Improved solar hydrogen production by engineered doping of InGaN / GaN axial heterojunctions

InGaN-based nanowires (NWs) have been investigated as efficient photoelectrochemical (PEC) water splitting devices. In this work, the InGaN/GaN NWs were grown by molecular beam epitaxy (MBE) having InGaN segments on top of GaN seeds. Three axial heterojunction structures were constructed with different doping types and levels, namely n-InGaN/n-GaN NWs, undoped (u)-InGaN/p-GaN NWs, and p-InGaN/p-GaN NWs. With the carrier concentrations estimated by Mott–Schottky measurements, a PC1D simulation further confirmed the band structures of the three heterojunctions. The u-InGaN/pGaN and p-InGaN/p-GaN NWs exhibited optimized stability in pH 0 electrolytes for over 10 h with a photocurrent density of about –4.0 and –9.4 mA/cm, respectively. However, the hydrogen and oxygen evolution rates of the Pt-treated u-InGaN/p-GaN NWs exhibited a less favorable stoichiometric ratio. On the other hand, the Pt-decorated p-InGaN/p-GaN NWs showed the best PEC performance, generating approximately 1000 μmol/cm hydrogen and 550 μmol/cm oxygen in 10 h. The band-engineered p-InGaN/p-GaN axial NWsheterojunction demonstrated a great potential for highly efficient and durable photocathodes. © 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement


Introduction
Photoelectrochemical (PEC) water splitting is a promising process for hydrogen generation, which utilizes semiconductor materials to realize energy conversion from solar energy to hydrogen chemical energy.Among various semiconductor materials, III-nitrides have shown a great potential for higher performance PEC applications.In particular, InGaN has attracted significant attention because of its tunable bandgap that can cover the entire visible part of the solar spectrum [1][2][3][4].Furthermore, by carefully controlling the In:Ga composition ratio of the InGaN material with its band edges capable of straddling the oxygen and hydrogen redox reactions [5][6][7][8], InGaN-based photoelectrodes can fulfill the thermodynamic criteria for water splitting (1.23 eV).Another unique feature of InGaN is its feasibility to grow in the form of nanowires (NWs), free from structure-related defects, on a Si substrate [9,10]; thus, it can be cost-effective and scalable for commercialization.The three-dimensional (3D) geometry of NWs would also result in a larger surface-to-volume ratio compared to planar structures and bulk electrode counterparts.Consequently, this helps to enhance the photogenerated charge carrier separation and collection processes [6,11], and increase the available sites for electrochemical reactions [12,13].Despite the progress, there is still an urgent need to understand the basic properties concerning their photoelectrochemical (PEC) performance, such as the intentional design of doping and bandgap engineered heterojunction for facilitating extraction of the photogenerated carriers.
Because of different surface charge properties, n-type and p-type semiconductors might exhibit different stabilities in aqueous electrolytes [14].Under illumination, electron-hole pairs would be generated and separated with help from surface band bending at the semiconductor/electrolyte interface.For n-type semiconductors, there is an upward band bending that helps holes accumulate at the n-type semiconductor/electrolyte interface and oxidize the water into oxygen.The photogenerated holes can behave as oxidizing agents, which can lead to the oxidation of the semiconductor photoelectrode surface, leading to associated photocorrosion.In contrast, the downward band bending in p-type semiconductors results in the accumulation of electrons under illumination, which serves as cathodic protection against photocorrosion [15].Thus, p-type semiconductors are more promising for high-stability PEC devices [16].Magnesium is a typical p-type dopant for III-nitride materials [16][17][18][19].It has advantages including low formation energy and efficient dopant incorporation into NWs [19][20][21], and Mg-doped InGaN-based NWs have already been applied to PEC devices [2,17,22].The previous study has been done on investigating the influence of doping concentration on tuning the surface Fermi level.However, the band alignment on the axial direction remains to be further studied since it is relevant to the carrier extraction for a PEC device [23,24].
The InGaN NWs that were grown by molecular beam epitaxy (MBE), however, requires the formation of GaN seeds prior to the growth of InGaN, which lead to the formation of axial GaN/InGaN heterojunctions.In addition to the surface charge properties, the interfacial charge kinetics at the GaN/InGaN heterojunction interface is crucial for improving the charge extraction/collection efficiency [25].In this study, we investigate the influence of the variation of the doping profile of GaN and InGaN on the PEC performance, as well as the stability of the water splitting process.By varying the polarity of GaN and InGaN by Si-and Mg-doping, we aim to control the interface charge kinetics at the axial heterojunctions that can improve the overall PEC performance.The morphological and optical properties of NWs with different GaN/InGaN axial heterojunction designs are also studied.Mott-Schottky measurements were used to quantify the ionized carrier concentrations carefully, the stability of various photoelectrodes was evaluated, and finally, the O 2 and H 2 gas evolution were quantified.These characterizations are essential building blocks for complete performance evaluation of the doping-engineered heterojunction based photoelectrodes for PEC water splitting.

Experimental details
Figure 1(a) shows a schematic of the general structure of the InGaN-based NW photoelectrodes grown by a Veeco Gen930 plasma-assisted MBE system.The InGaN NW is typically grown on GaN seeds on bare silicon substrates and decorated with a suitable metal co-catalyst.Three different structures of different Si-doped (n-type) and Mg-doped (p-type) GaN seeds and InGaN NWs, and hence different heterojunction designs, were proposed in this study.A photoanode composed of an n-type InGaN grown on n-GaN decorated with Ir metal co-catalyst was compared to another two photocathodes with unintentionally doped (utype) u-InGaN grown on p-GaN and p-InGaN grown on p-GaN structures.The u-InGaN/p-GaN and p-InGaN/p-GaN NWs were decorated with Pt metal co-catalysts to improve the surface charge extraction.Prior to the MBE growth, 20% HF soak, H 2 O rinse, and final hightemperature treatment of 850 °C in the growth chamber were applied to the silicon substrates to remove the native oxide layer.During growth, the nitrogen plasma source was operated at 350 W with a chamber pressure of 1.5 × 10 −5 Torr.In the case of p-InGaN/p-GaN NWs, 50 nm of p-type GaN seed was grown on a p-Si substrate at 660 °C with a Ga beam equivalent pressure (BEP) of 5 × 10 −8 Torr, and 150 nm of p-type InGaN NW was grown at 480 °C with a Ga and In BEP of 3 × 10 −8 and 1.5 × 10 −8 Torr, respectively.The p-type doping of GaN was realized by Mg doping at a cell temperature of 300 and 200 °C for GaN seeds and InGaN NWs, respectively.The u-InGaN/p-GaN sample was grown under the same conditions without Mg doping into InGaN.The reference n-type sample (n-InGaN/n-GaN) was grown on n-type Si (100) substrates using previously reported conditions [13] with a Si cell temperature of 1100 °C for n-type doping.
The morphology of the InGaN/GaN NWs was determined by an FEI Nova NanoSEM 630 field-emission scanning electron microscope (SEM) at an accelerating voltage at 4 kV.The optical properties of the InGaN/GaN NWs were examined by photoluminescence (PL) spectroscopy (LabRAM ARAMIS, Horiba Jobin Yvon) at room temperature using a 325-nm He-Cd laser as an excitation source.
As for the PEC device characterizations, electrochemical impedance spectroscopy (EIS) techniques were performed in the dark to estimate the ionized carrier concentrations inside the NWs.Accordingly, the band structures of the three heterojunctions were simulated with PC1D [26].It was assumed that the background doping concentrations of the n-InGaN/n-GaN NWs and p-InGaN/p-GaN NWs were homogeneous and equal to the values estimated by Mott-Schottky analysis.The unintentionally doped InGaN segments of the u-InGaN/p-GaN NWs were assumed to have an n-type background doping level of 5 × 10 16 cm -3 , whereas the doping level of the p-type GaN seeds was in accordance with the Mott-Schottky results.Other physical parameters in the PC1D model were either extracted from the literature [27][28][29][30][31][32][33] or specified from the SEM and PL results.
The corresponding PEC performance was studied by a systematic PEC characterization, including open-circuit potential (OCP) measurements, linear scan voltammetry (LSV), gas evolution measurements, and stability tests.All the PEC studies were performed in a threeelectrode cell using InGaN/GaN-NWs as the working electrode (WE), a Pt coil as the counter electrode (CE), and an Ag/AgCl as the reference electrode (RE).The reference potential is + 0.197 V vs RHE in pH 0 solution and + 0.61 V vs RHE in pH 7 solution.The u-InGaN/p-GaN and p-InGaN/p-GaN samples were tested in 1 M sulfuric acid (H 2 SO 4 ) solution (pH 0), whereas a 0.1 M potassium phosphate buffer solution (pH ~7) was used in the case of the n-InGaN/n-GaN sample.The three-electrode cell was connected to an Ametek PARSTAT4000 + potentiostat system.EIS characterizations (Nyquist and Mott-Schottky measurements) were conducted in dark conditions, whereas for the other PEC measurements, a solar simulator (HAL-320, Asahi Spectra) provided the simulated solar illumination, which was fixed at 600 mW/cm 2 during the whole study.For gas evolution measurements, a leak-tight glass reactor with a quartz window was utilized, which was connected to a closed high-purity nitrogen gas circulation system and purged before illumination.The evolved gases were collected and quantified using two separate SRI 310C gas chromatographs.In the case of hydrogen gas, a HayeSep Q column and nitrogen carrier gas were used, whereas a molecular sieve (5Å) column and helium carrier gas were used for the oxygen gas measurements.

Results and discussion
Figures 1(b)-1(d) show the morphology of the different InGaN/GaN NW photoelectrodes.The average height of the InGaN/GaN NWs is approximately 210-260 nm.The height of n-InGaN/n-GaN NWs is slightly longer than that of the u-InGaN/p-GaN and the p-InGaN/p-GaN NWs since the Si dopant induce the vertical growth [19,34].With introducing the Mg dopants, the InGaN NWs exhibit an increased lateral growth and slightly lower density over the substrate surface area.The average diameter of the p-InGaN/p-GaN NWs was ~45 nm, while that of the u-InGaN/p-GaN NWs is ~30 nm.This diameter increase has been commonly observed for Mg-doped III-nitride NWs [19,34,35].The optical properties of the different InGaN/GaN NWs were evaluated using room-temperature PL, as shown in Fig. 1(e).The PL spectra of the three samples are dominated by broad peaks because of near-band-edge emission and the transmission between the conduction band and the shallow doping levels [33,34].The PL peak emission energy is centered at about 2.15 eV for the u-InGaN/p-GaN NWs, while those of the pure Si-doped and Mg-doped InGaN/GaN NWs shift to lower energy values.The u-InGaN/p-GaN sample was supposed to have the lowest carrier concentration among the three samples as only the GaN segment in it was intentionally doped with Mg.The incorporation the band-edg bandgap of bandgaps are while overcom ) would introd hotons to be lo sion peaks [17 /GaN NWs to d ments [11,22].The ionized carrier concentrations of the different InGaN/GaN NWs were quantified using Mott-Schottky plots, which were depicted by extracting the fitted capacitance of InGaN NWs from the Nyquist plots, as shown in Fig. 2. Details of EIS measurements can be found in our previous articles [13,20].The positive slope of the curve for n-InGaN/GaN NWs and the negative slope of the curve for the u-InGaN/p-GaN and p-InGaN/p-GaN NWs were due to successful Si and Mg doping, correspondingly.A modified Mott-Schottky equation has been demonstrated in the literature [13,17,36], which considers the 3D geometry of NWs and hence better estimates the ionized carrier concentration within NW structures.The estimated ionized carrier concentrations of the u-InGaN/p-GaN and p-InGaN/p-GaN NWs were in the range of 1 × 10 16 and 7.6 × 10 15 cm -3 , respectively, while that of n-InGaN/n-GaN NWs exhibited the highest concentration of 5.7 × 10 17 cm -3 .As III-nitride NWs are naturally ntype, p-type dopants must usually go through a compensation step before the NWs exhibit ptype characteristics [17,18,20].The Mg doping may induce more structural defects, which can behave as recombination centers and decrease the ionized dopant concentration [17,20].
The OCP measurements were further used to identify the conduction type of NWs in Fig. 3(a) [10].The samples were tested under dark and light illumination conditions with an illumination period of 30 s.The rapid photoresponse indicates the excellent material quality of the NWs.Upon illumination, the photogenerated electron-hole pairs would induce a shift in the Fermi level of the majority carriers.The OCP shifts toward more anodic potentials for a p-type semiconductor and shifts oppositely for an n-type semiconductor.The more significant the OCP shift, the higher the potential change within the space charge layer caused by Fermi level pinning at the NW surface [10,17].The OCP results are consistent with the Mott-Schotty estimation.It is clear that both u-InGaN/p-GaN and p-InGaN/p-GaN NWs possess ptype conduction characteristics.The higher OCP difference (ΔOCP) of the p-InGaN/p-GaN NWs demonstrates that the net generation rates of electron-hole pairs in the Mg-doped NWs are higher than those of u-InGaN/p-GaN NWs.In contrast, the ΔOCP of the n-InGaN/n-GaN NWs is larger than those of the other two samples, indicating higher Si dopant concentrations than the Mg dopant concentrations.
The PEC performance of the InGaN/GaN NW photoelectrodes with different heterojunction designs was investigated using LSV and the applied bias photon-to-current conversion efficiency (ABPE).All samples exhibited a negligible dark current (Fig. 3(b)).Compared to the u-InGaN/p-GaN NWs, the p-InGaN/p-GaN NWs show a higher photocurrent density (~-17 mA/cm 2 at an applied bias of -1.0 V vs RHE, reversible hydrogen electrode).These results are reasonable despite the higher estimated ionized dopant concentration of u-InGaN/p-GaN NWs, because the Mott-Schottky measurements assumed that the dopants distributed homogeneously in the NWs.However, the actual axial inhomogeneity of the carrier distribution may affect real functional PEC devices under illumination.The NWs would also suffer carrier losses because of the electron-hole recombination at the p-n junction in the u-InGaN/p-GaN NWs [36].Although the ionized dopant concentration of the n-InGaN/n-GaN NWs was the highest among the three samples, the photocurrent density (~0.15 mA/cm 2 at 0.5 V vs RHE) was very small, which can be partially attributed to the poor carrier collection/extraction efficiency [23] as well as the ntype semiconductor surface instability [15,16].In Fig. 3(c), the ABPE conversion efficiency was calculated accordingly using Eq. ( 1) by assuming the Faradaic efficiency (ƞ) as 1.The highest ABPE was recorded for the p-InGaN/p-GaN NWs at 0.65% at an applied bias of -0.83 V vs RHE, which is about 2.5 times higher than that of the u-InGaN/p-GaN NWs and more than 50 times higher than that of the n-InGaN/n-GaN sample.The results are even higher than the typically reported ABPE (<0.5%) for similar GaN/InGaN nanostructures [1,[37][38][39][40][41][42][43]   The reliability of the photoelectrode is crucial for PEC water splitting, which can also be reflected in the gas production quality.The stability of the InGaN/GaN NW photoelectrodes with different heterojunction designs was measured by conducting chronoamperometry (CA) tests.The Pt-decorated p-InGaN/p-GaN and u-InGaN/p-GaN NWs demonstrated superior chemical stability in 1 M H 2 SO 4 under high-intensity illumination (6 suns) for over 10 h without noticeable degradation (Fig. 4(a)).A lower potential (-0.453 V vs RHE for p-type and + 0.81 V vs RHE for n-type) was applied to overcome the ohmic losses on the external circuit during the CA and gas evolution measurements [44].The n-type sample degraded rapidly within 2 h because of the strong self-oxidation [14].The amount of the gas evolved during the CA tests was continuously recorded and analyzed using a gas chromatography system.The leak-tight reactor used during these measurements is presented in Fig. 4(b).The amount of the gas produced by the n-InGaN/n-GaN photoanode was too small to be measured, which can be associated with its poor stability.In contrast, the p-type samples show superior chemical stability and evolution rates in the pH 0 electrolytes.The current density of the two photocathodes increase for the first two hours.The plausible explanation is the interfacial changes at the InGaN surface [4,45].Then, the current density of the u-InGaN/p-GaN and p-InGaN/p-GaN saturated at approximately -4.0 and -9.4 mA/cm 2 , respectively.These photocathodes exhibit a superior longterm photoactivity, demonstrated by interrupting the light illumination after 9 h, as shown in Fig. 4(a).In spite of the increase in the dark current densities of these two photocathodes at the end of the measurements because of the unexpected corrosion reaction, the dark current densities are still negligible compared to the photocurrent densities.Although a stable current is produced by the u-InGaN/p-GaN sample and the progressive evolution of gases, the oxygen and hydrogen do not exhibit a good stoichiometry.In Fig. 4(c), the overall stoichiometric ratio is H 2 :O 2 ~0.8, which can be partially attributed to the recombination of the photogenerated carriers at the p-n junction, which can reduce the carrier extraction efficiency.Another reason for the imperfect stoichiometry reaction of th gases with a GaN photocat linearly increa and 550 µmo decorated GaN The small ily tunnel through, while the transport of holes in the same direction will be hindered by the large valence band offset, leading to the accumulation of holes on the surface of the NWs [25].The occurrence of the water oxidation reaction on the surface of the n-type NWs may lead to the formation of Ga 2 O 3 , which can be dissolved in aqueous electrolytes, leading to the degradation and instability of these photoanodes [15,46].In the case of u-InGaN on the p-GaN heterojunction, the conduction band in the u-InGaN is tilted in a direction near the Fermi level of the p-GaN, forming a large potential barrier at the conduction band offset and moving the photogenerated electrons toward the surface, which may explain the cathodic behavior of these NWs [47].A similar energy band diagram can be seen for the p-type InGaN/GaN photocathodes but with a perfect downward surface band bending.A large conduction band offset is formed drifting the photogenerated electrons to the NW/electrolyte interface, and a small valence band offset is formed, allowing the separated holes to pass through to the NWs.In this case, hydrogen reduction occurs on the NW surface, and water oxidation takes place on the Pt counter electrode.As compared to the other two photoelectrodes, this heterojunction design demonstrates long-term stability and the best gas evolution efficiency, which can be explained as follows.In the PA-MBE InGaN NWs grown under nitrogen-rich conditions, all the exposed surfaces (top polar (000-1) c-plane and non-polar (10-10) m-plane side walls) are nitrogen-terminated surfaces [14].During the PEC water splitting, these surfaces are negatively charged, which, together with the downward surface band bending, can reduce the energy barrier for injecting holes into the electrolyte and generate hydrogen from the water oxidation reaction.These N-terminated surfaces were also reported to protect the InGaN NWs against photocorrosion and oxidation, leading to the observed long-term stability and perfect hydrogen and oxygen evolution [14].

Conclusions
In summary, we demonstrated the effect of the axial doping engineering of InGaN/GaN NWs on the morphology, optical properties, and PEC performance.Ionized dopant concentrations were estimated by Mott-Schottky measurements, and the corresponding band diagrams were studied by simulations.The optimized band structures of the p-InGaN/p-GaN NWs led to enhanced PEC performance.By incorporating Pt co-catalysts, a photocurrent density of about -17 mA/cm 2 at -1.0 V vs RHE was obtained.In addition, the Pt-treated p-InGaN/p-GaN axial NWs exhibited high stability over the 10 hr test period in 1 M H 2 SO 4 electrolyte while maintaining a considerably high current density of -9.4 mA/cm 2 at -0.453 V vs RHE, and a hydrogen evolution rate of 107 ± 1.8 µmol/cm 2 /hr.
Fig. Chron pH 0 e used f (c) hy In order to with different simulated usin 5(c) was con illuminated w of the NW.T during the PE electrolyte.T considering t barrier at the direct contact band bending charge carrie equilibrium.T the water oxi NW to the co conduction ba