Probing the Surface Heterogeneity of Polycrystalline Zinc Oxide by

J. Phys. Chem. C 2008, 112, 10931–10937

10931

Probing the Surface Heterogeneity of Polycrystalline Zinc Oxide by Static Adsorption Microcalorimetry. 2. The Adsorption of Carbon Monoxide Xinyu Xia, Jennifer Strunk, Raoul Naumann d’Alnoncourt, Wilma Busser, Lamma Khodeir, and Martin Muhler* Laboratory of Industrial Chemistry, Ruhr-UniVersity Bochum, D-44780 Bochum, Germany ReceiVed: January 25, 2008; ReVised Manuscript ReceiVed: May 6, 2008

The adsorption of CO on polycrystalline ZnO powder samples was investigated as a function of the pretreatment by applying static adsorption microcalorimetry and temperature-programmed desorption (TPD). Mainly weak adsorption sites were found to be present on the exposed ZnO surfaces but also a minor amount of highly active sites (>0.1 µmol/m2). On the majority sites, the adsorption of CO is weak, as reflected by the isotherms and the TPD profiles, with a maximum heat of adsorption of about 40 kJ/mol, which decreases with increasing coverage and also with an increasing amount of surface hydroxyl groups. The standard adsorption entropy is derived to amount to -102 J mol-1 K-1. These weak adsorption sites are hardly influenced by the pretreatment in flowing oxygen, hydrogen, or helium. On the highly active sites, as probed by small doses of CO in the calorimetric experiments, different exothermic (>100 kJ/mol) and endothermic surface reactions occur in addition to CO adsorption, depending on the gas atmosphere applied during the pretreatment. These results clearly indicate that the small amount of highly active sites accounts for the catalytic properties of ZnO. 1. Introduction It has been known for a long time that ZnO surfaces are strongly heterogeneous with respect to adsorption and catalytic reactions.1 For polycrystalline ZnO particles, Taylor et al.2 found multiple stages of hydrogen desorption, Krylov et al.3 reported broad activation energy distributions in the dehydration of isopropanol, and Kolboe4 determined several sets of adsorption sites according to the isopropanol thermodesorption profiles. Detailed information on structure sensitivity is accessible by studying single crystal faces.5 Grunze et al.6 investigated the interaction with oxygen, hydrogen, and water and found by applying infrared spectroscopy that only ZnO powders with a significant fraction of exposed polar surfaces show pronounced Zn-H and O-H bands upon hydrogen adsorption. Cheng et al.7 showed that the polar Zn-terminated ZnO(0001) surface and stepped nonpolar ZnO surfaces have higher activity than the nonpolar flat ZnO(101j0) surface in methanol decomposition, with formaldehyde being formed only over the polar ZnO(0001) surface. Some polar organic compounds including alcohols,8 aldehydes, carboxylic acids,9,10 terminal alkynes,11 and CH3SH and (CH3)2S212 adsorb dissociatively on the ZnO(0001) surface, but on the O-terminated ZnO(0001j) surface, only molecular adsorption occurs. In addition to the structural differences of these crystal faces, lattice defects such as oxygen vacancies also play an important role in the structure sensitivity of ZnO surfaces. For example, the yield of products of methanol decomposition was found to be higher over the reduced ZnO(101j0) than that over the stoichiometric ZnO(101j0) surface.7 The formation of methanol over polycrystalline ZnO/Al2O3 decreased remarkably in the presence of only a very low amount of carbon dioxide in the synthesis gas mixture, implying that oxygen vacancies are the active sites for the hydrogenation of carbon monoxide over ZnO.13 * To whom correspondence should be addressed. Phone: +49 234 32 28754. Fax: +49 234 32 14115. E-mail: [email protected]. URL: http://www.techem.rub.de.

Carbon monoxide is a useful probe molecule to study the structure sensitivity of ZnO surfaces in catalytic methanol synthesis because it is not only a reactant of this reaction but also a σ donor, which is suitable to characterize Zn2+ ions with different electronic environments.14 An early comparison of CO adsorption energies on different ZnO single-crystal faces was performed by Gay et al.15 based on a Clausius-Clapeyron analysis of ultraviolet photoelectron spectra, which yielded 50 kJ/mol for all of the investigated (0001), (0001j), (101j0), and (112j0) surfaces. This is a rather surprising result from a chemical point of view as there is no mechanism which would bind CO to the oxygen anions terminating the ZnO(0001j) surface. Recently, it was possible to determine CO binding energies by combing helium atom scattering (HAS) with thermal desorption spectroscopy (TDS).16 By applying this He-TDS method, which monitors the reflectivity of the surface for thermal helium atoms, the desorption of CO bound to the perfect parts of ZnO surfaces can be monitored. Staemmler et al.16 report a higher CO binding energy on the clean ZnO(0001) surface (26.9 kJ/mol) compared to that on the clean ZnO(0001j) surface (17.0 kJ/mol), in agreement with theoretical calculations. Furthermore, the sensitivity to a hydrogen pretreatment is different. After saturating the Zn-ZnO surface with adsorbed atomic hydrogen, it became repulsive to CO adsorption, while the CO binding energy on the O-ZnO surface was nearly unchanged (19.2 kJ/mol).16 Actually, for a clean, hydrogen-free ZnO(0001j ) surface, a (1 × 3) reconstruction is visible, which is rapidly converted even under ultrahigh vacuum (UHV) conditions into the stable hydroxylated (1 × 1) reconstruction.17 It was concluded that most of the previous experiments on the O-ZnO surface were indeed carried out on OH-ZnO,5 in agreement with a recent theoretical analysis of the stability of the polar ZnO surfaces.18 Thus, on the (1 × 1)-ZnO(0001j) surface, CO is actually adsorbed on top of the OH groups with the C-atom down being weakly held by electrostatic interactions.19,20 On the basis of the CO TPD data on the clean mixed-terminated ZnO(101j0) surface, a binding energy of 30.5 kJ/mol was obtained,21 which

10.1021/jp800756m CCC: $40.75  2008 American Chemical Society Published on Web 07/01/2008

10932 J. Phys. Chem. C, Vol. 112, No. 29, 2008 is comparable to the value of 26.9 kJ/mol observed on the Znterminated ZnO(0001) surface.16 Energetic heterogeneity on polycrystalline ZnO surfaces was also demonstrated by broad peaks in temperature-programmed desorption (TPD) spectra22 and by the decrease of the isosteric heat of adsorption with coverage derived by the ClausiusClapeyron method.23 Giamello et al.24,25 reported in a calorimetric study that the applied pretreatment affected the heat of adsorption of CO on polycrystalline ZnO; 44, 29, and 7 kJ/mol were found on ZnO samples pretreated in oxygen, in vacuum, and in hydrogen, respectively. They claimed that CO adsorption on ZnO can be described by the Langmuir isotherm and that the homogeneity of the ZnO surfaces is not changed by these different pretreatments.24,25 Despite these studies, the extent of surface heterogeneity, the precise initial value, and the coverage dependence of the heat of adsorption of CO on polycrystalline ZnO are still under debate. First, the values mentioned above scatter strongly. This may be attributed first of all to the weak interaction between CO and ZnO, which results in large error bars in both calorimetric and volumetric measurements, especially at room temperature. In addition, the amount of surface hydroxyl groups and active sites may vary depending on the pretreatment used in different studies. In this contribution, both TPD and microcalorimetry were applied to study the adsorption of CO on polycrystalline ZnO because the coverage regimes complement each other; in a TPD measurement, with a large CO uptake during dosing at lower temperatures, the overall properties of the ZnO surfaces at high coverages are addressed, while in a microcalorimetric measurement, small CO doses are utilized, thus probing at low coverages both the strongest adsorption sites and highly active sites inducing chemical reactions. ZnO samples pretreated under different conditions are compared in order to study the influence of coadsorbed species such as hydroxyl groups and of oxygen vacancies. The interaction of polycrystalline ZnO powder samples with carbon dioxide was reported in the preceding first part.26 2. Experimental Section 2.1. Samples. Two different commercially obtained zinc oxide powder samples were applied, which had both been synthesized by flame processes; the pyrogenic ZnO sample supplied by Degussa (AdNano, labeled ZnO-A) had a specific surface area of 12.1 m2/g after heating at 450 °C for 4 h. The second ZnO sample synthesized by oxidation of zinc vapor was supplied by Nanophase Technologies (labeled ZnO-N) and had a specific surface area amounting to 10 m2/g after the same thermal treatment. Its synthesis and specific surface area are close to the ZnO Kadox 25 sample formerly supplied by New Jersey Zinc Co. (8.5-10 m2/g), which has been used frequently in the literature as the polycrystalline ZnO sample for the investigation of CO adsorption.25,27 Coprecipitation of zinc and aluminum nitrates with sodium carbonate at a constant pH value of 7 followed by aging, filtration, washing, drying, and calcination at 320 °C yielded a ZnO/Al2O3 mixed oxide with 15 mol% Al2O3.13 For all measurements, the powder samples were pressed, crushed, and sieved applying the sieve fraction of 250-355 µm. 2.2. Microcalorimetric Measurements. The microcalorimeter setup and the measurement procedures were the same as those described recently.28,29 The adsorption temperature was T ) 30 °C, and the volume of the adsorption cell was V ) 129 ( 1 cm3, which was calibrated in each measurement. In this setup, the heat flow was measured by a Tian-Calvet micro-

Xia et al. calorimeter, and the pressure was recorded every second by means of a pressure transducer. The uptake was derived from the pressure measurements assuming ideal gas properties. For the calorimetric measurements, three kinds of samples were analyzed after ex situ pretreatment at 450 °C in different flowing gases at ambient pressure followed by transfer without contact with air. All flow rates are reported as normal mL/min (NmL/min, normal conditions 0 °C, 1013 mbar). The mass of the sample used for each measurement was around 0.5 g. An oxidized sample, that is, a sample with stoichiometric surface composition, was obtained by flushing the pretreatment reactor with 10% O2 in Ar using a flow rate of 25 NmL/min during the whole duration of the pretreatment. A heated sample was obtained by flushing the reactor with helium using a flow rate of 25 NmL/min, resulting in more oxygen vacancies than that in oxidized ZnO.30 A reduced sample was obtained by subjecting the heated sample additionally to a H2 flow at 250 °C for 1 h, followed by cooling in helium. Each of the ex situ pretreated samples had been stored in a sealed pyrex capsule before it was transferred into the adsorption cell in the microcalorimeter setup. In order to study the effect of different amounts of surface hydroxyl groups on CO adsorption, a ZnO-A sample pretreated at 250 °C for 4 h and a hydrated ZnO-A sample were also analyzed. The latter was obtained by exposing the ZnO-A sample oxidized at 450 °C to water vapor in the calorimeter cell followed by baking and evacuation at 120 °C for 72 h. The adsorption of CO was performed by successive dosing of CO (purity: 99.997%). Nitrogen (purity: 99.9999%) was applied for pressure compensation. The adsorption measurements on the same sample were repeated several times in order to test the reversibility, with an evacuation of the sample overnight at 30 °C between the measurements. 2.3. TPD Measurements. Details of the TPD setup were reported elsewhere.31 It included a gas supply unit with seven gas lines, a heated U-tube reactor, and a calibrated quadrupole mass spectrometer for fast quantitative online gas analysis. In the CO TPD experiments, 0.20 g of ZnO-N were used. The oxidized sample was obtained by heating at 450 °C in a 10% O2/Ne flow for 4 h; the reduced sample was obtained by heating at 300 °C and keeping at this temperature in a 2% H2 flow overnight, followed by keeping it in pure H2 for 1 h. The heated sample was obtained by heating at 450 °C in He flow for 4 h. The desorption of CO from fully covered ZnO was investigated in the following way. CO was adsorbed at 300 K in a flowing mixture of CO in He (10% CO, 10 NmL/min), and then, the sample was cooled to 78 K rapidly in flowing CO/He by applying liquid nitrogen cooling; at 78 K, the sample was purged with pure He (10 NmL/min) for 10 min and heated by 5 K/min up to 320 K in the same He flow. The desorption of CO from partially covered ZnO followed the above procedures except for the last step; the sample was heated by 5 K/min up to 210 K in pure He flow (10 NmL/min), then it was cooled again to 78 K, and finally, it was heated in pure He flow with a linear heating ramp continuously up to the final temperature. The adsorption energy can be derived reliably following our recently established method. By using a uniform distribution of adsorption sites with different energy, the following equation was derived to describe first-order TPD kinetics32

β

dΘ ) (ka p - Ad Kdiff)(1 - Θ) dT

(1)

where β is the heating rate, Θ is the fractional coverage on an energetically heterogeneous surface, T is the temperature (in K), ka is the adsorption rate constant (in Pa-1 s-1), assumed to

Adsorption of Carbon Monoxide

J. Phys. Chem. C, Vol. 112, No. 29, 2008 10933

Figure 1. Isotherms of CO adsorption on ZnO. Upper plot A: ZnO-A (1) nonpretreated; (2) oxidized at 250 °C; (3) oxidized at 450 °C; (4) heated in He at 450 °C; (5) reduced at 450 °C; (6) ZnO-N oxidized at 450 °C; (7) ZnO-A oxidized at 450 °C for 4 h and then hydrated. Lower plot B: repeated adsorption experiments with ZnO-A oxidized at 450 °C.

be independent of the adsorption sites, p is the CO pressure, Ad is the pre-exponential factor of desorption (in s-1), assumed to be independent of the adsorption sites, and Kdiff is a coefficient reflecting the diffusion equilibrium between adsorbates on sites with different binding energies (εd)

(

exp Kdiff )

)

(

εd, max - Θεδ εd, max - exp RT RT εδ 1 - exp -(1 - Θ) RT

[

]

)

(2)

where εd,max is the maximum εd in a uniform distribution, εδ is the width of this uniform distribution, and R is the gas constant. The pressure of the adsorptive gas is related to the desorption rate in a continuously operated stirred tank reactor (CSTR) model by

p)

nmmARTa dΘ ˙ dT V

(3)

where nm is the density of adsorption sites on the surfaces (in mol/m2), m is the sample mass, A is the specific surface area, Ta is the ambient temperature, and V˙ is the volumetric flow rate of the carrier gas (in Nml/min). 3. Results and Discussion 3.1. The Influence of Residual OH Groups on CO Adsorption. Figure 1 compares the adsorption isotherms of CO on ZnO samples after different pretreatments. The uptake of CO on the hydrated sample, on the sample pretreated at

Figure 2. CO TPD profiles obtained with ZnO-N. Upper plot A: from fully covered oxidized and reduced samples and from the partially covered heated sample. Heating rate: 5 K/min. He flow rate: 10 NmL/ min. Lower plot B: from partially covered heated samples. Initial coverage: (1) 1.18 µmol/m2, (2) 1.14 µmol/m2, (3) 1.02 µmol/m2, (4) 0.73 µmol/m2; flow rate: (1-3) 10 NmL/min, (4) 50 NmL/min; heating rate: (1) 10 K/min, (2 and 4) 5 K/min, (3) 2 K/min. Solid lines: experimental data; dashed lines: theoretical values of eq 1 with εmax ) 38.5 kJ/mol, ∆S0 ) -102 J mol-1 K-1, and εδ/nm ) 4.2 (kJ/mol)/ (µmol/m2).

250 °C, as well as on the nonpretreated sample is rather low and much smaller compared to the samples pretreated at 450 °C (Figure 1A, top). This observation clearly indicates that the adsorption sites for CO are preferably occupied by surface hydroxyl groups. In addition, residual strongly bound carbonate species may block adsorption sites for CO but not on the exposed nonpolar ZnO(101j0) faces with a heat of CO2 adsorption below 120 kJ/mol.26 For the samples pretreated at 450 °C, the coverage on the reduced sample is slightly lower than that on the heated and on the oxidized sample at the experimental pressure. Isotherms obtained by repeated adsorption show that the adsorption is fairly reversible (Figure 1B, bottom); therefore, only a small fraction of the exposed surfaces have been involved in irreversible overlapping reactions discussed in detail in section 3.2. The linear increase of the equilibrium coverage with pressure indicates that the fractional coverage is very low, no matter whether the surface is homogeneous or heterogeneous.33 Assuming that the Zn2+ adsorption site density on the ZnO(101j0) surface is 9.8 µmol/m2 results in an estimated fractional coverage of less than 0.02. The CO TPD profiles of the ZnO-N sample are shown in Figure 2. In the profiles from the fully covered samples (Figure 2A, top), there is no separation between physisorption and chemisorption. The broad continuous peak shape reflects the energetic heterogeneity of the adsorption sites. This heterogeneity can be attributed to the existence of structural or electronic heterogeneity of the surface prior to adsorption, decreasing Lewis acidity with increasing CO coverage, and increasing repulsive interactions between the adsorbed CO molecules. In

10934 J. Phys. Chem. C, Vol. 112, No. 29, 2008

Xia et al.

a careful FTIR study of CO adsorption on ZnO Kadox 25 at 77 K and at room temperature, it was observed that the CO stretching frequency νCO shifted gradually to lower wavenumbers with increasing coverage, indicating lateral interactions.27 The coverage dependence of the CO adsorption energy and the corresponding shift of νCO were recently reproduced experimentally and by DFT calculations on ZnO(101j0).19,34 The TPD profiles of the fully covered oxidized and reduced samples are quite similar, and the profile of the partially covered heated sample also follows the profiles of the fully covered samples at higher temperatures. Whereas the shape of the profiles is virtually identical, the heights differ slightly, indicating deviations in the amount of the various adsorption sites but not strongly differing adsorption energies as a function of the pretreatment. The coinciding TPD profiles provide strong evidence that the majority of the exposed adsorption sites is hardly influenced by the redox potential of the gas atmosphere applied during the pretreatment. As a consequence, the differential heats of adsorption of CO on polycrystalline ZnO are essentially independent of the pretreatment atmosphere, in contradiction to the calorimetric results by Giamello and Fubini.25 When performing TPD experiments with powder samples in a flow setup, surface heterogeneity is revealed by changing the initial coverage, the effect of readsorption can be determined by changing the flow rate of the carrier gas, and the effects of enthalpy and entropy (or pre-exponential factors) are discriminated by changing the heating rate. The broad peak shape and the shift of the peak position as a function of the flow rate (traces 2 and 4 in Figure 2B, bottom) indicate that readsorption cannot be omitted. As supported by modeling the peaks with eq 1, the experiments were performed under so-called free readsorption conditions,35 ka > 510 Pas and kanmmARTa/V˙ , 1. Under these conditions eq 1 changes to

β

( ) ( )

εa dΘ ∆S0 p0 exp ) - Kdiff exp dT p RT R

Figure 3. Fractional coverage of CO on adsorption sites with different adsorption energy as a function of pressure (100 Pa, 1 kPa, 10 kPa) and OH coverage. Simulating parameters: εmax ) 48.5 kJ/mol for OH coverage equal to 0, ∆S0 ) -102 J mol-1 K-1, and εδ/nm ) 4.2 (kJ/ mol)/(µmol/m2). Solid lines: OH coverage of 2 µmol/m2; dashed and dotted lines: OH coverage of 3 µmol/m2.

(4)

with standard adsorption entropy ∆S0 ) R ln[(Aap0)/Ad], adsorption energy ε ) εd - εa, and ka ) Aa exp(-εa/RT). Aa and εa are the pre-exponential factor and the activation energy of adsorption, respectively, and p0 is the standard pressure. The following thermodynamic parameters were obtained based on eq 4: the maximum adsorption energy εmax ) 38.5 kJ/mol and the standard adsorption entropy ∆S0 ) -102 J mol-1 K-1. The εδ and nm depend on each other, with εδ/nm ) 4.2 (kJ/mol)/ (µmol/m2). The low value of εmax indicates a very weak interaction between CO and the majority part of the ZnO surfaces, and the value of ∆S0 is typical for weak CO adsorption without backbonding. The ratio εδ/nm is also meaningful; on the one hand, on the fully covered sample, nm ) 5.0 µmol/m2, derived from the effluent CO mole fraction. This coverage is smaller than 9.8 µmol/m2, the density of Zn2+ sites on the ZnO(101j0) surface, because a fraction of the Zn2+ sites is occupied by hydroxyl groups. Consequently, εδ is about 21 kJ/mol, and the adsorption energy of CO on the least-active sites is derived to be εmin ) 18 kJ/mol. On the other hand, this value can be applied to estimate the effect of the amount of surface hydroxyl groups on CO adsorption. Considering the “Condensation Approximation”,36,37 εmax is equal to the differential heat of adsorption of the least strongly bound hydroxyl groups on ZnO. Therefore, the value of εδ/nm ) 4.2 (kJ/mol)/ (mol/m) also means that if 1 µmol/m2 of hydroxyl groups is removed from Zn2+ sites, then εmax for CO adsorption will increase by 4.2 kJ/mol. This is illustrated in Figure 3.

Figure 4. Simulation of isotherms (upper plot A) and of qdiff (lower plot B) as a function of OH coverage at T ) 30 °C. The simulation parameters are the same as those for Figure 3 based on the TPD results obtained with the oxidatively pretreated ZnO-N sample. Solid lines: coverage of hydroxyl groups ) 2.8 µmol/m2; dashed-dotted line: 2.5 µmol/m2; dotted line: 2.2 µmol/m2; triangles: Kadox ZnO pretreated at 400 °C;25 circles: ZnO-N pretreated at 450 °C; squares: Kadox ZnO pretreated at 500 °C.27

With such a model and assuming a residual OH coverage of nOH ) 2.5 µmol/m2 for the samples pretreated at 450 °C,26 several isotherms of CO adsorption on ZnO from our work and from the literature can be well evaluated, and the differential heat of adsorption (qdiff) of CO adsorption on ZnO can be successfully simulated, as shown in Figure 4. It illustrates how the occupation of high-energy sites by OH groups decreases both the coverage and qdiff of CO adsorbed on ZnO. In the TPD spectra from the fully covered samples, there is a small peak at T ) 197 K, and according to its narrow width and as it is clearly separated from the broad continuous CO desorption profile, it reflects special surface sites with a rather homogeneous energy distribution, which are less involved in surface diffusion. By line shape analysis, that is, by fitting the Wigner-Polanyi equation with the known time-dependent pressure, coverage, and adsorption/desorption rate using the above ∆S0, ∆H ) -34 kJ/mol is obtained. These sites may be

Adsorption of Carbon Monoxide

Figure 5. Differential heat of adsorption of CO on oxidized ZnO as a function of the uptake. (1-5): ZnO-A 450 °C oxidized first-fifth adsorption; (6): ZnO Kadox 15 oxidized at 400 °C;25 dashed-dotted line: values according to the volumetric results with OH coverage of 2.5 µmol/m2.

located on the highly crystalline nonpolar surface, as suggested by the inhibited surface diffusion of the adsorbates, the medium adsorption energy, and absent influence of the gas-phase composition used during the pretreatment. The existence of this small peak means that the distribution of the adsorption energy slightly deviates from a uniform distribution. Nevertheless, its influence on our simulation can be ignored because we focus on the sites with higher energy than this peak, both in the TPD experiments from partially covered samples and in static adsorption. Interestingly, on ZnO(101j0), a phase transition from the high-coverage (1 × 1) adlayer of CO to a presumably (2 × 1) phase was observed by HAS at low temperatures.5 The corresponding desorption maximum at 120 K obtained by HeTDS corresponds to an activation energy of desorption of 30.9 kJ/mol.5 3.2. Surface Reactions on Active Sites during Adsorption. Figure 4 predicted the range of qdiff in agreement with the volumetrically obtained isotherms. However, in the calorimetric measurements, much more complicated variations of qdiff were observed at low coverages. Figure 5 reports qdiff of CO on several ZnO samples with the stoichiometric surface composition due to a pretreatment in oxygen. Compared to Figure 4, the initial qdiff is much higher and not reproducible, that is, it decreases strongly during the first repeated measurements. The higher initial qdiff values of CO on oxidized ZnO can be attributed to the oxidation of CO. The formation of CO2 and carbonate when exposing the oxidized ZnO sample to CO has also been found in the literature.13,27 Therefore, the qdiff obtained in the first dose of each measurement are an average heat of CO adsorption, CO oxidation, and carbonate formation. As a consequence, they are not reversible and decrease in each repeated adsorption measurement. The effect of CO oxidation can only be observed with a small CO dosing pressure. When the dosing pressure is large enough, as was the case in the study by Giamello and Fubini,25 the reactions over a few highly active sites are obscured by adsorption; thus, qdiff is more stable, as shown in Figure 4. From their reported isotherms and heats of adsorption (44 kJ/mol), the standard adsorption entropy is -121 and -118 for the oxidized Kadox and ex-carbonate ZnO sample, respectively, which is close to the value obtained in our work, and the entropy values also indicate that the overlapping surface reactions are not significant for the oxidized samples. The low qdiff value of about 10 kJ/mol in Figure 5 may be due to experimental limitations when the uptake of CO becomes

J. Phys. Chem. C, Vol. 112, No. 29, 2008 10935

Figure 6. Dependence of qdiff on the CO dosing pressure observed for ZnO-A; as the dosing pressure (circles) increases, qdiff (filled squares) increases in successive dosing experiments.

Figure 7. Differential heats of CO adsorption on reduced ZnO as a function of the uptake. Upper plot A: on reduced ZnO, first and repeated second adsorption after evacuation; lower plot B: on ZnO/Al2O3 obtained by coprecipitation after reduction in H2 orCO.

very small, or it may originate from simultaneously occurring endothermic reactions. This point is addressed in Figure 6, which illustrates that qdiff depends on the CO dosing pressure. During a sequence of CO dosing experiments, the CO dosing pressure was lowered and increased again. Despite the increase in coverage, qdiff increased with increasing dosing pressure. This observation indicates that there is an endothermic process overlapping the exothermic CO adsorption. Its relative extent becomes stronger when CO adsorption contributes less to the overall released heat, thus interfering with the value of qdiff. This endothermic process is more pronounced after the pretreatment in helium, as discussed subsequent to Figure 8. A similar phenomenon has also been observed during CO2 adsorption on ZnO.26

10936 J. Phys. Chem. C, Vol. 112, No. 29, 2008

Xia et al.

Figure 8. Differential heat of CO adsorption on sample ZnO-A pretreated in helium at 450 °C as a function of the uptake. (1-5): five sequential adsorption experiments.

Figure 7 shows the heats of adsorption of CO on reduced ZnO. In the measurement directly after the pretreatment, qdiff is about 10-30 kJ/mol, whereas there is no significant difference between qdiff of CO on the oxidized sample and that on the reduced sample at a coverage above 0.1 µmol/m2. This observation proves again that only a minor part of the ZnO surfaces can be modified by the pretreating gas, as reflected in the TPD profiles (Figure 2). This result is different from Giamello and Fubini’s work,25 who reported that the qdiff of CO amounted to just 7 kJ/mol for a coverage up to 0.4 µmol/ m2. This value should also be attributed to some overlapping reactions because a value of ∆S0 ) +11.5 is derived from this value and the corresponding isotherm, which is obviously not a reasonable entropy change for adsorption. During the repeated measurements of CO adsorption on the reduced ZnO-A sample, an extremely high value of qdiff > 170 kJ/mol was observed, as shown in Figure 7, upper plot. It may arise from the generation of formate by hydroxylation of adsorbed CO. French et al.38 calculated the enthalpy of formate generated from CO2 over ZnO

1 CO2(g) + H2(g) f HCO2(ads) + 285 kJ/mol 2

(5)

On the basis of the heat of reaction of the water-gas shift reaction of 41 kJ/mol, the heat of dissociative adsorption of water on ZnO of 60-150 kJ/mol,39,40 and that of hydrogen on ZnO of 63 kJ/mol,41 the heat of the following reaction can be estimated to amount to roughly 208-298 kJ/mol

CO(g) + OH(ads) f HCO2(ads)

(6)

It has to be noted that the heat obtained in adsorption measurements is normalized by the amount of adsorbed CO instead of hydroxylated CO. Therefore, the reaction heat of CO hydroxylation over ZnO cannot be derived from our experimental data. Formate species were found by FTIR spectroscopy at 100 °C when evacuating ZnO (Kadox) at 400 °C and exposing it subsequently to 1.33 kPa of CO for 30 min.42 It is expected that this exothermic effect observed during the first CO adsorption experiment for ZnO-A is even stronger for a ZnO/ Al2O3 sample prepared by coprecipitation, which has a higher degree of surface hydroxylation. Indeed, this is the case as shown in Figure 7, lower plot. Figure 8 shows the heats of adsorption of CO on ZnO-A heated in helium 450 °C. The initial qdiff value in the first adsorption is about 60 J/mol, which is between the corresponding value of the reduced and the oxidized sample, but with increasing coverage, negative qdiff values were detected during

Figure 9. Isochoric kinetics of CO adsorption after the first dose on ZnO-A samples after different pretreatment methods. (1) oxidized sample; (2) sample heated in helium; (3) reduced sample. The smooth traces are the best-fitting results with ka ) 5.0 × 10-7 Pa-1 s-1 for trace 1, 4.3 × 10-7 Pa-1 s-1 for trace 2, and 0.82 × 10-7 Pa-1 s-1 for trace 3.

the first and second adsorption experiments. An extremely high value occurred at the beginning of the second adsorption. After repeating the measurements several times, the qdiff reached a value of 20-40 J/mol when the coverage was about 0.1 mu µmol/m2. Thus, the surface of ZnO heated in helium is not merely a state between the oxidized and reduced sample. The strong exothermic reaction may belong to the formation of formate, whereas it is more difficult to rationalize the endothermic process for which a positive entropy change is necessary. Possible processes that may account for the positive entropy change include surface relaxation or diffusion of hydrogen from the bulk to the surface43,44 associated with an increase in configurational or vibrational entropy. Surface reactions during CO adsorption on ZnO at low coverages are also reflected in the kinetic measurements. Figure 9 shows the kinetics of CO adsorption on ZnO-A after the first dose of each measurement, with the initial uptake of n0 ) 0. It takes over 1 h for the uptake to approach steady state. By fitting a linearized form of the Wigner-Polanyi equation,45 ka is obtained with an order of magnitude of 10-7 Pa-1 s-1. In comparison to other typical adsorption systems such as CO2 on ZnO and CO on Cu and Au catalysts measured under similar temperature and pressure conditions, ka is lower by 2 orders of magnitude. In these systems, the time to reach equilibrium is less than 10 min. It is also smaller than ka derived from the TPD measurements, clearly indicating that these surface reactions are significantly slower than CO adsorption. In conclusion, different exothermic and endothermic reactions occur in addition to CO adsorption depending on the pretreatment and the temporal course of the exposure to CO. These reactions were found to be difficult to reproduce as exchange reactions with the bulk may also occur.46 The ZnO sample heated at 450 °C in inert gas is the most active sample of all three investigated ones, with the most complex surface processes on the minority sites. The existence of these highly active sites provides further evidence for the postulated structure sensitivity of polycrystalline ZnO powder applied in heterogeneous catalysis. More attention should be paid to these minority sites instead of characterizing the overall properties of the exposed ZnO surfaces to achieve a better understanding of the reaction mechanisms catalyzed by ZnO. Further studies employing FTIR spectroscopy are in progress to identify the products into which adsorbed CO is converted.

Adsorption of Carbon Monoxide 4. Conclusions By combining microcalorimetric and TPD methods, reliable heats of adsorption of CO on polycrystalline ZnO were obtained. The adsorption sites of CO on the exposed ZnO surfaces were found to consist of a majority of less active sites and a minority of highly active sites (>0.1 µmol/m2), on which adsorbed CO molecules were converted to different products. Taking the competing occupation of surface sites by hydroxyl groups into account and by a combined analysis of TPD and isotherm data, the heat of CO adsorption on the majority part of properly pretreated polycrystalline ZnO surfaces was derived to amount to about 40 kJ/mol, and the standard entropy was found to be about -102 J mol-1 K-1. The differential heats of adsorption decreased slightly with increasing CO coverage and were found to depend more on the temperature of the pretreatment than on its gas-phase composition. On the highly active sites, adsorbed CO reacted irreversibly, as indicated by the high initial differential heats of reaction, and it was possible to modify the extent of these reactions by the redox potential of the gas phase applied in the pretreatment. Although the interaction between CO and these minority sites was not reflected in the TPD profiles probing the overall surface properties, it was possible to access it calorimetrically. However, the overlapping adsorption and reaction processes did not allow exact discrimination of the heats of adsorption and reaction. Acknowledgment. The authors thank Bernd Meyer, Yuemin Wang, and Christof Wo¨ll for fruitful discussions and the Deutsche Forschungsgemeinschaft (DFG) for financial support within the Collaborative Research Center (SFB 558) “MetalSubstrate Interactions in Heterogeneous Catalysis”. References and Notes (1) Henrich, V. E.; Cox, P. A. The Surface Science of Metal Oxides; Cambridge University Press: Cambridge, U.K., 1994. (2) Taylor, H. S.; Liang, S. C. J. Am. Chem. Soc. 1947, 69, 1306. (3) Krylov, O. V.; Fokina, E. A. Kinet. Katal. 1960, 1, 542. (4) Kolboe, S. J. Catalysis 1969, 13, 199. (5) Wo¨ll, C. Prog. Surf. Sci. 2007, 82, 55. (6) Grunze, M.; Hirschwald, W.; Hofmann, D. J. Cryst. Growth 1981, 52, 241. (7) Cheng, W. H.; Akhter, S.; Kung, H. H. J. Catal. 1983, 82, 341. (8) Vest, M.; Berlowitz, P. J.; Kung, H. H. Stud. Surf. Sci. Catal. 1988, 38, 577. (9) Vohs, J. M.; Barteau, M. A. J. Phys. Chem. 1989, 93, 8343. (10) Barteau, M. A.; Vohs, J. M. Stud. Surf. Sci. Catal. 1989, 44, 89. (11) Vohs, J. M.; Barteau, M. A. J. Phys. Chem. 1987, 91, 4766. (12) Halevi, B.; Vohs, J. M. J. Phys. Chem. B 2005, 109, 23976. (13) Kurtz, M.; Strunk, J.; Hinrichsen, O.; Muhler, M.; Fink, K.; Meyer, B.; Wo¨ll, C. Angew. Chem., Int. Ed. 2005, 44, 2790.

J. Phys. Chem. C, Vol. 112, No. 29, 2008 10937 (14) Solomon, E. I.; Jones, P. M.; May, J. A. Chem. ReV. 1993, 93, 2623. (15) Gay, R. R.; Nodine, M. H.; Henrich, V. E.; Zeiger, H. J.; Solomon, E. I. J. Am. Chem. Soc. 1980, 102, 6752. (16) Staemmler, V.; Fink, K.; Meyer, B.; Marx, D.; Kunat, M.; Gil Girol, S.; Burghaus, U.; Wo¨ll, C. Phys. ReV. Lett. 2003, 90, 106102-1. (17) Kunat, M.; Gil-Girol, S.; Becker, T.; Burghaus, U.; Wo¨ll, C. Phys. ReV. B 2002, 66, 081402-1. (18) Meyer, B. Phys. ReV. B 2004, 69, 045416. (19) Meyer, B.; Marx, D. J. Phys.: Condens. Matter 2003, 15, L89. (20) Kunat, M.; Meyer, B.; Traeger, F.; Wo¨ll, C. Phys. Chem. Chem. Phys. 2006, 8, 1499. (21) Wang, Y.; Xia, X.; Urban, A.; Qiu, H.; Strunk, J.; Meyer, B.; Muhler, M.; Wo¨ll, Ch. Angew. Chem., Int. Ed. 2007, 46, 7315. (22) Griffin, G. L.; Yates, J. T., Jr J. Chem. Phys. 1982, 77, 3751. (23) Hart, P. M. G.; Serba, F. Trans. Faraday Soc. 1960, 56, 551. (24) Garrone, E.; Ghiotti, G.; Giamello, E.; Fubini, B. J. Chem. Soc., Faraday Trans. 1 1981, 77, 2613. (25) Giamello, E.; Fubini, B. J. Chem. Soc., Faraday Trans. 1 1983, 79, 1995. (26) Xia, X.; Strunk, J.; Busser, W.; Naumann d’Alnoncourt, R.; Muhler, M. J. Phys. Chem. C 2008, 112, 10938. (27) Scarano, D.; Spoto, G.; Bordiga, S.; Zecchina, A.; Lamberti, C. Surf. Sci. 1992, 276, 281. (28) Naumann d’Alnoncourt, R.; Kurtz, M.; Wilmer, H.; Lo¨ffler, E.; Hagen, V.; Shen, J.; Muhler, M. J. Catal. 2003, 220, 249. (29) Xia, X.; Naumann d’Alnoncourt, R.; Strunk, J.; Litvinov, S.; Muhler, M. J. Phys. Chem. B 2006, 110, 8409. (30) Wilmer, H. Kinetische Untersuchungen zur Bedeutung von MetallTra¨ger-Wechselwirkungen fu¨r die CO-Konvertierung und Methanolsynthese. Ph.D. Thesis, Ruhr-University, Bochum, Germany, 2003. (31) Strunk, J.; Naumann d’Alnoncourt, R.; Bergmann, M.; Litvinov, S.; Xia, X.; Hinrichsen, O.; Muhler, M. Phys. Chem. Chem. Phys. 2006, 8, 1225. (32) Xia, X.; Strunk, J.; Litvinov, S.; Muhler, M. J. Phys. Chem. C 2007, 111, 6000. (33) Xia, X.; Litvinov, S.; Muhler, M. Langmuir 2006, 22, 8063. (34) Wang, Y.; Meyer, B.; Yin, Y.; Kunat, M.; Langenberg, D.; Traeger, F.; Birkner, A.; Wo¨ll, Ch. Phys. ReV. Lett. 2005, 95, 266104. (35) Cvetanovic´, R. J.; Amenomiya, Y. AdV. Catal. 1967, 17, 103. (36) Roginskii, C. Z. Acta Physicochim. USSR 1945, 20, 227. (37) Rudzinski, W.; Panczyk, T. J. Phys. Chem. B 2000, 104, 9149. (38) French, S. A.; Sokol, A. A.; Bromley, S. T.; Catlow, C. R. A.; Rogers, S. C.; King, F.; Sherwood, P. Angew. Chem., Int. Ed. 2001, 40, 4437. (39) Nagao, M.; Morimoto, T. J. Phys. Chem. 1969, 73, 3809. (40) Nagao, M.; Yunoki, K.; Muraishi, H.; Morimoto, T. J. Phys. Chem. 1978, 82, 1032. (41) Jones, P. M.; May, J. A.; Reitz, J. B.; Solomon, E. I. Inorg. Chem. 2004, 43, 3349. (42) Shido, T.; Iwasawa, Y. J. Catal. 1991, 129, 343. (43) Van de Walle, C. G. Phys. ReV. Lett. 2000, 85, 1012. (44) Shi, G. A.; Saboktakin, M.; Stavolaa, M.; Pearton, S. J. Appl. Phys. Lett. 2004, 85, 5601. (45) Xia, X.; Naumann d’Alnoncourt, R.; Strunk, J.; Litvinov, S.; Muhler, M. Appl. Surf. Sci. 2007, 253, 5851. (46) Polarz, S.; Strunk, J.; Ischenko, V.; van den Berg, M. W. E.; Hinrichsen, O.; Muhler, M.; Driess, M. Angew. Chem., Int. Ed. 2006, 45, 2965.

JP800756M