Enhanced photoelectrochemical performance of InGaN-based nanowire photoanodes by optimizing the ionized dopant concentration

InGaN-based nanowires (NWs) have attracted increasing interest for applications in photoelectrochemical (PEC) water splitting due to their chemical stability and tunable bandgap.^1–6 By tuning the In composition during growth, the bandgap of InGaN can cover the entire solar spectrum^1,7–9 while straddling the water splitting redox potentials.^10,11 Moreover, due to their three-dimensional (3D) geometry, NW-based photoelectrodes provide large surface-to-volume ratios that offer more benefits to the carrier separation process^8,12 and increase the available surface area for electrochemical reactions.¹³ Therefore, InGaN-based NWs are potential candidates for efficient PEC water splitting systems.

Several approaches have been applied to improve the PEC performance of InGaN-based photoelectrolysis systems, such as controlling charge carrier transport,¹⁴ engineering the bandgap,¹⁵ or combining with metal cocatalysts.⁶ In PEC water splitting, following light irradiation of an n-type semiconductor photoanode and separation of the photogenerated electron-hole pairs, the holes diffuse to the semiconductor/electrolyte interface, while the electrons collect at the metal counter electrode through the ohmic contact. To improve the collection efficiency of the photogenerated electrons, optimization of the charge carrier concentration is crucial. In the case of nanostructured photoelectrodes, the high density of surface states may also lead to severe band bending when they are in a direct contact with the electrolyte. Controlling the dopant concentration has been reported to help reduce the surface band bending to obtain more efficient carrier extraction because a proper dopant concentration can assist to achieve better band alignment.^16–18 For example, by controlling the Mg-dopant concentration, the apparent quantum efficiency (AQE) of p-GaN/InGaN NWs was enhanced to approximately 12.3%.³ On the other hand, Si has been widely used for n-type doping in InGaN-based materials. It has been found that the Si dopants can reduce the coalescence between nanowires and promote a more uniform morphology.¹⁹ In addition, due to the 3D geometry of NWs, the strain can be relaxed through the sidewalls, enhancing the Si solubility limit and achieving a higher dopant concentration for the NWs than that of the planar structures.^20,21

Although the impact of the Si doping concentration on the PEC performance of InGaN-based NWs has been studied, it has been only reported for planar structures,^14,18 which can be attributed to the lack of measuring techniques compatible with the NW geometry. For example, the dopant concentration of III-V materials is commonly measured using Hall-effect,²² current-voltage,^16,23 and field-effect²⁴ measurements. However, in the case of NW-based structures, these measurements are usually performed on single NWs, which make it difficult to obtain the average doping concentration in the NW ensemble. Furthermore, complicated electron-beam lithography steps are necessary, which require high accuracy levels and can significantly increase the fabrication cost. Compared with these techniques, PEC measurements based on electrochemical impedance analysis provide a new way to accurately estimate average dopant concentrations.^25,26 Owing to the full contact with the electrolyte, this method can be applied to various nanostructured semiconductors, including NWs.

Here, by employing non-destructive PEC techniques, such as open-circuit potential (OCP) and Mott-Schottky measurements, we successfully investigate the effect of Si doping on the PEC performance of InGaN-based NW photoanodes. For the sake of accuracy, all measurements were repeated three times and the average values were taken with a proper statistical error. InGaN-based NWs were grown on Si substrates by plasma-assisted molecular-beam epitaxy (PA-MBE). To further study how the Si dopant influences the PEC performance of InGaN-based NWs, the latter was doped with Si at different effusion cell temperatures during PA-MBE growth. Scanning electron microscopy (SEM) was used to study the morphological changes of the InGaN-based NWs upon introducing the Si dopant. The optical emission as a function of Si incorporation was also investigated using room temperature photoluminescence (PL) and Raman spectroscopy measurements. We found a close relationship between the optimization of the Si dopant concentration in n-type InGaN-based NW photoanodes and their PEC performance, which will be discussed in detail in PEC characteristics.

InGaN-based NWs were grown on 2 in. Sb-doped n-type Si (100) substrates in a PA-MBE (VEECO GEN 930) chamber. The Si substrates were cleaned with methanol, isopropanol (IPA), and deionized (DI) water. The surface native oxides were removed by dipping the Si substrates into 20% hydrofluoric acid for 1 min. The Si substrates were then mounted on a 2-in. VEECO In-free block and introduced to the PA-MBE load-lock chamber. On the trolley adjoining the load-lock heater, the substrates were pre-outgassed at the temperature of the heater (200 °C) for 1 h to remove water from the surface. Then, the substrates were moved into the buffer chamber, after which the temperature was linearly increased to 650 °C to remove any residual organic contaminants. Finally, the substrates were transferred into the growth chamber, and the substrate temperature was ramped up to 800 °C for surface deoxidation. For the Si-doped n-type GaN and InGaN, a Ga beam equivalent pressure (BEP) of 5 × 10⁻⁸Torr was introduced on the sample surface at the substrate temperature of 680 °C. Then, the Si and N₂ plasma cell shutters were opened immediately. After 30 min growth of Si-doped GaN seeds, the temperature was decreased to 520 °C for InGaN NW growth, and then, In (BEP of 1.5 × 10⁻⁸Torr) and Ga (BEP of 3 × 10⁻⁸Torr) fluxes ware applied for 2 h. During PA-MBE growth, the plasma power of 13.56 MHz RF was kept at 350 W, with a N₂ flow rate of 1.0 sccm. The Si cell temperature (T_Si) was varied from 1100 to 1180 °C. An undoped reference sample was grown at the same growth temperature of 680 °C for GaN and 520 °C for InGaN but without the Si dopant.

The morphology of the PA-MBE-grown InGaN-based NWs was characterized using an FEI Nova Nano 630 field-emission SEM at an acceleration voltage of 4 kV. PL and Raman spectra were obtained at room temperature using Horiba Jobin Yvon LabRAM ARAMIS microphotoluminescence and micro-Raman spectroscopy systems. The samples were excited with 473 nm and 325 nm lasers for PL and Raman measurements, respectively.

PEC characterization studies were conducted in a three-electrode cell, where the InGaN-based NWs acted as the working electrode, a Pt coil acted as the counter electrode, and an Ag/AgCl electrode acted as the reference electrode [E_ref = 0.197 V vs normal hydrogen electrode (NHE)], in 0.1 M potassium phosphate buffer solution (pH 7). All PEC measurements were conducted using an Ametek PARSTAT4000+ potentiostat system. Mott-Schottky measurements were performed in dark conditions, while a solar simulator (Asahi Spectra, HAL-320) was used for OCP and linear scan voltammetry (LSV) measurements (scanning rate at 0.1 V/s). Gas evolution was performed in a leak-tight glass reactor with a quartz window connected to a closed gas circulation system which was purged with high purity nitrogen before illumination. The gases evolved during the water splitting experiment were collected and quantified using two separate SRI 310c gas chromatographs (GCs). In the case of hydrogen gas, a HayeSep Q column and nitrogen carrier gas were used, while a molecular sieve (5A) column, with helium carrier gas, was used for the oxygen gas measurements.

Figures 1(a)–1(e) show the morphology of all InGaN-based NW samples grown under the same growth conditions but different T_Si. All InGaN-based NWs were grown vertically with an almost uniform cross-sectional area. The NW density of the Si-doped samples (∼330 μm⁻²) is slightly higher than that of the undoped sample (∼320 μm⁻²). While the length of the InGaN-based NWs was sensitive to increasing T_Si, their diameter remained unchanged at approximately 20 nm. The length of the vertically aligned InGaN-based NWs increased from 220 nm to 340 nm with increasing T_Si, as summarized in Fig. 1(f). At high Si dopant flux (T_Si≥1150 °C), adjacent NWs slightly coalesced due to the improved lateral growth.¹⁶ This coalescence may induce crystal defects that act as charge-carrier trapping centers and reduce the carrier concentration. In general, the morphological changes of the NWs upon the addition of dopant are typically attributed to the modification of the growth kinetics, specifically the adatom diffusion length.^20,27 As shown in Fig. 1(f), the addition of an increased amount of the Si dopant increased the vertical growth rate, which is beneficial for device applications.

The room-temperature PL spectra are shown in Fig. 2(a). For the undoped InGaN-based NWs, a single broad PL emission peak centered at approximately 620 nm was observed. After introducing the Si dopant at low T_Si (1100 °C), the PL emission was slightly blue shifted, and the full width at half maximum (FWHM) was smaller than that of the undoped sample, demonstrating a better crystal quality. This blueshift can be attributed to the band-filling effect and indicates an enhanced free-carrier concentration upon Si doping.²⁸ With increasing T_Si to 1120 °C, another PL emission shoulder begins to appear at shorter wavelengths, which becomes dominant at 1150 °C. However, at the highest Si flux, the PL emission was dominated by a redshifted peak at approximately 630 nm. Similar emission profiles were previously observed for Si-doped III-nitride materials.^29,30 The PL peak shifts were attributed to the inhomogeneous Si doping level along the radius and potentially along the nanowires due to varying indium compositions. The PL emissions indicate that Si doping affects not only the morphology of the InGaN-based NWs but also the In incorporation homogeneity. According to the PL spectra, the average In composition of the InGaN-based NWs was estimated to be approximately 38% (±2%), as measured by Vegard's law^31–35 

where the bandgap of InN and GaN is 0.7 eV and 3.4 eV, respectively, and b is the bowing parameter, which is 1.5 eV.

To elucidate the incorporation of the Si dopant in InGaN-based NWs, the optical properties were further investigated using Raman spectroscopy. Figure 2(b) depicts the Raman spectra of all InGaN-based NW samples, where a bare Si substrate was used as the reference. Typically, the InGaN-based host lattice shows two dominant peaks. The peak located between 729 cm⁻¹ and 734 cm⁻¹ represents the E₁(LO) phonon mode, while the other peak located between 550 cm⁻¹ and 567 cm⁻¹ represents the E₂(high) phonon mode. The downshifting of the E₁(LO) and E₂(high) phonon lines with increasing T_Si has been reported previously.^20,36 The E₂(high) line is related to the strain in the InGaN-based NWs. Upon incorporating Si dopants in the InGaN lattice, the compressive stresses in the radial direction of the NWs are relaxed, and tensile stresses in the growth direction are induced.^36–38 This leads to the slight downshifting of E₂(high). As T_Si increases, the E₂(high) peak broadens due to Si-induced lattice imperfections,^37,39 which may also led to the observed broadening of the PL spectra of heavily doped InGaN NWs as shown in Fig. 2(a). Apart from the signal from the Si substrate (∼521 cm⁻¹), the E₁(LO) phonon line, rather than the A₁(LO) line, is dominant due to the coupling of light into and out of the InGaN-based NWs, mainly through the sidewalls.²⁰ However, for InGaN-based NWs doped with high Si fluxes (high T_Si), the coalescence between adjacent NWs may cause more light coupling through the top facet of the NWs. Thus, the E₁(LO) phonon lines shift towards lower wavenumbers. Both the E₁(LO) and E₂(high) phonon line shifts indicate the successful incorporation of the Si dopant in the InGaN-based NWs.

OCP and Mott-Schottky measurements were performed to further analyze the influence of Si doping on the free-carrier concentration and hence the band bending of InGaN-based NWs. The OCP was measured under dark and light conditions with relatively long dark periods (200 s) to reach electrochemical equilibrium. The conductivity type of the semiconductor photoelectrodes can be identified from the OCP shift under illumination.⁴⁰ Under dark conditions, free charge carriers transport from the n-type semiconductors to the electrolyte to establish electrochemical equilibrium and thus cause an upward band bending at the semiconductor/electrolyte interface. With high enough illumination intensity, the photogenerated charge carriers shift the electron quasi-Fermi level upward, and hence, the OCP shifts to a more negative potential (downward).⁴¹ As shown in Fig. 3(a), upon illumination, the OCP of all samples was shifted downward, implying the successful incorporation of Si dopants. The difference between the OCP in the dark and under illumination (ΔOCP) reflects the photogenerated charge carrier density—in another word, it identifies the concentration level of ionized carriers, which are mainly generated from dopants. As demonstrated in Fig. 3(b), ΔOCP increased with T_Si up to 1120 °C and then decreased when T_Si reached 1150 °C. At higher Si fluxes (T_Si > 1150 °C), Si-related defects, which can behave as carrier recombination centers, would be induced during PA-MBE growth, causing significant carrier loss and hence decreasing the associated ΔOCP.^41,42 These defects could be originated from Si-induced lattice imperfections with increasing T_Si (Si dopant concentration), which were commonly evidenced by the broadening of the E₂(high) peak as shown in Fig. 2(b).

After verifying the conductivity type of the InGaN-based NWs, the ionized carrier concentrations were quantified by Mott-Schottky measurements. A Mott-Schottky plot is a kind of electrochemical impedance spectrum that can be used to analyze the relationship between the inverse square of the capacitance and the applied potential by applying the Mott-Schottky equation. In our experiments, the capacitance of the InGaN-based NWs was determined from Nyquist plots, which is a frequency-dependent impedance scheme. A representative Nyquist plot is shown in Fig. 4(a) for the sample doped with Si at 1120 °C. In the frequency range of 50 Hz to 50 kHz, the Nyquist plots of all samples were characterized by two semicircles corresponding to the two resistances of the SiN_x layer formed at the interface between NWs and the Si-substrate during PA-MBE growth and the InGaN-based NW/electrolyte interface, respectively.⁴⁰ Thus, the Nyquist plots were fitted using an R(RC)(RC) equivalent circuit consisting of two Randles circuits connected in series with a resistance.^6,43,44 The first resistance (R₁) is associated with the interfacial SiN_x layer, and the second resistance (R₂) represents the interfacial transfer resistance at the InGaN-based NW/electrolyte interface. The series resistance (R_s) represents the ohmic losses in the system. Furthermore, the variation of the Si dopant distribution along the radial direction of the NWs may also induce additional surface resistance.²⁰ The Nyquist plots were obtained in a sufficient range of applied potential from −1.0 to +1.0 V vs Ag/AgCl in the dark. The corresponding Mott-Schottky plots are presented in Fig. 4(b).

Typically, the Mott-Schottky equation is deduced from the one-dimensional Poisson's equation at the interface between the semiconductor electrode and the electrolyte, which can be expressed as

where ε, A, e, and N are the semiconductor dielectric constant, the surface area, the electronic charge, and the ionized carrier concentration of the semiconductor electrode, respectively. E_applied is the applied potential with respect to the reference electrode (Ag/AgCl), and E_fb is the flat-band potential. Because this equation was developed for planar semiconductor electrodes, it has some limitations when describing the behavior of NW-based electrodes due to their 3D geometry. A modified Mott-Schottky equation was introduced by solving Poisson's equation in three dimensions, considering the geometry of the NW-based electrodes.^28,45 This modified equation considers the cylindrical symmetry of the depletion region from the surface to the center of the NWs. At a distance x from the NW surface with a radius R, the potential difference in the space charge region V_sc can be given by

The total charge in a single cylindrical NW of length L can then be calculated as $Q=eNLπ(R2−x2)$⁠, so that the expression for the total capacitance of the NW ensemble is

where D_NW is the density of NWs and ε_r = 12 is the relative dielectric constant of InGaN.⁴⁶ From these equations, V_sc and C_total can be calculated for specific positions with the same distance (radius) to the center of the NWs. Thus, the ionized dopant concentration (N) is the only parameter that is fit by experimental data in the Mott-Schottky equation for NWs. For simplicity, we assume that the Si-dopant distribution inside the NWs is homogeneous. The linear fitting of the Mott-Schottky results of a representative sample doped at T_Si = 1120 °C is shown in Fig. 4(c). The fitting curves showed good agreement with experimental results and gave a proper range for the ionized dopant concentration. By combining the modified equations and the fitting results, the ionized dopant concentrations were estimated and are summarized in Fig. 5(a). The ionized dopant concentration gradually increased from 1.8 × 10¹⁷ cm⁻³ to 2.1 × 10¹⁸ cm⁻³ with increasing T_Si and then decreased when T_Si was larger than 1120 °C. This behavior is consistent with the morphological characteristics and the ΔOCP profile shown in Fig. 3(b). When T_Si exceeded 1150 °C, the NWs exhibited slight coalescence, which may lead to the formation of charge carrier trapping centers,^47,48 thus reducing the ionized dopant concentration. However, compared to the undoped NWs, the overall ionized dopant concentration was enhanced upon Si doping.

According to the estimated values of E_fb and N from the Mott-Schottky measurements and the bandgap energies calculated from the PL spectra, the conduction band (E_cb) and valence band (E_vb) edge potentials of the InGaN-based NWs can be calculated as

where K_B is the Boltzmann constant (1.38 × 10²³ J/K) and T is the absolute temperature (295 K). The density of states in the conduction band N_c is given by $Nc=22me*KBTh23/2$⁠, where h is Planck's constant (6.626 × 10⁻³⁴J/s) and m_e* is the effective mass calculated as $me*=xmeInN+1−xmeGaN$⁠. The effective masses of InN and GaN are 0.10m_o and 0.151m_o, respectively, where m_o is the electron mass (9.11 × 10⁻³¹kg); thus, N_c was calculated to be 1.15 × 10¹⁸ cm⁻³.^34,49,50 Specific parameters of the band structure are listed in Table I, and the corresponding band diagram is depicted in Fig. 5(b). The conduction band minimum of all samples is more negative than the reduction potential (0 V vs NHE), and the valence band maximum is more positive than the water oxidation level (1.23 V vs NHE). The estimated values of E_cb and E_vb are consistent with the previously reported values for InGaN-based electrodes with similar In compositions.^1,2 According to the values of E_cb and E_vb, the bandgap of all samples straddles the redox potentials of water splitting. Owing to their higher carrier concentrations, the Si-doped InGaN-based NWs are expected to exhibit better PEC performance than the undoped InGaN-based NWs.

Both the charge-carrier concentration and the positions of E_cb and E_vb with respect to the water splitting redox potentials are essential for understanding the PEC performance. The relationship between the current density and the charge-carrier concentration was investigated by LSV, which is shown in Fig. 6(a). The enhanced carrier concentration upon optimizing the Si dopant level had an obvious effect on the performance of the InGaN-based NW photoanodes. The dark-current densities were negligible for all samples [not shown in Fig. 6(a) for simplicity]. All the Si-doped samples exhibited higher current densities than the undoped one. At the optimized Si doping level (T_Si = 1120 °C) and the highest charge-carrier concentration (2.1 × 10¹⁸ cm⁻³), the current density attained its maximum (1.42 mA/cm² at 1.597 V vs NHE), which was almost 9 times higher than that of the undoped sample. Lower current densities were obtained for the less doped and the heavily doped samples, which can be attributed to the small charge carrier concentration and the generation of carrier trapping centers (crystal defects), respectively.

Applying an external bias during the PEC measurements helps the photogenerated charge carriers to migrate from the working electrode to the counter electrode and reduces the ohmic loss. However, only when the applied bias is smaller than the thermodynamic water splitting potential (1.23 V vs NHE), the whole process can be considered as a PEC process instead of an electrolysis process. According to the LSV results, the applied bias photon-to-current conversion efficiency (ABPE) was calculated and is depicted in Fig. 6(b). The ABPE is a diagnostic efficiency to represent the development of the photoanode performance with respect to the applied bias. ABPE is given by the following equation, where J is the current density, V_app is the applied bias, and P_light is the power density of the incident light:⁶ 

With the optimization of the Si doping, the ABPE of InGaN NWs was 8 times higher than that of the undoped sample. The ABPE of Si-doped InGaN NW photoanodes can be further improved by the decoration with metal particle co-catalysts^6,51,52 and surface passivation⁶ due to the enhanced charge transfer kinetics at the surface.

Furthermore, we have tested the device stability by the chronoamperometry (CA) technique for the samples with different Si dopant concentrations and measured the hydrogen and oxygen gas evolution accordingly. The CA test was performed in 0.1 M potassium phosphate buffer solution (pH 7) containing methanol as a hole-scavenger to attract the accumulated holes on the surface of the photoanode and reduce the associated corrosion. To further improve the charge carrier separation, the InGaN NWs were decorated with Ni metal co-catalysts fabricated by the simple dip-coating process.⁶ As compared to the other samples, the one grown at the optimized Si-doping level (T_Si = 1120 °C) was showing a reasonable current stability with time. The results of this sample are then shown in Fig. 6(c) along with the gas evolution measurements. The undoped and the less- or heavily doped samples showed a faster decay of the current density with time, which was commonly seen for the bare InGaN photoanodes.^51,53 As shown in the inset of Fig. 6(c), the hydrogen and oxygen gases were steadily evolved during the testing period with a good stoichiometry. However, the stability of the Si-doped InGaN NWs and accordingly the amount of the produced gases can be further improved by the conformal coating of visible light-transparent metal oxide thin layers such as TiO₂ or NiO_x.^51,53

In conclusion, we introduced a precise technique for quantifying the Si doping level in NW-based photoelectrodes. The impact of Si doping on the morphological, optical, and PEC properties of the InGaN-based NWs was investigated. Increasing the Si doping level enhanced the vertical growth rate, which was reflected in the increased length of the InGaN-based NWs, accompanied with a slight coalescence for the heavily doped NWs. The successful incorporation of the Si dopants was confirmed by the shift of the E₁(LO) and E₂(high) phonon lines towards a lower wavenumber and the downward shifting of the OCP. The net ionized dopant density measured by Mott-Schottky plots was very sensitive to increases in T_Si, and a maximum dopant concentration of 2.1 × 10¹⁸ cm⁻³ was achieved at T_Si = 1120 °C. At this optimized doping level, E_cb and E_vb straddled the water splitting redox potentials, achieving almost 9 times higher current density than that produced by the undoped InGaN-based NWs. Better stability and stoichiometric evolution of hydrogen and oxygen gases were also demonstrated after the optimization of the Si-doping level. Our findings systematically demonstrate that optimization of the Si doping level in InGaN-based NW photoanodes is indispensable for enhancing their PEC performance.

We acknowledge the financial support from the King Abdulaziz City for Science and Technology (KACST) under Grant No. KACST TIC R2-FP-008. This work was partially supported by the King Abdullah University of Science and Technology (KAUST) baseline funding, BAS/1/1614-01-01, and MBE equipment funding C/M-20000-12-001-77.