Langmuir 1989,5, 1393-1393 sistent with the parallel orientation of 2000-A-thick HT films on Si. The dearth of active sites to induce perpendicular orientation by reaction with the edges of the MoS, crystallites causes parallel orientation and no chemical bonding to the substrate. In AT films, however, the 3.5-eV MoS, peak appears to grow in from higher loss energy. This is a speculative observation and, if true, would imply that the d,z orbital of MoS, is involved in bonding to the substrate, perhaps through Mo-S-0 or Mo-0 linkages. Thus, the native oxide covered surface of Si (and presumably S i c and Si,N,) is sulfated upon initial exposure to the plasma during deposition. Elemental S may
1393
be deposited also. In HT films, MoS, is deposited on top of or beside the sulfates and is unaffected by them. In AT films, the initial MoS, may be bonded through the sulfates or Mo-0 linkages.
Acknowledgment. I thank R. Bauer for preparing the samples, S. A. Jackson for performing the AES and EELS measurements, and J. L. Childs for the XPS measurements. Funding for this work was provided by the Defense Advanced Research Projects Agency and the United States Air Force Space Division Contract No. F04701-85-C-0086. Registry No. MoS,, 1317-33-5;Si, 7440-21-3.
Control of Metal Distribution in Supported Catalysts by pH, Ionic Strength, and Coimpregnation Eli Ruckenstein* and Prakash Karpe Department of Chemical Engineering, State University of New York at Buffalo, Buffalo, New York 14260 Received January 25, 1989. I n Final Form: May 19, 1989 A model which accounts for transient processes comprising diffusion, ionic migration, and adsorption is developed for a cylindrical pore. The model employs the Poisson-Boltzmann equation for the electrical potential set up by the pH and ionic strength dependent charge of the surface of the pore. The governing equations are solved by the finite element method. The numerical calculations have been carried out for the case of H,PtCl,/y-alumina. The effects of the ionic strength and pH of the impregnating solution and of the addition of coimpregnating species on the catalyst loading and distribution profiles are investigated. The catalyst loading decreases monotonically with increasing ionic strength and passes through a maximum when the pH varies. Different coimpregnants yield a variety of catalyst distribution profiles, such as uniform and bell-shaped profiles. The model is able to successfully explain qualitatively the experimental observations reported in the literature. The results are also compared with those obtained when the ionic dissociation and electrokinetic effects are neglected as well as with those in which the ionic effects are included but the electrokinetic ones are neglected. Substantial differences are found to exist between their predictions. 1. Introduction
The supported noble metal catalysts are often prepared by impregnation, i.e., by contacting the support pellets with a solution of an appropriate compound of the active metal. Either dry impregnation, in which the initial state of the support is dry, or wet impregnation, in which the support pellets are first wetted in an excess amount of solvent (usually water), is employed.' The technique of wet impregnation has been applied to the manufacture of industrially important catalysts, such as Pt/A1,0,, Pd/A1,0,, Ni/A1,0,, etc. In wet impregnation, the metal complex, present in ionic form, diffuses into the support pores and is adsorbed on the support surface. The adsorption profile of the metal inside the pellet influences the activity, selectivity, and life span of a ~ a t a l y s t ; therefore, ~.~ the control of the catalyst load-
* Author to whom the correspondence should be addressed. (1) van den Berg, G. H.; Rijnten, H. Th. In Preparation of Catalysts ZI; Delmon, B., Grange, P., Jacobs, P., Poncelet, G., Eds.; Elsevier: Amsterdam, 1979; p 959. (2) Becker, E. R.; Nuttall, T. A. In Preparation of Catalysts ZI; Delmon, B., Grange, P., Jacobs, P., Poncelet, G.,Eds.; Elsevier: Amsterdam, 1979 p 959.
0743-7463/89/2405-1393$01.50 f 0
ing and of the distribution profile constitutes an important aspect of catalyst design. Several models, reviewed in ref 4, have been proposed to explain the various physicochemical phenomena involved in the wet impregnation process. These models assume Fickian diffusion of the active species inside the pores and Langmuir adsorption for their adsorption on the support surface. They neglect, however, the occurrence of ionic dissociation and electrokinetic phenomena inside the substrate and, thereby, fail to explain the influence of the pH and ionic strength of the impregnating solution on the catalyst distribution. The scope of this paper is to examine the role of the ionic dissociation and electrokinetic phenomena during the catalyst preparation by wet impregnation. A first attempt in this direction was made in a recent paper from this lab~ratory.~ The present paper contains a greater variety of cases and includes the treatment of the coimpregnation as well as a com(3) Machek, V.; Hanika, J.; Sporka, K.; Ruzicka, V.; Kunz, J.; Janacek, L. In Preparation of Catalysts IZI; Poncelet, G.,Grange, P., Jacobs, P., Eds.; Elsevier: Amsterdam, 1983;p 69. (4) Lee, S.Y.;Aris, R. Catal. Reu. Sci.-Eng. 1985,27,2, 207. (5) Karpe, P.A.; Ruckenstein, E. Colloid Polym. Sci. 1989,267, 145.
0 1989 American Chemical Society
1394 Langmuir, Vol. 5, No. 6, 1989 parison with the case in which ionic dissociations are considered but electrokinetic effects are neglected. The commonly employed oxide support materials, such as silica and alumina, develop a charge on the surface of their pores when they are contacted with aqueous solutions. The protonation-deprotonation equilibria of the surface hydroxyl groups of the oxides give rise to positive and negative sites on the surface, the development of these charged sites being strongly dependent on pH and ionic strength. The adsorption of the impregnating ions on the surface of the pores alters the charge of the surface. The pH a t which the net surface charge is zero is referred to as the point of zero charge and is denoted by PZC. At a pH < PZC, the surface of the substrate has a net positive charge, and at a pH > PZC a net negative one. As defined above, the point of zero charge is also the isoelectric point (IEP). Thus, by varying the pH of the impregnating solution, the distribution of the metal on the support can be controlled in a desired way. This has been experimentally demonstrated by Benesi et al., for the cases of Pt(NH3),C1,/A1,03 and Pt(NH,),Cl,/SiO,, by Komiyama et al. for NiC1, on yAl,03,' by Heise and S c h ~ a r z *for ' ~ the case of H,PtCl,/ Al,03, by Contescu and Vassl' for Pd(NI3$,Cl2/yA1,03, and by Chu et al. for NiCl, on y-A1203. The pH-dependent charge distribution on the surface sets up an electrical potential inside the pores of the support. An electrical potential gradient inside the pores is generated by the pH gradient along the pore, which is due to the diffusion of H+ ions into the pores, and by the diffusion of other ions present in the solution. The ionic strength of the impregnating solution influences the electrical potential which, in turn, affects the diffusion and migration of the ions into the support pores. Hence, the amount and profile of the deposited metal can also be controlled by varying the external ionic stren th of the impregnating solution. Heise and Schwarz'.' P have experimentally demonstrated that the total uptake of the metal by the support decreases when the ionic strength of the impregnating solution increases. Further, introducing a coimpregnating species that competes with the impregnant for the adsorption sites makes it possible to affect the catalyst distribution in such a way as to obtain different shapes of the distribution profile. The effects of various inorganic and organic acids as competing species on the deposition of Pt on alumina have been extensively investigated experimenta11y.2,13-15a
The purpose of the present paper is to develop a theoretical framework that would account for the electrochemical phenomena that occur during the preparation of catalysts by wet impregnation. The effects of variation of the pH and ionic strength of the impregnating (6) Benesi, H. A.; Curtis, R. M.; Studer, H. P. J. Cata. 1968,IO,328. (7) Komiyama, M.; Merrill, R. P.; Harnsberger, H. F. J. Catal. 1980,63,35. (8) Heise, M. S.;Schwarz, J. A. J. Colloid Interface Sci. 1985,107, 237. (9) Heise, M. S.;Schwarz, J. A. In Preparation of Catalysts ZV; Delmon, B., Grange, P., Jacobs, P. A., Poncelet, G., Eds.; Elsevier: Amsterdam, 1987;p 1. (10) Contescu, Cr.; Vass, M. I. Appl. Catal. 1987,33,259. (11) Chu, P.; Petersen, E. E.; Radke, C. J. J. Catal., in press. (12) Heise, M. S.;Schwarz, J. A. J. Colloid Interface Sci. 1986,113, 55. (13) Maatman, R. W.;Prater, C. D. Ind. Eng. Chem. 1957,49(2), 253. (14) Shyr, Y.S.; E r s t , W. R. J. Catal. 1980,63,425. (15) (a) Jianguo, W.;Jiayu, Z.; Li, P. In Preparation of Catalysts IIfi Poncelet, G., Grange, P., Jacobs, P., Eds.; Elsevier: Amsterdam, 1985; p 57. (b) Berube, Y.G.; De Bruyn, P. L. J. Colloid Interface Sci. 1968, 28, 92.
Ruckenstein and Karpe
Figure 1. Cylindrical pore model.
solution, the influence of the nature of the substrate, the effect of coimpregnating species on the distribution profile, and the amount adsorbed are examined within this framework. The model has been illustrated for the case of impregnation of y-alumina with Pt. An aqueous solution of H2PtCI, is considered as the impregnating medium. NaI is used to adjust the ionic strength, while the acid HI and the base NaOH are used to vary the pH of the impregnating solution. Since I- is only weakly adsorbed,15bits adsorption is neglected here. NaN03, NaCl, and sodium citrate are considered as the coimpregnating species.
2. Theory The governing equations that describe wet impregnation are based on the following assumptions: (1) The porous structure of the catalytic support pellet comprises a large number of long, narrow, uniform, cylindrical pores. (2) The irregularity in shape and length of the pores can be accounted for by introducing a tortuosity factor, T.', (3) The pores are initially filled with water. (4)The volume of the impregnating solution is much larger than the total void volume of the support. ( 5 ) The adsorption of impregnating species on the active sites of the surface of the pores is isothermal. (6) The diffusivities of the ionic species and the dielectric constant of the medium are independent of the species concentrations. Any additional premises that may be deemed germane to the development of the model will be stated whenever required. A single cylindrical pore of half-length 1 and radius a is considered, and a cylindrical coordinate system ( x , r ) is located at the center of the open mouth of the pore as shown in Figure 1. The half-length I of the tortuous pores is estimated from the relation 1 = TR where R, is the effective radius of the pellet. Since t%e pore isopen at both ends, the concentration of each species is the same at both ends of the pore. Hence, the governing equations can be applied over half-length only. 1. Species Conservation Inside the Pore. Due to the presence of the electrical potential, the axial flux of an ionic species inside the pore is governed by the NernstPlanck equation:
JZi= -Diaci(x,r,t)/ax- MjziciFax*(x,r,t)/ax (1) where subscript i denotes the ionic species i, F is the Faraday constant, x*(x,r,t)is the electrical potential, ci(x,r,t) and zi are the local concentration and the charge number of species i (zihas a positive sign for the cation and a negative sign for the anion), and t is time. Regarding the ionic species i, the following convention has been adopted: i = 1 represents the impregnant (assumed here, as discussed later, to be HPtC1,- ion at pH 0
0.00
l--!!LJ
0.00
0.06
0.10
REDUCED P O R E LENGTH
Figure 2. Dependence of catalyst adsorption profile on neutral site adsorption equilibrium constant K (no other electrolyte added): (a) K , = 1.5 X lo6 cm3/mol, K , = 1.0 X lo4 cm3/mol, (c) K, = 1.0 X lo3 cm3/mol, (d) K , = 1.0 X 10' cm3/ mol, (e) K , = 0 cm3/mol. The values of the other parameters are given in Table I.
b)
on the kind of double-layer model used to estimate these values. IyerZgreports ranges of K,, values for C1- and NO3- ions which are based on a model that accounts for the Stern layer. No values of K,, for HPtCb- ions adsorbed on alumina have been reported in literature. There are, however, experimental datalsgwhich can be used to evaluate the values of the various equilibrium constants consistent with the double-layer model employed here. The equilibrium experimental data for HPtC1,- and NO; on y-alumina of Heise and Schwartzgwere employed for this purpose. The values of the equilibrium constants K , and Kb for alumina and K,, for NO3- in the ranges reported in the above references were used as initial guesses, while the value of Kc, for HPtC1,- was taken in the range of values of Kc2 for C1- as an initial guess since it was noted' that their adsorption strength is similar. The total equilibrium concentration of adsorbed HPtC1,- ( N , Ncl)was calculated on the basis of the double-layer model employed here. The values of the equilibrium constants that yielded the best fit with the data were determined by the Simplex method.33 The values of K i (for i = 1, 2) were taken to be about 2 orders smaller than the corresponding values of KCj.Since the adsorption of an active ionic species (i = 1, 2) on the neutral sites (MOH) involves van.der Waals interactions and that on the positive sites (MOH2+)both van der Waals and ionic interactions, the values of K j are expected to be smaller than those of KCi. However, as seen from Figure 2, changes in the value of K , by 5 orders of magnitude do not affect the adsorption profile of the impregnant significantly. A similar procedure was followed for estimating the values of K,, for the C1- and for the citrate ion from the experimental data of ref 1 and 15, respectively. For the divalent PtC12- ion, the values of K , and K,, were taken equal to those of the monovalent HPtC1,- ion. This can be justified by the fact that the two charges of PtCl,,-
+
(32) Vordonis, L.; Akratopulu, A.; Koutaoukos, P. G.; LycourghiOtis, A. In Preparation of Catalysts IV; Delmon, B., Grange, P., Jacobs, P., Poncelet, G., Eds.; Elsevier: Amsterdam, 1987; p 309. (33) Beveridge, G. S.; Robert, S. S. Optimization: Theory and Practice; McGraw Hill: New York, 1970.
0.0
0.1
0.2
0.3
0.4
REDUCED PORE LENGTH
Figure 3. Variation of catalyst distribution profile with the external concentration of the impregnant (col) (in the absence of an additional electrolyte) by considering the effect of electrokinetic and ionic dissociation phenomena (-), by neglecting the electrokinetic phenomena (-), and by neglecting both electrokinetic and ionic dissociation phenomena 7- - -): la) col = 0.25 X M. I , = 0.75 X M. DH, = 3.6: (b) cor = 0.5 X 10-~ M, zo = 0.5 io-^ M, PH = 3 . 3 , ( ~ ) " ~=~1.0 , X M, zo = 1.0 X M, Io = 2.0 X M, pHo = 3.0; (dl col = 2.0 X M, pHo = 2.7. The values of the other parameters are given in Table I.
are sufficiently far apart for only one of the charges to appreciably interact with the surface. The estimated values of all the equilibrium constants are given in Table I. 4. Results and Discussion
1. Effect of External Concentration of the Impregnant (col) and Importance of Electrokinetic Effects. The effect of the external concentration of the impregnant, col on the adsorption is illustrated in Figure 3. When col is increased, the concentration profile of the adsorbed impregnant evolves into an egg-shell shape. The depth of penetration and the amount adsorbed increase with increasing col. Figure 3 also underlines the importance of the electrokinetic phenomena during wet impregnation (compare the dotted curves, where calculations are made by equating the electrical potential to zero in eq 10-15 with the solid curves which account for the electrokinetic phenomena). Figure 3 also contains a comparison with the results of the commonly used model based on Fickian diffusion and Langmuir adsorption (where both the ionic dissociation and electrokinetic effects are neglected). The model is discussed in detail by Lee and A r k 4 The importance of both the ionic and the electrokinetic phenomena is clearly evident when the dashed curves (calculated on the basis of the model of ref 4) and the solid curves in the figure are compared. The calculations have been carried out by assuming complete dissociation of H,PtCl, for pHo 13.6 and partial dissociation for pHo
