Introduction

TiO2 has been extensively investigated as a photoanode for photoelectrochemical (PEC) water splitting owing to its high photocatalytic activity, proper band-edge positions, superior photo-chemical stability, low-cost and non-toxicity1,2,3,4,5,6,7. Much research effort has been devoted to combat the large band gap of TiO2 (~3.0–3.2 eV)8,9 by using cationic10 or anionic doping (or alloying, especially for high doping concentration)11,12,13. Although monodoping (single element doping) reduces the band gap of TiO2 to some extent, it increases charge carrier recombination10,14 and eventually decreases the overall PEC performance. Even without doping, the performance of TiO2-based PEC water-splitting cells is greatly limited by its rapid charge recombination due to the high trap densities in TiO24.

Recently, a new donor–acceptor codoping concept was proposed to improve the PEC water-splitting performance of TiO2 photoanode because codoping TiO2 with donor–acceptor pairs can reduce charged defects and their associated recombination, improve material quality, enhance light absorption and increase solubility limits of dopants15,16,17,18,19. According to density-functional theory calculations, various donor–accepter pairs, such as (W, C)15,19, (Mo, C)15,16,19, (2Nb, C)15,18, (2Ta, C)15,18, (W, 2N)15,20, (Ta, N)15, (Nb, N)15, (Sb, N)21, (Cr, N)22, (Zr, S)15 and (Nb, P)15, have been proposed as good candidates for TiO2 codoping on the basis of the modified band gaps and band-edge positions. Nevertheless, up to date, there is only indirect experimental evidence for the effects of codoping for PEC water splitting. For example, a number of metal/anion codoping pairs were shown to improve the photocatalytic activities of TiO2 for organic pollutant decomposition due to enhanced visible-light absorption and charge separation22,23,24,25,26,27. Codoping ZnO with (Ga, N) led to highly conductive p-type ZnO film28, and improved PEC performance because of enhanced visible-light absorption and crystallinity29. Codoping LiFePO4 with (Si, F) showed enhanced electrical conductivity and thus led to improved cathode rate performance for Li-ion batteries.30

To our best knowledge, there is no direct experimental demonstration on the effect of codoping TiO2 for PEC application because of the lack of suitable doping methods for donor (metallic) dopants. Previous methods of doping TiO2 with metallic donors typically introduced the dopants during the synthesis stage of TiO2 with methods, such as sol-gel23,26,31, hydrothermal32, sputtering27 and solution combustion33. All these in-situ doping methods change the crystallinity/morphology compared with undoped TiO2, preventing the isolation of the doping effect from crystallinity/morphology effects on the PEC performance. Here we report a novel ex-situ sol-flame doping method to dope TiO2 nanowires (NWs) with metallic dopants (donor), and C (acceptor) is subsequently doped into TiO2 by conventional carbon monoxide (CO) annealing. The TiO2:(W, C) NWs double the saturation water oxidation photocurrent of undoped TiO2 for PEC water splitting and outperform the monodoped samples (TiO2:W and TiO2:C). Moreover, such significant PEC enhancement is because the codoping TiO2 with (W, C) significantly improves the electrical conductivity of TiO2 and modestly enhances its activity for oxygen-evolution reaction (OER) as well.

Results

Ex-situ codoping of TiO2 NWs

The ex-situ codoping method for TiO2 NWs is illustrated in Fig. 1. First, rutile TiO2 NWs with an average length of 2.8 μm and diameter of 100 nm are grown on fluorine-doped tin oxide (FTO) glass substrates by the hydrothermal method34, and the W-dopant precursor solution is prepared separately by the sol-gel method (see Methods section). Second, the TiO2 NWs are dip-coated with the W-precursor solution (0.05 M) and then dried by air blowing, which leaves a uniform thin layer of W precursor on the surface of TiO2 NWs. Next, the W-precursor-coated TiO2 NWs are annealed in a 1,100 °C flame for 2 min that promotes the rapid diffusion of W into TiO2 NWs. Finally, the W-doped TiO2 NWs are annealed in gaseous CO at 550 °C for 30 min in a tube furnace. As such, TiO2 NWs are successfully doped with the (W, C) donor–acceptor pair.

Figure 1: Schematic illustration of the ex-situ doping method.
figure 1

(a) The as-synthesized TiO2 NWs grown by the hydrothermal method are dip-coated with W-precursor solution and then dried by air (or N2) blowing. (b) The W-precursor-coated NWs are annealed in a high-temperature flame (1,100 °C for 2 min) to enable rapid W diffusion. (c) The W-doped TiO2 NWs are annealed in a tube furnace at 550 °C for 30 min under CO gas flow for C doping.

Full size image

Our ex-situ sequential doping method has several advantages. First, the most unique aspect of our doping method is the combination of sol-gel process and flame annealing (named as sol-flame) for doping W. Flame provides high temperature and an ultrafast heating rate to enable rapid dopant diffusion occurring in a very short time (a couple of minutes or less). Such brief high-temperature annealing minimizes morphology and crystallinity change of TiO2 NWs. The flame-annealed TiO2:W NWs have almost identical morphology (Supplementary Fig. S1a,b) and crystallinity (Supplementary Fig. S1e,f) with that of as-synthesized (undoped) TiO2 NWs. When the W-precursor-coated TiO2 NWs are annealed in a conventional furnace, the slow heating rate of the furnace limits the annealing temperature to be not more than 600 °C to protect the FTO substrates. As a result, WO3−x particles are formed on the surface of TiO2 NWs instead of diffusing into TiO2 (Supplementary Fig. S1c). Furthermore, even with rapid thermal annealing at 1,100 °C using the furnace (that is, preheating the furnace to 1,100 °C first and then quickly inserting the sample), cracks are generated in glass substrate due to thermal stress (Supplementary Fig. S1d), showing the advantage of flame annealing. Second, the ex-situ doping approach, that is, incorporating dopants after NW growth, can be broadly applied to various metal-oxide NWs regardless of their synthesis methods. Third, the sequential doping process enables independent control and optimization of doping conditions for each dopant (Supplementary Fig. S2). It should be noted that our doping methods are applicable to other donor–accepter pairs as well. We have tested other codoping pairs, including (Nb, C) and (Ta, C), and they also exhibit better PEC performance than the Nb, Ta or C monodoped TiO2 and undoped TiO2 (Supplementary Fig. S3).

Morphology and dopants distribution

The morphology and dopant profiles of TiO2:(W, C) NWs are illustrated in Fig. 2. The cross-sectional and top views (Fig. 2a) show that the TiO2:(W, C) NWs have an average diameter of 100±10 nm, length of 2.8±0.1 μm and surface coverage density around 6.3 × 109 NWs per cm2. The energy-dispersive X-ray spectrometer mapping, together with transmission electron microscope (TEM) image, (Fig. 2b), clearly shows that both W and C are doped over the entire TiO2 NW. The surface concentrations of W and C dopants are estimated to be ~10.2±2.0 at.% and 13.1±3.0 at.%, respectively, via a quantitative analysis of the low loss and core-loss electron energy loss spectroscopy (EELS) spectra (Supplementary Fig. S4). The W and C dopant radial depth profiles within TiO2 NWs are analysed by the secondary ion mass spectrometry (SIMS) measurements for which we use larger TiO2:(W, C) NWs with an average diameter of ~300 nm for higher secondary ion yield (Methods section and Supplementary Fig. S5). Figure 2c shows the radial depth profiles of W and C dopants, which are normalized by Ti and O counts, respectively. For both W and C, their depth profiles show that the dopant concentration is high near the surface and then decreases along the radial direction towards the centre. The relatively symmetric shape of W and C dopant profiles along the diameter of the NW indicates homogeneous radial diffusion of W and C. The SIMS profiles show that the radial diffusion depths for both W and C are about 60 nm, so our TiO2 NWs (~100 nm diameter) used for PEC tests are fully doped with W and C dopants, which is confirmed by EELS analysis (Supplementary Fig. S4).

Figure 2: Morphology and dopant-distribution characterization.
figure 2

(a) Top-view and cross-sectional scanning electron microscopy images. Scale bar, 500 nm (top view) and 1 um (cross-sectional). (b) TEM and energy-dispersive X-ray spectrometer mapping images. Scale bar, 100 nm. (c) SIMS analysis of radial depth profiles for W and C.

Full size image

PEC water-splitting performance

The PEC water-splitting performance of the codoped TiO2:(W, C) NWs is evaluated by measuring photocurrent density–potential (J–V) curves using a standard three-electrode configuration under air mass (AM) 1.5 G simulated solar light illumination (100 mW cm−2), and compared with undoped and monodoped (W and C) TiO2 NWs (Fig. 3a). The monodoped TiO2 NWs are prepared by using the identical doping conditions as the TiO2:(W, C) samples. First, the photocurrent density of the undoped TiO2 NWs is 0.75 mA cm−2 at 1.23 VRHE, which is comparable to previously reported values.34,35 Second, both monodoped TiO2:W and TiO2:C NWs, in comparison with the undoped TiO2 NWs, show much higher photocurrent density values, that is, 1.22 and 1.41 mA cm−2, respectively, at 1.23 VRHE. More importantly, the codoped TiO2:(W, C) NWs have the highest photocurrent density of 1.81 mA cm−2 at 1.23 VRHE, which is about 2.4 times higher than that of undoped TiO2 NWs, supporting the effectiveness of using donor–acceptor pairs to enhance PEC performance as suggested by previous theoretical calculations15,19. To the best of our knowledge, this is the first experimental demonstration for a donor–acceptor codoped TiO2 system that exhibits enhanced PEC water-splitting performance than monodoped TiO2. Finally, the stability of the codoped TiO2:(W, C) NWs is evaluated by measuring the photocurrent–time (J–t) response (Fig. 3b). In the given 30-min experimental time interval, no obvious degradation of the photocurrent is observed. Significantly, even after over 6 months of storage in a strong base (2 M KOH) solution, the J–V curve of the codoped TiO2:(W, C) NWs remains unchanged (Supplementary Fig. S6), indicating the excellent chemical stability of the TiO2:(W, C) NWs.

Figure 3: Comparison of PEC performance.
figure 3

(a) Photocurrent-potential (J–V) curves of undoped, W-doped, C-doped and (W, C)-codoped TiO2 NWs measured in a three-electrode setup (TiO2 working, Pt-wire counter, SCE reference electrode, scan rate of 50 mV s−1) and 1 M KOH electrolyte under simulated solar light illumination (AM 1.5 G, 100 mW cm−2). (b) The photocurrent density-time response shows good stability of TiO2:(W, C). (c) IPCE spectra measured at 1.23 V versus RHE. The inset shows a magnified view in the wavelength region of 400–500 nm.

Full size image

The incident photon-to-current conversion efficiencies (IPCEs), equivalent to external quantum efficiency, of four samples (undoped, W-doped, C-doped and W/C-codoped TiO2 NWs) are measured at 1.23 VRHE where the photocurrent density values for all the samples are saturated. The IPCE values are plotted versus wavelength in Fig. 3c to compare the wavelength-dependent photocurrent response for the four TiO2 NWs. First, the codoped TiO2:(W, C) NWs have the highest IPCE values (IPCEmax ~80% at 380 nm) throughout the whole wavelength region below the band gap (~420 nm), and then the IPCE values are lowered in the order of C-doped (IPCEmax ~66% at 385 nm), W-doped (IPCEmax ~64% at 390 nm) and undoped TiO2 (IPCEmax ~32% at 395 nm). Second, the codoped TiO2:(W, C) NWs shows noticeable though very weak IPCE response at the visible-light region (>420 nm).

Discussion

IPCE of a PEC photoelectrode is determined by the efficiencies of three fundamental processes involved in PEC water-splitting reaction as illustrated in equation (1), that is, charge generation efficiency (), charge transport efficiency within the material (ηtransport) and charge transfer (collection) efficiency at the electrode/electrolyte interface (ηtransfer).8

To identify the key factors responsible for the high IPCE for the TiO2:(W, C) NWs, the individual contributions of the above three factors are investigated separately. First, the light-absorption properties of the four TiO2 samples (undoped, W-doped, C-doped and W/C-codoped) are measured (see Methods section) and compared over a wavelength range from 300 to 650 nm (Fig. 4a). Figure 4a shows the light-absorption spectra of the four TiO2 NW arrays grown on quartz substrates. Although the TiO2:(W, C) NWs show slightly higher visible-light absorption tails in the wavelength region of 420–580 nm, its integrated light absorption (~64%) over 300–580 nm is comparable to the other three samples, indicating that codoping within our doping conditions has little impact on the light absorption or charge generation efficiency ().

Figure 4: Comparison of optical, charge transport and charge transfer properties.
figure 4

(a) Light absorption (plus scattering) of undoped (black), W-doped (blue), C-doped (green) and (W, C)-codoped TiO2 NW (red) arrays grown on quartz. The integrated light absorption (plus scattering) values calculated by integrating over the wavelength, are 59.5%, 60.7%, 61.1% and 64.0% for undoped, W-doped, C-doped, and (W, C)-codoped TiO2 NWs, respectively. (b) Charge transport efficiencies. (c) Charge transfer efficiencies. (d) Current–voltage curves of single NW. The inset shows the scanning electron microscopy image of single TiO2:(W, C) NW with metal (Pt) contacts at both ends. Scale bar, 1 μm.

Full size image

Furthermore, the charge transport efficiency (ηtransport) and charge transfer efficiency (ηtransfer) are decoupled and quantified by using H2O2 as a hole scavenger approach (see Methods section)36. The estimated charge transfer and transport efficiencies of the four TiO2 samples are compared in Fig. 4b. For the estimated charge transfer efficiency (Fig. 4b), because the rutile TiO2, especially the (110) terminated surface, has high catalytic activity towards OER,37,38, all of the four TiO2 samples show relatively high charge transfer efficiencies exceeding 70% throughout the whole potential range. Furthermore, the codoped TiO2:(W, C) NWs sample has the highest charge transfer efficiency (99%) at 1.23 VRHE, in comparison with the undoped (89%), W-doped (91%) and C-doped (96%) samples. The observed charge transfer efficiency difference suggests that the presence of donor–acceptor pair effectively reduces the surface recombination of charge carriers, especially at the low potential region (<1.0 VRHE). The OER activity is also improved on monodoped rutile TiO2 (110) surface that is consistent with the recent theoretical predication based on density-functional theory calculations.39 Previous experimental studies also suggest that the monodoping can induce high surface acidity and increase of reactant adsorption,40,41 enhancing the charge transfer process. Finally, though the codoped TiO2:(W, C) NWs improves the charge transfer efficiency approximately by 10%, the improvement is not large enough to explain its 2.4 times higher saturation photocurrent density value.

The third and last factor affecting the IPCE value is the charge transport efficiency, for which both electron and hole transport is important and mostly affected by the bulk recombination process. As shown in Fig. 4c, the codoped TiO2:(W, C) NWs sample has the highest estimated charge transport efficiency (50–80%), on average over two times higher than that of the undoped TiO2 NWs. The transport efficiency for monodoped TiO2:W and TiO2:C samples are in between those of the codoped and undoped samples, very similar to the PEC performance shown in Fig. 3a. The electron-transport property is further compared by measuring the conductivity of individual TiO2 NWs (see Methods section). Figure 4d shows linear current–voltage (I–V) curves for four individual TiO2 NWs, indicating good ohmic contact. The calculated electrical conductivity values of the undoped, W-doped, C-doped and W/C-codoped samples are ~0.0001, 0.37, 0.46 and 2.26 S cm−1, respectively. The W-doped, C-doped and W/C-codoped samples have over three orders of magnitude higher conductivity than the undoped sample, suggesting that doping TiO2 with W and/or C significantly improves the electron-transport property. It should be noted that codoping gives higher conductivity improvement than the monodoping cases, indicating the defect passivation effect by codoping of W (donor) and C (acceptor). On the other hand, the improvement of charge transport efficiency is much lower than the increase of the electrical conductivity, suggesting that the bulk recombination, likely hole transport, is still the bottleneck for the overall charge transport process. Considering all the three factors, that is, light absorption, charge transport and transfer properties, the most pronounced effect of codoping for TiO2 is to improve the charge transport property, that is, increasing electrical conductivity, and a minor effect is to improve the surface OER activity.

We further synthesized W/C-codoped TiO2 NWs with higher doping concentration (W/C no. 2) than the optimally codoped TiO2: (W, C) NWs (W/C no. 1, the sample discussed above) by using higher molar concentration of W-precursor solution (0.1 M) and longer duration of CO gas annealing (550 °C for 1 h). As shown in Fig. 5a–c, the W/C no. 2 has much higher W (17.5 at.%) and C (18.3 at.%) surface doping concentrations than the W/C no. 1 (optimal sample for PEC, W: 10.2 at.%, C: 13.1 at.% at the NW surface). Even though the W/C no. 2 has such high W- and C-doping concentrations, there is no nanoparticle formation at the TiO2 NW surface and the high-resolution TEM images at the region of A, B and C (Fig. 5a) also confirms that the W/C no. 2 remains to be highly crystalline without dopant segregation. In comparison with the W/C no. 1, the W/C no. 2 exhibits inferior PEC water-splitting performance with lower saturation photocurrent density and lower IPCE values (especially near the band-edge wavelength region, 360–400 nm) (Fig. 5d). Finally, both W/C no. 1 and W/C no. 2 exhibit <1% IPCE values in the visible-light wavelength region (420–550 nm). To identify the reasons for the lower performance of W/C no. 2, we again determine the individual efficiency of the three processes that governs the PEC water-splitting performance, that is, light absorption (charge generation), charge transport and charge transfer efficiency, as shown in Fig. 5f–h. First, the visible-light absorption (>420 nm) of the W/C no. 2 with higher W- and C-doping concentrations is obviously higher than that of the W/C no. 1 sample (Fig. 5f). Nevertheless, the IPCE in this visible-light region is nearly zero. Second, the charge transfer efficiency of W/C no. 2 is lower than that of W/C no. 1, especially at lower applied potential region (Fig. 5g). It is likely due to decreased OER activity by higher W and C concentration at the TiO2 surface10. Third, the charge transport efficiency of W/C no. 2 is consistently lower than that of W/C no. 1 over the entire applied potential range (Fig. 5h). This large decrease of charge transport efficiency in W/C no. 2 likely comes from the decreased conductivity due to the enhanced dopant scattering effect at much higher doping concentration.42 As a result, though the TiO2 NWs with higher doping concentration of W and C dopants exhibit increased visible-light absorption, their IPCE in the visible light range is still very weak (about 0.5%). In addition, the TiO2 NWs with higher doping concentration of W and C dopants exhibit lower charge transport and charge transfer efficiencies, and hence lower PEC performance than those of the optimal sample.

Figure 5: Characterization and PEC water-splitting performance of TiO2 NWs with high W- and C-doping concentrations.
figure 5

(a) TEM and high-resolution TEM images of W/C no. 2 (W: 17.5 at.%, C: 18.3 at.% at the NW surface). Scale bar, 50 nm (TEM) and 5 nm (high-resolution TEM). (b) STEM image. Scale bar, 50 nm. (c) Concentration profiles of W and C dopants along the NW diameter in W/C no. 2. Comparison of W/C no. 1 (blue) and W/C no. 2 (red) with (d), three electrodes J–V curves (W/C no. 1, W: 10.2 at.%, C: 13.1 at.% at the NW surface), (e) IPCE, (f) light absorption (quartz substrate), (g) charge transfer efficiency and (h) charge transport efficiency. The data of undoped TiO2 NWs (black) are added for reference.

Full size image

In summary, we report the first experimental demonstration that codoping TiO2 NWs with W and C indeed greatly enhances the PEC water-splitting performance compared to the monodoped and undoped TiO2 NWs. Our novel ex-situ sol-flame doping method effectively dopes TiO2 NWs with W without changing the morphology and crystallinity of TiO2 NWs and without damaging the delicate FTO substrate. The codoped TiO2:(W, C) NWs exhibit improved saturation photocurrent density than monodoped TiO2:W and TiO2:C and 2.4 times higher saturation photocurrent density than the undoped TiO2 NWs, and the enhancement is attributed to mainly charge transport efficiency improvement and minor charge transfer efficiency improvement. We believe that the effectiveness of our doping method and codoping strategies, together with theoretical guidance, will enable further improvement of the PEC water-splitting performance of many other metal-oxide photoanodes.

Methods

Synthesis of rutile TiO2 NWs array

The pristine TiO2 NW arrays were synthesized on TiO2 nanoparticle-coated FTO substrates using the hydrothermal method, and the details were reported in our previous work34. Briefly, for a typical synthesis, 0.6 ml of titanium (IV) butoxide (Aldrich Chemicals, 97%) was added into 50 ml of an aqueous HCl solution (25 ml of deionized (DI) water+25 ml of concentrated HCl (38%)) under magnetic stirring. For the growth of larger diameter TiO2 NWs that were used for the SIMS analysis, larger amount of Ti-butoxide (1 ml) was added and the growth substrate was quartz. After stirring for another 5 min, the solution was poured into a Teflon-lined stainless steel autoclave (100 ml capacity), and pieces of the TiO2 nanoparticle-coated FTO substrates, which have been cleaned for 10 min using an ultrasonicator in a mixed solution of DI water, acetone and isopropanol (IPA), were immersed in the solution. The autoclave was sealed and heated to 170 °C in an oven, and held at 170 °C for 7 h. The autoclave was then removed from the oven and allowed to cool to room temperature naturally on the bench top. The obtained products were washed with DI water and finally annealed at 450 °C for 1 h in air.

Preparation of W-dopant precursor solution

The W-dopant precursor solutions with metal salt concentrations of 0.01–0.2 M were prepared by using tungstic acid (H2WO4, 99%; Sigma-Aldrich Chemicals), hydrogen peroxide (H2O2, 30%; Sigma-Aldrich Chemicals) and 2-methoxyethanol (C3H8O2, >99%; Acros Organics) with a three-step sol-gel method: (1) dissolving the desired amount of H2WO4 into 6–10 ml of H2O2, (2) removing excess H2O2 and water by evaporation of the solution on a hot plate (130 °C) until the amount of the remained solution was about 1 ml and (3) mixing the 1 ml solution with 2-methoxyethanol by sonication (5 min) and stirring (1 h). After aging for 6–12 h, the prepared precursor solution was used for dip-coating the as-synthesized TiO2 NWs.

Doping TiO2 NWs with W and/or C

(i) To dope W into TiO2 NWs, the prepared W-precursor solution was coated over the entire surface of TiO2 NWs by the dip-coating method43,44. After drying by air (or N2) blowing and subsequent air-annealing at 80 °C for 10 min, the W-precursor-coated TiO2 NWs were annealed in a premixed flame of 1,100 °C for 2 min with a stoichiometric fuel-to-oxidizer equivalence ratio (Φ) of 1.0 (the flow rates of CH4 and air were 2.0 and 25.5 SLPM (standard liters per minute), respectively).43,44 The stoichiometric condition and usage of CH4 minimizes the possibility of forming oxygen vacancy and carbon deposition on the TiO2 NWs. (ii) For the C doping, the as-synthesized or W-doped TiO2 NWs were annealed in a tube furnace (550 °C for 30 min) with 30 sccm CO gas flow. For codoping of W and C, sequential doping process of (i) and (ii) was used. The dependence of PEC performance on the doping conditions is discussed in detail in Supplementary Fig. S2.

Material characterizations and dopant depth profiles

The morphologies of undoped and doped TiO2 NWs were investigated using a scanning electron microscopy (FEI XL30; Sirion). The dopant distribution and surface concentration were investigated using a TEM (F20 FEI Tecnai, 200 kV) equipped with an energy-dispersive X-ray spectrometer and an EELS (Supplementary Fig. S4). SIMS measurements were conducted to analyse the radial depth profiles of W and C in TiO2:(W, C) NWs. To increase the secondary ion yield for SIMS measurement, larger diameter TiO2 NWs (~300 nm) were grown on quartz substrates using larger amount of Ti-butoxide (1 ml) addition. After doping with W and C, the TiO2:(W, C) NWs were sonicated in IPA. The NW suspension was then drop-cast onto a piece of Si (100) wafer to disperse TiO2 NWs on the Si substrate. To remove the solvent residue, the NW-deposited Si substrate was washed with DI water and dried in a vacuum. The SIMS measurements were carried out using a Cameca NanoSIMS-50. A Cs+ primary ion beam, accelerated at 16 keV with a 3.6 pA beam current, was used to sputter the surface of the NWs. The 12C, 18O, 28Si, 48Ti, 48Ti16O and 184W16O secondary ions and ionic fragments were collected and analysed in a mass spectrometer. The rastered area was 1 × 1 μm2 and the gated/analysis area was the centre 0.25 μm2 area to ensure that the collected counts only come from the NWs (Supplementary Fig. S5). The sputtering rate was estimated by using the TiO2 NW diameter and the depth profile of 28Si.

Optical measurements

As FTO and the underlying glass can contribute to background absorption in the longer wavelength regime, TiO2 NWs for the optical measurements were grown on the quartz substrate instead. The wavelength-dependent optical absorption properties were obtained with an integrating sphere using a xenon lamp coupled to a monochromator (Model QEX7; PV Measurements, Inc.). The absorption plus scattering was calculated from these measurements with the formula A+S=1−T−R, where A is the absorption, S is the scattered light that was not accounted for in the measurement, T is the total transmittance and R is the total reflectance.

Electronic transport properties measurement

To measure electronic transport properties of single TiO2 NWs, the NWs were sonicated in IPA and then drop-cast onto a SiO2/Si substrate45. Platinum (Pt) metal contacts were deposited on the both ends of each TiO2 NWs by using the focused ion beam (FEI Strata DB235)46. Current–voltage (I–V) measurements were conducted at room temperature using a semiconductor analyser (Model 4200-SCS; Keithley) with tungsten probes.

PEC measurements

The PEC performance of undoped and doped TiO2 NWs was measured using a potentiostat (Model SP-200; BioLogic) in a three-electrode configuration (a saturated calomel reference electrode (SCE) and a Pt-wire counter electrode) under illumination of simulated solar light (AM 1.5 G, 100 mW cm−2) using solar simulator (Model 94306A, Class AAA; Oriel). Before the measurement, the solar simulator intensity was calibrated with a reference silicon solar cell and a readout metre for solar simulator irradiance (Mode 91150V; Newport). The illuminated area of the working TiO2 electrode was 0.63 cm2 defined by a mask. A volume of 1 M KOH or 1 M KOH with 0.5 M H2O2 solution was used as the electrolyte. During the PEC measurement, the electrolyte was deaerated by Ar purging to remove the dissolved oxygen. In a typical J–V measurement, the voltage was swept linearly from −1.0 to 2.0 VSCE at a scan rate of 50 mV s−1. The IPCE was measured at 1.23 VRHE (=~0.2 VSCE) using a specially designed IPCE system for solar cell (QEX7; PV measurements), with three-electrode configuration. A 75-W Xe lamp equipped with a monochromator (CM-110, 1/8; SP Spectra Product) was used to generate a monochromatic beam. The incident light intensity was calibrated by a standard silicon photodiode. The charge transport and transfer efficiencies were estimated as functions of applied potential by using H2O2 as a hole scavenger under AM 1.5 G simulated solar light illumination36. The key assumption for this approach is that the oxidation kinetics of H2O2 is very fast and its charge transfer efficiency is 100%, so the ratio of photocurrent density measured in H2O and H2O2 gives the charge transfer efficiency (ηtransfer) for H2O (equation 2). The charge transport efficiency (ηtransport) was further calculated by dividing photocurrent density in H2O2 by the total light-absorption efficiency (), which is obtained from integration of light absorption (Fig. 4a) with respect to the AM 1.5 G solar light spectrum (equation 3).

Additional information

How to cite this article: Cho, I. S. et al. Codoping titanium dioxide nanowires with tungsten and carbon for enhanced photoelectrochemical performance. Nat. Commun. 4:1723 doi: 10.1038/ncomms2729 (2013).