A. ZnO CASESTUDY
ZnO exhibits varied catalytic properties, being active for hydrogenation and dehydrogenation reactions, dehydration of alcohols, methanol synthesis, and other reactions. ZnO is a wide-bandn-type semiconductor with surface states present in the band gap. It can be doped with cations, and defects can be generated by treatment under oxidizing or reducing conditions (which change the availability of electrons at the surface).
Because of these properties, ZnO has been considered to be an ideal material for testing the electronic theories of catalysis, specifically the band model (based on the collective properties of the solid) or the localized site
320 A. ZECCHINA et al.
model (based on the reactivities of sites with atomically defined structures (390). These theories are beyond the scope of this review (391, 392). We state only that recently the localized site model has gained credibility, whereas the band model and the direct correlation between catalytic activity and semiconducting properties have been discounted.
The semiconducting properties of ZnO have nevertheless proved to be ad- vantageous because they permit the extensive application of modern surface spectroscopies (such as XPS, UPS, EELS, and LEED) for the investigation of the surface structures since the charging problems that are usually en- countered with oxides are avoided (393). For this reason, ZnO is one of the most thoroughly investigated oxides.
ZnO can be prepared as single crystals or microcrystalline powders. The single crystals often have needle-like shapes and preferentially expose pris- matic (10 ¯10) and (11 ¯20) faces. When examined by LEED, these neutral faces appear unreconstructed (393). In contrast the Zn and oxygen-rich positively and negatively charged (0001) faces prepared by cleavage of single crystals show a distinct tendency toward extensive reconstruction and accumulation of metal impurities.
Highly dispersed powders can be prepared in several ways. When pre- pared by combustion of metallic zinc, the resulting very pure ZnO (with
∼10 m2g−1specific surface area) is constituted of microcrystals characterized by well-defined elongated prismatic habits (Fig. 14) exposing preferentially unreconstructed (10 ¯10) and (11 ¯20) faces (394). When prepared by decom- position of ZnCO3, the resulting high-surface-area powder (∼50 m2g−1) is constituted of very small microcrystals with ill-defined shapes.
Because of the high purity, the excellent morphological definition, and the good optical properties of the microcrystals, the surface and catalytic prop- erties of ZnO powders prepared by Zn combustion have been investigated extensively by spectroscopic techniques. The results of these investigations are well suited to comparison with results obtained with single crystals and less well-defined samples.
On the virgin sample, the surfaces of the microcrystals are fully covered by adsorbed water and CO2; consequently, they do not show any adsorption by diatomic molecules such as CO and H2.
Activation at temperatures near 673 K is sufficient to clean the prismatic faces, whereas the sites located on the other (high index) faces, edges, steps, and corners remain covered by adsorbed impurities (mainly hydroxyl group).
For this reason, the samples activated at 673 K can be considered as ideal for investigation of the properties of the (10 ¯10) and (11 ¯20) faces of ZnO, the abundance of which is on the order (10 ¯10)>(11 ¯20). Samples treated at higher temperaturesin vacuohave been less thoroughly investigated because ZnO has a tendency to lose oxygen and become nonstoichiometric.
FIG. 14. HRTEM image of a ZnO microcrystal, where (10 ¯10) planes are evident. The arrows show the presence of terraces and steps along the prismatic direction and many small facets (having the same symmetry as the predominant one) at the prism terminations [reproduced from Scaranoet al.(394) with permission of Elsevier Science Publishers].
To a first approximation, the (10 ¯10) surface can be considered as a 2D square array of threefold coordinated zinc as shown schematically in Fig. 15a.
The structure of the (11 ¯20) face is very similar (Fig. 15b).
1. ZnO/CO
The IR spectra of increasing doses of CO adsorbed at 77 K on ZnO pre- treated at 673 K under vacuum are illustrated in Fig. 16a. These spectra are characterized by strong absorptions in the 2195- to 2165-cm−1range (395).
These absorptions, centered at frequencies higher than that of CO gas, cor- respond to polarized CO speciesσ bonded at the carbon end on Zn sites (394). In this review we comment only on the most important features of the spectra. For details, the reader is referred to Gayet al.(396).
Figure 16a shows that the spectrum of adsorbed CO changes gradually with coverage. In particular, atθ → 0 a single peak (probably due to the
322 A. ZECCHINA et al.
FIG. 15. Representation of the structure of (10 ¯10) and (11 ¯20) faces of ZnO. On both faces threefold coordinated Zn2+species are exposed. Zn2+are white; O2−are gray.
FIG. 16. (a) IR spectra of CO adorbed at 77 K on ZnO samples; species A are dominant in the 0≤θ≤0.25 coverage interval; species C corresponds to the complete occupation of the vacant Zn2+sites (θ = 1); and species B, B, and Bare dominant atθ ∼=0.5, 0.7, and 0.9, respectively. Weak bands labeled with one or two asterisks correspond to Zn2+–CO complexes located on edges and corners. (b) IR spectra of CO adsorbed at 77 K on hydrogen-precovered ZnO samples. All spectra were collected according to the procedure described in Scaranoet al.
(394).
superposition of two unresolved components), tailing into the low-frequency region, is observed at 2190 cm−1 (peak A). With increasing CO coverage, peak A gradually shifts to lower frequencies and then declines as new bands (B, B, and B) appear and disappear in succession. For θ→1, a single narrow component at 2168 cm−1 (C band) dominates. Like peak A, the C band is a composite, as can be seen at intermediate coverages, for which two components are detected. In our opinion, the composite structure of peaks A and C reflects the presence of two types of structurally similar Zn2+
sites located on the (10 ¯10) and (11 ¯20) faces, respectively. The stepwise evo- lution of the spectra of adsorbed CO as a function of coverage is a unique characteristic of ZnO.
The five bands arise from the stepwise occupation of the surface Zn2+sites and correspond to different surroundings of the CO molecules.
Species A is dominant in the 0≤θ≤0.25 coverage interval, and species C corresponds to the complete filling of the vacant Zn2+sites (θ=1). Spe- cies B, B, and Bare dominant at coveragesθ∼=0.5, 0.7, and 0.9, respectively.
At the lowest coverages, the frequency of adsorbed CO (singleton fre- quency) is observed at 2190 cm−1(A band), i.e., shifted by 47 cm−1with re- spect to CO gas. This is in agreement with results of high-resolution electron energy loss vibrational studies (397). This large upward shift [when compared with that observed for the only other divalent ion without d electrons that has been investigated (Mg2+)], can be explained in terms of the strong polarizing electrostatic field associated with the threefold co- ordinated Zn2+ions present on the (10 ¯10) and (11 ¯20) faces. On the basis of UPS results characterizing CO on single-crystal faces (396), the presence of a σ-type orbital overlap involving the 5σ-orbital of CO and an empty dangling sp3 hybrid with acceptor character centered on Zn2+ must be considered as well. As documented by Gay et al.(396), the CO axis is slightly tilted with respect to the surface plane. Direct support for the presence of a σ bond derives from the experimental adsorption enthalpy at zero cover- age (∼50 kJ mol−1) (396), which is definitely higher than that observed for CO on MgO (15–20 kJ mol−1), for which the electrostatic interaction plays the determining role (see Section IV.A.2).
An increase in the coverage from 0 to 1 shifts the peak downwards from 2190 to 2168 cm−1because lateral interactions between the adsorbed CO species increase. The total shift observed for the full monolayer (ν˜tot=
−22 cm−1) is the sum of two types of dipole–dipole interactions, static and dynamic. By using the method of the diluted isotopic mixtures (see Sec- tion IV.A.2), it has been determined that the total shift is the sum of two contributions (ν˜total=ν˜dynamic+ν˜static where ν˜dynamic=6 cm−1 and ν˜static= −28 cm−1).
324 A. ZECCHINA et al.
In comparing the shifts observed for CO on MgO and on NiO, we note that both the dynamic and static parts are intermediate between those observed for CO on MgO (3.5 and−11.3 cm−1) and on NiO (27 and−43 cm−1). This observation suggests that the dynamic and static shifts are minimal when the CO–surface interactions are purely electrostatic and increase with increasing σ- and π-orbital overlap contributions to the adsorbate–adsorbent bond (the largest shifts are found for CO on metals with partially filleddbands, for whichπ-overlap contributions are maximum). It is difficult to establish whether the static shift arises only from through-space effects (dipole–dipole interactions) or whether through-solid effects operate as well. A distinct and gradual decrease of the isosteric heat of adsorption with coverage of CO on prismatic faces has been ascribed to repulsive adsorbate–adsorbate interactions (396).
IR spectra of CO adsorbed on prismatic faces of ZnO show that the ap- parent half-widths of peaks A and C are∼4 and∼1.5 cm−1, respectively.
The decrease of the half-width with coverage [also clearly observed for CO on NaCl (100)] reflects the increasing collective character of the vibration observed on moving from low to high coverage, which becomes increas- ingly less influenced by inhomogeneous broadening effects associated with the presence of surface defects that interrupt the surface periodicity. The half-width of peak A (dominant at the lowest coverages) is larger than that observed for CO on NaCl (100) (see Section IV.C). This result can be ac- counted for both by the presence of a chemical bond between CO and Zn sites (which influences the lifetime of the excited states) and by inhomoge- neous broadening caused by site heterogeneity. The relative weights of the two contributions are not known.
The weak bands labeled with one or two asterisks in Fig. 16 are also note- worthy. These bands are observed at 2178 (shoulder) and 2184 cm−1forθmax, and their frequencies move gradually to a common position at 2190 cm−1 as the coverage is decreased. We assign these two bands to Zn2+–CO com- plexes formed on the same prismatic faces (therefore, both bands shift to the common 2190 cm−1frequency at zero coverage) but located near edges and corners (which explains the reduced shifts induced by increasing adsorbate–
adsorbate interactions because the dipolar effects near the edges and corners are approximately one-half and one-fourth of the total effect characteristic of the sites located at the centers of the faces) (394). If this assignment is correct, the intensity ratio between the main peak and the peaks at 2178 and 2184 cm−1(at full coverage) may provide a qualitative estimate of the real extension of the flat portions of the prismatic faces and terraces (qualita- tive but not quantitative because the intensity of the high-frequency bands could be influenced by intensity borrowing effects, as found for metallic surfaces). Note that the intensity ratio of the bands at 2184 and 2178 cm−1 relative to the main peak is larger for ZnO ex-carbonate, in agreement with
its more disordered morphology compared to ZnO microcrystals resulting from combustion of metallic zinc. These considerations indicate again that the IR spectra of adsorbed probes may provide information about the sur- face morphology. It also becomes evident that surface inhomogeneity has a strong influence on the frequencies and shapes of the IR peaks of adsorbed probe molecules.
In conclusion, CO is an excellent probe molecule for the properties of coordinatively unsaturated surface Zn2+sites, both isolated sites or those grouped into 2D patches. It is also emphasized that the prismatic faces also expose coordinatively unsaturated oxygen ions; their presence, however, is indirectly detected only via their influence on the electrostatic field at the Zn2+center.
2. ZnO/H2
When H2is used to probe ZnO surface sites, hydrides and hydroxyl groups are formed at temperatures as low as 80 K, as indicated by the immediate appearance of two medium strong bands at 3500 cm−1 (OH groups) and 1710 cm−1 (ZnH groups) (398–400). These bands are asymmetric on the low- and on the high-frequency side and are probably composed of several contributions. The formation of these bands is a clear indication that Zn ions and coordinatively unsaturated oxygen ions participate in the adsorp- tion process. Presumably, the species formed initially is a polarized form of molecular hydrogen adsorbed on top of the coordinatively unsaturated Zn2+ions. This species is transient in the 77–273 K temperature range (400) because it is sufficiently polarized to undergo a nucleophilic attack by the unsaturated oxygen ions in adjacent positions (heterolytic dissociation), as shown in Scheme 9. This reaction leads to formation of a hydride–hydroxyl pair. A Lewis acid–base pair operates in the dissociative adsorption of H2. It is thus evident that CO is suitable for probing the Lewis acidity, whereas H2 detects acid–base pairs. Because of the similarities of C–H and H–H bonds (both of theσ type), it is expected that ZnO can activate not only hydrogen but also alkanes (398, 401).
3. Model of Adsorption Sites for CO and H2on ZnO
In the preceding discussion it was implicitly assumed that the sites that are reactive for hydrogen chemisorption are the same ones that adsorb CO
SCHEME9
326 A. ZECCHINA et al.
SCHEME10
(located on the prismatic faces). If the sites located on the prismatic faces are involved in both hydrogen and CO adsorption it would be expected that adsorption of CO on prismatic faces precovered by hydrogen would be at least partially inhibited. The result of an experiment to test this possibility is illustrated in Fig. 16b, which shows that (i) the overall intensity of adsorbed CO is greatly reduced; (ii) the spectrum of adsorbed CO changes with cov- erage whereas atθ∼=0 the ˜ν(CO) is observed at 2197 cm−1and the value at the maximum coverage is 2183 cm−1; and (iii) the spectrum of adsorbed CO is complex (with at least three components clearly detected and intensities that depend on coverage).
Although the distinct reduction of the CO intensity caused by pread- sorption of hydrogen demonstrates beyond doubt that prismatic faces are reactive for both CO and hydrogen adsorption, the complex (coverage- dependent) structure of the residual band of CO after preadsorption of hydrogen needs further consideration. First, we infer that hydrogen does not block the Zn sites completely and that a fraction of sites remain co- ordinatively unsaturated. The coordinatively unsaturated ZnO pairs on the prismatic faces are all structurally equivalent. It must therefore be postulated that the presence of ZnH and OH groups formed by dissociative adsorption of H2creates repulsive forces that limit further H2adsorption, thus leaving a fraction of the Zn sites (one-half in our model) unoccupied. This postu- late accounts partially for the low coverage (∼10%) observed by Dent and Kokes (399). Furthermore, the frequency of adsorbed CO at zero cover- age (2196 cm−1; i.e., higher than the 2190 cm−1observed for CO on clean ZnO) suggests that the residual reactive Zn sites have acquired an enhanced tendency to polarize CO.
A plausible structure of the adsorbed hydrogen phase is characterized by rows of ZnHOH groups separated by rows of empty ZnO sites. In this ordered structure, each ZnH of the original ZnHOH pair has two empty Zn sites in nearest-neighbor positions along a perpendicular row, as shown in Scheme 10. These unoccupied sites are available for further adsorption of CO, as shown in Scheme 11.
In conclusion, three ZnH hydride species can be distinguished which dif- fer from each other in the number of CO molecules (0, 1, or 2) adsorbed in nearest-neighbor positions. An analogous situation occurs for the OH
SCHEME11
groups. Moreover, as the CO groups are aligned in rows, three arrangements of the CO molecules (with 0, 1, or 2 CO molecules in nearest-neighbor posi- tions) can be distinguished, the relative amounts of which change with cov- erage. Figure 17 demonstrates the effect of CO adsorption on the ˜ν(ZnH) mode. The ˜ν(ZnH) frequency shifts downward upon CO dosage in two clear and well-defined steps, a result that closely fits the scheme illustrated previ- ously. An opposite but smaller shift has been simultaneously observed for the ˜ν(OH). The physical reasons for the opposite signs of the frequency shifts of the ˜ν(ZnH) and ˜ν(OH) modes have been discussed (140, 402, 403). The observations are consistent with the shift of the C–O vibration in the two steps, as shown in Fig. 16b.
B. EXAMPLES OFREACTIONSCATALYZED BYZnO
Dent and Kokes (404) investigated the hydrogenation of ethylene cat- alyzed by ZnO; the available data were reviewed by Gates (405). Interest in this reaction is motivated by the relative simplicity of its mechanism and the possibility to detect kinetically relevant intermediates spectroscopically.
When the reaction is carried out with D2, the sole product is CH2DCH2D.
This result is in contrast to what is observed with metallic catalysts. How- ever, since ZnO and other active oxides (e.g.,α-Cr2O3) behave similarly, it
FIG. 17. IR spectra of CO adsorbed at 77 K on hydrogen-precovered ZnO samples; the effect of CO adsorption on the ˜ν(ZnH) modes is represented (unpublished data).
328 A. ZECCHINA et al.
is inferred that the result must be related to the structure of the adsorbed hydrogen (resulting from heterolytic and not homolytic splitting). The mech- anism shown in Scheme 12 has been proposed for this reaction. The ethylene molecules initially are adsorbed at empty Zn sites and then interact with hy- dride groups (obtained upon H2preadsorption) in adjacent positions, giving ethyl groups.
SCHEME12
FIG. 18. Effect of the ethylene absorption on the ˜ν(OH) and ˜ν(ZnH) modes, parts (a) and (b), respectively. Spectra A and B report the regions before and after ethylene adsorption, respec- tively. (Unpublished results).
The IR results discussed previously clearly suggested the first step. How- ever, the hydrogen migration illustrated in the second step is not demon- strated directly by the spectroscopic results. In particular, the H2–CO interaction experiments illustrated previously point to an opposite situation (the absence of hydrogen migration at room temperature). The closely corre- lated spectroscopic perturbations of hydride and hydroxyl groups induced by CO adsorption on the surrounding vacant Zn sites are much more readily ex- plained in terms of static interactions induced by CO adsorption on hydride–
OH pairs permanently located in adjacent position than on groups located at variable distances. Coadsorption of hydrogen and ethylene is illustrated in Fig. 18. The effects caused by ethylene adsorption on the ˜ν(ZnH) and ˜ν(OH) modes are similar to those caused by CO; this result clearly favors a similar interaction mechanism with formation of aπ-complex between ethylene and vacant Zn sites adjacent to the ZnH–OH pairs on (10 ¯10) faces and strongly suggests that oxygen sites are not directly involved in this interaction.
In summary, the formation of the Zn–ethyl intermediates is the result of the interaction of a ZnH (hydride) species with an ethene molecule adsorbed on an adjacent Zn center without direct intervention of an oxygen center and without hydrogen migration.
330 A. ZECCHINA et al.