, in emu- Faraday Trans., 1992,88,355. ent gas containing SO, must account for these perturbations.27 M. A. Kolher, N. W. Cant, M. S. Wainwright and D. L. Trimm, J. Catal., 1989, 117, 188. 28 G. J. Millar, C. H. Rochester and K. C. Waugh, J. Chem. SOC., Faraday Trans., 1991,87, 1467. References 29 Y. Fu, Y. Tian and P. Lin, J. Catal., 1991,132,85. 1 H. Bosch and F. Janssen, Catal. Today, 1988,2. 30 M. Waqif, 0.Saur, J. C. Lavalley, S. Perathoner and G. Centi, J. 2 R. A. Rajadhyaksha and H. Knozinger, Appl. Catal., 1989, 51, Phys. Chem., 1991,95,4051. 81. 31 B. Kartheuser, B. K. Hodnett, A. Riva, G. Centi, H. Matralis, M. 3 N. Y. Topsoe, J. Catal., 1991,128,499. Ruwet, P. Grange and N. Passarini, Znd. Eng. Chem. Res., 1991, 4 T. J. Dines, C. H. Rochester and A. M. Ward, J. Chem. SOC., 30,2105.Faraday Trans., 1991,87, 1473.32 K. Tanabe, M. Misono, Y. Ono and H. Hattori, Stud. Surf: Sci. 5 M. M. Kantcheva, K. Hadjivanov and D. G. Klissurski, J. Catal., 1989,51, 199. Catal., 1992, 133, 8643. 33 T. Yamaguchi, Appl. Catal., 1990, 61, 1. 6 G. Ramis, G. Busca, F. Bregani and P. Forzatti, Appl. Catal., 34 (a) C. Morterra, G. Ghiotti, E. Garrone and E. Fisicaro, J. 1990,64,259. Chem. SOC.,Faraday Trans. I, 1980,76,2102; (b)C. Morterra, G. 7 M. Iwamoto, H. Yahiro, S. Shundo, Y. Yu-u and N. Mizuno, Cerrato, C. Emanuel and V. Bolis, J. Catal., 1993, 142, 349. Appl. Catal., 1991,69, L-15. 35 0.Saur, M. Bensitel, A. B. Mohammed Saad, J. C. Lavalley, C. 8 G. Zhang, T. Yamaguchi, H. Kawakami and T. Suzuki, Appl. P. Tripp and B. A. Morrow, J. Catal., 1986,99, 104. Catal. B: Enuiron., 1992, 1, L-15.36 M. Waqif, J. Bachelier, 0. Saur and J. C. Lavalley, J. Mol. 9 A. A. Siddiqi and J. W. Tenini, Hydrocarbon Proc., 1981,60, 115. Catal., 1992,72, 127. 10 R. Prasad, L. A. Kennedy and E. Ruckenstein, Catal. Rev. Sci. 37 A. A. Tsyganenko, D. V. Posdnyakov and V. N. Filimonov, J. Eng., 1984,26, 1. Mol. Struct., 1975,29, 299. 11 H. Kobayashi, N. Takezawa and C. Minochi, J. Catal., 1981,69, 38 H. Knozinger and P. Ratnasany, Catal. Rev. Sci., Eng., 1984, 26, 487. 163. 12 N. M. Dobrynkin, A. A. Davydov, A. A. Budneva, V. V. Popovs-39 P. Nortier, P. Fourre, A. B. Mohammed Saad, 0.Saur and J. C. kii, V. A. Rogov and V. F. Serebryakov, Kinet. Catal., 1992, 33, Lavalley, Appl. Catal., 1990,61, 141. 133. 40 E. Garbowski and M. Primet, J.Chem. SOC., Chem. Commun., 13 G. J. Millar, C. H. Rochester and K. C. Waugh, J. Chem. SOC., 1991, 11. Faraday Trans., 1991,87,2785. 41 A. A. Tsyganenko, K. S. Smirnov, A. M. Rzhevskij and P. P. 14 R. M. Friedman, J. J. Freeman and F. W. Lytle, J. Catal., 1978, Mardilovich, Muter. Chem. Phys., 1990,26,35.55, 10. 42 M. Bensitel, 0. Saur, J. C. Lavalley and B. A. Morrow, Muter. 15 S. F. Tikhov, V. A. Sadikov, G. N. Kryukova, E. A. Paukshtis, Chem. Phys., 1988,19,147.V. V. Popovskii, T. G. Starostina, G. V. Kharlamov, V. F. Anuf-43 C. Johnston, N. Jorgensen and C. H. Rochester, J. Chem. SOC.,rienko, V. F. Poluboyarov, V. A. Razdobarov, N. N. Bulgakov Faraday Trans. I, 1988,84,2001.and A. V. Kalinkin, J. Catal., 1992,134, 506. 44 J. C. Duchet, J. C. Lavalley, D. Ouafi, J. Bachelier, D. Cornet, C. 16 B. R. Strohmeier, D. E. Leyden, R. S. Field and D. M. Hercules, Aubert, C. Moreau, P. Geneste, M. Houari, J. Grimblot and J. J. Catal., 1985,94, 514. P. Bonnelle, Catal. Today, 1988,4, 97. 17 V. K. Kaushik, Ch. Sivaraj and P. Kanta Rao, Appl. Surf. Sci., 45 J. Valyon, R. L. Schneider and W. K. Hall, J. Catal., 1984, 85, 1991,51, 27. 277. 18 H. S. Gandhi and M. Shelef, J. Catal., 1973, 28, 1. 46 J. S. Lee, M. H. Yeom and D. E. Park, J. Catal., 1990,126, 361. 19 J. W.London and A. T. Bell, J. Catal., 1973,31, 32. 47 E. Jobson, A. Baiker and A. Wokaun, J. Chem. SOC., Faraday20 G. Busca, J. Mol. Catal., 1987,43, 225. Trans., 1990,86, 1131. 21 R. Bechara, A. Aboukais and J. P. Bonnelle, J. Chem. SOC., Faraday Trans., 1993,89, 1257. 22 M. C. Kung and H. H. Kung, Catal. Rev. Sci. Eng., 1985,27,425. Paper 4/01935K; Received 30th March, 1994

ISSN:0956-5000    

DOI:10.1039/FT9949002815

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

J. CHEM. SOC. FARADAY TRANS., 1994, 90(18), 2821-2826 Adsorption on MCM-41 Mesoporous Molecular Sieves Part 1.-Nitrogen Isotherms and Parameters of the Porous Structure J. Rathousky* and A. Zukal J. Heyrovsky Institute of Physical Chemistry, Academy of Sciences of the Czech Republic, Dolejskova 3, 18223 Prague 8,Czech Republic 0.Franke and G. Schulz-Ekloff Institute of Applied and Physical Chemistry, University of Bremen, 28359 Bremen, Germany Nitrogen adsorption isotherms have been measured on a series of aluminosilicate and titanosilicate MCM-41 molecular sieves, whose mean pore radius varied from ca. 0.9 to ca. 2 nm. By means of comparison plots the nature of the adsorption on these materials was found to depend strongly on their pore size. With pores of radius of around 1 nm rnultilayer coverage of the pore walls occurs. If the radius is increased to 1.5-1.8 nm the mechanism of adsorption changes into a two-stage one, i.e.the multilayer coverage of the pore walls is suc-ceeded by the spontaneous filling of the pore volume by capillary condensation without hysteresis. In even greater pores the usual capillary condensation with hysteresis occurs. The estimation of pore structure parameters was based on standard methods of adsorption isotherm pro- cessing, including the calculation of the pore size distribution from the desorption branch of the hysteresis loop. With the smaller pore materials, where the Kelvin equation does not hold, the estimation was based on the cylindrical pore model. A detailed knowledge of the pore structure of MCM-41 materials was thus obtained.The synthesis of a new family of mesoporous molecular sieves using rodlike micelles of cationic surfactant molecules as tem- plates has been reported recently.'g2 These silicate and aluminosilicate materials were designated as MCM-41. Recent results on silicate MCM-41 have revealed that ran- domly oriented rod-like micelles interact with silicate species to yield two or three monolayers of silica encapsulating the external surface of the mi~elles.~.~ Subsequently, these com- posite species spontaneously assemble into the long-range ordered structure characteristic of MCM-41 and then the sili- cate species in the interstitial spaces of the ordered organic- inorganic phase continue to condense.After the removal of the organic part of these organic-inorganic composites uni- formly sized pores are left. This mechanism is most probably operative also in the formation of MCM-41 materials of other chemical composition. MCM-41 molecular sieves have been synthesized with a regular, hexagonal array of uniform channels of approx- imately hexagonal cross-section. If their shape is modelled by a cylinder, their radius varies from 0.5 nm to ca. 5 nm. The most regular arrangement and the best uniformity of pores are usually observed for the smaller pore size materials (r < 2 nm). Those with larger pores display somewhat irregular, yet essentially hexagonal pore arrangements and a slightly deformed pore shape.* MCM-41 molecular sieves are promising materials for both various applications and fundamental ~tudies.~,~ On these molecular sieves capillary condensation without hyster- esis in a system of uniform cylindrical pores was first observed. In two studies'** published so far on this adsorp- tion phenomenon only one sample was investigated with mean pore sizes of 1.46 nm ' and 1.66 nm,' respectively.Data presented in this study were obtained on a series of six samples whose mean pore radius varied from ca. 0.9 to ca. 2 nm. Nitrogen adsorption isotherms were measured because with this adsorbate the methods of analysis of adsorption data and of the determination of parameters of the porous structure of solids are satisfactorily verified.For the analysis of adsorption isotherms comparison were used, i.e. the amount adsorbed on the solid under investigation was plotted against that adsorbed on a reference non-porous adsorbent at the same equilibrium pres- sure. By this method the adsorption on both materials is compared directly without any normalizing factor. The parameters of the porous structure of MCM-41 materials were determined using standard methods and were then used for discussing the observed adsorption processes. Experimental Materials The sources of silica, aluminium and titanium were Ludox AS-40 (Du Pont, 40 wt.% colloidal silica in water), Al(OH), (J. T. Baker, 98%) and tetrabutyl orthotitanate (TBOT, Merck, p.a.). The quaternary ammonium surfactant com-pounds [C,H2,+ ,(CH,),NX, X = OH, C1, Br] were obtained from Fluka (25% water solutions).These compounds are designated as DDTMAX and HDTMAX for n = 12 and 16, respectively. Tetraethylammonium hydroxide solution (TEAOH, Merck, 20 wt.% solution in water), sodium hydrox- ide (Janssen, p.a.), propan-2-01 (Merck, p.a.), toluene (Merck, p.a.) and mesitylene (Merck, p.a.) were also used. All chemi- cals were used as received. Synthesis Procedures Because of the need to obtain materials with pore sizes varying from 1 to 2 nm differing in their chemical composi- tion (aluminosilicates, titanosilicate), three synthesis pro-cedures were adopted, uiz. the procedure I with samples AIMS-1 to -3, the procedure I1 with AIMS-5 and -6, while the titanosilicate TiMS-4 was prepared using procedure 111.All aluminosilicate sieves contained 3.1 mol% of A1,0,, the tita- nosilicate TiMS-4 contained 4.3 mol% of TiO, (determined by AAS). The structure of all sieves was also checked using X-ray diffraction (XRD) and transmission electron micros- copy (TEM). The characteristic feature of their diffractograms is that they exhibit from two to three reflections only at small angles 28 (< 6"). From the comparison with published data',, it follows that they can be indexed on a hexagonal lattice typical of MCM-41. The transmission electron micrograph in Fig. 1 Transmission electron micrograph of AIMS-2 Fig. 1 confirms that the typical example of aluminosilicate MCM-41, AIMS-2, contains an almost regular, hexagonal array of channels with a mean pore diameter of ca.3 nm. Procedure I The reaction mixture was prepared as follows: 0.31 g of Al(OH),, 0.3 g of sodium hydroxide and 1 g of deionized water were put into a 60 ml glass beaker and brought to the boil with stirring until a clear solution resulted. Then 9.26 g of TEAOH solution were added and the solution was cooled. In a separate 400 ml polypropylene beaker 9.26 g of Ludox AS-40 were stirred with a magnetic stirrer (at ca. 600 rpm). These two mixtures (sodium aluminate and the silica suspension) were combined at room temperature by adding the aluminate solution to the silica suspension. The gel was stirred for 5 min (to achieve good homogeneity, an agitation rate of up to 1000 rpm may be needed), mixed with an amount of the 25% aqueous solution of surfactant (see below) corresponding to the composition of the reaction mixture presented below (e.g.10.55 g with HDTMAOH), and stirred for another 5 rnin (500 rpm). Then the gel was reacted with stirring (150 rpm) in a 250 ml polypropylene autoclave at 104 "C for 24 h. The resulting solid product was recovered by filtration, washed with water, extracted with ethanol for 4 h in a Soxhlet apparatus and finally calcined in air at 600°C for 22 h. The reaction mixture of samples corresponded to an oxide molar ratio -of lA1,0, : 31.01Si02 : 2.2(surfactant),O : 3.16(TEA),O : 1.89Na20 : 615-802H20. The content of water depended on the molecular weight of the surfactant used.The following surfactants were used : with AlMS-1, DDTMABr ; with AIMS-2, HDTMAOH; with AIMS-3, HDTM ABr. Procedure I1 The reaction mixture was prepared as follows: 0.62 g of Al(OH),, 0.6 g of sodium hydroxide and 1.5 g of deionized water were put into a 250 ml glass beaker and brought to the boil. After a clear solution resulted, 18.52 g of TEAOH solu- tion were added and the solution was cooled. Then 22.49 g of the solution of HDTMACl were added. Aluminate, which precipitated at first, dissolved again on warming. In a separate 400 ml polypropylene beaker 18.52 g of Ludox AS-40 were agitated with a magnetic stirrer (at ca. 500 rpm). The two mixtures (sodium aluminate and the silica suspension) were combined at room temperature by adding the aluminate solution to the silica suspension.Immediately a gel formed which was stirred for 30 min (at 500 rpm). Then 2.1 g of toluene (with AIMS-5) or 3.2 g of mesitylene (with AIMS-6) was added and the mixture was homogenized for 1 rnin at an agitation rate of up to lo00 rpm. Later the gel was reacted with stirring (150 rpm) in a 250 ml polypropylene J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 autoclave at 104°C for 4 h. The resulting solid product was recovered by filtration and treated as in procedure I. The reaction mixture corresponded to an oxide molar ratio of 1A1,0, : 31.06Si02 :2.21(HDTMA),O : 3.16(TEA),O : 2.20(toluene), or 3.35(mesitylene), :1.89Na20:806H20. Procedure I I I The reaction mixture was prepared as follows: 19.26 g of Ludox AS-40 were put into a 400 ml polypropylene beaker and stirred by a magnetic stirrer at an agitation speed of ca.1200 rpm. Afterwards, 18.52 g of TEAOH solution and sub- sequently 16 g of HDTMACl (1/3 of the total amount) were added. To the gel formed the other 2/3 of the surfactant and 1.8 g of TBOT, diluted with 1.9 g of propan-2-01, were added simultaneously. All vessels, except that with TBOT, were washed with 10 g of deionized water to achieve a quantitative transfer of reagents. As the vessel with TBOT could not be washed with water because of the hydrolysis, an extra 0.05 g of TBOT was always added to make up for the losses. All components were cooled down to 10°C in an ice bath and also the gel was formed at the same temperature.The gel was agitated for 1 rnin and then reacted with stir- ring (150 rpm) in a 250 ml polypropylene autoclave at 104 "C for 24 h. The resulting solid product was recovered by fil- tration and treated as in procedure I. The reaction mixture corresponded to an oxide molar ratio of 31.01Si02 :0.64(TBOT), : 4.53(HDTMA),O : 3.04(TEA),O :4.50(C3H,),0 :972H20. Preparation of Reference Adsorbents As the nature of the surface of the reference adsorbents should be identical or at least similar to that of the invest- igated solid, they were prepared by the thermal destruction of appropriate aluminosilicate or titanosilicate MCM-41 materials at lo00 "C for 2 h. Methods The adsorption isotherms of nitrogen (Linde, purity 99.9996%) at -196°C were measured with an Accusorb 2100E instrument (Micromeritics).The temperature of the liquid-nitrogen bath was measured by a thermistor probe. Each sample was degassed at 330°C for at least 20 h until a pressure of Pa was attained. With TiMS-4, AIMS-5 and AIMS-6 the scanning behaviour was investigated, i.e. desorp-tion steps were started before the saturation pressure p" was reached. Powder X-ray diffraction data were obtained on a Seifert 3000 P diffractometer in the Bragg-Brentano geometry arrangement using Co-Kor radiation with a graphite mono- chromator and a scintillation detector. A TEM image was obtained on a Phillips EM 420 trans- mission electron microscope, operated at 120 kV, equipped with an LaB, cathode.Results Reference Adsorbents When the adsorption on reference aluminosilicate adsorbents was related to the unit surface area all isotherms were practically identical. For the analysis of adsorption data the isotherms on reference adsorbents prepared from AIMS-2 (available in the largest amount) and TiMS-4 were used. The BET surface area of the aluminosilicate reference adsorbent was 25.5 m2 g-' and the constant, c = 45.8. With the titan- osilicate reference adsorbent the respective values were 6.1 m2 J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 40 P 0 0.2 0.4 0.6 0.8 1 PIP" Fig. 2 Adsorption isotherms of nitrogen at -196°C on AlMS-1 (a) and AlMS-6 (b): (0)adsorption, (m) desorption g-' and 16.3.The BET equation was valid for p/p" from ca. 0.05 to 0.40, and less precisely up to about 0.5-0.6. MCM-41 Materials With the exception of AlMS-1 (Fig. 2), the adsorption iso- therms on all alumino- and titano-silicate sieves were ana- logous. Their shape is unusual with a sudden increase in adsorption occurring at p/p" =-0.25 (Fig. 2 where AlMS-6 was shown as a typical example). The position of this increase can be characterized by which is the relative pressure corresponding to the point of inflection (Table 1). Adsorption isotherms can be approximated by the BET equation from p/po z0.1 (with all sieves) to an upper limit which increased with the serial number of the sample from ca. 0.20 (AlMS-1) to ca. 0.35 (AlMS-6). Monolayer capacities, n,, c constants and surface areas, ABET,are summarized in Table 1.Discussion The regular porous structure, which is created using rod-like micelles as templates, is termed the primary structure. In the later stages of crystal growth, a secondary mesoporous struc- Table 1 Adsorption of nitrogen at -196 "C on MCM-41 molecular sieves (the BET equation) sample n,,,/mmol g - c ABET/mZg-' (p/p")inf AlMS- 1 8.377 78.3 817.5 - AlMS-2 10.982 76.0 1071.7 0.29 AlMS-3 9.953 55.3 971.3 0.33 TiMS-4 8.474 59.4 826.9 0.34 AIMS-5 8.416 62.9 821.3 0.39 AlMS-6 10.403 61.3 1015.2 0.43 n,, monolayer capacity; c, constant in the BET equation; ABET, BET surface area; (~/p')~,,~,relative pressure at the point of inflection in the region of capillary condensation.ture is formed with irregular pores of practically all sizes. It is closely related to the crystal morphology. Application of the Comparison Plots The surface areas of both reference adsorbents were small, which clearly shows that the porous structure of the original molecular sieves was totally destroyed and the thermal treat- ment caused the sintering of the material. Therefore, adsorp- tion isotherms on them can be used for the construction of comparison plots, in which the adsorption in the porous system of MCM-41 materials is compared with that on the flat surface. The treatment used is analogous with that of ref. 10. Fig. 3 shows the dependence of n us. nref (where n is the adsorption on the adsorbent studied, while nrefis that on the reference adsorbent at the same equilibrium pressure) obtained by the transformation of the adsorption branches of isotherms up to p/p" = 0.8.The low-pressure part of all plots of n us. nref can be approximated by a straight line going through the origin. With AlMS-1, a sharp knee and a plateau with a small slope follow. In contrast, both AlMS-2 and AlMS-6 are characterized by a steep upward swing passing gradually into a plateau. With AlMS-2 this steep increase, occurring in the reversible part of the isotherm, has been already explained as a consequence of capillary condensation without hysteresis in primary me sop ore^.^ With AlMS-6, in contrast, it occurs in the hysteresis region.By scanning the hysteresis loop (Fig. 4) it was found that the adsorption hys- teresis occurs during the filling of primary mesopores. The Kelvin capillary condensation mechanism is obviously oper- ative here. The direct proportionality indicates that in the first stages of adsorption the formation of adsorbed layers on MCM-41 is the same as that on a non- porous adsorbent. The slope B, (Table 2) determines the surface area of the studied solid A = B1Aref, where Arefis the surface area of the reference adsorbent. As the ratio ABET/B1 (Table 2) is very close to Aref,the similarity of the mechanism 30 A7.. . ..... -. . ._.. ... .. -20.-I -0 E E--. 0 0.2 0.4 0.6 0.8 nre,/mmol g -' Fig. 3 Comparison plot of the adsorption of nitrogen on AIMS-1 (a),AlMS-2 (b) and AlMS-6 (c) 2824 30 25 r I (5, E --.E C 20 15 I I I 0.35 0.45 0.55 0.65 0.75 P/P" Fig.4 Scanning of the hysteresis loop of the adsorption isotherm on AIMS-6: (0)adsorption, (m) desorption of adsorption on MCM-41 and the reference adsorbents is confirmed. By a detailed inspection of the n us. nref plots, small but systematic deviations from direct proportionality were found to occur. From this it follows that the surfaces of MCM-41 materials and reference adsorbents are not totally identical as regards the nature of the adsorbent-adsorbate interactions. This fact is apparent from the variation of c (from the BET equation, Table l), which for MCM-41 materials varies between 55.3 and 78.3, while for the reference absorbents it decreases to 45.8 (aluminosilicate) or as low as 16.3 (ti tanosilica te).J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 The plateau observed in n us. nref plots of AlMS-1, AlMS-2 and AlMS-3 at around p/p" = 0.4 obviously formed after the primary mesopore structure had been totally filled. The small slope of this plateau is caused by the multilayer coverage of the surface of the secondary mesopores. If this part of the n us. nrefplot is approximated by the straight line n = C, + Clnref (2) then C, gives the adsorption capacity of primary pores and C, the surface area of secondary mesopores, Ame2= C,A,,,. The values of C,, C, and Am,* together with the volumes of the primary pores Vmel= Coum0,(umol being the molar volume of nitrogen in the liquid state) are given in Table 2.With TiMS-4, AlMS-5 and AlMS-6, eqn. (2) cannot be applied because primary mesopores are filled by Kelvin capil- lary condensation, while on the reference adsorbent multi- layer coverage of the surface occurs. On the basis of the parameters Vmel and Ame2, the mean radius of the primary mesopores, rmel,can be simply estim- ated. For a cylindrical pore: The calculated values of rmel are presented in Table 2. Determination of the Parameters of the Porous Structure of MCM-41 Molecular Sieves As the method of comparison plots cannot be used with all samples for the determination of the volume of primary pores and the surface area of secondary ones, the distribution of the volume and the surface area of mesopores was calculated from the desorption branch of the hysteresis loop by the method of Dollimore and Heal.' The calculation directly provided V,, =f(rme) and A,, = g(rme), where V,, and A,, are the volume and surface area of pores of radius r 2 r,,,,.Table 2 Adsorption of nitrogen at -196"Con MCM-41 molecular sieves (comparison plots) AlMS- 1 33.76 24.7 9.102 3.46 0.317 88.2 0.87 AIMS-2 44.99 23.8 21.337 2.58 0.743 65.8 1.48 AlMS-3 3 8.47 25.2 19.103 4.54 0.665 115.8 1.55 TiMS-4 163.42 5.06 - - - - - AlMS-5 33.26 24.7 AlMS-6 40.37 25.2 ~~~~ ~ B,, C,and C, are constants of the linear parts of the comparison plot; V,,, and rmelare the volume and radius of primary mesopores; Arne, is the surface area of secondary mesopores.Table 3 ~~ AIMS-1 ~ 0.344 124.4 AlMS-2 0.578 168.9 AIMS-3 0.525 163.7 TiMS-4 0.345 286.0 AIMS-5 0.646 480.1 AlMS-6 1.056 864.4 Porous structure parameters of MCM-41 molecular sieves 74.5 743.0 0.368 51.1 1020.6 0.768 74.6 896.7 0.686 35.1 791.8 0.704 41.4 779.9 0.716 94.8 920.4 0.895 (v,,),,,and (A,,),,, are the volume and surface area of all pores with r 2 1.8 nm; (A*),,', (I/*),,', (n*),,, volume, capacity and mean radius of pores with r -= 2.5 nm; (A*),,, ,surface area of pores with r 2 2.5 nm. 10.565 0.99 22.049 1.50 19.695 1.53 20.21 1 1.78 20.556 1.84 25.695 1.94 and (r*),,' are the surfacea area, J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 These distribution functions end at I,, = 1.8 nm (the lower closure point of the hysteresis loop), the corresponding values of V,, and A,, being equal to the total surface area (A,,),,, and volume (V,,),,, . At the lower closure point of the hysteresis loop all pores with r < 1.8 nm are filled by the adsorbate. Their volume V,, .8 can be calculated according to where no.4 is the adsorption at the lower closure point of the hysteresis loop and (nief)0.4is the adsorption of the reference adsorbent at p/p" = 0.4 related to the unit surface area. The dependence of the mesopore volume on their radius V,, = Vme(rme)calculated by the method of Dollimore and Heal'' can be converted to an increasing function of pore radius using which gives the volume of pores whose radius lies between 1.8 nm and T,,.Then the total volume of pores of r d r,, equals The dependence of V us. rmeis shown in Fig. 5. With TiMS-4, AlMS-5 and AlMS-6 the distribution of primary mesopores is partly included. It must be considered with caution as the 1.o 0.8 0.6 01 $ 0.4I-0.2 0.0 1.5 2.0 2.5 3.0 3.5 4.0 r/nm Fig. 5 Pore size distribution I/ vs. rme: (a) AIMS-1, (b) AIMS-2, (c) AIMS-3, (d)TiMS-4, (e) AIMS-5, cf)AIMS-6 Table 4 Adsorption of nitrogen at application of the Kelvin equation is not straightforward at p/po +0.4.' Nevertheless, the conclusion can be reached that with AlMS-5 and AIMS-6 primary mesopores are not uniform, their radius lying in an interval of the width of ca.0.5 nm. This conclusion is also supported by TEM (Fig. 1). With all of these samples r,, = 2.5 nm can be chosen as the upper limit of the pore size of primary mesopores. Based on a chosen limiting pore size, a set of porous structure parameters can be obtained (Table 3). The volumes of primary mesopores (V*),, were determined directly from the plot of V us. rme [eqn. (6),Fig. 51. With AlMS-1, -2 and -3, (V*),,, is rather larger than Vmeldetermined using com- parison plots. It is clear why: the volumes determined using comparison plots relate only to the primary mesoporous structure, while (V*),,, relates to all pores of T < 2.5 nm.(For this reason all parameters based on the limit of 2.5 nm are designated by an asterisk). Using (V*),,', the adsorption capacity of primary mesopores can be simply calculated according to (n*),,' = (V*)mel/~mo,.The surface area of primary mesopores, (A*)mel,is given by the difference A,,, -(A*),,*, where the surface area of secondary mesopores, (A*),,, ,was directly determined from the distribution curve, A,, = g(rme),as the surface area of all pores of r > 2.5 nm. The mean radius of the primary pores (r*),,' was estimated analogously to rmelaccording to eqn. (3). All parameters of the porous structure are summarized in Table 3. With AlMS-1, -2 and -3, rmel and (r*)mel are in rea- sonable agreement; the values of rme1calculated using com- parison plots are obviously nearer the reality.The surface area of secondary mesopores is relatively small with all samples varying between 50 and 100 m2 g-'. With increasing radius of the primary mesopores, however, the total surface area of the mesopores, (A,,),,,, calculated from the desorp- tion branch of the hysteresis loop increases and approaches the total surface area, ABET. Mechanism of Adsorption in Primary Mesopores It can be shown that trends of the values of parameters derived from adsorption isotherms and comparison plots are reasonable, being in agreement with the concepts of observed adsorption phenomena. The values of the relative pressure, (pip"),,and the adsorp- tion, ne, corresponding to the end of the direct proportion- ality between n and nref were determined from comparison plots (Table 4).In order to compare individual samples the quotients n,/n, were calculated, expressing n, in terms of the monolayer capacity, n, . Analogously, the mean pore radius of the primary mesopores was expressed as the quotient (r*)mel/~m, where the mean statistical thickness of the nitro- gen monolayer, t, = 0.354 nm. It is evident from Table 4 that (pip"), and n,/n, increase with increasing (~*),,~/t,. The region of equilibrium pressure -196°C on MCM-41 molecular sieves AIMS- 1 2.8 0.19 9.708 1.16 0.83 AIMS-2 4.2 0.26 15.256 1.39 0.66 AIMS-3 4.3 0.30 14.429 1.45 0.68 TiMS-4 5.0 0.30 13.059 1.54 0.62 AlMS-5 5.2 0.34 13.151 1.56 0.6 1 AIMS-6 5.5 0.38 16.861 1.62 0.59 (r*)mel, mean radius of pores with r -= 2.5 nm; t,, mean statistical thickness of nitrogen monolayer; (plp"),, n, and relative pressure, adsorption and adsorption in primary mesopores at the end of the direct proportionality between n and nref;n,, monolayer capacity; (n*)mel, capacity of pores with r < 2.5 nm.where nitrogen molecules are adsorbed similarly on both the surface of MCM-41 sieves and the flat surface widens with increasing radius of the primary pores. The extent of the filling of primary pores by multilayer coverage of their walls is expressed by the ratio (ne)mel/(n*)mel(Table 4), while the difference 1 -(ne)mel/(n*)melgives the proportion that is filled by the capillary condensation.[In contrast to n, ,the amount of (ne)mel= relates only to the surface of primary pores.] From the first row of Table 4 it follows that with AIMS-1, primary pores are filled almost entirely by the formation of the monolayer. In contrast, with AIMS-2 and -3, both containing larger pores, the cooperative adsorbate-adsorbate interactions significantly manifest themselves. They cause the capillary condensation by which the pores fill spon- taneously in the reversible part of the isotherm. With TiMS-4, AIMS-5 and AIMS-6 the number of mesopores that are large enough for Kelvin capillary condensation with hys- teresis, gradually increases. Conclusions The nature of the adsorption in mesopores of MCM-41 molecular sieves strongly depends on their pore size._While with pores of radius of around 1 nm the multilayer coverage of the surface is effective, it turns into a two-stage process when the radius is increased to 1.5-1.8 nm. In these pores multilayer coverage is followed by capillary condensation without hysteresis. In even larger pores (r > 1.8 nm) the usual Kelvin capillary condensation takes place. For estimation of porous structure parameters, both the nature of the adsorption and the limits of the methods used J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 must be taken into account. The values of the pore structure parameters obtained provide detailed knowledge of these new materials. The authors are grateful to the Volkswagen Foundation for financial support (Grant 1/69 159). References 1 C. T. Kresge, M. E. Leonowicz, W. J. Roth, J. C. Vartuli and J. S. Beck, Nature (London), 1992,359, 710. 2 J. S. Beck, J. C. Vartuli, W. J. Roth, M. E. Leonowicz, C. T. Kresge, K. D. Schmitt, C. T-W. Chu, D. H. Olson, E. W. Schep-pard, S. B. McCullen, J. B. Higgins and J. L. Schlenker, J. Am. Chem. SOC., 1992,114, 10834. 3 C-Y. Chen, H-X. Li and M. E. Davis, Microporous Muter., 1993, 2, 17. 4 C-Y. Chen, S. L. Burkett, H-X. Li and M. E. Davis, Microporous Muter., 1993, 2, 27. 5 P. Behrens, Adu. Mater., 1993,5, 127. 6 P. Behrens and G. D. Stucky, Angew. Chem., 1993,105,729. 7 0. Franke, G. Schulz-Ekloff, J. Rathousky, J. Starek and A. Zukal, J. Chem. Soc., Chem. Commun., 1993,724. 8 P. J. Branton, P. G. Hall and K. S. W. Sing, J. Chem. SOC., Chem. Commun., 1993,1257. 9 S. J. Gregg and K. S. W. Sing, Adsorption, Surface Area and Porosity, Academic Press, London, 1982, pp. 94, 84, 153. 10 M. Ribeiro Carrott, P. Carrott, M. B. Carvalho and K. S. W. Sing, J. Chem. SOC., Faraday Trans., 1991,87, 185. 11 D. Dollimore and G. R. Heal, J. Appl. Chem., 1964,14,109. Paper 4/014971; Received 14th March, 1994

ISSN:0956-5000    

DOI:10.1039/FT9949002821

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

J. CHEM. SOC. FARADAY TRANS., 1994, 90(18), 2827-2835 IR Study of Ethene and Propene Oligomerization on H-ZSM-5: Hydrogen-bonded Precursor Formation, Initiation and Propagation Mechanisms and Structure of the Entrapped Oligomers Giuseppe Spoto, Silvia Bordiga, Gabriele Ricchiardi, Domenica Scarano, Adriano Zecchina" and Enzo Borello Dipartimento di Chimica lnorganica, Chimica Fisica e Chimica dei Materiali, Universita di Torino, Via Pietro Giuria 7,l-10125 Torino, Italy The oligomerization reaction of ethene and propene on H-ZSM-5 has been studied by fast FTIR spectroscopy. Oligomerization proceeds through : (i)formation of short-lived hydrogen-bonded precursors by interaction of the alkene with the internal acidic Brsnsted sites, (ii) a protonation step and (iii) a chain-growth step.The relative strength of the hydrogen bonds in the ethenMH and propenMH n-complexes (precursors) is estimated on the basis of the downward shift of both the v(0H) and v(C=C) frequencies (-389 and -11 cm-' for ethene and -539 and -19 cm-' for propene). For both molecules, the protonation of the precursors is the rate-determining step of the oligomerization process. The chain-growth mechanism and the structure of the entrapped oligomers are discussed on the basis of computer graphic and molecular dynamics simulations. Mainly linear or low branched products are formed whose length and structure is essentially determined by the steric hindrance imposed by the zeolitic framework. In a recent paper devoted to a IR and UV-VIS investigation of the proton-catalysed oligomerization of acetylene, methyl- acetylene and ethylacetylene in H-ZSM-5 channels,' we have shown that the acetylenic hydrocarbons form hydrogen-bonded precursors by interaction with the Brsnsted-acid sites of the zeolite, which are then slowly protonated and oligo- merized to give short conjugated carbocationic chains.The lifetime of the hydrogen-bonded precursors was sufficiently long to allow the detection of their vibrational spectrum. On the basis of the shift and half-width of the perturbed OH stretching bands it was possible to calculate the strength of the hydrogen bonds between the triple bond and the Brsnsted sites, and to conclude that there is a clear relation- ship between the strength and the rates of protonation and oligomerization.Methyl and ethyl group substitution also favours nucleophilic attack. The effect of the steric constraints, constituted by the three- dimensional array of intersecting channels typical of the MFI structure of ZSM-5, was also studied in detail, leading to the conclusion that the positively charged chains are character- ized by an average oligomerization number that is essentially dictated by the distance between the channel crossings. This leads to conjugated products whose main TC -,n* electronic transitions are located in the 30 000-25 OOO cm-range. On the basis of these results, it is conceivable that ethene and propene oligomerization (which are known to occur in H-ZSM-5 channel^^.^) could be preceded by the formation of hydrogen-bonded complexes and that the strength of the hydrogen bonds could be evaluated by measuring the pertur- bation of the OH and C=C stretching bands.Similarly, it is also expected that the three-dimensional array of the inter- secting channels should have a distinct effect on the length and on the branching of the positively charged oligomers. This effect could be studied in detail by means of IR spectros- copy, taking advantage of the prolific literature concerning the IR spectra of saturated hydrocarbons4 and by compari- son with the 13C NMR results.' In this paper we report the IR spectra, recorded under reaction conditions by in situ FTIR spectroscopy, of the hydrogen-bonded precursors and growing oligomers formed by interaction of ethene and propene with H-ZSM-5 zeolite.For propene, the hydrogen-bonded species are found to have a very short life (of the order of few seconds); the vibrational spectrum is hence investigated by using fast scanning condi- tions. Experimental The H-ZSM-5 samples used in these experiments were syn- thesized in the ENICHEM, Centro di Ricerche di Bollate, laboratories. Most of the experimental data reported here were obtained on samples with high external surface areas (crystallites with dimensions in the 20-50 nm range) and low (ca. 20) Si :A1 ratios. Full characterization of this material is given in ref. 6. Before they were dosed with ethene or propene, H-ZSM-5 samples (in the form of self-supporting wafers suitable for IR transmission measurements) were outgassed under high vacuum (< 10-Torr) at 673 K for 3 h to remove water and other adsorbed impurities. The thermal treatment was carried out in the same cell as was used for the transmission IR measurements.Ethene and propene (Matheson, high-purity grade), pre- viously purified by repeated freeze-pumpthaw cycles, were dosed from a vacuum line permanently attached to the IR cell. The IR spectra were recorded (at 4 cm-' resolution) on Bruker IFS48 or IFS88 FT instruments. For the fast acquisi- tion of the interferograms the Bruker LC/CG IR software package was used, running on an Aspect 1000 computer and allowing the recording of ca. 7 interferograms s-(at 4cm-' resolution).Computer graphics and modelling of the structure of the zeolite and of the entrapped oligomers, and molecular dynamic calculations were performed with software programs from Biosym Technology Inc. running on a SGI4D35 work-station. Results and Discussion H-ZSM-5-Ethene System As the IR spectrum of the H-ZSM-5-C2H4 system changes with time because of the occurrence of oligomerization, it is useful to divide the whole set of experimental data into three 2828 0.5 h UJ +d.-C 3 r; V." v _. 1600 1400al tm e ns 0.0 3800 3600 3400 3200 3000 2800 wavenumber/cm -' Fig. 1 IR absorbance spectra showing the formation of hydrogen- bonded precursors by interaction of ethene with H-ZSM-5.(-) Pure H-ZSM-5 outgassed at 673 K; (----) after dosage of C2H4 (5 Torr); (-.-.) effect of increasing the C2H4 pressure to 10 Torr. The spectra in presence of ethene were obtained for contact times 6 10 s. At the bottom the spectrum of C2H4 in the gas phase is shown for comparison. parts (A-C). In part A the spectra obtained immediately after dosage of two different amounts of ethene (and for a total contact time not exceeding 10 s) are considered, with the aim of elucidating the structure of the primary interaction pro- ducts (precursors). In part B the sequence of spectra recorded at 6.8 s intervals for a total contact time of ca. 1.5 min is reported and discussed. This sequence is expected to give information on the first stages of oligomerization and on the structure of the shortest oligomers.Finally, part C is devoted to the discussion of the sequence of spectra obtained for contact times in the 1.5 min to several hours interval, where chain propagation is supposed to be the predominant phenomenon. This sequence is expected to give information about the mechanism of chain propagation under the space restrictions imposed by the zeolitic framework and about the structure of the oligomeric chains under conditions approaching pore filling. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Hydrogen-bonded Precursors (Contact Timed 10 s) The IR spectrum of pure H-ZSM-5 (previously outgassed at 673 K for 3 h under high vacuum) is shown in Fig. 1, together with the spectra obtained at room temperature in the pres- ence of two doses (5 and 10 Torr) of ethene (and for a total contact time of ca.10 s). In the same figure the spectrum of pure ethene in the gas phase is also shown for comparison. The spectrum of the pure zeolite is characterized by a triplet in the OH stretching region at 3746, 3660 and 3609 cm-' which is assigned as follows:6 (i) 3746 cm-: v(0H) of Si-OH groups (silanols) mainly located on the external surface of the zeolite particles. This peak is particularly intense because of the small dimensions (20-50 nm) of the microcrystals constituting the zeolite sample used in this experiment; (ii) 3660 cm-': v(0H) of A1-OH groups in defective (partially extra lattice) positions; (iii) 3609 cm-': v(0H) of structural -Si-(OH)-Al- groups (structural Brrnsted-acid sites).Dosage of ethene results in the gradual disappearance of the 3660 and 3609 cm-' absorptions with simultaneous for- mation of two much broader bands (FWHM = 200 cm-') at 3369 and 3220 cm-'. Note that the process is accompanied by the appearance of a clear isosbestic point at 3560 cm-' (Fig. 1). Analogous experiments (results not reported in detail for the sake of brevity) performed on zeolite samples charac- terized by different Si:Al ratios, where the intensity ratio of the bands at 3660 and 3609 cm-' (outgassed samples) and consequently of the bands at 3369 and 3220 cm-' (samples contacted with ethene) is different, imply that the band at 3369 cm-' originates from perturbation of that at 3660 cm-' (with a downward shift of 291 cm-') and the band at 3220 cm-' from perturbation of that at 3609 cm-' (downward shift of 389 cm-').On all samples the 3746 cm-' absorption is hardly affected. As far as the spectroscopic features of adsorbed ethene (which are compared in Table 1 with those of the ethene mol- ecule in the gas phase') are concerned, the following com- ments can be made: (i) the vasym(CH2) (B2J and vsy,(CH2) (B1J modes (which are observed, respectively, at 3106 and 2990 cm-' in the gas phase) are shifted to lower frequency after interaction, giving rise to new bands at 3095 and 2974 cm-;a small downward shift from 1444 (gas phase) to 1440 cm-' (adsorbed state) is also found for the 6(CH,) (B1J mode which also gains intensity; (ii) the v(CC) (Alg) and Table 1 Most relevant spectroscopic features of H-ZSM-5 [v(OHj region], C2H4, C,H,, C2HJH-ZSM-5 and C,H,/H-ZSM-5 (in cm-'); owing to the uncertainty in measuring the positions of the bands, only a few frequencies are reported for the n-complex of C,H, C2H4-OH n-complex C,H,-OH n-complex H-ZSM-5 C2H4 (gas) C,H, (gas) V Ai V Ai assignment' 3476 3476 0 3660 3369 29 1 2609 3220 389 3106 308 1 3095 11 - 3012 - - 2990 2979 2974 16 - 2960 - - - 2916 - - -_ 2852 - - 1623 1647 1612 11 - 1472 - - - 1448 - - 1444 1416 1440 4 - 1399 - - 1342 1340 2 Ref. 6 and 7.J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 6(CH,) (Alg) modes, which are IR-inactive in the free mol- ecule and are responsible for two Raman lines at 1623 and 1342 cm-', become IR-active in the adsorbed state, giving two weak bands at 1612 and 1340 cm-l.Both the shift to low frequency of the OH stretching bands and the perturbation of the vibrational modes of adsorbed ethene can easily be explained on the basis of the formation of the 1 : 1 n complexes HZCTCH2 a In particular, owing to the rehdced symmetry of adsorued C2H, and to the reduced density of charge of the carbon- carbon double bond, structures I and I1 account for the IR activation and for the small downward shift (Av = -11 cm-') of the v(C=C) mode. Formation of hydrogen-bonded n-complexes was pre-viously evidenced by IR spectroscopy for the interaction of ethene with hydrogen halides*-" and with the acidic Brsnsted sites of Y zeolite^.^'-'^ In the latter case the shifts of v(0H) were in the range 370-350 cm-l.The higher value observed on H-ZSM-5 for complex I(389 cm-') is caused by the higher acidity of the Brsnsted sites in the MFI structure. The different acidity of the two zeolites is also demonstrated by the fact that, unlike H-ZSM-5, ethene is not oligomerized by H-Y zeoliteg (only weakly adsorbed species are formed). Note that under the temperature and pressure conditions 2829 adopted here, no formation of hydrogen-bonded complexes with the silanols (responsible for the band at 3746 cm-') is observed. This type of species is found only by lowering the temperature or by increasing the gas pressure (spectra not illustrated for brevity). This behaviour is fully consistent with the low acidity of silanols.Based on the extent of the v(0H) shifts upon ethene adsorption, the Brsnsted acidity is in the order : -Si(OH)Al-> -Al(OH)( b -Si-OH) in agreement with what was found by adsorption of CO at 77 K. It has been reported that the hydrogen-bonded species I are associated with a formation enthalpy of CQ. 38 kJ mol-'. Of course not all this enthalpy is related to hydrogen-bond formation. In fact, AH can be roughly split into two terms, AHa and AHb: AHa is associated with the pure hydrogen-bonding interaction (involving the Bransted sites and the C=C bond), while AHb is due to the van der Waals interaction mainly associated with the CH, groups.AHb can be estimated from the adsorption enthalpy of n-hexane5 and is ca. 11 kJ mol-'. We infer that the enthalpy associated with the pure hydrogen-bonding interaction is of the order of 16 kJ mol-', in reasonable agreement with the literature data concerning quantum-chemical calculations of the protonation of alkenes by acidic OH groups of iso- morphously substituted zeolite^.'^" Protonation Mechanism and Structure of the Shortest Oligo-mers (Contact Time 0-130 s) The IR spectra (recorded at the temperature of the beam and at intervals of 6.8 s) of the C,H4-H-ZSM-5 system for contact times between 0 and ca. 2 min are shown in Fig. 2. k h T0.03 ?its) 1 L I I I II r I I I1 I r I 14 3000 2900 2800 1600 1400 wavenumber/cm-' wavenumber cm-' Fig.2 IR absorbance spectra showing the initial stages of C,H, oligomerization on H-ZSM-5. The spectral sequence (1-19), obtained by recording one spectrum every 6.8 s, covers the contact time interval 0-130 s: (a) v(CH,) and v(CH,) region; (b) v(C-C), 6(CH,) and 6(CH,) regon. The spectrum of gaseous ethene is shown at the bottom for comparison. After ca. 7 s of contact (curve 1 in Fig. 2) the spectrum is characterized by the peaks, already discussed, due to the hydrogen-bonded precursors (bands at 2974, 1612, 1440 and 1340 cm-'). In the successive spectra new weak features, assigned to saturated CH, and CH, groups, develop at 2960 and 2876 cm-l [v,,,,(CH,) and vsym(CH3)], 2940 and 2866 cm-' [vasym(CH2) and vsy,(CH,)], 1460 and 1382 cm-' [G,,,,(CH,) and G,,,(CH,)] and 1469 and 1442 cm-' [dasymm(CH2)and 6,,,(CH2)].The appearance of the charac- teristic modes of saturated (CH,) and CH,) groups is a clear indication that a protonation and polymerization process occurrs following the Scheme 1, where (because of the rela- H2C =;'CH2 I Y Y Scheme 1 tively small amount of defective Al-OH groups) only chains growing on Si-(OH)-A1 sites are shown. C,H, stands here for the possible isomers illustrated in Scheme 2, which are in chemical equilibrium. Similarly, C,H ,represents different isomers in chemical equilibrium, and so on. In agreement with ref. 14-18 structures (a), (b), (c) and (d)have deliberately \si.o.T' 0 been written in covalent form, even if a certain degree of ion- icity cannot be excluded (which could be enhanced by the 'solvating' effect of the zeolite framework). As already pointed out by Van den Berg et al.,' because of the con- straints imposed by the zeolite superstructure, the distribu- tion (in chemical equilibrium conditions) of the possible isomers in the channels is not necessarily identical to that observed in the gas phase or in solution (where the more branched and energetically more stable isomers are preferred). Detailed information concerning the reaction mechanism and the nature of the products can be gained by closer inspection of the spectral sequence of Fig. 2. From this, we can first infer that the number of oligomeric chains formed in the time interval considered here must be very small, as demonstrated by the negligible decrease of the bands of the hydrogen-bonded precursors (see in particular the peaks at 1612, 1440 and at 1340 cm-l).Furthermore, if the very early stages are considered (spectra 1-4 in Fig. 2 corresponding to contact times between ca. 7 and ca. 30 s) it can be concluded that as far as the first-formed products are concerned, there is no sign of -CH(CH,), and -C(CH,), chain branching (which should contribute an extra, characteristic peak at ca. 1368 cm-').19 It is also most noticeable that in the sequence J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 of Fig. 2, the intensity ratio I(CH,)/I(CH,) of the (CH,) and (CH,) stretching modes is constantly greater than 1.As the specific intensity of the (CH,) stretching modes is known to be always larger than that of the (CH,) groups,20-22 this observation [together with the absence of branched products containing -CH(CH,), and/or -C(CH,), groups] indicates that the only species formed in the 7-30 s time interval are small, linear -0-C,H,,+ oligomers. Moreover, as I(CH,)/I(CH,) is between 1 and 2, it is also inferred that structures of type (c) are less important. Whether spectrum 2 of Fig. 2 corresponds to a -0-CH,CH, species only or, more probably, to a mixture of -0-CH,CH, plus linear -0-(CH,),CH, species cannot be stated with confidence because a true reference compound is not available and because the considerations based on the intensity ratios must always be used with caution.In any case, due to the predominace of short, linear structures the 7-30 s interval can be considered as essentially characterized by the protonation reaction (initiation step). In principle the -O-c& species could also be under the branched forms (b) and (4,formed through isomerization of linear species. The experimentally ascertained absence of branched forms of the -O7c4H9 dimer in the very early stages of oligomerization implies that the isomerization rate is lower than the protonation (initiation) rate (as we shall see, the propagation rate is, however, larger). This observation is intriguing: at least for the smallest oligomers formed, for instance, at the channel intersections, no space restrictions limiting the isomerization rate to the (b) and (d) type oligo- mers are expected to exist.A plausible explanation for the experimentally observed low propensity to form branched oligomers can be given, however, if partial covalent character of the oligomers is considered. While it is well known that fully ionic C4H9+ species should readily give the (CH,),C+ carbocationic species, the same does not hold for more cova- lent structures, even in absence of steric constraints. In con- clusion, the experimental results essentially confirm those obtained from quantum-chemical ~alculations,'~-' suggest- ing a high degree of covalent character of the C-0 bond. In the section devoted to propene polymerization we shall add further arguments in favour of the hypothesis that the rate of chain propagation is larger than the rate of protonation (initiation) and isomerization.Based on these spectroscopic results, the evolution of the C,H,-H-ZSM-5 system in the 0-30 s interval is graphically represented in Plate 1 (where the zeolite framework is oriented along the [OlO] crystallographic direction). Con- cerning the zeolitic structure, in Plate 1 we have chosen a representation which is informative about the internal space accessible to molecules moving inside the channels and where the crystallographic features of the zeolitic lattice are omitted. This goal was achieved using the Connolly algorithm,, which allows the 'Connolly surface' to be constructed.This surface represents the boundary of the volume from which a probe molecule (H,O in this case) is excluded if it experiences van der Waals overlap with the atoms of the zeolite frame- work. In more detail, Plate l(a) shows the situation before admission of ethene (pure zeolite): the accessibility in the intersecting straight and sinusoidal channels (running along the [OlO] and [lo01 directions, respectively) of the Brnmsted- acid sites (represented as red balls for which the Li+ van der Waals radius was adopted) is evidenced. Plate l(b) shows the complete transformation of the Brmsted sites into the hydrogen-bonded 1 : 1 complexes through interaction with ethene. Finally, in Plate l(c) the situation after ca.30 s of contact is shown. In this last picture randomly distributed, short, linear -C,H, and -C,H, oligomers growing in the straight and sinusoidal channels are seen, together with the J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Plate 1 Computer graphic models showing: (a)the H-ZSM-5 channel array (Connolly representation) and distribution of Brmsted-acid sites (red balls); (b) the formation of 1 : 1 n-complexes upon interaction with ethene; (c) the formation of entrapped short and linear protonated species G. Spoto et al. (Facing p. 2830) J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Plate 2 Computer graphic model illustrating the location (as obtained by molecular dynamic calculations) of: linear (A), containing terminal isopropyl groups (B), and (CH,),CH oligomers (C) inside the silicalite framework Plate 3 Computer graphic simulations showing the distribution inside the H-ZSM-5 framework of unreacted precursors and oligomerization products after 130 s of contact with ethene.(a)View along the [OlO] crystallographic direction; (b) view along the [loo] direction J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 2831 vast majority of unreacted hydrogen-bonded precursors (in agreement with the experimental observations). The implicit assumption used to construct the previous representations is that the hydrogen-bonded precursors are all equivalent, because they are formed through interaction with Br~rnsted sites of very similar acidity. Following this simplified hypoth- esis, the fact that a very small fraction of these complexes has evolved into protonated species in this time interval is simply due to the kinetic effects associated with the high activation barrier for the protonation process and not to the existence of a small fraction of more acid sites.When the contact time increases beyond 30 s (Fig. 2, curves 5-19), the intensity of the v(CH,) and v(CH,) bands grows proportionally. Moreover, a weak peak at 1366 cm-' is observed to emerge from the background: this is a clear indi~ation'~that, as well as the formation of trimeric and longer oligomers of the linear type, branching occurs with possible formation of species like (e) and (f).Based on simple (n= 0,l .......) (n = 0,l .......) (8 1 (f 1 geometrical considerations, one suspects that the growth in the H-ZSM-5 channels of oligomers containing terminal iso- propyl (e) and tert-butyl (f) groups can suffer in different way from the space restrictions imposed by the framework (because of their different dimensions compared to the channel diameter). To clarify this point we decided to invest- igate the mobility (at different simulated temperatures) of en- trapped hydrocarbons of the type CH,(CH,CH,), CH(CH,), and (CH,),CH (as models of those formed inside the H-ZSM-5 channels) by means of molecular dynamics calcu- lation~.,~For the sake of simplicity, silicalite (i.e.a pure sili- ceous material having the same structure and channel diameter as ZSM-5) was chosen as the model for the zeolite structure.The result was that, while there are no constraints limiting the diffusion of isomers containing the isopropyl group (or less branched species), the (CH,),CH species cannot enter the channels and can only be located at the channel intersections, where it almost completely fills the available space. A graphical representation of the results obtained by molecular mechanics (by considering a fixed zeolite framework and by allowing the different neutral isomers to reach the lowest-energy configuration) is shown in Plate 2. It follows that the -0-C(CH,), species derived from isomerization of the -0-C4H, moieties, once formed cannot add further ethene molecules. On one hand they com- pletely fill the space at the channel intersections (preventing further C2H4 molecules from reaching the catalytic site) and, on the other hand, they are too large so they cannot grow and penetrate the channels.In other words, the only possible value of n for the hypothetical -0-(CH,CH,),C(CH,) species (f)is zero and the entrapped 0-C(CH,), species behave as dead oligomers. By comparison of the intensities of the v(CH,) and v(CH,) bands in the stretching and bending regions (Fig. 2) with those of model hydrocarbons,20-22 we infer that the oligo- mers formed for contact times in the 20-100 s interval are characterized by a CH,: CH, ratio in the (2-4) : 1 range. Although this stoichiometry is very approximate, some general conclusions can be derived.For instance, if we con- sider that the CH, :CH, ratio in isobutyl and isopentyl struc- tures is, respectively, 1 : 2 and 2 : 3 (in comparison with the value 2-4 : 1 observed experimentally), it is concluded that not only are the (CH,),C-groups, possibly formed at the channel intersections (and on the external surface as well), small or negligble in number, but also the oligomers (e) con-taining the isopropyl group must represent a minor fraction. After ca. 130 s of contact (considered somewhat arbitrarily as the time interval where initiation is more important than propagation) CH,:CH, finally reaches 4-5 : 1 [Fig. 2(a), curve 191. Note that the number of chains (equivalent to the number of hydrogen-bonded precursors consumed) formed at this time must still be very small, as demonstrated by the small (ca.10%) decrease in the 1612 cm-' band (characteristic of the ethene-OH n-complex). When the overall integrated intensity of the bands due to the stretching vibrations of the CH, and CH, groups (3050-2800 cm-') is plotted as a function of time, Fig. 3 is obtained. The integrated intensity in the 1395-1355 cm-' range is also shown in Fig. 3. Note that in the latter region only the CH, bending modes contribute; hence from the examination of this spectral interval we can obtain an insight as to the time evolution of the CH, groups. As these groups give a qualit- ative indication of chain termination, and hence of the approximate number of chains formed, from Fig.3 it can be inferred that in the 0-130 s time interval not only do we have growth of the oligomeric chains but also the parallel and con- tinuous formation of new species through the protonation of the unreacted hydrogen-bonded precursors. Based on the previous discussion, the situation after 130 s is represented graphically in Plate 3 (as obtained using the correct van der Waals surfaces of the polymeric chains). The following comments can be made: (a) notwithstanding the small dimensions of zeolitic pores and the sinusoidal behav- iour of some of them, linear polymeric chains can grow freely in the straight and the sinusoidal channels without any hin- drance. Polyethylene is flexible enough to follow any channel shape. (b) As noted previously, branched oligomers contain- ing the -CH(CH,), group fit exactly the space available in the channels, while the bulkier tert-butyl residues can be accommodated only at the channel intersections.Owing to the experimentally ascertained low branching rate, the con- centration of these species is, however, very low (and corres- pondingly only one of these groups is reported in the simulations for the sake of illustration). (c) The small channel dimensions prevent the presence of more than one chain inside the same channel portion. This means that when a living polymeric chain reaches a channel intersection where another polymerization centre is functional, it must stop its growth. This observation is in agreement with the spectro- scopic results which indicate the presence of chains with a limited number of carbon atoms.(d)When the growing oligo- meric chains reach portions of the framework where unre- acted precursors are present, owing to the limited space available they can interact strongly with the n-bonded ethene molecules. As will be shown in the following, this results in ethene being displaced from the complex. Structure of the Oligomers after Prolonged Contact Time The evolution of the spectra for prolonged contact time is illustrated in Fig. 4. The most relevant results are: (i) the bands associated with the (CH,) groups gradually take over and become the predominant features [in both the C-H stretching, Fig. *a), and bending, Fig. 4(b), regions]. At the end of the time interval shown in Fig.4, the intensity ratio between the CH, and CH, bands points towards an average CH, :CH, z 7-8 : 1 stoichiometry. This implies that if only linear polymers are considered, the average chain is made of ca. eight carbon atoms. It is interesting to note that a chain containing this number of carbon atoms almost exactly fills the space between two channel crossings (Plate 3). The pre- J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 2.0 1.5 r I 5 m m m v mV m 1.0 2 ms? 0 -0 +-h 13, +-. .-0 3.5 3I I 0 0.10 0.20 0.30 0.401 42 43 4 Fig. 3 Integrated intensity of the IR bands in the 3000-2800 and 1395-1355 m-’ regions as a function of the ethene-H-ZSM-5 contact time I 1I.. i . I I 3600 2900 2ioo 1600 1400 wavenum ber/cm -’ wavenumber/cm-’ Fig. 4 IR (absorbance) spectra of the ethene oligomerization products formed over a prolonged contact time. The spectral sequence covers 1:he 40(spectrum 1)-1040 (spectrum 26) s interval. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 vious hypotheses on the length of the chains have been made without considering the possible presence of isopropyl and tert-butyl terminations [the latter, containing three CH, groups, could account for a considerable fraction of the I(CH,) intensity even when present in very small amounts]. Consequently, the figures previously given for the chain lengths must be considered as underestimated. (ii) The rate of growth of the bands in the 3050-2800 cm-range (mainly due to CH, formation) decreases gradually (Fig.3 and 4); this is a clear indication of the progressive filling of the empty space where the oligomeric chains are forced to grow and is in agreement with the previously advanced hypothesis that the average number of carbon atoms in the chains is ca. eight. The growth of the chains is necessarily stopped at the channels intersections because they encounter either other growing oligomers or dead species that fill the voids at the intersections (as will be shown in the following, the number of sites at which protonation and oli-gomerization occur is ca. 20% of the total). (iii) The IR bands of the hydrogen-bonded precursors diminish progressively and at the end of the experiment they are reduced to ca.40% of their original value [see the intens-ity of the bands at 1612, 1440 and 1340 cm-' in Fig. 4(b)]. The decrease of concentration of the hydrogen-bonded pre-cursors is also demonstrated by the changes occurring in the OH stretching region illustrated in Fig. 5. As the reaction continues the absorption centred at 3220 cm-' gradually declines in favour of a new broad and complex band with a maximum at ca. 3470 cm-'. As will be demonstrated later, this new absorption is due to OH groups interacting with the polymeric chains. The residual presence of a considerable quantity of unre-acted hydrogen-bonded precursors even after prolonged contact is demonstrated by the results shown in Fig.6, where the effect of removing the gas phase is illustrated. When the sample is outgassed at room temperature, the bands charac-teristic of the n-complex at 1612, 1440 and 1340 cm-' and the broad band at 3220 cm-' disappear rapidly while the peak at 3609 cm-', characteristic of the unperturbed Si(0H)Al groups, simultaneously shows up. In addition, the spectrum in the OH stretching region shows a novel feature at ca. 3470 cm-' (i.e. at the same frequency as the band appearing after prolonged oligomerization time). The origin of this band can be readily understood if consideration is made of the results of the adsorption experiment of a saturat-I 10.0 , I ,J 3800 3600 3400 3200 wavenumber/crn -' Fig. 5 IR (absorbance) spectra illustrating the changes occurring in the v(0H) region during ethene oligomerization (full lines).Broken line: spectrum of pure H-ZSM-5. 2833 0.71 II 1 1I 1I V." 3800 3600 3400 3iOO wavenumber/cm -' Fig. 6 Effect of outgassing the ethene-H-ZSM-5 system at room temperature. Curve 1: after 1040 s of contact. Curves 2-4: after out-gassing for 1 (2), 5 (3) and 30 (4) s. Broken line: spectrum of pure H-ZSM-5. ed hydrocarbon (n-heptane) on H-ZSM-5 illustrated in Fig. 7. We note that upon heptane adsorption: (i) the band at 3609 cm-' (bridged OH groups with Brsnsted-acid character) is shifted to CQ. 3474 cm-(Av = -135 cm-') with broadening (FWHM x 160 cm-'); (ii) in the presence of an excess of n-heptane (spectrum not shown), the band at 3746 cm-' (silanols) is shifted to 3700 cm-' (Av = -46 cm-') and is broadened (FWHM z 100 cm-').This experiment undoubtedly shows that saturated hydrocarbons perturb the Si(0H)Al and -SOH oscillators and that the effect in term of induced shift and band broadening (change of half-width) is larger for the more acidic group (as expected). As the fre-quency of the OH groups perturbed by n-heptane is identical to that observed after polymerization, this experiment also demonstrates that the saturated hydrocarbon chains growing inside the zeolite channels interact with the framework acidic OH groups, displacing the weakly adsorbed C,H, molecules. The process can be represented as follows: H2CICH2 I From this observation two further important conclusions can be drawn.First, the observed decrement of the band of the hydrogen-bonded precursors (1612 cm-I) observed when the ethene is left to stand in contact with H-ZSM-5 for a long time is not totally due to the transformation of the n-complexes into the protonated and oligomerized form. Sec-ondly, by comparison of the intensity of the residual peak at 3474 cm-' with that obtained in the blank experiment with n-heptane (where all the Brnrnsted sites are perturbed) we can infer that, contrary to the initial expectations, a minority (only ca. 20%) of the Brnrnsted-acid sites have undergone protonation, so acting as a true catalytic centre where chain initiation and propagation are occurring. This also demon-strates that the oligomeric chains formed on one-fifth of the Brnrnsted sites are sufficiently long to perturb nearly all the other (unreacted) Brnrnsted sites, displacing adsorbed C2H4 and decreasing the concentration of precursors. The decrease 1.! hIn c..-C 4 v a, C -e2 a 0.0 3800 3600 3400 3200 3000 2800 wavenumber/cm-’ Fig.7 IR (absorbance) spectra showing the perturbation of the v(0H) bands of H-ZSM-5 upon dosage of n-heptane.Broken line: pure H-ZSM-5. Full line: after dosage with n-heptane. in concentration of the precursors induced by pore filling also justifies the decrement of the oligomerization rate. Finally, this confirms the view that protonation is the slowest process and that a high activation barrier is present.Propene-H-ZSM-5 System When the sample is contacted with propene (4 Torr) at room temperature and the IR spectrum is recorded following the T A J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 procedure already described for the ethene-H-ZSM-5 system (i.e. with acquisition times of the order of 7 s), strong bands in the 3050-2800 cm-’ range due to CH, and CH, groups of saturated oligomeric species are the only features observed. This indicates that protonation and oligomerization are so fast that the hydrogen-bonded precursors and protonated (initiation) intermediates escape observation. In order to detect these short-lived intermediates, we decided to record the spectra following the Bruker GC/LC (gas/liquid chromatography) fast acquisition program which allows the collection of 7 interferograms per second at ‘4 cm-’ resolution.The results are shown in Fig. 8 (where each spec- trum is the average of 10 interferograms). Only the difference spectra (obtained by using the spectrum of the pure zeolite as background) are reported : in this representation the bands appearing as negative peaks belong to species which are con- sumed, while those appearing as positive peaks belong to species which are formed during the adsorption and oligo- merization. The main results emerging from this experiment are: (i) the hydrogen-bonded precursor ti& ICHCH3 I is characterized by v(0H) = 3070 cm-’ (with a downward shift of 539 em-’ with respect to the unperturbed species) 17.5x 10” (arb.units) I n 1 L I \I I I 1 I 1t . 3600 3200 2800 1600 1400 wavenumber/cm-’ Fig.8 IR (absorbance, background-subtracted) spectra illustrating propene (initial pressure 4 Torr) oligomerization on H-ZSM-5. The spectral sequence covers the 0-90 s time interval. The time interval between two successive spectra is 1.5 s. Each spectrum corresponds to the average of 10 interferograms. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 strate that teristic of the band at 3470 cm-', the CH, -CH(CH,), meric chains of the type siderations are based hydrogen formed similar acidity). All the HM of 420 cm-'. These dat 2835 and a FW a undoubtedly demon- grams to investigate transient species residing on the surface the hydrogen bond formed by interaction of for times not longer than tenths of a second.This timescale with the Brsnsted-acid sites is stronger th an that indicates that the observed time evolution is associated with y interaction with ethene (as etive effect of the CH, group) xpected on the basis of kinetic effects only and that this spectroscopy cannot be con- the inducpropene formed b . This in turn explains sidered as time-resolved in the usual sense (i.e. on the time- this experiment. A strong the higher (with respect to ethene) pro perturbation tonation rate sition of the b is also charac- scale dictated by the molecular and electronic motions). inferred fto the v(wards w rom the examination of the po and due Conclusions C=C) mode of the precursors, ith respect to the free molecule (Av = -19 cm-l) di which is shifted down- The three steps in the oligomerization of ethene and propene on H-ZSM-5 (formation of hydrogen-bonded precursors, ).(ii) The hydrogen-bonded w seconds (Fig. 8). This is d precursors sappear ue both to their con- protonation and chain propagation) can be studied individ- after a fe(Table 1displacembefore). (the avera protonation-oligomerization) and to ually by IR spectroscopy provided that a sufficiently fast sumption (caused by ated by the gr by the chains growing owth of acquisition rate is used (of the order of 7 s for ethene and 1.5 s for propene). characteristic annels (as is clearly demonstrd by the saturated hydrocarboent of the propene molecules in the chperturbe of the Brsnst ed sites n chains, as discussed The relative strength of the hydrogen bonding in the C,H,-OH and C,H,-OH n-complexes (precursors) as deter- iii) At the end of the experim ent (i.e.after c a. 90 s), mined by the shift of \(OH) and v(C=C) is much higher for ge CH, :CH, ratio (as estima ted by the inte nsity of propene (because of the inductive effect of the methyl groups). and CH, bands) is ca. 1 : 1 and no signs of This also accounts for the much higher oligomerization rate. and -C(CH,), groups aby the absence of the characteristic ban re observed (as argued d at ca. 1366 c m-'). A For both molecules, the slowest step of the reaction is found to be the protonation of the precursors. o is almost exactly the same 1 : 1 rati as expected f or poly- The steric constraint imposed by the three-dimensional array of channels of the zeolite framework strongly influences the nature (degree of branching, distribution and length) of (CH2CHCH3),CH2CHI 2CH, the products.References be used 1 S. Bordiga, G. Ricchiardi, G. Spoto, D. Scarano, L. Carnelli, A. Zecchina and C. Otero Arean, J. Chem. SOC., Faraday Trans., the ethene case, the methyl g of the modes of the terminal groups. groups are effroups cannot . This means tthe intensity of the band at ca. 3470 cm-' due to the stretching modes of the unreacted groups perturbed by saturated chains can be inferred that also in this case only (ca. 25%) of Brsnsted-acid Unlike here as indicators of chain terminationnot possible to estimate the length of by means From sites) it fraction the oligomeric chains (which is initiated at different ectively hat it is Brsnsted-acid a small 2 3 4 5 6 1993,89, 1843.J. R. Anderson, T. Mole and V. Christov, J. Catal., 1980,61,477. V. Bolis, J. C. Vedrine, J. P. Van den Berg, J. P. Wolthuizen and E. G. Derouane, J.Chem. Soc., Faraday Trans. I, 1980,76,1606. L. J. Bellamy, The Infrared Spectra of Complex Molecules, Chapman and Hall, London, 2nd edn., 1980, vol. 11. J. P. Van den Berg, J. P. Wolthuizen, A. D. H. Clague, G. R. Hays, R. Huis and J. H. C. Van Hooff, J. Catal., 1983,80, 130. A. Zecchina, S. Bordiga, G. Spoto, D. Scarano, G. Petrini, G. Leofanti, M. Padovan and C. Otero Arean, J. Chem. SOC., in protonation.In other wordte is increased greatly with rethe internal available space of the protonic sites have ef spect to ethene s, although the proto- is already filled when fectively been involved , it is so 7 Faraday Trans., 1992,88,2959, and references therin. K. Nakamoto, IR and Raman Spectra of Inorganic and Coordi- nation Compounds, John Wiley, New York, 4th edn., 1986; G. Herzberg, Infrared and Raman Spectra of Polyatomic Molecules, D. Van Nostrand, New York, 1947. in protonengaged nation rahigh that only 25% case the previous con- 8 A. J. Barnes, J. B. Davies, H. E. Hollam and J. D. R. Howells, J. on the implicit precursors are all equivalation. Of course, also in this assumption te hat the nt (because they are 9 Chem. SOC.,Faraday Trans. I, 1973,69,246.L. Andrews, G. L. Johnson and B. J. Kelsell, J. Chem. Phys., through interaction with B rsnsted sites of very 1982,76,5767. 10 L. Andrews, G. L. Johnson and B. J. Kelsall, J. Chem. Phys., observations made are in mechanism shown in Scheme 3. The exFig. 8 also demonstrates the utility o agreement with periment illustrated the f fast acquisition pro- in 11 12 1982,104,6180. V. B. Liengine and K. Hall, Trans. Faraday SOC., 1966,62,3229. L. Kubelkova, J. Novarova, Z. DolejSek and P. Jiri, Collect. Czech. Chem. Commun., 1980,45,3101. 13 14 15 16 N. W. Cant and K. Hall, J. Catal., 1972,25, 161. P. Viruela-Martin, C. M. Zicovich-Wilson and A. Coma, J. Phys. Chem., 1993,97,13713. V. B. Kasansky, Acc. Chem. Res., 1991,24, 379. I. N. Senchenya, N. D. Chuvylkin and V. B. Kazansky, Kinet. Catal., 1985,26,926. CH2CHCH2CH2CH3 17 18 19 1. N. Senchenya and V. B. Kazansky, Kinet. Catal., 1987,28,490. V. B. Kazansky and I. N. Senchenya, J. Catal., 1989,119,108. N. B. Colthup, L. H. Daley and S. E. Wiberly, Introduction to Infrared and Raman Spectroscopy, Academic Press, New York, II H&=CHCHs 1975. CH3 CH3 I I CH2CHCH2CHCH2CH2CH3I \,SiCo'Al< I I H&=CHCH, - ............ 20 21 22 23 24 S. A. Francis, J. Chem. Phys., 1950,18,861. S. H. Hastings, A. T. Watson, R. B. Williams and J. A. Anderson Jr., Anal. Chem., 1952,24,612. H. Luther and G. Czerwony, Z. Phys. Chem., NF, 1956,6,286. M. L. Connolly, Science, 1983,221, 709. Insight 11 and Discover User Guides, Biosym Technology, San Diego, 1992; P. Dauber-Osgurthorpe, V. A. Roberts, D. J. Osgurthorpe, J. Wolff, M. Genest and A. T. Hagler, Proteins: Struct., Funct. Genet., 1988, 4, 31. Scheme 3 Paper 4/01492H; Received 14th March, 1994

ISSN:0956-5000    

DOI:10.1039/FT9949002827

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

J. CHEM. SOC. FARADAY TRANS., 1994, 90(18), 2837-2844 Acid Properties of a ZSM-20-type Zeolite Hendrik Kosslick,* Heinz Berndt, Hoang D. Lanh, Andreas Martin, Hans Miessner and Vu Anh Tuan Institute fur Ange wandte Chemie Berlin-Adlershot (ACA), 0-12484 Berlin, Rudo wer Chaussee 5, Germany JochenJanchen Schuit Institute of Catalysis, TU Eindhoven, Postbus 513,5600 MB Eindhoven, The Netherlands The acid properties of an HZSM-20 zeolite have been characterized, in comparison to an HY zeolite, by IR spectroscopy, temperature-programmed desorption (TPD) of ammonia, time-resolved temperature-programmed FTIR of ammonia desorption and by microcalorimetry. Also, catalytic properties were tested for n-hexane con- version. These studies show that for HZSM-20 there is a shift of the bridging OH band to lower wavenumbers, an increase in the ammonia desorption temperature and a higher initial heat of ammonia chemisorption compared to HY.Thus, HZSM-20 contains medium-strong Brsnsted sites, ranging in strength between zeolites HZSM-5 and HY. Obviously, the higher Si : Al ratio of HZSM-20 results in an enhanced number of strong acid sites with no second-neighbour aluminium atoms. This leads to a change in the distribution of Brsnsted-acid sites. Superacid Brsnsted sites could not be detected. Furthermore, the IR spectra of adsorbed pyridine and ammonia revealed the presence of at least two kinds of Lewis-acid sites. Some Lewis sites of HZSM-20 exhibit a weak acid strength and the strong Lewis sites are weaker than those of HY.The different acidic properties of both zeolites are reflected in their catalytic activity and selectivity in n-hexane conversion. The synthesis of ZSM-20 was reported in 1976,' but this zeolite type has become the subject of intensive studies only recently because it has potential applications in acid-catalysed reactions of hydrocarbons, such as the FCC process.2-8 The structure of ZSM-20 is related to faujasite-type zeolites (FAU).' ZSM-20 is now accepted to be an intergrowth of blocks of FAU and of EMT framework topology. EMT differs from FAU in its stacking." Additionally, differences in the T-0 bond length and T-O-T angles have been sug- gested.'' Until now, little information has been available concern- ing the acidic properties of zeolite HZSM-20 and related material^.'^-'^ Very strong Brsnsted-acid sites are proposed to be the origin of the high activity of this material. Superacid Brsnsted sites could originate from the zeolite structure or from an interaction between Lewis sites (cationic non-framework aluminium species) and Brsnsted sites.' 3*1 It has also been suggested that the high catalytic activity arises from a cooperative action of Brsnsted and Lewis sites.16 Recently, we reported the synthesis, dealumination, physi- cochemical and catalytic properties of a ZSM-20-type zeolite.' 7,18 This material exhibited a distinctly higher cata- lytic activity in the isomerization of rn-xylene and the conver- sion of ethylbenzene compared to the HY zeolite usually used. The reason for this difference is still sought.The aim of this study was to examine the nature, strength and concentration of acid sites of this material using methods such as FTIR spectrocopy using probe molecules, TPD of ammonia and microcalorimetry (differential heat of ammonia chemisorption) in order to clarify the origin of the strong acidity. Experimental Materials Zeolite ZSM-20 was synthesized under hydrothermal con-ditions at 373 K over 14 days from a reaction gel of the following composition: 1.25 Na,O * 1 A1203-30.2 SiO, 26.4a TPAOH 268 H,O. The synthesis product was withdrawn, washed repeatedly with distilled water until neutral and then dried at 393 K. The sample was then ion-exchanged with a 0.5 mol dmP3 solution of NH4N03 at 333 K (twice for 3 h).92% ion-exchange was reached, as determined by AAS." Subsequently, the sample was calcined in air at 773 K for 2 h under shallow-bed conditions (heating rate /3 = 10 K min-') to generate the protonic form. The crystallinity of the product was characterized by XRD to be a well crystallized HZSM-20 zeolite, as described in the literat~re.'~-~, The resulting Si :A1 ratio was cu. 4.3, as determined by 29Si MAS NMR.18 The HY sample (for comparison) was prepared from an industrial NaY zeolite (supplied by CKB Bitterfeld). The latter was treated in the same way as described for the HZSM-20 sample. 85% ion exchange was reached, which is typical.23 The Si : A1 ratio was cu. 3." Highly dealuminated samples of HZSM-20 and HY were included in the investigations for the special IR studies described below.The dealumination procedure of the samples has been described previously.' 8924 The samples were impreg- nated with an ethanolic solution of RhCl,, dried at 393 K in air and calcined subsequently in air at 573 K. The resulting solids contain 1 wt.% Rh. Characterization Methods The IR spectroscopic investigations of the samples were carried out on a Bruker IFS 66 FTIR spectrometer using self- supporting wafers of diameter 20 mm and weight 30 mg. All wafers were pretreated by heating (p = 10 K min-') to 673 K for 1 h under vacuum Pa). The OH vibrational spectra were recorded in the range 3400-3800 cm-'. Additionally, the IR spectra of adsorbed pyridine were recorded in the 1300-1700 cm-' region to discriminate between Brsnsted sites and Lewis sites.Also, the thermal desorption of pyridine was observed by IR spectroscopy to differentiate the strength of different acid sites.25 The CO stretching vibration of rhodium(1) dicarbonyl ions [Rh'(CO),] located on the Brsnsted sites of the dealuminat- + ed and Rh-loaded samples was used as a sensitive probe for the characterization of the strength of Brsnsted site^.^^,^^ CO (1.33 kPa) was adsorbed at 423 K and the spectra were recorded at room temperature (rt) after evacuation Pa). The TPDA experiments were carried out in a conventional flow system with thermal conductivity detection. Pressed zeolite samples of 150 mg were placed in a U-tube reactor. The samples were calcined for 1 h by heating them to 773 K (/3 = 10 K min-') under flowing helium.After cooling the samples to 373 K they were saturated with ammonia by passing a helium flow (2.5 dm3 h-') containing 2.7 vol.% ammonia until no more ammonia was adsorbed (checked by a thermoconductivity cell). Thereafter physically adsorbed ammonia was carefully removed by passing a dry helium flow through the reactor (2.5 dm3 h-' for 2 h) at the same tem- perature. Then the TPDA profiles were recorded whilst increasing the temperature to 823 K (/?= 10 K min-'). The final temperature was held for 45 min to complete the removal of ammonia from the zeolite. The reproducibility was better than 3%.The FTIR-NH,TPD measurements were carried out on a BIORAD FTS 60A spectrometer at a resolution of 2 cm-', accumulating 256 scans. The adsorption of ammonia was performed after an activation of the self-supporting wafers up to 873 K in vacuum Torr). For quantitative evalu- ations the spectra were normalized using the Si-0 overtone as an internal standard. The differential heats of ammonia chemisorption were measured at 423 K using a Calvet-type microcalorimeter (Setaram) which was connected to a standard adsorption apparatus. The samples (900 mg) were calcined at 673 K for ca. 15 h under vacuum (p < 1 mPa). The approach of the adsorption equilibrium was controlled by pressure measure- ment and followed by the thermokinetic curve.n-Hexane conversion over HZSM-20 and HY was carried out in a flow reactor under atmospheric pressure at 773 K with a contact time of W/F= 0-24 g h mol-'. The catalyst was activated in the reactor at 773 K in a dry flow of nitro- gen for 2 h. The reaction products were analysed by gas chro- matography. An alumina column was used for the analysis of cracking products and a Beton 34 column for the analysis of aroma tics. Results and Discussion The OH vibrational spectra of zeolite HZSM-20 and HY are shown in Fig. 1. The appearances of the OH vibrational spectra of HZSM-20 and HY nearly coincide. The spectrum of HZSM-20 reveals three main absorption bands at 3735, 3632 and 3552 cm-'. By analogy with faujasites they are assigned to the vibrations of silanol groups and of the Brsnsted-acid bridging OH groups in the large cavities and in the hexagonal prisms (high-frequency and low-frequency band), respectively.* 7-29 In comparison to zeolite HY the high-frequency band of HZSM-20 is shifted by 8 cm-' to lower wavenumber, indicating a higher acid strength of Brsnsted sites in HZSM-20.In contrast, the low-frequency band is shifted by 6 cm-' to higher wavenumber, indicating a decreased acid strength of weaker acid sites. In conclusion, the Brsnsted sites of HZSM-20 seem to be more differen- tiated with respect to strength in comparison to HY. Similar changes were observed in the OH vibrational spectra of dea- luminated HY zeolites,24 suggesting that these changes are probably due to the increased Si :A1 ratio of HZSM-20 in comparison to HY.30 The IR spectra of pyridine adsorbed on HZSM-20 and HY at 453 K are shown in Fig.2. Two absorption bands were observed, belonging to vibrations of pyridine located on dif- ferent acid sites. The band at 1540 cm-' is assigned to the pyridinium ion formed by the interaction of pyridine and the protons of Brsnsted sites (BS band). The band at cu. 1445 cm-' is assigned to pyridine coordinatively bound to Lewis J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 3632 1 3552 al C* s s 3700 3500 waven umber/cm -' Fig. 1 IR spectra of dehydrated HZSM-20 and HY samples in the OH stretching region, recorded at rt 1540 I al Cco e 2n co HY I 1445 HZSM-20 1 I I 1600 1500 1400 wavenumber/cm-' Fig.2 IR spectra of HZSM-20 and HY after pyridine adsorption at rt and partial desorption at 453 K under vacuum, recorded at rt J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 sites (LS band).3' Furthermore, the LS band of HZSM-20 is resolved into two components with maxima occurring at 1445 and 1440 cm-'. This observation points to the presence of at least two types of Lewis sites.'* It was proposed that these bands correspond to chemisorbed pyridine and labile bound species, re~pectively.~~ The thermal desorption of pyridine was followed by IR spectroscopy in order to differentiate between the strength of the different kinds of acid sites in HZSM-20 (Fig. 3). The adsorption on Brarnsted sites is very strong, as shown by the high thermal stability of the pyridinium ion up to 613 K.In contrast, the intensity of the LS band decreases rapidly with increasing temperature. At rt the LS band is dominated by the absorption band at 1440 cm-', but after heating the sample this band disappears indicating a low acid strength of the corresponding Lewis sites. At 453 K the LS band is resolved into two components with maxima arising at 1450 and 1440 cm-'. After additional temperature increase the LS band at 1450 cm-' still remains in the spectrum. This illus- trates the high strength of at least one kind of Lewis site. Interestingly, the decrease in the relative intensity of the BS band with increasing temperature is lower for HZSM-20 than for HY (Table 1).This may indicate stronger Br~rnsted sites in HZSM-20 than in HY. In contrast, the opposite is found for the intensity of the LS band, showing a stronger decrease for HZSM-20. Fig. 4 shows the TPDA profiles of HZSM-20 and HY. It is known that the TPDA profiles of HY zeolite are hardly res~lved.~~*~~The desorption from weak and strong acid sites are not well separated because the desorption of ammonia Table 1 Relative intensity of the BS and LS band in the IR spec-trum of pyridine adsorbed on HZSM-20 and HY as a function of the desorption temperature desorption BS band temperature/K HZSM-20 HY 453 K 100 100 523 K 95 96 613 K 86 76 LS band HZSM-20 HY ~~ 100 100 58 78 49 64 shape. It is resolved into two maxima at 533 (low-tem-perature peak) and 683 K (high-temperature peak).The change in the concentration of acid sites in faujasite-type zeo- lites can be determined by TPDA.23 The ratio of the adsorp- tion peak areas of HZSM-20 and HY amounts to 0.62 +_ 0.03 demonstrating a relatively lower concentration of acid sites in HZSM-20 than in HY. Temperature-programmed FTIR spectroscopic studies on HZSM-20 reveal that the main part of ammonia desorbed up to 723 K comes from the decompo- sition of ammonium ions located on Brarnsted sites (see below). The higher resolution of the TPDA profile of HZSM-20 than that of HY indicates a different distribution of the strength of the Brarnsted sites in HZSM-20.The FTIR difference spectra of ammonia adsorbed on zeolite HY and HZSM-20 are shown in Fig. 5. In the N-H stretching range broad but less well resolved bands appear between 2400 and 3400 cm-'. Below 1800 cm-' resolved N-H bending modes of ammonia adsorbed on dif- ferent Brarnsted and Lewis sites appear. This spectral range is from acid sites with different strengths are superimpo~ed.~~ dominated by an absorption band at ca. 1450 cm-',which is The TPDA profile of HZSM-20 shows a somewhat different usually asigned to ammonium ions formed by the proto- LS I Q1 C Le n8 (I) i&I L I JLl 1 I I I 1600 1400 wavenumber/cm -' Fig. 3 IR spectra of HZSM-20 after (a) pyridine adsorption at rt, partial desorption at (b) 453 K, (c) 523 K and (d) 613 K under vacuum, recorded at rt nation of ammonia molecules located at Brsnsted-acid sites.The sharp band at cu. 1620 cm-' and the band at 1300 cm-' are assigned to bending modes of ammonia coordinatively bound to Lewis From the similarity of the IR spectra of adsorbed ammonia on HY and HZSM-20 it is concluded that, generally, both zeolites contain the same types of acid sites. The intensities of the bending modes of ammonium ions are higher for HY than for HZSM-20. This is in qualitative agreement with the concentration differences found by conventional ammonia TPD as well as with the different A1 framework content of the zeolites, i.e. the lower Si : A1 ratio of HY. I 533 n c.-C .-5 c n U I 683 HZSM-20 473 573 673 773 TIK Fig.4 TPDA profiles of the HZSM-20 and HY samples (heating rate fi = 10 K min-') J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 1 .o 0.8 0.6 0.4 0.2 0.0 -0.2 b) -0.4 I ~IIIIII~I~I 4000 3600 3200 2800 2400 2000 1600 wavenumber/cm-' Fig. 5 FTIR spectra of HZSM-20 (a)and HY (b)after adsorption of NH, at 373 K on the activated sample and subsequent evacuation at 373 K (top spectrum) and at increasing temperatures To analyse the behaviour of ammonia adsorbed on differ- ent sites at increasing desorption temperature, time-resolved IR spectroscopy was applied.24 The spectra shown in Fig. 5 were recorded continuously during the TPD run, using the spectra of the initial zeolite without NH, recorded during the same temperature ramp as the reference.With increasing temperature, the intensities of the bands due to adsorbed ammonia decrease and, simultaneously, the hydroxy vibra- tion bands reappear. At higher temperatures a band at 1300-1335 cm-' becomes visible in the spectra of both the types of zeolites. The corresponding NH, species remain on the sample even at 773 K, so we assign them to NH, bound to strong Lewis sites. A more quantitative picture can be obtained from the determination of desorption absorbances. Calculating the dif- ferences of the NH, and NH4+ absorbances in subsequent spectra (A, -AT+AT) we obtained difference spectra showing the desorption of NH, from the different adsorption sites in the corresponding temperature interval (T + AT).Owing to the different desorption behaviour of ammonia located at various sites, it is possible to obtain a better resolution of different absorption bands. As a result, from the band at 1440-1450 cm-' a distinct shoulder at 1490 cm-' could be resolved, which is assigned to weakly bound NH4+ (Fig. 6). Integrating the desorption absorbances we can obtain the relative amount of ammonia desorbed from the correspond- ing adsorption site. Even taking into account that the absorption coefficients of ammonia adsorbed on different sites may vary, these relative amounts of desorbed ammonia are now directly comparable with the conventional NH, TPD experiments, with the advantage that we can now calcu- late the desorption profiles for the different adsorbed species independently.8 1.0 C e SI% 0.5 0 0 lu e 5:a C .-E. 0.05 gU -0.10 I (I~ 2( 0 1800 1600 1400 wavenum ber/cm -' Fig. 6 FTIR spectra of NH,-loaded HY (A) and corresponding desorption absorbances (B): (a) 423-436 K, (b) 573-586 K, (c) 677-686 K FTIR-NH, TPD profiles obtained for HY in the range of N-H bending vibrations are shown in Fig. 7. The main part of ammonia desorbed from Bransted sites belongs to the 1450 cm-' absorption band. The desorption of ammonia characterized by the 1450 cm-band occurs in a broad tem- perature interval between 423 and 675 K. The desorption maximum is reached at CQ. 533 K. The desorption of NH, 10 a 0000 Cm $6 i2 zlm C ?JO\ .-t. g4 v) U 2 0 373 473 573 673 773 TIK Fig.7 FTIR-TPD curves of ammonia desorbed from different sites of HY: A, 1675 cm-'; 0,1620 cm-'; V, 1490 m-'; 0,1450 cm-'; 0,1380 cm-'; +, 1330-1300 cm-' '.\ J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 species with absorption bands at 1675 and 1490 cm-' pro-ceeds even at low temperatures (below 473 K), suggesting the assignment of these bands to weakly bound species. A com-parison with the behaviour of the hydroxy bands during the TPD run provides further arguments for assignment of the NH, bands. Fig. 8 shows the evaluation of the hydroxy bands. Additionally, a parallel is observed between the disap- pearance of these absorption bands and the reappearance of the high-frequency hydroxy stretching vibration band at low temperature, where 1450 cm-' ammonia is hardly desorbed (Fig.8). This parallel supports the assignment of the 1675 and 1490 cm-' bands to weakly bound ammonium ions, which interact with Bronsted sites in the large cavities. Above 423 K the reappearance of the high- and low-frequency bands of structural hydroxy groups correlates strongly with the disap- pearance of the 1450 cm-' band. Above 475 K distinct differ- ences in the reappearance of high- and low-frequency bands are not found. Ammonia bound to Lewis-acid sites, which is characterized by an absorption band at CQ. 1620 cm-' is desorbed in a low-temperature interval of 423-523 K, indicating the low acid strength of these sites.Additionally, a relatively large fraction of ammonia (absorption band at 1300-1330 cm-') is bound to strong Lewis sites. It is desorbed at high tem-perature between 573 and 773 K. Even at 773 K some ammonia remains at these sites. For HZSM-20 a similar FTIR-NH, TPD pattern was found (Fig. 9). Again the NH, adsorption from Brsnsted sites dominates, belonging to the curve of the 1440 cm-I band. Ammonia desorption occurred over the same tem-perature interval as found for HY. However, the temperature of maximum desorption was ca. 20 K 'higher, indicating a mean stronger Brsnsted acidity of HZSM-20, but it was ca. 60 K lower than the corresponding temperature observed on the strongly acidic HZSM-5 zeolite (Fig.10). Also, weakly bound ammonium ions characterized by absorption bands at 1675 and 1490 cm-' are present and are desorbed between 373 and 473 K. HZSM-20 also contains weak Lewis sites that give rise to an absorption band at 1620 cm-' and desorb ammonia at low temperature. The maximum temperature of 10 8 a C 46 sn m C 0.-4-e42 -0 2 0 373 473 573 673 773 TIK Fig. 8 FTIR-TPD curves of the reappearance of OH groups during ammonia desorption from HY: A, 1675-1490 cm- '; 0,1620 cm- '; V, 1450-1380 cm-'; V,3560 cm-'; 0,3650 cm-'; A, 3745 cm-' lo r------a c z6 s % C .-4-E4 s -0 2 0 373 473 573 673 773 T/K Fig. 9 FTIR-TPD curves of ammonia desorbed from different sites of HZSM-20: A, 1685 an-'; 0,1620 cm-';V, 1490 cm-'; 0,1440 cm-'; 0,1330-1300 cm-' ammonia desorption from strong Lewis sites, represented by the absorption band at 1300-1330 cm-', is slightly below 673 K. Hence, strong Lewis sites in HZSM-20 exhibit a lower acid strength than the corresponding sites in HY.From the TPD curves it is apparent that the amount of ammonia desorbed from strong Lewis sites is lower in HZSM-20 than in HY. In summary, the results of time-resolved FTIR-NH, TPD experiments reveal that Brsnsted sites in HZSM-20 exhibit a mean stronger acid strength than those of HY. Both zeolites under study contain weak and strong Lewis sites, but the concentration of Lewis sites in HZSM-20 is lower than in HY.Additionally, the acid strength of strong Lewis sites is lower in HZSM-20. These findings indicate distinct differ- I I 373 473 573 673 773 TIK Fig. 10 FTIR-TPD curves of ammonia desorbed from Brsnsted sites of HZSM-5 (band at 1470 cm-'): A, NH, form; 0,activated NH, form; V, H-form 2842 c 1601140 I I I 0 1 2 3 [ammonia]/mmol g-l Fig. 11 Differential heat curve of ammonia chemisorption from HZSM-20 (0)and HY (0)measured at 423 K ences in the nature, concentration and strength of the acid sites between the zeolites. Microcalorimetric determination of the differential heat of chemisorption of ammonia (Q) on the zeolites as a function of the sorbed amount is a further suitable method for the char- acterization of the acid properties of zeolite^.^*-^' The initial heat of ammonia chemisorption (Qo) is assumed to be a measure of the acid strength of Brsnsted sites. The heat curve of ammonia chemisorption on HZSM-20 (Fig.11) is similar to that of dealuminated HY.23 The initial heat of ammonia chemisorption on HZSM-20 amounts to ca. 130 kJ mol-'. The comparison of Qo measured on different zeolites shows that the acid strength of Brsnsted sites of HZSM-20 is higher than that of HY, but lower than that of HZSM-5. The sites may be classified as medium-strong acidic sites (Table 2). The heats of chemisorption of ammonia on strong acid sites of zeolites usually exceed 80 kJ mol-'. Sorption heats below 55 kJ mol-' are expected to belong to physisorbed ammonia. Ca.1.2 mmol g-' of adsorbed ammonia on HZSM-20 exhibits a Q value of 80 kJ mol-', i.e. it is located on strong Brsnsted sites. From this, a portion of 0.5 mmol g- 'possesses a higher acid strength than that of HY. An additional amount of 2 mmol g-' ammonia exceeds a heat of 55 kJ mol-'. Hence, besides strongly bound ammonia (Q > 80 kJ mol- ') a large fraction of the ammonia is only weakly bound (80 > Q/ kJ mol-' > 55). The total amount of chemisorbed ammonia exceeding a heat of 55 kJ mol-' amounts to 3.2 mmol g-'. This value is in agreement with the Si: A1 ratio of 4.3 as determined by 29Si MAS NMR. An exact calculation of the total concentration of Brsnsted sites is, however, difficult due to the tailing of the heat curves and the continuous transition from chemisorption to physisorption.The observed differences in the initial heats of ammonia chemisorption on several zeolites correlate well with the shifts of the wavenumbers of the high-frequency hydroxy bands (Table 2). Despite the discussion on the physical nature of Table 2 Initial heats of ammonia chemisorption (Q,,),wavenumbers of the high-frequency hydroxy bands, and maximum temperatures of ammonia desorption from Brernsted sites (Tmax)of different zeolites ~ ~~~~~~ Tm*x/Kwavenumber zeolite Q&J mol-' /cm-' TPDA FTIR-TPD HY 125-1 10 3644 523 543 HZSM-20 135-125 3632 533, 683 563 HSAPO-5 145-135 3625 - - HZSM-5 150-140 3610 730 630 HMOR 160-140 3605 823 - J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 this shift it seems that it is at least a heuristic indicator of acidity changes . 2-44 The acidity of Brsnsted-acid sites of HZSM-20 is found to be medium-strong again. Differences in the acid strength of Brransted sites could arise from structural effects, i.e. different bond lengths and angles in the framework42 as well as different 'cage' effect^.^' In order to check for possible structural effects, HZSM-20 and HY were highly dealuminated to observe the acid strength of Brsnsted sites located on isolated framework alu- minium atoms (Alf). Moreover, the dealuminated samples were carefully extracted in order to avoid any influence of non-framework aluminium species on the remaining Brsnsted sites.The acid properties were characterized by FTIR spectroscopy in the spectral range of the CO stretching vibration of rhodium(1) dicarbonyl ions [Rh*(CO),] as a + sensitive probe for the strength of Brsnsted The frequencies of the carbonyl stretching vibrations are the same for both types of zeolite, indicating a similar strength of the isolated Brsnsted sites, as shown in Fig. 12. Structural effects influencing the acid strength of Brsnsted sites in HZSM-20 could not be detected. However, a structural effect cannot be fully excluded due to the unknown distribution of rhodium ions over the different cation sites in the structure of HZSM-20. Theoretical calculations suggested that the acid strength of Brsnsted sites is related to the aluminium distribution of the zeolite framework.Only Brsnsted-acid -Si-(OH)-Al-framework groups with no second-neighbour Al, in the oxygen four-rings are responsible for the strong Brransted acidit~.'~*~~.~'They are located in the large cavities and are easily accessible for reacting molecules. The numbers of alu- minium atoms with 0, 1, 2 and 3 Al, as second neighbours in 2053 2118 I 1 I 21 25 2050 1975 wavenumber/cm-l Fig. 12 Comparison of IR spectra in the CO stretching region of rhodium-loaded (1 wt.%) highly dealuminated HZSM-20 (a) and HY (b) (Si : A1 > 100) after CO adsorption (1.33 kPa), recorded at rt J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 the oxygen four-rings have been calculated by Beagley et With increasing Si :A1 the number of aluminium atoms with 0 Al, neighbour peaks with increasing Si :A1 ratio at 32 Al, per unit cell.The HZSM-20 zeolite investigated contains ca. 40 Al, per unit cell. At this Al, concentration most of the oxygen four-rings contain an isolated Al, atom. In contrast, zeolite HY has a much higher Al, concentration (>50 Al, per unit cell). Therefore, the number of aluminium atoms with no second-neighbour Al, atom in the oxygen four-rings of HZSM-20 is considerably higher than that of HY. This could explain the higher Brransted-acid strength of zeolite HZSM-20 in comparison with zeolite HY. The samples were tested in n-hexane conversion in order to evaluate the influence of the different acidity of HZSM-20 and HY on their catalytic properties. The n-hexane conver- sion as a function of contact time (W/F)over HZSM-20 and HY is shown in Fig.13. Note that HZSM-20 is more active than HY. Increasing the contact time from 4 to 24 g h mol-' the n-hexane conversion on both zeolites is increased, but over HZSM-20 a distinctly higher degree of conversion is found. This is in qualitative agreement with the enhancement of the Brernsted acidity of HZSM-20, but this alone cannot fully explain the observed substantially high catalytic activity. The contribution of very strong Brransted sites and/or Lewis sites should be considered, but the characterization of acidity presented here gives no evidence for the presence of very strong or 'superacid' Brransted sites in the HZSM-20 sample under study. However, it cannot be excluded that the high catalytic activity found for this material might be a result of the cooperative action of an enhanced number of isolated strong Brransted sites with Lewis sites as proposed by Zho- bolenko et Interestingly, the distribution of the reaction products of n-hexane conversion over HZSM-20 and HY is very different (Fig.14). Over HY deep cracking of n-hexane is observed. The main reaction products are propene, methane and ethane. In contrast, over HZSM-20 deep cracking is dimin- ished and a relatively high amount of aromatics is formed. Generally, the increased Brernsted acidity in faujasite-type zeolites should lead to deeper cracking and only to a low increase in the formation of aromatic^.'^ Hence, it is con- cluded that the increased acid strength of Brransted sites alone is unlikely to be responsible for the different distribu- tion of reaction products found over the two types of zeolite.This indirectly points to the contribution of Lewis sites to the catalytic properties of zeolite HZSM-20. Differences in the number and strength of Lewis sites in HZSM-20 and HY (W/F)/gh mol-I Fig. 13 n-Hexane conversion over HZSM-20 (a) and HY (6) as a function of contact time 2843 50 1 40 30 YO 20 10 n c.1 c2 c3 C3 aromatics -HZSM-20 HY Fig. 14 Distribution of reaction products in n-hexane conversion over HZSM-20 and HY were found.In summary, the catalytic activity and selectivity in n-hexane conversion confirm the differences found in the number and strength of Brsnsted- and Lewis-acid sites between HZSM-20 and HY. Conclusions Zeolite HZSM-20 displays a medium-strong Brernsted acidity, ranging in its strength between HY and HZSM-5. This classification is deduced from the observed initial heat of ammonia chemisorption and from the wavenumber of the high-frequency band of Brernsted hydroxy groups in the large cavities. Owing to the higher Si :A1 ratio of zeolite HZSM-20 in comparison to zeolite HY, the number of strong Brsnsted sites with no second-neighbour aluminium atoms is increased, leading to a change in the distribution of acid sites of different strength.This is reflected in the shift of the bridg- ing OH band to lower wavenumber (3632 cm-'), the higher temperature of NH, desorption from Brransted sites and the higher heat of ammonia chemisorption (ca. 130 kJ mol-') by zeolite HZSM-20. The heat curve of ammonia chemisorption reveals that ca. one-third of the ammonia molecules located at Brransted sites are strongly bound (Q > 80 kJ mol-I). Surprisingly, the main part is weakly bound with chemisorption heats ranging from 80 to 55 kJ mol- '. This observation is confirmed by the high- wavenumber shift of the low-frequency band in the OH spec- trum of HZSM-20 compared to HY. Besides Brernsted sites, Lewis sites were also found. Two kinds of Lewis site of different strength seem to exist.In the IR spectrum of adsorbed pyridine on HZSM-20 two charac- teristic bands at 1450 and 1440 cm-' appear. They are assigned to the interaction of pyridine with stronger and weaker Lewis sites. Additionally, the strong Lewis sites in HZSM-20 are weaker than the strong sites in HY, as shown by FTIR-NH, TPD. In summary, it is concluded that the increased acidity of zeolite HZSM-20 is due to a change in the distribution of the strength of acid sites, where the number of strong Brransted sites with high acid strength is significantly increased. There- fore, in comparison to zeolite HY the intrinsic acidity of HZSM-20 is improved but the number of Brsnsted-acid sites is decreased. The strength of isolated strong acid sites is similar in both types of zeolites.Superacid sites could not be detected. The moderately increased acidity cannot explain the observed high catalytic acitivity of this material alone, because mildly dealuminated HY samples have a similar strength of Brernsted sites but not such a high activity.'l Thus, it is tentatively concluded that the origin of the high 2844 J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 catalytic activity of HZSM-20 and the difference in the dis- 24 H. Miessner, H. Kosslick, U. Lohse, B. Parlitz and V. A. Tuan, tribution of reaction products could be a cooperative action J. Phys. Chem., 1993,97,9741. of Brernsted and Lewis-acid sites as proposed by Beyerlein16 25 A. Martin, U. Wolf, H. Berndt and B. Liicke, Zeolites, 1993, 13, 309.and Zh~bolenko.~~ 26 I.Burkhardt, E. Jahn and H. Miessner, J. Phys. Chem., 1992, %, 2259. The technical assistance and work of Mrs. U. Wolf, Mrs. K. 27 M. Stocker, S. Ernst, H. G. Karge and J. Weitkamp, Acta Chem. Struve, Mrs. W. Ziesche and Mr. D. Schielinski is gratefully Scand., 1990,44,519. acknowledged. 28 P. A. Jacobs and J. B. Uytterhoeven, J. Chem. SOC., Faraday Trans. I, 1973,69, 373. 29 J. W. Ward, J. Catal., 1970,17, 355. References 30 D. Barthomeuf, J. Phys. Chem., 1979,83,249. 31 T. R. Hughes and H. M. White, J. Phys. Chem., 1967, 71, 2192; 1 J. Ciric, US Pat., 3,972,983, 1976; 4,021,331, 1979. L. M. Parker, D. M. Bibby and G. R. Burns, J. Chem. SOC., 2 S. Ernst, G. T. Kokotailo and J. Weitkamp, Zeolites, 1987, 7, Faraday Trans., 1991,87,3319.180.32 B. H. Chiche, F. Fajula and E. Garrone, J. Catal., 1994, 146,3 F. Lemos, J. .M. Lopes, F. R. Ribeiro and E. G. Derouane, Appl. 460. Catal., 1989,49, 175. 33 B. M. Lok, B. K. Marcus and C. L. Angell, Zeolites, 1986,6, 185. 4 D. E. W. Vaughan and K. G. Strohmaier, US Pat., 5 133 953, 34 A. Baranski, T. Dal and J. Galuszka, Bull. Pol. Acad. Sci., Ser. 1992. Chim., 1979,27, 363. 5 R. B. Lapierre, R. D. Partridge and S. S. F. Wong, Eur. Pat. 35 H. G. Karge, V. Dondur and J. Weitkamp, J. Phys. Chem., 1991,Appl., 94862, 1983. 95, 283. 6 H. Litterer, Ger. Pat. Appl., 3 334673, 1985. 36 A. V. Kiselev and V. I. Lygm, lnfrared Spectra of Surface Com- 7 R. M. Dessau, US Pat., 4,384,153, 1983. pounds, Wiley, New York, 1975. 8 J.M. Lopes, F. Lemos, F. R. Ribeiro, N. Dewaele, E. G. 37 K. Nakamoto, Infrared Spectra of Inorganic and Coordination Derouane and M. Guisnet, J. Mol. Catal., 1988,48, 373. Compounds, Wiley-Interscience, New York, 1986. 9 V. Fiilop, G. Borbely, H. K. Beyer, S. Ernst and J. Weitkamp, 38 H. Stach, J. Janchen and U. Lohse, Catal. Lett., 1992,13,389.J. Chem. SOC., Faraday Trans. I, 1989,85,2127. 39 J. C. Vedrine, A. Auroux, V. Bolis, P. Dejaifve, C. Naccache, P. 10 J. M. Newsam, M. M. J. Treacy, D. E. W. Vaughan, K. G. Wierzchowski, E. G. Derouane, J. B. Nagy, J-P. Gilson, J. H. C. Strohmaier and W. J. Mortier, J. Chem. SOC., Chem. Commun., van Hooff, J. P. van den Berg and J. Wolthuizen, J. Catal., 1979,1989,493. 59, 248. 11 J. L. Lievens, J. P.Verduijn, A-J. Bons and W. J. Mortier, Zeo-40 A. L. Klyachko, G. 1. Kapustin, T. R. Brueva and A. M. Rubin- lites, 1992, 12, 698. stein, Zeolites, 1987,7, 119. 12 R. B. Borade and A. Clearfield, J. Phys. Chem., 1992, %, 6729. 41 A. G. Pelmenschikov, G. Morosi and A. Gamba, J. Phys. Chem., 13 Y. Sun, P-J. Chu and J. H. Lunsford, Langmuir, 1991,7,3027. 1991,95, 10037. 14 B. L. Su, J. M. Monoli, C. Potvin and D. Barthomeuf, J. Chem. 42 J. G. Fripiat, F. Berger-Andre, J. M. Andre and E. G. Derouane, SOC., Faraday Trans., 1993,89,857. Zeolites, 1983, 3, 306. 15 A. Corma, V. Fornes and F. Rey, Appl. Catal., 1990,59,267. 43 A. G. Pelmenschikov, E. A. Paukshtis, V. I. Pavlov, E. N. Yur-16 R. A. Beyerlein, G. €3. McVicker, L. N. Tacullo and J. J. Ziemiak, chenko, K.G. Ione, G. M. Zhidomirov and S. Beran, J. Phys.J. Phys. Chem., 1988,92, 1967. Chem., 1989,93,6725.17 H. Kosslick, M. Richter, V. A. Tuan, R. Fricke and W. Storek, 44 S. Beran, J. Phys. Chem., 1990,94,335.J. Chem. SOC.,Faraday Trans., 1992,88,2421. 45 W. J. Mortier and R. A. Schoonheydt, Surface and Solid State 18 A. Martin, U. Wolf, H. Kosslick and V. A. Tuan, React. Kinet Chemistry of Zeolites, in the series Prog. Solid State Chem., 1985,Catal. Lett., 1993,51, 19. 16, 86 ff. 19 S. Ernst, G. T. Kokotailo and J. Weitkamp, Innouation in Zeolite 46 E. Dempsey, J. Catal., 1974,33,497; 1975, 39, 155. Materials Science, ed. P. J. Grobet, W. J. Mortier, E. F. Vansant 47 R. J. Mikovsky and J. F. Marshall, J. Catal., 1976,44, 170. and G. Schulz-Ekloff, Elsevier, Amsterdam, 1987, p. 29. 48 B. Beagley, J. Dwyer, F. R. Fitch, R. Mann and J. J. Walters, 20 J. Dwyer, D. Millward, P. J. OMalley, A. Araya, A. Corma, V. J. Phys. Chem., 1984,88, 1744. Fornes and A. Martinez, J. Chem. SOC., Faraday Trans., 1990, 49 V. L. Zhobolenko, L. M. Kustov, V. B. Kazansky, E. Loeffler, 86,101. Ch. Peuker and G. Ohlmann, Zeolites, 1990,10,304.21 E. G. Derouane, N. Dewaele, 2. Gabelica and J. B. Nagy, Appl. 50 A. Corma and A. V. Orchilles, J. Catal., 1989,115, 551. Catal., 1986,28, 285. 51 J. Dwyer, K. Karim, W. J. Smith, N. E. Thompson, R. K. Harris 22 V. A. Tuan, R. Fricke, K. Szulzewsky and H. Kosslick, Crystal and D. C. Apperley, J. Phys. Chem., 1991,95,8826.Res. Technol., 1991,26, 161. 23 U. Lohse, B. Parlitz and V. Patzelova, J. Phys. Chem., 1989, 93, 3677. Paper 4/02549K; Received 29th April, 1994

ISSN:0956-5000    

DOI:10.1039/FT9949002837

出版商:RSC    

年代:1994

数据来源: RSC