Abstract text (incl. figure legends and references)

In this study, we investigate the stability and activity of Pt single atoms (SAs) deposited on TiO2 via a facile dark-deposition approach [1]. We aim to identify stable and active Pt SAs on both anatase nanosheets and ~7 nm thick polycrystalline anatase films sputtered on electron transparent SiO2 TEM grids to get an atomistic understanding on the SA anchoring as well as their stability upon photocatalytic water splitting. To identify stable species, anatase nanosheets where loaded with Pt SAs and further step-by-step leached by cyanide to reduce the SA loading while monitoring the photocatalytic activity as well as the atomistic configuration by aberration corrected high-angle annular dark field scanning transmission microscopy (STEM) [2]. In addition, we exploit the TiO2 model thin films in combination with identical location STEM investigations to visually follow the photoinduced destabilization over 24 h of operation time. Here, a machine learning-based algorithm is used to segment SAs as well as their agglomerates at various regions of interest. Within this approach we get quantitative information about the abundance and arrangement of the SAs as function of photocatalytic operation time.

We show that neither the majority of anchored Pt SAs nor metallic Pt0 species are responsible for the overall observed high photocatalytic activity. In fact, >90% of the deposited Pt can be removed by cyanide leaching without significant loss in photocatalytic activity (Fig. 1). In this way, we could identify an optimized SA loading quantity for anatase nanosheets, which is accompanied by a remarkable turnover frequency of ~4.87 x 105 h-1 for H2 evolution.

Identical location microscopy (see Fig. 2) after five different photocatalytic operation times reveals that the majority of initial deposited Pt is in a SA or multimere state. Within 1 min, ~90% of the SAs and multimeres agglomerate rapidly to 2D rafts. This agglomeration process continues, however, at a much slower speed within the next 24 h. Here, mainly the 2D rafts coarsen to nanoparticles with an average size of ~4 nm, whereas the photocatalytic activity stays unchanged and still 7% of the Pt loading maintains its SA state. In addition, the observed SA agglomeration is heavily influenced by the presence of methanol as a hole scavenger, elucidating the role of charge carriers on the stability of Pt SAs on TiO2. Moreover, identical location microscopy allows to follow various additional phenomena, like dissolution of 2D rafts, as well as morphological changes and ripening of agglomerated Pt species revealing the high dynamics in the system under photocatalytic conditions.

In conclusion, high resolution STEM on nanostructures and thin model systems is an ideal tool for studying Pt SA photocatalysts, their local arrangement as well as stability after different operation times. The correlation with photocatalytic data as well as spectroscopy (e.g. XPS, XAAS) allows to establish structure property relationships further improving the material efficient SA approach for photocatalytic applications.

[1] Cha, G., Mazare, A., Hwang, I., Denisov, N., Will, J., Yokosawa, T., ... & Schmuki, P. (2022). Electrochimica Acta, 412, 140129.

[2] Qin, S., Denisov, N., Will, J., Kolařík, J., Spiecker, E., & Schmuki, P. (2022). Solar RRL, 2101026.