摘要:

G. H. TWIGG I59 THE EFFECT OF CRYSTAL PARAMETER ON HYDROGENATION AND DEHYDROGENATION BY 0. BEECK AND A. W. RITCHIE Received 7th February, 1950 Through measurement of the surface of oriented and unoriented evaporated porous nickel films by the B.E.T. method, using neon, krypton, methane and butane, and by measuring the number of crystallographic sites through the adsorption of hydrogen at liquid nitrogen temperature, direct prooi has been obtained that (110) oriented nickel films do also expose (110) planes t o the gas phase. A study of the hydrogenation of benzene over oriented and unoriented nickel and iron films showed no difference in rate in contrast to the previously reported 5-fold rate for the hydrogenation of ethylene over (110) oriented nickel films in comparison to unoriented films.It is concluded that the observed slow hydrogenation, which kinetic considerations show, proceeds through the adsorbed state of benzene, is not influenced by the crystal geometry of the substrate. In contrast, the dehydrogenation of cyclohexane over unoriented platinum, was found t o be 10 times faster than over partially ( I 10) oriented platinum films, indicating that the geometrical factor is of great importance in this reaction, the kinetics of which is in agreement with the assumption that cyclohexane loses 6 hydrogen atoms simultaneously when colliding with the catalyst surface. Following the hypothesis of Balandin, this reaction should be facilitated by (111) faces, which should be more abundant in unoriented than in (110) oriented films. I.Experimental Proof for Surface Exposure of (110) Planes in Gas-induced Orientation of Nickel Films.-In an earlier publication Beeck, Smith and Wheeler have shown that by controlling the pressure of an inert gas during evaporation of metals such as nickel, iron and others, unoriented and oriented porous metal films could be produced a t will. Completely oriented nickel films were obtained with an inert gas pressure G f about I mm., the (110) plane, the least dense of the planes, lying parallel to the backing and the two remaining planes showing random distribution. Films evaporated in high vacuum were found to be unoriented. The oriented nickel films were found to have twicz the surface for hydrogen adsorption per gram of metal and were about ten times as active as catalysts for the ethylene hydrogenation than the unoriented films.The oriented films therefore had five times the intrinsic activity of unoriented films. Oriented and unoriented films were found to have approximately the same thickness fGr the same weight, that is the same density. This demands that the oriented films have smaller pores (or a pore size distribution favouring smaller pores) since the internal surface for hydrogen adsorption per gram of metal is twice that of unoriented films. Beeck, Smith and Wheeler, Proc. Roy. SOC. A , 1940, 177, 62.I 60 THE EFFECT OF CRYSTAL PARAMETER Several arguments were presented 1 (which will not be repeated here) why the crystallites of gas evaporated nickel films showing ( I 10) orienta- tion parallel t o the backing should also actually expose ( I 10) planes t o the gas phase.No direct experimental proof was offered at t h a t time, although all indirect proof, especially the catalytic experiments, pointed t o the correctness of this deduction, particularly the fact that films evaporated in an inert gas pressure of 10 mm. were unoriented but had the intrinsic activity Gf oriented films. This was explained by the assumption that under conditions of high gas pressure small crystallites of nickel would form in the gas phase with (110) planes exposed and t h a t these crystallites would then be deposited randomly t o form the un- oriented film of not only the intrinsic catalytic activity of oriented films but also of an especially large specific surface. It is the purpose of Part I of this paper t o bring a direct proof that gas-induced orientation of evaporated nickel films produces ( I 10) planes as the metal-gas interface.Experimental It is generally accepted that when hydrogen is chemisorbed on a clean metal surface, such as a nickel surface, that each hydrogen atom occupies one crystallo- graphic site. but is also borne out by adsorption experiments and by the fact that the heat of adsorption is high, 30 kcal. per mole for the sparsely covered nickel surface and 15 kcal. for the completely covered ~urface.~ Thus measurement of the number of hydrogen atoms adsorbed affords a means of measuring the number of crystallographic sites irrespective of size. Beeck, Ritchie and Wheeler (see also 5 , have shown recently that such measurements have to be done cautiously, having in mind that hydrogen is also absorbed into the interior of the metal structure.Initial adsorption a t liquid nitrogen temperature of - 196" C and 0.1 mm. pressure represents a true measure of the number of sites available for adsorption, al- though, as has been shown,3 the total sorption at room temperature as measured shortly after admission of the gas is in close agreement with the value for ad- sorption at - 196" C . The total surface, irrespective of the number of crystallographic sites, can be measured by the Brunauer, Emmett, Teller (B.E.T.) method.6 However, this method will give correct results only when no chemisorption of the gas employed takes place. The author and his co-workers have found 5 that nitrogen which is generally used in the B.E.T.method cannot be used with clean nickel, iron and many other metal surfaces because nitrogen is adsorbed on nickel at - 196' C with a heat of adsorption of 10 kcal. for the sparsely covered surface decreasing to 5 kcal. for the completely covered surface.3$ 5 The nitrogen mole- cule occupies two crystallographic sites and is probably only partially dissoci- ated. This chemisorption, which cannot be detected by the usual B.E.T. procedure using nitrogen a t liquid nitrogen temperature, causes the surface to appear too large by a factor of 1-55 in the case of nickel. Accordingly, gases have to be used whose adsorption is of known van der Waals' type at the tem- perature of measurement. Neon, krypton and butane were used in this investiga- tion at such temperatures as to make the number of molecules in the gas phase of nearly the same order of magnitude as the number adsorbed on the small surface studied.This is necessary in order t o obtain the desired high accuracy and can be achieved by using suitably small saturation pressures. The use of krypton at - 196" C for the measurement of small surfaces comparable with those of the metal film has been previously described by Beebe.' By measuring the number of sites relative to the total surface in cm.3 for oriented and unoriented nickel films, the average size of the site for each type of film can be established. This has been shown by electron diffraction,** 2 Rupp, 2. Elektrochem, 1929. 35, 586. 3 Germer, 2. Physik, 1929, 54, 408.4 Beeck, Ritchie and Wheeler, J . Colloid Sci., 1948, 3, 505. 5 Beeck in Advances in Catalysis, Vol. I1 (Academic Press, New York, 1950). 6 Brunauer, Emmett and Teller, J . Amer. Chem. SOC., 1938, 60, 309. 7 Beebe. Beckwith and Honk, ibid., 1945, 67. 1554. Beeck, Cole and Wheeler, this Discussion.0. BEECK AND A. W. RITCHIE I61 It will be seen that for unoriented films the surfaces measured by the B.E.T. method using molecules of different sizes agree well within the limits of experi- mental error. The oriented films, however, with their larger surfaces and smaller pores gave surfaces markedly dependent of the size of molecule used, making it necessary to extrapolate to the size of the hydrogen molecule since the chemi- sorption of hydrogen is used for comparison.For this purpose the area of the hydrogen molecule was taken as I Az. This figure may be somewhat arbitrary because adsorbed hydrogen atoms have been shown 3~ to migrate from site t o site and the pore size necessary for this fast migration (as distinguished from the slow diffusion into the crystal lattice) would be difficult to evaluate. It will be seen, however, that whatever the choice of this figure, within the possible limits, it has little bearing on the quantitative conclusions t o be drawn. The surface measurements with neon were carried out at 16" K in a Collins helium cryostat. A full description of the experimental details will be pub- lished later. The cross-sectional areas of methane, krypton and neon were calculated by means of the formula A = 1-52 x 10-l~ ( ~ % f / p ) ~ l 8 , where p is the liquid density and M the molecular weight.The area of n-C,Hlp was calculated by the method of Livingston * assuming that the molecule lies flat on the surface as a cylinder with the diameter of 4-75 8, (the distance of nearest approach between two hydrocarbons) and with a length of 1.29 A per carbon atom. The procedure of producing the evaporated metal films has already been described in detai1.l Results The experimental results are presented in Table I and in Fig. 2. Fig. I shows the B.E.T. plots for Expt. 7, 8 and g, where the hydrogen adsorption a t - 196' C and the B.E.T. surface measurements with neon, krypton and butane were done on the same oriented film. The hydrogen adsorption was carried out first after experimental proof was obtained that adsorbed hydrogen had no measurable effect on the B.E.T.surface measurements. Expt. 5 and 6 were TABLE I - No. __ I 2 3 4 5 6 7 8 9 I0 -. - Film Type Unoriented Unorient ed Unoriented Unoriented Oriented Oriented Oriented Oriented Oriented Dep. Temp. -183" C Unorient ed Ads. Temp. O C - 196 - 196 - 196 - 196 -257 - 196 -257 -196 - 78 - 78 Gas Kr Kr CH4 n-C4H,o Kr Ne Kr Ne Kr n-C*H,o Uni layer Molecules Per IOO mg. Ni x 10-18 6-15 5-40 11.70 15-00 12.08 5-29 3.68 5-85 3'48 20.00 Area per Molecules A2 14.6 14.6 15-68 24'5 14.6 14.6 24'5 14.6 9'95 9'95 Surface Area m.2/g. 9.00 8-55 8.46 8-53 17-09 14-92 17-64 12-95 19-90 5'37 H2 Ads. per IOO mg. Ni dolecules x 10-18 Area Per H Atom Site A2 6.18 6.18 6-14 6-18 6-42 7'25 6-37 4.68 7-18 3'9 made with two different oriented films, the hydrogen adsorption at - 196'C preceding the B.E.T measurements.Numbers 2 , 3 and 4 are averages of measure- ments on several different unoriented films. In these experiments the hydrogen adsorption at 23'C was used, while Expt. I was preceded by a hydrogen ad- sorption measurement a t - 196" C. The average value for the area per H atom site for the unoriented films is 6.17 A2 and is independent of the size of molecule used for the B.E.T. measurement. 8 Livingston, J . Amer. Chem. SOC., 1944, 66, 569. I7I 62 THE EFFECT OF CRYSTAL PARAMETER Extrapolation t o I Az for the oriented film is obviously best done by a curve through the points from Expt. 7, 8 and g , where the same film was used. This happens t o be a straight line and extrapolation leads to the value 8-65 A2 for the (110) oriented films.15 .- = to 2 0 E! \ + X w 5.0 - IZ 0.05 PIPS FIG. I. 0.10 Discussion The X-ray values for the short and long distances in the face-centred cubic nickel lattice are 2.48 A and 3.51 A, respectively. The (1x0) site area is therefore 2-48 x 3-51 = 8-70 A2 in unexpectedly good agreement with the experimental value of 8-65 Biz. The area of the (100) site is 2-482 = 6-15 Aa, again in excellent agreement with the experimental value of 6.17 Aa. Since planes involving higher indices would undoubtedly adsorb more than I H atom per site, it may be safely concluded that (110) oriented films exhibit only (110) planes to the gas phase. In the case of the unoriented films the agreement with the (100) site aiea may be more fortuitous since the average size of the three major planes in- cluding the (111) plane is If, therefore, only a small number of planes along the cube edges were (110) planes and another small number at the corners were (111) planes, one would expect an overall value close to that of the (100) plane.It was also shown l that relatively heavy high vacuum films (film weights of several hundred mg. were used in these experiments) showed a tendency to slight ( I 10) orientation with increasing thickness, if deposited at 23" C. Since such films showed also a slightly increased activity per unit weight, it must be concluded that heavy high vacuum films expose a certain un- known fraction of (110) planes also for this reason.One film (Expt. 10) = 5-84 A2 very close to the (100) value.0. BEECK AND A. W. RITCHIE 7.0 - 6.0 - 5.0 - Ne KI n-C,)r, I I I 1 4 u- != E 0 v) 2 5 4 d Q The dependence of surface on pore size makes it also necessary to adjust the ratio of the intrinsic activities of the two film types. It was earlier reported that the intrinsic activity of oriented nickel films for the hydrogenation of ethylene was about five times greater than for unoriented film. If it is assumed that the surface available for ethylene hydrogenation in oriented films is the same as that measured by methane adsorption (which would be the lower limit), the new factor becomes 5 x 1-4 = 7, as can readily be deduced from Fig. 2 , since the earlier comparison was based on hydrogen adsorption only.Oriented evapor- ated nickel films are, therefore, at least about seven times more active than unoriented films. 11. Hydrogenation of Benzene and Dehydrogenation of Cyclo- hexane.-After having demonstrated that ( I 10) planes are exposed in oriented evaporated nickel films and are at least five, but possibly seven times mole active (see Part I) as catalyst for the hydrogenation of ethylene than the planes exposed by unoriented films (which may either be (100) planes only or contain some (110) and (111) planes also) it was tempting to test such films for the hydrogenation of benzene or the dehydrogenation of cyclohexane for the latter of which Balandin9 has proposed his " multiple adsorption " hypothesis. According to this hypothesis dehydrogenation of cyclohexane takes place on the octa- hedral face, i.e.(111) planes, of a face-centred cubic metal lattice. He further specified that the lattice constant must lie between those of nickel and palladium. Since at the temperature necessary for dehydrogena- tion (about 300'C) nickel and iron films sinter to the extent that they lose nearly completely their internal surface * only the hydrogenation of benzene at 2 3 O C and at about 6ooC was studied over these two metals. Dehydrogenation of cyclohexane was carried out over tungsten, nickel and platinum films. No systematic study was attempted in these ex- periments. 9 Balandin, 2. physik. Chem. B , 1929, 2, 289 ; ibid.. 1929, 3, 167.1 64 THE EFFECT OF CRYSTAL PARAMETER Experiment a1 The experimental technique was that used previously except that metal valves were used instead of greased stopcocks in order t o avoid solution9 of benzene and cyclohexane in the grease.The benzene was Merck Reagent, thiophene-free and was used without further purification, except for removal of air by repeated freezing at liquid nitrogen temperature and pumping to high vacuum. The cyclohexane was an A.P.1.-N.B.S. spectrographic standard sample of better than 99 yo purity. All metal films were sintered a t the reaction tem- perature at which they were t o be used and the hydrogen adsorption at 23" C was used for a measure of the surface available for the reactions t o be studied. Results Benzene Hydrogenation.-Both oriented and unoriented nickel films were active for the benzene hydrogenation a t 58" C.The initial rates were slow, i.e. of the order of I mm. pressure decrease per minute for a benzene-hydrogen ratio of 1/3 at the initial total pressure of about 240 mm. The reaction was found to be independent of the benzene pressure and proportional to the o-44th power of the hydrogen pressure at 58" C and proportional t o the o.56th power at 100' C. The rate a t 19" C was six times lower than that a t 58" C, correspond- ing t o an activation energy of 8-7 kcal. for this temperature interval. The specific reaction rate per unit surface was found t o be greater for the oriented films than for the unoriented film by a factor of 1.24. Comparative experiments on nickel and iron were carried out a t 23OC on unoriented films and the specific activity of nickel was found to be 2-3 times that of iron.All measurements are tabulated in Table 11. TABLE I1 I 2 3 4 5 6 - .-. Y cd ; - Ni Ni Ni Ni Fe Nj - Oriented ,I Un- oriented *, ,, 8 , 58 58 57 58 23 21.5 - Ej o c E v A, 181.0 183.0 171.5 159'5 318.5 342.0 h v g c o v A, - 57'5 58.0 58.5 40'5 67.0 63.0 8.42 10.10 6-91 7'51 11.90 5'95 6-31 7'56 5-18 5'63 I I 'go 5'95 h ." .* +. 9 2 0) cd .A Y 4 0'234 0.239 0.199 0.181 0.039 0.091 2 8 4 ~. - Evaporated at 0" C Evaporated a t oo C Film evaporated at -183~ C Evaporated a t 0" C ,, 3 9 3, I 9 * Corrected to Ph2 = 318.5. Cyclohexane Dehydrogenation.-The cyclohexane dt ydrog iation was investigated over tungsten and platinum films and was attempted over palladium. Those three metals sinter much less than nickel (see Table 11). In spite of this fact tungsten was found t o be inactive a t 324°C. Platinum and palladium were both very active, although proper evaluation in the case of palladium was impossible due t o the high rate of solution of hydrogen in this metal.The measurements on platinum are set forth in Table 111. lo Beeck, this Discussion.Run -- I 2 3 4 5 6 7 8 9 I 0 0. Film Type Unoriented I , Partially oriented BEECK AND A. W. RITCHIE Reaction Temp. "C 324 324 324 324 324 324 324 325 322 324 * Calculated (see text). TABLE I11 P o C yclohexane (-.) 5 7'0 58.0 29'5 29.0 32'5 32-0 32.0 57'0 57'0 57'0 - dpfdt -_ IOO mg. Corrected to 57-0 mm. Cyclohexane 14.8 7'48 10.60 10.41 6.63 8-06 1-06 0.57 0.85 7.98 Hz Ads. 23O C After Sintering Molecules x 10-18 Relative Activity 13.0 11'0 I 1.6 I 1-5 12'1 12.2 11'0 1'4 0.8 1.1 -f H, adsorption before sintering. As shown in Fig.3, where the logarithm of cyclohexane pressure is plotted against time, the initial reaction rate over platinum is first order with respect t o the partial pressure of cyclohexane which allows for easy normalizing of the measured rates. From the temperature coefficient of the rates at 326OC and 356" C an activation energy of 9.3 kcal. was calculated. In Expt. 2 and 10 adsorp- tion of hydrogen was measured a t 23' C before sintering, and in agreement with previous findings there is little difference in the ad- sorption of oriented and unoriented films. In Expt. 3 to 10, hydrogen adsorption a t 23OC after sintering was measured showing that one- third of the surface is still available after sintering at 325" C.In Expt. I , 2, 7, 8, g and 10 hydrogen ad- sorption was not measured after sintering, and the values were calculated for I , 2 and 7 from a plot of the figures in column 5 against those in column 6 for Expt. 3 to 6. The values for the oriented films 8, 9 and 10 were assumed t o be the same and adjusted through the adsorption measure- ments of the unsintered films z and 10. From the ratio of the averages in column 8 it is seen that the 0 10 20 30 FIG. 3. MINUTES unoriented films are more active by a factor I I * ~ / I * I = - 11 than the oriented films. This figure is likely to be much greater since electron diffraction studies showed the " oriented " films to be partially oriented only. All that can be said with regard to palladium films is that a high-vacuum film which was not strictly unoriented but showed mixed orientation under electron diffraction study was for the same film weight 2-5 times more active than a (110) oriented film deposited in I mm.of nitrogen.I 66 ALLOY CATALYSTS Ih’ DEHYDROGENATION Discussion Because oi the zero-order reaction rat2 with regard to benzene it must be concluded that the benzene hydrogenation proceeds through the adsorbed state of benzene and is therefore possibly not subject to influence by the size of the crystallographic sites but is simply a function of total surface available which, as has been shown in Part I, is 1.4 times larger for oriented than for unoriented films for the same hydrogen adsorption. This should make the rate on oriented films 1-4 times faster, A smaller factor would be expected since benzene may not be able to enter the smallest pores.Within the experimental limtis the rates may be regarded as equal. If the same argument were applied to the comparison of un- oriented iron and nickel films, iron should be 2-7 times more active than nickel if one assumes that the large (100) face is exposed. Actually the activity for the same hydrogen adsorption is half as large. Further dis- cussion of this point will be postponed until after Part 111. The tenfold activity of unoriented films even over only partially oriented films suggests that Balandin’s hypothesis may be operating if the assumption is made that unoriented films expose, in part at least, (111) planes. In the later stage of this reaction the adsorption of benzene poisons the reaction, reducing the initial rate of a second run on the same film by a factor 2 0 . It is likely that the inactivity of tungsten films (the surface of which after sintering at 325O C is still about I / Z of that at 2 3 O C) is due to poison- ing by a reaction product with the surface. Experimental proof of a highly unsaturated residue was obtained by admitting small amounts of cyclohexane to both platinum and nickel films at 25oOC. In both cases 4 hydrogen molecules appeared in the gas phase for each cyclohexane molecule admitted. Even at room temperature cyclohexane is strongly adsorbed on tungsten and nickel with evidence that dissociation to benzene and hydrogen takes place. This strongly suggests that poisoning by residues is responsible for the inactivity of tungsten. For further discussion of the results presented in Parts I and I1 of this paper in connection with other aspects of metal film catalysis see (10). Shell Development Company, California. Emeryville,

ISSN:0366-9033    

DOI:10.1039/DF9500800159

出版商:RSC    

年代:1950

数据来源: RSC

摘要:

166 ALLOY CATALYSTS Ih’ DEHYDROGENATION ALLOY CATALYSTS IN DEHYDROGENATION * BY GEORGE-MARIA SCHWAB Received 24th October, 1949 In homogeneous Hume-Rothery alloys the activation energy of the formic acid dehydrogenation increases in proportion t o the square of the increase of the electron concentration caused by the multivalent solute metal. Amongst the intermetallic Hume-Rothery phases, the y-phase shows a maximum activation energy. In general, activation energy and electric resistance run parallel. It is concluded that the catalytic activation consists in an entrance of substrate electrons into empty levels of the first Brillouin zone of the metal. It is suggested that in alloys containing polar or covalent bonds the conductivity electron concentration is lowered. Catalytic experience confirms this view.The frequency factors of the dehydrogenation reaction on different catalysts increase logarithmically with increasing activation energy. The mechanical hardness of alloys runs parallel with the activation energy, and it is concluded that the hardness is a function of the saturation of the Brillouin zone. This is con- firmed catalytically for martensite. * Contribution from the Dept. of Inorganic, Physical and Catalytic Chem- istry of the Institute Nicolaos Canellopoulos, Piraeus, Greece.GEORGE-MARIA SCHWAB 167 It has often been suggested 1, that the catalytic action of metals, where the cycle is not a simple oxidation-reduction one, is connected with their electronic structure. Only recently, direct proofs for this view are appearing, e.g.in Bockris's connection between overvoltage and work function. In this case, however, it is not yet clear if a discharge of ions or a catalytic recombination of atoms is rate-determining. A direct com- parison of catalytic action and electronic structure has been found in our laboratory by the systematic study of the dehydrogenating action of alloy catalysts. In the present work, the activation energy of the dehydro- genation of formic acid vapour has been measured in presence of homo- geneous alloys of varied electron concentration and intermetallic phases. Kinetics.-In all cases of simple catalytic reactions, the Arrenhius equation k = k , . e-qiRT, represents fairly well the experimental temperature dependence of the velociy constant k, q being known as activation energy.Only in zero- order reactions proceeding on entirely saturated active surfaces (centres), is k free from a temperature-dependent adsorption factor, so that q may be considered as the true activation energy required by the adsorbed molecules. The formic acid decomposition has been shown to be of this type.4 As K O or its logarithm B contains the unknown active pro- portion of the geometrical surface, a determination of q is possible only by measuring the temperature coefficient or log K = B - q/4*57T, . - (1) d log k d(IIT) = 414'57. Special care is needed in such measurements so as not to confuse accidkntal activity alterations with the true temperature coefficient of k . Only complete coincidence of temperature-velocity curves, re- peated at rising and falling temperatures, can yield reliable values of .4. As a secondary result, empirical values of the fre- quency number, k , (or of B ) , are obtained, whose dependence on various factors form the object of separate study (see below). Experimental '$ f [ The experimental arrangement developed for this purpose and used in most of the work, is shown in Fig. I . ~ Liquid formic acid is con- tained in the vessel A which com- municates with the evaporator B, filled with glass particles or fine capillaries. In that particular zone of B which, by means of the electric furnace C, is maintained just at the boiling point, the liquid is evaporated smoothly and the vapour a t atmo- spheric pressure passes over the catalyst in E. Undecomposed acid f /I ==r FIG.I . is recirculated by G, H, and rhe gas- eous reaction products (CO +Ha) purified in J and K, enter a flowmeter at L. MM is the outer tube of a 1 Goldschmidt, Ber., 1927, 60, 1263. a Russell, Nature, 1926, 117, 47. Bockris, ibid., 1947, 159, 539. Hinshelwood, J . Chem. SOC., 1923, 123, 1014. and ref. Schwab and Theophilides, J . Physic. Chem., 1946, 50, 427.168 ALLOY CATALYSTS IN DEHYDROGENATION thermocouple, removable by a ground joint a t the top of the apparatus. By this arrangement, the reaction velocity can be directly obtained for every temper- ature of the catalyst in runs a t rising and falling temperatures. Always, after a short period of " formation " of the catalyst, coincident curves are obtained and evaluated in the form of log k - 1/27 diagrams according to eqn. (I) and The alloys were prepared from chemically pure metals in porcelain crucibles under suitable molten salts in a gas blast-furnace, and the phases present were found by X-ray analysis.Ductile metals were used as foils, brittle alloys as pieces, and the total geometrical surface of the pieces estimated by measuring their dimensions. Homogeneous alloysJ. ?-Silver dissolves most of the metals of Groups 11-V of the Periodic Table t o form Hume-Rothery's a-phases, without a change of the cubic face-centred lattice type, up to an electron concentration (EC) = 1.33 in the period VB (Cd, In, Sn, Sb) and EC = 1-1 in the period VIB (Hg, T1, Pb, Bi). Thus, in these systems, it is possible t o alter the EC a t will without a radical change of the chemical and crystallographic character of the catalyst.The results of the kinetic measurements show that the activation energy using pure silver (17.6 kcal./mole) is always increased, by a few kcal./mole by " addi- tion " of Cd, In, Sn, TI, Hg, but by amounts up t o 20 kcal./moIe by Sb, Pb, Bi. The increase is not always strictly proportional t o the concentration of the solute metal, but for small additions the results can be approximately represented by x being the atomic fraction of the solute metal, n its Group number or valency, and A a constant, amounting t o 11 in the 5th and IOO in the 6th period if q is expressed in kcal. Similar results were obtained with elements of the 4th period ( A - 100) and with solutions of thallium, lead, cadmium 8 and iron Q in gold.Hence, in homogeneous a phases the activation energy is increased when the EC is raised by dissolution of a multivalent solute, and this increase, exactly as is the electric resistance, is proportional t o the square of the EC increase per solute atom. Likewise, within the domain of the densest hexagonal e phase of Ag-Sb,los 7 the activation energy rises with increasing Sb content or EC. Heterogeneous alloys and intermetallic phases .-In binary alloys con- sisting of two phases (Ag + Cu,ll cc + y, y + E and E + I) Cu-Sn,', lo a + and fl + y Au-CdJ8 A1 + eAg-A1,ll Au + Fe,9 Cu + Cu,Mg, Cu2Mg + CuMg2,0) the activation energy is generally found t o lie between those of the component phases. The intermetallic phases or compounds themselves show different values.For the system Cu-Sn 7, lo these are shown in Fig. 2, and for Au-Cd 8 (2). 4 q u + Ax(n - I)', - (3) A - hardness : kg. mm.-2 B - - 0 - act. energy: kcal. mole -1 C . . . . . . resistance: Qcm. 104 FIG. 2. in Fig. 3. Generally, the cubic face-centred phase fl, the hexagonal densest phase E and the hexagonal layer phase I) show values not very different from 6 Schwab and Holz, 2. anorg. Chew., 1944. 252, 205. 8 Schwab and Pesmatjoglou, J . Physic. Chew., 1948, 52, 1046. S Schwab and Petroutsos, ibid., 1950, 54 (in press). lo Schwab and Karatzas, 2. Elektrochem., 1944, 50, 204. 11 Schwab and Schwab-Agallidis, Ber., 1943, 76, 1228. Schwab, Trans. Faraday SOC., 1946, 42, 689.GEORGE-MARIA SCHWAB that of the corresponding saturated a phase, while the more complicated cubic phase y always gives a very distinct maximum of the activation energy.As Fig. z shows, and as can be shown for numerous other systems,12 this exceptional behaviour of the y phases concerns also the electric resistance and the hardness. FIG. 3. Theory of the electron influence.-According t o the theory of Mott and Jones l3 the upper limit of stability of a purely metallic (Hume-Rothery) phase is given by that electron concentration at which the co-ordinated material waves of the fastest electrons just begin to be reflected by the most intensely reflecting lattice planes according to Bragg's law. At this limit, the surface of a sphere, constructed with the reciprocal half wavelength of the fastest electron as radius, touches from within planes of a polyhedron enclosed by planes parallel t o the above lattice planes at distances from the centre of the sphere reciprocal t o the respective lattice distances.This polyhedron is called the volume of the first Brillouin zone in the k-space. This theory permits a satisfactory theoretical calculation of Hume-Rothery's electron concentrations of the stable phases a, 8,. r and 5. It is important that the volume of the first Brillouin zone at the stability limit is not filled by the sphere ; the corners of the polyhedra, amounting to about 30 yo of the total zone volume a, /3 and Q, remain empty. It is in these corners that electrons, acquiring energy from an external field, can move, and this accounts for the fact that even saturated phases a, ,8 and c are still good conductors, although poorer than a pure monovalent metal with a half-empty zone.Its zone, as represented in the k-space, is, because of the complicated lattice structure, of a highly polyhedric, nearly spherical shape, and thus in the state of saturation it is almost completely (i.e. 88.5 yo) filled by the sphere. Little room is left for accelerated electrons, and this a t once accounts for the high resistance of y alloys. Now, our experimental results for the activation energies, for homogeneous alloys as well as for intermetallic phases, can be expressed in terms of the zone theory as follows. The activation energy increases within the domain of one phase according as the Brillouin zone approaches the saturation limit, and, comparing different phases, it is highest in the y phase where there is the highest degree of zone completion. Thus, generally, empty levels in the zone favour the catalytic reaction. We are led to the conclusion that, t o form an activated adsorbed formic acid molecule, electrons from it must intrude into the metal, and that this entry needs more thermal energy the more the metal is already saturated with electrons. Most probably the two protons of the formic acid molecule occupy extra lattice sites in the catalyst surface, while their electrons are " dissolved " in the electron gas of the metal.A remarkable exception is the y phase. l2 Scbwab, Exflerientia, 1946, 2, 103. l3 Mott and Jones, The Theory of the Properties of Metals and Alloys (O.U.P. 1936). F"170 ALLOY CATALYSTS IN DEHYDROGENATION This picture of the activated state of dehydrogenation (and, of course, also of hydrogenation) agrees fully with the views generally accepted for the dis- solution and activated adsorption of hydrogen.It gives, at the same time, a concrete interpretation, based on experiments, to the concept of a homopolar substrate-catalyst bond, by supposing a sharing of substrate electrons. Theory of polar and covalent alloys.-The zone theory of metals and alloys is based on the assumption that the valency electrons of the metal atoms form a degenerated gas distributed according to Fermi statistics over the levels of the Brillouin zone. This is the extreme case of a purely metallic bond, as it occurs in typical metals and Hume-Rothery alloys. The same picture, however, has been successfully applied by Mott and Jones l3 to the description of lattices, which, from the crystallochemical point of view, are not purely metallic.E.g. in graphite, the four electrous per atom can be placed in one common Brillouin zone, and likewise in selenium, the six electrons per atom. However, many arguments suggest that in graphite three electrons per atom are localized in covalent bonds, and only one is a conductivity electron, and that in selenium two electrons per atom are localized in covalent bonds between neighbouring atoms of endless chains parallel to the c-axis, and only four electrons are con- ductivity electrons. This picture suggests a different treatment, viz. the electrons belonging to covalent bonds are placed in fully complete zones and only the conductivity electrons in a Brillouin zone of smaller k-volume.In fact, it has been shown * that a zone containing one electron per atom can be constructed from reflecting planes of the graphite lattice, and most probably a similar procedure is possible for selenium and other semi-metals. This viewpoint is important in discussing the degree of saturation of the conductivity zone in alloys other than of the Hume-Rothery type, containing covalent or polar bonds in addition to the metallic bond. In these, probably, the number of free electrons to be placed in a conductivity zone, taken from X-ray data, is less than the total number of valency electrons of the constituent atoms, because electrons are used in forming covalent bonds or anions.From this point of view, a relatively low activation energy will be an argument in favour of non-metallic bonds. In the Au-Fe system, on dissolution of a few atomic per cent. of iron in the gold lattice, the activation energy and specific resistance are higher due to the two valency electrons contributed by the iron atom. At higher percentages, however, the magnitude of both properties decrease again, indicating, in agreement with magnetic data, that valency electrons now enter the gaps of the incomplete iron shell and dis- appear from the conductivity zone. Gold and antimony form a compound AuSb, having the lattice type of iron pyrite FeS,. Its activation energy was found to be nearly equal to that of a saturated solution of antimony in gold. This reveals a relatively low degree of saturation of the first Brillouin zone.Apparently, the covalent bond between the pair of antimony atoms uses most of the valency electrons of antimony. In the compound Cu,Sb, the antimony atoms are isolated from each other. Nevertheless, this compound shows the same activation energy as the Sb-poorer 6 phase, an energy only slightly higher than that of pure copper. Hence, even here, the antimony electrons do not fill the conductivity band, the antimony atoms being anions or a t least neutral. In the system Cu-Mg, the ordered cubic compound Cu,Mg has a much increased activation energy. This, together with the lattice data, show that the binding is almost entirely metallic, as in Hume-Rothery alloys. The other compound CuMg, shows an activation energy much lower even than that of copper itself.As the lattice data indicate a purely metallic bond, the first Brillouin zone must have an extremely low degree of saturation. This is not due to chemicaE bonds in this case, for the X-ray diagram of this phase reveals that the poly- hedron representing the zone volume is far from spherical in shape and very sparsely filled by the inscribed sphere. The result with Cu,Mg was confirmed by the behaviour of the isomorphous compound Au,Pb. This, also, does not exhibit an increase of activation energy, as compared with lead-saturated gold. Frequency numbers .-Hitherto, only the activation energy, the term q in eqn. ( I ) , has been considered. The actual absolute reaction velocity is deter- mined by this and the temperature-independent frequency factor k,, or its log- arithm B.This factor, in a zero-order reaction, ought to be a lattice frequency of v - 1013 sec.-l, multiplied by the number of moles covering the unit surface. Expressed in flow rate of the reaction products, this product has been calculated t o be loll to 1o12 ~ m . ~ min.-l crn.-,. The experimental evidence l4 shows that l4 Schwab, Proc. X I I d . Congr. Pure Appl. Chem; (London, 1947). Different alloy systems have been examined.GEORGE-MARIA SCHWAB 171 this value gives an upper limit only for fully active surfaces. In real systems, the measured B-values are much less, and the smaller this is, the lower is the activation energy. The empirical relation B = B , + q / h . - (4) has been shown 1* to be a general law of heterogeneous catalysis.It indicates that a low activation energy is always, at least in part, compensated by a low B-value. Eqn. (4) is explained by assuming that centres with high activation energies at a catalytic surface are more frequent than very active centres with low activation energies. The magnitude h is a parameter determining the dis- tribution law of the active centres in the surface. It often l5 depends on the temperature a t which the distribution equilibrium has been established. Fig. 4 shows that eqn. (4) is valid for alloy dehydrogenation catalysts. B has been plotted against q, and two straight lines represent the results approximately, one for the cubic face-centred metals (a phases) and another one for other lattice types. We assume, a t least preliminarily, that this does not signify a different h-value, but that it is due to the greater hardness of alloys other than a phases.This results in a greater roughness of the broken surfaces of the pieces, while R refers to the geometrical minimum surface. Mechanical hardness.-The increase of hardness with increasing activa- tion energy is a general feature of dehydrogenating alloys and very generally the hardness l* shows the same dependence electron concentration and lattice type as do electrical resistance and activation energy. This is illustrated in Fig. z which contains the three curves for the Cu-Sn system. It has been shown that the hardness increase is due to an increased formation energy of atom dislocations and i t was suggested that this in turn is due to the greater resistance against increases in local atomic distance (zone volume decreases) found in phases with a nearly full Rrillouin zone. From this point of view, the question whether an observed high hardness is of electronic or of structural nature, can be checked by measuring the activation energy of dehydrogenation. The hardness of carbon steel or of martensite may be due to the increased electron concentration of martensite as compared with ferrite (a iron), as similar hardness increases are observed in other alloy systems of similar EC. A direct proof is to be seen in the fact that hardened steel dehydrogenates fc;.mic acid with an activation energy of 25 kcal./mole whereas soft pure iron or annealed steel requires 20 kcal./mole only. Thus, i t appears possible to utilize the dehydrogenation catalysis on the basis of its clear-cut electronic mechanism as a useful auxiliary method for alloy research. Dept. of Inorganic, Physical and Catalytic Chemistry. Institute Nicolaos Canellopoulos, Piraeus, Greece. l5 Schwab and Pesmatjoglou, Hedvall-Festskrift, Goteborg, 1948, 533. Schwab, Trans. Faraday SOC., 1949. 45, 385.

ISSN:0366-9033    

DOI:10.1039/DF9500800166

出版商:RSC    

年代:1950

数据来源: RSC

摘要:

THE PARAHYDROGEN CONVERSION ON PALLADIUM -GOLD ALLOYS* BY A. COUPER AND D. D. ELEY Received 30th January, 1950 The activation energy of the parahydrogen conversion on palladium-rich alloys is 3-5 kcal./mole and increases abruptly between 40 and 30 atomic per cent. palladium to the value of 8-5 kcal. found for gold-rich alloys. Magnetic measure- ments have identified this composition as corresponding to the complete filling of the d-band, or atomic d-orbitals, with electrons. Vacant d-orbitals are there- fore essential for the low-temperature catalysis, and it is suggested that they bond the chemisorbed hydrogen atoms M-H. The reaction goes through a triatomic H, complex formed by entry of an H, molecule from the van der Waals layer, H- -H- -H M-H + p H 2 -+ ‘M’ -+ OH, + M-H.\ / It is postulated that each site exerts two bonds to the activated complex, employing “ atomic ” d and hybrid d2.56 s P * . * ~ metallic orbitals. The values of temperature-independent factor found agree with earlier calculations for this mechanism. The d-band of palladium may also be filled by electrons from dissolved hydrogen atoms which cause a similar increase in activation energy. This work 1 was initiated in Bristol University in 1945 in an attempt to relate the activity of transition metal catalysts to their electronic structure. In the past various efforts have been made to provide a theoretical basis for catalyst activity, by relating catalytic activity to lattice spacing. Thus, on the experimental side there are the papers by Balandin,2 Long et aZ.,3 Twigg, Herington and Rideal,’~ 6 and Beeck and his co-workers.6B 7 On the theoretical side, calculations of the chemi- sorption of hydrogen molecules by the Eyring-Polanyi method indicate a marked effect of the lattice-spacing in the surface on the activation energy of the process.** @ Recent advances in our knowledge of the metallic state now make it possible to begin a more fundamental approach to the problem, and to attempt to relate catalyst activity to the nature and occupation of the electronic energy levels of the catalyst.All those metals active as hydrogenation catalysts fall into the group of transition elements, with the exception of copper which, in general, has a higher activation energy than the others. The atoms of these ele- ments possess partly empty d-shells and the theory of the bulk metal may be treated by either of two methods. The first method, due to Mott and * Part I in a series of papers on the Electronic Basis of Catalyst Activity.1 Couper and Eley, Nature, 1949, 164, 578. 2Balandin, 2. physik. Chem. A , 1927, 126, 267; B, 1929, 2, 2 8 9 ; B, 1929, 3, 167. Long, Frazer and Ott, J . Amer. Chem. Soc., 1934, 56, 1101. Twigg and Rideal, Trans. Faraday SOC., 1940, 36, 533. Herington, ibid.. 1941. 37, 361. Beeck, Rev. Mod. Physics, 1945, 17, 61. 6 Beeck, Smith and Wheeler, Proc. Roy. SOC. A . 1940, 177, 62. * Sherman and Eyring, J. Amer. Chem. SOC., 1932, 54, 2661. Sherman, 9 Okamoto, Horiuti and Hirota, Sci. Papers Inst. Phys. Chem. Res., Tokyo, Sun and Eyring, J . Chem. Physics, 1934, 3, 49.1936, 29, 223. T 72A. COUPER AND D. D. ELEY I73 Jones,l0 is an extension of the free-electron theory of Bloch, and con- siders that the " valency " electrons lie in two energy bands, a broad s-band overlapping a narrow d-band. The s-band can accommodate at most two electrons per atom and the d-band 10 electrons per atom, and thus there is a relatively high density of energy states in the d-band. The electrons fill up both bands to the same energy level, Hence in the element palladium, of the 10 valency electrons approximately 9-4 enter the d-band and 0.6 enter the s-band. Thus, there are 10 - 9-4 = 0-6 positive holes in the d-band. Alloying this metal with approximately 60 atomic per cent.* of silver or gold lowers the paramagnetism to zero, in agreement with the above model, which requires that the s-electron of the added monovalent element goes to fill the 0.6 holes in the d-band of the palladium.Besides paramagnetism, positive holes in the d-band give rise to high heat capacity, electrical resistance, and other properties. On this view we should be inclined to associate the catalytic activity of the transition metals with their partly empty d-band,ll and a promising line of investigation should be the catalytic activity of an alloy such as that of palladium with gold, where changes in lattice spacing are small. That it is not entirely possible to separate lattice spacing from electronic structure effects in metals is shown by the valence bond treatment due to Pauling.12 This author, also arguing from the magnetic data, considers that the bonds between atoms in a transition metal are electron pair bonds using hybrid metal orbitals built up from 2-56 4d orbitals, the 5s and 2.22 5p orbitals, viz. (d)2.5s The remaining 2-44 d orbitals are " atomic " orbitals.Of the valency electrons available in the transi- tion metal?, 5-78 are supposed to be used for bonds, so that in palladium with 10 electrons 4-22 are left to enter the atomic orbitals, leaving 2 x 2-44 - 4-22 = 0.66 unpaired electrons in the atomic orbitals. The 5-78 bonds resonate between the 12 neighbours surrounding the centre atom. Thus, what are holes in the d-band on the free electron theory are unpaired &electrons on the Pauling theory. These actual numbers were developed to explain the chromium-copper series, but are approximately applicable to palladium.In any attempts to apply such theories to catalysis, it is easiest to assume, in the first instance, that the electronic structures of the bulk metal persist unchanged into the surface layer where catalysis occurs. Investigations of the theory of surface energy levels have been made on the basis of the free electron model 1 3 9 l4, l 5 but are not yet sufficiently developed to apply to transition mctals. From the point of view of va.lence bond theory this means assuming that there is no change in orbital hybridization at the surface, either alone or in the presence of the reacting substrate. While this work was in progress, our attention was drawn by the note by Dowden l6 to the interesting work of Rienacker, who has made extensive investigations of the catalytic properties of alloys, though not from the theoretical viewpoint adopted here.His latest paper with Sarry,]' will be discussed later. Magnetic and crystallographic data support the suitability of palladium-gold and palladium-hydrogen systems for the present work. Nickel has the complicating feature of ferromagnetism, and some of the other alloys show departures from the simple theory. 10 Mott and Jones, The Properties of Metals and AZZoys (London, 1936). * All the alloy compositions in this paper are given as atomic percentages. l1 Eley, Research, 1948, I, 304. l2 Pauling, Physic. Rev., 1938, 54, 899 ; J . Amer. Chem. Soc., 1947, 69, 542. l3 Lennard- Jones, Trans. Faraday Soc., 1932, 28, 333. 1* Shockley, Physic.Rev., 1939. 56, 317, and references cited therein. l5 Pollard, ibid., 1939, 56y, 324. Is Dowden, Resewch, 1948, I, 239. 17 Rienacker and Sarry, 2. anorg. Chem., 1948, 257, 41. Pauling and Ewing, Rev. Mod. Physics, 1948, 20, 112.174 PARAHYDROGEN CONVERSION ON GOLD ALLOYS In palladium-gold the evidence l8 indicates that the lattice constant in- creases from 3-88 A (Pd) to 4-07 A (Au) almost linearly, all the alloys being homogeneous, face-centred cubic systems. No information is available on phase-separation in palladium-gold, but since no special heat treatment was given we may assume that all the alloys are in the " disordered " state, a view suppcrted by the electrical resistance data given in the next section. For palladium-hydrogen there is the complica- tion of two phases, a and fl.lB* For the a phase the lattice constant varies scarcely at all from o to 56 atomic per cent.hydrogen, being about 3.88 A. Between 3 and 10 atomic per cent. hydrogen, however, the /? phas,: appears, with a lattice spacing varying upwards from 3-97 A. The existence of this second phase is little understood. Experimental The palladium-gold alloy wires, 42 S.W.G., 0.1032 mm. diameter, were ob- A satisfactory check on their homogeneity As shown in The catalyst tained from Johnson Matthey, Ltd. and composition is given by values of the specific resistance. Fig. I , these agree very well with the earlier data of Geibel.21 FIG. I.-The specific resistance of the alloy wires used as catalysts, compared with the data of Geibel. W. Geibel ; 0 Alloys used as catalysts.consisted of a 15 cm. length of wire sealed down the axis of a cylindrical Hysil reaction vessel. This was connected via a liquid air trap and mercury cut-off t o an apparatus essentially the same as that used in the work on tungsten wiresz2 The apparatus was baked out initially and the resistance-temperature relation established for the catalyst wire. The reaction vessel was then immersed in liquid oxygen, parahydrogen admitted, and at zero time the catalyst filament was heated electrically t o the required temperature. The hydrogens used were all purified by passage through a palladium thimble and stored over sodium films. Analyses were made with a micropirani gauge.23 l8 Mundt, Ann. Physik, 1934, 19 (5), 721. l9 Linde and Borelius, ibid., 1927, 84 (4), 747.2o Kruger and Gehm, ibid., 1933, 16 (5), 174. 21 Geibel, 2. anorg. Chem., 1911, 69, 38. 22 Eley and Rideal, Proc. Roy. SOC. A , 1941, 178, 429. 23 Bolland and Melville, Trans. Faraday SOC., 1937, 33, 1316.A. COUPER AND D. D. ELEY I75 A new method was used to clean the filaments from oxygen, since heating in vucuo to a high temperature was precluded by the low melting points of the alloys. The reaction vessel was surrounded by a coil of five turns of wire connected to an oscillator, immersed in liquid oxygen, and filled with 0.1 mm. hydrogen gas. On switching on the oscillator a current of 2.9 A a t a frequency of 15 Mc./sec. passed through the coil, giving a luminous discharge which was continued for 5 min. The work of Johnson 24 and tests with the homogeneous parahydrogen conversion 25 showed that the concentration of hydrogen atoms in the discharge was quite adequate to clean the filament. It was essential to heat the wire in vacuo subsequently at 7oo-goo0 K to remove dissolved hydro- gen atoms.This method was used with all the wires from pure gold to go atomic per cent. palladium, namely, Au, 10 yo Pd, 20 yo Pd, 30 yo Pd, 55-z.y0 Pd, 70 % Pd. The solubility of hydrogen atoms in go yo Pd and Pd was so high that it was found better to clean these wires by heating in molecular hydrogen at 600° K, followed by a period of outgassing a t 7oo-goo0 K. Activation Energy and Frequency Factor in Palladium-Gold Alloys .- A series of experiments was made for each wire at 1-2 mm. Hg pressure over as wide a range of temperatures as possible.In many cases the results were checked with a second specimen of wire. The reaction velocities were expressed in terms of the first-order constant k,, defined below, which has frequently been f0und,~59 2 2 and again in the present investigation, to describe accurately the course of the reaction. Here I C k , = - In -O t C,’ where C,(C,) is the concentration of parahydrogen in excess of the equilibrium value at time zero ( t ) . A set of typical results in Fig. 2 shows that the effect FIG. ;?.-Typical Arrhenius lines for the parahydrogen conversion on Pd/Au alloys of smaller Pd content. 0 55.2 atomic yo Pd. x 30 atomic % Pd. n 40 I 1 % Pd. 17 20 I , % Pd. of temperature is given by the Arrhenius equation, k , = Be -E/RT.In general, the activation energy E may be obtained with a maximum uncertainty of 0.3 kcal., with a corresponding uncertainty in B of a factor 4. To correct for differences in reaction volume V and catalyst area A , we employ Bo = B( V / A ) ’ 24 Johnson, Proc. Roy. SOC. A , 1929, 123, 603. 25 Farkas, Orthohydrogen, Parahydrogen and Heavy Hydrogen (Cambridge, 1935).176 PARAHYDROGEN CONVERSION ON GOLD ALLOYS the value of the frequency factor for unit volume and unit area, usually given in units of cm. min.-l. The results are collected in Table I. TABLE I.-ACTIVATION ENERGY E AND FREQUENCY FACTOR Bo Atomic Yo Pd in Alloy Temp. OK 170-330 170-330 170-330 1 70-3 30 180-350 200-500 150-330 150-330 150-330 170-330 150-350 200-400 400-800 3 70- 1000 3 70-1000 300-750 500-800 500-770 E cal.mole-' 3640 3960 3980 3140 3980 3470 3170 2820 3 I90 3520 2960 3180 8000 8950 7850 8600 1 7500 17500 B min.-1 4.68 x 102 1-20 x 103 4-90 X 10' 6-74 x 10 8.51 x 1oz 1-15 x 10 1-02 x 103 7-58 x 10' 6-30 x 102 2-04 x 10 2-95 X I0 1.51 x 10' 1-31 x 102 6.45 x 10 1-25 x 102 5-83 x 10' 7'59 x I 0 2 2-34 x 10' Bo cm. mh-1 2-75 x 105 7-05 x 105 2-59 x 105 3-96 x 104 4.82 x 105 5-40 x 103 4-80 x 105 3.56 x 105 4-10 x 105 2-99 x 105 1-14 x 104 1-82 x 104 1.18 x 105 7'41 x 104 6.35 x 104 3-72 x 104 2.06 x 104 8-31 x 102 * indicates not plotted in Fig. 3, 4. Notes.-(a) First specimen-first determination. (b) First specimen- second determination. (c) Second specimen. (d) After '' activation " by oxidation and reduction. ( e ) First specimen, heated a t 600° K in hydrogen and then in zlucuo.(f) Fjrst specimen, heated a t gooo K in hydrogen and then in vucuo. (g) First specimen, discharge applied, outgassed a t 800' K. (h) Second specimen, discharge applied, outgassed at 750" K. ( j ) Second specimen. ( k ) Second specimen. (I) Gold wirc, 60 cm. long. (m) Gold foil in vessel, walls a t catalyst temperature. 2 3 4 5 6 -.I i FIG. 3.-The parahydrogen conversion on the Pd/Au alloys (points omitted). 4. 30 at. yo Pd. I. Pure Au. 3. 20 #t 6. 55'2 ,, 9. Pure Pd. 7. 70 at. % Pd. 2. 10 at. yo Pd. 5. 40 I, 8. 90 3 ,A. COUPER AND D. D. ELEY r € kcd , \ \ \ \ - \ \ \ '\ I77 0 - 0 ' . FIG. 4.-Activation energy and Bo as a function of composition. The broken line x denotes the paramagnetic susceptibility in arbitrary units (Vogt, Ann.Physik., 1932, 14 ( 5 ) , I). The catalysts fall into two groups in a most striking fashion, as shown particularly in Fig. 3, or in Fig. 4 in which E and log,, BO are plotted as a function of atomic composition. The data in Table I are completely described by the footnotes, but one or two further points are of importance. Thus, as noted by Farkas,26 a palladium wire heated a t 573" K successively in oxygen, hydrogen and vacuum, has a more than usually active surface. The effect is seen t o be due to a decrease in E, and there is also a decrease in Bo below the usual value, which would rule out the view that this treatment increases the surface area of the catalyst. Arguing from our results, described later, on the poisoning effect of hydrogen dissolved in palladium, we suggest tentatively that the effect of the initial oxygen treat- ment is to remove dissolved hydrogen atoms.This would imply that the wires prepared simply by heating in hydrogen, that is, the first three examples, still contained, even after prolonged outgassing, a little (perhaps 0.5 atomic per cent.) dissolved hydrogen. The surface oxygen film seemed t o vary somewhat in its reducibility by molecular hydrogen, from alloy t o alloy. Thus, palladium remained inactive until it was heated t o a t least 320° K in hydrogen, while the oxygen film on 90 yo Pd was removed a t 250' K. 55-2 yo Pd could not be activated by molec- ular hydrogen even a t 850° K and the discharge method was essential for these A paramagnetic conversion was detectable on a gold foil at low temperatures, alloys.and will be discussed in another paper. 26 Farkas. Tvans. Faraday Soc., 1936, 32, 1617.178 PARAHYDROGEN CONVERSION ON GOLD ALLOYS Reaction Order.-Let the reaction vessel, volume V , having its walls a t T, (here go" K) , contain n molecules of hydrogen, and the surface area of catalyst be A cm.2 and contain f adsorption sites per cm.a. We suppose that a fraction u of the sites are occupied, and that k,(k,) are the first-order constants for the change para + ortho (ortho + para). It may be shown that and substituting n = pV/kTV, where p is the pressure of hydrogen gas, ke = ( k , + k,) ! P' Some time ago Eley remarked that the data then available supported the view that (I ot pl/n , the Freundlich isotherm, over wide ranges of pressure.We have now made specific investigations on this point for wire catalysts, and have found that tungsten wires and the 40 yo Pd wire accurately obey the Lang- muir isotherm, so we write a = bP /(I + b p ) and We have tested pure gold and 40 yo Pd a t their average working temperatures of 600" K and 350" K respectively. The gold catalyst gave a truly first-order reaction, that is, k , independent of p so that bp -g I, and the 40 yo Pd catalyst was very nearly first-order at I mm. Hg pressure. Attempts t o obtain a pressure dependency experiment on pure palladium failed for the reason below. Hydrogen-charged Palladium Catalysts.-In Fig. 5 we show the effect FIG. 5.-Variation of conversion velocity on palladium with pressure. 0 1.2 mm. A 6.1 mm.of simply raising the pressure from the standard value of 1-2 mm. t o 6.1 mm. pressure. This effect can only be due to an increased concentration of dissolved hydrogen atoms in the palladium, and so a palladium wire was charged wth atoms produced by the electrodeless discharge until no further solution took place, as judged by measuring the resistance of the wire at go" or 298" K. Thus, in one series the resistance of the wire increased from 2.542 D to 2-766 0 a t 298" K. The data of Sieverts and Danz 28 allow us t o calculate that this wire has a hydrogen content of 8 atomic per cent., though the hydrogen is not necessarily distributed uniformly. It was found possible to carry out a series of parahydrogen con- versions on this wire in the range 290-400" K, since the wire did not lose dis- solved hydrogen a t any appreciable rate. The effect of temperature on k , The result is an increase in activation energy t o 6500 cal.=7 Eley, Trans. Faradny SOC., 1948, 44, 216. Sieverts and Danz, 2. Physik. Chem. B , 1937, 38, 61.A. COUPER AND D. D. ELEY 1 79 at 1-2 mm. Hg pressure is shown in Fig. 6, from which we calculate the ap- proximate values E - 11,000 cal. and BO = 1-17 x 109 cm. min.-l. FIG. 6.-Effect of hydrogen dissolved in palladium upon the activation energy. 0 Pd with dissolved hydrogen. Pd outgassed a t 600' K in vacuo. The applicability of the Sieverts resistance against dissolved hydrogen relationship was confirmed in some solubility experiments on the 70 Pd wire. The solubility effects described above were reversible, but after a certayn amount of this kind of working, irreversible changes in the wire occurred.Discussion The Activation Energy, E.-From Fig. 4 we see that the activation energy for the conversion is about 3-5 kcal. for pure palladium and remains the same for all alloys down to 40 at. yo Pd. Between 40 at. yo Pd and 30 at. yo Pd, the activation energy increases to 8-5 kcal., which is also the value for the 20 at. yo Pd and 10 at. yo Pd alloys, beyond which it incieases to the value of 17.5 kcal. for pure gold. The exact point at which the d-band of palladium becomes filled has an uncertainty to an atom percentage of about 5 yo. The recent magnetic work of Sieverts and Danz 28 on the palladium-hydrogen system would put the point of zero susceptibility at 64 yo, which would agree closely with the results of our catalytic work.Clearly we are justified in associating a low activ- ation energy with the presence of holes in the d-band (free electron theory) or vacant atomic d-orbitals (valence bond theory). It is also clear that a bulk concentration of d-vacancies of 5 atomic per cent. or even less suffices to maintain the activity of the catalyst. There must therefore be a mechanism by which d-vacancies corresponding to a layer 33 atoms thick can be made available to hydrogen atoms adsorbed on the surface. Such behaviour is presumably implicit in the free electron theory, or in the resonance of valence bonds. Catalytic activity, a surface property, is quite different from magnetic susceptibility, a bulk property, which decreases linearly with the concentration of gold or dissolved hydrogen, as shown in the dotted line in Fig.4. The lattice spacing change over 40 to 30 at. % Pd will only be 0.02 A. We may use Eyring's calculaticns 8 to give an idea of the magnitude of the change in activation energy re- sulting from such a change in lattice spacing. In the neighbourhood of optimum spacing the result is 0.01 kcal., and at the worst 0.8 kcal., which is much less than the observed change of 5.0 kcal. The results are a little different for the palladium-hydrogen system in that the initial activation energy for pure palladium of 3-5 kcal. is in- creased to 6.5 kcal. by the solution of a very small amount of hydrogen.180 PARAHYDROGEN CONVERSION ON GOLD ALLOYS Pd 40 % Pd Au An overall concentration of 8 at.yo hydrogen, produced by electrical discharge, gave E - 11.1 kcal., which is larger than the E value for 60 at. yo Au. In the first place we are of the opinion that the hydrogen atoms are concentrated in the outer layers of the wire. Such a skin effect always occurs in the solution of hydrogen in palladium unless special precautions are taken. 2@ Thus, an overall concentration of 8 at. yo in our wire is easily shown to correspond to 64 at. yo in the outermost 30,000 atomic layers. Secondly, it is quite possible that the presence of the /3 phase with relatively large spacing gives rise to an in- crease in E additional to that arising from the filling of the d-band. A non-homogeneous distribution of hydrogen in the wire would explain the disrupting effect of continual solution and removal of hydrogen.It is a matter for some satisfaction that these results are qualitatively similar, both for E and B", to the results recently published by Rienacker and Sarry l7 for the parahydrogen conversion on copper-platinum foils, except that the change in E in their work occurs at 16 at. yo Pt, that is. some way from the 40 yo value, and is less abrupt.* Thus, some of the remarks we make here may be applicable to their work also. The Frequency Factor Bo.-In a previous paper Eley 27 gives a calculation of B" for the exchange mechanism, viz., converFion occurring by an exchange of atoms between a parahydrogen molecule in a van der Waals' layer and a hydrogen atom in an underlying complete chemisorbed layer, indicated by MH.For a dilute van der Waals' layer, as found here, palladium at 297" K and a reaction volume of 400 ~ m . ~ , the calculated value was B = zoo. Thus, the value of B" is zoo x 300 = 6 x 104 cm. min.-l. This value is calculated far f = 1.2 x 1015 sites per cme2 and it is better to take a more exact value off so as to allow for the variation from palladium to gold. Accordingly we use the formula B" = 8.61 x 10-l~ f/T4, which follows from eqn. (20) of ref. 27. We obtain the values in the Table below, taking as f the mean values for the (110) and (100) planes. PHZ + HM-+ MH + OH,, TABLE II.-CALCULATED VALUES OF B" 250 1 1.15 x 10l5 6-26 x 104 -4 x 105 300 1-07 x 1016 5-32 x 104 1-1 x 104 600 1-02 x 1 0 ~ 5 3-58 x 104 2.1 x 104 I I I. The calculated B" values are shown in Fig.3 by a dotted line. The agreement is good, considering the difficult nature of the problem and bears out the conclusions of the earlier work based on data available in the literature. However, later, so far unpublished, calculations show that the Bonhoeffer-Farkas mechanism may possess the same value of B", so that we cannot have a decision on mechanism on this ground alone. In this connection we have to qualify the remarks in an earlier paper &Levels and Catalysis .-The only certain conclusion from the present experiments is that &vacancies are essential in bonding the activated complex of the conversion reaction. It is, however, well worth pcstulating a detailed reaction path so as to take the analysis further. The original exchange mechanism so for the conversion involved 29 Kruger and Gehm, Ann.Physik, 1933, 16 ( 5 ) , 190. * Rienacker and Gaubatz (A.C.S. Abstr., 1941, 35, 3944) gave magnetic data which indicate that the d-band is filled at about 50% Pt in Pt-Cu alloys. 30 Rideal, Proc. Camb. Phil. SOC., 1938, 35, 130. (ref. 2 2 J p. 449).A. COUPER AND D. D. ELEY 181 the adsorption of the hydrogen molecule over empty sites in the chemi- sorbed layer, but we have found no evidence for these sites (cf. remarks in a review 31), so we prefer to formulate the mechanism using a single site in the metal surface. This involves a departure from the usual Langmuir postulate, at least for activated complexes. H H, H- -H- -H Hz H M 'M/ M We write ! + ' \ / ' + 1 and for the equivalent formation of hydrogen deuteride, H D, H- -D- -D HD D There is a little evidence that the d-vacancies, i.e.unpaired d-electrons, are concerned with the initial MH bond. Thus, copper, which has no unpaired d-electrons, has a heat of chemisorption of hydrogen of only g kcal./mole 32 compared with average valuzs of 30 kcal. for tungsten 33 and 24 kcal. for nickel.' The hydrogen on copper, and presumably gold also, will then be held by the d 2 n 5 ' I spz*2z orbitals * relatively weakly became a large part of the bonding power will be concerned with the metal itself. For palladium the hydrogen atom is either bonded by the unpaired &electron, the simplest view to take, or the atomic d-orbitals are used to form a metal hybrid surface orbital containing more d than the usual 2.56.This notion that the atomic d-electrons hold the hydrogen is in accordance with the qualitative evidence that hydrogenation catalysts adsorb hydrogen strongly. The H, activated complex will then be held by a d-bond and a metal hybrid bond to the metal site, and resonance between the three or more canonical states will be fundamental in lowering the energy of the complex. When the d-band becomes full, there may possibly occur a change over to the Bonhoeffer-Farkas mechanism for the conversion, 2MH + 2M + H, when the activation energy will be related to the heat of desorption of the chemisorbed hydrogen as molecules. The B" factor would not neces- sarily show any marked change in value, according to a transition state analysis of the recombination mechanism, to be published later.It seems quite certain that the Bonhoeffer-Farkas mechanism must set in at a certain relatively high temperature for all metals, and measurements of chemisorption on the alloys which we have planned should enable this temperature to be calculated. The value of E = 17.5 kcal. for pure gold requires a little discussion. Unlike the alloys, for gold the log k, against I/T plot is markedly curved, E increasing continuously as T increases. The upper temperature of 800° K for the gold wire is approaching the temperature at which the homogeneous conversion of parahydrogen through hydrogen atoms is observed. In this region we shall have to anticipate yet a third mechanism for the parahydrogen conversion, that is, a reaction in the gas phase set off by hydrogen atoms desorbed from the gold surface, Au-H+Au+H; H + pH, --f OH, + H.Returning now to the exchange mechanism which is the fundamental process at low temperatures, the question whether or not the H, complex 31 Eley, Quart. Rev.. 1949. 3, 209. 32Ward, Proc. Roy. SOC. A , 1931, 133, 506. 33 Roberts, ibid., 1935, 152, 445. * Hereafter called metal hybrid orbital.182 PARAHYDROGEN CONVERSION ON GOLD ALLOYS may be formed will depend largely on steric conditions within the chemi- sorbed layer. Fig. 7 shows that linear and possibly’ triangular complexes are sterically possible on the (110) plane of palladium, but scarcely so on the (100) plane. The linear complex would be analogous to the complex (tto) p/une 0 o(=T-J(-J 000 000 000 n e 00 00 00 00000 FIG.7.-Hydrogen atoms on sites in the IIO and IOO planes of palladium. found in the gaseous reaction.s4 It may be hoped that other simple reactions will fit into a similar picture, and in this connection the least dense lattice planes of nickel and other metals have been stated to be the most effective lattice planes for the hydrogenation of ethylene.a Besides lowering the energy of the activated state, the resonance energy of the initial state may influence the reaction. Following up an earlier suggestion,s5 Eley s1 has shown how the chemisorption of benzene most probably takes place by opening of the double bond. In this way we get a low heat of adsorption due to loss of resonance energy, and a consequent exchange reaction with hydrogen sensitive to benzene pressure and with a low activation energy of g kcal.By contrast, the reaction with ethylene is zero order in the ethylene and has an activationenergy of about 20 kcal. 3-atom complexes are easily formed on the (110) plane. 34 Eyring and Polanyi, 2. physik. Chem. B, 1931, 12, 279. ss Smith and Meriwether, J . Amer. Chem. SOL, 1949, 71, 413.A. COUPER AND D. D. ELEY 183 As pointed out in our note the zo-fold decr2ase in B" from palladium to gold (loo-fold from platinum to copper 17), is in accordance with Dowden's suggestion that the entropy of activation may be influenced by the density of electronic states. In our calculation of B" we, in effect, as- sumed that the ratio of electronic weights in initial and activated states was unity'. Until we have further investigated the much more marked changes in B" that arise from other factors, as in the effect of dissolved hydrogen, we reserve any' further comment on this point. Bond Type in Chemisorbed Hydrogen.-We shall conclude with an explicit statement of our views on the hypothesis 16* 36* 37 that chemi- sorbed hydrogen atoms are ionized on the surface of transition metals.On the theoretical side, our objections are essentially the same as those of Emmett and Teller.38 Let I denote ionization potential, E electron affinity, 4, work function, r the spacing from the chemisorbed ion to the surface, and e the electronic charge. Then there are three possible processes, for which we wish to calculate A E a d s . the internal energy change on chemisorption from hydrogen molecules.D H ~ is the dissociation energy of a hydrogen molecule. W + +H2 -+ W-. . . H + I t;. I ,"v. zp ' (A) eV. Bond. ~ ~ W- . . . H+ 1-41+0 = 1-41 13.59 4.54 - 2-52 W+ . . . H- 1'41$-2*08=3'49 - 4-54 0.7 1-02 W- . . . HZ 1-41+0~60=2~01 16.2 4-54 - 1-78 W + 4H2 -4 W+. . . H- DE2,ev. A E : 4 : * p 4-46 10.99 4-46 7-28 4-46 9-88 W + H 2 -+ W-. . . H i Data and results are listed in Table I11 and from this it is quite clear that all the ionization processes are prohibitively endothermic, involving &??ads. - 10 eV -230 kcal. No assistance is to be expected from the entropy change. TABLE I11 On the experimental side, Bosworth's 38 value for the W-H contact potential of - 1-04 V corresponds to the small dipole of 0.4 D, i.e. about one-tenth of an electronic charge (negative) on the H atom.Oatley's 4O value for Pt-H of + 1-17 V corresponds to a similar small positive charge, on the hydrogen atom. Both bonds are essentially covalent bonds of low polarity, such as a single electron l6 or electron pair lS bond might be. It is our belief that many misconceptions have been based on Coehn's demonstration *l of the positive nature of hydrogen atoms dis- salved in palladium. Calculation, however, shows that the positive charge per hydrogen atom is only one-fiftieth of an electronic charge.42 36 Schmidt, Chem. Rev., 1933, 12, 363. 37 Nyrop, The Catalytic Action of Surfaces, 2nd Edn. (London, 1937). 38 Emmett and Teller, 12th Report Comm. on Catalysis (New York, 1940). *O Oatley, Proc. Physic. SOC., 1939, 51, 318. 41 Coehn and Specht, 2.Physik., 1930, 62, I . 4 2 Duhm, ibid., 1935, 94, 434. p. 68. 39 Bosworth, €'roc. Camb. Phil. SOC., 1937, 33, 394.184 SOME REACTIONS OVER ALLOY CATALYSTS Dissolved Hydrogen as a Catalyst Poison.-The results of this paper help to throw light on a number of obscure points in the literature. Thus, we have Willstatter's 43 observations on the loss of activity of hydrogenation catalysts which might be reactivated with oxygen, and the loss of activity of platinum cathodes with continual discharge of hydrogen.44 Films of platinum and palladium formed by sptttering in vacuo are good catalysts, but formed in hydrogen are quite inactive.46 The active films did not contain occluded oxygen, which argues against the Rogjnsky hypothesis 46 that pure metals are inactive, but may be activated by the merest traces of dissolved oxygen or other gas. Most striking of all are the results of Farkas 27 on the parahydrogen conversion on the inlet and outlet sides of a palladium disc and a palladium tube through which hydrogen is diffusing. The activation energy for the conversion is notably higher on the inlet side where the concentration of dissolved hydrogen atoms will naturally be higher than on the outlet side. This experiment definitely shows that only energy states in a piece of metal in the immediate neighbourhood of the surface can influence the catalytic effect. On this view the activation of platinum catalysts by oxidation, e.g. platinum foil by anodic polarization (Faraday *') or by chromic acid, or platinum black by oxygen 48 is due, at least in part, to removal of hydrogen atoms dissolved or " embedded " in the surface. That the effect of dissolved hydrogen is a complex one is shown by the fact that while E is raised, Bo is also raised powers of ten above the normal value. Abnormally high values of Bo may be associated with the presence of a distribution of sites of activity,27 but further comment is reserved for a future paper. Our best thanks are due to the Chemical Society and the Anglo- Iranian Oil Company for financial help in the purchase of the alloy wires, to the Royal Society for an optical pyrometer, and to the D.S.I.R. for a research grant to A. C. The Chemistry Department, The University, Bristol, 8. Leitz, ibid., 1921, 54, 113. 43 Willstatter and Jacquet, Ber., 1918, 51, 767. 44 Hammett, J . Amer. Chem. Soc., 1924, 46, 7. 45 Bredig and Allolio, 2. physik. Chem., 1927, 126, 41. 46 Roginsky, J . Phys. Chem. U.S.S.R., 1941, 15, I. 47 Faraday, Experimental Researches in Electricity (London, 1914). Willstatter and Waldschmidt- Ablesowa and Roginsky, 2. physik. Chem., 1935, 174, 449.

ISSN:0366-9033    

DOI:10.1039/DF9500800172

出版商:RSC    

年代:1950

数据来源: RSC

摘要:

184 SOME REACTIONS OVER ALLOY CATALYSTS SOME REACTIONS OVER ALLOY CATALYSTS BY D. A. DOWDEN AND P. W. REYNOLDS Received 2nd February, 1950 Analysis of the " electronic factor " in heterogeneous catalysis suggests that activity of Ni-Fe alloys in multiple bond hydrogenation should increase rapidly in the region where the energy density of electron levels a t the Fermi surface rises ; moreover, that reactions controlled by the rate of transfer of an electron from metal to substrate should, in the Ni-Cu alloys, decrease in rate as the 3d-band begins to empty. Experiments on styrene hydrogenation over Ni-Fe catalysts and hydrogen peroxide decomposition on Ni-Cu alloy foils provide general confirmation of these predictions. Methanol and formic acid decompositions on the Ni-Cu alloys decrease in speed as the 3d-bandD.A. DOWDEN AND P. W. REYNOLDS 185 holes are filled, following the pattern of multiple bond hydrogenation over supported Ni-Cu catalysts and further emphasize the role of the electronic factoi- in catalysis. Catalysis at the interface between solids and fluids is a complex phe- nomenon, the energetics of which are functions of a number of para- meters as yet insufficiently defined to provide the industrial chemist with a basis for fruitful calculation. However, it has proved useful to divide the parameters (somewhat arbitrarily) into two sets, one set being called the " geometric factor " the other the " electronic factor ",2 and to seek experimental correlations between the catalytic activity of solids and their other physical properties supposed to depend chiefly on the same factors.Thus the edge of the unit cell of metal catalysts with cubic lattices might be the gauge of the geometric factor and the magnetic properties the measure of the electronic factor. The early qualitative success of covalent bond theory as applied to the study of the activated complex,a and the newness of the theory of the solid state, resulted in most emphasis being placed on the geometric factor almost to the complete neglect of the effect of electron characteristics. have re-established the practical value of the electron characteristics of solids as a guide to their catalytic activity whilst semi-empirical theory has provided a firmer footing for future endeavour. According to a new view 6 9 7 the activity of a metal depends on the values of the electronic exit work function 8, the energy density of electron levels at the Fermi surface [g(E),,-,] and the gradient of the latter [dg(E) /dE],,-6 = G.Reactions proceeding at a velocity affected by the rate of formation or the concentration of a chemisorbed positive ion or radical appear to be favoured by an increase in the value of each of these (when G is positive). When negative ions assume the dominant role reaction velocity should increase as 0 and g ( E ) decrease and when G decreases (especially to negative values). In particular, for the series of binary solid solutions containing one of the group 8 elements (Ni, Pd or Pt) together with one of the Group I elements (Cu, Ag or Au) and the binary alloys of iron, cobalt and nickel, the efficacy in multiple bond saturation or in dehydrogenation depends largely on the number and the characteristics of the holes in the d-bands of the alloys.The activity diminishes as the number of holes falls to zer0.*~8 For instance8 the specific activity of a series of supported nickel-copper catalysts in styrene hydrogenation at 20' C decreased from a maximum at pure nickel to zero at the alloy of approximately the equi- atomic composition ; this fall was paralleled by the gradual decline of ferromagnetic properties to zero at the same composition. Now the low-temperat ure electronic specific heat data of Keesom and Kurrel- meyer * indicate, not only a sharp fall in g ( E ) at about the equi-atomic composition in the nickel-copper series of solid solutions but a sharp rise in the iron-nickel series somewhere between the binary alloys of com- position 80 Ni + 20 Fe and pure nickel.Theory suggests therefore that activity in multiple bond hydrogenation over the nickel-iron alloys should Recent researches 4s 5 s 1 Griffith, Recent Advances in Catalysis, Vol. I (Academic Press Inc., New Roginsky and Schultz, Z. physik. Chem. A , 1928, 138, 21 ; Russell, Nature, Sherman and Eyring, J . Amer. Chem. SOC., 1932, 54, 2661. Rienacker, 2. anorg. Chem., 1938, 236, 252 ; ibid., 1939, 242, 302 ; ibid., 5 Dowden and Reynolds, Nature, 1949, 164, 50. York, 19481, p. 91. 1926, 117, 47. 1941, 248, 45 ; J . prakt. Chem., 1941, 158, 95. Couper and Eley, Nature, 1949, 164, 578. Dowden, Research, 1948, I, 239 ; J .Chem. SOC. (in press). * Reynolds, J . Chem. SOC. (in press). 9 Keesom and Kurrelmeyer, Physica, 1940, 7, 1003.186 SOME REACTIONS OVER ALLOY CATALYSTS increase rapidly at some point in this composition range. By way of contrast the decomposition of hydrogen peroxide, on the basis of the Haber-Weiss lo. l1 mechanism for the ferrous ion catalyzed reaction, should involve electron transfer with anion formation according to the equation H,O, + metal electron - OH- + OH and requires that the decomposition rate over alloy catalysts should de- crease as holes appear in the d-band. This paper presents experimental results on styrene hydrogenation over supported nickel-iron catalysts and methanol, formic acid and hydrogen peroxide decomposition over nickel-copper foils, which are in general accord with these conclusions.Experimental Undoubtedly the best metal surfaces for catalytic work are those prepared by evaporation in vacuo.lz The difficulty of preparing homogeneous alloy films of this type leads instead, in these preliminary studies, to the choice of metal foils and supported metals closer to the orthodox, industrial model. Reactants : HYDROGEN was prepared by electrolysis and purified by diffusion through a palladium thimble preceded and succeeded by liquid nitrogen traps. Hydrogen peroxide and methanol were both A.R. grade, the latter very low in sulphur content (3 parts per 1 0 6 ) . Pure styrene was vacuum- distilled, made up in methanol solution and treated with activated Raney nickel t o remove residual poisons.A go yo solution of pure formic acid was dried with phthalic anhydride and distilled to yield a formic acid feed containing only 0.2 yo of water. NICKEL-IRON CATALYSTS.-The nickel and iron were co-precipitated as carbonates upon a specially purified kieselguhr (specific area - 5 m.2 g.-l) using aqueous solutions of the pure nitrates and ammonium carbonate ; each catalyst contained a total weight of metal equal t o 15-5 yo of the weight of kieselguhr present. The products were dried, calcined for 4 hr. a t 400' C , sieved through a IOO B.S.S. mesh and small lots (- I g. in thin layers) reduced by pure hydrogen during 40 hr. a t 500' C. All measurements were done on the reduced catalysts which were kept either under pure hydrogen or for very short times under the de-oxygenated liquid substrates.Specific areas were estimated by the Brunauer-Emmett-Teller l3 method with argon (area 14-4 x 10-l6 cm.2) as the sorbate. X-ray examinations confirmed the presence of metal crystallites with face-centred cubic structure and showed the crystal size dis- tribution t o be similar in each catalyst. Specific activities were therefore calculated using the specific areas from gas absorption. NICKEL-COPPER FOILS.-Nine alloys, which together with the pure metals covered the whole composition range a t 10 yo intervals, were prepared by Johnson Matthey and Co. Ltd. from spectroscopically standardized nickel and copper ; they and the pure metals were rolled into foils of 0.003 cm. thickness spectro- scopic analysis of which showed only faint impurity lines of silver, calcium, silicon, magnesium, iron and cobalt.The area of each foil was measured by a modification, due to Wooten and Brown,14 of the Brunauer-Emmett-Teller l3 method using ethylene (area 17-55 x 10-l6 cm.2) as the substrate. It was found that the area by gas adsorption was only 85 yo of the geometric area in accord with some recent observations of Davis, De Witt and Emrnett.ls However, this factor was the same for all the foils and since microscopic methods showed little variation in grain size over the series, the geometric area was used to obtain the specific activity. The magnetic susceptibilities of the copper-rich alloys were obtained with a Gouy balance and checked, in part, with a Suck- smith balance. Normal cleaning of the metal surfaces by treatment with pure hydrogen followed by extensive outgassing, both at 500' C, resulted in the deposition of a metal film on the walls of the Pyrex container.This treatment was in con- sequence abandoned in favour of a high frequency discharge in hydrogen lo Haber and Weiss, Proc. Roy. SOC. A , 1934, 147, 332. l1 Weiss, Trans. Faraday SOC., 1935, 31, 1547. I* Beeck, Smith and Wheeler, Proc. Roy. SOC. A , 1940, 177, 62. l3 Emmett, Brunauer and Teller, J . Amer. Chem. SOC., 1937, 59, 1553. l4 Wooten and Brown, ibid., 1943, 65, 113. l6 Davis, De Witt and Emmett, J . Physic. Chem., 1947, 51, 1232.D. A. DOWDEN AND P. W. REYNOLDS 187 (- 2 mm. Hg) prolonged intermittently for 12-24 hr. in a system containing a liquid nitrogen trap and alternating with pumping out a t I O - ~ mm.Hg. The sole internal check of the success of this technique lies in the reproducibility of the results so obtained. Foils used in the decomposition of hydrogen per- oxide were pre-treated by immersion in 5 yo nitric acid for 30 sec. followed by thorough washing with distilled water. As used the foils all possessed their bright metallic sheen without sign of thick (> 300 A) oxide film formation. Decomposition of Hydrogen Peroxide and Hydrogenation of Styrene .- The rate of evolution of oxygen from neutral hydrogen peroxide (H,O, = 0.058 g . ern.-.) and the rate of adsorption of pure hydrogen by styrene in methanol (10 ~ m . ~ of a 10 yo solution) were followed with a soap-film flow-meter at atmo- spheric pressure and temperatures of zoo C t o 80" C in a static system shaken at high speed (2000 c./min.). The hydrogenation over -0.1 g.catalyst was observed at the maximum shaking rate where the reaction velocity was almost independent of this rate. The decomposition of hydrogen peroxide over the foils (area 16 cm.2) showed for a given foil a maximum a t an intermediate shaking rate which coincided with that giving the best contact, assessed visu- ally, between foil and substrate ; the activity of each foil was measured a t its optimum shaking rate. Decomposition of Formic Acid and Methanol.-The decomposition of the vapours of formic acid and methanol over nickel-copper alloy foils was effected at atmospheric pressure in an apparatus similar to that described by Schwab l6 but modified to permit pre-treatment of the foils with hydrogen in the electrode- less discharge and to permit vacuum out-gassing.The flow of reactant vapours was maintained at a constant value (go ~ m . ~ sec.-l) for a foil area of 40-120 cm.2. and the exit gases, substantially carbon monoxide or carbon dioxide and hydrogen measured with a soap-film flow-meter. The reactor temperature could be varied from 20" C to 520" C as recorded by a thermocouple, sheathed in thin glass and placed in contact with the foil. Blank runs gave small gas rates which necessitated very minor corrections to the rates of the metal catalyzed reactions. Results Styrene Hydrogenation.-Fig. I contains the specific activity of the nickel-iron catalysts, in styrene hydrogenation a t 20' C plotted against the ZIG.effect of nickel content on the rate of hydrogenation of styrene by alloy catalysts. Curve A : Ni-Fe catalysts, H, uptake ~ m . ~ min.-l cm.-2, x 104. Curve B : Ni-Cu catalysts, H, uptake ~ m . ~ min.-l cm.-2, x 104. Curve C : Number of holes per atom in the +band. Curve D : Coefficient of the electronic specific heat term ; X 103, cal. mole-1 deg. -2. atomic fraction of nickel in the metal crystals. Each point is the mean of a t least ten tests with fresh catalysts and the vertical line represents the probable Schwab, J . Physic. Chevn., 1946, 50, 427.188 SOME REACTIONS OVER ALLOY CATALYSTS error. The pattern is completed by inclusion of results for a'Forresponding series of nickel-copper catalysts taken from an earlier paper and t o give the pure nickel members of both series the same specific activity.Also depicted is the variation with nickel content of the number of holes per atom in the alloy 3d-bands 17, and the density of electron levels g ( E ) , taken t o be proportional to the electronic specific heat c,oefficient according t o Keesom and Kurrelmeyer. Separate experiments showed that hydrogenation stopped a t the ethyl benzene stage. Hydrogen Peroxide Decomposition.-Fig. z shows the specific activity of the nickel-copper foils. Each point is the mean of a t least ten experiments with new foils and the vertical lines indicate the probable error. normalized FIG. 2.-Specific activity of Ni-Cu alloys in hydrogen peroxide, formic acid and methanol decomposition. Curves A, B, C : 0, evolution from H,O, a t 80" C, 70" C and Go" C respectively, cm.3 min.-1 cm.-2.Curve D : Total CO, + H, exit gases from HCOOH at 253" C ~ m . ~ min.-l cm.-a. Curve E : Total CO + H, exit gases from CH30H a t 253" C ~ m . ~ min.-l cm.-2, Curve F : Magnetic susceptibility of the foils in arbitrary units. x 104. The rates recorded are the constant values observed immediately after the attainment of steady shaking (within 15-30 sec. of zero time) and while the foils retained their pristine sheen without change of colour. In runs prolonged to 15 min. or more, the rates frequently showed a slow rise from the initial constant values but this was always accompanied by a loss of lustre and the appearance of a dirty red-brown colour indicative of thick oxide films ; such effects have been neglected as not representative of metal or metal plus the thinnest films.Omission of the pre-treatment with dilute nitric acid did not change the character of the results but caused some loss of reproducibility. The activity a t 60" C and 70" C of alloys containing more than 40 yo of nickel was too small t o estimate. The measured magnetic susceptibilities are given in arbitrary units ; a small field-dependence due to ferromagnetism, is already apparent a t 30 at. yo nickel and becomes greatly magnified with increasing nickel content. The apparent activation energies (kcal. /mole) and frequency factors (mole- cules sec.-1 cm.-2) are respectively for pure copper 10.3, 1.0 x 1oZ3 ; g o at. % copper, 10-9, 2-1 x 1oZ3; 80 at. yo copper, 11.0, 2.4 x 1 0 ~ ~ ; 70 at.yo copper, 23-7, 1.3 x 1 0 ~ ~ . Formic Acid and Methanol Decomposition.-Included in Fig. z are the specific activities of the foils in the decomposition of the vapours a t 253' C . Each point on the formic acid curve is the mean of six determinations on the same foil with intermediate treatment in the hydrogen discharge and the probable error is shown. The activation energy of formic acid decomposition (250' C-300" C) varies from 30 kcal./mole over pure nickel to 36 kcal./mole over pure copper but although our variation is of the same kind as found by Rienacker4 in a static system, our values are approximately twice as great suggesting poisoning of our foils.189 D. A. DOWDEN AND P. W. REYNOLDS Discussion Since there are no phase changes or magnetic transitions at room temperature in the face-centred nickel-iron alloys covering the range 3oNi + 70Fe to pure nickel, Fig.I shows quite clearly, in accord with expectation, the general parallelism between the ra.te of hydrogenation of styrene and the variation of electron-level density g(E). Together with the earlier results in the nickel-copper series, it yields the complete activity pattern as the Fermi surface approaches and passes under the peak of g(E) in the overlapping 3d- and 4s-bands and emerges into the 4s-band. In the nickel-iron alloys a rising g ( E ) accompanies an almost linear decrease in the number of d-band holes but in the nickel-copper alloys g ( E ) remains almost constant, until the holes are quite full at the critical composition (0.61 atomic fraction of copper) and then decreases rapidly.The rise in activity between the alloy 75Ni + 25Fe and pure nickel must therefore be attributed to the great increase in g(E), whereas the decline between pure nickel and the alloy 40Ni + 60Cu is due both to the disappearance of 3d-band holes and to the fall of g(E). The order- disorder changes which are only sluggish at 500' C around the composition Ni,Fe do not appear to affect the activity since the activity curve possesses no symmetry with respect to this composition. Fig. 2 shows that the decomposition rates of formic acid and methanol vapours over Ni-Cu alloys decrease as the +band holes fill up and the trend is quite similar to that for styrene hydrogenation. On the other hand, the rate of hydrogen peroxide decomposition falls off as the 3d- band begins to empty, as is to be expected for a process controlled by the rate of transfer of an electron from the metal to the substrate.The correlation between the electron characteristics of metals and their catalytic properties depends inevitably on the physical data avail- able. Especially important to the topics of this paper are the values of 8, g(E) and G, whose variations across the two alloy series are known in form but not in detai1.l' Simple band-theory, using early values for the saturation magnetization of the ferromagnetic alloys, gave a clear- cut picture of the d-band holes in the nickel-copper solid solutions and although recent studies l* of associated phenomena are not always in exact accord with this model, the divergences are not sufficient to invalidate our conclusion.Again, it is possible that the number and properties of d-band holes differ in the bulk and the surface but if so the effect is not sufficient to disturb significantly the relationship with bulk properties. Beeck and co-workers,19 in a purely geometric interpretation of ethylene hydrogenation by pure metals, correlate activity with crystal parameter (closest spacing in the plane of least packing density) and find 3-75 A to be its optimum value. For the face-centred cubic alloys of nickel here examined, this suggests an increase in activity as iron and copper are alloyed with nickel in direct contradiction of the results given. Many re- actions, 49 2 0 p * not only multiple-bond saturation, appear to depend on the presence of d-band holes in approximately the same way and the pattern, viewed broadly, is independent of any particular interatomic distance in the substrates.The effect of intramolecular environment on the rate of a catalyzed reaction a t an unsaturated group is illustrated by the styrene hydrogen- ation. Under mild conditions the reaction proceeds only as far as ethyl- benzene ; more severe treatment is required to produce ethylcyclohexane, yet each step in the reaction is favoured thermodynamically and exhibits 17 Mott and Jones, Properties of Metals and Alloys (Oxford Univ. Press, 1936), 1st ed., chap. 6. l8 Stoner, Reports Prog. Physics, 1946-47, I I , 43. l9 Beeck, Rev. Mod. Physics, 1945. 17, 61. so Tammann, 2. anorg. Chem., 1920, 1x1, go.190 SOME REACTIONS OVER ALLOY CATALYSTS the same dependence of rate on the holes in the d-bands (ethylbenzene can hardly differ from benzene % * in this respect).From the geometric standpoint the resistance of the aromatic ring to hydrogen addition might be associated with differing multiple-bond lengths ; electronically the resonance energy could be the significant factor. In truth only the sum- mation of such influences can represent the facts. Although the results on hydrogen peroxide decomposition are in har- mony with the Haber-Weiss lo* l1 mechanism, i t is necessary to consider an alternative formulation of Bray and Gorin 21 for the initiation of the ferrous ion catalysis, i.e. Fe2+ + H20, - FeOa+ + H,O. Adapted to alloy catalysts this becomes metal + H202 - (metal, 0) + H20, where the symbols in parentheses denote a chemisorbed complex. Cer- tainly thin oxide films must be formed on the alloys during reaction and both theory and the work by Pilling and Bedworth,22 on the high-tem- perature oxidation of nickel-copper alloys, suggest that the uptake of oxygen should be most rapid for the copper-rich alloys.However, on completion of the oxide film, the short-range forces of the covalent bonds in the ferroxyl analogue will depend only on the oxide layer, not on the base metal, and the apparent dependence on the electron structure of the alloy must be accounted for. None of the recorded phase bound- aries 23 in the nickel oxide-cupric oxide system can account for the changes in activity, moreover Hickman 24 has found by electron diffraction that the low temperature oxidation of nickel copper-alloys produces only cuprous oxide and cupric oxide in the oxide layer over a wide range of alloy composition. Thus the use of the Bray and Gorin mechanism for the alloy catalysis requires one or more postulates : (a) the oxide film is not complete, (b) there is a change in the structure or composition of the oxide film as the 3d-band begins to empty, (c) the electron levels in the oxide film undergo some systematic change parallel to that occurring in the metal; none of these appears to be true. The electron-transfer process of the Haber-Weiss mechanism is not likely to be greatly affected by a thin oxide film. Formic acid decomposition might be formulated in several ways since its behaviour over the alloy series suggests electron donation to the metal (cf. Schwab),25 HCOOH - HCOOH+ + e, HCOOH+ - CO,+ + Ha and C02+ + e - C02, HCOOH + H+ - H, + HCOOf and HCOO+ - CO, + H+, followed by or reaction with a positive ion, e.g. together with its free radical counterpart. Imperial Chemical Industries Limited, Billingham Division, Billingham, Co. Durham. 21 Bray and Gorin, J. Amer. Chem. SOC., 1932, 54, 2124. 92 Pilling and Bedworth, Ind. Eng. Chem., 1925, 17, 372. *s Rigamonti, A.C.S. Abstr., 1947, 41, 7192 ; unpublished work from this z4 Hickman, Metals Tech., 1948, 15, (Tech. Publ. 2483). 25 Schwab, Trans. Faraday Soc., 1946, 42, 689. *6 Peschard, Compt. rend., 1925, 180, 1836. lab oratory.

ISSN:0366-9033    

DOI:10.1039/DF9500800184

出版商:RSC    

年代:1950

数据来源: RSC

摘要:

GENERAL DISCUSSION GENERAL DISCUSSION Dr. D . D . Eley (Bristol) (partly communicated) : Firstly, I shall give new evidence in favour of my assumed mechanism. Dr. Couper and Mr. Hulatt in my laboratory have carefully examined the effect of pressure on the first-order constant K, of parahydrogen conversion on wires of W, Pd-Au and Pt. IjK, = A $- Bp meaning that the adsorbed reactants are obeying the Langmuir isotherm, which on the simplest view means that they are being adsorbed on a uni- form surface. The results fit over the whole range of surface fractions covered o to I. The effect of temperature on the constants A and B for Pt leads to a value of 2300 cal. for the heat of adsorption uniform from fl = o to I. This is rather larger than the value of 1000 cal. cal- culated by Lennard-Jones in 1932 for the van der Waals’ adsorption of H, on Cu. But it does suggest a second layer held by predominantly dispersion forces on top of the primary’ layer held by predominantly exchange forces1 It is indeed difficult to see how the Trapnell-Rideal mechanism can give a Langmuir isotherm for the hydrogen atoms over the whole range of surface covered with apparent heat of 2300 cal., yet this is what would be required by the above results. The interchange mechanism was based on the observations that on and Ni3 the conversion had a rate closely similar to the rate of atomic All these data satisfactorily fit the equation exchange between gaseous D , and pre-adsorbed hydrogen, Again it was found4 that a considerable part of the chemisorbed oxygen film on W may be removed, corresponding to a fall in the negative value of the contact potential irom 1-76 to 1-22 V, before any parahydrogen conversion sets in.Both these results are easily reconciled with these interchange mechanism but not, I think, with that advanced by Trapnell and Rideal. I might also refer to the result that on a surface partially poisoned by oxygen the rate of conversion is reduced in proportion ap- proximately as the amount of chemisorbed exchangeable hydrogen. For these reasons the interchange mechanism is still in my view the probable mechanism for the parahydrogen conversion for cases such as W at low temperatures. An alternative explanation of the loosely ad- sorbed hydrogen is that it is held in a second layer, but the published details of the experimental methods and results are insufficient to make a discussion worthwhile at present.If we assume that the surface orbitals are the ordinary metallic orbitals one can a t least see the possibility of a small amount of exchange interaction between the metal and a second adsorbed layer, which might account for the extra few kcal. of energy beyond that expected for strictly dispersion forces. Dr. B. M . W . Trapnell (RoyaE Institution, London) (communicated) : Dr. Eley has unfortunately only provided us with quantitative heat data for platinum, and it is possible that hydrogen chemisorption by platinum is so strong even at high coverages that it is effectively irreversible, in which case the interchange mechanism would be the more important.With tungsten, the experimental result indicating the impossibility of the interchange mechanism is that the reaction order is between 0.1 and 0.5 at -rooo C and I mm. If this is to be interpreted in terms of varying coverages in a van der Waals’ adsorbed layer, use of the Langmuir isotherm shows that the fraction covered must lie between 0.45 (order 0.5) and 0.8 (order 1.0). Now for an adsorbed layer of a small molecule D g + H-M -+ M-D + HD. Trans. Faraday Soc., 1929, 25, 447. Eley, Proc. Roy. SOC. A , 1941, 178, 452. A. and L. Farkas, J . Amer. Ckem. SOC., 1942, 64, 1594. 4 Eley and Rideal, Proc. Roy. SOC. A , 1941. 178, 429.1 92 GENERAL DISCUSSION to be so densely packed under these conditions the heat would have to be some 7000 cal. My results show, however, that at this heat, the adsorbed amount corresponds to considerably less than I atom per site, so that there is no second layer present at all, let alone a concentrated one.Even at 2900 cal. heat, the adsorption cannot exceed one atom per site by more than a few per cent. We must conclude that the fractional order or reaction is not explicable by the interchange mechanism : it is, of course, readily explicable by the older mechanism. Provided the heats of chemisorption of H, and D, are not very different, the older mechanism would predict near equality of conversion and exchange rates, as the expression Eley’s other objections are less easy to understand. K ( I - e) R=- 1 +. would be nearly the same in the two cases. And as regards the experi- ments on oxygenated wires, we cannot, I think, discuss these until we know whether the oxygen was mobile under the conditions of evaporation, the heat of chemisorption of hydrogen into a partial oxygen layer and so on.Sir Charles Goodeve (B.1.S.R.A ., London) (contributed) : A useful approach to this problem of chemisorption and surface reaction lies through a consideration of metallic solutions. Hydrogen and nitrogen, in par- ticular, are soluble to some extent in most metals and exist in solution as single atoms (or ions) lying interstitially in the metallic lattice. In- deed, in some metals the concentration of the non-metallic element can be increased up to a point or points, where phase changes or shifts in the lattice dimensions or a slight rearrangement take place.These new phases are commonly called hydrides or nitrides, despite the fact that the re- sulting phases are essentially metallic ; e.g. they have metallic electrical conduction. The high heats of solution (of the atom) signify chemical bonding, but the high diffusion coefficients of these atoms compared with those of other atoms in the solid metallic lattice, show that this bonding energy is related to the solid state, rather than to a position, or to a phase. The whole range of compositions can be looked upon as states of solid solution. I know of no evidence so far, that the state of chemisorption of these elements on a surface differs in any way from that of solution except in so far as the different co-ordination number of an atom just at the surface may lead to a lower, or a higher energy level.In other words, if one breaks a metallic crystal containing hydrogen or nitrogen in solution, the atoms of these elements so disclosed on the new surfaces will be in the same state as atoms produced by chemisorption, except that their surface concentration may be greater or less. that more attention should be paid to penetration from the chemisorbed state into the substrate. Trapnell’s contention that this can be ignored in the para- hydrogen conversion on tungsten should be supported by quantitative data on the solubility and/or the diffusion coefficient of hydrogen in tungsten showing that these are very low at the particular temperatures used. Another consequence of this approach is the conclusion that the lateral mobility of chemisorbed atoms must be as great as, and probably many times greater than, the diffusion rate of these in solution in the metal. Finally, experimental studies of surface reactions based on this ap- proach can lead to a much more rigorous comparison with theory.In the simplest of heterogeneous catalytic reactions one has at least three processes in series, the adsorption, the surface reaction and the desorption, with the resulting latitude in fitting in a particular theory. If, however, Such a picture gives strong support for Beeck’s reminder 5 This Discussion.GENERAL DISCUSSIOK 193 one chooses conditions whee the reservoir of the reactants is in solution and the rate of diffusion is known to be high enough to maintain the surface concentration at a fixed level, then a very much simplified situation results.with a solution of nitrogen in iron (€-iron nitride) which when placed in a vacuum (or in an atmosphere of nitrogen) at 350 to 50oOC gives up its gas at an easily measurable and reproducible rate. The reaction is found to be bimolecular with respect to K concentration in the solution and to have a constant energy of activ- ation of 42 kcal. There is ample evidence to prove that diffusion to the surface is much faster than the over-all liberation of gas. From a knowledge of the behaviour of N atoms in the crystal we know that there is a repulsion between the atoms and that this rises rapidly when two approach to a distance less than 5 A. In order for a physically adsorbed N, molecule to be formed the N-N distance has to come nearly down to I L4 and the Fe-N distance of 1-94 A must be nearly doubled.It is, therefore, consistent with our general knowledge of energy-spatial relationships to associate the 42 kca!., with the peak of a barrier between the chemical and physical adsorbed states which will lie somewhtre near a N-N distance of 1-5 A. What is much more important is that, on the assumption that the heat treatment of the particles used gave smooth surfaces and that the nitrogen atom concentration on the surface is the same as in the bulk, a calculation of the expected surface reaction velocity agrees well (within a factor of two) with the observed velocity. Such agieement in experiments of this type is probably fortuitous, particularly as the theory ignores the effect of surface contamination, say, of oxide.The agreement does, however, suggest that experiments should be done on other systems using metals with hydrogen or nitrogen where similar simple conditions may lead to further tests of theory. It might even be possible to deduce an expected value of the energy of activation from force constants of the N-N and the Fe-N link. At all events, further experiments should show how far this approach can be usefully followed. Dr. B. M. W. Trapnell (Royal Institution, London) said : I should like to discuss two points in connection with my paper.7 The first is the suggestion by Dr. Beeck that the reversible uptakes of gas which we have observed are not chemisorption, as we have claimed, but are solution, which is known to proceed with a low heat.The suggestion implies that the conversion might proceed through evolution of dissolved gas rather than chemisorbed gas. Now we have investigated the solution of hydrogen in tungsten films, and have found that the dissolved amount is so small that the reversible uptakes of gas must almost entirely be chemisorption. Moreover, if we consider the rate of evclution of gas from surface and bulk under specified conditions, clearly the rate of evaporation of chemisorbed hydrogen will be much larger than the rate of evolution of dissolved hydrogen as the solution process requires an appreciable energy of activ- ation whereas chemisorption is non-activated, and because the chemi- sorbed amount is some 20 times the dissolved amount. The conversion therefore will proceed almost entirely through the reversibility’ of chemi- sorption, as we have said.Secondly, it should be noted that the 8 values of Fig. I were derived assuming that when the heat of adsorption is zero, 8 = I. This now appears a rather artificial method of procedure, and in the detailed pub- lication of the work, oxygen chemisorption will be used as the basis of determination. Dr. 0. Beeck (Emeryville, California) said : While I am inclined to Such experiments have been done The new method alters none of our conclusions. Goodeve and Jack, Faraday SOC. Discussions ~ 9 4 8 , 4, 82. 7 Rideal and Trapnell, this Discussion. r\1 94 GENERAL DISCUSSION agree that the probable mechanism of the para hydrogen conversion by tungsten surfaces proceeds through condensation and re-evaporation of chemisorbed gas, I believe that this process is still somewhat more com- plicated than the authors assume.that on many metal films, in- cluding tungsten, hydrogen will be dissolved to some degree in the metal structure, after the surface is covered with a monolayer, and that this solution is of the activated type with a heat of sorption considerably lower than the lowest heat of adsorption. This phenomenon will even take place at very low temperatures and is most likely responsible for the para-ortho hydrogen conversion or the hydrogen-deuterium exchange a t low temperature. Solution of hydrogen can be shown to be responsible for the easily reversible part of the hydrogen isotherm on nickel and can be shown to take place even if the surface is irreversibly covered with carbon monoxide.By subtracting the hydrogen isotherm on a carbon monoxide covered nickel surface from the isotherm obtained with hydrogen alone, the resultant true adsorption isotherm is completely flat over the pressure range from I O - ~ to 10-1 mm. While the mechanism for solution is not clear in detail as yet, it must be assumed to proceed through a second layer adsorption on that very small part of the surface where the heats of adsorption are the lowest. Taking this view, the ideas of Rideal and Trapnell as w-ell as those of Couper and Eley may in fact be very closely related. Dr. P. A. Wright (Bedminster Smelting Works, Bristol) (communicated) : A possible explanation of the difference in surface coverage achieved in the experiments of Roberts and of Rideal and Trapnell lies in the 8 yo of sites left vacant after random covering of the surface by H, molecules.If we imagine that these sites are filled by H, molecules lying “ end on ”, we shall then achieve a surface coverage of “ 108 ” yo ; this latter addi- tion would be expected to yield a smaller heat of adsorption and conse- quently some degree of mobility. Dr. F. S. Stone (Bristol) (partly communicated) : With reference to Eucken’s observations on the reaction of cyclohexene with adsorbed hydro- gen, it is of interest to quote details of a parallel case of the time-dependence of the adsorbed state of oxygen. Carbon dioxide can only be taken up on our copper oxide catalyst provided adsorbed oxygen is present with which it can form a CO, c~mplex.~ If oxygen has been pre-adsorbed, carbon dioxide will subsequently adsorb to give a ratio of 0-7 C0,/1 0, on the adsorbent.If a I/I mixture of CO, + 0, is admitted, the gases are taken up in a I/I ratio, and if a 2/1 mixture of CO, + 0, is admitted, the gases are taken up in a 1-5/1 ratio. With mixtures in which the CO, is in large excess, e.g. a 6/1 mixture, the uptake ratio is 2/1, indicating total capture of oxygen by CO,. It therefore appears that quite a large fraction of the oxygen adsorbed loses its ability to react with CO, in a very short time, since merely by increasing the ratio of CO,/O, admitted the percentage of oxygen reacting increases sharply. A similar loss of yield is found in the reaction of CO with adsorbed oxygen.In neither case, however, does the total loss exceed 60 Yo even over long periods of time. This behaviour is analogous to that reported for oxygen on porous nickel films.1° In the present case the adsorbed oxygen is lost either by oxidation or by diffusion into the porous interior: indeed, the uptake of oxygen on cuprous oxide,9 manganese oxide l1 and nickel oxide,12 obeys a diffusion equation. Prof. W. E. Garner (Bristol) said : The nature of the adsorption of hydrogen on nickel appears to vary with the method of preparation of It has now been conclusively shown a Beeck in Advances i~ Catalysis, Vol. 11. Garner, Gray and Stone, this Discussion. Beeck, Smith and Wheeler, Yroc. Roy. SOC. A , 1940, 177, 62. l1 Sakai, Kurimura and Okura, quoted in ref.( I ) . l2 Roginsky and Tsellinskaya, J . Phys. Chem. U.S.S.R., 1947, 21, 919.GENERAL DISCUSSION I95 the nickel. Euchen finds that the heat of adsorption on metal prepared from oxide is 6 kcal. a t IOO-ZOO~ K and is 17 kcal. above 300° K, whereas Beeck finds a heat of adsorption of 30 kcal. at -183" K. A possible ex- planation of these discrepancies is that there are two types of atomic hydrogen films, one formed within the nickel lattice and the other outside. The formation of a hydrogen film within the lattice is supported by the negative surface potential of atomic hydrogen films. It is suggested that the presence of foreign atoms of greater dimen- sions than the metal atoms, e.g. oxygen, in the surface of the nickel causes a compression of the atoms in the surface of the metal over considerable areas, thereby hindering the penetration of the surface by the hydrogen atoms.Thus in Euchen's experiments at low temper?ture, the hydrogen film is on the outside of the lattice, and on raising the temperature, this passes into the surface. In Beeck's films, being relatively free from foreign atoms, there is no hindrance to the formation of the internal film even at liquid-air temperatures. Mr. A. E. J. Eggleton (Imperial College, London) (communicated) : Using an apparatus l3 similar to that of Roberts, the accommodation co- efficient a of neon on iron has been determined. After flashing the wire (0.05 mm. diam.) to I Z O O O C in vacuo or in neon at a pressure of ca. 0.1 mm. Hg for periods up to I hr. a was 0.2 and there was no drift with time after the thermal disturbance due to flashing had died away (8 min.) as is the case with tungsten.(Clean metal wires, using Roberts' technique, have values of a of 0.06; when contaminated by adsorbed gas a is normally > 0 . 1 5 . ) Admission of hydrogen to a pressure of I x I O - ~ mm. had no effect on the value of a (ct. clean W). Flashing the wire at lower temperatures gave higher va!ues of a, up to 0.29. In one experiment the wire, after preliminary outgassing, was heated a t IIOOOC for 45 min. in 0.1 mm. of hydrogen, left overnight at 7ooOC i n uacuo, followed by 20 min. at 1200~ C in z x I O - ~ mm. H, and a further 30 min. in uacuo and finally 20 min. at IZOOO C in neon which had been left standing over charcoal immersed in liquid nitrogen for I hr.(The hydrogen was circulated past the wire and through liquid nitrogen traps during the reduction.) After this treatment sufficient of the iron had evaporated to form a visible film on the walls of the tube surrounding the wire but a did not fall below 0.2 and again admission of hydrogen had no effect. Heating the wire to higher temperatures or for longer periods was not possible since the evaporation from the iron wire seriously alters the physical constants of the wire. In view of these results compared with those of Roberts and my own results on tungsten it seems improbable that a clean iron surface can be prepared by reduction of iron oxide with hydrogen or by outgassing of iron powders at high temperatures in uacuo and that workers who have used this method for preparing iron surfaces for adsorption, are in fact dealing with a surface that is already contaminated.More recently a clean surface has been prepared by subjecting the iron wire to positive ion bombardment in a gas discharge.l4 Treatment with positive neon ions (1000 V and I mA) reduced the value of a from 0.2 to 0.16. However, if the wire was first subjected to a discharge in nitrogen for 15 min. and then to a neon discharge for z min. the value of a was reduced to approximately 0.07. There was a small drift with time in- dicating slow contamination. Admission of a small amount of hydrogen caused a to rise to approximately 0-12. These results are of some interest in connection with the remarks of Prof. Garner. It seems most likely that failure to clean the iron wire by flashing treatment alone was due to the rapid diffusion of impurities Heating the wire in hydrogen before flashing had no effect. 13 Proc.Roy. SOC, A , 1930, 129, 146. Cp. Mitchell, this Discussion (Introductory paper).196 GENERAL DISCUSSIOK from the interior of the wire to the surface after the cleaning treatment. Now it is suggested that the treatment with positive nitrogen ions forms a barrier which prevents the diffusion of impurities from the interior. Subsequent treatment with neon ions removes the surface layer of nitrogen exposing a clean surface. The nitrogen may act simply by blocking the gaps between iron atoms, or, as suggested by Prof. Garner, by com- pression of a layer of atoms in the lattice, in this case below the surface.Oatley l5 found similar behaviour with platinum : the work function of a surface prepared by bombardment with oxygen followed by’ argon was the same as that of a surface cleaned by flashing. Treatment with argon alone gave a value approaching that of a clean surface, but which rapidly altered, indicating contamination. In connection with Dr. Bremner’s communication l6 it may be noted that the bombardment treatment only’ raises the temperature of the wire to 400’ C. The first values of a are obtained about 5 min. after the treat- ment has ended and there is a continual slow drift lasting 40 min. or more. It seems very unlikely’ that the low value of a and the slow drift can be due to thermal disturbance particularly since the same wire in a con- taminated condition had been flashed at IZOOO C after which a was 0.2 and there was no drift at all after thermal disturbance had died away ( N 8 min.).Mr. P. R. Rowland (Guy’s Hospitul, London) said : When discussing the positions of adsorbed atoms on a metal surface evidence taken from the adsorption of hydrogen is perhaps the most unfortunate to consider since on adsorption it is probable that the orbital electrons of hydrogen enter the metal leaving protons which imbed themselves in the electron gas of the metal. It is difficult to attribute a size to :uch particles which cannot themselves be regarded as simply substituting for metal atoms at adsorption sites in the Langmuir sense but may wander through the lattice with high mobility. Electronegative atoms on the other hand are more likely to attach themselves to the surface by bonds resembling those in solid salts, i.e.to surround themselves with as many metal atoms as possible by substitution of metal atoms at Langmuir sites. It is inter- esting to note that in the presence of electronegative atoms the most stable metal crystal faces are those on which every adsorption site is identical so that the adsorbed atoms all lie in one plane. This suggests that such adsorbed layers together with the metal surface atoms may be regarded as two-dimensional compounds or adsorption complexes. Mr. P. Zwietering (GeEeen, NetherEands) (communicated) : The results of Dr. Maxted’s measurements 179 l8 were invariably expressed as the activ- ity A of the catalyst depending on the amount of poison c added to the reaction mixture.The greater part of this amount c will be adsorbed on the catalyst; however, a small portion must remain in solution to build up the equilibrium concentration corresponding with the amount adsorbed on the catalyst. Therefore we may write : G = a8 + j3p, . - (1) where -8 = fraction of surface covered with poison, and p = equilibrium conc. in the solution. The relation between p and 8 can simply be expressed by the equation of Langmuir : so that p = - - - 8Y 1 - 8 PYe c = c c e + - I - - 8 ’ l5 Pvoc. phys. SOC., 1939, 51, 318. 17 Trans. Furaduy Soc., 1945, 41, 406. l6 This Discussion. l8 This Discussion.I97 GENERAL DISCUSSION From this it follows : Assuming that the activity A of the catalyst is proportional to the fraction of the surface not covered by poison we may write From (2) it follows then A / A o = I - 0, or 0 = I - A/Ao.Consequently there must exist a linear relationship between and (I - A/A,). GA / A 0 - . ~ _ _ I - A / A o In Fig. I and z the experimental figures, obtained from the published Obviously the experi- The deviations at curves, have been plotted according to eqn. (3). mental data are very well described by this equation. high values of A / A 0 must be attributed to the limited ac- curacy with which the experi- mental data can be obtained from the published curves. It follows from the theory that the slope a of the straight lines must be deter- mined by the total area of catalytic surface, so that for the series I to IV a propor- tionality may be expected between a and the amount of catalyst used, whereas for the series V and VI, a must be proportional to the surface area.Finally also A , must be determined by the area of the catalytic surface, so that a must also be proportional to A , . In Fig. 3, 4 and 5 the values for CI derived from the linear plots have been plotted against the above-mentioned quantities. The result does not entirely come up to ex- Pectations. Instead of the .\ '-8 '-6 \. Tbiophen \ '-4 \ FIG. I. &pected straight lines through the origin, linear plots are found which intersect the ordinate at a constant negative value of ao. Probably this negative value of m0 can be explained if the adsorption of the reactants and/or the reaction products is also taken into account. In this case eqn.( ~ a ) must be replaced by where 6, = fraction of surface covered with poison and 8, = fraction of surface covered with reactants and/or reaction products. From this it follows :198 GENERAL DISCUSSION where may be considered as a constant. is a correction term for the slope which in a first approximation oil FIG. 2. / /y Surface urea oc t 2.0 . I-0 /x Wlof Pi 0.05 “O./ 0*/5 - cm3n/PT Alonola y er 4 FIG. 3. FIG. 4. FIG. 5. Prof. G.-M. Schwab (Athens) (communicated) : We made some years ago a series of measurements of the action of poisons on platinum catalysts, using the hydrogenation of cinnamic acid, poisoned by ethyl mercaptan. We found that on progressive poisoning (within the steep Schwab and Photiades, Ber., 1944, 77, 296.GENERAL DISCUSSION I 99 quasi-linear branch of Dr.Maxted’s curves) the activation energy increases considerably. From this we concluded that there was an energy dis- tribution between active centres and were able to deduce from this dis- tribution a surface equilibrium temperature coinciding in fact with the temperature of preparation of the catalyst. In view of these and Maxted’s present results it is interesting to observe that a poison eliminates the most active centres first, whereas sintering decreases their extent in the same proportion as it does the total surface. A possible explanation would be that, during preparation, a certain crystal habitus is established, wherein the ratio of different faces in the surface depended on the method of preparation (and temperature), and that during sintering this habitus is preserved, perhaps simply by the elimination of the smallest crystals without a change in shape of the larger ones.Prof. E. Wicke (Gottingen) said : With our calorimetric values of the adsorption heat of H, on reduced Ni powder, i.e. Q(0) = 21.2 - 18.0 d kcal./mole H, ( 6 = covered surface fraction ranging from o to 0.5) one may calculate adsorption equilibrium data for atomic adsorption using the well-known formula (based on Langmuir’s theory) : P+ P+ + exp {1/2R(-- Q(e)/T + AS)}’ e = where A S is the entropy difference of gaseous (molecular) and adsorbed (atomic) hydrogen, the latter being nearly negligible. These calculations are in complete agreement with measured isotherms, isosteres and isobars in a temperature range 50-250’ C where adsorption equilibrium is estab- lished.If one adsorbs C2H4 and H, together on a Ni surface (reduced Ni powder) around oo C and at such a low concentration ( 0 < 0.2) that the pressure in the gas phase is negligibly small (9 < I O - ~ mm. Hg), then no hydro- genation occurs at all. I f one pre-adsorbs C2H4 and immediately lets HB impinge on the surface from the gas phase, hydrogenation, though in- complete, takes place, decreasing in velocity and completeness with in- creasing time between pre-adsorption of C,H4 and inlet of H, ; this poison- ing, as is well known, is caused by the polymerization of C2H4. If finally H, is pre-adsorbed, and C,H, introduced from the gas phase, the hydro- genation is fast and nearly complete. But again the reaction proceeds more slowly, with increasing time after the pre-adsorption of H,.This sort of poisoning, not previously reported, is caused by the H atom “ dis- sipation ’’ along the Ni surface rendering the distance from one another too large for fast hydrogenation at lower temperatures. This effect is especially marked with pre-adsorption around IOOO C (where there is rapid “ dissipation ” of H atoms), followed by C,H4 a t about oo C. I do not see how these results can be understood in terms of the mechanisms of C2H4 hydrogenation hitherto proposed, but they confirm conclusively the idea of the catalytically active atom-pair state. Dr. D. D. Eley (Bristol) (communicated) : I should like to congratulate Dr. Reeck on the numerous stimulating contributions he has made to ethylene hydrogenation and to catalysis as a whole.It may be worth- while at this time to mention one or two points of detail. I am very glad to see that Dr. Beeck now believes the reaction is one between adsorbed ethylene and adsorbed hydrogen on the surface left bare from acetylenic complexes, as I do not believe it is possible to reconcile the reaction be- tween gaseous ethylene and adsorbed hydrogen with the usually observed kinetics (zero order in ethylene, first in hydrogen). Since he wishes to retain his old mechanism for tungsten, I think it will be of great interest to check on the kinetics in this case. I doubt, however, that there can be any hydrogen chemisorbed in the ethylene film on tungsten, as the para-hydrogen conversion is completely poisoned up to relatively high temperatures.20 Eley and Rideal, Proc.Roy. SOC. A , 1941, 178, 429. Hence there is no sign of solution.2 00 GENERAL DISCUSSION I wonder whether it is worth pursuing the idea that percentage d- character determines interatomic spacing, and thus correlating percentage d-character with catalyst activity (Boudart) or heat of chemisorption (Beeck). The strength of the chemisorption bond is determined not only by the extent of orbital overlap (therefore yo d-character), but also by the number of metallic electrons involved, and other factors. To a first approximation, all these are allowed for in my calculation by the way the sublimation energy is used as a measure of the strength of the metallic bond, The orders are : ,H2 Rh < Ni < Fe < Ta < W, Cr Obs.Heat \C,H,Rh<Ni<Fe<W,Cr<Ta Cr, Ta < Fe < Ni < W < Rh Cr < Fe, Ni < Rh < Ta < W. % d Heat Calc. (Eley) My calculations put Cr at the wrong end and Rh too high. The 04 d series, if it were completely reversed and W placed at the top, would exactly fit the experiment. This is what Dr. Beeck has done. I t seems to me Beeck believes that the surface bond is difleerent from the metallic bond and can only use the d orbitals left oueY from the metallic bond, i.e. the higher the yo d in the metal-metal bond, the weaker the metal-gas bond. To reconcile the position of W in the series, Beeck notes an un- published result that in evaporated films, its lattice parameter is different from the wire state, i.e.face-centred cubic with a greater distance which leads to a lower Yo d character. If this is so, I find it most difficult to understand why both W wires (Roberts) and W films (Beeck) should have the same heat of chemisorption for hydrogen. In my view it is incorrect to distinguish the surface orbitals M-H from the " under-surface " orbitals M-M. Dr. J . J . A. Blekkingh (Rotterdam) (comnzunicated) : A nickel hydro- genation catalyst in contact with hydrogen has the property of transmitting molecular hydrogen to double bonds in normal hydrogenation reactions and also of changing molecular hydrogen into atomic hydrogen. Adsorbed molecular hydrogen can be physically replaced by other substances ; this is not so with absorbed atomic hydrogen. The amount of atomic hydrogen in a nickel catalyst can be in proportion of I atom H to about 3 atoms of nickel.It is because of the presence of atomic H that there is-amongst other things-a shifting of the double bonds in unsaturated fatty acid derivatives, a side-reaction (isomerization) occurring in addition to the normal hydro- genation. This side-reaction takes place already at room temperature. The isomerization with atomic H in the nickel catalyst can be represented diagrammatically by the following steps: (I) A partial hydrogenation, in which I atom H is added to I CH group of a double bond. There are two possible radicals. (ii) A partial dehydrogenation, in which I atom H is given to the nickel. A double bond is formed again. There are three possibilities for the posi- tion of the double bond.A radical with a free valency is formed. -CH2-C€€=CH-CHZ-- / \ In this way we have found that a double bond can shift along the whole chain rather quickly. 21 Pauling, PYOC. Roy. SOC. A , 1949, 196, 343.GENERAL 1) ISCUSS ION 2 0 I It may be that the second step is not a partial dehydrogenation, but again a partial hydrogenation. In this case we may speak of atomic hydrogenation, with addition of H + H in the trans position to the double bond (addition to both sides of the plane of the double bond). The ordinary hydrogenation may be called nzolecular hydrogenation, with addition of H, = 2H in the cis position to the double bond (addition to one side of the plane of the double bond). In practice, we suppose, isomerization with atomic hydrogenation (including I : 4 additicn to conjugated double bonds) and molecular hydro- genation.take place more or less together, and this may be the reason for different hydrogenation velocities with different nickel catalysts, or with oriented and unoriented nickel films. Selectivity of fatty oil hydrogenation may be favoured by the formation of conjugated double bmds in linoleic and linolenic, and successive atomic hvdrogenation to isolated double bonds. In our opinion atomic H will not be able to poison the nickel hydro- genation catalyst, at least not above room temperature. Mr. P. F. Tiley (Bristok) said: The results of Beeck and Ritchie on the use of krypton, butane and other g-ses in the B.E.T. surface area measurements cf metal films cannot escape comment in view of the results which have been obtained with these gases on adsorbents in other physical forms.For example, I would refer to the work of Davis, De Witt and Emmett 22 who found from B.E.T. studies with krypton and butane on silver foil and monel ribbon that the areas determined were only 80-90 yo of the geometric areas. In the same work the authors found that, on adsorbents such as glass spheres, zinc oxide, tungsten powder and silica gel, the surface areas as measured by krypton and butane adsorption were less than the nitrogen values by some 30-40 yo. Beebe, Beckwith and Honig1 observed a similar phenomenon on comparing krypton and nitrogen adsorptions on anastase. In all the results quoted, the values for the cross-sectional areas of the adsorbate molecules were those c$- culated by the usual formula, viz : N,, 16.2 ; Kr, 15-2 ; C4H10, 29.7 A2 (Dr.Beeck uses Kr, 14.5 ; C4H10, 24-5). The results of Emmett and Beebe cannot be interpreted in terms of porz size nor is chemisorption of nitrogen likely on these adsorbents. The problem which faces us (discussed fully by Emmett 23 and Livingston a =) is that, having determined the number of adsorbate molecules in the monolayer by the B.E.T. isotherm, it is still necessary to employ a suit- able value for the cross-sectional area in order to calculate the absolute surface area. The value of 16 ( f 0.5)A2 for N, has been generally ac- cepted in view of confirmation by other methods of surface area deter- minatioq. Emmett suggested that the empirical values of approx.2 1 and 44 A? for Kr and C4HI0 produce more satisfactory results, but as yet no fundamental explanation of these discrepancies has been established. The surface area measurements of Dr. Beeck are extremely interesting as they are among the first to have been made on evaporated metal films and may throw some light on this problem. However, until this problem is satisfactorily elucidated it is felt that the determination of absolute surface area by B.E.T. methods must still involve uncertainties of the order of 20-30 74, particularly if gases other than nitrogen are used. Dr. F. C. Tompkins (Inzperial College, London) (communicated) : I think that Mr. Tiley expects too much of B.E.T. surface area determin- ations. Accepting all the postulates of the B.E.T. model there is no reason why we should introduce the cross-sectional area of the adsorbate mole- cule-it is the lattice parameter, or the area of the crystallographic site, which is important.When the adsorbate molecule is larger than the site, one has to make decisions based on energy considerations as t o what 22 Davis, de Witt and Emmett, J . Physic. Chem., 1947, 51, 1232. *3 Beebe, Beckwith and Honig, J . Amer. Chem. SOC., 1945, 67, 1554. 2 4 Livingston, J . Colloid Sci., 1949, 4, 447. G *202 GENERAL DISCUSSION is the stable packing in the layer-if it does happen to be hexagonal close-packing it is then and only then legitimate to use the cross-sectional area of the molecule. Normally up to 70 to 80 yo 1st layer coverage, it is a good approximation at low temperatures to assume localized ad- sorption-when there are attractive lateral interactions there can be, and often is, a transition to hexagonal packing; but where there are repulsive interactions the localized array is approximately retained. Furthermore, analysis shows that the R.E.T.eqn. merely corrects the Langmuir eqn. for 2nd higher layer adsorption by the term (fis - p ) / p s . Consequently where the B.E.T. c term is high we are virtually using only the 1st layer adsorption on a uniform surface for the evaluation of v,, and the above considerations are important. Finally, the evaluation of v m by the B.E.T. eqn. carries with it the assumption of the equality of thc internal partition functions of the adsorbed molecules in the 1st and 2nd layer-provided the same gas is used for different adsorbents and the heats of adsorption are approximately the same no trouble arises in comparing areas.With different gases on the same adsorbent, the ratio of such partition functions though of the order of unity will be somewhat different and, particularly where c is not large, different om values can be expected. It seems therefore profitless to con- cern oneself with the precise values to be taken for the cross-sectional area of molecules in the adsorbed state. The trouble is more acute when a non-uniform surface is considered-the transition from 1st to 2nd layer adsorption is then less abrupt and all the uncertainties and inac- curacies associated with low c values on uniform surfaces are again present. Mr. P. F. Tiley (Bristol) (cornrnunicated) : In the last few years, both theory and experiment have demonstrated the shortcomings of the B.E.T.model, particularly with regard to heats of adsorption. How- ever, the value of the standard B.E.T. technique, in providing a reason- able estimate of surface area, has not been disproved. In so far as an assessment is possible, the errors (using N, as adsorbate) are not as great as would be expected from Dr. Tompkins’ contentions concerning the invalidity. of the B.E.T. isotherm for non-localized adsorption and the irrevelance of the dimensions of the N, molecule in localized adsorption. With krypton, which may be used for low surface areas, consistently good agreement with the N, values is obtained, using the “ coIrected ” value for UKr m 20 Hi2.Whether the reason for this “ correction ’’ lies in the method of evaluation of urn, or in the packing of the adsorbate molecules is, I suggest, a matter for conjecture. It is felt that, whilst Beeck and Ritchie have drawn rather bold conclusions from their experi- mental results, yet the logical outcome of Dr. Tompkins’ arguments is to underrate the empirical value of the B.E.T. method. In connection with these low pressure B.E.T. studies at liquid air temperatures, few workers seemed to have mentioned the question of thermo-molecular flow. From the experimental details given, i t is diffi- cult in some cases to discover if there is justification for thus ignoring this effect. Mr. Porter in Section IV of this Discussion gives an excellent quantitative treatment of this phenomenon.Dr. F. C. Tompkins (Imperial College, London) (communicated) : It is my contention that the B.E.T. method cannot be expected to give results of absolute surface area determinations to better than 10-20 yo. I know of no experimental data of measurements obtained by methods not involving adsorption where the agreement with the B.E.T. method is of higher concordance. Nor can I be convinced by the fact that N, and Kr (using an arbitrary value of UKr) give consistent values ; 0, and A for example often give quite different values than does N,. My’ criticism was concerned with the excessively high accuracy implied by giving values of (TN? as 16.2 A (or 16 f 0.5 A,) ; it would be more consistent if this were written as 16 f z A.GENERAL DISCUSSION 203 Prof.A. R. Ubbelohde (Belfast) said : The terminology of the electron band theory of metals has not been adapted in any quantitative way to catalysis, and there is a risk that rather vague concepts may hinder real progress in interpreting catalytic phenomena. Starting with the Pauling theory as an alternative to the electron band theory of metallic binding, it is possible to formulate a more con- crete model for the binding of atoms such as H, 0, N to the metal atoms M of the catalyst. This " pseudometallic bonding " model postulates that a group of atoms such as (M-H,) behaves as regards external elec- tronic behaviour like the atom n places to the right in the periodic system. For example, the grouping Pd-H behaves electronically like Ag, and in particular forms metallic bonds to neighbouring atoms similar to those which would be formed by Ag.There is considerable experimental foundation for the pseudometallic bonding of groups (M-H,). For example, when hydrogen is dissolved in alloys Pd-Ag, in certain ranges of composition 23 the electrical resistance is actually lowered, as would be expected of the " pseudosilver group " (Pd H), compared with Pd, when alloyed with Ag. Again, the wetting of metals by mercury, which almost certainly involves the formation of metallic bonds between Hg and the surface atoms of the metal, is changed by dissolution of hydrogen in the sense to be expected from the pseudo- metallic model. Thus whereas Hg wets Pd only slowly, it wets Ag with avidity and recent experiments have shown that it likewise amalgamates with great avidity with Pd H,.,, forming I ' palladium hydride " amalgams.Mercury also wets iron after the surface has been cathodically charged with hydrogen. This concept of the pseudometallic bonding of hydrides can explain a number of features of hydrogenation catalysis, such as the poisoning by mercury. It leaves the possibility open that on isolated peaks on the catalyst hydrogen might be bonded by an ordinary covalent bond, less stable than the pseudometallic mode of binding in which (M-H,) is alloyed with the neighbouring atoms. Properties of ions such as VO++ and UO,++ suggest that the grouping M-0, should also be able to act as a pseudometallic unit similar to the metal zn places to the right in the periodic system.And the concept can be extended in obvious ways, when experimental evidence suggests that such extension is fruitful. Mr. A. S. Porter and Dr. F. C. Tompkins (Imperial College, London) (communicated) : The admission of mercury vapour at oo C to an evapor- ated iron film covered with hydrogen has been found to cause rapid de- sorption of most of the hydrogen, fo'i'owed by a further slow desorption. After I to z hr., desorption was complete (within I yo of the amount of gas originally put on the film). On admission of H, again, no further adsorption of hydrogen could be detected on this film, even at - 195' C and I O - ~ cm. H,. Similarly, a film prepared in the presence of mercury vapour was found to take up less than 1/2ooo of the normal amount of hydrogen. It is clear, from these observations, that any Fe-H-Hg alloys which may exist are far less stable than the Pd-H-Hg alloys mentioned by Prof.Ubbelohde. Mr. D. A. Dowden (Bzllinghawz) said : In a discussion of the possible correlation of the catalytic activity of metals with their electronic struc- ture, both the electronic models in current use should be exhausted of their possibilities since it appears 2 5 that some situations can best be described by one of these models and with difficulty or not at all by the other. The band theory suggests effects connected with changing electronic exit work function, inner potential, electron-level density, etc., and em- phasizes electron mobility while Pauling's transitional metal valencies encourage ideas more closely allied to the chemical, directed bond.25 See Ubbelohde, J . Chem. SOC., 1950, 1143. ,ti Mott, Proc, Physic. Soc., 1949, 62, 416.204 GENERAL DISCLSSION Mott 27 has given a qualitative explanation of the variation in cohesive energy of the transitional metals as the d-bands are filled up. Where the metals possess vacant d-levels the increased cohesion can be ascribed to the decrease of the average electron energy on removal of s, p-electrons into lower-lying d-levels. If a chemisorbed particle is considered, in a crude approximation, as just another lattice surface atom then the change in chemisorptive bond strength of a given substrate at the surface of transitional metals should parallel the change in cohesive energies of the metals. An old rule 28 expresses this trend by including a factor, pro- portional to the square root of the sublimation energy of the metal, in the binding energy of the chemisorbed substrate.This interpretation assumes that the electron levels remain almost unchanged in the series of metals and focuses attention on the behaviour at the top of the Fermi distribution. In fact there may be variations at the bottom of the distribution leading to changes in the centre of gravity of the occupied and hence to changes in cohesive energy and heats of chemisorption but the parallelism between these two remains unchanged. Prof. A. R. Ubbelohde (Belfast) said : The discontinuous change in catalytic activity of Pd H,, or of Pd Au,, at the critical compositions which are just no longer paramagnetic, may have more than one cause. A factor for which insufficient allowance may have been made is the highly strained two-phase structure of the alloys up to the critical compositions, when they contain hydrogen.For example, as more hydrogen is dissolved in Pd, ci and f i Pd H phases with approximate composition Pd Ho.05 and Pd H 0 . 6 5 coexist in the solid, and the hydrogen-rich phase merely grows at the expense of the other as the total hydrogen content increases. A significant point for catalytic activity is that the coexistence of these two phases which have different specific volumes induces a high state of strain in the solid. This strain persists, as can be shown by X-ray studies and in other ways, till all the solid has been converted to the /3 phase, after which it is much smaller.Two-phase structures are also formed by PdAu, with x < 0.6, when these alloys have dissolved hydrogen. The resultant lattice strains may likewise affect catalytic activity. Mr. D. A. Dowden (Billinghanz) (communicated) : Experiment indicates that the idiosyncrasies of particular alloy systems, remarked by Prof. Ubbelohde, do not disturb the general activity pattern of a decline in hydrogenation activity as holes in d-bands are filled up. The notion of " a hole in the d-band " or its counterpart " a vacant d-orbital " should be neither less useful nor more confusing than was the idea of " an in- complete Lewis-Kossel octet " a generation ago. The fall in magnetic susceptibility of palladium as it dissolves hydrogen is the over-riding criterion of activity changes in the presence of hydrogen ; phase changes and strain effects appear to be of secondary importance.It should have been a clear indication that in some temperature ranges reactions involving palladium must be hydrogen-poisoned. These solu- tion effects occur only on a smaller scale in catalysts of nickel and platinum consequently the nickel-copper alloy series, being free from order-disorder transitions and immiscibility gaps is, despite its ferromagnetism, the best for general study. The geometric path of adsorbed hydrogen atoms during the transition from the chemisorbed to the dissolved state must surely include inter- stitial sites in the surface lattice just as solution involves interstices in the bulk. These surface sites, being directly accessible from the gas phase, are in the chemisorption zone, possess the largest number of surface- metal nearest neighbours, should have the largest heats of adsorption and 27 Mott and Jones, Properties of Metals and Alloys (Oxford, 1936).28 Eucken, 2. Elektrochem., 1922, 28, 6. 29 Seitz and Johnson, J . A@l. Physics, 1937, 8, 84, 186 and 246GENERAL DISCUSSION 20 5 should be filled first. Such a surface phase might undergo transitions analogous to those of the bulk phases. More important, as Prof. Ubbelohde implies, since the ratio palladium/hydrogen atoms -I, for instance in a stoichiometric, surface phase on a (100) plane then that phase has electronic properties more like a layer of copper or silver than palladi~rn.3~ Further adsorption above this phase, e.g.above the palladium atoms, can occur only by much weaker bonds. In fact because of diffusion to the interior this surface phase is usually non-stoichiometric under catalytic condition and is never as inactive as copper except at very low temperatures. No doubt these details could be resolved by a czreful study of solution and chemi- sorption in series of binary alloys of the type described. Dr. G. C. A. Schuit ( K ~ n i n k l i j k e l S h e l l Laboratorium, Amsterdam) (communicated) : Dr. Beeck’s remarkable correlation between the character of the metallic bond according to Pauling and his velocity constants for the hydrogenation of ethylene can be still improved upon by taking into account the valency of the metal as stated by Pauling. To effect this, 10 log k is plotted against valency x yo d-character.sl It is seen that the decrease in activity for Ta, which originally could not be accounted for by the decrease in d-character, is effected by the decrease in valency and that Ta falls on the same straight line as the metals with valency 6.Cr, which, according to Pauling, should have a valency between 6 and 3 , indeed seems to be situated in valency between these two extremes. Only W appears definitely to fall out of the correlation, but as Dr. Beeclc reported the metal-film structure to be face-centred, and thus abnormal, this deviation seems explainable. Both an increase in valency and in d-character tend to strengthen the total bonding strength between the metal atoms ; this correlation seems to indicate that the catalytic activity runs parallel to the strength of the bonding between the metal atoms.Dr. D. D. Eley (Bristol) said : The calculations on heats of chemi- sorption3z assume that the metal orbital used to bond the surface hydrogen is, in all cases, the ordinary metallic (here symbolized as d2.56~fi2.2s) orbital. This would point to a second extreme hypothesis to explain our results, viz. the chemisorbed H atoms are held by d2.56~fi2-22 and the atomic d orbital is only required to bind the H3 activated complex strongly. Obviously, to decide between these closely similar possibilities requires further work. In reply to Prof. Ubbelohde, the question of phase separation was most carefully considered by us with the conclusions given in our paper, viz, that there is no phase separation in the Pd-Au system, but that a part of the increase in activation energy in the Pd-H system may arise from the large spacing of the ,&phase.We do not want to ignore his suggested effects of lattice strain on electron energy levels, but we believe the main results described in our paper may be described in terms of the simple band or orbital picture. Prof. G.-M. Schwab (Athens) (communicated) : It would be of interest to study the parahydrogen conversion on Hume-Rothery alloys of Group I metals where d-gaps are absent and the occupation of the conductivity band is an important factor. The parahydrogen conversion ought to fall in one class with hydrogenation and dehydrogenation reactions] and an increase of electron concentration or formation of a y-phase, it is predicted, should increase the activation energy.I agree with Dr. Eley’s view, that the adsorption of hydrogen “ a s a positive ion ” does not really mean complete ionization, but rather a covalent bond by means of electrons shared between the proton and the conductivity band. 3O See Beeck, this Discussion 3l Data taken from Pauling, Proc. l20-v. SOL. A , 1949, 23, 196. 32 Eley, this Discussion.2 06 GENERAL DISCUSSION Prof. R. M. Barrer (Aberdeen) (communicated) : I wish to support the observation by Couper and Eley concerning the frequently supposed positive charge on hydrogen dissolved in Pd. The calculation of the +ve charge per H-atom of 1/50 of the protonic charge must if correct lead to the conclusion that nearly all the H-atoms are substantially neutral.I pointed this out two years ag0,~3 and indicated also that the frequently postulated protonic character of the dissolved hydrogen was hardly compatible with the well-known clustering of the hydrogen atoms in the metal when a critical concentration of dissolved hydrogen was reached. I f the H atoms are protonic, they should rep4 one another. I would also like to raise a point concerning Table I in the paper by Eley and Couper. All the larger apparent energies of activation E, are measured in a much higher temperature range than the small energies and there is an abrupt change in the temperature range at the point where E, also changes abruptly. Since one has only apparent energies of activation to go by, and since such apparent energies may themselves be dependent on temperature (the last two paragraphs in the paper by Rideal and Trapnell 34 emphasize just this point) I would like to ask whether we can in fact be sure that true energies of activation would vary in any similar way ? This it seems may have to -be answered satisfactorily before the electronic effect on the true energy of activation can be regarded as established. Prof.J. H. de Boer (Geleen and DeZf.) said : The decomposition of H,O, is enormously catalyzed by the presence of small amounts of ions of e.g. Co2+, Cu2+ etc. Is it to be excluded that a small amount of Cu2+ are formed locally when H,O, is in contact with copper? If this were true, curves A, B, and C of Fig. z of Dowden and Reynold’s paper would denote a decrease of surface attack by H,O, on copper, by increasing amounts of nickel.This might be important for corrosion problems; the explanation of such a behaviour would follow the same lines as given by these authors. Mr. D. A. Dowden (Billingham) (communicated) : Early in our experi- ments we considered the possibility’ noted by Dr. de Boer, that the whole or a part of the activity in hydrogen peroxide decomposition observed on the copper-rich foils might be due to a homogeneous catalysis by metal ions dissolved from the foils. However, analysis showed no significant differences in metal content between fresh and spent substrate. More- over, if an otherwise normal experiment was halted during the steady state (well before the exhaustion of the hydrogen peroxide) the foil re- moved and the experiment continued, no activity was observed in the second stage.These tests dispose of the possibility of a homogeneous catalysis. Mr. D. A. Dowden (Billingham) (communicated) : It has been suggested by Pro€. H. S. Taylor that some of our results might arise from bulk diffusion phenomena and indeed the effects noted by him occur quite frequently in the comparison of liquid with vapour-phase hydrogenations on the industrial scale. However, in these experiments the inter-phase contact is immeasurably better than can be found in a normal liquid- phase converter. Inspection of the shaking vessel (photography with a flash of duration one microsecond) confirmed the intimate mixing at shaking speeds where the reaction rate was already independent of shaking speed and showed the absence of any peculiar resonance effects.Attention should be drawn to the fact that the vapour-phase reaction over the same catalysts shows activity trends with increasing copper content which parallel those found in the liquid-phase reaction. This indicates con- clusively that the activity pattern is a result of variation in electronic structure and not of complicated diffusion processes. Since the absolute activities of the iron-nickel alloys are of the same order of magnitude as 33 Faraday SOC. Discussions, 1948, 4, 77, 120. 34 This Discussion.GENERAL DISCUSSION 20 7 those of the nickel-copper alloys and all linear dimensions, concentrations and shaking speeds remain constant, the same conclusion applies to the activity pattern of the iron-nickel alloys. Dr.N. Uri (Manchester) (communicated) : Dowden and Reynolds point out that it is necessary to consider the reaction Fe2f + H202 + FeO*+ + H,O as an alternative to the Haber-Weiss 35 reaction Fez+ + H,02 -+ Fe3 + OH + OH- which adapted to alloy catalysts becomes metal + H202 + (metal, 0) + H20. Although this mechanism, first suggested by Bray and Gorin,36 has been recently revived by Kolthoff and Medalia,37 who show that it is in agree- ment with the reaction kinetics of the hydrogen peroxide decompxition, it would appear that the formation of Fe02+ ion is unlikely for the following reasons. (i) Whereas the Haber-Weiss reaction can be interpreted as one in- volving the breaking of the 0-0 bond in H202 and electron transfer from the Fez+ ion to one of the OH radicals, the Bray-Gorin reaction would require a complicated rearrangement of atoms. In the latter case one would expect a considerable activation energy.Experimentally, however, the activation energy is found to be comparatively small, viz. of the order 5-10 kcal.:s and, indeed, corresponds closely to the endothermicity of the Haber-Weiss reaction (in accordance with the energy values estimated by Evans and Uri 39 and Evans, Baxendale and Uri 40 for the 0-0 bond in H,O,, the ionization potential of ferrous ion and the electron affinity in solution of the OH radical). (ii) Baxendale, Evans and Park38 have shown that OH radicals pro- duced in the Haber-Weiss reaction initiate the polymerization of vinyl compounds. No one has as yet observed the inclusion of iron atoms in the polymers, although these systems have been thoroughly investigated.In this connection I would like to mention that I have recently postulated the formation of radicals derived from tungstic and molybdic acid : 41 it would also be interesting to establish experimentally whether such radical endings could be detected in polymers formed in systems like Fez+ + H2W0, + H202 + vinyl monomer. Prof. G.-M. Schwab (Athens) (communicated) : It is rather satisfactory that two so entirely independent approaches to the influence of the electronic factor, as Mr. Dowden’s and ours, lead to so complete agree- ment of results. The only difference is that Mr. Dowden, working with electron-defect metals, considers the electron gaps in these elements as decisive, whereas we, working mainly on Group I metals, have to do with the empty conductivity levels.There, in the concentration range where the iron gaps are filled up, Mr. Dowden would probably expect a decrease in formic acid dehydrogenation, whereas an increase is found. However, this difference can easily be explained by thc fact that gold is a much more powerful catalyst than iron or copper, in a way that its conductivity band is of much more importance than the iron I ‘ gaps ”. 35 Haber and Weiss, Proc. Roy. SOC. A , 1934, 147, 334 ; Naturwiss., 1932, 20, 94896 Bray and Gorin, J . Amer. Chem. Soc., 1932, 54, 2124. A sphere of contact is given in our study of iron-gold alloys. 37 Kolthoff and Medalia, J . Polymer Sci., 1949, 4, 377. Y * Baxendale, Evans and Park, Trans.Faraduy Soc., 1946, 42, 155. 39 Evans and Uri, ibid., 1949, 45, 224. 40 Evans, Baxendale and Uri, ibid., 1949, 45, 236. 41 Uri, J . Physic. Chem., 1949, 53, 1070.2 08 GENERAL DISCUSSION As for the formic acid reaction on nickel and copper, the most reliable values for the activation energies, as found in our laboratory, are 15 and 18 kcal. respectively, which is one-half of the figures given by him. The explanation by poisoning is probably correct. In Rienaecker’s early work values as high as 26 kcal. instead of 17 for silver and zg kcal. instead of 12 for gold were given. One remark is in order concerning the frequency factors of the hydrogen peroxide decomposition. Their logarithms, expressed for I 5 cm. surface and a product flow in ~m.~/min.are 2-1 for q = 10 kcal. and 8.8 for q = 24 kcal. For formic acid vapour over any catalyst whatsoever the respective values are 3-1 and 7.3 kcal. Thus, as to the general relation- ship of activity and activation energy, the two reactions, externally so different and internally s3 opposite, do not differ essentially. Lastly, it is of interest to note that Wagner and Himmler found that the completion of the electron band decreases the (de)hydrogenation, but enhances the nitrogen adsorption, thus indicating the adsorption of nitrogen in the form of an anion or at least at the expense of the metal electrons. Mr. D. A. Dowden (BiZZingham) (counntunicated) : It is especially im- portant now, in view of the successes of the carbonium-ion theory in other fields, to keep an open mind about the presence of positive ions at metal sur:aces.As Huang and Wyllie 4 2 have pointed out, simple image force calculations provide no conclusive answer to this problem. In systems where hydrogen chemisorption is accompanied by a positive dipole layer there is no difficulty in applying an old established model used in the study of thermionic emission43 and considering a small fraction of the surface to be covered by positive ions 01 strong positive dipoles. More- over, any chemisorbed hydrogen, whatever the polarity of the bond, may function as a proton under those conditions where the presence of another species of large proton affinity (e.g. olefines, alcohols) can lead to the pro- duction of strongly adsorbed carbonium and oxonium ions.Bremner 44 has given evidence suggesting that the hydrogenolysis of esters and al- cohols is controlled by reactions involving carbonium ions at the surface of metallic copper catalysts. If this analysis is correct then one must admit the presence of adsorbed species which are essentially positive ions not only at copper surfaces but at all active metal surfaces. Prof. J . H. de Boer (Geleen a ~ d EeZft) said : Prof. Schwab’s remarks that there is a linear relationship between the logarithm of the frequency factor and the activation energy reminds me of a similar relatimship between the work function I$ of a metal and the logarithm of the constant A of the Richardson equation for thermionic emission : i = A TZe-Ed/kl’ in cases of adsorption of gases on metal surfaces.An increase of 4 by the chemisorption of, e.g. oxygen, is accompanied by a large increase of A , whilst a decrease of + by the adsorption of, e.g. alkalimetals or thorium or, in many cases, hydrogen, gives rise to a substantial decrease of A . Richardson himself noticed already that both A and I$ depend amongst other things, upon the degree of occupation with adsorbed material 19 and that a relationship : The importance of this in ammonia catalysis is obvious. 40 = 45, + k In Ae holds. The cause of the change of the constant A was later ascribed by Schottky and by others to the temperature dependence of the dipoles which the adsorbed atoms have formed on the metal surface. This tem- perature effect gives rise to a temperature dependent term in A+, which 4 2 This Discussion.43 de Boer, Electron Emission nnd Adsorption Phenomena (Cambridge, 1935). Bremner, Research, 1948, I , 281.GENERAL DISCUSSION 209 term when substituted in Richardson’s equation cancels its proportion- ality with T against the T of kT, the result being that part of the exponent is independent on temperature and will automatically be combined with A . The effect gives indeed the right order of magnitude for the changes in A which are 0bserved.l If now in a catalytic reaction the formation of a positive ion plays a dominant role, and we can alter the work function of the metal independ- ently, we may expect : (i) An increase of the work function to give a decrease of the energy of activation and a decrease of the frequency factor ; (ii) a decrease of the work function to give an increase of the activation energy and an increase of the frequency factor.If the formation of a negative ion on the surface plays a dominant role, we expect : (iii) an increase of the work function to increase the activation energy and to increase the frequency factor ; and (iv) a decrease of the work function to decrease the activation energy and to decrease the frequency factor. I would not be surprised if double layers of either positive or negative sign play a role at the surface of alloys. I do not think it improbable that, e.g. in a silver-cadmium alloy, the distributicn of the cadmium atoms and the silver atoms at the surface results in a small double layer with its positive side pointing away from the metal.The formic acid decomposition on such a surface would then represent case (ii), giving an increase of both the activation energy and the frequency factors, as Prof. Schwab has fcund. Mr. P. Rowland (Guy’s Hospital, London) said : I wish to call atten- tion to the work of Gwathmey and his collaborator^.^^ The essential properties of heterogeneous catalysts are mainly the chemical properties of the surfaces of crystallites and it is therefore valuable to study the chemistry of the surfaces of single crystals. Two main types of experi- ment have been carried out on electro-polished single crystal spheres, slices, etc. ( a ) The rates of reaction of the different faces with O,, Cl,, etc., have been studied and found to vary markedly with the indices of the surface ~lanes.4~ (b) Catalytic reactions have been studied, e.g.the reaction of H, with 0, in the neighbourhood of 400’ C on copper and In many cases rearrangement of surface atoms of the metal occurs during the catalytic reaction to give microfacets parallel t o certain crystal planes, e.g. the reaction between H, and 0, on copper exposes mainly ( I I I) micro- facets. Gwathmey has further shown that different faces have different catalytic activities (e.g. for H, + 0, on copper at 4ooOC the activities of (001) and (111) are in the appropriate ratio I : 2 ) and has shown some correlation between the catalytic properties of a face and its oxidation rate. such as The author has extended the oxidation technique by studying reactions c u (cryst.) + I&) -+ CuJ,(g) under conditions such that bulk halide is not built up.48 These reactions etch the surface so that only certain microfacets appear (i.e.(OOI), ( O I I ) , (111) and (012)). It is the author’s opinion that these microfacets appear because they are stabilized by chemisorbed monolayers of halogen. It is also suggested that in the reactions between H, and 0, studied on metal surfaces by Gwathmey, the rearrangement of the surface metal atoms is 45 Cp. de Boer, Electron Emission and Adsorption Phenomena (Cambridge, 193 5 ) . 46 Benton and Gwathmey, J . Physic. Chem., 1942, 46, 969. 47 Gwathmey and Leidheiser, J . Amer. Chem. SOC., 1948, 70, 1200, 120G. 48 Kowland, Nature, 1949, 164, 1091.2 I 0 GENERAL DISCUSSION ordered by relatively stable adsorbed layers of oxygen on the faces con- cerned.There is strong evidence that in the presence of electronegative substances at moderate temperatures the mobile surface atoms of a metal rearrange to give definite microfa~ets.~~ These experiments have an important bearing on the production of metal catalysts either by reduction of oxy-compounds or by leaching out one component of an alloy as in the manufacture of Raney nickel. Thus, in the later stages of the reduction of CuO with hydrogen, con- ditions closely resemble those in the reaction of H, with 0, on copper crystals. In both cases copper atoms migrate freely through the vapour phase.;O It will be expected, therefore, that the copper crystallites pro- duced will present mainly (111) facets as in Gwsthmey’s experiments. Also, unless exceptionally pure hydrogen is used nearly all metals will retain a monolayer of oxygen and their catalytic properties are those of surfaces so contaminated.On the argument given above it is pr. bable that, beside altering the work function, etc., these layers stabilize the facets of the crystallites and inhibit sintering (e.g. pure copper sinters at room temperature and is non-catalytic 51). Similar considerations apply to experiments on “ flashed ” tungsten wires. The cleaning process at high temperature should leave a smooth surface 5 2 but heating in the presence of oxygen should produce microfacets and hence a change in the surface. There is strong evidence that etching in solution is also controlled by chemisorbed 1ayers;S and this may be important in the production of Raney nickel by leaching out A1 from Ni + A1 alloy. The nickel atoms must re-aggregate since they are too widely spaced in the original alloy to form a structure after removal of the aluminium. The adsorbed layers will play an important part in determining which facets predomin- ate in the nickel crystallites and will also remain to modify the catnlytic properties of the metal. Studies of the electrochemical properties of metal crystals are therefore essential to an understanding of Raney type catalysts. The chief criticism of the single crystal method is the impossibility of obtaining a clean surface. However, as indicated above, most metal catalysts are contaminated with oxygen. In any case, the state of the suiface which is important is that which obtains during catalysis and it is not practical (or even possible in some cases) to work with reagents which are completely free from oxygen. The chemistry of catalysts is therefore, in the main, the chemistry of contaminated surfaces and for this reason the study of the behaviour of known crystal surfaces, under well-defined conditions of contamination (e.g. oxidation, H, + 0, catalysis, etc.) is of fundamental importance. The results of Gwathmey and of the author together with those of electron emission experiments show that the adsorptive properties (and hence the effective chemical properties) of metal surfaces are very’ dependent on the crystal plane concerned. By applying single crystal techniques in conjunction with those of electron diffraction and electron emission further insight may be gained into the processes involved in the production of metal catalysts, the types of surface which predominate and the chemical properties of those surfaces. Mr. D. A. Dowden (BiEZingham) (communicated) : Our magnetic ap- paratus could not be used to determine Curie points and it was not there- fore of use as a test of homogeneity in the ferromagnetic alloys. The supported alloy catalysts are not completely homogeneous because the susceptibilities of the nickel-copper series from 0.6 to 0.9 atomic fraction of copper are somewhat larger than for the corresponding annealed foils 4 9 Chalmers, King and Shuttleworth, Pvoc. Roy. SOC. A , 1948, 193, 465. 50 See also Dowden and Reynolds, this Discussion. 51 Beeck, Smith and Wheeler, PYOC. Roy. SOC. A , 1940, 177, 62. s 2 But only if alternating current is used ; see Johnson, Physic. Rev., 1938, 54, 459. 53 Tamman and Sartorius, 2. anorg. Chem., 1928, 175, 97.GENERAL DISCUSSION 21 I and show a slight field depend:nce between 2000 and 4000 oersted. Nevertheless the close parallelism between the decline in susceptibility (at a fixed field strength) and activity in the region where the 3 d-band is filling establishes the main point for the bulk of the catalyst. The inferences drawn from the results on the iron-nickel series must be true in a similar approximation.

ISSN:0366-9033    

DOI:10.1039/DF9500800191

出版商:RSC    

年代:1950

数据来源: RSC

摘要:

GENERAL DISCUSSION 21 I 111. ADSORPTION AND CATALYSIS ON OXIDES INTRODUCTORY PAPER BY W. E. GARNER The principal catalytic reactions occurring on oxides are those con- cerned with hydrogenation, dehydrogenation, dehydration and oxidation, and subsequent molecular rearrangements such as isomerization, polymer- ization, etc. The basic processes therefore involve the exchange of hydrogen, oxygen and water, and in some cases carbon monoxide, between the adsorbent and the substrate. It is clear that the nature of these elementary processes must be understood before the kinetics of oxide catalysis can be placed on sound foundations. With this in mind, con- siderable attention has been paid in recent years to the study of the basic processes by physical and chemical methods. The Substrate.-The oxides act as catalysts in a region of temper- ature where there is believed to be relatively little mobility of oxygen atoms in the lattice itself, although, as shown by the rates of oxidation of metals at low temperatures, there may be appreciable movement of metal ions under the electxical fields which are set up during the oxida- tion.lS On the surfaces of oxides, however, there must be greater freedom of movement, and the mobility of the oxygen atoms must be considerably greater on the surface than in the bulk of the oxide.on the interaction between mixtures of oxides shows that the first process on raising the temperature is the flow of the more mobile oxide over the surface of the less mobile oxide. on the ex- change of 0 l 8 with the oxygen atoms of A1,0,, MgO and Cr,O, show that there is a very considerable mobility of oxygen from the first layer of atoms into the adsorbed state and vice versa. In this Discussion, Winter describes further work on oxygen exchange with thorium oxide (440-540° C), chromium oxide ( 3 ~ 3 - 5 2 7 ~ C) and mag- nesium oxide. In all cases the exchange reaction is kinetically first order.The interesting result is obtained that for chromium oxide the activation energy is 29.5 kcal. up to 410°, and 1-4 kcal./mole above this temperature. The rate-determining step is ascribed to desorption below 4 1 0 O and to exchange between adsorbed oxygen and the oxygen atoms of the lattice above this temperature. If this interpretation is correct, then the oxygen ions in the surface of Cr,O, are mobile at catalytic temperatures.The oxide catalysts may be classified into two groups, semiconductors and insulators. In semiconductors there is a higher mobility within the lattice than occurs with insulators, as a result of which lattice defects are the more readily produced. There are two principal kinds of defect in oxides, one due to an excess of metal and the other due to an excess of oxygen within the lattice, and both kinds may arise during the oxidizing and reducing processes that occur in catalytic reactions at high temper- atures. However, within the range of temperatures at which catalytic The work of Huttig Experiments by Winter Mott and Gurney, Electronic Processes in Ionic Crystals, 1940. Anderson, Trans. Faraday SOC., 1948, 44, 163.Hiittig. 2. anorg. Chem., 1935, 224, 225 ; see also this Discussion. Houghton and Winter, Nature, 1949, 164, 1130.212 INTRODUCTORY PAPER processes are carried out, the formation of defects by oxidation and reduction is confined, in general, to the surface layers, since the diffusion of defects within the lattice possesses a high temperature coefficient. The importance of these defects for catalysis lies in the fact that around the defects there are produced a variety of surface energy levels, some of which may lie in the range required ( a ) for electron exchange reactions between the absorbent and the lattice and ( b ) for the facilitation of chem- ical bonding of molecules and atoms with the surface. Dowden6 has analyzed the position from this point of view, and has indicated the ways in which defects can facilitate adsorption prccesses.Since the basic reactions occurring on the surface of semiconductors may modify the semiconductivity, measurement of this property during adsorption and catalytic reactions yields useful 7 Chemisorption.-There has been much discussion 3s to whether chemisorption and catalytic reactions occur on the whole available sur- face or on active centres. As Eucken * points out, it is important to separate adsorption from catalysis, since there may be reasons of a statis- tical character why, on a completely covered surface, only a part of the adsorbed film is effective catalytically. There seems to be little doubt from the studies of tungsten and nickel surfaces, that the whole surface is involved in the chemisorption of hydrogen, and there also seems to be little doubt in the case of some of the catalytic reactions involving hydrogen.There is, however, no reason why this should be generally true. The energetics of the catalytic activation of hydrogen have been discussed by Polanyi,g who finds that on a good hydrogen catalyst the strength of the hydrogen bond with the surface must lie within narrow limits. This is probably true generally with regard to the relationship between the strength of the bonds of chemisorbed substances and cata- lytic activity. The conditions for ths formation of a chemical bond with the characteristics required for catalysis may be just as likely to be met at a defect, a Frenkel step, or a dislocation as at a plane surface.There is therefore no a priori reason why catalysis should not proceed as rapidly at defect structures as on the plane surface, except that the numbers of defects are necessarily limited compared with the total number of surface atoms. Many observers working with oxide surfaces have found it necessary to postulate the presence of " active " centres. A striking example is that of catalysis on alumina-silica, where the activity is pro- portional to the number of structural acid groups on the surface.1° One of the stages in a catalytic reaction involves the dissociation of molecules, which requires the breaking of strong chemical bonds (H,, 102, and O,, 117 kcal.). This is, however, balanced by the simultaneous formation of strong bonds with the surface of the solid.There is, how- ever, a lower limit of temperature below which the dissociation occurs very slowly. This varies from case to case ; for example, it occurs at - 1 5 0 O K for hydrogen on nickel8 and at N 3oo°K for hydrogen on manganous chromic oxide.ll There is considerable evidence available (cf. Eucken8) that hydrogen may be chemisorbed as molecules as well as atoms. Taylor has summarized the evidence bearing on the dissoci- ation of hydrogen and other gases on solid surfaces.l2 In many of the catalytic reactions involving oxides, adsorbed hydrogen would appear to be dissociated into atoms at temperatures below those at which catalytic reactions involving hydrogen occur. Therefore, in Dowden, J . Chem. SOC., 1950, 242. Garner, Gray and Stone, this Discussion ; Gray, this Discussion.Anderson and Bevan, this Discussion. Eucken, this Discussion. Polanyi, Sci. J . Roy. Coll. Sci., 1937, 7, 21. lo Millikan, Mills and Oblad, this Discussion. Taylor and Williamson, J . Amer. Chein. Soc., 1931, 53, 2168. l2 Taylor, ibid., 1931, 53, 578 ; Advances in Catalysis (1g47), p. I .W. E. GARNER 213 the majority of cases, the dissociation of hydrogen may be ruled out as a rate-determining step. On oxides, hydrogen may be reversibly or irreversibly chemisorbed . The occurrence of reversibly adsorbed hydrogen on oxides is of intersst because it resembles the adsorption of hydrogen on metals. The ad- sorption of hydrogen on metals can be visualized as giving M-H linkages, the hydrogen being bonded possibly with the electrons of the d-shell.On zinc oxide, which possesses interstitial zinc atoms, the reversible adsorption of hydrogen of this character on the defects offers no diffi- culty, nor does its adsorption on the lower oxides of the transition elements, where not all of the metal atoms are directly linked to oxygen. In pro- moted oxides, where one of the oxides is easily reduced to metal (for ex- ample, Cu in CuO . ZnO) or on promoted iron-ammonia catalysts, the metallic type of bonding with hydrogen must be present.6 Oxides showing reversible adsorption of hydrogen at low temperatures, on heating may evolve hydrogen, which is re-adsorbed irreversibly on raising the temperature to give OH groups on the surface.13 This explains the fact that at still higher temperatures, the hydrogen can be desorbed only as water.The adsorption of hydrogen atoms on oxides may thus be effected in at least two ways, in one of which it is held as &I-€3 and in the other as M-OH. Many catalytic pi-ocesses may involve the switch of hydrogen from one position to the other. Further investigation may, however, show that hydrogen is adsorbed in other ways in addition to the two mentioned above. indicates that the oxygen mole- cules adsorbed on Cu,O surfaces undergo the following series of reactions : Work described in this Disucssion 1 2 3 0, + 0-0 + 2 0 + 2 0 - . I ? Stage I, giving a chemisorbed molecule, is apparently a rapid process, as is also the case with hydrogen, but it has not been found possible to decide whether z or 3 is the rate-determining step.There is consider- able mobility of the adsorbed oxygen atoms on cuprous oxide, so that on porous surfaces some of the oxygen may escape reaction with other gases, possibly due to diffusion down narrow channels. Thus, in the reaction between CO, and adsorbed oxygen, the gases must be admitted simul- taneously with the carbon dioxide in large excess in order that each oxygen atom may enter into reaction. h similar time dependence of the reactivity of adsorbed hydrogen is found by Eucken in studies of the reduction of cyclohexene, which he ascribes to the separation of the hydrogen atom pairs formed on dissociation by a process of surface diff usicn. The chemisorption of carbon moncxide is very similar to that of hydrogen. The heats of adsorption and behaviour on desoiption indicate that there are two types of bonding which can be represented as M-CO and MOC0,.14 In this connection it will be noted that carbon dioxide can react with adsorbed oxygen to give surface carbonate ions held by electrovalency to the surface.I n the chemisorption of hydrocarbons on oxide surfaces, there is a rupture of carbon-carbon and carbon-hydrogen bonds. There are two theories of the cleavage process, in one of which radicals are produced which adhere to the surface by homopolar bonds ; according to the other, carbonium ions are produced which adhere to the surface by electro- valency. In order to explain the nature of the final products of hydro- carbon alkylation, isomerization and cracking, and the yields in which the products are obtained, it has been convenient in recent years to postulate the formation of fugitive intermediates containing the carbonium 13 Garner and Kingman, Trans.Faraduy SOC., 1931, 27, 322. l4 Garner, J . Chem. Soc., 1947, 1329.INTRODUCTORY PAPER ion. The presence of a positively charged carbon atom in the hydrocarbon chain is regarded as facilitating scission, and the transference of hydrogen atoms and methyl groups along the carbon chain. There appears as yet to be no independent physical evidence supporting the carbonium mechanism, which is highly desirable. Promotion and Poisoning .-The types of promotion and poisoning are at least as numerous as the ways in which solid substances can interact with one another, Mixtures of solids can give rise to solid solutions or chemical compounds. Also, additives can modify the net- work structure of glasses and colloidal aggregates, or can produce various types of interface when adsorbed on a solid surface.Solid solutions are of considerable importance in the semiconducting oxides, especially if the cations and anions entering into solid solution have different valencies from those of the lattice. It is possible to vary the number and nature of the anionic and cationic vacancies by the addi- tion of appropriate ions which will lead to modifications in catalytic activity. Again, vacancies can be poisoned by the adsorption of strongly adsorbing ions on the defect structures. The alkali metals may play this role in a number of catalysts. In the fornation of chemical compounds between oxides (as in the formation of spinels from oxides), Huttig has shown that labile inter- mediate states are first formed on the surface of the less mobile oxide which are more effective in catalysis than the final compound.The characteristics of mixed oxide films on the surface of solids and their nucleation to give a second solid phase form an interesting subject for further research. The new surface lattice structures may be geometrically more uxful for catalysis than those of either component, and they may, on energetic grounds, enter more readily into chemical reaction with adsorbents. The nature of the defects may be modified by compound formation, and from this point of view Anderson has studied the semi- conductivity of the spinels. The importance of the structural acids produced by the interaction of alumina and silica gel is emphasized in this Discussion in the employment of cracking catalysts.Selwood and his co-workers have employed magnetic susceptibility measurements to study the structure of the transition metal oxides, for example, of chromium, manganese, nickel and iron, when dispersed on supporting oxides such as magnesia, y-alumina and rutile. It was shown that above a certain surface concentration the transition metal oxides aggregate into islands of several layers thick which have high magnetic susceptibilities and high catalytic activity. The support may confer its own structure and induce its own valency on the transition metal oxide : thus nickel on magnesia has a valency of two, on y-alumina a valency of three, and on rutile a valency of four.Selwood and Lyon l5 extend this work in this Discussion to other oxides on the same supports. The stabilization of different valencies on the supported oxide obviously must play an important part in catalysis, and the magnetic susceptibility results indicate that the d-shell is involved in the surface reactions. Griffiths, Chapman and Linders l6 have studied the promotion of molybdenum oxide catalysts by silica, making measurements cf changes in composition, surface area, crystal structure, electrical conductivity and thermoelectric potential of the catalysts. It is shown that the area and activity of the catalyst for hydrocarbon decomposition is at a maximum when the maximum amount of silica is incorporated in the lattice.There is some reduction to metal with hydrogen when the silica exceeds 3 yo. The molybdenum metal produced is not, however, believed to play much part in the catalysis, which is mainly due to MOO,. Evidence is brought forward that MOO, is an electron-excess conductor. 15 Selwood and Lyon, this Discussion. 16 Griffiths, Chapman and Linders, this Discussion.W. E. GARNER Insulators as Catalysts .-Insulators, of which silica-alumina and aluminium chloride are examples, bring about the dehydration, isomeriza- tion, polymerization and the cracking of organic molecules. Since the mechanisms of the processes which are involved resemble those occurring in the homogeneous phase with strong acids, it is generally concluded that the sites responsible for the catalysis are strongly acidic.It is also on similar grounds maintained that organic carbonium ions, and oxonium ions, are formed as intermediates during cracking processes, and that an exchange of protons occurs between the adsorbed molecules and the surface. Water present as OH and H ions is regarded as being the source of the protons, since the activity decreases with fall in the water content of the catalyst. In agreement with these hypotheses, the acidity of the surface, as determined by chemical methods, in many cases runs parallel with the catalytic activity and also, as anticipated, poisons of a basic nature are found to lead to a decrease in activity. These questions are discussed in a number of papers in this Dis- cussion. Millikan, Mills and Oblad l o have studied the nature of the structural acidity and conclude that structural acids do, in fact, exist at catalytic temperatures but are unstable at the higher temperatures.They suggest that in some cases structural acids are not actually needed for all catalytic cracking reactions. Since hydrocarbons are weak Lewis bases, all that is needed is an active polar complex of aluminium oxide and silica capable of undergoing change of co-ordination. It is suggested that these complexes are formed at the points of contact of silica and y-alumina grains. May, Saunders, Kropa and Dixon l7 have studied the cracking of I : I-diary1 ethanes on silica-alumina catalysts, using the production of ethyl toluene and methyl styrene as a measure of the total cracking. It is shown that the rate of cracking decreases as the electro- negativity of the hydrocarbon increases, the order being the same as in Hammett 's reactivity series. This supports a carbonium ion mechanism in which there are interchanges of protons between structural acids and the adsorbed molecules. Tamele la ascribes the activity to changes in the co-ordination number of an alumininm-silica complex by the addition of water. This complex in a fully dehydrated state is a Lewis acid and an electron pair acceptor. He ascribes various methods of determining acid strengths of catalysts, and in particular the titration of the solid immersed in benzene with n-butylamine, using p-dimethylaminobenzene as indicator. The activity of the catalysts for a number of processes runs parallel with the acidity. In the polymerization of isopropylene the activity varies linearly with the surface acidity. The University, Bristol. l7 May, Saunders, Kropa and Dixon, this Discussion. Tamele, this Discussion.

ISSN:0366-9033    

DOI:10.1039/DF9500800211

出版商:RSC    

年代:1950

数据来源: RSC

摘要:

W. E. GARNER CATALYTIC ACTIVITY AND COMPOSITION OF OXIDE SYSTEMS BY G. F. HUTTIG Received (in German), 19th January, 1950 Intermediate states in the transformation from an oxide mixture to a chemical compound are important in the theory of catalysts. The transforma- tion does not proceed discontinuously but through individual characteristic states ; it takes place by the diffusion of the mobile component, i.e. the oxide2 16 CATALYTIC ACTIVITY OF OXIDE SYSTEMS of lower melting point, over the surface of the less mobile component, the former later becoming strongly bound and localized. Other stages of activation include internal diffusion of the mobile oxide, the formation of disordered crystalline aggregates and finally the filling of crystal defects. The various stages are linked with the changes of catalytic activity during activation for binary oxide systems.The activity of a catalyst composed of several oxides depends primarily upon its chemical nature, which can be determined by qualitative and quantitative analysis. As a general rule it is valid to suppose that an observable catalytic activity can only be expected if there is some chemical affinity between a t least a component of the catalyst and a component of the substrate which, however, in the conditions under consideration, will just be sufficient to bring about a chemical reaction between the catalyst (or part of it) and the substrate. The complexity of catalytic phenomena in this field is even greater, since systems of the same chemical nature show different catalytic characteristics according to their previous history.The cause of this may be the difference in the magnitude of the reactivity. We shall define the chemical reactivity as the free energy which I mole of the catalyst system in the state considered contains in excess of that in its thermo- dynamically stable state. In general there is little error if one equates the chemical reactivity to the thermodynamic activity which, according to the above consideration, is related to the difference in the total energy. The cause of activity in catalysts consisting of a single oxide lies either in the high degree of dispersion, or irreversible lattice defects, or less often, in the existence of a thermodynamically unstable modification. For catalysts consisting of several oxides, a further possible cause of the activity is the potential energy of the oxide mixture compared with that of its reaction products.Thus, for example, although each oxide alone shows no activity, a mixture of ZnO and Cr,O, will react slowly (perhaps after a very long period) t o form the compound ZnCr,O,, and here the activity of the mixture is larger than that of the product by an amount equal to the affinity of this reaction. From experience we know that the process above does not proceed discontinuously, but passes through a number of intermediate states, which can no longer be regarded as being the original mixture nor yet the final product. A mixed catalyst will thus change continuously in the catalytic furnace to a greater extent than a single oxide.Since it is a fundamental postulate of thermo- dynamics that a series of successive states indicate also a series of free energies in decreasing ordei of magi:itude, it is possible to reduce the changes which occur in a catalyst consisting of an oxide mixture, to the following scheme. The arrangement of decreasing reactivity is of course related only to the total free energy content per mole of catalyst, and not a t all to how this free energy is distributed over the individual molecules. Thus, in principle, we can have in any state, molecules or regions with a high free energy content and others with small energy content. Regions of especially high free energy content are expected a t or near the phase boundaries. Every active substance must take part in every reaction or every chemical equilibrium which has an affinity larger than the reactivity by an amount which would be attributed under similar conditions to the stable meta1c.l This is, of course, also valid for the transient re- actions and processes which take place between catalyst and substrate.Thus, under similar conditions the catalytic activity will be the greater, 1 Thus for example, a stable ZnO in contact with liquid water is in equilibrium with stable Zn(OH), at 39" C, whereas ZnO of thermochemical activity of 930 cal. is in equilibrium with it at 140' C . Hiittig and Moldner, 2. anorg. Chem., 1933, 2 1 1 , 368.G. F. HGTTIG the higher the reactivity. It is therefore not impossible that the total catalytic activity of any one of the intermediate states may be greater than that of the original mixture, since in some of the intermediate states, the activity of the sites accessible to the substrate, and perhaps their number, is greater than in the original mixture.It is a well-known experimental fact that many catalysts require a certain time before reaching their optimum catalytic activity. Mixture Intermediate States Stable Compoud (fi) j - - - - - - - _ _ + @ Activity : A > u ~ > c I ~ > u ~ > - - - A o = e -+ The Ageing of the Catalyst --f FIG. I. This definition of the activity in terms of energy is insufficient fcir an unambiguous description of such a state. It is possible for two catalysts identical in chemical composition and free energy content to behave differently in regard to catalytic activity because of differences in kind and spatial arrangement of lattice defects or constitutional character- istics.In modern chemistry, especially using colloid principles, one can obtain an almost infinite number of states in one and the same chemical compound. Such states (in some cases confined to the surface) can be caused by partial or complete differences in the position of the atoms or mdecules 01 ions. This causes a number of singularities in the fields of force between atoms, which in turn affect in different ways the extent and selectivity of the catalyst. With single oxides there exist as many series of states as there are reactions in which an oxide is involved. Every active state of the solid will still have the characteristics of the solid from which it has been formed.It is clear that the possibilities of preparation will increase as we pass from a single oxide to systems built up from two or more oxides. The variations in preparation most frequently made use of include the following: the stable oxides are ground separately or together to a more or less finely dispersed system, or one can use as starting material reactive oxides with certain properties, or alternatively compounds are put into the catalyst furnace (e.g. hydroxides or carbonates) which at the temperature employed are trans- formed into oxides. Often, small alterations in the method of prepara- tion (as for instance the capacity of the furnace, or the method of stirring, or mixing z, may cause essential differences in the catalytic properties of the resultant preparation.The gaseous atmosphere in which the catalyst is formed has also a considerable influence. To obtain a certain selective activity, it is especially favourable to prepare the catalyst in an atmosphere of the same gas (e.g. by thermal decomposition of an oxy- salt) whose change is to be effected catalytically. Thus, for example, one would prepare the iron catalyst, required for ammonia synthesis, by reduction oi an iron oxide with the mixture (3H, + INJ. The phenomena encountered in this field are comparable in their enormous complexity with those found in organic chemistry. In the literature of catalysis, the development of theory has dealt relatively little with this aspect, in contrast to the predominant part given to it in kinetic problems.The Hiittig and Heinz, 2. anorg. Chertz., 1948, 255, 2 2 3 .218 CATALYTIC ACTIVITY OF OXIDE SYSTEMS majority of the innumerable empirical observations found in the patent literature are almost ignored by those engagcd in fundamental research. An attempt by Huttig 3 to introduce some theoretical order, on the basis of a comprehensive system of reaction types, has remained almost un- noticed. An example of the fundamental dependence of the properties of the catalyst upon its history is shown in Fig. 2 . Two different compounds I TI FIG. 2. were prepared, both having the composition IZnO + 1Fe203. One compound (a) was prepared by mechanical disintegration and ad- mixture of the two oxides, while the other compound (b) was formed by precipitation of the Zn and Fe hydroxides from a common solu- tion.As is well known, these hydroxides are transformed into oxides a t moderate temperatures. Various samples of each compound were heated in the same way, t o different high temperatures T , and, after cooling, their magnetic mass susceptibility x was measured. In Fig. z the temperatures of pre- heating are plotted as abscissae and corresponding values of x x 106 as ordinates. Graphs (u) and (b) refer to the compounds mentioned above. It can be seen that the magnetic character of the two series of compounds is completely different, and the same might be expected of their catalytic properties. In the theory of catalysts, those intermediate states in the transforma- tion from the oxide mixture to the chemical compound will be of special importance.Preliminary information about this is obtained from the experimental results given in Fig. 2. Various samples of the mixture IZnO + ICr203 were heated for the same time in a stream of pure nitrogen to different high temperatures. After cooling, determinations were made of the solubility of ZnO at 40' C in 10 N hydrochloric acid and that of Cr,O, in boiling hydrochloric acid cf maximal concentration. The amounts of ZnO and Cr,O, dissclved a t any instant are plotted as a function of the preheating temperature (ordinates are on a different scale for ZnO and Cr2O3). In the upper half of Fig. 3 , the full 1iI:e cor- responds to the amount of Cr,03 dissolved and the dotted line to that of ZnO. In the lower half of this Figure are given the results of X-ray examination of the compounds prior to solution. Firstly, the transformation of the mixture into the chemical compound does not occur discontinuously but through individual, characteristic, inter- mediate states which, however, are not substances with specific X-ray diagrams.For example in Fig. 3, such intermediate states predominate for the two compounds preheated between 550° C and goo0 C. The mode and extent of catalytic capacity are characteristic of the corresponding system of mixed catalysts. Secondly, in the course of this transfcrmation, both oxides fulfil different functions, often marked by opposite behaviour. Thus increase in the solubility of one component always corresponds to a decrease in the solubility of the other, or vice versa.It is only when the final product ZnCr20, is present that any further alteration is associated with Hiittig, Kolloid-Z., 1941, 94, 258. From these results, one can draw the following conclusions.G. F. HUTTIG 219 a decrease in the solubility of both components. We shall therefore distinguish between a " mobile " component, that with the lower melting point (in this case, Cr,O,), and the " less mobile " component (ZnO). 400 800 b * ' ' ' ' ' l ' ' i t------f +-A <--- (ZnO + Cr203) ZnO ZnCr,O, + Cr203 + ZnCr,O, FIG. 3. - - - ZnO --Cr,03 It is clear that the existence of these intermediate states and the phenomena associated with them, are of special interest in the theory of catalysts composed of two or more oxides. Naturally one cannot expect to solve this problem simply on the basis of catalytic and related phenomena but rather on the broadest possible phenomenological basis.I . . . , I . 260 4bO ' 6b 0 8& I , . . , , , 1 I , , ' , I I 1 a 1 6 1 c Id1 e I f FIG. 4 The experimental methods employed here are the same as those re- ferred to in Fig. 3. The change of some arbitrarily chosen properties of the system IZnO + 1Cr,O, are shown in Fig. 4. The results unless other- wise stated have been obtained by Hu ttig and his co-workers. Graph (I) shows the catalytic activity of the system for the CH30H decomposition ; (2) refers to the same process observed by Jander and Weitendorf, while (3) gives the intensity of the strongest line of Cr,O, in the X-ray diagram, and (4) the strongest line of ZnCr,O,.2 2 0 CATALYTIC ACTIVITY OF OXIDE SYSTEMS Graph ( 5 ) represents the magnetic susceptibilities, and (6) the amounts of water taken up in constant time intervals aiicl water vapour pressures ; (7) shows the adsorption capacity compared with Pb(NO,), dissolved in CH,OH as observed by Starke by means of radioactive indicators. Graph (8) gives the adsorption capacity for eosin solution ; (9) shows the velocity with which compounds become oxidized to ZnCrO,.As these results were obtained by different authors with compounds of not always the same previous treatment, they can only serve a quali- tative purpose. Especial use may, however, be made of the adsorption isotherms of any suitable gas as the interpretation of these by Langmuir and by Zsigmondy, allows us to estimate quantitative and quzlitative changes in the surface and also in the distribution of the radii of capillaries.Systems which have been investigated from this point of view are : A1,0, with Cr,03 ; with Fe,O, ; with SiO,. BaO with Fe,O, ; MOO, ; WO,. Be0 with Cr,O, ; Fe,O,. CaO with AI,O, ; Cr,O, ; Fe,O, ; SiO, ; TiO,. CdO with Cr,O, ; Fe,O,. COO with MnO,. Cr,O, with Fe,03 ; SO, ; TiO,. CuO with Al,O, ; Cr,O, ; Fe,O, ; Mn,O, ; WO, ; ZnO. Cu,O with Cr,O, ; Mn203. Fe,O, with SiO, ; TiO,. MgO with Al,O, ; Cr203 ; Fe,O, ; GeO, ; SiO, ; TiO, ; V,O, ; AVO,. Na,O with SO,. NiO with Al,O, ; Fe,O,. PbO with Fe,O, ; SiO, SrO with Fe,O,. ZnO with Al,O, ; Cr,O, ; Fe,O, ; SiO, ; TiO, ; WO,. Only in a few cases have systems of salts or organic materials been examined with regard to their intermediate states.However, it should bd pointed out that the investigation of catalysts and of powder metal- lurgy, now in a most important stage of their development, have the same scientific interest in regard to the intermediate states. The problem is now one of setting up general rules and laws and, if possible, a model on the basis of the large number of observations. In particular, one can see that at low temperatures (i.e. the initial stages oj the transformation) only those changes which are a function of the surface occur. Only later are changes in the crystal lattice observable. In addition, it seems feasible to distinguish 6 stages in the life-time of an oxide mixture (denoted by the letters (G) t o (f) for the system ZnO + Cr,O, ; cp.also foot of Fig. 4). STAGE ( a ) : This starts almost always with the mixing of oxides a t room temperature and is slightly changed by a moderate increase of temperature. It is mainly distinguished by the mixing which causes an intimate surface contact of the two compounds, such that the surface available to the larger molecules is considerably reduced. Here the main characteristic is a mutual covering of surfaces ; we shall term this the " covering " stage. STAGE ( b ) : This stage is one of activation and is due t o the formation of Zwztter-molecules and to molecular covering of surfaces. The mole- cules of the mobile component (e.g. Cr,O,) diffuse over the surface of the less mobile component (e.g. ZnO), covering it with a molecular layer. Observations by Roulston also indicate the possihility of surface migra- tion of loosely held particles.The original interpretation of the apparent Cf. e . g . the report of the Symposium held by the Sylvania Electric Pro- ducts Inc., Bayside, New York, on '' The Physics of Powder Metallurgy '' (Aug., Roulston, PYOC. Camb. Phil SOC., 1941, 37, 440. 1949).G. F. HUTTIG 221 surface migration of atomic groups in reactions in the solid state was due to Hedvall,6 who assumed that it was concerned with sites of energy disturbance, which only become appreciably mobile structures at the crystal surface. In systems where stage (b) predominates a relatively high catalytic activity is to be expected. STAGE (c) : This comprises a stage of deactivation of the Zwilter- molecules and of molecules covering the suriaces.The foreign molecules migrating over the surface of the less mobile component finally become strongly bound and localized. This process can be regarded as similar to Taylor’s switch from van der Waals’ adsorption to activated adsorpticn. Thirsk and Whitmore observed film formation and the transition into ordered arrangements in the system MgO + Fe,O,. It was observed by Holm* that coagulation of the oxide film on metals led to structures with regular geometrical boundaries. In any case a decrease in chemical reactivity of the molecules involved occurs in this state and with it a de- crease in catalytic activity. From the viewpoint of colloid chemistry we can consider this “ solidification ” as the protective action associated with some protective colloids (Thiesen O ) .According to Schenk 10 molc- cules of Ag,S which have migrated on solid ZnS are less reactive than solid Ag2S. l2 This is a stage of activation due to internal diffusion : as the temperature is further increased, molecules of the mobile oxide diffuse into the crystal lattice of the less mobile oxide. The loosening necessarily associated with this causes an increase in catalytic activity and for the first time the molecules of the less mobile component actually contribute to this activity. STAGE (e) : This stage is the formation of less ordered crystalline aggregates of the addition compound. STAGE (f) : Finally there is the filling of the crystal defects. The processes occurring in the two latter stages are linked with diminution in catalytic activity.It is possible that a part of the action of the mixed catalyst is based upon the Hedvall effect through the continuous transformation within the rigid system, causing an increase in the catalytic effect. Related to this is Schwab’s l 1 ~ l3 treatment of the existence of “ coupled gas-solid- body-catalysis ”. Such an effect has actually been found for the reactions BaO + CuSO, and BaO + ZnSO,. It would be very desirable to extend these investigations not only to processes which in the classical sense go to completion, but to the more subtle transformations among the inter- mediate states. Of the results obtained by varying the ratio of the two components in the mixture, we may note the observation on the ZnO + Fe,O, system. Here minute amounts of Fe,O, added to ZnO promote great catalytic activity, increased adsorption capacity and increased magnetic mass susceptibility.The effect is more marked with smaller than with larger concentrations. The discussion has so far assumed that the two oxide components possess for each other a large affinity which is sufficient for the formation of a classical compound. However, this is not always so ; for example, Be0 + Fe,O, do not chemically combine and do not form solid solutions. STAGE (d) : 6 Hedvall, Chalmers tekn. Hogskolas Handl., 1942, No. 4. * Holm, J . Res. Nut. Bur. Stand., 1942, 28, 5Gg. lo Schenck, Chernie, 1943, 56, 279. 11 Schwab and Karatzas, J . Physic. Chern., 1948, 52, 1053. 12 Cf. the lectures given in the Discussion of the Deutsche Bunsen-gesellschaft on contact catalysis, especially the lecture by R.Fricke which refers to the research of Rienacker. Thirsk and Whitmore, Trans. Faraday Soc., 1940, 36, S62. See also Schwab, Trans. Faraday SOC., 1947. 43, 715. Thiessen, Kolloid-Z., 1942, 1 0 1 , 241. 13 Schwab and Schwab-Agallidis, Kemisk Tidskrift, 1946, 58, 161.2 2 2 STRUCTURE OF OXIDE CATALYST SYSTEMS In spite of this, changes particularly in catalytic activity, may be observed at elevated temperatures. Analogous stages labelled a, b and G, are trav- ersed as before, after which the process comes to a halt. Thus, in this case, the less mobile oxide remains essentially unchanged, while the mobile oxide, or part of it, is found in some crystallographic form on the surface of the less mobile oxide.In a similar manner, we can explain the specific action which a promoter is known to exert on a catalyst (e.g. an oxide). Recently attempts have been made to incorporate the phenomena of intermediate states into a general theory of chemical kinetics.14 The usual definition of reaction velocity, i.e. the amount of reaction products formed per unit time, cannot be applied. Instead, one has to replace the velocity by the reciprocal of the time T required to reach a charac- teristic state with certain properties at a given temperature. Where different reaction mechanisms are not superimposed we may expect Arrhenius' relation to be valid, - In T = - q/RT + In k o . Thus, in 1946 Huttig, Ehrenberg and Kittel calculated from some results on the system ZnO + Fe,O, a t lower temperatures, where the processes are practically determined by surface diffusion, a value q(= activation energy) = 30,000 cal. For higher temperatures where the diffusion of Fe,O, into the ZnO lattice is the rate-determining step, q = 70,000 cal. According to the most recent views 1 4 9 * it might be permissible to identify these values with the activation energy for diffusion of Fe,O, over the surface and diffusion into the interior of the ZnO lattice, respectively. Where the intermediate states are compounds which can actually be prepared, it follows from isothermal considerations, that the last-formed state must be formed at a lower velocity than all the previous states. A discussion of the role of q and K O in the velocity of formation has been given by Roginski.lB A comprehensive description of the theory of mixed catalysts up to 1943, was given by the author in the Handbook of Cataly~is.~7 Institute j i i r Anorg. wad Physikal. Chemie, Graz, Rechbauerstrasze 12. l4 Huttig, Z. Elektrochem., 1950, (in press). 15 Hiittig, Ehrenberg and Kittel, 2. anorg. Chem., 1936, 228, 112. der Technischen Hochschule. Roginski, J . Phys. Chem, RUSS., 1947, 21, 1143. Contribution by Hiittig in the Handbook of Catalysis, edited by Schwab.

ISSN:0366-9033    

DOI:10.1039/DF9500800215

出版商:RSC    

年代:1950

数据来源: RSC

摘要:

2 2 2 STRUCTURE OF OXIDE CATALYST SYSTEMS STRUCTURE OF OXIDE CATALYST SYSTEMS BY P. W. SELWOOD AND LORRAINE LYON Received 4th Jalzuary, 1950 Magnetic methods, in general, reveal the oxidation state and the atomic environment of active oxides of the transition group elements. Susceptibility isotherms have been determined for all these elements, and their interpreta- tion falls into three categories: change of exchange interaction, change of oxidation state, and intercation covalency. There is no obvious relation between magnetic susceptibility and catalyst activity. The susceptibility is a tool for the elucidation of structure, which in turn, determines activity. Nevertheless, i t is not infrequently possible to pre- dict catalyst activity from susceptibility measurements, especially when these are considered together with surface area determinations.P. W.SELWOOD AND LORRAINE LYON 223 These methods are applicable not only t o supported oxides but t o the same oxides in self-supported, or gel-like systems. The susceptibility isotherm method has been shown to yield informa- tion concerning the structure of supported transition group oxides. l This information includes ( a ) the oxidation state of the metal ion in the supported oxide, (b) the possible existence of overlapping wave-functions leading to partial covalent bonds between adjacent positive ions, and (c) the " paramagnetic neighbourhood," i.e. the approximate number of paramagnetic neighbours possessed by each paramagnetic ion. Susceptibility isotherms obtained for the several oxides of the transi- tion elements, when supported on a diamagnetic high area oxide such as y-alumina, fall into several classes, depending on which of the factors mentioned above is dominant, or on which combination of factors the susceptibility depends.Thus, the supported system chromia + alumina 2 gives a typical result in which all the change of susceptibility depends on a modified exchange interaction, or paramagnetic neighbourhood. The system Fe203 + A1203 gives an isotherm the shape of which depends both on the paramagnetic neighbourhood, and the existence of inter- cationic bond^.^ The system consisting of manganese oxide supported on alumina, through moderate temperature ignition of alumina impreg- nated with manganous nitrate, has an isotherm dependent on change of oxidation state, as well as change of paramagnetic neighbourhood. The purpose of the present paper is to extend these observations to supported oxides of vanadium and of ruthenium, and to several other systems not previously investigated.The method, in brief, consists of preparing oxides on high area supports by methods which are, in general, familiar to catalytic chemists, and then to measure the magnetic sus- ceptibility of the product as a function of concentration of the supported oxide, and as a function of temperature. Experimental Magnetic, surface area, and X-ray methods have all been described earlier in papers from this Laboratory. Preparation and analysis of samples is more conveniently described seriatim below. Results Preparation of Samples.-VANADIA + ALuMINA.-samples were prepared by impregnation of y-alumina with ammonium metavanadate solutions of vary- ing concentrations.The alumina had a surface area of 224 m."g. (BET, nitrogen). In general, 10 g. of alumina was impregnated with IOO cm.3 of solution for I hr. The yellow samples were then suction-filtered, dried a t 110' C, and reduced in hydrogen for 60 hr. One series was reduced at 800' C , another at 650" C. All samples were analyzed as follows : the sample was dissolved in I/I sulphuric acid, with heating. The solution was titrated with N/IO potassium permanganate. The sample was then reduced with sulphur dioxide, after which titration with permanganate was repeated. The two titrations gave, respectively, the average oxidation state and the total vanadium. A reduction temperature of 800' C gave samples which contained about 80 yo of the vanadium in the + 3 state; the 650' reduction gave about 70 yo of the vanadium in the + 3 state.In both cases there was no evidence to suggest that the remaining vanadium was in any state but + 4. It may be wondered why reduction was not continued until all the vanadium was in the same ( f 3) 1 This is the eighth paper on the susceptibility isotherm from this Laboratory. The seventh, by Hill and Selwood, appeared in J . Amev. Chem. SOC., 1949, 71, 2522. 2 Eischens and Selwood, ibid., 1947, 69, 1590. 3 Selwood, Ellis, and Wethington, ibid., 1949, 71, 2181. 4 Selwood, Moore, Ellis, and Wethington, ibid., 1949, 71, 693.224 STRUCTURE O F OXIDE CATALYST SYSTEMS oxidation state.But complete reduction to the + 3 state is difficult without causing irreversible changes in the alumina support, and in the distribution of the vanadia. Results on the samples reduced at Sooo C are shown in Table I and those at 650" in Table 11. These Tables give the measured susceptibilities at several TABLE I-MAGNETIC DAT-4 FOR VANADIA + ALUMINA SERIES REDUCED AT 800°c A 298' K 229'5° K 1.8 83.0 2.6 2'5 2'9 69.0 3'2 71.2 86.4 3'2 66.6 2.8 56-6 2'9 3'3 41.2 3'4 4'0 31'9 69-7 67'4 5.6, 5'1, 4'9 48.3 38.9 86 84 90 84 84 2 I0 80 90 I80 222 2 80 3 40 I_ 1'1 56.0 1'3 1.6 46.8 50.0 50-1 48.5 2'1 2-2 4'2, 4.1, 3'9 76.5 2'3 49'1 1-9 40.0 34'0 2'4 30'7 2.6 30'4 3'2 25.7 2'0 1'5 71.6 1'7 59'7 60.0 2'4 2-8 61.9 82- j 2-7 57'2 2'4 49'0 2.5 41'4 3'1 38.8 3'0 2-2 57.5 5.31 4.8, 4'4 34.8 3'6 29'0 2-3 2-6 3'1 84.0 103 87'5 3 *6 83.5 4'4 95'5 6.j, 6.0, 5.8 I02 4'3 3-7 73-0 3'3 53.6 3'7 45'7 3.8 43'1 4'4 35'2 87.7 2.5 3'3 4'1 4'7 4'9 *5-I 5'2 5'5 6.7 8.8 9.5 13'4 * Sample slightly ferromagnetic. TABLE II.-MAGNETIC DATA FOR VANADIA + ALUMINA SERIES REDUCED AT 65oOC A P 2'5 2'4 2-5 2'7 3'8 2'7 2'4 298' K 229.5' K 183.9' K ~27.5' K 2'5 3'3 4'1 *4'9 *5.1 5 ' 2 5'7 1.5 72'5 1-8 64.4 57-8 2-1 3.099 2-98, 2-97 5-69 5.1, 4'9 56.0 86-4 3-0 63-1 2'7 52-6 2.1 98.2 2'5 2-7 73'1 85-4 3.54, 3.70, 3'72 67-8 6.5, 6.0, 5.8 I02 3'9 80.0 3'6 66-6 24 36 74 } 84 210 90 90 40'4 1'7 { 41'7 76.5 4'2, 4.1, 3'9 1'9 42.0 1'7 35'2 * Sample slightly ferromagnetic.P. W. SELWOOD AND LORRAINE LYON 22 5 temperatures, the calculated susceptibilities per gram of vanadium, the Weiss constant in 'C, and the magnetic moment in Bohr magnetons.For instance, the sample containing 2-5 yo vanadium had a susceptibility at 25" C of 1.1 x 10-6 per gram. The susceptibility per gram of vanadium was then 56 x 10-6. The reciprocal susceptibilities thus obtained gave a straight line when plotted against temperature. From this plot the Weiss constant is found, and the moment is then given by Normally, the magnetic susceptibility is independent of field strength, although measurements are invariably made at several fields. In a few cases the vanadia + alumina samples exhibited a slight field strength dependence of susceptibility. Where found, this is indicated in the Tables by including measured susceptibilities at 3630, 4710, and 5450 oersteds, in that order. In such cases the susceptibility per gram of vanadium has been calculated by first extrapolating the measured susceptibilities to zero reciprocal field.But this dubious procedure is further discussed below. FIG. I .-Susceptibility isotherms for vanadia + alumina. Susceptibility isotherms are shown in Fig. I for the 800' C reduction vanadia + alumina series. Chemical analyses for oxidation state, which are not very accurate, are as follows. For the 650' reduction series the average oxidation state was independent of concentration (at -70 y0V+3) except in the region, referred t o below, of the peculiar peak in the susceptibility isotherm. In the region of this peak the average oxidation state was N 66 y'V+3.For the 800' reduction series the average oxidation state was also independent of concentra- tion (at N 80 yOVt3) except in the region of the peak where i t was again ~ 6 6 y ~ V + ~ Vanadia + Ruti1e.-Samples were prepared by the ignition of high area rutile with ammonium metavanadate solutions of varying concentrations. In general, 10 g. samples of rutile were soaked for one hour in IOO ~ m . ~ of solution. H226 STRUCTURE OF OXIDE CATALYST SYSTEMS - 50 "0 - Y ;lc") 3 -40 5 F z 9 .- k Y . a 3 -30 L 300"k - Percent Vanadium ,4 ,6 The samples were then filtered with suction, dried at IOO'C and reduced in hydrogen at 350° C for 18 hr. This reduction temperature was found to yield products containing about 95 yo V+4 and 5 yo Vf3. Lower reduction temper- atures did not give consistent results.It will be noted that the high concentra- tion of V+* found does not constitute evidence for valency inductivity because more V+a could be obtained by raising the reduction temperature. Analyses were performed as described for the vanadia + alumina series. The data obtained are shown in Table 111 and Fig. 2. A few attempts were made to obtain higher total vanadium concentrations in the vanadia + rutile TABLE III.-MAGNETIC DATA FOR VANADIA + RUTILE SERIES 3'3 3'8 4'1 5'0 6.6 x I d 298' K 1-05 29.1 26.8 26.0 23-1 23-6 1-12 1 - 1 1 1-25 1-65 FIG. I - I 1-48 1-59 1-58 1-80 34'0 34'1 27.8 24'4 183.9' K - 1.58 I 'go I .96 1.90 41'5 43'0 43'2 34'2 29-2 2'1 I 127.5' K 2'22 58.5 55'0 51.0 43'4 35'4 2'37 2.38 2-46 2.62 A +34 +34 + 38 + 60 + IO( 2.-Susceptibility isotherms for vanadia + rutile.series. But these gave erratic results, as was not unexpected in view of the strong dependence of susceptibility on the vanadiumloxygen ratio in this range, as reported by Hoschek and Klemm.5 Hoschek and Klemm, 2. anorg. Chem., 1939, 242, 63.P. w. SELWOOD AND LORRAINE LYON 227 Chromia + Rutile.-This supported oxide system was investigated to see if any valency inductivity occurred. It was realized that, because of the easy oxidation to chromate, the results might be difficult t o interpret. Samples of high-area rutile were impregnated with chromium nitrate solu- tions ot several concentrations. The impregnated samples were filtered, dried at 110" C, and then ignited for 4 hr. at 200-250°, except as noted below. The resulting samples ranged from yellow-brown t o dark brown as the chromium concentration increased.They were analyzed by fusion with sodium peroxide. The mixture was dissolved and filtered from some hydrous titania, neutralized and boiled t o destroy excess peroxide, acidified and titrated with permanganate after the addition of excess ferrous ammonium sulphate. The magnetic susceptibilities first obtained showed the surprising result of an isotherm which turned up with increasing chromium concentration. The reason for this was obviously because the lower chromium concentrations con- tained increasing proportions of chromate, which is diamagnetic. It was found that somewhat similar results could be obtained by impregnating y-alumina with chromium nitrate, and igniting, but omitting the usual reduction step.Chromia-rutile samples were prepared through an ignition step a t 165" C for 7 days. These samples were then leached with water until the washings showed no trace of chromate. Two of the resulting samples contained 5-6 yo and 2-6 % of chromium. Some of the preparations showed a slight trace of ferromagnetism, which was not unexpected in view of the well-known occurrence of ferromagnetism in some intermediate oxide of chromium. The magnetic susceptibilities of these two samples indicated gave moments as follows : for the 5.6 yo sample, p = 3.8, A = - 28" ; for the 2.6 yo sample, p = 2.8, A = - 27". These results seem to indicate that the 5-6 yo sample had all its chromium in the + 3 state, and that the 2-6 yo sample had all its chromium in the + 4 state.Copper Oxide + Magnesia.-The system copper oxide + alumina had already been studied in this Laboratory. It was thought worth while t o pre- pare one sample of copper oxide on each of the two supports, magnesia and rutile. A solution of copper nitrate was used for the impregnation of high-area magnesia. The samples were filtered, dried and ignited a t 190" C for 60 hr. The black sample was analyzed by solution in nitric acid, followed by con- version to the sulphate, and electrolysis. The magnetic susceptibilities were as follows : Temperature . . 229" K 184" K 127.5" K It contained 21-7 yo copper. x x I C ~ per g. sample . 3'9 4'8 6.4 x x 106 per g. Cu . 19-4 23'5 30'9 From these data the magnetic moment was found to be 1.6 and the Weiss constant 40" Copper Oxide + Rutile.-Preparation and analysis were similar to those described above.Magnetic results were as follows : Temperature . . 229" K 184" K 127.5" K The sample contained 2-9 yo copper, and was green. x x 106 per g. sample . 1'1 1-22 1'5 x x I G ~ per g. Cu . 31.0 38.0 51'7 These data give a moment of 2 and a Weiss constant of about 26". The data are less accurate than usual because of the low concentration and the presence of only one unpaired electron per C U + ~ ion. Ruthenia + Alumina.-Choice of a solution suitable for use in the im- pregnation step is somewhat more complicated for ruthenium than for other elements studied. Ammonium chlororuthenate, the sulphate, and the chloride were each used. Effective impregnation was achieved only with the chloride, as follows.The metal was fused with sodium hydroxide and sodium peroxide, the fusion mixture was dissolved in hot water, unreacted metal was removed by filtration, and hydrous ruthenium dioxide was precipitated by digestion after addition of alcohol. The washed oxide was then dissolved in conc. HCl. After impregnation on high-area alumina the samples were ignited a t several temperatures from 400" to 660OC. Analyses of the samples, which proved difficult, was achieved by dissolving the samples in I / I sulphuric acid, followed by precipitation of the ruthenium at a controlled pH = 5. The ruthenium was finally weighed as metal, after reduction. Samples were thus prepared ranging in ruthenium content from about 2 to about 17 yo.Magnetic measurements were made a t room temperature and228 STRUCTURE OF OXIDE CATALYST SYSTEMS at - 1 5 0 O C, but in all cases the samples proved t o be diamagnetic with a sus- ceptibility indistinguishable from that of the pure support. Silica-supported Samples-Almost all the work previously reported has involved alumina or titania as the diamagnetic support. Some measurements were made on silica-supported samples.e There seemed no justification for an extensive series for any of these systems, but a few representative samples were made for copper, nickel, chromium, and iron oxides, all prepared through an impregnation on silica. The first, silica I , was obtained by hydrolysis of dilute sodium silicate solution with 10 yo HCl. The susceptibility was - 0.2 x 10-5, approximately independent of temperature.The second batch of silica, silica 11, was obtained by the hydrolysis of ethyl orthosilicate in alcohol with the addition of water and a little HCI. The surface area was 845 m.Z/g., and the susceptibility - 0.3 x 10-6. A pale green sample of copper oxide on silica 11, prepared and analyzed in the usual way, contained 7-9 yo copper. The copper was found to possess a magnetic moment of 1-9, and the Weiss constant was 30'. A black sample of nickel oxide on silica 11, similarly prepared and analyzed, contained 4.3 yo nickel and showed a moment of 2.0 and Weiss constant of 26'. Another sample contained 4-4 yo nickel, p = 2.1, A = 26'. A sample of silica I1 supported chromia prepared by impregnation with chromic acid solution, followed by ignition and reduction contained 9-5 yo chromium.The moment was 3-7 and the Weiss constant 190' A sample of iron oxide supported on silica I was prepared by impregnation with concentrated ferrous ammonium nitrate solution. The deep red ignited sample showed no measureable ferromagnetism. Three samples similarly prepared on silica I1 showed a slight ferromagnetism. The data for these four samples of supported iron oxide are shown in Table IV. Two batches of silica were prepared. TABLE IV I Susceptibility x 106 i % Fe I I I 1 300° K 1 23OK 1 184O K I I -1 3'2 1.1 { 2'5 { 2'0 2'9 I00 1-03, 1-13, I ' I O * 127 - 3-89, 3'76, 3'74* 156 3'9 130 1-82, 1-76, 1-80 191 3'848 3-77, 3-70 I95 6.33, 6.22, 6-08 244 4'6 150 2-12, 2.26, 2-19 228 4'57, 4'55.4'55 230 7-10 286 7-35, 7.161 -~ 70 4'2 } 5 0 4'9 5 0 4'9 }IOO 6.0 * These susceptibilities obtained a t 3630, 4710, and 5450, respectively. The susceptibilities per g . of iron were found by extrapolation to zero reciprocal field. The above data are given in some detail because they illustrate an im- portant point which is often overlooked in magnetochemical studies. Electron Microscope Investigation.-Magnetic susceptibility data on supported oxides give a clear indication that exchange effects persist t o quite low concentrations. The surface areas show that a t these low concentrations there is insufficient supported oxide t o cover the support by any continuous aggregation. It was concluded early in this work that the supported oxides must be aggregated into microcrystals which are separated from each other by large areas of bare support.One picture of the surface which is consistent with the view expressed, is that the surface consists of large uncovered areas with rather regular aggregations of supported oxide in somewhat the way that mesas appear on a desert. If Selwood, J . Amer. Chem. SOL, 1948, 70, 883.P. W. SELWOOD AND LORRAINE LYON 229 this view were correct, then it might be possible t o obtain electron microscope pictures of these aggregates. An electron microscope study of several supported chromia + alumina samples failed to yield any indication of such structures. A similar study on uranium dioxide supported on alumina likewise gave negative results. Discussion Interpretation of New Data.-All the new magnetic data fall into patterns previously established for other supported oxide systems.The vanadia + alumina susceptibility isotherms are much like those pre- viously reported for chromia + alumina, and for several other supported oxide systems. There is one major difference, however, and that is the astonishing peak shown at about 5 yo vanadium. This peak is related to an apparent increase of, moment, but, as will be pointed out below, this effect is probably spurious. The peak is almost certainly due to a trace of ferromagnetic component, or at least to the great increase of susceptibility reported by Hoschek and Klemm for the stoichiometric ratio VO,.,,, which corresponds to the formula V,O,, near which the peak occurs. Absence of the peak for the vanadia + rutile system is doubtless due to the high V+4 con- centration, which makes it difficult for the system to revert to V,O,.Otherwise the isotherm is the normal chromia + alumina type. The system chromia + rutile seems to show evidence of valency inductivity, which puts it somewhat in the class of manganese oxide supported on alumina (as prepared by low-temperature ignition). Copper oxide supported on magnesia and on rutile gives a normal moment and a normal degree of dispersion. The situation is not unlike that previously reported for copper oxide on alumina.' Ruthenia + alumina is an addition to the small group consisting of supported oxides of molybdenum and of rhenium which give zero moment. If our interpretation is correct, this means that valency bonds between adjacent rutheniums are such that the moment is quenched entirely.The several oxides supported on silica appear to offer no feature not previously encountered. Copper, nickel and chromium give moments and dispersions almost exactly similar to those found for the same oxides on alumina. Iron oxide gives the same rather low moment as found on alumina, but this is not true if the sample shows some ferromagnetism. It appears from these results that the main observational character- istics of the susceptibility isotherm, as a tool in catalyst research, have now been found. Some details of interpretation await further develop- ment. This is especially true of the need for a more quantitative inter- pretation of the relation between moment, Weiss constant, and structure. It may be worth while to summarize the several cases in which valency inductivity does or does not appear.The cases reported from this Laboratory are as follows. VALENCE INDUCTIVITY Valency inductivity plays no part in this system. Found Not found Chromia + titania Manganese + alumina Iron + titania Nickel + alumina Nickel + titania Vanadia + alumina Vanadia + titania Copper + alumina Copper + titania Silver + alumina. These results suggest that the effect will not be found if the radius of the supported ion (in its induced form) differs from that of the corresponding diamagnetic ion in the support by more than about 30 yo. 7 Selwood and Dallas, J . Amer. Chem. SOL, 1948, 70, 2145.230 STRUCTURE O F OXIDE CATALYST SYSTEMS We shall conclude this section with a remark concerning the failure of electron microscopy to show any aggregates of the supported oxides.It may be that the precise conditions for viewing the aggregates have not yet been obtained. But it will be noted that the magnetic and surface area data do not require any particular shape for the aggregates. It is possible that the supported oxides are present as long fibres, only a few atoms thick. This view is that commonly held for the structure of so-called hydrous oxides, to which class several of the supporting oxides undoubtedly belong. The Ferromagnetic Correction.-Attention is drawn t o the high apparent susceptibilities and moments in each case (e.g. vanadia + alu- mina, iron oxide + silica) in which a trace of ferromagnetism was present.Traces of ferromagnetism are often present in inorganic solids; it is particularly common to find them in solids containing iron. Such sub- stances may contain magnetite or y-iron sesquioxide in proportions not over a few parts per million, but this is more than enough to give a measureable dependence of susceptibility on field strength. Our policy in all this work on solid oxides has been to make measurements at three fields, so that a trace of ferromagnetism would be detected. The im- portance of this procedure cannot be over-emphasized. But it has also been our policy to reject samples in which such ferromagnetism is found, unless, as for vanadia + alumina, the occurrence of the ferromagnetism is of primary interest for its own sake. A procedure commonly used for samples containing traces of ferro- magnetism has been to plot the apparent susceptibilities as a function of reciprocal field strength, and to extrapolate to I / H = o for the true susceptibility. This procedure seems to give reliable results if all the sample at all times during the measurement is maintained in a saturating field. For spherical particles the demagnetization factor is so large, that the field must apparently be at least 7000 oersteds. The Faraday method for susceptibility measurement generally fulfils this condition. But the Gouy method suffers from the major systematic error that part of the sample is in a field which is far from saturating, indeed it may be near zero. but it has not received adequate attention. The presence of a trace of ferro- magnetism in samples measured by the Gouy method will definitely lead to erroneous and very high apparent susceptibilities, even though the usual correction procedure is employed. The apparent high moments reported for certain concentrations of vanadia + alumina, and for one sample of iron oxide + silica should be ignored. They are included to draw atten- tion to this frequently overlooked source of error. This difficulty has been pointed out by Knappwost,81 This work was done under contract with the Office of Naval Research. Department of Chemistry, Northwestern University, Evanston, Illinois, U.S.A . 6 Knappwost, 2. Physik. Chem. A , 1941, 188, 246. S Knappwost, i b i d . A , 1942, 191, 261.

ISSN:0366-9033    

DOI:10.1039/DF9500800222

出版商:RSC    

年代:1950

数据来源: RSC

摘要:

THE USE OF lSO IN STUDIES OF THE REACTIVITY OF SOLID OXIDES BY E. R. S. WINTER Received 22nd February, 1950 An experimental technique is outlined for following continuously with a mass spectrometer l 8 0 exchange between solid oxides and oxygen gas. Pre- liminary results are presented and discussed for thorium and chromic oxides. The apparent activation energy for oxygen exchange is 27 -f 0.5 kcal./mole for thorium oxide between 440° and 540' C. With chromic oxide two processes are operative, one at temperatures below 4 1 0 O C with an activation energy of 29.5 f 0.5 kcal./mole and one above 4 1 0 O C with an activation energy of I f 4 kcal./mole. The high activation energy is ascribed tentatively t o a limiting process governed by the rate of desorption of oxygen from the oxide surface.The low activation energy is believed t o be associated with the true exchange reaction. The kinetics of these two limiting cases are discussed. As is well known, considerable progress has been made in the investiga- tion and understanding of some heterogeneous reactions by the employ- ment of the isotopes of hydrogen and of ortho- and para-hydrogen.1 Recently the carbon isotopes have been used to study the transformations of certain organic compounds upon catalyst surfaces, a while Joris and Taylor have investigated some aspects of ammonia synthesis, using 15N upon iron and tungsten catalysts. These and other investigations have shown that many catalytic reactions involve dissociative adsorption of the reactants and it is clearly of interest to determine to what extent such a step is important in other heterogeneous catalytic reactions.Suitable reactions for such a study include oxidations taking place upon oxide surfaces, since if dissociative adsorption does occur it is very likely to involve some exchange of oxygen between the reactants and the surface of the oxide. Such exchange could be detected and followed by use of the isotopes of oxygen. Similar exchange might be found in the decompositions and transformations of oxygen-containing organic com- pounds upon solid oxides. In general we would expect a detailed study of oxygen exchange to give useful information upon the nature and reactivity of oxide surfaces and upon heterogeneous reactions. This paper presents some preliminary observations upon the exchange of 1 8 0 between gaseous oxygen and certain solid oxides.Experiment a1 The reaction system is shown in Fig. I.* T is a cold b g e r designed t o prevent any condensible impurities from reaching the mass spectrometer. During the present work it was not found necessary t o use this trap, and after a few experiments it was discarded. A weighed amount of the oxide was placed in the bulb R (capacity about 85 ml.) and outgassed overnight (10-6 t o 10-8 mm. Hg) a t a suitable temperature. The furnace temperature was then rapidly ad- justed t o that a t which the exchange reaction was t o be studied and allowed to 1 Eley, Advances in Catalysis (Academic Press Inc., 1948)~ p. 157 ; Quart. Rev., 1949, 3, 209. Beeck, Otros, Stevenson and Wagner, J . Chem. Physics, 1948, 16, 255. Joris and Taylor, ibid., 1939, 7, 893.* Reproduced by courtesy of the Editors of the J . Chem. Soc.232 l*O IN REACTIVITY OF SOLID OXIDES reach constant conditions over I&-z hr. The enriched oxygen gas was intro- duced in known amount through T, and a very small quantity of the gas (< 10-5 g./hr.) continuously bled off through the capillary leak L into the ionization region of the mass spectrometer. A record was taken of the change with time of the l 8 0 content of the gas. The apparatus is described more fully elsewhere ; the method has been used by Urey and Brandner 5 who discuss in sufficient detail the errors inherent in the use of a mass spectrometer in this fashion t o obtain kinetic data. The results are quoted as velocity constants calculated in min.-l. The mass spectrometer is a Nier-type 60" sector instru- ment ; the masses were brought into focus by voltage scanning.Ta moss specfromefer source FIG. I. The oxides used were : a sample of thorium oxide of ordinary reagent grade and unknown history ; chromic oxide prepared by the method given by Anderson et al. ; 6 magnesium oxide prepared by heating the carbonate in air to about 850OC. The chromium and magnesium oxides were of A.R. quality. The enriched oxygen contained in general 1-2 t o 1-3 yo l80, although a sample con- taining about I yo l80 was used for some experiments. that the exchange consists of two processes occurring with different speeds, and that it is possible t o separate the fast reaction from the slower. We are here concerned only with the former, which we consider t o be exchange between the gas and oxygen ions lying in the surface of the oxide.The slow reaction, which has been found t o continue a t approximately constant rate through at least five ionic layers, we believe t o be a measure of the rate of diffusion of oxygen through the solid lattice. The surface exchange reaction has been found for all oxides so far examined t o be kinetically first order with respect t o residual l80 content in the gas, i.e. we may define the experimental rate constant, k , by the equation Kinetics of the Exchange Reaction.-We have shown elsewhere 4 p An investigation of this reaction is proceeding. - da/dt = k,(a - am), . * (1) where a is the atom fraction of l 8 0 in the gas a t time t and a, that at the end of the surface exchange.We may regard the exchange as being represented by the equilibrium ki k-1 where 0, and 0, refer t o the reacting species from the solid and gas respectively. A plausible assumption, which we make use of here, is that 0, refers to chemi- sorbed, most probably mobile, oxygen atoms, and 0, to oxygen ions in the oxide surface. There is no evidence as to what proportion of the surface oxygen 1 8 0 , + 1 6 0 8 + loog + 1 8 0 a , Winter, J . Chem. Soc., 1950 (in press). Brandner and Urey, J . Chem. Physics, 1945, 13, 351. Bevan, Shelton and Anderson, J . Chem. SOC., 1948, 1729. Houghton and Winter, Nature, 1949, 164, 1130.E. R. S. WINTER 23 3 ions is active in this connection and indeed the proportion probably depends upon the temperature of the exchange reaction and upon the thermal history of the oxide specimen.Extended speculation upon the details of the exchange mechanism is not profitable as yet since correlation is first needed with other experimental figures, notably with the heats and rates of adsorption and de- sorption of oxygen a t the temperatures involved, and with studies of semi- conductivity such as those of Anderson et aLe and Garner et aL8 Nevertheless, the preliminary results so far obtained do indicate the power of this method of approach t o the problems of catalysis and of gas-solid interaction, and emphasize the need for other experimental work so that a proper interpretation can be made of the present observations. We can divide the process with some certainty into three stages : (a) chemi- sorption of oxygen, ( b ) exchange with the oxide surface, (c) desorption of oxygen ; (a) probably involves the formation of atoms and, (c) conversely, their desorption as molecules.We can find no published data which would enable us t o decide between (a) and (c) as rate- determining processes and experimental work is therefore proceeding along these lines in these laboratories. Preliminary results indicate that chemisorption of oxygen is. for chromium and magnesium oxides, and at our temperatures, much more rapid than the observed rates of exchange ; we therefore examine here the cases which arise when (b) or (c) are rate-determining. notation already introduced we use the following : Any of these stages may be rate-determining.CASE (I). EXCHANGE REACTION RATE-DETERMINING.-In addition t o the a , = the initial 1 8 0 content of the gas ; /lo = initial 1 * 0 content of the exchangeable portion of the oxide surface fi = the corresponding l80 content at time t from the start of the reaction ; (assumed to be the normal lSO abundance) ; fi, = the value of B a t the completion of the surface exchange (the a and na = number of exchangeable oxygen ions in the oxide surface/g. of oxide, w = wt. of oxide used in g. ; terms being expressed as atom fractions of leg) ; (i.e. of the species 0,) ; nu = quantity of oxygen gas (expressed as the number of atoms) in the gas phase in the reaction system (including the amount involved in the C, term) ; C, = concentration of the species 0, in the reaction zone, expressed here as a number of atoms/g.of oxide. We may then write the velocity of the exchange reaction, given as the number of atoms undergoing exchange in the reaction system in unit time, as v = - np daldt = nuke (a - cxa) . . = klw(na) (Cg)a(I - S) - k-lw(ns) (C,) (I we may also put with sufficient accuracy for our purpose so that Now at any time during the reaction while a t the end of the reaction and since we have put k, = L1, Eqn. (2)-(6) give us kl = k-l, v = klw(%)(Cu) (a - 8)s (a0 - a)=, = wn,(P - P o ) , (a0 - % ) n u = wn,(B, - P o ) . a , = Pa- * - - k,C, = her*). . . CASE (2). DESORPTION RATE-DETERMINING.-The kinetic law governing the desorption is not known although it seems probable from the worker of Garner, Gray and Stones (cf.also Goodeve and Jack9) that it will be second order, being governed by the speed of recombination of the atoms a t the surface : 0 + 0 + 0,. Garner, Gray and Stone, Proc. Roy. SOC. A , 1949, Iw, 294. H I Gray, ibid. 1949,197,314. Goodeve and Jack, Faruday SOC. Discussions, 1948,4,82.234 l*O IN REACTIVITY OF SOLID OXIDES If this is the case the rate of desorption, expressed as atoms in unit time, is or in general if the desorption depends upon the nth power of the concentration of chemisorbed atoms, where k , and k , are velocity constants. 7J' = K,W(C,)2, . (8) v' = k,w(C,)n, . * (9) n&,(oc - am) = k,w(C,)"(oc - j3) . . (10) We can then write which, using eqn. (4)-(6). gives k3(C,)" = nsker*). . The proper use of eqn. (7) and (11) depends upon determination of C, ; this is being attempted by adsorption measurements, making the assumption that C, will at least be proportional to the quantity of chemisorbed oxygen/g.of oxide. Results and Discussion The results are expressed as values of KO, where The reasons for this method of presentation are given below. Eqn. (7) and (11) involve two quantities, namely C, and n,, of which the former is difficult to measure. Both would be expected t o depend upon the previous thermal history of the oxide ; thus heating to a high temperature, in air or in vucuo, will produce sintering which will reduce the number of active sites in unit area of the surface and so lower both C, and n,. Whether C, and n, are associated with the same active sites is not a t present determinable.Similarly, different times and tem- peratures of outgassing will leave on the surface varying amounts of any strongly adsorbed impurities, which again could lead to changes in n, and C,. In the latter connection it is regrettable that the present experi- mental technique does not permit the use of films of oxide formed in situ in an all-glass apparatus by oxidation of metal films, such as have been used by Garner, Gray and Stone.* This is because at least 1oZo atoms of exchangeable surface oxygen atoms are needed, and this number of atoms will (as clcse-packed ions of radius - 1-4 A) occupy an area of about 8 m.2 In what follows we discuss the rate of oxygen exchange making allowance for probable or known changes in n, and C,. ( a ) Variation of Exchangeable Surface (n,) with Temperature .- With thorium oxide (as the results of Table I reveal) (cf.also y-alumina 4), there was no significant variation of n,, calculated from eqn. ( 5 ) , over the temperature range studied, viz., 446" to 535" C (y-alumina 4ooo-5g2'' C). There was no alteration in n,, determined a t a fixed temperature, when the temperature of outgassing of these two cxides (normally 590°C in both cases) was lowered by 75" C. Chromium oxide showed greater lability, as is seen from Table I, referring t o experiments in which 18-20 hr. outgassing was given. Allowing for experimental error it is evident that n, remains independent of the reaction temperature over the range 323" to 527" C ; a sharp de- crease in ns occurs when the outgassing temperature is increased from 567" to 600" C, whereas between 387O and 567" C, n, is independent of outgassing temperature. This change in n, is unlikely to be due t o alter- ation in the amount of chemisorbed normal oxygen remaining on the oxide at the end of the outgassing since when the chromium oxide was outgassed for only 19 hr.a t 495"C, it gave a value of 1-9 x 1of0 for n,. If, as seems very probable, n, refers t o highly reactive centres on the oxide surface, this drop in n, will be paralleled by a drop in catalytic activity and determinations of n, should provide a sensitive test of the onsetE. R. S. WINTER 235 of sintering. with magnesium oxide.1° Simila.r, but more extensive, changes of n, have been found TABLE I Oxide Temp. O C Outgassing 387 455 455 455 455 455 455 527 567 600 600 5 90 5 90 5 90 5 90 Exchange Reaction 386 323 345 365 391 410 441 527 345 367 374 445 472 502 535 nr x 10-20 atomslg.1 :98 1-94 1-92 1-73 1-99 2-04 1-92 1-28 1-58 5.8 6.1 5'9 6.4 2'00 2-20 (b) Dependence of Rate of Exchange upon Oxygen Pressure.-In- sufficient work has yet been done to determine properly this relationship ; the range of experimental conditions is limited by the difficulty of making capillary leaks which will work satisfactorily at high pressures, and by the need to adjust the oxygen/oxide ratio in the reaction system so that neither (ao - M,) nor (a, - /Ip) are too small. The latter restriction has been imposed hitherto by difficulties in le0 analysis and will eventu- ally be removed by the use of greater enrichments of l80.We have already reported that values of k , for magnesium and chromium oxides are independent of the oxygen gas pressure in the reaction system ; this is also approximately true for y-al~mina.~ The figures given in Table I1 show that K O (eqn. (12)) increases with oxygen pressure €or magnesium oxide (cf. also y-alumina where a similar effect is observed 4, but is ap- proximately constant for chromic oxide ; this probably means that, at least for the first two oxides, C, is still increasing over the pressure range used. It should be noted that these experiments upon pressure-depend- ence were performed before it was found that there are two possible limit- ing processes (cf. (d) below) ; the figures for magnesium oxide refer to the region of low activation energy, and those for chromium (and aluminium) oxide to that of high activation energy.We do not con- sider the independence of k , and oxygen pressure previously reported ' and shown in Table I1 to have any significance. TABLE I1 I Temp. O C I I I Oxide 40'7 79'7 83'5 69.5 139'4 0.028 0.029 0.029 0.037 0.035 kri 0-01 24 0.0161 0.0170 0.023 0'022 Reaction 491 491 493 386 385 Outgassing 1 0 Houghton and Winter (in course of publication).2 36 l8O IN REACTIVITY OF SOLID OXIDES (c) Influence of Outgassing Temperature upon Reaction Velocity. -The figures inset against the experimental points in Fig. z are the tem- peratures of outgassing in "C. With the exception of the points for 387O C outgassing, which are those given in Table 11, and the two points for 600°C outgassing which are discussed in the next section, all the runs were performed at oxygen pressures between 5-7 and 6.4 cm.Hg so that the effect of pressure may be disregarded (cf. below). I t is seen that k o increases slightly as the outgassing temperature is raised from 387' to 455" C, but that a further increase to 567' C has no effect (cf. below for discussion of the effect of outgassing at 600" C). We may remark that y-alumina also appears to be relatively insensitive to outgassing tem- perature,4 but that the sample of magnesium oxide used by us is very sensitive. lo ( d ) Temperature Coeficient of the Velocity of Exchange.-The results plotted in Fig. z clearly indicate that there are two limiting processes for exchange with chromium oxide ; only one has so far been found with thorium (and aluminium 4, oxide, but an extensive examin- ation of magnesium oxide lo has demonstrated the presence of two pro- cesses here also, with activation energies very similar to those for chromium oxide.The activation energy for thorium oxide obtained from Fig. 2 is 27 & 0.5 kcal./mole., while for chromium oxide the two values are 29-5 & 0-5 and I f 4 kcal./mole. The large error estimated for the last figure is due to the rapidity of the exchange reaction in this temper- ature range. In the absence of experimental data from other methods of approach we tentatively assign the lower activation energy for chromium oxide to the exchange reaction proper and the higher activation energy to the desorption of oxygen from the oxide surface.This is done on the grounds that (a) by analogy with somewhat similar systems the energy of activation for the desorption of chemisorbed oxygen is is likely to be moderately high ; (b) the exchange reaction involves the switch of electrons between two very similar nuclei, in similar but not identical environments, and is likely to need a smaller activation energy than (a). Work upon the adsorption and desorption of oxygen is in pro- gress, together with an examination of the influence of lattice defects upon the initial exchange velocity. If the above interpretation of the results is correct, then the rate con- stants in the region of low activation energy should be plotted using eqn. (7), but those obtained in the region of high activation energy should be expressed in terms of eqn.(11). In practice eqn. (12) has been used throughout. However, with thorium oxide the four experiments were performed at oxygen pressures between 6.5 and 7-0 cm. Hg, so that C , will not vary appreciably, assuming other conditions are fixed, unless it is very sharply dependent upon oxygen pressure. The latter seems un- likely since C , must be closely related to the quantity of oxygen adsorbedE. R. S. WINTER 237 under the conditions of the reaction, and preliminary observations of oxygen adsorption do not reveal a strong pressure dependence. Also, the values of n, obtained in the thorium oxide experiments remained sensibly constant (Table I) so that under these conditions and assumptions eqn. (11) reduces to eqn. (12). Considering the results for chromium oxide, excluding for the moment the four runs involving 387" and 600' C outgassing, we note from Table I that n, remains approximately constant for all the experiments.Since these experiments were performed at oxygen pressures lying between 5-23 and 6.4 cm. Hg, while R , for this oxide (in the region of high activation energy) shows no variation with oxygen pressure, it is evident that on the above argument both eqn. (7) and (11) reduce to eqn. (12). Now, both n, and C, are measures of the chemical reactivity of the oxide surface, and it is reasonable to assume that n, =f(C,) ; we will also assume that if n, is unchanged by a change in outgassing conditions, then C,, referred to a fixed temperature of reaction, is also unaltered. If this is true the reason for the two results for 387'C outgassing falling below the rest of the results in Fig.3 is that the value of k , (eqn. (11)) is lower, and con- versely k , is constant (at a fixed temperature) for outgassing temperatures between 455' and 567" C. Referring now to the two results for 600' C outgassing, we see that although n, has fallen appreciably, the points (plotted according to eqn. (12)) lie entirely in accord with the others. Assuming that k , has not changed (at constant reaction temperature) we see from eqn. (11) and (12) that this agreement will occur if This argument is, however, open to criticism since we have no direct knowledge of Fz, and it is quite possible that this does change when sintering sets in. The above discussion shows that a considerable amount of further work is necessary upon oxygen exchange and more particularly upon the adsorption and desorption of oxygen upon these oxides before an unam- biguous interpretation of the results, and an evaluation of R1 and k , is possible.This will then permit the measwement of the activation energies associated with the two limiting processes, since those quoted here are affected to an unknown extent by the variation of C, with temperature. It is pleasing that this investigation completely confirms the views of Anderson 6, 11 that at temperatures below about T,/2 only the surface of these solids reacts with the gas phase, the exchange with underlying oxygen ions being very low, In this connection it is interesting to note that the inflection point in the activation energy plot €or chromic oxide (Fig. 2) is about 410" C, a figure in good agreement with that observed by Anderson et al.6 from conductivity measurements upon sintered discs of chromic oxide. Anderson made no comment upon the inflection points he found for several oxides and spinels in his plots of log (conductivity) against I / T , but if our views are correct these changes of slope are connected with a change in the limiting process concerned. (C,). n, Acknowledgement is made to Prof. H. V. A. Briscoe for his interest and encouragement, to Messrs. I.C.I. €or financial assistance towards the expenses of running the mass spectrometer, and to the Central Research Fund of London University for a grant for the purchase of equipment. Department of Inorganic and Physical Chemistry, Imperial College, London, S. W.7. 11 Anderson, Faraday SOC. Discussions, 1948, 4, 163.

ISSN:0366-9033    

DOI:10.1039/DF9500800231

出版商:RSC    

年代:1950

数据来源: RSC

摘要:

ELECTRONIC CONDUCTIVITY AND SURFACE EQUILIBRIA OF ZINC OXIDE BY D. J. M. BEVAN AND J. S. ANDERSON* Received 30th Jannary, 1950 The electronic conductivity of zinc oxide at 500-1000~ C, in air, varied widely according to the mode of preparation of the oxide. At low oxygen pressures the conductivity is quasi-metallic, and fairly reproducible for different samples, irrespective of their origin. It is suggested that the differences between samples arise largely from the variation in relative importance of surface-conduction and bulk-phase conduction processes. For any one sample, the conductivity in oxygen at pressures greater than 0-01-0-1 mm. varies with oxygen pressure in readily reversible fashion, according t o the law K = A . p,,-t, as required for impurity conduction arising from the reaction 02- = i02 + ze occurring a t the surface.At low oxygen pressures, however, the conductivity becomes independent of oxygen pressure ; equili- brium is attained under these conditions by a process which is slow, structure sensitive and attended with hysteresis. I t is suggested that this involves a diffusion process, which gives rise to a heterogeneous surface a t low pressures. The bulk-phase conductivity of zinc oxide is probably high, but electrons are trapped by chemisorption of oxygen on the surface. It has been found that the conductivity of ferric oxide a t 500-600° displays similar features, a pressure-independent conducting state preceding the dis- sociation to Fe,O,. The measurements reported in this paper were undertaken, in the first place, to clear up some anomalies in the behaviour of zinc oxide as an electronic semiconductor.In the course of the work it became apparent that the semiconductivity was controlled, under the conditions of the experiments, almost entirely by surface reactions. An investigation of the phenomena over a wide range of conditions disclosed some features not previously described, and although these cannct yet be fully explained they are undoubtedly relevant to the problem of the nature of an oxide surface under reducing conditions. A s a typical electronic (excess) semiconductcr, zinc oxide has pre- viously been studied by a number of workers. and Miller,2 working with powder compacts, found the oxide to have a very low conductivity at ordinary temperatures, with a high temperature coefficient (activation energy of the conduction process of 0.71 eV 2 ) .Fritsch,3 to eliminate as far as possible the effects of contact resistances between particles, fused or sintered zinc oxide at high pressures and high temperatures ; his material mas described as markedly coloured, and was a relatively good conductor, with a very small or negative (quasi-metallic) temperature coefficient. Although the properties of this oxide may have been modified-by introduction of impurities or by freezing in some stoichiometric unbalance during the sintei ing process-they may, alter- natively represent true bulk phase properties of zinc oxide. In the latter case the low conductivity of polycrystalline mzterial, as ordinarily pre- pared, must be ascribed to resistances arising at the surface of the grain.Jander and Stamm * Present address : Atomic Energy Research Establishment, Harwell. Jander and Stamm, 2. anorg. Chem., 1931, 199, 165. Miller, Physic. Rev., 1941, 60, 890. Fritsch, Ann. Physik., 1935, 22, 375. 238D. J. M. BEVAN AND J. S. ANDERSON 239 In the investigations cited, the conductivity was measured in air over a narrow or wide 1g temperature range. Wagner,4 in developing the statistical concept of lattice defect equilibria, showed that the semi- conductivity of oxides must, in general, be a function of the oxygen pressure ; in the case of ZnO, if the conductivity be proportional to the concentration of impurity centres (neutral zinc atoms, or electrons trapped near interstitial zinc ions) then K ap&.The assumption, which may not be justified, is thereby made that the electronic mobility and the activation energy E are independent of the stoichiometric excess of zinc. The conductivity of zinc oxide was, indeed, found by von Baumbach and Wagner to follow the expected law over a limited range of experi- mental conditions (Poz = 10 - 1000 mm. ; that, whereas the results of Jarder and Stamm and of Miller are typical of the behaviour of zinc oxide in air, at low oxygen pressures (Po2 < I O - ~ mm.) the conductivity is high and has a small positive or even a negative temperature coefficient. It became necessary, therefore, to make a systematic study of the variation of conductivity with oxygen pressure, and of the range over which Wagner's hypothesis is valid.This paper summarizes the results that have been obtained. T = 550' and 65oOC). More recently we have found Experimental Preparation of Material.-Preliminary measurements were made on zinc oxide prepared from the oxalate For the preparation of further specimens, very pure zinc was deposited electrolytically from zinc sulphate (Merck, pro analysi) . The acid solution was first treated with hydrogen sulphide t o precipitate heavy metals, a few milligrams of pure copper sulphate being first added so that copper sulphide could serve as a scavenger. The acidity of the solution was adjusted, and electrolysis commenced a t low current density and 2.9 V potential difference, to deposit metals more noble than zinc. After 3 hr. a fresh aluminium cathode was substituted, and the potential difference raised to 4.6V.The zinc was finally washed with distilled water, stripped from the cathode, melted in a Pyrex tube, and finally distilled a t 600' and I O - ~ mm. pressure. The product was converted to oxide by three processes. The residue was sub- sequently ignited in air a t 950°, a fresh surface being periodically exposed by stirring the powder. At 950' it may be expected that diffusion should take place freely enough to preclude the persistence of any unoxidized zinc. This oxide had a pronounced yellow colour, which showed no tendency to revert to white, even after the most prolonged ignition at 900-1100'. (ii) A portion of the metal was dissolved in the minimum quantity of pure HC1 ; basic zinc carbonate was then precipitated by adding an excess of freshly sublimed ammonium carbonate and was ultimately converted to the oxide a t 800'.This oxide was distinctly off-white in colour. (iii) A further portion of metal was dissolved in HCl and precipitated as zinc oxalate. Huttig and his co-workers have shown that the physical properties and chemical reactivity of zinc oxide may vary markedly, according to the mode of preparation of the material. Although the three samples prepared by the methods cited should have been of similar chemical purity, they varied widely both as to the magnitude and the activation energy of their semiconductivity (Fig. I ) , although the conductivity in oucuo was markedly reproducible for all samples. In the course of the measurements the oxides were heated t o 950-1000'.This should have permitted the healing of lattice disorder, but not necessarily of bulk recrystallization. Each sample was quite stable in behaviour, and the as described in our previous paper.6 (i) The metal was burned in oxygen in a silica tube. The oxalate, ignited in air a t Soo', gave a pure white oxide. 4 Wagner, 2. physik. Chem. B, 1933, 22, 181. 5 von Baumbach and Wagner, ibid., 1933, 22, 199. 6 Bevan, Shelton and Anderson, J . Chem. Soc., 1948, 1729. 7 Rosencranz, 2. physik. Chem. B, 1931, 14, 407. * Huttig, Kostelitz and Feher, 2. anorg. Chem., 1931. 198, 2 0 6 ; Huttig, I<olloid-Beihefte, 1934, 39, 277 ; Huttig and Goerts, 2. anorg. Chem., 1937, 231, 249.240 ELECTRONIC CONDUCTIVITY OF ZINC OXIDE differences in colour and conductivity may perhaps be associated with differ- ences in particle size, and with the consequent variation in importance of surface and bulk phase conduction.12 I I 10 9 8 I4 13 I I I 1 I I FIG. I.-Conductivity of ZnO in air and in vacuum. of ZnC,O,. 2. ZnO from ignition of metal. I. ZnO from ignition 3. ZnO from ignition of basic zinc carbonate. Conductivity Measurements .-The general technique of conductivity measurements has been described in our previous paper.6 To provide atmospheres of steady, controlled oxygen content, down t o Po, = I O - ~ - I O ~ mm., a flow method was used. Oxygen, or commercial nitrogen containing 0.5 yo of oxygen, was allowed to flow from storage reservoirs A (Fig. z ) , through any selected one of a battery of capillary leaks B, and through n GAS RESERVDlW T- I POT ElkirW VARIABLE LEAK STSTEM ER GAUG Ld FIG.z the conductivity apparatus C to the pumping system. The steady-state pressure was measured by means of a multi-range McLeod gauge. With the leaks used it was possible to expose the specimen for prolonged periods to well-defined and reproducible oxygen pressures between 2.0 and 4 x 10-6 mm. By suitable manipulation of taps and by-pass connections a smooth and rapid change of oxygen pressure could be achieved. For higher oxygen pressures to I atm., static conditions were permissible, since a t such pressures, the ' I gettering " action of metal parts is unimportant. Satisfactory constancy of temperature between 500° and gooo was attained by means of a Sunvic energy regulator controlling the primary of the auto- transformer which supplied the heating current of the furnace.Temperatures were measured with a chromel-alumel thermocouple and potentiometer.D. J. M. BEVAN AND J. S. ANDERSON 241 Preparation of Pellets.-Plates 25 mm. x 5 mm. x 1-2 mm. were pressed at 5-10 t o n s / k 2 . " Fully- aged '' specimens were sintered at IOOOO for up to 20 hr. and can have undergone little subsequent recrystallization. To study phenomena associated with changes in specific surface, later pellets were sintered a t 500-600' for 1-4 hr. ; these had sufficient strength for careful manipulation. General Technique.-The dependence of conductivity upon oxygen pres- sures was first systematically studied a t each of a series of temperatures from 500-800°. It was observed that the behaviour under low oxygen pressures was complicated by irreversible processes of ageing-changes in specific surface, or in the area and perfection of inter-crystalline contacts.The effect of repeated cyclic variations of oxygen pressure a t 592' C was studied accordingly, using a plate of zinc oxide that had previously been sintered at 600' in air for only 4 hr:, so that subsequent ageing processes would show up. To standardize conditions as far as possible, one sample of oxide was separated by sedimentation into fractions of fairly uniform particle size, and the fraction < 8 p was used for subsequent work. With this, the rate of attaining equilibrium a t low oxvgen pressures was investigated, and pressure-conductivity isotherms were obtained under conditions approaching true equilibrium.These were too fragile t o use without sintering. Results and Discussion Fig. 3 summarizes the pressure-conductivity relations obtained at a series of temperatures, using the same plate of zinc oxide throughout. At all temperatures, and over a considerable range of pressures, the con- ductivity K is given by the relation (I) K ; , ~ = APmB, . * (1) where B = 0.25 f 0.02. The result of Wagner and von Baumbach is thus confirmed, and found to be valid down to PO, = 0-01 mm. or lower. /oy uxyyen pressure, mm 1 7 4 1 72 I 10 I 1 2 I FIG. 3. Under these conditions Wagner's interpretation appears formally valid ; the concentration of conduction electrons could be regarded as determined entirely by the shift of equilibrium in the reactions (2u) (as envisaged by Wagner, involving interstitial zinc) or ( 2 b ) (involving vacant oxygen sites) : where Znt+, 0;- represent ions on lattice positions, Zn: represent an interstitial zinc atom, and [7 a vacant oxygen site. Whatever the details of the mechanism, the ionizable impurity centres presumably consist of electrons trapped in the nejghbourhood of the lattice defects.242 ELECTRONIC CONDUCTIVITY OF ZINC OXIDE Equilibrium in reaction ( 2 ) is rapidly attained, as is shown by the establishment ~f a steady and reproducible conductivity within 15-30 min.of changing the oxygen pressure, even at temperatures below 600' C. The melting point of zinc oxide is about ~IOO', and the Tammann tem- perature would accordingly be about gooo C.I t appears probable, therefore, that under all conditions summarized in Fig 3, the rapidity of equilibrium implies that the processes of reaction ( 2 ) can involve only the superficial sheets of atoms in, at most, a layer a few unit cells thick in the zinc oxide crystal. The phenomena will, therefore, be regarded as essentially surface processes. At pressures below about I O - ~ mm, (for the sample used in experi- ments summarized by Fig. 3) the conductivity is no longer uniquely determined by the shift of equilibrium in reaction ( 2 ) : the conductivity becomes independent of the pressure of oxygen, and but little dependent upon temperature. * This state corresponds to the reproducible con. ductivity in vucuo, with a very small temperature coefficient, shown in Fig.I. Its attainment differs completely from the ready reversibility observed at high pressures; the transition to the vacuum condition is I FIG. 4 slow, and attended with considerable hysteresis, with both increasing and decreasing pressures. The resulting step in the log K - log p isotherms is a structure-sensitive property, which shows signs of shifting to lower oxygen pressures on successive heatings. Because the conductivity at higher oxygen pressures has a large temperature coefficient, the magnitude of the difference between high-pressure and low-pressure conducting states diminishes as the temperature is rzised, and at high temperatures (e.g. 800') there is a smooth transition, with no hysteresis, between the pressure-dependent and the pressure-independent conducting states.The effect of successive cycles of pressure changes upon the lower pressure state is shown in Fig. 4. In this experiment the plate was main- tained for I hr. at each pressure in cycle I, and for 2 and 3 hr. at each pressure in cycle I1 and I11 respectively. Instead of eliminating hyster- esis effects, the more prolonged experiments indicate the occurrence of an irreversible process which displaces the step towards lower pressures, and produces a smaller but significant increase in conductivity at higher pressures. The latter fact, taken in conjunction with what is known of * In a series of experiments such as those of Fig. 3, fortuitous changes of contact resistance, etc. (due to shrinkage of the plate) make it difficult to derive a ( K , T ) , plot from the ( K , p ) T isotherms given, or to obtain satisfactorily repro- ducible isotherms when the apparatus is cooled to room temperature between experiments.D.J. M. BEVAN AND J. S. ANDERSON 243 the phenomena of fritting and reaction processes ( e g the work of Huttig, Jander and others) indicates that one factor, at least, in the irreversible ageing is a surface recrystallization, which improves the conductivity of intergranular resistances and effectively changes the relative importance oi surface conduction and bulk-phase conduction. Hysteresis in the transition from pressure-independent conduction is more marked with decreasing pressures than with increasing-pres: ures. To study the rate of the processes involved, a plate of oxide was brought to the pressure-independent state by heating it at a high temperature (800") in vacuum (Po2 < z x 10-6 mm.).The temperature was then lowered to 608" and the pressure of oxygen was increased to, and held steady at, a value which just sufficed to bring about a decrease in con- ductivity. The change of resistance was observed as a function of time in three such experiments, at PO, = 4-5 x IO-~, 1-05 x I O - ~ and 4-4 x I O - ~ mm. respectively; the results are summarized in Fig. 5. This change of resistance was not merely that arising from the rela- tively rapid shift in equilibrium (2) ; the rate-determining process was that responsible for the transition from the low-pressure state to the high-pressure state, The resistivity rose asymptotically to a steady value, and the form of the resistance-time curves is suggestive of a diffusion- controlled process.It is of interest to compare these results with observations made recently by Garner, Gray and Stone on the rate of increase in resistivity of cuprous oxide, a positive hole conductor, as oxygen was stripped from the surface by evaporation into a hard vacuum. The resistivity of cuprous oxide increased according to the relation AR = KtB which can be interpreted as indicating that positive holes are destroyed by a bimolecular process in which oxygen atoms, mobile over the surface, recombine to form oxygen molecules which are subsequently desorbed. The fundamental processes observed on our zinc oxide are just the converse of this : conducting centres are destroyed by the dissociation of adsorbed oxygen molecules, and conversion of oxygen atoms to oxide ions (eqn.( 3 ) ) . * (3) 0, adsorbed + 4e $ z 0 2 - 1 - . 0, gas + 0, adsorbed These processes are operative over the entire pressure range, and are probably rate-controlling for the linear segments of the log R - log p curves ; the kinetic data of Fig. 5 refer to a slower process, superimposed 13 -/- 7ik e, hour~ 4 6 4 1p /2 FIG. 5. on these, when the population of chemisorbed oxygen atoms becomes small. The results described in the preceding sections may be summarized as follows : whereas equilibrium between a zinc oxide surface and oxygen Garner, Gray and Stone, Proc. Roy. Soc. A , 1949, IW, 294.244 ELECTRONIC CONDUCTIVITY OF ZINC OXIDE is readily mobile at temperatures such that mobility within the crystal is still very small (evidence of conductivity at PO, > I O - ~ mm.), a different surface state is attained at low pressures by a much slower process.The characteristics of this process (e.g. pressure of onset of the step, hysteresis) are structure sensitive, and change as grain growth, and fritting proceed. Whereas the conductivity in air varies widely for different zinc oxide preparations, the conductivity is surprisingly repro- ducible when the surface is in the low pressure state, and is comparable with that reported by Fritsch for macrocrystalline material. It appears to us that these results can be interpreted only in terms of a difference in semiconducting properties between the surface and the three-dimensional crystal.The conductivities measured were those of granular aggregates, and even apart from the fact that the true conducting cross-section between grain and grain is unknown, and changes as fritting proceeds, the resistance may reside almost entirely in barriers at the contacts between grains if there is, in fact, a difference in stoichiometric composition between the surface and the interior of each particle. In such a case the overall resistance is likely to be highly sensitive to chemical reactions and chemisorptive processes involving the surface layers of atoms. As has been stated, the linear segments of the curves in Fig. 3 and 4 clearly reflect the maintenance of equilibrium in reaction ( 3 ) at the surface of the zinc oxide ; this equilibrium leads directly to the observed relation K = Ap-1, irrespective of the attainment of inner equilibrium in the crystal.The propagation of reaction (3) into the interior of the crystal when equilibrium is disturbed, proceeding through the creation and migra- tion of lattice defects as envisaged by Wagner, is contingent upon suf- ficiently free diffusion. As such, it certainly occurs at higher tem- peratures and over a much wider range of oxygen pressures than were used in our experiments. Thus Kitchener and DancylO have informed us that the stoichiometric excess of zinc in solid solution in zinc oxide at 1300' is quantitatively controlled by the oxygen pressure as is required by Wagner's model. At the much lower temperatures used in this work the problem is rather to interpret what happens when the surface layers are partially stripped by' evaporation of oxygen. That the log R - log P curves are quite linear over the whole range studied shows that the surface, in presence of I atm.of oxygen, is still not saturated. What is not assured is that in either the surface or the bulk phase there is a perfect stoichiometric balance between zinc atoms and oxygen atoms. Several possible models for the relation between surface and bulk phase suggest themselves ; an assessment of their validity must be made in the light of further work. (i) If the crystal of zinc oxide approximates to the ideal composition ZnO, de-oxygenation of the surface in vacuum must give rise to surface atom layers that are richer in zinc than is the interior of the crystal lattice.On this model the conductivity of the surface in vacuo should be greater than that of the bulk phase, but the consequences of the model are worth considering in relation to dissociation, reduction and surface reactions proceeding at very low partial pressures of oxygen. The re- sulting surface state could equally well be regarded as attainable by the chemisorption of zinc atoms on the zinc oxide surface ; the concentration of ionizable impurity centres would then be the surface concentration of zinc atoms Nzn, which could be related to the dissociation equilibrium of the oxide (4), by an adsorption isotherm of the Langmuir type (5) ZnO, + Zn, + 902, . - (4) 10 Kitchener and Mrs. Dancy (private communication).D. J. M. BEVAN AND J. S. ANDERSON 245 The conductivity K measures the concentration of free electrons, and so is proportional to ~ N X , i.e. whence R, the measured resistance of the specimen at the pressure p is related to the limiting resistance Ro at very low pressures, by the ex- pression No such relation can give rise to a stepped curve, but (7) represent reasonably well the form of the 800° isotherm, with a smooth transition from the pressure-dependent to the pressure-independent states.(ii) Several lines of evidence suggest that the three-dimensional crystal of zinc oxide is a good semiconductor, with a quasi-metallic temperature coefficient of conductivity. If-as has generally been considered-this conductivity is an impurity conduction process, it follows that the in- terior of the crystal must be oxygen-deficient to some slight extent (ZnOl_,) whereas at the surface the chemisorption of oxygen forms almost stoichiometric, non-conducting zinc oxide.The low conductivity of a sintered plate of oxide would then be attributable entirely to barrier resistances at the contacts between grains. The displacement of equi- librium between oxygen molecules and chemisorbed oxygen would then lay’ bare the substrate of conducting oxide. Over the free surface attain- ment of equilibrium could readily be reached ; the behaviour of material in the inter-granular bridges might be associated with the step in the log R - log p curves. (iii) The assumption that zinc oxide is an impurity semiconductor may be at fault; the conductivity may be intrinsic, arising from the excitation of electrons from the d levels of the Zn2+ ions to the conduction band.At the surface the chemisorption of oxygen would, in this case, trap electrons from the conduction band by converting oxygen atoms to 0 2 - ions, on surface lattice points. This behaviour would parallel the effect of sulphur upon the conductivity’ of a-Ag,S, as interpreted by Reinhold l1 (though in this instance the high diffusion rates and the tendency of silver towards true bivalency lead to a bulk effect, and not merely a surface reaction). Whatever model be accepted for the conductivity of zinc oxide, the behaviour described in this paper, at low oxygen pressures, seems to require some heterogeneity of surface. Hysteresis in attaining equilibrium at low pressures might in itself be explainable in terms of diffusion along inner surfaces-eg.in the material bridging inter-granular contents. The ultimate equilibrium conductivities should, however, fit the K = A . p-t relationship. This is not the case, and as an alternative hypothesis we would suggest that the slow process associated with the step of Fig. 3 and 4 represents the creation, by surface diffusion, of a heterogeneous surface. This could either be regarded as consisting of regions of a normal zinc oxide surface with zinc-rich regions (2-dimensional metal), or as regions of exposed, conducting substrate with patches richer in chemi- sorbed oxygen. Whether the change is regarded as brought about by an increase in population of chemisorbed zinc atoms, or a decrease in population of chemisorbed oxygen atoms, the process could be envisaged as a separation, by a diffusion-controlled process, into two surface phases. This model is suggested purely as a working hypothesis, to be rejected or modified in the light of further experimental evidence. To examine how far the experimental facts were typical of other instances, a few experiments were carried out on or-Fe,O,, at temperatures such that no chemical conversion to magnetite could occur. The results, indicated in l1 Reinhold and Schmitt, 2. physik. Chem. B, 1939, 44, 75. R2 - Ro2 = const. x p ~ * . - (7) This point is considered further below.246 SURFACE REACTIONS O F COPPER OXIDE Fig. 6, show that in this case also, at low oxygen pressures, the con- ductivity becomes independent of the oxygen pressure and (at the lower /os pressure, mm. -5 - 4 -3 -2 -I 0 1 2 FIG. 6.-Conductivity of ferric oxide as function of oxygen pressure. temperatures) higher than is explicable in terms of the equilibrium which governs the conductivity at high oxygen pressures. Irrespective of the validity of our tentative model this state ot the surface is not without significance for the properties of oxides under the conditions obtaining during their action as catalysts for a variety of gas reactions. I I University of Melbourne, Australia.

ISSN:0366-9033    

DOI:10.1039/DF9500800238

出版商:RSC    

年代:1950

数据来源: RSC