Light-driven water oxidation by a dye-sensitized photoanode with a chromophore/catalyst assembly on a mesoporous double-shell electrode

Dye-sensitized photoelectrochemical cells (DSPECs) provide a strategy for solar energy conversion by integrating light absorption, photoelectron conversion and electron transfer to drive electrochemical reactions including water splitting.^1–12 Charge separation is a critical factor in these DSPEC devices for electron transfer to carry out electrochemical reactions. Multiple strategies have been developed to slow the rates of back electron transfer (BET) in these dye-sensitized systems in order to enhance charge separation and photoelectrochemical efficiencies.^13–15 Construction of core/shell oxide, layer-by-layer structures, by atomic layer deposition (ALD) has led to decreased charge recombination (CR) (back electron transfer) and substantially improved photocurrent densities due to proper band alignment and interfacial barriers.^16–24 

A typical photoanode in a DSPEC device consists of a mesoporous nano-TiO₂ or SnO₂ semiconductor serving as a high-surface-area film for efficiently loading molecular chromophores with catalysts to create an extermal charge-transfer layer.^25–32 As a way to enhance water-splitting efficiencies in DSPEC devices, atomic layer deposition (ALD) has been used to form nano ITO(tin-doped indium oxide)/ALD TiO₂ or SnO₂/ALD TiO₂ core/shell-based photoanodes on fluorine-doped tin oxide (FTO) substrate.³³ The latter configuration results in greatly improved photoelectron conversion efficiencies by decreasing BET rates.^34–36 However, the overall efficiencies in these devices are still low due to rapid recombination by back electron transfer. Overcoming the problem of large BET rates is a major challenge in maximizing water-splitting performance.

In this study, atomic layer deposition (ALD) has been used to prepare derivatized electrodes on the surface of mesoporous Al₂O₃ films on FTO to give the electrodes FTO//Al₂O₃//ALD ZnO$|$ALD TiO₂ with the ZnO$|$TiO₂ double shell added by atomic layer deposition. The final electrodes were surface-modified by the addition of the molecular ruthenium(II) polypyridyl chromophore, RuP²⁺ [Ru(4,4′-PO₃H₂-bpy)(bpy)₂]²⁺: bpy = 2,2′-bipyridine, Fig. S1(a), and the Ru-based water oxidation catalyst C1 according to the Sun group [Ru(bda)L₂: bda = 2,2′-bipyridine-6,6′-dicarboxylate, L = 4-O(CH₂)₃PO₃H₂-pyr, pyr = pyridine, Fig. S1(b)].^37–44 The resulting photoanode structure, FTO//Al₂O₃//ALD ZnO$|$ALD TiO₂$|$–(RuP²⁺)₂–C1, was further explored to include an optimized tunneling barrier with a built-in electric field at the ALD ZnO$|$ALD TiO₂ interface to decrease BET and enhance photoinduced charge transfer. The mesoporous Al₂O₃ film is insulated with the external ZnO$|$TiO₂ film providing a mesoporous double-shell electrode surface to the conducting FTO substrate. As shown in Fig. 1, the combined layered structure, FTO//Al₂O₃//ALD ZnO$|$ALD TiO₂, provides the basis for improving the performance of a dye-sensitized photoanode toward light-driven water oxidation. This work used a rare assembly of a two-layer ALD layer system on an insulating core as a charge collection electrode, which resulted in greatly long recombination lifetime.

Nanosecond transient absorption (TA) measurements were used to investigate charge recombination dynamics to explore the role of the structure within the photoanode. The available results show that optimizing the double-shell structure by ALD results in remarkable enhancements in charge recombination lifetime in the mesoporous double-shell based photoanodes relative to the mesoporous single-shell based photoanodes. This would demonstrate the decreased BET arising from interfacial built-in electric fields and the tunneling barrier. Based on the photophysical results, photocurrent densities for light-driven water oxidation, 2H₂O → O₂ + 4H⁺ + 4e⁻, were also substantially enhanced in the mesoporous double-shell (FTO//Al₂O₃//ALD ZnO$|$ALD TiO₂) based dye-sensitized photoanode, compared with that in the mesoporous single-shell (FTO//Al₂O₃//ALD ZnO, FTO//Al₂O₃//ALD TiO₂) based photoanodes, with O₂ detected with a generator–collector dual working electrode.

Figure 1(b) shows scanning electron microscopy (SEM) images of a mesoporous Al₂O₃//ALD ZnO$|$ALD TiO₂ film on FTO substrates with a film thickness of ∼4.5 µm. The electrode was prepared from an ∼4.5 µm mesoporous Al₂O₃ film on a FTO substrate with a typical doctor-blading method followed by ALD deposition of the ZnO and TiO₂ shells. Figures S2(a) and S2(b) display TEM images of a Al₂O₃//ALD ZnO$|$ALD TiO₂ core–double shell structure with a total shell thickness of 6.2 nm and a Al₂O₃//ALD ZnO core–shell structure with a shell thickness of 3.3 nm.

The effect of shell thickness on electrode performance for light-driven water oxidation in DSPEC devices was also investigated. The comparison showed that, FTO//Al₂O₃//ZnO(∼3 nm)$|$TiO₂(∼3 nm), the mesoporous double-shell electrode gave the highest photocurrent densities compared with single-layered structure and double-layered one with other thicknesses for light-driven water oxidation, and it was this structure that dominated subsequent studies. The working electrode is an equivalent of the FTO/ALD ZnO$|$ALD TiO₂ double-shell electrode, considering about the insulating Al₂O₃ film that serves as a surface for dye and catalyst loading in subsequent studies. The effect of shell thickness on photocurrent densities under DSPEC experimental conditions for light-driven water oxidation is shown in Fig. 2. For creating the surface assemblies, the molecular dye RuP²⁺ was co-loaded with the molecular water oxidation catalyst C1 at a molar ratio of 2:1 as previously described.^37,38,45 The samples were excited by 1 sun illumination with light on-light off cycles in pH 4.6, 0.1M acetate buffer with 0.4M NaClO₄ supporting electrolyte at an applied bias of 0.25 V vs. Ag/AgCl as the reference electrode. Effect of film thickness of the ZnO shell (0, 3, and 6 nm) and the TiO₂ shell (0, 1.5, 3, 4.5, 6, and 7.5 nm) was explored with the maximum photocurrent density obtained in the configuration FTO//Al₂O₃(core)//ZnO(∼3 nm)$|$TiO₂(∼3 nm)$|$–(RuP²⁺)₂–C1.

The photocurrent densities obtained for the developed double−shell structures prepared by ALD are comparable to those obtained previously based on nanoparticle oxide core−ALD shell structures.^34,38 The ALD double−shell electrodes result in a greatly enhanced photocurrent response compared to single−shell electrodes prepared by ALD. In ALD double−shell electrodes, there are optimized tunneling barrier shells with the built−in electric field at the ALD ZnO$|$ALD TiO₂ interface that inhibit back electron transfer, enhancing charge recombination lifetime. With increasing shell thicknesses, the photocurrent densities are layer thickness dependent, indicating a tunneling mechanism in the process. Photocurrent densities first increase because of a decrease in the BET rate by increasing the tunneling barrier,^17,38,46 and then decrease due to diffusion-limited BET from electrons localized in trap states in the shells⁴⁶ and from reduced loading of the molecular dye due to decreased surface areas with smaller pore size.⁴⁷ 

Figure S3 shows a UV–visible absorption spectrum of RuP²⁺ on different mesoporous shells to estimate the surface coverage of RuP²⁺. The surface coverage (Γ) of RuP²⁺ for FTO//Al₂O₃//ZnO(∼3 nm)$|$TiO₂(∼3 nm)$|$–(RuP²⁺)₂–C1 was determined by subtracting the background spectra [Fig. S4(b)],^47,48 which gave Γ(RuP²⁺) 6 × 10⁻⁸ mol/cm².

The Faradic efficiency for O₂ formation in FTO//Al₂O₃//ZnO(3 nm)$|$TiO₂(3 nm)$|$–RuP²⁺–C1, was determined by a previously reported generator–collector dual working electrode method,^38,49 as shown in Fig. 3(a). In these experiments, the samples were illuminated for 10 min with the input of a one sun lamp. Photoelectrochemical O₂ evolution at the photoanode generator was monitored using a parallel fluorine-doped tin oxide (FTO) collector cathode with the two separated by 1 mm, with a pre-established collection efficiency of 70%.^38,49 The Faradic efficiency (η) for O₂ generation was evaluated from current density (j)−time plots by using Eq. (1). In Eq. (1), Q_collector refers to the total charge that passed at the collector electrode; Q_generator is the total charge passed at the generator electrode. The Faradic efficiency (η) for FTO//Al₂O₃//ZnO(3 nm)$|$TiO₂(3 nm)$|$–(RuP²⁺)₂–C1 was 75.5%

Ultraviolet Photoelectron Spectra (UPS) and Kubelka-Munk function transformed spectra from absorption measurements on FTO//Al₂O₃//ZnO(3 nm) and FTO//Al₂O₃//TiO₂(3 nm) films are shown in Fig. S4 with the energy band data shown in Table S1. From the comparison, it can be seen that the Fermi level for the ALD TiO₂ layer (−0.09 V vs. NHE (Normal Hydrogen Electrode)) is lower than that for the ALD ZnO layer (−0.34 V vs. NHE), due to which the built-in electric field can form at the ALD ZnO$|$ALD TiO₂ interface (Fig. S5). Based on these results, there is an internal driving force for electron transfer from the TiO₂ layer to the ZnO layer, which creates an internal barrier to back electron transfer through the shells to increase recombination lifetimes and enhance photoanode performance for light-driven water oxidation. Given the difference in conduction band potentials between the TiO₂ shell (−1.08 V) and the MLCT (Metal-to-Ligand Charge-Transfer) excited state RuP^3+/2+* (−0.81 V), photoinduced electrons from exited states RuP^3+/2+* are favored energetically tunneling through the 3 nm TiO₂ layer, creating a strong barrier to back electron transfer, consistent with earlier results on core-shell electrodes.⁴⁶ Following electron excitation and transfer through the TiO₂ shell, electrons will subsequently be transferred through the 3 nm ZnO layer to the conducting FTO substrate [Fig. 1(a)] favored energetically by the interfacial electric field.

The data obtained by analysis of the electronic spectra for the ZnO(6 nm) and TiO₂(6 nm) films (Fig. S6 and Table S2) show that electronic energy levels of metal oxide thin layers vary with thicknesses, which further confirms that photoinduced electrons from RuP^3+/2+* are injected through the ALD thin layer by electron tunneling with barriers that vary with thickness.

Nanosecond transient absorption (TA) experiments were carried out to analyze the details of back electron transfer from the electrodes to RuP³⁺ following electron excitation and injection by RuP²⁺ with a 425 nm laser in pH 4.6, 0.1M acetate buffers with 0.4M NaClO₄ supporting electrolyte. As shown in Fig. 3(b), normalized transient absorption (TA) difference spectra (ΔOD)–time decay traces were obtained due to BET and analyzed by application of the Kolrausch–Williams–Watts (KWW) stretched exponential function [Eq. (2)]. The bi-exponential KWW fitting model for back electron transfer describes absorbance-time changes as the sum of two exponential decay functions

with a characteristic distribution lifetime, τ, and distribution width, B. A₁ and A₂ are the initial absorbances at time 0 for the two amplitudes in the exponential sum in Eq. (2).^2,5,50–53 The average charge recombination (CR) lifetime, t_CR, was calculated from Eq. (3) with $Γ$⁠, the gamma function $Γx=∫0∝tx−1e−tdt$⁠, and τ and B derived from KWW fit to the kinetic traces.^54,55 The obtained fitting parameters for the decay traces are shown in Table I,

Kinetic absorbance decay traces for back electron transfer to RuP³⁺ in the assemblies are shown in Eqs. (4)–(7): Eqs. (4) and (5) for assembly 1, FTO//Al₂O₃//ALD ZnO$|$ALD TiO₂$|$–RuP²⁺; Eq. (6) for assembly 2, FTO//Al₂O₃//ALD TiO₂$|$–RuP²⁺; and Eq. (7) for assembly 3, FTO//Al₂O₃//ALD ZnO$|$–RuP²⁺.

The kinetic fitting results show that the kinetics of back electron transfer for all three electrodes in Eqs. (4)–(7) is complex, following KWW distribution kinetics. The complexity in the kinetics is a characteristic feature of the heterogeneous, thin-film structure with short term components arising from recombination between surface-trapped electrons and holes at RuP³⁺, t_CR1, and long term components from recombination between electrons hopping through a distribution of trap states within the reduced metal oxide and holes at RuP³⁺, t_CR2.^56–59 For the assembly FTO//Al₂O₃//ALD ZnO$|$ALD TiO₂$|$–RuP²⁺, the contribution to the long-term component for charge recombination arises from slow back electron transfer from reduced ZnO [ZnO (e^–)] through the shells to RuP³⁺, and the relatively short-time component results from rapid BET from reduced TiO₂ [TiO₂ (e^–)] to RuP³⁺. For the assembly FTO//Al₂O₃//ALD TiO₂$|$–RuP²⁺ and FTO//Al₂O₃//ALD ZnO$|$–RuP²⁺, the long term component is due to slow BET from electrons trapped in a distribution of localized states within the oxide layer, and the short term component is from rapid recombination between surface-trapped electrons and holes at RuP³⁺. Further work still needs to be done to illustrate the recombination phases in kinetic absorbance decay traces of back electron transfer to RuP³⁺ in the assemblies.

Based on the recombination lifetime t_CR in Table I with 2 ms for t_CR1 and 51 ms for t_CR2 for the optimized assembly in FTO//Al₂O₃//ZnO(3 nm)$|$ALD TiO₂(3 nm)$|$–RuP²⁺, higher photocurrent densities are obtained for light-driven water oxidation than the other two. The origin of the effect arises from an increase in recombination lifetime due to the significant optimized barrier with a built-in electric field and the extended separation distance for back electron transfer. In addition, the difference between ALD TiO₂ and ALD ZnO/TiO₂ at time scales longer than 2 ms where the recombination seems to be faster for the bilayer system compared to the single layered one could be due to the surfaces’ charge accumulations which may screen the built-in electric field effect.

The electrode, FTO//Al₂O₃//ALD ZnO$|$ALD TiO₂, with a mesoporous double-shell structure on an FTO conducting substrate, was developed for chromophore-catalyst assembly for light-driven water oxidation. Photocurrent density for photoanode assembly, FTO//Al₂O₃//ZnO$|$TiO₂$|$–(RuP²⁺)₂–C1, were largely enhanced comparable with that for mesoporous single-shell photoanodes, revealing a Faradaic efficiency of 75.5% for O₂ evolution. The resulting assembly is important in decreasing charge recombination time. Nanosecond time-dependent transient absorption (TA) measurements show that the mesoporous ALD double-shell structure results in greatly decreased back electron transfer based on an optimized electron tunneling barrier with the interfacial built-in electric field.

Supplementary material includes the detailed experimental section, TEM pictures, and XPS information.

The research on mesoporous double-shell based photoanode for light-driven water oxidation was supported by the University of North Carolina Energy Frontier Research Center (UNC EFRC): Center for Solar Fuels, an Energy Frontier Research Center supported by the U.S. Department of Energy (DOE), Office of Science, Office of Basic Energy Sciences under Award No. DE-SC0001011. The transmission electron microscopy (TEM), scanning electron microscopy (SEM), and ultraviolet photoelectron spectroscopy (UPS) experiments were performed at the Chapel Hill Analytical and Nanofabrication Laboratory, CHANL, a member of the North Carolina Research Triangle Nanotechnology Network, RTNN, supported by the National Science Foundation, Grant No. ECCS-1542015, as part of the National Nanotechnology Coordinated Infrastructure, NNCI.