Photoelectrochemical water splitting in an organic arti ﬁ cial leaf †

Photoelectrochemical water splitting is demonstrated in an organic arti ﬁ cial leaf composed of a triple junction polymer solar cell for light absorption and charge generation and low-overpotential catalytic electrodes for hydrogen and oxygen evolution. For small area solar cells (<0.1 cm 2 ), a solar to hydrogen conversion e ﬃ ciency of 5.4% is obtained using RuO 2 catalysts. Using earth-abundant NiMoZn and Co 3 O 4 catalysts for hydrogen and oxygen evolution, the e ﬃ ciency is 4.9%. For larger area (1.7 cm 2 ) solar cell devices the solar to hydrogen e ﬃ ciency with RuO 2 catalysts reduces to 3.6% as a consequence of an increased overpotential for water splitting. This shifts the operating point of the water splitting device beyond the maximum power point of the solar cell and reduces the photocurrent.


Introduction
Storage of solar energy is important to counterbalance the intermittency of solar electricity supply and demand.Capturing solar energy in chemical bonds of molecular fuels is most effective in terms energy density and the successful construction of a direct articial system for efficient solar fuel generation is an important challenge for science and engineering.Solar fuels are attracting considerable attention recently and solutions are emerging on how this can be achieved. 1,2Solar to chemical energy conversion requires three concerted steps: (1) absorption of light, (2) creation of charges (electrons and holes) with an appropriate chemical potential to enable (3) catalytic chemical reactions in which the charges are used to oxidize and reduce compounds in endothermic reactions such as the splitting of water and the reduction of carbon dioxide. 1 To enable solar energy production in yields exceeding the energy conversion of natural photosynthesis (typically <1%) with cheap and abundant materials is a tremendous challenge. 3Photoelectrochemical (PEC) water splitting requires a theoretical potential of E 0 H2O ¼ 1:23 V, but in practice occurs at a potential (V H 2 O ) higher than E 0 H2O due to overpotential losses (h) occurring at the electrodes ðV Depending on the type of the electrodes and catalysts, the electrolyte, and the current density, overpotentials for both hydrogen and oxygen evolution vary and the electrolysis potential lies typically in the range of 1.4-1.9V. 4,5 The principle of solar energy driven water splitting has previously been described using inorganic solar cells with suitable catalysts for hydrogen and oxygen production.Early publications include the monolithic photoelectrochemical water splitting devices of Lin et al. 6 based on a triple stack amorphous silicon solar cell with RuO 2 and Pt electrodes and of Turner et al. 7 based on tandem GaAs/InGaP 2 tandem photocathode and a Pt electrode.One of the most efficient water splitting devices was published by Licht et al., 8 who reached a solar to hydrogen conversion efficiency of h STH ¼ 18.3% with a Si/Al 0.15 Ga 0.85 As tandem cell in combination with Pt and RuO 2 electrodes for hydrogen and oxygen evolution.Based on a similar concept using a GaInP/GaInAs dual junction cell in an optical concentrator system with a polymer electrolyte membrane electrolyser Wittstadt et al. demonstrated solar water splitting with 18% efficiency in an integrated system. 9ore recently, Nocera et al. 10 reported h STH ¼ 2.5% for a wireless stand-alone device based on a triple junction solar cell with earth-abundant nickel-molybdenum-zinc (NiMoZn) and cobalt oxide cubane (CoO) catalysts for hydrogen and oxygen evolution.van de Krol et al. reached h STH ¼ 4.9% based on a semi-transparent tungsten doped bismuth vanadate (W:BiVO 4 ) photoanode that was optically and electrically connected to an amorphous silicon tandem solar cell with cobalt phosphate and Pt catalysts. 11Grätzel et al. have reported h STH ¼ 12.3% based on two series connected lead perovskite solar cells and a NiFe layered double hydroxide for hydrogen and oxygen evolution. 12e were interested to explore the feasibility of an organic articial leaf by combining organic solar cells with suitable catalysts for water splitting.Organic and polymer solar cells produce electrical power directly by converting sunlight.7][18] Designing an organic articial leaf capable of producing hydrogen directly from sunlight requires the selection and optimization of a number of parameters.To minimize losses, the solar cell should operate close to the maximum power point and for efficient water splitting the maximum power point voltage (V max ) of the cell should be designed such that its value matches with the potential for water splitting, V H 2 O , under the relevant working conditions, related to choice of electrodes, electrolyte, and current density.With a required V max z V H 2 O ¼ 1.4-1.9V, single junction or series connected tandem polymer solar cells are generally not sufficient for the water splitting reaction to take place.However, a series connected triple junction polymer solar cell can provide the required potential at its maximum power point as we have shown recently. 16n important aspect of photoelectrochemical water splitting devices is the selection of appropriate catalysts for oxygen and hydrogen evolution reactions.The operating potential during water splitting mainly depends on this selection.The best catalysts in terms of lowering the overpotential are based on precious transition metals or their oxides.0][21][22][23] Besides overpotential, the compatibility of the hydrogen and oxygen evolution catalysts with each other in the selected electrolyte is very important.In general, catalysts operate better in highly acidic or highly alkaline media compared to more neutral pH conditions.Highly acidic or alkaline conditions, however, affect the stability of the catalyst and the sealing of an integrated photoelectrochemical device when kept in contact with the electrolyte for a long time.Preferably, the hydrogen and oxygen evolution catalysts are formed from earth-abundant materials and should be able to operate at near neutral pH conditions.
Here we present the design and characterization of a photoelectrochemical articial organic leaf that integrates an organic triple junction solar cell with catalysts for hydrogen and oxygen evolution.We demonstrate photoelectrochemical water splitting with a solar to hydrogen conversion efficiency of 5.4% with low-overpotential ruthenium oxide (RuO 2 ) catalysts and of 4.9% efficiency with earth-abundant cobalt oxide (Co 3 O 4 ) and NiMoZn catalysts.To ensure a low overpotential, the catalyst to solar cell area ratio is about 15-20.In an alternative third conguration we use a larger area triple junction solar cell with RuO 2 similar sized catalyst surface areas, resulting in h STH ¼ 3.6%.

Catalysts and electrolytes
6][27] The electrocatalytic properties of RuO 2 have been studied for a signicant time, 28 not only for oxygen evolution but also as employing the oxide as a hydrogen evolution catalyst. 29,30Ruthenium oxide can be deposited on various substrates mainly via two methods: electrodeposition or thermal decomposition. 28As indium tin oxide (ITO) is commonly employed as electrode in polymer solar cells, it is convenient to apply the catalyst layers directly onto ITO to create a monolithic device.Even though it is possible to electrodeposit RuO 2 on ITO, 31,32 we observed that the electrodeposited layers were not very stable when used for oxygen evolution.Another method for depositing ruthenium oxide is thermal decomposition of RuCl 3 from an aqueous solution of RuCl 3 , followed by annealing at elevated temperatures. 25In our experiments, the RuO 2 catalyst on transparent conductive substrates tended to delaminate in time especially during active gas evolution periods, but attached better to metal substrates.
In this study, ruthenium oxide was deposited onto titanium substrates through thermal decomposition of RuCl 3 . 25The details are described in the Experimental section.The procedure gives reproducible and stable performance over a couple of hours.The activity of RuO 2 both as oxygen and hydrogen evolution catalysts is shown in a Tafel plot in Fig. 2. RuO 2 on a Ti substrate gives an overpotential of less than 130 mV for hydrogen evolution and less than 315 mV for oxygen evolution for current densities up to 10 mA cm À2 in a 1.0 M KOH electrolyte.RuO 2 is actually a remarkably good electrocatalyst for hydrogen evolution as a result of an activation that occurs under reductive conditions and that is considered as an intrinsic property of the oxide. 28,33or the selection of earth-abundant catalysts, we considered that choice for the catalyst for oxygen evolution is more important than the catalyst for hydrogen evolution, because the overpotentials for oxygen are generally higher.One of the earthabundant oxygen evolution catalysts, which has been extensively studied lately, is cobalt oxide.Two well-known forms of this oxide are cobalt borate (CoBi) 34 and cobalt phosphate (CoPi). 35These catalysts are deposited electrochemically and have been used for photoelectrochemical water splitting. 10,11obalt oxide can also be deposited on conductive surfaces in the form of nanoparticles for oxygen evolution. 36In this method the synthesized nanoparticles are dispersed in methanol and deposited on ITO through drop casting and thermal annealing (see Experimental section for details).The cobalt oxide catalyst can operate in an electrolyte of 0.1 M potassium borate (KBi) at pH 9.2, which is crucial for the catalyst stability.The electrochemical activity of cobalt oxide nanoparticles is shown in Fig. 2. The catalyst gives an overpotential up to 410 mV for oxygen evolution for current densities up to 10 mA cm À2 .This performance is comparable to the CoBi and CoPi catalysts. 34,35ue to reproducibility and ease of processing, cobalt oxide nanoparticles were selected as the oxygen evolution catalyst for the desired water splitting device.
The selection of a suitable earth-abundant hydrogen evolution catalyst also requires ease of processing, reproducibility and ability to operate in an electrolyte of near neutral pH conditions.8][39][40] Among the available options of hydrogen evolution catalysts, many nickel compounds are used under harsh conditions with pH values above 13 or below 1. [37][38][39][40] However, the NiMoZn catalyst developed by Nocera et al. 10 was shown to work well in KBi  electrolyte where Co 3 O 4 nanoparticles can also operate.Hence, the NiMoZn alloy was selected as the hydrogen evolution catalyst.The catalyst is made by electrodeposition onto a nickel substrate (see Experimental section for details). 10The Tafel plot for NiMoZn in 0.1 M KBi for hydrogen evolution shows that the catalyst requires very low overpotentials of about 100 mV for current densities up to 1 mA cm À2 , but that the overpotential increases signicantly when the current density is increased.
Photoelectrochemical water splitting with RuO 2 /RuO 2 catalysts at low current densities For achieving a high h STH , a larger catalyst surface area provides smaller current densities and hence smaller overpotential.A triple junction polymer solar cell with an active area of 0.0676 cm 2 was connected to RuO 2 catalysts for oxygen and hydrogen evolution submerged in a 1.0 M KOH electrolyte.The catalyst areas of 1.1 and 1.3 cm 2 , respectively, were obtained by drop casting of the RuCl 3 precursor solution.Fig. 3a and b show the time evolution of the voltage and current density of photoelectrochemical water splitting device under illumination by simulated AM 1.5G light.The three curves in Fig. 3a correspond to the J-V characteristics of the triple junction cell measured just before, during, and directly aer a 20 min water splitting experiment.The power conversion efficiency (PCE) of the solar cell just before the water splitting experiment is $6.7% with a V max ¼ 1.44 V.The PCE of 6.7% is less than the previously reported value of 9.6%, 24 mainly because the batch of PCDTBT polymer used in this study gave inferior performance.
The voltage and current density measured during water splitting follow the J-V curve of the solar cell until the operating point stabilizes (Fig. 3a).The stabilization takes roughly 15 min and it is mainly due to charging of the double layer on the catalyst surfaces and back reaction of the reaction intermediates on or in the neighbourhood of the catalysts (Fig. 3b). 41In this specic case it takes longer because of the very low current density on the catalyst surfaces.Aer reaching stabilization, a slight decrease in the operating current density is observed for the following 5 min.The decrease is attributed mainly to the degradation of the solar cell during operation, as evidenced from the solar cell performance just aer the water splitting measurements (Fig. 3a).Solar to hydrogen conversion efficiency can also be affected by the slow degradation of the RuO 2 catalysts because it is known that the electrolyte used (1.0 M KOH) does not allow for stable catalyst performance in the long run. 42n our experiments, however, the decrease in catalyst performance over time turned out to be marginal.
Fig. 3b shows that the stable operation takes place at around 1.49 V at a solar cell current density of 4.40 mA cm À2 .At an operating potential of V op ¼ 1.49 V, the total overpotential for hydrogen and oxygen evolution is 0.26 V, in excellent agreement with the value expected from the Tafel plots for the corresponding current densities on the catalyst surfaces of 0.03 V and 0.23 V for hydrogen and oxygen evolution reactions, respectively (shown with the O symbols in Fig. 2).It is important to highlight that, the operating point during water splitting is very close to the maximum power point of the solar cell (V max ¼ 1.44 V), which enhances the efficiency of the PEC water splitting device.The solar cell efficiency in the operating point is estimated as V op Â J op /P in ¼ 6.6%, close to the maximum PCE of 6.7%.Assuming 100% Faradaic efficiency, the photocurrent in the operating point J op (4.40 mA cm À2 ) allows to estimate the solar to hydrogen efficiency as: h STH ¼ 1.23 Â J op /P in ¼ 5.4%.Separate electrochemical experiments in which evolved gassed were collected, showed a 1 : 2 volume ratio of O 2 to H 2 and Faradaic efficiencies higher than 88%.In this PEC device with low current densities at the electrodes, hydrogen and oxygen bubbles during water splitting were visualized using a high resolution camera.

Photoelectrochemical water splitting with Co 3 O 4 /NiMoZn catalysts at low current densities
To realize solar to hydrogen conversion with the earth-abundant catalysts, a 0.0676 cm 2 triple junction polymer solar cell was combined with Co 3 O 4 and NiMoZn catalysts with surface areas tuned to 1 cm 2 .Fig. 3c shows the J-V curves of the solar cell just before, during, and aer 20 min of water splitting with Co 3 O 4 / NiMoZn catalysts submerged in a 0.1 M KBi electrolyte at pH ¼ 9.2.The simultaneous measurement of the transient current density and voltage during water splitting is shown in Fig. 3d.In this case the stabilization of the operating point takes about 3 min.We attribute the reason of shorter stabilization period to the different nature of catalysts and the electrolyte.Another reason can be the difference between the geometrical and actual surface areas of the catalysts.The RuO 2 catalysts are placed on very rough titanium substrates while both Co 3 O 4 and NiMoZn layers are deposited on smoother surfaces.Higher roughness would mean higher surface area, which will require more time to form the double layer on the catalyst surfaces.During 20 min of PEC water splitting, a slight decrease in the operating current density is observed together with a slight increase in operating voltage.These are mainly due to the degradation of the solar cell that is more prominent in the case of this specic solar cell (Fig. 3c).
The triple junction solar cell combined with earth-abundant catalysts had a PCE of 6.5%.The operating point of the water splitting device stabilized at V op ¼ 1.56 V.This operating voltage is again close to V max ¼ 1.45 V.The V op ¼ 1.56 V suggests a total overpotential of 0.33 V, which is less than the value of $0.46 V expected from the Tafel plots.The difference might be related to different concentrations of dissolved hydrogen and oxygen in the electrolyte during Tafel plot measurements.Aer 15 min we nd J op ¼ 3.98 mA cm À2 , providing a PCE of 6.1% and h STH ¼ 4.9% in the operating point.

A larger area organic leaf
For a more realistic estimation of h STH , the catalyst surface areas should be in the same range as the solar cell surface.Therefore, we constructed an organic leaf with a larger area triple junction solar cell integrated with RuO 2 /RuO 2 catalyst for hydrogen and oxygen evolution on Ti substrates (Fig. 4a).The solar cell area was 1.7 cm 2 and the catalyst surface areas are $1.2 cm 2 each.To reduce the potential drop over the ITO front surface of this larger area solar cell, the current was collected by a boundary metal contact (Fig. 4b, le).This metal border also denes the nominal solar cell area.Aer completion, the solar cell was sealed with a second glass plate and a modied epoxy resin (Fig. 4b, middle) and then titanium substrates with the RuO 2 catalysts are attached at the back side with a glue and graphite conductive adhesive (Fig. 4b, right).
Fig. 3e shows that the l.7 cm 2 triple junction solar cell has a PCE of 6.1%, which is somewhat less than the PCEs of the small area cells (compare panels (a) and (c) in Fig. 3) due to a slight reduction in short-circuit current and ll factor.For PEC water splitting, the triple junction solar cell was coupled to the two RuO 2 catalysts for oxygen and hydrogen evolution in 1.0 M KOH electrolyte.Evolution of hydrogen and oxygen was easily observed by the eye (Fig. 3f).Fig. 3e shows the J-V-voltage characteristics measured during water splitting.Aer 15 min operation, the operating point of the articial leaf is V op ¼ 1.67 V and J op ¼ 2.94 mA cm À2 .The current stabilization in this conguration takes less than a minute and is much faster than for the smaller area cells due to the high current density on the catalyst surfaces.Aer stabilization, the operating point does not signicantly change over the course of a 20 min measurement (Fig. 3f).The slight decrease in current density can be attributed to the degradation of the solar cell (Fig. 3e).
The maximum power point of this specic solar cell is at V max ¼ 1.40 V while the operating point during water splitting is now at V op ¼ 1.67 V.The latter is a direct consequence of the increased current density experienced by the catalysts and results in higher overpotentials both for hydrogen and oxygen evolution (see B markers in Fig. 2) and a higher operating potential.The total overpotential of 0.44 V in the operating point is 0.04 V higher than the value expected from the Tafel plots (Fig. 2).At V op ¼ 1.67 V, the power output of the solar cell is 4.91 mW cm À2 , which is signicantly less than the maximum power of 6.05 mW cm À2 that can be delivered by the cell.As V op is now signicantly larger than V max , the photocurrent is reduced considerably to J op ¼ 2.94 mA cm À2 .As a result, h STH is 3.6%.The loss in h STH from 5.4% to 3.6% by increasing the solar cell area is signicant, and solely due to the increased overpotential.This results in a signicant reduction of the operating current for this particular triple junction cell.This demonstrates that designing an efficient articial leaf requires a subtle balance between V op and V max .For V op > V max , a signicant loss in photocurrent can be expected.When V op < V max , the photocurrent density actually is increased but because the efficiency of a solar cell is always a trade-off between current density and voltage, operating a cell too far below from the maximum power point represents an avoidable loss.
We also constructed a large area organic articial leaf with earth abundant catalysts.In this case the triple junction polymer solar cell with a $1.2 cm 2 area was integrated with Co 3 O 4 and NiMoZn catalysts operating in 0.1 M KBi electrolyte.The 20 min water splitting experiments revealed a h STH ¼ 1.3% (Fig. S1,  ESI †).This signicantly lower efficiency is partially due to a large operating voltage of V op ¼ 1.83 V (Fig. S2, ESI †), but also due to faster degradation of this specic solar cell used in this experiment (Fig. S1, ESI †).
The stability of the RuO 2 /RuO 2 and Co 3 O 4 /NiMoZn catalysts used in the PEC water splitting devices were tested with twoelectrode measurements at applied potentials of 1.65 V and 1.85 V for 20 min.The applied potentials were selected with respect to the operating potentials of large area articial leafs.The results (Fig. S3, ESI †) show that both catalyst couples do not show substantial degradation on this time scale.The catalyst stability is important for future and long term application.Several investigations have addressed the stability of the catalysts under the conditions used in this work.The intrinsic activity and stability of RuO 2 for oxygen and hydrogen evolution reactions in alkaline electrolytes has been described in detail, showing that dissolution of RuO 2 during oxygen evolution is limiting the stability but that it is table during hydrogen evolution. 43For Co 3 O 4 nanoparticle a high electrocatalytic stability in alkaline conditions has been reported for oxygen evolution. 44Likewise, the NiMoZn electrode in 0.1 M KBi at pH ¼ 9.2 showed no appreciable degradation during more than 150 h operation. 45 further level of integration is shown in Fig. 5, where a larger area organic leaf is fabricated with dual RuO 2 catalysts electrodeposited on ITO.At this level of integration it is no longer possible to measure J-V characteristics, but the evolution of hydrogen and oxygen evidences the functionality of the device.

Conclusions
We have demonstrated efficient photoelectrochemical solar to hydrogen conversion using an organic articial leaf based on an organic triple junction solar cell and transition metal and metal oxide electrocatalysts.A solar to hydrogen conversion efficiency of h STH ¼ 5.4% was obtained with RuO 2 catalysts and h STH ¼ 4.9%with earth-abundant Co 3 O 4 /NiMoZn catalysts.In these  examples the overpotential was kept low, by a large catalyst to solar cell catalyst to solar cell area ratio of $15.For a lower surface area ratio ($0.7), h STH is reduced to 3.6%, mainly because the 0.18 V increase in overpotential that originates from the $20 times higher electrocatalytic current density moves the operating point of the articial leaf too far away from the maximum power point of the solar cell and results is a signicant decrease in photocurrent density.
This work demonstrates that for efficient articial leafs, balancing the nature and surface area of the catalysts with the materials used in the solar cell is crucial.In this respect organic solar cells offer an advantage for designing photoelectrochemical water splinting devices because a wide choice in organic semiconductors is available, which enables tuning the maximum power point voltage (V max ) to coincide with the operating point (V op ) determined by the thermodynamic potential for water splitting and the overpotentials dened by catalysts.By optimizing these parameters, a signicant progress in the performance of organic articial leaves can be achieved.Further improvements of the present system should also focus on improving the stability of triple junction polymer solar cells and the catalysts.

RuO 2 catalysts
RuO 2 was prepared through thermal decomposition of RuCl 3 .RuCl 3 was dissolved in ultra-pure water to a concentration of 0.2 M and 200 mL of this solution was placed onto a pre-cleaned and air plasma treated titanium substrate to form a catalyst area of about 1.3 cm 2 .The substrate was then dried on a 90 C hot plate for 20 min and then oxidized in a 350 C oven for 3 h.The same procedure was used for oxygen and hydrogen evolution RuO 2 catalysts.X-ray diffraction (XRD) conrmed the formation of crystalline RuO 2 on Ti (Fig. S4, ESI †).

Co 3 O 4 nanoparticles
The procedure for the synthesis of the Co 3 O 4 nanoparticles was adapted from literature. 36Under constant magnetic stirring at 45 C, cobalt(II) acetate tetrahydrate (0.5 g) was dissolved in a mixture of ultra-pure water (2 mL) and ethanol (23 mL).Aer 15 min, ammonia (25+%, 3.3 mL) was added dropwise.The reaction mixture was heated to 80 C and kept there for 3 h under reux to enable the formation of the nanoparticles.To this crude product mixture, acetone (100 mL) was added to start precipitation.To improve separation, the mixture was centrifuged for 20 min at 2000 rpm.Aer decanting the solvent, methanol (12 mL) and acetone (120 mL) were added to the precipitate, followed by centrifugation at 2000 rpm for 20 min.The precipitated particles were redispersed in methanol (>25 mL).The Co 3 O 4 nanoparticles were characterized with UVvis spectroscopy and transmission electron microscopy and found to be 3-5 nm in size (Fig. S5 and S6, ESI †).

Co 3 O 4 electrode
A solution of the Co 3 O 4 nanoparticles (250 mL, circa 3 mg mL À1 ) was placed onto a pre-heated ITO coated glass slide at 110 C to form a catalyst area of around 1 cm 2 .Aer 5 min, the covered ITO slide is heated with a heat gun with a power of 2000 W for several minutes to improve binding of the nanoparticles to the substrate.The Co 3 O 4 electrode was characterized with scanning electron microscopy (SEM), energy-dispersive X-ray (EDX), and XRD (Fig. S7-S9, ESI †).

NiMoZn electrode
The procedure for the preparation of a NiMoZn electrode was adapted from literature. 10 In ultra-pure water (100 mL), nickel(II) sulfamate hydrate (1.309 g), sodium molybdate dihydrate (0.460 g), zinc chloride (6 mg), sodium pyrophosphate (3.460 g), and sodium bicarbonate (7.5 g) were dissolved under constant magnetic stirring.To circa 30 mL of this solution, a few drops of hydrazine hydrate were added just before deposition.The substrate, a nickel foil, was pretreated in diluted sulfuric acid at À2.0 V versus Ag/AgCl for 3 min, without iR correction.Electrodeposition was carried out in a one-compartment electrochemical cell for 60 min at 0.0775 A cm À2 versus Ag/AgCl.Two nickel electrodes were used as working and counter electrode.Aer deposition, the catalyst lms on the counter electrode were allowed to leach overnight in 10 M KOH.The overpotential of the NiMoZn electrode is signicantly smaller than that of the Ni foil (Fig. S10, ESI †).

Electrochemical measurements
Tafel plots were constructed from the cyclic voltammetry (CV) measurements performed in the stationary current mode with 50 mV steps taking 300 s per step.Two cycles were performed for each catalyst and the second cycle was used to form the Tafel plots.In all cases, a Pt disk and Ag/AgCl electrode were used respectively as counter electrode and reference electrode.A three-compartment cell was used during measurements, where the compartments for working and counter electrodes are separated with a ne porosity glass frit.The compartment with the reference cell was connected to the working electrode compartment using a Luggin capillary.For Co 3 O 4 , the measuring range was selected starting above 0.60 V versus Ag/ AgCl, for NiMoZn beneath À0.7 V. Owing to a different electrolyte and different catalyst performance, these values were selected as above 0.25 V and below À0.95 V when RuO 2 is used as oxygen and hydrogen evolution catalysts respectively.The electrolytes were not stirred during measurements.
Tafel plots were constructed by characterizing each catalyst in a three electrode CV measurement.The measured potential (E meas ) was converted into the overpotential h via: vs. RHE is the potential difference between the reference and the reversible hydrogen electrode (RHE), E 0 is water oxidation or hydrogen reduction potential (1.23 V or 0.00 V), i is the current and R u the uncompensated resistance between the working and the reference electrode.The value for R u was determined by electrochemical impedance spectroscopy.The measurement was done in the frequency range from 10 kHz to 100 mHz.All measurements were done with respect to an Ag/AgCl (3 M KCl) reference electrode in media with different pH such that: E ref. vs. RHE ¼ 0.210 + 0.059 Â pH.

Device preparation
The small area triple junction solar cells were prepared by rst spin casting PEDOT:PSS (Clevios® P VP AI 4083, H. C. Starck) in air onto pre-cleaned glass substrates with indium tin oxide (ITO) patterns (Naranjo Substrates).Then, the PEDOT:PSS layer was dried at 140 C for 10 min.On top of the dried PEDOT:PSS the PCDTBT:[70]PCBM blend was spin cast in nitrogen atmosphere.The PCDTBT active layer was dried on a hot plate for 10 min at 70 C. Aerwards, the ZnO, pH-neutral PEDOT and PMDPP3T:[60]PCBM layers were spin cast sequentially to form the middle cell.This process was then repeated to form the back cell.For optimal performance, the ZnO layer was deposited in nitrogen atmosphere, while pH-neutral PEDOT and PMDPP3T:[60]PCBM layers were spin cast in air.The triple junction devices were completed by thermal evaporation of 1 nm LiF and 100 nm Al at 3 Â 10 À7 mbar.
The front cell was spin cast from a warm solution of PCDTBT and [70]PCBM (1 : 4 w/w) in chlorobenzene at 7 mg mL À1 polymer concentration.The middle and the back cells were spin cast from a solution of PMDPP3T and [60]PCBM (1 : 3 w/w) in chloroform containing 7.5% (v/v) o-DCB at 4 mg mL À1 polymer concentration.ZnO nanoparticles of $5 nm diameter were spin cast from a solution of 10 mg mL À1 ZnO in isopropanol (IPA).pH-neutral PEDOT was prepared by diluting Neutral pH PEDOT NT5/CH03311/BH from Agfa with ultra-pure water at a 1 : 1 volume ratio and adding 0.2 mL/mL IPA to improve the wetting on ZnO nanoparticles.The solution was then ltered with a 5.0 mm Whatman Puradisc FP30 syringe lter.
The large area solar cell was manufactured by rst etching away a stripe of ITO from a full-ITO covered substrate to form two separate electrode areas.The etching was performed using zinc dust and hydrochloric acid.Aerwards the substrates were cleaned thoroughly and a 100 nm of aluminum frame was evaporated to improve charge collection and specify the solar cell area.Deposition of the remaining layers was performed as explained above.The back electrode of LiF (1 nm) and Al (100 nm) was evaporated in such a way that it makes contact with the other half of the ITO.As a result, both holes and electrons can be collected from the ITO layers.
The large area stand-alone device was prepared by rst sealing the solar cell inside a nitrogen atmosphere with a UVcuring resin.This resin was allowed to cure for 15 min under a 365 nm UV lamp.Subsequently, the titanium substrates with RuO 2 catalysts were glued at the back side of the solar cell onto the glass cover using a two-component glue (Bison-Kombi Snel®).These catalyst layers were connected to the positive and negative poles of the solar cell with graphite conductive adhesive.

Characterization
The characterization of the solar cells was rst done inside the glove-box with nitrogen atmosphere.A Keithley 2400 sourcemeasurement unit was used to measure current density to voltage (J-V) characteristics of the devices.The illumination was carried out with $100 mW cm À2 white light from a tungstenhalogen lamp ltered by a Schott GG385 UV lter and a Hoya HMC 80A 72 mm daylight lter.No mismatch correction was performed.The measurements were performed inside a glove box with a nitrogen atmosphere.The triple junction devices were exposed to UV illumination (with a Spectroline EN-160L/F 365 nm lamp from Spectronics Corporation) for about 10 min to provide an ohmic contact between the ZnO and pH-neutral PEDOT layers before being measured.To prevent parasitical charge collection due to the high lateral conductivity of pHneutral PEDOT, triple junction devices were measured with a mask slightly smaller than the actual device area, which is determined by the overlap of the ITO and Al electrodes.
Solar to hydrogen conversion efficiencies were determined using a home-built setup.As the water splitting experiments took place in air, the solar cells were placed in a nitrogen lled box and connected to the catalysts through external cables.The solar cell was illuminated with white-light from a tungstenhalogen lamp ($100 mW cm À2 ) ltered by a Schott GG385 UV lter and a Hoya HMC 80A 72 mm daylight lter.The solar cell was placed such that the generated short-circuit current in this setup corresponded to the short-circuit current measured inside the glove box, which corresponds to AM 1.5G power standards.A Keithley 2600 source-measurement unit was used for simultaneous measurement of current and voltage during water splitting.
The small-area (0.0676 cm 2 ) solar cells provide low current densities on the catalyst surfaces ($1.2 cm 2 ), which makes it difficult to observe the hydrogen and oxygen bubbles during water splitting with by eye.Using a high resolution camera during water splitting traces of bubbles that are not apparent to eye were easily visualized.For the RuO 2 /RuO 2 catalysts movements of tiny gas bubbles were observed, while in the case of Co 3 O 4 /NiMoZn catalysts, bubbles growing on the Co 3 O 4 catalyst surface were seen.Water splitting experiments with the standalone large area device were made by placing the device inside a glass container lled with the electrolyte and taking into account for the AM 1.5G illumination conditions.

Apparatus
Electrodepositions and cyclic voltammetry were carried out with an Autolab PGSTAT 30 controlled by the GPES soware package, equipped with an Aldrich® glass reference electrode (General purpose, reference, Ag/AgCl, 3 M KCl).Electrochemical impedance spectroscopy was done in the same setup with the Autolab PGSTAT 30 controlled by the FRA soware package.
The air plasma treatment was carried out in a Femto PCCE low pressure plasma system (Diener Electronic).A 270 W plasma was applied for 2 minutes.

Fig. 1
Fig. 1 Device layout of triple junction organic solar cell, the organic compounds used in the small band gap (PMDPP3T:[60]PCBM) and wide band gap (PCDTBT:[70]PCBM) photoactive layers, and schematic energy level diagram (close to open circuit) showing the operation of the triple junction cell coupled to the water splitting reactions.

Fig. 2
Fig. 2 Tafel plots of RuO 2 in 1.0 M of KOH and of Co 3 O 4 and NiMoZn in 0.1 M potassium borate (KBi) at pH 9.2.The markers indicate the expected overpotentials during the operation of the small scale RuO 2 / RuO 2 PEC cell (O), the small scale Co 3 O 4 /NiMoZn PEC cell (,), and the large scale RuO 2 /RuO 2 PEC cell (B).

Fig. 3
Fig. 3 (a, c, e) J-V curves of the triple junction solar cells before, during and after water splitting measurement of 20 min.(b, d, f) Simultaneous measurement of operating voltage and current density of the solar cell during photoelectrochemical water splitting, (a, b) 0.0676 cm 2 solar cell connected to RuO 2 /RuO 2 catalysts in 1.0 M KOH.(c, d) 0.0676 cm 2 solar cell connected to Co 3 O 4 /NiMoZn catalysts in 0.1 M KBi.(e, f) 1.7 cm 2 solar cell connected to RuO 2 /RuO 2 catalysts in 1.0 M KOH.The light source is not chopped and the electrolyte is not stirred during measurements.The inset in panel (f) shows the hydrogen and oxygen evolution from the RuO 2 catalysts on the Ti substrates.

Fig. 4
Fig. 4 (a) Schematic device layout of the organic artificial leaf.(b) Photographs of the actual device, from left to right: front side, back side before, and back side after applying the water splitting electrodes and catalysts.

Fig. 5
Fig. 5 Artificial organic leaf under illumination based on a triple junction organic solar, with RuO 2 covered ITO electrodes for hydrogen (right) and oxygen (left) evolution.