Passivation layers for nanostructured photoanodes: ultra-thin oxides on InGaN nanowires

An experimental strategy for systematically assessing the influence of surface passivation layers on the photocatalytic properties of nanowire photoanodes by combining photocurrent analysis, photoluminescence spectroscopy and high resolution transmission electron microscopy with a systematic variation of sample structure and the surrounding electrolyte is demonstrated. Following this approach we can separate the impact on recombination and transport processes of photogenerated carriers. We apply this strategy to analyze the influence of ultra-thin TiO2, CeO2 and Al2O3 coatings deposited by atomic layer deposition on the photoelectrochemical performance of InxGa1−xN/GaN nanowire (NW) photoelectrodes. The passivation of surface states results in an increase of the anodic photocurrent (PC) by a factor of 2.5 for the deposition of 5 nm TiO2. In contrast, the PC is reduced for CeO2- and Al2O3-coated NWs due to enhanced defect recombination in the passivation layer or increased band discontinuities. Furthermore, photoelectrochemical oxidation of the InxGa1−xN/GaN NW photoelectrode is attenuated by the TiO2 layer and completely suppressed for a layer thickness of 7 nm or more. Due to efficient charge transfer from the InxGa1−xN NW core a stable TiO2-covered photoanode with visible light excitation is realized.


Introduction
In recent years semiconductor photocatalysis and especially the possibility of photoelectrochemical water splitting on semi-conductor surfaces has drawn interest as a clean method to produce hydrogen by conversion of solar energy. 1,2 For a pho-toelectrochemical water splitting cell both cathode and anode materials need to be optimized, and at least one of the elec-trodes needs to be a photoelectrode. Typical photoanode materials for the demanding 4-hole oxidation of water are metal oxides like TiO 2 , as they are stable in aqueous solutions and efficiently enhance photo-activated electrode reactions. 1,3 However, those materials leak in their high band gap of typically more than 3 eV, meaning only the ultraviolet regime of the solar 25 spectrum can be used to drive the photocatalytic reaction.
The In x Ga 1 x N material system (InGaN) excellently matchesthe solar spectrum, as the band gap can be shied from the ultraviolet into the visible regime by adjusting the In-content xin the alloy, 4,5 thus allowing the absorption in a broad spectral 30 range of the solar spectrum.
Planar group III-nitrides have been investigated with respectto their photocatalytical properties [6][7][8][9][10][11][12] and it has been shownthat nanostructured anode materials, namely InGaN nanowires (NWs), can further enhance the solar-to-hydrogen conversion due a larger active surface area. [13][14][15][16] Molecular beam epitaxy makes it possible to grow such NWs with high aspect ratio, high crystal quality, and low defect density in precisely designed structures. 17 However, InGaN NW photoelectrodes suffer from a high loss of photo-generated carriers due to non-radiative surface recombination. A method to suppress such recombination processes is to improve the passivation of surface states by deposition of ultra-thin lms. 18,19 Moreover, such surface coatings can have several additional effects on the lifetime and efficiency of a photoanode, depending on material parameterssuch as chemical composition, material quality or coating thickness. TiO 2 has been shown to be an excellent protection layer for different types of (nanostructured) photoanodes. [20][21][22][23][24][25][26][27][28][29][30][31] In this work we show that the performance of InGaN/GaNNW photoanodes under visible light excitation is significantly enhanced by deposition of an ultra-thin TiO 2 coating. The requirements for a good passivation layer are assessed by comparison of TiO 2 , CeO 2 and Al 2 O 3 as coating materials. We systematically investigate the different contributions of loss processes to the photoelectrochemical performance of coated InGaN/GaN NW photoanodes, and the inuence of the coating thicknesses. Since photoexcitation of the InGaN at energies smaller than the band gaps of the coatings is possible and the band edge 10 positions of the coatings straddle the redox levels for water splitting, an efficient charge transfer from the InGaN core to the coating material allows the realization of a coated photoanode with visible light excitation in the InGaN core.  (T 405nm  Al2O3 =99%; T 405nm  TiO2 =98%;   T 405nm  CeO2 =97%). Hence, it can be assumed that a comparable amount of carriers is excited in the InGaN part of uncoated and coated samples. Excitation of carriers in the GaN base of the nanowires does not have to be considered.

Photoelectrochemical investigation
For photoelectrochemical investigation (photoluminescence, photocurrent and open circuit potential measurements) the sample was mounted in a cell with a continuous electrolyte flow, realized by a peristaltic pump.
As standard electrolyte a 1 M phosphate buffer solution at pH 8 with a conductivity of 100 mS/cm was used. For a hole scavenger containing electrolyte 1 M of Na 2 SO 3 was added to the solution. A pH-value of 8 and a conductivity of 100 mS/cm were adjusted for comparability.
An external bias was applied by connecting the sample as working electrode in a threeelectrode setup with an Ag/AgCl reference electrode and a platinium counter electrode.
Ohmic contacts on the backside of the conductive silicon substrates were deposited by thermal evaporation of ~50 nm Al and ~150 nm Ag. A potentiostat was used to apply a controlled bias.
For photo-excitation of the NW structures a 405 nm laser diode was fiber-coupled to the chamber through a quartz window on top. The incident excitation power at the electrolyte surface was determined to 29 mW using a power meter.
For photoluminescence (PL) measurements the light emitted by the nanowire sample is fibercoupled to a QWave USB spectrometer (rgb lasersystems). Two pass filters at 435 and 425 nm were used to suppress the laser line.
Simultaneous measurement of the photocurrent (PC) was facilitated using a Keithley 2700 multimeter connected between potentiostat and working electrode. A shutter was placed in front of the laser diode, to make a switching between dark and illuminated intervals in All measurements were performed at room temperature.

Sample properties
The investigated samples consist of InGaN/GaN NWs (cf. scanning electron microscopy image (SEM) and schematic structure in Fig. 1a), which were coated with 2-15 nm thin lms of TiO 2 , CeO 2 and Al 2 O 3 respectively using atomic layer deposition (ALD) (cf. high resolution transmission electron images (HRTEM) and schematic structure in Fig. 1b). All NW samples show complete conformal coverages with homogeneous film thickness. Upon film deposition the shape of the nanowire remains unchanged, only the diameter is increased by twice the film thickness, resulting also in a decrease in average distance between the nanowires of the ensemble. Nevertheless, the nanowires remain sufficiently separated to allow free access of the electrolyte to their lateral facets.
For photoexcitation of carriers in the InGaN part of the NWs a wavelength of 405 nm was used. In Fig. 2 transmission measurements of 5 nm thin ALD lms of the three coating materials deposited on quartz glass are shown. At 405 nm all coating materials are almost completely transparent (T405 nm,Al 2 O 3 1⁄4 99%; T405 nm,TiO 2 1⁄4 98%; T405 nm,CeO 2 1⁄4 97%). Hence, it can be assumed that a comparable amount of carriers is excited in the InGaN part of uncoated and coated samples. Excitation of carriers in the GaN base of the NWs does not haveto be considered. To analyze the stability of the NW PC, repeated PC measurements at a xed anodic bias of 700 mV for illumination intervals of 30 min separated by dark intervals of 10 min were performed (Fig. 3b). The bias of 700 mV was chosen as it resultsin a stable photocurrent without degradation of the NWs or the respective coatings. In the transient measurement a decrease in PC over time is observed which stabilizes after approximately1 h, id est after 1-2 illumination intervals, at a lower PC than the initial one (dotted red line in Fig. 3b).
The application of an anodic bias to InGaN in electrolytes under illumination is known to cause oxidation of the NWs. The formation of an (In x Ga 1 x ) 2 O 3 layer causes a large band offset atthe In x Ga 1 x N/(In x Ga 1 x ) 2 O 3 interface and a suppression of carrier transfer into the electrolyte, which leads to the perma-nent decrease of the PC 32 (red dotted line in Fig. 3b). Once the oxide film is closed no further oxidation of the InGaN occurs and a stable PC of the InGaN NW electrodes is observed. The overshoot at the beginning of each illuminated interval (Fig. 3b) is typical for immediate transport of surface near carriers into the electrolyte aer rst light excitation, the decay is caused by the recombination and trapping of following carriers.

Influence of 5 nm oxide coatings
The PC of InGaN nanowires coated with 5 nm thin films of TiO 2 , CeO 2 or Al 2 O 3 was analyzed in comparison to an uncoated reference sample (Figure 4b-d) at a fixed anodic bias of 700 mV using the same measurement sequence as above. For each of the measurements the second illuminated interval is shown, to discuss the transients in their stable condition where no further anodic oxidation occurs.
All coated samples still show significant PCs. However, the influence of the coating strongly depends on the coating material. In comparison to the reference sample an increase of photocurrent intensity by a factor of 2.5 after deposition of 5 nm TiO 2 is observed ( Figure   4a). In contrast, CeO 2 and Al 2 O 3 coverages with the same thickness lead to a PC decrease by a factor of 0.9 (Figure 4b,c), despite the fact that a passivation of surface states would be expected for all three materials. To elucidate the origin for the different influence of the three coating materials on the PC, we investigated the transport of photo-excited carriers through the coating layer.
To contribute to the photocurrent carriers excited in the InGaN part of the nanowire need to be transported into the electrolyte and can undergo different recombination processes, schematically shown in Figure 5 for uncoated and coated samples.
In the case of uncoated nanowires the photocurrent is mainly reduced by radiative recombination in the InGaN core (process 1) and by non-radiative surface recombination (process 2) [25]. For coated nanowires both can be affected by formation of the InGaN NW/oxide interface (process 1', 2'). Moreover, defect recombination within the ultrathin coating layer (process 3) and surface recombination at the oxide/electrolyte interface (process 4) result in further losses. Radiative recombination in the coating can be neglected, as no band-to-band excitation within the coating takes place and the simultaneous transfer of electrons and holes from the InGaN is suppressed due to band discontinuities.
To result in an enhanced photocurrent PC' the reduction of the recombination rate due to passivation of surface states (process 2 -> process 2') must overcompensate the additional loss processes 3 and 4. In the following, the 4 processes are investigated separately for the different coating materials.  The assumption of suppressed band transfer for Al 2 O 3 coated InGaN nanowires is supported by the fact that PL and PC are independent of the applied bias in this case. [4,5,20,22,23,26].

Figure 5 (a-c) Photoluminescence spectra of TiO 2 -, CeO 2 -and Al 2 O 3 -coated nanowires normalized to the spectra of uncoated reference pieces from the same sample. Spectra are taken at 700 mV anodic bias. (d) Band alignment of the coating materials compared to In x Ga 1-x N with an In-concentration of x=0.25 from PL measurement (solid lines) and x=0.1 from XRD measurement (dotted lines); band position values for T = 0 K
Photoluminescence intensity and band alignment are comparable for TiO 2 -and CeO 2 -coatings on InGaN nanowires. However, a significant increase of the PC was observed for the TiO 2while a decrease was found for the CeO 2 -coating, suggesting non-radiative recombination in the CeO 2 -coating (process 3 in Figure 5), most likely at structural defects. This conclusion is supported by HR-TEM analysis of the TiO 2 -and CeO 2 -layers, shown in Figure 7.
The TiO 2 layer consists of an amorphous matrix with embedded grains that have a diameter of approximately the same size as the film thickness. In contrast, the CeO 2 film is polycrystalline, showing differently orientated grains with a grain-size equal to the film thickness and large-angle grain boundaries. The Al 2 O 3 film is amorphous. As CeO 2 is known for its high defect ratio due to oxygen vacancies and Ce 3+ -defects [27], the

Ce-oxidation state was estimated by EELS measurements (see supplemental information 3),
showing a high degree of Ce 3+ -defects, especially at the surface. Both, the large angle grain boundaries and the oxygen vacancies/Ce 3+ -defects could give rise to an enhanced defect recombination rate in the CeO 2 -coating compared to the TiO 2 -film.
The surface recombination of photo-generated holes as minority carriers (process 4) with electrons in surface states was investigated for all different coating materials. For this purpose, the PC was measured in the presence of Na 2 SO 3 that was added as a "hole scavenger" (hs) to the electrolyte solution. As such, it suppresses surface recombination by fast trapping of photogenerated holes before they undergo non-radiative surface recombination. The effect of the hs on the PC was comparatively analyzed for all samples including the reference. In Figure 8 the PCs measured in the presence of the hs, normalized to a PC measured without hs on the same sample is shown. The magnitude of this ratio is a measure for the surface recombination velocity of the respective sample without hs. The presence of the hs has also impact on the bias-dependence of PC and PL. In Figure 9 this is exemplarily shown for the sample coated with a 5 nm TiO 2 film. Figure 9a demonstrates that the main influence of the hole scavenger on the photocurrent is a shift of the onset potential (minimum potential to observe a PC) to the cathodic direction. This indicates that the open circuit potential (OCP) is also shifted (measured OCPs are marked by vertical lines: -330 mV → -605 mV) and therefore a higher carrier separation e.g. higher photocurrent at a given potential is gained.
The shift of OCP can be explained by an enhanced occupation of electron surface states in the presence of the hs, since the recombination with photogenerated holes is suppressed.
In Figure 9b the influence of a hole scavenger on the bias-dependent PL intensity is shown.
In this case two effects of the hole scavenger are observed: A general offset to lower PL intensities and a shift in the direction of cathodic bias. The latter is a consequence of decreasing OCP, leading to strongly reduced PL intensities only in the cathodic regime. In the anodic regime however, only a small decrease of PL intensity is observed as the curves are already in saturation.

Influence of coating thickness
The carrier recombination in the TiO 2 -coating was further investigated by analyzing the evolution of the PC with the thickness of the coating (2, 5, 7, 10 and 15 nm). In Figure 10a the resulting dependence of the PC on the thickness is shown. At t = 0 the resulting photocurrent increases with decreasing film thickness, were a 2 and 5 nm coating lead to an increase and a 7-15 nm coating lead to a decrease of PC compared to the uncoated reference sample. This trend can be attributed to an enhancement of defect recombination in the coating for increased layer thickness (process 3).
However, the transient PC behavior strongly depends on the film thickness: For film thicknesses < 7 nm a degradation of the PC with time can be observed. The intensity loss is highest for the uncoated reference and the 2 nm TiO 2 coated sample and less pronounced for the sample with a 5 nm TiO 2 coating. Films with thicknesses ≥ 7 nm show very stable photocurrents in the contrast. We attribute the permanent degradation to the anodic oxidation of the InGaN for the thin films, while the higher film thicknesses serve as protection-layer.
The observed differences in long term stability lead to the fact, that after some measurement time even nanowires with TiO 2 -film thicknesses of ≥ 7 nm lead to increased PCs compared to the uncoated sample. Therefore, a film thickness of 5-7 nm seems to be the best compromise between long term stability and photocurrent enhancement.
Measurements in hole scavenger (Figure 10b) result in a strong increase of PC for the reference and the 2-5 nm TiO 2 coated samples. Films of 7-10 nm show almost the same PC as without hs. When the hs is added to the electrolyte the anodic oxidation is suppressed and therefore the PC is stable for all samples. Conclusion he inuence of ultra-thin TiO 2 , CeO 2 and Al 2 O 3 coatings on the photoelectrochemical performance of InGaN NWs on GaN NW templates was comparatively analyzed. All investigated coating materials were deposited by ALD. They are transparent for the 30 investigated excitation wavelength of 405 nm and their band edge positions straddle the redox levels for water splitting.It was demonstrated that the presence of a 5 nm TiO 2 coating enhances the anodic PC of uncoated InGaN/GaN NWs by a factor of 2.5, which is attributed to the passivation of surface states.
In contrast, for 5 nm coatings with CeO 2 or Al 2 O 3 the passivation of surface states could also be proved by analysis inhole scavenger electrolyte, but did not lead to an enhancementof the PC, as other recombination processes were enhanced. For Al 2 O 3 , the high band offset to InGaN suppresses the transfer of photocarriers and results in an enhancement of radiative recombination. In the case of CeO 2 additional carrier losses are introduced by increased defect recombination in the layer.
While an enhancement of the PC was observed for TiO 2 coatings with thicknesses of 2 nm and 5 nm, an increased thickness of 7 nm to 15 nm results in a decrease of the PC due to recombination in the coating.
For longer operation times the anodic PC of uncoated InGaN/GaN NWs stabilizes on a decreased level as a conse-quence of photoanodic oxidation. While this degradation effectis reduced for a coating thickness up to 5 nm it is completely suppressed for a thickness of 7 nm and more.
Consequently, ultra-thin TiO 2 coatings with a thickness of 5 nm and 7 nm provide a stable enhancement of the photo-current of InGaN/GaN NWs for photoelectrochemical application. In addition, due to efficient transfer of photocarriers, the resulting composite nanostructures consisting of a photoactive 1 InGaN core and a functional TiO 2 coating provide the possibility of triggering photochemical reactions at the TiO 2 surface with subbandgap excitation.

ALD deposition of oxide films
The NWs were coated with thin films of CeO 2 , TiO 2 and Al 2 O 3 with different layer thickness using atomic layer deposition (ALD) (cf. TEM image and schematic in Figure 1b). Uncoated reference pieces for each sample were comparatively analyzed.
For ALD of CeO 2 Ce(thd) 4 and ozone were used as precursors at a deposition temperature of 270 °C [21]. 400 ALD-cycles were used for a nominal film thickness of 5 nm (HR-TEM: 5.8 ± 0.7 nm). The deposition of CeO 2 was verified by EELS analysis. The size of the indirect band gap was estimated from a Tauc plot obtained by transmission measurements of a 5 nm CeO 2 film on quartz glass as reference and is about 2.9 eV (see supplemental information 1) (Lit: 3.2 eV [22]).
From transmission measurements on a 5 nm TiO 2 film on quartz glass as a reference an indirect band gap of 3.1 eV was determined from Tauc plot (see supplemental information 1) (Lit.: 3.2 eV [23]).