10.2.1 Molecular Photocathodes for H₂ Evolution Based on Low Bandgap Semiconductors

This part is focused on p‐type inorganic semiconductors grafted with molecular catalysts [9] as photocathodes for H₂ evolution. The inorganic semiconductors currently in the limelight are silicon (Si, 1.1 eV bandgap), gallium phosphide (GaP, 2.2 eV bandgap), indium phosphide (InP, 1.3 eV bandgap), cuprous oxide (Cu₂O, 2.1 eV bandgap) and gallium indium phosphide (GaInP₂, 1.8 eV bandgap). They all possess a low bandgap responsible for visible‐light absorption [10, 11].

In the following, we will present the different photoelectrodes reported so far with a specific focus on the various methodologies used to construct such hybrid systems.

10.2.1.1 Molecular Catalysts Physisorbed on a Semiconductor Surface

The simplest method that has been used is to simply physisorb a molecular catalyst onto the semiconductor surface. This can be done by drop‐casting or soaking method. Nann et al. [12] used an iron sulfur carbonyl catalyst [Fe₂S₂(CO)₆], which can be considered as a mimic of the catalytic subsite of [FeFe] hydrogenase enzyme and therefore capable for catalytic proton reduction, to fabricate an InP‐based photocathode (Figure 10.3). The photocathode fabrication involves the layer‐by‐layer coating (10 layers) of InP quantum dots (QDs) onto gold support using thiol chemistry. The modified photocathode was immersed in a toluene solution of the synthetic mimic overnight, to allow spontaneous attachment of the diiron mimic onto InP QDs. Photoelectrolysis at −0.4 V vs Ag/AgCl in sodium tetrafluoroborate aqueous electrolyte in a three‐electrode system produced ∼250 nA cm⁻² current density under illumination with a 395 nm LED array_. The current density was stable for 1 h.

The same [Fe₂S₂(CO)₆] catalyst has also been physisorbed on planar silicon [13] and nanostructured silicon such as porous silicon [14, 15], porous silicon nanoparticles [16], silicon nanowires [13] and diatom‐derived silicon [17] to fabricate photoelectrodes for H₂ production. Photoelectrolysis was carried out in 0.5 M sulfuric acid (∼pH 0.3) as an electrolyte in a three‐electrode system under illumination with an Abet solar simulator (1 sun irradiation). A comparison of the studies using [Fe₂S₂(CO)₆] catalyst based on the approximate current density and faradaic efficiency is shown in Table 10.1. Stable photocurrents were observed for porous silicon and silicon nanowire photocathodes coated with [Fe₂S₂(CO)₆] catalyst for a reasonable amount of time. Although planar silicon coated with [Fe₂S₂(CO)₆] catalyst showed higher photocurrents than porous silicon with [Fe₂S₂(CO)₆] catalyst, stability was a major concern with planar silicon. In the case of particle systems coated with [Fe₂S₂(CO)₆] catalyst such as InP QDs, porous silicon nanoparticles and diatom‐derived silicon, lower photocurrent values and modest stability were observed when compared with robust porous silicon and silicon nanowire systems, respectively. The reason perhaps the low electronic conduction between particles and varied amount of particles attached on the electrode surface. However, silicon‐based molecular photocathodes improved the current density and, therefore, H₂ production, by several orders of magnitude when compared with previous work by Nann et al. [12].

Photocathodes coated with [Fe2S2(CO)6]	Applied potential (V vs RHE)	Current density (A cm−2)	Faradaic efficiency (%)	References
Gold electrode coated with InP QDs	−0.14	0.25 × 10−6	60	[12]
Planar silicon	−0.24	5.00 × 10−3	77	[13]
Porous silicon	−0.24	2.80 × 10−3	90	[14, 15]
Porous silicon nanoparticles	−0.13	2.20 × 10−6	14	[16]
Diatom‐derived silicon	−0.24	6.00 × 10−6	12	[17]
Silicon nanowire	−0.24	17.00 × 10−3	99	[13]

Chorkendorff and coworkers used the same strategy to decorate naked H‐terminated p‐Si(100) photocathodes with various sulfide clusters of molybdenum and tungsten (Figure 10.4) [18, 19]. Excellent photocurrent densities (12–14 mA cm⁻² at saturation) were observed under red light irradiation. Clearly, molybdenum‐based clusters (8 mA cm⁻² at 0 V vs. reversible hydrogen electrode (RHE)) are more efficient than tungsten‐based clusters (8 mA cm⁻² at 0.3–0.35 V vs RHE) to promote catalysis at low applied potentials. The presence of Cu in the tetranuclear clusters does not influence the results, whereas Co‐based clusters yield unstable performances. Slightly better performances are obtained when using silicon nanopillars [19]. Except for electrodes based on Co/W and Co/Mo clusters, stable performance over 1 h was observed with less than 5% loss of the photocurrent density over the first 30 min.

10.2.1.2 Covalent Attachment of the Catalyst to the Surface of the Semiconductor

To better control the stability of the catalyst linkage, molecular technologies based on semiconductor surface chemistry have been developed. Two approaches have been used. The first one consists in directly functionalising the surface of the naked semiconductor, whereas the second relies on the coating of the semiconductor with a layer of transparent metal oxide onto which the catalyst is covalently grafted.

The surface chemistry of silicon and other II–V semiconductors is rich and allows for the creation of stable covalent E−C (E = Si, Ga, …) bonds. For example, UV light can be used to induce the coupling of alkene to p‐Si(111) and p‐GaP(100) surfaces. This allowed Moore and Sharp to decorate such surfaces by protected pendant amino groups [20]. Further deprotection of the amino groups followed by amide coupling with a phthalimide‐functionalised nickel bis(diphosphine) catalyst, initially developed for a covalent attachment onto carbon nanotubes (CNTs) [21, 22], gave semiconductor substrates bearing the nickel complex at the surface as shown by grazing angle‐attenuated total reflectance (GATR) FTIR and XPS spectroscopies. GATR‐FTIR spectra of the functionalised substrates also display a vibrational band characteristic of surface oxide layers likely formed during surface functionalisation. No photoelectrochemical activity is reported for this construct, likely due to the presence of this passivating layer. In another study, the same methodology was used to graft a cobalt or iron porphyrins at the surface of p‐GaP(100) substrate (Figure 10.5) [24]. These electrodes proved active for light‐driven hydrogen evolution with photocurrents up to 1.3 mA cm⁻² at 0 V vs RHE in 0.1 M phosphate buffer (pH 7) and 1 mA cm⁻² at 0.35 V vs RHE for the cobalt porphyrin system. The current density was found to be stable for >5 min.

Rose and coworkers were successful in immobilising a nickel bis(diphosphine) complex (Figure 10.5) starting for a chlorinated p‐Si(111) wafer by a stepwise surface modification procedure exploiting palladium‐catalysed coupling chemistry [23]. Unreactive sites are blocked with methyl residues, which provide a good stability upon cycling in CH₃CN in the presence of trifluoroacetic acid (TFA). A 2.5 × 10⁻¹⁰ mol cm⁻² surface concentration of the Ni catalyst was estimated from XPS and electrochemical characterisations. Under 33 mW cm⁻² broadband LED irradiation, the onset HER potential was found at −0.06 V vs. normal hydrogen electrode (NHE), only 40 mV more negative than a control Pt‐coated p‐Si electrode. The covalently assembled semiconductor–molecular catalyst construct has an onset for light‐driven HER shifted 200 mV more positive than a methylated p‐Si photoelectrode in contact with a solution of the same catalyst.

An alternative method consists in coating the surface of the semiconductor by a layer of a transparent conducting or semiconducting oxide, typically TiO₂. Then the catalyst can be attached onto the metal‐oxide surface through the covalent binding of phosphonate or carboxylate groups. This stepwise functionalisation permits to take benefit of the three‐dimensional structuration of the metal‐oxide layer to increase catalyst loading. In addition, the deposition of a protection layer can be well controlled with such a stepwise strategy. A thin Ti layer or atomic layer deposition (ALD)‐deposited TiO₂ layers have been used in this context. Actually, this strategy was first used by Chorkendorff and coworkers as an alternative of the drop‐casting method presented above. Planar n⁺p silicon electrode was first protected by a 7‐nm‐thick metallic Ti layer and then coated with a 100‐nm‐thick TiO₂ film. A [Mo₃S₄] cluster bearing phosphonate functional groups was then attached onto the TiO₂ surface to yield an efficient photocathode developing 20 mA cm⁻² current density at 0 V vs. RHE and proven stability over 1 h (Figure 10.6) [25].

Turner and coworkers used a similar strategy to graft a cobaloxime catalyst onto TiO₂‐protected GaInP₂ semiconductor [26]. ALD was first used to protect the GaInP₂ surface against oxidative corrosion. A 35‐nm‐thick TiO₂ film proved optimal to obtain good photoelectrocatalytic performances after soaking the electrode in an ethanolic solution of a picolinic acid derivative of a cobaloxime. Further stabilisation of the electrode construct was demonstrated when additional layers of TiO₂ (10 cycles) were deposited using ALD (Figure 10.6). Current density up to 9 mA cm⁻² was measured at 0 V vs. RHE under 1‐Sun irradiation. The photocathode showed an initial decay of the photocurrent over the first 4 h and the current density was then sustained for 16 h at ∼5 mA cm⁻² corresponding to a turnover frequency (TOF) of 1.9 s⁻¹ in 0.1 M sodium hydroxide (pH 13).

Similarly, a molecular nickel bidiphosphine catalyst was attached through phosphonate linkage onto a silicon photocathode protected by an organic dimethoxyphenyl monolayer embedded into an aluminium–zinc oxide layer capped with a 20‐Å‐thick TiO₂ layer [27]. The passivation of the silicon surface with organic and inorganic coatings improved the stability as well as charge extraction towards the interface with the electrolyte where the catalyst operates. The onset potential was ∼200 mV more positive for such a semiconductor/organic coating/AZO/TiO₂ molecular catalyst system compared with semiconductor/organic coating/TiO₂ molecular catalyst system.

10.2.1.3 Covalent Attachment of the Catalyst Within an Oligomeric or Polymeric Material Coating the Semiconductor Surface

Although it is the last presented here, this strategy has been the first to be implemented to construct a H₂‐evolving photocathode based on silicon. In 1984, Mueller‐Westerhoff and Nazzal [28] covalently attached a 1‐methyl‐ferrocenophane catalyst to polymethylstyrene and coated this polymer onto a p‐type silicon photocathode (Figure 10.7). The polystyrene film was also used to protect the silicon surface from corrosion. H₂ evolution was observed with a saturating current density of 230 mA cm⁻² under 9 suns for 5 days in neat boron trifluoride hydrate electrolyte, the high acidity of which probably prevents passivation of the silicon electrode. H₂ evolution was also observed in aqueous acidic electrolytes such as perchloric acid (HClO₄), fluoroboric acid and hydrochloric acid. This field experienced a renaissance 5 years ago with the work of Moore and coworkers [29]. These authors investigated cobaloxime catalysts on GaP semiconductor photocathodes for H₂ evolution. Two different polymeric frameworks, polyvinylpyridine (PVP) [30, 31] and polyvinylimidazole (PVI) [32, 33], were investigated, both of which are able to coordinate cobaloxime in axial positions thanks to their pyridine or imidazole residues. Table 10.2 gathers the results obtained for based on H‐bridged cobaloximes [30, 31] as well as BF₂‐annulated cobaloximes [34]. Parameters such as light intensity [31], pH [34], orientation and crystal faces of GaP [33] have been varied. Table 10.2 can be summarised that H‐bridged cobaloxime shows greater efficiency when PVP was used a grafting polymer at pH 7, whereas BF₂‐annulated cobaloxime shows greater efficiency with PVP grafting polymer at pH 4.5 (Figure 10.7). This observation is to be confirmed with PVI grafting polymer, but is in line with previous observations that BF₂‐annulated cobaloximes require more acidic conditions to mediate electrocatalytic H₂ evolution [35, 36].

Cobaloxime	Grafting polymer	pH	Illumination (100 mW cm−2)	Orientation and crystal faces	j at 0 V vs RHE (mA cm−2)	References
H‐bridged	PVP	7	Newport Oriel apex illuminator model 71228 light source	100	∼2.80	[30]
BF2‐annulated	PVP	7	Solar light PV cell test simulator model 16S‐300‐005V4.0 xenon lamp	100	∼0.90	[31]
H‐bridged	PVP	7	PV cell testing 16S 300 W solar simulator	100	1.20	[34]
H‐bridged	PVP	4.5	PV cell testing 16S 300 W solar simulator	100	∼1.15	[34]
BF2‐annulated	PVP	7	PV cell testing 16S 300 W solar simulator	100	0.56	[34]
BF2‐annulated	PVP	4.5	PV cell testing 16S 300 W solar simulator	100	∼1.15	[34]
H‐bridged	PVI	7	100 W Oriel solar simulator	111A	∼0.89	[33]
H‐bridged	PVI	7	100 W Oriel solar simulator	111B	∼0.89	[33]
H‐bridged	PVP	7	100 W Oriel solar simulator	100	∼1.3	[32]
H‐bridged	PVI	7	100 W Oriel solar simulator	100	∼1.2	[32]

In conclusion, light‐harvesting semiconductor‐tethered molecular catalyst systems seem promising for H₂ production, especially as modern techniques allow to overcome surface passivation issues. However, the fabrication of bulk semiconductors often depends on rare earth elements such as In, Ga or Ge, which is not always economically viable; organic photovoltaic solar cells may be a cost‐effective alternative with identical designs of semiconductor systems.

10.2.2 H₂‐evolving Photocathodes Based on Organic Semiconductors

Organic semiconductor materials (OSCs) are conducting polymers, oligomers or self‐assembled discrete molecules that use abundant materials and can be deposited using low‐cost solution‐based processes. Additionally, OSCs display high optical absorption in the visible region of the solar spectrum. They generally possess much more reducing excited states than inorganic semiconductors and their electronic properties can be easily tuned through structural modifications [37]. These materials allowed to develop organic solar cells reaching 10% of power conversion efficiency (PCE) [38]. Light‐induced charge separation is achieved at the interface between the two OSCs. Upon absorption of a photon in one of the OSCs, an exciton (i.e. a pair of electrostatically bound electron and hole) forms and then migrates to the junction with the other OSC. Here, the exciton dissociates, resulting in the appearance of a hole in one OSC called the donor and an electron in the other OSC called the acceptor. In photovoltaic cells, electrons and holes then migrate to the electrodes where charges are collected. A key issue is the structuration of the two materials as excitons can only diffuse over a few tens of nanometres during their lifetime, that is, before spontaneous recombination takes place. Therefore, the two OSCs must be either deposited as two very thin layers, possibly with an interlayer containing both compounds, yielding a mixed interlayer junction, or as a blend of the two materials forming molecular p–n junctions all over the bulk layer (bulk heterojunction, BHJ). In general, BHJs are formed between a polymer or an array of small molecules, acting as the light absorber, and a fullerene derivative [39]. To use such organic junctions in the context of direct photoelectrochemical water splitting, different conditions should be met: (i) the organic layer in contact with a liquid electrolyte must maintain the capacity of light absorption and generating charges as well as the internal charge separation and transport to the active surface; (ii) an appropriate photovoltage must be reached to allow charge transfer between the organic layer and an acceptor located in the electrolyte or at the surface of the photoelectrode and (iii) the electrochemical reaction at the organic semiconductor/liquid electrolyte interface must be facilitated by an appropriate catalyst. Early reports described photocathodes made of conducting polymers, such as polyacetylene (CH) [40], polypyrrole [41], polyaniline [42], poly(3‐methylthiophene) [43] or poly(3‐hexylthiophene) (P3HT) [44, 45]. However, these photoelectrodes did not contain any junction between OSCs nor HER catalysts and charge separation arise from the Schottky junction between p‐type OSC and electrolyte. Hence, only low photocurrents (few to tens of microamperes per square centimetre) were obtained in aqueous electrolytes, depending on the solution pH, irradiation intensity, excitation wavelength and film thickness. Another step in the photoactive cathode design was to blend the p‐type P3HT‐conjugated polymer, with low bandgap and a high degree of intermolecular order leading to high‐charge carrier mobility [46], as the light‐harvesting unit and a fullerene derivative, such as phenyl‐C₆₁‐butyric‐acid methyl ester (PCBM), as electron acceptor in a BHJ structure in order to promote electron transfer. The resulting photocathode was used in aqueous saline solution [47]. Again, in the absence of any catalyst, very low photocurrents were observed and the faradaic yield for H₂ evolution was determined to be 0.3% [47].

Implementation of HER catalysts was found to be key to reach higher efficiencies. With Pt nanoparticles loaded onto the top of an organic p/n bilayer film of small molecules (metal‐free phthalocyanine, H₂Pc/fullerene, C₆₀), photocurrent values reached several cents of microamperes per square centimetre and the faradic efficiency for hydrogen production in aqueous media approached or even surpassed 90% (Figure 10.8) [48]. H₂ production at Pt‐coated C₆₀/water interface uses electrons from the conduction band of the C₆₀ layer. Although the bottom of the conduction band of C₆₀ is less negative than H₂ generation, the formation of the two‐electron reduced C₆₀ ²⁻ species was evidenced spectroscopically after a 20‐min induction period [48].

Al‐doped ZnO (40‐nm thickness) was used to cover P3HT/PCBM and acted as electron‐selective layer to a Pt/C catalyst [49]. The resulting photocathode delivers more than 1 mA cm⁻² at 0 V vs. RHE under neutral pH conditions with 100% Faradic efficiency after equilibrium between the Zn²⁺/Zn⁰ concentration equilibrium in the AZO film is reached.

The photoelectrode shown in Figure 10.8 has been combined with photoanode based on OSC to oxidise thiol [50] or based on TiO₂ to oxidise water [51] (Figure 10.9). By the use of these photoanodes, H₂ could be generated without additional electric bias. Of note, it is positive to invert the coating order and obtain an (indium tin oxide) ITO/n‐type OSC/p‐type OSC architecture with photoelectrochemical activity for water oxidation [52] or hydrazine oxidation [53, 54]. In the latter case, concomitant H₂ generation occurs at the counter Pt electrode without any additional electric bias.

Amorphous molybdenum sulfide (a‐MoS _x ) can be used as an earth‐abundant HER catalyst instead of Pt to develop photocathodes based on the same P3HT/PCBM BHJ [55]. It has been demonstrated recently that a‐MoS _x is a coordination polymer based on discrete trinuclear molybdenum disulfide clusters [56]. Photocathodes prepared by the direct spray deposition of mixtures of a‐MoS _x and TiO₂ on the organic BHJ‐active layer (Figure 10.10) deliver a photocurrent of 200 μA cm⁻² at 0 V vs. RHE and up to 300 µA cm⁻² at −0.4 V vs. RHE with a 0.6 V anodic shift of the H₂ evolution reaction onset potential, a value close to the open‐circuit potential of the P3HT:PCBM solar cell. The presence of TiO₂ nanoparticles in the MoS₃ film optimises both electron extraction at the PCBM interface and electron transport within the catalytic layer [55], as also observed in other architectures with compact TiO _x layer separating the P3HT:PCBM BHJ layer and the catalytic film [57]. Different interfacial layers were investigated to improve the charge transfer between P3HT:PCBM and a‐MoS _x . Metallic Al/Ti interfacial layers led to an increase in the photocurrent by up to 8 mA cm⁻² at 0 V vs. RHE. A 50‐nm‐thick C₆₀ layer also works as an interfacial layer, with a current density reaching 1 mA cm⁻² at 0 V vs. RHE [59]. A common issue is the long‐term stability of the light‐driven activity under operation. In a different architecture, the use of an insoluble cross‐linked PEDOT:PSS hole selective underlayer between the transparent conducting oxide substrate and the P3HT:PCBM BHJ layer proved to be quite efficient to avoid delamination of the organic layer and to allow sustained PEC activity for more than 3 h with faradaic efficiency of 100% in pH 2 aqueous media [57]. Alternatively, reduced graphene oxide, molybdenum, tungsten and nickel oxide underlayers proved efficient for sustained photoelectrochemical activity in acidic aqueous electrolytes [60, 61]. However, photoelectrodes based on P3HT:PCBM BHJ core do not allow to store much energy as the onset potential for H₂ evolution is marginally positive with respect to the equilibrium potential of the H⁺/H₂ couple. We then switched to planar p/n heterojunctions of small molecules and oligomers as this technology allows the use of a wider variety of donor and acceptor molecules. The use of α‐sexithiophene (α‐6T) and boron subphthalocyanine chloride (SubPc) as donor and acceptor OSC materials, respectively, in combination with C₆₀ as electron‐extracting layer (Figure 10.10) indeed resulted in a significant increase of both the onset potential and the solar‐to‐hydrogen power saved [59, 60] for H₂ evolution of OPV‐based photocathodes [58].

10.2.3 Dye‐sensitised Photocathodes for H₂ Production

Dye‐sensitised photocathodes for H₂ production are typically composed of three major parts: (i) a semiconductor (SC) layer for charge separation, (ii) a molecular dye for light harvesting and (iii) a catalyst for converting H⁺ to H₂. Upon visible‐light irradiation, two successive electron transfers take place (Figure 10.11): first, hole injection from the excited dye to the semiconductor valence band occurs and generates the reduced dye, which, in turn, transfers the electron to the H₂‐evolving catalyst. Obviously, two successive light‐induced cycles are required for hydrogen production to occur.

The development of such H₂‐evolving dye‐sensitised photocathodes follows on from the knowledge acquired on p‐type dye‐sensitised solar cells (p‐DSSCs) [62, 63]. Indeed, both systems rely on similar requirements: (i) a p‐type semiconductor fulfilling precise specifications such as transparency in the visible‐light domain, good carrier mobility, porosity to insure a high molecular loading, a valence band potential suitable for electron transfer to a molecular photosensitiser (Figure 10.11); (ii) the robust grafting of molecular photosensitisers onto the surface of the semiconductor and (iii) the choice of an efficient electron acceptor in order to avoid the detrimental recombination of the photoreduced dye with the holes generated in the semiconductor (Figure 10.11). However, two main differences also arise for the construction of H₂‐evolving dye‐sensitised photocathodes: first, the final electron acceptor is no more the electrolyte but the H₂‐evolving catalyst; as a consequence, the efficiency of the process does not rely anymore on a single electron transfer but on multielectronic catalysis, thus introducing kinetic limitations. Second, the system should operate in aqueous electrolytes, which introduces new constraints both regarding stability and electronics because of the pH dependence of the surface state of the semiconductor. Such additional complexity renders the development of efficient dye‐sensitised H₂‐evolving photocathodes more challenging. Most of the systems reported to date are assessed under a three‐electrode configuration using a reference electrode and a Pt counter electrode and not in a complete device. Examples of complete devices, such as tandem‐dye‐sensitised photoelectrochemical cells, are scarce, owing to the difficulty to find suitable working conditions, matching the requirements of both the anode and the cathode activities.

Currently, the range of known transparent and stable p‐type metal oxides that can be prepared from inexpensive materials and by techniques available to most laboratories is limited to NiO [64, 65], CuO [66], CuCrO₂ [67, 68], CuAlO₂ [68, 69] and CuGaO₂ [68, 70]. To date, the majority of reported photocathodes for H₂ production rely on NiO [5, 71], the preparation of which has been subject to optimisation for p‐DSSCs [64]. Nonetheless, NiO remains sub‐optimal as a p‐type semiconductor due to low hole mobility, an absence of hole traps and low permittivity [72]. Alternatively, quasi‐metallic oxides such as ITO also proved to be suitable porous materials for the construction of dye‐sensitised photocathodes [73–75].

A wide range of dyes suitable for sensitisation of p‐type semiconductors have been reported in recent years, and some structural prerequisites for good performance have been elucidated, thanks to intensive studies on p‐DSSCs [63]. It is important that the dye offers a large driving force for hole injection and long‐lived charge separation. The former can be optimised by modifying the electronic structure so that the oxidation potential of the dye (HOMO) is significantly more positive than the valence band of the NiO. The latter has been addressed by designing dyes based on a donor acceptor motif [72], also called “push–pull” dyes. The electron‐rich donor moiety (HOMO) is anchored to the NiO, whereas the electron poor acceptor (LUMO) is located away from the NiO, thus disfavouring recombination. The bridge between the two moieties is as long as possible and rigid to avoid bending of the acceptor unit towards the NiO surface. A range of chemical groups are available for anchoring molecules to metal‐oxide surfaces such as carboxylic acid, phosphonic acid, hydroxamic acid and silatranes [76]. Carboxylic and phosphonic acids are commonly utilised.

There is a wide range of molecular catalysts reported for proton reduction to hydrogen, which may be incorporated into a photocathode system. The catalyst may be dissolved in the electrolyte solution (Figure 10.11a) or attached to the electrode surface. In most cases where a catalyst is attached to the surface, it is functionalised with the same anchoring groups as the dyes used in DSSCs (Figure 10.11c). In some molecular systems, the catalyst is covalently attached to the dye and does not require a separate anchoring group (Figure 10.11b). It is desirable that the catalyst operates at a high rate (TOF), with minimal energy requirements (overpotential), and has high stability (turnover number, TON). One of the best performing catalysts for H₂ production is Pt. However, Pt is rare and there is not enough in the Earth's crust if all the vehicles in use today were to transition to a hydrogen economy [77], as such there is a large research effort to develop systems that use earth‐abundant elements [9, 78, 79]. This chapter focuses on catalysts using relatively abundant metals and Pt will not be discussed.

10.2.3.1 Dye‐sensitised Photocathodes with Physisorbed or Diffusing Catalysts

One of the earliest examples of a dye‐sensitised photocathode for H₂ production was reported in 2012 [80]. An organic dye, denoted P1, well‐established in the field of p‐DSSCs, was used as the sensitiser (Figure 10.12). P1 consists of a triphenylamine (TPA) donor unit, linked to two dicyanovinylene acceptor units via thiophene bridges, and a carboxylate anchoring group. The H₂‐evolving photocathode was built by simply dropping a cobaloxime catalyst solution onto the sensitised NiO surface and allowing it to dry. Cyclic voltammetry (CV) indicated that electron transfer from the reduced dye to the catalyst is energetically feasible. Chronoamperometry experiments (−0.4 V vs. Ag/AgCl) were performed under chopped light irradiation (>400 nm) and showed −4 µA photocurrent density with P1 alone compared with −15 µA in the presence of the cobaloxime. H₂ production was detected using a modified Clark‐type electrode. In both cases, the photocurrent decays over tens of seconds, but the decay is much faster with the catalyst and this is most probably due to rapid leaching of the cobaloxime into the electrolyte.

Later in the same year, a photocathode allowing H₂ production without any applied bias was reported in conjunction with a photoanode [82]. It consisted of a NiO film sensitised with an organic donor–bridge–acceptor dye, relying on a TPA unit bis‐functionalised with perylene monoimide (PMI) acceptors through sexithiophene bridges. This dye‐sensitised photocathode allows H₂ production with unity Faradaic efficiency, without any additional catalyst, which is unprecedented in the field. It produced a stable 3.9 µA cm⁻² photocurrent density and up to 0.6% incident photon to current efficiency (IPCE). In a tandem configuration with a BiVO₄ photoanode, H₂ was produced with 80% Faradaic efficiency without any applied bias. Furthermore, the dye‐sensitised film is stable in HCl and NaOH solutions for days; the observed stability is likely to be the result of the hexyl chains protecting the surface from the solution.

This stability effect was also recently reported for a photocathode sensitised with a related TPA–quaterthiophene–PMI push–pull dye, inspired from the thylakoid membranes of photosystems I and II (Figure 10.12) [81]. The presence of hexyl chains on the organic dye creates a hydrophobic layer over the electrode surface protecting not only NiO from acid dissolution but also the carboxylate anchoring of the dye from surface desorption. As a consequence, the electrode was stable at pH 0, which is the optimal pH for the [Mo₃S₄]⁴⁺ H₂‐evolving catalyst present in solution. Under these conditions, the dye‐sensitised photocathode maintained a photocurrent density of −183 µA cm⁻² for 17 h at 0 V vs. NHE with a Faradaic efficiency for H₂ production of 49%.

An alternative strategy for improving stability is the application of a thin layer of Al₂O₃ by ALD to prevent dye desorption. In one example, a PMI‐based dye‐sensitised NiO film was protected by 30 ALD cycles of Al₂O₃ [83]. A further advantage of Al₂O₃ over layers lies in the insulating properties of Al₂O₃ that can slow down charge recombination between the reduced dye and the NiO surface (see another example in Ref. [84] described below), as demonstrated with femtosecond transient absorption spectroscopy. The resultant electrode was stable to air and light over 45 days and could be assessed in H₂SO₄ aqueous solutions. Addition of either a nickel bis(diphosphine) or a cobaloxime H₂‐evolving catalyst to the solution gave significant photocurrent enhancement and led to hydrogen production with Faradaic efficiencies of 98% and 60% for the Ni and Co catalysts, respectively. When films without ALD treatment were assessed during 2 h with the Ni catalyst in solution, a lower Faradaic efficiency (80%) was measured and complete loss of the dye was observed, whereas the dye remained intact on ALD‐treated electrodes.

10.2.3.2 Dye‐sensitised Photocathodes Based on Covalent or Supramolecular Dye–Catalyst Assemblies

Reports on supramolecular or covalent dye–catalyst dyad assemblies grafted onto NiO are scarce. In principle, this is advantageous because charge transfer is no longer diffusion‐limited as for catalysts in solution, and the dye and catalyst are not competing for surface sites as occurs for co‐adsorption strategies.

The first example of a H₂‐evolving dye–catalyst assembly anchored onto NiO was reported in 2013 [84]. A cyclometalated tris‐heteroleptic ruthenium photosensitiser was employed. The phenylpyridine ligand was anchored to NiO through a carboxylate group and pendant pyridines on the bipyridine ligands were coordinated to a cobaloxime catalyst (Figure 10.13). CV showed that hole injection from the excited Ru dye to NiO and electron transfer from the reduced Ru dye to the cobaloxime (see Figure 10.11) were both energetically favourable. The anchoring shows good stability in water and phosphate buffer (pH 7) attributed to the increased electron density from the phenylpyridine at the carboxylate group. Upon irradiation, the onset potential of the catalytic wave is shifted 40 mV more positive, indicating that solar energy can be stored by the system. The IPCE spectrum matches the absorption spectrum of the dye and is enhanced in the presence of the cobaloxime. A single ALD layer of Al₂O₃, coated onto NiO before grafting the dye–catalyst assembly, was shown to more than double the photocurrent density; this beneficial effect was attributed to reduced hole–electron recombination across the NiO surface. The photocathode produced 290 nmol H₂ with a Faradaic efficiency of 45%, at −0.2 V vs. NHE for 2.5 h (pH 7 phosphate buffer). Shifting the applied bias positive to 0.1 V reduced the unproductive dark current to give a Faradaic efficiency of 68%.

The first fully noble metal‐free dyad was composed of a push–pull organic dye covalently attached to a cobalt diimine–dioxime catalyst by a copper‐catalysed azide–alkyne cycloaddition (“click” chemistry) (Figure 10.13) [85]. It was demonstrated that the dye could transfer electrons to the Co catalyst both on thermodynamic grounds and by the observation of the photoaccumulation of Co(I) under visible‐light irradiation in the presence of triethanolamine (TEOA) as electron donor. In MES buffer (pH 5.5) at 0.14 V vs. RHE, a photocurrent of −15 µA cm⁻² was observed, and H₂ was produced with a Faradaic efficiency of 10%.

In 2016, a layer‐by‐layer strategy was reported to construct a H₂‐evolving photocathode based on a [Ru(bpy)₃]²⁺ derivative photosensitiser and a nickel bis(diphosphine) catalyst. Both of them are functionalised by phosphonic acid groups to allow a covalent anchoring of the Ru dye onto NiO and the chelation of Zr⁴⁺ cations (Figure 10.14a) [86]. The assembly was built through a step‐by‐step procedure, the NiO surface being first covered by a monolayer of phosphonated photosensitiser, before soaking in a ZrOCl₂ solution, and finally in the phosphonic-acid-substituted catalyst solution. The best performing electrode was constructed with one layer of the phosphonic acid functionalised [Ru(bpy)₃]²⁺ and two layers of Ni catalyst. It displayed a photocurrent of −9 µA cm⁻² at 0.3 V vs. RHE and a Faradaic efficiency for H₂ production of 10%. Post‐photoelectrocatalysis characterisation of the electrode showed that no Ru was lost from the surface. The Ni catalyst was harder to detect due to the underlying NiO dominating XPS or inductively coupled plasma‐optical emission spectrometry (ICP‐OES) measurements. However, the CV of the electrode was unchanged after photoelectrolysis, suggesting little modification of the surface composition.

The following example is conceptually very similar to the previous one but relies on a conducting inverse opal nanostructured ITO electrode material (Figure 10.14b) [75]. Additionally, a phosphonic acid‐substituted dianiline (DA) was inserted between NiO and the Ru photosensitiser to provide longer‐lived charge separation, as confirmed with nanosecond transient absorption measurements: for the ITO‐DA‐Zr‐RuP₂ ²⁺ assembly in the absence of catalyst, a long‐lived signal at 510 nm characteristic for the reduced dye is observed, in agreement with hole injection into ITO. For the assembly including NiP, this signal can still be observed but with a lower yield, indicative of electron transfer to the catalyst within the first 50 ns of the experiment. Under photoelectrocatalytic conditions, the photocathode produced H₂ with 53% Faradaic efficiency and with a quantum yield of 0.85%, in MES buffer pH 5.1 at an applied potential of −0.25 V vs. NHE.

10.2.3.3 Dye‐sensitised Photocathodes Based on Co‐grafted Dyes and Catalysts

Co‐grafting dye and catalyst onto the electrode is an appealing alternative to covalent assembly of the two components, which often requires tedious synthetic steps. This strategy relies on electron hopping from the reduced dye to the catalyst on the surface for hydrogen production to occur. However, in sharp contrast with photoinduced intramolecular electron transfer processes within dyad assemblies, this process remained underexplored until the recent studies by the group of L. Hammarström. The first demonstration of electron hopping between molecules on NiO was reported in 2012 [87]. The dye used was Coumarin‐343 due to its high reduction potential and fast hole injection into NiO. The selected catalyst was a derivative of a known Fe phosphine complex (Fe1), reported to produce H₂ with a TOF of 10 000 s⁻¹ (Figure 10.14a). The radical anion of Fe1 could be observed within the 50 ns time resolution of the equipment and the charge recombined with holes in NiO over 100 µs. Interestingly, these processes were unaffected by the dyeing order of the dye and catalyst. In the absence of Fe1, no signals were observed on the nanosecond timescale, in agreement with the previously reported femtosecond hole injection from dye to NiO followed by recombination on the picosecond timescale. To rationalise these data, it is postulated that electrons migrate across the surface via a sub‐nanosecond self‐exchange mechanism between dye molecules. Due to low mobility in NiO, holes are localised; this self‐exchange mechanism may thus transport the electrons away from the holes to Fe1, explaining the nanosecond lifetime for the Fe1 anion despite dye–NiO charge recombination occurring on the picosecond timescale.

Four years later, the photoinduced dynamics of C343 on NiO was investigated with a different catalyst, a Fe carbonyl benzene dithiolate complex (Fe2 in Figure 10.15a) [88]. For co‐adsorbed C343 and Fe2, hole injection is observed from C343 to NiO on c. 200 fs timescale, as for C343 alone. However, for C343 alone, the reduced dye lifetime is hundreds of picoseconds, whereas with Fe2 the reduced C343 has a half‐life of 10 ps. In this case, the neutral dye signal recovers and the reduced catalyst signature is observed and persists for up to 30 µs. The electron transfer from dye to catalyst is three orders of magnitude faster than ‘hole‐hopping’ observed on TiO₂ [92, 93], and this may be due to the large driving force for electron transfer from dye to catalyst in this case. Unreactive C343 anion cannot be observed, suggesting that there are no dye molecules unable to transfer electrons to Fe2, either directly or by rapid hopping from dye to dye.

Fe2 showed poor binding to NiO and to improve on this, a novel catalyst was prepared with a phosphonate anchoring group (Fe3 in Figure 10.15a) [89]. Transient absorption spectroscopy of C343 on NiO (λ _exc = 440 nm) shows hole injection in <150 fs, in agreement with the previous work, and charge recombination over 9 ps. A co‐sensitised sample of C343 and Fe3 on NiO shows quite different kinetics. Hole injection is similar to C343 alone, but the signal for the C343 anion decays much faster and via several processes between 1 and 1000 ps. A signal corresponding to reduced Fe3 can be observed from 1 ps and persists for >2 ns. The decay of the C343 anion is assigned to the reduction of Fe3, and therefore, this reduction occurs via a combination of different processes. However, the reduction of Fe3 mostly takes place between 1 and 10 ps and with an electron transfer yield of 70%. This electron transfer occurs more quickly and in greater yields than previously observed for C343 with Fe2, and this may be related to the flexibility of the linking group. C343 co‐grafted with Fe3 on NiO was measured on a longer timescale in order to elucidate the fate of the reduced Fe3. The signal assigned to reduced Fe3 decays via multiple processes with time constants from ∼2 µs to >10 ms. The longer components, which outcompete the lifetime of reduced Fe2, are again attributed to the flexible linker in Fe3. Lifetimes of tens of milliseconds are important as protonation of the reduced Fe3 is expected to occur on the millisecond timescale. Indeed, the system produced H₂ (acetate buffer pH 4.5, −0.3 V vs. Ag/AgCl) with a Faradaic efficiency of 50% and a photocurrent of −10 µA cm⁻². IR spectroscopy showed that deactivation is, at least in part, due to the loss of the CO ligands and desorption of Fe3 from the surface, whereas the C343 is quite stable.

A photoelectrode was assembled with a bis‐phosphonate derivative of the well‐known [Ru(bpy)₃]²⁺ dye in combination with a cobaloxime catalyst possessing a pyridine dicarboxylate anchoring group (Figure 10.15b) [90]. Under a bias of −0.4 V vs. Ag/AgCl, a cathodic photocurrent of −13 µA cm⁻² was observed. Hydrogen production was not quantified, but the photocathode was combined with a molecular photoanode under a two‐electrode configuration. The tandem cell produced greater photocurrent than the photocathode alone but less than the photoanode alone, indicating that the photocathode activity is limiting the device performance.

Another example of a co‐grafted photocathode was prepared with CdSe QDs as the light absorber combined with a cobaloxime catalyst [119]. The QDs were capped with thioglycolic acid, which provides a carboxylate function for grafting onto NiO, whereas the cobaloxime contained a phosphonate anchoring group. The electrode was constructed from an open porous NiO structure where nanoflakes were grown from a nickel nitrate precursor, in order to have a high surface area and promote high loadings of sensitiser and catalyst. Indeed, when co‐grafted with QDs and cobaloxime, the photocurrent density was 1.7 times greater when compared with a classical doctor‐bladed NiO preparation. The electrode displayed stable photocurrents up to 110 µA cm⁻² at 0 V vs. NHE in neutral aqueous solution and produced hydrogen with 81% Faradaic efficiency. When a cobaloxime without a phosphonate anchoring group was used, the photocurrent rapidly decayed to 66 µA cm⁻², thus highlighting the benefit of co‐grafting the catalyst. Other QDs/NiO‐based photoelectrodes have been prepared with deposited MoS _x catalyst [94] or molecular cobalt dithiolene catalyst in solution [95].

In 2015, a tandem water‐splitting cell was reported based on molecular dyes and catalysts [91]. The photocathode consisted of the P1 dye and a cobaloxime catalyst, functionalised with an alkyl–phosphonate anchoring group, co‐adsorbed onto NiO (Figure 10.15c). The photocathode displayed a photocurrent of −45 µA cm⁻² on the timescale of the linear scan voltammetry (LSV) at applied potentials <−0.1 V vs NHE. This dropped to −20 µA cm⁻² at longer times. Hydrogen production was confirmed with a Faradaic efficiency of 68%. The photocathode was integrated into a tandem cell with a photoanode composed of an organic dye and a Ru‐based water oxidation catalyst co‐adsorbed onto TiO₂. The tandem cell gave a Faradaic efficiency of 55% for H₂ without any applied potential and a solar to hydrogen efficiency of 0.05%. In one example of a molecular photocathode not using NiO, a [Ru(bpy)₃]²⁺ derivative with pendant pyrrole groups was polymerised onto a carbon electrode, followed by electrodeposition of MoS _x [96]. The electrode gave a photocurrent of −25 µA cm⁻² with an applied bias of 0.1 V vs Ag/AgCl at pH 0.3. Under these conditions, H₂ was produced with 98% Faradaic efficiency. In terms of the quantity of MoS _x , a TON of 31 was obtained, equating to a TOF of 16 h⁻¹.

Proof of concept for dye‐sensitised H₂‐evolving photocathodes has been made in the last 5 years. However, low current density and poor durability remain as major challenges. A commercial device should perform efficiently for many months or years and yet most reports detail activity on the scale of hours.

10.3.1 Photocatalytic Systems Consisting of a Molecular Catalyst and a Semiconductor Photoelectrode

As already mentioned in the section about H₂ evolution, semiconductors with a narrow bandgap can absorb visible light, and it is expected that a combined system with molecular electrocatalysts will be useful even for selective CO₂ reduction (Figure 10.16).

In the initial stage of the studies, combinations of semiconductor photoelectrodes with solutions containing a molecular electrocatalyst were developed for CO₂ reduction. A current was obtained at potentials more positive than those of the theoretical redox potentials under irradiation of the semiconductor photoelectrodes. This indicated that the catalytic reduction reaction at the molecular catalyst progressed by utilisation of highly reducing conduction band electrons that were produced by photoexcitation of the semiconductor photoelectrodes. It was reported that macrocyclic Ni^II and Co^II complexes were found to be good electrocatalysts for CO₂ reduction to CO with p‐Si (1.1 eV) [100, 101], p‐GaAs (1.4 eV) [102] and p‐GaP (2.2 eV) [103] photoelectrodes. Recently, Kubiak and coworkers examined in detail a photoelectrochemical system using p‐Si as a semiconductor photoelectrode and Re(bpy‐Bu ^t )(CO)₃Cl (bpy‐Bu ^t  = 4,4′‐di‐tert‐butyl‐2,2′‐bipyridine, Figure 10.17a) as a molecular catalyst for CO₂ reduction to CO [106]. Briefly, when the p‐Si electrode was selectively photoexcited (λ _ex = 661 nm) in an Ar atmosphere in acetonitrile containing the Re complex and a supporting electrolyte, the potential of the reduction of the Re complex shifted in the positive direction by ∼500 mV compared with that of the corresponding reaction using a platinum electrode. Under a CO₂ atmosphere, the system afforded CO with a Faraday efficiency of 97% ± 3%, and its quantum efficiency for light‐to‐chemical energy conversion reached 61%. An interesting feature of this system was that direct hydrogen production from the p‐Si electrode surface could be controlled by the addition of water and variation of the concentration of the molecular catalyst, which afforded control of the CO/H₂ production ratio [107]. Fe [108] and Mn [109] complexes combined with silicon‐based photoelectrodes were also successfully applied as molecular catalysts for CO₂ reduction.

Immobilisation of molecular electrocatalysts on semiconductor electrodes was examined with respect to efficient electron injection from the photoexcited electrode to the electrocatalyst. Grätzel and coworkers recently reported a CO₂ reduction system using polycrystalline Cu₂O in the form of a p‐type semiconductor decorated with a Re(bpy)CO₃Cl molecular catalyst with a phosphonate anchoring unit (Figure 10.18) [110]. The photocathode was prepared by electrodeposition of Cu₂O followed by sequential modification with 20 nm of Al‐doped ZnO (AZO) and 100 nm of TiO₂ by ALD [111]. Cu₂O is generally photoelectrochemically unstable; however, decomposition of the Cu₂O electrode was effectively suppressed by protecting the surface with a TiO₂ layer, whereas the AZO layer between these acted as an n‐type buffer layer for constructing a p–n junction with the Cu₂O layer, thereby enhancing charge separation. Moreover, they synthesised a mesoporous scaffold with a thickness of 4.5–5 µm with 18 nm TiO₂ particles on the surface of the flat photoelectrode in order to support a sufficient amount of the Re catalyst. Photoelectrochemical measurements revealed that this photoelectrode generated substantial photocurrents exceeding 2.5 mA cm⁻² under simulated sunlight in a CO₂‐purged acetonitrile solution containing 0.1 M tetrabutylammonium hexafluorophosphate as an electrolyte. The mesoporous scaffold gave a 40‐fold increase in current compared with that of the flat photocathode. Control experiments were performed in the absence of Cu₂O and light. Six hundred millivolt more negative potential was required to observe catalytic current when a mesoporous layer of TiO₂ on FTO was used as the electrode. This indicates that the photovoltage was produced by photoexcitation of the Cu₂O layer. Notably, both the photocurrent and obtained photovoltage showed an improvement over a previously reported system that used a TiO₂/AZO/Cu₂O photoelectrode in an acetonitrile solution containing a dissolved Re molecular catalyst (2.1 mA cm⁻² and 560 mV) [104]; that is, the immobilisation of the molecular catalyst on the photoelectrode surface resulted in enhancement of the photoelectrochemical activity for CO₂ reduction. They suggested the following reasons for this: (i) electron transfer from the semiconductor electrode to the catalyst immobilised on the electrode surface should be more efficient compared to that in the combined system of the photocathode with the homogeneous Re catalyst in solution; (ii) immobilisation of the Re catalyst on the high surface area of the mesoporous TiO₂ scaffold could decrease light absorption by the catalyst, that is, the so‐called inner filter effect, resulting in effective formation of photogenerated electrons on the photocathode. The system displayed >80% Faradaic efficiency for CO generation from CO₂ in the first hour, although the photocurrent gradually decreased. The authors noted the possibility of catalyst deactivation on the electrode, which was possibly caused by structural changes in the bipyridine ligand of the Re catalyst.

Aiming to couple CO₂ reduction with water oxidation, Morikawa, Sato and colleagues immobilised a Ru complex polymer ([Ru(N^N)(CO)₂] ⁿ , RCP, N^N = diimine ligand) as a catalyst layer on a zinc‐doped indium phosphide photoelectrode (InP–Zn, E _g = 1.35 eV) [105]. This polymer was synthesised from the monomer shown in Figure 10.17b by photoelectrochemical polymerisation using the InP–Zn photoelectrode. Ru−Ru bonds were formed by a reductive reaction using electrons generated by photoexcitation of the InP–Zn photoelectrode, and then the pyrrole groups were oxidatively polymerised by anodic polarisation. This method was originally developed by Deronzier and coworkers [112]. The RCP‐modified p‐InP–Zn electrode generated a cathodic photocurrent under visible‐light irradiation in a CO₂‐purged aqueous solution, and the onset potential was 0.8 V more positive than that of the dark cathodic current. The formation of formic acid as a product of CO₂ reduction was confirmed using the RCP‐modified InP–Zn photocathode at a potential of −0.6 V vs Ag/AgCl under visible light (λ > 400 nm), and the Faradaic efficiency was 62.3%. These results suggest that photoexcited electrons in the conduction band of InP–Zn transferred to the RCP, resulting in subsequent reduction of CO₂ by the reduced RCP. Notably, the photocathode with the RCP, which was prepared by cathodic polymerisation only, showed only 34.3% of the current efficiency for formate formation. The authors noted that the two‐step polymerisation that formed the polypyrrole structure reinforced the contact between the RCP and p‐InP–Zn, resulting in a higher Faradaic efficiency. They also pointed out the possibility of the formation of gaseous products, for example, CO and H₂, as the reason for the decrease in the Faradaic efficiency, whereas no detailed analysis of the gas phase component was performed. They subsequently reported a photocathode using Cu₂ZnSnS₄ (CZTS, E _g = 1.5 eV) as a p‐type semiconductor, which consists of abundant and relatively cheap elements, with the same RCP catalyst [113]. CZTS modified with the RCP exhibited a much higher cathodic photocurrent than unmodified CZTS in CO₂‐purged water, which suggests that the photoexcited electrons in CZTS were transferred to the RCP during the CO₂ reduction reaction. Furthermore, both a higher photocurrent and production of formic acid were obtained by adding an equimolar amount of a Ru complex with phosphonate groups as anchor (Figure 10.17c) in the polymerisation process on the photocathode. The authors suggested that stronger contact with the RCP on the photoelectrode could be obtained by chemisorption of the phosphonate anchors, and thus a higher activity for CO₂ reduction was achieved. A photoelectrode in which S was partially substituted by Se (Cu₂ZnSn(S,Se)₄ – CZTSSe) was combined with the RCP. The photocurrent and formic acid generation were improved due to higher hole conductivity obtained by Se substitution. In both systems using CZTS and CZTSSe‐based photocathodes, the Faradaic efficiencies were ∼80%.

Morikawa and coworkers successfully applied these photocathodes with an n‐type semiconductor photoanode, which could use water as an electron source for the reduction of CO₂. The reaction proceeded via step‐by‐step excitation of both semiconductor electrodes, known as a Z‐scheme‐type electron transfer, using simulated solar light containing UV wavelengths. When the RCP‐modified semiconductor photocathodes were used for CO₂ reduction, the valence bands of the semiconductors of the photocathodes were too ‘shallow’; that is, they were more negative compared with the redox potential of water oxidation, and thus this reaction could not take place at the counter electrode without applying an external bias. To solve this problem, they utilised oxide‐based n‐type semiconductor electrodes (either TiO₂ or SrTiO₃) as the photoanode for water oxidation (Figure 10.19). First, a photoelectrochemical cell consisting of an RCP/InP–Zn photocathode and a Pt/TiO₂ photoanode was assembled [114]. The electrodes were partitioned by a Nafion film, and CO₂ and Ar were introduced into the cathodic and anodic sides, respectively. The cell was then irradiated with simulated sunlight (AM 1.5) from the anode side without bias. The TiO₂ absorbed only the UV light and the InP–Zn absorbed the remaining light transmitted through the photoanode. Formic acid was generated linearly from the cathodic side and, at the same time, oxygen was formed at the anodic side. Isotopic labelling experiments clearly indicated that formic acid and oxygen were derived from CO₂ and water, respectively. In this system, the conduction band edge potential of TiO₂ was 0.5 V more negative than the valence band edge potential of InP. Thus, photogenerated electrons in the conduction band of TiO₂ could move to the valence band of the excited InP through the outer circuit, resulting in Z‐scheme‐type electron transfer. The efficiency of the conversion of sunlight energy to chemical energy was 0.03%.

The same group achieved much higher energy conversion efficiency by changing the photoanode from TiO₂ to SrTiO₃ [115]. As the bandgap of SrTiO₃ (3.2 V) is larger than that of TiO₂ (3.0 eV), shorter wavelengths of light would be absorbed. The SrTiO₃ photoanode generated an anodic photocurrent at a much more negative potential compared with that of the TiO₂ electrode, which indicated that the conduction band potential of SrTiO₃ was located at a more negative potential than that of TiO₂, and the driving force of electron transfer to the photocathode increased. The photocurrent generated from the photoelectrochemical cell using SrTiO₃ and RCP/InP–Zn was seven times higher than that from the cell using the TiO₂ photoanode, and the energy conversion efficiency reached 0.14%.

The same group also reported a monolithic stand‐alone device composed of an amorphous Si/SiGe triple junction tandem photovoltaic cell and the RCP catalyst. This system was highly active for the reduction of CO₂ to HCOOH with 4.6% energy conversion efficiency, where the RCP catalyst was separated from the light absorber (the photovoltaic) with a carbon cloth substrate [116].

10.3.2 Dye‐sensitised Photocathodes Based on Molecular Photocatalysts

The use of molecular photocatalysts immobilised on electrodes for CO₂ reduction is attractive because the molecular structure may be tuned in order to optimise redox properties. They can also display high catalytic efficiency and selectivity for CO₂ reduction. As is the case for the H₂‐evolving photoelectrodes described in Section 10.2.2, molecular photocatalysts need to be fixed to a p‐type semiconductor electrode (e.g. NiO) to drive CO₂ reduction with irradiation and using electrons injected from the valence band of the semiconductor (Figure 10.20).

In 2014, Inoue and colleagues reported a system in which a supramolecular photocatalyst composed of a zinc porphyrin sensitiser and a Re catalyst unit (Figure 10.21a) was immobilised on NiO with carboxylic groups as anchors [6]. This supramolecular photocatalyst strongly absorbs visible light, which is attributed to the Soret band of the Zn porphyrin at 430 nm. They first demonstrated the photocatalytic activity of this system under irradiation with visible light (λ _ex = 430 nm) in a DMF–triethylamine (TEA, 4 : 1, v/v) mixed solution. The TON of CO formation and the quantum yield were 14% and 0.12%, respectively. They then examined the immobilisation of the supramolecular complex on NiO in order to perform photoelectrochemical CO₂ reduction. The fluorescence lifetimes (τ) of both the Zn porphyrin mononuclear complex and the supramolecular complexes immobilised on NiO were shortened to about 30 ps from that in solution (τ < 2 ns), which suggested efficient electron injection from NiO to the excited state of the Zn porphyrins. Nanosecond laser flash photolysis experiments showed the formation of a radical anion of the Zn porphyrins with a lifetime of 180 ns. CO₂ reduction was performed using the photoelectrode with the Zn porphyrin–Re supramolecular photocatalyst in CO₂‐saturated DMF solution containing tetrabutylammonium hexafluorophosphate (0.1 M) as a supporting electrolyte under irradiation at λ _ex = 430 nm. This gave CO as a main product and a TON of 10 based on the immobilised complex, although the Faradaic efficiency was only 6.2%. An alternative photocathode, which was prepared by co‐adsorption of the supramolecular complex with Zn porphyrin as an additional photosensitiser at a ratio of 24 : 1 (mononuclear sensitiser: supramolecular sensitiser), gave a TON of 122 based on the Re catalyst unit. This result was believed to imply that electrons are concentrated in the Re complex unit in the supramolecular photocatalyst by electron hopping between molecules.

Ishitani and colleagues employed a supramolecular photocatalyst, which consisted of a Ru(II) photosensitiser and a Re(I) catalyst with methylphosphonate groups as anchors (RuRe, Figure 10.21b), for photoelectrochemical CO₂ reduction by immobilisation on a NiO electrode [7]. This photocatalyst showed high activity for CO₂ reduction in an aqueous solution (quantum yield for CO formation = 13%, TON = 130) [117] as well as in organic solvent. As the valence band edge potential of NiO was more negative than the reduction potential of the excited state of the Ru photosensitiser unit, electron transfer from NiO to the light‐absorbing Ru complex was expected to proceed. This process was clearly observed as a cathodic photocurrent when using a model mononuclear Ru complex on NiO in acetonitrile electrolyte (0.1 M Et₄NBF₄) under irradiation by visible light, which selectively excites the Ru complex. CO₂ reduction was examined using the photoelectrode modified with RuRe at an applied potential of −1.2 V vs. Ag/AgNO₃ in a DMF–TEOA (5 : 1, v/v) mixed solution. A cathodic photocurrent and corresponding selective CO generation were observed, and the TON of CO formation based on immobilised RuRe was 32 after 22 h irradiation. This photoelectrode was later successfully employed for CO₂ reduction in an aqueous solution [118]. In a CO₂‐purged aqueous solution containing 50 mM NaHCO₃ (pH = 6.6), a photocathodic response was clearly observed under irradiation at λ _ex = 460 nm, where only RuRe was photoexcited. The photoelectrode was irradiated for 12 h with an applied bias of −0.7 V vs. Ag/AgCl, giving 361 nmol of CO (TON_CO = 32 based on the amount of RuRe adsorbed on the NiO electrode). Although a small amount of H₂ was produced as a by‐product, the total selectivity of CO formation was 91%. This photocathode was then successfully combined with a CoO _x /TaON photoanode, which was reported as an efficient photoanode for water oxidation under visible‐light irradiation, in a functional photoelectrochemical cell (Figure 10.22). This cell produced 79 nmol of CO with a small amount of H₂ (6 nmol) and O₂ (77 nmol) from the photocathode and photoanode, respectively. The cell was operated under visible‐light irradiation (λ _ex > 400 nm) with an applied electrical bias of 0.3 V and a chemical bias of 0.10 V originating from the pH difference between the electrolytes for each electrode. The light energy conversion efficiency of the reaction was 1.6 × 10⁻⁵, considering the electrical and chemical bias. This is the first example of a visible‐light‐driven CO₂ reduction system using water as the reductant, based on molecular photocatalysts hybridised with semiconductor photocatalysts.