Effect of the Microstructure of the Semiconductor Support on the Photocatalytic Performance of the Pt-PtOx/TiO2 Catalyst System

The influence of the semiconductor microstructure on the photocatalytic behavior of Pt-PtOx/TiO2 catalysts was studied by comparing the methanol-reforming performance of systems based on commercial P25 or TiO2 from sol–gel synthesis calcined at different temperatures. The Pt co-catalyst was deposited by incipient wetness and formed either by calcination or high-temperature H2 treatment. Structural features of the photocatalysts were established by X-ray powder diffraction (XRD), electron spin resonance (ESR), X-ray photoelectron spectroscopy (XPS), optical absorption, Raman spectroscopy and TEM measurements. In situ reduction of Pt during the photocatalytic reaction was generally observed. The P25-based samples showed the best H2 production, while the activity of all sol–gel-based samples was similar in spite of the varying microstructures resulting from the different preparation conditions. Accordingly, the sol–gel-based TiO2 has a fundamental structural feature interfering with its photocatalytic performance, which could not be improved by annealing in the 400–500 °C range even by scarifying specific surface area at higher temperatures.


S1. Experimental
Raman spectra were recorded with a dynamically aligned Bio-Rad (Digilab) dedicated FT-Raman spectrometer equipped with a Spectra-Physics Nd-YAG-laser (1064 nm) and a high sensitivity liquid-N2 cooled Ge detector. The laser power used was about 250 mW in the samples. The resolution of the Raman instrument was ca. 4 cm −1 and a backscattered geometry was used. For each spectrum, 256 individual spectra were averaged.
A Philips CM12 instrument equipped with a high-resolution camera was used to acquire and elaborate the TEM images. Powdered samples were dispersed in 2-propanol under ultrasound irradiation and the resulting suspension dropwise was deposited on a holey carbon-coated support grid.
The ESR experiments were performed with a Bruker Elexsys E500 X-band spectrometer. A typical microwave power of 1 mW and 1 G magnetic field modulation at ambient temperature were used. The magnetic field was calibrated with an NMR field meter. Signal intensity, linewidth and g-factor (spectroscopic splitting factor) values were used to characterize the samples. The knowledge of the g-factor can give information about a paramagnetic center's electronic structure.
X-ray photoelectron spectroscopy (XPS) measurements were carried out using an EA125 electron spectrometer manufactured by OMICRON Nanotechnology GmbH (Germany). The photoelectrons were excited by both MgK (1253.6 eV) and AlK (1486.6 eV) radiation. Spectra were recorded in the Constant Analyzer Energy mode of the energy analyser with 30 eV pass energy resulting in a spectral resolution around 1 eV. For XPS experiments the samples in the form of fine powder were suspended in isopropanol. Drops of this suspension were placed on standard OMICRON sample plates; after evaporation of the solvent catalyst coatings with sufficient adhesion and electric conductivity were obtained. Effects of possible electric charging were compensated by adjusting the binding energy of the main component of the C 1s envelope (hydrocarbons) to 285.0 eV. By this choice both the Ti 2p3/2 and the O 1s binding energies coincided with the range expected for TiO2, confirming the reliability of the calibration. Chemical states of the elements were deduced from high-resolution spectra using XPS databases [1,2]. Quantification was performed using a combination of CasaXPS [3] and XPSMultiQuant [4,5].
The photocatalytic reaction in the UV-visible region was carried out in a cylindrical quartz reactor (140 mm in height and 60 mm in diameter) equipped with magnetic stirrers, gas inputs and outputs. Osram HQL de luxe 125W lamps were used as light sources operated. Nitrogen gas with 20 cm 3 /min flow rate was continuously bubbled through the reactor. The initial concentration of methanol was 6 v/v% in distilled water. The reaction was carried out at room temperature. The amount of catalyst and the reaction volume was 0.140 g and 260 cm 3 , respectively. For experiments in the visible light region, a homemade LED light source operating at four distinct wavelengths (400, 420, 440, 450 nm) is available but this part of the reactor system was not used in the present experiments. In this latter part of the equipment, the LED light source is immersed in a UV-Consultig Peschl UV-Reactor System 1 photoreactor equipped with gas inputs and outputs. The reaction setup is depicted in Figure S1. GC (Agilent 7820A) equipped with SUPELCO Carboxen 1010 column and TCD detector was used to follow the H2 production. The internal standard of the GC analysis was argon gas added to the vapor-gas mixture before the GC sampling valve. The H2 production was expressed as H2 production rate (mmol h −1 ). The reaction was monitored for 270 min. After the photocatalytic reaction, the samples were recovered from the aqueous methanol solution by centrifuging, washing with 3 × 50 cm 3 absolute ethanol followed by drying under N2 flow. Then the recovered samples were also characterized by ESR, XPS and/or TEM techniques. Figure S1. Scheme of the photocatalytic reactor system. The LED light illuminated part of the reactor system (reactor on the left side) was not used in the present experiments.

S2.1 Raman Spectroscopic Measurements
The results of Raman spectroscopic measurements ( Figure S2) are consistent with the XRD results ( Table 1 in the main text); our sol-gel prepared TiO2 samples contained exclusively anatase as a crystalline phase while P25 contained anatase and rutile phases. Figure S2. Raman spectra of the bare TiO2 samples. a: P25; b: sol-gel precipitation followed by calcination at 400 °C (SG400); c: sol-gel precipitation followed by calcination at 450 °C (SG500), d: sol-gel precipitation followed by calcination at 500 °C (SG500). Figure S3 shows the TEM images of TiO2 samples prepared by sol-gel precipitation followed by calcination. The polycrystalline anatase structure with a particle size of 10-40 nm was obtained.

S2.3. Evaluation of the Optical Absorption Spectra
In order to estimate the optical band gaps of the four TiO2 materials, the standard procedure of processing the diffuse reflectance spectra [6] was followed. First, reflectance data were transformed into the Kubelka-Munk function , where R is the measured reflectance), which represents the wavelength dependence of the absorption coefficient. As the major crystalline phase in all four materials is anatase with indirect bandgap [7], the quantity was plotted as the function of the photon energy (Ephoton) and the intersection of its linear fit in the absorption edge region with the baseline was determined (Tauc plot method D [8]). Note that in the case of electronic transitions at interfaces, where momentum non-conservation can occur [9] or in amorphous semiconductors [8], the same formula describes the absorption edge. The Tauc plots are shown in Figure S4A. The resulting band gap values are very close to each other and are around 3.00-3.15 eV.
According to the theory of disordered semiconductors [8], defects (i.e. deviations from the normal coordination in terms of positions and chemical identity of atoms) inevitably generate localized electronic states near the band edges of the ideal material, which -in terms of the density of states -form more or less exponential tails into the bandgap. In addition, deep localized states also occur in the bandgap region. Optical excitations involving these localized states are necessarily observable in the absorption spectrum. As a result, the absorption spectrum of disordered semiconductors contains three distinct regions: (i) at photon energies much lower than the bandgap transitions between the tail states give a weak absorption tail which is sometimes exponential; (ii) around the bandgap energy, another exponential segment exists due to transitions involving tail and non-localized band states which are known as the Urbach tail; and (iii) above the bandgap, energy absorption is due to interband transitions and the spectrum follows the Tauc rule. The magnitude of the absorption in the region (i) may be regarded as a tentative indicator of the extent of disorder in the material.
The logarithmic representation of the Kubelka-Munk functions of the materials studied in this work is shown in Figure S4B. It is obvious either from the absorbance data ( Figure 3 of the main paper) or the semi-logarithmic plots of the Kubelka-Munk functions normalized to each other (insert of Figure S3B) that absorption in the weak absorption region (i.e., well below 3 eV) is the weakest for the SG500 material, only slightly higher for P25, considerably higher for SG450 and the highest for SG400. This means that while the SG500 and the P25 samples are relatively free from defect-induced deep localized states, the densities of state tails penetrating deeply into the bandgap are prominent features of the electronic structure of SG400, and SG450 represents an intermediate situation. In all materials, a pronounced linear part can be identified in the logarithmic representation of the Kubelka-Munk function near the bandgap energy. This exponential tail of the absorption edge, also known as the Urbach tail, is a consequence of the disorder-induced smearing of the band edges. The slope of the linear segment of the semi-logarithmic plot of the absorption edge is a direct measure of the extent of the disorder [10] or the width of the disorder-induced tails of the density of states [8]: the lower the slope (or the higher its reverse, the Urbach width or Urbach energy), the higher is the disorder in the system. Indeed, the Urbach energy is lowest for the P25 sample, slightly higher for the SG500 and significantly increases through SG450 to SG400. This order again confirms that disorder-related localized electronic states dominate the optical response of the SG400 material and even in SG500 their role is more prevalent than in P25.
It must be noted that even if these disorder-induced electronic states may offer optical absorption well into the visible range, the generated electron-hole pairs remain localized either at or near the surface of the semiconductor or even in its bulk and their chance for participation in a photocatalytic process remains very low. On the other hand, these states can act as electron traps and/or recombination centers, which further diminishes the photocatalytic performance of the material. Figure S5 shows the TEM images of the fresh sol-gel based Pt-PtOx/TiO2 samples. According to Figure S5, TiO2 was consistent with polycrystalline anatase having a particle size of 10-40 nm. All the samples (of SG400PtH2red, SG450PtH2red and SG500PtH2red) with co-catalyst obtained by hydrogen reduction were very similar to each other (see Figure  S5A, C, E) with a few features in the 2-4 nm range giving darker contrast, which were identified as reduced Pt particles, whose presence was also confirmed by XPS. When the Pt was formed by calcination, Pt containing nanoparticles cannot be observed using conventional bright-field TEM [11] due to the little contrast difference between PtOx and TiO2 (also see Figure S5B, D, F).

S3.1 TEM Measurements
According to the TEM images of recovered samples, Pt-like nanoparticles can be observed even in the case of PtOx/TiO2 sample formed by calcination in air ( Figure S6).

S3.2 XPS Measurements
While the Pt 4f core-level spectra of the Pt-loaded TiO2 samples recorded after 1h X-ray exposure are compared in Figure 7 of the main text, Figure S7 shows spectra obtained immediately after the beginning of the XPS measurement, so the in-radiation reduction of the Pt content is minimal. Materials, it is evident that in the case of the catalysts prepared by reduction, the Pt content is very easily reducible even under the UHV conditions of the XPS experiment (probably due to the slight heating effect of the X-ray source). Nevertheless, while in the case of the SG400 and SG450-supported reduced samples, a significant decrease of the Pt 2+ contribution is evident during the measurement, the SG500-supported reduced sample is more reduced initially and the further reduction remains limited. This indicates that in the Pt-loaded SG400 or SG450 catalysts, either the initial reduction was less complete (which is not very probable, as when a Pt-oxide fraction is not reduced at 400 °C in hydrogen, it should be stable upon the more gentle conditions of the XPS experiment) or the supports prepared at lower temperatures facilitate the reoxidation of Pt upon air exposure to a higher extent compared to the SG500 material. On the other hand, the PtP25 sample formed by reduction was indeed quite oxidized at the beginning of the XPS experiment and was reduced under the measurement conditions to a much smaller extent than its SG-based counterparts, which suggests that the P25 material has a more pronounced stabilization effect on the ionic Pt species than the SG-based materials.
While in general, the spectral changes of the Pt-loaded SG samples formed by calcination during the XPS experiment remained much more limited (see Figure 7B of the main text and Figure S7B of the Supplementary Materials), it appears that not only the Pt 2+ but even the Pt 4+ state is more stable on the SG400 or SG450 supports than on the SG500. In the case of the P25-based catalyst activated by calcination, only a certain decrease of the Pt 4+ content is evident during the XPS measurement and the metallic Pt, which was detected on the calcined SG-based catalysts even at the very beginning of the XPS experiment, and does not appear even after long measurement times.

S3.3 ESR Measurements
As the line shape measured for the Pt-loaded photocatalyst samples was not perfect Lorentzian, we carried out a series of simulations to find the best fit, the quality of the simulation was characterized by R (coefficient of correlation). We performed fits with isotropic (Lorentzian, Gaussian, superposition of Lorentzian and Gaussian, dispersion distorted Lorentzian) and rhombic (Lorentzian, Gaussian) models. The unresolved rhombic symmetry model gives the best results (see Table S1). The ESR results on selected sol-gel based Pt-PtOx/TiO2 samples before and after the photocatalytic reaction are compared in Figure S8. The g values of the freshly prepared and used samples are slightly different. The line width of the reduced samples is significantly larger than that of the calcined ones, although an increase for the latter is evident during the photocatalytic reaction. The most likely cause of line broadening is the so-called g-strain (owing to small sample inhomogeneity). It may be noted that a correlation is apparent between the pre-annealing temperature of the sol-gel based support and the extent of the line broadening during the photocatalytic process.
In addition, the anisotropy of the g-values seems to be slightly more pronounced in the reduced samples. From the values of g, it is not possible to clearly determine the origin of the sharp line, although it is still very probable that the signal could be attributed to a single electron trapped in an oxygen vacancy [12].
The differing anisotropy and line widths suggest that the reductive or oxidative co-catalyst activation may result in the formation of vacancies with somewhat different environments. It is interesting to see that the line width difference between the recovered samples formed by H2 treatment and calcination significantly decreases during the photocatalytic reaction which means that the nature of the defect sites became more similar after the photocatalytic action. This observation coincides with the XPS results, which also indicate that the surface state of all recovered sol-gel based photocatalysts is very similar. In turn, the similar Pt chemical states and the similar defect structure of the semiconductor indeed explain the similar catalytic behavior of the SG catalysts formed by H2 reduction or calcination.