Dye–catalyst dyads for photoelectrochemical water oxidation based on metal-free sensitizers

Dye-Sensitized Photoelectrochemical Cells (DS-PECs) have been emerging as promising devices for efficient solar-induced water splitting. In DS-PECs, dyes and catalysts for water oxidation and/or reduction are typically two separate components, thus limiting charge transfer efficiency. A small number of organometallic dyes have been integrated with a catalyst to form an integrated dye–catalyst dyad for photoanodes, but until now no dyads based on metal-free organic dyes have been reported for photoanodes. We herein report the first example of dyad-sensitized photoanodes in DS-PEC water splitting based on metal-free organic dyes and a Ru catalyst. The di-branched donor–π–acceptor dyes carry a donor carbazole moiety which has been functionalized with two different terminal pyridyl ligands in order to coordinate a benchmark Ru complex as a water oxidation catalyst, affording water oxidation dyads. The two dyads have been fully characterized in their optical and electrochemical properties, and XPS has been used to confirm the presence of the catalyst bonded to the dye anchored to the semiconductor anode. The two dyads have been investigated in DS-PEC, showing an excellent faradaic efficiency (88% average across all cells, with a best cell efficiency of 95%), thus triggering new perspectives for the design of efficient molecular dyads based on metal-free dyes for DS-PEC water splitting.


Introduction
In the search for an alternative to fossil fuels, hydrogen is gaining increasing interest in the scientic community and the automotive-aerospace sector, because its combustion generates only water with a zero-carbon footprint. [1][2][3] Hydrogen is today mainly produced from methane via the steam-reforming reaction. 4 It is, therefore, critical to develop sustainable and clean processes to produce hydrogen. In this scenario, solar-driven water splitting has been playing an increasing role, with particular reference to the direct conversion from water and sunlight via photocatalysis (PC) or photoelectrochemical cells (PEC). 5-7 A PEC is a device composed of two electrodes soaked in a water-based electrolyte, where at least one of them is photoactivated. The photoelectrodes typically consist of a wide bandgap semiconductor (SC), which can be sensitized adequately by a dye to extend the absorbed spectrum into the visible region. 8,9 In the case of a DS-PEC with only a photoactive anode, a platinum wire is used as a passive cathode. The sensitizer is adsorbed or chemically bonded onto the SC surface, while the WOC, typically a ruthenium complex, could be either dispersed in aqueous media 10,11 or adsorbed onto the SCs surface. 12,13 Such arrangements are not optimal for the whole process. In the former case, the device efficiency strongly depends on the diffusion of the catalyst near the electrode. In contrast, in the latter one, the adsorption of the catalyst on the SC surface may compete with the absorption of the dye, limiting the light-harvesting capability. An innovative way out is to design an integrated dye-catalyst supramolecular system, also referred to as a dyad, where the molecular sensitizer and the catalyst are bound together as a single molecular unit. In this way, the drawbacks of the two previous arrangements are circumvented, providing more efficient light-harvesting and a faster charge transfer between the two components. Despite its great potential, only a few examples of molecular dyads for photoanodes have been so far reported in the literature. These examples are limited to organometallic compounds in the dye part of the dyad (such as Ru complexes or porphyrins). [14][15][16] The absence in the literature of organic dyes in dye-WOC dyads for photoanodes is surprising, considering the emerging role of metal-free dyes in the elds of solar energy conversion. 8,[17][18][19] Only two examples of dyads containing organic dyes have been so far reported, but both referred to photoactive cathodes. 20,21 We thus decided to investigate a dyad where a metal-free sensitizer is covalently bonded to a typical WOC, oen used as a catalyst benchmark (a derivative of [Ru(bda)(pic) 2 ] (bda ¼ 2,2 0 -bipyridine-6,6 0 -dicarboxylate; pic ¼ 4-picoline)). 22 As a metal-free sensitizer, we have selected a di-branched donoracceptor D-(p-A) 2 molecule that we have previously investigated, as a class, in the eld. 10,23,24 The donor moiety was functionalized with a pyridine ligand, which is then used to coordinate the Ru centre of the WOC complex. To study the effect of different spatial arrangements between the dye and the catalyst, two pyridine ligands, with a para and meta relative position between the coordinating nitrogen atom and a methylene linker to the dye, were envisaged, affording the CBZ-4Py and CBZ-3Py precursors, respectively (Fig. 1). The methylene linker of the picolyl functionality was introduced to break the pconjugation path between the dye and the catalyst. This ensures that the dyad is constituted by two separated p-conjugated entities, thus avoiding perturbation of the optical, electrochemical, and catalytic properties of the two components, and circumventing charge recombination processes. 16,25 The two corresponding molecular dyads CBZ-4Py + Ru and CBZ-3Py + Ru were obtained through a two-step procedure directly on the electrode and fully characterized in their optical and electrochemical properties.
When used in water splitting PEC, the investigated dyads showed high faradaic efficiencies (FE), with the meta-pyridine dyad CBZ-3Py + Ru showing the highest FE.
As the metal-free D-(p-A) 2 component of the dyads, we selected a carbazole (CBZ) derivative, which has been functionalized with pyridine linkers. The CBZ donor building block has been selected as a versatile sulphur-free alternative to the commonly used phenothiazine scaffold. 8,18,26 The central pyrrole-like ring ensures the strong electron-donor character. 27 The presence of the two benzofused benzene rings can be exploited to easily build the two (p-A) arms carrying the terminal cyanoacrylic acceptor-anchoring groups. Moreover, the NH group of CBZ can be easily functionalised with the pyridine linker needed to couple to the WOC. CBZ-based sensitizes have been successfully employed in PC and PEC water oxidation and hydrogen generation. 10,23 As WOC, we selected a close derivative of the ruthenium-based benchmark for DS-PEC, [Ru(bda)(pic) 2 ], which performed well in many literature examples. 28

General information
Spectroscopic and electrochemical investigation of dyes. Absorption spectra were recorded with a V-570 Jasco spectrophotometer. UV-O 3 treatment was performed using Novascan PSD Pro Series -Digital UV Ozone System. The thickness of the layers was measured utilizing a VEECO Dektak 8 Stylus Proler. Cyclic Voltammetry (CV) was carried out at a scan rate of 50 mV s À1 , using a Bio-logic SP-240 potentiostat in a three-electrode electrochemical cell under Ar. The working, counter, and the pseudo-reference electrodes were an FTO glass for the dyes in solution or a sensitized 3 mm-thick TiO 2 lm, a Pt wire and an Ag/Ag + electrode (0.01 M AgNO 3 , 0.1 M TBAClO 4 in ACN) or an Ag/AgCl electrode (3 M KCl), respectively. The preparation and sensitization of the 3.5 mm thick TiO 2 lm are described below. The Pt wire was sonicated for 15 min in deionized water, washed with 2-propanol, and cycled for 50 times in 0.5 M H 2 SO 4 before use. The Ag/Ag + and the Ag/AgCl pseudo-reference electrodes were calibrated by adding ferrocene (10 À3 M) to the test solution aer each measurement (potentials measured versus Fc/Fc + and converted to NHE by addition of +0.63 V; 29 NHE converted to vacuum by addition of À4.6 V). 30 Preparation of photoanodes. The photoanodes have been prepared as described below, adapting a procedure reported in the literature. 31 All the containers used were in glass or Teon and were treated with EtOH and 10% HCl before use to exclude metal contamination. FTO glass was cleaned in a detergent solution for 15 min using an ultrasonic bath, rinsed with pure water and EtOH. Aer treatment in a UV-O 3 system for 18 min, a transparent active layer of 0.8 cm 2 was screen-printed using Dyesol 18NR-T active transparent TiO 2 paste. The coated lms were thermally treated at 125 C for 6 min, 325 C for 10 min, 450 C for 15 min, and 500 C for 15 min. The heating ramp rate was 5-10 C min À1 .
FTO plates coated with 3.5 mm transparent TiO 2 lm, prepared as described above, were treated in a UV-O 3 system for 20 min at room temperature, then immersed into a 2 Â 10 À4 M solution in EtOH of the dye + catalyst precursor for 3 h at room temperature in the dark. The stained substrates were rinsed with EtOH and dried with a stream of dry nitrogen. The UV-vis spectra, CV and LSV were recorded in comparison with a bare 3.5 mm transparent TiO 2 lm.
X-ray photoelectron spectroscopy (XPS) measurements. XPS was performed on Thermo Scientic K-Alpha X-ray photoelectron spectrometer with a monochromatic Al Ka X-ray source under high vacuum (2 Â 10 À8 mbar), using an electron ood gun to avoid sample charging. The data tting was performed by Thermo Avantage v5.9911 soware, using mixed Gauss-Lorentz functions. IPCE, LHE and APCE measurements. IPCE measurements were performed using a monochromator (JASCO) illuminating on the active area of the working electrode (0.80 cm 2 ). The light intensity was monitored using a reference Si cell photodiode (THORLABS, S120VC) and corrected to calculate the IPCE values. The photocurrent was measured using an AUTOLAB PGSTAT302N potentiostat in a three-electrode purposely designed photoelectrochemical cell and the data collected with GPES electrochemical interface (EcoChemie). LHE spectra have been recoded, using the same samples, with a spectrophotometer Jasco V-570. APCE were calculate according to the relationship: Oxygen evolution quantication by collector-generator technique. The dyad-sensitized photoanode was illuminated under an applied bias, thus acting as an O 2 generator. An FTO electrode (previously cleaned via 15 min sonication in EtOH) was sandwiched to the photoanode (with both conducting sides facing inward), in order to be used as the collector, i.e. the electrode at which the reduction of the evolved O 2 takes place. The sandwiched device was held together by 4 layers of unstretched paralm (ca. 400 mm spacing), sealed by pressing at 65 C for 60 s. The two lateral sides of the C-G sandwich were le open, allowing for the lling of void space between electrodes with electrolyte solution by capillary forces. Both the photoanode generator and the FTO collector were contacted using Cu tape (also covered by the paralm layers), and respectively connected to the two working electrodes of a bipotentiostat (Autolab PGSTAT302N). An Ag/AgCl electrode and a Pt wire were used as the reference and the counter electrode. The sandwich was then immersed in degassed Na 2 SO 4 0.1 M solution (pH 5.8) and illuminated with a 200 W xenon lamp (using a LOT-Oriel xenon white light source, equipped with a 420 nm cut-off lter to minimize TiO 2 contribution and an IR lter to avoid cell warming). The generator electrode was held at 0.3 V vs. Ag/ AgCl bias ($0.5 V vs. NHE), while the collector at À0.8 V vs. Ag/ AgCl ($À0.6 V vs. NHE), identied in literature as the optimal O 2 reduction potential. 32 In a typical experiment, the working electrodes' currents were recorded for 300 s in the dark, then during 300 s of illumination, and in the dark for an additional 300 s. The procedure allows the diffusion of oxidation products across the solution. The faradaic efficiency for O 2 production, h O 2 , can be calculated by: where Q coll is the integrated current measured at the collector electrode, Q gen is the integrated photocurrent measured at the generator electrode, and h coll is the collector efficiency. The latter must be quantied under the specic set-up and experimental conditions used; in our case 82%, in good agreement with the one reported by the Finke 33 group on a very similar setup (i.e. with a small gap between the generator and collector).

Synthesis
The synthesis of the dyes CBZ-4Py and CBZ-3Py functionalised with the two pyridine linkers is presented in Scheme 1. The Suzuki-Miyaura cross-coupling between commercially available 9H-3,6-dibromocarbazole and the protected pinacol ester of 5formyl-2-thienylboronic acid 35 affords intermediate 2 with protected formyl groups. The deprotonated derivative is then submitted to alkylation with the proper picolylbromide to give the pyridine derivatives 3a and 3b. Subsequent deprotection of the formyl groups and nal Knoevenagel condensation with cyanoacetic acid, in the presence of piperidine, yields the desired sensitizers with the pyridine ligand. The new sensitizers have been fully characterised with 1 H and 13 C NMR, as well as melting point, FT-IR, and HRMS to check the required structures and purities (see ESI †).
A stoichiometric amount of the dye was added to [Ru(bda)(DMSO)(MeCN)(pic)], 14,16 but, unfortunately, the desired product could not be characterized due to good solubility only in coordinating solvents, thus decomposing the Ru complex of the dyad. To overcome this issue, the dyad was completed directly on the electrode surface, starting from the synthetic precursors. A TiO 2 electrode was soaked into a 2 Â 10 À4 M EtOH/DMSO (9 : 1) dye solution in the presence of a stoichiometric amount of the Ru precursor. To eliminate the unreacted Ru precursor, the photoelectrode was then rinsed with EtOH. Since the Ru precursor does not carry any anchoring group, the nal step ensures that the presence of Ru in the rinsed electrode can only be due to the formation of the desired complex with the functionalized dye, in turn, anchored to the SC via the acceptor-carboxylic functionality.
Cyclic voltammetry (CV) proles (Fig. S2, ESI †) showed irreversible behaviour for the oxidation processes of CBZ-3Py and CBZ-4Py. Differential Pulsed Voltammetry (DPV) (Fig. S3, ESI †) was thus used to determine the oxidation potential from the current peak 38 and then calculate the HOMO energy levels, which were estimated to be around À6 eV. This value is lower than the electron-donating level of the WOC (À5.8 eV), 39 and therefore dye regeneration can take place. The LUMO levels have been derived from electrochemical HOMO values and optical bandgaps, calculated as described above. The so calculated LUMO energies are also very similar, and both are at higher energy than that of the CB of TiO 2 (À4.0 eV vs. vacuum), thus ensuring efficient electron injection from the excited dye to the SC. Therefore, the electrochemical study guarantees proper energy levels alignments for electron-hole donation among the dye and other components of the cell.

Characterisation of the photoanodes
The electrochemical properties of the dyad-sensitized anodes have been investigated by CV and compared with those of the dye and the reference catalyst [Ru(bda)(pic) 2 ] (Fig. S4 and S5, ESI †). The CV prole of the CBZ-4Py + Ru-sensitized electrode presented a redox peak at E 1/2 ¼ +0.70 V vs. NHE, which well compares with the reported value for the reference Ru complex (+0.65 V vs. NHE). 39 The CBZ-3Py + Ru-sensitized electrode showed a less intense, but still reversible, redox peak at E 1/2 ¼ +0.77 V vs. NHE. It is slightly higher than the reported potential for the redox benchmark, but still comparable to it and other similar systems. 13,25 In contrast, the dye-sensitized electrode evidenced only the beginning process of the dye oxidation, in agreement with the CV prole of the corresponding N-alkylcarbazole derivative 10 not carrying the terminal pyridine functionality (Fig. S6, ESI †). The electrochemical study of the dyad-sensitized electrodes thus supports the presence of a [Ru(bda)(pic)(ligand)] complex anchored, through the dye, to the SC. X-ray photoelectron spectroscopy (XPS) analysis was performed to conrm the formation of the anchored dyads. The XPS survey spectra (Fig. S7, ESI †) show the presence of sulphur and nitrogen peaks on the TiO 2 surface, thus supporting the presence of the dye portion. Ruthenium peaks are not clearly resolved in the survey spectra due to the proximity to the carbon peak. The core-electron binding energy of C 1s and Ru 3d 5/2 are plotted in Fig Meanwhile, Ru 3d 3/2 peaks are hidden by the C 1s peak. 16,41 The presence of the Ru complex is further supported by the N 1s peak shi to higher binding energy (core-level shown in Fig. S8, ESI †), which is more pronounced in the CBZ-3Py + Ru dyad (0.3 eV). Finally, the valence-band XPS (Fig. S9, ESI †) contains the overlapping O 2p and Ru 4d peaks. The intensity increase in the Ru 4d region (<3.5 eV) further conrms the presence of the coordination complex.

Photoelectrochemical properties
The ability of the dyad-sensitized PECs to perform the photoinduced water splitting reaction was measured through photoelectrochemical measurements under white light illumination (200 W Xe lamp, 420 < l < 800 nm). The bias applied was chosen according to the linear sweep voltammogram (LSV) of the sensitized TiO 2 lms (Fig. S10 †) by selecting the potential at which the best light/dark ratio was recorded. The obtained photocurrents are shown in Fig. S11 and S12 (ESI), † and Fig. 4 (inset) for CBZ-4Py + Ru and CBZ-3Py + Ru sensitized electrodes, respectively. Each dyad-sensitized electrode has been checked against dye-sensitized electrodes (without WOC). In all cases, the dye-sensitized electrode was much less efficient compared to the dyad-sensitized electrode ( Fig. S11 and S12, ESI, † and inset of Fig. 4). The CBZ-3Py dye-sensitized electrode had a negligible photocurrent of 3.5 mA cm À2 , while the corresponding dyad CBZ-3Py + Ru, aer a 10 min chronoamperometry, showed 25 mA cm À2 , corresponding to an 8fold increase. In the case of CBZ-4Py (8.6 mA cm À2 ) and the corresponding dyad CBZ-4Py + Ru (30 mA cm À2 ), a performance increase for the latter was observed, but was less pronounced (Fig. S12, ESI †). These values well match the behaviour of other organic sensitizers reported in the literature. 16,42,43 Long-term measurements were performed to evaluate the stability of the dyad sensitized electrodes and are shown in ESI ( Fig. S13 and S14 †).
The O 2 evolution was measured through the collectorgenerator technique, 38 as recently reported by many studies on DS-PEC water splitting. 8,12,14,32,44 In this technique, a second FTO working electrode (the collector) is positioned at a xed distance (ca. 400 mm) from the photoanode (the generator) and kept at a reducing potential (À0.8 V vs. Ag/AgCl, $À0.6 V vs. NHE). In this way, the collector reduces the oxygen gas as soon as it is formed, thus transforming a very-difficult-to-evaluate variable (affected, among other things, by leakage and air contamination) into a simple electric signal. The direct quantication of oxygen production FE is possible from the ratio between the charge owing through the collector and the one through the generator. The FE for CBZ-4Py + Ru and CBZ-3Py + Ru sensitized electrodes were 58 AE 12% and 88 AE 9% (average of 4 measurements on different samples), respectively (Fig. S15 †  and 4).
To gain a better insight on the performances of the dyads to convert photons to electrons the incident photon-to-current efficiency (IPCE) was also measured (Fig. 5). The recorded IPCE well matched the trend of the measured faradaic efficiency. The intensity of the signal, related to the produced current density, was higher for the CBZ-3Py + Ru sensitized electrode, with a maximum efficiency of $16% at 450 nm. The IPCE of the TiO 2 lms sensitized with the dyads were compared with those of the corresponding lms sensitized with the dye only (without WOC, Fig. S16 and S17 †). The light harvesting efficiency (LHE) of both dyads sensitized electrodes were also measured to calculate the absorbed photon-to-current efficiency (APCE), shown in Fig. S18 and S19 (ESI). † It is worth noting, as an addition evidence of the formation of the dyads, that the IPCE of the devices sensitized with the dye by itself (e.g., without the integrated catalyst) showed a negligible efficiency, in agreement with the hereabove described photocurrent experiments (see Fig. S11 and S12 in ESI † and inset of Fig. 4).
The notably higher efficiency of the 3Py derivative, compared with the 4Py analogue, can be ascribed to the fact that in the former dyad the rotation about the single bond (between the methylene bridge and the quaternary carbon atom of pyridine) allows the catalyst portion to come closer to the dye. Therefore, a more efficient charge transfer between the dye and the catalyst fragment of the dyad is ensured (see the simulated movie in ESI †).

Conclusions
In conclusion, a new series of metal-free organic dyes, bearing a coordinating pyridine functionality, has been synthesized, fully characterised and used to sensitize photoanodes for photoelectrochemical water oxidation. These sensitizers have been used to coordinate a proper Ru precursor to yield the rst example ever of a molecular dyad based on a metal-free organic dye integrated to a benchmark Ru complex for water oxidation. Optical, electrochemical, and XPS experiments conrmed the formation of the dyad and, in particular, the presence of the Ru catalyst anchored to the TiO 2 photoanode via the dye linker. Both dyad-sensitized photoanodes showed state-of-the-art FE in the O 2 generation process, 8,[45][46][47] with the CBZ-3Py + Ru dyad reaching an average efficiency of nearly 90% and a best cell efficiency of 95%. Moreover, IPCE measurements conrmed the validity of the approach showing efficiencies among the best recorded for similar compounds and noticeably supported the evidence of the formation of dyads when compared to isolated dyes. Further studies, including a detailed computational investigation, are currently underway to further support and extend the potential of the herein introduced molecular dyad approach to water splitting.

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