J . Chem. SOC., Faraday Trans. I , 1984, 80, 1579-1594 Metal Dispersions on Zirconium Phosphates Part 2.-Hydrogen Reduction of Silver(1)-exchanged a-Zirconium Phosphate BY SOOFIN CHENG-~ AND ABRAHAM CLEARFIELD* Department of Chemistry, Texas A&M University, College Station, Texas 77843, U.S.A. Received 2 1st September, 1983 The kinetics of the reduction by hydrogen of Ag+ ions in a-zirconium phosphate and the formation of silver particles have been studied. Reduction proceeded under relatively mild conditions. The time dependence of the percentage of Ag+ ions reduced in this reaction is always expressed by an S-shaped curve with a maximum reaction rate at ca. 15 % conversion. An initial induction period was observed and is suggested to result from the extremely slow rate of initiating silver nuclei.The remainder of the reaction rate was found to conform to an Elovich-typed equation. The initial rate was found to be directly proportional to the Ag+ loading and to the surface area of the support but proportional to the half-power of the hydrogen pressure. It is also suggested that the reduction reaction occurs on the surface with the reaction itself determining the rate after the induction period. Supported metals are widely used as catalysts. Metal is usually introduced to the support as cations, either from aqueous solutions or from suspensions, by one of several processes such as impregnation, ion exchange, deposition or coprecipitation, followed by drying and hydrogen reduction. Such catalysts, which contain very small metal crystallites, have certain definite advantages over the bulk metal, which can be in the form of films, wires or powders, because the high dispersion of the metal leads to a high surface area and to increased resistance to sintering.However, the dispersion of metal particles and the activity of the catalyst strongly depend on the conditions under which the catalysts have been prepared and on the particular support ad0pted.l Many efforts have been made to develop finely dispersed metal particles by reduction of transition-metal ions in zeolites.2+ The possibility of these catalysts being used as highly active and/or bifunctional catalysts is the incentive for such work. Kinetic and physicochemical studies have been reported on the formation of metal particles in Nevertheless, incomplete characterization is an inextricable problem owing to the diversity of sites existing in zeolites and to the limitation of analytical methods employed in the studies.There is theoretical and practical interest in studying the redox behaviour of transition-metal ions in zirconium phosphates. They form synthetic hydrogen- containing analogues of clays and possess cages akin to those of zeolites. As a comparison, zirconium phosphates provide a simpler system for studying metal-particle formation. a-Zirconium phosphate, Zr(HPO,), H,O, is a crystalline ion-exchanger. It has a layered structure with an interlayer distance of 7.56 A. Ion exchange occurs by replacement of the orthophosphate hydrogen by cations which then occupy positions between the layers.' The flexible interlayer distance in zirconium phosphate makes it possible to detect the formation of small metal clusters in between the layers by X-ray powder diffraction methods.In contrast, structural differences in zeolites t The work reported in this paper formed a part of the Ph.D. Thesis of S. Cheng. 1579 52-21580 METAL DISPERSIONS ON ZIRCONIUM PHOSPHATES are not easily detected using X-ray diffraction techniques because of their rigid structures. From a practical point of view, studying the redox behaviour of cations in zirconium phosphate (often referred to as ZrP) will lead to a better understanding of the durability and applicability of this material in various catalytic reactions. In previous work8 the reduction of CulI by hydrogen in copper-exchanged a-zirconium phosphate, ZrCu(PO,),, was found to proceed in two stages.Below ca. 150 Torr* the product was ZrCuH(PO,),, as CuI1 was reduced to Cul. Then, further reaction with hydrogen at pressures above 150 Torr yielded copper metal and A-zirconium The rates of both reactions were reported to conform to an Elovich-type equation. In the present work, silver reduction is investigated. Since silver reduction is only a one-stage process, the study of this simpler system is expected to provide more information on the metal-formation step. The influence of hydrogen pressure, temperature, AgI loadings and the surface area of a-zirconium phosphate on the process is examined. EXPERIMENTAL SAMPLE PREPARATION A series of zirconium phosphate samples with different crystallinities was prepared by refluxing the zirconium phosphate gel in H3P04 solution for periods of time as described by Clearfield et aZ.l0 These samples are labelled ZrP (4.5: 48), (6:48), (9: 48), (12: 50) and (12: 336), where the first number represents the molar concentration of H3P04 acid and the second number the duration of reflux in hours.The crystallinity increases with the reflux time and concentration of acid. Silver-exchanged zirconium phosphate samples were prepared as described elsewhere. l1 Samples for reduction were obtained by dehydrating the air-dried, silver-exchanged zirconium phosphates at CQ. 100 "C in a vacuum overnight and were kept in a vacuum desiccator. CHEMICALS Reagent-grade silver acetate (Fisher Scientific Co.) was used as the source of Ag+.Ultrahigh-purity (99.999 %) hydrogen was obtained from Matheson Gas Products. Zirconium phosphate gel was obtained from Magnesium Elektron Inc. Reagent grade phosphoric acid (Fisher Scientific Co.) was used without further purification. Distilled deionized water was used throughout. APPARATUS As mentioned in a previous paper,* the apparatus in which the reductions were carried out consists of a U-tube Pyrex reactor connected to a recirculatory loop which forms part of a vacuum system. The temperature of the reaction zone inside the reactor was read via a thermocouple held inside the sample and connected to a Doric digital trendicator model no. 400A. A side-arm of Pyrex tubing was attached for collecting samples in situ, thus avoiding any exposure to the atmosphere.In order to minimize the dead space, most of the assembly (loop) was made of capillary tubing. The total volume of the system was ca. 100 cm3 and that of the reactor was ca. 27 cm3. The reactor was heated by means of a well-insulated tubular furnace; the heating rate and the temperature were regulated using a Weathermeasure temperature controller. A Pyrex tube for blowing cold air was inserted inside the heating zone of the furnace to conduct away heat evolved from the reaction. The temperature control of the reaction zone in the oven with this air circulation was ca. 1 "C. The pressure changes occurring during the process of reaction were monitored by a Validyne differential-type digital manometer connected to a linear single-pen recorder.The precision of the manometer was f 0.1 Torr. REDUCTION PROCEDURES A weighed amount of dehydrated sample (ca. 0.2 g) was placed inside the reactor and spread over supported glass wool to achieve maximum surface exposure. The reactor was mounted onto the vacuum system and heated at the reduction temperature for > 5 h under vacuum to * 1 Torr = 101325/760 Pa.S. CHENG AND A. CLEARFIELD 1581 remove any absorbed moisture which may have been picked up during the transfer process. The reduction was started by bleeding-in a known amount of hydrogen which was kept in the outer chamber at a known pressure. The hydrogen uptake was followed by recording the decrease in pressure with time. The end of the reaction was determined when no further change in pressure could be detected for ca.30 min. Part of the reduced sample was examined by X-ray powder diffraction methods immediately after the reaction. The rest of the sample was transferred to the side-arm tube and sealed off under vacuum pressure for further analysis. The amount of Agl reduced was then calculated from the amount of hydrogen consumed. INSTRUMENTAL The powder diffraction patterns were obtained using a Seifert-Scintag Pad I1 X-ray diffractometer with Cu Ka radiation. The average silver particle size was determined from line-broadening measurements on the (1 1 1) reflection of silver metal with the aid of Scherrer's equation.12 These values were not corrected for errors resulting from instrument or strain. The data were used only as a basis of comparison.The silver-particle dispersion was determined by means of scanning electron microscope with a Jeol model JSM-35 oscilloscope. RESULTS The reduction of anhydrous samples of Ag+-substituted a-zirconium phosphate, i.e. ZrAg,H,-,(PO,), where 0 < x d 2, was found to be an exothermic reaction. As hydrogen was consumed, the colour of the samples changed from white to sepia or olive, indicating the reduction of Ag+ ions to Ag atom aggregates. X-ray powder diffractions patterns of the reduced samples contained reflections at 28 = 38.2 and 44.4", which correspond to the Ag (1 1 1) and (200) planes, respectively. The reduction rate was found to be a function of the initial hydrogen pressure, the temperature of the reaction zone, the Ag+ ion loading, and the surface area of the zirconium phosphates.Kinetic studies were mainly carried out on highly crystalline zirconium phosphate (12:336) in order to avoid difficulties in temperature control caused by the abrupt release of enormous amounts of heat during the reduction of the high-surface-area samples. FORMATION OF METALLIC SILVER PARTICLES X-ray powder diffraction patterns of the completely reduced samples (12: 336) showed reflections due to silver metal and c-ZrP [composition Zr(HP0,),]7i rather than the A-ZrP (same composition) observed on reduction of CuII-ZrP. In addition, the high-surface-area samples gave different results. Instead of the &ZrP phase, reflections characteristic of an Ag-I1 phasell with composition ZrAg,~,,H,~,,(PO,), were observed along with the reflections due to silver metal.However, if the samples were only partially reduced, diffraction patterns characteristic of metallic silver and the anhydrous phases of Ag+-containing zirconium phosphate phases, which could be Ag-I [ZrAg,,,,H,.,,(PO,),], Ag-I1 or Zr(AgPO,),,* depending on the extent of reduction, were obtained. The Ag-I phase had an assumed composition of The sizes of the silver particles formed by the reduction were zvaluated by X-ray line-broadening measurements based on the width at half height of the Ag (1 1 1) peak. Table 1 shows the average silver particle size of the samples obtained by reducing Zr(AgPO,), (12: 336) at various temperatures and at different initial hydrogen pressures. The particle size does not seem to be a function of these two reaction conditions.It will be shown later that all the silver comes to the surface. Apparently ZrAg0.22Hl.78(Po4)2.'1 * The X-ray pattern of the fully exchanged phase exists over a range of compositions, ZrAg, H2JPO&, with x = 1.9-2."1582 METAL DISPERSIONS ON ZIRCONIUM PHOSPHATES 400 Table 1. Average silver particle size of the samples obtained by reducing Zr(AgPO,), (12: 336) at different temperatures (pH, = 580 Torr) or hydrogen pressures (T = 150 "C) (A) - - av. particle av. particle T/OC size/A p/Torr size/A 100 0 62 392 124 525 100 393 191 424 145 426 354 54 1 190 389 583 446 - - - - I I 1 1 1 I l l , , surface area of ZrP support/mZ g-' Fig. 1. Average Ag particle size of the samples obtained by (A) reducing Ag+-exchanged a-ZrP (1 2 : 336) with various Ag+ loadings ( T x 100 "C, p H 1 x 585 Torr) and (B) reducing Zr(AgPO,), as a function of surface area of the supports (T x 100 "C pH2 x 280 Torr).J.Chem. SOC., Faraday Trans. 1, Vol. 80, part 6 S. CHENG AND A. CLEARFIELD Plate 1 I: %- - n a m m (u .. 3 w r( I M n en (Facing p . 1582)J. Chem. SOC., Faraday Trans. I , Vol. 80, part 6 Y Plate 2 S . CHENG AND A. CLEARFIELDS. CHENG AND A. CLEARFIELD I583 the very small surface area of the support [2.2 m2 g-l for (12:336)] relative to the amount of silver deposited leads to the formation of relatively large particles with a wide distribution range. As the loading of Ag+ in the exchanger was reduced there was an almost parallel reduction in the average particle size as a function of loading [fig.1 (A)]. The effect of increasing the surface area of the support is shown in fig 1 (B). Again it is seen that the particle size decreases as the space available on the surface increases. The silver-particle distribution was examined by scanning electron microscopy. Plates 1 (A) and (B) show the silver particles on ZrP( 12 : 336) samples which contained 6.64 and 1.18 m equiv g-l of Ag+, respectively, before reduction. Reduction was carried to completion. The ZrP crystals have a shape resembling hexagonal platelets13 but the composite with silver is more ellipsoidal. The highly loaded sample gives a wide range of silver particle-size distributions [plate 1 (A)]; the particle diameter varies from ca. 0.6pm to a few hundred Angstroms. However, the size distribution on the low silver-loaded sample is relatively uniform, as shown in plate 1 (B).In order to understand further the formation and distribution of silver particles on the surface of a-ZrP, single crystals of a-ZrP with a diameter of ca. 0.1 cm were exchanged with silver acetate, reduced by H, and examined by the scanning electron microscope. For comparison, a single crystal of a-ZrP without Ag content was exampled by the same technique. Without a silver coating, both crystal faces are clean and smooth. For the single crystal containing Ag, the micrographs [plates 2(A) and (B)] show two kinds of silver-particle distribution, one on the side and the other on the base of the platelet. Plate 2(A) shows one of the cracks on the side of the platelet and the very thick layers of large silver particles which accumulate on its surface.However, a very uniform distribution of fine silver particles can be observed on the basal surface of the hexagonal platelet, as shown in plate 2(B). After the surface of the crystals had been examined, they were sliced through the side so that the interior layers could be revealed. Although no differences were discernible between the surface and interior layer of the pure a-ZrP crystal, as expected, marked differences were noticeable for the silver-coated crystal. Large silver particles were found to aggregate on the side surface of the platelet. The particle sizes become smaller and smaller as the particles approach the edge of the crystal. Silver particles also accumulate along cracks existing in the interior layer.This reveals that the silver particles can be located anywhere, providing space is available. KINETICS OF REDUCTION The reduction of silver-exchanged a-ZrP can be expressed as Zr(AgPO,), + H, -+ Zr(HPO,), + 2Ag. The kinetic curves for the reduction of dehydrated Zr(AgPO,), (1 2 : 336) as a function of temperature are shown in fig. 2 (A). These curves are typically S-shaped. The value of the percentage conversion, a, increases rapidly after the initial induction period and reaches a maximum increasing rate at a point between 5 and 20% conversion, depending upon temperature. After this the rate of increase of a slows down gradually. This variation is shown in fig. 2(B), where da/dt is plotted a a function of time. This type of kinetic curve has also been observed in other reduction s y s t e m ~ .~ ~ - ~ ~ However, attempts to represent the curves by a simple algebraic equation have not, on the whole, been successful. The best fit was reported with the Elovich equation, as it applies to most of the data in the middle part of the curves.8v l4 In the present study the Elovich equation was found to fit the curves in the range a = 0.1-0.8, where a is the extent of conversion. This portion of the reaction will be treated separately from the initial induction period or accelerating rate period (ARP).1584 100 80 h 60 G .- E 2 40 > 20 04 METAL DISPERSIONS ON ZIRCONIUM PHOSPHATES tlmin Fig. 2. (A) Kinetic curves for the reduction of (12:336) Zr(AgPO,),; (B) time dependence of the reduction rate at various temperatures.A, 86; a, 104; 0, 121 and 0, 132 "C. INDUCTION PERIOD The reduction reactions are characterized by an initial induction period which extends to a x 0.1 and is not fitted by the Elovich equation. The range of the induction period was found to depend upon the reduction temperature, hydrogen pressure, the Ag+ loading and the surface area of the support. Similar induction periods haveS. CHENG AND A. CLEARFIELD 1585 commonly been observed in solid reaction ~ y s t e m s . l ~ - ~ ~ Garner et aZ.18 suggested that the induction periods were mainly due to an abnormally slow rate of nuclei growth. RATE OF REDUCTION REACTION The Elovich equation, which fits the middle part of the kinetic curves, was originally applied to the rate of adsorption of gases on solid surface.21q22 The rate decreases exponentially with an increase in the amount (or fraction) a of gas adsorbed: daldt = Re-aa. (2) The exponent a is constant for a given sample and for fixed experimental conditions.The other constant, R, represents the initial rate when a = 0. The integrated form of eqn (2) can be written as with a = (2.3/a) log (t+ to) - (2.3/a) log to to = l/(aR). (3) (4) With a correct choice of to, the plot of a as a function of log(t+t,) should give a straight line with a slope of 2.3a. Then, the initial rate, R, is obtained from eqn (4). The disposable parameter, to, is found by trial and error; if to is too small the curve of a against log (t+ to) is convex, and if to is too large the curve is concave.23 The resultant data for the temperature-dependent kinetic curves (shown in fig.3) are collected in table 2. A general expression for any rate law is r = -dC,/dt = kC$ Cb,... CE lo3 KIT Fig. 3. Arrhenius plot of In R against 1/T for the reduction of (12: 336) Zr(AgPO,),.1586 METAL DISPERSIONS ON ZIRCONIUM PHOSPHATES Table 2. Values of zo, a, R and R' for the reduction of Zr(AgPO,), (12: 336) at various temperatures; pH2 M 580 Torr Elmo1 (Ag+) g-* T/OC 2, a R (% min-I) (catalyst) min-' 85.8 7.0 0.0150 9.50 3.82 x 10-4 104.0 2.0 0.0193 25.90 1.04 x 10-3 121.0 1.0 0.0203 49.30 1.98 x 10-3 132.0 0.5 0.0165 121.20 4.88 x 10-3 tlmin t/min Fig. 4. (for legend see opposite).S. CHENG AND A. CLEARFIELD 1587 t/min Fig. 4. Kinetic curves for (A) the reduction of Ag+-exchanged (1 2 : 336) a-ZrP with various Ag+ loadings [ T x lOO"C, pH, x 585 Torr; a, 0.324; A, 1.18; 0, 3.09 and 0, 6.64 mequiv g-l (a-ZrP)], (B) the reduction of (12: 336) Zr(AgPO,), at various hydrogen pressures [T x 100 "C; pH2 = 0, 280; 0, 376; A, 478 and 0, 587 Torr] and (C) the reduction of Zr(AgPO,), with various surface areas of the support [T x 100 "C, pH2 E 280 Torr; 0, ZrP (9: 48); A, ZrP (12:50) and 0, ZrP (12:336)].where C, is the concentration of the reactant A and CB,. . ., CN are the concentrations of either the reactants or products. For the reduction of Ag+-exchanged ZrP the rate law can be written in the following form: R' = - d[Ag+]/dt = k[Ag+]'pE2 Sn.. . (6) where p is the partial pressure of hydrogen, S is the surface area of the support and R' is the initial rate obtained from the product of R values from the Elovich plots and the initial concentrations of Ag+ in ZrP.According to the Arrhenius equation, the rate constant, k , is a function of temperature in the form k = Aexp(-E,/RT) (7) where A is the frequency factor, E, is the activation energy of the reaction and R is the ideal gas constant. Therefore the initial rate of the reduction is temperature-dependent : R' = Aexp(-Ea/RT)[Ag+l1pg2S n.... (8) The temperature effect was studied by keeping all other variables constant. Thus eqn (9) (8) reduces to with the constant B = A[Ag+Izpg2 Sn. A plot of In R' against 1/T should be linear, and the slope gives the value of -E,/R (fig. 3). The activation energy obtained has a value of 15.2kO.2 kcal mol-l.? The values of I, rn and n in eqn (6) were then determined by examining the kinetic curves obtained by varying the concentration of R' = Bexp (- E,/RT) t 1 cal = 4.184 J.1588 METAL DISPERSIONS ON ZIRCONIUM PHOSPHATES 7 9- .z 7 - E 8- Table 3.Values of to, a, R and R for the reduction of Ag+-exchanged ZrP at various Agf loadings, initial hydrogen pressures and surface areas of the support expt. no. Ag+/ T pH2 mequiv g-l S ZrP /"C /Tom (a-ZrP) /m2 g-l to (12:336) 101 280 6.64 2.2 5.0 (12:336) 102 376 6.64 2.2 4.0 (12:336) 103 478 6.64 2.2 3.8 (12:336) 99 587 6.64 2.2 3.0 (12:336) 101 593 3.09 2.2 17.0 (12:336) 99 586 1.18 2.2 50.0 (12:336) 102 584 0.034 2.2 43.0 (12:50) 100 282 6.64 5.0 2.0 (9:48) 101 281 6.64 7.8 2.0 U 0.0224 0.020 6 0.020 5 0.023 7 0.008 04 0.00420 0.005 16 0.0148 0.0105 R (% min-l) 8.93 12.10 12.80 14.10 7.32 4.76 4.51 33.80 47.60 R'lmol (Ag+) g-l (catalyst) min-l 3.59 x 10-4 4.87 x 10-4 5.15 x 10-4 5.67 x 10-4 5.26 x 10-5 1.53 x 10-5 1.36 x 10-3 1.92 x 10-3 1.78 x iO-* 11 I I I I I I I I 1 2 3 4 5 6 7 8 9 ' p",/lo* tom 3 [ Ag+l /mequiv g-' (a-ZrP) S/mZ g-' Fig.5. Logarithmic plot for variation of initial reduction rate with (a) the initial hydrogen pressure, (b) Ag+ loading and (c) the surface area for the support.S. CHENG AND A. CLEARFIELD 1589 one component and keeping those of the other components constant. Fig. 4(A), (B) and (C) show the kinetic curves as a function of Ag+ loading, initial hydrogen pressure and the surface area of the support, respectively. These curves have been fitted to Elovich plots and the resultant values of to, a, R and R' are listed in table 3.In experiments no. 1-4 the initial hydrogen pressure varied from 280 to 587 Torr. The value of rn in eqn (6) is equal to the slope of the straight line when log R' is plotted against logpH*, as shown in fig. 5(a). The experimentally determined value is 0.59. In experiments no. 4-7, the loading of Ag+ ions has been varied from 6.64 to 0.324 mequiv (Ag+) g-l (a-ZrP). Fig. 5 (b) shows that by plotting log R' against log [Ag+] a straight line results with a slope of 1.2, which is the value of I in eqn (6). The effect of the surface area of the support is obtained from experiments no. 1, 8 and 9. The surface area has been varied from 2.2 to 7.8 m2 g-l (a-ZrP) and initial hydrogen pressure has been kept at 280 Torr.The plot of log R' against log S is shown in fig. 5(c). The slope has a value 1.4, which is the value of n in eqn (6). Eqn (6) can thus be written in the form R' = -d[Ag+]/dt = k'[Ag+]1*2&fBS1.4 where k' is the rate constant independent of the Ag+ loading, the hydrogen pressure and the surface area of the support. DISCUSSION The reduction of Ag+ ions in a solid system is generally found to proceed easily under mild conditions. Several studies on the reduction of silver(1) oxide with hydrogen have been r e p ~ r t e d . ~ ~ - ~ ' The reduction reaction has been found to proceed at a measurable rate at temperatures above 40 "C with an activation energy of 8.0-1 5.0 kcal mol-l. More recently, several articles dealing with the reduction- reoxidation behaviour of silver(1) ions in zeolite systems have been p ~ b l i s h e d .~ ~ ~ ~ 28-30 None of the kinetic curves reported in those studies showed the initial induction period characteristic of the present system. In the research performed by Iwamoto et aL5 the activation energy was determined by plotting the logarithm of the initial rates of reduction, calculated from the amount of hydrogen .consumption for 10 s, against the reciprocal absolute temperatures between 273 and 741 K. The value obtained was 23.4f1.6 kJmol-l or 5.6k0.4 kcal mol-l, which is much smaller than that found in the present study (15.2 cal mol-l). This large difference is attributed not only to the different methods used in determining the initial reduction rates but also to the possibility that the reaction mechanisms and rate-determining steps are different in these two systems.On the other hand, the reduction of the Ag+ ions in a-zirconium phosphate was found to proceed much more rapidly than in X-type zeolites. For example, Iwamoto et aL5 reported that the zeolite-X sample consumed an amount of hydrogen corresponding to the complete reduction of Ag+ ions within 10 min at 450 "C. In contrast, complete reduction of Zr(AgPO,), (12: 336) could be achieved in a few seconds at much lower temperatures. Beyer et aL6 found a marked temperature dependence for the reduction of Ag+ in zeolites. At temperatures lower than 430 K, they could only achieve a maximum reduction of ca. 60% of the silver ions. They reported that two different mechanisms were operative at high and low temperatures.Above 430 K the reaction is first-order in Ag+ concentration and independent of the hydrogen pressure and has an activation energy of 97.6 kJ mol-1 (24.2 kcal mol-l). The migration of cations from SI sites of Y zeolites is thought to control the rate. At low temperatures the reaction is dependent1590 METAL DISPERSIONS ON ZIRCONIUM PHOSPHATES on hydrogen pressure with an activation energy of 40 kJ mol-l (9.5 kcal mol-l). The reaction was suggested to be a catalysed reaction with the regeneration of the surface-active sites as the rate-limiting step. Some of the impurities in the Y zeolites were thought to be active sites, although this could not be proved. The reduction of silver(1) in zirconium phosphates was found to proceed at relatively low temperatures, even at room temperature ! Moreover, the reaction rate was strongly dependent on the surface area of the supports.For a high-surface-area sample, such as Zr(AgPO,), (4.5:48) with a surface area of ca. 30 m2 g-l, complete reduction was achieved in 2 h at 21 "C with B hydrogen pressure of 440 Torr. However, for a low-surface-area sample, Zr(AgPO,), (1 2 : 336) (surface area ca. 2.2 m2 g-l), reduction of 30% of the Ag+ ions was achieved after 7 h at 43 "C andpH2 = 580 Torr. The kinetics of the reoxidation of the silver particles will be reported in a subsequent paper. However, our preliminary studies show high-surface-area samples reoxidize much faster than the low-surface-area samples. The metallic silver supported on ZrP is converted to Ag+ ions which diffuse back into the lattice of the support even at room temperature.This reoxidation phenomenon probably accounts for the difference observed between the X-ray powder diffraction patterns of the completely reduced samples of high- and low-surface area a-ZrP. Although the patterns were taken immediately after complete reduction, air oxidation of silver must occur and perhaps is accelerated by the heat produced from the X-ray beam. As a result the high- surface-area samples, where reoxidation of silver is rapid, give patterns of Ag+ partially exchanged phases instead of the c-ZrP phase. SILVER-PARTICLE SIZE AND DISTRIBUTION Silver-particle size can be changed by varying the amount of Ag+ loading or the surface area of the supports. The dispersion of the silver particles was not uniform; neither was the particle-size distribution.This can cause considerable deviation when the average particle size is determined by X-ray diffraction line-broadening measurements. Thus the data should be used only as a basis of comparison and as information required to understand the reaction mechanism, rather than as absolute values. Several factors can explain the discontinuity observed in the linear relationship between particle size and Ag+ loading when the loading is very small [fig. 1 (A)] and also the discontinuity between particle size and surface area of the support [fig. 1 (B)]. The electron micrographs show that the particle-size distribution is relatively more uniform when the Ag+ loading is low (plate 1) or the surface area of the support is higher.Different size distribution curves can lead to different average sizes. Further- more, any reoxidation which occurs on the high-surface-area samples tends to reduce the silver-particle size REACTION MECHANISM Gas-solid reactions involve a series of complicated steps which can be summarized as follows: (1) gas diffusion to a surface, (2) adsorption of gas molecules on a surface which can be a solid reactant or solid product, (3) movement of adsorbates on the surface of the solid phase, (4) adsorption on active sites, (5) reaction itself and (6) diffusion of products from the active sites. The unraveling of these complex reaction mechanisms presupposes an understanding of the various physical stages reached during the course of the total reaction.It is also clear that the relative contribution of the different steps to the overall reaction may change during the course of the reaction. It is generally accepted that hydrogen gas diffuses into the internal porous cavities of zeolites and that the reduction process proceeds both on the surface and in the internal pores.30331 However, the degree of hydrogen-gas diffusion into theS. CHENG AND A. CLEARFIELD 1591 interior of zirconium phosphate crystallites is considered to be negligible. If hydrogen gas did diffuse into the interior so that reduction proceeded inside the cavities, then nucleation of silver atoms in the cavities should result in the formation of silver clusters uniformly distributed in the interlamellar regions.As a result, the interlayer distance should become larger. However, this is not what is actually observed, i.e. the observed c-ZrP phase, Zr(HPO,),, has an interlayer spacing of only 7.41 A,9 and the silver collects on the surface of the crystals and along cracks and fissures. Therefore hydrogen diffusion into the interior of ZrP crystallites is thought to be minimal. The reason may be that the windows of the cavities of a-ZrP are too small for hydrogen-gas diffusion. The maximum free diameter of the windows in the a-ZrP crystallites is 2.64 A. The interlayer distance for Zr(AgPO,), is 7.76 A compared with 7.56 A for a-ZrP. Thus the windows in the former phase are only slightly larger. However, there are two Ag+ (radius 1.26 A)32 in each cavity of dimensions 5.3 x 7 A.Thus the interlamellar region is quite crowded and hydrogen-gas diffusion will be greatly restricted. The most likely mechanism, therefore, is that hydrogen is adsorbed on the surface of Zr(AgPO,), crystallites and the reduction reaction proceeds on the surface.8 The proposed reaction sequence is as follows: (1) H, molecules are adsorbed on the surface and diffuse to the active sites; (2) the reduction reaction proceeds on the surface: Ag atoms and protons are formed; (3) Ag atoms leave the active sites and form clusters; (4) protons diffuse into the interior and Ag+ ions diffuse out, enabling the reaction to proceed continuously on the surface. It is proposed that the initial induction period in this reaction is mainly due to an abnormally slow initial rate of silver-nuclei growth.Several investigationslap 3 3 9 34 have examined nuclei formation in solutions and on crystals and indicated that the formation of nuclei of a new phase is kinetically difficult. Nevertheless, after the induction period the number of nuclei increases at a rate proportional to a power of the time which is usually larger than 2. In other words, if the rate-determining step in our system is the nucleation of silver atoms, then the observed reaction rate must increase after the induction period and its dependence on time must be expressed by a curve having relatively high curvature. We can imagine that somewhere near the end of the induction period processes other than nucleation become rate-deter- mining; a concomitant rate deceleration occurs.Because the initial rates derived from Elovich plots are nearly proportional to the square root of the hydrogen pressure, the reaction mechanism must involve the dissociative adsorption of hydrogen molecules. The active sites thus have to be species which are able to bond with hydrogen atoms. Because it is well accepted that silver metal cannot chemisorb hydrogen,35 the autocatalytic reaction involving the spillover of H, on silver is excluded. The only other possible active species are the ion-exchange sites on the surface where the Ag+ ions are situated. The mechanism thus proposed is as follows:1592 METAL DISPERSIONS ON ZIRCONIUM PHOSPHATES k3 ,”:\\ - 0----Ag The protons formed in reaction (14) must diffuse into the lattice while unreacted Ag+ ions diffuse out to the surface in order to continue the reduction processes.The diffusion of H+ and Ag+ ions in zirconium phosphate is expected to occur more rapidly than the reduction reaction on the surface. Jerus and Clear~ield~~ have studied the kinetics of the reaction between gaseous HC1 and Zr(NaPO,), : Zr(NaPO,), + HCl, --+ Zr(HPO,), + NaCl. (15) This reaction should occur on the surface since all of the NaCl is present on the surface and there is no mechanism for chloride-anion diffusion between the layers of ZrP. Thus the HCl at the surface must ionize in order to form protons which then diffuse into the interior. Sodium ions are displaced and diffuse out to the surface; this counter-diffusion of ions was found to be rate-determining.36 A comparable reaction rate should be found in the present study if the counter-diffusion of cations is the rate-determining step.However, the half-life of reaction (1 5 ) in the same temperature and pressure range is one half of that observed in the present study. In addition, the activation energy of the migration of Na+ ions in Zr(NaPO,), is 10.3 kcal m ~ l - ’ , ~ ~ which is significantly smaller than that of the Ag+ reduction reaction (1 5.2 kcal mol-l). Therefore the rate-determining step of the reduction reaction should be some process other than the migration of the cations. Reaction (14) is actually the nucleation of silver atoms, which is slow during the induction period but very fast thereafter. Hence the rate-determining step, after the induction period, is most likely to be the process depicted by reaction (1 3). The rate equation is thus expressed as (see Appendix) rate = k, [k, k,/(k-, k-,)]ip~2[(0-Ag+),] (16) where (0-Ag+), represents the Ag+ ions at the surface ion-exchange sites.Since the surface concentration of Ag+ ions is proportional to the loading of Ag+ ions and to the surface area of the support where c is a constant. Upon substituting into eqn (16) the following equation results: with rate = k’p&,[Ag+], k’ = ck3[k, k,/(k-, k-,)];. This means that the initial reduction rate is directly proportional to the concentration of Ag+ loading and the surface area of the support, but only to the square root of the initial hydrogen pressure. Eqn (18) is similar to the empirical rate equation, eqn (10). The only discrepancy between eqn (10) and (18) is in the order of the surface-area term; 1.4 in eqn (10) and 1 .O in eqn (1 8).One reason that an order greater than oneS. CHENG AND A. CLEARFIELD 1593 is observed may result from the different activities on the sides and on the bases of the ZrP platelets. Electron micrographs show that large silver particles accumulate on the side faces, while fine silver particles cover the surface of the bases. This indicates that the side surface is more active than the basal surface towards the reduction processes. The probable reason is that the windows of the six-sided cavities of ZrP crystallites have a large free distance on the sides perpendicular to the layers; the diffusion of the cations, therefore, is easier in the direction parallel to the layers.As a result, the reduction reaction proceeds more readily at the edges of the layers which compose the side surface of the platelets and is therefore not a linear function of surface area. Another reason for the difference is that the concentration of surface-active sites varies as a function of the surface area. For crystalline a-ZrP, the surface area is increased mainly by decreasing the crystal size. However, the length of the edge on the base and the thickness of the platelets do not vary proportionally as the crystal size changes. Electron micrographs show that the ratio of the surface area of the sides to that of the bases increases as the size of the crystals decreases. Consequently, the dependence of the reduction rate on the surface area of the support should have a power greater than one.Two possible reasons can explain the decrease in the reduction rate as the reaction proceeds. One is a decrease in the concentration of the surface cations, which are the active sites for dissociative adsorption of hydrogen. As the reaction proceeds, the concentration of Ag+ ions in the interior decreases and the proton concentration increases. Since the rate of the diffusion process is proportional to the concentration gradient, the rate of counter-diffusion of Ag+ ions and the protons in the ZrP lattice should decrease as the reduction reaction proceeds. Actually, Jerus and Clea~Iield~~ have shown that the activation energy of cation migration increases as the loading of cations decreases.When the reduction reaction is nearly complete, the cation diffusion is probably so slow that it determines the rate. That explains why the Elovich equation can never be fitted to the decelerating rate in the kinetic curves. The other reason may be the increased diffusional resistance to H, of the silver particles formed on the surface as the particle size increases. As the Ag particles grow and cover the surface, hydrogen gas must travel through the pores or spaces between silver particles to get to the surface of the ZrP crystallites and then diffuse to the active sites buried under the silver particles. That a thick layer of silver particles covers the surface has been shown by the electron micrographs of the reduced samples. We acknowledge financial support of this study by the National Science Foundation under grants CHE79-16160 and CHE81-14613.We thank the Texas A&M EM Center for help in obtaining the electron micrographs. APPENDIX If reaction (1 3) is the rate-determining step, the reaction rate can be expressed as rate = k, [I2]. (20) Application of the steady-state treatment to the intermediates I, and I, gives eqn (21) and (22), respectively : (21) (22) d[I,lldt = k,[I11 [(O-Ag+),l- (k-,[IzI+ k, [I211 = 0 4Illldt = kl([O-Ag+),lP,* - k-,[I,I- k,[I11 [(O-Ag+),l+ k-,[I2I2 = 0. An assumption is made to simplify eqn (21); i.e. k-,[I,] %- k,.1594 METAL DISPERSIONS ON ZIRCONIUM PHOSPHATES Because reaction (13) is the rate-determining step, it seems to be reasonable to assume that its rate constant, k,, is negligible compared with the value k-&].Therefore eqn (21) simplifies (23) to or (24) Eqn (23) can be written in the form k-2[I,I2 = k,[I,1"0-Ag+),l. (25) Upon substituting expression (25) into eqn (22), the following equation results : k111(O-Ag+)slPH2 - kl [Ill = 0. (26) Thus [Ill = (kl/k-l) "O-Ag+)slP,, (27) (28) and [I21 = (kl k,lk-l~-2)~PL2[(o-Ag+~,l. The rate expression is therefore rate = k,(k, k,/k-, k-,):p~,[(O-Ag+),]. (16) K2 [Ill [(O-Ag+)sl- k-2[I,I2 = 0 [I21 = (kZ/k-ZY [41: W-Ag+),l~. J. R. Anderson, Structure of Metallic Catalysts (Academic Press, New York, 1975), p. 163. Kh. M. Minachev and Ya. I. Isakov, ACS Monogr., 1976, 171, 552. P. A. Jacobs, Carboniogenic Activity of Zeolite (Elsevier, Amsterdam, 1977). R. G. Herman, J. H. Lunsford, H. Beyer, P.A. Jacobs and J. B. Uytterhoeven, J. Phys. Chem., 1975, 79, 2388. M. Iwamoto, T. Hashimoto, T. Hamano and S . Kagawa, Bull. Chem. Soc. Jpn, 1981, 54, 1332. H. Beyer, P. A. Jacobs and J. B. Uytterhoeven, J. Chem. Soc., Faraday Trans. I , 1976, 72, 674. A. Clearfield, in Inorganic Zon Exchange Materials, ed. A. Clearfield (C.R.C. Press, Boca Raton, Fla, 1982), chap. 1. A. Clearfield, D. S. Thakur and H. Cheung, J. Phys. Chem., 1982,86, 500. A. Clearfield and S . P. Pack, J. Znorg. Nucl. Chem., 1975, 37, 1283. lo A. Clearfield and J. A. Stynes, J. Znorg. Nucl. Chem., 1964, 26, 117. l1 A. Clearfield and S. Cheng, J. Znorg. Nucl. Chem., 1980, 42, 1341. l2 S. Cheng, Ph.D. Thesis (Department of Chemistry, Texas A&M University, 1982), p. 36. l3 A. Clearfield and G. D. Smith, Znorg. Chem., 1969, 8, 431. l4 T. Inui, T. Ueda, M. Sushin and H. Shingu, J. Chem. Soc., Faraday Trans. I , 1978, 74, 2490. l5 A. Y. Rozovskii, V. D. Stytsenko and V. F. Tret'yakov, Kinet. Catal., 1973, 14, 953. l6 H. H. Voge and L. T. Atkins, J. Catal., 1962, 1, 171. l 7 N. F. H. Bright and W. E. Garner, Nature (London), 1934, 133, 570. l8 W. E. Gamer, W. R. Southon, J. Chem. Soc., 1935, 2, 1705. 2o W. E. Gamer, A. S. Gomm and H. R. Hailes, J. Chem. SOC. 1933, 2, 1393. 2* (a) Ya. Zeidovich, Acta Physicochim. USSR, 1934, 1, 449; (b) S. Roginskii and Ya. Zeldovich, Acta 22 S. Yu Elovich and G. M. Zhabrova, Zh. Fiz. Khim., 1939, 13, 1761. 23 H. A. Taylor and N. Thon, J. Am. Chem. Soc., 1952,74,4169. 24 J. M. Dunoyer, C . R. Acad. Sci., 1942, 214, 556. 25 B. D. Averbukh and G. I. Chaparov, Zh. Fiz. Khim., 1949, 23, 37. 26 J. A. Allen, Austr. J. Chem., 1963, 16, 193. 27 A. van Tiggelen, L. Vanreusel and P. Neven, Bull. SOC. Chim. Belg., 1952, 61, 651. 28 K. Tsutsumi and H. Takahashi, Bull. Chem. SOC. Jpn, 1972,45, 2332. 29 G. R. Eulenberger, J. G. Keil and D. P. Shoemaker, J. Phys. Chem., 1967, 71, 1812. 30 Kh. M. Minachev, G. V. Antoshin, E. S. Shipro and T. A. Navruzov, Izu. Akad. Nauk SSSR, Ser. Khim., 1944, 9, 2081. 31 M. Kermarec, M. F. Guilleux, D. Delafosse, M. Briend-Fauer and J. F. Tempere, Proc. 8th. Int. Symp. Reactions in Solids, ed. J. Wood, 0. Lindquist and C. Helgesson (Plenum Press, New York, 1977), p. 689. 32 B. E. Douglas and D. H. McDaniel, Concepts and Methods of Inorganic Chemistry, (Blaisdell Publishing Co. New York, 1965). 33 A. Ya. Rozovski, V. D. Stytsenks and V. F. Tret'yakov, Kinet. Katal., 1978, 18, 991. 34 J. Haber, A. Kostowska and J. Stoczynski, Proc. 8th Znt. Symp. Reactions in Solids, ed. J. Wood, 35 G. C. Bond, Heterogeneous Catalysis (Oxford University Press, Oxford, 1974). 36 P. Jerus and A. Clearfield, J. Znorg. Nucl. Chem., 1981, 43, 21 17. Groen, Nature (London), 1954, 174, 1836. Physicochim. USSR, 1934, 1, 554. 0. Lindquist and C. Helgesson (Plenum Press, New York, 1977), p. 331. (PAPER 3/ 1665)