Photoelectrochemical CO2-to-fuel conversion with simultaneous plastic reforming

Solar-driven conversion of CO2 and plastics into value-added products provides a potential sustainable route towards a circular economy, but their simultaneous conversion in an integrated process is challenging. Here we introduce a versatile photoelectrochemical platform for CO2 conversion that is coupled to the reforming of plastic. The perovskite-based photocathode enables the integration of different CO2-reduction catalysts such as a molecular cobalt porphyrin, a Cu91In9 alloy and formate dehydrogenase enzyme, which produce CO, syngas and formate, respectively. The Cu27Pd73 alloy anode selectively reforms polyethylene terephthalate plastics into glycolate in alkaline solution. The overall single-light-absorber photoelectrochemical system operates with the help of an internal chemical bias and under zero applied voltage. The system performs similarly to bias-free, dual-light absorber tandems and shows about 10‒100-fold higher production rates than those of photocatalytic suspension processes. This finding demonstrates efficient photoelectrochemical CO2-to-fuel production coupled to plastic-to-chemical conversion as a promising and sustainable technology powered by sunlight. A versatile solar-driven hybrid photoelectrochemical platform has been developed for the simultaneous conversion of greenhouse gas CO2 and waste plastics into value-added fuels and chemicals with high efficiency and selectivity.


Catalyst design and assembly
The versatility and product tunability of our PEC system for CO 2 R depends on the nature of the catalyst, for which we chose three different types of systems. The CoP L molecular catalyst can electrocatalytically reduce CO 2 to CO with a high selectivity (~90%) over a broad potential range 29 . The inorganic bimetallic Cu 91 In 9 alloy operates as a selective CO-forming catalyst at a low overpotential range (-0.3 to -0.5 V versus RHE), but produces a mixture of CO and H 2 (syngas) at more negative potentials 23,30 . The enzyme FDH reduces CO 2 to formate at its W active site with a high selectivity over a wide range of potentials 18,31 .
The CoP L molecular catalyst was synthesized and characterized following a previously reported procedure ( Supplementary Fig. 1a) 29 and then immobilized on an activated graphite foil substrate using multiwalled carbon nanotubes (MWCNT, geometrical surface area of 0.84 cm 2 ; see Methods for details). The surface coverage of the active molecular catalyst on the electrode was estimated to be 24 nmol cm -2 by inductively coupled plasma-optical emission spectroscopy (ICP-OES) analysis. X-ray photoelectron spectroscopy (XPS) of the Co 2p region (taken from a CoP L catalyst film) shown in Supplementary Fig. 1b HCOOH can be used directly in fuel cells, as a H 2 storage chemical 17 or as a useful synthon for cascade chemical synthesis 18 . The commercial production of syngas, CO and HCOOH often relies on energy-intensive processes, such as the steam reforming of fossil fuels 4,19 or methyl formate hydrolysis (for HCOOH) (refs. 4,20).
Conventional PEC systems for CO 2 reduction (CO 2 R) often rely on the oxygen evolution reaction (OER; E° = 1.23 V versus the reversible hydrogen electrode (RHE)) at the anode, which requires most of the overall energy input 4,21 . The operation of a single-light-absorber PEC system under unassisted, bias-free conditions therefore results in low fuel production rates due to the stringent thermodynamic requirement 4,6,22 . There are several reports on efficient and bias-free CO 2 R using PEC devices 6,18,23 , but these systems rely on dual-lightabsorber tandem structures to provide the necessary photovoltage for simultaneous water oxidation and CO 2 R. The replacement of the OER with less thermodynamically demanding anodic reactions, such as the oxidation of plastic-derived oxygenated substrates would not only make the overall process effective even with a single-light absorber 24 , but also valorize plastic waste, and thereby mitigate pollution. Although CO 2 R along with the oxidation of simple alcohols, such as glycerol, hydroxymethylfurfural (HMF) and glucose (derived from cellulose), has been recognized using solar-driven or electrochemical processes 4,25,26 , the use of plastics remains elusive.
Here we demonstrate a PEC system that operates under zero applied voltage for solar-driven CO 2 R coupled with the reforming of real-world, pretreated polyethylene terephthalate (PET) plastics to glycolic acid (GA, exists as glycolate under alkaline conditions; used as a feedstock in the pharmaceutical and cosmetics industries 27 ) in different pH environments, which makes the individual chemical processes conducive in the respective chambers. Lead halide perovskite devices (PVKs) are used in the system as perovskites are efficient light absorbers and can be integrated on encapsulation as a part of the photocathode indicates the successful attachment of the molecular catalyst on MWCNT 29 . The metallic Cu 91 In 9 CO 2 R cat was fabricated on a Cu foil support (geometrical surface area of 0.84 cm 2 ) via a reported galvanostatic electrodeposition technique 23 . The alloy composition is an optimized composition for syngas production at moderate overpotentials 23 . The powder X-ray diffraction (PXRD) pattern of Cu 91 In 9 revealed the existence of Cu 91 In 9 alloy phases along with some residual Cu and Cu 2 O (from aerial oxidation) phases ( Supplementary  Fig. 2a). The uniform microporous structure (average pore size of ~35 µm) of the Cu 91 In 9 alloy dendrites on the Cu foil was evident from the field-emission scanning electron microscopy (FESEM) image in Supplementary Fig. 2b. The inset of Supplementary Fig. 2b shows the energy dispersive X-ray (EDX) spectra that confirms the elemental stoichiometry, which was further confirmed by ICP-OES analysis. The bright-field scanning transmission electron microscopy (STEM) image and the corresponding elemental map ( Supplementary Fig. 2c 18 . A bimetallic Cu 27 Pd 73 alloy system (geometrical surface area of 2 cm 2 ) was employed as the anode oxidation catalyst. Previous work has already demonstrated that Pd-rich alloys are well-suited for substrate oxidation reactions, particularly the oxidation of ethylene glycol (EG), which is a hydrolysis product of PET plastics (PET hydrolysate) 24 . The Cu 27 Pd 73 catalyst was synthesized via a galvanostatic electrodeposition procedure using an activated Ni foam as the scaffold (Ni foam|Cu 27 Pd 73 ; see Methods for details). The PXRD pattern of the Cu 27 Pd 73 catalyst (deposited on a graphite foil scaffold instead of Ni foam to eliminate the interference of Ni with the alloy peaks) shown in Supplementary Fig. 3a reveals the shift in the Pd(111), Pd(200) and Pd(220) reflections towards a higher 2θ value, which indicates alloy formation 24,32 . The FESEM image ( Supplementary Fig. 3b) of the Cu 27 Pd 73 catalyst shows a flower-like morphology and EDX confirmed the elemental stoichiometry of the system ( Supplementary Fig. 3c), which was further corroborated by ICP-OES analysis. The transmission electron microscopy (TEM) image of Cu 27 Pd 73 also demonstrates its flower-like morphology ( Supplementary Fig. 3d), and high-resolution TEM (HRTEM) was used to estimate its atomic spacing as 0.22 nm, which corresponds to the (111) reflection ( Supplementary Fig. 3e). The bright-field STEM image and corresponding EDX maps (Supplementary Fig. 3f,g) further corroborate the morphology and elemental stoichiometry of the Cu 27 Pd 73 catalyst. The XPS survey spectra show an absence of surface contamination, whereas the Cu 2p and Pd 3d regions confirm the existence of both metals on the surface of the oxidation catalyst ( Supplementary Fig. 3h-j).

Electrochemical analysis for CO 2 R
Prior to the two-electrode PEC experiments, the CO 2 R cat systems were electrochemically characterized. Cyclic voltammetry (CV) analyses indicate the activity of the CO 2 R cat systems at moderate overpotentials ( Supplementary Fig. 4).
Controlled potential electrolysis (CPE) was subsequently carried out with each CO 2 R cat system individually using a three-electrode configuration with a Ag/AgCl reference electrode and a Pt mesh counter electrode. A two-compartment cell separated by a proton exchange membrane was used for all the electrochemical measurements (see Methods for details). The reaction medium was a CO 2 -saturated aqueous 0.5 M KHCO 3 electrolyte (pH 7.2) for CoP L and Cu 91 In 9 , and a CO 2saturated 3-(N-morpholino)propanesulfonic acid based electrolyte solution (MOPS, pH 6.4) for FDH. The more acidic pH for FDH is helpful to maintain an optimal local pH for enzyme function 33,34 . A constant potential of -0.7 V versus RHE was applied for both the molecular catalyst and the alloy system and -0.8 V versus RHE was used for the FDH biocatalyst. The applied potentials were selected according to the CV overlap plots (of the PVK|CO 2 R cat photocathode under illumination and the 'dark' Ni foam|Cu 27 Pd 73 anode), which determine the active operating conditions during the two-electrode PEC experiments at zero applied voltage (see 'PEC characterization of the electrodes' below for details). All the applied potentials were iR corrected and the duration of the electrochemical measurements was two hours.
Product quantification after the electrochemical experiments was carried out using gas chromatography for CO and H 2 , and ion-exchange chromatography for formate. The CoP L catalyst produced CO with a Faradaic efficiency (FE) of 85 ± 2%, whereas the metallic Cu 91 In 9 yielded syngas (FE CO = 48 ± 8% and FE H 2 = 46 ± 4%) and FDH produced formate with a FE of 96 ± 2% (Fig. 2a). The electrochemical analyses reveal that the product and its distribution from the CO 2 R reaction can be tuned using different types of molecular, metal and biological catalyst systems, which is consistent with previous reports 6 Isotopic labelling experiments were conducted under similar experimental conditions using 13 CO 2 -saturated, 13 C-labelled electrolytes (H 13 CO 3 -) to confirm the origin of the CO 2 R products (see Methods for details). After two hours of CPE, labelled 13 CO was detected as the only major product of CoP L and Cu 91 In 9 with no traces of 12 CO, as evident from the gas-phase infrared spectra in Fig. 2b. Similarly, 13 C-formate was detected as the sole product in case of FDH using 1 H NMR spectroscopy (Fig. 2c). These results confirm that gaseous CO 2 is the only carbon source in the products.

PEC characterization of the electrodes
For the PEC measurements, the PVK light absorber (sandwiched between the electron-and hole-transporting layers (ETL and HTL) respectively, followed by encapsulation) 6 was integrated with the respective CO 2 R catalyst to form the PVK|CO 2 R cat photocathode (see Methods for details). The unique architecture of the PVK-based photocathodes and the encapsulation procedure provided an effective and versatile platform for interfacing a diverse range of catalysts (see Methods for details). The PVK|CO 2 R cat was back-illuminated through the fluorine-doped tin oxide (FTO)-coated glass substrate during the experiments.
The operating conditions for the two-electrode PEC system under no applied voltage can be determined using CV overlap plots of the photocathode (PVK|CO 2 R cat systems with CoP L , Cu 91 In 9 or FDH) and the anode (Ni foam|Cu 27 Pd 73 ) (refs. 23,24). For this purpose, the CV scans for the Ni foam|Cu 27 Pd 73 anode (with a 0.5 M EG substrate in 1 M KOH; EG is the model substrate for PET plastic) were taken in a threeelectrode PEC configuration (reference electrode, Ag/AgCl; counter electrode, Pt mesh) using a two-compartment cell ( Supplementary  Fig. 5a). Similarly, individual CV scans under continuous solar irradiation (air mass 1.5 global (AM 1.5G)) were also taken for the different PVK|CO 2 R cat photocathode systems. As the reaction medium for the CoP L and Cu 91 In 9 catalysts, a CO 2 -saturated 0.5 M KHCO 3 electrolyte (pH 7.2) was used, and for FDH, a CO 2 -saturated MOPS-based electrolyte (pH 6.4) was employed. The overlap between the CV scans revealed the approximate working potential at the anode during the PEC measurements (at zero applied voltage). Therefore, as observed from Supplementary Fig. 5b-d, the positive working potentials were 0.39 V (J overlap = ~6.7 mA cm -2 ), 0.36 V (J overlap = ~4.9 mA cm -2 ) and 0.24 V (J overlap = ~1.7 mA cm -2 ) versus RHE when using PVK|CoP L , PVK|Cu 91 In 9 and PVK|FDH photocathodes, respectively. Consequently, considering the average open-circuit voltage provided by the PVK to be ~1.08 ± 0.01 V ( Supplementary Fig. 6), the working potentials experienced by the catalyst attached to the photocathode during the PEC experiments were estimated to be about -0.7 V, -0.7 V and -0.8 V versus RHE for CoP L , Cu 91 In 9 and FDH, respectively, which are used for the electrochemical characterizations (see above).
At the low positive potentials, the Cu 27 Pd 73 alloy anode efficiently catalyses the oxidation ( Supplementary Fig. 5a) of EG (obtained from pretreated PET) into GA. The alkaline conditions (1 M aqueous KOH) at the anode facilitate the oxidation process due to the involvement of OH ads species 24 and are also suitable for the pretreatment and depolymerization of PET plastics, as discussed in previous reports 9,24 . However, the neutral-acidic pH ranges at the cathode are necessary to maintain the stability-activity relationship of the different CO 2 R cat species involved 6,23,34 . Time (h) 8 10 Applied voltage (V)

Solar-driven PET reforming coupled to CO 2 -to-fuel production
The solar-driven PEC experiments were performed in a two-compartment, two-electrode configuration separated by a bipolar membrane using the bimetallic Cu 27 Pd 73 alloy oxidation catalyst (on a Ni foam scaffold) and the three different types of CO 2 R cat (CoP L , Cu 91 In 9 and FDH) integrated to PVK photocathodes ( Fig. 1; see Methods for details). The anolyte consisted of 1 M aqueous KOH (N 2 purged, pH ~14) and the photocathode compartment contained electrolyte solutions with near-neutral pH ranges. For PVK|CoP L and PVK| Cu 91 In 9 , CO 2 -saturated 0.5 M aqueous KHCO 3 (pH 7.2) was used as the catholyte. For the biological FDH electrodes, given the sensitivity of enzymes towards pH and salt concentrations and the local pH effects 33,34 , a CO 2 -saturated MOPS-based electrolyte (pH 6.4) was used as the catholyte following a previously reported protocol 18 . The bipolar membrane allowed for the optimization of individual oxidation and reduction processes simultaneously in different pH media, and generated a chemical bias between the electrodes due to the pH difference (∆pH ≈ 6-7) between the compartments 35,36 . The PVK|CO 2 R cat devices were back-illuminated through the FTO-coated glass substrate with simulated solar light (AM 1.5G) and all the PEC experiments were carried out at room temperature. Prior to the experiments using real-world, pretreated PET bottles, studies with EG (0.5 M) as the model substrate (at the anode) were first carried out with the Cu 27 Pd 73 ||PVK|CO 2 R cat (CoP L , Cu 91 In 9 or FDH) PEC systems. The purpose of using a model substrate was to realize the maximum efficiency of the systems in the absence of the depolymerized by-products from the PET pretreatment. The CV scans and chronoamperometry (CA) traces (at zero applied voltage) for tests with the EG model substrate are shown in Supplementary Fig. 7 (a complete set of chopped, light and dark CV scans are shown in Supplementary   Fig. 8), and the details on product yields and FEs are shown in Supplementary Fig. 9 and tabulated in Supplementary Tables 1 and 2. The PEC CO 2 R product distribution at zero applied voltage during CA (with EG at the anode) was consistent with that obtained electrochemically (discussed above), with Cu 27 Pd 73 ||PVK|CoP L , Cu 27 Pd 73 ||PVK|Cu 91 In 9 and Cu 27 Pd 73 ||PVK|FDH selectively forming CO, syngas and formate, respectively.
The forward chopped CV scan for Cu 27 Pd 73 ||PVK|CoP L using an alkaline pretreated PET solution as the anolyte (consisting of both EG and terephthalate (TPA)) is shown in Fig. 3a. The corresponding CA trace at zero applied voltage (Fig. 3b) revealed an average steady-state photocurrent density of 2.4 ± 0.3 mA cm -2 , which is consistent with the current (J overlap ) anticipated from the overlap of the CVs of PVK|CoP L under continuous illumination and that of Ni foam|Cu 27 Pd 73 under dark conditions ( Supplementary Fig. 5b). The time-dependent minor decrease in the photocurrent density may be attributed to a number of factors, such as device voltage fluctuations due to light chopping, minor catalyst poisoning and CO 2 mass transport limitations 24,37 . The amount of CO evolved from the PVK|CoP L photocathode (Supplementary Video 1) after ten hours of PEC experiments with pretreated PET at the anode was 263 ± 99 µmol cm -2 ( Fig. 3c and S9 S11 S11 S12 S12 S13 S14 S15 S16 S16 S17 S17 S17 S18 S19 S19 S19 H 2 CO HCOO -

Fig. 4 | Comparison with representative photocatalytic (PC) and PEC systems.
Comparison of our Cu 27 Pd 73 ||PVK|CO 2 R cat (CO 2 R cat is CoP L , Cu 91 In 9 or FDH) PEC system with other selected PC and PEC systems for CO 2 R. The rates are normalized to the irradiation area. Our single-light-absorber PEC system (under zero applied voltage with plastic reforming) exhibits ~10-100 times higher rates than those of PC CO 2 R processes and comparable product formation rates with bias-free, dual-light-absorber PEC tandems (see Supplementary Table 5 for further details, corresponding references and abbreviations).
Article https://doi.org/10.1038/s44160-022-00196-0 with molecular catalysts. After ten hours of PEC experiments, the anolyte was analysed using high-performance liquid chromatography (HPLC) to quantify the oxidation product(s). GA was identified as the major oxidation product with a yield of 31 ± 7 µmol ( Fig. 3e and Supplementary Table 4). The corresponding FE GA was 98 ± 3% (Supplementary Table 4). The minor drop in the average steady-state transient currents and product yields during the experiments using PET (as compared to pure EG) may be attributed to the presence of TPA in the solution obtained after pretreatment, which can block catalytic active sites 9,24 . Control experiments with only TPA as the substrate (without EG) resulted in a drastically lower photocurrent density (0.02 ± 0.01 mA cm -2 ) with negligible production of CO and H 2 after ten hours of the PEC experiment ( Supplementary Fig. 10).
The overall high FE of the oxidation reaction to produce GA is due to the low working positive potential during the PEC operation (at the anode as determined from the overlap plots, as discussed above) and the lack of possible intermediates of the C 2 substrate (EG, either in the pure form or from pretreated PET) during the 4eoxidation process 24 . The Cu 27 Pd 73 oxidation catalyst retained its morphology after 10 h of PEC measurement, as observed from the postcatalysis FESEM characterization in Supplementary Fig. 11. The ratio of the oxidation to total reduction products (GA:CO + H 2 ) agrees with the theoretically expected ratio (0.5, 4e − :2e − process) and is estimated to be ~0.45 (for both EG and the pretreated PET substrate).
To obtain further insights on the formation of GA, the adsorption energies of EG and possible reaction intermediates on the surface of Cu 27 Pd 73 were calculated using density functional theory (see Supplementary Fig. 12 for details). GA was found to have a more positive adsorption energy compared with those of EG or glycolaldehyde, consistent with its preferential formation.
The Cu 27 Pd 73 ||PVK|Cu 91 In 9 system was used for solar syngas production with pretreated PET at the anode. Consistent with the electrochemical experiments, Cu 91 In 9 produced about a 1:1 mixture of CO and H 2 at the working potentials offered by our PEC system with no external voltage (see 'Electrochemical analysis for CO 2 R' section above). Thus, the moderate-to-high reductive conditions (equivalent to -0.7 V versus RHE) that prevail in our PEC process aided the production of syngas over the bimetallic surface. The PEC experiments with the Cu 27 Pd 73 ||PVK|Cu 91 In 9 system were performed under the same conditions as those for the molecular catalysts (see above). Using pretreated PET at the anode, the CA traces revealed an average steady-state photocurrent density of 2.4 ± 0.1 mA cm -2 after ten hours. The amounts of CO and H 2 formed after ten hours were 212 ± 148 and 240 ± 75 µmol cm -2 , respectively (Fig. 3c and Supplementary Table 3) with a FE CO of 43 ± 24% and a FE H 2 of 49 ± 18%, as shown in Fig. 3d. About a 15% drop in the FE CO was observed from two to ten hours, which may be due to some minor phase separation of Cu and In in the alloy during catalysis, evident from post-catalytic STEM mapping (Supplementary Fig. 13). GA was identified as the main oxidation product from PET ( Fig. 3e and Supplementary Table 4) with a yield of 52 ± 18 µmol and a FE GA of 96 ± 9%. The [oxidation product] to [CO + H 2 ] ratio was estimated to be ~0.45 ± 0.02 (theoretically expected ratio of 0.5, 4e − :2e − process). A representative wavelength-dependent external quantum efficiency (EQE) spectrum (action spectrum) of the Cu 27 Pd 73 ||PVK|Cu 91 In 9 system ( Supplementary  Fig. 14) was determined at zero applied voltage and a PET substrate used to monitor the PEC device efficiency. The spectrum exhibited a plateau at ~80-90% in a wavelength range from 420 to 730 nm, which is comparable with those of other PVK devices and indicates no limitation through light absorption 24,38 .
Moving beyond the molecular and solid-state CO 2 R systems, FDH was used to construct a Cu 27 Pd 73 ||PVK|FDH solar-driven plastic oxidation-CO 2 R bio-photoelectrochemical system operating at zero applied voltage. After ten hours, the average steady-state photocurrent density obtained for the bio-photoelectrochemical system was 0.9 ± 0.1 mA cm -2 with pretreated PET at the anode. The lower photocurrent density for FDH may be due to a low enzyme loading and limited availability of the active sites 18 and is consistent with the CV overlap (J overlap ) of the individual PVK|FDH photocathode (under continuous illumination) and that of Ni foam|Cu 27 Pd 73 (discussed in the 'Electrochemical analysis for CO 2 R' section above). After the ten hours of bio-photoelectrochemical operation (at zero applied voltage) with pretreated PET substrate, 121 ± 87 µmol cm -2 formate (Fig. 3c and Supplementary Table 3) was obtained at the photocathode with a FE formate > 95% (Fig. 3d). The TON formate and TOF formate were 3.4 × 10 5 and 9 s -1 , respectively, assuming all the drop cast enzymes were immobilized and electroactive (Supplementary Table 3). The corresponding oxidation product generated at the anode ( Fig. 3e and Supplementary Table 4) was GA (14 ± 8 µmol) with a FE GA of 96 ± 14%. The [oxidation product] to [formate] ratio was estimated to be 0.43 ± 0.01 (theoretically expected ratio of 0.5, 4e − :2e − process). Similar to other CO 2 R cat systems for PEC operation, the results with the PET substrate are comparable to those obtained using the EG model substrate (see Supplementary Tables 1  and 2 for details).

Comparison with representative systems and future scope
Although there are several reports on PEC systems that use CO 2 R to give products such as syngas, formate and so on, they are distinct from our system. Most such processes rely on the OER as the anodic reaction 6,18,23 , which raises the thermodynamic demand. As a result, the PEC systems reported so far either require an external energy input in the form of an electrical bias 7,39-41 , or dual light-absorber tandems 6,22,23 to drive the reactions, but often with low production rates ( Fig. 4 and Supplementary Table 5).
The single-light-absorber Cu 27 Pd 73 ||PVK|CO 2 R cat PEC device performs CO 2 R with high efficiency and selectivity, accompanied by selective reforming of real-world PET plastics without the need for an external voltage input. The product formation rates (µmol cm -2 h -1 ) achieved by the system are comparable to (or higher than) most existing bias-free PEC tandems 6,18,23 (Fig. 4) and ~10-100-fold higher than various photocatalytic CO 2 R processes based on heterogeneous photocatalysts ( Fig. 4 and Supplementary Table 5) 42,43 . Moreover, the added advantage of our system is the versatility and tunability of the substrate and product scope, which may be further explored and broadened in future development. Depending on the working potentials and nature of the CO 2 R cat , the system can effectively produce CO, syngas or formate with a high selectivity, as demonstrated. We also note that in the case of solar-driven reforming systems (for example, PEC reforming), the solar-to-fuel (STF) efficiency is not a decisive metric for comparison when the water oxidation is replaced by the oxidation of organics at the anode. This is because for organic oxidations (the oxidation of EG from pretreated PET in our case), the ∆G° values are low as opposed to the large ∆G° for water oxidation (~237 kJ mol -1 ) when coupled to proton reduction 8 . This results in a substantially reduced STF efficiency (Supplementary Table 6) and does not provide an accurate estimate for the economic potential of solar-driven waste-reforming processes.
The separation between the individual cathodic and anodic compartments allows for a better optimization of the individual reaction processes, and the overall system can benefit from the use of flow set-ups, thermoelectric units, gas-diffusion (photo)electrodes and concentrated light systems 44,45 . These can potentially result in improved current densities, yields and efficiencies, and ultimately pave the way for scaling and commercial implementation.

Conclusions
An efficient and versatile PEC system was developed by combining solar-driven CO 2 R with plastic reforming to form value-added products employing a single-light absorber with no applied voltage. Three different types of CO 2 R cat systems (a CoP L molecular catalyst, a bimetallic Cu 91 In 9 alloy and a FDH biocatalyst) were integrated with a perovskite light absorber to form the photocathodes. A bimetallic Cu 27 Pd 73 alloy Article https://doi.org/10.1038/s44160-022-00196-0 was used as an oxidation catalyst to reform PET plastic to GA with >90% FE. The overall PEC system, defined as Cu 27 Pd 73 ||PVK|CO 2 R cat had a tunable product distribution that produced CO, syngas or formate with a high selectivity and noteworthy product formation rates, in combination with PET reforming at the anode. This work presents a unique demonstration in which solar-driven, selective CO 2 R is combined with plastic waste valorization. This advancement in PEC systems is not only an important stepping stone towards diversifying the scope of solar-fuel synthesis with an improved efficiency and selectivity, but also a key indicator towards sustainable commercial implementation. , sodium tetrachloropalladate (Na 2 PdCl 4 , 98%, Sigma-Aldrich), Cu foil (99.9%, Alfa-Aesar), caesium chloride (CsCl, >99.5% trace metals basis, Sigma-Aldrich), dldithiotreithol (DTT, >99.5%, Sigma-Aldrich), MOPS sodium salt (>99.5%, Sigma-Aldrich), MOPS (>99.5%, Sigma-Aldrich), polystyrene latex microspheres (PS beads, 2.5 wt% dispersion in H 2 O, diameter 750 nm, Alfa Aesar), sodium hydrogen carbonate (NaHCO 3 , >99.998% trace metal basis, Puratronic), sodium hydrogen carbonate-13 C (98 atom% 13 C, Sigma-Aldrich), titanium dioxide nano particles (P25, anatase:rutile 80:20, diameter 21 nm, Evonik Industries), tris(hydroxymethyl)aminomethane hydrochloride (TRIS-HCl, >99.0%, Sigma-Aldrich) and a commercial sparkling water PET bottle (Highland Spring, sourced from Sainsbury's UK) were used without further purification unless otherwise stated. W-formate dehydrogenase (FDH) from Desulfovibrio vulgaris Hildenborough (DvH) was purified according to previous reports 18,46 .

Preparation of PVK devices
The inverse structure triple-cation mixed halide PVK devices were first fabricated according to previously reported protocols with slight modifications 38,47 . In brief, a HTL of NiO x was deposited on a FTO-coated glass substrate by spin coating a solution of Ni(NO 3 ) 2 ·6H 2 O (1 M) and ethylenediamine (1 M) in EG, followed by annealing the FTO-coated glass at 573 K. Next, a second HTL was deposited over the NiO x by spin coating a F4TCNQ-doped PTAA solution inside a N 2 -filled glove box. This second HTL ensured a more effective charge extraction, and thereby increased both the photocurrent and photovoltage over those of previous reports 47 . For the deposition of the perovskite layer, a precursor solution of caesium formamidinium methylammonium perovskite was prepared by first making 1,000 µl of a FAMA 0.22 Pb 1.32 I 3.2 Br 0.66 solution in 510 µl of dimethylformamide, 340 µl of dimercaptosuccinic acid and 150 µl of 1-methyl-2-pyrrolidone, to which 48 µl of caesium iodide in dimethyl sulfoxide (1.5 M) was added. Next, the PVK film was deposited onto the PTAA layer via a two-step spin-coating technique: first, 10 s at 1,000 r.p.m. and next 35 s at 6,000 r.p.m. Chloroform was used as an antisolvent ~7 s before the end of the spin coating. The PVK layer was then annealed for 30 min at 373 K. The ETL was next deposited over the perovskite layer via spin coating a solution of PCBM (35 mg ml -1 in chlorobenzene) at 3,000 r.p.m. for 45 s. This was followed by the deposition of a PEIE film onto the PCBM-coated perovskite via spincoating the PEIE solution (3.9 µl ml -1 in isopropanol) at 3,000 r.p.m. for 30 s. The PEIE layer helps to prevent the interfacial degradation due to a reaction with the metal contact. A 100 nm conductive Ag layer was finally deposited by metal evaporation through a patterned mask to ensure an active perovskite area of ~0.5 × 0.5 cm 2 .

Encapsulation and preparation of the CO 2 R photocathodes
The encapsulation step is critical to prevent the degradation of the PVK devices in aqueous medium. For encapsulation, graphite powder was mixed homogeneously with epoxy (Araldite standard 2 component epoxy) in a 3:4 graphite:epoxy mass ratio to form a conductive graphite epoxy (GE) paste. The paste was then evenly spread over the Ag contact layer of the perovskite device. This paste provided a specific advantage in terms of the reaction scope as it enabled the integration of perovskite devices with a broad range of catalysts. The specific CO 2 R cat film (fabrication protocols are discussed in detail below) was then placed over the GE and the device was allowed to dry for 24 h for the GE to harden. Photographs of the individual CO 2 R cat films are shown in Supplementary Fig. 15. Thereafter, the wiring was done and the edges of the device were sealed using Araldite 5-min Rapid 2 component epoxy 47 .
The final PVK-based photocathodes are denoted as PVK|CO 2 R cat (CO 2 R cat is CoP L , Cu 91 In 9 or FDH). Photographs of the integrated PVK|CO 2 R cat photocathodes are shown in Supplementary Fig. 16.

Fabrication of the CoP L molecular catalyst on a graphite foil
The CoP L molecules were synthesized according to a recently reported procedure 29 . For the fabrication of the catalyst films, 5 mg of the MWCNT was dispersed in 2.34 ml of dimethylformamide using probe sonication for 10 min (80% amplitude). Thereafter, 600 µl of the freshly prepared 1000 µM CoP L (M W = 2,202 g mol −1 ) solution in dimethylformamide and 60 µl of Nafion (5% v/v in lower aliphatic alcohols and water) was added to the dispersion. The mixture was further bath sonicated for 20 min. The resulting catalyst ink (100 µl) was then dropcast onto activated graphite foil and dried overnight under ambient conditions to form the CoP L film. The film was then attached to the GE encapsulant of the perovskite device, as mentioned above.

Preparation of the Cu 91 In 9 catalyst
The bimetallic Cu 91 In 9 catalyst was prepared by a galvanostatic electrodeposition method in which dynamic H 2 bubbles were used as a template to form macroporous structures of the catalyst. CuSO 4 ·5H 2 O and In 2 (SO 4 ) 3 ·H 2 O were used as precursor salts. The deposition bath contained 70:30 Cu 2+ :In 3+ (total concentration 0.025 M) in 1.5 M H 2 SO 4 . A three-electrode configuration was used in which a Cu foil was used as the counter electrode, a Ag/AgCl electrode was used as the reference and the Cu foil substrate was used as the working electrode. During the galvanostatic deposition, a -3 A cm −2 current density was applied for 60 s. The catalyst was rinsed with Milli-Q water and then dried under a N 2 stream.
Article https://doi.org/10.1038/s44160-022-00196-0 After annealing at 500 °C for 20 min with a heating rate of 1 °C min −1 , the resulting electrodes had a geometrical surface area of ~0.28 cm 2 and a thickness of ~10 µm. The electrode was then combined with the PVK device by attaching to the GE, as discussed in 'Encapsulation and preparation of the CO 2 R photocathodes' above. Prior to the experiments, FDH (100 pmol) was preincubated for 5 min with DTT in a TRIS buffer solution (5 µl, 20 µM FDH, 50 mM DTT, 20 mM TRIS-HCl, pH 7.6) before immobilization on the Ti|IO-TiO 2 electrode under a N 2 atmosphere. After 10 min of incubation, the Ti|IO-TiO 2 |FDH electrode was immersed in the electrolyte solution (16 ml, 86 mM MOPS, 50 mM CsCl, 50 mM NaHCO 3 , pH 6.4, pre-purged with CO 2 ).

Preparation of the Cu 27 Pd 73 oxidation catalyst
The Cu 27 Pd 73 alloy catalyst was synthesized by a dynamic H 2 -bubbleassisted galvanostatic electrodeposition method using an activated Ni foam as the scaffold. CuSO 4 ·5H 2 O and Na 2 PdCl 4 were used as the precursor salts for Cu and Pd, respectively. The electrolyte was prepared by dissolving total 0.02 M precursor salts in 30:70 Cu 2+ :Pd 2+ in 1.5 M H 2 SO 4 . A three-electrode configuration was used for the electrodeposition with a Ag/AgCl electrode as the reference electrode, a Pt foil as the counter electrode and the activated Ni foam scaffold as the working electrode. Galvanostatic deposition was carried out by employing a -1.5 A cm −2 current density for 60 s. After preparation, the catalyst was washed in Milli-Q water and dried under a gentle N 2 flow.

Material characterization
The PXRD measurements of the samples electrodeposited on the graphite scaffold were performed using a Paralytical X'Pert Pro (Cu Kα radiation) diffractometer with a 2θ range from 30 to 80° at a scan rate of 1° min -1 . The FESEM images were acquired using a TESCAN MIRA3 FEG-SEM instrument equipped with an Oxford Instruments Aztec Energy X-maxN 80 EDX system. The TEM, high-resolution TEM, bright-field STEM and elemental mapping were performed using a Thermo Scientific Talos F200X G2 TEM (FEI, operating voltage 200 kV). The UV-vis spectra were recorded using a Varian Cary 50 UV-vis spectrophotometer. The XPS of the catalysts was performed at the Maxwell Centre, University of Cambridge, with a near ambient pressure XPS system using a SPECS XR 50 MF X-ray source, µ-FOCUS 600 X-ray monochromator and a differentially pumped PHOIBOS 150 1D-DLD near ambient pressure analyser. The peak positions were calibrated with respect to the C 1s peak at ~286 eV and a Casa-XPS software was used for the curve fitting and deconvolution. The ICP-OES measurements were performed on a Thermo Scientific iCAP 7400 ICP-OES DUO spectrometer at the Microanalysis Service, Yusuf Hamied Department of Chemistry, University of Cambridge.

Pretreatment of real-world PET bottles
The real-world sparkling water bottle made of PET plastic was pretreated using an alkaline pretreatment method 24 . Briefly, the PET bottle was first cut into small pieces and then dipped in liquid nitrogen. Thereafter, the pieces were pulverized in a grinder. The grinded PET bottle was then added to 1 M aqueous KOH (concentration 50 mg ml -1 ) and heated to 80 °C for 100-120 h under stirring to ensure sufficient depolymerization of the PET. The solution was then kept unperturbed at room temperature to cool and allow the unreacted cloudy mass of PET to settle down. The final concentration of EG after pretreatment was ~11.8 ± 4.4 mg ml -1 , as determined using HPLC. For the PEC experiments (at zero applied voltage) using real-world PET, the supernatant was directly taken as the anolyte.

Electrochemical and PEC measurements
The electrochemical and PEC measurements were performed with PalmSens Multi EmStat3+ (multichannel potentiostat that consisted of four separate channels) and Ivium CompactStat potentiostats. The experiments were conducted in a two-compartment cell separated by a bipolar membrane. Unless mentioned explicitly, for all the experiments the electrolyte in the cathodic chamber consisted of 0.5 M aqueous KHCO 3 (purged with CO 2 with 2% CH 4 as the internal standard for 30 min; pH 7.2) for the CoP L and Cu 91 In 9 catalysts. In the case of the FDH catalyst, the electrolyte was prepared according to previously reported protocols 18 and comprised 86 mM MOPS, 50 mM NaHCO3 and 50 mM CsCl (CO 2 purged; pH 6.4). Similarly, for all the measurements the anolyte was either 1 M aqueous KOH (with a 0.5 M EG model substrate) or alkaline (1 M aqueous KOH) pretreated PET bottles (discussed above) and was purged with N 2 (with 2% CH 4 as the internal standard) before the experiments.
The dark CV scans for Ni foam|Cu 27 Pd 73 (with 0.5 M EG substrate) were performed in a three-electrode configuration with a Pt mesh as the counter electrode and Ag/AgCl (saturated NaCl) as the reference electrode at a scan rate of 25 mV s -1 . Similarly, the CV scans for the individual photocathodes (PVK|CoP L , PVK|Cu 91 In 9 or PVK|FDH) were also taken under a continuous solar illumination at a scan rate of 10 mV s -1 (PVK layer was back-illuminated through the FTO glass). For such PEC measurements in a two-compartment set-up, a Newport Oriel 67005 solar light simulator was used, equipped with an AM 1.5G solar filter. The light intensity was calibrated to 100 mW cm -2 (1 sun) before each PEC measurement. From the combination of the individual CV curves of Ni foam|Cu 27 Pd 73 and PVK|CO 2 R cat (CoP L , Cu 91 In 9 or FDH), the overlap potentials (the theoretical working potential for the PEC operation under a standalone, zero applied voltage) were determined for each individual system.
Consequently, to corroborate the working potential and product selectivity of the cathodic products under a zero applied voltage solardriven PEC operation, electrochemical tests were first performed with the different CO 2 R cat electrocatalysts at their respective working potentials determined from the overlap plots. For this purpose, a twocompartment cell was used with a 0.5 M aqueous KHCO 3 electrolyte in both compartments (CO 2 purged before the electrochemical measurements) and the CO 2 R cat film (graphite|MWCNT|CoP L , Cu|Cu 91 In 9 or Ti|IO-TiO 2 |FDH), Pt mesh and Ag/AgCl were used as the working, counter and reference electrodes, respectively. CPE was performed at the particular working potential determined from the overlap plots (iR-corrected) and the products were analysed after 2 h. The pH of 1 M aqueous KOH was estimated to be 14. To further confirm the CO 2 R products, electrochemical isotope labelling experiments were carried out using 13 C-labelled electrolytes and 13 CO 2 purging under identical conditions. Unless otherwise mentioned, all the potentials were converted into the RHE scale from the Ag/AgCl scale according to equation (1): The solar-driven PEC measurements were performed in the twocompartment, two-electrode configuration with a Ni foam|Cu 27 Pd 73 (effective area ~2 cm 2 ) anode and a PVK|CO 2 R cat photocathode (CoP L , Cu 91 In 9 or FDH) separated by a bipolar membrane. The CV scans (chopped, light and dark) were performed at a scan rate of 10 mV s −1 . The long-term CA scans were performed for 10 h at zero applied voltage under a chopped light irradiation (50 min on, 10 min off) and the photocurrents obtained were normalized to the photoactive area of the PVK. All the experiments were conducted at room temperature with stirring in both compartments.
For the EQE measurements, a LOT MSH-300 monochromator, Thorlabs PM100D power meter with a Thorlabs S302C thermal power sensor and an Ivium CompactStat potentiostat were used. The wavelength (full-width at half-maximum 15 nm) was varied between 300 and 800 nm in 25 steps every 30 s and the EQE values were determined using equation (2), where h is the Planck constant, c the speed of light, J the photocurrent density, e the electronic charge, λ the wavelength and P λ the wavelength-dependent light intensity flux: Article https://doi.org/10.1038/s44160-022-00196-0 The STF efficiencies for the different CO 2 R products were estimated using equation (3) as in previous reports 18,23,48 , where J is the average steady-state photocurrent density, \∆E°\ the difference in standard potentials, FE the Faradaic efficiency for the CO 2 R product and P the solar power density (the values are tabulated in Supplementary Table 6):

Product detection and quantification
The evolved gas from the photocathode (CO and H 2 ) was detected and quantified using a Shimadzu GC-2010 Plus gas chromatograph by manual injection from the cell headspace (2% CH 4 was used as an internal standard) 47 . The formate was detected and quantified using ion-exchange chromatography (Metrohm 882 compact ion-exchange chromatography system) equipped with a Metrosep A Supp 5-150/4 column (eluent: 3.2 mM aqueous Na 2 CO 3 , 1 mM aqueous NaHCO 3 ). The chromatographic separations for the oxidation products were conducted using a Waters HPLC system equipped with a Phenomenex Rezex 8% H + column at a column temperature of 75 °C. The samples were analysed in the isocratic flow mode (flow rate: 0.5 ml min −1 , 0.0025 M aqueous H 2 SO 4 ) using a Waters breeze system equipped with refractive index (RIS-2414) and diode array UV-vis (λ = 254 nm) detectors. The FEs of the products formed were calculated using equation (4), where Z is the number of electrons transferred, n the number of moles of product formed, F the Faraday constant (96,485 C mol -1 ) and Q passed the total amount of charge passed during the same time interval: The TON and TOF of the CO 2 R catalysts (CoP L and FDH) were calculated using equations (5) and (6), respectively, where n product is the number of moles of product formed, n catalyst the number of moles of catalyst present on the active area and t the duration of the experiment: TOF (s −1 ) = TON t (6) For the 13 C-isotope labelling experiments, 13 CO was detected using infrared spectroscopy (Thermo Scientific Nicolet iS50 infrared spectrometer) in the gas-phase transmission mode. The head space from the cell was transferred to an air-tight evacuated infrared cell (path length, 10 cm; equipped with KBr windows) after the experiment for the detection of 13 CO. The 13 C-formate formed in solution was detected using 1 H NMR spectroscopy (Bruker DPX 400 spectrometer) in D 2 O.

Computational calculations
Density functional theory calculations were performed on the Quantum Espresso 6.4 code using the projector augmented wave method and pseudopotentials with plane waves and density cutoffs of 40 and 200 Ry, respectively. The revised Perdew-Burke-Ernzerhof exchange functional was used with a Fermi-level smearing of 0.01 Ry. Optimizations were converged if the forces were below 0.001 Ry Bohr -1 . A bulk Cu 1 Pd 3 face-centred cubic metal alloy was constructed and optimized allowing the unit cell to relax using a 11 × 11 × 11 Monkhorst-Pack k-point grid. The optimized Cu 1 Pd 3 bulk structure was used to construct (111) surfaces in a 4 × 4 slab (11.02 × 11.02 Å) that consisted of 4 layers (64 atoms, 16 atoms per layer) with a vacuum spacing of 30 Å. Geometry optimization and single-point calculations were performed using a 4 × 4 × 1 Monkhorst-Pack k-point grid with the bottom two layers being kept frozen. A second slab model with the surface copper atoms being coordinated by hydroxyl (OH) groups was constructed. Molecules were calculated in a 15 × 15 × 15 Å cubic unit cell using a 1 × 1 × 1 Monkhorst-Pack k-point grid. Adsorption energies were calculated based on the electronic energies between the slab surface, the adsorbate in vacuum and the adsorbate coordinated on the slab surface.