J. Chem. SOC., Furaduy Trans. I , 1989, 85(4), 855-867 Effect of Form on the Surface Reactivity of Differently Prepared Zinc Oxides Vera Bolis, Bice Fubini" and Elio Giamello Dipartimento di Chimica Inorganica, Chimica Fisica e Chimica dei Materiali, Universita' di Torino, Via P. Giuria 7, 10125 Torino, Italy Armin Reller Anorganisch-chemisches Institut, Uniuersitat Ziirich, Winterthurerstrasse 190, 8057 Zurich, Switzerland The relationships between the form of crystallites and their adsorption properties have been investigated on ZnO samples of various origins by means of electron microscopy, X-ray diffraction and adsorption calorimetry. Three polycrystalline ZnO samples have been studied: one obtained by ignition of zinc metal (Kadox) and the others obtained by decomposition of zinc carbonate and oxalate.The higher surface reactivity found on Kadox ZnO in comparison with that found on ex-salt ZnO is not due to the preferential development of crystal planes at the surface but to the presence of better defined single microcrystals with sharp edges. In all the cases examined the preparation route leads to a difference in morphology and consequently a difference in reactivity. Carbon monoxide and hydrogen have been used as surface probes for active sites : CO is coordinated onto cations exposed at the edges between the (0001) and (1010) planes; H, is adsorbed in different forms, one of which, type I, occurs on sites located at the same edge as CO. Prereduction of ZnO reduces the adsorption activity towards both gases; the extent of this reduction also depends upon the actual morphology.The reactivity of a solid in general, including surface reactivity in catalysis, depends decisively on its form, which comprises compositional, structural and morphological features. This is particularly relevant in the surface chemistry of zinc oxide. On the one hand, studies of single crystals have shown a dependence of its adsorption properties upon the exposed surface on the other hand, differences in adsorption properties have been found on polycrystalline materials of various origins.*-' In the latter case differences in adsorption capacity and interaction energy have been ascribed by some authors, including ourselves, t o different distributions of the exposed crystal faces on the samples, resulting from the preparation route Since zinc oxide is an active component in mixed catalysts, such as Cu/ZnO and Cr,O,ZnO,lO, l1 for various reactions, e.g.the synthesis of methanol and higher alcohols, the influence of the preparation route on the final state of the ZnO surface is of essential importance in catalyst preparation. Zinc oxide crystallizes in a wurtzite-type structure and is usually made up of hexagonal prisms, where the hexagonal planes, (0001) and (OOOT), are located perpendicular to the c axis and the prismatic planes, (1010) and (1 120), are parallel to it. These low-index planes are the most commonly exposed faces in polycrystalline material^.^ On the polar faces, (0001) and (OOOT), Zn2+ cations and 02- anions, respectively, are the more outwardly exposed.On the non-polar prismatic planes, (1010) and (1 120), both zinc and oxygen ions are located in the same plane. The polar faces, where surface reconstruction produces particular anion/cation vacancies, have been 855856 Surface Reactivity of Zinc Oxides regarded by several authors as the active sites in catalytic reaction^.^. ' 9 8~ l1 In fact, at room temperature both CO and H, are selectively adsorbed on this type of face.2*12-15 A clear-cut relationship between the extension of polar faces, the adsorption capacity towards CO and H, and catalytic activity has thus been but never proved, on the basis of parallel investigations on morphology and surface properties. Lavalley et al.,' however, recently reported that kinks and edges (at the intersection of polar and non-polar planes) should be taken into account as possible active sites instead of the polar planes.Moreover, the actual shape of the microcrystalline particle (i.e. the microstructure of the crystal faces present and thus of the edges formed) can be an essential cause of the existence of specific sites. Therefore, in addition to the crystal morphology (i.e. the type and relative abundance of the various exposed faces), the abundance of ' morphological ' defects such as surface steps, edges, kinks and corners, which cause the presence of ions that are in particularly low coordination at the surface and not available on extended faces, must be taken into account. In fact, the number of these defects (negligible in the case of single crystals) can be very high in finely divided polycrystalline materials and thus can influence the reactivity of the solid by introducing particularly active sites.Conditions used in the preparation route could play an important role in this respect. The aim of this paper is to compare directly adsorption features (the number of active sites and the binding energy) and morphological aspects [as elucidated using X-ray diffraction (X.r.d.), scanning electron microscopy (SEM), transmission electron microscopy (TEM) and high-resolution electron microscopy (HREM)] of ZnO prepared by different procedures : the widely studied Kadox ZnO, obtained by the ignition of Zn metal, and samples obtained by the decomposition of salts (carbonate and oxalate), a procedure which is close to that adopted in the preparation of mixed catalysts. Carbon monoxide and hydrogen have been chosen as test molecules in adsorption work for three main reasons.(i) The adsorption of these gases, at least on Kadox ZnO, has been studied widely by infrared spectroscopy,', ' 9 1 3 7 1 5 9 l6 adsorption microcalor- imetry4. 6 * and temperature-programmed de~orpti0n.l~. (ii) Both of these gases are selectively adsorbed at room temperature on only one (CO) and two (H,) kinds of sites, and assignments of the adsorbed form to a well defined surface arrangement have been proposed. Carbon monoxide is coordinated onto the coordinatively unsaturated Zn2+ cations (c.u.~),,' l2 whereas hydrogen is dissociatively adsorbed onto the so-called type I sitesl3'l7 yielding surface hydroxyls and hydrides.These sites are proposed to be located at the (0001) face14 or at the edge between the (0001) and (1010) faces." Hydrogen is also adsorbed onto the so-called type I1 sites, for which a bridged form requiring suitably oriented pairs of Zn2+ and 0,- ions has been pr~posed.~ In addition, previous calorimetric investigations by some of us indicated the possibility that some adsorbed hydrogen could slowly diffuse into the bulk.4 (iii) Hydrogen and carbon monoxide are the starting materials for syngas reactions. In addition to the morphological aspects, the semiconducting nature of ZnO (n-type) can also play a crucial role in its reactivity. Zinc donor centres in excess, and consequently the presence of a relevant electronic population in the conduction band, will obviously influence any interaction involving electron exchanges between the solid and the adsorbed m01ecule.'~ The effect of reducing pretreatments on adsorption has thus been investigated.A drastic change in reactivity towards CO upon reduction was found by some of us in a previous study of one ZnO sample.6 The study of the influence of reduction upon adsorption of both CO and H, has been extended to all the samples examined in order to verify whether reduction affects samples obtained by different preparation routes in the same way.V. Bolis, B. Fubini, E. Giamello and A . Reller 857 Experimental Three different ZnO samples have been investigated: zinc oxide obtained from the ignition of the metal in an oxidising atmosphere (Kadox 25, New Jersey Zinc Co.), zinc oxide obtained by decomposition of zinc carbonate and zinc oxide obtained by the slow decomposition of zinc oxalate.The B.E.T. surface area of the samples (measured at 77 K by means of a Carlo Erba Sorptomatic apparatus, N, iunit) were 10, 24 and 18 m2 g-l, respectively. The heat of adsorption of CO and H, and related quantities were measured at room temperature by means of a Tian Calvet microcalorimeter connected to a volumetric apparatus as described previously.20 The samples were pretreated following three different procedures. (i) The sample was slowly heated at 673 K under vacuum and then 75 Torr of 0, (1 Torr z 0.133 kPa) was admitted and removed three times at the same temperature. The sample was then cooled and transferred into the calorimeter under oxygen.(ii) The sample was slowly heated up to 673 K in vacuo and then transferred, in vacuo, into the calorimeter. (iii) The sample was slowly heated to 673 K, then cooled to 473 K. At this temperature 75 Torr of H, was admitted and removed three times. The sample was then evacuated at the same temperature (473 K) in order to ensure the elimination of any adsorbed hydrogen irreversibly held at the surface at lower temperatures. The origin of the ZnO sample will be designated as k-, c- or 0-, indicating, respectively, Kadox, ex-carbonate or ex-oxalate. The type of pretreatment undergone by the sample will be indicated by the terms -ox (oxidised), -un (untreated) or -red (reduced), following the ZnO forrnula, e.g. k-ZnO-red indicates Kadox zinc oxide submitted to the reducing pretreatment.Adsorption was performed by admission of successive small doses of the adsorptive (CO or H,) up to an equilibrium pressure of 75 Torr in the case of CO and 25-30 Torr in the case of H, (first run). The calorimetric cell was completely evacuated after the first adsorption run in order to evaluate the reversibility of the process. Furthermore, in some cases a second adsorption run has been performed. Owing to the small size of the calorimetric vessel (containing at maximum 3-4 g of powder) and the relatively low surface area of the samples examined, it was difficult to perform accurate volumetric measurements, particularly in the case of the ex-oxalate zinc oxide, which has a poor adsorption capacity. For this reason data relative to the adsorption of H, onto the reduced ex-oxalate zinc oxide are not available. The three different zinc oxides were characterized with respect to their structural, compositional and morphological features.X-Ray diffractometry was carried out using a Guinier-IV camera as well as a Siemens Kristalloflex diffractometer, both with Cu Ka radiation. In order to confirm the purity of the samples, i.e. the absence of zinc carbonate and, respectively, zinc oxalate, thermogravimetric measurements were performed using a Perkin-Elmer TGS-2 thermomicrobalance. For a detailed analysis of both the morphology and microstructure, including the identification of the cry- stallographic faces constituting the crystallite surfaces, scanning electron microscopy (SEM) using a Cambridge Instruments Stereoscan mk 111 instrument as well as high- resolution electron microscopy (HREM) and selected-area electron diffraction (SAED) using a Jeol 200CX instrument equipped with top-entry stage were applied.For the HREM studies the samples were dispersed in hexane and disposed on copper grids coated with a ‘holey’ carbon film. In order to obtain diffraction patterns of isolated ZnO crystallites, optical diffraction studies were performed by laser-beam diffraction of high- resolution electron micrographs (fringe patterns) on an optical bench with standardized geometry.858 Surface Reactivity of Zinc Oxides Results Surface Morphology The initial thermogravimetric measurements of all three samples gave no evidence for a weight change of > 0.3% within the temperature range 300-900 K.Therefore the presence of traces of zinc carbonate or zinc oxalate can be excluded. In X-ray powder diffractograms, as well as in the patterns obtained from the Guinier IV camera, one observes slight differences between the samples. Whereas for c-ZnO and o-ZnO a slight broadening of the reflections is observed, sharp reflections are registered for k-ZnO. These results can be explained by the presence of smaller crystallites making up c-ZnO and o-ZnO, compared with k-ZnO, or by the presence of poorly crystalline materials. In addition, no indication can be derived from X-ray studies for a possible prevalence of a particular type of morphology. Thus no unambiguous identification of crystallographic faces making up the surfaces of the crystallites could be obtained.A more informative answer, however, could only be achieved from electron-microscopy studies. As is shown in the scanning electron micrographs [plate 1 (a)-(c)] the micromorphology of the three materials can be characterized as follows : for c-ZnO, agglomerates of crystallites with dimensions 4 1 pm are observed. In the case of o-ZnO rod-like agglomerates made up of densely packed crystallites are observed. The characteristic shape of these agglomerates can be described as pseudomorphs of the precursor material, i.e. the zinc oxalate crystals. For k-ZnO the crystallites appear to be randomly agglomerated. As a consequence this material can be easily dispersed. In summary, the scanning electron micrographs give evidence for the presence of very small ZnO crystallites constituting the three samples.Owing to the limited resolution of SEM, however, the exact shape of these crystallites cannot be defined. Therefore, the high-resolution micrographs presented in plate 2(a)-(c) must be examined. First one should note that high-resolution electron microscopy cannot yield quantitative distributions of particle sizes or the exact numbers of crystallographic faces constituting the surfaces of the differently prepared samples. The analysis of the micrographs, however, yields information on the shape and dimensions, as well as on the crystallographic faces actually making up the surfaces of isolated particles. In plate 2 the characteristic features of the three ZnO samples are summarised.Plates 2(a) and (b) present an agglomerate of c-ZnO microcrystals and, as an inset, the high-resolution image of few isolated crystals. The dimension of these particles lies in the range 30-60 nm. The formation of partial large arrays of single- crystal material can be explained by the fact that, during the decomposition of the initial zinc carbonate, relatively high local partial pressures of CO, exist. As has been shown in the decomposition of various carbonates,2' such conditions allow sintering of the product phase, i.e. the zinc oxide. This result is supported by the fact that no sharp edges confining the crystallites are observed. The evaluation of the fringe patterns in the high- resolution micrographs obtained gives evidence that prismatic and hexagonal faces contribute to the surfaces.A predominance of prismatic faces can be derived from the inspection of larger arrays of this sample. The micrographs of o-ZnO [plates 2(c) and ( d ) ] support the abovementioned role of product formation governed by the course of the decomposition of the initial zinc oxalate: agglomerates made up of crystallites with equal dimensions in the range 30-60 nm exhibit the shape of the precursor crystals, i.e. they represent well defined pseudomorphs. Again the isolated crystallites of ZnO exhibit no well defined edges, but rather disc-like shapes. An analysis of the fringe patterns indicates that prismatic and hexagonal faces are present. The k-ZnO sample consists of perfectly formed single crystals with dimensions in the range 2G300 nm [plates 2(e)-(g)].This feature shows the most distinct difference between the samples c-ZnO and o-ZnO. The presence of such well defined single crystals with sharp edges must be a consequence of the basically different method of preparation,J . Chem. SOC., Faradaj) Trans. 1, Vol. 85, part 4 Plate 1 Plate 1. Scanning electron micrographs of microcrystalline ZnO, obtained from the decomposition of (a) zinc carbonate (c-ZnO) (b) zinc oxalate (0-ZnO) and by (c) the ignition of metallic zinc in an oxidising atmosphere (k-ZnO). V. Bolis et cil. (Fucing p . 8 5 8 )J . Chem. Soc., Faraday Trans. 1, Vol. 85, part 4 10 nm Plate 2(a)(b). For caption see overleaf. Plate 2 (a) (b) 100 nm V. Bolis et al.J . Chem. SOC., Furaduy Trans. 1 , Vol. 85, part 4 (C) 500 nrn - 30nm - Plate 2(c)(d).For caption see overleaf. Plate 2(c)(d) V. Bolis et al.J. Chem. SOC., Faraday Trans. 1, Vol. 85, part 4 (e 1 100 nm- 10 nm Plate 2. Electron micrographs of c-ZnO [(a) and (b)], o-ZnO [(c) and (d)] and k-ZnO [(eE(g)]. As insets, lattice images as well as respective optical diffraction patterns are presented for c-ZnO and k-ZnO in order to confirm the crystallographic faces making up the surfaces of crystallites. V. Bolis et al.V. Bolis, B. Fubini, E. Giamello and A . Reller 859 i.e. the combustion of finely dispersed metal. As one may see from the micrographs, the surfaces of these single-crystal particles are made up of prismatic and hexagonal faces. The pronounced predominance of hexagonal faces, which was regarded as a decisive feature in earlier 7 - 8 - l5 cannot be supported by these results.Adsorption of Carbon Monoxide Two runs of adsorption of CO were performed on the ZnO samples. During the first run a small fraction of CO ( < 10 YO) underwent oxidation to carbonate-like species and was irreversibly held at the surface after evacuation. The extent of this process was found to vary in the order o > c > k, and, obviously, ox > un > red. In the second run the adsorption of CO was entirely reversible with respect to simple 0-coordination of the molecule on the acidic cationic centres (Znf:,). Since the adsorption of CO was performed in order to evaluate the population of this type of site, it seems more convenient to compare the second runs obtained for the different samples.In fig. l(a>-(c) the volumetric isotherms of adsorption (adsorbed amounts, n,, us. equilibrium pressure) are reported for the three k-, c- and o-samples. For each sample three curves relative to the standard pretreatment (ox, un, red) are also shown. Inspection of the figures indicates that the number of surface sites interacting with CO depends on both the origin of the sample and the reducing treatment it has undergone. The amounts adsorbed per unit surface area decrease in the order k > c > o and, for each sample, in the order ox > un > red. In fig. 2 the integral heat (Qint) evolved during the adsorption is plotted as a function of adsorbed amount (n,) for all the samples. In all cases the integral heat depends linearly on the adsorbed amount, as all the experimental points fall on straight lines passing through the origin.Under such circumstances the adsorption enthalpy values, AadsH, are given directly by the slope of each line. In the case of the k- and o-samples all points fall on the same line independent of the redox treatment. Instead, three separate lines are found for the c-sample. However, the -AadsH value found for samples pretreated in an oxidising atmosphere is the same (46 kJ mol-') in all three cases. The adsorption data (adsorption enthalpy and adsorbed amounts) for all the samples examined are summarized in table 1. The adsorptive capacity of the k-ZnO-ox sample is two and four times higher than that of c-ZnO-ox and o-ZnO-ox, respectively. Thermal treatments in uacuo and in a reducing atmosphere cause a decrease in the number of active sites in all samples: on k- and o-ZnO the thermal treatment in vacuo eliminates 10 YO of the sites, the reducing one 40 Oh.c-ZnO is more sensitive to thermal treatments as it loses 60 and 70 % of the total sites, respectively. k-ZnO, even in the less-active form (red), exhibits a number of active sites that is higher than that of the other samples in whatever form. In the case of k- and o-ZnO, the energy of interaction is not affected by the reducing treatments (fig. 2), whereas in the case of c-ZnO (as previously reported by some of US)^ the energy of interaction decreases with progressive reduction. The value of - Aads H falls to 29 kJ mo1-' in the case of the untreated sample and to 8 kJ mol-' in the case of the reduced one [fig.2(b)]. Adsorption of Hydrogen Fig. 3 reports volumetric (n, vs. p ) (a) and calorimetric (Qint us. p ) (b) isotherms for the adsorption of H, onto k-, c- and o-ZnO in the oxidised form. The data refer to the first adsorption run. In all cases the process is pressure-dependent in the whole range examined. On k-ZnO, which also in this case shows the highest activity, at ca. 20 Torr the process860 1 . o 0.75 N 'E 4 \ g0.5 a? 0.25 Surface Reactivity of Zinc Oxides ~ I I I ( a ) 0 2'0 6'0 0 20 40 60 8 0.5 0.25 0 20 LO 60 plTorr Fig. 1. Volumetric isotherms of adsorption of CO on (a) k-ZnO (0, ox; @, un; 0, red), (6) c-ZnO (A, ox; A, un; A, red) and (c) o-ZnO (0, ox, 0, un; W, red). is not yet accomplished, whereas both c- and o-ZnO at the same pressure are nearly saturated.Adsorption values deduced from the volumetric isotherms [fig. 3 (a)] indicate that the amounts adsorbed per unit surface area decrease in the same order as that found for CO adsorption, i.e. k > c > 0. Table 2 reports the quantitative data of hydrogen adsorption ('p = 20 Torr) for the whole set of samples. As in some cases the heat of interaction decreases with coverage, the two limiting values of the variation range for the enthalpy are reported [columns 3(a)-(c)]. Two main facts can be observed by inspection of table 2. (i) Similar to the findings for carbon monoxide, the adsorption capacity of k-ZnO-ox is two and fourV . Bolis, B. Fubini, E. Giamello and A . Reller 86 1 3 N ‘E 5 c, - 2 \ E d 01 1 3 N ‘E h 1 0 0.2 0.4 0.6 Fig.2. Integral heat of adsorption us. adsorbed amounts of CO on ( a ) k-ZnO (0, ox; @, un; e, red), (6) c-ZnO (A, ox; A, un; A, red); and (c) o-ZnO (0, ox; 0, un; ., red). times higher than that of c- and o-ZnO, respectively (n,). Furthermore, the numbers of sites on the different samples decrease upon undergoing the reducing treatment. The relative decrease of active sites upon reduction (n,) is the same for the k-, c- and o-ZnO samples. (ii) The adsorption enthalpy decreases with coverage on k- and o-ZnO-ox samples (see also fig. 4), whereas in the case of c-ZnO-ox the differential heat is constant. Furthermore, in this latter case the heat of adsorption is unaffected by the reducing treatments [table 2, column 3(6)]. The same does not apply for k- and o-ZnO, for which the enthalpy range depends upon the treatment [table 2, columns 3 ( a ) and (h), respectively].More detailed information about the variation of the heat of adsorption862 Surface Reactivity of Zinc Oxides Table 1. Adsorption enthalpy and amount adsorbed measured a t p = 20 Torr for CO adsorbed on ZnO samples (a) k-ZnO (b) c-ZnO (c) o-ZnO _ _ _ ~ co/ co/ co/ molecule - AadsH/ molecule - AadsH/ molecule -‘adsH/ nm-, nlU/nzb kJ mol-1 nm-2 nla/nZb kJ mol-l nm-2 n,a/n2b kJ mol-1 ox 0.22 1/1 46 0.11 0.5/1 46 0.05 0.2/1 46 un 0.19 1/0.9 46 0.05 0.3/0.4 29 0.04 0.2/0.9 46 red 0.16 1/0.7 46 0.04 0.3/0.3 8 0.03 0.2/0.6 46 a n, illustrates the variation in adsorptive capacity from sample to sample in the same oxidation state (number of CO molecules referred to those adsorbed on k-ZnO taken as unity).b n z illustrates the variation in adsorptive capacity due to reduction for the same sample (number of CO molecules referred to those adsorbed on the ‘ox’ form taken as unity). I I I I 1.5 (61 6 * - N N ’& 1.0 ‘E 1 h 3 \ s 0.5 0 10 20 30 40 0 10 20 30 40 plTorr p/Torr Fig. 3. (a) Volumetric (n, us. p ) and (b) calorimetric (Qint us. p) isotherms of adsorption of H, on 0, k-ZnO-ox; 0, c-ZnO-ox; and 0, o-ZnO-ox. Table 2. Adsorption enthalpy and amount adsorbed measured at p = 20 Torr for H, adsorbed on ZnO samples (a) k-ZnO (b) c-ZnO (c) o-ZnO H2/ - AadsH/ molecule - ‘ads H,/ - Aads H / molecule H2/ molecule nm-, nla/n2b kJ mol-’ nm-, nla/nzb kJ mol-’ nm-, nlU/nzb kJ mol-l ox 0.64 1 / 1 60-14 0.33 0.5/1 ca.35 0.11 0.2/1 75-50 const. un 0.46 1/0.7 55-30 0.23 0.5/0.7 ca. 35 0.08 0.2/0.7 ca. 59 const. const. red 0.30 1/0.6 ca. 55 0.24 0.6/0.7 ca. 35 - - - const. const. ~~ ~ a n, illustrates the variation in adsorptive capacity from sample to sample in the same oxidation state (number of H, molecules referred to those adsorbed on k-ZnO taken as unity). n4 illustrates the variation in adsorptive capacity due to reduction for the same sample (number of H, molecules referred to those adsorbed on the ‘ox’ form taken as unity).V. Bolis, B. Fubini, E. Giamello and A . Reller 863 Fig. 4. Differential heat of adsorption (qdiff/kJ mol-l) of H, on 0, k-ZnO-ox; a, c-ZnO-ox; and 0, o-ZnO-ox us. amount adsorbed (n,/pmol m-,). with coverage is given in the differential-heat plot reported in fig.4 for the oxidised samples. Arrows in the figure indicate the adsorbed amount corresponding to 20 Torr equilibrium pressure. On k-ZnO the differential heat of adsorption decreases from an initial value of 60 kJ mol-1 to 14 kJ mol-'. A process associated with a very low heat of interaction occurs and becomes more important than the initial process at increasing coverage. Although in the case of o-ZnO the total adsorption is much lower than in the case of k-ZnO, the differential-heat curve lies above that of k-ZnO at low coverage (high-energy) sites. On c-ZnO the heat of adsorption is 35 kJ mol-l over the whole range examined. The thermokinetics of H, adsorption are typical of an activated process, in agreement with the fact that hydrogen is dissociatively chemisorbed on ZnO.'3 a, 1 3 7 l5 Under our experimental conditions the evolution of heat corresponding to the adsorption of a dose of hydrogen lasted typically between 1 and 2 h.Only in the case of k-ZnO-ox is the presence of much slower processes (lasting ca. 6-7 h) observed, as already r e p ~ r t e d . ~ Chemisorption of hydrogen at room temperature consists of a reversible and an irreversible part. In this aspect too the behaviour of the three oxidised samples is different. A large fraction of adsorbed H, (corresponding to ca. 50% of the total heat evolved) is not removed from k-ZnO-ox upon evacuation at room temperature. Irreversible adsorption is lower in the case of o-ZnO (20 % of the total heat evolved) and is practically absent on c-ZnO.Progressive prereduction of the surface somehow inhibits the irreversible phenomena, in parallel with the disappearance of the low-energy process : all the hydrogen on k-ZnO- red is reversibly adsorbed, and the heat of adsorption, 55 kJ mol-l, does not vary with coverage. Discussion The reactivity of zinc oxide surfaces towards CO and H, strongly depends upon form and reduced state of the material. The origin of the sample determines both the form of the powder particles (as evidenced by electron-microscopy results) and the Zn/O ratio, which can be more or less near to the stoichiometry according to the preparation route, as reported by Parravano et aZ.,, In order to evaluate the role of form in determining the surface activity of the sample, it is convenient to compare adsorption results obtained for the different samples864 Surface Reactivity of Zinc Oxides pretreated in oxygen (ox).Stoichiometry differences in pristine samples, caused by the different preparation routes, should be lessened by the standard oxidising thermal treatment. The effect of the reduced state of the sample upon its surface reactivity will be discussed separately. Influence of Form on Adsorption Properties Zinc oxide obtained by direct oxidation of the metal is a material that exhibits unique features if compared with zinc oxides obtained by the slow decomposition of compounds such as zinc carbonate or oxalate. The most relevant differences in form evidenced by X-ray diffraction and electron-microscopy measurements can be ascribed to the two distinct preparation routes.The ex-metal is the most crystalline sample, made up of well separated and defined particles, whereas the ex-salt powders are mainly constituted of densely packed aggregates built up of compact small particles (the specific surface areas of these samples are higher than that of Kadox). The well defined crystallites of Kadox are caused by the growth of ZnO in an oxidising atmosphere at a high temperature. Thus the kinetics of formation of the oxide are very fast in comparison with the kinetics of decomposition of a salt. In the latter case the oxide starts to grow very slowly on a salt particle that is still intact at T < 800 K. Therefore materials with less-defined particle shapes but relatively high surface areas are formed. Similar results have been found in the case of a-Al,O,.23 The main result from a comparison of data on the three samples is that the differences in form of the particles arising from the different preparation routes lead to remarkable differences in adsorption capacity. This feature appears to be important in determining the surface reactivity of k-ZnO : in spite of the fact that the ex-metal zinc oxide bears the lowest surface area among the three samples it exhibits the highest specific surface activity, as shown by the adsorption measurements. The enthalpy values reported for CO adsorption are consistent with a simple coordination process,12 and the constancy of the adsorption enthalpy (46 kJ mol-I) with coverage indicates that CO adsorption involves the same type of equivalent and non- interacting sites on all the samples.In fact, the TEM and HREM observations reported indicate that, contrary to what is generally accepted, k-ZnO does not exhibit a preferential development of polar faces with respect to c- and o-ZnO but shows a better definition of the border of the crystallites, i.e. the edges between polar and apolar planes. The differences in adsorptive capacity between k-ZnO and the two ex-salt samples can thus be explained not in terms of a different distribution of exposed faces but by taking into account the high crystallinity of the ex-metal oxide and thus the presence of well defined intersections of crystallographic planes, i.e. well defined crystal edges. Preferential sites for the reversible adsorption of CO are likely to be highly unsaturated Zn2+ cations exposed at the intersection of polar and apolar planes.This is in agreement with the assignment proposed by Lavalley et al.' In the case of c- and o-ZnO, the low adsorptive capacities observed suggest that, in spite of the poor definition of the microcrystals, which should imply a high abundance of structural defects, the cationic centres are less coordinatively unsaturated than those in k-ZnO. The correspondence between the number of active sites observed on different samples for CO and H,, respectively (the n, values reported in the first rows of tables 1 and 2), suggests that the edge sites also play an important role in hydrogen adsorption, in agreement with results reported in ref. (8).However, interaction of H, with ZnO is more complicated than the simple coordination observed in the case of CO, as indicated by the variation of adsorption enthalpy and thermokinetics with coverage observed in some cases.V. Bolis. B. Fubini, E. Giamello and A . Reller 865 The different phenomena occurring when H, is contacted with ZnO are as follows. (i) A reversible but slightly activated adsorption related to the dissociation of hydrogen molecules with corresponding formation of Zn---H and 0---H species, as previously reported.4 This kind of adsorption, usually indicated as type 1,13 is the only process observed on c-ZnO. It also takes place on k-ZnO, although it is overlapped by other phenomena:4 it is practically absent on o-ZnO. As mentioned above and suggested by the present results, the sites responsible for type I adsorption are the Zn2+ and 0,- ion pairs that are probably present at the intersection of polar and apolar planes.8 (ii) The highest energetic adsorption, slow and irreversible at room temperature, observed at low coverage on k- and o-ZnO, (fig.4) is ascribed to the dissociative homolytic adsorption onto the so-called type I1 sites proposed by Ghiotti et aL7 on the basis of infrared data. This adsorption leads to the formation of Zn---H--- Zn and 0---H---0 bridged species located at the (1010) prismatic faces, where a suitable orientation of Zn” and 02- ion pairs is available.’ (iii) The slow, irreversible and low-energy process (not detectable by infrared spectroscopy) observed in the case of k-ZnO (fig.4, high coverage) has been assigned previously by some of us to the diffusion of molecular hydrogen into the open structure of Zn0.4 This fact could also be related to the presence of structural defects such as steps, kinks and edges that would favour the entrance of the small hydrogen molecules into the structural channels. However, the diffusion of hydrogen is more likely to depend upon the surface concentration’ of H, and not upon the equilibrium pressure. Two neighbouring adsorbed hydrogen atoms must recombine in order to penetrate as a molecule into the bulk. The coverage obtained at the same equilibrium pressure is much greater on k-ZnO than on the other samples. This justifies the occurrence of diffusion only on k-ZnO and the progressive disappearance of the diffusion process itself observed on k-ZnO-un and k-ZnO-red, where the total amounts adsorbed are lower than in the case of k-ZnO-ox (table 2).Finally, note that the two products obtained by the decomposition of the parent zinc compounds exhibit different reactivities. The different adsorption capacities shown by the two ex-salt zinc oxides suggest that several factors have to be taken into account in order to prepare an active ZnO catalyst via precursor salt decomposition, i.e. the reaction conditions, such as temperature and atmosphere, the morphology of the precursor and the nature of the salt. In the case of the decomposition of zinc oxalate carbon monoxide is produced, and the zinc oxide is formed in a more reducing atmosphere than that obtained by decomposition of the carbonate.Influence of the Reducing Treatments on the Activities of the Samples The data reported in tables 1 and 2 clearly indicate that drastic changes in adsorption features occur when the solids are submitted to partially reducing treatments. The variation observed upon ZnO reduction depends, in principle, on both the form of the individual crystal particles and the electronic population of the bands : a clear distinction between these two effects is beyond the aim of the present paper and would require a detailed analysis of the conductivity of each sample. The discussion is limited, therefore, to the most relevant quantitative effects of the reduction on the chemisorption activity of ZnO, which can be summarized as follows: (i) a decrease in the total number of active sites and (ii) a selective inhibition of particular mechanisms of adsorption.The coordination of carbon monoxide involves the donation of electrons to the coordinatively unsaturated surface cations : a surface with an increased electron density is obviously less active towards this kind of adsorption. The dissociative adsorption of hydrogen requires sites involving pairs of zinc and oxygen species : the progressive elimination of oxygen by reducing treatments thus lowers the potential number of active sites. In the case of c-ZnO the decrease in the number of sites coordinating CO also866 Surface Reactivity of Zinc Oxides involves a progressive decrease in the interaction energy (- A,,,H decreases from 46 to 29 and 8 kJ mol-1 for the ‘ un ’ and ‘ red ’ samples, respectively).Since the adsorption enthalpy is a measure of the acid-base interaction, not only the number of acid sites but also their acid strength is affected by the reducing treatments. A similar decrease in the acid strength of the surface sites has not been observed in the k- and o-ZnO samples. The reducing treatments in these cases lower the number of acid sites, although to a lesser extent than in c-ZnO, but leave their acid strengths unchanged. As far as H, adsorption is concerned, the reduction involves not only a decrease in the total amount adsorbed, but also remarkable modifications to the adsorption features. Calorimetric data indicate a different distribution of the various processes. In particular, fewer sites for type I adsorption are available on reduced k- and c-ZnO samples: this decrease parallels that of CO adsorption on these two samples, in agreement with the assignment of the same sites of adsorption for CO and H2(I).Furthermore (as mentioned above), the slow diffusion of H, into the bulk observed on k-ZnO-ox is inhibited upon reduction. Conclusions The insights obtained from the experimental studies presented permit a discussion emphasizing the interdependence between the individual form of a solid and its effect on surface chemistry on processes involving heterogeneous catalysis. The precursor material and the preparation route leading to solid products with characteristic morphological features (and therefore with localised and defined numbers of active sites) are among the important factors influencing specific catalytic activity.In the case of ZnO the predominance of hexagonal faces was regarded as a decisive feature explaining the comparatively high activity of Kadox [see e.g. ref. (7) and (S)]. The experimental findings presented here do not support this assumption, which therefore must be discarded or at least revised. The decisive feature is instead the high perfection of the k-ZnO crystallites, and therefore the presence of well defined edges between the hexagonal and prismatic faces. In comparison with the ex-salt samples, these edges establish a relatively high density of coordination sites for CO. As the energy of interaction of CO is the same in all cases, the variations of the total amounts adsorbed observed for samples of different origins are ascribed to the presence of a different number of equal sites, essentially depending on ‘ form ’ factors.Furthermore, in contrast to the results indicated by Natta in the earliest work on ZnOlO and to what is generally accepted,18 a high surface reactivity of a zinc oxide sample is not necessarily related to a high surface area. An abundance of specific sites, which strongly depends upon the preparation route, is the most important feature for an active zinc oxide surface. In addition, the surface reactivity is influenced by the redox pretreatments undergone by the sample. This work has been supported by the Italian ‘Minister0 della Pubblica Istruzione’ (Progetto Nazionale ‘Struttura e Reattivita’ delle Superfici’). We are indebted to Dr J. C . Lavalley for kindly supplying the ex-oxalate zinc oxide sample. References 1 W. Hotan, W. Gopel and R. Haul, SurJ Sci., 1979, 83, 162. 2 H. R. Gay, M. H. Nodine, E. I. Solomon, V. H. Henrich and H. J. Zeiger, J . Am. Chem. Soc., 1980, 3 S. Akther, K. Lui and H. H. Kung, J . Phys. Chem., 1985, 89, 1958. 4 B. Fubini, E. Giamello, G. Della Gatta and G. Venturello, J . Chem. SOC., Furuduy Truns. 1, 1982, 78, 102, 6752. 53.V. Bolis, B. Fubini, E. Giamello and A . Reller 867 5 M. Bowker, H. Houghton, K. C. Waugh, T. Giddings and M. Green, J. Catal., 1983, 84, 252. 6 E. Giamello and B. Fubini, J . Chem. Soc., Faraday Trans. I , 1983, 79, 1995. 7 G. Ghiotti, F. Boccuzzi and R. Scala, J . Catal., 1985, 92, 79. 8 J. C. Lavalley, J. Saussey and T. Rais, J . Mol. Catal., 1982, 17, 289; C. Chauvin, J. Saussey, J. C. 9 B. Fubini, V. Bolis and E. Giamello, Thermochim. Acta, 1985, 17, 283. Lavalley and G. Djega-Mariadassou, Appl. Catal., 1986, 25, 59. 10 G. Natta, in Catalysis, ed. P. H. Emmet (Reinhold, New York, 1955), vol. 3. 11 K. Klier, Adu. Catal., 1982, 31, 243. 12 C. H. Amberg and D. A. Seanor, in Proc. Third Znt. Congr. Catal. (North Holland, Amsterdam, 1964), 13 R. J. Kokes, A. L. Dent, C. C. Chang and L. T. Dixon, J . Am. Chem. SOC., 1972, 94, 4429. 14 F. Boccuzzi, E. Garrone, A. Zecchina, A. Bossi and M. Camia, J. Catal., 1978, 51, 160. 15 G. L. Griffin and J. T. Yates Jr, J. Chem. Phys., 1982, 77, 3744; 3751. 16 L. A. Denisenko, A. A. Tsyganenko and V. N. Filimonov, React. Kinet. Catal. Lett., 1984, 25, 23. 17 G. L. Griffin and J. T. Yates Jr, J . Catul., 1982, 73, 396. 18 M. Bowker, H. Houghton and K. C. Waugh, J . Chem. Soc., Faraday Trans. I , 1981,77,3023; 1982,78, 19 G. Heiland and H. Luth, in The Chemical Physics of Solid Surfaces and Heterogeneous Catalysis, ed. 20 B. Fubini, RPc. Gen. ThPrmique, 1979, 18, 297. 21 M. Maciejewski and H. R. Oswald Thermochim. Acta, 1985, 85, 39. 22 G. Carnisio, F. Garbassi, G. Petrini and G. Parravano, J. Catal., 1978, 54, 66. 23 D. Scarano, A. Zecchina and A. Reller, work in preparation. paper I. 22. 2573. D. A. King and D. P. Woodruff (Elsevier, Amsterdam, 1984), vol. 3, part B. Paper 8/01620H; Received 25th April, 1988