J . Chem. SOC., Faraday Trans. I , 1982, 78, 1515-1523 Influence of Silica and Alumina Supports on the Temperature-programmed Reduction of Copper(I1) Oxide BY STEPHEN J. GENTRY AND PETER T. WALSH Health and Safety Executive, Research and Laboratory Services Division, Sheffield Laboratories, Red Hill, Sheffield S3 7HQ Received 9th June, 1981 Temperature-programmed reduction has been used to sty..dy the reduction of copper oxide supported on silica and alumina. It has been shown that copper oxide exists in two distinct phases when supported on silica, dispersed cupric oxide and a copper silicate species. The relative amounts of the two phases depend on surface area and copper loading, while the reducibility of the silicate phase is dependent on calcination temperature. No evidence of support interaction is found for a-Al,O,, while the complex t.p.r. profiles obtained for copper on y-Al,O, suggest complex suppport interaction. Previous work in these laboratories applied the technique of temperature- programmed reduction (t.p.r.) to an investigation of the redox behaviour of Cul* ions in X and Y zeolites1 and to a study of the promoting effect of transition metal ions on the reduction of bulk cupric oxide.2 It was found that t.p.r. was a convenient and sensitive method for characterising the materials used and also enabled mechanistic and kinetic information to be obtained on the reduction processes.In the present work t.p.r. has been used to study the influence of the support on the nature of the copper species present in copper oxide on silica and alumina.There have been several studies39 of bimetallic catalysts containing copper. The reducibility and alloying behaviour of copper and nickel on silica has received a t t e n t i ~ n ; ~ however, most of the work has concentrated on the reduced state of the metals. Mixed-oxide catalysts containing copper oxide have found use as catalytic converters for carbon monoxide,6 nitric oxide7 and hydrocarbons.8 Consequently there have been many studies of the copper-oxide-support system, particularly with high-surface-area y-A1203. Various techniques have been employed in these investigations. Physical means of characterisation have included magnetic susceptibility measurement^,^-^^ X-ray diffraction,l2. l3 differential thermal analysis,12 electron paramagnetic resonance spectroscopy,13~ l4 X-ray K-absorption spectr~scopy,~~~ l5 optical and infrared spectro~copy~~~ l6 and X-ray photoelectron spectroscopy.13~ 1 7 9 Chemical methods have utilised a static hydrogen-reduction technique,lg and the catalytic activity with respect to the dehydrogenation of isopropyl alcoholg and the oxidation of carbon monoxide.lo In addition to the studies based on the catalytic applications of the copper-support systems, the high degree of dispersion of the copper attainable on these supports has lead to investigations using e.p.r. spectr~scopy~~~ 2 o y 21 of the effect of a magnetically dilute environment on the paramagnetic Cu2+ species. Not all the techniques described above are particularly sensitive to the interactions that can occur between the catalyst and the support when low loadings are used and the catalyst is highly dispersed.Therefore we have used t.p.r., which has been shown 15151516 REDUCTION OF SUPPORTED COPPER OXIDE to be a sensitive technique for catalyst characteri~ation,~ to investigate the interaction between copper oxide and four supports (two samples of amorphous silica having different surface areas, a-Al,O, and y-Al,O,). The materials were also examined by X-ray diffraction and X-ray photoelectron spectroscopy in order to complement, where possible, the t.p.r. data and also to compare the three methods of characterisation. EXPERIMENTAL APPARATUS A N D PROCEDURE The apparatus has been previously2* described. The reactor consisted of two concentric silica tubes.The gas mixture passed down through the inner silica tube (8 mm diameter) containing the sample on a coarse silica sinter and then up through the outer silica tube (14 mm diameter, 180 mm long). A thermocouple well (4 mm diameter) was positioned in the sample bed along the axis of the reactor. The reducing gas (a H,+N, mixture) was dried by magnesium perchlorate before passage through the reference arm of the katharometer and immediately after passage through the reactor. In all experiments a linear heating rate of 13.7 K min-l from 300 to 930 K was used, together with a 10% H2+N2 mixture at a flow rate of 20 cm3 min-’ and a constant weight of material, ca. 200 mg for the supported copper oxide. Two parameters are obtained from a t.p.r. profile; the temperature at which the maximum rate of reduction occurs, T,, and the area underneath the profile corresponding to the amount of hydrogen consumed in the reduction process.Prior to reduction each sample, unless otherwise specified, was pretreated in the reactor under flowing (20 cm3 rnin-l), dried air at various temperatures for a period of 1 h as specified in the Results section. PREPARATION OF MATERIALS The four supports used are listed below, together with their specific surface area as measured by the nitrogen B.E.T. method after treatment in N, at 430 K for 1 h. Also listed is a shorthand notation: amorphous silica (Gasil35, Unilever), 330 f 2 m2 ggl, ‘SiO, (330)’; amorphous silica (Gasil 200, Unilever), 540 & 2 m2 g-l, ‘ SiO, (540)’; a-alumina (Laboratory Reagent grade, B.D.H.), 25 k 2 m2 g-l, ‘a-Al,O,’, and y-alumina (Laboratory Reagent grade, B.D.H.), 70 f 2 m2 g-l, ‘ y-Al,O,’. Each support was impregnated by slurrying with an aqueous solution of cupric nitrate (AnalaR grade, B.D.H.).The loading was varied by adjusting the molarity of the cupric nitrate solution. The slurry was evaporated to dryness at 420 K in an oven and then calcined at the appropriate temperature in the t.p.r. reactor. The loading is expressed as the percentage of copper by weight of the total. Unsupported cupric oxide was prepared by calcining cupric nitrate (AnalaR grade, B.D.H.) in an oven at 630 K and then transferring to the t.p.r. reactor. The t.p.r. experiment was then conducted without further treatment. CHARACTERISATION OF MATERIALS X-RAY D I F FR A c T I o N (X.r.d.) diffractometer using the Debye-Sherrer method.Powder X-ray spectra of the supports and supported copper oxide were recorded on a Philips X-RAY PHOTO E LE c T RO N s P E c T R 0s c o P Y (X.P.S.) X-ray photoelectron spectra were obtained with a VG ESCA 3 using A1 Km radiation. Samples were mounted on double-sided adhesive tape. The measured binding energies were corrected for possible charging effects by assigning the binding energy of the C (1s) peak arising from sample contamination as 285.0 eV.22S. J. GENTRY AND P. T. WALSH 1517 SCANNING ELECTRON MICROSCOPY A N D X-RAY MICROPROBE ANALYSIS The homogeneity of the distribution of the copper species on the support was examined using a Cambridge Stereoscan 600 together with a Link Systems energy-dispersive X-ray analyser.COLOUR The colour of the materials was noted, as this gave an indication of the degree of dispersion of the copper or whether there was any chemical interaction between the support and the copper. RESULTS X-ray microprobe analysis confirmed that the method of impregnation produced a homogeneous distribution of copper throughout the support. The t.p.r. profiles for similar loadings (ca. 1.2%) of copper on the four supports and the profile for unsupported cupric oxide are shown in fig. 1. All the samples were calcined at 900 K for 1 h. Copper oxide supported on silica shows low- and high-temperature reduction regions (cf. unsupported cupric oxide). For Cu/SiO, (330) the low-temperature region consists of two processes consuming approximately three times as much hydrogen as the single high-temperature process.However, for Cu/SiO, (540) a single low- temperature peak is observed, similar in area to the high-temperature peak. When copper is supported on a-Al,O, only low-temperature reduction processes occur, and the t.p.r. profile is similar to that obtained in the low-temperature region for Cu/SiO, 400 600 800 TI K FIG. 1.-Effect of support on t.p.r. profile. Samples calcined at 900 K for 1 h. Sample weight ca. 200 mg. (a) 1.22% Cu/SiO, (330); (b) 1.24% Cu/SiO, (540); (c) 1.22% Cu/a-Al,O,; ( d ) 1.25% Cu/y-Al,O,; (e) unsupported CuO, sample weight 16 mg. The numbers under each peak denote the amount of hydrogen consumed (pmol) in the reduction process. The peak heights have been attenuated for clarity.1518 REDUCTION OF SUPPORTED COPPER OXIDE (330).The Cu/y-Al,O, t.p.r. profile is a very broad peak with reduction occurring continuously between ca. 430 and 920 K. The total amount of hydrogen consumed by all the samples was the same, within experimental error, as that calculated for the reduction of CuO to CuO. The Cu/SiO, (540) system was examined in more detail. Fig. 2 shows the effect of calcination temperature on the t.p.r. profiles of 1.24% Cu/SiO, (540), and also, for comparison, the profile of partially decomposed unsupported cupric nitrate (obtained by calcining at 570 K for 1 h). Profiles (a)-(e) show that T, for the low-temperature reduction process remains unaffected by calcination temperature, whilst Tm for the higher-temperature reduction process increases as the calcination temperature is raised.The amount of hydrogen consumed by the low-temperature process progres- LOO 600 800 TIK FIG. 2.-Effect of calcination temperature on t.p.r. profile of 1.24% Cu/SiO, (540) calcined for 1 h, sample weight ca. 200 mg: (a) 420, (b) 520, (c) 610, ( d ) 770, (e) 900 K. Also shown (f) unsupported cupric nitrate calcined at 570K for 1 h, sample weight 16mg. The numbers under each peak denote the amount of hydrogen consumed bmol) in the reduction process. The peak heights have been attenuated for clarity. sively decreases to ca. 20 pmol, whilst the amount for the high-temperature process is constant. Above a calcination temperature of 610 K the total consumption of hydrogen is the same, within experimental error, as that calculated for the reduction of CUO to CUO.The effect of copper loading on the t.p.r. profile is shown in fig. 3. Tm does not vary significantly for either the low- or high-temperature reduction processes. Fig. 4 shows the effect of the copper loading on the amount of hydrogen consumed in the low- andS. J. GENTRY AND P. T. WALSH 1519 1 I l I I I 400 600 800 TIK FIG. 3.-Effect of loading on t.p.r. profile of Cu/SiO, (540), calcined at 770 K for 1 h: (a) 0.25, (b) 0.59, (c) 1.25, ( d ) 1.91, (e) 4.67, (f) 6.63%. Sample weight ca. 200 mg. The numbers under each peak denote the amount of hydrogen consumed (pmol) in the reduction process. 2 4 6 copper loading (wt. 5%) FIG. 4.-Effect of loading on hydrogen consumption of Cu/SiO, (540), calcined at 770 K for 1 h : (0) low-temperature process; ( x ) high-temperature process; (0) total hydrogen consumption; (- - -) theoretical total hydrogen consumption.1520 REDUCTION OF SUPPORTED COPPER OXIDE high-temperature reduction processes.The total hydrogen consumption is also shown in comparison with the theoretical value, calculated assuming the stoichiometric CuO+H, + Cu+H,O. reaction The slope of the curve corresponding to the low-temperature process appears to change at ca. 2% loading whilst the amount of hydrogen consumed still increases with copper loading. The curve corresponding to the high-temperature process rises with increased loading until the amount of hydrogen consumed reaches a constant value above ca. 4 % loading. The total hydrogen consumption is directly proportional to the copper loading throughout the experimental range and deviates from the theoretical value by a maximum of 5%.Copper loadings of ca. 1% were approaching the detectable limit of the X.r.d. technique for particle sizes > 5 nm. For the supported samples presented in fig. 1 CuO was only detected on 1.22% Cu/SiO, (330). Of the samples detailed in fig. 3 CuO was detected only on 6.63% Cu/SiO, (540), calcined at 770 K. No other diffraction lines were observed in any of these samples. Loadings of ca. 1% were at the limit of detectability of the X.P.S. technique, and consequently gave poor spectra. However, for the samples presented in fig. 1 no C~(2p,/,) chemical shifts were observed for any of the supported materials relative to CuO, which has a measured binding energy of 936.2 & 0.2 eV [corrected to a C (1s) binding energy of 285.0 eV].Also the binding energy of the Cu(2p3,,) peak for 6.63% Cu/SiO, (540) calcined at 770 K was the same, within experimental error, as that for CUO. For the samples shown in fig. 1 the colours are as follows: Cu/SiO, (330), blue-grey; Cu/SiO, (540), pale blue; Cu/a-Al,O,, blue-grey ; Cu/y-Al,O,, pale blue. The 1.24% Cu/SiO, (540) samples remained pale blue for all calcination temperatures. When the loading was increased the Cu/SiO, (540) samples calcined at 770 K gradually changed colour from pale blue to turquoise. DISCUSSION Fig. 1 shows that the reduction of copper oxide is clearly affected by the support material. All the supports give rise to reduction processes occurring at lower temperatures than in unsupported CuO (1.t.processes). Cu/a-Al,O, is the only material studied which reduces in a single step; all the other materials give rise to additional reduction processes at higher temperature than that for unsupported CuO (h.t. processes). However, for Cu/y-Al,O, the h.t. and 1.t. processes are not clearly resolved. The total amount of hydrogen consumed in each sample is equal to the theoretical amount for the overall reduction CulI + CuO. There is no evidence, however, that the reduction processes, most clearly resolved in the Cu/SiO, samples, correspond to a stepwise reduction of CulI to Cul (1.t.) and Cul to Cuo (h.t.), as observed in the reduction of cupric ions in zeolites., Rate measurements and magnetic susceptibility studieslS carried out during the isothermal hydrogen reduction of CuO dispersed on several supports were consistent with the direct reduction of CulI to CuO. Furthermore, X.P.S.combined with an infrared absorption studylS following carbon monoxide adsorption failed to identify Cul as an intermediate in the hydrogen reduction of Cu/y-Al,O,. It is unlikely that the large variations in T, with calcination temperature for the h.t. process in SiO, (540) reported here (fig. 2) could arise from a stabilisation of Cul, since this species would itself be formed from CuI1 in the 1.t. process which is clearly independent of calcination temperature.S. J. GENTRY AND P. T. WALSH 1521 We conclude that silica and alumina supports do not stabilise Cul as an intermediate in the reduction of supported CuO, in line with previous observations' of a one-step reduction of CuO supported externally on X zeolite.Cu/A1,0, SAMPLES The reduction behaviour of Cu/a-Al,O, and Cu/y-Al,O, is markedly different. The t.p.r. profile of Cu/a-Al,O, [fig. 1 (c)] consists of a major peak at 493 K and a shoulder at ca. 533 K. Cu/y-Al,O,, on the other hand, has a broad t.p.r. profile [fig. l(d)] spanning 430-920 K. These results are similar to those of Voge and Atkins,lg who ascribed the non-uniformity of CuO supported on y-Al,O, compared with a-Al,O, to chemical interactions between the oxide and y-Al,O,. Cu/a-Al,O, is blue-grey, whereas Cu/y-Al,O, is a pale blue colour, indicating the presence of relatively large CuO particles in the former sample.Banerjee et aL2, in a t.p.r. study of nickel oxide supported on a-Al,O, found that the form of the Arrhenius plots for hydrogen reduction approached that for massive nickel oxide as the loading was increased. Furthermore, the activation energy for the initial reduction process, postulated as the interaction of gaseous molecular hydrogen with low-energy sites on the NiO lattice, was found to increase with loading. This, it was suggested, would be compatible with the increase in crystal size of NiO. It is thus apparent that a-Al,O, functions solely as a dispersing agent for copper oxide, whereas a considerable range of oxide-support interaction occurs in Cu/y-Al,O,. Copper oxide and alumina are known to form a mixed oxide having a spinel-type s t r u c t ~ r e , ~ ~ and a substantial amount of evidence had been accumulated for the existence of an aluminate phase when cupric ions are dispersed on y-A1,0,.99 1 1 - 1 5 9 17-19 y-Al,O, possesses a defect spinel structure, whereas a-Al,O, has a hexagonal close-packed structure. Thus it is to be expected that a spinel would be formed more easily on y-Al,O,. Indeed Friedman et aL.l3 and Wolberg and Roth15 found that the formation of a copper aluminate phase from impregnation of y-Al,O, with cupric nitrate was dependent on surface area, calcination temperature and copper concentration. Cu/SiO, SAMPLES From the results presented in fig.2 and 3 it can be seen that the position of the 1.t. peak is uninfluenced by either calcination temperature or loading.It is clear from fig. 1 (a) and (b), however, that, although the 1.t. peak is in essentially the same region, some structure is introduced as the surface area of the support is reduced. Since increasing the loading, increasing calcination temperature and decreasing support surface area would each be expected to increase the average particle size of the oxide, it appears that the 1.t. reduction process must be essentially independent of particle size. It is suggested that the degree of dispersion in the Cu/SiO, (540) samples is such that the dominant reduction mechanism is one of nucleation rather than a decaying rate mechanism, such as the contracting-sphere model appropriate for the bulk For similar copper loadings Cu/SiO, (330) is blue-grey, similar to Cu/a-Al,O,, whereas Cu/SiO, (540) is pale blue, suggesting that CuO exists as larger particles on the lower-area support as expected.The structure observed in the 1.t. reduction profile of Cu/SiO, (330) (and also in Cu/a-Al,O,) may then be attributed to a change in mechanism. The higher-temperature shoulder (Tm = 549 K) may then arise from the contribution of a decaying rate stage in the reduction of the larger CuO particles. The position of the high-temperature peak is again independent of surface area (fig. l), indicating that a similar interaction occurs between CulI and the two silica supports. For each sample the ratio of the peak areas in the h.t. and 1.t. processes is equal to the ratio of CU" combined with the support (which may be described as1522 REDUCTION OF SUPPORTED COPPER OXIDE a copper silicate species) and in a dispersed CuO phase.For Cu/SiO, (540) this ratio is significantly greater than for Cu/SiO, (330). Thus the higher degree of dispersion attainable on the higher-area support increases the relative amount of silicate species produced. The amount of silicate produced varies with loading. From fig. 3 and 4 it can be seen that below 2% Cu the amounts of both species increase approximately equally with increased loading. However, beyond this level the amount of copper silicate formed increases more slowly until a maximum is reached at ca. 4%. Above this level all additional copper is present as the 1.t. dispersed oxide form. It thus appears that an equilibrium is established between dispersed CuO, support and copper silicate.Below a loading of ca. 2%, copper exists equally as dispersed oxide and silicate, whilst a maximum level of silicate, corresponding to ca. 2% Cu for Cu/SiO, (540), is reached as the total loading is increased. This contrasts with the Cu/y-Al,O, system,', where the crystalline phase of cupric oxide, analysed by X.r.d., forms only when the adsorption sites of the support, which interact with the copper to form the aluminate, are saturated. The degree of interaction between copper oxide and silica is strongly dependent on calcination temperature as shown in fig. 2. Two features may be identified. First, the area of the low-temperature peak at T, z 535 K diminishes to a constant value, equivalent to 20 pmol H,, for calcination temperatures above 610 K.Partially decomposed copper nitrate essentially reduces simultaneously with copper oxide [fig. 2(f)]. Consequently at the lower calcination temperatures used (< 610 K) the peak at T, z 535 K contains a component for the reduction of undecomposed nitrate, and therefore the area is larger than the theoretical value for CuO reduction. Secondly, a high-temperature peak of approximately constant area emerges, shifting from 566 K after calcination at 420 K, to 910 K after calcination at 900 K, suggesting that the equilibrium concentration of the copper silicate species is established at as low a temperature as 420 K. However, the interaction becomes markedly stronger as the calcination temperature is increased. Wolberg and Roth15 found that the formation of a copper aluminate phase from copper dispersed on y-Al,O, was similarly dependent on surface area, copper concentration and calcination temperature.The behaviour of the Cu/SiO, system in having both 1.t. and h.t. processes, ascribed here to dispersed CuO and a CuO-SiO, interaction, respectively, contrasts with that of the Ni/SiO, system.26 Here only poorly dispersed NiO is formed and a NiO-support interaction only occurs when a silica-alumina support was used. The existence of an equilibrium between CuO and copper silicate may influence the activity of Cu/SiO, catalysts. If the dispersed CuO has a greater specific activity than the copper silicate phase (which may be expected, particularly for oxidation reactions, since dispersed CuO is more easily reduced), then for loadings below 2% on the higher surface area silica support only half of the available copper is in a catalytically active form.Increasing the loading to more than 2% increases the fraction of copper present as the dispersed oxide but reduces the degree of dispersion. Hence there would be an optimum loading and support surface area for maximum catalytic activity. CONCLUSIONS T.p.r. is a sensitive analytical technique for the characterisation of supported copper oxide. The deviation of the experimental points shown in fig. 4 from the theoretical curve is at most 5% over a range of conditions which precluded detailed analysis by either X.r.d. or X.P.S.S. J . GENTRY AND P. T. WALSH 1523 It has been shown, using t.p.r., that supported copper oxide can exist in two forms.These are a dispersed cupric oxide phase and a combined cupric-oxide-support phase. For Cu/ y-Al,O, the cupric-oxide-support interaction is not as clearly demonstrable as for the Cu/SiO, system, whilst no such interaction occurs for Cu/a-Al,O, under the conditions used. A detailed study of the Cu/SiO, system showed that (i) the relative amounts of dispersed cupric oxide and copper silicate are influenced by the surface area of the support and the copper loading and (ii) the reducibility of copper silicate is markedly affected by the calcination temperature. S. J. Gentry, N. W. Hurst and A. Jones, J. Chem. Soc., Faraday Trans. 1, 1979, 75, 1688. S. J. Gentry, N. W. Hurst and A. Jones, J. Chem. SOC., Faraday Trans. I , 1981, 77, 603. H. E. Swift, F.E. Lutinski and W. L. Kehl, J. Phys. Chem., 1965, 69, 3268. J. H. Sinfelt, J . Catal., 1973, 29, 308. S. D. Robertson, B. D. McNicol, J. H. de Baas, S. C. Kloet and J. W. Jenkins, J. Catal., 1975,37,424. S. Sourirajan and M. A. Accomazzo, Can. J. Chem., 1960, 38, 1990. ’ S. Sourirajan and J. L. Blumenthal, Proc. 2ndInt. c’ongr. Catal., 1960 (Editions Technip, Paris, 1961), p. 2521. S. Sourirajan, M. A. Accomazzo and K. Nobe, Proc. 2nd Int. Congr. Catal., 1960 (Editions Technip, Paris, 1961), p. 2497. P. W. Selwood and N. S. Dallas, J . Am. Chem. SOC., 1948, 70, 2145. lo J. Mooi and P. W. Selwood, J. Am. Chem. SOC., 1952,74, 2461. l 1 P. E. Jacobson and P. W. Selwood, J. Am. Chem. SOC., 1954, 76, 2641. l 2 E. D. Pierron, J. A. Rashkin and J. F. Roth, J. Catal., 1967, 9, 38. l 3 R. M. Friedman, J. J. Freeman and F. W. Lytle, J. Catal., 1978, 55, 10. l5 A. Wolberg and J. F. Roth, J. Catal., 1969, 15, 250. Is K. P. de Jong, J. W. Geus and J. Joziasse, J. Catal., 1980, 65, 437. l 7 A. Wolberg, J. L. Ogilvie and J. F. Roth, J. Catal., 1970, 19, 86. l 8 G. Ertl, R. Hierl, H. Knozinger, N. Thiele and H. P. Urbach, Appl. Surf. Sci., 1980, 5, 49. l9 H. H. Voge and L. T. Atkins, J. Catal., 1962, 1, 171. 2o R. J. Faber and M. T. Rogers, J. Am. Chem. Soc., 1959, 81, 1849. 21 Y. Matsunaga, Bull. Chem. SOC. Jpn, 1961, 34, 1291. 22 T. Robert, M. Bartel and G. Offergeld, Surf. Sci., 1972, 33, 123. 23 A. K. Banerjee, S. R. Naidu, N. C. Ganguli and S. P. Sen, Technology, 1969, 11, 162. 24 H. Hahn, G. Frank, W. Klingler, A. Storger and G. Storger, 2. Anorg. Allg. Chem., 1955, 279, 241. 25 J. Haber, J. Less-Common Metals, 1977, 54, 243. 26 M. Houalla, F. Delannay and B. Delmon, J. Chem. SOC., Faraday Trans. I , 1980, 76, 1766. P. A. Berger and J. F. Roth, J. Phys. Chem., 1967, 71, 4307. (PAPER 1 /928)