Interfacial charge transfer in Pt-loaded TiO2 P25 photocatalysts studied by in-situ diffuse reflectance FTIR spectroscopy of adsorbed CO

The efficiency of photocatalytic systems is strongly depending on the charge carrier transfer from the excited semiconductor to co-catalyst particles attached on its surface. In this study, we investigated the influence of photo-induced charge transfer in photoplatinized TiO2 P25 photocatalysts by diffuse reflectance FTIR spectroscopy of CO molecules adsorbed on the Pt co-catalyst under well-defined gas phase conditions. In contrast to aqueous conditions, where shifts of the CO stretching vibration of up to 50 cm have been reported, the observed shifts under gas phase conditions are very small (< 1 cm). This demonstrates that the difference in dielectric properties between aqueous electrolytes and vacuum are critical for the development of prominent shifts of adsorbed CO bands upon trapping of photogenerated charge carriers on co-catalyst particles. The experimental findings are discussed in terms of an electrostatic Stark effect, charge screening, co-adsorption, coverage-dependent shifts of the vibrational bands of adsorbed CO and photocatalytic surface reactions.


Introduction
Solar-driven chemical energy conversion, production of chemicals and remediation of pollutants have attracted substantial research efforts over the last decades [1][2][3]. The main challenge of the transition towards a solar-powered economy is the intermittency of solar light, which requires a scalable technology to convert, store and transport solar energy in a convenient form. In this regard, production of solar fuels (e.g. H 2 , CH 4 , alcohols) by means of photocatalytic water splitting and CO 2 reduction can be a promising solution [3][4][5][6][7][8]. However, most photocatalytic systems have rather low solar-to-chemical energy conversion efficiencies. The performance of a photocatalytic system can be improved by loading suitable co-catalysts which increase selectivity and lower over-potentials of the desired redox reactions [9,10]. Also, co-catalysts aid separation and utilization of photogenerated charge carriers which determine the activity of a photocatalytic material. Consequently, a detailed understanding of these processes is of utmost importance for the optimal design of such systems.
Light-induced processes in inorganic semiconductor photocatalysts can be studied by mid-infrared spectroscopy. This technique can access both the molecular species present at the semiconductor surface and the dynamics of majority charge carriers [11][12][13][14][15][16]. Small probe molecules such as CO are sensitive to the electronic structure of adsorption sites [17] and, therefore, are widely used to probe for instance Lewis and Brønsted acid sites in solid catalysts and to study the oxidation state and distribution of supported metal nanoparticles by means of vibrational spectroscopies [19]. Moreover, light-induced changes of the electrochemical potential at the semiconductor/liquid interface can be monitored in-situ via the shifts of vibrational bands of CO adsorbed on metal co-catalyst particles [20,21]. Shen et al. recently reported an 11 cm −1 red shift of the vibrational band of CO adsorbed at the gas/solid interface, when platinized TiO 2 was exposed to UV irradiation [22]. The observed shift was ascribed to the transfer of photogenerated electrons from titania to the Pt nanoparticles. This effect can in principle be used to understand how loading, dispersion, and oxidation state of the metal co-catalyst affect charge carrier transfer and utilization in co-catalyst loaded photocatalytic systems.
To the best of our knowledge, the work of Shen et al. [22] is the only study reporting such prominent light-induced shifts of the adsorbed CO band for gas/solid systems. In the present work, we systematically studied this effect for platinized TiO 2 P25 photocatalysts by steady-state and time-resolved diffuse reflectance FTIR spectroscopy (DRIFTS) under well-defined conditions. Contrary to literature, we did not observe any prominent shifts of the vibrational bands of CO adsorbed on platinized photocatalysts at the time scale ranging from sub-seconds to minutes. The largest light-induced blue shift of ca. 0.5 cm −1 was observed for the 2112-2114 cm −1 band of CO adsorbed on isolated Pt atoms present in Pt-TiO 2 sample.

Pt-TiO 2 synthesis
0.5 wt.% Pt was photodeposited on commercial Aeroxide P25. Briefly, 500 mg TiO 2 were dispersed under continuous stirring in 90 mL 10 vol.% aqueous CH 3 OH containing 2.5 mg Pt in the form of H 2 PtCl 6 . The dispersion was transferred in a double-walled glass reactor with a top-mounted quartz window, degassed and irradiated with the full spectrum of a 500 W Hg(Xe) lamp (Newport, USA). The lamp was equipped with an IR water filter and a full spectrum turning mirror. The suspension was irradiated for 1 h and a grey-brownish solid was separated by centrifugation, washed twice with demineralized water and dried under vacuum at room temperature.

Oxidative treatment of Pt-TiO 2
In order to remove organic adsorbates present on Pt-TiO 2 the sample was pre-treated as follows. The material was placed in an in-situ low-temperature chamber (Harrick Scientific), evacuated to the lowest stable pressure (ca. < 10 −3 mbar) and kept under dynamic vacuum for 0.5 h. Then, 100 mbar O 2 were added followed by heating the sample to 523 K and irradiation with the 325 nm line of a continuous-wave He-Cd laser (Kimmon Koha). The light intensity at the optical window of the DRIFTS cell was ca. 10 mW/cm 2 . This treatment removed a substantial amount of hydrocarbons and oxygenates present on the untreated material.

Characterization
The morphology of the samples and the Pt particle size distribution were studied by bright field transmission electron microscopy (FEI Technai G2, 200 kV). Diffuse-reflectance UV-vis spectra of the presented samples have been reported in a former publication of our group [23].

Infrared spectroscopy
All diffuse-reflectance infrared Fourier transform (DRIFT) measurements were carried out on a Bruker Vertex 70v FTIR spectrometer equipped with a liquid-nitrogen cooled MCT detector and a diffusereflectance accessory (Praying Mantis, Harrick Scientific). Powder samples were placed in a low-temperature chamber (Harrick Scientific) equipped with two KBr windows for IR signal collection and one quartz window for UV irradiation of the samples. The spectra were recorded at 4 cm −1 resolution in the 3950-600 cm −1 range by averaging 100 individual scans at room temperature (20°C), unless otherwise stated. Dry KBr (Sigma Aldrich, IR spectroscopy grade) was used as reference for the survey DRIFT spectra. For CO adsorption experiments, pure CO gas was introduced into the evacuated in-situ DRIFTS cell, typically to a partial pressure of 2 mbar, and let equilibrate for several minutes. The sample in the dark or before exposure to CO was used as the reference for difference spectra. The 325 nm line of a continuous wave Cd-He laser (Kimmon Koha) was used as the light source. Light intensity at the quartz window of the DRIFTS cell was ca. 10 mW/cm 2 . The exposure time was controlled with an optical shutter (Thorlabs). A low-pass IR filter (cut-off frequency 3950 cm −1 ) was placed in front of the detector compartment to block stray UV light and prevent spectral back-folding (aliasing).
Besides this, a broad absorption feature emerged in the region below 2000 cm −1 . Its intensity increased toward lower wavenumbers and could be fit with an exponential function describing IR absorption due to free conduction band electrons (CBE) [11][12][13][14]: where A is the proportionality coefficient and v is the frequency of IR irradiation in cm −1 . After several minutes under UV light, bands at 2115 cm −1 and 2057 cm −1 emerged in the DRIFT spectrum pointing to photocatalytic decomposition of organic contaminant species into CO. Intensities of these bands slowly increased over time under irradiation. In the dark, the 2115 cm −1 band was stable for several hours while the 2057 cm −1 band lost ca. 70% of intensity within the first 10-15 min and then decayed further at slower rates (Fig. 2). These bands can be ascribed to CO adsorbed on isolated Pt atoms (2115 cm −1 ) and metallic Pt particles (2057 cm −1 ), respectively [31][32][33][34].
The components at 2087 cm −1 and 2067 cm −1 were ascribed to CO linearly adsorbed on metallic Pt particles while the band at 1836 cm −1 is characteristic for CO adsorbed on Pt metal in bridged configuration [32][33][34]. Exposure of this sample to 325 nm light induced no apparent shift of the adsorbed CO bands (Fig. 3A) even under prolonged (0.5 h) irradiation. When the spectra recorded under UV irradiation were referenced to the dark spectrum, minor intensity losses at 2099 cm −1 , 2054 cm −1 and 1860 cm −1 became apparent (Fig. 3B, red spectrum). These spectral changes did not relax in the dark completely and some intensity loss at 2114 cm −1 was permanent, pointing to partial desorption of CO under illumination. Repetitive exposure to UV light did not result in accumulation of these effects, though. Similar spectral changes were observed when untreated Pt-TiO 2 with adsorbed CO was heated from 293 K to 323 K ( Figure S2, Supporting Information).

Oxidized Pt-TiO 2 under 325 nm irradiation
In order to remove organic adsorbates present on untreated Pt-TiO 2 , the material was subjected to an oxidative treatment as described in Section 2. The dark DRIFT spectrum of the oxidized sample is shown in Fig. 1, red curve. Intensities of the CH x (3000-2900 cm −1 ) and carboxylates (1535 cm −1 , 1438 cm −1 ) bands became substantially lower after the treatment [26,30]. When oxidized Pt-TiO 2 was exposed to UV light, the bands of adsorbed water and surface hydroxyls decreased in intensity and the broad CBE absorption feature appeared in the spectrum below 2000 cm −1 . Prolonged irradiation (1 h 20 min) of oxidized Pt-TiO 2 still led to the formation of bands at 2114 cm −1 and 2057 cm −1 ( Figure S3, Supporting Information). However, compared with the untreated sample, these bands developed much slower and had significantly lower intensities suggesting that persistent adventitious carbon containing adsorbates present on TiO 2 or from the background gas in the in-situ cell were involved in formation of these CO bands. When compared with bands resulting from intentionally dosed CO at 2 mbar, these contributions are ∼200 times less intense and thus, were neglected in the further experiments.

Effect of 325 nm irradiation on CO adsorbed on oxidized Pt-TiO 2
When oxidized Pt-TiO 2 was exposed to 2.0 mbar CO, several bands appeared in the ν(CO) region (Fig. 4A). Positions and structure of the bands arising from CO adsorbed on metallic Pt particles in the linear (2087 cm −1 , 2067 cm −1 ) and bridged (1836 cm −1 ) configuration were similar to those observed in the untreated sample (Fig. 3A). This suggests that the oxidative treatment did not cause any prominent sintering, reconstruction or oxidation of the Pt nanoparticles. The main difference between the untreated and oxidized samples was a shift of the 2114 cm −1 band to 2112 cm −1 after the oxidative treatment. No apparent shift of the adsorbed CO bands was observed when oxidized Pt-TiO 2 was exposed to 325 nm irradiation (Fig. 4B). First exposure of   oxidized Pt-TiO 2 to UV light led to a loss of intensity at 2114 cm −1 and 2099 cm −1 while the components at 2028 cm −1 and 1840 cm −1 became more prominent. Most of these spectral changes did not recover in the dark (Fig. 4B, blue curve) suggesting light-induced desorption of small quantities of CO. Repeated exposure to 325 nm irradiation did not lead to any further spectral changes.

Shift of the 2112 cm −1 band (CO adsorbed on atomic Pt species) under 325 nm irradiation
The 2120-2100 cm −1 spectral region exhibited complex band rearrangements when untreated or oxidized Pt-TiO 2 with adsorbed CO was exposed to 325 nm irradiation (Figs. 3B and 4B). Overlapping bands of CO adsorbed on isolated Pt atoms and metal nanoparticles complicated the analysis of these light-induced effects. Hence, we separately studied the effect of UV irradiation on CO adsorbed on isolated Pt atoms. To prepare this situation, oxidized Pt-TiO 2 was equilibrated with 2.0 mbar CO for one hour and then evacuated and kept under dynamic vacuum at room temperature for another hour. In this way, CO desorbed from Pt nanoparticles (i.e. 2087-2067 cm −1 and 1836 cm −1 bands almost completely disappeared) and the sample only exhibited the well-resolved band at 2112 cm −1 with only a minor component at 2084 cm −1 (Fig. 5). The 2112 cm −1 band was stable under dynamic vacuum for hours which agrees with its assignment to CO adsorbed on atomically dispersed Pt [31].
When this sample was exposed to 325 nm irradiation, the apparent band maximum shifted from 2112.3 cm −1 to 2112.9 cm −1 and its full width at half maximum increased from 8.3 cm −1 to 8.5 cm −1 (Fig. 5). These spectral changes reversed in the dark and could originate either from charging of Pt atoms or from the light-induced heating of the material. To test the latter hypothesis, we heated the sample under static vacuum in the dark. The 2112 cm −1 band experienced a red shift, when the temperature increased from 273 K to 353 K. However, a small blue shift was observed at temperatures above 353 K ( Figure S4, Supporting Information). At 383 K, the position and full width at half maximum of the 2112 cm −1 carbonyl band were similar to those observed under 325 nm irradiation (cf. Fig. 5). Hence, the shift of the 2112 cm −1 band observed under UV irradiation cannot be unambiguously attributed to the interaction of adsorbed CO with photogenerated charge carriers. Moreover, the changes of carbonyl bands induced by above-bandgap irradiation of Pt-TiO 2 P25 took place at the time scale of minutes while dynamics of photogenerated electron absorption happened at a shorter time scale ( Figure S5, Supporting Information) and charge transfer between oxide semiconductors and cocatalysts has been reported to occur at sub-μs time scales [35,36]. This discrepancy further supports the conclusion that the changes of the vibrational bands of adsorbed CO observed under above-bandgap irradiation cannot be unequivocally ascribed to charge carrier (electron) transfer between titania and Pt surface species alone.
3.5. Discussion of excitation induced ν(CO) band shifts at the solid-gas interface in Pt-TiO 2 P25 Under 325 nm irradiation, Pt-TiO 2 P25 did not exhibit any prominent shifts of the adsorbed CO bands which can be unambiguously attributed to persistent charge transfer to the Pt co-catalyst particles. On the other hand, CO adsorbed on metal electrodes [37][38][39] or semiconductor-supported Pt particles [18,19] immersed in an aqueous electrolyte show shifts up to 50 cm −1 /V, when they are exposed to above-bandgap excitation or when an external electric bias is applied to the system. In order to understand this discrepancy one should consider that, in general, the effect of the applied electric field on the CO band is more prominent at the liquid/solid [18,19,37,38] interface than in gas/ solid [40][41][42] systems, relating to the relative permittivities of water (ε r ≈ 80) and vacuum or air (ε 0 (vac.) = 1, ε r (air) ≈ 1). The influence of adding or removing charge to or from the adsorption site (here Pt) on the vibrational frequency of adsorbed CO is described in the Blyholder model in terms of changes in the electron density in the 2π* orbital of the CO molecule upon back-donation from filled metal d-orbitals, which is a localized description [17]. In aqueous liquid/solid systems the electrical bias induces shifts of the CO band position in the order of 28-50 cm −1 /V, while the effect of the applied electrical field on the vibrational frequency of CO adsorbed at the gas/solid interface is in the order of 10 −6 cm −1 /(V/cm) [40][41][42]. For both systems, the influence of the electric field on the vibrational frequencies of adsorbed CO can be described by the electrostatic Stark effect [43]. However, different relative permittivities of vacuum and aqueous electrolytes [39] and the presence of the electrochemical double layer at the liquid/solid interface leads to very prominent ν(CO) shifts observed under electrochemical conditions. On the other hand, the electric field strength in vacuum is limited by the break-down voltage which would lead to band shifts in the order 10 −2 cm −1 [42]. Interestingly, theoretical studies performed with CO adsorbed on small Cu and Au clusters predicted shifts of ν(CO) in order of tens of cm −1 upon addition or removal of one electron [44,45].
By using classical electrodynamics and approximating an average Pt nanoparticle with a solid sphere of radius R = 1 nm ( Figure S6, Supporting Information), one can calculate the electrical field strength E(x) at the surface of such a particle, when one electron (Q = 1.602 × 10 −19 C) is added to or removed from the system: where ε 0 = 8.854 × 10 −12 F/m is the vacuum permittivity. This rough approximation returns E(x) ≈ 1.5 MV/cm, which would induce a shift of ν(CO) of about 15 cm −1 . However, we did not observe such prominent band shifts in our experiments. This can be explained by screening of the electrical charge by the d-band of the Pt metal [46] which was not taken into account by Eq. (2). A Pt nanoparticle with 1 nm radius ( Figure S6, Supporting Information) contains about 250-300 atoms. Position and structure of the vibrational bands of adsorbed CO suggest that these particles are metallic. Therefore, the electron gas can efficiently screen the transferred electrical charge resulting in a substantially weaker effective electrical field on the surface of the Pt nanoparticles than that estimated with Eq. (2) [46]. On the other hand, screening cannot explain why the bands of CO adsorbed on isolated Pt atoms were not prominently affected by above-bandgap irradiation , under the 325 nm irradiation (red) and the difference between these two spectra (blue).
( Fig. 5). Such small metal clusters (or atoms) do not develop d-bands but can rather be described by molecular orbitals. Therefore, trapping of a charge carrier by these species would prominently affect the interaction of adsorbed CO molecules with Pt clusters and, consequently, the position of ν(CO). However, the absence of a prominent band shift of CO adsorbed on isolated Pt atoms suggests high energy penalties of transferring one electron to an isolated Pt metal atom which effectively can hinder such a charge transfer.

Conclusions
The influence of UV irradiation on the vibrational bands of CO adsorbed at the gas-solid interface of platinized TiO 2 P25 was studied by DRIFT spectroscopy. No prominent shifts of the vibrational bands of CO adsorbed on metallic Pt nanoparticles were observed under above bandgap (UV, 325 nm) irradiation. Exposure of Pt-TiO 2 to UV light caused a blue shift of ca. 0.5 cm −1 of the 2112-2114 cm −1 band corresponding to CO adsorbed on isolated, atomic Pt species. A similar rearrangement of this carbonyl band was observed upon heating of the material from 293 K to 383 K. Therefore, the effect of above-bandgap excitation on the adsorbed CO band positions cannot be unambiguously attributed to charge transfer between the oxide support and Pt species. Based on these results we conclude that charge transfer between TiO 2 -P25 and a photodeposited Pt co-catalyst does not induce prominent band shifts of CO adsorbed at the gas/solid interface. This contrasts large shifts observed for liquid/solid interfaces and demonstrates that the dielectric properties of the surrounding medium are crucial for the formation of electric fields of sufficient strength to induce prominent stretching vibration band shifts of adsorbed CO molecules. Our work highlights the use and limitations of CO as a charge sensitive probe molecule in the investigation of photocatalytic systems. In such studies, it is essential to take into consideration parameters like particle size (atomic species vs. large nanoparticles), sample temperature, as well as the electronic properties of the interface which is largely governed by the dielectric properties of the surrounding medium (vacuum or gas vs. aqueous environments).