Chemical approaches to artificial photosynthesis: A molecular, dye-sensitized photoanode for O₂ production prepared by layer-by-layer self-assembly

In artificial photosynthesis, the direct bandgap approach, pioneered by Fujishima and Honda, with UV-driven water oxidation at a single electrode and H₂ generated at a Pt cathode, was remarkable for the simplicity of its design.^1–4 In designing visible light photoelectrodes for water oxidation or CO₂ reduction, the use of modified chemical approaches offers significant advantages by the use of modular approaches with individual components investigated separately, maximized in performance, and integrated in appropriate electrode architectures.^5,6 This approach has provided a basis for the Dye-Sensitized Photoelectrosynthesis Cell (DSPEC), which integrates photoelectrodes for water oxidation and water or CO₂ reduction.^7–14 In a DSPEC photoanode, light absorption occurs at a light-absorbing chromophore with the catalyst bound to the surface of a wide bandgap semiconductor, typically with TiO₂ as the photoanode.^15–20 We introduce here a “layer-by-layer” approach by Atomic Layer Deposition (ALD) for the preparation of DSPEC photoanodes with multiple chromophores and light absorbing units on nano-crystalline metal oxide surfaces. The procedure is based on phosphonate–metal ion ester bridging and the use of a layer-by-layer procedure with the use of ALD for the stepwise preparation of reactive assemblies.¹⁵ It is an extension of an earlier series of experiments in which we investigated the surface electron transfer between electron donors and acceptors in molecular assemblies on electrode surfaces with variations in composition introduced by ALD.^15,21

In the current series of experiments, we have utilized nano-particle indium tin oxide (ITO) films on fluorine-doped tin oxide (FTO) electrodes. Surface film assemblies were prepared by a series of stepwise, bridge-forming reactions with acid–base coupling on the oxide surfaces of electrodes, in this case, on conductive indium tin oxide (ITO) nano-particles on fluorine-doped tin oxide (FTO). The electrodes, nanoITO|–MV²⁺–ALD MO_x–RuP₂²⁺–ALD M′O_x–WOC, with MV²⁺ [N,N′-((CH₂)₃PO₃H₂)₂-4,4′-bipyridinium](PF₆), the metal to ligand charge transfer absorber RuP₂²⁺ ([Ru^II(4,4′-(PO₃H₂)₂-2,2′-bipyridine)₂(2,2′- bipyridine)], with the water oxidation catalyst WOC, ([Ru(bda)(4-O(CH₂)₃P(O₃H₂)₂-pyr)₂] (pyr = pyridine; bda = 2,2′-bipyridine-6,6′-dicarboxylate), and the bridging groups ALD MO_x and ALD M′O_x, derived by ALD addition of TiO₂, ZrO₂, and Al₂O₃, followed by aqueous hydrolysis, were prepared by step-by-step addition with characterization of the assemblies by spectral and electrochemical measurements.

The resulting photoanodes, with a separate Pt cathode for water reduction to H₂, led to solar water splitting (Scheme 1). When comparing assemblies for water oxidation, we have systematically varied the ALD bridging groups that link MV²⁺ and RuP₂²⁺ and the water oxidation catalyst WOC. From the results obtained, the electrode FTO|nanoITO|–MV²⁺–ALD TiO₂–RuP₂²⁺–ALD Al₂O₃–WOC was the most efficient one toward water oxidation. Based on photophysical measurements, the efficiencies are limited by a relatively slow electron transfer from the –RuP₂^2+*– excited state to the internal electrode. In the most efficient cell, excitation with a one sun light source (white light >400 nm filter, 100 mW/cm²) led to visible light water splitting with an incident photon to current efficiency (IPCE) of 2.2% at 440 nm.

Nano-particle films (10–20 nm) of ITO on fluorine-doped tin oxide (FTO) substrates (1 cm²) were used as the electrode for the photoanode. The electrode mimics the size, mesoporous morphology, and surface area of nano-crystalline TiO₂ electrodes that are used in dye-sensitized solar cells.²² The high surface areas available in these micrometer-thick, thin film electrodes allow for the binding of sufficiently large amounts of the light absorber for high levels of visible-light absorption with an absorbance of 0.8 for the assemblies shown in Fig. 1 at the metal-to-ligand absorption maximum at 440 nm. High conductivity and visible transparency of nanoITO make it an ideal substrate for investigating the electrochemical, photochemical, and photophysical properties of surface-bound molecules and assemblies.²³ 

The assembly structure shown in Fig. 1 was prepared by stepwise, sequential addition to nanoITO electrodes of (1) the phosphonate-derivatized methyl viologen-based electron acceptor, MV²⁺, followed by (2) ALD formation of a precursor that, upon hydrolysis, forms bridging assembly units with Ti(IV), Zr(IV), or Al(III), (3) the addition of the Ru^II poly-pyridyl based chromophore, RuP₂²⁺, (4) the use of ALD to add a second bridging unit, and (5) the addition of the water oxidation catalyst, WOC. The components were synthesized and characterized by established literature procedures. The multiple phosphonate groups on each molecular unit enable attachment to the oxide surface of the nanoITO electrode with assembly formation by stepwise vacuum deposition of ALD linking groups by procedures described in detail elsewhere.^11,15

In the procedure, mesoporous 2.3 μm thick, nanoparticle nanoITO films of 10–20 nm nanoparticles were deposited on conductive fluorine-doped tin oxide glass with a doctor-blading procedure, as reported previously. In the first step, the electrodes were prepared on the surface of the nanoITO films in 0.1M HClO₄ aqueous solutions that contained 2 mM MV²⁺ for 12 h. After copious rinsing by 0.1M HClO₄, the slides were dried in air under N₂ and placed in a chamber of the ALD for a first cycle of ALD deposition. In forming the ALD bridging units, the organometallic precursors for the three bridging units were tetrakis(dimethylamino)zirconium, tetrakis(dimethylamino)titanium, and trimethylaluminum.

In assembly formation, after the initial ALD step, the slides were immersed in 2 mM RuP₂²⁺ for 12 h, which led to rapid hydrolysis of the bridging precursor and, after hydrolysis, to external bridging units for the formation of the extended assemblies. Following the addition of the chromophore, the derivatized slides were treated by ALD with TiO₂, ZrO₂, or Al₂O₃ to create activated substrates for the formation of a second bridging site for the addition of the catalyst. Following the bridge formation by a second ALD activation cycle, the catalyst was added to the assembly by soaking in methanol solutions containing the WOC at 4 mM in methanol for 24 h to give the assembly. Following the preparation of the final electrode assembly, it was stored in a glove box for further use.

To determine the ratio of the WOC and chromophore in the electrode, samples of nanoITO|–RuP₂²⁺– were fabricated, followed by ALD treatment of different precursors with UV–visible analysis, as shown in Fig. S1. The experimental procedures for the characterization of the assemblies were consistent with evidence for a 1:1 ratio between RuP₂²⁺ and the catalyst, consistent with results obtained in a previous study.¹⁵ There was no evidence for peripheral binding of the hydrated Ti(IV), Zr(IV), or Al(III) sites to the bpy ligand that was not involved in the assembly formation, Fig. 1. Although they may be present as capping groups during the formation of the final assembly, we assume that they undergo solvolysis over the extended period for catalyst addition.

UV–visible measurements were used to monitor the assembly formation, with results as shown in Fig. 2 for nanoITO|–MV²⁺–ALD TiO₂–RuP₂²⁺ and nanoITO|–MV²⁺–ALD ZrO₂–RuP₂²⁺. The spectra show that the formation of the assemblies results in the same MLCT absorption at 450 nm independent of the ALD bridge. Based on the expression, Γ(mol/cm²) = A_λ/(ε_λ1, 000), the extent of chromophore loading was Γ = 6.3 × 10⁻⁸ mol/cm² and Γ = 6 × 10⁻⁸ mol/cm² for nanoITO|–MV²⁺–ALD ZrO₂–RuP₂²⁺ and nanoITO|–MV²⁺–ALD TiO₂–RuP₂²⁺, respectively. The values are comparable to complete surface loading of the dye [Ru(4,4′-bpy) (bpy)₂] on oxide surfaces.^24,25

The final assemblies, FTO|nanoITO|–MV²⁺–ALD MO_x–RuP₂²⁺–ALD M′O_y–WOC, were employed as photoanodes in photoelectrochemical cells for water oxidation. In these experiments, the final assemblies were used as working electrodes with a platinum mesh as the counter electrode with a Ag/AgCl (3M NaCl) reference electrode. The measurements were conducted at an applied bias of 0.3 V vs Ag/AgCl with the potential maximizing the electrode output. Experiments were conducted at a pH of 4.65 in 0.1M HAC/AC⁻ solutions in 0.4M LiClO₄. The light source was a one sun illuminator (100 mW/cm², 400 nm long pass filter).

From the results shown in Fig. 3(a), the assemblies nanoITO|–MV²⁺–ALD TiO₂–RuP₂²⁺–ALD MO_x–WOC for Ti(IV), Zr(IV), and Al(III) exhibited photocurrent densities above 120 μA/cm² at the end of three 10 s light on and off cycles. The photocurrent for nanoITO|–MV²⁺–ALD TiO₂–RuP₂²⁺–ALD TiO₂–WOC was the highest with oxide bridges between units playing a minor role. From Fig. 2(b), for the assemblies, nanoITO|–MV²⁺–ALD ZrO₂–RuP₂²⁺–ALD MO_x–WOC, a similar trend was observed but with a slightly lower photoresponse.

The electron acceptor, MV²⁺, has a profound effect on the performance of the assemblies without the introduction of MV²⁺ in the assemblies, nanoITO|–RuP₂²⁺–ALD MO_x–WOC, as shown in Fig. S2, and there was no evidence for photocurrents following excitation of the chromophore, apparently due to rapid back electron transfer at the surface following excitation and surface quenching:

Transient absorption measurements were also used to explore electron transfer within the assemblies. In Fig. 3(a), the transient-absorption-difference spectra for the assemblies FTO|nanoITO|–MV²⁺–ALD MO₂(IV)–RuP₂²⁺(M = Zr, Ti) obtained at the indicated delay times after 488 nm pulsed-laser excitation at pH = 4.65 in 0.1M HAC/Ac⁻, 0.4M in LiClO₄ at an applied bias of E_app = 0.3 V vs Ag/AgCl are shown. As noted above, under the conditions used in the water oxidation experiments, the actual charge on the assembly at the interface may be decreased by 2.

Based on the results shown in Fig. 4, nanosecond TA spectra for the two assemblies are consistent with excitation and rapid quenching for both assemblies, as shown by the appearance of the tailing negative absorption feature near 375 nm, assigned to nanoITO(e⁻)^22,26,27, and the ground state bleach at 450 nm, assigned to RuP₂³⁺. The bleach at 650 nm is due to emission from the unquenched RuP₂^2+* excited state. The appearance of RuP₂³⁺ for both assemblies is consistent with photoexcitation followed by oxidative quenching of the RuP₂^2+* excited state, Eq. (2), with electron transfer through MV²⁺ as a redox bridge:

In comparing the transient decay data in Fig. 4(c), the extent of intra-assembly quenching is enhanced for ALD TiO₂ as a bridging unit compared to ALD ZrO₂. In the ALD ZrO₂ case, the excited-state persists much longer than ALD TiO₂, as shown by the emission bleach at 650 nm and the absorption growth at 350 nm, both indicative of RuP₂^2+*. This result indicates that ALD TiO₂ provides a more favorable pathway for electron transfer from RuP₂^2+* to MV²⁺ through the use of accessible electronic states. Importantly, by 5 µs after excitation, both ALD linkages indicate the formation of nanoITO(e⁻) and RuP₂³⁺, with smaller magnitudes observed for ALD ZrO₂, consistent with a lower extent of intra-assembly quenching. Overall, these results support our water oxidation experiments, which show a lower photocurrent for ALD ZrO₂ linkages.

Incident photon conversion efficiency (IPCE) was evaluated for the assembly FTO|nanoITO|–MV²⁺–ALD TiO₂–RuP₂²⁺–ALD Al₂O₃–WOC, with the results shown in Fig. 5(a). Based on the IPCE profiles, the excitation overlaps the MLCT absorption profile for the chromophore as expected. Based on these results, an IPCE value for the assembly, nanoITO|–MV²⁺–ALD TiO₂–RuP₂²⁺–ALD Al₂O₃ –WOC, reached 2.2% at the MLCT maximum at 440 nm with the spectral profile consistent with light absorption dominated by the MLCT chromophore. The value obtained was considerably lower than the values for SnO₂/TiO₂ core/shell based photoanodes with values around 1 mA/cm²–2 mA/cm².²⁸ The lower performance for nanoITO|–MV²⁺–ALD TiO₂–RuP₂²⁺–ALD Al₂O₃–WOC, given the available results, might be due to the relatively rapid electron–hole recombination.

To test the stability and efficiency of the electrodes, toward O₂ generation, collector–generator (C–G) experiments were carried out by using an apparatus described elsewhere.²⁹ In these experiments, O₂ was generated at a collector electrode and monitored at a fluorine-doped tin oxide (FTO) electrode, separated by 1 mm. Short 10 min illumination periods, with a low intensity light source (100 mW cm⁻² and 400 nm filter), resulted in the photocurrent response, as shown in Fig. 5(b). At the end of a photolysis cycle, the generator current decays instantaneously with the collector current analyzed for O₂ production. The Faradaic efficiency (FE) was evaluated by Eq. (3) with Q_Collector and Q_Generator, the total charge passed at the collector and generator electrodes, respectively. The constant, 0.7, is the collection efficiency of the collector electrode,

During the 10 min illumination cycles, a decrease in photocurrent density occurs, presumably due to the loss of the assembly from the electrode. At the end of a 10 min photolysis period, the assembly, nanoITO|–MV²⁺–ALD TiO₂–RuP₂²⁺–ALD Al₂O₃–WOC, gave a photocurrent density of 45 μA/cm² after a 10 min water oxidation period with a Faradaic efficiency of 71% obtained by integrating the data in Fig. S3. This value is lower by a factor of more than 10 when compared to benchmark DSPEC photoanodes and core/shell SnO₂/TiO₂ electrodes in conventional DSPEC cells.^28,30 However, in the current cell configuration, the decrease in cell efficiency arises from relatively slow oxidative quenching by the –MV²⁺– redox spacer, and, with a redesign, it should be possible to significantly enhance the cell output.

The results described here are an important extension of earlier results in which utilization of the molecular spacer/ALD approach led to the preparation of molecular assemblies. The results shown here are a demonstration that this approach can lead to molecular dye-sensitized photosynthesis cells for water oxidation. The strategy used here was based on a modular layer-by-layer method with bridging structures prepared by an ALD approach in which the direct assembly of electron quenchers, light absorbers, and catalysts is accomplished at electrode surfaces. Although the results described here, with an IPCE value of 2.2% at 440 nm with an O₂ production Faradaic efficiency of 71%, are unimpressive, the overall approach has potentially considerable value in the design of new families of efficient assemblies for water splitting or CO₂ reduction.

See the supplementary material for details of the experimental section.

The data that support the findings of this study are available within the article (and its supplementary material).

This work was supported by the U.S. Department of Energy (DOE), Nuclear Energy University Program award, under Contract No. DE-NE0008539. This work also made use of IPCE and nanosecond transient absorption instruments in the AMPED EFRC Instrumentation Facility established by the Alliance for Molecular PhotoElectrode Design for Solar Fuels, an Energy Frontier Research Center (EFRC) funded by the U.S. Department of Energy, Office of Science, and Office of Basic Energy Sciences under Award No. DE-SC0001011. This work made use of the instrumentation at the Chapel Hill Analytical and Nanofabrication Laboratory (CHANL), a member of the North Carolina Research Triangle Nanotechnology Network (RTNN), which is supported by the National Science Foundation (Grant No. ECCS-1542015) as part of the National Nanotechnology Coordinated Infrastructure (NNCI).

The authors declare no conflicts of interest.