Core/Shell Nanowire Arrays with Enhanced Photoactivity

There are currently great needs to develop low-cost inorganic materials that can efficiently perform solar water splitting as photoelectrolysis of water into hydrogen and oxygen has significant potential to provide clean energy. We investigate the Si/TiO2 nanowire heterostructures to determine their potential for the photooxidation of water. We observed that highly dense Si/TiO2 core/shell nanowire arrays enhanced the photocurrent by 2.5 times compared to planar Si/TiO2 structure due to their low reflectance and high surface area. We also showed that n-Si/n-TiO2 nanowire arrays exhibited a larger photocurrent and open circuit voltage than p-Si/n-TiO2 nanowires due to a barrier at the heterojunction. Direct solar energy conversion to storable fuels such as hydrogen offers a promising route toward less reliance on fossil fuels.1-4 For example, photoelectrolysis of water to generate H2 on a semiconductor/electrolyte interface has the attractive advantages of clean processing and energy savings over steam reforming of natural gas. One of the most critical issues in solar water splitting is the development of a photoanode with high efficiency and long-term durability in an aqueous environment. TiO2 has been extensively studied as a photoanode due to its high resistance to photocorrosion.5-10 However, its conversion efficiency of solar energy to hydrogen is still low (less than 4%)6 due to its large bandgap (3.0∼3.2 eV). TiO2 requires an external bias to reduce water for H2 production to overcome the chemical over potential.5,11 On the other hand, Si (Eg ) 1.12 eV) can absorb sunlight efficiently. However, it is challenging to use Si for photoelectrolysis since it readily corrodes in water. Moreover, it is thermodynamically impossible for Si to oxidize water spontaneously due to its high valence band maximum (VBM) energy. Therefore, a composite semiconductor electrode composed of a semiconductor heterojunction has been proposed to compensate for these shortcomings. In these cases, the photoanode is composed of a small band gap semiconductor that is protected by a stable semiconductor.12-14 Semiconductor heterojunctions can absorb a different region of the solar spectrum.15-18 The advantage of composite structures is that each semiconductor needs to satisfy one energetic requirement: matching the conduction band minimum (CBM) or VBM with either the H2 reduction or O2 oxidation potential. Single semiconductor materials typically cannot satisfy the requirements of suitable bandgap energies for efficient solar absorption and meantime with band-edges aligned with both the H2 and O2 redox potential of water.3,19 Here, we prepared TiO2 coated Si nanowire arrays and studied their photo-oxidative properties. We observed that Si/TiO2 core/shell nanowire arrays showed higher photocurrent than the planar Si/TiO2. A semiconductor heterojunction of n-Si/n-TiO2 or p-Si/n-TiO2 has different band bending properties near the junction.12,20 The n/n heterojunction has a potential energy barrier between the two semiconductor regions that reflects minority holes in the TiO2 similar to the back surface field in solar cell.21 Using photocurrent and open circuit voltage measurements, we show that the n/n heterojunction is more promising for photoelectrochemical (PEC) cell application. Highly oriented Si nanowire arrays on the silicon wafers were synthesized by an aqueous electroless etching method.22 These Si nanowire arrays significantly suppress reflection that has the potential to provide a higher efficiency of PEC cell. We prepared n-type and p-type Si electroless etched nanowire (EENW) arrays from n-Si(100) (P doped, 0.6∼0.8 Ωcm) and p-Si(100) (B doped, 1∼5 Ωcm) wafers with dopant concentrations of ∼1016 cm-3. A clean silicon wafer was immersed into the etching solution containing 0.04 M AgNO3 (99.999%, Aldrich) and 5 M HF (49%, Honeywell) at room temperature. The length of the EENW was controlled by the etching time (0.2 μm/min growth rate). * Towhomcorrespondenceshouldbeaddressed.E-mail:[email protected]. NANO LETTERS 2009 Vol. 9, No. 1 410-415 10.1021/nl8032763 CCC: $40.75  2009 American Chemical Society Published on Web 12/01/2008 TiO2 was grown both on the Si EENW arrays and the planar Si wafer by a home-built atomic layer deposition (ALD) system with TiCl4 (99.990%, Alfa) and pure water as the precursors. Si samples were cleaned with a buffered HF solution to remove the native oxide layer right before loading into the ALD vacuum chamber. The ALD system deposits polycrystalline anatase TiO2 layer with an average growth rate of 1.2 Å per cycle on the planar Si wafer and 0.7 Å per cycle on our Si EENW arrays at 300 °C. On the Si EENW array, diffusion of the gas precursors is reduced due to the high density of the nanowire arrays, which leads to slower growth rate than the planar substrate. Care was taken to ensure that the TiO2 thickness on the EENW arrays and planar substrates were equivalent. Typical cross-sectional scanning electron microscope (SEM) and the transmission electron microscope (TEM) images of Si EENW and the ALD TiO2 coated Si EENW (Si EENW/TiO2) are shown in Figure 1a-d. The Si EENW and Si EENW/TiO2 arrays are produced vertically on the Si wafer with high density. Si EENW maintains the same crystallinity as the starting Si wafer,22 and their diameters are in the range of 20∼200 nm. Figure 1d shows polycrystalline TiO2 deposition on the EENW. Figure 1e,f shows the top view SEM images of Si EENW/TiO2 arrays and TiO2 thin film on the Si(100) wafer with average thickness of 35 nm. The nanocrystalline nature of the TiO2 coating is similar for deposition on the nanowire surface and on the flat wafer surface. The X-ray diffraction (XRD) characterization indicates that the ALD grown TiO2 layer has an anatase structure both on the Si(100) wafer and on the Si EENW array, as shown in Figure 2. For photoelectrolysis of water, anatase TiO2 has the advantage of a flat band potential (Ufb) that is 200 mV more negative than that of the rutile TiO2. This allows anatase TiO2 to have sufficient cathodic potential for hydrogen reduction from water.23 The carrier concentration (ND) of TiO2 layer by ALD was determined by capacitance versus voltage measurement. TiO2 thin film (90 nm) was deposited on the highly doped n-Si wafer (As doped, 0.001∼0.004 Ωcm), and Ni was deposited on the TiO2 thin film. From the Mott-Schottky relation, ND was found to be 2.76 × 1017 cm-3. The TiO2 becomes n-type semiconductor because of defects such as oxygen vacancies and titanium interstitials, and their carrier concentration varies from ∼1017 to ∼1020 cm-3 depending on synthesis. 24-26 Photocurrent measurements were performed in a 1 M KOH electrolyte with three electrodes configuration (EG&E Pinceton Applied Research Potentiostat, VersaStat II): Si/TiO2 photoanode as a working electrode, Pt gauze as a counter electrode, and a saturated calomel electrode (SCE, Pine Research Instrumentations, AFREF1) as a reference electrode. All three electrodes are in a glass cell which has a 1 in. quartz window, and Ar gas was bubbled through to remove the dissolved oxygen during the measurement. The current versus potential measurements were carried out at a 10 mV/s sweep rate. A constant light intensity of 100mW/ cm2 from a 450 W Xe lamp (Oriel, 6266) illuminated our samples, and a liquid filter (Oriel, 6123NS) was used to avoid solution heating by infrared light. Figure 3a shows the photocurrent versus bias potential characteristics for Si/TiO2 composite photoanodes. Under illumination, oxidation of water takes place on the photoanode 4OH+ 4hf 2H2O+O2 E 0 ) + 0.160 V (vs SCE) The photocurrent versus bias potential curves have three regions: low photocurrent density region at negative bias Figure 1. Characterization of Si EENW and Si/TiO2 core/shell structures. (a) Cross-sectional SEM of 20 μm long Si EENW arrays, demonstrating vertical alignment and high density. (b) Typical TEM image of a Si EENW. (c) Cross-sectional SEM images of Si EENW arrays coated with TiO2 by ALD at 300 °C. (d) Typical TEM image of Si EENW/TiO2 core/shell nanowire, showing that polycrystalline TiO2 covers the Si EENW. (e) Top view SEM images of Si EENW/ TiO2 arrays. (f) Top view SEM image of ALD grown TiO2 thin film on a Si wafer, showing nanometer size grains of TiO2. Figure 2. X-ray diffraction pattern of TiO2 layer grown by ALD at 300 °C. 25∼40 nm TiO2 deposited on Si EENW arrays (red line), 35 nm TiO2 thin film on a Si planar substrate (black line). Both TiO2 layers index to polycrystalline anatase. Nano Lett., Vol. 9, No. 1, 2009 411 potential (region I), plateau of the photocurrent density at more positive bias potential (region III), and increasing photocurrent density region (region II) between regions I and III. No photocurrent passes through the semiconductor and electrolyte interface when the negative bias voltage is close to the flat band potential because any created excess holes and electrons are recombined before holes transfer into the electrolyte (region I).27 The photocurrent plateau appears as the bias potential sweeps to more positive direction (region III), where the photocurrent is limited by the number of the holes excited by illumination. Our planar Si/TiO2 samples show comparable photocurrent density to those reported in the literature. The reported value9 for a 15 μm thick film of P-25 TiO2 on Ti foil is 0.1 mA/ cm2 under the same illumination conditions (Xe lamp, 100 mW/cm2) even though the thickness of our ALD grown TiO2 film is only 35 nm. At region III, 20 μm long Si EENW/ TiO2 samples show 2.5 times higher photocurrent density than planar Si/TiO2 samples for both n-type and p-type Si. The Si EENW/TiO2 composites have higher photocurrent mainly because of lower reflectance and larger surface area than the Si planar/TiO2. Figure 3a also shows that n-Si/n-TiO2 composites have 20∼25% higher photocurrent density and more negative onset potential than those of p-Si/n-TiO2 for both nanowire and planar structures. Higher photocurrent is expected for the n/n junction since band bending at the junction helps charge separation. The band bending of the semiconductors at the junctions are shown in Figure 3b,c. The Fermi energy (EF) of our n and p type silicon are -4.25 and -4.97 eV (relative to the vacuum level), respectively. EF of TiO2 was calculated28,29 from the reported electron effective mass in anatase (me* ) 1me) and the measured carrier concentration (ND ) 2.76 × 1017 cm-3). Figure 3b illustrates the charge flow in n-Si/n-TiO2 junctions under illumination. The e-/h+ pairs are created inside both the Si and TiO2 because the TiO2 shell is transparent under visible light. Visible light can be harvested by the core Si. Under the illumination, Fermi energies of the electrons and holes, quasi-Fermi energies (EF*), differ from EF in dark,19 and the quasi Fermi Figure 3. (a) Photocurrent density versus bias potential (vs SCE) of Si/TiO2 photoanodes: n-Si EENW array coated by TiO2 (red), p-Si EENW array coated by TiO2 (blue), n-Si(100) planar substrate coated by TiO2 (green), p-Si(100) planar substrate coated by TiO2 (purple). Schematic representation of band energies and charge transfer (b) for n-Si/n-TiO2 and (c) for p-Si/n-TiO2 under the illumination. Si EENW/ TiO2 samples have 2.5 times larger photocurrent density than Si planar/TiO2. n-Si/n-TiO2 photoanodes have more negative onset potential than p-Si/n-TiO2 both for planar and nanowire structures. 412 Nano Lett., Vol. 9, No. 1, 2009 energies of minority holes (pEF*) in the n-Si and n-TiO2 are shown. The photogenerated hole in TiO2 (TiO2h +) moves toward the n-TiO2/electrolyte interface and oxidizes OHto oxygen, while photogenerated electrons in the n-TiO2 (TiO2e -) move away from the front surface due to the schottky barrier at the interface with the electrolyte. In addition to this charge separation, the interface between the n-Si and n-TiO2 reduces the loss of holes in the TiO2 region which results in an increase of the photoanodic current. The potential barrier seen by the holes at the n-Si/ n-TiO2 junction reflects holes back into the TiO2 layer (Figure 3b). This is analogous to the back surface field of a solar cell which has shown larger photocurrent and larger output voltage by adding a heavily doped region adjacent to the contact.31 To complete the circuit, the photogenerated electrons in the n-Si (Sie) move to the counter electrode where the reduction reaction takes place. The photogenerated hole in n-Si (Sih) moves toward the n-Si/n-TiO2 junction and recombines with the TiO2e -. Therefore, n-Si/n-TiO2 core/ shell structure shows the largest increase in photocurrent since its band alignment at the junction helps reduce recombination under illumination. In the case of p-Si/n-TiO2 junctions, the flow of electrons and holes at the junction of p-Si/n-TiO2 is opposite to the desirable direction. Figure 3c shows that TiO2h + can move either to the electrolyte or to the p-Si in p/n junction. Therefore, the p-Si/n-TiO2 junction has smaller observed photoanodic current density than n-Si/nTiO2. The larger photocurrent in the n-Si/n-TiO2 leads to a larger negative onset potential. This is a result of the n/n junction’s effective charge separation that leads to a larger short circuit current (Jsc). Both n-Si EENW/TiO2 and p-Si EENW/TiO2 have similar dark current values (5 μA/cm2). Also, higher Voc is expected for n-Si/n-TiO2 since the EF and band energies of the n-Si and n-TiO2 shift upward at open circuit under the illumination. The Voc of n-Si/n-TiO2 photoanode is Voc ) Voc(TiO2/electrolyte) + Voc(Si/TiO2). For the p-Si/n-TiO2, the band energies of the n-TiO2 shift upward but the band energies of the p-Si shift downward. The photovoltage at the p-Si/n-TiO2 junction is in the opposite direction to the photovoltage at the n-TiO2/electrolyte interface due to the downward band bending. The Voc of p-Si/n-TiO2 photoanode is Voc ) Voc(TiO2/electrolyte) Voc(Si/TiO2). We can take the advantage of the higher Voc of the n/n junction for the PEC cell. Enhanced Jsc and Voc will provide a higher efficiency PEC cell. Also, for the solar water splitting, larger Voc provides high enough cathodic potential to reduce the water to hydrogen. It is important that the flat band potential of the semiconductor is lower than the hydrogen reduction potential. For example, Fe2O3 (Eg ) 2.1 eV) and WO3 (Eg ) 2.6 eV) have been studied due to higher stability and lower band gap than TiO2. However, both need an external bias voltage to complete the water splitting since their CBMs are lower than hydrogen reduction potential by 0.2 and 0.1 V.6 Therefore, the n/n composites have the potential for the spontaneous photoelectrolysis of water. Stability is another important requirement for the PEC cell. By coating the Si with TiO2, we can make the photoanode stable in the 1 M KOH aqueous solution. The n and p type Si EENW/TiO2 core/shell structures have shown constant photocurrent levels while testing for one hour (Supporting Information Figure S1). In contrast, planar Si wafers and Si EENW arrays generate vigorous H2(g) bubbles in the KOH electrolyte and as a result they are etched. Figure 4 demonstrates the photocurrent density depending on the length of the n-Si EENW/TiO2 arrays. We prepared 5, 10, and 20 μm long n-type Si EENW/TiO2 arrays and planar n-Si/TiO2. We observed that the longer Si EENW/ TiO2 arrays have higher photocurrent although the photocurrent density normalized by the length of the nanowires decrease as shown in Figure 4b. All of the three Si EENW/ TiO2 arrays had significantly lower reflectance than Si planar/ TiO2 sample (Supporting Information Figure S2). The Si EENW/TiO2 sample can effectively trap the light by extending the path length due to multiple reflection in a high density array structures similar to textured surfaces.32 When the wires are 5∼20 μm long, their reflectance values are nearly the same. All of the three different length Si EENW/TiO2 arrays have about 1∼2% of reflectance in the UV region (200∼350 nm), and about 5% of reflectance in the visible region (450 ∼900 nm). In addition to low reflectance, the high surface area of Si EENW is also expected to contribute to the higher photocurrent since it increases the interface area with the Figure 4. (a) Variation of photocurrent density versus potential depending on the length of n-Si EENW/TiO2 arrays: 20 μm (blue line), 10 μm (green line), 5 μm (red line) long NW arrays and n-Si planar/TiO2 (black line). (b) Relationship between photocurrent density versus the length of n-Si EENW/TiO2, illustrating longer wire arrays have higher photocurrent. The axis to the right is the current density normalized by the length of the nanowire. Nano Lett., Vol. 9, No. 1, 2009 413 electrolyte as well as the overall amount of TiO2. Lower reflectance and higher surface area of the Si EENW/TiO2 contribute to higher absorption and higher photocurrent density than the planar samples. In order to ascertain the contribution of the core Si, we measured the photocurrent and Voc of the Si/TiO2 photoanodes under the visible light illumination only. The light was passed through a 441.6 nm edge filter to cut off the UV region from the Xe lamp so that carriers are generated only in the Si. We observed that there was no photocurrent under the visible light for all Si/TiO2 composites. Photogenerated holes in Si cannot be transferred to the valence band of TiO2 since there is a significant barrier at the junction (Figure 5c,d). Instead, the photogenerated e-/h+ pairs in the Si recombine so that there is no net charge flux.19,33 Therefore, photo-oxidation cannot take place at the TiO2 surface unless carriers are photogenerated in the TiO2 shell. The Voc shifts under visible light illumination only as shown in Figure 5a,b. For n-Si/n-TiO2, the photogenerated holes in the Si move toward the TiO2 and recombine with electrons in TiO2 due to an electric field in space charge region. The charge separation shifts the EF* and the band energies of the n-Si upward, so that the space charge region diminishes. When flat band is attained, there will be no more charge separation. Therefore, the Voc of the n-Si/n-TiO2 photoanode becomes more negative (Figure 5a,c). For p-Si/ n-TiO2, the band bending of the p-Si diminishes similar to n-Si/n-TiO2, but the reduced band bending shifts the band energies of p-Si downward which reduces the Voc under visible light (Figure 5b,d). From the change of the Voc under visible light, we confirmed that the core Si absorbs the visible light and it contributes shift of the band energies. In conclusion, we compared the photocurrent density of the planar Si and Si EENW coated by ALD TiO2 thin film. The Si EENW/TiO2 has 2.5 times higher photocurrent density than the planar Si/TiO2 due to lower reflectance and higher surface area. We also observed an increase of the photocurrent by using n-Si/n-TiO2 heterojunctions because n/n junctions enhance the charge separation and minimize recombination. The n/n heterojunction is a promising structure for solar water splitting since the photovoltage at the junction can compensate the lower energy level of the conduction band of the shell semiconductor. Also, the n/n heterojunction could potentially increase the efficiency of the photovoltaic cell due to a higher open circuit voltage and higher photocurrent. Acknowledgment. This work was supported by the Director, Office of Basic Energy Sciences, Chemical Sciences, Materials Sciences and Engineering Division, of the Figure 5. Open circuit voltage (Voc) versus elapsed time for (a) n-Si EENW/TiO2 and (b) p-Si EENW/TiO2 arrays under visible light illumination only (<420 nm, ON) and in dark (OFF), and schematic diagram of band bending and Voc for (c) n-Si EENW/TiO2 and (d) p-Si EENW/TiO2 arrays under visible light illumination, demonstrating that n-Si EENW/TiO2 increase Voc the while the p-Si EENW/TiO2 decrease the Voc under visible light. 414 Nano Lett., Vol. 9, No. 1, 2009 U.S. Department of Energy under Contract No. DE-AC0205CH11231. We thank Professor Nate Lewis and his group for helpful discussion and assistance in the initial testing of our samples. Supporting Information Available: This material is available free of charge via the Internet at http://pubs.acs.org.