Two-Channel Decomposition of Methanol on Pt Nanoclusters

Feb 21, 2013 - (54) Huberty, J. S.; Madix, R. J. An FTIR Study of the Bonding of. Methoxy on Ni(100): Effects of Coadsorbed Sulfur, Carbon Monoxide...
0 downloads 0 Views 3MB Size
Article pubs.acs.org/JPCC

Two-Channel Decomposition of Methanol on Pt Nanoclusters Supported on a Thin Film of Al2O3/NiAl(100) Chen-Sheng Chao, Yu-Da Li, Bo-Wei Hsu, Won-Ru Lin, Hsin-Chung Hsu, Ting-Chieh Hung, Chao-Chuan Wang, and Meng-Fan Luo* Department of Physics and Center for Nano Science and Technology, National Central University, 300 Jhongda Road, Jhongli 32001, Taiwan S Supporting Information *

ABSTRACT: The decomposition of methanol on Pt nanoclusters grown from vapor deposition onto an ordered Al2O3/ NiAl(100) thin film was investigated under ultrahigh vacuum conditions with various surface probe techniques. The Pt clusters had mean diameter near 2.3 nm and height 0.4 nm before their coalescence; consisting of phase fcc, the clusters grew with their facets either (111) or (001) parallel to the θAl2O3(100) surface, depending on the temperature of growth. More than half the adsorbed monolayer of methanol on the Pt clusters decomposed via two channels: dehydrogenation to CO and C−O bond scission. The dehydrogenation was dominant and induced first at low-coordinated Pt sites, at 150 K on Pt(001) clusters and 200 K on Pt(111) clusters, whereas both lowcoordinated and some terrace Pt sites exhibited reactivity, despite the cluster size. On average, one CO was produced per surface Pt site, for a monolayer of methanol on either Pt(111) or Pt(001) clusters. In the other reaction, scission of the C−O bond occurred primarily in methanol itself and began about 250 K; the intermediate methyl preferentially formed methane on combining with atomic H from dehydrogenated methanol. No preferential reaction site for the C−O bond scission is indicated, but this process showed a remarkable dependence on the size and lattice parameter of the clusters: the probability of C−O bond scission decreased when the size increased and the lattice parameter decreased.



INTRODUCTION The catalytic decomposition of methanol (CH3OH) on platinum (Pt) is extensively investigated because the principal reaction is applied in direct methanol fuel cells (DMFC), which offer a prospect of direct conversion of methanol to electricity with great efficiency.1−5 The decomposition of methanol as a source of hydrogen also attracts intensive research. As the performance of a DMFC or the production of hydrogen is governed largely by the catalyzed reaction, knowledge of the detailed reaction kinetics and the correlation between reactivity and structure of the Pt catalyst is desirable. Many model systems such as Pt single crystals have been investigated, in either liquid or vapor phase, to shed light on the reaction mechanism and to correlate reactivity and structure; abundant valuable results are reported.6−20 Important information on the mechanism might be lacking because a gap exists between real Pt catalysts and single-crystal model systems. To bridge the gap and to gain insight, we conducted our experiments on a realistic model system, oxide-supported Pt nanoclusters. With scanning tunneling microscopy (STM), infrared reflection absorption spectroscopy (IRAS), temperature programmed desorption (TPD), reflection high-energy electron diffraction (RHEED), and synchrotron-based photoelectron spectroscopy (PES), we studied the decomposition of methanol on Pt nanoclusters formed on vapor deposition onto an ordered Al2O3/NiAl(100) thin film under ultrahigh vacuum © 2013 American Chemical Society

(UHV) conditions. The morphology and structure of the Pt clusters were characterized with STM, RHEED, and IRAS, and the catalyzed reactions with IRAS, TPD, and PES. The Pt clusters grew in the (111) orientation (denoted as Pt(111) clusters) at 300 K and in the (001) orientation at 430 K. Dehydrogenation of methanol to CO represented the dominant reaction, as observed earlier on Pt single crystals,6,10,12,15,20 whereas the oxide-supported Pt clusters exhibited superior reactivity. The dehydrogenation initiated about 150−200 K and at the Pt sites of small coordination number, sharing 30−65% of the total surface sites of the clusters at various coverages; eventually more than half the adsorbed monolayer of methanol underwent decomposition; on average one surface Pt yielded one CO, independent of the proportion of the low-coordinated sites. The reactivity was associated with the reactive sites, including both lowcoordinated and some terrace sites, but depended little on the cluster size. Earlier work on Pt single crystals showed that the terraces of the Pt(111) surface are inert;11,16,17 even on the more reactive Pt(001) surface, only a small fraction of adsorbed monolayer methanol reacts.15 Received: August 13, 2012 Revised: December 25, 2012 Published: February 21, 2013 5667

dx.doi.org/10.1021/jp3080282 | J. Phys. Chem. C 2013, 117, 5667−5677

The Journal of Physical Chemistry C

Article

In addition to the dehydrogenation, the scission of a C−O bond proceeded at a higher temperature, about 250 K, leading to adsorbed methyl (CH3) intermediate. Possible dehydrogenation intermediates formaldehyde and formyl are excluded as precursors to the C−O bond rupture. Instead of dehydrogenating, methyl preferred to combine with atomic hydrogen from dehydrogenated methanol, to form methane and then to desorb. This reaction path was reported earlier for a (1 × 1) Pt(110) surface,12,13 whereas the present clusters lacked facet (110); the temperature of onset of the C−O bond scission on the (1 × 1) Pt(110) surface was also significantly lower.12,13 We argue here that the reactivity toward scission of the C−O bond results largely from the nanoscale size of the cluster and the lattice parameter expanded by the Pt−oxide interaction, as the reactivity decreases when the size increases and the lattice parameter decreases with the coverage. Methanol on a similar model system, Pd clusters/Al2O3/NiAl(110), was shown to decompose also via these two channels,21−25 but neither the corresponding reaction sites nor the precursors to scission of the C−O bond resemble the present ones.

TPD spectra were recorded by ramping the sample at 3 K/s and monitoring the various masses on a quadruple mass spectrometer (Hiden), which was shielded and placed near (about 2 mm) the sample. IRAS spectra were recorded with a Fourier-transform infrared spectrometer (FTLA 2000, HgCdTe detector cooled with liquid nitrogen) with external optics aligned for incident angle 75° from the sample normal. The IRAS spectra are presented as a ratio of data of the sample and the oxide surface (or Pt clusters) measured at the same surface temperature (100 K), and are typically the average of 256 scans at resolution 4 cm−1. The PES experiments were performed at beamline U5-spectroscopy at National Synchrotron Radiation Research Center in Taiwan.33 The total energy resolution, including the beamline and energy analyzer, was estimated to be near 0.1 eV; the photon energies were fixed at 383 eV. The beam was incident normal to the surface; photoelectrons were collected at angle 58° from the surface normal. All photoelectron spectra presented here were first normalized to the photon flux. The binding energy (BE) is referred to the substrate bulk Al 2p core-level at 72.9 eV.34−36

EXPERIMENTAL SECTION Our experiments were performed in UHV chambers with a base pressure 4 × 10−10 Torr. A NiAl(100) sample (MaTeck GmbH) was polished to a roughness less than 30 nm and an orientation accuracy better than 0.1°. To obtain a clean surface, the sample was subjected to alternating cycles of sputtering and subsequent annealing before each experiment. The cleanliness of the sample was monitored with Auger electron spectra, lowenergy electron diffraction, and STM. An ultrathin θ-Al2O3 film was formed on oxidation of a NiAl(100) alloy surface at 1000 K; the formation of Al2O3 thin films is described elsewhere.26−29 To achieve a homogeneous crystalline Al2O3 surface with no NiAl facets,29,30 we refrained from protracted annealing after oxidation of the oxide films. The content of amorphous oxide surface29 was negligible. The grown θ-Al2O3 thin film had a thickness of 0.5−1.0 nm.26,29 The sample was then quenched to 300 K (unless otherwise specified) for vapor deposition of Pt from an ultrapure Pt rod heated by electron bombardment in a commercial evaporator (Omicron EFM 3). The rate of deposition of Pt was fixed about 0.1 ML/min, calculated according to the coverage prepared at 300 K.31,32 The coverage was estimated from the volume of the Pt clusters observed with STM; 1 ML corresponds to density 1.5 × 1015 atoms/cm2 of fcc Pt(111) surface atoms. After the deposition, the sample was cooled to the desired adsorption temperature (100 K, unless otherwise specified). Methanol gas was dosed with a doser pointing to the sample, with a background pressure 2−5 × 10−9 Torr. The methanol and methanol-d4 (Merck, purity 99.8%) was additionally purified with repeated freeze−pump−thaw cycles. We report methanol exposures in Langmuir units: 1 L = 10−6 Torr s. Atomic hydrogen (H) or oxygen (O) were prepared by cracking molecular hydrogen (H2) or oxygen (O2) over a hot filament. The position of the filament was approximately 60 mm from the sample. We use a relative measure of H (or O) exposure that derives from the H2 (or O2) pressure and the duration of operation of the filament. STM images (recorded with a RHK UHV 300 unit), constant-current topographies, were obtained at 90 K with a sample bias voltage typically 2.8−3.2 V and a tunneling current 0.1−0.2 nA. The STM tip consisted of an electrochemically etched tungsten wire. For RHEED, an electron beam of energy 38 keV was incident at a grazing angle 2−3° to the surface.

RESULTS AND DISCUSSION Our Pt clusters were grown by deposition from a vapor onto a thin film of Al2O3/ NiAl(100) at either 300 or 430 K. The clusters formed at 300 K have a mean diameter near 2.3 nm and height near 0.4 nm. With an increased Pt coverage, the mean diameter and height alter little, but the cluster density increases before cluster coalescence occurs (above 1.2 ML). The growth features result largely from the limited diffusion length and the small stable nucleus for deposited Pt atoms, which result from the strong Pt−Pt and Pt−oxide interactions.31,37 The strong Pt−oxide interaction reflects also on the charge transfer from the Pt clusters to the oxide and thus partial oxidation of Pt.32 These Pt clusters are structurally ordered, indicated by the RHEED patterns, have a fcc phase, and grow in preferred orientations of two kinds: major orientation (111) (with facets (111) parallel to the θ-Al2O3(100) surface) and minor orientation (001). The clusters formed at 430 K have a similar size37 but grow only in orientation (001). The Pt(001) facets have a square atomic mesh and match structurally better with the quasi-square oxygen mesh of the θ-Al2O3(100) surface.28,31,38 STM images and RHEED patterns for the Pt clusters are found in the Supporting Information and also earlier work.32,37,39 The surface sites of the Pt clusters were characterized with probe molecule CO and IRAS. Figure 1a shows IRAS spectra in the region of the C−O stretching mode (νCO) for CO adsorbed on Pt clusters (1.1 ML) on Al2O3/NiAl(100) at 300 K and annealed to selected temperatures (for 20 s) as indicated. As the diffusion of CO on Pt clusters at 100 K is restricted until the sample temperature exceeds 250 K, evident from the altered CO lines in the IRAS spectra, CO adsorbed at 300 K is expected to overcome readily the diffusion barrier and to occupy the energetically most stable sites. On adsorbing 0.2 L CO, an absorption line centered about wavenumber 2036 cm−1 appeared; with increasing exposure, the intensity increased and the wavenumber increased to 2054 cm−1 at CO 7.5 L, at which saturation occurred. Increasing the sample temperature to 350 K, the spectrum altered little. The line shape at this stage of CO saturation became asymmetric, comprising at least two distinct absorption features: a broad line near 2030 cm−1 and a narrow line near 2050 cm−1, as illustrated in Figure 1b. CO is known to adsorb preferentially on top of Pt atoms, i.e., atop





5668

dx.doi.org/10.1021/jp3080282 | J. Phys. Chem. C 2013, 117, 5667−5677

The Journal of Physical Chemistry C

Article

associated with a wavenumber greater than that for CO on a step. On annealing a sample above 350 K, both the absorption and the wavenumber decreased because CO desorbed; on annealing to 450 K, the line became more symmetric and the wavenumber shifted to 2041 cm−1, as most remaining CO was at low-coordinated sites and with greater adsorption energies, confirming our assignment above. Figure 1c shows the evolution of the CO IRAS with increased Pt coverage. At a coverage of 0.2 ML, the signals at smaller wavenumbers, for CO on low-coordinated sites, predominated (the top line); the result is confirmed by the desorption of CO centered at temperature 500 K (Figure 1d), assigned previously to CO from step sites of stepped Pt(111).42,43,45,46 With increased Pt coverage, the terrace sites were occupied more than the low-coordinated sites, indicated by enhanced signals near 2050 cm−1; the wavenumber increased and eventually the line shape became asymmetric at 1.1 ML (the third line). The CO desorption feature also shifted consistently toward a lower temperature (Figure 1d), as more CO desorbed from the terrace sites.42,43,45,46 The temperature of onset of CO desorption from Pt clusters at a coverage of 1.1−7.3 ML is near that from stepped Pt(111)45 and Pt clusters/SiO2;46 this result indicates that the impact of partial oxidation of Pt32 on the CO−Pt bond and thus on the absorption wavenumber is minor. The significant shift of the wavenumber of infrared absorption from 2054 to 2083 cm−1 with Pt coverage increased from 1.1 to 7.3 ML is likely due to an enhanced long-range CO dipole−dipole interaction, rather than to varied surface structures. As the corresponding CO desorption lines are similar, as shown in Figure 1d, the composition of terrace and low-coordinated sites on the clusters varied insignificantly at coverage 1.1 − 7.3 ML, except that intensity increased with the coverage. At coverage 7.3 ML, the absorption at wavenumbers 2083 and 2063 cm−1 becomes comparable to that observed from stepped Pt single-crystal surfaces.42−44 The CO probe experiments on clusters formed at 430 K yield similar results: the CO IRAS spectra exhibit features for CO on both terrace and low-coordinated Pt sites, exemplified in Figure 1e. The wavenumbers (νCO) also increased with increased coverage, as observed from the clusters formed at 300 K. In Figure 1e, the wavenumber 2070 cm−1 for CO on the terrace is almost the same as that for CO on the single-crystal counterpart.15 Comparison of TPD and IRAS spectra from CO on clusters showed that the formation of Pt carbonyls (Pt(CO)n)47,48 on the low-coordinated sites, such as for cobalt carbonyls formed on low-coordinated cobalt,49,50 is negligible. Figure 2a presents a plot of integrated intensity of the CO IRAS spectra as a function of temperature and figure 2b shows the TPD spectrum for CO on 1.1 ML Pt clusters. On annealing the sample to 450 K (for 20 s), the IRAS intensity decreased to about one-third of the original intensity and the wavenumber of the absorption line shifted to about 2040 cm−1 (Figure 1a), indicating that the remaining CO is on the low-coordinated site. The fraction of remaining CO, as evaluated by integration of the TPD spectrum from 450 to 600 K, is about 45% of the total desorbing CO, but integration of the TPD spectrum neglects the desorption of CO during the annealing and cooling periods of measurements of the IRAS. This additional desorption corresponds to 10−15% of all desorbing CO, which was estimated by comparing the integrated TPD spectrum for the remaining CO to that of the shadowed area in Figure 2b. The remaining CO amounted to only 30−35%, similar to that

Figure 1. CO IRAS spectra from (a) 1.1 ML Pt/Al2O3/NiAl(100) exposed to CO at 300 K and subsequently annealed to varied temperatures as indicated, (b) CO on 1.1 ML Pt/Al2O3/NiAl(100), (c) CO on Pt clusters at varied coverage on Al2O3/NiAl(100). Panel d shows CO TPD spectra for CO on Pt clusters at varied coverage on Al2O3/NiAl(100) and (e) CO IRAS spectrum for CO on 7.3 ML Pt deposited on Al2O3/NiAl(100) at 430 K. For panels a−d the Pt clusters were formed at 300 K, and for panels b−e the surfaces were saturated with 7.5 L CO at 100 K and annealed to 300 K. The green curves in panels b and e represent signals from CO on lowcoordinated Pt sites and the blue one centered about 2052 cm−1 denotes signals for CO on terrace Pt sites; the red curve indicates the sum of the two fitted curves.

sites;6,12,15,40−43 lack of absorption between 1800 and 1900 cm−1 or at an even smaller wavenumber gave no indication of CO on the bridge and hollow sites.41−43 We assign the absorption line near 2030 cm−1 to CO adsorbed on top of lowcoordinated Pt, such as at a step or corner of a cluster, and that near 2050 cm−1 to CO on top of terrace Pt. As the lowcoordinated sites are energetically favored,44,45 they are occupied first when CO is adsorbed (top two spectra in Figure 1a); the terrace sites are filled when further CO is adsorbed. Earlier work on stepped Pt surfaces (with (111) terraces)42−44 and Pt clusters46 also indicated that CO on terraces is 5669

dx.doi.org/10.1021/jp3080282 | J. Phys. Chem. C 2013, 117, 5667−5677

The Journal of Physical Chemistry C

Article

Figure 3. TPD spectra for (a) CD3OD (m/z = 34 u), CO (m/z = 28 u), D2 (m/z = 4 u), and CD4 (m/z = 20 u) from 1.1 ML Pt clusters/ Al2O3/NiAl(100) exposed to 3.0 L CD3OD at 100 K; (b) CH4 (m/z = 16 u) from 1.0 ML Pt clusters/Al2O3/NiAl(100) exposed to 3.0 L CH3OH at 100 K; and (c) CD3OD (m/z = 34 u) from Pt at varied coverages (as indicated) on the Al2O3/NiAl(100) after exposure to 3.0 L CD3OD. The inset in panel b shows a CH4 (m/z = 16 u) TPD spectrum from 4.0 L CH4 adsorbed on 1.0 ML Pt clusters/Al2O3/ NiAl(100). The Pt clusters were grown at 300 K.

Figure 2. (a) Temperature-dependent integral intensity of the CO IRAS spectra from 1.1 ML Pt/Al2O3/NiAl(100) exposed to 6.0 L CO at 300 K and (b) the CO TPD spectrum for 1.1 ML Pt/Al2O3/ NiAl(100) saturated with 6.0 L CO at 300 K. 6.0 L CO is sufficient to saturate the surface.

cannot be distinguished from the background in the present work. Except these methanol cracking patterns, no signals for the desorption of CO, D2 or CD4 were observed for methanold4 adsorbed on Al2O3/NiAl(100).51 The CO and D2 signals indicated dehydrogenation of methanol. The CO produced from dehydrogenated methanol (denoted as COm) desorbed above 350 K, and desorption continued to about 600 K. This COm desorption occurred in a temperature range similar to those from Pt single crystals and Pt clusters supported on other oxides;6,13,15,42,43,45,46 it differs little from that of molecularly adsorbed CO (Figure 2b). As adsorbed methanol must decompose to COm far below 350 K, COm moved to the preferred sites and desorbed as molecularly adsorbed CO. D2, from dehydrogenated methanol-d4, desorbed above 270 K, and the desorption extended to above 600 K. This desorption feature resembles recombinative desorption of atomic hydrogen (H) on Pt clusters (not shown), verifying that dehydrogenation proceeds below 270 K. The IRAS spectra presented below revealed a more precise temperature for the dehydrogenation. The desorption of methane-d4 (CD4), according to the bottom spectrum of Figure 3a, began about 300 K and continued to about 600 K. As the adsorbed methanol yielded the same desorption of methane (CH4) between 300 and 600 K, shown in Figure 3b, the observed signals at m/z = 20 u result from CD4 rather than D2O. This observation indicates an alternative channel of decomposition, C−O bond scission: the C−O bond of methanol broke, methyl radical (CD3) formed, and the intermediate combined with atomic deuterium (D) from dehydrogenated methanol-d4 to desorb as CD4. The experiments of atomic hydrogen coadsorption presented below confirm that dehydrogenated intermediates formaldehyde and formyl were not involved. The methane molecularly adsorbed on Pt clusters desorbed at a much lower temperature regime (the inset of Figure 3b): the maximum of desorption occurred at about 200 K and extended to about 400 K, similar to the

estimated from the IRAS spectra; consistent with the CO IRAS spectrum, the remaining CO also exhibited desorption features for CO on the low-coordinated sites: the onset temperature shifted to about 430 K and the desorption peak to about 475 K.45,46 Similar results were obtained for varied Pt coverage: Pt carbonyls thus did not form on the low-coordinated Pt. This comparison also indicates that CO adsorbed on terrace and low-coordinated Pt gives comparable infrared absorption intensities, which allowed us to use the absorption intensity as a measure of the number of surface sites. The total CO IRAS signals are consistently proportional to the intensity of the corresponding CO TPD spectra (demonstrated in our Supporting Information). The variation of CO IRAS spectra with Pt coverage (Figure 1c) thus indicates that the proportion of low-coordinated sites decreased with coverage from about 65% (at 0.2 ML Pt) to 30% of total sites (at 1.1 ML). The decomposition of methanol on these oxide-supported Pt clusters was studied with TPD, IRAS, and PES. Figure 3a shows TPD spectra of methanol-d4 (CD3OD), CO, D2, and CD4 for Pt clusters (1.1 ML) on Al2O3/NiAl(100) exposed to CD3OD (3.0 L). Adsorbed methanol and methanol-d4 show the same desorption behavior but the latter gives D2 signals more clearly than H2. The large desorption feature of CD3OD between 110 and 150 K is attributed to the desorption of multilayer methanol-d4, resembling what was observed on varied surfaces,6,13,15,51 and the other with maximum about 190 K and extending to 270 K is due primarily to the desorption of monolayer methanol-d4. The CO and D2 spectra show cracking patterns in the corresponding temperature range. The CD4 signals (m/z = 20 u) in this temperature range resulted largely from the background, as these signals also appeared for methanol-d4 adsorbed on the oxide. According to earlier experiments on (1 × 1) Pt(110),12,13 D2O (m/z = 20 u) from decomposed CD3OD might desorb between 150 and 200 K but 5670

dx.doi.org/10.1021/jp3080282 | J. Phys. Chem. C 2013, 117, 5667−5677

The Journal of Physical Chemistry C

Article

desorption of methane on Pt single crystals.52 The methane coadsorbed with CO exhibits a similar desorption behavior. The apparent difference implies that the rate-limiting step for the desorption of methane (from decomposed methanol) is either the combination of methyl radicals with atomic deuterium or the C−O bond scission. This result is striking, as scission of the C−O bond in the decomposition of methanol on Pt was indicated only on a (1 × 1) Pt (110) surface,12,13 whereas our Pt clusters lack Pt(110) facets. Detailed discussion of this issue continues below. Figure 3c shows stacked CD3OD TPD spectra from CD3OD (3.0 L) on Al2O3/NiAl(100) and Pt clusters at varied coverages. As CD3OD (1.0 L) nearly saturated the surface, all spectra exhibit the desorption of multilayer CD3OD. The intensity of monolayer CD3OD desorption (centered about 190 K) decreased with increased Pt coverage, because there were more Pt clusters on the surface, more methanol reacted and thus less methanol desorbed. At 2.4 ML Pt clusters, the surface was almost covered with Pt clusters and the monolayer CD3OD adsorbed only on these clusters (bottom of Figure 3c). The intensity of monolayer CD3OD desorption from 2.4 ML Pt clusters amounted to about half that from the oxide (top Figure 3c). Methanol adsorbed on the oxide did not react but only desorbed; if the same quantity of monolayer methanol is assumed on the Pt clusters, the comparison indicates that about half the methanol adsorbed on Pt clusters decomposed and the other half desorbed. This estimated probability 0.5 for the reaction is probably a lower limit because more methanol is expected to adsorb on Pt clusters than on the oxide as the sample with Pt clusters has a greater surface area. The amount of desorbed COm, which directly measures the amount of dehydrogenated methanol, agrees with this estimate. For Pt coverage ≥2.4 ML, the reacting methanol was 1.4−2.1 times the desorbing methanol.53 As the dehydrogenation to CO predominated, this comparison implies that 60−70% of monolayer methanol on Pt clusters reacted. We recorded IRAS spectra in experiments to explore in detail the channel of methanol dehydrogenation. Figure 4a displays a series of IRAS spectra of C−O stretching absorption bands for methanol on Pt (1.1 ML) on Al2O3/NiAl(100) annealed to selected temperatures. The top line for methanol (3.6 L) adsorbed on the sample at 100 K is broad and asymmetric; it comprises a line at 1048 cm−1, due to multilayer methanol, and an extension to about 1020 cm−1, attributed to monolayer methanol. The feature resembles closely those for methanol on Au clusters and on thin film Al2O3/NiAl(100).51 With increased temperature, the intensity attenuated greatly and the absorption feature became invisible above 300 K, as adsorbed methanol either desorbed or decomposed. The signals at smaller stretching frequency (1020−1030 cm−1) at elevated temperatures are not readily assigned to methoxy or other intermediates, because they were observed for methanol on the oxide;51 also, the methoxy features were not identified in the C−H absorption region (νsCH3 and νasCH3).21,54−56 Figure 4b shows the corresponding IRAS spectra of the C−H absorption regime; the narrow absorption line at 2827 cm−1 is assigned to νsCH3 of CH3OH and the broad absorption centered about 2948 cm −1 to ν as CH 3 + 2δ as CH 3 of CH3OH.56,57 Consistent with the results of C−O stretching absorption and the above TPD experiments, the absorption intensity decreased remarkably upon desorption of multilayer methanol (annealing to 150 K). Relative to the C−O stretching signals, the C−H absorption signals are small; for monolayer

Figure 4. IRAS spectra of (a) C−O stretching and (b) C−H absorption regions for 3.6 L CH3OH on 1.1 ML Pt clusters on Al2O3/ NiAl(100) at 100 K and annealed to selected temperatures; (c) corresponding CO IRAS spectra and (d) plot of the integrated intensities of C−O stretching absorption for CO and methanol as a function of temperature. The Pt clusters were formed at 300 K. For panel d, the error bars indicate reproducibility, based on the results of varied Pt coverages (0.2, 0.4, 0.8, 1.1, and 1.8 ML); the CO intensities obtained from any Pt coverage are normalized to the maximal one (350 K) and the νCO intensities of methanol are normalized to the one at 150 K.

methanol (150 K), the signals are near the noise level and for submonolayer methanol (200 K), no absorption feature is distinguishable. Because of the limited sensitivity, the absorption features for methoxy, expected near 2810 cm−1 (for νsCH3) and 2920 cm−1 (for νasCH3 + 2δasCH3),21,56,57 were not identified. The dehydrogenation of adsorbed methanol to CO is evident through the CO IRAS spectra as shown in Figure 4c. No signals in the C−O stretching regime for COm were observed when the sample was exposed to methanol at 100 K and annealed to 150 K (the top and second lines in Figure 4c). On annealing to 200 K (for 20 s), C−O stretching absorption appeared near 2022 cm−1, indicating formation of COm. As adsorbed CO diffused readily to the preferred sites only above 250 K and as on 5671

dx.doi.org/10.1021/jp3080282 | J. Phys. Chem. C 2013, 117, 5667−5677

The Journal of Physical Chemistry C

Article

decreased to about 1975 cm−1 for coadsorbed methanol, which was observed also on Au nanoclusters.51 On annealing to 150 K, multilayer methanol desorbed, as the effect of methanol coadsorption became negligible, the wavenumber of CO signals increased above 2010 cm−1 and their intensity increased. The CO intensity at 150 K exceeded that from contaminating CO, indicating that methanol dehydrogenation had already occurred. On annealing to 200 K, the intensity continued to increase and the absorption line shifted further to 2025 cm−1. The observed wavenumber 2010 or 2025 cm−1 indicates again that the reaction at low temperature occurred on the lowcoordinated Pt sites. Elevating further the sample temperature varied the CO spectrum in a manner similar to that for methanol on Pt(111) clusters: the intensity of CO signals increased to a maximum at 350 K and attenuated above 350 K. The intensities of CO signals as a function of temperature, obtained from varied Pt coverages (1.1−7.3 ML), are plotted in Figure 5b. The result shows consistently greater production of COm than on Pt(111) clusters (Figure 4d) at temperatures 150−350 K. For instance, about 35% and 55% of the maximal COm quantity are given at 200 and 250 K on Pt(100) clusters whereas only about 20% and 44% on Pt(111) clusters. The numbers of the low-coordinated sites cannot account for the difference, as the difference remained even when the proportions of the low-coordinated sites on Pt(111) and Pt (001) clusters were comparable. A smaller activation energy is indicated for dehydrogenation of methanol to CO on Pt (001) clusters. The onset temperature for dehydrogenation of methanol to CO on Pt(111) clusters is near that on Pt(111) and ion-eroded Pt(111) single crystals,11,20 as in both cases the dehydrogenation is initiated on low-coordinated Pt; in contrast, the dehydrogenation temperature on the Pt (001) clusters is below that on a Pt(001) single crystal.15 We speculate that dehydrogenation is induced first on the low-coordinated sites for Pt(001) clusters but on the terrace sites for Pt(001) single crystals. In addition to the reaction temperatures (activation energies), the experiments reveal also the reactive sites and probability of dehydrogenation. Using the intensity of CO IRAS spectra from molecularly adsorbed CO as a measure of the number of surface sites, we plot the produced D2 and COm (from both TPD and IRAS measurements) as a function of the number of surface sites (Figure 6a−c). Both D2 and COm increase in proportion to the total surface sites, rather than the low-coordinated sites. This result excludes the cluster-size dependence for dehydrogenation of methanol, as the Pt coverage ranged from 0.2 to 7.3 ML, corresponding to morphological evolution from nanoclusters to films, and indicates that the low-coordinated sites are not the only reactive sites. The IRAS line shapes and intensities of COm support this implication. The line shapes were typically asymmetric, for instance, the COm lines at 350 K in Figures 4c and 5a, and the intensities amount to 65−80% of the CO saturation level (Figures 1b,d). Contrasting that the lowcoordinated sites share only 30−45% of the total surface sites in most cases, this observation indicates that COm occupies not only the low-coordinated sites but also terrace sites of some proportion. The unoccupied terrace sites can be filled on adsorbing further CO, which renders the CO spectrum almost the same as that at CO saturation (Figure 7). If COm does not diffuse away from the reaction sites, the terrace sites taken by COm are reactive and the remaining available terrace sites are inactive. This assumption is verified because by further

Pt(111) single crystals only the defects are reactive,11,16,17 the observed small wavenumber for COm implies that dehydrogenation at 200 K was initiated at the low-coordinated sites. The effect of dependence of CO coverage on wavenumber cannot account for the small wavenumber, because 0.2 L CO adsorbed on Pt clusters at 100 K gave absorption about 2050 cm−1. We exclude also the effect of coadsorbed methanol and atomic hydrogen because the coadsorbed methanol decreased much at 200 K and atomic hydrogen altered little the wavenumber of C−O stretching. As with increased temperature more COm formed, the absorption line became more intense and the wavenumber increased. At 350 K, the intensity attained a maximum and the wavenumber shifted to about 2037 cm−1; elevating further the temperature decreased the intensity as COm desorbed; the absorption line vanished above 500 K, agreeing with the CO TPD spectra (Figures 2b and 3a). These features were observed generally for Pt coverage varied from 0.2 to 1.8 ML. Figure 4d plots the variation of intensity of C−O stretching absorption for COm and methanol with temperature. The error bars indicate reproducibility, based on results of Pt coverage varied from 0.2 to 1.8 ML. A general trend is obvious that COm formed above 200 K, the COm quantity increased with temperature and the maximum appeared about 350 K; in contrast, adsorbed methanol decreased with the increased temperature. Dehydrogenation of methanol on Pt(001) clusters began at an even lower temperature. Figure 5a displays the CO IRAS spectra for methanol on Pt(001) clusters annealed to selected temperatures. The small CO signals in the top line for the Pt clusters came from contaminating CO; on adsorbing methanol (3.0 L), the CO signals attenuated and their wavenumber

Figure 5. (a) CO IRAS spectra for Pt (7.3 ML, deposited at 430 K) on Al2O3/NiAl(100) exposed to 3.6 L CH3OH and annealed to selected temperatures; (b) plot of the integrated intensity of CO IRAS spectrum as a function of temperature, obtained from methanol on varied Pt coverages (from 1.1, 3.5, and 7.3 ML) deposited at 430 K. For panel b, the error bars indicate reproducibility; the CO intensity obtained from each Pt coverage is normalized to the maximal one (350 K). 5672

dx.doi.org/10.1021/jp3080282 | J. Phys. Chem. C 2013, 117, 5667−5677

The Journal of Physical Chemistry C

Article

Figure 6. Plots for (a) CO and (b) D2 integrated intensities of TPD spectra and (c) CO integral intensities of CO IRAS spectra (from 3.6 L CD3OD adsorbed on Pt clusters at varied coverages 0.3, 0.4, 0.5, 1.1, 2.4, 3.5, and 7.3 ML for TPD and 0.2, 0.4, 0.8, 1.1, 1.8, 3.5, and 7.3 ML for IRAS measurements) as a function of surface Pt sites. The intensity of CO IRAS spectrum from molecularly adsorbed CO serves as a measure of the number of surface sites. The black notation indicates Pt clusters formed at 300 K and the blue notation at 430 K.

Figure 8. CO TPD spectra for (a) 3.6 L CH3OH adsorbed on 1.1 ML Pt/Al2O3/NiAl(100) at 100 K (black line) and 6.0 L CO adsorbed on 1.1 ML Pt/Al2O3/NiAl(100) at 300 K (gray line) and (b) 3.0 L CO adsorbed on H-precovered (5.0 L, black) and pristine (gray) 1.0 ML Pt clusters on Al2O3/NiAl(100) at 100 K; the inset of panel b shows the corresponding CO IRAS spectra for the samples before thermal desorption. The Pt clusters were grown at 300 K.

constantly observed at varied Pt coverage, and also on both Pt(111) and Pt(001) clusters. The disparity arises because the “invisible” COm adsorbed with the C−O axis greatly tilted, producing small infrared absorption signals. Such adsorption configuration is allowed possibly because the coadsorbed H from dehydrogenated methanol modified the adsorption potential. The CO TPD experiments for CO adsorbed on Hprecovered Pt clusters verify this speculation (Figure 8b). The CO desorption intensity (black) from CO on H-precovered Pt clusters is about 70% greater than that on pristine Pt clusters (gray), despite the fact that the intensities of their CO IRAS spectra differ by about 20% (the inset of Figure 8b). An alternative possibility is that molecularly adsorbed CO dissociated more readily into atomic carbon and oxygen than COm, so the desorbing CO from molecularly adsorbed CO decreased; the coadsorbed H hence decreased the probability of dissociation of COm on Pt clusters. This mechanism is unfavorable because CO prefers to dissociate on the lowcoordinated Pt sites and above 450 K,45 but the desorbed CO signals above 450 K (Figure 8b) were little enhanced by coadsorbed H. The comparable CO TPD spectra from CO m and molecularly adsorbed CO indicate that each surface Pt yields on average one COm for monolayer methanol. This large productivity agrees with the above TPD experiments in that monolayer methanol more than 50% reacts. In comparison with previous results on single-crystal counterparts, the supported Pt clusters are much more productive. For instance, the terraces of Pt(111) clusters are active whereas those of Pt(111) single crystals are inactive;11,16,17on the hex Pt(001) single crystal surface, only 0.085 monolayer of methanol underwent dehydrogenation; on the more reactive (1 × 1) Pt(001) surface, the reaction is enhanced by a maximal factor 3.4,15 still not comparable with the present case. The enhanced

Figure 7. CO IRAS spectra for 3.0 L CH3OH adsorbed on 1.1 ML Pt/ Al2O3/NiAl(100) at 100 K and annealed to 350 K (the top) and consequently exposed to 6.0 L CO at 100 K and annealed to 350 K (the bottom). The Pt clusters formed at 300 K.

exposure to methanol at 100 K and by annealing to 350 K, the CO desorption signals for the COm-covered Pt clusters (such as the COm line at 350 K in Figure 4c) increased significantly but the CO IRAS altered little. The unoccupied terrace sites remained free of CO, confirming that these sites are inactive and indicating that newly produced COm did not diffuse to the inactive terrace sites. The CO TPD spectra reflect the quantity of COm more precisely than the corresponding CO IRAS spectra. The intensities of CO IRAS signals from decomposed methanol and molecularly adsorbed CO (at saturation level) differ by 20− 35%, but the intensities of their CO thermal desorption spectra are nearly the same, exemplified by Figure 8a. Such disparity is 5673

dx.doi.org/10.1021/jp3080282 | J. Phys. Chem. C 2013, 117, 5667−5677

The Journal of Physical Chemistry C

Article

methanol C 1s was observed on oxide-supported Au nanoclusters.36 As CO dissociated into elemental carbon on Pt clusters above 350 K or about 500 K indicated earlier,45 the CHx intermediate was produced at this low temperature. Increasing the temperature to 300 K enhanced such alteration: the energy of C 1s of methanol shifted further to BE 288.2 eV, and the energy of the other C 1s signals shifted to 285.0 eV because of the increased proportion of CHx contributing signals at 284.8−284.0 eV. On annealing to 350 K, no methanol C 1s signals remained, whereas the line of C 1s near 285.0 eV grew, broadened and shifted to smaller energy, attributed to more CO and CHx. We adsorbed atomic hydrogen and oxygen isotope (18O) on samples to probe the species formed from the C−O bond scission channel. Atomic hydrogen is expected to combine with the hydrocarbon from decomposed methanol and to desorb as methane, and the atomic oxygen could combine with elemental carbon and desorb as CO. Figure 9b shows the TPD results for CD3OD on Pt clusters (1.0 ML) annealed to 270 K and then exposed to atomic hydrogen. We observed desorption signals of CD4 (m/z = 20 u) and CD3H (m/z = 19 u) and a cracking pattern of CD3H (m/z = 17 u). No desorption of CDH3, CD2H2 and CH4 was indicated because the TPD signals m/z = 15 and 16 u, corresponding to CH4 and the cracking patterns of CH4, CD2H2, and CDH3, were absent. As the ratio of intensities of signals at m/z = 19 u (CD3H) to signals m/z = 17 u is about unity, almost equivalent to the ratio of measured methane to its cracking pattern, the signals at m/z = 17 u are predominantly contributed by the cracking pattern of CD3H, rather than the desorption of CD2H2. Accordingly, only CD3 was yielded, and CD3 was not further dehydrogenated to other hydrocarbon species but combined with atomic hydrogen to form methane and to desorb. This result indicates also that the C−O bond breaks via primarily methanol and perhaps methoxy to a small extent, but not via formaldehyde and formyl. Figure 9c shows the TPD result for CD3OD (3.0 L) on Pt clusters (1.0 ML) annealed to 350 K and then exposed to atomic oxygen isotope (18O). Significant desorption of CO (m/z = 28 u) occurred between 350 and 600 K as in other conditions, but only a small amount of C18O (m/z = 30 u) desorbed in the same temperature regime, indicating a small amount of elemental carbon that combined with atomic oxygen isotope and desorbed. As the quantity of elemental carbon was small and as no other intermediate hydrocarbon species existed, the probability of CD3 being dehydrogenated to elemental carbon was small; these results also reflect that methane formed by combination of atomic carbon and hydrogen to only a slight extent. The origin of the C−O bond scission is intriguing as it was previously observed only on (1 × 1) Pt(110)12,13 and as our Pt clusters lack Pt(110) facets. The origin differs because on a (1 × 1) Pt(110) surface the C−O bond breaks above 140 K12,13 but on the present Pt clusters above 250 K. We blocked the Pt atop sites by adsorbing CO to evaluate the effect on the C−O bond breaking. For CO preadsorbed Pt clusters, the CH4 desorption between 300 and 370 K vanished (the lower panel in Figure 10a), whereas that above 370 K remained and altered little. The atop CO cannot entirely inhibit scission of the C−O bond but decreases the probability; the unblocked sites, the bridge and hollow sites, are likely reactive. The onset temperature of CH4 desorption increased to 370 K, as the formation of methane on atop Pt was hindered. As molecularly adsorbed methane desorbed mostly about 200 K (inset of

productivity of the Pt clusters arose from the increased number of active sites and also possibly from the decreased size of the adsorption-site ensemble15 for the dehydrogenation. The partially oxidized Pt was not responsible for the large productivity because the partially oxidized Pt was around the oxide-Pt interface but the productivity was sustained at large coverages, such as 7.3 ML Pt. For the alternative channel of methanol decomposition, the photoelectron experiments showed that the scission of the C− O bond began about 250 K. Figure 9a shows the PES spectra

Figure 9. (a) C 1s photoelectron spectra from 1.0 L CH3OH adsorbed on 0.9 ML Pt clusters/Al2O3/NiAl(100) at 120 K and annealed to selected temperatures; (b) TPD spectra for CD4 (m/z = 20 u), CD3H (m/z = 19, 17 u), and CH4 (m/z = 16 u) from 3.0 L CD3OD on 1.0 ML Pt clusters/Al2O3/NiAl(100) annealed to 270 K and exposed to 3.0 L atomic hydrogen; and (c) CO (m/z = 28, 30 u) TPD spectra for 3.0 L CD3OD on 1.0 ML Pt clusters/Al2O3/NiAl(100) annealed to 350 K and exposed to isotopic atomic oxygen (18O, 3.0 L). The Pt clusters were formed at 300 K.

from methanol (1.0 L) adsorbed on Pt clusters (0.9 ML) and annealed to selected temperatures. The feature initially centered about BE 287.7 eV is assigned to C 1s of methanol on Pt clusters,20,32 and that about 285.2 eV to C 1s of CO on Pt clusters, as molecularly adsorbed CO gives a similar BE. The C 1s line of CO became evident above 200 K and continued to increase with increasing temperature. On annealing to 250 K, the energy of methanol C 1s increased to 288.0 eV and the line for CO broadened. This shift of methanol C 1s was not due to the formation of formaldehyde indicated earlier on oxide films,58 as most formaldehyde desorbed already from Pt clusters at 250 K.59,60 The shift resulted from the coexistence of methanol with new species from decomposed methanol, elemental carbon or CHx, indicated by the broadening of the C 1s near 285.2 eV. A similar BE shift of the energy of 5674

dx.doi.org/10.1021/jp3080282 | J. Phys. Chem. C 2013, 117, 5667−5677

The Journal of Physical Chemistry C

Article

reactive sites. In addition, the methane desorption structure described above (Figure 10a) remained similar even for Pt coverages at which the probability decreased. The decreased probability of C−O bond scission is associated with the electronic structures altered by the altered size and lattice parameter of the Pt clusters. Figure 10c plots the CD4 production per site (corresponding to the probability of C− O bond scission) and the lattice parameter as a function of Pt coverage. The saturation or the decreased probability occurs near 1.2−2.0 ML Pt, at which the cluster size increases evidently because of cluster coalescence; above 3.0 or 4.0 ML, the clusters mostly merge into films (according to the STM observation), so the continuously decreased probability must be associated with the electronic structures affected by the increased film thickness, accompanied by a decreased lattice parameter (Figure 10c). As the lattice of the clusters expanded relative to the bulk value (3.92 Å) because of the strain from the Pt−oxide interface, due to the lattice mismatch, the lattice parameter decreased with increased cluster size or film thickness. The reactivity toward scission of the C−O bond hence originated from the particular electronic structure resulting from the nanoscale size of the cluster and the expanded lattice; when the size increased and the lattice constant decreased to a value near the bulk value, the reactivity decreased. This argument concurs with the fact that the singlecrystal Pt(111) with defects or steps, or stepped Pt(332), remained inert toward the C−O bond scission of methanol.11,16,17,20,61 These reactive Pt clusters exhibited the Pt 5d states different from those of Pt single crystals. Previous work showed that the BE of Pt 5d-derived states have two well separated features,32 in contrasting with the Pt 5d states of the Pt single crystals, for which the two features merge into a single peak.62 The features of the Pt 5d-derived states evolved with coverage and eventually merge, as seen for the Pt single crystals. Methanol decomposition on a similar model system, Pd clusters on Al2O3/NiAl(110), was extensively studied.21−25 The Pd clusters in that work were evidently larger than the present Pt clusters (average diameter 5.5 nm and height about 1.0 nm) and exhibit a morphology of a well shaped crystallite: they grew in the (111) orientation, and predominantly exposed (111) facets with a small fraction of (100) facets.21−23,63 As the Pd clusters contained about 3000 Pd atoms each, the electronic and catalytic properties of facets either (111) or (100) resembled their single-crystal counterparts.25 Dehydrogenation to CO represented the dominant reaction channel and scission of the C−O bond was slow on the Pd clusters,21,22 like methanol decomposition on our Pt clusters, but the reaction sites differ. On the Pd clusters, dehydrogenation proceeds mainly on (111) terraces,25 whereas on our Pt clusters, dehydrogenation occurs preferentially on the low-coordinated sites; breaking of the C−O bond is induced by the edges and corners of the Pd clusters,23 the low-coordinated Pd, but no preferable sites are indicated on the Pt clusters, and the reactivity is associated with the cluster size and lattice parameter. These differences indicate intrinsically disparate catalytic properties of Pt and Pd toward decomposition of methanol. In addition, the breaking of the C−O bond on the Pd clusters results in formation of hydrocarbon species (CHx), likely from precursors methoxy, formaldehyde and formyl, and eventually atomic carbon after complete dehydrogenation;24,25 in contrast, the breaking of the C−O bond on our Pt clusters yields mainly methyl (CH3), implying no precursors formaldehyde and formyl; the methyl intermediates prefer to

Figure 10. (a) CH4 TPD spectra for 3.6 L of CH3OH adsorbed on pristine (upper panel) and CO preadsorbed (lower panel) 1.0 ML Pt clusters; (b) plot of intensity of CD4 desorption (from 3.6 L CD3OD adsorbed on Pt clusters at varied coverages 0.3, 0.4, 1.1, 2.4, 3.5, and 7.3 ML) as a function of the number of surface sites; (c) probability of scission of the C−O bond (CD4 production per site, solid squares) and lattice parameter of clusters (solid circles) as a function of Pt coverage. The number of surface sites in panel b was measured by the intensities of CO IRAS spectra from molecularly adsorbed CO; the dashed line and dotted line in panel c were drawn to guide the eye for variation of the reaction probability and lattice parameter respectively. The black notation indicates the Pt clusters formed at 300 K and the blue notation at 430 K.

Figure 3b) and no methanol remained above 350 K (either with or without CO preadsorbed), the formation of methane is the rate-limiting step for methane desorption above 370 K. The increased desorption temperature reflects the activation energy for the formation of methane (combination of CH3 with H) on the bridge or hollow sites of Pt clusters. Accordingly, the methane desorption in the temperature range 300−370 K, the vanishing part due to the CO blocking, is associated with scission of the C−O bond and the formation of methane on the atop sites of Pt clusters. In Figure 10b we plotted the intensity of CD4 thermal desorption as a function of the surface sites (measured by the intensities of CO IRAS spectra from molecularly adsorbed CO). The CD4 produced at the same surface number is slightly greater on Pt(001) than on Pt(111) clusters, but the trend is the same: the CD4 in the beginning increased proportionally with the surface sites but saturated above a particular number of surface sites. This trend differs much from those for production of CO and D2 (Figure 6a−c). As the increased surface sites correspond to increased adsorbed methanol on the Pt clusters, the saturation at greater coverages indicates a decreased probability for scission of the C−O bond. The decreased probability is not due to an altered number of some particular sites. The quantity of CD4 varies commensurately with neither the low-coordinated nor the terrace sites; the C−O bond still breaks at 7.3 ML Pt clusters, for which nearly no Pt−oxide interface sites exist: the Pt−oxide interface sites are not the only 5675

dx.doi.org/10.1021/jp3080282 | J. Phys. Chem. C 2013, 117, 5667−5677

The Journal of Physical Chemistry C

Article

(2) Williams, K. R.; Burstein, G. T. Low Temperature Fuel Cells: Interactions between Catalysts and Engineering Design. Catal. Today 1997, 38, 401−410. (3) Burstein, G. T.; Barnett, C. J.; Kucernak, A. R.; Williams, K. R. Aspects of the Anodic Oxidation of Methanol. Catal. Today 1997, 38, 425−437. (4) Beden, B.; Lamy, C.; Leger, J. R. Electrocatalytic Oxidation of Oxygenated Aliphatic Organic Compounds at Noble Metal Electrodes; Bockris, J. O. M., Conway, B. E., White, R. E., Eds.; Plenum Publishers: New York, 1992; Vol. 22, p 566. (5) Rostrup-Nielsen, J. R.; Nielsen, R. Fuels and Energy for the Future: The Role of Catalysis. Catal. Rev., Sci. Eng. 2004, 46, 247−270. (6) Sexton, B. A. Methanol Decomposition on Platinum (111). Surf. Sci. 1981, 102, 271−281. (7) Sexton, B. A.; Rendulic, K. D.; Hughes, A. E. Decomposition Pathways of C1-C4 Alcohols Adsorbed on Platinum (111). Surf. Sci. 1982, 121, 181−198. (8) Sexton, B, A.; Hughes, A. E. A Comparison of Weak Molecular Adsorption of Organic Molecules on Clean Copper and Platinum Surfaces. Surf. Sci. 1984, 140, 227−248. (9) Ehlers, D. H.; Spitzer, A.; Lüth, H. The Adsorption of Methanol on Pt(111), an IR Reflection and UV Photoemission Study. Surf. Sci. 1985, 160, 57−69. (10) Akhter, S.; White, J. M. A Static SIMS/TPD Study of the Kinetics of Methoxy Formation and Decomposition on O/Pt(111). Surf. Sci. 1986, 167, 101−126. (11) Gibson, K. D.; Dubois, L. H. Step Effects in the Methanol Decomposition of Methanol on Pt(111). Surf. Sci. 1990, 233, 59−64. (12) Wang, J.; Masel, R. I. Structure Sensitivity of Methanol Decomposition on (1 × 1) and (2 ×1) Pt(110). J. Vac. Sci. Technol. A 1990, 9 (3), 1879−1884. (13) Wang, J.; Masel, R. I. C-O Bond Scission during Methanol Decomposition on (1 × 1) and (2 ×1) Pt(110). J. Am. Chem. Soc. 1991, 113, 5850−5856. (14) Franaszczuk, K.; Herrero, E.; Zelenay, P.; Wieckowski, A.; Wang, J.; Masel, R. I. A Comparison of Electrochemical and Gas-Phase Decomposition of Methanol on Platinum Surfaces. J. Phys. Chem. 1992, 96, 8509−8516. (15) Kizhakevariam, N.; Stuve, E. M. Promotion and Poisoning of the Reaction of Methanol on Clean and Modified Platinum (100). Surf. Sci. 1993, 286, 246−260. (16) Diekhöner, Butler, D. A.; Baurichter, A.; Luntz, A. C. Parallel Pathways in Methanol Decomposition on Pt(111). Surf. Sci. 1998, 409, 384−391. (17) Panja, C.; Saliba, N.; Koel, B. E. Adsorption of Methanol, Ethanol and Water on Well-Characterized Pt-Sn Surface Alloys. Surf. Sci. 1998, 395, 248−259. (18) Lu, C.; Thomas, F. S.; Masel, R. I. Chemistry of Methoxonium on (2 × 1) Pt(110). J. Phys. Chem. B 2001, 105, 8583−8590. (19) Peng, T. L.; Bermasek, S. L. The Internal Energy of CO2 Produced by the Catalytic Oxidation of CH3OH by O2 on Polycrystalline Platinum. J. Chem. Phys. 2009, 131, 154701. (20) Matolínová, I.; Johánek, V.; Myslivecek, J.; Prince, K. C.; Skála, T.; Škoda, M.; Tsud, N.; Vorokhta, M.; Matolín, V. CO and Methanol Adsorption on (2 × 1) Pt(110) and Ion-Eroded Pt(111) Model Catalysts. Surf. Interface Anal. 2011, 43, 1325−1331. (21) Schauermann, S.; Hoffmann, J.; Johánek, V.; Hartmann, J.; Libuda, J. Adsorption, Decomposition and Oxidation of Methanol on Alumina Supported Palladium Particles. Phys. Chem. Chem. Phys. 2002, 4, 3909−3918. (22) Schauermann, S.; Hoffmann, J.; Johánek, V.; Hartmann, J.; Libuda, J.; Freund, H.-J. The Molecular Origins of Selectivity in Methanol Decomposition on Pd nanoparticles. Catal. Lett. 2002, 84, 209−217. (23) Schauermann, S.; Hoffmann, J.; Johánek, V.; Hartmann, J.; Libuda, J.; Freund, H.-J. Catalytic Activity and Poisoning of Specific Sites on Supported Metal Nanoparticles. Angrew. Chem. Int. Ed. 2002, 41, 2532−2535.

combine with atomic hydrogen to form methane, instead of dehydrogenating to atomic carbon. The disparate processes confirm the different catalytic origins of the C−O bond scission as described above.



CONCLUSION We investigated the decomposition of methanol on Pt nanoclusters formed by vapor deposition on thin-film Al2O3/ NiAl(100) with STM, IRAS, TPD, RHEED, and synchrotronbased PES. The Pt clusters have a mean diameter near 2.3 nm and height near 0.4 nm; the size varies little before cluster coalescence; these clusters have a fcc phase and grow in orientation (111) or (001), depending on the growth temperature. The proportion of low-coordinated Pt sites ranges from 30% to 65% of total surface sites, for clusters at varied coverages. More than half the adsorbed monolayer methanol decomposes and reacts through either dehydrogenation to CO or scission of the C−O bond. In the dominant channel, adsorbed methanol begins to decompose to CO at the lowcoordinated Pt sites, at 150 K on Pt(001) clusters and 200 K on Pt(111) clusters, whereas both low-coordinated and some terrace Pt sites are reactive, independent of the cluster size. On average, one surface Pt produces one CO for monolayer methanol on either Pt(111) or Pt(001) clusters, despite the proportion of low-coordinated sites. In the other reaction path, the scission of the C−O bond of methanol occurs primarily without involving the possible dehydrogenation intermediates formaldehyde and formyl. The process proceeds at a higher temperature, near 250 K, and the CH3 intermediates combine with atomic hydrogen to form methane and to desorb above 300 K. The CO preadsorbed on the atop sites fail to terminate this reaction path; the C−O scission continues but gives methane formation and desorption at even higher temperatures, 370−600 K. The scission of the C−O bond shows an evident dependence on the size and lattice parameter of the clusters: as the size increases and the lattice parameter decreases with coverage, the probability of scission of the C− O bond decreases.



ASSOCIATED CONTENT

* Supporting Information S

Additional text describing the evolution of Pt cluster size with coverage and RHEED patterns from Pt clusters. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: mfl[email protected]. Tel: +886-3-4227151 ext. 65349. Fax: +886-3-4251175. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS National Science Council provided support (NSC-98-2112-M008-015-MY2 and NSC-100-2112-M-008-010-MY3) for the work; we thank Dr. Y.-J. Hsu (National Synchrotron Radiation Research Center of Taiwan), Dr. C.-M. Lin, G..-R. Hu, C.-W. Lin, Y.- J. Chang, and W.-S. Liao for technical assistance.



REFERENCES

(1) Hamnett, A. Mechanism and Electrocatalysis in the Direct Methanol Fuel Cell. Catal. Today 1997, 38, 445−457. 5676

dx.doi.org/10.1021/jp3080282 | J. Phys. Chem. C 2013, 117, 5667−5677

The Journal of Physical Chemistry C

Article

(24) Bäumer, M.; Libuda, J.; Neyman, K. M.; Rö sch, N.; Rupprechter, G.; Freund, H.-J. Adsorption and Reaction of Methanol on Supported Palladium Catalysts: Microscopic-Level Studies from Ultrahigh Vacuum to Ambient Pressure. Phys. Chem. Chem. Phys. 2007, 9, 3541−3558. (25) Freund, H.-J. Model Systems in Heterogeneous Catalysis: Selectivity Studies at the Atomic Level. Top. Catal. 2008, 48, 137−144. (26) Gassmann, P.; Franchy, R.; Ibach, H. Investigation on Phase Transitions within Thin Al2O3 Layers on NiAl(001) − HREELS on Aluminum Oxide Films. Surf. Sci. 1994, 319, 95−109. (27) Blum, R.; Ahlbehrendt, D.; Niehus, H. Growth of Al2O3 Stripes in NiAl(001). Surf. Sci. 1998, 396, 176−188. (28) Frémy, N.; Maurice, V.; Marcus, P. Initial Stages of Growth of Alumina on NiAl(001) at 1025 K. J. Am. Ceram. Soc. 2003, 86, 669− 675. (29) Zei, M. S.; Lin, C. S.; Wei, W. H.; Chiang, C. I.; Luo, M. F. Epitaxial Growth of Al2O3 on NiAl(100) by Gas Phase Oxidation and Electrooxidation. Surf. Sci. 2006, 600, 1942−1951. (30) Luo, M. F.; Chiang, C. I.; Shiu, H. W.; Sartale, S. D.; Kuo, C. C. Patterning Co Nanoclusters on Thin Film Al2O3/NiAl(100). Nanotechnol. 2006, 17, 360−366. (31) Luo, M. F.; Wen, W. H.; Lin, C. S.; Chiang, C. I.; Sartale, S. D.; Zei, M. S. Structures of Co and Pt Nanoclusters on a Thin Film of Al2O3/NiAl(100) from Reflection High-Energy Electron Diffraction and Scanning-Tunnelling Microscopy. Surf. Sci. 2007, 601, 2139− 2146. (32) Luo, M. F.; Ten, M. H.; Wang, C. C.; Lin, W. R.; Ho, C. Y.; Chang, B. W.; Wang, C. T.; Lin, Y. C.; Hsu, Y. J. TemperatureDependent Oxidation of Pt Nanoclusters on a Thin Film of Al2O3 on NiAl(100). J. Phys. Chem. C 2009, 113, 12419−12426. (33) Hong, I.-H.; Lee, T.-H.; Yin, G.-C.; Wei, D.-H.; Juang, J.-M.; Dann, T.-E.; Klauser, R.; Chuang, T. J.; Chen, C. T.; Tsang, K.-L. Performance of the SRRC Scanning Photoelectron Microscope. Nucl. Instrum. Methods A 2001, 467, 905−908. (34) Maurice, V.; Frèmy, N.; Marcus, P. Hydroxylation-Induced Modifications of the Al2O3/NiAl(100) Surface at Low Water Vapour Pressure. Surf. Sci. 2005, 581, 88−104. (35) Moulder, J. F.; Stickle, W. F.; Sobol, P. E.; Bomben, K. D. Handbook of X-ray Photoelectron Spectroscopy; Physical Electronics, Inc.: USA, 1995. (36) Sartale, S. D; Shiu, H. W.; Tien, M. H.; Chiang, C. I; Luo, M. F.; Lin, Y. C.; Hsu, Y. J. Adsorption and Decomposition of Methanol on Au Nanoclusters Supported on a Thin-Film of Al2O3/NiAl(100). J. Phys. Chem. C 2008, 112, 2066−2073. (37) Sartale, S. D; Shiu, H. W.; Tien, M. H.; Huang, Y. J; Luo, M. F. Scanning Tunneling Microscopy Study of Growth of Pt Nanoclusters on Thin Film Al2O3/NiAl(100). Surf. Sci. 2006, 600, 4978−4985. (38) Jaeger, R. M.; Kuhlenbeck, K.; Freund, H.-J.; Wuttig, M.; Hoffmann, W.; Franchy, R.; Ibach, H. Formation of a Well-Ordered Aluminium Oxide Overlayer by Oxidation of NiAl(110). Surf. Sci. 1991, 259, 235−252. (39) Luo, M. F.; Wang, C. C.; Hu, G. R.; Lin, W. R.; Ho, C. Y.; Lin, Y. C.; Hsu, Y. J. Active Alloying of Au with Pt in Nanoclusters Supported on a Thin Film of Al2O3/NiAl(100). J. Phys. Chem. C 2009, 113, 21054−21062. (40) Greeley, J.; Marvikakis, M. Competitive Paths for Methanol Decomposition on Pt(111). J. Am. Chem. Soc. 2004, 126, 3910−3919. (41) Martine, R.; Gardner, P.; Bradshaw, A. M. The AdsorbateInduced Removal of the Pt{100} Surface Reconstruction. Part II: CO. Surf. Sci. 1995, 342, 69−84. (42) Xu, J.; Yates, J. T., Jr Terrace Width Effect on Adsorbate Vibrations: A Comparison of Pt(335) and Pt(112) for Chemisorption of CO. Surf. Sci. 1995, 327, 193−201. (43) Mukerji, R. J.; Bolina, A. S.; Brown, W. A. A RAIRS and TPD Investigation of the Adsorption of CO on Pt{211}. Surf. Sci. 2003, 527, 198−208. (44) Yoshinobu, J.; Tsukahara, N.; Yasui, F.; Mukai, K.; Yamashita, Y. Lateral Displacement by Transient Mobility in Chemisorption of CO on Pt(997). Phys. Rev. Lett. 2003, 90, 248301.

(45) Sun, Y.-N.; Qin, Z.-H.; Lewandowiski, M.; Shaikhutdinov, S.; Freund, H.-J. CO Adsorption and Dissociation on Iron Oxide Supported Pt Particles. Surf. Sci. 2009, 603, 3099−3103. (46) Lundwall, M. J.; McClure, S. M.; Goodman, D. W. Probing Terrace and Step Sites on Pt Nanoparticles Using CO and Ethylene. J. Phys. Chem. C 2010, 114, 7904−7912. (47) Stakheev, A. Y.; Shpiro, E. S.; Jaeger, N. I.; Schulzekloff, G. FTIR Evidence of the Formation of Platinum Carbonyls from Pt Metal−Clusters Encaged in KL Zeolite. Catal. Lett. 1995, 34, 293− 303. (48) Taimoor, A. A.; Pitault, I.; Meunier, F. C. Correlation between Deactivation and Pt-Carbonyl Formation during Toluene Hydrogenation Using a H2/CO2 Mixture. J. Catal. 2011, 278, 153−161. (49) Beitel, G. A.; Laskov, A.; Oosterbeek, H.; Kuipers, E. W. Polarization Modulation Infrared Reflection Absorption Spectroscopy of CO Adsorption on Co(0001) under a High-Pressure Regime. J. Phys. Chem. 1996, 100, 12494−12502. (50) Risse, T.; Carlsson, A.; Bäumer, M.; Klüner, T.; Freund, H.-J. Using IR Intensties as a Probe for Studying the Surface Chemical Bond. Surf. Sci. 2003, 546, L829−L835. (51) Hu, G.-R.; Chao, C.-S.; Shiu, H.-W.; Wang, C.-T.; Lin, W.-R.; Hus, Y.-J.; Luo, M.-F. Low-Temperature Decomposition of Methanol on Au Nanoclusters Supported on a Thin Film of Al2O3/NiAl(100). Phys. Chem. Chem. Phys. 2011, 13, 3281−3290. (52) Oakes, D. J.; McCoustra, R. S.; Chesters, M. A. Dissociative Adsorption of Methane on Pt(111) Induced by Hyperthermal Collosions. Faraday Discuss. 1993, 96, 325−336. (53) The ionization cross-section for varied species and at varied partial pressure has been considered. (54) Huberty, J. S.; Madix, R. J. An FTIR Study of the Bonding of Methoxy on Ni(100): Effects of Coadsorbed Sulfur, Carbon Monoxide and Hydrogen. Surf. Sci. 1996, 360, 144−156. (55) Sim, W. S.; Gardner, P.; King, D. A. Structure and Reactivity of the Surface Methoxy Species on Ag{111}. J. Phys. Chem. 1995, 99, 16002−16010. (56) Barros, R. B.; Garcia, A. R.; Ilharco, L. M. The Decomposition Pathways of Methanol on Clean Ru(0001), Studied by ReflectionAbsorption Infrared Spectroscopy (RAIRS). J. Phys. Chem. B 2001, 105, 11186−11193. (57) Pratt, S. J.; Escott, D. K.; King, D. A. Multilayer Growth and Chemisorbate Reactivity of Methanol on Pd{110}. J. Chem. Phys. 2003, 119, 10867−10878. (58) Zhou, J.; Mullins, D. R. Adsorption and Reaction of Formaldehyde on Thin-Film Cerium Oxide. Surf. Sci. 2006, 600, 1540−1546. (59) Henderson, M. A.; Mitchell, G. E.; White, J. M. The Decomposition of Formaldehyde on Pt(111): A TPD and HREELS Study. Surf. Sci. 1987, 188, 206−218. (60) Attard, G. A.; Ebert, E. D.; Parsons, R. Formaldehyde Decomposition and Oxidation on Pt(110). Surf. Sci. 1990, 240, 125−135. (61) Hu, Y.; Griffiths, K. NO-Methanol Interaction on the Surface of Pt(332). Surf. Sci. 2007, 601, 2467−2472. (62) Gland, J. L.; Sexton, B. A.; Fisher, G. B. Oxygen Interactions with the Pt(111) Surface. Surf. Sci. 1980, 95, 587−602. (63) Hansen, K. H.; Worren, T.; Stempel, S.; Lægsgaard, E.; Bäumer, M.; Freund, H.-J.; Besenbacher, F.; Stensgaard, I. Palladium Nanocrystals on Al2O3: Structure and Adhesion Energy. Phys. Rev. Lett. 1999, 83, 4120−4123.

5677

dx.doi.org/10.1021/jp3080282 | J. Phys. Chem. C 2013, 117, 5667−5677