Article https://doi.org/10.1038/s41467-023-41976-2 Facet-engineered TiO2 drives photocatalytic activity and stability of supported noble metal clusters during H2 evolution Received: 21 October 2022 Yufen Chen 1,2, Lluís Soler 1,2 , Claudio Cazorla 3 , Jana Oliveras4, Neus G. Bastús4, Víctor F. Puntes4,5,6 & Jordi Llorca 1,2 Accepted: 25 September 2023 Metal clusters supported on TiO2 are widely used in many photocatalytic Check for updates applications, including pollution control and production of solar fuels. Besides 1234567890():,; 1234567890():,; high photoactivity, stability during the photoreaction is another essential quality of high-performance photocatalysts, however systematic studies on this attribute are absent for metal clusters supported on TiO2. Here we have studied, both experimentally and with first-principles simulation methods, the stability of Pt, Pd and Au clusters prepared by ball milling on nanoshaped anatase nanoparticles preferentially exposing {001} (plates) and {101} (bipyr- amids) facets during the photogeneration of hydrogen. It is found that Pt/TiO2 exhibits superior stability than Pd/TiO2 and Au/TiO2, and that {001} facet- based photocatalysts always are more stable than their {101} analogous regardless of the considered metal species. The loss of stability associated with cluster sintering, which is facilitated by the transfer of photoexcited carriers from the metal species to the neighbouring Ti and O atoms, most significantly and detrimentally affects the H2-evolution photoactivity. Anatase titania (TiO2) has been widely employed for photocatalytic has been revealed to be more efficient than the {101} facet9,10 because it production of hydrogen (H2) due to its suitable conduction and contains 100% 5-fold-coordinated titanium atoms (5c-Ti), which ren- valence band edges and superior photostability. Huge efforts, though, ders more active sites and stronger interactions with the adsorbates have been devoted to optimizing the photocatalytic performance of during the photocatalytic reactions11, whereas the {101} surface facet TiO2 by modifying its physical properties and/or blending it with metal possesses 50% 5c-Ti and 50% 6c-Ti5,12. Conversely, the {101} facet species1–4. It is generally accepted that the {101} TiO2 facet is thermo- showed higher photocatalytic H2 evolution than the {001} facet in an dynamically more stable than the {001} facet because of its lower aqueous methanol solution10,13; in this latter case, however, platinum average surface energy, 0.44 J/m2 and 0.90 J/m2, respectively5,6. The (1 wt%) was deposited onto the TiO2 samples as a co-catalyst. The co- photocatalytic activity of TiO2 is highly facet and shape dependent and catalyst plays an important role in the photocatalytic phenomenon, extensive studies have investigated the influence of specific surface not only by improving the overall photocatalytic performance, but also coordination environments on it7,8. by showing different trends in the H2 evolution rates depending on the Determining which anatase TiO2 facet is photocatalytically the facets exposed by TiO214. It has been demonstrated that the presence most active in fact remains contentious. For instance, the {001} facet of an optimum ratio of {101}/{001} facets in TiO2 anatase is capable of 1 Department of Chemical Engineering, Universitat Politècnica de Catalunya, Eduard Maristany 16, EEBE, Barcelona 08019, Spain. 2Institute of Energy Tech- nologies and Barcelona Research Center in Multiscale Science and Engineering, Universitat Politècnica de Catalunya, Eduard Maristany 16, EEBE, Barcelona 08019, Spain. 3Department of Physics, Universitat Politècnica de Catalunya, Campus Nord, B4-B5, Barcelona E-08034, Spain. 4Institut Català de Nanociència i Nanotecnologia (ICN2), CSIC and The Barcelona Institute of Science and Technology (BIST), Campus UAB, 08193 Barcelona, Spain. 5Institució Catalana de Recerca I Estudis Avançats (ICREA), 08010 Barcelona, Spain. 6Vall d’Hebron Research Institute (VHIR), Hospital Universitari Vall d’Hebron, Passeig de la Vall d’Hebron, 129, Barcelona 08035, Spain. e-mail: lluis.soler.turu@upc.edu; jordi.llorca@upc.edu Nature Communications | (2023)14:6165 1 Article https://doi.org/10.1038/s41467-023-41976-2 reducing the recombination of photogenerated carriers due to the dynamic conditions. Our study reveals that, during photoreaction, the formation of a heterojunction between the two different facets, which noble-metal clusters may either undergo substantial rearrangements drives the migration of electrons and holes to the {101} and {001} and coalescence or remain stable and practically maintain their as- facets, respectively15–17. This optimal {101}/{001} ratio is highly depen- synthesized morphology, depending on the TiO2 nanoshape. This dent on the studied reaction6,17–19. Taken together, these results sug- behaviour is rationalized by state-of-the-art first-principles simulations gest that the existing ambiguities seem to be caused by many different that effectively account for photoinduced electronic excitations in factors that come into play when evaluating the catalytic activity (e.g., crystals. Thus, the present work shows that the anatase TiO2 nanosh- distinct surface terminations, noble metals employed as co-catalysts14, ape mainly drives the stability and photocatalytic performance of and even different reactions). noble metal clusters in H2 production, a finding that has the potential To date, most efforts to improve the photocatalytic activity of to be generalised to other types of reactions and probably also to other TiO2 have been directed toward synthesizing highly exposed anatase supports (e.g., rutile TiO2 and CeO2). surfaces9,20 while stabilizing the shape21,22 of the crystal. It is well known that the photocatalytic activity of noble metal-loaded TiO2 heavily Results depends on the interactions between the co-catalysts and TiO2 sup- TEM images confirmed that anatase TiO2 plates and bipyramids were port, which are specific to the metal species and anatase surfaces23–26. succesfully synthesized (Supplementary Fig. 1). Both anatase nanosh- Therefore, to decipher the role of facets in a metal-supported het- apes exhibited {001} and {101} terminations, with plates preferentially erogeneous photocatalytic system, both the metals and facets should exposing {001} facets and bipyramids preferentially exposing {101} be carefully considered. Many previous studies have successfully facets (see the methods section for details). The size distribution is synthesized and evaluated various combinations of noble metals and shown in Supplementary Fig. 2; the average side length and thickness exposed TiO2 facets, yet another key factor has been largely neglected: of TiO2-001 (anatase plates) are 31.1 and 9.8 nm, respectively, and the stability of the metal nanoparticles during the photoreaction25,27. those of TiO2-101 (anatase bipyramids) are 12.3 and 23.6 nm, respec- Single metal atoms and clusters anchored on exposed anatase tively. XRD (Supplementary Fig. 3) and Raman spectroscopy (Supple- facets have been thoroughly analysed due to their abundant active mentary Fig. 4) demonstrated that the crystallinity and morphology of surface area, low-coordination environment and distinctive electronic TiO2 were preserved after ball milling and calcination (500 °C, 1 h)38,40. properties28,29. These ultra-small metal entities are prone to aggregate After the mechanochemical procedure, the specific surface area when irradiated with light under reductive conditions due to their high (SSABET, Supplementary Table 1) was measured to be, in both cases, ca. surface free energy, thereby reducing their surface-to-bulk atoms ratio 55 m2 g−1. and, in turn, photocatalytic performance30. While the agglomeration or sintering of ultra-small particles during high-temperature treatments Photocatalytic H2 evolution has been commonly interpreted in terms of Ostwald ripening and To evaluate the photocatalytic properties of the different metal particle diffusion31,32, the underlying causes of cluster coalescence in clusters-TiO2 architectures, 1 h, 4 × 1 h light on-off cycle experiments, photoreactions still remain obscure. Despite various attempts to iso- and 20 h (stability) tests of H2 production under UV light (365 ± 5 nm) late single atoms/clusters during photocatalytic processes, such as were independently performed (Fig. 1 and Supplementary Fig. 5). The zeolite confinement33, thiolate-protection34,35, and polymer scheme of the continuous flow photoreaction system is shown in stabilization36, there is little holistic understanding of the aggregation Supplementary Fig. 6; products were continuously monitored by phenomena occurring in different metal-supported systems under the micro-gas chromatography. The temperature of the photocatalysts same photocatalytic conditions. was monitored during the photoreaction and it was always below Here, we used a mechanochemical strategy to obtain different 54 °C. As shown in Fig. 1a, H2 was produced immediately after UV metal clusters supported on different TiO2 facets and investigated radiation was shed on the gaseous reactants, achieving different rates their long-term stability in photocatalytic hydrogen production. of H2 evolution depending on the noble metal and anatase nanoshape. Mechanochemical approaches represent a promising alternative to The Pt/TiO2-101 photocatalyst exhibited a remarkable performance conventional wet chemistry methods to produce supported single with a much higher value of H2 evolution than that of others. Mean- atoms and clusters37–39. In particular, Au, Pd and Pt clusters (1 wt%) while, the Pd/TiO2-101 photocatalyst showed negligible photoactivity were dispersed onto anatase plates or bipyramids preferentially enhancement compared to bare TiO2, whereas the Pt/TiO2-001 and Au/ exposing {001} or {101} facets, respectively (Supplementary Fig. 1) by a TiO2-101 samples behaved very similarly during the reaction. Overall, one-step ball milling procedure and their room-temperature photo- the rate of H2 production for the as-prepared photocatalysts catalytic performance on the hydrogen evolution reaction in gas phase decreased in the order: Pt/TiO2-101 ≫ Pd/TiO2-001 > Pt/TiO2-001 ~ Au/ from a water/ethanol mixture was systematically explored under TiO2-101 ≫ Au/TiO2-001 > Pd/TiO2-101 > TiO2-001 ~ TiO2-101. Fig. 1 | Photocatalytic performance of as-prepared photocatalysts. a 1 h test of photocatalytic H2 production. b Normalized H2 evolution rate for 20-h test. Nature Communications | (2023)14:6165 2 Article https://doi.org/10.1038/s41467-023-41976-2 Fig. 2 | XPS results of Au/TiO2, Pd/TiO2 and Pt/TiO2 photocatalysts. a–c XP Au 4f, Pd 3d and Pt 4f spectra of fresh Au/TiO2-001, Au/TiO2-101, Pd/TiO2-001, Pd/TiO2-101, Pt/TiO2-001 and Pt/TiO2-101 photocatalysts. d–f Corresponding ratios of metal oxidation states extracted from XP spectra in (a–c). Interestingly, an increasing trend of H2 evolution was observed for the experiments (below 54 °C)43. Evidence of the unstable Au was also Pt/TiO2 samples, which can be attributed to a progressive Pt reduction observed in a shell-like gold structure around dihexagonal pyramidal under photoreaction conditions. This phenomenon has been reported cadmium selenide nanocrystals44,45. On the other hand, metals sup- by Dessal et al.41, and later studied by operando X-ray absorption ported on anatase plates are always more stable than those on anatase spectroscopy by Piccolo et al.42. It is important to recall that the pho- bipyramids, regardless of the metal considered. toactivity trend may be influenced not only by the preferential facet exposed by anatase but also by a more complex situation derived from Characterization of photocatalysts {001}/{101} interface contributions. The oxidation states of Au, Pd and Pt clusters over different anatase Cycling experiments (Supplementary Fig. 5a) demonstrated good nanoshapes were clarified by X-ray photoelectron spectroscopy (XPS). reproducibility of all photocatalysts in short-term photocatalytic pro- Remarkable differences in the relative distribution of atomic fractions cesses, but long-term stability experiments (20 h) showed important for each supported metal were identified, as shown in Fig. 2. For Au/TiO2 differences among them (Supplementary Fig. 5b). For comparison photocatalysts, the fraction of oxidized Au (binding energy of 4f7/2 for purposes, we normalized the H2 evolution rates in Fig. 1b. Our results reduced Au: 83.1 eV; Au+: 84.2 eV; Au3+: 85.8 eV) was higher in the {101} reveal that Pt clusters loaded on anatase plates and bipyramids case than in the {001} surface. It has been reported that oxidized Au exhibited remarkable and identical durability. On the other hand, the species strongly interact with the surface of TiO2, thus accelerating Au/TiO2 photocatalysts showed poor stability of H2 production rates the e- transfer between the metal and support interface during for both anatase plates and bipyramids (especially in the bipyramids photocatalysis46. These results align with the higher H2 evolution rate case). Meanwhile, the stability of the Pd/TiO2-001 sample was in obtained for the Au/TiO2-101 compared to Au/TiO2-001. Interestingly, in between those of the Pt and Au systems (Pd/TiO2-101 was not con- the Pd/TiO2-001 sample, a large amount of reduced Pd (3d5/2 at 335.1 eV, sidered in the normalized analysis due to its negligible H2 production). 78.86%) was measured, whereas the dominant Pd species in Pd/TiO2-101 Based on these experimental findings, we may conclude that the was in the oxidized form (54.05% of Pd2+ at 336.3 eV and 37.35% of Pd4+ variability in the photocatalytic performance of functionalized anatase at 337.2 eV). This significant difference in the most abundant Pd oxi- does not solely respond to individual factors like the noble metal dation states can be directly ascribed to the distinct photocatalytic species or anatase nanoshape but also the interactions between them. performance observed for the {001} and {101} surfaces; the reduced Pd In terms of stability, the measured trend is Pt/TiO2-001 ~ Pt/TiO2- clusters have been proved to be more favourable for photocatalytic H2 101 > Pd/TiO2-001 > Au/TiO2-001 > Au/TiO2-101, from which it is clear production than the high-valent states38. In the case of Pt clusters, we that the metals exhibit different photocatalytic stability, namely, Pt/ observed that the {001} facet induces more Pt4+ (4f7/2 at 73.2 eV) than TiO2 ≫ Pd/TiO2 > Au/TiO2. These results suggest that minor rearran- the {101} facet, which is less active than the Pt0 (70.8 eV) and Pt2+ gements of the Pt species take place during the photocatalytic process (72.1 eV) species47. Based on the XPS results, we may conclude that whereas important migration of Au species occurs, being Pd species an different anatase nanoshapes induce different oxidation states in the intermediate case. Deactivation caused by metal diffusion into the metal clusters and that this effect depends on the metal species and TiO2 lattice can be safely excluded due to the low temperature of the determines the photocatalytic performance. XPS analyses were Nature Communications | (2023)14:6165 3 Article https://doi.org/10.1038/s41467-023-41976-2 Fig. 3 | HAADF-STEM images of fresh and after photoreaction samples. a, b Au/TiO2-001. c, d Au/TiO2-101. e, f Pd/TiO2-001. g, h Pd/TiO2-101. i, j Pt/TiO2-001. k, l Pt/TiO2- 101. Scale bar: 10 nm. The bright dots refer to the corresponding metal clusters in each photocatalyst. conducted for the post-reacted Au/TiO2-001, Au/TiO2-101, Pd/TiO2-001, corresponding cluster size distribution histograms are shown in Sup- Pd/TiO2-101, Pt/TiO2-001 and Pt/ TiO2-101 photocatalysts, as shown in plementary Figs. 13 and 14. The mean size of Au clusters in fresh Supplementary Fig. 7. In both Au/TiO2-001 and Au/TiO2-101, Au was Au/TiO2-001 and Au/TiO2-101 were about 1.5 nm and 1.6 nm, respec- reduced after long-term UV irradiation. The metallic Au species (Au0) in tively. They both underwent particle growth after photoreaction, Au/TiO2-001 and Au/TiO2-101 increased from 69.9% and 51.5% to 80.9% resulting in larger particles of about 1.8 nm and 3.3 nm, respectively. It and 69.4%, respectively. The oxidation states of Pd components in is evident that the Au species loaded on the {101} surface suffered from Pd/TiO2-001 exhibited similar proportions before and after photo- more intense agglomeration during the photoreaction among all cat- reaction, whereas Pd/TiO2-101 also underwent reduction, as was simi- alysts, which is perfectly consistent with the worst stability of Au/TiO2- larly observed for Pt/TiO2 photocatalysts. The change in the oxidation 101. The HRTEM images also show that the size of the Pd clusters on states of metal clusters during photocatalysis can be accounted for the the {001} facet increased slightly after photoreaction, in accordance reduction of the metal species induced by the photogenerated elec- with its moderate stability. By contrast, Pt clusters on both the {001} trons under the ethanol/water and UV irradiation conditions, as and {101} surfaces showed similar size distributions before and after observed in previous studies2,42. the photoreaction, according with their superior stability. Therefore, HAADF-STEM and HRTEM images were acquired to investigate deactivation can be correlated well with metal sintering, which, in turn, the morphological transformation of the supported metal clusters depends on the facet exposed by TiO2. from freshly synthesized to after 20-h photoreaction (Fig. 3 and Sup- plementary Fig. 8, and Fig. 4 and Supplementary Fig. 9, respectively). First-principles DFT calculations Enlarged HRTEM images of Pt/TiO2 samples are shown in Supple- To elucidate the mechanisms of metal-cluster stability on TiO2 sur- mentary Figs. 10 and 11 for a better visualization of the Pt clusters. faces, we performed first-principles calculations based on density Representative HAADF-STEM-EDX analysis for Pd/TiO2 (Supplemen- functional theory (DFT). We estimated the adsorption energy and tary Fig. 12) clearly shows the correlation between the metal and the agglomeration tendency of Au, Pd and Pt atoms and nanoclusters on structures identified as metal clusters in the samples. The HAADF- the {001} and {101} surfaces of anatase (Fig. 5a–c and Supplementary STEM and HRTEM images indicate that there is no preferential Fig. 15) with the finding that Pt renders the largest adsorption energy deposition of metal clusters on different titania facets, neither in the and lowest clustering tendency (especially in the {001} facet), followed fresh samples nor in the post-reacted ones. This is attributed to the dry by Pd and Au atoms. The adsorption energy of Pt and Au nanoclusters mechanochemical preparation method used, which, in contrast to wet was very similar to those obtained for individual atoms, thus hereafter chemistry methods48–51, does not yield any preferential deposition. The we reasonably extend the conclusions obtained for single atoms to Nature Communications | (2023)14:6165 4 Article https://doi.org/10.1038/s41467-023-41976-2 Fig. 4 | HRTEM images of fresh and after photoreaction samples. a, b Au/TiO2-001. c, d Au/TiO2-101. e, f Pd/TiO2-001. g, h Pd/TiO2-101. i, j Pt/TiO2-001. k, l Pt/TiO2-101. Scale bar: 5 nm. Metal clusters are indicated by arrows. nanoclusters. The ranking of adsorption energies perfectly correlates clearly indicates a tendency for Au and Pd agglomeration. Specifically, with the amount of charge that the metal atoms transfer from their a photoinduced change in the preferred adsorption sites of Au and Pd highly localized d orbitals to the anatase surface (Fig. 5b). The larger atoms on the {101} surface spontaneously occurred during the geo- the amount of exchanged charge the stronger the metal adsorption on metry optimizations (i.e., from a preferred “hollow” adsorption site TiO2 due to the appearance of attractive electrostatic forces on the under mild photoexcitation conditions to a preferred “O-top” surface and electronic hybridizations between d metal and p oxygen adsorption site under medium and intense photoexcitation condi- orbitals close to the Fermi energy level (Fig. 5c). These results suggest a tions, Fig. 5g, h). In these two cases, the total amount of charge decrease in metal-cluster stability according to the sequence Pt > Pd > exchanged between the metal atoms and TiO2 surface are minute (<0.1 Au when photoexcitation is absent. e- as referred to in dark conditions) and the most notable electronic To theoretically assess the stability of metallic clusters on TiO2 effects are intra-metallic s → p (Au/TiO2-101) d → s, p (Pd/TiO2-101) surfaces under photoexcitation conditions, we adopted a DFT charge redistributions. Pt/TiO2-101, on the other hand, behaves com- approach that allows for effective simulation of photoexcited pletely different from the analogous Au and Pd systems and very carriers52,53. For the {001} facet, it was found that the stability ranking of similarly to Pt/TiO2-001, namely, photoexcitation does not induce any nanoclusters was the same than determined under no illumination change on the preferred metal adsorption site and the amount of conditions, namely, Pt > Pd > Au. This result also can be understood in d charge donated to the anatase surface is substantial. Therefore, the terms of the amount of d charge donated by the metal atoms to the stability and photoactivity of Pt/TiO2-101 appear to be much superior anatase surface, which in all the cases increases upon the intensifica- to that of Au/TiO2-101 and Pd/TiO2-101, and similar to that of Pt/TiO2- tion of photoexcitation (Fig. 5d, e and Supplementary Fig. 16). In par- 001. These theoretical DFT results are in consistent agreement with ticular, Pt atoms transfer the largest amount of charge to their our experimental observations and reveal a central role played by neighbouring Ti and O atoms, thus enhancing the reactivity of the photoinduced metal-anatase charge transfer on the stability and H2 surrounding anatase surface (Fig. 5f). These theoretical results, production rate of TiO2-based photocatalysts. therefore, may explain in part the superior stability and H2 production Finally, we also performed first-principles DFT calculations to get rate observed for Pt-based photocatalysts. It is worth noting that insights into the possible causes of the differences in photocatalytic analogous results were obtained on Au nanoclusters adsorbed on the hydrogen activity experimentally observed for the Pt/TiO2-101 and {001} facet (Supplementary Fig. 17). Pt/TiO2-001 systems (Fig. 1 and Supplementary Fig. 18). Specifically, we As for the {101} facet, it was found that Au and Pd atoms can easily followed the computational strategies explained in the previous move on the TiO2 surface when photoexcited (Fig. 5g, h), a result that works54–57, since these allow to effectively estimate the performance of Nature Communications | (2023)14:6165 5 Article https://doi.org/10.1038/s41467-023-41976-2 Fig. 5 | First-principles simulations of noble metal blended anatase TiO2 based promoting electrons from the valence to the conduction band (Methods). f Charge on density functional theory. a DFT adsorption energies; (a) stands for single distribution difference analysis for Au and Pt atoms adsorbed on anatase TiO2 and atom and (c) for nanocluster. b Metal atoms charge donation; (d) stands for the density of excited electrons expressed as a function of the electronic Fermi highly localized electronic orbitals. c Partial density of electronic states for Au/TiO2- smearing. g–i Metal atoms charge donation under photoexcitation conditions on 001 and Pt/TiO2-001. d, e Metal atoms charge donation under photoexcitation the {101} facet; results are referred to non-photoexcitation conditions. The dashed conditions on the {001} facet; results are referred to non-photoexcitation condi- lines indicate photo-induced metal adsorption site transitions. tions and expressed as a function of the electronic Fermi smearing employed for catalytic materials for the hydrogen evolution (HER) and oxygen evo- anatase plates or bipyramids preferentially exposing {001} or {101} lution (OER) reactions through the calculation of adsorption free facets, respectively. We systematically investigated the photocatalytic energies, ΔG, of intermediate molecules involved in the oxidation of H2 evolution for different metal clusters/anatase hybrids, and the water and the reduction of hydrogen molecules (i.e., H*, O*, OH* ranking of H2 production rate was Pt/TiO2-101 ≫ Pd/TiO2-001 > Pt/ and OOH*). TiO2-001 ~ Au/TiO2-101 ≫ Au/TiO2-001 > Pd/TiO2-101 > TiO2-001 ~ TiO2- As it can be appreciated in the Supplementary Fig. 18a, in terms of 101. XPS analyses revealed that the {001} and {101} facets induce dif- HER performance the two Pt/TiO2-101 and Pt/TiO2-001 systems are ferent oxidation states of metal co-catalysts during the milling process, very much alike, and in principle very efficient, since the estimated thus affecting their photocatalytic performance. Furthermore, the proton adsorption free energies, ΔGH*, are quite small in absolute value long-term stability tests concluded a decreasing order of Pt/TiO2-001 ~ in both cases (i.e., 0.15–0.20 eV) and very similar in size to that of the Pt/TiO2-101 > Pd/TiO2-001 > Au/TiO2-001 > Au/TiO2-101. Specifically, archetypal HER catalyst Pt (111). On the other hand, in terms of OER the metal species behaved differently in terms of photocatalytic sta- activity (Supplementary Fig. 18b), Pt/TiO2-101 turns out to be much bility under same conditions, Pt/TiO2 ≫ Pd/TiO2 > Au/TiO2, and the superior than Pt/TiO2-001 since the OER overpotential estimated for anatase plates were always more stable than the anatase bipyramids the former system is much smaller and comparable in size to those of irrespective of the considered co-catalyst. HAADF-STEM and HRTEM archetypal OER catalysts like SrCoO3, RuO2 and PtO2 (i.e., of the order images unambiguously depicted the phenomena of metal species 0.1 V). Therefore, based on these DFT simulations, the likely cause for aggregation before and after 20 h of photoreaction. Moreover, theo- the observed highest photocatalytic efficiency in H2 production of Pt/ retical DFT simulations confirmed the crucial role of photoinduced TiO2-101 (Fig. 1) is that in terms of the oxygen evolution reaction (OER) metal-anatase charge transfers on the stability and H2 production rate this system is much more efficient than Pt/TiO2-001. of TiO2-based photocatalysts, which depends on the different metal/ In summary, a simple mechanochemical strategy was employed to facet interfaces, in consistent agreement with the experimental synthesize noble metal clusters (Au, Pd or Pt) supported on TiO2 observations. Nature Communications | (2023)14:6165 6 Article https://doi.org/10.1038/s41467-023-41976-2 Methods acetylacetonate was mixed with 0.2943 g, 0.2937 g or 0.2939 g of TiO2, Reagents respectively, leading to a ball to powder weight ratio (BPR) of 45 and Gold (III) acetate (Alfa Aesar, 99.9%), Palladium (II) acetate (Acros metal loading of 1 wt%. Milling was performed at 15 Hz for 10 min. No Organics, 47.5% Pd), Platinum (II) acetylacetonate (Acros Organics, signals of the precursors (acetate or acetylacetonate) were detected by 98%) were used as metal precursors. Titanium (IV) fluoride (TiF4, 99%), either Raman spectroscopy or XPS (C 1 s signal) in the samples pre- titanium (IV) chloride (TiCl4, 99%), 1-octadecene (90%) (1-ODE), pared. The resulting metal/TiO2 samples were denoted as Au/TiO2-001, 1-octadecanol (1-ODOL, 97%). Oleylamine (OLAM, 70%) and oleic acid Au/TiO2-101, Pd/TiO2-001, Pd/TiO2-101, Pt/TiO2-001 and Pt/TiO2-101. (OLAC, 90%) were obtained from Sigma-Aldrich. Absolute ethanol was The obtained photocatalysts were used for the photocatalytic experi- obtained from Scharlau. Milli-Q water was routinely used. ments without any further treatment. ICP-OES analyses (Supplemen- tary Table 2) showed that all samples contained a similar metal loading The preparation of TiO2 photocatalysts of 0.9 ± 0.1 wt%, which is similar to the theoretical value of 1.0 wt%. TiO2 nanoshapes exhibiting (001) and (101) facets were prepared fol- lowing a surfactant-assisted synthesis reported by Gordon et al.13. Photocatalysts characterization Highly monodisperse anatase tetragonal bipyramidal TiO2 nanocrys- X-ray diffraction (XRD) data were collected on a PANalytical X’Pert tals (NCs) exposing primarily {101} facets were produced using a mix- diffractometer using a Cu Kα radiation source (λ = 1.541 Å). In a typical ture of TiF4/TiCl4 precursors in the presence of OLAM as a co- experiment, the 2θ diffraction (Bragg) angles were measured by surfactant. Similarly, the use of 1-ODOL as a co-surfactant leads to TiO2 scanning the goniometer from 10° to 100°. The samples were prepared nanoplates with a high percentage of {001} facets. TiO2 NCs were by centrifugation to precipitate the NCs. The supernatant was dis- prepared as follows: 30 mmol of co-surfactant (OLAM or 1-ODOL), carded, and samples were dried at room temperature. Peak positions 10.2 mL of 1-ODE, and 0.48 mL (1.5 mmol) of OLAC were degassed at were determined using the X’Pert HighScore program after baseline 120 °C for 1 h. After, 1.5 mL of a mixed TiF4/TiCl4 stock solution 1:1 was correction. High-angle annular dark field scanning transmission elec- added at 60 °C and the solution was quickly heated to 290 °C. After tron microscopy (HAADF-STEM) and high-resolution transmission 10 min, 8 mL of the stock solution was pumped into the flask at electron microscopy (HRTEM) images were acquired on a FEI Tecnai 0.3 mL min–1 using a syringe pump. Afterwards, the heating mantle was G2 F20 instrument equipped with a field emission gun operating at an removed, and the flask was left to cool naturally to ambient tem- accelerating voltage of 200 kV. At least one hundred of Au, Pd and Pt perature. After the synthesis, a mixture of 2-propanol and methanol is particles in each photocatalyst were collected to estimate the size added to precipitate the NCs. Centrifugation at 4951 × g for 10 min is distribution. X-ray photoelectron spectroscopy (XPS) was conducted used to recover them. This washing process was repeated twice. For on a SPECS system. The apparatus was equipped with an XR-50 X-ray the mixture of TiF4/TiCl4 precursors, TiF4 and TiCl4 stock solutions source (Al anode at 150 W), a PHOIBOS 150 EP hemispherical energy were mixed at equal volume in the glovebox. TiF4 stock solution analyser and an MCD-9 detector. The spectra were corrected with the consists of 0.2 M TiF4 and 1.0 M OLAC in 1-ODE. TiCl4 stock solution position of the adventitious carbon 1 s signal at 284.8 eV. Thermo- consists of 0.2 M TiCl4 and 1.0 M OLAC in 1-ODE. The TiF4 stock solu- gravimetric analyses (TGA) were performed on the TA Instrument with tion is stirred on a hot plate set to 80 °C to promote the dissolution of a Q50 system under nitrogen atmosphere (10 °C/min). The actual TiF4. Once dissolved, the TiF4 stock solution is orange-brown, and the loading amount of Au, Pd and Pt was quantified by Inductively Coupled TiCl4 stock solution is dark brown. Plasma-Optical Emission Spectrometry (ICP-OES, Perkin Elmer Optima The TiO2 anatase suspensions were evaporated overnight at room 3200RL). Surface area measurements were conducted on an automatic temperature. The obtained anatase powders were analysed through Micromeritics ASAP 2020 analyzer using N2 adsorption isotherms and thermogravimetric measurements (TGA) to assess the temperature of BET (Brunauer−Emmett−Teller) surface area analysis methods. Sam- complete mass loss of residual surfactants on the surface (Supple- ples were degassed under vacuum at 200 °C for 4 h before adsorption mentary Fig. 19). Both materials were calcined in air at 500 °C for 1 h measurements. The average pore diameter distributions were derived (2 °C min−1) and the resulting materials were labelled as TiO2-001 (for from the desorption branches of the isotherms based on Barrett- plates exhibiting predominantly {001} facets) and TiO2-101 (for bipyr- Joyner-Halenda (BJH) model. amids exhibiting predominantly {101} facets). To facilitate an intuitive nomenclature of the samples that reflects the information of the pri- Photocatalytic hydrogen production marily titania facet exposed, we named the samples as TiO2-{hkl}. The The photocatalytic hydrogen evolution reaction was conducted in a XRD patterns (Supplementary Fig. 3) were used to confirm the for- 40 mL tubular glass photoreactor at room temperature and atmo- mation of anatase without any impurities or rutile phase. TEM images spheric pressure under dynamic conditions. An UV lamp equipped (Supplementary Fig. 1a, b) unambiguously showed the synthesized with four LEDs emitting at 365 ± 5 nm (SACOPA, S.A.U.) was employed shapes of anatase plates and bipyramids. At least one hundred TiO2 as the light source. An argon gas (Ar) stream (20 mL min−1) was passed nanoparticles in each sample were considered for estimating the par- through a saturator (Dreschel bottle), which contained a liquid mixture ticle size/shape distribution (Supplementary Fig. 2). On the basis of the of water (H2O, 150 g) and ethanol (EtOH, 17 g). The resulting gaseous obtained structural information, the percentages of preferentially mixture of H2O, EtOH and Ar was directly introduced into the photo- exposed (101) facets in TiO2 bipyramids and (001) facets in TiO2 plates reactor. 2.0 mg of each photocatalyst was dispersed in 0.5 mL absolute were estimated to be 91–97% and 50–76%, respectively. The detailed ethanol and treated by ultrasonication for 10 min to form a homo- calculations are shown in Supplementary Fig. 20. The Fourier Trans- geneous suspension. Afterwards, the slurry was dropped onto a cir- form (FT) images of fresh Pd/TiO2 provide direct evidence of crystal cular cellulose paper (from Albet, thickness 0.18 mm, area 2.54 cm2) planes of anatase {001} facets and {101} facets (Supplementary Figs. 1c and dried at 50 °C for 1 h. The impregnated paper was placed upside and 1d, respectively). down in the middle of the glass photoreactor with two separated tubular sections, along with an O-ring (Supplementary Fig. 6). The light The preparation of M-TiO2 photocatalysts source was positioned at the bottom of the photoreactor. By aligning a The mechanochemical preparation of the photocatalysts was per- synthetic quartz glass cylindrical lens from the light source to the formed in a high-energy RETSCH Mixer Mill MM200, using a 10 mL cellulose paper loaded with photocatalyst, 80.5 ± 0.5 mW cm−2 of UV stainless-steel vial and one stainless-steel milling ball (d = 15 mm, irradiance reached the sample. The cellulose paper was totally stable m = 13.54 g), as described elsewhere38,40. Briefly, 0.0057 g of gold during drying/wetting, UV-exposure and gas permeation. The tem- acetate, 0.0063 g of palladium acetate or 0.0061 g of platinum perature of the photocatalyst under the UV irradiation was monitored Nature Communications | (2023)14:6165 7 Article https://doi.org/10.1038/s41467-023-41976-2 directly with a K-type thermocouple in contact with the sample. The Gibbs free energies were estimated with the formula: gas hourly space velocity (GHSV) was 26,000 h−1 and the contact time was 0.14 s. Prior to the photoreaction, the system was purged with Ar 4GX = 4E X + 4E ZPE  T4S gas (20 mL min−1) for 30 min. The products evolved from the outlet of photoreactor (hydrogen and acetaldehyde in equimolar amount, where ΔEX corresponds to the molecular adsorption energy (that is, C2H5OH!H2 + CH3CHO) were analysed on-line every 4 min with a calculated under zero-temperature conditions) and ΔEZPE and ΔS are micro-gas chromatograph (GC, Agilent 490)58,59. the difference in zero-point energy and entropy, respectively, between the molecule absorbed on the anatase surface and in the correspond- First-principles simulation techniques ing gas phase. The contribution from the anatase catalysts to both We performed spin-polarized first-principles calculations based on ΔEZPE and ΔS are very small and hence were neglected. The value of the density functional theory (DFT) for anatase TiO2 surfaces functiona- ΔEZPE and ΔS terms were taken from the previous works56,57. lised with Au, Pd and Pt co-catalysts. The PBE functional60 was used as implemented in the VASP software package61. A “Hubbard-U” scheme62 Data availability with U = 3 eV was employed for a better treatment of the localized Ti d The data that support the findings of this study have been included in electronic orbitals (Hubbard-like corrections were not applied on the the main text and Supplementary Information. The first-principles noble metal atoms). The value of the lattice parameters, however, were density functional theory calculations presented in this study have constrained to their corresponding experimental values of been deposited in a public repository at https://dataverse.csuc.cat/ a0 = b0 = 3.78 Å and c0 = 9.51 Å63 since these are not correctly repro- dataset.xhtml?persistentId=doi:10.34810/data756. All other relevant duced by the PBE + U approach (Supplementary Methods and Sup- data supporting the findings of this study are available from the cor- plementary Fig. 21). We used the “projector augmented wave” method responding authors upon request. to represent the ionic cores64 by considering the following electrons as valence: Au 5d, and 6s; Pd 4d; Pt 5d and 6s; Ti 3d, 4s, and 3p; and O 2s and 2p. Wave functions were represented in a plane-wave basis trun- References cated at 650 eV. For integrations within the first Brillouin zone, a 1. Chen, X., Liu, L., Yu, P. Y. & Mao, S. S. Increasing solar absorption for Monkhorst-Pack k-point grid was employed with a density equivalent photocatalysis with black hydrogenated titanium dioxide nano- to 16 × 16 × 12 for the anatase unit cell. Geometry relaxations were crystals. Science 331, 746–750 (2011). performed by using a conjugate-gradient algorithm that allowed for 2. Lee, B. H. et al. Reversible and cooperative photoactivation of cell shape variations; the geometry relaxations were halted when the single-atom Cu/TiO2 photocatalysts. Nat. Mater. 18, 620–626 forces on the atoms fell all below 0.005 eVÅ−1. By using these technical (2019). parameters we obtained zero-temperature energies converged to 3. Hejazi, S. et al. On the controlled loading of single platinum atoms within 0.5 meV per formula unit. as a co-catalyst on TiO2 anatase for optimized photocatalytic H2 All the DFT geometry relaxations in the present work were per- generation. Adv. Mater. 32, e1908505 (2020). formed at the PBE + U level with U = 3 eV and by constraining the size of 4. Long, J. et al. Gold-plasmon enhanced solar-to-hydrogen conver- the lattice vectors to their experimental values; the positions of the sion on the {001} facets of anatase TiO2 nanosheets. Energy Environ. atoms, however, were allowed to fully relax in all the cases. Subse- Sci. 7, 973–977 (2014). quently, the energy, charge density, and optoelectronic properties of 5. Liu, G. et al. Titanium dioxide crystals with tailored facets. Chem. anatase TiO2 were estimated with the range-separated hybrid HSE06 Rev. 114, 9559–9612 (2014). exchange-correlation functional without performing further relaxa- 6. Cao, Y., Li, Q., Li, C., Li, J. & Yang, J. Surface heterojunction between tions (Supplementary Methods and Supplementary Fig. 21). (001) and (101) facets of ultrafine anatase TiO2 nanocrystals for The slab supercells were constructed as a 4 × 4 × 4 repetition of highly efficient photoreduction CO2 to CH4. Appl. Catal. B Environ. the anatase unit cell and a vacuum region of 25 Å thickness was con- 198, 378–388 (2016). sidered in all the simulations. Adsorption energies were calculated 7. Wei, T. et al. Facet-regulating local coordination of dual-atom with the formula Eads = EM@anatase – Eanatase – EM, where EM@anatase cocatalyzed TiO2 for photocatalytic water splitting. ACS Catal. 11, represents the energy of the blended noble metal-TiO2 system, Eanatase 14669–14676 (2021). the energy of the anatase slab, and EM the energy of the isolated noble 8. Jiang, M. P. et al. Magnetic-field-regulated TiO2 {100} facets: a metal atom or cluster. To estimate accurate energy band gaps, density strategy for C-C coupling in CO2 photocatalytic conversion. Chem of electronic states and charge redistributions at affordable compu- 6, 2335–2346 (2020). tational expense, we employed the hybrid HSE06 exchange- 9. Yang, H. G. et al. Anatase TiO2 single crystals with a large percen- correlation functional65 and adopted the equilibrium geometries tage of reactive facets. Nature 453, 638–641 (2008). determined at the PBE + U level. In order to simulate photoexcitation 10. Pan, J., Liu, G., Lu, G. Q. & Cheng, H. M. On the true photoreactivity effects in TiO2-based photocatalysts, we employed an effective DFT order of {001}, {010}, and {101} facets of anatase TiO2 crystals. approach, equivalent to those employed in previous works52,53, that Angew. Chem. Int. Ed. 50, 2133–2137 (2011). consists in constraining the partial occupancies of each electronic 11. Li, C. et al. Facet-dependent photoelectrochemical performance of orbital by adjusting the width of the corresponding Fermi smearing. In TiO2 nanostructures: an experimental and computational study. J. particular, a Fermi smearing of σ (eV) in the constrained DFT calcula- Am. Chem. Soc. 137, 1520–1529 (2015). tions implies considering the electronic occupation function f(E,σ) = 1/ 12. Katal, R., Masudy-Panah, S., Tanhaei, M., Farahani, M. H. D. A. & (exp[(E−EF)/σ]+1) instead of the usual step function f(E) = 1 for E ≤ EF Jiangyong, H. A review on the synthesis of the various types of and f(E) = 0 for E > EF, where EF represents the Fermi energy level. anatase TiO2 facets and their applications for photocatalysis. Chem. Regarding the theoretical estimation of the performance of cat- Eng. J. 384, 123384 (2020). alytic materials for the hydrogen evolution (HER) and oxygen evolu- 13. Gordon, T. R. et al. Nonaqueous synthesis of TiO2 nanocrystals tion (OER) reactions, we followed the computational strategies using TiF4 to engineer morphology, oxygen vacancy concentration, explained in the previous works54–57. These strategies rely on the cal- and photocatalytic activity. J. Am. Chem. Soc. 134, 6751–6761 culation of adsorption Gibbs free energies, ΔG, of intermediate mole- (2012). cules involved in the oxidation of water and the reduction of hydrogen 14. Pellegrino, F. et al. Formic acid photoreforming for hydrogen molecules (i.e., H*, O*, OH* and OOH*). The molecular adsorption production on shape-controlled anatase TiO2 nanoparticles: Nature Communications | (2023)14:6165 8 Article https://doi.org/10.1038/s41467-023-41976-2 assessment of the role of fluorides, {101}/{001} surfaces ratio, and 36. Weng, B., Lu, K. Q., Tang, Z., Chen, H. M. & Xu, Y. J. Stabilizing platinization. ACS Catal. 9, 6692–6697 (2019). ultrasmall Au clusters for enhanced photoredox catalysis. Nat. 15. Roy, N., Sohn, Y. & Pradhan, D. Synergy of low-energy {101} and Commun. 9, 1543–1553 (2018). high-energy {001} TiO2 crystal facets for enhanced photocatalysis. 37. Jin, H. et al. Simple and scalable mechanochemical synthesis of ACS Nano 7, 2532–2540 (2013). noble metal catalysts with single atoms toward highly efficient 16. Liu, L. et al. Engineering coexposed {001} and {101} facets in hydrogen evolution. Adv. Funct. Mater. 30, 4–9 (2020). oxygen-deficient TiO2 nanocrystals for enhanced CO2 photo- 38. Chen, Y. et al. Enhanced photoproduction of hydrogen on Pd/TiO2 reduction under visible light. ACS Catal. 6, 1097–1108 (2016). prepared by mechanochemistry. Appl. Catal. B Environ. 309, 17. Yu, J., Low, J., Xiao, W., Zhou, P. & Jaroniec, M. Enhanced photo- 121275 (2022). catalytic CO2-reduction activity of anatase TiO2 by coexposed {001} 39. Michalchuk, A. A. L. & Emmerling, F. Time-resolved in situ mon- and {101} facets. J. Am. Chem. Soc. 136, 8839–8842 (2014). itoring of mechanochemical reactions. Angew. Chem. Int. Ed. 61, 18. Lee, T. Y., Lee, C. Y. & Chiu, H. T. Enhanced photocatalysis from e202117270 (2022). truncated octahedral bipyramids of anatase TiO2 with exposed 40. Chen, Y. et al. A straightforward method to prepare supported Au {001}/{101} facets. ACS Omega 3, 10225–10232 (2018). clusters by mechanochemistry and its application in photo- 19. Liu, N. et al. (101)-(001) Surface heterojunction-enhanced anti- catalysis. Appl. Mater. Today 21, 100873 (2020). bacterial activity of titanium dioxide nanocrystals under sunlight 41. Dessal, C. et al. Influence of Pt particle size and reaction phase on irradiation. ACS Appl. Mater. Interfaces 9, 5907–5915 (2017). the photocatalytic performances of ultradispersed Pt/TiO2 catalysts 20. Fu, C. et al. Spontaneous bulk-surface charge separation of TiO2- for hydrogen evolution. J. Catal. 375, 155–163 (2019). {001} nanocrystals leads to high activity in photocatalytic methane 42. Piccolo, L. et al. Operando X‑ray absorption spectroscopy investi- combustion. ACS Catal. 12, 6457–6463 (2022). gation of photocatalytic hydrogen evolution over ultradispersed Pt/ 21. Yang, X. H., Li, Z., Sun, C., Yang, H. G. & Li, C. Hydrothermal TiO2 catalysts. ACS Catal. 10, 12696–12705 (2020). stability of {001} faceted anatase TiO2. Chem. Mater. 23, 43. Schubert, J. S. et al. Elucidating the formation and active state of Cu 3486–3494 (2011). co-catalysts for photocatalytic hydrogen evolution. J. Mater. Chem. 22. Luan, Y., Jing, L., Wu, J., Xie, M. & Feng, Y. Long-lived photo- A 9, 21958–21971 (2021). generated charge carriers of 001-facet-exposed TiO2 with 44. Meyns, M. et al. Growth and reductive transformation of a gold shell enhanced thermal stability as an efficient photocatalyst. Appl. around pyramidal cadmium selenide nanocrystals. J. Mater. Chem. Catal. B Environ. 147, 29–34 (2014). 20, 10602–10605 (2010). 23. Ohyama, J., Yamamoto, A., Teramura, K., Shishido, T. & Tanaka, T. 45. De La Cueva, L. et al. Shell or dots - precursor controlled mor- Modification of metal nanoparticles with TiO2 and metal-support phology of Au-Se deposits on CdSe nanoparticles. Chem. Mater. interaction in photodeposition. ACS Catal. 1, 187–192 (2011). 28, 2704–2714 (2016). 24. Al-Azri, Z. H. N. et al. The roles of metal co-catalysts and reaction 46. Yu, X. et al. A microorganism bred TiO2/Au/TiO2 heterostructure for media in photocatalytic hydrogen production: performance eva- whispering gallery mode resonance assisted plasmonic photo- luation of M/TiO2 photocatalysts (M = Pd, Pt, Au) in different alcohol- catalysis. ACS Nano 14, 13876–13885 (2020). water mixtures. J. Catal. 329, 355–367 (2015). 47. Liu, H. et al. One-step solvothermal formation of Pt nanoparticles 25. Wei, T., Zhu, Y., Wu, Y., An, X. & Liu, L. M. Effect of single-atom decorated Pt2+-doped α‑Fe2O3 nanoplates with enhanced photo- cocatalysts on the activity of faceted TiO2 photocatalysts. Langmuir catalytic O2 evolution. ACS Catal. 9, 1211–1219 (2019). 35, 391–397 (2019). 48. Wei, Z. et al. Noble metal-modified faceted anatase titania photo- 26. Zhou, P., Luo, M. & Guo, S. Optimizing the semiconductor–metal- catalysts: octahedron versus decahedron. Appl. Catal. B Environ. single-atom interaction for photocatalytic reactivity. Nat. Rev. 237, 574–587 (2018). Chem. 6, 823–838 (2022). 49. Zhang, Q. et al. The dependence of photocatalytic activity on the 27. Chen, Y. et al. Enhancement of photocatalytic performance with the selective and nonselective deposition of noble metal cocatalysts on use of noble-metal-decorated TiO2 nanocrystals as highly active the facets of rutile TiO2. J. Catal. 337, 36–44 (2016). catalysts for aerobic oxidation under visible-light irradiation. Appl. 50. Kobayashi, K., Takashima, M., Takase, M. & Ohtani, B. Mechanistic Catal. B Environ. 210, 352–367 (2017). study on facet-dependent deposition of metal nanoparticles on 28. Yang, X., Wang, A., Qiao, B. & Li, J. U. N. Single-atom catalysts: a new decahedral-shaped anatase titania photocatalyst particles. Cata- frontier in heterogeneous catalysis. Acc. Chem. Res. 46, lysts 8, 542 (2018). 1740–1748 (2013). 51. Liu, L. & Corma, A. Evolution of isolated atoms and clusters in cat- 29. Liu, J. Catalysis by supported single metal atoms. ACS Catal. 7, alysis. Trends Chem. 2, 383–400 (2020). 34–59 (2017). 52. Paillard, C., Torun, E., Wirtz, L., Íñiguez, J. & Bellaiche, L. Photo- 30. Rigby, K. & Kim, J. H. Deciphering the issue of single-atom catalyst induced phase transitions in ferroelectrics. Phys. Rev. Lett. 123, stability. Curr. Opin. Chem. Eng. 40, 100921 (2023). 87601 (2019). 31. Hansen, T. W., Delariva, A. T., Challa, S. R. & Datye, A. K. Sintering of 53. Peng, B., Hu, Y., Murakami, S., Zhang, T. & Monserrat, B. Topological catalytic nanoparticles: particle migration or Ostwald ripening? phonons in oxide perovskites controlled by light. Sci. Adv. 6, Acc. Chem. Res. 46, 1720–1730 (2013). eabd1618 (2020). 32. Yuan, W. et al. Direct in situ TEM visualization and insight into the 54. Liu, Z., Wang, B., Chu, D. & Cazorla, C. First-principles high- facet-dependent sintering behaviors of gold on TiO2. Angew. Chem. throughput screening of bulk piezo-photocatalytic materials for Int. Ed. 57, 16827–16831 (2018). sunlight-driven hydrogen production. J. Mater. Chem. A 10, 33. Liu, L. et al. Generation of subnanometric platinum with high sta- 18132–18146 (2022). bility during transformation of a 2D zeolite into 3D. Nat. Mater. 16, 55. Liu, Z., Wang, B. & Cazorla, C. Strain engineering of two-dimensional 132–138 (2017). piezophotocatalytic materials for improved hydrogen evolution 34. Abbas, M. A., Kamat, P. V. & Bang, J. H. Thiolated gold nanoclusters reaction. ACS Sustain. Chem. Eng. 10, 16924–16934 (2022). for light energy conversion. ACS Energy Lett. 3, 840–854 (2018). 56. Mom, R. V., Cheng, J., Koper, M. T. M. & Sprik, M. Modeling the 35. Xiao, F. X. et al. Light-induced in situ transformation of metal clus- oxygen evolution reaction on metal oxides: the infuence of ters to metal nanocrystals for photocatalysis. ACS Appl. Mater. unrestricted DFT calculations. J. Phys. Chem. C. 118, 4095–4102 Interfaces 7, 28105–28109 (2015). (2014). Nature Communications | (2023)14:6165 9 Article https://doi.org/10.1038/s41467-023-41976-2 57. Man, I. C. et al. Universality in oxygen evolution electrocatalysis on Author contributions oxide surfaces. ChemCatChem 3, 1159–1165 (2011). Y.C., L.S., J.O., N.G.B., V.F.P. and J.L. designed and performed the 58. Martínez, L., Soler, L., Angurell, I. & Llorca, J. Effect of TiO2 experimental studies. C.C. performed the computational studies. L.S. nanoshape on the photoproduction of hydrogen from water- and J.L. supervised the work. ethanol mixtures over Au3Cu/TiO2 prepared with preformed Au-Cu alloy nanoparticles. Appl. Catal. B Environ. 248, 504–514 (2019). Competing interests 59. Xing, C. et al. A direct Z-scheme for the photocatalytic hydrogen The authors declare no competing interests. production from a water ethanol mixture on CoTiO3/TiO2 hetero- structures. ACS Appl. Mater. Interfaces 13, 449–457 (2021). Additional information 60. Perdew, J. P. et al. Restoring the density-gradient expansion for Supplementary information The online version contains exchange in solids and surfaces. Phys. Rev. Lett. 100, 1–4 (2008). supplementary material available at 61. Kresse, G. & Furthmüller, J. Efficient iterative schemes for ab initio https://doi.org/10.1038/s41467-023-41976-2. total-energy calculations using a plane-wave basis set. Phys. Rev. B - Condens. Matter Mater. Phys. 54, 11169–11186 (1996). Correspondence and requests for materials should be addressed to 62. Dudarev, S. L., Botton, G. A., Savrasov, S. Y., Humphreys, C. J. & Lluís. Soler or Jordi Llorca. Sutton, A. P. Electron-energy-loss spectra and the structural stabi- lity of nickel oxide: an LSDA+U study. Phys. Rev. B 57, 1505–1509 Peer review information Nature Communications thanks the anon- (1998). ymous reviewers for their contribution to the peer review of this work. 63. Jeremy, K. et al. Structural-electronic relationships in inorganic A peer review file is available. solids: powder neutron diffraction studies of the rutile and anatase polymorphs of titanium dioxide at 15 and 295 K. J. Am. Chem. Soc. Reprints and permissions information is available at 109, 3639–3646 (1987). http://www.nature.com/reprints 64. Blöchl, P. E. Projector augmented-wave method. Phys. Rev. B 50, 17953–17979 (1994). Publisher’s note Springer Nature remains neutral with regard to jur- 65. Krukau, A. V., Vydrov, O. A., Izmaylov, A. F. & Scuseria, G. E. Influ- isdictional claims in published maps and institutional affiliations. ence of the exchange screening parameter on the performance of screened hybrid functionals. J. Chem. Phys. 125, 224106 (2006). Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, Acknowledgements adaptation, distribution and reproduction in any medium or format, as This work was supported by projects MICINN/FEDER PID2021- long as you give appropriate credit to the original author(s) and the 124572OB-C31 and GC 2021 SGR 01061. Y.C. (CSC No. 201806920042) source, provide a link to the Creative Commons license, and indicate if acknowledges the China Scholarship Council for Ph.D. scholarship changes were made. The images or other third party material in this support. L.S. and C.C. are grateful to MICINN Ramon y Cajal program article are included in the article’s Creative Commons license, unless for individual fellowship grant agreements RYC2019-026704-I and indicated otherwise in a credit line to the material. If material is not RYC2018-024947-I, respectively. J.L. is a Serra Húnter Fellow and is included in the article’s Creative Commons license and your intended grateful to ICREA Academia program. N.G.B. and V.P. acknowledge use is not permitted by statutory regulation or exceeds the permitted financial support from RTI2018-099965-B-I00, AEI/FEDER,UE and use, you will need to obtain permission directly from the copyright 2017-SGR-1431. ICN2 is supported by the Severo Ochoa program from holder. To view a copy of this license, visit http://creativecommons.org/ Spanish MINECO (SEV-2017-0706) and is funded by the CERCA Pro- licenses/by/4.0/. gramme/Generalitat de Catalunya. The authors thank E. Molins, I. Matas and M. Benito from ICMAB to kindly analyze BET areas. © The Author(s) 2023 Nature Communications | (2023)14:6165 10