Natural photosynthesis stores sunlight in chemical energy carriers, but it has not evolved for the efficient synthesis of fuels, such as H2. Semi-artificial photosynthesis combines the strengths of natural photosynthesis with synthetic chemistry and materials science to develop model systems that overcome nature’s limitations, such as low-yielding metabolic pathways and non-complementary light absorption by photosystems I and II. Here, we report a bias-free semi-artificial tandem platform that wires photosystem II to hydrogenase for overall water splitting. This photoelectrochemical cell integrated the red and blue light-absorber photosystem II with a green light-absorbing diketopyrrolopyrrole dye-sensitized TiO2 photoanode, and so enabled complementary panchromatic solar light absorption. Effective electronic communication at the enzyme–material interface was engineered using an osmium-complex-modified redox polymer on a hierarchically structured TiO2. This system provides a design protocol for bias-free semi-artificial Z schemes in vitro and provides an extended toolbox of biotic and abiotic components to re-engineer photosynthetic pathways.
Semi-artificial photosynthesis bridges the rapidly progressing fields of synthetic biology and artificial photosynthesis, and offers a platform to develop and understand solar fuel generation1,2,3,4. Synthetic biology has vastly opened up the way that nature can be manipulated to streamline functionality and to build artificial biological systems, but its complex machineries and metabolic pathways limit engineering flexibility5. Artificial photosynthesis utilizes synthetic, often biomimetic, components to convert and store solar energy, but it is often constrained by inefficient catalysis and costly and/or toxic materials6. Semi-artificial photosynthesis aims to integrate the high efficiency and selectivity of enzymes with the controllability of synthetic materials to photocatalyse endergonic reactions in the absence of competing processes7. It also allows the construction of biologically inaccessible pathways with a high level of control and flexibility3. The catalytic activity of redox enzymes can be harnessed when adsorbed on electrodes by protein film electrochemistry and protein film photoelectrochemistry (PF-PEC)8,9. A key challenge is to design biotic–abiotic interfaces that effectively wire together the biological and synthetic components to operate at their optimum.
Solar-driven water splitting into H2 and O2 is the most prominent model reaction in artificial photosynthesis10. Inefficient catalysis (particularly, kinetically slow O2 evolution and the formation of partially oxidized side products) is a major limitation in synthetic systems, and results in the requirement of large overpotentials and energy conversion losses6. Oxygenic organisms convert solar energy using a photosynthetic Z scheme that contains two light absorbers, photosystem I (PSI) and photosystem II (PSII)11. In this tandem configuration, the first excitation in PSII drives water oxidation to O2 and produces a proton gradient, whereas the second excitation in PSI generates a low potential electron to drive CO2 fixation into sugars12. Alternatively, H2 can be produced from microalgae and cyanobacteria via electron transfer from ferredoxin to a [FeFe]-hydrogenase ([FeFe]-H2ase), which reduces protons to H2 (ref. 13). The efficiencies for photobiological H2 production are low for several reasons14,15. First, PSII and PSI overlap in light absorption and compete for a small fraction of the solar spectrum. Second, high light intensities limit the efficient electron flux upstream and downstream of PSII. Third, in vivo H2 production relies on O2-sensitive [FeFe]-H2ases, which prevents sustained water splitting. Fourth, CO2 fixation is preferred over proton reduction, which leads to low H2 yields. Overcoming these limitations offers scope to enhance H2 production with biological components.
We previously reported a PEC water-splitting system with a PSII photoanode wired to a [NiFeSe]-H2ase cathode3. However, this system relied solely on light absorption by PSII and required an externally applied voltage due to the low electrochemical potential of the electrons that leave PSII. This limitation can be resolved by introducing a second light absorber to further promote the energetics of the electrons to be delivered to H2ase16. To generate sufficient driving force for overall water splitting, while maximizing solar energy harvesting, complementary dual-absorber/tandem systems can be assembled, which show theoretical limits for a solar-to-hydrogen efficiency of up to 25%17. PSII18,19 and BiVO4 (ref. 20) photoanodes wired to PSI photocathodes have been reported to produce electricity, but no chemical fuel. Tandem systems that contain PSII have not been combined with enzymatic fuel synthesis21,22,23.
Here, a semi-artificial system for the unassisted, light-driven water splitting with PSII and H2ase is presented. This PEC system does not require an external energy input as dual light absorption is realized by a tandem photoanode that consists of PSII wired to dye-sensitized titanium dioxide (TiO2), which provides a sufficient voltage to reduce protons using a H2ase cathode. This PEC design was inspired by dye-sensitized solar cells24,25 and it allows PSI to be replaced by a rationally designed diketopyrrolopyrrole (dpp) dye with an absorption complementary to that of PSII. An efficient electronic communication between PSII and dpp was achieved by using the redox polymer poly(1-vinylimidazole-co-allylamine)–Os(bipy)2Cl (POs), which bypasses possible limitations from an inefficient interfacial electron transfer. Simultaneously, the hydrogel character of the redox polymer provides a solvated environment for the biocatalyst. A hierarchically structured inverse opal TiO2 (IO-TiO2) scaffold was employed to provide a high surface area for the effective integration of polymer/PSII.
Assembly of the tandem PSII–dye photoanode
The components and assembly of the PEC tandem cell are depicted in Fig. 1 and the principle of operation as a semi-artificial Z scheme is shown in Supplementary Fig. 1. Hierarchically structured IO-TiO2 electrodes were assembled on a TiO2-layer-protected fluorine-doped tin oxide (FTO)-coated glass substrate via a modified co-assembly method (Supplementary Fig. 2)3. The TiO2 protection layer was used to prevent the direct contact of the electroactive components (PSII and POs) with the FTO. The optimal thickness of the IO-TiO2 film was determined to be 20 μm, based on preliminary electrochemical screening, and was utilized throughout this work4. Scanning electron microscopy (SEM) revealed a diameter of 750 nm for the TiO2 macropores with connecting channels of 150 nm and a mesoporous skeleton with a porosity of approximately 50 nm. The macroporosity permits the penetration of the bulky PSII and POs (~20 nm and 16 nm in size, respectively)3,4, whereas the mesoporous structure provides an enhanced effective surface area for the guest adsorption.
Dye sensitization of the IO-TiO2 photoanodes (geometrical surface area, A = 0.25 cm2) was performed by soaking IO-TiO2 in a tetrahydrofuran (THF) solution of dpp26,27 overnight (Supplementary Fig. 3). The resulting IO-TiO2|dpp electrodes had a dpp surface loading of 72 ± 4 nmol cm−2 (estimated by spectrophotometry (Supplementary Fig. 4)), consistent with previously reported loadings of molecular species on mesoporous metal oxide scaffolds28. The organic dpp dye was selected for its complementary absorption spectrum to PSII and for its ability to act as an efficient visible light photosensitizer for TiO2 in aqueous media, chemisorbed via its phosphonic acid anchoring group27. For comparison, electrodes were also prepared using the ruthenium bis(2,2′-bipy)(4,4′-bis(phosphonic acid)-2,2′-bipy) dibromide dye (RuP; bipy, 2,2′-bipyridine)29,30,31,32, which is commonly used as a benchmark in aqueous dye-sensitized schemes (Supplementary Figs. 1c and 3).
The macroporous voids of the IO-TiO2|dpp electrodes were subsequently filled with a blend of PSII and POs to give the fully integrated IO-TiO2|dpp|Pos−PSII tandem photoanode. The PSII chosen for this study was isolated from the thermophilic cyanobacterium Thermosynechococcus elongatus as a well characterized33,34, highly active and relatively robust PSII variant35,36. Loadings of 24 ± 4 nmol cm–2 and 143 ± 25 pmol cm–2 were determined by inductively coupled plasma optical emission spectroscopy (ICP-OES) and Ultraviolet–visible (UV–vis) spectrophotometry for POs and PSII, respectively (Supplementary Fig. 4). The redox polymer POs mediates the electron transfer between PSII and the IO-TiO2|dpp surface, which enhances their electrical connection and physically stabilizes PSII on the electrode4,37. The integration of all the biotic and abiotic components in the hybrid photoanode was further confirmed by high angular annular dark field scanning transmission electron microscopy (HAADF–STEM) (Supplementary Fig. 5).
The UV–vis absorption spectra of dpp, POs and PSII in solution and individually adsorbed on the IO-TiO2 electrodes were recorded (Fig. 1 and Supplementary Fig. 6). IO-TiO2|PSII displays absorption maxima (λmax) at 450 nm (Bx/y Soret bands) and 680 nm (Qy band)38, whereas IO-TiO2|dpp showed a broad absorption from 475 to 575 nm. Thus, the absorption spectra of the co-sensitized IO-TiO2 electrode demonstrate a panchromatic light absorption and highlight the light harvester’s complementarity desired for a semi-artificial Z scheme. For comparison, the spectrum of RuP (λmax = 457 nm) significantly overlaps with that of PSII (λmax of the Bx/y band at 450 nm). The IO-TiO2|POs spectrum consisted of a broad and weak absorption, in line with the modest molar absorption of POs (extinction coefficient ε = 8.72 mM–1 cm–1 at 531 nm)4, which is not expected to substantially affect the light conversion efficiency in the fully assembled IO-TiO2|dpp|POs−PSII tandem system.
The photocurrent response (J) of IO-TiO2|dpp|POs−PSII was recorded by PF-PEC at an applied potential (Eapp) in a three-electrode configuration. Stepped potential chronoamperometry under a periodic illumination with ultraviolet-filtered simulated solar light was used (1.5 AM filter, light intensity flux (irradiance) Ee = 100 mW cm–2, λ > 420 nm (Fig. 2)). Photogenerated electrons in PSII are transferred to the electron acceptor plastoquinone B (QB) (Fig. 1)9. The holes are collected at the oxygen-evolving complex, where water is oxidized to liberate protons and gaseous O2. The conduction band (CB) of IO-TiO2 receives electrons from the photoexcited dpp which is thereby oxidized (dpp+) to give rise to an anodic photocurrent. The Os3+ complex embedded in POs mediates the electrons between the reduced QB and oxidized dpp to close the electric circuit between the two light absorbers.
The photoanodic current onset potential (Eonset) of approximately −0.5 V versus the standard hydrogen electrode (SHE) is consistent with the reported anatase TiO2 CB edge of approximately −0.6 versus SHE39 (Fig. 2 and Supplementary Fig. 7). The IO-TiO2|dpp|POs−PSII tandem electrode exhibits a shift of more than 0.5 V towards a negative potential compared to single-absorber photoanodes with immobilized PSII3,4,40, which makes it a promising candidate for bias-free overall water splitting. Potential-independent steady-state photocurrents (80 μA cm−2) were observed at Eapp > −0.2 V versus SHE (Fig. 2b) and attributed to water oxidation3,4. Prolonged irradiation at Eapp > −0.2 V versus SHE results in an irreversible drop in photocurrent, most probably due to PSII photodegradation (irreversible light-induced D1 subunit damage)4.
Control experiments that omitted one component of the tandem photoanode exhibited only a marginal photoactivity. The small background photoresponse for IO-TiO2|dpp|PSII and IO-TiO2|dpp|POs (Fig. 2), and similarly for IO-TiO2|dpp (Supplementary Fig. 7a,b) can be assigned to the stoichiometric electron transfer from photoexcited dpp (dpp*) to TiO2 without the regeneration and photocatalytic turnover of the dye. A low photocurrent in the absence of POs (IO-TiO2|dpp|PSII) supports an insufficient direct electronic interaction between PSII and dpp without the redox polymer4. No significant photocurrents were observed in the absence of dpp (IO-TiO2|POs−PSII (Fig. 2) and IO-TiO2, IO-TiO2|POs and IO-TiO2|PSII (Supplementary Fig. 7a,b)), consistent with the more positive reduction potentials of PSII’s QB and POs relative to the CB of TiO2, which results in an unfavourable electron transfer. The presented semi-artificial system therefore demonstrates the successful assembly of a functional biotic–abiotic interface for a controlled electron transfer in an artificial Z scheme. A PSII tandem system based on IO-TiO2|RuP was also assembled and exhibited a similar behaviour (Supplementary Fig. 7c,d). To maximize the performance of the tandem systems, screenings of the dye loading (Supplementary Fig. 8), POs−PSII ratio (Supplementary Fig. 9) and IO-TiO2 thickness (Supplementary Fig. 10) were conducted.
Photocurrent action spectrum
The photocurrent response as a function of irradiation wavelength (the photocurrent action spectrum) was recorded for IO-TiO2|dye|POs−PSII and relevant control samples (Fig. 3a) to characterize the complementary light absorption of the tandem photoanode (Supplementary Fig. 1). In a typical experiment, the wavelength was decreased from 760 to 420 nm (λscan) at Eapp = 0.5 V versus SHE while measuring the photocurrent. The action spectra were corrected to equal the photon flux at each wavelength and normalized (Supplementary Figs. 11 and 12).
In agreement with the photocurrent responses under full visible light irradiation (Fig. 2), IO-TiO2, IO-TiO2|POs, IO-TiO2|PSII and IO-TiO2|POs−PSII gave negligible photocurrents on monochromatic light illumination across all wavelengths (Fig. 3a). PSII-free electrodes loaded with dpp (IO-TiO2|dpp, IO-TiO2|dpp|POs) gave small photocurrent responses concurrent with the absorption spectrum of dpp, consistent with the assignment of background current due to dpp photooxidation (see above). For the functional IO-TiO2|dpp|POs−PSII tandem system, the photocurrent onset was observed at λscan = 620 nm, with a maximum photocurrent at approximately 560 nm. This photoresponse is consistent with the spectral overlap of PSII with dpp and the required simultaneous excitation of both photoactive components41. The absence of photocurrent at λscan > 620 nm is consistent with the requirement of dpp excitation for the electron injection into the CB of TiO2 and for POs oxidation. An external quantum efficiency21 of 2.7% was obtained at λmax = 560 nm (Ee = 6 mW cm−2).
Dual-wavelength action spectra were also recorded by coupling excitation by the scanned monochromatic light (λscan) to simultaneous irradiation at a fixed wavelength to continuously excite either PSII (λconst = 660 nm (Fig. 3b and Supplementary Fig. 12b)) or dpp (λconst = 523 nm (Supplementary Fig. 12c)). Continuous excitation of PSII (thereby probing dpp) in IO-TiO2|dpp|POs−PSII (Fig. 3b) led to an action spectrum profile (with regard to λscan) similar to the UV–vis spectrum of dpp and the single-wavelength excitation experiment. In comparison to the latter, a maximum at λscan = 550 nm was also observed, but with a 40% higher photocurrent magnitude. The overall photocurrent cross-section using dual excitation (Fig. 3b) was approximately two times higher compared to the cross-sections of individual components (Fig. 3a), which confirms the functional and efficient dual-absorber tandem mechanism in IO-TiO2|dpp|POs−PSII (Supplementary Fig. 1).
Continuous excitation of dpp (which probes PSII) (Supplementary Fig. 12c) resulted in a general increase in the photocurrent across all wavelengths (760–420 nm) compared to the single-wavelength excitation action spectrum. A new photocurrent maximum was detected at 680 nm, which corresponds to the PSII Qy band. A photocurrent maximum at 550 nm remained, which corresponded to the higher intensity excitation of the PSII/dpp spectral overlap region and led to a higher photocurrent compared to the single-wavelength experiment. Absorption at 480 nm, which corresponded to the excitation of the PSII β-carotene, and at ≤420 nm, which corresponded to the excitation of the Bx and By bands, were also observed. Action spectra of the RuP-sensitized photoanodes recorded for comparison (Supplementary Figs. 13 and 14) also correlated with the UV–vis absorption spectrum of RuP (Supplementary Fig. 6) and exhibited analogous features.
Bias-free overall water splitting via an artificial Z scheme
The negative Eonset and broad absorption spectrum of the IO-TiO2|dpp|POs−PSII tandem photoanode make it a suitable light absorber for bias-free (unassisted) overall water splitting. To achieve this long-standing goal16,42, the photoanode was wired to a previously reported indium tin oxide (ITO)-based IO-ITO|H2ase cathode3, which utilizes a reversible biological electrocatalyst for H2 production integrated in a hierarchically structured ITO electrode. The Desulfomicrobium baculatum [NiFeSe]-H2ase was used for its high proton reduction activity, O2 tolerance under reductive conditions and marginal inhibition by H2, which offer advantageous properties for water splitting compared to the O2-sensitive [FeFe]-H2ases available in algal H2 production43. ITO has been shown to be a suitable electrode material for the wiring of [NiFeSe]-H2ases in a direct electron-transfer regime, and the IO-ITO|H2ase electrode exhibited high current densities for proton reduction (>400 μA cm–2) and Eonset of −0.35 V versus SHE (pH 6.5, N2 atmosphere)3. Comparison of the voltammetric responses of IO-TiO2|dpp|POs−PSII and IO-ITO|H2ase measured individually (Supplementary Fig. 15) indicated that Eonset of the anodic (−0.50 V versus SHE) and cathodic (−0.35 V versus SHE) current responses overlap by approximately 0.15 V (refs 42,44). Thus, a two-electrode PEC cell that consists of the two enzyme-modified electrodes should be capable of bias-free solar-driven water splitting, assuming only minor resistive solution/membrane losses45. A comparison of the absolute currents indicates that IO-TiO2|dpp|POs−PSII should primarily limit the current response when wired to IO-ITO|H2ase.
A semi-artificial PEC device was therefore assembled that consisted of an IO-TiO2|dpp|POs−PSII photoanode connected to a IO-ITO|H2ase cathode separated by a glass frit membrane in a two-electrode, two-compartment cell. Figure 4 demonstrates the ability of the system to achieve bias-free solar-driven water splitting. Chronoamperometry measurements with longer irradiation times (Fig. 4a) were performed to minimize the charging effects below an applied voltage (Uapp) of 0 V. At more positive voltages, the charging effects decreased and photocurrent responses stabilized. On irradiation with ultraviolet-filtered simulated solar light, a current density of 28 ± 5 μA cm–2 was achieved at Uapp = 0 V (Fig. 4b). Voltage independent steady-state photocurrents (122 ± 21 μA cm−2) were reached at Uapp = 0.3 V. The photocurrent magnitudes were similar to those of a two-electrode system with a Pt cathode instead of an IO-ITO|H2ase (Supplementary Fig. 16), consistent with photocurrent limitation by IO-TiO2|dpp|POs−PSII. A two-electrode system with an IO-TiO2|H2ase cathode was also assembled and exhibited a similar behaviour, albeit with less charging due to the matched Fermi levels of IO-TiO2 (Supplementary Fig. 17).
The overall water splitting with the IO-TiO2|dpp|POs−PSII || IO-ITO|H2ase PEC cell was studied at Uapp = 0.0 and 0.3 V (Supplementary Fig. 18a). At zero bias (Uapp = 0 V), the initial photocurrent decayed from 130 μA cm–2 to 5 μA cm–2 after one hour of irradiation, which led to an average half-life time (τ1/2) of 6.5 minutes. At Uapp = 0.3 V, the photocurrent decayed from 140 μA cm–2 to 15 μA cm–2 after one hour of irradiation with a τ1/2 of ~8 min. These lifetimes are similar to those of previously reported PSII-based photoanodes3,4, and are consistent with the stability of PSII in vivo (τ1/2 of ~20 minutes)9. The relative stability of the IO-TiO2|dpp|POs−PSII system can be attributed to the efficient electron transfer through TiO2−dpp−POs−PSII, the physical stabilization of PSII by the polymer and the reduced accumulation of excited states in chlorophyll a within PSII46. However, it is important to emphasize that the current hybrid enzyme system is a proof-of-concept device, and its practical applicability is intrinsically limited by the photodegradation pathways of PSII in vitro.
After one hour of continuous light irradiation at Uapp = 0.0 V, H2 was detected (0.06 μmol H2 cm–2) with a Faradaic efficiency (ηF) of 76%, but a reliable O2 analysis was prevented by the detection limit of the apparatus. At Uapp = 0.3 V, O2 and H2 were quantified (Fig. 4c) with ηF = 88 ± 12% and 82 ± 10%, respectively (Supplementary Table 1), and a solar-to-hydrogen conversion efficiency of 0.14 ± 0.02% was obtained47. A PSII-based turnover frequency of 2.5 ± 0.3 mol O2 (mol PSII)−1 s−1 was calculated based on the quantified O2 and PSII3,4. Previously, similar ηF values were reported for benchmark PSII-photoanodes, diffusional-mediated IO-ITO|PSII3 and IO-ITO|POs−PSII4, but they required a significantly higher driving force (Uapp = 0.9 V and Eapp = 0.5 V versus SHE, respectively). Negligible photocurrents were detected in the control experiments (Supplementary Fig. 18b and Supplementary Table 1). O2 evolution was also confirmed using a rotating ring-disc electrode (RRDE) set-up (Supplementary Fig. 19).
The reported enzyme-based tandem PEC system that consists of an IO-TiO2|dpp|POs−PSII photoanode connected to a IO-ITO|H2ase cathode achieves the long-standing goal of a bias-free in vitro system for overall water splitting using PSII (O2 generation) connected to H2ase (H2 generation). This semi-artificial design addresses key limitations in biology as the PEC wiring of PSII to H2ase via an abiotic dye allows for (1) a panchromatic solar light absorption by using a synthetic green-light absorber (in contrast to non-complementary absorption by PSI), (2) a quantitative use of electrons extracted from PSII for H2 production (which thereby avoids inefficient metabolic pathways) and (3) separation of H2 and O2 gases in separate compartments (as opposed to inhibiting an O2-sensitive H2ase).
The tandem system produced H2 and O2 from water with high Faradaic efficiencies in a 2:1 ratio and presents an effective strategy to construct biotic–abiotic interfaces. Future work will involve investigating other dyes and replacing TiO2 with a semiconductor with a more negative CB potential to enhance the driving force for a more efficient catalysis or CO2 reduction chemistry. Moreover, our study provides a blueprint to advance future semi-artificial systems capable of bias-free solar fuel and chemical synthesis and a toolbox to develop proof-of-concept model systems for solar energy conversion.
2-(N-morpholino)ethane sulfonic acid (Alfa Aesar), tetrabutylammonium hydroxide (Sigma Aldrich), CaCl2 (Breckland Scientific), MgCl2 (Fisher Scientific), KCl (Alfa Aesar), KOH (Breckland Scientific), polystyrene (PS) beads (750 nm diameter, 2.6% w/v suspension in H2O (Polysciences Inc.), TiO2 nanoparticles (NPs) (Aeroxide P25 TiO2 NPs, 21 nm diameter, 80/20 anatase/rutile w/w (Evonik Industries)), ITO NPs (<50 nm diameter (Sigma Aldrich)), FTO-coated glass slides (8 Ω sq−1 (Sigma Aldrich)) and Parafilm (Sigma Aldrich) were purchased from commercial suppliers and used without further purification unless otherwise noted. Methanol, absolute ethanol, 2-propanol, dimethyl sulfoxide and THF (high-performance liquid chromatography grade) were purchased from Sigma Aldrich. PSII was isolated from the thermophilic cyanobacterium T. elongatus according to a previously reported procedure35, with an average oxygen-evolving activity of approximately 5,300 µmol O2 h−1 mg−1 of chlorophyll a. A stock PSII solution that contained 2.6 mg chlorophyll a ml−1 (83 μM PSII) was stored in a liquid N2 Dewar. [NiFeSe]-H2ase from D. baculatum was purified using a previously published method48, with a specific activity of 2,115 μmol H2 min–1 mg–1. Stock solutions of H2ase (8 μM in 20 mM Tris/HCl buffer, pH 7.0) were stored in 10–20 μl aliquots at −40 °C in an anaerobic glovebox and used immediately after thawing.
Polymer and dye synthesis
POs (ref. 37) was synthesized according to previously reported procedures4. In brief, an ethanolic solution of cis-[OsCl2(bipy)2] and a poly(1-vinylimidazole-co-allylamine) backbone (1/1.65 weight ratio) was stirred for 5 d at 90 °C. POs was then precipitated by the addition of diethyl ether, collected by centrifugation, washed thoroughly with diethyl ether, and dried under vacuum to obtain a reddish powder. An aqueous solution of POs (10 mg ml−1) was used in all the experiments. The dpp-based dye was synthesized using a previously reported procedure27. Briefly, pseudo-Stobbe condensation of 1-bromo-4-cyanobenzene with diethyl succinate was followed by lactam N-alkylation and then desymmetrization of the intermediate via a Suzuki–Miyaura cross-coupling. The phosphonic acid anchoring group was then added via Hirao cross-coupling using diethyl phosphite followed by hydrolysis. RuP29 was prepared as previously reported. Both dpp and RuP were characterized as previously reported.
The surface morphology of the electrodes was analysed by SEM (acceleration voltage 5 kV, working distance 5 mm (Philips SFEG XL30)), energy-dispersive X-ray spectroscopy attached to the SEM, STEM (TitanX 60–300) with a HAADF detector (acceleration voltage 300 kV) and powder X-ray diffraction (Empyrean 2 (PANalytics)). A centrifuge (5804 Eppendorf), furnace (ELF 11/14B/301 (Carbolite)) and ultraviolet/ozone cleaner (ProCleaner Plus (BioForce Nanosciences)) were used for the electrode preparation. UV–vis absorption spectra were recorded on an FS5 spectrofluorometer (integrating sphere reflectance mode (Edinburgh Instruments)) and spectrophotometer (Varian Cary 50 (Agilent)) using cuvettes with an optical path length of 1 cm (transmittance mode). With the integrating sphere, the measurement was performed by concentrating the light reflected from the electrode on the detector using a polytetrafluoroethylene-coated sphere (120 mm in diameter). The relative reflectance was measured with respect to the reflectance of the reference standard white board, which was taken to be 100%.
Preparation of IO-TiO2 electrodes
The IO-TiO2 electrodes were fabricated according to a method adopted from a previously reported procedure for the synthesis of IO-ITO3,4. FTO-coated glass slides (2 × 1 cm) were cleaned by sonication in two 30 min steps in 2-propanol and absolute ethanol. First, to ensure no direct contact of the electroactive components (PSII and POs) with the FTO layer, it was coated with a layer of mesoporous TiO2 (mesoTiO2). TiO2 NPs (50 mg) were dispersed via sonication for 20 min in a MeOH/water mixture (300 μl, 5:1 v/v). The suspension (10 μl) was deposited onto a 0.50 cm2 area defined by a Parafilm ring on an FTO slide and doctor bladed to give a 3-μm-thick mesoTiO2 layer. The IO-TiO2 layer was then deposited on top of the mesoTiO2 layer. TiO2 NPs (30 mg) were dispersed in a water/MeOH mixture (300 μl, 4:1 v/v) via sonication (3 h). The PS bead dispersion (1 ml) was centrifuged (10,000 revolutions per minute (r.p.m.) for 3 min), and the supernatant was removed. The pellet was redispersed in MeOH (1 ml) before being centrifuged again (10,000 r.p.m. for 0.5 min). The supernatant was removed and the TiO2 NP dispersion was added to the PS pellet. The pellet was dispersed into the solution by sonication (10 min at <10 °C). The resulting PS-TiO2 NP dispersion was drop cast (5 μl) onto a 0.25 cm2 area defined by a Parafilm ring on an FTO slide. After evaporation of the solvent, the electrodes were annealed at a 1 °C min−1 ramp rate from room temperature to 500 °C and sintered for 20 min to give a 20 μm thick IO-TiO2 film. The electrodes were allowed to cool to room temperature and cleaned with an ultraviolet/ozone cleaner (15 min) and characterized by SEM and HAADF–STEM, elemental mapping using energy-dispersive X-ray spectroscopy and powder X-ray diffraction (Supplementary Fig. 2).
Preparation of IO-TiO2|dye electrodes
IO-TiO2 electrodes with a pore diameter of 750 nm, a 20 μm film thickness and a geometrical surface area of 0.25 cm2 were used in all the experiments, unless stated otherwise. The IO-TiO2|dye-modified electrodes were prepared by soaking IO-TiO2 electrodes in solutions of dpp or RuP (0.15 mM in THF or H2O, respectively) overnight in the dark. To remove excess dye prior to the enzyme/polymer deposition, the IO-TiO2|dpp electrodes were rinsed with THF, followed by water and then air dried. Similarly, the IO-TiO2|RuP electrodes were rinsed with water and air dried.
Preparation of IO-TiO2|dye|POs–PSII electrodes
The IO-TiO2|dye|POs−PSII electrodes were prepared by depositing a blend of PSII (1 µl and 2.6 mg chlorophyll a ml−1) stock solution and POs (1 µl, 10 mg ml−1) onto the IO-TiO2|dye electrode (20 µm thick) and incubating the electrodes in the dark for 15 min at room temperature. Prior to the electrochemical studies, the IO-TiO2|dye|POs−PSII electrode was rinsed (3 × 500 µl) with the PSII buffer electrolyte solution (composition given below) to remove loosely bound species from the electrode surface.
Determination of PSII, POs and dye loading on IO-TiO2
The amount of PSII on the IO-TiO2 surface was quantified by scratching off the IO-TiO2 from the FTO glass substrate and washing it with MeOH (500 µl) to extract chlorophyll a (which originated from PSII) from the electrode surface into a centrifuge vial. The vial was centrifuged (10,000 r.p.m. for 1 min), and the UV–vis spectrum of the supernatant was recorded (Supplementary Fig. 4b). The band with an absorption maximum of λmax = 665 nm assigned to chlorophyll a (ε = 79.95 (chlorophyll a mg)−1 ml cm−1)49 was used to calculate the amount of PSII monomers49 assuming 35 chlorophyll a molecules per PSII monomer33. The Os-complex loading in the POs was determined by ICP-OES obtained by washing off the POs from the IO-TiO2 electrode with aqueous concentrated HNO3 solution (70 wt%) and measuring the concentration of the Os metal ions relative to the Os ICP standard (1 mg Os ml−1 in 20% HCl (Ricca Chemical)). The dpp/RuP loadings were quantified by scratching off the IO-TiO2|dye from the glass substrate and washing with tetrabutylammonium hydroxide (0.1 M) in MeOH (500 µl) to extract the dye from the electrode surface into a centrifuge vial. The vial was centrifuged (10,000 r.p.m. for 1 min) and the UV–vis spectrum of the supernatant was recorded (Supplementary Fig. 4a). The amount of dye desorbed into the solution was estimated using the Beer–Lambert Law (Supplementary Fig. 4c,d).
Preparation of IO-ITO|H2ase electrodes
[NiFeSe]-H2ase was immobilized on ozone-cleaned IO-ITO electrodes (20 μm film thickness, geometrical surface area, A = 0.25 cm2) by depositing the enzyme solution (5 μl) on the electrode surface, followed by incubation for approximately 5 min. The loading of [NiFeSe]-H2ase (40 pmol) was adjusted to the electrode thickness, as reported previously3.
Chronoamperometry and cyclic voltammetry measurements were performed using an Ivium Compactstat potentiostat and a gas-tight two-compartment glass cell with a water jacket for temperature control (T = 25 °C). A three-electrode set-up was employed with an IO-TiO2 (or IO-ITO) working electrode, a Ag/AgCl (3 M KCl) reference electrode and a Pt wire counter electrode separated by a glass frit in another compartment. The cell was filled with a PSII electrolyte solution (12 ml, pH 6.5) that consisted of CaCl2 (20 mM), MgCl2 (15 mM) and KCl (50 mM). All the current densities are reported with respect to the geometrical surface area of the electrodes (μA cm–2) . Experimentally measured potentials are reported versus SHE using the conversion ESHE = EAg/AgCl + 0.197 V (25 °C).
Bias-free PEC overall water splitting and O2 and H2 quantification were studied using a two-electrode configuration with an IO-TiO2 photoanode that contained PSII and an IO-ITO|H2ase cathode. In all the experiments that involved the IO-ITO|H2ase cathode, 2-(N-morpholino)ethane sulfonic acid (40 mM) was added to the PSII electrolyte solution as it was found to retain the H2ase electroactivity and only caused negligible photocurrents from 2-(N-morpholino)ethane sulfonic acid oxidation.
PEC experiments were performed using a Xe lamp (150 W (Newport)) solar light simulator (Ee = 100 mW cm−2, AM 1.5 G filter, λ > 420 nm filter (LOT Quantum Design)). Action spectra were recorded with a Xe lamp (300 W (Newport)) solar light simulator (LOT Quantum Design) coupled to a monochromator (MSH300 (LOT Quantum Design)). For the dual excitation experiments, an Ivium ModuLight light-emitting diode module (λ = 460/523/660 nm; 4 mW cm−2) was used as a second light source. Light intensity was measured as a function of wavelength with a thermal sensor (S302C (Thorlabs)) and power meter console (PM100D (Thorlabs)).
IO–TiO2|dye|POs−PSII electrodes were exposed to dark and light cycles in the PF-PEC measurements. To avoid an overestimation of the photocurrent, the photocurrent response was defined as the baseline-corrected (dark-current subtracted) photocurrent peak shoulder edge after light exposure9. Action spectra were corrected to equal the photon flux at each wavelength and normalized (taking the peak at 500 nm as unity). Error bars are ±s.d. of the sample estimated from at least three experiments. All the data processing was performed using OriginPro 9.1 software.
RRDE experiments were performed using a Ag/AgCl (3 M KCl) reference and glassy carbon counter electrodes. The disk/ring apparatus was embedded in a cylindrical polyether ether ketone housing. A Pt ring electrode surrounded a glassy carbon disk electrode on which a mesoTiO2 layer was deposited prior to depositing additional components (dpp, POs and PSII). The electrodes were placed in a single-compartment cell under continuous purging with N2. The ring/disk electrode was rotated at 400 r.p.m. for the O2 evolution measurements and the collection efficiency of the ring electrode was calibrated under conditions similar to those of the experiment (Supplementary Fig. 19).
Quantification of O2 was performed using a fluorescence-based O2 sensor (Ocean Optics FOSFOR-R probe (Neofox) inside an anaerobic glovebox (Belle Technology) to avoid the ingress of atmospheric O2. The probe was placed inside the anodic compartment headspace of a two-compartment cell protected from direct irradiation. The results are reported as the average of six measurements and the background signal was subtracted from all the measurements. The reported O2 values were corrected for dissolved O2 using Henry’s law (KH(O2) = 769.23 l atm mol−1 at 25 °C). H2 was analysed by taking aliquots of the headspace gas (50 μl) after electrolysis and quantified using gas chromatography (carrier gas N2, flow rate 0.7 ml min−1, molecular sieve column, thermal conductivity detector (Agilent 7890)). Calibration runs were performed to quantify the volume of H2 evolved. The reported Faradaic efficiencies were corrected by subtracting the dpp dye background current estimated from control experiments from the experimental data.
The data that support the presented plots within this paper and other findings of this study are available at the University of Cambridge data repository with the identifier https://doi.org/10.17863/CAM.25834.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work was supported by an ERC Consolidator Grant MatEnSAP (682833), the UK Engineering and Physical Sciences Research Council (EP/L015978/1 and EP/G037221/1, nanoDTC and a DTA studentship), the Christian Doppler Research Association, the OMV Group and a Royal Society Newton International Fellowship (NF160054), the Cluster of Excellence RESOLV (EXC 1069) funded by the Deutsche Forschungsgemeinschaft and the European Union’s Horizon 2020 MSCA ITN-EJD 764920 PHOTOBIOCAT. The HAADF–STEM was carried out at the National Center of Electron Microscopy (NCEM), which is supported by the Office of Science, Office of Basic Energy Sciences of the US Department of Energy under Contract no. DE-AC02-05CH11231. Work at the Molecular Foundry was supported by the Office of Science, Office of Basic Energy Sciences of the US Department of Energy under Contract no. DE-AC02-05CH11231. We thank J. Fontecilla-Camps and C. Cavazza for providing the H2ase enzyme, V. Hartmann for his contribution to the PSII preparation and N. Plumeré, C. Creissen, S. Kalathil and N. Heidary for valuable discussions.
About this article
Nature Communications (2018)