In situ charge carrier dynamics of semiconductor nanostructures for advanced photoelectrochemical and photocatalytic applications

3.1 TiO₂ photoanodes

Since the first report of the photoactivity property of TiO₂ by Fujishima and Honda [52], this photoelectrode archetype has been thoroughly studied over the past few decades. TiO₂ photoelectrodes have several advantages, such as having a high stability, a low cost, and being environmental-friendly. Therefore, TiO₂ has been widely used in various PEC applications, including PEC water splitting [53], [54], [55], [56], [57], [58], PEC fuel cells [59], PEC dye degradation [60], [61], [62] and PEC sensing [63], [64], [65], [66]. Cowan et al. reported the charge carrier dynamics within a TiO₂ photoanode under PEC operating conditions [38]. Nanocrystalline rutile TiO₂ films (nc-TiO₂) were deposited on the conductive fluorine-doped tin oxide substrate. A linear-sweep voltammetry measurement revealed that the onset potential of anodic photocurrent was approximately −0.8 V (vs. Ag/AgCl). The photocurrent reached a saturated value of 50 μA at −0.5 V. In situ TA spectroscopy was used to quantitatively analyze the carrier dynamics, providing a better understanding of charge transfer scenarios during the PEC reaction. In Figure 2A, the TA data were collected at 10 μs and 5 ms after the excitation of a 355 nm UV light (pump) at −0.9 and 0 V. At −0.9 V (a potential less anodic than the onset potential), a broad featureless absorption decay within a 10 μs to 5 ms timescale was observed. Under UV excitation, the absorption band of the photoholes and trapped photoelectrons were located at 450 [67], [ 68] and 770 nm [69], respectively. Therefore, the overlapped absorption of the photoholes and trapped photoelectrons led to the broad featureless absorption decay. However, at a largely anodic bias (0 V), a strong absorption band around 460 nm was found, which could be attributed to the long-lived photoholes. A similar spectral feature attributable to the photoholes was also observed on platinized nc-TiO₂ (Pt-nc-TiO₂) [70]. This absorption can still be observed at 5 ms after the UV excitation. In contrast, the absorption of trapped photoelectrons around 800 nm completely decayed within 5 ms after UV excitation. In Figure 2B, the decay profile was probed at 460 nm at 0 V. This decay profile can be fitted by two decay components: (1) power law decay and (2) stretched exponential decay. The fast component (1) can be attributed to electron–hole recombination, while the slow component (2) can be assigned to the oxidation reaction between photoholes and water at the photoelectrode surface. The slow component has a lifetime of 30 ms, suggesting that the formation of long-lived photoholes is a prerequisite for the water splitting reaction [71]. From the inset of Figure 2B, the TA profile of the photoholes showed a longer lifetime than the profile of the trapped photoelectrons. Meanwhile, the profile of the trapped photoelectrons was highly similar to the fast decay component of the TA profile of photoholes. Therefore, it can be concluded that the early stage of photoholes absorption is dominated by electron–hole recombination, and the oxidation reaction between photoholes and water contributes to the late stage of the decay profile. In Figure 2C, the absorption of trapped photoelectrons gradually increased when the applied bias was anodically shifted from −0.9 to 0 V. The electron density in the interband trap states decreased under a more anodic bias condition, and this decrease contributed to the prolonged lifetime of the trapped photoelectrons. The decay profiles of photoholes were also recorded under different applied biases. As shown in Figure 2D, these profiles again displayed two decay components. At a potential more anodic than −0.5 V, no appreciable change in the TA profile of photoholes could be found. This feature was consistent with the observed photocurrent saturation at −0.5 V. Noticeably, the electron–hole recombination (fast component) showed a decreased rate at a more anodic bias. This would result in a larger proportion of photoholes that could react with water, therefore enhancing the overall water oxidation efficiency. Figure 2E depicts the plausible charge dynamics pathways for an nc-TiO₂ photoanode. After UV excitation, the photoholes would either recombine with the trapped photoelectrons or live long enough to react with water. Under an anodic bias condition, the trapped photoelectrons would be detrapped and transfer to the counter electrode, preventing them from recombining with photoholes. Meanwhile, the remaining photoholes would further react with the water, undergoing a water oxidation reaction.

Figure 2:

(A) In situ transient absorption (TA) spectra of nc-TiO₂ recorded at −0.9 and 0 V in 0.05 M NaOH at 10 μs and 5 ms after UV excitation (355 nm, 250 μJ cm⁻²). The Pt-nc-TiO₂ spectrum was adopted from the literature [70]. (B) TA profile of the photoholes (460 nm) at 0 V. Inset highlights the difference in the TA profile between the photoholes (460 nm) and trapped photoelectrons (800 nm). (C–D) TA profiles of trapped photoelectrons (C panel) and photoholes (D panel) under different applied biases. (E) Proposed charge dynamics mechanism of nc-TiO₂. Reprinted with permission from a study by Cowan et al. [38]. Copyright 2010 American Chemical Society. (F) TA profile of photoholes for nc-TiO₂ (460 nm) at 0 V in 0.5 M NaClO₄ under UV excitation (355 nm, 35 μJ cm⁻²) and its correlation with the fractional yield of photoholes. (G) TA profiles of photoholes under different applied biases. (H) Correlation of the two η _sep values derived from incident photon-to-current conversion efficiency (IPCE) and TA measurements with the applied bias under UV excitation (355 nm and 3.7 mW cm⁻² for IPCE; 355 nm and 35 μJ cm⁻² for TA). Reprinted with permission from a study by Cowan et al. [40]. Copyright 2013 Royal Society of Chemistry.

Another study conducted by Cowan et al. researched the reasons for the low incident photon-to-current conversion efficiency (IPCE) of nc-TiO₂ by analyzing the in situ TA data [40]. An equation commonly used to evaluate the factors affecting IPCE value can be shown by η IPCE ( λ ) = η LH ( λ ) η sep ( λ ) η col ( λ ) . Here, light harvesting efficiency (η _LH) can be measured by UV–visible spectroscopy, which was 0.35 for an nc-TiO₂ electrode with a thickness of 1.1 μm. The charge collection efficiency ( η col ( λ ) ) is approximated to be ∼1 by considering that the electron diffusion length is larger than the photoanode thickness under a sufficiently large applied bias [72]. The charge separation efficiency ( η sep ( λ ) ) can then be derived from the equation. In Figure 2F, the TA profile of photoholes probed at 460 nm can be well fitted to a stretched biexponential function ( Δ A ( t ) = A 1 e ( − t / τ 1 ) b 1 + A 2 e ( − t / τ 2 ) b 2 ). The absorption decay profiles of photoholes under various applied biases are further shown in Figure 2G. The fractional yield of photoholes, defined as [h ⁺] _t /[h ⁺] _t=0, was also computed from the slow decay component to represent the empirical value of η _sep. As the applied bias anodically shifted from −0.185 to −0.060 V (vs. Ag/AgCl), the magnitude of the fast component increased due to a decrease in the electron–hole recombination rate. The suppression of the electron–hole recombination ensures the survival of long-lived photoholes, which are beneficial for water oxidation reactions. The magnitude of the slow component also increased as the photoanode was anodically biased from −0.185 to −0.060 V. At a bias more anodic than −0.060 V, no significant change in the TA signal was observed, which was again consistent with the photocurrent saturation recorded at −0.060 V. In Figure 2H, the good agreement between the empirical η _sep from TA profiles and the computed η _sep from the IPCE measurements demonstrated the validity of both approaches for determining η _sep. In both experiments, η _sep was found to increase with the applied bias from −0.185 to −0.06 V prior to plateauing at a bias more anodic than −0.06 V. This result demonstrates that the charge separation efficiency is the main reason accounting for the photocurrent generation in a PEC water splitting reaction.

In order to enhance the PEC performance of TiO₂, numerous strategies have been reported, such as foreigner element doping [55], [ 73], metal particle deposition [53], [ 56], type-II/Z-scheme heterojunction introduction [54], [57], [74], and photosensitizer incorporation [75], [ 76]. Pesci et al. studied oxygen-deficient, hydrogen-treated rutile TiO₂ (H:TiO₂) nanowire photoanodes [39] and conducted in situ TA measurements under various biases to establish a plausible charge dynamics mechanism. Figure 3A shows the band structure of air-annealed (A:TiO₂) and H:TiO₂. After the hydrogen treatment, the concentration of oxygen vacancies (V_O) increases. Note that the energetic states of V_O lie within the bandgap, which are around 0.75 and 1.18 V below the conduction band edge. These V_O states can serve as electron donors (EDs) to increase the carrier concentration. In the PEC system under an applied bias, the increased donor concentration can also assist in suppressing electron–hole recombination because the photovoltage drop across the depletion layer is greatly enhanced. In Figure 3B, the linear-sweep voltammograms are in a good agreement with the TA amplitude change probed at 500 nm and 10 ms after excitation under various biases. Since photoholes play a decisive role in the water oxidation activity, the results of Figure 3B further validate that the TA signals in the 425–550 nm region were contributed to by the photohole dynamics of rutile TiO₂. Figure 3C compares the in situ TA profiles between H:TiO₂ and A:TiO₂. At −0.85 V (vs. Ag/AgCl), both H:TiO₂ and A:TiO₂ showed power law decay kinetics with a short carrier lifetime and indistinguishable kinetics profiles for the photoholes (500 nm) and trapped photoelectrons (800 nm) at the ms–sec timescale. This feature is characteristic of the pronounced electron–hole recombination. As the applied bias shifts anodically to −0.6 V, there is a relatively large increase in the absorbance for H:TiO₂ at the early timescale (1–10 µs). Such an absorbance increase reflects an increased yield of photoelectrons and photoholes, indicating the suppression of fast recombination under an anodic bias. In contrast, A:TiO₂ shows a minimal change in the charge carrier yield as the applied bias is anodically varied. This outcome suggests that the early charge separation process is essential to enhancing the photoactivity of H:TiO₂. Another conspicuous feature from Figure 3C is that the decay kinetics of photoelectrons and photoholes are decoupled at the ms–sec timescale. This observation indicates that processes other than electron–hole recombination can occur under anodic bias conditions. In Figure 3D, the TA profiles of photoholes (500 nm) and trapped photoelectrons (800 nm) at −0.6 V for H:TiO₂ and A:TiO₂ are compared. For H:TiO₂, the trapped photoelectrons decay by more than 50% in the early stage (10 μs to 1 ms), whereas the photoholes remain almost unchanged at the same timescale. The fast decay of trapped photoelectrons follows a power law kinetics model (τ = 14 ms) and is attributed to the electron transfer to the external circuit. However, the slow decay of photoholes can be well fitted with a single stretched exponential function (τ = 0.15 s) and is assigned to the hole injection into the electrolyte. This timescale is at the same magnitude with the photohole lifetime (τ = 0.03–0.4 s) required for effective water oxidation on anatase TiO₂ photoelectrodes [77]. Conversely, on A:TiO₂, the photohole decay profile at −0.6 V can be fitted by a combination of stretched exponential and power law functions, implying there are at least two probable photohole transfer pathways. The fast decay component is assigned to electron–hole recombination, whereas the slow decay component is attributed to a photohole injection into the solution. These two events compete with one another and deteriorate the water oxidation activity.

Figure 3:

(A) Proposed charge dynamics mechanism of A:TiO₂ and H:TiO₂ at an anodic bias after UV excitation. k _ct, k _r and k _et respectively stand for the rate constants of the charge transfer into the solution, the charge recombination, and the electron transfer. E _1O and E _2O respectively denote the two states of V_O at 0.75 and 1.2 eV below the conduction band edge. (B) Linear-sweep voltammograms and in situ transient absorption (TA) amplitude change at 500 nm and 10 ms after UV excitation (λ = 355 nm, 70 μJ cm⁻²) for A:TiO₂ and H:TiO₂ in 1 M NaOH under various applied biases. (C) TA profiles of A:TiO₂ and H:TiO₂ at different applied biases. (D) TA decay profiles of photoholes (500 nm) and trapped photoelectrons (800 nm) for H:TiO₂ and A:TiO₂ at −0.6 V. Reprinted with permission from a study by Pesci et al. [39]. Copyright 2013 American Chemical Society.

3.2 BiVO₄ photoanodes

BiVO₄ is an n-type semiconductor with a narrow bandgap (E _g = 2.4 eV) that is suitable for visible light absorption. In addition, the relatively cathodic valence band edge (E _VB = +2.5 V vs. NHE) [78] is able to drive many oxidation reactions. However, BiVO₄ suffers from a short hole diffusion length that makes the overall solar-to-fuel efficiency low. Zhang et al. reported a black phosphorene (BP)–deposited BiVO₄ (BP/BiVO₄) nanostructure photoanode, in which the BP layer could facilitate photohole extraction as a result of the p–n band alignment, promoting the overall water oxidation efficiency. The morphology of the prepared BiVO₄ and BP/BiVO₄ nanostructures is revealed in Figure 4A, showing a thickness of approximately 1 µm. In order to elucidate the veritable charge dynamics of BP/BVO, in situ TA measurements were performed. Note that the TA signals of the photoholes of BP/BVO are situated around 400–700 nm [79]. To acquire information on the photohole dynamics, the TA profiles were probed at 500 nm under open-circuit potential (OCP) and 0.8 V (vs. Ag/AgCl). In Figure 4B and C, the collected TA spectra are fitted by a biexponential decay model ( Δ A ( t ) = A 1 e ( − t / τ 1 ) + A 2 e ( − t / τ 2 ) ) composed of a fast decay component (τ ₁) and a slow decay (τ ₂). The corresponding fitting parameters are summarized in Figure 4D. The fast component (τ ₁) can be ascribed to the trapping of photoholes at the band edge of BiVO₄, and the slow component (τ ₂) can be attributed to photohole trapping at the electrode/electrolyte interface. Under an applied bias of 0.8 V, both pure BiVO₄ and BP/BiVO₄ displayed a reduced τ ₁, which can be understood by the fact that an applied bias can facilitate photohole transportation. The τ ₂ value of pure BiVO₄ remained unchanged at 0.8 V, while the τ ₂ value of BP/BiVO₄ doubled. The twofold increase of τ ₂ indicated that the photoholes were efficiently extracted from BiVO₄ to the electrode/electrolyte interface by BP deposition. The increased yield of the photoholes was expected to promote the water oxidation performance. The charge separation efficiencies (η _sep) were also measured and shown in Figure 4E, which showed a 4.4% enhancement via BP deposition under 1.23 V (vs. RHE). This result further corroborates the beneficial function of BP as a photohole extractor.

Figure 4:

(A) SEM images of BiVO₄ and black phosphorene (BP)/BiVO₄. (B–C) In situ TA profiles of the photoholes (500 nm) for BiVO₄ and BP/BiVO₄ at OCP (B panel) and 0.8 V (C panel) in 0.5 M KPi. (D) Fitting parameters of the recorded TA profiles. (E) η _sep of BiVO₄ and BP/BiVO₄ at different applied biases. Reprinted with permission from a study by Zhang et al. [42]. Copyright 2019 Springer Nature.

Another strategy for enhancing the solar-to-fuel efficiency of BiVO₄ is to introduce a type-II band alignment, which can achieve a pronounced charge carrier separation. WO₃-modified BiVO₄ (WO3/BiVO₄) is a common type-II heterostructure model studied in PEC water splitting systems. The band alignment allows photoelectrons to flow from the BiVO₄ to the WO₃ domain, leaving the photoholes in the BiVO₄ domain to conduct a water oxidation reaction. Selim et al. [80] reported the interfacial charge dynamics of WO₃/BiVO₄ by using in situ TA spectroscopy. Figure 5A and B shows the TA spectra under 1.23 V (vs. RHE) and OCP at 10 µs and 10 ms after excitation. Note that the samples were excited from the BiVO₄ side with the aim of exclusively studying the dynamics of the charge transfer from BiVO₄ to WO₃. Under 1.23 V, the TA profile of pure BiVO₄ exhibits a maximum absorption at 550 nm, whereas the TA profile of WO₃/BiVO₄ shows an additional absorption band at a long wavelength that disappears by 10 ms. Previous work has shown that the photoelectrons of WO₃ exhibit a broad absorption around 800 nm [45], [ 81]. Because WO₃ was not excited in the current TA setup, the photoelectron signal of WO₃ at a long wavelength region was attributed to the photoelectron transfer from BiVO₄ to WO₃ at a µs timescale. However, at the OCP there was no noticeable difference in the TA decay between BiVO₄ and WO₃/BiVO₄. Importantly, the lack of a photoelectron signal for WO₃/BiVO₄ suggested that an applied bias was necessary for prompting the photoelectron transfer from BiVO₄ to WO₃. In Figure 5C and D, the TA profiles of photoholes (500 nm) of pure BiVO₄ and WO₃/BiVO₄ were compared under different applied biases. These kinetics traces could be fitted with a combination of a power law and a single exponential function. The photohole yield increased with an increasing anodic bias, with WO₃/BiVO₄ showing a more pronounced enhancement. Note than the anodic bias could create a wider depletion layer at the electrode/electrolyte interface, which increased the band bending to promote photohole accumulation at the photoanode surface. For pure BiVO4, which showed biphasic TA kinetics, two pathways of photohole dynamics were considered: trap-mediated bimolecular recombination at an early timescale (μs–ms) and water oxidation process coupled with back electron–hole recombination at a slow timescale (ms–sec). Compared with BiVO₄ (τ = 0.18 s at 1.2 V), WO₃/BiVO₄ showed an increased lifetime of photoholes (τ = 0.29 s at 1.2 V), demonstrating the suppression of bimolecular recombination by the introduction of WO₃. The plausible charge dynamics mechanism is shown in Figure 5E and F. In pure BiVO₄, photoholes suffer from bimolecular recombination, and thus the water oxidation performance is limited. With the introduction of WO₃, photoelectron transfer from BiVO₄ to WO₃, prohibiting charge carriers from bimolecular recombination. This results in the formation of long-lived photoholes at BiVO₄, which has been considered as a requisite for driving efficient water oxidation on BiVO₄ [82].

Figure 5:

(A, B) In situ transient absorption (TA) spectra of BiVO₄ and WO₃/BiVO₄ at 1.23 V (A panel) and open-circuit potential (OCP) (B panel) in a 0.1 M phosphate buffer recorded at 10 μs and 10 ms after UV excitation (355 nm, 500 μJ cm⁻²). (C, D) TA profiles of photoholes (500 nm) of BiVO₄ and WO₃/BiVO₄ at different applied biases. (E, F) Proposed charge dynamics mechanism of BiVO₄ and WO₃/BiVO₄, in which (i) represents water oxidation at the BiVO₄ surface, (ii) denotes the electron transfer from BiVO₄ to WO₃ and (iii) stands for the photoelectron extraction from the photoanode to the external circuit. Reprinted with permission from a study by Selim et al. [80] Copyright 2013 Royal Society of Chemistry.

3.3 WO₃ photoanodes

WO₃ is another well-reported n-type semiconductor for use as a photoanode in PEC water splitting thanks to the visible light-responsive band structure (E _g = 2.5 eV). However, the fast charge recombination at the surface defect sites (W⁵⁺ and surface states) hinders the overall PEC performance. Several strategies have been proposed to address this issue, such as hydrogen treatment [83], [ 84], co-catalyst deposition [85], [86], [87] and passivation layer introduction [88], [ 89]. Kim et al. reported an efficient surface passivation approach for decreasing the photoelectron trapping event for WO₃ using an Al₂O₃ overlayer [44]. The photoanode was fabricated by depositing a 5.3-nm-thick Al₂O₃ layer on WO₃ thin films (WO₃/Al₂O₃). The deposited Al₂O₃ not only promoted photoelectron transfer to the external circuit by passivating the electron traps but also boosted the water oxidation kinetics by arousing the photohole trapping process. The photoelectron trapping event was first examined with a steady-state absorption spectroscopy under two distinct applied biases, one above the photocurrent onset (1.3 V [vs. Ag/AgCl]) and the other one below the photocurrent onset (0 V [vs. Ag/AgCl]). In Figure 6A, the absorbance change at 800 nm of pure WO₃ and Al₂O₃/WO₃ was measured under a bias condition. Here, the absorption at 800 nm could be attributed to photoelectron trapping signal of WO₃ [90], which remained unchanged at 1.3 V, despite the deposition of Al₂O₃. However, at 0 V there was a notable absorbance increase indicative of photoelectron accumulation on both pure WO₃ and Al₂O₃/WO₃. Importantly, the photoelectron accumulation of WO₃ (absorbance at 800 nm) was largely reduced upon the deposition of Al₂O₃. This result indicates that Al₂O₃ could passivate the electron traps of WO₃ to promote photoelectron transfer to the external circuit. This would further mediate the photohole dynamics of WO₃ to affect the overall PEC performance.

Figure 6:

(A) Absorbance change at 800 nm for pure WO₃ and Al₂O₃/WO₃ at different applied biases in Ar-purged 0.1 M HClO₄. (B) Transient absorption (TA) profiles of photoholes (480 nm) at open-circuit potential (OCP) and 1.3 V after UV excitation (355 nm, 250 μJ pulse⁻¹). (C) TA spectra of pure WO₃ and Al₂O₃/WO₃ at 1.3 V and 2 μs after UV excitation. A, B represents the subtracted spectrum. (D) Proposed charge dynamics mechanism of pure WO₃ and Al₂O₃/WO₃ under a bias condition. (E) Deconvoluted TA spectra of pure WO₃ and Al₂O₃/WO₃ at 1.3 V and 2 μs after UV excitation. Reprinted with permission from a study by Kim et al. [44]. Copyright 2013 Royal Society of Chemistry.

In situ TA spectroscopy was further performed to investigate the photohole dynamics. In Figure 6B, the TA profiles of trapped photoholes (480 nm [81]) at OCP and 1.3 V for pure WO₃ and Al₂O₃/WO₃ were compared. The yield of the trapped photoholes (i.e., the TA intensity) of Al₂O₃/WO₃ was largely increased at 1.3 V, while the increased photohole yield of pure WO₃ was less pronounced. This comparison suggests that Al₂O₃ passivation can enhance the number of trapped photoholes for WO₃. To confirm this argument, the TA spectra of pure WO₃ at 1.3 V and 2 µs after UV excitation were subtracted from those of Al₂O₃/WO₃ to reveal the photohole population as a result of the Al₂O₃ passivation. As Figure 6C shows, the spectral subtraction (A, B) reveals a substantial increase in the photohole trapping signal at 480 nm. This outcome corroborates that the increased photohole trapping also contributed to the better PEC performance of Al₂O₃/WO₃. Figure 6D depicts the proposed charge dynamics mechanism for pure WO₃ and Al₂O₃/WO₃. For pure WO₃, trapped photoelectrons dominated the charge transfer dynamics to hinder the PEC performance. Upon Al₂O₃ deposition, the electron traps of WO₃ could be passivated, accompanied by an increased population of trapped photoholes, both of which were beneficial for PEC water oxidation. In Figure 6E, the deconvoluted TA spectra could be utilized to quantify the number of trapped photoelectrons and photoholes for pure WO₃ and Al₂O₃/WO₃. The trapped photohole population of Al₂O₃/WO₃ was three times higher than that of pure WO₃. On the other hand, the number of trapped photoelectrons of Al₂O₃/WO₃ was merely 20% of the number of Al₂O₃/WO₃. This result confirmed the benefit of Al₂O₃ deposition for enhancing the PEC water oxidation activity of WO₃ by mediating interfacial charge dynamics.

3.4 α-Fe₂O₃ photoanodes

α-Fe₂O₃ (hematite) is a well-reported promising photoanode material with several advantages. The bandgap of α-Fe₂O₃ (E _g = 2.1 eV) [91] is capable of visible light absorption, and the valence band edge (E _VB = +2.38 V vs. NHE) [92], [ 93] is thermodynamically suitable for water oxidation reactions. However, there are several intrinsic properties that deteriorate the solar-to-fuel efficiency of α-Fe₂O₃. The drawbacks include limited photon harvesting due to the high light penetration depth [94], and the incapability of hydrogen production because of an unfavorable conduction band edge (E _CB = +0.28 V vs. NHE) [95], as well as sluggish water oxidation kinetics resulting from a short photohole diffusion [96]. The issue of photohole dynamics is particularly critical, thus invigorating broad interests in studying the charge dynamics of α-Fe₂O₃. Barroso et al. [47] reported a perspective that summarized the recent studies on interfacial charge dynamics for α-Fe₂O₃ by employing in situ TA spectroscopy. Figure 7A shows the in situ TA spectra at 800 nm for Si-doped α-Fe₂O₃ photoanode. Under 0.5 V (vs. RHE) (around the flat-band potential), the spectrum exhibits a positive absorption at 580 nm, which rapidly decays within a few ms due to fast electron–hole recombination. As the applied bias anodically shifts from 0.8 to 1.6 V, the TA lifetime notably increases by three orders of magnitude. A broad, long-lived absorption centered at approximately 650 nm and a narrow, negative bleaching at 580 nm can also be noticed. To gain insights into these origins, TA profiles at different applied biases were further probed at 650 and 580 nm. In Figure 7B, the TA absorption at 650 nm exhibits biphasic decay kinetics at a bias more anodic than 0.8 V. A fast decay component is observed at a timescale <10 ms, followed by a slow decay feature beyond 10 ms. The fast decay component can be attributed to electron–hole recombination, which is suppressed under more anodic bias conditions. The depressed charge recombination is due to the formation of a depletion layer rendered by the applied bias. However, the slow decay term can be attributed to the formation of long-lived photoholes. In Figure 7C, the TA bleaching at 580 nm can be ascribed to the electronic transition from the valence band to the shallow localized states below the conduction band. Note that the negative bleaching reflects a photoelectron trapping event at these intraband states, while the positive absorption corresponds to a photohole trapping process. At 0.5 V, a nearly unbiased condition, the intraband states are filled with electrons and thus function as photohole traps. Upon UV excitation, the rapid photohole trapping process induces an ESA, producing positive TA signals at 580 nm observed at 0.5 V. Under anodic bias conditions (0.8 to 1.6 V), these intraband states are oxidized to become electron traps. The trapping of photoelectrons on the intraband states then give rise to negative TA bleaching signals.

Figure 7:

(A) In situ transient absorption (TA) spectra of Si-doped α-Fe₂O₃ photoanode at different applied biases after UV excitation (355 nm, 200 μJ cm⁻²). (B, C) TA profiles at 650 nm (b panel) and 580 nm (C panel) at different applied biases. (D) Origins of the two TA signals at 650 and 580 nm. (E) Correlation between the photocurrent and TA amplitude at 650 and 580 nm at 100 ms after UV excitation under different applied biases. (F) Dependence of the square of the TA amplitude at 580 nm and 100 μs after UV excitation on the applied bias. (G) Proposed charge dynamics mechanism of α-Fe₂O₃ under different applied biases. Reprinted with permission from a study by Barroso et al. [47]. Copyright 2013 Royal Society of Chemistry.

The origins of the two TA signals at 650 and 580 nm are depicted in Figure 7D. Attention was first given to the long-lived absorption feature at 650 nm. Under an anodic bias condition, a depletion layer was formed at the photoanode surface, which facilitated charge carrier separation to enable photohole accumulation. The recorded long-lived TA absorption with a considerably long lifetime (>1 s) supported the argument. Figure 7E further shows the correlation between the amplitude of the slow decay term at 650 nm and the PEC activity of the α-Fe₂O₃ photoanode under different applied biases. The two traces exhibited a similar onset potential and a bias-dependent relation. This consistency confirmed that the slow decay term at 650 nm under an anodic bias resulted from the long-lived photoholes, which was decisive to the water oxidation activity. On the other hand, the TA bleaching at 580 nm under different applied biases was also analyzed. In Figure 7F, the amplitude of the bleaching at 580 nm was found to increase with the square root of the applied bias. This relation is similar to the square-dependence of the depletion layer thickness on the potential drop across the depletion layer [97]. Such a similarity demonstrates that the TA bleaching at 580 nm derived from charge carrier generation within the depletion layer. Figure 7G depicts the conceived charge dynamics mechanism of the α-Fe₂O₃ photoanode at different anodic biases. At 0.5 V, a nearly unbiased condition, the Fermi level of α-Fe₂O₃ is close to the conduction band edge, resulting in electron donation to the intraband trap states. Upon photoexcitation, the photoholes are rapidly captured by these states (<1 µs), resulting in the positive TA absorption signals at 580 nm. This TA absorption rapidly decays as the photoelectrons are subject to recombination with the trapped photoholes. At 1.1 V, a moderate bias, a depletion layer is formed at the photoanode surface, sweeping out the trapped photoelectrons and emptying the intraband trap states inside the depletion layer. Under this situation, the intraband states serve as electron traps, capturing photoelectrons (<1 ms) to cause TA bleaching. The trapped photoelectrons subsequently recombine with the photoholes on the ms timescale. At 1.6 V, a highly anodic bias, the strong built-in field drifted most of the trapped photoelectrons to the external circuit. Therefore, the photoholes at the valence band accumulate and diffuse to the photoanode surface, driving water oxidation reactions on the timescale of a few seconds.

4.1 Cu₂O photocatalysts

As a p-type semiconductor, Cu₂O offers a narrow direct bandgap (E _g = 2.0–2.2 eV) [100], [ 101] and an appropriate band structure (E _CB = −1.08 vs. NHE; E _VB = +0.93 V vs. NHE) [102] suitable for various solar-driven photocatalytic applications. Since Tennakone et al. [103] first demonstrated photocatalytic carbon dioxide reduction on Cu₂O in 1989, Cu₂O has been an appealing visible-responsive photocatalyst for converting greenhouse gases into value-added chemicals. However, the stability of Cu₂O has been a concern, since it can be easily oxidized or reduced by photoexcited charge carriers. Additionally, for p-type Cu₂O, the downward band bending at the Cu₂O/electrolyte would obstruct the photohole transfer to the surface, which would hinder the oxidation reaction. As depicted in Figure 8A, the overall photocatalytic performance is then restrained by electron–hole recombination induced by trapped photoholes. To achieve practical use of Cu₂O photocatalysts, appropriate surface modification to mediate interfacial charge dynamics is indispensable. Pastor et al. [98] reported a visible light-driven photocatalytic carbon dioxide reduction system by using RuO _x -deposited Cu₂O films (Cu₂O/RuO _x ) as a photocatalyst. After the deposition of cocatalyst RuO _x , the photoelectron lifetime of Cu₂O was prolonged and the carbon dioxide reduction performance was improved. As illustrated in Figure 8B, RuO _x was deposited on the surface of Cu₂O in order to extract photoholes from the valence band. The yield of long-lived photoelectrons can be increased due to the suppression of electron–hole recombination. In situ TA spectroscopy was performed for the purpose of realizing the charge carrier dynamics of Cu₂O/RuO _x . Pure Cu₂O film was first measured in an Ar-purged electrolyte under different experimental conditions (with the addition of an Na₂SO₃ as hole scavenger or AgNO₃ as electron scavenger). In Figure 8C, the TA spectra of Cu₂O without adding an electron/hole scavenger exhibited a broad negative bleaching from 475 to 750 nm and two positive absorption signals below 475 nm and beyond 800 nm. The magnitude of the positive signal was increased beyond 850 nm by the addition of an Na₂SO₃ hole scavenger and decreased by the introduction of AgNO₃ electron scavenger. Therefore, the absorption beyond 850 nm can be attributed to the photoelectrons of Cu₂O. Similarly, the absorption below 475 nm was ascribed to the photoholes of Cu₂O since it was largely enhanced by the addition of an AgNO₃ electron scavenger. Note that such electron/hole scavenging processes were ultrafast (<10 µs) because the corresponding TA signals already emerged at 10 µs after UV excitation.

Figure 8:

(A, B) Proposed charge dynamics mechanism of pure Cu₂O (A panel) and Cu₂O/RuO _x (B panel). (C) In situ transient absorption (TA) spectra of pure Cu₂O in an Ar-purged electrolyte with the addition of 0.1 M Na₂SO₃ or 0.1 M AgNO₃ at 10 and 100 μs after UV excitation (355 nm, 1.2 mJ cm⁻²). (D) TA profiles of photoelectrons (950 nm) for pure Cu₂O and Cu₂O/RuO _x in an Ar-purged electrolyte with the addition of 0.1 M Na₂SO₃. Reprinted with permission from a study by Pastor et al. [98]. Copyright 2014 Royal Society of Chemistry.

The effects of RuO _x deposition on the charge dynamics of Cu₂O were further examined. In Figure 8D, the TA profiles of photoelectrons at 950 nm for pure Cu₂O and Cu₂O/RuO _x are compared. For pure Cu₂O, the addition of Na₂SO₃ only caused a fairly limited increase in the TA amplitude, suggesting that photohole scavenging was ineffective as a result of a pronounced charge recombination. With the deposition of RuO_x, the TA amplitude was largely increased, regardless of Na₂SO₃ addition, indicating an effective suppression of charge recombination. It can be concluded that RuO _x deposition can serve as a hole acceptor to facilitate photohole transfer from Cu₂O to RuO _x . This charge dynamics mediation increases the yield of long-loved photoelectrons for Cu₂O (> 100 µs), expediting the oxidation reaction kinetics to improve the overall photocatalytic activity.

4.2 Dye-sensitized TiO₂ photocatalysts

Among the different semiconductor photocatalysts, TiO₂ is the most extensive photocatalyst for carbon dioxide reduction by virtue of its good stability, nontoxicity, and low price of production. Additionally, the intrinsic oxygen vacancy can activate carbon dioxide molecules in dark and irradiation conditions, making TiO₂ an attractive photocatalyst candidate for carbon dioxide reduction [104]. Kim et al. reported a tin porphyrin (SnP)-sensitized TiO₂ photocatalyst and studied the charge dynamics mechanism associated with hydrogen production by employing in situ TA spectroscopy [99]. The general reaction mechanism is plotted in Figure 9A. In this sensitized photocatalytic system, SnP is first excited by visible light to produce the triplet state ³SnP*, which then receives electrons from the ED, ethylenediaminetetraacetic acid (EDTA), to become SnP∙⁻. Afterward, an interfacial electron transfer from SnP∙⁻ to the conduction band of TiO₂ occurs. Finally, a hydrogen production reaction takes place at TiO₂ at the expense of the separated electrons. It is worth mentioning that SnP was not chemically attached to the TiO₂ surface. Therefore, whether or not the electrons of ED can swiftly transfer to the reactive sites of TiO₂ determines the photocatalytic activity. Figure 9B compares the in situ TA profiles spectra for SnP, SnP/EDTA, and SnP/EDTA/TiO₂ at 30 µs after visible excitation. SnP itself did not show substantial TA signals because of the fast charge recombination (<30 µs). In the presence of EDTA as ED, the collected TA profile showed a positive absorption signal at 680 nm, which could be assigned to SnP∙⁻ as a result of the electron transfer from ED to ³SnP*. This positive TA signal was depressed upon the addition of TiO₂, which could be accounted for by the electron transfer from SnP∙⁻ to TiO₂. Figure 9C further shows the TA profiles collected at 680 nm for the three cases. SnP itself decays within 10 µs. For SnP + EDTA, the decay of the TA absorption was significantly retarded, the degree of which became more pronounced with an increasing EDTA concentration (as shown in Figure 9D). This feature reflects the production of long-lived SnP∙⁻, which can survive for minutes. However, for SnP + EDTA + TiO₂, the TA absorption decay was accelerated to reach a lifetime <30 μs. Effective hydrogen production was achieved on the present SnP-sensitized TiO₂ system, even though SnP was not chemically attached to TiO₂. The success was made possible because the lifetime of SnP∙⁻ was sufficiently long to enable an interfacial electron transfer to TiO₂ for hydrogen production.

Figure 9:

(A) Proposed charge dynamics mechanism of SnP-sensitized TiO₂. (B) In situ transient absorption (TA) spectra for SnP, SnP + EDTA and SnP + EDTA + TiO₂ at 30 μs after visible excitation (532 nm). (C) TA profiles at 680 nm for SnP, SnP + EDTA and SnP + EDTA + TiO₂. (D) TA profiles at 680 nm for SnP + EDTA under different EDTA concentrations. Reprinted with permission from a study by Kim et al. [99]. Copyright 2010 Royal Society of Chemistry.