A bright outlook on organic photoelectrochemical cells for water splitting Materials

Photoelectrochemcial (PEC) water splitting is increasingly attracting attention as a means to generate clean, renewable hydrogen fuel from solar energy. This Highlight article covers the key advances that have been made over recent years in organic photoelectrochemical (OPEC) cell research, and identi ﬁ es a pathway forwards combining state-of-the-art materials and device engineering from both the organic photovoltaics and inorganic PEC water splitting communities. We discuss the advantages of using buried junction device architectures in OPEC photoelectrodes and identify the need for new materials for OPEC water splitting cells in order to improve their e ﬃ ciency and operating stability. We present an overview on promising new absorbers and device architectures employed in the parallel ﬁ eld of organic photovoltaics with a critical view on requirements for OPEC water splitting. We also elaborate on progress made with organic multijunction cells that we see as promising photocathode designs for voltage-unassisted water splitting. Finally, we see the urgent need for benchmarking rules, in terms of performance and stability parameters, as this emerging ﬁ eld of OPEC water splitting progresses.


Introduction
In order to reach commercial application, photoelectrochemical (PEC) water splitting cells must produce hydrogen with high efficiency and long operating stability at low cost. Achieving these three requirements with one photoabsorbing material remains challenging, and there is intensive research in various aspects of PEC systems using traditional inorganic materials such as compound semiconductors (III-V, Si, etc.), 1 transition metal oxide (BiVO 4 , Fe 2 O 3 , TiO 2 , Cu 2 O) or chalcogenides (CZTS, CIGS). 2 More recently, organic semiconductors have been attracting interest within the PEC eld due to their unique opportunities to carefully tune the visible light absorption and frontier molecular orbital energies, as well as morphological Ludmilla Steier obtained her PhD in Chemistry and Chemical Engineering atÉcole Polytechnique Fédérale de Lausanne (EPFL) in 2016 working in the group of Michael Grätzel on thin lm materials for photovoltaics and solar fuel production. She then joined the group of James Durrant at Imperial College London where she obtained a Marie Skłodowska Curie Fellowship in 2017. With a broad expertise in fabrication and characterisation of materials, Ludmilla is interested in understanding device physics, materials properties and especially charge transfer across interfaces in order to achieve more efficient and stable devices for solar energy conversion and storage. Sarah Holliday studied Chemistry at the University of Edinburgh before joining the group of Iain McCulloch at Imperial College London in 2011. Here she carried out her PhD research in the design and synthesis of non-fullerene acceptors for organic photovoltaics. She then worked with Christine Luscombe at the University of Washington as a postdoctoral researcher studying the photooxidative stability of organic photovoltaics, as well as developing new materials to improve the performance of CZTS based solar cells. She is particularly interested in the scale-up of low-cost, printed solar energy technologies for real-world applications, especially in the context of energy access globally.
properties. Furthermore, the compatibility of organic semiconductors with low-cost, large-area printing routes offers great potential for these materials to meet the requirements for efficient, stable and low-cost PEC devices.
As the eld of organic photoelectrochemical (OPEC) cells is relatively young but growing rapidly, there is an opportunity now to highlight key works in the evolution of OPEC water splitting cells and focus these towards the most promising directions for future research. With a clear vision of the requirements for OPEC cells and knowledge in PEC water splitting with inorganic semiconductors, this Highlight review seeks to advocate for the concept of buried bulk heterojunctions (BHJ) as the most advantageous and promising route towards efficient OPEC water splitting cells. With this conclusion, we elaborate on some of the most recent trends and achievements in research on buried BHJs in organic photovoltaics (OPV) and suggest how these advances can directly contribute to improved materials and electrode design in OPEC photocathodes. Furthermore, we suggest key benchmarking rules in the OPEC water splitting community in order to create criteria of comparison between different materials and cell designs.

Thermodynamic requirements for water splitting
The minimum free energy needed for splitting water into hydrogen and oxygen is DG ¼ +237.18 kJ mol À1 at 25 C, which is equal to a required voltage of 1.23 V (¼DG/nF for n ¼ 2 electrons with F being the Faraday constant).
DG ¼ DH f À TDS ¼ 286.03 kJ mol À1 À 0.163 kJ mol À1 K À1 T [K] However, in practice, this voltage is not sufficient to split water. When overpotential losses at the anode (h a ), cathode (h c ) and in solution due to low ionic conductivity (h sol ) are taken into account, it becomes clear that the nal applied voltage must be substantially higher than 1.23 V. The overpotentials at the anode and cathode are highly dependent on the catalyst used and can easily increase the voltage required to drive the water splitting reaction to 1.5-1.8 V, even with the best catalysts used. Hence, it is important to focus on PEC cell designs that can provide photovoltages to sustain standalone (unbiased) water splitting.
What can be concluded from PEC water splitting with traditional inorganic semiconductors?
Since the research eld of OPEC water splitting is relatively young and unexplored, knowledge transfer from research in PEC water splitting with traditional inorganic photoelectrode materials can prove useful. In this paragraph we identify conceptual challenges in inorganic PEC systems and see great opportunities for organic semiconductors to succeed in overcoming those hurdles in OPEC water splitting cells.
The research directions of the inorganic PEC water splitting community can conceptually be distinguished by the two types of junctions used: (i) semiconductor/electrolyte solution junctions and (ii) buried junctions wherein the absorber is sandwiched between charge selective contacts and coated with a suitable catalyst. Photoelectrodes based on earth-abundant materials such as BiVO 4 , hematite or TiO 2 make use of the semiconductor/electrolyte junction to drive charge separation and simultaneously use the photoabsorber as the water oxidation catalyst. With the voltage requirement in mind, a big part of the inorganic PEC community is working towards a tandem photocathode-photoanode design shown in Fig. 1a. 3 At the photoanode, the n-type semiconductor has to provide enough driving force to spatially separate the charges created upon illumination and efficiently drive the catalysis so that holes are extracted to solution and electrons to the contact before they have time to recombine. However, water oxidation, which requires the accumulation of 3 holes, 4 is a very slow reaction (ms to s), while recombination timescales are on the order of ps to ms. 5 When the holes at the surface are consumed before they can be extracted, the semiconductor cannot build up the necessary photovoltage, V Ph1 to drive the reaction and higher internal electric elds will be required to minimise recombination. A similar challenge is faced by the photocathode though the rate of reaction is faster for hydrogen evolution than for water oxidation. Indeed, photovoltages in these photoelectrodes are signicantly lower than the Shockley-Queisser limit (Fig. 1b). 6 Commonly used strategies to increase the build-up of photovoltage in the photoabsorber include the use of (i) cocatalysts to improve or enable photocatalysis to occur, 7,8 (ii) surface state passivation overlayers 9-12 or (iii) capacitive charge storage overlayers. 13,14 Although these strategies improve photovoltage generation in these photoelectrodes, their full potential has still not been exploited due to the inherent photovoltage limitation of the semiconductor/electrolyte solution junction. In order to achieve photovoltages closer to the Shockley-Queisser limit, the redox couple has to be carefully chosen with respect to the Fermi level of the semiconductor to maximise the built-in voltage, and the reaction has to be on timescales that are competitive with charge recombination in the material. On the other hand, the implementation of well-designed buried pn junctions into photoelectrode designs signicantly improved photovoltage generation and utilisation (Fig. 1b). 6 This approach is adapted from photovoltaics where pn junctions are carefully engineered and combined with suitable charge selective contacts to minimise photovoltage losses through recombination. Coupling such a buried pn junction to suitable electrocatalysts lis the photovoltage limitation in a semiconductor/liquid junction type device and allows for the utilisation of photovoltages close to the Shockley-Queisser limit. Furthermore, through the spatial separation of electrons and holes in such a buried pn junction type photoelectrode charge carrier lifetimes are signicantly prolonged to reach timescales required for catalysis. However, using traditional inorganic materials from the photovoltaic eld, the design of suitable pn junctions is oen assimilated with careful engineering of a defect-free junction using high quality materials (e.g. III-V compound semiconductors) that provide sufficiently long carrier diffusion lengths to reconcile maximum light absorption with efficient charge extraction. 1 These materials are signicantly limited by cost, however, and there are currently only a few examples of buried junctions using low-cost inorganic semiconductors such as Cu 2 O as photoabsorbers. 15,16 On the other hand, engineering a donor-acceptor junction with organic semiconductors can be achieved at low-cost via established large-area printing techniques for the fabrication of organic photovoltaics (OPVs). In the following we will introduce the concept of a bulk heterojunction (BHJ), before we outline how it entered the eld of OPEC water splitting cells.

Bulk heterojunction organic photovoltaics
Organic semiconducting materials are based on molecular frameworks of sp 2 hybridised carbon atoms, whereby the extended delocalisation of pi-electron orbitals between neighbouring atoms results in the formation of an effective bandgap E g between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO). In highly conjugated molecular or polymeric systems, this bandgap is small enough such that an electron can be photoexcited from the HOMO to the LUMO upon absorption of incident solar photons, leaving behind a hole to form an electron-hole pair (exciton). Organic semiconductors benet from excellent light harvesting properties due to their large extinction coefficients that arise from the signicant overlap between the ground and excited states of the molecule, as well as their broad and tuneable absorption bands, which can be tailored through chemical design to match the incident solar spectrum. These properties, combined with the low-cost, relative earth abundance and solution processability of organic semiconductors, makes these materials excellent candidates for light harvesting in PEC cells as well as organic photovoltaic applications.
Photoelectrodes based on single organic semiconducting materials such as polyaniline or polythiophenes were investigated in the past, but these gave very low photocurrent densities on the order of a few microamperes per square centimetres, which can be partly explained by the large exciton binding energies in organic semiconductors. 17,18 Due to the relatively low dielectric constant of organic semiconductors (3 r $ 2-4), the large Coulomb attraction between charges is poorly screened. This results in high rates of charge recombination following photoexcitation, which reduces the available photocurrent. In order to overcome this problem of charge recombination, the OPV community in the 1980s developed an approach of combining two different organic semiconductors with offset HOMO and LUMO energy levels, such that an electron which is photoexcited in the material with smaller electron affinity (the 'donor') is transferred to the second material (the 'acceptor'), which facilitates charge separation at this interface as demonstrated in Fig. 2a. 19 Since the exciton diffusion length in organic semiconductors is only around 10 nm, 20 these two materials are typically blended on the nanoscale in order to increase the donor-acceptor interfacial area, creating what is known as a bulk heterojunction (BHJ) OPV device. 21,22 The performance of organic semiconductors is determined by many different factors that can be controlled during the design and synthesis of the polymer or small molecule. For example, the planarity of the semiconducting backbone, which controls both the degree of conjugation and crystal packing of the material, can be modied by increasing the number of fused rings or bridging atoms between rings. In addition, the judicious choice of electron-withdrawing (uorine, carboxylate, imide etc.) or electron-donating (thiophene, alkyloxy etc.) functional groups can tailor the frontier energy levels of the molecule and therefore the optical bandgap, as well as the available voltage of the device. In addition to the optoelectronic properties, device performance is equally dependent on the morphology of the interpenetrating network of donor and acceptor materials within the blend. As such, the density and length of solubilising side-chains on the organic semiconductor can be tuned in order to optimise the phase behaviour and crystal packing of each material, which in turn can affect the charge generation and transport properties of the bulk heterojunction. Finally, the molecular weight properties and purity of the material can be strongly affected by the polymerisation conditions, which in turn can inuence the device performance.
One of the earliest and most widely studied donor polymers in the eld of OPV is the simple homopolymer poly(3hexylthiophene) (P3HT) (Fig. 6). Typically it is blended with the fullerene derivative phenyl-C 61 -butyric acid methyl ester (PCBM) as the acceptor (see Fig. 7), giving a power conversion efficiency of 3-5% in OPV devices. 24 The well studied nature of this blend within the OPV community, combined with its reasonably high hole mobility 25 and demonstrated aqueous stability, 26 made this an obvious choice for early organic photocathode studies. Indeed, the vast majority of organic photoelectrode studies to date have focused on these P3HT:PC 61 BM blends, as discussed in more detail below.
The rise of organic photoelectrochemical cells

Organic semiconductor/electrolyte junction versus buried junction architectures
In the early stages of research, OPEC water splitting in aqueous solution was demonstrated with polymer photocatalyst/ electrolyte solution junctions, yielding photocurrent densities on the order of only a few microamps or less. 17,27,28 Striving for higher current output, polymer absorbers were combined with known hydrogen evolution reaction (HER) catalystsa known strategy from PEC water splitting with inorganic materials. While p/n bilayer junction designs still yielded only a few tens of microamps of photocurrent, even with Pt surface modication, 29 the breakthrough leap in photocurrent densities from the microamp to milliamp scale was achieved introducing BHJs into the photoelectrode design and along with HER catalysts and charge selective layers. Bourgeteau et al. rst introduced the PEDOT:PSS/P3HT:PCBM/TiO 2 :MoS x photocathode design ( Fig. 3a), inspired by the idea of TiO 2 electron extraction layers in the OPV eld. 30 With the aim to protect the P3HT:PCBM blend from oxygen and water, they developed a thick catalyst layer of MoS x intermixed with TiO 2 nanoparticles that could achieve photocurrent densities of 180 mA cm À2 at 0 V vs. RHE and a photocurrent onset potential for hydrogen evolution of $0.45 V (Fig. 3b). The photovoltage achieved in the OPEC device was increased by addition of TiO x , approaching the open circuit potential (V OC ) of the equivalent OPV device. However, the observed photocurrent densities from hydrogen evolution were more than 50Â smaller than the short circuit current obtained in the OPV device.
In a subsequent study, Bourgeteau et al. developed an OPEC photocathode that successfully extracted the full photovoltage and photocurrent of an equivalent OPV device by placing a highly conducting LiF (1.2 nm)/Al (100 nm)/Ti (30 nm) interfacial layer between the BHJ and the HER catalyst (Fig. 4a). 31 Using this strategy, they could achieve photocurrent densities of À8 mA cm À2 at 0 V vs. RHE with a noble metal-free OPEC photocathode. It is worth noting that these relatively thick metallic interfacial layers not only provide the necessary conductivity and band alignment to efficiently extract photogenerated electrons from the organic blend to the catalyst, but act also as a mirror to reect otherwise transmitted photons back into the device. Reective contacts of 100 nm Ag have been shown to enhance the short circuit current by 1/3 in OPV devices. 32 Transmission losses should therefore be considered when optimising OPEC photoelectrodes, as they can account for large losses in photocurrent. 33 This work also highlights that light harvesting and catalytic components can be optimised separately because the photocurrent onset potential for hydrogen evolution is determined by the sum of the onset potential for dark electrolysis (carried out by the electrocatalyst) and the photovoltage of the junction (Fig. 4b). Using a Pt/C electrocatalyst instead of MoS x , the photocurrent onset could be shied to $200 mV more positive potentials due to a smaller overpotential for hydrogen evolution of the Pt/C electrocatalyst (compare the dark current onset potentials during dark electrolysis for Pt/C and MoS x in Fig. 4b). This demonstrates that such a buried BHJ photoelectrode device can be truly competitive against a PV + electrolyser wired system.
Meanwhile, Guerrero et al. studied charge extraction from P3HT:PCBM to a redox couple in acetonitrile solution. Using a ferrocene/ferrocenium redox couple, they showed that all charges generated by the P3HT:PCBM blend in an equivalent OPV device could successfully be extracted to the hole scavenger in solution for this particular OPEC device. 32 This work might suggest that a buried junction approach using charge selective layers and HER or OER catalysts may not be required for efficient charge extraction in OPEC water splitting cells. However, it is important to note that charge extraction from such an OPEC cell using a direct BHJ-solution interface can only be efficient as long as the electrolyte solution is able to regenerate the oxidised photoabsorber (in this case P3HT) before charge recombination within the BHJ annihilates the oxidised excited state. The  language of "regeneration" in this case comes from the picture of a photoelectrochemical dye-sensitised solar cell (DSSC), wherein an electron is injected from the LUMO of the excited dye molecule to the conduction band of TiO 2 . 34 The redox couple in the electrolyte then regenerates the oxidised dye. The open circuit voltage, V OC , will depend solely on the redox potential of the electrolyte and the at band potential of the TiO 2 but such a PEC cell can only function when this redox potential is more negative (higher in energy) than the HOMO of the dye. This picture can be adapted to the PEC cell Guerrero et al. used where P3HT is the analogue to the dye molecule and PCBM/ZnO electron acceptors to the TiO 2 (Fig. 5). While such a cell can reach relatively high efficiencies, as has been demonstrated for DSSCs, it should be highlighted that these systems may suffer from known charge recombination losses especially if "regeneration" is not fast enough. The Fc/Fc + redox reaction is a 1-electron transfer and not proton coupled; hence it is suitable for such a device. However, water splitting is a much more complex multi-electron reaction and is therefore much slower. This challenge of charge extraction via a slow reaction is even more severe in single organic semiconductor photoelectrodes. An example gives the work of Bornoz et al. on a poly[benzimidazobenzophenanthroline] (BBL) photoanode. 35 Using sulte as a sacricial hole acceptor they could obtain photocurrent densities on the order of 230 mA cm À2 while ten times lower photocurrent densities were extracted from direct water oxidation. Even with a sacricial agent, the obtained photocurrent densities were rather low which is likely due to limitations in charge separation in the BBL. Though the exciton binding energy in BBL is relatively low (E ¼ 0.29 eV, 3 r ¼ 8.3), 36,37 the generation of free charges is still not barrier-free.
Another limitation of a direct semiconductor/liquid interface is that the photovoltage is governed by the redox potential of the electrolyte as shown in Fig. 5. Hence in case of photoanodes for water oxidation, the HOMO level of the donor has to be tailored to the redox potential of the electrolyte, which still remains challenging for materials chemists as we will discuss in the section of low band gap donor polymers.
Additional requirements such as the chemical stability of the organic semiconductor in the electrolyte medium, resistance against the formation of unwanted surface states (e.g. dangling bonds) and stabilised adsorption of both reactants and reaction intermediates at the semiconductor surface must also be considered in this case. In particular, surface adsorption will certainly be an issue that is encountered when using hydrophobic organic semiconductors in aqueous electrolytes. With these limitations in mind, the outlook proposed herein is for the OPEC eld to focus on buried BHJ photoelectrode designs that make use of a charge selective layer in-between BHJ and catalyst in order to eliminate the recombination pathway from the blend to solution.

Stability
Since these rst demonstrations of efficient proton reduction with organic photocathodes, the OPEC community has focused on the buried junction approach. Interestingly, the reasoning behind the buried BHJ approach was based on introducing an anti-corrosion layer (to protect the blend and PEDOT:PSS) rather than to provide a recombination-blocking layer as we propose is important to minimise charge recombination between the blend and the electrolyte solution. 30 Indeed, many of the hole selective layers (HSLs) explored in OPEC cells are corrosion-sensitive (e.g. PEDOT:PSS, polyaniline, MoO 3 , CuI). 26,[38][39][40] For example, MoO 3 is commonly used as HSL in OPV devices but encounters problems with intercalation and irreversible reduction into sub-stoichiometric phases of MoO 3Àx -OH x and MoO 2 , which have unfavourable energy alignment. Hence the strategy of using oxide protective layers was a logical adoption from the inorganic PEC water splitting eld, which is essential when using corrosion-sensitive materials such as Cu 2 O. 15 Promising recent examples of oxide protective layers for OPEC devices include the atomic layer deposition (ALD) of amorphous TiO 2 onto P3HT:PCBM photocathodes with a CuI HSL, which showed constant hydrogen evolution for more than 3.5 h. 41 This was a signicant improvement on previous results using amorphous TiO 2 deposited by pulsed layer deposition (PLD). 40 Thus far, there are only few examples of photoelectrodes showing operating stability on the order of several hours (Table 1). Operational stabilities of up to 3 h for P3HT:PCBM based OPEC devices have been achieved using cross-linked PEDOT:PSS as the hole selective layer and solutionprocessed TiO x , however the photocurrents measured in this case were extremely low. 42 So far, the most promising HSL materials in terms of stability are NiO and WO 3 , which are able to sustain milliampere range photocurrents over the course of up to 10 h. 43,44 While the use of protective coatings has proven effective to some extent, it is important to consider that relying on inherently corrosion-sensitive materials in PEC water splitting cells is not favourable for large-scale production, since any imperfections in the protective layer (pinholes, cracks, etc.) would lead to increased production costs due to high failure rates, reducing the economic viability of the entire production process.
In order to achieve standalone efficient water splitting with durable OPEC devices, it is therefore important to explore new materials and architectures that can provide improved performance and stability. Fortunately, there is a vast source of literature on high performance organic semiconductors and devices from the organic photovoltaics (OPV) community that can be adapted to OPEC applications. In the following, we will discuss recent progress in OPV materials, including the emergence of low-bandgap polymers and non-fullerene acceptors for improved light absorption and photovoltage, as well as new tandem device architectures that could be applied to organic PEC devices.
Thus far, one promising recent study by Francàs et al. explored the low bandgap polymer PCDTBT in an organic photocathode blended with PCBM and achieved impressive operational photocathode stabilities of >5 h, as well as exhibiting signicantly higher photovoltages compared to P3HT:PCBM blends. This demonstrates the huge potential to combine the progress in materials design and device engineering between the elds of OPV and PEC research, in order to develop high efficiency, stable and low-cost OPEC devices.

Low-bandgap donor polymers
Because of the limited visible wavelength absorption of fullerene acceptors, much of the focus of OPV materials development has been to design lower bandgap donor polymers in order to harvest more of the solar spectrum. Initial efforts in this area focused on strategies to planarise the molecular backbone and thereby improve the p-orbital overlap. This included the introduction of bridging atoms such as uorene and indacenodithiophene, fused ring systems such as benzodithiophene and thienothiophene, and increasing the degree of quinoidal character through replacing phenyl ring systems with thiophene-based systems. In more recent years, the strategy of push-pull hybridisation has been widely adopted, whereby the backbone is composed of alternating electron-rich and electronpoor moieties which results in molecular orbital mixing of these components and a reduced effective bandgap, as demonstrated in Fig. 1b. 23,[45][46][47] Using the diverse toolkit of synthetic organic chemistry to tune the energy levels of the molecule, this strategy has proven extremely successful in extending the absorption of donor polymers into the near-IR region of the spectrum, with bandgaps below 1.3 eV having been reported. 48 Meanwhile, the maximum available open circuit from the OPV cell is determined by the energetic offset between the HOMO of the donor and the LUMO of the acceptor, and can therefore be increased by designing donor polymers with lowlying HOMO energies, some examples of which are shown in Fig. 6. The paradox of trying to maximise light absorption by reducing the bandgap, whilst maximising the voltage by trying to lower the HOMO, has been one of the main dilemmas in materials design for OPV electron donor polymers. In the context of OPEC cells, the large voltage of 1.5 V that is needed to drive unassisted water splitting means that it may be necessary in this case to compromise on the photocurrent in order to maximise photovoltage, meaning that a deep HOMO is more of a priority in these applications. Stabilisation of the HOMO in push-pull polymers can be achieved by various strategies such as incorporating electron-withdrawing substituents along the backbone, as exemplied by the development of the uorinated PTB7 polymer by Liang et al., which has an ionisation potential of 5.15 eV compared to 5.04 eV for the unuorinated counterpart. 49,50 It should be noted that uorination is a particularly useful strategy in this regard since it typically results in stabilisation of both the HOMO and LUMO, thereby allowing for a narrow bandgap to be maintained whilst increasing the available voltage. [51][52][53] However, it should also be noted that a sufficiently high-lying LUMO on the donor is required to provide an energetic driving force for charge transfer to the acceptor LUMO. Further stabilisation of the donor HOMO was achieved by replacing the electron-donating alkoxy side chains of PTB7 with alkylthienyl side chains in PTB7-Th, also known as PCE-10, resulting in an increased V OC from 0.76 V to 0.81 V. 54 A larger ionisation potential of 5.44 eV was then obtained by incorporating thioalkyl chains on the alkylthienyl groups, resulting in 0.89 V with fullerene-based acceptors. 55 Similarly, Li et al. recently published the benzodithiophene based polymer PDTB-EF-T, wherein a deep HOMO of À5.5 eV achieved through the synergistic electron-withdrawing effect of the uorine groups on the alkylthiophene side-chains and ester groups on the thiophene co-monomers. 56 In this example, side-chain engineering was used to optimise the interchain interactions and achieve a favourable morphology, resulting in an outstanding ll factor of 0.76 and power conversion efficiency of 14.2%. The medium bandgap benzodithiophene-co-benzodithiophenedione based push-pull polymer PBDBT developed by the Hou group, (also known as PCE-12) has also demonstrated excellent performance when spectrally matched with low bandgap non-fullerene acceptors (as discussed below). The large ionisation potential (5.33 eV) of this polymer in particular may offer promise for high photovoltage OPEC applications. 57 Thirdly, a number of oligothiophene-based push-pull polymers have been developed by the Yan group and others that have strong, temperature-dependent aggregation properties, resulting in the formation of robust, near-ideal morphologies with small but phase-pure polymer domains. [58][59][60][61] Although this class of polymers present challenges in reproducibility and scale up due to the high-temperature solution processing required, many of these polymers such as PffBT4T-2OD and its longer chain analogue PffBT4T-C9C13 deliver excellent photovoltaic performance of up to 13% PCE, with high open circuit voltages of up to 1.1 V that could prove promising for OPEC applications. 62 As well as selecting materials to maximise photocurrent and photovoltage in photoelectrode applications, it is equally important to consider the stability of these new push-pull polymers in transferring this research to PEC applications. Unfortunately, it has been shown that many of the highperformance donor polymers suffer from signicant degradation under illumination in ambient conditions. [63][64][65] Photooxidation and degradation of the polymer backbone is one factor that contributes to this instability, most notably in the case of BDT-TT polymers, [66][67][68] and it has been shown to be accelerated by the use of high-boiling point solvent additives such as 1,8diiodooctane in deposition of the active layers. [69][70][71] Another source of instability for OPV devices is related to the electrode or interfacial materials, for example in the use of easily oxidised, low-workfunction metal electrodes 72 or the acidic, hygroscopic poly(3,4-ethylenedioxythiophene)-polystyrene sulfonate (PEDOT:PSS) hole transport layer, as discussed above. 73 These factors will certainly need to be taken into consideration when transferring this research to OPEC applications, as indeed will the stability of these materials under operation in aqueous conditions, which has not yet been studied.

Non-fullerene acceptors
Until very recently, the vast majority of organic solar cell research utilised a fullerene derivative, such as the archetypal phenyl-C 61 -butyric acid methyl ester (PC 61 BM) or its C 70 analogue PC 71 BM, as the electron accepting material. Whilst fullerenes have excellent electron transport properties and large electron affinities, making them compatible with many donor polymers, they also have several limitations. Most signicantly, C 60 fullerenes in particular have very limited absorption in the visible and near-IR region of the solar spectrum due to their high molecular symmetry, which limits the contribution of the acceptor towards light absorption within the active layer. In addition, the very deep LUMO energy of PC 61 BM and PC 71 BM signicantly limits the available open circuit voltage. Since the optoelectronic properties of fullerenes cannot easily be modi-ed through changes to the chemical structure, it has therefore only been possible until recently to tune the V OC through designing polymers with deeper HOMO energies, which oen contradicts the need for a low optical bandgap as mentioned above.
Over the past few years, chemists have made considerable breakthroughs in developing alternative, non-fullerene based acceptors (NFAs) with improved visible wavelength absorption and tuneable frontier energy levels, which have resulted in rapid advances in OPV device performance. [74][75][76][77] The most successful materials in this regard (some example of which are shown in Fig. 7) are based on the same push-pull hybridisation strategy employed in the design of semiconducting polymers, which allows for precise control over the optoelectronic properties, with the phase separation properties optimised through choice of solubilising side-chains and tuning of the backbone planarity. For example, by replacing the uorene core moiety of FBR for a more extended and electron-rich indacenodithiophene unit in O-IDTBR, the bandgap can be decreased from 2.14 to 1.68 eV. 64,78 Further extension of the core in the form of dithieno-indacenodithiophene, as well as modication of the electron withdrawing end groups, gave an even smaller bandgap of 1.59 eV for the small molecule ITIC. 79,80 Recent interest in semitransparent OPV devices has inspired chemists to develop increasingly large, fused ring systems with ultra low bandgaps as demonstrated by the 1.25 eV bandgap of 3TT-FIC 81 or IEICO-4F (1.24 eV), 82 although it should be noted that the added synthetic complexity of such molecules may be less advantageous to commercial scale-up.
As discussed above, the greater priority in PEC cell applications is to maximise photovoltage in order to have sufficient overpotential to drive unassisted water splitting, in this context, some of the wider bandgap, uorene-based NFAs may be more promising for OPEC applications, especially when used in conjunction with polymers that have low-lying HOMO energies. For example, the relatively small electron affinity LUMO of FBR (3.57 eV) results in open circuit voltages of up to 1.13 V when combined with the oligothiophene-based polymer PffBT4T-2DT by Baran et al. 83 Elsewhere, small molecule NFAs have already delivered many instances of open circuit voltages > 1 V and up to 1.2 V. 84 Clearly, the synthetic versatility and degree of optoelectronic control of this new generation of acceptors offers a distinct advantage over their fullerene predecessors. Indeed, the ability to tune both the donor and acceptor simultaneously has successfully liberated the eld of OPV from its previous paradigm, whereby the donor polymer was historically responsible for the majority of light absorption as well as maximising the V OC through maintaining a deep HOMO. The opportunity to optimise both materials in the blend has recently led to a greater overall performance in OPV devices, and certainly presents interesting opportunities to explore NFA blends in the context of OPEC devices.
As well as contributing to enhanced photovoltaic performance, non-fullerene acceptors have been shown to offer improved stability in OPV applications. 63,64,78,85 This is largely due to the formation of a more stable morphology compared to fullerene-based acceptors, which have a tendency to diffuse through the BHJ blend and aggregate over time. [86][87][88] While many NFA based blends have demonstrated excellent stability in ambient conditions and under illumination, the stability of these blends in aqueous environments has not yet been explored. An important part of the work in transferring this research to the eld of PEC cells will therefore require the aqueous stability of these materials to be studied as a rst instance, as well as modifying their cohesive and adhesive properties under operation in PEC architectures.

Device architectures and interfacial engineering
Inverted versus conventional architectures In addition to active layer materials, the performance of an organic photovoltaic device is strongly affected by the choice of Fig. 7 Examples of some of the high performance fullerene and non-fullerene acceptors employed in organic photovoltaics research. device architecture and interfacial layers. In the conventional OPV architecture depicted in Fig. 8, a transparent conducting oxide (typically indium tin oxide) is used as the high work function, transparent anode, with a selective hole transport layer such as the solution processed PEDOT:PSS deposited between the anode and the active layer. A low workfunction metal such as Al is typically used as the cathode in this case, with an additional interlayer of Ca or LiF to facilitate electron injection. This conventional architecture has some disadvantages, in particular with regards to stability due to the reliance on easily oxidised, low workfunction metal electrodes 89 and acidic, hygroscopic PEDOT:PSS. 90 An alternative, inverted architecture (Fig. 4) was therefore adopted wherein ITO is instead used as the cathode, with an electron transport layer such as ZnO or TiO x solution deposited on top. 91,92 MoO 3 is oen used as the hole selective interlayer in this case and stable metal such as Ag can be used as the top contact, allowing for considerably reduced degradation rates. As well as offering improved stability, the inverted architecture is increasingly used due to the enhanced photovoltaic performance that it offers for many BHJ blends. This may be partly due to the increased light absorption achieved through the use of high reectance Ag electrodes, but also because of differences in vertical phase separation within the blend. In each conguration, differences in surface energy at the electrode can favour the build-up of a polymer-rich or acceptor-rich layer at the interface. This can signicantly affect the charge carrier mobilities and the extent of charge recombination at this interface, which therefore impacts short circuit currents and ll factor, highlighting the importance of tailoring the interfacial properties for enhanced performance. 93,94 Since the majority of high performance OPV materials are now optimised for the inverted architecture, one important aspect of developing OPEC devices will be to investigate how the vertical phase separation behaviour and device performance change when using a conventional architecture for photocathodes. At the same time, this will involve investigating alternative charge selective layers that can allow for preferential vertical phase separation within the active layer, whilst improving the interfacial cohesion and stability of the device.

Multijunction device designs
As well as advances in interfacial engineering, the development of high performance OPEC devices can certainly benet from recent progress in tandem and multijunction architectures as a route to achieving higher photovoltages whilst harvesting more of the solar spectrum. [95][96][97] In single junction OPV devices, the V OC is limited by the bandgap of the donor and acceptor materials, which leads to inherent energy losses due to the thermalisation of high-energy photons (E > E g ). By combining 2 or more sub-cells with complementary bandgaps and connecting them in series, as demonstrated in Fig. 9, the total theoretical V OC becomes the sum of the individual V OC s of the subcells and can oen be sufficient to provide the necessary voltage of >1.8 V to drive water splitting. This architecture allows the use of strongly absorbing catalysts and larger electrolyte solution volumes in OPEC devices, since light absorption has only to be optimised within the photocathode. In addition, catalysts with reective properties could be used to enhance efficiency. It is worth mentioning that although there have been some examples of organic photoanodes, 35,98 we think that OPEC cells based on tandem-photocathodes are more promising than a photocathode-photoanode tandem cell. As pointed out earlier, water oxidation is a complex and slow reaction, making efficient hole extraction difficult. In addition, many organic materials are sensitive towards oxygen and suffer from degradation upon exposure. Furthermore, the tandem architectures shown below are good starting points to explore these systems for OPEC water splitting.
In addition, by selecting materials with complementary absorption across the spectrum, larger photocurrents can be achieved from multijunction devices due to enhanced spectral coverage (Fig. 10).
Tandem and triple junction OPV devices have gone through remarkable advances in performance in recent years, accelerated by the emergence of highly tuneable non-fullerene acceptors as well as new medium and wide-bandgap donor polymers, which present new opportunities to tailor the spectral coverage and voltage of the complementary subcells. With regards to OPEC devices, the opportunity to produce much higher photovoltages >1.8 V using multijunction devices is particularly important in terms of meeting the high voltage requirements for water splitting. Since the number of available donor and acceptor materials combinations for organic multijunction devices has multiplied in recent years, the performance of organic multijunction devices is limited more by the availability of charge recombination layers which can deliver sufficiently high photon transmittance and efficient charge recombination, whilst avoiding any damage or degradation of the active layer beneath. The introduction of semitransparent, ultrathin metal layers provided signicant advances in this area due to the ability to manipulate the optical eld distribution as well as providing balanced current densities between subcells, as demonstrated by the MoO 3 /Ag/ polyelectrolyte layer developed by Zou et al. 99 Recently, Che et al. published a tandem OPV device based on a vacuumdeposited small molecule:C 70 bottom cell and a solutionprocessed PTB7-Th:non-fullerene acceptor top cell (Fig. 11) which gave impressive power conversion efficiencies of 15% for relatively large device areas (2 mm 2 ) and open circuit voltages of up to 1.6 V. 100 A nearly optically and electrically lossless charge recombination layer was employed, consisting of a layer of Ag nanoparticles adjacent to a vacuum thermal evaporation (VTE)-deposited BPhen:C 60 electron-ltering buffer layer, with a hydrophilic capping PEDOT:PSS interlayer deposited on top to protect the BPhen:C 60 from damage during solution deposition of the top cell. As well as the exceptionally high PCE, this example offers particular promise for photocathode applications due to its optimal performance in the conventional device architecture, unlike many of the high performance OPV devices in the literature.
Homo-tandem devices, wherein 2 or more subcells of the same active layer materials are stacked in series or parallel, have also attracted recent interest as a means to enhance the light trapping and photovoltage, as well as offering simplied processing due to the use of only one active layer blend. 101,102 Esiner et al. fabricated triple junction OPV devices with 2 identical photoactive layers based on the wide bandgap polymer PTPTIBDT-OD that could produce open circuit voltages up to 1.74 V, which they demonstrated to be large enough to split water in an electrochemical device. [103][104][105] In another example by Chen et al., a double junction homo-tandem device was fabricated with the oligothiophene polymer P3TEA and a perylene diimide based non-fullerene acceptor SF-PDI 2 , giving an exceptionally high V OC of 2.1 V and 10.8% PCE. 106 The high photovoltage of this OPV device was also used to split water in an electrolytic cell, demonstrating the potential to implement homo-tandem devices in photocathodes if the materials are Fig. 9 Schematic of a simplified energy diagram of a tandem photocathode comprised of two sub-cells with complementary absorption spectra that are connected in series via a recombination layer and coated with a HER catalyst. A dark electrode is used for water oxidation. This architecture allows the use of strongly absorbing catalysts and larger electrolyte solution volumes between photocathode and anode, since light absorption only needs to be optimised within the photocathode. In addition, catalysts with reflective properties could enhance efficiency. Using a dark anode for oxygen evolution can help to overcome many hurdles in PEC water splitting, such as the susceptibility of many organic semiconductors towards degradation under oxygen exposure. shown to be stable and the design can be adapted to a conventional architecture, buried junction OPEC conguration.

Benchmarking OPEC performance
Finally, with all these new research opportunities it is important to emphasise that progress in the eld of OPEC water splitting should be documented in a consistent way. Below we will highlight a few important points that should be considered to make published results more comparable with each other.
(1) In order to assess operational stability and faradaic efficiency, OPEC cells should be operated under stated illumination (typically AM 1.5G simulated 1 sun intensity) for several hours whilst simultaneously recording gas production by gas chromatography or Clark electrodes. It is also advisable to carry out these measurements under a nitrogen atmosphere whilst stirring the electrolyte to avoid the build-up of pH gradients and bubble adhesion to the surface of the photoelectrode. The potential applied during the chronoamperometric measurement should be chosen with respect to the reaction measured. For proton reduction, values of 0 V vs. RHE are commonly used, however more positive values can be selected if the photovoltage of the photocathode is sufficiently high. Ideally, these measurements should be performed at the operating potential of an ideal 2-electrode tandem device. For a photoanodephotocathode tandem this potential would be equivalent to the intersection point, E op , of their J photo -V curves as shown in Fig. 12a. In a single-electrode tandem device this would be the intersection point between the J photo -V curve of the photoelectrode and the electrolysis curve of the dark electrode.
(2) When stating photocurrent densities, a comparison at 0 V vs. RHE for photocathodes and 1.23 V for photoanodes has been helpful to compare literature results thus far. It is noted that all J-V curves should be referenced to the RHE scale, taking into account the varying electrolyte solutions and their pH used. Apart from comparing photocurrent densities at these potentials, it is useful to state the photocurrent at the operating potential of an ideal 2-electrode tandem cell as mentioned above. This can be achieved by determining the potential V x at the intersection of the tangents (a) and (b) as shown in Fig. 12b.
(3) Similarly, the onset potential, V O , for HER or OER can be determined from the intersection of tangent (b) with the potential axis (at zero current). This is a more consistent way of referencing the photocurrent onset potential than determining it by an arbitrary value of 0.1 mA cm À2 as has been proposed earlier. 31 Indeed, depending on the catalyst used, some catalysts transform into pre-catalytic species by changing oxidation states. This is usually seen as a redox peak in the linear sweep and might misguide the interpretation of the onset potential. If J-V curves show 2 slopes and it is not clear how to determine the reaction onset potential, it is advisable to measure gas evolution at potentials between these two slopes to conrm faradaic efficiency.
(4) Although the use of noble-metal catalysts is incompatible with a low-cost OPEC cell design, research on novel OPEC BHJs should not be hampered by this criterion. As discussed above, a buried BHJ design offers the advantage to disconnect light harvesting and catalysis components, which means that each component can be optimised separately. However, it is also important to consider the need for mild deposition procedures for earth-abundant catalysts with low overpotentials for HER and OER.
Following these benchmarking guidelines will benet the OPEC community and help the eld to advance more rapidly by unifying the structure and research language within the literature. Table 2   protected CuI/P3HT:PCBM/TiO 2 /RuO 2 photocathode shows the best stabilised OPEC performance. Some of the recent examples show rather resistive J-V curves and hence, the cross-over potential, V x , and photocurrent, J x , cannot be determined.

Conclusions
In summary, although a direct semiconductor/electrolyte solution junction would be highly desirable for a simple and lowcost PEC design, decades of research on PEC water splitting with inorganic materials has revealed some serious shortcomings of such systems. In particular, these devices suffer from signicant photovoltage losses due to water splitting being too slow to compete with charge recombination in the absence of a strong electric eld. Buried junction PEC devices have been proven to achieve generally higher photovoltages but are oen prohibitively expensive when fabricated with traditional inorganic materials. OPEC water splitting devices, meanwhile, have great potential to realise a cost-effective buried junction type photoelectrode, standing on the shoulders of decade-long research in both the OPV and PEC elds.
The rapid advances in materials design within the eld of OPV has provided a vast number of high performance donor polymers and non-fullerene acceptors with carefully tailored chemical structures that can now deliver both high photocurrents and open circuit voltages > 1 V. Opportunities to adopt these new materials to OPEC photoelectrodes means that signicantly more efficient OPEC devices are now in reach. In addition, the use of improved double or triple BHJ tandem photocathode architectures and interlayer materials could be especially promising for high-efficiency OPEC devices providing photovoltages of >1.8 V necessary for the water splitting reaction.
Finally, with all these new research opportunities it will be important to establish a systematic structure to compare OPEC water splitting cells amongst each other and with their inorganic counterparts, and to easily identify benchmark systems. Towards this goal, we suggest here some benchmarking guidelines and emphasise the need to implement operational stability measurements over several hours into the protocol.
With more research on buried junction OPEC devices implementing some of the most promising blend materials used so far in the OPV eld we expect to see rapid advances in Fig. 12 (a) Schematic relating band gaps and photovoltages for an idealised two-absorber tandem photoelectrolysis device in which the anode is the top absorber. Free energy losses in each absorber (V loss ) and catalytic overpotentials for each half-reaction (h) are fundamental loss pathways. Current density-potential curves for electrocatalysts (grey) and an idealised PEC photoanode (blue) and photocathode (orange) are shown with their relative power contributions (shaded areas) in an ideal tandem operating at E op being the intersection of the two J-V curves. Reprinted from ref. 6 with permission from Elsevier. (b) Useful benchmark parameters in a current density-potential (J-V) curve of typical OPEC cells showing the current density at 0 V vs. RHE, the photocurrent J x at the cross-over potential V x and the photocurrent onset potential V O . For simplicity, a photocathode J-V curve was chosen. OPEC benchmark performances. Simultaneously, the eld should move towards reporting stabilised performances. Existing organic blend materials should be screened in order to nd candidates for improved stability in aqueous environments under applied potential and light bias. Furthermore, materials design should include synthesis strategies to improve material performance and stability and be tailored to OPEC water splitting needs. Identifying stable and high-performance materials for OPEC photocathodes will pave the way for stable tandem architectures that could demonstrate stable and efficient voltageunassisted water splitting with low-cost devices.

Conflicts of interest
The authors declare no conict of interest.