ISSN:1359-6640    

DOI:10.1039/FD99192FX001

出版商:RSC    

年代:1991

数据来源: RSC

摘要:

General Discussions of the Faraday Society/ Faraday Discussions of the Chemical Society Date Subject Volume 1907 Osmotic Pressure Trans. 3* I907 Hydrates in Solution 3* 1910 The Constitution of Water 6* 191 1 High Temperature Work 7* 1912 Magnetic Properties of Alloys 8* 1913 Colloids and their Viscosity 9* 9*1913 The Corrosion of Iron and Steel 1913 The Passivity of Metals 9* 1914 Optical Rotatory Power 1o* 1914 The Hardening of Metals 10; 1915 The Transformation of Pure Iron It 1916 Methods and Appliances for the Attainment of High Temperatures in a Laboratory 12* 1916 Refractory Materials 12* 1917 Training and Work of the Chemical Engineer 13* 1917 Osmotic Pressure 13* 1917 Pyrometers and Pyrometry 13* 1918 The Setting of Cements and Plasters 14* 1918 Electrical Furnaces 14* 1918 Co-ordination of Scientific Publication 14* 1918 The Occlusion of Gases by Metals 14* 1919 The Present Position of the Theory of ionization 15* 1919 The Examination of Materials by X-Rays IS* 1920 The Microscope: Its Design, Construction and Applications 16* 1920 Basic Slags: Their Production and Utilization in Agriculture 16* 1920 Physics and Chemistry of Colloids 16* 1920 Electrodeposition and Electroplating 16* 1921 Ca pi 11arity 17* I92 I The Failure of Metals under Internal and Prolonged Stress 17* 1921 Physico-Chemical Problems Relating to the Soil 17* 1921 Catalysis with special reference to Newer Theories of Chemical Action 17* 1922 Some Properties of Powders with special reference to Grading by Elutriation 18* I922 The Generation and Utilization of Cold 18* 1923 Alloys Resistant to Corrosion 19* I923 The Physical Chemistry of the Photographic Process 19* 1923 The Electronic Theory of Valency 19* I923 Electrode Reactions and Equilibria 19* I923 Atmospheric Corrosion.First Report 19* I924 Investigation on Oppau Ammonium Sulphate-Nitrate 20* 1924 Fluxes and Slags in Metal Melting and Working 20* I924 Physical and Physico-Chemical Problems relating to Textile Fibres 20* I924 The Physical Chemistry of Igneous Rock Formation 20* 1924 Base Exchange in Soils 20; 1925 The Physical Chemistry of Steel-Making Processes 21 1925 Photochemical Reactions in Liquids and Gases 21* 1926 Explosive Reactions in Gaseous Media 22* I926 Physical Phenomena at Interfaces, with special reference to Molecular Orientation 22* I927 Atmospheric Corrosion.Second Report 23* 1927 The Theory of Strong Electrolytes 23* 1927 Cohesion and Related Problems 24* I928 Homogeneous Catalysis 24* I929 Crystal Structure and Chemical Constitution 25* I929 Atmospheric Corrosion of Metals. Third Report 25* 1929 Molecular Spectra and Molecular Structure 26* I930 Colioid Science Applied to Biology 26 Faraday Discussions Date 1931 1932 1932 1933 1933 1934 I934 I935 1935 I936 1936 I937 1937 I938 I938 I939 1939 1940 1941 1941 I942 1943 I944 1945 1945 1946 I946 I947 1947 1947 I947 1948 I948 I949 I949 I949 1950 I950 I950 1950 I95 I 1951 1952 I952 1952 1953 I953 1954 1954 I955 1955 1956 1956 I957 1958 1957 1958 1959 1959 I960 I960 1961 1961 I962 1962 1963 Subject Photochemical Processes The Adsorption of Gases by Solids The Colloid Aspect of Textile Materials Liquid Crystals and Anisotropic Melts Free Radicals Dipole Moments Colloidal Electrolytes The Structure of Metallic Coatings, Films and Surfaces The Phenomena of Polymerization and Condensation Disperse Systems in Gases: Dust, Smoke and Fog Structure and Molecular Forces in (a) Pure Liquids, and (h)Solutions The Properties and Functions of Membranes, Natural and Artificial Reaction Kinetics Chemical Reactions Involving Solids Luminescence Hydrocarbon Chemistry The Electrical Double Layer (owing to the outbreak of war the meeting was abandoned, but the papers were printed in the Transactions)The Hydrogen Bond The Oil-Water Interface The Mechanism and Chemical Kinetics of Organic Reactions in LiquidSystemsThe Structure and Reactions of Rubber Modes of Drug Action Molecular Weight and Molecular Weight Distribution in High Polymers (Joint Meeting with the Plastics Group, Society of Chemical Industry) The Application of Infra-red Spectra to Chemical Problems Oxidation Dielectrics Swelling and Shrinking Electrode Processes The Labile Molecule Surface Chemistry (Jointly with the Societe de Chimie Physique at Bordeaux) Published by Butterworths Scientific Publications, Ltd Colloidal Electrolytes and Solutions The Interaction of Water and Porous Materials The Physical Chemistry of Process Metallurgy Crystal Growth Lipo-proteinsChromatographic Analysis Heterogeneous Catalysis Physico-chemical Properties and Behaviour of Nuclear Acids Spectroscopy and Molecular Structure and Optical Methods of Investigating Cell Structure Electrical Double Layer HydrocarbonsThe Size and Shape Factor in Colloidal Systems Radiation Chemistry The Physical Chemistry of Proteins The Reactivity of Free Radicals The Equilibrium Properties of Solutions on Non-electrolytesThe Physical Chemistry of Dyeing and Tanning The Study of Fast Reactions Coagulation and Flocculation Microwave and Radio-frequency Spectroscopy Physical Chemistry of Enzymes Membrane Phenomena Physical Chemistry of Processes at High Pressures Molecular Mechanism of Rate Processes in Solids Interactions in Ionic Solutions Configurations and interactions of Macromolecules and Liquid Crystals Ions of the Transition Elements Energy Transfer with special reference to Biological Systems Crystal Imperfections and the Chemical Reactivity of Solids Oxidation-Reduction Reactions in Ionizing Solvents The Physical Chemistry of Aerosols Radiation Effects in Inorganic Solids The Structure and Properties of Ionic Melts Inelastic Collisions of Atoms and Simple Molecules High Resolution Nulcear Magnetic Resonance The Structure of Electronically Excited Species in the Gas Phase Volume 27 * 28* 29 29* 30* 30: 31 31* 32* 32* 33* 33* 34* 34* 35" 35* 35* 36* 37* 37* 38 39* 401 41 42* 42 A* 42 B* Disc.I* 2 Trans. 43* Disc. 3 4* 5* 6 7* 8* Trans. 46* Disc. 9* Trans. 47' Disc. 10* I I* 12* 13 14 15* 16* 17* 18* 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33* 34 35 Faraday Discussions Date Subject Volume 1963 Fundamental Processes in Radiation Chemistry 36 1964 Chemical Reactions in the Atmosphere 37 1964 Dislocations in Solids 38 391965 The Kinetics of Proton Transfer Processes 401965 Intermolecular Forces 1966 The Role of the Adsorbed State in Heterogeneous Catalysis 41* 1966 Colloid Stability in Aqueous and Non-aqueous Media 42* 1967 The Structure and Properties of Liquids 43 1967 Molecular Dynamics of the Chemical Reactions of Gases 44 1968 Electrode Reactions of Organic Compounds 45 1968 Homogeneous Catalysis with Special Reference to Hydrogenation and 46Oxidation 1969 Bonding in Metallo-organic Compounds 47 1969 Motions in Molecular Crystals 48 1970 Polymer Solutions 49 * 1970 The Vitreous State 50* 197 1 Electrical Conduction in Organic Solids 51 1971 Surface Chemistry of Oxides 52 1972 Reactions of Small Molecules in Excited States 53 1972 The Photoelectron Spectroscopy of Molecules 54 1973 Molecular Beam Scattering 55 1973 Intermediates in Electrochemical Reactions 56 1974 Gels and Gelling Processes 57 1974 Photo-eff ects in Adsorbed Species 58 1975 Physical Adsorption in Condensed Phases 59 1975 Electron Spectroscopy of Solids and Surfaces 60 1976 Precipitation 61 1977 Potential Energy Surfaces 52 1977 Radiation Effects in Liquids and Solids 63 1977 Ion-Ion and Ion-Solvent Interactions 64 1978 Colloid Stability 65* 1978 Structure and Motion in Molecular Liquids 66 1979 Kinetics of State Selected Species 67 1979 Organization of Macromolecules in the Condensed Phase 68 1980 Phase Transitions in Molecular Solids 69 1980 Photoelectrochemistry 70 1981 High Resolution Spectroscopy 71 1981 Selectivity in Heterogeneous Catalysis 72 1982 Van der Waals Molecules 73 1982 Electron and Proton Transfer 74 1983 Intramolecular Kinetics 75 1983 Concentrated Colloidal Dispersions 76 1984 Interfacial Kinetics in Solution 77 1984 Radicals in Condensed Phases 78 1985 Polymer Liquid Crystals 79 1985 Physical Interactions and Energy Exchange at the Gas-Solid Interface 80 1986 Lipid Vesicles and Membranes 81 1986 Dynamics of Molecular Photofragmentation 82 1987 Brownian Motion 83 1987 Dynamics of Elementary Gas-phase Reactions 84 1988 Solvation 85 1988 Spectroscopy at Low Temperatures 86 1989 Catalvsis bv Well Characterised Materials 87 1989 Charie Trakfer in Polymeric Systems 88 1990 Structure of Surfaces and Interfaces as studied using Synchrotron Radiation 89 1990 Colloidal Dispersions 90 1991 Structure and Dynamics of Reactive Transition States 91 * Not available; for current information on prices etc., of available volumes, please contact the Marketing Oficer, Royal Society of Chemistry, Thomas Graham House, Science Park, Milton Road, Cambridge CB4 4 WF, stating whether or not you are a member of the Society.

ISSN:1359-6640    

DOI:10.1039/FD991920X003

出版商:RSC    

年代:1991

数据来源: RSC

摘要:

Furuduy Discuss., 1991, 92, 13-29 Giant Palladium Clusters: Synthesis and Characterization Michael N. Vargaftik and Ilya I. Moiseev N. S. Kurnakov Institute of General & Inorganic Chemistry, the USSR Academy of Sciences, Moscow I1 7907, USSR Dmitry I. Kochubey and Kirill I. Zamaraev" Institute of Catalysis, Siberian Branch of the USSR Academy of Sciences, Novosibirsk 630090, USSR A series of palladium clusters containing from four to several hundred Pd atoms in the metal skeleton has been prepared and characterized with respect to structure and chemical properties, including catalytic activity. For smaller clusters good agreement was observed. between single-crystal X-ray and EXAFS structural data. Giant clusters approximating to Pd~61Lm(OCOCH3)lgo(L=phen, bipy) and Pd561phen,o6o&jo (x= PF6, Clod, BF,, CF3C02), were characterized with TEM, SAXS, EXAFS, NMR and magnetic succeptibility data.These clusters contain a close- packed metal core and ligands L and X that are located at the periphery of a cluster. They are very soluble in water and polar organic solvents and can be considered as a bridge between low molecular clusters and particles of colloidal metals. Giant Pd clusters were found to be active homogeneous catalysts for various organic reactions, e.g. oxidative acetoxylation of alkenes and alkylarenes; oxidation of alkenes, formic acid and alcohols; dehydration of alcohols and formation of acetals. The kinetics and mechanism of the homogeneous oxidation of alkenes and HCOzH in solutions of giant clusters were elucidated.Since Faraday's discovery of metal sols,1 colloidal metals have found numerous applica- tions in catalysis. They exhibit high catalytic activity in many reactions in solution and are also used for preparation of highly dispersed supported-metal catalysts.' However the structure of colloidal metal particles that are responsible for the catalytic activity and the detailed mechanism of catalytic reactions over these particles are still far from being clear. Recent progress in the chemistry of metal clusters has opened up a new approach to the chemistry of colloidal metal particles. However, there is still a gap between our knowledge of the structure of low molecular metal clusters, which were characterized reliably with X-ray diffraction, and colloidal metal particles.In this paper we report on the synthesis and characterization of a family of palladium clusters, which contain from four to several hundred metal atoms in their skeleton. The latter can be considered as a bridge between traditional low molecular metal clusters and colloidal metal particles. Catalysis of AIkene Oxidation with Pd(OAc), in Solution and with Supported Pd Palladium( 11) complexes have been found to oxidize alkenes stoichiometrically in acetic acid solution to give vinyl or ally1 esters and Pd metal.3 Pd"+ C,H2, + AcOH --* Pdo+ C,,Hz,-10Ac+2H' (1) 13 Giant Palladium Clusters Table 1 Interatomic distances, R in 1and 2 found from X-ray and EXAFS data X-ray EXAFS cluster distance R/A R-S/A RIA 1 Pd-C( carbon yl) Pd-O(carboxyl) 1.99 2.12 1.48 1.76 1.92 2.20 Pd-Pd 2.66 2.32 2.64 Pd-Pd 2.91 2.67 2.99 2 Pd environment" Pd -C(carbonyl)Pd-O( carbonyl) Pd -Pd 2.22-2.45 3.1 1-3.20 2.67-2.69 Pd-Mo 2.72-2.74 Pd -Pd 3.80 Mo environmentb Mo-C(carbony1) 1.97-2.01 Mo-C(Cp) Mo-Pd 2.29-2.36 2.72-2.74 EXAFS data from Pd K-edge spectrum.EXAFS data from Mo K-edge spectrum. Reoxidation of Pdo to Pd" with oxygen Pdo+2H++$02 + Pd"+H20 (11) enables the Pd to act as catalyst for the oxidative acetoxylation of alkenes C,H2,+$O2+AcOH + C,H2,-,0Ac+H20 (111) This reaction has been used in industry for production of vinyl acetate from ethene and allyl acetate from propene via vapour-phase processes with supported palladium catalyst., Formation of vinyl and allyl acetates by this reaction over supported Pd catalyst might be considered to proceed via reactions (I) and (11).However, the selectivity of reaction (111) with respect to the formation of various oxygenated products is different for the supported catalyst and Pd(OAc)* in solution, e.g. with propene, allyl acetate is formed predominantly over the supported catalyst, whilst a mixture of allyl, isopropenyl and n-propenyl acetates3 plus acetone is formed in solutions of Pd( OAc).' These facts suggest that the active centres of heterogeneous supported Pd catalyst for selective oxidation of propene to allyl acetate are different from Pd" complexes and may be, perhaps, some clusters of palladium in an oxidation state lower than two.SynthesisLow-molecular-weight Pd Clusters In search of compounds responsible for the catalytic activity in reaction (111), some low-molecular-weight palladium clusters were synthesized. In particular Pd' cluster Pd,(CO),(OAc), (1)was obtained in the single-crystal form by reduction of Pd(OAc), with CO in acetic acid solution.6 According to X-ray analysis, the nearly rectangular metal skeleton of 1bears two pairs of OAc- and CO bridging ligands [Fig. 1(a)].Pd-Pd and Pd-C interatomic distances obtained from the EXAFS spectrum [Figs. l(b) and l(c)] of the cluster were found to agree with those from X-ray data (Table 1). The EXAFS experiments and calculations are described below.M. N.Vargaftik et al. 2.5 5.63 8.75 11.9 15.0 k/ld;-' 1.o 0 Fig. 1 (a)The structure according to X-ray data6, (b)Pd K-edge EXAFS spectrum and (c) RDA function for the Pd,(CO),(OAc), cluster A series of tetranuclear palladium( I) carbonylcarboxylates Pd4( CO),( OCOR),, R = Et, Ph, CD3 ,CF, ,CC13, CH2Cl, was obtained both by the same procedure as that used for the synthesis of 1 and by the substitution of OAc- ligands in 1with RCO, using the corresponding carboxylic acids.' Their similarity to 1was ascertained by the similarity of the main pattern in the curves of radial distribution of atoms (RDA) calculated from EXAFS data for all the clusters mentioned.8 1 was found to react readily with Na,[CpMo(CO),] in THF solution to give the octanuclear cluster Na2{ Pd,[ CpMo( C0),l4} (2).X-Ray analysis of a single crystal of 2 showed the cluster molecule to have a planar metal skeleton in the form of Pd square inscribed within the Mo square (see Fig. 2).9 Interatomic distances for this cluster are presented in Table 1. Carbonyl ligands of 1 undergo ready substitution with other neutral ligands, e.g. phosphines, carbenes or 1,l.O-phenanthroline (phen). In particular, reaction of 1with diphenyldiazomethane resulted in Pd,( CPh2)4( OAc), (3): Pd4( CO),( OAc), + 4Ph2CN2 Pd4( CPh2)4( OAc), + 4CO + 4N2 (IV)---+ Our attempts to prepare a single-crystal sample of 3 failed. On the basis of elemental analysis and molecular weight data, IR, NMR and EXAFS spectra, 3was suggested to consist of a square Pd, metal skeleton with two pairs of Ph2C and OAc bridging ligands (Fig.3)." In the case of another neutral ligand, phen, substitution resulted in Giant Palladium Clusters Fig. 2 The structure of {Pd4[CpMo(CO)&J2- cluster anion according to X-ray data' + Fig. 3 (a) Pd K-edge EXAFS spectrum, (b) RDA function for the surrounding of Pd atoms and (c) the structure of Pd4(CPh2),(OAc)4 cluster suggested by EXAFS data" Pd,(CO),phen,(OAc), (4): Pd,( CO),( OAc), +4 phen =Pd,( CO),phen,( OAc), +2 CO (V) X-Ray data for the single-crystal of 4 showed that the cluster had a nearly tetrahedral Pd, metal skeleton with two carbonyl bridges at two edges and four phen ligands coordinated bidentately at four vertices (Fig.4)." Note, that OAc- ligands are ousted from the inner sphere of the cluster, apparently because of steric hindrances, to function as outer-sphere anions. M. N. Vargaftik et al. 0 Fig. 4 The structure of [Pd4phen4(C0)2](0Ac)4 cluster according to X-ray data" 2 was found to catalyse dehydration of alcohols under mild conditions (293-353 K) in the absence of strong acids. The reaction was shown to proceed via an unusual mechanism including carbene species. l2 1 was shown to be inert as a catalyst for reactions of alkenes. However, an active catalyst of oxidative substitution of H atom of alkenes was formed when phen was added to a solution of 1. E.g. propene and 2-methylpropene were oxidized selectively by O2 at 20°C and 1 atm in a methanol solution which contained 1 and phen in the molar ratio phen :Pd < 1:1 :l3 CH2=CHCH3+20,+ MeOH ---* MeOCH2CH=CH2+ H20 (VI)CH2=C(Me)CH3+;O2+ MeOH + MeOCH2C(Me)=CH2+ H20 Oxidative acetoxylation [reaction (III)] of ethene and propene by O2to give vinyl and ally1 acetates, respectively, was found to proceed when an alkene-0, gas mxiture was contacted with a solution of 1 in AcOH with phen added (phen: Pd = 1:2);5 toluene was oxidized selectively by the same solution to give benzyl a~etate.~ The catalyst in these reactions appeared to be a phenanthroline complex of low- valence palladium.However, 4 proved to be inactive in these reactions, presumably because of steric inaccessibility of Pd atoms for reagent molecules. Catalytic activity of the solutions which contained 1+phen, appeared only if [phenlo: [Pdl0< 1:1.The highest activities were achieved for phen :Pd in the range 1:2-1 :4, after complete loss of CO by cluster (4).5 Giant Pd Clusters Catalytically active solutions were found to be also formed via reduction of Pd(OAc), with H2 and other reductants in the presence of L = phen, bipy or their derivatives.' With H2 as the reductant, we have studied the stoichiometry of the reactions that lead to the formation of the catalytically active Pd compounds. We have also isolated these compounds and characterized them with EXAFS' '-I6 and other spectroscopic techniques. The primary isolated product of Pd(OAc), reduction with H2 in AcOH solution was found to be an X-ray-amorphous substance of the composition Pd,phen(OAc),H, as ascertained by elemental analysis.The formation of this substance from Pd(OAc), was Giant Palladium Clusters -20 -40 -60 S(ppm) Fig. 5 ‘HNMR spectra of 5 in the hydride region: (a)solution of 5 in MeCN; (b) solution of the analogue of 5 (L=bipy instead of phen) in MeCN; (c) solid 5 under magic angle spinning. Chemical shifts S are measured with respect to TMS found to require 1.3 f 0.05 mol of H2 per Pd atom: 4Pd,(OAC)6+3phen+ 15H2=3Pd,phen(OAc),H,+ 18AcOH (W Palladium(I1) acetate is known to be a trimer Pd,(OAc),. On the basis of these data, the product of the reaction was tentatively suggested to be a hydride complex [Pd,phen(OAc),H,],] (5). The presence of hydrogen atoms within complex (5) was ascertained by ‘H NMR spectra which exhibit lines in the interval of the chemical shifts that is typical for hydrides (Fig. 5).’4915 5 demonstrates high reactivity.16 As a rule, 5 and its analogues decompose to give low-molecular-weight Pd complexes under the action of various electrophilic reagents, e.g.tetrachloro-o-benzoquinone,NO, CO and 1,3-dimethyl-l-r1itrourea.’~The only exception was the reaction with oxygen. Upon interaction with 02,5 lost hydrogen atoms to give H20 and a new cluster (6)of Pd,phen(OAc), composition was f~rrned.~~”~ 6 is stable in air and is soluble in water and polar organic solvents. It can be precipitated from aqueous solutions by adding sodium or potassium salts MX or M2X or acids HX or H2X (X-=Cl-, ClO,, SO:-, BF,, PF,, CF,CO,, CH,CO,).When anions X- different from CH,CO, were used, OAc- ligands in the initial cluster (6) were substituted by these ligands. Structural characterization shows that 6 and its analogues with various anionic ligands are giant clusters of Pd. In particular 6 and its analogue (7) with X=PF, were characterized by high- resolution transmission electron microscopy, electron diffraction, EXAFS, small-angle X-ray scattering, NMR, magnetic susceptibility and molecular mass measurements. Structural Characterization of Giant Pd Clusters Experimental High-resolution Transmission Electron Microscopy ( TEM) TEM data were obtained with a JEM-100 CX electron microscope (ca. 2 A line resol- ution), at the minimal electron beam intensities to diminish the destructive influence of the electron beam upon cluster samples.The image scale was calibrated against the 3.345A interplanar distance of graphite. Samples for TEM study were prepared by pouring a drop of solution of the cluster in MeCN, MeOH or AcOH onto carbon M. N. Vargaftik et al. supports, followed by evaporation of the solvent in vacuo. Since 5 is air-sensitive, its samples were prepared and examined under an Ar atmosphere. In experiments with 6 and 7 it was ascertained that electron micrographs of each particular region of the sample studied gave virtually identical images for exposure times from 10s to 10min. Electron Diflraction (ED) ED data were obtained for the same samples with clusters 5-7 deposited on carbon films as in the TEM experiments; the same electron microscope was used.The exposure time was 1-3 min. When obtaining each diffractogram, an image region was chosen which contained at least 10-20 cluster particles. Interplanar distances were calibrated against those for a film of crystalline gold photographed using the same apparatus and the same operating mode. EXAFS EXAFS spectra of the K-edge of palladium X-ray absorption of clusters 1-7 were obtained at the EXAFS Station of the Siberian Center of Synchrotron Radiation, Nov~sibirsk.'~"The storage ring VEPP-3 with electron beam energy 2 GeV and average stored current of 70 mA was used as the radiation source. The energy of the synchrotron radiation quanta was monitored with the help of the two crystal cut-off Si"") mono-chromator.Note, that it was checked that higher harmonics (3hv etc.) for the energies around the Pd K-edge (24.6 keV) were absent in the spectra of the VEPP-3 radiation source. X-Ray absorption spectra were recorded in the transmission mode using two ionization chambers, Le. the monitoring one and the full absorption one. The monitoring chamber contained 0.5 atm Ar and was located in front of the sample, while the full absorption chamber contained 1 atm Xe and was located behind the sample. The samples of clusters 1-7 were prepared in the form of pellets, which exhibited at the Pd K-edge a jump of absorption Ap X =0.8. The pellets consisted of ca. 80 YOof the cluster material plus ca.20 YOof Apieson-L hydrocarbon binder. Pd foil with a known X-ray absorption edge energy, distances between Pd atoms and coordination numbers was used as a reference sample to calibrate the position Eo of the absorption edge for the clusters, as well as the phase corrections 6 for Pd-Pd distances for a given interval of k values. Solid Pd(OAc), with known was used to calibrate the phase correction 6 for Pd-L distance, where L is the oxygen or nitrogen atom of a ligand in the first coordination sphere of the metal atom. For each sample X-ray absorption data were analysed in the form of kX(k) and k3x(k), where x(k) is the oscillating part of the absorption coefficient p, for two different intervals of wavenumbers k =2.5-15.0 A-1 and k =4.0-17.0 A-'.The background was removed by extrapolating absorption in the pre-edge region onto the EXAFS region in the form of Victoreen's polynomials. Three cubic splines were used to construct the smooth part of p. The inflection point of the edge of the X-ray absorption spectrum was used as the initial point (k =0) of the EXAFS spectrum. The RDA function was calculated from the EXAFS spectra in kX(k) or k3x(k)forms using Fourier analysis."* An alternative fitting proced~re,'~~ was also used to determine interatomic distances from the EXAFS spectra. Whenever both procedures were used to analyse the same EXAFS spectrum, they gave very similar values for the same Pd-Pd and Pd-L distances. For the same sample and the same procedure of analysis (Fourier analysis or fitting) very similar values were obtained for the same Pd-Pd and Pd-L distances from EXAFS spectra in kx(k) and k"x(k) forms.Theoretical parameters from McKale table^"^ were used when determin- ing RDA functions using a fitting procedure. The phase correction S for Pd-Pd and Pd-L distances for a given interval of k values, which has to be known when one looks Giant Palladium Clusters for Pd-Pd and Pd-L distances from RDA curves obtained via Fourier analysis, was found experimentally from RDA curves for the reference materials, i.e. Pd foil with the known Pd-Pd distances and the complex Pd(OAc)2 with known structure. More details about the EXAFS Station of the Siberian Center of Synchrotron Radiation and the techniques that are used there to record and analyse EXAFS spectra can be found else~here.”~ Small-angle X-Ray Scattering (SAXS) SAXS data were obtained by a standard procedure with the use of the small-angle X-ray chamber KRM-1.” NMR NMR spectra were recorded with a Bruker CXP-300 spectrometer with and without magic angle spinning adapter.The frequency of spinning was 2 kHz. Molecular Mass The molecular mass for 6 was estimated from the rate of sedimentation in aqueous solution under ultracentrifugation with the use of an MOM-3180 (Hungary) ultracen- trifuge at the maximum rate of rotation 58 000 s-I. The molecular mass, M, was calculated using the formula based on the Stokes-Einstein law: where T~ is the viscosity of the solvent; So is the ‘true’ (i.e.extrapolated to zero concentration of the solute) sedimentation coefficient; v is the partial specific volume of the solute; po is the density of the solvent. The moving-boundary method was applied and the same technique was used as that used earlier for the Aus5 clu~fer.’~ Magnetic Susceptibility The magnetic susceptibility of solid samples was measured by Faraday’s method in the (0-9) x lo3Oe interval of field strength at 77-300 K.16 Results and Discussion Palladium Hydrido Cluster Considerable broadening of the hydride signal in the ‘H NMR spectra of 5 in solution [Fig. 5(a) and (6)J may be explained e.g. by assuming that this substance is a large polynuclear palladium cluster similar in structure to a fragment of palladium hydride or palladium metal.Weak temperature-independent paramagnetism (specific suscepti- bility xy =0.5 x which is typical for metals, was detected for 5 in the temperature range 77-300 K.16 TEM showed the particles of 5 to have a metallic core of nearly spherical shape and 20k5 8, in diameter [Fig. 6(a)],in accordance with SAXS data [Fig. 6(6)].Additional information on the structure of 5 was obtained from EXAFS. The RDA calculated from EXAFS data (see Fig. 7) has a shoulder, which, if attributed to a Pd-N bond, corresponds to interatomic distance 2.1 kO.1 A, that is typical for Pd-N distances in palladium complexes. The RDA has also a peak from the distance between neighbouring Pd atoms, H =2.6* 0.1 A, which is somewhat shorter than that for bulk Pd metal (2.74 A).Maxima corresponding to the distances between more remote Pd atoms were not detected, presumably because of some disorder in the arrangement of Pd atoms, which M.N.Vargaftik et al. Fig. 6 (a) TEM microphotographs and (b) diameter distribution curves for the metal cores according to the SAXS data for 5, 6 and 7 t Fig. 7 RDA curve for cluster 5 obtained from EXAFS spectrum could be induced, e.g. by irregular inclusions of H atoms in the cluster skeleton or for some other reason. The size of metal core in 5 found from TEM and SAXS data allowed us to estimate the value n in the idealised [Pd,phen(OAc),H,], formula as lo2, supposing that the packing density of Pd atoms in the core of 5 is the same as that in bulk Pd metal.22 Giant Palladium Clusters 0.0066 n m I 5 Q 0 2 c 6 8 0.042 n cro I e W Q 0 2 c 6 a (R -WA Fig. 8 (a),(b) Pd K-edge EXAFS spectra and (c), (d) RDA curves for the surroundings of Pd atoms for 6 in the initial [(a),(c)] and relaxed [(b),(d)]forms 'Palladium-56'1' Cluster The molecular mass of 6was estimated as (1.O f 0.5)x lo5from the data on the sedimenta- tion rate in aqueous ~olution.'~ The size of 6 was evaluated from SAXS, TEM and electron diffraction data. In TEM metallic cores of the 6molecules were observed as nearly spherical particles of 26f3.5 A diameter [Fig. 6(a)]. In the electron diffractogram of the 6, there were several dif€use rings with an arrangement of maxima close to those for the metallic palladium.In view of the EXAFS data (see below), which suggest that the structure of the metal core in 6 is notably different from that for the bulk Pd metal, this diffractogram is assumed to be caused by the destructive influence of the electron beam on cluster molecules exposed in the electron microscope, resulting in the loss of the ligands and relaxation of the structure of the cluster metal core to that typical for the bulk Pd metal. Nevertheless, no agglomeration of the palladium particles was observed under the conditions used in TEM and ED experiments. From the half-width of diffraction rings, the size of the particles responsible for the diffraction pattern was found to be ca. 25 A, in agreement with TEM (26f3.5 A) and SAXS data.The structure of the native cluster 6 was characterized with EXAFS data [Fig. 8(a) and (c)]. The RDA curve, which characterizes the local environment of Pd atoms in 6 is given in Fig. 8(c). One of the peaks in the RDA curve of Fig. 8(c) was found to correspond to a short interatomic distance (presumably to R =2.1 fO.1 A) between Pd and N atoms. The four other intense peaks correspond to the four shortest Pd-Pd distances (Table 2). The set of Pd-Pd distances obtained from EXAFS (the ratio of the interatomic distances are 1 :1.2 :1.4 :1.6) is seen to be consistent with the icosahedral packing of Pd atoms in the metal core of 6 (the ratios of the interatomic distances expected for the four nearest-neighbour atoms of the icosahedral skeleton are also 1 :1.2 :1.4: 1.6) and to deviate notably from the patterns of distances expected for f.c.c.and h.c.p. packings. M. N.Vargaftik et al. Table 2 The four shortest Pd-Pd distances in the metal skeleton of 6 found from EXAFS data, compared with those expected for various packings of Pd atoms data type Pd-Pd distance/A EXAFS 2.60*0.04 3.1k0.1 3.66k0.1 4.08*0.1 -packings" f.c.c. 2.60 -3.66 -4.50 h.c.p. 2.60 -3.66 -4.50 icosahedron 2.60 3.10 3.66 4.10 -a In calculations, the shortest Pd-Pd distance was taken to be 2.60A for all packings. With the known character of packing of Pd atoms and the distances between the nearest-neighbour Pd atoms in the metal core of 6 (Table 2), one can estimate the number Nz of palladium atoms in the cluster molecule. As was found for a sphere 25 %i in diameter, Nx is ca.570. On the basis of this value for Nxand of the chemical composition of 6, the cluster was approximated by the formula 15Pd570~30phen63~3(0Ac)~90~lo~The molecular mass corresponding to this formula (M = 83 200) agrees with the value of M = (1 * 0.5) x lo5,which was estimated from the rate of sedimentation in solution. The value found, Nx= 570, matches quite well the idealised five-layer icosahedron, which contains, according to the well known formula Nx= 1/3( 10m3+ 15rn2+ 11 rn + 3), M (the number of layers) = 5, Nx= 561 metal atoms. Taking into consideration the data of the chemical analysis, the overall Pd561phen60( OCOCH3)180 idealised formula can be suggested for the 6.Thus, we arrive at the conclusion that it is really a giant cluster containing more than 500 palladium atoms in its metal core. The idealised formula seems to correspond to some average size and composition of the cluster, rather than to a certain fixed size and composition. Examination of molecular models showed that bidentate phen ligands, because of steric hindrance, may be coordinated only at the edges and vertices of the icosahedron. This examination has shown that at the outer layer of the idealised icosahedron which contains 252 metal atoms, ca. 60 bidentate phen ligands may be arranged. As a result, almost the whole surface of the metal skeleton is screened sterically by bulky phen ligands. Acetate anions may be located only in the outer sphere of the cluster (see Fig.9), as found by X-ray structural analysis for the low-molecular-weight cluster [Pd4phen4(CO)2]10Ac)4 (see Fig. 4). Structural Non-rigidity of Giant Clusters Icosahedrai metal particles of nm scale were occasionally observed among other small ones (f.c.c., decahedral etc.) obtained by the condensation of the vapour of noble metals on a cool support.2o Owing to the comparatively small difference betwen f.c.c. and icosohedral packings, the icosahedral metal skeleton of the Pd giant cluster was expected to be readily transformed to the f.c.c. one upon various perturbations, such as heating (e.g. under the beam of the electron microscope) or ligand substitution. In agreement with this expectation a noticeable exothermic effect accompanied with a mass loss, was observed at 120-130 "C upon heating 6 in a solid form (Fig.10). The RDA curve [Fig. 8(d)] calculated from EXAFS data [Fig. 8(6)] for 6 after its thermolysis at 130 "C for 2 h, indeed showed disappearance of the peaks at 3.1 and 4.1 A, which are characteristic of icosahedral packing. The shortest Pd-Pd distance found for sample 6 after its Giant Palladium Clusters Fig. 9 Idealised model of cluster 6; 1 = Pd atoms coordinated with phen ligands; 2 = Pd atoms accessible for coordination with OAc- anions or molecules of substrate or solvent; 3 =van der Waals shapes of coordinated phen molecules I 20 so 100 150 200 300 T/OC Fig. 10 Derivatogram curves for 6 and 7 M.N.Vargaftik et al. Table 3 Reactions catalysed by clusters 6 and 7 reaction reagents products conditions yield/% ref. C2H4 C3H6 C6HSCH3 oxidative acetoxylation CH,=CHOAc 60-100 "C, 1 atm of O2+alkene, AcOH soln. O2+alkene, AcOH soln. CH~=CHCH~OAC 60-100 "C, 1 atm of C6HsCH2OAC 80-100 "C, 0.2-1.0 95-99 96-98 95-98 22,23 5,16 .5 atm of 02,AcOH soln. oxidation of formic acid HCOZH C02+ H20 20-70°C, AcOH soln. 100 23 oxidation of alcohols C2HSOH CH3C02C2HS 60 "C, 0.2- 1.O atm 30 22 CH3CH(OC2H5)2CH3CHO of 02,C2HsOH soln. 50 20 CH3CHOHCH3 CH3COCH3 20 "C, 0.2-1.O atm 100 22 of 02, CH3CHOHCH3 formation of acetals CH3CHO +CzHsOH CH3CH(OC2H512 20-50 "C, C2H50H soln. 90 22 thermolysis, became nearly equal to that for bulk palladium (2.77 A).The set of Pd-Pd distances calculated from EXAFS data (2.77, 3.85, 4.84A) fitted the f.c.c. structure of the metal skeleton well. Work is now in progress to clarify the kinetics of relaxations of the structure of cluster 6 under various conditions. Similar effects were observed upon substitution of OAc- ligands with other anions. In particular, the giant cluster 7, with the idealised formula Pds61phen60( PF6)60060 (28 f5 A diameter of metal core), was obtained by precipitation of aqueous solution of 6 with KPF616 and characterised by TEM, SAXS and EXAFS. An exothermic effect upon heating was observed also for this substance (Fig. 10). However, even the freshly prepared sample of 7 exhibited the RDA curve, which contained the set of Pd-Pd distances characteristic of f.c.c.packing (2.79, 3.87, 4.75 A), as well as some additional peaks (at 4.08 and 4.42 A). The attribution of the last two peaks is still unclear. Small variations in preparation procedure (temperature, pH of solution, time of precipitation) as well as heating or ageing at room temperature resulted in some changes in the RDA curve. Similar results were obtained for the clusters formed upon precipitation of 6 with other anionic ligands, C104-, SO2-, BF4-and CF,CO;. The mixtures of several types of giant clusters with different packings of metal core, though close in their size and chemical composition, are assumed to be formed as a result of ligand substitution. Catalytic Properties of Giant Clusters Polynuclear hydrido complexes 5 were found to exhibit catalytic activity for hydrogena- tion of unsaturated compounds, dimerization, and positional isomerization of lower alkene~.'~These reactions are typical of those catalysed by hydrido complexes of transition metals." All the reactions are inhibited by dioxygen, which converts 5 into the giant cluster 6, which has different catalytic properties (see Table 3).Giant Palladium Clusters Table 4 The values of the constants of kinetic equation (1) for oxidative acetoxylation of ethene and propene in solutions of 6 and 7 at 60 "C substrate cluster k/min-' &/lo3 K,~/lo4 KII, ethene 7 8.2k0.7 5.8*0.3 3.0k0.2 1.3k0.1 propene 7 5.6*0.5 CQ. 30 1.2zk0.1 4.8k0.5 progene 6 3.3k0.3 ca.30 5.2k0.3 0.7k0.05 Mechanism of Oxidative Reactions Catalysed by Giant Clusters Giant palladium clusters exhibit diverse catalytic activities. For oxidative acetoxylation of alkenes and HC02H oxidation catalysed by 6 and 7 the kinetics and mechanism were in more detail than for other reactions. Oxidative acetoxylation of ethene and propene in solutions of 6 and 7 was found to obey the kinetic where ro is the initial rate of formation of vinyl acetate from ethene and that of ally1 acetate from propene, respectively; KI, KIIand KIIIare the Michaelis constants (Table 4). The oxidation of HC02H catalysed with 6 and 7 was found to follow similar kinetics to those of ethene and ~ropene.~~ The Michaelis-Menten character of the reaction kinetics suggests that the formation of the products is preceded by reversible coordination of alkene, O2and AeOM molecules with the cluster.A comparison of the values of KI (Table 4) shows propene to be coordinated more weakly than ethene. In homogeneous oxidation of alkenes with Pd" complexes, the oxidation products appeared, in the absence of dioxygen, in stoichiometric quantities with respect to the Pd" In contrast to this, in the case of giant Pd clusters, the oxidation of alkenes was not observed in the absence of 02.This fact further supports the assumption that the cluster-catalysed oxidations could not be regarded as the result of alternating oxidation-reduction reactions of Pd" and Pd" with substrate and oxidant molecules, respectively. The surface of the cluster core is assumed to be screened with bulky phen ligands. Examination of the idealised molecular model of 6 showed that only ca. 20 palladium atoms of the 252 located at the surface of the cluster skeleton, are sterically accessible for C2H4 or AcOH molecules.In this steric situation, a smaller value of the stability constant for the rr-complex of the giant cluster with propene as compared to that with ethene, may be explained by the larger size of the propene molecule. The structure of the real molecules of giant clusters is assumed to more or less deviate from that of a perfect polyhedron. The number of accessible coordination sites can be evaluated in this case experimentally using a poisoning technique, i.e.by oxidation of the substrates in the presence of some ligands capable of strong bonding with Pd atoms at the surface of the metal skeleton. The data obtained for the oxidation of C2H4, C3H6 and HC02H showed that bulky ligands, e.g. PPh3 and phen had no effect on the rates of the reactions. In contrast to these, smaller ligands, e.g. C2H5SH and SCN-, efficiently suppressed the cluster-catalysed oxidation. Complete inhibition of ethene oxidation was achieved by adding ca. 50 molecules of small ligands per cluster molecule. For the oxidation of propene and M. N.Vargaftik et al. Table 5 Kinetic isotope effects for the oxidation of ethene, propene, and formic acid by dioxygen in the solutions of 6 and 7 in AcOH at 60 “C catalyst substrate (cluster) ksubstr-H/ ksubstr-D kCH3C02H/ kCD3C02D ethene 7 1.1 *O.l 1.1 f0.1 propene 6 2.2f 0.2 1.O f0.05 propene 1 3.6f 0.2 1.O f 0.05 propene Pd black 1.0*0.1 2.0f0.2 formic acid 6 1.1 *0.1 1.Of 0.1 formic acid 7 1.1 fO.1 l.O*O.l formic acid, only cu.15 ligand molecules per cluster were required for complete inhibition of the reaction. Considering the facts mentioned above, as well as the values of the kinetic isotope effects (KIE) found for the reactions studied (see Table 9,the rate-determining step of ethene oxidation was proposed to be an oxidative addition (with the opening of the ?r-bond) of a ?r-coordinated C2H4molecule to a Pd-Pd group of the cluster to form the o,c-coordinated . .Pd-CH2CH2-Pd ..group. Subsequent splitting of the C-H bond in this group is assumed to be fast and facilitated owing to the intermediate formation of the Pd=C multiple bond: H /H2C=CH2 slow H2C--CH2 fast H2C-C H fast1 -I1 -IIII --Pd-Pd-Pd--Pd....Pd-Pd--Pd....Pd-Pd-H,C=CH H (VIII) /I I -Pd-Pd....Pd-Because of the enhanced stability of the allyl group, the oxidation of propene is prcposed to proceed via another rate-determining step, which includes the oxidative addition of propene molecule to a Pd-Pd group to split the allyl-H bond, as shown by eqn (IX). Formation of the surface wallyl group appears to favour the ‘allyl’ reaction path over the ‘vinyl’ one, ensuring high selectivity of propene oxidation to allyl acetate. The overall catalytic cycle proposed for the reaction of alkene oxidation over giant palladium clusters as catalysts, is presented in Scheme 1.Giant Palladium Clusters COCH, Scheme 1 References 1 M. Faraday, Philos. Trans. Roy. Soc. London, 1857, 147, 145. 2 J. B. Nagy, I. Bodart-Ravet and E. G. Derouane, Faraday Discuss., Chem. Soc., 1989,87, 189. 3 1. I. Moiseev, M. N. Vargaftic and Ya. K. Syrkin, Dokl. Akud. Nauk SSSR, 1960, 130, 820; 133, 377. 4 K. Fujimoto and T. Kunugi, J. Jpn. Petrol. Inst., 1974, 17, 739. 5 I. I. Moiseev, Sou. Sci. Rev., Chem. Rev., Harwood, London, 1982, Sect. B, 4, 139. 6 I. I. Moiseev, T. A. Stromnova, M. N. Vargaftik, G. Ya. Mazo, L. G. Kuz’mina and Yu. T. Struchkov, J. Chem. SOC.,Chem. Commun., 1978, 27. 7 T. A. Stromnova, M.N. Vargaftik and I. I. Moiseev, J. Organometal. Chem., 1983, 252, 113. 8 T. A. Stromnova, N. Yu. Tikhonova, D. I. Kochubei and I. I. Moiseev, to be published. 9 T. A. Stromnova, I. N. Busygina, S. B. Katser, A. S.Antsyshkina, M. A. Porai-Koshits and I. I. Moiseev, J. Chem. Soc., Chem. Commun., 1988, 114. 10 T. A. Stromnova, I. N. Busygina, D. I. Kochubey and I. I. Moiseev, Mendeleev Commun., 1991, 1. 11 M. N. Vargaftik, T. A. Stromnova, T. S. Khodashova, M. A. Porai-Koshits and I. I. Moiseev, Koord. Khim., 1981, 7, 132. 12 T. A. Stromnova, I. N. Busygina and I. I. Moiseev, VII International Symposium on Homogeneous Catalysis, Lyon, 1990, Abs, p. 103. 13 T. A. Stromnova and M. N. Vargaftik, Izv. Akad. Nauk SSSR. Ser. Khim., 1980, 478. 14 M.N. Vargaftik, V. P. Zagorodnikov, I. P. Stolarov, D. I. Kochubey, V. M. Nekipelov, V. M. Mastikhin, V. D. Chinakov, K. I. Zamaraev and I. I. Moiseev, Izv. Akad. Nauk SSSR, 1985, 2381. 15 M. N. Vargaftik, V. P. Zagorodnikov, I. P. Stolarov, I. I. Moiseev, V. A. Likholobov, D. I. Kochubey, A. L. Chuvilin, V. I. Zaikovsky, K. I. Zamaraev and G. I. Timofeeva, J. Chem. Soc., Chem. Commun., 1985,937. 16 M. N. Vargaftik, V. P. Zagorodnikov, I. P. Stolarov, I. I. Moiseev, D. I. Kochubey, V. A. Likholobov, A. L. Chuvilin and K. I. Zamaraev, J. Mol. Catal., 1989, 53, 315. 17 (a)D. I. Kochubey, Yu. A. Babanov, K. 1. Zamaraev, R. V. Vedrinskii, G. N. Kulipanov, L. N. Mazalov, A. N. Skrinskii, V. K. Fedorov, B. Yu. Helmer, and A. T. Shuvaev, X-Ray Spectral Method of Sfructural Study of Amorphous Solids, Novosibirsk, Nauka, Siberian Branch, 1988; (b) A.C. Skapski and M. L. Smart, J. Chem. Soc., Chem. Commun., 1970,658; (c)W. Blau, E. Zschech and J. Bergmann,Nucl. Instr. Meth. A., 1987, 261, 166; (d) A. G. McKale, B. W. Weal, A. P. Paulikas, C. K. Chan and G. S. Knapp. J. Am. Chem. Soc., 1988, 110, 3763. 18 D. I. Svergun and L. A. Feigin, X-Ray and Neutron Small-Angle Scattering, Moscow, Nauka, 1986. M. N. Vargaftik et al. 19 G. Schmid, R. pfeil, R. Boese, F. Bandermann, S. Meyer, G. H. M. Calk and J. W. A. van der Velden, Chem. Bet., 1981, 114, 3634. 20 L. D. Marks, Ultramicroscopy, 1985, 18, 445. 21 B. R. James, Homogeneous Hydrogenation, Wiley, New York, 1973. 22 V. P. Zagorodnikov and M. N. Vargaftik, Izv. Akud Nauk SSSR. Ser. Khim, 1985, 2652. 23 V. P. Zagorodnikov, M. N. Vargaftik, A. P. Lyubimov and I. I. Moiseev, Izv. Akad. Nauk SSSR. Ser. Khim., 1989, 1495. 24 N. M. Zhavoronkov, Yu. A. Pazdersky, M. K. Starchevsky, P. I. Pasychnik, M. N. Vargaftik and I. I. Moiseev, 2.Anorg. Allg. Chem., 1989, 576, 284. Paper 1/025975; Received 28th May, 1991

ISSN:1359-6640    

DOI:10.1039/FD9919200013

出版商:RSC    

年代:1991

数据来源: RSC

摘要:

Faraday Discuss., 1991, 92, 31-44 Chemistry of Ag, Aggregates in Aqueous Solution: Nsn-metallic Oligomeric Clusters and Metallic Particles Arnim Henglein,* Paul Mulvaney and Thomas Linnert Hahn-Meitner-Institut Berlin GmbH, Bereich S,lOOO Berlin 39, Germany Since the discovery of long-lived oligomeric silver clusters in aqueous poly- anion solutions it has become possible to investigate the chemical reactivity of silver aggregates over a wide size range, especially in the range where the transition from the atom to the metal takes place. Both oligomeric and metallic particles react with nucleophilic reagents such as NH3, CN-and SH-.The reactivity is attributed to the coordinative unsaturation of silver surface atoms. In the case of the metallic particles, the changes in the shape of the surface plasmon absorption band can be used to detect the interaction of nucleophilic reagents with surface atoms. A mechanism is proposed for oxidative corrosion catalysed by chemisorbed molecules.Oligomeric clusters are strong reducing agents which can transfer an electron to 02,nitro-compounds, carbon tetrachloride etc. The standard potential of the silver microelectrode, Ag, * Ag,-, + Ag++ e-, as a function of the agglomeration number n is also discussed to explain the stability of the clusters. Electron acceptors often catalyse the conversion of oligomeric clusters into larger metallic particles. Electrons are readily photo-emitted from oligomeric clusters. As the transition to the metal occurs with increasing size, the quantum yield of emission strongly decreases.The emitted electrons appear as eiq in the aqueous solvent. Two developments in the colloid chemistry of silver have arisen during the past few years: 1. The investigation of the reactions of surface atoms in particles several nanometres in diameter that have metallic properties and 2. the investigation of extremely small oligomeric clusters, the size dependence of their reactivity and properties in the transition range between the atom and the metallic state being of special interest. The reactions studied include electron-transfer processes. Small silver particles in aqueous solution are very sensitive towards oxygen; all experiments have therefore to be carried out under anaerobic conditions.Small silver particles can transfer an electron to many organic and inorganic acceptors A: Ag,+A + Ag:+A-6) and the description of such processes has to be given in terms of the standard potentials of the two redox systems AgL/Ag, and A/A-. Very little is known about the standard ionization potential Eo(Agi/Ag,) as a function of the agglomeration number n. On the other hand, strong electron donors D may transfer an electron to a silver particle, Ag,+D -+ Ag,+D' (ii) and these processes are described using the standard electron afinity potential Eo(Ag,/AgJ which is also not known in detail. Finally, a silver particle in a solution 31 Chern is try of Silver Aggregates containing excess silver ions may be regarded as a metal microelectrode, the equilibrium being: (iii) Such a metal electrode potential is of interest in the description of processes where D donates an electron to a silver ion on the surface of a silver particle Agn-, such as in the photographic development process.It is not possible to discuss all three potentials in the present paper. What is known about them has recently been compiled' and we make use of this data. Another interesting feature in the chemistry of small silver particles is their interaction with nucleophilic compounds.2 The atoms on the surface of a metal are coordinatively unsaturated, this effect being more pronounced the smaller the particle. In other words, unoccupied orbitals are present on the surface into which nucleophiles can donate electrons.Typical nucleophiles are the functional groups of polyanions such as poly- phosphate, poly( vinyl sulfate), and polyacrylate. They may weakly complex silver particles and in this way act as stabilizers since the negatively charged chains of the polyanions repel each other thus preventing the bonded particles from colliding with each other. However, the nucleophiles studied also include ammonia, cyanide anions, hydrogen sulfide anions and thiolates, i.e. compounds which are known to form strong complexes with free silver ions in aqueous solution. Both electron exchange and nucleophilic addition to silver particles often lead to changes in the optical properties of the particles. Therefore, by recording the optical properties, one can often draw conclusions about the state of the particle surface.For example, the intensity, wavelength position and half-width of the well known surface plasmon absorption band, which lies at 380nm in the case of metallic silver particles in water, is very sensitive towards chemical modifications of the surface. Another reaction described here is photoelectron emission. In the case of particles in aqueous solution, the photoemitted electron is hydrated and can readily be detected by its strong absorption at 700 nm. Photoelectron emission from compact silver elec- trodes occurs with extremely small quantum yields (<10-4).3 It is shown here that emission can occur with yields greater by many orders of magnitude in the transition range between the metallic and atomic states.Experimental y-Radiolysis and electron-pulse radiolysis were used in some of the redox investigations. These methods allow one to produce a known concentration of reducing or oxidizing radicals in the aqueous solution and to let the radicals transfer an electron to, or inject a positive hole into, the colloidal metal particles. The methods have important advan- tages in comparison with chemical redox experiments: they are very reproducible, a minimum of disturbing impurities is introduced, and the reactions are initiated homogeneously (i.e. without local concentration gradients when reagents are mixed). Ionizing radiation produces hydrated electrons and hydroxy radicals. The hydrated electrons are very strong reducing agents; they can, for example, reduce Ag+ ions.The oxidizing OH radicals can be converted into reducing organic radicals in the presence of an alcohol: 'OH +R-CH20H 3 H,O+ RCHOH (iv) Thus, by exposing a deaerated silver salt solution (some mol dmd3) containing ca. 0.1 mol dm-3 alcohol to y-rays, the silver is reduced to form colloidal aggregates. A polyanion (10-5-10-3 mol dmT3) in the solution stabilizes the aggregates. Depending on the conditions, either metallic silver particles several nanometres in diameter or very much smaller oligomeric clusters are stabilized. A. Henglein, l? Mulvaney and T. Linnert 40 1.5 .. I 1.0 'i .I20 z ms: 0.5 3 X Q) 5 10 e 024 a -20 -40 300 400 so0 A/nm Fig.1 0, Surface plasmon absorption band of a 1x rnol dm-3 silver sol. Changes in absorption (a) after electron donation and (6) positive hole injection by free radicals When the reducing radicals transfer an electron to metallic silver particles the Fermi potential of the latter shifts to more negative values. It has been shown previously that the shift is often large enough to enable the stored electrons to form H2 from the aqueous solvent, i.e. the colloidal particles act as catalysts for the two-electron reduction of water by radical^.^ In the present studies, radicals are again used to shift the Fermi potential; however, our interest is now focussed on the accompanying changes in the optical properties of the particles.Radicals can also be produced photolytically. The solution contains ca. lo-* mol dm-3 acetone and 0.5 mol dm-3 alcohol. UV light which is absorbed by the acetone initiates radical formation (via the triplet state of acetone with subsequent H-atom abstraction from the alcohol). The solutions were evacuated in a glass vessel carrying a side arm with an optical cuvette. The absorption spectrum was measured at various times after irradiation in the field of a 6oCo y-source. Reagents were added to the irradiated solution by sucking a certain amount of deaerated reagent solution into the evacuated vessel. Pulse radiolysis was carried out with 3.6 MeV electrons as described previ~usly.~ The equipment for laser flash photolysis has also been de~cribed.~ Results and Discussion Metallic particles Electron Donation and Optical Absorption Fig.1 (spectrum 0) shows the surface plasmon absorption band of a 1 x lo-" mol dm-3 silver sol which had a mean particle size of 3 nm. The sol was prepared by exposing a deaerated 1 x mol dmA3 AgClO, solution containing 0.2 mol dm-3 propan-2-01 and mol dm-3 poly(viny1 sulfate) to a dose of lo5rad of y-radiation. Chemistry of Silver Aggregates J lrvwmi 2468 Fig. 2 Change in (a) electrical conductivity and (b) in the 440 nm absorption of the silver sol after a pulse which produced 4 x mol dm-3 reducing radicals In the experiment of Fig. l(a), the silver sol was exposed to a pulse of high-energy electrons. The solution was under an atmosphere of nitrous oxide.N20 reacts rapidly with the hydrated electrons generated: N2O+eZq+H,O + N,+OH'+OH-(4 Under these circumstances, hydroxy radicals are the only reactive species from the radiolysis of the aqueous solvent. They react with the dissolved propan-2-01 during the pulse to produce 4 x lov6mol dm-3 2-hydroxypropan-2-yl radicals; these organic radicals diffuse to the colloidal particles and transfer electrons to them: Agn+ x(CH3)&OH -+ Ag:-+ x(CH3)*CO+ xH' (vi) This process was followed quantitatively in the millisecond range by recording the accompanying increase in the electrical conductivity of the solution. A typical conduc- tivity vs. time curve is shown in Fig. 2(a);one can see that the process is complete after ca.5 ms. No further change in conductivity up to 10 s after the pulse was observed (which means that the stored electrons lived for a long time). Fig. 2( b) shows the change in the 440 nm absorption of the solution. The absorption decreases at this wavelength when the radicals react with the colloidal particles. Fig. l(a) shows the changes in optical absorption at varous wavelengths (measured 5ms after the pulse). It can be seen that negative signals are present at longer wavelengths and positive signals at shorter ones. In other words, a blue shift of the plasmon absorption band occurs upon electron donation to the particles. Similar experiments were carried out with a solution in which positive holes were injected by the radicals into the silver particles. This solution did not contain organic additives, the OH radicals generated could therefore react directly with the colloidal particles [Fig.l(b)] One can see that positive hole injection leads to an increase in absorption at longer wavelengths and to a decrease at shorter ones, i.e. to a red shift of the plasmon absorption band. Fig. 3(a) shows the absorption spectrum of a sol before and after various degrees of positive hole injection. In this experiment, the solution was exposed to various doses of y-radiation, which produced much greater concentrations of OH radicals than in the A. Henglein, €? Mulvaney and T. Linnert 0300 400 3SO 600 650 A/nm Fig. 3 (a) Absorption spectrum of 1 x mol dm-3 silver sol before and after oxidation by OH radicals.The percentage of Ag atoms oxidized is noted on the curves. (b) Absorption of a 1x loA4mol dm-3 silver sol after addition of various amounts of Ag+ ions (as perchlorate). A, 8x B, 3 x C, 2 x mol dm-3 Ag+ above single-pulse experiment. It is seen that the injection of positive holes by the OH radicals causes a decrease, broadening and red shift of the plasmon absorption band. The broadening and red shift are also observed when silver ions are added to a silver sol as can be recognized from Fig. 3(6). The silver ions are obviously strongly bound to the silver particles, their positive charge being shared with all the atoms in the particle. The changes in the intensity and shape of the plasmon absorption band can be explained quantitatively by the change in N,, the density of the electron gas in the metal particles as electrons are donated or removed by free radical attack.Or, in other words, the properties of the plasmon absorption band are determined by the position of the Fermi level. The above results on colloidal particles are thus understood in the same terms as the results of experiments on the electromodulation of the optical properties of compact metal electrodes.6 According to the optical absorption coefficient of a collection of uniform spheres, very small compared to the wavelength of light and imbedded in a medium of refractive index no,is where N is the particle concentration, V the volume per particle, Chemistry of Silver Aggregates the wavelength of maximum absorption, and is the plasma wavelength.is a constant related to the bandwidth at half maximum absorption, w, w = A;/ A, = ( E~+2n;)c/2a (5) where u=Nee2R/mu (6) is the particle conductivity and m the effective electron mass (0.995 mo),and R is the mean free path of the electrons in the colloid, normally taken to be 1/R = (11 r + 1/ roo), where roois the mean free path in bulk silver (520 A). u is the electron velocity at the Fermi level. A decrease in N, by positive hole injection decreases (+ [eqn. (6)] and hence leads to an increase in bandwidth [eqn. (5)] as observed in Fig. 3. The decrease in N, also causes an increase in the wavelength of the absorption maximum [Fig. l(b)]. Electron donation leads to an increase in N, and hence to a blue shift of the maximum [Fig.1(a)]. The absorption coefficient at A = A, increases with N,, i.e. with decreasing bandwidth w. The change in N, in the experiments on electron donation [Fig. l(a)] can be calculated as the number of transferred electrons is known from the increase in conduc- tivity {to which practically only the protons formed [reaction (iv)] contribute} and using the literature values for V, no, e0 and u.'.~Good agreement between calculated and measured absorption spectra was obtained which shows that the change in Ne is indeed the main factor that determines the optical changes upon electron donation and positive hole injection. Within a small wavelength range around 315 nm (3.9 eV) in Fig. 1, weak bleaching signals are observed upon electron accumulation [curve (a)]and weak absorption signals upon positive hole injection [curve (b)].These effects are attributed to the influence exerted by the changing electron density in the metal particles on the discrete interband transitions. Direct d-band to Fermi level transitions occur at hv > 3.9 eV, and the filling of conduction band states at the Fermi level by injected electrons would result in a shift of the transition to higher photon energies, i.e. to a bleaching in the difference spectrum. Conversely, positive hole injection causes the onset of interband transitions to shift to lower energies, i.e. to absorption in the difference spectrum. React ions with Nucleoph ilic Reagents Fig.4 shows the plasmon absorption band of a silver sol before and after addition of various amounts of sodium hydrogen sulfide. It can be seen that even small amounts of SH-ions decrease the intensity and broaden the band significantly. At concentrations above 5 x lo-' mol dm-3 the band does not change substantially. These findings show that the surface plasmon oscillation in the silver particles is strongly affected by modifications to a small fraction of surface atoms. When the silver sol containing SH-ions (5 x mol dm-3) was exposed to air, the broadened absorption band disappeared almost instantaneously and the spectrum of colloidal silver sulfide, Ag2S, appeared. In other words, the silver particles were rapidly oxidized by O2 when SH-ions were present.Similar observations have been made using other nucleophiles such as CN-and NH:. The following mechanism for the nucleophile-catalysed oxidation of silver was proposed.: Ag,& + X === Ag:-dy'+ X (vii) A. Henglein, P. Mulvaney and T. Linnert 37 **O I 300 400 500 hfnm Fig. 4 Plasmon absorption band of a 1x loF4mol dm-3 silver sol before and after addition of various amounts of NaSH. (a) 10, (b) 20, (c) 50 pmol dm-3 SH-(viii) In the first step [reaction (vii)] a molecule of the nucleophile donates an electron pair into an unoccupied orbital on the surface, the surface atom (which is designated by script letters) acquiring a slight positive charge 6 + and the interior of the silver particle a corresponding negative charge 6-.A certain number of surface atoms are 'pre- complexed' or 'pre-oxidized' this way until an equilibrium is reached where the negative charge accumulated inside prevents further electron donation. In the second step [reaction (viii)], oxygen picks up the excess negative charge of the interior and dissolves the metal as AgX molecules. The decrease and broadening of the plasmon absorption band by nucleophilic reagents is not yet understood in detail. One may think of two opposing effects: The negative charge inside the pre-complexed particles increases N, in eqn. (6), and the increased value of 0 should yield a blue shift of the band. On the other hand, the 6 +6-dipolar structure on the surface might be equivalent to a decrease in R [eqn (6)], and this should decrease 0, Le.cause a red shift. Generally, a small red shift was observed, indicating that the second effect was prevailing. The effective decrease in (T may also be responsible for the increase in bandwidth [eqn (5)]. If the above mechanism is correct, one could use organic electron acceptors instead of oxygen in the second step [reaction (viii)] to oxidize the pre-complexed silver particles. This was indeed found for certain quinones and nitro compounds. Fig. 5 shows a typical example. The spectrum of a 1 x lop4mol dm-3 silver sol is shown here before and after Chemist.y of Silver Aggregates 1.5 1.0 8 e s3 0.5 h/nm Fig. 5 Plasmon absorption band before and after addition of NaSH, and after subsequent ac iition of AQS.The solution was finally exposed to air addition of 2 x mol dm-3 NaSH. The spectrum after subsequent addition of 3 x mol dm-3 anthracene quinone sulfonic acid (AQS) is also shown; it contains the band at 400nm of the hydroquinone and the broad plasmon band has disappeared. This shows that the quinone has oxidized the silver atoms. The amount of hydroquinone formed corresponds, within the limits of error, to the amount of silver present. The hydroquinone is not stable towards oxygen. When the solution was finally exposed to air, the hydroquinone band disappeared rapidly and the 325 nm band of the quinone increased (above the weak background absorption of colloidal Ag2S formed). The equilibrium of eqn. (vii) can be shifted to the left-hand side by deposition of excess electrons on the colloidal particles. This can be done by generating reducing radicals in the solution.Fig. 6 shows the broad spectrum (0) of a 1x mol dm-3 silver sol containing 5 x lo-' mol ~m'-~ NaSH. The solution which also contained 0.5 mol dm-3 propan-2-01 was y -irradiated under an N20 atmosphere; 2-hydroxypropan- 2-yl radicals were produced which donated electrons to the colloidal particles. The concentration of stored electrons is indicated in the legend of Fig.6. One can see that excess electron donation indeed leads to an increase and narrowing of the band. This experiment illustrates the important fact that the extent to which nucleophilic reactions occur on the surface of a metal particle depends on the position of the Fermi level.It may be mentioned that the stored electrons are finally discharged by water molecules to yield H2 upon aging after the y-irradiation. However, this process takes place over several hours at the pH of the solution (pH9). The absorption band becomes less intense and broader and finally spectrum 0 is reached again. This shows that the changes produced by changes in the electron density on the particles are reversible. A. Henglein, P. Mulvaney and T. Linnert 1.5 1.0 1P) I 9 e5!3 0.5 300 400 500 h/nm Fig. 6 Absorption band of a 1 x mol dm-3 silver sol pre-complexed by 5 x rnol dm-3 SH-before and after electron transfer by reducing organic radicals. The concentration of the stored electrons is (a) 1.3 x and (b) 4.3 x mol dm-3 Non-metallic Clusters Optical Absorption and Electron Transfer Fig.7 shows the absorption spectrum (0) of a completely reduced silver perchlorate solution, in which small clusters absorbing at 300, 340 and 370nm are present.The stabilizer was sodium polyacrylate. The spectra after 20 and 35 days are also shown. One can see that the clusters live for many days. When they disappear, an absorption band at 400 nm develops, which is attributed to larger metallic particles. Neither the exact nature of the clusters nor the nature of the optical transition which produces the strong bands (absorption coefficient ca. lo4dm3 mol-I cm-')l0 are fully understood. Pulse radiolysis investigations have yielded some information about the kinetics of cluster formation and reasons have been given for the agglomeration number of the clusters being smaller than 10."~'2 The numerous chemical reactions observed may be summarized as follows: 1.Upon exposure of a cluster solution to air, the cluster absorption bands decay, the 370 nm band disappearing within seconds, the 340 nm band within minutes and the 300 nm band within hours. This shows that the bands are produced by different clusters. The clusters are oxidized by 0, to yield Ag+ ions. 2. Upon addition of lop3mol dm-3 CCl, to a cluster solution, the cluster bands disappear and AgCl is formed, the rate of reaction decreasing in the above order as in the reaction with 02.2This shows that the clusters are strong reducing agents.Chemistry of Silver Aggregates 0250 300 350 400 450 A/nm Fig. 7 Absorption spectrum of a silver cluster solution before and after (a)20 and (b) 35 days 3. Upon addition of 1x mol dm-3 nitromethane to a cluster solution, the 370 nm band disappears very rapidly but the 340 and 300 nm bands are unaffected. Nitromethane is an electron acceptor like CC14, although its one-electron redox potential is more negative (CCl,: ca. -0.7 V; CH3N02ca. -1.0 V).* The 370 nm cluster is obviously a stronger reducing a ent than the 340 and 300nm clusters. 4. When 1 x 10-$ mol dm-3 nitrobenzene is added to a cluster solution, all the bands disappear and an absorption band at 380 nm develops. Nitrobenzene, whose one- electron redox potential is -0.4 V, catalyses the conversion of oligomeric clusters into larger metallic particles.The following mechanism was proposed: l3 Ag, +C6H5NO2 --+ Agm-,+Ag++C6H,NO;- (ix) C,H,NO;-+Ag, --.* C,H,NO,+Ag, (XI Agn+Ag+ --.* Agn+l (xi) A nitrobenzene molecule picks up an electron from a small cluster Agm, the radical anion formed diffuses until it encounters a larger cluster Agn; it transfers the electron to it, and the Agi species then combines with the Ag+ ion formed in reaction (ix). Note that, according to this mechanism, larger clusters generally have a more positive redox potential than smaller ones. 5. As pointed out above, three redox potentials have to be considered in electron- transfer reactions on clusters.Although already p~blished,'*'~ Fig. 8 is shown here, which describes the metal electrode potential of eqn (iii) as a function of the agglomer- ation number n. It is clear that the size dependence of the potential is very complex A. Henglein, P. Mulvaney and T. Linnert -2 .o T+ + Ag, + Agn-, + Ag + e-b Y CI 8 -0.5 \ a 9 t 0-+0.5 -Ag,' ' ' '-.---, Ag + Ag++ e-n-ao+' ' ' ' ' ' ' ' ' ' ' +1.0,0 5 10 15 agglomeration (n) Fig. 8 Standard redox potential of the silver microelectrode as a function of the agglomeration number for small n. Note that the silver atom is strongly reducing, the standard potential of the system Ag'/Ago being -1.8 V. An oscillation exists in the n = 1-3 range, dimeric silver having a substantially less negative potential than its neighbours Ago and Ag,.In the early stages of Ag+ reduction (i.e. when a large excess of Ag+ ions is still present), a long-lived complex Agi+ is formed which is stable towards oxygen.2 This complex may be regarded as dimeric silver complexed by two Ag+ ions. It certainly has a more positive redox potential than naked Ag, which explains its stability towards 02. 6. Photoelectron-emission from the oligomeric clusters with large quantum yields (>lo%) was observed." Smaller yields of ca. 2% were observed for larger clusters having agglomeration numbers between 10 and 20, which absorb close to 380 nm (which, however, were very sensitive to O2 in contrast to the larger metallic particles that also absorb at 380 nm).The yield of photoelectron-emission for real metallic particles (3 nm) was found to be <0.1%. These findings demonstrate that the reactivity of silver aggregates drastically changes with size in the transition range between the atom and bulk material. Two reasons can be given for the increasing emission yield with decreasing aggregate size. 1. The Fermi potential is generally more negative in the smaller aggregates. 2. The smaller aggregates have a lower density of states in their 'conduction band'. An electron lifted into a high energy level, from which it can tunnel into an acceptor level in the aqueous solvent, finds it more difficult to thermalize to the Fermi level from where the transition into water cannot occur (for a discussion of unoccupied and occupied electronic levels in water, see ref.15). Reaction with Nucleophilic Reagents When a nucleophilic reagent is added to a cluster solution, the sharp absorption bands disappear rapidly and finally a strong absorption in the 380-420 nm range appears. In the experiment of Fig. 9, ammonia was the added nucleophile. Obviously, nucleophilic reagents promote the growth of clusters to yield larger metallic particles. As mentioned above, the clusters are stabilized by the polyanion present in the solution during their Chemistry of Silver Aggregates 2.0 a2 0,2 1.0 23 0 I I I 300 400 SO0 A/nm Fig. 9 Absorption spectrum of a cluster solution before and at (a)10 min and (b) 3 h after the addition of 0.7 mol dm-3 NH3.pH of the sol before addition, 10.5 formation. The charged groups of the usual polyanions are weak nucleophiles, which bind clusters only weakly.Upon addition of a stronger nucleophile, a competition occurs, the weaker bond to the polyanion being broken. As a consequence, the cluster is detached from the polymer. It can encounter other clusters in the solution and larger particles are formed this way. In some cases, this cluster-cluster interaction to form larger particles is rather slow and may require hours. During the cluster growth new absorption bands appear and disappear upon further growth. A typical example is shown in Fig. 10, where 4x mol dm-3 KCN was added to a cluster solution One can see that the original small cluster bands disappear within seconds and that a new band at 370 nm is present after 1 min.However, this band then decays and a new band at 440 nm appears after 20 min. When the sample is aged further this band also vanishes and finally the plasmon absorption band of metallic particles is present. The intermediate bands are attributed to larger clusters, which are formed from the smaller ones. As they carry CN-groups on their surface they cannot bind to the stabilizing polymer chains and therefore undergo further agglomeration. The great reactivity of clusters towards nucleophilic reagents is ascribed to the fact that practically all atoms in a cluster are ‘surface’ atoms, their coordinative unsaturation being most pronounced.One may expect that this property of the clusters makes them especially efficient as catalysts in various chemical systems. Conclusisn In the theoretical calculations on the plasmon absorption band of silver particles, one generally uses the electronic properties of the bulk metal and the refractive index no of A. Henglein, P. Mulvaney and T. Linnert 2.0 6) e 1.0 2 % 300 400 500 h/nm Fig. 10 Spectrum of a cluster solution before and (a) 1 and (b)20 min after the addition of 4 x lod4mol dm-3 KCN the matrix in which the particles are imbedded. In cases, where an additional component is present which interacts with the particles, the calculated and measured bands some- times do not agree; this is often accounted for by choosing another value of no for the surface region.However, as was already noted by Berry and SkillmanI6 in their early studies on the optical absorption of silver particles in dye solutions, the changes in the optical absorption could not be explained by changes in no. In fact, these authors concluded that ‘a reasonable explanation is that the electronic properties of the small spheres differ from those of the bulk’. In our present paper, the change in the electronic properties of the particles is regarded as the essential cause of the optical effects accompanying the chemisorption of various additives, although it has to be conceded that changes in no should also be considered in more detailed studies. As far as we know, a quantitative theoretical model for the effects of chemisorption of molecules on the electronic properties of the metal particles has not yet been developed.Such a theoretical treatment would be highly desirable. References 1 A. Henglein, Ber. Bunsenges. Phys. Chem., 1990, 94, 600. 2 A. Henglein, T. Linnert and P. Mulvaney, Ber. Bunsenges. Phys. Chem., 1990, 94, 1449. 3 J. K. Sass, R. I(. Sen, E. Meyer and H. Gerischer, Surf: Sci,, 1974, 44, 515. 4 A. Henglein and J. Lilie, J. Am. Chem. Soc., 1981, 103, 1059. 5 M. Haase, H. Weller and A. Henglein, J. Phys. Chem., 1988,92, 4706. 6 W. N. Hansen and A. Prostak, Phys. Rev., 1968, 174, 500. 7 W. T. Doyle, Phys. Rev., 1958, 111, 1067. 8 R. H. Doremus, J. Chem. Phys., 1965,42, 414. 9 U. Kreibig, J. Phys. F, 1974, 4, 999. Chemistry of Silver Aggregates 10 T. Linnert, P. Mulvaney, A. Henglein and H. Weller, J. Am. Chem. Soc., 1990, 112, 4657. 11 P. Mulvaney and A. Henglein, J. Phys. Chem., 1990,94,4182. 12 P. Mulvaney and A. Henglein, Chem. Phys. Lett., 1990, 168, 391. 13 A. Henglein, Chem. Rev., 1989,89, 1861. 14 A. Henglein, in Modem Trends of Colloid Science in Chemistry and Biology, ed. H-F. Eicke, Birkhauser Verlag, Basel, 1985 p. 126-146. 15 A. Henglein, Faraday Discuss. Chem. Soc., 1977, 63, 80. 16 C. R. Berry and D. C. Skillman, J. Appl. Phys., 1971,42, 2818. Paper 1/02587B; Received 10th April, 1991

ISSN:1359-6640    

DOI:10.1039/FD9919200031

出版商:RSC    

年代:1991

数据来源: RSC

摘要:

Faraday Discuss., 1991, 92, 45-55 Small Alloy Particles formed by Co-reduction of Soluble Precursors with Alkalides or Electrides in Aprotic Solvents James L. Dye* and Kuo-Lih Tsai Department of Chemistry and Center for Fundamental Materials Research, Michigan State University, East Lansing, MI 48824, USA Soluble compounds of transition metals and post-transition metals in dimethyl ether or tetrahydrofuran are rapidly reduced at -30 "Cby dissolved alkalides or electrides to produce metal particles with crystallite sizes from <3 to 15 nm. When two different metal salts are used, alloys or intermetallic compounds form as indicated by X-ray photoelectron spectroscopy. Confirmation by electron diffraction can be made in the case of air-stable samples. Stoichiometric amounts of the alkalide [e.g.K+( 15-crown-5)*K-] or electride [e.g. K+(15-cr0wn-5)~e-l are used, and can be prepared separ- ately or in situ. Elements used in this work range from Ta to Te. Especially with active metals, organic decomposition products are formed along with the metal particles. Introduction There is much scientific and technological interest in the preparation and properties of nanoscale metal particles. A number of methods have been developed to access this size range,' including evaporation in vacuum2 or into organic mat rice^,^ chemical vapour deposition,4" mild reduction:-* and thermal decomposition of precursor^.^ Of particular relevance to this work are low-temperature reductions of metal salts, especially in homogeneous solutions.The use of alkali metals, often in combination with aromatic compounds to form aromatic radical anions and dianions was pioneered by Ftieke and co-workers." The method is now commonly used to produce reactive metals for organometallic syntheses. A more recent development is reduction with alkali-metal organoborohydride solutions, such as NaB(Et),H, that have been shown to yield both single metals and alloys of the iron-group elements and the noble metals." We demonstrated recently the feasibility and generality of homogeneous reduction of transition metal and post-transition metal compounds with solutions of two relatively new classes of compounds, alkalides and electrides, in dimethyl ether (Me,O) or tetra- hydrofuran (THF) solutions.The method has potential applicability to metal and alloy formation in the form of powders, on inert supports, and in the pores of zeolites. The rapid reduction to the zero oxidation state produces metal particles that form colloids and aggregates with crystallite sizes that range from <3 nm to ca. 15 nm. When com- pounds of two different metals are used, intermetallic compounds or alloys can form rather than simple mixtures of the separate metals. In this paper we report the formation and identification of such compounds and alloys for two test cases and the preliminary results for a number of other binary systems. Alkalides and Electrides ' These two new classes of compounds 12-15 owe their parentage to alkali-metal solutions in ammonia, amines and polyethers.? The species e& and M- (alkali-metal anion) t The first observation of metal-ammonia solutions was described in the unpublished notes of Sir Humphrey Davy in 1808 (see ref.16a). The first published account was in 1864.'6h 45 Alkalide- Electride Reduction 1 2 18-crown-6 cryptand[ 1.1.11 ( 18C6) (Clll) have been extensively studied in ~olution.'~t Indeed, reductions that involve such species (dissolving-metal reductions) have been used in the laboratory and industrially for many years. The unique feature of alkalides and electrides is the ability to produce stoichiometric crystalline salts in which trapped electrons or alkali-metal anions serve as the anions. This is achieved" by using powerful cation complexants such as crown ethers,*' 1, or cryptands?l 2, to protect the alkali-metal cation from reduction by M- or e-.$ To first order, most solid alkalides and electrides may be viewed as close-packed large (8-10 A diameter) cations with M-or e- in the holes that are produced by such packing.The organic complexant not only permits isolation of solvent-free crystalline alkalides and electrides, but also serves to greatly enhance the solubility of alkali metals in such aprotic solvents as Me,O and THF. The relevant equilibria are: M(s) * M,f,lv+e,,v (1) in which L is the complexant and n is 1 or 2. Reaction (3) lies far to the right-hand side, which tends to drive reactions (1) and (2) to the right also. The effect on metal solubility is dramatic.For example, potassium is not perceptibly soluble in Me20 (lo-' mol dmP3 solutions would be easily discernible by eye because of the intense colours of eLOl, and MLolv). In the presence of stoichiometric amounts of 15C5, at least 0.5 mol dm-3 solutions of K+(15C5)2 K- can be formed at -25 "C. The solubility is presumably even higher than this since no precipitate forms upon cooling this solution to -78 "C. This dissolution process is slow, however, and potassium powders or films require about 1h at -25 "C to form concentrated solutions. By contrast, pre-synthesized alkalides and electrides dissolve rapidly. Feasibility Studies A brief report that describes the use of alkalides and electrides to prepare nanoscale metal particles was published recently.22 In it we demonstrated the feasibility of aprotic reductions of compounds of R,Au, Cu and Te, to give metal particles that could be t A number of conferences under the generic name, 'Colloque Weyl', have been held that deal with metal soIutions.18 .$ The abbreviation nCm will be used for crown ethers; n is the total number of atoms in the ring and m is the number of oxygen atoms, Cnmo represents a cryptand with n, m,o oxygens in each of the three strands. J.L. Dye and K.-L. Tsai Fig. 1 Kontes double tube (H-cell) used to prepare metal and alloy particles. The precursor compound was introduced into compartment A, the alkalide or electride (or metal +complexant) were put into B, solvent was distilled in and the solutions were mixed rapidly washed with air-free methanol without substantial surface oxidation.Elemental Si, Zn, Ga, In and Sb showed surface oxidation when exposed to methanol, but the sub-surface composition contained the elements, as indicated by the X-ray photoelectron spectros- copy (XPS) of argon-ion sputtered samples. Metallic samples of Fe, Ni, Mo and W could be produced, but a methanol wash resulted in complete oxidation. Samples of TiCl, were apparently reduced, as indicated by the disappearance of the blue colour of the alkali-metal solution until an excess was added, but complete oxidation of unwashed samples occurred before XP spectra could be obtained. When mixtures of precursor compounds were used, alloy or compound formation occurred.This was demonstrated for Au-Zn, Au-Cu, Cu-Te and Zn-Te by their XP spectra, and the compounds AuZn and AuCu were identified by their selective-area electron diffraction (SAD) patterns. The details of these experiments and the extension to other binary systems are the subjects of this paper. Experimental The aprotic solvents Me20 or THF were distilled from solutions of excess Na-K and benzophenone into stainless-steel (Me20) or glass (THF) storage vessels, from which 10-30 cm3 were withdrawn by distillation as needed. A commercially available? H-cell (Fig. 1) was used to carry out the reduction. The solid or liquid sample of the com- pound(s) to be reduced was added to one arm and either a presynthesized sample of the alkalide or electride or the alkali metal and a stoichiometric amount of complexant was added to the other arm in the He- or Ar- filled drybox.In the feasibility studies described previously, small amounts (a few mg) of reactants were used to produce 10-2-10-3 mol dm-3 solutions. This has now been scaled up to millimole amounts when precursor solubility permits. The potasside, K’( 15C5)2K-was used as the reductant because of its high solubility and ease of preparation. To carry out homogeneous reactions, the desired compound23 and the alkalide were separately dissolved in MezO i-Kontes ‘double tube’, Kontes Co., Vineland, NJ. Alkalide- Electride Reduction Table 1 Approximate solubility ranges of precursor halides in dimethyl ether at -20 "C a low (<0.005 mol drn-,) moderate (0.005-0.025 mol dm-,) high (>0.025 mol drn-,) wc12 MoI~ wc1, FeCl,( CQ.0.02 mol drn-,) TaC15(CQ. 0.5 mol drn-,) TeCl, ,TeBr, NiC12 Ptc14 cuc12 LaI,( cu. 0.006 mol drn-,) MoC15(>0.3 mol drn-,) AuC13( >0.03 mol drn-,) AgCl InCl, ZnI, GaCl, TiCl,, TiBr, a Moderate and high solubilities are suitable for this method. When no concentration is specified, the solubility is in the range indicated, but no attempt was made to estimate it. LaI, may be more soluble, but a slow reaction with the solvent occurred to yield a precipitate. or THF and the solutions were mixed rapidly. Reduction was immediate to give colloidal solutions when dilute and precipitates when concentrated. Synthesis The major challenges of this method are related to the solubility and solution stability of the precursor compounds, the stability of the alkalide or electride solution, and the removal of by-products of the reduction reaction. Solubility data are scarce for THF and virtually non-existent for Me20. Table 1 gives approximate solubility ranges of some compounds of interest in this work.When chlorides are used, the ionic products, such as KCl and K+(15CS),Cl-, are only slightly soluble in Me20 and a second washing solvent is required. When water or methanol cannot be used because of oxidation of the metal, liquid ammonia is often effective in removing the ionic byproducts while avoiding oxidation. After precipitation of mixtures of the metal or alloy and the ionic byproducts, the Me20 was removed by distillation and NH3 was distilled into the H-cell.By successive decantation through the frit and back-distillation of ammonia, it was possible to wash the metallic precipitate five or six times. In some cases, KCl and/or K+(15C5)2Cl- were entrained with the precipitate and could not be removed by washing the initial product. When this occurred, the material was removed from the H-cell in the glove box, ground with an alumina mortar and pestle, reintroduced into a clean H-cell, and again washed with liquid ammonia. This procedure effectively removed the chlorides. For less reactive elements such as Pt, Au, Cu, Te, it was possible to wash the precipitate with oxygen-free methanol or water after opening the cell in a nitrogen-filled glove bag.When the metal particles were too small and loose to filter easily they were separated from the washing solvent by centrifugation in the glove bag. Characterization Powder X-ray diffraction (XRD) provides direct evidence for metal particle formation when the crystallite size is >3 nm. A Rigaku D/max-RBX rotating anode diffractometer with a graphite monochromator (Cu-Ka radiation) and scintillation counter were used. The width of the strongest XRDline can be used with Scherrer's equation24 to determine the average size for crystallites between 3 and 20nm. For X-ray amorphous metals, common in this work, XRD provides only an upper size limit of ca. 3 nm, but cannot J. L. Dye and K.-L.Tsai be used to verify metal formation. The by-product salts are always crystalline, so that XRD provides a convenient measure of the efficiency of removal of these salts. By contrast, organic decomposition products are not usually detectable by XRD. XPS, including the observation of Auger lines, provides the most general way to identify the presence of metals or alloys. A Perkin-Elmer PHI 5400 ESCA/XPS spec- trometer system was used with Al-Ka or Mg-Ka radiation. A sample neutralizer and an argon-ion gun were used when necessary. Even when the initial by-products of reduction could not be removed by washing, metal formation could be verified by XPS studies of the mixture. When the surface was oxidized, argon-ion sputtering could be used to probe the sub-surface composition.Finally, oxidation of the metal could be followed by XPSbefore and after exposure to air. Charging problems could be minimized by using the neutralizer and the position of the C 1s peak at 285.0eV as an internal standard. The effect of sample charge could be virtually eliminated by evaluating the modified Auger parameter, a’,and by using two-dimensional chemical state Transmission electron microscopy (TEM) was used to obtain particle sizes and the morphology of aggregation. A JEOL 1OOCX-I1 transmission electron microscope operat- ing at 100 KV was used for imaging, energy dispersive spectra (EDS) and selective-area electron diffraction (SAD). It was not possible to prevent air oxidation with the instrument used so reactive metals could not be observed directly in the TEM.But the particle size distribution of the products of oxidation could be studied. It was also possible with this instrument to obtain elemental analyses of microscopic crystallites by EDS. This allowed us to verify the presence of both components of an alloy and to identify regions that contain only one metallic element. For products that could be briefly exposed to air, SAD provided identification of the structure (for simple structures). By using all of the characterization techniques described above, it was often possible to identify the major phase or phases resulting from reduction and to obtain particle size information. But the high reactivity of nanoscale metal particles and the presence of organic complexants and solvents often resulted in the inclusion of organic decomposi- tion products. We have not yet examined these products extensively, but have used FTIR spectroscopy to demonstrate their presence and to identify major IR-active groups on the surface of the metal particles.Results and Discussion Single Elements The preparation of small metal particles of a single element has been described pre- vio~sly.~~While nearly any soluble metallic compound can be reduced by alkalides or electrides in Me,O or THF, the less reactive noble metals, Au, Pt etc., are easiest to isolate and characterize. Because AuCl, is very soluble in Me20 and can be easily obtained as the anhydrous compound, and because the product, Au metal, is unreactive, gold was used to develop the methodology.Crystallite sizes (and particle sizes) were ca. 10nm so that XRD identification was possible. When dilute solutions mol dm-3) of AuC13 were reduced, colloidal gold that precipitated only slowly was formed in Me20. The particle size was smaller when more dilute solutions were used. After removal of Me20 by distillation, the residue again formed a colloidal solution when taken up in water. Fig. 2 shows typical XRD lines used to determine average particle sizes for Cu, Te and Au. Also shown are the size distributions for Cu and Au from electron microscopy. There is very good agreement between the average crystallite size and the particle size in both cases. The average particle size of Te from the reduction of TeBr, was 12nm, according to the XRD linewidth.Electron micrographs showed that the Te particles are rounded cylinders of width ca. 6 nm and length ca. 25 nm. Alkalide- Electride Reduction -Im mY 18 40 'I -3 -2 -1 0 1 2 3 24 48 72 96 120 144 168 >192 2O/ degrees particle size/A Fig. 2 The most intense XRD line for (a)Cu, (b)Te, (c) Au, produced by reduction of precursors with alkalides or electrides. Calculated particle sizes are 57, 117 and 94 A, respectively. Particle size distributions in 8, are given for (d)Cu (711 particles counted), and (e) Au (2500 particles counted) Both Au and Cu could be used to test the XPS method of identification. Removal of the ionic byproducts was achieved by washing with oxygen-free methanol The XPS data indicated little or no surface oxidation, but exposure to air oxidized the copper to CUO.In addition to the non-oxophilic metals, Au and Pt, the post-transition elements Sb and Te could also be prepared as non-oxidized particles that could be washed with methanol. The iron-group elements, Fe, Cr, Ni, Co could be prepared as metal particles with some surface oxidation, but removal of the by-products with methanol resulted in complete oxidation. Recently, we used liquid ammonia as a washing solvent for the product of reduction of FeCl, ( Fe2Cl,) by K'( 15C5)2K- in Me20. Metallic iron results from this procedure as demonstrated by both XRD and XPS studies. Presumably the other elements in this group could also be washed with NH3.The most challenging test of this method is the preparation of high-melting oxophilic metals such as Ta, W, Mo and Ti. We have not yet been able to demonstrate the formation of metallic Ti. The other three metals can be obtained as X-ray amorphous solids. We have demonstrated that millimolar amounts of TaCl, can be reduced in Me20 with K+( 15C5)2K-. The reduction product can be washed free of ionic byproducts with liquid ammonia and used as a highly reactive precursor for subsequent reactions. The samples are pyrophoric, as anticipated for such an oxophilic element as tantalum. Its reactive nature also results in some decomposition of the organic complexant and/or solvent so that ammonia-insoluble organic compounds are included in the precipitate.These have not yet been identified. J. L. Dye and K.-L. Tsai Alloys and Compounds It is clear that a wide variety of elements can be prepared as small metallic particles by reducing dissolved compounds in Me20 or THF with M-or e&. Because of the small particle size (and probably even smaller initial particles) it was of interest to examine the products of reduction of homogeneous mixtures of two or more metal salts. If intermetallic compounds could be formed rather than mixtures of metals, this would open up an opportunity to synthesize new materials, since many intermetallic compounds that are stable at low temperatures decompose at high temperatures. A major advantage of the present method is the ability to control overall stoichiometry simply by adjusting initial compositions.All binary systems studied so far yielded some evidence for alloy or compound formation. But all have been X-ray amorphous, thus preventing identification of the compounds by XRD. The photoelectron and Auger peaks can be used to distinguish intermetallic compounds from a mixture of the pure metals, but cannot provide positive identification. When the sample can be briefly exposed to air without oxidation, it is possible to use electron microscopy, EDS and especially SAD to identify compounds. Au-Zn The identification of the known compound AuZn by XPS and SAD was described previously.22 The compound forms both when ZnI, is present in excess and when AuCl, is at the higher concentration.Although only the Au XPS peaks were shown in ref. 22, a shift of the Zn peak also occurred upon compound formation. The electron diffraction lines agreed with the simple cubic structure of AuZn (CsCl type). Au-CU Formation of a compound in the Au-Cu system was suggested but not proven by the XPS data. The Au 4f7/2 and 4f5/2 XPS lines are near those of pure Au, and the Cu 2p3/2 XPS line at 932.3 eV is close to that of Cu (932.4) and Cu20 (932.2).27 However, the modified Auger parameter, a',which is free of charging errors, is 1849.9 eV, cu. 1.6 V below those of copper metal and CuO but close to the value for Cu20 (1849.5 eV). Thus, the photoelectron and Auger data are not definitive in this case. The SAD results are, however, clear and unambiguous.The diffraction pattern is of the same type as that of AuZn and corresponds to a simple cubic structure with each Au atom at the centre of a cube of Cu atoms and vice versa. The patterns of the two known higher temperature phases of AuCu (tetragonal and othorhombic) would be very different from that observed. The cubic unit cell has a = 2.95 A compared with 3.196 A for Tellurides When particles of Au, Cu or Te were separately prepared by reduction of AuCl,, CuC1, and TeBr,, respectively, :he XRD pattern always showed peaks of the metal (Fig. 2). Reduction of mixtures gave, however, X-ray amorphous products unless an excess of one component was used. This suggested the formation of alloys or compounds with either random occupancies or crystallite sizes <3 nm.The reduction of a mixture of CuC1, and TeBr, with K'(15C5),eP in Me20 yielded a powder that could be washed with air-free methanol without significant oxidation. The Cu 2p3/2 and 2pIl2 binding energies from the XPS of the product, 832.5 and 952.3 eV respectively, are indistinguish- able from those of metallic copper, and the L3M45M45 Auger peak is at nearly the same energy as for pure CU.,~ The modified Auger parameter, a',1850.5 eV, is slightly lower than the average literature value of 1851.5eV. Note, however, that one paper lists a Alkalide- Electride Reduction 96.0 92.0 88.0 84.0 binding energy/eV Fig. 3 XP spectra of the reduction product of a mixture of TiC1, and AuCl, in Me20 (a) and after washing with air-free methanol (b) value of 1850.3 for this parameter.29 Thus, the XPS and Auger data for the binary system Cu-Te do not distinguish between compound formation and mixtures of the pure metals.They do show, however, that the original compounds were reduced. The case for compound formation in the Au-Te system is somewhat stronger. The Au 4f,,, and 4f5/2 peaks in the binary system are nearly twice as wide as for pure Au, although the peak positions are essentially the same as those of the pure metal. Also the value of a’for Te in the mixture (1064.5 eV) is at least 0.4 eV lower than that of elemental teli~rium.~~ We conclude, on the basis of the XRD results and the XPS data, that alloy or compound formation occurs in the Au-Te system and probably also in the Cu-Te system.Ti-Au When a solution of TiCI4 in Me20 is mixed with a solution of an alkalide or electride, immediate bleaching of the blue colour occurs until an excess of the reducing agent is added. Apparently this very oxophilic element reacts with the complexant or with Me20, resulting in reoxidation. The particle size of the final product is <3 nm as indicated by both XRD and TEM. While this technique could be useful in providing small oxide or alkoxide particles, it does not yield metallic titanium. The reduction of a mixture of TiCl, and AuC1, yielded the XPS results shown in Fig. 3. The unwashed product [3(a)]gave broadened Au 4f7,* and 4f5,, peaks at the positions observed with pure gold, (83.7 and 87.4eV) and shoulders at ca.84.8 and 88.4 eV. Upon washing the sample with methanol, the XPS pattern [Fig. 3(b)]reverted to that of pure Au. The possible formation of a Ti-Au compound is suggested by these results, but further work, especially a study of the Ti XPS data, needs to be done to verify this conclusion. The results could be significant for catalysis if it were possible to produce small noble metal particles on a finely divided Ti02 or Ti(OR)2 support. J. L. Dye and K.-L. Tsai 7G.O 715.0 705.0 28.0 26.0 24.0 22.0 binding energy/eV Fig. 4 XPS results for the Ta-Fe system after reduction of FeC1, and/or TaC& with K+(15C5I2K-in Me20 at -30 “C; product washed with anhydrous NH3: (a) Fe XPS from 2: 1 FeCl, to TaCl,; (b)Fe XPS from FeC13 alone; (c) Ta XPS from 2 :1 FeCl, to TaCl, :(d)Ta XPS from TaC1, alone Ta-Fe Refractory and oxophilic metals, such as Ta, W and Mo present severe challenges to the method described in this paper.They require strong reducing agents, have multiple oxidation states and have such a high affinity for oxygen that small particles are expected to be very reactive. The ‘intertness’ of the bulk metals is due to a compact oxide layer rather than to an inherent lack of reactivity. In spite of these difficulties, our initial results showed that Mo and W metals could be produced by reduction of MoCl, and WC16, respectively, but, as expected, washing with methanol resulted in complete oxidation. Considerable effort has been expended in an attempt to prepare millimole quantities of small tantalum particles for use as precursors in ternary nitride synthesis. Purified TaCl, is very soluble in Me20 and reacts immediately in an H-cell (Fig.1) with K+(15C5)2K- prepared in situ.Since 5 moles each of K and 15C5 are required per mole of TaCls, copious amounts of KC1 and Kf( 15C5)2C1- are produced and are only slightly soluble in Me20. Most, but not all, of the by-products can be removed by washing with anhydrous liquid ammonia. The product is Ta metal together with some unidentified organic material. The fine black powder is pyrophoric and reacts immediately with water and methanol. It can, however, be handled and stored in an inert-atmosphere box without apparent change.The XRD pattern shows the product to be amorphous, but the XPS peaks [Fig. 4(d)] provide identification as Ta metal. Pyrophoric iron can be readily prepared by reduction of FeCl, in Me20. When solid FeCl, was present in contact with the saturated solution during reduction with an excess of K+(15C5)2K-, crystalline iron was produced as indicated by the XRD pattern. The by-products could be removed by washing with liquid ammonia. More dilute solutions yielded X-ray amorphous iron. The XPS peaks [Fig. 4(b)] show the presence of both metallic iron and oxidized iron on the surface of the particles. In an attempt to prepare the known compound Fe2Ta, mixtures of FeCl, and TaCl, in a 2: 1 mole ratio were reduced with K+(15C5)*K- in Me,O and the product was washed with NH3.The product has not yet been identified, but the XPS data show the presence of Ta, Fe and K in a cu. 1:2 :3 mole ratio. The Fe and Ta XPS patterns are shown in Fig. 4(a) and (c). The latter shows a shift of the Ta 4f7/2 peak from 21.6 eV in Ta metal to 23.8 eV. Oxidation of the mixture in air shifts the peak to 24.8 eV. The Fe 2p3/2 peak at 706.3 eV (Fe metal =706.8 eV) shifts to 709.7 eV upon oxidation (Fe203= 71 1 eV). The K 2p3/2 peak at 292.6 eV does not shift upon oxidation. A substantial Alkulide- Electride Reduction carbon peak at 285.0 eV decreases dramatically upon argon ion sputtering. The results show that some type of complex alloy or compound formation occurs in the K-Ta-Fe system, but the product or products have not been identified.Conclusions In spite of the preliminary nature of these experiments, it is clear that reduction of pure compounds in solution in MezO or THF with dissolved alkalides and electrides leads to nanoscale metal particles. The strong reducing ability of alkalides and electrides makes the method very general and permits metal production across the transition metal and post-transition metal series, from Ta to Te. It is difficult to positively identify alloys and compounds with particle sizes too small to yield an XRD pattern and with reactivities too high to load into the available TEM without oxidation. While the less reactive compounds AuCu and AuZn could be characterized, a number of other binary systems yielded powdered samples whose XPS and Auger patterns suggested alloy or compound formation, but did not provide identification.Additional experiments, such as sintering the products to improve crystal- linity, anaerobic TEM studies, and thermal studies (DSC and TG) are in progress. This research was supported by the U.S. National Science Foundation Solid State Chemistry Grant No. DMR 87-14751 and by the MSU Center for Fundamental Materials Research. We acknowledge use of instrumentation in the Center for Electron Optics and the Composite Materials and Structure Center at Michigan State University. J.L.D. carried out some of the experiments described in this paper at Cornell University under a John Simon Guggenheim Fellowship. We are grateful to Francis J.DiSalvo and Scott H. Elder for their encouragement and assistance. References 1 See, e.g. MRS Bulletin, December, 1989 and January, 1990 that are devoted to the methods used for small particle synthesis. 2 W. B. Phillips, E. A. Desloge and J. G. Skofronick, J. Appl. Phys., 1967, 39, 3210. 3 K. J. Klabunde, Y.-X.Li and B. J. Tan, Chem. Muter., 1991, 3, 30 and references therein. 4 P. G. Fox, J. Ehretsman and C. E. Brown, J. Catal., 1971, 20, 67. 5 T. W. Smith and E. Wychick, J. Phys. Chem., 1980,84, 1621. 6 J. J. F. Scholten, J. A. Konvalinka and F. W. Beekman, J. Catal., 1973, 28, 209. 7 R. M. Wilenzick, D. C. Russel, R. H. Morris and S. W. Marshall, J. Chem. Phys., 1967,47, 553. 8 D. W. Mackee, J. Phys. Chem., 1967,67, 841. 9 F. Fievet, J.P. Lagier and M. Figlarz, MRS Bulletin, 1989 (Dec.), 29. 10 R. D. Rieke, Science, 1989, 246, 1260 and references therein. 11 H. Bonnemann, W. Brijoux and T.Joussen, Angew. Chem. Int. Edn. Engl., 1990,29, 273. 12 J. L. Dye, Prog. Inorg. Chem., 1984, 32, 327. 13 J. L. Dye, and M. G. DeBacker, Annu. Rev. Phys. Chem., 1987, 38, 271. 14 J. L. Dye, Pure Appl. Chem., 1989,61, 1555. 15 J. L. Dye, Science, 1990, 247, 663. 16 (a)P. P. Edwards, Adu. Inorg. Chem. Radiochem., 1982,25, 135; (b)W. Weyl, Ann. Phys. Chem., 1964, 197, 601. 17 J. C. Thompson, Electrons in Liquid Ammonia, Oxford University Press, 1976. 18 Meral-Ammonia Solutions, Colloque Weyl I, ed. G. Lepoutre and M. J. Sienko, W. A. Benjamin, New York, 1964; Metal-Ammonia Solutions, ed.J. J. Lagowski and M. J. Sienko, IUPAC, Butterworths, London, 1970; Electrons in Fluids, ed. J. Jortner and N. R. Kestner, Springer-Verlag, Berlin, 1973; Colloque Weyl ZV. Electrons in Fluids -The Nature of Metal-Ammonia Solutions, J. Phys. Chem., 1975, 79, (26); Can. J. Chem., 1977, 55, (1 1); J. Phys. Chem., 1980, 84, (10); Colloque Weyl VI. The 6th Inr. Conf. on Excess Electrons and Metal-Ammonia Solutions, J. Phys. Chem., 1984, 88. 19 B. Van Eck, L. D. Le, D. Issa and J. L. Dye, Inorg. Chem., 1982, 21, 1966. 20 C. J. Pedersen, J. Am. Chem. SOC.,1967,89, 2495, 7017. 21 J.-M. Lehn, J. Inclusion Phenom., 1988, 6, 351 and references therein. 22 K.-L. Tsai and J. L. Dye, J. Am. Gem. SOC.,1991, 113, 1650. J. L. Dye and K.-L. Tsai 23 See ref. 22 for purity and sources of precursor compounds. 24. H. P. Klug and L. E. Alexander, in X-ray Diflraction Procedures for Poiycrystalline and Amorphous Materials, Wiley, New York, 1962, pp. 491-538. 25. C. D. Wagner, L. H. Gale and R. H. Raymond, Anal. Chem., 1979,5,466. 26. C. D. Wagner and A. Joshi, J. Electron Spectrosc. Relat. Phenom. 1988,47, 283. 27 Handbook of X-Ray Photoelectron Spectroscopy, Perkin-Elmer Co., 1979. 28 W. F. McClune, Powder Diflraction File, Set 30-608; JCPDS, International Centre for Diffraction Data: 1601 Park Lane, Swarthmore, PA 19001, 1989, p. 771. 29 J. C. Fuggle, E. Kaline, L. M. Watson and D. J. Fabian, Phys. Rev. B, 1977, 16, 750. Paper 1/02598H; Received 23rd May, 1991

ISSN:1359-6640    

DOI:10.1039/FD9919200045

出版商:RSC    

年代:1991

数据来源: RSC

摘要:

Faraday Discuss., 1991, 92, 57-67 Catalytic Hydrogenation of Cyclohexene: Liquid-phase Reaction on Rhodium M. Boudart" and D. J. Sajkowskit Department of Chemical Engineering, Stanford University, Stanford, CA 94305, USA The turnover rate u, for the liquid-phase hydrogenation of cyclohexene on Al,O,-supported clusters of rhodium containing cu. 12 atoms was found to be the same as that on supported rhodium clusters containing 150 atoms or more. The smallest rhodium clusters were examined by chemisorption of H2 and CO, by infrared spectroscopy of adsorbed CO, and by extended X-ray absorption fine structure, which indicated an average coordination number of five for the first shell. The results show that ligand effects for metallic clusters containing only a few atoms have no observable effect on the measured rate of the structure-insensitive reaction under study.At the Faraday Discussion on heterogeneous catalysis in 1950, Otto Beeck expanded' on observations just made by one of us2 According to these observations, the rate of hydrogenation (per unit area) of ethene on evaporated metal films could be correlated not only with the lattice parameter of the metals' (a geometric effect) but also with the percentage d character of the metallic bond according to Pauling3 (an electronic effect). As stated in our 1950 paper,2 'both interpretations, far from being mutually exclusive, lead to the same prediction concerning the activity of the metals involved'. Thus the geometric and electronic effects needed a new definition, since lattice parameter and electronic structure were clearly linked as pointed out by Pauling in his theory of metals3 The new definition, as applied to catalysis by metals, was given by Soma-Noto and Sachtler? who distinguished between an ensemble effect and a ligundeffect to replace geometric and electronic effects.In the ensemble effect, the metal provides a multiple site consisting of several surface atoms. The ensemble is required in a rate-determining step of the reaction. By contrast, the ligand effect is similar to that operating in homogeneous catalysis by transition-metal coordination complexes, where the ligands to the central metal atom control its catalytic activity. The idea of ensemble versus ligand effect has been seminal.As adapted to metallic clusters of decreasing size, it becomes related to the idea of structure-sensitive and structure-insensitive reaction^.^ A structure-sensitive reaction is affected by surface crystalline anisotropy as it requires surface ensembles. The surface concentration of these ensembles changes with metal particle size, especially below cu. 5 nm. Hence structure-sensitive reactions are affected by particle size. By contrast, a structure-insensitive reaction requires only a simple site, perhaps a single surface atom, and is affected neither by crystalline anisotropy nor by particle size down to ca. 1nm. Therefore, to assess the importance of a ligand effect for a reaction catalysed either by a bimetallic cluster or by a unimetallic cluster of a size below 1 nm, a structure insenstive reaction should be used.Such a reaction is the hydrogenation of cyclohexene in the liquid phase or in the gas phase, as catalysed by palladium'' and The mechanism of hydrogenation of cyclohexene on metallic surfaces first proposed for t Present Address: Amoco Oil Company, Amoco Research Center, P.O. Box 3011, Naperville, IL 60566, USA. 57 58 Catalytic Hydrogenation on Rhodium Clusters ethene by Horiuti and PolanyiI4 has been confirmed over the years by ab initio quantum-mechanical calculations in the case of ethene15 and by experimental work with the Wilkinson RhCl( PPh3)3 soluble catalyst in the case of cyclohexene.'6 The availability of turnover rates for the hydrogenation of cyclohexene on the Wilkinson rhodium catalyst as well as recent on the preparation of y-alumina-supported rhodium clusters containing only a few atoms led us to the present study.We have prepared y-A1203- supported rhodium clusters containing more than 100 atoms and as few as ca. 12 atoms, characterized these clusters by extended X-ray absorption fine structure (EXAFS) and other techniques, and measured the turnover rate for cyclohexene hydrogenation on our rhodium catalysts, all in order to assess the catalytic activity of the smallest clusters and to compare it with that of a Wilkinson catalyst. Experimental and Results Preparation of Catalysts A first set of three catalysts containing 0.89, 1.68 and 3.66 wt.% rhodium was prepared by impregnating very pure 77-alumina (150 m2 g-', Exxon Research and Engineering Company) with rhodium nitrate solution (Engelhard) by the method of incipient wetness.The samples were dried in air at 378 K for 12 h, then reduced by heating at 0.1 Ks-' in flowing H2 (4.9 x lo3mol g-' s-l) at atomospheric pressure to 723 K and holding at this temperature for 1 h. Dihydrogen was obtained from a Matheson Hydrogen Gen- erator, model 8325, and purified by flowing through molecular sieve and manganese traps." The samples were passivated by evacuating at Pa and 723 K for 1h, then cooling to room temperature under vacuum, and finally flowing a 1% 02-99% He mixture (Matheson) over the sample for 1h. Samples were then sieved (44 pm, 325 mesh) and stored in a desiccator.Another catalyst containing 0.45 wt.% rhodium was prepared by impregnating y-A1203 (American Cyanamid, 200 m2 g-I, specific pore volume of 0.58 cm3 g-', 99.99% purity with measured impurities in ppm of Na20 == 20, SO4=10, Fe =40, NiO and COO each 10) by the method of incipient wetness. The precursor RhC13 3H20 (Engelhard, 99.99% purity) was dissolved in deionized water (resistivity == 15 Ma cm) prior to impregnating the y-A1203. The incipient wetness point of the y-A1203 was 0.7 cm3 H20 -1 g . The sample was dried at 380K in air for 12h, then reduced in flowing H2 (5 x lo3mol Hz g-' s-') at a heating rate of 0.2 K s-' to 573 K over 1 h and holding at this temperature for 1 h, evacuated at Pa and 573 K for 1 h, and then cooled in vacuum to room temperature. The sample was passivated by flowing a mixture of 1 mol% O2in helium (Matheson) at 5 x lo3mol g-' s-' for 600 s.The sample was then exposed to air and stored in a desiccator. This catalyst preparation is very similar to that used by van Zon et all8and differs from that of Via et a1.l' only by our use of a reduction temperature that is 200 K lower. The rhodium content of this catalyst as well as the other three samples of the first set was determined by atomic absorption (Galbraith Laboratories). EXAFS EXAFS data were collected at the Stanford Synchroton Radiation Laboratory's beamline VII-3, with ring currents of 20-80 mA, and energy of 3 GeV. Passivated samples were pressed into self-supporting wafers (0.1 g crnd2) and treated in a controlled atmosphere cell described by Gallezot et aL2' The samples were re-reduced as described below in the cyclohexene hydrogenation section.All EXAFS data were collected at room tem- perature and under hydrogen at atmospheric pressure. M. Boudart and D. J. Sajkowski 59 10 -10 10 -10 I 50 90 110 K/nm-' R/ lo2pm Fig. 1 EXAFS function x(K)K us. K and RSF for Rh powder (A and E), 0.89% Rh/q-A1203(B and F), 1.68% Rh/q-A1203(C and G), 3.66% Rh/q-AI2O3(D and H) Analysis of the EXAFS data followed the methods previously described.21322 The EXAFS function x(K) was multiplied by K and then Fourier-transformed from K =30 to 150nm-' to give the radial structure function (RSF). The plot of x(K)K vs.K is shown in Fig. 1, together with the corresponding RSF for a physical mixture consisting of 10 wt.% rhodium powder (99.95% purity, AESAR) diluted with the same v-Al2O3 used in the catalyst preparation (A and E), for 0.89% Rh/77-A1203(B and F), for 1.68% Rh/77-A1203(C and G), and for 3.66% Rh/77-A1203 (D and H). The same information is given in Fig. 2 for 0.45% Rh/y-A120,. Data from the sample of Rh powder were used to obtain back-scattering amplitudes and phase shifts for curve fitting. The data shown for the 0.45% sample are an average of five scans. Only a single scan was performed for the other samples. Average coordination numbers, N1,and interatomic distances, R1, for the first coordination shell, were obtained by the methods of curve fitting22 (Table 1).Curve fitting was done on the x(K)K us. K data obtained by back-transforming the RSF data between 170 and 330 pm and then fitting in a range of K of 30-150 nm-'. Fig. 3 shows 60 Catalytic Hydrogenation on Rhodium Clusters 0 -lo] 50 100 150 0 500 1C K /nm-' R/pm Fig. 2 EXAFS function x(K)K (A) us. K and RSF (B) for 0.45% Rh/y-Al,O, Table 1 EXAFS parameters for Rh/A1203 catalysts Rh (wt.%) metal exposed (%) Nl ~~ ~ R,/pm 0.45 100 5.O 27 1 0.89 100 7.2 270 1.68 80 8.7 271 3.66 69 10.0 268 a Uncertainty is ~20%. Uncertainty is *2 pm. Klnm-' Klnm-' Fig. 3 Curve fitting of EXAFS function obtained by back-transforming the RSF for (a) 0.45% Rh/y-A1203 and (6) 3.66% Rh/7-A1203 the quality of the fit for the 0.45% Rh/y-A1203 and the 3.66% Rh/q-AI2O3samples. It is representative of the curve fitting obtained for the other two catalysts.Transmission Electron Microscopy (TEM) The instrument was a Hitachi H500H electron microscope operated at 100 keV. Speci- mens were prepared by placing a drop of solution containing the catalyst dispersed in methanol onto a perforated carbon film supported on a copper grid. The samples were not re-reduced prior to performing the electron microscopy. Micrographs are shown elsewhere.23 M. Boudart and D. J. Sajkowski 61 Table 2 H2 and CO chemisorption on 0.45% y-Al2O3and 3.66%/77-A1203 0.45 1.47 1.93 3.66 0.69 0.67 The purpose of the TEM work was to ascertain whether the samples contained large Rh particles or rafts.Even on specimens prepared from 3.66% Rh/77-A1203,no clusters larger than ca. 2 nm were visible in the electron micrographs. This was also true for the 0.89% Rh/ 77-A1203 specimen and therefore inferred to be true for 1.68% Rh/ 77-A1203. It was concluded that the dispersion of the metal on all samples was adequate. Chemisorption The percentage metal exposed for all samples was measured by hydrogen chemisorption and found to be 100, 100, 80 and 69%, for the 0.45% Rh/7-A1203 and the 0.89, 1.68 and 3.66% Rh/ q-A1203 catalysts, respectively (Table 1). The amount of hydrogen chemisorbed was obtained by extrapolating room-temperature isotherms in the range 10-40kPa to zero pressure. A stoichiometry of one chemisorbed hydrogen atom per rhodium surface atom was assumed, except when the ratio H/Rhtotalwas larger than unity, in which case the amount of metal exposed was taken as 100%.Chemisorption was performed in the volumetric apparatus described by Han~on.~~ Passivated catalysts were re-reduced using the procedure given below. Isotherms obtained at room temperature for the chemisorption of H2 and CO are shown el~ewhere.~~ Values of the ratio of adsorbed hydrogen atoms, and CO molecules, to the total number of rhodium atoms are given in Table 2 for two samples. The percentage metal exposed (Table 1) was obtained from H2 chemisorption with the assumption of a stoichiometry of 1 H atom per surface Rh atom, except for samples with a stoichiometry of more than 1 H atom per total number of Rh atoms (100% metal exposed).Fourier-transform Infrared (FIIR) Spectroscopy of adsorbed CO FTIR spectra of carbon monoxide adsorbed on rhodium were obtained with an IBM 98 spectrometer at a resolution of 2 cm-'. The IR cell and vacuum system have been described by Rivera-Lata~.~' IR spectra were obtained on self-supporting wafers (15 mg cm-2) that were made from the passivated catalysts. The same re-reduction procedure used prior to the chemisorption experiments was used to prepare the catalysts for the IR measurements. Spectra were obtained on two samples while under vacuum (used for background removal) and after 0.5 h of exposure to CO and 1.3 kPa, and a further 600 s evacuation at Pa. Spectra were obtained with the catalysts at room temperature.Fig. 4 shows the spectra for the 0.45% (A) and 3.66% (B) samples after evacuation and background removal. They are very similar to those obtained at 1.3 kPa CO. In each case the converage by CO should be at or near saturation. Four peaks are present in both spectra and have been observed in many previous IR studies of CO adsorbed on rhodi~m.~~-~~ Table 3 summarizes the frequencies and assignments of the peaks in Fig. 4. They are in good agreement with published work. There was a shift to lower frequency of the r11-p2CO band on the 0.45% Rh/r-A1203. In addition, the unequal intensities of the bands assigned to the geminal carbonyl species indicate the bond an le between carbon monoxide molecules bound to the rhodium was greater than 90°?' These bands also appear with unequal intensities in the spectra Catalytic Hydrogenation on Rhodium Clusters I I I 1 I 2100 1900 1700 2100 1900 1700 wavenumber/cm-' wavenumberlcm-' Fig.4 Infrared spectra for CO adsorbed on 0.45% Rh/y-A1203 (A) and 3.66% R/7pA1203 (B) Table 3 IR frequencies (cm-') of adsorbed CO on 0.45% Rh/y-A1203 and 3.66% Rh/v-Al203 subcarbonyl Rh/A1203 (YO) sym. asym. q'-p' co q1-p2co 0.45 2092 2022 2060 1845 3.66 2093 2025 2063 1878 shown by van't Blik et uZ.,~' but have equal intensities in many of the other published spectra. Cyclohexene Hydrogenation Hydrogenation rates were measured with the slurry reactor system described by Cheng and Boudart.I2 The only modification was to eliminate the possibility of contamination of the catalysts with vacuum grease by using Teflon seals and fittings.The following sequence was used to re-reduce the passivated catalysts prior to collecting chemisorption, EXAFS and kinetic data: first, evacuation at Pa and 373 K for 1 h, then reduction under static atmospheric dihydrogen at 373 K for 1 h, then evacuation for 30 min at 373 K, and finally cooling to reaction temperature under vacuum. The re-reduction temperature of 373 K appears adequate since reduction at higher temperatures did not increase the amount of chemisorbed hydrogen, as also seen in recent temperature-programmed reduction of Solvents used were heptane (Baker Analyzed, HPLC), cyclohexane (Baker Instra- Analyzed), methanol (Baker Analyzed, Photrex), ethyl acetate (Baker Analyzed), and benzene (Baker Analyzed, Photrex).They were degassed to remove dissolved dinitrogen and dioxygen by sparging with helium (Liquid Carbonic). Helium had been purified by passing it over manganese and molecular sieve traps, the latter held at liquid-nitrogen temperature. Cyclohexene (Aldrich, 99% purity) was purified by flowing it through activated alumina (Baker Analyzed) followed by degassing using freeze-thaw cycles as described by Cheng and Boudart.12 Cyclopentene (Wylie Chemicals, 99.996 purity), M. Boudart and D. J. Sajkowski 63 Table 4 Effect of solvent and metal exposed on turnover rate (s-l for cyclohexene hydrogenation at 298 K and pH2 = 101 kPa ~~ Rh (wt.%): 0.45 0.89 1.68 3.66 H2solubility“/ rate constant,bk/ metal exposed (%): 100 100 80 69 mol m-3 10-~cm s-’ cyclo hexane 9.4 13.7 13.3 11.7 3.83 5.59 n-heptane -20.0 21.0 15.0 4.70 6.59 ethyl acetate -12.6 16.7 11.3 3.54 6.53 methanol -14.6 15.8 11.3 3.54 6.13 benzene -1.4 ----” At 101 kPa and 298 K.Assumes surface concentration of rhodium atoms equals 10’’ cm-2. used to determine turnover rates for cyclopentene hydrogenation, was also purified by this method. Hydrogen (Liquid Carbonic, 99.99% pure) was purified by passing it over a molecular sieve trap at 77 K and a manganese trap. Turnover rates for the hydrogenation of cyclohexene in each solvent at 298 K, 101 kPa dihydrogen and a cyclohexene concentration of 0.14 mol dm-3 are shown in Table 4.Turnover rates are per rhodium surface atom as determined by hydrogen chemisorption. The order with respect to cyclohexene was zero in all solvents for conversions <%YO. The order with respect to dihydrogen was approximately unity in the pressure range of 20-101 kPa in each solvent. The apparent energy of activation was 20.9 kJ mol-’ in the temperature range of 288-310 K in cyclohexane. Table 4 also lists the solubility of H2 in each solvent, as calculated from data contained in ref. 44. Turnover rates for the hydrogenation of cyclopentene in cyclohexane were also measured at atmospheric pressure of H2 and 298 K. They were comparable to the rates reported in Table 4 for the hydrogenation of cyclohexene.This allows us to rule out cyclohexene disproportionation to benzene and cyclohexane that is observed as a side reaction during hydrogenation of cyclohexane on nickel. 12,13 Discussion Size and Structure of the Rhodium Clusters To study the effect of metallic crystal size on rate of reaction, the clusters should expose more than 50% of This was the case for our rhodium clusters: 69-100% metal exposed. These values assume, however, one hydrogen atom chemisorbed per surface rhodium This stoichiometry seems quite safe. Recent reports of hydrogen atom to total rhodium atom ratios significantly in excess of refer to clusters that are small enough to have 100% metal exposed. Thus a ratio of the number of adsorbed hydrogen atoms to the total number of rhodium atoms as large as 1.7 has been obtained on a 0.57% Rh/y-Al,O, sample which had a value of N1equal to 3.7 as determined by EXAFS.18.53 Scholten et al.54have pointed out that chemisorption data in the literature and their own work indicate the hydrogen uptake per rhodium atom increases as the crystallite size is reduced by decreasing the rhodium loading on alumina.An apparent exception is the work of Via et aZ.,” who report a value of N, equal to 1.5 from EXAFS but a hydrogen atom to total rhodium atom ratio of only 1.1 on a 0.5% Rhy-A1203 sample. Clearly, for clusters with 100% of metal exposed, EXAFS data are required to determine cluster size. The values of N, reported in Table 1 are significantly less than 12, the value for the face-centred cubic lattice of bulk rhodium.The largest value of N1= 8.0 for the 3.66% Rh/q-A1203 suggests an average cluster size of C1.3 nm for all our catalysts assuming a spherical geometry for the clusters.55 This is also in agreement with our estimate of Catalytic Hydrogenation on Rhodium Clusters the upper bound of the cluster sizes from TEM, namely 2 nm. Therefore the results of hydrogen chemisorption, EXAFS, and TEM all indicate that the rhodium clusters on our catalysts are very small (<2 nm) and are suitable for a test of the effect of particle size on catalytic activity. Note also that the values of R, reported in Table 1 for the catalysts do not differ appreciably from that of bulk rhodium, R, =268.7. This has also been reported in previous EXAFS studies conducted on rhodium provided that the measurements were made under atmospheric pressure of dihydr~gen.'~~''~~~ Our clusters are apparently not as small or have a different morphology from those prepared by Via et (11.'~ and van't Blik et al.5i753 These authors reported values of N, of 1.5 and 3.7, respectively, for rhodium supported on yA120, and metal loadings (1 wt.%.Arguments against a raft morphology as suggested for rhodium on alumina by Yates et al.56 have been presented by other^.'^-^^*^' At least, our work indicates that the 0,45%/y-A1203sample with the smallest value of Nl contains clusters with a dozen rhodium atoms of the average. Effect of Cluster Size on Rate of Cyclohexene Hydrogenation Turnover rates, v,, in Table 4 vary little for different solvents (excluding benzene) for all four catalysts in spite of a change in the percentage metal exposed from 100 to 69%.This has also been found for Pt6-' Pd," and Nil2*I3 in the gas phase and in the liquid phase, and for Pd" in the liquid phase. The near constancy of v, as the metal loading varies by a factor of ca. four and the amount of surface rhodium varies by a factor of ca. three also confirms the absence of mass and heat transfer limitations under our conditions.5s959 The conclusion that cyclohexene hydrogenation does not depend on particle size over rhodium has also been reached by Campelo et aL6' and Marques Da CruzV6* Campelo et al. used A1PO4-SiO2 as a support with H2 at 560 kPa, at 313 K and in methanol as solvent, but did not state their range of rhodium particle size.Marques Da Cruz used Rh supported on y-Al2O3 and SiOz to investigate the hydrogenation of cyclohexene in the gas phase. The rhodium particles had a particle size between 1.71 and 19.6 nm. This corresponds to a percentage metal exposed of ca. 5-50%, below the range reported in this investigation. The magnitude of v, in the gas phase is about the same as that reported here for the liquid phase. This similarity of gas- and liquid-phase turnover rates for cyclohexene hydrogenation has also been noted on Pt6-' and Pd." Variations of v, in the solvents, excluding benzene, follow the same trend as the solubility of H2 in each solvent. Since the turnover rate is a first-order pressure of H2, it is reasonable to assume the same dependence on solubility of H2.Indeed, Table 4 shows that rate constants, calculated by dividing the average turnover rate in each solvent by the solubility of H2, are approximately independent of the solvent. A first-order dependence of the turnover rate on solubility of H2 has also been observed for the hydrogenation of cyclohexene over Pt by Madon et a1.' This result and the zero-order dependence of the rate on the concentration of cyclohexene can be explained by the rate-determining step, as recently defined,62 being the dissociative adsorption of H2 from a precursor state within a saturated overlayer of adsorbed hydrocarbon.' For benzene as a solvent, v, is much lower than expected from its solubility of HZ.This inhibition can be explained by benzene successfully competing with cyclohexene for adsorption on the rhodium surface. This was verified experimentally as v, decreased by a factor of four when the reaction was conducted with 0.15 mol dmd3 benzene and 0.14 mol dm-3 cyclohexene in cyclohexane as the solvent. Benzene inhibition of rhodium-catalysed hydrogenation of cyclohexene in the gas phase has also been observed by Marques Da Cruz,6' Conan et a1.63 and Jarrell et al.64 Benzene was not an inhibitor for the hydrogenation of cyclohexene on Pt' but did inhibit rates on Pd" and M. Boudart and D. J. Sajkowski 65 Table 5 Turnover rate, q, for cyclohexene hydrogenation, with soluble Rh catalysts and Rh/ v-AI2O3, at 298 K hydrogen cyclohexenecatalyst pressure/ kPa conc./mol dm-3 solvent ut/s-' ref.-C4HsO 0.11 63 1 C6H6 0.06 64 4.1 C6H, 0.33 65 --0.22" 66 2.5 C6H6 0.64 67 1.3 CZH80H 3.07 16 metallic rhodium m/??-A1203 101 0.14 C6H12 13.7 this work " Maximum turnover rate obtained from rate expression presented in ref. 68. '2-Aminopyridine also added to solution, T = 303 K. Table 6 Turnover rate (s-') for cyclohexene hydrogenation on Pt, Pd, Ni, and Rh at 298 K and pH2= 101 kPa a catalyst metal exposed (YO) gas phase liquid phase' supported Pt 14-100 2.72-2.94' 0.55-0.66d Pt(223) 2.8' supported Pd 11-76 2.38-3.98f 1.35-1 .72f supported Ni 36-100 1.53-2.44' 0.32-0.66' supported Rh 5-52/69-100' 5.13-7.17' 1.16- 1.36k " The rate was zero order with respect to cyclohexene in all cases.Cyclohexane solvent. 'Ref. 6. Ref. 8. 'Ref. 10. Ref. 11. Ref. 13 Ref. 12. Refers to YO metal exposed, gas phase/liquid phase, respectively. Ref. 61. This work. J A notable difference between the soluble rhodium catalysts and Rh/q-AI2O3 is their behaviour in benzene. As discussed, benzene acts as an inhibitor for supported rhodium. In contrast, turnover rates for Wilkinson's catalyst are higher in benzene than in n-hexane or cyclohexane. l6 A comparison of turnover rates for the hydrogenation of cyclohexene by soluble rhodium catalysts with Rh/ q-A1203 is shown in Table 5.'6-63-68This list is intended to be representative of the range of published turnover rates for the hydrogenation of cyclohexene by soluble rhodium catalyst but is not comprehensive.The turnover rate on Rh/q-A1203 is at least 40 times greater than that on Wilkinson's catalyst RhCl( PPh3)3. Finally, a summary of results obtained for the rate of hydrogenation of cyclohexene is shown in Table 6. Turnover rates are not very different for Ni, Rh, Pd, and Pt, as expected for structure insensitive reaction^.^ A comparison of turnover rates for cyclo- hexene hydrogenation by Group VIII B metals at 298 K and 10.1 kPa dihydrogen in the gas and liquid phase is shown in Table 3. Again it is seen that rates on a given metal are about the same in the gas and liquid phases. The absence of a strong dependence of the turnover rate on the metal used to catalyse the reaction is consistent with the reaction being considered as structure insensitive,' besides showing no effect of particle size on the rate.Catalytic Hydrogenation on Rhodium Clusters Conclusions The changes in the surface atomic structure (ensemble effect) of metallic clusters containing 1-100 atoms has been discussed abundantly together with the change in their electronic structure (self-ligand eff e~t).~~ In particular, the lattice parameter of bare metallic clusters, cu. 1nm in size, is contracted as compared to that of the bulk metal but relaxes to values similar to that of the bulk metal when the clusters are covered with hydrogen or hydrocarbon^.^^-^^ For a structure insensitive reaction such as the hydrogenation of cyclohexene, no effect of particle size (i.e.surface atomic structure) on turnover rate is expected and none was found. The salient new finding in the present work was that the turnover rate for that reaction remains essentially the same for clusters of rhodium containing only a dozen atoms. However, it was found by EXAFS that the lattice parameter of the latter, in the presence of hydrogen, is the same as that of bulk rhodium, within experimental error. Thus the electronic and atomic structures of these covered clusters seem to be closely related to those of larger ones. Correspondingly, no effect of cluster size on turnover rate was found down to the smallest size investigated. A similar result has been for silica supported bimetallic clusters of palladium and gold.Clusters consisting of a core of ca. 13 atoms of gold decorated with ca. 10 palladium atoms were found to exhibit turnover rates for hydrogenation of ethene in the gas phase and of cyclohexene in the liquid phase differing by only a factor of two from those of small palladium clusters. While bare metallic clusters with fewer than 100 atoms may well exhibit different physical and chemical properties dependeing on how many atoms they contain? it seems that the turnover rate for structure insensitive reactions that proceed on essentially saturated surfaces is independent of cluster size down to the smallest size measured today. It appears that the structural differences of the few available remaining vacant sites are ‘erased’ by the almost complete overlayers of adsorbed intermediate^.^^ This work was supported by NSF Grant NSF-CPE 82-19066.For their assistance in various parts of this work, we thank Eric Birnbaum, H. C. Chalal, Klaus Heinemann, G. M. Matis, L-C. de MCnorval, Fred Rumpf and Nicole Somorjai. References 1 0. Beeck, Discuss. Faraday SOC.,1950 8, 118. 2 M. Boudart, J. Am. Chem. Soc., 1950,72, 1040. 3 L. Pauling, Proc. R. SOC.London A, 1949, 196, 343. 4 Y. Soma-Noto and W. M. Sachtler, J. Catal., 1974, 32, 315. 5 M. Boudart, in Proc. 6th Intl. Congr. Catalysis, ed. G. C. Bond, P. B. Wells, and F. C. Tompkins, The Chemical Society, London, 1977, vol. 1, p. 1. 6 E. Segal, R. J. Madon and M. Boudart, J. Catal., 1978, 52, 45. 7 G. Leclerq and M. Boudart, J.Catal., 1981, 71, 127. 8 R. J. Madon, J. P. O’Connell and M. Boudart, AIChEJ., 1978,24,904. 9 D. J. O’Rear, D. G. Loffler and M. Boudart, J. Cataf., 1985,94, 225. 10 S. M. Davis and G. A. Somorjai, J. Cutaf.,1980, 65, 78. 11 E. E. Gonzo and M. Boudart, J. Catal., 1978, 52, 462. 12 W.-C. Cheng and M. Boudart, J. Catal., 1987, 106, 134. 13 C. M. McConica and M. Boudart, J. Card., 1989, 117, 33. 14 J. Horiuti and M. Polanyi, Trans. Faraday SOC.,1934, 30, 1164. 15 N. Koga, C. Daniel, J. Han and K. Morokuma, J. Am. Chem. SOC.,1987 109,3455. 16 J. Halpern, Inorg. Chim. Acta, 1981, 50, 11. 17 G. H. Via, G. Meitzner, F. W. Lytle and J. H. Sinfelt, J. Chem. Phys., 1983, 79, 1527. 18 J. B. A. D. van Zon, D. C. Koningsberger, H. F. J. van’t Blik and D.E. Sayers, J. Chem. Phys., 1985, 82, 5742. 19 C. R. McIlwrick and C. S. G. Phillips, J. Phys. €, 1973, 6, 1208. 20 P. Gallezot, R. Weber, R. A. Dalla Betta, and M. Boudart, 2.Naturforsch, TeiI A, 1979, 34, 40. 21 R. S. Weber, Ph.D. Dissertation, Stanford University, 1982. M. Boudart and D. J. Sajkowski 22 M. G. Samant, Ph.D. Dissertation, Stanford University, 1985. 23 D. J. Sajkowski, Ph.D. Dissertation, Stanford University, 1986. 24 F. V. Hanson, Ph.D. Dissertation, Stanford University, 1975. 25 F. J. Rivera-Latas, Ph.D. Dissertation, Stanford University, 1986. 26 A. C. Yang and C. W.Garland, J. Phys. Chem., 1957, 61, 1504. 27 J. T. Yates, T. M. Duncan, S. D. Worley and R. W. Vaughan, J. Chem. Phys., 1979, 70, 1219. 28 R.R. Cavanagh and J. T. Yates Jr., J. Chem. Phys., 1981, 74,4150. 29 D. J. C. Yates, L. L. Murrell and E. B. Prestridge, J. Catal., 1979, 57, 41. 30 H. C. Yao and W. G. Rothschild, J. Chem. Phys., 1978, 68, 4774. 31 H. Arai and H. Tominaga, J. Card., 1976, 43, 135. 32 C. A. Rice, S. D. Worley, C. W. Curtis, J. A. Guin and A. R. Tarrer, J. Chem. Phys., 1981, 74, 673. 33 S. D. Worley, C. A. Rice, G. A. Mattson, C. W. Curtis, J. A. Guin and A. R. Tarrer, J. Phys. Chem., 1982,86, 2714. 34 S. D. Worley, C. A. Rice, G. A. Mattson, C. W. Curtis, J. A. Guin and A. R. Tarrer, J. Chem. Phys., 1982, 76, 20. 35 H. F. J. van’t Blik, J. B. A. D. van Zon, T. Huizinga, J. C. Vis, D. C. Koningsberger and R. Prins, J. Phys. Chem., 1983,87, 2264. 36 M. Primet, J.Chem. SOC., Faraday Trans. 1, 1978, 74, 2570. 37 C. W. Garland and J. R. Wilt, J. Chem. Phys., 1962, 36, 1094. 38 T. L. Brown and D. J Darensberg, Znorg. Chem., 1967,6, 971. 39 B. E. Koel and G. A. Somorjai, in Catalysis: Science and Technology, ed. J. R. Anderson and M. Boudart, Springer-Verlag, 1985, vol. 7, ch. 3. 40 P. A. Theil, E. D. Williams, J. T. Yates Jr. and W. H. Weinberg, Su$ Sci., 1979, 84, 54. 41 H. F. J. van’t Blik, J. B. A. D. van Zon, D. C. Koningsberger and R. Prins, J. Mol. Card., 1984,25,379. 42 H. P. Wang and J. T. Yates Jr., J. Catal., 1984, 89, 79. 43 J. C. Vis, H. F. J. van’t Blik, T. Huizinga, J. van Grondelle and R. Prinw, J. Catal., 1985, 95, 333. 44 Solubility Data Series, ed. C. L. Young, Pergamon Press, Exeter, vol.5/6, 1981. 45 R. van Hardeveld and F. Hartog, Surf: Sci., 1969, 15, 189. 46 M. Boudart., Adv. Card. Rel. Subj., 1969, 20, 153. 47 D. J. C. Yates and J. H. Sinfelt, J. Catal., 1967, 8, 348. 48 D. J. C. Yates, L. L. Murrell and E. B. Prestridge, in Growth and Properties od Metal Clusters, ed., J. Bourdon, Elsevier, Amsterdam, 1980, p. 137. 49 D. J. C. Yates, L. L. Murrell and E. B. Prestridge, J. Catal., 1979, 57, 41. 50 S. E. Wanke and N. A. Dougharty, J. Card., 1972, 24, 367. 51 H. C. Yao, S. Japar and M. Shelef, J. Catal., 1977, 50, 407. 52 H. F. J. van’t Blik, J. B. A. D. van Zon, D. C. Koningsberger and R. Prins, J. Mol. Card., 1984,25,379. 53 H. F. J. van’t Blik, J. B. A. D. van Zon, T. Huizinga, D. C. Koningsberger and R. Prins, J.Am. Chem. Soc., 1985, 107, 3139. 54 J. J. F. Scholten, A. P. Pijpers and A. M. L. Hustings, Catal. Rev.-Sci. Eng., 1985, 27, 151. 55 H. Poppa, Vacuum, 1984,34, 1018. 56 D. J. C. Yates, L. L. Murrell and E. B. Prestridge, J. Catal., 1983, 57, 41. 57 S. B. Newcomb, W. M. Stobbs and J. A. Little, Phys. Star. Solidi, 1983, 76, 191. 58 R. M. Koros and E. J. Nowak, Chem. Eng. Sci., 1967, 22,470. 59 R. J. Madon and M. Boudart, Znd. Eng. Chem. Fundam., 1982,21,438. 60 J. M. Campelo, A. Garcia, D. Luna and J. M. Marinas, Gazz. Chim. Zral., 1982, 112, 221. 61 G. Marques Da Cruz, Thtse de Docteur-IngCnieur, UniversitC Pierre et Marie Curie, Paris, 1983. 62 M. Boudart and K. Tamaru, Catal. Lett., 1991, in the press. 63 J. Conan, M. Bartholin and A. Guyot, J.Mol. Card., 1975, 1, 375. 64 M. S. Jarrell, B. C. Gates and E. D. Nicholson, J. Am. Chem. SOC., 1978, 100, 5727. 65 P. E. Kreter and D. W. Meek, Inorg. Chem., 1983, 22, 319. 66 W. Reiman, W. Abboud, J. M. Basset, R. Mutin, G. L. Rempel and A. K. Smith, J. Mol. Curd., 1980, 9, 349. 67 M. H. J. M. de Croon, P. F. M. T. van Nisselrooij, J. H. A. M. Kuipers and J. W. E. Coenen, J. Mol. Catal., 1978, 4, 325. 68 J. A. Osborn, F. H. Jardine, J. F. Young and G. Wilkinson, J. Chem. SOC. A, 1966, 12, 1711. 69 A. Kaldor, D. M. Cox, and M. R. Zakin, in Advances in Chemical Physics, ed. I. Prigogine, S. A. Rice, Wiley, New Yotk, 1988, vol. 70, part 2, p. 211. 70 P. Gallezot, A. Bienenstock, and M. Boudart, Now. J. Chem., 1978, 2, 263. 71 P. Gallezot, J. Chim. Phys., 1981, 78, 881. 72 P. Gallezot, Zeolites, 1982, 2, 103. 73 M. Boudart and R. Davis, in Proc. TOCAT Z Symposium on Catalysis, Tokyo, 1990, Catalytic Science and Technology, Kodansha, Tokyo, 1991, vol. 1, pp. 129-134. 74 M. Boudart, J. Mol. Card., 1985, 30, 27. Paper 1/02589I; Received 28rh May, 1991

ISSN:1359-6640    

DOI:10.1039/FD9919200057

出版商:RSC    

年代:1991

数据来源: RSC

摘要:

Faraday Discuss., 1991, 92, 69-77 Preparation of Pt-Ru Bimetallic Particles on Functionalized Carbon Supports by Co-exchange Anne Giroir-Fendler, Dominique Richard and Pierre Gallezot Institut de Recherches sur la Catalyse -CNRS, 2 avenue Albert Einstein, 69626 Villeurbanne Cedex, France Bimetallic Pt-Rucatalysts have been prepared over a full range of composi- tion from pure platinum to pure ruthenium by co-exchange on functionalized carbon supports. This method leads to small particles, homogeneous in size and composition, as shown by electron microscopy. These catalysts have been used in the hydrogenation of cinnamaldehyde. A maximum of selec- tivity towards cinnamyl alcohol as well as activity is observed for a ruthenium content of 30% on active charcoal.On graphite maximum activity is observed for a ruthenium content of 45 %. The study of bimetallic catalysts is of wide interest since the addition of a second metal often improves the activity and selectivity of metallic catalysts. This is the case for the selective hydrogenation of cinnamaldehyde (CAL) to cinnamyl alcohol (COL). The addition of other metals such as iron,'*2 tin: germanium: cobalt' or ruthenium' to platinum, has been shown to increase the selectivity to COL. However, in order to interpret the catalytic results, it is necessary to control the preparation of these catalysts to obtain bimetallic particles homogeneous in size and composition. In fact, in the case of the Pt-Rusystem previously ~tudied,~ some catalysts were heterogeneous and phase segregation was suspected for a composition close to the one with the highest selectivity, thus hampering any attempt to understand the reason for this high selectivity.Different methods of preparing bimetallic catalysts are described in the literature. The most frequenctly used is ~o-impregnation~.~ followed by reduction. The second metal is sometimes added by decomposition of an organometallic compounds-10 on a previously reduced monometallic catalysts. The co-exchange method has been used to prepare bimetallic catalysts on silica" or zeolite,12 but not, to our knowledge, on carbon supports. The cation-exchange method has proved successful for producing metal particles with a high di~persion'~ on graphite or on active charcoal previously functionalized by an oxidizing treatment with sodium hypo~hlorite.l~*'~ Thus, the aim of the present work is to prepare bimetallic Pt-Ru catalysts with particles homogeneous in size and in composition.The activity and selectivity of these catalysts of composition varying between pure Pt and pure Ru will then be studied in the hydrogenation of cinnamal- dehyde as a function of composition. Experimental Materials The carbon materials used were a graphite (HSAG 12 from Lonza, 470 m2 g-') and an active charcoal (50/S from CECA, 1450 m2 g-'). The metal salts [Pt(NH3)4]C12 and [Ru( NH3)6]C13 were purchased from Johnson-Matthey. Cinnamaldehyde (from Merck) was used as received and stored under nitrogen to avoid any oxidation.Hydrogen (99.995 YOpurity) was supplied by 1'Air Liquide. 69 Reparation of Pt-Ru Bimetallic Particles Cationic Exchange Both carbon supports were functionalized according to a previously described method', before the ion-exchange of the metal. The platinum salt, [Pt(NH3)JC12 is dissolved in water and passed through an anion-exchange resin (Amberlite IRN-78), in the hydroxyl form, in order to substitute OH- for Cl- anions since the hydroxy anion provides a more efficient exchange." The resin is then rinsed three times with water to complete the collection of the ammino complex. The support is suspended in ammonia (1mol dm-,) solution, and stiried under nitrogen. The Pt(NH,);+ solution is added dropwise and the stirring is maintained for 15 h thereafter. The support is then filtered (Millipore MF type filters) and washed thoroughly with water and dried at 373 K under flowing nitrogen.In the case of ruthenium, the exchange proceeded similarly, starting from [Ru( NH3)6]C13. For the co-exchange, Pt(NH3)qf was added dropwise to the suspension of the carbon support in ammonia (1 mol drn-,), followed immediately by Ru( NH,);', also dropwise. The suspension was stirred for 15 h. The dried catalysts were reduced in flowing hydrogen for 2 h at 473 K. Some of the samples were reduced in a calorimeter. Transmission Electron Microscopy TEM examinations were performed using a JEOL lOOCX microscope. For direct observation, powders were ultrasonically dispersed in alcohol.A drop of the catalyst suspension was then deposited on an amorphous carbon film supported on a copper grid, and dried under an IR lamp. The charcoal grains were also embedded in an epoxy resin and cut in thin sections with an ultramicrotome to check the metal distribution inside the charcoal grain. Analytical Electron Microscopy The composition of the metal particles was determined by energy dispersive X-ray emission (EDX) with a SiLi diode attached to a field-emission gun, scanning transmission electron microscope (FEG-STEM) VG-HB501. The spatial resolution of analysis is as high as 1.5 nm so that the composition of individual particles can be measured. Quantita- tive analysis used the Pt-La, and the Ru-Ka emission line. Although they are not the most intense, they have been chosen since they do not interfere with any other line which may be present in the spectrum.Metal Analysis Metals on carbon supports such as active charcoal and graphite are usually analysed by atomic absorption after mineralization of the support. To achieve mineralization, the catalyst is treated with a mixture of fuming acids and heated. This procedure is not satisfactory for Ru since volatile oxides tend to form and escape. Therefore X-ray fluorescence was used for the analysis of both metals in the catalysts. However, atomic absorption was used for the determination of the metal in the filtrate after the exchange. X-ray fluorescence requires a calibration curve established from standards very similar to the samples to be analysed.The standards were prepared by impregnation of both supports with the required amount of H,PtCl, and RuCl, in a benzene-ethanol mixture in a rotatory evaporator. Calorimetry A differential calorimeter of the Tian-Calvet type (Therm Analyse) was used to monitor the heat flux exchanged during the reduction of some catalyst precursors on graphite. A. Giroir-Fendler, D. Richard and P. Gallezot Table 1 Composition of active-charcoal-supported catalysts Ru/ Pt in solution (atom) metal loading (wt. %) Ru Pt catalyst cont. (atom %) Ru Pt Ru/Pt in catalyst (atom) 0.00 0.0 4.0 0 100 0.00 0.13 0.3 4.5 9 91 0.10 0.35 0.6 3.4 25 75 0.33 0.45 0.75 3.25 32 68 0.47 0.83 1.2 2.8 46 54 0.85 1.16 1.5 2.5 55 45 1.22 2.36 2.2 1.8 70 30 2.33 7.40 2.3 0.6 88 12 7.33 4.5 100 0 Table 2 Composition of graphite-supported catalysts exchange solution (wt.Yo)a Ru/ Pt catalyst Ru/ Pt in solution (wt. Yo) in catalyst Ru Pt (atom) Ru Pt (atom) 0.75 6.5 0.23 0.58 2.25 0.5 1.5 6.5 0.47 1.52 2.0 1.5 1.o 2.0 1.o 0.77 0.83 1.86 2.0 3.0 1.2 1.6 1.1 3.O 5.O 6.5 1.5 3.85 0.85 9.0 2.0 2.0 2.0 2.25 0.59 7.3 3.0 2.0 3.0 2.8 0.45 11.5 4.0 2.0 4.0 3.17 0.41 15.7 5.O 2.0 5.0 4.68 0.43 19 a Relative to supporting material in suspension. Catalytic Tests The hydrogenation of cinnamaldehyde (0.1 mol) was run at 333 K under 4 MPa of hydrogen in a batch reactor. The catalysts were heated at 373 K, for 2 h in the reaction mixture before the introduction of the substrate.The detailed procedure has already been described e1~ewhere.l~ Results and Discussion Exchange and Reduction Processes Table 1 gives the compositior, of the starting solutions and of the active-charcoal supported catalysts. After ion-exchange the filtrates were checked for both metals and only traces were found indicating that all the metal has been retained on the support. Functionalized active charcoal can fix up to 11 wt. % of platinum by cationic exchange. Thus, the exchange is almost complete and there is no preferential exchange of one metal since the Ru/R ratios are similar in solution and in the catalysts (Table 1). However, with a graphite support, the amount of metal found by X-ray fluorescence on the solids is generally smaller than expected if the exchange was complete, particularly for Pt (Table 2).Atomic absorption determination of the metals in the filtrates after Preparation of Pt- Ru Bimetallic Particles 15 8 hb:'3 10 d v 5 0 Fig. 1 Atomic ratio (Ru/Pt) in the catalyst us. the atomic ratio (Ru/Pt) in solution. 0,On active charcoal; +, on graphite, Pt 2 wt.%; 0, on graphite, Pt 6.5 wt.% the exchange, gives the balance. This is consistent with previous observations showing that functionalized graphite can only fix up to 3-4 wt. % of Pt by cationic exchange.13 Moreover, Fig. 1, where the atomic ratio Ru/Pt found for the catalyst is plotted vs. the atomic ratio in the solution, clearly shows that Ru is more easily exchanged than Pt since (Ru/ Pt),,, is always larger than (Ru/ Pt),ol.Fig. 2 shows the thermograms recorded during the reduction of ion-exchanged R, Ru and Pt-Ru graphites. Pt is reduced at a lower temperature than Ru, a single peak is observed at 435 K. For Ru two peaks are observed, one at 475 K and the other at 568 K; the first can be attributed to the reduction of Ru3+ to Ru2+ and the second to the reduction of Ru2+ to Ru. In the case of Pt-Ru samples, there is a large peak at 373 K, indicating that both Pt and Ru ions are reduced at a lower temperature. The easiest reduction of Ru in presence of Pt is not unexpected but enhanced reduction of Pt is noteworthy. A similar co-enhanced reduction of Pd and Ni co-exchanged in NaY zeolite has been observed and was attributed to the presence of PdNi ion pairs.16 Although this interpretation needs to be substantiated it can be applied in the case of our co-exchanged Pt-Ru catalyst where the two elements can be confined in the form of ion pairs stabilized by the functional groups of graphite; Characterization of Catalysts The sizes of the Pt-Ru particles on active charcoal are very homogeneous over the full range of composition studied.Since the active charcoal is in the form of thick grains, direct TEM 'observation provides information only for the particles which are in the outer part of the grains. To check the particle size and the distribution of the metal within the grain it was necessary to observe very thin slices of the grains by cutting the active charcoal with an ultramicrotome.Thus in the catalyst shown in Fig. 3(a) (Ru-Pt, 25-75 atom %) no difference was observed between the outer and inner parts of the grains, indicating that the particles are uniformly distributed in the grain. Particles have an average size of 2 nm (individual sizes in the range 1.5-2.5 nm). The particles prepared A. Giroir-Fendler, D. Richard and P. Gallezot 100 200 300 100 200 300 100 200 300 T/ "C Fig. 2 Thermograms recorded during the reduction of the catalysts: (a)Pt/graphite, (b)Pt-Ru (66-34 atom %)/graphite, (c) Pt-Ru (25-75 atom %)/graphite, (d)Ru/graphite Fig. 3 TEM micrograph Pt-Ru catalyst (a)on active charcoal, (b)on graphite with a higher Ru content are smaller than 2nm, which is consistent with the fact that Ru alone usually gives smaller particles than Pt.14 TEM observation of the different bimetallic catalysts on graphite shows a homogeneous dispersion with particles size in the range 1-2 nm.Fig. 3( b) is a micrograph of a Pt-Ru catalyst on graphite (8-92 atm %). As in the case of monometallic particles on the particles are mainly located at the edges of the graphite sheets or they decorate the steps of graphite adlayers. 74 Preparation of Pt-Ru Bimetallic Particles Table 3 Global and local composition of the catalysts area analysed by STEM-EDX X-ray fluorescence (%) single particles collection of particles Ru Pt Ru (YO) average Ru (%) RU (Yo) average Ru (YO) on active charcoal 49 51 43-68 49.5 50- 52 51.5 25 75 19-45 32 15-40 25 on graphite 9 91 --6-15 10 24 76 14-30 21 15-22 18 35 65 13-50 -30-50 75 25 60-84 73 67-75 70 Table 3 gives the overal! composition measured by X-ray fluorescence and the local composition measured by STEM-EDX analysis. Whatever the size of the analysis window (collection of particles or individual particles) the composition of the active- charcoal supported catalyst is homogeneous and consistent with the composition given by X-ray fluorescence. A trend towards enrichment in Pt is observed when the particle size increases.Analysis carried out on thin sections of grains also shows that the composition of the particles does not depend upon their position within the grain.EDX analysis of the graphite-supported catalysts agrees very well with the global analysis of the sample and shows that the composition of individual particles are close to the average. It is thus possible to conclude that all particles are bimetallic and homogeneous in composition. Catalytic Properties The hydrogenation of CAL in the presence of Pt and Ru on the two carbon supports has been reported previou~ly.'~ On active-charcoal support Ru gives mainly the saturated aldehyde whereas platinum gives a mixture of hydrocinnamic aldehyde and COL. On the graphite support Pt is highly selective to COL. The higher selectivities on graphite were interpreted in terms of electron transfer from the graphite support to the metal particles.Fig. 4 shows the plot of the initial selectivity towards COL as a function of the amount of Ru in the catalyst. In the case of active charcoal the initial selectivity exhibits two maxima at ca. 35 and 75% of Ru. The first maximum corresponds to an initial selectivity close to 60 YO, although neither Ru nor Pt gives COL, at least initially. In the case of graphite the dependence of the initial selectivity upon the Ru content is less complicated; the Pt alone is also quite selective, as shown previously, and the selectivity decreases with Ru content especially above 60 YORu. In Fig. 5 the activity of the catalysts is plotted DS. the Ru content for both supports. On active charcoal, a broad maximum is observed at 30atom% of Ru whereas, on graphite, the activity shows a maximum at ca.45 atom YOof Ru. The catalytic properties of the Pt-Ru catalysts are very dependent on the composition of the catalyst and the nature of the carbon support. In the case of active-charcoal supported catalyst, it has been shown, by EDX analysis that Pt and Ru were always associated in particles homogeneous in size and composition. The increase in selectivity and activity which is observed when, starting from pure Pt the Ru content increases, A. Giroir-Fendler,D. Richard and P. Gallezot a0 0220 40 60 80 0 20 40 60 80 Ru content (atom %) Ru content (atom YO) Fig. 4 Initial selectivity towards COL Sco,vs. the Ru content of the catalyst (a)on active charcoal, (b)on graphite 0.4 7 0.3 v) h Y .cI3 O** 0.1 I I I I I o 20 40 60 a0 0 20 40 60 80 Ru content (atom %) Ru content (atom YO) Fig.5 Activity of the bimetallic catalyst in the hydrogenation of cinnamaldehyde (a)on active charcoal, (b)on graphite are well accounted for by a mechanism of activation of the carbonyl bond where both Pt and Ru atoms at the surface of the particle are involved (Fig. 6). Ru atoms, which are more electropositive than Pt atoms, act as an adsorption site for the oxygen atom of the carbonyl. The latter thus becomes more reactive toward hydrogen dissociated on the Pt atoms. This interpretation accounts well for the coincidence of the activity and selectivity maxima. These maxima correspond to an optimum concentration of Ru on the surface.However, the second maximum at 68 % Ru does not correspond to a maximum in activity. In this case, the increase in selectivity may come from a smaller rate for the double-bond hydrogenation. The increase in electron density on Pt, arising from an electron transfer from the Ru, decreases the probability of activation of the ethylenic bond, this activation requires, as a first step, an electron transfer to the d band of the metal. Note that the second maximum is observed for a composition where a change in the structure of the Pt-Ru alloy occurs. Indeed, the phase diagram of the Pt-Ru binary system shows that for (65 % Ru the structure of the alloy is f.c.c. and that for higher Ru content the structure is hexagonal.Preparation of Pt-Ru Bimetallic Particles m Fig. 6 Scheme of the carbonyl activation In the case of the graphite support, the selectivity of pure Pt is already quite high. This can be accounted for by a support effect previously evidenced: l7 an electron transfer from the graphite to the Pt particles decreasing the probability of the C=C bond hydrogenation. When the Ru content is increased, the selectivity is maintained and the activity increases, this is because the mechanism of activation of the C=O bond discussed above is operating. For >60 % Ru, the selectivity and the activity decrease, because the concentration of Ru on the surface becomes too large. Conclusion The following conclusions can be drawn. 1.The co-exchange method leads to very small bimetallic particles homogeneous in size and composition. In the case of active charcoal the particles are uniformly distributed in the charcoal grain. In the case of the graphite support, these particles are preferentially located along the edges of the graphite sheets and on the steps of graphite adlayers. 2. The activity and selectivity of these particles in the hydrogenation of CAL depend upon the composition and are at their maximum for an intermediate composition where a concerted mechanism increases the C=O hydrogenation rate. We thank B. Blanc for help in preparing the X-ray fluorescence standards, F. Mdis from the Service Central d’Analyse (C.N.R.S.) for analysing them and A. Chambosse and H. Urbain for the atomic absorption analysis.References 1 D. Richard, P. Gallezot, J. Ockelford and A. Giroir-Fendler, Cutul. Lett., 1989, 3, 53. 2 D. Goupil, P. Fouilloux and R. Maurel, React. Kine?, Catal. Lett., 1987. 35, 185. 3 Z. Poltarzewski, S. Gaivagno, R. Pietropaolo and P. Staiti, J. Catal., 1986, 102, 190. 4 S. Galvagno, Z. Poltarzewski, G. Nieri and R. Pietropaolo, J. Chem. Soc., Chem. Commun., 1986, 1729. 5 P. Fouilloux, in Heterogeneous Catalysis and Fine Chemicals, ed. M. Guisnet et al., Elsevier, Amsterdam, 1988, pp. 123-129. 6 C. H. Bartholomew and M. Boudart, J. Catal., 1972, 25, 173. 7 G. Blanchard, H. Charcosset, M. T. Chenebaux and M. Primet, in Preparation of Catalysts, ed. B. Delmon et ul., Elsevier, Amsterdam, 1979, Vol. 11, p.197. A. Giroir-Fendler, D. Richard and P. Gallezot 8 E. Kern-Talas, M. Hegedus, S. Gobolos, P. Szedlacsek and J. Margitfalvi, in Preparation of Catafysts, ed. B. Delmon et af.,Elsevier, Amsterdam, 1987, vol. IV, pp. 689-698. 9 0.A. Ferretti, L. C. Bettega de Pauli, J. P. Candy, G. Mabillon and J. P. Bournonville, in Preparation of Catalysts, ed. B. Delmon et af., Elsevier, Amsterdam, 1987, vol. IV, pp. 713-723. 10 H. Miura, H. Taguchi, K. Sugiyama, T. Matsuda and R. D. Gonzalez, J. Catal., 1990, 124, 194. 11 S. Alerasool and R. D. Gonzalez, J. Catal., 1990, 124, 204. 12 T. Tatsumi, Y. G. Shult, T. Sigiura and H.Tominaga, Appf. Catal., 1986, 21, 119. 13 D. Richard and P. Gallezot, in Preparation of Catalysts, ed. B. Delmon et al., Elsevier, Amsterdam, 1987, vol. IV, pp. 71-81. 14 A. Giroir-Fendler, D. Richard, P. Gallezot, in Heterogeneous Catalysis and Fine Chemicals, ed. M. Guisnet et af.,Elsevier, Amsterdam, 1988, pp. 171-178. 15 D. R. Lowde, J. 0. Williams, P. A. Attwood, R. J. Bird, B. D. McNichol and R. T. Short, J. Chem. Soc., Faraday Trans. I, 1979, 75, 2312. 16 J. Schaefer Feeley and W. M. Sachtler, Zeolites, 1990, 10, 738. 17 D. Richard, P. Fouilloux and P. Gallezot, in Proc. 9th Int. Congr. Catalysis, ed. M. J. Phillips and M. Ternan, The Chemical Institute of Canada, Ottawa, 1988, vol. 3, pp. 1074-1081. Paper 1/02599F; Received 24th May, 1991

ISSN:1359-6640    

DOI:10.1039/FD9919200069

出版商:RSC    

年代:1991

数据来源: RSC

摘要:

Faraday Discuss., 1991, 92, 79-107 GENERAL DISCUSSION Dr. A. Masson (ENSCP, Paris, France) opened the discussion: In your general talk, you have raised the question, why, until now, it has not been possible to obtain the same resolution using electron microscopy with very small clusters as is obtained with the bigger ones? This question is directly related to the problem of the temperature and stability during synthesis and observation of these tiny clusters. I think STM is a more appropriate method. At present it is possible to obtain atomic resolution for two- dimensional clusters at least for metal clusters on graphite and semiconductor surfaces. Prof. A. Howie (University of Cambridge) responded: I agree that the scanning tunnelling microscope has great pQtential for the study of the atomic and electronic structure of very small clusters on conducting supports.In the case of insulating supports the atomic force microscope should be useful. Both these instruments are simpler and more readily compatible with cluster beam equipment than is the STEM. The problem of specimen damage is also less serious. On the other hand, the imaging and spectroscopy modes of the STEM are relatively well understood, it can provide direct analytical information, operates over a wider range of magnifications and can deal with clusters buried below the surface. I would also like to draw attention to the rapidly developing capability of the STM for single atom transfer to surfaces or for repositioning of them there.' Most of this work seems to aim at information storage and the construction of atomic-scale switches or electronic devices.However it may also be possible eventually to assemble on a well defined surface site, a cluster with a precisely controllable number of atoms and then to measure some of its properties as a function of its size. 1 D. M. Eigler, C. P. Lutz and W. E. Rudge, Nature, (London) 1991, 352, 600. Prof. Sir John Meurig Thomas (The Royal Institution, London) commented: You raised a point in your lecture about why no one had used electron diffraction in the gas phase to determine the structure of well defined small (and isolated) metallic particles. One of the reasons why this has proved difficult in the past is that it has been very difficult to form highly uniform (monodisperse) particles.Of late, however, some very promising developments have occurred whereby it may prove possible to carry out gas-phase electron diffraction on the femtosecond time-scale. Zewail and co-workers,*'* in their sophisticated adaptation of flash photolysis which they have been carrying out on reacting beams, involving four or five atoms in radical or molecular form, have demonstrated that it is, in principle, possible to conduct electron diffraction measure- ments on a cluster of four or more atoms (metallic or otherwise) and to do so repeatedly at very short time intervals, so that structural changes may be monitored while such clusters undergo spontaneous changes. 1 A. Williamson and A.H. Zewail, Proc. Natl. Acad. Sci. USA, 1991,88, 5021 2 J. M. Thomas, Nature (London), 1991, 351, 694. Dr. P. A. Buffat (EPFL, Lausanne, Switzerland) said: In his lecture, Prof. Howie has emphasized the contribution that electron diffraction on unsupported clusters can make to our knowledge of the crystallography of such small particles. He was calling for an experiment on metal clusters such as those done a few years ago on noble gases. From the discussion following his talk, it appears that high resolution electron micro- scopy is unable to describe the structure fully, in particular because only a few spatial 79 General Discussion frequencies are transmitted through the limited bandwidth of the microscope and because inelastic events introduce continuous structural changes under the electron beam.It is worth adding at this point that it is very difficult to produce sources of metal clusters intense enough for electron diffraction. A first attempt has been made by Yokozeki and Stein' on silver clusters and a departure from the f.c.c. structure was reported, however the new structure could not be identified. A more detailed study has just been completed by Hall et al.; again on unsupported silver clusters produced by a helium-gas-aggregation technique (cluster size ca. 1-6 nm). The radial distributions of scattered intensities were calculated by the Deb ye relation for several cluster structures and diameters and they were fitted to the experimental intensities in order to determine which structures were present in the beam and in what proportions.For some operating conditions of the source only f.c.c. were present, for others a mixture of f.c.c., decahedra and icosahedra was found. It should be pointed out that with Hall's experimental setup, a cluster receives a dose of less than one electron during its passage under the electron beam. 1 A. Yokozeki and G. D. Stein, J. Appl. Phys., 1978,49, 2224. 2 B. D. Hall, M. Flueli, R. Monot and J. P. Borel, Phys. Reu. B, 1991, 43, 3906; B. D. Hall, Thesis no. 954, EPFL, Lausanne, Switzerland, 1991. Prof. Howie responded to the two comments: Pulsed electron diffraction methods certainly offer realistic opportunities for structural and other studies of metallic clusters in molecular beams.As noted by Dr Buffat (see also the discussion following his own paper), some structural information about small Ag clusters has already been obtained using 100 keV electrons in 15 ms pulses of ca. 5 x lo9 electrons focussed in a 300 pm spot.' Each pulse illuminates perhaps lo5spinning clusters and the superimposed results from many such pulses are interpreted on a statistical basis. With an improved cluster source and tighter mass-selection of clusters, this method could be further developed. To collect diffraction information from individual clusters as suggested by Prof. Thomas, the electron beam would have to be focussed to a spot of 1 pm or below and, to freeze their rotational motion, the pulse duration should be less than ca.lo-'* s. Electron diffraction patterns from extended surfaces have been successfully recorded with lO-'Os pulses of lo5 electrons in a 0.8 mm spot2 and substantial improvements, perhaps extending to the femtosecond region, are possible. It may not be feasible (or desirable in view of the electronic excitations which could be generated even on the femtosecond scale) to arrange for any metallic particle to interact with more than a few electrons during a single pulse. A complete diffraction pattern would therefore not be available but a CCD array could collect not only the angular distribution averaged over many pulses, but also angular correlations in the diffraction data.3 1 B. D. Hall, M. Fiueli, D. Reinhard, J.-P. Borel and R. Monot, Reu.Sci. Instrum. 1991, 62, 148. 2 H. E. Elsayed-Ali and G. A. Mourou, Appl. Phys. Lett. 1988, 52, 103. 3 J. M. Rodenburg and 1. A. Rauf in EMAG-MICRO 89 ed. P. J. Goodhew and H. Y. Elder Inst. of Physics, Bristol, 1990, pp. 119-122. Prof. R. W. Joyner (Uniuersity of Liuerpool) opened the discussion of the paper by Zamaraev: There is much fascinating chemistry in this paper and the further data on the giant Pd 561 cluster are intriguing. Before the interpretation of the EXAFS data can be accepted certain features should be clarified, particularly since most calculations predict an icosahedral to f.c.c. transition for cluster sizes >ca. 100 atoms: ' (a) The change in EXAFS data on heating, shown in Fig. 8, is attributed to an icosahedral to f.c.c.transition. It seems moie likely that the data show breakdown into smaller clusters, since the magnitude of the non-nearest-neighbour peaks in the Fourier transform decrease markedly on warming. (6) It is not clear why the unrelaxed cluster does not show peaks corresponding to d3a and d2q (where a is the nearest-neighbour distance). As can be seen by General Discussion inspection of Fig. 9 these distances should be observed for large icosahedral clusters, as well as in f.c.c. structured materials. They are quite readily observable in much smaller platinum clusters (d ca. 1 nm).* (c) It is really necessary to have quantitative values of Pd-Pd and Pd-ligand coordination numbers, so that the structural proposals can be fully assessed.1 See e.g. M. B. Gordon, F. Cyrot-Lackman and M. C. Desjonqueres, Surface Sci., 1979 68,359; 1979, 80, 159. 2 See e.g. P. Johnston et al, Faraday Discuss. Chem. SOC.,1990 89, 91. Prof. K. I. Zamaraev (Institute of Catalysis. Novosibirsk, USSR) responded: Before answering your specific questions I should like to make some more general comments. (1) Though the effect of EXAFS has been known for a long time, systematic applications of EXAFS spectroscopy to structural studies have started only recently. In this sense EXAFS is a rather young technique, though of course a very powerful one. However from previous experience with various new instrumentation techniques it is known, that when using them one can often face various a priori unexpected pitfalls, and usually it takes time to rationalise these pitfalls and find how to avoid them.I think that in this respect EXAFS spectroscopy is no exception. That is why in our characterization of giant Pd clusters with EXAFS we thought it appropriate to rely more on those characteris- tics of EXAFS spectra which looked to us clearer and avoid making conclusions from those characteristics which looked to us less clear. (2) We collected data on the ratios of those Pd-Pd interatomic distances, that could be found from the RDA peaks that were of sufficient magnitude with respect to the noise. In our case these were typically the distances up to ca. 4.5-4.7 A, but no more than that. The exact distance up to which the analysis could be done varied somewhat from one sample to another.Thus, we did not measure the distances exceeding 4.5-4.7 A from the EXAFS spectra of our giant Pd clusters because the background noise was too great (see Fig. 8 in our paper). (3) I also think that further studies are needed before it will be possible to interpret reliably the intensities of the EXAFS spectra of small metal cores and the magnitudes of their Fourier transforms. In Fig. 1 the magnitudes of Fourier-transform EXAFS spectra are compared for small Pd (25A) and Fe (1214) particles on one hand, and bulk Pd and Fe on the other hand. The data for Pd are taken from our studies, while the data for Fe are taken from Zhao and Montano.' For both metals the peaks for small particles are seen to be dramatically (much more than one could expect from the anticipated values of the coordination numbers in small particles) smaller than those for bulk metals.Zhao and Montano have tentatively attributed this effect to the inelastic scattering of electron waves on the boundary of small particles. However, more detailed studies seem to be necessary to describe this effect quantitatively. The data of Fig. 1 clearly demonstrate one of the pitfalls one can face when calculating coordination numbers for atoms in small metal particles directly from the magnitudes of the EXAFS spectra. Until the factors that determine the magnitudes of the peaks are better under- stood, we hestitate to make any far-reaching conclusions about the structure of our giant clusters on the basis of these magnitudes.(4) Note, that both the formula Pd561phen60(OAC)180 of our giant clusters and its structure presented in Fig. 9 in our paper are idealised rather than exact ones. As indicated in the paper, they correspond to some average size and composition of the cluster, rather than to certain fixed ones. We have some evidence that the interatomic distances found for this cluster (Table 2 in our paper) also should be considered as average values for two or more distances that are close to each other, but are not the same. This evidence is provided by the data of Fig. 2. In Fig. 2(a) the log-ratio of EXAFS functions of the first Pd-Pd shell for Pd561phen60(OAc),80 cluster in the relaxed f.c.c. form and for Pd foil are presented as a function of k2. If only one fixed value General Discussion 0.1 1 .o" c .I )"I\9 i i,i\ 25 A particlesE '" \\I"-+-...,-..,,.-..--.-,0 1.5 3.0 4.5 6.0 (R-&)/A 0 2 4 6 8 10 radius/W Fig. 1 Intensities of the RDA peaks for bulk and small particles of (a)Pd and (b) Fe. The RDA curve for 25 8, Pd particles refer to a Pd561phen60(OAc)180 cluster in the f.c.c. relaxed form. The RDA curves for bulk Fe and 12 %, Fe particles are taken from ref. 1 existed for the first Rpd-pd distance, then the log-ratio function in Fig. 2(a) would be a straight line. But it is seen to notably deviate from a straight line. This means that in fact more than one interatomic distance is 'hidden' within the first most intense peak from the nearest-neighbour Pd atom in the RDA curve of Fig.8(d) in our paper. In Fig. 2(6) simulations of the log-ratio function are given. They were made for the model of two equally probable different values for the first Rpd-pd distance. Various curves correspond to various values of the difference AR between the two values of Rpd-pd. The experimental curve is seen to agree well with the simulation curve for AR = 0.04 A. We do not think that this agreement between the experimental and simulation curves should be used to insist that the precise structure of the first peak is composed of two equally intense peaks separated by a distance of 0.04 A. Perhaps, other models for the deconvolution of the first peak can give still better agreement with experiment.The only conclusion we make at the moment from the data of Fig. 2 is that the structures of the giant cluster as an ideal icosahedron (in the unrelaxed form) or an ideal cuboctahedron (in the relaxed form) are indeed idealised approximations. The devi- ations from such a simplified picture are revealed even in the peak from the first coordination shell of Pd atom in the relaxed cluster. Still bigger deviations can be General Discussion I I I I I I0.0 30 60 90 120 150 180 ’ 0.0 I30 60 90 I 120 150 180 k2/A-2 Fig. 2 The log-ratio of EXAFS functions of the first Pd-Pd shell for Pd56,phen60(OA~)180 in the f.c.c. relaxed form. x(k) is the EXAFS function for the cluster and x,(k) is the EXAFS function for the foil: (a)experimental function and (b) simulated function.Simulations were made for the model of two equally probable different values for the first Rpd-pd distance. The difference AR between the two values of RPd-Pdwas varied in order to achieve the best fit between the simulated and experimental functions. In calculations the Debye- Waller factor was chosen to be equal to g2=0.007 A’ expected for more distant shells. At the moment we do not have a good criterion for selection of a reliable model that would describe these deviations. Under these circum- stances we hesitate to make any far-reaching conclusions about the structure of our clusters from the magnitudes of the peaks from the second and more distant coordination shells.(5) As suggested by EXAFS data, our giant Pd561 clusters can exist in two extreme forms: an icosahedral and an f.c.c. one. It seems that the difference in free energy between the two states is not large. The studies to specify conditions favouring this or that structure are still in progress. However, as a preliminary conclusion, it seems possible to say that for the cluster with the idealised icosahedral Pd561phen60(OAc)180 initial structure the transition to the f.c.c. structure is favoured by warming and, perhaps, also by the presence of oxygen. Note also that there exist several different versions of an icosahedral structure. The structure we propose for our Pd561phen60(OA~)180 cluster in the unrelaxed form is close to the face-centred variant of the model proposed by Watson and Weinert,* for massive icosahedral metal alloys, and actually observed for these alloys by many authors (see, General Discussion e.g.the references in ref. 2). Thus the icosahedral structure is known not only for very small metal particles, as suggested, but also for massive crystalline metal alloys. After these more general comments I can answer your specific questions: (a) You are right that the magnitudes of the non-nearest-neighbour peaks decrease markedly upon warming. But I do not think that this phenomenon should be attributed to the decomposition of our cluster into smaller pieces. From Fig. 1(a) it is seen that the distribution of the relative magnitudes of various peaks in the RDA curve of the cluster after it has been warmed is close to that in the RDA curve for Pd foil.The latter certainly contains the unbroken metal skeleton. (b) According to the version of the icosahedral structure proposed by Watson and Weinert: the peak which you place at J3a should be located at ca. 1.6~.This is the peak in the RDA curve for the unrelaxed structure of our cluster, that corresponds to Pd-Pd distance R =4.08*0.1 A [see Fig. 8(c) and Table 2 in our paper]. The peak at J2a = 3.68 A is present in our RDA curve (the distance 3.66 A in Table 2). We do not interpret still more distant peaks in this Fig. because their magnitudes are comparable with that of the noise. (c) Of course, the structural features of the giant clusters would be clearer if quantitative numbers of Pd-Pd and Pd-ligand coordination numbers were found.But coordination numbers are calculated from the magnitudes of the peaks of the RDA curves. I have already explained in the first part of my response, why I think that, at the moment, these calculations are not reliable enough for our giant clusters. 1 J. Zhao and P.A. Montano, Phys. Rev. B, 1989,40, 3401. 2 R. E. Watson and M. Weinert, Muter. Sci. Eng., 1986, 79, 105. Dr. P. Gallezot (CNRS, Villeurbanne, France) asked: Did you check the structure of the giant clusters after catalytic reactions? I would expect that, for instance, oxidation reactions would greatly modify their structure. How do the catalytic properties of the giant clusters compare with those of naked palladium particles of comparable size? Prof.Zamaraev responded: We did not check the structure of the giant clusters after catalytic reactions. But I agree, that during oxidation reactions it can be greatly modified. For example, if one precipitates the cluster with phen and OAc- ligands after a catalytic oxidative acetoxylation of an alkene is complete, and then dissolves the cluster again and starts the reaction again, some of the initial catalytic activity is lost. This indeed suggests that at least some of the cluster,molecules are decomposed or modified during the reaction. Catalytic properties of the giant cluster in oxidative acetoxylation of alkenes are similar to those of supported catalysts with finely dispersed palladium. But I am not sure whether a supported metal particle can be treated as a naked one.Perhaps, it is more appropriate to consider it also as a cluster, which is stabilised by the functional groups of the support as ligands. Prof. C. N. R. Rao (Indian Institute of Science, Bangalore, India) asked: (1) What are the coordination numbers you find from EXAFS? I thought that you should find more structure-specific features in EXAFS characteristic of the icosahedral structure? (2) Did you not find other clusters (smaller than Pd561)? Prof. Zamaraev replied: (1) At the moment we hesitate to interpret the magnitudes of our EXAFS in terms of coordination numbers because of the complications outlined in my response to the question of Prof. Joyner. The main argument in favour of the existence of the cluster isomer with an icosahedral structure is the characteristic pattern of four Pd-Pd distances as indicated in Table 2 of our paper.General Discussion (2) As indicated in our paper, besides giant clusters with phen or bipy+anionic ligands we have isolated and characterised also a family of much smaller clusters of the Pd, family. But we have not eludicated in our systems any other clusters that would cover the gap between Pd, and Pd,,’, except, of course, the fact that according to TEM and SAXS data the clusters with the idealised Pd5,’ nuclearity have a certain size (nuclearity) distribution. Dr. R. E. Benfield (University of Kent) said: Another large palladium cluster has been reported by Prof.G. Schmid (Essen).’ This is of idealised formulation Pd561phen360200and so differs somewhat from your compounds. X-ray diffraction has suggested an f.c.c. structure.’ We have collected palladium EXAFS data on this material [2] and preliminary analysis shows a cuboctahedral cluster structure, not the icosahedral one you have found for Pd561phen60(OA~),80. The obvious difference between the two compounds is in the nature of the ligands; the Essen compound lacks acetate groups. Can you comment on the possible influence of the ligand sphere on the metal cluster structure? 1 G. Schmid, Polyhedron, 1988, 7, 2321. 2 R. E. Benfield, R. J. Newport, S. J. Gurman, G. Schmid, work in progress. Prof. Zamaraev replied: We certainly know the works of Prof.Schmid in the area of cluster chemistry. As you have mentioned, the obvious difference between his and our Pd,,, cluster is the nature and number of ligands. Especially, the different number of the key phen ligands in the two clusters should be noted. In ref. 16 of our paper we have shown that according to EXAFS, substitution of OAc- ligands for PF,+02- ligands in our Pd5,,phen6,(OAc) cluster indeed modifies greatly the structure of its metal core. In ref. 16 we have tentatively interpreted the EXAFS data for our Pd561phen60(PF6)60060 cluster in terms of a mixture of icosahedral and cuboctahedral structures. Though further studies seem to be needed to clarify the precise structure of Pd561phen60(PF6)60~50cluster, the conclusion that it is different from that of the icosa- hedral Pd561phen60( isomer seems to be reliable enough.I should like to mention that in contrast to the EXAFS data, the X-ray diffraction data obtained in our TEM experiments with Pd561phen60( cluster have suggested OAC)~~~ an f.c.c. structure for this cluster. We explain this by a possible reconstruction of the cluster skeleton under the electron beam, which may be similar to its reconstruction upon heating. Prof. A. Henglein (Hahn-Meitner-Institute, Berlin, Germany) asked: What are the optical properties of your complexes when they are in solution? Do they show the absorption of colloidal palladium? Prof Zamaraev replied: Our giant clusters when they are in solution, have a dark brown colour. Their absorption from the metal core in the UV-VIS is similar qualitatively to that for colloidal Pd.In addition bands from the ligands are observed. Prof. E. Matijevid (Clarkson University, New York, USA) asked: Did you try to use non-destructive techniques, such as light scattering, to study the size of your ‘giant clusters’ in solution? Prof. Zamaraev responded: In our preliminary experiments we have failed to detect the size of our clusters in solution with light scattering. Perhaps, now we should repeat our attempts using more sophisticated instrumentation. However, we have measured the sizes of our clusters in solution with SAXS. The same sizes were obtained as for solid samples described in this paper. General Discussion Prof.A. M. Bradshaw (Fritz-Haber-Institute, Berlin, Germany) commented: there seems to be a problem here in establishing the structure of the giant cluster using EXAFS alone, which is perhaps not surprising in view of the localised nature of the primary excitation in this experiment. Might not wide-angle X-ray scattering be of more help in establishing definitively that it is not f.c.c.? Prof. Zamaraev responded: I think you are right that we have to look more attentively into the possibility of using wide-angle scattering to characterize our clusters. Perhaps, its combination with EXAFS can indeed bring more light on their structure. Prof. Thomas asked: Since your giant clusters are water soluble, you do not get a ‘Tyndall cone’ when you pass light beams through your solutions? But what are their long-term stabilities? In time, no doubt, these giant clusters must aggregate or dispropor- tionate.What precisely happens? Prof. Zamaraev responded: There is some scatter in the lifetimes of our Pd561phen60(OA~)180cluster in solutions. In acetic acid solution it is typically stable for several months, but sometimes for more than a year. As a solid it is stable under air at least for two years. Heating of this cluster in acetic acid solution leads to its dispropor- tionation into bigger Pd metal particles and Pd(I1) complexes. Professor P. P. Edwards (University of Birmingham) commented: This is a very nice example of using new chemical routes to the synthesis of very small particles of colloidal metal, or ‘giant clusters’.However, I would ask if you could amplify specifically the experimental technique which leads you to conclude such a precise nuclearity e.g. 570. Certainly, the low resolution TEM (Fig. 6 in your paper) could not allow one to give such precision. Also from our own work over the past six years (including measurements on the 400 keV microscope) we have great difficulty in attaining such high precision as quoted in this paper. Could you also give an estimate of the ‘nuclearity distribution’ that corresponds to your measured size distribution [Fig. 6( b)] with d ranging from ca. 15-40 A? A final point, during our entire programme on colloidal metals (of some six years duration) we have always found, even within narrow size distributions, a range of particle morphologies as revealed by high-resolution electron microscopy (see Fig.3 and 4). You clearly do not, I wonder if you could comment? Prof. Zamaraev responded: (1) As indicated in our paper, the Pd561 nuclearity of our cluster is an approximate one. It was chosen on the basis of the following data. With the known (from EXAFS) character of the packing of Pd atoms and the average distance between the nearest-neighbour Pd atoms in the cluster, we have estimated the number NEof Pd atoms in the cluster molecule. As found for a sphere of 25A in diameter (which is approximately the average size suggested for our cluster by TEM, SAXS and XRD data), NE is ca. 570. On the basis of this value for NEand of the chemical composition of the cluster which was found from elemental analysis, it was approximated with the formula Pd570phen63( OAc) The molecular mass corresponding to this formula (M = 83 200) agrees well with the value M = (1*0.5) x lo5, which was estimated from the rate of sedimentation of our cluster from solution.The value found, Nz=570, matches well the ideal five-layer icosahedron which contains Nz = 1/3 (lorn3+ 15m2+ 11rn +3) =561 metal atoms for rn = 5. Note, that for the ideal four-layered icosahedron the nuclearity is Nz = 309. For the cluster with our General Discussion Fig. 3 Direct imaging by high-resolution microscopy of a small (cu. 35 A diameter) decahedral silver particle. The structural model employed to support this assignment is also given; this consists of 1427 silver atoms.Reproduced with permission from D. G. Duff, A. C. Curtis, P. P. Edwards, D. A. Jefferson, B. F. G. Johnson, A. I. Kirkland and D. E. Logan, Angew. Chem., Int. Ed. Engl., 1987, 26, 676 Fig. 4 High-resolution image of an icosahedral gold particle, together with a simulated image for a 923 atom particle. Reproduced with permission from A. I. Kirkland, D. A. Jefferson, D. Tang and P. P. Edwards, Proc. R. SOC.,London, 1991, 434, 279 chemical composition this would mean a molecular mass M =45 800, which still does not differ markedly from the value found experimentally. For a six-layered icosahedron the nuclearity is N = 923. This would mean a molecular mass M = 137 000, which again does not contradict the experimental value of M.However, on the grounds of similar calculations we can rule out the three-layered icosahedron with Nx= 147 and M = 21 800, which is too small, and the seven-layered icosahedron with Nx= 1415 and M = 210 000, General Discussion which is too big. Thus, strictly speaking we cannot distinguish unambiguously from our data between the cases of four-, five- and six-layered icosahedra, though we can reject icosahedra that are bigger or smaller. Of course, the idealized formula Pd561phen600AC180 which corresponds to the ideal five-layered icosahedral structure, characterizes some average nuclearity and composi- tion of the molecules of our cluster, rather than a precise fixed size and composition. We have chosen the number of layers in the icosahedron rn = 5, since first it is the average for the set of the values m =4,5 and 6 which agree with the experimental value of the molecular mass of the cluster, and secondly, it agrees best with the average size of our cluster as found from TEM data.Note, that for an equal number m of the layers of metal atoms, ideal cuboctahedra contain the same number of atoms as icosahedra. Therefore, the above-mentioned considerations should be valid for clusters with cuboctahedral packing as well. In other words, they should be equally valid for our cluster in both unrelaxed icosahedral form and relaxed cuboctahedral form, provided that the distances between the nearest- neighbour Pd atoms in the two forms are about the same.What makes our Pd cluster different from normal colloidal particles, is a distinct ligand environment with definite stoichiometry. However, a set of more or less imperfect metal polyhedra with a certain size distribution, rather than a single size perfect polyhedron as in molecular clusters with small nuclearity, seems to coexist when the number of metal atoms amounts to several hundred. In this respect our cluster resembles, perhaps, organic polymers, where for a given chemical composition the polymer can contain a set of different molecules of varying length and isomeric structure, rather than identical molecules of the same length and isomeric structure. (2) The average diameter and the width at the half-height of the size distribution for our cluster according to TEM data are 26k3.5 A and ca.8 A, respectively. These data agree well with the view of our cluster as a species which contains on average five layers of Pd atoms with the width of the distribution over the number of the layers *1 layer. According to SAXS data, the width at the half-height of the size distribution for our cluster is cu. 13 A (after subtraction of the instrumentation broadening). The difference between the widths of the size distribution obtained from TEM and SAXS was attributed to the fact that the size distribution found from TEM is expected to characterize the metal core of our cluster, while size distribution found from SAXS may contain also a contribution from the size of the ligand shell.(3) I think that the reason for a rather uniform spherical morphology of our clusters is, perhaps, that their growth is strongly controlled by the ligands. In any case, this is a precise chemical composition that makes our cluster different from most other colloidal particles. The following comments were made later in the meeting but are printed here as they are closely related to the preceding discussion. Prof. Zamaraev said: In our paper we have described the procedure for preparation of giant Pd clusters of distinct chemical composition, starting with Pd" acetate+ phenanthroline solutions in acetic acid which are treated first with H2 and then with 02.The giant cluster with the idealized formula Pd561phen60(OA~)180 was prepared and characterized by various spectroscopic techniques. Substitution of various other anionic ligands for OAc- ligand in this cluster was found to generate a family of giant Pd clusters.An obvious question arises, whether the same procedure can be used to prepare giant clusters of other metals. I should like to present some preliminary data that show what happens when Pt" acetate is used instead of Pd" following the same preparation procedure. General Discussion The Pt clusters formed in acetic acid solution were characterized with EXAFS, TEM and SAXS. The details of cluster preparation and characterization will be given else- where.' I shall describe only the main results of this study. Under otherwise identical conditions, the size and structure of the Pt particles formed depends strongly on the [Pt] :[phen] ratio.Fig. 5 shows the RDA curves found from Pt L3 EXAFS spectra for solid particles precipitated from solutions with [Pt] :[phen] = 2 :1 and 8 :1. For [Pt] :[phen] =2 :1 the solid particles were found (from elemental analysis) to contain 71% Pt, the rest being phen and acetate ligands. The average size of the particles was (22 *3) A according to both TEM and SAXS. For [Pt :[phen] = 8 :1 the solid particles contained 97%Pt and their average size was (30f3) B according to SAXS. RDA curves for the two samples are quite different. For [Pt] :[phen] =8: 1 the pattern of RDA peaks is practically the same as that for Pt foil, except that the magnitudes of all peaks are notably smaller for the cluster.Note that similar effects were observed for our Pd561 cluster of ca. 25 A size and Fe particles of ca. 12 A size (see Fig. 1 in this general discussion) and tentatively attributed to the inelastic scattering of the electron waves on the boundary of a small metal particle. For [Pt] :[phen] = 2: 1 the pattern of RDA curves is quite different from that for Pt foil. In particular, a very intense Pt-ligand peak is observed, while the magnitude of the peak from even the nearest-neighbour Pt atoms is notably smaller than that for the cluster obtained with [Pt]: [phen] =8: 1. Thus, the environment of Pt atoms in the particles formed with [Pt] :[phen] =2 :1 is much richer in ligand molecules and much less rich in Pt atoms compared with the particles formed with [Pt] :[phen] = 8 :1.In Fig. 6 Pt L3 adsorption edges are compared for Pt metal and Pt clusters prepared at various [Pt] :[phen] ratios, namely 8 :1, 4 :1 and 2 :1. For the cluster prepared at [Pt] :[phen] =8 :1 the shape of the edge coincides with that for the bulk Pt metal, while for clusters prepared at [Pt] :[phen] =2 :1 it is notably different from that for the bulk metal and similar to the shape of the edge for the complexes where Pt is in an oxidized state. For [Pt] :[phen] =4: 1 the shape of the edge is intermediate between those for [Pt] :[phen] =2 :1 and bulk Pt metal. Note, that although according to EXAFS data clusters prepared with [Pt] :[phen] = 2 :1 have a dramatically different structure from that of Pt metal, in TEM microphoto-graphs they look like normal small metal particles.This shows clearly that the dark spots in TEM microphotographs, that many people consider to be small metal particles, may in fact be species that, in addition to metal atoms, contain also a large amount of ligand material. Fig. 7 shows how the average size and chemical composition of Pt clusters prepared from the solution with the same initial [Pt] :[phen] ratio 2: 1 vary with the H2 reduction time prior to quenching with 02.Analysis of these data has shown that as the time of reduction is varied from r =2-4 h the mean diameter of the clusters increases from 15 f3 8, to 22 f3 A,while the width at half-height of the size distribution remains almost the same A,,2 = 10-12 A.Elemental analysis of the clusters formed at different r has revealed two important features concerning their composition. First, for fixed initial [Pt] :[phen] = ratio 2 :1 and fixed r the number of ligands per one Pt atom varies from one preparation series to another, though not over a very broad range. Secondly, as r increases, the relative content of Pt in the clusters increases, while that of the ligands decreases. The magnitudes of these variations of the composition for the clusters prepared in solutions with the same initial value of [Pt] :[phen] =2 :1, but at different r can be seen from the formulae in Fig. 7, i.e. [Pt,phen(OAc),-,I, at T = 2 h and [Pt,phen(OA~)~.~-,], at r =4 h. These data show that the same preparation procedure described in our paper, fv;; notably different results with platinum acetate. For Pd, a giant cluster (ca 25 diameter) of definite chemical composition, though with a certain size distribution among 90 General Discussion 0.10 hulY.- a $ W 4 Y .n 4 0 2.0 4.0 6.0 8.0 (R -WA 0.60 n CI .n a $ W 4 Y .CI 2 E Fig. 5 RDA curves calculated from RL3 EXAFS spectra of platinum clusters preci itated from solutions with [Pt]:[phen] (a)2:l and (b) 8:l.(a)Cp,=71wt.%; dm,,,=(22*3) 1,TEM and SAXS data. (b) C, = 97 wt.%; d,,,, = (30 * 3) A, SAXS data General Discussion E Fig. 6 Pt L3 absorption edges for the bulk Pt metal and Pt particles prepared at various [Pt] :[phen] ratios; (a) 8 :1 and bulk Pt metal; (b)4: 1 and (c) 2: 1 (a) (6) Fig.7 TEM microphotographs of the particles with the average compositions [Pt,phen(OAc),-,], and [Pt4phen(OAc)o,5-l], that precipitate from solution in acetic acid with [Pt] :[phen] =2; 1 at and (b) ~=4two different H2 reduction times. (a) ~=2 h (a) Cp,=67wt.%; d,,,,=(15*3) A; A1/2=12A. (b) CR=71 wt.%; dm,,,=(22*3)A; Al,,=10A its molecules, is formed. The idealized formula, that corresponds to the average size and composition of this cluster, is Pd561phen600A~180. For Pt a broad range of species with various average sizes and relative content of the metal and ligands are formed. The precise average size and chemical composition of Pt-containing particles under given preparation conditions do not vary much and can be intentionally controlled by varying these conditions. 1 M.N. Vargaftik, I. 1. Moiseev, D. 1. Kochubey and K. I. Zamaraev, to be published. Dr. J. S. Bradley (Exxon Research and Engineering Co., New Jersey, USA) asked: In the preparation of the PdS6, cluster you initially reduce palladium acetate with General Discussion hydrogen in the presence of phenanthroline or bipyridyl, which presumably reduces the palladium to the zero-valent state. You then pass oxygen through the solution. This last step is reminiscent on the one hand of the passivation of highly dispersed heterogeneous supported metal catalysts, and, on the other, of the addition of capping reagents to colloidal semiconductor clusters (the work of Steigerwald and others) to prevent further cluster growth.Do you think that oxygen plays such a role, as a passivator or terminating agent to prevent further cluster aggregation? If so, do you think that the surface of the cluster is oxidised to Pd"? If this is so, perhaps the addition of carbon monoxide to the cluster would reveal the oxidised state of the surface by the frequency of the C-0 stretching mode. Have you observed CO adsorption by the palladium clusters? Prof. Zamaraev replied; The giant Pd561phen600A~180 cluster is formed through the 20A Pd hydrido cluster as a precursor. The latter has the composition [Pd,phen(OAc),H,], with n ca. 100 and is formed upon reduction of Pd" acetate by H2 in the presence of phenanthroline according to eqn.(VII) in our paper. The role of o2is to enable the assembly of the framework of the giant Pd561phen600A~180 cluster by removing hydrido ligands from the precursor cluster. The negative charges of the 180 OAc ligands in the giant cluster must be balanced by positive charge of the same magnitude on the metal core of the cluster. This means on average a charge of ca.+ 1/3 per one Pd atom of the cluster. However the actual distribution of this positive charge over the Pd561 core is still unknown. I agree with you that most probably it should be localized near the surface of the metal core where the phenanthroline and acetate ligands are located, rather than inside the metal core. Probing with CO may indeed prove useful for revealing the oxidation state of the surface Pd atoms in the cluster and we plan to do such experiments.Prof. Edwards asked: Could I once again ask for experimental clarification of the data for 'magic numbers' in these colloidal particles. You make reference to cluster chemistry, where, for example, the availability of single crystal data leads to an unambiguous derivation of cluster nuclearity. Obviously, what you have here is a particle size distribution and a corresponding particle nuclearity distribution. To summarize my earlier comments, following six years of study of colloidal particles of Cu, Ag, Au, Pt and Pd we find no evidence (even under detailed HREM investigation) of a 'magic number' system in which all particles possess such structure.One generally finds, even with narrow size distributions, a vast array of different structure types and particle morphologies. Great caution has to be exercised in evaluating particle nuclearity. Prof. Zamaraev responded: What makes Pd colloidal particles prepared (and isolated) via our technique different from colloidal particles of this and other metals prepared via other techniques, is their exact chemical composition, i.e. [Pdgphen(OAc),], or [Pd9 bipy (OAc),] as ascertained by elemental analysis and stoichiometry of their formation. Under controlled experimental conditions the precise stoichiometry ([Pd]:[phen]: [OAc] =9: 1:3) was very well reproduced for our Pd particles, just as is always the case for the chemistry of normal metal clusters of small nuclearity with various ligands. Note that substitution of various other avionic ligands for OAc- ligands in our Pd particles proceeded also in a quantitative manner, i.e.with a certain stoichiometry. These experimental facts show clearly that we isolate our colloidal Pd particles as species with definite metal-to-ligand stoichiometry. That is why we call these clusters. They are giant clusters, since their average size is ca. 25 A. Under controlled experimental conditions this size is also very reproducible as is the chemical composition of our species. The way we arrived at a still more specific idealised Pd~6#hen6o(OAC)1,0 formula for our giant cluster was explained in my reply to your previous question. General Discussion A different picture emerged when we used the same procedure to prepare colloidal Pt particles.For Pt a broad range of species with various average sizes and relative content of the metal and ligands was formed. However, the exact average size and chemical composition of Pt-containing particles under given preparation conditions did not vary much, and it could be intentionally controlled by varying these conditions. I think that an important message from the chemistry of Pd and Pt colloidal particles with phen (or bipy) +acetate (or other anionic) ligands is that synthesis and stabilization of small metal particles should be always considered as chemical processes, in which participation of ligands seems to be extremely important at all stages.The nature of the ligand species can vary dramatically from one preparation technique to another, e.g. for our technique, when colloidal particles were precipitated from solutions, phen (or bipy) molecules and OAc anions (or other anions) served as ligands. For small metal particles supported on oxides, carbon etc., the functional groups of the surface of the supports (e.g., OH, C02H, C=CR2 etc.) as well as adsorbed species (e.g.hydrogen or oxygen) can serve as ligands. For particles prepared via matrix isolation, the atoms of the matrix can play the role of the ligands. In our preparation procedure we controlled the interaction between the metal atoms and the ligands more carefully than before. That is why, I think, we have succeeded in preparing Pd particles of a definite composi- tion, shape and average size.But you are certainly right to state that our particles demonstrate a particle size distribution (around the average size) which means a corresponding particle nuclearity distribution. In my response to your previous question I have already explained this aspect of the structure of our Pd cluster in terms resembling those used in polymer chemistry. I do not think that at the moment I can add anything else substantial to that explanation. Prof. Rao opened the discussion of the paper given by Prof. Henglein. He said: I have two questions: (1) How do you determine n in Ag, ? (2) Could you tell me about the photoelectron emission experiments? Prof. Henglein replied: (1) We do not yet know the exact value of n for the clusters that produce the strong absorption bands in the 300-370nm range.As our cluster solutions do not exhibit EPR signals, we have to conclude that they contain an even number of atoms. They probably represent certain magic numbers, ie. clusters of especially high stability. Our kinetic studies using pulse radiolysis'72 showed that the clusters must be smaller than the smallest silver particles that develop the plasmon absorption band. This band was found to appear at a cluster size of 12 reduced silver atoms.' 1 A. Henglein and R. Tausch-Treml, J. Colloid Znterface Sci., 1981, 80, 84. 2. P. Mulvaney and A. Henglein, Chem. Phys. Lett. 1990, 168, 391. (2) In the photoelectron emission experiments, the solution of the clusters or of the metallic particles is illuminated with a laser flash (at 308 nm where all these species have some absorption). The emitted electrons appear as hydrated electrons in the aqueous solvent where they live for ca.10 FS. These hydrated electrons can readily be traced quantitatively because of their high absorption at 700 nm. Dr. P. K. Wrona (University of Warsaw, Poland) asked: In your calculation of the Ag/Agf (monomeric Ag) did you take into account that monoatomic silver in solution could be highly solvated? Prof. Henglein replied: In contrast to the silver ion which is highly hydrated, the free silver atom is expected to have a very low free energy of hydration. Neutral atoms General Discussion (such as the noble-gas atoms) and neutral non-polar molecules have very small free energies of hydration.I think it is reasonable to compare the silver atom with these species with respect to its interaction with water. I believe that the hydration energy of the silver atom is smaller than 0.1 eV. The error in the redox potentials in Fig. 8 in our paper should be *0.1 eV. Prof. Thomas asked: The striking characteristics of dimeric silver stand out in your work. How widespread, if at all, is such behaviour of dimeric metals elsewhere in the periodic table? Prof. Henglein replied: The electrochemical potentials for n = 1-3 in Fig. 8 were calculated using the thermodynamic data of the silver species involved (ref. 1). These data came from mass spectrometric studies on the molecular composition of silver vapour.' Such data are also available for a few other metals, and we have calculated electrochemical potentials for these metals.It seems that a strong oscillation of the redox potential as a function of the agglomeration number exists only for silver (Fig. 8). In the case of lead, for example, one can see the oscillation but the amplitude is very 1 K. Hilpert and K. A. Gingerich, Ber. Bunsenges. Phys. Chem., 1980, 84, 739. 2 A. Henglein, Chem. Rev. 1989, 89, 1961 (see Fig. 3). Prof. Matijevie asked: It is well known that silver forms well defined polynuclear complexes with halides, such as Ag2Cl+, Ag3C12+, which define the solubility of silver in solutions of halide salts.Do you see an analogy between this behaviour of silver ions and the polymerisation of metallic silver revealed to the low number of solute species? Prof. Henglein replied: The complexes you mention are complexes in which the oxidation state of silver is + 1. In our case, the clusters are formed by complete reduction of Ag+ ions. I do not see an analogy between our silver clusters and the silver ion complexes. Prof. Edwards commented: Your standard redox potentials (Fig. 8) are most interest- ing. Does the value for Ago/Ag+ (agglomeration, n = 1) imply that Ago, once formed, would spontaneously reduce the host solvent? We have often wondered how it can be that Ag colloids form; that is, is the first process the formation of a single Ago atom (in an aqueous environment), or rather the agglomeration via reduction of ionic-type aggregates Agi, Ag:+ etc.? Do you expect Ago once formed, to be heavily solvated? Could you contrast the value of Ago/Agf with that for Nao/Na'? Prof.Henglein responded: The free silver atom can undergo only one-electron transfer reactions. The initiating step of hydrogen formation would be: Ago + H20-+ Ag++ H + OH-. This reaction, in which the free H-atom appears, is endoergic by 1 eV. It cannot occur. The build-up of colloidal silver from Ago atoms has been studied by pulse radiolysis.'92 After the formation of Ago by a pulse of radiation, the reaction Ago+ Ag+ --+ Agz occurs in the microsecond range. Agt dimerizes within ca. 100 ps, and the Ag:+ species formed undergoes association and dismution reactions to build up larger particles in the mil- lisecond to second range.I do not expect the free silver atom to be strongly hydrated. Neutral atoms (such as the noble-gas atoms) are known to have very small free hydration energies. General Discussion That the free silver atom is a strong reducing species has also been shown by pulse radiolysis.' It rapidly transfers an electron to many organic and inorganic acceptors. Eg. Ago+C1CH2CO2H--* Ag++C1-+CH2C02H Ago+Cu2+ + Ag++ Cu' Note that electron transfer in the latter reaction occurs in the opposite direction to that expected from the conventional standard potentials of the systems Ag+/ Ag and Cu2+/Cu+. With a standard potential of Ago of -1.8 V, the free silver atom is almost as reducing as a sodium electrode.1 R.Tausch-Treml, A. Henglein, and J. Lilie, Ber. Bunsenges. Phys. Chem., 1978, 82, 1335. 2 A. Henglein and R. Tausch-Treml, J. Colloid Interface Sci. 1981, 80, 84. Prof. M. W. Roberts (University of WaZes College of Curdim commented: Perhaps I can draw a parallel with bulk single crystals of Ag and Zn and your observations of 'chemisorption enhanced corrosion' of your sols. Although dioxygen bond cleavage at Zn(0001) and Ag(ll1) surfaces is very slow at 295 K, in the presence of ammonia dioxygen dissociation at a Zn(0001) surface is very fast. Furthermore the kinetics show all the characteristics of the involvement of a precursor complex involving dioxygen and ammonia.Can you think of your reactions with sols in the same way? Prof. Henglein replied: The silver particles carrying nucleophilic molecules (such as NH,, CN-, and SH-) on their surface are oxidized not only by dioxygen but also by many organic and inorganic electron acceptors [such as nitrobenzene, methyl viologen, and hexacyanoferrate( HI)].' We explain this by electron transfer from the silver particles to the acceptors. It cannot be excluded that this electron transfer occurs through an intermediate complex but we do not have any experimental evidence for this. 1 P. Mulvaney, T. Linnert, and A. Henglein, J. Phys. Chem., in the press. Dr. H. Miessner (Zentrumfur Heterogene Katalyse, Berlin, Germany) said: Small silver clusters are known to be stabilised also by solid matrices e.g.in the micropores of zeolites. Could you give a comment on the similarities and/or the differences of the properties of small Ag clusters in relation to their environment? Prof. Henglein replied: When we found the strong absorption bands of silver clusters in aqueous solution, we compared our data with those obtained in matrix studies such as clusters in zeolites and in solid noble-gases at low temperature. It was not possible to correlate any of the cluster bands in solution with bands in the solid environments. The assignment of optical absorption bands to structures of clusters in zeolites and frozen matrices are rather vague for n >3 and the absorption coefficients are not known. Our clusters have absorption coefficients of ca.20 000 M3 mol-' cm-'. Dr. B. Mile (University of WaZes CoZlege of Cardim commented: Referring to Fig. 8 in your paper relating changes in the standard redox potential with the agglomeration number n it may be that the potential for even-numbered clusters, Ag,, Ag,, Ag6, will all show minima while those for odd-numbered clusters Ag, ,Ag,, Ag, will show maxima i.e. Ag, is not unique but represents the start in a series of oscillations of potential between even- and odd-numbered particles. Of course such oscillations are well known for the ionisation potential of gas phase clusters of increasing size. Prof. Henglein replied: It seems probable that the oscillation in the n = 1-3 range in Fig. 8 is not the only one.Weaker oscillations due to odd-even effects may exist at higher values of n. We do not know the exact behaviour of the curve at larger agglomeration numbers (as the thermodynamic properties of the larger clusters are not known) and have therefore dashed it for n > 3. General Discussion Dr. Gallezot opened the discussion on the paper by Prof. J. L. Dye (Michigan State University, USA) Your technique is very promising for the preparation of bimetallic clusters at low temperatures. Have you tried to stabilize the colloids by using appropriate ligands (soluble ligands, polymers, solid supports) ? Prof. Dye replied: We have not yet attempted to stabilize the colloids. When dilute ( to mol dm-3) precursor solutions are used, colloid formation is nearly always observed.Aggregation seems to occur more rapidly in dimethyl ether (Me,O) than in tetrahydrofuran. Sometimes, upon evaporation of Me,O, colloidal solutions form when the residue is taken up in methanol or water. Prof. Henglein asked: How can methanol oxidize metals such as Fe or Ni? Do you observe the formation of methane? Prof. Dye responded: The oxidation may not be by methanol. In order to wash the metal particles we find it necessary to centrifuge the solutions inside a glove bag. Under these conditions it is very difficult to exclude oxygen completely. Prof. Rao asked: With such a small particle size how can you consider the reduced metallic product to be an alloy? Prof. Dye replied: The word ‘alloy’ may not be correct for very small particles.Perhaps we should simply call them intermetallic compounds. In the case of solid solutions, however, the distinction becomes more difficult. Dr. Bradley commented: We have found that uniform bimetallic colloidal clusters can be obtained from mixtures of metal salts, when (i) the two metals are miscible in the concentration range used and (ii) when one metal salt is easily reduced to the metal and can then function as a catalyst for the reduction of the second. Copper and palladium is a case in point. Prof. Dye commented: We do not know how general the formation of compounds is, or whether one can get uniform miscibility in some cases. Our electron diffraction results show that reduction of a mixture of zinc and gold salts in homogeneous solution yields the compound AuZn, while copper and gold form the corresponding compound AuCu with a cubic CsCl structure.Dr. J. T. Gauntlett (ICI, Runcorn) said: In my limited experience of attempts to make mixed metal colloids, they frequently fail due to the different reduction potential of the metal salts. Is your apparent success due to the use of a ‘battering ram’ reagent which reduces everything rapidly? Prof. Dye replied: Yes. One of the potential advantages of the present method is that reduction is so fast that sequential reduction seems not to be a problem. Prof. Edwards commented: This type of approach is clearly of considerable potential. I suppose it is formally akin to that of ‘dissolving metal reductions’ in organic chemistry: but now with colloids! I wonder if there is any qualitative difference between reduction via one -electron systems (e.g.e,) and two -electron systems (e.g. Na-) of the sort you have here? We have also had some success recently in using metal solutions to reduce zeolites (at low temperature) to form new clusters, such as Naf’ etc.’ It would be interesting to try the same type of reductions with your ‘electrides’ or ‘alkalides’. 1 P. Anderson and P. P. Edwards, Angew. Chem., Innt. Ed. Engl., 1991, 30, 1501. General Discussion Prof. Dye replied: Generally, we carry out the reduction by titrating to an end point at which the blue colour of excess alkalide or electride is observed. We have not found any difference in the products formed by using alkalides compared with those that are reduced with electrides.Indeed, the formation of reduced zeolites by exposure to aprotic alkalide or electride solutions would be worth trying. When we used metal-ammonia solutions, only amide was produced. Dr. Bradley said: In the reduction of mixed metal salts, and you give the example of Au-Zn, you report the formation of AuZn regardless of the relative concentrations of the two salts. Two questions: What happens to the excess metal salt? In the case of metal salt mixtures, for metals which are miscible over the entire range of compositions, does the stoichiometry of the resulting colloid reflect that of the salt mixture? Prof. Dye replied: With AuZn we were somewhat surprised to note the formation of only the one-to-one complex since other compounds such as Au,Zn and AuZn, are possible. The XPS results suggest that the excess salt forms the respective metal rather than another compound.Of course, there could be coincidental overlap of compound XPS signals with those of the pure metal. We have not yet found an example of solid solution formation. This is not surprising since miscibility is often found only at high temperatures. Prof. Henglein asked: What is the stoichiometry of your reductions? When you reduce a more electronegative metal such as Cd or Zn, do you need a large excess of the reducing agent? Prof. Dye replied: As far as we can tell, the reduction is stoichiometric. Any excess reductant imparts a stable blue colour to the solution.At CQ. -3V, the reduction potential of the solvated electron is sufficiently negative to reduce virtually any transition metal ion. Furthermore, the reduction is very rapid. Prof. H. Kuroda ( University of Tokyo, Japan): In the case of small bimetallic particles, the distribution of the constituent metals in a particle is often not uniform. I would suggest carrying out the characterization of the structures of your bimetallic alloy particles by means of EXAFS spectroscopy. If you perform a careful analysis of the EXAFS data of the constituent metals, you could get some useful information to judge whether the two metals are forming a uniform alloy or not. Prof. Dye replied: EXAFS is the method of choice to determine the particle size for particles that are small enough to have average coordination numbers reduced from those in the bulk. It is even more powerful for bimetallic particles because both the nature of the neighbours and their number can be determined.The occasional use of EXAFS in the US is not easy however. We hope to be able to collaborate with someone who uses the method routinely. Prof. Roberts said: The Au4f spectra in Fig. 3 of your paper could be interpreted in a different way; namely, that an electronegative species [CI-(a)]is chemisorbed and this would be accompanied by the observed shifts in the binding energies. The shifts do not necessarily imply Ti-Au formation. The ease of removal (washing in methanol) also supports the presence of a chemisorbed species and not the formation of a Ti-Au alloy as suggested.Prof. Dye replied: Although chemisorption of electronegative species can indeed, shift XPS peaks, this explanation seems unlikely in our case. A shift of the gold XPS pattern never occurs with gold alone, even when the chloride reaction products are present. Also in the gold-zinc and gold-copper cases, there is no effect of chloride. Genera1 Discuss ion Prof. Roberts commented further: My comment was prompted by the fact that at both magnesium' and lead surfaces2 the Mg 1s and Pb 4f peaks are shifted by some 3.8 eV and 2 eV respectively to higher binding energies when chemisorbed chloride is present. I would have anticipated the Au f binding energy to behave similarly.1 C. T.Au and M. W. Roberts Surface Sci., 1985,149,L18. 2 P.G.Blake, A. F. Carley, V. Di Castro and M. W. Roberts, J. Chem. SOC.Furuduy Trans. 2, 1986,82,723. Dr. J. A. Creighton (University of Kent) asked: Have the authors considered using UV-VIS absorption spectroscopy to characterise these colloidal metals produced by their method? The advantage of using alkalides or electrides over methods using more conventional chemical reducing agents is that colloids of the more electropositive metals can be produced. It is obviously desirable to characterise these in situ in the H-cell, and UV-VIS absorption spectroscopy lends itself very well to this. A compilation of absorption spectra of 52 of the colloidal metallic elements has been published.' The absorption spectra of the bimetallic colloids would probably also indicate whether they are alloy particles or simply mixtures of particles of the two separate elements.1 J. A. Creighton and D. G. Eadon, J. Chem. Soc., Furuday Trans., 1991,87, 3881. Prof. Dye replied: We have so far obtained spectra only of gold and copper colloids in THF. To extend this to other metals we need to find a way to stabilize the colloids in dimethyl ether. The suggestion is a good one. Prof. Henglein commented: When I read your paper for the first time, I found it surprising that you would get AuZn. One would think that reduced Zn would rapidly reduce Au+ until all gold ions are reduced and then the reduction of Zn2+ would follow to give Zn metal.However, one has to keep in mind that on the atomic scale redox processes between metal species can occur in a direction opposite to what is expected from conventional electrochemistry. (For example, Ago reduces Cu*+). Prof. Dye replied: The reduction process is probably nearly simultaneous, rather than sequential. Dr. Masson commented: I have a general comment concerning the bimetallic particles you are able to obtain so beautifully by your technique, but it seems to me that the main problem of the activity of these particles is related to a spatial distribution problem. How for instance, is it possible to follow the segregation problems? I wonder if it is possible because all the techniques employed are without spatial resolution. Prof.Dye responded: As described in the paper, the XPS results suggest, but do not prove, compound or ailoy formation. They do show, however, that we are not just forming separate pure metal particles. In the cases of CuAu and ZnAu the electron diffraction results are definitive and prove compound formation. Dr. D. J. Greenslade (University of Essex) asked: Surely we need to get mass spectroscopists interested in developing techniques for colloids comparable to those applied to biological molecules, such as FAB techniques? Prof. Dye replied: I agree, it would be very desirable to determine cluster sizes by mass spectrometry using either FAB or laser desorption to get the clusters into the gas phase. We intend to investigate this possibility. Prof.Rao opened the discussion of Prof. Boudart's paper: How can we be so sure that the Rh clusters have 12 atoms? I am not sure EXAFS can give such exact answers; it can only give a coordination number. General Discussion Prof. Boudart (Stanford University, USA) replied: It is easy to go from an average coordination number for a metallic cluster to an average number of atoms per cluster on the basis of hard sphere models, as discussed in ref. 45 in our paper. Prof. Joyner said: This is an important paper, but its importance is that of the dog that did not bark in the night in the Sherlock Holmes story, 'Silver Blaze'. The lack of differences in the range of hydrogenations studied by Boudart may conceal a number of significant points. (i) I have advocated that heterogeneous catalytic activity may be considered within a molecular framework: the catalyst is described as a surface molecule which is modified by embedding in the matrix.' Alkene hydrogenation is clearly very similar on single metal atoms and on metal particles.Yet embedding a single atom in a particle has marked electronic consequences. Is Professor Boudart not surprised that these are not more clearly mirrored in catalytic performance? (ii) You suggest that hydrogen activation is rate determining on heterogeneous catalysts. Calculations suggest that on the homogeneous Wilkinson catalyst this step is unactivated and therefore unlikely to be rate determining.* Again, I welcome your comments. 1 R. W. Joyner, Catai.Today., in the press. 2 N. Koga et al., 3. Am. Chem. SOC.,1987, 109, 3455, Prof. Boudart replied: (i) Indeed, I am very surprised about the main result of our paper. It suggests to me that embedding a single metal atom in a cluster has electronic consequences resembling those found with a Wilkinson catalyst, at least for alkene hydrogenation. (ii) The brilliant theoretical work of Koda et al. is for phosphine complexes, not phenyl phosphine complexes present in a Wilkinson catalyst. More calculations are in order before the point you raise can be settled. Prof. Joyner added: It is worth noting that metal/support and support/metal interac- tions are not symmetric, at least as regards range of operation. A metal/support interaction, in which the electronic properties of a metal are modified by a support is of very short range only.Because of the strong screening effect in metals, this interaction extends usually for only CCI. 3 A. Through the Schottky barrier effect, however, the metal may modify the support over a much longer distance. The range may exceed 30 A and be limited either by bond-bending or electron tunnelling through the support.' 1 J. C. Frost, Nature (London), 1988, 334, 587. Dr. Bradley said to Professor Boudart: You mentioned the possible importance of an organic or carbonaceous overlayer in heterogeneous catalysis. Could you be more precise about the nature of this overlayer? Do you mean by this a chemisorbed layer of reactive (or spectator) species, or perhaps a layer of physisorbed reactant molecules? The reason for my question is that in a liquid phase high resolution 13CNMR experiment on the adsorption of ethene on colloidal platinum clusters, my colleague John Millar has observed, in addition to free ethene in solution, a second ethene species, with a chemical shift only 3 ppm upfield from the resonance for the free ethene.This shift is not sufficient to be ascribed to chemisorbed ethene. When exposed to hydrogen, both resonances disappear as ethane is formed. Could this second form of ethene be simply physisorbed on the metal particle, and if so, is this the sort of overlayer to which you refer? Prof. Boudart replied: The organic overlayer on metal surfaces during hydrogenation of ethene could be due to the so-called half-hydrogenated state, a reactive intermediate, or to a spectator species, namely ethylidine.But, in the case of the latter species it General Discussion cannot be imagined in the case of cyclohexene. I cannot rule out an overlayer of physisorbed alkene, but I believe the species that we invoke to explain structure insensitivity must be more strongly held to the surface. A clear-cut case of adsorbate- induced structural change has just been reported in the case of CO on a Pd tip observed by field ion microscopy.’ 1 A. Gaussmann and N. Kruse, Cutal. Let?., 1991, 10, 305. Prof. Zamaraev commented to Dr. Bradley: You mentioned that the NMR line of ethene is shifted when a colloidal metal is added to the solution. In connection with this observation I should like to draw attention to the paper by Nekipelov and Zamaraev,’ where even very weak outer-sphere coordination of organic molecules to metal complexes in solution was shown to perturb notably the NMR spectra of these molecules.In most cases the conditions of fast exchange between the free state of the organic molecule in solution and its state in the outer coordination sphere of the metal complex were fulfilled. Under these conditions the observed NMR lines were the average ones for the molecule in the free and coordinated states, and they were shifted with respect to the ‘free-state’ lines observed in the absence of the metal complex. I suggest that the data reported by Dr. Bradley can, perhaps, be explained in a similar way, i.e.by a coordination of ethene . molecules to colloidal particles and fast exchange of ethene between this coordinated state and the free state. And to Prof. Boudart: From the studies of catalysis over single crystals, it is known that certain reactions (e.g. ammonia synthesis over iron and rhenium,2) proceed with dramatically different turnover rates over different crystallographic faces of one and the same metal. At the same time for certain other reactions (e.g. hydrogenation of cyclo- hexane described by Prof. Boudart in his paper), the turnover rate is rather insensitive to the initial structure of the catalyst. I should like to draw your attention to a possible explanation of such different behaviour which can be given following the ideas that have been formulated many years ago by Bore~kov,~ Note, that and further substantiated in his more recent m~nograph.~ the regularity is valid for catalysis which resembles Newton’s third law, i.e.the action is equal to the counteraction. If the catalyst has sufficient power to perturb the reacting molecules via intermediate chemical interactions to such an extent that they become converted into molecules of reaction products, then via the same intermediate chemical interactions the molecules of the reaction mixture should be equally able to perturb the structure of the catalyst and, in the extreme case, to reconstruct the latter to the state that corresponds to the minimum of the free energy of the overall ‘catalyst and reaction mixture’ system under given reaction conditions.For some catalytic processes over single crystals of metals such reaction-induced reconstructions of the surface atomic layer of the catalyst have indeed been observed experimentally. The chances for such reconstructions are expected to be greater for stronger intermediate chemical interactions between the catalyst and the molecules of the reaction mixture and for higher reaction temperatures. When reconstruction occurs up to the above-mentioned state that corresponds to the minimum of the free energy of the overall system, then the expected intrinsic difference between the turnover rates for the catalysts that initially have been in different states, is gradually eliminated. And after a certain period of time (the duration of which is controlled by the rate of the reconstruction) the turnover rate of the catalytic reaction is expected to reach the same steady state value, whatever the initial state of the catalyst. This steady state value corresponds to the steady state of the reconstructed surface of the catalyst under given reaction conditions.Thus, the same reaction over the same catalyst can in principle demonstrate both structure-sensitive and structure-insensitive behaviour, depending on whether the reac- tion rate is measured before or after the reaction-induced reconstruction is over. General Discussion Note, that a similar phenomenon can be expected for homogeneous catalysis. When reactants are added to the solution of a homogeneous metal complex catalyst, its chemical state also can be substantially perturbed.For example, molecules of the reacting compounds or reaction products can substitute all the ligands that have been coordinated initially to the metal atom. If this happens, the turnover rate is insensitive to the initial composition of the metal complex catalyst. Another possibility is the transformation of the initial metal complex in the reaction mixture into a very active colloidal metal. In this case the reaction rate under certain conditions again can, perhaps, be not very sensitive to the initial composition of the complex. I think that one shall have all these possibilities in mind when discussing structure sensitivity or insensitivity of the turnover rate for a particular reaction.1 V. M. Nekipelov and K. I. Zamaraev, Coord. Chem. Rev., 1985, 61, 185. 2 G. A. Somorjai, Proceedings of the 8th International Congress on Catalysis, Verlag Chemie, Beriin, 1984, vol. 1, p. 113. 3 G. K. Boreskov, J. Phys. Khim., 1958, 32, 2739. 4 G. K. Boreskov, Heterogeneous Catalysis, Nauka, Moscow, 1986. Dr. Masson said: It would appear that symmetry or coordination number are not good hypotheses to explain structure sensitivity, at least for the chemical reactivity you described on Rh. Do you think that these hypotheses could have a general significance? I should like to mention the work done at ATT laboratories by M. Jarrold concerning the specific reactivity of C2H4 on silicon (13) and silicon (14) clusters, a typical example of geometry and electronic effects on sticking coefficients. Prof.Boudart replied: Structure insensitivity in surface catalysis is difficult to explain. By contrast, structure sensitivity is easy to identify as a manifestation of crystalline anisotropy. The most striking example is in ammonia synthesis on clusters or large single crystals of iron. In both systems the large activity in the reaction has been ascribed to surface atoms with coordination number equal to seven.' 1 M. Boudart and G. Djiga-Mariadassou, Kinetics of Heterogeneous Catalytic Reactions, Princeton Univer- sity Press, 1984, pp. 168-169 Prof. Roberts said: Can I return to the comments of Dr. Bradley and Prof. Zamaraev. It is now well established that there are very efficient routes to products involving surface species which are not strictly chemisorbed and present at very low concentrations.Can you provide some information on the concentration of the 'unusual ethene' molecules present? Dr. Bradley replied: In experiments on the adsorption of 13C-ethene on colloidal platinum (a 10 A colloid stabilised by isobutylaluminoxane') we observed by 13CNMR a resonance shifted upfield by only 3 ppm from the resonance of free dissolved ethene. The second species was clearly ethene (from the 'H coupled spectrum), somehow associated with the metal particles (from its short TI)and was exchanging with the free ethene, whose TI was thus similarly shortened. The concentration was not measured, and we will determine the concentration in spin counting experiments, but from the relative intensities of the free and shifted resonances (ca.3 :1) I would say that it was present in significant amounts. 1 J. S. Bradley et al., J. Catal., 1991, 129, 530. Prof. Roberts commented: I agree very much with your point of view that catalytic reaction pathways are likely to involve very low concentrations of reactants. This is General Discussion one of the conclusions of our coadsorption studies (see my comments to Prof. Murrell). I just wondered whether your second ethene species was the reactive one? Prof. Edwards said: Given your very interesting conclusions regarding the apparent insensitivity of various processes to the state of aggregation, and indeed the heterogeneous/ homogeneous interface, I wonder if colloidal metals or sols may in fact be responsible for certain homogeneous catalytic reactions? As we have seen in this conference, the presence of colloidal metals, especially at low diameters, would be difficult to identify (or discriminate) positively in such liquid-state reactions.Clearly, experiments such as light-scattering etc. would be of very little use in this size regime. Prof. Boudart replied: To rule out homogeneous catalysis in a heterogeneous colloidal solution is as difficult to rule out as the opposite situation. Your point is well taken and should be considered seriously on a case by case basis. Prof. Zamaraev said: Prof. Edwards has raised an important issue: how chemists who study homogeneous catalysis with metal complexes in solutions can actually distinguish whether their reaction is catalysed by a soluble metal complex (homogeneous catalysis) or by colloidal metal particles that can be formed upon decomposition of the metal complex (heterogeneous catalysis).Note that chemists who study catalysis with colloidal metals in fact also face the same problem: how to prove that their reaction is provided by colloidal particles rather than by metal complexes that can be formed upon decomposition of colloidal particles. Specialists in catalysis have worked out various approaches to solve this dilemma. One of them is to look for the difference in selectivity towards various reaction products with a metal complex and a colloidal metal as catalysts. For example, as indicated in our paper, for oxidation of propene with oxygen in acetic acid solutions in the presence of the giant Pd561 cluster (which is a ca 25 A colloidal particle with a rather narrow size distribution) allyl acetate is predominantly formed, while in the presence of Pd" acetate a mixture of allyl, isopropenyl and n-propenyl acetates plus acetone is formed.This allows one to reject the possibility that catalytic properties of the giant cluster arise from Pd" acetate that can be formed upon decomposition of the giant cluster. Note also that a true homogeneous catalyst usually demonstrates a notably better reproducibility in the values of reaction rate than a catalyst whose activity arises from its decomposition into a colloidal metal. This difference in reproducibility can also be used to distinguish between the two types of catalysis.Dr. J. W. Couves (The Royal Institution, London) asked: The amplitude of EXAFS oscillations are dependent on both the coordination number and the Debye-Waller factor. What is the effect of the variation of Rh particle size on the Debye-Waller factor of these particles? Prof. Boudart replied: The Debye-Waller factors for the smallest and largest clusters listed in Table 1 were 49 and 15 pm2 respectively in the order expected for corresponding values of coordination numbers, 5.0 and 10.0 respectively. The above values are given in ref. 23 in our paper. Prof. Bradshaw said: If I have understood your conclusions correctly, the interatomic separation corresponds to that of the bulk even for clusters containing as few as 12 atoms, as long as they are covered with hydrogen or hydrocarbons.The corresponding bare clusters, on the other hand, show a contraction. Is it not surprising that the General Discussion chemisorption of such species is sufficient to give rise to the bulk lattice parameter in view of the difference between chemisorption bond energy and cohesive energy? Prof. Boudart replied: The phenomenon of contraction or expansion of the metal lattice parameter for clusters finds its exact parallel at the surface of large single crystals. In those systems, the interplanar distance at the surface is contracted if the surface is bare and relaxes to its bulk value with a monolayer of chemisorbed hydrogen.As you correctly imply, the explanation of these well established observations is not straight- forward. Dr. Gallezot said to Prof. Bradshaw: I do not agree with you that the binding energies of adsorbates with surface metal atoms are weak with respect to metal-metal ones. This is true for the associative adsorption of unsaturated hydrocarbons via n-bonding to the metal surface. Thus we have shown that benzene adsorption on platinum cluster hardly perturbs the structure.' However, if you consider metal- hydrogen or metal-carbon monoxide bond enthalpies they are quite similar to metal-metal bonds as far as platinum metals are concerned. This accounts for the modifications of the structure of Pt-clusters induced by CO adsorption that we have observed.2 1 G.Bergeret and P. Gallezot, J. Catal., 1981, 72, 294. 2 P. Gallezot, Zeolites, 1982, 2, 103. Dr. Benfield communicated: We have some new results (shown on our poster) which illustrate how uncertain it is whether some long-established reactions are homogeneous or heterogeneous in nature. In the Wurtz reaction, organic halides are reductively coupled with sodium to form a carbon-carbon single bond. Although this reaction was first reported in 1855l its mechanism remains ill-defined.2 It is often classed with Frankland-and Grignard-type reactions which are considered to proceed via homogeneous organometallic routes. The Wurtz reaction is accompanied by the appearance of a blue colour.This was described by Wurtz himself' but, remarkably, has not been characterised spectroscopi- cally until this year. In the Centre for Materials Research at Kent, we are studying a modification of the Wurtz reaction in which dichloroorganosilanes are coupled with alkali metals to form polysilanes. We have obtained strong UV-VIS and EPR spectro- scopic evidence that the blue colours generated in Si-Si and C-C bond-forming polymerisation reactions with Na and K arise from colloidal alkali metal particles formed during the rea~tion.~ Of course, this does not prove that the colloids play an active role in the reaction mechanism. But another important piece of evidence is the unusual polymodal molecular weight distribution of the product polysilanes.If the reaction is homogeneous, it is necessary to invoke at least two competing, non-interactive mechanisms to account for this. We have shown that a single heterogeneous polymerisation mechanism can explain the observed molecular weight distrib~tion.~ We think that alkali metal colloids may prove to be involved in the mechanism of all Wurtz-type coupling reactions, and are currently designing some experiments to test this idea. 1 A. Wurtz, Justus Liebigs Ann. Chem., 1855, 96, 364. 2 J. March, Advanced Organic Chemistry: Reactions, Mechanism and Structure, McGraw-Hill, London; 2nd edn., 1977, pp. 407-412. 3 R. E. Benfield, R. H.Cragg, R. G. Jones and A. C. Swain, Nature (London), 1991, 353, 340.4 R. G. Jones, R. E. Benfield, R. H. Cragg and A. C. Swain, J. Chem. SOC., Chem. Commun., 1992, 112. Prof. Edwards said: This is an interesting observation. As you indicate in your paper* alkali colloids are usually unstable in air. I would draw your attention to your suggestion 104 General Discussion that the ‘anisotropy’ in the proposed conduction electron spin resonance (CESR) from Na colloids arises from particles in an anisotropic environment. I do not quite understand this; I do not believe I know of many examples of anisotropy in CESR (one case may be beryllium metal with a complex, anisotropic Fermi surface). An alternative possibility may be sodium particles of different sizes and having a range of g-values. This observa- tion needs to be carefully investigated. Of course another possibility is radical formation.I wonder if one can unambiguously assign the resonance to colloidal sodium? One possible experiment would be low-temperature CESR, where one might anticipate the onset of asymmetry when the skin depth reduces to dimensions below the particle size (see ref. 2). Another telling experiment might be with K colloids, where large g-shifts would be anticipated. 1 R. E. Benfield, R. H. Cragg, R. G. Jones and A. C. Swain, Nature (London), 1991,353, 340. 2 R. N. Edmonds, M. R. Harrison and P. P. Edwards, Annu. Rep. hog, Chern., 1985, 82, 265. Dr. Benfield responded: We are investigating the anisotropy in the ESR spectra in more detail. All the literature on CESR seems to describe metal particles in symmetric solids or frozen solutions.Exactly the experiments you suggest are already in progress. However, UV-VIS spectroscopy suggests that our Na-colloids are only 2nm in diameter, and this is probably too small to give an asymmetric CESR lineshape even at liquid helium temperatures. Dr. Miessner opened the discussion of Dr. Gallezot’s paper: Activated carbon usually has a pore structure with micropores smaller than 2nm. Do you think that the selectivity you found with your catalysts might be interpreted also in terms of shape selectivity, i.e. that the large molecules of cinnamaldehyde can reach the active hydrogenation sites inside the micropores only by the aldehyde group at the end of the molecule? Dr.Gallezot replied: This is a very interesting point. Indeed we have shown recently’** that high selectivity to unsaturated alcohol can be obtained on zeolite-supported metal catalysts. More specifically, the microporous structure of Y-zeolite imposes molecular constraints on the mobility and orientation of the cinnamaldehyde molecule which can only adsorb end-on, i.e. via the C=O group, on the encaged metal clusters. However, in the case of methyl-crotonaldehyde, the steric constraints are weaker for this less bulky molecule and there is no selectivity improvement. This shows how shape selective effects depend upon the relative size of the pores and of the organic substrate; supramolecular catalytic systems should be tailored for specific reactions.I do not think that these effects play any role in the present active charcoal where the pores are large with respect to the cinnamaldehyde molecule. However active charcoal with narrow pore distribution (molecular-sieve carbons) could certainly be used to induce shape-selectivity effects with suitable organic substrates. 1 P. Gallezot el a/., Catal. Lett., 1990, 5, 169. 2 D. Blackmond et aL, J. Catal., 1991, 131, 401. Prof. Joyner asked: (1) You observe very different activities on the two carbon supports and these also change in a markedly different way with alloying, [see Figs 4(a) and (b) in your paper]. Could this be due, at least in part, to the very different chemical nature of the supports, since active charcoal will have a high coverage of oxygen- containing functional groups, which will not be expected on the graphite support? It is possible that the aldehyde is weakly adsorbed on the activated charcoal and hydro- genated spilt-over hydrogen. (2) You appear to envisage a homogeneous electronic interaction between metal and support.We showed some years ago that, because of the screening properties of General Discussion the metal, such interactions are highly inhomogeneous and short range.' For spherical particles, diameter 1.5 mm, only a small fraction of the surface can be significantly modified electronically. We also showed that a consequence of this interaction is a very marked dependence of catalytic activity on particle radius, between R3 and RS,i.e. a very high structure sensitivity.Have you looked for, or observed, such behaviour? 1 R.W. Joyner, D. K. Saldin, J. B. Pendry and S. R. Tennison, Sur$ Sci., 1984 138, 84. Dr. Gallezot replied: (1) Your question deals with the influences of the supports and of the second metal on the chemoselectivity and you suggest that our data could be interpreted in terms of adsorption of the unsaturated aldehyde molecule on the acidic groups of the active charcoal and hydrogenation by spilt-over hydrogen. For the sake of clarity let us first consider the effect of supports on platinum. As stated in the paper, both supports have been functionalized by NaClO and they both contain acidic groups. The density of acid sites titrated by NaOH is even larger on graphite (6 mmol m-2) than on active charcoal (1.1 mmol m-*) so that the molecule would be adsorbed (according to your suggestion) on both supports thus making no difference.Anyway, it would be difficult to believe that hydrogenation by spilt-over hydrogen could be chemoselective. If so there will be no difference between metals or with particle morphology in contradiction with published results (see ref. 14 and 17 in our paper and ref. 1.). We maintain that the very large difference in selectivity to unsaturated alcohol observed between charcoal and graphite is due to the higher density of states on Pt clusters located on graphite steps owing to the electron-donating interac- tion with the extremities of basal planes. The higher density of states (evidenced by an expansion of the Pt-lattice and by a decrease of the ratio of the adsorption coefficients of toluene and benzene) decreases the probability of activating the C=C bonds thus increasing the selectivity.Concerning the effect of alloying on the selectivity we do not see how the spill over assisted hydrogenation could account for the selectivity data as a function of the alloy composition. In this case, the interpretation of the data should take into account both the support effect, which still exists because the bimetallic clusters are still very small and selectively located on graphite steps, and the effect of ruthenium as discussed in our paper. (2) I am not sure that in the present state of theory(ies), where even the electronic structure of isolated, naked clusters is under debate, one is able to predict with certainty what would be the perturbation induced by foreign atoms from supports or adsorbed species.Meanwhile we have experimental facts showing that the electronic properties of the clusters are modified when they are interacting with graphite steps. Obviously this would not happen if the particles were not very homogeneous in dispersion and location with an average diameter as small as 1.3 nm. 1 Catalysis of Organic Reactions, ed. W. Pascoe, Marcel Dekker, New York, 1992, p. 1. Prof. Rao asked: Is there not alloy formation between Pt and Ru? If so how would the metal distances be affected by alloy formation vis-a-vis the effect of the carbon support? Dr.Gallezot replied: Our results did show that Pt and Ru atoms are associated in the same clusters. Since these bimetallic clusters are as small as Pt clusters and since they are also interacting with graphite steps we may expect an expansion of the metal-metal distances. However no radial electron distribution study has been done on these san;ples. General Discussion Prof. J. N. Murrell (University of Sussex) commented: The last two papers we have discussed suggest that heterogeneous catalysis will tell us little about the individual properties of metal clusters. In one example the catalytic activity appears to be indepen- dent of cluster size, and in the other the observations seem to depend on a subtle interplay of substrate and cluster.Whilst this is undoubtedly an important area of research I do not see where the clear messages about cluster structure and properties can come from. Dr. Gallezot said: I would not dare pronounce a definitive judgement on any area of science from two papers. Anyway it was not our purpose to learn something on cluster structure from catalytic measurements but just the opposite, namely by starting from well characterized clusters try to learn something on the mechanism of hydro- genation leading to chemoselectivity. I would agree that there are only a few cases where catalytic reactions can be used to characterize unambiguously the structure of clusters. A good example is provided by competitive hydrogenation reactions which can probe the electronic structure of metals.' 1 see e.g.P. Gallezot er aL, J. Catal., 1986, 102, 456; P. Gallezot et al., J. Catal., 1990, 123, 341. Prof. Roberts said: I am pleased that John Murrell has given me the opportunity to put forward the view that as far as the reactivity of molecules at metal single crystals is concerned then it is very difficult to predict the chemistry. I will give three or four examples to illustrate this point from our own studies: (a) Dioxygen bond cleavage at a Zn(0001) surface per molecular impact is highly inefficient,' on the other hand in the presence of NH3(g) (which alone is chemically unreactive with both Zn(0001) and a Zn(0001)-0 overlayer surfaces) the efficiency of bond cleavage increases by a factor close to lo3with the generation of amide, hydroxyl and oxide species: (b) bent chemisor- bed imide species NH(a) are formed by a highly selective oxydehydrogenation reaction' when a Cu(ll0) surface is exposed to an ammonia rich dioxygen-ammonia mixture, the NH(a) concentration approaches a monolayer and little oxygen chemisorption occurs; (c) the chemisorption of CO at both atomically clean and an oxidized aluminium surface is weak and of very small coverage; however a coadsorbed mixture of CO(g) and O,(g) leads to the facile formation3 of surface carbonate and carbidic carbon at lower temperatures; (d) oxygen can exhibit a dual role -both activating an adsorbate which otherwise is unreactive and in the same system also providing stability to the adlayer when present at the surface in low concentration^;^ (e)dissociative chemisorption of NO (in contrast to CO) is facile at a Cu( 11 1) surface' and leads to the formation of N20.It is such examples as these that drew our attention to the role of surface transients and precursor states in providing very efficient low energy pathways to products even when the concentrations of these transients are present at immeasurably low concentra- tion~.~'~Such species are likely to be relevant to chemistry under dynamic conditions with the obvious inference that we have to distinguish between preadsorption and coadsorption. It is the latter that is likely to be most relevant to the molecular events involved in heterogeneously catalysed reactions since they simulate more closely the dynamic conditions.So far we have used the probe-molecule approach to search for evidence but a more direct spectroscopic approach needs to be developed to characterise the transients under reaction conditions. We can learn much here from developments in gas-phase reaction kinetics but with the added severe restriction imposed by the substrate surface, that the concentration of intermediates present may possibly be no more than 10" cm-'. 1 A. F. Carley, M. W. Roberts and Song Yan, J. Chem. SOC.,Chem. Commun., 1988, 267; Catal. Lett., 1988, 1, 265; J. Chem. SOC.,Faraday Trans., 1990, 86, 2702. 2 B. A. Afsin, P. R. Davies, A. Pashuski and M. W. Roberts, SurJ Sci. 1991, 259, 2724. 3 A. F. Carley and M. W. Roberts, J. Chem. SOC.,Chem. Commun. 1987, 355. General Discussion 4 P. G. Blake, A. F. Carley, V. Di Castro and M. W. Roberts, .I.Chem. SOC.,Faraday Trans. I, 1987,82,723. 5 D. W. Johnson, M. H. Matloob and M. W. Roberts, J. Chem. SOC.,Faraday Trans. I, 1979,75,2143. 6 M. W. Roberts, Chem. SOC.Rev., 1989, 18,451. 7 P. G. Blake and M. W. Roberts, Catal. Lett., 1989, 3, 399. Prof. Matijevi6 said: There are indeed cases where homogeneous us. heterogeneous catalysis is being disputed. An example is the Pd/Sn chloride catalyst for electroless plating. The controversy as to whether the catalytic effect was due to complex solutes or finely dispersed colloidal particles was resolved by documenting that in all cases (regardless of the method of preparation), the latter were responsible for the plating process.' The task was particularly difficult in the cited example, because the liquid is black and dense. Thus, optical techniques are useless and the separation of fine particles by centrifugation is rather difficult and time consuming. 1 E. Matijevic, A. N. Poskanzer and P. Zuman, Plating, 1975 62, 958.

ISSN:1359-6640    

DOI:10.1039/FD9919200079

出版商:RSC    

年代:1991

数据来源: RSC

摘要:

Faraday Discuss., 1991,92, 109-119 Structural and Electronic Properties of Finely-divided Supported Pt-group Metals and Bimetals Dominique Richard,? John W. Couves and John M. Thomas l%e Davy Faraday Research Laboratory, The Royal Institution, 21 Albemarle Street, London WlX 4BS, UK Finely dispersed platinum and platinum-ruthenium particles, characterized initially by transmission electron microscopy, are shown by X-ray absorption and X-ray scattering techniques to retain their face centred cubic structure. The electron density of these particles of colloidal dimension, judging from white-line intensities, seems to be enhanced by the polymeric macro-ligand to which they are bound. Colloidal dispersions of noble metals have been the object of many studies ever since they were first investigated in the middle of the last century'-3 and they continue to be of great fundamental and practical intere~t.~ Their high degree of dispersion is an obvious advantage in heterogeneous catalysis, but it is not yet clear to what extent the electronic properties of such metals and bimetals may be both reliably measured and modified in a controllable manner.Much scope exists for adjusting' their properties by the use of so-called 'protective' polymer matrices6 or by using an appropriate electron-accepting or-donating support. Such stabilized and adjusted metallic particles of colloidal dimensions have already been shown to exhibit unusual catalytic behaviour in liquid-phase hydrogenations6" and it is well known that the bimetallic particles of Pt-Re and Pt-Ir are powerful catalysts for the reforming of hydrocarbons.X-ray absorption studies by Sinfelt et aL,* have been particularly useful in clarifying the properties of these catalysts and similar structural studies have been made of the Pd-Au system.9 In this paper we discuss the preparation and the detailed characterization of both Pt and Pt-Ru colloidal dispersions. The catalytic interest of the Pt-Ru system being already known for the selective synthesis of isoalkanes," the selective hydrogenation of a$-unsaturated aldehydes" or for use in fuel cell electrodes.12 Previous studies of the structure of R and Pt-Rucolloids employing X-ray absorption (EXAFS and XANES) and scattering techniques have been reported by Duff et aL13 Experimental Preparation of Colloids Colloidal suspensions of Pt were prepared according to the method of Hirai,3 modified by Duff et al.:13 this entails14 reduction of H2PtC16 by methanol in the presence of a soluble polymer support.Typically 0.5 mol dm-3 of H2PtC16 (from Johnson Matthey Chemicals plc) and 3g of polyvinylpyrrolidone (PVP, weight-averaged molecular weight 10 000 from Aldrich) were used, dissolved respectively in 25 cm3 of doubly distilled water and 25 cm3 of methanol (HPLC grade from BDH). t Permanent address; IRC-CNRS, 2, Avenue A. Einstein, F-69626, Villeurbanne, France. 109 110 Colloids of Pt-group Metals and Bimetals Table 1 samplereference Pt (atom %) Pt (wt. YO) polymer medium Qglllcose(mmol) PVP 10 000 MeOH/H20 PVP 10 000 MeOH/H20 PVP 44000 MeOH/H20 PVA MeOH/H,O PVP 10 000 propan-2-01 83 90 PVP 10000 MeOH/H20 0.5 63 77 PVP 10 000 MeOH/H20 1.25 51 67 PVP 10 000 MeOH/H20 1.75 43 59 PVP 10 000 MeOH/H20 1.75 35 50 PVP 10000 MeOH/H20 2.5 13 22 PVP 10000 MeOH/H20 3.75 PVP 10000 MeOH/H20 0.5 PVP 10 000 MeOH/ H20 5.0 Other polymer supports, PVP 44 000 (molecular weight) and PVA (polyvinylalcohol) were also used as was propan-2-01 as the reducing agent in place of methanol.The conditions of preparation of the different samples reported herein are summarized in Table 1. Preparation of Ru colloids (using RuC13 3H20 as starting material) was attempted using the same procedure. Glucose (from Sigma) was added to the mixture with a molar ratio Ru/glucose in the range of 1to 10 (Table l), in an attempt to increase the reducing power of the solution, since, under these conditions methanol is not a strong enough agent to ensure the reduction of Ru sa1ts.l' Bimetallic colloids were prepared in the same way with various proportions of Pt and Ru there being a total amount of 0.5 mmol of metal and an appropriate quantity of glucose to keep the Ru/glucose molar ratio equal to 10.X-Ray Absorption Spectroscopy X-Ray absorption spectra around the Pt L3 and L2-edge (11551 and 13260 eV respectively) were recorded on station 7.1 at the SERC Daresbury Laboratory Synchrotron Radiation Source, using a Si (1 11) double crystal monochromator. The EXAFS was recorded in equal k-steps in the range k =2-16 A-'.Measurements were made in the transmission mode on concentrated colloidal samples using a specially constructed variable pathlength cell (to allow optimization of the absorption spectra), with thin mylar windows (50 microns). Ru K-edge (221 17 eV) X-ray absorption spectra were recorded on station 9.2 at the SERC Daresbury Laboratory utilizin the Si (220) monochromator, the EXAFS data being recorded in the range k =2-16 1-l. EXAFS data reduction (energy calibration, pre-edge/ post-edge background subtrac- tion and edge-step normalization) were performed using the EXCALIB and EXBROOK suite of programmes. Curve-fitting of the k3-weighted data of the Pt-L3 and Ru-K edge EXAFS was performed using the EXCURV901' program using the curved wave approximation.The quantitative analysis of the Pt L3 and L2-edge X-ray absorption threshold resonances was performed using the approach suggested by Mansour et all6in which the edge position is determined by the inflection point of the edge step. The pre-edge background is subtracted from the whole of the data, and the data is then normalized to the edge step. The L2-edge data is then adjusted so that the EXAFS oscillations D. Richard, J. W. Couves and J. M. Thomas % 10 nW 0246810 particle size/nm Fig. 1 (a)TEM micrograph of Pt colloids (PVP 10000, MeOH); (b) histogram of particle size distribution observed on the L3 and L,-edges are coincident. The areas of the threshold resonances were then determined for both edges in the range -10 to 15 eV (0 eV being the edge position).X-Ray Scattering The X-ray scattering data were collected on station 9.1 of the SRS facility at Daresbury Laboratory. Data were collected in a reflexion mode (8-28) ranged from 1 to 130" (28) with a step of 0.1", using a Si(ll1) double crystal monochromator and a A =0.5551 A. The radial electron distribution (RED) was calculated from Fourier analysis of X-ray scattering data as described previously. l7 Transmission Electron Microscopy Specimens for electron microscopy were prepared by allowing a drop of the diluted colloidal suspension, ultrasonically dispersed, to dry on an amorphous carbon film supported on a copper grid. TEM examinations were performed using either a JEOL 1200FX microscope (operated at 200 kV) or a JEOL lOOCX (100 kV).Analytical Electron Microscopy The composition of the metal particles was determined by energy dispersive X-ray emission (EDX) with a SiLi diode attached to a field-emission gun, scanning transmission electron microscope (FEG-STEM) VG-HB501. The spatial resolution of analysis is close to 1.5 nm so that the composition of individual particles could be measured. Results and Discussion Preparation Fig. 1-3 show the micrographs and size distribution of different samples enumerated in Table 1. Sample Pt(1) presents particle size in the 2-4.5 nm range with an average size of 3.4 nm. Sample Pt(2) presents slightly smaller sizes (average 3.1 nm), whereas sample Pt(4) presents not only small particles (ca.3 nm) but also some which are much larger thus yielding an average of ca.4 nm. We note that, using propan-2-01 as a reducing agent instead of MeOH, leads to larger particle size, a situation that is different from what has been observed previously by Hirai6 in the case of rhodium colloids. 112 Colloids of Pt-group Metals and Bimetals (a) 30 20 Yo 10 n -0246810 particle size/nm Fig. 2 (a)TEM micrograph of Pt colloids (PVP 44 000, MeOH); (b)histogram of particle size distribution 204 -0246810 particle size/nm Fig. 3 (a)TEM micrograph of Pt colloids (PVP 10 000, propan-2-01); (b)histogram of particle size distribution Electron micrographs of the bimetallic colloids indicate that the latter are generally not as homogeneous in size as the monometallic ones.There are particles in the 2-4 nm range but also some larger ones with a size up to 10nm. Fig.4 shows one of these samples, Pt-Ru(4) K(43 %Pt, 57 %Ru in atom). Platinum Colloids For the colloid of pure Pt, both RED and EXAFS analysis indicated the presence of particles with the f.c.c. structure essentially indistinguishable from that of the bulk metal. Fig. 5 gives the RED of a monometallic Pt colloid which is compared to a theoretically derived distribution of spherical particles containing 767 atoms (3.0nm size). Some of the Pt-Pt distances are elongated with respect to the distances in bulk platinum, and some are the same or even slightly shorter.This lattice distortion can be attributed to a ligand effect arising from the polymer to which the particles are bonded. A ligand effect, causing lattice expansion has already been observed in the case of large gold D. Richard, J. W. Couves and J. M. Thomas Fig. 4 A typical electron (transmission) micrograph of Pt-Ru colloids 200 1 I 7.36 150 (a) 4.80 hI A 100-2.81 $ 50-& .-P 0-Y (d -2 -50 -loo! . ' . ' . ' . ' " ' ' . "V . ' . 1 0 12 3 4 5 6 7 8 9 10 r/ A 1000 -(b) 7.34 I 800 A E s 600cr k 2 400.-U-2 200 0 0 12 3 4 5 6 7 8 9 10 r/ A Fig. 5 (a)Radial electron distribution of Pt colloid [Pt( 1b)]. (b) Pair distribution in a model Pt particle (spherical particle, 767 atoms, diameter 30.5 A) Colloids of Pt-group Metals and Bimetals + -20..4.5.-..1.a/4.0.-1.6 11-3.5-1.4-3.0-1.2 .-2.5-1.o.-2.0.-0.8 ,_ .-1.5 0.6.-1.o-0.4.-0.5-0.0 r Fig. 6 EXAFS (k3-weighted) (upper) and magnitude of Fourier transform (lower) for (a) Pt 4 mm foil (experimental solid line, theory dashed line); (b)Pt colloid, Pt(1b) (Experimental solid line, theory dashed line) Table 2 Pt-foil 2.77 12 0.01 3.92 6 0.014 Pt(la) Pt(W Pt(3)W2) 2.75 2.75 2.75 2.74 11.2 11.1 11.2 11.4 0.015 0.015 0.015 0.017 3.87 3.90 3.87 3.86 3.45 3.73 3.98 3.71 0.019 0.018 0.017 0.019 W4) 2.74 11.4 0.018 3.86 3.78 0.020 clusters by Fairbanks et aL'* and lattice distortion caused by a strong interaction of raft-like platinum particles with a graphite support has been observed by Gallezot et al.19 EXAFS data of all the monometallic pt colloids are consistent with the particles retaining the f.c.c.structure (Fig. 6 and Table 2). The first shell distances of all the samples show a slight lattice contraction, compared to the bulk pt metal, of 0.6 to 0.9 %. This is in line with theoretical calculations of such lattice contractions for particles of this size,*' and with other experimental measurements of similar dispersion~.'~ Owing to the highly correlated nature of the co-ordination number and the Debye-Waller factor, estimated values of the co-ordination numbers (based on average particle size measurements from electron microscopy) were used in the EXAFS curve-fitting pro- cedure.In this way the Debye-Waller factor can be used to investigate the disorder of the colloidal systems. The higher Debye- Waller factors observed for the colloidal particles compared with those of bulk Pt are a result of the large vibrational amplitude in the colloidal particle and arise possibly from the relaxation of co-ordinatively unsatur- ated surface Pt atoms at certain faces of the fine particle. D. Richard, J. W. Couves and J. M. Thomas 1.20--h Ls \ +o v E-1.00--0 10 20 30 40 50 60 70 80 Energy/eV Fig. 7 White line (normalized data) of different samples; (a)Pt(NHJ4C1,, H20; (b)Pt( 1b); (c) Pt-Ru( 1); (d)Pt metal bulk Variations in Debye-Waller factor within the Pt colloidal samples is a result of the degree of polydispersity within the samples.Samples Pt(la), Pt(1b) and Pt(2) have a relatively small degree of polydispersity (k0.8A of the mean value). This produces a Debye-Waller factor of ca. 0.015 A2, which can be rationalised solely by the increase in the vibrational amplitude and relaxation of the surface structure. Colloids Pt(3) and Pt(4) have significantly higher Debye- Waller factors, a consequence of the bimodal distribution of particles as observed by electron microscopy (Fig. 4). L2 and L3 X-ray absorption edges arise from electronic transitions from the 2pIl2 and 2p3/2 core states respectively. The white line (X-ray absorption threshold resonance) observed at these edges is a result of transitions from the core states to unoccupied d312 and dSl2 states for the L2 and L3 edges respectively. Brown et aL21showed that Pt metal has predominately J=5/2 character, which results in the white line on the L3 edge having higher intensity than that of the L,-edge, there have been many reports showing a correlation of white line intensity with number of d-electron~.~~-~~ This qualitative approach (Fig.7) shows that the relative white line intensity (to Pt metal) of the L2 and L3 edges varies with d-electron occupancy. Raw data for samples Pt(1b), Pt-Ru( 1) and well defined standards are shown in Fig. 7a. Note that the relative intensity of the white lines recorded for the monometallic Pt particles falls between that of Pt2+ and Pt', indicating that the Pt colloids have a higher degree of unoccupied d-states compared to the bulk platinum. The extraction of information from white line intensities is not however straightforward due to ambiguities associated with determining the precise area of the white line.Mansour et aL16 suggested a systematic approach to extracting an estimate of the number of unoccupied d-states in terms of the fractional change of d-band occupancy compared to bulk Pt. The method uses differences in areas compared to a standard material, which can be determined to a greater accuracy than the absolute area under the white line. This approach has been utilized to investigate the Pt and bimetallic particles. The values obtained from applying the method of Mansour et al.are shown in Table3. The total number of unoccupied d-states, hT, is considerably larger in the case of the Pt colloid (0.39 hole) than for bulk Pt2*(0.3 hole). Ruthenium Colloids With the samples of pure Ru, there are strong indications that the salt has not been totally reduced to metallic particles during the preparation. No particles were observed Colloids of Pt-group Metals and Bimetals Table 3 sample relative L3area relative L2area hT Pt(bu1k metal) Pto2 1.oo 1.14 1.oo 1.22 0.30" Pt en 1.15 1.28 Pt acac 1.10 1.19 PtW) 1.05 1.16 0.39 " Ref. 21. 30 4.65 20 Ox 105 m go P).-Y 2 -10e -20 r/ nm Fig. 8 Radial electron distribution of 'Ru colloid' [Ru(2)] by TEM, which means that, if they exist, they must be very small (<1nm).RED also reveal that no organized structure is present in the sample at long distances (Fig. 8), which is contrary to what is observed in the case of Ru particles on a carbon support. However, two peaks are observed at 2.7 A and 4.65 A. The second of these is very large which could indicate that even if Ru is not reduced, there is a trend towards the formation of small clusters. Ru K-edge EXAFS shows a weak first backscattering shell at 1.8 A (not corrected for phaseshift) and a second shell at 2.55 A (not corrected for phaseshift) no other shells were observed in the Fourier transform of the 'Ru colloid' (Fig. 9). The amplitude envelope of the k3-weighted data is not consistent with a Ru atom being the backscatterer.However, it suggests that the shell is a mixed one of relatively weak backscatterers. This would be consistent with the Ru3+ being largely unreduced and in the form of the hydrated ion. In an aqueous solution of RuC13 3H20, as well as the fully hydrated ion [Ru(H20)J3+, there will be species such as [RU(H~O)~C~]~+ and [Ru(H20)4C12]+ present. This adds complexity to the fitting of EXAFS oscillations. The long contact may result from a Ru-Ru contact which was observed from the RED measurements and is associated with cluster formation. Platinum-Ruthenium Colloids The RED of a bimetallic colloid appears to be very close to the one for pure Pt. In order to obtain element-specific distance distributions of the bimetallic colloidal particles, an anomalous scattering study was attempted, but this did not give any evidence for D.Richard, J. W. Couves and J. M. Thomas 4 0.7- .CIB 0.6.-5 0.5.- 0.4- 0.3- 0.2- 0.1- 0.0.: t-:: 0 1 2 : 3 : 4 : 5 , 6 , 7 , 8 ’ 9 ’ J 10 r/ 8, Fig. 9 Magnitude of Fourier transform for (a)Ru(1) (solid line); (b)Pt-Ru(3) (dotted line); (c) Pt-Ru( 5) (dashed line) Table 4 sample r,/A N, a,/A2 r2/A N2 %/A Pt-Ru(1) 2.76 11.2 0.012 3.87 3.45 0.012 Pt-Ru(2) 2.75 11.2 0.015 3.91 3.44 0.019 Pt-Ru(2) 2.75 11.2 0.014 3.86 3.50 0.015 Pt-Ru(4) 2.75 11.2 0.014 3.90 3.78 0.016 Pt-Ru(6) 2.75 11.2 0.016 3.92 3.93 0.017 the modification of the structure of the colloidal particles upon addition of Ru to the structure of the Pt colloids.EXAFS analysis of the Pt L,-edge (Table4) indicates that the bimetallic colloids retain a structure consistent with the f.c.c. structure of Pt bulk metal, just as is observed with the monometallic Pt colloids. The lattice contraction found in the bimetallic colloids is similar to that of the monometallic Pt colloids, indicating that they are of similar dimensions. The Debye-Waller factors of the first shell are all slightly smaller than those observed for the various preparation of Pt colloids. This may be an indication of smaller variations of vibrational amplitude compared to bulk Pt or that there is a lower degree of surface relaxation. Ru K-edge EXAFS show a similar low r backscatterer, however, the amplitude of this backscatterer, in the Fourier transform, is significantly reduced (Fig.9), compared to that of the ‘Ru colloid’. This could result either from a reduction of the co-ordination number of this shell or from an increase in the Debye- Waller factor caused by an increase in static disorder within the first backscattering shell. The magnitude of the Fourier transform of the first backscattering shell is also dependent on the amount of Ru in the bimetallic colloid. Simultaneously with the decrease in magnitude of the first backscattering shell, two other shells appear at 2.30 and 2.80A (not corrected for phaseshift). These could result from a variety of possibilities: Ru-Pt contacts; reduced Ru on the surface of the Pt colloidal particles; or from adsorption of low co-ordinated Ru hydrated ions.The presence of Ru should stabilize the surface of the Pt particles of the core, as observed from the decrease in Debye-Waller factor of the first backscattering shell from the Pt L,-edge EXAFS. This suggests that the bimetallic particles are present in a ‘segregated’ form, i.e. an f.c.c. Pt Colloids of Pt-group Metals and Bimetals Table 5 sample relative L3 area relative L2 area hT Pt(bu1k metal) Wb) 1.oo 1.05 1.00 1.16 0.30" 0.39 Pt-RU(1)Pt-RU(3) 1.07 1.05 1.16 1.17 0.38 0.36 Pt-Ru( 4) 1.04 1.15 0.36 Ref. 21. core with an external coating of Ru. This is consistent with the results of analytical microscopy which show a Ru-deficient particle compared to the global composition.It is also consistent with the observation that, under our preparation conditions, the Ru salt alone is not reduced. It appears that the reduction of the Ru salt to yield colloidal metal is facilitated by prior production of minute particles of Pt. The analysis of the L2/L3 white line areas of the Pt-Ru colloids is shown in Table 5. As with the monometallic Pt colloids, the areas are all larger than that of bulk Pt. There is, however, no correlation with Ru content. Applying Mansour et aL's method (Table 5) it can be seen that the presence of Ru on the pt particle decreases the total number of unoccupied d-states. Also, there seems to be a relationship between the Ru content of a colloidal particle and the number of unoccupied d-states.This trend was not observed in the simple analysis of the absolute white line areas, perhaps owing to the inherent errors of the method. No error estimates have been made on the values of unoccupied d-states determined for the Pt and Pt-Ru colloids, but it is likely that these will be high. However using this systematic approach we were able to show that adsorption of Ru onto the surface of the Pt colloidal particles decreases the average number of unoccupied d-states. This is interpreted as charge transfer from the Ru to the Pt. Conclusion TEM, X-ray absorption spectroscopy and X-ray scattering has been used to characterize colloidal Pt particles produced by reduction with methanol. It has been shown that the particles are of 2-4nm in diameter and that they possess the f.c.c.structure. EXAFS analysis shows that there is a slight contraction of the first shell distance, with an accompanying increase in the Debye-Waller factor. This is a result of relaxation of atoms at the surface of the Pt particle. RED measurements also indicate a structural lattice distortion. The electronic properties of these particles are considerably modified compared to bulk Pt. The density of unoccupied d-states, determined from analysis of the X-ray absorption threshold resonance, is seen to be significantly lower compared with the bulk. This arises from the small particle size and interactions with the polymer support. Bimetallic colloidal particles consist of a R f.c.c. core with an adsorbed Ru surface layer.The effect of this layer of Ru is to reduce the density of unoccupied d-states of the Pt core. Modification of the catalytic selectivity of these platinum colloids2* is a result of these changes in the electronic properties. The authors thank the SERC for a grant for the use of the SRS facility at the Daresbury Laboratory. D.R. also thanks the CNRS for allowing his stay at the DFRL and NATO for providing a complementary grant (1990-86C89FR). The authors are also grateful to Dr G. Sankar for many helpful discussions of the EXAFS data. D. Richard, .I.W. Couves and J. M. Thomas References 1 F. Selmi, Nuov. Aun. Sci. Natur. di. Balogne (Z), 1845, 4, 146. 2 M. Faraday, Philos. Trans. R. SOC.,1857, 147, 145. 3 T.H. Graham, Philos. Trans. R. SOC.,1861, 151, 183. 4 J. M. Thomas, Pure Appl. Chem., 1988,60, 1517. 5 J. Turkevich, R. S. Miner Jr and L. Babenkova, J. Phys. Chern., 1986,90, 4765. 6 H. Hirai, J. Makromol. Sci.-Chem., serie A, 1979, 13, 633. 7 A. Harriman, G. R. Millward, P. Neta, M. C. Richoux and J. M. Thomas, J. Phys. Chem., 1988,92,1286. 8 J. H. Sinfelt, Bimetallic Catalysts, J. Wiley, 1983. 9 J. B. Michel and J. A. Schwartz, in Preparation of Catalysts ZV, ed. G. Poncelet et al., Elsevier, Amsterdam, 1987, pp. 669-687. 10 T. Tatsumi, Y. G. Shul, T. Sugiva and H. Tomimoya, Applied Catal., 1986, 21, 119. 11 A. Giroir-Fendler, D. Richard and P. Gallezot, Faraday Discuss. Chem. SOC.,1991,92, 111. 12 A Hammett, B. J. Kennedy and F. E. Wager, J.Catal, 1990, 124,30. 13 D. G. Duff, P. P. Edwards, J. Evans, J. T. Gauntlett, D. A. Jefferson, B. F. G. Johnson, A. 1. Kirkland and D. J. Smith, Angew. Chem. In?. Edn. Engl., 1989, 28, 590. 14 P. R. van Rheenen, M. J. McKelvy, W. S. Glausinger, J. Solid State Chem., 1987, 67, 151. 15 N. Binstead, S. J. Gurman and I. J. Ross, J. Phys. C, 1984, 17, 143; SERC Daresbury Laboratory Update 1990. 16 A. N. Mansour, J. W. Cook and D. E. Sayers, J. Phys. Chem., 1984,88, 2330. 17 G. Bergeret and P. Gallezot, Proc. 8th In?. Congr. on Catalysis Berlin, Verlag Chemie, Weinham 1984, p. 659. 18 M. C. Fairbanks, R. E. Benfield, R. J. Newport and G. Schmid, Solid State Commun., 1990, 73, 431. 19 P. Gallezot, D. Richard and G. Bergeret in Novel Materials in Heterogeneous Catalysis, ed.R. T. K. Baker and L. L. Murrell, A.C.S. Symposium Series, 437, 1990, pp. 150-159. 20 S. N. Khanna, J. P. Bucher, J. Buttet and F. Cyrot-Lackmann, Surf: Sci., 1983, 127, 165. 21 M. Brown, R. E. Peierls and E. A. Stern, Phys. Rev. B, 1877, 15, 738. 22 F. W. Lytle, P. S. Wei, R. B. Greegor, G. H. Via and J. H. Sinfelt, J. Chem. Phys., 1979, 70, 4849. 23 P. Gallezot, R. Weber, R. A. Della Betta and M. Boudart, 2. Nuturforsch. Teil A, 34, 40. 24 M. G. Samant and M. Boudart, J. Phys. Chem., 1991,95,4070. 25 F. W. Lytle, J. Catal., 1976, 43, 376. 26 B. Moraweck, P. Bondot, D. Goupil, P. Fouilloux and A. J. Renouprez, J. Phys. (Paris) C8, 1986,47, 297. 27 C. N. R. Rao, J. M. Thomas, B. G. Williams and T. G. Sparrow, J. Phys. Chem., 1984,88, 5769. 28 D. Richard, J. W. Couves and J. M. Thomas, in preparation. Paper 11032576; Received 1st July, 1991

ISSN:1359-6640    

DOI:10.1039/FD9919200109

出版商:RSC    

年代:1991

数据来源: RSC

摘要:

Faraday Discuss., 1991, 92,121-128 Small Metallic Particles studied by Optical and Electron-optical Spectroscopy Hellmart Seiler, Ulrich Haas" and Bernd Ocker Institut fur Physik, Universitat Hohenheim, Garbenstrape 30, D-7000Stuttgart 70, Germany Karl-Heinz Kortje Institut fur Zoologie, Universitat Hohenheim, Garbenstrape 30, D-7000 Stuttgart 70, Germany Electron energy loss spectroscopy (EELS) in a transmission electron micro- scope (TEM) and photoacoustic spectroscopy (PAS) has been used to study the electronic structure of small Ag, In and Au particles, deposited on thin carbon foils, quartz supports or as self-supporting layers, respectively. In the low energy-loss region the EELS spectra of Ag particles with diameters of ca. 50 nm are different from the spectra of the homogeneous Ag layers, mainly owing to the influence of the carbon foil.By decreasing the packing density of the Ag particles, a shift of the plasma losses to lower energies is observed in the energy region 10eV. The optical behaviour of homogeneous metallic and small particle layers on thin quartz supports was studied by PAS in the ultraviolet/visible (UV/VIS) spectral region. The homogeneous layers, e.g. of Ag, show, in the UV range, the well known change of reflectance due to an interband transition of bare silver. For Ag particle layers surface plasma resonances are excited. With increasing pack- ing density of the particles the absorption peak observed is red-shifted and becomes broader, whereas the results of EELS in the low energy-loss region revealed a shift to higher energies.To obtain additional information investigations of In and Au particles were performed. As early as 1857 Michael Faraday produced gold in coiioidal form. It is only quite recently that aggregates of atoms or molecuies to small particles, microcrystals or clusters have become of interest in fundamental and applied research. Intensive experimental and theoretical work from many branches of science and technology is concerned with free clusters as well as with small particles deposited on various support materials or embedded in diverse matrices.'-5 Many physical properties of clusters prove to be strongly size dependent and they seem to bridge the gap from atoms and small molecules to bulk material.Furthermore, clusters are important in new areas, e.g. in the develop- ment of physics of microstructures or in the field of heterogeneous catalysis. Therefore, a special interest exists in the investigation of metal particles. To understand how the electronic states are dependent on the amount of aggregation of metal atoms to small particles, electron-optical and optical methods can be The size and shape of small particles have been studied by high-resolution TEM, scanning electron microscopy (SEM), and scanning transmission electron microscopy (STEM):-'5 whereas the electronic properties were studied by EELS and image energy filtering in a STEM. Most investigations of small particles by EELS have been devoted to volume and surface plasmon excitations.' 5-22 The plasmon energy is proportional to 121 Op tica 1 and Electron -optical Spectroscopy Fig.1 SEM micrograph of Ag particles the square root of the free electron density, therefore, a variation of plasmon energy can be attributed to a variation of free electron density of the metallic aggregates. In studying plasmon excitation of small particles by means of EELS, the effect of the supporting-film substrate has to be taken into ac~ount.'~-~~ Another characteristic property of a metal is its optical response, Le. the interaction with light. Most metals show a strong absorption in the UV/VIS caused by interband transitions and a strong absorption in the far infrared. The reason for the latter absorption is that these very low-energy photons are able to excite electrons from the continuum of states just below the Fermi energy to states just above Fermi energy.The influence of particle size on many of the optical properties can be described by the theory of Mie.24 Investigations in the UV and IR region of the spectrum have been carried out by optical, photothermal and photoacoustic spectros~opy.'~~~~~~'~~~-~~For small metallic particles the excitation of surface plasmons could be ob.~erved.~,~',~~-*~ Experimental We investigated homogeneous layers and small particle layers of Ag, In and Au. Ag and Au were chosen because of their chemical stability and the possibility of obtaining particles nearly spherical in shape. Indium was taken as an example of a simple metal. All thin homogeneous layers and the In particle layers were prepared by evaporation of the bare metal in high vacuum.For EELS investigations Ag layers were self- supporting, obtained by evaporating the metal on thin glass slides, and subsequently H.Seiler et al. 26 I 1 0 30 60 energy loss/eV I I I 1 0 30 60 energy loss/eV Fig. 2 (a) EEL spectrum of self-supporting Ag foil, thickness 70nm. (b)’EEL spectra of Ag particles. I: High packing density, particle size 50 nm median diameter. 11: Low packing density, particle size 20 nm median diameter picking up on electron microscopical grids after floating the layers from the glass support with distilled water. The homogeneous layers of In and Au as well as the particle layers of Ag, In and Au were deposited on thin carbon foils (15 nm thick).The supports used for PAS studies were thin quartz plates. The gas aggregation technique was employed to produce cluster layers of Ag and Au. The metals were evaporated in an atmosphere of N2 at a pressure of ca. 100 Pa. The amount of deposited material was monitored by a quartz oscillator mounted in a corresponding position to the supports. The investigations by EELS were performed in transmission in a ZEISS EM 902 TEM with an integrated imaging electron energy spectrometer. The EM 902 TEM was operated at 80 kV with 1eV energy resolution, and an aperture of 5.7 mrad. The magnification was 20 000, giving information about a definite area of the specimen.Material analysis of very thin foils (d <50 nm) in high vacuum is possible by measuring the edge structure. The optical absorption behaviour was studied in the UV/VIS range from 200 nm to cu. 800nm with an EG & G, Princeton Applied Research Model 6001 photoacoustic spectrometer, as described elsewhere.29 The samples were investigated at normal pressure and room temperature (293 K). Optical and Electron-optical Spectroscopy Fig. 3 TEM micrograph of In particles Results and Discussion EELS Silver The Ag layers produced in the inert-gas atmosphere are close-packed smoke particles that look as dark as soot, whereas the silver layers evaporated in high vacuum burnish brightly. The observation of cluster layers with SEM shows small silver particles with diameters ca.50 nm and less (Fig. 1). The electron diffraction patterns of homogeneous Ag layers and of cluster layers are similar. In order to avoid interference in the spectra of Ag and carbon, we investigated self-supporting Ag layers. Fig. 2(a) shows in the low loss region a main maximum at 26 eV and besides the first sharp loss at 4 eV there are broader losses at 8.5, 19 and 34 eV. According to Raether6 the interband transition of Ag is expected at 3.9 eV, the bulk plasmon loss at 3.7 eV, and the surface plasmon loss at 3.6 eV, which cannot be resolved here. Fig. 2(b) shows a comparison between the spectra of Ag particles on carbon foil with high (I) and low (11) packing density. In the low loss region, spectrum I shows the same loss energies as the self-supporting Ag foil [Fig.2(a)]. In spectrum 11,compared to I, there seems to be a shift of the 8.5 and 4 eV peaks to lower energies, 5.5 and 3.0 eV, respectively. At 25 eV one can clearly see the main peak of carbon. Ouyang et aL3' investigated the surface plasmon loss of Ag at ca. 3.5 eV. For particles with diameters 10-20nm they found a decrease of the loss energy with decreasing particle size. A dependence of the energy of low-energy loss spectra can also be attributed to a chemical H.Seiler et al. 0 30 60 energy losslev 1 11.5 0 30 60 energy loss/eV Fig. 4 (a) EEL spectrum of In layer, thickness 90 nm, on carbon foil. (b) EEL spectra of In particles. I: High packing density, particle size 200 nm median diameter.11: Low packing density, particle size 40 nm median diameter change of surface or the strong dependence of surface plasmon excitation on the local geometry of the particles.” Indium Owing to the ‘liquid-like’ condensation of In vapour on carbon, we obtained discon- tinuous island films on the substrate. We could detect separated single particles on the films (Fig. 3). The particle size could be controlled by the amount of condensed indium on the carbon film. A thin continuous In layer could be obtained by cooling the carbon substrate to liquid-nitrogen temperature. Fig. 4(a) shows the EEL spectrum of an In foil. In the low loss region there is a multiple plasmon generation with energies ca.11.5 eV. Fig. 4( b) shows a comparison between spectra of particles with high (I) and low (11) packing density. In the low energy loss region both spectra have at 23 eV an overlap with the main peak of carbon and show at 11.5 eV the excitation of the bulk plasmon of In. In contrast to the homogeneous film, we can detect a new peak at 8 eV (I, high packing density) which shifts down to 4.5 eV (11, low packing density). Optical and Electron-optical Spectroscopy I 1 0 45 90 energy loss/eV energy loss/eV Fig. 5 (a) EEL spectrum of Au layer, thickness 20 nm, on carbon foil. (b) EEL spectra of gold particles. I: High packing density, particle size 1 nm median diameter. 11: Low packing density, particle size 1 nm median diameter Gold Homogeneous and particle layers of Au look like the corresponding Ag layers.Whereas gold layers, evaporated in high vacuum, are bright reflecting, the investigated layers produced in the inert-gas atmosphere are very close-packed particles that look as dark as soot, with diameters of ca. 1 nm. The homogeneous Au foil [Fig. 5(a)] shows in the low loss region a main maximum at 24.5 eV and shoulders at 18 and 33 eV. The weak absorption shoulder at 54 eV corresponds to the 02, 3 edge and the shoulder at 6 eV may be caused by surface plasmon excitation. Fig. 5(b) shows Au particles on carbon foil with different packing densities. For high packing density of the particles [Fig. 5(6), I] the EEL spectrum shows in the low loss region the same loss energies as the homogeneous Au foil, but with higher resolution of the losses at 6, 18, 33 and 54eV.By decreasing the packing density [Fig. 5(b),111 these losses are again as weak as for the homogeneous foil. The shift of the main peak to lower energies is caused by the overlap with the main peak of carbon at 23 eV. H.Seiler et al. 4> E : .-M v) d P, 540 ~-~/;m-250 300 350 400 450 4.5 4.0 3.5 3.0 E/eV > E : EM ._v) d P, 4 4.5 4.0 3.5 3.0 E/eV Fig. 6 Photoacoustic spectra of Ag particles. (a)High packing density; (b) low packing density PAS The PA signal of a substance is in principle determined by the transfer of absorbed light energy to heat by radiationless processes. Spectra of homogeneous Ag, In and Au layers show in the UV/VIS the change of absorptivity due to interband transitions.The absorption behaviour of Ag cluster layers differ in this spectral region, however, and change obviously for decreasing packing density of the small silver particles. Fig. 6 shows the PA signal of Ag cluster layers with different packing densities [for Fig. 6(a) twice that for Fig. 6(b)].The spectra obtained for lower packing densities are similar to the results for a statistically roughened bare surface of Ag and for small Ag particles embedded in a dielectric matrix with low packing density.'"' This absorption behaviour is caused by the lowest surface plasmon mode of spherical Ag particles. A shift to higher energies for decreasing packing densities of the particles is observed, which is in accordance with results of optical absorption spectra of Ag microcrystals.8 Relatively weak and broad absorption signals are obtained for In and Au particle layers.Probably, the statistically roughness of these particle layers is not sufficient to induce surface plasmon resonances by the optical excitation method employed. For In, a further contribution may arise from the strong contamination of the sample surface by oxygen and carbon, as proved through an investigation of the In layers by Auger electron spectroscopy.3' The financial support of the Deutsche Forschungsgemeinschaft (B.O.) is gratefully acknowledged. References 1 W. P. Halperin, Rev. Mod. Phys., 1986, 58, 533. 2 Elemental and Molecular Clusters, ed.G. Benedek, T. P. Martin and G. Pacchioni, Springer Series Mater. Sci. 6, Springer-Verlag, Berlin, 1987. 3 M. A. Duncan and D. H. Rouvray, Spektr. Wiss. 1990, 2, 62. Optical and Electron-optical Spectroscopy 4 R. S. Berry, Spektr. Wiss. 1990, 10, 72. K. H. Meiwes-Broer and H. 0. Lutz, Phys. Mutter, 1991,47, 283. 6 H. Raether, Surface Plasmons on Smooth and Rough Surfaces and on Gratings, Springer Tracts Mod. Phys. 11 1, Springer-Verlag, Berlin, 1988. 7 U.Kreibig and L. Genzel, Sut$ Sci. 1985, 156,678. 8 H. Abe, W. Schulze and B. Tesche, Chem. Phys., 1980,47,95. 9 L. D. Marks, Ultramicroscopy, 1985, 18, 445. A. Howie and R. H. Milne, Ultramicroscopy, 1985, 18, 427. 11 C. Colliex, J. L. Maurice and D.Ugarte, Ultramicroscopy, 1989, 29, 31. 12 J. Liu and J. M. Cowley, Scanning Microsc., 1988, 2, 65. 13 Y. Saito, A. Takemoto, N. Tanaka and K. Mihama, Roc. XIth Znt. Congr. Electron. Microsc. Z, 1986,169. 14 H. Seiler, U. Haas and K-H.Kortje, Proc. XIZth Int. Congr. Electron. Microsc. I, 1990, 250. D. Ugarte and C. Colliex, 2.Phys., 1989, 12, 333 16 M. De Crescenzi, P. Picozzi, S. Santucci, C. Battistoni and G. Mattogno, Solid State Commun., 1984, 51, 811. 17 P. E. Batson, Surf: Sci., 1985, 156, 720. 18 M. Achiche, C. Colliex, H. Kohl, A. Nourtier and P. Trebbia, Ultramicroscopy, 1986, 20, 99. 19 2. L. Wang and J. M. Cowley, Ultramicroscopy, 1987, 21, 77; 347. D. Ugarte, Proc. XZZth Znt. Congr. Electron Microsc. I, 1990, 42. 21 H. Seiler, U.Haas, K-H. Kortje and B. Ocker, Microscopy, Microanalysis, Microstnrct., 1991, 2, 191. 22 A. Howie and R. H. Milne, J. Microsc., 1984, 136, 279. 23 F. Ouyang and M. Isaacson, Ultramicroscopy, 1989, 31, 345. 24 G. Mie, Ann. Phys., 1908, 25, 377. M. Bauer and H.D. Breuer, in Photoacoustic and Photothermal Phenomena, ed. P. Hess and J. Pelzl, Springer Series in Optical Sciences, Springer-Verlag, Berlin, 1988, vol. 58, pp. 214-216. 26 T. Inagaki, J. P. Gourdonnet and E. T. Arakawa, in Photoacoustic and Photothermal Phenomena, ed. P. Hess and J. Pelzl, Springer Series in Optical Sciences Springer-Verlag, Berlin, 1988, vol. 58, pp. 156-163. 27 U. Kreibig, B. Schmitz, and H. D. Breuer, in Photoacoustic and Photothermal Phenomena, ed. P. Hess and J. Pelzl, Springer Series in Optical Sciences, Springer-Verlag, Berlin, 1988, vol. 58, pp. 217-219. 28 H. Abe, K-P. Charli, B. Tesche and W. Schulze, Chem. Phys., 1982, 68, 137. 29 U. Haas and H. Seiler, 2.Naturforsch., Teil A, 1984, 39, 1242. F. Ouyang, P. E. Batson and M. Isaacson, Roc. XZZth Znt. Congr. Electron, Microsc. I, 1990, 274. 31 H. E. Bauer, personal communication. Paper 1/02591K, Received 28th May, 1991

ISSN:1359-6640    

DOI:10.1039/FD9919200121

出版商:RSC    

年代:1991

数据来源: RSC