Pesquisar no resumo do documento
Applied Catalysis A: General 235 (2002) 71–78
Sol–gel synthesis of titania–alumina catalyst supports
E.Y. Kaneko a,b, S.H. Pulcinelli a,∗, V. Teixeira da Silva c, C.V. Santilli a a Instituto de Quı́mica, UNESP, P.O. Box 355, 14800-970 Araraquara, SP, Brazil b Área Interunidades, IFSC/USP, P.O. Box 365, 13560-000 Sao Carlos, SP, Brazil
c Instituto Militar de Engenharia, Pça Gal Tiburcio 80, 22290-970 Rio de Janeiro, Brazil
Received 22 January 2002; received in revised form 12 April 2002; accepted 12 April 2002
Titanium oxide is a good candidate as new support for hydrotreating (HDT) catalysts, but has the inconvenience of presenting small surface area and poor thermal stability. To overcome these handicaps TiO2–Al2O3 mixed oxides were proposed as catalyst support. Here, the results concerning the preparation, characterization and testing of molybdenum catalyst supported on titania–alumina are presented. The support was prepared by sol–gel route using titanium and aluminum isopropoxides, chelated with acetylacetone (acac) to promote similar hydrolysis ratio for both the alcoxides. The effect of nominal complexing ratios [acac]/[Ti] and of sol aging temperature on the structural features of nanometric particles was analyzed by quasi-elastic light scattering (QELS) and N2 adsorption isotherm measurements. These characterizations have shown that the addition of acac and the increase of aging temperature favor the full dispersion of primary nanoparticles in mother acid solution. The dried powder presents a monomodal distribution of slit-shaped micropores, formed by irregular packing of platelet primary particles, surface area superior to 200 m2 g−1 and mean pore size of about 1 nm. These characteristics of porous texture are preserved after firing at 673 K. The diffraction patterns of sample fired above 973 K show only the presence of anatase crystalline phase. The crystalline structure of the support remained unaltered after molybdenum adsorption, but the surface area and the micropore volume were drastically reduced. © 2002 Published by Elsevier Science B.V.
Keywords: Sol–gel process; Molybdenum catalysts; Thiophene hydrodesulfurization
Molybdenum supported on alumina promoted by nickel or cobalt constitutes the most important cata- lysts for petroleum hydrotreating (HDT) . These catalysts are used mainly to saturate unsaturated hy- drocarbons in order to remove sulfur (hydrodesulfuri- zation—HDS), nitrogen (hydrodenitrogenation— HDN), oxygen (hydrodeoxygenation—HDO) and metals (hydrodemetallation—HDM) from different
∗ Corresponding author. Tel.: +55-16-201-6638; fax: +55-16-222-7932. E-mail address: firstname.lastname@example.org (S.H. Pulcinelli).
petroleum streams . The more stringent environ- mental regulations that have been recently enacted throughout the world has increased the need for more active HDT catalysts and this has led to the study of a variety of compositions that differ from standard sulfide catalysts. Basically the approach has been to explore: (a) new supports; (b) noble metal catalysts; (c) zeolite-containing combinations; (d) new compo- sitions. Therefore, development of more active and more selective catalysts has become a challenge for refiners [2,3]. Among the new supports, titania  has attracted attention in view of the higher HDS activity displayed by molybdenum sulfide supported on this oxide . It has been found, for example, that
0926-860X/02/$ – see front matter © 2002 Published by Elsevier Science B.V. PII: S0926 -860X(02 )00236 -3
72 E.Y. Kaneko et al. / Applied Catalysis A: General 235 (2002) 71–78
for thiophene HDS, molybdenum catalysts supported on titania are four times more active than those sup- ported on alumina [4,5]. However, titania supports generally present not only low specific surface area (50 m2 g−1) when compared to alumina (200 m2 g−1) but also poor thermal stability of the anatase phase at high temperatures [5–7]. To overcome these handi- caps and taking advantage of the good properties of titania (high activity) and alumina (excellent texture, mechanic and thermal properties), TiO2–Al2O3 mixed oxides have been proposed as HDT catalyst [5,8].
The sol–gel process has been used for the last 20 years to prepare different alumina [9,10] and titania  powders with uniform particles size and high surface area. The possibility of fine tuning the texture of these supports by careful control of the sol–gel process parameters (hydrolysis ratio, complexing ratio, aging temperature and acidity) has been demon- strated [9–11]. However, the sol–gel routes to pre- pare titania–alumina mixed oxides as HDT supports have not been developed yet . Within the material science area, parallel work has been in progress di- rected at development of sol–gel strategy to produce titania–alumina membranes , however, the texture of these materials is inadequate for using them as catalyst support.
The work described in the present article combines the strategies frequently used to controlling the tex- ture of sol–gel-derived transition metal oxides, to pre- pare titania–alumina mixed oxides for application as HDT catalyst support. The effect of the modification of the reactivity of precursor (titanium tetraisopropox- ide) by acetylacetone (acac) as complexing ligand, of the temperature of sol aging and of the powder firing on the texture of titania–alumina powder was inves- tigated. The performance of this new support in the thiophene HDS reaction is reported.
2.1. Samples preparation
All experiments were performed under atmospheric conditions and all chemicals were used without further purification. A solution of titanium tetraisopropoxide (Strem Chemical, 98%) in isopropanol (1 mol dm−3) was added slowly to acac (Fluka, 99%) and the
mixture was stirred at 298 K for 15 min. An exother- mic reaction of complexation occurred leading to a yellow solution. Different amounts of acac were used to adjust the nominal complexing ratios [acac]/[Ti] to 0, 0.5, 1 and 2. Aluminum di(isopropoxide) ace- toacetic ester chelate (Strem Chemical, 98%) was mixed to the solution containing the complexed tita- nium precursor to yield the molar ratio Ti/(Ti + Al) equal to 0.5. Hydrolysis of the clear solution was then performed by dropwise addition of aqueous para-toluene sulfonic acid (PTSH) (Vetec, 99.9%) solution (0.1 mol dm−3). The resulting solution ex- hibited a pH of 2.4, nominal hydrolysis ratio (h = [H2O]/[Ti + Al]) of 100 and acidity ratio (H+ = [H+]/[Ti + Al]) of 0.2. The solution was kept under stirring and then aged under reflux at different temper- atures (333, 343, 353 and 363 K) for 16 h. The aged suspensions were cooled down to room temperature and dialyzed, i.e. aliquots of 15 ml of so-prepared sus- pensions were put inside acetylcellulose membranes tubing (12–14000 MW) and then submitted to static dialysis against 100 ml of bi-distilled water for 4 days, until pH reached 7, to eliminate as much as possible the soluble species. The water of dialysis was changed each 8 h. The colloidal particles were isolated from suspension by freeze-drying at 268 K and pressure of ∼1 mHg. The powders were heat-treated at different temperatures (473, 673, 773 or 973 K) for 2 h.
The molybdenum catalysts were prepared by the ad- sorption method. Succinctly, ammonium heptamolyb- date aqueous solution (0.004 mol dm−3) was contacted with the support for 36 h. After this adsorption period, the solution was evaporated, the obtained solid was dried at 383 K for 18 h and then fired at 673 K for 3 h. The volume of ammonium heptamolybdate solution and the mass of titania–alumina powder in the mix- ture was calculated in such a way to obtain a catalyst containing four molybdenum atoms nm−2.
The hydrodynamic diameter of particles in the studied sols were determined from quasi-elastic light scattering (QELS) measurements, performed with a solid state laser of 25 mW (λ = 532 nm) and a BI-1000AT Brookhaven photocorrelator. The X-ray powder diffraction (XRPD) measurements have been performed on dried and fired samples with a Siemens
E.Y. Kaneko et al. / Applied Catalysis A: General 235 (2002) 71–78 73
D5000 diffractometer, using the graphite monochro- mated Cu K radiation. The BET surface area and the porous texture of catalyst supports have been evalu- ated by N2 adsorption isotherms obtained at 77 K in a relative pressure range between 0.001 and 0.998, us- ing an ASAP 2010 (Micrometrics) equipment. Before each measurement, samples were degassed at 353 K under vacuum (≈1 m Hg) for time enough (12 < t < 18 h) to observe the absence of significant change in vacuum stability. Micropore size was calculated from the adsorbed volume (V) versus the statistical thick- ness of the adsorbed layer (t) curves, also called t-plot . The thickness of this layer was calculated for each relative pressure from the Harkins and Jura  equations, following the algorithm proposed in .
Before catalytic testing, the samples were sulfided in situ using a 10% (v/v) H2S/H2 gas mixture (AGA, 99.999%) at 673 K during 2 h. Immediately after the sulfiding step, the reaction was started by passing a 10/1 molar ratio of H2/C4H4S reactant mixture through the reactor. This ratio was obtained by flowing pure H2 (1.2 l h−1) through two saturators connected in series maintained at room temperature by a water bath and filled with thiophene (Merck, 99%). The reaction conditions were adjusted to maintain low conversions of thiophene (<10%) and absence of diffusional effects. Products were analyzed on-line by gas chromatography using a HP 2900 gas chromatograph.
QELS experiments were performed on synthesized sols obtained after aging in PTSH aqueous solution (pH = 2.4). The hydrodynamic size distribution was calculated supposing a set of particles with a multimodal distribution. The effects of the aging tem- perature and complexing ratio on the particles size distribution are shown in Fig. 1a and b, respectively. The analysis of these data shows that sols are made of monomodal distribution of nanometric colloidal particles, the mean diameter of which increases when [acac]/[Ti] increases up to 0.5 or the synthesis tem- perature increases up to 343 K.
The effect of the aging temperature and of com- plexing ratio on the powder texture of titania–alumina dried powders is verified by changes in the N2
Fig. 1. Evolution of the hydrodynamic diameter distribution of particles with the (a) aging temperature ([acac]/[Ti] = 1) and (b) complexing ratio (T = 343 K).
adsorption–desorption isotherms shown in Fig. 2a and b, respectively. The plateau at high p/p0 range, the high volume of gas adsorbed at low relative pres- sure (p/p0) and the absence of hysteresis, in the curve corresponding to the sample synthesized at 343 K with different complexing ratio, is typical of materi- als presenting micropores (type I isotherm, according to IUPAC classification ). Above this tempera- ture and for [acac]/[Ti] = 1 (Fig. 2b) a considerable increase in the adsorption occurs at high p/p0 range, indicative of the mesoporosity and isotherms present a hysteresis loop types H3 and H2 for samples synthe- sized at 353 and 363 K, respectively. This hysteresis loop evolution indicates that the regularity of cross section along the longitudinal direction of pores in- creases by increasing the temperature. In an idealized description, this is equivalent to the transformation
74 E.Y. Kaneko et al. / Applied Catalysis A: General 235 (2002) 71–78
Fig. 2. Evolution of N2 adsorption–desorption isotherms with the (a) aging temperature ([acac]/[Ti] = 1) and (b) complexing ratio (T = 343 K).
from slit-shaped towards inkbottle-shaped mesopores , that may correspond to the changes of plate-like to corpuscular particles.
Quantitative information about the micro and mesopore contribution on the overall texture of titania–alumina powders was obtained from t-plot. Non-porous solids are characterized by a linear re- lationship between V and t, in which the specific surface area (St) is proportional to the slope of the line (for N2 adsorption, St = 15.47V/t). The capillary condensation of adsorbate in the mesopores leads to a positive deviation of V–t curves. On the other hand, for solids containing micropores the thickness of adsorbed layer is limited by the core radius of the
pores and the corresponding t-plot presents a negative deviation from linearity. An example of this negative deviation is observed in the t-plot corresponding to samples aged at 343 and 363 K, shown in Fig. 3. Two straight lines are obtained for sample prepared at 343 K and the t-values corresponding to intersection of them (ti) allows to calculate the mean micropore size (d = 2ti). The volume of micropore (V) was estimated by extrapolating the less step line above the intersection to the ordinate. The external volume (Ve) or the volume of mesopore (Vm) can be calcu- lated by subtracting the total pore volume (Vp) from V. The total (St = SBET) and the external (Se) areas were calculated from the slope of straight lines, and
E.Y. Kaneko et al. / Applied Catalysis A: General 235 (2002) 71–78 75
Fig. 3. The t-plot corresponding to samples prepared with [acac]/[Ti] = 1 and aged at 343 and 363 K, illustrating the proce- dure to determining the mean pore size (d = 2ti), surface area (S) and volume (V) of micropore.
the subtraction of them gave the micropore area (S). The average size of micropores was calculated by the following equation :
d = fV S
Table 1 Textural parameters of dried titania–alumina powders and supports fired at 673 K prepared from powders aged at different temperaturesa
Experimental parameter Surface area (m2 g−1) Pore volume (cm3 g−1) Mean pore size (nm)
Average pore size (nm)
SBET S Vt V 4V/S 2V/S
[acac]/[Ti] aged at 343 K 0/1 285 271 0.16 0.14 1.1 2.0 1.0 1/2 274 262 0.14 0.12 0.8 1.8 0.9 1/1 234 199 0.15 0.10 1.1 2.0 1.0 2/1 238 228 0.12 0.10 1.2 1.8 0.9
Aging temperature (K) for [acac]/[Ti] = 1 298 35 23 0.01 0.01 0.9 1.8 0.9 333 172 166 0.09 0.08 1.0 2.0 1.0 343 234 199 0.15 0.10 1.1 2.0 1.0 353 277 204 0.28 0.11 1.2 2.2 1.1 363 290 0.30
Agingb temperature (K) for [acac]/[Ti] = 1 333 262 (57) 202 (29) 0.31 (0.11) 0.10 (0.02) 1.2 (1.2) 2.0 (2.8) 1.0 (1.4) 343 223 (98) 200 (80) 0.21 (0.11) 0.13 (0.06) 1.2 (1.3) 2.6 (3.0) 1.3 (1.5) 353 212 (98) 211 (66) 0.19 (0.16) 0.14 (0.04) 1.3 (1.3) 2.7 (2.4) 1.3 (1.2) 363 235 (175) 224 (151) 0.23 (0.19) 0.19 (0.12) 1.4 (1.4) 3.4 (3.2) 1.7 (1.6) Al2O3 180 (160) 0.34 (0.35) 7.0 (6.8)
a The values in parenthesis correspond to the molybdenum-impregnated supports. b Supports fired at 673 K.
where the constant f depends on the geometric form of the pores, assuming values of 4 and 2 for cylindrical and slit-shaped pores, respectively.
The values of porous texture parameters are sum- marized in Table 1. It is found that the average width (d) of slit-shaped pore is in good agreement with the mean size of micropores obtained from the intersection of the two straight lines of the t-plot. This result indicates clearly that the micropore are slit-shaped. Furthermore, it is of interest to note that the majority of parameters characteristic of porous texture increases by increasing the synthesis temper- ature or by decreasing the complexing ratio. As the more pronounced change in texture is induced by synthesis temperature, we have choice to study the effect of this experimental parameter on the catalytic activity of titania–alumina-supported molybdenum catalyst.
The effect of firing at 673 K and subsequent ad- sorption with four molybdenum atoms nm−2 on the surface area, pore volume and average pore size of titania–alumina support are presented in Table 1. A dramatic reduction of surface area and porous volume is verified after molybdenum impregnation on the titania–alumina support indicating that molybdenum
76 E.Y. Kaneko et al. / Applied Catalysis A: General 235 (2002) 71–78
Fig. 4. Effect of firing temperature (curves a–d) and aging (curves e–f) on the XRPD patterns of titania–alumina powder prepared with [acac]/[Ti] = 1.
is preferentially being deposited at the external sur- face, thus blocking the micropores. The effect of impregnation on the porous texture increases as the synthesis temperature decreases and results in the significant rise in the micropores contribution to the total porosity. Furthermore, the comparison of dried and fired samples reveals that the initial characteris- tics of porous texture are not considerably affected by thermal treatment. These results evidence the good thermal stability of titania–alumina support.
Further information about the effect of aging and firing temperatures on the thermal stability of titania–alumina support is shown in XRPD patterns presented in Fig. 4. Below 673 K, the XRPD pat- terns present broad peak profiles, typical of mate- rials containing nanocrystalline domains. The peak intensity increases, while their width decreases as the firing temperature increases indicating crystallite growth. The diffraction pattern of the sample fired at 973 K shows only the presence of anatase crys- talline phase. The XRPD patterns of impregnated supports (not presented) show similar behavior. Fur- thermore, a small increase in crystallinity is observed when the aging temperature is increased from 298 to 343 K.
The catalytic activities of the titania–alumina- supported molybdenum catalyst and that of commer- cial alumina support were evaluated in the thiophene HDS reaction carried out at different temperatures. Fig. 5a and b show the dependence of the mean
selectivity to butane formation with the temperature and the Arrhenius plot of reaction rate, respectively. The different curves correspond to titania–alumina supports prepared at different aging temperature. The apparent activation energy, calculated from the slopes of the curves, is essentially the same either for the supports obtained at different synthesis temper- atures (81 kJ mol−1 ± 5 kJ mol−1) or for the molyb- denum supported on commercial alumina sample (74 kJ mol−1 ± 5 kJ mol−1).
The presented experimental data show that the ag- ing temperature and the complexing ratio of titanium precursor affect the process of formation and aggre- gation of titania–alumina powders in different ways. The drastic decrease of the mean hydrodynamic di- ameter of particles from 70 to 4 nm, when the ratio [acac]/[Ti] increases from 0 to 0.5, is not followed by a considerable reduction of the surface area of dried powder (Table 1). This feature suggests the presence of aggregates in the sol prepared with [acac]/[Ti] = 0 and that the addition of a small amount of acac pro- motes the desaggregation of particles.
The power of dispersion of such complexing agent can be evaluated by comparing the mean hydrody- namic diameter measured in the sol with the aver- age size (lvs) of primary particles in the powder given
E.Y. Kaneko et al. / Applied Catalysis A: General 235 (2002) 71–78 77
Fig. 5. HDS reaction temperature dependence of the (a) mean selectivity to butane formation and (b) rate of reaction conver- sion of thiophene (Arrhenius plot) for supports prepared with [acac]/[Ti] = 1 at different aging temperatures.
lvs = f ρSBET
where ρ is the density of powder measured by helium picnometry (ρ = 2.7 ± 0.1 g cm−3) and f the shape factor, assuming values of 6, 4 and 2 for spherical (or cubic), equiaxial road and platelet geometry, respec- tively. Because the non-negligible contribution of ad- sorbed liquid layer to the size of particle measured by QELS, it may be higher than the true size of primary particles. This condition was only observed by using f = 2 in Eq. (2), that results in lvs = 2.3 nm for the length of plate-like particles in the sample prepared with [acac]/[Ti] = 0.5. Using f = 6 or 4 results in lvs higher or similar to the mean hydrodynamic size, what is an absurd. Furthermore, the plate-like geometry of primary particles is consistent with the presence of
slit-shaped micropores revealed from the N2 adsorp- tion analyses (Table 1) and with the morphology of titania–alumina sols previously observed by electron microscopy . Otherwise, the similarity between the average length of platelets and the hydrodynamic diameter indicates that particles are fully dispersed down the primary particle size. These findings indi- cate also that the acac ligand has a role of capping agent , that modifies the surface properties and prevents the aggregation of primary particles.
The overall effect of aging temperature on the char- acteristic of titania–alumina sol is in agreement with the results found by Yoldas  for bohemite sol . The sol aged at room temperature has a collapsed structure of folded bohemite particles and increasing the aging up to 323 K favors the formation of an ex- panded structure of unfolded bohemite particles . This transformation results from peptization, that leads to the rupture of interlamellae hydrogen bonds  and irreversible fracture of particles into smaller platelets (Fig. 1a) increasing considerably the surface area (Table 1). However, there are some interesting differences between our results and the Yoldas  finding. The first is the decrease in the tendency to peptization, indicated by the growth of hydrodynamic size (Fig. 1a) as the aging temperature increases up to 343 K, and the second is the absence of crystalline pseudo-bohemite phase in isolated powder (Fig. 4). The former can results from the well-known  decrease of the complexing power of the capping agent (acac) as the temperature increases. The later effect confirms the improvement of the thermal sta- bility of amorphous pseudo-bohemite achieved in titania–alumina system .
Otherwise, the presence of alumina hinders the anatase–rutile transformation (Fig. 4), increasing the thermal stability of the porous texture. In fact, an increase of the surface area and porosity is observed after firing the sample aged at 333 K (Table 1). This behavior can result from stress arising during dehy- dration process, leading to break down of particles and aggregates . On the other hand, for titania–alumina aged at 343, 353 and 363 K a slight decrease of the surface area and porosity and an increase of the mean pore size are observed after firing at 673 K (Table 1). This behavior results from the sintering process. This general feature is in agreement with previous finding  showing that the presence of alumina inhibits
78 E.Y. Kaneko et al. / Applied Catalysis A: General 235 (2002) 71–78
the crystallite growth, increasing the stability of the anatase phase.
It is noteworthy that despite of the employed sup- port, i.e. titania–alumina or alumina, the observed activation energy (Fig. 5a) is practically the same meaning that in all of the studied catalysts the active phase is alike and is constituted basically by MoS2. This result implies that the expected synergic effects associated to the titanium species formed during sul- fiding (e.g. TiSx)  either did not occur or, if so, the amount of such TiSx species is too low to improve the catalytic activity.
The performance of the titania–alumina-supported molybdenum catalysts is somewhat inferior to that pre- sented by the Mo/Al2O3 sample (Fig. 5b), meaning that the MoS2 dispersion on the former is lower than on the later. Since all of the catalysts were synthe- sized using the same methodology and were sulfided using the same temperature condition, the explanation for different dispersions has to lie on the nature of the supports. In fact, while the titania–alumina supports are essentially microporous and present a pronounced decrease in surface area after molybdenum incorpo- ration as shown in Table 1, the alumina is basically mesoporous and did not present such a decrease in sur- face area after molybdenum adsorption. The decrease in surface area observed for the titania–alumina sam- ples suggests that during metal adsorption there is a blockage of the microporous, resulting in large MoS2 particles, which present low catalytic activity. In fact, the butane selectivity results in Fig. 5a indicate that the titania–alumina samples indeed have larger MoS2 particles due to the lower value . Nevertheless, the textural results suggest that by increasing the ag- ing temperature micropores can be transformed into mesopores, which may improve the performance of titania–alumina catalyst as a support for HDT cata- lysts. One other way to overcome the porous blockage setback consists in changing the metal incorporation from the adsorption to the incipient wetness method, which leads to a better dispersion of the active phase.
This study has demonstrated that titania–alumina mixed oxides having a monodisperse distribution of nanometric particles, surface area superior to
200 m2 g−1 and mean pore size of about 1 nm can be prepared by the sol–gel route. The acac, used as complexing ligand, has a role of capping agent pre- venting the aggregation of the primary nanoparticles. The decrease of the complexing power of the capping agent (acac) as the temperature increases favors the aggregation of particles leading to control the pore size distribution. The slit-shaped micropores, formed by irregular packing of platelet primary particles, are partially blocked by molybdenum sulfide leading to catalysts with a somewhat lower activity than that of alumina-supported catalyst.
We would like to acknowledge Sandra S.X. Chiaro of CENPES-PETROBRAS for his scientific interest on this study and to FAPESP and CNPq (Brazil) for the financial support.
 E. Olguin, M. Vrinat, L. Cedeño, J. Ramirez, M. Borque, A. Lopez-Agudo, Appl. Catal. A 165 (1997) 1.
 M. Breysse, J.L. Portefaix, M. Vrinat, Catal. Today 10 (1991) 489.
 F. Luck, Bull. Soc. Chim. Belg. 100 (1991) 781.  C. Pophal, F. Kameda, K. Hoshino, S. Yoshinata, K. Segawa,
Catal. Today 39 (1997) 21.  J. Ramirez, L. Ruiz-Rairez, L. Cedeño, V. Harle, M. Vrinat,
M. Breysse, Appl. Catal. A 93 (1993) 163.  L.G. Karakchiev, T.M. Zima, N.Z. Lyakhov, Inorg. Mater. 36
(2000) 709.  S. Luo, L. Gui, X. Fu, Y. Tang, Mater. Res. Soc. Symp. Proc.
436 (1994) 445.  J.A. Montoya, P. del-Angel, T. Viveros, J. Mater. Chem. 11
(2001) 944.  B.E. Yoldas, Am. Ceram. Soc. Bull. 54 (1975) 289.
 A.C. Pierre, D.R. Uhlmann, J. Am. Ceram. Soc. 70 (1987) 28.
 E. Scolan, C. Sanchez, Chem. Mater. 10 (1998) 3217.  L. Shi, N. Wong, J. Mater. Res. 14 (1999) 3599.  S.J. Gregg, K.S.W. Sing, Adsorption, Surface Area and
Porosity, Academic Press, London, 1982 (Chapters 1 and 6).  W.D. Harkins, G. Jura, J. Am. Chem. Soc. 66 (1944)
1362.  G.E.S. Brito, C.V. Santilli, S.H. Pulcinelli, Colloid Surf. A97
(1995) 217.  Y. Okamoto, H. Tomioka, T. Imanaka, S. Teranishi, J. Catal.
66 (1980) 93.