Synergistic ultraviolet and visible light photo-activation enables intensified low-temperature methanol synthesis over copper/zinc oxide/alumina

Although photoexcitation has been employed to unlock the low-temperature equilibrium regimes of thermal catalysis, mechanism underlining potential interplay between electron excitations and surface chemical processes remains elusive. Here, we report an associative zinc oxide band-gap excitation and copper plasmonic excitation that can cooperatively promote methanol-production at the copper-zinc oxide interfacial perimeter of copper/zinc oxide/alumina (CZA) catalyst. Conversely, selective excitation of individual components only leads to the promotion of carbon monoxide production. Accompanied by the variation in surface copper oxidation state and local electronic structure of zinc, electrons originating from the zinc oxide excitation and copper plasmonic excitation serve to activate surface adsorbates, catalysing key elementary processes (namely formate conversion and hydrogen molecule activation), thus providing one explanation for the observed photothermal activity. These observations give valuable insights into the key elementary processes occurring on the surface of the CZA catalyst under light-heat dual activation.

comparison between as-prepared CZA and reduced CZA sample. UV-vis absorption spectra provide key information on the photo-excited regions for the different components. Light with energy E > 3.2 eV (wavelength λ< 390nm) is sufficient to activate ZnO via band-gap excitation (ZnO absorption in CZA is masked by the strong metallic Cu absorption). The Cu localized surface plasmon resonance (LSPR) is located at around 560 nm. A strong damping of Cu plasmon is observed in the Cu-based catalysts due to its resonance with the Cu interband transition region 2 (~590 nm, appears as a strong "background" absorption next to plasmon resonance peak). The absorption of CuO and Cu 2 O (wavelengths of 826 nm and 516 nm, respectively) also resides on either side of the Cu LSPR peak.  Supplementary  Fig. 13a); however, under light irradiation (350-800 nm) the surface reaction was accelerated with steady state attained in 50 min ( Supplementary Fig. 13b). With reference to the steady-state DRIFTS spectra collected at 50 °C, a higher MeOH concentration during reaction and a lower concentration of surface intermediates are apparent with light irradiation at 225 °C ( Supplementary Fig.13c), which suggests that formate hydrogenation step was accelerated and a "new" reaction equilibrium was accordingly achieved during light-assisted MeOH production. Similarly, CO formation was shown to be accelerated on the Cu surface under light irradiation ( Supplementary Fig. 13d). Based on the peak separation information provided in Biesinger's study, we first conduct curve-fitting in reference Cu LMM spectra with fixed sub-peak positions relative to the one with highest kinetic energy (HKE) (take Cu 2 O for example, HKE-3.79, HKE-4.88, HKE-7.58, and HKE were taken as the sub-peak positions, and position of HKE might change in different samples) and fixed range of FWHM parameter (initially 0.5-3.5, adjusted to 0.5-5.5 when necessary). Cu, Cu 2 O, and CuO species were deconvoluted into seven, four, and four sub-peaks, respectively. Good fitting results were obtained with reference spectra as shown in Supplementary Figs. 14a-c. Afterwards, the peak height relations were recorded for further curve-fittings of sample spectra. (taken Cu 2 O species again for example, the height for the above-mentioned four sub-peaks is I×0.143, I×0.928, I, I×0.344, respectively) Finally, the Cu speciation was accomplished with the deconvolution of Cu LMM spectra with fixed peak separation, relative peak height ratio, and FWHM range. Due to a large number of curves involved in the Cu LMM analysis, Cu 2p spectra ( Supplementary Fig. 14d) were always taken as a reference to cross-check the [Cu(I)+Cu(0)]/Cu(II) ratio.  1f-j), ZnO nanoparticles were highly aggregated and interconnected with small-sized Cu species, which is supporting the dynamic feature of the particles and the migration of ZnO to the Cu surface under reaction conditions. In addition, the interactions among all species seem to be promoted since particle boundaries can hardly be distinguished and a porous cluster morphology, which may represent a stable structure of active catalyst was finally obtained. After introducing light into the system ( Supplementary Fig. 3), there is a reconstruction (particle jointing) in the microstructure of the catalyst (denoted by red arrow in Supplementary  Fig. 3h).
The presence of Cu metal and wurtzite ZnO were identified in the XRD patterns of the reduced sample ( Supplementary Fig. 2a). Al2O3 exhibits amorphous-like structural features.
Overlapping of the Cu2O and ZnO peaks makes it difficult to determine the presence of Cu2O in the XRD pattern. The small amount of preserved Cu oxide on the Cu particle surface is thought to be vital for the optimized chemical properties of the Cu-ZnO interface during CO2 hydrogenation 13 . Compared to the Al-free catalyst (CZ), XRD pattern showed that alumina incorporation promotes an interaction between the different metal components, which is supported by the presence of weak aluminates 14 peaks in the XRD patterns of CA and CZA. The CO2-TPD profiles ( Supplementary Fig. 2b) show that, without ZnO, CA provides limited CO2 adsorption sites, while the presence of defective ZnO favours CO2 adsorption. A downward shift the in CO2 desorption temperature by at least 150 °C is observed when comparing the profiles of ZA and CZA, where the desorption peak across the 400-450 °C region in ZA is shifted to 250 °C in CZA. The decrease in temperature is believed to arise from the destabilisation of CO2 on the ZnO surface in the presence of Cu. This is essential for MeOH synthesis whereby the destabilisation of surface adsorbed CO2 at low temperatures is critical for controlling the selectivity towards MeOH. Interaction between the Cu and ZnO is also apparent from the binding energy (BE) shifts in Cu 2p (negative-shift) and Zn 2p/O 1s (positive-shift) spectra ( Supplementary Figs. 1c, d). This provides evidence of electron transfer from the ZnO to Cu in the CZA catalyst 15 . As shown in Fig. 1a-c and Supplementary Fig. 1

Supplementary Note 2 Selectivity vs. Temperature and activation energy for RWGS reaction:
Under the dark condition, the selectivity toward MeOH decreased due to promotion of the RWGS reaction as temperature increased. This can be correlated to the exothermic and endothermic natures of MeOH production and the RWGS reaction, respectively. Under illumination, CO production was enhanced across the considered temperature range ( Supplementary Fig. 8c), while a significant increase in MeOH-production was only detected at 200-250 °C. As illustrated in Supplementary Fig. 11, this was attributed to the temperaturedependency of the formate conversion rate. Finally, a slightly altered although similar MeOH selectivity vs. temperature trend (when compared to the dark condition) was observed depending on the degree of light promotion in CO and MeOH production.
The calculated apparent activation energy for the RWGS reaction is shown in Supplementary  Fig. 8d. A decrease in the apparent activation energy from 15.1 to 11.9 kcal/mol was observed under the irradiated condition. As revealed by Robatjazi et al. 16 , generated electrons on accessible unpopulated adsorbate orbitals can activate the CO2 reactant by either forming transient negatively charged ions or polarised species 16,17 , thus lowering the activation energy for the overall RWGS reaction 16,18,19 . In this study, a similar mechanism, where the electrons generated from Cu LSPR (or potentially transferred from excited ZnO) transiently occupy the orbital states of surface CO2 to promote its activation and subsequent processes before decaying into thermal energy into the lattice, is used to explain the observed CO production performance under light irradiation ( Supplementary Fig. 8c). In addition, CO production was found to be promoted across the considered temperature range under irradiation, which differs from MeOH production.

Supplementary Note 3 Contributions from photothermal heating and photocatalytic processes:
As elucidated in the main text, the photothermal CO2 hydrogenation over Cu/ZnO/Al2O3 is essentially a light-assisted thermocatalysis where photochemical processes are synergistically contributing to the methanol synthesis. Absorbed photon energy (visible light) via the Cu LSPR can remain trapped inside the metal nanostructure and cause local heating of the Cu metal lattice surroundings, which may contribute to the CO2 hydrogenation reaction. However, while the contribution from localised photothermal heating cannot be completely disregarded, to evaluate the extent of localised heating a control experiment where the reduced CZA sample was irradiated under the same reaction conditions (illumination without external heating) for two hours was conducted. There was a small temperature increase from 24 °C to 26 °C, indicating that heat generation within catalyst powder due to illumination is insignificant. Additionally, when comparing the MeOH yield and selectivity at 225 °C under both the dark (non-illuminated) and illuminated (350-800 nm light) conditions, while there is an enhanced MeOH yield the MeOH selectivity remains essentially unchanged. This implies that the enhanced MeOH yield under irradiation primarily derives from the photocatalytic process and not thermal effect. If the thermal effect was having a significant influence, any attributable increase in MeOH yield would be at the expense of MeOH selectivity which is not the case here.
The near-linear dependence of the MeOH production rate on light intensity is an indicator of a charge-carrier-driven 17, 20 reaction which differs from a thermally-driven reaction. The relationship suggests that the excited electrons can interact efficiently with surface adsorbates and a photocatalytic mechanism is responsible for the observed photo-enhancement. However, the distinctive roles of electrons from Cu LSPR-excitation and ZnO bandgap excitation in promoting the CO2 hydrogenation reaction over Cu/ZnO/Al2O3, and the interplay between the two remain unknown.

Supplementary Note 4 Discussion on reaction mechanism over different catalysts:
The distinctive MeOH-production capacity reflected in the DRIFTS spectrum (predominantly at 1031 cm -1 in Supplementary Fig. 9) and the activity results ( Supplementary Fig. 8) for CA, ZA (ZA show negligible methanol performance), and CZA can be accounted for by three different reaction pathways on three different catalysts: (i) CA -a minor amount of carbonate, formate, and methoxy can form on the Cu surface. The limited CO2 adsorption sites and ineffective methoxy formation on the pure Cu surface deliver a poor MeOH production capability compared to CZA; (ii) ZA -MeOH preferably forms from the strongly reducing atmosphere (CO/H2 mixture) and CO2 is thought to be a catalyst poison as it can potentially oxidise the active sites (O vacancies in ZnO1-x). 21 In addition, the hydrogenation of Zn-formate to methoxy has been reported as a very slow surface reaction 6 . This may account for the negligible MeOH production/methoxy formation on ZA; (iii) CZA -the relatively better MeOH-production benefits from the presence of the Cu-ZnO interfacial interaction, facilitating the formation of formate and methoxy species. These two are widely deemed as stable active intermediates for MeOH generation by Cu-containing catalysts. 7,11,22,23,24 In terms of CO production, the CO can form on both the ZnO and Cu surface (gaseous phase CO at 2113, 2176 cm -1 for the three samples). Two possible pathways are available for CO formation: (i) CO forms mainly on the Cu surface from the dissociation of Cu-CO3*, as a distinct peak is presented at 2077 cm -1 and 2094 cm -1 indicating adsorbed CO* species on the CA and CZA samples, respectively. The peak intensity of gaseous CO decreases in the following order: CA>CZA>ZA, indicating that Cu alone favours CO-production, as is consistent with a higher CO selectivity (lower MeOH selectivity) observed for the CA sample ( Supplementary Fig. 8b); (ii) The second pathway for CO production follows the decomposition of ZnO-HCOO*on the ZnO surface 6 , supported by the presence of a weak peak at 2164 cm -1 (the adsorption of CO on ZnO is weak). Supplementary Fig. 11 illustrates the following temperature-dependent characteristics of the CO2 hydrogenation reaction: i) At T <150 ºC, CO2 is chemically adsorbed on ZnO in the form of bicarbonate or carbonate depending on temperature, which can then convert to formate, while its further hydrogenation is kinetically unfavourable; ii)

Supplementary Note 5
At the onset temperature of methanol-production (200 °C), formate species begin to slowly convert to carbonyl hydrides (formaldehyde, methoxy, etc.) and then to methanol. All the active species (including carbonate) are observable due to a very low overall reaction rate; iii) At temperature in the range of 225-250 ºC, formate accumulates on the catalyst surface and its conversion represents the rate-limiting step in methanol synthesis. No carbonyl hydrides are observed in the DRIFTS spectra. CO2 is chemically bonded to the ZnO surface in the form of carbonate due to its higher thermal stability compared to bicarbonate and hydroxyl groups. The attached formate species can be activated and hydrogenated with the aid of light-generated electrons; iv) At temperatures greater than 250 ºC, the energy supplied from heating alone is sufficient to overcome the activation energy barrier of formate conversion. At such a high temperature, the thermal catalytic reaction is dominant. The observed temperaturedependence of the photocatalytic performance could be related to kinetic differences of formate conversion across the temperature range studied. Formate appeared as the dominant surface species over CZA at lower temperatures (Fig. 2b). However, the amount of formate is temperature-dependent and its conversion represents one of the kinetically-limiting steps at 200-250 ºC, which could be promoted through electronic transitions in the photo-excited adsorbate-nanoparticle systems 25,26 .
On the other hand, adsorbed CO was widely detected on the Cu-containing catalyst from 50-400 ºC (Supplementary Fig. 9) with photo-induced CO production clearly demonstrated across the full temperature range ( Supplementary Fig. 8c), indicating a non-volcanic CO production behaviour, contrary to MeOH synthesis.

Supplementary Note 6 Valence band/DFT alignment:
The XPS valence band spectra provide the valence band electronic states of Cu and ZnO. The valence band electronic states between 4.0 eV to 8.0 eV mainly originate from the O 2p orbitals of ZnO 27 , while the peak at ~10.5 eV is attributed to Zn 3d in Zn or ZnO 27 . Al2O3 valence band features were not indicated as they completely overlap with the ZnO valence band 28 and Al2O3 only contributes up to 10 % of the CZA catalyst with a minor contribution to the overall catalytic activity.
DFT calculations have been performed to determine the molecular orbital energy levels of various possible reaction intermediates. The molecular orbital energy levels are aligned against the Cu 3d DOS and the XPS VB spectra of CZA to identify the possible catalyst active sites that favour electron interaction between CZA catalyst surface and the surface adsorbed intermediate species.
In terms of chemical bond formation, while valence electrons are the ones involved, electronic interactions usually occur between valence band electrons (ground state) excited into the conduction band (excited state) of the catalysts and those in the higher energy molecular orbitals of the reactant or intermediate species. Chemical bond cleaving occurs when electrons are injected into the antibonding orbitals of the adsorbate molecules to achieve zero bond order. However, theoretical calculations to determine the energy levels of excited electrons for catalysts are non-trivial, especially when the catalysts are made up of multiple components of various structures and crystal phases. Clearly, this approach has its limitations, since only relatively rudimentary comparisons may be made by considering the gas phase adsorbate orbital levels unbound to the substrate. Given the structural complexity inherent in dealing with a multicomponent substrate, devising a suitable model substrate to explore the adsorption of reactants, products, and intermediates is not at all trivial and poses a major challenge for the computational modelling of such catalysts 29 . Hence, valence band analysis is used as an indicator to indirectly probe the catalyst behaviour and its electronic interactions with reactant and intermediate species.
Initial inspection of Fig. 3 reveals that for all of the reactants, products and intermediates, the lowest unoccupied molecular orbital (LUMO) levels all lie approximately within the range of the highest CZA valence bands below the Fermi level, implying that occupation of these orbitals is possible and thus implying a possible binding interaction with the substrate, in agreement with the observed surface species elucidated from the experimental studies. Closer inspection of the alignment of unoccupied levels of key adsorbates with substrate bands can further reveal insights into a potential mechanism of activation of reactant species.
From an electron and energy level perspective, electrons at the same energy level will be promoted by the same extent energetically upon excitation by the same amount of energy. Therefore, valence band electrons with a similar energy level as the critical molecular orbitals of reactant species can be promoted to the antibonding orbitals of the reactant species upon excitation, allowing electronic interactions between the catalyst and the reactant that lead to bond formation between the catalyst surface and the reactant (orbital overlapping), while weakening or breaking the native bonds of the reactant molecule.
By aligning the Fermi levels of the CZA catalyst valence band spectrum, Cu 3d orbitals PDOS, and the molecular orbitals of reactant species and intermediates, it is apparent that the majority of the species either interact with ZnO O 2p (~4-8 eV), Zn 3d (~10 eV), and Cu 3d (~2-4 eV), in particular the d x 2 -y 2 and d z 2 orbitals (~3 eV) for Cu 3d. The electronic interaction between ZnO/Cu and the CO2 antibonding orbitals, as a result of exotic ZnO defects and photoexcitations, offers an explanation for the observed extent of CO2 activation, and the subsequent increase in catalytic performance, i.e. ZA < CZA < CZA + mono-excitation < CZA + dual-excitation.

Supplementary Note 7
The deficiency in lattice oxygen and formation of chemisorbed carbon/oxygen groups on ZnO are predictable under CO2/H2 (1:3 gas flow ratio) reaction conditions. When compared with untreated ZnO/Al2O3 (ZA), the reduced and spent CZA catalysts displayed an increase in the intensity of the peak at around 991 eV in the Zn Auger spectra (Supplementary Fig. 15a). The intensity increase may originate from either oxygen vacancies or metallic Zn 30 . The presence of metallic Zn was reported by Behrens et al. 13,31 to be unlikely during methanol-synthesis (at around 250 ºC) and is supported by the Zn 2p spectra (Fig. 4b) (ZnO (1021.7 eV) vs. Zn (1021.4 eV)). A richness in oxygen vacancy (OV) and surface oxygen groups (OS, hydroxyl groups and carbonates, for example) 32 is apparent in the reduced and spent CZA catalyst in comparison with the untreated ZA. A small increase in the oxygen vacancy ratio in spent catalyst illuminated catalyst could be used to explain the observed changes in adsorption feature A and B since the hybridization between Zn 4p/4sp and O1s is dependent on the neighboring coordination environment.