In Situ Investigation of Charge Performance in Anatase TiO2 Powder for Methane Conversion by Vis–NIR Spectroscopy

The intrinsic behavior of photogenerated charges and reactions with chemicals are key for a photocatalytic process. To observe these basic steps is of great importance. Here we present a reliable and robust system to monitor these basic steps in powder photocatalysts, and more importantly to elucidate the key issue in photocatalytic methane conversion over the benchmark catalyst TiO2. Under constant excitation, the absorption signal across the NIR region was demonstrated to be dominated by photoexcited electrons, the absorption of photoexcited holes increases toward shorter wavelengths in the visible region, and the overall shapes of the photoinduced absorption spectra obtained using the system demonstrated in the present work are consistent with widely accepted transient absorption results. Next, in situ measurements provide direct experimental evidence that the initial step of methane activation over TiO2 involves oxidation by photoexcited holes. It is calculated that 90 ± 6% of photoexcited electrons are scavenged by O2 (in dry air), 61 ± 9% of photoexcited holes are scavenged by methane (10% in argon), and a similar amount of photoexcited electrons can be scavenged by O2 even when the O2 concentration is reduced by a factor of 10. The present results suggest that O2 is much more easily activated in comparison to methane over anatase TiO2, which rationalizes the much higher methane/O2 ratio frequently used in practice in comparison to that required stoichiometrically for photocatalytic production of value-added chemicals via methane oxidation with oxygen. In addition, methanol (a preferable product of methane oxidation) is much more readily oxidized than methane over anatase TiO2.


I. Xe lamp excitation intensity
The focal point of the Xe-lamp-lens configuration was determined to be between 10 and 14 cm, with the beam diameter varying from c.a. 0.5 cm to 0.3-0.4 cm to 1.5 cm as the distance is varied from 10 to 12 to 14 cm. The diameter of the sample holder is c.a. 0.6 cm, thus the beam diameter of the (approximately) focused Xe lamp output is roughly matched with the sample diameter.
The Xe lamp is positioned such that the total distance travelled by the Xe lamp output from the lens to the sample is c.a. 12±2 cm. Because 1) the distance between the lens and the sample can only be estimated, 2) the Xe lamp is sometimes dismantled for another experiment then re-assembled for the present experiment, and 3) the power density varies significantly with distance from the filterlens assembly, the optical power of the filtered and focused Xe lamp was measured at 3 different distances (10, 12, and 14 cm) from the end of its optics holder. The results are summarised in Table   S1. After passing through a combination of the 325-385 nm bandpass, 365 nm bandpass, and the focusing lens, the maximum power density varies from c.a. 1 to 15 mW/cm 2 .
Given that 1) the Xe lamp illumination is incident on the sample at an angle, and 2) the maximum of the Xe lamp output is unlikely to be perfectly overlapped with the small internal measurement light, the actual power density at the sample is likely to be smaller than those calculated in Table S1. The final effective power density of the 365 nm excitation light is therefore estimated to be c.a. 1 mW/cm 2 .

III. Raw reflectance data for TiO2 measurements
Raw %R spectra for anatase TiO2 powder in the presence of air, methanol, methane, and 4/1 Methane/O2 are respectively shown in Figure S1, Figure S2, Figure S3, and Figure S4. The slight change in overall %R going from an Argon atmosphere to air and 4/1 methane/O2 is due to slight change in reactor height and/or orientation upon opening and closing the reactor gas valves.

IV. Supplementary information for BaSO4 measurements
BaSO4 powder was loaded into the reaction chamber sample holder and pressed flat using a spatula.
Prior to measurement under each environment, the reactor chamber was purged with 150 ml/min of the relevant gas for 10 minutes, then the outlet and inlet reactor valves were closed to seal in the slightly pressurised gas. Measurements were first performed under 100% Argon, followed by (dry) air, then 10% methane in Argon.
V. Photoinduced absorbance spectra      VI. Slope analysis of photoinduced absorbance spectra plotted on log-log scale Figure S10 shows the %abs spectra in Figure S5 Aa) re-plotted on a log-log scale. Three out of the four spectra appear to be approximately linear on a log-log scale in the NIR region, but some curvature in the plots are apparent when compared to the straight-line fit. The slopes of the approximately linear plots were evaluated. Results of linear fit through the data between 1000 and 2600 nm are shown in Table S2, and is the source of the slope value of 0.63±0.03 reported in the main text. Figure S11 shows the %abs spectra in Figure S6 a) re-plotted on a log-log scale. Results of linear fit through the data between 1000 and 2600 nm are shown in Table S3, and is the source of the slope value of 0.61±0.06 reported in the main text.
To test the effect of expressing the photoinduced absorption amplitude in different units, the %R data was also processed using the KM and log(1/r) transformations, with the relative reflectance ( ) being: The KM and log(1/r) transformations were performed on the same dataset as that used to obtain the %abs spectra in Figure S5 Aa) and Ab), which in turn is the same dataset used to obtain Figure S6 a). The KM transformed spectra for the dataset obtained under Argon are shown in Figure S12, and results of linear fit through the data between 1000 and 2600 nm are shown in Table S2. The corresponding difference spectra (spectrum obtained under Argon subtracted by the spectrum obtained under air) for the KM transformed data are shown in Figure S13, and results of linear fit through the data between 1000 and 2600 nm are shown in Table S3. Finally, the log(1/r) transformed spectra for the dataset obtained under Argon and the Argon-air difference spectra are respectively shown in Figure S14 and Figure S15, with the results of linear fit through the data between 1000 and 2600 nm shown in Table S2 and Table S3, respectively.     The spectra in Figure 4 c) in the main text were normalised (Figure 4 d)) and plotted on a log-log scale in Figure S16. are respectively plotted as a function of wavelength in Figure S17 a), b), c), and d) for individual repeats. All traces in Figure S17 a) and c) were averaged to obtain the average traces in Figure 5 a) and c), respectively. Figure 5 b) correspond to trace 4 in Figure S17 b). Traces 1-3 in Figure S17 d) were averaged to obtain the average trace in Figure 5 d). Statistical analysis of the %abs ratios between 1200 and 2500 nm in Figure S17 a), c), and d) are respectively detailed in Table S4, Table   S5, and  and 4, respectively. The mean and SD of these four values are respectively 0.075 and 0.017, which is the origin of the estimate of 0.08±0.02 for 〈 % (( + )/ )〉 1200−2500 .