Tantalum Nitride‐Enabled Solar Water Splitting with Efficiency Above 10%

Designing photoanode semiconducting materials with visible‐light absorption and minimal charge‐carrier recombination for achieving efficient solar‐to‐hydrogen (STH) conversion is challenging. Here, hybrid Ta3N5 nanorods and thin films are developed on transparent GaN/Al2O3 substrates. A Ta3N5 photoanode with a loaded cocatalyst achieves the best current density, i.e. 10.8 mA cm−2, at 1.23 V versus the reversible hydrogen electrode under simulated AM 1.5G solar illumination. In a tandem configuration with dual‐CuInSe2 photovoltaic cells, this semi‐transparent photoanode achieves a reproducible STH energy conversion efficiency of ≈12% (the highest among photocatalytic materials), and remains at more than 10% for 6.7 h of tandem device operation. Detailed transient absorption spectroscopy and theoretical analysis indicates that this high performance originates from efficient light absorption and hole utilization inside the Ta3N5 material. The results show the feasibility of suppressing dominant optical and charge‐carrier‐ recombination losses by using nanostructured visible‐light‐absorbing materials for practical STH conversion.


Introduction
Water splitting driven by abundant solar light is a potential method for eco-friendly and sustainable hydrogen fuel production with a zero carbonfootprint to meet rising global energy demands and address climate change concerns. [1,2]A high solar-to-hydrogen (STH) energy conversion efficiency is one of the key criteria for the large-scale production of clean hydrogen fuel via sunlight-induced overall water splitting (OWS). [2]The design of photovoltaic electrolysis and wireless monolithic tandem devices has enabled STH energy conversion efficiencies of up to 30%. [3,4]owever, the development of these technologies is hampered by fabrication and scale-up difficulties, and high costs. [2][7] The key components of this tandem device are serially wired wide and narrow energy bandgap photoabsorbing semiconductors, and a counter electrode.During device operation (see Figure S1, Supporting Information), the incident solar spectrum is partially absorbed by the top photoabsorber, and the rest of the transmitted spectrum is harvested by the bottom photoabsorber to generate charge carriers for OWS.Assuming an external quantum efficiency (EQE) of 100%, the STH energy conversion efficiency of a tandem device can theoretically reach 26.86% for top and bottom bandgap energies of 1.72 and 1.1 eV, respectively.A wide range of materials, e.g.silicon, [8,9] lead halide perovskites, [10][11][12] chalcogenides, [10,11] and metal oxides, [5,12] have been extensively explored for photoanodic/photocathodic/photovoltaic applications (see Table S1, Supporting Information) to achieve unbiased OWS.Although a STH energy conversion efficiency of 17.6% has been achieved with a silicon-perovskite-based tandem device, [8] lead-based perovskites have disadvantages in terms of environmental issues and stability.The STH energy conversion efficiency of a single-junction metal oxide/nitride (Ta 3 N 5 , BiVO 4 , Fe 2 O 3 , and WO 3 ) [5,6,10,13] photoanode-based tandem device is less than 7%.The design of photoanodes with narrow bandgap energies (1.3-2.35eV; Figure S1, Supporting Information), good semi-transparency, and minimum charge-carrier losses is imperative and remains challenging for achieving efficiencies greater than 10%.
Ta 3 N 5 (with a bandgap energy of 2.1 eV) [14,15] meets the above prerequisites for photoanodes and can potentially achieve STH energy conversion efficiencies of up to 15.26%.There are many literature reports on enhancement of the performances of Ta 3 N 5 -based photoanodes by doping with other elements, [16,17] using a selective charge-transport layer, [18][19][20] and nanostructuring. [14,21,22]These design strategies suppress optoelectrical losses to deliver a current density at 1.23 versus the reversible hydrogen electrode (V RHE ), fill factor, and onset potential of up to 10 mA cm −2 , 42%, and 0.4 V RHE , respectively.These photoanodes are not suitable for the development of PEC-PV tandem devices because of the non-transparent substrate, and consequently give half-cell-STH energy conversion efficiencies of up to 4.07%. [19,23]Previously, our group prepared planar Ta 3 N 5 thin films by conventional sputtering on a transparent GaN/Al 2 O 3 substrate, which was integrated with dual-CuInSe 2 photovoltaic cells in a tandem configuration. [10]This Ta 3 N 5 -CuInSe 2 tandem device facilitated bias-free OWS with a STH energy conversion efficiency of ≈7%; this is higher than those achieved with metal oxide-based tandem devices.The difference between the theoretical and achieved STH efficiencies probably originates from optical and charge-carrier-recombination losses from the Ta 3 N 5 photoanode.Recently, charge-carrier losses have been further suppressed by developing a tandem device based on a pin-hole-free Ta 3 N 5 thin film. [24]The initial STH energy conversion efficiency was ≈9%; it decreased to ≈4% within 15 min of device operation.These results all indicate that material and structural design of Ta 3 N 5 is essential for boosting the STH energy conversion efficiency above 10% for potential commercial applications.
Herein, we describe the production of hybrid nanostructured Ta 3 N 5 -nanorods (NRs) and Ta 3 N 5 -thin films (TFs) on trans-parent GaN/Al 2 O 3 substrates by using glancing angle deposition (GLAD) and sputtering methods.The Ta 3 N 5 -NRs are polycrystalline, well separated, and vertically oriented on Ta 3 N 5 -TF/GaN/Al 2 O 3 substrates.With cocatalyst-loaded NRs, the hybrid Ta 3 N 5 photoanode achieved the best transmittance of 80% and a current density of 10.8 mA cm −2 at 1.23 V RHE (under simulated AM 1.5G solar illumination).The Ta 3 N 5 photoanode in tandem with dual-CuInSe 2 photovoltaic cells enabled unbiased solar water splitting with a record STH energy conversion efficiency of 12.1% (among well-studied photocatalytic-material-based tandem devices); the efficiency remained above 10% for up to 6.7 h of device operation.Transient absorption spectroscopy and theoretical analysis were used to investigate important material features such as the dominant charge-carrier relaxation processes, doping density, rate constant, and charge transport.The results show that the high performance of the tandem device is attributable to efficient light absorption and hole utilization by the Ta 3 N 5 photoanode.

Material and Structural Properties
Figure 1 shows the material characterization results for the fabricated Ta 3 N 5 -based photoanode.Initially, the TaO x N y -NRs and interlayer TF were deposited on a GaN/Al 2 O 3 substrate by GLAD and magnetron-sputtering techniques, respectively (see Experimental Section and Figure S2 of the Supporting Information for details).The X-ray diffraction (XRD) patterns in Figure S3 (Supporting Information) show that subsequent nitridation converted the amorphous TaO x N y to crystalline Ta 3 N 5 .The GLAD technique is easily scalable for large-scale production.For example, the GLAD method can simultaneously produce 20 TaO x N y -NRs specimens in a single batch (see Figure S4, Supporting Information).At the laboratory scale, the sample area is limited by the diameter of the quartz tube and nitridation furnace.However, such limitations of the nitridation process can be easily overcome for area scale-up, as shown by the mature application of high-temperature ammonia treatment in the automotive industries.The scanning electron microscopy (SEM) and highangle annular dark-field (HAADF) scanning transmission electron microscopy (STEM) images in Figure 1a,b show the formation of well-separated, vertically-oriented, and regular Ta 3 N 5 -NRs on a Ta 3 N 5 -TF/GaN/Al 2 O 3 substrate.The fabricated samples were loaded with FeNiCoO x cocatalyst by photo-electrodeposition (PED) and dip coating methods.After cocatalyst loading by PED, the Ta 3 N 5 -NRs surface is covered with a thin (≈4 nm), homogenous, and amorphous FeNiCoO x cocatalyst layer (see the highresolution transmission electron microscopy (HR-TEM) image of Figure S5a, Supporting Information).Dip coating technique enhances the thickness of the cocatalyst layer up to 8 nm (Figure 1c; Figure S5, Supporting Information).In addition, the HR-TEM images show fringes with interplanar spacings of 0.52, 0.37, and 0.29 nm, which correspond to the {020}, {022}, and {112} planes, respectively.The polycrystalline Ta 3 N 5 -NRs consist of aggregated single-crystal Ta 3 N 5 domains.The size of these grain domains is ≈25 nm (see the white area in Figure S5d, Supporting Information).The orientations of different planes can be uniquely identified by electron diffraction (Figure 1d).The STEM and energy-dispersive X-ray spectroscopy (STEM-EDS) mappings in Figure 1e indicate a uniform distribution of primary Ta and N elements.The STEM-EDS mappings show uniform loading of the FeNiCoO x cocatalyst across the Ta 3 N 5 -NR surface; this promotes water oxidation and stability, as reported in the literature. [16]

Performance Evaluation
Figure 2 shows the performance characteristics of the Ta 3 N 5 photoanode and PEC-PV tandem device in three-electrode and twoelectrode configurations, respectively.Details of the performance evaluation are provided in the Experimental Section.Prior to evaluation, the solar simulator was well calibrated to the AM 1.5G solar irradiance spectrum (Figure S6, Supporting Information).The tandem device consists of a top semi-transparent Ta 3 N 5 photoanode in series with the bottom CuInSe 2 photovoltaic elements and Ni/Pt electrode (Figure 2a). Figure S7 (Supporting Information) shows the working principle/charge flow diagram of the Ta 3 N 5 -CuInSe 2 tandem device for unassisted solar water splitting.High STH energy conversion efficiency of the tandem device was achieved by optimizing the transmittance spectra and current-potential characteristics of the Ta 3 N 5 photoanode by tuning the Ta 3 N 5 -TF thickness.Deposition of Ta 3 N 5 (NRs/TF) on the GaN/Al 2 O 3 substrate shifts the onset for transparency from 400 to 600 nm.However, the transmittance above 600 nm de-creases with increasing Ta 3 N 5 -TF interlayer thickness from 150 to 450 nm (Figure S8a-d, Supporting Information).The Ta 3 N 5 -NR length and GaN layer thickness are ≈1.6 and 4.5 μm, respectively.The decrease in transmittance with an increase of Ta 3 N 5 -TF interlayer thickness originates from enhanced light reflection (Figure S9, Supporting Information) rather than the presence of impurities like Ta 5 N 6 , Ta 2 N, and TaN (Figure S10, Supporting Information).Under simulated AM 1.5G solar illumination, the current density of the Ta 3 N 5 photoanode (at 1.23 V RHE ) improved from 0.2 to 10.5 mA cm −2 with increasing Ta 3 N 5 -TF thickness from 0 to 150 nm (Figure S8e, Supporting Information).This current gain is attributed to improved contact formation between the Ta 3 N 5 -NRs and the GaN layer, which gives efficient electron transport.However, the photocurrent density decreased slightly to 9.5 mA cm −2 with further increases in the Ta 3 N 5 -TF thickness to 450 nm.At an optically and electrically optimized Ta 3 N 5 -TF thickness of 150 nm, the current-potential (J-V a ) characteristics (Figure 2b) of the Ta 3 N 5 photoanode show an onset potential V on of 0.75 V RHE (corresponding to J = 0.8 mA cm −2 ) for water oxidation.Above V on , J rises sharply with increasing V a to reach 10.5 mA cm −2 at 1.23 V RHE , which is 84.5% of the reported theoretical limit for Ta 3 N 5 (ref.[21]).The incident photon-to-current efficiency (IPCE; Figure S8f, Supporting Information) is in the range of 90% to 80% from 300 to 540 nm, and drops sharply to zero at the light-absorption edge at 600 nm (2.06 eV).This indicates efficient solar energy harvesting by Ta 3 N 5 to produce holes for an efficient oxygen evolution reaction.The integrated current (9.4 mA cm −2 ) obtained from the IPCE is consistent with the measured J-V a data.The transmitted solar spectrum passing through the photoanode powers the dual-CuInSe 2 PV cells for self-biasing and the hydrogen evolution reaction.
Under simulated AM 1.5G solar illumination, the two seriesconnected (dual) CuInSe 2 cells extract a J of 19.4 mA cm −2 at 0.2 V RHE , with a sufficiently high V on (above 1.3 V RHE ) to drive water reduction at the Pt electrode (Figure 2b).The reasons for the use of CuInSe 2 cells for the tandem device are discussed in Figure S11 (Supporting Information).Owing to the limited light transmittance of the optimized photoanode, the dual-CuInSe 2 cells behind the photoanode (Figure 2b) give a reduced current of 11.4 mA cm −2 at 0.2 V RHE .Here, the current density is the ratio of the measured current from dual-CuInSe 2 cells to the photoactive area of the photoanode (twice the area of an individual CuInSe 2 cell).The J of the dual-CuInSe 2 cells is, therefore, half that of a single CuInSe 2 PV cell, which is typically ≈40 mA cm −2 .The reduction (by 40%) in the current density of CuInSe 2 PV cells when a photoanode is placed in front of them is consistent with the integrated current density obtained from EQE measurements (see Figure S12, Supporting Information).For tandem-device operation, the J-V a curves of the photoanode and dual-CuInSe 2 cells behind the photoanode intersect (Figure 2b) and they operate with a matched current density J op of 9.5 mA cm −2 (at V op 1.16 V RHE ).Such a high current density leads to a record STH energy conversion efficiency () of 11.7% for OWS activity (see Experimental Section).If PV elements with a higher onset potential and/or fill factor are used, the STH efficiency can be further enhanced to 13% with an anodic shift of the operating potential.Figure 2c shows that a STH energy conversion efficiency of more than 10% is maintained for up to 1.4 h of tandem device operation in the two-electrode configuration.Optimization of the cocatalyst deposition process improves and extends the photoanode stability to a relatively long time of 6.7 h (Figure S13, Supporting Information).This improvement in photoanode stability is attributed to an increase in the thickness of FeNiCoO x cocatalyst layer after dip coating method (Figure S5, Supporting Information).Quantitative analysis of individual cocatalyst elements from XPS measurement is challenging for nanostructured photoanode (see Figure S14, Supporting Information for details) and is currently out of the scope of present manuscript.Moreover, high-resolution XPS (HR-XPS) spectra of Fe 2p, Ni 2p, and Co 2p core levels showed enhanced peak intensity after dip-coating.This observation is in a good agreement with the increase in FeNiCoO x thickness, revealed by HR-TEM (Figure S5, Supporting Information).
Figure S15 (Supporting Information) shows that the cocatalyst is partially dissolved in the electrolyte solution during water oxidation and degrades the performance during extended hours of operation.X-ray photoelectron spectroscopy (XPS) confirmed the formation of Ta 3 N 5 (Figure S16, Supporting Information).In the absence of a cocatalyst, the Ta 3 N 5 photoanode gives a significantly low current density of 0.085 mA cm −2 at 1.23 V RHE (Figure S17, Supporting Information).The current degradation (by 50% within 4.2 mins) is attributed to the formation of insulating TaO x or TaO x N y on Ta 3 N 5 -NRs surface (evident from XPS analysis).Such oxide formation suppressed the hole extraction from Ta 3 N 5 to electrolyte for water oxidation. [25]In the presence of a cocatalyst, the nitrogen deficiency observed after stability tests originate from exposure of the Ta 3 N 5 -NRs surface to the electrolyte as a result of partial cocatalyst dissolution.The XPS results are consistent with previous literature reports. [14,19,26][29] Moreover, Yanbo Li et al. reported leaching of thermodynamically unstable Fe (VI) active centers contributes to the dissolution of cocatalyst during water oxidation. [30]We believe that the stability of Ta 3 N 5 photoanode can be enhanced by development of 1) (non-dissolving) stable co-catalyst in high alkaline solution and 2) an oxide-based surface protective layer on Ta 3 N 5 for hole transport.
On the basis of charge flow (using current-time (J-t curves)), Figure S18 (Supporting Information) evaluates the generation of hydrogen/oxygen gases and the STH energy conversion efficiencies of tandem devices.Figure 2d shows that hydrogen and oxygen gases (stoichiometric 2:1 ratio) are produced with an average STH energy conversion efficiency of 11.07% for nearly 7 h at a faradaic efficiency of 98% (Figure S19, Supporting Information).Figure S20 (Supporting Information) shows the performance optimization progression of Ta 3 N 5 -NRs photoanodes to achieve an STH energy conversion efficiency of 12.1% (the highest among metal oxide/nitride-based tandem devices).In addition, the results show reproducible device statistics with STH energy conversion efficiencies greater than 10%.
Figure S21 (Supporting Information) shows the PEC characterization of optimized and non-optimized Ta 3 N 5 -based photoanodes.Such measurements (see Experimental Section for details) correlate the device parameters with the performance characteristics of the photoanode.The results indicate that the optimized sample showed a lower onset potential but higher current density at 1.23 V RHE compared with those obtained with nonoptimized samples.Specifically, the current density is ≈10.4 and 7.7 mA cm −2 at 1.23 V RHE for the optimized and non-optimized photoanodes, respectively.As a result, the applied bias-to-photon efficiency for the optimized photoanode (1.3%) is higher than that for the non-optimized photoanode (0.5%).This performance gain for the optimized photoanode is attributed to higher charge separation, injection efficiency (using a hole scavenger H 2 O 2 electrolyte [31] ), and lower charge-transfer resistance (shown by electrochemical impedance spectroscopy (EIS)).A Mott-Schottky (MS) plot indicated a flat-band potential of 0.1 V RHE and ntype doping in Ta 3 N 5 .Determination of the doping density n d from MS analysis can be unreliable because of the frequencydependent slope of 1/C 2 versus V a (C is capacitance) and difficulty in determining the Ta 3 N 5 -NRs surface area.

Charge-Carrier Dynamics and Transport
The performance of Ta 3 N 5 -CuInSe 2 tandem device is governed by competing charge carrier dynamics and transport properties inside Ta 3 N 5 material.Probing such features are essential to reveal parameters responsible for high STH energy conversion efficiency of tandem devices.For this, transient diffuse reflectance spectroscopy (TDRS) and semiconductor device modeling were employed.Figure 3 presents the analysis of TDRS/charge carrier dynamics for the determination of parameters of the bestperforming Ta 3 N 5 photoanode.Details of the measurements are provided in the Experimental Section.The absorption signal S(t) was measured in diffuse reflectance mode by using the expression: Here, R and R 0 are the intensity of the diffusely reflected light with and without excitation, respectively.We used pump photon energy of 3.1 eV to generate electrons and holes in the conduction band (CB) and valence band (VB), respectively.After photoexcitation, the absorption signal S(t) decay using a probe photon energy of 0.15 eV was obtained.Our previous studies on Ta 3 N 5 indicate that the probe photon energy of 0.15 eV was absorbed by excess mobile electrons in CB to produce an absorption signal. [22,24]Figure 3a shows that the absorption signal S(t) and the S(t) decay rate increase with increasing pump fluence intensity P FL .The results indicate that S(t) decays faster with increasing P FL in the early, nanosecond time range.In the late, sub-microsecond time range, the S(t) decay is relatively independent of P FL and follows power-law (At − ) decay.Among several theoretical models for describing power-law decay, such S(t) decay features can originate from the bimolecular recombination of mobile charge carriers and hole-trapping/detrapping processes.A schematic diagram of such processes is shown in Figure S22 (Supporting Information).Bimolecular recombination is defined as the band-to-band recombination of mobile electrons in CB and mobile holes in VB.Bimolecular recombination introduces P FL -dependent S(t) decay and hole detrapping from exponential tail trap states accounts for power-law decay.The linear rise in the maximum absorption signal S m (at t = 0.4 ps) with P FL up to 3 μJ (Figure S23, Supporting Information) indicates that the photogenerated mobile-charge-carrier density ∆n 0 is probed without recombination within the time resolution of the system (140 fs).Here, S m = ∆n 0 , where  is a proportionality constant.Figure 3b shows that the amplitude A of the power-law decay increases rapidly, and almost reaches saturation, with increasing P FL .The exponent  is almost independent of P FL (see inset ofFigure 3b).In n-type doping (as per Hall-effect measurements), [32] the absorption signal decay originates from the band-to-band bimolecular recombination of mobile electrons and holes, and trapping and detrapping of holes via exponential shallow tail trap states A correspond to the proportionality constant and trapped-hole density, respectively.c) Time evolution of S(t), simulated densities of trapped holes p t (t), mobile electrons ∆n(t), and holes ∆p(t).d) Mapping of trapped holes p t (E, t) at energy E, ∆n(t), and ∆p(t) at various delay times t for P FL = 2.4 μJ.The evaluated material parameters are provided in Table S2 (Supporting Information).
(shown in Figure 3c).Owing to the detrapping process, the time evolution of the trapped-hole density p t A is given by p [33] which leads to power-law decay of S(t).As shown in Figure 3b, the n-type doping density n d , characteristic energy E 0 , and a parameter given by N t /(k t N m )  were determined.The initial values of the trap density N t and rate constant k t were estimated from the decay kinetics at P FL = 3 μJ, while keeping N t /(k t N m )  constant.The parameters were refined for good calibration of the numerical model the measured data at various P FL values (Figure S24, Supporting Information).Table S2 (Supporting Information) provides the initial and refined parameters obtained from TDRS analysis.
Figure 3c,d show the charge-carrier dynamics behind the S(t) decay kinetics at P FL = 2.4 μJ.The results indicate that the decay of the mobile photogenerated hole density ∆p(t) is significantly faster than that of the electron density ∆n(t).In addition to bimolecular recombination with electrons, the mobile holes are depleted and accumulated in shallow trap states.This leads to a steep rise in the trapped-hole density p t (t) with time t.After t = 100 ps, the holes are detrapped from shallow trap states to VB and eventually participate in band-to-band recombination with mobile electrons.As a result, relatively slow electron decay (relative to that of mobile holes) follows powerlaw-decay kinetics with time.The TDRS analysis gave a bimolecular recombination rate constant k r = 1.3 × 10 −9 cm 3 s −1 , which is in the range for well-recognized direct-bandgap semiconductors such as lead halide perovskites. [34]The evaluated n d = 2.5 × 10 19 cm −3 corresponds well to the reported value from Hall-effect measurements. [10]The trap states feature E 0 = 0.13 eV, N t = 1.3 × 10 20 cm −3 , and k t = 2.6 × 10 −9 cm 3 s −1 .The exact origin of such a high density of shallow trap states is currently unknown; it could originate from intrinsic grain defects, grain boundaries, and/or surface defects.Moreover, the charge carrier relaxation processes and extracted parameters are almost similar to that reported Ta 3 N 5 thin film, [24] indicating the Figure 4 shows the detailed optical and electrical simulation results for the optimized Ta 3 N 5 photoanode under AM 1.5G solar illumination.The details of simulations are provided in the Experimental Section.The results (Figure 4a) indicate that the normally incident light is significantly absorbed (95%) with minimal reflection (5%) and negligible transmission loss at wavelengths below 560 nm.The calculated theoretical current limit of 12.83 mA cm −2 for a bandgap energy of 2.08 eV shows that such optical characteristics limit the maximum current density to 11.5 mA cm −2 .Figure 4b shows that the current density at 1.23 V RHE (J a ) decreases to the measured value of 10.5 mA cm −2 at a charge-carrier diffusion length (L d ) of 99 nm.The simulation results are consistent with, and validated by, previous reports (see Figure S26, Supporting Information). [17,21]The J a significantly improved from 2.25 to 10.5 mA cm −2 with increasing L d from 3 to 99 nm (more than radius of nanorods).Both the diffusion length and the hole extraction rate from the NR surface limit the onset potential V on (ref.[21,35]).The rate at which the current rises with the potential V a (above V on ) decreases with increasing series resistance.At the V op in Figure 4c, the current J op is limited by a charge-carrier-recombination loss of 16%, in addition to an optical loss of 10%.The energy band diagram (Figure 4d) across the NR diameter shows a space charge region (SCR) with an upward band bending around the NR surface and quasi-neutral flat-band region (FBR) in the bulk NRs.The photogenerated minority holes diffuse from the FBR and selectively drift under the electric field of the SCR toward the NR surface.In contrast, the upward band bending drives electrons away from the SCR to the FBR, and electrons further diffuse to reach the GaN layer.Owing to high n d , the energy band diagram suggests that the SCR region is short (≈5 nm) as compared to FBR and charge carrier transport is majorly governed by diffusion process.Consequently, the holes and electrons (Figure 4e) are accumulated and depleted at the NR surface, respectively.As a result, the charge-carrier-recombination rate is maximum at the center of the NR, and decreases on approaching the NR surface, as shown in Figure 4f.Such charge-carrier distribution and recombination rate profiles lead to a slight decrease in the extracted flux (the difference between the generation rate G and recombination rate R) in the FBR compared with that in the SCR of Ta 3 N 5 -NRs.Our results indicate that the high current density of Ta 3 N 5 photoanode is achieved by suppressing charge carrier recombination and optical losses with the design of Ta 3 N 5 nanorods (e.g., nanorod's radius should be less than the charge-diffusion length).

Conclusions
We achieved a record STH energy conversion efficiency of 12.1% for unbiased and overall solar water splitting by using a tandem stack consisting of a Ta 3 N 5 photoanode and dual-CuInSe 2 photovoltaic elements.The obtained efficiency at the laboratory scale (Figure S27 and Tables S4-S8, Supporting Information) was significantly higher than the best reported values (up to 9%) for well-explored visible-light-absorbing Ta 3 N 5 , BiVO 4 , and Fe 2 O 3 photoanode based lead-free tandem technologies.Owing to decoupled optical and electrical path lengths, the developed photoanode efficiently absorbs solar light in the visible solar spectrum range (up to 600 nm) and uses most of the photogenerated minority holes for water oxidation.Although the STH energy conversion efficiency is ideal for commercialization, the stability of the photoanode requires further improvement by protecting the Ta 3 N 5 -NRs surface from self-photooxidation either by a hole-storage layer (e.g., gallium nitride, [18] AlO x , [36] or ferrihydrite [37] ) or an appropriate cocatalyst (e.g., NiCoFe-Bi) design. [38]A nanostructured photoanode of a visible-lightabsorbing oxysulfide (Y 2 Ti 2 O 5 S 2 [33,39] ) with a long charge-carrier lifetime of up to 6 ns could be a promising candidate for achieving highly efficient, durable, and stable photocatalytic activity.Additionally, TDRS and theoretical analysis identified the dominant charge-relaxation processes (i.e., bimolecular recombination and trapping/detrapping of holes), performance-limiting parameters, energetics, and charge transport for the optimized Ta 3 N 5 photoanode.These could be used to analyze or design narrow-energyband photocatalytic materials.This work represents a major step toward achieving highly efficient STH production for practical applications.

Experimental Section
][42] Initially, GaN/Al 2 O 3 substrates (10 mm × 15 mm × 0.3 mm) were pre-sputtered to deposit uniform Ta thin films, as per a previously reported method. [10]The thickness of the Ta thin film (TF) varied from 100 to 300 nm.The prepared Ta/GaN/Al 2 O 3 samples were mounted on the substrate holder of the GLAD system, which was positioned over the magnetron at an optimized distance of 12 cm and a glancing deposition angle of ≈86°to the substrate normal (see Figure S2, Supporting Information).The magnetron was equipped with a Ta target of purity 99.99% (Toshima Co., Japan).A mixture of Ar (flow rate: 15 sccm), O 2 (2.5 sccm), and N 2 (2.5 sccm) gases was used to drive reactive sputtering of TaO x N y .Pre-sputtering (≈25 min with a locked magnetron shutter) was performed to achieve a constant deposition rate of the sputtered material at a radiofrequency magnetron power of 325 W. Once a constant deposition rate had been achieved, the magnetron shutter was unlocked to initiate the GLAD of TaO x N y at a substrate rotation speed of 90 rpm and working pressure of 0.48 Pa.Well-separated, uniform, and vertically aligned TaO x N y NRs were fabricated on the Ta/GaN/Al 2 O 3 substrate.The as-grown amorphous TaO x N y -NRs and Ta phases were converted to crystalline Ta 3 N 5 by nitridation at an NH 3 flow rate of 50 sccm and a temperature of 1000 °C for 90 min.The prepared samples were naturally cooled to obtain semitransparent Ta 3 N 5 -NRs/Ta 3 N 5 -TF/GaN/Al 2 O 3 electrodes.
Cocatalyst Loading: A FeNiCoO x cocatalyst was loaded by using a combination of photo-electrodeposition and dipping techniques to promote water oxidation at the Ta 3 N 5 surface.For photo-electrodeposition, the Ta 3 N 5 electrode was inserted in a solution containing 2 mM NiSO 4 .6H 2 O (99.99% trace metal basis, Sigma Aldrich), 0.5 mM Co(NO 3 ) 2 .6H 2 O (99.99% trace metal basis, Sigma Aldrich), and 0.8 mM FeSO 4 .7H 2 O (99.99% trace metal basis, Sigma Aldrich) in 0.25 M potassium borate (K 2 B 4 O 7 .4H 2 O) buffer at pH 10.The FeNiCoO x cocatalyst was photoelectrodeposited on the Ta 3 N 5 -NRs surface for 11 min at a current density of 10 μA cm −2 under AM 1.5G simulated solar illumination.A standard three-electrode configuration was used for the cocatalyst deposition, with the Ta 3 N 5 specimen, Ag/AgCl, and Pt wire as the working, reference, and counter electrodes, respectively.After photo-electrodeposition, the electrode was washed with deionized water and dried in a N 2 stream.A dip-coating technique facilitated a uniform increase in the thickness of the FeNiCoO x cocatalyst on the NRs surface.For dip coating, Fe(III) 2ethylhexanoate (10 or 15 μL; Wako) and Ni(II) 2-ethylhexanoate (10 or 12.5 μL; Wako) were added to hexane (10 mL; high-performance liquid chromatography grade, Wako) to prepare a FeNi complex solution.Co(II) 2-ethylhexanoate (20 μL; Wako) was mixed with hexane (10 mL) to prepare a Co-based solution.Depending on whether the amount of Fe(III) 2-ethylhexanoate used was 10 or 15 μL, the Fe:Ni:Co ratio was 1:1:2 or 1:0.83:1.34 to improve the photoanode stability (Figure 2c; Figure S13, Supporting Information).Loading of the cocatalyst sample was achieved by sequential processes, namely dipping in the FeNi complex solution for ≈15 s, natural drying in air, dipping in the Co-based solution for ≈5 s, and natural drying in air.This dip-coating process was performed four times and then calcination was performed at 145 °C for 50 min in air.This led to the formation of a uniform FeNiCoO x cocatalyst layer on the surface of the Ta 3 N 5 -NRs (Figure 1c).

Material and Structural Characterization:
The crystal structure of the Ta 3 N 5 electrode was examined by X-ray diffraction (XRD; Cu Ka radiation, Smart Lab., Rigaku, Japan).The optical transmission spectra (Figures S8a  and S9, Supporting Information) were recorded with an ultraviolet-visible spectrometer (V-670, JASCO).The morphologies were determined by SEM (SU8020, Hitachi, Japan).The fine structure, crystallinity, and elemental composition were determined by using a TEM instrument (FEI Titan 80-300, Themofisher Scientific, USA) equipped with a Schottky field-emission gun operated at 300 kV.TEM images and nanobeam diffraction patterns were recorded with a Gatan UltraScan CCD camera (Gatan, Inc., USA).Cross-sections for TEM analysis were prepared with a Zeiss Auriga gallium focused ion beam system (Carl Zeiss AG, Germany) by using an in situ lift-out technique with Pt welding.Specimens for high-resolution TEM imaging were thinned by using an argon polisher (Fishione NanoMill Model 1040; E.A. Fischione Instruments, Inc., USA).The areas for elemental composition analysis were selected from high-angle annular dark-field (HAADF) images formed using a 0.5 nm per pixel probe and mapped using a Bruker XFlash 6-30 EDS detector (Bruker Co., USA) with a 30 mm 2 active area chip and energy resolution of 129 eV.The elemental composition maps were acquired within a 10 keV channel with a resolution of 0.59 nm per pixel and a dwell time of 16 384 μs.Drift correction was applied every 10 s.The orientations of the Ta 3 N 5 clusters were identified by using Single Crystal 4 Software; the previously reported Ta 3 N 5 crystal structure was used as a reference. [43]The surface chemical states of the samples were identified by XPS (PHI Quantera II spectrometer, ULVAC-PHI, Inc.) with Al K and Mg K X-ray sources.The binding energies in the obtained XP spectra were calibrated to the C1s peak of carbon at 284.4 eV.The position of the VBM of the Ta 3 N 5 -NRs was determined from PESA results (Model AC-3, Riken-Keiki Co., Ltd., Japan).
Photoelectrochemical (PEC) Characterization: PEC characterization of the photoanode was performed with a standard three-electrode configuration in a 1 M KOH aqueous solution electrolyte (pH 13.6) by using a potentiostat (HSV-110, Hokuto Denko, Japan).A Pt wire electrode, Ag/AgCl electrode, and the fabricated photoanode were used as the counter, reference, and working electrodes, respectively.The applied potential V Ag/AgCl between the photoanode and the Ag/AgCl electrode was converted to the potential V a on the reversible hydrogen electrode scale (V RHE ), by using the Nernst equation: Here, V 0 Ag∕AgCl = 0.198 V (at 25 °C) is the standard potential of the Ag/AgCl electrode.A cooling plate (As One Cool Plate, model SCP85) was used to prevent electrolyte heating and maintain the PEC cell at a constant temperature during measurements.A commercial solar simulator equipped with an AM 1.5G filter (XES-40S2, SAN-EI Electric Co., Ltd.) illuminated the photoanode, which was calibrated by using a spectroradiometer (EKO Instruments, LS-100; Figure S6, Supporting Information).Additionally, a standard silicon solar cell (certified by the National Renewable Energy Laboratory) was used as a secondary standard to check the intensity of the solar simulator before each experiment.Current-potential (I-V a ) curves were recorded at a potential scanning rate of 10 mV s −1 from the negative to positive bias.The measured current was normalized to the illuminated photoactive area (≈0.7-0.85 cm 2 ) of the photoanode to obtain the current density J.The photoactive area of the photoanode was determined by using Image J software.The IPCE was measured at an operating potential of 1.16 V RHE in the wavelength range 300-650 nm (with a 10 nm step) by using a tunable monochromatic light source (CT-10, Jasco Co., Japan).The intensity of the monochromatic light was calibrated by using a standard reference cell (silicon photodiode, S228/e009, Hamamatsu Photonics).The IPCE was evaluated by using the following equation: Here,  (nm) is the wavelength of the monochromatic incident light, J P (mA cm −2 ) is the photocurrent density, and P (mW cm −2 ) is the power density of the incident light.A gas chromatographic system (Model 3000, Inficon Co., Ltd.) was used to measure the amounts of evolved O 2 and H 2 gases from the photoanode and the Pt counter electrode, respectively, under simulated AM 1.5G solar illumination at an operating potential of 1.16 V RHE .Electrochemical impedance spectroscopy (EIS; SP-300 Potentiostat-Biologic equipment) was performed on the semi-transparent Ta 3 N 5 photoanodes at a potential of 1.16 V RHE with an AC amplitude of 10 mV and frequencies ranging from 0.1 to 100 kHz.The solar cell (individual CuInSe 2 cell with a total photoactive area of 0.45 cm 2 ) was characterized in ambient air without encapsulation by using a potentiostat (SP-300 Potentiostat-Biologic equipment) as a source meter and the same solar simulator as was used in the two-electrode configuration described above.The EQE was measured at an operating potential of 0 V in the wavelength range 300-1200 nm (with a 10 nm step) by using a calibrated tunable monochromatic light source (CT-10, Jasco Co., Japan).The short-circuit current was obtained from the EQE data.The integrated photocurrent from the solar cell was calculated by the integration of the EQE data over the standard AM 1.5G solar spectrum.
Construction of Tandem PEC Device: CuInSe 2 -based photovoltaic (PV) cells (Solar Frontier KK, Japan) were used to fabricate a two-seriesconnected CuInSe 2 element (dual-CuInSe 2 ), which was terminated with a Pt/Ni electrode, in accordance with the previously reported work. [10]In this configuration, the photogenerated electrons inside the dual-CuInSe 2 promote the hydrogen evolution reaction at the interface between the Pt/Ni electrode and the aqueous electrolyte solution.The PEC-PV tandem device was constructed by placing the semi-transparent Ta 3 N 5 photoanode in front of the dual-CuInSe 2 PV cells.The solar light (below 600 nm) was absorbed by the photoanode, and the rest of the transmitted light (above 600 nm) was absorbed by the dual-CuInSe 2 .The Ta 3 N 5 photoanode and dual-CuInSe 2 /Ni/Pt electrode were serially connected by lead wires via an amperemeter, and the current from the tandem device with time was recorded.A photoactive area of the PV cells without photoanode filtration of the solar spectrum of 0.9 cm 2 gave the current density versus potential characteristics shown in Figure 2b.The current density of the PV cells behind the photoanode was determined by normalization of the current at the photoactive area (0.7-0.85 cm 2 ) of the individual fabricated Ta 3 N 5 photoanode after encapsulation.The STH energy conversion efficiency  for the light-induced OWS reaction was calculated by using the following equation: Here, P in = 100 mW cm −2 and  F are the power density of the simulated AM 1.5G solar illumination and faradaic efficiency, respectively.Transient Diffuse Reflectance Spectroscopy: Transient diffuse reflectance spectroscopy (TDRS) is essentially similar to transient absorption (TA) spectroscopy.In the TDR measurements, the TA signal was detected in diffuse reflection mode because of the opaque nature of the target samples.The pump and probe lights were irradiated from Ta 3 N 5 layer side of Ta 3 N 5 photoanode.All the measurements were performed in air at 297 K.The TA intensity in diffuse reflection mode is expressed as percentage absorption (absorption (%)) and is calculated as 100 × (1 -R/R 0 ), where R and R 0 are the intensities of the diffusely reflected light with and without pump excitation, respectively.Femtosecond TDRS measurements (t < 3 ns) were performed using a Ti:sapphire laser with a regenerative amplifier (Spectra-Physics, Solstice; wavelength 800 nm, pulse width 100 fs, pulse energy 3.5 mJ per pulse, and repetition rate 1 kHz) as the light source.The output from the laser was split into four paths for the excitation of two optical parametric amplifiers (OPAs; Spectra-Physics, TOPAS Prime), white-light-continuum generation was achieved by focusing the fundamental light (800 nm) on a sapphire plate, and second-and thirdharmonic generations of the fundamental light (800 nm) were obtained by using -BaB 2 O 4 crystals.The second-harmonic light (400 nm) was used for the pump light, and the pump light intensity was varied by using neutral density filters from 0.075 to 4.5 μJ per pulse.For the probe light, infrared (IR) light of wavelength 8210 nm (0.15 eV) was generated from an OPA equipped with a difference-frequency-generation crystal.The delay time of the probe pulse relative to the pump pulse was controlled up to 3 ns by changing the optical path length of the pump pulse.The time resolution of the system was ≈140 fs.A liquid-nitrogen-cooled HgCdTe photodetector (Kolmar Technologies, KMPV11-1-J1) was used to detect the diffusely reflected probe light from the Ta 3 N 5 -NRs/TF/GaN/Al 2 O 3 .Both the pump and probe light were irradiated from the Ta 3 N 5 -NRs side.The diameter of the pump beam on the sample was ≈1 mm and the irradiated area of the pump beam was evaluated using a beam profiler (Newport LBP2-HR-VIS2).
In TDRS measurements for t > 3 ns, continuous-wave IR light of wavelength 8079 nm (0.15 eV) from a quantum cascade laser (Thorlabs, QD8050CM1 AC303) was used as the probe light source.The pump light, of wavelength 400 nm, was identical to that used in the measurements for t < 3 ns described above.The diffusely reflected probe light from the sample was detected by a liquid-nitrogen-cooled fast HgCdTe photodetector (Kolmar Technologies, KV104-0.25-A-2/11, bandwidth 80 MHz).The signal from the detector was preamplified by a voltage amplifier (Femto, DHPVA-200), and then amplified with a voltage amplifier (Femto, DUPVA-1-60), subsequently being processed and recorded with a digital oscilloscope (Lecroy, WaveRunner 6200 A).The pump-induced signal (AC signal) was selectively extracted by using the AC-coupled mode of the amplifier (Femto, DUPVA-1-60).The DC offset of the signal from the detector was independently recorded with a digital multimeter (National Instruments, USB-4065) and used to calculate the absorption value (Absorption (%)).Using this process, small TA signals (<0.01%) were detected with a time resolution of a few nanoseconds.
Photoluminescence (PL) Spectroscopy: Picosecond time-resolved PL spectroscopy was performed on the Ta 3 N 5 -NRs by using a streak camera (Hamamatsu Photonics, StreakScope C4334) equipped with a monochromator.A Ti:sapphire laser with a regenerative amplifier (Spectra-Physics, Solstice, wavelength 800 nm, pulse width 100 fs, pulse energy 3.5 mJ per pulse, and repetition rate 1 kHz) was used as a light source.Secondharmonic light (400 nm) generated from the fundamental light (800 nm) by using a -BaB 2 O 4 crystal was used for excitation.The excitation light was irradiated from Ta 3 N 5 layer side of Ta 3 N 5 photoanode.All the measurements were performed in air at 297 K.
Numerical Model for Charge-Carrier Dynamics: Figure 3 and Figure S22 (Supporting Information) present numerical models of relaxation of charge carriers behind measured absorption decay kinetics with time t.Pump light of energy 3.1 eV (greater than the bandgap energy of 2.1 eV) was absorbed with penetration to a depth of 34 nm from the top of the Ta 3 N 5 -NRs to produce initial electron and hole densities ∆n 0 of 1.23 × 10 20 cm −3 at a pump fluence intensity P FL of 3 μJ per pulse, as per the Lambert-Beer law.As previously discussed, the bimolecular bandto-band recombination (rate constant k r ) of photogenerated electrons and holes accounts for the increase in the absorption decay rate with increasing P FL .Power-law absorption decay originates from the band-to-band recombination of electrons and holes detrapped from shallow tail trap states.These shallow trap states, with a density N t , have an exponentially distributed energy E, which is given by N t g(E) = N t exp(−E/E 0 )/E 0 , where N t is the total trap density and E 0 is the characteristic energy.In addition to recombination with electrons, the photogenerated holes are trapped (rate constant k t ) and detrapped (rate constant k d ) via exponential trap states.For heavy n-type doping, the Fermi energy level E f is close to the CBM of E c , which indicates that the shallow tail states of VB are filled with electrons.The time evolution of the density of trapped holes p t (E,t) at energy E and time t is given by where k d = k t exp(−E/k B T) and N m is the effective density of states of VB.The total trapped-hole density p t (t) across all the trap states is given by p t (t) = ∫p t (E,t)dE.The dynamics of the hole density ∆p(t) and electron density are given by where n d is the n-type doping density.For simplicity, it is considered that the trap states were initially filled with electrons or the trapped-hole density p t (E,0) and p t (0) is zero; ∆p(0) and ∆n(0) are equal to ∆n 0 , which is calculated by using an absorption coefficient ( a ) of 2.91 × 10 5 cm −1 (ref.[44]) and P FL .Initial estimates of the material parameters were obtained by fitting the analytical expression for the trapped-hole density, p t A (t) = [N t /((1 + n d /∆n 0 )Γ(1 − )sin()(k t N m )  )]t − (previously derived for k r = k t , [33] ) to the measured data in Figure 3b.Here, the exponent  = k B T/E 0 , where k B and T are the Boltzmann constant and temperature, respectively.The numerical model is well fitted to the measured data for absorption signal decay for various P FL values (Figure 3c; Figure S24, Supporting Information).
Optical Simulations: Optical simulations of the Ta 3 N 5 photoanode were performed to obtain absorption, reflection, and transmission spectra at the wavelength  of the normally incident light.These characteristics determine the upper absorption limit of the current density, which can be obtained by assuming 100% extraction of photogenerated charge carriers, after accounting for the reflection and transmission (optical) losses.During the water-splitting operation, the normally incident AM 1.5G solar light propagates through the quartz glass, electrolyte, and Ta 3 N 5 -NRs before reaching the GaN substrate.The photoanode, therefore, consists of air/glass/electrolyte/Ta 3 N 5 /GaN.The dimensions of the Ta 3 N 5 , which consists of a thin-film interlayer, NRs, and GaN layers are similar to those obtained from the SEM image (Figure 1a) of the fabricated photoanode.The literature values for the refractive index n r and extinction coefficient k i [21] were used for the optical simulations.Light absorption is negligible, therefore the extinction coefficients for glass and the electrolyte were set at 0; the refractive indices are 1.47 and 1.33, respectively.The linearly polarized light (along the x-direction) in transverse electric field mode at a power of 1 W and wavelength  is normally incident at the glass/air interface.The periodic-boundary condition with the continuous electric field at the opposite periodic plane is considered to include light-scattering effects from the adjacent NRs.The wave optics module of COMSOL Multiphysics was used for physics-controlled meshing scheme discretization of the modeled device.The wave optics module was used to solve the Maxwell equations for the propagation of electromagnetic waves in the frequency domain to obtain the electric field at each discretized node.The phase-elimination method [45] was used to simulate the optical metrics (Figure 4a), i.e., absorption (A), reflection (R), and transmission (T) spectra versus wavelength.The integrated current density J I was evaluated as Here, S() is the power flux density (W cm −2 nm −1 ) at wavelength  of the AM 1.5G solar spectrum, h is Planck's constant, and c is the speed of light in a vacuum.In addition,  = 280 nm is the lower limit of the solar spectrum and  g is the absorption edge at 596 nm; X() corresponds to the IPCE and absorption in Figure 4a.The function X() was set at 1 (100%) to calculate the theoretical maximum limit.The Lambert-Beer law was used to calculate the charge-carrier generation rate G(z): A () S ()  hc  abs exp (− abs z) d (8)   Here,  abs = 4k i / is the absorption coefficient, z is the position along the direction of incident light from the top of the NRs, and A() corresponds to the absorption spectrum in Figure 4a.The scaling constant c 0 is adjusted so that the simulated current density at 1.23 V RHE from electrical simulations is equal to J I from the absorption spectrum.Electrical Simulations: Electrical simulations were performed to identify performance-dominating factors and quantify the electrical losses of the fabricated semi-transparent Ta 3 N 5 photoanode.In addition, these simulations provide information on the energetics, distributions of charge carriers, and recombination inside the Ta 3 N 5 -NRs at the operating potential V op .The energetics and charge transport are symmetrical in the x-y radial plane, therefore to reduce computational costs the 3D Ta 3 N 5 -NR was simplified to a 2D model spanning the direction of the incident light (or length in the z-direction) and the NR diameter in the x-direction.As shown in the SEM image in Figure 1a, the NR base interfaced with a Ta 3 N 5 thin film of thickness approximately 150 nm on the GaN layer.The GaN layer was modeled as an ohmic contact with selective electron extraction from Ta 3 N 5 .However, the Ta 3 N 5 surface creates a Schottky junction with the electrolyte.The boundary condition at the Ta 3 N 5 surface is therefore defined as a pseudo-metal contact with a certain work function and surface recombination velocity (or transfer rate) for charge carriers, which encapsulates the electrolyte and cocatalyst properties.In the oxygen evolution reaction, the overall hole-transfer kinetics from Ta 3 N 5 to the electrolyte is governed by the selective hole-transfer rate across the FeNiCoO x cocatalyst.The work function is considered to be in the middle of the redox potential, in accordance with the Nernst equation, at pH 13.6.The charge-carrier recombination and mobility were modeled as bimolecular band-to-band recombination and a constant effective mobility.The generation rate G obtained from optical simulations was used for generation of electrons and holes in Ta 3 N 5 for the AM 1.5G solar spectrum.The physicscontrolled algorithm of the COMSOL Multiphysics package was used to discretize the model with extremely fine meshing to obtain smooth variations in the simulated energy band diagram and distribution of charge carriers along the x-direction in the SCR.A steady-state solution at each applied potential V s on the ohmic GaN contact was achieved by performing self-consistent numerical calculations by using the fundamental Poisson's, current-drift-diffusion, and continuity equations for electrons and holes. [46][49] The currentpotential characteristics were simulated by varying the diffusion length, hole-extraction rate, and series resistance.The potential axis V s (in volts) was adjusted to V b on the V RHE scale by using V b = V RO /2 + V s , where V RO = 1.23 V is the difference between the reduction and oxidation potentials of water.The impact of the series resistance R s on the simulated performance characteristics was included by rescaling the potential axis as V a = V b + JR s , where J is the simulated current density.The material parameters used for the electrical simulations are provided in Table S3 of Supporting Information, which were obtained from TDRS analysis, values reported in the literature, and model calibration.

Figure 1 .
Figure 1.Material and structural properties.a) SEM image (side view) of Ta 3 N 5 -NRs/Ta 3 N 5 -thin film/GaN.b,c) High-resolution transmission electron microscopy images of FeNiCoO x /Ta 3 N 5 -NRs.d) Spot electron diffraction pattern of selected area in (b).e) STEM-EDS elemental mappings of selected area in (b) for Ta, N, O, Fe, Ni, and Co elements.In (c-e), the boundary colors correspond to selected areas in (b).

Figure 2 .
Figure 2. Performance evaluation of tandem device.a) Schematic diagram and working principle of tandem device comprised of serially connected semitransparent Ta 3 N 5 photoanode with dual-CuInSe 2 photovoltaic cells and Pt/Ni electrode.b) Current-potential (J-V a ) characteristics of dual-CuInSe 2 cells (dashed line: standalone; solid line: behind photoanode) and Ta 3 N 5 -NR photoanode.J op and V op (1.16 V RHE ) are the operating current and potential for unbiased solar water splitting.c) Evolution of current J with time t of tandem device (two-electrode configuration).The photoanode stability was improved from 1.4 to 6.7 h (for  > 10%) by optimizing the cocatalyst deposition process.d) Stochiometric (2:1) production of hydrogen and oxygen gases.The solid and dashed lines are estimates from the stabilized J-t curve and an average  = 11.07%at a faradaic efficiency of 98% (Figure S19, Supporting Information).

Figure 3 .
Figure 3. Decay analysis of TDR signal of Ta 3 N 5 nanorods.a) Pump fluence intensity (P FL )-dependent TDR signal S(t) with delay time t at pump photon energy of 3.1 eV and probe photon energy of 0.15 eV.S(t) follows power-law decay: At − (dotted lines) after ≈30 ns, where A and  are defined as the amplitude and exponent.b) Variation in A (at t = 0.4 ps), and  with P FL . and p tA correspond to the proportionality constant and trapped-hole density, respectively.c) Time evolution of S(t), simulated densities of trapped holes p t (t), mobile electrons ∆n(t), and holes ∆p(t).d) Mapping of trapped holes p t (E, t) at energy E, ∆n(t), and ∆p(t) at various delay times t for P FL = 2.4 μJ.The evaluated material parameters are provided in TableS2(Supporting Information).

Figure 4 .
Figure 4. Detailed optical and electrical simulations.a) Absorption (A), reflection (R), transmission (T), and IPCE spectra at V op = 1.16 V RHE against wavelength of normally incident light.The IPCE spectrum shadowed part of the absorption spectrum.b) Measured (solid) and simulated (dashed) current potential curves.c) Normalized losses (optical loss, charge-recombination loss) and extracted contribution to J op compared with the theoretical current limit of 12.83 mA cm −2 for bandgap energy of 2.08 eV.Changes in (d) energy band diagram, e) high-resolution electron density n and hole density p, f) generation rate G and recombination rate R with x along the NR diameter (200 nm from top of NR) at V op = 1.16 V RHE .In (d) the arrow directions represent electron and hole transport.SCR, FBR, E fn , and E fp correspond to the space charge region, flat-band region, and quasi-Fermi energy levels, respectively, for electrons and holes.