J. Chem. SOC., Faraday Trans. 1, 1984,80, 803-81 1 Redispersion of Cobalt Particles Supported on Titanium Dioxide BY SEIJI TAKASAKI, HIDEO SUZUKI, QORU TAKAHASHI, SHUJI TANABE, AKIFUMI UENO* AND YOSHIHIDE KOTERA Department of Materials Science, Toyohashi University of Technology, Tempakucho, Toyohashi-shi 440, Japan Received 6th June, 1983 The particle size of cobalt in a Co/TiO, catalyst increased with increasing reduction temperature up to 600 OC. At 700 OC, where the phase transformation of TiO, from anatase to rutile occurred, the cobalt particles redispersed to individual crystallites. This redispersion has been confirmed by measuring the specific activity of the cobalt catalyst for the hydrogenation of propene. The relation between the phase transformation of TiO, and the redispersion of cobalt has been studied kinetically using high-temperature X-ray diffraction spectroscopy.By adapting the one-electron approximation to metallic electrons, Kubol has re- vealed that the spacing between quantized electronic states becomes large in very fine particles. From his calculations the thermal and electronic properties may show considerable deviation from the normal bulk values when the size of the metal particle is < 100 A. Hence, studies have been focused on the relationship between the size of the metal particle and its catalytic behaviour. Many studies concerned with this relationship have been published and Boudart, has proposed that the catalytic reaction is either structure sensitive or structure insensitive. The hydrogenation of ethylene, cyclopropane or benzene on Pt/Al,033 is considered to be structure insensitive, while the hydrocracking, isomerization or dehydrocyclization of 2- methylpentane4 would be structure sensitive.Adsorption on size-controlled metal catalysts has also been studied extensively. Tauster was the first to observe a strong metal-support interaction (SMSI) when he studied the adsorption of hydrogen on platinum dispersed on various metal oxides. Bakers tried to clarify the nature of the interaction between platinum and titanium dioxide during the adsorption of hydrogen on platinum. He observed that the platinum particles increased in size with increasing reduction temperature up to 500 "C, but they redispersed to smaller sizes when they were reduced at 600 "C.This redispersion was explained in terms of the platinum- catalysed formation of Ti40,.6* ' Dumesic8 has observed the redispersion of iron particles supported on titanium dioxide when the catalyst was reduced by hydrogen at 600 "C. He assumed that the iron atoms migrated into the bulk oxide, which is reduced to Ti,O,. In the present work we have also observed the redispersion of cobalt particles in a Co/TiO, catalyst prepared by an alkoxide techniq~e.~ Redispersion was observed when the catalyst was reduced by H, at 7OO0C, the temperature of the phase transformation of TiO, from anatase to rutile. A catalyst prepared by a conventional impregnation method did not show redispersion at any reduction temperature, although the phase transformation of the support occurred at 700 OC. The sizes of the cobalt particles and the crystallites were measured using a transmission electron 803804 REDISPERSION OF C O PARTICLES SUPPORTED ON TiO, microscope and X-ray diffraction spectroscopy, respectively.The results show that the redispersion of the cobalt particles is in fact decomposition of the particles to the individual crystallites which make up the particle. The specific activities of the reduced catalysts for the hydrogenation of propene also indicate that the cobalt particles in the catalyst reduced at 700 OC redispersed to become much finer particles. The activation energies for the anatase to rutile phase transformations in the catalysts prepared by alkoxide and impregnation methods and in pure TiO, were measured by high-temperature X-ray diffraction spectroscopy.There was a consider- able difference between the activation energies of the two catalysts. This difference is attributed to the activation energy of the redispersion of cobalt particles in the catalyst prepared by the alkoxide technique. EXPERIMENTAL CATALYST PREPARATION The catalyst used was Co/TiO, and was prepared by hydrolysis of a mixed solution of tetraisopropyl titanate and cobalt nitrate dissolved in ethylene glycol. The cobalt nitrate was heated at 90°C to remove water, i.e. 90 O C Co(NO,), 6H20 - Co(NO,), * H,O and the tetraisopropyl titanate and ethylene glycol were distilled before use. Cu. 4.0 g of cobalt nitrate was dissolved in 100cm3 of ethylene glycol at 9OOC. Depending upon the desired concentration of Co in the catalyst, an appropriate amount of this solution was poured into the tetraisopropyl titanate in an atmosphere of dry N,.The solution was then stirred at 90 "C for several hours. Water was then added to the mixed solution in the ratio 4: 1 by volume to the amount of propyl titanate. A gel was obtained and it was dried by heating under reduced pressure. The fine powder obtained was then calcined at 500 OC in air for 4 h, followed by reduction for 4 h at various temperatures in a stream of hydrogen. The loading of Co ions in the catalyst could be controlled by varying the concentration of cobalt nitrate. The concentration of cobalt ions in the catalyst was measured by X-ray fluorescence spectroscopy after the extraction of the cobalt ions with hot nitric acid.In the present experiment the concentrations were 1.96, 4.11, 6.02 and 9.68 wt %. The catalyst containing 4.25 wt %cobalt was also prepared by an impregnation method using an aqueous solution of cobalt nitrate and titanium dioxide powder obtained by hydrolysis of the tetraisopropyl titanate. COBALT PARTICLE AND CRYSTALLITE SIZES The size of the cobalt particles and the size distribution were monitored by a transmission electron microscope (TEM, Hitachi H-300) operated at an accelerating voltage of 75 kV. The sample was first ground in an agate mortar and then smpended in water or ethyl alcohol using a supersonic wave. Some of the finest part of the suspension was pipetted onto a microgrid covered with a collodion film (400 mesh, Nissin Film Co.). The micrographs were obtained using an instrumental magnification of 50000 and they were enlarged 4 times when printed.The size distribution curves were obtained by measuring the dimensions of ca. 500 particles for each catalyst. The size of the cobalt crystallites was also measured using an X-ray diffractometer (Rigaku Denki Co., Geigerflex) operated at 40 kV with a filament current of 15 mA using a Zr filter for Mo Ka radiation. The average size of a cobalt crystallite was determined using Scherrer's equation for the diffraction peak at 28 = 20.0" assigned to (1 1 I) planes of cobalt. For these measurements the scanning speed was lowered to 0.25O min-'.J. Chem. Soc., Faraday Trans. I , Vol. 80, part 4 Plate 1 Plate 1. Transmission electron micrographs of Co/TiO, catalyst prepared by the alkoxide technique.The catalyst has been reduced by hydrogen at 500 OC. Wt % Co: (a) 1.96, (b) 4.11, (c) 6.02 and (d) 9.86. TAKASAKI, SUZUKI, TAKAHASHI, TANABE, UENO AND KOTERA (Facing p . 804)J. Chem. Sue., Faraday Trans. 1, Vol. 80, part 4 Plate 2 Plate 2. Transmission electron micrographs of 4.1 1 wt % Co/TiB, catalysts reduced at various temperatures: (a) 400, (b) 500, (c) 600 and (4 700 "C. TAKASAKI, SUZUKI, TAKAHASHI, TANABE, UENO AND KOTERAs. TAKASAKI et al. 805 HYDROGENATION OF PROPENE ON A SIZE-CONTROLLED COBALT CATALYST The hydrogenation of propene was taken as a model reaction to investigate the relationship between the size of the metal particle and its catalytic behaviour. The reaction was carried out using ca.1 .O g of the catalyst packed into a quartz reactor connected to a closed circulating system. The initial pressures of both hydrogen and propene were 100 mmHg and the gases were analysed by gas chromatography using a column packed with Polapak Q. The hydrogenation of propene was carried out at 50, 100 and 150 "C. In this temperature range catalytic hydrogenation often accompanies the hydrocracking to produce hydrocarbons with lower carbon numbers. Therefore, the rate of propane formation was measured for the initial 5 or 10 min of the reaction, where hydrocracking was not observed. Before the hydrogenation, the catalyst in the reactor was reduced by hydrogen at the desired temperature for 4 h and then evacuated at the same temperature for 2 h, followed by cooling to the reaction temperature.PHASE TRANSFORMATION OF TiO, FROM ANATASE TO RUTILE The activation energies of the phase transformation of titanium dioxide from anatase to rutile, both in the presence and the absence of cobalt ions, were measured by high-temperature X-ray diffraction spectroscopy in the temperature range 650-720 "C. The sample, prereduced by hydrogen at 500 "C, was placed in a sample holder made of stainless steel.lo The holder had a thermocouple in it and it was placed inside the oven of the X-ray apparatus. The oven was sealed by mica or aluminium foil through which the X-ray passed. Hydrogen gas was introduced into the oven at the flow rate of 100 cm3 min-l. For the kinetic measurements of the phase transformation, the changes in peak heights at 28 = 29.4" assigned to anatase and 28 = 32.0" assigned to rutile were monitored using Co Ka radiation with an Fe filter.The X-ray diffracto- meter was operated at 30 kV with a filament current of 10 mA. At the desired temperature, the gas introduced into the oven was changed from N, to H, and the measurements were started, although only a few percent of the anatase had been transformed to rutile. RESULTS PARTICLE SIZE AND THE SIZE DISTRIBUTION Some of the photographs obtained for the catalysts-prepared by the alkoxide technique and reduced at 500 OC are shown in plate 1. The concentration of cobalt in the catalyst was varied from 1.96 to 9.86 wt %. The size distributions obtained by measuring ca. 500 particles in each photograph are shown in fig. 1. The effect of the reduction temperature on the size distribution was studied using the 4.1 1 wt % Co/TiO, catalyst in the temperature range 400-700 OC.The photographs obtained are shown in plate 2 and the particle size distributions are given in fig. 2. The mean size of the cobalt particles in the reduced catalysts was calculated using the equation d = Eni di/Cni where ni is the number of particles of size di(A). The results are plotted against the reduction temperature in fig. 3. The size of the cobalt crystallite was also measured using the X-ray diffractometer. The catalyst was dipped into collodion oil immediately after reduction to prevent exposure to air and it then underwent X-ray analysis. The crystalline size was calculated using the equation D = O.~L/COSP where A is the wavelength of the radiation from Mo Ka and 2/? is the width of the peak at half height, following diffraction from the (1 11) plane of the cobalt crystallite. The results are shown in fig.3. However, the catalyst prepared by a conventional impregnation method contains cobalt particles with sizes from 20 to several hundred ingstroms. The size distributions of the catalysts reduced by hydrogen at 500 and 700 O C are shown in fig. 4.806 50 0 !! 30 6 2 ti 10 0 50- a 2 5 30- 2 c.’ a 10- 0 L REDISPERSION OF C O PARTICLES SUPPORTED ON TiO, 200 400 600 0 particle size/A ( b ) 20 10 H 600 bL 800 1000 particle size/A ( d ) n 200 400 600 0 200 400 600 800 1000 1200 particle size/A particle size/A Fig. 1. Particle size distributions of cobalt in Co/TiO, catalysts prepared by the alkoxide technique.Wt % Co: (a) 1.96, (b) 4.1 1, (c) 6.02 and (4 9.86. SIZE EFFECTS ON THE HYDROGENATION OF PROPENE After prolonged reaction, ethane and methane were formed by the cracking of propene and/or propane, and so the rate of propene hydrogenation was estimated by measuring the amount of propane formed in the initial stage of the reaction. The rate of hydrogenation is expressed in terms of a turnover number: the amount of propane formed per minute and per unit surface area of cobalt metal in the catalyst. The specific surface area of cobalt was estimated using the equation S = 5/pd where p is the density of metallic cobalt (i.e. 8.96 g ~ m - ~ ) , d is the size of the cobalt particle observed by TEM, and it is assumed that the cobalt crystallites are cubes. The rates thus obtained are plotted against the cobalt particle size for 4.1 1 wt % Co/TiO, in fig.5 . From the rates thus calculated the activation energies for hydrogenation were obtained (fig. 6). PHASE TRANSFORMATION OF TiO, FROM ANATASE TO RUTILE Kinetic studies of the phase transformation of TiO, from anatase to rutile were carried out using a high-temperature X-ray diffractometer. The rates of the phase transformation were estimated by measuring changes in the anatase and rutile peaks. TAKASAKI et al. 807 Q) 2 2 30 a, 20 10 W DD c.' c 2 a 50 401 0 particle sizelA particle sizelii 40 501 ( d ) 0 200 4 00 particle size/A particle sizelii Fig. 2. Cobalt particle size distribution in 4.1 1 wt % Co/TiO, catalyst reduced at various temperatures: (a) 400, (b) 500, (c) 600 and (d) 700 OC.5 00 4 00 5 N Q) ." Z 300 3 5 200 8 -u a, - .- c1 m a 100 0 1 I I I 4 00 500 600 700 reduction temperature/°C Fig. 3. Changes in mean particle size of cobalt (m) and cobalt crystalline size (0) with varying reduction temperature.808 F .- 1- - - - REDISPERSION OF Co PARTICLES SUPPORTED ON TiO, - -. 7 30 20 E 2 *-) a 10 w g 4 10 0 30-- & 20- 10- I 200 400 600 800 particle size/A ( 4 I 1 I 1 -1 10 \ '\ 0 I00 200 300 400 mean particle size/A Fig. 5. Effects of mean particle size of cobalt upon the specific activity of the cobalt catalyst for propene hydrogenation. Reaction at a, 50; 0, 100 and a, 150 OC.s. TAKASAKI et al. 809 40 d I - 2 6 2 20 \ 5 c m .- I u 0 I I I 0 / / i 200 400 600 mean particle sizelA Fig.6. Activation energy for propene hydrogenation catalysed by Co/TiO, catalysts. - 2 -Y E: - -4 -6 1 1- I I I I 1 I I 0 1.04 1.08 lo3 KIT Fig. 7. Arrhenius plots for anatase to rutile trznsformations in m, pure TiO, and in catalysts prepared by e, the alkoxide method and +, the impregnation method.810 REDISPERSION OF CO PARTICLES SUPPORTED ON TiO, heights. The height of the peak is assumed to be proportional to the concentration of the species to which it corresponds. The catalysts used were the 4.1 1 wt % Co/TiO, sample prepared by the alkoxide technique and the 4.25 wt% Co/TiO, sample prepared by the impregnation method. The phase transformation of TiO, in pure titanium dioxide was also investigated. From the rates thus obtained the activation energies for the phase transformation were estimated (fig.7). DISCUSSION As has been mentioned in a previous paper,g the size of nickel particles in a Ni/SiO, catalyst prepared by the alkoxide technique was controllable by changing the concentration of nickel in the catalyst. In the present work, the size of cobalt particles in the Co/TiO, catalyst prepared by the alkoxide technique was uniform only when the concentration of cobalt was < ca. 6 wt %. As can be seen in fig. 1, the particle size distribution becomes broad for 6.02 and 9.86 wt % Co/TiO, catalysts. Accordingly, this discussion will be concerned with the 4.1 1 wt % Co/TiO, catalyst, which has a relatively sharp particle size distribution curve. The reduction temperature is observed to have a significant effect on the size distribution.The size of the cobalt particles increased with increasing reduction temperature up to 600 OC, while at 700 OC the particle size decreased suddenly from ca. 425 to 100 A (see fig. 3). This phenomenon is called redispersion and is not unique to our work. Bakers and Dumesic8 have already reported the redispersion of platinum and iron particles, respectively, supported on TiO,. The redispersion of cobalt particles appears to be decomposition of the cobalt particle to individual crystallites, since the size of the particle is almost the same as the size of the crystallite when the redispersion occurs at 7OOOC (see fig. 3). The mechanism of the redispersion is still ambiguous. According to Baker et aZ.s the redispersion is caused by the platinum catalysed formation of Ti,O,.The formation of Ti,O, was shown by electron diffraction and, later, by e.s.r. rneas~rements.~ In the present work, no Ti,O, formation was observed by electron diffraction but the phase transformation of TiO, from anatase to rutile was observed using X-ray diffraction when the catalyst was reduced at 700 OC. The relationship between the redispersion of cobalt particles and the phase transformation will be discussed later. The redispersion of cobalt particles was confirmed by measuring the catalytic activity for propene hydrogenation. As can be seen in fig. 5, the catalyst reduced at 700 "C shows the highest turnover number of all the catalysts, suggesting that the cobalt particles in the catalyst reduced at 700 OC are the finest.The fraction of cobalt reduced to metal does not depend upon the reduction temperature used. Magnetic measurements have shown that the fraction of cobalt metal in the catalyst reduced at 500 "C was ca. 77%, while that in the catalyst reduced at 700 "C was ca. 78%. Fig. 6 gives the activation energies of the catalyst reduced at various temperatures for the propene hydrogenation. Again, the lowest activation energy was observed for the catalyst reduced at 700 "C. This leads to the conclusion that the cobalt particles in Co/TiO, catalyst prepared by the alkoxide technique are redispersed when the catalyst is reduced by hydrogen at 700 "C and the redispersion consists of decomposition of the cobalt particle into the individual crystallites of which it is composed.Note that the catalyst prepared by the conventional impregnation method did not redisperse when the catalyst was reduced at 700 OC. As suggested by Baker et aL6 the redispersion of the metal particles is involved with the change in the structure of the TiO,. In the present work the phase transformation of TiO, from anatase to rutile was observed when the catalyst was reduced at 700 OC, the temperature at which the redispersion of cobalt particles occurred. Therefore, thes. TAKASAKI et at. 81 1 activation energies of the phase transformation were measured for the catalysts prepared by the alkoxide technique and the impregnation method and for pure titanium dioxide. The results are shown in fig. 7, indicating that the values of the activation energy decreased in the sequence pure TiO, > alkoxide technique > impreg- nation method.In a previous paper, it was reported1' that the rate equation for the transformation is expressed in terms of the first order of the anatase concentration and the activation energy for pure TiO, was ca. 504.0 kJ mol-l. In the present experiments the activation energy for pure TiO, is 550.6 kJ mol-l, which is in good agreement with the earlier value; the activation energies for the catalysts prepared by the alkoxide technique and impregnation method are 500.6 and 350.7 kJ mol-l, respectively. According to Shannon et a1.12 the activation energy for the transformation is governed by the nature and amount of impurities introduced. In the present case, the added cobalt ions seem to decrease the activation energy for the phase transformation. Although the amount of cobalt ions added to both catalysts is almost the same, there are considerable difference in the activation energies. We believe that the difference in the activation energies can be attributed to the activation energy for the redispersion of the cobalt particles in the catalyst prepared by the alkoxide technique, since in the catalyst prepared by the impregnation method no redispersion occurred. We thank Dr J. Tasaki, Y. Murase and T. Mori of the Japanese Government Industrial Institute, Nagoya for measuring the magnetization and for the transmission electron micrographs. R. Kubo, J. Phys. SOC. Jpn, 1962, 17,975. M. Boudart, Proc. 6th Int. Congr. Cutul. (The Chemical Society, London, 1976), p. 1 . J. M. Dartigues, A. Chambellan and F. G. Gault, J. Am. Chem. SOC., 1976, 98, 856. S. J. Tauster and S. C. Fung, J. Catal., 1978, 55, 29. R. T. Baker, E. B. Prestridge and R. L. Carten, J. Catal., 1979,56, 390; 1982, 59, 293. * M. Boudart, Adv. Catal., 1960, 20, 153. ' T. Huizinga and R. Prins, J . Phys. Chem., 1981,85, 216. * B. J. Tatarchulk and J. A. Dumesic, J. Catal., 1981, 70, 308. l o A. Ueno, S. Okuda and Y. Kotera, 4th Int. Con$ on the Chemistry and Uses of Molybdenum, l1 Y. Suzuki and Y. Kotera, Bull. Chem. SOC. Jpn, 1962, 35, 1353. l* R. D. Shannon and J. A. Pask, J. Am. Chem. SOC., 1965,77, 391. A. Ueno, H. Suzuki and Y. Kotera, J . Chem. SOC., Faraday Trans. I , 1983, 79, 127. Colorado, 1982, p. 250. (PAPER 3/933)