摘要:

Faraday Discuss. Chem. SOC., 1989, 87, 133-147 Mechanism of Sulphur Dioxide Oxidation over Supported Vanadium Catalysts Bair S. Balzhinimaev, Alexei A. Ivanov, Olga B. Lapina, Vyacheslav M. Mastikhin and Kirill I. Zamaraev* Institute of Catalysis, Siberian Branch of the USSR Academy of Sciences, Novosibirsk 630090, USSR The mechanism of SOz oxidation to SO, over industrial vanadium catalysts has been elucidated on a molecular level using 51V, "0 and 23Na n.m.r., i.r. and relaxation kinetic methods. The following reaction scheme has been shown to describe quantitatively the whole set of experimental data: slow side-reaction 2VXV I 0 fast side-reaction 1 v; ,vv 0 2 catalytic cycle According to this scheme three types of binuclear vanadium( v) complexes are involved in the catalytic cycle: the oxocomplex, sulphite complex and peroxocomplex.The rate constants for all the steps of the catalytic cycle have been determined. The data obtained provide guidelines for further improvement of vanadium catalysts as applied to particular conditions of their operation. Oxidation of SO, to SO3 over supported vanadium catalysts is known to be the key stage of modern industrial processes for sulphuric acid production. The great practical importance of the oxidation of SO, over vanadium catalysts has stimulated numerous mechanistic studies of this process. Contributions to this field made by Boreskov should be particularly stressed.''2 He and his co-workers studied in detail the kinetics of SO2 oxidation to SO3 under both steady-state and non-steady-state conditions.Relaxation methods have been used to study the kinetics in the latter case.3 These data, together with the e.s.r. data concerning the rates of vanadium( v) r* vanadium( ~ v ) transformation in the catalyst during the reaction4 have provided a deep insight into the mechanism of this reaction. 133134 Mechanism of SO, Oxidation In recent years SO2 oxidation over vanadium catalysts has been extensively studied also with such powerful spectroscopic techniques as 51V, ''0 and 23Na n.m.r. [see review articles in ref. (5)-(7) and references therein) and i.r.;8 additional relaxation kinetic studies have been carried out for various steps of this reaction.' These new data when analysed together with the data of earlier allowed us to elucidate the mechanism of catalytic oxidation of SO, on a molecular level.In fact, the insight into the mechanism of this heterogeneous reaction is now almost as deep as that transitionally available only for homogeneous catalysis. Vanadium catalysts for SO, oxidation consist of sulphate and pyrosulphate com- pounds of vanadium and alkali metals, usually potassium, which are supported on porous materials such as silica or silica-alumina. Under reaction conditions, i.e. at 700-800 K, the active component of the catalysts exists as a melt forming a very thin liquid layer on the surface of the support. The thickness of this liquid film increases with the increase in the amount of the active component in the catalyst and the decrease of the surface area of the porous support.For typical industrial catalysts the liquid film is only 102-103 8, thick. The oxidation reaction s0,+1/20, r* so3 (1) proceeds on the active sites located in the bulk of the liquid film and perhaps also on the boundary between the film and the surface of the solid support (see below). Catalyst Characterization with 'lV, "0 and 23Na N.M.R. Experimental N.m.r. experiments were performed using a Bruker CXP-300 spectrometer (7.04 T for V n.m.r., 2.14 T for 1 7 0 and 23Na n.m.r.). 5'V n.m.r. spectra were recorded at 78.86 MHz in the frequency range 250 kHz with a 1 ps radiofrequency pulse and with a pulse repetition frequency of 10 Hz. The scan accumulation number was from lo3 to lo5. The chemical shift of the 51V spectra was measured relative to the signal of an external V0Cl3 reference."0 and 23Na n.m.r. spectra were recorded at 770 K using Pyrex sample tubes with an outer diameter of 10mm; the high-temperature probe head was cooled with flowing water, and a BE-45 Bruker magnet with a 40mm pole gap and with I9F field stabilization was used. I7O n.m.r. spectra were recorded at 12.21 MHz in the frequency range 30 kHz. The pulse duration was 30 ps and the pulse repetition frequency was 50 Hz. The chemical shift of the "0 spectra was measured relative to the signal at 298 K of an external H 2 0 reference. 5 x 104-105 scans were used for signal accumulation. Na n.m.r. spectra were recorded at 23.8 MHz in the frequency range 20 kHz. The pulse duration was 5 ps and the pulse repetition frequency was 10 Hz. The 23Na chemical shifts were measured relative to the signal of a solid Na2S04 external reference (whose linewidth at 770 K was <lo0 Hz).Before 51V, I7O and ""a n.m.r. measurements were recorded the samples were contacted at 770 K for 70-80 h with an equilibrium SO3 r* SO, +502 mixture which was generated by partial dissociation of SO3. SO, was preliminary introduced into the samples at room temperature and at a pressure of 52.6 kPa. The samples were then sealed with the gas medium and used for n.m.r. measurements. 5 1 23 Solid-state 5'V N.M.R. Solid-state "V n.m.r. was used to characterize vanadium species that are present in the catalysts at room temperature when the film of the active component is solid rather than liquid. At elevated temperatures when the film melts, the "V n.m.r.lines became too broad and thus could not be recorded.B. S. Balzhinimaev et al. 135 0 -500 -1000 Fig. 1. "V n.m.r. spectra of vanadium compounds with various environments of oxygen atoms. ( a ) VzOs, distorted octahedron; ( b ) KV03 , distorted tetrahedron; (c) K,VO,, regular tetrahedron. The starting point of these studies was the measurement of the 51V n.m.r. spectra of various solid vanadium compounds in which the Vv atom is bound to oxygen anions and/or oxygen atoms of sulphate and pyrosulphate ligands. A library of 51V n.m.r. "fingerprints' from such compounds has been compiled and certain correlations have been elucidated between the chemical shifts and the shape of n.m.r. lines on one hand and the structure of the local environment of vanadium atoms on the Fig.1 shows some typical 51V n.m.r. spectra for vanadium in distorted octahedral (vanadium pentoxide and oxosulphate vanadates), distorted tetrahedral (metavana- dates) and regular tetrahedral (orthovanadates and pyrovanadates) environments of oxygen atoms. Vanadium in distorted octahedral environments demonstrates spectra with almost axial symmetry. Vanadium in a distorted tetrahedral environment is charac- terized by a fully anisotropic chemical-shift tensor S (the corresponding computer- simulated spectrum for the best-fit parameters is shown in fig. 1 by the dotted line). For vanadium in a regular tetrahedral environment the spectrum is seen to be isotopic. Consider now the "V n.m.r. spectra of the catalysts. Before treatment in the catalytic reaction the spectra of various industrial vanadium catalysts were found to be sig- nificantly different and to depend on the catalyst preparation procedure.However, the treatment of the catalysts under catalytic reaction conditions was found to smooth these differences out. This indicates that the active component in all the catalysts studied is the same and is actually formed in the course of the catalytic reaction. Two types of vanadium(v) compounds have been revealed in oxidized catalysts by V n.m.r. First, from a comparison of the "V n.m.r. spectra of the catalysts with those from our library it follows that oxosulphate vanadates K3V02( SO,)( S,O,) (a1, = -1260 ppm, 8, = -320 ppm) and K3VO2(S0J2 (SII = -1040 ppm, S, = -320 ppm) are present in industrial catalysts as the main components.The relative amount of these 51136 Mechanism of A SO2 Oxidation B 0 -500 -1000 -1500 I 1 I I 0 -500 -1000 -1500 6 (PPm) Fig. 2. (A) Comparison of the "V n.m.r. spectrum of an industrial catalyst ( a ) treated under reaction conditions, with the spectrum of K3V02(S04)(S207) ( 6 ) . (B) The 51V n.m.r. spectra for 4% VzOs supported on Si02 with specific surface area (m'g-'): ( a ) 20, ( 6 ) 170, ( c ) 300. two oxosulphate vanadates depends on the catalyst preparation procedure. A typical spectrum of an industrial catalyst containing K3V02(S04)(S207) is shown in fig. 2. The posiiion and shape of the lines from these compounds are typical for vanadium surroun- ded with a distorted octahedron of oxygen atoms. In the 51V n.m.r. spectra of industrial catalysts (fig.2) there is also a line with a chemical shift of -540 ppm. We attribute this line to vanadium which interacts with the support. Indeed, for V205 supported on Si02 we observed the same line [see fig. 2(B)]. The intensity of this line increased strongly both when the surface area of Si02 was increased, the amount of supported V205 remaining the same, and when the amount of supported V205 was increased, the surface area of Si02 remaining the same. This clearly indicates that the compound responsible for this line is formed upon interaction of the starting vanadium compound with the support. The position and shape of the line is typical for vanadium surrounded by a tetrahedron of oxygen atoms. The role of the elucidated vanadium compounds in catalysis will be discussed in subsequent sections.1'0 N.M.R 0 n.m.r. has been used to characterize the catalysts at 770 K, i.e. under conditions typical of their operation in industry. Fig. 3 shows the 170 n.m.r. spectra of a V205-K2S207 melt, which is the active component of the catalysts. These spectra were recorded at 770 K for melts with various K:V ratios (see fig. 3). As seen from fig. 3, the spectrum of the pure K2S2O7 melt contains only one line ca. 100 Hz wide at 170 ppm, despite the fact that in the S2072- anion there are two types of oxygen atoms. The presence of only one line suggests fast exchange between terminal and bridging oxygen atoms which may occur via the reaction 17 s20: r* so:-+so,. (2) Reaction (2) changes the volume of the melt. Because of this the characteristic time, r,B.S. Balzhinimaev et al. 2K,VO(SO,), Ks 137 v<so4. (4) I 0 0 II p 4 , II SO4=V\ I s04’1 so4 so4 2000 1500 1O:l 1 5 : 1 5O:l 200 :l P I I I 1 600 400 200 0 I 1 I \ S(ppm) 1 2 3 4 C,./mol dm-3 Fig. 3. The I7O n.m.r. spectra of V205-K2S20, melts at various K : V ratios and dependences of the I7O chemical shift (0) and linewidth (0) on vandium concentration.138 Mechanism of SO, Oxidation ( b ) O@ 0 0 oQo P 0 Fig. 4. Schematic representation of the structures of vanadium species present in V,05-K,S,0, melts: ( a ) at small (K : V > 10) concentrations; ( b ) at large ( K : V = 3-5) concentrations. Remember that according to "V n.m.r., upon freezing of the melt in industrial catalysts with K : V = 3, for which C, = 4 mol dm-3 and thus equilibrium (3) must be shifted to the right, K,V02( SOJ2 and K3V02(S04)(S207) entities are formed in which vanadium atoms are in distorted octahedral environments of oxygen atoms.Such entities can perhaps be formed via reactions of the type Note that the change of the slope for the Av us. C, curve in fig. 3 occurs in a narrower interval of C, values than is expected for a simple dimerization reaction. This can be explained by further association of vanadium species forming larger oligomers of the type shown in fig. 4( 6). The large size of these oligomers makes their rotational diffusion very slow. The internal rotation of its fragments can also be hindered because of the branching and linking of the oligomeric chain. Owing to these two factors the -"V n.m.r. lines of these species are too broad to be detected.Fig. 5 demonstrates how the catalytic activity is changed upon variation of vanadium concentration in the V205-K2S207 melt supported on two different solids: Pyrex glassB. S. Balzhinimaev et al. 139 5 r C,/mol dmp3 Fig. 5. Catalytic activity in terms of the rate constant kk (see scheme 1 and table 1) versus vanadium concentration for V20,-K2S20, melts supported on Pyrex glass (0) and silica (0). Calculated from the data of ref. (7). T = 770 K. tubes with total surface area S = 0.01-0.04 m2 and silica with specific surface area S = 200 m' g-'. An increase in the activity at high vanadium concentrations suggests that dimeric and oligomeric vanadium species are much more active than monomeric species.For the melt supported on Pyrex glass the increase in the catalytic activity occurs within a much narrower interval of C, values than expected for the situation when active sites are formed via a simple dimerization reaction. In agreement with the I7O n.m.r. data, this suggests that at high C, vanadium complexes form not only dimers but also oligomers and that catalytically active sites are located in both dimers and oligomers. Note that sharp changes in the dependences of n.m.r. parameters of the melt (fig. 3 ) and of its catalytic activity on Pyrex glass (fig. 5) on vanadium concentration occur at notably different regions of C , , namely C, =r 1 mol dm-3 in the first case and 3 mol dm-' in the second case. This can be explained if one assumes that dimerization (oligomeriz- ation) follows route (4)-(5a) and also takes into account that catalytic activity was measured with the samples where the thickne!s, h, of the melt layer supported on Pyrex glass and SiO, was ca.3 x lo4 and 5 x lo2 A, respectively, while "0 n.m.r. spectra were recorded with samples containing a ca. 2 cm thick layer of a pure melt. From the values of the diffusion coefficient of SO3 in the melt (see below) one can estimate that for the layers with h = 5 x 10,-3 x lo4 A diffusion of SO3 is fast enough (characteristic time T ~ = 10-3-3 s) to provide equilibration of the compositions of the melt with that of the gas phase, while for the layer with h = 2 cm it is too slow ( T~ == lo7 s) for such equilibration to occur. In this case one may expect an enhancement of the dimerization (oligomerization) process via reaction ( 5 a ) for thick layers, since in this case SO3 from the gas phase has no time to penetrate into the bulk of the melt and thus shift equilibrium (4)-(5a) to the left.For the V205-K2S207 melt supported on Si02 the formation of the catalytically active species starts at notably lower vanadium concentration than for the melt supported on Pyrex glass. However, for both supports this process becomes completed at about the same C, (ca. 4 mol dmP3), nearly the same catalytic activity being achieved at C, * 4 mol dm-3. These facts can be explained if we assume that dimeric vanadium species can be additionally stabilized on the surface of Si02, e.g. via ligand-substitution reactions140 of the type Mechanism of SOz Oxidation K'i 0 0 0 0 I mflmnmmm;r + 2 KHS04 For this reaction to occur, two closely located OH groups must be present on the silica surface.The existence of such groups has been proved recently by experiments in which anchoring of monomeric metal complexes to those neighbouring groups was shown to result in the formation of oligomeric complexes." The possibility for vanadium to coordinate to the surface sites of SiO, has been demonstrated with "V n.m.r. (see fig. 2). In terms of the proposed model, the dependence of the catalytic activity versus C, for the melts supported on SiO, can be interpreted as follows. Starting from C,= 0.2 mol dm-' dimeric vanadium species attached to the surface of SiOz are formed. These species are about as active as dimeric and oligomeric species in the bulk of the melt and make the main contribution to the catalytic activity up to C, == 2 mol dm-3.Starting from C, = 2 mol dm-' catalytically active dimeric and oligomeric species are also formed in the bulk of the melt. 23Na N.M.R. If one substitutes potassium in the melts for sodium, it is possible to observe the 23Na n.m.r. spectra of these melts.5 The dependence of the chemical shift of the 23Na signal on the sodium-to-vanadium ratio suggests that some of the sodium cations can be coordinated to sulphate or pyrosulphate anions which are already bound to vanadium atoms, forming a sort of outer-sphere ionic complex with vanadium sulphate (pyrosul- phate) compounds. Relatively small changes in the chemical shift and width of "Na n.m.r.lines upon addition of V2OS agree with this modeL5 Relaxation Kinetic Studies Experimental Relaxation kinetic studies have been carried out using the device represented schemati- cally in fig. 6 [see ref. (3) and references therein]. Reaction mixtures 02-SO2-SO3 of various composition cr 0, were fed to reactor 1 using He as carrier gas. The reactor was made of two coaxial 10 cm long Pyrex glass tubes of 7.0 and 6.3 mm diameter which made the gap between the walls of the tubes 0.35 mm. The inner surface of the outer tube and the outer surface of the inner tube were covered with V2O5-K2S2O7 melts with K:V from 3 to 5. The amount of melt corresponded to a film thickness of ca. 3.5 x lo4 A. Variations in the composition of a reaction mixture were recorded using an MSKh-6 time-of-flight mass spectrometer with a stroboscopic discrete signal converter providing its output signal either to a recorder or to an oscillograph in the case of veryB.S. Balzhinimaev et al. at mospher e 141 He t 02 -f - 0 0 0 0 0 0 0 0 0 - K& \ +o2tso2+sog 5 0 0 0 0 0 0 / I recorder 7 gas admission mass spectrometer Fig. 6. Device for the relaxation kinetic studies. 1, Inner Pyrex tube; 2, outer Pyrex tube; 3, layers of V2O5-K2S20, melt; 4, heater; 5, valve with a switch time of 0.1 s. fast changes in the signal. To prevent SO3 polymerization and condensation, the gas-admission system and the feeding lines were heated to 430-440K. In order to decrease the memory effect towards sulphur oxides and to prevent their corrosive action, the surfaces of the admitting valve, the components of the ion source and the drift tube of the analyser were covered with gold.This permitted the analysis time to be reduced to times as short as 0.2s. The device was used to examine the kinetics of two types of processes: (i) diffusion of 02, SO, and SO3 in the melt, and (ii) individual steps of the catalytic reaction. These two types of process are revealed when studying rapid and slow relaxations of the composition of the reaction mixture after the catalytic reactor, induced by jumpwise changes in the composition of the reaction mixture before the reactor. As shown in ref. (3), (8) and (9), rapid relaxation processes result from changes of the concentration of the reactant gases in the melt and are controlled by their diffusion in the melt.Slow relaxation processes result from changes in the chemica! composition of vanadium complexes in the melt and are controlled by the chemical reactions forming the catalytic cycle. Typical relaxation data for both types of processes are given in fig. 7. Rapid Relaxation Processes Fig. 7( a ) shows rapid relaxations of the partial pressure, Pso,, of SO2 in the gas phase after the catalytic reactor, induced by variations in the composition of the gas phase before the reactor. Section 1 of the relaxation curves corresponds to the steady-state content of 02, SO, and SO3 in the melt, which was established during its treatment in142 1.0 5 4 \ CIA v 0.5 Mechanism of SO2 Oxidation 1 0 2 4 6 8 t l s -150- 0 40 80 120 t l s Fig. 7.Rapid ( a ) and slow ( b ) relaxation of the partial pressure of SO2 in the gas phase after the catalytic reactor upon jumpwise switches of the composition of the reaction mixture before the reactor. The solid and dotted lines are, respectively, the relaxation curves for the reactor loaded with the V205-K2S20, melt and the empty reactor. T = 770 K. the initial reaction mixture of constant composition P& = 0.4, Pgoz = 1.23, Pioo, = 3.1, PEe = 95.27 kPa. Section 2 corresponds to the relaxation of Pso, after a rapid switch from the initial reaction mixture to that composed of He-0, with Po2 = 1, PHe = 99 kPa. Section 3 corresponds to the relaxation of Pso2 back to the initial steady state upon switching the composition of the reaction mixture back to the initial one.From the relaxation curve of fig. 7(a) one can calculate the diffusion coefficient of SO2 and its solubility in the melt. The dotted line in the curve corresponds to the relaxation of Psoz in the empty reactor having no V205-K2S207 melt, and characterizes the resolution time, T,, of the device depicted in fig. 6. T, is seen to be notably shorter than the characteristic times for the relaxation of Pso, in the presence of the melt. Similar relaxation curves were obtained for Po2 and Pso,. The diffusion coefficients, D, of O,, SO, and SO3 and their solubility in the melt at a partial pressure of 1 kPa (i.e. effective Henry coefficients H) found from these curves, are summarized in table 1. The very high value of Hso, indicates chemical binding of SO3 to some sites in the melt, presumably via the reaction SO3 + SO:- r* S205-.The very low value of Dso3 suggests that in the melts studied with K: V = 3-5, &Of- anions are bound to oligomeric vanadium complexes [see fig. 4(b)]. The diffusion of such complexes is very slow because of their large size, and it controls the transport of SO3 in the melt. Slow Relaxation Processes Fig. 7(b) shows by way of an example slow relaxation of the partial pressures of SO2 upon a jumpwise switch of the gas-phase composition from POT = 4.6, PHe = 95.4 kPa to P& = 0.74, Pgo2 = 0.75, P& = 3.1 and P i e = 95.41 kPa. The inflexion in the relaxation curve clearly indicates the formation of some intermediate complex between SO, and vanadium species in the melt and allows one to calculate the characteristic times of its formation and decomposition.As shown by i.r. spectroscopy, SO2 exists in this intermediate complex in the form of the sulphite anion, SO:- (characteristic bands at 1083, 930 and 652 cm-'). Fig. 8 shows a typical relaxation curve for the SO:- anion obtained using a conventional Bruker IFS-113v spectrometer with a specially designed cell which allowed us to studyB. S. Balzhinimaev et al. 143 Table 1. Kinetic parameters for SO2 oxidation values at 770 K enthalpy or activation energy constants a b /kJ mol-' Ho2/mol cm-3 kPa-' Hso,/mol cm-3 kPa-' Hso3/ mol cm-3 kPa-' D02/cm2 s-' DSOZ/cm2 s-' D,,,/cm's-' kk/cm3 mol-' s-' k?,/cm3 mol-' s-' k)/cm3 mol-' s-' k_2/s-' k:/cm3 mol-' s-' kY3/cm3 mol-' s-' kq/S-l kl-,/cm3 mold' s-l K,/cm3 mol-' a5/ kPa-' K , K J kPa-' K3 K4/ kPa 3 x lo-" 2 x 1 0 - ~ 10-~ lo-" 1.5 x lo-' 1 o4 5 x lo3 4~ lo3 3.2 x lop8 0.45 0.05 10 1 o2 0.1 0.11 5 3 x 1 0 - ~ 6.3 x 1 0 - ~ 2.5 x lo-' 1.6 x lo4 6.1 x lo3 5~ lo3 lo-" 4.5 x 3.6 0.5 6.3 0.1 1.9 x 10' 0.12 0.12 3.8 6.3 27 61 -50 66 42 121 79 142 42 50 -59 -109 -42 -79 -63 42 (' From relaxation studies of various reaction steps.steady-state experiments. From the overall set of relaxation and 5 10 t/min Fig. 8. Relaxation curve for SO:- anion obtained by an i.r. method upon admission of SO3 ( Pso, = 0.7 kPa) upon the V2O5-K2S20, melt, prereduced up to the steady-state under PsO, = 10 kPa.144 Mechanism of SO2 Oxidation lo5 A thick layers of the V205-K2S207 melt supported on a silicon plate, under reaction conditions.Prereduced melt was used in this study instead of the preoxidized melt used in experiments of fig. 7 ( b ) . The following observations made using relaxation techniques and steady-state kinetic studies are important for elucidation of the mechanism of the catalytic reaction. (i) The rate of the oxidation with O2 of the V" complexes formed upon prereduction of the catalyst with SO, is much less than the rate of the overall catalytic reaction under steady-state conditions. This means that the catalytic reaction does not procees via the stepwise mechanism consisting of alternating steps of Vv reduction with SO2 to VIV and subsequent oxidation of VIv back to Vv with 02.4 Moreover, the rate of the catalytic reaction under both steady-state and non-steady-state conditions is proportional to the concentration of Vv and does not correlate with the amount of V"' in the catalysts.'2 This suggests that only Vv species are involved in the catalytic cycle.' (ii) Ca.two rather than four VIV atoms per O2 molecule are oxidized to Vv (though very slowly) upon absorption of O2 by prereduced catalysts. One O2 molecule is evolved into the gas phase per SO, molecule reacted upon admission of SO3 to preoxidized catalysts (at the initial period of O2 -+ SO, substitution process in the melt). Both these facts suggest that peroxide fragments of the vv/ \vv type are formed in the melt in the presence of O2.I3 (iii) The involvement of binuclear vanadium( v) fragments in catalysis is supported also by the stoichiometry of VIv interaction with SO, (two VIv atoms disappear to form, fragments)'? as well as by the increase of the catalytic activity apparently, at high concentrations of vanadium, which enhances the formation of such fragments (see fig.5 ) . Note that these catalytically active fragments can be either dimeric vanadium complexes or binuclear sites of oligomeric vanadium complexes [see fig. 4(b)] in the bulk of the melt and on its boundary with the support. (iv) Desulphonation of the melts, i.e. elimination of SO, from SO:- and/or S20;- ligands via sufficiently long bubbling of He through the melts, results in their deactivation. This suggests that a certain concentration of SO:- and/or S20;- ligands is necessary in binuclear Vv fragments to make them active in ~atalysis.'~ (v) The rate of the catalytic reaction is also suppressed by an excess of SO3 in the gas phase.' This suggests the existence of some optimum concentration of SO:- and/or S @ - ligands in the coordination sphere of binuclear vanadium(v) fragments forming the active sites of the catalysts.All the structural and kinetic data obtained so far for SO, oxidation over vanadium catalysts can be interpreted quantitatively in terms of the following reaction scheme. Binuclear vanadium(v) fragments formed via reactions ( 5 a ) , ( 5 b ) or ( 6 ) loose one of the bridging 0,- ligands, e.g. via the reaction 0;- so:- \ VV/ VV 0 0 0 0 II p, II II ,o, II l o 1 I I L:v, ,v:L + s20;- L:v v:L + 2so:- L L L L (7) 0 V V which produces the coordinatively unsaturated binuclear / \ fragments that serve as the active sites of the catalysts.B.S. Balzhinimaev et al. fast side-reaction 0 s207 v<' ',.v s,o:- 145 I 0 VV' ' v; / V V 0 2 vv&03 so2 x 1 0 vv v"' ' The catalytic cycle can be represented as follows: slow side-reaction Scheme 1. Note that all the three complexes involved in the catalytic cycle can participate in the fast side reaction ( 5 a ) , although for the sake of simplicity in scheme 1 this reaction is shown only for complex vv/ \vv. The cycle is seen to involve three types of binuclear 0 0 so3 vanadium compounds: an oxocomplex, vv/ \vv, a sulphite complex, vv/ \,w 0 2 and a peroxocomplex, vv/ \vv. A vacancy in a bridging position of these complexes seems to be necessary for the accomplishment of stage ( 3 ) , which we assume to proceed via coordination of O2 to this vacancy and subsequent transfer of two electrons from coordinated SO:- to 02, forming a peroxide complex and evolving SO3: This reaction scheme provides the following rate equation where Cv,, Co,, Cso, and Cso3 are the concentrations of binuclear vandium fragments, 02, SO2 and SO3 in the melt.Remember that SO3 exists in the melt in the chemically bound form as S20;- species. For industrial catalysts with a sufficiently high concentra- tion of vanadium nearly all the vanadium exists in the form of such fragments, so that146 Mechanism of SO, Oxidation Cv2= C v / 2 , where Cv is the total concentration of vanadium in the melt. The subscript numbers refer to the corresponding steps of scheme 1. The first multiplier on the right-hand side of eqn (8) reflects the kinetics of the catalytic cycle itself. The second multiplier reflects the side reaction ( 5 a ) or ( 5 6 ) , which reversibly poisons the active sites of the catalyst.K , is the equilibrium constant for this reaction. The third multiplier characterizes the fraction of vanadium in the active Vv oxidation state. 0 is the fraction of vanadium in the inactive V" state. cp takes into account the reversibility of the reaction 2 SO, + 0, r* 2 SO3. K , is the equilibrium constant of this reaction. Eqn (8) represents the kinetics of SO, oxidation in the form usual for reactions in liquids. To transform this equation into the usual form for heterogeneous catalytic reactions, one has to express the concentrations, C,, of O,, SO, and SO3 in the melt via their partial pressures, P,, in the gas phase: C, = HIPi, where HI is the Henry coefficient.In this case eqn (8) transforms into where k, = k:Ho2, k, = kkHso,, a, = K , H s o ; . The rate constants in eqn (9) were determined in two different ways: (i) from relaxation studies of the kinetics of various steps of the catalytic cycle [e.g. fig. 7(6) and 81 and (ii) from the overall set of relaxation kinetic studies and conventional steady-state kinetic studies at various temperatures and compositions of the reaction mixture. The rate constants and activation energies obtained in these two different types of experiments were found to agree quite well and are summarized in table 1. Eqn (9) has been demonstrated to explain quantitatively all the regularities observed so far for the kinetics of SO2 oxidation.8 For example, under conditions that are typical for commercial plants producing sulphuric acid, reaction (3) is the rate-determining step.In this case eqn (9) can be simplified as follows where K , = k,/ k-, is the equilibrium constant for stage ( 2 ) . The term K 2 Pso,/ (1 + K 2 Pso,) characterizes the fraction of active sites that exist in the form of the sulphite V V species. In agreement with the experimental data, the reaction rate is seen to be first order with respect to 02. However, under very low concentrations of SO, and high concentrations of O,, reaction (2) becomes the rate-determining step. In this case eqn (9) transforms into i.e. the reaction rate is zero order with respect to 0, and first order with respect to SO,.Eqn ( 9 b ) has been demonstrated indeed to describe well the experimental data obtained for very low SO, and high 0, pressures.8 Good agreement of eqn (8) and (9) with numerous other experimental data has also been demonstrated.8 Thus, a combination of detailed (on the level of elementary steps) kinetic studies with a thorough characterization of the catalyst at different stages of its preparation and performance with a set of spectroscopic methods, has allowed us to gain deep insight into the mechanism of SO, oxidation over industrial vanadium catalysts. These data are important not only from the scientific, but also from the practical point of view. First, they make it possible to optimize the operation of the industrial reactors on the basis of a reliable kinetic model that seems to reflect the real mechanism of the catalyticB. S. Balzhinimaev et al. 147 reaction on the molecular level. Secondly, they provide guidelines for further improve- ment of vanadium catalysts applied to particular conditions of their operation. We thank Dr Yu. A. Lokhov for his valuable help in the i.r. experiments. References 1 G. K. Boreskov, Catalysis in the Production of Sulfuric Acid (Goskhimizdat, Moscow, 1954), in Russian. 2 G. K. Boreskov, Heterogeneous Catalysis (Nauka, Moscow, 1986), in Russian, chap. 7, pp. 201, 202, 228, 229. 3 B. S. Balzhinimaev, V. E. Ponomarev, G. K. Boreskov a n d A. A. Ivanov, React. Kinet. Catal. Lett., 1984, 25, 219. 4 G. K. Boreskov, G. M. Polyakova, A. A. Ivanov a n d V. M. Mastikhin, Dokl Akad. Nauk SSSR, Ser. Khim., 1973, 210, 626. 5 K. I. Zamaraev and V. M. Mastikhin, Colloids Su$, 1984, 12, 401. 6 V. M. Mastikhin a n d K. 1. Zamaraev, Z. Phys. Chem., 1987, 152, 59. 7 V. M. Mastikhin, 0. B. Lapina, B. S. Balzhinimaev, L. G. Simonova, L. M. Karnatovskaya and A. A. 8 A. A. Ivanov and B. S. Balzhinimaev, React. Kinet. Card Lett., 1987, 35, 413. 9 B. S. Balzhinimaev a n d A. A. Ivanov, Relaxation Methods in Heterogeneous Catalj,sis: E.yperimenta1 Ivanov, J. Catal., 1987, 103, 160. Results (Preprint, Institute of Catalysis, Novosibirsk 1985), in Russian. 10 N. H. Hansen, R. Fehrmann and N. Bjerrum, Inorg. Chem., 1982, 21, 744. 11 A. L. Chuvilin, B. L. Moroz, V. I . Zaikovskii, V. A. Likholobov a n d Yu. I. Yermakov, J. Chem. Soc., Chem. Commun., 1985, 733. 12 V. E. Ponomarev a n d A. M. Ketov, React. Kinet. Catal. Lett., 1981, 18, 229. 13 B. S. Balzhinimaev, V. M. Mastikhin and A. A. Ivanov, Rasplauy, 1987, 1, 100 (in Russian). Paper 9/00383E; Receiued 23rd January, 1989

ISSN:0301-7249    

DOI:10.1039/DC9898700133

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

Furuduy Discuss. Chem. SOC., 1989, 87, 149-160 Interaction of Chromocene with the Silica Surface, and Structure of the Active Species for Ethene Polymerization Adriano Zecchina,” Giuseppe Spoto and Silvia Bordiga Dipartimento di Chimica Inorganica, Chimica Fisica e Chimica dei Materiali dell’ Universita di Torino, V. P. Giuria 7, 10125 Torino, Italy The anchoring process of Cr(Cp)? on silica hydroxyl groups occurs by elimination of CSH, and formation of SSi-0-CrCp mononuclear species. These anchored species are then able to adsorb incoming Cr(Cp), to give catalytically inactive dimeric species. These reactions are heavily diffusion- controlled. Interaction of CO and NO with mononuclear anchored species gives well defined dicarbonylic and dinitrosylic compounds, while the reac- tion of CO with dinuclear (catalytically inactive) species gives much more complex polycarbonylic compounds.The SSi-0-CrCp mononuclear surface species are the active sites for ethene polymerization. The chain- initiation mechanism probably consists of the formation of a metallocyclic structure. Many mechanistic and structural problems associated with ethene polymerization on supported chromium catalysts [Phillips Cr/ Silica and Union Carbide Cr( Cp)’/ Si02 catalysts]’-’ are far from being completely understood. In particular the structure of the active centres and of the mechanism of chain initiation and propagation on the Cr/Silica (Phillips) catalyst have been debated for over 30 years.‘” Less attention has been paid to the Cr(Cp),/SiO, (Union Carbide) catalyst, as only a few contributions can be found in the more recent literature.’ Our group has extensively studied the surface chemistry of the Cr/SiOz system in the hope of elucidating the structure of the catalytic centres and understanding the major features of the polymerization mechanism.’ However, even on simplified versions of the Phillips catalyst, many problems remain unresolved and require further experi- mental efforts.In view of the similarity of the Phillips and Union Carbide catalysts (same support, same metal in identical, or similar, oxidation state) we thought that, in addition to continuing work on the Cr/ Silica system, some general information concern- ing common (and hence general) features could be achieved from a comparison of the two catalysts.In this paper we report our first spectroscopic results on the structure of the active sites in the Union Carbide catalyst and on the polymerization of ethene occurring on them at ca. 320 K. Experimental The Cr(Cp),/Si02 catalyst was prepared directly in a suitably designed i.r. cell by room-temperature vacuum sublimation and subsequent adsorption from the gas phase of Cr( Cp), onto an SiO, pellet (Aereosil, surface area 380 m’ g-I) previously outgassed at 973 K under high vacuum (lo-’ Torrt). t 1 Torr = 101 325/760 Pa. 149150 Chromocene on Silica and Ethene Polymerization 1.0 e, C f sl % 0.1 4( 10 3500 3000 2500 2000 1500 1000 500 wavenumber/cm-' Fig. 1. ( a ) 1.r. spectrum of a silica sample outgassed at 973 K. ( h ) 1.r. spectrum of the same silica sample contacted with Cr( Cp), vapour.The i.r. spectra of the adsorbent and of the adsorbed species have been recorded in transmission mode with a Bruker IFS 113V F.t.i.r. spectrometer (4 cm-' resolution). The u.v.-visible-n.i.r. spectrum has been recorded in the diffuse reflectance configur- ation on a Varian 2390 spectrophotometer equipped with a diffuse reflectance attachment. Results and Discussion The Anchoring Process In fig. 1 the spectrum of a silica sample activated in L'acuo at 973 K is shown together with that of the same sample after interaction with Cr(Cp), (from the gas phase). The following can be seen: (i) the narrow peak at 3748 cm-', which is due to the stretching mode of isolated surface hydroxyl groups of silica, is greatly but not completely eroded; (ii) a low intensity and broader (AV,,,,== 100 cm- ' ) band centred at 3590 cm-' is formed. Owing to its position and half-width, this band can safely be attributed to perturbed (hydrogen-bonded) OH groups; ( i i i ) new peaks appear at 3099 cm-' (weak and complex) and 1424 cm-' (very weak).On the basis of the literature datax-'" they are assigned to the v ~ . ~ and vCc- modes, respectively, of a Cp ring. In some experiments, additional weaker peaks were observed in the 3000-2800 and 2100-1950 cm-' regions. Because they were not present when fresh chromocene was used, we conclude that they are associated with small variable amounts of impurities derived from the decomposition of Cr(Cp),. As the reactivity of the SiO,/Cr(Cp), system towards CO, NO and C2H4 was the same in either presence or their absence, they will not be further discussed.In a few experiments a large excess of chromocene was dosed by increasing the time of exposure of the sample to the Cr(Cp), vapour; however, we never succeed in consuming all the OH groups of the surface.A. Zecchina, G. Spoto and S. Bordiga 151 The disappearance of the silanols and the appearance of the Cp- vibrations can undoubtedly be explained by the reaction: SSi-OH +Cr(C,H,), - ZSi-OCrC5H5+C5H6 ( 1 ) whi -h has already been proposed for the interaction of surface silanols with chromocene in organic s ~ l u t i o n s . ' ~ ~ The incomplete consumption of the OH groups, even under an excess of reactant, is not a consequence of incomplete availability of the hydroxyls towards the interaction with gaseous rnolecules (for instance because of their location in narrow pores).In fact it has been known for many years"." that the free OH groups of Aereosil can interact with adsorbed species. In the following, we shall develop considerations which demonstrate that this is a kinetic (diff usion-controlled) effect. In fact the heavy Cr(Cp), vapour impinging the pellet from both sides (see scheme 1, where a section of the pellet is shown) is immediatelj adsorbed (both chemically and physically) onto the surface of the microparticles, and can thus form a penetration front which divides the pellet into two parts. The first part is fully saturated, or even super-saturated, by the Cr(Cp), vapour and is characterized by the presence of both anchored [following reaction( 1 ) ] and weakly and/or physically adsorbed chromocene.In this saturation region all the OH groups have been consumed. The second part is characterized by absence of penetration of the vapour: the surface of the silica microparti- cles is completely free from adsorbed chromocene and the surface silanols are unreacted. Between these two regions a narrow interface can exist where the Cr concentration changes abruptly from supersaturation values to zero. On this basis it is evident that the intensity of the unreacted OH groups gives information about the extension of the Cr( Cp),-free region of the pellet. In the boundary layer, dividing the super-saturated and clean regions, unreacted OH and anchored chromocene coexist: only in this narrow region does the possibility exist of finding unreacted silanols interacting with adjacent anchored Cr(Cp), via hydrogen bonding. If the boundary layer is very thin, the hydrogen bonding will involve only a very small fraction of hydroxyl groups (as observed i n the experiment). As a matter of fact, if after Cr(Cp), adsorption the silica pellet is cut perpendicular to the main faces, the presence of an inner white part and of external red-coloured layers is immediately noted.Moreover, the boundary between the regions is sharp. On this basis we consider the previously advanced hypothesis as the correct one, as it explains simultaneously: ( i ) the incomplete OH consumption; (ii) the small hydrogen-bonding effects; ( i i i ) the existence of incomplete diffusion and the appearance of regions which are red and white. The presence in the external layers of both reacted and unreacted chromocene can be confirmed by u.v.-visible-n.i.r.reflectance spectroscopy (a typical spectrum is shown in fig. 2). In fact, by comparison with the known spectrum of molecular Cr(Cp), in homogeneous conditions,'3 the peaks at ca. 29 000 (shoulder) and 24 000 cm-' are assigned to the L- M (charge transfer) and d-d transitions of unreacted or weakly perturbed chromocene. The remaining bands at 36000 (strong) and ca. 18 000cm ' (weak and broad) are assigned to analogous L - M and d-d transitions of the anchored clean pellet Scheme 1 front152 Chromocene on Silica and Ethene Polymerization I I 1 5 4 3 2 1 wavenumber/ lop4 cm-' Fig.2. U.v.-visible-n.i.r. reflectance spectrum of Cr(Cp), adsorbed on Si02 outgassed at 973 K. species. Owing to the lower strength of the total ligand field in the anchored species (vide infra), the d-d transition occurs at lower frequency. The existence of weakly adsorbed chromocene in the external layers of the pellet is also confirmed by a simple final outgassing experiment. In fact, if after adsorption at room temperature, the sample is outgassed under mild conditions (323 K), a fraction of weakly adsorbed chromocene is desorbed and is found unmodified as a dark-red (the colour characteristic of the chromocene microcrystals) condensation ring on the internal wall of the liquid-nitrogen trap of the vacuum line. At the same time the intensity of the i.r. modes of the Cp- ring decreases and the band due to weakly perturbed OH groups disappears. On the basis of all these considerations, scheme 2 is proposed for the external regions of the silica surface after interaction with Cr(Cp), where the weakly adsorbed Cr(Cp), is supposed to interact weakly with the anchored, highly coordinatively unsaturated, CrCp moiety to form more fully coordinated species. The weakness of this interaction explains the observed downward shift of the d-d transition of the anchored species with respect to Cr(Cp),.In the boundary regions (scheme 3) several species coexist, including hydrogen- bonded species and coordinatively unsaturated Si -0-CrCp groups. Cr(Cp), does not penetrate the inner regions (scheme 4) because of incomplete diffusion. Schemes 2 and '4 predominate.Scheme 2A. Zecchina, G. Spoto and S. Bordiga 153 Scheme 3 /H 0 I /H 0 I /H 0 I Scheme 4 Desorption of Cr(Cp), under mild conditions increases the fraction of coordinatively unsaturated species with respect to fully coordinated species. As will be shown later, this process is far from being complete at 323 K. Moreover, the process cannot be forced to completion by using higher outgassing temperatures because decomposition processes begin to occur and these complicate the situation instead of simplifying it (results not reported for the sake of brevity). The Structure of the Adsorbed Species (as probed by CO and NO) Adsorption of CO on a sample previously contacted with Cr(Cp), at room temperature gives the spectrum illustrated in fig.3( a ) , while the adsorption of CO on a sample which was first contacted with Cr(Cp)2 and then evacuated at 323 K, in order to remove part of weakly adsorbed chromocene, is shown in fig. 3(6). On the basis of the considerations developed above, the two samples differ in the relative amounts of the species (a) and ( b ) in scheme 5 whose concentrations are C, >> Cb for the first case and C,, 3 C, for the second case. On this basis, it is expected that the spectrum of the first sample should be dominated by the bands associated with the carbonylic complexes mainly derived from ( a ) by reaction with CO, while the spectrum of the second sample should be dominated by the bands of the carbonylic complexes derived from species (6) and/or from ( a ) by expulsion of Cr(Cp), [a process which could be favoured by lower surface concentration of weakly bound Cr(Cp),] (vide infra).A band pair at 2054 and 1985 cm-' predominates in fig. 3(6), while a sextet at 1920, 1895, 1832, 1775, 1628 and 1579 cm-' is more abundant in fig. 3(a). As the intensity of the components of the doublet at 2054 and 1985 cm-' decreases or increases in a strictly parallel way by decreasing or increasing the CO pressure in the 40-0 Torr interval, it is inferred that they are associated with a single simple reversible complex, having cis- dicarbonylic structure, derived from ( a ) and/or (6) species as shown below: tCO -CO SSi-0-CrCp SSi-O-CrCp(CO)2 SSi-O-CrCp..-CpCrCp SSi-O-CrCp(CO),+ CpCrCp. Although it is probably less important, the second reaction must be considered also, because in a CO atmosphere the equilibrium of the ligand-substitution reaction could be shifted to the right simply by mass-action effects when the surface concentration of weakly adsorbed Cr(Cp), is not too large.(2) tC'O -CO154 0.5 e, E: -2 2 D 0.0 Chromocene on Silica and Ethene Polymerization I li 1 A 2200 2000 1800 1600 1400 wavenumber/ cm- I 0.5 0.0 2200 2000 1800 1600 1400 wavenumber/cm-' Fig. 3. ( a ) 1.r. spectrum obtained after dosing 40 Torr CO onto a freshly prepared Cr(Cp),/SiO, sample. ( b ) Spectrum obtained after dosing 40 Torr CO onto a Cr(Cp),/Si02 sample previously outgassed for 1 h at 323 K. I n both spectra the peak at 1424cm-' is due to the ucc mode of a Cp- ring. On the basis of the intensity ratio of the two peaks, the angle formed by the two oscillators is ca.85 '.I4 The sextet of bands at 1920, 1895, 1832, 1775, 1628 and 1579 cm-' derives from the interaction of CO with the ( a ) dimers in an excess of Cr(Cp),. As the ( a ) dimers are not active in ethene polymerization (vide infra), we shall not discuss into detail the structure of the carbonylic complexes derived from them. We mention only that: ( i ) the sextet is the sum of two correlated triplets (1920, 1832, 1628 cm-* and 1895, 1775, 1579 cm-I), presumably belonging to two different (polycarbonylic) species; ( i i ) the ratio between the two triplets changes slightly from one sample to the other and with time, Unlike the doublet at 2054 and 1985 cm-', the two triplets at lower frequencies do not change in intensity on decreasing the equilibrium pressure of CO.It is thus inferredA. Zecchina, G. Spoto and S. Bordiga 155 1 .o 0, C -e 0.5 0, 2 0.0 n ' * c 0 I I3c 0 1 3 ~ ~ ~ / 2200 2000 1800 1600 wavenumber/cm-' Fig. 4. 1.r. spectra recorded after contacting Cr(Cp),/SiO, samples previously outgassed at 323 K with: ( a ) 40 Torr "CO; ( b ) 40 Torr I3CO; ( c ) 40 Torr of a '2CO/'3C0 1 : 1 mixture. that the corresponding carbonylic species are more stable. It is also worth mentioning that the rate of formation of the dicarbonylic species is larger than that of the polycar- bonylic species. Further insight into the structure of the carbonylic species is obtained from the isotopic '2CO/"C0 substitution experiments (fig. 4). It can be seen that when the 1:l mixture is used, the doublet at 2054 and 1985 cm-' splits into three doublets (two bands of the '2CO-'2C0 complex, two bands of the mixed '2CO-'3C0 species and two bands of the '3CO-'3C0 species) with 1 : 2 : 1 approximate intensity ratios.These results definitely confirm the cis-dicarbonylic nature of the species hypothesized in reaction (2). The isotopic substitution pattern of the two triplets derived from the dinuclear species ( a ) is more complex and it will not be discussed in detail. It is sufficient to mention here that the spectrum of the 1 : 1 mixture is not the sum of the spectra of the pure "CO and "CO species. This indicates considerable coupling between the CO oscillators within these species. In particular, the splitting of the two higher-frequency peaks at 1920 and 1832 cm-' suggest that two preferentially coupled CO groups are present in the polycarbonylic polynuclear complex.In a separate study we shall advance a more detailed hypothesis on their structure. We can anticipate here that we are probably156 Chromocene on Silica and Ethene Polymerization 1 .o 0 c .f! s % 0.0 1 , 2000 1900 1800 1700 1600 1500 wavenumber/cm-' Fig. 5. 1.r. spectrum obtained after contacting a Cr(Cp),/SiO, sample with 0.5 Torr NO. dealing with a salt-like dimeric species [ 2Si-O-Cr(CO),Cp]- [ Cr( Cp)J+ and/or [ Cr( CO),]-[ Cr( Cp)J+ similar to that described in ref. ( 15) and ( 16). The interaction of NO with adsorbed Cr(Cp), (fig. 5) is simpler than that of CO, as only two main peaks (with constant intensity ratio) are always seen at 18 13 and 1707 cm-' (with a shoulder at 1793 cm-') on both concentrated and diluted samples.This result can be explained by assuming that the two peaks belong to a cis-dinitrosylic complex derived from (a) and (6) as follows: +NO SSi-0-CrCp - ~Si-O-CrCp(NO)2 %3-O-CrCp-.CpCrCp - fSi-0-CrCp( NO),+ CpCrCp. Unlike CO, NO (being a stronger ligand) displaces all the weakly bound Cr(Cp), (even in the presence of an excess of it) without giving further dinuclear polynitrosylic products. It is worth noting that the frequency and intensity ratio of this doublet are very similar to those observed for cis-dinitrosylic complexes formed on Cr"/~ilica.'~ The previous assignment is consequently strongly reinforced. +NO The Ethene Polymerization Centres As is well ethene polymerizes on SiOz samples functionalized with Cr(Cp),, and the reaction can be easily investigated by i.r.spectroscopy following the evolution (in the presence of ethene) of the i.r. spectrum of the growig polymeric chain.A. Zecchina, G. Spoto and S. Bordiga 157 4000 3500 3000 2500 2000 1500 wavenumber/cm- ' Fig. 6. 1.r. spectra obtained at increasing CzH4 contact time (last spectrum recorded after 20 min) illustrating the polymerization reaction on a Cr(Cp)2/ SiO, sample previously outgassed at 323 K. The spectrum of the SO2 pellet (activated at 973 K) before anchoring chromocene is also reported for comparison. A typical spectrum is shown in fig. 6 for a sample previously contacted with Cr(Cp), and then outgassed at 323 K to remove a fraction of the weakly adsorbed chromocene.It is remarkable that the rate of the reaction (not very large under this ethene pressure) is much lower on samples which did not undergo the outgassing procedure at 323 K. This observation indicates that Cr(Cp), bound weakly to the anchored (6) structures acts as a poison. This is not unexpected because the Cp- ligand of chromocene can fill the coordination vacancies that are necessary for the polymerization. Moreover, in the conditions investigated here (it is conceivable that under pressures of the order of several atmospheres the situation could be different) ethene, unlike CO and NO, is not strong enough a ligand to displace the Cr(Cp), from the ( a ) species. It is noticeable that the polymerization is also totally poisoned by CO and NO.Following these considerations, we conclude that the polymerization centres are the mononuclear coordinatively unsaturated ( b ) species. For the reasons discussed above, the ( b ) centres are very scarce on freshly prepared samples (because of diffusion problems) and are present only in the boundary layers between supersaturated and clean regions of the pellet. This explains the low polymeriz- ation activity (at least at the low ethene pressures used in this investigation) of the freshly prepared samples. Outgassing at 323 K partially removes the weakly adsorbed poisoning chromocene: this explains the increase of the polymerization activity observed after the thermal treatment. Unfortunately, as mentioned before, removal of all the weakly adsorbed Cr(Cp), is not possible without causing undesired chemical transforma- tions.This makes the problem of building a system with maximum catalytic activity apparently impossible. A way to overcome this problem is illustrated in fig. 7, where the effect of anchoring Cr(Cp), in an ethene atmosphere is reported. The experimentChromocene on Silica and Ethene Polymerization 4000 3500 3000 2500 2000 1500 wavenumberjcm- ' Fig. 7. 1.r. spectra recorded at increasing contact time (last spectrum after 20 min) illustrating the effect on the ethene polymerization rate of anchoring chromocene on silica in presence of 70 Torr C,H, following the experimental procedure described in the text. The spectrum of the virgin Si02 sample (outgassed at 973 K) is also reported for comparison.As in fig. 6, the peaks in the 2100-1950 cm-' region are due to adsorbed impurities (see text). Notice that the absorbance smle is the same as in fig. 6. was performed as follows: a silica pellet outgassed in vacuo at 973 K in the usual way was brought in the immediate vicinity of a Cr(Cp), film sublimed on the internal wall of the i.r. cell and kept a t low temperature (ca. 263 K) to avoid vaporization (and hence interaction through the gas phase with the OH groups of the silica pellet). Then C2H, was dosed by filling the cell with 60Torr of gas. The temperature of the film was then allowed to increase rapidly up to 323 K in order to cause fast sublimation of the metallocene on the adjacent silica pellet in the presence of ethene. After this operation the i.r.spectra were recorded at fixed time intervals. From the results reported in fig. 7, the following interesting facts emerge: ( i ) the surface OH groups are nearly completely consumed ensuring maximization of the anchoring reaction and hence formation of the maximum number of potentially active centres); (ii) the polymerization activity is much increased with respect to the previous experiments. The explanation of these phenomena is as follows. The anchoring process occurs in the presence of ethene in the usual way by elimination of C,H, and formation of the ( b ) structure. However, when ethene is present, the freshly formed ( b ) structures begin to act immediately as catalytic centres with subsequent fast growth of the polymeric chain attached to the Cr atom (scheme 6).The incoming Cr(Cp), finds the coordination vacancies already saturated by the growing polymeric chain. Consequently, it cannot be adsorbed on the pre-existing anchored chromocene sites and cannot poison the catalytic centres. Moreover, in absence of these sites, the excess chromocene can now more freely migrate inside the pellet and react with almost all the OH groups of the microparticles. In conclusion, the higher consumption of OH groups and the absence of any negative effect of unreacted chromocene is simultaneously explained on the basis of the same concept.A. Zecchina, G. Spoto and S. Bordiga 0 1 ,7i\ 0 C2H4, I Cr-Polymeric chain / + Scheme 6 @ A Cr 159 J Scheme 7 The Polymerization Mechanism As shown in fig. 6 and 7, the main features of the growing polymeric chain are represented by peaks at 2918 and 2851 cm-' (asymmetric and symmetric stretching frequencies of CH2 groups) and peaks at 1472 and 1465 cm-' (bending modes of the same groups).This spectrum is very similar to that observed on Cr"/Si02 (Phillips catalyst). However, unlike this system, no broad bands are observed at ca. 2750 cm-', which were assigned to an agostic interaction.* It is remarkable that, even during the first polymerization stages, no definite signs of the presence of groups other than CH2 were found. These observations suggest that the chain initiation consists of the formation of a metallocyclic structure following the mechanism in scheme 7 and that the propagation corresponds to an insertion of the C2H4 molecules into the ring following the mechanism in scheme 8 with the formation of long, doubly anchored, chains without end groups.Following this idea chain termina- tion (not observed in our low-pressure, low-temperature experiments) can be represented by scheme 9. The major difficulty with this reaction scheme is the following: how can ethene be inserted into the metallocycle when the Cr atom is apparently fully saturated (in fact, as the C5HS usually behaves an an 77' ligand occupying three coordination vacancies, the Cr is sixfold coordinated). A plausible answer lies in the well established fluxionality of the Cp- ligand (especially with Group VI metal derivatives)I6 which can change from 7' to q3 configur- ation (ring slippage). In fact this transformation could allow the ethene to be coordinated Scheme 8160 Chromocene on Silica and Ethene Polymerization Scheme 9 by the metallic centre before insertion into the metallocycle.Moreover, we have to consider also the fact that the Group VI metals give a rich variety of seven-coordinated compounds. '*,I9 It is quite conceivable that both these factors operate simultaneously here, thus giving a highly versatile catalytic character to the metallic centre of structure ( 6 ) . We thank Prof G. Ghiotti and Prof E. Garrone for useful discussions. This research was supported by Ministereo della Pubblica Istruzione, Progetti di Rilevante Interesse Nazionale. References 1 M. P. McDaniel, Adu. Catal., 1985, 33, 47. 2 G. Ghiotti, E. Garrone and A. Zecchina, J. Mol. Catal., 1988, 46, 61 and references therein. 3 F. J. Karol, G. L. Karapinka, C. Wu, A. W. Dow, R. N. Johnson and W. L. Carrick, J. Polym. Sci., 4 F. J. Karol, G. L. Brown and J. M. Davidson, J. Polym. Sci., Part A I , 1973, 11, 413. 5 F. J. Karol and C. Wu, J. Polym. Sci., Part A I , 1975, 13, 1607. 7 B. Robenstorf and R. Larsson, J. Mol. Catal., 1981, 11, 247 and references therein. 8 V. T. Aleksanyan, B. V. Lokshin, G. K. Borisov, G. G. Deviatykh, A. S. Smirnov, R. V. Nazarova, J. A. Koningstein and B. F. Gachter, J. Organornet. Chem., 1977, 124, 293. 9 V. T. Aleksanyan, I. A. Garbuzova, V. V. Gavrilenko and L. I. Zakhakin, J. Organomet. Chem., 1977, 129, 293. Part A l , 1972, 10, 2621. 10 V. T. Aleksanyan and V. B. Lokshin, J. Organomet. Chem., 1977, 131, 113. 11 A. Zecchina, G. Ghiotti, L. Cerruti and C. Morterra, J. Chem. Phys., 1971, 68, 1479. 12 G. Ghiotti, E. Garrone and F. Boccuzzi, J. Phys. Chem., 1987, 91, 5640. 13 K. R. Gordon and K. D. Warren, Inorg. Chem., 1978, 17, 987. 14 P. S. Braterman, Metal Carbonyl Spectra (Academic Press, New York, 1975). 15 B. Longato, B. D. Martin, Y. R. Norton and 0. P. Anderson, Inorg. Chem., 1985, 24, 1389. 16 R. D. Fisher, Chem. Ber., 1960, 93, 165. 17 A. Zecchina, E. Garrone, C. Morterra and S. Coluccia, J. Phys. Chem., 1975, 79, 978. 18 J. M. O'Connor and C. P. Casey, Chem. Reu., 1987, 87, 307 and references therein. 19 R. P. A. Sneeden, Organochromium Compounds (Academic Press, New York, 1975). Paper 8/04954H; Received 19th December, 1988

ISSN:0301-7249    

DOI:10.1039/DC9898700149

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

Furuduy Discuss. Chern. Soc., 1989, 87, 161-172 GENERAL DISCUSSION Dr R. Burch (University of Reading) began the discussion of the paper by Waller et al.: In the very last sentence of your paper you comment that the activity (of Cu catalysts) for reverse shift is a better guide to methanol synthesis activity than the N,O-determined copper area. I should like to add that in our experience the N,O-determined copper area is also a poor guide to the activity of supported copper catalysts for methanol synthesis from CO-C02-H2 mixtures. We have recently reported’ a strong support effect when Cu/ZnO catalysts are compared with Cu/Si02 catalysts. Furthermore, we have observed2 that there is a synergy between Cu and ZnO even when they are physically separated. We have concluded that ZnO may play a much more active role in the synthesis of methanol than has generally been realised.Do you consider that a similar active participation of ZnO in the reverse shift reaction could account for your observa- tions? 1 R. Burch and R. J. Chappell, Appl. Catal., 1988, 45, 131. 2 R. Burch, R. J. Chappell and S. E. Golunski, Catal. Lett., 1988 1, 439. Prof. F. S. Stone (University of Bath) replied: Our results certainly do not suggest that N,O-determined copper area is a poor guide to methanol synthesis activity. It is more a question of relatively small uncertainties among the various ways of carrying out the N 2 0 experiment. We have no evidence to suggest that active participation of ZnO can account for our observations of the beneficial effects of ageing or the slight decline in activity for the catalyst derived from the 205 min aged precursor indicated in fig.7-9. Dr K. C. Waugh ( I . C. I. C&P Group, Billingham, Cleveland) (communicated): I should like to make some observations on Dr Burch’s comment of Prof. Stone’s elegant paper. Dr Burch reported in his comments that the methanol synthesis activity of a physical mixture of silica-supported copper and silica-supported zinc oxide was greater than that of the individual components when tested alone. The result is as intriguing as it is difficult to explain. However, I believe that Dr Burch’s rationale that it betokens some unique synergy, albeit at a distance, between copper and zinc oxide, is both premature and fallacious. The main reason for declaring it to be premature is that Dr Burch has looked at Cu/Si02, ZnO/Si02 mixtures only.In the absence of his having tested Cu/Si02 physically mixed with MnO/ S O z or A1203 or MgO/ SO, etc. he cannot logically conclude as to a ‘unique’ synergy. He has countered the suggestion that there might be some migration of the zinc oxide to the copper during the reaction by separating the Cu/Si02 after the reaction and showing its post- and pre-reaction activities to be the same within experimental error. This is exactly as I would have thought since the suggestion pre-supposes some form of synergy between the copper and the zinc oxide which we and other groups have shown not to exist.’-3 It was the culminating conclusion of a vast body of research in I.C.I. that the methanol synthesis activity was a linear function of the copper metal area and that this conclusion held regardless of the nature of the support or lack of it,’ fig.1. We should not and cannot abandon this work when faced with Dr Burch’s interesting result. Rather this information should be used as a basis of explanation. Taking it as read that the methanol synthesis activity is a function of the copper metal area the only explanation of Dr Burch’s work is in a re-distribution of the copper 161162 General Discussion 12 .o 1 1 .o 10.0 9.0 8 .O 7.0 6 .O 5 .o 4 .o 3 .o 2 .o 1 .o 10 20 30 40 Cu-metal area/m2 g-’ Fig. 1. Methanol synthesis activity as a function of copper-metal area. a, CuO/ZnO/A1,0, (60: 30: 10); +, CuO/ZnO/AI20, (45: 37: 18); 0, CuO/SiO,; A, CuO/A1203; A, CuO/MgO; 0, CuO/MnO; H, CuO/ZnO.over the surface of the ZnO/Si02 during the course of the reaction; post-reaction examination of this material was not done. However, even if it had been done the result might well have been misleading, since in order to separate the Cu/SiO, from the ZnO/ Si02, it would have been necessary first to passivate both materials by controlled oxidation, a process which could induce sintering in the newly formed Cu/ZnO/SiO,. Necessarily, therefore, the experiment which should have been carried out, is the in situ measurement of the post-reaction copper-metal area by whatever means. The absence of such a measurement allows us endless speculation. 1 K. C. Waugh, Appl. Catal., 1988, 43, 315. 2 M. Bowker, R. A. Hadden, H. Houghton, J.N. K. Hyland and K. C . Waugh, J. Catal., 1988, 109, 263. 3 W. X . Pan, R. Cao, D. L. Robberts and G. L. Griffin, J. Catal., 1988, 114, 440. Dr R. Burch (University of Reading) responded: In reply to Dr Waugh’s comments we should like to make the following remarks. ( a ) We have not used the expression ‘unique’ synergy. Indeed it would be foolish to do so since we have data showing that, for example, Ga,O, also promotes Cu catalysts for methanol synthesis. ( b ) The question raised concerning the transfer of Cu to ZnO/Si02 catalysts can be answered easily. Post-reaction chemical analysis of the ZnO/ Si02 particles failed to detect any Cu on these particles. Therefore, it is not appropriate to suggest, as Dr Waugh has done, that ‘the only explanation’ of our results is ‘a re-distribution of the Cu over the surface of the ZnO/SiO,’.Indeed, our analytical measurements indicate that this is the least likely explanation of our results. ( c ) Dr Waugh comments that they and other groups have shown that synergy between Cu and ZnO does not exist. Currently, this appears to be a minority view since several groups’-5 have independently reported support effects with Cu catalysts for methanol synthesis. Furthermore, the work on Cu/rare earth intermetallics”’ clearly shows no correlation between Cu surface area and methanol activity. 1 B. Denise, R. P. A. Sneeden, B. Beguin and 0. Cherifi, Appl. Catal., 1989, 30, 353. 2 W. R. A. M. Robinson and J. C. Mol, Appl. Catal., 1988, 44, 165.General Discussion 163 3 J. C . Frost, Nature (London), 1988, 577.4 M. S. W. Vong, M. A. Yates, P. Reyes, A. Perryman and P. A. Sermon, 9th Int. Congr, Catal., Calgary, 5 J. Barbier, Th. Fortin, Ph. Courty and P. Chaumette, Bull. Chim. SOC. France, 1987, 925. 6 G. Owen, C. M. Hawkes, D. Lloyd, J. R. Jennings, R. M. Lambert and R. M. Nix, Appl. Catal., 1987, 7 R. M. Nix, T. Rayment, R. M. Lambert, J. R. Jennings and G. Owen, J. Catal., 1987, 106, 216. 1988, p. 545. 33, 405. Dr R. B. Moyes (University of Hull) said: According to fig. 7 the activity for the test reaction reaches a maximum with ageing time in the mother liquor of the precursor to the catalyst. I should like Prof. Stone to speculate on the structural possibilities which might explain this result. Prof. Stone answered: The ageing process in the mother liquor will not involve a change in overall Cu/Zn ratio of the malachite once all the aurichalcite has disappeared.However, it will lead to a more uniform distribution of the Cu” and Zn2+ ions in individual crystallites, and I see that as advantageous for achieving high Cu dispersion in Cu/ZnO after decomposition and reduction. Extended ageing will presumably lead to ripening in the zincian malachite suspension. The latter is likely to be deleterious for surface area in the final catalyst, and there is evidence for this in the measured areas reported in the paper. I ascribe the maximum in the activity of the catalyst with precursor ageing time to these effects. Prof. J. B. Moffat ( University of Waterloo, Ontario, Canada) then said: I believe that you have put your finger on an aspect of catalyst preparation which is too infrequently studied.In your work on the effects of precursor ageing, you observed, at least during the first 30 min of the process, the production of more finely divided solid. My belief is that d G / d A cannot be negative. Would you care to comment on this from a surface chemistry point of view? In other words, what is the driving force for a reduction in particle size? Prof. Stone replied: Prof. Moffat asked about the paradox of a decreased particle size on ageing. The driving force for this is the change in chemical composition of the malachite phase on ageing, the increased zinc content leading to a more stable structure, whose small crystallites form rapidly. Prof. M. Ichikawa (Hokkaido University, Sapporo, Japan) commented: It is important to choose the catalyst precursors for the unique metal centres of CuO-ZnO in methanol synthesis as reported by Prof.Stone et al. However, I would like to concentrate much more on the further structural changes of the catalysts upon admission of the reactants such as CO + H2 and how to reach the static states of the catalysts in the working conditions, resulting in the production of the new metal centres involving the methanol synthesis. According to the previous work by Ueno and co-workers’ by EXAFS evaluation for Cu/ZnO catalysts, the size and morphology of Cu clusters on ZnO are essentially affected (changeable) by the presence or absence of CO and/or H2. Thus, I wonder whether the well characterised metal centres on your ZnO-CuO catalysts prepared from some particular precursors will retain their characteristic structures and chemical compositions under the working conditions of syngas.1 K. Tohji, Y. Udagawa, T. Mizushima and A. Ueno, J. Phjx Chem., 1985, 89, 5671. Prof. Stone added: Prof. Ichikawa is quite right to stress that changes in the size and morphology of the copper crystallites in the Cu/ZnO catalyst under reaction164 General Discussion 20 LO 60 80 100 0 t/min Fig. 2. Activity of 8% Cu/ZnO in CH30H decomposition at 429 K (a) and CO/ H2 at 523 K (b) [with X-ray diffraction patterns before ( c ) and after (d) use] showing catalysis-induced appearance of Cuo(lll) at 42.5" 28 (shaded). conditions need to be borne in mind. On long-term usage there is a decrease in methanol synthesis activity.However, the precursor ageing effect still shows through, even after those considerations have been applied. Both factors are therefore important. Dr P. A. Sermon (Brunel University) (communicated): (i) We have observed with Cu/ZnO that activity for methanol synthesis (and decomposition) changes with time on stream (fig. 2), producing additional Cuo (fig. 3). (ii) Is there a clear relationship between activities in C02/H2 at 1 bar, and CO/C02/H2 at 50 bar, confirming the pivotal role of C02? (iii) In your fig. 1 what percentage of your catalysts is sufficiently crystalline to be X-ray detectable ? Prof. Stone replied: With regard to the first point raised by Dr Sermon, whilst this increase in Cuo may occur in the absence of C 0 2 , it is very unlikely that with C 0 2 present in the feed there is an increase in the amount of Cuo with time.This is not in any sense to deny that CO may not be a better reductant for copper oxide than H2. On Dr Sermon's second point, I would argue that there is indeed a clear relationship in that there is a good correlation between the results in fig. 6 and 7 for the catalysts derived from precursors aged up to 140min.General Discussion 165 01 I I I I I I I 30 60 90 120 150 180 time/min Fig. 3. Reduction of CuO at 423 K with 6 kPa H2 (0) and CO (0). Clearly CO is a faster and more effective reductant for CuO to Cu (22.5% CuO remaining) than H2 (32.5% CuO remaining). As to the proportion of our catalysts which are X-ray detectable, the EM results [our plate l ( b ) ] imply that all the material in the unaged precursor is sufficiently crystalline to be seen by X-rays.There is evidence, however, from the intensities of the malachite pattern that some of the material in the samples aged for 30min may be undetectable by X.r.d. There is also some indication for this from the particularly high SA obtained after calcination of this precursor (our table 1). Prof. V. Ponec (Gorlaeus Laboratory, Leiden University, The Netherlands) made the first comment on the paper by Lambert et al: My comment concerns fig. 1B. It looks here as if Cu were already present but not yet active and needed some induction period to be activated to start to produce CH30H. CH, is produced first and this means H 2 0 is (probably) produced too.H 2 0 can convert Cu to CU"+ which is then stabilised by Nd20+,, the oxide. Moreover, one can imagine, also, other ways by which Gun+ can be formed to become the active centre of CH30H synthesis. My question is, what were the reasons (not mentioned in your paper) which led you to ignore the last possibility completely? Drs R. M. Nix and R. M. Lambert ( University of Cambridge) replied: The 'induction period' prior to onset of methanol synthesis is the period during which extensive solid-state transformations are occurring. Specifically, the conversion of the intermetallic precursor into an intimate intergrowth of elemental copper and rare-earth oxide crystal- lites. This activation process proceeds via the incorporation of hydrogen into the lattice of the alloy to yield both ternary alloy hydrides and, by a hydrogenolysis process, also the binary rare-earth hydride (the exact mechanisms and proportions of hydride forma- tion being very dependent upon the intermetallic precursor concerned).This behaviour is well illustrated by the representative in situ X.r.d. spectra shown in fig. 1A. Further- more, it has been shown that there is a direct correlation between the rise in synthesis activity and the concurrent growth of the (X.r.d.-visible) copper and rare-earth oxide phases.'166 General Discussion Methane is a by-product of the activation process and not of a catalytic reaction. It is generated during the oxidation of the various metal/alloy hydrides by reactions such as 2NdCu - H, + 3CO -+ 3CH,+ 2Cu + Nd203.Generation of water is not therefore a significant factor and, indeed, the use of water vapour as the oxidizing media for alloy activation is detrimental to the ultimate catalytic activity; the favoured activation process is an in situ activation in a syngas feed following low-temperature pretreatment in pure hydrogen. A wide range of results points to the activity of these materials being associated with very highly dispersed copper in the poorly crystalline rare earth oxide matrix (such copper is evident at levels up to 25 at.'% by EDAX)., This copper could, in principle, be present as individual atoms or ions (i.e. CU"+) but at the levels concerned it is more likely to be present as small ( < l o A) clusters; this proposition is supported by EXAFS data on Cu/Th02 catalysts.' 1 R.M. Nix, T. Rayment, R. M. Lambert, J. R. Jennings and G. Owen, J. Catal., 1987, 106, 216. 2 G. Owen, C. M. Hawkes, D. Lloyd, J. R. Jennings, R. M. Lambert and R. M. Nix, Appl. Catal., 1987, 33, 405. 3 J. C. Frost, Nature (London), 1988, 334, 577. Prof. A. Baiker (ETH Zurich, Switzerland) said: I have two questions, the first concerns the deactivation of the Nd/Cu derived catalyst in the presence of CO, shown in fig. 2B). I am wondering whether the authors have an explanation for this behaviour. It may be interesting to mention that we have not observed a similar deactivation due to the presence of CO, with catalysts prepared by in-situ activation from amorphous Cu7Zr3 alloys.' Do you have any suggestion for the different behaviour of these catalysts? The second question concerns the occurrence of segregation of the constituents during the transformation of the alloy to the final catalyst, e.g.during the oxidative decomposition of the alloy. Segregational phenomena have been reported for a number of amorphous alloys (see paper 15 in this series). Do you observe segregation, and if so, what are its consequences on the preparation of the alloy-derived catalyst? 1 D. Gasser and A. Baiker, Appl. Catal., 1989, 48, 279. Drs R. M. Nix and R. M. Lambert responded: The deactivation of the NdCu-derived catalyst upon exposure to process gas containing CO, is a feature that is common to all the RE/Cu alloy-derived catalysts'.' and also to Cu/ThO, catalysts3 The extent of deactivation and also the degree of recovery after removal of CO, from the feed varies significantly: from complete deactivation with zero recovery (e.g.Ce/Cu catalysts) to only partial deactivation and complete recovery (e.g. Cu/Th02 catalysts). In the case of the highly basic rare-earth oxide containing catalysts, isotopic labelling experiments indicate the formation of a surface carbonate and the extent of deactivation/recovery of NdCu and CeCu catalysts has been associated with the decomposition temperatures of this ~ a r b o n a t e . ~ If this reasoning can be extrapolated to catalysts obtained from amorphous Zr/Cu alloys, then the different behaviour may arise from the significantly weaker interaction of CO, with ZrO,. In this context it is interesting to note that recent work of colleagues of ours at I.C.I.has shown that catalysts obtained from ternary RE/Cu/Zr alloys are significantly less susceptible to C 0 2 poisoning than the binary RE/ Cu catalysts but can exhibit comparable activities.''5 The concept of surface segregation is well defined when looking at near-surface transformations (such as the oxidation of amorphous alloy ribbons) and, indeed, we have observed apparent surface segregation of NdO, during oxidation of ultra-thin alloy films on copper single-crystal substrates. In the case of the bulk intermetallic compounds, however, the solid is largely transformed into a microporous, intergrowth of crystallites and it is not clear to us exactly what meaning should be attached to 'surface segregation' in such systems.General Discussion 167 1 J.R. Jennings, R. M. Lambert, R. M. Nix, G. Owen and D. G. Parker, Appl. Caral., 1989, 50, 157. 2 G. Owen, C. M. Hawkes, D. Lloyd, J. R. Jennings, R. M. Lambert and R. M. Nix, Appl. Card. 1987, 3 J. C. Frost, Nature (London), 1988, 334, 577. 4 R. M. Nix, R. W. Judd, R. M. Lambert, J. R. Jennings and G. Owen, J. Catal., 1989, 118, 175. 5 G. Owen, C. M. Hawkes, D. Lloyd, J. R. Jennings, R. M. Lambert and R. M. Nix, Appl. Card., in press. 33, 405. Prof. J. Cunningham (University College Cork, Republic of Ireland) addressed the authors: My question to Drs Badyal and Nix seeks clarification concerning the suggestion on pp. 123-124 that 'a substantial amount of copper is present in another form, specifically a form that is undetected by both HREM and XRD, and also inert or inaccessible to N 2 0 titration'.Not only is that suggestion reminiscent of references to 'missing-copper' made by Herman et al. in their studies of Cu/ZnO catalysts,' but also it echoes reservations expressed in the paper by Waller et al. concerning the adequacy of N20- determined copper area. Against this background it would be helpful to have clarification from Drs Badyal and Nix as to (i) whether they consider the missing copper in the system to take the form of small particles dispersed within the oxide phase and (ii) whether they associate the N,O-measured copper surface area wholly or in part with that fraction of the particles whose surfaces partially obtrude through the oxide surface. 1 R. G. Herman, K. Klier, G. W. Simmons, B. P. Finn and H. B.Bulko, J. Card., 1978, 56, 407. Drs R. M. Lambert and R. M. Nix (University of Cambridge) replied: As indicated in our response to Prof. Ponec, we believe the excess copper ( i e . that not present in the form of the larger Cu crystallites visible by X.r.d.) to exist in the form of small copper clusters entrained in the rare-earth metal oxide matrix. Indeed, some of the highest levels of activity were exhibited by catalysts derived from NdCu alloys (exten- sively pretreated in pure hydrogen at < 100 "C) which contained no X.r.d.-visible copper particles. The results of N20 titrations on a range of alloy-derived catalysts have been described elsewhere:' the values obtained are consistent with the X.r.d.-visible copper representing only a fraction of the total amount present but also require that the excess copper is non-titratable under standard conditions (60 "C).This either could be due to almost complete encapsulation of the small clusters under reaction conditions or might arise from electronic modification of the redox properties of the small clusters as a result of coordination to the oxide matrix. Certainly, the extent to which the synthesis mechanism is associated with the rare-earth oxide surface as opposed to the small copper clusters has still to be resolved. 1 R. M. Nix, R. W. Judd, R. M. Lambert, J. R. Jennings and G. Owen, J. Card., 1989, 118, 175. Prof. M. W. Roberts ( University of Wales College of Cardin (communicated ): Could you comment on the origin of the Nd 3d5,, peak that develops at low binding energy (977 eV) on exposure of the overlayers to oxygen even at low (1.5 L) exposure.What was the O(1s) binding energy and how did this vary with oxygen coverage? Does the stoichiometry, NdO, where x = 1 .O, imply that the fast initial oxygen interaction occurs throughout the bulk of a Nd overlayer even when this is many layers thick? How was x estimated? Was it related to the calculated concentration of the higher oxidation state of neodymium, Nd2+, and the latter estimated from the Nd 3d5,, spectrum (your fig. 3)? Drs R. M. Lambert and R. M. Nix replied: The low-energy satellite peak in the Nd 3d5,, spectra at the higher O2 exposures has an energy shift and relative intensity characteristic of Nd3': such satellites are observed in many of the 3d spectra of the light rare-earth metal elements and arise from final-state configuration interaction.In neodymium metal itself the main peak corresponds to a poorly screened 4f" final state and there is negligible intensity in the well screened f"" satellite.168 General Discussion For the five-monolayer Nd film the 01s peak initially grew at constant binding energy (ca. 529.25 eV), but a small shift was observed concurrent with the appearance of the Nd satellite, ultimately to give a peak at 528.7 eV (similar behaviour was observed for Nd films of different thickness). The oxygen stoichiometry at various stages in the oxidation process can be estimated by comparison with the 0 1 s signal intensity at saturation exposure. Using this method an overall stoichiometry of NdO,-l is arrived at after 2.5 L 02.Some degree of oxygen concentration gradient through the film might be expected but all the experimental evidence suggests that this is small during the fast initial oxygen interaction with this and all other ultra-thin films ( < 5 monolayers of Nd). Prof. Roberts made the further comment in response to the authors’ reply: The Nd(3d5,,) spectra for the interaction of a 5 ML Nd film with oxygen (fig. 3A) show evidence for the presence of possibly three different states of Nd in the early stages of oxidation (after 0.6 L and 1.5 L exposure): Nd’, Nd3+ (as evidenced by intensity at 978 eV, the low energy satellite from Nd203) and a species characterised by a binding energy of 983 eV, possibly Nd2+. Under such conditions of inhomogeneity, estimating the stoichiometry of the oxidised film by comparing the intensity of the O( 1s) peak with that at saturation (presumably taken to correspond to Nd203) is not appropriate. My question was therefore to ascertain what quantitative procedure of Nd( 3d5,,) spectra analysis was followed to establish that the stoichiometry of the metal oxide overlayer corresponds to NdO, with x = 1.0.Prof. A. Zecchina (Turin University, Italy) said: The presence of TiO, islands is supposed to lead electronic charge transfer to neighbouring ruthenium atoms, which therefore bond to CO with stronger energy. Could you be more detailed on this point which is crucial for understanding the SMSI effect? In particular, why does electronic charge flow from TiO, to Ru and not vice versa? Is it a problem of semiconductor-metal junction? Is this the only way to explain the strengthening of the Me-CO bond? Drs R.M. Lambert and J. P. S. Badyal replied: Titanium deposited in an ambient atmosphere of oxygen at a pressure of 1 x lop6 Torr results in the laying down of an ultra-thin TiO, film where ‘as deposited’ stoichiometry corresponds to x = 2 at monolayer completion. X.P.S. measurements yield a value of 459.2eV for the T i ( 2 ~ , , ~ ) binding energy: this corresponds to Ti02. Annealing of TiOz films to temperatures characteristic of SMSI behaviour results in a 3.2 eV decrease in binding energy, which is consistent with the transformation TiOz --* TiO. This is further supported by LEED and Auger measurements.’ In the case of TiO? submonolayer quantities, we observe simple site-blocking on CO chemisorption; in the case of TiO, the presence of reduced titanium ions may be expected to lead to charge transfer to the neighbouring ruthenium metal with a comcomitant increase in CO binding energy, consistent with our observations.1 J . P. S. Badyal, A. J. Gellman, R. W. Judd and R. M. Lambert, Cural. Leu. 1988, 1, 41. Prof. J. B. Moffat then remarked: Miyazaki’ has carried out BEBO calculations of the interaction of a number of diatomic molecules including CO on various one- component metals and has predicted that a molecular state should exist with CO and all metal surfaces although the depth of the energy well, not surprisingly, varies with the metal. In all cases, an activation energy was predicted to separate the molecular state from the totally dissociated state.You have shown that on neodymium/coqper intermetallic compounds oxidation of the rare-earth component proceeds by dissociative chemisorption of CO and that at low Nd coverages and low temperature (<200 K) molecular chemisorption on exposed copper was also evident.Genera 1 Discussion 169 Do you inevitably see the molecular state and is this a necessary precursor to the dissociated state? How, if at all, is the observation of the molecular state related to the catalytic process? In connection with the titanium/ruthenium studies and your comment regarding hydrogen spillover do you have information on how the Ru-H energy compares with that of Ti-H? 1 E. Miyazaki, J. Caral., 1980, 65, 84.Dr Lambert, Dr Badyal and Dr Nix replied: The molecular CO state observed at low temperatures and low Nd coverages in the TPD studies on the model systems is associated with chemisorption on exposed copper surface; the binding characteristics of this species are not significantly perturbed by the presence of pre-adsorbed neodymium (which is itself oxidized by dissociative adsorption during the initial stages of exposure). At initial Nd coverages greater than a monolayer, however, (i.e. when there is no exposed copper present) CO is still rapidly and dissociatively adsorbed by the neodymium at 300 K and there is no evidence for a molecular precursor state. Further molecular adsorption on the oxidized neodymium was not observed under high vacuum conditions at either 77 or 300 K but will certainly be an important feature of the high-pressure catalytic chemistry.We have observed two TPD features on exposing H2 to Ti/Ru(0001).' The low- temperature feature due to atomic hydrogen associated with the bare Ru(0001) patches showed an initial increase [our fig. 4(6)] due to hydrogen spillover from the TiH, islands; these TiH, species give rise to the appearance of a sharp hydrogen desorption feature at high temperature which increases in intensity with titanium precoverage. 1 J. P. S. Badyal, A. J. Gellman and R. M. Lambert, J. Catal., 1988, 111, 383. Dr A. R. Gonzalez-Elipe (Instituto de Ciencias de Materiales de Sevilla, Seville, Spain) said: In relation to the formation of hydride species in your Ti/Ru and TiO,/Ru systems I would like to mention that in previous work using i.r., H'-n.m.r., e.p.r., TPR and X.P.S." on M/Ti02 catalysts (where M: Rh, Pt or Ni) we have postulated the formation of such species in 'real catalysts' through 'spillover' of hydrogen atoms from the metal to the reduced TiOz support according to the reaction: TiVi'+ Rhs-H -+ (Ti-H)3++ Rh,.In addition we have presented evidence that such species could produce (i) an additional suppression of the H2 and CO adsorption in the SMSI ~ t a t e , ~ - ~ (ii) an enhanced mobility of the reduced TiO, support leading to 'decoration' of the metallic particles6 and (iii) the reduction of CO to give CH,OH.' Owing to the rather different experimental conditions used in your work with 'model systems' I would like to ask you a few questions. First, I would like to know what are the conditions to generate hydride species in your TiO, islands deposited on Ru(0001).Its formation is reported in your paper but you do not give details on this point. My second point is do you observe an additional suppression of CO adsorption in your TiO,/Ru system when hydride species are present? Finally, I would like to know your opinion on the role of such hydride species in hydrogenation or hydrogenolysis reactions occurring on these M/Ti02 systems. 1 J. C. Conesa, P. Malet, G. Munuera, J. Sanz and J. Soria, J. Phys. Chem., 1984, 88, 2986. 2 J. Sam, J. M. Rojo, P. Malet, G. Munuera, M. T. Blasco, J. C. Conesa and J. Soria, J. Phys. Chem., 3 A. R. GonzLlez-Elipe, G. Munuera, J. P. Espinos, J. Soria, J.C. Conesa and J. Sanz, Proc. 9th Inr. Cotig. 4 J. Sanz and J. M. Rojo, J. Phys. Chem., 1985, 89, 4974. 5 A. Munoz, A. R. GonzLlez-Elipe, G. Munuera, J. P. Espinos and V. Rives Arnau, Specrrochim. Acra, 1985, 89, 5427. Catal., Canada 1988, ed. M. J. Phillips and M. Terman, vol. 3, p. 1392 Part A, 1987,43, 1599.170 General Discussion 6 G. Munuera, A. R. Gonzalez-Elipe, J. P. Espinos, J. C. Conesa, J . Soria and J. Sanz, J. Phys. Chem., 1987, 91, 6625. Drs Badyal and Lambert replied: We do not claim to have generated a hydride species on TiO, phase on Ru(0001); however, we were able to generate a Ti-H species following hydrogen exposure to Ti/ Ru(0001).' This species is extremely efficient in CO dissoci- ation and we suggest that such strongly bound hydrogen species may be responsible for the enhanced catalytic behaviour of Ru/Ti02 catalyst which have been reduced at high temperature.1 J. P. S. Badyal, A. J. Gellman and R. M. Lambert, J. Catal., 1988, 111, 383. Prof. A. K. Datye (University of New Mexico, Albuquerque, U.S.A.) had the following question and comment: The influence of TiO, on adsorbed CO in fig. 6 ( a ) is shown to be more pronounced than a simple site-blocking effect. However, reduced TiO, species are known to wet metal surfaces and spread on them. Hence the question is: How was the T i 0 surface coverage measured and was it measured before or after the CO adsorption. Does adsorption and dissociation of CO lead to any restructuring of these oxide overlayers? Your results imply that the TiO,H,. species may be responsible for the altered metal behaviour in the SMSI state.However, the increase in CO hydrogenation activity is not more than a factor of 3 after high-temperature reduction. The major difference in activity say between Pt/Ti02 and Pt/ S O z is seen even before high-temperature reduction on the fresh catalyst. Drs Badyal and Lambert replied: As explained in our paper, the T i 0 surface coverage was measured by Auger spectroscopy, calibration being provided by the known growth morphology of this phase on Ru(0001).' No restructuring of the oxide phase appears to be induced by CO desorption or dissociation. Regarding the differences between Pt/Ti02 and Pt/Si02, we too have observed similar differences in behaviour for equivalent metal loadings in the case of Ru/Si02 and Ru/Ti02 before high-temperature reduction; our results indicate clearly that this initial difference is due to differences in metal dispersion.2 1 J.P. S. Badyal, A. J. Gellman, R. W. Judd and R. M. Lambert, Catal. Left., 1988, 1, 41. 2 J. P. S. Badyal, R. M. Lambert, J. C. Frost, C. Riley and K. Harrison, J. Catal., submitted. Prof. A. K. Cheetham (University of Oxford) had a question for Prof. K. I. Zamaraev (Institute of Catalysis, Novosibirsk, U.S.S.R.): The last slide of your talk, and the abstract of your paper, shows a three-step catalytic cycle that achieves the oxidation of SOz to SO3 without the involvement of VIv species. What is the evidence that step 3, which involves the loss of bridging SO, and the incorporation of 02, proceeds via a single reaction of Vv species rather than a two-stage process, for example - S O , 0 2 v v ___, VlV ___, vv Prof.Zamaraev replied: Note that the first stage of the two-stage process proposed by Prof. Cheetham actually corresponds to the upward direction of reaction 4 of scheme 1 from our paper. As indicated in our paper, the rate of oxidation with O2 of the VIv complexes formed in reaction 4 is much less than the rate of the overall catalytic reaction under steady-state conditions. This means that the catalytic reaction does not proceed via the stepwise mechanism consisting of alternating steps of Vv reduction with SO:- anion to V" and subsequent oxidation of V" back to Vv with Oz. Moreover, the rateGeneral Discussion 171 CO poisoning expt 40 0 1 2 3 CO adsorbed/ 10' mol Fig.4. Ethene polymerization rate over chromocene/silica catalyst as a function of amount of CO adsorbed. (Initial C2H, pressure = 100 Torr; temperature = 40 "C; (Cr = 4.5 wt%, CO/Cr = 0.33% .) of the catalytic reaction under both steady-state and non-steady-state conditions is proportional to the concentration of Vv and does not correlate with the amount of V" in the catalysts. This suggests that only the Vv species are involved in the catalytic cycle. Under these circumstances it seems more likely that step 3 of our three-step catalytic cycle proceeds uia a transfer of two electrons from SO:- ligand to O2 ligand in the coordination sphere of the binuclear V v complex r SO:- s0:- (see p. 13 of our paper) rather than via the two stage process -so, 0, vv ___* V'V L vv proposed by Prof.Cheetham. Prof. J. H. Lunsford (Texas A&M Uniuersily, U.S.A.) began the discussion of Prof. Zecchina's paper: Attempts to identify active surface species by spectroscopic techniques are often frustrated by the fact that only a small percentage of a potentially active phase is involved in the catalytic cycle. We have found that a chromocene-on-silica catalyst, prepared by the sublimation technique, is active for ethene polymerization, but the sites responsible for this activity are extensively poisoned by the addition of carbon monoxide as shown in fig. 4. These results suggest that <0.33'/0 of the Cr is active for the polymerization reaction. The addition of 0.023 Torr of CO to a Cp,Cr/Si02 wafer initially yielded weak infrared bands at 2004, 1970 and 1831 cm-'.The integrated area of the band at 2004 cm-' was only 0.36% of the total area observed for the carbonyl bands with excess CO in the cell. Thus, it appears that the infrared bands reported by172 General Discussion Zecchina et al. in their fig. 3 do not reflect the active chromium species, but by observing the sample following the addition of very small amounts of CO it may be possible to probe the active site. Prof. Zecchina replied: Your results are very interesting because they confirm that under low ethene pressure the active sites are very scarce. In our model the small number of active sites simply derive from the fact that they are located at the narrow boundary between the free and the supersaturated (self-poisoned) regions.In this respect it is most interesting to consider that 0-CrCp(CO), complexes derive not only from the SSi-0-Cr-Cp 'active centres', but also from the 2%-0-Cr-Cp . . . CpCrCp inactive centres (reaction 2). On this basis, the bands of fig. 3 are more useful for elucidating the basic chemistry of the surface species than for probing the active sites. However, they are also relevant for establishing the structure of the active centres because they show that the Cr(Cp), poisoning the sites can be displaced by incoming ligands (actually CO). There is no reason for not extending this concept to ethene, especially when it is used under high pressure (as in the industrial process). Dr A. F. Masters (University of Sydney, New South Wales, Australia) said: It is probably unnecessary to invoke an qs=q3 equilibrium of the cyclopentadienyl ring of the metallocycle in the mechanism of scheme 8, as discussed in the text.The q5 e 37 cyclopentadienyl equilibrium relieves electronic saturation. Your metallocycle is formally electronically unsaturated (14 e) and is easily able to interact with ethene from an electron-counting point of view. My first question relates to the identity of the intermediate (6) of scheme 5 , and hence to the related intermediates, ( a ) of scheme 5 , and those of scheme 3. It seems that the only evidence for intermediate (6) is the decrease in the intensity of the 0-H infrared absorption at 3748 cm-', and the assumption that this reduction in intensity is caused by the reaction of a surface OH group with CrCp,, generating dicyclopentadiene. Is there any other evidence for ( b ) ? For example, have silanols been reacted with chromocene in homogeneous solution? With regard to the infrared spectra of your fig. 3, were these and your other spectra obtained via transmission or reflectance? If your assignments are correct, the group of six bands between 1920 and 1579 cm-' would presumably arise from the interaction of CO with supported chromocene [cf, eqn (2)]. Do you have any suggestions as to the identity(ies) of the product(s)? What CO pressure was used for these reactions? Finally, what percentage of chromium-containing species in your silica matrix would you estimate as representing the active catalyst? Prof. Zecchina replied to each of these comments: The reaction of Cr(Cp), with silanols in solution with formation cyclopentadiene is well known from the papers of Karol et al. (quoted in the references). The i.r. spectra have been performed in transmission by using thin silica wafers. With regard to the two triplets in the 1900-1579 cm-' range, formed under 40 Torr of CO at room temperature, we can only mention that they have carbonyl frequencies very unusual for Cr"-Cr" complexes. Negatively charged multicarbonyl species have to be invoked. Their detailed structure is, however, unknown: further research is needed. The active sites for ethene polymerization under low pressure are present only in the narrow boundary layer between the free and the supersaturated regions, so their total number is small. However, following our model, the number of active sites depends upon the ethene pressure. Under real industrial conditions (200 psi) ethene can displace the extra Cr( Cp), , poisoning the potentially active sites.

ISSN:0301-7249    

DOI:10.1039/DC9898700161

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

Faraday Discuss. Chem. SOC., 1989,87, 173-187 Acid-Base and Oxidation Catalysis on Heteropolysalts with Surface Acid Layers Katarzyna Bruckman, Jerzy Haber and Ewa M. Serwicka" Institute of Catalysis and Surface Chemistry, Polish Academy of Sciences, 30-239 Krakow, ul. Niezapominajek I, Poland Physicochemical and catalytic properties of heteropolyacids of the series H3+rlPV,,Mo,2-,,040 ( n = 0-3), both pure and supported on the potassium salt K3PMo,2040, have been investigated. Thin acid coats formed on such a support display modified properties and enhanced thermal stability. In particular, it is postulated that the change in the acidic properties of the supported acids is a consequence of their modified hydration ability resulting from the epitaxial relationship with the support.Results of catalytic experi- ments for the oxidation of acrolein, methanol and alkanes are presented and compared for both series of the catalysts. Possible mechanisms of all these processes are proposed on the basis of experimental data and quantum- chemical calculations. Relatively weak thermal stability of the heteropolyacids with the Keggin structure (fig. 1) represents a major drawback in their potential application in catalysis. Goodenough and co-worker~'-~ in a recent extensive study of acrolein oxidation over the KxH3-xPMo12040 (0 < x < 3) system found that the best catalytic performance could be obtained for x = 2.5. They established that at this composition the catalyst consisted of the H3PMo12040 acid phase stabilized in the form of an epitaxial, isostructural layer at the surface of the K3PMo12040 particles.The latter, in contrast to the acid phase, is a water-insoluble compound, crystallizing in a cubic lattice, thermally stable up to 1000 K. An interesting question of far-reaching implications is whether it is a general phenomenon that heteropolyacids with the Keggin structure may be supported on the appropriate heteropolysalts, thus becoming stabilized in the form which otherwise decomposes under conditions of catalytic reaction. Heteropolyacids derived from H3PMo12040 by substitution of one to three molybdenum atoms by vanadium appear, in view of their similarity with the parent acid4 particularly promising objects for such a study. Furthermore, the literature data indicate the advantageous properties of such compounds in the selective oxidation processe~.~-~ For this reason the research into their thermal stability and stabilization effects is of great potential significance.Since all water-insoluble salts of these acids with the alkali cations (K+, Cs+, Rb', NH;,) are isomorphous with K3PMo,2040, this compound was chosen to serve as a support for each member of the series of acids studied. Experimental Materials The H3+nPV,Mo12-n040~~H20 ( n = 0-3) heteropolyacids were prepared according to the method of Tsigdinos and Hallada.' The exact water content of the crystalline acids hydrates was found to be H3PMoI2O4,,.28 H20, H4PVMol 1040-32 H20, H5PV2M010040.3 1 H20 and H6PV3M0904,,-30 H20, as determined by TGA. These samples are hereafter referred to as H3, H4, H5 and H6.The K3PM~12040 support, abbreviated to K3, was prepared by the method described by Tsigdinos"' from stoichiometric quantities of 173174 Heteropolysalt-supported Heteropolyacids Fig. 1. Structure of the Keggin unit. H3PMo12040 and K2C03. Its B.E.T. surface area was 160 m2 g-'. On the assumption that one Keggin anion occupies 144 A', catalysts corresponding formally to one monolayer coverage with the acid component, denoted hereafter H3/ K3, H4/K3, H5/ K3, H6/K3, were prepared by impregnation of the support with the desired quantity of aqueous solution of the respective acid, as described previously." All catalysts, pure and supported, were subjected to thermal treatment at 623 K for 3 h. The B.E.T. surface areas were as follows: for the pure, calcined acids 1-3 m2 g-', for the supported samples 5-8 m2 g-', and for the support 70 m2 g-'.Techniques X-Ray Diffraction Powder X-ray diffraction data were obtained with a DRON 2 diffractometer, using Cu K, radiation. Th erma 1 Analysis Thermogravimetric analysis was carried out in static air in a Setaram Micro ATD M5 thermal analyser. Samples of ca. 15 mg were heated at 5 K min-' up to 873 K. Electron Microscopy Microstructural characterisation of the surface topography of the catalyst particles was carried out with a JEOL 100 CX instrument equipped with an ASID 4D high-resolution scanning accessory operating at 100 keV. Raman Spectroscopy Raman spectra were acquired on a DFS-24 spectrometer equipped with an ILA-120 ( A = 647 nm, linear polarisation) laser source.Electron Spin Resonance E.s.r. spectra were recorded at room temperature and at 77 K with an X-band SE/X (Technical University Wroclaw) spectrometer. DPPH and n.m.r.-marker were used for the determination of g factors.K . Bruckman, J. Haber and E, M. Serwicka Acidity Measurements 175 In the temperature-programmed desorption experiments a 100 mg sample was first outgassed at 573 K for 0.5 h at 5 x Torr.? Then pyridine (20 cm3) was introduced at 293 K and the adsorption carried out for 0.5 h. The sample was then outgassed at 423 K for 1 h and the t.p.d. experiment was carried out in the temperature range 423-723 K at a heating rate of 10 K min-'. Desorbing products were analysed with a mass spectrometer. In the i.r. experiments the sample pretreatment and pyridine adsorp- tion were carried out in a similar way and the intensity of the pyridinium ion band was taken as a measure of the Bronsted-acid site concentration.The spectra were recorded with an IR-20 spectrometer. Catalytic Testing Reactions were investigated under continuous-flow conditions in a standard glass tubular microreactor. Products were analysed chromatographically. Results Catalyst Characterization Scanning Electron Microscopy SEM analysis was undertaken to obtain a- better insight into the catalysts morphology. Plate 1 shows the SEM images of the H4, K3 and H4/K3 calcined catalysts. The picture for other pure and supported acids is identical with those presented here as an example. There is a clear difference between the crystal habit of the K3 support and the pure acid phase. The potassium salt is composed of small, well formed round or hexagonal crystallites, of ca.1 p m diameter, whereas the acid sample consists of agglomerates of irregular, cracked crystallites varying in size and habit. The surface topography of the supported acids closely resembles that of the K3 support. Also, the defect-free, clean single crystallites dominate. No separate irregular crystallites, characteristic of the unsupported acid phase, can be seen, although the amount of the acid deposit constitutes almost 30% of the catalyst weight. Laser Raman Spectroscopy In view of the relative insensitivity of the i.r. spectra of the primary Keggin unit to the type of the secondary structure and nature of the counter cation, Raman spectroscopy proved to be particularly useful for elucidation of these phen~rnena.l*-'~ Our Raman investigation of the pure and K,-supported heteropolyacids demonstrated that the acid deposit represents structurally a new quality, combining the properties of the acid and the upp port.'^ Fig.2 shows as an example the Raman spectrum of the K3 support, together with those of pure and supported calcined H5. The characteristic feature of the K3 spectrum is a split band in the 230-250 cm-' region. The bands observed in this range are caused by deformation vibrations both of the terminal M=O groups and of the entire framework, and are sensitive to the cationic environment of the Keggin unit . The other characteristic fragment of the spectrum falls in the range 900-1050 cm-' where the valence vibrations of the individual M=O groups, the PO, and the breathing modes of all 12 M=O groups are expected. The spectrum of the calcined acid differs from that of the support in both regions.In the 230-250 cm-' range the band at lower 12-14 t 1 Torr = 101 325/760 Pa.176 )r Y .- 2 Y C .- Heteropolysalt-supported Heteropolyacids 200 250 950 1000 wavenumber/ cm- ' Fig. 2. Laser Raman spectra of calcined (623 K) samples: (-) H,, (- - -) K3, (- - -) H5/K3. wavenumbers becomes more intense and the other appears only as a shoulder. In the 900-1050 cm-' region the absorption maximum is shifted towards 1010 cm-'. A similar shift was observed by and assigned to a dehydrated form of acid. The Raman spectrum of the supported acid layer in the low-frequency range resembles that of the acid, whereas the 900-1050 cm-' bands are similar to those observed for the K3 support.Diflerential Thermal Analysis This technique is particularly useful for determination of the thermal stability of the acids, since, following the final loss of constitutional water, an exothermic peak appears around 700 K, associated with the irreversible destruction of the Keggin Accord- ing to d.t.a. results the decomposition temperatures of pure acids are as follows: H3, 706 K; H4, 715 K; H5, 685 K; and H6, 669 K. For each member of the supported-acid series a distinct shift of the exothermic peak position towards higher temperatures is visible, i.e. H3/K3, 727 K; H4/K3, 733 K; H5/K3, 725 K; and H6/K3, 700 K. This phenomenon clearly indicates thermal stabilization of the acid layer deposited on the potassium salt.X - Ray Digraction The X-ray data confirm the stabilizing effect of the support on the heteropolyacid structure." Fig. 3 shows as an example the diffraction patterns of ( a ) pure support K3, ( b ) overheated H5, (c) overheated H5/K3 and ( d ) a mechanical mixture of overheated H5 and K3 in ratio as in H5/K3. The additional heat treatment of the previously calcined catalysts was performed for 1 h at 673 K, i.e. close to the decomposition conditions determined for the H5 acid from the d.t.a. measurements. The pure support gives a typical powder spectrum of the K3PMo12040 cubic lattice. Overheated H5 shows peaks due to MOO, and some other, as yet unidentified, product of the decomposition of theK .Bruckrnan, J. Haber and E. M. Serwicka 177 I T I 0 I ( c ) LO 30 20 10 2810 Fig. 3. Powder X-ray diffraction patterns of (a) K3, (b) H5 overheated at 673 K, (c) H5/K3 overheated at 673 K, ( d ) mechanical mixture of H5 and K, overheated at 673 K in ratio as for H5/K3. 0, Diffraction pattern of MOO,; x, diffraction pattern of unidentified phase. Keggin unit (anhydrous, undecomposed H5 is practically amorphous to X-ray). The overheated supported H, shows only the pattern characteristic for the cubic lattice of the support, in agreement with the idea of formation of the epitaxial, isostructural layer of the acid on top of the K3 particles. Finally, a mixture of the overheated H5 and K3 in a quantitative ratio corresponding to that of the supported HS/K3 sample shows both the peaks characteristic for the K3 phase and those found in decomposed H5.Infrared spectra corresponding to the X-ray diagrams" confirm that the primary Keggin unit of the acid phase supported on K3 becomes stabilized against thermal decomposition. Electron Spin Resonance It has been demonstrated that e.s.r. can be successfully employed to follow various stages of the heteropolycompound dehydration and/or d e s t r ~ c t i o n . ~ ~ ' ' - ' ~ We have shown that the e.s.r. spectra observed for the K,-supported H3 sample calcined for 5 h at 673 K are characteristic of the preserved, undecomposed Keggin anions, whereas in the pure H3 phase treated in the same way the e.s.r. signal of Mo5+ in the MOO, matrix appear^.'^"'^ In fig.4 similar evidence is presented for the H,-supported catalyst. Initial spectra of both samples are similar, with gl = 1.976, A, = 7.43 mT, gll = 1.927, All = 20.28 mT [fig. 4( a)] and are typical of a V4+ ion in a hydrated, undecomposed acid phase.'* Prolonged calcination of the pure H, acid results in a spectrum dominated by a signal with clearly178 Heteropolysalt-supported Heteropolyacids Fig. 4. E.s.r. spectra of ( a ) H, and/or H,/K3 samples, ( b ) H, overheated at 673 K, ( c ) H,/K3 overheated at 673 K. Spectra recorded at 77 K. resolved gll= 1.913 and All = 18.79 mT [fig. 4(6)], typical of the vanadium-containing products of irreversible destruction of the Keggin units.I8 E.s.r. of the supported sample treated this way shows mainly a signal with well distinguished parallel compounds gll= 1.931, All = 18.79 mT [fig.4(c)], characteristic of V4+ ions in the undecomposed, partially dehydrated acid phase.'' Acidity Measurements Experiments with pyridine adsorption demonstrated that the deposited layers, when compared to the bulk acids, show modified acid-base properties. Fig. 5 shows the general trends for acidity measured by the amount of pyridine retained in the catalyst after outgassing at 423 K and detected by i.r. in the form of the pyridinium ion.*' ForK . Bruckman, J. Haber and E. M. Serwicka H3/K3 HLIK3 Hs/K3 Hb/K3 K3 I 1 I I I 179 Fig. 5. Acidity of calcined samples as determined from intensity of the i.r. 1540 cm-' band of the adsorbed pyridine. both types of catalysts investigated (bulk and the supported) a fall in Brgnsted acidity is observed as the vanadium substitution increases.Simultaneously, the deposition on the K3 support results in a general increase of the acidity. No significant number of Lewis-acid centres could be detected after pyridine adsorption. Parallel experiments with the t.p.d. of pyridine trapped in the catalyst after outgassing at 423 K, allow one to infer the nature of the acidic centres observed. From bulk, calcined acids pyridine is desorbed with a number of overlapping peaks, indicating the presence of centres of different acid strength [fig. 6 ( a ) ] . Also in this experiment the trend of decreasing acidity with increasing number of vanadium atoms is visible, parallelled by a shift towards relatively weaker acid sites.The insert in fig. 6 ( a ) shows the desorption profile from the H6 sample calcined at a temperature 150 K lower than the standard treatment. This clearly shows that the occurrence of the strongest acid sites is associated with the amount of water retained by the bulk of the acid phase. The maxima of pyridine desorption peaks from the supported samples occur at the same temperature for all samples [fig. 6( b ) ] , indicating similar strength of the acid sites. Here also the total amount of desorbed pyridine decreases as the number of vanadium atoms in the supported acid increases. Catalytic Testing Met h a n ol Oxida t ion Our recent study2' of the oxidation of methanol over pure and supported catalysts showed that in the temperature range 523-573 K introduction of vanadium into the Keggin anion influences the selectivity pattern of both series in a similar way, whereas their activity changes in the opposite manner.Formaldehyde, the product of oxidative dehydrogenation, and dimethyl ether, the product of dehydration, appear as major products for both series. Fig. 7 and 8 show the catalytic performance of both series at 533 K. Substitution with vanadium shifts the product distribution towards formaldehydeHeteropolysalt-supported Heteropolyacids 423 523 623 723 T/ K Fig. 6. T.p.d. of pyridine from the calcined catalysts. (a) Pure acids: (-) H3, (- - -) H,, ( . . . . - ) H,, (--.-) H6. ( b ) Supported acids: (-) H3/K3, (---) H4/K3, ( - - . - - ) H,/K3, (- ' -) H6/ K3. K H3/K3 Ht,/K3 H5/K H6/K < 100 - - 80 - - I s - 2 60 - 1 A > c .- .- 5 40 - 20 - Fig.7. Catalytic activity in methanol oxidation at 533 K on 0, pure acids, 0, supported acids. Conditions: He : O2 : CH30H = 76.14: 16.2 : 7.65. PCH30H = 58.2 Torr.K . Bruckman, J. Haber and E. M. Serwicka 181 1 .o T 9. 0, 5 0.5 d rn I 1 I - 1 I 1 I . I H3 H& H5 H6 Fig. 8. The ratio of CHzO and CH30CH3 selectivities in CH30H oxidation at 533 K. 0, Pure acids, A supported acids. i Table 1. Activity and selectivity to maleic anhydride in pentane oxidation" activity/ selectivity to catalyst T/ K pmo1(C2H5) s-' gi:id phase maleic anhydride ( O/O ) 96.2 147.5 217.7 121.5 178.5 53 1 .O 607.5 46 68 63 38 43 47 66 C5H12 : O2 : He = 1.8 : 25 : 73.2. for both the pure and supported acids. Simultaneously the activity of the pure acids decreases, whereas that of the supported samples increases.Acrolein Oxidat ion The results of the oxidation of acrolein to acrylic acid conducted at 623 K have shown" that the maximum yield of acrylic acid on pure heteropolyacids is observed for the thermally most stable H4 sample, although all vanadium-containing catalysts are better than the unsubstituted sample. The supported catalysts generally perform better than their unsupported counterparts, and the optimum catalytic properties are obtained for the samples with the highest vanadium content. Alkane Oxidation Vanadium-substituted heteropolyacids were shown recently to be surprisingly selective in the catalytic oxidation of n-pentane to maleic anhydride.* K,-supported acids show similar properties.The catalytic performance of pure and supported H5 is summarized in table 1. The data show that the activity expressed per gram of the acid component182 Heteropolysalt-supported Heteropolyacids is distinctly higher for the supported sample. Also, as the temperature increases, the selectivity of the pure acid shows signs of deterioration, whereas on the supported catalyst a steady increase is observed. Maleic anhydride is the only selective product of pentane oxidation on the heteropoly- catalyst, while on the industrial pyrophosphate catalyst a mixture of the maleic and phthalic anhydrides is formed.8 Oxidation of n-butane is also known to give maleic anhydride on heteropolyacid catalysk6 In order to obtain some insight into the nature of the activation of alkane molecules on the surface of the Keggin unit, the oxidation of n-butane, n-pentane and n-hexane were investigated.In each case maleic anhydride was the only selective product, although the activity in the n-hexane oxidation was distinctly smaller. Discussion All techniques employed to characterize pure and supported heteropolyacids clearly prove that the intimate interaction between the members of the H3+,, PV,Mo,,-,,O,, series and the K3PMo,2040 support produces surface acid coats of a new quality. The phenomenon has a general character and is independent of the composition of the supported acid layer. The most significant effect is the observed thermal stabilization of the deposited material. Analysis of the Raman spectra gives an indication as to the structure of the surface layer in the supported catalyst.The absence of the band at ca. 1000 cm-’ typical of the ‘anhydrous’ acid phase indicates that a certain number of water of crystallization molecules are retained in the structure. On the other hand, the similarity of the bands in the 230-250 cm-’. region, sensitive to the cationic environment, to those found for the pure acids indicates that the surface layer consists of an acid hydrate rather than of a solid solution involving potassium cations. Therefore, in agreement with earlier findings,’ we propose that the stabilization effect is a consequence of the formation of an epitaxial acid layer derived from H3+, PV, Mo,,-,,O~~-X H20, isostructural with the cubic lattice of the K3PMo1,04, support particles. Electron-microscopic data’ showed that in the partially dehydrated H,PMo,,O,, a phase exists that is isostructural with the potassium salt of this acid.Brown et al.” found that the 12-tungstophosphoric acid hexahydrate is isomorphous with the insoluble cubic salts of various 12- heteropolyanions. They established that the cationic positions in such a hydrate are occupied by ( HSO2>+ diaquahydrogen ions, and the structure can readily accommodate the extra protons expected from the stoichiometry of related heteropolyacids. In view of this, it seems reasonable to visualise the surface layer in the K,-supported catalysts as hexahydrates of the respective acids. Although, on their own, such hexahydrates are unstable and difficult to obtain, the epitaxial relationship with the support provides convenient conditions for their stabilization.Measurements of the acidity revealed that the supported acid layer shows acid-base properties different from those of the pure acids. Before discussing the details, it is necessary to recall that numerous investigations of the state of protons in the heteropoly- compounds concluded that basically two types of protons can be distinguished; non- localized hydrated protons and non-hydrated less mobile protons.’ The smaller the degree of hydration, the smaller is the proton mobility, and, consequently, a lower acid strength of the constitutional protons may be expected. This is clearly visible when comparing the results of t.p.d. of pyridine from the H, acid samples calcined at various temperatures (fig.6). The sample that was pretreated at lower temperature, and therefore retained more water, shows the presence of a significant amount of strong-acid centres, whereas the sample calcined in more drastic conditions lacks such acidic sites almost completely. We may thus conclude that a correlation exists between the acidity of protons and their degree of hydration.K . Bruckman, J. Haber and E. M. Serwicka 183 The second important factor that has to be taken into account is the increasing charge of the Keggin anion as the degree of vanadium substitution increases and, as a result, the increasing total number of constitutional protons available. However, a quantitative analysis of the i.r. and t.p.d. data shows that acidity, measured by the amount of pyridine retained in the catalysts after desorption of weakly adsorbed pyridine at 423 K, shows a falling trend with an increasing number of vanadium atoms.Thus, the number of acid sites strong enough to be seen by pyridine under these conditions represents only a fraction of the constitutional protons present in each sample and decreases on gradual substitution with vanadium. According to t.p.d., strongly adsorbed pyridine detects 12% ( H3), 8% ( H4), 6% (H,) and 4% ( H6) of the constitutional protons in the bulk, calcined acids, whereas the respective numbers for the supported samples are 45% (H3/K3), 32% (H4/K3), 20% (H,/K,) and 13% (H6/K3). This signifies that the majority of protons present in the calcined samples constitute weak-acid centres, firmly bound to the Keggin anions, their amount increasing with number of vanadium atoms.For the series of pure, calcined acids a shift towards weaker acid centres as the substitution with vanadium increases is also visible within the range of acidic sites detected by the t.p.d. of pyridine [fig. 6(a)]. Such behaviour shows that the more vanadium there is, the lower is the degree of rehydration responsible for the appearance of the strong acidity in the calcined acids. In the supported samples [fig. 6(6)] the falling trend in acidity within the series is also visible, but the n u m i v of sites detected by pyridine represents a larger fraction of the total number of constitutional protons expected from the sample's stoichiometry. The maxima of the desorption peaks occur at the same temperature, indicating the same strength of the acid centres responsible and, therefore the similar degree of hydration of the protons involved.The lack in the supported samples of stronger acid sites, observed for instance in the H, and H4 members of the pure acid series, indicates that the supported layers cannot accommodate more than a certain, well defined number of water molecules. It seems reasonable to assume that this maximum number corresponds to the number characteristic for the cubic acid hexahydrates, i.e. that the most hydrated protons in these samples are present as diaquahydrogen ions. There is no such limitation for the dehydration of the deposited acid coat, therefore here also most of the constitu- tional protons represent weak-acid centres, in amounts increasing with the number of vanadium atoms.Catalysis and Mechanistic Studies Detailed characterization of the physical and chemical properties of the catalysts investigated is essential for an understanding of the observed trends in their catalytic behavior. For the mechanistic analysis, however, it is important to realise that all chemical reactions discussed above are governed by HOMO-LUMO interactions. There- fore, suitable quantum-chemical calculations should help in elucidating the respective reaction mechanisms. According to Taketa et al.,23 the HOMO of the (PMo,,0,J3- ion is composed mostly of the 2p lone-pair orbitals of the bridging oxygens, Ob. The LUMO is a mixture of the Mo 4d and Ob 2p orbitals and is antibonding with respect to the Mo-Ob bond.Methanol Oxidation We have used the reaction framework proposed by MoffatZ4 and extended by Farneth et for the quantum-chemical interpretation of the methanol t.p.d. data:?'184 Heteropolysalt-supported Heteropolyacids CH30HT.*'OKU * CHl...OKu+ HzO CH3+..*OKU+~KU -, CH,O+ U + . - . ~ K U . (4) Our MNDO calculations gave the following HOMO and LUMO for the organic species involved. (a) CH30H. *HOMO = -0.83 p,(O) + 0.28 p x ( c ) - 0.34 S ( H&) 'PLUMO= -0.68 p,(C)+O.52 s(HkH)-0.26 s(HgH)-0.26 ~ ( H $ ~ ) - 0 . 2 9 s(HOH). S ( H:H) The HOMO represents a bonding orbital localized mainly on the oxygen atom, while the LUMO is the first orbital of the antibonding s stem, associated mainly with the CH3 charge of 0.19. group. The calculated C-0 bond length is 1.39 K and the carbon atom carries positive (6) CH30Hl.The HOMO'S highest electron density is associated with the CH3 group, while the LUMO has comparable contributions from all structural elements but the CH3 group. The C-0 bond is elongated to 1.46 A and the carbon atom increases its positive charge to 0.23, as intuitively predicted by Highfield and M ~ f f a t . ~ ~ (c) CH30-. *HOMO = -0.73 p,( 0 ) - 0.43 pz( 0 ) -4- 0.21 S ( H&) - 0.43 S ( HgH) + 0.21 S ( 'PLUMO= -0.24p,(0)-0.68 p,(C)-0.39 p,(C)+O.23 S(H;~)-O.~~S(H:.,) + 0.23 S ( HcH). The HOMO carries the highest electron density at the oxygen atom, while the LUMO is localized predominantly on the CH3 group. After taking into account all conceivable HOMO-LUMO interactions the detailed mechanism of the methanol interaction with the Keggin anion, presented in fig.9, has been proposed. Formaldehyde appears as a result of nucleophilic attack of a bridging oxygen lone pair (Keggin's HOMO) on the C-H bond of the methoxy group (configur- ations 111 and IV). Dimethyl ether is formed in a competitive reaction between the methoxy group and a methanol molecule, according to eqn (3). Substitution with vanadium results in changes of the charge distribution within the Keggin anion. In particular, from the "0 n.m.r. data reported by Maksimovskaya et a1.,27 it is known that the electron density on the bridging oxygens of the Mo-0-Mo type systematically increases. This is equivalent to an increased capability of performing a nucleophilic attack on the C-H bond of the methoxy group shown in configuration 111 and therefore should enhance the reaction path leading to formaldehyde.This picture is fully consistent with the catalytic data, showing that substitution with vanadium shifts the spectrum of the methanol oxidation products towards formaldehyde (fig. 8).K . Bruckman, J. Haber and E. M. Serwicka H \ /H H C II I + H 'c=o t / H H I 0 - 'Mo Mo /'*\Mo/ It 0 It 0 It 0 185 + 2 e + 2 e s XI Fig. 9. Mechanism of interaction of methanol with the Keggin unit. In order to explain the trends in activity, it is necessary to bear in mind that the activity towards methanol conversion will depend on the ability to absorb methanol by the secondary structure of the heteropolyacid catalysts.' From the acidity measurements [fig. 6 ( a ) and (6)] it follows that the acidity of the calcined acids decreases as the number of vanadium atoms increases, as will their ability to absorb polar molecules such as methanol.Therefore, the diminishing trend in catalytic activity observed for the bulk acids reflects primarily their absorptive capacities. The situation is different for the supported acid layer. Here the secondary structure is similar throughout the series, and more rigid owing to the epitaxial relationship with the support. Thus, no significant differences in the absorptivity should be expected. Consequently, the activity will depend on the efficiency of the generation of methoxy groups, i.e. on the methanol transformation in steps I and I1 of fig. 9. In view of the high proton affinity of the methanol molecule, even weak Brprnsted-acid centres can ensure methanol protonation. Therefore, the growing activity of the supported series reflects the increasing number of weak-acid sites resulting from substitution of Mo by V atoms.186 Heteropolysalt-supported Heteropolyacids Acrolein Oxidation It has been demonstrated that in the catalytic oxidation of acrolein, substitution with vanadium has a beneficial effect, but the low thermal stability of pure acids is responsible for the poorer performance of the H5 and H, samples." The HOMO-LUMO interactions of the acrolein molecule with the 12-molybdophos- phate anion have been discussed in detail by Serwicka et al.3 Here it has been demon- strated that the selective insertion of oxygen into the unsaturated aldehyde molecule is initiated by nucleophilic attack of a bridging oxygen lone pair on the carbonyl carbon for which, according to the quantum-chemical calculations, the LUMO of acrolein has the largest coefficient.The increasing selectivity to acrolein observed on substitution with vanadium is associated, according to the argument presented above, with an increased ability of the bridging oxygens to perform nucleophilic attack. Alkane Oxidation It is a striking feature of the catalysts investigated that maleic anhydride is the only oxygenated product of all the alkane molecules studied, i e . butane, pentane and hexane. No traces of such possible intermediate products as butene or butadiene could be detected, making the mechanism of consecutive dehydrogenation improbable.Con- versely, this may be taken as a hint that a unique mechanism is operating, consisting of a concerted abstraction of two hydrogen atoms from carbons 1 and 4, and simultaneous linking of these carbon atoms by oxygen to form a cyclic tetrahydrofuran-like structure. There are several possible sites at the Keggin unit, composed of three neighbouring bridging oxygen atoms, where such an operation could be visualised. All these oxygen atoms contribute two lone pairs of electrons to the HOMO of the Keggin unit. The HOMO and the LUMO of butane and higher alkanes can be deduced from the data available for propane28 in view of the same type of carbon-orbital hybridisation as well as the similarity of the C-C and C-H bonds throughout the series of higher alkanes.Following this reasoning, the HOMO of the butane molecule would be the last of the bonding set, essentially associated with the (T C-C bonds, while the LUMO would be the first antibonding orbital of the 7r type, involving mainly the CH3 and CH2 groups. The reaction can be visualised as an attack of the lone pairs of electrons of two adjacent bridging oxygens of the Keggin unit on the LUMO of the hydrocarbon molecule at carbon atoms 1 and 4. As the LUMO is localized mainly on the C-H bonds and is antibonding, such an attack would result in abstraction of H atoms to form surface OH groups. Simultaneously, the electrons of the C-H bonds would shift to the LUMO of the Keggin unit, which involves Mo 4d and Ob 2p orbitals. This would render carbon atoms 1 and 4 positive and susceptible to the attack of the lone pairs of the third adjacent bridging oxygen, closing the five-membered ring into a tetrahydrofuran-like structure.In the case of pentane or hexane cleavage of the methyl or ethyl group is required. Conversion of pentane is greater than that of hexane because the cleavage at the a position is easier than at the p position. The postulated tetrahydrofuran-like structure would easily undergo further dehydrogenation and oxidation to maleic anhydride. Analysis of the relevant interatomic distances in alkane molecules and in the Keggin unit shows that the geometrical fit easily allows the appropriate orbital overlap in consecutive stages of the concerted reaction. References 1 J. B. Black, N. J. Clayden, P. L. Gai, J .D. Scott, E. M. Serwicka and J. B. Goodenough, J. Catal., 1987, 106, 1. 2 J. B. Black, J. D. Scott, E. M. Serwicka and J. B. Goodenough, J. Catal., 1987, 106, 16. 3 E. M. Serwicka, J. B. Black and J. B. Goodenough, J. Caraf., 1987, 106, 23.K . Bruckman, J. Haber and E. M. Serwicka 187 4 H-G. Jerschkewitz, E. Alsdorf, H. Fichtner, W. Hanke, K. Jancke and G. Ohlmann, Z. Anorg. Allg. 5 M. Akimoto, H. Ikeda, A. Okabe and E. Echigoya, J. Catal., 1984, 89, 196. 6 M. Ai, J. Catal., 1984, 85, 324. 7 M. Misono, Catal. Rev. Sci. Eng., 1987, 29, 269, and references therein. 8 G. Centi, J. Lopez Nieto, C. Iapalucci, K. Bruckman and E. M. Serwicka, Appl. Catal., 1989, 46, 197. 9 G. A. Tsigdinos and C. J. Hallada, Inorg. Chem., 1968, 7, 137. Chem., 1985, 526, 73. 10 G. A. Tsigdinos, Znd. Eng. Chem. Res. Dev., 1974, 13, 267. 1 1 K. Bruckman, J. Haber, E. Lalik and E. M. Serwicka, Catal. Lett., 1988, 1, 35. 12 C . Rocchiccioli-Deltcheff, R. Thouvenot and R. Franck, Spectrochim. Acta, Part A, 1976, 32, 587. 13 L. P. Kazanskii, Zzv. Akad. Nauk SSSR, Ser. Khim., 1975, 3, 502. 14 E. N. Yurchenko, J. Mol. Struct., 1980, 60, 325. 15 Chau Dieu Ai, P. Reich, E. Schreier, H-G. Jerschkewitz and G. Ohlmann, Z. Anorg. Allg. Chem., 1985, 16 K. Bruckman, J. Haber, E. M. Serwicka, E. N. Yurchenko and T. P. Lazarenko, Catal. Lett., to be 17 E. M. Serwicka, 2. Phys. Chem. N. F., 1987, 152, 105. 18 R. Fricke, H-G. Jerschkewitz and G. Ohlmann, J. Chem. SOC., Farada Trans. 1, 1986, 82, 3479, and 19 E. M. Serwicka, Z. Phys. Chem. N. F., in press. 20 E. M. Serwicka, K. Bruckman, J. Haber, E. A. Paukshtis and E. N. Yurchenko, Catal. Lett., to be published. 21 K. Bruckman, J. Haber, E. M. Serwicka and J-M. Tatibouet, J. Catal., to be published. 22 G. M. Brown, M-R. Noe-Spirlet, W. R. Busing and H. A. Levy, Acta Crystallogr., Sect. B, 1977,33, 1038. 23 H . Taketa, S. Katsuki, K. Eguchi, T. Seiyama and N. Yamazoe, J. Phys. Chem., 1986, 90, 2959. 24 J. G. Highfield and J. B. Moffat, J. Catal., 1985, 95, 108. 25 W. E. Farneth, R. H. Staley, P. J. Domaille and R. D. Farlee, J. Am. Chem. SOC., 1987, 109, 4018. 26 E. M. Serwicka, E. Broclawik, K. Bruckman and J. Haber, Catal. Lett., in press. 27 R. I. Maksimovskaya, M. A. Fedotov, V. A. Mastikhin, L. I. Kuznetsova and K. I. Matveev, Dokl. 28 W. L. Jorgensen and L. Salem, The Organic Chemist’s Book of Orbitals (Academic Press, New York, 526, 86. published. references therein. Akad. Nauk SSSR, 1978, 240, 117. 1973), p. 181. Paper 9/00124G; Received 4th January, 1989

ISSN:0301-7249    

DOI:10.1039/DC9898700173

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

Furaduy Discuss. Chem. SOC., 1989, 87, 189-198 Preparation, Characterization and Catalytic Activity of Monodisperse Colloidal Metal Borides Janos B.Nagy,* Isabelle Bodart-Ravet and Eric G. Derouane Laboratoire de Catalyse, Facultks Universitaires Notre-Dame de la Paix, 61, Rue de Bruxelles, B-5000 Namur Belgium A new method, proposed for the preparation of colloidal ( d = 20-70 A) monodisperse particles of Ni2B, Co,B and Ni-Co-B, has been examined from the mechanistic point of view. The metal boride catalysts are prepared by reduction with NaBH, of Nil' and Co" ions solubilized in the inner water cores of reversed micelles composed of cetyltrimethylammonium bromide- hexan-l-ol-water. The curve of particle size as a function of either Ni" or Co" concentration shows a clear-cut minimum.This behaviour can be explained, provided a critical number of metal ions is assumed for the initial nuclei formation. The catalytic properties of the metal boride particles have been tested for the room temperature liquid phase hydrogenation of croton- aldehyde. On Ni,B particles the C=C double bond is preferentially hydro- genated, while on Co2B particles a non-negligible hydrogenation of the C=O group also occurs. Small metallic clusters can be stabilized by polymeric and surfactants or by special ligands including The present paper essentially deals with the prepar- ation, characterization and catalytic activity of very small ( d d 100 A) colloidal nickel boride, cobalt boride and mixed nickel-cobalt boride particles. Quite recently, some review articles were published on the preparation of monodisperse colloidal metal Small particles of controlled size and narrow size distribution have been synthesized by reducing, with hydrogen, hydrazine or sodium borohydride, metal salts dissolved in the inner water cores of microemulsions.8-'9 The catalytic activity of nickel boride is generally superior to that of Raney nicke1.20-22 This higher activity is attributed to a higher electron density on the nickel atom owing to electron transfer from boron to nickel.22 In addition, nickel boride is more resistant to sulphur poisoning, because the inhibitors are less strongly adsorbed than they are on Raney Nickel boride is essentially active in the hydrogenation of C=C double bonds and leaves the C=O and C z N groups q~asi-intact.~,-*~ The catalytic activity of cobalt boride is much lower than that of nickel b ~ r i d e .~ ~ - ~ ' This is probably due to its lower d-electron density, which results in a stronger adsorption of the reagents." Its selectivity, however, is higher for the C=O and C r N In order to understand the mechanism of formation of the monodisperse colloidal metal particles, it is indispensable to study the physico-chemical properties of the reaction medium. In this paper, a detailed analysis is given on the preparation of monodisperse Ni2B, Co,B and mixed Ni-Co-B particles in the CTAB (cetyltrimethylammonium bromide)-hexan- l-ol-water microemulsion. These particles can be prepared by reducing the Ni" and/or Co'l ions by NaBH, under a nitrogen or argon atmosphere.It is emphasized that the size of the water core, the lability of the interface and the concentra- tion of metallic ions are the determining parameters in preparing colloidal metal boride particles of controlled size. The catalytic activity of these metal boride particles is illustrated for the hydrogenation of crotonaldehyde. 189190 Monodisperse Colloidal Metal Bodies Y hexanol B mol of CoCl2 Y hexanol Y hexanol A mol N i C l 2 B mol CoCl2 Z H20 Fig. 1. Preparation of the mixed Ni-Co-B particles. 0 c x c 1 . x = B / ( A + B ) . Experimental Materials The commercial products, hexan-1-01 (Aldrich, 98%), CTAB (Serva, 99% ), NiCI2-6Hz0 (Merck, p.a.), CoC12.6H20 (Merck, p.a.), NaBH, (Aldrich, 99"/0), crotonaldehyde (Merck, 99.8%), and ethanol (Merck, p.a.), were used without further purification.Preparation of the Particles from the Microemulsion CTAB-Hexan-1-ol-Water The metal boride particles were synthesized from several reversed microemulsions of different compositions and, for each composition, the metal-ion concentration was varied. The bimetallic particles were obtained from a constant total metal ion concentration ( 5 x lo-* mol kg-I), varying the [Co"]/([Ni"] + [Co"]) ratio from 0 to 1 in steps of 0.1. The metal boride particles were prepared by adding dropwise an excess of aqueous NaBH, solution at 0 "C with vigorous stirring. The synthesis was carried out in a glove box under an argon atmosphere to prevent the oxidation of the particles. Fig. 1 shows the preparation scheme of the mixed Ni-Co-B. The expected microemulsion composition is achieved after complete mixing of the reactants.At the end of the reaction, the temperature is raised up to room temperature until complete hydrolysis of the excess NaBH, has occurred. Electron Microscopy The average size of the metal boride particles was measured using a Philips EM301 electron microscope in the transmission mode. For these measurements the particlesJ. B.Nagy, I. Bodart-Ravet and E. G. Derouane 191 were ultrasonically dispersed in butanol and deposited on grids covered with Formvar. The method of Feret was used to obtain the mean diameter of the particles.29 Hydrogenation Tests The crotonaldehyde hydrogenation tests were conducted at atmospheric pressure (760 * 5 Torr)? and at room temperat-ire (23* 1 "C) in a slurry-type static reactor with continuous stirring by following volumetrically the consumption of hydrogen.The catalysts were synthesized under argon flow. For the activity measurements the reaction was carried out either in ethanol or in an ethanol (90 wt "/o) microemulsion (10%) mixture, [crotonaldehyde] = 4 x lo-' mol kg-I, [M,B] = 5 x mol kg-I. Results and Discussion Solubilization Sites of Metal Ions and Sizes of Inner Water Cores of Reversed Micellar Aggregates In earlier W O ~ ~ S ~ * ~ ~ ~ ~ , ' ~ - ~ ~ l 3 Cn.m.r. measurements have shown that, in the reversed micellar system CTAB-hexan-1-ol-water, the Ni" and Co" ions are located in the inner water cores of the micelles quite close to the interface. Indeed, on average, one hexanol molecule is included in the first coordination shell of Co" ions, while one or more hexanol molecules participate in that of Nil' ions.The size of the micellar cores has been determined by an indirect method based on l9Fn.rn.r. measurements of probe molecules.3o The average radii of the aggregates ( rM) containing the precursor ions are important parameters in the understanding of the formation of colloidal particles. Preparation of Monodisperse Colloidal Metal Boride Particles Monodisperse colloidal nickel and cobalt boride particles are synthesized by reducing, with NaBH,, the metallic ions solubilized in the water cores of the microemulsions. The NaBH,/MCl, ratio was held equal to 3 because larger particles were obtained for a lower value, the particle size remaining constant above that ratio.I3 The composition of the particles was determined by XPS as being, respectively, Ni,B and Co2B.I6 In each case, the size of particles (25-706) is much smaller than th$t obtained by reduction of Ni" or Co" in water (3000-4000 A) or in ethanol (2500-3000 A) and the size distribution is quite narrow ( * 5 A).The average size of the particles decreases with decreasing size of the inner water core (decreasing water content), while a minimum is detected as a function of either Co'l or Nil' concentration (fig. 2). These observations can be understood if one analyses the nucleation and growth processes of the particles. Principles for the Formation of Colloidal Particles To form a stable nucleus a minimum number of atoms is required." At the very beginning of the reduction, nucleation only occurs in those water cores which contain enough ions to form a nucleus.At this moment, the micellar aggregates act as 'reaction cages' where the nuclei are formed. On the other hand, because the microemulsions are dynamic, the water cores rearrange rapidly. The other ions brought into contact with the existing nuclei essentially participate in their growth process. As the latter process is faster than nucleation, no new nucleus is synthesized at this moment. As all the nuclei are formed at the same time and grow at the same rate, monodisperse particles are obtained. In summary, the particle size depends on the number of nuclei formed at the very beginning + 1 Torr = 101 325/760 Pa.192 Monodisperse Colloidal Metal Bodies [Co"]/lO-* mol kg-' Fig.2. Variation of the average diameters of the cobalt boride particles as a function of Co" ion molal concentration. CTAB-hexanol-H1O: ( a ) 37%-45%-18%; ( b ) 38%-47%-15%; ( c ) 40%- 50% - 10%. of the reduction and this number is a function of the number of water cores (containing enough ions to form a stable nucleus) reached by the reducing agent before the rearrangement of the system. Quantitative Model for the Formation of Colloidal Particles The first step in the calculation of the essential parameters which control the particle size is to study the distribution of the ions in the microemulsion water cores. From the average radii of the microemulsion water cores ( rM)6*771331s-19 and the total volume of water (V,) per kg of microemulsion, the number of water cores per kg of reversed micelles ( NM) is computed, the solubility of water in the hexanol organic phase being negligible. Illsing NM and the initial concentration of metal ions expressed in mol kg-I allows one to determine the average number of ions per water core (n,ons).6 Poisson statistics give the probability (pk) of having k ions per water core ( k is an integer taking the values 0, 1,2,3 etc.), provided the average number of ions per water core ( i = nions) is g i v e n : The number of nuclei formed (N,) when the ions solubilized in 1 kg of solution are reduced is proportional to the number of aggregates containing enough ions for nucleation.If the minimum number of ions required to obtain a nucleus is i, N , can be calculated from the following relation: .s xT=,pk is the probability of having i ions or more per aggregate; hence N M is the number of water cores containing i ions or more; F is a proportionality factor taking into account the proportion of aggregates reached by the reducing agent before rearrange-J. B.Nagy, I. Bodart-Ravet and E. G. Derouane 193 ment of the system can occur. In eqn (2) we do not know the values of i and F, but we can calculate all the other parameters. Indeed, the number of nuclei ( N , ) is the number of particles prepared and the latter is given by the ratio of the total weight of catalyst and the weight of one particle. The experimental and computed data are reported in table 1 for the mixed Ni-Co-B particles. For all the particles synthesized, we have calculated the proportionality factor F, varying systematically the value of the minimum number of ions required to form a nucleus (i).Only if i takes the value 2 is F reasonably constant. The order of magnitude of F is always lop3. This means that at the very beginning of the reduction, i.e. when the nuclei are formed, only one aggregate per thousand leads to the formation of metal boride particles. The minimum in the particle size as a function of the concentration of ions can easily be explained (see fig. 2). For a constant microemulsion composition, at low ion concentration, only few water cores contain the minimum number of ions (two) required to form a nucleus, hence few nuclei are formed at the very beginning of the reduction and the size of the metal boride particles is relatively large.When the ion concentration increases, the distribution of precursor ions in the microemulsions is very different and the number of nuclei obtained by reduction increases faster than the total number of ions (fig. 3). This results in a decrease of the catalyst particle size. When more than 80% of the water cores contain two or more ions, the number of nuclei formed remains quasi-constant with increasing ion concentration. Hence, the size of the particles increases again. We have synthesized in the same microemulsion nickel boride and cobalt boride particles. In the first case, F = 3.2 x lop3, in the second case, 1 7 . 4 ~ lop3. As for these experiments, the rearrangement rate of the micelles is constant in the first approximation, the difference between the value of F probably being due to the different solvation of the two types of ions at the The Ni" ions are multiply coordinated with hexanol at the interface and their mobility is lower, hence the probability of collision between the two reduced Ni atoms required to form a nucleus is also lower. In other words, the rate of nucleation is higher for cobalt boride than for nickel boride particles.From the same micellar system, bimetallic particles of Ni-Co-B have also been prepared (table 1 ) . The average particle size and the width of the size distribution have been measured by electron microscopy. No coherent values are obtained for F if the particles are considered to be homogeneous bimetallic catalysts. On the other hand, knowing the values of F for Ni2B and Co2B in this micellar system, we have calculated the expected sizes for the case where a mechanical mixture of separate particles of the monometallic borides is formed.Nevertheless, in most cases, the experimental size distributions are too narrow to correspond to a mechanical mixture of monometallic catalysts. Hence, the particles are probably bimetallic, but not completely homogeneous. Indeed, the nucleation rate is higher for Co" ions than for Nil' ions (see above), the nuclei are formed preferentially from Co" ions and the particles contain more nickel at the surface. Fig. 4 shows quite well the method of reduction in water-in-oil microemulsions. After a fast diffusion of the reducing agent, nucleation occurs in the water droplets that contain at least two metal atoms.The nucleus is stabilized by the adsorbed surfactant molecules. The growth of the particles requires an exchange between different water cores. Finally, surfactant-protected monodisperse particles are formed which can be used directly or by depositing them on a support. Hydrogenation of Crotonaldehyde The monodisperse Ni2B, Co2B and Ni-Co-B particles were tested for the liquid phase hydrogenation of crotonaldehyde at room temperature: [M2B] = 2.5 x mol kg-',Table 1. Parameters for the formation of the monodisperse colloidal Ni-Co-B particles" 0.0 5.1 0.0 13.7 0.86 3.57 32 3.27 1.35 24.22 2.82 0.8713 3.2 '$ 0.9428 5.7 5 0.2 4.1 1 .o 14.9 0.67 4.58 28 3.27 0.9 1 35.93 5.36 0.4 3.1 2.0 15.7 0.57 5.39 22 3.27 0.44 74.32 13.0 0.9708 13.4 0.5 2.6 2.6 16.0 0.54 5.69 18 3.28 0.24 136.70 25.3 0.9774 25.9 9 0.6 2.0 3.1 16.2 0.52 5.91 23 3.28 0.51 64.3 1 12.4 0.9813 12.6 2 0.8 1 .o 4.1 16.6 0.49 6.27 17 3.28 0.2 1 156.20 31.9 0.9862 32.3 5 0.0 5.1 16.7 0.48 6.40 21 3.28 0.39 84.10 17.5 0.9877 17.7 - 5 1 .o 2 E 2 2 CTAB 18.0 wt %-hexan-1-01 70.0%-water 12.0%.(I Homogenous distribution of nickel and cobalt is supposed. ' Values given for the system containing three quarters of the total amount of water. " Values given for the system before reduction of metal ions. Values given for 1 k of solution. '' W, is calculated on the basis of homogeneous Ni-Co-B particles: MW (NiCoB) = (1 - x ) MW(Ni,B)+x MW(Co2B) with MW(Ni2B) = 128.23 g and MW(Co2B) = 128.68 g. f~ is calculated with M,(Ni2B) = 7.9 g cmP3 and M,(Co,B) = 8.1 g ~ m - ~ .J.B.Nagy, I. Bodart-Ravet and E. G. Derouane 195 [Co"]/ lo-' rnol kg-' Fig. 3. Variation in the number of nuclei formed per aggregate and of the probability to have two or more ions per aggregate as a function of Co" concentration in the microemulsion 37% CTAB-45% hexan-l-ol-18% water. [crotonaldehyde] = 4.0 x lo-' mol kg-', pH, = 760 * 5 Torr and T = 23 i 1 "C. The tests were carried out under continuous strong stirring in order to avoid the diffusional limitations on the catalytic activity. In table 2 the initial rates of hydrogenation are compared for Ni2B and Co,B particles synthesized in situ in ethanol and in the reversed micellar system (CTAB 18Y0-hexanol 70 O/O -water 12 "/0 ) . The Ni2B particles synthesized in ethanol are quite large ( d = 2500-3000 A) and they aggregate strongly.The hydrogenation occurs essentially on the C=C double bond, yielding butanal. No doubly hydrogenated product butanol is detected. Nevertheless the reaction is not very selective as only 20-25% of butanal is formed. A side reaction probably occurs between butanal and crotonaldehyde, but the reaction products have not been identified. The micellar system has a small inhibitory effect on the reaction (A and B in table 2), due essentially to the adsorption of B(OH),, Br-, Cl-- and CTAt ions.'" The micellar system can also decrease the aggregation of the particles, leading to an overall increase in the initial rates (A and C). Finally, if the particles are prepared from the micellar system, their size is small ( d = 32 f 5 A) and consequently the hydrogenating activity is the highest (A and D).The hydrogenation on Co,B particles leads to the C=O double bond hydrogenation product, butenol. However, the doubly hydrogenated product butanol is also observed in a non-negligible amount. For example, after some 80% overall conversion, the reaction mixture contains 40 rnol '/o butenol, 20 rnol '/o butanol and 20 rnol '/o croton- aldehyde. The micellar system again shows an inhibiting effect on the reaction (E and F) due to the strong adsorption of the above-mentioned species. In addition, the aqueous layer accompanying the adsorbent species can also have a negative effect by decreasing the crotonaldehyde concentration near the surface. As the CozB particles are not aggregated in the different reaction media, no increase in activity is observed owing to the inhibition of aggregation.Moreover, the much smaller particles obtained from microemulsions196 Monodisperse Colloidal Metal Bodies NaBH4/water diffusion o f BH, (fast 1 no nucleation I nucleation (slow) -4 I w a w- exchange of water content t e r rearrangement) ’ J ’ nucleus stabilized by surfactant absorption particle growth (fast) surf ac t an t co s ur fac t a n t + - surfactant - sta bi I ized particle organic phase Fig. 4. Model for the preparation of monodisperse colloidal particles from water-in-oil micro- emulsions. ( d = 21 f 5 A) are not able to attain the activity of the reference compound ( E and G). This is probably due to the much stronger inhibition of the crotonaldehyde adsorption, which is supposed to occur on the C=O double bond.’6 On the mixed Ni-Co-B particles, the reaction occurs essentially on nickel, leading to butanal.The activity varies linearly as a function of the amount of nickel in the c a t a l y ~ t * ~ ” ~ (table 3). This phenomenon can be explained by the much lower activity of Co,B than Ni2B.*‘ Nevertheless, a small amount (5- 15% ) of butanol is also formed on Co sites from butanal obtained on Ni sites, but the initial hydrogenation rate only characterizes the C=C bond hydrogenation.J. B.Nagy, I. Bodart-Ravet and E. G. Derouane 197 Table 2. Initial hydrogenation rates of crotonaldehyde on Ni2B and Co2B particles reaction medium synthesis/ ethanol micellar solutionu reaction medium (wt Y o ) (wt O/O ) R,/"/o conv min-' NizB catalyst A ethanol 100.0 0.0 Bh ethanol 90.0 10.0 C" ethanol 90.0 10.0 D micelle 90.0 10.0 E ethanol 100.0 0.0 F ethanol 90.0 10.0 G micelle 90.0 10.0 Co,B catalyst 13.1 11.7 25.5 48.0 4.0 1.7 2.8 Composition in wt %:CTAB 18%-hexan-l-o1 70%-water 12%.The Ni,B particles are aggregated. " The Ni,B particles are not aggregated. Table 3. Initial hydrogenation rates (Ri) of crotonaldehyde on mixed Ni-Co-B particles as a function of cobalt atom fraction x 'uncleaned' particles 'cleaned' particles X R , / Bad\(' 0/0 conv. min-' g-' (%) 0.0 0.2 0.4 0.5 0.6 0.8 1 .o 16.0 57 12.9 63 5.0 59 3.2 61 8.3 59 0.6 58 0.2 59 74.3 75.2 10.1 25 .O 10.5 1.4 0.5 0.60 2 0.59 7 0.06 11 0.18 5 0.10 3 0.01 8 0.004 10 Percentage of adsorbed boron with respect to the amount of boron remaining in the solution after reduction ("B n.m.r.results). In order to increase the catalytic activity of the monodisperse particles care has been taken to eliminate the adsorbed ions from the surface. Indeed, it has been shown by B n.m.r., that some 60% of the total amount of boron remaining in solution is adsorbed on the metal boride particles. In addition, a certain amount of Br- and C1- ions can also be found on the surface, solvated by a water layer. CTA+ ions are finally retained as counterions together with Na' cations. The presence of these ions was detected using XPS? The monodisperse Ni2B, Co2B and Ni-Co-B particles were prepared in a glove-box: 100 g of micellar system containing 5.0 x lo-* mol kg-' metal chloride. The micellar system is first eliminated by suction, then the remaining particles are washed twice with 12 cm3 HCl (0.1 mol dm-3), three times with 40 cm3 H 2 0 and twice with 40 cm3 acetone- ethanol ( 1 : 1 volume ratio).Finally they are dried at room temperature under vacuum in the glove box. Despite the potential oxidation during the different preparation steps, the 'cleaned' particles are more active than the uncleaned ones (table 3). The residual boron content 1 1198 Monodisperse Colloidal Metal Bodies is higher on the cobalt-containing particles, showing that the negative species are more strongly adsorbed on this catalyst. Finally, it is also possible to determine the initial rate per unit surface, using the B.E.T. values of the particles.16 Conclusion This paper describes the preparation of monodisperse colloidal metal particles when the average diameter is rather small (< 100 A).The role of microemulsions is emphasized and a mechanism is proposed for the formation of colloidal particles. This medium allows the preparation of colloidal nickel and/or cobalt boride particles, whose size is influenced by the metal-ion concentration expressed with respect to water and by the number and size of the micellar aggregates. An important factor in controlling the particle size is the nucleation process at the very beginning of the reduction, and the importance of the compartmentalization of the reaction medium is emphasized. The room-temperature hydrogenation activity of Ni2B is enhanced in presence of the microemulsion, because the particle aggregation is impeded.On the other hand, Co,B is less active, essentially because of the strong adsorption of B(OH),, C1-, Br- and CTA' ions. When the particles are cleaned, the inhibition is decreased and a significantly higher activity is observed. References 1 G. A. Ozin, Chemrech, 1985, 488. 2 H. Hirai, Y. Nakao and N. Toshima, J. Macromol. Sci. Chem., 1978, A12, 1117; 1979, A13, 727. 3 P. Gallezot, Caral. Rev. Sci. Eng., 1979, 20, 121. 4 J. H. Fendler, Chem. Rev., 1986, 87, 877. 5 M. Haruta and B. Delmon, J. Chim. Phys., 1986, 83, 859. 6 J. B.Nagy, E. G. Derouane, A. Gourgue, N. Lufimpadio, I. Ravet and J-P. Verfaillie, in Surfactants in Solution, Modern Aspects, Proc. Sixrh Int. Symp., New Delhi, 1986, ed.K. Mittal (Plenum Press, New York, in press). 7 J. B.Nagy, Colloids Surf, 1988, in press. 8 P. Stenius, J. Kizling and M. Boutonnet, U.S. Par. 4 425 261 (1984). 9 A. Wathelet, Me'rnoire de Licence (Facultes Universitaires Notre-Dame de la Paix, Namur, 1984). 10 A. Claerbout, Ph.D. thesis (Namur), in preparation. 11 M. Boutonnet, J. Kizling, P. Stenius and G. Maire, Colloids Surf.', 1982, 5 , 209. 12 K. Kurihara, J. Kizling, P. Stenius and J. H. Fendler, J. Am. Chem. Soc., 9883, 105, 2574. 13 J. B.Nagy, A. Gourgue and E. G. Derouane, Srud. Surf: Sci. Caral., 1983, 16, 193. 14 D. Rosier, J. L. Dallons, G. Jannes and J. P. Puttemans, Acra Chim. Hung., 1987, 124, 57. 15 I . Ravet, J. B.Nagy and E. G. Derouane, Srud. Suif Sci. Catal., 1987, 31, 505. 16 I. Ravet-Bodart, Ph.D. thesis (Facultes Universitaires Notre-Dame de la Paix, Namur, 1988). 17 I. Ravet, A. Gourgue, Z. Gabelica and J. B.Nagy, Proc. 8th Int. Congr. Catalysis, West-Berlin, 1984 18 I. Ravet, N. B. Lufimpadio, A. Gourgue and J. B.Nagy, Acra Chim. Hung., 1985, 119, 155. 19 1. Ravet, A. Gourgue and J. B.Nagy, Surfactants in Solutions, ed. K. L. Mittal and P. Bothorel, (Plenum Press, New York, 1987), p. 697. 20 R. Paul, P. Buisson and N. Joseph, C.R. Acad. Sci. Paris, 1951, 627. 21 H. C. Brown and C. A. Brown, J. Am. Chem. Soc., 1963, 85, 1003; 1005. 22 Y. Okamoto, Y. Nitta, T. Imanaka and S. Teranishi, J. Catal., 1980, 64, 397. 23 D. G. Holah, I. M. Hoodless, A. N. Hughes and L. Sedor, J. Catal., 1979, 60, 148. 24 J. A. Schreifels, P. C. Maybury and W. E. Swartz Jr, J. Caral., 1980, 65, 195. 25 T. W. Russel, R. C. Hoy and J. E. Cornelius, J. Org. Chem., 1972, 37, 3552. 26 Y. Nitta, T. lmanaka and S. Teranishi, Bull. Chem. Soc. Jpn, 1980, 53, 3154. 27 M. Kajitami, J. Hasegawa, E. Kasai and A. Sugimori, Tetrahedron Lett., 1979, 2407. 28 S. W. Heinzman and B. Ganem, J. Am. Chem. Soc., 1982, 104, 6801. 29 J. R. Anderson, Structure of Metallic Catalysts (Academic Press, New York, 1975). 30 T. Nguyen and H. H. Ghaffarie, C.R. Acad. Sci. Paris. Ser. C, 1980, 290, 113. 31 P. C. Hiemenz, Principles of Colloid and Surface Chemistry (Marcel Dekker, New York, 1977), p. 234. (Verlag Chemie, Weinheim, 1984) vol. IV, p. 871. Paper 8/0505OC; Received 19rh December, 1988

ISSN:0301-7249    

DOI:10.1039/DC9898700189

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

Faraday Discuss. Chem. SOC., 1989, 87, 199-205 GENERAL DISCUSSION Dr J. A. H. MacBride (University of Durham) (communicated): In addition to my comments on the paper by Bond et af., two further comments arise from the mechanisms presented by Dr Serwicka. First, when molecular orbitals have contributions from two or more inequivalent atoms which are potential sites for reaction, in which the MO is the 'frontier' orbital (e.g. the HOMO of a nucleophile), the atom with the highest coefficient in the ground-state is not necessarily the site of fastest reaction. This arises, of course, because the MOs may be so perturbed as reaction proceeds that the ground-state calculations do not predict the relative activation barriers correctly. ' Fur- thermore, this approach is inapplicable if equilibrium between the products (of some step) is attained, as might be the case for methylation on bridging or terminal oxygen atoms of the Keggin structure.The conclusion that the Keggin-oxygen is behaving as a nucleophile in the rate-determining step of oxidation also seems open to question. The second point concerns the mechanism of oxidation of alkenes to maleic anhy- dride. It is not necessary to infer a furan derivative as intermediate since maleic acid forms its anhydride' (in the absence of dehydration catalyst) at 200 "C (and its trans isomer at a somewhat higher temperature) so that cyclisation after oxidation is not improbable. Preference for C,-terminally oxidised products (all leading to maleic anhydride) may be due to preferential (radical) oxidation of positions adjacent to an initial non-terminal double bond (ally1 positions) and to preferential formation of a double bond (by oxidation-dehydration, or dehydrogenation) conjugated with an initial carbonyl function: I I *CH2 CH II * CH2 I I - II CH2 CH I I c=o I X I c=o I X C02H I 9 CH -H$ HC'4 II CH I - II p HC.II 0 CO2H *radical stabilised positions X = alkyl, H, or OH 1 See for example: I. Fleming, Frontier Orbitals and Organic Reactions, (John Wiley and Sons, 1976), p. 23 ff. 2 J. D. Roberts and M. Caserio, Basic Principles of Organic Chemisfry, (W. A. Benjamin Inc., 2nd edn, 1977) p. 847. Dr E. M. Serwicka replied: Any mechanism postulated should take into account the available experimental evidence. In the case of methanol interaction with the Keggin unit, adsorbed protonated methanol [11, fig.9, p. 1851 has been detected with PAS-FTIR' 199200 General Discussion and 13C m.a.s. n.m.r.* identified methylation of bridging oxygen sites (111, fig. 9). Additionally, e.s.r. i.r.3 and TPD4 data on methanol interaction with the Keggin unit indicate that major reaction steps have threshold temperatures. At T < 453 K methanol chemisorption with the electron transfer to the Keggin unit occurs to give an e.s.r. signal typical of a reduced but not oxygen-deficient Keggin anion, with the characteristic Keggin i.r. spectrum virtually unperturbed (V, fig. 9). There is no measurable formal- dehyde evolution at this stage (TPD). At higher temperatures, desorption of formal- dehyde (TPD) occurs which leaves behind the Keggin unit containing a bridging oxygen vacancy (i.r.) at which a trapped electron gives a characteristic e.s.r.~ i g n a l ~ - ~ (VI, fig. 9). Any mechanism alternative to that proposed by us would have to deal with these facts. The mechanism you suggest (see comments on the paper by Bond et al.) is basically similar to ours, except for the transfer of electrons rather than the transfer of protons in the initial step. Our intention was to emphasise that the Keggin unit has strong acidic properties and the proton affinity of methanol will be the driving force to form the adsorbed methyl oxonium moiety. Configuration IV is the only one in our scheme that lacks direct experimental support and as such remains speculative. However, transition I11 + V in fig.9 requires a proton abstraction and a transfer of two electrons to the Keggin unit. This concerted operation may be presented as proceeding in two steps, the first being an attack of the Keggin bridging oxygen lone pair (HOMO) on the methoxy group LUMO localized predominantly on the CH3. In view of the antibond- ing character of the latter this weakens the C-H bond and enables a facile proton transfer to give a virtual configuration, IV, which will rapidly reform, releasing excess electrons to the catalyst via the carbon-oxygen v-system, to end up as configuration V. Concerning the applicability of a HOMO-LUMO approach we are aware of its limitations but in the present case the theoretically predictable reaction pathways agree well with those observed experimentally, indicating the usefulness of such considerations.1 J. G. Highfield and J. B. Moffat, J. Catal., 1985, 95, 108. 2 W. E. Farnett, R. H. Staley, P. J. Dorraille and R. D. Farlee, J. Am. Chem. Soc., 1987, 109, 4018. 3 E. M. Serwicka, unpublished results. 4 E. M. Serwicka, E. Broctawik, K. Bruckman and J. Haber, Catal. Lett., 1989, 2, 351. 5 E. M. Serwicka, J. B. Black and J. B. Goodenough, J. Catal., 1987, 106, 23. Dr J. C. Vedrine (Institut de Recherches sur Catalyse, CNRS, Villeurbanne, France) said: You have made theoretical calculations of LUMO and HOMO orbitals of CH,OH adsorbed on bridging oxygen atoms of your Keggin unit which helps you to propose a reaction mechanism for oxidation into HCHO. However you have to make the assump- tion that terminal >Mo=O oxygens are inactive.I would have liked a theoretical approach without such an assumption. My question is therefore to know if you have made the calculation for CH,OH adsorption on Mo=O oxygen. If so, are the molecular orbitals very different? Prof. R. A. van Santen (University of Technology, Eindhoven, The Netherlands com- mented: I would like to point out that the predicted reactivity of bridging 0 species agrees with the results of Pauling's valency rules. I would like, to ask whether it would not be better to study clusters containing Mo as well as V, since the acidic proton coordinated to the bridging oxygen is generated by vanadium. If one were to apply thermodynamic arguments, which of the oxygens (bridging or monocoordinated) would be most active? Prof.J. Haber replied: From the solution "0 n.m.r. study' it can be inferred that introduction of vanadium leads to an increased electron density on bridging oxygens as compared to the unsubstituted 12-molybdophosphate anion but of the two options it is the bridging oxygen of the Mo-0-Mo type rather than Mo-0-V type thatGenera 1 Discussion 20 1 carries higher charge. Therefore in our mechanistic scheme we include Mo centres only. Nevertheless, calculation of the electronic structure of a V-substituted Keggin anion would certainly improve the validity of discussed HOMO-LUMO interactions. 1 R. I . Maksimovskaya, M. A. Fedotov, V. A. Mastikhin, L. T. Kuznetsova a n d K. I . Matveev, Dokl. Akad. Nauk SSSR, 1978, 240, 117. Prof. V. Ponec ( Gorlaeus Laboratories, Leiden University, The Netherlands) suggested: According to your mechanism all oxygen in formaldehyde molecules originates from the oxide lattice and not from CH30H.Perhaps, the validity of this conclusion can be checked by using CH,"OH. The exchange reaction of H2'80 with the oxide probably spoils it, but the 160/'80 ratio in CH'O could also be informative with such a complica- tion, for example, if one of the extreme values (zero or a) were found. Prof. Haber responded: In view of the heteropolyacid ability to exchange rapidly bulk oxygen with water vapour' (at 523 K 80% of 0 atoms within 1 h), it would be difficult to design a sensible experiment with labelled molecules. In our opinion, the use of 180-labelled catalyst, rather than CH3180H could provide the unequivocal answer.1 M. Misono, K. Sakata and Y. Yoneda, Proc. 7th Int. Cong. Catal., Tokyo, 1980, p. 1047. Prof. J. B.Nagy (FUNDP, Namur, Belgium) asked: You have emphasized that the methyl group should be attached on the bridged oxygen rather than on the terminal oxygen in heteropolysalts. Is the difference in energy between these two forms high enough? If not, at higher temperatures, a random distribution of the methyl groups between the two types of oxygen could occur due to the mobility of the methyl groups. Prof. Haber answered: On increasing temperature only bridging oxygens are extracted from the Keggin anions (i.r., e.s.r.). This makes us believe that the mobility of methyl groups, also at elevated temperatures, is limited to the network of the Keggin bridging sites.Dr G. J. Hutchings ( University ofliverpool) said: In the proposed mechanistic scheme for methanol conversion over heteropolyacid catalysts (your fig. 9) it is proposed that the methanol interacts with the Keggin unit to produce a methyloxonium compound (111). I t is proposed that intermediate I11 is subsequently deprotonated by an adjacent oxygen to form a methylene oxonium ylide type intermediate (IV). In our studies of methanol conversion to hydrocarbons over the pentasil zeolite H-ZSM-5 we have proposed similar intermediates in the formation of the initial carbon-carbon bond.',' By analogy with organometallic chemistry it would be considered that structures such as intermediate IV would be very reactive to gas-phase oxygen. In this case it would be gas-phase oxygen and not lattice oxygen that would be incorporated into the product formaldehyde.It is possible to comment on this possibility? 1 G. J. Hutchings, M. V. M. Hall, F. M. Gottschalk and R. Hunter, 1. Chem. Soc., Faradajv Trans. I , 1987, 83, 571. 2 G. J. Hutchings, L. Jansen van Rensburg, W. Pick1 and R. Hunter, 1. Chem. Soc., Faraday Trans. I , 1988, 84, 131 1. Prof. Haber responded: Indeed, our mechanism is formally similar to that proposed in your studies of the C-C bond formation on ZSM-5. However, a basic question may be raised as to whether the species presented as IV in fig. 9, interacting with a transition- metal oxide system can be considered as analogues of ylide structures appearing at the surface of a zeolite catalyst. At variance with oxides of main group elements, empty levels in the d-band are present in the transition-metal oxide catalyst.Therefore, any202 General Discussion excess negative charge on the carbon atom will be immediately transferred to the d-band making the transient ylide-like structure short-lived. The resulting methene group will be much less prone to the attack by molecular oxygen. Prof. G. Centi (Department of Industrial Chemistry and Materials, Bologna, Italy) remarked: My question regards the role of V in the heteropolyacids, with specific reference to the selective transformation of a1 kanes. We are working on this problem, but our preliminary results indicate a more complex situation than that which you outlined. When all the vanadium is inside the Keggin structure in substitution of Mo the heteropolyacid is not very active in butane oxidation. The activation of butane requires the presence of some V sites in a cationic position, external to the Keggin unit.These results indicate a direct participation of V sites in the mechanism of activation of butane, and not only in the modification of electron density of bridging oxygen as you suggested. In our opinion the mechanism is not very different from what is observed for V-P oxides' on which experiments with labelled molecules' clearly indicate that the H atoms are abstracted from carbon atoms 2 and 3 and not 1 and 4. Have you other evidence regarding your hypothesis of the mechanism of alkane activation on V-heteropolyacids? 1 G. Centi, F. Trifiro, J. R. Ebner, V.H. Franchetti, Chem. Rev., 1988, 88, 55. 2 M. A. Pepera, J. L. Callahan, M. J. Desmond, E. C. Milberger, P. R. Blum and N. J. Bremer, J. Am. Chem. SOC., 1985, 107, 4883. Prof. Haber replied: The suggested way of activation of alkanes over heteropolyacids is an attempt to employ reasoning along the line of possible HOMO-LUMO interactions. At present there is no experimental evidence that it is indeed the operating reaction pathway. Nevertheless, as you certainly remember, our common work on selective oxidation of pentane [ref. (8), p. 1871 has shown that there is a striking difference in the behaviour of a V-substituted heteropolyacid and a standard V-P-0 catalyst. The former gives exclusively maleic anhydride, the latter a mixture of phthalic and maleic anhydrides.In view of this, one has to be very cautiouc- in usir;g a per analogiam argument in mechanistic considerations concerning both s j stems. Clearly, a detailed study, preferably with labelled molecules, is required to elucidate this problem. Finally, Prof. J. B. Moffat (comrnunicared ): The equation provided by you 0 A n - G U V , O - G cos 6 + uV20, -An cos 6 appears to be a modified version of the Young-Dupre equation. However, it is question- able whether the assumptions implicit in the Young-Duprk equation, namely those which lead to the disappearance of the Gibbs surface chemical potentials, are valid in the system under study. Would you care to comment'? Prof. Haber, in response, said: From the thermodynamic point of view the situation at the solid/solid/gas line is fully analogous to that assumed for the solid/liquid/gas line and the same surface free-energy relationship as well as equilibrium condition operates in both cases.However, a rigorous derivation of the equivalent Young equation is not an easy exercise in the former case. As in the description of thin-oxide layers we are interested mainly in predicting the evolution of the system with time, it is more convenient to use the concepts of the work of cohesion and adhesion and describe spreading as proceeding when the energy of adhesion of the mobile phase to the immobile support is greater than its energy of cohesion. Prof. W. Palczewska (Polish Academy of Sciences, Warsaw, Poland) asked the first question on Prof. B.Nagy's paper: Could you specify the kind of reaction used andGeneral Discussion 203 Table 1.Size of the NizB particles synthesized from CTAB-hexan- l-ol-water microemulsions [Ni(II)] x 102/mol kg-’ d a / A u b / A n ‘’ CTAB( 12.0% )-hexanol (80.0% )-water (8.0% ) CTAB( l8.O%)-hexanol (70.0°/0)-water (12.0%) 2.5 31 4 773 5.0 32 4 1001 Average diameter of the particles; ” standard deviation; ( number of measured particles. the temperature range for the catalytic activity of the metal-boride particles? Did you examine the influence of boron on the catalytic activity? Prof. J. B.Nagy replied: We essentially checked the hydrogenating activity of the boride particles in the liquid phase at room temperature. In order to increase the temperature, the particles should be deposited on a support to prevent coagulation.We are currently working on the problems using Pt and Pt-ReO, particles.’ Two patents are also available in the literature using deposited collodial monodisperse particles. In the first case,’ the deposition of the particles on a support follows their preparations in the microemulsion. In this case, however, a broad metallic size distribution is obtained, due to the aggolomeration during impregnation or transfer onto the surface of the support and a requisite thermal stability cannot be achieved.3 The second method3 consists of reducing the metal ions directly on the surface of the support after impregna- tion of the microemulsion. In the latter case, the microemulsion components are eliminated by evaporation and helium purge. The influence of boron on BOT on the surface of the metal-boride particles was not examined systematically with respect to the catalytic activity of the boride particles and, unfortunately, I cannot comment further on this topic. 1 A.Claerbout, PhD. Thesis (Namur, in progress). 9 P. Stenius, J. Kizling and M. Boutonnet, U.S. Patent, 1984, 4425261. 3 H. Abrevaya and M. Targos, U.S. Patent, 1987, 4714692. Prof. J. M. Thomas (The Royal Institution, London) then said: The dispersity of the particles shown on your transparency is 5-10 A. Are they really monodisperse? Prof. J. B.Nagy answered: Indeed, the dispersion of the particlesoshown by the transparency is within 5-10 A. The smallest distribution obtained is 4 A as shown by table 1 [ref. (16) of our paper]. It has to be noticed that the size-distribution is comparable to previously reported values [ref.( 1 1 ) of our paper]. Dr P. A. Sermon (Brunel Uniuersity, Uxbridge) (communicated): Your colloidal routes to metal borides in microemulsions use excess NaBH,. Were you able to: (i) vary the Me/B ratio, (ii) remove CTAB and NaBH, contaminants, and were the surfaces so produced homogeneous? Prof. B.Nagy responded: (i) The excess NaBH, used over MC12 (M = Ni and/or Co) is ca. three-fold. When a higher excess was used, it did not influence the particle size, whereas a lower excess led to imperfect reduction of the metal ions. The excess is necessary, because part of the NaBH, reagent is hydrolysed even at a temperature close to 0°C.204 General Discussion Ek Fig. 1. X.P.S. spectrum of Ni,B particles ( Ni2p3,? of Ni,B and NiO) synthesized from CTAB( 18%)- hexan-l-ol(70% )-water( 12%) microemulsion with [Ni(ii)] = 5 x 10 mol kg-'. NiO has & = 855.2 eV, Ni2B has Eh = 851.7 eV.(ii) The strongly adsorbed contaminants are CTA+, Br-, C1- and B(OH), ions. The washing procedure to eliminate most of these ions is described in the text. Nevertheless, BOT contaminant still remains on the surface, together with the NiO and COO oxides, stemming from the easy oxidations of the highly reactive boride particles. The Ni2B, Co2B and Ni-Co-B particles thus prepared are amorphous, the mixed particles do contain both Ni and Co atoms, and they do not correspond to a mechanical mixture of separate Ni2B and Co2B particles. Two pieces of evidence strongly support this hypothesis.( a ) The nucleus of the particle is essentially made of Co2B, hence the outer rim does contain more Ni2B entities (see text); ( b ) the catalytic hydrogenation of crotonaldehyde as a function of bulk Ni content does not follow the strict proportionality curve, but it shows a slightly higher activity, especially for low Ni content, which is attributed to the surface enrichment in Ni2B [ref (15) of our paper]. These results also show clearly, that the particles are not homogeneous. Prof. M. W. Roberts (University of Wales College of Cardim asked: Do you have any information regarding the oxidation state of 'nickel' in the nickel borides from the Ni(2p) spectra? This together with the valence-band density of states might throw some light on the high catalytic activity of NiB.Prof. B.Nagy replied: The X.P.S. spectra of Ni in Ni2B show a certain degree of charge transfer from boron to nickel.',* Indeed, the binding energy of boron in Ni2B is increased ( E = 187.7 eV) with respect to the element B (& = 186.2 eV3, and E b = 187.3 ev"). In contrast, the binding energy of Ni is decreased: Eb = 851.7 eV with respect to Ni metal (Eb = 852.1 eV3 and E b = 852.7 eV5) (fig. 1). The higher electron density on Ni in Ni2B could also explain the higher activity of Ni2B with respect to Ni metal. 1 I . Bodart-Ravet, Ph.D. Thesis (Facultes Universitaires Notre-Dame d e la Paix, Namur, 1989). 2 Y. Okamoto, Y. Nitta, T. Imanaka and S. Teranishi, J. Chem. SOC., Faradaj? Trans. I , 1979, 75, 2027. 3 P. C. Maybury and W. E. Swartz Jr, J. Catal., 1980, 65, 195. 4 V. G . Aleshin, T. Y. Kosolapova, V. V. Nemoshkalenko, T. I . Serebryakova a n d N. G. Chudinov, J. Less-Common Met., 1974, 67, 173. 5 B. P. Lochel a n d H. H. Strehblow, J. Electrochem. Soc., 1984, 131, 713. Prof. P. B. Wells (Hull University) commented: The presence of C1-, Br-, BOY ions adsorbed on the metal particles can be distinguished by using the hydrogenation reactionGeneral Discussion 205 of either buta-1,3-diene or isoprene. In the presence of C1-, Br- ions 1,4-addition is favoured, while 1,2-addition predominates without these ions. Prof. B.Nagy responded: Thank you for your comment concerning the detection of C1- and Br- ions adsorbed on the metal-boride particles. The hydrogenation of buta- 1,4- diene (or of isoprene) could be used and the products would be but-2-ene, in the presence of C1- and Br- ions (1,4 additions), and but-1-ene in the absence of these ions (1,2-addition). It certainly would be very interesting to check.

ISSN:0301-7249    

DOI:10.1039/DC9898700199

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

Furuday Discuss. Chem. SOC., 1989, 87, 207-213 Role of Surface Atomic Arrangements of well defined Phosphates in Partial Oxidation and Oxidative Dehydrogenation Reactions Jacques C. Vedrine," Jean Marc M. Millet and Jean Claude Volta Institut de Recherches sur la Catalyse, associe' h 1' UCB, Lyon I, LP-CNRS, 2 Avenue Albert Einstein, F 69626 Villeurbanne Ce'dex, France Various vanadium and iron phosphates have been synthesized and character- ized physically before and after catalytic reaction has taken place. Vanadium phosphate was used for butane oxidation into maleic anhydride and iron phosphate for isobutyric acid dehydrogenation to methacrylic acid. Vanadium phosphate catalyst in its working state corresponds to a mixture of the pyrophosphate phase (VO)2P207 and a given VOP04 phase (a,,, p or 8).31P MASNMR spectra showed that the most efficient catalysts con- tained VOP04 entities in interaction with (VO)zP207, while an X-ray elec- tronic radial distribution study showed that VOP04 entities are present but not necessarily organized in sufficiently long-range order to be detected by classical X-ray diffraction. The iron phosphates Fe2Pz07, FePO, and Fe7( PO4), were transformed under catalytic reaction into Fe,( PO4), and an unknown phase. The latter phase was tentatively assigned from Mossbauer spectroscopy to Fe,P,02, with an Fe3+/Fe2+ ratio equal to two. The best catalysts including the industrial catalyst were those containing the Fe,P,O,, phase. A similarity has been found in the behaviour of vanadium and iron phosphates.The best catalysts correspond to those which exhibit short- or long-range order between the reduced pyrophosphate phase and a more oxidized form as VoPo4 or Fe6P6023. It is suggested that the ease of transformation of two PO4 groups, at least locally, into one P207 group is the clue to obtaining an efficient catalyst. This is a case of topotactic transformation. 1. Introduction Metallic oxides constitute one of the widest families of current industrial catalysts. They correspond to insulators such as alumina, silica-alumina, MgO, zeolites etc., and of semiconductors such as transition-metal oxides. The simple idea that the best geometrical fit between a reactant molecule and surface atomic arrangements enhances catalytic properties has largely been accepted for many years. This has recently been shown'.* unambiguously in our laboratory for Moo3 single crystals.Since that time similar results were obtained in other l a b o r a t o r i e ~ , ~ ~ ~ and the idea has become widely accepted. For instance for Moo3 the (100) face was found to be selective for acrolein in propene oxidation reaction while the (010) face gave carbon oxides exclusively. For methanol conversion on the same catalyst the (100) face was observed to be both dehydrogenating and dehydrating, while the (010) face was only dehydr~genating.~ The purpose of this paper is to present some of our data which show that for several phosphate catalysts the best ones undergo facilitated oscillation of surface structure between two topotactic phases during catalytic reaction, i.e. specific surface atomic arrangements. 2. Experimental Two phosphate families were prepared: vanadium and iron phosphates. 207208 Surface Atomic Rearrangements Table 1. Chemical composition of iron phosphate catalysts sample formula surface area method of preparation /m*g-' P/Fe A FePO, precipitation, FePO4.2H,O, calcination 723 K 6.4 1.02 B Fe2P20, reduction of A in N2-H2-H20 at 1023 K 6.1 0.97 C Fe7(P0J6 precipitation and reduction in N2-H2-H20 0.4 0.86 2.1 Vanadium Phosphates Two procedures were In an aqueous medium V2OS was reduced by HCl then H,PO, was added to give a P/V ratio slightly greater than one. After removal of the HCl solution the VPO paste was dried under vacuum at 80°C and then pelletized. In an alcoholic medium VzOs and H3P0, (P/V> 1) were refluxed with isobutyl alcohol as a reducing agent for 24 h under nitrogen.The blue solid obtained was filtered, washed with isobutyl alcohol and dried under vacuum at 80°C. The blue paste was then hydrolysed and dried at 100 "C for 6 h. After preparation the samples were activated under the reactant mixture, namely 1-2 wt '/o butane in air at 400 "C with 2000 h-' WHSV. 2.2 Iron Phosphates We have chosen three samples within the Fe203-Fe0-P205 phase diagram, namely Fe2P2O7 (Fez+), FePO, (Fe3') and Fe,(PO,), (Fe2+, Fe3+) since it is known that the oxidation mechanism involves two oxidation states of the cations. FePO, was prepared by calcining phosphosiderite, FeP04.2H20, synthesized by successive evaporations of an aqueous solution of iron chloride and phosphoric acid.Fe2P20, was obtained by reduction of FePO, at 750 "C in an N2-H2-H20 gas mixture.' Fe7(P0,), was prepared by reduction of a precipitate of the appropriate stoichiometry obtained with iron nitrate and diammonium hydrogenophosphate.8 Chemical compositions of the materials were determined by atomic absorption. The data are summarized in table 1. 2.3 Catalytic Reactions Butane oxidation was performed in a fixed-bed stainless-steel tube reactor (i.d. == 1 cm) placed in a salt melt environment.' The catalyst was pressed into small platelets ca. 0.5 cm in diameter and a few mm in height and introduced into the metallic tube. The amount of catalyst was ca. 15 g over ca. 20 cm. 1-2 wt YO butane in air was used as a feed and the analysis was carried out by on-line gas chromatography. The reaction temperature was 400°C.A steady state was reached after a few hours under high conversion conditions. Oxidative dehydrogenation of isobutyric acid was performed in a flow microreactor containing 50-100 mg of catalyst. The total flow rate was 1 cm3 s-l with partial pressures (in kPa) equal to I BA : H 2 0 : O2 : N2 = 5.94 : 77.2 : 4.32 : 15.4. The side- products were C02, propene and acetone. Catalytic features were measured under steady-state conditions at 380 "C (653 K). 3. Results and Discussion 3.1 Vanadium Phosphate Materials Vanadium phosphate is the best known catalyst for the oxidation of butane into maleic anhydride and has been studied extensively."'J. C. Vedrine, J. M. M. Millet and J. C. Volta 209 Table 2.MASNMR and catalytic data for several vanadium phosphate catalysts [from ref. (5) and (16)] for butane oxidation at 400 "C into maleic anhydride (MA) ~ ~~ ~ samples 6 (PPm) conversion ( '/o ) selectivity (9'' ) MA yield ('% ) catalyst catalyst - -30 and -40 74 -40 (sh) and -30" 89 75 74 56 66 " Weak intensity (sh = shoulder). The precursors which have been prepared correspond to the VOH P04-0.5H20 phase, as shown by X-ray diffraction. However, after activation and catalytic reaction (VO)2P207 was observed as the primary component. Depending on the preparation and activation conditions, several VOPO, phases at low concentration were also seen in XRD measurements. These could be one of the four known VOPO, phases ( a , p, y and ;5),"." which differ in the way the VO, octahedra and PO, tetrahedra are joined. A sample prepared following the aqueous method exhibited poor catalytic perform- ance with only 30% yield in maleic anhydride.The XRD pattern corresponds to broad peaks assigned to (VO)2P207. X-Ray radial electronic distribution ( RED)'s shows that the catalyst contains the (VO)2P207 phase, but it also shows bond lengths corresponding to p-VOPO, entities. RED technique is sensitive to local order, i.e. it gives information about amorphous compounds undetectable by classical XRD. Samples prepared in an alcoholic medium were observed to contain (VO)2P207 and to a lesscr extent the VOPO, phases all, y and 8. These samples exhibit much better catalytic properties for maleic anhydride. A 3'P MASNMR study was carried out on pure (VO)2P207, the a l l - , /3-, y- and 6-VOPO, phases and on several samples exhibiting different catalytic behaviour.531h (VO):P207 was observed to give a very broad spectrum owing to the paramagnetism of V4+ ( d ' ) cations.Pure a l l - , p-, y- and 6-VOPO, phases were observed to give a single sharp "P peak at -19.5, -11.8, -18.8 and +1.6ppm, respectively, with respect to H,PO,. For the actual catalysts, analysed after catalytic reaction, the same spectra were observed in agreement with XRD data. However, in addition, or sometimes in place of, the peak typical of a given VOPO, phase a new narrow band appeared that was shifted towards lower frequency with respect to the pure VOPO, phases. This is presumably due to the interaction between VOPO, groups and (VO)2P207, owing to the paramagnetism of the latter phase. As this interaction depends on the distance between both phases, it corresponds to a short-range interaction.Poor catalysts exhibit the VOPO, phase peak unshifted with respect to the pure phase and analogous in nature to that evidenced by XRD analysis. On the other hand, the best catalysts exhibited a shifted "P narrow peak (-20 to -40 ppm) and the pure phase peak was no longer observed. The structure of the VOPO, groups could not be determined unambiguously from P peak position because of the paramagnetic shift which depends on the distance from the paramagnetic species. However, they could be identified in many cases by XRD analysis. The best catalyst corresponds to a structure in which a l l and/or S groups are present (see table 2).Bordes" has described the complicated structures of the different vanadium phos- phate phases. She differentiated two groups of phosphates designated 1 and 2 (fig. 1). In group 1 the structural unit is composed of a single VO, octahedron with each equatorial oxygen sharing a corner of one phosphate PO, anion of VOHP0,-2H20. a'-, all- and p-VOPO, belong to this group. In group 2 one has a pair of edges sharing octahedra, with each equatorial oxygen sharing a corner of one phosphate anion. (VO)2P207, VOHPO4-0.5H20 and the 6- and y-VOPO, phases belong to this group. Any transforma- tion within a group corresponds to a topotactic reaction and therefore to a small activation energy since a limited number of bonds must be broken.210 Surface Atomic Rearrangements Fig.1. Classification of the structure of vanadium phosphates groups 1 ( a ) and 2( b ) from ref. ( 12). LVOP04*2H20 VOHPO4*4H2O fi-VOPO, l-VOH PO4.O.5 H,O La,-VOPO, VO( H, PO,) ( face-sharing) p -vo POJ a,,-VOP04 VO(PO?), ' y-VOPO, ' ( vo ) 2 P207 The conclusion that we draw from our study is that the actual catalyst corresponds to (VO)2P,07 in the bulk, but at the surface there are some localized VOPO, entities. This may correspond to crystallographic shear (cs) planes, as suggested by Bordes," or to the presence of disorder in the crystalline structure, as proposed by Centi and Trifiro.', Our X-ray radial electronic distribution and "P MASNMR data give new insight to help elucidate how the catalyst works. We are of the opinion that local desorder created during the preparation and for activation of the catalyst allows one to have local sites particularly efficient for butane oxidation, namely the redox V"'-V5+ couple, in a specific atomic arrangement which facilitates the butane to butene transformation suggestedgh"' to be the first and rate-determining step of the reaction.This conclusion is in agreement with the data of Contractor et al.,', who studied butane oxidation in a riser-type reactor. They observed that the partial oxidized catalyst gives high selectivity for maleic anhydride in the absence of air. The reduced catalyst is reoxidized in the separate regenerator zone to its optimum state with an average vanadium valence of ca. +4.l. 3.2 Iron Phosphates Iron-phosphate-based catalysts have been claimed to exhibit interesting catalytic proper- ties for the oxidative dehydrogenation of isobutyric acid to methacrylic acid." Within the Fe203-FeO-P,0, phase diagram it is possible to find some well defined phases bearing either iron(rr1) ions or iron(1r) ions or both.We have chosen three phases with an Fe/ P ratio close to unity, namely FePO,, Fe2P,07 and Fe,( PO,), . These three samples were studied for the isobutyric acid oxidative dehydrogenation reaction. The main results are summarized in table 3. The three catalyst samples were studied by X-ray diffraction and Mossbauer tech- niques before reaction and 48 h into catalytic reaction. It was observed that the starting materials exhibit XRD patterns and Mossbauer spectra similar to those given in the literature for pure phases.After catalytic reaction, the same techniques allowed us to observe the presence of Fe,( PO,), phase for both Fe,( PO,), and FePO, starting materials. However, it was necessary to introduce an extra iron(ri) doublet into the simulated Mossbauer spectra to fit the experimental spectra, in addition to the two iron( 1 1 ) sites corresponding to Fe,( PO,), . This third iron( I I ) site has the Mossbauer parameters 6 = 1.19, w = 0.26 and A = 2.73 mm s I and cannot be attributed to any known phase. Its intensity corresponds to only 4% of the total intensity. The same analysis has beenJ. C. Vedrine, J. M. M. Millet and J. C. Volta 21 1 Table 3. Main features of iron phosphate catalysts used in isobutyric acid oxidative dehydrogenation at 380°C (653 K ) selectivity (%) rate of formation samples coz propene acetone MAA" /10 mol-' m-' ~~ ~~ A FePO, 1 8 25 66 B Fe2P207 3 14 16 67 c FeAPO,), 9 7 52 32 107 70 184 " Methacrylic acid.carried out for Fe,P,O, after catalytic reaction. There was an iron(i1) site with the following Mossbauer parameters: S = 0.49, u' = 0.41 and A = 0.48 mm s-' which cannot be assigned to any known phase. Its intensity represents 5% of the total intensity. For Fe,( no detectable change with the starting material was observed in XRD and Mossbauer analyses. In another experiment Fe2P207 was progressively oxidized at 723 K (450 "C) for 5 h under various oxygen pressures. A detailed Mossbauer study" has shown that besides the Fe2P2O7 spectrum, two iron sites appear, one iron(ii1) and one iron(ii) in a ratio Fe3+: Fe2+ of 2.2: 1.6.Their Mossbauer parameters were very close to the new sites observed for Fe'+ in Fe,P,O, and for Fe" in FePO, after catalytic reaction (uide supra). We tentatively suggest that both sites existed in both samples after catalytic reaction but were not observed because of overlap with the main Fe2+ or Fe3' sites, respectively, owing to their low concentration. We further suggest that they correspond to a still unknown phase with the following Mossbauer parameters: Fe": Fez+: 6, = 0.47 f 0.05 and A , = 0.68 f 0.02 mm s-'; 62 = 1.20f 0.05 and A, = 2.73 * 0.02 mm s-' with Fe3+ : Fe" = 1.9 f 0.3. This phase corresponds to Fe6P6OZ3. A detailed XRD analy- sis of the powder spectrum is in progress to try to determine the structure of this new phase.It is worth noting at this stage that the best catalytic performances (see table 3) were obtained when this new phase was present in small amounts. Moreover this phase was also observed in the industrial catalyst. It is therefore tempting to suggest that its presence is important for high catalytic performance because it allows an easy and reversible transformation A(>, 3Fe2P2O7 T--+ Fe6P6013 -0: where two P207 anions are transformed into four PO, anions. This obviously holds true because of the close similarity in their crystallographic structures. One may then consider the catalyst surface as a kind of living, even breathing, structure oscillating from one form to the other by inserting one oxygen atom per two P atoms and vice versa.Even if vanadium and iron phosphates are rather different materials we are of the opinion that they behave in similar fashion in catalytic reaction conditions with a more or less easy oscillation between two topotactic structural states with local short-range or long-range order. 4. Conclusion This study of vanadium and iron phosphates analysed before and after catalytic reaction clearly shows that under catalytic reaction conditions short- or long-range order in212 Surface Atomic Rearrangements atomic structural arrangements oscillates from a reduced state to an oxidized state. This corresponds to the reversible transformation of two PO, ions into one P207 ion with elution of oxygen. When such a transformation is easy, for example in the case of a topotactic reaction, the catalyst exhibits high catalytic performance.If it is not easy, lower catalytic performance will be obtained. For instance, the topotactic transformation of p-(VO)2P207 into y- or 6-VOPO, within the same classification group” should result in a better catalyst than the transformation from group 1 to group 2, such as p-(VO),P,07 into p-VOP04. Such a transformation is greatly dependent on the starting material and activation and reaction conditions. The transformations may occur locally in a short-range order as in a crystallographic shear plane or at extended zone defects or in long-range order. The former state can be observed by X-ray radial electronic distribution and MASNMR studies, while the latter one may only be detected by X-ray diffraction.These transformations may be visualized as a breathing motion. If oxidation or reduction is too strong it may correspond to irreversible transformation, thus rendering the catalyst less efficient. The state of the art corresponds to finding the best preparation and activation procedures which would result in a catalyst able to breathe easily as described above under catalytic reaction conditions. The addition of dopants or of another phase such as an alkali-metal pyrophosphate in the actual industrial catalysts, lX which as we have shown contains the Fe,P,O,, phase, is most probably to favour short-range disorder and therefore catalytic properties. This disorder arises from the close structural fit between the additional phase and the starting pyrophosphate phase.As a more general conclusion it may be emphasized that phosphate-type catalysts behave in a specific way with respect to other catalytic systems such as bismuth molybdates or tin-antimony oxides (solid solution) because of this specific structural oscillation such as to, 3Fe2P207 t Fe,P,023 -0, or - 1 / 2 0 , + 1 / 2 0 : 2 y - or 6-VOPO, r--+ p-(VO)2P207 corresponding to topotactic transformations. In other words, not only are the stereochemistry of the reactant molecule and the atomic structure at the surface important parameters, but also the redox properties of the surface and local topotactic structural transformation under catalytic reaction conditions. References 1 J. C. Volta, W. Desquesnes, B. Moraweck and G. Coudurier, React.Kinet. Catal. Lett., 1979, 12, 241; 2 J. C. Volta and J. L. Portefaix, Appl. Catal., 1985, 18, 1; J. C. Vedrine, G. Coudurier, M. Forissier and 3 J. M. Tatibouet and J. E. Germain, J. Catal., 1981, 72, 365. 4 J. M. Tatibouet, J. E. Germain and J. C . Volta, J. Catal., 1983, 83, 24. 5 M. David, Thesis (Lyon 1986). 6 T. C. Yang, K. K. Rao and I . Der Huang. US Patent, 4 392 986 (1987), assigned to Exxon Corp. 7 J. T. Moggins, J. S. Swinnea and R. Steinfink, .I. Solid. State Chem., 1983, 47. 279. 8 A. Modaressi, Thesis (Nancy, 1982). 9 ( a ) T. Shimoda, T. Okuhara and M . Misono, J. C‘hem. Soc. Jpn, 1985, 58, 2163; ( h ) K. Miyamoto, T. Nitadori, N. Mizuno, T. Okuhara and M. Misono, Chem. Lett., Cheni. Soc. Jpn, 1988, 303. 10 B. K. Hodnett, Card Rev. Sci. Eng., 1985,27,373, and references therein; B. K. Hodnett and B. Delmon, Appl. Catal., 1985, 15, 141. 11 P. Courtine, in Solid State Chemistry in Catalysis, ed. R. K . Grasselli and J. F. Brazdil, AC’S Sjwp. Ser., 1985, 279, 37. J. C. Volta and B. Moraweck, J. Chem. Soc., Chem. Comtnun., 1980, 338. J. C. Volta, Muter. Chem. Phys., 1985, 13, 365; Catal. Today, 1987, 1, 261.J. C. Vedrine, J. M. M. Millet and J. C. Volta 213 12 E. Bordes, Catal. Todaj7, 1987, 1, 499; 1988, 3, 163; E. Bordes, Thesis (Compiegne, 1979). 13 G. Busca, G. Genti and F. Trifirb, Appl. Catal., 1986, 28, 265. 14 R. M. Contractor, H. E. Bergna, H. S. Horowitz, C. M. Blackstone, B. Malone, C. C. Torardi, B. Griffiths, U. Chowdhury and A. W. Sleight, Catal. Today, 1987, 1,49; R. M. Contractor and A. W. Sleight, Card Today, 1988, 3, 175. 15 G. Bergeret, M. David, J. P. Broyer, J. C. Volta and G. Hecquet, Catal. Today, 1987, I , 37. 16 M. David, F. Lefebvre and J. C. Volta, in Proc. XI Simp. iberoamer. Catal., Guanajuato, June 1988, ed. F. Cossio, G. delAngel, 0. Bermudez and R. Gomez (IMP, Mexico, 1988), p. 365. 17 G. Centi and F. Trifiro, Catal. Today, 1988, 3, 151. 18 E. S. Cavaterra, G. C. Petrini, L. Moreschini, L. B. Dalloro and P. S. Pomodoro, US Patenf, 3 948 959 (1976), assigned to Montedison. 19 J. M. Millet, C. Virely, M. Forissier, P. Bussiere and J. C. Vedrine, Hyperjne fnteracrions, 1988, in press. Paper 9/00122K; Received 4th January, 1989

ISSN:0301-7249    

DOI:10.1039/DC9898700207

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

Faraday Discuss. Chem. SOC., 1989, 87, 215-225 Nature of Active Species of (VO),P,O, for Selective Oxidation of n-Butane to Maleic Anhydride Gabriele Centi and Ferruccio Trifiro* Dipartimento di Chimica Industriale e dei Materiali, V. le Risorgimento 4, 401 36 Bologna, Italy Guido Busca Istituto Chimico, Facolta Irigegneria, Fiera del Mare, Pad. D, Genova, Italy Jerry Ebner Monsanto Chemical Company, 800 N. Lindbergh, St. Louis, Missouri 631 67, U.S.A. John Gleaves Department of Chemical Engineering, Washington University, Campus Box 11 98, St. Louis, Missouri 63130, U.S.A. TEM, EXAFS, FTIR, temporal analysis of products (TAP), stopped-flow desorption (SFD) and catalytic measurements of (VO)2P207 are reported. The reduced interaction between (020) planes of (VO)7P207 in samples prepared in an organic medium induces a charge localization on the V atoms of the coupled trans-vanadyl present in this plane, enhancing their catalytic reactivity in butane oxidation.Contiguous surface Brdnsted sites (P-OH) also participate in the mechanism of selective oxidation. C-containing residues are present in relevant amount on the surface during catalytic experiments and give rise to a specific fouling of the active sites, but their possible role as co-catalysts in the transfer mechanisms of single activated species is also discussed. The vanadium-phosphorus oxide catalysts have an unmatched ability to perform with high selectivity the complex transformation of n-butane to maleic anhydride ( 14-electron oxidation, abstraction of 8 hydrogen atoms and insertion of 3 oxygen atoms in the molecule).This reaction also represents the only example of a commercial process for the selective functionalization of an alkane.' The literature contains numerous studies on aspects of the mechanism of this reaction.'72 The catalytic activity/selectivity has been shown to be optimal in long-run catalytic tests, when the bulk phase is vanadyl pyr~phosphate.'-~ In this text the term equilibrated samples274 indicates samples which have been held for a long time (at least 700 h ) in the reaction environment. This conditioning has been shown to be fundamental' for the characterization of the active phase, because it avoids misunderstandings, some- times found in the literature, connected to the characterization of transient situations in non-equilibrated samples.The catalytic performances of vanadyl pyrophosphate are closely related to the method of preparation employed. For example, changing the solvent system from aqueous to organic results in vanadyl pyrophosphate catalysts with higher surface areas and altered crystallite morphologies.6-'' The changes in catalytic performance resulting from these morphological transformations are indicative of the structure sensitivity of the reaction.9 This paper further explores the structural and surface characteristics of equilibrated VPO catalysts and the importance of structure sensitivity in butane oxidation. 215216 Nature of Active Species of (VO)2P207 Experiment a1 Catalyst Preparation The catalysts used in this study were prepared in aqueous or organic media according to well established literature procedure^^.^.'^ and will be hereinafter called VP-aq or VP-org, respectively.The aqueous preparation involves synthesis of the blue [ VOHPO4I2- H20 precursor by reduction of vanadium( IV) with aqueous HC1 and precipitation by addition of 85% H3P04 followed by filtration and drying. In the organic preparation the V2OS is refluxed in isobutyl alcohol saturated with HC1 to reduce and dissolve the vanadium, and the blue [VOHPO4I2-H20 precursor complex is precipitated by addition of 99% H3P04, refluxing and distilling to remove solvent. The structure of the aqueous precursor compound [VOHPO4I2-H20 has been established' ',12 and it has been utilized as a model compound for EXAFS studies. The model compound VO(H2P04), was synthesized using the aqueous method with a starting P/V ratio of 2.2/1.The structure was verified on the basis of chemical analysis and X-ray diffraction (XRD) data (ASTM 21-1436).13 The [VOHPO4I2-H20 precursors from the aqueous and organic preparations were formed into active catalyst by calcination in air at 400 "C for 1-2 h. The final catalyst was formed in situ by running the butane oxidation reaction at 1.5% butane, air and 1000 g.h.s.v. The equilibrated catalyst samples from both preparations had ca. 750 h continuous on-stream time in the butane-oxidation reaction. Analysis of the samples showed the average vanadium oxidation state to be 4.01 kO.01. Characterization of the Catalysts EXAFS14 data were collected using an Elliot GX-21 15 kW rotating anode source and a Ge-511 monochromator crystal.All the spectra were processed using standard XAS analytical methods." EXAFS Fourier-transforming and modelling techniques were used to assess the short-range (0-4 8,) structure around the vanadium atoms. Reference photoelectron-scattering amplitudes and phase shifts that were needed in the modelling were determined from standard compounds with well known structures or from theoreti- cal calculation^.^^'^' High-resolution transmission electron microscope's (TEM) images were obtained using a 300 keV Philips 430ST TEM. This instrument has a spherical aberration limit on continuous contrast transfer ( i e . a maximum point resolution) at CQ. 1.9 8, and a damping limit just below that, as inferred from diff ractograms of poorly crystallized light-element ( e.g.carbonaceous) material. Periodicities in TEM images were measured by digitizing the TEM negatives and determining the position of local maxima in fast Fourier transform (FFT) power spectra. The structure model of (VO)2P207 was obtained by transforming the asymmetric units of Gorbunova and Linde"" from the listed P62,a format into the P6c2, format of Bordes,lg6 and then replicating them fourfold to fill the cell. The model was tested by calculating structure factors for electron scattering and looking for the expected Pbc2, extinctions. Temporal analysis of products (TAP), stopped-flow desorption (STD), Fourier- transform infrared (FTIR), X-ray diffraction (XRD) and catalytic measurements were performed as reported e l ~ e w h e r e .~ - ~ , ~ Results and Discussion Nature of the Microstructure of the Active Phase XRD analyses of samples prepared by both the aqueous and organic methods were consistent with the presence of a single crystalline phase, ( V0)2P20, .6*7.'1s19 However, the [VOHPO4I2-H20 precursor from the aqueous preparation leads to well shapedG. Cenri et al. 217 crystals of (VO)2P207, in contrast to the organic preparation which leads to lamellar crystals with rose-like or similar patterns of platelets.' The XRD data show a decrease in the sharpness and intensity of the reflection at 3.87 A (020) in going from PV-aq to PV-org samples. Indeed, the related (010) reflection of the precursor of PV-org has a much lower intensity than expected from the calculated powder pattern.The effect is due to the presence of alcohol that reduces interlayer forces between the (010) sheets of [VOHPO4I3-H20, inducing their preferential exposition on the surface. The transfor- mation from the precursor phase to the (V0)2P207 occurs via a topotactic mechan- ism,6-831 ' which preserves the morphological appearance of the parent precursor in the final activated sample. It has been suggestedE910*2" that structural defects form that are associated with the apparent disorder of the stacking of the (020) planes in (VO)2P207 prepared in an organic medium. It also should be considered that in the structure of (vo)2P20719 the resolution factor ( R = 0.081) is indicative of problems with the structural refinement, and the X-ray powder diffraction pattern of the catalyst does not match in detail that of the single-crystal data.The nature of the microstructure of the VP-org sample was therefore analysed using TEM. Plate 1 is the 1 5 0 0 0 0 ~ magnification TEM image of the orgmic-based, equilibrated PVO catalyst. Note that two distinct morphological type's, piatelets and rods, are present in the sample. The Fourier-transform power spectrurx c;f ctr2e of the platelet crystals shown in plate 1 shows ca. 40 periodicities, with a continuous contrast transfer cloud superimposed over most of them. Spacings were checked against several known PVO structures, and are consistent with previous TEM The best match was the (020) orientation of the (VO)2P207 structure. A 7 x 7 periodicity mask was constructed in Fourier space to remove most of the contrast except for that due to lattice periodicities.Plate 2 shows a direct comparison between the indexed (VO)2P207 structure and the periodicity of the filtered image. The positions of the regions of relative brightness and darkness are supportive of the structural model for ( VO)2P,07, with the bright regions corresponding to regions of low electron column density (tunnels), and vice versa. In this case, all four large, eight medium and four small tunnels through each unit cell appear where they should. Examination of the cross-fringe images of the rod-shaped structures suggests this crystal is also (VO)2P207. The rod axis appears to be along the (020) layering direction. These results provide further evidence that the active catalyst is only (VO)2P207 and the spatial arrangement of atoms is in full agreement with the G ~ r b u n o v a ~ ~ ~ structure.Local Structure at the Surface The local bonding around the vanadium centres in relation to preferential exposition of the (020) plane was examined using the EXAFS technique, above the vanadium X-ray adsorption threshold. The structurally characterized compounds [VOHPO4I2- H2012721 and VO(H2P04)2'3 were used as standards. The Fourier transforms of the experimental EXAFS results for these two compounds are shown in fig. 1. All peak positions are shifted to shortened radii by 0.35 A owing to the presence of back-scattering phase shifts. When theoretical back-scattering phase shifts and amplitudes are included in calculating model spectra, the positional agreement between the calculated structure and the experimental data is quite good.The short V=O bond present in the two structures at 1.57 and 1.60 A, respectively, is clearly seen in both spectra and the peak associated with the four longer equatorial V-0 bonds (ca. 0.4 8, > V=O) is the most intense. The Fourier transforms of the experimental EXAFS results for the PV-aq and PV-org catalysts are shown in fig. 2 ( a ) and ( b ) , respectively, with the corresponding XRD patterns for these catalysts. PV-aq is well ordered in the layered direction, as reflected by the sharpness and intensity of the (020) XRD reflection. PV-aq does not show the peak characteristic of the V=O bond, which is present in the PV-org catalyst.218 Nature of Active Species of (V0)2P20, 0' 61 , 1 I 1 1 : 0 2 .- c 2 E 4: I 8 6 4 2 0 0 2 4 6 8 I I 1 0 2 4 6 8 R I A Fig.1. Fourier transforms EXAFS spectra of ( u ) [VOHPO4I2-H20 and ( b ) VO(H2P04)? model compounds with the respective simulated spectra. The EXAFS results suggest that the charge on vanadium is delocalized through axial O-V.-.O-V interactions in PV-aq and thus there is no discrete V=O bond. The long axial V-0 bond distances in the Gorbunova'ga structure are reflected in the aqueous- based catalyst. In contrast, the (VO),P,O, catalyst derived from the organic preparation does show the discrete V=O bond distance and the XRD data indicate poorer ordering in the layered stacking direction as compared with the aqueous-based catalyst. The EXAFS data thus suggest that going from the three-dimensional structure of (V0)2P207 in PV-aq to the two-dimensional type present in PV-org results in a localization of the charge on the surface vanadium as a consequence of the reduced interaction between the (020) planes. The surface structure of the (020) plane is characterized by the presence of paired vanadium pseudo-octahedra orientated trans to one another.' The pyrophosphate sublattice connects each of the equatorial oxygen atoms.The greater electronegativity of P polarizes the V-0-P bond and generates positive charge on vanadium. The result is a medium-strength Lewis-acid site, as shown by FTIR spectra of ammonia, pyridine and acetonitrile adsorbed on it.'2723 A comparison of the FTIR spectra of acetonitrile adsorbed on the PV-aq and PV-org samples shows that in the latter the localization of the charge owing to the reduced interaction between (020) sheets induces an increaseG.Centi et al. 219 Q Y [OZOJ- h 4 W t? ? > a a (d U CT h W ? k CI n N h 5 h N CI v Y n 4 h d v h d W U v) E Y U a220 Nature of Active Species of (V0)ZP20, Table 1. Specific rate per square metre of surface area of maleic anhydride formation from n-butane catalyst T / K r (maleic anhydride)/ lo9 mol s-' m-' PV-aq 660 PV-aq 690 PV-org 570 PV-org 600 0.29 1.02 1.58 2.58 Reagent composition: 1.6% n-butane, 10.2% 02, 88.2% N2. 0" .... ... @ p 0 0 O H Fig. 3. Surface crystal structure of the (020) plane of (V0)2P20, in the Lewis-acid strength. The two bands originating from the Fermi resonance between the v(CN) and the combination of the 6(CH,)+v(CC) are shifted from 2282 and 2300 cm-' (in PV-aq) to 2300 and 2328 cm-' (in PV-org), indicating an increase from medium-strong to very-strong Lewis acidity.FTIR conclusions thus parallel the EXAFS data, showing a localization of the charge around V in the PV-org sample. This effect has a considerable influence on the catalytic behaviour in n-butane selective oxidation. Table 1 shows that the specific rate of maleic anhydride formation per square metre of surface area is greatly enhanced in the PV-org sample with respect to the PV-aq sample. The charge localization on the V atoms thus modifies their reactivity in the activation of butane, the rate-determining step of the reaction as shown by kinetic measurement~.~~ FTIR measurementsZ2 indicate the presence of Brgnsted acidity on the vanadyl pyrophosphate, characterized by a v(0H) band at 3660cm-* which is rather stable to evacuation, and a shoulder near 3600 cm-' (unstable) and assigned to P-OH and P-(OH)2 groups, respectively. Pyridine adsorbed on PV-org gives rise to the bands of the pyridinium cation, as well as those of pyridine chemisorbed on Lewis-acid sites.The stability of these bands indicates the strong acid nature of the Brgnsted sites on (V0)2P207. No large differences were found between the PV-org and PV-aq samples. These surface Brgnsted-acid sites derive from the topotactic mechanism of condensation of two hydrogen phosphate groups in [VOHPO4I2-H20 to form (V0)2P207. Cu. one half of the hydrogen phosphate groups should remain as such on the surface (fig.3).G. Centi et al. 22 1 1700 1550 1400 ti/cm-' Fig. 4. FTIR spectra of pyridine adsorption on ( a ) PV-org and ( b ) K-PV-org (0.98 wt '/o K) evacuated at 450 K. Table 2. Effect of doping with K of (V0)2P20, (PV-org) on the selectivity to maleic anhydride (MA) and carbon oxides (CO,) in the butane oxidation (80% con- version) at 608 K selectivity ( O h ) K (wt Yo) MA C O Y 0.0 68 32 0.00041 38 62 0.1 1 29 71 0.39 11 89 0.98 5 95 In order to verify the possible role of the Bronsted-acid sites in the catalytic behaviour of vanadyl pyrophosphate, doping experiments with K were performed. K-doping was performed in a water-free medium in order to avoid the detrimental contamination of the surface with water and to produce a highly selective doping of the Bronsted sites.Fig. 4 shows the FTIR spectra of pyridine adsorbed on PV-org and on PV-org doped with small amounts of potassium (0.98'/0 wt) (K-PV-org). The selective inhibition of Bronsted acidity (bands at 1640 and 1542 cm-') by K-doping is clearly evident, whereas the Lewis acidity [ v(8a) vibration at 1610 cm-'1 is not affected. The additional bands at 1598 and 1440 cm-' in K-PV-org are due to the interaction of pyridine with K' ions. Table 2 shows that doping with K leads to a selective inhibition of the formation of222 Nature of Active Species of (VO),P,07 maleic anhydride even for a very small amount of potassium added. These results thus show a specific role of P-OH groups in the mechanism of selective synthesis of maleic anhydride from n-butane and in the architecture of the active site.TEM and EXAFS results indicate that the microdomains of Vv phosphate indicated by some authors' as the active sites for butane selective oxidation are absent in the equilibrated catalyst. A further confirmation of this evidence was obtained using SO,, which inhibits the reactivity of surface Vv sites by the formation of a stable VIv s ~ l p h a t e . , ~ On non-equilibrated PVO samples co-feeding SO, increases the selectivity to maleic anhydride, especially at high conversion, whereas this effect is not present in equilibrated catalysts.25 This indicates (i) that the presence of surface Vv sites is detrimental because they are involved mainly in the consecutive oxidation of maleic anhydride and (ii) that the Vv sites are absent or present in very limited numbers on the surface of equilibrated PVO catalysts.In agreement, the rate of V" oxidation to Vv of the catalyst (air at 400 "C) decreases in going from non-equilibrated to equilibrated samples from ca. 2.6 x mol h-* of Vv formed in fresh non-equilibrated catalysts to ca. 8 x mol h-' of Vv in equilibrated samples [ 1 g (VO),P,07 contains ca. 6.5 x lo-' moll. In-situ Surface Modifications and Dynamics of the Surface of (VO)2P20, The surface nature and the reactivity of the (VO),P,O, during catalytic experiments is dramatically different from that of the clean surface. We have investigated these aspects using two novel systems of analysis, temporal analysis of products and stopped-flow desorption ( SFD),27*28 in combination with FTI R and reactivity measure- ments.The TAP reactor is a new device for the study of reaction dynamics of solid- catalysed vapour-phase reactions, based on a real-time mass quadrupole analysis of the products formed when transient pulses of reagents using high-speed valves are introduced in a microreactor. When the formation of CO, by introducing an O,/butane mixture over an equilibrated PV-org sample is monitored in the TAP reactor, it appears that two processes occur together in their formation: a fast process and a slower process.' Using 1802 and monitoring the oxygen isotope distribution in the carbon dioxide formed as a function of time4 it is shown that the fast process is C018018 formation using chemisorbed oxygen-18. Double-pulse feeding the hydrocarbon and after a calibrated time the oxygen (in the range 0.1-1000ms), suggest that when insufficient oxygen is channelled to the adsorbed intermediates forming maleic an- hydride, other C-containing products are formed.These remain strongly held on the surface. The release of carbon dioxide by the slower process is mainly attributable to the combustion of these surface C-containing residues (hereinafter called SCR). TAP multiresponse experiments (fig. 5 ) show the production of CO, from the reaction of butane and oxygen with a clean PV-org sample. During the first half of the experiment the catalyst is exposed to a number of equally intense butane pulses. After 5.15 s the butane is turned off and the catalyst is fed with a similar number of equally intense oxygen pulses.During the first part of the experiment, carbon dioxide forms at the expense of lattice oxygen or of long-lived adsorbed oxygen species. The formation of CO, passes through a maximum as a consequence of the increasing amount of long-lived SCR (formed in the interaction of butane with the vanadyl pyrophosphate surface) and of the consumption of available surface oxygen. When the butane is turned off and the catalyst is fed with pure oxygen CO, production begins again at the expense of the SCR. Specific tests were performed in order to titrate the amount of SCR on an equilibrated PV-org sample. The amount of residue can be determined by titrating the surface with oxygen and monitoring the quantity of COX. In parallel experiments, the change in the rate of butane depletion as a function of the amount of SCR removed from the surface was monitored.Three indications derive from these tests: ( i ) in steady-state conditions, only ca. 5-10% of the active sites are not hindered by SCR (theG. Centi et al. 0 2 - butane 223 h 0 0, h C K * .e v) * .- v) (d E 10 8 6 4 2 0 0 2000 4 000 6 000 8000 10000 tlms Fig. 5. TAP multipulse experiments with PV-org. rate constant of butane depletion at 300 "C decreases from ca. 30 x to ca. 2 x dm' s-' g-'), (ii) only ca. one third of the surface is covered by SCR, (iii) there is a fairly good linear relationship between the amount of SCR and the decrease in activity. This indicates that these residues are not present on the surface in three-dimensional patches or filaments and indicates a selective deactivation of the active centres.Stopped- flow desorption (SFD) experiments" give interesting further information about the possible role of these strongly held residues. This technique consists of the analysis of the products that are desorbed in a non-reactive flow from the surface of a catalyst subjected to switching from continuous-flow conditions to those for desorption. The products desorbed from the PV-org catalyst after furan oxidation, together with maleic anhydride, indicate the presence on the surface of relevant amount of crotonaldehyde, a product of self-hydrogenation of furan. The hydrogen atoms extracted from furan in the synthesis of maleic anhydride act as hydrogenation sites of a second adsorbed furan molecule even in the presence of gaseous O2 and of an oxidation catalyst.TAP experiments with furan' show the tendency of furan to be strongly adsorbed on the catalyst, forming long-lived C-containing residues. FTIR spectra28 of furan and other C4 compounds adsorbed on (VO)2P207 are in agreement with this observation. In conclusion, it seems possible to propose that strongly held surface residues may play the role of co-catalysts acting as hydrogen-transfer agents. Other possible roles of these surface residues may be suggested. Preliminary experi- m e n t ~ ~ ' ~ ~ indicate a significant increase in the selectivity to maleic anhydride from butane when these residues are present. Two possible other co-catalysis effects may be envisaged: (i) electron-transfer agents and (ii) activated oxygen-transfer agents.The former reaction is critical in multi-electron reactions like butane oxidation. The forma- tion of localized electrons able to activate oxygen near the adsorbed intermediate forming electrophilic oxygen radical ions may give rise to a non-selective attack of the hydrocar- bon. The activation of oxygen far from the adsorbed intermediate with a shuttling towards the hydrocarbon along selective routes can give a specific point of attack to the adsorbed organic molecule. Various observation^^'^ suggest desorption, hopping or224 Nature of Active Species of (VO)2P207 walking on the surface of (VO)2P207 of the organic intermediates to be unlikely. Furthermore, TAP experiments have evidenced considerable surface of acti- vated oxygen species on the surface of (VO)2P207 and their possible role in the mechanism of oxygen insertion, whereas different sites (lattice oxygen) are involved in the hydrogen-abstraction properties.Therefore, if sufficient activated oxygen is not channelled into the adsorbed intermediate, this can evolve towards other non-selective products. The transformation of butane to maleic anhydride is a delicate balance between the rates of surface transfer of the single species (e, H, 0) and transformation of the intermediate. It seems that a significant co-catalysis role, mainly in the transfer mechan- isms, can be attributed to the C-containing residues present on the surface of (VO)2P207. Experiments are in progress to clarify these preliminary suggestions.Conclusions The architecture of the active centres of (VO)2P207 for selective oxidation of butane comprises different aspects. The reduced interaction between the (020) planes of (V0)2P207 in samples prepared in an organic medium induces a charge localization on the V atoms of the coupled trans-vanadyl present in this plane, enhancing their catalytic reactivity in butane oxidation. Contiguous surface Bronsted sites (P-OH) also partici- pate in the mechanism of selective oxidation. C-containing residues form in relevant amounts on the surface during catalytic experiments and give rise to a specific fouling of the active sites, but their possible role as co-catalysts in the transfer mechanisms of single activated species is also discussed. This suggests an ensemble behaviour of the active surface in the mechanism of butane transformation to maleic anhydride that further stresses the importance of the characterization of equilibrated catalysts which have different properties in comparison with non-equilibrated catalysts.References 1 Selective Catalytic Oxidation of C, Hydrocarbons to Maleic Anhydride, ed. B. K. Hodnett, Catal. Today, 1987, 1. 2 J. R. Ebner, V. Franchetti, G. Centi and F. Trifiro, Chem. Rev., 1988, 88, 5 5 . 3 J. R. Ebner and J. T. Cleaves, in Oxygen Complexes and Oxygen Activation by Transition Metals, ed. A. E. Martell and D. T. Sawyer (Plenum Press, New York, 1988), p. 273. 4 G. Centi, F. Trifiro, G. Busca, J. R. Ebner and J. T. Cleaves, in Proc. 9th Int. Congr. Catal., ed. M. J. Philips and M.Ternan (The Chemical Institute of Canada, Ottawa, 1988), p. 1538. 5 M. A. Pepera, J. L. Callahan, M. J. Desmond, E. C . Millberger, P. R. Blum and M. J. Bremer, J. Am. Chem. SOC., 1985, 107, 4883. 6 G. Busca, F. Cavani, G. Centi and F. Trifiro, J. Catal., 1986, 90, 400. 7 E. Bordes, in Petroleum Division Reprints of the Symposium Hydrocarbon Oxidation, 194th ACS Meeting, 8 H. S. Horowitz, C. M. Blackstone, A. W. Sleight and G. Tenfer, Appl. Catal., 1988, 38, 193. 9 R. A. Schneider, U.S. Patent 4043 943 (1977). New Orleans, 1987, p. 792. 10 F. Cavani, G. Centi and F. Trifiro, J. Chem. SOC., Chem. Commun., 1985, 492. 1 1 J. W. Johnson, D. C. Johnston, A. J. Jacobson and J. F. Brody, J. Am. Chem. SOC., 1984, 106, 8123. 12 C. C. Torardi and J. C. Calabrese, Inorg. Chem., 1984, 23, 1310. 13 S. A. Linde, Y. E. Gorbunova, A. V. Lavrov and V. G. Kuznetsov, Dokl. Akad. Nauk SSSR, 1979,224, 14 B. R. Stults and E. C. Marques, Monsanto Internal Report, 1983. 15 S. P. Cramer, K. D. Hodgson, E. I. Stiefel and W. E. Newton, J. Am. Chem. SOC., 1978, 100, 2748. 16 B. K. Teo, P. A. Lee, A. L. Simons, P. Eisenberger and B. M. Kincaid, J. Am. Chem. SOC., 1977,99,3854. 17 P. A. Lee, B. K. Teo and A. L. Simons, J. Am. Chem. SOC., 1977, 99, 3856. 18 P. Fraundorf and J. R. Ebner, Monsanto Internal Report, 1987. 19 ( a ) Y. E. Gorbunova and S. A. Linde, Sou. Phys. Dokl., 1979, 24, 138; ( b ) E. Bordes, P. Courtine and J. W. Johnson, J. Solid State Chem., 1984, 55, 270. 20 D. C. Johnston and J. W. Johnson, J. Chem. SOC., Chem. Commun., 1985, 1720. 21 M. E. Leonowicz, J. W. Johnson, J. F. Brody, H. F. Shannon and J. M. Newsam, J. Solid State Chem., 1985, 56, 370. 141 1 .G. Centi et al. 225 22 G. Busca, G . Centi, F. Trifiro and V. Lorenzelli, J. Phys. Chem., 1986, 90, 1337. 23 G. Busca, G. Centi and F. Trifirb, J. Am. Chem. SOC., 1985, 107, 7757. 24 G. Centi, G. Fornasari and F. Trifiro, Ind. Eng. Chem. Prod. Rex Dev., 1985, 24, 32. 25 G. Centi, G. Golinelli and F. Trifiro, Appl. Caral., in press. 26 J. T. Gleaves, J. R. Ebner and T. C. Keuchler, Catal. Rev. Sci. Eng., 1988, 30, 49. 27 G. Busca, G . Centi and F. Trifirb, in Catalyst Deactivation, ed. B. Delmon and G . F. Froment (Elsevier, 28 G. Busca and G. Centi, J. Am. Chem. Soc., 1989, 111, 46. Amsterdam, 1987), p. 427. Paper 8/05047C; Received 20th December, 1988

ISSN:0301-7249    

DOI:10.1039/DC9898700215

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

Faraday Discuss. Chem. SOC., 1989, 87, 227-237 Carbon Monoxide Hydrogenation Selectivity of Catalysts derived from Ruthenium Clusters on Acidic Pillared Clay and Basic Layered Double-hydroxide Supports Thomas J. Pinnavaia,* M. Rameswaran, Emmanuel D. Dimotakis, Emmanuel P. Giannelis and Edward G. Rightor Department of Chemistry and Center for Fundamental Materials Research, Michigan State University, East Lansing, Michigan 48824, U.S.A. Acidic pillared clays, e.g. alumina pillared montmorillonite (APM), and basic layered double hydroxides, e.g. hydrotalcite (HT), provide well defined surface environments for dispersing metal-cluster carbonyl complexes. In the present work, FTIR spectroscopic studies have been used to elucidate the surface organometallic chemistry of R U ~ ( C O ) , ~ on APM and HT.For APM as the support, cluster binding occurs initially by protonation to form HRu7(CO)T2 cations on the intracrystalline gallery surfaces of the clay. Further reaction results in the grafting of mononuclear sites of the type [Ru(CO), (OAIZE)~],, ( x = 2 , 3 ) to the pillar surfaces. The reaction of R u ? ( C O ) , ~ with HT affords chemisorbed HRu,(CO);, anions which can be transformed t o surface-bound [Ru(CO), (OM=):],, ( M = Al, Mg) com- plexes analogous to the grafted species on APM. The reduction of the grafted complex o n both supports results in active ruthenium catalysts for CO hydrogenation. Ru-APM exhibits very high selectivity for isomerized hydrocarbons (branched alkanes and internal alkenes). The isomerized products arise from the unique texture and bifunctional nature of Ru-APM; the clay-embedded ruthenium catalyses Fischer-Tropsch chain propagation, and the intracrystalline Bronsted acidity of the clay host catalyses alkene rearrangements through carbenium-ion mechanisms.In contrast, the Ru-HT system gives very different product distributions containing a high fraction of oxygenates, specifically methanol and lesser amounts of C,-C, alcohols. The high alcohol selectivity, which i s atypical for CO hydrogenation over Ru, is ascribed in part to the inhibition of CO dissociation on the metal particles by decoraments provided by the highly basic support. Catalysts derived from smectite clays were introduced in the petroleum industry when the fixed-bed Houdry process came on stream in 1939.' Following World War 11, clay-cracking catalysts were replaced by mixed-metal oxides and, eventually, by synthetic zeolites with improved steam stability and shape selectivity.However, recent develop- ments in the intercalation of smectite clays by robust polyoxometal oligomers has rekindled interest in clays as catalyst precursors.' This resurgence of interest is due partly to the possibility of developing materials known as metal-oxide pillared clays having unique two-dimensional (2D) galleries and zeolitic microporosity. For instance, montmorillonite clays interlayered with alumina or zirconia aggregates are highly selective acid catalysts for petroleum cracking,'-' alcohol dehydration7-10 and other acid-catalysed reactions."." Recently, reactive transition-metal centres have been introduced into the galleries of metal-oxide pillared clays in an effort to enhance the catalytic versatility of these materials.Initial approaches to metal immobilization on pillared clays have utilized impregnation techniques. ''*I4 Often, the latter methods lack the specificity needed to form clean metal crystallite~.'~-'' On the other hand, more specific metal immobilization 227228 C 0 Hydrogenation Selectivity methods based on specific surface organometallic reactions18 can afford well defined molecular species on the support which can be converted to well characterized catalytic materials. Initial studies of alumina pillared montmorillonite" indicate that the intercalated oxide reacts with metal cluster carbonyls to afford surface-grafted species analogous to those formed on the surfaces of bulk alumina.The layered double hydroxides (LDHs) of the type [ ( M ~ ~ ~ , M . ~ ' ( ~ H ) , I [ A ~ - I . ~ , ~ . z H ~ o are complementary to smectite clays insofar as the layers are 2D hydroxycations and the gallery species are anions.20*21 LDHs pillared by polyoxometallate anions have been reported recently,22 but most catalytic studies to date have focused on non-pillared hydrotalcite derivatives wherein M" = Mg, M"' - - Al, and An- = CO:-. Upon decompo- sition at elevated temperatures, these materials afford highly basic oxides for vapour- phase aldol reactions and alkene isomerizations. 2 3 ~ 2 4 In the present work, we compare the surface organometallic chemistry of R u ~ ( C O ) ] ~ on an acidic pillared clay and a basic hydrotalcite support.Since the selectivity of a catalyst can depend greatly on the support material, we have also examined the Fischer- Tropsch selectivity properties of the ruthenium catalysts derived from these support systems. Ruthenium was especially suitable for this latter objective, because it is the most active metal known for syn-gas c o n v e r ~ i o n . ~ ~ * ~ ~ Also, the unsupported metal is known to be selective toward formation of straight-chain hydrocarbon products with little or no selectivity for oxygenate f~rmation.'~ Thus, deviations from straight-chain hydrocarbon selectivity and low oxygenate yields could be correlated directly with support effects. Experimental Alumina Pillared Montmorillonite Alumina pillared montmorillonite was prepared by the reaction of sodium montmorillo- nite (Crook County, Wyoming) with aluminum chlorohydrate solution (Chlorohydrol Reheis Chemical Company) containing A1 1304(0H)24( H20)Tl oligomers according to previously described procedures.28329 The product was air-dried on a glass plate and subsequently dehydroxylated to the alumina pillared form by heating at 623 K under vacuum for 2 h. The N2 B.E.T.surface area was 300 m2 g-I, and the X-ray basal spacing was 1.85 nm. Chemical analysis indicated the unit-cell formula of the pillared clay to be [Al(OH )2.8012.87[A13.1 1 Fe,.42Mg0.481( Si7.92AlO.08)020(OH ) 4 . Hydrotalcite Synthetic hydrotalcite was prepared by a coprecipitation method.23 A 3 : 1 molar ratio of Mg(N03)2.6H20 and A1(N03)3.9H20 in distilled water was added to an aqueous solution of NaOH and Na2C03 until the pH of the mixture was 10.0. The resulting slurry was heated overnight at 338 K.The product was then recovered by centrifugation, washed with distilled water, and dried in air at room temperature. The X-ray basal spacing was 0.776 nm. Chemical analysis indicated the unit-cell formula for this material to be [Mg,Al2(0H),,](CO3)~4H20. Surface Reactions of RU~(CO),~ The reaction of Ru,(CO) 1 2 with alumina pillared montmorillonite was carried out according to previously described methods.29 A solution of R U ~ ( C O ) ~ ~ (0.02 mmol) inT. J. Pinnavaia et al. 229 Table 1. Carbonyl stretching frequencies for ruthenium complexes on alumina pillared clay (APM) and hydrotalcite (HT) supports compound frequency/cm- ’ [HRU,(CO);~]-APM [Ru(CO),~(OA~~),],,-APM 2070s, 2000s [HRU,(CO),,I-HT 2074vs, 201054 1984s, 1947vs [Ru(CO),(OM_),],,-HT(M = Mg or Al) 2047s, 1965s 2 128s, 2099s, 2077vs, 2060s, 2030s 40cm3 CH,Cl, was added under an argon atmosphere to 0.50g alumina pillared montmorillonite which had been previously dried under vacuum at 298 K for 4 h.The reaction mixture was allowed to stir for 20 h and then transferred to a nitrogen-filled glove box. The product was filtered, washed with CH2C12 and dried under a stream of argon. The ruthenium loading was 1.0 wt ‘/o, and the X-ray basal spacing was 1.85 nm. The surface reaction of Ru,(CO),~ and hydr~talcite”~” was carried out in a manner analogous to that described above for alumina pillared montmorillonite.Hydrotalcite (1 g) was dried under vacuum at room temperature for 4 h. A solution of Ru,(CO),~ (0.084 mmol) in 100 cm3 of degassed CH2C12 was added, and the reaction mixture was stirred for 20 h at 298 K under an argon atmosphere. The red reaction product was filtered and washed with a small amount of CH2C12. Chemical analysis indicated the ruthenium loading to be 0.45 wt %. Fischer-Tropsch Synthesis The catalytic hydrogenation of carbon monoxide was carried out in a stainless-steel single-pass tube reactor. The reactor tube was fitted with a quartz glass liner and a quartz glass frit to contain the catalyst. All reacting gases were ultra-high purity grade and were purified further by passing them through a manganese/silica adsorbent to remove oxygen, Linde 4A for a molecular sieve to remove water, and an alumina adsorbent at 201 K to remove metal carbonyl contaminants.Samples were analysed using a Hewlett-Packard 5890 Gas Chromatograph equipped with a flame-ionization detector, a thermal conductivity detector and automatic gas-sampling valves. The catalyst supported on alumina pillared montmorillonite was reduced in flowing hydrogen at 673 K for 2 h. Since hydrotalcite begins to decompose at a temperature near 573 K,23 the reduction of ruthenium in this support system was carried out at 548 K under flowing hydrogen for a reaction time of 16 h. The reduction temperatures were achieved at a ramp rate of 5 K min-’. Hydrogenation of carbon monoxide over the reduced catalysts was carried out under differential reaction conditions ( ( 5 % conver- sion) at gauge pressures of between 0 and 200 lbf and at 548 K.Resu 1 t s Surface Organometallic Chemistry The reaction of alumina pillared montmorillonite (APM) with Ru3(CO) 12 in CHzClz solution results in the formation of gallery-bound HRu3(CO)T2 cations on the gallery surfaces.” The infrared frequencies for the terminal carbonyl groups of the protonated cluster (cf: table 1) lie very near ( * 5 cm-’) the frequencies observed for the PF, salt of ? 1 Ibf in-’ = 6.894 76 x lo3 Pa.230 CO Hydrogenation Selectivity HRu,(CO)T2 .I9 Under the conditions of the protonation reaction, the clay retains 4.5 wt YO water. This adsorbed water undoubtedly plays a role in determining the Brgnsted acidity of the gallery regions. As the [HRu,(CO)T2]-APM is allowed to age at 298 K, the protonated cluster reacts further to form pillar-grafted mononuclear ensembles of the type [ Ru( CO),( 0Al=)Jn, where x = 2 , 3 and 0 A l ~ represents aluminate groups on the pillaring aggregates.Conversion of the protonated clusters to the pillar-grafted ensembles is complete within 24 h at 298 K. The vibrational frequencies of the terminal CO groups on the grafted ensembles are provided in table 1. These frequencies lie near those for analogous species formed on the surfaces of bulk The reactions of hydrotalcite (HT) with R U ~ ( C O ) ~ ~ in CH2Cl2 solution results in the formation of surface-bound H Ru3( CO) anions. The infrared vibrational frequencies of the cluster anion, presented in table 1, are in good agreement with those observed for authentic salts of HRu3(CO)LI .,' Upon exposure to air the HRu,(CO),-HT reacts further to form grafted [ Ru( CO),( OM-),]n (M = Mg, Al) species, analogous to the grafted complexes formed on APM. The carbonyl stretching frequencies for the HT-bound complex are provided in table 1 .Carbon Monoxide Hydrogenation The reduction of [Ru(CO),(OAI=),],-APM with hydrogen at 673 K results in the formation of ruthenium crystallites in the acidic microporous support. X-Ray diffraction measurements indicate that the 1.85 nm basal spacing of APM is not altered in the reduction reaction. Previously reported transmission electron microscopy indicate the ruthenium particle size to be <5.0nm. The reduction of [ R u ( C O ) . ( O A ~ ~ ) ~ ] ~ - H T in hydrogen at 548 K also leads to the formation of ruthenium crystallites which can be observed by electron microscopy.TEM analysis indicates the average ruthenium particle size to be 5.9 nm. The LDH support matrix remains crystalline after the reduction step, as evidenced by the presence of 001 reflections in the X-ray diffraction pattern. Also, I R spectroscopy indicated the presence of carbonate in the support matrix. The product distribution obtained for the hydrogenation of carbon monoxide over Ru-APM at 1309 kPa and 548 K follows Anderson-Schulz- Flory (ASF) statistics over the product range C,-C,. As can be seen from fig. 1, the methane yield is much higher than expected based on ASF statistics with a chain-propagation probability of (Y = 0.500*0.002.The high methane yields are presumed to result from hydrogenolysis of higher hydrocarbons. The carbon-number distribution for CO conversion in the highly basic Ru-HT catalyst system also follows ASF statistics. As shown by the data in fig. 2 for products in the C2-C8 range, the chain-propagation probability, (Y = 0.447 f 0.002, is similar to the value observed for the APM support system. Also, the methane yield is much higher than expected based on ASF statistics. Despite the similarities in reactivity and chain-propagation probabilities for ruthenium dispersed on acidic APM and basic HT supports, two dramatic differences in product selectivity are observed for these support systems: ( i ) Ru-AMP exhibits very high yields of isomerized products (branched alkanes and internal alkenes) relative to normal products (n-alkanes and terminal alkenes), whereas Ru-HT affords low ratios of isomerized to normal product ratios, and (ii) Ru-APM yields only trace amounts of oxygenated products (< 1 "/o), whereas Ru-HT affords substantial yields of alcohols in the C,-C4 range.Table 2 provides the ratio of isomerized to normal-chain hydrocarbons obtained for Ru-APM and Ru-HT. For C4-C9 products produced over the acidic APM support, 64-90% of the hydrocarbon chains have been isomerized. In contrast, the C,-C, productsT. J. Pinnavaia et al. 23 1 7 0 60 50 h 8 40 E f 30 '0 x lz 20 10 0 1 2 3 4 5 6 7 8 carbon no. 0 - 1 - 2 - 3 --. 3= v - 4 2 -5 -6 - 7 Fig. I . CO hydrogenation product distribution and Anderson-Shulz- Flory plot for Ru- APM catalyst at 548 K and 1309 kPa.formed over the basic HT support are only 30-38% isomerized. The extent of isomeriz- ation in the Ru-APM system is highly dependent on the H2: CO ratio. As shown in fig. 3, the isomerized to normal hydrocarbon products in the C,-C, range increase substantially with decreasing H., : CO ratio. It is particularly significant that Ru-HT produces substantial yields of methanol, while no methanol was observed with Ru-APM. The selectivity towards alcohol forma- tion depends on the reaction pressure and temperature. The pressure dependence is indicated by the data in table 3. Note that the total alcohol yield increases with increasing pressure. Although the methane yield decreases with increasing pressure, the fraction of total alcohol present as methanol (ca.7 5 % ) changes little over the pressure range investigated. Also, decreasing the temperature of CO hydrogenation over Ru-HT from 548 to 533 K, at 826 kPa increases the alcohol yield from 13.2 to 30.8%. The same temperature change causes the fraction of total alcohol present as methanol to increase slightly from 75 to 8 1'7'0. Discussion Surface Organometallic Chemistry R U ~ ( C O ) , ~ binds initially to partially hydrated APM as the protonated species HRu3(CO)r2. The protonation of the ruthenium cluster carbonyl attests to the strong Bronsted acidity of the pillared clay support. Normally, acid strengths equivalent to 98% H2S04 or trifluoroacetic acid are required for protonation of the cluster in homogeneous s o l ~ t i o n .~ ~ ~ ~ ~ The acidity of the pillared clay is believed to arise from the partial thermal dehydration and dehydroxylation of the intercalated Al 1304(OH)24+x( H20)\7,_",'+ ion to form alumina aggregates and ionizable p r o t o n ~ . ~ ~ - ~ ~232 CO Hydrogenation Selectivity 60 50 40 h 8 3 U v 30 -f! E: V T3 c" 20 10 0 -1 - 2 -3 ; 1 3- - 4 f v - 5 -6 -7 1 2 3 4 5 6 7 carbon no. Fig. 2. CO hydrogenation product distribution and Anderson-Shulz-Flory plot for Ru-HT catalyst at 548 K and 1309 kPa. Table 2. Isomerized/normal hydrocarbon product ratios" for CO hydrogenation over Ru supported on APM and HT (1309 kPa, 548 K) no. Ru-APM Ru-HT 4 1.79 0.45 5 5.32 0.5 1 6 6.8 1 0.63 7 9.75 - 8 6.94 - 9 2.98 - " This ratio is defined as sum of branched hydro- carbons and internal alkenes divided by the sum of n-alkanes and terminal alkenes.Upon ageing, or upon exposure to air, the chemisorbed HRu3(CO)T2 ion is converted to the pillar-grafted complex [ RU(CO).(OA~E)~]~. Grafting to the intercalated alumina pillars rather than the external surfaces of the clay layers is supported in part by the fact that unpillared montmorillonite is capable of binding only trace amounts of the ruthenium complex. Conversion of the pillar grafted ruthenium ensembles to metallic ruthenium crystallites is readily accomplished by reduction with hydrogen at 673 K. Earlier electron microscopy studies have demonstrated that the ruthenium crystallites are embedded within the microporous pillared clay particles with very little rutheniumT.J. Pinnavaia et al. 233 1 ’ I I I I 0.0 0.5 1 .o 1.5 2.0 H*/CO Fig. 3. Dependence of isomerized/normal hydrocarbon ratios for C4-C6 products on the H,/CO reactant ratio for conversion over Ru-APM catalyst at 548 K and 1309 kPa. 0, C,; 0, C,; 0, C,. Table 3. Pressure dependence of CO hydrogenation selectivity over Ru-HT‘ hydrocarbon and total alcohol yields (wt %) alcohol distribution pressure total /kPa C, C? C3 C4 C5 C,, alcohols MeOH EtOH PrOH BuOH 101 85.5 8.8 5.7 tr tr tr tr tr tr tr tr 482 65.2 8.3 7.0 4.5 2.8 1.9 10.0 78 22 tr tr 826 63.5 7.5 7.4 4.5 2.4 1.5 13.2 75 20 5 tr 1171 54.1 7.6 9.1 6.0 3.3 2.5 17.6 78 18 4 tr 1309 52.2 7.3 8.8 5.7 3.2 2.8 20.1 74 17 4 4 “ Temperature = 548 K; H,/CO = 2.0; CO Conversion<S0/~; GHSV = 1000-3000 h-’; Time on stream>24 h.immobilized at external surfaces.29336 The embedding of ruthenium crystallites within the clay particles has important catalytic consequences, as will be discussed below. The organometallic chemistry of R u ~ ( C O ) , ~ on a basic hydrotalcite support is complementary to the acid-mediated chemistry observed on alumina pillared montmorillonite. On the basic support the ruthenium cluster carbonyl undergoes reduc- tive decarbonylation to the H Ru3( CO) 1, anion. Equivalent reductive decarbonylations in homogeneous solution require the presence of very strong bases such as potassium hydr~xide.”~ The basicity of the layered double hydroxide arises from the hydrolysis of the surface carbonate anion,44 not by reaction of the lattice hydroxyls.234 CO Hydrogenation Selectivity R H+ Oxidation of HRu,(CO),,-HT in air results in the formation of grafted [ R~(CO),(OMEZ)~],, analogous to the grafted complexes formed on APM.However, since HT is non-microporous, the ruthenium grafted complexes are restricted to occupy- ing external surfaces only. This latter result is verified by the retention of the 0.77 nm X-ray basal spacing characteristic of the carbonate intercalate of the layered-double hydroxide. Upon reduction in hydrogen at 548 K, ruthenium crystallites with an average particle size of 5.9 nm are observed at the external surfaces of the support by electron microscopy. CO Hydrogenation Selectivity Ruthenium dispersed on conventional silica and alumina supports is highly selective for the catalytic hydrogenation of carbon monoxide to linear terminal alkenes and alkane^.*'^^^ In contrast, ruthenium supported on alumina pillared clay exhibits a very high selectivity towards isomerized products (branched hydrocarbons and internal alkenes) relative to normal products (n-alkanes and terminal alkenes).Between 64 and 90% of the products formed in the C,-C, range are isomerized derivatives (c$ table 2). The high yields of isomerized products obtained for the Ru-APM catalysts system most likely arises from the high Brmsted acidity of the pillared clay support. At Fischer-Tropsch reaction temperatures, the adsorption of carbon monoxide on clean ruthenium surfaces is known to involve a dissociative mechani~rn.~~ When the metal is dispersed on a conventional support, the ruthenium alkyl intermediates are converted to terminal straight-chain alkenes or normal alkanes. However, if the metal is dispersed on an acidic support, the terminal alkenes are capable of forming carbenium ions.The carbenium ions may rearrange via protonated cyclopropane intermediates to branched- chain h y d r o ~ a r b o n s . ~ ~ - ~ ~ The following scheme summarizes the elementary steps involved in converting metal alkyl intermediates to internal alkenes (paths B-D) and to branched-chain hydrocarbons (paths E-G). We propose that the pathways represented in the above scheme are responsible for the high isomerization selectivity of Ru-APM. The importance of terminal alkenes as intermediates is supported by an increase in selectivity towards isomerized productsT.J. Pinnavaia et a1 23 5 Fig. 4. Schematic representation of defect sites occupied by ruthenium in a Ru-APM catalyst. The shaded particles represent Ru crystallites, the filled circles represent alumina pillars and the slabs represent the silicate layers of the clay. with decreasing hydrogen-to-CO ratio (cf: fig. 3 ) . As indicated by pathways A and B in the scheme, the formation of terminal alkenes should be facilitated by a decrease in the hydrogen-to-carbon monoxide ratio. The selectivities towards isomerized products observed in the present work are unusual but not unprecedented. Ruthenium as a syn-gas catalyst supported on the hydrogen-exchanged form of dealuminated zeolite Y affords substantial yields of isomerized hydrocarbons. 27,4950 Nevertheless, the high selectivity of Ru-APM towards isomerized Fischer-Tropsch products even at low conversions, where the fugacity of alkene products is low, remains exceptional.We attribute the selectivity of Ru-APM, in part, to the special textural features of the microporous support. Pillared clays contain numerous defects which arise from distortions of the host layers and mis-matching of layer edges, as illustrated schematically in fig. 4. In the case of Ru-APM, the size of the embedded metal crystallites (1.0-5.0 nm) is larger than the size of the pillared galleries (ca. 0.85 nm). Thus, the ruthenium aggregates appear to be stabilized by the clay defect sites. The encapsulated ruthenium particles are readily accessible to hydrogen and carbon monoxide. When chain propaga- tion is terminated and a terminal alkene is released from the metal site, the alkene is obligated to migrate through the acidic microporous structure of the pillared clay where carbenium-ion formation and chain isomerization can occur.Consequently, Ru- APM is a bifunctional catalyst which serves as a support for the intracrystalline dispersion of ruthenium crystallites for Fischer-Tropsch chain propagation and, at the same time, as an acidic microporous medium for carbenium-ion formation and isomerization. The selectivity of ruthenium supported on basic hydrotalcite also is highly unusual owing to the substantial oxygenate (alcohol) yields obtained at relatively low reaction pressures. Although ruthenium normally adsorbs CO by a dissociative mechanism under syn-gas reaction conditions, basic supports such as MgO and alkali-metal-promoted alumina have been observed to form alcohols.51~5' I t is especially noteworthy, however, that Ru-HT is a much more effective support matrix for improving oxygenate yields over dispersed ruthenium than are MgO or alkali-metal-promoted aluminas.The latter supports require reaction pressures of 8000 kPa to achieve alcohol yields of 20 wt %. In contrast, analogous alcohol yields are obtained with Ru-HT at substantially lower reaction pressures (1300 kPa). The oxygenate selectivities of syn-gas catalysts are known to depend on the support matrix. For instance, in the case of rhodium, basic supports promote methanol formation, whereas non-basic supports provide high yields of hydrocarbons.The methanol selec- tivity of rhodium on basic supports is related directly to the surface coverage of236 CO Hydrogenation Selectivity non-dissociatively adsorbed CO on the metal surface.53 In general, mediation of catalytic selectivity by a support matrix may involve an electronic effect on the dispersed metal ~ r y s t a l l i t e ~ ~ or, alternatively, a support could provide decorants for the surface modification of the metal c r y ~ t a l l i t e s . ~ ~ - ~ ~ Since electronic effects of the support extend over only two to three atomic layers of clean metal, it is unlikely that such effects are operative for the Ru-HT catalyst system where the average crystallite size is 5.9 nm. We propose that the mechanism for oxygenate selectivity in the Ru-HT catalyst arises from the decoration of the metal crystallites by the basic support material.A similar process is responsible for oxygenate formation over rhodium. In this case, the function of the oxide decorament is ( i ) to block metallic Rho surface sites and suppress the production of CH, units through CO dissociation, and (ii) to stabilize non-reduced Rh"+ sites, including the possibility of Rh203 on the surface of the The unreduced Rh"' sites are responsible for methanol formation, whereas the metallic centres on the decorated metal are responsible for C l oxygenate formation.57 The latter species are most likely formed through a metal acyl reaction pathway.58 The increase in oxygenate selectivity for Ru-HT with increasing pressure ( c j table 3) and decreasing temperature is consistent with the proposed inhibition of CO dissociation by support decoraments.This research was supported in part by the National Foundation, Division of Materials Research, through grant DMR-85 14154, and the Michigan State University Center for Fundamental Materials Research. Fellowship support for E.P.G. and E.G.R. was pro- vided by Exxon Corporation and ECC International, respectively. References 1 H. H. Voge, ACSSymp. Ser., 1983, 222, 3410. 2 T. J. Pinnavaia, Science, 1983, 220, 365. 3. J . Shabtai, R. Lazar and A. G. Oblad, Stud. Su[j:1: Sci. Catal., 1981, 7, 828. 4 R. J. Lusier, J. S. Magee and D. E. W. Vaughan, Preprints of the 7th Canadian Symp. on Catalysis, Edmonton, Alberta, Canada, October, 1980. 5 M. L. Occelli, 9th Eng. Chem.Prod., Res. Dev., 1983, 22, 553. 6 M. L. Occelli, S. D. Landau and T. J. Pinnavaia, J. Catal., 1984, 90, 250. 7 R. Birch and C. 1. Warburton, J. Catal., 1986, 97, 51 1. 8 A. K. Galway, J. Card, 1970, 19, 330. 9 M. L. Occelli, R. A. Innes, F. S. S. Hwu and J . W. Hightower, Appl. Catal., 1985, 14, 69. 10 Y. Morikawa, T. Goto, Y. Moro-Oka and Ikawa, Chem. Lett., 1982, 1667. 11 M. L. Occelli, J. T. Hsu and L. G. Galya, J. Mol. Catal., 1985, 33, 371. 12 J. A. Ballantine in Chemical Reactions in Organic and Inorganic Constrained Svstem, ed. R. Setton (Reidel, New York, 19861, pp. 197. 13 V. N. Parulekar and J. W. Hightower, Appl. Catal., 1987, 35, 249. 14 V. N. Parulekar and J. W. Hightower, Appl. Catal., 1987, 35, 263. 15 B. J. Tatarchuck and J. A. Dumesic, J. Catal., 1981, 70, 308.16 M. A. Wheeler and M. T. Bettman, J. Catal., 1974, 473, 124. 17 V. K. Jones, L. R. Neubauer and C . H . Bartholomew, J. Phys. Chem., 1986, 90, 4382. 18 Y. I . Yermakov, B. N. Kuznetsov and V. A. Zakharov, in Cataly.sis by Supported Complexes (Elsevier, 19 E. P. Giannelis, E. G. Rightor and T. J. Pinnavaia, J. Am. Chem. Soc., 1988, 110, 3880. 20 S. Miyata, C l a j ~ Clay Mineral., 1980, 28, 50. 21 R. M. Taylor, Clay Mineral., 1984, 19, 591. 22 T. Kwon, G. A. Tsigdinos and T. J. Pinnavaia, J. Am. Chem. Soc., 1988, 110, 3653. 23 W. T. Reichle, S. Y. Kang and D. S. Everhardt, J. Catal., 1986, 101, 352. 24 T. Nakatsuka, H. Kawasaki, S. Yamashita and S. Kohjiya, Bull. Chem. Soc. Jpn, 1979, 52, 2467. 25 M. A. Vannice, Catal. Rev. Sci. Eng., 1976, 14, 1.53. 26 M.A. Vannice, J. Catal., 1975, 37, 469. 27 D, L. King, J. Catal., 1976, 15, 38. 28 T. J. Pinnavaia, M. S. Tzou, S. D. Landau and R. H . Raythatha, J. Mol. Catal., 1984, 27, 3942. 29 T. J. Pinnavaia in Heterogeneous Cataljqyi.s, ed. B. Shapiro (Texas A&M Press, College Station, Texas, Amsterdam, 1981 ). 19841, pp. 141.T. J. Pinnavaia et al. 237 30 M. Rameswaran, E. G. Rightor, E. D. Dimotakis and T. J. Pinnavaia, Proc. 9th Int. Cong. Caral., 1988, 31 M. Rameswaran and T. J. Pinnavaia, Prepar. Pap. Am. Chem. SOC. Diu. Fuel Chem., 1988, 33, 942. 32 A. Theolier, A. Chopin, L. D’Ornelas, J. M. Basset, G. M. Zanderighi, R. Ugo, R. Psaro and 33 J. Evans and G. S. McNulty, J. Chem. Soc., Dalton Trans., 1984, 1123. 34 M. Deeba, B. J. Streusand, G. L.Schrader and B. C. Gates, J. Catal., 1981, 69, 218. 35 ( a ) B. F. G. Johnson, J. Lewis, P. R. Raithby and G. Suss, J. Chem. SOC., Dalton Trans., 1979, 1356; ( b ) E. R. Eady, P. F. Jackson, B. F. G. Johnson, J. Lewis, M. C. Malatesta, M. McPartlin and J. H. Nelson, J. Chem. SOC., Dalton Trans., 1980, 383. 2, 783. C. Sourisseau, Polyhedron, 1983, 2, 119. 36 E. G. Rightor and T. J. Pinnavaia, Ultramicroscopy, 1987, 22, 159. 37 J. Knight and M. J. Mays, J. Chem. Soc. A, 1970, 711. 38 B. Delley, M. C. Manning, D. E. Ellis, J. Barkowitz and N. C. Trogler, Inorg. Chem., 1982, 21, 2247. 39 M. L. Occelli and R. M. Tindwa, Clays Clay Mineral., 1983, 31, 22. 40 J. Shabti, F. E. Massoth, M. Tokarz, G. M. Tsai and J. McCauley, Proc. 8th Int. Congr. Card (Verlag Chemie, Berlin, 1984), vol. IV, p. 735. 41 D. Plee, A. Shultz, G. Poncelet and J. J. Fripiat, Stud. Surf: Sci. Catal., 1985, 20, 343. 42 D. Tichit, F. Fajula and F. Figureras, Stud. Surf: Sci. Catal., 1985, 20, 351. 43 P. T. B. Tennakoon, W. Jones and J. M. Thomas, J. Chem. SOC., Faraday Trans. I , 1986, 82, 3081. 44 E. P. Giannelis and T. J. Pinnavaia, Inorg. Chem., submitted. 45 C . S. Kellner and A. T. Bell, J. Caral., 1981, 75, 251. 46 J. Weitkamp, Ind. Eng. Chem. Prod., Res. Dev., 1982, 21, 549. 47 B. C. Gates, T. R. Katzer and G. C. Schuit, in Chemistry of Catalytic Process (McGraw-Hill, New 48 G. M. Kramer, G. B. McVicker and J. J. Ziemiak, J. Card., 1985, 92, 355. 49 R. Oukaci, A. Sayari and J. G. Goodwin, J. Catal., 1987, 106, 318. 50 T. Tatsumi, Y. G. Shul, T. Sugivra and H. Tominaga, Appl. Caral., 1986, 21, 119. 51 A. P. Risch and J. A. Rabo, J. Catal., 1978, 52, 157. 52 A. Bossi, F. Garbassi and G. Petrinni, Preprinrs of 7th Int. Cong. Card., Tokyo, E4, 1980. 53 J. R. Katzer, A. W. Sleight, P. Gajardo, J. B. Michael, E. F. Gleason and S. MacMillan, Faradajl 54 R. W. Joyner, J. B. Pendry, D. K. Saldin and S. R. Tennison, S u r - Sci., 1984, 138, 84. 55 G. van der Lee, B. Schuller, H. Post, T. L. F. Favre and V. Ponec, J. Catal., 1986, 98, 522. 56 H. Y. Luo, A. G. T. M. Bastein, A. A. J. Mulder and V. Ponec, Appf. Caral., 1988, 38, 241. 57 G. van der Lee and V. Ponec, .I. Caral., 1986, 99, 511. 58 W. M. H. Sachtler, D. F. Shriver and M. Ichikawa, J. Card., 1988, 99, 513. York, 1979), pp. 20-22. Discuss. Chem. SOC., 1981, 72, 121. Paper 8/049931; Received 15th December, 1988

ISSN:0301-7249    

DOI:10.1039/DC9898700227

出版商:RSC    

年代:1989

数据来源: RSC

摘要:

Faraday Discuss. Chem. SOC., 1989, 87, 239-251 Metallic Glasses in Heterogeneous Catalysis Alfons Baiker Department of Industrial and Engineering Chemistry, Swiss Federal Institute of Technology ( ETH), ETH-Zentrum, CH-8092 Zurich, Switzerland Metallic glasses exhibit some unique properties which make them interesting materials in catalysis. Recently their use as catalyst precursors has been advanced and several efficient catalysts have been prepared by various pretreatments of the metallic glasses. An understanding of the solid-state reactions occurring during the transition of the amorphous precursor to the active catalyst was found to be crucial for successful application of these materials. This paper describes the present knowledge and discusses prin- cipal possibilities and problems involved in the application of metallic glasses in catalysis. The use of metallic glasses in catalysis was first reported at the beginning of this decade.’-3 Early investigations in this field were of a rather phenomenological nature, focusing mainly on the catalytic behaviour with little attention to the surface and bulk characterisa- tion of the materials used.Reviews which cover these activities have been written by S ~ h l o g l , ~ Yoon and Cocke,’ and Shibata and Masumoto.6 It was only recently that researchers started to give intensive consideration to the characterisation of the amor- phous alloys used in the catalytic reaction. These investigations showed that presumably in most applications of metallic glasses in catalysis the surface of the metastable amorphous alloy undergoes chemical and structural changes under reaction conditions.These observations, coupled with the fact that as-quenched alloys exhibit very low surface areas, and consequently low activity, led several investigators to use metallic glasses as catalyst precursors rather than as catalysts. The aim of this paper is to review briefly the knowledge gained and to show where metallic glasses may be of interest in catalysis. Properties and Preparation of Metallic Glasses Metallic glasses’ can be regarded as congealed metallic melts, rigid but devoid of crystalline order. Another name sometimes used is amorphous metal alloys, which underlines the fact that such materials are all alloys and never pure metals. The great variety of metallic glasses reported in the literature’ fall into a few well defined categories: (i) late transition metal + metalloid; (ii) early transition metal + late transition metal or Group IB metal; (iii) alkaline-earth metal +Group IB metal; (iv) early transition metal + alkali metal; and (v) actinide+ early transition metal. For catalysis only metallic glasses of categories (i) and (ii) have been applied so far (table 1).Various factors have been suggested to influence the glass-forming tendency, among which are the thermodynamics (phase diagram), kinetics, metastable phases, stoichiometry, concentration of valence electrons, atomic sizes of constituents and electronegativity. Also a ‘confusion principle’ has been proposed, to the effect that complex mixtures of constituents have greater glass-forming tendency than binary mixtures.Frequently, the search for alloys that readily yield metallic glasses is aided by the fact that eutectic compositions are favoured. 239240 Metallic Glasses in Heterogeneous Catalysis Table 1. Catalytic reactions studied over metallic glasses or catalysts derived therefrom major products alloy ref. C , -C, hydrocarbons C , -C3 hydrocarbons C I -C, hydrocarbons C I -C, hydrocarbons C -C, hydrocarbons methanation methanation methanation C I -C, hydrocarbons C I -C, hydrocarbons C , , C, hydrocarbons methanol synthesis methanol synthesis methanol synthesis methanation acetylene ( + )-apopinene buta-l,3-diene ethene, buta-1,3-diene buta- 1,3-diene buta-1,3-diene buta- 1.3-diene cis-cyclododecene cyclohexene, n-hexene phenylethyne, a-pinene cyclohexene, benzene ethene, propene, isoprene cis-but-2-ene, buta-l,3-diene ethene ethene ethene ethene, isoprene oct-1 -yne, oct-4-yne, phenylacetylene hex-1-ene phenylethyne methyl formate carbon monoxide ethane, cyclopropane 5-aminopentan-1-01 1 3 24 25 26 27 28,29 30 31,32 33 34 35 36 37 38 39 16,17 40 41 42 43 44 2, 14 15 45 46 47,48 49 50 51 18 52 53 54 55 56 57 58 47 59A.Baiker 24 1 Several principal can be used for the production of metallic glasses, including vapour and sputter deposition, plating and melt quenching. Among these methods melt-quenching is the one most widely used. In melt-quenching the melt of the constituents is so rapidly quenched that there is insufficient time for crystallites to nucleate and grow.The common feature of all devices that have been designed to quench molten alloys is that the melt must be converted very rapidly from a jet or stream of droplets into a thin layer in contact with a ‘chill block’ to produce thin foil or ribbon of 10-60 p m thickness. The two most widespread techniques are melt-spinning and melt-extraction. In both, the ‘chill-block’ is a rapidly spinning copper wheel. Melt- spinning produces thin ribbons, melt-extraction makes fine wires. It is important to note that the glassy metals used so far in catalysis have mainly been produced by melt-spinning. Motivation for using Metallic Glasses in Catalysis Metallic glasses possess several properties’”’ which make them interesting materials in catalysis. Ideally, the surface of amorphous materials should be devoid of any long-range ordering of the constituents and exhibit a high density of low-coordination sites and defects.The important role of low-coordination sites in catalysis, such as terrace, step and kink sites, has been demonstrated by Somorjai’ ’ on crystalline materials. Metallic glasses possess high flexibility with regard to fine-tuning of the electronic properties,I2 mainly due to the fact that thermodynamic constraints are less severe in supercooled liquids than in crystalline materials. Metallic glasses are ideally chemically homogeneous and structurally isotropic. Metallic glasses are highly reactive owing to their metastable structure and they undergo solid-state reaction frequently more easily than their crystalline counterparts.This property is of importance in their use as catalyst precursors. Metallic glasses exhibit good conductivity for electricity and heat, making them particularly interesting for application in electrocatalysis. Metallic glasses prepared by melt-quenching exhibit a planar morphology ideal for investigation with modern characterisation tools such as photoelectron spectroscopy and scanning tunnelling microscopy. The manufacture of foils or small ribbons may facilitate the design of new reactor conceptions. Limitations with regard to the application of metallic glasses in catalysis orginate from their metastable structure and the intrinsically low surface area, which corresponds to the geometrical area of as-quenched materials. Catalytic Studies Presently available reports on catalysis over metallic glasses are listed in table 1 .Two different types of investigations can be distinguished depending on the degree of pretreatment of the amorphous metal alloys, namely, catalytic studies on fresh unrecon- structed surfaces of metallic glasses, and studies which were centred on the role of pretreatment and its influence on the catalytic properties of the metallic glasses. Studies on Fresh Unreconstructed Surfaces of Metallic Glasses Relatively little catalytic work has been carried out so far under conditions where the surface of the metal alloys can be regarded as unreconstructed, i.e. where the chemical composition and structure of the surface can be assumed to be in the state characteristic242 Metallic Glasses in Heterogeneous Catalysis of the freshly quenched material.In principle, such investigations can be performed only at temperatures far below the crystallisation temperature of the alloy and require special precautions to eliminate possible contamination of the alloy during its transfer from the site of fabrication to the catalytic reactor. In their initial work, Smith et al.' showed that Pd-Si glasses, made by splat cooling,' are active catalysts for the deuterogenation of cis-cyclododecene at room temperature. Pd-Si glasses produced more trans-isomerisation, more dideutero-saturate and less extensive exchange than crystalline Pd. Later these studies were extended by including Pd-Ge glasses" and comparing the catalytic behaviour of the metallic glasses with their crystalline counterparts.The different selectivities found for the above hydrogenation over amorphous and crystalline Pd systems was attributed to the different surface topography of the metallic glasses and the crystalline alloys. The glassy surfaces were suggested to be free of atomically flat terraces and to be highly populated with protuberances approximating kinks and ledges, as compared with the surface of large crystallites, where the terraces predominate. The kinks and ledges on crystalline surfaces are of discrete dimension^,'^ whereas the glassy surface is assumed to present protuberances with a continuum of coordination numbers. In contrast to the results above, Giessen et a1.15 found no significant differences in the selectivity behaviour between glassy and crystalline phases of Pd,,,Si,,, during hydro- genation reactions of n-hexene, phenylethyne, a-pinene, and cyclododecene.Catalytic selectivities with regard to cis- trans isomerisation, double-bond migration and the stereochemistry of addition were approximately the same regardless of whether a glassy or crystalline catalyst was used. Minor differences were observed in hydrogen-deuterium exchange. The reason for these contrasting results is not clear. More recently, it has been suggested that the hydrogenation of cis-cyclododecene is not very suitable for characterising differences in the structure of these surfaces, since an isomer (trans-cyclododecene) is produced in the reaction which has a rate of hydrogenation different from that of the parent compound.These differences in rates of hydrogenation mask the actual amounts of isomerisation that are occurring and, therefore, conceal the kinds of catalytic sites available on the surface. With this in mind, Smith and c o - ~ o r k e r s ~ ~ ~ " tested the molecule ( + )-apopinene (6,6-dimethyl-1 R,5 R- bicyclo[3.1 . l ] hept-2-ene) as a surface probe to distinguish the relative percentages of terraces, ledges and kinks available on the metallic surface. This probe molecule is more useful than the cis- cyclododecene used earlier because its isomerisation product (-)-apopinene has an identical rate of hydrogenation on a symmetrical surface. The crystallised alloys showed a higher ratio of isomerisation to deuteration than the parent amorphous alloys.I t was suggested that the isomerisation of ( + )-apopinene reflects the total number of ledge, kink and terrace sites, whereas hydrogenation reflects only the number of kinks. Based on these arguments, Smith and co-workers concluded that the surface structure of an amorphous alloy is not two-dimensionally random (flat), but is three-dimensionally random (hilly or rolling). Molnar et aLx studied the selective half-hydrogenation of phenylacetylene, oct- 1-yne and oct-4-yne over Pd-Si and Pd-Ge glassy and crystalline catalysts, and for comparison, over splat-cooled Pd, reduced Pd02 and Pd foil. They found that terminal alkynes comminute Pd structures and expose new active sites. These sites are different on the rapidly cooled catalysts compared to the regularly crystallised catalysts.Although no significant changes were detected in alkyne-hydrogenation selectivities after several hydrogenations, marked changes were revealed by ( + )-apopinene. On the terminal acetylene-treated foils and on the reduced PdO?, the rates of hydrogenation and isomeri- sation of ( + )-apopinene increased, but the ratio of the two rates remained almost the same. In contrast, the splat-cooled catalysts showed a higher rate increase for isomerisa- tion than for hydrogenation. Molnar et al." suggested that the effect of the terminalA. Baiker 243 alkynes is to expose sites of lower coordinative unsaturation, especially ledges and kinked sites by comminution of the Pd structures. These newly exposed sites are assumed not to influence alkyne reactions.It is interesting to note that heat treatment of the amorphous alloys was previously" reported to have a similar effect on the surface structure of the amorphous alloys. Direct insight into the surface structure of metallic glasses has recently been obtained by using scanning tunnelling microscopy (STM). Wiesendanger et ~ 1 . ' ~ performed STM measurements on glassy Rh25Zr75 prepared by melt-spinning. They found that the surface is made up of flat areas, identified as disordered regions, and hill structures, which they attributed to nanocrystals embedded in the amorphous matrix. Similar surface mor- phology was recently found by Walz et al.") for amorphous Feg1Zr9 prepared by melt-spinning. They concluded from their STM investigations that the surfaces of the Fe-Zr samples were at least partially crystallised.Schlogl et aL21 investigated the surface morphology of amorphous Feg1Zr9 alloys using both scanning electron microscopy (SEM) and STM. They also found pronounced anisotropy of the surface structure of the melt-quenched material. The surface was reported to consist of disc-shaped structures several hundred ingstroms in size, with clear valleys separating them. The valleys were attributed to inhomogeneous cooling of microdroplets of the liquid alloy. It could be clearly shown that the smooth surface of the droplets consists of irregular corrugations with an average distance between the maxima of several nanometres and an amplitude of below 1 nm. It has been suggested that the latter structure originates from frozen waves excited on the surface of the formerly liquid microdroplets preserved by the rapid quenching rate of ca.lo6 K s-I. Insight into the electronic structure of amorphous metal alloys surfaces has been gained by studying their interaction with probe molecules using photoelectron spectros- copy (UPS and XPS). Hauert et al.'? studied the chemisorption of CO on Ni-Zr metallic glasses using UPS. Chemisorption of CO was investigated as a function of exposure, temperature and alloy composition. Molecular and/or dissociative CO chemisorption was observed depending on the alloy composition. Molecular chemisorption was pre- dominant on nickel, whereas on zirconium dissociative chemisorption was prevalent. However, Hauert et a1.22 observed that the ratio of molecular to dissociative CO chemisorption was not directly related to the surface 'composition of the Ni-Zr alloys.This phenomenon was attributed to Zr modifying the local electronic structure at the Ni-atom sites in such a way that the chemisorption behaviour of these sites is profoundly different from elemental nickel. More recently, Baiker et al.'3 investigated the adsorption of nitrogen and CO on amorphous Feg1Zr9 using UPS and XPS. The studies showed that both molecular and dissociated nitrogen were present on the surface after exposure to dinitrogen at 79 K. Upon addition of small amounts of hydrogen to the surface the dissociated nitrogen species was desorbed completely, presumably as ammonia. The results of the UPS and XPS studies, including binding energies and line profiles, agreed well with the results of similar investigations carried out on single-crystal surfaces of iron.Similarly, no differences were observed in the CO adsorption normalized to the number of iron sites in the surface (amount adsorbed, kinetics, chemical shift) when comparing CO adsorp- tion on polycrystalline iron and amorphous Fe9,Zr9. This indicated that the local electronic structure of the adsorption sites on the amorphous alloy surface are similar to those of elemental iron. The similarity of the nitrogen adsorption, which is known to be structure-sensitive, may be taken as an indication that both the polycrystalline and the 'amorphous' surfaces exhibit similar microstructures. The similarities in the microstructure of the surfaces may have been produced by the extensive sputter-cleaning of both samples before adsorption experiments.All the experimental investigations discussed above indicate that the surfaces of metallic glasses prepared by melt-quenching are rather inhomogeneous, containing244 Metallic Glasses in Heterogeneous Catalysis disordered regions and crystalline-like regions. These are probably created during the quenching process. Thus the ideal amorphous metal surface, being isotropic and showing uniform short-range ordering, may in reality be difficult to produce, and even more difficult to maintain under conditions where catalytic reactions are performed. This phenomenon imposes severe limitations to all applications of metallic glasses, where a stable disordered surface structure is demanded.The final answer concerning the short-range ordering of amorphous surfaces cannot yet be provided. High-resolution STM of the topography and the local tunnelling barrier height of the amorphous areas could be promising in providing information about short-range order in metallic glass surfaces if the electronic surface structure and the geometric surface structure can be correlated. l Y Studies on Pretreated Metallic Glasses In most catalytic applications of metallic glasses, pretreatment of the as-quenched materials (e.g. in a reducing-gas atmosphere) was found to be crucial to obtain high catalytic activities. Several factors may contribute to this behaviour, the most important being: (i) the surfaces of metallic alloys exposed to air are likely to be covered with a superficial layer of inactive metal oxides; (ii) the surface area of as-quenched materials is very small (usually <0.1 m’g-’) and is therefore easily deactivated in the presence of contaminants. Several different procedures have been applied to improve the catalytic properties of as-quenched materials, including reduction in hydrogen 1,325726~3 1*39743350 or in other reducing-gas atmospheres ( e.g.H2/C0,24,28-30,35*37 H2/CO-,,37338 H2/N253*54) as well as etching in acid solutions ( HCl,33 H N O ~ , ~ ~ ’ ~ ~ . ~ ~ , ~ ~ HF4’,51 ) followed by oxidation and reduction. It seems likely that in most cases where such pretreatments were applied, the original surface structure of the amorphous alloy was altered. Thus comparative studies of the catalytic behaviour of the pretreated amorphous metal alloys and their crystalline counterparts do not generally provide a reasonable basis for answering the question of whether or not the amorphous surface is more active than the corresponding crystalline surface.In any case, it is interesting to note that in most studies [exceptions are reported in ref. (15) and (36)] the amorphous samples were found to exhibit improved or better compared to their crystalline counterparts. The reason for this behaviour is in many cases not clear, since too little effort has been expended on characterising the chemical and structural properties of the amorphous and crystalline alloy surfaces. Several factors, such as degree of ordering and dispersion of the active component, electronic properties, formation of new phases, nucleation and growth of crystalline domains, segregation phenomena and textural properties, will be differently influenced during pretreatment, depending on whether an amorphous or a crystalline alloy is used as starting More recently, highly active catalysts were prepared from glassy metals by selectively oxidising the more electropositive constituent of the material.After reduction, finely dispersed transition-metal particles which are embedded in an amorphous or partially crystalline oxide matrix of the more electropositive constituent were obtained. In principle, this method for the preparation of supported metal catalysts from metal alloys is not new: it has been applied previously by Shamsi and Wallace6’ to crystalline intermetallic alloys.The use of amorphous metal alloys as precursors may offer several advantages, such as higher flexibility in composition, homogeneous distribution of constituents on a molecular scale and higher reactivity. These advantages emerge from the intrinsic properties of the materials outlined above. Next, we shall discuss some crucial factors influencing the structural and chemical properties of catalysts prepared from metallic glasses. catalytic behaviour, i. e. either higher activity 1-3,13,24-26,28.29,3 1,34,38,43,44.50-52,54-57 selectivity,23 13.1 8,25.39,42,43,57A. Baiker 245 Factors influencing Structural and Chemical Properties of Catalysts derived from Metallic Glasses Important properties of metallic glasses influencing the structural and chemical proper- ties of the catalyst derived from them are: (i) chemical composition; (ii) chemical and structural homogeneity; (iii) thermal stability and crystallisation behaviour; (iv) dissol- ution of gases; and (v) segregation phenomena.These factors together with the condi- tions used for the chemical transformation of the precursor are most crucial to obtain catalysts with the desired properties. Chemical Composition The chemical composition influences virtually all properties discussed subsequently and is therefore a controlling factor in the preparation of catalysts from metallic glasses. Note that the flexibility in the composition of metallic glasses is not as large as one would anticipate from the fact that thermodynamical constraints are less stringent for metastable solids.Chemical and Structural Homogeneity Ideally an amorphous metal alloy should be chemically and structurally isotropic. Chemical and structural anisotropies lead to non-uniform propagation of the solid-state reactions occurring during the transformation of the amorphous precursor to the active catalyst. Such inhomogeneities are frequently due to either too slow cooling rate,32 or surface contamination during exposure to air32959 (surface oxide layer). Thermal Stability and Crystallisation Behaviour The thermal stability is a severe limitation if the metallic glass is to be used in the as-quenched state for catalysis; however, that is not necessarily the case if the glassy alloy is used as a catalyst precursor. The thermal stability is mainly influenced by the chemical composition of the metallic glass and the medium to which it is exposed.It has been shown that the crystallisation temperature can be significantly lowered in the presence of a hydrogen a t m o ~ p h e r e ~ ’ ” ~ ~ ~ ~ ” ~ or an adsorbed organic compound.61 Metallic glasses have been found to crystallise by nucleation and growth processes. The driving force is the difference in free energy between the glass and the appropriate crystalline phase(s). Depending on the composition, crystallisation may occur by: ( i ) primary crystallisation, where one crystal phase with a composition different from the amorphous matrix is produced; (ii) polymorphic crystallisation, where one phase with the same composition as the glass is crystallised (occurs only in concentration ranges near the pure elements or compounds) or (iii) eutectic crystallisation, where two crystalline phases grow concomitantly by a discontinuous reaction.Most metallic glasses can crystallise by two or more different reactions. The route by which crystallisation occurs depends not only on the thermodynamic driving force (difference in free energy), but also on the kinetics of the possible routes. In the case of pretreatment of the metallic-glass precursor in reactive gas atmospheres, solid-gas-phase reactions are likely to influence the expected crystallisation behaviour. During rapid solidification, as well as during annealing treatments, surfaces are expected to catalyse nucleation as the crystalline phase replaces a portion of the surface, thus reducing the total energy required for nucleation.An important factor for crystallisa- tion is the oxygen content near the surface. Oxygen may stabilise a number of crystalline phases, thus increasing the driving force for crystallisation. Selective oxidation of one of the components, e.g. the metalloids at the surfaces of metal-metalloid glasses, is likely to result in excessive crystallisation of the metal (e.g. ~ ~ p p e r , ~ ~ * ~ ~ * ~ ~ palladium, 30.38246 Metallic Glasses in Heterogeneous Catalysis and iron44v53 in binary zirconium alloys). Selective oxidation is likely to exhibit the strongest influence on surface crystallisation. Even at temperatures far below any crystallisation event in the bulk glass, primary crystallisation of the transition metal has been observed in metal-metalloid glass.h2 It should also be noted that the crystallisation behaviour of melt-spun ribbons may be different on both ribbon sides.32 Nucleation for primary crystallisation of the transi- tion metals is observed to occur on both sides of the glassy ribbons, while other crystallisation reactions have been observed to prefer usually either the free surface or the contact side of the ribbon.6' This phenomenon may lead to different structural and chemical properties of the two ribbon sides, and consequently also to large anisotropy in the catalysts prepared from such Dissolution of Gases in Metallic Glasses The dissolution of gases is frequently different in metallic glasses than in their crystalline counterparts owing to the marked differences in the structural and electronic properties.As regards the metallic glasses, present knowledge concentrates mainly on the absorption of hydrogen. Since almost all catalysts prepared so far from metallic glass precursors were exposed to a hydrogen-containing atmosphere, either during activation or reaction, understanding the interaction and solid-state reactions induced by hydrogen seems to be crucial. Maeland et aLh3 have shown that the solubility (absorption capacity) of hydrogen in metallic glasses with the general formulae Ti,-,Cu, and Zr,-,Cu, ( x = 0.3-0.7) is larger than in corresponding crystalline alloys. Besides its possible direct influence on the catalytic properties of metallic glasses as a hydrogen source, the absorption of hydrogen generally enhances the formation of metal hydrides, which have been shown to be crucial intermediates in the preparation of catalysts from a r n o r p h o u ~ ~ ~ and crystalline64365 metal alloys.Unfortunately, no similar studies are presently available for the solution of other gases in metallic glasses. Segregation Phenomena The surface composition of metal alloys is often different from that of the bulk. Major driving forces for surface segregation revealed by model calculations" are: different surface free energies of the components and size mismatch in the case of clean surfaces, as well as different heats of chemisorption and reaction of components in the presence of adsorbates. Surface segregation induced by selective oxidation is well known for crystalline and amorphous alloys of the type A-B, where A is an early transition metal or rare-earth metal (e.g.Zr, Ti, lanthanides or actinides), and B a Group VIII (e.g. Ni, Fe, Pd) or Group IB metal (e.g. Cu, Au). Upon exposure of the alloy to oxygen, component A (the more electropositive element) is oxidised and is enriched at the ~ u r f a c e . ~ ~ * ~ ~ - ' ~ As a result of this, phase separation may occur, and the remaining atoms of component B cluster together and precipitate. The phase separation is crucial for the formation of oxide-supported metal particles. Similar segregation phenomena may also occur by adsorption or absorption of hydrogen. However, the enthalpies of hydride formation are much smaller than those of oxide formation.The results on hydrogen-induced surface segregation are rather controversial. The exclusion of oxygen traces in the bulk and surface of the alloys is crucial for proper studies of surface segregation induced by hydrogen. No significant segregation effect has been measured in crystalline Cu30Zr70 and Cu70Zr30,67 LaNi, , 7 ' Mn,Zr, Cr2Zr, V2Zr68 and amorphous Ni26Zr76, C U ~ ~ Z ~ ~ ~ and Fe24Zr76 ,72 whereas strong surface segregation was reported for amorphous Pd-Zr alloys after exposure to a hydrogen atmo~phere.~'Faraday Discuss. Chem. SOC., 1989, Vol. 87 Plate 1. High-resolution electron micrograph and electron diffraction pattern of catalyst ( a ) prepared from amorphnus Pd,,Zr,, . Note the extremely small domain sizes of the phases Pd and ZrO, (baddeleyite), which are also reflected by the partial absence of well defined reflection maxima.Evaluation of the electron diffraction patterns gives evidence for the presence of metallic palladium [ ( 1 11) reflection has highest intensity]. A. Baiker (Facing p. 247)A. Baiker 247 -7.0 h LL v c - -8.0 t 0 \ \ \ -9.0 I I I I I I I I I I I 2.8 3.0 3.2 3.4 3.6 lo3 K/T Fig. 1. Comparison of CO oxidation activities of palladium-on-zirconia catalysts prepared by in situ activation from amorphous Pd,,Zr,, ( a ) and by conventional impregnation of zirconia with a palladium salt ( b ) , respectively. Arrhenius plots of the turnover frequencies are plotted. Conditions were: reactant-gac mixture, 1700ppm CO, and 1700ppm O7 in nitrogen; flow rate, 150 cm' (s.t.p.) min-'; amount of catalyst, ( a ) 0.37 g; ( b ) 1.24 g.Examples illustrating the Potential of Metallic Glasses as Catalyst Precursors There are several examples reported in the literature which demonstrate the potential of metallic as catalyst precursors~'7,10.~5,~~,~~.4'),51.53.54,56.~9.74 For ill us tratio n we may consider the preparation of a palladium-on-zirconia catalyst'" for the oxidation of CO. The catalyst was prepared from amorphous Pd,Zr, by exposure to CO oxidation conditions at 550 K. Under these conditions the amorphous precursor was transformed to a catalyst containing well dispersed palladium particles embedded in a zirconium dioxide matrix. It is interesting to note that the metallic glass precursor was virtually inactive, but the activity developed during the in situ activation, finally reaching a steady state after the transformation was complete.The solid-state reactions occurring in the metallic glass during in situ activation resulted in a large increase in the B.E.T. surface area from 0.02 to 45.5 m2 g-'. The palladium metal surface area of the as-prepared catalyst determined by CO chemisorption was 6.9 m' g-', which corresponds to a palladium dispersion of ca. 6 % . Fig. 1 compares the intrinsic activity of palladium in the catalyst prepared from the metallic glass with the corresponding activity of palladium in a palladium-on-zirconium catalyst prepared by conventional impregnation (incipient wetness) of zirconium dioxide with a palladium salt [( NH4)'PdCl4]. Note the markedly higher turnover frequency measured for the PdlZrO, catalyst prepared from the metallic glass as compared to the conventionally prepared catalyst.The reason for this behaviour is not completely understood so far; however, there is clear indication that the enhanced activity of the palladium in the catalyst prepared from the metallic glass precursor has to be sought in the extremely large interfacial area (metal-metal oxide) of this catalyst. The structural features of the catalyst prepared from the metallic glass are illustrated by the high- resolution electron micrograph shown in plate 1. Note the small intergrown crystalline domains leading to a large interfacial area between palladium and zirconia. This structural feature is characteristic for as-prepared catalysts and is possibly not obtainable by conventional preparation techniques.I t has been found recently that such large248 Metallic Glasses in Heterogeneous Catalysis interfacial areas originating from small intergrown crystals may enhance the formation of solid solutions with hydrogen" and oxygen.75 Photoelectron spectros~opy~~ indicated that the electronic properties of the palladium in the Pd/Zr02 catalyst prepared from the metallic glass were similar to that of a pure palladium foil. This further supports the important role of the interfacial area between palladium and zirconia, particularly in view of the fact that CO oxidation over palladium was found to be structure-insensitive in several investigation^.^^ Conclusions Metallic glasses may be used in the as-quenched state or as catalyst precursors in heterogeneous catalysis.The motivation for using them in the as-quenched state is based on the ability to tailor the electronic properties and the special surface structure which ideally exhibits no long-range ordering of the constituents. The latter property seems, however, partly questionable, at least for surfaces of metallic glasses prepared by melt-spinning, as recent investigations using STM have indicated. Surfaces of metallic glasses already tend to undergo structural relaxation at temperatures far below the crystallisation temperature, and the presence of adsorbed reactants is likely to enhance this relaxation. Thus, in order to make use of the special surface structure of metallic glasses, catalytic reactions have to be performed at low temperatures, which is a severe limitation, in particular in the light of the intrinsically small surface area of these materials.At present, the use of metallic glasses as catalyst precursors appears to be more promising. Several efficient supported-metal catalysts have been prepared by exposing the precursor alloy to an oxygen-containing gas atmosphere at higher temperature. The aim of these pretreatments is to transform the alloy into a supported-metal catalyst by oxidizing the more electropositive component. Catalysts prepared as such have been shown to possess reasonably large B.E.T. surface areas and metal surface areas. A major difference in their structure compared to conventionally prepared supported-metal catalysts is that both the active metal as well as the oxidic support are made up of small disordered and/ or intergrown crystalline particles.As a result of this, the interfacial area between metal particles and support is extremely large, leading to very strong metal-support interactions. Such structures are for most systems difficult to prepare by conventional preparation methods. Thus, the preparation from amorphous alloys offers great potential. The solid-state reactions occurring during pretreatment of the metallic glasses are complex, and the structure of as-prepared catalysts is influenced by several factors determined by the intrinsic properties of the metallic glass and the conditions of pretreatment. Relevant intrinsic properties are: chemical composition, chemical and structural homogeneity, thermal stability and crystallisation behaviour, dissolution of gases and segregational phenomena.Better understanding of all these phenomena is a necessary prerequisite for progress in the use of metallic glasses as catalyst precursors. Although reports on the application of metallic glasses in electrocatalysis are still scarce, these materials seem to offer interesting properties for such applications, par- ticularly since large surface areas are not necessarily required in electrocatalysis. Many metallic glasses exhibit high corrosion resistance, and that coupled with the ability to form homogeneous solid solutions supersaturated with various elements should be beneficial in electrocatalysis. Metallic glasses have been used for electrocatalysis of sodium ~hloride,~' sea water7* and for the hydrogenation of carbon monoxide," the electro-oxidation of methanol8' and the electrocatalytic evolution of hydrogen and oxygen from alkaline so1ution.x'*x3 With regard to the preparation of metallic glasses, it is important that new methods are developed that enable the materials to be fabricated in a form better suited forA. Baiker 249 catalytic purposes.Promising routes include novel chemical technique^.^^-^^ The desired techniques should provide the possibility to deposit the amorphous metal alloys at high dispersion on a support material. Besides these more practical aspects, it should be pointed out that studies performed on metallic glass surfaces are likely to aid us in answering some long-standing funda- mental questions in catalysis.Metallic glasses were suggested to be ideal model systems' for the study of several catalytic problems, among them, the role of bimetals and multimetallics, the role of short-range ordering, the electronic and geometric structure of defects, the influence of' promotors, and surface segregation and clustering. Thanks are due to A. Reller (University of Zurich) for the high-resolution electron micrograph and to H. J. Giintherodt, P. Oelhafen and B. Walz for valuable discussions. Financial support of our work by Lonza AG and the Swiss National Science Foundation is gratefully acknowledged. References 1 H. Komiyama, A. Yokoyama, H. Inoue, T. Masumoto and H. Kimura, Sci Rep. Res. Inst. Tohoku Uniu. Ser. A, 1980, 28, 217. 2 G .V. Smith, W. E. Brower, M. S. Matyaszczyk and T. L. Pettit, in Proc. 7th Int. Congr. Caral., ed. T. Seiyama and K. Tanabe (Elsevier, New York, 1981), vol. A, pp. 355-363. 3 A. Yokoyama, H. Komiyama, H. Inoue, T. Masumoto and H. Kimura, J. Catal., 1981, 68, 355. 4 R. Schlogl, in Rapidly Quenched Metals, ed. S. Steeb and H. Warlimont (Elsevier, Amsterdam, 1985), 5 C. Yoon and D. L. Cocke, J. Non-Crysr. Solids, 1986, 79, 217. 6 M. Shibata and T. Masumoto, in Stud. Surf Sci. Catal. (Prep. Catal. I V ) , ed. B. Delmon, P. Grange, 7 F. E. Luborsky, Amorphous Metallic Alloys (Butterworths, London, 1983). 8 C. Suryanarayana, Rapidly Quenched Metals: A Bibliography, 1973- 1979 (IFI/Plenum, New York, 1980). 9 H. J . Giintherodt, in Rapidly Quenched Metals, ed. S. Steeb and H.Warlimont (Elsevier, Amsterdam, VOI. 11, pp. 1723-1727. P. A. Jacobs and G . Poncelet (Elsevier, Amsterdam, 1987), vol. 31, pp. 353-372. 1985) vol. 11, pp. 1591-1598. 10 R. B. Diegle, J. Non-Cryst. Solids, 1984, 61 & 62, 601. 1 1 G. A. Somorjai, Caral. Rev. Sci. Eng., 1978, 18, 173. 12 P. Oelhafen, in Glass?! Metals I I , ed. H. Beck and H. J. Guntherodt (Springer, Berlin, 1983), pp. 283-321. 13 W. E. Brower, M. S. Matyjaszczyk, T. L. Pettit and G. V. Smith Nature (London), 1983, 301, 497. 14 G . A. Somorjai, Adu. Catal., 1977, 26, 1. 15 B. C . Giessen, S. S. Mahmoud, D. A. Forssyth and M. Hediger, in Rapidly Solidified Amorphous and Crysralline Aflop, ed. B. H. Kear, B. C. Giessen and M. Cohen (Elsevier, New York, 1982), pp. 255-258. 16 G. V. Smith, 0. Zahraa, A.Molnar, M. M. Khan, B. Rihter and W. E. Brower, J. Catal., 1983, 83, 238. 17 A. Molnar, G. V. Smith and M. Bartok, J. Catal., 1986, 101, 540. 18 A. Molnar, G. V. Smith and M. Bartok, J. Caral., 1986, 101, 67. 19 R. Wiesendanger, M. Ringger, L. Rosenthaler, H. R. Hidber, P. Oelhafen, H. Rudin and H. J. Giintherodt, Sur$ Sci., 1987, 181, 46. 20 B. Walz, R. Wiesendanger, L. Rosenthaler, H. J. Giintherodt, M. Diiggelin and R. Guggenheim, Mater. Sci. Eng., 1988, 99, 501. 21 R. Schlogl, R. Wiesendanger and A. Baiker, J. Caral., 1987, 108, 452. 22 R. Hauert, P. Oelhafen, R. Schlogl and H. J. Guntherodt, Solid State Commun., 1985, 55, 583. 23 A. Baiker, H . Baris and R. Schlogl, J. Caral., 1987, 108, 467. 24 A. Yokoyama, H. Komiyama, H. Inoue, T. Masumoto and H.Kimura, Scr. Metall., 1981, 15, 365. 25 A. Yokoyama, H. Komiyama, H. Inoue, T. Masumoto and H. Kimura, ACS Symp. Ser., 1982, 196,237. 26 M. Peuckert and A. Baiker, J. Chem. SOC., Faruday Trans. 1, 1985, 81, 2797. 27 Y. Shimogaki, H. Komiyama, H. Inoue, T. Masumoto and H. Kimura, Chem. Lett. ( J p n ) , 1985, 661. 28 A. Yokoyama, H. Komiyama, H. Inoue, T. Masumoto and H. Kimura, Chem. Lett. (Jpn), 1983, 195. 29 A. Yokoyama, H. Komiyama, H. Inoue, T. Masumoto and H. Kimura, J. Non. Cryst. Solids, 1984, 61 30 M. Shibata, N. Kawata, T. Masumoto and H. Kimura, Chem. Lett. (Jpn), 1985, 1605. 31 G. Kisfaludi, K. Lazar, Z. Schay, 1 . Guczi, C. Fetzer, G. Konczos and A. Lovas, Appl. Sur$ Sci., 1985, 32 G. Kisfaludi, Z. Schay, L. Guczi, G. Konczos, L. Lovas and P. Kovacs, Appl.Swfi Sci., 1987, 28, 1 1 1 . 33 G. Kisf'dudi, Z. Schay and L. Guczi, Appl. Surf Sci., 1987, 29, 367. & 62, 619. 24, 225.250 Metallic Glasses in Heterogeneous Catalysis 34 H. Yamashita, M. Yoshikawa, T. Funabiki a n d S. Yoshida, J. Caral., 1986, 99, 375. 35 M. Shibata, Y. Ohbayashi, N. Kawata, T. Masumoto a n d K. Aoki, J. Catal., 1985, 96, 296. 36 S. J. Bryan, J . R. Jennings, S. J. Kipling, G . Owen, R. M. Lambert and R. M. Nix, Appl. Catal., 1988, 37 D. Gasser a n d A. Baiker, Appl. Catal., 1989, 48, 279. 38 A. Baiker a n d D. Gasser, J. Chem. Soc., Faradaji Trans. 1, 1989, 85, 999. 39 G . Carturan, G. Cocco, E. Baratter, G. Navazio a n d C. Antonione, J. Caral., 1984, 90, 178. 40 S. Yoshida, H. Yamashita, T. Funabiki and T. Yoezawa, J. Chem.Soc., Farada). Trans. 1 , 1984,80, 1435. 41 H. Yamashita, M. Yoshikawa, T. Funabiki a n d S. Yoshida, J . Chem. Soc., Faraday Trans. 1, 1985, 81, 2485. 42 J. C. Bertolini, J. Brissot, T. Le Mogne, H . Montes, Y. Calvayrac a n d J. Bigot, Appl. Surj Sci., 1987, 29, 29. 43 A. Baiker, H. Baris a n d H. J. Guntherodt, J. Chem. Soc., Chem. Commun., 1986, 930. 44 H. Yamashita, M. Yoshikawa, T. Funabiki a n d S. Yoshida S. Yoshida, J. Chem. Soc., Faraday Trans. 45 T. Takahashi, Y. Nishi, N. Otsuji, T. Kai, T. Masumoto a n d H. Kimura, Can. J. Chem. Eng., 1987, 65, 46 S. Yoshida, H. Yamashita, T. Funabiki and T. Yonezawa, J. Chem. Soc., Chem. Commun., 1982, 964. 47 H. Yamashita, M. Yoshikawa, T. Funabiki and S. Yoshida, J. Chem. Soc., Farads), Trans. I , 1986, 82, 1771.48 H. Yamashita, T. Kaminade, T. Funabiki a n d S. Yoshida, J. Muter. Sci. Lett., 1985, 4, 1241. 49 H. Yamashita, M. Yoshikawa, T. Funabiki a n d S. Yoshida, J. Chem. Soc., Faraday Trans. I , 1987, 83, 2883. 50 A. Baiker, H. Baris a n d H. J. Guntherodt, Appl. Catal., 1986, 22, 389. 51 H. Yamashita, M. Yoshikawa, T. Kaminade, T. Funabiki and S. Yoshida, J. Chem. Soc., Faradajl Trans. I , 1986, 82, 707. 52 S. S. Mahmoud, D. A. Forsyth, a n d B. C. Giessen, Marer. Rex Soc. Sjimp. Proc. (1986), vol. 58, pp. 53 A. Baiker, R. Schlogl, E. Armbruster a n d H. J. Guntherodt, J. Catal., 1987, 107, 221. 54 A. Armbruster, A. Baiker, H. J . Guntherodt, R. Schlogl a n d B. Walz, in Stud. Sur-f Sci. Catal. (Prep. Catal. I V ) , ed. B. Delmon, P. Grange P. A. Jacobs a n d G .Poncelet (Elsevier, Amsterdam, 1987), vol. 31, pp. 389-400. 40, 173. I , 1987, 83, 2895. 274. 131-138. 55 R. Lamprecht, PI?. D. Thesis (University of Basel, 1987). 56 A. Baiker, D. Gasser a n d J. Lenzner, J. Chem. Soc,., Chem. Commun., 1987, 1750. 57 H . Yamashita, T. Kaminade, M. Yoshikawa, T. Funabiki a n d S. Yoshida, C , Mol. Chem., 1986, I , 491. 58 G. Kisfaludi, K. Matusek, Z. Schay and L. Guczi, 6th Int. Svmp. Heterogeneous Catal. (Varna, Hungary, 1987). 59 A. Baiker, H. Baris, F. Vanini a n d M. Erbudak, in Proc. 9th int. Congr. Catal. (Cam/-vsis: Theory and Practice), ed. M. J. Phillips and M. Ternan (Chem. Inst. Canada, 19881, vol. 4, pp. 1928-1935. 60 A. Shamsi and W. E. Wallace, Ind. Eng. Chem. Prod. Res. Dev., 1983, 22, 582. 61 W. Kowbel a n d W. E.Brower, J . Catal., 1986. 101, 262. 62 U. Koster, Z. Metallkde., 1984, 75, 691. 63 A. J. Mealand, L. E. Tanner a n d G . G. Libowitz, J . Less-Common Met., 1980, 74, 279. 64 R. M. Nix, T. Rayment, R. M. Lambert, J . R. Jennings and G . Owen, J. Caral., 1987, 106, 216. 65 D. L. Cocke, M. S. Owens, and R. B. Wright Appl. Surf Sci., 1988, 31, 341. 66 A. R. Miedema, Z. Metall., 1978, 69, 455. 67 F. Vanini, S. Buchler, Xin-nan Yu, M. Erbudak, L. Schlapbach and A. Baiker, Surf Sci., 1987, 189/190, 11 17. 68 L. Schlapbach, Nato AS1 Ser. B, 1986, 136, 397. 69 F. Spit, K. Blok, E. Hendriks, G . Winkels, W. Turkenburg, J . W. Drijver a n d S. Radelaar, in Proc. 4th Int. Cor~f.. Rapidlj, Quenched Met., ed. T. Masumoto a n d K. Suzuki ( J p n Inst. Met., Sendai, 19821, pp. 1635-1640. 70 S. Shina, S. Badrinarayanan and A. P. Shina, J. Less-Common Metals, 1986, 125, 1179. 71 L. Schlapbach, A. Seiler, F. Stucki and H. C. Siegmann, J . Less-Common Metals, 1980, 73, 145. 72 S. M. Fries, H. G . Wagner, S. J. Campell, U. Gonser, N. Blaes a n d P. Steiner, J. Phys. F, 1985, 15, 1179. 73 P. Oelhafen, R. Lapka, U. Gubler, J. Krieg, A. DasGupta, H . J. Guntherodt, T. Mizoguchi, C. Hague, J. Kubler a n d S. R. Nagel, in Pro(*. 4th Int. Con/: Rapid!,. Quenched Met., ed. T. Masumoto and K. Suzuki ( J p n Inst. Met., Sendai, 19821, pp. 1259-1265. 74 Y. Shimogaki, H. Komiyama, H. Inoue. T. Masumoto a n d H . Kimura, J. C’hem. Eng. Jpn, 1988,21, 293. 75 A. Baiker, D. Gasser, A. Reller a n d R. Schlogl, in preparation. 76 T. Engel a n d G . Ertl, Adv. Catal., 1979, 28, 1; S. Ladas, H. Poppa a n d M. Boudart, Sut$ Sci., 1981, 77 M. Hara, K. Hashimoto a n d T. Masumoto, J. Appl. Electrochem., 1983, 13, 295. 78 N. Kumagai, A. Kawashima, K. Asami a n d K. Hashimoto, J . Appl. Electrochem., 1986, 16, 565. 79 M. Enyo, T. Yamaguchi, K. Kai and K. Suzuki, Electrochim. Acta, 1983, 28, 1573. 102, 151.A. Baiker 25 1 80 A. Yokoyama, H. Komiyama, H. Inoue, T. Masumoto and H. M. Kimura, Scr. Metall., 1981, 15, 365. 81 A. Kawashima, T. Kanda and K. Hashimoto, Muter. Sci. Eng., 1988, 99, 521. 82 G . Kreysa and B. Hakansson, in Dechema-Monographien (VCH, Weinheim, 1986), vol. 102, pp. 361-373. 83 L. Brossard, R. Schulz and J. Y. Huot, Znt. J. Hydrogen Energy, 1988, 13, 251. 84 J. van Wonterghem, S. Morup, C. 1. W. Koch, W. W. Charles and S. Wells, Nature (London), 1986, 85 J. van Wonterghem, S. Morup, S. W. Charles, S. Wells and J. Villadsen, fhys. Rev. Lett., 1985, 55, 410. 86 J-F. Deng and X-p. Zhang, Appl. Caral., 1988, 37, 339. 322, 622. Paper 8/04941 F; Received 16th December, 1988

ISSN:0301-7249    

DOI:10.1039/DC9898700239

出版商:RSC    

年代:1989

数据来源: RSC