Insight into the charge transfer in particulate Ta3N5 photoanode with high photoelectrochemical performance

Charge transfer has been demonstrated to have a fundamental role in particulate Ta3N5 electrode for achieving high efficient photoelectrochemical water oxidation.


Introduction
Photo-induced water splitting paves a promising way for the production of renewable solar fuels by converting solar energy to hydrogen directly. To achieve this conversion process, photocatalysis (PC) and photoelectrocatalysis (PEC) are viable technological choices. [1][2][3][4] Many kinds of materials have shown photoresponse in PC but failed in PEC due to the issue of electrode fabrication. 2,5,6 In PC, an efficient charge separation in particles can usually lead to fast reaction at the semiconductor/liquid interface with the assistance of a suitable cocatalyst. 7 In PEC, however, photogenerated electrons must transfer through the lm and be collected by the substrate to match the surface reaction. Thus, the charge transportation in the lms also plays a fundamental role in the whole PEC process, 8 and the interfaces of particle-particle (PP) and particle-substrate (PS) are critically important for charge transportation. To ensure good connections at these interfaces, semiconductor is usually epitaxially grown on a conductive layer in situ by, for instance, hydrothermal, 9 chemical bath, 10 and vapour deposition methods. 11,12 Electrodes of low-dimensional structure 13,14 or host-guest structure 15,16 are intentionally designed to improve the charge transportation efficiency in the electrode so as to decrease the electron-hole recombination. However, most previous researches focus on charge transfer in the semiconductor particles since it determines the generation of separated electron-hole pairs, [17][18][19] and little attention has been paid on clarifying the charge transfer process in the lm, which also plays a determinable role during the photoelectrochemical reaction.
Transient photovoltage (TPV) can provide direct insight into the charge transfer process through the electrode. Upon illumination, Dember photovoltage is generated 20 which stems from the diffusion difference of photogenerated electrons and holes. Typically, the gradient distributed light in the lm excites photogenerated electron-hole pairs in gradient concentration 21,22 which will cause them to diffuse from a high concentration region (surface, high light intensity) to the low concentration region (bulk, low light intensity) at different velocities. Then the electron-hole pairs dri apart and electric eld builds up; the decay process of the photovoltage can reveal the charge transfer process in the lm.
Cyclic voltammetry (CV) in the dark can also reveal the electron transfer in electrodes. 23 For the cathodic reaction, electrons must transfer through the lms before exchanging with an efficient redox couple, e.g. Fe(CN) 6 3À /Fe(CN) 6 4À . The magnitude of the cathodic current and potential of the reduction peak can reveal the electron transportation in the lm. In addition, a particulate electrode with well-connected lm can result in porous structure which provides a large electrochemical surface area for reaction. This area is proportional to the capacitance of the Helmholtz layer which can be determined from CV 24 and so we can evaluate the charge transportation in the lm by an electrochemical method. Suitable materials and fabrication methods of the electrodes are important for us to pinpoint the issue of charge transportation. In terms of materials, semiconductors with long free paths are the best candidates because it permits us to focus on the charge transfer at the PP interfaces regardless of the charge diffusion in the semiconductor crystals. Ta 3 N 5 is reported to have a diffusion length of $10 3 nm, 25 indicating a long life time of photogenerated charges. Moreover, Ta 3 N 5 has demonstrated an outstanding photocatalytic oxygen evolution activity, 26 indicating an effective charge separation in Ta 3 N 5 crystals. Also, the excellent light harvesting ability of Ta 3 N 5 makes it an appealing material for photoelectrochemical water splitting with a potential solar energy conversion efficiency of 15% under AM 1.5 G sunlight. 27 For the fabrication of the electrode, a bottom-up method, that is by depositing as-prepared semiconductor particles on the conductive substrate to form a particulate electrode, can make it more facial so as to regulate the charge transfer at the PP and PS interfaces without inuencing the intrinsic properties of the semiconductor such as light absorption, carrier concentration etc. Electrophoretic deposition (EPD) method is an alternative choice. A wide range of semiconductors can be fabricated into electrodes with controllable thickness by EPD, such as Fe 2 O 3 , 28 BiVO 4 , 29 Ta 3 N 5 , 30 TaON, 31,32 while postnecking treatment has been found to be powerful in improving the PEC response. [30][31][32][33][34] Some explanation to the possible function of necking treatment has been proposed, 31 but more stringent evidences are required to identify the effects of this modication.
Herein, we take Ta 3 N 5 as a demonstration to show the inuence of charge transportation on the PEC performance of a particulate Ta 3 N 5 electrode. By optimizing the substrate, precursor concentration, temperature of necking treatment and cocatalyst loading, we have achieved the highest photocurrent for electrodes fabricated by EPD. The TPV based on Dember photovoltage and CV measurements are used to clarify the critical effect of necking treatment on charge transfer at the interfaces of particle-particle and particle-substrate.

Experimental
Ta 3 N 5 powder synthesis Ta 2 O 5 powder (Amresco Chemical, $99.99%) was immersed in water, then it was dried and annealed in air at 800 C for 2 h, prior to being nitrided in ammonia ow (250 sccm) at 950 C for 15 h.

Ta 3 N 5 electrode fabrication
The as-synthesized Ta 3 N 5 powder (50 mg) was dispersed into 50 mL acetone (Kemeol, $99.5%) under ultrasonic treatment for 10 min. Then 20 mg iodine was dissolved into the suspension to make it suitable for EPD. Ti foil (1 cm Â 2 cm) was used as substrate aer washing in 1 M HF aqueous, pure water and anhydrous ethanol. A piece of uorine-doped tin oxide (FTO, Nippon Glass Sheet) glass (2 cm Â 3 cm) is used as the counter electrode with the conductive layer facing towards the Ti foil at a distance of 1 cm. The Ta 3 N 5 crystals were deposited to the Ti foil at a bias of 20 V for 1 min.

Characterization
The absorption spectra from 350 to 800 nm were taken on Cary 5000 UV-VIS-NIR spectrophotometer (JASCAO) equipped with an integrated sphere. X-Ray diffraction (XRD) patterns were recorded on Rigaku D/Max-2500/PC powder diffractometer operating at 40 kV and 200 mA with Cu-Ka radiation (l ¼ 0.154 nm) at a scanning rate of 5 min À1 . The morphology of the electrodes was imaged by a Quanta 200 FEG scanning electron microscope (SEM). High-resolution transmission electron microscopy (HRTEM) images were obtained on Tecnai G2 F30 S-Twin (FEI Company) with an accelerating voltage of 300 kV. X-Ray photoelectron spectroscopy (XPS) was recorded on VG ESCALAB MK2 spectrometer with monochromatic Al-Ka radiation (12.0 kV, 240 W). All the bonding energies were corrected with reference to the C 1s (284.8 eV) signal.

Transient photovoltage measurement
The transient photovoltage (TPV) was measured with a pulsed laser (355 nm, 5 ns) using a Ta 3 N 5 device (see ESI for details †). The average power was 122 mW unless otherwise stated. The signals were read from an oscilloscope (Tektronix TDS 3012C).

(Photo)electrochemistry measurement
Cyclic voltammetry was measured without illumination in an electrolyte of NaOH (1 M) or NaOH (1 M)-K 3 Fe(CN) 6 (0.25 M) aqueous solution at a scan rate of 100 mV s À1 in the range of À1.4 to 0.4 V vs. SCE. For the electrochemical area measurement, the scan rate was varied from 20 to 500 mV s À1 in the range of À0.1 to 0.2 V vs. SCE in 1 M NaOH aqueous solution.
The photocurrent was recorded under simulated light (AM 1.5 G, 100 mW cm À2 ) at a scan rate of 50 mV s À1 from À0.8 V to 0.6 V vs. SCE in 1 M NaOH aqueous solution (pH 13.6).
The incident photon-to-current conversion efficiency (IPCE) was measured under monochromatic light irradiation provided by a tungsten lamp equipped with a monochromator (CROWNTECH, QEM24-D 1/4 m Double).
Faradaic efficiency was tested by recording the photocurrent and the generated O 2 simultaneously. The O 2 evolution was evaluated by gas chromatography (GC, Agilent 7890a GC) with a 5Å molecular sieve column. Argon carrier gas with a velocity of 10.0 mL min À1 was used to purge the working electrode compartment to carry the evolved gases to GC for analysis. The quantity and retention time of the gases were calibrated with a series of standard gas samples.
All the (photo)electrochemical tests were conducted on an Ivium potentiostat/galvanostat in a three-electrode system with a quartz window. A piece of platinum foil (2 cm Â 2 cm) was used as counter electrode and saturated calomel electrode (SCE, 0.241 V vs. RHE) used as the reference electrode. The potential was converted to the reversible hydrogen electrode (RHE) scale by the Nernst equation as below: (1)

Results and discussion
The morphology of the electrodes was rst characterized by SEM. The SEM image of a cross-section view shows that the asprepared electrode has a thickness of 5-7 mm (Fig. S1 †). Topview images show that the surface of porous raw Ta 3 N 5 ( Fig. 1a) is covered with an amorphous layer aer treating with TaCl 5 solution (Fig. 1b). The following heat treatment seems to retrieve the smooth surface of Ta 3 N 5 but boundaries between particles become less obvious compared to those on raw-Ta 3 N 5 electrode (Fig. 1c). The HRTEM image conrms the existence of a 1-5 nm amorphous layer (Fig. 1d). This amorphous tantalum species may bridge the adjacent Ta 3 N 5 particles. The XRD patterns ( Fig. S2 †) show no change of the Ta 3 N 5 crystals with necking treatment, however, the surface change was revealed in XPS spectra (Fig. 1e). There is no prominent difference between raw-Ta 3 N 5 and heated-Ta 3 N 5 electrodes but for the TaCl 5 -Ta 3 N 5 electrode, another peak at 28.1 eV was observed, in agreement with the reported Ta 4f 5/2 signal in tantalum oxide. 35,36 It is inferred that the amorphous layer in Fig. 1b may be tantalum oxide. Since it was covered on the raw-Ta 3 N 5 electrode, the intensity of O 1s (530.7 eV) became stronger, and the intensity of Ta 4f 7/2 (24.6 eV) became weaker aer treating with TaCl 5 solution, indicating the screening effect of the amorphous tantalum oxide. It is also inferred that Ta 4f 7/2 at 24.6 eV arises from Ta-N bonds. Aer necking treatment, the intensity of O 1s (530.7 eV) ascribed to surface absorbed oxygen species is prominently decreased compared to the TaCl 5 -Ta 3 N 5 electrode, with Ta 4f 5/2 ascribed to Ta-O shied from 28.0 to 27.6 eV, and Ta 4f 7/2 ascribed to Ta-N bond shied from 24.6 to 25.0 eV for the necked-Ta 3 N 5 electrode. It is suggested that a nitrogendoped tantalum layer has been formed. Because the Ta-N bond is more covalent than the Ta-O bond, nitrogen doping of tantalum oxide during necking treatment will shi the binding energy of Ta 4f to lower energy. [36][37][38] Despite the small changes of the morphology of Ta 3 N 5 particles, the PEC performance shows a substantial difference with necking treatment (Fig. 2a). The raw-Ta 3 N 5 electrode exhibits notoriously low photoresponse, and so does the TaCl 5 treated electrode. However, aer necking treatment, the photocurrent of the necked-Ta 3 N 5 electrode increases to 1.56 mA cm À2 from 9 mA cm À2 . For the necking treatment, the TaCl 5 concentration and post-heating temperature have great inuence on the photoresponse of the necked-Ta 3 N 5 electrode ( Fig. S3 †). The optimized TaCl 5 concentration is 15-20 mM (concentration at 20 mM gives the best repeatability) and the best post-heating temperature is 600 C for the rened necking treatment. At high temperature, the particles will have strong connection with each other at the PP and PS interfaces. However, NH 3 has strong capability of reduction at high temperature, and can damage the normally used FTO substrate. 28,31 Some metal based candidates were chosen and Ti foil gave the best result (Fig. S4 †). The benet of calcination is also conrmed by the improved photoresponse of heated-Ta 3 N 5 compared to the raw-Ta 3 N 5 electrode. Further cocatalyst loading, e.g. CoPi, 39 on necked-Ta 3 N 5 electrode efficiently accelerates the water oxidation to an optimized current of 6.1 mA cm À2 at 1.23 V vs. RHE (Fig. 2b), which is higher than most reported results fabricated by the EPD method. 30,31,40,41 In Fig. 3, IPCE action spectra show the quantum efficiency at different irradiation wavelengths, which are in agreement with the absorption of Ta 3 N 5 . The raw-Ta 3 N 5 electrode shows very low IPCE, but aer the rened necking treatment, it dramatically increases to more than 20%. Also, the IPCE is much higher by loading CoPi or increasing the bias to accelerate the consumption of photogenerated charges.
The rened necking treatment plays a key role in boosting the photoanodic performance of the particulate Ta 3 N 5 electrode and it is also essential for the cocatalyst functioning Ta 3 N 5 photoanode. Thus we endeavour to pinpoint the exact role of necking treatment by analysing the reaction at the semiconductor/liquid interface and charge transfer at PP and PS interfaces on the particulate Ta 3 N 5 electrode.
For an electrode made from powder, necking treatment may have an inuence on both the intrinsic property of Ta 3 N 5 particles and the connection among particles. In order to elucidate the possible change of Ta 3 N 5 crystals, the absorption spectra and photocatalytic activity (see ESI for details †) were measured. The absorption spectrum of Ta 3 N 5 shows little change aer necking treatment (Fig. S5a †), indicating that the necking treatment does not inuence the intrinsic light harvest in Ta 3 N 5 . From the absorption spectra, the marginal photocurrent of the electrode was evaluated to be $12.6 mA cm À2 under simulated sunlight (AM 1.5 G, 1 sun) (Fig. S5 †). The inuence of necking treatment to the Ta 3 N 5 particles were further ascertained by photocatalytic O 2 evolution. In a PC reaction (Fig. 4a), light is harvested in the Ta 3 N 5 crystal and photogenerated charges dri to the active sites to take part in the surface reaction. The proportion of photogenerated charges that reach the surface reaction sites to those generated in Ta 3 N 5 particles upon illumination is   referred as charge generation efficiency (F gene ), which is basically the charge separation efficiency in the Ta 3 N 5 crystal. To preventing possible confusion with the separation efficiency in the electrode (as will be mentioned below), we dene it as generation efficiency for the crystal. This generation efficiency is largely determined by the bulk property of the crystal. The proportion of photogenerated holes that take part in the (electro)chemical reaction to those arriving at the reaction sites is referred as injection efficiency (h inj ), which is inuenced by the surface property of Ta 3 N 5 . When the surface is not preferred for catalytic water oxidation, although sufficient holes reach the reaction sites, they cannot be consumed in time and h inj will be less than unity. In the presence of efficient electron scavengers, the inuence of electrons in the crystal is limited and the PC reaction can be used to evaluate the intrinsic properties of the Ta 3 N 5 particles, such as the charge generation in the crystals and hole injection on the surfaces. 42 The PC activities of the Ta 3 N 5 powder from the raw-Ta 3 N 5 electrode and necked-Ta 3 N 5 electrode were evaluated with AgNO 3 as sacricial reagent. 26 As is shown in Fig. 4b, the amount of released O 2 for necked-Ta 3 N 5 has slightly decreased, implying that necking treatment has limited inuence on the intrinsic property of Ta 3 N 5 .
Then we focussed on the impact of necking treatment to the electron transfer at the PP and PS interfaces in the Ta 3 N 5 electrode. In a PEC reaction (Fig. 4a), electrons and holes should be considered simultaneously since the electrons need to transfer through the Ta 3 N 5 lms during the PEC water oxidation. The transportation of photogenerated charges will be another factor that limits the PEC performance.
As it is reported, the nal photocurrent is determined by the following expression: 43,44 where J 0 is the theoretical photocurrent, h LH is the light harvest efficiency, h inj is the charge injection efficiency as dened above, and h sep is separation efficiency in the electrode as dened in ref. 43. The absorption and PC water oxidation measurement have revealed that necking treatment has little inuence on light harvest (h LH ) and surface reaction (h inj ). Hence, the difference of charge separation (h sep ) of the electrode should be responsible for the huge difference of the PEC activity. Taking H 2 O 2 as hole scavenger, 43 the charge separation efficiency is calculated. From  Fig. 4c, it is revealed that the charge separation in raw-Ta 3 N 5 electrode is inefficient, but necking treatment improves h sep to more than 30% at the bias of 1.3 V vs. RHE. Interestingly, we found that CoPi/necked-Ta 3 N 5 electrode has an even higher h sep of 60% at 1.3 V vs. RHE.
A higher separation efficiency means less electron-hole recombination in the electrode. For a photoelectrocatalytic reaction occurring on a particulate electrode (Fig. 4a), the recombination may occur via two routes: (i) recombination in the particles through, for example, bulk defects or surface states; (ii) recombination at the interfaces of particles. Thus the measured separation efficiency should be dependent on the charge generation in particles and transfer among particles under illumination: where F gene is the charge generation efficiency as mentioned in Ta 3 N 5 crystal, relating to recombination process (i), and F trans is the transportation efficiency for photogenerated charges transferring through PP and PS interfaces before arriving at the conductive substrate from the origin, 45 relating to recombination process (ii). CoPi has been reported to suppress the surface states, and more long-lived photogenerated holes can survive in the particles. 46 Thus, the F gene is increased and this leads to the cocatalyst promoted charge separation for the CoPi/necked-Ta 3 N 5 electrode (Fig. 4c). For the necked-Ta 3 N 5 electrode, the previous absorption (Fig. S5 †) and PC activity (Fig. 4b) indicates a similar F gene . Thus, the dramatically improved h sep for necked-Ta 3 N 5 electrode should stem from the improved F trans upon necking treatment.
To consolidate the conclusion, prototypical devices fabricated from Ta 3 N 5 electrodes (Fig. 5a, see ESI for details †) were used to delve into the charge transportation in the electrode by TPV. In Fig. 5b, the necked-Ta 3 N 5 based device shows higher transient photovoltage and faster decay process than the raw-Ta 3 N 5 based device. A linear response of current-bias (Fig. S6 †) conrms the ohmic contact at the interfaces of PS. 20 Thus, the detected photovoltage plausibly stems from the Dember effect and this was further conrmed by the relationship between the direction of the induced laser and electric eld. For the Dember effect, holes always stay closer to the top layer, and the direction of the electric eld is always consistent with the direction of the laser beam (Fig. S7a †), and the intensity of the Dember voltage is determined by the amount of separated charges. Light of stronger intensity will induce more separated charges and hence a higher photovoltage (Fig. S7b †). When the intensity of the induced light is the same, and the same amount of charges are generated, a higher photovoltage indicates better charge separation. Thus the higher photovoltage of the necked-Ta 3 N 5 device indicates that it shows better charge separation which is in accordance with the result in Fig. 4.
When light is removed, the electron-hole pairs will recombine, leading to a dynamic decay of the Dember photovoltage at the time scale of ms (Fig. 5b, solid line). Two plausible recombination processes mentioned above are involved in the decay process. Dual exponential curves (Fig. 5b, dashed line) are tted with a fast decay process (recombination in Ta 3 N 5 crystals, lifetime of s 1 ) and a slow one (recombination at interfaces, lifetime of s 2 ). The two electrodes have approximately the same s 1 values, conrming again that the Ta 3 N 5 particles on both electrodes have similar charge generation efficiency (F gene ) as the PC result reveals in Fig. 4a. However, the necked-Ta 3 N 5 electrode has lower s 2 , indicating that the electron-hole pairs more readily recombine through the interfaces in the TPV devices.
In order to clarify the result of TPV further, an equivalent circuit (EC) based on the transmission line model 47 (Fig. 5c) was used to simulate the decay process in the device. Because there is no semiconductor/electrolyte interface, only three processes are considered: (i) the electron transport resistance in the Ta 3 N 5 crystals (R crystal ), (ii) the electron transport resistance among particles (R) and particles-substrate (R contact ), (iii) a capacitive charging to the porous Ta 3 N 5 matrix (C bulk ). For the Ta 3 N 5 crystal, the conduction band or the defect energy level can accommodate electrons which can behave like a capacitor (C crystal ). C bulk can be estimated to be at the order of 10 À5 F cm À2 , while C crystal is around 10 À8 F cm À2 (see ESI for details †). Thus, C crystal is much smaller compared to C bulk and the series of C crystal make it even smaller, so that we can ignore C crystal . So the equivalent circuit in Fig. 5c can be reduced to that in Fig. 5d, including the resistance among and in Ta 3 N 5 particles. The time constant (s 0 ¼ R bulk C bulk ) of electronic decay in the simulated circuit corresponds to the lifetime (s) determined by TPV. As the bulk capacitance (C bulk ) of the two devices (raw-Ta 3 N 5 and necked-Ta 3 N 5 ) are estimated to be of the same order of magnitude based on eqn (S1) (ESI †) the change in time constant is reected in the change of resistance in the device. Comparing the lifetime of s 2 in the two types of Ta 3 N 5 electrodes as tabulated in Fig. 5b, the smaller s 2 of the necked-Ta 3 N 5 electrode means that necking treatment decreases the charge transfer resistance (R bulk ) in the Ta 3 N 5 lm.
The amorphous layer of nitrogen doped tantalum oxide that bridges the Ta 3 N 5 particles is a plausible route for electron transportation. As necking treatment is performed at high temperature, efficient connection at interfaces of PP will be formed. Additionally, nitrogen doping in the tantalum oxide layer can improve its conductivity 48 and facilitate the transportation of the photogenerated charges. Indeed, it is reasonable that low charge transfer resistance should lead to the high photovoltage and fast decay process since the Dember photovoltage is caused by driing apart of electron-hole pairs, the better the conductivity is, the easier they can be dissociated/ recombined, and vice versa.
The efficiency of electron transportation across the interfaces of PP and PS was further probed using the benchmark redox couple of ferri/ferrocyanide. 23 The exchange of electrons between the solution and electrodes is fast (Fig. S8(a) †), allowing for the characterization of electron transfer by CV. The reduction reaction occurring on the Ta 3 N 5 electrode (without illumination) can provide a direct measure on the electron transportation. In darkness, the electrochemical reduction occurs by transferring electrons from the Ta 3 N 5 electrode to K 3 Fe(CN) 6 . Better electron transportation in the lm will lead to more facial electron exchange. Moreover, the improved electron transfer in the lm will decrease the ohmic polarization of the electrode and decrease the overpotential for the reduction of K 3 Fe(CN) 6 . The CV reveals that Ta 3 N 5 is poor for hydrogen evolution (Fig. S8(b) †) but active for K 3 Fe(CN) 6 reduction ( Fig. S8(c) †). This allows us to focus on the electron exchange between the electrode and K 3 Fe(CN) 6 regardless of the inuence of hydrogen evolution in the potential window of À0.4 to 1.2 V vs. RHE. Fig. S8(c), † shows that the reduction peak of K 3 Fe(CN) 6 shis positively on Ta 3 N 5 electrode, indicating that Ta 3 N 5 is superior to Ti substrate in electrocatalytic reduction of K 3 Fe(CN) 6 . When Ta 3 N 5 lm has a good contact with the substrate, more electrons will be exchanged with K 3 Fe(CN) 6 on the Ta 3 N 5 particles than on the Ti substrate (Fig. S9 †). Thus, the results in Fig. 6a can be related to electron transfer in the Ta 3 N 5 lm. It is found that the necked-Ta 3 N 5 electrode has a reduction current of À19.7 mA cm À2 at 0.65 V vs. RHE, while that for the raw-Ta 3 N 5 electrode is À18.1 mA cm À2 at 0.38 V vs. RHE, which is similar to that of the substrate (Fig. S8(c) †). The more positive potential along with higher current of the reduction peak indicates more efficient electron exchange on necked-Ta 3 N 5 electrode. Because electrons should transport through the particulate Ta 3 N 5 lm, it is concluded that necked-Ta 3 N 5 electrode is superior to raw-Ta 3 N 5 electrode in electron transportation (Fig. S9 †). The reduction peak potential of raw-Ta 3 N 5 electrode (0.38 V vs. RHE) is close to the substrate (0.37 V vs. RHE), indicating that more electrons are leaked out through the substrate other than through Ta 3 N 5 because of the weak connection at the PS interfaces. Additionally, the CV of CoPi/ necked-Ta 3 N 5 electrode is similar to that of necked-Ta 3 N 5 The simplified circuit of (c). R contact is the serial resistance at the particle-substrate interfaces; R bulk is the total resistance in Ta 3 N 5 films including the resistance in Ta 3 N 5 particles (R crystal ) and at particle-particle interfaces (R); C bulk is the capacity of the parallel-plate capacitor built by Ti and FTO. electrode ( Fig. S8(d) †), conrming that CoPi has little impact on the charge transfer in the electrode.
To sum up, necking treatment can facilitate charge transfer at the PP and PS interfaces in the Ta 3 N 5 lm which is benecial for the collection of photogenerated electrons, and depressing the recombination at the interfaces.
For particulate electrode, only the particles having good connection with the conductive layer can provide the area for electrochemical reaction. If an electrode made from powder has good conductivity in the lm a porous structure with large electrochemical active area is expected. By eliminating the inuence of the diffusion layer with concentrated electrolyte, we can get the relative area of the electrode from its charging capacitance of the Helmholtz double layer based on the following expression: 8,24 where v is the scan rate, and C is the capacitance of the double layer which is proportional to the surface area S.
Here, we measured the charging current (j) in a potential window (À0.1 to 0.2 V vs. SCE) where there is no faradaic current at different scan rates without illumination (Fig. S10 †). The j-v plot of each electrode is shown in Fig. 6b and the slope should be proportional to the surface area based on eqn (4). Taking the area of bare Ti substrate as unit, it is found the area only increases a little with TaCl 5 treatment or heating. However, the rened necking treatment leads to 22 times larger electrochemical surface area. Based on the analysis above, it is inferred that for raw-Ta 3 N 5 electrode, the poor connection in the particulate Ta 3 N 5 layer contributes little to the electrochemical active area. Only the layer of Ta 3 N 5 contact with the substrate can be used for reaction (Fig. S9, † le) and thus the electrochemical area is limited. Aer necking treatment, the connection at the interfaces of particle-particle (PP) and particlesubstrate (PS) are improved substantially and even the layer of Ta 3 N 5 far from the substrate can be used for reaction (Fig. S9, †  right). Thus the necked-Ta 3 N 5 electrode provides a much larger electrochemical surface area for reaction.
Inspired by the results above, another example was further raised to support the conclusion. We found that the necking treatment can improve the photocurrent to 6.0 mA cm À2 at 1.6 V vs. RHE for the Ta 3 N 5 electrodes made from traditional thermal oxidation-nitridation method (Fig. S11, see ESI for details †).
As for the role of CoPi, it can improve the charge separation by enhancing the charge generation in the Ta 3 N 5 crystal as shown above. Further, the surface charge injection process is also greatly accelerated with CoPi (Fig. S12 †). Because of the cocatalyst promoted charge separation and injection, the CoPi/ necked-Ta 3 N 5 electrode can provide a photocurrent of 11.2 mA cm À2 at 1.6 V vs. RHE. Moreover, the fast consuming of photogenerated holes can protect Ta 3 N 5 from being oxidized. 44 Thus, the stability of the necked-Ta 3 N 5 electrode is substantially improved with the assistance of CoPi as shown in Fig. 7a. The near unit faradaic efficiency of the CoPi/necked-Ta 3 N 5 electrode (Fig. 7b) conrms that the photocurrent is originated from O 2 evolution reaction. However, the photocurrent decayed noticeably with prolonged time (Fig. 7a). The failure of the Ta 3 N 5 electrode is suspected to be the result of the large surface area of the particulate electrode which makes it difficult to wholly cover Ta 3 N 5 with effective co-catalyst in the porous structure. Many unconsumed photogenerated holes were accumulated and destroyed the intrinsic Ta 3 N 5 . Further efforts to improve the stability is still in progress.

Conclusions
Efficient Ta 3 N 5 photoanode is fabricated on Ti foil by EPD method with rened necking treatment. Further loading cocatalyst CoPi gives a photocurrent of 6.1 mA cm À2 at 1.23 V vs. RHE under simulated sunlight (AM 1.5 G). To the best of our knowledge, this is the highest photoresponse for an electrode made by EPD. The benet of necking treatment is proved to stem from the high-temperature treatment and the formation of a nitrogen-doped tantalum oxide layer, which can improve the charge separation efficiency. The TPV shows higher Dember photovoltage and faster decay process for necked-Ta 3 N 5 electrode, which suggests that necking treatment can decrease the charge transfer resistance at particle-particle interfaces. This Fig. 6 (a) Cyclic voltammograms on raw-Ta 3 N 5 (orange), TaCl 5 -Ta 3 N 5 (pink), heated-Ta 3 N 5 (blue) and necked-Ta 3 N 5 (black) electrodes for the reduction of ferricyanide in NaOH + K 3 Fe(CN) 6 aqueous solution. (b) The plots of capacitive current to scan rate for the Ti foil (grey), raw-Ta 3 N 5 (orange), TaCl 5 -Ta 3 N 5 (pink), heated-Ta 3 N 5 (blue) and necked-Ta 3 N 5 (black) electrodes. The slopes are shown beside the curves. will promote the collection of photogenerated charges, decrease the recombination at the interfaces and improve the charge separation efficiency. CV measurement further conrms the benet of necking treatment in promoting electron transfer and providing higher electrochemically active area for surface water oxidation.