Dynamic charge collecting mechanisms of cobalt phosphate on hematite photoanodes studied by photoinduced absorption spectroscopy

Abstract

Reaction sites, surface states, and surface loaded electrocatalysts are photoinduced charge storage sites and critical to photoelectrochemical (PEC) performance, however the charge transfer mechanisms involved in the three remain poorly understood. Herein, we studied the charge transfer processes in hematite (Fe₂O₃) without/with loaded cobalt phosphate (CoPi) by operando photoinduced absorption (PIA) spectroscopy. The loaded CoPi receives trapped holes in surface states at low potential and directly captures holes in the valence band at high potential. Through the dynamic hole storage mechanisms, loaded CoPi on Fe₂O₃ facilitates spatial charge separation and serves as a charge transfer mediator, instead of serving as a catalyst to change the water oxidation mechanism (constant third-order reaction). The spatial separation of photoinduced charges between Fe₂O₃ and CoPi results in more long-lived holes on the Fe₂O₃ surface to improve PEC water oxidation kinetically. The dynamic charge collecting mechanism sheds light on the understanding and designing of electrocatalyst loaded photoanodes.

---

Introduction

Photoelectrochemical (PEC) water splitting is a promising strategy to convert solar energy into clean hydrogen fuel.¹ In PEC water splitting, the water oxidation process by a photoanode is generally regarded as a bottleneck and strongly relies on the surface hole density, which is related to various hole storage sites.^2–4 A high-performance photoanode is usually composed of a conductive back contact, light-absorbing semiconductor and surface loaded electrocatalyst. The multiple interfaces (semiconductor|electrocatalyst, semiconductor|solution, and electrocatalyst|solution) and charge storage sites (reaction sites, surface states of the photoanode, and the loaded electrocatalyst) complicate the surface reaction mechanisms.^5,6 Therefore, understanding the charge transfer mechanisms among reaction sites, surface states of the photoanode and the loaded electrocatalyst in photoelectrocatalysis is critical.

Among various photoanodes, hematite (Fe₂O₃) has received lots of attention owing to its abundant, nontoxic and scalable characteristics.^7,8 There have been many research studies on revealing hole storage sites related to PEC properties in the bare Fe₂O₃ system, including reaction intermediates,^9–11 multi-hole cooperation clusters^4,12 and filling surface trap states.¹³ In order to promote PEC performances, the OER catalysts, such as cobalt phosphate (CoPi)¹⁴ and NiFe(OH)_x electrocatalysts,^15–17 are usually loaded on the Fe₂O₃ surface as new hole storage sites. The loaded electrocatalyst negatively shifts the onset potential and improves the photocurrent, while it definitely influences or fully changes the properties of hole storage sites resulting in complicated charge transfer processes. In particular, the promoted role of the electrocatalyst on the Fe₂O₃ surface is a subject of debate, particularly CoPi, a common and effective covering layer.^6,14,18–21 There appear two opposite explanations about the role of CoPi including a catalyst and spectator by (photo)electrochemical and spectroscopic techniques.^18,22,23 Boettcher claimed that potential-sensing electrochemical atomic force microscopy (AFM) in operando gave the direct evidence of loaded CoPi as a catalyst by tracking surface potential variation of CoPi in photoelectrocatalysis.^18–20 Nevertheless, in situ spectroscopy likewise offered direct proof of the CoPi layer as a spectator through similar spectra shapes and charge transfer dynamics regardless of transient or steady state spectroscopy.^14,24–26 Furthermore, spectroscopic results also revealed a crucial point that the contribution of the CoPi layer to photocurrent was lower than 5%.²² However, these research studies were mainly devoted to revealing the reaction sites of the water oxidation reaction of Fe₂O₃ modified with an electrocatalyst and ignore interfacial dynamic processes.

In principle, both the surface states of Fe₂O₃ and loaded electrocatalyst will compete for photoinduced holes with reaction sites on the photoanode surface. The reduction reaction of photoinduced holes by surface states or the electrocatalyst is dependent on their thermodynamic driving force, and should show distinct dynamics/kinetics. Generally, compared with free holes in the valence band (VB), the trapped holes in surface states possess a lower PEC activity and tend to recombine with photoinduced electrons. If the loaded electrocatalyst plays a leading role in surface catalysis, a surface reaction kinetics similar to electrocatalysis is expected, vice versa. But whether the trapped holes interact with the loaded electrocatalyst and its role in photoelectrocatalysis is unclear. More importantly, it is also not studied whether there is charge transfer between the loaded electrocatalyst and reaction sites of Fe₂O₃ and its crucial roles if loaded electrocatalysts do not serve as a catalyst. Thus, the dynamic interaction between the electrocatalyst and Fe₂O₃ and its role in PEC performances are quite important and deserve to be studied deeply.

In this work, the charge transfer processes among surface states, reaction sites and the coated electrocatalyst for Fe₂O₃ without/with CoPi loading were studied by photoinduced absorption (PIA) spectroscopy. Loaded CoPi on the Fe₂O₃ surface does not remove the mid-gap surface states. The dynamic transition in collecting holes for loaded CoPi from surface states to the VB following increasing potential is discussed. The roles of the dynamic charge collecting mechanism in PEC water oxidation are revealed. This dynamic mechanism sheds light on the understanding and designing of electrocatalyst loaded photoanodes.

Results and discussion

Determining the distribution of surface states in bare Fe₂O₃ and CoPi/Fe₂O₃

Fe₂O₃ photoanodes were prepared by a combined method of solvothermal and high-temperature calcination.²⁷ A subsequent electrodeposition was applied to obtain CoPi/Fe₂O₃ photoanodes.¹⁴ The cyclic voltammetry (CV) curve of CoPi/Fe₂O₃ shows negligible current at the applied bias of 0.6–1.6 V vs. RHE in the dark (Fig. S1†), while the CV curve of CoPi displays obvious current for cobalt oxidation or water oxidation. The negligible current of CoPi/Fe₂O₃ suggests a dense and non-porous Fe₂O₃ film.¹⁶ For the PEC performances, loaded CoPi on the Fe₂O₃ surface negatively shifts onset potential, improves photocurrent, and enhances transient spikes (Fig. 1a and b), in line with previous reports.^14,23,28

Fig. 1 Potential-dependent transient photocurrents for (a) bare Fe₂O₃ and (b) CoPi/Fe₂O₃ photoanodes under chopped illumination (AM 1.5 G: 100 mW cm⁻²). The applied potential is shown by a dashed line in the right axis. DOS of surface states in (c) bare Fe₂O₃ and (d) CoPi/Fe₂O₃ as a function of potential under illumination (456 nm LED: 5 mW cm⁻²) and in the dark.

In order to determine the charge transfer processes among reaction sites, surface states and the surface loaded electrocatalyst, the potential distribution of mid-gap or surface states is first investigated by electrochemical impedance spectroscopy (EIS) under illumination and in the dark. The EIS results (Fig. 1c and d) fitted with the equivalent circuit (Fig. S2†) show that bare Fe₂O₃ and CoPi/Fe₂O₃ photoanodes exhibit a similar distribution of density of states (DOS = C/e, where C is the capacitance and e is the elementary charge.) in the 0.7–1.4 V_RHE range. Loading CoPi on the Fe₂O₃ surface shifts the distribution of surface states about 100 meV towards lower potential. Moreover, CV results in Fig. S3† after PEC water oxidation also display similar reduction waves in the 0.8–1.4 V_RHE range for bare Fe₂O₃ or in the 0.6–1.1 V_RHE range for CoPi/Fe₂O₃. Both EIS and CV results demonstrate that there are similar distributions of surface states in bare Fe₂O₃ and CoPi/Fe₂O₃ photoanodes in the 0.8–1.4 V_RHE range. Therefore, although loading CoPi eliminates fractional surface states and releases Fermi-level pinning revealed by fast decay in chopped OCP results (Fig. S4†), the surface states in the 0.8–1.4 V_RHE range still exist in CoPi/Fe₂O₃.

The surface states in Fe₂O₃ always act as hole trap states and have been described as water oxidation intermediates^9,11 or recombination sites,²⁹ because Fe₂O₃ is an n-type semiconductor with a Fermi level close to the conduction band (CB) edge. Thus, the surface states in the 0.8–1.4 V_RHE range are tentatively assigned as hole trap states for both bare Fe₂O₃ and CoPi/Fe₂O₃ photoanodes. In principle, with the existence of surface states, the photoinduced holes in the VB might be consumed by indirect paths (water oxidation by trapped holes: h⁺_{surface states} + H₂O → O₂, or trap-mediated recombination) for bare Fe₂O₃, besides direct paths (water oxidation by VB holes: h⁺_VB + H₂O → O₂, or recombination). When CoPi is loaded on Fe₂O₃, the surface states can bring new charge transfer paths between the photoinduced holes, surface states and the loaded CoPi. Namely, the photoinduced holes in Fe₂O₃ can move to CoPi via surface states (sequential hole transfer) besides the direct transfer of VB holes to CoPi, which may play a vital role in CoPi/Fe₂O₃ photoelectrocatalysis.

Charging sites of CoPi/Fe₂O₃ at low and high potentials

Next, the photoinduced charge transfer processes were studied to reveal the direct or sequential hole transfer processes in CoPi/Fe₂O₃ with the existence of surface states. Based on the position of the energy levels, the oxidation ability of trapped holes (surface states: 0.8–1.4 V_RHE) should be lower than that of free holes (VB: 2.5 V_RHE). A logical speculation is that a probe with <1.4 V_RHE redox potential can be oxidized by trapped holes, otherwise there will be no oxidation as schematically shown in Fig. 2a. Thus, CoPi (Co^2+/3+: ∼1.10 V_RHE) and NiOOH (Ni^2+/3+: ∼1.38 V_RHE) electrocatalysts were chosen as probes and deposited on the Fe₂O₃ surface (Fig. 2b). The role of surface states in charge transfer was first studied by PIA spectroscopy at low potential (0.8 V_RHE). The wavelength-dependent PIA responses of CoPi, bare Fe₂O₃, CoPi/Fe₂O₃ and NiOOH/Fe₂O₃ at 0.8 V_RHE are shown in Fig. 2c. Bare Fe₂O₃ and CoPi display negligible PIA intensity, illustrating that the bulk/surface recombination is too fast to probe. In contrast, CoPi/Fe₂O₃ presents an appreciable PIA intensity, indicating the effective hole transfer from Fe₂O₃ to CoPi. Moreover, the consistence of PIA spectra of CoPi/Fe₂O₃ and spectroelectrochemical absorption (SECA) spectra of CoPi (Co^2+/3+, for calculation details, see Fig. S5†) in Fig. 2c indicates that CoPi is the main hole storage site at 0.8 V_RHE. Amazingly, no PIA response is observed in NiOOH/Fe₂O₃ at 0.8 V_RHE, corresponding to no hole transfer from Fe₂O₃ to NiOOH and in line with negligible photocurrent in Fig. S6.† Since free holes in the VB of Fe₂O₃ can oxidize both CoPi and NiOOH, the distinct PIA response should result from the CoPi oxidation by the long-lived trapped holes in the surface states of Fe₂O₃. Thus, at low potential, the photoinduced holes in Fe₂O₃ mainly undergo sequential transfer processes from the VB to surface states then to CoPi. The spatial separation between holes in CoPi and electrons in Fe₂O₃ prolongs charge lifetimes (Fig. S7†) by suppressing charge recombination at 0.8 V_RHE. Moreover, transient absorption spectra of CoPi/Fe₂O₃ at 100 ms at 0.8 V_RHE in Fig. S7e† display a broad absorption which is quite similar to the absorption spectra of oxidized Co³⁺, demonstrating that the long-lived transient absorption signals in CoPi/Fe₂O₃ at lower potential should originate from cobalt oxidation of Co^2+/3+ rather than passivating surface states or increasing band-bending.¹⁴ A combination of negligible photocurrent (Fig. 1b) and appreciable PIA (Fig. 2c) at 0.8 V_RHE demonstrates that the oxidized CoPi cannot contribute to the water oxidation processes. Therefore, at low potential, the loaded CoPi on the Fe₂O₃ surface serves as the charging site for photoinduced holes, and prolongs the lifetime of holes by suppressing recombination, but with no catalysis effect for water oxidation.

Fig. 2 (a) Schematic of charge transfer between surface states and probes. Red and blue arrows refer to the transfer processes of photoinduced holes and electrons, respectively. ‘×’ denotes no charge transfer. (b) CV and corresponding redox potentials of CoPi and NiOOH. (c) Wavelength-dependent PIA spectra of bare Fe₂O₃, CoPi, CoPi/Fe₂O₃ and NiOOH/Fe₂O₃ at low potential of 0.8 V_RHE; left axis is PIA spectra and right axis is the SECA spectrum of CoPi (Co^2+/3+: 1.3 V_RHE – 1.0 V_RHE). (d) Wavelength-dependent PIA spectra of bare Fe₂O₃, CoPi and CoPi/Fe₂O₃ at high potential of 1.4 V_RHE; left axis is PIA spectra and right axis is the SECA spectrum of CoPi (Co^3+/4+: 1.7 V_RHE – 1.4 V_RHE). Pump wavelength in PIA is 456 nm (5 mW cm⁻²).

Then, the charge transfer processes were studied by PIA spectroscopy at high potential of 1.4 V_RHE to reveal the main charging sites during the OER. In wavelength-dependent PIA spectra (Fig. 2d), bare Fe₂O₃ and CoPi/Fe₂O₃ display similar spectral shapes from 500 to 900 nm, while CoPi still shows no photo-response. The PIA spectra shape is in line with the absorption spectra of photoinduced holes in Fe₂O₃ and different from absorption properties of oxidized cobalt species in CoPi (Co^2+/3+ shown in Fig. 2c and Co^3+/4+ shown in Fig. 2d).^12,29,30 These results suggest that the charging sites of bare Fe₂O₃ and CoPi/Fe₂O₃ photoanodes at high potential (1.4 V_RHE) may be mainly on the Fe₂O₃ surface. Moreover, the similar transient absorption results of bare Fe₂O₃ and CoPi/Fe₂O₃ photoanodes further confirm the assignment of same charging sites (Fig. S8†).^14,24

Besides the spectral shape, an obvious increase of PIA amplitude for CoPi/Fe₂O₃ is observed relative to that of Fe₂O₃ in Fig. 2d. In order to uncover the origin of the increased PIA amplitude, a quantitative comparison of ratios of PIA amplitudes (PIA_CoPi/Fe2O3/PIA_Fe2O3 ≥1.8) and photocurrents (J_CoPi/Fe2O3/J_Fe2O3 ∼ 1.8) for the Fe₂O₃ and CoPi/Fe₂O₃ photoanodes is displayed in Figs. S8a† and S8b. The ratios display the proportional relationship between photocurrent and hole density (J ∝ [h⁺]¹). However, CoPi/Fe₂O₃ at this potential (1.4 V_RHE) and illumination (5 mW cm⁻²) displays a third-order kinetics (J ∝ [h⁺]³) which will be further discussed in detail below, suggesting that the increased PIA amplitude should lead to a much higher ratio of photocurrents (1.83³ ≈ 5.8). This disagreement indicates that the increased PIA amplitude is not only derived from the increased holes in Fe₂O₃, but also from the effective hole storage in loaded CoPi. Since the PIA data are obtained from quasi-steady-state PIA spectra measured with a continuous pulsed light on, the initial states of cobalt ions in CoPi/Fe₂O₃ for PIA at 1.4 V_RHE should be the Co³⁺ state due to the sequential and direct hole transfer processes (Fig. S9c†). The increased PIA amplitude in CoPi/Fe₂O₃ (Fig. 2d) is derived from the oxidation process of Co^3+/4+ by photoinduced holes in the VB of Fe₂O₃ besides increased holes in Fe₂O₃, since the trapped holes in surface states cannot oxidize Co^3+/4+ (redox potential: 1.47 V_RHE). The Co^3+/4+ oxidation process via direct hole transfer contributes little to the photocurrent of CoPi/Fe₂O₃ at 1.4 V_RHE, which should be less than ∼1 μA by evaluation from the PIA amplitude in the I–t curve in Fig. S10.† Loaded CoPi on the Fe₂O₃ surface still exhibits negligible catalysis at high potential. Therefore, at high potential (1.4 V_RHE), the main charging sites of photoinduced holes in the CoPi/Fe₂O₃ photoanode are the reaction sites on the Fe₂O₃ surface and Co^3+/4+ in the CoPi electrocatalyst. Moreover, the same as that at low potential (0.8 V_RHE), the spatial separation between holes in CoPi and electrons in Fe₂O₃ also prolongs charge lifetimes by suppressing charge recombination (Fig. S11†) at high potential (1.4 V_RHE).

Potentiodynamic charge storage processes

Wavelength-dependent PIA spectra of bare Fe₂O₃, CoPi and CoPi/Fe₂O₃ demonstrate that loaded CoPi displays two distinct hole-storage mechanisms, where loaded CoPi on the Fe₂O₃ surface receives trapped holes in surface states and free holes in the VB at low and high potentials, respectively. In order to further uncover potential-dependent charge storage mechanisms, the PIA spectra were recorded at a fixed probe wavelength and varying potentials (0.6 V_RHE ≤ potential ≤ 1.6 V_RHE). Since the probe wavelength of > 650 nm in PIA spectra has been assigned to photoinduced holes on the Fe₂O₃ surface by the hole scavenger and potential-dependent TAS experiments (Fig. S8†),^15,31 the probe wavelength of 694.3 nm was used to monitor photoinduced hole dynamics on the Fe₂O₃ surface. The reasonability is confirmed by the linear relationship of PIA intensity and photoinduced hole density in Fe₂O₃ in Fig. S10.† Moreover, the linear relationship between absorbance and high-valent cobalt species (holes) in Fig. S9† demonstrates that the probe wavelength of 694.3 nm is also suitable for monitoring holes in CoPi.

Fig. 3a–c show the potential-dependent PIA spectra of bare Fe₂O₃, CoPi and CoPi/Fe₂O₃ photoanodes. There are three regions for a PIA spectrum, including baseline in the dark (0–0.5 s), photoinduced species accumulation (0.5–6 s) and species consumption after turning the light off (6–10 s). The variation trend of PIA amplitudes with applied potential is extracted and shown in Fig. 3d. CoPi shows no PIA response at 0.6–1.6 V_RHE, suggesting that PEC charging of CoPi is only through collecting extraneous charges from Fe₂O₃. For the bare Fe₂O₃ photoanode, the PIA amplitudes increase monotonically to a saturation intensity. The invariable PIA intensity above 1.2 V_RHE can originate from the signal overlap of inactive holes (surface states) and active holes (VB), which is a common phenomenon for the Fe₂O₃ photoanode by the solvothermal method.²⁹ Increasing the external potential results in higher percentage of active holes, but doesn't influence PIA intensity. However, potential-dependent PIA amplitudes of CoPi/Fe₂O₃ display a complicated trend, first increasing (0.7–1.0 V_RHE) and then decreasing (1.0–1.6 V_RHE) resulting in a peak at 1.0 V_RHE. Moreover, at the same potential, the PIA amplitudes of CoPi/Fe₂O₃ are greater than those of bare Fe₂O₃. Because the extinction coefficients of Co²⁺/Co³⁺ (ε₁ = 7332 M⁻¹ cm⁻¹) and Co³⁺/Co⁴⁺ (ε₂ = 1833 M⁻¹ cm⁻¹) in CoPi are larger than the extinction coefficient of photoinduced holes (ε = 637 M⁻¹ cm⁻¹) in bare Fe₂O₃ (Fig. S10†), the increased PIA amplitude in the whole applied potential range (0.6–1.6 V_RHE) should originate from hole transfer from the Fe₂O₃ surface to CoPi. Combined with all PIA results in Fig. 2 and 3, the complicated variation of PIA amplitudes in CoPi/Fe₂O₃ can be attributed to the transition from oxidation of Co²⁺/Co³⁺ by trapped holes in surface states at low potentials to oxidation of Co³⁺/Co⁴⁺ by free holes in the VB at high potentials. That's to say, the difference of extinction coefficients should be the origin of the complicated trend in potential-dependent PIA amplitudes for CoPi/Fe₂O₃. Furthermore, the potential-dependent PIA spectra of CoPi/Fe₂O₃–air and CoPi/Fe₂O₃–Ar (Fig. S12†) with modified surface states display a similar change trend, suggesting that the potential-dependent PIA variation in electrocatalyst-coated Fe₂O₃ can be related to the distribution of surface states, free holes in the VB and oxidation states of the electrocatalyst.

Fig. 3 PIA responses at various potentials for bare Fe₂O₃ (a), CoPi (b) and CoPi/Fe₂O₃ (c). (d) PIA amplitudes following increasing potential extracted from Fig. 2a–2c. Average lifetimes (τ_ave) of (e) accumulated and (f) dissipative processes for bare Fe₂O₃ and CoPi/Fe₂O₃ photoanodes in (a and c). Pump wavelength is 456 nm (5 mW cm⁻²); probe wavelength is 694.3 nm; colors of curves at different potentials in (a−c) are identical.

Effect of stored holes in CoPi on charge transfer

After knowing the potentiodynamic hole storage mechanisms of loaded CoPi, the dissipation paths of holes in CoPi are further studied to uncover the function of CoPi in PEC water oxidation. By comparing the potential-dependent PIA responses of bare Fe₂O₃ and CoPi/Fe₂O₃ in Fig. 3a and c, the PIA spectra of the CoPi/Fe₂O₃ photoanode display slower accumulation and dissipation processes. In order to specifically compare the kinetic change, the lifetimes of PIA accumulation and dissipation processes for bare Fe₂O₃ and CoPi/Fe₂O₃ photoanodes are fitted using a bi-exponential function in Fig. 3e and f (fitted results are shown in Table S1†). At 0.7–0.9 V_RHE, drastic recombination leads to no accumulation processes in bare Fe₂O₃. After loading CoPi on the Fe₂O₃ surface, the sequential hole transfer through surface states to oxidize Co^2+/3+ results in the long accumulation lifetimes of 2.4∼3.8 s. With potential above 1.0 V_RHE, the accumulation processes of the bare Fe₂O₃ photoanode show an average lifetime of ∼0.45 s, which is faster than the shortest accumulation lifetime (∼0.62 s) in the CoPi/Fe₂O₃ photoanode. Due to concentrated electrolyte (1 M NaOH) and weak illumination (5 mW cm⁻²) in PIA measurements, the slow accumulation lifetimes in the CoPi/Fe₂O₃ photoanode shouldn't be due to the CoPi deposition caused by the local pH change under light irradiation.³² Moreover, with the increase of applied potential, the first increasing then decreasing accumulation lifetimes in Fig. 3e also exclude diffusion-controlled CoPi deposition. Thus, this slow accumulation process is attributed to slow hole hopping from the CoPi/Fe₂O₃ interface to the CoPi bulk after cobalt oxidation at the millisecond timescale. The prolonged accumulation lifetimes indicate that loaded CoPi induces new charge separation paths near the Fe₂O₃ surface for promoting charge separation via cobalt oxidation by photoinduced holes. The smaller accumulation lifetimes with potential above 1.2 V_RHE than that with potential below 1.0 V_RHE suggest that the direct hole transfer to oxidize Co^3+/4+ is faster than the sequential hole transfer to oxidize Co^2+/3+.

For dissipation processes, except the short lifetime (∼0.6 s) near the onset potential, bare Fe₂O₃ displays an almost constant dissipation lifetime about 1.2 s, in line with the understanding of the water oxidation reaction by free holes in the VB.³ Namely, the increased potential for bare Fe₂O₃ is used for suppressing recombination and increasing density of free holes in the VB, rather than changing the enthalpy or activation energy. In contrast, the dissipation lifetimes of CoPi/Fe₂O₃ exhibit first an increased then decreased trend, resulting in the longest lifetime about 2.9 s at 1.0 V_RHE. Due to the weak contribution to water oxidation of CoPi (Fig. 2c and d), the slow PIA decay should be attributed to the recombination process between stored holes in CoPi and photoinduced electrons on the Fe₂O₃ surface. The dissipation lifetimes of 1.6 ∼ 2.9 s (0.7–1.0 V_RHE) directly demonstrate the effective suppression of charge recombination by Co³⁺ states. When external potential increases to 1.6 V_RHE, the dissipative average lifetime (∼1.6 s) is still longer than the dissipation process of bare Fe₂O₃ (∼1.2 s), indicating that the oxidized Co⁴⁺ states also suppress the charge recombination effectively. The suppressed charge recombination is also supported by the TAS and TPC results (Fig. S7 and S11†).

Then, the charge recombination between oxidized CoPi and Fe₂O₃ was investigated at high potential, to confirm the roles of oxidized Co⁴⁺ in PEC processes. In principle, if the Co⁴⁺ concentration is high enough in CoPi for the CoPi/Fe₂O₃ photoanode at the ground state, photogenerated hole transfer from Fe₂O₃ to CoPi under illumination will be blocked. In consequence, more recombination events of photoinduced electrons and holes in CoPi will be observed, which will lead to decreased PIA amplitude. In order to confirm this hypothesis, CoPi was loaded on the porous Fe₂O₃. Since the conductive substrate (FTO) is partially bare, there exist triphasic interfaces of FTO/CoPi, FTO/Fe₂O₃ and CoPi/Fe₂O₃, as shown in Fig. S13a.† When potential is applied to CoPi/porous-Fe₂O₃, CoPi can be directly oxidized to high-valent cobalt species (Co⁴⁺) resulting in typical redox waves (Fig. S13b†). Then the charge transfer processes of CoPi/porous-Fe₂O₃ are studied using PIA. CoPi/porous-Fe₂O₃ shows similar potential-dependent PIA spectra to that of the above dense CoPi/Fe₂O₃ under low light intensity irradiation of 5 mW cm⁻² (Fig. S14†). As expected, the PIA spectra in Fig. 4a display a negative amplitude at high potential (1.5 V_RHE) and under illumination (30 mW cm⁻²). Additionally, the chopped photocurrent at 1.5 V_RHE in Fig. 4b shows a disappearing transient spike, which can be understood by electron shunt transfer. Photoinduced electrons don't move from the porous Fe₂O₃ surface to the bulk then to FTO, but from the porous Fe₂O₃ surface to CoPi then to FTO resulting in a positive current. The oxidized CoPi can effectively capture photoinduced electrons, resulting in a flow of electrons at the moment of suddenly turning the light on/off. Because the flow of electrons induced by CoPi is opposite to the flow of back electrons, the disappearing transient spikes are observed. Thus, both PIA and photocurrent results of porous Fe₂O₃ with loaded CoPi directly demonstrate the possibility of charge recombination between Co⁴⁺ species and photoinduced electrons of Fe₂O₃. Therefore, at high potential, loaded CoPi on dense/porous-Fe₂O₃ serves as a mediator to fast capture holes in the VB of Fe₂O₃ followed by slow consumption by photogenerated electrons of the Fe₂O₃ surface, which can slow down the recombination process and produce more long-lived holes for water oxidation.

Fig. 4 Potential-dependent PIA spectra (a) and chopped photocurrents (b) for CoPi/porous-Fe₂O₃. Pump wavelength is 456 nm (30 mW cm⁻²); probe wavelength is 694.3 nm; colors of curves at different potentials in (a and b) are identical.

To further understand the roles of electrocatalyst-mediated charge transfer processes in the PEC water oxidation reaction, we correlate reaction rate and hole density for Fe₂O₃ without/without CoPi loading at increased potential. The light-intensity-dependent PIA spectra were recorded for bare Fe₂O₃ and CoPi/Fe₂O₃ photoanodes. In Fig. 5a and b, PIA amplitudes for the two photoanodes increase with light intensity increasing at a fixed potential, suggesting more holes on the Fe₂O₃ surface and promotional oxidation for CoPi. According to previous reports,^4,12 the photoinduced hole can directly be regarded as a reactant, and its reaction parameters can be quantified by rate analysis (r = J = k × [h⁺]^β, where r is the reaction rate and equivalent to current density (J), k is the rate constant, [h]⁺ is the hole density and β is the reaction order). Due to the linear relation of PIA intensity and hole concentration in Fig. S15,† the reaction orders at changed potential from 0.6 V_RHE to 1.6 V_RHE are calculated and shown in Fig. 5c and d. Although PIA amplitudes of Fe₂O₃ under fixed illumination display saturation above 1.2 V_RHE due to surface states, there is an effective rate analysis at the fixed potential and changed illumination in Fig. 5a. This result demonstrates that there is a linear scaling relationship of photoinduced hole density distribution between the VB and surface states in the present study. With potential increase, the reaction order of Fe₂O₃ displays first an increase to a maximum value (from zero to fourth) and then a decrease to a constant (from fourth to third) value in Fig. 5e, implying that surface states participate in or modulate the surface reaction. After loading CoPi on the Fe₂O₃ surface, in line with the scaling relationship of hole density with PIA (Fig. S15d†), a similar linear shape for the reaction order–potential relation is obtained. Particularly, the maximum (fourth) and constant (third) reaction orders for CoPi/Fe₂O₃ arise at lower potential than bare Fe₂O₃ (Fig. 5e), illustrating similar functions of surface states and coated CoPi electrocatalyst. Furthermore, Fig. 5f demonstrates that the relationship of hole density (electro-induced absorption (EIA) intensity) and current in the CoPi electrocatalyst obeys an Eying-like equation (log(J) ∝ [h⁺]) reported by Hong Nhan Nong,³³ rather than rate analysis that shows a high-order reaction (β ≈ 9) for electrocatalytic current from microampere to milliampere. These results demonstrate that surface states and the loaded electrocatalyst can directly receive and store a part of holes from the VB of hematite, but don't affect the reaction kinetics of reaction sites.

Fig. 5 Light-intensity-dependent PIA spectra of (a) bare Fe₂O₃ and (b) CoPi/Fe₂O₃ at increased potential. Pump wavelength is 456 nm; probe wavelength is 694.3 nm. Rate law analysis for (c) Fe₂O₃ and (d) CoPi/Fe₂O₃. The colors of lines at various potentials in (c and d) are identical. (e) Reaction order of the two photoanodes as a function of potential. (f) Log(current) versus hole density from EIA spectra for the CoPi electrocatalyst. The inset shows the rate law analysis with corrected hole density for the CoPi electrocatalyst.

Discussion

The charge transfer processes for three kinds of hole storage sites at the interface of Fe₂O₃, CoPi and water under illumination and at applied potential are summarized in Scheme 1. As discussed above, there are surface states for bare Fe₂O₃ with a potential distribution in the 0.8–1.4 V_RHE range. These surface states are occupied by intrinsic electrons, which can only be removed by illumination (trapping holes). Due to limited oxidation activity, the surface states play a role in aggravating recombination of photoinduced charges resulting in large onset potential of bare Fe₂O₃. Moreover, the holes of surface states at low potential (≤0.8 V_RHE) can only oxidize Co²⁺ to Co³⁺ rather than Co³⁺ to Co⁴⁺ and water oxidation, which is the main charge transfer path in the CoPi/Fe₂O₃ photoanode (the left panel of Scheme 1). This process is a sequential hole transfer from the VB to surface states (ps ∼ ns)³⁴ then to CoPi (∼0.6 s), followed by recombination with photoinduced electrons (≥1.3 s) resulting in dissipation of hole energy. Slower recombination in CoPi/Fe₂O₃ is the origin of the negative shift of onset potential.

Scheme 1 Kinetic schematic of photoinduced charge transfer at the interface of Fe₂O₃ and CoPi under illumination and at applied potential from energy (top panel) and space (bottom panel) scenarios. The potential distribution of surface states (orange), Co^2+/3+ (purple) and Co^3+/4+ (grey) are displayed as Gaussian curves. Colored (solid) lines “—” are representative of redox potentials of cobalt species, hydrogen generation and oxygen generation. Red and blue arrows refer to transfer processes of photogenerated electrons and photoinduced holes, respectively (the dashed and solid lines indicate the transferred amount of photoinduced charges).

At high potential (≥1.4 V_RHE), the space charge region and band-bending of Fe₂O₃ will be broadened and increased, respectively (the right panel of Scheme 1). Under quasi-steady-state illumination and at high potential, photoinduced holes can completely fill the surface states and oxidize Co²⁺ to Co³⁺via the sequential hole transfer process. Although surface states on the bare Fe₂O₃ surface can be completely occupied by trapping holes, the photocurrent is still lower than the theoretical value due to inevitable recombination loss by back electrons (Fig. S11†). Because of the delayed recombination process at high potential, more high-energy and long-lived holes in the VB will be generated to directly contribute to water oxidation (OER: ∼0.2 s), as well as Co³⁺ oxidation (Co^3+/4+: ∼0.8 s). Nevertheless, although CoPi acts as the hole storage site, CoPi still doesn't contribute to PEC water oxidation due to low Co⁴⁺ density. However, the stored holes in CoPi are not static spectators but dynamic charge mediators, which can again capture photoinduced electrons for a long time. Owing to spatial charge separation of photogenerated electrons (Fe₂O₃ surface) and trapped holes (CoPi electrocatalyst), CoPi/Fe₂O₃ displays longer recombination lifetimes resulting in enhanced PEC performances.

As shown by the spectroscopic results, the delayed recombination between photogenerated electrons of Fe₂O₃ and trapped holes of CoPi electrocatalyst due to spatial charge separation will improve PEC performances of CoPi/Fe₂O₃, which can be applied to all potentials with photo-response. There are connections and differences between this new viewpoint and previous reports. Although surface state passivation to decrease recombination is common understanding,^26,35,36 the EIS results in Fig. 1d reveal that loaded CoPi leads to more surface states, demonstrating negligible contribution of surface state passivation. Using TAS and PIA, Durrant found consistent spectral shapes and long-lived holes after loading CoPi on the Fe₂O₃ surface, which are attributed to enhanced electron depletion by broadening the space charge region.^14,22,24,25 Besides reproducing the reported results, our potential- and light-intensity-dependent spectra further reveal that loaded CoPi on the Fe₂O₃ surface displays potentiodynamic hole storage mechanisms and acts as a hole storage site of delayed recombination. On the other hand, although Boettcher claimed the direct catalysis evidence of CoPi by tracking surface potential of loaded CoPi in PEC water oxidation,¹⁸ the microscopic measurement of AFM might not reflect the overall character of loaded CoPi electrocatalyst. Nevertheless, because of the ensemble-averaged measurements in spectroscopy, present experiments can't completely exclude that a loaded electrocatalyst (like CoPi) on the Fe₂O₃ surface has no contribution to water oxidation. Instead, we are convinced that a large proportion of the electrocatalyst only acts as a mediator that stores photoinduced holes and recombines with photogenerated electrons. Considering the key role of photogenerated electrons in PEC performance of Fe₂O₃, several strategies of influencing photogenerated electrons on/near the surface were employed, which display considerable PIA difference (Fig. S16†), such as decreasing density of electron trap states by surface covalent doping with fluorine (Fig. S16a†),³⁷ reducing reverse recombination of electrons via surface insulator modification (SnO₂) (Fig. S16b†)³⁸ and enhancing separation efficiency of photoinduced charges by constructing a built-in electric field using a homojunction (Fig. S16c†).¹⁵ These results demonstrate that controlling recombination of photogenerated electrons plays a remarkable role in charge dynamics, highlighting the importance of regulating charge transfer processes at the semiconductor–electrolyte interface.

Conclusions

We studied the charge transfer mechanisms among surface states, reaction sites and the coated electrocatalyst in Fe₂O₃ without/with CoPi by PIA spectroscopy. Surface states in Fe₂O₃ are not removed by loading CoPi, but serve as a mediator. At low potential, photoinduced holes of the VB are preferentially trapped into surface states and trapped holes further move to CoPi, presenting a sequential hole transfer. Photoinduced holes at high potential are prone to direct water oxidation by holes of the VB on the Fe₂O₃ surface, resulting in constant and decreasing PIA signals for Fe₂O₃ and CoPi/Fe₂O₃ photoanodes, respectively. Namely, there appears a dynamic transition in collecting holes for CoPi from surface states to the VB with increasing potential. More importantly, loaded CoPi did not serve as a catalyst to change third-order water oxidation reaction kinetics on the Fe₂O₃ surface, but promotes spatial separation of photoinduced charges resulting in long-lived holes for facilitating PEC water oxidation performance. The dynamic charge collecting mechanism not only deepens the understanding of the roles of the loaded electrocatalyst in PEC, but also helps the designing of the microstructure of electrocatalyst loaded photoanodes.

Data availability

All experimental and characterization data are available in the ESI.†

Author contributions

D. L., X. W. and C. L. conceived the research. D. L. carried out the experiments, analyzed data and wrote the manuscript. R. W. and H. Y. analyzed data. H. Z. provided the homojunction sample. X. W. and C. L. discussed the data and revised the manuscript. All authors have given approval to the final version of the manuscript.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

We acknowledge financial support from the National Key R&D Program of China (2021YFA1500600), the National Natural Science Foundation of China (21872143), the Fundamental Research Center of Artificial Photosynthesis (FReCAP), and the DICP Foundation of Innovative Research (DICP I202122).

Notes and references

1. I. Roger, M. A. Shipman and M. D. Symes, Nat. Rev. Chem., 2017, 1, 1–13.
2. C. Ding, J. Shi, Z. Wang and C. Li, ACS Catal., 2016, 7, 675–688.
3. S. Corby, R. R. Rao, L. Steier and J. R. Durrant, Nat. Rev. Mater., 2021, 6, 1136–1155.
4. C. A. Mesa, L. Francàs, K. R. Yang, P. Garrido−Barros, E. Pastor, Y. Ma, A. Kafizas, T. E. Rosser, M. T. Mayer, E. Reisner, M. Grätzel, V. S. Batista and J. R. Durrant, Nat. Chem., 2020, 12, 82–89.
5. C. Li, Z. Luo, T. Wang and J. Gong, Adv. Mater., 2018, 30, 1707502.
6. D. R. Gamelin, Nat. Chem., 2012, 4, 965–967.
7. D. A. Grave, N. Yatom, D. S. Ellis, M. C. Toroker and A. Rothschild, Adv. Mater., 2018, 30, 1706577.
8. D. Zhou and K. Fan, Chin. J. Catal., 2021, 42, 904–919.
9. O. Zandi and T. W. Hamann, Nat. Chem., 2016, 8, 778–783.
10. Y. Zhang, H. Zhang, A. Liu, C. Chen, W. Song and J. Zhao, J. Am. Chem. Soc., 2018, 140, 3264–3269.
11. B. Klahr and T. Hamann, J. Phys. Chem. C, 2014, 118, 10393–10399.
12. F. Le Formal, E. Pastor, S. D. Tilley, C. A. Mesa, S. R. Pendlebury, M. Gratzel and J. R. Durrant, J. Am. Chem. Soc., 2015, 137, 6629–6637.
13. B. Klahr, S. Gimenez, F. Fabregat-Santiago, T. Hamann and J. Bisquert, J. Am. Chem. Soc., 2012, 134, 4294–4302.
14. M. Barroso, C. A. Mesa, S. R. Pendlebury, A. J. Cowan, T. Hisatomi, K. Sivula, M. Gratzel, D. R. Klug and J. R. Durrant, Proc. Natl. Acad. Sci. U. S. A., 2012, 109, 15640–15645.
15. H. Zhang, D. Li, W. J. Byun, X. Wang, T. J. Shin, H. Y. Jeong, H. Han, C. Li and J. S. Lee, Nat. Commun., 2020, 11, 4622.
16. J. Qiu, H. Hajibabaei, M. R. Nellist, F. A. L. Laskowski, S. Z. Oener, T. W. Hamann and S. W. Boettcher, ACS Energy Lett., 2018, 3, 961–969.
17. P. Shadabipour and T. W. Hamann, Chem. Commun., 2020, 56, 2570–2573.
18. M. R. Nellist, F. A. L. Laskowski, J. Qiu, H. Hajibabaei, K. Sivula, T. W. Hamann and S. W. Boettcher, Nat. Energy, 2017, 3, 46–52.
19. M. R. Nellist, J. Qiu, F. A. L. Laskowski, F. M. Toma and S. W. Boettcher, ACS Energy Lett., 2018, 3, 2286–2291.
20. G. M. Carroll, D. K. Zhong and D. R. Gamelin, Energy Environ. Sci., 2015, 8, 577–584.
21. J. Xiao, L. Fan, Z. Huang, J. Zhong, F. Zhao, K. Xu, S.-F. Zhou and G. Zhan, Chin. J. Catal., 2020, 41, 1761–1771.
22. Y. Ma, A. Kafizas, S. R. Pendlebury, F. Le Formal and J. R. Durrant, Adv. Funct. Mater., 2016, 26, 4951–4960.
23. J. Li, W. Wan, C. A. Triana, Z. Novotny, J. Osterwalder, R. Erni and G. R. Patzke, J. Am. Chem. Soc., 2019, 141, 12839–12848.
24. M. Barroso, A. J. Cowan, S. R. Pendlebury, M. Gratzel, D. R. Klug and J. R. Durrant, J. Am. Chem. Soc., 2011, 133, 14868–14871.
25. Y. Ma, F. Le Formal, A. Kafizas, S. R. Pendlebury and J. R. Durrant, J. Mater. Chem. A, 2015, 3, 20649–20657.
26. C. Zachaus, F. F. Abdi, L. M. Peter and R. van de Krol, Chem. Sci., 2017, 8, 3712–3719.
27. Z. Wang, G. Liu, C. Ding, Z. Chen, F. Zhang, J. Shi and C. Li, J. Phys. Chem. C, 2015, 119, 19607–19612.
28. B. Klahr, S. Gimenez, F. Fabregat-Santiago, J. Bisquert and T. W. Hamann, J. Am. Chem. Soc., 2012, 134, 16693–16700.
29. C. A. Mesa, L. Steier, B. Moss, L. Francas, J. E. Thorne, M. Gratzel and J. R. Durrant, J. Phys. Chem. Lett., 2020, 11, 7285–7290.
30. J. Yuan, Y. Yuan, J. Zhang, H. Xu, Z. Mao and Y. Ma, ChemSusChem, 2022, 15, e202102313.
31. S. R. Pendlebury, A. J. Cowan, M. Barroso, K. Sivula, J. Ye, M. Grätzel, D. R. Klug, J. Tang and J. R. Durrant, Energy Environ. Sci., 2012, 5, 6304–6312.
32. Y. Surendranath, D. A. Lutterman, Y. Liu and D. G. Nocera, J. Am. Chem. Soc., 2012, 134(14), 6326–6336.
33. H. N. Nong, L. J. Falling, A. Bergmann, M. Klingenhof, H. P. Tran, C. Spöri, R. Mom, J. Timoshenko, G. Zichittella, A. Knop-Gericke, S. Piccinin, J. Pérez-Ramírez, B. R. Cuenya, R. Schlögl, P. Strasser, D. Teschner and T. E. Jones, Nature, 2020, 587, 408–413.
34. M. Barroso, S. R. Pendlebury, A. J. Cowan and J. R. Durrant, Chem. Sci., 2013, 4, 2724.
35. J. E. Thorne, J.-W. Jang, E. Y. Liu and D. Wang, Chem. Sci., 2016, 7, 3347–3354.
36. J. Lee, D. Seo, S. Won and T. D. Chung, Sustainable Energy Fuels, 2021, 5, 501–508.
37. S. Chen, J. Bai, X. Nurimaimaiti, J. Wang, Y. Zhang, T. Zhou, J. Li and B. Zhou, Nano Energy, 2020, 78, 105396.
38. T. Hisatomi, H. Dotan, M. Stefik, K. Sivula, A. Rothschild, M. Gratzel and N. Mathews, Adv. Mater., 2012, 24, 2699–2702.

---

Footnote

† Electronic supplementary information (ESI) available: Detailed experimental section and characterization methods. See DOI: https://doi.org/10.1039/d2sc05802b

---