Optimization of Photoanodes for Photocatalytic Water Oxidation by Combining a Heterogenized Iridium Water-Oxidation Catalyst with a High-Potential Porphyrin Photosensitizer

Graphical Abstract

Slow and steady wins the race: Photoanodes containing a Cp*Ir (Cp*=pentamethylcyclopentadienyl) water-oxidation catalyst and a porphyrin photosensitizer are assembled. Silatrane and tetrahydropyranyl-protected hydroxamic acid anchors are used to bind the molecules to the SnO₂ electrode. The photosensitizer/catalyst ratio is varied to optimize the electrodes. Photocatalytic water oxidation is observed over 20 h with a Clark-type electrode.

Abstract

The development of water-splitting dye-sensitized photoelectrochemical cells has gained interest owing to their ability to generate renewable fuels from solar energy. In this study, photoanodes were assembled from a SnO₂ film sensitized with a combination of a high-potential CF₃-substituted porphyrin dye with a tetrahydropyranyl-protected hydroxamic acid surface-anchoring group and a Cp*Ir (Cp*=pentamethylcyclopentadienyl) water-oxidation catalyst containing a silatrane anchoring group. The dye/catalyst ratios were varied from 2:1 to 32:1 to optimize the photocatalytic water oxidation. Photoelectrochemical measurements showed not only more stable and reproducible photocurrents for lower dye/catalyst ratios but also improved photostability. O₂ production was confirmed in real time over a 20 h period with a Clark electrode. Photoanodes prepared from 2:1 and 8:1 dye/catalyst sensitization solutions provided the most active electrodes for photocatalytic water oxidation and performed approximately 30–35 turnovers in 20 h.

Introduction

The advance of climate change has resulted in a demand for renewable fuel generation from alternative energy sources.1 Solar energy is a promising source, as the large solar flux reaching the earth's surface (1.2×10⁵ TW) surpasses annual global energy demands by nearly four orders of magnitude.2 The generation of useable fuels from solar energy can be accomplished with water-splitting dye-sensitized photoelectrochemical cells (WS-DSPECs), which are inspired by the processes that occur in natural photosynthesis, in which photocatalytic water oxidation occurs.2b, 2d, 3 WS-DSPECs operate through the coupling of light-driven water oxidation at the photoanode with reductive fuel-forming reactions at the cathode. To perform the required photocatalytic water oxidation, light-harvesting dye molecules and water-oxidation catalysts are attached to the surface of a semiconductor electrode. Upon photoexcitation, the dye molecule injects an electron into the conduction band of the semiconductor, and the oxidized dye molecule is left on the surface. This oxidized dye molecule can either oxidize a neighboring dye molecule through lateral hole transfer or oxidize a neighboring catalyst molecule.4 Once the catalyst gains four oxidizing equivalents sequentially, water oxidation can occur to form O₂, protons, and electrons. The protons and electrons can then be used at the cathode to form fuels such as H₂.5

The design of operational WS-DSPECs is challenging because all of the water-splitting processes must occur more quickly than charge recombination. Previous strategies to slow charge recombination include the application of atomic layer deposition (ALD),5b, 6 the adoption of core–shell assemblies,7 the alteration of the linker length,8 and the variation of the surface-anchoring group.9 Additionally, the molecules must be stable under photoelectrochemical conditions to avoid deactivation or surface desorption. Common methods to provide surface stability include the use of ALD,5b, 6b–6d, 10 hydrophobic polymer overlayers,11 electroassembly,12 and stable surface-anchoring groups.2b, 9, 13

Of these strategies, the alteration of the surface-anchoring group for a dye or catalyst molecule has several advantages. First, the anchor can be used to favor electron injection from the dye molecule, for example, carboxylic acid and hydroxamic acid anchoring groups mediate facile electron injection into metal oxide surfaces; therefore, they are suitable anchors for dye molecules.9, 13a, 14 Secondly, the anchor can provide surface stability under aqueous conditions to prevent desorption. Excellent anchors for this purpose are hydroxamic acid and silatrane groups, which form strong surface bonds under aqueous conditions and are stable under acidic, neutral, and basic conditions.9, 13b, 15 Carboxylic acids and phosphonic acids are only stable at pH<4 and pH<7, respectively; therefore, they are less robust for WS-DSPECs.9, 16 Furthermore, some anchors, such as silatrane groups, can decouple the catalyst molecule from the semiconductor electronically and, thus, decrease the chances of back electron transfer to the catalyst.6c, 13b, 13c, 15 In addition, silatranes simplify the introduction of the catalyst, because they do not chelate to the metal center and unwanted side reactions are avoided.13b, 13c Hydroxamic acids can also be protected with a tetrahydropyranyl (THP) group, and this strategy decreases the number of synthetic steps for the anchor because the molecule deprotects readily upon surface binding.17 Of the anchors, silatranes and hydroxamic acids are the most robust and have the best properties for use in WS-DSPECs.

We have previously incorporated a silatrane anchor into a Cp*Ir catalyst (Ir, Scheme 1), which performs electrochemically driven water-oxidation catalysis.13c The catalyst operates at 1.35 V versus the normal hydrogen electrode (NHE) at pH 5.8 on nanoporous indium tin oxide (nanoITO) conductive electrodes and is postulated to remain monomeric and retain the Cp* ligand during catalysis.13c The catalyst has also been assembled in photoanodes with perylene-3,4-dicarboimide (PMI) dyes protected through the ALD of Al₂O₃. The observation of higher photocurrents for photoanodes with the catalyst than for the dye-only photoanodes suggests that successful catalyst oxidation occurs; this was also supported by femtosecond transient absorption spectroscopic measurements.6c Unfortunately, oxygen detection was not confirmed,6c as was also the case for other photoanodes with Cp*Ir-type water-oxidation catalysts.18 Thus, improvements to the design of photoanodes containing these catalysts are necessary to permit O₂ detection.

Scheme 1

Proposed surface binding of P and Ir to SnO₂ electrodes. The THP-protected hydroxamic acid in P and the silatrane anchoring group in Ir deprotect upon binding to the SnO₂ electrode.

Furthermore, we recently designed new, high-potential CF₃-substituted porphyrin dye molecules.19 These porphyrins are preferred over pentafluorophenyl-substituted porphyrins because they avoid potential nucleophilic attack at the para C−F of the C₆F₅ group.19, 20 They are also electrochemically wellbehaved and show reversible features for the formation of the radical cation and dication species. These porphyrins are appropriate dye molecules for WS-DSPECs because they can inject electrons into the conduction bands of semiconductor electrodes and their high reduction potentials can result in the oxidization of catalyst molecules.19 For example, the porphyrin core of P (Scheme 1) has an excited-state potential (S₁) of −0.39 V versus NHE and a potential of 1.53 V versus NHE for the formation of the radical cation species, and these properties allow it to inject electrons into the SnO₂ conduction band upon photoexcitation and also oxidize Ir.19

Using silatrane and THP-protected hydroxamic acid anchoring groups, we have designed active photoanodes for photocatalytic water oxidation with the molecular iridium wateroxidation catalyst Ir and the high-potential porphyrin P (Scheme 1). Herein, we report the assembly, characterization, and photocatalytic activities of these photoanodes and are able to confirm light-driven O₂ evolution with a Cp*Ir wateroxidation catalyst for the first time.

Results and Discussion

The photoanodes were assembled with catalyst Ir and porphyrin P (Scheme 1). Catalyst Ir contains a silatrane anchoring group, whereas P contains a THP-protected hydroxamic acid anchoring group; both readily deprotect upon surface binding.13c, 17 As mentioned above, the anchoring groups ensure water stability upon binding to metal oxide electrodes (pH 2–11) and promote relevant electronic coupling for each molecule (facile injection for P and poor electronic coupling for Ir). SnO₂ is preferred for the reasons described previously. Codeposition successfully bound both Ir and P to SnO₂ if the electrode sensitization solution contained P and Ir in a suitable ratio. A toluene solution of the mixture was heated to 60 °C for approximately 24 h to promote anchoring-group deprotection and surface binding (Scheme 1); the solution was also kept under nitrogen and in the dark to limit undesired reactivity. The P/Ir ratio in the sensitization solution was varied to obtain different concentrations of P and Ir on the SnO₂ electrode. Further details on the sample preparation can be found in the Supporting Information. Samples prepared from 2:1, 8:1, and 32:1 P/Ir sensitization solutions, labeled as SnO₂-(2:1), SnO₂-(8:1), and SnO₂-(32:1), were studied for photocatalytic water oxidation.

The photoanodes were characterized by several spectroscopic methods, including UV/Vis spectroscopy, attenuated total reflectance Fourier transform infrared (ATR-FTIR) spectroscopy, and X-ray photoelectron spectroscopy (XPS). The UV/Vis spectra of Ir and P before surface binding are shown in Figure S3. Porphyrin P exhibits typical porphyrin absorption features, that is, a strongly absorbing Soret band at λ=420 nm and four Q bands at λ=515, 548, 590, and 645 nm. The retention of the absorption bands of P after surface binding indicates that the molecular structure has not changed significantly (Figure S4). As the SnO₂ absorption masks the Ir absorption bands at λ=229 and 280–320 nm (broad),13c the UV/Vis spectrum of SnO₂-Ir looks identical to that of SnO₂ and shows no absorption features for Ir (Figure S5). Not surprisingly, the UV/Vis spectra of SnO₂-(2:1), SnO₂-(8:1), and SnO₂-(32:1) resemble that of SnO₂-P (Figure S4). To characterize the electrodes further, ATR-FTIR spectroscopic measurements of SnO₂-P, P, SnO₂-Ir, Ir, SnO₂-(2:1), SnO₂-(8:1), and SnO₂-(32:1) were performed. The ATR-FTIR spectra of SnO₂-P and free P have good overlap and show C=O stretches at =1670 cm⁻¹, N−H bending at =1521 cm⁻¹, and C=C stretching at =1592 cm⁻¹ (Figure S10). Similarly, the ATR-FTIR spectra of SnO₂-Ir and Ir have overlapping features and show bands for the C−H stretches of the Cp* ligand and the heterocyclic ligand at = 2969, 2923, and 2879 cm⁻¹, C=O stretching at =1668 cm⁻¹, N−H bending at =1521 cm⁻¹, and C=C stretching at =1590 cm⁻¹, as observed previously for Cp*Ir bound to TiO₂ surfaces (Figure S9).13c The ATR-FTIR spectrum of SnO₂-(2:1) has overlapping features from those of both SnO₂-P and SnO₂-Ir (Figure S11); SnO₂-(8:1) and SnO₂-(32:1) have analogous ATR-FTIR spectra to that of SnO₂-(2:1).

To characterize further the natures of the bound Ir and P on the electrodes, XPS was used to analyze their Ir, F, and N contents (Figure S14–16). As only Ir contains iridium and only P contains fluorine, we were able to detect the presence of both molecules on the SnO₂ electrodes by XPS. Iridium was detected on SnO₂-Ir, SnO₂-(2:1), SnO₂-(8:1), and SnO₂-(32:1) through its signature Ir 4f_7/2 and Ir 4f_5/2 peaks, which provided further confirmation that Ir was bound to the electrodes (Figure S14). The F 1s signals detected for SnO₂-P, SnO₂-(2:1), SnO₂-(8:1), and SnO₂-(32:1) are indicative of P surface binding (Figure S15). Additionally, an untreated SnO₂ control sample did not show any Ir or F features.

Cyclic voltammetry (CV) measurements were also performed in 0.1 m KNO₃ at pH 6. The cyclic voltammogram of SnO₂-Ir shows a strong catalytic wave at 1.2 V versus NHE, which we assign to water-oxidation catalysis (Figure S21), whereas SnO₂-P does not show this wave (Figure 1). Furthermore, as the Ir concentration on the surface increases, a higher catalytic current is observed, and this indicates that the surface ratios have varied (Figures 1 and S21). To estimate the quantity of P and Ir on the electrodes, XPS and surface desorption measurements were performed. The F/Ir ratios of the electrodes were determined through XPS analysis (see Supporting Information for details), and the F/Ir ratios enabled the calculation of the P/Ir ratios (Table S1). As silatranes and THP-protected hydroxamic acids have different binding affinities to SnO₂, the P/Ir ratios on the surfaces differed from the ratios in the sensitization solutions. The lower P/Ir ratios on the electrodes compared to those in the sensitization solution indicate that the silatrane anchoring group of Ir has a stronger binding affinity than the THP-protected hydroxamic acid anchor of P. The surface loadings of P and Ir were determined through surface desorption experiments. Porphyrin P was desorbed from the electrodes and quantified by UV/Vis spectroscopy, and a loading of approximately 110 nmol cm⁻² was found for a saturated surface coverage. As expected, the loading of P was lower in the photoanodes with Ir coadsorbed and varied in the range 48–63 nmol cm⁻². The amount of Ir on the surface was determined from the P/Ir ratio from the XPS analysis and the known quantity of P desorbed from the electrode and was found to vary in the range 21–40 nmol cm⁻² (Table S1).

Figure 1

Cyclic voltammograms of SnO₂-P, SnO₂-(2:1), SnO₂-(8:1), and SnO₂-(32:1) in 0.1 m KNO₃ at pH 6.

To study the photoanode response under illumination and at an applied bias, chronoamperometry (CA) was performed in 0.1 m KNO₃ at pH 6. First, the effect of the applied bias was studied for SnO₂-P, SnO₂-(2:1), SnO₂-(8:1), and SnO₂-(32:1) (Figure 2). Biases of 0.2, 0.4, 0.6, and 0.8 V versus NHE were applied to the photoanodes, and the detection of photocurrent upon illumination is indicative of dye photoexcitation and electron injection into the conduction band of SnO₂, as was confirmed previously for derivatives of P.19 As the bias increased, the photocurrent also increased. As the Ir^III/IV redox couple occurs at 0.9 V versus NHE,13c biases higher than 0.8 V were not tested to avoid the electrochemical oxidation of Ir. The maximum photocurrent was achieved with 0.8 V bias versus NHE; therefore, this value was maintained for the other experiments.

Figure 2

Chronoamperograms of SnO₂-P, SnO₂-(2:1), SnO₂-(8:1), and SnO₂-(32:1) in 0.1 m KNO₃ at pH 6 with applied biases of 0.2 (black), 0.4 (green), 0.6 (blue), and 0.8 V (red) vs. NHE. The samples were illuminated (≈100 mW cm⁻²) from 10–20 s. A 420 nm long-pass filter was used to avoid direct band-gap excitation of SnO₂.

To compare the photocurrents produced from the photoanodes and to test for photoelectrochemical reproducibility, CA measurements were performed with a 0.8 V applied bias. The photoanodes were then exposed to alternating dark and light periods of 20 s to test their responses to light over 120 s (Figure S23). The highest photocurrent was observed for SnO₂-P, probably because of the higher dye loading on the electrode. The CA experiment was then immediately repeated a second time to test the reproducibility (Figure 3). The photocurrent reproducibility depended on the P/Ir ratio. The SnO₂-(2:1) and SnO₂-(8:1) photoanodes had excellent reproducibilities and displayed identical photocurrent responses in the repeat CA experiments (Figure 3). However, as the dye loading increases, the reproducibility is degraded. For SnO₂-(32:1), the photocurrent response is not the same for the repeat experiment and is slightly lower than the original value (Figure 3). The largest decrease in photocurrent occurs for SnO₂-P, for which the photocurrent response dropped by approximately 50 % in the repeat CA experiment. We propose that the reproducible, stable photocurrents for SnO₂-(2:1) and SnO₂-(8:1) result from productive photo-oxidation events, that is, fast Ir oxidation from oxidized P leads to photocatalytic water oxidation, whereas SnO₂-P photodegrades under illumination and bias as the lack of productive photo-oxidation events result in a lower current upon CA repetition. This decrease in photocurrent was also observed in a previous study on photoanodes with Ir and perylene photosensitizers on TiO₂ electrodes.6c The SnO₂-(32:1) sample showed a less significant drop in photocurrent, which suggests that some catalyst oxidation may occur, but some photodegradation of P also occurs because the photocurrent is not identical upon CA repetition.

Figure 3

Chronoamperograms at 0.8 V vs. NHE. The samples were illuminated (≈100 mW cm⁻²) from 20–40, 60–80, and 100–120 s. The experiments were repeated (labeled “repeat”).

Additionally, several control CA measurements were performed to rule out current responses from other oxidation events (Figure S24). First, a CA measurement of SnO₂-Ir was collected, and no current response was observed; therefore, Ir oxidation does not occur if P is not present. Bare SnO₂ electrodes also showed no current response under illumination; to avoid direct band-gap excitation, a 420 nm long-pass filter was used during all of our experiments. Silatrane and THP-protected hydroxamic acid anchors can deposit some of the protecting group on the SnO₂ as triethanolamine (TEOA) and tetrahydro-2H-pyran-2-ol (THPOH), respectively. Electrodes sensitized with triethanolamine (SnO₂-TEOA) and tetrahydro-2H-pyran-2-ol (SnO₂-THPOH) were prepared and showed no photocurrent response (Figure S24). Lastly, SnO₂ electrodes exposed only to the sensitization solvent, toluene, were tested to ensure that the photocurrent was not caused by toluene photocatalytic oxidation.21 Again, no photocurrent response was detected (Figure S24).

To further assess the oxidative events occurring on the photoanodes, nanosecond transient absorption spectroscopy (nsTAS) was used. Lateral hole transfer between dye molecules is known to occur in sensitized photoanodes. Recently, we characterized the influence of the solvent and the photosensitizer on the hole/trap ratio.4 In those studies, the absorption of the pump light by the porphyrin photosensitizer resulted in lateral hole transfer across the surface of the porphyrin monolayer. Eventually, the hole transferred to a porphyrin dyad “trap center”, which mimicked a catalytic center. This resulted in a change to the kinetic signature from a transient absorption (owing to the porphyrin) to a bleach (owing the dyad) at the same probe wavelength. The Ir catalyst does not have an oxidized spectral signature within the visible region. Therefore, we were only able to probe the transient species of P and not those of Ir. A spectral map of SnO₂-P, constructed from excitation at λ=515 nm (Figure S26), showed transient absorptions at λ=450, 560, and 610 nm. The results of probing the kinetics at λ_probe=450 nm are shown in Figure 4 (see the Supporting Information for more details).

Figure 4

Kinetics traces determined through nsTAS with excitation at λ=515 nm and probed at λ=450 nm. The best fits (red) of Equation (1) are overlaid for SnO₂-P (black), SnO₂-(32:1) (blue), and SnO₂-(16:1) (green).

We hypothesize that two principal processes are in competition: (1) the recombination of photoexcited electrons back to the oxidized P from the SnO₂ and (2) hole transfer from the photoexcited P to Ir on SnO₂ to return P to its ground state. The transient absorption kinetics were modeled as a triple exponential with an offset [Eq. 1]:

(1)

where A_τ represents the contribution of each exponential process; τ was fixed at 10, 100, and 1000 ns; and ΔmOD accounts for any longtime offset in the system. To avoid overparameterization, we chose to fit the results with a multiexponential function that spans three orders of magnitude with the variation of only the time constant for each. The lifetimes were held at 10, 100, and 1000 ns. No physical meaning is attributed to each lifetime, but this method allows for the accurate determination of the total integrated area underneath the transient trace. From the best-fit parameters, we then determined the total lifetime, τ_total, by calculating the weighted average of the three exponentials. The results are shown in Table 1 and are consistent with our expectations. As the P/Ir ratio decreases, a decrease in the overall τ_total occurs owing to the increased probability of direct hole transfer to a catalytic center. We were unable to observe the kinetics for the SnO₂-(2:1) and SnO₂-(8:1) samples, probably because the signal is below the sensitivity of our nsTAS instrument. Consequently, SnO₂-(16:1) was prepared to show the trend in the decreasing τ_total.

Table 1. Best-fit nsTAS data from Equation (1).

Composition	A10ns [mOD]	A100ns [mOD]	A1000ns [mOD]	ΔmODo [mOD]	τtotal [ns]
SnO2-P	5.3	5.9	3.0	1.06	768.2
SnO2 (32:1)	5.8	4.9	0.6	0.75	307.9
SnO2 (16:1)	4.2	4.2	0.0	0.00	165.3

As the photoelectrochemical CA experiments and nsTAS kinetics suggested that Ir oxidation might occur, O₂ evolution measurements were performed with a Clark electrode to validate this by monitoring for photocatalytic water oxidation that results from Ir oxidation. Clark-electrode O₂ assays are not trivial to run and are time-intensive if an electrochemical bias is applied to the samples under illumination, especially if low quantities of O₂ (<1 μmol) are expected, as is the case here. The Clark electrode is extremely sensitive to stirring and temperature changes; oscillations in the traces can occur from the application of a bias or illumination of the electrochemical cell; therefore, some baseline and trace variability occurs over time. The difficulty of the detection of low quantities of O₂ above this baseline noise level results in the need for time-intensive assays. A custom glass cell with zero headspace was designed and constructed to improve the signal/noise ratio for the measurement of dissolved O₂. Our system was left to equilibrate for 1 h before the assays to minimize the baseline noise level. O₂ evolution was monitored over approximately 20 h with 0.8 V applied bias under approximately 100 mW cm⁻² illumination with a 420 nm long-pass filter to avoid the direct band-gap excitation of SnO₂. Of the photoanodes, SnO₂-(2:1) and SnO₂-(8:1) showed significant O₂ evolution, which is indicative of photocatalytic water oxidation (Figure 5). Owing to noise in the data, the average amount of O₂ produced over the last 0.5 h was used to calculate the turnover numbers. Over 20 h, SnO₂-(2:1) completed approximately four catalytic turnovers with an approximate turnover frequency (TOF) of 0.2 h⁻¹, and SnO₂-(8:1) completed approximately five turnovers with an approximate TOF of 0.3 h⁻¹. A threefold increase of the illumination intensity resulted in a significant increase in evolved O₂ and faster TOFs for both photoanodes (Figure 5); SnO₂-(2:1) performed approximately 30 turnovers with an approximate TOF of 1.4 h⁻¹, whereas SnO₂-(8:1) performed similarly with approximately 35 turnovers and an approximate TOF of 1.7 h⁻¹. To our knowledge, this is the first time that O₂ generation has been confirmed for a molecular iridium water-oxidation catalyst in a dye-sensitized photoanode performing photocatalytic water oxidation. Faradaic efficiencies were calculated from the comparison of the quantity of O₂ measured by the Clark electrode to the theoretically produced O₂ from the charge passed during each assay; the faradaic efficiencies were approximately (80±10) % for the photoanodes investigated. SnO₂-(32:1), SnO₂-P, SnO₂-Ir, and SnO₂ did not show O₂ evolution (Figure S27).

Figure 5

Clark electrode O₂ production measurements of SnO₂-(2:1) (A and C) and SnO₂-(8:1) (B and D) over 20 h. A bias of 0.8 V vs. NHE was applied under illumination. The illumination for (C) and (D) was three times more intense. A 420 nm-long pass filter was used in all cases. The Clark electrode is extremely sensitive to applied electrochemical bias, light exposure, stirring rate, and temperature. The temperature was controlled at 25 °C to minimize oscillations, but they are unavoidable if an electrochemical bias and light are both required for the assays. The dip in trace (A) at approximately 6 h was caused by the stir bar briefly hitting the electrode.

Incident photon-to-current efficiencies (IPCE) were measured for SnO₂-(2:1), SnO₂-(8:1), and SnO₂-(32:1) and are shown in Figure 6. The efficiencies are less than 2 %, which is typical for many photoanodes containing ruthenium dyes and wateroxidation catalysts.3c On the basis of the faradaic efficiencies from the O₂ evolution measurements and the maximum IPCE of approximately 0.9 % for SnO₂-(2:1) and SnO₂-(8:1), the overall yields for photocatalytic water oxidation are approximately 0.7 %. The low efficiencies are thought to arise from fast back electron transfer from the conduction band of SnO₂ to the oxidized P or Ir, as this is a common problem for photoanodes.2b, 3c

Figure 6

IPCE measurements for SnO₂-(2:1), SnO₂-(8:1), and SnO₂-(32:1).

To understand the processes that occur during photocatalysis and the reasons for the less than ideal yields, photostability measurements were performed. SnO₂-(2:1), SnO₂-(8:1), SnO₂-(32:1), and SnO₂-P were illuminated for approximately 22 h at a bias of 0.8 V versus NHE (chronoamperogram in Figure S25), and the samples were analyzed before and after illumination. The UV/Vis spectra of the samples before and after the photostability measurements are shown in Figure 7.

Figure 7

UV/Vis spectra before (subscript “0”) and after (subscript “f”) illumination and the application of 0.8 V vs. NHE for (A) SnO₂-(2:1), (B) SnO₂-(8:1), (C) SnO₂-(32:1), and (D)  SnO₂-P. (E) A photograph of the photoanodes after approximately 22 h of illumination with a bias of 0.8 V vs. NHE.

Interestingly, there is a trend in the photostabilities of the samples: as the amount of P present on the photoanode increases, the photodegradation also increases. As shown in Figure 7, the UV/Vis spectra of SnO₂-(2:1) before illumination and after approximately 22 h are nearly identical. As the P/Ir ratio increases, more significant photodegradation is observed in the UV/Vis spectra. This photodegradation is especially substantial for SnO₂-P, in which the oxidized P cannot cause productive oxidation events because no Ir is present to oxidize. The photodegradation of the samples can also be seen visually (Figure 7 E), and the color of the sample fades if more P is present. The UV/Vis spectra of the electrolyte after the photostability tests did not show the desorption of P into the electrolyte, and this suggests further that photodegradation occurs and not dye desorption (Figure S7).

XPS measurements were also performed after the photostability measurements. The retention of the Ir and F signatures in the photoanodes confirms that the molecules remained surface-bound during the 22 h photoelectrochemical experiments (Figures S17 and S18). This further highlights the practicality of the use of silatrane and hydroxamic acid anchoring groups in photoanodes. A slight shift to higher binding energies for the Ir 4f peaks of SnO₂-(2:1) and SnO₂-(8:1) suggests that the oxidation of Ir by P^.+ had occurred; this shift was not observed for SnO₂-(32:1). In addition, ATR-FTIR spectroscopy measurements after illumination show the presence of a new C=O stretch at =1730 cm⁻¹, which suggests that oxidation of the porphyrin ring occurs (Figures S12 and S13). A similar process occurs in the enzyme heme oxygenase, in which the heme tetrapyrrole ring reacts with O₂ to form a new C=O group with eventual conversion into biliverdin.22 The XPS results also show a new O 1s peak and changes to the N 1s peaks after the photostability tests and, therefore, support the introduction of a new C=O group into the tetrapyrrole ring of P and a change in the ring structure (Figures S19 and S20). Together, these experiments support the idea that P undergoes structural changes during photodegradation, and this is clearly an additional reason for the lower efficiencies in the photoanodes.

Finally, to demonstrate that the photoanodes had been structurally optimized, two samples were prepared and analyzed. First, photoanodes were prepared on TiO₂ surfaces. As the porphyrin in P cannot inject into the conduction band of TiO₂,19 which is approximately 500 mV higher than that of SnO₂, these samples should not be operable. The UV/Vis spectra and CV data of TiO₂-(2:1) are similar to those of SnO₂-(2:1) (Figures S8 and S22); however, TiO₂-(2:1) does not show any current response upon illumination (Figure 8), as P is unable to inject electrons into the conduction band; therefore, O₂ evolution is not observed (Figure S28). The second sample was a Zn-metalated version of P (Zn-P), which should be unable to oxidize Ir as it has a potential of 1.25 V versus NHE.19 The CA measurements revealed a current response for the illumination of SnO₂-(Zn-P+Ir) because Zn-P can inject (S₁=−0.85 V vs. NHE) into the conduction band of SnO₂ (Figure 8); however, no O₂ evolution was detected because the oxidation of Ir cannot occur (Figure S28). These data highlight the importance of combining an appropriate semiconductor electrode with a suitable high-potential porphyrin for the assembly of functional photoanodes.

Figure 8

Chronoamperograms of the structural controls TiO₂-(2:1) and SnO₂-(Zn-P+Ir) with a 0.8 V vs. NHE applied bias in 0.1 m KNO₃ at pH 6. The samples were illuminated (100 mW cm⁻²) from 20–40, 60–80, and 100–120 s.

Conclusions

Taken together, the photoelectrochemical repetition experiments, nanosecond transient absorption spectroscopy (nsTAS) kinetics, O₂-evolution measurements, and photostability studies suggest that the photosensitizer P can oxidize the Ir catalyst on our photoanodes. The samples with lower P/Ir ratios, SnO₂-(2:1) and SnO₂-(8:1), had less photodegradation and better photoelectrochemical reproducibility. The higher relative content of Ir present on the surfaces of SnO₂-(2:1) and SnO₂-(8:1) increased the probability of oxidization of Ir by P^.+,4a which resulted in the evolution of O₂. The O₂ production and yields are limited not only by well-studied back electron transfer events2b, 3c but also by photodegradation in the sample, which increased if more P was bound to the electrodes. Even though photodegradation occurred, P and Ir remained bound to the electrodes over extended time periods, and this behavior highlights the advantages of silatrane and tetrahydropyranyl-protected (THP-protected) hydroxamic acid anchoring groups for the construction of robust water-splitting dye-sensitized photoelectrochemical cells (WS-DSPECs). These results emphasize the importance of optimizing the relative concentrations of the dye and catalyst molecules on semiconductor electrode surfaces to make the most active WS-DSPECs as well as the need for the development of more stable dye molecules for photoanodes.

Experimental Section

Details of the instrumentation, sample preparation, synthetic procedures, experimental procedures, additional electrochemical and spectroscopic data (UV/Vis, XPS, ATR-FTIR, photoelectrochemistry), and additional O₂-evolution measurements can be found in the Supporting Information.

Acknowledgements

This work was supported by the U.S. Department of Energy, Chemical Sciences, Geosciences, and Biosciences Division, Office of Basic Energy Sciences, Office of Science (DE-FG02-07ER15909). Further support was provided by a generous gift from the TomKat Foundation. The authors thank Dr. Min Li and the Materials Characterization Core on Yale West Campus for help with the XPS measurements and also Coleen T. Nemes for help performing the IPCE measurements.

Conflict of interest

The authors declare no conflict of interest.