摘要:

hfra-red Studies of Carbon Monoxide and HydrocarbonsAdsorbed on Silica-Supported NickelBY J. B. PERIResearch and Development Dept ., American Oil Company,Whiting, Ind., USA.Received 3rd January, 1966Infra-red studies of NH3, HC1, CO, C2H4, and butene adsorbed on transparent aerogels weremade to elucidate the surface chemistry of nickel/silica and the nature of adsorbed species. Inaddition to the hydroxyl band found on silica at 3750 cm-1, most “ dry ” nickel/silica aerogelsexhibited a band near 3620 cni-1, apparently caused by silanol groups near nickel ions. Reductionof ail nickel ions by 1-12 treatment proved very difficult. Nickel ions remaining on “reduced”gels were apparently the acid sites that adsorbed NH3 strongly. Chemisorption of HC1 showedthat reactive oxide ions, not found on silica, were also present.Bands between 1800 and 2210 cm-1showed at least 7 types of adsorbed CO, varying in relative importance with the extent of reduction.Three distinct ionic sites for CO adsorption appeared to exist. Bands near 2200 cm-1 arose fromCO on the sites which held NH3 most strongly, implying binding to a nickel ion. Reaction ofHCl with these sites, however, showed that they also contain oxide ions.Adsorption of ethylene on “ reduced ” nickel/silica produced 1- and 2-butenes which werequickly converted to butane on subsequent addition of H2. Spectra ascribed by others to associ-atively adsorbed ethylene (CH2-CH2) and to ethyl groups on reduced nickel probably insteadshow butenes and butane.I ICatalysts prepared by reduction of metal salts on a silica gel support are of bothpractical and theoretical interest, but, despite intensive investigation, the nature of suchcatalysts remains poorly understood.Nickel/silica catalysts have probably been themost thoroughly studied.1 The silica support, through its high surface area, promoteshigh dispersion of the reduced metal and may otherwise alter its catalytic properties.Major differences exist between nickel/silica catalysts, depending on method ofpreparation and pretreatment.1 Reduction of supported nickel can be much moredifficult than would be expected from the reduction of pure nickel oxide. Nickel ionscan significantly increase the acidity of pure silica 2, 3 and cause catalytic activityfor polymerization of ethylene and isomerization of olefins.2Many infra-red studies of CO and/or hydrocarbons adsorbed on nickel/silica andother catalysts have been reported,4-19 but results to date leave many questionsunanswered.Little attention appears to have been given to possible changes in thespectrum of silica caused by nickel. Spectra of adsorbed CO on nickel/silica showseveral types of adsorption, which are inadequately understood at present. Bandsabove 2140 cm-1 are of particular interest. A band near 2200 cm-1 was ascribed byEischens and Pliskin 4 to an active intermediate in catalytic oxidation of CO havingthe probable structure Ni. * .O-=C=O. A similar band in the range 2175-2190 cm-1on supported nickel oxide and also, at higher pressures, even on reduced nickel/silica,was attributed by O’Neill and Yates 5 to CO loosely attached to oxide ions in thesurface.Similar bands have since been reported for CO adsorbed on several oxideswhich do not contain nickel. Amberg and Seanor 6 have concluded that such bandsreflect adsorption through ion-dipole interaction in the electrostatic force fieldprovided by oxide ions. Bands near 2200 cm-1 have also been attributed to CO ions,12122 ADSORPTION ON SILICA-SUPPORTED NICKELranging up to CO2+ (presumably generated by nietal ions in the surface),7 and toweak a-complexes with metal ions in the suiface.8Spectra of ethylene on nickel/Cabosil were in certain cases thought by Eischensand Pliskin9 to represent saturated C2 species attached to metallic nickel, but theyalso noted weak olefinic C-H stretching bands and some polymerization. Thepowder samples used in their work scattered radiation badly, however, and handi-capped infra-red study.Later studies 10, 11 of ethylene on nickel/porous glass werealso handicapped by poor transmission and complicated by catalytic activity of theporous glass support.12Additional infra-red study of adsorbed CO and hydrocarbons on nickel/silicaseemed clearly warranted. Prospects for increased sensitivity and resolution in suchstudy were improved by preparation of transparent plates of nickel/silica aerogeland by the availability of a new prism-grating spectrometer.EXPERIMENTALAPPARATUS A N D PROCEDURESMost of the apparatus has been described elsewhere.13914 A conventional glass vacuumsystem was used.?he Perkin-Elmer models 12C and 112 (LiF and CaF prisms) used inprevious studies were supplemented by a Beckman model IR-9 (prism-grating) spectrometer,in which the cell illustrated in fig. 1 was permanently mounted. The sample holder wasmade from Vycor tubing split lengthwise (semicircular in cross-section). It fit closely in thecell to ensure reproducible positioning of the sample. The holder was moved verticallybetween the window section and the oven section by an external magnet that acted on aPyrex-enclosed magnet which was connected to the holder by a silica fibre, as shown. A cellused with the Perkin-Elmer 12C spectrometer permitted concurrent gravimetric study,13but did not allow spectroscopic study below 2000 cm-1.Procedures were generally as previously described.l3* 14 Calcination of aerogel platesin oxygen was typically for 11.1 at 200 torr.Reduction was normally carried out, withvariations as described below, by admitting H2 to the cell at 50 torr and heating at 400°Cfor 1-3 h. The H2 was completely replaced at least once during this period. Typicalaerogel samples weighed 0.2-0.3 g and provided a surface area of 25-30 m2/cm2 in the infra-red beam. Spectra were normally recorded, and gases admitted, after the samples had cooledto room temperature. Spectrometer settings varied considerably for different samples andspectral regions, but slit widths were usually less than 3X normal.MATERIALSNickel/silica aerogels were prepared in various ways in attempts to ensure uniformdistribution and maximum surface area of the nickel on the silica support.Five diEerentcatalysts were studied. These were prepared as follows.NS-1 (4.5-5-1 % Ni).-Silica hydrogel plates 14, 15 were washed repeatedly, first with dis-tilled H20, then with ethanol, and finally with n-pentane. Then they were allowed to equili-brate with a n-pentane solution of nickel oleate. Excess nickel oleate was washed out withfresh pentane; n-pentane was removed above its critical temperature. The dry aerogelplates were greenish-brown to black as removed from the autoclave, and transparent mediumbrown after calcination at 500-600°C in air.NS-2 (9.8 % Ni).-Silica hydrogel plates were washed with H20 as above and allowed toequilibrate with an ammoniacal solution of nickel acetate.The solution was then replacedwith methanol, and the methanol was removed above its critical temperature. The finishedplates were very dark brown. The nickel was slightly extracted and reduced during theautoclaving.NS-3 (23.1 % Ni).-Silica hydrogel plates were washed with € 3 2 0 and soaked in asolution made by dissolving 50 g of Ni (NO3)2.6H2O in H28 to make 500 ml, adding 170 mlof concentrated (29.9 % NH3) NH4OH while stirring, and filtering. The plates were ageJ. B. PER1 123in this solution for 6 h at 100°C in a sealed autoclave. Then the solution was replaced withmethanol, and the methanol was removed as above. The finished plates were transparentgreen. Slight colour variations showed that distribution of the nickel was not completelyuniform throughout the plates.NS-4 (24-0 % Ni).-These plates were prepared in the same way as NS-3, but the methanolwas replaced with ethanol, and the ethanol was replaced with n-pentane before the finalautoclaving. These plates were transparent and uniformly green throughout.The surfacearea of NS-4, as measured by nitrogen adsorption after drying at 5OO0C, was 575 m2fg.(Areas of the other aerogel plates were not measured, but they were certainly all in the rangefrom 500 l o 800 m21g.) 27 II 'fa-c----- - bFIG. 1 .-Infra-red cell. (a) Aerogel sample ; (b) sample holder ; (c) silica rod at 45" to plane ofsample (fitting diagonally in cell) ; (d) cell body of 1-in.square silica tubing (CaF2 windows waxedto cell) ; (e) furnace ; (f) Pyrex-enclosed magnet ; (g) silica fibre.NXCKEL/CABOSIL ("7 % Ni).-Five grams of Cabosil (G.L. Cabot Inc.) was added to2 g of Wi(NO3)2.6H2O in 25 ml of distilled H20 and the mixture was dried on a hot plate.To prepare a transparent disc, 0.25 g of the dry powder was pressed in a la in. die under12 tons pressure.Airco Assayed Reagent CO was used without further purification. Ethylene (PhillipsResearch Grade) was passed over P2O5 and degassed by freezing and evacuation beforeadmission to the infra-red cell. (Chromatographic analysis showed traces of ethane andmethane as impurities.) Hydrogen was purified by passage through a trap filled with activatedcharcoal and cooled with liquid nitrogen.Nickel oleate was prepared from nickel formateand oleic acid. All other reagents were commercial materials of high purity, further purifiedin some cases, as previously described.13~ 14RESULTSSPECTRAL CHARACTERISTICS OF NICKEL/SILICAThe spectra of nickel/silica aerogels predried by evacuation above 400°C generallyresembled those of comparably dry silica aerogel ;15 the isolated hydroxyl band foun124 ADSORPTION ON SILICA-SUPPORTED NICKELon dry silica at 3750 cm-1 was always present. Some differences were noted, however.An unusual hydroxyl band, near 3620 cm-1, as illustrated in fig. 2, characterizedNS-2, NS-3, and NS-4 but was not seen for NS-1 (which was lowest in Ni content)or nickel/Cabosil. It was much less prominent on NS-2 than on NS-3 and NS-4.This band was prominent on oxidized samples and remained, very little altered, aftertreatment with H2 at 500°C.3500 3 6 0 0 3 7 0 0 3003 3900wavenumber (cm-1)FIG.2.-Hydroxyl-stretching bands for NS-3. (a) Dried at 500°C and reduced in H2 at 400°C;(6) after exchange with D2 at 500°C ; (c) after heating in H20 vapour and drying at 600°C ; ( d ) aftersubsequent drying at 800°C.The hydroxyl groups on NS-3 causing the 3620cm-1 band exchanged H withgaseous D2 at 500°C at essentially the same rate as the groups responsible for the3750cm-1 band, but the corresponding OD groups exchanged more slowly withNH3. (The NH3 was first adsorbed at room temperature and then removed byheating and evacuation at 200 and 400°C.) The relative intensity of the two bandswas unaffected by long calcination in 0 2 at 600°C or by " steaming " in 15 torr ofH20 vapour at 600°C followed by evacuation at 600°C as shown in fig.2. Subsequentevacuation at 8OO"C, however, reduced the 3620cm-1 band much more than the3750 cm-1 band, and shifted the frequency as also shown in fig. 2. For NS-3 driedat 6OO0C, hydroxyl combination bands were found at 4535, 4305, and (weaker)4120 cm-1. Only a band at 4520 cm-1 remained in this region after the plate hadbeen dried at 800°C. Because the 4520-35 cm-1 band apparently arises from com-bination of the 3750 cm-1 stretching vibration with an out-of-plane deformation ofsilanol groups,15 the 3620 cm-1 band may be similarly related to the 4305 cm-1and/or 4120 cm-1 bands.Strong bands that normally occur near 1640 and 1870 cm-1in the spectrum of dry silica appeared near 1680 and 1840 cm-1 for NS-3 and NS-4.These bands are thought to be caused by combinations and/or overtones of Si-0vibrations of the silica latticeJ. B. PER1 125ADSORPTION OF NH3 AND HCIAdsorption of NH3 indicated that nickel ions were left in the surface after '' re-duction ". NS-1, NS-3 and NS-4 all held NH3 strongly after they had previouslybeen treated with H2 at 400-450°C for 1 h or more. (NS-2 was not studied.) Notall of the adsorbed NH3 was removed by evacuation for 1 h at 300°C. Major bandsnear 3400, 3300 and 1630 cm-1 indicated that it was retained mostly as NH3. Noevidence was found for NHZ (no band near 1460cm-1). The adsorbed NH3resembled that held on Lewis acid sites on dry silica-alumina 9 rather than physicallyadsorbed NH3 or -NH2 on silica.15Chemisorption of HC1, which does not occur on pure dry silica,ls showed thepresence of oxide ions different from those normally found on silica.Treatmentof NS-3 or NS-4 (reduced and predried at 450°C) with HCl at room temperatureproduced an intense and very broad hydroxyl-stretching band extending from3750cm-1 to below 28OOcin-1. A small band also appeared near 1630cm-1,indicating formation of some H20.ADSORPTION OF coSpectra obtained when CO was adsorbed on NS-4, previously calcined in 0 2 anddried by evacuation at 500"C, are shown in fig. 3. In the range from 4.000 to 1300 cm-1new bands appeared only between 2080 and 2210 cm-1.Above 0.1 torr, the mostintense band occurred at 2198 cm-1. Below 10 torr its intensity was roughlyproportional to the CO pressure, showing that many adsorption sites capable ofholding such CO remained empty. Below 0.1 torr, a smaller band at 2210 cm--1became increasingly evident and at 0.015 torr was the only band left. The bands at2198 and 2210cm-1 were reversibly restored on re-addition of CO and remainedunchanged on long standing; they clearly represent CO weakly held in two distinctways. A third band was seen near 2140 cm-1. Although this frequency is near thatof gaseous CO, the CO causing this band was held fairly strongly. Since its intensitydid not increase in direct proportion to the pressure, this band apparently representsCO on sites that were largely filled at low pressures.A weak band was also noted at2080 cm-1. (The spectra shown in fig. 3-5 include, at higher pressures, a smallcontribution from gaseous CO, evidenced by resolved fine structure. This contribu-tion does not significantly alter the contours of the bands caused by adsorbed CO.)After this same NS-4 sample had been heated in 50 torr of H2 for 1 h and evacuatedfor 1 h at 3OO"C, it was cooled to room temperature, and CO was again admitted.The spectra then recorded are represented in fig. 4. Comparison with fig. 3 revealsthat the band near 2198 cm-1 was slightly reduced in intensity, and the band at2210 cm-1 was nearly eliminated. The 2140 cm-1 band, however, increased several-fold in intensity, remained fairly intense after evacuation for 1 h at room temperature,and was not completely removed by evacuation (4 h) at 200°C.New weak bandsappeared at 2180 and 2105 cm-1, and the band near 2080 cm-1 was stronger. Athigher CO pressures (up to 27 torr) a band was also seen near 2050 cm-1.The same sample was then heated for 1 h at 450°C in 50 torr of H2, dried byevacuation, cooled, and CO was added as before. The spectra showed further changesin band intensities. After additional reduction in fresh Hz for 1 h at 450"C, evacuationfor .?z h, and cooling, the aerogel was re-exposed to CO. Spectra then recorded areshown in fig. 5. The major effects of reduction at 450°C were a marked decrease inthe band at 2135 cm-1 and an increase in the 2050-2080 cm-1 region.A band(not shown) also appeared near 1960cm-1, and weaker bands were noted near1850-70 cm-1126 ADSORPTION ON SILICA-SUPPORTED NICKELSeveral further treatments with H2 at 450°C were carried out in the course ofexperiments, to be described below, involving adsorption of NH3. When CO wasadded to the cooled sample after these treatments, the bands near 2200 cm-1 persisted,very little diminished in intensity. The bands near 2050-80 cm-1 remained strong,and the band at 2135 cm-1, weak. The band at 1960 cm-1 was weaker than afterinitial reduction at 450°C.2000 2 too 22cowavenumber (cm-1)FIG. 3.-Adsorption of CO on unreduced NS-4.(a) Before CO addition; (b) 9.4 torr; (c) 2-8torr ; ( d ) 1-45 torr ; ( e ) 0.7 torr ; (f) 0.19 torr ;(9) 0.05 torr.C- U02 0 0 0 2100 2200wavenumber (cm-1)FIG.4.-Adsorption of CO on NS-4 reduced at300°C. (a) Before CO addition ; (b) 27 ton ;(c) 9 7 torr ; (d) 3-2 torr ; ( e ) 0.6 torr ; (f) 0.18torr ; (9) 0.05 tom ; (h) after evacuation for 1 h.The effect of prior chemisorption of HC1 on adsorption of CO on this sample wasthen investigated as also described below. Afterwards the sample was again exposedto H2 at 450°C (3 h, 3 changes of H2). Subsequent CO adsorption on the cooledsample gave a band near 2200 cm-1 which was much weaker than initially, whilebands near 2050 and 2135 cm-1 were much stronger. The bands near 2050 and2135 cm-1 appeared at slightly lower frequencies after brief evacuation. The NS-4sample was then heated in H2 for 2 h and dried by evacuation for 1 h at 550°C.After it had been cooled and CO admitted, the spectra shown in fig.6 were obtained.The band near 2200 cm-1 was almost absent, but a band at 2145 cm-1 remainedfairly strong. At least 3 bands were present between 2040 and 2100 cm-1, thestrongest being at 2050 cm-I. After evacuation for I h, bands were left at 2148 and2040 with a shoulder at 2055 cm-1. The hydroxyl band at 3620 cm-1, although lessintense, still remained after this series of treatments. When the sample was removeJ. B. PER1 127from the cell, a mirror-like deposit of metallic nickel, covering the area illuminatedby the infra-red beam, was found on the face nearest the infra-red source.NS-1 was apparently reduced more readily than NS-4.After an NS-I samplehad been treated with W2 (50 torr) for 3 11 at 400°C (with 4 changes of M2), evacuated,and cooled, addition of CQ (2 torr) produced bands very similar in intensity andfrequency (2075, 1930 cm-1) to those reported by others.99 16 A weak band near2200cm-1 was also noted, however. This was the only band observed when COwas adsorbed on oxidized NS-1 before reduction in H2.-I0-f- d2000 2100 2200wavenumber (cm-1)FIG. 5.-Adsorption of CO on NS-4 reducedat 450°C. (a) Before CQ addition ; (b) 13.1torr; (c) 4-05 torr; (d) 0.29 torr; (e) afterevacuation €or 1 h.I 1 I 1wavenumber (cm-1)30 2100 2200FIG. 6.-Adsorption of CO on NS-4 reducedat 550°C. (a) Before CO addition; (b) 5-7torr ; (c) after evacuation for 15 min.EFFECT OF PRIOR ADSORPTION OF NH3 ON ADSORPTION OF COAfter the NS-4 sample had been treated in Hz at 450°C as described above andcooled, NH3 was admitted (5 torr) and adsorbed.Shortly afterwards the cell wasbriefly evacuated, CO was admitted (6.4 torr) and a spectrum was recorded. Thesample was then heated to 100°C and the cell evacuated for 3 h to remove CO andsome adsorbed NH3. After the sample had been cooled, CO was readmitted (6.4 torr)and a spectrum recorded. The procedure was repeated after evacuation at 200, 300and 400°C. Adsorbed NH3 initially prevented the appearance of all the bands tobe expected from the adsorbed CO, except those near 205Qcm-1. As NH3 wasprogressively removed, the other bands reappeared when CO was subsequently added.The band near 2140 cm-1 was restored after evacuation at 200°C (I h), and that at2198 cm-1 was largely restored after evacuation at 300°C128 ADSORPTION ON SILICA-SUPPORTED NICKELEFFECT OF PRIOR ADSORPTION OF HC1 ON ADSORPTION OF coAfter further reduction of the NS-4 sample in H2 at 450"C, CO adsorption gavespectra generally resembling those in fig.5. After a 15-min evacuation had removedmost of the CO (a band remained near 2060 cm-I), HCl was admitted (5.5 torr),I I I Ibii2000 2100 2 2 0 0wavenumber (cm-1)FIG. 7.Effect of chemisorption of HCl onsubsequent adsorption of CO. (a) NS-4after exposure to HCl and evacuation for15 min ; (b) after addition of CO (5.7 torr).and the cell was again evacuated for 15 min.Spectra were obtained before and after ad-mission of CO (5.7 torr).As illustrated infig. 7, prior chemisorption of MCl markedlychanged the appearance of the bands pro-duced by adsorbed CO. The strong bandexpected near 2200 cm-1 was almost gone,but the band near 2135 cm-1 was greatlyincreased.When MCl was chemisorbed on a freshsample of NS-4 (previously calcined in 0 2and dried by evacuation at 500"C), subse-quent CO adsorption produced a moderatelystrong band at 2135 cm-1 and a weak bandnear 2200 cm-1, resembling the bands offig. 7 except that the former was relativelyless intense. After evacuation at 500°C andre-adsorption of CO, the band near 2135 cm-1was less intense, while the band near 2200cin-1 was considerably more intense thanbefore.After a final evacuation at 600"C,the aerogel plate appeared chalky, havinglost its original transparency.ADSORPTION OF ETHYLENE ANDBUTENE1-Butene immediately isomerized to 2-butenes on reduced NS-1 aerogel, but unlessH2 was added the catalyst was quicklypoisoned by tightly held hydrocarbon pre-sumably formed by dehydrogenation of thebutene. The activity of the catalyst wasrestored by evacuation at 400°C or by admission of H2 at room temperature.Hydrogenation of butene occurred extremely rapidly at room temperature when H2was present. Nothing identifiable as an active intermediate was seen in spectrarecorded during these reactions.When ethylene was added in small doses to reduced NS-1 in the absence of H2,weak C-H stretching bands were sometimes noted.To permit study in greaterdetail, five aerogel plates were stacked in a quartz cell and, after reduction in H2at 400°C and a 15-min evacuation, were exposed to a small dose of ethylene (1.2 cm3s.t.p.). No C-H stretching bands were evident at this point. On subsequentadmission of 132 (without prior evacuation of the cell), strong saturated C--Hbands were produced. Brief evacuation removed these bands. Re-addition ofthe initial dose of ethylene then produced C-H stretching bands resembling thosepreviously assigned to 2-butenes adsorbed at low coverage on y-alumina.13 Addi-tion of H2 (18 torr) immediately gave C-H stretching bands assignable to saturJ . B. PER1 1 29ated hydrocarbon. These bands were almost entirely removed by evacuation €or5 min.Re-addition of the initial dose of ethylene again gave the ‘‘ 2-butene ”C--W bands, and re-addition of H2 again gave the saturated C-€i bands.Similar study was made with a single plate of NS-3 in cell C13(P.E. 12C spectro-meter, LiF prism). The gel was heated in H2 at 400°C, cooled in H2, evacuatedbriefly, and exposed to ethylene. Fig. 8 illustrates spectra obtained before and afteri 29200 I l : r ! I I t I I I l I I , !2800 2900 3000 3 100wavenumber (cm-1)FIG. 8.-Adsorption of ethylene on reduced NS-3. (a) After reduction at 400°C ; (b) < 15 rninafter addition of ethylene (1 cm3 s.t.p.) ; (c) 2 h later.2 9 2 0 2960 IO L L 2 - L L U U d ~ L2800 2900 3000 3100wavenumber (cm-1)FIG. 9.-Hydrogenation of “ ethylene ” adsorbed on reduced NS-3.(a) After original reductionat 400°C ; (b) after 5 min evacuation following (c) in fig. 8 ; (c) < 15 rnin after H2 addition (23torr) ; (d) after 5 min evacuation.the ethylene had been added. Spectrum (b), recorded within 15 rnin after the ethylenehad been added (1 cm3 s.t.p.) shows C-H stretching bands near 3003, 3083 and3105 cm-1 in addition to bands below 3000 cm-1. After 2 h the bands above 3000cm-1 appeared quite different. As shown by spectrum (c), the principal bandswere then near 2920, 2970 and 3020 cm-1. The spectrum closely resembles spectr130 ADSORPTION ON SILICA-SUPPORTED NICKELpreviously obtained for butenes on y-alumina.13 Concurrent weight readingsestablished that adsorption of ethylene had increased by about one-third (from1.2 to 1-6 mg) between spectrum (b) and spectrum (c).Evacuation of the cell for 5 min removed about 25 % of the adsorbed hydro-carbon.As shown by spectrum (b) in fig. 9, the remaining hydrocarbon stillresembled 2-butenes. Spectrum (c), recorded shortly after M2 (23 torr) had beenC2 3 2900 3000 3100wavenumber (cm-1)FIG. 10.-Adsorption of ethylene on unreduced WS-4. (a) Before addition of eLylene; (1;) im-mediately after additicn of ethylene (3-9 torr); (c) 17 h later (3.4 torr); (d) 4 days after largeraddition (8.3 torr) ; (e) after a 15-min ‘‘ freeze-out ” with liquid N2 ; (f) gas phase (sample out)after (b) (3-9 torr).added, shows that the adsorbed olefin was rapidly hydrogenated to form a saturatedhydrocarbon.Weight readings recorded at the same time indicated desorption ofabout two-thirds of the adsorbed hydrocarbon. The C-H stretching bands re-sembled those of adsorbed n-butane. As shown by spectrum (d), these bandswere markedly decreased by evacuation for 5 min. Weight readings showeddesorption of about half of the remaining adsorbate.Spectra obtained when ethylene was admitted to “ bare ” nickel (evacuatedat 400°C after reduction) were similar to those shown for “ hydrogenated ” nickel,but the C-H stretching bands below 3000cm-1 were somewhat weaker with thebare nickel. Spectra obtained by addition of 1-butene to reduced NS-3 appearedidentical to those obtained 1 h or nnore after ethylene had been added, and sub-sequent changes on evacuation and on addition of H2 were also the same. Fhysicaladsorption of n-butane gave bands very similar to those obtained from hydrogen-ation of adsorbed ethylene or butene.Ethylene adsorption was then studied on fresh unreduced NS-4 aerogel (pre-viously calcined in 0 2 and dried by evacuation at 500°C) with the Beckman IR-J.B. PER1 131spectrometer. Small additions of ethylene were initially adsorbed as ethylene,producing weak bands at 3005, 3083 and 3105cm-1 (similar bands occur in theRaman and infra-red spectra of liquid ethylene 20). With more adsorbed ethyleneand longer contact time, strong bands appeared in the 2850-3050 cm-1 range andcontinued to increase slowly for several days. The ethylene pressure in the cellsteadily decreased.Typical spectra obtained during this process are shown infig. 10. These bands, and others at 1630, 1455 and 1355 cm-1, indicated extensiveformation of 1 - and 2-butenes and apparently some hydrocarbon of higher molecularweight on standing. After 4 days' contact the gas phase contained, in addition toethylene, 1.65 % 1-butene, 3.4 % cis-=l-butene, 7.5 % trans-2-butene, and 0-63 %ethane (as shown by chromatographic analysis).Ethylene adsorption on unreduced nickel/Cabosil (dried by evacuation at 500'C)was then studied briefly. Spectra recorded 1 h after admission of ethylene (30 torrin cell) showed moderately strong C-H stretching bands resembling those of ad-sorbed butenes. These bands had increased somewhat after an additional 16 hand showed evidence for formation of some hydrocarbon of higher molecularweight.DISCUSSHONThe acid sites created by nickel in the surface of silica were thought by Uchidaand lmai 2 to be protonic.The spectra of strongly adsorbed NH3 and unsaturatedhydrocarbons indicate, however, that incompletely co-ordinated nickel ions (Lewisacids) rather than protons are responsible for the acidity of " dry " nickel/silica.On a hydrated surface the nickel ions are undoubtedly shielded through co-ordination with I-120 and hydroxyl groups. When HzO is removed by drying athigh temperatures, nickel ions are apparently exposed on the surface. Rather thanfree protons being retained at sites where nickel replaces silicon, silanol groupsare apparently left near nickel ions possibly as shown in fig.11 (a). A low stretchingfrequency (e.g., 3G20 ; normal 3750 cm-1) for these silanol groups could resultfrom the inductive effect of the nickel ion on the oxygen atom of the group. Suchgroups should be more acidic than normal silanol groups. The 3620cm-1 bandcould conceivably be caused by Ni-OH or by bicarbonate groups rather than bysuch Si-OH groups, but the high stability of the groups argues against thesealternatives.As demonstrated by chemisorption of HCl, unusually reactive oxide ions alsoexist on the nickel/silica surface. They apparently directly adjoin nickel ions,possibly as also indicated in fig. ll(a). Not all N 2 + ions need be held as shown,however. When HC1 is cliemisorbed, new hydroxyl groups can be created andchloride ions can be attached to a nickel ion, as shown in fig.ll(b). The presenceof new hydroxyl groups and chloride close to each other and to the original hydroxylgroup can explain the broad H-bonded hydroxyl stretching band and the formationof some H20. The Ni2+ ion occupies an almost complete octahedral co-ordinationsite. (The chemistry of the site is assumed to make additional chemisorption ofHC1 dificult.)The merits and defects of past explanations of CO bands near 2200 em-1 havebeen thoroughly discussed elsewhere.6-8 As found here, the sites on which ad-sorbed CO gives bands near 2200 cm-1 are also the sites which hold adsorbed NH3most strongly. These sites must contain incompletely co-ordinated nickel ions,because NH3, as NH3, could hardly be held so strongly by an oxide ion alone.Possibly the bands near 2200 em-1 arise from CO held as shown in fig.1 l(c), com-pleting the tetrahedral co-ordination of a Ni2+ ion, through a weak u bond. Th132 ADSORPTION ON SILICA-SUPPORTED NICKELexistence of two bands (2198 and 2210cm-1) near 2200cm-1 probably reflectsattachment of CO to Ni2+ ions that differ slightly in local environment.Bands near 2200cm-1 for CO on other oxides probably also reflect weak a-bonding to exposed metal ions. Little 8 points out that borane carbonyl (H3B : CO)exhibits a C-0 band at 2164 cm-1 and suggests that weakly a-bonded complexeson sites more electron-deficient than BH3 could give higher frequencies. Althoughpolar environments generally decrease the infra-red frequencies of polar groups,HSiSiFIG.1 1.-Co-ordination of Ni2+ on a silica surface. (a) On " dry " silica ; (b) incomplete octa-hedral co-ordination produced by chemisorption of HCl ; (c) tetrahedral co-ordination completedby CO ; (d) octahedral co-ordination completed by CO.factors which are unimportant in usual solvent effects may operate on a polar orionic surface. Weak attachment by one end of a CO molecule to an atom held ina solid lattice introduces a new bond to be compressed when the C--0 bond isstretched. This could somewhat increase the apparent force constant and frequencyof the CO bond. Inability of an adsorbed CO molecule to re-orient to minimizeits energy of interaction with an intense local field might slightly decrease the C-0internuclear distance.This, too, could shift its frequency to a higher value.17The bands near 2140cm-1 are also of major interest. Although, at -195"C,physically adsorbed CO gives bands at 2130 and 2150 cm-1,18 such weak adsorptioncannot explain the present results. Because the bands near 2140 cm-1 are enhancedby mild reduction of the original aerogel, they could conceivably represent COheld on partially reduced nickel ions (Ni+) or possibly even on single nickel atoms.The interconversion of sites giving a CO band near 2200 cm-1 and those giving aband near 2135 cm-1 produced by chemisorption of HC1 does not indicate reduc-tion of Ni2f in the usual sense, however. More probably, modification of the co-ordination of Niz+ ions, resulting from HCl chemisorption, changes the way in whichCO is subsequently held, possibly as shown by fig.ll(d). The adsorbed CO, inthis case, completes the octahedral co-ordination of a Ni2+ ion. EnhancemenJ. B. PER1 133of the 2140 cm-1 band by prior reduction in H2 might result from some Ni2+ siteswhich held hydride or hydroxyl ions in much the same way as chloride is held infig. ll(b). Conceivably electron pairs could also substitute for anions in fillingco-ordination positions.d-Electrons should become more available on partial reduction of Ni2+ or prob-ably from filling its co-ordination shell with chloride and hydroxyl ions. Backbonding of d-electrons from nickel would be expected to lower the C-0 stretchingfrequency and increase the strength of the Ni-C bond.We might thus expectthat (as observed) CO giving bands near 2140 cm-1 would be more tightly heldthan CO giving bands near 2200 cm-1. The similarity to the frequency of gaseousor physically adsorbed CO appears to be coincidental.Bands in the vicinity of 2050-80 cm-1 and at lower frequencies represent surfacecarbonyl structures formed by interaction of CO with reduced nickel atoms, smallclusters of atoms, or small nickel crystallites.9s 19 As indicated by the variety ofbands and the changes observed with changing CO pressure, both interactions ofadsorbed CO molecules with one another and variations in environment of thenickel atoms to which they are bonded (including effects of various exposed crystalfaces) must affect the frequencies.Evidently, in addition to at least three distincttypes of ionic sites (bands near 2210, 2198 and 2140cm-1) at least four types ofmetallic sites exist. Detailed characterization of these sites is clearly not possibleat present, and no distinction can be drawn between linear and bridged forms ofadsorbed CO with regard to bands near 1960 cm-1. Stable volatile carbonyls mustalso be formed to some extent.Although much effort was directed toward obtaining spectra of hydrocarbonsadsorbed on metallic nickel, this goal was apparently not realized. Attention wasinstead attracted to complications arising from incorporation of nickel ions in thesilica surface. Retention of unsaturated hydrocarbons by surface sites that con-tain both metal ions and oxide ions is hardly surprising.13s 14 Formation ofbutenes from ethylene over incompletely reduced nickel/silica might have beenexpected, because the reaction is known to occur over unreduced nickel/silica.2The spectra obtained here and those of Eischens and Pliskin 9 for ethylene onnickel/silica are so similar that they probably show the same adsorbed species.These are principally ethylene and butenes before H2 addition, and mostly butaneafterwards.Such re-interpretation is supported by the olefinic C-H stretchingbands reported (but de-emphasized) by Eischens and Pliskin, and by the bands atlonger wavelengths which they ascribed to diadsorbed ethylene (CH2-CM2) andto ethyl groups (CH2-CH3). Spectra of 2-butenes and butane 21 show bands similarto the bands at longer wavelengths.The samples of nickel/Cabosil studied by Eischens and Pliskin probably con-tained some unreduced nickel, because their reduction conditions were rather mild.Their failure to observe a band near 2200 cm-1 for CO adsorbed on typical reducednickel/Cabosil samples could reflect poisoning of ionic nickel sites (e.g., by ad-sorbed HzO), or simply that few such sites were present.Sites responsible for bandsnear 2200 cm-1 are not the only ionic sites on nickel/silica, and few, if any, of theformer are needed to catalyze dimerization of ethylene. Sample NS-1 afterreduction held NH3 strongly and formed butenes from ethylene, but it showed aband near 2200 cm-1 only at fairly high CO pressures.The results of Eischens and Pliskin were not duplicated in all respects. TheC--H stretching band intensities did not usually increase greatly when H2 was1 II134 ADSORPTION ON SILICA-SUPPORTED NICKELadmitted to bare nickel/silica holding adsorbed ethylene after evacuation.Aftersaturated C-H stretching bands had been removed by evacuation they were notsubsequently restored by re-addition of hydrogen. If, in the earlier work, tracesof olefin had been left after brief evacuation the discrepancies can mostly be explained.Spectral differences could also result from differences in pressure, contact time, andthe extent of formation of hexenes or products of higher molecular weight. Thespectra shown by Eischens and Pliskin evidently represent coverage of a larger frac-tion of the total surface with adsorbed hydrocarbon than is represented, for example,in fig.8 and 9.Infra-red studies of nickel/silica to date thus appear to shed no direct light onthe nature of ethylene adsorbed on reduced nickel. The results of such studiesare, however, evidently consistent with the formation of highly carbonaceousresidues on clean nickel.12CONCLUSIONSModification of adsorptive and catalytic properties of reduced rnetal/silicacatalysts by unreduced metal ions held in the silica surface has received too littleattention in previous infra-red studies. For nickel/silica such modification canbe quite significant. Unambiguous interpretation of spectra of surface groups andadsorbed molecules has rarely been possible to date.No direct evidence yetestablishes that any of the adsorbed species observed spectroscopically plays animportant role as an active intermediate in a catalytic reaction. No spectra havebeen obtained for CQ or hydrocarbons on nickel/silica which cannot be attributedto reasonably stable surface species or conventional molecules, often held by icnicsites on the support. Considerable caution seems warranted in interpretation ofspectra of adsorbed molecules on metal/silica catalysts. The evident complexityof oxide surfaces suggests that much additional work will be needed before adequateunderstanding is achieved.The author thanks Prof. R. L. Burwell, Jr., for helpful discussions and J. Mekichfor assistance in the experimental work.1 Schuit and Van Reijen, Adv. Catalysis (Academic Press, N.Y.), 1958, 10, 243.2 Uchida and Imai, Bull. Chem. SOC. Japan, 1962,, 35,989 ; 1962, 35, 995.3 Barth and Pinchok, J. Physic. Chem., 1964, 68, 655.4 Eischens and Pliskin, Adu. CataZysis (Academic Press, N.Y.), 1957, 9, 662.5 O’Neill and Yates (D. J. C.), Spectrochim. Acta, 1961, 17, 953.6 Amberg and Seanor, Proc. 3rd Int. Congr. Catalysis (North-Holland Publ. Co., Amsterdam,7 Gardner and Petrucchi, J, Physic. Chem., 1963, 67, 1376.9 Eischens and Pliskin, Adv. Catalysis (Academic Press, N.Y.), 1958, 10, 1 .10 Little, J. Physic. Chenz., 1959, 63, 1616.11 Little, Sheppard and Yates @. J. C.), Proc. Roy. SOC. A, 1960, 259, 242.12 Little, Klauser and Amberg, Can. J. Chem., 1961, 39, 42.13 Peri, Actes Deuxieme Congr. Pnt. Catalyse (Editions Technip, Paris, 1961), p. 1333.14 Peri, Proc. 3rd Int. Congr. Catalysis (North-Holland Publ. Co., Amsterdam, 1965), p. 1100.15 Peri, paper to be submitted to J. Physic. Chern.16 O’Neill and Yates (D. J. C.), J. Fhysic. Chern., 1961, 65, 901.17 Seanor and Amberg, J. Chem. Physics, 1965,42, 2967.18 Smith and Quets, J . Catalyris, 1965, 4, 163.19 Yates (J. T.), Jr. and Garland, J. Physic. Chem., 1961, 65, 617.20 Herzberg, Molecular Spectra and MoZecular Structure, vol. II (Van Nostrand Co., New York,21 Arner. Fetroleum Imt. Res. Proj. 44 Spectra, serial numbers 438, 907, 91 1 ,22 McKee, J. Arner. Chem. SOC., 1962, 84, 1 109.1965), p. 450.8 Little, book in preparation.1945), p. 326

ISSN:0366-9033    

DOI:10.1039/DF9664100121

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Conversion and Equilibration Rates of Hydrogen on NickelBY D. D. ELEY AND P. R. NORTONChemistry Dept., No ttingliam UniversityReceived 27th January, 1966The pH2 conversion rate has been compared with that for OD% and H2+D2+2HD on nickelwires rigorously cleaned in vaciio of 10-10 torr. The rates are significantly higher than on evapor-ated films, prepared under similar conditions but still possibly contaminated. The results at 1 torrfall into three temperature ranges (a) 330-400"K, a fractional order (ca. 0.6) reaction, activationenergy 5-7.5 kcal mole-1. There is evidence for a zero order HzfD2 reaction setting in as thepressure is lowered to 0.3 torr ; (6) 200-300"K, a fractional order reaction, 0.05-0.40, with a loweractivation energy, 1-5-27 kcal mole-1. In (a) and (b), pH2, oD2 and H2fD2 rates are very similarto each other. On decreasing the temperature to (c) 77-200"K, the activation energies for thep22 and oD2 rates decrease to zero, so that at 77°K these two reactions go lo3 times faster thanH2+D2.The mechanisms advanced are for (a) Bonhoeffer-Farkas, (b) Eley-Rideal, and (c) a para-magnetic conversion induced probably by unpaired d electrons in the Ni and proceeding througha Harrison-MacDowell mechanism.1.n this paper we re-examine the pM2 conversion and &+D2 reaction over awide temperature range on nickel, to resolve the diversity of previous results.1 Inaddition, the oD2 conversion is examined, so as to decide whether the conversionmechanism at 77°K is chemical or paramagnetic.We refer to postulated mechanismsby Roman numerals.2pH2 (or oD2) + P+P +OH:! (or pD2). (1)The spin reversal is catalyzed by the paramagnetic site P.39 4 The rate dependson the concentration of molecularly adsorbed pH2, and therefore occurs fastestat low temperatures.5The Bonhoeffer-Farkas 6 chemical mechanism ; the M2 + D2 reaction will occur ifthe H and D atoms diffuse over the surface, or if groups of 4 adjacent sites 7 areavailable.pHz.+MH. M-tMH. MH2-tM. MH+oPT2. (IIIa)The Rideal mechanism 8 which involves empty sites adjacent to chemisorbed(IIIb)(IIIC)pHz+M. M->MI<. MH+M. M+oH~. (11)H atoms.pH2 + MH-tMH3 +MN 4- OH^.MH . MIS. MH+M€I. MI32. M-tMHzMH. M.The Eley mechanism 8 9 9 which postulates a triatornic complex.The Boreskov variant 1" of (111~1, b) which postulates migration of an H atomover adjacent chemisorbed H atoms, until an H2 molecule is formed on a site oflow desorption energy, when it evaporates.MH2+MDr+MHD+MHD.(W']The Schwab Inechanism,ll molecular mixing apparently based on 2 metal sites.135An earlier paper 12 gives a treatment of the 4-site Polanyi model.136 CONVERSION OF HYDROGEN ON NICKELDiagnostic tests of mechanisms are mainly based on a comparison of rates. Thus,any chemical mechanism, (11)-(IV), should show decreasing rates pH2 > H2 + D2 > oD2due to the effect of zero-point energy on the activation energy.13 But if a para-magnetic mechanism is dominant on a given surface at 77°K the rate order willbe pH2 > oD2 9 H2 + Dz, since the last reaction is not catalyzed by the paramagneticmechanism (I).In addition, pressure dependencies give reaction orders which inturn give information about the surface coverage of adsorbed layers. The deter-mination of activation energy may also be helpful, since often a paramagneticmechanism in a van der Waals layer shows a negative apparent activation energy.5EXPERIMENTALThe reaction system was essentially that of Eley and Rideal14,2 modified to achieveultra-high vacuum conditions and to permit analysis of €€D by an M.S. 10 mass spectro-meter. A cylindrical reaction vessel, two liquid-nitrogen traps and a main mercury cut-offin series were connected via another mercury cut-off to a gas preparation system, and amicro-Pirani gauge for analysis of the ortho-para species.15 The system was pumped bya rapid two-stage mercury pump, and it was bakeable up to the cold trap next to this pump.Pressures were measured on the punip side of the main cut-off by an ion-gauge, protectedby a cold trap. After a 24-h bake of the reaction space at 400"C, vacua of 10-9-10-10 toriwere obtainable.A length of 8 mm tubing containing a Nietrosil leak connected the reactionspace to the head of an A.E.I. M.S. 10 mass spectrometer, the whole being bakeable at300°C to give residual pressures of ca. 10-10 torr. This system allowed analysis of hydrogensamples of up to 10 torr pressure. Independent experiments established that no isotopefractionation occurred at the leak inlet.The hydrogen gases were purified by a palladium-silver thimble, and orthodeuteriumand parahydrogen prepared by contacting the gas with a charcoal catalyst at 20°K.Thehydrogen gases were stored in bulbs with greased taps, and cleaned regularly by adsorptionon charcoal at 77°K. Checks were made with the M.S. 10 periodically and the impurity levelfound to be regularly less than 1 part in 108. Gases swept in from the preparation linethrough the two cold traps into the reaction vessel showed no trace of H20, Hg, 0 2 or N2.The pressure in the vacuum system remained constant at 10-9 torr when opened to the gasline pumped to 5 x 10-8 torr.The cylindrical reaction vessel contained a polycrystalline Ni wire (Johnson MattheyGrade I) held axially between heavy tungsten leads. An elaborate cleaning treatment wasadopted which gave a high and reproducible catalytic activity for the Ni filament.First,the Ni wire was heated electrically to 1000-1 100°C in a 10-7 torr vacuum, until a thick nickelfilm was formed on the glass walls. At this point the wire was cooled, the tube cut offand the film removed with dil. HCl. This procedure was repeated until no further gasevolution was detectable on heating the wire to 11OO"C, when the system was baked at400°C to 10-9 torr. The wire was then oxidized at 800°C in 10-1 ton oxygen, reducedat 900°C in 1 torr hydrogen, and then outgassed at 900°C. The oxidation-reduction treat-ment was repeated until the wire activity became reproducible. Prior to a catalytic test,the wire was annealed for 1 h at 800"C, and the system rebaked to remove oxygen condensedin the liquid nitrogen traps.Before each run the wire was flashed at 890-910°C in a vacuumof 10-9 torr, this temperature giving the best compromise between efficient cleaning andexcessive evaporation.Catalytic runs were carried out by two methods : (i) heating the whole reaction vesselemploying the temperature bath used in the film work ; a 10 cm length of 32 s.w.g. Niwire was used ; (ii) electrical heating of 20 cm of 42 s.w.g. wire with the walls at 77°K or273"K, taking care at 77°K to choose the pressure range so as to avoid the Busch effect.16.17Methods (i) and (ii) gave identical results showing that the whole catalytic activity was dueto the wire surface. When catalytic films were examined, the same catalyst wire, after theabove pretreatment, was used as evaporation source, with the walls at O"C, and thereafterthe nickel film was sintered for 2 h at 370°KD.D. ELEY AND P. R. NORTON 137RESULTSAs in earlier work the time course of the three reactions is described by thefirst-order equation :1 x o - xt Xr-xeqk , = - In - eq, min-I,where XO, xt and Xeq are the fractions of parahydrogen (orthodeuterium or hydrogendeuteride, mass 3) present at times zero, t, and at equilibrium.From the measured pressure and effective temperature of the reaction volume,12, the number of molecules in the reaction space is calculated, and hence the absolutereaction rate k,. A is the catalyst area taken as the geometric area of the well-annealed wire.Apparent activation energies E are derived fromk, = nk,/GOA molecules cm-, sec-l.k, = B, exp-(EIkT), molecules cm-2 sec-l.TEMPERATURE DEPENDENCE OF THE RATEThe spin isomerizations studied were limited by technical considerations to therange 77-32OoK, while the response time of the M.S.10 allowed H2+D2 equilibra-tion to be followed from 77 to 1000°K. In the present paper we restrict considera-tions to an upper temperature of 400"K, since above this temperature catalystpoisoning by hydrogen sorption becomes a major feature of the results.Fig. 1 shows that the orthodeuterium and parahydrogen conversion rates areclosely similar on a nickel wire over 77-300°K. The H2+D2 rate, however, is muchless at the lower temperatures, becoming nearly equal to the first two rates onlyabove 200°K.The H2+D2 rates over 200-350°K were derived by extrapolationfrom the data of fig. 2, on a Ni wire using the temperature bath method. Fig. 2shows the results for the above system for the temperatures 200-300°K on a muchexpanded scale, in which range the rates decrease in the order pH2 > H, + D2 > oD2.TABLE 1kinetic data at p = 0.8-1.3 torrparameter T range, "K P H ~ OD2E 77- 1 50 0 0E 200-300 1.54 2-68E 330-400 2 5loglo k m 77 16.06 15.909 9 273 17-32 16.859 ) 373loglo Bm 77- 1 50 16.15 16-009 200-300 18.60 19.00Y9 330-400 9 18.60-- --H2+D22.42.367.5713.017-1218.1019.0419.0422.3If the H, +D2 reaction is studied to higher temperatures, there is an increase inactivation energy at about 290°K.This is shown in fig. 3, where the close agree-ment between results for both heating methods from 200 to 290°K proves con-clusively the absence of spurious effects arising from wall reactions (evaporatedfilms) or the Busch effect.16.17 The fact that the rates at 0-2, 1-2 and 3.76 torrgive E = 7.5 kcal (300-400°K) shows that the increased activation energy ove138 CONVERSION OF HYDROGEN ON NICKELthis higher temperature range did not arise from the Busch effect. By using measuredpressure dependencies all rate data were corrected to p = 0-8-1.3 torr, and theresults collectzd in table 1. E is in kcal mole-1, km and Bm in molecules cm-2 sec-1.Bm = Bzp& includes the pressure dependency (table 2).103/TFIG. 1 .-Arrhenius plots for II), ortho, para H2 ; 0, ortho, para D2 and X , H2+D2, over 77-300°K.I I I3 4 51 0 3 1 ~FIG.2.-Arrhenius plots for a, ortho, para HZ ; 0, H2+D2 ; 0, ortho, para D2 over 200-300"Kon a 32 swg Ni wire by cooling bath methodD. D. ELEY AND P. R. NORTON 139PRESSURE DEPENDENCE OF THE RATEPressure dependencies are expressed by the Freundlich equation,k, = k:pL,,the value of n being derived from the slope of log km against logp plots as shownin fig. 4, for two temperatures, 77 and 273°K. For H2+D2 the mass spectrometric103/~FIG. 3.-Hz+D2, equilibration rates over 200400°K on an electrically heated 42 s.w.g. wire,0, 0.2 torr ; A , 1.2 torr ; 0, 3-76 torr ; 9, cooling bath 32 s.w.g. wire, 1.2 torr.method also allowed results to be obtained at 350"K, as shown in table 2.The ratedata for H2 + D2 at 0.2 and 1 -2 torr in fig. 3 give n = 0.6 (300-400"K), which agreesroughly with the average values for the data in table 2.TABLE 2.-PRESSURE DEPENDENCE OF THE RATET"K PH2 ODZ H2+D277 0.14 (10-1-10 torr) 0.14 (10-1-10 ton) 0.30 (10-3-10 torn)273 0.22 ,, 0.05 ,, 0.40 $9350 - - 0.0 (0*03-0*3 torn)350 - - 0.35 (0-3-1-0 torr)350 - - 0.76 (1.0-10 torn)CONVERSION, EQUILIBRATION AND EXCHANGE ON NICKEL FILMSIn fig. 6 the dotted curves through the points represented as crosses and opencircles show absolute rate data, based on the geometric dred, for a nickel film pre-pared as described earlier. If we assume the hydrogen uptake at 273°K and 1 torrcorresponds to a monolayer, we estimate a roughness factor of 5-10 for this film,which would displace the above curves downwards by this factor.Clearly theactivation energies and frequency factors for the film are lower than those for thenickel wire (curves through open squares and closed points). It has been foundthat k, and E values for nickel films depend on the evaporation and annealing con-ditions of the film. Evaporation of nickel from a tungsten wire appears to givehigher apparent activation energies E for the reactions, possibly due to contaminationof the nickel films by carbon monoxide from the tungsten filament140- L t l l l l l l l ' f l ' l ' 'CONVERSION OF HYDROGEN ON NICKEL16.1ii0-0- O--<17.4116.516.1M 0416.5ooPo/- 15.6-12.0 - I3I II 0 2l o g p , torrFIG.5.-Pressure dependencies of H2+D2 equilibration, El, 77°K (log k,, scale 12-13.2), 0, 273°K(15-6-17.4) ; A, 350°K (17.4-18-4)D. D. ELEY AND P. R. NORTON 141A nickel flm prepared as above in a 10-9 torr vacuum was contacted with 1 torrD2(H2) at 273°K for 10 min, then pumped, followed by admission of ca. 1 torrH2 (D2). The rate of formation of HD was then followed according to the methodof Eley,l* as developed by several other authors.19-21 If 100 % of the surface ex-changes D atoms at 273"K, only 3 % of the nickel surface exchanges rapidly at77"K, as shown in table 3. In accordance with the earlier workers, it is concludedthat the nickel film has a heterogeneous surface.1 0 3 1 ~FIG. 6.-A comparison of Arrhenius plots for wire ; 0, ortho, para H2 ; 0, H2+D2 ; and filmsevaporated from this wire under identical conditions : x , ortho, para H2 ; 0, H2fD2.At 273°K it is not clear if we are dealing with a true exchange reaction.Onadmitting 1 torr H2 to a Ni-D film at 273"K, 80 % of the D was displaced as D2into the gas phase over the first few seconds, and thereafter reacted to give HD byAt 77"K, H2 did not displace D2 from an Ni-D film, within the limit of sensitivityof the mass spectrometer, which is rather low at p < 10-4 torr. After exchange wasH2 + D2+2 HD.TABLE 3.-EXCHANGE AND EQUILIBRATION ON A NICKEL FILM AT 273°Kexpt. 273°K 77'K comments% exchange of H 100 3 D2 film first displacedkm, arb. units 100 3.5 n = 0.5km, geom.area 1016 3 . 5 ~ 1014 exch. and equil. bothmolecules cm-2 sec-1E, H2+ Dz, kcal mole-1 0.85 0.85by H2 at 273°Kpoisoned by N2 at 77°142 CONVERSION OF HYDROGEN ON NICKELcomplete at 77°K the introduction of further H2 (D2) over the D (H) adsorbed filmresulted in a rapid equilibration reaction which, however, could be poisoned byadsorption of N2 gas. Since N2 does not chemisorb on Ni at 77"K,22 we concludethat the poisoning is due to physically adsorbed N2 blocking sites for the exchangereaction, probably sites which will adsorb molecular H2. Within the limits oferror, the observed rate constant km for equilibration was independent of gas-phasecomposition from 50 M : 50 D to 90 D : 10 H.DISCUSSIONNATURE OF CHEMISORBED HYDROGENThe initially chemisorbed hydrogen bonds to Ni as atoms with an Ni-H bondstrength of 60 kcal mole-1, and with a Nidf-Hd- dipole of about 0.13 D.23 As aresult the hydrogen film raises the work function of the metal and also its resi~tivity.24~25This negative hydrogen may sit in the interstices between the metal atoms.26927The heat of adsorption decreases with surface coverage of H atoms,2* until,approaching 8H = 1, reversible adsorption sets in and there is a decrease in workfunction (positive variation of the surface potential) and the resistivity of the metal.This reversibly adsorbed species is taken to have a positive polarity, and field emissionstudies show that this species is preferentially found on the loosely packed planesof the transition metals, such as the (110) plane of f.c.c.Ni.27 The first type ofchemisorbed H has been called type A, the second type C (for a review, see ref.(1)).The type C adsorbed hydrogen may be an intermediate in the activated process offorming type A chemisorbed H. Type C hydrogen may be atomic, or molecular.W Tvr>e A Distance of hydrogen speciesfrom surface --*FIG. 7.-Potential energy diagrams for type A (2M-€33 and C (M . . . H2) chemisorption, afterDowden. Ea is the activation energy for adsorption ; q is the heat of chemisorption.Dowden29 has expressed the situation in terms of the potentiaf energy curves offig. 7. While the type A hydrogen is commonly believed to be held on the Nisurface dsp hybrids, Dowden has suggested that the type C (molecular) hydrogenis held on the empty atomic d orbitals.The clearest distinction between the twotypes of chemisorbed hydrogen was found by Hickmott 30 in his desorption kineticstudy of W-H. Here the strongly chemisorbed p hydrogen is found to be atomic,and the weakly bound a hydrogen to be molecular. The a hydrogen is detectableat 10-2 tom and 194°K on tungsten. Because of the strong decrease of heat oD. D. ELEY AND P. R. NORTON 143adsorption with coverage, p hydrogen may evaporate from the atomic layer evenat 77°K. The Mignolet and Sachtler picture provides a stereochemical reconcili-ation of these views. They show that the lattice planes which adsorb type C (a)hydrogen all show two interpenetrating sub-lattices of surface sites, A and €3. TheA sites which would normally be occupied by the next layer of metal atoms duringcrystal growth are filled by type A chemisorbed H.When these A sites are half-filled, atoms may start to fill the B sites, and a neighbouring A-B pair of atomsforms a kind of stretched hydrogen molecule. It has recently been found thatchanges in thermoelectric power may sensitively reflect changes in the Fermi levelof the metal which follow the adsorption of hydrogen in negative and positivespecies.31It is clear that the character of the reversible weakly bound positive hydrogenwhether molecular 29 or atomic,32 is still uncertain for any particular metal exceptW.30 If molecular, presumably the chemisorption has a charge-transfer char-acter.33 In addition to the above " homogeneous " surface, Sweett and Rideal28found evidence for 1.5 % sites of higher adsorption potential on evaporated nickelfilms.These active sites may be the trap sites observed in hydrogen adsorptionon field emission points of nickel.34CONVERSION A N D EQUILIBRATION, 330-400°KThe results obtained, viz., E(pH2)>5 kcal mole-1, E(H2fDz) = 7.57, and afractional order for this latter around 0.6, agree with the earlier studies in thistemperature range of Fajans 13 (nickel tube), Schwab 11 (nickel foil) and Kiperman 35(nickel powder). The pH2 conversion was too fast to follow accurately with themethod used and its E value may be in the range 5-7 kcal mole-1. We have nottherefore been able to establish the isotope effect on the activation energy whichFajans found to be 1.4 kcal mole-1 for E(H2 + D2) - E(pH2).The frequencyfactor loglo 12,==22 is somewhat larger than that of about 21 found for earlier wires,36and is much larger than that of 19 found for transition metal films in a temperaturerange near 293"K.Z According to Sweett and Rideal at 0.3 torr and 400°K a nickelfilm has a coverage of H atoms of about 8-05. I t is therefore most likely thatconversion and exchange are both going through the Bonhoeffer-Farkas mechanism11, when 8 = 0.5 would correspond37 to a pressure exponent (assuming Langmuiradsorption) of n = 0.5, similar to that found. On this view the true activationenergy is given by 379 38E~ = ~ + q n = E+q(i-e)where q is the heat of adsorption, assumed constant.The assumed constancy of4 with 0 makes these considerations crude, but if we approximate q as the 8 = 0.5value of Sweett and Rideal, 17.4 kcal mole-1E, = 7.6 + 17*4(0.6) = 18 kcal mole - '.This agreement between Et and q would mean that the true activation energy forchemisorption, Ea in fig. 7, is zero. The finding that km (Hz+D2) is independentof the D content of the gas, disagrees with the results of Schwab and Killrnann,lzwhich they used to provide support for mechanism (IV).It is always possible that the rate of mechanism (In is limited by the rate ofsurface diffusion of H(D) atoms. Gomer34 has found for the rate of approach,2, of 2 H atoms on Ni= (~/2)8210'210-700*/4.6~~~-2 sec-144 CONVERSION OF HYDROGEN ON NICKELwhich for N = 1015, 8 = 0-5, T = 400°K gives 2 = 1022 cm-2 sec-1.This is muchfaster than the observed km (H2+D2) of 1018 molecules cm-2 sec-1, and it is there-fore unlikely that surface diffusion is limiting the rate.Referring to table 2 and fig. 5, the H2+D2 reaction at 350°K tends to zeroorder as the pressure is lowered, which we suppose is due to desorption of hydrogenfrom the homogeneous surface, leaving the reaction to occur by the Bonhoeffer-Farkas mechanism on the 1.5 % of strongly adsorbing sites mentioned earlier.Because of their high heat of adsorption these sites must be supposed to remainsaturated even at 0.03 torr. These strongly adsorbing sites must contain at least4 adjacent free valencies if the H2+D2 reaction is to occur rapidly, or there mustbe rapid H(D) atom migration between sites.However, until further results haveconfirmed the present scanty data, we should not put too much weight on the zero-order region.C 0 NVER S I ON, EQUILIBRATION AND EX CHANGE, 200-300°KThe close similarity of activation energies, 2-68 kcal mole-1, (oD~), 2.36 (H2 + D2),and 1.54 (pH2) and frequency factors, loglo Bm = 18.60 - 19-04 molecules cm-2 sec-1,suggest that all three reactions are going through the same mechanism. Earliervalues for pH2 on a Ni wire are E = 2.6 (loglo Bm = 19-93) 36 ; for pH2 on a Nifilm, E = 3.5 (loglo Bm = 18.90) and for H2+D2 on a Ni film, E = 2.44 (logloBm = 17*95).2 Ni supported on Si02 gave E = 3.26 kcal mole-1 (loglo Bm = 19.32)at 293"K,21 and a Ni film gave E = 1.4 kcal mole-1 at 194OK.39 The isotope effecton loglo km at 273°K is oD2 16.90 ; H2 + D2, 17.12 ; pH2, 17-32 which seems reason-able.The approximate constancy of Bm means that the main effect in this decreasein rate is the decrease in apparent activation energy AE, which from oD2 to pH2is 1.14 kcal mole-1. This could arise from the known difference in zero-pointenergy of 1.8 kcal mole-1 between gaseous H2 and D2, but possible contributionsmay also arise from Ni-H and Ni-D bonds and from the transition complexes,and it is not possible to be more specific at present.The sudden change in activation energy from about 7 to 2 kcal mole-1 at about330°K must denote a change in mechanism. As we lower the temperature weexpect the primary atomic layer of H to tend to become complete, i.e., B - t l .Thiswould give rise to a zero-order mechanism (11), as was found for a nickelJiZm at293"K.Z In fact, the orders found here for the wire at 273°K were 0.05 (oD~),0.40 (H2+D2), 0.22 (pH& the H2+D2 value being anomalously large. At thesame time as the temperature falls below 330"K, loglo Bm changed from 22.3 to19.04 for H2+D2, and presumably a similar change also occurred for the othertwo reactions. Many of these facts may be qualitatively reconciled if we suggestthat a mechanism (111) has set in, involving type C, second layer (en) adsorbedmolecular hydrogen. The sites must be relatively strongly adsorbing, since theobserved n values correspond to 8~ values of 0.95, 0.60 and 0-78. They cannot bethe sites of easy desorption postulated by Boreskov, so that (IIIc) is ruled out.Mechanism (IIIa) and (b) will result in an immediate exchange of atoms betweena D2 molecule and neighbouring chemisorbed H atoms.Since at this temperaturethere is still a good degree of mobility,34 i.e., diffusivity of atoms, in the primarylayer it should be possible to exchange 100 % of the chemisorbed H as was foundat 273°K.We have further considered the possible application of Boreskov's mechanism,in which the rate-determining step is H atom diffusion acoss the Ni-H, reactingto give a desorbed H2 molecule at a site of low desorption energy. This woulD. D. ELEY AND P. R. NORTON 145require an order of 0.5 (except in special circumstances occurring only in Pd-H 40)which is greater than the values found here.Secondly, while the activation energyfor this type of surface atomic diffusion is about 2 kcal mole-1, the low desorptionenergy sites occur only 1 in 108 with a very low desorption rate of loglo k = 10at 148"K, lower by 5 than the loglo k, values.EQUILIBRATION, EXCHANGE AND CONVERSION AT 77-200°KThe activation energy for the H2+D2 reaction remains constant at 2-4 kcalmole-1 from 300 down to 77"K, so presumably the mechanism remains (IIIa) or(b), occurring in a molecular hydrogen held on special strongly adsorbing sites.The frequency factor remains at its relatively low value of logloBrn = 19-04, inagreement with this view. The fall in amount of exchangeable hydrogen in theadsorbed atomic film to 3 % at 77°K can then be associated with the known lossof mobility of the atomic film.34 It is difficult to explain the fact that the observedorder has only fallen from 0.40 (273°K) to 0.30 (77°K).The main result of this paper is that the pH2 and oD2 loglo km against !/Tgraphsbend away from the line for I-T2+Dz, so that at 77°K the loglo km values are 16.06(pHZ), 15-90 (oD~) and 13.0 (Hz+D2).The two spin isomerization rates arewithin a factor 2 of each other, and 1000 times the equilibration rate, and this,together with their zero apparent activation energy, points to the operation of aparamagnetic mechanism (I).In principle, the surface of nickel, which has a bulk saturation moment of 0-6unpaired electrons per atom should be very paramagnetic. The band theorywould suppose that 60 % of the surface atoms each correspond to d9, Ni+, and con-tain one unpaired electron.However, the chemisorption of a monolayer ofhydrogen results in electron pairing and a loss of at least some surface paramag-netism.42 Furthermore, diamagnetic metals, such as Cu and Au, give a conversion,attributed to surface free valencies.36 In Kummer's 43 view, the surface para-magnetism of Ni, Cu and Au powders are similar and only correspond to oneunpaired electron per 50- 100 atoms.Sandler and Durigon 44 consider that chemisorption of hydrogen, but not oxygen,on Ag powder at 500°C produces paramagnetic centres. On the Pauling valence-bond view,45 if we assume a hybrid bond structure in the surface, as in the bulk,30 % of the surface nickel atoms possess the configuration &(empty d)2 (d2sp3)6in which there are two unpaired electrons in the atomic d shell, and presumablyone or more unpaired electrons in those hybrid d2sp3 orbitals which project from thesurface (the number of projecting orbitals depend on the lattice plane).To bedefinite we shall adopt this latter model, and assume the surface paramagnetismdue to unpaired electrons in d2sp3 bonding orbitals has been removed by chemisorbedhydrogen (H-, type A). The unpaired d orbitals will then hold the reversiblyadsorbed molecular hydrogen (Hi type C). Since the electron transfer in this casewill be far from complete we shall assume the unpaired d electrons are not coupledand that the nickel surface contains 30 % of atoms with two unpaired electrons withparallel spin (Hund's rule), giving p = 2 4 1 3 ) = 2.83 Bohr magnetons.Formetals, such as Cu, Au, Ag, which do not chemisorb hydrogen, we see no reasonto modify our view that unpaired electrons will be available in the hybrid bondingorbitals projecting from the surface, although a pairing up of surface bonds mayoccur between adjacent atoms. In the latter case, surface paramagnetism wouldonly remain at special sites, related to the '' residual conversion " of Kummer andSandler146 CONVERSION OF HYDROGEN O N NICKELIt is useful to consider quantitative calculations based on Wigner's theory,3and for this purpose we classify mechanisms according to the nature of the ad-sorption process.46 A careful assessment 44 has shown that mechanism (Ib) givesa ratio k (obs.)/k (calc.) of about 10 (originally given as 100, see appendix) for 5rare earth oxides, which is perhaps as good an agreement as is possible at present.A hydrogen molecule, mass nz, moment of inertia Z, proton magnetic moment p P ,in contact with a paramagnetic centre of moment pa, separation of centres rs cmfor a time ts sec, has a transition probability cptp = nTo1G(T), where Wol = 24p~$In2t~/h2mr~.G(T), a sum over states, is 0.210 at 77"K, 0.254 at 90"K, henceWol = 1-28 x 1016(pztz/r:), c.g.s.units with pa in Bohr magneton, rs in A.(la) Conversion occurs during simple collision with the surface, a fraction FaUsing W7igner's approximation for a homogeneous of which is paramagnetic.collision, ts = rs/3u, where the molecular velocity ZI = (3kT/r?z)+, thenand k m = ['/(2xr~kT~]cpl;, molecules cm-2 sec-1.At 77°K and 1 torr(Wol)coI, = Sp~,u~I.n2/9h2kTr~ = 1.152 x 10-9(p3Tr,6 c.g.s. units,it, = 8.84 x 10g(F,,u%)/rf molecules cm-2 sec-I.(Ib) Conversion occurs47 while the molecule is vibrating with frequency v =4.5 x 1011 sec-1 over a " sunken " (i.e., strongly adsorbing) site, where the numberof sites no crn-2 may be approximated by 101s Fa. The fractional coverage of sites8 is often taken as unity (zero order conversion), and each vibration is regarded asa collision with cp = (W&ollG(T) (cf. Sandler 48). Henceand at 77"K,k , = cpnoev,k , = 1.41 x lo1 ' ( F a 6 p ~ ) / r f molecules cm-2sec- '.(Ic) Molecular hydrogen, van der Waals (&w) or type C (On), mobile in a2-dimensional film over the whole surface with velocity c, during a mean time ofsojourn tm, traversing n, paramagnetic sites of radius r.If Znt is the average pathlength of the hydrogen molecule across the paramagnetic site 49With a rate of desorption p, thenIf q is the heat of desorption of a hydrogen molecule,where we take the total number of sites per cm2, ns = 1015, so thatSubstituting t: = P/c2 = 21.2/c2 in Wol, and Zm = 4r/x50,n, = tmF,c/l,.k m = VnrnB.tnz = ( l / v ) exp (q/kT) and = uS@v exp (-q/kT),n,P = F,crzS8/Z,.k, = 2 4 p ~ ~ ~ I 7 ~ ~ 2 r ~ l h ~ n i r ~ c ' G(T) = Facn,@x/4r.Putting c2 = 2kT/m for two-dimensional mobility,Values of k , are czlculated for Fa = 0-3, /la = 2-83 Bohr magnetons, I' = 2 A ,rs = 2A, T = 77°K.k, = 4.66 x 1018(rF,8p~)/r,8T9D.D. ELEY AND P. R. NORTON 147Since the reactions are nearly zero order, (Ia) is eliminated immediately, and weassume 8 = 1 for (Ib) and (Ic), and clearly this gives an upper value so far as 8 isconcerned. Observed and calculated values are compared belowpH2, 77"K, 1 torrkm (obs.) = 1-15 x 1016 molecules cm-2 sec-1,k , (calc. la) = 3.35 x 108 9 , !,km (calc. Ib) = 5.35~ 1013 ,, 9 ,km (calc. Ic) = 7.24 x lo15 ,, 9 9Mechanisms (Ib) and (Ic) may be distinguished, as for NdzQ3,47 where FaOmay be shown to be a small fraction of the total area, by a measurement of hydrogenadsorption. In the present instance the maximum calculated rate, as given by(Ic), is approximately equal to the observed rate, within the uncertainty of assigningparameters such as r, and tla, as was concluded for mechanism (Ib) with the rareearth oxides.The isotope effect predicted by Wigner theory is given byIt is easiest to refer to the km(pH2)/km(oD2) ratios found experimentally for theParamagnetic gases.These are 5.4 for 0 2 at 83"K, and 10 for 0 2 and NO at 193and 293OK.51~52 The ratio found for nickel is 1.44 at 77"K, and according to fig. 1even less at higher temperatures. The gaseous 0 2 and NO ratios will involve allthe above terms, except the 8H/oD ratio. While it is usual for deuterium to be morestrongly adsorbed than hydrogen, so that OH/BD would tend to reduce the isotoperatio ; in fact the differences from this source must be very small, since the commonreaction order of 0.14 ~ ~ l d point to OH = 8D = 0.86.The low isotope ratiosof 2 for Nd203 at 193°K and 1.5 for haemin at 83°K 48 show that the effect is notconfined to transition metals in bulk. Type C chemisorbed hydrogen is less likelyon these two solids.In summary, the 77°K spin isomerization reactions are clearly paramagneticin origin. An upper calculated value for the Wigner mechanism variant (Tc), onparamagnetic nickel atoms is nearly equal to the observed rate, and the agreementis possibly as close as may be expected.The authors' thanks are due to the S.R.C. for a studentship held by P. R. N.ERRATUMIn calculating Wigner transition probabilities for para-ortho Hz, in ref. (49) and incertain (not ref.(2)) papers by one of the present authors, a proton magnetic moment of1/1840 Bohr magnetons has been used, in place of the correct value of 2-79/1840. In themost detailed instance for Nd3+ ion in ref. (47), p. 214, and ref. (46), p. 162, the correctvalue of 4p at 90°K is (2.79)zx 8.66 x 10-14. The knZ (obs.)/knr (calc.) ratio in all cases islowered by (2.79,2 Le., roughly a factor of 10. So for Nd2O3, and the other rare earthoxides 5 1 Ic,, (obS.)/km (calc.) now approximates to 10. In ref. (52)' p. 20 and ref. (46),p. 164, for y-AlaO3, q= 1 . 9 2 ~ 10-14 should be 1.5 x 10-13, and km (obs.)/km ; (calc.)5-2 x 102. In ref. (46), copper phthalocyanine, a recalculation gives km (obs.)/km (calc.) =0.02, and for the metals at 90"K, Au 10, Al 100, Zn 0.6.(The calculations for these metalsinvolve highly speculative estimates of the number of active centres.148 CONVERSION OF HYDROGEN ON NICKELIn most cases the effect of this correction is to improve the agreement between observedand calculated rates, which is now perhaps as good as may be expected for the rare earthoxides, where the concentration of sites has been estimated on the basis of adsorptionmeasurements.1 Bond, Catalysis by Metals (Academic Press, London, 1962), p. 155.2 Eley and Shooter, J. Catalysis, 1963, 2, 4.3 Wigner, 2. physik. Chem. B, 1933, 23, 28.4 Malckar and Teller, Proc. Roy. SOC. A , 1935, 150, 520.5 Bonhoeffer, Farkas, A. and Rummel, 2. physik. Chem. By 1933,21,225.6 Bonhoeffer and Farkas, A,, Trans.Faraday Soc., 1932, 28, 242.TfiPolanyi, J. SOC. Chem. Ind., 1935, 123T.8 Rideal, Proc. Cambr. Phil. Soc., 1939, 35, 130.Eley, Trans. Faraday SOC., 1948,44, 216.p. 1095.10 Boreskov and Vassilevitch, Actes 2nd Congr. h t . Catalyse (ed. Technip, Paris, 1961), vol. 1,11 Schwab and Killmann, Actes 2nd Congr. Int. Catalyse (ed. Technip, Paris, 1961), vol. 1, p, 1047.12 Schwab and Killmann, Bull. SOC. Chim. Belg., 1958, 67, 305.13 Fajans, 2. physik. Chem. B, 1935,28,239, 252.14 Eley and Rideal, Proc. Roy. SOC. A , 1941, 178, 429.15 Bolland and Melville, Trans. Faraday SOC., 1937, 33, 1316.16Bonhoeffer and Farkas, Trans. Faraday SOC., 1932,28, 561.17 Busch, Ann. Physik, 1921, 64, 401.18 Eley, Proc. Roy. Sac. A, 1941, 178, 452.19 Farkas, A. and Farkas, L., J. Amer. Chem. SOC., 1942,64, 1594.20 Gundry, Actes 2nd Congr. Int. Cutalyse (ed. Technip, Paris, 1961), vol. 1, p. 1083.21 Schuit and Van Reijen, Ado. Catalysis, 1955, 10, 242.22 Hayward and Trapnell, Chemisorption, 2nd ed. (Butterworths, 1964), p. 75.23 Eley, Disc. Fmaduy Soc., 1950, 8, 34.24 Mignolet, Disc. Fwaday SOC., 1950, 8, 105.25 Sachtler, J. Chem. Physics, 1956, 25, 751.26 Mignolet, Bull. SOC. Chim. Belg., 1958, 67, 281.27 Sachtler and Dorgelo, 4th Int. Conf. Electron Microscopy, 1958, 1, 51.28 Sweett and Rideal, Actes 2nd Congr. Int. Cutalyse (ed. Technip, Paris, 1961), vol. 1, 175.29 Dowden, Chemisorption (ed. Garner) (Butterworths, 1957), p. 3.30 Hickmott, J. Chem. Physics, 1960, 32, 810.31 Eley and Petro, to be published.32 Gundry and Tompkins, Trans. Furaday SOC., 1956,52, 1609.33 Mignolet, Bull. Chim. SOC. Belg., 1958, 67, 358.34 Gomer, Wortman and Lundy, J. Chem. Physics, 1957,26, 1147 ; 27, 1099.35 Kiperman and Davydova, Kinetics and Catalysis, 1961, 2, 687.36 Couper, Eley, Hulatt and Rossington, Bull. Soc. Chim. Belg., 1958, 67, 343.37 Eley, Nature, 1962, 194, 1076.38 Sandler, J. Chem. Physics, 1953, 21, 2243.39 Singleton, J. Physic. Chem., 1956, 60, 1606.40 Scholten and Konvalinka, Bristol Con$ Chemisorption and Catalysis, 1965.41 Hayward, Herley and Tompkins, Surface Sci., 1964, 2, 156.42 Selwood, Adsorption and Collective Paramagnetism (Academic Press, London, 1962), p. 104.43 Kummer, J. Physic. Chem., 1962, 66, 1713.44 Sandler and Durigon, Truns. Faraday Soc., 1966, 62, 215.45 Pauling, Proc. Roy. Sac. A, 1949, 196, 543.46 Eley, Coloquio sobre Quimica Fisica de procesos en superficies sdlidas ( X X V Aniversario del47 Ashmead, Eley and Rudham, Trans. Faraday SOC., 1963,59,207.48 Sandler, Can. J. Chem., 1954, 32, 249.49 Harrison and MacDowell, Proc. Roy. SOC. A, 1953,220, 77.SO Ashmead, Ph.D. Thesis (University of Nottingham).51 Farkas, A, Farkas, L and Harteck, Proc. Roy. Soc. A, 1934,144,481.52 Farkas, L. and Garbatski, J. Chem. Physics, 1938, 6, 260.C.S.I.C., Octubre, 1964), Madrid, (1969, p. 157

ISSN:0366-9033    

DOI:10.1039/DF9664100135

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Role of Chemisorption in Simple Catalytic ReactionsBY V. PONEC, Z. KNOR AND S . &RN*Institute of Physical Chemistry, Czechoslovak Academy of Sciences,Mgchova 7, Prague 2Received 13th January, 1966Hydrogen and oxygen sorption and their interaction have been studied on evaporated films ofPt, Pd, Rh, Ni, Fey Mo, Cu and Mn. Hydrogenation of cyclopropane has been followed on films ofNi, and sorption and interaction of hydrogen and nitrogen have been measured on films of Fe.Oxygen adsorbed on nickel does not react at 78°K with hydrogen adsorbed subsequently, but restrictsthe hydrogen adsorption roughly proportional to the degree of the surface coverage by oxygen. Atroom temperature there is extensive interaction of the preadsorbed oxygen with subsequently admittedhydrogen only if part of the surface had been left uncovered by oxygen, otherwise atomization of H2in the gas phase is necessary.Platinum behaves in the same way even at 73°K. With molybdenumthere is no interaction between the preadsorbed oxygen and admitted hydrogen at 273°K even ifhydrogen is present in the adsorbed state, atomization is again necessary.Products of the interaction at 273°K differ with various metals in a characteristic way. With Niand Mo the reaction is stopped after all oxygen had already reacted and the reaction cannot berepeated on the same surface at this temperature. With Pt, Pd and Rh, however, the sequence ofoxygen and hydrogen sorption can be repeated without limitation. Cu behaves differently. Thereis no hydrogen adsorption and no interaction with oxygen even on surfaces not fully precovered byoxygen.We conclude that hydrogen enters the reaction with oxygen always in the adsorbed state.There seems no connection between the efficiency of the metal in this reaction and the presence orabsence of hydrogen causing a " positive " effect in the change of the film conductivity. Twoconditions are required if a metal is to be active in the 024-H~ reaction : (i) all reactants must beadsorbed, (ii) their heat of adsorption has to be low.In hydrogenation of cyclopropane on nickel a fast reaction takes place only on a very smallportion of the surface, while on the remaining part the dissociated and dehydrogenated species areremoved slowly. Thus, various adsorbed particles play a different role in the catalytic reaction.Further results show that adsorption is a necessary prerequisite for a catalytic reaction, but itselfdoes not represent a sufficient condition.Tn all systems studied, adsorption of both components islargely competitive.We have been studying processes on clean surfaces of metals in the form ofvacuum evaporated films. The possible existence of several kinds of adsorbedspecies or of different types of chemisorption bonds in the adsorption of one andthe same molecule has been indicated previously.1-5 Consequently, we directedour study to the following questions, which form the borderline between the problemsof adsorption and of catalysis: (i) which kinds of bonds or of adsorbed speciescan be identified by measuring adsorption isotherms, isobars, kinetics of adsorp-tion and, particularly, of electric quantities, such as resistance and work function ?(ii) what is the role of various kinds of bonds or of the strength of bond (the valueof which can be estimated by calorimetric measurements) in simple reactions ?(iii) which molecules, participating in the reaction, enter into it in the adsorbedstate? In this paper we survey results which might be helpful in answering thesequestions.EXPERIMENTALThe following metals were studied: Ni, Pd, Mo, Rh, Cu, Fe, Mn.The films wereevaporated from filaments of the respective metals, except for Cu and Mn which were14150 ROLE OF CHEMISORPTIONevaporated from electroplated layers of W and Mo. We studied, first, the adsorption andinteraction of hydrogen and oxygen, but results were obtained also with nitrogen andcyclopropane.All experiments were carried out in high-vacuum systems the descriptionand function of which, together with the procedure of evaluation of the data have beengiven.6-8 After 3-4 days of evacuation, degassing of the filament and baking out of theapparatus, the evaporation of the film took place at a vacuum of 10-9 to 10-8 torr ; withCu and Mn it was 10-8 to 10-7 torr.The film resistance changes were measured in a vessel in which the sorption and re-actions of atomized hydrogens could also be followed. The resistance was measuredby a d.c. Wheatstone bridge. The work function changes were measured by the diodemethod, the current-voltage curves being ascertained before and after the adsorption ofgas.7 The vessel used was constructed in principle according to the design described inref.(9). Gas was admitted always to a cold cathode that was cleaned before the measure-ment of each current-voltage curve by repeated flashing at ca. 2300°K. Heats of ad-sorption were measured 10 in a Beeck-type calorimeter 11 (heat capacity 1.4 calldeg., sensi-tivity of 1 x 10-3 cal corresponding to 1 crn of deflection of the measuring device 10) usingan a.c. method which permitted the automatic recording of the heat effects.12 The surfacearea was measured using Kr or Xe.13 Catalytic activity in the hydrogenation of cyclo-propane was followed manometrically in a closed apparatus 14 at low pressures.RESULTS AND DISCUSSIONADSORPTIONComparison of the extent of chemisorption of gas with that of physical adsorptionof an inert gas allows estimation of the degree of surface coverage by the chemisorbate.With further assumptions, and with a suitable choice of chemisorbate, this comparisonmay also give an estimate of the " valency " of the metal surface atoms.As shown in table 1, hydrogen covers the surface of most metals to a high degree.In palladium, hydrogen dissolves 193 20 (at 78°K almost up to the ratio of Pd/H = 1)TABLE SURFACE COVERAGE OF VARIOUS METALS AT COMPLETE ADSORPTION OFmetalsFeNoNiRhPtlatticebody-centredcubecentredcubeface-centredcubeface-centredcubeface-centredcubebody-HYDROGEN 15-18no.of sites on1 cm2 of thesurfacex 1015(110) 1-72(100) 1-21*(211) 1.98(110) 1-44(loo) 1-02'(211) 1.72(111) 1-86(100) 1-61(110) 1.14(111) 1-62(100) 1.41(110) 0.99(111) 1.50(100) 1.31(1 10) 0.92mean no. of sites on1 cm2 at equal dis-tribution of 511planes x 10131-431 *371.541.331 924no. of H atomsadsorbed on 1 cm*x 10150-761.181.20-991 -04* According to Brennan, Hayward and Trapriel175 the number of atoms actually exposed bythis plane (and accessible to gases) is double that which corresponds to the number of atoms in asingle surface layerv. PONEC, z. KNOR AND s. E E R N ~ 151and only with Mn is the adsorption at 78°K very small and increases 21 substantiallyonly with increasing temperature.On copper, gaseous hydrogen is adsorbed atT<30OoK only when atomized 21 in the gas phase.A rapid chemisorption of hydrogen at 78°K is followed, as a rule, by a slowchemisorption of small extent (5-10 % of the total amount adsorbed.5~ 8, 16, 179 22, 23)The kinetics of the slow chemisorption is described by the logarithmic equation:be = 2.3 log/(t+to)+C, where C, to and b are constants and to is usually close tozero. The kinetics are related to the film porosity: on more dispersed films, theconstant b is higher.8, 23 The adsorption kinetics at 78°K are similar for metalsof a different electronic structure (see below) such as Ni,59 23 Fe,17* 22 on the onehand, and Mo8 and Rh16 on the other. At around 273"K, the adsorption ofhydrogen is practically instantaneous on all transition metals.The hydrogen isobars 16 on these metals differ to a greater extent.Accordingto the type of isobars, schematically shown in fig. 1, metals can be divided into twoloo 200 300 'K )[30 200 300TFIG. 1 .-Schematic illustration of two types of isobars ; type I, for N i , 5 , 2 3 Fe,17*26, Co 24 ; type 11,for Mo,* Rh.16,26 E, ratio of the amount adsorbed at given temperature to that directly measuredat 78°K ; Aemax, maximum divergence of isobar branches.groups. All metals without exception show a " divergence " (AE) of the twobranches of the isobars, the upper branch being always nearly reversible. Thedivergence suggests that a part of the film surface with sites of high heats of adsorptioncannot be covered at 78°K and that this activated process can only take place athigher temperatures.This process competes with a decrease in the extent of ad-sorption with increasing temperature, caused by desorption of hydrogen fromsites with a lower heat of adsorption. As a result, a maximum appears on thelower branch of the isobars. The film structure (i.e. probably the difficult accessi-bility of some adsorption sites) is responsible for at least a part of the activatedadsorption. The value of of the hydrogen isobars on nickel changed inparallel with the film dispersity and was greatest with the most dispersed films.23The most characteristic difference of transition metals for hydrogen adsorptionis revealed by changes of their electric properties.Changes of resistance of filmswere studied in the greatest detail. The types of dependences obtained can bedivided into three groups schematically shown in fig. 2 : (i) those with a maximum(hydrogen on Fe,191 279 28 C0,27 Ni,lS 239 279 289 297 Pd19- 20), followed by aminimum on some metals (Ni) at 78°K. (ii) Those with decreasing slope, and afte152 ROLE OF CHEMISORPTIONthe minimum-at 78°K-also with an ascending branch (Pt).179 30 (iii) Those withpermanent increase (d(AR)/dO)>O) showing at 8+1 at T = 78°K a greater resistanceincrease (hydrogen on M o , ~ Rh,16 Mn 173 27 ; also Ti 28, Cr 27). The work functionchanges, accompanying the hydrogen adsorption,39 319 329 29 are similar.I ' 3aeFIG. 2.-Schematic illustration of types of resistance changes AR/R with the extent of adsorption8 for various metals (see text).Oxygen, as contrast hydrogen, effects an increase of film resistance with Ni,17Pt,17 Rh,16 No 8 and Fe 17 of the type shown in fig.3a and with Mn of the typegiven in fig. 3b. Nitrogen is chemisorbed on iron to a small extent only, but itcauses a relatively large change of resistance.17 The dependence of resistance changeon the extent of adsorption is here practically a linear one (at 273°K). Cyclo-propane 14 behaves similarly to oxygen. Also other strongly adsorbed gases be-have in an analogous way.33-35 Therefore the positive effect, ie., the decrease ofthe film resistance with increasing adsorption on clean or on already partially coveredsurfaces is exhibited almost exclusively by hydrogen which is distinguished by itssmall dimensions.The rare existence of the positive effect occurs with other gases only if the ad-sorption tends to become physical in nature.Thus, according to Hansen andLittmann,36 xenon exhibits the positive effect on extremely thin films composed ofisolated islands. (With these films it is possible to accept the explanation suggestedby Broeder et aZ.37) From other gases described, signs of the positive effect werefound with CQ 389 33 at 78°K and higher pressures.38~ 33 However, a weak ad-sorption (which for CQ and Xe is most likely of the molecular type) is not alwaysaccompanied with a decrease of film resistance. We found 7 9 17 with nitrogenthat a weak reversible adsorption at 78°K on a surface already covered by chemi-sorption at 273"K, caused an increase of the resistance, although the change of thework function points to a polarization of the type Me-Ni.Displacement of hydro-gen from an iron surface by carbon monoxide proved that a part of the hydrogev. PONEC, z. KNOR AND s. C E R N ~ ‘ 153desorbed at 273°K first passes into a weakly bonded state giving a positive effect,and only then desorbs.38According to this mechanism, “ positive ” hydrogen may be weakly bound bymeans of d-orbitals of atoms on which possibly CO is already adsorbed. Thismakes it possible also to understand that at 78°K simultaneous adsorption of COand H2 is possible on the same surface and that the extent of adsorption for bothgases is practically the same as on the clean surface.The increased concentration04 08 0 1 2 3 40 a02FIG. 3.-Dependence of the relative resistance change (a, 6) on the extent of hydrogen sorptionat 298°K. For clean surface of Ni-film (a) and a surface partially covered by oxygen pre-adsorptionDependence of the hydrogen sorption (@HJ on the extent of oxygen preadsorption (@02) at 78°K(c) and 298°K (d). @ express the extent of sorption in multiples of hydrogen “monolayers”(complete hydrogen adsorption at 78°K).(b). 8 = 1 corresponds to the maximum hydrogen uptake in each case.of “ positive ” hydrogen is shown in the results of Siddiqi and Tompkins 39 (theirfig. 2). The time course of resistance changes on admission of hydrogen in theregion of the maximum of the curves on nickel, etc.17~ 29 and, furthermore, the factthat hydrogen, during desorption, first of all passes through the state in which itis bound by a bond causing the positive effect, suggests that such a bond constitutesan intermediate stage not only for desorption but also for adsorption.This wouldimply, according to the interpretation proposed below, a confirmation of the modelsuggested by Gundry and Tompkins 5 to explain the kinetics of adsorption.The positive effect and its occurrence together with the negative effect, at variousdegrees of coverage of the same transition metal have not as yet been explainedsatisfactorily. The positive effect of hydrogen cannot be explained by interactionwith impurities (especially with oxygen).On iron 17 and nickel,409 41 hydrogendoes not react with oxygen at low temperatures and in the platinum+hydrogensystem 17 the time course of the resistance changes in the region of the positiveeffect has a distinctly different character in the presence or absence of oxygen onthe platinum surface. Nor does it seem correct either to relate the positive effectto hydrogen dissolution and to the formation of H+ species as has been suggested,4154 ROLE OF CHEMISORPTIONbecause with palladium-in which hydrogen dissolves and the electron structureof which is similar to that of nickel-the dissolved hydrogen causes an increase ofresistance.Igs 20 Finally, neither the existence of species of the H f , H i type or ofmore complex species of various charges,ls 43 etc., is a convincing explanation.A simultaneous co-existence of two differently charged species-ions-on the surfaceof a conducting metal does not seem possible because adsorbed particles are mobileat temperatures above 150OK.4 The explanation of the positive effect by speciesof the H+ type meets with other difficulties.For example, nickel is a metal of pre-dominating electron-conductivity, whilst iron or molybdenum are metals of pre-dominating hole-conductivity.45~ 46 Thus, the similarity of the curves showing thedependence of the resistance changes on the extent of adsorption for Ni and Fewould signify that hydrogen is adsorbed at low coverages on nickel as a negativeion and on iron as a positive ion. But, at the same time, hydrogen at low coveragesalways increases the work function of the above metals and with iron and nickelthe strength and polarity of bond are very similar (see table 2).With molybdenum,formation of the positive hydrogen ion, which would have to be assumed over thewhole extent of adsorption, is still less probable, because with molybdenum the bondis more polarized in the sense of Mo+H-. (Placing of hydrogen under the surfaceatoms is unlikely, because of the great strength of the molybdenum lattice; thus,chemisorbed gases would penetrate to a far smaller depth than with iron and nickel.)It seems, therefore, one must look for an explanation of the positive effect in termsof the electron structure of the transition metal.19, 23TABLE 2.4OMPARISON OF STRENGTH AND POLARITY OF HYDROGEN ON SOME hlETALSNi Fe M Oheat of adsorption at 840,work function change A$maroom temperature (kcal mole-1) 26-30,47 26 10 32 47 40 10at 8- 1, [V] - 0.35 31 - 0.26 7 - 0.43 31 -0.28 7 - 0.20 7+ 3- +work function change A&=;the atomized hydrogen sorption ;78°K; compared with clean sur-face, PI -0.06 * 7 -0.15 * 7film resistance change (AR/A8) ;atomized hydrogen sorption + + +* after 30 min sorption the value depends on the amount of added atomized hydrogen.Transition metals are distinguished by a characteristic electronic sfructure.45p 47-50Electron orbitals of ns and (n-l)d atoms with f.c.c.metals and of as, np and(n- 1)d atoms with b.c.c. metals hybridize and give rise to delocalized orbitals andto corresponding bands of energetic levels, as for metals of a simpler electronstructure.51 It appears, however, that it is not a single broad band of levels thatis formed, but that there are several more or less distinguishable and mutually over-lapping sub-bands,4s the energy levels of which predominantly belong to orbitalsof one kind, e.g., with f.c.c.metals, to orbitals of practically only s or only d char-acter.52 Thus, one often speaks approximately of separate s- or d-bands of transi-tion metals. With b.c.c. metals there are also orbitals localized on individualatoms in addition to the spd band which-as indicated by the positive sign of thv. PONEC, z. KNOR AND s. EERNY 155Hall constant-clearly divides further into sub-bands.45 Atomic localized orbitals areformed from dz2 and d3c2-yz orbitals of isolated atoms. Electrons of not fully occupiedd-bands or electrons on localized d-levels are responsible for the main part of themagnetic moment of metals.52 The following two specific effects have been sug-gested to explain the high resistance of transition metals : (i) the scattering of con-ducting electrons by jumping into holes in the d-band, which-due to their higheffective mass-do not themselves participate in the conduction process.52 (ii) Thechange of the electron mobility by interactions with unpaired spins of electronsin the d-bands or in localized d-levels.49~ 53 These effects can assert themselvesalso in the adsorption of gases.On the basis of the semi-empirical theory of transi-tion metals the following model is suggested.On all metals, excepting platinum, hydrogen is adsorbed at low surface coverageby means of s-orbitals (f.c.c.metals) or spd-orbitals (b.c.c. metals) by a covalentbond. Thus, the participation of these orbitals in electrical conduction is restrictedand the metal resistance increases. If the metal has at its disposal for bondingalso electrons on not fully occupied d-levels, irrespective of whether the electronsbelong to localized or delocalized orbitals, then hydrogen is bound at a highersurface coverage with these orbitals by a weaker bond which reduces the film re-sistance by means of the effects (i) and (ii). The comparison of the electronegativitiesof metals 47 and hydrogen suggests that the covalent bond is, most probably, weaklypolarized in the Me+H- sense.It can be inferred, therefore, that hydrogen, whichcauses,31, 37 in most cases, an increase of the work function, is above the metalsurface. The hydrogen bond in the region where the film resistance is decreased,e.g., on Ni, also causes a decrease of the work function.29~ 31 This can mean eitherthe bond polarization is in the reverse direction or, more probably, that orbitalsof d-character extend to a lesser distance than s-orbitals of higher quantum number,which leads to the placing of hydrogen, polarized as Me+H-, under the surface level.With platinum, which has few reactive s-orbitals and a high work function, andoxidizes with great difficulty, hydrogen is bound with Me-Hf polarization first withd-orbitals and only at higher surface coverage also with s-orbitals which are, onplatinum, less advantageous for the chemisorption bond.This is the reason whythe change of film resistance in hydrogen adsorption on platinum is reversed ascompared with other transition metals.The adsorption of atomized hydrogen on a surface already covered by the ad-sorption of non-atomized hydrogen, leads to a decrease of the value of work function(table 2) and, eventually, even to a decrease of the work function below the valueat a clean surface.29 The explanation can be sought in the hydrogen bond (whichis the site of slight negative charge) with the metal atoms under the outermost metallayer already occupied by the preceding adsorption.Since these atoms in the lowerlayers of metal have a greater number of neighbouring atoms than the surface atoms,the hydrogen bond with them is weaker and therefore the adsorption of atomizedhydrogen, at lower temperatures, is reversible.21 The assumption of bonding withatoms at different distances from the surface of metal explains why if adsorptionof oxygen is followed by uptake of atomized hydrogen, interaction takes place,whilst with admission of gases in the reverse order the sorbed atomized hydrogendoes not enter into interaction.21Oxygen (and also most likely other strongly adsorbed gases) bonds with metalatoms so strongly that it tears them off the metal, thus the film resistance is moremarkedly increased than by hydrogen adsorption.54~ 36 The bond with d-orbitalsof these atoms is either impossible or it does not reveal itself in the film resistancevalues156 ROLE OF CHEMISORPTIONINTERACTION I N THE ADSORBED LAYERThe competitive character of adsorption of gases, which can react together athigher temperatures, has been found by us (in addition to Hz+CO+Fe system38)also in other systems.Hydrogen is displaced by oxygen from molybdenum surface ; 8adsorption of hydrogen on iron precludes the slow chemisorption of nitrogen,l7 etc.On the basis of our own experiments 409 4 1 9 21 on the interaction of pre-adsorbedoxygen or hydrogen with the other gas in the gaseous phase, we concluded thathydrogen interacts with oxygen in the adsorbed state. The same condition probablyholds also for oxygen when it interacts with adsorbed hydrogen.Conditions forinteraction were examined in the following experiments.Oxygen was pre-adsorbed, e.g., on nickel 40941 ; hydrogen subsequently admittedat 78°K did not react with the pre-adsorbed oxygen ; it merely restricted the subsequentadsorption of hydrogen to an extent roughly proportional to the part of the surfacewhich had been left unoccupied by oxygen (fig. 3c). At room temperature an ex-tensive interaction took place with the pre-adsorbed oxygen,403 41 but only if somepart of the surface had been left uncovered by oxygen adsorption. Interactionat 298"K, and its absence at 78"K, is evidenced by two facts : (a') change of the filmresistance and of the work function at 298°K is, after admission of hydrogen tosurface with pre-adsorbed oxygen, the reverse of the normal change with clean sur-face (fig. 3a).(b) The extent of hydrogen consumption was several times greater onsurfaces with pre-adsorbed oxygen (if the extent of oxygen pre-adsorption was suf-ficiently great) than on clean surfaces of the same area (fig. 34. If no part of thesurface is left free, consumption of hydrogen decreases almost to zero and no inter-action can be detected by measuring the film resistance. However, under analogousconditions, Quinn and Roberts 55 found a certain interaction by the work functionmeasurement. If the surface of nickel is completely covered by oxygen, reactionof hydrogen with pre-adsorbed oxygen can take place at 273°K if hydrogen is atom-ized in the gaseous phase.21 Platinum behaves in the same way, the only differ-ence being that the reaction can be detected but at 78°K.With M o , ~ Mn 17 andFe 17 no signs of interaction are found at 273"K, even if hydrogen is present on thesurface in the adsorbed state. On these metals, hydrogen can react with the layerof pre-adsorbed oxygen 179 21 only if it had been atomized in the gaseous phase andit again participated in the reaction after it had been sorbed by film.8, 41 Forsorption of oxygen by the film containing irreversibly bound pre-adsorbed hydrogen,the film resistance change is shown in fig. 4, and indicates that oxygen can react inthe adsorbed state. As the adsorption is undoubtedly faster than interaction itseems probable that oxygen also enters the reaction in the adsorbed state.The extent of interaction of hydrogen with pre-adsorbed oxygen at 273°K differsfor various metals.On Fe 17 and Mn 17 the extent is small ; on nickel 40, 41 andmolybdenum,8 reaction stops after the pre-adsorbed oxygen had reacted to a degreecorresponding to a ratio H/O not much greater than unity. On platinum,l7palladium19 and rhodium,16 oxygen and hydrogen can mutually react at 273°Kwithout limitation and the product of reaction desorbs from their surface. Oncopper,21 no adsorption of gaseous hydrogen takes place and therefore hydrogendoes not interact with pre-adsorbed oxygen. However, after hydrogen had beenatomized in the gaseous phase, it reacts easily with pre-adsorbed oxygen. Thisreaction can be repeated without restriction which means that, with Pt, Rh, Cu and Pd,the product desorbs at 273°K from the surface and makes the continuation ofreaction possible.Where reaction Metals clearly differ in the products of interaction at 273°Kv.PONEC, z. KNOR AND s. E E R N ~ 157can be repeated without limitation, i.e., on Pd, Pt, Rh and Cu, the most probableproduct is desorbable water. With other metals, the reaction proceeds probablyonly to the stage of formation of the OH group. This is indicated by the indirectindications just mentioned, and also by other results.56~ 57 The formation of theOH group at 273°K also agrees well with the fact that water on the surface of nickeland iron decomposes at 273°K with formation of hydrogen which is liberated intothe gaseous phase.s*$ 59 The fact that even with atomized hydrogen a desorbableproduct does not form on Ni, Fe, Mo and Mn at 273"K, shows that the additionof further hydrogen to the OH group, which is the most probable product of inter-action, is a difficult process and that water, which can be produced under suitableconditions (e.g., nickel catalysts are used in industry for removal of traces of oxygenat low temperatures), is probably formed by a mechanism other than that of themere addition of H+OH.5 10 15t, minFIG.4.-Time course of film resistance change and of oxygen sorption on Pt film covered by pre-adsorption of irreversibly bound hydrogen.The overall order of the activity of metals : Pt, Pd, Rh; Cu (H atoms) > Ni > Mo,Fe, Mn, Cu; as derived from the temperature and other conditions under whichhydrogen + oxygen interaction takes place, is generally in good agreement withthe catalytic experience.There is no direct connection between the catalytic activityand the presence or absence of the positive effect on the film resistance againstcoverage plots, i.e., between catalytic activity and the various kinds of chemisorptionbonds of hydrogen. The same conclusion holds also for other catalytic reactionsof hydrogen.605 71 Nor can other parameters, such as the shape of the isobars, thekinetics of adsorption, number of oxygen layers, extent of adsorption, etc., bedirectly correlated with catalytic activity. (This is not surprising if the above ex-planation of the adsorption kinetics and isobars is accepted.) The most activemetals are evidently those which adsorb oxygen and hydrogen most weakly.How-ever, as e.g., with copper, and similar non-transition metals, the low heat of ad-sorption of reacting gases is not a sufficient condition for the reaction to procee158 ROLE OF CHEMISORPTIONeasily at a low temperature. Nor is the adsorption of reactants a sufficient con-dition for a catalyst to be active for the reaction, as it is shown by results obtainedwith Mo, Fe and Mn. However, we arrive at the above order of metal activitiesin oxidation (and other reactions) of hydrogen if the two requirements are com-bined: (i) adsorption under reaction conditions of the reactants is the necessaryprerequisite; (ii) the high activity requires a low heat of adsorption of participantsentering reaction.The high reactivity of more weakly bound species has been postulated.7~ 61962In our experiments with cyclopropane we also found that only the part of the nickelsurface not covered by strongly bound products of cyclopropane chemisorptionis active in fast hydrogenation.According to results obtained by Thomson andWislilade 63 and Stephens 64 the same conclusion probably holds for the hydro-genation of ethylene.Sachtler and Fahrenfort 65 first proved directly from the hydrogen-deuteriumexchange on gold that the adsorption of both reactants is a necessary prerequisitefor a catalytic reaction; this is also probably the interpretation of the hydrogen-deuterium exchange on other non-transition elements.67 Sachtler and de Boer,66and also Kazansky and Voevodsky,68 further proved experimentally that hydrogenenters oxidation in the adsorbed state on gold66 and palladium.68 A similarconclusion results from our experiments for Ni,40* 41 Mo 8 and Pt.17 Since on thesemetals the adsorption of oxygen during the reaction is possible (this also holds forgold according to the latest results 69, the mechanism requiring all participants toenter the reaction in the adsorbed state again seems to be the most probable.Thegeneral character of this conclusion suggests the reason why the Rideal type ofmechanism, where some of the reaction participants react by an impact from thegaseous phase or from the physically adsorbed state, has-contrary to the preceding(Bonhoeffer-Farkas) mechanism-never been directly proved experimentally butwas based on kinetic arguments only.Beeck70 first suggested and proved a correlation between the high heat of ad-sorption and low catalytic activity.A similar correlation follows from resultsobtained by other authors in spite of their own different interpretations of theirresults.609 71 There are also several theories of the catalytic activity of solids, basedon a correlation with thermodynamic parameters (for review, see ref. (71)). How-ever, these correlations do not apply, e.g., to metals which do not adsorb one or moreof the reactants, such as, e.g., Au, Ag, Cu, Pb, Sn and other non-transition elementsin hydrogenation or oxidation reactions, or Pt, Pd and Ni in ammonia synthesis,etc.This clearly shows that energetic factors alone cannot serve as a guide for arational choice of catalysts. Nevertheless, Pt and Rh are generally, for a numberof reactions among the most active metals and adsorb gases comparatively weakly(see the general rule of Tanaka and Tamaru 72), whilst Mo, W, Ta, Ti, etc., belongto the least active group of metals and adsorb all gases very strongly. Nickel andcobalt, etc., occupying a position between these extreme groups, are closer to thegroup of active metals which is in accordance with the values of heats of adsorption.NOW, one may ask what is the basis of the low heat/high activity correlation.The reasons might be as follows : (i) high heat of adsorption means a low potentialenergy of adsorbed species and if the exothermic effect of the overall reaction isnot high, the surface reaction among adsorbed particles is highly endothermic.Since the endothermicity represents the minimum possible activation energy, thenalso the surface reaction is accompanied by a high activation energy.(ii) Re-actants form a product with an exothermic or a small endothermic effect, so thatthe final step-the desorption of products-must be strongly endothermicv. PONEC, z. KNOR AND s. EERNS 159According to the general rule of Tanaka and Tamaru72 the higher is the heat ofdesorption of the products, the higher is the heat of adsorption of the reactantsand thus again the correlation of the activity with the heat of adsorption of theinitial reactants is obtained.(iii) On metals with a higher heat of adsorption ofmore complex molecules such as hydrocarbons, the degree of surface coverage bya layer of non-reactive products of dissociative chemisorption is doubtlessly greaterthan on metals with a lower heat of adsorption.70~ 64 The metals then differ in thearea of the working surface and its rate of the reaction on this surface where againeffects (i) and (ii) are operative.From the possible operation of mechanisms (i)-(iii) there result definite con-clusions for the relation between the absolute value of the heat of adsorptionand the activation energy. The greatest differences in energies of activation on thevarious metals is expected with mechanism (i) ; according to mechanism (iii) metalsdiffer but in the value of the frequency factor.However, the activity of some metals cannot be explained along the lines above.Among these is the recombination of H-atoms60 and other gases, and the decom-position of formic acid.73 Hydrogen atoms are weakly adsorbed by a number ofnon-transition elements, nevertheless the activity of these elements in recombinationis low.Therefore, the correlation of catalytic activity with heat of adsorption showsa maximum at the optimum heat of adsorption, i.e., for the lower heats of adsorption(or of bond strengths or other similar parameters), the activity of metals increaseswith the heat of adsorption, and decreases for the region of higher heats of ad-sorption.This can be explained as follows. On transition metals where the lowheat/high activity correlation holds, hydrogen recombines and desorbs in a singlestep with an activation energy close to the heat of adsorption of hydrogen with zero“ intrinsic ” activation energy (fig. 5 4 . On non-transition metals it is not the de-sorption itself but the recombination of atoms in the adsorbed state that is difficult(see scheme 5b) and is connected with a high intrinsic (non-endothermic) activationenergy. The different rate-determining steps for transition and non-transitionU(4 (b)FIG. 5.-Schematic illustration of the course of potential curves for hydrogen adsorption on transi-tion (a) and non-transition (b) metals. 1, Physical adsorption ; 2, weak adsorption of atoms boundwith d-orbitals of metals ; 3, strong chemisorption of atoms.Chemisorption on non-transitionmetals endothermic (2’) and exothermic (3’) ; E&s, activation energy of desorption. Curve 2 doesnot represent the “equilibrium ” d-orbital-H bond of p. 155.elements explains the existence of two branches in the correlation of catalyticactivity with the heat of adsorption. In formic acid decomposition, again a goodvolcano-shaped correlation with the heat of formation of formates was obtained,In order to explain the maximum, we must again suppose a different rate-determinin160 ROLE OF CHEMISORPTIONstep for metals on the right and left side of the correlation. It seems improbablethat in the formation of the activated complex adsorption only should play therate-determining role with metals of the left side of the correlation,73 because, e.g.,on silver an adsorbed layer is easily formed by adsorption of formic acid.74 But theassumption of a different rate-determining step for metals on the left and rightside of the heat-activity correlation itself, is acceptable.Thus we conclude that, in addition to the necessity of adsorption of reactants,there exists another prerequisite limiting the operation of the correlation of highactivity with low heats of adsorption, viz., that such correlation can hold goodbut for a group of metals with the same rate-determining step in the given reaction.The low heat/high activity correlation does not hold where Bronsted empirical rulerelating the activation energy with the heat of adsorption operates as is so with non-transition elements on the left side of correlations in hydrogen recombination, formicacid decomposition and possibly in further reactions.In conclusion, an electronic factor may operate when the above rules hold.Asuitable electronic structure makes possible a rapid adsorption of reactants withouta noticeable activation energy, though even strong bonds must dissociate duringadsorption (fig. 5a). The electronic structure of metals is, furthermore, decisivefor the strength of the adsorption bond and consequently operates through rule(ii) concerning the magnitude of the heat of adsorption, and also renders it possiblethat the surface reaction proceeds without a high intrinsic activation energy.Acombination of all effects of the electronic factor lead to complications in correlatingactivities with electronic structures. It seems, therefore, that any correlation of thecatalytic activity with parameters characterizing merely the electronic structure ofmetals, will be found and understood only by analyzing step by step the mechanismof the overall catalytic reaction as indicated above.1 Sachtler and Dorgelo, Bull. SOC. chim. Belg., 1958, 67, 465.2 Mignolet, Bull. SOC. chim. Belg., 1958, 67, 358.3 Mignolet, J. Chim. Physique, 1957, 54, 19.4 Dowden in Chemisorption, ed. Garner (Butterworths, London, 1957), p. 3.5 Gundry and Tompkins, Trans. Faradizy SOC., 1956, 52, 1609.6 Knor and Ponec, Coll. Czech. Chem. Comm., 1961,26, 529.7 Knor and Ponec, Coll.Czech. Chem. Comm., 1966, 31, in press.8 Ponec, Knor and &rnS;, Coll. Czech. Chem. Comrn., 1965, 29, 3031.9 Pritchard, Trans. Faraduy SOC., 1963, 59,437.10 CemS;, Ponec and HlBdek, J. Catalysis, 1966, 5, 27.11 Beeck, Cole and Wheeler, Disc. Faraday Soc., 1950, 8, 314.12 Hlhdek, J. Sci. Instr., 1965, 42, 198.13 Ponec and Knor, Coll. Czech. Chem. Comm., 1962, 27, 1091.14 Knor, Ponec, Herman, DolejBek and cernf, J. Catalysis, 1963, 2, 299.15 Knor and Ponec, COIL Czech. Chem. Comm., 1961, 26, 961.16 Ponec, Knor and CernS;, Coll. Czech. Chem. Comm., 1965,30, 208.17 Ponec, Knor and cerny, unpublished.18 Boreskov and Vasilevitch, Kinet. katuliz, 1960, 1, 69.19 Knor, Ponec and eernS;, Kinet. kataliz, 1963, 4, 437.20 Suhrmann, Wedler and Schumicki in Structure andProperties of Thin Films (Willey, New York,21 Ponec, Knor and tern$, J.Catalysis, 1965, 4, 485.22 Porter and Tompkins, Proc. Roy. Soc. A , 1953, 227, 529.23 Ponec and Knor, Actes 2nd Congr. Int. Catalyse, (Technip, Paris, 1961) p. 195.24 Ponec, unpublished.25 Gundry and Tompkins, Trans. Faraday Soc., 1957, 53, 218.26 Lanyon and Trapnell, Proc. Roy. SOC. A , 1955,227, 387.27 Suhrmann, Hermann and Wedler, 2. physik. Chem., 1962,35, 155.28 Zwietering, Koks and van Heerden, J. Phys. Chem. Solids, 1959, 11, 18.1959), p. 278v. PONEC, z. KNOR AND s. ~ E R N J 16129 Tretyakov and Balobnev in Mechanizm vzaimodeystviya metullov s gazarni (Izd. Nauka, MOSCOW,30 Suhrmann, Wedler and Gentsch, 2.physik. Chem., 1958, 17, 350.31 Culver and Tompkins, Adv. Catalysis, 1959, 11, 67.32 Crossland and Pritchard, Surface Sci., 1964, 2, 217.33 Wedler and Fouad, 2. physik. Chem., 1964,40, 12.34 Suhrmann, Schwandt and Wedler, 2. physik. Chem., 1962, 35, 47.35 Gryamov, Shimulis, Yagodovsky, Dokl. Akad. nauk U.S.S.R., 1960, 160, 1132.36 Hansen and Littmann, 2. Elektrochem., 1963, 67, 970.37 Broeder, van Reijen, Sachtler and Schuit, 2. Elektrochem., 1956, 60, 838.38 Cukr, Merta, Addmek and Ponec, Coll. Czech. Chem. Comm., 1965,30,2682.39 Sidiqqi and Tompkins, Proc. Roy. SOC. A , 1962,268,452.40 Ponec and Knor, Qll. Czech. Chem. Comm., 1962,27, 1443.41 Ponec, Knor and Cern3, Proc. 3rd Int. Congr. Catalysis, Amsterdam (North Holland Co.,42 Sachtler and Dorgelo, 2. physik. Chem., 1960, 25, 69.43 Suhrmann, Adv. Catalysis, 1955, 7, 303.44 Gomer, Disc. Faraday SOC., 1959, 28, 23.45 Goodenough, Physic. Rev., 1960, 120, 67.46 Frank, Appl. Sci. Res. B, 1957, 6, 379 ; 1958, 7, 41.47 Hayward and Trapnell, Chemisorption (Butterworths, London, 1965).48 Mott and Stevens, Phil. Mag., 1957, 2, 1304.49 Mott, Adv. Physics, 1964, 13, 325.50 Vonsovsky and Izyumov, Usp. Fiz. Nauk U.S.S.R., 1962, 77, 379 ; 1962, 78, 4.51 Reitz, Adv. Solid State Physics, 1955, 1, 1.52 Mott and Jones, The Theory of the Properties of Metals and Alloys (Oxford, 1936).53 Coles, Adv. Physics, 1958, 7, 40.54 Sachtler and van Reijen, J. Res. Inst. Cat., Hokkaido Univ., 1962, 10, 87.55 Roberts, Proc. 3rd Int. Congr. Catalysis, Amsterdam (North Holland Co., Amsterdam, 1965),56 Farnsworth, Schlier and Tuul, J . Physics Chem. Solids, 1958, 9, 57.57 Campbell and Thompson, Trans. Faraday SOC., 1961, 57, 279.58 Suhrmann, Heras, Heras and Wedler, Ber. Bunsenges. physik. Chem., 1964, 68, 511.59 Lazarov, lecture held at the Faculty of Sciences, Charles University, Prague, 1964.60 Bond, Catalysis by Metals (Academic Press, New York, London, 1962).61 Trapnell and Rideal, Disc. Faraday Soc., 1950, 8, 114.62 Dowden, Bull. SOC. chim. Belg., 1958, 67, 439.63 Thomson and Wishlade, J. Chem. Soc., 1963,4278.6'1 Stephens, J. Physic. Chem., 1958, 62, 714.65 Sachtler and Fahrenfort, Actes 2nd Congr. Int. Cutalyse, Paris (Technip, Paris, 1961), p. 831.66 Sachtler and de Boer, J. Physic. Chem., 1960,64, 1579.67 Fensham, Tamaru, Boudart and Taylor, J. Physic. Chem., 1955, 59, 806.68 Kazansky and Voevodsky, Dokl. Akad. nauk U.S.S.R., 1957,116, 633.69 Kulkova and Levchenko, Kinet. kataliz, 1965, 6, 765.70 Beeck, Disc,Faraday SOC., 19S,O, 8, 118.71 Ponec and Cerny, Rozpravy Cs. akad. v6d (Disc. of Czechoslovak Acad. of Sci.), 1965, 75,72 Tanaka and Tamaru, J. Catalysis, 1963, 2, 366.73 Fahrenfort, van Reijen and Sachtler, 2. Elektrochem., 1960, 64, 216.74 Tamaru, Trans. Faraday Soc., 1959,5§, 824.75 Brennan, Hayward and Trapnell, Proc. Roy. SOC. A , 1960,256,81.1964), p. 50.Amsterdam, 1965), p. 353.p. 365.no. 5 (in English)

ISSN:0366-9033    

DOI:10.1039/DF9664100149

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Nature and Reactivity of Nickel and Oxidized Nickel SurfacesBY M. W. ROBERTS* AND B. R. WELLS*Dept. of Chemistry, The Queen's University of Belfast, Belfast, N. IrelandReceived 17th January, 1966The role of oxygen in determining the surface characteristics of nickel catalysts has been in-vestigated by work function, photoelectric and chemical reactivity studies. The possible limita-tions of any one particular method are discussed. Surfaces which have been exposed to oxygen(10-4 m min) at 23" exhibit the characteristics expected of a surface oxide although the chemicalreactivity is not that associated with NiO. Oxide structures of low work function can form, andthese were shown to have specific adsorption characteristics.The adsorption of carbon monoxidc and oxygen and the catalytic oxidation of carbon monoxideby heavily oxidized nickel films were studied.Two distinct states of adsorbed oxygen were recog-nized. A substantial decrease in work function occurred on adsorbing carbon monoxide, the de-crease being greater the higher the temperature. At 170" and 1.5 mi the decrease was 4 . 8 5 eVthe major portion of this arising from a space charge term. The oxide work function remainedvirtually unchanged during catalysis at 23". This was due to a low energy state of adsorbed oxygeninhibiting the adsorption of carbon monoxide thereby preventing the development of a space-charge.The traditional approach to the study of the nature and reactivity of surfaceshas been based mainly on adsorption measurements ; these have involved kinetic,thermodynamic and more recently spectroscopic studies. The adsorbates that havebeen most extensively used are nitrogen, oxygen, carbon monoxide and hydrogen,and distinct correlations of activity with the nature of the metal were obtained.1This approach is, however, limited and recently more sophisticated physical methodshave been used with more care taken to define the initial crystallographic and chemicalnature of the surface. In particular, flash filament,2 field emission,3 Auger emission 4and diffraction methods 5 have provided more detailed information.In this laboratory we have been concerned with the study of oxygen interactionwith metal surfaces from two points of view ; first, the mechanism of the reactionparticularly with reference to the onset of iiicorporation or oxidation and secondlythe possible influence of surface oxygen on the chemical reactivity of the substratc.We have used Ni+02 as a model system for our investigations.The intrinsicchemical reactivity of a solid surface depends on the nature of the szrface orbitalswhich in turn depend on both the structure and the chemical environment of thesurface atom. Therefore the problem of whether one is studying catalysis with ametal or a modified surface (possibly oxide) is important since any attempt toforinulate models for adsorbed species depends on the electron distribution at thesurface.Much theoretical and experimental effort has recently been made to understandthe nature of surface free valencies (the physicist's surface states).Eley and hisco-workers 6 have used the paramagnetic conversion of parahydrogen as the surfacesensitive parameter while electron transfer processes occurring during adsorptionhave been followed by conductivity changes. In the present work we have usedphotoelectric, work fmctiai and chemical reactivity studies to explore the natureof a typical catalyst (nickel) present in a reaction either as the metal or oxide. Anoutline of the theory underlying the electrical measurements of particular relevanceto this work is given (see also the reviews by Green,7 Pluinmer and Wolkenstein 29).* Now at: School of Chemistry, University of Bradford.16M. W. ROBERTS AND B. R. WELLS 163EXPERIMENTALChanges of work function during gas interaction with metal films were determined bythe capacitor method 9 while the perturbation of the photoelectric properties of nickelribbons was studied by monitoring the photoelectric yield and the energy distribution ofthe phot oelectrons.10CAPACITOR MEASUREMENTSMETAL+ METAL sYsTEMs.-Xn order to determine volta potentials a potential Y is appliedbetween the two metals such that the electrical field in the gap is equal to zero.Underthese conditions the electrochemical potential of electrons in the two metals is the same.The method of ascertaining when the field is reduced to zero is to vary the gap betweenthe metals by vibrating one of them and to adjust V(to V*) such that the a.c. signal recordedis a minimum. Under these conditions -eoV* = #1,2, where 41,~ is the volta potentialbetween metals 1 and 2.but not on the reference electrode (2) and the voltage Vagain adjusted to give no a.c.sjgrtalthen the change A YO in the compensating voltage is equal to the change in the work functionof metal (1) due to gas adsorption.SEMICONDUCTOR+ METAL sYsmM.-The electrochemical potential & of an electron ina semiconducting oxide is the work necessary to transfer the electron from infinity to theinterior or” the oxide; this is made up of two term: (a) the chemical potential Ek whichdepends on the chemical nature of the seniconductor, and (b) the effect of surface dipolesor charge in the surface region of the semiconductor. This is the electrostatic contributionto the electrochemical potential and is equal to e&’, $’ being the inner potential term andeo the electronic charge.Therefore,METAL AND CHEivIISORBED GAS-IMETAL SYSTEM.-lf a gas iS chemisorbed On metal (1)EF = Ef. + eo($ + x), (1)where 4’ has been divided into the outer potential due to the free charge as the semiconductorsurface and x a surface dipole term; it can be shown that the electrochemical potentialcorresponds to the F e d level for electrons. The choice of reference level for definingpotentials is arbitrary; thus while EF is defined as above, within the oxide it is more usualto define electrical potentials with respect to the band structure of the senliconductingoxide, e.g., by taking the mid-gap as reference level. The potential difference between apoint B in the interior of the oxide and a point at the surface is equal to the inner potentialdifference (4~-C$s) which in units of KT is usually denoted byThe term Y is sometimes referred to as the “ surface potential ” or barrier height and isthe extent to which the bands are bent at the surface with respect to the bulk.Considernow the work function of an oxide semiconductor where the bands are flat, i.e., where thereis no electrostatic potential term due to some portions of the semiconductor being morecharged than others. In this case, eqn. (3) holds :-6; = Ek+eoX, (3)where 4; is the electronic work fuiiction of the neutral semiconductor. If, however, thebands are not flat, e.g., due to the presence of adsorbed species (or surface states) then thework function of the oxide q5e is given by eqn.(4) :where ($B-&) is the potential shift in the space charge region and A1 is the change in thedipole term1 64 REACTIVITY OF OXIDIZED NICKEL SURFACESOXIDIZED METAL+METAL SYsTEM.-If the change in the work function of a metal duringoxidation is being studied then not only must the oxide-gas interface be considered but alsothe metal-oxide interface. Two distinct cases exist : (i) when the oxide thickness is greaterthan the Debye length of the oxide then the work function of the oxide is monitored;(ii) if the oxide thickness is less than the Debye length then one measures the work functionof the metal substrate as influenced by the thin oxide film. The generation of a metal-oxide interface can therefore lead to difficulties in the interpretation of work functionchanges, one consequence being the occurrence of a space charge at the metal-oxideinterface.PHOTOEMISSIONENERGY DISTRIBUTION OF PHOTOELECTRONSInformation about the energy levels of the elemental semiconductors germanium andsilicon has been obtained from studies of the external photo-effect.If light of a givenenergy hv impinges on a surface, electrons will escape with a range of energies the distribu-tion of which can be determined by applying a negative potential to the collector until thephotocurrent ceases. This stopping-potential is the same for all metallic cathodes butis different for semiconductors. For metals, the Fermi level represents the highest filledstates at 0°K; at any other temperature, T electrons will have energies of the order ofkT above the Fermi level.Since we are concerned with photoemission from surfaces atroom temperature the small kT effect can be neglected. For a semiconductor the photo-electrons will be considered to originate from the valence band so that the highest filledlevels are 6 eV below the Fermi level. The stopping potentials VO, Vh for electron emissionfrom metals and semiconductors, are given by eqn. (5) and (6) respectively, where 4c isthe work function of the collector.so thateVo = 4c- hv,A shift in VO (to Vh) during the interaction of oxygen with a metal surface therefore impliesa change in surface structure probably due to perturbation of the energy levels by theformation of an oxide.However, the data of Allen and Gobeli 11 suggest that for a thinoxide film ( ~ 2 0 A ) on silicon the mean free path of the electrons within the oxide, andemitted from the silicon, is only about 7 A. If degradation of the energy of photoelectronsfrom the Ni+ 0 2 system is appreciable by inelastic scattering, then any quantitative inter-pretation of 6 is difficult but the conclusion that a surface oxide is formed would still bevalid since chemisorbed oxygen is unlikely to change VO. In Auger emission, a similarproblem exists; Hagstrum,4 however, believes that for oxygen on nickel the change inthe electron energy distribution is more likely to be the consequence of a change in thesurface band structure rather than degradation of energy from inelastic scattering processes.Energy distribution data also enables the Fermi level to be determined from the ob-served potential V, at which the photocurrent is saturated sincewhere 6 is the work function of the emitter.PHOTOTHRESHOLDFor a metal or a metal+ chemisorbed gas the photoelectric threshold gives the positionof the Fermi level in each case; with a semiconductor the threshold energy is the lowestphoton energy able to excite electrons from filled levels over the surface and, assumingno emission from surface states, corresponds to the top of the valence band.With aflat-band semiconductor no problem arises from the volume effect in photoemission butif a space charge exists within the semiconductor then the valence band edge will havM.W. ROBERTS AND B. R. WELLS 165different energies at different depths so that photothreshold value depends on the electronescape depth. The source region of photoelectrons depends on two factors; first, theattenuation of the incident light by the solid and secondly the range of the excited electrons.It is usually assumed that the attenuation of the light depends exponentially on depthwhile the range of the excited electron depends on the dominant electron scattering mechan-ism. These two factors can give rise to escape depths ranging from -2OA in silicon to-2OOw in Cs3Sb so that again the quantitative significance of 6 and threshold must beconsidered carefully. But if photoemission is dominated by electrons emitted from aparticular depth then interpretation is simplified.Non-equilibrium conditions can exist when a seniconductor surface is illuminatedsince light of the appropriate energy can create excess hole-electron pairs.Under theseconditions one refers to a quasi-Fermi level for the system. Two ways in which the workfunction of an oxide can change under illumination are (a) photodesorption of oxygenresulting from the diffusion of holes to the surface and (b) perturbation of the space-chargeregion. The latter will be reflected by a change AY in the surface barrier height Y (eqn.(2)). If steady illumination is maintained a state of pseudo-equilibrium will be attainedwith an associated quasi-Fermi level ; under these conditions A Y is referred to as the surfacevoltage.This should be measurable by the capacitor method as a change in the voltapotential between the oxide and the reference electrode. It was this approach that ledBrattain and Bardeen 12 to invoke a space charge region in the surface of germanium andby analogy we conclude that any perturbation by light of the surface properties of theNif02 system would also indicate a surface layer that had the essential characteristics ofa semiconducting oxide. The surface oxide must, however, be at least a Debye lengththick before it can be expected to have the properties of the bulk oxide.RESULT§ AND DISCUSSIONTHE Ni+02 SYSTEMAdsorption measurements 139 9 with nickel films have indicated that at 23" andan oxygen pressure of 10-2mm the surface coverage corresponds to more thanone oxygen adatom per nickel atom.This implies that penetration of the latticehas occurred but to establish that an oxide had formed a knowledge of some otherparameter, sensitive to the atomic nature of the surface, was required. In fig. 1are shown work function changes A$ during oxygen interaction with nickel filmsand also the variation of the photoelectric yield from nickel ribbons during oxygeninteraction at low temperature. Thus, although the coverage at - 195°C is high(about one oxygen adatom per surface nickel) and no oxygen is desorbed on warmingto 23" the electrical characteristics (photoelectric and capacitor, fig. 1) are closeto those of the original clean metal, whether in the form of a film or polycrystallineribbon.There has therefore occurred substantial surface rearrangement inducedthermally but without the generation of a truly clean surface. Some evidence forthe latter was obtained by recooling the film to -195" and again adsorbing oxygen(fig. la). This resulted in only between one-sixth and one-quarter of the initial volumeof oxygen being taken up but with a large increase in work function.That at least two processes occur during oxygen interaction with nickel filmsat 23" was first indicated by capacitor data 14 ; the following is a summary of thepresent position.(i) The photoelectric yield-oxygen exposure pattern at 23" has a distinct minimumafter an exposure of 1-2 x 10-6 mm min and a subsequent maximum at 3-4 x 10-6mm min before eventually decreasing again at longer exposures.10 (ii) The processresponsible for the recovery in the yield at 23" is activated since if the ribbon ismaintained at about - 160" (fig.lb) then only a decrease in yield is observed duringoxygen interaction. This is compatible with positive decays in work function 9166 REACTIVITY OF OXIDIZED NICKEL SURFACESduring adsorption studies of oxygen on nickel films at 23" but not at -195".Threshold data 15 indicate that the work function increase at - 160" was >1.0 eV(cp. 1-2 eV for 0 2 on sintered nickel films 16 at - 195"). (iii) After an oxygen ex-posure of >lO-4 mm min the value of 6 (eqn. (7)) is between 0-3 and 0-4 eV and thisrefers to the energy of the highest filled states with respect to the Fermi level pro-vided our assumption that a sufficient number of maximum energy electrons arecollected.(iv) Work function changes estimated from the shifts in I/, (eqn. (8))TIME Cmin31.1FIG. l.--(a) Work function change A+ on adsorbing oxygen on a nickel film at -195" 0, followedby warming to 23" in vacuo (D, and re-adsorbing oxygen at -195".(b) Variation of photocurrent during oxygen interaction with a polycrystailine nickel ribbonmaintained at about -160", 0 ; followed by warming in vnrim (p<10-7mm) to 23", @. h :2150 A.after an oxygen exposure of - 10-4 rnm min are between 0.2 and 0.3 eV. Thesechanges are substantially smaller than those observed by the capacitor methodwith both films (-0.8 eV) and ribbons ( ~ 0 . 6 eV). (v) When the interaction ofoxygen was studied 16 with nickel films at 160" (the oxygen being added as a numberof doses) an initial increase in the work fmction (capacitor method) of 0.3 eV wasfollowed first by a decrease to a value of 0.4 eV below the clean metal value and thenby an increase to -0.3 eV above the metal value (cp.also data at 137"C, fig. 2).(vi) Heating a chemisorbed oxygen layer on nickel ( A ~ E - 1.4 eV> from - 195to 23" in vawo (fig. la) results in a value of q 5 ~ 0 (clean metal value), and subsequentheating 16 to 150" gives a A 4 value of - + 1-0 eV. These results are interpreted 16 interms of the following scheme :Ni,G/> Ni ++02-+Ni-06- '11.Ni--Od- represents chemisorbed oxygen, Ni,O the low work function oxide formedby incorporation of oxygen and Ni,O a high work function oxide which is notdistinguishable from (Ni,O -I- 06- chemi).The forinula Ni,O(nz> 1) describes theNi + 0 2 system when substantial incorporation has occurred resulting in a workfunction lower than the initial metal value; it is therefore analogous to Ni3O oM. W. ROBERTS AND B. R. WELLS 167Ni70 proposed by Park and Farnsworth 17 (fig. 2b) but a more detailed comparisoiiis not possible in view of the polycrystalline nature of our film. Therefore bycontrolling the temperature and oxygen pressure the various stages Ni-06-, Ni,Oand NimO could be isolated 16; also at a given temperature and oxygen pressure,different work function changes will be observed depending on how the reactionwas studied.This was attributed 18 to the existence of non-stoichiometry in thesurface region, possibly a consequence of chemisorbed oxygen acting as an anchor 16thereby restricting the mobility of a surface nickel atom and preventing the in-corporation of oxygen by a rotational mechanism. There is direct experimentalevidence for the influence of gas impingement rate on the extent of oxygen in-corporation, Bloomer and Cox 19 having shown that for the interaction of oxygenwith barium films the uptake increasedstant for pressures >,lo+ mm.at pressures below lO-7mm but was con-0 . 50 I O .O. O O. 9.FIG. 2.-(a) Change in the work function of a nickel frlm at 137" while the oxygen pressure wasincreased from 4 10-6 to 10-3 mm, 0. The oxygen was added as a number of small doses f .Subsequently hydrogen was adsorbed to -10-2 mm at 23", B.(b) Structure of Ni30 proposed by Park and Farnsworth 17 for oxidized nickel single crystal sur-faces at room temperature.(e, Ni ; 0, 0).Chemical reactivity studies 16 clarify our views on the nature of nickel surfaceswhich have incorporated oxygen. Although the electrical characteristics can beregenerated at 23" and are identical to those of the metal (fig. l), the surface isessentially inactive to carbon monoxide adsorption, the activity being more like thatexpected of oxidized nickel (table 1). Similarly, a surface (NimO) which has beenregenerated at 150" is inactive in carbon monoxide adsorption whereas hydrogenis adsorbed extensively at 22" ; a regenerated surface, in contrast to the clean metal,is therefore very selective in its adsorption behaviour.This may be a generalphenomenon of significance in catalysis since electronegative species other thanoxygen (e.g., S, F) could also give rise to surfaces with specific adsorption properties.Studies by Burshtein 20 have indicated that low work function surfaces are formedfrom the interaction of fluorine with iron. The high activity of regenerated surfacesTABLE I.-REACTLVITY OF SURFACES AT 22" TO H2 AND COsurface eH2 420Ni, 0 2 to 10-3 inm at - 195", warmed to 22" in uucuo 0.64 0.130-69 0.10- N 0.07Ni, 0 2 to 10-3 n m at - 195", heated to 150" in vacuo (NinZO)Ni, 0 2 to 10-2 m, 22" wino) 0-1-0.2 GO-10Ni, 0 2 to 10 mm at 170", evacuated (NiO)Ni, 0 2 to 1.5 mm at 170", evacuated (NiO)-0.01 168 REACTIVITY OF OXIDIZED NICKEL SURFACESto hydrogen (table 1) also emphasizes its limited use as an adsorbate for studyingthe nature of surfaces.Further information about the nature of nickel surfaces exposed to oxygenat 22" was obtained from photoelectric energy distribution data obtained withpolycrystalline ribbons ; in particular, for an exposure of - 10-4 mm min no photo-electrons characteristic of emission from nickel could be detected (- 10-14 A).Thus,provided the photoelectrons do not loose appreciable energy by inelastic scattering,these results suggest that the electron escape depth lies within the surface oxideand/or that any photoelectrons originating in the underlying nickel are absorbedby this oxide.At 22" and 10-2mm, there are present approximately 2 oxygenatoms per surface nickel atom; on the basis of Ni3O being formed, our oxide couldextend about 14A below the surface. The stopping-power of such thin oxidefilms (- 10 A) for photoelectrons originating from the underlying metal is sur-prising, but Allen and Gobeli 11 came to a similar conclusion where a surface oxide(-20 A) on silicon absorbed 94 % of the photoelectrons emitted from the under-lying silicon.SINTERINGCatalytic activity has frequently been found to be dependent on the temperatureat which metal films have been pre-heated. This heat treatment (sintering) mayin some cases alter the product distribution; the latter has usually been attributedto the onset of surface mobility of metal atoms effecting a change in the crystallo-graphic nature and therefore reactivity of the surface.The chemical consequencesof sintering under relatively " poor " vacuum conditions (e.g., a partial pressure of10-7 mm of 0 2 ) must, however, not be overlooked and fig. 26 illustrates the effectof sintering nickel at 137" on the work function when oxygen is present in the gasphase. In regions A and B the oxygen pressure is 4 10-6 mm but depending on thetime of exposure, oxygen either remains on the surface as Q6- or for longer ex-posures a low work function oxide Ni,O (m> 1) is formed (region B). Subsequentexposure to oxygen leads to this " incorporated structure " chemisorbing oxygen(C) and the total uptake would correspond to 3-4 oxygen '' monolayers '' at 18-3 mm.This surface is nevertheless very reactive to hydrogen at 23", at 10-2mm since thework function decreased by - 1.1 eV (fig. 2).This decrease compares with 0-3 eVfor a film oxidized at 10 mm at 170". The hydrogen coverage in the latter casewas only a few percent (table 1) which is what would be expected of bulk NiO.Therefore, on the basis of hydrogen adsorption and associated work function data,nickel films do not exhibit the low reactivity expected of bulk NiQ unless thay havebeen extensively oxidized (P(O2) >I mm at - 170").Polycrystalline ribbons that have been exposed to oxygen at 23" (- 10-5 mm min)behave in an analogous manner 21 (table 2). On heating to -500' at 10-8 mm,TABLE 2.---INFLUENCE OF HEATING A POLYCRYSTALLTNE NICKEL RIBBON WHICH HAS BEENEXPOSED TO 0 2 AT 23" TO 500" AT 10-8 mmexpt. threshold of oxygen relative threshold after heat i/io afterclean surface exposure photocurrent 0 2 exposure treatment heating(i/io) after0 2 exposure(eV) (mm min) (eV> (min)4 5-07 1 .4 ~ 10-5 0.36 5.23 3 1 *958 5-05 >Ix 10-2 0.21 > 5.3 0.5 0.391.0 0.692.0 1.135.0 1-55threshold afterheating(eV)4.555.1 14-914.724-6M. W. ROBERTS AND €3. R. WELLS 169the relative photocurrent (i/$ increased by nearly a factor of 10, i.e., to nearly twicethat observed with the clean metal (io). There was associated with the increasein yield a substantial decrease in the threshold energy. The possible role of theselow work function surfaces in adsorption and catalysis has already been illustratedby the specific adsorption characteristics of Ni,O.OXIDIZED NICKELConductivity measurements have provided evidence for electron exchangeprocesses during the adsorption of gases on oxides. However, care should be takenwith the detailed interpretation of such data. For example, if a gas adsorbed onthe surface of an oxide which has n-type bulk conductivity increases the conductivitythen it would be concluded that a net positive charge is present on the surface givingrise to an increase in electron density in the space charge region.In the absence ofany other evidence we can, however, not be certain that the surface region of thestarting material was indeed n-type.If it had been p-type then the increase inconductivity would reflect an effective negative surface charge.Similarly, the interpretation of changes in chemical reactivity effected by bulkdoping of oxides in terms of the expected shift in the Fermi level of the oxide mustalso be considered cautiously. In the flat-band semiconductor case (eqn. (3)),with uniform doping little alteration of x will occur so that A$:, the change in thework function will be equal to AEp the change in the Fermi level. On the otherhand, in the presence of surface states and a space-charge (eqn. (4)) the work functioncould be virtually insensitive to bulk doping. Bardeen first pointed out that surfacestates of sufficient density could control the surface potentials; only recently hasany diagnostic experimental work been attempted when Allen and Gobeli 22 showedthat bulk doping Si from n to p-type only produced a shift in the Fermi level of0.2 eV.This small shift compares with a change of - 1.2 eV in the bulk levels.With these facts in mind, we investigated the perturbation of the work function ofNiO by both adsorbed CO and 0 2 and also during the catalytic oxidation of COusing a stoichiometric mixture. The latter would indicate whether the Fermi levelchanged with any variation in surface coverage that might occur during the courseof the reaction.OXIDATION OF NICKEL AT 170"Two distinct types of behaviour were observed during the oxidation of nickelfilms at 170" (fig. 3). The first type (a) corresponds to an instantaneous increase in- 0.5a2,- 0-a0101 I blFIG. 3.-The oxidation of nickel films at 170" and 1-5 mi; P = evacuation to -10-6 mm.work function of about 0.7 eV followed by a slow decrease associated with oxidegrowth.The second type (b) differs in that the fast increase is followed by an almos170 REACTIVITY OF OXIDIZED NICKEL SURFACESequally fast decrease. We interpret these data as reflecting the initial formationof an essentially metastable oxide which under conditions not yet defined can revertto an oxide of much lower work function. There is possibly a link here with theability of nickel to form a low work function oxide at elevated temperatures (fig. 2a).Such a large decrease is apparently unlikely to arise from a redistribution of thespace charge within the oxide.23 Evacuating to 10-6 mm at 170" resulted in thework function decreasing by between 0.05 and 0.1 eV.ADSORPTION OF OXYGEN AND CARBON MONOXIDE ON OXIDIZED NICKEL.FILMS WHICH HAVE ADSORBED 0 2 TO 10-2mm AT 23"CARBON MONOXIDE.-A film which adsorbed 80 x 10-3 cm3 of 0 2 at 23" adsorbedat the same temperature 5 x 10-3 cm3 of CO at 10-1 mm.Although the CO coverageis small there was an associated decrease in work function of about 0.5 eV. Onincreasing the CO pressure to 1.5 mm there was a further decrease of 0-2 eV. Similardata were obtained in other similar experiments.FILMS OXIDIZED AT 1-5mm AT 170"OxYGEN.--The adsorption of oxygen (p"1.5mm) at -195" caused the workfunction to first increase (fig, 4) and then decrease to a value about 0-25 eV belowthe oxide value.On heating to 23" the work function increased (correspondingI I04 0Time (min)FIG. 4.-Adsorption of oxygen on oxidized nickel at - 195" and 1.5 mm followed by warming to 23".to the initial fast decrease at - 195") and then remained about 0.1 eV above theinitial oxide value. There are therefore two distinct states of adsorbed oxygen,a molecular low energy state which can form only at low temperature and a morestrongly bound state (06-) which is stable at 23"; the latter increased the workfunction by about 0.1 eV. These results are in general agreement with the adsorptionand conductivity data of Kuchynka and Klier 24 who used " clean " NiO.CARBON MoNox1uE.-At -80" and a pressure of 10-1 mm, the decrease in workfunction was 0.35 eV but on increasing the pressure to 1.5 mm a further decreaseof 0.1 5 eV occurred : the CO uptake at 10-1 mm was small (-3 x 10-3 cm3) anM.W. ROBERTS AND B. R. WELLS 171corresponded to about 10 % of the krypton monolayer value. At 23" and 10-1 mm,the uptake was appreciably less but the work function remained approximatelythe same. Analysis of the gas phase indicated that COa was desorbed even at-80" (a few % of a monolayer). At 23" the desorption of COz could be extensivecorresponding to as much as would be expected from a monolayer of chemisorbedoxygen, but the participztion 25 of lattice oxygen in the CO extraction process (eqn.(9)) excludes this simple picture :At 170" and a carbon monoxide pressure of 1.5 mm, the work function decreasedinstantaneously by about 0.85 eV (fig. 5a) ; this was followed by a slow smallincrease.Desorption of CQ2 was appreciable at this temperature but correspondedto no more than about a '' monolayer " of oxygen being involved. On evacuationO2-+C0(g)+Ni2+ = CO,(g)+Ni"+ UO2- (9)- -B e d+I0[a1FIG. 5 4 a ) Change in work function on adsorbing CO at 170" and 1.5mm on oxidized nickelfollowed by cooling to 23", reheating to 170" and evacuation (P), A; subsequently 0 2 was ad-sorbed, @.(b) Influence of hydrogen adsorption, e, on the work function of oxidized nickel followed byreadsorbing oxygen, (8, and re-admitting hydrogen, ; P = evacuation.and admitting oxygen to a pressure of 1-5 mm, the work function increased instant-aneously by about 0.7 eV, i.e., about 80 % of the initial decrease in work functionwas nullified.The latter was essentially an irreversible process since evacuationto 10-6 mm at 170" caused the work function to decrease by only 0.15 eV. Thereis also evidence (fig. 5a) for a temperature-dependent reversible adsorption processinvolving carbon monoxide; the effect is small (-0.1 eV) and gives rise to anincrease in work function on cooling from 170 to 23", or a decrease on evacuationat 170". The studies of Klier and Jiratova25 enabled four states of adsorbed COto be recognized, two involving the CQ molecule itself and two C02 (ads) species;adsorbed oxygen substantially decreased the concentration of CO (ads). Fromthe results shown in fig.5d we can correlate our CQdf with the CQ (ads) states M andD (nomenclature of M and J). The generation of nickel (eqn. (9)) may be associatedwith Cod+ the coverage of which therefore increases with temperature giving riseto successively lower work function values (0.45, 0.7 and 0.85 eV at -80, 23 and170" respectively at 1.5 rnm pressure of CQ).The dipole moment M of adsorbed species on metals is usually related to thechange in work function A# by the relation A# = 271nM/~. If we use this model,then the dipole moment estimated for the adsorption of CO on oxidized nickelfilms at 23" and 10-1 mm is about 30-40 D (n-5 x 1013 cm-2, A4-O-4 eV, E = 10)172 REACTIVITY OF OXIDIZED NICKEL SURFACESThe above relationship is therefore not valid and this is likely to be a consequenceof the substrate being a semiconducting oxide so that the work function changeA$ is made up of two terms (eqn.(lo)),where Ax is the change in the surface dipole and A& the change in the surfacepotential (space charge region). The latter is the major contribution since theabove calculation is based on all the adsorbed CO giving rise to Cod+, whereas alarge fraction is likely to be utilized in forming ((202) ads species. Although theCO (ads) coverage at 170" and 1.5 mm on our " NiO " is not known it is likely tobe small (cp. Klier and Jiratova) so that S(A4>=0.8 eV is again mainly due to thespace charge term arising from CO (ads). The latter is almost completely annulledby adsorbing oxygen (fig.5a) which agrees with Klier's observation that pre-adsorbedoxygen decreased CO (ads) substantially.Similar space charge contributions to A 4 are also likely to apply for the adsorp-tion of H2 and CO on a nickel surface exposed to oxygen at 23" and 10-2mm(6A4 - 1.8 and 0-5 eV respectively), since there is overwhelming evidence that thesurface has all the essential features of an oxide. The oxide in this case is probablyno more than 10-20A (possibly less) so that the usual theory based on a uniformspace charge probably does not hold and Bardeen26 has suggested that in suchsystems one must consider the surface species as individually screened. Similarconsiderations probably apply to the re-adsorption of oxygen at - 195" after thermalregeneration at 23" (fig.la).Work function changes with CO are relevant to the mechanism of reductionof the oxide ; for comparison, changes observed 16 with hydrogen at 23 and 170"are also shown in fig. 5b. A feature of the hydrogen data is that after the initialdecrease there occurred an almost equally fast increase. The latter was observedif the temperature was >loo" and since the desorption of water is extensive onlyabove 120" the increase in work function was attributed to this process. The cyclecould be repeated on re-adsorbing oxygen and then hydrogen. Since with CO at170" there was neither an appreciable increase in work function nor extensivedesorption of C02 (equivalent to - 1 layer) it is likely that the slow step in the reduc-tion is desorption of a surface complex; decomposition of the COj complex ofDell and Stone 27 occurs only above 300".A 4 = Ax+ A$s, (10)CATALYTIC OXIDATION OF CARBON MONOXIDEThe work function change observed during the oxidation of carbon monoxideusing a stoichiometric mixture of CO and 0 2 and oxidized nickel as a catalyst willdepend on the rates of various processes. The change might depend on the rateof adsorption of CO (donor molecule) or 0 2 (an acceptor), the rate of conversionof adsorbed CO into a surface complex (CO3), or its desorption as CU2.In fig. 6is shown the change in work function during the oxidation of CO at 23" using threeCO+ 0 2 mixtures (2 : 1) at a total pressure of 5 x 10-2 mm followed by one highpressure mixture (1.5 mm) ; the work function remains essentially constant duringcatalysis. Small changes were observed (-0.1 eV) during the initial and finalstages of conversion of any one mixture, but these are probably due to oxygen ad-sorption at the initial stages and carbon monoxide adsorption at the final stages ofconversion (mixtures 1-3); with the 4th mixture a small decrease in work functionduring the initial stage of catalysis was followed by a slow increase. After thesecatalytic studies the system was evacuated, the work function increased by about0.1 eV (Cod+ removed), and carbon monoxide admitted to the surface at a pressurM.W. ROBERTS AND B. R. WELLS 173of 1.5 mm. The work function decreased instantaneously by 0.5 eV and then moreslowly by a further 0.2 eV.Since oxygen alone would have increased the workfunction by about 0.1 eV the presence of gas phase oxygen causes a new steady stateat the surface which is different from that with only CQ (g). We conclude from theconstancy of the work function during catalysis (cp. during CO adsorption) thatc 3Timu (rnin)FIG. 6.-Work function of oxidized nickel film, v, during the catalytic oxidation of CO at 23"using four stoichiometric mixtures (2 CO : 1 0 2 ) . Adsorption of CO, A ; P = evacuation.Ic0 * LY > CI-uo x0 2 0 4 0 bOT i m e (min)FIG. 7.-Catalytic oxidation of five stoichiometric mixtures (C0+02) by oxidized nickel film at23" ; C02 admitted (10-1 mni) after 4th mixture.the adsorption of oxygen is a faster process than that of carbon monoxide; more-over, this oxygen effectivity poisons the surface for adsorbing COa+ so that no space-charge is generated at the oxide-gas interfqce. This is essentially the same con-clusion that we came to from adsorption studies of the individual gases CO and 0 2at 170" (fig.5a). These results also provide direct experimental evidence that theadsorbed oxygen which prevents the formation of C06+, and likely to be involvedin the catalytic reaction, is a weakly adsorbed form since it is removed by evacuationfor a few minutes at 23" (fig. 6)I74 REACTIVITY OF OXIDIZED NICKEL SURFACESThe activity of these oxidized films (Po,>l mm, 170", 1 h) for CO oxidationat 23" was studied and fig. 7 shows the percentage conversion-time relationshipfor five stoichionietric mixtures.The CQ was removed from the gas phase by aliquid-introgen trap. The activity decreased by about 10 % with each mixturebut the addition of CO;! to a pressure of 10-1 min after the 4th mixture did notaccelerate the decrease in activity. The activation energy was about 2 lical mole-1(cp. Dell and Stone 27) and the rate of CO2 formation by our oxidized films at 23"and a total pressure of 1-5 mm was -4.5 x 10-2 ~ 1 ~ 1 3 min-1 m-2. This compareswith 4 x 10-4 cm3 min-1 m-2 calculated from the data of Dell and Stone at 20" anda total pressure of 2-8 mrn while a typical rate observed with Teichner's 28 activecatalyst was 10-2 mm min-1 m-2 at a pressure of 1.7 mm. As was also shown byTeichner the initial rate of CO2 formation was not proportional to the initial totalpressure PO, e.g., in two different experiments at Po values of -10-2 and 1 mmthe rates were 6.4 x 10-3 and 3 x 10-2 cm3 min-1 m-2 respectively.This suggeststhat the concentration of adsorbed species participating iii catalysis at 22" is virtuallyunaltered above 10-2 mm.Since the admission of CQ2(g) after mixture 4, fig. 7, did not accelerate thede-activation of the surface it is unlikely that poisoning is due to the CO;! formedin the catalytic reaction. A more likely explanation is the accumulation of surfacespecies which can only be formed from adsorbed oxygen and carbon monoxide andnot from C02(g) probably the adsorbed state L suggested by Klier and Jiratova.1 Trapnell, Proc. Roy. SOC. A , 1953, 218, 566. Dowden, J. Chem. Soc., 1950, 242.2 Ehrlich, Adu. Catalysis (Academic Press Inc.), 1963, 14, 255.3 Gomer, Field Emission and Ionization (Oxford University Press), 1961.4 Hagstrum, Takeishi, Becker and Pretzer, Surface Sci., 1964, 2, 26.5 Farnsworth, A h . Catalysis (Academic Press Inc.) 1964, 15. MacRae, Surface Sci., 1964, 1,6 Eley, Coloquio sobre Quitnica Fisica de procesos en Srperfcies Sdidas (25th Aniversario del7 Green, in Modern Aspects of Electrochemistry, ed. Bockris, 1959, 2 (Butterworths).8 Hummer, in The Electrochemistry of Semicoizductors, ed. Holmes (Academic Press, 1962).0 Quinn and Roberts, Trans. Faraday Soc., 1964, 60, 899.10 Quinn and Roberts, Trans. Faraday Soc., 1965, 61, 1775.11 Allen and Gobeli, J. Appl. Physics, 1964, 35, 597.12 Brattain and Bardeen, Bell System Tech. J., 1953, 32, 1.13 Brennan, Hayward and Trapnell, Proc. Roy. Soc. A , 1960, 256, 81.14 Quinn and Roberts, Proc. Chem. Soc., 1962, 246.15 Quinn and Roberts, unpublished data.16 Roberts and Wells, submitted to the Faraday Society.17 Park and Farnsworth, J. Appl. Physics, 1964, 35, 2220.18 Quinn and Roberts, Nature, 1963, 200, 648.19 Bloomer and Cox, Brit. J. Appl. Physics, 1965, 16, 1331.20 Burshtein and Shurmovskaya, Surface Sci., 1964, 2, 210.21 McKee and Roberts, unpublished data.22 Allen and Gobeli, Physic. Rev., 1962, 127, 150.23 Fromhold, private communication.24 Kuchynka and Klier, Colf. Czech. Chem. Comm., 1963, 28, 148.25 Klier and Jiratova, Proc. 3rd Int. Congr. Catalysis (Amsterdam, 1964).26 Barden, Surface Sci., 1964, 2, 381.27 Dell and Stone, Trans. Faradcly Soc., 1954, 50, 501.28 Coue, Gravelle, Raue, Rue and Teichner, Proc. 3rd Int. Congr. Catalysis (Amsterdam, 1964).29 Wolkenstein, Ado. Catalysis, (Academic Press Tnc.), 1960, 12, 189.319.C.S.I.C., October, 1964)

ISSN:0366-9033    

DOI:10.1039/DF9664100162

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

GENERAL DISCUSSIU NProf. C. H. Amberg (Carleton University, Ottawa) said: Although I find Peri'sarguments convincing as far as the interaction of CO with his supported Ni catalystsis concerned, I should like to urge some caution in using CO adsorption as a criterionfor surface " acidity ". In Amsterdam, Seanor and I 1 reported the occurrence ofhigh-frequency CO bands on ZnQ catalysts. It was observed in that instance that thefrequency increased (to a final value of 2212 cm-1) with increasing pre-oxidation of thesurface. In the extreme case it is not to be expected that positive zinc ions remainedexposed in the surface. All the CO surface species were weakly and reversibly adsorb-ed, is., they could be completely removed by briefly outgassing the system at roomtemperature.These facts taken in conjunction with absolute intensity measurementsof the bands in question led us to interpret the adsorption in terms of dipole forces, asmentioned by Peri, one of the possible configurations then being that of the oxygen(the positive end of the CQ dipole) pointing towards the surface.The complexity of the situation is underlined even further by the 2200 cm-1 bandreported earlier by Little and myself 2 for CO adsorbed on Cr203/Al203. Althoughthe same type of species as that described for ZnQ undoubtedly contributed to thisband, as judged from experiments at room temperature, an activated adsorption(300") considerably intensified the 2200 cm-1 band and produced a much more stronglyadsorbed surface species.The latter might conceivably be interpreted in terms ofPeri's model. It would indeed be interesting to repeat his HCl and NH3 pretreatmenton Cr203/A1203 in order to test how far one might go in making generalizations fromthe particular systems studied.Dr. N. D. Parkyns (Gas Council, London) said: Does Dr. Peri know the crystallitesize of his nickel on silica aerogel samples? Our own experience with nickel onalumina aerogels is that very high transparencies seem to be associated with very lownickel crystallite size. For example, alumina aerogel containing 10 % by weight ofnickel had a transmission of about 70 % at 2000cm-1. However, the presence ofnickel was absolutely undetectable by X-ray diffraction, indicating a crystallite sizeof less than 20A.A second point is the apparent absence of a nickel carbonyl gas-phase band at 2060 cm-1. Here again in our nickel/alumina aerogel samples, at COpressures above 3 torr this sharp band is the dominating one between 2000-2100 cm-1.It may be that the silica gel structure screens the nickel atonis from forming therelatively bulky carbonyl molecule. Alternatively, it may indicate that the nickelatoms are present in some stable non-metallic form.Dr. J. Erkelens and Dr. Th. J. Liefkens (Unileuer Res. Lab., Vlaardingen) said: Inrelation to the work done by Dr. Peri, I mention some of our results obtained with1-hexene adsorbed on a nickel-on-silica catalyst. On adsorption of 1-hexene both on ahydrogen-covered and on a bare surface four infra-red bands appeared in the 3 pregion at 2963, 2926, 2876 and 2857 cm-1.After introduction of hydrogen, mainlyan increase in the intensity of the CHz bands occurred, the intensity decreased afterevacuation of the hydrogen. After a second introduction of hydrogen, however, theintensity of the bands increased again. A similar phenomenon was observed forethylene, both on a hydrogen-covered and on a bare nickel surface.1 Amberg and Seanor, PPOC. 3rd Int. Congr. Catabsis (North Holland Publ. Co., Amsterdam,2 Little and Amberg, Can. J. Chem., 1962, 40, 1997.1965), p. 450.17176 GENERAL DISCUSSIONAs it was expected that a hydrogenation-dehydrogenation reaction would takeplace, we replaced hydrogen by deuterium. When deuterium was added to 1 -hexeneadsorbed on a hydrogen-covered surface, almost exclusively CH bands and hardly anyC--D bands appeared in the 4 p region.Bands due to C-I3 vibration appearedafter 16 h. On evacuation of the deuterium the intensity of all bands decreased.After the introduction of a second dose of deuterium, the bands due to C-D vibra-tions increased very strongly. When deuterium was added to 1-hexene adsorbed on abare nickel surface, the reaction with deuterium took place immediately.Dr. W. J. Thomas (Uniuersity CoZZege, Swansea) (communicated) : With referenceto the adsorption of NH3 on silica-supported nickel (previously reduced by Hz), thework of Peri 1 showed that not all of the NH3 was removed by evacuation for 1 h at300°C. The gas was retained mostly as adsorbed NH3 resembling that held on Lewisacid-sites on dry silica-alumina.2 Of some relevance to this work is an infra-red studyof the adsorption and oxidation of NH3 on silica, silica-supported platinum and ironoxide, details of which are to be published shortly.4The background spectra of compressed discs of Cabosil (silica produced by G.L.Cabot Inc.) and platinum-supported on Cabosil showed very similar features. Atroom temperature, introduction of NH3 into the evacuated cell containing the catalystsample immediately produced well-defined absorption bands at 3424, 3344,2958 anda shoulder at 3280 cm-1. Evacuation of the cell at room temperature for less than1 min removes the bands entirely, suggesting a very weak adsorption of NH3 on silicaand silica-supported platinum.Spectra were also obtained at 72, 90 and 130°C. Atemperature of 150°C completely removes the bands at 3424, 3280 and 2958 cm-1,but the band at 334.4 cm-1, which corresponds to a stretching frequency of gaseousammonia, remains. When the sample is cooled to room temperature the spectrumis identical to the original spectrum obtained before evacuation and heating.Such experiments suggest the when ammonia is heated to temperatures up to 150°Cin the presence of either silica or silica-supported platinum, no decomposition ofammonia occurs. The main features of the absorption band at 2958 cm-1 is verysimilar to one previously reported 5 in a study of the low-temperature adsorption ofNH3 on porous glass. This band was interpreted in terms of an OH vibration in ahydrogen bonded complex such as Si-OH- - -NH3.The bands at 3424 and 3280cm-1 must then be assigned to N-H stretching frequencies of the adsorbed species.It is of interest to speculate as to the mode of attachment of the species to thesurface since this may afford a clue to the mechanism of the oxidation of ammonia onsuch catalysts. If the ammonia molecules attach to the surface via a hydrogenatom, then one may expect the perturbed N-H vibrations to appear at lower fre-quencies than those assigned to N-H vibrations in the free gas. Such a displace-ment will depend upon the magnitude of the interaction between surface and substrate.Since the feature at 3424 cm-1 is higher than the mean value of v3NH3, ca. 3414 cm-1,the mode of attachment does not appear to be through a hydrogen atom.Attachmentvia the nitrogen atom through its lone pair of orbital electrons interacting with thefree surface hydroxyl sites would probably cause an electron redistribution around thecentral N atom sufficient to raise slightly the NH frequencies. Thus, a single adsorbedspecies can account for the observed spectra including the ill-defined shoulder at3280 cm-1.1 Peri, this Discussion.2 Eischens and Pliskin, Adv. Catalysis (Academic Press, N.Y.), 1958, 10, 1.3 Griffiths, Ph.D. Thesis (Univ. Wales, 1964).4 Hallam, Griffiths and Thomas, to be published.5 Yates, Sheppard and Angell, J. Chem. Physics, 1955, 23, 1980GENERAL DISCUSSION 177Dr. M. V. Mathieu (Lyon, France) said: Although alumina 1 and zeolites 2 presentaccording to their structure, different OH stretching band, silica is more stable; itsfree OH band is always near 3750 cm-3 (porous glass, aerogel, xerogel, Cabosil).The:half-width of this band is very small (10-15 cm-1). Therefore it seems probable thatthe band at 3620 cm-1 is caused by Ni-OH. After our observations, the pure nickel-hydroxide had its most intense OH band exactly at this frequency (another weak bandat 3580 cm-1). The preparation of the sample supports this assumption. SampleNS 1 and nickel-Cabosil are the only samples prepared without ammonia; also theconditions do not allow the formation of nickel hydroxide. The thermal stability ofthis band is very great (nickel hydroxide is completely decomposed at 350°C invacuum 3), but the Ni2+ dispersed in silica is very stable to reduction factors.Theband at 4305 cm-J could be assigned to a combination Y O H + ~ N ~ - ~ H , this latter beingnot very far from GSi-OH. Also, Ni2+ ions can chemisorb ammonia as a Lewisacid. Eyraud 4 made the same observation with the chemisorption of water on cupricions on alumina and found that two molecules of water were chemisorbed on onecupric ion.Peri's paper also calls the attention to a third point, that one must be very careful ininterpreting the infra-red spectra of adsorbed molecule. The species chemisorbed atmoderate temperature are often not the true active catalytic complex. For example,the adsorption of C02 on thoria is strongly bonded in a bidentate carbonate form (itneeds two adjacent sites) below 100°C, but above 150"C, in a monodentate carbonateform.Now, the CO chemisorbed on thoria can only be oxidized above 250°C in amonodentate carbonate which is the active complex. To support this assumption,this monodentate carbonate is stable up to 350°C and thoria is a catalyst above thistemperature.Prof. N. Sheppard (Norwich) said: In Cambridge, and more recently in Norwich,I and my colleagues (Dr. B. A. Morrow, Dr. J. W. Ward and Dr. M. Clark) havecarried out infi-a-red studies of the surface species obtained by the chemisorption ofethylene on nickel, as well as on palladium and platinum. Like Dr. Peri we havealso found that it is a difficult matter to reduce nickel oxide to the metal; thereforemuch of our work has been carried out by the reduction of nickel nitrate in a mannersimilar to that originally employed by Eischens and Pliskin.We have used presseddiscs of nickel nitrate in Cab-0-Sil with a relatively low percentage of the metalOur samples are therefore most closely related to Dr. Peri's sample NS-1 which hedemonstrates can be more easily reduced to nickel from the evidence of CO adsorptionstudies. Our spectra show a number of qualitative features in common with thosereported by Dr. Peri on this latter sample (but not illustrated in his paper) viz., arelatively weak spectrum after initial adsorption, a stronger spectrum after theaddition of hydrogen, and evidence for dimerization. Because our spectra have manyfeatures in common when initial adsorption is brought about on each of the threemetals Pt, Pd, and Ni we are confident that they do represent spectra of hydrocarbonsadsorbed on the metals.Our spectra have been studied over a rather wide tempera-ture range (- 80" to + 150°C) and this has helped greatly in their interpretation.The spectra from initial adsorption on platinum are the most straightforward andare most readily discussed first. Over the whole temperature range two dominantabsorption bands, A and B, occur at 2920 and 2880 cm-1 and we consider that this(< 10 %I.1 Peri, J. Physic. Chem., 1965, 69, 220.2 Rabo et al., this Discussion.3 Merlin and Teichner, Compt. rend., 1953, 236, 1892.4 Eyraud et al., Compt. rend., 1962,254, 688178 GENERAL DISCUSSIONCH2-CH2‘\ M simple spectrum can be interpreted with confidence in terms of an Msurface species (M = metal).A weak band C at 2795 cm-1 that also occurs in allcases is very probably the expected overtone of an angle deformation frequency ofMCI& groups. A further weak band at 3010 cm-1 of variable intensity which occursmore pronouncedly at -78°C may be caused by an ethylenic surface species, possibly/C H 4 HM but no bands in the vicinity of 3080 cm-1 were found which of type Mcould be assigned to C=CH2 groups. At room temperature addition of hydrogenleads to only a weak residual spectrum from surface species but much ethane appearsin the gas phase-at higher temperatures (90-1 50°C) hydrogenation leads to spectra ofsurface species of the n-butyl type and a minor proportion of butane (in addition toethane) in the gas phase. The higher temperatures therefore clearly lead to a limiteddegree of dimerization or polymerization on platinum, but up to 100°C the predomi-nant species contain two carbon atoms.The increases in intensity of the infra-redbands from the combined gas-phase and surface spectra on hydrogenation leads to the/ \CH2-CH2\M species which domin- conclusion that the associatively adsorbed Mates the infra-red spectrum is not necessarily the only one present. There may also beother surface species, perhaps occurring on different crystalline faces, which aredeficient of CH bonds. These may be of the ‘‘ surface carbide ” type.Spectra obtained on initial adsorption on palladium again have the dominant A,B, C pattern of bands at somewhat different frequencies from Pt (2910,2863 and 2775cm-1).These spectra and those obtained after hydrogenation are otherwise verysimilar to the spectra on platinum except that the details indicate that a similar degreeof dimerization occurs at 100-1 50°C lower temperatures on palladium.The spectra on our nickel on Cab-0-Sil samples reveal a much more complicatedsituation. A, B, C spectra on initial adsorption are found only at the lowest tempera-ture studied (- 78°C) and then only after a brief period of time. At room tempera-ture little of the associated surface species can be present and the initial spectrumobtained was at all temperatures very dependent on the time of adsorption.Hydro-genation at room temperature gives stronger bands due to surface alkyl groups whichalso indicate that extensive dimerization has occurred. We conclude that whereas theadsorption of ethylene on silica-supported platinum leads to fairly clear-cut spectrawhich are caused by associatively adsorbed surface species, the situation with nickelis much more complicated. In the latter case a variety of chernisorbed states areformed, and some of these involve dimerization to C4-containing species.Dr. A. N. Webb (Texaco Research Center, Beacon, New York) said: Dr. Peri’sfinding that prior adsorption of ammonia eliminates the CO species characterized bythe bands above 2100 cm-1 suggests that the adsorption sites involved are acidic innature.This property could account for the observed extensive polymerization ofolefin. The relatively low intensity of the metallic Ni-CO band near 2060 crn-1(fig. 5 and 6 ) indicates only about one exposed reduced nickel atom per 200 total. It ispossible that the static low-pressure reduction treatments were insufficient ; however,Sieg et al. 1 have shown by magnetic studies that nickel on Cabosil and silica gelsupports is easily reduced.1 Sieg, Constabaris and Linquist, presented at 145th Meeting Amer. Chein. Soc., New York,/(September, 1963) ; abstr. p. 19-1GENERAL DISCUSSION 179The aerogel preparations of Dr. Peri are different from conventional silica sup-ported samples in other respects. The latter do not yield the high frequency CObands, and do not produce extensive olefin polymerization even when deliberatelyoxidized.In the oxidized state they are inert toward hydrocarbon adsorption. Apossible explanation for the acidic properties and irreducibility of the aerogel sup-ported nickel is the formation of a mixed oxide phase. In any case the nickel-silicaaerogel system does not appear to be suitable for studying the spectra of moleculesadsorbed on metallic nickel.Dr. R. P. Eischens (Texaco Research Center, Beacon, New York) (communicated) : Inhis paper Dr. Peri indicates that he observes the polymerization which would be ex-pected for ethylene on an acidic surface. I agree with this portion of his interpretation.However, his attempts to compare these results with results observed on metallicnickel may be misleading.With his Cabosil-nickel samples it should be easy todistinguish between ethylene on metallic nickel and ethylene which has undergone anacid-catalyzed polymerization. The differences are independent of any specificinterpretation of the spectra. On metallic nickel the strongest bands are foundbelow 2900 cm-1 while in Peri’s spectra the strongest band is found near 2920 cm-1.When the nickel is free of hydrogen prior to ethylene chemisorption, the infra-redexperiments indicate an adsorbed species having a hydrogen/carbon ratio which isclose to one.1 This is in accord with the chemisorption experiments of Jenkins and13ideal.2 In conjunction with our infra-red work we observed the production of about0-7 molecules of gaseous ethane for each ethylene adsorbed on bare nickel.Hence,the acid catalyzed polymerization can not account for ethane formation and for thelow hydrogen/carbon ratios in the adsorbed species.The polymerization of ethylene on nickel which Dr. Pliskin and I observed shouldnot be confused with the acid-catalyzed polymerization. On a properly reduced barenickel surface the dissociative ethylene chemisorption produces fragments which tendto bond to each other. This type of polymerization is not detectable until the hydro-carbon is reconstituted by treatment with hydrogen. Prior to hydrogen treatment the2925 cm-1 band, which is characteristic of methylene groups in paraffinic hydrocarbons,is relatively weak. After hydrogen treatment the 2925 cm-1 band is strong relativeto the methyl band at 296Ocm-1. This indicates a high CH2/CH, ratio which weinterpret as evidence for polymerization. It is apparefit that this pattern of spectra1changes is not encountered in Peri’s experiments.Dr.J. B. Peri (Amer. Oil Co., Whiting, Indiana) said : In reply to Mathieu, theevidence does not exclude assignment of the 3620 cm-1 band to OH groups heldby nickel ions. On the other hand, the evidence favouring such assignment is hardlycompelling. The fact that reasonably pure silicas always show an isolated OHband near 3750 em-1 does not establish that other frequencies are not possiblefrom Si-OH groups on surfaces containing many foreign ions. Equally, thereis 110 reason to expect the same frequency to be exhibited by OH groups in nickelhydroxide and those held by Ni ions in a silica surface.The small half-widths ofboth the bands could in fact be considered as weak support for assignment to Si-OH.I find it easier to reconcile the high stability of these OH groups with attachmentto silicon rather than nickel ions. Further evidence is needed.I agree with Amberg that more evidence is needed before concluding that CQbands near 2200 cm-1 always reflect weak attachment to metal ions. It does seemstrange, however, that such bands should in some cases reflect attachment to oxideions and in other cases attachment to metal ions. Perhaps Zn2f ions are exposed1 Eischens and Pliskin, Adv. Catalysis, 1958, 10, 2.2 Jenkins and Rideal, J . Chern.Soc., 1955,2490180 GENERAL DISCUSSIONon an oxidized ZnO surface. High-temperature treatment may expose metal ionson an oxide surface as a result of condensation and removal of surface hydroxylgroups. Possibly this type of process can explain Amberg’s observations.In reply to Sheppard, possibly spectra of simple Cz hydrocarbons attached tometal surfaces have been obtained in some instances. On the other hand, assign-ments to such structures are hardly unambiguous and can be confidently acceptedonly if polymerization can be excluded. On nickel/silica the situation is evidentlycomplex and yau acknowledge some polymerization even on platinum/silica.Further study of the interaction of olefins with well-dried, but unreduced, metal/silica samples and of adsorption on unsupported metals is needed.In reply to Eischens, differences are to be expected in the spectra obtained forethylene on nickel/silica, depending on sample composition, pre-treatment, tem-perature, ethylene pressure, and time of contact.I do not insist that acid-catalyzedpolymerization is necessarily always important on nickel/silica. On the otherhand, the spectrum shown by him fur ethylene on bare nickel (fig. (3) in his ref. (1))hardly shows the strongest band to be below 2900cm-1. It appears at least ashigh as 29 10 cm-1. A band also appears near 3070 cm-1 which could show C=CH2groups. Eischens (his ref. (1)) shows that the carbon/hydrogen ratio is variableand that a surface carbide holding no hydrogens is obtained in some cases.Amixture of hydrogen-rich and carbidic surface species may exist, and the speciescausing the infra-red bands need not have the composition determined for theadsorbed phase generally. Ethane formation need not have involved the speciescausing the infra-red bands observed. Confirmation from experiments on unsup-ported nickel would help to clarify the situation.I definitely agree with Webb as to the unsuitability of my nickel/silica aerogelsfor studying spectia of molecules adsorbed on metallic nickel. This is a majorconclusion of my paper. I remain sceptical, however, that “ conventional ” silicasupported samples are always as inert before reduction as suggested. I find thatethylene does slowly polymerize on unreduced, dry nickel/Cabosil.In reply to Parkyns, I have no direct evidence on nickel crystallite size, but therelatively weak bands corresponding to CO on metallic nickel suggest that thecrystallites are fairly large.Spectra 6 and c in fig. 4, e.g., may contain some con-tribution near 2060 cm-1 from gaseous nickel carbonyl. The path length in my cellis, however, apparently considerably shorter than in his cell, which would helpexplain weaker bands for gaseous nickel carbonyl.Although both alumina 1 and zeolites 2 can, depending on their structure, exhibitvarious OH stretching bands, silica is more stable. Its free OH stretching band isalways found near 3750 cm-1 (porous glass, aerogel, Xerogel, Cabosil). The half-width of this band is small (10-15 cm-1). It therefore seems probable that the band at3620 cm-1 is caused by Ni-OH.According to our observations, pure nickelhydroxide had its most intense OH band at exactly this frequency (with another weakband at 3580 cm-1). The preparation of the sample supports this interpretation.Sample NS-1 and nickel/Cabosil are the only ones prepared without ammonia and theconditions of their preparation do not permit formation of nickel hydroxide. Thethermal stability of this band is great (nickel hydroxide is completely decomposed at350°C under vacuum 9, but Ni2f dispersed in silica is stable against reduction factors.The band at 4305 cm-1 could be assigned to a combination V O H + ~ N ~ - O H , the latterbeing not very far from 6~1-0~. Also Ni2+ ions can chemisorb ammonia as a Lewis1 Peri, J .Physic. Chem., 1965, 69, 220-231.2 Rabo et al., this Discussion.3 Merlin, Teichner, Compt. rend., 1953, 236, 1892GENERAL DISCUSSION 181acid. Eyraudl made the same observation for chemisorption of water on cupricions held on alumina and found that two molecules of water were chemisorbed on onecupric ion.Peri's paper calls attention to a third point-one must be careful with the infra-red spectra of adsorbed molecules. The species chemisorbed at moderate tem-peratures are often not the true active catalytic complexes, e.g., the adsorption ofCO2 on thoria occurs differently at different temperatures. Below lOO"C, C02 isstrongly bonded in a bidentate carbonate form (it needs two adjacent sites), above150°C, in a monodentate carbonate form.Now, the CO chemisorbed on thoriacan only be oxidized above 250'C, evidently in a monodentate carbonate whichis the active complex. To support this assumption, we have noted that this mono-dentate carbonate is stable up to 350°C and that thoria is a catalyst above thistemperature.Dr. H. Heyne and Prof. F. C. Tornpkins (Imperial CoZZege) (comnzunicated) : Wehave recently examined the spectra of CO chemisorbed on reduced and oxidized Ptsupported on Aerosil. The pellets were prepared by the Eischen technique andreduction was effected at 200°C for 4 h in 10 torr pressure of hydrogen, followed byoutgassing at 200°C for 2 h. Oxidation was carried out at 300°C with 20 torr oxygenfor 1 h. CO was chemisorbed at room temperature at 10 torr pressure.Three bands were observed with peak intensities at 2070-2080,2120 and 2170 cm-1,the latter band requiring about 1 h to develop its maximum intensity.The 2070cm-1 band was entirely similar to that found when the reduced Pt chemisorbed CO,e.g., the wave number of the band maxima and the extinction coefficients were thesame, as was its removal as gaseous C02 when 1 torr oxygen was added at roomtemperature; the band is thus assigned to CO chemisorbed on (a group of) metalsites. The 2120 cm-1 band was ascribed to chemisorption on Pt ions in the oxidelayer and designated Pt2+(CO) by analogy with the bands at 2146, 2135, and 2122cm-1 bands observed for the CO stretching frequency in the Pt (11) carbonyl halogenides[Pt(CQ)X2]2, where X = C1, Br, I respectively.The species Pt2+(CO) is stable tooxygen at room temperature, but is reduced by hydrogen, i.e., the band at 2070 cm-1is generated while the 2120 cm-1 disappears-a linear relationship is found betweenthe change of peak extinctions of these bands. The 2170 cm-1 band is only developedat higher CO pressures and disappears on evacuation showing that the heat of adsorp-tion of CO on these sites is much smaller than that on Pt and Pt2+ sites. The band isassigned to CO chemisorbed over oxide ions (cp. Amberg and others). At roomtemperature, oxygen is chernisorbed on the reduced Pt and there is no oxide formation-addition of CO rapidly removes the chemisorbed oxygen giving gaseous C02 andonly the 2070 cm-1 band develops.We would note that the three bands observed by Peri when CO was chemisorbedon partially reduced supported NiO and assigned to structures which may be designatedNi(CQ) at 2050-2080 cm-1, Ni+(CO) at 2140 cni-1 and Ni2+ (2170 cm-1) have similarpeak wave numbers, 2070-2080 cm-1 (Pt(CQ)), 2120 cm-1 (Pt2+(CQ)) and 2170 cm-1(02-(CO)), as those observed by us on an " oxidized " Pt surface.Prof.J. Tarrkevich and Dr. Y. Fujita (Princeton University) said : We wish to report onthe photochemical preparation of methyl radicals adsorbed on silica and their stabiliza-tion for days at room temperature. The methyl radical was prepared by photolysis ofmethyl iodide adsorbed on Vycor glass using 2537A radiation. The radicals weredetected with the Varian e.s.r. spectrometer. The spectra for the methyl radical,trideuteromethyl radical and 54 % 0 3 methyl radical are given in fig.1. The methylradical shows four sharp lines due to the protons (nuclear spin 3). The intensity ratio1 C. Eyraud et al., Compt. rend., 1962, 254, 688182 GENERAL DISCUSSIONof these lines is 1-0 : 3.0 : 3.3 : 1-0 compared to the theoretical intensity of 1 : 3 : 3 : 1.The hyperfine splitting is 22.6 gauss coinpared with McConnell's theoretical value of225. The width of the individual lines was 1-0 gauss at room temperature and 0.8gauss at liquid nitrogen temperature.FIG. 1.-E.s.r. absorption derivative spectra of methyl (la) and deuteromethyl(16) and C13 methyl (lc)radicals at room temperature-iC1 f ebacevac0 5 10 15time, minFIG. 2.-Interaction of gases with methyl radicals at room temperature.The trideuteromethyl radical shows seven lines due to the deuterons (nuclear spinThe intensity ratio is 1-0 : 3.3 : 6.8 : 7-5 : 6.5 : 3.2 : 1-4 compared to the theoreticalvalues of 1 : 3 : 6 : 7 : 6 : 3 : 1.The hyperfine splitting was 3-54 gauss while thetheoretical value based on the ratio of gyromagnetic constants of the proton to thedeuteron is 3.47GENERAL DISCUSSION I83The 54 % C13 46 % 0 2 methyl radical showed a spectrum of twelve lines withseparation and intensities corresponding to those calculated by Cole et a2.1 Theseparation between the ultimate lines is 67.5 gauss giving a difference of 38.5 gauss.Theory (1) predicts for this difference a value of 38 gauss for a planar CH3 freelymoving radical 4-10 or 60-66 gauss for a planar radical fixed in one position butrotating around its symmetry axis and a difference of 300 gauss for a tetrahedral CH3configuration On the basis of the observed spectra we must conclude that CH3 inthe pores of silica at room temperature and even at liquid-nitrogen temperature isplanar and is free to move either on the surface of the silica or in the cavity of the pore.The intensity of radical production increased linearly with irradiation reaching6 x 1014 spins in 30 min. The decay of the signal at room temperature had a half-lifeof 100 h.This decreased to less than 5 min at 90°C. The reactions of the methylradical with H2, D2, 0 2 , NO, CH31 were studied (fig. 2).The methyl radical has been successfully prepared, stabilized and characterized,physically and chemically, at room temperature.This investigation indicates thatsurface may stabilize reactive intermediates and thus influence the course of catalyticreactions. The method used offers the possibility of investigating these adsorbedreactive intermediates by preparing them within small pores and having the reactivegas diffuse into the pore from the gas phase at a rate controlled by the gas pressure.Dr. G. Lienard (kcole Roy. Militaire, Brussels) said: I am surprised that it ispossible to prepare evaporated films less clean than the metallic wires used for theirevaporation. The criterion of cleanliness is based on the difference of rates for the0-p hydrogen reaction either in presence of the wires or of the evaporated films.Instead of speaking of cleanliness or dirtiness of the surface, I see two other explana-tions : (i) The structure of the wire and the films are different, e.g., the proportion offaces, and the reaction mechanism-or the kinetics-may not be the same on allcrystal faces and so the catalytic properties of nickel seem to be different.(ii) Thesecond explanation is based on the author’s observation that it is necessary to repeatthe oxide-reduction operations followed by an outgas of the metal to obtain a constantcatalytic activity. We have observed catalytic activity variations with other metallicflms (Mo, W, Rh, Te, Ni, Re, Pt, Pd, Ti, . . .). The catalytic activity variations havebeen measured in a flow system and it varies under the influence of the reagents them-selves-until it stabilizes after a few hours at a constant value.The work on thesystem C&+D2 on molybdenum films has been published.1Thus, the criterion of constant catalytic activity does not constitute a check of thesurface purity, but only a test of experimental reproducibility. The catalyst is not themetal alone but the metal plus the surface states. If the reaction proceeds with anadsorption-desorption mechanism, it is not evident that all the surface states arerapidly in equilibrium with the gaseous phase. In fact, in a clean system-in ultrahigh vacuum conditions and high purity of gases and metal4 think a few p.p.m.impurity in the system might cause an important variation of the catalytic activity.This effect on the catalytic activity is stiil not well known and very few authors havemeasured this effect directly.With the nickel wires described here, one may supposedifferent surface states to occur from the repeated oxido-reduction treatments.Dr. I?. R. Norton (Nottingham Unicersity) said: In answer to point 1 of Dr.Lienard, it was not our intention to attribute the entire difference in catalytic activitybetween wires and films to contamination; indeed we believe that the topographicinhomogeneity of the film surface is the underlying cause for these variations, possiblythe easy accessibility of all sites on a smooth wire surface leading to higher rates. The1 Cole, Pritchard, Davidson and McConnell, Mol.Physics, 1956, 1, 406.1 J. Chim. Physique, 1964, 61, 1174184 GENERAL DISCUSSIONactivity of a film within 2 min of evaporation for the H2 + D2 reaction has been shownto be sensibly constant over long periods at 10-10 torr after an initial decrease attribut-able entirely to sintering effects. This does not fit with the fact that a filamentpoisons in some 30-60min in a 10-10 torr vacuum (to 30 % of original activity).Allowing for this factor, we certainly have always considered point 1 as may be seenin any of Eley's papers.In answer to your second point, the oxidation-reduction treatment was only apreliminary treatment to free the wire from carbon-containing materials and sub-sequently, after reduction and annealing, high temperature outgassing was used as afinal cleaning method.No variation of activity was found with conditions of 0 2treatment provided that subsequent reduction, outgassing and anncaling were com-plete. With films, a deviation from first order kinetics was found which was attribut-able to reversible hydrogen poisoning and all k, values were measured after this initialreversible poisioning had been allowed to go to saturation. It is not surprising that acarbon-containing adsorbate modifies the catalytic activity since it is these speciesthat we have to remove before obtaining a clean surface as judged by maximum cataly-tic activity and rapid chemisorption of H2 at 77°K up to a monolayer.Prof. D. D. Eley (Nottingham Uniuersity) said: Dr. Petro and I1 have discussedthe change in conductivity 0 and thermoelectric power S of metals due to the presenceof adsorbed gas, in terms of the modern theory of metals.2 In this theory CT = const/A, where I is the mean free path and A the area of the Fermi surface in the metal.Inother words, we are not concerned so much with the number of conductivity electronsin the metal, but the number at the Fermi-surface. Since in transition metals 3= f(Nd(G)), where-f(Nd(G)) is a function of the density of states at the Fermi surface,we visualize two possibilities. (a) The Fermi level is above the maximum of the densityof states-energy diagram. Then formation of Md+---Hd- will lower the Fermi level,increase A and also increase f(Nd(G)). Hence 0 = const A/f(Nd(G)) and couldincrease if A is dominent, and decrease iff(Nd(G)) is dominant.The latter wouldappear to be the case for low coverages in Pd and Ni. (b) The F e d level is belowthe maximum in the density of states diagram. Hence formation of M"--Hd- willdecrease A and decreasef(Nd(G)), and if we assume the latter is still the dominatingfactor (T will increase.Since according to Mignolet 4 the polarization on platinum is Ptd+-Hd-, then anincrease in 0 on adsorption referred to by Dr. Ponec might imply (b) above is operative.This may be surprising, since Pt has a similar position to Ni and Pd in the periodictable. Still essentially we are concerned with a surface effect, and the density of statesdiagram on the surface will not necessarily be the same as in the bulk metal, althoughthis is the simplest assumption to make.Dr.V . Ponec (Znst. Physic. Chem., Prague) said : With regard to the remarks of Eley,when orbitals of a metal atom are extracted by chemisorption bond from the participa-tion in the metal-metal bonds this process always increases the resistance. Thissituation is described by Sachtler as shortening of a mean free path of current carriers.There is not sufficient proof that only d-orbitals can bind, e.g., hydrogen. Theparticipation of other orbitals in chemisorption is probable. Then, the Fermi surfaceis influenced by chemisorption, if it is essentially the transfer of a charge only.However, the chemisorption is predominantly covalent as mentioned in our paper. Itis not possible to predict the slope of N(E) curve at the Fermi level for nickel with any1 Eley and Petro, Nature, 1966, 209, 501.2 Ziman, Electrons and Phonons (Oxford, 1960), p.391.3 Mott and Jones, Theory of the Properties of'Metals and Alloys (Oxford, 1936), p. 192.4 Mignolet, J. Chem. Physics, 1957, 54, 19GENERAL DISCUSSION 185certainty and to deduce from this prediction how the mean free path is changed by thechange in the slope of this curve.Prof. D. D. Eley (Nottingham Uniuersity) said: To establish whether hydrogen isadsorbed as molecules or atoms on a metal surface, even when this has a uniform heatof adsorption and otherwise obeying the criteria for the Langmuir adsorption isotherm,it will be necessary to secure accurate adsorption data over a wide range (severalpowers of 10) of pressure.It is doubtful if most data obtained so far are accurate orextensive enough for a proper test. It is possible to be misled by linearizing theLangmuir isotherm ; there are two ways of so doing, eqn. (2) and (3) below :molecular atomic1 1 1V Vmbp K +- (2a) - ---In these equations, Y is the volume adsorbed at s.t.p. and Vm the monolayervolume. For a strict test we need to use the least-squares method to fit the nonlineareqn. ( l a ) and ( l b ) to the data, and to compare the standard deviations to find which plotgives the best fit. But it is usual to linearize eqn. (l), as in (2) or (3) and to decide theresult by least squares or inspection. The plots of eqn. (2) 1/V against l/p or l/p*emphasizes the low pressure data, and those of eqn.(3), p / Y against p , or p+/ V againstp t , emphasizes the high pressure data. An example of the uncertainties of thisprocedure is shown in fig. ( l a ) and l b below. They refer to data of Ward for theL I I I0.5 1-0 1-5FIG. 1adsorption of hydrogen on reduced copper powder A, as presented by Trapnell in hisbook. Trapnell points out that the graph of p*/V against p* gave a non-linear plot,when compared with that of p / V against p reproduced here in fig. la. But the samedata plotted in fig. l b as l / V against l / p ) gives a reasonably good straight line, whe186 GENERAL DISCUSSIONwe compare it to the systematic deviation at low pressures in fig. la. We have hadto omit one point from fig. l b at l/pf = 7.25 for reasons of scale.Tt lies above thestraight line as drawn.We do not attempt to draw any conclusions about these data, which could onlycome from a statistical analysis as outlined above, but merely to emphasize thedanger inherent in the visual inspection of linear graphs derived from non-linearequations, and the need for a much more critical examination of isotherms withrespect to associative and dissociative adsorption. Superficially, fig la and btogether might point to atomic adsorption at low pressures going over to molecularadsorption at higher pressures, if they were borne out by more experiments over widerranges of pressure.Prof. F. C. Tompkins (Imperial College) (communicated) : Apart from the wordsof caution of Eley concerning the analysis of an adsorption isotherm in terms of aLangmuir-type equation in the two pressure ranges, there is the more fundamentalexperimental results that the heat of adsorption of gases on metals is almost invariablya function of coverage ; this renders any deductions about the power of p extremelyhazardous.It is, for example, easy to show that in molecular adsorption, a decreaseof heat of adsorption with coverage can lead to an isotherm which, on analysis as aLangmuir isotherm in the low-pressure range, can give the exponent of p as 3 insteadof unity. The Langmuir isotherm contains two disposable parameters V . and b ; theyhave physical significance and should not only have acceptable magnitudes, butshould also be proved to be the same when both the low-pressure and high-pressureanalysis are used.Prof.C. Kernball (Queen's Uniceusity, Belfast) said: Prof. Eley is right when hesays that care is necessary in testing the fit of experimental points to any particularequation. Nevertheless, it is important not to overemphasize the applicability of theLangmuir isotherm. There are unlikely to be many cases where adsorption data overa really wide range of pressure can be fitted accurately to the Langmuir isotherm.The point that I want to stress is that the correct pressure function to be used wheneverdissociative adsorption of hydrogen is occurring is the square root and not the firstpower of pH2. This statement holds irrespective of the type of isotherm selected.Dr. J. Miiller (Inst. Inorgan.Chern., Czechoslovak Acad. Sci., Prague) said: Inconnection with Dr. Roberts' paper, we have found 1 that the chemisorbed layer ofoxygen on nickel powder increases the B.E.T. Vm value of nitrogen by about 10-20 "/oin comparison with that on the clean surface. Further, at - 196°C solely physicaladsorption of nitrogen proceeds on the clean surface, whereas on an oxygen-coverednickel a part of nitrogen is more strongly adsorbed. These phenomena may beexplained by the change of the number of nitrogen molecules in thc monolayer dueto the electrostatic forces between chemisorbed oxygen layer that is polarized andinitially electroneutral nitrogen molecules. The occurrence of the electrostatic forcesin a similar case has been suggested also by Campbell and Duthie.2The influence of pre-adsorption (chemisorption) on the subsequent physicaladsorption and surface area determination is complex : in addition to the change ofthe number of physically adsorbed molecules in the monolayer the change of thesurface area itself has to be taken into consideration, too.It concerns, e.g., thedecrease of the surface area due to the blocking up of the pores in the solid or due tothe sintering occurring during pre-adsorption. Finally, the increase of the surfacearea caused by the incorporation and/or rearrangement of the surface layer cannot beexcluded.1 Miiller and Regner, Coll. Czech. Chern. Comm., 1965,30, 3399.2 Campbell and Duthie, Trans. Favaday SOC., 1965, 61, 558GENERAL DISCUSSION 187My second remark concerns the nature of oxygen adsorption on nickel. We havefound 1 , 2 in good agreement with Roberts' work function measurements 3 ~ 4 thatno incorporation of oxygen into the crystal lattice of nickel occurs at - 196°C.At- 78, 22 and higher temperatures, however, incorporation occurs in increasingextent. At 22°C and pressures of 10 torr of oxygen the extent of oxygen adsorptionamounts to about 2 oxygen atoms per 1 surface atom of nickel.From our measurements it also follows that incorporation of oxygen from theadsorbed layer proceeds if no oxygen is present in the gas phase. This finding is inagreement with Roberts' measurement fig. 1. In such cases it is convenient to speakabout the thermal regeneration of oxygen-covered nickel surface.The extent of thethermal regeneration found after 1 h evacuation (10-5 torr) at different temperaturesis given in the table :temp. of evacuation in "C - 196 - 78 + 22 +165 +315extent of thermal regeneration in%I, % 0 5 10 14 48From measurements of oxygen adsorption at - 196°C we found that a greaterpart of the nickel surface is covered by weakly bound (physically adsorbed) oxygenwhich can be removed by 1 h evacuation at - -78°C :surface coverage 6 = O/Ni ; T = - 196°C ; Pfinal = 40 torrtotal adsorption chemisorption physical adsorptionsampleB3 1 -73 0.3 1 1 4 2A4 1.33 0.16 1.17These results agree with those of Davydova and Kiperman 5 who found that at- 196°C only 15-20 % of nickel surface is covered by strongly bound (chemisorbed)oxygen. Possible explanations of these findings are given in ref.(3).Dr. D. Brennan (University of Liverpool) said: In reply to the postulation ofrelatively thick lower oxide phases during the initial stages of adsorption of oxygen bynickel at about 300"K, it would appear that insufficient account has been taken of thesaturation coverage data. At 300"K, the heat of adsorption is similar to the heat offormation of NiO, but the coverage at saturation is much less than this stoichiometrywould necessitate if at the same time all the nickel atoms originally involved in theformation of the lower oxide were still to be implicated. In advancing our hypothesis,we have attempted to explain certain striking experimental observations in terms ofdefinite surface structures.Descriptions in terms of the formation of various oxidesof continuously variable composition, unspecified structure and arbitrarily assignedproperties are in a sense merely restatements of the problems, not solutions of them.Dr. D. F. Klemperer (University of Bristol) said: When oxygen is admitted to anickel film, particles bearing both positive and negative charge are thrown from thesurface.6 These currents, whose sizes initially compare with the true photoelectriccurrent, may persist for as much as an hour at room temperature and their presence iseasily demonstrated by making photoelectric-type experiments with the light switchedoff. In addition, illumination can stimulate the emission of electron showers with1 Miiller, Kinetika i kataliz, 1966, 7, 188.2 Miiller, J.Catalysis, in press.3 Quinn and Roberts, Trans. Faraday SOC., 1964, 60, 899.4 Roberts and Wells, this Discussion.5 Davydova and Kiperman, Kinetika i kataliz, 1965, 6, 137.6 Anderson and Klemperer, Proc. Roy. SOC. A, 1960, 258, 350188 GENERAL DISCUSSIONwork function values which range to below 2 eV.l Perhaps it is these electrons whichoriginate from the quasi-Fermi levels and pseudo-equilibrium states of Roberts andWells. In any event, such effects must be distinguished when the method of plottingphotoelectric current against time is used during gas admission.A further point concerns the scheme in which areas of Ni,O (low work function)and Ni,O (high work function) form simultaneously from a precursor designatedNi-0". These three states emerge clearly from the photoelectric data of Andersonand Klemperer 2 who found that the low work function areas were as much as 0.48 eVpositive with respect to clean nickel films (6 = 5-14 eV) and that high work functionareas had a minimum work function of 5.71 eV.They concluded that both types ofnickel area could take up more oxygen precursor but that the low work function areaswere unable to nucleate the oxide structure, perhaps because of their unfavourablecrystallographic orientation.The low work function (oxygen-contaminated) areas might be expected to exhibitunusual adsorption characteristics and it is interesting, therefore, to see this realizedfor two gases. The case of hydrogen suggests that surface potentials,3 electricalconductivity changes 4 and heats of adsorption 5 in the nickel-hydrogen system canindeed depend on the excellence (or otherwise) of vacuum technique employed.Contamination with oxygen may even be the reason why industrial nickel catalystsare better than clean films in catalysing the reaction between carbon moncxide andhydr ogen.6Dr.M. W. Roberts (Queen's University, Belfast) said : There are many features ofsimilarity (see ref. (9), (lo), (14) and (18) of our paper) between our studies and those ofAnderson and Mlemperer; there are, however, some differences in detail. Dr. C. M.Quinn showed that with polycrystalline nickel ribbons no significant emission ofelectrons could be detected during oxygen interaction in the absence of light.Thecurrent we observe is therefore wholly due to the process of photoemission. Mr.McKee has shown that the same is true with copper. It therefore appears that nickelfilms have unusual properties in this respect and it would be worthwhile investigatingthe phenomenon.Areas of low and high work function were postulated by Anderson and Klemperer ;the problem has now been resolved further since the different states have been isolatedby controlling the substrate temperature and oxygen pressure. For example, chemi-sorbed oxygen is in general unstable with respect to bulk nickel above -195";clearly, the next stage is to study individual crystal planes.Dr. Klemperer deduced that areas of high and low work function existed from theoccurrence of split Fowler curves. Since a Fowler analysis gives essentially a thres-hold value then it is strictly incorrect to refer to it as a work function under the presentconditions where an oxide is present.The threshold could reflect emission from sur-face states (provided a sufficient density was present) or from the " valence band " andneed not reflect the work function of the system. It is therefore not valid to compareKlemperer's Fowler values with our capacitor data ; they would only be in agreementunder particular circumstances and then only fortuitously.Prof. S. J. Teichner ( T i m . Rech. Catalyse, ViZZeurbanne) said : It would appear thatthe comparison of the properties of a NiO formed on metallic nickel (Ni/NiO) with1 Grunberg and Wright, Pruc.Roy. Soc. A , 1955,232,403.2 Anderson and Klemperer, Proc. Roy. Suc. A , 1960, 258, 350.3 Sachtler, J. Chem. Physics, 1956, 25, 751.4 Oda and Arata, J. Physic. Chem., 1958, 62, 1471.5 Brocker and Wedler, this Discussion.6 Tompkins, Nature, 1960, 186, 3GENERAL DISCUSSION 189those of a NiO prepared by decomposition of a nickel salt or hydroxide (NiO) isdifficult. The behaviour of these two materials is sensibly different.ll 2 As anexample, the amount of CO adsorbed on Ni/NiO is much smaller than the amount of0 2 . On NiO this situation is reversed. The complex CO3 which is stable on Ni/NiOup to 300°C is decomposed on NiO at room temperature merely by an excess of CO.Also the processes of formation of this complex on two materials are different. OnNi/NiO the C03 complex may be formed by an interaction of CO with preadsorbedoxygen. On NiO this interaction leads to adsorbed C02 and the CQ3 complex isformed only by the interaction of 0 2 with preadsorbed CO. On NiO the inhibitingeffect of COa in the oxidation of CO at room temperature is commonly observed bymost authors. As Dr. Roberts and Dr. Wells do not observe this effect on Ni/NiO,it would appear that the mechanism of the room temperature oxidation of CO isdifferent on the two catalysts. Thus, the comparison of the results of Klier for theadsorption of CO at room temperature on NiO with those of Dr. Roberts and Dr.Wells for CO on Ni/NiO at 170°C is also difficult.Dr. M. W. Roberts (Queen’s Uniuersity, Belfast) said: Prof. Teichner has empha-sized rightly that care is necessary in any comparison of data obtained with oxidizednickel prepared by different methods. One of the objectives of our studies of theNi+O2 system is to determine how, and under what conditions, does nickel oxidizeto nickel oxide and when is the latter NiO. This problem is fundamental to any studywith oxidized nickel and particularly so with that prepared by decomposing nickelhydroxide (Teichner’s preparation) where the fresh oxide increased in activity aftersuccessive CO oxidation experiments and regeneration at 200O.1 Dell and Stone usedoxidized nickel granules (prepared via the oxalate) while we used oxidized nickelfilms, so that there are four systems for comparison (Teichner’s, Stone’s, Klier’s andours) and not just two (NiO and Ni/NiO). The studies of Dr. Klier with NiO areconsistent with our conclusions from work function and adsorption studies ; we claimno more than this.1 Dell and Stone, Trans. Faraday Soc., 1954, 50, 501.2 Teichner and Morrison, Trons. Faraduy Soc., 1955,51,961. Teichner, Marcellini and Rue, Adu.Catalysis, 1957, 9, 458. Marcellini, Ranc and Teichner, Proc. 2nd Int. Congr. Catalysis, (ed.Technip, Paris, 1960), p. 289. Gravelle and Teichner, J . Chim. Physique, 1964,61,527,533,625,1089, 1098. Rub and Teichner, Bull. Soc. Chim., 1964,2791,2797. Cou6, Gravelle, Ru6 andTeichner, Proc. 3rd Int. Congr. Catalysis, (ed. North-Holland Publ. Co., Amsterdam, 1964),p. 748

ISSN:0366-9033    

DOI:10.1039/DF9664100175

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Thermodynamic Factors in Adsorption and Catalysis -Equilibria in the Adsorbed PhaseBY C. KEMBALLDept. of Cnemistry, The Queen’s University of Belfast, N. IrelandReceived 7th October, 1965Chemisorption and catalysis of complex molecules or mixtures involve the establishment of somebut not necessarily all the possible equilibria between gas and surface and between different kinds ofspecies on the surface. Each such equilibrium established necessarily exerts a thermodynamicinfluence on the amounts of the various adsorbed species which take part in the kinetically controlledprocesses associated with the adsorption or the catalysis. Problems of this kind are discussed andemphasis is placed on the importance of using the correct functions of pressure in expressions todescribe the amounts of radicals and atoms on surfaces.The use of the concept of a virtual pressureis also advocated for situations where some but not all of the molecules in the gas phase are equili-brated with the surface.Adsorbed atoms and radicals are frequently encountered in studies of chemisorp-tion and catalysis. They may be formed by the dissociative adsorption of somemolecular species or by the establishment of equilibria between two or more gases andthe surface phase. When such equilibria are set up, the amounts of each kind ofadsorbed species will be related to some function of the pressures of the various gasespresent. The object of this paper is to attempt to set out in general terms some of thethermodynamic factors which may be relevant in studies of adsorption and catalysiswhen two or more gases are present.We shall stress the importance of using thecorrect functions of the pressures of the gases in quantitative discussions of adsorptionand of the kinetics of catalytic reactions. There are often cases when some but notall of the gases present are in equilibrium with the adsorbed species and the conceptof a virtual pressure can be useful in considering such cases. In other examples theremay be some limitation on the extent to which the equilibria between the variouspossible species on the surface are attained and situations of this kind are most impor-tant in relation to the selectivity of catalytic processes.In this paper we shall not attempt to discuss in detail the complications whicharise as surfaces become covered with atoms, radicals or other chemisorbed species.Sorn;e allowances will have to be made for the effects of coverage and we shall assumethat equations of the Langmuir type can be applied.Thus, we are ignoring effects ofsurface heterogeneity and changes of heats of adsorption with coverage.1. BASIC LANGMUIR EQUATIONS AND THE CHEMISORPTION OFHYDROGENIn this section, we set out briefly some of the standard equations which will berequired. The Langmuir isotherm for the adsorption of a single gas iswhere OA is the fraction of the surface sites covered at pressurep~. If the adsorptionis weak 6 A will be proportional to P A ; if strong, the coverage becomes independent1 9C. KEMBALL 191of pressure and for intermediate strengths of adsorption the relationship (1.1) may,over a limited range of conditions, be approximated bywhere 0 < n < 1.areandand like eqn.(1.1) these may be replaced by simpler expressions of the typefl*= P i , (1 -2)When two gases compete for the same sites on a surface, the appropriate isothermsOA = apA/(l -!- a p A -!- bPB),= bpB/(l -!- apA+ bpB)7(1.3)(1 -4)e*K PiPB". (1.5)When dissociative adsorption occurs, as with the chernisorption of hydrogen asatoms, the kinetic equilibrium for the adsorption is given byklpHz(l -eH)2 = (1 -6)where kl and k2 are constants. Rearrangement of eqn. (1.6) leads to the form of theLangmuir isotherm appropriate to the dissociative adsorption of a diatomic gaswhereWe want to stress the difference between eqn.(1.7) and eqn. (1.1). It is not correctto attempt to described the coverage of the surface by chemisorbed atoms by insertingthe pressure of hydrogen moZecuZes in eqn. (1 .I)-the appropriate pressure function isthe square root of the pressure. The basis for eqn. (1.6) was the kinetic equilibriumbetween adsorption and desorption but consideration of the thermodynamic equili-bria which are set up yields the same type of result. It is worth examining thisapproach which can be applied readily in more complex cases. The following equili-bria are established :gas phase H2+2Hdr 11surface phase (H2)u=2(H)uThe chemisorbed hydrogen atoms are in equilibria not only with the molecularhydrogen in the gas phase but also with the pressure of hydrogen atoms in the gasphase even aIthough the latter is extremely low at room temperature.For the gas-phase dissociation we haveand hence,Substitution from eqn. (1.10) into eqn. (1 .l) clearly gives an equation similar in formto eqn. (1.7).= &/pH,, ( 1 -9)pH = K*pft,. (1.10)2. AMMONIA, HYDROGEN A N D NITROGENIt is useful to consider what happens on metal catalysts in the decomposition orsynthesis of ammonia because it is one of the best known applications of the conceptof a virtual pressure. Temkin and Pyzhev 1 introduced the idea of a virtual pressur192 THERMODYNAMIC FACTORS I N CATALYSISof nitrogen in order to account for the observed kinetics ot the catalytic decompositionof ammonia.The idea may be represented by the following diagramgas phase NH3, Hz+[Nz] dr 1;surface phase NH2, NH, H+NC'where the dotted lines indicate virtual equilibria. Equilibrium is assumed to beestablished between the various chemisorbed species and the gas-phase ammonia andhydrogen. It follows that the amount of chemisorbed nitrogen can be represented asbeing in equilibrium with that pressure of nitrogen gas which would satisfy the gas-phase equilibrium,with the existing pressures of hydrogen and ammonia. If the adsorption of speciesother than nitrogen atoms is assumed to be weak, the Langmuir isotherm is obtainedby substituting the appropriate pressure function in eqn. (1. I), to yield~ N H J + N ~ + ~ H ~ , (2.1)@N = aNpNH3/Pfi2/(1 aNPNHJ/P$Z),which, by analogy with eqn.(1.2), may be approximated asThe kinetics of the decomposition of ammonia for which the rate-determining step isassumed to be the desorption of two nitrogen atoms as a molecule of nitrogen gas willthen be given by the expressionwhere R is the rate of reaction. Many examples are known293 where the observedkinetics take a form closely similar to that of eqn. (2.4).The virtual pressure of nitrogen can be very large. The equilibrium constant ofreaction (2.1) is about 6 x 103 atm2 at 400°C and so in the presence of partial pressuresof an atmosphere of ammonia and of hydrogen, the virtual pressure of nitrogen willbe some 6000 atm. If the pressure of hydrogen is reduced, the virtual pressure ofnitrogen increases.The idea of the virtual pressure is useful in understanding how it is possible toconvert iron in iron nitride Fe4N at, say, 400°C by treatment with ammonia.Thedirect conversion 2 in the presence of nitrogen gas would require pressures of over 103atm but virtual pressures in excess of this are readily attainable even with low pressuresof ammonia. The thermodynamic potential of the adsorbed nitrogen atoms risesand becomes sufficiently high for the formation of the bulk nitride provided that therate of desorption of nitrogen is sufficiently slow.3. CHEMISORPTION OF METHANEThe main point reached in the discussion in 5 1 and 2 is worth restating briefly.Once an adsorption equilibrium has been attained the amount of any radical on thesurface will be related to the pressure (real or virtual, small or large) of that radical inthe gas phase.This pressure in turn may be expressed as the appropriate function ofthe real pressures of the stable molecular gas or gases which are in equilibrium withthe surface and which give rise to the radical.A more complex adsorption can now be examined. Let us suppose that the dis-sociative adsorption of methane gives rise to the adsorbed species CH3, CH2, CH, Cand H. In order to discuss equilibrium it will be necessary to know the pressure oC. KEMBALL 193hydrogen in the gas phase as well as the pressure of methane. The appropriatepressure functions for each of the adsorbed species will be as follows :adsorbedspecies CH3 CH2 CH C HThe isotherms for each species are obtained by use of equations similar to eqn.(1.3)and (1.4) but as we are considering radicals we must substitute the appropriate pressurefunctions (and not merely the pressure of stable gas molecules) for PA andpB. Thus,assuming that all species are competing for the same surface sites, the isotherms will beeCH3 = a 1 P C H ~ P ~ Z * / ~ ) (3.1)eCHz = aZPCH4Pi:/x, (3.2)eCH = a3PCH4Pi:/X~ (3 3)eC = a4PCH4Pi:/X, (3 -4)eH = c p & ~ / ~ , (3 - 5 )= + a IPCH.qP;:+ a2pCH4pH,' + a3pCH4pG:+ a 4 P C € 1 4 P i . f cPl$z. (3.6)whereSome simplification of the isotherms may be possible in practice if one or more of thespecies is weakly adsorbed, e.g., if (C), is very small, the term a4 p ~ ~ , p ~ , 2 will be negli-gible compared with the other terms in eqn.(3.6).Suppose that theadsorption of methane in the presence of hydrogen forms the adsorbed species CH3and H but that dissociation to form CH2, CH and C occurs so slowly that the rate canbe neglected. The consequence will be that a pseudo-equilibrium is set up involvingonly (CH& and (ma. Such partial or pseudo-equilibrium may not be important inthe case of methane but they can be of profound importance for higher hydrocarbonsand more complex molecules.Another possibility is the establishment of a partial equilibrium.4. ETHANE AND HIGHER HYDROCARBONSThe adsorption of ethane exemplifies phenomena which may be expected withhigher hydrocarbons and because the establishment of some of the adsorptionequilibria lead to catalytic processes.A great variety of possible situations can beenvisaged for the dissociative adsorption of ethane depending on the extent to whichdifferent dissociation processes are involved. We define five such situations chosensomewhat arbitrarily as ideal cases in order to illustrate the various possibilities. Aswith the adsorption of methane, the pressure of hydrogen in the gas phase has aninfluence on the adsorbed species.The following cases are considered :(a) the dissociative adsorption of up to n hydrogen atoms in each molecule of ethaneoccurs rapidly and reversibly. Thus, if n = 2, the adsorbed species C2H5, C2H4and H will be in equilibrium with the gas phase but more highly dissociatedspecies will not be formed in appreciable quantities during the course of anexperiment ;(b) the dissociation of all hydrogen atoms in each molecule occurs rapidly andreversibly.This leads to the complete range from (C2H& to (C& and alsoin equilibrium with the gas phase 194 THERMODYNAMIC FACTORS IN CATALYSIS(c) as in (b) but the rupture of the C-C bond also occurs reversibly on the surfacegiving rise to CH3, CH2, CH and C as additional adsorbed species.(d) as in (c) but equilibrium is established between (C), on the surface and carbon inthe bulk of the adsorbent or catalyst, e.g., as a metal carbide;(e) as in (d) but methane is formed in the gas phase so that surface phase finallybecomes equilibrated with gas phase ethane, hydrogen and methane.The representation of the adsorption isotherms for each of the adsorbed species incases (a) and (b) presents no difficulty. The isotherms will be similar to those formethane in eqn.(3.1)-(3.6) except that the pressure formation for (C2H& will bep~~~+p;;~* and corresponding expressions will apply to the more highly dissociated C2species.In case (c), the situation is similar to that described in 6 2 for ammonia, hydrogenand nitrogen. The various single carbon species for (CH& to (C), are in equili-brium with the gas-phase ethane and hydrogen and hence with a virtual pressure ofmethane which would satisfy the gas-phase equilibriumC2H6 + H2 G2CH4. (4.1)Schematically, the equilibria in case (c) aregas phase C2H6 + H2+ [CH4]4: 11 / C2 species +C1 species surface phaseand the actual pressure of methane in the gas phase is substantially below the virtualpressure as with nitrogen in the decomposition of ammonia.The appropriate pressurefunctions controlling the amounts of the single carbon species are as follows :The equilibrium constant of reaction (4.1) is 1012 atm at 25°C * and so in the presenceof atmospheric pressures of ethane and hydrogen the virtual pressure of methane willbe 106 atm. It follows that the amounts of the various single carbon species on thesurface are likely to be substantial under case (c) and the fraction of the surfaceoccupied by C2 species will be greatly reduced.The ability of ethane and more particularly higher hydrocarbons to form carbideswith certain metals as in case ( d ) follows from the considerations just discussed forcase (c).If the rate of desorption of single carbon species as methane is sufficientlyslow, the thermodynamic potential of (C), may be very large and sufficient to causecarbide formation if the atoms can pass into the bulk from the surface. On the otherhand, if methane is formed in the gas phase the catalytic hydrogenolysis of ethane byreaction (4.1) will occur and the system will approach the position defined by case (e).Our discussion of ethane has been largely theoretical and it is interesting to see towhat extent the cases described are relevant to experimental investigations of adsorp-tion or catalysis with hydrocarbons. The exchange reaction on metal films betweenethane and deuterium4 has been shown to involve the interconversion of adsorbedethyl radicals and adsorbed ethylene molecules and consequently the situation on theli Unless otherwise stated, thermodynamic data are taken from publications of the AmericanPetroleum Institute, Research Project 44C.KEMBALL 195surface may correspond to case (a) with n = 2. On the other hand, more highlydissociated C2 species may be present on the surface even although they are notrehydrogenated sufficiently rapidly to play any significant part in the mechanism of theexchange reaction. Such a situation would be intermediate between case (a) andcase (b). Miyahara 5 has argued that adsorbed C2H3 and CzH2 are involved in themechanism in addition to C2H5 and CzH4 but his proposal has been criticized.6Iron films 7 show no activity for the catalytic exchange of ethane and deuteriumbut they bring about the hydrogenolysis of ethane at higher temperatures.It seeinslikely that the situation on the iron films may approximate to case (c) with the rate ofthe reversible adsorption of ethane drastically reduced because of the extensive cover-age of the surface by single carbon species. If this description is correct, it impliesthat some limitation on the extent of the various dissociative processes is essential fora metal to retain catalytic activity for the exchange of ethane.An interesting intermediate state of behaviour is found for the exchange ofn-hexane with deuterium on rhodium films.7 Initially the catalyst shows activity butthe activity declines with time and the behaviour has been described as " self-poison-ing ".There may be a gradual formation of strongly adsorbed hydrocarbon specieswhich reduce the surface available for the reversibly adsorbed species responsible forthe exchange process. By analogy with the cases described for ethane, such stronglyadsorbed species may be formed by rupture of C-C bonds 011 the surface, as in case(c), or by stripping hydrogen atoms away from the carbon skeletons, as in case (b).It is perhaps remarkable that examples of self-poisoning with the exchange reactionsof hydrocarbons on metal catalysts are not encountered more frequently. In otherwords, it is surprising that with so many metals the desirable amount of dissociationfor the exchangz process can occur sufficiently rapidly without leading to furtherdissociation or breakdown of the molecules and consequent blocking of the surface.Self-poisoning of exchange reactions has been found to occur fairly extensively withcompounds, other than hydrocarbons, such as alcohols,* arnines 9 and ether.10One of the drawbacks of discussing the adsorption of ethane in terms of fivedistinct cases is that it tends to obscure the likelihood that one case will changegradually into the next in many practical examples. Thus, in practice, it will beunusual to find the complete equilibrium with all the Cz species (case (b)) attainedbefore some rupture of C--C bonds occurs to lead on to case (c).The merging ofthe various cases is seen clearly in the work of Roberts who studied the adsorption ofethane on clean rhodium 11 and iridium 12 surfaces.Gaseous methane was formedover iridium at temperatures as low as 27°C and over rhodium at 0°C.The hydrogenation of ethylene is often regarded as a simple non-selective type ofcatalytic reaction but this is not strictly correct. The equilibrium constants at 25°Cfor the following reactionsandC2H4 + 2H2+2CH4,are 1017.7 atm-1 and 1029.7 atm-1 respectively and, in the presence of excess hydrogen,the formation of methane and not ethane would be expected. Thus for a catalyst towork satisfactorily for the hydrogenation of olefins, it must exhibit some selectivity innot bringing about processes of the type associated with cases (c) and (e). Considera-tions of the possible smface equilibria help us to understand why slower rates ofhydrogenation are sometimes found when ethylene is admitted to a catalyst beforehydrogen than when the two gases are admitted together.When ethylene is admitte196 THERMODYNAMIC FACTORS I N CATALYSISalone more highly dissociated C2 species may be formed by the types of surface reac-tions associated with case (b) but some polymerization may also be possible. Theequilibrium constants for the reactions2C2H4+C4Hg (but-1-ene), (4.3)and 3C2H4eC4H12 (hex-1 -ene), (4.4)are 1011 atm-1 and 1020 atm-2 respectively at 25°C. Once highly dissociated orpolymeric species are formed on the surface a considerable time may be required fortheir removal after the admission of hydrogen.The possible equilibria for adsorbed ethane have a direct bearing on the catalytichydrogenolysis of ethane to form methane by reaction (4.1).Early investigations 13showed that the rate of reaction over a supported nickel catalyst was dependent on asubstantial negative power of the hydrogen pressure of as much as -2-5. Themagnitude of this inhibition by hydrogen was more than could be explained by com-petition between hydrogen atoms and the hydrocarbon for the surface of the catalyst.Later work has indicated that the negative power arises in part through the need toform a Cz species fairly well stripped of hydrogen atoms to take part in the rate-determining step involving the rupture of the C-C bond. Kemball and Taylor 14argued that the rate-determining step was the rupture of the bond in an adsorbedethylene molecule but their method of working out the expected kinetics was onlyapproximate and a much more thorough kinetic analysis was carried out by Ciminoet aZ.15 Their approach is worth examining in detail because it provides a usefulexample which is relevant to this discussion.Cimino et aZ.15 assumed that the rate-determining step involved a hydrogen mole-cule and an adsorbed species CzH,.The coverage of the surface by this species wasthen expressed in terms of what we have called the pressure function p c , ~ , / & - ~ ) ’ ~ inthe Langmuir isothermwhich was then used in the approximate formBy selecting values of x and using the observed order of the reaction with respect toethane they were able to calculate the orders with respect to hydrogen in good agree-ment with those observed experimentally.They found evidence for values of x of2 or 4 on different iron catalysts and they suggested that x = 0 on supported nickel,i.e., on nickel the rate-determining step is the rupture of the bond in a C2 species-aconclusion which has been substantiated by recent work.16 This implies that thesubsequent desorption of methane is relatively fast compared with the breakdown ofthe C2 species on the surface. Anderson and Baker 17 came to different conclusionsabout the rate-determining step of the hydrogenolysis over nickel films in that theybelieved that the desorption of methane and not the rupture of C-C bonds on thesurface was rate-determining.The same conclusion was reached by Galwey 18 whostudied the adsorption and subsequent hydrogenolysis of several hydrocarbons onsupported nickel.A comparison of eqn. (4.5) with one of the eqn. (3.1)-(3.6) shows one weakness inthe analysis by Cimino et aZ.15 They make no allowance for competition from anytype of adsorbed species other than C2H, for the sites on the surface. This mayintroduce a substantial error particularly if, as suggested by Sinfelt et aZ.16 theadsorption of CzH, is endothermic. Some of the negative dependence of the rate oC . KEMBALL 197hydrogenolysis may well be due to direct competition for the surface by chemisorbedhydrogen and a term such as c&,f ought to be included in the denominator of ther.h.s.of eqn. (4.5).One of the main stumbling blocks to a more extensive application of the types ofthermodynamic factors discussed in this paper is the difficulty of allowing for mutualcompetition of species for sites on the surface. Such competition and the furthercomplications associated with moderately strongly adsorbed species often mask therelatively straightforward application of the thermodynamic factors. A simpleexample will illustrate the point. Let us suppose that we are interested in the coverageof a surface by phenyl radicals when the gases present are benzene and hydrogen.For simplicity we shall assume only three kinds of adsorbed species C&5, C6H5 andH. It follows thatand the operation of the thermodynamic factors is obvious. If C6H6 only is stronglyadsorbed, we have'C6HJ &2! 9 (4.10)but if the phenyl radicals are strongly adsorbed, the influence of the pressure ofhydrogen ceases to affect their concentration and the thermodynamic factors arecompletely masked.The examples quoted have been concerned almost entirely with catalysis on metalsbut the same principles can be applied to other systems as well.Whenever thedesorption of one of the products of a catalytic reaction is slow, the virtual pressureof this substance may increase and bring about side reactions which would not other-wise occur. The equilibriaand(4.1 1)(4.12)were examined by Kemball and Rooney 19 in the catalytic cracking of cyclopentene onsilica-alumina. More cyclopentane and much more methylcyclopentane wereobserved in the gas phase then would have been expected for the known pressures ofcyclopentene, hydrogen and methane in the gas phase.This provided evidence thatthe virtual pressures of both hydrogen and methane were higher on the surface than theactual pressure and that the desorption of each was a slow reaction-a conclusionwhich was supported by the high activation energy observed for the formation ofboth gases as products of the cracking reaction.5. THE FISCHER-TROPSCH SYNTHESISA detailed discussion of the Fischer-Tropsch synthesis necessarily requiresconsideration 20 of the various equilibria involving carbon monoxide and hydrogen inorder to find the upper limits to the possible production of higher hydrocarbons andalcohol.We want to show that thermodynamic arguments can also be useful in thediscussion of the mechanism of the reaction and in this way, we can demonstrate thatthe action of a successful catalyst has to be extremely selective. By making use of the198 THERMODYNAMIC FACTORS I N CATALYSISconcept of virtual pressure one can discuss the steps in the mechanism even althoughthe intermediate products of these steps are not necessarily found in the gas phase.Two of the most thermodynamically favoured reactions of carbon monoxide andhydrogen areCO+3&+CH4+&0, (5.1)and 2 c 0 2H2+CH4 + c02, (5.2)which have equilibrium constants at 227°C of 1010 and 1012 atm-2 respectively.Neither of these reactions is likely to be important for the mechanism of the synthesisbecause the union of two molecules of methane to form ethane is not going to takeplace unless the virtual pressure of methane is very high as the equilibrium constant ofreaction (4.1) at 227°C is about 107.4 atm.Any catalysts which were active forreactions (5.1) or (5.2) would be poor synthesis catalysts.We now examine the possibility that the surface intermediate responsible for thesynthesis is not methane but methanol. The equilibrium constants 20 at 227°C for thereactionsCO +2H2+CH30H, (5.3)2CH30H+C&I58H -I- H20, (5.4)and CH30H+ C~H~OH+CH~CHZOH+ HzO, (5.5)are 10-2.2 atm-2, 107.2 and 106.2 respectively. Thus, we shall not expect substantialamounts of methanol to be formed although the virtual pressure might be sufficient topropagate synthesis by the favourable steps (5.4) and (5.5).On the other hand, thereis experimental evidence that radioactive methanol 21 is not incorporated into thesynthesis products as readily as radioactive ethanol.We now assume that the species which gets incorporated into the growing moleculeis not methanol but carbon monoxide and we write the equilibria as2CO + 4H2fC2H50H + H20CO +2H2 + C~HSOH+C~H~OH + HzO(5.6)(5.7)The equilibrium constants for these reactions at 227°C are 102,s atm-4 and 104.0atm-2 respectively and so they represent thermodynamically feasible routes for thesynthesis. We are not implying that ethanol necessarily has to appear in the gasphase-this will depend on the relative rate at which it is formed by (5.6) and reinovedby (5.7).Saturated hydrocarbons can then be formed by the hydrogenolysis of thealcohols or olefins by their dehydration. Qf course, the actual entities which takepart in the mechanism on the surface are much more likely to be radicals rather thanmolecules and proposals for the mechanism in terms of radicals have been given.20Our thermodynamic arguments provide support for the idea that once the C-0 bondis broken the synthesis ceases except when the C-0 bond is broken at the union oftwo carbon atoms with the elimination of water. A carbon skeleton without anoxygen atom has lost the potential for growth.6. CONCLUSIONSThe main points which we should like to stress are the following :(a) it is important to use the right pressure functions in eqn.(1.11, (1.3) and (1.4) inplace of PA and p~ when the adsorbed species are atoms or radicals. This isessential even although the Langmuir isotherm may only give an approximaterepresentation of the effects of surface coverage on strength of adsorption C . KEMBALL 199(6) when some but not all of the gas phase compounds are in equilibrium with asurface the concept of a virtual pressure (which may be very large) of any gas non-equilibrated but capable of being formed from equilibrated gases is useful inconsidering conditions on the surface ;(c) catalytic activity is frequently associated with some selective behaviour of thecatalyst in regard to the kinds of dissociative adsorption of the reactant andproducts which are brought about ;(d) even in complex catalytic processes, thermodynamic considerations may help toeliminate some individual processes as possible steps in the mechanism. Hereagain, the concept of virtual pressure can be employed with advantage.1 Temkin and Pyzhev, Actaphysicochim., 1940, 12, 327.2 Frankenburg, Catalysis, ed. Emmett (Reinhold, New York, 1955), vol. III, p. 171.3 Bokhoven, van Heerden, Westrik and Zwietering, Catalysis, ed. Emmett (Reinhold, New York,4 Anderson and Kemball, Proc. Roy. Soc. A, 1954,223, 361.5 Miyahara, J. Res. Inst. Catalysis, Hokkaido Univ., 1956, 4, 143 ; 1957, 5, 115.6 Kernball, J. Res. Inst. Catalysis, ,%kkaido Univ., 1957, 4, 222.7 Gault and Kemball, Trcms. Faraday SOC., 1961, 57, 1781.8 Anderson and Kemball, Trans. Faraday Soc., 1955,51,966.9 Kemball and Wolf, Trans. Faraday Soc., 1955,51, 11 11.1955), vol. 111, p. 265.10 Clarke and Kemball, Trans. Faraday SOC., 1959, 55,98.11 Roberts, Trans. Faraday Soc., 1962, 58, 1159.12 Roberts, J. Physic. Chem., 1963, 67, 2035.13 Morikawa, Benedict and Taylor, J. Amer. Chem. SOC., 1936, 58, 1795.14 Kemball and Taylor, J. Amer. Cltem. Soc., 1948, 70, 345.15 Cirnino, Boudart and Taylor, J. Physic. Chem., 1954, 58, 796.16 Sinfelt, Taylor and Yates, J. Physic. Cfiem., 1965, 69, 95.17 Anderson and Baker, Proc. Roy. Suc. A , 1963, 271, 402.18 Galwey, Proc. Roy. SOC. A, 1963, 271,218.19 Kemball and Rooney, Proc. Roy. Soc. A , 1960,257, 132.20 Storch, Golumbic and Anderson, The Fischer-Tropsch and ReIated Syntheses, (Wiley, New York,21 Kummer and Emmett, J. Amer. Chem. SOC., 1953, 75, 5177.1951) p. 9 and 569

ISSN:0366-9033    

DOI:10.1039/DF9664100190

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Adsorption and Co-ordination of Unsaturated Hydrocarbonswith Metal Surfaces and Metal AtomsBY G. C . BONDResearch Laboratories, Johnson Matthey and Co., Ltd.Exhibition Grounds, Wembley, Middx.Received 14th January, 1966A survey is first made of the co-ordination of olefins and chelating diolefins to metal atom inorganometallic complexes : with d* and dl0 complexes, the observed stability sequences for ethyleneco-ordination (e.g., N i V PdII<PtII ; AgI< PdII< RhI) are interpreted in terms of the spatialdirection and extension of the &orbitals. Measurements of the power of diolefins to inhibit theisomerization of I-pentene catalyzed by methanolic RhCl3 confirm their stronger co-ordinationwhich is attributed to an entropy effect. The principal phenomena of homogeneous catalysisby organometallic complexes are consistent with the above stability sequences, optimum activityoccurring with RhI and PdII complexes : comparison of these phenomena with those in hetero-geneous hydrogenation catalysis leads to the working hypothesis that the reactive state of adsorbedunsaturated hydrocarbons is also a v-complex.A molecuIar orbital model appropriate to a face-centred cubic metal is outlined, and the direc-tion of emergence of the eg and t2g orbitals at the (loo), (110) and (111) planes obtained: the degreeof occupation of the orbitals is discussed, and possible models for .rr-adsorbed states of olefins,diolefins and alkynes are sketched.The (111) plane is the least suited for their adsorption.Abrief attempt is made to apply molecular orbital concepts to mechanisms in hydrogenation catalysis.The commercial justification for fundamental research in catalysis lies in thehope that the knowledge thereby acquired will assist the development of catalystswhich are cheaper, more active, more selective and longer lived than those hithertoin use. One approach to the first two targets for supported metal catalysts is toattempt to reduce the size of the metal crystallite so that virtually every metal atomis employed. It is, however, important to know whether there is a practical limitto this, and whether all the characteristic catalytic properties of a metal are pre-served at very small crystallite sizes : or whether indeed any novel properties thenappear. The ultimate degree of metal dispersion is expected to be achieved inmetal-impregnated zeolites, although little is as yet known of their characteristics.That a single metal atom possesses all the attributes necessary for catalysis isestablished by the many metal-containing species now known to catalyze homogene-ously the activation of molecular hydrogen and the isomerization, polymerization,hydrogenation and oxidation of unsaturated hydrocarbons.Tn several instances,by the application of physical measurements, it has proved possible to describethe active species with great assurance. A critical comparison of the phenomenaof homogeneous and heterogeneous catalysis should therefore enable us to decidewhether, or to what extent, an assembly of metal atoms differs in its properties fromthe individual metal atom.Moreover, if an analogy is established between thetwo fields, mechanistic information gained on homogeneous reactions should assistthe unravelling of heterogeneous mechanisms. The purpose of this paper is to ex-amine the present position in this area.20G. C. BOND 201HOMOGENEOUS A N D HETEROGENEOUS REACTIONS OFUNSATURATED HYDROCARBONSCO-ORDINATION AND CHEMISORPTION OF UNSATURATED HYDROCARBONSDESCRIPTIVE CHEMISTRYAlthough it is not universally established that the attachment of an olefin or alkyneto a metal atom in either a homogeneous or a heterogeneous system necessarilyprecedes its reaction, nevertheless we consider the available information on theco-ordination and chemisorption of unsaturated hydrocarbons before discussingpossible reaction mechanisms.We deal first with their co-ordination to metal atomsin complexes.Mono-olefins co-ordinate to a variety of metal atoms.1 The bonding is par-ticularly stable in the d8 square planar complexes of the second and third rowmetals (RhI, IrI, Pdn and Ptn) : their structural features have been examined bya number of physical methods, and only thermochemical information is wanting.Co-ordinated ethylene is, however, kinetically labile in Zeise’s salt 2 and in certainRhI complexes,3 rapidly exchanging with C2D4. Ag* and RUE also form com-plexes of moderate stability with mono-olefins.1 Octahedral d6 species (e.g., Rhm)do not in general form olefin complexes, nor are stable “ simple ” complexes 4formed with any of the first row metals, although a number of complexes formulatedas (PR&Ni (olefin) have been described.5 To the left of group 8, the presence ofother n-bonding ligands (e.g., n-CgH5, CO, PR3) is necessary to stabilize the metal-olefin bond.Chelating (i.e., non-conjugated) diolefins form much stabler complexes than domono-olefins, even when the two double-bonds are at the ends of a long flexiblechain (e.g., as in 1,7-octadiene, see below).Thus stable ‘‘ simple ” diolefin com-plexes are known with CoI, Nin, CuI, IrI and Osu, as well as with all the speciesforming mono-olefin complexes. This is despite the fact that in cyclic diolefins(e.g., 1,5-cyclo-octadiene) the planes of the two double bonds cannot be normal tothe co-ordinate bond axis.Iron is notable for its preference for conjugated diolefins ;attempts to prepare RhI, Pdn and Pt* complexes of 1,3-cyclo-octadiene lead tocomplexes of 1,5-cyclo-octadiene .6n-Allylic complexes may be regarded as a form of olefin complex: they areformed most extensively by CoI and Pdn, but isolated examples are known in-volving Rum, Rhm and Ptn. The allyl group occupies two co-ordination sites.Complexes of Cro, Feo and NiO containing only allyl groups have low stability,and various Nin-ally1 species have been proposed as intermediates in the oligomer-ization of butadiene.5 Butadiene does not, as might be expected, form a four-centre n-ally1 complex with palladium chloride,7 although multicentred complexeshave been recognized in other systems.A complex in which methylene is co-ordinated to IrI is known.8Comparatively few complexes of alkynes with elements of the transition serieshave been recorded. Tert-butylacetylene (tba) can displace ethylene from[(C2H4)PtC12]2 to give [(tba) PtCf212 and complexes of the general formula(Pq53)zPt (alkyne) have been prepared 1 : clearly the mode of bonding is differentin these two types of complex. Ethylene is displaced from [(C2H&RhC1]2 by1-pentyne and 2-pentyne,9 but the products have not been adequately character-ized. Rhodium carbonyl chloride, [Rh(C0)2C1]2, reacts with disubstituted alkynesto give, amongst other products, polymeric species of the formula [(alkyne)Rh(CO)Cl],.lo A number of complexes of W, Fe, Co and Ni containing alkynesas well as CO or cyclopentadiene ligands have been described 11 : in the binuclea202 ADSORPTION AND CO-ORDINATIONspecies, the C=C bond is at riglit angles to the metal-metal bond and is somewhatlonger than normal.12Despite the almost total lack of quantitative information on the bond strengthsin these complexes, certain qualitative trends are emerging.In groups 8 3 and lB,the stability of mono-olefin complexes increases from Nin to Pdn to Ptn, althoughCuI complexes are stabler than those of AgI 13 : relevant to this is the existence 14of the complex trans-(C2H&PtC12 below - 20°C, while no analogous Pd complexis known. By the same token, Rh may be taken to form stronger complexes thanPd, for the complexes [(C2H&RhC1]2 and (CzH&Rh acac are quite stable.Thedifferent C=C stretching frequencies in ethylene complexes support these views.3There is little quantitative information on the effect of olefin structure on the stabilityof complexes, although undoubtedly ethylene complexes are the most stable. Onlyfor the AgI complexes are any quantitative stability measurements available, butthe situation is confused by the combination of steric and electronic effects.ls 13Olefins are known to interact with the surfaces of many of the transition metals,although generally substantial disruption of the molecules occurs 15 : heats ofadsorption therefore refer to no particular process and are of limited value. Thepositive surface potential found 16 for ethylene on a Ni film also related to speciesof unknown structure, and infra-red studies 17 have been less helpful than mighthave been hoped in resolving these problems.Virtually nothing is known of theinteraction of alkynes and of diolefins with metal surfaces.It is in any event doubtful whether species formed in such circumstances are thosewhich participate in catalytic reactions, and more emphasis should perhaps be placedon indirect assessments of chemisorption strengths arising from kinetic analysis ofreacting systems. This information is of two kinds: (i) the sequence of chemi-sorption strengths of an olefin on a series of metals, and (ii) the sequence of chemi-sorption strengths of different unsaturated hydrocarbons on one metal.(i) Thesequence of chemisorption strengths of ethylene at 50°C on a number of alumina-supported metals has been derived from an analysis of the ethylene-deuteriumreaction,ls and isPt EIr > Pd> Rh> Ru- 0s.The placings of Pd and Kh are reversed below room temperature. The positionof Fe, Co and Ni relative to these metals is uncertain, but on their general charac-teristicsl they would be placed in the neighbourhood of Pd and Rh. Cu wouldfall at the end of the sequence. (ii) Alkynes and diolefines are more strongly chemi-sorbed during their hydrogenation on metal surfaces than are olefins, but smalldifferences in heats of adsorption may cause large differences between areascovered.15 Alkynes and diolefins are chemisorbed with comparable strengths onPd and Ni.There is one further area in which the properties of olefin-metal complexes andmetal surfaces show common phenomena. The olefin-metal bond in complexesis comparatively weak, and olefin is often readily displaced by diolefins and alkynes :many other ligands, including phosphines, amines, nitriles, cyanide ion and carbonmonoxide, can, however, also cause olefin displacement 19 and these include mole-cules which are catalyst poisons for hydrogenations.Again no quantitative in-formation is available, but a causal connection is strongly suggested.HOMOGENEOUS AND HETEROGENEOUS CATALYSIS : AREAS OF ACTIVITYHomogeneous catalysis of the reactions of unsaturated hydrocarbons is chieflyexercised by compounds of the elements of group 8.The simplest reaction is thaG. C. BOND 203of olefin isomerization (double-bond migration or cis-trans isomerization) whichcan occur in the absence of hydrogen.20 Salts and complexes of Rh and Pd areparticularly effective, while those of Pt, Ir and Ru, and of the first row metals, aremuch less so, except for 1,3-cyclo-octadiene which isomerizes in the process ofcomplex formation with both PdCl2 and PtC12 to the less stable but more stronglycomplexing 1,5-cyclo-octadiene.6 Complexes of AgI are apparently unreactive,but photocatalyzed reactions of olefins co-ordinated to CuI have, however, beenreported21More widely disseminated is the ability to catalyze homogeneous hydrogenation.In the first row metals, Fe(C0)S and notably the pentacyanocobaltate ion, Co(CN): -,are homogeneous hydrogenation catalysts, although the latter only affects diolefinsor “ activated” olefins.Salts of Nin and Crm also have weak hydrogenationactivity,z2 as does PdCl2 when suitably promoted? The rhodium phosphine,(R3P)3RhICl, and the inadequately defined Ptlr-Snn complex are probably themost active yet discovered, although of major interest are the Ir complexes(Pr3)2Ir(CO)Cl and (PR&Ir(CO)C124 : a related 0 s complex also exhibits someactivity.24 No clear pattern of activity has yet emerged, for activity depends sosubstantially on the ligands surrounding the central atom. The conditions requiredto achieve successful homogeneous hydrogenation seem more rigorous than thoseneeded for isomerization, and in particular the presence of n-bonding ligands isrequired : sirnple salts have at best very low activity.Unsaturated hydrocarbons are readily polymerized by a number of group 8metal salts and complexes.Cyclic oligomerization of butadiene to 1,5,9-cyclo-dodecatriene occurs in the presence of Ni complexes,5 while its polymerization totrans-poly- 1,4-butadiene is catalyzed by RhCl3 and I~Cl~2-5 Ethylene is selectivelydimerized to butenes by RhC13 in alcohol.26 Alkynes are dimerized in the presenceof PdC12 to give cyclobutadiene derivatives and are polymerized by RhC13 10 : oligo-merization of 1-heptyne occurs with a number of group 8 metal chlorides in thepresence of a hydridic reducing agent such as lithium hydride.27The oxidation of ethylene to acetaldehyde in the presence of PdC12 is not strictlycatalytic, but this and a number of related processes (such as the synthesis of vinylacetate from ethylene and acetic acid) can be made to perform continuously if aCun salt is added.No other metal salts have been reported as significant homo-geneous oxidation catalysts. Much work has also recently been reported on thecarbonylation of unsaturated hydrocarbons catalyzed by PdC12.The following provisional generalizations may therefore be made. (i) Activityresides chiefly in salts and complexes of Rh and Pd and to a much smaller extentin those of Ir and Pt. (ii) Complexes of the first-row metals are active in reactionsof diolefins but not of mono-olefins. (iii) Complexes of CuI and AgI are unreactive.This pattern is consistent with what is known of complex stabilities.Mono-olefincomplexes of the first-row metals are insufficiently stable to be reactive while thoseof the third-row metals are too stable. Maximum activity is therefore found inthe second row (Rh and Pd). Considerations other than stability may determinethe lack of reactivity of group 1B complexes : because the d-shell is then full, thereare no vacant directed orbitals available for use and so a second reactant may beunable to engage in the complex.The behaviour of heterogeneous catalysts in these reactions is as follows. Allthe group 8 metals are active catalysts for the hydrogenation of mono-olefins, di-olefins and alkynes,l the characteristics of these reactions being determined by thedifferent strengths of adsorption of the hydrocarbons.The first-row metals areless active than the others: Cu catalyzes the reduction of diolefins and alkynes204 ADSORPTION A N D CO-ORDINATIONhigh selectivity resulting from the inability of the catalyst to adsorb the mono-olefin formed : Ag and Au are inactive. Olefin isomerization, except as an integralpart of hydrogenation, is generally thought not to occur.28 Polymerization occursto a limited extent except with acetylene during its hydrogenation. Oxidation ofolefins is non-selective over the group 8 metals, and only over Ag is a partially-oxidized product formed.CONCLUSIONSTwo broad conclusions emerge from this comparison between the reactivityof co-ordinated hydrocarbons and that of chemisorbed hydrocarbons. (i) Theareas of maximum activity are coextensive and are largely confined to group 8 :ability to form olefin complexes of moderate stability parallels homogeneous activityexcept for complexes of the group 1B metals, and we may suppose that for highreactivity on metal surfaces species of similar stability are needed.(ii) There are,however, some significant differences between the two fields, and these must reflectthe different properties of single metal atoms and of assemblies. For example,(a) Fe, Co and Ni are better olefin hydrogenation catalysts than expected on thebasis of the stability of their olefin-metal complexes. (b) Polymerizations are cata-lyzed by complexes and not by surfaces, perhaps because in the former there aremore co-ordination positions available than at the latter.( c ) Selectivity inhomogeneous oxidation is achieved because the complex disrupts at the criticalstage, a possibility which does not exist in a heterogeneous system. These differ-ences, however, arise from the different environments of metal atoms in the twosystems and do not necessarily mean different forms of bonding.The case for considering that the bonding of olefins to metal surfaces to give areactive species similar in nature to that in olefin-metal complexes rests on thefollowing points. (a) There is a similarity between the sequence of ethylene chemi-sorption strengths quoted above and the sequence of stabilities of ethylene com-plexes.(b) Olefins are less strongly chemisorbed or co-ordinated than either alkynesor diolefins. ( c ) Olefins are displaced from olefin-metal complexes by ligands whichare also catalyst poisons. It would be unexpected if two different kinds of bondingof olefins to metal atoms led to such a close parallelism of behaviour.Arguments against the participation of a n-bonded olefin in the exchange of abicyclononane 29 may be criticized on two grounds. (i) The observed phenomenaare not established to be general since they refer only to a Pd catalyst, and thereare strong grounds for believing that n-allylic species contribute significantly inexchange reactions on this metal. (ii) n-Olefin states are probably only weaklybonded and may therefore only appear at high coverages, whereas the coverageby hydrocarbon species in the exchange of saturated hydrocarbons is low.Weshall therefore adopt as a working hypothesis the idea that the reactive state in olefinhydrogenation and related reactions is a n-bonded olefin.INHIBITION OF THE RhC13-CATALYZED ISOMERIZATION OF 1 -PENTENE BYDIOLEFINS A N D ALKYNES 9Many qualitative observations show that diolefins and alkynes co-ordinate tometal atoms more strongly than do mono-olefins. We have attempted to gainquantitative information in the following manner. The isomerization of 1 -penteneproceeds smoothly in the presence of methanolic solutions of RhC13 at 6OoC, equi-librium being attained in about 30 min (see fig. 1 ; 1-pentene, 0.9 M ; RhC13,0.005 M).We have found this reaction to be strongly inhibited by diolefins and alkynes : resultG. C.BOND 205for 1,3-pentadiene ( N 83 % trans, - 17 % cis) are also shown in fig. 1, and its effectat 0.001 M (i.e., pentadienelpentenez 10-3) is easily detectable. We now calculatethe relative strengths with which olefin and diolefin compete for the catalyst, asfollows : let us assumeC+O~CO+C+PC+D$CD1 3245where C = catalyst, 0 = olefin, P = product and D = diolefin. Assuming furtherthat [O]O [Clo, [D]o r [C], and [C] N zero (subscript zeros signifying concentra-tions before attainment of equilibria), we obtaintcoi-~o coiOP I - KD CDlo-[CDI'where KO = kl/k2 and KD = k4/k5. Hence we may calculate the fraction of catalystin the form of the olefin complex (to which the rate is taken to be proportional)II t I I 1YO:, 10 20 30. 4 0 5 0 6 0time, minFIG.1 .-Inhibition of 1-pentene isomerization (catalyzed by methanolic RhCl3 at 60°C) by variousconcentrations of 1,3-pentadiene ; see text for details.for any value of K ~ / K D as a function of the initial diolefin concentration [D]o. Itappears (fig. 2) that the diolefin inactivates more than an equal number of catalystmolecules, and we take this to mean that only a fraction of RhC13 added is cata-lytically active. We assume about 10 % of the RhCls is reduced to an active RhIspecies. Curve C (fig. 2) shows the calculated dependence of rate on initial diolefinconcentration for KO/& = 10-3 and the experimental points for both 1,3- and1,4-pentadiene lie close to this curve.Points for isoprene lie close to a calculatedcurve for which KO/& = 2.8 x 10-3.In a similar way we have evaluated approximate KO/& values for 1,7-octadiene(10-2), 1,5-cyclo-octadiene (10-2) and norbornadiene (5 x 10-4). The inhibiting effectsof 1- and 2-pentyne have also been examined, and analogous KO/& values are respec-tively 3-7 x 10-3 and 2.3 x 10-3 : they thus compete about as strongly as isoprene206 ADSORPTION AND CO-ORDINATIONMOLECULAR ORBITAL DESCRIPTION OF THE BINDING OF UNSATURATEDHYDROCARBONS I N COMPLEXESThe major contribution to the bonding of an olefin to a d8 metal atom probablyarises from two sources : (i) a a-bond formed by electron donation from the filledbonding orbital of the olefin to the metal atom’s vacant d’2-,,2 orbital and (ii) a x-bondformed by back-donation from a filled tzg orbital (which may have been hybridizedwith a p orbital for greater extension) into the vacant n antibonding orbital of theolefin : this tends to reduce the accumulation of negative charge on the metal atom.I I \r 1 1OgO025 O*OOS 0 * 0 0 7 5 0-010initial diolefin concentration, MFIG. 2.-Decrease in the rate of 1-pentene isomerization by various concentrations of 1,3-penta-diene (0) and of 1,4-pentadiene (0).A, curve calculated for [Clo = 0.005 M and Ko/KD = 0 ;By curve calculated for [C]O = 0.0005 M and &/KD = 0 ;C, curve calculated for [Clo = 04005 M and Ko/KD = 10-3.Ethylene molecules in CsHsRh(CzH& rotate about the co-ordinate bond axis, theenergy barrier being 6 kcal mole-1 : the rotation is facilitated by the participationof the t2g orbital in the xy plane (dZg) in the n-bond when the ethylene is approxim-ately in the plane of the complex.Molecular orbital theory is not able to offer aconfident interpretation of the relative strengths of olefin-metal bonds in RhI, Nin,Pda and Ptn complexes, but two factors are likely to be important. (i) A contrac-tion of the orbitals on passing from group 82 to 83 caused by the increase of nuclearcharge, resulting in less overlap and a weaker bond. (ii) A greater extension inspace of orbitals on moving down a vertical group : if this is not compensated byincreased bond-length (which seems to be the case), greater overlap and strongerbonding will result. Many other factors may contribute.The same description holds for diolefin d8 complexes.From our work on RhIspecies, diolefin complexes are more stable than the mono-olefin complex by 3-4kcalmole-1, and we conclude this must represent an entropy effect measuring thegreater difficulty of simultaneously breaking two olefin-metal bonds.The lower stability of the dlo complexes (CuI, AgI and AuI) is described by thenecessity of using the spherically-symmetrical vacant s orbital for the a-bond : thG. C. BOND 207overlap is less than with a directional dx2-y2 orbital and hence the bonding willbe weaker.No detailed molecular orbital treatment of the bonding in n-allylic complexesappears to have been given, although it seems likely that both 0- and n-bondingoccur, with the dzy orbital taking part.No explanation for the popularity of CoIand Pdn allylic complexes has yet been offered. In alkyne complexes which arestructurally analogous to olefin complexes, it is difficult to see how both sets ofn-orbitals can interact with the metal atom orbitals to account for their stability.In complexes where both sets are evidently engaged, the structure of the moleculeis substantially altered.12MOLECULAR ORBITAL MODEL FOR THE METALLIC STATEAmong the properties of metals requiring description are thermal and electricalconductivity, magnetic properties, bond length and crystal structure, bond strength(latent heat of sublimation, melting point, etc.) and chemisorptive and catalyticproperties.Two extreme models have been used : the collective electron or bandmodel and localized bonding or valence bond model.15 The former in its simpleform takes no interest in crystal structure : the latter, although essentially qualitative,has been extended to account for the regular alteration of crystal structure on passingacross the transition series 30 and for certain aspects of alloying behaviour.31 Onthe basis of both models, there must be both localized and collective electrons intransition metals, the former accounting chiefly for magnetic properties and thelatter for bonding and conductivity.Recently,however, the possibility has been appreciated of describing the formation of electronbands in terms of the overlap of molecular orbitals which are directional becauseof the existence of the crystal field.329 33 This model is particularly suited to ourpresent purpose because it provides a detailed (if not necessarily valid) picture ofthe bonding potentialities of surface atoms.34 Consideration is restricted to theface-centred A1 (f.c.c.) structure, which results from the overlap of each of thetwelve lobes of the t 2 , orbitals (four in each of the three planes) with one lobeprojecting from a near-neighbour (fig.3). Thus, we obtain the necessary co-ordina-tion number of twelve, and the t 2 , electrons occupy a collective, comparativelynarrow band. The remaining d-electrons occupy six e, orbitals directed towards(but not bonding with) those from next-nearest neighbours along the Cartesianaxes (fig.3): the eg electrons are therefore localized, although existing in twoenergy levels (or narrow bands) as a result of intra-atomic exchange-splitting (Hund’srule).33 We may therefore confidently predict the direction of emergence of thetwo kinds of orbital at various faces of any f.c.c. crystal.Difficulties arise when we consider the degree of occupation of the various orbitalsand bands. Trost 32 assigns the nine valence electrons of the cobalt group metalsso that both the tzg and eg bands are half-filled, and the ten electrons of the nickelgroup so that the tzg band is half-filled and the e, band empty. His assignment issomewhat arbitrary, and we prefer the more realistic picture of Goodenough33(fig. 4) who divides the 0.55 d-electron holes of nickel between the t 2 , band (0.41)and the e, band (0.14).Both bands are completely filled in group 1B. No exactdivision of the 0.75 d-electron holes in cobalt or of the 0.95 holes in iron has beengiven, but the e, band is probably vacant in iron and almost so in cobalt (fig. 4).No quantitative information on the noble group 8 metals is available, and we assumethat they are similar to the first row metals.Neither of these models is readily applicable to surface phenomena208 ADSORPTION AND CO-ORDINATIONWe now attempt to describe the adsorbed states of unsaturated hydrocarbonspecies on the three low index planes of a f.c.c. metal on the assumption that forthe relevant metals the e, band is sufficiently unoccupied to participate in a-bondingand the tzg band sufficiently filled to take part in n-bonding.Expected changesin chemisorption strengths and derived catalytic properties are then sketched inFIO. 3.-The disposition of atoms about a central atom in a face-centred cubic metal. Near-neighbours (open circles) are bonded by overlap t 2 g orbitals of (&, etc.) and next-nearest neigh-bows (filled circles) by overlap with eg orbitals (dx2-y2 and dz2).FIG. 4.4eparation of the d-band into t 2 g and e, sub-bands (following Goodenough 33).terms of the various degrees of occupation of the t2, and e, bands in horizontalseries : differences in the behaviour of vertical groups must be ascribed as before todifferent extension of the orbitals in space.MOLECULAR ORBITAL DESCRIPTION OF CHEMISORBED STATESThis model for the bonding of unsaturated hydrocarbons to metallic surfaceswill not permit bonding as strong as is possible in a square planar dS complexG.C. BOND 209especially with group 8 3 metals where the e, band is almost filled. Perhaps, however,this is required, for excessively strong bonding can only lead to low reactivity, andcertainly the activity of most complexes in homogeneous catalysis is low in com-parison with heterogeneous counterparts.Representations of the emergence of orbitals at the (loo), (1 10) and (1 11) facesof a f.c.c. crystal are shown in plan and in elevation in fig. 5-7. At the (100) surface,an e, orbital projects vertically from each surface atom and four tzs orbitals disposedas in fig.5. The former could overlap with the 1s orbital of a hydrogen atom whichPLAN SECTION THROUGH a...a9aSCALEFIG. 5.-Diagrammatic representation of the emergence of orbitals at the (100) face of a face-centred cubic metal. Filled arrows : e, orbitds in plane of paper ; hatched arrows : t 2 , orbitalsin plane of paper; open arrows : t 2 , orbitals emerging at 45" to plane of paper. The brokencircle shows the position of an atom in the next layer above the surface layer. In both the planand section, an e, orbital emerges normal to the plane of the paper from each atom. The scaleapplies to nickel.alternatively could settle in the octahedral hole where 1s orbital could be overlappedby five e, orbitals 34 giving presumably a strongly bound state.At the (1 10) plane,a tzg orbital projects vertically from each atom and a further four as shown in fig. 6 :PLAN SECTION THROUGH b .br--./' '\r \S C A L E t l F - - +-- -0 1 2 3 4 9 1FIG. 6.-Diagrammatic representation of the emergence of orbitalsyat'the (1 10) face of face-centredcubic metal. Filled arrows : e, orbitals in plane of paper or emerging at 45" ; hatched arrows :t z g orbitals in plane of paper ; open arrows : t 2 , orbitals emerging at 30" to plane of paper. In theplan, a t 2 g orbital emerges normal to the paper from each atom : in the section, it is an eLI orbital.There are also two eg orbitals emerging at 45" to the plane, providing another typeof site for a hydrogen atom. The (1 11) plane is peculiar in that no orbitals emerg21 0 ADSORPTION AND CO-ORDINATIONeither vertically or from the section through adjacent atoms (fig.7). The triadsof emergent e, and ~2~ orbitals have trigonal symmetry : a further type of hydrogensite is also apparent.PLANCSECTiOt. THROUGH c cbehind b c f p r cFIG. 7.-Diagrammatic representation of the emergence of orbitals at the (1 11) face of a face-centred cubic metal. Filled arrows : e, orbitals emerging at an angle of 36" 16' to the plane ofthe paper ; other arrows : as in fig. 6. Note : no orbitals emerge normal to the surface.ADSORPTION OF MONO-OLEFINSOn the basis of the assumed degrees of occupation of the e, and tzg orbitals,n-adsorbed mono-olefins should be readily adsorbed on the (100) plane as shownin fig.8 : moreover it should be able to rotate easily. On the (1 10) plane, themolecule should only absorb with the axis of the co-ordinate bond at 45" to thesurface, unless the functions of the eg and tzg orbitals can be interchanged. On theFIG. 8.-Representation of n-adsorbed ethylene on the (100) face of nickel : (a) plan ; (b) sectionshowing an orbital overlap (compare fig. 5).(1 1 I ) plane, three sets of three adjacent orbitals are coplanar, wherein the sequenceof the orbitals is eg, tZg, e, and hence an olefin might n-bond with the orbitals re-versing their normal function. The sequence t2,, e,, t2g is not coplanar. Thereis, however, a major difficulty in that the e, orbitals make an angle of only about35" with the surface, and this might well mean a substantial repulsion between theadsorbed molecule and adjacent surface atoms.There is good evidence for the existence of a-diadsorbed olefin29 regardless ofits relevance in olefin hydrogenation.Geometric considerations show that it ad-sorbs across the near-neighbour spacing.15 It is uncertain whether both the effand f2, orbitals are comparably suited for covalent bond formation in the group 8G. C . BOND 21 1metals : we will assume this is so, although in group 8 2 the tzg orbital will perhapsbe preferred. There are two possibilities for a-diadsorbed olefin: (i) using thevertical eg orbitals at the (100) surface and (ii) using either the vertical t 2 , orbitalsor the 45"-emergent e, orbitals at the (110) surface.The (111) plane again seemsunsuited.The co-ordination of carbon monoxide to a metal atom bears a marked resemb-lance to that of olefins in that the bond has both a- and n-constituents. .It hasbeen demonstrated35 by LEED that carbon monoxide is strongly adsorbed on the(100) face of nickel and that the CO/Ni ratio is 1 : 2 ; this is presumably a bridgestructure. It is more weakly adsorbed on the (110) face, the CO/Ni ratio being1 : 1, i.e., a linear structure. According to our model, the bridge structure isaccommodated on the (100) face if the next nearest neighbour distance is the onespanned and if t 2 g orbitals may be used for o-bonding (fig. 9). No suitably pro-jecting orbitals exist to span the longer spacing on the (110) plane, although the useIblFIG.9.-(a) Representation of bridge-adsorbed CO on the (100) face of nickel. The section isthe same as in fig. 5: irrelevant orbitals are omitted for clarity. (6) Representation of linear-adsorbed CO on the (1 10) face of nickel. The surface is viewed at 45"C, the CO being in the planeof the paper : irrelevant orbitals are again omitted.of e, and t 2 g orbitals as for an olefin permits a linear structure at 45" to the surfaceof this plane (fig. 9). Two such structures per atom seem possible, although noton adjacent atoms, and indeed, the infra-red spectrum of carbon monoxide onrhodium/alumina has been interpreted to suggest this possibility.36 Our modeladditionally suggests the feasibility of a linear species vertically on the (1 00) surface,and results have been obtained which suggest the existence of linear species attachedto atoms spanned by a bridge structure.37ADSORPTION OF DIOLEFINS AND ALKYNESThe adsorption of diolefins must involve the simultaneous interaction of bothdouble bonds with the surface if their stronger adsorption than mono-olefins isto be explained.According to our model, this requires the surface atom to havetwo emergent e, orbitals, a situation obtaining on the (1 10) face (fig. 6 ) : there arealso four t2g orbitals suitably disposed. We envisage two possibilities (see fig. 10)for the adsorption of butadiene on this plane, involving respectively one and (lessconveniently) two adjaceot surface atoms. Such sites would also accommodatea four-centre allylic species.The adsorption of a diolefin containing one or moremethylene groups between the double-bonds across near-neighbours on the (100)plane also appears possible. The (111) plane again presents its former difficulties21 2 ADSORPTION AND CO-ORDINATIONFor alkynes it is again necessary to postulate the simultaneous interaction ofthe two n-bonds with surface orbitals to explain their strong adsorption, but thegeometric requirements are different from those for diolefins. The (1 10) face onlyhas the necessary disposition of eg and tzg orbitals : the alkyne molecule sits acrosstwo adjacent atoms, as seen in fig. 11. The (100) face lacks suitably directed egorbitals, although adsorption simply through n-bond formation is a possibility.(a) ( b)FIG.10.-Representation of butadiene adsorbed on (a) a single atom and (b) two adjacent atomson a (110) plane. For simplicity C-C bond lengths are drawn as equal.FIG. 11.-Representation of acetylene adsorbed on the (110) plane of nickel (compare fig. 6).ADSORPTION AND CATALYSIS ON ALLOYSBecause the eg band is narrower than the tzg band, its degree of occupation willchange more rapidly, and on moving from group 82 to 8 3 it will alter from almostempty to almost full. This should be reflected in changes in cheniisorption strengthand catalytic activity. For hydrogenation reactions, nickel is usually superior tocobalt and platinum to iridium : palladium is not always better than rhodium butits true activity may sometimes be masked by toxic dissolved hydrogen.We cannotuse observed activity relations to pronounce on the role of the e, band, but we notethat on alloying a group 8 3 metal with a 1B metal the eg band will be filled beforethe t Z g band, and hence significant activity changes may be expected for only smalladditions of the group 1B metal. There are several cases in the literature wherethis effect may be operative.HYDROGENATION MECHANISMS AND ADSORBED INTERMEDIATESDiscussion of reaction mechanisms in heterogeneous catalysis is usually con-ducted without regard to detailed considerations of surface geometry or to thesteric probability of the steps invoked.1 We attempt an initial approach to thisproblem in relation to the hydrogenation of olefins and alkynesG. C. BOND 21 3HYDROGENATION OF OLEFINSThere are two distinct types of site possible for adsorbed hydrogen atoms on the(100) plane.Since an octahedral hole clearly cannot accommodate an olefin mole-cule while it can accept a hydrogen atom, it follows from the observed kinetics(showing no, or weakly competitive, hydrogen adsorption) that such a hydrogenatom cannot be reactive, for such sites are expected to be always filled. Hydrogenatoms bonded directly over metal atoms by means of a single eg (dt2) orbital musttherefore be those responsible for reaction. Such an atom is about level with then-bonded olefin, whose hydrogen atoms are presumably displaced away from thesurface as in olefin-metal complexes. cis-Addition of the hydrogen atom to forman ethyl radical bonded through either an eg or a t Z g orbital is therefore envisaged.Although an " octahedral " hydrogen atom may not be able to leave its site un-aided to attack the olefin, it could participate in reactions of the type :The concept of non-equivalent sites, for which there is some experimentalevidence, demands a careful specification of reaction mechanisms.It is, e.g., possibleto recognize three distinct types of site for a hydrogen atom on the (110) plane.HYDROGENATION OF ALKYNESA difficulty arises in considering the hydrogenation of alkynes, for the (110)face which contains sites where both n-orbitals may be engaged by adjacent atomsis not the preferred one for olefin chemisorption : and conversely the (100) facewhich is preferable for olefin chemisorption is not able to bind an alkyne by bothits n-orbitals simultaneously.If reaction were to occur on the (110) face and ifconsiderations of orbital availability were to encourage the desorption of the olefinas soon as it was formed, a new factor would enter into selectivity. This bringsus to a dilemma: the absence of olefin hydrogenation until the near exhaustion ofthe alkyne requires either that the alkyne can by its stronger adsorption preventolefin adsorption on all available crystal planes or that planes suitable only forolefin adsorption are entirely absent. Both of these alternatives seem improbable.A feature of alkyne hydrogenation is the preference for cis-addition of hydrogenespecially to disubstituted alkynes. This point is taken care of in the present model,for as with co-ordinated olefins the substituted groups will be displaced away fromthe surface and hence the molecule; will be predisposed to the cis-configuration.Furthermore, on the (110) plane, sites of low energy for hydrogen chemisorptionare available below the plane of the molecule, facilitating cis-addition.Vinyl radicals are intermediates in the hydrogenation of alkynes, and an-bondedspecies (e.g., HC=CH2) have been proposed.Related species are known in organo-/ I * *metallic complexes (e.g., an-co-ordinated ally1 group 38). Care is, however, required,for the axis of the n-bond is normal to the plane of the molecule in which the a-bond lies : the use of molecular models indicated that orbitals from two adjacentmetal atoms would be required to adsorb a an-bonded vinyl radical in comfort.CONCLUSIONSIt would be wrong to claim too much for the molecular orbital approach tochemisorption and catalysis outlined in this paper.There are too many uncer-tainties and assumptions involved, nevertheless the proper description of bindin214 ADSORPTION AND CO-ORDINATIONin solids and at their surfaces must be in terms of molecular orbitals. Restrictionsare then imposed on the nature of surface species which may be validly written bythe spatial arrangement, symmetry and degree of occupation of the emergent orbitals.I am grateful to Miss Margaret Hellier and J. A. Dawson for carrying out thework described on pp. 204-205 and to Dr. D. E. Webster of the University of Hull forhelpful discussions and a critical reading of a draft of this paper.1 Bond and Wells, Adv. Catalysis, 1964, 15, 92.2 Cramer, Inorg.Chem., 1965, 4,445.3 Cramer, J. Amer. Chem. SOC., 1964,86,217.4 Bond, Platinum Metals Reu., 1964, 8, 92.5 Wilke, Angew. Chem., 1963,2, 105.6Rinehart and Lasky, J. Amer. Chem. SOC., 1964, 86, 2516. Frye, Kuljian and Viebrock,7 Shaw, Chem. and Ind., 1962,1190.8 Mango and Dvoretzky, Abstracts of papers, 150th meeting of the h e r . Chem. SOC., September,9 Bond, Dawson and Miss Hellier, unpublished work.10 Maitlis and McVey, J. Organometal. Chem., 1965, 4, 254.11 Guy and Shaw, Adv. Inorg. Chem. Radiochem., 1962, 4, 78.12see, e.g., Bailey, Churchill, Hunt, Mason and Wilkinson, Proc. Chem. SOC., 1964, 401.13 Andrews and Keefer, Molecular Complexes in Organic Chemistry (Holden-Day Inc., San14 Chatt and Wilkins, J. Chem. SOC., 1952, 2622.15 Bond, Catalysis by Metals (Academic Press, London and New York, 1962).16 Mignolet, Disc. Faraday SOC., 1950, 8, 105.17 Eischens and Pliskin, Adu. Catalysis, 1958, 10, 1. Little, Sheppard and Yates, Proc. Roy. SOC.18 Bond, Phillipson, Wells and Winterbottom, Trans. Faraday Soc., 1966, 62,443.19 Cramer and Parshall, J. Amer. Chem. Soc., 1965,87, 1392.20 Bond and (Miss) Hellier, J. Catalysis, 1965, 4, 1. Spaxke, Turner and Wenham, J. CataZysis,1965, 4, 332; Harrod and Chalk, J. Amer. Chem. SOC., 1964, 86, 1776; Davies, Arcstral. J.Chem., 1964,17,212.21Trecker, Henry and McKeon, J. Amer. Chem. SOC., 1965, 87, 3261. Srinivasan, J. Amer.Chem. SOC., 1964,86, 3318.22Tulupov, Zhur. Jiz. Khim., 1957,31, 519; 1963, 37, 698.23 Maxted and Ismail, J. Chem. Soc., 1964, 1750.24 Vaska, Inorg. Nuclear Chem. Letters, 1965, 1, 89. Vaska and Rhodes, J. Amer. Chem. SOC.,1965,87,4970.25Rinehart, Smith, Witt and Romeyn, J. Amer. Chem. SOC., 1962, 84, 4145. Canale, Hewett,Shryne and Youngman, Chem. and Ind., 1962, 1054.26 Cramer, J. Amer. Chem. Soc., 1965, 87, 4717.27 Luttinger and Colthup, J. Org. Chem., 1962, 27, 3752.28 See, however, Phillipson and Wells, Proc. Chem. SOC., 1964, 222.29 Burwell and Schrage, J. Amer. Chem. SOC., 1965, 87, 5253.30 Altmann, Coulson and HumeRothery, Proc. Roy. SOC. A, 1957,240, 145.31 Robins, J. Less-Common Metals, 1959, 1, 396.32 Trost, Can. J. Chem., 1959, 37, 460.33 Goodenough, Magnetism and the Chemical Bond (Interscience, New York, 1963).34 Dowden, in Coloquio sobre Quimica Fisica de Processos en Superficies Sdlidas (Liberia Cien-35 Park and Farnsworth, J. Chem. Physics, 1965, 43, 2351.36 Yang and Garland, J. Physic. Chem., 1957, 61, 1504.37 Garland, Lord and Troiano, J. Chem. Physics, 1965, 69, 1195.38 Powell, Robinson and Shaw, Chem. Comm., 1965,78.Inorg. Chem., 1965,4, 1499.1965, paper S90.Francisco, etc., 1964), chap. 4.A, 1960,259,242.tifica Medinaceli, Madrid, 1963, p. 177

ISSN:0366-9033    

DOI:10.1039/DF9664100200

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Isotopic Exchange between Deuterium and Cycloalkane onPalladium CatalystsBY ROBERT L. BURWELL, JR., AND KLAUS SCHRAGEDept. of Chemistry, Northwestern University,Evanston, Illinois, U.S.A.Received 6th December, 1965Results of isotopic exchange between deuterium and the hydrocarbons, bicyclo[3.3. llnonane, cis-and trans-l,2-dimethylcyclopentane, on palladium-alumina catalysts at about 50°C are reported. Anintermediate common to the exchange of both dimethylcyclopentanes gives nearly completely ex-changed cis- and trans-l,2-dimethylcyclopentane in the ratio of one to about four. In addition tothis two-set exchange process, there is a one-set exchange process unaccompanied by epimerization inwhich nearly all of the hydrogen atoms in one set exchange. Other processes lead to the exchange of asmaller number of hydrogen atoms.The isotopic distribution patterns of exchanged bicyclononane are inconsistent with the occurrenceof one-set exchange via alternation between monoadsorbed alkane and n-complexed olefin but sup-port instead alternation between mono-adsorbed and di-adsorbed alkane.The mechanism of two-setexchange is considered in the light of these experiments. Several forms of the r-ally1 mechanism andof the roll-over mechanism are consistent with observations.Isotopic exchange between cyclopentane and deuterium on palladium-aluminacatalysts at about 50°C involves at least five processes: (i) two-set exchange in whichmost of the hydrogen atoms on both sides of the ring exchange, (ii) &-exchange inwhich cyclopentane-dg is the main product, (iii) one-set exchange in which nearly allof the hydrogen atoms on one side of the cyclopentane ring exchange, (iv) and (v), twoprocesses which lead to exchange of only a few hydrogen atoms and one of whichexchanges mainly one hydrogen atom.1-3 Because of uncorrelated relative yields ofthe various processes on a series of palladium-alumina catalysts, one concludes thateach process occurs on a separate set of sites.1The activation energies for these processes decrease from 22 kcal/mole for process(i) to about 11 for processes (iv) and (v).The kinetic orders in hydrogen are about-0.9 for process (i), - 1.2 for (ii), -0.8 for (iii), and roughly -0.4 for (iv) and (v).1In this paper we report the results of isotopic exchange between deuterium andcis- and trans- 1,2-dirnethylcyclopentane and bicyclo[3.3.llnonane on the catalystscharacterized above and we apply these results to the mechanism of these reactions,particularly to those of one-set and of two-set exchange. Some of the results onbicyclononane have been the subject of a preliminary communication.4EXPERIMENTALCatalysts IIIa, V and VII were 5 % palladium-on-alumina prepared by impregnating 60-80mesh Harshaw y-alumina with palladium chloride,ln 3 IIIa and V were from the same batch.The catalysts were preliminarily reduced at 140°C. Once in the apparatus, they were reducedover night in hydrogen at 350°C. This treatment at 350°C was repeated periodically withvery little effect upon the selectivity for processes (i)-(v) or for total activity.After initialreduction at 350°C, catalysts IIIa and V were evacuated for one day at 350°C with a mercurydiffusion pump backed by a liquid-nitrogen trap and then retreated with hydrogen overnight21216 ISOTOPIC EXCHANGE ON PALLADIUMat 350°C. Uncontrolled minor variations in the conditions of reduction made rather sub-stantial differences in the relative numbers of the various types of sites particularly thoseassociated with processes (iv) and (v).We used a flow apparatus in which dilution with helium was used to obtain reduced partialpressures of deuterium. Hydrocarbon vapour was introduced by passing the gas througha thermostatted tube containing Chromosorb P wetted with the liquid hydrocarbon.Withsolid hydrocarbons, a tube loosely packed with solid hydrocarbon was substituted. In thislast case, we knew only the upper limit of the partial pressure of hydrocarbon since we hadno independent measure of the degree of saturation achieved by this technique. Fischer andPorter Teflon needle valves were used throughout to avoid contamination by stopcock grease.Except for the results with bicyclononane, the results have been corrected for isotopicdilution using an average of the inlet deuterium content (0.997) and the exit content com-puted from the mass spectrometric results. The large content in dimethylcyclopentane-d14necessarily leads to very low accuracy in the small content in dl3. The large C13 correctionfor do leads as usual to lowered precision in d1.1 ,2-Dimethylcyclopentane was made by reaction of methyl magnesium iodide with 2-methylcyclopentanone, dehydration of the resulting carbinol by distillation from p-toluene-sulphonic acid monohydrate, and hydrogenation of the dimethylcyclopentene on 5 %platinum-on-charcoal (Baker) at 200 atm of hydrogen.The resulting mixture was separatedinto cis- and trans-l,2-dimethylcyclopentane by preparative gas chromatography on DowCorning Silicone 200. The cis contained 0.35 % trans and 0-15 % of an unidentified impurity.The impurity in the trans amounted to 1.4% and consisted of cis and small amounts of both1,3-dimethylcyclopentanes. The impurities did not interfere or could be corrected for.We are indebted to Dr. J. A. Marshall and Mr.C . J. V. Scanio for the ethylene ketal ofbicyclo[3.3.1]2-nonen-9-one which was converted to bicyclononane.1TABLE 1 .-ISOTOPIC DISTRIBUTION PATTERNSrun no.temp. "CD2 flowPHC, torrdPD~, torrDO %D1%0 2 %D3 %0 4 %D5 %0 7 %D9 %DlO %Dll %Dl2 %0 6 %0 8 %Dl3 %Dl4 %trans94-51 70.3811.1210-446-311-249-202-3 14-066-042a069,068-1 10a4031.701cis transIlIa 686048.87068029.498 94.510.091 0.35-410 1-135*381 0.445,442 -310*460 -250~536 a1950756 -3401 *078 so701.154 -0401 -500 -0701 -740 -0602-5 13 *0801 1 -772 -05049.669 2-10cis0.000.1 1-56-5 1-60*61-71-971 -461 a491 -992.232.072.1 184-58corr. obs.trans cisIIIa 660.00a 1 1-35-43-58*7 1*921 0321.711 *772.182.743.004.3379.9796-380.235e l 5 5-340-160el65-1 10-1 10-145-070-165-430-040-1 101.39VII 13750397528f98- 1340.538-108-198-105-121-080-080-291-034-223-009-09 1-000VII 1395014.4 g8f97.765~432-05 1-120-080-105*08 8*058-384a086-447-050-334-000100a catalysts IIIa and VII, ref.(1). b The listed material was the feed. The product was separatedinto cis and trans and analyzed by mass spectroscopy. C deuterium flow rate in mmolelh. d partialpressures of deuterium and hydrocarbon in torr. e It was assumed that all DO of epimeric product wasimpurity in original feed. Accordingly, DO was set equal zero.fupper limit. An additional flowof 93-7 mmole/h of helium was presentR. L. BURWELL AND K. SCHRAGE 217RESULTSIn table 1 we give isotopic distribution patterns for the products of runs with bi-cyclononane at 50°C and two different pressures on catalyst VII and we give patternsfor cis- and trans- 1,2-dirnethylcyclopentane at 60°C on catalyst IIIa corrected forinitial content in the product of epimerization and for isotopic dilution. We alsoshow the original analysis for the run with trans-dimethylcyclopentane. We employDi to mean % cycloalkane-df. Fig. 1 presents distribution patterns obtained onnumber of deuterium atoms exchangedFIG. 1 .-Isotopic exchange between deuterium and cis-l,2-dimethylcyclopentane at 50°C ; partialpressure of hydrocarbon = 30 torr; 13.5 g of catalyst V.The data are normalized to perdeutero =unity and represent initial exchange. Total % exchange in recovered cis and flow rates of hydro-carbon in mmole/h were: run 111, 1450 torr deterium, 6-12%, 1.36; run 108, 720 torr, 516% 2.2;run 110,207 torr, 416%, 5-56 ; run 114 with cycIopentane, 570 torr, 1.75 %, 9-55. The hydrocarbonpressure was 110 torr in run 114.catalyst V with cis-l,2-dimethylcyclopentane at 50°C at three different pressures ofdeuterium. The pattern of the trans-product is shown for 720 torr deuterium. In therange D7-0l3, the pattern for trans at 1450 torr is about 25% above and that at 207torr, about 25 % below the 720 torr run. All runs are normalized to 0 1 4 = 1.00 andthey have been corrected for isotopic dilution and initial content in the particularepimer. Runs at 70" gave larger 0 1 4 .For reference, in fig. 1, we also show a runwith cyclopentane on this catalyst.Since exchange of the dimethylcyclopentanes is accompanied by epimerization, theproducts of these runs were separated before mass spectroscopy into cis and trans bygas chromatography using 33 % dimethylsulpholane on firebrick. In analysis using15 % dimethylsulpholane on 60-80 mesh Chromosorb P in a 1/8thin. column 12 m longwith a columntemperature of 25"C, 650 ml/hof nitrogen and a sample size of OaIpl, wefound that exchanged cis- and trans- 1,2-dimethylcycIopentane separated from un-exchanged material. For example, the retention time of exchanged trans was 83min, of unexchanged, 88 min.The exchanged hydrocarbon tails in a way which reflects the isotopic distribution.Exchanged methylcyclopentane and cyclopentane could also be separated but, parti-cularly with cyclopentane, not so well.Thus, gas chromatography could be used tocheck mass spectroscopy to within the experimental error. Others have observe21 8 ISOTOPIC EXCHANGE O N PALLADIUMseparations of deuterated and undeuterated hydrocarbons and that the former havethe smaller retention time.5In trying to determine which exchange processes of dimethylcyclopentane correspondto those of cyclopentane, we have plotted the logarithms of summed values of certainDi's against P D ~ to determine the kinetic order in deuterium of the various reactions.The unnormalized data of fig.1 with the addition of a run at P D ~ = 100 torr wereused. All exchange processes are inhibited by deuterium. For D12-D14, iz = -0-87.This should correspond to process (i) of the introduction although, in comparing actualvalues of n, one should remember that the kinetic orders listed in the introduction weremeasured on another catalyst. Since the ratio, (Dla-D14)/trans, is constant to withinour experimental error, the epimerization has the same kinetic order. For 0 8 - 0 1 1(reaction (iii)), n = - 0.5. Good fits to the log-log plot were obtained. D1- 0 7 gavea poor fit with a value of n of about - 0.35. As is evident from fig. 1, D1 declines withincreasing P D ~ even more slowly than 0 2 - 0 7 .TABLE 2.-EPIMERIZATION AGAINST EXCHANGE IN 1,2-DIMET€IYLCYCLOPENTANESrtlnno.IIIa66IIIa68TIIa65ITIa67-09v110V108Vllltemp.OC6060707050505050pD27107107107101002077201450cis- D12- D i p cis btrans trans-Ds-Dy40.26019-304 5-30-350.20-24a for runs with cis-l,2-dimethylcyclopentane as feed, cis-D12-D14 divided by total trans formed.b for rum with trans-l,2-dimethylcyclopentane, trans-D4-D14 divided into total cis formed.In table 2, we compare the yield of epimerized dimethylcyclopentane with highlyexchanged but unepimerized.With catalyst IIIa we have values for both 1,2-di-methylcyclopentanes as starting materials, with V, only for cis. As highly exchangedtrans, we take ds-dl4; as highly exchanged cis, d12-di4, i.e., molecules in which un-equivocally hydrogen atoms from more than one set have been exchanged.We definea set as those hydrogen atoms which can be interconnected by cis-eclipsed con-formations.3 As shown in fig. 2a, cis-l,2-dirnethylcyclopentane has a set of 3 (shownas h) and a set of 11 (shown as H). The trans has two identical sets of 7 each.The ratio, (cis-012-014)/trans, is larger with catalyst V than with IIIa. We do notknow why. However, the precision in these experiments is not high. Also, we maymiss some exchanged cis which ought to be counted. Some cis-&-dlO may representmaterial in which hydrogen atoms from more than one set are involved. This will beless serious in runs with trans since a wider range of exchanged species is reckoned ashighly exchanged than with cis, Ds'D14 against 012-014.Further, catalyst IIIa hasrelatively larger numbers of sites giving two-set exchange.1 Thus, on IIIa, cis- D11-Oi4is a larger fraction of exchanged cis: run IIId66,43 %, run V108,28 %.MECHANISM OF ONE-SET EXCHANGEThe results of isotopic exchange with bicyclo[3.3.1 Jnonane, table 1, are inconsis-tent with the recent suggestion that isotopic exchange occurs via alternation betweenmonoadsorbed alkane and olefin n-bonded to a single metal atom.6p7 The rings inbicyclononane, fig. 2b, are large enough so that prohibition of eclipsed conformationR. L. BURWELL A N D K. SCHRAGE 219at a bridgehead has relaxed. Fig 2b shows one cyclohexane ring in the boat form inwhich H(1) arid H(2) are eclipsed.By twisting, H(2) and H(3) can become eclipsed.Thus, all eight H-atoms can exchange by alternation between eclipsed di-adsorbedand mono-adsorbed. Although the bridgehead bicyclononene might well be isolated,it is of much too high an energy to serve as an intermediate when n-complexed.8\/"acis-1,ZDimethylcyclopentane Bicyclo[ 3.3.1 lnonaneThus, the eclipsed diadsorbed intermediate has the correct geometry and the olefinn-complex does not. Further, we suspect that the C-* bond more commonly repre-sents binding between carbon and two or more surface atoms rather than a single one.MECHANISM OF TWO-SET EXCHANGEIn the experiments with dimethylcyclopentane, we have been able to examine thedegree of isotopic exchange, both in the separated epimer and in the unisomerizedreactant. Thus, the connection between epimerization and exchange appears moreclearly than in former experiments where only the total mixture of reactant and epimer(or racemized product) could be examined mass spectrographically.3~ 9 In isotopicexchange of dimethylcyclopentanes the highly exchanged reactant and the epimerformed during reaction appear to be derived from a common intermediate whichforms trans- and cis-1,2-dimethylcyclopentane in the ratio of three or four to one.This conclusion is supported by the data of table 2 on catalyst IIIa.It accords withthe observation that the patterns for cis and trans in runs with cis as feed which havebeen normalized to 0 1 4 = 1-00 nearly coincide at D12 and 013.This is more strikingwhen trans is the feed as follows from run 68 in table 1. Here normalized cis epimerand trans reactant nearly coincide beyond D,. The division of the common inter-mediate into cis and trans products should be independent of the partial pressure ofdeuterium and, as shown in table 2, the ratio of epimer to highly exchanged reactant is,indeed, independent of pressure.In the runs on catalyst V, D12 + Dl3 + 0 1 4 for the trans product from cis-dimethyl-pentane reactant decreased from 91 % at P D ~ = 207 torr to 82% at 1450 torr. Avery large fraction of epimerization is associated with extensive multiple exchange,yet not all. The remaining trans is distributed in quantities which rise slowly andmonotropically from D1. We do not understand the origin of this material and we sus-psct that it originates from still another process and, perhaps, still other types of sites.The early suggestions as to the mechanism of two-set exchange involved adding tothe Horiuti-Polyani mechanism a symmetric species which could react to form speciesabsorbed either in the h-set or in the H-set.As applied to cyclopentane, the inter-mediate served as a bridge between the two sides of the ring. Applicable inter-mediates were 1,l -di-adsorbed cyclopentane 2 and 1 -mono-adsorbed cyclopentene.220 ISOTOPIC EXCHANGE ON PALLADIUMWe have argued that one-set and two-set exchange involve different sets of sites1 butthere is an independent argument that no dissociatively adsorbed cyclopentane canserve as a symmetric intermediate to give large values of D5 and Dlo (as in fig.1)on one set of sites. Consider 1,l-diadsorbed cyclopentane as an example. As firstformed, it will be Cs(H4*)(dd*) in which one hydrogen atom has been removed fromthe, as yet, unexchanged side. This can revert with equal probability to Cs(H4D)(d4*)and C5(H4*)(d5). If 0 s is to kept large, 0 6 must also be large, which is contrary to fact.Consider a set of sites which promote alternation between mono- and di-adsorbedcyclopentane and upon which there is a small probability of roll-over of the eclipseddi-adsorbed species as shown in fig. 3. Such a set of sites would form mainly d5 anddg since the two hydrogen atoms attached to the carbon atoms bonded to the surfacewould be transferred during roll-over to the set of hydrogen atoms away from thesurface.Such a process may well be the origin of &exchange. Two or more roll-overs could lead to the formation of dlo but D5 and Dlo could not both be large. 1-Mono-adsorbed cyclopentene has the defects of both examples discussed above. Ifits formation is of low probability, the new species are ds and dg, not d10. In fact, nosymmetric intermediate formed dissociatively from cyclopentane can give large 0 5 andDlo on one set of sites.n II * * p\* *; '\*h* *FIG. 3.-RoU-over mechanism.However, most of species discussed above can account for large D5 and Dlo if theprobability of their formation is zero on one set of sites (which form mainly ds) andlarge on another (which form, then, mainly 40).We may eliminate 1,l-di-adsorbedcyclopentane as not providing epimerization but we need further data to select among1 -mono-adsorbed cyclopentene, roll-over, and a mechanism which has been particu-larly supported by Rooney.69 79 9 As shown in fig. 4, this mechanism does not in-volve the transfer of the carbon-surface bond from one side of the ring to the other.Rather, it involves two new species, ncomplexed olefin and n-ally1 which interconvertby two methods in one of which the hydrogen is added or removed from below theplane of the ring, in the other, from above.We believe that we havecorrectly identified the hydrogen atoms which exchange to form dg. What are thetwo pairs which exchange as units to give d~o and &? 1,1,3,3-Tetramethylcyclo-hexane has an isolated trimethylene unit as part of a ring.At lower temperatures itexchanges the three hydrogen atoms predicted by the conventional Horiuti-Polanyimechanism and two more, (not three).6 Bicyclononane has two trimethylene units.We suggest, therefore, that the additionally exchangeable pairs are g,g and h,h of fig.2b. Further, whatever process gives two-set exchange (here multi-set exchange) mustbe substantially contained within the trimethylene unit. Otherwise, Dlo could not becomparable in size to 012.These considerations are inconsistent with alternation between di-adsorbed alkaneand mono-adsorbed alkene,3 but consistent with the n-alkene-n-ally1 model. How-ever, one may object that it is difficult to envision an adequately low activation energyThe results with bicyclononane in table 1 are useful hereR.L. BURWELL AND K. SCHRAGE 22 1for step b of fig. 4. Because the bonds involving hydrogen are short, the H-H bondwould have to be stretched considerably before much binding in the sense of theproduct can occur. This objection may be obviated in two ways. First, the processmight occur at a surface step, for example, at the step of a screw dislocation. In sucha process, the addition of hydrogen from the top would not be sterically so difficult.FIG. 4.-~-Allyl mechanism.Indeed, one might drop the reaction which involves the hydrogen molecule and addatoms both from above and below as shown in fig. 4c. Alternatively, followingSiegel, one might substitute formation of 3,3-di-adsorbed alkene 10 for step b of fig.4.Either of the following sequences would accommodate the data.H H H H H H HC---C=C- + - C - ~ C ~ C - - + -Cz~--C--- (1)II I 11H H H H H H H HC=C- + H C=C- + -C=C H + -C=C\ / (2)C\ I /C\ I /C\ /CIn some molecular examples, the a-ally1 of the second step is known to isomerize asshown.11 Although a model involving 3,3-di-adsorbed alkene accords with the iso-topic results, one might object on kinetic grounds. Increase in the partial pressureof deuterium inhibits all of the exchange processes and with dimethylcyclopentanesand bicyclononane (table 1) it inhibits two-step exchange most of all. However, thedifference in kinetic orders between one- and two-step exchange is only about 0.4.It is difficult to believe that a reaction depending upon an intermediate with twohydrogen atoms fewer than di-adsorbed alkane should be so little relatively inhibitedby hydrogen.Further, cyclopentane itself shows a much smaller difference betweenthe kinetic orders of one-set and two-set exchange on catalyst V. A possible explana-tion of the lesser difference is that more free sites are needed for the larger molecules,not for bonding, but just to prevent steric interference during that process whichgives two-set exchange.In an isddted trimethylene unit, neither repeated alternation between n-alkene andn-ally1 nor repeated roll-over can exchange one of the hydrogen atoms on the middlecarbon atom. To get “ isolation ” with bicyclononane in the roll-over mechanism,one must assume that two-set exchange occurs consequent to migration of materialhighly exchanged on one-set sites (predominantly to ds with bicyclononane) to two-set sites and that the relative stability of bridgehead di-adsorbed is different on the tw222 ISOTOPIC EXCHANGE ON PALLADIUMsites. On one-set sites, as we have seen, desorption is slow compared to the rateof formation of bridgehead diadsorbed.On two set sites, to contain a trimethyleneunit, we must assume that formation of bridgehead di-adsorbed is comparable in rateto desorption.Let us convert the transition state of the roll-over mechanism into an intermediateas shown in fig. 5. Then the difference between ease of formation of bridgehead di-adsorbed on one- and two-set sites is not a necessary assumption.If R is a hydrogenatom, reiteration of the process will exchange all hydrogen atoms in cyclopentane orfive of six in an isolated trimethylene unit. If R is alkyl, exchange is accompanied byepimerization or racemization.*FIG. 5.-Roll-over mechanism with five-co-ordinate species as an intermediate.In fig. 5, the coordination is approximately that of a trigonal bipyramid but itmight instead be represented as in the dimer of aluminum trimethyl.12 The troughsin the (1 10) faces of f.c.c. metals might be a suitable location for a transition state oran intermediate of the roll-over type. Or, the two rows of sites might be on oppositefaces of a step. Further, one can imagine possible structures of this type at projectingedges. Such locations might be favoured by the low coordination number of themetal atoms at such sites.In conclusion, we note that much may be learned about the mechanism of surfacereactions by suitable studies of the consequences of isotopic tracer studies on a seriesof molecules of appropriate structure particularly when the reactions which are studiedhave useful stereochemical consequences. However, complete specification of mech-anism will be difficult on the basis of such studies alone. We need to know much moreabout the detailed geometry of surfaces and about the nature of bonding to metallicsurfaces.This work was supported by the Petroleum Research Fund of the AmericanChemical Society.1 Schrage and Burwell, 1966,SS. J. Amer. Chem. SOC.2 Anderson and Kemball, Proc. Roy. SOC. A , 1954,226,472.3 Burwell, Shim and Rowlinson, J. Amer. Chem. SOC., 1957,79,5142.4 Burwell and Schrage, J. Amer. Chem. Soc., 1965,87, 5253.5 Falconer and CvetanoviC, Anal. Chem., 1962,34, 1064.6 Rooney and Webb, J. Catalysis, 1964, 3,488. Rooney, J. Catalysis, 1963, 2, 53.7 Bond and Wells, A h . CataZysis, 1964, 15, 91.8 Eliel, Stereochemistry of Carbon Compounds, (McGraw-Hill Book Company, New York, 1962).9 Gault, Rooney and Kemball,J. CataZysis, 1962,1,255.10 Siegel, A h . Catalysis, 1965,16, in press.11 Powell, Robinson and Shaw, Chem. Comm., 1965,78.12 Wells, Structural Inorganic Chemistry, 3rd ed., (Oxford University Press, London, 1962). p. 757.p. 301

ISSN:0366-9033    

DOI:10.1039/DF9664100215

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Nature and Reactivity of Intermediates in Hydrogenation ofButa-l,3-Diene Catalyzed by Cobalt and Palladium-GoldAlloysBY B. J. JOICE, J. J. ROONEY,” P. B. WELLS AND G. R. WILSONDept. of Chemistry, The University, HullReceived 3 1st Jdnudry, I966The gas-phase hydrogenation of buta-1,3-diene has been studied in a static system using alumina-supported cobalt, a series of cobalt powders, and pumicesupported palladium-gold alloys as catalysts.But-1-ene and but-2-enes, which were formed by 1,2- and 1,4-addition of hydrogen respectively toadsorbed diene, were initial products under all conditions, and yields of n-butane were either very lowor zero. The distributions of butenes were very dependent on the physical nature of the cobaltcatalysts. Well-sintered metal, consisting of large particles, gave selective formation of trans-but-2-ene, whereas, cobalt prepared in a finely divided state at lower temperatures was more active per unitsurface area, but gave high yields of but-l-ene and much lower trans to cis ratios in the but-2-enes.These variations are attributed to differences in the catalytic properties of different crystal faces andthe change in reaction mechanism is believed to be mainly due to a change in the ease of formationand stability of 1-methyl-n-ally1 intermediates.Stable n-allylic complexes are formed readily on the surface of palladium and various amountsof gold in the alloys have a small effect on the mechanism of buta-1,3-diene hydrogenation.Thebut-1-ene yields and activation energies exhibited sharp maxima which coincided in the 60-75 % goldrange and an explanation in terms of variations in the nature of the palladium &orbitals is attempted.The temperature dependencies of the product distributions were influenced between 80 and 130°Cby the p- to a-phase change of the hydrides of palladium and palladium-rich alloys and the effect isdiscussed.The concept that the chemisorption bond may, in appropriate cases, have n-character has been introduced in recent years to interpret several features of hydro-carbon reactions with hydrogen or deuterium on transition metal catalysts.Themajor distinction between this and previous theories is that multi-centred bondingbetween hydrocarbon species and individual metal atoms in the surface is considered.Thus, reactions such as dehydrogenation and hydrogenation of paraffins, olefins,diolefins, acetylenes, aromatic hydrocarbons, etc., may occur by the interconversionof a variety of 6- and 7c-bonded complexes.*-4 The n-complexes may include boththe metal-olefin and metal-arene types.Besides providing mechanistic insights, this theory leads to important generalconclusions about surface reactions.First, catalytic activity and selectivity areconsidered to be functions of the chemical properties of individual surface atomsrather than functions of the bulk properties of the metals. Secondly, there may be acorrelation between metal-adsorbate bonding on the one hand, and metal-ligandbonding in organometallic compounds on the other ; a preliminary attempt at sucha correlation has shown some success.4 Thirdly, the role of the geometric factor incatalysis is given a new significance.Because of metal-metal bonding the arrangementand number of atoms which are nearest neighbours to a given surface atom influencethe energy levels and electron occupation of the valency-shell atomic orbitals of thelatter, and thus its bonding properties, as well as determining the number of co-ordinating positions available for the intermediates in a catalytic reaction. A view,* present address : Dept. of Chemistry, David Keir Building, Stranmillis Road, Belfast, 9.22224 HYDROGENATION OF BUTADIENEpreviously held, that metal-metal spacings and thus the type of crystal face exposedplay a vital role because only some of these match the geometrical requirements formultiple a-bonding of adsorbates may be of limited validity.The present theory also proposes that interpretation of catalytic behaviour mustultimately take into account, as does co-ordination chemistry, fundamental electronicproperties which influence the nature and strength of metal-ligand bonds.Theseproperties include, among others, the number of electrons in the valency shell of theisolated atom, the effective nuclear charge, d+s and d+p promotion energies andionization potentials. Such properties distinguish the co-ordination chemistry, andpossibly to some extent, the catalytic behaviour of one metal from that of another.Progress in heterogeneous catalysis is limited by uncertainties concerning thenature of metal-metal bonds in solids and by ignorance of the precise arrangements ofatoms which constitute active centres in catalysts.Therefore, a fruitful approach inthis field at the present time may be one in which comparisons are made of the forma-tion, stability and reactivity of the same ligands in organometallic complexes and inhomogeneously and heterogeneously catalyzed reactions, the comparisons being madefor as wide a range of metals as possible. By using co-ordination chemistry as aguide, a clearer picture of some of the fundamental factors governing heterogeneousreactions should emerge.The hydrogenation of buta-l,3-diene (subsequently referred to as butadiene) is aparticularly suitable reaction to test the validity of this approach.This is the simplestconjugated diene and it has been used to prepare a variety of organometallic compoundswhich exhibit several types of CT- and n-bonds. Furthermore, something is alreadyknown of the properties of all the group 8 metals as catalysts for the heterogeneoushydrogenation of this compound,S and homogeneous hydrogenation of butadiene andother dienes, catalyzed by some transition metal complexes is being studied byseveral workers.697There are several reports of the liquid phase hydrogenation of butadiene,s-11with nickel, palladium and platinum as catalysts and ethanol as solvent, but only onereport 11 contained a discussion of mechanism which was based on an ionic model.In 1963, Meyer and Burwell reported details of the gas-phase reaction of butadienewith deuterium using a palladium-alumina catalyst.This was followed by a moredetailed study by Bond, Wells et d 4 ~ 5 9 13-15 of the catalytic activities of each of thegroup 8 metals and copper €or the gas-phase reaction of butadiene with hydrogen;deuterium was also used when examining cobalt, nickel, copper, palladium andplatinum.14, 15Iron, cobalt, nickel, copper and palladium are the only metals which selectivelyhydrogenate butadiene to butenes.5,14 The yield of n-butane in the initial productsis zero for iron, cobalt and copper, 0-3 % for nickel and palladium, and considerablefor all the other group 8 metals.4, 5 , 13 The butene distributions obtained by selectivehydrogenation are far removed from the thermodynamic equilibrium compositionsand they remain constant, or nearly so, until the near removal of butadiene.Theyare also independent of the initial hydrogen pressure. These features suggest thatbutene desorption occurs in preference to isomerization, i.e., the butenes are formedentirely by 1,2- and 1,4-addition of hydrogen, a conclusion which is supported byisotopic studies. The reaction of butadiene with deuterium using cobalt, nickel andcopper,ls and palladium 14 has revealed a close similarity in the deuterium contentsand distributions in each of the three n-butenes.Some typical butene distributions for alumina-supported cobalt, nickel, copperand palladium catalysts are given in table 1, which shows two important features.First, the importance of 1,4-addition varies widely (from 15 % over copper to 70 % oveB.J. JOICE, J. J. ROONEY, P. B. WELLS AND G. R. WILSON 225cobalt), and secondly, the trans/& ratio in the but-2-enes is much higher for cobaltand palladium ( > equilibrium) than for nickel and copper ( < equilibrium).In the adsorbed state a butadiene molecule must take up one of two possibleconformations.CH2=CH CH-CH// \\CH2 CH2\CH=CH2I 11Conformational interconversion may or may not occur and the same is true of thehalf-hydrogenated state, i.e., adsorbed C4H7 species. For butadiene molecules in thegas phase, rotation about the central carbon-carbon bond occurs but is hindered bysteric repulsion of the vinyl groups in conformation 11. Smith and Massingill 16have estimated the ratio I : I1 to be 95 : 5 approximately at ambient temperatures.Thus, at the instant that adsorption occurs the population of I is expected to exceedthat of 11 on the surface.TABLE 1 .-TYPICAL INITIAL B UTENE DISTRIBUTIONS OBTAINED FROM BUTADIENE HYDROGENATIONIN A STATIC SYSTEM USING ALUMINA-SUPPORTED COBALT, NICKEL,COPPER AND PALLADIUM CATALYSTSInitial butadiene pressure = 50 mm ; initial hydrogen pressure = 150 mm ;initial butane yield = zero.temp.butene composition (%)("C) B-1 t-B-2 C-B-2 metalc o 125 29 65 6Ni 77 49 34 17c u 100 85 6 9Pd 21 60-2 3 7.0 2.8Pd 18 68-4 30.1 1.7Co and Ni reduced at 400" and 250" respectively.When 1,4- addition produces a high yield of trans-but-Zene compared to cis-but-2-ene we propose that the nature of chemisorption is such that (i) the conformationsof adsorbed C4H7 do not interconvert and (ii) for adsorbed diene either the conforma-tions are non-interconvertible, or they interconvert but maintain an equilibriumsimilar to that which is operative in the gas phase.The notation for the adsorbedspecies is shown in fig. 1 . Table 1 shows some results for palladium, obtained byA0.orb.t w I I - I - DaIhyI -TT- d l y l (- 5%)FIG. 1 .-Notation suggested for the adsorbed species in butadiene hydrogenation when conforma-tional interconversion of the half-hydrogenated state does not occur.Leszczynski and Wells where translcis ratios in the but-Zenes approach closely thevalue of 19, which would signify virtually complete non-interconvertibility of theconformations on this metal at room temperature226 HYDROGENATION OF BUTADIENEAt the other extreme a trans/cis ratio of about unity signifies that conformationalinterconversion of adsorbed C4H7 and perhaps C4H6 occurs readily.The mechanismshown in fig. 2 allows this possibility to the extent that the key a-bonded C4H7 species,CH3--CH--CH=CH2, is formed during reaction. The two 1 -methyl-n-ally1 com-plexes are believed to be less stable in this case and are either true intermediatesor represent a more transient state attained by C4H7 species before the addition ofthe second hydrogen atom.I8-1/CH./ n cn2 ,/ =;+.c;=,c*. =- cr' -ig'c"-c*, H, I-'-',,'! . * . cn,= C M c H;= c nw 1 \cu-c";,,:;cHf" -- 'FC", w_ C H , / a'" - c " 1 'C",' ,'''I * \\ 1 ).:2 !I ',%\? C H = C Y , ---+j <\ ' CY,C" c*=c*,-K- 8-111 :W - c*<' 4 Cv\cn, - <-'-' ..;. I: 1".... '.+. &U.-c** II - I C H I ~ 'cn, --- 1FIG. 2.-Notation suggested for the adsorbed species in butadiene hydrogenation when conforma-tional interconversion of the half-hydrogenated state occurs.Two points are noted. Addition of the first hydrogen atom to adsorbed butadieneis believed to occur exclusively at a terminal carbon atom in agreement with the generalbehaviour of this compound in addition reactions. Secondly, for those butenylspecies other than n-allylic complexes, addition of the second hydrogen atom takesplace to the carbon atom which is a-bonded to the surface.It follows that if l-methyl-n-allylic complexes are not formed during hydrogenation only 1 ,Zaddition of hydrogencan occur.Comparisons of fig. 1 and 2 with table 1 suggest strongly that x-allylic complexesare formed readily on cobalt and palladium surfaces but not on those of nickel andcopper. This concurs with the organometallic properties of these metals 17 and withthe previous finding that palladium has outstanding ability to form n-allylic complexesduring the catalyzed exchange of cycloalkanes with deuterium.192 Our contentionis therefore that the compositions of the product butenes yield information about themode of chemisorption of their precursors.We have successfully modified cobalt and palladium catalysts in that their stereo-selectivity for the formation of trans-but-2-ene has been diminished.Modificationhas been achieved (i) by varying the conditions of preparation of the cobalt catalystsand (ii) by alloying palladium with gold. These features and other accompanyingfeatures are discussed in terms of the mechanisms described above.EX PER1 M EN'T ALCATALYSTSA stock of alumina-supported cobalt catalysts containing 10 % (wt/wt) metal was pre-pared by depositing the required amount of A.R. cobalt nitrate by evaporation of an aqueoussolution in which the support (Peter §pence type A alumina) was immersed. The drymaterial was then calcined, followed by reduction of metal oxide under 300 mm of hydrogenin a static system at 329°C. The hydrogen was changed after 6 h and the reduction con-tinued for a further 14h.Samples from the stock were reduced again at 200°C under100 mm of hydrogen for 2 h before being used.Cobalt powder P1 was prepared by the reduction of an aqueous solution of cobalt sulphatB. J. JOICE, J. J. ROONEY, P. B. WELLS AND G. R. WILSON 227by hypophosporous acid. The procedure, developed by Holbrook and Marsh,l8 consistedof adding hypophosphorous acid to the metal ions in a boric acid+ sodium hydroxide buffersolution at pH9.0. Hydrogen was evolved after heating to 90°C for 20min and a blackprecipitate of cobalt deposited. Spectrophotometric analysis showed a phosporus contentof 2 %.Cobalt powders P2-PS were prepared by reduction of grey cobalt oxide (B.D.H. Ltd.)in a stream of hydrogen for 24 h at elevated temperatures.X-ray powder photographswere obtained for all the powders, surface areas of P2, P4 and P8 were measured by theB.E.T. method, and P2 and P8 were examined by electron microscopy.Stocks of pumice-supported palladium, gold, and palladium-gold alloys were preparedby deposition of palladium and gold chlorides (Johnson, Matthey and Co. Ltd.) from acidsolution on 18-30 B.S.S. mesh pumice (Hopkin and Williams Ltd.). After evaporation todryness, the salts were reduced to metal in a stream of hydrogen for 24 h at room temperature,then at 200°C for 2 h, followed by further reduction for 3 h under 100 mm of hydrogen at300°C in a static system. Finally, the preparations were heated in cacuo for 3 h at 450°C.Samples from the stock were reduced again at 200°C under 100 mm of hydrogen for 2 hbefore being used.These catalysts contained 5 % (wtlwt) metal.APPARATUS, MATERIALS AND EXPERIMENTAL METHODSThe apparatus consisted of a conventional high vacuum system. Each catalyst samplerested on the bottom of a cylindrical 100 ml Pyrex reaction vessel and pressure changes duringreaction were measured manometrically.Butadiene (Distillers Company Ltd.) contained no detectable impurities either by G.L.C.or mass-spectrometric analysis.Hydrogen was purified by passage over a Pt-silica catalyst at 300°C and then dried, or bydiffusion through a palladium thimble.Gas samples withdrawn from the reaction vessel were analyzed using a G.L.C. unitcontaining a 25 ft column of 40 % (wtlwt) hexane-2,5-dione supported on 30-60 B.S.S.meshfirebrick. The column was operated at room temperature with nitrogen as carrier gas,and the detector was a katharometer.Mass-spectrometric analyses were obtained using a modified A.E.I. M.S.3 spectrometer,with an electron beam energy of 12.0 eV.RESULTSCOBALT CATALYSTSThe distributions of isomeric n-butenes obtained from butadiene hydrogenationwere dependent on the previous temperature treatment of the catalyst. For example,when cobalt-alumina was prepared by reduction of the supported oxide by deuteriumat 329"C, subsequent hydrogenation of diene at 144°C gave but-1-ene as the mostabundant isorner, whereas, a catalyst sample which had received further treatmentat 414°C under 200 mm deuterium for 20 h, gave trans-but-2-ene as the major product.The following results are typical (reaction temp.= 144°C ; initial PD2/PC4H6 =I= 211 ;analyses after 20 % removal of diene).I________ ___ reduction products (%)temp. ("C) B-1 t-B-2 C-B-2 n-butanereaction A 329 51 31 17 1reaction B 414 28 56 16 0The yields of cis-but-2-me and n-butane were only slightly influenced by the moredrastic treatment. Both types of butene distribution were independent of conversionup to 60 % removal of butadiene. In reaction A the average number of deuteriumatoms present in each butene was 2-00+0*01 ; it was also observed that the '' hydroge228 HYDROGENATION OF BUTADIENEexchange ” reaction [Dz + H (ads) +€ID + D (ads) J and the butadiene exchangereaction [C4H6 + D (ads) +C4H5D + M (ads)] occurred at similar rates.In reaction Bthe average deuterium content of each butene was 1.39f:O-Ol but here the rate ofbutadiene exchange exceeded the rate of “ hydrogen ” exchange.Orders of reaction measured before and after the treatment at 414°C were similar,being zero or slightly negative in butadiene and 1-0-1-5 in hydrogen. Apparentactivation energies were 12-2* 1.2 kcal mole-1 (94-153°C) and 8.9 f 1.0 kcal mole-1(100-1 87°C) before and after this treatment respectively.These results suggest either that a change in the physical structure of the cobaltparticles accompanied the treatment at 414°C or that a change in the support hadoccurred which altered the properties of the metal. To decide between these alterna-tives the catalytic activity of unsupported cobalt, in powder form, was investigated asa function of preparation temperature.P 2 P3 P4PSPb p7 p 0I Typo B Bchoviou200 4 0 0 6 0 0reduction temp.(“C)FIG. 3.-Butene distributions and selectivities (S = C~HS/(C~H~+ C4H10)) for the hydrogenation ofbutadiene at 110°C over a series of cobalt powders Pl-Ps, prepared at various reduction temperatures.Initial butadiene pressure = 50 mm ; initial hydrogen pressure = 100 mm ; analyses after 12f 1 %removal of butadiene.S, 0 ; but-1-ene, 0 ; trans-but-2-ene, A ; cis-but-2-ene, 0.The effect of reduction temperature on the physical characteristics of a series ofcobalt powders is shown in table 2 and in plate 1, the products obtained from butadienehydrogenation at 110°C are shown in fig.3. The powders produced at lower tem-peratures were also more active per unit weight or per unit surface area than thoseproduced at higher temperatures.Powders Pl-P3, which catalyzed the preferential formation of but-1-ene (type Abehaviour), had relatively high surface areas and consisted of small particles of metalwhich were mostly or completely in the a-phase (c.p.h.) ; micrograph (a) in plate 1shows the presence of large numbers of particles below 50 A in size.Powders P5-P8, which catalyzed the preferential formation of trans-but-2-ene(type €3 behaviour), were sintered ; their surface areas were relatively low and electronmicrographs revealed an almost complete absence of small particles (micrograph (b)is typical). The metal was mostly or completely in the P-phase (f.c.c.).Powder P4PLATE 1.-Typical electron micrographs of the edges of cobalt particle clusters : (a) powder P2prepared at 298°C ; (b) powder P8 prepared at 580°C ; width of micrographs = 0.75 p.[To face page 228B. J. JOICE, J. J. ROONEY, P. B. WELLS AND G. R. WILSON 229reduced at intermediate temperatures, exhibited intermediate physical characteristicsand catalytic behaviour. Fig. 3 shows, as for cobalt-alumina, that the yields of cis-but-2-ene and n-butane are not greatly influenced by the reduction temperature.Behaviour similar to that shown in fig. 3 has also been observed using nickelpowders where there is no change of phase as the reduction temperature is increased.TABLE 2.-PHYSICAL CHARACTERISTICS OF COBALT POWDERSP1 p2 p3 P4 PS p6 P7 P8reductiontemperature(’C) 97 298 324 415 443 467 514 580weight used (g) 0.263 0.034 0.021 0.183 0.202 0.176 0.277 0.303surface area (m2 g-1) 144 4-8 0.9phase of cobalt pure a pure a pure *a mixture p p purep pure p* some cobalt oxide present.a+P (trace a) (trace a)PALLADIUM-GOLD ALLOYSThe initial mixture of reactants used throughout this work consisted of 100 mmof hydrogen and 50mm of butadiene unless otherwise stated.Although freshlyprepared catalysts were highly active they were somewhat unstable but after about5 runs at room temperature the rate of reaction had declined to a steady value and theproduct distributions became reproducible and virtually independent of conversionup to some 80 % removal of butadiene.The selectivity for butene formation wasvery high on all the catalysts and only decreased slightly with conversion. Somebutene distributions and selectivity values for low and high conversions on threecatalysts of widely different gold content are given in table 3.TABLE 3 .-DEPENDENCE OF BUTENE DISTRIBUTIONS AND SELECTIVES ONCONVERSION ON PALLADIUM-GOLD ALLOYStemp. Pd conversion butenes (%)(“0 ( %I ( %) B-1 t-B-2 C-B-2 S19 100 22.4 48.1 47.5 4.4 1 -00019 100 76.4 48.8 46.2 5.3 0.96720 50 17.9 47-6 46-9 6-3 1.00020 50 85.8 48.1 44.9 7.0 0.98 119 25 20.5 49.4 43.5 7.2 0.99119 25 74.0 49.0 43.3 7.6 0.989Orders of reaction were determined from initial rates, the initial pressure of onereactant being held constant at 50mm while the pressure of the other was variedbetween 50 and 200 mm.Orders in hydrogen were between 0-7 and 0-9 while theorder in butadiene was zero in all cases. During these estimations of order inhydrogen the products were analyzed at 50 % conversion and found to be virtuallyindependent of hydrogen pressure.Activation energies were evaluated from initial rates in the range 048°C but,because of its low activity, the value for the 95 % gold alloy (all compositions areexpressed as atomic percentages) was obtained for a slightly higher, though over-lapping temperature range. A 100 % gold catalyst was almost completely inactiveat 200°C. The activation energies and corresponding initial yields of but-1-ene inthe total butenes are given in table 4.While the activation energies did not vary within experimental error between 0 and60 % gold a sharp increase of -4.5 kcal mole-1 was noted when the gold content230 HYDROGENATION OF BUTADIENEreached 65 %.At 70 % gold the value was still high but on increasing to 75 % asharp drop of -5.0 kcal mole-1 was found and the lowest activation energy was thatfor the 95 % alloy. A corresponding maximum in the yields of but-1-ene was alsoobserved in the 60-70 % gold range.TABLE 4.-oRDERS OF REACTION, ACTIVATION ENERGIES, AND BUT-1 -ENE YIELDSAT ROOM TEMPERATURE FOR PALLADIUM-GOLD ALLOYSPd( %)1008050403530255order(H2)0.90.80.90.90.70.7--order(C4H6)0.00.00.00.0-0.10.0--E(kcal mole-1)10.510.211.010.314-914.69.48.9B- 1(%I48.555.050.059.561.060.050.051.0The temperature dependencies of the product distributions were then examined byanalyzing reaction mixtures after 50 % conversion at several temperatures between18 and 200°C.Yields of the individual butenes are plotted against temperature forthree catalysts in fig. 4. The major effect of increasing the reaction temperature,4 0 8 0 120 160 100T "CFIG. 4.-Temperature dependence of butene distributions in the hydrogenation of butadiene onpalladium-gold alloys.0 % gold, A ; 65 % gold, B; 75 % gold, C.but-I-ene, 0 ; trans-but-2-ene, @ ; cis-but-Zene, 0.using pure palladium, was an increase in the yield of cis-but-2-ene at the expense of thetrans-isomer, while the yield of but-1-ene remaified almost constant.A feature ofthese results was the sharp dip through a minimum and corresponding rise througha maximum in the yields of but-1-ene and trans-but-Zene respectively between 100and 130°C. Similar results were also obtained with the alloys but as the gold contentincreased the dip and rise in the yields of but-1-ene and trans-but-2-ene became lesB. J. JOICE, J. J. ROONEY, P. B. WELLS AND G. R. WILSON 231sharply defined and broadened over a much wider temperature range. The minimumand maximum in the appropriate curves also shifted to slightly lower temperatures,especially with alloys of higher gold content. The only alloy which did not exhibitthese features was that containing 95 % gold where the yields of but-1-ene and cis-but-2-ene fell and rose in a linear fashion respectively, while the yield of trans-but-2-eneremained virtually constant, as the temperature was raised.Yields of cis-but-2-eneoften rose more steeply in the same temperature range where the dip and rise in thepercentages of the other isomers occurred than at higher or lower temperatures.The yields of but-1-ene were also about 10 % higher over the whole temperaturerange for the 60-70 % gold alloys than for those of higher or lower gold content.Maxima in the yields of n-butane were also found between 90 and 130°C andthese are shown for some of the catalysts in fig. 5. While the maximum was highestfor 100 % palladium it decreased steadily and broadened as the gold content increaseduntil it was barely perceptable, within experimental error, at 75 % gold.4 0 8 0 I 2 0 I60 200T ("C)FIG. 5.-Temperature dependence of n-butane yields in the hydrogenation of butadiene on palladium-gold alloys.0 % gold, 0 ; 20 % gold, A ; 50 % gold, 9 ; 65 % gold, 0.Isothermal plots of the ratios of trans-but-2-ene to cis-but-2-ene as a function ofalloy composition are shown in fig.6. An increase in gold content at a fixed tem-perature resulted in a continuous decline in this ratio over the whole range of com-position. When the temperature was raised at a given composition there was also adecrease in the ratios, the effect being most marked for 100 % palladium, and becoiningless as the alloys became richer in gold.We believe that the peculiar changes in product yields in the temperature range80 to 130°C might be associated with the p- to a-phase change in palladium andpalladium-gold hydrides so hydrogen solubility in these catalysts was briefly studied.A weighed quantity of each catalyst (-2.5 g) was maintained at 200°C in 150 mm ofhydrogen for 2 h, cooled at room temperature, and allowed to equilibrate with freshhydrogen at a pressure of 75 mm.The temperature was then raised by 1°C every2-3 min and the pressures recorded up to 160°C; the sample was then cooled at thesame rate and again the pressures noted. Since the system was effectively one ofconstant volume these pressures, when corrected to s.t.p., were a measure of the hydro-gen evolved or absorbed by the catalysts.A slight evolution of hydrogen from 100 %palladium, 0.016 atoms per palladium atom, was found up to 89"C, but at this tem-perature rapid evolution commenced and finally ceased at 109°C ; the total hydroge232 HYDROGENATION OF BUTADIENEevolved in this range was 0.38 atoms per palladium atom. The alloys behavedsimilarly with slow evolution of a comparable quantity of hydrogen, but in the criticaltemperature range the amount of hydrogen released decreased continuously withincreasing gold content and at 70 % gold or more evolution could not be detected.The temperature range for rapid evolution of hydrogen from the alloys was between72 and 97°C and therefore slightly below that for 100 % palladium. The usualhysteresis effect in the evolution and absorption of hydrogen by palladium and thealloys was also observed.2 0 b 0 I00PdFIG.6.-Dependence of isothermal ratios of trans-but-2-ene to cis-but-2-ene on composition in thehydrogenation of butadiene on palladium-gold alloys.40°C, 0 ; 80°C, A ; 100°C, 0 ; 160°C, 0.DISCUSSIONCOBALT CATALYSTSCobalt-alumina reduced at 4 14°C behaved exactly as reported previously.1sThe present results again show that butenes are formed on this catalyst solely by 1,2-and 1,4- addition of hydrogen to butadiene and that the syn- and anti- conformationsof the C4H7 intermediates do not interconvert. However, the product distributionsobtained using cobalt-alumina reduced at 329°C are similar to those obtained usingnickel-alumina 15 (table 1) and we conclude that although the same modes of hydrogenaddition are operative conformational interconversion of the C4H7 intermediatesoccurs readily.These conclusions are supported by two further observations (i) thatthe kinetics before and after the treatment at 414°C are similar and (ii) that eachbutene from reaction A had the same deuterium content (D.N. = 2-00), as did eachbutene from reaction B (D.N. = 1.39). Had the but-2-enes produced in eitherreaction been formed in two stages, i.e., by the primary formation of but-1-enefollowed by its isomerization to but-2-ene before desorption, the deuterium numberof the but-Zenes would have differed from that of but-1-eneB. J. JOICE, J. J. ROONEY, P. B. WELLS A N D G . R. WILSON 233Powdered cobalt prepared at temperatures in the range 90-330°C behaved likecobalt-alumina prepared at 329°C (type A behaviour in fig.3), whereas, powdersprepared at temperatures in the range 440-580°C behaved like cobalt-alumina preparedat 414°C (type B behaviour in fig. 3). Thus conformational interconversion of theC4H7 intermediates occurs readily on powders P1-P3 which were mostly or completelya-cobalt in a finely divided state (micrograph (a)). On the other hand, conformationalinterconversion did not take place readily at the surfaces of powders p5-P~ whichwere mostly or completely ,&cobalt in a sintered state of relatively low surface area(table 2 and micrograph (b)).The change of mechanism provided a change in product distributions that is inthe sense expected. According to the mechanism in fig.1, the predominant C4H7intermediate is syn- 1 -methyl-n-ally1 which on addition of a hydrogen atom gives but- 1 -ene and trans-but-2-ene, although the latter will be somewhat favoured if the methylgroup sterically hinders hydrogen addition. However, if n-allylic complexes are notformed readily, as provided by the mechanism in fig. 2, addition of hydrogen to C4M7intermediates is expected to give mainly but-1-ene. Consequently, a gross change inthe butene distributions as the mode of chemisorption changes is both expected andobserved.The present results (fig. 3) also closely resemble those which have been obtainedfrom the homogeneous hydrogenation of butadiene by pentacyano-cobalt catalysts.6Here but-1-ene amounted to 80 % of the total butenes when the CN/Co ratio washigh but as this ratio decreased below a critical value of -6 trans-but-2-ene becamethe predominant product. Yields of cis-but-2-ene were also low in all cases andindependent of variations in the nature of the catalyst.The mechanisms suggestedfor these homogeneous hydrogenations 19 have also many features in common withthose which we have postulated.The change from type A to type B behaviour (fig. 3) is not a consequence of thea- to P-phase change in cobalt because a similar change in catalytic behaviour hasbeen observed for a series of nickel powders which show no phase change. Thus,the product distribution is apparently a function of the particle size of the metal,i.e., n-allylic bonded intermediates are more important at the surface of well-sinteredcobalt than on cobalt prepared below 330°C.This marked distinction in behaviourmay be due to considerable differences in the catalytic properties of different crystalplanes, which might also explain why the activities per unit surface area, activationenergies, and the results using deuterium also differed widely, as the physical charac-teristics of the metal altered. Well-sintered cobalt particles may expose low-indexplanes, whereas, the smaller particles formed below 330°C probably expose both highand low-index planes and may also possess a greater number of structural defects.There is therefore the possibility that the type of bonding of the intermediates variesfrom one plane to that of another and an examination of butadiene hydrogenationat particular planes of single crystals is being undertaken at present to test this.PALLADIUM-GOLD ALLOYSThe present results for pure palladium agree well with those previously re-ported5, 129 14 and again demonstrate the marked ability of this metal for formingn-allylic complexes in heterogeneous reactions of hydrocarbons.Thus the mechanismof butadiene hydrogenation is mainly that shown in fig. 1 and it is significant thatsuch ready conversion of dienes to n-allylic complexes is also an important featureof the homogeneous organometallic chemistry of palladium.20Since pumice-supported gold was virtually inactive, even at 200"C, the palladiumatoms are the active centres in the alloys and the results are a measure of the chang234 HYDROGENATION OF BUTADIENEin catalytic properties of this metal with progressive modification by gold.In theregion 0-60 % gold there was very little variation in the major features of the productdistributions and apart from the 65 and 70 % gold alloys the yields of but-1-ece didnot alter significantly over the whole range of composition. Although the trans/cis-ratio in the but-2-enes decreased at lowcr temperatures by a factor of -2 from 0 to95 % gold (fig. 6) this was due to a relatively minor increase in the yield of cis-but-2-eneat the expense of the trans-isomer. Thus, the relative proportions of 1,2- and 1,4-addition of hydrogen were virtually constant so that the presence of gcld has no majoreffect on the ability of palladium to form n-allylic complexes.The slight decreasein stereoselectivity for trans-but-2-ene formation indicates that the stability of then-allylic complexes was not greatly diminished by increasing gold content so that thefraction of the C4H7 intermediates which undergo conformational interconversionat room temperature is small. The stereoselectivity also decreases with temperatureat a given composition but this could also be partly due to a decrease in the relativeproportions of anti- and syn-butadiene in the gas-phase and on the surfaces.The activation energies (table 4) show several features which are similar to thoseobtained by Couper and Eley 21 for parahydrogen conversion on palladium-goldalloy wires.Thus, the activation energies fcr both reactions are independent of goldcontent up to 60 % and the average value of 10-5+0-5 kcal mole-1 for butadienehydrogenation in this range, obtained under conditions where both palladium andthe alloys form the p-phase liydride (see later), is close to the value of - 11 kcal mole-1for parahydrogen conversion on a palladium wire, which had been charged withhydrogen. Couper and Eley also noted a sharp increase in activation energy of some5 kcal mole-1 as the gold content increased from 60 to 70 % and we have found anincrease of the same magnitude from the 60 to 65 % gold alloy. In this region,60-70 % gold, both the paramagnetism and solubility of hydrogen drop to zero. Themajor difference between the two investigations is that while the activation energiesfor parahydrogen conversion remained high for alloys of gold content above 70 %,those for butadiene hydrogenation decreased sharply between the 70 and 75 %alloys.Consequently, the activation energies for butadiene hydrogenation exhibita peak of some 5 kcal mole-1 in height in a narrow range of composition.Couper and Eley suggested that the slow step in parahydrogen conversion is theformation of an activated complex between an adatom of hydrogen and a hydrogenmolecule. The above similarities indicate that the slow step in butadiene hydro-genation may also be activation of molecular hydrogen on surfaces which are coveredby adsorbed diene, and that the same orbitals of the palladium atoms are responsiblein both systems.These authors argued that the increase in activation energy, as theparamagnetism of the alloys decreased to zero at -60 % gold, showed that d-orbitalvacancies help to lower the energy of the activated complex. The present resultssupport this view since the maximum in but-1-ene yields, which paralleled the maxi-mum in activation energies in the 60-75 % gold range, reveals that 1,4-addition ofhydrogen and thus n-allylic bonding also occurs with maximum difficulty in thiscomposition region. Now n-allylic bonding should occur more readily if the pallad-ium atoms have appropriate dn-orbitals which are not fully occupied and the sameorbitals may also participate in activating molecular hydrogen.The sharp drop in both but-1-ene yields and in activation energies between 70 and75 % gold (table 4) is interesting because the palladium atoms at the latter compositionare apparently behaving like those in the 0-60 % range, in contrast with those of the65 and 70 % gold alloys, in spite of the fact that the gold-rich alloys are no longerparamagnetic.The siniple electronic theory of catalysis based on the presence orabsence of holes in the d-band is not completely adequate to account for these result€3. J. JOICE, J. J. ROONEY, P. B. WELLS AND G. R. WILSON 235because the theory makes no distinction between palladium and gold atoms when thereis 65 "/o or more of the latter. Also, a strongly adsorbing species such as butadienemay have as much influence on the d-orbitals of the surface palladium atoms asneighbouring gold atoms.Although the surface atoms in alloys containing 75 % ormore gold may have no d-vacancies some of the palladium atoms seem to be sufficientlymodified by adsorbed diene to behave in much the same way as atoms in alloys of lessthan 65 % gold. Apparently chemisorbed hydrogen cannot exert this influence, asthe activation energies for parahydrogen conversion remain high in the gold-richregion, which might suggest that donor bonding between dn-orbitals on the metal andempty anti-bonding x-orbitals on butadiene has an important influence.In order to account for the behaviour of the 65 and 70 % gold alloys we suggestthat there may be a tendency towards phase-separation in this narrow range where thecomposition is given by the formula PdAu2, or very close to it.Since gold canbehave as a quasi-halogen the palladium atoms may tend to adopt an ordered environ-ment and have essentially the electronic configuration associated with a square-planararrangement as in PdC12. If this were so the dn-orbitals would be fully occupiedand perhaps less readily influenced by adsorbed diene in a direction which wouldfavour easier activation of hydrogen or .n-allylic bonding.The product distributions, especially with palladium and palladium-rich alloys,are apparently influenced by the nature of dissolved hydrogen. Thus the minimain but-1-ene and maxima in trans-but-2-ene yields (fig. 4) and the maxima in n-butanepercentages (fig. 5) occurred in the same temperature range where the p- to a-phasechange of palladium and palladium-gold hydrides was found at roughly the samehydrogen pressures used in the hydrogenation experiments. Moreover, the decreasein size of the maximum in n-butane yields with increasing gold content of the alloysparalleled the smooth decrease in the quantity of liydrogen rapidly evolved when theP-phase hydrides became unstable. These changes in product distributions areunderstood if the surface concentration of hydrogen atoms goes through a maximumin the region where the p- and u-phase hydrides co-exist.Such an increase inhydrogen concentration would favour the chance of adsorbed butene hydrogenatingto butane via the formation of adsorbed C4H9 species before desorption occurred.The surface concentration of butyl groups must also have been at a maximum in thesame temperature region so adsorbed butenes would also have the highest chance ofisomerizing by the mechanism of alkyl reversal before leaving the surface.However,this may not be the complete explanation because these changes in product dis-tributions were observed over too broad a temperature range with alloys of higher goldcontent, and were still present with the 75 % alloy, even though rapid evolution ofhydrogen was not detected with alloys containing more than 65 % gold.Thanks are due to S.R.C. for grants to B. J. J. and G. R. W.NOTE ADDED IN PROOFThe following are the weights of the pumice-supported palladium-gold alloysrequired to maintain an initial rate of hydrogenation of butadiene of 2 % per minuteat 20°C (Pc,H, = 50 mm ; PH, = 100 mm).weight (g) 0.14 0.8 3.3 3.0 2-0 4.0 2.8 50If these weights are assumed to be a measure of surface areas the activation energies(table 4) show that the frequency factors for the 35 and 30 % alloys must be largerthan those of the 40 and 25 % palladium alloys by a factor of N 103.Pd (%> 100 80 50 40 35 30 25 236 HYDROGENATION OF BUTADIENE1 Gault, Rooney and Kemball, J. Catalysis, 1962, 1, 255.2 Rooney, J. Catalysis, 1963, 2, 53.3 Garnett and Sollich, J. Catalysis, 1963, 2, 350.4 Bond and Wells, Ado. Catalysis, 1964, 15, 92.5 Bond, Webb, Wells and Winterbottom, J. Chem. Soc., 1965, 3218.6 Kwiatek, Mador and Seyler, Reactions of Co-ordinated Ligands, A.C.S., (Adv. Chem. Series,7 Frankel, Enken, Peters, Davison and Butterfield, J. Org. Chem., 1964, 22, 3292.8 Paal, Ber., 1912, 45, 2221.9 Lebede’v and Yabubchik, J. Chem. Suc., 1928,2190.no. 37), 1963, p. 201.10 Young, Meier, Vinograd, Bollinger, Kaplan and Linden, J. Amer. Chem. SOC., 1947, 69, 2046.11 Reiche, G r i m and Albrecht, Brennstof-Chernie, 1961, 42, 5 .12 Meyer and Burwell, J. Amer. Chem. Soc., 1963, 85, 2881.13 Wells, Chem. and Ind., 1964, 1742.14 Leszczynski and Wells, Przenysl. Chem., 1964, 43, 508.1s Phillipson and Wells, two papers submitted to J. Chem. Soc.16 Smith and Massingil, J. Amer. Chem. Soc., 1961, 83, 4301.17 For pertinent reviews see Guy and Shaw, Adv. Inorg. Chem. Radiochem., 1962,4,78 ; Bennett,18 Holbrook and Marsh, personal communication.19 Kwiatek and Seyler, J. Organometal Chem., 1965, 3, 421.20 Churchill, Chem. Comm., 1965, 625.21 Couper and Eley, Disc. Faraday Soc., 1950, 8, 172.Chem. Rev., 1962, 62, 611

ISSN:0366-9033    

DOI:10.1039/DF9664100223

出版商:RSC    

年代:1966

数据来源: RSC

摘要:

Geometrical Isomerization of Penta- 1,3-Diene Catalyzed byCobaltBY P. B. WELLS AND G. R. WILSONDept. of Chemistry, The University, HullReceived 3 1 st January, 1966Cobalt powder and alumina-supported cobalt catalyze the reaction :cis-penta-1 , 3-diene +trans-penta-l,3-diene.Molecular hydrogen is not required as a reactant. Isomerization takes place at temperatures above75"C, no other isomers of pentadiene have been detected, and self-hydrogenation is negligible.The isomerization of isotopically labelled cis-penta-l,3-diene at 160°C is described in detail. Theresults are consistent with a mechanism in which the first step is the dissociation of the dioleh byloss of hydrogen from the methyl group, and the overall process is the 1 : 5-transfer of hydrogen.When unsaturated hydrocarbons react with hydrogen on transition metal catalystsa large number of adsorbed states may participate in the mechanism.1 Economyof hypotheses requires the restriction of the proposed adsorbed states to the minimumnumber necessary for an adequate interpretation of the results, but by so restricting thediscussion an accurate account of the mechanism may not be given.This problemhas arisen in the isomerization reactions (i.e., cis-trans isomerization and double-bondmigration) that aliphatic olefins undergo at group 8 metal surfaces.In the past, mechanism (1) has been generally accepted.2-7 Here, olefin adsorptionis assumed to be associative, and hydrogen atom addition to adsorbed olefin precedeshydrogen atom abstraction from adsorbed alkyl.Consequently, this is termed theaddition-abstraction mechanism.R (3-43 R CH, H\ / -H \ // I \CH3\ ' +H -- H - C - C - H C - C' ' H " 6 7\HI H /" T c \ H -H tads9rbed alkyl[R .olkyl](As written here, the olefin is adsorbed as a metal-olefin complex1 ; but, for thispurpose, the essential features of the mechanism are not altered if adsorbed olefin iswritten as a species adsorbed to two metal atoms by two a-bonds.) The addition-abstraction mechanism interpreted the observation that isomerization occurred onlyin the context of hydrogenation, i.e. the presence of molecular hydrogen was consideredto be necessary.3 Tracer studies showed that adsorbed butyl was an intermediatein the formation of butane, and consequently there was good reason to associate23238 PENTADIENE ISOMERIZATJONisomerization with the decomposition of the butyl group.Other workers haveproposed concerted " hydrogen switch " mechanisms in which dissociative adsorptionof olefin and hydrogen atom addition occurred simultaneously.~9 9In 1962, it was reported that the dissociation of adsorbed olefin to give adsorbedn-allylic intermediates occurred readily as one of a series of consecutive dissociativesteps in the exchange of substituted cyclopentanes with deuterium catalyzed by pallad-ium films.10 It was suggested that analogous processes might occur in other reactions,and butene isomerization was quoted as an example. Mechanism (2) shows how iso-merization may proceed by an " abstraction-addition " mechanism, the abstractionstep being the dissociative adsorption of olefin to give adsorbed hydrogen and anadsorbed .n-allylic species.19 11R/ +ti-HrcT. ,C' ' H [R =alkyl]Butene isomerization and hydrogenation using deuterium as a tracer has beenstudied with each of the six noble group 8 metals as catalyst 6912 ; butene isomerization,butene hydrogenation, and the manner in which deuterium appears in both buteneand butane, was studied simultaneously. The conclusion was : " there is rarely anydefinite evidence for n-allylic intermediates : this is not to say that they do not exist,but rather that the observations may be rationally interpreted without them.',lHowever, the addition-abstraction mechanism is sufficiently versatile to account forthe observations, and so the relevance of the abstraction-addition mechanism cannotbe ascertained from this type of experiment.Clearly, if the dissociative adsorptionof olefin occurs, and if its consequences are to be detected, a new approach to theproblem must be made.An important prediction of the abstraction-addition mechanism is that isomeriza-tion should occur in the absence of molecular hydrogen if the rate of isomerizationis fast compared to the rate of removal of adsorbed hydrogen atoms (e.g., by com-bination). In 1964, we reported that butene isomerization did indeed take place inthe absence of molecular hydrogen at the surfaces of cobalt wire and cobalt-alumina,and circumstantial evidence supported the abstraction-addition mechanism.13Unfortunately, this observation does not provide proof of the mechanism, becausedissociative adsorption of olefin may be taking place on part of the surface to give anunreactive residue, and the hydrogen so liberated may then initiate isomerizationon the remainder of the surface by the addition-abstraction mechanism.We decided that studies of the cobalt-catalyzed isomerization of some isotopicallylabelled hydrocarbons in the absence of molecular hydrogen might enable a choice tobe made between the possible mechanisms.The reaction chosen for study was thecis-trans isomerization of penta-l,3-diene, using CH3--CD=CD--CD=CHD as thelabelled compound. This paper records our first experimentsP. B. WELLS AND G. R. WILSON 239EXPERIMENTACATALYSTSCobalt powder was prepared by reduction of grey cobalt ox de (B.D.H.) at 415°C in astream of hydrogen for 24 h.The powder weighed 0.07 g, its surface area measured by theargon adsorption method was 4.8 m2 g-1, and it consisted of a mixture of the a- andp-phasesof cobalt.The support used in the preparation of cobalt-alumina catalyst was Peter Spence type Aalumina. According to the manufacturers, this nlaterial consists of very small crystallites,less than 50 A in size, the surface area is 275 m2g-1, and the only distinguishable form ofalumina present is boehmite. They give the formula A1203. H20. The general methodof preparation of cobalt-alumina has been described elsewhere.14 The two samples used inthis work are designated Co-A1 and Co-A2; they each weighed 0.51 g and contained10 % by weight of cobalt.The reduction of the oxide to the metal was carried out at 414°C ;partial dehydration of the alumina occurred during the reduction. Catalysts prepared inthis way contained adsorbed hydrogen that was not removable by pumping the catalyst.This hydrogen exclianged rapidly with molecular deuterium to give IFD and €32 in the gasphase at temperatures above 100°C, and 1021 hydrogen atoms per g catalyst were exchange-able at 130°C. We conclude, from the magnitude of this quantity,15 that this exchangeablehydrogen was associated with the support. This adsorbed hydrogen was completelyexchanged for deuterium before the catalysts were used for the isomerization reactionsdescribed below.APPARATUSThe catalyst rested on the bottom of a 100 ml Pyrex reaction vessel which was connectedto a conventional high-vacuum system.Pressures were measured using a mercury mano-meter. Hydrocarbons extracted from the vessel were analyzed by gas-liquid chromato-graphy : the 10 ft column contained a saturated solution of silver nitrate in benzyl cyanidesupported on crushed firebrick. Reaction mixtures were separated into pure cis- and trans-pentadiene fractions using preparative gas-liquid chromatography, the stationary phase beingas cited above. The pure fractions were condensed from the carrier gas stream into U-tubesimmersed in liquid air.Mass-, infra-red and proton magnetic resonance spectra of pure samples of cis- and trans-pentadiene were obtained using (i) a modified A.E.I.M.S.3 mass spectrometer (gaseoussamples) ; (ii) a Unicam SP.100 infra-red spectrometer (gaseous samples), and (iii) a Perkin-Elnier 4.0 megacycle n.m.r. spectrometer and a Varian 100 megacycle spectrometer (liquidsamples in carbon tetrachloride solution).INTERPRETATION OF SPECTRAThe composition of a mixture of deuterated pentadienes was calculated from the massspectrum on the basis of the usual assumptions (i) that C-H and C-D bonds had equalchances of breaking in the fragmentation process, and (ii) that all hydrogen and deuteriumatoms in the molecule were equivalent. These assumptions are not likely to introducesubstantial errors because the fragmentation of C5H8 at the standard electron beam energyof 12.0 eV was slight (CsHZ = 100, C5H; = 15, C S H ~ = 1, C S H ~ and lighter fragmentswere not observed).However, the inadequacy of the assumptions was evident in onerespect. For the analysis of samples of pentadiene containing about 4 deuterium atoms,errors in the fragmentation correction accumulated at mass 68, so that pentadiene-do appearedto be 2-5 % of the mixture, whereas, from the general form of the distributions, it was expectedto contribute less than 1 %. Consequently, we put the contribution from pentadiene-doequal to zero and normalized the remainder of the distribution to sum to 100. For thisreason, these product compositions do not contain values for pentadiene-do.The infra-red spectrum of " light " trans-pentadiene (fig. la) contains a group of fiveabsorption bands; a very weak band at 2740 cm-1, and four strong bands at 2880, 2930,3010 and 3075 cm-1.The first three bands have been assigned to C-€3 stretching vibration240 PENTADIENE ISOMERIZATIONof the methyl group, whereas the other two bands are attributed to stretching vibrations ofbonds in which hydrogen atoms are bonded to olefinic carbon atoms. Four absorptionbands for C-D stretching vibrations corresponding to the four strong bands quoted abovewere observed at 2080,2125,2240 and 2325 cm-1 (see fig. l c and Id).I " ' " ' ' '3200 2800 2400 2000frequency (cm-1)FIG. 1 .-Infra-red spectra of pentadiene : (a)" light " trans-pentadiene, pressure in gas cellP = 13 mm; (b) labelled cis-pentadiene, P =41 rnm ; (c) and (d) labelled trans-pentadiene pro-duced by the 60 and 30 % isoinerization of thecis-isomer respectively, P = 15 mm and 7 mm.!L ! 3 4 5 b 7 8 97FIG.2.-Proton magnetic resonance spectra ofpentadiene : (a) " light " trans-pentadiene ;(b) " light " cis-pentadiene ; (c) labelled cis-pentadiene ; ( d ) labelled trans-pentadieneproduced by the 60 % isomerization of thecis-isomer. Spectra (a) and (b) were obtainedusing a 40 megacycle spectrometer, and (c)and (d) using a 100 megacycle spectrometer.Proton magnetic resonance spectra of " light " trans- and cis-pentadiene are shown infig. 2a and 2b respectively; the assignment given in fig. 2a is equally relevant to the otherspectra in this figure. All spectra were integrated by the usual procedures.The mean distribution of hydrogen in a given sample of deuterated pentadiene wascalculated from a knowledge of (i) the mean hydrogen content of the sample, obtained bymass spectrometry, and (ii) the integrals obtained from the p.m.r.spectra. Infra-red spectraprovide qualitative confirmation of the conclusions so obtained.MATERIALSPenta-l,3-diene as supplied by Koch-Light Laboratories Limited contained approximately65 % of the trans-isomer, 25 % of the cis-isomer and 10 % of an unidentified isomer. Purefractions of cis- and trans-pentadiene were obtained by preparative gas-liquid chromato-graphyP. B. WELLS AND G. R. WILSON 241Deuterium (98.7 atom % D) was prepared by the electrolysis of deuterium oxide and waspurified by diffusion through a heated palladium thimble.Isotopically labelled cis-pentadiene was prepared as follows.A solution of penta-l,3-diyne in ether was prepared by the method of Armitage, Jones and Whiting,l6 and the di-acetylene was extracted by preparative scale gas-liquid chromatography (using a 2 ft columnof hexane-2,Sdione supported on firebrick). Gas-phase deuteration of the diacetylenewas carried out in a static system using palladium supported on a-alumina as catalyst attemperatures between 20 and 80°C. Reactions did not proceed in well-defined stages, andproducts representing all possible stages of reduction were obtained simultaneously. Purecis-pentadiene was extracted from the mixture by preparative scale gas-liquid chromatography.The isotopic composition of the product was as follows :dl d2 d3 d4 d5 d6 d7 dg D.N.0.2 6.9 27.8 46.8 16.5 1.2 0-4 0.2 3.79Its p.m.r.spectrum is shown in fig. 2c and its infra-red spectrum in fig. lb. The deuteriumnumber D.N. represents the mean number of deuterium atoms present in the molecule. Themean formula of the material is given in table 1, where analysis I has been obtained fromone integration using the 100 megacycle spectrometer, and analysis I1 has been obtainedfrom a large number of integrations using the 40 megacycle spectrometer. The analysesagree moderately well and the almost complete absence of deuterium from the methyl groupis confirmed by the absence of absorption bands in the infra-red spectrum at 2080 and2125 cm-1 (see fig.lb).TABLE 1 .-ISOTOPIC COMPOSITION OF THE LABELLED CIS-PENTADIENE (REACTANT)[X = H or D]analysis I analysis I1total H content of 4 x 3 2.74 2.91 f0.03H content of each =CX- 0.21 0- 13 f0.02total H content of =CX2 0-84 0.91 fO.01Thus, the material that we obtained was not the specific substance CH3--CD=CD--CD=CHD, but we considered that our product was labelled sufficiently well for a meaningfulstudy of its isomerization to be undertaken.RESULTSThe interconversion of cis- and trans-pentadiene occurred at a measurable ratein the absence of molecular hydrogen when either isomer was contacted with thecatalysts under the conditions described. No isomers of C5Hg other than cis- andtrans-penta- 1,3-diene were detectable in the product.The cobalt powder rapidly catalyzed the isomerization of pentadiene at 160°Cand equilibrium proportions of the two isomers were obtained from 25 mrn of trans-pentadiene after a contact time of 10 min.Measurable initial rates of 2.8 % min-1were obtained at 116°C. The activity at 160°C was independent of the time for whichthe catalyst was pumped before use, for periods up to 3 h. 60 % of the activity(measured at 120°C) remained after the catalyst had been heated in va'cuo for 3 h at400°C: the loss of activity was probably associated with a decrease in the surfacearea of the powder.The identical alumina-supported catalysts Co-A1 and Co-A2 were active for theisomerization above 75°C. For example, 22.7 mm trans-pentadiene was convertedto an equilibrium mixture of cis and trans isomers over Co-Al in 160 min at 136°Cin a reaction that exhibited an initial rate of 3.7 % min-1. Reactions were stronglypoisoned by traces of oxygen in the reactant.Activity declined slowly in a serie242 PENTADIENE ISOMERIZATIONof reactions. The initial rate was approximately zero order in pentadiene pressure(temperature 142"C, pressure range 10-85 mm), and the activation energy for theconversion of the trans to the cis-isomer was 15 f 3 kcal mole-1 over the range 77-135°C. The initial rate of isomerization of the cis-isomer was about three times asrapid as that of the trans-isomer.The composition of the equilibrium mixture of cis- and trans-pentadiene wasmeasured at six temperatures in the range 133-230°C ; the concentration of the trans-isomer was 76-2 f 0.5 % at 133°C and the value decreased as expected,l7 to 73.0 k0.5 %at 230°C.When pentadiene remained in contact with Co-A1 at 160°C for several hourstrace quantities of pentenes were formed : their isomeric composition was not obtainedaccurately but trans-pent-2-ene was the major product.Preferential formation oftrans-pent-2-ene also occurs in the hydrogenation of cis- or trans-pentadiene over thesame catalysts at the same temperature.18Table 2 shows the extent to which hydrogen atoms in the hydrocarbon exchangewith deuterium which is associated with the support. Only a trace of deuteriumentered the initial products, but the deuterium content of the pentadiene increasedconsiderably as the hydrocarbons remained in contact with the catalysts.Clearly,transfer of adsorbed deuterium from the support to the metal was taking place, butthe rate of exchange of hydrogen for deuterium in the hydrocarbon was much slowerthan the rate of isomerization.TABLE THE EXCHANGE OF HYDROGEN ATOMS OF " LIGHT " PENTADIENE WITH DEUTERIUMATOMS ASSOCIATED WITH THE CATALYST SUPPORT. THE ANALYSES SHOW THE COMPOSITIONOF CIS-PENTADENE PRODUCED BY THE ISOMERIZATION OF 100 mm Hg SAMPLES OF TRANS-PENTADIENEcomposition isotopic composition of cis-pentadiene (%)temp. time of mixture("C) (min) (%cisisomer) do di dz 4 d4 ds D.N.135 1 2 99.5 0.5 0.0 0-0 0.0 0.0 0.005135 18 6 98-3 1.7 0.0 0.0 0.0 0.0 0.0171 60 61 25 84.8 13.9 1.1 0.2 0.0 0.0 0.167160 287 25 32.9 40.6 20.3 5.5 0.7 0.0 1.005160 1840 25 29.9 36.1 23.5 8.5 1-5 0.5 1.172Two series of reactions were carried out in which labelled cis-pentadiene wasisomerized over Co-A1 or Co-A2 at 160°C.In the first series, small pressures( N 15 mm) of reactant were employed, and the distribution of deuterium in the trans-pentadiene produced was almost independent of conversion until the reaction was 47 %towards equilibrium. The following initial distribution was obtained by extrapolationto zero conversion :dl d2 d3 d4 d5 d6 d7 ds D.N.0.4 5.8 24.8 43.5 21.2 3.7 0-5 0.1 3-93The exchange reaction with deuterium from the support was more noticeable in thisreaction, and in the reaction to be described below, than was expected from the resultsshown in table 2.In the second series of experiments, larger pressures (- 11 5 mm)of labelled cis-pentadiene were employed, and the results are shown in fig. 3. Theextrapolated initial distribution is similar to that shown above. Comparison of thep.m.r. spectra of the reactant (fig. 2c) and the product extracted after 60 % isomeriza-tion (fig. 2 4 shows that there has been a considerable redistribution of hydrogenbonded to the terminal carbon atoms. The mean formula of the product is giveP. B. WELLS AND G. R. WILSON 243in table 3 (as in table 1, analysis I has been obtained using the 100 megacycle spectro-meter, and analysis 11 has employed the 40 megacycle spectrometer). By 60 %isomerization, molecules of pentadiene must have adsorbed, isomerized and desorbedmany times.The net effect is observed when tables 1 and 3 are compared. The-t2 08 0u*- 10Y 82 .-500 10 2 0 3 0 4 0 5 0 6 0i% isomerizationFIG, 3.-The dependence of the isotopic composition of trans-pentadiene upon conversion a t 160°CInitial pressure of reactant (labelled cis-pentadiene) = 115 mm. Catalyst Co-A2 was used for thereaction taken to 60 % conversion, whereas the other reactions were catalyzed by Co-A1.hydrogen content of the methyl group has decreased considerably, whilst that of theterminal methylene group has increased. The hydrogen content of the internalmethine groups has increased only slightly, assuming that each of the three such groupsof the product have the same isotopic composition. The infra-red spectra shown inTABLE 3 .-bOTOPIC COMPOSITION OF LABELLED TRANS-PENTADIENE (PRODUCT)[X = HorD]total H content of -CX3 1-55 1-51 f0.02total H content of =CX2 1-11 1.27 f0.02H content of each =CX- 0.26 0.22 f0.02analysis I analysis I1fig. l c and Id confirm these results.Trans-pentadiene extracted after 30 and 60 %isomerization contained comparable amounts of deuterium in the methyl group, andfurther, deuterium and hydrogen are both bonded to olefinic carbon.DISCUSSIONCobalt powder and cobalt wire catalyze the cis-trans isomerization of butene andpenta-l,3-diene7 these reactions occuring in the absence of molecular hydrogen. Thi244 PENTADIENE ISOMERIZATIONis primd facie evidence either that the abstraction-addition mechanism is operative, orthat adsorbed hydrogen produced by the dissociative adsorption of pentadiene (togive an unreactive hydrocarbon residue) initiates reaction by the addition-abstractionmechanism. Initiation of the addition-abstraction mechanism by adsorbed hydrogenin equilibrium with absorbed hydrogen can be discounted because (i) the solubility ofhydrogen in cobalt is very l0~,19 and (ii) the cobalt wire was an active catalyst althoughit had never been treated with hydrogen.13Hydrogenation studies have shown that diolefins adsorb very strongly at nietalsurfaces,20 and the zero order of the pentadiene isoinerization, measured by the initialrate method, probably indicates that the surface was fully covered by adsorbedhydrocarbon over the range of pressures studied.A change in the distribution of deuterium took place when labelled cis-pentadienewas converted to trans-pentadiene over cobalt-alumina.The increase in the fractionof pentadiene containing six or more deuterium atoms is evidence that deuterium wasentering the methyl group ; this was confirmed by the infra-red spectra of the trans-pentadiene after the reaction had proceeded 30 and 60 % towards isomeric equili-brium (see fig. l c and Id). At the latter conversion, cis-pentadiene of general formulaA (below) had been converted to trans-pentadiene of general formula B.%,i: OO.79 H0-21 ' 0 - 7 9/An acceptable mechanism for the isomerization of labelled cis-pentadiene mustaccommodate the following observations (i) that hydrogen of the methyl group isexchanged for deuterium, (ii) that deuterium of the methylene group is exchanged forhydrogen, and (iii) that exchange of deuterium for hydrogen at the internal methinegroups is slight, amounting to not more than 0.09 hydrogen atoms per position,assuming that the isotopic composition at each methine group is the same in both thereactant and the product.In addition, molecules of pentadiene exchanged on average0.7 atoms of hydrogen for deuterium which was formerly associated with the support.Two mechanisms are discussed.ABSTRACTION-ADDITION MECHANISMThis mechanism may be written asX X x X(1)adsorbed cis-pen t adienP. B. WELLS AND G. R. WILSON 245x X\ 4C - c - 2a)adsorbed penta-1 ,Cdiene(v)adsorbed trans-pentadienecis-Pentadiene is assumed to interact associatively with the surface in the first instance.Structure (I) shows one of two conformations in which the diolefin may adsorb.Evidence that diolefins adsorb by the formation of two metal-olefin bonds has beendiscussed.1920 The first step in the isomerization involves the abstraction of a hydro-gen atom from the methyl group of (I) to give structure (11).Migration of adsorbedhydrogen atoms is expected to be rapid under these conditions, so the addition stepto form (111) may in principle involve either a hydrogen atom or a deuterium atom;since the relative chances of acquisition of H or D in this step are not known theadsorbed “ hydrogen ” atom has been denoted by the symbol Z. Species (111) isadsorbed penta- 1,4-diene.Assuming that (111) does not undergo conformationalchange, the next step must be the abstraction of Z which will regenerate (11) or give (IV).Addition of an adsorbed hydrogen or deuterium atom, as Z, to (IV) regenerates (III)or gives (V) which is adsorbed trans-pentadiene. Thus, the cis-trans isomerizationof pentadiene by this mechanism is effectively a 1 : 5-transfer of hydrogen. If thistransfer occurs a sufficient number of times, the hydrogen and deuterium atoms of theterminal methyl and methylene groups should equilibrate, and the ratio of the integralsfor the protons in these positions, measured from the p.m.r. spectra, should fall from3-2 in the reactant to about 1-5 in the product. Values of 1.4 and 1.2 were observedfor the trans-pentadiene (table 3) which suggest that this isotopic equilibrium wasachieved.According to the abstraction-addition mechanism the adsorbed “ hydrogen ”atoms Z are entirely H in the initial stages of reaction but, when the “ hydrogen ”atoms bonded to the terminal carbon atoms attain isotopic equilibrium, the chanceof 2 being H will be lower and its chance of being D will be appreciable.Further,when exchange of “ hydrogen” atoms between the metal and the support occurs,as it did in the present work, the composition of Z will be enriched in the isotope thatis associated with the support. IT, as suggested above, the “ hydrogen ” atoms ofthe terminal methyl and methylene groups in the product have attained isotopi246 PENTADIENE ISOMERIZATIONequilibrium, then they can be written as -CZ3 and =CZz respectively. Thus, thecomposition of Z after 60 % isomerization is readily determinable from table 3 as52-55 % H (from analysis I) or 50-63 % H (from analysis U).A feature of our abstraction-addition mechanism is that the isomerization processdoes not require the fission of any of the three C-X bonds, and consequently thedeuterium located at these positions is not expected to exchange with the adsorbed“ hydrogen ”.Experimentally, only slight exchange occurred, showing that otherprocesses which may give exchange at these positions (see below) are of minor im-portance. Our observations are consistent with this mechanism.ADDITION-ABSTRACTION MECHANISMIsomerization by this mechanism can be written as(1) (Wadsorbed cis-pentadieneil -H It0adsorbed trans-pentadiene adsorbed penta-l,+dieneThe addition of a hydrogen atom Z to (I) can occur in two ways to give species (VI)and (VII) and subsequent abstraction of a hydrogen atom X from (VI) or H from themethyl group of (VII) gives respectively adsorbed trans-pentadiene (VIII) and ad-sorbed penta- 1,4-diene (III).The formation of penta- 1,4-diene and its subsequentisomerization back to 1,3-diene provides the route whereby the exchange of hydrogenin the methyl group occurs according to this mechanism. The important feature ofthis mechanism is that all molecules of trans-pentadiene necessarily contain theinitiating entity Z bonded to the methyl-substituted carbon atom (see species (VIII)above).Information about the isotopic composition of Z can be obtained from anexamination of the behaviour of the terminal methylene group during isomerization.The processes to consider are+ z -Y- 2 I-xc YCY, - -xc --CYzt - -xc f C Y t * Y iIf, in the addition of Z, the chance of acquiring H is greater than the chance ofacquiring D then the hydrogen content of the terminal methylene group will increaseP. B. WELLS AND G. R. WILSON 247Experimentally, such an increase was observed, and in consequence, trans-pentadieneproduced by this mechanism must contain 2 as a hydrogen-rich entity. However, thereis no evidence for this from the p.m.r. spectrum of the trans-pentadiene. Formula €3is calculated on the assumption that the isotopic composition of each of the threeinternal methine groups is the same, However, even if the total increase in the H-content of the methine groups is attributed to hydrogen (H) bonded to the methyl-Me\/substituted carbon atom, the composition of this group would be C= whichHO.3 8DO. 62is deuterium-rich, and not hydrogen-rich as required by the addition-abstractionmechanism. This is an important failure of this mechanism.We consider finally the origin of the initiating species Z, with special reference tothe reactions catalyzed by cobalt powder. As stated above, initiation of the addition-abstraction mechanism on the powder apparently requires the dissociation of somepentadiene to give unreactive residues and adsorbed hydrogen atoms. We havefound no evidence for such residues.No induction periods or progressive poisoningduring isomerization were evident. Slow deactivation of the catalysts did occur withuse, which might have been due to the progressive formation of residues ; alternatively,it may have been caused by adventitious oxygen in the reactant.To sum up, we consider that the abstraction-addition mechanism interprets ourobservations successfully, whereas the addition-abstraction mechanism does not.FURTHER DISCUSSION OF THE ABSTRACTION-ADDITION MECHANISMThree further matters merit brief mention. First, a simplification to the abstrac-tion-addition mechanism may be valid which eliminates the necessity of postulatingpenta-1,4-diene as an intermediate. Species (II) and (IV) may be considered torepresent canonical forms of a resonance structure in which electron delocalizationextends over all five carbon atoms.Cis-trans isomerization would then be mostlikely to proceed by 1 : 5-transfer of hydrogen in one step, because addition of hydrogento the centre carbon atom would involve loss of resonance energy. However, notest of this can be made from the present work.Secondly, adsorbed penta- 174-diene, if formed, does not readily undergo thefollowing conformational interconversion, because the subsequent isomerization of(IX) to 1,3-diene would provide a molecule with deuterium exchanged for hydrogenat the centre carbon atom.Lastly, the successful geometrical isomerization of pentadiene by the abstraction-addition mechanism depends upon the conformational characteristics of the adsorbeddiene.Any 173-diene may adsorb in two conformationally distinguishable ways248 PENTADIENE ISOMERIZATIONand for each of the penta-1,3-diene isomers only one conformational situation willlead to isomerization.(4 Me no isomerization k- \Me(b) Meisomerizat ion(trans to cis)Meisomerization 0-L Me (cis to trans)-"pno isomerization (d)L=) - I-)From studies of penta-193-diene hydrogenation using this cobalt catalyst we haveconcluded that penta-193-diene prefers to adsorb as the transoid conformation, ratherthan as the cisoid conformation (probable preference -4 : l).lS If this is also trueduring pentadiene isomerization then, from the conformational standpoint, cis-pentadiene will isomerize more efficiently by the abstraction-addition mechanism thantrans-pentadiene.The authors acknowledge their gratitude to the Science Research Council for theaward of a maintenance grant (to G. R. W.) and for a grant for the purchase of themass spectrometer; to Dr. J. Feeney and Miss A. Heinrich of Varian AssociatesLimited, Walton-on-Thames, who obtained the p.m.r. (100 megacycle) spectra, andto Dr. D. E. Webster of this department for assistance in the interpretation of p.m.r.spectra.1 Bond and Wells, Adv. Catalysis, 1964, 15, 91.ZTwigg, Proc. Roy. SOC. A , 1941, 178, 106.3 Taylor, CataZysis (ed. Emmett, Reinhold, New York, 1957) 1957, 5, 296-379.4 Wagner, Wilson, Otvos and Stephenson, J. Chem. Physics, 1952, 20, 338, 1331.5 Hamilton and Burwell, Proc. 2nd Int. Congr. CataZysis, (Technip, Paris, 1961), 1, 1002.6 Bond, Phillipson, Wells and Winterbottom, Trans. Faraday SOC., 1964, 60, 1847.7 Bond, Catalysis by Metals, (Academic Press, London, 1962), pp. 252 ff.8 Taylor and Dibeler, J. Physic. Chem., 1951, 55, 1036.9 Turkevich and Smith, J. Chem. Physics, 1948, 16, 466.10 Gault, Rooney and Kernball, J. Catalysis, 1962, 1, 255.11 Rooney and Webb, J. Catalysis, 1964, 3, 488.12 Bond, Webb, Wells and Winterbottom, Trans. Faraday SOC., in press.13 Phillipson and Wells, Proc. Chem. Soc., 1964, 222.14 Joice, Rooney, Wells and Wilson, this discussion.15 Hall, Leftin, Cheselske and O'Reilly, J. Catalysis, 1963, 2, 506.16 Armitage, Jones and Whiting, J. Chem. SOC., 1952, 1993.17 Egger and Benson, J. Amer. Chem. SOC., 1965, 87, 3311.18 Wells and Wilson, unpublished work.19 Sieverts and Hagan, 2. physik. Chem. A, 1934, 169, 237.20 Bond, Webb, Wells and Winterbottom, J. Chem. Soc., 1965, 3218

ISSN:0366-9033    

DOI:10.1039/DF9664100237

出版商:RSC    

年代:1966

数据来源: RSC