Solar H2 generation in water with a CuCrO2 photocathode modified with an organic dye and molecular Ni catalyst

H2 generation using a Ni catalyst on dye-sensitised CuCrO2 highlights the benefits of using delafossite semiconductors for solar fuel production.


Introduction
Articial photosynthesis offers a platform to produce a storable energy supply from fossil fuel-free resources. [1][2][3][4] This sustainable, carbon-neutral approach can produce a 'solar fuel' such as H 2 or carbon-based molecules from water or CO 2 using solar light. This process can be realised using semiconductor electrodes modied with suitable electrocatalysts in a photoelectrochemical (PEC) cell. [5][6][7][8][9] Electrodes featuring a molecular catalyst have advantages over 'conventional' heterogeneous systems as their 'single site catalysis' is atom-efficient, 10,11 they offer tunability and selectivity for challenging chemical transformations, [12][13][14][15] and can be rationally designed to enhance activity. [16][17][18][19] Their molecular nature also enables kinetic and mechanistic studies to reveal how they operate under various conditions, outlining routes to improvement. [20][21][22][23] Despite these advantages, the development of molecular-based photocathodes is held back by severe material limitations as state-of-theart electrodes currently lack the requirements of visible light absorption, mesoporosity, p-type conductivity, and/or stability in aqueous solution. 5,[24][25][26][27] To bypass these limitations, a modular approach can be adopted where a visible light-absorbing dye and a molecular catalyst are co-anchored to a stable wide bandgap semiconductor platform. 7,[28][29][30] In this dye-sensitised photoelectrochemical (DSPEC) system, the p-type semiconductor serves as the anchor site for the dye, which typically permits ultra-fast hole extraction following visible light excitation of the dye and minimises energy loss. The photoreduced dye is subsequently responsible for electron transfer to the coimmobilised electrocatalyst, where the reduction half-reaction takes place. The separation of light harvesting, charge transport, and catalysis allows the components to be individually tuned for optimal performance, where the rate of each transfer step inuences the overall device efficiency. 30 A suitable pair of photoelectrodes in a tandem DSPEC cell could provide an efficient and inexpensive means of solar fuel production, exploiting simple and adaptable preparation techniques. [31][32][33][34][35] The requirements for a robust DSPEC photocathode material are high p-type conductivity, propensity to anchor molecular moieties, high surface area, and a valence band position capable of readily accepting a hole from the photoexcited dye. 29,30,36 Several DSPEC photocathodes have already been reported with the majority relying on NiO, 18,[37][38][39][40][41][42][43] and the only other examples being modied ITO 44 and CuGaO 2 . 34 NiO is stable and easily synthesised in mesoporous form, [45][46][47] but suffers from the drawbacks of low charge carrier mobility and fast charge recombination between valence band holes and the reduced dye. 28,[48][49][50] Despite many efforts and different approaches to enhance the PEC properties of dye-sensitised NiO photoelectrodes, [51][52][53] improvements in performance are hindered by these limitations and there is a crucial need for better alternatives.
Wide bandgap Cu(I)-based mixed metal oxides such as Cu I M III O 2 delafossites (M ¼ Co, B, In, Sc, Cr, Al, Ga) have been employed in p-type dye-sensitised solar cells (p-DSSCs), 54,55 but their incorporation in solar fuel devices is limited. [56][57][58] The sole example of their use with a co-immobilised dye and molecular catalyst in solar fuel generation was reported for CO 2 reduction to CO with an anchored precious metal-based Ru-Re dyad on a CuGaO 2 delafossite electrode. 34 Delafossite CuCrO 2 has shown promise in p-DSSCs but application has yet to be extended to DSPEC cells despite it showing clear benets such as a low-lying valence band, high hole mobility, and simple and scalable synthesis. [59][60][61][62][63] In this study, we report solar H 2 generation with dyesensitised CuCrO 2 and demonstrate the feasibility of solar fuel synthesis with a CuMO 2 delafossite using precious metalfree dye/catalyst molecules. This was achieved by rst modifying CuCrO 2 with a phosphonic acid-bearing diketopyrrolopyrrolebased organic dye (DPP-P) and characterising the PEC reduction of a soluble electron acceptor in aqueous conditions. Then, a tetraphosphonated Ni-bis(diphosphine), [Ni(P 2 N 2 ) 2 ] 2+ , molecular catalyst (NiP) was co-immobilised to determine the PEC activity for the reduction of aqueous protons (Fig. 1a). The resulting hybrid DSPEC photocathode produces H 2 at moderate applied voltages with good photocurrents. Direct comparison with a corresponding NiO photocathode highlights the benets of CuCrO 2 and encourages the search for new DSPEC cathode materials.

Results and discussion
Synthesis and characterisation of CuCrO 2 Scalable and straightforward procedures for preparation of CuCrO 2 make it a highly accessible material, and its metal oxide character ensures that molecular species can be easily attached to the surface using anchoring groups such as phosphonic acids or carboxylic acids. [59][60][61][62][63] In this study, CuCrO 2 lms were grown directly on ITO-coated glass following a previously established sol-gel route. 59,60 In brief, a mixture of Cu(acetate) 2 , and triethanolamine (0.4 M) in absolute ethanol was spin-coated on an ITO-coated glass substrate. These samples were annealed in air at 400 C for 45 min before repeating the spin-coating and annealing steps to obtain a total of 6 layers. Post-annealing was carried out under N 2 at 600 C for 45 min to form the delafossite structure. NiO lms (2 mm thick) were prepared for comparison using a previously reported hydrothermal growth method. 37 CuCrO 2 crystallises in a rhombohedral unit cell (space group R 3m) and is a wide bandgap p-type semiconductor (E g ¼ 3.1 eV) exhibiting a low-lying valence band and high hole mobility. 63,64 The structure consists of 'innite' [CrO 2 ] layers of edge-sharing [CrO 6 ] octahedra linked by linear O-Cu-O dumbbells and the ptype conductivity stems predominantly from Cu + vacancies in the crystal lattice. 65,66 Favourable mixing of Cr 3d states with O 2p states increases the covalent nature of this interaction in the valence band, hence holes are more delocalised than in other corresponding delafossite structures, accounting for the intrinsic high hole mobility. 64,66 X-ray diffraction (XRD) analysis conrmed the rhombohedral delafossite structure for CuCrO 2 (Fig. S1 †) and scanning electron microscopy (SEM) images showed individual rods with a length of 73.3 AE 16.5 nm and thickness of 20.7 AE 3.7 nm, leaving a pore diameter of 16.7 AE 4.8 nm (Fig. 2a). The CuCrO 2 lm (resulting from 6 layers) was approximately 500 nm thick. N 2 gas adsorption isotherms showed type IV behaviour consistent with a mesoporous material and gave a BET surface area of 25 m 2 g À1 (Fig. S2 †), which is similar to that obtained with other mesoporous structures. 47 The direct bandgap of CuCrO 2 was estimated from a Tauc plot as 3.1 eV (Fig. S3 †) and the atband potential, E  , of +1.0 V vs. RHE with Mott-Schottky analysis from consecutive impedance scans (Fig. S4 †). This is 0.25 V more positive than the E  of our NiO electrodes. 37 See Experimental section for more details about synthesis and characterisation of the electrodes.

Components of the molecule-loaded CuCrO 2 photoelectrode
As dye and catalyst species, we selected DPP-P and NiP respectively, both recently synthesised in our group (Fig. 1a). 67,68 For the most suitable light absorber, a dye with sufficient driving force to reduce the H 2 evolution catalyst as well as a thermodynamically accessible reduction potential for the extraction of holes by CuCrO 2 is required. Diketopyrrolopyrrole (DPP) chromophores have recently displayed high activity with NiO in p-DSSCs and are considered suitable candidates due to their high photostability, simple synthesis and modication, and lack of precious metal elements. 36 DPP-P absorbs strongly in the visible range (3 496 nm ¼ 2.6 Â 10 4 M À1 cm À1 , DMF) 67 and is expected to undergo reductive quenching when immobilised on a p-type semiconductor due to fast hole injection originating from the proximity and good electrical communication between the dye and semiconductor. [69][70][71][72][73][74] In this pathway, the rst step upon dye excitation is the reduction of DPP-P* by hole injection into the valence band of CuCrO 2 , followed by oxidation of DPP-P À by the catalyst, which ultimately performs the chemical reaction. NiP, a Dubois-type Ni-catalyst 75,76 featuring four phosphonic acid anchoring groups, has previously demonstrated reduction of aqueous protons both in solution and when immobilised on a semiconductor surface whilst maintaining molecular integrity during photocatalysis. 5,6,67,68 DPP-P has a reduction potential in the excited state of +1.57 V vs. RHE and the reduced dye has an oxidation potential of À0.7 V vs. RHE, thus DPP-P À can provide sufficient driving force for the reduction of NiP to a catalytically active state (onset of catalytic current for NiP ¼ À0.21 V vs. RHE). 68 The respective electrochemical potential of each component and the hole and electron transfer pathways for the fully assembled DPP-P/NiP-modied CuCrO 2 electrode is shown in Fig. 1b and the corresponding energy diagram with possible recombination routes in Fig. S5. †

Photoelectrochemistry of CuCrO 2 |DPP-P
To evaluate the compatibility of DPP-P with CuCrO 2 and to ensure this interface could function without the kinetic limitations imposed by immobilisation of a molecular catalyst, PEC measurements were conducted on dye-sensitised electrodes in the presence of a soluble electron acceptor. These photocathodes were prepared by soaking CuCrO 2 electrodes in a DPP-P solution (1 mM, DMF) for 15 h. The UV-Vis spectrum of the electrodes with immobilised DPP-P displays an absorption maximum at approximately 500 nm, consistent with the electronic transition of the free dye ( Fig. 2b and c). 67 Linear sweep voltammetry (LSV) and chronoamperometry experiments were carried out in an aqueous Na 2 SO 4 electrolyte solution (0.1 M, pH 3) at room temperature in a N 2 -purged one-compartment threeelectrode electrochemical cell using a Pt counter electrode and a Ag/AgCl/KCl sat reference electrode. UV-ltered simulated solar light was used for all PEC measurements (100 mW cm À2 , AM 1.5G, l > 420 nm). In control experiments without the acceptor, the bare CuCrO 2 electrodes displayed a small cathodic dark current, which has previously been attributed to the reduction of Cu 2+ impurities to Cu + with oxygen deintercalation (Fig. 3a). 77 Irradiation of the unmodied and DPP-P modied electrodes resulted in only minor photocurrents without a soluble acceptor (|j| < 3 mA cm À2 , 0.0 V vs. RHE) (Fig. 3a).
Addition of the electron acceptor 4,4 0 -dithiodipyridine (DTDP, 5 mM) in the electrolyte solution allows for estimation of a maximal attainable photocurrent as DTDP is known to be easily reduced in solution (E red,DTDP ¼ À0.06 V vs. RHE). 37 The electron acceptor allows the photoreduced dye to dispose of photo-electrons and to regenerate the ground state, thereby limiting the effects of reductive dye decomposition and charge recombination, and dramatically enhancing the photocathodic response for CuCrO 2 |DPP-P. An absolute photocurrent response of z160 mA cm À2 (0.0 V vs. RHE, Fig. 3a) was observed, which indicates efficient light-induced hole injection from the dye to the valence band of CuCrO 2 with reduction of the acceptor by DPP-P À . For comparison, a NiO electrode sensitised in the same manner displayed a lower maximum photocurrent (|j| z 80 mA cm À2 , 0.0 V vs. RHE), suggesting lower susceptibility to recombination between the reduced dye and holes in CuCrO 2 (Fig. S6 †). Thus, DPP-P displays excellent electronic communication with CuCrO 2 , which suggests that coanchoring of a catalyst could be a viable approach to exploit the reductive power of DPP-P À for solar H 2 production.

Photoelectrochemistry with CuCrO 2 |DPP-P/NiP
Catalyst and dye molecules were co-immobilised on CuCrO 2 electrodes through soaking in a solution of NiP (0.5 mM) and DPP-P (1 mM), in DMF for 15 h. The loading of DPP-P was quantied by UV-Vis spectroscopy following desorption in alkaline solution and the amount of immobilised NiP determined using inductively coupled plasma optical emission spectroscopy (ICP-OES) measurements. This resulted in a 2 : 1 ratio of dye to catalyst on the electrodes (Table S1 †). Coimmobilisation of NiP and DPP-P on CuCrO 2 resulted in a ve-fold enhancement in photocurrent compared to the bare electrode (|j| ¼ 15.1 mA cm À2 , 0.0 V vs. RHE) (Fig. 3b). This increased response is attributed to the ability of DPP-P À to reduce NiP and ultimately protons. 67 This is supported by the incident photon-to-current efficiency (IPCE) spectrum, which displays a maximum photocurrent at the same wavelength as the absorption maximum of DPP-P (l max ¼ 500 nm, Fig. S7 †). For comparison, CuCrO 2 electrodes showed low efficiency and no peak at this wavelength, demonstrating the essential role of the sensitiser.
H 2 generation was studied using controlled potential photoelectrolysis (CPPE) under constant light illumination with an applied potential of 0.0 V vs. RHE. The CPPE trace of the CuCrO 2 |DPP-P/NiP electrode showed high stability over a 2 hour period (Fig. S8 †) with 94 AE 10 nmol of H 2 generated, corresponding to a turnover number of the NiP catalyst (TON cat ) of 126 AE 13 and a faradaic efficiency (FE) of 34 AE 8%. Possible explanations for the modest FE are the dark current originating from Cu 2+ reduction and oxygen deintercalation, 77 as well as capacitive currents due to the mesoporous structure or from electrons trapped in surface states. 56,[78][79][80] The FE is lowered by probable photobleaching/decomposition and desorption of the dye species, and is overall comparable to previously reported dye-sensitised photocathodes (Table 1). Control experiments without dye (CuCrO 2 |NiP) or catalyst (CuCrO 2 |DPP-P) produced no detectable hydrogen, conrming that the full assembly is required for catalysis. A comparable NiO|DPP-P/NiP electrode modied in the same manner only yielded 35 AE 2 nmol of H 2 aer 2 hours, with a FE of 31 AE 8%, demonstrating the superior performance (2-3 times) of CuCrO 2 (Table 1). Accurate quanti-cation of the Ni-catalyst loading on NiO was not possible by ICP-OES (same element in catalyst and substrate) or by UV-Vis spectroscopy following desorption (low molar absorption of NiP).
Post-electrolysis characterisation of CuCrO 2 |DPP/NiP electrodes using ICP-OES showed that the amount of NiP retained on the surface aer 2 h of CPPE was 54% of the initial loading (Table S1 †). This is in part due to the relatively low surface area exhibited by the delafossite particulates (25 m 2 g À1 ), which accounts for low loadings of catalyst and dye, and allows for their easy desorption into the media. Nanostructuring of the surface would ensure higher loadings of dye and catalyst species, enhancing both stability and activity in the future. Alternate methods such as atomic layer deposition (ALD) 52,81-83 or polymeric assembly [84][85][86][87] could also be employed as additional stabilisation methods.

Comparison with state-of-the-art
Limited improvements in photocathode development for DSPEC proton reduction are largely due to p-type materials with low performance. Since the rst report in 1999 towards p-type DSSC, dye-sensitised NiO electrodes have generated a range of benecial research on dye architecture and electrolyte composition. 48,88,89 Despite this, their performance remains signicantly lower than their n-type counterparts, highlighting the limitations of NiO and the need for a better alternative. Table 1 highlights relevant examples as a comparison for our system.
The TON cat is a good measure of catalytic activity for a molecular catalyst-based system but remains unreported in most cases. A TON cat > 125 aer 2 h for our CuCrO 2 system in water compares favourably with the currently highest reported value of z20 for a NiO DSPEC photocathode. 43 With NiP as the catalyst, an ITO electrode produced higher photocurrents and more H 2 , 44 but PEC activity has only been demonstrated for an applied potential of +0.05 V vs. RHE. CuCrO 2 allows for a much higher working voltage due to the onset potential being situated at +0.75 V vs. RHE and therefore shows greater suitability for energy storage and implementation in tandem DSPEC cells. This photocurrent onset is also more favourable than other commonly used narrow bandgap p-type semiconductors such as GaP, 90,91 and p-Si, 5,92 highlighting the benets of moving to dye-sensitised systems for H 2 generation.
CdSe-sensitised NiO produces the highest amount of H 2 of these electrodes over the duration of 2 hours of CPPE, 42 but a large portion of the photocurrent stems from the bare quantum dots. Despite this, sensitisation with quantum dot species is a viable approach to further enhance the H 2 producing capability of a CuCrO 2 -based photocathode in the future. In comparing these properties, it is clear that material alteration can have a great inuence on activity, and that transferring from NiO to CuCrO 2 has advantages for DSPEC applications.

Conclusions
We have introduced CuCrO 2 co-sensitised with an organic dye (DPP-P) and molecular catalyst (NiP) for DSPEC H 2 generation under aqueous conditions. CuCrO 2 |DPP-P/NiP showed a photocurrent onset at +0.75 V vs. RHE and a photocurrent density of 15 mA cm À2 at 0.0 V vs. RHE with a TON cat of 126 AE 13 achieved in controlled potential photoelectrolysis under UV-ltered simulated solar light irradiation. The molecule-loaded delafossite electrode therefore surpasses the performance of benchmark NiO electrodes in side-by-side comparison. We also show that the phosphonated organic DPP dye allows for high performance in aqueous conditions on an electrode and is able to electronically cooperate with NiP, which enabled us to assemble a fully precious metal-free DSPEC photocathode. The photocathode displays a higher photovoltage than other current state-of-the-art materials such as p-Si and GaP, making it well suited for coupling with a photoanode in tandem water splitting. Co-immobilisation of a dye and a CO 2 reduction catalyst on this p-type semiconductor may allow photocathodic production of carbon based fuels and chemical feedstocks.
The synthesis of CuCrO 2 by sol-gel techniques is straightforward and scalable. Nanostructuring would enhance the molecular loading and provide another avenue to increase photocurrents and the H 2 producing capability of the photocathode. Material alteration, for example through Mg 2+ doping, 62 could also improve the activity by further enhancing conductivity and therefore charge extraction through the lm. Other methods to improve the separation between catalyst and the delafossite surface would also . The slides were annealed in air to 400 C for 45 min with a ramp rate of 10 C min À1 in a chamber furnace (Carbolite Gero). These steps were repeated to form 6 layers. The nal annealing step involved heating in a N 2 atmosphere to 600 C for 45 min with a ramp rate of 5 C min À1 using a tube furnace tted with a quartz tube, end seals, and insulation plugs (Carbolite Gero). The electrodes were le to cool to room temperature and used as-prepared without any additional treatment.

Material characterisation
XRD measurements were conducted using a PANalytical BV X'Pert Pro X-ray diffractometer. SEM images were taken using a FEI Phillips XL30 sFEG microscope. UV-Vis absorption spectra were obtained using a Varian Cary 50 spectrophotometer in transmission mode.

N 2 gas adsorption measurements
Adsorption isotherms were carried out using a Micromeritics 3 Flex (Micromiretics, Norcross, GA, USA) with N 2 as the adsorbate. Samples were prepared on glass slides then scraped from the surface. Degassing for 10 h at 110 C was required prior to measurements, which were carried out in liquid N 2 . The BET specic surface area was obtained by tting N 2 isotherms using the Microactive soware.

Mott-Schottky analysis
Electrochemical impedance spectroscopy (EIS) measurements were conducted using an IviumStat potentiostat at 25 C using a 3-necked round-bottomed ask under dark conditions. A three-electrode setup using a Pt mesh counter, Ag/AgCl/KCl sat reference, and a CuCrO 2 working electrode (0.25 cm 2 active area) was used with an electrolyte solution of Na 2 SO 4 (0.1 M, pH 3). The frequency range was 10 kHz to 0.01 Hz, with an excitation voltage of 10 mV. Nyquist plots obtained in the potential range 1.1 V to 0.3 V vs. RHE (15 mV step) were tted using ZView® (Scribner Associates Inc.) to a Randles circuit (inset Fig. S4 †) to obtain interfacial capacitance (C sc ) values. The Mott-Schottky equation, used to obtain an estimate of the atband potential through a plot of 1/C sc 2 against the applied potential. A negative slope indicated p-type character and the x-intercept is equal to E  + k B T/e. 37

Electrochemical measurements
Cyclic voltammetry was used to determine the reduction potential of the DPP-P dye, E (S/S À ) , from the half-wave potential. This was performed in a 3-electrode setup with a glassy carbon working electrode, Pt-mesh counter electrode, and a Ag/AgCl/ KCl sat reference electrode with a scan rate of 50 mV s À1 . The electrolyte solution consisted of tetrabutylammonium tetra-uoroborate (0.1 M) in dry DMF with the addition of DPP-P (around 0.1 M). Addition of the E 00 to E (S/S À ) provides an estimate for the excited state reduction potential, E (S*/S À ) .

Modication of electrodes with dye and catalyst species
Molecular species were co-immobilised through soaking in a bath consisting of DPP-P (1 mM) and NiP (0.5 mM) in DMF for 15 h. For CuCrO 2 |DPP-P and CuCrO 2 |NiP electrodes the concentration was 1 mM but all other conditions kept the same. All electrodes were rinsed with DMF and H 2 O then dried in air and stored in the dark before use.
Quantication of loaded DPP-P and NiP DPP-P was desorbed from CuCrO 2 |DPP-P/NiP electrodes using a solution of 0.1 M tetrabutylammonium hydroxide 30-hydrate in DMF (1 mL) and the absorption at 500 nm was determined using UV-Vis spectroscopy. A calibration curve was used to t values and determine the loading for 4 different electrodes. NiP was quantied by ICP-OES aer digestion of CuCrO 2 |DPP-P/NiP electrodes (1 cm 2 lm area) in aqueous HNO 3 (70%, 1 mL) overnight and dilution to 10% v/v with MilliQ® water. CuCrO 2 |DPP-P/NiP electrodes pre-and post-electrolysis were analysed along with blanks for nitric acid, CuCrO 2 , and CuCrO 2 |DPP-P in triplicate. Errors represent standard deviation from the mean. 37

PEC measurements
Photoelectrochemical measurements were carried out using an Ivium CompactStat potentiostat in a one-compartment three-necked custom made cell equipped with a at borosilicate glass window. A three-electrode setup was used with a Pt-counter electrode, a Ag/AgCl/KCl sat reference, and the working electrode consisted of the CuCrO 2 platform with an illuminated area of 0.25 cm 2 conned using electrical tape. All measurements were conducted using aqueous Na 2 SO 4 electrolyte solution (0.1 M, pH 3) and the cell was purged with N 2 for 15 min prior to experiments. Frontside illumination was used for all experiments using a calibrated Newport Oriel solar light simulator (150 W, 100 mW cm À2 , AM 1.5G) tted with a UQG Optics UV Filter (l > 420 nm) and IR water lter. CPPE experiments were carried out in a custom twocompartment airtight electrochemical cell separated by a Naon membrane and featuring a at quartz glass window. The volume of electrolyte solution in the working compartment was 12 mL with a gas headspace of 5 mL while the counter compartment consisted of 4.5 mL solution and a 3.5 mL headspace. Prior to electrolysis, the gas headspace was purged for 30 min with 2% CH 4 in N 2 . An Agilent 7890A series gas chromatograph with a 5Å molecular sieve column and a thermal conductivity detector was used to quantify the amount of H 2 produced. The oven temperature was kept constant at 45 C and the ow rate was 3 mL min À1 . The partial pressure of H 2 was calculated to account for dissolved H 2 and this was added to the overall amount of hydrogen generated to obtain the faradaic efficiency. All CPPE experiments were carried out in triplicate with an applied potential of 0.0 V vs. RHE.

IPCE measurements
IPCE spectra were recorded in a N 2 -purged three-necked onecompartment custom cell with a at borosilicate glass window. A three-electrode setup with Pt counter, Ag/AgCl/KCl sat reference, and working electrode was used with pH 3 Na 2 SO 4 electrolyte solution (0.1 M). Monochromatic light was provided using a 300 W Xenon lamp solar light simulator coupled to a monochromator (MSH300, LOT Quantum design) and the intensity calibrated to 0.8 mW cm À2 for each wavelength. The potential was maintained at 0.0 V vs. RHE for all wavelengths and photocurrents were recorded in triplicate with different electrodes (0.25 cm 2 active area) for both CuCrO 2 and CuCrO 2 -|DPP-P/NiP arrangements.

Conflicts of interest
There are no conicts to declare.