摘要:

GENERAL, DISCUSSIONProf. A. Cimino (University of Perugia) said : Whilst I fully agree with the generalthermodynamic consideration presented by Kemball, I should like to comment on theexample quoted in his paper regarding the application to the ethane hydrogenolysis.He suggests that the large negative exponent, experimentally found for the hydrogenpressure dependence of the rate of hydrogenolysis, could arise from the presence of ahydrogen pressure term in the denominator of the expression for coverage. Theinclusion of such a term could indicate the deduction made by Taylor, Boudart andmyself, regarding the breaking of the C-C bond and the nature of the radicals on thesurface. I should like, however, to draw the attention to the comparison of theresults obtained on the same type of surface, nickel, for the hydrogenolysis of propane 1and of ethane.2 The large difference found in the hydrogen exponent, -2.6 and - 1.1respectively, suggests that it is the hydrocarbon itself which determines the hydrogenexponent.This conclusion can be understood in a simple way with the hypothesisput forward by Boudart, Taylor and myself. Moreover, the behaviour of differentsurfaces, such as iron plus 0-6 % (molar) Li20, and iron plus 0.6 % (molar) K20 3would also be rationalized. One can conclude that whilst the use of a complete setof pressure functions is certainly right, the simpler treatment presented in our paper isadequate, even though the hydrogen pressure term is neglected. I think therefore thatthe objection raised by Kemball does not in fact apply to the ethane hydrogenolysis.Prof.C. Kemball (Queen’s Uniuersity, Belfast) said: I am glad that there is asubstantial measure of agreement between the views expressed by Prof. Cimino andthose in my paper. We both agree that, in general, the kinetic expression for reactionslike the hydrogenolysis of ethane may have to include terms to allow for the presenceon the surface of species other than the one which leads to the rupture of the C-Cbond. Thus, if C2H, is the required species, other species such as C2Hv, CH, or evenH might be competing for the surface. Such competition is likely to be important ifthe adsorption of C2H, is endothermic, as suggested by Sinfelt et aZ.4We differ in that Prof. Cimino believes that we can ignore competition fromadsorbed hydrogen atoms in hydrogenolyses on iron and nickel and I am not convincedthat we can do so.I suggest that the higher negative hydrogen exponent of -2.6 forthe hydrogenolysis of propane on nickel compared with - 1.1 for ethane could becaused by the need to free more sites from hydrogen to accommodate the largermolecule. The explanation of these different exponents given by Cimino et aZ.5assumes that the value of x in the adsorbed species C3Hz required for the breakdown ofpropane is zero by analogy with ethane. This assumption requires further justification.Prof. R. L. Burwell (Northwestern University, Evanston) said: I should like toenter a small caveat with regard to the application of Dr. Bond’s scholarly paper tocatalytic reactions which proceed upon the surfaces of group 8 metals.Most of ourpresent knowledge of the structure of organometallic compounds of transition metalsrelates to mononuclear compounds. However, the surface of metals provides the idealopportunity for the formation of polynuclear complexes in which a monoadsorbedspecies is bonded to two or more surface atoms (e.g., a methyl group positioned1 Morikawa, Trenner and Taylor, J. Amer. Chem. SOC., 1937, 59, 1103.2 Kernball and Taylor, J. Amer. Chem. SOC., 1948,70, 345.3 Chino, Boudart and Taylor, J. Physic. Chem., 1954, 58, 796.4 Sinfelt, Taylor and Yates, J. Physic. Chem., 1965, 69, 95.5 Cimino, Boudart and Taylor, J. Physic. Chem., 1954,58,796.24250 GENERAL DISCUSSIONbetween four atoms in the (100) face of a f.c.c.metal and bonded to all four by partialbonds) and in which a diadsorbed species can be bonded to two different surfaceatoms or to two different sets of surface atoms. Now, the intermediates of hetero-geneous catalytic reactions may be bonded to just one atom. I suspect, however,that the opposite will commonly be true and that the relatively large rates of hetero-geneous reactions will result from the fact that the surface of a metal permits formationof a wide variety of surface complexes and, in particular, of those which are particularlysuitable to fast reactions.The one case in which we have detailed information about polynuclear complexessupports this word of caution. The complexes of acetylenes for which X-ray struc-tures are available are polynuclear.In fig. 1 1 (Bond's paper), acetylene is shown asn-complexed to two surface atoms by two sets of 0- and n-bonds at right-angles to oneanother. Such bonding requires the acetylene essentially to retain its original sphybridization and linear geometry. Two dinuclear complexes are known in whichthe geometry resembles this. Both have two cobalt atoms and three carbonyl groupsattached to each cobalt atom. The carbon-carbon unit of the original acetylene ispositioned as in fig. 11 but the angle of the substituent to the original C = C unit is138" for the complex of diphenyl acetylene 1 and about 120" for that of perfluorocyclo-hex-1- yne- 3- ene.2Similar conclusions derive from structures of polynuclear complexes.With fourcobalt atcms as in fig. l a with a dihedral angle between the two triangles of 118", thestructure resembles an olefin a-bonded to the two Co atoms and n-complexed to theouter two Co atoms as shown in fig. lb. The additional cobalt atom behind the planeof the paper is like that in front.3 Carboiiyl groups are not shown. Bonding of anacetylene to an equilateral triangle of iron atoms is similar.49 5R .RFIG. l(a) FIG. l(b)The bond strengths of the carbon-carbon and nitrogen-nitrogen shown in table 1 6are relevant.TABLE 1 .-BOND STRENGTHS IN kCd/mOletriple double singleN-N 226 100 39N-H 93 c-c 194 143 83C-H 99The replacement of a n-bond by two a-bonds to the surface should be favourable withcarbon and much less favourable with nitrogen even ignoring the question of variationof bond strength in *-C with hybridization.1 Sly, J.Amer. Chem. SOC., 1959, 81, 18.3 Dahl and Smith, .I. Amer. Chem. Soc., 1962, 84,2450.4 Blount, Dahl, Hoogzand and Hiibel, J. Amer. Chem. Soc., 1966,88,292.5 Dodge and Schomaker, J. Organometallic Chem., 1965, 3, 274.6 Cottrell, The Strength of Chemical Bonds (Butterworths, London, 2nd ed., 1958).2 ref. (12) of Bond's paperGENERAL DISCUSSION 25 1We must be prepared for many different types of binding of acetylenes to the surfacebut available evidence suggest that the adsorbed acetylene is best written as ana, /I-diadsorbed alkene in which the olefinic bond is possibly also n-complexed tosurface atoms. This matter is important here mainly in that it implies a geometry forthe adsorbed species.Molecular analogies are less definitive for adsorbed alkene since polynuclear analoguesare unavailable at present. However, even if Zeise’s salt should be better representedas involving olefin n-complexed at one coordination site in four coordinate platinum(€1) rather than as doubly a-bonded in five coordinate platinum (II), the conclusion thatthe same structure will be found on the surface of norrnally twelve coordinate platinum(0) is uncertain.This is particularly true in view of the probability that there is sub-stantial doubly a-bonded character in the complexes of acetylenes and olefins whichoccupy two coordination sites in octahedral [((C6H&P)2ClCO]Ir complexes.1Again, the main question is whether adsorbed olefin has a geometry closer to alkaneor to olefin.At present, prediction of the structures of surface complexes frommolecular analogues must be taken as stimulating and useful but not definitive.Dr. G. C. Bond (Johnson-Matthey Ltd., London) said: Prof. Burwell doubts whetherit is possible to apply information concerning the binding of olefins in, for example,Zeise’s salt to the problem of olefin chemisorption on metal surfaces because in theformer there is only one metal atom whereas in the latter there are many. There isnow no doubt that individual metal atoms have catalytic properties, and although ametal atom in an assembly may have properties which differ in kind as well as in degreefrom those of the isolated atom, it is nevertheless reasonable to examine a model inwhich a surface metal atom is regarded in isolation.I stated in the introduction tomy paper that it is most important to obtain more experimental evidence to establishthe extent to which this model is valid. Prof. Burwell is right to question whethern-adsorbed alkynes would have a linear structure ; I have never intended to be dog-matic on this point. We both agree that structural determinations on alkyne com-plexes are likely to assist our understanding of alkyne chemisorption, but a satisfactorymolecular orbital description of the complexes has not yet been achieved.Dr. G. Blyholder (University of Arkansas) said: I agree with Dr. Bond that amolecular orbital approach is useful in considering surface species.Prof. Bur wellhas questioned the appropriateness of applying molecular orbital ideas developed forsingle metal atom species to species on a metal surface, which contains a large numberof metal atoms. It therefore seenis appropriate to present two cases where a molecularorbital view explains or rationalizes observations of surface structures on metals.The first case is that of the variety of C--0 stretching frequencies observed for COadsorbed on Ni.2 The second case involves the structures observed when primaryalcohols are adsorbed on Fe and Ni. Briefly, the infra-red spectra of the surfacespecies indicate that on Fe the predominate surface species has an alkoxide structurewhile on Ni the alcohol decomposes to give chemisorbed CO and to a lesser extent anacyl structure. The general principles based on molecular orbital considerations forthe stability of individual organometallic compounds have been found sufficient toexplain the differences in the stable structures observed for the surface complexes onFe and Ni.3Dr.G. C. Bond (Johnson-Matthey Ltd., London) (communicated) : There is consider-able indirect evidence that the reactive state of olefins during their hydrogenation in a.n-adsorbed state,4 but there are possible reasons why the a-diadsorbed state should1 Cramer and Parshall, J . Amer. Chem. Soc., 1965, 87, 1392.2 J. Physic. Chem., 1964, 68, 2772.4 Jardine and McQuillin, J. Chem. SOC. (C), 1966, 458.3 J. Physic. Chem., 1966, in press252 GENERAL DISCUSSIONexist in hydrocarbon exchange processes.The former state is likely to be the weakerand only appear at high surface coverages by hydrocarbon, a situation not existingduring exchange reactions. Furthermore the ratio H(ads)/hydrocarbon (ads) is muchhigher in exchange then in hydrogenation, and hence the n-adsorbed state may not beviable during exchange processes. A possible approach to this apparent dilemmawould be through a detailed study of multiple exchange accompanying hydrogenation.We have observed in one instance with an iridium catalyst the hydrogenation of propy-lene leads to propane containing on average 3.8 D atoms per molecule,l and hence byappropriate choice of reactant and conditions it might be possible to combine thefeatures of both reactions in one system.Prof.G.-M. Schwab (Physik.-Chem. Inst. Univ., Munich) said: The hope that,besides the interesting viewpoints of Dr. Bond, the overall electronic factor is still afruitful line to be fullowed, can be demonstrated by our last results on doped supports.The activation energy of the oxidation of carbon munoxide over undoped nickel-oxidehas been measured by different authors as 16 kcal/mole. When nickel-oxide isdeposited in very thm layers on a silver carrier, this activation energy rises, beginningat some 500 up to values as high as 45 kcal/mole for a thickness of 100 A. At stilllower thicknesses, it decreases a little because of the incoherence of the layer. Theformation of the thicker layers has been controlled so that they are coherent.Appar-ently, electrons have been transferred from the silver into the boundary layer of thesemiconductor, resulting in a bending down of the electron bands. This brings theFermi level near the conduction band, and the effect is the same as if the thin layerwere doped. This effect of a metallic carrier can further be proved by other experi-ments in which the silver support has in turn been doped producing an increase ordecrease of the valence electron concentration, and this doping had a considerableeffect on a supported semiconductor layer as catalyst.Dr. L. L. van Reijen (Amsterdam) said : I would ask Prof. Schwab to what extentthe effect of the silver layer on the activity of NiO for the oxidation of CO is alsoreflected in the rate of the reaction.The reason for this question is that in the mostelementary description of the electronic effect the rate of the reaction should beproportional to the concentration of electrons or holes immediately below the catalystsurface. This concentration depends on the position of the Fermi level according to arelation : c = N exp (El - E F ) / ~ T . Here N and El are constants characteristic ofN O and independent of the presence of the silver layer. In this way the electroniceffect should consist in a change in energy of activation and no corresponding change inthe frequency factor of the reaction. In other words, the change in energy of activa-tion should be reflected in the rate of the reaction without compensation effects.Dr, G. Ertl (Inst.Techn. Hochschule, Munchen) (communicated) : I agree with Dr.van Reijen, that the reaction velocity should depend on the factor exp (El -EF)/~T.But generally the Fermi level EF is not independent on the temperature.2 At very lowtemperatures it coincides with the energy of the valence band EV (p-type) or conductingband EC (n-type), at very high temperatures the semiconductor becomes intrinsic andthe position of EF is nearly in the middle of the band gap. At moderate temperatures(which is the interesting region) EF varies approximately linear with temperature :EF WEV + aT (for p-type semiconductor), or EF ~ E c -PT (for n-type semiconductors).Only the factors a and p depend on the doping. The reaction velocity is thereforeproportional exp (El - EV)/kT.exp (a/k) forp-type and exp (El - EL)/kT . exp (-p/k)for n-type materials. For a given semiconductor, therefore, the activation energy isconstant for all p-type samples and has another constant value for all n-type samples.1 Bond, Phillipson, Wells and Winterbottom, Trans. Fmuday Soc., 1964, 60, 1847.2 see, e.g., Azaroff and Brophy, Electronic Processes in Materials, (New York, 1963), p. 208GENERAL DISCUSSION 253The frequency factor changes with exp (a/k) or exp (-/3/k) and therefore with dopingor band bending. Such a behaviour has been found for example by Roginskij et aZ.1for the decomposition of ethanol over germanium with different dopings.Prof. Dr. G.-M. Schwab (Physik.-Chem. Inst. Univ., Munich) said: The reactionrates observed with nickel oxide films of different thickness on a silver support allobey the compensation effect well.However, the temperature 8 of common velocitylies well above the interval of measurements. Hence, the reaction velocities increasewith decreasing activation energy as expected by Dr. van Reijen. The change infrequency factor present here, as in all other similar cases, must be due to a moregeneral reason.Dr. G. C . Bond (Johnson-Matthey Ltd., London) said: It is universally found thatcopper catalysts are much less active than those containing group 8 metals for thehydrogenation of mono-olefins. Under conditions where the reactants are adequatelyadsorbed (i.e., at low temperatures), the magnitude of the activation energy preventsrapid reaction : where the adsorbed molecules are sufficiently activated (i.e., at hightemperature) their concentration has fallen and a low rate still obtains.The highselectivity for mono-olefin formation observed in the hydrogenation of alkynes anddienes is due to the fact that the mono-alefin is very much less strongly adsorbed than itsprecursors.Ir. A. Rozendaal (Unilever Res. Lab., Vlaardingen, Netherlands) said : In reply to Dr.Bond, in the literature many discrepancies exist regarding the adsorptive properties ofcopper. For instance, Dr. Pritchard argued in this Discussion and elsewhere 4 that-contrary to the view of many other investigators 2s3,5y6-copper dissociatively adsorbshydrogen at room temperature.In connection with the different views held, I thoughtDr. Bond’s statement that mono-olefins are not adsorbed on copper, might lead toconfusion and recalled the experiments of Trapnell2 and Allen and Mitchell.3 ButZ agree with him that under reaction conditions mono-olefins may be much lessadsorbed than higher unsaturated hydrocarbons.Dr. P. Cossee (Amsterdam) said: I agree with Bond when he states that thedetailed description on an atomic scale for adsorption on metal surfaces is auseful approach and that anything we may learn by comparison with homogeneouscomplexes may help us greatly in heterogeneous catalysis. The question whether weshould use localized orbitals or orbitals involved in energy bands is less important inthis qualitative picture.The major advance of Bond’s treatment is that local sym-metry is observed and adequately taken into account.A qualitative molecular orbital picture of the binding between the n-ally1 groupand transition metals will be published. There it is argued that an allyl group interactswith the metal either with one or with three carbon-atoms and thus forms a 0- 01 an-bonded allyl group, respectively. The latter may be more or less asymmetricallyplaced, which may be reflected in its n.m.r. spectrum, but it remains a n-bonded group.I therefore support the reluctance expressed in Bond’s paper to accept the notion ofan allyl group which is an-coordinated to one metal, as has been proposed by others.7Is it not surprising that some correlation is found between affinity for complexingolefins and particular catalytic reactions, e.g., the hydrogenation of olefins ?1 Frolov, Krylov and Roginskij, Dokl. Akad.Nauk. S.S.S.R., 1959, 126, 107.2 Trapnell, Proc. Roy. SOC A , 1953, 218, 566.3 Allen and Mitchell, Disc. Furuday Soc., 1950, 8, 357.4 Pritchard, Nature, 1965, 205, 1208.5 Beeck, Smith and Wheeler, Proc. Roy. Soc. A , 1940, 177, 62.6 Kington and Holrnes, Trans. Faraduy Soc., 1953, 49, 417.7 Powell, Robinson and Shaw, Chem. Comm., 1965,78254 GENERAL DISCUSSIONIt is not obvious that of the three steps involved in a hydrogenation :H HI+ M-IIIM+II -H4 M4H2-CH3IM-I1 ~complex formationrearrangementM-CH2-CH3 + I32 +M-H + C2H6(1) should be the rate-determining step. It would seem that (2) or (3) are at least aslikely to become the limiting factor.In that case a correlation with characteristics ofmetal to carbon a-bonds would be expected.Ir A. Rozendaal (Urzilever Res. Lab., Vlaardingen, Netherlands) said : Dr. Bondstates in his paper that, over copper catalysts, diolefins and alkynes are selectivelyhydrogenated to mono-olefins and ascribes this selectivity to the inability of coppercatalysts to adsorb the mono-olefins. However, Trapnell 1 and Allen and Mitchell 2found that copper does chemisorb ethylene. Trapnell 1 deduced this from the rela-tively high heat of adsorption of ethylene on copper (18.2 kcal/rnole), whereas Allenand Mitchell 2 found an increase in the electric resistance when ethylene was admittedto a copper film. How can these measurements be reconciled with Bond’s statementthat the selectivity of copper catalysts in the hydrogenation of diolefins and alkynes isdue to non-adsorption of the mono-olefin?Dr.G. C. Bond (Johnson-Matthey Ltd., London) said: If the approach I haveadvocated in my paper has any merit, it lies in its ability to predict the location ofdifferent types of adsorbed state. I was unable fully to develop all the implications ofthe molecular orbital approach in my paper, but this point may be briefly illustrated intwo ways. It has long been thought possible (see Eley and Norton, this Discussion)that on metal surfaces strongly bound atomic (type A) hydrogen is bound by means ofemergent hybrid dsp orbitals (which correspond to my t Z g orbitals) and the moreweakly bound and perhaps molecular (type C) hydrogen by atomic d orbitals (whichcorrespond to my e, orbitals).The diagrams presented in my paper show the direc-tion of emergence of the two kinds of orbital and hence we have the opportunity tospecify more closely than heretofore the geometric detail of mechanisms involvingthese species. Furthermore adjacent interatomic sites on the (1 11) plane of a face-centred cubic metal, although geometrically equivalent, are not orbitally equivalent :this realization lends some weight to the old idea of Sachtler and Mignolet that weakhydrogen chemisorption on this plane might involve sites not stoichiometricallypermissible for strongly bound atoms.I agree with Dr. Cossee that cis-ligand transfer is likely to prove an importantgeneral mechanism in homogeneous catalysis and that it has no obvious counterpartin heterogeneous mechanisms.However, I have the impression (unconfirmed byprecise evaluation) that homogeneous hydrogenation reactions are somewhat slowerthan corresponding heterogeneous reactions, and, although we cannot go further untilrate-determining steps in the former have been definitely assigned, the possibilityexists that hydrogen atom transfer steps in the latter are even more efficient than cis-ligand transfer.Prof. N. Sheppard (Norwich) said: In his paper Dr. Bond has raised the importantquestion, implicit in the title of this Discussion, of to what extent chemisorption is1 Trapnell, Proc. Roy.SOC. A, 1953, 218, 566.2 Allen and Mitchell, Disc. Faraday SOC., 1950, 8, 357GENERAL DISCUSSION 255relevant to catalysis. It has always to be borne in mind that equilibrium studies ofchemisorption (including the infra-red spectroscopic methods with which I am prim-arily concerned) will provide neaningful i-dormation about the surface species, butthese may be particularly stable ones which are not necessarily the reaction inter-mediates involved in catalytic processes. Therefore if a kinetic study of a particularcatalytic reaction leads to the postulation of a reaction intermediate, there is notnecessarily a contradiction if an equilibrium chemisorption study leads to the identifica-tion of a different surface species. However, this is not to say that chemisorptionstudies can tell us nothing about catalysis, for when a sufficiency of experimentalinformation has been obtained on related systems it is to be expected that the reactivesurface species on one surface (say on one metal) may bc the stable species on anotherrelated one.With the above reservations in mind, i wish to make some comments on the extentto which infra-red spectra of ethylene chemisorbed on silica-supported metals provideevidence for x-bonded surface ethylenic species.The starting point here must be theinfra-red spectra of authentic metal coordination compounds containing n-bondedethylene. The best known molecule of this type is Zeise’s salt K[PtC13(C2H4)]E-I20and Dr. D. B. Powell and 1 1 made a detailed analysis of the i.-r.spectrum of thiscompound with this point in mind. We concluded that the spectrum stronglysupported a perturbed near-planar ethylene structure of the type postulated by Chattand Duncanson2 on theoretical grounds. The spectrum showed G-H bond-stretching frequencies (little different from those of ethylene) between 3 100 and 2950cm-1 and a C=C stretching frequency near 1520 cm-1.Dr. D. J. C. Yates has recently sent me a report of work by himself and his col-leagues of the Esso Laboratories which has now been published 3 in which .theyinvestigated the infra-red spectra of ethylene adsorbed on the (presumably isolated)metal ions in zeolites. Adsorption on different metal ions led to remarkably differentintensities of the C-H and C=C bands, but the frequency patterns observed werevery similar to those found with co-ordination compounds of the type of Zeise’s salt.This shows that n-bonded surface species of the type envisaged by Dr.Bond can occurin chemisorption systems.On the other hand, infra-red spectra that I and my colleagues have obtained forethylene chemisorbed on the silica-supported metals platinum, palladium and nickel(see my discussion comment on Dr. Peri’s paper) are of a quite different type. Inparticular all the strong G--H bands occur below 2950 cm-1, and there is no sign ofthe characteristic bands of C=C& groups that would be expected for n-bondedcomplexes. It is possible that such bands are present with very weak intrinsicintensities, or that these species occur in only very small amounts on all three metals.However, at this stage we are forced to the conclusion that the z-complex hypothesis isan unnecessary one for the interpretation of the infra-red spectra of ethylene on thesemetals.The spectra observed are much more readily interpreted in terms of theclassical cr-bonded surface species.Dr. G. C. Bond (Johnson-Matthey, Ltd., London) said: I cannot accept Prof.Sheppard’s failure to detect x-adsorbed olefin by infra-red spectroscopy as conclusiveevidence that it does not exist. It would undoubtedly be a weakly-held state andmight perhaps only exist when both hydrogen and the olefin are present in the gasphase : an important role of the former in hydrogenation catalysis is to prevent theformation on the surface of strongly-held hydrocarbon species, which might block the1 Powell and Sheppard, Spectruchirn.Acta, 1958, 13, 69.2 Chatt and Duncanson, J . Chem. Suc., 1953, 2939.3 Carter, Yates, Lucchesi, Elliott and Kevorkian, J. Physic. Chem., 1966, 70, 1126256 GENERAL DISCUSSIONformation of any weaker state. The infra-red examination of adsorbed hydrocarbonspecies still does not appear to have been performed under reaction conditions. Bythe use of a flow system in a double-beam instrument, one cell of which contained thecatalyst, it should be possible to cancel out gas-phase absorption and hence to recordthe spectra of surface species while the reaction proceeds.Dr. G.-M. Sckwab (Physik.-Gem. Inst., Univ., Mztniclz) said: I think there is noessential contradiction between the electronic factor and the notion of orbitals emerg-ing from the surface.In materials poor in electrons, i.e., metals with half-filled zonesand p-type semiconductors, a deficiency in electrons is present and the orbitals emerg-ing from the surface will, at least in part,be unoccupied. Such orbitals are in a positionto form covalent bonds with molecules donating electrons, like H2 or CO. On theother hand, materials having many mobile electrons, e.g., metals with more than half-filled conductivity zones or n-type semiconductors, will have these orbitals mostlyoccupied and will be able to form covalent bonds principally with molecules thatpossess considerable electron affinity. Thus, the view point of the electronic factormay be valid also for localized surface bonds.Prof.D. D. Eley (Nottingham Uniuersity) said: Schwab has asked how band theorywill be reconciled with surface orbital theory. Surely this will be through the theoryof surface states, mentioned by Lennard-Jones 1 in his original paper on adsorption,and developed by Tam,2 Shockley3 and others, especially with regard to semi-conductors.Dr. D. A. Dowden (I.C.I. Ltd., Billingham) said: There is little real differencebetween the “ old ” and the “ new ” electronic theories of heterogeneous catalysisexcept in emphasis. The first notions were concerned with simple correlations betweenthe catalytic activity of solids and the electronic structure of the solids as interpreted bysimple band-theory . The inadequacies of simple band-theory were already evidentin 1950 ; it was realized that recourse must ultimately be had to the physical propertiesof the solid itself rather than to their shadowy representation in a theory.The bondsin chemisorption were taken to be of electron-transfer or covalent type and thedependence of bond energy on the electronic structure of the solid was described,intuitively, in terms of band-energy and level-density.Friedel described in 1952-1954, for a simple metal, the rather similar conditionswhich lead in dilute alloys to shielding of the foreign metal atom by the piling-up ordispersing of electrons in the conduction band (compare Comer’s paper) or by thesplitting-off of a localized bound level from the bottom of the band. The importantparameters were the valency difference between solute and solvent, the ionizationpotential and electron affinity of the solute atom, the energies of the top and bottomof the conductivity band and the density-in-energy of the electron levels at the Fermisurface and their gradient with energy.In one model of chemisorption, parts of the unsaturated orbitals of surface atoms,projecting from the surface, but also bonding with interior atoms, were allowed toform molecular orbitals with the orbitals (of appropriate symmetry) of adsorbedspecies.Then, because the ionization potential of most adsorbates lies close to orbelow the bottom of the d-band in transitional metals, the metallic orbitals are fulland bond formation can only result after elevation of an electron to the Fermi surfaceor to an anti-bonding level.This displacement controls the change in saturationmagnetization, the surface potential, in some measure the conductivity, etc. Thesymmetry of the orbitals projecting from the surface suggests the type of bond whichmay form (sigma, pi, delta, etc.), but to estimate their energy and to decide whether1 Lennard-Jones, Frans. Faraday SOC., 1932, 28, 333.2 Tamm, Phys. Zeit. Soviet Union, 1932, 1, 933. 3 Shockley, Physic. Rev., 1939, 56, 317GENERAL DISCUSSION 257they will in fact form is another matter. The disparity between “ old ” and “ new ”is more apparent than real and is over-emphasized by the lack of knowledge aboutmetals, the stress on simple correlations and the appraisal of the problem of chemi-sorption from too restricted a view point.Dr.J. J. Rooney (Queen’s University, Belfast) said : The initial exchange of trans-1,1,3,4-tetramethyl-cyclopentane with deuterium on palladium at 80°C 1 gave thefollowing distribution of deuteroisomersdi dz d3 d4 d5 d6 d7 ds dio dii dl2 dl38.1 1.8 1.1 1.8 6.6 0.2 0.4 0.5 0.8 1.1 9-6 66.9 1-1Although isomerization also occurred the predominant product was the trans-isomer.11‘ the mechanism in fig. 3 of Burwell and Schrage’s paper is responsible, “ roll-over ” of either the 1,2- or 3,4-diadsorbed species, followed by “ roll-over ’’ of the2,3-diadsorbed species, and finally “ roll-over ” of the 1,2- or 3,4-diadsorbed speciesare required in order to obtain the trans-&-isomer initially. In view of the con-siderable steric hindrance to “ roll-over ” of any of these intermediates, and especiallyof the 2,3-diadsorbed species, it seems most unlikely that this mechanism could beresponsible for the high yield of the trans-dlz-isonier.A maximum in the dlo-isomerwould also be expected and this was not found. The steric hinderance due to themethyl groups on carbon atoms 2 and 3 is even more severe if the “roll-over”mechanism of fig. 5 in paper 14 is considered.An alternative possibility to the “ roll-over ” mechanism (their fig. 3) as an explana-tion for the small maximum in the &-isomer in the initial distribution for the exchangeof cyclopentane with deuterium on palladium is desorption and readsorption of someintermediate olefin. If a molecule takes this reaction path once the maximum numberof hydrogen atoms which may be replaced initially by the ap process is eight.An objection to Siegel’s modification of the n-olefin-n-ally1 mechanism is that n-and a-bonded butenyl do not interconvert readily during hydrogenation of buta-l,3-diene on palladium, otherwise highly selective formation of trans-but-2-ene would notbe obtained.There is therefore no reason to believe that a similar interconversion isimportant during the exchange of paraffins with deuterium on this metal at the sametemperatures.Dr. P. B. Wells (Hull Uniuersity) said : Prof. Burwell has referred to an “ organo-metallic zoo ”. I do not wish to use this nomenclature unfairly, but a characteristicof animal evolution is that species resemble their predecessors.The exchange ofbicyclononane with deuterium is a nicely conceived experiment and, if the assumptionscontained in the paper are accepted, the results support structure (I) and rejectstructure (11) as the mode of adsorption of bridgehead bicyclononene. I would suggesthowever, that structures1 Gault, Rooney and Kemball, J. Catalysis, 1962, 1, 255258 GENERAL DISCUSSION(I) and (11) are equally acceptable, and that the actual structure formed may dependupon the nature of the precursor. In the paper by Joke et al. we have written adsorbedbutene, which is formed by hydrogen atom addition to adsorbed 1-methyl-n-allyl, ina manner analogous to structure (11).Adsorbed butene was written in this waybecause such a reaction does not involve rehybridization of the two relevant carbonatoms, and this seemed to us more reasonable from a chemical standpoint, than theformation of di-a-adsorbed butene. However, in paraffin exchange, adsorbed olefinis formed from a a-bonded alkyl group and similar chemical intuition suggests thatdi-a-bonded olefin would be formed, at least initially, because no hybridizationchange would be required.Secondly, Prof. Burwell rejects structure (11) with the phrase “ bridgehead bicyclo-nonene . . . is of much too high energy to serve as an intermediate when n-complexed.”What evidence is there for this statement ? The stabilities of metal-olefine and metal-acetylene complexes do not always run parallel to the stabilities of the hydrocarbonligands ; indeed, some stable complexes are known in which the ligand is too unstableto exist in the free state.Thus, I would like to ask : What criteria does he adopt todecide the stability or otherwise of adsorbed species in the event of their being 71-bonded (as in structure (II))?Prof. R. L. Burwell (Northwestern University, Evanston) (communicated) : Wellsenquires about the criteria upon which we reject a n-complex of bridgehead bicyclo-nonene. First, formation of bridgehead diadsorbed bicyclononane should involvelittle more difficulty than formation of, say, diadsorbed cyclooctane. Secondly, ifincreased strain favoured formation of a n-complex by more than the strain energy,then n-complexed bridgehead bicyclo[2,2,1 Jheptene or n-complexed bridgeheadadamantene should readily form even though the bridgehead olefins might be unstablealone. In fact, the exchange patterns reveal no sign of such species.From models,bridgehead bicyclononene should be substantially strained. Therefore, by interpola-tion between adarnantene and bicycloheptane at one extreme and cyclo-octane at theother, we concluded that eclipsed diadsorbed alkane is preferred.This argument lacks the complete force of a geometric theorem. Dr. JamesMarshall of Northwestern is attempting to synthesize the bridgehead bicyclononene.If he is successful, we shall obtain its heat of hydrogenation and, thus, the strainenergy. This would make our argument more quantitative.Prof. C.Kernball (Queen’s Universify, Belfast) said : Would Prof. Burwell agreethat the mechanisms shown in fig. 3 and fig. 5 of his paper are sufficiently different tojustify calling them by different names to avoid confusion? Since the one in fig. 3 isdesignated “ roll-over ” that in fig. 5 might well be called “ rock and roll ”. Dr.Rooney suggests that the small maximum at dg in the exchange of cyclopentane couldresult from desorption and re-adsorption of cyclopentene. Has he looked into thethermodynamics of the dehydrogenation to see whether it is likely that sufficientolefin could desorb to make a significant contribution to the exchange reaction ?Dr. J. J. Rooney (Queen’s University, Belfast) said: In reply to Kernball, thecontribution which desorption and readsorption of olefin may make to the initialexchange of cyclopentane does not depend on the equilibrium constant for thereaction, C$€10+C$38+H2. The key factor is the rate of dissociation of metal-olefin complexes relative to the rate of paraffin desorption.Olefin once formed willimmediately chemisorbed again in preference to paraffin so there may be little escapefrom the van der Waals layer. I agree that a small fraction of the surface complexesturnover so the point at issue is the degree of residual bonding to the suiface duringthis event. In the symmetrical intermediate which Prof. Burwell has drawn, the twohydrogen atoms which are not bonded to the surface are -1 A closer to the surfacGENERAL DISCUSSION 259than the carbon atoms which are bonded.I find it difficult to believe that these carbonatoms could be close enough to the surface in such a situation to allow anything butthe weakest bonding which would be indistinguishable from physical interaction.6-eH’i I I‘H II I SProf. R. L. Burwell (Northwestern University, Evanston) (communicated) : Rooneyraises a poir,t about steric hindrance in the roll-over mechanism. The problem appearsin many proposed mechanisms. To generalize, the problem of steric interference ariseswhenever a species adsorbed on a substantially densely packed plane contains a planaror linear array of atoms some of which are bonded to surface atoms and some not.This may forbid the formation of surface analogies to existing molecular complexeswhich involve one or a small cluster of metal atoms.In fig.3 and 5 the two carbon atoms which are bonded to the surface and the twoatoms or radicals attached thereto lie in a straight line. Steric hinderance wouldpresumably be greater in fig. 5 since a methyl or higher alkyl group replaces one of thehydrogen atoms of fig. 3. The exchange pattern of trans-l,1,3,4-tetramethylcyclo-pentane is interesting but the problem of steric hindrance fully appears in fig. 5 as weapply it to the epimerization of 1,2-dimethylcyclopentane.The objections advanced by Rooney to the mechanisms of fig. 3 and 5 inevitablyextend to that of fig. 4. All atoms attached to the three carbon atoms of the allylicgroup must lie in the same plane as the three carbon atoms.When the allylic functionis part of a cyclopentane ring, steric interference would be even worse than in fig. 3owing to the hydrogen atoms which project below the plane of the cyclopentane ring.As applied to the tetramethylcyclopentane, steric hinderance would be at least asserious as in fig. 5. Reaction (6) of fig. 4 which would require a methyl group to moveeven closer to the surface presents formidable difficulties. Similar problems of sterichindrance arise in formation of n-complexes between surface atoms and unsaturatedmolecules such as trans-2-butene, 2-butyne and toluene.Steric interaction between substrate and surface must be a real phenomenon sincethe stereochemistry of many hydrogenation reactions clearly depends upon it. Asimple example is the formation of cis- rather than trans-2-butene in the hydrogena-tion of methylallene.1On the other hand, many reactions which would seem to involve serious sterichindrance proceed at substantial rates.As an extreme example, both di-tert-butylacetylene and trans-di-tert-butylethylene hydrogenate on metallic catalysts atroom temperatures.2 Disubstituted norbornenes (fig. 6a) also hydrogenate readilyenough.3 Recently, Klabunovskii and Balandin have reported the hydrogenation ofa series of tripticenes in which large steric hindrance should be present in adsorption ondensely packed planes.4 Thus, argument based on steric hindrance is conclusive onlywith respect to densely packed planes. Missing atoms in such planes may eliminate1 Meyer and Burwell, J.Amer. Chem. Sac., 1963, 85, 2881.2 Puterbaugh and Newman, J. Amer. Chem. Sac., 1959, 81, 1611.3 See Table I, Burwell, Chern. Rev., 1957, 57, 902.4 Klabunovskil, Balandin and Mamedzade, DokEady Physic. Chern., 1965, 162, 426, is a leadingreference260 GENERAL DISCUSSIONsteric interaction. Klabunovskii and Balandin have suggested that reaction mayoccur at protuberances. Alternate rows on (1 10) faces of f.c.c. metals and edges mayfunction in this fashion as mentioned in our paper and shown in fig. 6b. We need toknow more about the detailed geometry of surfaces before questions of this nature canbe settled. However, species such as di-tert-butylacetylene and ethylene might beexcellent probes for the structure of surfaces since isomerization cannot complicatethe picture.Rw HR(6)FIG.6(a).-Disubstituted norbornene.FIG. 6(b).-Roll-over mechanism on an edge.a (4As an alternative to the roll-over mechanism which we suggest as one possibleintermediate for the formation of C S H ~ D ~ in the exchange of cyclopentane, Rooneysuggests that cyclopentene might desorb and then readsorb. At low concentrationsof cyclopentene, the activation energy for its formation must at least equal the heatof hydrogenation of cyclopentene, 27 kcal/mole. If there is any activation energy inthe adsorption of cyclopentene, this figure must be increased. The observed activa-tion energy for formation of CsH2D8 is 20 kcal. Thus, cyclopentene cannot functionas a gas-phase intermediate.If the intermediate is cyclopentene held perpendicularto the surface by van der Waals forces, the heat of adsorption must be at least 7 kcal.If the heat of adsorption is this large, the difference between our fig. and Rooney’sintermediate becomes small. Finally, the sites which lead to C5H2Ds constitute aset separate from those which lead to other products.Prof. D. A. Cadenhead (State University of New York at Bufalo) said: In additionto considering possible differences between surface and bulk composition in the palla-dium-gold alloys, Dr. Rooney should give more consideration to the role of hydrogenin his studies of the hydrogenation of buta 1,3-diene on these alloys. From workcarried out at Buffalo 1 and from other sources,2 it is evident that these catalystsshould be regarded as ternary palladium-gold-hydrogen systems in which the hydrogen,as well as the gold, plays a role as an electron donor.Care should be taken in drawingany conclusion with regard to changes in the activation energy in the 60-65 atom %gold compositional region. Even though the solubility of hydrogen is low in thisregion, it is still possible that the critical composition of d-band filling could be achievedprior to 40 atom % gold. What is required here is a paramagnetic characterization ofthe specific alloys used for these studies. Moreover, as Dr. Rooney has pointed outhimself, treatment of this catalytic behaviour in terms of d-band theory is somewhatover-simplified and a move towards a more atomic approach is required.Dr.J. J. Rooney (Queen’s University, Belfast), said: I agree that differencesbetween surface and bulk composition in the palladium-gold alloys may exist but we1 Masse, Ph.D. Thesis (State University of New York at Buffalo, February, 1966).2 Dickens, Linnett and Palczewska, J . Catalysis, 1965, 4, 140GENERAL DISCUSSION 26 1have no evidence about this point. I also agree with Prof. Cadenhead that dissolvedhydrogen in the palladium-gold alloys may exert an important influence on the kineticsof diene hydrogenation, but we were reluctant to speculate on this as we think thatany explanation for the increase of activation energy in the 60-65 atom % gold regionmust also account for the equally sharp decrease of activation energy in the 70-75atom % gold region.However, there was no detectable formation of a bulk hydridewith the 70 % alloy although an appreciable quantity of hydrogen dissolved in the65 % alloy, On the other hand, the variations in product yields with increasingtemperature indicated that hydrogen " absorbed " in the surface layers had someinfluence on the reaction up to, and including, the 75 atom % gold alloy. If this typeof hydrogen is responsible for the observed changes in kinetics in the 60-75 atom %gold region this would mean that the filling of the d-band by the gold electrons has nodirect significance by itself. The important factor would be the presence of a dia-magnetic palladium hydride in the surfaces of the alloys in a narrow range of composi-tion just where the formation of a bulk hydride becomes impossible.This explanationhas the advantage that postulation of superlattice formation with the compositionPdAu2 can be avoided.Prof. W. M. H. Sachtler (Amsterdam) said: The catalytic activity of Au-Pdalloys, as reported by Rooney et al., shows a dependence on alloy composition. I amnot sure, however, whether the rise in activation energy by 4.5 kcal for the alloyscontaining 65 % of gold should be attributed to a specific effect of holes in the d-band,since this concept leaves unexplained that (i) the activation energy falls by 5-7 kcal forthe alloy containing 95 % of gold, where the holes in the d-band should still be filled,and (ii) the rate of the catalytic reaction is lower for the gold-rich alloys with lowactivation energy than for the palladium-rich alloys with higher activation energy.An " overcompensation effect " does not seem to fit in with the simple electronic factorpicture .I wonder whether these results could be rationalized by assuming that some of thealloys were twa-phase systems.This was found to be the case for Cu-Ni alloys equili-brated at 2OO0C.1 In general, a simple series of solid solutions will be stable only foralloys, where the TAS term in the free energy of alloy formation is larger than theAH term. In the Cu-Ni case, a AH value of as low as 0.46 kcal is sufficient to causephase separation at moderate temperatures. For the Au-Pd system I do not knowthe entropy data, but the enthalpies were recently measured by Darby 2 who reporteda value of - 1.98 kcal for the alloy Auo.bPd0.4.Dr.J. K. Clarke and Mr. J. J. Byrne (University College, Dublin) (communicated) :The work described by Dr. Sachtler requires comment since it questions the keyassumption of homogeneity usually made in catalysis by alloys. While information onfree energies suggests that a homogeneous copper-nickel alloy is mestastable attemperatures less than ca. 300" we would ask if phase separation in samples of highsurface area/volume ratio is established by the experimental findings quoted.3 Thuslattice parameters and photo-electric work functions reported may be explicable onkinetic grounds.In the early stage of interdiffusion of the two metals, following the core-shell modelof diffusion of Rudman,4 Barnes 5 and others, the copper shell covering particles ofnickel becomes quickly permeated by nickel so that after a short time no pure copperremains.At concentrations approaching 15 % nickel, however, interdiffusion1 Sachtler and Jongepier, J. CataZysis, 1965, 4, 665.3 Sachtler and Dorgelo, J . CataZysis, 1965, 4, 654, 665.4 Fisher and Rudman, J. Appl. Physics, 1961, 32, 1604.5 Barnes, Nature, 1950, 166, 1032.102 Darby, Acta Met., 1966, 14, 265262 GENERAL DISCUSSIONbecomes slower by up to two orders of magnitude at 200-300" I p 2 causing a nickel-leanalloy to persist in the surface shell. In consequence, work functions and lattice para-meters measured may be those characteristic of a system in which the alloying processhas been arrested as described. In the absence of a completely definitive surfaceproperty, more precise measurements of lattice parameters are desirable before phaseseparation at alloy surfaces can be regarded as established.We are presently engagedin precise measurement of lattice parameters for this system for different methods ofalloy preparation and on the catalytic behaviour of the samples.Prof. W. M. H. Sachtler (Amsterdam) said: I agree with Dr. Clarke and Mr.Byrne that in general both thermodynamic and kinetic effects must be considered inorder to understand the composition of the phases obtained by interdiffusion of twometals or co-reduction of two metal salts. In our copper-nickel films, to which Dr.Clarke and Mr. Byrne refer, the compositions of the two co-existing phases after inter-diffusion were derived from the lattice parameters and Vkgard's law.These compositionswere found to be identical-within the error of the experiment-with the compositionsof the two minima of the free energy against composition plot. We therefore thinkthat in this particular case the phases obtained were the thermodynamically stable ones.Prof. D. D. Eley (Nottingham University) said: While accepting the assertion ofphase separation in Cu-Ni alloys, I question the statement that this occurs in the Au-Pdalloys. According to evidence available, the two metals form a continuous series ofsolid solutions.3Prof. W. M. H. Sachtler (Amsterdam) said: In reply to Eley, handbooks onmetals and alloys describe both the Au-Pd and the Cu-Ni systems as continuousseries of solid solutions, although it is clear from thermodynamical data that for theCu-Ni system the continuous series of solid solutions does not represent trueequilibrium at moderate temperatures. For this system both AH and A S have beenmeasured, and the plot of the resulting free energy against the composition shows twopronounced minima at 200°C. Only by more sophisticated work on alloy films,prepared under ultra-high vacuum, has the two phase system predicted by thermody-namics been confirmed experimentally. This example shows that phenomenologicaldata on the structure of alloys do not necessarily refer to their equilibrium state.Alloys are usually prepared at a high temperature where in many cases TAS is largeas compared with AH. At these high temperatures the continuous series solutions isstable. After cooling it may be preserved as a metastable state, because atom mobilityis poor at low temperatures. As, however, the mobility is much larger for atoms inand near the surface than for atoms in the bulk, it seems possible that under theconditions of heterogeneous catalysis the zone near the surface acquires the structurepredicted by thermodynamics, while the bulk of the material may remain in its meta-stable state. For an analysis of catalytic data, therefore, the knowledge of the equili-brium structure may be of greater value than the conventional phenomenological data.Dr. J. M. Thomas (Univ. COX North Wales, Bangor) said : Dr. Holscher has raisedthe question of whether the composition of the palladium-gold alloys used by Dr.Rooney, Wells et al. is the same in the surface layer as in the bulk. In certain favour-able circumstances, electron probe microanalysis 4 may prove helpful in assayingsurface composition. This technique should be particularly useful when the solidcatalysts have crystallite areas much greater than the cross-sectional area of theelectron beam (ca. 1 pm2).1 Pines and Smushkov, Fiz. Tverdogo Tela (Engl. trans.) 1959, 1, 858.2 Correa da Silva and Mehl, Trans. Amer. Inst. Min. Met. Eng., 1951, 191, 155.3 Hansen, Constitution ofthe Binary Alloys, 2nd ed., McGraw Hill (New York, 1958), p. 224.4 Birks, Electronic Probe Microanalysis, (Interscience, New York), 1963

ISSN:0366-9033    

DOI:10.1039/DF9664100249

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Forms of Oxygen Bonds on the Surface of Oxidation CatalystsBy G. K. BoreskovInstitute of Catalysis of the Siberian Division of the U.S.S.R. Academy of Sciences,NovosibirskReceived 28th Janutlry, 1966The study of the isotopic exchange of oxygen on oxides and metals affords additional informa-tion about the form of the oxygen bond at the surface of catalysts. As a result of high temperaturetreatment of oxides, there appears on the surface a form possessing the catalytic activity for isotopicexchange of oxygen at very low temperatures. One may suppose that under these conditions theexchange of oxygen proceeds through an intermediate formation of many-atom, strictly oriented,complexes.For oxides and metals having an equilibrium content of oxygen in the surface layer, the activityin a homomolecular exchange may serve as a measure of the reactivity of the surface oxygen and iswell correlated with the catalytic activity in a series of oxidation reactions.The exchange in the non-equilibriated mixture of isotopic forms of molecularoxygen (homomolecular exchange)studied first by Winter 1 9 2 is the simplest reaction which proceeds with the parti-cipation of oxygen.It can be brought about by a dissociative-associative mechanismas well as through formation of a three-atom or four-atom complex on a catalystsurface. The activity of various catalysts towards homomolecular exchange permitsone to evaluate the probability of the formation of intermediate forms on a surface.The isotopic exchange of molecular oxygen with oxygen bonded to a catalyst (hetero-geneous exchange) allows one to judge the reactivity of this oxygen and its uniformityor non-uniformity.Most information may be obtained by a simultaneous studyof both homomolecular and heterogeneous exchange by carrying out experimentswith different isotopic composition of " catalyst oxygen " and of the gas phase,as well as with non-equilibrium ratios of the concentration of the various isotopicvarieties of molecular oxygen (1602, 1 8 0 2 , 1 6 0 1 8 0 ) . In this case, both the overallnumber of atoms of the isotope 1 8 0 and its distribution between the symmetricaland unsymmetrical molecules varies in the gas phase. From these findings onecan determine the participation of catalyst oxygen in the homomolecular exchange,e.g., the rate of exchange by a mechanism without the participation of the catalystoxygen (I) and with the participation of one (11) and two (111) atoms of oxygen bondedto a catalyst.3Fig.1 shows the change of the overall content of atoms 1 8 0 and the fraction ofmolecules 1 6 0 1 8 0 in the gas phase with the homomolecular exchange according toeach of these three mechanisms. Methods for the investigation of the isotopicexchange are described in detail in ref. (4). We restrict consideration here to thediscussion of the main results.l6O2 + 1s0,~2160180,O X l u c CATALYSTS WITH A N EQUILIBRIUM CONTENT OF OXYGENThe rate of both homomolecular and heterogeneous exchange of oxygen of oxidesessentially depends on the conditions of pretreatment. For numerous oxides it26264 OXYGEN BONDS ON OXIDATION CATALYSTSwas established that there were two states differing in their catalytic activity towardsthe isotopic exchange.The first is obtained when oxides are heated at elevated tem-peratures in oxygen and corresponds to an equilibrium content of oxygen in the surfacelayer of oxide. In this state, oxides possess a stable and reproducible activity forthe homomolecular exchange reaction which occurs at relatively high temperatures.160180timeFIG. 1.-The variation of the fraction of atoms 1 8 0 and the molecules 160180 in the gas phasewhen the isotopic exchange of oxygen proceeds by mechanisms I, 11, III (ao>0*5).The second state is usually attained after a high-temperature heating of oxides invacuum.It is often characterized by a high catalytic activity in isotopic exchangeeven at very low temperatures ; this activity is unstable and heating in oxygen usuallyeliminates it completely.Fig. 2 shows in co-ordinates of log K and 1/T the results of investigating thecatalytic activity in the homomolecular exchange of various oxides and metals pre-heated in oxygen. The specific catalytic activities per unit of oxide surface varyover a wide range. It is interesting to compare the rate of the homomolecularexchange reaction with that of molecular oxygen with the oxygen of the oxidesurface. This comparison is complicated because the surface oxygen of numerousoxides is non-uniform; e.g., with NiO, C03O4, Mn02, CuO one observes a sharpfall of the rate of heterogeneous exchange with extent of exchange, the order ofmagnitude being several % of the amount of surface oxygen.For these oxidesthe rate of homomolecular exchange was compared with the maximum initial rateof the isotopic exchange with oxygen of the catalyst surface. For nearly all oxidesheated in oxygen, i.e., containing the equilibrium amount of oxygen in the surfacelayer, the rate as well as the activation energy and the order for a homomolecularexchange and heterogeneous exchange of oxygen are very close, except for yA1203for which the rate of heterogeneous exchange even at high temperatures is much lessthan that of the homomolecular exchange.From the coincidence of rates of homomolecular and heterogeneous exchangeone may conclude that on an oxide, with the equilibrium content of oxygen, thesimplest catalytic reaction, e.g., the isotopic exchange in molecular oxygen proceedswith participation of oxide oxygen.For the oxides with uniform bonding energyof the surface oxygen, all the surface oxygen participates in the reaction ; on oxideswith the non-uniformly bonded oxygen only the most active " mobile " oxygenparticipates.It was possible to establish how many atoms of surface oxygen participate inthe exchange of molecular oxygen for some oxides ; e.g., with vanadium pentoxideG. K. BORESKOV 265molybdenum trioxide and tungsten trioxide, two atoms of the oxide surface takepart in each act of homomolecular exchange.5-7 With these oxides the isotopicexchange proceeds through reversible surface dissociation to form molecularoxygen.With y - A l 2 0 3 , the homomolecular exchange proceeds without par-ticipation of catalyst oxygen, and partially with participation of one oxygen atomof the surface catalyst in each act of exchange.8A, Ti02 ; A, Moo3 ; @, Fe2(Mo04)3 ; 0 , V2Os ; 0, Cs203 ; x , CeOz ; CI ZnO ; 0 Fez03 ;FIG. 2.-The catalytic activity of different oxides and metals, after heating in oxygen, in thehomomolecular exchange of oxygen.9 Nd203; IJ CuO; 0 NiO; Ul Pd; @I Mn02; 0 Ag; Gd203; (3 Pt; Q c0304.From the closeness of rates of the heterogeneous and homomolecular exchangefor oxides with an equilibrium content of oxygen, one may conclude that the type ofintermediate interaction between the oxygen and the catalyst is the same for theseoxides as that of an oxygen bond in the surface layer of oxide, or it is so close thattransitions between them may occur very rapidly.The existence of other forms of atomic oxygen on the surface is very unlikelyunder these conditions ; in any case, the rate of formation of such a form must bemuch less than that of oxygen bonded on the surface layer of oxide, otherwise theexchange in molecular oxygen would occur through this form, and its rate wouldbe higher than that of exchange between oxygen and oxide.Thus, one concludesthat the catalytic activity in the homomolecular isotopic exchange may serve as ameasure of the bonding energy and the reactivity of the surface oxygen of the oxidewith the equilibrium content of oxygen.OXIDES HEATED I N VACUUMWinter found that magnesium oxides and zinc oxides treated at high temperaturesmanifest catalytic activity for the homomolecular exchange of oxygen at relativelylow temperatures.9 According to Barry and Stone,l* this reaction proceeds rapidlyon zinc oxide treated in vacuum even at - 193°C. In our work 11 this result wasconfirmed for a specimen heated at 800" and then in vacuum at MOO266 OXYGEN BONDS ON OXIDATION CATALYSTSUnder the action of oxygen the catalytic activity decreased at - 193°C very slowly,but rapidly at room temperature, so that after several hours it became immeasurablysmall. On the specimen activated by such treatment only a homomolecular exchangeoccurred with no heterogeneous exchange.This means that under these conditionsthe homomolecular exchange occurs without the participation of oxide oxygen.It may proceed by a dissociative-associative mechanism to form adsorbed atomsof oxygen differing in their properties from the oxide oxygen, or by the formationand decomposition of three or four-atom oxygen complexes on the catalyst surface.Hiroty and Chono give evidence in favour of the latter hypothesis.12 The causeof the appearance of low-temperature catalytic activity of zinc oxide is the excesszinc which is produced as a result of high temperature treatment. This is confirmedby the fact that zinc oxide is activated by heating at 450" in zinc vapour.13The development of catalytic activity at low temperatures by high temperaturetreatment was found also for aluminium,l4 nickel,l3 gadolinium oxides 15 and someother oxides.The specimen of aluminium oxide ( y - A l 2 0 3 ) exposed to air dis-played catalytic activity in the homomolecular exchange of oxygen at temperaturesabove 350°C. After treatment in vacuum at 600°, it acquired a very strong activity;at 25" with an oxygen pressure of 10 torr the rate of exchange exceeded 1012 mole/cm'sec. This high activity is unstable and under the action of oxygen decreases rapidly.However, in contrast to zinc oxide, the low-temperature activity of aluminiumoxide does not disappear completely under the action of oxygen, but attains a fixedstable value. This final value is lower by more than two orders of magnitudecompared with the initial one (at 25" and 10 torr, 2 x 109 mol/cm2 sec) but remainsunchanged under the action of oxygen for over 1000 h.The same activity is attainedimmediately when aluminium oxide is heated at 600°, not in vacuum, but in oxygenat a pressure of 10 torr.Fig. 3 (line 1) shows the values of the stable activity of aluminium oxide, at-tained after the treatment in vacuum and subsequent heating in oxygen at 6OO0C,at various temperatures. Between 150 and 250°C the activation energy is 13 kcallmole, but at lower temperatures it has a negative value of about - 4 k 1 kcal/mole.This supports the evidence that the mechanism of molecular exchange is differentat low temperatures. Below 150" exchange proceeds through the active complex,whose heat of formation is positive while the entropy is very small, possibly dueto the requirement of a complicated orientation of its components.Fig.3 also shows the results of the investigation of the isotopic exchange ofaluminium oxide oxygen with molecular oxygen (line 11). The activation energyof the heterogeneous exchange is 37 kcal/mole, and its rate at the temperature in-vestigated is much less than the rate of a homomolecular exchange. When theextent of exchange increases up to 40 % of the monolayer, the rate of the hetero-geneous exchange remains constant; this indicates that the surface oxygen of alu-minium oxide is uniform.From the detailed analysis of the kinetics of the heterogeneous exchange andhomomolecular exchange one may conclude that the mechanism of exchange withthe participation of two atoms of oxide oxygen does not take place for aluminium.The homomolecular exchange proceeds either without the participation of oxideoxygen or (to a lesser degree at elevated temperatures only) with the participationof one atom of oxygen of aluminium oxide.The latter process leads to a hetero-geneous exchange. The mechanism of exchange with the participation of oneatom of oxide oxygen is probably the dissociative adsorption of oxygen with sub-sequent desorption by recombination into molecules with ions of the oxide lattice.As one possibility one may suppose that the oxygen adsorption takes place on thG. K. BORESKOV 267anion defects of the surface; one atom of oxygen fills the defect while the other,migrating over the surface, is desorbed after recombination with an oxygen ionof the oxide lattice.It is also possible that the adsorption of molecular oxygenand its exchange with oxide oxygen is realized in a single stage through a three-atom intermediate complex.I r!tIIAxI1I1 L12 5 4 51 0 3 / ~ * ~FIG. 3.-The homomolecular exchange (I) and the isotopic exchange with oxygen of oxide @I)for y-Al203 after treatment in vacuum and subsequently with oxygen at 600°C.The homomolecular exchange without the participation of oxide oxygen seemsto proceed at some particular surface sites, produced in the high-temperaturetreatment, which differ in chemical composition and structure. These localitiesare different from atoms of surface oxygen participating in the heterogeneousexchange.This is confirmed by a different dependence of rates of these processeson the condition of pretreatment. Thus, increase in the degree of surface dehydra-tion of aluminium oxide accelerates the rate of the homomolecular exchange to agreater extent than that of the heterogeneous exchange, while treatment with oxygendecreases the rate of homomolecular exchange but increases the rate of the hetero-geneous exchange. Moreover, the oxygen of the entire surface participates in theheterogeneous exchange, while the sites active for the homomolecular exchange coveronly a small part of the surface as confirmed by experiments in which they arepoisoned with oxygen and water vapours.Further, one may conclude that these active sites may be at least of two types.The first type (sites A) are formed as a result of treatment at high temperatures invacuum and are poisoned by oxygen.This supports the evidence that type Asites are fmmed as a result of removing oxygen from the surface of aluminiumoxide. The second type (sites B) are formed both by the treatment in vacuumand by high-temperature treatment in oxygen. They are stable to oxygen treat-ment but are readily poisoned by water vapour. Hence, one may conclude thatsites B result from considerable dehydration of the surface. Fig. 4 shows theextent to which the surface of y-aluminium oxide is covered with hydroxyl groups andthe activity towards a homomolecular exchange for a specimen treated at varioustemperatures. The rapid increase of catalytic activity begins after the loss o268 OXYGEN BONDS ON OXIDATION CATALYSTShydroxyl groups from about 70 % of the surface. As Peri has established 16 at thisdegree of dehydration, the surface of y-aluminium oxide begins to show “non-regularities ” of the surface structure owing to the non-uniform distribution of theresidual oxygen over the surface. One may suppose that the following form ofnon-regularity due to the removal of the remaining hydroxyl groups may arise :H H -0 (-HzO) 0 +A1 A1 A1 --+ A1 Al A1On these sites, or similar ones that are characterized by the presence of twoneighbouring exposed atoms of aluminium on the surface, the isotopic exchangemay proceed as follows :- -0 f 4-02 0- -0- --o - 0 2 + 00 1 0 0 0 0 1 0A1 A1 Al -+A1 Al A1 -+ A1 Al Al/ \ /I\ / \ / \ / I \ / \ / \ / I \ / \If the fraction occupied by such sites is small and adsorption and desorption of oxygento form the symmetrical three-atom complex proceeds is fast enough, then the schememay explain the rapid homomolecular exchange and the absence of heterogeneousexchange.FIG.“C4.-The effect of the degree of occupation of the surface with hydroxyl groupscatalytic activity of y-Al2O3 in the homomolecular exchange of oxygen at 20°C.upon theThe experimental data available are insufficient to make a final judgment aboutthe mechanism of oxygen exchange at various sites of aluminium oxide in the varioustemperature regions.One may assert that at low temperatures (which is character-ized by a negative value of the activation energy), the exchange does not proceedthrough dissociative adsorption but through formation of a complicated complexof definite configuration containing several atoms of oxygen.For the exchangeat sites A, where there is partial removal of oxygen from the surface, there is nG. K. BORESKOV 269evidence to choose between intermediate dissociative adsorption and formation ofmany-atom oxygen complexes.The study of the isotopic exchange of oxygen on gadolinium oxide led to similarresults.14 Fig. 4 shows the Arrhenius diagram of the dependence of the rate of ex-change on Gd203 on the temperature. The right-hand side of the diagram (linesBCD) characterizes the activity of a specimen treated in vacuum at 700°C.A strong catalytic activity is observed even at low temperatures up to -78"C,but the homomolecular exchange is not accompanied by heterogeneous exchange.In the temperature range from -30 to -78" the activation energy is negative andequals - 4.5 kcal/mole, and the homomolecular exchange seems to proceed throughan active complex with the positive heats of formation but with a very small entropyas for y-Al2O3.Between -30 and 200°C the activation energy of the exchange is+5 kcal/mole. At temperatures below 200" the activity remains stable and un-changed even after long times (100 h) at 20". Heating in oxygen, however, attemperatures about 200°C causes a disappearance of low-temperature catalyticactivity.Contrary to y-Al203, no residual activity was observed. This fact maybe connected with the small surface area, being for Gd2O3, 4.5 m2/g instead of124 mz/g for y-Al203. The catalytic activity of gadolinium oxide heated in oxygenabove 200°C corresponds to the straight line ABE of fig. 4 where the activationenergy is 10 kcal/mole. In this case along with the homomolecular exchange therealso proceeds the heterogeneous exchange with approximately the same rate.For the specimen withan equilibrium content of oxygen heated in oxygen at the temperatures above 200°Cthe homomolecular exchange proceeds with the participation of oxide oxygen andseems to be related to its reversible surface dissociation. After heating in vacuumat 700" there is a low-temperature activity which is probably connected with thepartial loss of oxygen from the oxide surface.On the active sites formed, the ex-change proceeds by different mechanisms above and below -30°C. In the low-temperature range the exchange probably takes place through a strictly orientedcomplex containing several oxygen atoms.Thus, for Gdz03 there are three types of exchange.METALSOxygen isotopic exchange may be used also for the characterization of theoxygen bond on the surface of metal catalysts. In the Institute of Catalysis, ahomomolecular exchange and the exchange with sorbed oxygen was studied onplatinum, silver, palladium and nickel films. As noted previously, there is sorbedan amount of oxygen on platinum and silver corresponding to several monolayers.17The overall amount exceeds the solubility of oxygen in the metal, and it is probablethat the sorbed oxygen penetrates into the surface layers of metal to some depth.The homomolecular exchange of oxygen on platinum films proceeds with anobservable rate at 200".The activation energy is 16+2 kcal/mole, and the orderwith respect to oxygen pressure is close to 0.5. A study of the exchange with thesorbed oxygen provides evidence for the non-uniformity of oxygen on platinumsurface. The sorption heat varies between 13 and 24 kcal/mole over 25 % of themonolayer, while the activation energy increases by 5 kcal/mole. The rate of ahomomolecular exchange is close to the initial rate of the isotopic exchange withadsorbed oxygen measured at insignificant degrees of exchange.Hence, one mayconclude that the homomolecular exchange reaction proceeds on a small area,which is only 3 % of the overall platinum surface.18The rate of a homomolecular exchange of oxygen on silver films is one orderof magnitude less than that on pIatinum.19 The study of the exchange with sorbe270 OXYGEN BONDS ON OXIDATION CATALYSTSoxygen indicated that in contrast to platinum, the surface oxygen is uniform at leastover most of the surface. The concentration ratio of 1 6 0 3 8 0 and 1 8 0 2 in the oxygenleaving the surface of silver is equal to the equilibrium one for isotope content ofthe sorbed oxygen. This supports the conclusions of Sandler and Hickam20 fora high mobility of the sorbed oxygen and the absence of large differences in bondingenergy.The rate of the homomolecular exchange is the same as that of the hetero-geneous exchange; thus the sorbed oxygen exchanging at 200" is in the form ofatoms. If molecular oxygen is still present on the silver surface, the rate of itsexchange with the oxygen atom must be very high. At 250" the kinetics of exchangeis complicated and an additional amount of oxygen begins to participate in theexchange with a lower rate. One may suppose that by heating at 250" the sorbedoxygen is converted to a state having a stronger bond, and the character of bondmay approach that in silver 0xide.1954*3 32I30 30 40 50 601/Tx 103FIG. 5.-The homomolecular and heterogeneous exchange of oxygen.On palladium films, oxygen sorption proceeds with a high but continuouslydecreasing rate.The rate of the isotopic exchange also decreases with increase ofamount of oxygen adsorbed. After 20 h the amount of oxygen bonded reaches30 monolayers. Electron diffraction investigations carried out subsequently haveshown the formation of PdO on the surface of the palladium crystallites. The rateof the heterogeneous exchange rapidly decreases with the extent of exchange whichindicates that the oxygen on the surface is very non-uniform. The homomolecularexchange proceeds on a very small part (2 %) of the surface of the palladium film.21On a freshly prepared nickel film the homomolecular exchange proceeds witha very high rate (at 23°C 4.2 x 1013 molelcmz sec), but after 1 h it is 6 times smaller.As the oxygen is absorbed the rate of the homomolecular exchange continues todecrease according to an exponential law with respect to time of sorption of oxygen(fig.6). The rate of sorption decreases according to the same law, so that thG . K. BORESKOV 27 1ratio of the rate of the homomolecular exchange and the rate of oxygen sorptionis maintained constant (fig. 7). One may assume, therefore, that both the rates ofthe homomolecular exchange and of oxygen sorption are proportional to the con-centration of the same catalytically active sites which are probably the unoxidizedI0 I 2 3 4 57, P AFIG. 6.-The variation of the rate W of the homomolecular exchange of oxygen on nickel film at23°C with the time T of oxygen sorption.w I6 tg 24log xFIG.7.-The variation of the rate W of the homomolecular exchange of oxygen and the rate V ofoxygen sorption on a nickel film at 250°C with increase of amount of adsorbed oxygen.X = the number of monolayers of sorbed oxygen.atoms of nickel. After the absorption of large amounts of oxygen correspondingto hundreds of monolayers, the specific activity of the film approaches the activityof nickel oxide. We note that the initial catalytic activity of nickel exceeds theactivity of platinum by several orders of magnitude. For gold films there was n272 OXYGEN BONDS ON OXIDATION CATALYSTSobservable catalytic activity in the homomolecular exchange of oxygen at 350 and400°C.In table 1 we compare the characteristics of the catalytic properties of metals withrespect to the homomolecular exchange.Here W is the rate of the homomolecularexchange at 250°C mole/cmz sec; H is the fraction of the surface on which theTABLE 1 .-THE CATALYTIC ACTIVITY OF METALS WITH RESPECT TO THE HOMOMOLECULAREXCHANGE OF OXYGENmetal W H 4 E uniformity of oxygenon metal surfacePt 1 - 8 ~ 1012 0-03 6~ 1013 13 16 A2 0.5 non-uniformPd 4 . 2 ~ 1010 0.02 2.1 x 1012 - 30f3 0.3 sharply(oxidized) non-uniformAg 2 . 4 ~ 1011 1 2 . 4 ~ 1011 13 31 f 2 0.3 uniformhomomolecular exchange takes place; W/H is the rate of exchange per unit areaof the working surface (mole/cm2sec); q is the heat of oxygen adsorption at thesites where the exchange takes place (kcalfmole) ; E is the activation energy of thehomomolecular exchange (kcal/mole) ; and n is the order of the reaction with respectto oxygen.The specific catalytic activity of films decreases as follows :Pt > Ag > Pd > Au ;if one compares the sites of the surface where the homomolecular exchange pre-sumably takes place, silver and palladium change their places. The heat of oxygensorption on silver and on the sites of the platinum surface responsible for the homo-molecular exchange are equal but the rates of exchange on these metals stronglydiffer. Silver differs from platinum and palladium by its high mobility and uni-formity of oxygen on the surface. It is possible that this difference in the natureof the oxygen bond to these metals is due to the fact that silver has no unfilled d-orbitals.CORRELATION BETWEEN OXYGEN BONDING ENERGY AT THE SURFACE OFCATALYSTS AND THEIR CATALYTIC ACTIVITY I N OXIDATIONSThe above results of the isotopic exchange of oxygen on a number of oxidesand metals, the simplest catalysts for oxidations, indicate that there are variousforms of oxygen bonding on the surface of these catalysts.It would be interestingto establish the importance of these forms for the various catalytic oxidation reactions.Unfortunately, the active sites formed as a result of high temperature treatmentare not sufficiently stable, and it is not clear whether the oxygen sorbed on themparticipates in other catalytic reactions in addition to the isotopic exchange ofoxygen.The correlation between the oxygen-bonding energy at the catalyst surfaceand the catalytic properties in oxidations for catalysts with the equilibrium con-tent of oxygen in the surface layer is more clear.The correlation is based on theassumption that for a certain group of catalysts the variation of the active complexenergy may essentially be determined by the variation of the energy of one of thebonds which either break or is formed. If the formation of an active complexin the oxidation reaction includes the rupture of the oxygen-catalyst bond, thenone may expect a linear dependence of the activation energies on the variation ofthe bonding energyG. K. BORESKOV 273The nature of this dependence is illustrated in the following example. We supposethat when oxidation proceeds, the oxidized substance reacts with oxygen filling apart 8 of the active sites of the catalyst surface, oxygen being sorbed on the un-occupied fraction 1-8 of these sites.We assume for simplicity that the rate ofthe first stage is proportional to 8 and of the second to (1 - 0), i.e.,Wl = W f d P i ) , (1)W2 = K2(1 - e > f . ( ~ i ) . (2)Here KI and K2 are the rate constants of the first and the second reactions, whilefi(p~) andfi(Pi) express the dependence of the rate of these reactions on the com-position of the reaction mixture.In the stationary state when the rates of removal and the addition of oxygenWhen the first reaction takes place, the oxygen-catalyst bondthe second reaction takes place it is formed. Making use ofcorrelation, one may suppose that the activation energy of theand of the second one isEl = Eo,+a,q,E2 = E02--a,q,is broken, and whenthe Bronsted-Polanyifirst reacion is(4)where q is the energy of oxygen-catalyst bond, and a1 and a2 are constants lyingbetween 0 and 1.Substituting (4) and (5) in (3) and (l), we find the stationary coverage of oxygenon the active sites to beand the rate of the catalytic reaction to beThese equations show that the reaction rate increases with increase of q at smallvalues of q, passes through a maximum and then decreases.The optimal value ofq corresponding to the maximum of the rate is determined from the ratio :It corresponds with the maximum coverage of the active sites with oxygenPreviously, the search for a correlation between the catalytic activity and thebonding energy of the reactants with a catalyst have been quite successful.2o-23For the bonding energy of oxygen, one used the heat of conversion of oxides tovarious degrees of oxidation to each other or to a metal.These values may essenti-ally differ from that for bonding energy of oxygen at the surface of oxides andmetals. As shown above, oxygen sorbed on the surface of oxides and metals wit274 OXYGEN BONDS ON OXIDATION CATALYSTSthe equilibrium content of oxygen in the surface layer is very similar in bondingenergy to that of oxygen of the oxide surface. For these catalysts the homomolecularisotopic exchange proceeds with the participation of the most reactive oxygen at thesurface. The rate of the homomolecular exchange may, therefore, serve as a measurefor the reactivity of the surface oxygen and for a non-uniform surface to characterizethe most active part.Thus one may expect the catalytic activity in oxidations to vary similarly withcatalytic activity in the homomolecular exchange of oxygen.For a series of oxida-tion reactions this assumption received experimental support. Fig. 8 shows the3414 Upr C & ~ F s O , 6 p l ~ O CuO ZnOFIG. &--The catalytic activity of oxides of the elements of the fourth period in the homomolecularexchange of oxygen (l), oxidation of hydrogen (Z), oxidation of methane (3) and nitrogen oxidedecomposition (4) at 300°C.specific catalytic activities of oxides of elements of the fourth period in the homo-molecular exchange of oxygen,4 oxidation of hydrogen,27 oxidation of inethane 2 8and nitrogen oxide decomposition.zg All these reactions were carried out underoxidizing conditions where the composition of catalyst surface approached anequilibrium state corresponding to the conditions for the isotopic exchange toproceed.A distinct similarity in the change of a specific catalytic activity of oxideswith respect to the above reactions may then be noted. A similar relation was alsoseen for the oxidation of carbon oxide and of ammonia as well as for promotedvanadium catalysts in the oxidation of hydrogen and sulphur dioxide3The activity of catalysts in the exchange of molecular oxygen is thus decisivefor their activity in many other reactions too.This simple dependence may becomemore complicated by the effect of other factors. The catalytic activity may belimited by the stability of oxide or metal phases formed during the catalytic reaction.In some cases an essential contribution to the energy of an active complex in thelimiting stage of reaction may be bonding energy of the reactant to the catalyst, etc.It is more difficult to establish a correlation between the catalytic propertiesand the reactivity of the surface oxygen in complex oxidation reactions when thG. K. BORESKOV 275yield of the desired product is determined by concurrent and consecutive reactions.In fig. 9 an example is given of the results of the study of the oxidation of methanolwith excess of oxygen on the oxides of the fourth period elements. The measurementsof catalytic activity were carried out in a flow-system with circulation so that the uni-formity of temperatures along the catalyst layer was ensured despite a high heat ofreaction.The specific catalytic activity of oxides in terms of the overall rate ofmethanol oxidation is seen to be similar to the reactivity of oxygen in the above reac-tions. The selectivity of the processes decreases sharply with increase of reactivity ofP IXIFIG. 9.-The specific catalytic activity K and the selectivity q for formaldehyde formation of theoxides of the fourth period elements in methyl alcohol oxidation.the catalyst oxygen (curve I1 of the fig. 8). Thus, in order to obtain the intermediateproducts of oxidation, the reactivity of oxygen at a catalyst surface must be highenough to ensure the required rate of oxidation of the initial substance, and suf-ficiently low to prevent oxidation of products of incomplete oxidation.Similarresults were reported by Sachtler and De Boer 30 in the oxidation of olefines toaldehydes, or of diolefines, with the aid of molybdenum catalysts. They character-ized the reactivity of catalyst oxygen by the temperature at the beginning of theirreduction by hydrogen. We believe that the specific catalytic activity in the homo-molecular exchange of oxygen may be a more convenient and more general charac-teristic for the reactivity of oxygen catalysts with the equilibrium content of oxygenin the surface layer.1 Winter, J.Chem. Soc., 1954, 1522.2 Winter, J. Chem. Sac., 1955, 3824.3 Muzikantov, Popovsky and Boreskov, Kinetika i Kataliz, 1964, 5, 624.4 Azisiak, Boreskov and Kasatrina, Kinetika i Kataliz, 1962, 3, 81.5 Muzikantov, Popovsky, Boreskov and Mikichur, Kinetika i Kataliz, 1964, 5, 745.6 Klier, Novakova and Jiru, J. Catalysis, 1963, 2, 479.7 Novakova, Klier and Jiru, Proc. 5th Int. Symp. on the Reactivity of solids, Munich, 1964 (pre-8 Muzikantov, Popovsky, Boreskov, Goldstein and Shubnikov, Methods of Investigation of9 Winter, Adv. Catalysis, 1958, 10, 196.10 Barry and Stone, Proc. Roy. Soc. A , 1960, 255, 124.11 Boreskov, Gorgoraki and Kasatrina, DAN S.S.S.R., 1963, 150, 570.print).Catalysis and Catalytic Reactions, Novosikirsk, 1965, 1, 159276 OXYGEN BONDS O N OXIDATION CATALYSTS12Hirota and Chono, J. Catalysis, 1964, 3, 196; Sci. Papers I.P.C.R., 1964, 58, 115.13 Gorgaraki, Boreskov, Kasatrina and Sokolovsky, Kinetica i Kataliz, 1964, 5, 120.14 Boreskov, Muzikantov, Popovsky and Goldstein, DAN S.S.S.R., 1964, 159, 1354.15 Sokolovsky, Sazonov and Boreskov, Methods of Investigation of Catulysts and Catalytic16 Peri, J. Physic. Chem., 1965, 69, 220.17 Temkin and Kulkova, DAN S.S.S.R., 1955, 105, 1021.l8 Asih and Boreskov, DAN S.S.S.R., 1963, 152, 1387.19 Boreskov, Hasin and Starostina, DAN S.S.S.R., 1961, 164, 606.20 Sander and Hickam, Proc. Int. Congr. Catalysis, Amsterdam, 1965, 1, 227.21 Hasin, Boreskov and Starostina, Methods of Investigation of Catalysts and Catalytic Reactions,22 Makishima, Joneda and Saito, Act. 2nd Congr. Catalyse, 1961, 1, 617.23 Sachtler and Fahrenfort, Act. 2nd Congr. Catalyse, 1961, 1, 831.24 Roiter and Golodetz, Ukr. Him. Jurn., 1963, 29, 667.25 Tanaka and Tamaru, J. Catalysis, 1963, 2, 366.26 Boreskov, Adv. Catalysis, 1964, 15, 285.27 Popovsky and Boreskov, Izd. Akad. Nauk S.S.S.R., 1960, 10, 67.28 Andrushevich, Popovsky and Boreskov, Kinetika i Kataliz, 1965, 6, 860.29 Urieva, Popovsky and Boreskov, Kinetika i Kataliz, 1965, 6, 1041.30 Sachtler and De Boer, Proc. 3rd Int. Congr. Catalysis, Amsterdam, 1965, 1, 252.Reactions, 1965, 1, 151.Novosibirsk, 1965, 1, 342

ISSN:0366-9033    

DOI:10.1039/DF9664100263

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Electron Spin Resonance Study of Rearrangements in theCo-ordination of Cr5+ and V4f Complexes due toChemisorptionBY L. L. VAN REIJEN AND P. COSSEEKoninklijke/Shell-Laboratorium, Amsterdam(Shell Research N.V.)Received 5th January, 1966The surface chemistry of Cr5+ complexes in chromia/silica catalysts for the polymerization ofethene is investigated. To this end not only the chromia/silica system is studied, but also the relatedcombinations chromia/alumina, vanadia/silica and vanadialalumina. Electron spin resonancespectra are used to follow the rearrangements in the co-ordination of the Cr5+ and V4+ ions, causedby contact with a number of different gases or vapours : HzO, air, CO, C2H4, HC1 and NH3. Itis found that the atomic arrangement at the surface of some of the systems studied is very flexibleand easily influenced by, for instance, contact at room temperature with H20, HCl or NH3.It issuggested that such a flexibility-encountered especially in chromia/silica-is a prerequisite for thecatalytic activity. The active species in the fresh catalyst is Cr5+ in tetrahedral co-ordination.A mechanism is proposed according to which in the initiation step of the polymerization the co-ordination number is increased from four to five. An empty sixth co-ordination site may then beavailable to accommodate ethene monomers.Early in the history of the application of electron spin resonance to catalyststudies, the unusual valency state Cr5+ was detected in chromia/alumina andchromia/silica systems.1-3 As Cr5+ was found exactly under those conditions wherethese substances are used as catalysts for the polymerization of ethene (Phillipsprocess), it was a natural step to ascribe the catalytic activity to this particularspecies.2, 49 5 The most direct evidence available thus far to support this hypothesishas been given by Boreskov and co-workers.4, 6 For chromia on silica, silica/aluminaand alumina these authors established a proportionality between the polymerizationactivity and the intensity of the e.s.r.signal of Cr5+ in the fresh catalyst.Yet there is a certain ambiguity in these results. In the first place, there is asubstantial difference in specific activity between Cr5+ supported on silica and Cr5+supported on alumina, the former being far more active. In the second place, recentspectroscopic and electron spin resonance investigations of various well-definedsystems containing Cr5+ have shown that Cr5+ may occur in different co-ordinations.In highly alkaline solutions and in compounds like Sr5(CrO&.OH, pentavalentchromium is tetrahedrally co-ordinated in the anion (cr04)3-.7-10 'In highly acidicsolutions and in compounds like (NH4)2CrOCls, chromium has the square pyra-midal co-ordination, well known for V4+ in many vanadyl complexes.11-14 As bothof these Cr5+ species have also been found in chromia/alumina systems,lo thequestion arises which of the two co-ordinations is involved in the catalytic activity.The present investigation was undertaken to settle this point by an attack alongtwo lines.(a) An extensive study of the chemistry of the Cr5+ surface complex by investigat-ing the interaction of the activated chromiafalumina and chromia/silicasystems with a number of different gases and vapours : H20, air, CO, C2H4,27278 E.S.R.OF Cr5+ AND V4f SURFACE COMPLEXESHCl and NH3. Analogous experiments have also been performed withvanadialalumina and vanadia/silica systems.(b) Extension of the range of temperatures at which e.s.r. spectra have beenmeasured downwards to the temperature of liquid hydrogen (-253°C).This study of the surface chemistry of the Crs+ complexes may be considered as acontinuation of work by Kazansky and Pecherskaya,ls who measured the changesin the e.s.r. spectra of supported chromias due to the interaction with H20, 0 2 ,saturated hydrocarbons, C2H4 and acetaldehyde.The application of electron spinresonance at -253°C is a new feature of the present work. It is essential for thedifferentiation between the two Cr5+ species. It has been shown 10 that the e.s.r.signal of Cr5+ in tetrahedral co-ordination is influenced by spin-lattice relaxation.Therefore it shows a temperature-dependent line width, whereas the square pyramidalspecies does not. When the two species are present simultaneously, both will beseen in an e.s.r. spectrum at -253°C ; the square pyramidal one dominates at roomtemperature.EXPERIMENTALPREPARATION OF SAMPLESThe supports are commercial products : silica from Mallinckrodt, alumina from A.I.A.G.,type RA. The chromium and vanadium-containing samples were made by impregnationwith aqueous solutions of CrO3 and NH4VO3 respectively, followed by drying at 100°Cand calcination in air at 500°C.All experiments for a particular metal-support combinationwere started from a single batch of the calcined sample. The amounts of chromium andvanadium were in all cases 2 % by weight.Vacuum treatment at 500°C was applied only after a first calcination at 500°C : it hasbeen established for all samples that a second calcination at 500"C, after the vacuum treat-ment, duplicates the original sample. The gases used for treatment of the samples weredried carefully. Generally the catalyst was in contact with the slow-streaming gas for 1 h.The vapours H20, CH30H and CH3COH were carried over the catalyst in a stream ofnitrogen.ELECTRON SPIN RESONANCEAll samples were measured in quartz tubes (length 10 cm, int.diam. 0.3 cm). Thesetubes were side tubes of a reactor, in which calcination, vacuum treatment and subsequenttreatments were performed. After the pretreatment the catalyst was moved to the side-tube and the side-tube sealed off. E.s.r. measurements were made in the standard Varianequipment at 9-13 x 109 c/sec. Measurements at -253°C were performed by placing thesample tube in a Dewar filled with liquid hydrogen. The Dewar was then in turn insertedinto the microwave cavity. The intensities of the e.s.r. signals were measured by comparisonwith a reference, consisting of a mixture of vanadyl disalicilalethenediamine with an inertoxide.RESULTSBefore proceeding to a detailed discussion of the behaviour of Cr5+ in aluminaand silica surfaces, it is useful to give some attention to chromium in other valencystates. Only a relatively small fraction of the chromium (never exceeding 20 %)is in the pentavalent state.Under the conditions of our experiments the remainderof the chromium is either Cr6+ or Cr3f. Apart from the fact that the Cr5+ species ispresumed to be responsible for catalytic activity there is still another reason why it isstudied in preference to the others : owing to its one unpaired 3d-electron it is veryeasily detected by electron spin resonance, while details of its co-ordination can alsL. L. VAN REIJEN AND P. COSSEE 279be obtained by this method.Cr6+ is, on the contrary, diamagnetic and, hence, isnot detected at all, whereas the e.s.r. signals of Cr3f are spread over a wide range offield strengths and give rise only to a weak background. The easy detectability ofCr5+ might give rise to the notion that this is the only species influenced by theinteraction with the environment. This is certainly not correct. Fortunately thereare good reasons to believe that after catalyst preparation Cr6f is stabilized at specificsites of the surface and Cr5+ at others. Thus, Cr5+ is not a mere intermediate in theroute from Cr6+ to Cr3f. It may, for instance, be relatively stable under conditionswhere interconversions between the other species certainly occur.Various discussions about the valency states of chromium other than Cr5+ arealready available in the literature.13 2, 103 16-22 A survey of our own experiences inthis field will be published soon.In the present investigation four different combinations have been studied :Cr/SiO2, Cr/A1203, V/SiO2 and V/A1203.Although there is a certain similarity inthe behaviour of these systems, the most clear-cut results have been obtained forCr/SiO2. There are two reasons for this : in the first place the Cr5+ species in thesilica surface display well-resolved e.s.r. spectra with narrow lines ; secondly, the Cr5+species at the silica surface are distinctly more sensitive to the composition of the gasphase than Cr5+ in alumina and V4f. Therefore in the following a rather thoroughdiscussion will be given of the results for CrjSiOZ, while for the other combinationsonly some striking features will be mentioned.Cr/SiO2AFTER VACUUM TREATMENT AT 500°CThe e.s.r.spectra at 20°C and -253°C are given in fig. 1. The main feature ofthe spectrum is the strongly anisotropic peak with g1 = 1-975 and 911 = 1.898.- 100 GAUSStn IFIG. 1.-Electron spin resonance of Cr/SiO2 after vacuum treatment at 500°C.(a) spectrum at -253"C, magnification 1 x ; (b) spectrum at 20°C, magnification 16OxRelative values for the magnification at which these spectra were obtained are givenin the figures. Whereas according to the Curie law this ratio should be equal tothe inverse of the ratio of the absolute temperatures, i.e. about 15, a ratio of 160 isfound. This indicates that the spectrum at room temperature is substantiallybroadened with respect to that at -253°C.This is further confirmed by comparingthe ratio of the integrated intensities of the two spectra. This is about 50, whichshows that at room temperature part of the intensity has disappeared so far into thewings of the spectrum that it is lost on integrating the derivative spectrum280 E.S.R. OF Cr5+ AND V4+ SURFACE COMPLEXESAn analysis of this phenomenon has been given earlier9 10 The line broadeningat room temperature is ascribed to spin-lattice relaxation. A spin-lattice relaxationof the right order of magnitude is only to be expected for Cr5+ surrounded by atetrahedron of ligands.Further speculation about the surface configuration of Cr5+ in silica is possible.It is reasonable to assume that the (cro4)3- tetrahedron is linked to the Si02 net-work by sharing corners or edges with Si04 tetrahedra. Four different ligands ofthe Cr5+ may be distinguished : 0 2 - ions shared by a Cr5+ and a Si4+, 0 2 - ionspointing outwards, OH- ions pointing outwards and H20 molecules pointing out-wards.The symmetry of the ligand field is determined primarily by the polarizationof the bonds to the oxygens. According to this criterion the ligands fall into threecategories : a strongly polarized bond for an 0 2 - ion pointing outwards, a weaklypolarized bond for a H20 molecule and intermediately polarized bonds for the othertwo types of ligands. In this way it is possible to form: (i) a tetrahedron with apredominantly trigonal symmetry, formed by one strongly and three intermediatelypolarized ligands (see fig.2(a) and 2(b)) ; (ii) a tetrahedron with a predominant mirrorsymmetry, formed by three differently polarized ligands (see fig. 2(c)).Si( 0 ) (b) ( 0 )FIG. 2.-Crs+ in tetrahedral co-ordination at the surface of Cr/SiO2.(a) and (b) trigonal symmetry ; (c) mirror symmetryThe symmetry of the configuration will be reflected in the symmetry and thetemperature dependence of the electron spin resonance.* For (i) nearly symmetricale.s.r. spectra are to be expected, broadening rapidly with temperature in the range-250 to -200°C. For (ii) an anisotropic e.s.r. spectrum is to be expected with amore moderate increase in line width with temperature.When again consideringfig. 1, it is probable that the differences in line shape at -253 and 20°C are due to thesimultaneous presence of the two types. Careful analysis of the line shapes will berequired to make quantitative estimates.AFTER TREATMENT WITH H20, HCl OR NH3The e.s.r. spectra are given in fig. 3, the parameters in table 1. It is seen that theproducts of these treatments are essentially different, showing a strong anisotropyafter NH3, a weak anisotropy after H20 and a reversed anisotropy (2-gIr ~ 2 - g ~ )after HCl. A common feature of the three spectra, however, is the temperature-dependence. The line widths are independent of temperature and the intensitiesfollow the Curie law. As suggested in earlier work,9, 10 this is a strong argument infavour of a square pyramidal co-ordination of Cr5+ in these surfaces (see fig. 4).* The rate of spin-lattice relaxation is determined by the separation of the lower orbital doubletof the tetrahedral configuration ; nearly cubic or nearly trigonal symmetry allow only small separa-tions, giving rise to rapid relaxation; mirror symmetry allows much greater separations withcorrespondingly slower relaxation ratesL.L. VAN REIJEN AND P. COSSEE 28 1TABLE 1 . E . s . ~ . PARAMETERS OF Cr5+ IN Cr/SiOa SYSTEMS81 A 11 x 104 A ~ X 104(cm-1) (cm-1) IIvac. 500°C 1.898 1.975vac. 5OO0C, HC1 20°C 2.002 1 *970VX. 500"C, H20 20°C 1.952 1 -972 44 11vac. 5OO0C, NH3 20°C 1-925 1 -992 46 19The oxygen anions occupy roughly five of the six corners of an octahedron ; the bondof Cr5+ with the lower oxygen is enhanced by n-bonds between two occupiedp-orbitalsof the oxygen and two empty &orbitals of Cr5+ with the proper symmetry."II 100 GAUSSHl int------(I00 GAUSS V( 0 1 ( b l (C)FIG. 3.Electron spin resonance of Cr/SiOz after vacuum treatment at 500°C ; spectra at -253°C.(a) after contact with H20 vapour at 20T, magnification 6x ;(b) after contact with HCl at 20°C, magnification 2x ;(c) after contact with N H 3 at 20°C, magnification 0.3 x .In recent years, several e.s.r.studies have been made of square pyramidal complexesof pentavalent chromium. In table 2 the e.s.r. parameters of these complexes aregiven. It is satisfactory to see that for at least two of our three species there is a goodanalogy.The spectrum found after treatment with H20 is similar to that of Cr5+complexes formed by reduction of CrO3 in aqueous solutions of various acids likeacetic acid, trifluoroacetic acid and oleum. The spectrum after treatment with HClTABLE 2.E.S.R. PARAMETERS OF Cr5+ IN SOME COMPOUNDSgL ~ l l x i o 4 ~ ~ ~ 1 0 4 lit.(cm-1) (cm-1)0 IICrO3 in CF3COOH 1.961 1 a977 Kon 14CrO3 in oleumCsHsNHCr 0 C141.986 Symons 13 {z; 1.975 39 9 Van Reijen 101.979 Kon 231.976 36 9 Van Reijen 10is similar to that found for the CrOC14 configuration in CSHSNHCrOC14. Apparently,by the interaction with HCl the oxygen ligands in the equatorial plane of fig. 4 havebeen replaced by chlorines. No adequate analogy in homogeneous media is availablefor the result of the treatment with NH3.Two possibilities may be considered.Either the surface complex obtains its strong anisotropy from the presence of (NH4)282 E.S.R. OF Cr5+ AND V4+ SURFACE COMPLEXEScations in the surface layer, or the NH3 has reacted at the surface to form NH;, NH2-or N3- ligands. It is reasonable that such ligands might be formed in contact with ahighly charged positive ion like Cr5+. Examples have been described for the ionsOs8+ (complex (Os03N)-), Re7+ and M06*.241 1 1 1FIG. 4.-Cr5+ in square pyramidalco-ordination.FIG. 5.-Electron spin resonance of Cr/SiO2,enriched in 53Cr, after vacuum treatment at500°C and contact with N H 3 at 20°C; spec-trum at -2253°C.The interpretations given thus far can be refined by studying the hyperfine structurein e.s.r.spectra of samples enriched in the isotope 53Cr (nuclear spin 3/2). Anexample is presented in fig. 5. The hyperfine structure parameters have been assembledin table 1. They correspond well with those known for Cr5+ complexes studiedhitherto (see table 2).QUANTITATIVE ANALYSIS OF THE Cr5+ SPECIESA satisfactory quantitative analysis of the Cr5+ present requires the measurementof the e.s.r. spectra at -253°C. At higher temperatures the rapidly growing linewidth of some of the possible species leads to the disappearance of their contributionto the spectra. This is also the reason why at room temperature sometimes smallquantities of unimportant components are detected, while at -253°C only the majorspecies are visible.An example is given in fig. 6 for the treatment with CO at 20°C.In the room-temperature spectrum, three different types are found side by side; inthe spectrum at - 253"C, pentavalent chromium in tetrahedral co-ordination pre-dominates by far. Similar situations are encountered after treatment in air at 500°Cand after treatment with saturated hydrocarbons, C2H4, CH30H or CH3COH.The main results of the quantitative analyses are given in table 3. I t is seen thata high concentration of (cro4)3- is formed by evacuation at 500°C. This (cro4)3-is relatively stable against treatments with dry air at 500"C, CO at 200°C and C2H4 at100°C. It is strikingly sensitive, however, to contact with gaseous H20, HCl andNH3, even at room temperature.Water vapour leads to disappearance of the greater part of all the Cr5+ present.Only small quantities of Cr5+ are retained, in the square pyramidal co-ordination;the rest probably disproportionates into Cr3f and Cr6+.HC1 and NH3 lead tomuch more stable Cr5+ species, also in the square pyramidal co-ordinationL. L. VAN REIJEN AND P. COSSEE 283The remarkable instability of the (CrO@- configuration at silica surfaces and theease of formation of certain square pyramidal configurations containing Cr51- the ionis a new and surprising result of the present work. In the later discussion the im-portance of these phenomena for the polymerization activity of Cr/SiOZ catalystsare evaluated.1.2FIG. 6.-Electron spin resonance of Cr/SiOz after vacuum treatment at 500°C and contact withCO at 20°C.(a) spectrum at 20°C, magnification 60 x ; (b) spectrum at -253"C, magnification 1 xHere another aspect of the surface chemistry of Cr5+ complexes on silica will bedeveloped : the reversibility and interchangeability of the various rearrangements.When applying a treatment with HCl and NH3 to a Cr/SiOz that is calcined in air at500°C and to a Cr/SiO2 that is first evacuated at 500"C, similar products are obtainedin similar quantities.A treatment at room temperature, first with HCl 'and thenwith NH3, gives the result of the NH3 treatment alone; a treatment with NH3 andTABLE 3.-QUANTITES OF Cr5+ IN CrlSi02 AFTER VARIOUS TREATMENTS(relative amounts)vat. 500°Cair 500°Cvac. 500"C, H20 20°Cvac.500"C, HCl 20°Cvac. 500"C, N H 3 20°Cvac. 500"C, CO 20°Cvac. 500"C, CO 200°Cvac. 500"C, C2H4 100°C:cro4)3-10018nilnilnil593018iCrO)J+<14425140tlt l(1then with HCl gives the result of the HCl treatment alone. Apparently, the sensitivityof the Cr/SiO:! system to surface rearrangements is not at all characteristic of thehighly dehydrated surface that is obtained after evacuation at 500°C. This pointhas been further elucidated by studying Cr5+ in a freshly impregnated Cr/SiO2 systemthat was only dried in vacuum at 100°C. The original product and the results of thetreatments at 20°C with HCl and NH3 are shown in fig. 7. By comparison withfig. 3 it is seen that the species obtained are completely analogous to those formedafter high-temperature treatments.Only the amounts of Cr5+ are smaller than thoseof fig. 3. Several of the treatments lead to a substantial reduction of the Cr6+ in thecatalyst to Cr3+. Evidently, the Cr5+ species is relatively stable as compared wit284 E.S.R. OF Cr5+ AND V4+ SURFACE COMPLEXESCr6f. The Cr3f formed after treatment with HCl is responsible for the broad andsymmetric band that forms the background for the Cr5+ signal in fig. 7(b). Similarbackgrounds are also found after treatments with CH3OH and CH3COH at roomtemperature and after treatments with CO and CzH4 at elevated temperatures.(a I","1FIG. 7.-Electron spin resonance of Cr/SiOz, dried in vacuum at 100°C ; spectra at -253°C.(a) without further treatment, magnification 12 x ;(b) after contact with HC1 at 20°C, magnification 2Ox ;(c) after contact with N H 3 at 20"C, magnification 8 x .Cr/A120 3Cr5+ in alumina surfaces always gives rise to much broader e.s.r.spectra than Cr5+in silica surfaces. In consequence, the re-arrangements in the surface cannot beanalyzed so accurately as for the silica system. An example is given in fig. 8, wherethe spectra are given after vacuum treatment at 500°C and after subsequent admissionof water vapour at 20°C. Although the anisotropy of the two spectra is different,FIG. 8.-Electron spin resonance of Cr/Al2O3 after vacuum treatment at 500°C ; spectra at - 253°C.(a) without further treatment ; (b) after contact with H20 vapour at 20°C.the temperature dependence shows that both species are in a square pyramidal co-ordination.Apparently this kind of co-ordination is well-stabirized in the surfaceof A1203. It does show rearrangements due to contact with HzO, HCl or NH3 atroom temperature, but the effects are less pronounced qualitatively and quantitativelythan in the silica system.It is not quite clear why the Cr5+ spectra for A 1 2 0 3 are broader than those for SiO2.Probably the line width is due to dipole-dipole interactions with the Cr3f ions alsL. L. VAN REIJEN A N D P. COSSEE 285present in the sample. In alumina most of the Cr3+ is present in solution in theA1203 lattice; in silica it is present mainly in a separate chromium-rich phase.Presumably Cr5+ and Cr3f are better separated spatially in silica than in alumina.V/SiO2The study of the V4+ ion in silica and alumina is attractive from two points ofview: (a) the square pyramidal co-ordination of V4+ in vanadyl compounds has:been well identified, for instance, by X-ray-diffraction single-crystal work and in-vestigated extensively spectroscopically and by electron spin resonance ;25 (6) owingto the spin I = 712 with its large magnetic moment of the V nucleus, e.s.r.spectra ofV4+ compounds give much detail.263 27l 1 1 1 1 I I ! nIb)FIG. 9.-Electron spin resonance of V/SiO2 after vacuum treatment at 500°C ; spectra at -253°C.(a) without further treatment, magnification 32 x ;(b) after contact with H20 vapour at 20"C, magnification 8 x .The various possible co-ordinations of V4f in silica are best illustrated by thee.s.r.spectra of fig. 9. After a vacuum treatment at 500"@, the spectrum is that of anisotropic species. The line width is strongly temperature-dependent. After admissionof water vapour at room temperature an anisotropic spectrum is found with a linewidth independent of temperature.According to the analysis given for Cr5+ the first species is (vo4)4-; the secondhas the square pyramidal co-ordination of the vanadyl ion. These conclusions arefurther substantiated in fig. 10 and table 4. Fig. 10 shows the e.s.r. spectrum of VC14well-dispersed by rapid freezing of a dilute solution in a mixture of suitable saturatedhydrocarbons. It is completely analogous to the spectrum of fig. 9(a). Table 4shows the parameters obtained from the spectrum of fig.9(b) and compares themwith those obtained for a rapidly frozen dilute solution of vanadyl ions. The corres-pondence of the two sets of parameters is very satisfactory286As shown in fig. 9, the (VO4)4- and the (V0)Zf Configuration are formed at silicasurfaces under the same conditions as the corresponding configurations of Crs+ are,viz., (vo4)4- after evacuation at 50O0C, (VO)2+ after subsequent admission of watervapour. Here the similarity ends. The amount of (VO)2+ after contact with watervapour is of the same magnitude as that of the (vo4)4- present originally. Apparentlythe (V0)Zf configuration is much more stable at the silica surface than the (CrO)3+configuration. The (V0);- configuration is much less stable than the (cro4)3-configuration to contact with CO and CzH4.After treatment with these gases mostof the V4f disappears and the remainder has exclusively the square pyramidal co-ordination.E.S.R. OF Cr5+ AND V4f SURFACE COMPLEXESFIG. 10.-Electron spin resonance of VCl4 in mixture of isopentane and 3-methylpentane;spectrum at -253°C.V/Al203In the surface of the alumina support, V4f is always found in the same vanadylconfiguration. The parameters can be accurately deduced from the e m . spectra(see table 4). They differ only slightly from those for the vanadyl configuration onsilica. Apparently in alumina surfaces the vanadyl configuration is stable, whereasthe (vo4)4- configuration is unstable. The amounts of V4+ are equal and small(1 % of total vanadium) after vacuum at 500°C and after subsequent contact withH20 vapour at 20°C.The amounts of V4f rise by a factor of 5-10 after contact at20°C with HCl, NH3 and even with CO.TABLE 4.-E.S.R. PARAMETERS OF v4'91 A 11 x 104 A~ x 104(cm-1) (cm-1)g IIV/SiO2 1 -922 1-982 182 72V/A1203 1.916 1.989 169 66solution of (VO)2+* 1.930 1.984 176 69* dilute solution of VoSo4 in mixture of H20 and CH30H, pH = 2, measured at -193°C.DISCUSSIONIn order to evaluate the significance of our experimental results for the mechanismof the polymerization of ethene, we consider a model for this reaction suggested byVoevodsky during the 3rd International Conference on Catalysis.21 According tothis model, the initiation step consists of the introduction of a vinyl group in thL.L. VAN REIJEN AND P. COSSEE 287surface complex (see fig. ll(a)). Evidence for this step had already been presentedin the literature on the Phillips polymerization process.28 An attractive feature isthat no change in the valency of the chromium ion is required. The propagation stepin this model consists of the introduction of an ethene molecule between the Cr5+ ionand the growing polymer (see fig. 1 l(b)).CH2- c*H2FIG. 1 1 .-Polymerization of CzH4 on Cr/SiOz, according to Voevodsky.It is tempting 5 to compare this mechanism with one proposed earlier for thestereospecific polymerization of propene by TiC13.29-31 A close analogy between thetwo mechanisms can be created by changing the co-ordination of the intermediates offig.1 l(b) into essentially octahedral. The propagation step may then be represented asin fig. 12.5"'CHI-,_ f" ii"f ,CHZC"2 /II 'FH2' I0-Cr-- i",: y 2 - O-L,O I ___c o-&/e,,zO'A q C"Z 0'1 ---';H20 0FIG. 12.-Polymerization of C2H4 on Cr/SiOz; mechanism analogous to that for the Ziegler-Natta polymerization.Two points about the Cr/SiOz system should be stressed in particular from ourexperiments. (a) In the fresh catalyst, after pretreatment in vacuum or with CO, theCr5+ ion is surrounded by a tetrahedron of oxygen anions. (b) The (CrO#- surfacecomplex is rather unstable and in contact with suitable reagents it either decomposesor transforms into a square pyramidal complex.These results suggest that the actual polymerization mechanism might be asynthesis of Voevodsky's and our proposals.Before admission of C2H4 the activespecies is tetrahedrally co-ordinated. Owing to the initiation step, the configurationchanges from tetrahedral to penta-co-ordinated (see fig. 13). The propagation steptakes place in this penta-co-ordinated complex, where an empty site is available toaccommodate the incoming monomer molecules.The evidence for our model is only circumstantial. No new e.s.r. signals havebeen observed after contact of a catalyst with C2H4. However, it is doubtful whethera complex like that of fig. 13(b) would show an e.s.r. signal. Although the co-ordination is fivefold, the local symmetry is different from that in the chromyl com-plexes and it is not certain whether the ?c-bonding with one of the oxygens can developso well as in the other complexes.It is quite possible that spin-lattice relaxation forsuch complexes is very fast and that electron spin resonance can be observed only athelium temperatures288 E.S.R. OF Cr5+ AND V4+ SURFACE COMPLEXESFrom the results obtained for the other systems, Cr/A1203, V/SiO2 and V/A1203,it is evident that on A1203 both for Cr and V the square pyramidal configuration isstrongly preferred to the tetrahedral configuration. For V/SiO2 the tetrahedralconfiguration is found, but the results of the treatments with CO and C2H4 show thatit is less stable than the corresponding complex of Cr5+. By contrast, the squarepyramidal co-ordination of V4+ in V/SiO2 is more stable than the correspondingconfiguration of Cr5+ in Cr/SiO2.This is seen in particular from the results of thetreatment with H20 at room temperature.+ C2H4-VFIG. 13.-Proposed initiation step for the polymerization of C2H4 on Cr/SiO2.The activity of Cr/A1203, V/SiO2 and V/Al203 for the polymerization of etheneis inferior to that of Cr/SiO2. How do these results fit in with our experimentalevidence and with the model proposed for the polymerization reaction ? Accordingto our views, the square pyramidal configurations in the surface of A1203 are toostable to become involved in the formation of vinyl complexes as required for theinitiation step of the polymerization reaction. On the other hand, the tetrahedralcomplex in the surface of V/SiO2 is too reactive towards C2H4 and is transformed intoa stable square pyramidal complex. Apparently, for Cr/SiO2 a satisfactory balanceis established where the (cro4)3- is sufficiently stable to be present in the fresh catalyst,sufficiently unstable to react with C2H4, and not easily de-activated by transitior,to a stable chromyl configuration.However, the catalyst maker is still confronted with a problem of surface chemistry :why is it that Cr5+ and V4+ differ in their inclination to choose either one co-ordinationor the other and how is the choice influenced by the surroundings of the ion? Thisis not a new problem; it is encountered in the study of ions like Cr6+, V5+ and V4+in aqueous solutions.For instance, Cr6+ is always found in tetrahedral co-ordination ;V5+ is generally found in tetrahedral co-ordination, but at low pH the octahedrallyco-ordinated (VO2)+ ion is formed.32, 33 V4f is frequently found in the squarepyramidal vanadyl configuration, but at increasing pH precipitates are formed,presumably also containing V4+ in tetrahedral co-ordination.34 The evidence aboutCr5+ is less systematic.In highly alkaline solutions the (cro4)3- ion is found,whereas in acid solutions the co-ordination is square pyramidal. In all these solutionsthe pH apparently is a governing factor in making a choice between 02-, (OH)-and H20 as ligands. At a high pH, the 0 2 - ligand is preferred. These ions nearlyalways give rise to tetrahedral MO4 anions, probably owing to the great mutualrepulsion of 0 2 - ligands.At a low pH, 0 2 - will be replaced by (OH)- and H20,which leads to neutral or slightly positive octahedral complex ions. Where the sizeof the central ion is slightly too small to accommodate six ligands, n-bonding to oneof the oxygens stabilizes a complex with five ligands occupying five of the six cornersof an octahedron.It is tempting to compare the co-ordination in aqueous solutions with those atthe surface of an oxide lattice. This is particularly so, because the experimentsdescribed here have shown that the flexibility of a surface is very great. AdmissioL. L. VAN REIJEN AND P. COSSEE 289of gases even at room temperature leads easily to rearrangements in the structure ofthe surface. At an oxide surface the building stones for the formation of the transitionmetal ion complexes are the same as in aqueous solutions: 02-, OH- and H2O.The rules, however, are different.As in solutions, there is a strong trend towards asymmetric and high co-ordination, especially of the highly charged metal ions. Incontrast with solutions, the choice between 0 2 - , OH- and H20 is limited, because thetotal number of hydrogens is determined by the degree of dehydration of the surface.Also, local stoichiometry should be maintained much more rigorously at surfaces.A detailed analysis of our experimental evidence along these lines is premature. Theaspects mentioned above, however, may be illustrated by one example. Evacuationleads to a decrease in the number of available ligands.Assuming that the trend ofeach metal ion is to remain symmetrically surrounded, it can do so by changing its co-ordination from octahedral to tetrahedral. Admission of H20 vapour will lead tothe reverse phenomenon ; this is found in the experimental work. The favourablerole of HC1 in the stabilization of square pyramidal complexes is probably due tothe same circumstance. After admission of HCl, one 0 2 - ligand is then replacedby one OH- and one C1-.It is a pleasure to acknowledge the skilful experimental assistance of J. Gaaf andG. van Bokhorst in the inorganic preparative work and of A. R. Korswagen inmeasuring the e.s.r. spectra.1 O’Reilly and MacIver, J. Physic. Chem., 1962, 66, 276.2 Cossee and Van Reijen, Proc. 2nd Int. Conf. Catalysis (Paris, 1960), p.1679.3 Kazansky and Pecnerskaya, Kinet. Catalysis U.S.S.R., 1961, 2,417.4 Bukanaeva, Pecherskd,*a, Kazansky and Dzisko, Kinet. Catalysis U.S.S.R., 1962, 3, 315.5 Habeshaw and Hill, Proc. 3rd Znt. Con$ Catalysis (Amsterdam, 1964), p. 975.6 Boreskov, Bukanaeva, Dzisko, Kazansky and Pecherskaya, Kinet. Catalysis U.S.S. R., 1964,7 Carrington, Ingram, Schonland and Symons, J. Chem. SOC., 1956, 4710.8 Scholder and Klemm, Angew. Chem., 1954, 66,461.9 Van Reijen, Cossee and Van Haren, J. Chem. Physics, 1963, 38, 572.10 Van Reijen, Thesis (Eindhoven, 1964).11 Gray and Hare, Inorg. Chem., 1962, 1, 363.12 Gray, Bernal and Hare, Inorg. Chem., 1962, 1, 831.13 Mishra and Symons, J. Chem. Soc., 1963, 4490.14 Kon, J. Znorg. Nuclear Chem., 1963, 25, 933.15 Kazansky and Pecherskaya, Kinet. CataZysis U.S.S.R., 1963, 4, 210.16 Poole, Kehl and MacIver, J. Catalysis, 1962, 1, 407.17 O’Reilly and Poole, J. Physic. Chem., 1963, 67, 1762.18 Poole and Itzel, J. Chem. Physics, 1963, 39, 3445.19 Poole and Itzel, J. Chem. Physics, 1964, 41, 287.20 Slinkin, Fedorovskaya and Rubinstein, Kinet. Catalysis U.S.S.R., 1963, 4, 199.21 Voevodsky, Proc. 3rd Int. Con$ Catalysis (Amsterdam, 1964), p. 88.22 Van Reijen, Sachtler, Cossee and Brouwer, Proc. 3rd Int. Con$ Catalysis (Amsterdam, 1964),23 Kon and Sharpless, J. Chem. Physics, 1965, 42, 906.24 Woodward, Creighton and Taylor, Trans. Faraday SOC., 1960, 56, 1267.25 Ballhausen and Gray, Znorg. Chem., 1962, 1, 111.26 Kivelson and Sai-Kwing Lee, J. Chem. Physics, 1964, 41, 1896.27 De Armond, Garrett and Gutowsky, J. Chem. Physics, 1965, 42, 1019.28 Smith, Ind. Eng. Chem., 1956, 48, 1161.29 Cossee, J. Catalysis, 1964, 3, SO.31 Arlman and Cossee, J. Catalysis, 1964, 3, 99.32Newman, La Fleur, Brousaides and Ross, J. Amer. Chem. SOC., 1958,80,4491.33 Newman and Quinlan, J. Amer. Chem. Soc., 1959, 81, 547.34 Rossotti, F. J. C., and Rossotti, H. S., Acta Chem. Scand., 1955, 9, 1177.5, 379.p. 829.30 Arlman, J. Catalysis, 1964, 3, 89

ISSN:0366-9033    

DOI:10.1039/DF9664100277

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Surface Co-ordination of Oxygen on Oxygen-Deficient Ti02and Moo3 as Revealed by E.S.R.-measurementsBY P. F. CORNAZ," J. H. C. VAN HOOFF, F. J. PLUIJM AND G. C. A. SCHUITDept. of Inorganic Chemistry, Technological University, Eindhoven, Insulindelaan 2,EindhovenReceiued 1 3 th January, 1966E.s.r. spectra were obtained from Ti02 (anatase) and Moo3 after evacuation at 500°C and sub-sequent addition of 0 2 at room temperature. The spectra observed for Ti02 are quite complex : byvarying the degree of reduction, the 02-pressure, the microwave power and the temperature fore.s.r. recording they were separated in a number of components. On degassed TiQ2 two of thecomponents are assigned respectively to an F-center and to Ti3+, both at the surface. A third signalremained unidentified.Addition of 0 2 causes the emergence of at least four components. Threeof these were assigned to 02-surface co-ordination complexes, viz., the L-form (similar to CN-co-ordination), the P-form (similar to C2H2-co-ordination) and the A-form (similar to H202). Thefourth signal is believed to be connected with 0- formed by H-abstraction from OH- presumablyby the peroxo-complexes.On Moo3 after evacuation four component signals were observed, one of which is assigned toan F-centre, a broader and more intense one to surface Mo~+, the remaining two signals remainingunidentified. On 02-addition none of the signals observed for Ti02 were seen to occur here, theinteraction with 0 2 being almost solely restricted to a decrease in intensity of the F-center andMos+-signals.The Bi203+Mo03 system after evacuation is similar to Moo3 except for theabsence of the Mos+-signal. On NiO no signals could be found either in the evacuated or in the" reoxidized " form. The possible significance of the presence or absence of the co-ordinated02-species for selective oxidation catalysis is discussed.Various transition metal oxides, e.g., V205 or MoO3, or their binary mixtureswith other oxides, e.g., SnOz+VzOs or Bi203+Mo03 act as catalysts for selectiveoxidation of unsaturated hydrocarbons. Naphthalene or o-xylene are convertedto phthalic anhydride over V205, while propene gives acrolein and 1-butene is con-verted to butadiene over Moo3 + Biz03 catalysts. However, these catalytic reactionsare exceptional as to their specificity : on other and often related systems, oxidationto COz and HzO is observed.There should therefore be some special propertyof the particular catalyst systems that determines the degree to which it allows aselective conversion in contrast to a " total " oxidation.Now, in some reactions, viz., the conversion of propene to acrolein or of 1-butene to butadiene the " activation " of the hydrocarbon by the catalyst appearsunderstandable. By the work of Adarns and Jennings 1 who operated with deuter-ated propene, and of Sachtler 2 who applied propene " tagged " with W, the firstreaction is the formation of an allylic intermediate by donation of a H-atom to thesurface of the oxide catalyst. This is equivalent to saying that the catalyst is" reduced " by the hydrocarbon.Work by Batist3 has proved that at least for the conversion of 1-butene tobutadiene the overall reaction can be divided into two separate processes, viz.,(i) the reduction of the CdtdYSt, Moo3 or the Bi203fMoO3 system (Bi/Mo = 48/52* on leave of absence from Laboratoire de Physique de 1'Ecole Polytechnique de 1'Universite290de Lausanne, SwitzerlandCORNAZ, VAN HOOFF, PLUIJM AND SCHUIT 29 1at/at) under the formation of butadiene, H20 and an oxygen-depleted oxide structure ;(ii) reoxidation of the oxygen depleted oxide by interaction with gaseous oxygen.Now, although we are relatively well informed about the mechanism of thereduction, our knowledge concerning the reoxidation of oxygen-depleted transitionmetal oxides is incomplete.Kazansky and Pariisky 4 who studied the adsorptionof 0 2 on slightly reduced anatase (TiO2) by e.s.r.-measurements reports two signals,a narrow one the nature of which was not clear and a broad one with marked aniso-tropy (911 = 2-021, gl = 2.011) which they ascribe to an “ adsorbed peroxide ”.The latter disappears rapidly if acetaldehyde is introduced; this is considered toprove the identification of the species connected with the broad signal.Kokes 5 studied the adsorption of 0 2 on partly reduced ZnO also by means ofe.s.r.-measurements. The evacuated ZnO shows a relatively sharp signal at g =1-96. This signal diminishes somewhat if 0 2 is added but never disappears. How-ever, a second broader and anisotropic signal is then formed (gl = 2.0102 and gll =2-0025).A similar signal is also observed after allowing 14202 to interact with theevacuated sample. Kokes therefore ascribes this signal to a ‘‘ peroxo-group witha covalent bond to the surface S-0-0- or electrostatically bound 0; species”.The important information contained in these papers is that e.s.r.-measurementsform a valuable tool for the identification of the adsorbed 02-species on the surfaceand that one of these species is comparable to a peroxo-radical. We have thereforetried to obtain some information via e.s.r.-measurements on the following systems :(a) the anatase + 0 2 system reinvestigated here because the description given byKazansky and Pariisky on the system is not very informative; (b) M003+02,the reduced Mo03-samples being obtained by either incipient reduction with 1-butene or evacuation at higher temperatures ; (c) the Bi/Mo oxidef02 system inwhich reduction was effected by 1-butene ; (d) the NiO+02 system.The lattersystem calls for an additional discussion. Haber and Stone 6 have reported thatNiO, evacuated at higher temperatures and subsequently contacted with 0 2 at roomtemperature became photoactive. If irradiated it dissociates some 0 2 , a processthat proved reversible after cutting-off the illumination. Jongepier 7 has shownthat the Haber-Stone effect is mainly caused by thermal effects, the sample beingheated by the incident light. Unpublished experiments by Th. P. M. Beelen andF. L. M. A. H. de Laat in our laboratory have confirmed this explanation, However,some results of Beelen gave the impression that the oxygen revealed in the Haber-Stone experiments must be present in some special situation on the surface.Freshlyevacuated NiO when contacted with 0 2 readily adsorbed this gas and then becamepho toactive.However, after some time the potential activity for photodesorption disappeared.Only on further additions of 0 2 did the effect become stable in time. This seems tosuggest a two-stage adsorption.0 2 -+X -+Y (possibly 02-)in which only X is liable to photodesorption and cannot be converted to a morestable form if the surface is largely occupied by it. In view of the measurementsby Kokes it might be considered that X also is a peroxo-radical.EXPERIMENTALMATERIALSTi02.-TiC14 (Riedel-de Haen 140-12) was distilled in a dry nitrogen current from avessel filled with copper turnings.After repeating this procedure twice the resultant purifiedcompound (b.p. 1 3 5 . X ) was added to water at 0°C. Subsequent to a hydrolysis wit292 OXYGEN ON Ti02 AND M o o 3ammonia the resulting precipitate was filtered, washed until no Cl- could be observedin the washwater and dried at 120°C; the product was white. It was then precalcinedaccording to the following methods.SAMPLE A : calcination at 500°C for 4 h in a current of dry oxygen ;SAMPLE B : as A, but in a current of air saturated with water vapour ( P H ~ o = 14 mm Hg) ;SAMPLE C : as for B, but the air being saturated with D20 vapour (PD,o = 15 mrn Hg).The colour of A remained white but B and C became weakly yellow.X-ray diagramsshowed all the samples to possess the anatase-crystal structure.All samples were then pretreated in the apparatus shown in fig. 1 under a vacuum of10-4 torr during 4 h at 500°C. This pretreatment led to a slight and somewhat variablereduction of the samples. Incipient reduction is demonstrated by the samples becomingfirst slightly yellow and later gray, Subsequent to the reduction the samples were trans-ferred to a number of quartz tubes of diam. 4mm. One of these* tubes was sealed offunder vacuum at room temperature. Three others were sealed off after addition of 0 2to the previously evacuated apparatus also at room temperature and at oxygen pressuresof respectively 0.5, 10 and 100 torr.The samples were then studied in the electron spinresonance apparatus.FIG. 1.-Apparatus for the preparation of samples to be investigated in the e.s.r. apparatus : A,sample tubes; B, furnace; C, high-vacuum line; D, manometer; E, Pirani gauge; F, oxygen-dosing system.Moo3 AND Bi203+M003 (Bi/Mo = 48/52 at. %).--The MoQ3, a commercial sample(Baker 1144), was pretreated as for Ti@, viz., evacuation at 500°C followed for some samplesby Oraddition at room temperature and various pressures. One sample was reducedby subjecting it to 1-butene (pressure 10 torr) for 4 h at 350°C in the apparatus mentionedabove. It was then cooled to room temperature, transferred to a quartz tube and sealedoff. From other experiments this sample could be expected to be slightly reduced.The Bi + Mo binary oxide system was prepared by mixing powdered NH4 paramolybdate(Merck 1182) and Bi nitrate (Merck 1860) in the required ratio and subsequent slow heatingof the mixture to 300°C.The product was cooled, powdered and subjected to a secondcalcination for 4 h at 500°C. It was reduced as for Ti02 but at a lower temperature (400°C)since it is more easily reduced under vacuum.From X-ray measurements it is a mixture of two compounds (see Bleyenberg, Lippensand Schuit 8 ) that may be described as Bi2(Mo04)3 and (BiO)~(Mo04) the latter beingidentical with the mineral koechlinite. It has a strong yellow colour and represents themost active Bi/Mo oxide catalyst for the oxidative dehydrogenation of 1-butene to buta-diene encountered by usCORNAZ, VAN HOOFF, PLUIJM AND SCHUIT 293NO.-Preparation of samples identical to that for TiO2, i.e., evacuation of NiO (Bake1153) at 500°C cooling and transferring samples to the quartz tubes under vacuum andafter addition of 0 2 .APPARATUS FOR MEASURING E.S.R.-SPECTRAThe apparatus was the Varian E.P.R.spectrometer type V4500 at a frequency of about9.3 GHz with a variable temperature accessory V4540. Spectra were obtained at tem-peratures varying between - 195°C (liquid N2) and + 20°C. The modulation frequencywas 100 kHz. This frequency is adequate for all signals except one narrow line, quoted(12) below, for which it was definitely too high resulting in a non-reproducible line shape.For all other signals the records are derivatives of the adsorption spectrum.RESULTSThe interpretation of the e.s.r.spectra is complicated by the simultaneous presenceof several signals, most of which are typical powder spectra with two or three g-values. Fig. 2 gives the characteristic shape of the absorption and its derivativefor such lines with the corresponding posi-tion of the principal g-values. Their peculiarshape is essentially given by the shape of theprobability distribution of the different orien-tations resulting in a given g-value (Bleaney,gKneubuhl 10).I I , II t ' II I ; :I : t I , I I E . S . R . RESULTS ON Ti02The separation of the spectrum into dif-ferent characteristic signals has been achievedby varying, besides the degree of reductionand the oxygen pressure, also extensively themicrowave power (study of saturation pro-perties of the different parts in a signal) andthe temperature of measurement (study ofthe temperature dependence of the differentcomponents). The results of this study aregiven in fig.3, 4 which list the 6 signalswhich are consistently found in all our wectrawith variable intensities. (No other signals axial symmetry l o w e r s y m m e t r yhave been found except of an extremely low 2 g-values 3 g-valuesintensity; these may be due to hyperfine FIG. 2.--Twical shapes of e.s.r.-powdercoupling effects). spectra for compounds with axial and lowerbility densities for giveng-values. Thelower c, see above) does not have an essential curves are the derivatives of the compositeinfluence on the spectra.Only the samples absorption spectra for various line-widths.C prepared with I 3 2 0 show some differenceto those prepared with H20 in what is believed to be hyperfine structure ofthe line d2. The effects are too small to draw conclusions. The degree of re-duction prior to oxgyen adsorption has an important influence. The weaklyreduced samples show remarkable differences from the strongly reduced ones. Thesedifferences, however, can most often be explained by differences in intensity andline width of the fundamental signals of fig. 2. It is important that it has beenpossible to ascertain the existence of several distinct 3-g-value signals. This facilit-ates their correlation to special models of adsorbed radicals by taking into accountThe mode of preparation (A, B and symletry. The are proba-1294 OXYGEN ON Ti02 AND Moo3a---.- 1 . . --_-I----205 2. a0 1.95 1-90 gFIG. 3.-Separate spectra as observed on evacuated TiOz-samples. In actual recordings some ofthese spectra occur simultaneously.-__ __ &*-,-----c---.-- - l - - - a - - - L . - - l -I205 2.00 1.9 5 1.93 gFIG. 4.-Separate spectra as observed on Ti02 samples after contacted with 0 2 . In actual re-cordings they may occur simultaneouslyCORNAZ, VAN HOOFF, PLUIJM AND SCHUXT 295their symmetry properties. The following signals were observed on evacuatedsamples.SIGNAL ( a ) : on yellow samples at temperatures below -140°C; absent on graysamples.SIGNA-L (b) : a three g-value powder signal on yellow samples s t T< - 140°C.On gray samples this signal is replaced by an asymmetric broad and very intenseline b’ (at T< -- 140°C only).SIGNAL (f) 4- (h) : apparently two separate signals with single g-values ; intenseon gray samples at room temperature, much weaker on yellow samples.The (f+ h) signal appears siniilar in properties to an F-center signal (long relaxa-tion time, especially for (12)).We tentatively associate it with such a center. Thesignal (b) is almost certainly due to Ti3-t. A Ti3f-ion in ideal octahedral environ-ment would not show any observable signal above liquid-Ha temperature becauseof spin lattice relaxation (van Reyen 11). But at the surface the symmetry of a Ti3+ion is lower and might be observable at liquid-Nz temperature.An interestingfact is given by this signal. At low concentration (yellow samples) the electronis fixed (3 g-values), but at higher Ti3+ concentration (gray) the signal does not showthis structure but only the mean g-value. This means that the electron is in rapidmovement from one Ti3f to another (motional or exchange narrowing) and cor-responds to the high conductivity of these samples reported by Kasansky andPariisky.4 In agreement at room temperature the Ti3f are completely ionizedand the signal (b) disappears. The signal (a) is not easily assigned but its almostequal temperature dependence as (b) may suggest that it could be associated to anotherform of Ti3-1-.On Oradmittance, signal (6) decreases in agreement with its assignment as surfaceTi3+.The (u) and (J’)+(h) signals decrease som-ewhat but remain observable evenat highest 0 2 pressures applied here. They are therefore presumably connected withcenters situated deeper in the bulk and are therefore not of immediate interest tous. However, our (f) signal is similar to the sharp signal reported by Kasanskyand Pariisky4 as formed when small amounts of 0 2 are admitted to the evacuatedsample. We therefore disagree with them as to the origin of the signal : accordingto us it is an artifact of the oxygen-depleted situation.The following new signals are formed after the addition of 0 2 (fig. 4-6).SIGNAL (c): a broad intense signal on gray samples at room temperature.Itbecomes narrower at lower temperature and at lower 0 2 pressure. Possibly it isalso weakly present on yellow samples.SIGNAL ( d ) : at room temperature on yellow samples, no change between 10and 100 torr 0 7 and almost no change with teniperature; a very narrow 3-g-valueline, easily saturated; also observable on gray samples but weaker and obscuredobserved both on yellow and gray samples but more intense on thelatter. Becomes observable only below - 120°C and then becomes much strongerthan the (c) signal at - 160°C.The signal (c) is similar to that reported by Kazansky and Pariisky 4 as a broadsignal; they do not mention the other signals. However, signals (e) and (h) arevery similar to those reported by Kokes 5 on addition of 0 2 on degassed ZnO.Signal (d) is not mentioned by either of these authors; however, since it is easilysaturated, and in gray samples masked by ( c ) , it may readily escape observation.To show how the different signals overlap in the recorded spectra fig.5 and 6by (c).SIGNAL (e296 OXYGEN ON Ti02 AND MOO3I Torr QFIG. %-Typical recordings of e.s.r. spectra observed on ‘ I gray ” samples of Ti02 ; spectra re-corded at room temperature, oxygen pressure variable ; assignments of spectra according tofig. 3 and 4.FIG. but recordinCORNAZ, VAN HOOFF, PLUIJM AND SCHUIT 297are given. Fig. 5 shows the spectra of a " gray " TiOa-sample exposed to oxygenat various pressures and all recorded at room temperature. Fig. 6 shows spectraobtained on the same sample but under a constant 02-pressure (100 torr) and variablerecording temperatures.j:RESULTS ON Moo3Neither the degassed Mo03-samples, nor the samples degassed and subsequentlycontacted with oxygen, show any spectra at room temperature.At -195"C,however, spectra become observable that change in character upon contactingthe sample with oxygen at various pressures. The spectra are relatively simpleand therefore are all shown in fig. 7. They show the following details.- 195 O C100 Oe03 Torr 02FIG. 7.-E.s.r. spectra observed for degassed MoO3, subsequently contacted with 0 2 at variouspressures ; a11 recordings at - 196°C.Evacuated Moo3 (similar to sample reduced by butene) :signal ( I ) : a sharp signal at g = 2.001 ;signal (kl) and (k.) : two sharp lines on both sides of Z(g = 2-035 and 1.960) ;signal (m) : a relatively broad signal at g = 1.943.On adding 0 2 both ( I ) and (m) decrease in intensity while (kl) and (k2) remainunaltered.At a pressure of 12 torr there appear two lines (g = 1.973 and 1.882)denoted by n1 and n2. At still higher pressures these disappear again. Through-out the procedure the (kl) + (k2) system remains unaltered.None of the spectra accompanying the interaction of degassed Ti02 with 0 2are observed with Mo03. Practically the sole evidence for a reaction is given bythe partial disappearance of signals ( I ) and (m)295 OXYGEN ON Ti02 AND Moo3RI-MOLYBDATEThe spectra of the degassed sample appear similar to those of degassed Moo3except that there is no signal (m).This is perhaps ascribable to a reaction such aswhich is inherently plausible because Bi3+ is reduced in the reduction. However,other explanations are possible. In view of the scarcity of information derivedby e.s.r.-methods from this sample it was not further investigated.3Mos+ + Bi3+ ->3IvIo6+ + Bi",NiONeither the degassed nor the re-oxidized sample show any trace of an e.s.r.-signal at temperature > - 195°C ; it was not further investigated.DISCUSSIONThe anatase-structure of Ti02 can be derived from a cubic close packing ofoxygen anions with half of the octahedral sites occupied by Ti4+-ions, the otherhalf remaining empty. The actual packing is tetragonally distorted so that fourof the metal-oxygen distance are equal to 1937A and two others to 1.964A(Wells 12).The situation at the surface of a crystal may be discussed in the follow-ing manner. Consider the outer anion-layer and compare this outer layer to asimilar one that is situated in the bulk of the solid. For reasons of stoichiometrythe outer layer must show point defects, either anion vacancies, or OH- ions insteadof 02--ions. The possible presence of H-atoms is readily explained from themethod of preparation of the oxide that normally involves a precipitation from anaqueous solution. In principle, the OH--ions can be decomposed by calcinationat an elevated temperature according to the reaction :Cl being an anion vacancy.However, to arrive at a H-free surface the calcination temperature must beso high that the structure is converted to the rutile-type.It can therefore be as-sumed that the surface layer contains 02-, QH- and anion vacancies. In the anionvacancies the Ti4+-ions become exposed. It may be assumed that the oxygensurrounding of such exposed cations undergoes a further distortion, so that theremaining oxygen anions are displaced from their original positions.An ideal NiO crystal possesses the cubic close-packing of oxygen anions withall octahedral sites occupied by Ni2+-ions (rock-salt structure). Since it is easier toexpel all H20 from the structure it is considered that the surface layers containonly 02--ions, anion vacancies and exposed Niz+-ions in incomplete and distortedresidues of the original octahedral surrounding.Moo3 (see Wells 12) possesses a layer structure.Again the central cationsMom are octahedrally surrounded by 02--ions. These octahedra, however, sharetwo of their edges with their neighbouring octahedra in a certain plane of reference,while perpendicular to this plane the octahedra are linked by sharing corners. Thestructure therefore becomes a double-layer structure. The flanking planes cfoxygen anions of these double layers are probaly close-packed, while point defectsoccur preferentially at the edges of the layers. Again, these can be anion vacanciesor OH--ions. The information concerning the crystal structure of the bismuthmolybdate catalyst is scanty. One of its components, koechlinite or (Bi0)2(Mo04),is reasonably well-known : it possesses a complicated structure related to2 OH--+O2-+H20+CI, (1CORNAZ, VAN HOOFF, PLUIJM AND SCHUIT 299(La0)2(Mo04) in which the Bi3f-ions occur in tetragonal metal-oxygen-metal layersand between these layers, sliects of M o V I in distorted tetrahedra! surroundings (SillCnand Eundborg).l3 The other component with composition &(MOO& possessesan unknown structure, and it is believed that also here MovI occurs in tetrahedraloxygen-surroundings.If these compounds lose some of their oxygen and thus become oxygen deficienteither by reduction or by pumping at a relatively low temperature, the parent struc-ture may be left virtually unaltered except that at some sites in the crystal, anionvacancies will be formed.The concentration of these vacancies may be highernear or at the surface because a concentration gradient will develop as a consequer,ceof the removal of oxygen from the surface.The electrons released by the dissoci-ation of 0 2 will in iirst instance be trapped in anion vacancies thus forming F-centers(anion vacancy with single electron). Another possibility is that they becomeassociated with the metal cations. The F-center is a state with well-defined pro-perties in the spectroscopic and electron spin resonance sense. It has an averageg-value close to, but. somewhat lower than that of free-electron value of 2.0023(Kittel 14). The electron trapped on a cation, in our cases invariably a d-electron,possesses a g-value that is considerably lower ( N 1-95) than the free electron be-cause of spin orbit coupling.There may be more than one g-value (anisotropicsignal) because the symmetry of the site may be different from the cubic one.Signals of this kind may, moreover, be unobservable because of line-broadeningconsequent to spin-lattice relaxation (see, e.g., van Reyen 11). The applicationof lower measuring temperatures may lead to the detection of the signal althoughfor some cases (cZ1 system in octahedral surrounding) one has to apply liquid-H2or He temperatures to observe it. Finally, since some of the signals to be expectedwill show a marked anisotropy, especially when it is connected with a surface state,the fact that our measurements have been performed on powders will lead to an" averaging over positions " (Bleaney 9).On these oxygen-deficient structures 0 2 is now allowed to adsorb.The generalpractice in discussing the interaction of Q2 is to consider a number of possibleanions such as O,, 0- as intermediates in the final formation of 0 2 - anions.Applicatioils of the ligand-field theory as applied to co-ordination complexes leads,however, to the belief that these intermediates and especially the 0; are insufficientlycharacterized by the electrostatic model. In the following we therefore try todiscuss more deeply the modes of bonding of 0 2 .In terms of the molecular orbital theory the 02-molecule contains a number ofm.0. that can be described by linear combination of the atomic orbitals (a.0.) of theatoms. Following Ballhausen and Gray 15 the 0-atom is first hybridized by formingtwo hybrids ++ = 2ss+2p, and Cp- = 2s,-2pZ (normalization factors omitted).Now the ma.formed are (A and B are the two 0-atoms) :1. Q = ++(A)+#+(W o-bonding orbital,(9 = ++(w-++(B) o*-antibonding orbital,2. n, = 2P,(A)+2P,(B) 1n-bonding and antibonding orbitals, inon-bonding orbitals 300 OXYGEN ON Ti02 AND Moo3The configuration of 0 2 is then a2+?(A)+?(B) nzn$nzn;. If the 02-molecule isallowed to interact with a transition metal ion by occupying an anion vacancy inan octahedral surrounding we can do this in two ways.LINEAR STRUCTURE LThis can be depicted as M - 0 ~ 0 ; one of the lone pairs 4- is used to form ao-bond in the ligand a~,+e,+t~, set that interacts with the 4s, 4p and e,-d-orbitalsof the cation.The two anti-bonding orbitals nz and n; are then forming n-bondswith two of the t2,-dorbitals of the cation. The situation is similar to that encoun-tered in complexes containing CN- ligands. If the cation is a dl-system therewill be two bonding n-orbitals that are largely &orbitals one of which is filled withan electron pair, the other being occupied singly.PERPENDICULAR SITUATION P0This can be depicted as M t I , similar in properties to the0unsaturates. The orbital that takes part in forming o-bonds withco-ordination ofthe cation is oneof the n-bonding orbitals of 0 2 , say n,. Its counterpart, the antibonding n$orbitalcan then engage in the formation of a co-ordination n-bond by interaction withone of the ta,-d-orbitals of the cation.This fixes the position of the 0 2 molecule andthe situation of the remaining ny and n; system is entirely different from the formercase since it cannot form a an-bond. If the transition metal ion is a dl-system,the n-bond formed with help of the n:-orbital will contain an electron pair whilethe single electron will remain in the n;-bond. The bond is less strong and will becharacterized by an anisotropic signal with three g-values and the centre at virtuallythe free electron position. The P-arrangement is realized for instance in substancessuch as K3CrvOg.ANGULAR ARRANGEMENT A. None of the two arrangements given above is adequate to explain a situationas encountered in a molecule such as H 2 0 2 and there must therefore exist anotherarrangement. The essential feature in this arrangement is the formation of anelectron pair bond by one electron from 0 2 and one from the cation that hencedonates an electron and in doing so becomes oxidized.An alternative formulationconsists in using an electron from an I;-center but since this will be situated nextto a Tim-ion the final situation is identical to that obtained by using the d-electronin Tim. This, therefore, is the background of descriptions such asorTi3+ + 023Ti4+0;,Ti4+ + e + 02+Ti4+0;.The structure of the resulting Ti-0-0 arrangement cannot be linear for the samereasons that €3202 possesses a " helical " structure. One way of describing thestructure is to assume sp3-hybridization of the 0-atoms. Every 0-atom thenpossesses four hybrids, say 41, 4 2 , 43 and 4 4 .The electronic arrangement of 0 2then becomes#1(A) + +1(B) = a-bond with two electrons ;42(A), 42(B) = two lone pairs with four electrons ;43(A), 43(B) = two lone pairs with four electrons ;+4(A), 44(B) = two singly-occupied sp3-hybrids on each of the two 0-atomsCORNAZ, VAN HOOFF, PLUIJM AND SCHUIT 30 1The bond with the metal-ion is formed with the help of one of the singly-occupiedhybrids, say C$4(A) leaving a single electron in the equivalent orbital on B. Since&(A) and 44A) form the tetrahedral angle of 109" the M-0-0 structure is notlinear. This and the fact that the other electron used in the M-0 bond originatesfrom an 02--ion initially present are the main features of the A-arrangement.Thelatter ensures that it also can be formed on non-transition metal ions.Summarizing, we expect the following types of 0 2 bonded to surface metal ions...00M-0 MF 11 M r O z ~ 0 :"\ *. .D.A P LWhich of the three arrangements occurs depends on the number of d-electrons inthe cation and has to be discussed for each type of cation separately.It is first necessary to '' signalize " the possible occurrence of a species with asingle electron on one of its orbitals that is not a peroxide-structure but whose presencemay be connected with that of the peroxide structures. If a surface contains OH--ions there might occur a reaction such aswhere 0; is a peroxide-structure of the type A or P. These might therefore bytheir presence form the 0--ion, i.e., they would cause the observation of an e.s.r.-signal without being directly connected with this signal.Actually it is not certainwhether 0- can cause an observable signal, but the a priori possibility should bekept in mind, especially when 0- is present on the surface. We now discuss thevarious types of signals reported in this paper.OH-+ 0; +O-+ OOH (2)Ti02For this compound at least three and possibly four signals have been observedsubsequent to addition of 0 2 while from the theory four possible situations areexpected. The question is now whether we can pair them off. In both the L-and P-arrangements, ligand electrons are donated to metal tzg-orbitals which is onlypossible if these are only partly occupied. Therefore P and L cannot be reconciledwith 0 2 adsorption on ZnO: whatever the state of the Zn-ion it certainly has itst2,-orbitals fully occupied.We therefore ascribe Kokes' signal C and thereforealso our signal (e) to the A-arrangement in agreement with Kokes' identification ofthe species as M-0-0 or M+O;. In the experiments of Kokes the signal isobserved at room temperature while we see it only at temperatures < - 120°C.We believe that on Ti02 the species responsible for the (e)-signal is only formedat lower temperatures and identify it therefore with the low temperature adsorptionreported by Rapoport.16 Actually, the A-arrangement being the least stronglybonded is expected to occur at lower temperatures than P- and L-.Both in Kokes' experiments and in ours there occurs the small sharp signal(h).In both cases it is similar to a signal already observed for degassed samples((J') in our notation). Kwan,l7 in discussing Kokes' results has reported that thereare indeed two signals at g = 2.002, one of which is connected with the peroxo-system and that disappears by heating at higher temperatures while the other oneremains intact. He ascribes this signal to 0-. We tend to agree with him in thatthere are two signals with equal g-value. Looking at fig. 5, the impression is gaine302 OXYGEN ON Ti02 AND Moo3that the (f>-signal decreases with increasing 02-pressure ~ h i l e the (h)-signal increasesalso suggesting that they belong to different species. It is also suggestive in con-nection with Kwan’s assignment that 0- could be formed according to reaction (2)and therefore should increase with the concentration of peroxy-ions.Having thus identified two of our signals we are left with two possible structuresP and L and two sets of signals ( c ) and (d).Now (c) is a broad signal and there musttherefore be an element in the model structure to explain the line-broadening. TheL-structure contains three electrons in two orbitals both of which are bonding com-binations of n-antibonding orbitals and tZ,-orbitals. This bonding level is there-fore orbitally degenerate and we should not expect to see any signal. However,as a consequence of the Jahn-Teller effect, the degeneracy is probably removed byan orthorhombic distortion of the already tetragonally distorted octahedron.Thisinfers that, due to spin-lattice relaxation, the signal should be broadened at highertemperatures and become sharper at lower temperatures. This description fits the(c)-signal. The sharpening of the signal at lower temperatures is barely visibleat the lower temperatures (see fig. 6 ) : unfortunately it is obscured considerablyby the growing intensity of the (e) signal.The P-arrangement must lead to a signal that is essentially a “ hole on the ligand ”signal and therefore similar, although not equal, to that of the A-arrangement.There are no obvious reasons to suspect line-broadening for this arrangement andthe combined evidence therefore leads to assigning the (d-) signal to a P-structure.Our final assignments therefore become :signal (c) = L-arrangement ;signal (e) = A-arrangement ;signal (d) = P-arrangement ;signal (12) = 0-.Moo3The e.s.r.signals observed on samples obtained by degassing on the one handand by a slight reduction on the other hand are virtually identical. The secondtype of sample might contain hydrocarbon and especialiy allyl-radicals since thelatter are assumed to occur as intermediates during the reduction. Since the allyl-rddicdl contains three electrons it therefore should, in principle, cause us to expectan extra signal on the samples reduced by 1-butene. However, according to thereaction mechanism given by Batist, Lippens and Schuit 18 the allylic intermediate,if present, occurs as four-electron an-complex with Mo and therefore should not bediscernible since the energy levels of the complex are such that the four electronsshould occur in pairs.The e.s.r. observations are therefore consistent with thismechanism although they cannot be considered to prove its correctness.The signals observed on oxygen deficient Moo3 are : (a) a relatively strong signal(m) (g = 1-94); (b) a sharp signal (Ij at g = 2.001 ; (c) what seems to be a signalconsisting of two lines on both sides but not equidistant from (I).Signal (nz) is a MoV signal; signal (1) might be an F-center. The nature of thecombination (kl) and (k2) is not readily apparent: they might represent pairs ofF-centers with zero field splitting.On addition of 0 2 both ( I ) and (m) decrease considerably, in agreement with theirassignment, indicating that the species responsible for the signals are mostly situatedat the surface.Some of the centres, however, might be situated in the interior,since the signals do not disappear completely at higher 02-pressures. The (kl) + ( k 2 )combination does not show any alteration : it must therefore be a structural defectin the bulk. Tentatively, we might assign it to two F-centers on different MoO3-layers but close enough to show zero-field splittingCORNAZ, VAN HOOFF, PLUIJM AND SCHUIT 303The remarkable observation is made that the decrease in signal intensity of(m) and ( I ) is not combined with the formation of any of the signals observed earlierfor peroxide fwmation. At relatively high 0 2 pressures a double signal consistingof two peaks develops that disappears again at still higher 0 2 pressures perhapsbecause of physical adsorption of 0 2 .We have no explanation for this signal andwe shall therefore limit the discussion to the explanation of the non-appearanceof the Ti02 peroxide signals.Now, Moo3 differs from Ti02 in two respects : (a) the Mo-ions occur in stringsand in these strings Mo-ions are relatively close to each other because of the struc-tural features of octahedra sharing edges. Moreover, the final state of the reduc-tion is Mom and not Mov. We therefore believe that electron jumps such as2 Mo5++Mo6++Moi+ occur easily. (b) In between the Mo-layers there are 0-layers that do not contain cations and that therefore are in close contact.Property(a) may cause all the structures postulated earlier to contain an extra-electron sothat their number of electrons would become even instead of odd. Property (b)enables the formation of intermediate structures in which the 0 2 molecules arebonded to two Mo-ions at the same time. This could be visualized in the followingmodels :L-arrangementA-arrangement. M-0-0-M from M=O=O ;M-0 from M-00.\0-M\P-arrangement0 0M- 111 from M- I/[ o\ 0.MIt is difficult to say which of the two alternatives occur preferentially.In both the Biz03 + Mo03 and NiO systems the signals observed, if any, proveto contain no information that helps in discussion of the consequences of the removaland addition of 0 2 . Clearly, electron spin resonance for the compounds mentionedis not an adequate method to study these reactions.FINAL REMARKSIt remains to be discussed whether the presence or absence of 02-surface co-ordination complexes may be relevant in explaining the selective oxidative dehydro-genation of olefins or related reactions over oxides such as hloO3.We believe that it is reasonably certain that the selective oxidative dehydrogena-tion of butene to butadiene can be represented by reduction of the oxide catalystfollowed by re-oxidation of the oxygen-deficient catalyst by air or oxygen.Thisbelief is based on the observation that the rate of the reduction is similar to that ofthe oxidative dehydrogenation and that the hydrocarbon products are exactly thesame for the two cases.Therefore, the only oxygen source needed for this particularreaction is 02--ions and the only thing expected from the gaseous 0 2 is to maintainthe concentration of 02--ions intact. There is no need for the presence of 0 2 -surface co-ordination complexes : they might even complicate the situation b304 OXYGEN ON Ti02 AND Moo3creating other types of oxidation. Therefore, the apparent absence of these com-plexes on M083, a selective but not very active catalyst for the oxidation of buteneto butadiene, appears well in line with the expectation. However, the absence ofe.s.r. signals does not mean that they are not present. It does indicate, however,that if present they are not of the radical type and may therefore be assumed to beless reactive.Moreover, since it should be an advantage if the 0 2 molecules rapidlyform 02--ions, the 02-structures connected with two Mo ions appear particularlyfavourable in this respect. Present evidence therefore goes into the direction toconsider Moo3 as a likely catalyst for reactions such as butene-tbutadiene.Dowden 19 has postulated that for surface oxidation to occur surface co-ordinationof 0 2 is essential. This apparently does not apply for the butene-tbutadiene re-action but this may be an exception. For other catalytic oxidations such as theconversion of naphthalene to phthalic anhydride, Dowden’s concept may be valid,and perhaps of interest in this connection is that Moo3 is inactive for this reaction.However, if the interaction of radicals, such as the allylic structure and peroxo-systems as described, is assumed to be inherently possible-and the idea appearsattractive-the multitude of 02-co-ordination complexes shown to be possible mightcause the analysis of the oxidation to be difficult. It will make it necessary, e.g.,to compare the relative activities and the various products of each type of 02-complex.One of us (P. F. C.) gratefully acknowledges financial assistance from HetEindhovens Hogeschoolfonds and the Nederlandse Organisatie voor Zuiver Weten-schappelijk Onderzoek Z.W.O., that enabled him to co-operate in this investigation.1 Adams and Jennings, J. Catalysis, 1963, 2, 63.2 Sachtler, Rec. trav. chim., 1963, 82, 243.3 Batist, Kapteyns, Lippens and Schuit, in preparation.4 Kasansky and Pariisky, Proc. 3rd Int. Congr. Catalysis (North Holland Publ. Co., Amsterdam,5 Kokes, Proc. 3rd Int. Congr. Catalysis (North Holland Publ. Co., Amsterdam, 1965), p. 484.6 Haber and Stone, Trans. Faraday Soc., 1963,59, 192.7 Jongepier and Schuit, J. Catalysis, 1964, 3, 464.8 Bleyenberg, Lippens and Schuit, J. Catalysis, 1965, 4, 581.9 Bleaney, Phil. Mag., 1951, 42, 441.10 Kneubiihl, J. Chem. Physics, 1960,33, 1074.11 van Reyen, Thesis (Eindhoven, 1964).12 Wells, Structural Inorganic Chemistry (Oxford, Clarendon Press), 3rd ed.13 Silltn and Lundberg, 2. anorg. Chem., 1943, 252, 2.14 Kittel, Introduction to Solid State Physics, 2nd ed. (John Wiley and Sons, New York, 1956).15 Ballhausen and Gray, Molecular Orbital Theory (W. A. Benjamin Inc., Amsterdam, 1964).16 Rapoport, Dokl. Akad. Nauk S.S.S.R., 1966, 153 (4), 871.17 Kwan, Proc. 3rd Int. Congr. Catalysis, p. 493.18 Batist, Lippens and Schuit, J. Catalysis, in press.19 Dowden, Col. Quimica Fisica Proc. Superficies Sdlidas, Cons. Sup. Inv. Sc. (Madrid, 1965),1965), p. 367.p. 177

ISSN:0366-9033    

DOI:10.1039/DF9664100290

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

GENERAL DISCUSSIONProf. Z. 6. Szabo (Budapest) said: Experiments carried out by us were to ascertainthe relationship between the quantity of the consumed (reacted) hydrogen and theso-called active oxygen content of the oxide. The question was whether the reactionof the excess oxygen or defect electron concentration was related to the sudden changein the conductivity of the sample or not (“end point of the titration of defectelectrons ”).Chromium(II1)-oxide ignited at 1 100°C and titanium dioxide (sintered at 950-105OOC) doped with small quantity of chromium(II1) oxide were used. Contrary toprevious practice addition of hydrogen was made in extremely small amounts(1 x 10-6 mole at about 1 x 10-2 torr pressure). After adding H2, time was allowedfor stabilization of the gas pressure and the conductivity of the sample.We havefound this method is more suitable to investigate surface processes that occur onaddition of hydrogen than measurements at high €32 pressure.The conductivity of pure chromium oxide markedly decreased at first, then itslightly increased, but it never regained its original value. On further additions ofhydrogen the increase that occurred after the decrease in conductivity becomessmaller, and finally it completely ceases (cf. fig. 1). The decrease in conductivity iscaused by the donating electrons during the chemisorption of hydrogen (1 /2H2 + 0 =H+) and reaction of the active oxygens of chromium(II1) oxide. The increase inconductivity is presumably related to the diffusion of electron holes from the bulk ofthe oxide and to the desorption of chemisorbed water H20+ = H20+ 0.If thechange of conductivity caused by the addition of hydrogen is plotted against thenumber of moles of reacted hydrogen, then the marked decrease of the conductivity ofchromium oxide ceases when the quantity of hydrogen is sufficient to react with theactive oxygen content that can be determined in another way.#FIG. 1.Detailed measurements were also carried out on titanium dioxide doped withchromium(II1)-oxide. The pure titanium dioxide contained no active oxygen ;active oxygen in great amounts was found, however, in titanium dioxide doped byCrzO3. Despite the fact that the conductivity of pure titanium dioxide by virtue ofthe n-conducting character increases on addition of hydrogen, the conductivity oftitanium dioxide doped by chromium dioxide decreases for small amounts of addedhydrogen.After addition of quantities of hydrogen, which in some cases approachedthe excess oxygen content of the oxide mixture, there occurred an increase in con-ductivity on further additions of hydrogen, as expected from the behaviour of pure30306 GENERAL DISCUSSIONtitanium dioxide. These measurements confirmed our previous assumption that thechromium ion after infiltrating to the surface layer of titanium dioxide is transformedinto a higher oxidation state, and as a consequence a defect-conducting surface layer isformed.On addition of hydrogen the defect electrons originating from chromium ions ofhigher valency are first destroyed thus causing a decrease of conductivity.The effectof hydrogen on pure titanium dioxide (increase of conductivity) is found only afterthis has taken place.Dr. F. S . Stone (University of Bristol) said : Boreskov has referred to the remarkablelow-temperature activity for homomolecular oxygen exchange which can be developedin zinc oxide and other oxides by suitable vacuum treatment. It has been suggestedthat 0 3 or 0 4 adsorbed complexes not involving lattice oxygen may be responsible.It would be useful to know if the analogue of this low-temperature activity exists withother reactions. It seems that the oxidation of carbon monoxide should be a favour-able case, since the adsorbed CO3 complex which has been proposed as an intermediatemight in some circumstances be formed analogously from 0 2 and CO.Dr.K. Klier and Dr. P. Jir& (Inst. Phys. Chem., CzechosIovak Acad. Sci., Prague)said : An important argument concerning the mechanism of homomolecular oxygenexchange in Boreskov’s paper is based on the coincidence of the rates of homolecularand heterogeneous exchange. Such coincidence cannot be expected on theoreticalgrounds except for one special type of mechanism which, however, does not operatealone with most of the oxides investigated in Boreskov’s paper. Namely, theoreticaltreatment of these reactions 1 , 2 adapted for the conditions of homomolecular exchangewhen atomic isotope concentrations in the gas and solid are equal, yields for the timechange, e.g.of 1 8 0 1 8 0 molecules, the following expression 1- d[’ 0‘ 80]/dt = (R + R‘ + R”)( l/a)( [ ’ OI8 01 - [’ 8018 01 to).whereas the time change of 1 8 0 atom concentration in the gaseous phase under theconditions of heterogeneous exchange is given by- d[’80]g/dt = (R‘ + 2Rr‘)[(2a + m)/2am]([’80]g- [’ 80]g,co),Here A, R’, and R” are the rates of processes I, 11, and I11 in Boreskov’s paper, a is thenumber of oxygen molecules in the gas phase and m the number of exchanging oxygenatoms in the solid. Under the condition nz 9 2a, fulfilled in many experimental investi-gations, the latter expression changes to- d[’ 13 O],/dt = (R72 + R”)(l/a)([’ 01, - [’ o],, to)-The constant R + R’ + R“ is the true rate of homomolecular exchange, irrespective ofisotopic labelling, and R’/2 + R” the rate of heterogeneous exchange giving the numberof oxygen molecules that exchange their oxygen with the solid in unit time byprocesses I1 and 111 (process I does not contribute to heterogeneous exchange).Thetwo rates can be equal only when X = R’ = 0, ix., when only process 111 (two atomsexchange with the solid in one step) alone operates. However, this situation occursonly with V205 catalysts 3 among those studied in Boreskov’s paper (fig. 8) whereasothers 3 (C0304, NiO, ZnO) exhibit prevailing rates R and/or R’. There is anotherdifference among the catalysts, viz., that some of them exchange all of their oxygen,others only surface oxygen in the heterogeneous exchange.1 Klier, NovAkovA and Jirh, J.Catalysis, 1963, 2,479.2 Muzykantov, Popovskii and Boreskov, Kinetiku i Katuliz, 1964, 5, 624.3 NovAkovB, Klier and J i d , Proc. 5th Int. Syrnp. Reactivity of Solids (Munich, 1964), p. 276GENERAL DISCUSSION 307In conclusion, the different oxides exhibit selectivity with respect to processes I, 11,and 111. Under such circumstances, question arises as to whether the homolecularexchange rate (R+ R’+R”) can be used as a measure of bonding energy of oxygen tooxides.Prof. G. K. Boreskov (fizst. Catalysis, Novosibirsk, U.S.S.R.) said: Dr. Jirii haddoubts as to the possibility of the equality of the rates of the homomolecular and hetero-geneous exchange of oxygen and based his opinion on the possibility that the homo-molecular exchange can proceed according to a different mechanism.In this connec-tions I would remark that equality of rates of exchange mentioned above can beobserved only when two conditions are fulfilled : (i) the oxygen content in the surfacelayer of oxide must be equilibrium, which can be achieved by preheating in oxygen.(ii) It is the initial rate of the heterogeneous exchange that is to be compared with thatof homomolecular exchange. This initial rate corresponds to the exchange of mostreactive oxygen of oxide surface. The experimental data obtained both in ourInstitute and by other investigators indicate that in observing the above conditions therates of the homomolecular and heterogeneous exchange are very close both forsimple oxides and binary systems of chromites and cobaltites and also for metals whichsorbed oxygen apart from y-Al2O3 on which homomolecular exchange proceedswith the rate greatly exceeding that for heterogeneous exchange.One can conclude from this experimental evidence that for all oxides withequilibrium content of oxygen in the surface layer except for y - 4 2 0 3 , the homo-molecular exchange proceeds with the participation of oxide oxygen.For certainoxides it has been established that two atonis of surface oxygen participate in theexchange (mechanism 111 in my paper). For most oxides the accuracy of the experi-mental data is so far insufficient for the choice between mechanisms I1 and HI.Irrespective of this f x t , the rate of the homomolecular exchange with the participationor surface oxygen atoms can be characteristic for the reactivity of this oxygen.Dr.P. Jirii and Dr. J. NovAkovh (Inst. Physic. Chem., Czechoslovak Acad. Sci.,Prague) said : In connection with Prof. Boreskov’s paper we would like to make someremarks and present preliminary results about the selectivity of Fe-Mo-oxide cata-lyst. Many of the oxide catalysts discussed in Boreskov’s paper (Fe203, V205, ZnO,Fe283-Mo03) contain a certain mount of bonded hydrogen which is hardlyremoveable as water by means of pretreatment in vacuum or in oxygen atmosphere atTABLE 1 .-SELECTIVITY OF DEHYDRATED Fe-MO-CATALYSTStemp. “C of treatment amount adsorbed of composition of desorbedMo-Fe-cat a1 ys t CH3OH at 20°C products atin vacuum, 2 h (ml/g, s.t.p.1 86°C20200350350in 0 2 atmosphere, prevailing part H2Oprevailing CH30H+little CH20little COprevailing CO1 -0.7 prevailing CH20+higher temperatures. This is also true for the active and selective Fe-Mo-oxidecatalysts for methyl alcohol oxidation.In this catalyst, hydrogen is very stronglybonded and it does not undergo exchange with molecular deuterium. We preparedby vacuum and temperature treatment a series of Fe-Mo catalysts with decreasingcontent of bonded hydrogen. The highest amount of water was removed from th308 GENERAL DISCUSSIONsample by combined treatment in an oxygen atmosphere at 350°C. During the pre-treatment there was no determinable change of surface or pore structure of thecatalyst. On this series of samples we sorbed a constant amount of methyl alcohol atroom temperature and very low equilibrium pressure. By means of mass spectro-metric analysis we then followed the composition of desorbed products at 86°C.Fromthe results given in table 1 it follows that the concentration of bonded hydrogen in thecatalyst plays an important role, probably in connection with its activity and selectivity.The same type of catalyst with a low concentration of bonded hydrogen has lost itsselectivity and leads to a marked oxidation of methyl alcohol. We believe that thisillustrates another complication in using the suggested correlations for complexcatalytic systems of practical importance.Dr. L. L. van Reijen (Amsterdam) said : Prof. Boreskov's review about the activityof oxides for homomolecular oxygen exchange reactions clearly demonstrates aremarkable phenomenon. After high-temperature evacuation most oxides releaseoxygen very reluctantly, showing the great stability of the surface anions in thesecompounds.Yet readmission of oxygen does not lead to a rapid restoration of theoriginal surface composition. One of the important consequences is that the steadystate of the oxide surface during oxygen isotope exchange reactions is ill-defined. Nowwe have found a striking contrast in the reactivity of chromia on silica and chromia onaluminia surfaces towards oxygen and towards water vapour in the sense that thereactivity towards water vapour is much greater. Hence my question is whether thereis any information on the influence of either water vapour in the gas phase or hydroxylions in the surface on the activity of oxides for oxygen exchange reactions.Prof.R. A. W. Haul (Technisclze Hochschule, Hannover) said : The catalytic activityof oxides for oxygen isotope exchange has been studied by Prof. Boreskov for a varietyof high surface-area materials. After heating in oxygen at elevated temperature, e.g.600"C, the oxides are said to possess a stable and reproducible activity. Nevertheless,the temperatures are relatively low with respect to lattice rearrangements. Thequestion, therefore, arises as to how far this factor influences the results, i.e., activationenergies and pre-exponentials of the rate constants for homomolecular (e.g., fig. 2)and heterogeneous (e.g., line 11, fig.3) exchange.We have been studying oxygen isotope exchange between gaseous oxygen andoxides at higher temperatures and with larger crystals or single crystal plates annealedin oxygen at high temperatures. Although our main objective so far has been thestudy of oxygen diffusion in the bulk oxide, surface exchange has to be taken intoaccount in order to properly evaluate diffusion coefficients. The mathematicaltreatment given earlier 1 enables one to calculate both reaction constants for the hetero-geneous surface exchange k = ko exp (Ek/RT) and diffusion coefficients. For CdOfrom exchange experiments in the temperature range 630-855°C the following resultswere obtained :2 ko = 1 -3 x 102 cm/sec ; EX: = 49 kcal/mole ; k ~p -&13 and increase ofk when CdO is doped with LizO.For Ti02 we found 3 at temperatures 730-1300°C :(I) crystals annealed at 1050"C, 72 h, ko = 8-4 x 102 cm/sec, & = 59.6 kcal/mole;(11) crystals annealed at 13OO0C, 72 h, ko = 0.8 x 102, Ek = 57.9 kcal/mole ; increaseof k with decreasing 0 2 pressure.Although experiments with high surface-area material at lower reaction tempera-tures are of particular interest in connection with catalytic problems it would beinteresting to link up these results with those obtained with better defined crystalsurfaces.1 Haul, Diimbgen and Just, 2, physik. Chem. 1962,31, 309.2 Haul and Just, J. Appl. Physics, 1962, 33, 487.3 Haul and Diimbgen, J. Physics. Chem. Solids, 1965, 26, 1GENERAL DISCUSSION 309Prof. G. K.Boreskov ( h t . Catalysis, Novosibirsk, U.S.S.R.) said: In reply to Haul,the investigations of the homo-molecular and hetero-exchange of oxygen, cited in mypaper, were carried out under conditions when the rate of the isotopic exchange insidethe oxides was negligible. The only exception is vanadium pentoxide for which therate of oxygen diffusion into the body of crystals is very great and in the process ofexchange the isotopic composition near the surface is very close to the compositionin the middle of crystals. The investigations of Haul have been carried out at highertemperatures, and are not in contradiction with our results.In reply to van Reijen, the catalytic activity of ?-A1203 for the homomolecularexchange of oxygen arises only after the removal of more than 70 % of hydroxyl ionsfrom the surface.The adsorption by the activated sample at 300°C of 5 x 1013molecules of water per cm2 leads to the total deactivation.In reply to Stone, we have for the present no success in determination of thecatalytic activity of oxides after vacuum treatment at elevated temperatures for theother reactions, but I agree with Dr. Stone that it would be useful.Prof. R. L. Burwell (Northwestern University, Evanston) said: As a by-product ofwork aimed at characterization of the chemistry of chromium on silica gel, Dr. DanielCornet in our laboratories has observed the formation of Cr5+ on silica by differentpreparative methods. The samples described below gave an e.p.r. signal of the typeof fig. la.Our measurements were at room temperature and under vacuum.Cr(NH3):f was put onto wide-pore silica gel (Davison grade 70) by ion exchange 1to about 0.14mmole/ggel. After heating in oxygen at 250"C, 10-13 % of thechromium was present as Cr5+. Extraction with water removed 90 % of the chromiumas CrQi -. If reduction with carbon monoxide at 450°C was substituted for the oxygentreatment, a blue gel resulted which we believe to contain Crz+. Treatment of theblue gel at -78°C gave a brown gel which was evacuated at -78°C. As judged bythe intensity of its e.p.r. signal, the Cr5+ content was 1 %. No Cr3+ could be detectedbut the wide line-width of Cr3+ limits the sensitivity for Cr3+ to about 10 %. Furthertreatment with oxygen at room temperature increases the Cr5+ content to about 4 %without causing a detectable signal from Cr3+.If the CrZ+ gel is initially treated with oxygen at room temperature, the samplecontains about 0.4 % Cr5+ and 70 % Cr3+ as judged by e.p.r.and 32 % CrOi- asjudged by extraction. Similar results were obtained using a gel prepared by ionexchange of silica gel with chromous acetate followed by drying at room temperature,all in the absence of oxygen. Absorption of oxygen by the Cr2+ gel is very rapid evenat - 78". The gel might well be useful in removing traces of oxygen from nitrogen orother unreactive gases.Prof. J. Turkevich (Princeton University) said: The paper by Ayscough, Eden andSteiner 2 points out the poor correlation between catalytic activity and the intensity ofthe e x .signal at g = 2 observed at room temperature and attributed to Cr(V).However, after poisoning with water vapour the signal intensity correlated well withthe activity before poisoning. V. B. Kazanski working last year at Princeton Univer-sity found good correlation between catalytic activity and the g = 2 signal before thereaction. The signal changed in form and dropped in intensity during the course ofpolymerization reaction, particularly at 50"C, while the reactivity was sustained.Furthermore, treatment of the catalyst with 30 mm of C2H4 at 150°C increased thecatalytic activity at 0°C by a factor of 30. However, the signal at g = 2 becamemarkedly smaller. There was no correlation between the catalytic activity and thesignal at g = 2 observed at room temperature in an active catalyst.1 Burwell, Pearson, Haller, Tjok and Chock, Inorg.Chern., 1965, 4, 1123.2 Ayscough, Eden and Steiner, J. Catalysis, 1965, 4, 2783 10 GENERAL DISCUSSIONTwo explanations may be offered for the disappearance of the signal at g = 2attributed to Cr(V) in the active state of the catalyst. One point of view assumes thatthe crystal field of ligands surrounding Cr(V) is such that the signal is not observed.Theory indicates that if the ligands have the symmetry of an elongated octahedron oran undistorted tetrahedron or octahedron no signal should be observed. The secondpoint of view is that the active catalyst is a Cr valence state different from 5.The first theory has been formulated in a paper submitted to the J.Catalysis byKazanski and Turkevich. The precursor of the catalyst site is a distorted tetrahedronCr(V> which gives a signal at g = 2 and which has been observed by a number ofworkers. On treatment with ethylene it goes over into a square pyrzmid with theethylene forming a vinyl group and the hydrogen going over to form a hydroxyl group.(D,Cr(V); 0 , S i ; 0, 0The con&uration has the symmetry of an elongated octahedron which gives no signalat room temperature. The proportionality between catalytic activity and the signalat g = 2 before the reaction may be due to the fact that a certain fixed fraction of thedistorted tetrahedron is transformed into the active form. In the polymerizationprocess the reacting ethylene approaches the Cr atom in the frze " six " position of theoctahedron, this changes the octahedron to cornpressed octahedron by tilting of thesquare toward the surface. This process is favoured by the Jahn-Teller effect whichin this case would indicate that an elongated octahedron with a degenerate groundstate would tend to reariange its atomic arrangement into a compressed octahedronand thereby remove the degeneracy.In the compressed octahedron the reactingethylene molecule would leave position " six " and go to position " four " thus beingincorporated into the growing polymer chain. The configuration around the chro-mium V again becomes an elongated octahedron, which does not give an e.s.r. signalat room temperature. The occurrence of an undistorted tetrahedral or octahedralconfiguration on the surface is highly unlikely and is not considered.Dr. J.Karra and Prof. J. Turkevich (Princeton Uniuersity) said: Recent experi-ments carried out this system by us indicate that the situation is more complex.Two specific samples studied by Kazanski were examined using resonance techniques.A sample of chromia on silica (A) evacuated at 500°C which had good polymerizingactivity, and another sample B which was sample A treated with 30 mni C2H4 at150°C and had 30 times the catalytic activity of sample A. Sample E had shown amuch smaller signal at g = 2 zt room temperature than sample A. We find at liquid-air temperature (fig. 1) two signals-one at g = 2 and another at g = 4. While thereis an inverse relationship between catalytic activity and intensity of signal at g = 2,there is a direct relationship between catalytic activity and infensity of signal ot g = 4.Cr(I1) gives a signal at g = 4 at liquid-nitrogen temperature and this signal is weak atroom temperature.We have also examined the signal at 4-2"K (fig. 24. FoGENERAL DISCUSSION 31 1sample A there is a sharp line of 30 gauss width at gm2-0 with a background of abroad line with a centre at g = 2.09. The signal at g = 4 disappeared. For thesample treated with ethylene (B), the sharp line at g252 sharpened to 20 gauss and thebroad line retained its form (fig. 2b). The fact that the sharp line at g = 2 wasobserved with practically constant width at room temperature, liquid-nitrogen andliquid-helium temperatures shows that the Cr(V) is in a square pyramidal (compressed9 = 2 q = 4Botorc Ethylano Truatrneritii;i, /---A t t u r Eriiyicqe TreatmentFIG.1.-E.s.r. signals (X-band) of chromium on silica at liquid nitrogen temperature before and afterethylene treatment.FIG. 2 . 4 ~ ) E.s.r. signal (X-band) of chromium (V) on silica before ethylene treatment (sample A) at4.2"K. 1 in. on horizontal axis = 200gauss. (b) Same as above but after ethylene treatment(sample B) at 4-2°K.octahedron) form and this may be the precursor of the active state. In the previousdiscussion we identified this as tetrahedral. The broad signal in the background is ofunknown origin. It may be a silica defect.At 1.6"K the appearance of the spectra was not appreciably changed.Protonresonance was observed at liquid helium temperature on sample B.A double resonance experiment was performed on this sample by saturatingsuccessively each portion of the e.s.r. complex with the microwave radiation andobserving the enhancement of the proton resonance. This enhancement would beinterpreted as indicating a '' cross talk " between the species responsible for the e.s.r.absorption and a proton species in its neighbourhood. A positive enhancement o312 GENERAL DISCUSSIONonly 50 % was found having the characteristics of a solid state (fig. 3) effect.1 Calibra-tion of the sensitivity of the equipment gave a 2000 % enhancement for 1 % DPPH inpolystyrene. The source of the slight enhancement in the chromium silica catalystwas not a paramagnetic species at 9x2-O but one corresponding to the broad line atgz2.09 that was observed in the e.s.r.spectrum. We were unsuccessful in carryingout double resonance at g = 4 because of electronic and circuit difficulties. Thesmall double resonance solid-state effect observed at g = 2-09 may be due to hydro-carbons adsorbed on defects in the silica.5 0 5 .FIG. 3.-The enhancement of proton n.m.r. signal against the magnetic field with e.s.r. saturation at250 mw microwave power. The enhancement E = (A-Ao)/Ao, where Ao is the unenhanced n.m.r.signal intensity of proton without e.s.r. saturation and A is the enhanced proton signal with e.s.r.saturation.In conclusion the evidence in the Princeton University laboratory seems to indicatethat the Cr(V) signals observed either at room temperature, liquid-nitrogen tempera-ture, liquid-helium temperature and even at 1.6"K are not directly associated with thespecies responsible for catalytic activity.The only signal that seems to correlate withthe activity on the basis of two samples is the signal at g = 4 observed at liquid-nitrogen temperature and presumably due to Cr(I1).Dr. €3. Steiner (Manchester CoZZ. of Sci. and Techn.) said: One problem in thestudy of the polymerization of ethylene over chromium oxide catalysts is to find a goodcorrelation between the chromium oxide species responsible for the catalytic action andthe rate of polymerization of ethylene. Two such correlations have been established.One by Kazanskii and co-workers 2 using a catalyst activated at a rather low tempera-ture and which therefore cannot have been extremely active.In this case the intensityof the e.s.r. signal of Cr5+, as obtained at room temperature and before polymerization,correlated well with the yield of polyethylene, this being a measure of the polymeriza-tion rate.The second and different correlation was obtained by Dr. Ayscough, Dr Eden andmyself.3 We did not succeed in establishing more than a qualitative relation between1 Jean-Loup, Erb and Uebersfeld, Compt. rend., 1958, 246, 2121.2 Bukanaeva, Pecherskaya, Kazanskii and Dzisko, Kinetika i Kntaliz. 1962. 3. 358.3 Ayscough, Eden and Steiner, J. Catalysis, 1965, 4,278GENERAL DISCUSSION 313the intensity of the Cr5+ room temperature signal, when measured before polymeriza-tion, and the polymerization rate.The latter was measured directly using a catalystactivated at 550"C, which was reactive enough to promote the polymerization at tem-peratures as low as - 30 to -40°C. We found, howevei, a good correlation betweensignal strength and rate in the following way. If after reaction catalysts of variousactivities were treated with water the Cr5+ signal increased. The number of unpairedspins equivalent to the increase in the peak area of the signal, as obtained by doubleintegration, correlated well with the polymerization rate. Before contact withethylene, water does not alter the strength of the Cr5+ signal although it does alter itsshape.1 We interpreted our results as due to the breaking by water of the chemisorp-tive bond of the ethylene or ethylene polymer to the catalyst.As a result the Cr5+room temperature signal, which is partially suppressed by the ethylene chemisorption,is restored.These results seem to agree with the mechanism of the polymerization now putforward by Cossee and van Reijen. According to this mechanism the first step is achemisorptive attachment of ethylene to tetrahedrally co-ordinated Cr5+. Thistetrahedrally co-ordinated Cr5+ complex gives a signal at 20°K only and thereforecould not be observed in the older work when the Cr5+ signal was measured at roomtemperature only. There should therefore be no correlation between the roomtemperature signal and the polymerization rate as we have found.On contact withethylene and on polymerization, the tetrahedral Cr5+ complex is transformed into asquare pyramidal one as follows :-Si-0 OH -Si-0 CH2-CH2-(CH2-CW2)%-2-CH=CH2\ /Cr i- x C ~ H ~ -+\ /Cr-OH - .OH/ \0 -Si-0/ \-Si-0On treatment with water, after polymerization, the following reaction occurs :-Si-0 OH\ /Cr-OH + CH3-CH2-(CH2-CH2)s-2--CH=CH2OHThere should result therefore a square pyramidal complex of considerable symmetrywhich should give a signal observable at room temperature. It is suggested that theformation of this complex is responsible for the increase in the room temperaturee.s.r. signal which we have observed. This increase should be proportional to thenumber of Cr5+ polymerization centres and hence to the polymerization rate.This result is further supported by independent evidence obtained from thekinetics of the polymerization reaction at temperatures between -30 and 0°C.Atthese temperatures the rate follows strictly zero-order kinetics. From the measuredactivation energy and using the transition state formulation of heterogenous reactionrates, an estimate of the number of active polymerization centres was obtained. Thiswas within a factor seven (smaller) of the number of unpaired spins calculated fromthe increase in the Cr5+ signal after polymerization and treatment with water asdiscussed above.Dr. J. Haber (School of Mining and Metallurgy, Krakow) said: van Reijen andCossee express the opinion that Cr+5 ions are not intermediate species in the reduction1 Pecherskaya and Kazanskii, Kinetiku i Kaializ, 1963, 4, 244./ \-Si-314 GENERAL DISCUSSIONof Cr+6 to Cr+3, but are formed as a stable product.We have investigated the beha-viour of chromium ions supported on various aluminas as a function of the temperatureof annealing and coricentration of chromium. The results seem to indicate that,depending on the conditions of experiment, the reaction series Crf6, Cr+5, Crf4, Cr+3proceeds to a smaller or higher extent. The comparison between chromium oxidessupported on various aluminas (a, 0, y ) is difficult because of large differences insurface areas af these carriers. If, however, the amount of chromium is calculatedper unit surface area and samples of similar surface concentration of chromiumcompared, they show similar behaviour.This seems to indicate that the propertiesof chromium ions are practically independent of the modification of alumina used as acarrier and are related rathei to the state of dispersion of these ions at the surface.Dr. P. Cossee (Amsterdam) said : Turkevich mentioned that in certain experimentsduring polymerization the Crs+--signal disappears, whereas activity is maintained.Which Crsf-signal is meant : the one visible at room temperature or the one whichcan only be observed at very low (Hz- or Me-) temperatures?In view of Hill's fear that a vinyl end group on the growing polymer could reactwith another active site, giving rise to cross-linking, I would like to ask him, whetherpropene can be polymerized with a Phillips catalyst?Dr.T. Hill (Brit. Hydrocarbon Cheni. Ltd., Grangemouth) said: I would like tomake a point in connection with the mechanism of polymer chain initiation suggestedby Voevodsky 1 and extended by Dr. van Reijen and Dr. Cossee at this Discussion.They envisage polymer growth from a metal alkenyl species rather than from a metalalkyl species such as is common for Ziegler-type catalysts. Despite the speculativenature of such proposals as applied to the chromium oxide-silica system, the distinctionbetween different proposed polymer chain initiating complexes is important. Forexample, a metal alkenyl complex must give rise to a growing chain with a terminaldouble bond which may then be able to interact with and displace another growingpolymer molecule from a nearby site by disproportionation, as in " monomer-grow-ing polymer " disproportionation :M,-CH2-(CH,),,-CH =CH, + M,-CH,---(CH,),--CH CH2M,-CH2-(CH2),-CH =CH-M2 + CH,-(CH,),-CH == CH2This kind of process would affect the kinetics of 2 polymerization and give rise to al/M, = A+BN,relation between active site concentration and molecular weight of the typewhere Mn = number average molecular weight, N = active site concentration,A , B = functions of rate constants for propagation and termination. Consequentlykinetic studies would be useful, and a closer coupling of e.s.r.studies with the studyof the reaction itself would be fruitful. Information about the number, nature andvariable properties of active chromium, obtained from e.s.r.might help in the inter-pretation of polymerization kinetics.In connection with the conclusion of Prof. Turkevich based on his e.s.r. work, 1mention a simple experiment carried out by us. Samples of a catalyst (2 % chromiumon silica, activated in dry air at 550°C) were exposed to carbon monoxide at 400°Cand 35 torr pressure, for variable periods. Each sample was outgassed to constantweight at 400°C after reduction, cooled to lOO"C, and exposed to high-purity ethyleneat a pressure of 40 torr without removal from the apparatus. Results are shown inthe table 1 .1 Voevodsky, Proc. 3rd Int. Congr. CatalyJis, (Amsterdam), 1964, 1,104.2 Gaylord and Mark, Polymer Reu. (Interscience. New York.1959), vol. 2, p. 204GENERAL DISCUSSION 315In line with other work, 1 this showed that a certain degree of reduction may bedesirable to develop activity at low ethylene pressure, but equally that excessive reduc-tion is undesirable. This result is difficult to reconcile with the need for reduction tothe 2+ state proposed by Turkevich. There may be something specific aboutreduction with ethylene, e.g., a concurrent complex formation, but in this case poly-merization occurred rapidly with no induction or acceleration in rate, suggesting thata preliminary slow reaction of ethylene with catalyst is not important.TAEL,E 1 .-EFFECT OF co REDUCTION ON SUBSEQUENT POLYMERIZATIONtime of reduction rate of ethylene polymerizationmin rng polymerlg catalyst min10304145506010515016.023.542.047.54.1 -721.611-53 5Dr.van Reijen has rightly stressed the circumstantial nature of the e.s.r. evidencein relation to catalytic activity. Much of the present confusion regarding relevantoxidztion states may arise from the fact that after catalyst activation and contactingwith ethylene all oxidation states from 6+ to 2+ may exist in somecases. Our published results on ethylene chernisorption 3 indicated the need forcareful distinction between ethylene strongly chernisosbed to form stable oxidizedcomplexes which were unconnected with polymerization and the minute and moresubtle sorption processes underlying polymerization. It seems likely that lessambiguity might result from combining a study of the polymerization with e.s.r.examination using catalysts containing much lower chromium contents than 2 %.Inreply to Dr. Cossee, propylene is polymerized but at a considerably reduced rate.Mr. A. EIlisom and Prof. K. S. W. Sing (Brunel Uniueusity, London) said: Supportedchromium oxide catalysts on oxidation give a strong e.s.r. signal, which is characterizedby a g-factor of 1.97 and a line-width of about 50 gauss ; y-phase resonance 2 has beenascribed by van Reijen and Cossee to Cr5+ ions stabilized in various coordinationswithin the surface layer of the host lattice. Although this theory is widely accepted,it is not able to account for certain properties exhibited by the chromium oxidesystem. Our view is that the 7-phase signal must be associated with the surface of athin layer, or clump, of chromium oxide rather than with Cr5+ dispersed in the oxidematrix.In the work which we shall report elsewhere,3 y-phase resonance was obtained onvarious samples of chromia-alumina, some of these samples having been obtained byimpregnation of alumina (with Cr3f or Cr6+) and others by co-precipitation of thehydrous oxides of chromium and aluminium.The e.s.r. derivative spectra shown infig. 1 were determined on a sample containing 1-8 wt. % Cr, which was prepared byimpregnation with CrOJ followed by calcination at 400°C. The peak-to-peak widthof the y-phase signal is not changed with temperature over the range studied, - 196 to300°C. In our view, this constancy of the e.s.r. line-width may be due to spin-spinrelaxation rather than spin-lattice relaxation. Similar e.s.r.signals were given by1 Habeshaw and Hill, Proc. 3rd Inf. Congr. Catalysis, (Amsterdam), 1963, 2, 975.2 O’Reilly and MacIver, J. Physic. Chem., 1962, 66, 276.3 Ellison, Oubridge and Sing, unpublished work316 GENERAL DISCUSSIONthis same sample after heat treatment at 200°, but as might be expected, no resonancecould be detected on drying the sample a t 50". On the other hand, another sample,also prepared by impregnation with Cr03 but having a higher loading of chromium(1 1-3 % Cr), became paramagnetic on drying at 50°, giving the e.s.r. spectrum shownFIG. 1.-Electron spin resonance of impregnated CrO3/A1203 (1-8 wt. % Cr), calcined in air at 400°C.Spectra obtained at : (a) 77"K, signal gain 63 ; (b) 153"K, signal gain 80 ; (c) 233"K, signal gain 80 ;(d) 348"K, signal gain 80 ; (e) 423"K, signal gain 80 ; (f) 573"K, signal gain 250.I4000 3500 ,3000 2500H, gaussFIG.2.-Electron spin resonance of impregnated CrO3lA1203 (1 1.3 wt. % Cr), dried in air at 50°C.Spectrum obtained at 77"K, signal gain 125.in fig. 2. In this case, the narrow y-phase signal is superimposed on a broader signalsimilar to that reported by Deren, Haber and Kosek 1 ; these workers also noted theunexpected appearance of a paramagnetic species in uncalcined samples containingcrystallites of Cr03, but concluded that the broad band takes the place of the y-phase1 Deren. Haber and Kosek, Bull. Acad. Polon. Sci., Skr. sci.chim., 1965, 13, 21GENERAL DISCUSSION 3176 0 0 .signal. It would seem that there was a critical loading of chromic oxide (ca. 4 % Cr),above which the system became paramagnetic even on drying at low temperature;solution adsorption measurements on the same alumina indicated 1 that this loadingcorresponded to complete coverage of the alumina surface with a single layer ofchromic oxide. This evidence suggested that the development of the y-phase signalwas associated with the presence of chromic oxide clumps.Other workers2,3 have demonstrated that the y-phase signal may be eitherremoved or considerably reduced in intensity by leaching the sample in water, leavingbehind Cr3+ in the dispersed form (the &phase). It is difficult to understand howCrs+ ions could enter the lattice sites at the low temperature of 50” or again leave thesesites on leaching with water.It seems to us significant that van Reijen and Cosseemention that the most clear-cut results were obtained for Cr/SiOa and that “theCr5+ species in the silica surface display well-resolved e.s.r. spectra with narrowlines ”, yet all the evidence 3 suggests that the degree of dispersion of Cr (certainly asCr3+) is greater in alumina than in silica.M400-rn ii7 3 0 0 - 2c----. A0 100 2 0 0 3 0 0 Ltemperature, O KFIG. 3 . - 1 / ~ ~ plotted against temperature for a coprecipitated sample of Cr203/&03 (1.3 wt. % Cr),previously calcined in air at various temperatures. Temperature of calcination : X , 200°C ; 0,400°C ; A, 600°C.(XA is atomic susceptibility of Cr.)In seeking an explanation for these results, it is important to note the non-linearityof the molecular field plots (fig. 3) with all our samples found to exhibit y-phaseresonance; in no case does the magnetic susceptibility follow the Curie-Weiss law,large positive intercepts being given on the temperature axis. To shcw that thisbehaviour is a function of the y-phase chromium, the reciprocal of the e.s.r. intensityis also plotted against temperature (fig. 4) to provide to a first approximation a mole-cular field plct for this phase. The deviations from the Curie-Weiss law, evident infig. 3 and 4, seem to rule out the possibility of magnetically isolated Cr5+ ions in1 Ellison, Oubridge and Sing, unpublished work.2 Gremillion and Knox, J.Catalysis, 1962, 1, 216.3 Matsumoto, Tanaka and Goto, Bull. Chem. SOC. Japan, 1965, 38,45318 GENERAL DISCUSSIONoctahedral or tetrahedral lattice sites as the paramagnetic species responsible for 1'-phase resonance. Although the systems have some ferromagnetic character, fig. 3 and4 indicate that they are not true ferromagnetics. Curved molecular field plots of thistype have been explained 1 9 2 in terms of direct exchange interaction between ions ofdifferent valency, i.e., Zener " double exchange " 3 ; this mechanism demands aFIG. 4.-1 / I plottedpreviously calcined/ P 5 t IJ d0 100 2 0 0 3 0 0temperature, O Kagainst temperature for a coprecipitated sample of Cr203/A1203 (1-3 wt. % Cr),in air at various temperatures.Temperature of calcination : X , 200°C ; 0,400°C ; A, 600°C. (I is e.s.r. signal intensity.)system in which electrons are able to move freely between ions of the same atom, butwith different valency, on equivalent lattice sites. It would seem possible, therefore,that the exchange coupled system at the surface of the oxidized chromium oxideclumps may be represented as-cr6+-02--cr3 +-02--~r6 +-Further confirmation of the Cr3+-Cr6+ double exchange system is provided bythe nature of the optical spectra of the oxidized, supported chromium oxides.4 Theabsorption bands associated with the crystal field transitions in the Cr5+ ion appear tobe absent, whereas the spectra of Cr3+ and Cr6+ ions are superimposed with intenseand continuous absorption over a range of high wavelengths.The appearance of the7-phase optical spectrum suggests collective-electron behaviour, consistent with theZener description.Dr. L. L. van Reijen (Amsterdam) said : Essentially two species of Crs+ are observedin supported chromia catalysts, the first observable by e.s.r. at liquid-hydrogen1 Anderson and Hasegawa, Physic. Rev., 1955, 100,675.2 Gennes, Physic. Rev., 1960,118, 141.4 Ellison, Thesis (Royal Institute of Chemistry, 1965).3 Zener, Physic. Rev., 1951, 82,403GENERAL DISCUSS I ON 319temperature and not at room temperature, the second observable at liquid-hydrogenand at room temperature with a normal ratio of intensities. There is no correlationbetween the quantities of the two species after the various treatments of the catalyst.Hence, when considering a correlation between polymerization activity and amountof Cr5+, one has to specify which of the two species is meant.As yet, no direct e.s.r.evidence has been found for the occurrence of complexes ofCr5+ with either ethylene or growing polymers. We did try to detect such complexesby e.s.r. at liquid-helium temperatures, but did not succeed. Hence the only wayavailable for investigating the role of Cr5-t- in the polymerization activity is to look fora correlation between the amount of Cr5+ and the activity. The evidence availableis fzr from conclusive and even conflicting. In order to make an active catalyst,calcination at 500°C seems to be an essential prerequisite.As there is sufficientevidence that at least Cr6f is not responsible for the polymerization activity, this isone of the reasons to propose Crs+ as the active species. According to variousauthors, however, high activities for the polymerization of ethylene from the gasphase can only be acquired by a pretreatment of the catalyst with dry CO or dryC2H4 at relatively high temperatures. This suggests that lower valeiices may beinvolved.We think the efiects described by Seiner fit in well with our own experience, see,e.g., table 3 : the square pyramidal Cr5+ is present after calcination, disappears aftercontact with CzH4 and reappears after contact with water vapour. It is also seenfrom table 3, however, that simultaneous changes take place in the amounts of tetra-hedral Cr5+.It is still a moot point to what extent there is a correlation between theamount of tetrahedral Cr5+ after contact with C~MJ and the amount of square pyra-midal Cr5+ formed after a subsequent tieatment with water vapour. If there is sucha correlation, Steiner’s data may provide an argument in favour of tetrahedral Cr5+as the active species.In answer to Haber, it is indeed difficult to assess whether Cr5+ is an intermediatein the route from Cr6f to Cr3+ or whether it is an independently stabilized species.Our reasons for preferring the latter alternative are the following : according to theexisting chemical evidence Cr5+ is unstable with respect to Cr6f and Cr3+. Cr5+complexes, prepared in homogeneous solutions generally disproportionate easily intoCr6+ and Cr3+.For supported chromias on the contrary, it appears that high-temperature treatment in oxygen never leads to a complete oxidation to Cr6+, evenwhere it is proved by susceptibility measurements that no lower valences can bepresent. It is also difficult to destroy the last traces of Cr5+ by reduction treatments,where again by susceptibility measurements it is known that nearly all of the Cr6fhas been transferred into Cr3+.Turkevich has observed e.s.r. spectra of well-defined compounds containing Cr2+showing resonances at g = 2.0 and between g == 4.0 and g = 6.0. He uses thisinformation to conclude that similar resonances observed for C2H4-treated silica-supported chromias must also be ascribed to Cr2+. We think this conclusion is notyet warranted.First, we have observed e.s.r. spectra with resonances at g = 2.0 andbetween g = 4-0 and g = 6.0 for various well-defined compounds containing eitherCr3+ in octahedral coordination or Crs+ in tetrahedral coordination. Further we haveobserved such resonances in many different Cr/SiOZ samples where the conditions ofthe pretreatment had been such that the presence of Cr2+ could be virtually excluded.Prof. S. 2. Roginskii (Moscow) said: In connection with the part played inchemisorption and catalysis by unpaired electrons of the catalyst surface,l I mention1 Roginskii, in Jubilee Vol. to Sernenof, Khimicheskaya Kiizetikn i Tsepnge Reaktsii (ChemicalKinetics and Chain Reactions), ed. ‘‘ Nauka ”, (Moscow, 1966)320 GENERAL DISCUSSIONthe results obtained in our laboratory for a system that was not reported in any of thepapers.Chemisorption and catalytic oxidation on active charcoals exhibit character-istic e.s.r. signals with a g factor about 2-003 and a 3-25 gauss width. The signalsseem to be due to carbon, rather than to impurities, as those signals are observed forvery pure samples, and after prolonged high-vacuum treatment at T> 1000°C.Reversible oxygen adsorption suppresses the signals completely (fig. 2)., :!Active charcoals containing mineral impurities are good catalysts for oxidation ofmany hydrocarbons to hydroperoxides (fig. 3). It may be seen from comparison ofcurves 2 and 1 that the catalyst eliminates the induction period required for auto-oxidation.With charcoal samples of constant weight the rate of oxidation increaseswith increasing volume of the liquid (curves 2 and 3), and is hindered by inhibitors ofthe liquid-phase oxidation (curve 4). Various results give indications of chain propa-gation of the heterogeneous reaction in the volume of the liquid. E.s.r. spectraobtained in the course of the reaction (fig. 2) show the amount of oxygen chemisorbedreversibly on radical centres, and suggest the following scheme for the main reactionstepsHeterog., [C,]* +02-+[Cn]*. . . 02+[R62] dissolv. (1)Homogen., R& dissolv. + HR+R02H+ (2)i + O ~ - + R 0 2 . . . GENERAL DISCUSSION 32 1[Cn]* denotes a characteristic feature of active centres on the catalyst surface,namely, the superstoichiometric (n>1) decrease with the number of unpaired Celectrons in chemisorption. This effect was first observed by Voevodskii et al.forsugar coal,l with n- 70. In our experiments n was considerably less and changed withthe coverage.6 I2 18FIG. 3.The nature of the superstoichiometric attenuation of the e.s.r. signal is not clear.It might be due to changes included by 0 2 in the electronic structure of whole‘‘ domains ” involving 12 and more C atoms. The part Played by mineral elementsseems to be that of favouring the formation of dissolved RO2 radicals in the reactionof RH with chemisorbed 0 2 .Dr. J. Karra and Prof. J. Turkevich (Princeton University) said: To avoid mis-understanding the arguments presented for the possible identification of the tripletsno.2 and no. 3 with 0+, we emphasize that the Of would not be the normal 0’found in the gas phase. For this species by Hund’s rule would be in an S state andtherefore shows one g value equal to that of a free electron. The Of that we arepostulating must be in an excited state (fig. 3a) to give the e.s.r. characteristics that weobserve.Dr. F. S. Stone (University of Bristol) said: With regard to the modes of oxygenchemisorption on Ti02 proposed in the paper by Schuit and his colleagues, a reactionbetween adsorbed hydroxyl and 0; giving 0- and H02 has also been postulated toaccount for the irreversible photo-adsorption of oxygen on Ti02.2 This suggeststhat prior of in situ illumination of a Ti02 +oxygen mixture with ultra-violet light maybe a way to develop selectively, and thus to identify, the e.s.r.signal (or signals) whicharise through interaction of oxygen with adsorbed hydroxyl.1 Blumenfeld, Voevodskii and Semenov, Primenmiye Elektrunnogu Paramagnitnugo Resunansa uKhimii(Use of the Electron Spin Resonance in Chemistry), Ac. Sci U.S.S.R.., Novosibirsk, 1962.Stone, Anales real SOC. espan. Fis. Quim. B, 1965, 61, 109 (Coloquio sobre Quimica Fisica deProcessos en Superficies Solidas, Madrid, 1965)322 GENERAL DISCUSSIONDr. R. L. Nelson (A.E.R.E., Harwelf) said: Schuit and co-workers, and others,including Kasansky and Pariisky, and Kokes, have observed paramagnetic oxygenspecies adsorbed on oxygen-deficient nonstoichiometric oxides. Tench and I haveobserved oxygen species on the stoichiometric oxides MgO and CaO.1 On unirradi-ated MgO there is negligible adsorptionof oxygen and no paramagnetic signals appear.Although macroscopic stoichioinetry is retained, fast neutron irradiation producesmicroscopic non-stoichiometry with the formation of lattice vacancies and displacedions at interstitial positions.When oxygen is added at 20°C to MgO which has beenirradiated in vacuo with fast neutrons, oxygen is adsorbed and a complex signalappears on the low field side of the I; centre (fig. 1). This complex e.s.r. signal-":p = 4 x G-3 TOT p - i A X 0 - 2 'orr p -I 8 r lo-' Torr p c 10- 70 rcorresponds to bulk F-centres and 0 to adsorbed oxygen species.FIG. 1.-Effect of oxygen pressure on the e.s.r.spectrum of neutron-irradiated MgO. Signal Fincreases in intensity as further oxygen is added, with changes in the relative importanceof the constituent peaks. Above an oxygen pressure of 0-2 torr, pressure broadeningof the signal occurs. The signal can be removed by evacuation of the oxygen at 20°Cand has been assigned to oxygen reversibly adsorbed on the MgO surface. Thevariation in the structure of the peak as oxygen is adsorbed suggests that different sitesexist on the surface for the adsorption of oxygen, possibly having different character-istic heats of adsorption.Fig. 2a shows the signal from the adsorbed oxygen species on neutron-irradiatedMgQ which had been partially annealed at 300°C before adsorption of oxygen at 20°C.The spectrum is considerably simplified, with two sharp lines on the low field side ofthe F-centre, and this implies that the annealing procedure has destroyed certain of thesites capable of adsorbing oxygen. In this case the signal due to oxygen species issuperimposed on a signal connected with the irreversible adsorption of oxygen.The use of oxygen enriched in the 1 7 0 isotope (58 atom % 170) has indicated thatthe reversibly adsorbed oxygen species is diatomic rather than monoatomic.2 Keplac-ing natural oxygen (fig.2a) with 170-enriched oxygen (fig. 2b) produced an 80 %decrease in the intensity of the signal due to reversibly-adsorbed oxygen. Hyperfine1 Tench and Nelson, to be published.2 Tench and Nelson, J. Chem. Physics, 1966,44, 1714GENERAL DISCUSSION 323lines due to interaction with the 170 nuclear spin (I = 5/2) are not observed and maynot be detectable due to broadening. A 58 % decrease in the signal would be expectedfor a monoatomic oxygen species and an 80 % decrease for a diatomic one.It appears reasonable to conclude that the oxygen signals in fig.1 arise fromparamagnetic diatomic species of oxygen, probably O;, adsorbed on different defectsites on the MgO surface. In agreement with the work of Cornaz et aZ.,l no oxygenspecies have been detected at 20°C on the surface of unirradiated or neutron-irradiatedNiO, although considerable oxygen adsorption occurs.;. 1 0 3FIG. 2.-E.s.r. spectra of partially annealed MgO with adsorbed oxygen ; (a) natural abundanceoxygen, (6) 170 enriched oxygen.Dr.D. Iyengar and Dr. M. Codell (Pittzian Dunn Laboratory, Frankford Arsenal),Dr. J. Kana and Prof. J. Turkevich (Princeton Uniuersity) said: We report results ofan e.s.r. investigation that we have carried out which supplements those presented inthe paper by Schuit et al. The rutile powder was obtained from the National LeadCompany, was evacuated at 500°C and cooled in a vacuum to room temperature.It was then subjected to a variety of treatments with oxygen, hydrogen, oxides ofnitrogen at different temperatures. Varian e.s.r. equipment was used. The validityof the g values obtained was established by making measurements at 9.5 and 35kMclsec. The signals at room temperature were usually weak and broad. Themeasurements were made at liquid air-temperature and at temperatures of 4.2 and1 -6°K. The followiiig signals were obtained.Signal no.1 with g = 1.959 and g = 1.946, with a width of 30 gauss (fig. 1) wasobserved on the degassed sample cooled in vacuum to rooin temperature. Thesignal became more symmetrical at liquid helium temperatures and showed a relaxa-tion time of 2.3 x lo-* sec. Because of its g values, its temperature dependence, andchemical behaviour, it was assigned to a 3d 2 electron of Ti(lI1). The crystal field wasof tetragonal symmetry with A = 22,000 em-1 and 6 of 7,000 cm-1. This wasinterpreted to indicate thtt the Ti(II1) was produced by removal or displacement of oneof two of the two axial oxygen atoms situated at 1.988A from the titanium (III)?The characteristics of the observed e.s.r.signal excludes the possibility of an interstitial1 Cornaz, van Hooff, Pluijm and Schuit, this Discussion.2 von Hippel, Kalnajas and Westplul, J. Physics Chem. Solids, 1962, 23, 781324 GENERAL DISCUSSIONTi(II1) or a Ti(II1) located next to an oxygen hole produced by removal of one or moreof the four planar oxygens situated at 1.944 A away from the titanium or by a displace-ment of any of these oxygen atoms. The width of the signal is determined by thespin lattice relaxation time and the width observed indicates that the Ti(I1I) areseparated by at least 6A from each other. The signal disappeared on admitting atroom temperature either oxygen or the oxides of nitrogen and re-appeared on evacua-tion at 500°C or on hydrogen reduction at temperatures greater than 350°C.Thissignal seems to be the same as that reported by the authors as b and b'.\if FIG. 1.-E.s.r. spectrum (X-band) of rutile invacuum outgassed at high temperature. Thespectrum is due to Ti(II1) ion at liquidnitrogen temperature. 1 in. = 60 gauss on thehorizontal axis.Signal 2 (fig. 2b) was a symmetrical triplet with gz = 2.024, g, = 2.002 and gz =1.980. This signal saturates easily with microwave power. It was produced byadmitting oxygen, or oxides of nitrogen at room temperature to evacuated sample.Its intensity was enhanced by gaseous oxygen. It was permanently eliminated byoxygen at 600°C. Its intensity was decreased by hydrogen treatment at temperaturesabove 360°C.The signal saturated with microwave power at liquid nitrogen tempera-ture and was not observed at liquid helium temperature. On the basis of the fact thatone of the g values is greater, the other less than and the third equal to the value of thefree electron, we have assigned this signal to either O+ or 0; preferring the assignmentto O+ for reasons to be given below. We have adopted for the theoretical discussionof our system, the treatment by Schlichter for the Cl; colour centre.1 The shift weobtain is &i/S where A is the spin-orbit coupling constant for oxygen which we take as108 cm-1. The 6 is the difference between the state of unpaired spin and both excitedwhich can couple with it. In our case, experiment shows that they are equal. For theOf it is the energy separation between unpaired spin state characterized by the atomicorbitalp, and the excited unoccupied statepZ on the one hand, and the occupied filledstateps on the other hand (fig.34. For the 0,' it is the energy separation between thestate of unpaired spin characterized by the molecular orbital n: and the lowest statetszP on the one hand and .,* on the other hand (fig. 3b). The value of 6 obtainedexperimentally is equal to 5,000 cm-1. The e.s.r. measurements cannot differentiatebetween O+ and 0;. The assignment of the triplet no. 2 to O+ is based on theobservation that the signal is enhanced by oxygen pressure rather than broadened aswe would expect for the 0; because of the exchange reaction0; + o,+o, + 0;.The triplet no.3 was the same structure as triplet no. 1 (fig. 2 4 gs = 2.022, g, =1.999, gz = 1.977. It was formed by treatment of the degassed sample at roomtemperature with oxygen but not with oxides of nitrogen. It did not saturate withmicrowave power and thus could be distinguished readily from the no. 2 triplet. Itbroadened with excess oxygen pressure. Because of its similar structure to signalno. 2 triplet, it was ascribed to 0; but it might also be the O+. The uniform shift in1 Schlichter, FrincipZes of'Mugnetic Resonance, (New York, Harper Rowe, 1963), p. 195GENERAL DISCUSSION 325the three g values could be due to a contribution of overlap of molecular orbitals in the0; system which is lacking in the 0+.1 Our signals, no. 2 and no. 3 seem to be identi-fied with Cornaz et dl.signals d and e and we would suggest that e3 is really d2 and thetrue e3 was not observed.2 2aFIG. 2.-(a) E.s.r. spectrum (X-band) of rutile dried at 500" for 2 h and outgassed for 24 h at roomtemperature. (b) E.s.r. (X-band) of rutile heated at 5OO"Cin oxygen for 3 h. (30th spectra are obtained atliquid-nitrogen temperature.) 1 in. = 60 gauss on horizontal axis.Another signal no. 4 (fig. 4) arises on treatment of the degassed sample with oxygenand then evacuating. In samples that lack the no. 3 triplet, the signal no. 4 can beseen as a superposition of an asymmetric signal on the triplet no. 2. The g values are2.0106 and 2.0216. Because it is broadened by oxygen gas and because we findY6XFIG. 3 . 4 4 Atomic orbital diagram of 0; ; (b) molecular orbital diagram of 0;.analogous g values for 0; on porous glass, g = 2-0023, 2-0144 and 2.0302 and for 0;on zinc oxide, 2.0133, 2.010 and 2.04, we ascribe the signal no.4 (with gx = 2.002hidden by the gv of signal no. 2), gz = 2.0216, g, = 2.0106 to 0;. The broadening byoxygen we ascribe to the rapid exchange reaction. This signal may be identified withCornaz' signal C.1 Schlichter, Zoc. cit., p. 215.1326 GENERAL DlSCUSSlONThe chemical interpretation of the processes observed is as follows. On vacuumdegassing because of production of anion vacancies the following reaction occurs :(no. 1)The Ti3+ gives the signal that is observed. The hole is not observed. The hole is notobserved. On admission of oxygen at room temperature(no.4)Ti4+ +Ti3 -I- + hTi3+ + O2-+Ti4+ + 0,while the hole (h) combines with gaseous oxygenor0 2 + h-, 0,'0 2 + h-+O+ + 0.(no. 3)(no. 2)The e.s.r. evidence for the existence of 0; or O+ is strong and their formation issurprising since the ionization potential of 0 2 and 0 is 12.88 and 13.55 eV respectively.i j- - - - - - - 4 ---;'I ---_----FIG. 4.-E.s.r. spectrum (X-band) of rutileoutgassed at room temperature followingthe addition of oxygen at room tempera-ture. The spectrum \yas obtained at liquidnitrogen temperature. 1 in. = 60 gauss onorizontal axis.However, rutile has a high dielectric constant which varies from 86 to 170 D. Theionization energy of an atom or molecule in the interior of the crystal varies inverselywith the square of the dielectric constant of the solid and this would lower the ioniza-tion of 0 2 or 0 to a small fraction of 1 eV.The effect on the surface will not be asgreat but certainly might be appreciable to make the ionization of 0 2 or 0 energeticallyfavourable.The observation that Ti02 may favour ionization of oxygen molecule adsorbed onthe surface suggests investigation of Ti02 as a favourable support for reactions involv-ing carbonium ion as an adsorbed species in catalytic reaction.Prof. G. C. A. Schuit (Eindlzoven) said: We were interested to learn that onrutile several e.p.r.-signals of adsorbed oxygen are observed which might correspondto some of our signals on anatase. As to the interpretation we believe that a mole-cular orbital description is more adequate and can give more details than the essen-tially ionic description (with crystal field effects) given in Karra and Turkevitch'scomment.For instance, it is not readily understandable why species such as Ofand 02 could exist on an ionic lattice simultaneously with 0 2 - (or OF). Actually,0the 0; structure postulated by Ivengar et ul. and our perpendicular M= I structure0are formally equivalent: in both cases an electron is donated to a Ti ion. How-ever, whereas in the ionic-crystal field approach this elect1 on becomes effectivelya metal electron this is not so for the m.0. treatment: depending on the relativGENERAL DISCUSSION 327energies of the orbitals in the 1.c.a.o. combination it might be still to some extenta ligand electron.Considerations of this type apply especially to n-type bondingbecause there the @orbital is an antibonding orbital of relatively high energy.This means that contrary to a-type bonding where lower lying ligand orbitals areinvolved, electron transfer from the ligand to the metal ion is possible to a certaindegree notwithstanding the high electron-affinity of 0 2 as also reflected in itsionization pctential.We shall briefly discuss the signals no. 1 to 4 reported by Turkevich. Signal 1 :We agree with the correspondence to our signal b but this signal was considered tobe the partially exchange narrowed b-signal which has other g-values than thoseof signal 1. This may be due to the different structures of anatase and rutile.Signal 2 corresponds definitely to our d-signal on anatase. We suggest thatits disappearance at 4°K is a result of saturation. Our d-signal was attributed tothe P (perpendicular)-structure which can be pictured as 0,’ in the sense outlinedabove. We do not notice any important dependence of this signal, on anatase.Signal 3 : this has a definitely different stiucture than our e-signal. Our e3peak has been clearly separated from d2, so its form is well established. But it maywell be that this signal, corresponding in our approach to the A (angular) situationis more sensitive to the local symmetry of the substrate, owing to its own lowersymmetry. This is suggested also by the fact that only one g-value is stronglychanged with respect to the d-signal (or signal 2 on Ivengar et d.). So we canconclude that some correspondence may in fact exist between signal no. 3 on rutileand signal e on anatase.Signal 4 : the correspondence to our signal c is good. Again we differ in inter-pretation by using opposite approaches.To the remark on the ionization energy which might have its counterpart in ourdescription as an activation energy, we mention the following experimental result :if 0 2 is admitted to outgassed Ti02 anatase at -160°C a slow conversion of signalb to e was observed. However, this reaction takes place rapidly and irreversiblyat about -80°C. In these particular experiments, signal d was not observed.In reply to Nelson, our e signal has been ascribed to a a-bonding with 0 2 , con-sequently, it is a good approximation to write it as 0,. It should be the only oneto occur on substances having either no d-orbitals (as Mg) or largely filled d-orbitals(Zn). Therefore the peroxide-signal of Nelson should resemble our e-signal.Indeed, its structure appears similar although the splittings are smaller. More-over, its position on the low field side of the F-centre ( g > 2 ) also corresponds tothe situation of our e-signal.In reply to Stone, we repair the omission of a reference to the work of Stone andco-workers who introduced the idea of a reaction between 0- and OH- at the sur-face. This reaction may even be more important on anatase than on rutile, takinginto account the probably higher concentration of OH--ions. The f+h signal,which was tentatively attributed to 0-, possibly formed by such a reaction, is doubleand the narrow part of it (h) could not be handled correctly with our 100Hz-modula-tion apparatus. It appears that it may be due, at least in part, to some of ourquartz sample-tubes. We can now use a superheterodyne adaptor permitting low-frequency modulation. This, in combination with the illumination studies pro-posed by Stone will probably shed more light on this form of adsorbed oxygen

ISSN:0366-9033    

DOI:10.1039/DF9664100305

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Studies of Cations in Zeolites : Adsorption of CarbonMonoxide; Formation of Ni ions and Nai' centresBY J. A. RABO, C. L. ANGELL, P. H. KASAI AND VERNER SCHOMAKER*Union Carbide Research Institute, Tarrytown, N.Y., U.S.A.Received 3 1st December, 1965Zeolite cations lying on the intracrystalline pore surface of the Linde X and Y molecular sievesare linked to only three oxide ions and consequently are not well shielded electrically. They there-fore create very large electrostatic fields, extending into the main zeolitic cavities, causing carbonio-genic catalytic activity. This activity follows in magnitude the changes in the field, and is inde-pendent of the presenceof OH groups ; nochange in activity is observedeven after 99 %of OH protonspresent after ordinary activation in vacuum at 500°C are removed.Bivalent cations exposed onthe surface of the main intracrystalline cavities adsorb carbon monoxide with infra-red frequenciesspecific to cation ; the adsorption follows Langmuir isotherms, suggesting that a single carbonmonoxide molecule can be independently attached to every surface cation. Certain transition-metal zeolites adsorb carbon monoxide much more tenaciously than alkaline earth cations, indicat-ing a highly significant difference in their ability to form co-ordination bonds.Univalent nickel ions can be prepared both on the intracrystalline surface and at fully co-ordinated positions by heating NiIIY with alkali metal vapour. The surface NiI ions are chemicallyvery reactive but thermally unstable ; those at fully co-ordinated positions are thermally stableeven at 400"C, and are inert to H2, NH3, and CO although they react with oxygen to form 0;.The alkali-metal X and Y zeolites react with alkali metal vapour to form coloured products ofnon-stoichiometric compositions containing the paramagnetic centres Na 2 + and Naa +.The Na 2 +centres are stable up to 500°C and they react reversibly with gases ; with oxygen they form 0,radicals.Many anhydrous zeolites are adsorbents, having an enormous intracrystallinesurface associated with a homogeneous system of pores and cavities. Syntheticzeolites Linde Molecular Sieve Type X and Linde Molecular Sieve Type Y can bemade to show 1, 3 a wide range of carboniogenic catalytic activities-i.e., activitiesapparently due to the formation of carbonium-ion-like intermediates-dependingon appropriate selection of Si/Al ratio and charge and size of zeolitic cation intro-duced by ion exchange into the sodium-containing, as-synthesized materials. Thecarboniogenic activity increases with increase in cation charge and with decrease incation size and has therefore been attributed to the unusually strong accessibleelectrostatic fields of the fraction of the cations that lie on the intracrystalline surfaceand, in essence, to polarization of the substrate molecules in these fields.We briefly review in this paper the existing knowledge of the structures of theX and Y sieves, recapitulate the arguments on their catalytic activity, and indicatehow we were led to predict or recognize a number of remarkable new properties,all intimately involving the zeolitic cations, including strong characteristic gasadsorption and special chemical activities, e.g., special susceptibilities toward re-duction. Each bivalent surface cation seems to be able to bind a single carbonmonoxide molecule.The resulting shift in the C-0 stretching frequency dependson the cation and can be used to follow the distribution of cations between" surface " and " hidden " sites. The unusual reducibility of surface cations is* present address : Department of Chemistry, University of Washington, Seattle, Washington,U.S.A.32J. A. RABO, c. L. ANGELL, P. H. KASAI AND v. SCHOMAKER 329illustrated by the reduction of Nin ions in Nin-exchanged Y-zeolite to NiI by addedalkali metal, but hidden Nin can also be reduced to NiI.Another surface effectis shown when alkali-metal zeolites react with alkali metal to form products thatcontain complex ionic centres of unusual composition, e.g., Na: +, that in the zeolitenevertheless have a high degree of thermal stability. These complex surface centresalso co-ordinate various gases strongly.Throughout the course of our work on the adsorption effects, catalytic pro-perties, and general chemistry of these zeolitic surfaces we have profited by depend-ing on the ionic model and simple physical principles for guidance and inspiration.We therefore believe it is appropriate to follow this line of thought in describingthe results in this paper, even though it is insufficient to deal accurately and com-prehensively with our systems and the phenomena.STRUCTURAL FRAMEWORKDespite differences in origin, composition, and chemical and physical properties,the mineral faujasite and the synthetic zeolites of interest in this paper-LindeMolecular Sieve Type X and Linde Molecular Sieve Type Y-are related in crystalstructure to the extent that the basic framework is the same for all three.Theirstructures have been described in their general aspects 1-4 but are all too little knownin their detail. Because the basic framework of these materials is primarily involvedin the variations in the properties that are here our principal concern, namely, the netelectric fields, valences, and electron affinities of the zeolitic (exchangeable) cations,its relevant aspects will now be reviewed.We must emphasize, however, that theonly anhydrous material that has been studied by X-ray diffraction is single-crystalfaujasite that had been repeatedly exchanged with calcium ion and then dehydrated.We of course cannot be sure of the extent to which we are justified in applying theresulting information to the synthetic materials. Any speculation on fine details ofthe structures, for example on the ordering of A1 relative to Si, immediately leads oneto recognize many probable differences among faujasite, X and Y. As a probableresult of the chemical as well as structural differences there are differences in certainproperties, both micro and macroscopic, between the Molecular Sieves X, Y and thefaujasite. The absorption capacity of X and Y is greater than the faujasite,sa the Yadsorbs a " plug gauging " compound (triethylamine) which is excluded from X,and the Y has a far superior steam stability than the X.5bIn any event, the faujasite framework is a highly rigid, notably open, negativelycharged, complex ring system of composition (Si,Al)O2 in which SiO4 and A104tetrahedra are linked together by shared oxygen ions.The stainless-steel and plastic-spaghetti model pictured in fig. 1 shows that the framework can also be regarded interms of larger units linked together by oxygens. Such larger units, e.g., are the(Si,Al)12018 hexagonal prisms and the (Si,A1)24036 truncated octahedra (" sodalitebaskets ").Also shown in fig. 1 are the three possible cation positions of greatestinterest here: SI, SII and SIII. Site SI is hidden from the zeolite surface, beingintimately surrounded by the ions of the framework, whereas SI-I and S I ~ areon (or in) the zeolitic surface, the SU next to 6-oxygen rings on the three-foldcubic axes and the Sm next to 4-oxygen rings on four-fold inversion axes. AnX-ray study 5 of a dehydrated Can-exchanged single crystal of faujasite (Si/Al -2.5)revealed that sites SI (16 per unit cell) and Sn (32 per unit cell) are occupied by thecalcium ions (and any remaining original zeolitic cations). The average populationdeduced for SI was about 1 and for SII about 3 : the calcium ions seem to stronglyprefer SI over SII.A cation at Sn, lying on an almost planar, scalloped 0 6 ring330 STUDIES OF ZEOLITE CATIONShas only three-fold oxygen-ion co-ordination (Ca-0 = 2*34A), whereas in SI,positioned between two puckered 0 6 rings, it has six-fold octahedral co-ordinationto oxygen ions (Ca-0 = 2.42A). At Sn, calcium also has three significant next-nearest oxygen neighbours (Ca-0 = 2-97 A).FIG. 1 .-Cation positions in faujasite.The essential difference between X and Y and between the charge distributionsaround SI and SII then follow. In our discussion Type Y is understood to be Nay.Exchanged forms in which sodium is exchanged with other cations are designatedhere as Cay, ZnY, etc. Consider a Cay with Si/Al = 2. A calcium ion at SI issurrounded by 12 tetrahedra, four of which, on the average, contain aluminiumand bear a net negative charge.This calcium ion can be said to balance two of thefour negative charges, leaving two of the charges-half a charge per A104-to bebalanced by the zeolite cations in SJJ. In SII the zeolitic cation is surrounded onone side by six tetrahedra, on the average two A104 and four SiO4, having an averagenet total of one negative charge. All the tetrahedra are equivalently related to SIand Sn sites. Consequently, half of the Sn have a full positive net charge centredon the cation, while the other half have a full negative net charge distributed on theoxygen ions. By the same argument, the Sn: sites in the CaX zeolite with SilAl = 1would be effectively neutral : the bivalent cation at SI would face six A104 groups,each of which would have a net negative charge of 2/3, whereupon each Sn: with itsthree A104 would have an effective negative charge of 3 x 2/3 = 2.The same con-clusions may be reached in another way: CaY has 32 calcium ions per unit cell,16 at SI and 16 at SII ; if the 32 units of balancing charge for the Sn: cations areconsidered to be localized on the 32 Sn: sites, here regarded as equivalent, each musthave one unit of charge, half enough to neutralize a calcium ion. On the otherhand, a CaX with Si/Al = 1 has 48 calcium ions per unit cell, of which 32 matchthe charges of the 32 SII sites they occupy.Quantitative indications on the strength of SII cations were obtained by Pickertet d , 1 who calculated the electrostatic field along the three-fold cubic axis in thezeolite cavity near a surface cation at SII for a particular fully ionic model at a Yzeolite with Si/Al = 2.Further calculations were later made by E. Dempsey ** present address : Socony Mobile Research Laboratory, Princeton, N.J., U.S.AJ. A. RABO, c. L. ANGELL, P. H. KASAI AND v. SCHOMAKER 331on the same assumptions for a X zeolite of Si/Al ratio 1.0 and for positions in acavity along a four-fold axis through a cation at Sm. The fields (table 1) are larger(i) in the Y than in the corresponding X, (ii) with bivalent than with univalent cations,and (iii) near Sm than near SII, provided that both are occupied. The electrostaticshielding by the framework is unusually small at Sn: and Sm, although still greatenough to account for substantial differences between Sn[ and Sm and between Xand Y.Yet, despite imperfections of the ionic model the actual fields in the cavitiesat distances as great as several 8, from a bivalent cation are probably still as greatas 1 VIA.TABLE 1 .-ELECTROSTATIC FIELDS (VIA) IN x AND Y ZEOLITES Qcation charge + 1Si[Al 1 : 1site SIIweight 4 .r,2.0 1.522.5 -6 13.0 -233.5 *084.0 -02+ 11 : lSIII62-471-23-62-3 1-12+ 12 : 1SII42.301 -27-7752-38+ 12: 1SIIr23.211.881.19-79-543.21 : lSII45-653.221 -971.146 9+22 : 1SII26-303.852.5 11.731-26a calculated by E. Dempsey for positions on a symmetry axis at the distances r from the centreof a cation SII (see text) for ideally ionic, fully exchanged models of the X and Y zeolites (unpub-lished research at Union Carbide Research Institute).b number of equivalent occupied sites per (Si/A1)24036 basket.Around every cavity there are four tetrahedrally arranged SII and six octahedrallyarranged S ~ I .Since the SI and Sn: are more than sufficient to accommodate thebivalent cations the less attractive Sm sites are probably only populated in the uni-valent-cation form of the zeolite.It is thus evident from the structure and the electrostatic field calculations thatthe order of preference for strong cations, such as the alkali and alkaline-earthcations, is SI > Sn > Sm. This also suggests that bivalent ions will replace univalentions at the most preferred sites and that the higher the valence of a cation at a surfacesite the higher will be both the electron affinity of the cation and the field near thecation.And the stronger the field, the greater will be the polarization of adsorbedmolecules and the tendency for reduction of the cation. Some of these questionsrelate more naturally to the electrostatic potential at a cation site than to the electro-static field at points near the site. Nevertheless, we have so far dealt with the fieId,because it is the more important by far for our concept of the carboniogenic activityof the surface cations in catalysis.CATION-SPECIFIC ADSORPTION OF CARBON MONOXIDECarbon monoxide adsorbed on bivalent-cation X and Y zeolites shows a cation-specific infra-red stretching band as well as two bands that are cation non-specificto a remarkable degree, occurring unchanged or almost unchanged also with uni-valent-cation and decationized preparations.6 The specific bands do not appearif even a small amount of water is present, presumably because the cations bind HzOmore strongly than CO.The non-specific bands were shown to correspond to adsorption at two differentkinds of ubiquitous sites, but these sites were not further characterized.The specificbands, however, were shown to have possibilities for the study of the cation-C332 STUDIES OF ZEOLITE CATIONScomplexes to which they are due and of the cation distribution and cation environ-ment in the zeolite. For example, the shift of the band from the gas frequencywas shown to be directly proportional, for a wide selection of closed-shell bivalentcations, to the calculated field at the expected position of the carbon atom in theM---C---0 complex.For the bivalent transition-metal cations Fen, Con and Nin,on the other hand, the shifts scattered rather widely from the shift against field curve,as if not only a simple charge-polarizability bonding but also special covalent orligand-field bonding were involved. The specific band appeared even when the degreeof exchange of added ion for original sodium ion was less than enough to fill all the SI,showing that these cations did not realize the full expected preference for the hiddensite. Finally, the specific band is harder to pump away than the non-specific andharder to pump away at low temperature than at room temperature.We have now determined the pressure dependence of the peak optical densitiesof the specific bands for vacuum-activated samples of Cay, ZnY, Bay, COY, NiYand CaX. We have also measured the total gas adsorptions for Cay, ZnY, BaYand NaY and have made use of earlier data on CO adsorption by CaX,7 NaY,8COY 8 and NiY.8The experimental extinctions (mg-1 cm2) and gas adsorptions (s.t.p.ml g-1) areshown in fig. 2, 3 and 4 for some of the samples, together with curves calculatedfor Langmuir specific adsorption plus linear non-specific adsorption. Table 2 liststhe compositions, Langmuir and non-specific adsorption constants, and some furtherdata for all the samples.The notation isfor the extinction curves, andfor the gas-adsorption curves, where K is the Langmuir combination constant, Dthe peak intensity of the specific band, n the amount of CO adsorbed, and C theslope of the non-specific adsorption.D = DonoKpJ(1 +Kp) (1)(2) n = (noKpI(1 +KP)) + cpTABLE 2.-ADSORPTION OF CARBON MONOXIDE ON SEVERAL BIVALENT-CATION ZEOLITES,EFFECTIVE SITES, NON-SPECIFIC ADSORPTION COEFFICIENTS c, ANDRELATIVE MOLECULAR OPTICAL DENSITIES DoSAMPLE COMPOSITIONS, LANGMUIR COMBINING CONSTANTS K AND NUMBERS no OFexchange K C* n0 n0CaY 84 0-35 0.0144 0.148 13.7 7-6 8ZnY 66 *50 *0142 -152 4.8 1 2.8 3BaY 80 -00308 -0130 *113 24.6 16-4 6.8COY 78 103 *0135 -612 2.36 1.6 6.2NiY 76 18.7 ,0178 -380 3.86 2-7 5.6CaX 98 0.108 -0558 -370 12.7 8.0 25.6a for the NaY sample, C was found to be 0.0138 ml/g mm.b with D the optical density per mg of a disc sample about 1 cm2 in cross section.c assuming 16 MI1 in SI and the remainder in SII.zeolite % mm-* cm3/g mm Do cm3lg siteslunit cell Mlrin S1lCAll the extinction data fit the Langmuir curves (eqn.(1)) well or fairly well, imply-ing adsorption on approximately equivalent, non-interacting or only weakly inter-acting sites. The combining constants are generally consistent with the frequencyshifts, for the closed-shell ions, but for Nin and Con they are relatively very muchstronger and their order is inverted. When allowance is made for a non-specifiJ. A. RABO, C . L. ANGELL, P. H. KASAI AND V. SCHOMAKER 333adsorption proportional to the CO pressure and nearly the same for all the Y samplesincluding the Nay, the gas adsorption data also follow Langmuir curves (eqn.(2))with the same respective combining constants as were determined from the specificextinctions. The CaX adsorption does not fit very well and the COY and NiYadsorptions do not extend to low enough pressures to afford more than a purely00400 1-FIG. 2.-Peak intensity against pressure for the specific infra-red stretching band of CO adsorbedon Cay and ZnY. The solid lines are calculated from eqn. (1). 0 and 0 are experimental values.P (mm Hg)-COYP (mm Hg)-NiYFIG. 3.-Peak intensity against pressure for the specific stretching band of CO adsorbed on COYand NiY. The solid lines are calculated from eqn.(1). 0 and are experimental values.qualitative check on the combining constants. Finally, the two kinds of datatogether afford relative determinations of the molecular extinction coefficients ofthe MnCO complexes and of the effective numbers of adsorption sites. Exceptfor CaX the extinction values take the same order as the combining constants.The numbers of effective sites all correspond at least roughly with the degreesof exchange and the postulated structures, i.e., with the assumption that all Mnions in excess of 16 per unit cell go into Sn, The correspondence is good for Zn334 STUDIES OF ZEOLITE CATIONSand for Cay, which is the only case for which there is independent comparablestructural information, viz., Dodge's study of Can-exchanged faujasite.For BaYthe number of effective sites appears to be much greater than the number of BalIions in excess of 16 per unit cell, but this was hardly unexpected, since the large BaIIradius might well keep barium ion from penetrating the sodalite basket and, thence,the hexagonal prism that encloses SI. For CaX, NiY and COY the reverse is true :either the numbers of bivalent surface ions in these zeolites disagree with our presentexpectations or the old gas-adsorption results are for some reason in error.F------ /. I./' - L-. ---L - - ' - __ II 0 0 130 I IDp (mm Hg)FIG. 4.-Pressure dependence of the adsorption of carbon monoxide on Nay, ZnY and Cay.The solid lines are calculated from eqn. (2). 0, 0, and A are experimental values.REMOVAL OF OH GROUPS I N ZEOLITE CATALYSTSIt has been suggested 3 9 1 that the large electrostatic field near Sn cations in thezeolite cavities is responsible for the carboniogenic catalytic activity of bi- andter-valent-cation X and Y zeolites.As outlined above, this field is greater formultivalent cations than for univalent cations and greater for a Y zeolite than foran X zeolite, all in agreement with the observed differences in catalytic activity.Unfortunately, all the X and Y preparations cited showed weight-loss on ignitioncorresponding to 0.1-1-0 wt. % of OH groups,l which raises the question of whetherthe hydroxyl protons were significant for the carboniogenic activity. The OHgroups in zeolites have been studied extensively,9-11 and it has been established thatY zeolites activated by heating in vacuum to 550°C usually contain three distinctstructural OH groups with infra-red stretching frequencies of 3745, 3640 and 3540cm-1.The small 3745 cm-1 band is chemically inert : it is not affected by addingwater or benzene, and its intensity and frequency are independent of the cation.The other two, much stronger bands are shifted both by water and benzene, buttheir frequencies are only slightly affected by change in cation?To check on the possible significance of the protons we have attempted by heattreatment to remove the OH groups from samples of CaY and CemY using the3640 and 3540 cm-1 bands to measure OH removal. In one series of experiments,the processed pellets of the zeolite samples were heated to the desired activationtemperature in vacuum in a specially designed quartz cell and subsequently trans-ferred within the same cell into a 1 mm cavity between two NaCl windows.Thisprocess was repeated after every heat treatment as long as any of the OH bandsJ. A. RABO, C. L. ANGELL, P. H. KASAI A N D V. SCHOMAKER 335with the exception of the chemically inactive 3745 cm-1 band, remained. Fig. 5shows that the OH groups persist up to about 650°C but are removed to less thanabout 1 % of their former number between 650 and 710°C. Samples of each ofthe zeolites so heat-treated were also exposed to water vapour at room temperatureand then activated in vacuum at 500°C; spectra were taken after every step. Theadded water did not restore the 3640 and the 3540 cm-1 bands, and it can be readilypumped off at 500°C. All these samples were also X-rayed before and after heattreatment, but no changes could be seen in the diffractometer traces.3750 3650 4550 34503? 55 2650 3 3 5 i ?4>"cm-1Fro.5-Infra-red spectra, OH region, of Cay and CeIIIY. - activated at 500°C ; - - - activated at 700°CTwo samples of each of three zeolites (Nay, Cay and Cemy), one activated atthe standard 550°C and the other (with the exception of NaY) at 700"C, were testedfor cumene cracking in a 5-cm3 reactor attached to a gas chromatograph. One-hundred-microgram samples of cumene were injected into the stream of heliumcarrier gas. In this system 12 at low temperatures the catalyst bed also acts as aninterfering chromatographic column ; it therefore had to be run at temperaturesabove 400°C in order to permit good resolution of the product peaks.With NaYless than 10 % of the cumene was converted to benzene and propylene at 450°Cwhile with CaY and CemY the conversion was greater than 99 % for both sets ofcatalyst samples. Because of the high conversion, it is not possible to say whetherthe Cay and CemY activities changed at all on OH removal; but they did notfall to the NaY level, which we take here as the standard for a zeolite catalyst lackingthe special activity characteristic of the addition of multivalent cations. It is there-fore clear that the enormous rise in activity on replacing univalent cations wit336 STUDIES OF ZEOLITE CATIONSbivalent cations is due directly to the cations themselves and is not mediated by the0.1-1.0 wt.% of structural OH groups that are present after normal activation :this experiment has lowered the range for equivocation on this point to less than0.01 wt. % of OH.Tervalent-cation zeolites, e.g. C e q , probably require special attention, however.While we think that most bivalent cations will be stable in the anhydrous X and Yzeolites in either SI or Sn, we expect that the effective electrostatic field of tervalentions may always become strong enough on dehydration to cause hydrolysis,MmY + € 3 2 0 +(OH-Mm)HIY. The resulting OH groups on the tervalent cationsmight be more significant proton donors than the OH groups normally presentin a bivalent-cation Y.The latter OH groups, according to the infra-red studies,gare not linked to the surface cations. In the CemY, on the other hand, there is indeedstrong evidence for OH groups linked to the cation (band at 3522cm-1). It isalso conceivable that tervalent ions when fully dehydrated may shift to new positions ;and since in Y zeolite the number of SI positions is larger than the number of ter-valent ions, possibly only a small fraction will remain at surface sites. Conse-quently, the catalytic activity may diminish in certain cases to an extent dependingon the activation conditions.REDUCTION OF ZEOLITES BY HYDROGENOur discussion of the structure of the X and Y zeolites suggests that the surfacecations should have uniquely high electron affinities and related tendencies towardreduction.Indeed, zeolites have been treated with various reducing agents,l39 14and a lack of thermal stability of certain cations has been demonstrated by Yates,lswho showed that Cd and Hg metal can be vaporized from the corresponding CdnXand HgnX zeolites by heating at 140-500°C in hydrogen.Several years ago we observed another interesting reaction between noble-metal-loaded bivalent-cation zeolites and hydrogen gas. Unlike the non-loaded forms,these zeolites consumed large quantities of hydrogen (fig. 6). This reaction firstattains appreciable rates at temperatures that show a strong inverse dependenceon cation electronegativity ; e.g., 250°C for NinrY and CODY, 400" for MgEY, and500" for BaDY.The highest hydrogen uptakes reached values of about a moleculefor every two multivalent Sn cations. Since the Y zeolites without noble metalin general did not consume any hydrogen at these temperatures (NiIIY and one ortwo other transition metals were exceptional), we attributed this effect to activationof the hydrogen on the very finely dispersed noble metal, followed by reaction withthe zeolite, probably indeed with the surface cations. Furthermore, since in thisreaction the H/Pt atomic ratios climbed to values as high as 20-30, we assumed thateither the noble metal forms a surface metal hydride 16 complex (PtH2) and migratesin the zeolite cavities, functioning as a hydrogen transfer catalyst, or the protonsformed migrate by jumping from oxide ion to oxide ion.We hoped to find thatthe multivalent zeolite cations in SII were reduced to the univalent state, while OHgroups were formed by addition of protons to framework oxide ions. However,we found ferromagnetic Ni" and Coo in the hydrogen reduced Pd- or Pt-loadedsamples of NiDY and ConY and were unable to obtain convincing e.s.r. evidenceof the presence of univalent cations (e.g., MgI, CaI) in the hydrogen-treated alkalineearth forms. In any case, since noble-metal-loaded alkali metal zeolites do notreact at least up to 500"C, the reaction with hydrogen does depend on the bivalentcations; they therefore surely play a role in the reaction, even though the ultimatesite of reduction has not as yet been definedJ. A. RABO, C.L. ANGELL, P. H. KASAI AND V. SCHOMAKER 337REDUCTION OF Nin IONS I N NinY BY ALKALI-METAL VAPOUR TOFORM NiIAfter the ideas discussed above had led to the hope that novel materials could beprepared by reducing multivalent surface (Sn) cations in Y zeolites to unusual lowervalence states, and after the discovery of large hydrogen uptake by noble-metal-loaded multivalent-cation pr' zeolites had given promise that this hope might berealized, we began experimentation with various reducing agents and many different0.080.060.0 40.020 . 0 0 40 .H2 1* r6 zAnaam a 350°C863e-1 450°C 4 48 50Qd~401pC342pcxI I I0.200.180.160'140'12 g .-d0.100'08 cd0.060.04c, Eo5:0.000 I00 200 30 0 400time, hexposure to hydrogen was 300°C in all cases.FIG.6.-Reaction of noble-metal-loaded zeolites with hydrogen. The initial temperature one, CaYx 1.25 wt. % Pt0, MgY x 0-5 wt. % PdA, CoYX0.5 wt. % PtM, BaYx0.1 Pdx, MgYfinal atom ratio H : Pt or Pd13-19.614.015-0zeolite preparations. Many attempts were made in collaboration with Dr. K. K.Miiller, but none of these systems led to such good results as the ones to be describedin the rest of this paper, although NiI was obtained and definitely characterized insome of the early experiments with NinY. The most remarkable results have beenobtained with sodium vapour or, in some cases other alkali-metal vapours. Thepresent section describes results on NPY, covering first the reduction techniqueand some of the key experiments, then the significant details of the resulting e.s.r.and optical spectra, which we ascribe to NiI, and finally some of the other propertiesof the reduced materials.REDUCTION TECHNIQUEIn a typical reduction of N i V , sodium was placed at the bottom of a quartztube, the zeolite sample, preactivated in vacuum at 575"C, was supported 2 in338 STUDIES OF ZEOLITE CATIONSabove it on a small wad of glass wool held by indentations in the tube, the top ofthe tube was connected to a pumping system and continuously evacuated, and afull-length heating mantle was brought to 585°C to vaporize the sodium and heat thezeolite bed.At this temperature, the vapour pressure of sodium is about 10mm.The usual reaction time was 1 h, after which the heating mantle was removed andthe tube allowed to cool.REDUCTION OF Nin(72 %)YIn the first experiment NinY was exposed to a large excess of sodium (Na/Ni> 4)for 1 h.In addition to a strong, broad background due to Ni", the product showeda small e.s.r. signal at g = 2.065 which we attribute to NiI. In the next experimentNa/Ni was only 0.1, and the tube was heated at 585°C for 16 h. In this case onlythe signal due to ferromagnetic Ni" was observed, which suggested that on excessiveheating the NiI ions disproportionate : [2 NiI]Y + N i q + Ni".In order to optimize the amount of the reducing metal it was considered thatthere should be only one sodium atom for every Sn nickel ion, to avoid any excessof sodium that would further reduce the NiI to the metal.For the first experimentswe assumed that only the Su nickel ions would be reduced. In the NinY used inthis study, 72 % of the original sodium ions were replaced by nickel ions, and weassumed that about 30 % of the nickel ions would take surface positions whilethe rest would occupy SI. Allowing for some loss of sodium through reaction withthe glass wool and the quartz tube, we chose Na"/Ni= = 0.4 as standard in reducingthe Ni (72 %)Y. The reduction was carried out at 575°C in 1 h. With this ratiothe zeolite turned from pink to green, and although the e.s.r. spectrum was still mainlydue to Ni", the NiI signal (fig. 7) was much larger than from the previous samples.Two Ni (72 %)Y samples were reduced at 575°C for 1 h in vacuum, one by sodiumand one by caesium.Both samples turned green and gave an e.s.r. signalat g = 2.065,but the sample treated with sodium had the more intense colour. The amounts ofalkali metal consumed were calculated from the amounts of hydrogen that wereevolved on leaching the reduced samples in acid : Na/Nitotal = 0.12 and Cs/Nitotal =0.14. On the basis of the e.s.r. signal, the amount of NiI in these samples was about3 % of the total Ni. Three-quarters of the alkali-metal content therefore remainedunreacted or was consumed in reducing Nin to Ni". If all the NiI occupied Sn,the fraction of the NiI ions at the surface positions was about 10 %. The e.s.r.signal and colour of these samples remained unchanged even after several monthsat room temperature.Samples of Ni (72 %)Y were also reduced at 250°C for 108 h, a long reactionperiod chosen to allow the alkali metals to vaporize even at this low temperature.With Na", KO and Cs", these samples turned respectively green, brown and yellow-gray, and e.s.r.indicated 1 % NiI, no NiI and 1.5 % NiI, while the amounts of Ni"were much smaller than in the samples treated at high temperature. After 4 daysat room temperature, the NiI(Cs") signal (the signal obtained by using caesium asreducing metal) vanished and the NiI (Na") signal decreased greatly. By acid leaching,Na?/Ni was 0.59 and Cs"/Ni was 0.54, much larger than was obtained at 585°C.REDUCTION OF Nin(5 %)YIn the experiments with Ni (72 %)Y we assumed that sodium reduced only theNin ions at SD and that the SI ions were unaffected.To check on the reducibilityof Nin ions at SI, samples of a Ni (5 %)Y were also treated with alkali-metal vapour.We recall that bivalent cations in general prefer SI to Sn. With only 8/3 nickelions per unit cell, Ni (5 %)Y therefore should have almost all its Ni at SI. SincJ. A. RABO, C. L. ANGELL, P. H. KASAI AND V. SCHOMAKER 339NiII should be more difficult to reduce at SI than at SU, a large excess of sodium wasapplied at 575°C for 1 h i r vacuum. The sample turned from pale pink to palegreen and developed a sharp e.s.r. signal (fig. 8) at g = 2.0?4, apparently shiftedfrom the g = 2-065 of surface NiI because of the difference in crystal field betweenSI and Sn. The relative sharpness of the signal indicates decreased interactionbetween nickel ions, as expected.By acid leaching, Na"/Ni was 4.0.gL=2O65FIG. 7.-E.s.r. spectrum of surface NiI ionsin Ni" (72 %) Y reduced by sodium vapour.FIG. 8.-Em. spectrum of the fully co-ordinated (hidden) NiI ions at SI in Nin(5 %) Y reduced by sodium vapour.In order to confirm that the novel stability of NiI is peculiar to the zeolite sub-strate, a y-alumina loaded with 5 wt. % of NiClz was treated with sodium vapourat 575°C for 1 h in vacuum. The brown product gave a Ni" e.s.r. signal but noNiI signal.E . S . R . SPECTRA OF NiI IN REDUCED NinYAs mentioned earlier, the symmetry at SI is slightly distorted octahedral, thesite being surrounded by 6 oxide ions. Therefore, NiI at SI and Cua (also 3d9)in a similar environment are expected to have similar e.s.r.spectra. The spectrumof CUE ion would be the more complex, however, because of hyperfine interactionwith the magnetic nuclei Cub3 and Cu65. In contrast, all but 1.2 % (NW) of theNi nuclei are non-magnetic. The e.s.r. spectrum of a sodium-reduced Ni(5 %)Ysample in which most of the Ni ions are believed to be at SI is shown in fig. 8. Thispowder-pattern spectrum is characteristic of a system with an axially symmetricg-tensor :These values are closely similar to those found for CUD in various tetragonallydistorted octahedral ligand fields." Their deviations from the free-electron value*The comparison should best be made against the spectra of CuIIY zeolites. Such spectrahave been reported by Nicula, Stamires and Turkevich (J.Chenz. Physics, 1965, 42, 3684), who,however, made no attempt to interpret the e.s.r. spectrum of dehydrated CuIIY in terms of specificcation positions. Our independent study with a thoroughly dehydrated CuIIY sample is still inprogress. Although a complete assignment of the spectrum has not yet been achieved, becauseof the complexity of the hyperfine structures, it is clear that the Cull ions are distributed betweenat least two different kinds of sites.911 = 2.30-2.40; 91 = 2*0963-0*003340 STUDIES OF ZEOLITE CATIONSshould then be given bywhere II is the spin-orbit coupling constant of the Cuu ion, and AEll and AE, arethe respective energy separations of the dxy and d,, (or dyz) orbitals from the lowestorbital, d x 2 - y 2 .Site Sn: lies outside a puckered six-oxygen ring and has C3v symmetry.In de-hydrated Cau-exchanged faujasite,S it lies 2-34A from the three nearest oxide ionsbut only 0-57 A above the plane defined by them (< 0-Ca-0 - 114"). For NiIions at this site there are then three d-orbital energy levels of which the dz2 level(now calling the z-axis the C3v axis and resorting to the limiting planar case for labels)and dxz, d,, level are clearly expected to be above the dxy, d x 2 - y 2 level. Let us supposethat the degeneracy between dxy and d x 2 - y 2 is removed by some further mechanism.The expressions given above are then applicable to the g tensor of NiI at Sn also(fig. 7). We believe that this spectrum is dominated by the NiI (SII) signal super-posed on a broad background due to nickel metal.Analysis of the NiI signal gives911 = 2-10-2.13 and g1 = 2.065+0-003.It is easy to understand that AE, should be larger for Sn: than for SI, and henceE 8A/AEj, ; A g l z 2A/AEL (1)FIG. 9.-E.s.r. spectra of NiI ions at SIand SII in Nil1 (5 %) Y reduced by sodiumvapour, ~ ; after exposure to carbonmonoxide, - - -.that Ag,-should be the smaller at the sur-face sites. A subtle point is the mechanismby which the degeneracy between the dx2-y2and dz2 at SI and the degeneracy betweendx2-y2 and dxy at Sn: are removed. Jahn-Teller distortion might be responsible.However, these degeneracies are most prob-ably removed by the extra negative chargeassociated with each A104 tetrahedron.The existence of a rather well-defined g1component for NiI ions at Sn might thenbe taken as evidence for a well-orderedarrangement of A1 and Si atoms within thehexagonal rings.That there are two uniquesites for NiI ions is further supported by thee.s.r. spectrum of another Ni ( 5 %)Y re-duced by Na vapour which shows NiI ionsat both sites (fig. 9) and, as will be discussedlater, by the differences of their temperaturestabilities and of their reactions with ad-sorbed gases.OPTICAL SPECTRA OF NiI I N REDUCED NiaYAnhydrous Nin(75 %)Y is pink, but on reduction turns bright green. Thereflection spectra of Ni (5 %)Y and Ni (75 %)Y were examined in the visible regionbefore and after reduction by Na vapour (fig.10 and 11). The Ni(5 %)Y spectrumshows no prominent peak either before or after the reduction and therefore maybe used as a standard for background. The broad absorption band centred at450 mp that is observed with Ni (75 %)Y before the reduction must then be attributedto NiII at Sn:. Upon reduction, this band diminishes while a new band appearsat around 650 nip. The new band we attribute to an 2Eg+2Tg transition of NiIat SU. As would be expected for iso-electronic d9 ions, CuIV and NiIY showsimilar optical spectra (fig. 12)J. A. RABO, c. L. ANGELL, P. H. KASAI AND v. SCHOMAKER 341The spin-orbit coupling constant for a free NiI ion is 600 cm-1. If one uses thereduced value A = 500 cm-1 for NiI in the zeolite, taking into account the efiectof the neighbouring oxide ions, the expression given above for Ag, and the valueof Asl observed for NiI at Srr: yield AE = 15,000 cm-1, in excellent agreement withthe observed optical transition.8e2 ncdcd1 1 1 , 1 , 1 1 , , [ 1 , 1 ~ 1 1 , 1 1 , , 1 1 ~ ,400 500 600mPFIG.lO.-Optical spectra of activated NiII (5 %) Y , --- ; and of Nil1 (5 %) Y reduced bysodium vapour, -.- - - I i - I j I 1 - I r-J-p----- 1 1 1 - -;I-, I 1 1-1-IIA 0 3 500 €00mPFIG. 1 1.-Optical spectra of activated Nil1 (75 %) Y , - - - ; and of Nin (75 %) Y reduced bysodium vapour, -.EFFECT OF HIGH TEMPERATURESince the NiI formed in NPY showed excellent stability at room temperature,we also studied it at high temperatures. It was observed occasionally that the e.s.r.signal of reduced Ni (5 %)Y increased on heat treatment.To investigate thi342 STUDIES OF ZEOLITE CATIONSphenomenon a Ni (5 %)Y sample with about 6 % of the NiII reduced to Nil wasevacuated and sealed in a tube. After each heat treatment the intensity of thee.s.r. signal was examined :temp. O C time, h Nilsignal100 18 slightly larger than original200 24 twice the original300 24 6.5 times the original350 18 7 times the original ** N i I / N i t , , ~ 4 4 .The last signal remained unchanged for months at room temperature. The samplewas then heated to 500°C for 24 h, whereupon it showed a large Ni" peak and aconsiderably diminished NiI peak.d, I I 1 1 I I I 1 1 1 1 1 1 1 1 1 I I t I I400 500 600 700mPexpected was therefore in the range 1-5 %.FIG.12.4ptical spectrum of Cd*Y activated at 300°C. The remaining water content to beA Ni (5 %)Y reduced by a large excess of sodium showed two e.s.r. signals,one large with gL = 2-094 and one small with gL = 2.063 (fig. 9), indicating thepresence of NiI both at SI and SU. On heating at lOO"C, the NiI(Sn) signal slowlyvanished while the other slightly increased. A sample of Ni (5 %)Y reduced bycaesium was heated in vacuum gradually to 560°C ; up to 400°C the NiI(S1) increasedgradually by about ten-fold, after which it remained unchanged at 460°C for 24 h.At 540°C the NiI(S1) decreased and the Ni" signal appeared.Several times NiI(Sa) was found in small quantities when Ni (75 %)Y was flashactivated at 585°C. However, acid leaching released no H2, indicating that thisNiI was produced by electron dislocation rather than by any change in overallchemical composition.The signal is thermally unstable, like the one produced bysodium vapour. The reduced Ni (72 %)Y showed a similar temperature dependence :the stability at room temperature was excellent, but at and above 100°C the NiI(Sn)signal and the strong green colour vanished and Ni" appeared.EFFECTS OF ADSORBED GASES ON NiIThe profound difference in co-ordination between NiI(Sn) and NiI( SI) lendsitself to study of their abilities to co-ordinate with various ligands. The expectatio1. A. RABO, C. L. ANGELL, P. H. KASAI AND V. SCHOMAKER 343is that NiI(S1) will have virtually none. Samples of reduced Ni (5 %)Y, Ni (75 %)Yand Ni (25 %)Y were used with 0 2 , CO, H2 and NH3.Before and after admittingeach gas the e.s.r. spectra were recorded.The reduced Ni ( 5 %)Y used for CO adsorption showed a large e.s.r. peakat g1 = 2.093 ascribed to NiI(SI), a small peak at gL = 2.063 ascribed to NiI(Sn),and a third peak not due to nickel. On admitting CO at room temperature theNiI(Sn) peak vanished while the NiI(S1) peak was unaffected. The reduced Ni (75 %)Yhad a strong NiI(Sn) peak at g1 = 2.063. On admitting CO this peak vanished,while a small peak, previously masked by the much larger one, appeared at gl =2-094. Since this g value is identical with the one observed with reduced Ni (5 %)Y,we ascribe it to NiI(S1).Oxygen was adsorbed on a Ni(5 %)Y reduced by caesium.Before admittingoxygen, the e.s.r. spectra indicated NiI(S1) as well as a trace of NiI(Sn). On ad-mitting oxygen at room temperature both peaks vanished; however, on pumpingfor 30 min the original peaks re-appeared with almost unchanged intensity. Onpumping further at 100 and 250"C, the NiI(S1) diminished to 1/40 in intensitywhile a peak due to 0; appeared. At the same time, the NiI(Sn) vanished. Onfurther heating in vacuum at 425"C, the NiI(S1) grew to about 80 % of the originalintensity, while 0; vanished. On heating above 500°C, NiI(S1) began to diminishand the signal assigned to ferromagnetic Ni" appeared.With reduced Ni (25 %)Y (Na"/Ni = 0.68) on exposure to 0 2 at 25"C, all signalsvanished; but on pumping at room temperature for an hour, 90 % of the NiI(S1)was recovered while the NiI(Sn) grew to twice the original intensity ; and on heatingat 100°C in vacuum, both NiI peaks vanished again and a large 0; signal appeared.Samples of reduced Ni (5 %)Y and Ni (75 %)Y were exposed to NH3 at room tem-perature; according to the e.s.r.spectra the NiI(Sn) was destroyed and metal wasformed. The NiI(S1) was not affected at all.Similar Ni(5 %)Y and Ni (75 %)Y samples were reduced with sodium withrespective Na"/Ni ratios of 11.6 and 0-96. On exposure to H2, the NiI(Sn) slowlydiminished even at room temperature and vanished at 100°C with concurrent Ni"formation, while the NiI(S1) was unchanged even after 5 days at room temperature.REDUCTION OF NaY AND NaX WITH SODIUM VAPOUR:THE COMPLEXES Naz+ AND Na:+In the course of studying the reduction of various cation-exchanged zeolites bysodium vapour, it was discovered that NaY upon exposure to sodium vapour at580°C turns bright red and shows a unique e.s.r. spectrum consisting of 13 peakswith intensities gradually increasing toward the centre (fig.13).Development of pink colour and a similar e.s.r. spectrum were observed earlier 17when NaY was exposed in vacuum to y-rays or X-rays. Both the colour and thee.s.r. spectrum were attributed to electrons trapped in the large zeolite cavities,each electron being shared among four sodium ions tetrahedrally arranged at thefour surrounding Sn. We believe that the new, red, non-stoichiometric NaYprepared with sodium vapour contains similar Naz + paramagnetic centres.The centres produced by radiation are stable at room temperature, but are quicklydestroyed at about 200°C.The chemically prepared centres, on the other hand,show remarkable stability : the red material evolves no sodium on heating to 500°Cunder high vacuum, and afterwards the colour and e.s.r. spectrum are not significantlychanged. Introducing oxygen to the chemically prepared material at room tem-perature results in instant disappearance of both the colour and the e.s.r. signa344 STUDIES OF ZEOLITE CATIONSascribed to Nai+ and in the appearance of a strong new e.s.r. signal attributableto the 0; radical. The bonding with oxygen is reversible: heating at 500°C invacuum removes the 0; and restores the Nai+. The g value (1.999+0.001) andthe hyperfine interaction with the sodium nuclei ( A = 32.3+0-2 gauss) are identicalfor the chemically prepared Na:+ centre and the irradiated material.FIO.13.E.s.r. spectrum of the Naa+ centre in Y zeolite.R - 100 gaussLFIG. 14.-E.s.r. spectrum of the Nad * centres in X zeolite (first half of spectrum).This success prompted us to treat NaX similarly. The NaX used has a Si/A1ratio of about 1.19 and a correspondingly larger content of NaI than the Nay,On exposure to sodium vapour, this NaX turns dark blue and shows a simple e.s.r.spectrum consisting of 19 lines (fig. 14). This spectrum corresponds to a centresimilar to that obtained with NaY but involving six equivalent sodium ions insteadof four. The g value is the same for the blue centre as for the red, but the hyper-fine interaction per Na nucleus is reduced to about 2/3.Each large cavity in thestructure is surrounded by six Sm arranged octahedrally in the directions of the fourJ. A. RABO, C. L. ANGELL, P. H. KASAI AND V. SCHOMAKER 345fold cubic axes as well as by the four Sn: arranged tetrahedrally in the directions ofthe three-fold cubic axes. The blue centre, we believe, is an electron shared amongsix sodium ions all situated at S ~ I : the SII of the cavity are then necessarily vacant.This assignment suggests that in NaX, in contrast to Nay, Sm is occupied in prefer-ence to SII in order to accommodate the increased number of sodium ions.Magnetic susceptibility measurements on NaY reduced by sodium vapour showedparamagnetism corresponding to 50-80 % occupation of the cavities by Naifcentres. On the other hand, estimates of spin concentration from the e.s.r.spectraof similar preparations gave only 1019 to 1020 spins/g (2-20 % occupation), dependingupon the sample. This is hardly good agreement, but is reasonable, since e.s.r.detects only well-defined centres whereas all of the paramagnetic centres contributeto the susceptibility.DISCUSSIONAs has been reviewed above, the " surface " cations in anhydrous X and Yzeolites are incompletely co-ordinated with oxide ions, so that for molecules in thezeolite cavities the (enormous) electric field of a surface cation is not effectivelyshielded out by the fields of all the other ions. Essential consequences of this specialaccessibility of the cations are the carboniogenic activities and the specific complex-ing of, e.g., carbon monoxide.The shielding is still expected to be appreciable,however, and different for X and Y, and the differences between X and Y in catalysisand gas adsorption seem to correspond.The low co-ordination of a surface cation by oxide ions should also lead tohigher (less negative) electrostatic potential at the surface cation site than at thesite of a fully co-ordinated cation (which we also call hidden cation) and so shouldcause a surface cation to have greater effective electron affinity than a fully co-ordinated cation, at least to the extent that the structures in question can be con-sidered to be fully ionic.This concept follows immediately from the Born-Haberconsiderations for building up an ionic crystal from its component atoms. Weused it, loosely, to explain our discovery that multivalent surface (Sn, S ~ I ) cationscould be reduced to subvalent states not ordinarily found in molecules or crystals,that this reduction is easier in Y zeolite than in X, and that it should depend on thesequence of electron affinities of the cation. Such thoughts led us to NiI(Sn) butleft us surprised to get NiI(S1) and Na: + : they are useful but incomplete. Withoutdoubt, polarization is much more important for a cation in a surface site than fora fully co-ordinated, highly symmetrical site. The bonds are probably covalentto various degrees. Evaluation of the Madelung sums, especially with respect tothe full process of introducing both electron and sodium ion, is necessary.Ourknowledge of the structures is insufficient: for a small cation at SII the primaryco-ordination probably becomes planar rather than pyramidal, and major unexpectedrearrangements may possibly occur.We now comment on the reduction process NinY +Na"+NiINaW in which abivalent cation is replaced by two univalent cations, of which the Nil remains atthe original Nin site. In the Madelung energy, the sum of terms involving a NiI,NaI pair and distant parts of the crystal will usually equal the corresponding sumfor the original Nin, while a repulsion term between each NiI and its NaI neighbouris added that has no explicit counterpart. The repulsion terms between neigh-bouring NiI, NaI pairs, moreover, will be greater than the counterpart terms forthe original NiII ions.In addition, each new NaI may or may not go into a siteequivalent to the Ni* site with respect to interaction with the nearby part of th346 STUDIES OF ZEOLITE CATIONSframework. Equivalence will prevail for reduction of Sn cations if the new NaIalso occupy Sn. The net change in crystal energy will then strongly oppose thereduction, but the process finds its support in the enormous difference between theionization potentials of NiI and Na" : Nii& + Na&+NiiaS + Naias + 13.1 eV. Inthe reduction of Nin at SI the new sodium ions again probably occupy Sa, whichis less favourable for the crystal energy than SI; the net change in crystal energyis therefore also less favourable than for the reduction of Niu in Sn.Reduction of surface ions may or may not require activation energy, but reduc-tion at SI requires that electrons penetrate the surrounding negatively charged oxideions, which will require significant activation energy.The chemical stability of univalent nickel will depend also on the rate of dis-proportionation, [Z NiIJY + N i q + Ni".Since nickel atoms probably migratefreely on the zeolite surface even at moderate temperatures, they will agglomerateand form the metal, which will make the overall reaction substantially irreversible.In any case, disproportionation requires electron jumps between NiI ions eitherdirect or indirectly, and increasing the distance between NiI ions, i.e., lowering theconcentration, will reduce the rate.At SI, NiI is much less mobile than at SH,and the electron jumps presumably also require higher activation energy. Further-more, metal formation will be prevented or strongly hindered because the mobilityof Ni" atoms will also be less at SI than at Sn : to move from SI to SH, they have topenetrate two six-oxygen rings, each too small to give them free passage. HiddenNiI is therefore expected to be much more stable than surface NiI toward dis-proportionation.DISCUSSION OF THE EXPERIMENTAL RESULTSSURFACE NiI IONSThe step Ni&face + Na" +Ni&,faaNa' is probably followed in the presence ofexcess alkali metal by the step Ni~,,fa,+Na"+NaI+Ni". However, the first re-action probably can be carried out stoichiometrically. This reflects qualitativelythe great difference in energy between the following two gas reactions :Ni:as + Na,",,-+Ni:,, + NaLa, + 13.1 eVNi~as+Na~a,-+Ni~a,+Na',,,+2.5 eV.Not only large surplus of alkali metal leads to Ni" formation, but also long ex-posure at high temperatures (100°C and up) through the disproportionation[2NiI]Y-+NiIW+Nio.Since our study of the surface ions was done with a ratherhighly exchanged (72 %) N i V , the rate of electron exchange between NiI and Ninis fairly fast and leads to disproportionation of the surface Nil. Much less forma-tion of Ni" by disproportionation would occur in a NiY containing less nickel.Nevertheless, the hidden sites will be preferentially filled by Nia ions.The lowstability of the surface NiI prepared at low temperature (250°C) is due to excesssodium stored in the zeolite cavity system on reduction. This sodium then reactswith the surface Nil even at room temperature.The experiment with reduced Ni(25 %)Y, which first showed an increase insurface NiI at room temperature on exposure to oxygen, may support the suggestionthat NiI is more stable relative to Nin in SU than it is in SI : the increase may be theresult of the reaction Ni6idden + Nii&a, -+Nizdden + Niiurfa,. This reaction is ex-pected only between adjacent ions and is therefore unlikely in the Ni (5 %)Y. Itsnon-appearance with Ni(75 %)Y, on the other hand, may result from dispropor-tionation of the highly concentrated surface ionsJ.A. RABO, C. L. ANGELL, P. H. KASAI AND V. SCHOMAKER 347Surface NiI has a remarkable chemical activity. It co-ordinates with H2, NH3,CO, olefins, 0 2 , and probably with many other ligands. With NH3 the metal isformed even at 25°C. This is not surprising because NH3 is a very strong ligandand likely to weaken the binding of the NiI cation to the zeolitic surface and there-by lower the activation energy of migration. In addition, ionization of NiI byelectron jump will be further strongly assisted by the strong-electron-donor ligand.The reaction of NiW with H2 to form metal is of special interest. It occursalready at 25"C, while the H2 reduction of NinY itself occurs only at and beyond200°C : the hydrogen molecule is more readily activated on NiI than on Nin.CARBON MONOXIDE co-ordinates much more strongly with surface NiI than withNin : adsorbed CO can be pumped off NirrY at 25°C but from NiI requires at least200°C.Since this desorption requires temperatures at which surface NiI normallydisproportionates it appears that Nil is stabilized by the carbonyl ligand. Thisstrong bonding with CO was expected since CO is a very weak electron donor thatfavours co-ordination with atoms in the zero-valent state or in negatively chargedcomplexes.The C-0 stretching frequency on Nin in NinY is 2217 cm-1, while on NiY,presumably on the NiI ions, it is 2188 cm-1. This suggests that the C-0 bond isweaker in NiICO than in NinCO. This is reasonable because NiI is expected to bea stronger electron donor than Nin, any electrons donated to the carbonyl groupgoing into antibonding orbitals.OXYGEN is very weakly adsorbed on or near surface NiI at 25°C-the reversibleloss of signal is presumably only a magnetic effect. At the same time, there is anunexplained stronger adsorption of oxygen somewhere else in the structure.At1OO"C, 0; ions are formed :Nil(SII)+ O2+[O;Ni"(SII)].To remove the 0; requires pumping at 200-300°C, and at this temperature the bareNiI ions disproportionate to Nin and Ni". The [O,NiL(Sn)] complex is evidentlymore stable than NiI(Sn) itself.HIDDEN NiI IONSThe sharp increase in the hidden-NiI concentration on heat treatment was un-expected. We now understand that two processes concur during the reduction ofNin (5 %)Y : reduction of hidden Ni* ions, and reaction between surface sodiumions and sodium atoms to form the Nai+ complexes.This is reasonable becausewith as low as 5 % cation exchange most of the zeolite cavities are still filled withsodium ions capable of forming the Na: + complex. This assumption is confirmedby recent experiments in which Ni (5 %)Y on reduction first turned pale red andshowed the characteristic spectra of the Nai + centres. The paramagnetic electronof the Nazf centre so formed evidently then transfers at high temperatures tohidden Nin by the reaction Na:++Nin(SI)+4Na++NiI(SI), which results in a 5-to 10-fold increase in the concentration of hidden NiI.The excellent yields of hidden NiI ions (- 50 %) are due to the structural pro-tection against disproportionation, which will also explain the enormous differencein heat stability between the hidden and the surface ions (-460'C against 100°C).Since hidden NiI at SI is octahedrally co-ordinated with six oxide ions and rathertightly surrounded by other ions besides, it cannot co-ordinate in classical fashionwith any other ligand.It does react with oxygen at and above lOO"C, however,giving up an electron to form superoxide ion : the 0; signal appears and the NiIsignal vanishes. On heating at 460°C in vacuum the reverse occurs, the NiI signa348 STUDIES OF ZEOLITE CATIONSreappearing and the 0; signal vanishing. Hidden NiI should not react in this waywith NH3, GO, or H2, at least to the extent that these molecules are not known toform chemically significant negative ions.It is also reasonable that this processrequires high temperatures : an electron has to penetrate a structure filled with oxideions. The reversibility of the reaction is evidently due to a fairly even balancebetween the effective ionization potentials of NiI(S1) and O;(SII) and the absenceof any competing processes. This is an interesting reaction, possibly the first ofits kind to be observed on a macroscopic scale.CONCLUSIONThe experimental results and arguments put forward in this paper indicate thatthe carboniogenic catalytic activity, cation-specific carbon monoxide adsorption,reduction of Nin ions, and formation of the Na:+ centres are all related to the largeelectrostatic field in the cavities or to the corresponding large electron affinity ofthe surface cations.Changing the field by changing the size of the bivalent cationsresults in the expected change in catalytic activity. Similarly, the combinationconstant for carbon monoxide follows the changes in field for the alkaline earthand zinc cations in both X and Y zeolites. However, the simple electrostaticargument does not explain the enormous difference in the combination constantbetween alkaline-earth and transition-metal cations. Furthermore, there is asuggestion that the simple electrostatic concept upon which we attempted to generalizethe structural distribution of bivalent cations, which has so far been established byX-ray diffraction only for calcium ions, may break down with transition-metalcations: in any case, fewer Con and Nin surface ions are implied by the CO-adsorption results than were expected.The large differences in combining constantbetween Nin and Con and the closed shell ions are probably specific effects of theiropen-shell structure. With the help of the significantly higher affinity of bivalentsurface cations as compared to oxide ions, carbon monoxide adsorption emergeshere as a quantitative method of measuring the fraction of cations at surface posi-tions and to some degree determining their co-ordination.The reduction of Ni* ions and the stability of the corresponding NiI ions canbe explained by arguments based on structure and electrostatic theory. But thechemical activity of NiI, particularly the highly increased bond energy with carbonmonoxide, shows that the properties of the open-shell electronic structure are ofgreat importance also.The Nap + centres bear similarities to the colour centres of solid-state physics.Their remarkable stability probably results, in first approximation, from the largermagnitude of the concerted electron affinities of sets of surface cations in a cavity,as compared to the electron affinity of a single cation. But full discussion of thecombination of cation orbitals occupied by the delocalized electron and of everysignificant contribution to the ionic energy sum seems to be required for a betterunderstanding of why and with just what effect these Molecular Sieves take upsodium vapour. The formation of simple, symmetrical Naz + centres in X zeolite wassurprising but has been rationalized satisfactorily except for a difficulty of structuralconception that remains for both the X and Y materials, namely, the lack of a plausibleassignment of sites for all the sodium ions,The authors express their appreciation to Dr. E. Dempsey for his permission topublish the data in table 1 and to Mr. Gary Skeels, Mrs. Maria Howell, and Mr.Paul Schaffer for their skilful and creative assistance in carrying out the experiments3. A. RABO, C. L. ANGELL, P. H. KASAI AND V. SCHOMAKER 3491 Pickert, Rabo, Deinpsey and Schomaker, Proc. 3rd Int. Cungr. Catalysis, 1965, p. 728.2 Breck, Abstr., 134th Meeting of the Amer. Chem. SOC., Chicago, Sept., 1958.3 Rabo, Pickert, Stamires and Boyle, Actes Duexierne Congr. Int. Catalyse, 1960, p. 2055.4 Broussard and Schoemaker, J. Amer. Chem. Sac., 1960, 82, 1091.5 Dodge, R. P., unpublished research at Union Carbide Res. Inst.5a Breck and Flanigen, unpublished research, Union Carbide Corp., Linde Div., Tonawanda Lab.6 Angel1 and Schaffer, J. Physic. Chern., in press.7 Neddenriep, R. J., unpublished research, Union Carbide Corp., Linde Div., Tonawanda Lab.8 Kruerke, U. and Rabo, J. A., unpublished research at Union Carbide Res. Inst.9 Angel1 and Schaffer, J. Physic. Chem., 1965, 69, 3463.10 Bertsch and Habgood, J. Physic. Chem., 1963, 67, 1621.11 Carter, Lucchesi and Yates, J . Physic. Chem., 1964, 68, 1385.1 ZEmmett, Adv. Catalysis (Academic Press), 1957, 9, 645.13 Castor, U.S. Patent no. 3,013,986.14 Breck, U.S. Patent no. 3,013,992.15Yates, J . Physic. Chem., 1965, 69, 1676.16 Rabo, Pickert and Schomaker, Pruc. 3rd Int. Congr. Catalysis, 1965, p. 1264.17 Kasai, J. Chem. Physics, 1965, 43, 3322.b Breck, U.S. Patent 3,130,007, (1964)

ISSN:0366-9033    

DOI:10.1039/DF9664100328

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Oxygen Chemisorption and the Catalysis of N20Decomposition on NiO-MgO and Related Solid SolutionsBY A. CIMINO,* M. SCHIAVELLO AND F. S. STONE $Received 3rd January, 1966By preparing at 1200°C solid solutions of formula NixMgl,&, where ions of stable valency(Mg2f) are interposed between nickel ions without change in structure, it has been possible to studyhow the chemisorption properties and the catalytic activity of the transition metal ions are influencedby isolation. Solid solutions of the type NixLiYMg~-,O have also been investigated.Chemisorption of oxygen has been studied from -78 to 550°C and catalysis of N20 decom-position from 300 to 500°C. The activity per nickel ion is found to increase as the ions are dilutedby magnesium ions, reaching a maximum at about 1 % nickel.Chemisorption occurs preferentiallyon nickel ions, and activity in the catalysis is paralleled by the occurrence in chemisorption studiesof a reversible type of oxygen adsorption which develops as nickel ions are diluted.The drift of charge across an array of Ni2+/Ni3+ ions, which gives rise to p-type semiconductivityin NiO, is considered to be much less relevant for the catalysis of N20 decomposition than thenature of the octahedral complex formed at the surface by chemisorbed oxygen. Several forms ofoxygen chemisorption are discussed, including a strong form which occurs on lithium-containingspecimens and inhibits the catalysis.Many of the researches on catalysis by metal oxides published since the 1950Faraday Society Discussion on heterogeneous catalysis have attempted to define arelationship between semiconductivity and catalysis.The theme was of sufficientgenerality and elegance to demand thorough investigation, and in the last 15 yearsabout a hundred papers prompted by this topic have appeared on zinc oxide andnickel oxide alone. ZnO was a good choice as an archetype, and the concept of freecarriers and donors, the latter mobile at high temperatures, continues to assist thedevelopment of ideas on the role of the chemisorbed state in catalysis and photo-catalysis on this oxide. Nickel oxide, on the other hand, has not yielded correspond-ing dividends. In the first place the simple band theory is not a good approximation,and secondly the chemistry of its lattice imperfections is still incompletely understood.The link between semiconductivity and catalysis is consequently much less direct andcurrently of limited heuristic value.This is unfortunate, since the quantitativeapproach provided by band theory cannot be matched at present by theories based onHeitler-London and crystal field approximations. It is possible to set up localizedmodels of co-ordination and charge transfer (or covalent ligancy) at oxide surfaces,as Dowden 1 and others 2 have done, but this still leaves the formidable problemof estimating the relative contributions from Coulombic, polarization and crystalfield stabilization energies, as recently pointed out by Dickens and Sutcliffe.3 Thereis a need to devise experiments in chemisorption and catalysis on oxides which arespecifically amenable to discussion in terms of the localized viewpoint.This can beachieved by investigating systems in which electronic and magnetic interaction is* Istituto di Chimica Generale, UniversitA di Palermo ; (present address : Tstituto di Chimicat Department of Physical Chemistry, University of Bristol ; (present address : Istituto di Chimica$ Department of Physical Chemistry, University of Bristol.Generale, Universita di Perugia).Generale, Universita di Palermo).3 5A. CIMINO, M. SCHIAVELLO AND F. S. STONE 351limited, as when true solid solutions are formed between the transition metal oxideand an isomorphous, weakly active matrix which is insulating and diamagnetic.4~ 5In order to throw light on the role of the chemisorbed species in catalysis, combinedstudies of adsorption and catalysis are required.In the present research we haveused magnesium oxide as a matrix, and distributed Ni2+ and Ni3f ions by formingat high temperature specimens of formula Ni,Mgl-,O and Ni,Li,Mgl-,-,O, with xbetween 0.001 and 0.1. N20 decomposition has been chosen as the catalytic reaction,partly in view of its relative simplicity but also because of the general interest inchemisorbed oxygen. Nitrogen is not chemisorbed, so the relevant chemisorbedstates involved are those of nitrous oxide and oxygen. Oxygen is often consideredto take different chemisorbed forms depending on the temperature and on the solidsubstrate, and one would expect the catalysis of N20 decomposition to be directlyaffected.We shall show that this is so. Chemisorption of oxygen has been measuredfrom -78 to 550°C on the solids of greatest catalytic interest, thereby spanningthe full range (300 to 500°C) studied in the heterogeneous catalysis. Chemisorptionreceives the greater attention in the experimental part of this paper as some of thecatalytic results are already in the course of publication elsewhere.5EXPERIMENTALSolid solutions Ni,Mgl,O and Ni,Li,Mgl,,O have been prepared 5 by impregnationof magnesium oxide with solutions of nickel nitrate and lithium carbonate, followed by afirst heating in air at 600°C and, after re-grinding, a second heating at 1200°C. Specimensprepared in this way have been examined by X-ray powder diffraction, optical reflectancespectroscopy, magnetic susceptibility and e.s.r.Surface areas have been determined byB.E.T. measurements with nitrogen or krypton at 77°K. The nickel and lithium contentsquoted are nominal ones, i.e., those based on the amounts of nickel and lithium originallyadded to the magnesium oxide during impregnation. Subsequent analysis confirmed thatthere was no loss of nickel on firing. Lithium, however, was lost. For instance, analysisby flame photometer showed that a specimen of Ni,Li,Mgl,,O with nickel and lithiumnominal concentrations of 1 % lost 95 % of its lithium on firing at 1200°C. In spite of suchlosses, the remaining small lithium contents exert profound effects on the properties of theNiO-MgO solid solutions, as will be seen later.The presence of Ni3+ ions in firedNi,Li,Mg~,,O specimens has been confirmed by e.s.r. studies.6RESULTSA. CHEMISORPTION OF OXYGENThe chemisorption of oxygen has been studied by the conventional manometricmethod using a fixed volume system suitable for pressures from 0.005 to 0-15 torr(Pirani gauge) and 0.15 to 0.6 torr (McLeod gauge). The specimen was containedin a silica bulb and after installation was pre-treated by outgassing for many hoursat 850°C and 10-6 torr. Thereafter a succession of adsorption experiments wasperformed on the given specimen, a standard outgassing procedure of 1 h at 850"being adopted between the runs. After a few such runs, reproducibility was attained,and results were then accumulated by conducting individual experiments at differentpressures and temperatures.The normal procedure with each catalyst was to scanthe full temperature range at 0.4 torr and then to complete a set of 16 individualexperiments at -78, 0, 200 and 500°C (four pressures between 0.05 and 0.5 torr foreach temperature). In addition, tests for reversibility of the chemisorption weremade at each temperature by observing the amount of gas which could be re-adsorbedafter evacuation. Evacuation was carried out at the same temperature and for thesame time as employed in the adsorption. Kinetics of chemisorption were no352 N20 DECOMPOSITION ON NiO-MgOstudied in detail ; the adsorption process, however, was normally rapid and unlessotherwise stated the ordinates on adsorption isobars refer to amounts adsorbedafter 30 min.The four solids selected for study as having the greatest significance for the catalyticwork described in 5 C were (i) pure magnesium oxide (MO), (ii) Ni,Mgl-,O withx-0.01 (MN l), (iii) Ni,Mgl-,O with x-0.1 (MN 10) and (iv) Ni,Li,Mgl-,,Owith x-0.01 and y (nominal) -0.01 (MNL 1 : l).* Their surface areas were 44, 12,12 and 0.82mz/g respectively, the pure MgO having also been heated to 1200°C.Coverages are expressed on the basis of a monolayer value for oxygen of 0.20 cm3 m-2.This coverage scale can be regarded as arbitrary, but it does, in fact, correspond toone in which the monolayer is regarded as one oxygen atom adsorbed on each cationof a surface made up of equal areas of (loo), (1 101 and ( 1 1 1) planes.7 All the solidspossess the same crystal structure and the lattice parameter varies by less than 1 %.-I-_ L___1-__1 1 1 I0 2 0 0 400 600temp., "CFIG.l . 4 x y g e n adsorption isobars at p 4 . 4 torr on magnesium oxide (MO) and on Ni,Mgl-,Osolid solutions (MN 1 and MN 10).Adsorption isobars constructed from experiments at 0.4 torr are shown in fig. 1for MO, MN 1 and MN 10. It is evident that at 0°C and above the solid solutionsare more active than pure magnesium oxide, confirming that the transition metal ionpossesses greater specific activity in oxygen chemisorption than the magnesium ion.The same result is observed also at the other pressures studied. The most interestingresult, however, is that below 400°C the solid solution containing 1 % Niz+ (MN 1)is more active than that containing 10 % Ni2+ (MN lo).The marked rise of the* The actual concentrations of the solid solutions are based on a scale referred to 100 magnesiumatoms ; for example, MN 10 is defined as a solid solution with a ratio M g : Ni = 100 : 10 (i.e.,x = 0.0909 in the formula NixMgl-xO) and MNL 1 : 1 a solution containing a nominal ratioMg:Ni:Li= 1OO:l:lA. CIMINO, M. SCHIAVELLO AND F. S. STONE 353chemisorption on MN 10 above 250°C should also be noted. The isobar for MN 1has been extended to 700°C in other experiments, but no similar point of inflexion hasbeen observed for this solid solution.Fig. 2 shows adsorption isotherms at ca. 200 and 500"C, the first temperaturebeing below the onset of the new trend for MN 10 and the second above it.It isapparent that most of the increased chemisorption on MN 10 at 500" is achievedalready at the lowest pressures, typical of a strong adsorption. The different pressureM N 10: 200"/---- -_ + - -0--MO: 204'I I I- I0. I 0-2 0.3 0.4pressure, tonFIG. 2.--Oxygen adsorption isotherms at 500" and -200°C for magnesium oxide (MO) andNixMgl-,O solid solutions (MN 1 and MN 10).dependence at 200" on MN 1 and MN 10, whose isotherms are seen to diverge appreci-ably, points to the fact that although the coverage is larger on MN 1, the chemisorptionis weaker. The apparently greater amount of adsorption on MO at -78" comparedto MN 10 (see fig. 1) is not significant, as the isotherms for this temperature run veryclose together.The existence of a weak form of oxygen adsorption at low temperatures is indicatedalready in fig.1 by the minimum in the adsorption isobar for pure magnesium oxideat ca. 50°C. The studies of reversibility amply confirm this observation (table 1)and show, furthermore, that the phenomenon is general. Thus, all the oxygenchemisorbed at 0" can be desorbed by pumping for 30min at this temperature.At 200" and above, however, it is quite otherwise. Each of the three oxides developscharacteristics of its own. With MgO there is a transition from complete reversibilityat 0" to only one-fifth reversibility at 200" : it is clear that at least half the gas chemi-sorbed at 0" becomes converted to a strongzr form at 200".More significantly,however, it follows that in the neighbourhood of 200" the specific adsorption on Ni2fions (i.e., the excess over that shown by the MgO matrix) is primarily in the form ofreuersibly-held oxygen. This is most clearly seen by comparing figures for MO andMN 1 in table 1. The excess of coverage shown by MN 1 at 200" and 0.4 torr is 0.29units; of this excess, 0.21 units is represented by reversibly-held gas. Finally, by500" the high-temperature chemisorption of MN 10 reveals itself by a lower absolutecoverage of reversibly-held gas as well as by the higher total coverage already notedin fig. 1. It is obviously a very strong form of chemisorption354 N 2 0 DECOMPOSITION ON NiO-MgOMeasurements of the rates of adsorption provide further information. On MN 10the adsorption of oxygen at -78, 0 and 200°C was virtually complete in 5 min ;at 500°, however, the rate was much slower than this (fig.3). The new chemisorptionprocess above 200" is therefore a slow process, and as such is compatible with itsbeing a strong adsorption. On MN 1 the rates at all temperatures were comparableto those at the lower temperatures on MN 10. On MO, on the other hand, the ratesof adsorption were faster at all temperatures than on the solid solutions. We con-clude that the extensive specific adsorption on Ni2+ ions shown by MN 1, althoughpredominantly a reversible adsorption, is nevertheless involving a greater activationenergy than0time, minFIG. 3.-Kinetics of oxygen adsorption at 500" and -200°C for p-0.4 torr on NixMgl-& solidsolutions (MN 1 and MN 10).It remains to present the results on MNL 1 : 1.The isobar at 0.44 torr is shownin fig. 4, together with that for MN 1 at 0.40 torr. The main observation is that theoxygen chemisorption coverage on MNL 1 : 1 is an order of magnitude greater thanon MN 1 or MN 10. It is also a more rapid adsorption, and at any given temperaturethe proportion of gas reversibly held is lower, falling to 8 % at 500°C compared with43 % on MN 1. All these results point to a much stronger chemisorption of oxygenthan on the Ni,Mgl-,O solid solutions.TABLE 1 .-OXYGEN COVERAGES DETERMINED IN CHEMISORPTION EXPERIhfEl"stotal coverage (%) at 0.4 torr(reversible coverage in parenthesis)catalyst - 7s0c 0°C ca.200'C 5oc"CMO 0.20 (0-20) 0.13 (0.13) 0.28 (0.06) 0.420.65 (0.28) MN 1 0.32 (0.32) 0.49 (0.49) 0.57 (0.27)MN 10 0.14 (0-14) 0.24 (0.23) 0.37 (0.18) 0.96 (0-25A. CIMINO, M. SCHIAVELLO AND IF. S. STONE 355B. CHEMISORPTION OF NITROUS OXIDEThe chemisorption of N20 was investigated on the same four specimens as usedfor the oxygen studies. Experiments above 50°C were not practicable on account ofthe small amount of adsorption and (except for MO) the onset of decomposition.2*oL__._1, O d 200 400 600temp., "CFIG. 4.-Oxygen adsorption isobar at p = 0.44 torr on NixLi,,Mg~-,,O(MNL 1 : l), with MN 1from fig. 1 for comparison.0.2 s!& E!?8 0-10. I 0 . 2 0 - 3 0-4 0 . 5 0pressure, torrFIG. 5.-N20 adsorption isotherm at 0 and 40°C on magnesium oxide (MO) and on Ni,Mg,_,Osolid solutions (MN 1 and MN 10).Taking the monolayer for N20 to be 0.41 cm3 m-2 (non-dissociative adsorption oncations), the coverages at 40" and 0.4 torr were about 0.1 5 % on both MO and MN 1356 N20 DECOMPOSITION ON NiO-MgObut only 0.05 % on MN 10.On MNL 1 : 1 the corresponding coverage was 0.8 %.The adsorption was reversible and strongly pressure-dependent. Isotherms at 0 and40" are shown in fig. 5, where it is seen that the adsorbed amounts decrease withincreasing temperature. The decrease, however, is not as great as would be expectedon the basis of additional isotherms measured in the region of predominantly physicaladsorption between - 80 and - 40°, where the heat of adsorption was approximately5 kcallmole.We infer from this that some chemisorption is occurring at tempera-tures above 0" and that the isotherms at 0 and 40" probably do not correspond totrue adsorption equilibrium. In cases of chemisorption the ability to desorb readilycannot be adduced as evidence that there is thermodynamic equilibrium ; for instance,we have seen in the oxygen measurements that the chemisorption on MN 1 and MN 10at 0" was reversible, but the adsorbed amounts were actually increasing with increasingtemperature.The evidence for chemisorption of N2O at 0" is supported indirectly by the factthat catalytic decomposition in the adsorbed film could be detected on MN 1, and onstanding at 40' the decomposition on MN 1, MN 10 and MNL 1 : 1 was appreciable.By contrast, no decomposition was observed on pure magnesium oxide.c.CATALYSIS OF N20 DECOMPOSITIONThe catalytic properties of MgO, Ni,Mgl-,O and Ni,Li,Mgl-,-,O in N20decomposition have been studied in detail at 300-500°C. Results obtained on anumber of specimens, which gave a first appraisal of the behaviour of these systems,are given elsewhere,s together with full particulars of the experimental procedure.A wider range of catalysts has since been investigated, and kinetics have also beenexamined further. The detail will appear in a separate publication.8 In this sectionwe give only a summary of the catalytic results, sufficient to integrate the presentationand to illustrate the relevance of the chemisorption data in understanding the catalyticbehaviour .The decomposition has been studied in a static system at approximately 60 torr.The reaction was followed by sampling, freezing out unreacted N20 and measuringthe amount of gaseous products manometrically or mass spectrometrically.Kineticstudies have shown that poisoning is a common occurrence when nickel ions arepresent. For this reason, comparisons of activity between the various catalysts havebeen primarily based on measurements of initial rates (< 1 % decomposition).In addition to pure MgO, catalytic studies have been made on solid solutionsNi,Mgl-,O with Ni : Mg ratios equal to 0.1, 0.01, 0.005 and 0.0015, accordinglydesignated MN 10, MN 1, MN 0.5 and MN 0-15 (i.e., x-0.1, 0.01, 0.005 and 0.0015respectively). Two solid solutions Ni,LiyMg~-,--,O corresponding to the designationsMNL 1 : 1 and MNL 10 : 1, i.e., x-0.01, y(nomina1) -0.01 and x-0.1, y(nomina1)-0.01 respectively, have also been investigated.The surface areas of all MO andMN catalysts were between 10 and 20 m2/g ; those of MNL were much lower.Fig. 6 summarizes in the form of Arrhenius plots the specific activities of the sevendifferent catalysts, the velocity constants referring to unit surface area. A number ofimportant conclusions can be drawn : (i) the least active catalyst is pure magnesiumoxide, (ii) the addition of nickel ions in solid solution at concentrations of 0.5, 1 and10 % increases the velocity constant at 400°C by an order of magnitude, and decreasesthe activation energy, (iii) the most active catalyst is not MN 10, but MN 1, and(iv) the incorporation of small amounts of lithium in the solid solutions markedlyincreases the activation energy E.The catalysts of greatest interest as representingextremes of behaviour are MO (least active and highest E), MN 1 (most active andlowest E), MN 10 (lowest activity per Niz+ ion of MN series) and MNL 1 : 1 (highesA. CIMINO, M. SCHIAVELLO A N D F. S. STONE 3 57E among the solid solutions). These were the four types selected for the chemisorptionstudies described in Q A and B.Mass spectrometric analysis of samples of the gaseous products during the first1 % of decomposition revealed N2/02 ratios greater than 2.00. This shows thatoxygen is being held on the surface. From the magnitude of the N2/02 ratio, andMNL 1:1MN 0.15L I I 11.4 1-6 I- 8( l / T ) x 10;FIG. 6.-Arrhenius plots for the specific velocity constants of N20 decomposition on magnesiumoxide (MO) and on NixMg1-,O and NixLiyMgl-x-yO solid solutions.knowing the number of N20 molecules admitted, the volume of the vessel and sampler,the area of the catalyst and the pressure of incondensable gas in the sample, it ispossible to estimate the coverage of adsorbed oxygen at the reaction temperatureduring CdtdlySiS.The results are shown in table 2. It may be seen that the coveragesat ca. 400°C are in the order MO< MN 1 <MN 10< MNL 1 : 1. The direct chemi-sorption experiments (§A), which have much higher accuracy, were carried out atoxygen pressures from 0-4 to 0.05 torr. The amounts of decomposition in the experi-ments of table 2 vary between 1 and 0.15 % ; these amounts at 60 torr (the initialN20 pressure during catalysis) correspond to partial pressures of oxygen of 0-3 and0.05 torr respectively, showing the significance of the pressure range selected for theexperiments of §A.A comparison may be made of the coverages determined bythe two methods by reference to fig. 1, 2 and 4, with appropriate interpolation. It isevident that not only is the sequence of coverages on the four solids the same as thatTABLE 2.-oXYGEN COVERAGES DETERMINED DURING CATALYSIScatalyst T % decomposition ratio N2/02 coverage (%I*MO 400 0.29 2-32 0.45MN 1 400 0.17 3.06 0.56MN 10 400 0.89 2.25 0.77MNL 1 : 1 450 0.27 2.16 3.0* see text for definition of coverage.1358 N20 DECOMPOSITION ON NiO-MgOobserved in the direct chemisorption experiments at high temperatures, but also themagnitudes are in satisfactory agreement.This is an important check, for it confirmsthat we are justified in applying the chemisorption data of § A in discussions of thechemisorbed situation during the catalysis.DISCUSSIONCOMPARISON WITH NICKEL OXIDEThe results prove conclusively that nickel ions isolated in a matrix of magnesiumoxide are specifically active in chemisorption and catalysis. Moreover, the activityper nickel ion is greater when present at dilutions of 0.5 or I % than at 10 %. Thecoverage of chemisorbed oxygen which is specific to the nickel ions of the surface isreadily seen from fig.1 and table 1. Assuming that the magnesium oxide matrixbehaves identically in MO and MN specimens, the excess coverage on MN 1 at 0°Cand 0.4 torr is 0.36% (0.23 % at 500"). For MN 10 the corresponding figures are0.11 and 0.54 %. For a random solution in which every Nizf ion in the surface isassumed to be active, the maximum excess coverage that could be expected is 1 % onMN 1 and 10 % on MN 10. Thus the nickel ion couerage on MN 1 is 36 % at 0" and23 % at 500", and on MN 10 it is 1 % at 0" and 5 % at 500".Typical literature values of nickel ion coverage on pure nickel oxide of comparablesurface area range from 1 to 15 %. For example, there are reported coverages at0.06 torr of 1 % at 20°, 10 % at 250" and 6 % at 450" ;2 at 0.03 torr, 15 % at 353" ;9at 0.1 torr, 15 % at 260O.10 On this reckoning, therefore, MN 1 is considerably moreactive per nickel ion than NiO itself.The respective behaviour of MN 1 and MN 10confirms the trend, and this in fact is a more valid comparison than that with NiOsince the two solid solutions had precisely the same specific area and pre-treatment.It is possible that the lower figures for coverage on MN 10 and NiO may reflect agreater difficulty in cleaning the surface of chemisorbed oxygen at the start of theexperiment : the stronger the chemisorption on an oxide the more likely it is that theprocess studied (and the coverage estimated) in a chemisorption experiment is one inwhich the residue of a chemisorbed layer is re-filled. This, however, would merelyendorse the view that oxygen chemisorption on MN if) and NiO is stronger (and lessreversible) than on MN 1.The difficulty of comparison with NiQ is acute when we come to discuss the catalyticactivities.Oxygen poisoning of NiO is so prevalent that little or no agreementexists about the specific activity or activation energy for N20 decomposition on NiO.Our own values for the activation energy, measured under conditions comparableto those in which the MN series of catalysts were studied, ranged from 23 to 36 kcal/m0le.5 Taken together with our results on MN 10 and MN 1, this again suggeststo us that activity increases with dilution in the range from 1-00 to 0.01 in Ni,Mgl-,O.Quite irrespective of the activity inCredSe, however, is the conclusion made possibleby our results that a well-developed p-type semiconductivity such as NiQ exhibits isnot necessary in order for the nickel ion to exhibit high activity in chemisorption andcatalysis in this system.In other words, the conclusion may be drawn that thecatalyst does not function well in this reaction simply because of its high concentrationof positive h Aes, but because other parameters contribute to the characteristics ofthe adsorbed complexes.CHEMISORBED STATE OF OXYGEN ON Ni,Mgl-,O AND Ni,Li,Mgl-,-,OBelow 50°C a type of oxygen chemisorption exists which is not specific to nickelor lithium ions, since it occurs also on the MgO matrix. It is a weak adsorptioA. CIMLNO, M. SCHIAVELLO AND F. S. STONE 359and not relevant to N20 decomposition.We consider that it is probably to beidentified with the molecular adsorption reported on similar insulator oxides by e.s.r.studies.11In the range of temperature from -78 to 400°C a specific adsorption developson nickel ions (fig. 1). It is an activated process, but is reversible at the temperatureof measurement. As such it demands our attention as a possibly significant catalyticintermediate. The appreciation of its catalytic importance, however, is made clearby examination of the properties of MN 1. MN 1 chemisorbs more oxygen at 200"than MN 10 and it preserves its content of reversible oxygen even to 500" (table 1).Parallel with this result is the observation that over this range of temperature MN 1is the more active catalyst in N20 decomposition (fig.6). We consider that thiscorrespondence of properties, which a priori is unexpected on account of the muchsmaller nickel content of MN 1, is sufficiently close for us to be able to attribute thehigh activity of MN 1 to the ability of its nickel ions to produce chemisorbed oxygenin this particular form. We return later in the discussion to consider the chemicalnature of this adsorbed state, since some of our other observations eliminate certainpossibilities.Let us consider the next highest range of temperature, above 400°C. This regionis characterized by the rise of adsorption on MN 10. Charge transfer is certain atthese temperatures, and we suggest that the activated process involved is the formationof 0 2 - ions accompanied by the transfer of electrons from centres deep in the bulk(thus distributing the excess charge and minimizing the repulsion energy).Suchtransfers presumably occur readily in NiO itself, as witnessed by the well-knownconductivity increases on chemisorption of oxygen. The more dilute the Ni2+02-Niz+array becomes due to interposing the Mgz+ ions the more difficult is the electrontransfer, not so much because the chains of adjacent transition metal ions are brokenas the fact that the lattice is becoming more ionic and the electron bands are narrowing.Such an electron transfer process would be much easier on MN 10 than MN 1, andthe high-temperature chemisorption of oxygen in the form of 0 2 - will be corres-pondingly more prevalent.The 0 2 - chemisorption is to be regarded as a poison forthe catalysis of N20 decomposition, and the reversible chemisorption of the previousparagraph can be looked upon as its precursor.The strongest-held chemisorbed state of oxygen that we have measured occurs onthe lithium-containing solutions, and it is prevalent over the whole temperature range.As such it can be expected to poison the surface towards the weak chemisorptions,and these conclusions are fully confirmed by the catalytic observations. The activa-tion energy rises from 18.5 kcal/mole on MN 1 to 30 kcal/mole on MNL 1 : 1. Theincorporation of lithium in solid solution in pure NiO is well-known to lead to astronger form of oxygen chemisorption than on NiO itself; comparisons havingbeen made by several investigators.9~ 10, 12 Conversion of Ni2+ to Ni3+ bycharge compensation from Lif in the MgO matrix could lead to a stronger Coulombicattraction between surface cations and chemisorbed oxygen.In addition, thefluxing power of lithium must be taken into consideration. The fact that the surfacearea is markedly decreased shows that mobility of both cations and anions is increased,and an accumulation of nickel ions, assisted by the attraction of lithium ions for Ni3+,could readily occur in the surface layers. Lithium ions do not easily enter the latticeof MgO, and this provides an additional reason for expecting abnormally high surfaceconcentrations of nickel ions in MNL 1 : 1 . We see this accumulation of nickel ionsas part of the explanation of the large magnitude of the oxygen uptake on MNL 1 : 1,and its trend above 300°C (cf.MN 10). However, there are other possibilities. Duringpreparation the entry of Lif ions into the surface layers will be accompanied by th360 N2O DECOMPOSITION ON NiO-MgOincorporation of oxygen ions to preserve electrical neutrality. Outgassing in vacuumat high temperatures will produce surface anion vacancies to be filled during thesubsequent chemisorption experiment. The ions which enter the surface anionvacancies may be either 0 2 - or O,, the latter being suggested by analogy with lithiumperoxide.ACTIVITY OF THE MgO MATRIXNelson, Tench and Wilkinson 13 have shown by e.s.r. studies that oxygen adsorp-tion occurs slowly on MgO above 250°C with simultaneous oxidation of transitionmetal impurities in the bulk ; electrons are presumably transferred to chemisorbedoxygen.The process which we have observed on pure MgO is too rapid to be identi-fied with this sort of electron transfer ; furthermore, the impurity concentration inour spectroscopically standardized MgO (c 5 p.p.m.) is about 30 times too small toaccount for the observed coverage. The amounts of oxygen which are chemisorbedon the MgO specimen agree with those reported by Houghton and Winter 14 in thecourse of their oxygen exchange studies. Charge transfer of the type Mg2+02-+Mg+O- is at least 5 eV endothermic in MgO, but the reaction Mg+0-+302-+O-MgZ+O- would be strongly exothermic. An overall adsorption process of the kind0-I-Mg2+-02--Mg2+-02----Mg2+-0--Mg2+-02--i I I I I 1 1 Ican be envisaged, the initial charge transfer step being favoured in the surface,perhaps on particular planes.Such an adsorption could account for the oxygenexchange. 14THE ACTIVE CENTREThe specific activity of MN specimens in the N20 decomposition begins to recedebelow MN0.5 (fig. 6). MN0.1 shows mainly the characteristics of the pure MgOsince the activity of the matrix, although weak, predominates over the small numberof intrinsically more active Ni2+ sites. With increasing concentration the activityof Ni2+ ions emerges to control the catalytic behaviour. However, the increasingconcentration also introduces adsorption characteristics, shown by the chemisorptionexperiments, which are detrimental for the catalytic activity.This is the reason forthe decrease in the activity beyond MN 1.The active centres must be capable of chemisorbing atoms of oxygen (as opposedto molecules) and we imagine that they are held as 0- ions. In speculating furtherthe analogue of our proposed chemisorption process on MgO is worth close attention.The energy of the charge transfer Ni2+02--Ni+O- in the parent oxide can be set atabout 3.5 to 4eV from the absorption spectrum of NO. The energetics of theendothermic part are not as unfavourable as on MgO. However, we should alsorecognize the ability of Ni2+ to form Ni3+. Thus a centre exists on MN capable ofallowing adsorption in the form shown, especially on (01 1 } planes, or aA. CIMINO, M.SCHIAVELLO AND F. S. STONE 36 1steps on (001 1 planes, and from which oxygen desorption is likely to be easy. Theproposed mechanism is then one in which two 0- ions, originally formed from NzOmolecules and separately chemisorbed, migrate together over not more than a fewlattice distances to form the adsorbed state indicated in the sketch. There aresufficient of these " edgewise " adsorbed configurations on nickel ions to account forour observed coverages in the chemisorption experiments, and we may also drawattention to the crystal field stabilization change which favours adsorption on them.2The catalysis of N2Q decomposition, viewed in terms of such a model, is seen todepend primarily upon a particular conformation of the chemisorbed oxygen complex.The adsorption properties of the surface towards N20 itself are only significant inso far as they should permit the initial interaction to be in the nature of a chemisorptionrather than a physical collision with the surface site. Our experimental data on theadsorption of N20, although referring only to lower temperatures, support this view.There is little difference between the weak chemisorptions on the MN catalysts, andthe results with MNL 1 : 1 indicate that polarization of the N20 molecule is a probablesource of the adsorption energy.The research described in this paper was a combined investigation carried out inthe Universities of Bristol and Palermo and was supported by a joint award underthe N.A.T.O. Research Grants Programme.The authors also acknowledge with thanks the assistance of Dr. R. I. Bickley,Dr. V. Indovina and Dr. F. Pepe in some of the experiments.1 Dowden, Chemisorption (ed. Garner) (Butterworths, London, 1959, p. 13. Dowden and2 Haber and Stone, Trans. Faraday SOC., 1963, 59, 192.3 Dickens and Sutcliffe, Trans. Faraday SOC., 1964, 60, 1272.4 Vrieland and Selwood, J. Catalysis, 1964, 3, 539.5 Cimino, Bosco, Indovina and Schiavello, J. Catalysis, 1966,5,271,6 Cimino, Cordischi, Porta and Valigi, Ricerca Sci., 1965, 35 (&A), 1153.7 Dell and Stone, Trans. Faraday SOC., 1954, 50, 501.8 Cimino, Indovina, Pepe and Stone, to be published.9 Keier and Kutseva, Izu. Akud. Nauk, Otdel. Khim. Nauk, 1959, p. 797.Wells, Actes IIme Congr. Int. de Catalyse (Technip, Paris, 1961), p. 1499.10 Wang, Huang and Lou, Sci. Sinica, 1965, 14, 319.11 Kasanskii, Nikitina, Pariskii and Kiselev, Dokladji Akad. Nauk, 1963, 151, 369.12 Winter, Disc. Fauaday SOC., 1959, 28, 183.13 Nelson, Tench and Wilkinson, Proc. Brit. Ceram. SOC., 1965, no. 5 , 181.14 Houghton and Winter, J. Chem. Suc., 1954, p. 1509

ISSN:0366-9033    

DOI:10.1039/DF9664100350

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Mechanism of the Catalytic Oxidation of Carbon Monoxideon Zinc OxideBY P. AMIGUES AND S . J. TEICHNERInstitut de Recherches stir la Catalyse, 39 Boulevard du 11 novembre 1918,69 Villeurbanne, FranceReceiced 28th January, 1966Pure zinc oxide is active at 260°C in the catalysis of the oxidation of carbon monoxide. Oxygenappears to be adsorbed in an ionic as well as a non-ionic form, while carbon monoxide and carbondioxide are adsorbed essentially in non-ionic forms. Two surface sites, probably Zn+ ions, arerequired to fix 0 2 and also COZ. Carbon dioxide is not an inhibitor in the reaction but mutualdisplacement of C02 and 0 2 is observed.The kinetic study of the reaction shows that the rate is proportional to the square root of theoxygen pressure and is independent of CO pressure.These results may be explained by assumingthat 0 2 is adsorbed weakly and CO strongly on two different types of sites. As, however, an in-hibiting effect is shown by non-stoichiometric mixtures with an excess of CO and an acceleratingeffect for mixtures with an excess of 0 2 , it is necessary to consider a competitive adsorption of COand 0 2 on one type of site and the reaction with either CO from the gas phase or with CO weaklyadsorbed on the sites of a different type. However, since the concentration of the ionic species0- (ads) practically does not vary on the surface of ZnO, the order + with respect to oxygen is inter-preted as involving the participation in the reaction of non-ionic oxygen 0 (ads). As with NiO,the slowest step of the oxidation of CO on ZnO does not involve the transfer of electrons. Thusno effect of doping ZnO by Li+ or Ga3+ on catalytic activity is expected and none is observed.The purpose of this paper is to show that a kinetic study combined with measure-ments of catalyst conductivity in the presence of reagents or products gives valuableinformation on the mechanism of the reaction.When the slowest step of the reaction mechanism is an electron transfer betweenreaction partners and the solidyl, 2 a consistent argument in favour of the elec-tronic theory of catalysis 3 9 4 may be advanced.Correlations between the electricconductivity and catalytic activity may be investigated mainly in two ways. First,a correlation may exist between the conductivity and the rate constant of the reac-tion or the apparent activation energy of the reaction.4 In these cases the con-ductivity is usually modified by doping the catalyst with altervalent cations.sp6However, many limitations are present when the correlation is based on apparentactivation energy 7 and, as is stressed in this paper, further information on the typeof surface species involved has to be obtained.In the second method, doping isnot involved. The electrical properties of the catalyst are studied during the re-action,* and from the trend of the conductivity change in the course of reaction,conclusions concerning the state of the solid may be proposed.Oxidation of CO on ZnO is a suitable test for such investigations mainly be-cause ZnO is a well studied catalyst.99 10 The two ways of approach are combinedin this work with a view to finding the relationship between the catalytic activityand electric conductivity, both initially and during the reaction.In this way a fairlygood insight into the mechanism of the reaction has been achieved.3 6P. AMIGUES AND S. J. TEICHNER 363EXPEKIMENTALAPPARATUSConductivity measurements and kinetic tests were performed in the same cell as de-scribed previously.10 In this method a powdered catalyst sample is placed in a quartzcyiindcr closed at the bottom by a platinum disc electrode. The upper electrode is moveableand a constant pressure may be applied through it to the sample. An a.c. bridge, Leedsand Northrup model 1554A2, is used for the conductivity measurements.The limitationsof this method have been discussed.lol11 A liquid-nitrogen trap is placed very close to thesample in order to condense CQ2 formed during the reaction. The cell may be placed inthe centre of an electric furnace whose temperature is controlled within 1°C. A fixedamount of catalyst (150 mg, 2.25 m2) was used throughout the tests and the initial pressureof stoichiometric (COf to,) mixtures was 3 torr. The kinetics were followed by pressuremeasurements.MATERIALSPure zinc oxide was prepared directly in the cell by decomposition of pure zinc hydroxidein vacuum (10-6 torr) at 35OOC.12 Lithium-doped zinc oxide (3 at. %) was obtained bydecomposition of Zn(OH)2 containing LiOH. Gallium-doped zinc oxide (3 at.%) wasprepared in the same way. Experimental details of these preparations have been givenpreviously.13RESULTS AND DISCUSSIONCHARACTERIZATlON OF ZnQIf the temperature of the dehydration of Zn(OH)2 in vacuum (10-6 torr) is raisedto 900T the conductivity of the sample during the cooling increases. This is2 2.51000fTFIG. 1.-The vzriation of the conductivity of the zinc oxide in vacuum when the temperature in-creases to 350°C and then decreases.characteristic of a metallic behaviour.10 Metallic zinc is, in fact, condensed on thecool parts of the cell. For this reason the temperature limit for dehydration o364 MECHANISM OF c8 OXIDATIONZn(O€€)z was fixed at 350°C (12 h). Fig. 1 shows an Arrhenius plot of the varia-tion of the conductivity of the sample during slow heating and slow cooling.Theactivation energy of conductivity during cooling is 0.7 kcal/mole. This value isnot modified if the dehydration of Zn(OH)a in vacuum is performed at 450°C. Thecatalytic activity of zinc oxide proceeds with a reasonable rate at 261"C, the tem-perature selected for the kinetic tests. The equilibria involved in the vacuumactivation of the catalyst are :ZnO +Zni + 402 (1)Zni+Zn+ + e - (2)Zn+ +Znj2 + + e - (3)Equilibrium (3) is probably involved to a very small extent at temperatures notexceeding 5OO"C.lOCHEMISORPTION OF OXYGENAt least two equilibria may be envisaged depending on the temperature range.14w 2 + e - +O(&:,402+2e-+0&,.The conductivity of vacuum-activated samples of ZnO decreased in oxygen (3 torr)at temperatures from 25 to 261°C (fig.2). However, after a subsequent evacuationat the same temperature as used for the adsorption, the conductivity did not returnto the value which was recorded after the first activation. This result means thatoxygen adsorption is, in part, irreversible. The reversibility is observed only when__ i50 I00 150 200time, minFIG. 2.-The variation of the conductivity of vacuum-activated zinc oxide exposed to a pressureof oxygen of 3 torr as a function of time ; temperatures : 25, 125, 150, 200 and 261°C.the adsorption and desorption temperatures are above 300°C. Below this tempera-ture only the electrons participating in the reversible chemisorption of oxygen areliberatsd during the desorption.For subsequent discussion it is convenient to introducethe following symbols : [e;], concentration of electrons related to the fractioP. AMIGUES AND S. J. TEICHNER 365of reversibly adsorbed oxygen, [e TI, concentration of electrons remaining free afterchemisorption of oxygen at pressure poZ, when the observed conductivity is 01 :[e,], the concentration of electrons remaining free after evacuation of oxygen whenthe observed conductivity is 00 is[eJ = ka,. ( 5 )Sincethen[ei] = ka,.Lei1 = [e,l - Cell9[e,] = k(a,-a,).The amount of the reversibly adsorbed oxygen in ionic form may depend on PO,.A relationship then exists between c and PO,.The temperature of vacuum activation of the catalyst (350°C) was lowered to261"C, which was the temperature selected for the reaction test, and oxygen wasadmitted at several pressures from 10-6 to 160 torr. Equilibrium readings of con-ductivity for each pressure are shown by curve A in fig.3. By progressively de-creasing the oxygen pressure to 10-6 torr, curve B was registered. The linear portions5 3.5 7 2.5 T 75 0 0.5 I 1'5 2log PO2FIG. 3.-The variation of the conductivity of vacuum-activated zinc oxide as a function of oxygenpressure. Curve A, adsorption ; curve B, desorption.of these curves are close together for pressures higher than 10-1 torr. The slope ofline A is -0-59 and that of line B is -0.5. These values, which are close to -3may be interpreted by assuming that for oxygen pressures higher than 10-1 torr thisgas is reversibly chemisorbed at 261°C as 0- (ads) ions :By the law of mass action,When the oxygen pressure is decreased from 160 to 10-6 torr (curve B) a conductivityelectron is liberated for each 0-(ads) ion desorbed to the gas phase.When allthe oxygen ions, previously adsorbed reversibly, are desorbed at 10-6 torr, theconductivity registered is 00 and the concentration of free electrons, [e,], is given402(g, + e- =O&. (9)[ O - ] = Kp&,[e-]. (10366 MECHANISM OF co OXIDATIONby eqn. (6). For any given intermediate pressure po2 the conductivity is a and theconcentration of free electrons is[e-] = ko. (1 1)The concentration of reversibly adsorbed oxygen ions is then equal to the concentya-tion of electrons liberated when the oxygen pressure decreases from pol to 10-6 torr :[O-] = [e;]= [e;]-[e-] = k(oo-a).(12)The relationship between the conductivity and PO, is established from eqn. (10) to(12) :(00 - O)/O = K&,. (1 3)For high values of PO,, O<OO and eqn. (13) becomeswhich accounts for the value (-3) of the slope of line B in fig. 3. For curve Athe oxygen pressure was raised by steps from 10-6 to 160 torr. The concentrationof irreversibly adsorbed oxygen ions increases with the oxygen pressure. This isshown by the fact that when vacuum is established after some intermediate pressureof oxygen, the value of GO depends on the previous pressure of oxygen. For example :Po2 (torr) 3 1 60log 00 3-50 Z.50log d 6-80 7.1Eqn. (14) then giveslog Q = log (o0lK) - 3 log PO,.(16)As the term oo]K decreases when PO, increases, the dependence of log o on PO, is nolonger linear, at least for low pressures; this is the case of curve A in fig. 3. Onthe other hand, if oxygen is evacuated progressively after the maximum pressure of160 torr has been attained (curve B), the value of 00 is then a minimum becauseit corresponds to a maximum irreversible adsorption of oxygen. The licear relation-ship between log CT and logpo, is therefore observed even in the low-pressure range.These results give support to the idea that at 261°C the reversible fraction of oxygenis adsorbed as 0- ions.15In the above treatment the dissociation of 0 2 molecules by interaction with twosurface sites has been ignored. In view of the subsequent discussion of a competitiveadsorption of 0 2 and COz, it is preferable to consider instead of eqn.(9) the followingequation :If the fraction of oxygen-covered surface is 8 and the fraction of bare surface is1 - 0, the concentration of free sites is proportional to 1 - 8 and the concentrationof 0-(ads) ions is proportional to 6. Assuming a Langmuir adsorption, the ratev1 of oxygen adsorption as 0- isthe rate of desorption u-1 beingO,(g) + 2 sites + 2e- +2O-(ads). (17)Vl = klPO2(1 - 6I2[e-l2, (1 8)U-7 = k-182. (19P. AMIGUES AND S. J. TEICHNER 361where ko- = (k~/k-l)*.ments of oxygen adsorption show that the coverage is sparse, soEqn. (20) may be simplified since the volumetric measure-0 = ko-p&[e-].tnstead of 0 the concentration of adsorbed 0- ions may be considered :[0-] = co.iM oreover?[O-1 = [ e , ] - [ e - ] = k(ao-a).At 261"C, [e-] is always low compared to [e;], so that eqn.(21), (22) and (12)then givewhich is of the same form as eqn. (15).When the pressure of oxygen admitted to the evacuated catalyst at 261°C ishigher than 10-2 torr the conductivity decreases monotonically with time until aQ = o o p ~ ~ / c k o - , (23)i- I-.------...-- -I 4al2cotime, minFIG. 4.-The variation of the conductivity of vacuum-activated zinc oxide as a function of timefor an initial PO, = 7 x 10-3 ton.FIG. 5.-The decrease of the oxygen pressure in the experimental cell as a function of timecorresponding to the conductivity values of fig. 4.stable value of a, as for fig.2, is observed. However, when the initial pressure issmaller'than 10-2 torr (small admission of oxygen to the cell), the conductivityfirst decreases, shows a minimum and then increases (fig. 4), the pressure in the celldecreasing steadily from the initial to the equilibrium value (fig. 5). Such a varia-tion of conductivity has already been observed 1 5 , ~ at temperatures higher than150°C. In our studies this phenomenon was not observed at 25"C, nor at 261°Cwhen the initial pressure was higher than 10-2 torr368 MECHANISM OF co OXIDATIONThe explanation proposed earlier 1S916 for the first decrease and a subsequentincrease of the conductivity involves a high rate of formation for the 0- ions andthen a slow migration of ionized donors from the bulk to the surface where theyreact with 0-, this last step shifting the ionization equilibrium (2) to the right.Wepropose here an explanation based on the fact that, when the initial dose is small(fig. 5), the adsorption proceeds under a decreasing oxygen pressure. If oxygenis adsorbed as more than one surface species whose influences on conductivity aredifferent and whose proportions depend on the pressure of the gas, one species maybe transformed into another when the pressure decreases. This accounts for avariation of the conductivity such as that observed with a minimum in fig. 4.It is assumed that oxygen is reversibly adsorbed at 261°C as an ionized species0- (ads). The following mechanism, which is consistent with the reaction charac-teristics given below, does not involve the diffusion of donors(Zn:) from the bulkto the surface. In fact, this process may be given a low probability considering therelatively low temperature and small coverage of adsorbed oxygen.Let us assumethat the rates of adsorption of 0- (211) and 0 (24 and their rates of desorption (respec-tively u-1 and 21-2) are given by Langmuir-type equations :v1 = klpo2(i-e0- - ~ ~ ) ~ [ e - ] ~ ,v - ~ = k-,O;-,21-2 = k-,e;.The high concentration of free electrons at the beginning of the introduction of theoxygen at a small pressure may give a high value to ul compared to 02. The firstquantities are therefore adsorbed mainly in an ionic form, i.e., with a decrease of con-ductivity (fig. 4). During this adsorptionpo, decreases (fig.5) and so do (1 - 80- - 80)and [e-1. Thus, u1 decreases much more than v2 and for a given value of PO,and of [e-J, 01 becomes equal to n . 1 . At the same time, 212 which decreases moreslowly than 211 is still higher than 21-2. If oxygen were reversibly adsorbed only as the0- species an equilibrium would be established for v1 = D - 1 . The conductivityand PO, would then stabilize at a minimum level. But if oxygen is also adsorbed asthe 0 species it continues to be adsorbed for 212 > u-2 when 211 = 0.-1 and the pressurepo, decreases. Once ul is smaller than u-1 a desorption of 0- occurs followed byits adsorption as 0 :The concentration [e-] and CT increase up to a value for which 211 = 21-1 and v 2 = 21-2.For higher initial pressures of oxygen (larger oxygen doses) the pressure PO, decreasewith time is much smaller and the conductivity does not show any minimum.17Since the oxygen pressure reaches a constant value in a short time, the transitionbetween the two chemisorbed species is not recorded.0-(ads)+O(ads)+ e - .(28)CHEMISORPTION OF CARBON MONOXIDEBelow 50°C the conductivity of the activated catalyst was not modified by CO.Above 50°C the conductivity increased slightly and, above 18O"C, C02 was foundin the gas phase. This shows that an extra reduction of the catalyst occurs inaddition to reaction (1). Fig. 6~ represents final conductivity values in an Arrheniusplot for pco = 3 torr. No significance can be attributed to the activation energyof conductivity because the increase of CT depends not only on the temperature buP.AMIGUES AND S. J. TEICHNER 369also on the amount of CO chemisorbed and the additional reduction of the solid.At 15O"C, where reduction is not observed (no C02 found), the conductivity of thecatalyst increased with pco (fig. 6 ~ ) . The chemisorption of CO, which is an electrondonor, is not purely ionic.18 Thus it is not possible to relate a given number ofelectrons to each adsorbed CO molecule. The slope of the straight line of fig. 6 ~ ,which is equal to 0.14, cannot lead to a discussion comparable to that developedfor oxygen.hEa.AvI 75M3I ' 5FIG. 6~.-Arrhenius plot for the variation of the conductivity of vacuum-activated zinc oxide atpco = 3 torr.FIG.6~.-The variation of the conductivity of vacuum-activated zinc oxide as a function of carbonmonoxide pressure at t = 150°C.As the variation of CT, in the presence of CO, is very small it is probable thatCO adsorption is essentially non-ionic. The formation of CO-+, though often as-sumed,l9 would not be compatible with a high concentration of electrons in anactivated ZnO :CO(g)+CO+(ads)+e-.If the sites involved in the covalent adsorption of CO are the Zn; ions, a secondaryionization according to reaction (2) may explain the small increase of the con-ductivity recorded above 50°C. If the adsorption of CO, below this temperature,takes place simultaneously on Zna and Zn;, equilibrium (2) is not modified and theconductivity does not vary.This hypothesis is consistent with the assumption oftwo types of CO chemisorption on ZnO.20 Finally, the reduction of ZnO by COat 261°C does not develop a metallic behaviour in the catalyst because during coolingthe conductivity decreases.CHEMISORPTION OF CARBON DIOXIDEThe chemisorption of C02 below pressures of 200 torr does not modify theconductivity of the activated catalyst in the temperature range 25-350°C. The re-sults discussed below give evidence of a competitive adsorption of oxygen and carbondioxide on the same sites, i.e., on Zn; ions or Zn,' atoms. This precludes the ad-sorption of C02 on lattice 0 2 - ions with the formation of CO:-.21 The formatio370 MECHANISM OF co OXIDATIONof a CO, ion by the capture of an electron 21 is also precluded as the conductivityis not decreased.When the concentration of conductivity electrons is not affected by chemisorptionof a gas, the absence of an ionic character for the bonding between the species andthe surface may, in general, be concluded.Conductivity measurements do notthen permit one to follow the chemisorption of this type of gas as a function ofpressure, time or temperature, and so no information is gained concerning thenature of chemisorption sites or adsorbed species. However, it is logical to expectthat even if the gas adsorbed first does not influence the electric conductivity, itsinteraction with a second gas which, when adsorbed separately does influence it,may cast some light on the nature of this interaction and on the nature of the sitesinvolved.This approach leads us to consider the influence on the conductivity of thecatalyst of a successive or simultaneous adsorption of the gases CO, 0 2 and C02.INTERACTION OF GASES(A) 0- (ads) + CO (9)The adsorption of 0 2 at 261°C decreased the conductivity of activated ZnO.This is due to the formation of 0- ions (see above). After 15 min of evacuationCO was introduced and an increase of conductivity to the value observed beforethe introduction of 0 2 was recorded, even for CO pressures as low as 10-3 torr.At the same time C02 was found in the liquid-nitrogen trap. This reduction wasalso observed with the stoichiometric ZnO (non-activated in vacuum, c = 10-9ohm-1 cm-1) whose conductivity at 261°C under a pressure of CO of 3 torr wasthe same (a = 10-1 ohm-1 cm-1) as after activation in vacuum at 350°C.Carbondioxide was found again in the trap and does not seem, therefore, to inhibit thereduction by CO. This inhibition, however, is observed at 150°C because thepresence of COZ in the trap was not recorded and the conductivity increase thendepended on the pressure of CO.If it is assumed that oxygen is adsorbed on activated ZnO as 0- and 0, the inter-action with CO can only be explained byCO (9) + 0- (ads) +C02 (in trap) + e-, (29)or CO (9) + 0 (ads) +C02 (in trap). (30)In the second case the increase of conductivity would be due to a shift to the rightin equilibrium (28).(B) 0- (ads) + CO2 (g) AND C02 (ads) + 0 2 (9)The activated sample of ZnO was first submitted at 261°C to an oxygen pressureof 3 torr, then a pressure of 100 torr of C02 was established without evacuation ofthe oxygen.Fig. 7 shows that the first decrease of conductivity due to the adsorp-tion of oxygen was followed by an increase when C02 was introduced. When asteady value of conductivity was recorded, dry ice was replaced by liquid nitrogenin order to condense C02. Only oxygen then remained in the gas phase and theconductivity decreased to the value recorded previously, before the introductionof c02.The action of C02 may be interpreted either by the formation of a neutral com-plex between 0- (ads) and C02, with liberation of electrons, or merely by displace-ment of 0- (ads) due to a competitive adsorption of the CO;! on the same site. Thevalue of Q after condensation of CO2 in the trap is similar to the value of CT afterthe chemisorption of oxygen; this gives an argument in favour of a displacemenP.AMIGUES AND S. J. TEICHNER 37 1of oxygen by COa. If conversely, the formation of a neutral complex is assumed,no noticeable liberation of surface sites would occur and the re-adsorption of oxygenafter condensation of C02 would be, at least partially, prevented. The adsorptionof C02 at 261°C is therefore reversible and essentially covalent.2240M0 CI51 I -..100 200 300time, minFIG. 7.-The variation of the conductivity of vacuuni-activated zinc oxide after the introductionof oxygen (3 torr), then carbon dioxide (100 torr) and finally after the trapping of CO2.200 400 60C 8 0 3I time, min0-5 I 1.5 2log P c o zFIG.8~.-The variation of the conductivity of vacuum-activated zinc oxide after the introductionof carbon dioxide (100 torr), then oxygen (3 torr), part A ; and finally after the trapping of C02,part B.FIG. 8~.-The variation of the conductivity of vacuum-activated zinc oxide as a function of carbondioxide pressure in the presence of oxygen (poZ = 3 torr).The reverse experiment on which oxygen (3 torr) was introduced to the catalystcontaining pre-adsorbed CO2 (100 torr) is shown in fig. 8 ~ . Part A of the curveaccounts for the influence of 0 2 in presence of COz and part B corresponds to 0 2remaining alone after trapping of C02. The condensation of this gas therefor372 MECHANISM OF co OXIDATIONliberates some parts of the surface which are active in the chemisorption of 0 2and confirms the competitive chemisorption of C02 and 0 2 .The sites involved in 0- and CO2 adsorption are probably Zn: ions or Znpatoms.For each C02 molecule the number of these sites may be either one or twoas in the case of oxygen. Two possibilities are therefore envisaged :(31)(32)The concentration of adsorbed oxygen and consequently the concentration ofconductivity electrons then depend not only on the oxygen pressure but also on thecarbon dioxide pressure. Fig. 8~ shows the plot of log Q as a function of log pco2for a constant PO, = 3 torr. The slope of the straight line is equal to 0.45.and desorption of each gas are equal. For oxygen,(33)for C02,(34)substituting ko- = (k-l/k-l)* and kco, = k21k-2 and taking the ratio of eqn.(33)and (34) :Combining with (33),As the surface coverage by oxygen is sparse,0- (ads) + C02 (9) +CO2 (ads) + $ 0 2 (9) + e-,20- (ads) + C02 (g)+CO2 (ads) + 0 2 (9) + 2e-.CASE I. CO2 IS ADSORBED ON ONE SITE.-At equilibrium the rates Of adsorptionk,p,,(l - O0- - 8co,)2[e-]2 = k - &- ;k2Pc02(1-00- - &o,) = k - 20c02;Oco, = kC02PC02~0-/ko- P&,Ce-I- (35)00- = kO-P~,Ce-l/(l + ko-P&,[e-l+ kC02PC02) (36)80- = ko-P8,re-l/(1-t- ~cozPcoz)~ (37)p-] = [eol-[e-I, (12)[O-] = kOo-. (38)eo-=[ eollk. - - (39)[e-I = [ e m +kco,Pco,)lkko-P&. (40)At equilibrium the concentration of adsorbed oxygen is given bybutThe adsorption of oxygen makes [e-] small compared to [e;], so that [e;] - [e-] M[e;] and thereforeSubstituting (39) into (37) :If the product ~ C O , ~ C O , were small compared to 1 (sparse adsorption of COz),the electrical conductivity would not depend on pco,.However, even for smallpressures of C02, the conductivity of the oxygen containing catalyst increases. Itwould therefore appear that ~ C O ~ P C O , and 1 are of the same order of magnitude.For higher pressures ~ C O , ~ C O , is is certainly much greater than 1 and the coverageby CO;? is not small. Eqn. (40) then becomesIf this equation were observed for a constant PO, the conductivity would be pro-portional to the first power of ~ c o , . However, it is seen in fig. 8~ that the slopeis 0-45 and not 1 .The hypothesis of the adsorption of C02 on one site (eqn. (31))must therefore be discarded.re-I = re,lkco2Pcoz/kko- P L (41P. AMIGUES AND S. J. TEICHNER 373CASE 2. CO2 IS ADSORBED ON TWO SITES.-Eqn. (33) for Oxygen iS not modi-fied, but for C02 we now haveIn this equation the exponent of Ocoz is unity and not 2. For oxygen, the rate ofdesorption depends on 0;- because two oxygen atoms have to join to produce amolecule. For carbon dioxide even adsorbed on two sites, the rate of desorptiondepends on Oco,. In the same way as before the following equation may be derivedwherere-3 = AP$o,/Ph, (43)A = k c o , [ e ~ ] / k k o - and k$/k-, = kco,.For a constant oxygen pressure the plot of log as a function of log pco, shouldthen give a straight line of slope equal to 8.This is the case of fig. 8 ~ . Chemi-sorption of one C02 molecule therefore displaces two oxygen atoms; this tendsto show that this chemisorption requires two surface sites and that these sites arethose active in oxygen adsorption. If the neutral adsorbed oxygen atom is alsopresent, eqn. (33) and (42) becomek,poz(l-60- -Oo-6co,)2[~-]2 = k-16&,k2Pco2(-l-~0- - ~ o - ~ c o , ) 2 = k-26C02but eqn. (43) is not modified.(C) CO2 (ads) + co (9)Finally, this interaction may cast some light on the sites for adsorption of CO.When carbon monoxide was introduced (3 torr) on a catalyst with pre-adsorbedC02 (100 torr) no variation of conductivity was recorded, although in the absenceof C02 the conductivity rose with the pressure of CO as seen earlier.The sites forthe adsorption of CO are therefore inhibited by C02. It is therefore concludedthat C02 and CO as well as CO;! and 0 2 are competitively chemisorbed on thesame sites, which are probably Zni or Zn:. The conclusions would be the sameif these sites were oxygen vacancies. It will be also shown below that CO and 0 2(reagents) equally compete for the same type of sites, as may be deduced from theprevious conclusion.OXIDATION OF CARBON MONOXIDEIt has been shown that the product of reaction CO2 is adsorbed on the same sitesas the reagent gases. With a view to preventing the inhibition of the reaction aliquid-nitrogen trap was placed very close to the catalyst. Pressure measurementsas a function of time allowed the reaction kinetics to be followed.Fig.9 shows kinetic curves for a stoichiometric mixture (curve A) and for mix-tures with an oxygen excess (curves B, C, D) or a carbon monoxide excess (curvesE, F). Differential and integral methods of determination of the overall order ofthe reaction gave the value 3 for stoichiometric mixtures at various initial totalpressures (maximum pressure 3 torr). The linear relationship,where PO is the total initial pressure and P pressure at time t enabled the rate con-stant to be calculated (k = 5 x 10-3 mm3 min-1 at 261°C) for all stoichiometricmixtures. For non-stoichiometric mixtures it is seen from fig. 9 that the rate of2(P& P+) = kt, (44374 MECHANISM OF co OXIDATIONthe reaction depends on the partial pressure of 0 2 and does not vary with the partialpressure of CO.The partial order determinations show that for these mixturesthe partial order with respect to 0 2 is 5, and that with respect to CO is zero. Theresult for CO is in agreement with that of other authors working with a ZnO catalystat 25OOC.239 24 The rate of the reaction is then expressed by2, = kP&P&-The order + with respect to 0 2 is interpreted by assuming a weak adsorption ofthis gas in dissociated form. The order zero with respect to CO would imply thatthe surface is covered by this gas; and a maximum in the adsorption isobar of COis observed by Kubokawa at 25O0C,20 a temperature close to that selected for thereaction in this work.U 100 200 300time, minFIG.9.-The kinetics of the COfO2 reaction for different mixtures COS.02. Curve A, 2 C0+02 ;curveB,2CO+1-1202; curveC,2C0+1-3602; curveD,2C0+2.1602;curveE,2C0+0-9202;curve F, 2 CO + 0.66 0 2 ; total pressure 3 tom.The two reactive species may be adsorbed on different sites and this would beconsistent with kinetic results. However, there is a presumption in favour of acompetitive adsorption of CO and 0 2 on the basis of the previous conductivityresults. Also an inhibition by CO is observed for CO-excess mixtures (fig. 9, curvesE, F) when the reaction goes to completion (high ratio ~co/Po,). On the otherhand, an accelerating effect by 0 2 is noted for 02-excess mixturss (fig. 9, curvesB, C) when the reaction is accomplished to about 80 % (high ratio po,/pco).Conductivity measurements during the reaction give supplementary supportfor the hypothesis of competitive adsorption.Fig. 10 represents conductivity valuesin the reaction for mixtures showing the kinetics of fig. 9. When the reaction mix-ture is admitted to the catalyst its surface is under the influence of CO and 0 2 whoseeffects on the conductivity are opposed (C02 is trapped). For a stoichiometricmixture the ratio po,/pco remains constant (0-5) during the course of the reaction.If one of the gases is in excess, its influence becomes preponderant as the reactiontends to completion and the ratio po,/pco starts to become appreciably differentfrom 0-5. The curves of fig. 10 for oxygen excess mixtures (curves B’, C’) show thatnear the end of the reaction ( P O , $ ~ c o ) oxygen adsorption becomes easier, as judgedfrom the conductivity decrease.At the same time the rate of the reaction increases(curves B and C, fig. 9). Thus, oxygen seems to be active in reaction when it is iP. AMIGUES AND S. J. TEICHNER 375an adsorbed state, but its chemisorption appears to be hindered by the initial rela-tively high pressure of CO (high ~ c o / ~ o , ratio). For mixtures with a CO excessan initial decrease of the conductivity (fig. 10) is followed by an increase when theratio pco/po, begins to become high. The catalytic activity is not very much affectedfor these mixtures compared to the behaviour of the stoichiometric one (fig. 9).too 200time, minFIG. 10.-The variation of the conductivity of vacuum-activated zinc oxide during the reaction withmixtures of fig.9. Curve A’, 2 C0+02 ; curve B’, 2 CO+ 1.12 0 2 ; curve C’, 2 CO+ 1.36 0 2 ;curve D’, 2 C0+2.16 0 2 ; curve E’, 2 CO+O.92 0 2 ; curve F’, 2 C0+066 0 2 .The interpretation of the conductivity results (fig. 10) may be sought in termsof the hypothesis that 0 2 and CO species are adsorbed on the same, or on different,sites. The chemisorbed species are cleared from the surface either by desorptionor by reaction. At the stationary state the rate of adsorption of each gas is thenequal to the rate of desorption or to the rate of reaction, depending on the relativemagnitude of these rates. For simplicity, it is first assumed that 0- is the reactivespecies. As far as conductivity results are concerned, the involvement of neutral0 species in the reaction does not interfere in these results because of the equi-librium (28).If CO and 0 2 are adsorbed on the same sites :(45‘)(46’376 MECHANISM OF co OXIDATIONk-108- and k-28~0 represent the rate of desorption of adsorbed species and k’8o- 8cotheir rate of reaction. An explanation of the results of electric conductivity (fig.10)may be given only for the competitive chemisorption of 0- and CO, their rates ofdesorption being higher than the rate of the reaction in the adsorbed state.17 Eqn.(45) and (46) are therefore selected, and not (45’) and (46’), with k-l8& %k’Oo- 8coand k-~Oco~k’80-8~0. This givesdOo-/dt = k,po,(l-Oo- -Oco)2[e-]2- k - l e i - , (47)d6co/dt = kzPco(1-00- -Oco)-k-zOco.(48)Expressing 00- for the stationary state:ko- p8,L-e-I1 + ko -PA, Ce -1 + kC0PCOYeo- =withko- = (kl/k-l)* and kco = kz/k-,.But oxygen adsorption being sparse, ko-p&,[e-] < 1, henceAt 261 “C,from (39) and (50),withThe conductivity being proportional to [e-1, it follows that60- = ko-P&,[e-l/(~ fkcoPc0).Oo- = [e,]/k;L-e-I = 4 1 + kcoPco~/P~2YA = [e,]/kko-.(49)0 = A’U + ~c0Pc0)/P~2. (52)Fig. 11 represents the variation with time of log c calculated from eqn. (52)and from experimental values of the pressure (fig. 9). The value of kco was arbit-rarily chosen to be equal to 28 in order to obtain the curves on fig. 11 not too distantfrom one another, as they correspond to the curves on fig.10. From (47) and (48)the value of OCO is established :OCO = kcoPco/(~ + kcoPco). (53)With kco = 23, 6c0 is close to I, if pco is not too low.It is seen from fig. 10 and 11 that values of log c calculated on the basis of thehypothesis advanced above agree fairly well with the experimental results. Somedeviations in fig. 11 which appear when the reaction is close to completion are easilyexplained. For CO-excess mixtures, for example, Q is not infinite (eqn. (52)) buttends to 00 as [e-] is no longer negligible compared to [e;] and tends to be equalto [e;]. The existence of adsorbed neutral oxygen which was supposed previouslydoes not change the results of fig. 11. Eqn. (49) would then becomebut as before,which leads to eqn. (50) and (52).1 + k O P C O + ko-P82Ce-I +kOPP.AMIGUES AND S. J. TEICHNER 377Although the conductivity results are explained by the competitive adsorptionof 0- and CO species, the reaction cannot proceed through 0- as in the scheme,The reaction rate would be represented byBut it has been shown that if the conductivity of ZnO varies with p& the coverage00 - is practically constant at 26 1 "C (eqn. (39)) and 8co w 1, so the rate would remainconstant (overall order zero).0- (ads) + CO (ads) - 4 0 2 (g) + e-. (55)v = coo-eco. (56)D'I I100 200 300time, minFIG. 1 1 .-Values of the conductivity calculated according to eqn. (52) and from the experimentalvalues of pco and p o z recorded for experiments of fig. 9. Curve A', 2 C0+02 ; curve B', 2 COf1.12 0 2 ; curve C , 2 C0+1.36 0 2 ; curve D', 2 COf216 0 2 ; curve E', 2 COf0-92 0 2 ; curve F',2 CO + 0.66 0 2 .If the neutral adsorbed oxygen is more reactive than the species 0- with whichit is in equilibrium (eqn. (28)), the order 8 with respect to this gas and the kineticresults (fig. 9), in particular the inhibition or acceleration for high or low pco/po,ratios, may be explained.The hypothesis that the reaction takes place between0 and CO adsorbed on the same sites must be discarded as it leads to the order4 with respect to PO, and - 1 with respect to pc0.17 The hypothesis that 0 and COreact, being adsorbed on diflerent sites is in agreement with the partial orders ofreaction but does not explain the acceleratingeffect on the reaction due to the ratiopo,/pco becoming high or an inhibiting effect due to the ratio pco/po, becoming high.However, all the experimental facts (conductivity and kinetics) are explainedif 0 2 and CO compete for one type of site without reaction and CO reacts from thegas phase by a Rideal mechanism.An alternative explanation is that the reactionproceeds between 0 adsorbed on sites S (in competition with CO on these sitesbut without reaction) and CO weakly adsorbed on sites S'378 MECHANISM OF co OXIDATIONThe previous assumptions for the stationary state being maintained, it is foundthatThe general equation for the rate of the reaction isConductivity and kinetic results have shown thatTherefore,1 + k c o p c o ~ k o p g , + k o - p ~ ~ [ e - ) .V =For kcopco $ I , this equation can be simplified toorThis equation accounts for order 8 with respect to 0 2 and 0 with respect to COif Ci% C;.This condition implies that the reaction takes place between 0 (ads)and CO from the gas phase. Eqn. (62) with kco = 28 when applied to the dataof fig. 9 (expt. B and C) shows that the rate of the reaction is increased when 75 %of the conversion is performed, which is in agreement with the kinetics of fig. 9.For a mixture with an excess of 0 2 (expt. B and C, fig. 9 and 10) pco decreasessteadily when the reaction goes to completion, but P O , has still a measurable value.When eqn. (63) is taken into account, though C; is higher than C;, Cj'p&,/pcomay be larger than C$&, and the rate increases.At the same time, for the samepercentage of conversion, the conductivity decreases greatly (fig. 10 and 11, expt.B and C) which is a further argument that CO in the adsorbed state reacts with0 (ads) as almost no CO is present in the gas phase.In short, oxygen is adsorbed in two forms. The concentration of 0- is almostindependent of p o , (though the conductivity does depend on PO,) therefore this specieswould not participate in the reaction. The active species would be neutral 0 (ads)whose concentration is proportional to ph2. The chemisorption of CO hindersthe adsorption of oxygen in the active form. This explains the inhibiting effectof an excess of CO on the reaction. The conductivity being dependent on pco(eqn. (51)), the variation of the rate of the reaction and of the conductivity aretherefore not independent.In particular, for a reaction mixture with an excess ofoxygen, the increase of the rate of the reaction (fig. 9) is related to the decrease of6 (fig. 10)P. AMIGUES AND S . J. TEICHNER 379Finally, the kinetic equations developed for pure ZnO are valid for gallium-or lithium-doped catalysts. The conductivities at the steady state for the reactionperformed at 261°C are respectively : 3.2 x 10-6 for ZnO+Li, 3.2 x 10-4 for ZnOand 1.25 x 10-3 ohm-1 cm-1 for ZnO + Ga. The apparent activation energy of thereaction for these three catalysts are: 15.9 for ZnO+Li, 15.9 for ZnO, and 17-2kcal/mole for ZnO+Ga. The doping does not seem therefore to have any in-fluence on the reaction. As the rate-determining step does not involve a chargetransfer because only neutral species are involved, the lack of relation betweenthe catalytic activity and the electronic structure is expected.The interest of Prof. M. Prettre in this work is gratefully acknowledged.1 Achalme, Compt. rend., 1912, 154, 352.2 Nyrop, J. Physic. Chem., 1935, 39, 643.3 Wolkenstein, ThCorie L?lectronique de la Catalyse sur les Semi-Conducteurs (Masson et Cie,4 Schwab and Block, J. Chem. Physics, 1954, 51, 664.5 Heckelsberg, Clarck and Bailey, J. Physic. Chern., 1956, 60, 559.6 Keier, Roginsky and Sozonova, Compt. Rend. U.R.S.S., 1956,106,859.7 Teichner, J. Catalysis, 1966, 5, 724.8 Boreskov and Matveiev, Problems in Kinetics and Catalysis, 1955, 8, 165.9 Keier and Chizhikova, Proc. Acad. Sci. U.R.S.S., 1958, 120, 1.Paris, 1961).Schwab, Angew. Chem., 1963, 2, 59.Keier and Chizhikova,Problems in Kinetics and Catalysis, 1960, 10, 77.10 Arghiropoulos and Teichner, J. Catalysis, 1964, 3, 477.11 Arghiropoulos, B., Thesis (Lyon, 1962).12 Aigueperse, Arghiropoulos and Teichner, Compt. rend., 1960, 250, 550.13 Aigueperse and Teichner, Ann. Chim., 1962, 7, 13 ; J. Catalysis, 1963, 2, 359.14 Barry and Stone, Proc. Roy. SOC. A , 1960, 355, 124.15 Kokes, J. Physic. Chem., 1962, 66, 99.16 Peers, J. Physic. Chem., 1963, 67, 2228.17 Amigues, P., Thesis (Lyon, 1964).18 Taylor and Amberg, Carz. J. Chem., 1961, 39, 535.19Nagarjunan and Calvert, J. Physic. Chem., 1964, 48, 17. Lee and Mason, Proc. 3rd Int.20 Keier and Chizhikova, Compt. Rend., U.R.S.S., 1958, 120, 830. Nagarjunan, Sastri andKubokawa. Bull. Soc. Chem. Japan,Glemza and Kokes, J. Physic. Chem., 1962, 66, 566.Congr. CataEyzis (North-Holland Publ. Co., Amsterdam, 1965), p. 748.Kuriacose, Proc. Nut. Acad. Sci. India, 1961, 27, 496.1960, 33, 555.21 Matsushita and Nakata, J. Chem. Physics, 1962, 36, 665.22 Kubokawa and Toyama, Bull. Chern. Soc. Japan, 1962,10, 1407.23 Matsuura, Kubokawa and Toyarna, Nippon Kagaku Zasshi, 1960. 81. 997.24 Doerfiler and Hauffe, J . Catalysis, 1964, 3, 156. 171

ISSN:0366-9033    

DOI:10.1039/DF9664100362

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Hall Effect Studies of Carbon Monoxide Oxidation overDoped Zinc Oxide CatalystsBY HAKZE CHON AND CHARLES D. PRATERSocony Mobil Oil Co., Inc., Research Dept., Central Research Division, Princeton,New JerseyReceived 18th January, 1966The role of the conduction band electrons in the chemisorption of oxygen and the oxidationof carbon monoxide on indium-doped polycrystalline zinc oxide was studied by a dual a.c. Halleffect method that employed a 52clsec ax. sample current and a 6Oc/sec a.c. magnetic field togive an 8 c/sec Hall voltage. This method seems to be particularly well suited for measuring theconcentration of current carriers in semiconductor catalysis studies.The results of this study support the conclusions that the chemisorption of oxygen on indium-doped zinc oxide in the temperature range 200-350°C is associated with an electronic charge transferfrom the zinc oxide to the surface oxygen species, and that 0- plays a prominent role in this ad-sorption.The chemisorption of carbon monoxide in the temperature range 300-35OoC, on indium-doped zinc oxide containing pre-chemisorbed oxygen, gives an increase in carrier concentrationcorresponding to one electron per carbon monoxide molecule chemisorbed. The measurementsof the changes in carrier concentration during carbon monoxide oxidation at 350°C by this dopedzinc oxide indicate that the conduction band electrons are intimately involved in the reaction, andthat 0- is an important transient surface species therein. A comparison of the results obtainedfrom reaction studies at 460°C over lithium-doped and indium-doped zinc oxide supplies addi-tional evidence that a strong correlation exists between the concentration of conduction bandelectrons in these materials and their level of catalytic activity for carbon monoxide oxidation.Schwab and Block 1 9 2 have used results obtained from electrical conductivitymeasurements to provide evidence that the electrons in the conduction band ofzinc oxide are directly involved in the oxidation of carbon monoxide over thiscatalyst.Other investigators 3-8 have made similar studies with regard to theadsorption of oxygen. For such studies, however, the use of the Hall coefficienthas a number of advantages over that of the electrical conductivity. For example,the Hall coefficient gives the number and sign of the current carriers, informationmore useful in catalytic studies than the product of the number and mobility of thecarriers obtained from electrical conductivity.Furthermore, the Hall coefficientis less affected by grain boundaries and other inhomogeneities in the sintered poly-crystalline materials used to satisfy the surface area requirements of the catalyticreaction. The Hall coefficient gives information mostly about the bulk propertiesof the solid even in the presence of such inhomogeneities, whereas the electricalconductivity gives information mostly about the grain boundaries.3~ 99 10 Halleffect measurement has been little used in catalytic studies, however, because itnecessitates difficult experimental procedures.Many of these difficulties can belessened by using alternating currents and magnetic fields instead of the directcurrents and constant magnetic fields usually employed. The a.c. method isparticularly useful for following the changes in carrier concentrations in the catalystas the chemical reaction proceeds.In the adsorption of oxygen.on zinc oxide, photo-adsorption 11 and electron spinresonance 12 studies have been used to provide evidence as to the nature of the38H. CHON AND C. D. PRATER 381adsorbed species. During the catalytic reaction, on the other hand, transient speciesthat are negligible in such adsorption studies may become prominent and playan important role in the reaction. Hence it is desirable to observe continuouslyand simultaneously the number of electrons in the conduction band and the com-position of the ambient gas phase.A combination of time-of-flight mass spectro-metry and the a.c. Hall technique makes such observations possible. In the studypresented in this paper, such a combination is used to examine the oxidation ofcarbon monoxide over zinc oxide catalyst containing indium impurities as electrondonors.EXPERIMENTALSeveral ax. methods for measuring the Hall voltage have been described.13-15 Wehave used a dual a.c. method with ax. sample currents and a.c. magnetic fields at differentfrequencies so that the Hall voltage appears at the sum and the difference frequencies.13The appearance of the Hall voltage at the new frequencies makes it possible to reject, byuse of selective amplifiers, any stray signals from the current and magnetic field circuits.For example, a signal will be introduced into the Hall voltage circuit if the two Hall probesare not on the same equipotential line of the electrical field providing the sample current.This mis-alignment voltage, which is generally much larger than the Hall voltage, can beeasily and quickly rejected in the dual a.c.method, but must be carefully balanced out inthe d.c. and single frequency a.c. methods. A block diagram of the dual a.c. Hall ap-paratus is shown in fig. 1. In our experiments a 60 c/sec, 1500 gauss, magnetic field andRECORDER 0I'--1 AMWFlER . -zqFIG. 1.-Block diagram of the dual a.c. Hall effect apparatus.The reference sample circuit canbe omitted when the signs of the carriers are known.a 52 c/sec sample current were used to obtain an 8 c/sec Hall voltage. This voltage wasamplified by a battery-operated 8 c/sec selective amplifier, preceded by a cathode followerinput stage and two matched twin-T resistance-capacitance filters set to reject 52 and 60clsec. The circuit for the selective amplifier, shown in fig. 2, employs a parallel twin-Tresistance-capacitance network tuned to reject 8 c/sec and to provide negative feedbackfor all other frequencies. The amplifier, with its cathode follower and two rejection filters,provides a gain of 725 at 8 c/sec and 4 x 10-4 at 52 c/sec. The 8 c/sec output of the amplifierand the 52 c/sec sample current were continuously and simultaneously recorded on a two-channel medium frequency recorder (Texas Instrument C.Oscillo-Writer). The sign ofthe carrier in experimental samples can be determined, when necessary, by comparing thephase of the 8 c/sec Hall signal of a reference Hall sample of known type with the phas3 82 HALE EFFECT I N ZINC OXIDEof the 8 clsec signal from the experimental sample. The electrical techniques are notdifficult to apply; the major problem is to keep the stray-signal pickup and noise in thesystem at a minimum.The Hall coefficient RH was calculated from the relationR, = 2l/,tW/i@d, (0where VH is the Hall voltage, i is the total current, 0 is the magnetic field strength, t is thesample thickness, W is the sample width, and d is the distance between the Hall probes.The factor 2 in eqn.(1) arises from the modulation producing the Hall signal. The carrierconcentration n is given by the relationn = 1.18/RHq = 0.59i0d/qVHtW, (2)where q is the electrical charge of the carrier and 1.18 = 3n/8 is the lattice scattering factor.6SJ7 %6SL7 6SJ7 %6SL7 6SL7-225 hLl VFIG. 2.-Input stage and 8 c/sec selective amplifier. Voltages are supplied by batteries. Allcapacitance are in micro farads. Precision resistors and capacitors with a tolerance of f l % areused in the twin T filters in the feedback circuits. The 60 clsec filter was manufactured by FreedTransformer Company.Powdered samples of spectroscopically pure zinc oxide and zinc oxide doped withindium and with lithium (supplied by New Jersey Zinc Company) were used to preparethe Hall samples for the oxidation and adsorption studies reported here.Samples pre-pared in our laboratory by impregnating spectroscopically pure zinc oxide with galliumnitrate were used in experiments to examine the effect of inhomogeneities on the Hallvoltage. The preparation of the Hall sample requires the development of considerableskill to obtain samples that are uniform, of adequate density, free of macroscopic cracks,and of sufficient mechanical strength to permit handling. Hall samples were preparedby pressing the powder, at 440 kglcm2, in a six-section split die shown in fig. 3. Electrodeswere strips of 0.001 in. thick platinum foil, partially imbedded in the sample as shown infig.4. The Hall samples were sintered at various temperatures between 850 and 1000°Cfor 2 h in a nitrogen atmosphere. To avoid contamination during the sintering, thesamples were surrounded by zinc oxide powder of the corresponding composition. After-ward, they were cooled slowly in a nitrogen stream to about 6Oo0C, and then rapidly toroom temperature. The average dimensions of the Hall samples after sintering were2.6 x 0.9 x 0.1 cm. A typical value of the surface area of the sintered Hall samples was0.1 m2/g, as determined by the B.E.T. method using the adsorption isotherm of kryptonat liquid-nitrogen temperature.The volume of thereaction chamber was made sufficiently small so that the number of oxygen and carbonmonoxide molecules at a pressure of 500p was of the same order of magnitude as the?he reaction chamber and auxiliary systems are shown in fig.4H. CHON AND C . D. PRATER 383number of electrons in the conduction band of the Hall sample. The sample holderfor the reaction chamber is also shown in fig. 4. Toreduce as much as possible the amount of metal ex-posed to the reaction mixture, the electrical leads to thesample were sealed in gIass, except for a small portionneeded for attaching the sample. Pressure measure-ments during adsorption and reaction were made witha thermistor gauge constructed from a matched pair ofthermistors and calibrated against a McLeod gauge foreach gas. Before either adsorption or reaction measure-ments were made, the sample was evacuated at 450°C toa pressure of 2 x 10-6 mm by a mercury diffusion pump.The heater for the feaction chamber was constructedwith a bifilar winding and was capable of producing atemperature of 550°C in the reaction chamber.After thevacuum pre-treatment, the reaction chamber was isolatedand a known amount of gas was admitted. For adsorp-tion and reaction studies, the pressure was continuouslymeasured by the thermistor gauge; and for reactionstudies, the compositional changes of the reaction mix-ture were continuously followed by a time-of-flight massspectrometer connected to the reaction chamber by acontinuous leak. The volume of gas removed by theleak was sufficiently small so that the number of moleculeslost had a negligible effect on the reaction.The transienttime for the mass spectrometry sample to pass from thereaction chamber to the mass Spectrometer ionizationchamber was about 10 sec. Any carbon dioxide desorbedfrom the catalyst during the reaction was removed fromthe gas phase by continuously trapping it in a cold fingerimmersed in liquid nitrogen.FIG. 3.-Hall sample forming dieassembly. The die is made in sixsections to allow electrodes to boembedded in the sample duringthe formation, and to allow forremoval of the sample by dis-assembling.FIG. 4.-Reaction chamber and auxiliary system : a, calibrated gas containers ; b, thermistorvacuum gauge elements; c, magnet pole piece; d, bifilar wound furnace; e, cold finger; f, glassleak ; g, sampIe ; h, electrode ; i, lead wires imbedded in glass.RESULTSSome experiments were made to determine how well the dual ax. Hall measure-ments on sintered polycrystalline zinc oxide reflect the bulk carrier concentrations384 HALL EFFECT IN ZINC OXIDEThe solid curve in fig. 5 gives the observed carrier concentration as a function ofthe temperature for a sintered, polycrystalline, indium-doped zinc oxide samplefor temperatures from 77 to 770°K.From the temperature dependence of thecarrier concentration, the donor concentration for this sample was calculated tobe 1.5 x 1018 cm-3 ; the indium impurity concentration, determined by spectro-graphic analysis, was given by New Jersey Zinc as 1.9 x 1018 cm-3. The dashedcurve in fig. 5 reproduces data obtained by Bogner 16 using a single crystal of indium-doped zinc oxide with an indium concentration of 0.9 x 1018 cm-3.The two curves22 1/T (“K-1) x 0-3FIG. 5.Electron concentration of indium-doped zinc oxide as a function of temperature forpolycrystalline and for single crystal Hall samples. The solid curve with open dots is the resultsobtained in this study with polycrystalline indium-doped Hall sample containing 1.5 x 1018 donors/cm3. The dashed curve gives the results obtained by Bogner 16 with a single crystal containing0 9 x lO1*/cm3 indium donors.are similar. The donor levels, as determined from the temperature dependence ofthe carrier concentration given by the solid curve in fig. 5, were 0-057 eV below thebottom of the conduction band. The insensitivity of the Hall effect measurementsto inhomogeneities is demonstrated by the results obtained for two gallium-dopedHall samples containing the same amount of gallium impurities but subjected todifferent preparation procedures so that a different degree of volume shrinkage wasobtained.Hall effect measurements on the two samples gave the same carrierconcentration, even though their electrical conductivities differed by a factor of 6.On the other hand, if the pressing and sintering are insufficient, the inhomogeneitybecomes sufficiently extensive to give an appreciable change in Hall voltage.Nevertheless, for zinc oxide there seems to be a range of conditions that givessufficient internal surface area for an adequate catalytic reaction, in which theinfluence of inhomogeneities on the Hall voltage is sufficiently small to be neglected,as shown by our measurementsH.CHON AND C. D. PRATER 385The expected increase in carrier concentration on doping with indium or galliumand the decrease on doping with lithium is seen by making a comparison of roomtemperature values : the carrier concentration in undoped zinc oxide samples isapproximately 3 x 1016 cm-3 ; in an indium-doped (45 p.p.m.) sample, 7.2 x 1017cm-3 ; in a gallium-doped (55 p.p.m.) sample, 1.2 x 1018 cm-3 ; in a lithium-doped(500 p.p.m.) sample, 2.5 x 1014 cm-3.Fig. 6 gives the changes observed in electron concentration with the passage oftime at 255, 365 and 400°C in a Hall sample exposed to oxygen at a pressure ofJ I 1 L20 40 60 80 100 120time (min)FIG.6.-The time course of the chemisorption of oxygen by indium-doped zinc oxide for a partialpressure of oxygen of 900 p.900 p. This Hall sample was formed from zinc oxide containing 45 p.p.m. indium.These data are adequately simulated by the Elovich equation, as shown in fig 7.The amount of oxygen adsorbed, as measured manometrically, was compared tothe change in the number of electrons in the conduction band in a series of experi-ments in which small incremental volumes of oxygen were successively adsorbed bythe Hall sample. The initial pressure for each adsorption increment was in the range20-40 p, and the final pressure before the next increment was admitted or the experi-ment terminated was always less than 3.5 p. The results obtained at 200, 300 and350°C are summarized in table 1 for three different Hall samples of indium-dopedTABLE 1 .-CHEMSORPTION OF OXYGEN ON INDIUM-DOPED ZINC OXIDEsample temp., ‘C number of oxygenatoms chemisorbeddecrease in thenumber of electrons AO/AeA 200 0.55 x 1016 0.51 x 1016 1 -08300 2-05 2.12 0.97350 2-66 2.79 0.95B 300 3.78350 4.91D 250 1.82300 4.183.334.281-813.631-131.151.011.1386 HALL EFFECT IN ZINC OXIDEzinc oxide.The data show that approximately one electron is lost from the con-duction band for each oxygen atom adsorbed at the temperatures studied. Whenwell-outgassed Hall samples were contacted with carbon monoxide, only smallincreases in carrier concentration were observed. These small changes are at-tributed to the interaction of the carbon monoxide with a small amount of chemi-sorbed oxygen remaining on the surface of the outgassed samples.On the otherhand, when carbon monoxide was contacted with samples on which an appreciablelog10 ( t - t t o )FIG. 7.-The fraction of electrons lost from the conduction band of indium-doped zinc oxideduring oxygen chernisorption plotted according to the integrated Elovich equation. For 255"C,to = 0002 min, for 365"C, t o = 0.8 min, for 4OO"C, to = 1 min.number of oxygen atoms was chemisorbed, substantial increases in the number ofelectrons in the conduction band were observed. The data presented in table 2show that, at 300 and 350"C, the correlation between the amount of carbon monoxideadsorbed, as determined manometrically, and the increase in number of electronsin the conduction band is one electron introduced into the band for each moleculeof carbon monoxide adsorbed.No desorption of carbon dioxide was observedduring the experiments at 300 and 350°C. At the end of the adsorption period inthe 300°C experiment, the temperature of the sample was increased to 400°C. Thisincrease led to the desorption of an amount of carbon dioxide equal to the amountof carbon monoxide adsorbed. No changes in carrier concentration were ob-served when carbon dioxide was adsorbed on these Hall samples at 300 and 350°Ceven though manometric measurements in the experiment at 300°C showed that theamount of carbon dioxide adsorbed was several times the amounts of oxygen orcarbon monoxide adsorption observed in the previous experiments,TABLE 2.-cHEMISORPTION OF CARBON MONOXIDE ON OXYGEN-PRE-CHEMISORBEDZnO (In)ACOlAe number of CO increase in theadsorbed number of electrons sample temp., "CA 300 0 .7 2 ~ 1016 0.62 x 1016 1-16B 300 2.69 247 1 -09350 3-52 2.92 1 *20C 300 5.5 1 4.84 1.1H. CHON AND C. D. PRATER 387The kinetics of the oxidation of carbon monoxide on indium-doped zinc oxidewas studied at 350°C by reacting an approximately stoichiometric mixture of oxygenand carbon monoxide over a Hall sample. The initial partial pressures of oxygen4 8 12 16 20 24time (min)FIG. 8-Electron concentration and ambient gas composition changes observed during carbonmonoxide oxidation over an indium-doped zinc oxide Hall sample at 350T.20 A0 60 80 LOO 120time (miil)FIG. 9.-The comparison of the changes in electron concentration observed on indium-dopedand on lithium-doped zinc oxide at 460°C.The minimum electron concentration in the indium-doped zinc oxide had an approximately stationary state lasting about 2 min. It cannot be shownon the time scale used in the figure.and carbon monoxide were 210 and 430 p. The volume of the static reactingsystem was such that these pressures correspond to 2.1 x 1017 oxygen and 4.3 x 1017carbon monoxide molecules per cm3 of Hall sample. The results obtained aresummarized by the graphs in fig. 8. A rapid increase in carrier concentratio388 HALL EFFECT IN ZINC OXIDEbegins when the oxygen partial pressure drops almost to zero.The electron con-centration after essentially all the carbon monoxide had reacted was 5.8 x 1017 cm-3of Hall sample. The changes in carrier concentration during reaction of oxygen+carbon monoxide mixtures over indium- and over lithium-doped zinc oxide at 460°Care compared in fig. 9. No ambient gas-phase analyses were obtained in theseexperiments. The volumes and initial partial pressures of the gases in these twoexperiments were the same. The initial electron concentration in the indium-doped sample was 1 . 4 6 x 1018 cm-3, and in the lithium-doped sample 7.7 x 1016 cm-3.A much larger rate of reaction over the indium-doped sample than over the lithium-doped sample is indicated by the large difference in the length of time required toreach the point where the rapid increase in electron concentration begins.DISCUSSIONThe a.c.Hall method gives a surprisingly good measurement of the electronconcentration in the conduction band of the polycrystalline indium-doped zincoxide samples used in this study, as shown by : (i) the similarity between the tem-perature dependence curve for the polycrystalline sample and the single crystal ;(ii) the nearness of the donor level concentration as measured by the Hall voltageto the impurity concentration, as measured spectroscopically ; (iii) the demon-strated insensitivity of the a.c. Hall method to grain boundary modifications in auseful range of sample surface areas, and (iv) the internal consistency of the ad-sorption and reaction results. Volger 9 has given a theoretical treatment of the Halleffect in an inhomogeneous conductor consisting of conducting cubes of dimension 1imbedded in a matrix of material with a much lower conductivity, so that each cubeis separated from an adjacent cube by a layer of thickness b<l.The Hall voltagefor this inhomogeneous conductor was shown to be essentially the same as thatof the cubes alone. Koppe and Bryan 10 have shown that, in a layer of conductingmaterial of uniform thickness exposed to a uniform magnetic field, cavities of anyshape have no influence on the Hall voltage. In both cases, however, large changesin conductivity can occur. Consequently, the dual a.c. Hall method appears toprovide a convenient experimental method to measure accurately the changes incarrier concentrations of semiconducting catalysts during adsorption and chemicalreaction.Barry and Stone 11 studied the adsorption of oxygen by zinc oxide from roomtemperature to 400°C and found that the rate of adsorption followed the Elovichequation.In our study, the rate of decrease of electron concentration in the con-duction band during adsorption follows the same law, as shown by the data givenin fig. 7. Many adsorption mechanisms that give an Elovich equation for adsorp-tion do not give a rate of disappearance of active sites that follow the same law 17and are eliminated from consideration by the above results when the “ active sites ”are the conduction band electrons. Weisz 1 8 has shown that the build-up of theboundary layer should introduce some function of the amount adsorbed into an ex-ponential in the expression for the rate of adsorption.An Elovich equation forthe rate of adsorption will result when the function is simply the amount adsorbed.Barry and Stone show that the formation of a charged boundary layer in the solidcan give an adsorption law identical with the Elovich equation. The boundarylayer mechanism can also give a rate of disappearance of electrons from the con-duction band that follows this law.Several investigators 39 11’ 12 have presented evidence for the presence of twokinds of adsorbed species, 0- and 02-, on zinc oxide. Barry and Stone11 givH. CHON AND C. D. PRATER 389evidence that 0 2 - occurs at high temperature, and suggest that it is eventually in-corporated into the lattice by an interaction with interstitial zinc. Thus, two pro-cesses are thought to contribute to oxygen adsorption: one involving the move-ment of electrons in the conduction band, and the other the migration of atomicspecies within the zinc oxide.Wagner 18 and others 199 20 have offered evidencethat the presence of trivalent impurity atoms in zinc oxide decreases the concentra-tion of interstitial zinc. Consequently, the contribution of interstitial zinc to oxygenadsorption and carbon monoxide oxidation should be reduced by using the indium-doped zinc oxide; thus the role of conduction band electrons in these processesmay be easier to determine. Better control of the concentration of donor levelsand less sensitivity to pre-treatment procedures is also obtained.A quantitative comparison between the number of oxygen atoms adsorbed andthe decrease in the number of electrons in the conduction band has been made foronly the initial parts of the adsorption isotherm.The approximate one-to-onecorrespondence between these numbers, as shown by the data in table 1, is consistentwith the hypotheses that the major portion of the adsorbed species on the surfaceis 0- in the temperature range 20O-35O0C, at least in those early parts of the iso-therm studied and for the relatively short times ( N 20 min) allowed for each adsorptionincrement.The results given in table 2 for the temperature 300 and 350°C show that approx-imately one electron appears in the conduction band for each carbon monoxidemolecule adsorbed by an oxygenated Hall sample, and are consistent with theexpectation that the adsorbed species with which carbon monoxide interacts is0-.Adsorbed carbon dioxide does not show any interaction with the conductionband electrons. If 0- is the only adsorbed species that contributes to the chargeconservation and is the only adsorbed species for oxygen, the conservation lawsfor charge and for atomic species givee+z-2y = eo+z0-2yo, (3)where e is the number of electrons cm-3 of Hall sample, y is the number of oxygenmolecules in the gas phase cm-3 of Hall sample, z is the number of carbon monoxidemolecules in the gas phase cm-3 of Hall sample, and the subscript denotes theinitial values of these numbers.Fig. 10 is a graph of e as a function of z-2y forthe carbon monoxide oxidation results given in fig. 8. The breakdown of the con-servation law given by eqn. (3) that occurs at the lower values of Ne shows thatthere is at least one transient species not accounted for in our conservation law.The direction of the departure shows that this transient species must have lesscharge per oxygen atom than 0-; it could be, e.g., either 0; or oxygen atoms.Although the present study has not progressed sufficiently for an extensivekinetic analysis of carbon monoxide oxidation to be made profitably, it is instructiveto see how well the data in fig. 8 can be simulated by simple and plausible models.This kind of exploration discloses much about the level of complexity required inmechanistic models that are to simulate particular features of the data.Three ofthe most striking features of the data given in fig. 8 are (i) the disappearance ofoxygen from the gas phase long before the carbon monoxide has been exhausted;(ii) the almost stationary state that exists in the electron concentration during alarge part of the time that oxygen is present in the gas phase; and (iii) the smallsize of the drop in electron concentration that occurs.The rate of adsorption for any adsorption mechanism can be expressed in a massaction form, provided the rate parameters are allowed to be functions of the con-centrations of adsorbed species. If the rate and extent of adsorption are determined1390 HALL EFFECT IN ZINC OXIDEby the build-up of a charged boundary layer, these rate parameters can be stronglydependent on the concentrations of the adsorbed species.The smallness of thechange occurring in the electron concentration during the reaction makes it morelikely that the data of fig. 8 can be simulated by a simple mass-action kinetics withconstant rate parameters.I 1 I 05 PO2-2yFIG. 10.-The data for CO oxidation given in fig. 8 plotted according to the conservation law givenby eqn. (3).Let us examine a model consisting of the three irreversible surface reactions :kik2k3O2 fe-0;0; +e-+20-0- + CO-CO, + e.(4)In the first two reactions in eqn. (4), the back reactions have been neglected becauseof the small change in electron concentration observed ; the electron concentrationat equilibrium for oxygen adsorption should have been much lower than that ob-served during the reaction.The third reaction has been made irreversible becausethe carbon dioxide adsorption experiments showed that the interaction representedby the back reaction was too small to be detected by our measurements. Since alarge change in free energy occurs in carbon monoxide oxidation, at least one suchessentially irreversible step is to be expected. For the set of reactions given byeqn. (4), the conservation laws for charge and atomic species gives us two equations :one relating the number of 0- cm-3 of Hall sample, designated [O-1, to e, y and z;the other relating the number of 0; cm-3 of Hall sample, designated LO;], to thesequantities.These equations are[O;] = z+e-2y+u,[0-] = 2y-2e-z-v+”, (6H. CHON AND C. D. PRATER 39 1where u is a constant given by 2y0 -ZO - eo, and N is the number of electrons in theconduction band of an oxygen-free surface. The last quantity is to allow for theeffects of small quantities of 0- that may remain on the surface after the pretreatment.It will be convenient to define a new quantity x such that x = e-y. In termsof the quantities x, y and z, the rate equations for the reactions in eqn. (4) aredy/dt = - k i y ( - ~ + y ) , (7)&/dt - k3~(N- v - z -2~1, (8)dx/dt = -kz(x+y)(z+ x - y + U) -I- k3z(N- u - z - 2 ~ ) . (9)Dividing eqn. (7) and (9) by eqn. (8) to eliminate the time variable, we havedx x+y ( v + z + x - y )- 1, -- - u-dZ z ( N - V - z - 2 ~ )X - k Y Y - dy = p--dz z (N-v-z-2x)’where ci = k2/k3 and p = ki/k3.The graphs y against z and x against z for theobserved values of x, y and z are shown in fig. 11 and 12, and values of a and pwere estimated from the most linear portions. Expressing the values of all quan-tities in units of 1017 cm-3 of Hall sample, the values of the constants, and initial2FIG. 11.-Oxygen, y , as a function of carbon monoxide, for the experimental data given in fig. 8.The slope from the straighthe portion is 0.538. Units are in 1017 cm-3 of Hall sample.values in eqn. (10) and (1 1) for the data of fig. 8 are : a = 1.548, p = 0.538, u = - 5.8,iV = 5.8, yo = 2.1, zo = 4.3 and xo = 3.6.A computer solution was obtainedfor eqn. (10) and (1 1) for the above conditions ; the results are given by the solidcurves ir, fig. 13. The experimental data are given by the points in fig. 13. Al-though the theoretical results do not agree in all details with the experimental results,the agreement between the two is surprisingly good. The dashed curves give thetheoretical results obtained from a model in which the first reaction in eqn. (4) isreplaced by the reaction 02+20, a reaction that does not involve the conductio392 HALL EFFECT IN ZINC OXIDEelectrons. Although this second model gives a slightly better agreement with thedata than the first model, the choice of one over the other is not justified becausethere are important differences between the theoretical results and the experimentaldata that must be explained : the amount of oxygen remaining is a much more linearFIG.12.-The quantity x as a function of z for the experimental data given in fig. 8. The slopefrom the straight-line portion is 0548. Units are in 1017 ~ n - 3 of Hall sample.zFIG. 13.-A comparison between the experimental data of fig. 8 and the computed results from thetwo sample models given in the text. The com-puter results obtained for the model with the first reaction Oz+e +OT are given by the solid curveand the computer results obtained for the model with the first reaction 0 2 +20 are given by dashedThe experimental data are shown by the points.curve. Units are in 1017 CM-3 of Hall sample.function of carbon monoxide and the e against z curve has a much flatter minimumthan predicted by either model.Further studies are in progress to obtain morecomplete adsorption and reaction data so that an extensive kinetic analysis of thereaction can be madeH. CHON AND C. D. PRATER 393Some information as to the dependence of the reaction rate on the concentra-tion of electrons in the conduction band can be obtained from the data given infig. 9. Let us take the time interval to the beginning of the rising part of the electronconcentration curve as a measure of some kind of reciprocal average reaction rate.This time interval was approximately 30 min for the lithium sample and 2 min forthe indium sample. On the other hand, the initial value of the electron concentra-tion was 7.7 x 1016 cm-3 for the lithium sample and 1.46 x 1018 for the indium.The ratio of the time intervals is 15, and the reciprocal of the ratio of the initialcarrier concentrations is 19.This indicates a crude proportionality between theaverage reaction rate and the initial concentration of electrons in the sample.We may summarize by saying that the data presented strongly support the con-clusion that the electrons in the conduction band of indium-doped zinc oxide areintimately involved in the oxidation of carbon monoxide over this catalyst at 350°C.Furthermore, there is much evidence that 0- is the charged atomic species presentin greatest abundance on the surface during the reaction, and is also intimatelyinvolved in the oxidation reaction. The dual a.c. Hall method appears to providean accurate method for measuring the carrier concentration in polycrystalline,semiconducting catalyst samples with sufficient surface area for simultaneouskinetics and Hall effect measurements.The authors thank R. M. Lago for his assistance with mass spectrometer opera-tion, J. P. Gibbons for his help in carrying out the experimentation, L. F. Morgenfruhfor building the necessary glass apparatus, and Dr. J. C. W. Kuo for his help withthe computer calculation.1 Schwab and Block, 2. physik. Chem., 1954,1,42.2 Schwab, Semiconductor Surface Physics (University of Pennsylvania Press, Philadelphia, 1957),3 Morrison, Adv. CatuZysis (Academic Press, New York), 1955, 7, 259.4 Mehick, J. Chem. Physics, 1957, 26, 1136.5 Cimino, Molinari and Cramarossa, J. Catalysis, 1963, 2, 315.6 Glemza and Kokes, J. Physic. Chem., 1962, 66, 566.7 Chizhikova and Keier, Problemy Kinetiki i Katuliza, Akad. Nauk S.S.S.R., 1960, 10, 77.8 Bevan and Anderson, Disc. Furaduy Suc., 1950, 8,238.9 Volger, Physic. Rev., 1950,79, 1023.10 Koppe and Bryan, Can. J. Physics, 1951,29,274.11 Barry and Stone, Pruc. Roy. Suc. A, 1960, 355, 124.12 Kokes, J. Physic. Chem., 1962, 66, 99.13 Russell and Wahlig, Rev. Sci. Instr., 1950, 21, 1028.14 Pel1 and Sproull, Rev. Sci. Znsir., 1953, 23, 548.15 Dauphinee and Mooser, Reu. Sci. Instr., 1955, 26, 660.16 Bogner, J. Physic. Chern. Solids, 1961, 19, 235.17 Low, Chem. Reu., 1960, 60, 267.18 Weisz, J. Chem. Physics, 1953, 9, 1531.19 Wagner, J. Chem. Physics, 1950, 18, 62.20 Gensch and Hauffe, 2, physik. Chern. , 1951, 196, 427.21 Schwab and Rau, 2. physik. Chem., 1956,9, 127 ; 1958, 17,257.p. 283

ISSN:0366-9033    

DOI:10.1039/DF9664100380

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

GENERAL DISCUSSIOhlMr. R. C . Hansford (Union Oil Company of California) said: Dr. Rabo states thatthe catalytic activity of Cay and C e T zeolites for cumene cracking must be duedirectly to the metal cations and not to hydrogen cations that may be present as OHgroups after normal activation. This conclusion is based on the absence of infra-redbands, characteristic for OH groups bonded to silicon or aluminum, after treatmentat high temperatwes. However, such evidence is not sufficient to rule out thepresence of hydrogen ions, which if sufficiently free, would not produce absorptionbands in the infra-red region.It is not clear whether some hydrogen ions associated with the crystal frameworkin some way are involved in the catalytic properties of these catalysts. Weisz andMiale 1 present data for the cracking of hexane over a wide variety of cationic formsof the same zeolite and over a number of other zeolites containing calcium or hydrogencations.The activation energy for all types of cations and zeolites reported appearsto be about the same, as reflected in the Arrhenius plots of the data. If one acceptsthis conclusion, in spite of the considerable temperature range over which the com-parison is based, it would seem unlikely that the same activation energy for hexar,ecracliing would be obtained if the metal cation were directly responsible for catalyticactivity. Certainly, one would expect a difference in this kinetic parameter for suchwidely different cations as H+, Ca2+, and Ce3+.Prof. 2.G. Szabo (Budapest) said: Dr. Rabo has convincingly shown that thevariation of ionic field strength is connected with strong variation of catalytic activity.He attributed the observed effect to the ions of unusual valency built-in the zeolitepores. 1 think that the oxygen ions of zeolite are also markedly influenced by thesebuilt-in ions. It means that the electron shell of oxygen ions is strongly polarized,i.e., deformed. At the Amsterdam Congress a paper was presented in which thiseffect was explained by a very plausible hypothesis, viz., the less the oxygen ions of anoxide catalyst are deformed, themore effective it was in dehydrogenation and vice versa.The more the electron cloud was deformed, the more dehydrating the catalyst was.We have also prepared alumina catalysts; after sintering them at different tempera-tures, the whole spectrum of catalytic selectivity was observed.The ultra-violetreflexion of these substances show markedly differing spectra. These differentdeflexions point to the fact that the electron shells are, to various extents, deformed.The consequence of this polarization is that the same substance operates in one case asa Lewis-acid, in the other as a Lewis-base. The change in selectivity is thus a con-sequence of the changing deformation of electronic shells.Dr. T. I. Barry (N. P.L., Teddington) said : Dr. Rabo and his colleagues have shownthe importance to catalytic activity of local electrostatic fields due to cations in exposedSn sites. However, there is little experimental evidence to show how the cations aredistributed between the various sites especially when more than one type of cation ispresent.At the National Physical Laboratory, Mr. Lay and I have used electron spinresonance to reveal the sites occupied by Mn2f in X 2 and other zeolites when a varietyof other cations are present. We have interpreted the results on the assumption thatMn2f ions, which have no strong site preference, will occupy those sites not stronglybound by other cations.J . Catalysis, 1965, 4, 527, 2 Barry and Lay, Nature, 1965, 208, 13 12.39GENERAL DISCUSSION 395For X the following conclusions can be drawn : (i) Small ions, such as Li+ andMg2+, and especially ions which usually form tetrahedral, partially covalent bonds,like Zn2+, do not readily displace other ions from SI.(This result for Mg2+ andZn2f is contrary to the ideas expressed by Rabo et al.) (ii) Highly electropositiveions, especially those of Pauling radii just less than or equal to 1-33 A, e.g., K+ arestrongly bound to SI (large ions cannot enter SI). (iii) Large, especially monovalent,ions are weakly bound in SII sites, whereas small ions cannot readily be displaced fromthese sites. (iv) The distribution of cations may be slow to reach equilibrium;sometimes several days are required at 20-600°C. (v) The bonding of Mn2+ ions isusually ionic with only about 10 % covalent character. Under certain circumstancesthe covalent character increases to 25 %. This change may imply structural rearrange-ments or hydrolysis of the metal ions or the lattice.Dr.F. S. Stone (University of Bristol) said : With reference to the optical absorptionof NiIY and CurrY illustrated in Dr. Rabo’s paper, it would be interesting to know theform of the curves at wavelengths beyond 700mp. Have maxima been observed?Optical absorption of Cu*(2Es - 2T2g) in octahedral co-ordination in dilute solidsolutions CuO-MgO occurs with a maximum at about 850-900 mp (1 1,500 cm-1).1* 2Now the Cu-0 and Ni-0 separations at SI in Cu*Y and NiY are presumablygreater than the metal-oxygen separation in the close-packed lattice of MgO, favouringa weaker crystal field in the zeolite. For the 2Eg-2T2g absorption of NiI in NiY tobe correctly assigned as occurring at 650 mp (15,000 cm-I), as stated in the paper, itwould appear that very large effective point charges of the oxygens surrounding SI inNil’ have to be invoked.Prof.C. Kemball (Queen’s University, Belfast) said : Some interesting results wereobtained recently in Belfast by Dr. C . G. Pope with a nickel X zeolite. The maininvestigation involved the study of the catalytic exchange of benzene with deuteriumon samples of 13X sieve in which the sodium ions had been partially replaced by Ninbut some results were also obtained after a reduction of the catalyst. A nickel Xzeolite contains about 3 Nin/cage was treated with 2 cm of deuterium for 30 min at300°C and then cooled to room temperature before evacuation. This reductionprocedure gave rise to three effects : (a) the development of a pronounced bluish-green colour which differed from the pale green colour of the hydrated Nin; (b) thedevelopment of a very broad e.s.r.spectrum ; (c) a substantial increase in the catalyticactivity for the exchange of benzene and an even greater increase in the conversion ofbenzer,e to cyclohexane.We believe that the results may be explained by the formation of nickel atoms fromthe nickel cations. We have some evidence thzt aggregation of the nickel atoms toform small crystallites is not important at 300°C because the reduction of the catalystis very easily reversed. Evacuation of the hydrogen or deuterium at 300°C and out-gassing for 30 min destroys the green colour and the enhanced catalytic activitypresumably because the products of the reduction recombine inside the zeolite toreform nickel cations.Enhanced colour and e.s.r.signal were obtained after treatment with hydrogen at350°C but in this case the colour was not destroyed on evacuation. Some irreversiblechange had taken place in the zeolite because rehydration at room temperature gave agrey solid instead of the usual pale green colour. Aggregation of nickel occurs 3 at400°C under the action of hydrogen and a similar process may begin, at 350°C. Itis possible that the three effects which we observed after treatment at 300‘C are not1 Chapple and Stone, Proc. Brit. Ceram. Soc., 1964, 1, 45.2 Schmitz-Du Mont and Fendel, Monats., 1965, 96,495.3 Yates, J. Physic. Chem., 1965, 69, 1676396 GENERAL DISCUSSIONdue to a single cause-the colour and the catalytic activity might be a consequence ofone type of reduction, perhaps to NiI, and the e.s.r. signal might be the result of somedifferent reduction process.Dr.J. A. Rabo (Union Carbide Res. Inst., Princeton) said: The observable opticalspectrum cf the NiIY zeolite is restricted between 330 and 670 mp because of lightscattering by zeolite particles and because of the MgO coating used in the lightintegrator sphere.In reply to Kemball, the NiIIY zeolite usually contains small quantities of nickelmetal, which is easily shown by its ferromagnetism, therefore one has to consider thepossible contribution of the metal in any catalytic properties.In accord with Barry we agree that the simple electrostatic rule that we derived, onthe arrangement of the zeolite cations, from an X-ray study of Ca and Na exchangedfaujasite, cannot be applied quantitatively to ions which strongly differ from theseions; a different cation arrangement is indicated in this paper, by cation specificcarbon monoxide adsorption, for the large barium and for several of the open shellions (NP, Con).Our conclusion on the catalytic site in zeolites emerges from several observations :(i) the carboniogenic activity of alkali and earth alkali X and Y zeolites increaseswith cations of higher valence and smaller size ; (ii) at least 35 % calcium exchange isrequired to observe a sharp increase in activity, in good agreement with the expectedappearance of the calcium ions at surface sites at this exchange level; (iii) carbonio-genic activity is usually inversely proportional to the amount of OH groups.Theactivity is the same or improved even when 99 % of OH groups are removed.The Na2f centres have excellent chemical stability ; their 13 line e.s.r. spectrum isunchanged even at liquid helium temperature. Beside the Naz+ centres we have alsoprepared KZ+ centies in Y zeolite ; their e.s.r. spectrum shows 13 lines due to inter-action between the electron and the four potassium nuclei : the overall signal is abouttwo-thirds as wide as the Na:+ signal, in good agreement with theory.In reply to Peri, I do not think that the reversible CO band at 2140 cm-1 indicatescarbon monoxide associated with NiI ions because NiI tends to disproportionate,which should prevent reversibility.The high frequency infra-red band (2210 cm-1)observed when carbon monoxide is adsorbed on surface Nin ions is probably due toM-C-0 configuration in which the lone pair electron is polarized away from thecarbon while the lone pair on the oxygen is shifted toward the oxygen. This wouldincrease the triple bond character and the C-0 bond strength. The wall effectsuggested by Peri cannot explain the shift in C-0 stretching frequency because thiswould require a force constant of approx. 5 for the M-CO bond.Dr. R. L. Nelson and Mr. R. W. Wilkinson (A.E.R.E., Harwell) (communicated):Fig. 1 in the paper of Cimino, Schiavello and Stone1 indicates that on spectro-graphically pure MgO oxygen is adsorbed to coverages of -0.1 % monolayer at 25°Cand -0.4 % monolayer above 40O0C, at oxygen pressures of -0.4 torr.Theyconclude that both the extent and the rate of this adsorption are too great to beexplained by simultaneous oxidation of transition metal impurities in the bulk.2 Inour experience with spectrographically-standardized MgO, quoted as - 5 p.p.m.transition metals, careful analysis 3 has indicated impurity levels of -25 p.p.m.transition metals, precautions being taken to avoid contamination during handling.The limits of spectrographic analysis of many transition metals are sufficiently highfor the total transition metal level to be several times this latter figure.1 Cimino, Schiavello and Stone, this Discussion.2 Nelson, Tench and Wilkinson, Proc.Brit. Ceram SOC., 1965, 5, 181.3 analyses carried out by Analytical Chemistry Branch, Chemistry Division, A.E.R.E., U.K.A.E.AGENERAL DISCUSSION 397Work on highly-purified MgO 1 with all transition metals below the limit of spectro-graphic analysis, has indicated oxygen coverages of 0.009 % monolayer at -78"C,0.005 % monolayer at 25"C, and 0.02 % monolayer at 40O0C, using oxygen pressuresof 0.01 torr. These coverages are considerably lower than those quoted by Cimino,Schiavello and Stone, even allowing for the different oxygen pressures used. OnMgO doped with 120 p.p.m. Fe, oxygen is reversibly adsorbed to a coverage of 0.14 %at 25°C. Valence changes in bulk impurity ions are not observed during simultaneousoxygen and e.s.r. measurements at 25"C, but are observed after adsorption at 400°C.Similar results are obtained for Cr-doped MgO.The room-temperature oxygenadsorption is fast at 0-01 torr (mainly complete within 2 min) and is almost certainlydue to the presence of impurity ions, since it is virtually absent in the highly purifiedmaterial.Dr. K. Klier and Dr. K. Kuchynka (Inst. Physic. Chem., Acad. Sci., Prague) said:In order to obtain information on the activity of isolated Niz+ ions with respect tooxygen chemisorption and nitrous oxide decomposition, Cimino, Schiavello andStone employed the method of diluting the Ni2+ cations in MgO matrix. Theyobserved that the dilution leads to increased specific adsorption of oxygen but also toits increased reversibility as well as to increased activity in nitrous oxide decomposi-o a d s1tion.The complex involved is suggested to be the Ni3+-0;& array from whichoxygen desorption is likely to be easy.On desorption of oxygen the charge on the oxygen anions should be returned tothe lattice. Therefore, during such a reversible process one could expect to observe achange in physical properties (electric conductivity, optical spectrum, e.s.r.) due to thechange of valence state of nickel ions. None of these changes takes place duringreversible oxygen adsorption on nickel oxide 2-4 whereas profound effects are observedduring irreversible adsorption. Thus it appears that the otherwise acceptable mech-anism of N20 decomposition, proposed by Cimino, Schiavello and Stone, need notnecessarily lead to the conclusion that charged adsorbed atomic species are responsiblefor the phenomena of the reversible adsorption of oxygen, observed with the samecatalysts.Moreover, the catalytic decomposition of N20 on pure nickel oxide proceeds in atleast two distinct reaction steps.These two steps are (a) the decomposition of theN20 molecule with the formation of chemisorbed oxygen atomic anion and liberation ofa nitrogen molecule and (b) the recombination of oxygen anions to a neutral moleculewith simultaneous return of electrons to the oxide lattice. Now, it is observed thatwhereas the decomposition of N20 molecule is a relatively easy process taking placeon pure NiO at temperatures below 0°C even in the 10-2torr pressure range, thesubsequent liberation of 0 2 molecule from the surface is a much more difficult process,requiring considerably higher temperat~res.5~6 This would be consistent with thetemperature range where catalytic decomposition of N20 was measured in the Cimino,Schiavello and Stone paper.This distinction would not be so marked, were theanionic adsorption-desorption process so easy and reversible as described in the paperby Cimino, Schiavello and Stone.1 Wilkinson, to be published.2 Marcellini, Ranc and Teichner, Actes 2nd Congr. Irzt. Catalyses (Technip Paris, 1960), p. 289.3 Kuchynka and Klier, Coll. Czech. Chem. Comm., 1963, 28, 148.4 Klier, Kinetika i Kataliz, 1962, 3, 65.5 Hauffe, Glang and Engell, 2. physik. Chem. A , 1952,201,223.6 Kuchynka, Coll.Czech. Chem. Comm., 1965,30, 613, 622398 GENERAL DISCUSSIONFurther evidence that isolated Niz+ ions are inactive for chemisorption of oxygen(20°C) with charge transfer comes from our observation that dehydrated NiA syn-thetic zeolite, where Ni2+ ions are planar-coordinated to three oxygen anions andwell separated, does not exhibit a change of optical spectrum on exposure to oxygen.After rehydration in air one obtains a spectrum identical with that of the originalNiz+-A-hydrated zeolite.We suggest, on the basis of these results, that isolated Ni2+ ions can adsorb (20°C)neutral oxygen molecules but that at least two neighbouring Ni2+ ions are necessaryfor its splitting and charge transfer according to the scheme0- 0-0 2 - I 02- I/ \ I / \IAdmittedly, the situation in NiO+MgO crystals may be different to that on NiAsieves and NiO.We should therefore like to ask whether there is any positive evidencebased on conductivity or optical reflectance spectra or e.s.r. measurements (mentionedin the experimental part of the paper by Cimino, Schiavello and Stone) that chargetransfer does occur during reversible oxygen adsorption on NiO + MgO crystals.Dr. J. M. Thomas (Univ. COIL North Wales, Bangor) said : In the paper by Cimino,Schiavello and Stone it is stated that, since the quantity of oxygen chemisorbed revers-ibly at low temperatures increases with increasing temperature, thermodynamicequilibrium could not have been attained during the measurement of adsorbedamounts. As this so-called reversibly-held oxygen appears to play a dominant rolein the catalysis of the N20 decomposition, it would be of advantage to obtain quanti-tative values for the enthalpy and entropy of adsorption.I would therefore ask theauthors whether it would be possible and worthwhile to arrive at such data by extend-ing the time of measurement well beyond the 30 min used by them in obtaining eachpoint on the adsorption isobars ?Dr. F. S. Stone (Bristol University) said: In reply to Dr. Thomas, although theadsorption isobars for the reversible oxygen chemisorption in our paper refer arbi-trarily to amounts adsorbed up to 30 min, we have extended a number of the measure-ments to much longer times. There is a very small slow uptake beyond 30 min, buteven where adsorption has been followed for 12 h, thermodynamic equilibrium is notattained.The behaviour for MN 1 up to 3 h can be seen from fig. 3 in the paper.Lack of thermodynamic equilibrium is common in chemisorption on oxides, presum-ably because of successive stages with increasing activation energies. The only suremethod to obtain enthalpies of adsorption in these circumstances is by adsorptioncalorimetry, but this technique is difficult at high temperatures.Dr. I. D. Gay and Prof. F. C. Tompkins (Imperial ColIege) (communicated): Wehave studied the catalytic decomposition of N20 on pure NiO at low pressures (0-51-1torr) and comparatively low temperatures (22-140°C). After 10-20 min the reaction isstrictly first-order but log (rate constant) decreased linearly with log (initial pressureof N20) due to the poisoning of the catalysts by the irreversible adsorption of oxygenon about 0.05 % of the surface sites.The reversible adsorption of gaseous oxygenhad no effect on the rates. The kinetic results were well interpreted solely in terms oftwo reactions,N,O(a)+N,(g)+O(4 (1)(2) O(4 + N2O(g)+N&) + 02(g)GENERAL DISCUSSION 399both of which proceed at an essentially uniform group of sites after the initial 10-20min reaction ; in the initial stages, however, theactivation energies are controlled by thenumber of oxygen adatoms on these sites. Reaction (2) is rate determining, and E2,the activation energy, increases during the initial 20 min linearly with the total amountof oxygen adatoms left on the surface.The low value for the overall 1st-orderreaction of the activation energy of 2-3 kcal/mole is consistent with the mechanismproposed. Under our experimental conditions, non-equilibrium processes thereforemake a significant contribution and the mechanism is different from that whichprevails at higher temperatures ( > 300°C).Dr. J. Dewing (I.C.I., Ltd., Runcorn) (communicated) : Dr. Stone has shown that inhis series of catalysts the one with the strongest oxygen adsorption has the highestArrhenius activation energy for nitrous oxide decomposition. Dr. Cvetanovit and Ipublished results of experiments on the decomposition of nitrous oxide on nickel oxideat low pressures and at conversions below 1 x.1 In the experiments the catalyst was inequilibrium with the reacting gas phase in a flow system and the partial pressure ofoxygen over the catalyst was below 10-6 torr.We found activation energies below10 kcal mole-1 and that doping the catalysts with lithium reduced the activation energyfurther. The low activation energy for nitrous oxide decomposition over nickel oxideat low pressures has been confirmed qualitatively by Winter.2It is to be expected that a sparsely covered surface, as used in our work, would havea higher differential heat of adsorption for oxygen than a similar surface reactingunder higher pressures of nitrous oxide and oxygen. Doping with lithium is nowknown to increase the strength of chemisorption of oxygen on nickel oxide. Wetherefore conclude that, under conditions of low pressure of both reactant and productoxygen, high heat of adsorption of oxygen is correlated with low activation energy forthe catalytic decomposition of nitrous oxide.This is directly contrary to Dr. Stone’sfindings.The explanation for the difference may lie in the higher oxygen coverage of thesample MNL 1 : 1 during reaction. If the reaction OCCUIS preferentially on a fewsites which compete for N20 and 0 2 the heat of desorption of 0 2 would contribute tothe overall Arrhenius activation energy in the form of an increase in the effectivesurface area with temperature. This assumes we are dealing with reversible oxygenadsorption on catalytically active sites. If this is so, the difference in activationenergy may change as the pressure at which the decomposition is studied is decreased.One may even predict that at very low pressure and low degrees of decomposition thelithium-doped catalysts will show lower activation energies than the undoped catalysts.Prof.A. Cimino (Uniuersity of Perugia) and Dr. F. S. Stone (University of Bristol)(communicated): We are aware of the study made by Dewing and CvetanoviC, andtheir results are quoted in the article by Cimino, BOSCO, Indovina and Schiavello 3 towhich we refer in the paper. At low N20 pressure the dpparent activation energy Eis much lower than otherwise obtained, and Winter 4 gave clear evidence that E forN2O decomposition over NiO contains a large term due to oxygen desorption. Inagreeing, therefore, with the last paragraph of Dewing’s remarks, we would particular-ly stress the difference in the conditions prevailing in our experiments (PN20- 60 torr)and those of Dewing and CvetanoviC 10-4 to 10-5 torr, 1 % decomposition).If appropriate to our specimens, the data of Nelson and Wilkinson on similarspectroscopically standardized MgO almost disposes of the need for an explanation of1 Dewing and Cvetanovib, Can. J.Chem., 1958, 36, 678.2 Winter, Disc. Faradgy Sac., 1959, 28, 183.3 Cimino, Bosco, Indovina and Schiavello, J. Catalysis, 1966,5, 271.4 Winter, Disc. Faraday SOC., 1959, 28, 183400 GENERAL DISCUSSION“ intrinsic ” oxygen chemisorption on pure oxide such as we have suggested. Ourconclusions, however, regarding specific chemisorption on nickel ions are unaffectedsince they depend on comparisons between specimens prepared from the same batchof magnesium oxide.We would agree with Klier and Kuchynka that charged atomic species have notbeen proved responsible for the reversible oxygen adsorption discussed in our paper,and our reference to 0- in outlining the mechanism was mainly schematic.Thenumber of sites involved is only a small fraction of the total and would not necessarilylead to observable changes in the physical properties. Such physical measurementsas we have made were not carried out in situ, but before using the solid as a catalyst.As for the problem of oxygen desorption and its role as a possible rate-determiningstep, it may not be correct to assign the same rate-determining step at both low and hightemperatures. Moreover the reversible adsorption of N20 reported in our papershows that N2O decomposition is not as easy as suggested.Prof.S. Z. Roginskii (MOSCOW) said: The true nature of the relation betweenmacroscopic electronic properties of oxides and their catalytic activity is of interest forthe electronic theories of catalysis. There are serious discrepancies in various resultsobtained by different authors for the same systems, in particular in those reported byTeichner and others and the conclusions made earlier by Schwab, by us, and byParravano and others. The difficulty of obtaining identical surfaces provides only apartial explanation. Complete insight might be obtained by additional experimentsthat would permit differentiation of the parts played by macroscopic and microscopicelectron parameters, and by means of a more critical treatment of the too simplifiedmodels used by most authors of the electronic theories of catalysis.The error theseauthors make is to believe that the mechanism of adsorption and catalysis on semi-conductors is simpler than for metals. Most unexpected is the correlation oftenobserved between the chemisorption and catalytic properties of semiconductors andof the forbidden band width.1Interesting possibilities seem to be opened up by the work of Cimino, Stone andChiavelli who compared individual nickel and magnesium oxides with their solidsolutions. We are working with similar systems and the results obtained also showthat the catalytic properties of nickel ions in solid solutions are qualitatively the sameas in individual compounds.This is an argument against the premise that thedecisive part is played by macroscopic electronic properties, and also against thepcssibility of a simple correlation between electroconductivity and catalytic activity.Determinations of the work function 4 lead to the same conclusion. The sign of thecharge changes of surfaces by the same molecules chemisorbed on p - and i- semi-conductors is the same as for n-semiconductors.2Several years ago I proposed a simple thermodynamical explanation of the linearrelation between the changes of the activation energies for chemisorption and catalysisand the changes of the work function caused by doping of solids.3 When thetransition complexes have a positive charge, the energy of their formation willdecrease linearly with increasing $(EA+ = Eo--aA4), while the reverse would takeplace for negative complexes. These effects provide an explanation for manyphenomena observed in the “modification” of catalysts and for the influence ofadditives on complex (parallel and consecutive) processes.The above is true forvarious mechanisms of adsorption, and for the whole diversity of models used in1 Krylov and Fokina, Kinetika i Kataliz, 1964, 5, no. 2. Roginskii, Acfes 2 Congr. Cafal., 1961,2 Enikeev, Roginskii and Rufov, Pruc. Int. Conf. Semiconductor Physics, (Prague, 1960).3 Rogjnskii, Doklady Acad. Nauk U.S.S.R., 1959, 130, 122.2, 1536GENERAL DISCUSSION 401theories for semiconductor catalysis.The coincidence or discrepancy may showwhether the Fermi level position is of importance in the given case, or not.Dr. J. Vi)lter (Institut anorg. Katalyseforschung, Berlin-Adlershof) said : Recentlywe studied the hydrogenation of benzene on metallic Ni, which was supported by Li+doped MgO.1 Though the reactions are different the catalysts are similar to thoseused by Stone ; we had 1 % Ni instead of 1 % NiO. We have found that doping withLi+ (and Ga3f) is ineffective for hydrogenation on Ni/MgO. Activation energy andspecific activity are unaffected. This is in good agreement with the results reported byStone and supports his conclusions that the link between catalysis and semiconduc-tivity is much less direct than expected some years ago.The influence of Li-dopingwe find is a drastic diminution of surface area of MgO. The same effect is nowreported by Cimino, Schiavello and Stone.Dr. C. D. Prater and Dr. H. Chon (Socony Mobil Oil Co. Inc., Princeton) said:We should like to question the firmness of the conclusions of Amigues and Teichnerthat (i) their data are inconsistent with the conclusion that 0- directly participates inthe oxidation of CO for the conditions in their experiments, and (ii) that competitiveadsorption between CO and 0 2 is required to explain this data. These two conclu-sions concern the form of the function 00 in eqn. (60) of their paper. We shall seeka function f, operationally indistinguishable from 80, that involves 0- instead of 0.Eqn.(60) is used to explain the changes in slope, occurring at about 75 % conversion,that are observed in curves B and C of fig. 9. Two kinds of interactions are involved-in one, surface oxygen interacts with gas phase CO (Rideal mechanism); in theother, surface oxygen and adsorbed CO interact as a two-dimensional surface gas.Hence, iff and 80 are operationally indistinguishable, the mechanism involving 0-is as valid for use in eqn. (60) as the one involving 0.With regard to 80, the analysis of Amigues and Teichner is essentially qualitative,and relies heavily on the following information : (i) the partial order of the reactionwith respect to oxygen is 3, and (ii) with respect to CO is zero ; (iii) an " accelerating "effect of oxygen is observed, as shown by a comparison of curves A, B, C, and D offig. 9 ; (iv) an " inhibiting " effect of CQ is observed, as shown by the small changesin the relative positions of curves A, F, and E of fig.9 ; (v) curve A of fig. 10 showsonly a small change in conductivity during the course of the reaction. In the analysisof their data, their assumption is that the adsorption steps are in equilibrium; thevalidity of this assumption is not demonstrated, however.For their data we shall give a counter-example that reproduces the essential featureson which the above conclusions are based. We examine the mechanismk iO2 +e+O,0, +e+20-O-+CO+CO,+e.For completeness, we have added the reverse steps to the first and second reactions,but we shall not assume that these steps are at equilibrium.Let the values of kz andkl be equal to the values of CI and p given in our paper. Let k3 = 10 and the equili-brium constants, K1 = k-l/kl and K2 = k-Z/kz, have the values 10.9 and 0-076, asobtained in our experiments. Let the free volume of the reactor be five times thatused by us, but let all other conditions be the same. The amounts of 0- and 0; arek- ik2k - 2k3f Volter and Kuhn, 2. anorg. Chem., 1965, 340, 216402 GENERAL DISCUSSIONeliminated from the rate equations by using mass and charge conservation equations ;no steady-state assumptions are used. The resulting rate equations, in terms of ournotation, ared y Id t = ( k I V ) [ K (u - 2y + z + e ) - ye],dz/dt = ( - kJV)(N, - v + 2y - 2e - z),(deldt) = ( k J V)[K I( u - 2y + z + e ) - ye] + k , [K2( N , - v - 2y - 2e - z ) ~ -e(v-2y + z + e)] + ( k 5 / V)z(Ne- v -I- 2y - 2e - z).The new quantity Y that appears in the rate expressions, is the ratio of the freevolume in the reactor, here assumed to be that used by us.-FIG.1 .-(a) Results from computer " experiments " at constant total pressure and varying carbonmonoxide to oxygen ratios. The values of the carbon monoxide to oxygen ratios and the designationsof the curves are those given in fig. 9 of the paper of Amigues and Teichner.Computer solutions were obtained for the rate equation for three cases. In thefirst, the CO to 0 2 ratio was varied, without changing the total pressure; in thesecond, the 0 2 partial pressure was varied without changing the CO partial pressure ;and, in the third, CO partial pressure was varied without changing 0 2 partialpressure.For the first, the CO to 0 2 ratios have the values given in fig. 9 ofAmigues arid Teichner. The results obtained from the computation are shown infig. l a and lb. The conversion results given in fig. la show the " accelerating " effectof 0 2 and that only sniall changes from curve A occur for the CQ-ratios used to obtaincurves E and F. It was possible to obtain the particular spread between curves A, E,and F seen in fig. 9 by making a change in the values of the rate parameters usedin the computation. Thus, the accelerating and inhibiting effects, points (iii) and (iv)above, are produced.The computed electron concentrations are shown in fig.lb. Allowing for thedifferences between conductivity and bulk electron concentration for this sample, andfor the differences in scales used in fig. 10 and 11 and in fig. lb, there is considerablGENERAL DISCUSSION 403similarity in the forms of the curves. Curve A of fig. l b has at least as great a resem-blance to curve A of fig. 10 as does curve A in fig. 11 of Amigues and Teichner. Thus,the agreement with point (v) of the results obtained from the counter-example seems tobe as satisfactory as the results obtained from the mechanisms used by Amigues andTeichner .IFIG. l.-(b) Electron concentrations for computer " experiments " of fig. l(a).timeFIG. 2 . 4 4 Computer " experiments " at constant carbon monoxide partial pressure and varyingoxygen partial pressures.The curve designated 1 X is for the stoichiometricratio ; the curve designated2 X has twice the oxygen partial pressure, and that designated 3 X has three times the oxygen partialpressure of the stoichiometric ratio.The results obtained from variations of oxygen partial pressure at constant COpressure are shown in fig. 2a. The values of the oxygen partial pressure are once, twiceand three times the stoichiometric amount. The slopes of the rapidly rising part ofthe curve in fig. 2a are in the ratio Jl : J2 : J3. Hence, the reaction order wit404 GENERAL DISCUSSIONrespect to oxygen is 3. The results obtained from variations of CO partial pressure atconstant 0 2 partial pressure are shown in fig.2b. Clearly, the reaction order withrespect to CO is approximately zero. Thus, we have reproduced points (i) and (ii)cited previously.r------timeFIG. 2 4 b ) Computer " experiments " at constant oxygen partial pressure and varying carbon qon-oxide partial pressure. Curve 1 X is for the stoichiometric ratio ; curve 1.5 X has one and a halftimes the carbon monoxide partial pressure and curve 2 X has twice the carbon monoxide partialpressure of the stoichiometric ratio.In generating our counter-example, we have neither sought the mechanism thatgives the best fit to the data presented, nor sought to account for every aspect of theirdata. Nevertheless, the ease with which the above counter-example was obtainedand the extent to which this crude attempt " fits '' the data, show that a more extensivescreening of other possible mechanisms is necessary before the conclusions of their con-tribution given in the first paragraph, can be firmly established.The necessity fordoing this becomes even more important when we consider the possibility of a build-upof a charged boundary layer under dynamic, non-equilibrium conditions. Thekinetic consequences of such possibilities must be considered when screening mechan-istic interpretations. Therefore, we conclude that, although the mechanism usedby Amigues and Teichner may be sufficient to explain their data, it is not unique;and mechanisms involving direct participation of 0- in the reaction are not ruled out.Prof. S.J. Teichner (Inst. Recherch. Catalyse, Villeurbanne) said: In reply to Prater,the reaction with stoichiometric mixtures (CO + 4 0 2 ) assumes an overall order of 3.This has been established by the differential method by plotting log (-dP/dt) as afunction of log P (P is the total pressure of the stoichiometric mixture). The slope ofthe straight line was 3. This overall order is verified by the integral method (eqn. (44)of my paper) which gives a linear plot for the whole of the reaction. The partialorders with respect to 0 2 and CO were determined in the following way. When thestoichiometric mixture had proceeded to 50 % the addition of an extra amount of 0 2increased the rate. Under the same conditions the addition of an extra amount of COdoes not modify the rate.Since only 0 2 accelerates the reaction, the overall order 3is correlated topo, and a partial order zero is indicated for COGENERAL DISCUSSION 405The simplest conclusion which can be deduced from these results is that of aLangmuir mechanism between the species adsorbed on different sites. However, fornon-stoichiometric mixtures, with an excess of 0 2 (curves B and C of fig. 9) anacceleration of the reaction is observed once 75 % of the mixture has reacted (a kneeon curves B and C). For a non-stoichiometric mixture with an excess of CO thedifferential method by plotting log ( - dP/dt) as a function of log poz (gas which is notin excess) gives a straight line whose slope is 0.7. The rate of the reaction decreasestherefore more rapidly when the reaction proceeds than would be predicted by theorder 3 with respect to 0 2 .A CO excess has then an inhibiting effect which is recordedfor pc0/po2 >2 and not for the stoichiornetric mixture.In conclusion, a competitive adsorption of CO and 0 2 must occur as has alreadybeen observed by conductivity measurements described in the paper. Eqn. (60) and(62) account for this phenomenon. But in order to observe the total order 3 forstoichiometric mixtures, the only two possibilities are that either CO reacts from thegas phase by a Rideal mechanism, or it reacts when adsorbed on different sites thanthose adsorbing oxygen.Now, considering the nature of adsorbed and reacting oxygen (0- or 0), the order3 with respect to 0 2 is compatible with the formation of 0- as a reactive species, e.g.,according to + 0 2 + e-+O;&, but in this case the coverage by 0- and therefore the ratewould depend onpo,.But the coverage by 0- is practically constant over the pressurerange 1.0-0.01 torr of 0 2 and it is over this range that the kinetic expressions should beconsidered. As a consequence, a partial order of zero with respect topo, should befound. It can be seen from fig. 3 that for poz = 1.0 torr, loga = -5.5 and forpo2 = 0.01 torr, log Q = -4.5, log 00 being equal to -2.5. According to eqn. (12)the coverage by 0- is practically unaltered in this pressure range: [e,], whichrepresents the concentration of free electrons which can be used in the adsorption ofoxygen to form 0-, is decreased by adsorption of 0 2 at 0.01 torr to 1/100 of its previousvalue ([e-]/[e;] = 0.01) and is decreased by adsorption of 0 2 at 1 torr to 1/1OOO of itsprevious value ([e-]/[e,] = 0.001).Therefore the concentration [O-]cc(l-O~Ol) atpo2 = 0.01 torr and [0-]cc(1-0~001) at poz = 1 torr. In the range of the pressurechange during the reaction with the stoichiometric mixture, or with excess oxygen,[Q-] is then practically the same and cannot therefore depend on&,. For this reasonthe uncharged species, which we write as 0, whose concentration on the catalystdepends on ph,, and does not depend on [e-1, is considered tobe the reactivespecies.This does not mean that 0- cannot react with CO. In fact, this is what happenswhenPo, is very low, e.g., when the evacuated catalyst having pre-adsorbed oxygen issubmitted to interaction with CO(g) (see the paper), or near the end of the reactionperformed with a non-stoichiometric mixture containing an excess of CO.Thefinding that for relatively high oxygen pressure and therefore low electron concentra-tion [e-1, the reaction does not proceed through 0- species but through less chargedor uncharged species is confirmed by fig. 10 of the paper by Dr. Prater and Dr. Chon.To sum up, if the participation in the reaction of different oxygen species is repre-sented as a function of oxygen pressure in the gas phase according to the scheme :OCrrev.) I O,,", 0-1 ,-I- I 3 increasing 0 2 pressureA I B I Cthe range in which the reaction is carried out by Dr$Prater and Dr.Chon is A andperhaps B, while in our work it is the range C . For this reason it is 80 and not 60-which is intervening in the eqn. (60)406 GENERAL DISCUSSIONFinally, if the variation of [0-] (and of [e-I) is very small in the PO, range from 1 to0.01 torr compared with the total amount of 0- fixed (which enables to consider [0-]as constant) this variation is readily seen from the variation of conductivity with po2(fig. 3). This allows us to follow conductivity changes during the reaction (fig. 10)and to recalculate these by eqn. (52) in fig. 11, for which the variation of o is neces-sarily connected with the variation of [0-] (and of [e-I).Dr. J. Haber (School of Mining and Metallurgy, Krakow) said: On admission ofsmall amounts of oxygen Amigues and Teichner observed first a decrease of theconductivity of ZnO and then its increase (fig. 4).The phenomenon is similar to thatobserved earlier by Kokes, but the interpretation is different. The authors explainthe conductivity against time curve in terms of the difference in the rates of adsorptionof 0- and 0 species (eqn. (24) and (25)). It seems, however, that the rate of a thirdprocess must also be taken into account, consisting in the transition of 0- to 0accompanied by the transfer of electrons to the solid. This process will equilibriatesurface concentrations of 0- and 0 in the course of adsorption. The increase of theconcentration of 0 should thus result in the increase of the concentration of 0-followed by the decrease of conductivity instead of its increase as observed in theexperiment.Results of our experiments on photoeffects at zinc oxide surfaces seem to indicatethat such changes of conductivity with time as shown in fig.4 may be rather due tothe incorporation of oxygen. We have observed that photodesorption of oxygen isreversible at temperatures up to 4OO0C, accompanied by an irreversible process atstill higher temperatures,l whereas photoadsorption is irreversible even at lowertemperatures. These results indicate that outward diffusion of interstitial zinc ionscan proceed at a measurable rate already at low temperatures, whereas their inwarddiffusion begins only at higher temperatures. This may be related to the potentialdifference in the barrier layer, accelerating the diffusion of zinc ions to the surface.There are two other points which may be mentioned in connection with the surfaceproperties of zinc oxide.Dere4 and Fryt 2 in our laboratory have shown that ondoping of zinc oxide with lithium the conductivity changes from n to p-type. Theyexplain this result by assuming that lithium ions are incorporated substitutionallywith one half of substituted zinc ions being moved to interstitial positions and thesecond half forming new layers of ZnO lattice. On the other hand we have found ondoped nickel oxide samples that on increasing the concentration of lithium the orderof the reaction of CO oxidation with respect to 0 2 and CO changes considerably.These facts indicate the complexity encountered by comparing the behaviour of pureand doped samples.Prof.R. L. Burwell (Northwestern University, Evanston) said: Amigues andTeichner report that exposure of zinc oxide to CO at above 180°C leads to the appear-ance of C02 in the gas phase. They ascribe formation of C02 to reduction of zincoxide. May I enquire whether the authors demonstrated that loss of CO was equalto gain in COz? Some years ago Prof. Taylor and I observed the same phenomenonat about the same temperatures 3 although the zinc oxide was prepared differently,by ignition of the oxalate at 400°C in air. We accumulated considerable evidencethat formation of COa resulted, not from reduction of zinc oxide, but from dispro-portionation.2 c o = c02+ [C]. (1)a1 Haber and Kowalska, hull.Acad. Polon. Sci., ser. sci. chim., 1965, 13, 463.2 Dereh and Fryt, Bull. Acad. Polon. Sci. ser. x i . chim., in press.3 Burwell and Taylor, J. Ameu. Chem. Soc., 1937, 59, 697GENERAL DISCUSSION 407On continued treatment with CO, the adsorbent became gray-brown. The colourchange became obvious only after some hours. The original white colour of the oxidewas restored by treatment with hydrogen by a process which we suggested involvedconversion of carbon to methane. The carbon was clearly not bulk graphite butprobably represented small atomic clusters which may almost be considered to beadsorbed carbon. We suggest that C02 formation in the experiments of Amiguesand Teichner may also have resulted mainly from reaction (1).Prof. S.J. Teichner (Inst. Recherch. Catalyse, Villeurbanne) said: In reply toBurwell, it does not seem that the disproportionation of CO occurs on ZnO at 261°Cduring the catalytic oxidation of CO. After several experiments carried out on thesame sample of ZnO (evacuated between the experiments) no C deposit was found onthe solid. In reply to Haber, when ZnO is doped with Li the n-type conductivity isconserved. Thep-type conductivity is only observed with ZnO in oxygen, at tem-peratures not exceeding 150°C (enrichment layer). The order of the reaction (CO +$ 0 2 ) is not modified on the Li- or Ga doped-catalysts used throughout this work.I do not sce the point relative to the virtual pressure of 0-. The electrons are pro-vided by the solid and the pressure function (see the paper by C .Kemball), for 0; aswell as for 0, is p&. This is implied in the kinetic expressions (24)-(27) of my paper.In reply to Amberg, metallic zinc may be formed by activation of ZnO in vacuum ata temperature exceeding 5OO0C,1 and by interaction with CO at 261 "C if pco = 120torr, but not for pco = 3 torr. When metallic zinc is formed the activity in the cata-lytic oxidation of CO decreases? In reply to Chon, the conductivities and activationenergies of conductivities of Li- and Ga-doped ZnO used in my work are summarizedin the table.samplea(ohm cm)-1 [a2500Cin oxygen (160 torr)1 5Oodev)a ( o h Cm)-1 [a2500Cin vacuum (10-6 torr)at 250°C 1 5Oodev) at 250°CZnO(Li) 5-68 x 10-4 0.24 8.3 x 10-11 1-15ZnO 2-26 x 10-2 0.066 1 .o ~ 10-9 1 -09ZnO(Ga) 3.18 x 10-2 0-065 1.OX 10-8 0.67In reply to Rabo, the vacuum-activated ZnO is not transparent to i.-r. even for athickness of 2 mg/cm2. This results from a high electron concentration of the solid.The re-oxidation of the sample gives an i.-r. transparent material but no OH bands arefound.Dr. H. Chon and Dr. C. D. Prater (Socony Mobil Oil Co., Inc., Princeton) (com-municated) : Teichner's conclusion that the doping does not seem to have any influenceon the reaction requires some reservation, because his samples show only smallchanges in semiconductivity after his addition of gallium. The conductivity ofZnO(Ga) in a vacuum (3.18 x 10-2 ohm-1 cm-1) given in his remarks is not muchdifferent from his value for pure ZnO (2.26 x 10-2 ohm-1 cm-I), in spite of theaddition of fairly large amounts of gallium (3 at.%). On the other hand, our HallEffect measurements on ZnO show that the carrier concentration of ZnO at roomtemperature can be increased nearly 100-fold by doping with 0.01 at. % of galliumor indium.Dr. Y. Fujita and Prof. J. Turkevich (Princeton Uniuersity) said: We have examinedthe e.s.r. signals of ZnO degassed at 500°C for 6 h and then treated with oxygen, oxidesof nitrogen, discharge in oxygen or hydrogen and irradiated with 2537L$ Hg lamp.We have also irradiated H202 in Vycor glass and obtained an e.s.r. signal analogous1 Archiropoulos and Teichner, J. Catalysis, 1964, 3, 477.2 Amigues, Thesis (Lyon, 1964)408 GENERAL DISCUSSIONto one found on ZnO. The zinc oxide obtained from New Jersey Zinc Co.andevacuated at 500°C and cooled in a vacuum gave two signals at g = 1.96 (6 gausswidth) and g = 2.003 (4 gauss width), (fig. 1). The intensity of these signals increasednearly according to Curie law as the sample was cooled to liquid-nitrogen temperature.FIG. 1.-E.s.r. spectrum (X-band) of zincoxide evacuated at 500°C. and cooled in avacuum ; 1 in. = 50 gauss on horizontalaxis.AFIG. 2.-E.s.r. spectrum (X-band) of zinc oxide evacuated at 50O0C, and cooled in vacuum andtreated with 10-3 mm of 0 2 at room temperature. The spectrum was obtained at liquid-nitrogentemperature. (a) 2db, (b) 5db, (c) IOdb, ( d ) 15db. 1 in. = 50 gauss on horizontal axis. Amplifica-The signal width at g = 2.0 did not change. These lines are similar to those observedby Kokes 1 and Sancier.2 We assign the g = 1.96 signal to Zn+ present in interstitialposition in the lattice, the g = 2.003 signal to an electron trapped on an ensemble ofoxygen atoms on the surface.We do not favour assignment of the g = 2.003signal to 0- because that should give an anisotropic signal for 0- in the highlyasymmetric crystal field on ZnO as it does for the 0;.tion is adjusted so as to show nearly same intensities.1 Kokes, J . Physic. Chern., 1962, 66, 99.2 Sancier and Freund, J. Catalysis, 1964, 3, 293GENERAL DISCUSSION 409Treatment with 10-3 mm 0 2 at room temperature produces a triplet with g = 2.05,g = 2.010 and g = 2.003. This signal is stable for months.Saturation studies andchemical treatment show that this triplet is due to one species (fig. 2). Oxygenbroadens the signal. We assign this signal to 0;. This signal is also produced bypassing a discharge through a sample of ZnO containing 10 mm Hg of 0 2 . Vycorglass treated with H202 irradiated with U.V. light also gives a signal stable for dayswith gl = 2.030, 9 2 = 2.014, and 9 3 = 2.0023 similar to the 0; observed on zincoxide (fig. 3). Irradiation of this ZnO sample with ultraviolet light decreases theintensity of this triplet. Heating to temperatures over 200°C changes the triplet to the2.003 singlet. Treatment with NO at 220°C produces the increased singlet at g =2-003 but no triplet, (fig. 4).4 4(lz = 2 . 0 3 0 2 YFIG.3.-The e.s.r. spectrum (X-band) of 0,radical produced on irradiation with U.V. light,at liquid nitrogen temperature, the Vycorglass treated with H202. The spectrum isobtained at room temperature.FIG. 4.-Change in the e.s.r. spectrum (X-band)of zinc oxide evacuated at 5OO0C, cooled invacuum, treated with 10-3 nun of NO andthen heat-treated at 220°C for 10 min.FIG. 5.-E.s.r. spectra of ZnO pure at 4.2% (a), and 1~9°K (b) : 20db, 1 in. = 100 gauss. (MiddleHydrogen does not change either signal at g = 2.0 or at g = 1 -96 at room tempera-ture and 1Omm pressure. After 5 min discharge in hydrogen the sample becameconductive. The signal at g = 1-96 could be observed while that at g = 2 could notbecause of bad signal-to-noise ratio. At liquid-helium temperature the signals atg = 1.96 and g = 2-003 were observed but at 1.6"K both disappeared (fig.5). Weoffer the following interpretation of the interaction of zinc oxide with oxygen. Onheating, zinc oxide decomposessignal is attributed to the defect of quartz.)2Zn0 +2Zn + 0 2 410 GENERAL DISCUSSIONThe interstitial zinc ionizesThe Zn+ gives the signal at g = 1.96. The free electrons are not detected by e.s.r.Those trapped on oxygen ensembles on surface given signal at g = 2-003. Onadmission of oxygen to the sample the free electron attaches itself to 0 2 to form 0;.On heating this dissociates to give an electron trapped on the oxygen ensemble onsurface with singlet signal at g = 2-003.Prof. C. H. Amberg (Carleton University, Ottaiva) said: There are several appar-ently conflicting views regarding the ZnO-catalyzed oxidation of carbon monoxide.I hope I shall not materially add to the confusion by asking some questions.First,if one can accept Turkevich's statement that peroxide ions (0;) are formed on theZnO surface during outgassing, then this might provide an alternative explanation toCO being oxidized on highly outgassed ZnO at room temperature, which was reportedin an earlier discussion.1 We interpreted this at the time as being due to the highconcentration of conduction electrons created by surface oxygen removal. Thedecay of the oxidation reaction might then be regarded in terms of the depletion of theperoxide ions. Secondly, is it not conceivable that the differences reported byTeichner et al.and Chon et al. on the subject of CO oxidation reside in the widelydifferent methods of sample preparation? The surface areas of the catalysts differedby a factor of approximately 150, with Amigues and Teichner using the higher surfacearea and thus the more highly defective oxide. We have all on occasions observedmetallic zinc depositing in our cells after prolonged outgassing. This is undoubtedlydue to the solubility of interstitial zinc in the zinc oxide lattice being exceeded. I ambecoming convinced that eventually many of the high-area zinc oxides in fact becometwcP-phase systems. Thus some of the secondary slow oxygen chemisorption stepsmight well be interpreted in terms cf the re-oxidation of the metallic zinc " puddles "which constitule the second phase.I should like to ask whether the possibility existsthat these re-oxidized regions enter into the catalytic process. One might expectconductivity measurements not to encompass in a simple manner the charge transferprocesses taking place at the surface of such regions.Dr. L. Degols (Ecok Roy. Militaire, Belgium) said: The evaluation by Volger ofthe Hall coefficient for polycrystalline samples composed of particles whose con-ductivity is higher in their internal part (bulk)-region 1- than in their surface region-region 2-results in the following equation :Zn +Zn+ + e.where 12/11 is the ratio of the dimensions of the two regions.Volger points out that, although A$!)<&), the first term is prominent in the casehe considered, because of the low value of Z z / l ~ .This statement may be suspectbecause of the low dimensions of the particles ( N 104 A for a surface of 0-1 m2/g).On the other hand, according to the electron theory of catalysis, adsorptionphenomena do not exert any influence on the bulk properties of the catalyst. Thus,the assumption that the variations of Hal1 coefficient in the course of adsorptioncorrespond to modifications of the bulk carrier density seems hazardous. A contribu-tion of the last term of Volger equation seems to provide a better explanation. Thevariations of the Hall coefficient should result from (i) the niodification of the lengthof the surface barrier and (ii) the modification of the density of the carriers in it.For further progress, complementary data are thus necessary, e.g., measurement1 See discussion in Amberg and Seanor, Proc.3rdlnt. Congr. Catalysis (North Holland Publ. Co.,Amsterdam, 1965), p. 450GENERAL DISCUSSION 41 1of the Hall effect at different frequencies of the injected current would be useful. Forsolids whose conductivity in the surface region is lower than in the bulk, it is commonlyadmitted that the barrier Iesistance is, at high frequencies, short circuited by the barriercapacitance. The limiting frequencies of the region of dispersion can be as low as10-4 Hz and as high as 10+10 Hz (paper to be published shortly).If the Hall coefficient depends only upon the bulk properties, no variations of thiscoefficient should be observed as a function of frequency.In the contrary case, thesemeasurements provide important information concerning the variations of the barrierduring adsorption.i n addition, it is quite possible that the type of barrier studied by Volger and thushis equation are not applicable in this case. Supplementary theoretical work is there-fore necessary. In any case, conclusions concerning the C0+02 mechanism are tobe considered with caution.Dr. H. Chon and Ds. C. D. Prater (Socony M. Oil Co. Inc., Princeton) said:Concerning the comments of Degols, we offer the following reply. Our experimentson the effects of volume shrinkage show that changes in particle size and in the surfaceregions responsible for most of the resistance at low frequencies do not change the2l-I10-90 -r(I20-I.GEQ 10- s c, .- > .dc1 y 05a0020. I -LA 10 1-30 2 41/Tx lo3FIG. 1.Hall voltage for the range of sample porosities used in our experiments. Two samplescontaining the same amount of gallium impurity were subjected to different preparationprocedures so that different degrees of volume shrinkage were obtained. Theseprocedures give different particle sizes and different grain boundary effects. Th412 GENERAL DISCUSSIONresults are presented in the accompanying figure. These two samples gave essentiallythe same Hall voltage, but differed by a factor of six in resistance. Furthermore, themethods gave results for the totul number of donors in the sample that agree within20 % with the number of atoms of impurity in the sample, as measured by spectro-scopic techniques. This indicates that the method is giving results characteristic ofthe sample as a whole.In regard to the possible role played by bulk electron concentration in adsorptionand reaction, there are no good reasons that, under appropriate conditions, theelectrons in the conductivity band of the bulk of the semiconductor should not be incommunication with the surface. A useful analogy is to compare the internalelectron gas with the external molecular gas. We may consider the internal electrongas to be separated from the external gas by the solid surface. We can have the rateof reaction either dependent on, or independent of, the partial pressure of an externalgas. According to the degree of surface coverage, the observed dependency will bedetermined by the conditions of the experiment and the nature of the adsorption site.In the same way, the bending of the conduction band at the surface can be such that the" partial pressure " of the internal electron gas can influence the rate of reaction, asobserved in our experiments. The bending of the band and the nature of the surfacestates can be such that no influences are observed.*1 Coutts and Revesz, J. Appl. Plzysics, to be published, (July 1966)

ISSN:0366-9033    

DOI:10.1039/DF9664100394

出版商:RSC    

年代:1966

数据来源: RSC