Origin of the overall water splitting activity of Ta3N5 revealed by ultrafast transient absorption spectroscopy† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c9sc00217k

A detailed transient absorption spectroscopy study efficiently correlates charge carrier dynamics with the overall water splitting efficiency in Ta3N5 photocatalyst.


Introduction
Considering ever increasing energy demand, there is an urgent need to promote and enhance the production of renewable energy sources. This approach will pay dividends in protecting our environment from a serious crisis. In this direction, using solar energy to generate H 2 from water using a photocatalytic process constitutes one of the methods. [1][2][3][4][5][6][7][8][9] Over the past decades, this approach has been intensively investigated to enhance the solar to hydrogen (STH) conversion efficiency. For the commercial use of the solar H 2 , a stable photocatalyst consistently demonstrating a higher STH efficiency is desired. To address this challenge, novel visible light absorbing photocatalysts are being synthesized and tested for photocatalytic water splitting activity.
Tantalum nitride (Ta 3 N 5 ) is an n-type, visible light absorbing nitride-based photocatalyst. The conduction band (CB) and valence band (VB) of Ta 3 N 5 are characterized by the Ta 5d and N 2p orbitals, respectively. 10,11 Ta 3 N 5 is commonly prepared by the nitridation reaction of tantalum (Ta) precursors such as Ta foil or tantalum oxide (Ta 2 O 5 ). The VB and CB positions in Ta 3 N 5 are thermodynamically suitable to drive water oxidation and reduction to generate O 2 and H 2 , respectively. Hence, Ta 3 N 5 is capable of overall water splitting (OWS) upon one-step photoexcitation. However, Ta 3 N 5 is usually known as a photoanode material with efficient O 2 evolution, 10,12-17 but not for either H 2 evolution or OWS ability.
Recently, OWS activity with an apparent quantum efficiency (AQE) of 70% is demonstrated in an ultraviolet light absorbing SrTiO 3 photocatalyst. 18 As a further advancement, Ta 3 N 5 forms one of the few promising visible light absorbing photocatalysts capable of OWS. Nevertheless, achieving OWS in Ta 3 N 5 prepared by nitriding Ta foil or Ta 2 O 5 still remains a challenge. Recently our group employed potassium tantalate (KTaO 3 ) as a Ta precursor for the nitridation reaction to prepare Ta 3 N 5 . In this approach, Ta 3 N 5 prepared by 0.25 hour nitridation of KTaO 3 demonstrated the highest OWS activity. This is the rst report on realizing visible light OWS activity without the need of sacricial reagents for Ta 3 N 5 . 19 However, the efficiency of OWS activity was found to decrease with increase in the nitridation time from 0.25 to 10 hour. These observations emphasize the need to elucidate the role of the Ta precursor and the nitridation time in determining the efficiency of OWS.
As stated earlier, the H 2 evolution in Ta 3 N 5 (prepared by the nitridation Ta foil or Ta 2 O 5 ) is apparently less known and typically smaller when compared to O 2 evolution. In a few reports that demonstrate H 2 evolution in Ta 3 N 5 , the size of Ta 3 N 5 particles was relatively small (between 5 to 8 nm (ref. 20) and in one case it is 2 nm (ref. 21)). In addition to the smaller size, Ta 3 N 5 particles were typically present in a composite form with either SiO 2 (ref. 20) or graphene oxide. 21 In these reports, the effect of size on the band bending/optical properties and the expected electronic interaction between the Ta 3 N 5 with the counterpart in the composite is not considered. In another report showing H 2 evolution, the surface of the Ta 3 N 5 was modied with magnesium layer. 22 Thus, without the modication of either structural/surface properties of Ta 3 N 5 or by inducing electronic interaction with another material, H 2 evolution was not realized. Based on these observations, it is agreeable to state that efficient generation of H 2 in Ta 3 N 5 is fundamentally less pronounced compared to O 2 . 10,21,23 However, the origin of this observation is unclear.
A key approach to examine the role of Ta precursor and the nitridation time on the efficiency of OWS is investigating the decay dynamics of photogenerated charge carriers. Whilst there has been extensive research/progress in the development of Ta 3 N 5 from the materials perspective, the dynamics of photogenerated charge carriers and its impact on the efficiency of the photocatalytic reaction have not been explored. The fundamental process in OWS in Ta 3 N 5 is the generation of electronhole pairs upon absorption of light. Then, these carriers must migrate towards the surface of the Ta 3 N 5 to get involved in the reaction with water. However, due to structural/electronic defects in Ta 3 N 5 , efficient charge transport towards the surface is affected by the undesirable trapping and/or recombination. To unveil these photoinduced processes in detail, femtosecond time-resolved diffuse reectance (fs-TDR) spectroscopy is employed, which is a pump-probe method. Here, excitation of Ta 3 N 5 with an ultrafast laser pulse leads to the formation of holes and electrons in the VB and CB, respectively. The dynamics of these transient species can be followed by monitoring their transient absorption spectrum using an optical probe pulse from visible to infrared. This technique has been widely used to examine charge carrier dynamics in various types of photocatalysts and to correlate the obtained photophysical parameters with that of photocatalytic efficiency. [24][25][26][27][28][29] Note that TDR is essentially similar to the transient absorption spectroscopy (TAS) except that the TA signal in the former is monitored in diffuse reection mode because of the opaque nature of the Ta 3 N 5 powder samples. The ultrafast dynamics of electrons and holes and their decay pathway was found to be strongly inuenced by the Ta precursor used to prepare Ta 3 N 5 and the nitridation time. Results from this systematic investigation help in understanding the structure-property relationship and provide rational insights on Ta 3 N 5 material design to improve the efficiency of OWS. As mentioned earlier, this work provides mechanistic insight on explaining OWS through charge carrier dynamics, while a comprehensive study of the morphological characterization and photocatalytic performance is reported elsewhere. 19 2. Results and discussion Fig. 1 shows the ground-state optical absorption spectra for the Ta 3 N 5 prepared by nitridation of Ta 2 O 5 or KTaO 3 in both visible and NIR region. The absorption peak at 545 nm (Fig. 1A) is attributed to the fundamental transition (N 2p to Ta 5d) in Ta 3 N 5 . Though the magnitude of absorption at 545 nm increases with nitridation time, the energetic position was not inuenced by the choice of Ta precursor or the nitridation time. As shown in Fig. 1B, an enhancement in the magnitude of NIR absorption is noticed with an increase in the nitridation time from 0.25 to 10 hour indicating the presence of defect states within the bandgap of Ta 3 N 5 . For Ta 3 N 5 prepared by 0.25 hour nitridation of KTaO 3 , noticing a virtually absent NIR absorption suggests a signicant reduction in the concentration of defect states. This conclusion is in good agreement with our previous study using scanning transmission electron microscopy (STEM), which revealed the single-crystal nature of this sample. 19 Previous reports have different opinions in relating the NIR absorption with the electronic nature of these defect states. 17,[30][31][32][33] Hence, we consider that the NIR absorption in Ta 3 N 5 is due to the formation of defects related to charged nitrogen vacancies and/or reduced Ta vacancy.
In the following parts, Section 2.1 will correlate the charge carrier dynamics in Ta 3 N 5 synthesized by the nitridation of Ta 2 O 5 and KTaO 3 (0.25 hour). In Section 2.2, the effect of KTaO 3 nitridation time on the charge carrier dynamics in Ta 3 N 5 will be presented. In the last Section 2.3, plausible approaches to enhance the efficiency of OWS in Ta 3 N 5 will be outlined. The efficiency of OWS was found to be highest in Ta 3 N 5 prepared by 0.25 hour nitridation of KTaO 3 . On the other hand, 0.25 hour nitridation of Ta 2 O 5 neither forms Ta 3 N 5 phase nor shows the OWS. To understand the origin of OWS activity, charge carrier dynamics is compared between the Ta 3 N 5 prepared by 0.25 hour nitridation of KTaO 3 and long-time nitridation of Ta 2 O 5 . We selectively monitor the dynamics of free electrons and VB holes at 3435 and 545 nm wavelengths, respectively. The responsible optical transitions probed at these wavelengths are depicted in Scheme 1. It has been consistently demonstrated that probing at 3435 nm (IR region) typically monitors the intra-band transition close to the CB levels. Thus, 3435 nm probe offers unique information on the dynamics of free/shallowly trapped electrons. 24,29,34,35 With regard to the hole dynamics, 545 nm probe is employed. As depicted in Fig. 1A, the absorption peak observed at 545 nm is due to the bandgap transition from N 2p (VB) to Ta 5d (CB). Monitoring the dynamics at 545 nm yields direct information on the occupancy of holes in the VB, where the recovery of the ground-state bleaching signal offers information on the hole dynamics.
2.1.1 The effect of Ta precursor on the free electron dynamics probed at 3435 nm. Fig. 2A shows the transients corresponding to free electrons (3435 nm probe) in Ta 3 N 5 prepared by the nitridation of KTaO 3 for 0.25 hour. In the Ta 3 N 5 prepared by the nitridation of KTaO 3 for 0.25 hour, there still exists a large amount of unreacted KTaO 3 phase. Hence, it is essential to disregard the possibility of free electron generation from the unreacted KTaO 3 . The onset in the optical absorption of KTaO 3 is z345 nm (z3.6 eV). The visible light excitation employed in this study does not allow bandgap transition in KTaO 3 to generate free electrons (Fig. S1 †). Hence, the free electron dynamics presented in Fig. 2A originates from the Ta 3 N 5 phase formed aer nitridation, not from the unreacted KTaO 3 .
Considering the TA signal is monitored in the diffuse reection mode in the current study, a direct comparison of the signal magnitude is not straightforward due to the likely differences in the scattering/surface properties of different types of powder samples under study. Hence, we primarily rely on the decay dynamics rather than the signal magnitude while comparing between two different samples. The TA signal magnitude at 3435 nm probe for Ta 3 N 5 prepared by 0.25 hour nitridation of KTaO 3 can be found in ESI (Fig. S2). † To understand the decay pathways of free electrons in Ta 3 N 5 (0.25 hour nitridation of KTaO 3 ), the effect of pump uence on the half lifetime, s 1/2 , (i.e., the time at which initial concentration of photogenerated electrons reach to half of its initial value) is compared. The s 1/2 value decreases from 200 to 5 ps upon increasing the uence from 0.014 to 0.12 mJ per pulse. A qualitatively similar observation is noticed also for hole dynamics probed at 545 nm ( Fig. S3 †). Thus, the dynamics of both electron and hole is very sensitive to pump uence from early ps time. In addition, matching of both free electron and hole This journal is © The Royal Society of Chemistry 2019 dynamics is noticed (Fig. S4 †). Fitting the transients in Fig. 2A to a second-order type recombination model showed good agreement ( Fig. S5 †). Taking these observations collectively, it is inferred that photogenerated carriers in Ta 3 N 5 (0.25 hour nitridation of KTaO 3 ) decay primarily via second-order type recombination of free (untrapped) electrons with holes. If the charge carriers decay via trapping to defect states (generally related to rst-order type recombination), uence-independent dynamics is expected, which is however not observed. Fig. 2B displays the normalized electron kinetics in Ta 3 N 5 prepared by the nitridation of Ta 2 O 5 . The TA signal magnitude at 3435 nm probe for Ta 3 N 5 prepared by nitridation of Ta 2 O 5 can be found in ESI (Fig. S2). † Despite increasing the pump uence by a factor of approximately nine, the electron dynamics in early 5 ps (Fig. 2B inset) is virtually unaffected suggesting the occurrence of electron trapping. A similar observation is noticed also at 580 nm pump (Fig. S6 †). The s 1/2 value in Fig. 2B decreases from 50 to 7 ps upon increasing the uence from 0.014 to 0.12 mJ per pulse. A matching between the free electron and hole dynamics was not observed (Fig. S7 †) and the transients could not be tted to the second-order type electron-hole recombination model (Fig. S5 †). Combining these observations, it is concluded that the major part of electrons in Ta 3 N 5 prepared by the nitridation of Ta 2 O 5 decays via trapping rather than second-order type direct electron-hole recombination.
2.1.2 Electron transfer process from Ta 3 N 5 prepared by 0.25 hour nitridation of KTaO 3 to the Rh cocatalyst. To reveal the mechanism of OWS, the effect of Rh cocatalyst loading and the water interface on charge carrier dynamics in Ta 3 N 5 prepared by 0.25 hour nitridation of KTaO 3 is carried out. Typically, metal cocatalyst particles are loaded on the surface of the photocatalyst to enhance the overall efficiency of photocatalytic reaction. In the case of Ta 3 N 5 , Rh metal cocatalyst particles are photodeposited on the surface by the photodeposition process. The Rh cocatalyst acts as an electron acceptor to facilitate the reduction reaction with water in generating H 2 . A faster electron (3435 nm probe) decay starting from z5 ps was noticed for Ta 3 N 5 prepared by 0.25 hour nitridation of KTaO 3 aer Rh cocatalyst loading (Fig. S8 †) suggests electron transfer process to the Rh cocatalyst. To further corroborate this process, the effect of water interface on the electron transfer to Rh cocatalyst is investigated. Considering water is prone to absorb IR probe light, 545 nm probe which corresponds to VB holes is employed. Fig. 3 compares the VB hole dynamics (545 nm probe) in the presence of water interface for Ta 3 N 5 prepared by 0.25 hour nitridation of KTaO 3 with and without Rh cocatalyst loading. Clearly, a longer hole lifetime is noticed from z5 ps aer loading Rh cocatalyst. As concluded from Fig. 2A, electrons decay majorly via recombination with VB holes. However, when electrons are transferred to Rh cocatalyst present on the surface, the number of electrons that recombine with VB holes is decreased. As a result, the efficiency of electron-hole recombination is reduced eventually resulting in a prolonged lifetime for VB holes due to slow recovery of the ground-state bleaching. A longer hole lifetimes is particularly benecial to promote water oxidation reaction to generate O 2 and to realise OWS. Observing electron transfer process corroborates the notion that electron trapping by the structural/electronic defects is an inefficient process, which is in agreement with the virtually absent ground-state NIR absorption (Fig. 1B) and the single-crystal nature of the Ta 3 N 5 Fig. 2 The fs-TDR time profiles corresponding to free electron dynamics probed at 3435 nm for Ta 3 N 5 nitrided from (A) KTaO 3 for 0.25 hour and (B) from Ta 2 O 5 . 440 nm pump is used to conduct these measurements. formed upon nitriding KTaO 3 for 0.25 hour as reported previously by using STEM. 19 2.1.3 On the inadequate H 2 evolution in Ta 3 N 5 prepared by the nitridation of Ta 2 O 5 . For Ta 3 N 5 prepared by the nitridation of Ta 2 O 5 , due to electron trapping within the bulk, efficient migration of electrons towards the surface cannot be anticipated. As a result, electron transfer to the Rh was indeed not observed (Fig. S8 †). Instead, hole transfer to Rh cocatalyst was observed as evidenced by the longer electron lifetime. We do not completely understand why electron transfer to the Rh cocatalyst was not observed here. From XPS study, the Rh cocatalyst was conrmed to be metallic but not as an oxide (Fig. S9 †). One of the plausible mechanisms is due to the occurrence of faster/efficient electron trapping is expected to reduce the electron mobility and the number of free electrons available for transfer to the Rh cocatalyst. This is the likely reason behind noticing inadequate H 2 evolution (compared to O 2 ) and consequently the absence of OWS activity in Ta  2.1.4 The effect of Ta precursor on the hole trapping process. Fig. 4 displays the TA spectra of Ta 3 N 5 prepared by the nitridation of KTaO 3 (for 0.25 hour) and Ta 2 O 5 in the visible region at different delay times starting from 0.6 to 200 ps. The TA spectra provide useful information on the presence of defect states within the bandgap. Before discussing these results, it is important to associate the spectral features with the possible transitions and/or defect states. Typically, the TA spectrum of Ta 3 N 5 shows three prominent features: (i) a bleaching (negative) signal around 545 nm, (ii) a second bleaching signal at 590 nm, and (iii) a broad positive TA signal between 650 to 750 nm. Let us discuss each one of this contribution.
The beaching signal at 545 nm is due to the depopulation of the ground state due to band gap transition ( Fig. 1A and Scheme 1) and is correlated to the VB free hole dynamics. The second bleaching (negative) signal at 590 nm probe cannot be explained unless considering the possibility of an optical transition from a lled defect state. To this end, the electronic transition from a positively charged nitrogen vacancies ðV N Þ to the CB is tentatively proposed. A previous prediction using rst principles density functional theory calculations is in good agreement with this assignment. 30,37 However, this transition is not discernible in Fig. 1A, which is likely due to low absorption coefficient typical to transitions involving defect states. By correlating the energetic positions of bleaching at 590 nm with 545 nm, the V N states are situated at 0.17 eV above the VB maxima. When holes from the VB are trapped to V N states, V N is converted to V N or V N , which reduces the density of V N states. As a result, possibility of the V N to the CB optical transition at 590 nm reduces and eventually shows bleaching signal (Fig. S10 †). We observe the 590 nm bleaching already at 0.6 ps delay time which indicates the ultrafast hole trapping process from the VB to V N . Thus, the presence of hole trapping process can be tentatively related to 590 nm bleaching. In an earlier report from Ziani et al. also observed the bleaching signal at 590 nm for Ta 3 N 5 (nitrided from Ta foil) thin lms. 38 In a recent microsecond TAS study, trapped holes are found to be probed at 590 nm in Ta 3 N 5 (nitrided from Ta 2 O 5 ). 34 Combining these observations, the presence of bleaching signal at 590 nm can be correlated to the formation of hole trapping nitrogen vacancies above the VB. The broad positive TA signal between 650 to 750 nm is tentatively assigned to shallowly trapped electrons (Fig. S11 †) and will be discussed in Section 2.2. Fig. 4A shows the TA spectra of Ta 3 N 5 prepared by the nitridation of KTaO 3 for 0.25 hour. The free VB hole dynamics shows time-dependent change as observed at 545 nm bleaching. The virtually absent second bleaching at 590 nm indicates that the density of V N defect states is too low to promote hole trapping. This conclusion is in good agreement with noticing a pronounced pump uence dependent free VB hole dynamics probed at 545 nm (Fig. S3 †). The occurrence of hole trapping is found to be an incompetent process in Ta 3 N 5 prepared by 0.25 hour nitridation of KTaO 3 . Fig. 4B shows the TA spectra of Ta 3 N 5 prepared by the nitridation of Ta 2 O 5 , which is different compared to Fig. 4A in terms of the second bleaching signal at 590 nm in addition to the one at 545 nm. Effect of pump uence on the dynamics of VB free holes (545 nm probe) is found to be less prominent particularly in early 5 ps time scale (Fig. S12 †). This observation suggests the presence of hole trapping V N defect states above the VB. Noticing the pronounced second bleaching signal at 590 nm with time indeed support the occurrence of hole trapping. In short, the presence of second bleaching at 590 nm indicates an enhancement in the formation of hole trapping V N defect states.
Conclusion from Section 2.1 is briey summarized as follows. Depending on the Ta precursor employed for the nitridation reaction to prepare Ta 3 N 5 , a distinctive electron and hole dynamics is rationalized. In Ta 3 N 5 prepared by the nitridation of Ta 2 O 5 , trapping of both electrons and holes is observed. In particular, due to efficient electron trapping, electron transfer to the Rh cocatalyst is not observed which is essential to realize H 2 evolution and consequently OWS activity.
In the case of Ta 3 N 5 prepared by the 0.25 hour nitridation of KTaO 3 , charge carriers are free and the decay is primarily via second-order recombination between CB electrons and VB holes. The decay of both electrons and holes by trapping is found to be ineffective. This conclusion is in good agreement with noticing virtually absent ground-state NIR absorption (Fig. 1B) and with the single-crystal nature of the Ta 3 N 5 without surface defects. 19 Hence, efficient migration of both electrons and holes toward the surface is ensured eventually prompting the OWS activity.

The effect of nitridation time on the charge carrier dynamics in Ta 3 N 5 prepared by the nitridation of KTaO 3
The efficiency of OWS is found to be high for 0.25 hour nitrided Ta 3 N 5 . However, the OWS activity was reduced by a factor of approximately ve with an increase in the nitridation time from 0.25 to 10 hour. 19 Hence, investigating the effect of nitridation time on the carrier dynamics is essential to understand the origin of this phenomena.
2.2.1 Comparing the free electron dynamics in Ta 3 N 5 prepared by 0.25 and 10 hour nitridation of KTaO 3 . The TA signal magnitude at 3435 nm probe for Ta 3 N 5 prepared by 10 hour nitridation of KTaO 3 can be found in ESI (Fig. S2). † As depicted in Fig. 5A, despite changing the pump uence by a factor of approximately nine, the free electron dynamics (3435 nm probe) in early 5 ps for Ta 3 N 5 prepared by 10 hour nitridation of KTaO 3 is barely affected. Comparing between Fig. 2A and 5A, a pronounced electron trapping in Ta 3 N 5 nitrided from KTaO 3 for 10 hour compared to 0.25 hour can be perceived. This notion is further corroborated by noticing: (i) a negligible inuence on the free electron dynamics upon loading Rh cocatalyst (Fig. S13 †), and (ii) a slow recovery of the VB free holes due to a reduction in the number of free electrons (because of trapping) available for recombination (Fig. S14 †). These observations indicate the occurrence of electron trapping in Ta 3 N 5 nitrided from KTaO 3 for 10 hour. This conclusion is further supported by noticing an enhancement in the groundstate NIR absorption (Fig. 1B) due to defect formation within the bandgap.

The effect of nitridation time on the hole trapping.
To investigate the effect of increasing the nitridation time on the hole dynamics, TA spectra of Ta 3 N 5 prepared by 10 hour nitridation of KTaO 3 is recorded (Fig. 5B). The presence of a second bleaching signal at 590 nm (which is absent in Fig. 4A) indicates the formation of hole trapping V N defect states. A similar observation is also noticed for Ta 3 N 5 prepared by 4 hours nitridation of KTaO 3 (Fig. S15 †). To rule out any potential contribution to the bleaching signal at 590 from the unreacted KTaO 3 , TA spectra of KTaO 3 is recorded. As expected, no such distinct bleaching signal at 590 nm reminiscent to that of Ta 3 N 5 is detected for KTaO 3 (Fig. S16 †).
To further rationalize the effect of nitridation time on the hole trapping process, the electron lifetime probed at 3435 nm is compared as a function of KTaO 3 nitridation time. As shown in Fig. 5C, the electron lifetime gets longer with an increase in the nitridation time. This observation can be explained as follows. Due to hole trapping, fewer holes are available for electrons to decay by recombination and hence a longer electron lifetime can be expected. Thus, from Fig. 5B and C, an enhancement in the efficiency of hole trapping process with an increase in the nitridation time is inferred.
Note that, we have not considered the effect of dark carrier (which depends on the relative position of Fermi level with respect to the CB and the defect density) on the decay dynamics of electrons and holes in Ta 3 N 5 prepared by the nitridation of KTaO 3 . This is due to experimental difficulty in determining the precise concentration of dark carrier or Fermi level position in these powder samples which is in a composite form with KTaO 3 . In this regard, a further systematic investigation of the effect of nitridation time on the relative shi in the Fermi level (with respect to the CB) and determining the corresponding dark carrier density would be helpful.
To put the results from Fig. 5 in perspective, it is instructive to understand the plausible mechanism by which defects are formed and their relation with the nitridation time. In this regard, a tentative model is proposed. The temperature employed during the nitridation reaction of KTaO 3 is 1173 K. The nitrogen source used for nitriding the Ta precursor is NH 3 gas, which decomposes into N 2 and H 2 under the experimental condition. Ideally, the Ta 5d orbital of the CB is present in Ta 5+ states, which ensures to maintain charge neutrality and avoid electron trapping. However, during the nitridation reaction, the time exposed to H 2 increases with nitridation time. This H 2 gas acts as a reducing agent to convert the Ta 5+ states to form Ta 4+ and/or Ta 3+ , which are proposed to act as electrons trapping centres. This notion is in agreement with a previous report demonstrating the formation of reduced Ta vacancies using synchrotron-excited X-ray photoelectron spectroscopy (XPS). 39 Similarly, the formation of nitrogen ðV N Þ vacancies which are associated to hole trapping centres becomes inevitable, particularly for nitridation time exceeding 0.25 hour. Maintaining the ideal Ta 3 N 5 stoichiometric ratio via synchronous substitution of nitrogen to oxygen (present in Ta precursor) is the key to control the defect formation.
Based on the conclusions drawn from Fig. 5, the formation of both electron and hole trapping centres within the bandgap of Ta 3 N 5 is pronounced with an increase in the nitridation time. The efficiency of charge transport from the bulk to the surface is negatively affected by carrier trapping. As a consequence, inef-cient electron transfer to the Rh cocatalyst is observed and the number of free carriers available at the surface of Ta 3 N 5 for photocatalytic reaction will be reduced. Due to these two combined effects, the efficiency of OWS is found to be reduced with increase in the nitridation time from 0.25 to 10 hour (Scheme 2). From our previous STEM study, the location of charge trapping defects in long time (>0.25 hour) nitrided Ta 3 N 5 is found to be predominantly on the surface of Ta 3 N 5 . 19 Hence, removing these surface defect states either by chemical treatment or by surface modication may further enhance the efficiency of OWS activity for long-time nitrided Ta 3 N 5 . A future work is necessary to unravel the electronic nature of the bulk and surface defect states in detail. The AQE value at 500 nm is ten times smaller compared to 420 nm. 19 Hence, one of the challenges is to enhance the AQE under visible light excitation. To discern whether the low AQE in visible light is due to inefficient charge generation, free electron dynamics (3435 nm probe) is compared between 580 and 440 nm excitation (Fig. S17 †). At 580 nm excitation, a similarity in terms of both carrier lifetime and pump uence dependence is noticed to that of 440 nm pump. In addition, electron transfer to the Rh cocatalyst is also noticed at 580 nm pump (Fig. S18 †). Besides that, Ta 3 N 5 is widely considered to have an indirect bandgap. Note that 580 nm pump is very close to the band edge of the Ta 3 N 5 , however, neither free electron lifetime nor electron photogeneration is affected. Hence, indirect bandgap nature of the Ta 3 N 5 does not sufficiently explain the lower activity under the visible light excitation. Alternatively, the absorption coefficient value of Ta 3 N 5 at 530 nm is lower by a factor of ve compared to 400 nm. 31,40,41 Upon 0.25 hour nitridation of KTaO 3 , the actual amount of Ta 3 N 5 phase formed on KTaO 3 is z2%. 19 Thus, a low AQE in the visible region is attributed to a combination of the lower absorption coefficient and the presence of only 2% of Ta 3 N 5 which can potentially absorb visible light, but not due to inefficient charge photogeneration. Thus, increasing the effective density of Ta  nitridation time as 0.25 hour) to expose the unreacted KTaO 3 surface in the direction of NH 3 ow forms one of the approaches.

Conclusions
A detailed ultrafast TAS study allowed us to elucidate a correlation between the synthesis procedurestructural defectscarrier dynamicsefficiency of OWS. Selectively probing the dynamics of electrons and holes allowed us to develop a phenomenological model that describes the role of Ta precursor employed to prepare Ta 3 N 5 and the effect of nitridation time in the context of OWS.
In addition to H 2 evolution being fundamentally low compared to O 2 , the OWS has not been realized for Ta 3 N 5 prepared by nitriding Ta 2 O 5 . The answer to this longstanding question was found to originate from the absence of electron transfer to the Rh cocatalyst by the virtue of efficient electron trapping.
When Ta 3 N 5 is prepared by employing KTaO 3 as a Ta precursor for nitridation reaction, OWS was observed for the rst time. For 0.25 hour nitrided Ta 3 N 5 , which shows the highest OWS activity, charge carriers decayed by second-order type electron-hole recombination but not by trapping. Due to single-crystal and defect-free nature of 0.25 hour nitrided Ta 3 N 5 , efficient migration of both electrons and holes towards the surface was achieved. Thus, to realize OWS, avoiding carrier trapping process was found to be indispensable.
When the nitridation time of KTaO 3 was increased from 0.25 to 10 hour, the efficiency of OWS reduced by a factor of approximately ve. This was attributed to the formation of defects which promoted both electron and hole trapping eventually affecting the efficient migration of charge carriers to the surface. These transient optical studies yielded key insights into the factors determining the efficiency of solar driven OWS by Ta 3 N 5 .

Synthesis of Ta 3 N 5 by the nitridation of KTaO 3
KTaO 3 particles were fabricated by a conventional solid-state reaction method. Ta 2 O 5 (99.9%; Kojundo Chemical Laboratory Co., Ltd.) and K 2 CO 3 (99.5%; Kanto Chemical Co., Inc.) were mixed at a Ta : K molar ratio of 1 : 1.05. Excess K was added to compensate for losses by volatilization at high temperatures. The mixture was thoroughly ground in an agate mortar for 90 min in the presence of a small amount of ethanol as a dispersant. Aer drying, the resulting mixture was transferred into an alumina crucible and calcined at 1173 K for 1 hour and then at 1423 K for 10 hour in static air. The KTaO 3 obtained in this manner was washed with ultrapure water at 343 K for 2 hour and centrifuged twice to remove any residual K 2 CO 3 . The powder was then completely dried by heating at 343 K overnight. Subsequently, the as-prepared KTaO 3 was subjected to a nitridation process to obtain Ta 3 N 5 grown on KTaO 3 . KTaO 3 (0.5 g) was transferred into an alumina tube and nitrided at 1173 K for 0.25, 4 and 10 hour under a ow of NH 3 gas at 100 mL min À1 .

Synthesis of Ta 3 N 5 by the nitridation of Ta 2 O 5
The commercially available Ta 2 O 5 (99.9%; Kojundo Chemical Laboratory Co., Ltd.) was subjected to the nitridation at 1173 K for 20 hours under a ow of NH 3 gas at 100 mL min À1 .

Photodeposition of a Rh cocatalyst on Ta 3 N 5
A Rh cocatalyst for H 2 evolution was loaded on Ta 3 N 5 photocatalysts by a previously-reported photodeposition process. 19 A Rh core was photodeposited using RhCl 3 $3H 2 O (Kanto Chemical Co., Inc.) as the metal precursor. This was accomplished by dispersing the photocatalyst powder in 150 mL of an aqueous methanol solution (10 vol%) containing the metal precursor. The pH of this solution was not adjusted but the temperature was maintained at 288 K by circulating cooling water. The suspension was evacuated to completely remove dissolved air and then exposed to visible light (l $ 420 nm) with continuous stirring. The photodeposition of Rh was conducted approximately for 3 hours.

Femtosecond time-resolved diffuse reectance (TDR) spectroscopy
In fs-transient diffuse reectance (fs-TDR) measurements, a femtosecond Ti:sapphire laser with a regenerative amplier (Spectra-Physics, Solstice, wavelength of 800 nm, pulse width of 100 fs, pulse energy of 3.5 mJ per pulse, and repetition rate of 1 kHz) was used as a light source. The output from the laser was split into four paths for the excitation of two optical parametric ampliers (OPAs: Spectra-Physics, TOPAS Prime), the whitelight-continuum generation by focusing the fundamental light (800 nm) into a sapphire plate, and the second and third harmonic generations of the fundamental light (800 nm) by using BBO (b-BaB 2 O 4 ) crystals. For generating the 440 or 580 nm pump pulse, light from one of the OPA was used. For the probe pulse, a white light continuum covering from 500 nm to 1600 nm and a 3435 nm probe light generated from the other OPA with a difference-frequency generation crystal was used. The time resolution of the system was about 140 fs. The powder samples are taken in 1 mm quartz cuvettes. The diameter of the pump beam on the sample was about 0.5 mm as observed with a charge-coupled-device (CCD) camera. Amplied Si photodetector is used to measure the TA spectra in the visible region. Liquid-nitrogen-cooled mercury-cadmium-telluride (HgCdTe) photodetector is used for IR probe (3435 nm) experiments. The diffusely reected light from the sample was passed through a grating monochromator (Princeton Instruments, Acton SP2150) for data acquisition. The transient absorption intensity of the TDR measurements is presented as percentage absorption, where absorption (%) ¼ 100(1 À R/R 0 ), using R and R 0 as the intensities of the diffusely reected light with and without excitation, respectively. A more detailed description of the fs-TDR setup is available elsewhere. 42,43 4.5 X-ray photoelectron spectroscopy (XPS) The XPS spectra were obtained using X-ray photoelectron ULVAC-PHI, INC PHI Quantera 2 spectrometer.

Conflicts of interest
There are no conicts of interest to declare.