About Degradation and Regeneration Mechanisms of NiO Protective Layers Deposited by ALD on Photoanodes

The use of high pH electrolytes requires protective layers to avoid corrosion in photoanodes based on semiconductors like silicon. NiO is one of the materials that complies the requirements of transparency, conductivity, chemical stability and catalysis on its surface in contact with the electrolyte. Here, these layers have been deposited by atomic layer deposition (ALD) at low temperatures, and their stability is analyzed over 1000 hours. Due to the layer structure characteristics, best overall performance has been achieved at 100 ºC deposition temperature. By electrochemical measurements progressive time dependent degradation under anodic working conditions is observed, attributed to the formation of higher nickel oxidation states at the electrode/electrolyte interface as a main degradation mechanism, resulting in an OER overpotential increase. Another minor degradation mechanism affects the optical surface quality and gives rise to a loss of photon absorption efficiency in the hundreds of hours scale. A regeneration process based on an in situ periodic cyclic voltammetries, bringing the electrodes to cathodic conditions every 3, 12 or 48 hours, has shown to partially recover the main degradation mechanism achieving 85% of stability over 1000 hours in a study with over 10 mA·cm -2 photocurrent densities. The degradation mechanisms of ALD-grown NiO protective layers over 1000 h at anodic alkaline conditions are identified and a recovery mechanism is presented


INTRODUCTION
As our society turns in the direction of a renewable based energy system, new challenges appear 1 . To store one of the most abundant energy sources, the sun, to be used when it is not available, photoelectrochemical (PEC) water splitting offers a clean and direct path [2][3][4] . The produced hydrogen can be either stored for further conversion into electricity with fuel cell technologies 5,6 or used in chemical process such as CO2 revalorization 7 and other chemical products fabrication.
TiO2 has been the most used protective layer for photoanodes during last years, always combined with a catalyst layer or particles such as IrO2, Pt, Ni, or Ni-Fe 19,26,29,30 . In a recent work 31 , we have found the significant role played by the conduction mechanisms trough TiO2 concerning the identified instability over long periods when in contact with alkaline environments at anodic oxidative potentials. The mechanisms related to the OHdiffusion inside TiO2 were analyzed, causing degradation in TiO2 layers working under anodic conditions. The preferential conduction paths observed and attributed to grain boundaries and defects within the crystalline structure are altered in such conditions, changing the conductivity through the protective layer. Therefore, following these findings, there is still room for alternative materials which can assure better performance for very long-term under OER conditions.
In this study, we use NiO layers grown by atomic layer deposition (ALD) on silicon p-n junctions to efficiently fulfil all the required parameters as protective layers for PEC water splitting in alkaline environments (chemical resistance, photoabsorber protection, conductivity, transparency and catalytic effectivity for the oxygen evolution reaction (OER)) in a single step at low temperature fabrication process. NiO has a 3.8-4 eV band gap 32 , with expected low visible spectra absorption coefficient. Moreover, it is also expected to be highly stable in alkaline media, forming nickel compounds in its surface, some of them more favorable for the OER water splitting reaction 33 . Likewise, NiO is known to have p-type semiconductor behavior 34 , although some works report up to three orders of magnitude lower conductivity than titanium dioxide 35,36 .
A standard silicon based frontal photoanode has been selected as supporting electrode for this study. ALD allows obtaining very conformal and pinhole-free layers, characteristics highly favorable for protective layers 37 . Low crystallization temperature is also a characteristic of ALD 14,38 , and is needed to avoid degradation of sensible photoabsorbers such as CIGS and CZTS as we demonstrated in other works 23,24 .
In this work we report the degradation and regeneration mechanisms over 1000 hours anodic photocurrent of the NiO protected layers deposited at low temperature (100 ºC) in a single-step ALD deposition. Due to higher crystallinity and better stoichiometry as deposition temperature increases (100 to 300ºC), which is expected to reduce acceptors density, the higher electrical conductivity has been found at the lower temperature. Although, at this lowest temperature, the optical transmittance is smaller limiting also the layer capability.
On the other hand, NiO regeneration capabilities are shown with an in situ periodic cyclic voltammetry process, bringing the electrodes to cathodic conditions after anodic periods of different duration. Under these procedures, a partial recovery has been shown, achieving 85% of stability over 1000 hours. This opens a route for potential self-regeneration strategies of the catalyst/protective layers.

MATERIALS AND METHODS
ALD NiO has been grown on laboratory standard p + n silicon buried junctions prepared as frontal photoanodes and simultaneously on p + degenerately doped silicon, to simulate direct injection in dark conditions. p + -Si samples were created by cutting a degenerately doped silicon wafer (0.001 ohm·cm) into 1x1 cm 2 pieces, and 50 nm Al were thermally evaporated as back contact. p + n-Si samples were fabricated as in previous works 14,31,39 , where a 0.5 cm 2 active area was lithographically defined by SiO2 passivation on a silicon n-type wafer (0.1-0.5 ohm·cm resistivity). Boron was implanted in the defined front surface and activated by rapid thermal annealing, creating a 200 nm p + region on top of the n-type substrate. As back contact, 1 µm Al/0.5%Cu was sputtered on top of 30 nm Ti to form a proper ohmic contact. p + -Si, and p + n-Si samples were sonicated for 5 min in a 1:1:1 isopropanol, acetone and DI water cleaning solution, followed by abundant rinsing and further 5 min sonication in DI water.
Before ALD deposition, sample's front surface was dipped in 0.1 M HF for 5 min, rinsed in DI water and immediately introduced in the ALD chamber.
A Cambridge Savannah 100 Atomic Layer Deposition system was used to grow NiO layers.
NiCp2 (nickelocene, bis(cyclopentadienyl)nickel) was selected as precursor and ozone (O3) as reactant 40,41 . NiCp2 container was heated at 80 ºC during the process. Successive pulses in N2 flow atmosphere were introduced to the chamber, 4 s pulses, 10 s maintained in the chamber ("expo" mode) and 20 s purges for NiCp2, and 20 s pulse and 40 s purge for O3. Under these conditions, layers have been grown at deposition temperatures of 100, 200 and 300 ºC for 1000 cycles, corresponding to roughly 50 nm layers, too thick for real protective layers but robust enough for manipulation during the degradation experiments. Samples were then soldered to a Cu wire using Ag paint and epoxy protected leaving the front area exposed (

Photoelectrochemical, Electrical and Morphological Characterization
NiO layers grown by ALD were fabricated at 100, 200 and 300 ºC deposition temperatures on top of p + n-Si, p + -Si, FTO and glass. This range was selected to investigate low deposition temperatures, as some photoabsorbers require avoiding high temperatures in order to prevent any alteration 23,24 . In Fig. 1 we present representative cyclic voltammograms of the first 100 cycles measured for these samples. In dark conditions, measurements were performed using layers deposited on top of p + silicon anodes, and, under 1 sun simulated illumination, using layers deposited on top of p + n-Si junctions. A clear dependence can be seen in both cases, with significantly higher currents when depositing at 100 ºC than at 200 ºC, further reduced when increasing to 300 ºC. Clear Ni 2+ /Ni 3+ peaks can be seen more cathodic than the OER onset potential, corresponding to Ni active sites 33,42,43 presenting steady increase with cycling. Slight activation is observed for the OER in the 1.8-1.9 V vs RHE range too. Both Ni peaks increase and initial activation are known to be caused by slight surface reorganization and residual Fe ion traces incorporated from electrolyte impurities, being beneficial for efficient OER 42 .
Transmittance on top of glass substrates was measured to be in the range of 40 to 80% (Fig. S. 2), reduced with lower deposition temperatures 36 . NiO is known to be less transparent when increasingly defective 36,44,45 . This is expected to be directly translated into reduced maximum saturation photocurrent for photoanodes under illumination. To exclude any influence of the electrolyte contact, Ni metallic contacts were deposited on top of p + -Si devices and I-V measurements were also carried out. In Fig. 2 it can clearly be seen conductivity to depend on deposition temperature as it has been seen in Fig. 1. Likewise, 100 ºC deposited layers are significantly more resistive than the p + -Si/Ni contact with a rectifying behavior. From SEM analysis (Fig. 3, a-c), we do not observe any significant differences upon deposition temperature. Samples present complete surface coverage, with a measurable thickness by cross section of around 50 nm, as predicted (Fig. 4). A granular rugose surface is observed for all deposition temperatures in the range of ~20 nm, close to the resolution limit of the SEM, slightly more pronounced for 300 ºC deposition temperature. Main differences are deduced from the XRD spectra. Bunsenite crystallographic structure is measured for all the deposition temperatures, but major orientation changes happen when increasing it. At 100 ºC a single exposed crystallographic facet is measured, with a diffraction angle of 37.37 º associated to the 111 direction. On the contrary, for the 200 and 300 ºC deposition temperatures, a preferential peak of 43.48 º is measured. Crystalline interplanar spacing (d) was measured to be reduced at higher temperatures, together with the full width at half maximum (β, FWHM). Using the Scherrer equation, crystalline domains (D) are measured to be 18.8, 14.7 and 24.1 nm for 100, 200 and 300 ºC. All this is information is resumed in Table 1. The variation with temperature are difficult to analyze, as there is a simultaneous change in the dominant orientation from 111 to 200. These trends are coherent with the results found for the sputter-deposited NiO films, which nucleate and grow preferentially in the (111) plane direction at room temperature if sufficient oxygen is available, whereas they show a (200) crystal orientation if deposition temperature is increased [46][47][48] . The slight shift to higher diffraction angles from 200 to 300 ºC can he attributed to compressive stress or to more stoichiometric NiO, reducing the cell volume (interplanar spacing, d) 49 .   HRTEM of NiO layers grown at 100 ºC and 200 ºC (Fig. 4)   Current mapping was measured by AFM. Again, lower conductivity is determined as deposition temperature is increased, following the same tendency as for cyclic voltammograms ( Fig. 1) and I-V measurements (Fig. 2). Regions between grains present no conductivity, less observed when increasing deposition temperature, probably by a better crystallization (such as it has been determined by XRD). For 100 ºC, all grains appear as conductive, with a +/-30% current dispersion. For 200 and 300 ºC, smaller grains are the ones appearing to be more conductive, although overall conductivity is lower compared to 100 ºC (observe the different current scale). NiO is known to present oxygen anions migration under applied potentials 54 , up to filament formation 55 , although is not observed in our potential range. Gathering all these results, we find a relation between the deposition temperature, the crystal orientation and the conductivity, determining the overall benefit in using 100 ºC deposited layers. As can be seen in Fig. 3d, increasing deposition temperature in ALD changes the preferential growth direction too. Higher deposition temperature gives more thermal energy, allowing the adsorbed atoms to diffuse longer distances and minimize surface energy to form more thermodynamically stable crystal structures, shifting from (111) XRD preferential plane orientation to the (200) one, improving crystallinity (Fig. 3d) When immersed in alkaline electrolytes and at anodic potentials, the NiO surface reacts with OHto form Ni(OH)2 and NiOOH, known to modify the OER catalyst performance of the NiO surface [60][61][62][63][64] . With the valence band maximum located below the water oxidation level, there is no resistance expected as NiO surface has an accumulation zone in contact with the catalytic region 59 .

Electrochemical and Photoelectrochemical Stability Characterization.
100 ºC sample was analyzed by electrochemical impedance spectroscopy (EIS), and two semicircles were observed ( Fig. 6a and 100 ºC sample was left 24 h in stability at 1.9 V vs RHE (Fig. 6b) and anodic current decay was observed from 3 towards 2 mA·cm -2 during first hours. This decay points toward a logarithmic dependence 65 of current on time (Fig. S. 4), and could suggest the formation of a self-passivation layer and its corresponding introduced charge transfer resistance, lowering the current. The increase in resistance is also observed in periodical EIS (Fig. 6c and Table S. 3).
Due to the temperature fluctuations in the laboratory affecting reaction kinetics, some variations are observed on the general current trend.  Thus, SEM images of 100 ºC sample revealed the presence of ~200 nm bumps formed on the surface after electrochemical testing ( Fig. 7a and b), which were confirmed by EDX to be

Fig. 7: a) SEM images of a NiO layer grown at 100 ºC on top of a p + -Si anode in 1 M KOH after finishing the stability measurement and b) cross section.
Stability measurements were performed using a p + n-Si photoanode, protected by a 100 ºC grown NiO layer (Fig. 8). Sample was polarized to 1.7 V vs RHE, the lowest potential to be considered in current saturation regime (where the current density is limited by the number of electron-hole pairs photogenerated in the p-n junction). Initially, this photoanode was cycled 6 times (voltage inversion from anodic to cathodic conditions) in 12 h intervals under 1 sun AM 1.5G illumination. After 500 h, the cycling period was increased to 48 h. Under these conditions, a higher current density decay can be observed from 12 to 9 mA·cm -2 . If the cycling is reset to 12 h, the photocurrent recovers up to 11 mA·cm -2 , and further reducing it to 3 h gives a slightly extra photocurrent reaching almost 12 mA·cm -2 at 1.7 V vs RHE (Fig. 8a,b). This test measurement was performed during 1000 h, where we stopped the experiment to be capable to further characterize the sample.
A photocurrent density constant decay of ~2 mA·cm -2 after 1000 h at 1.7 V vs RHE was observed. In Fig. S. 7 it is compared the current density during stability measurement at 1.7 V vs RHE with the slope at 8 mA·cm -2 (determined during the cyclic voltammograms) and current density values at the saturation potentials (1.8 V vs RHE). Also, some small fluctuations are visible (especially in Fig. 8b) due to lab night and day temperature variations.
These results reveal the more frequent application of cyclic voltammograms allows increased partial recovery of the passivation layer, and density current values are recovered.
Worthy information about these degradation and regeneration mechanisms can be deduced from a detailed analysis of the cyclic voltammograms measured during the 1000 h stability test (Fig. 8, c and d), presenting several remarkable characteristics.
First, an increase of the Ni 2+ /Ni 3+ redox peaks, which suggests a progressive increase of the nickel involved in the reaction, related bumps and flakes formed as observed comparing Fig.   3a and Fig. 7.
Second, a decrease of the electrochemical activity giving by a lowering of the slope in the 1. Third, a decrease of the saturation current, which must be attributed to a decrease of the amount of light reaching the photodiode. This is observed in Fig. S. 7 as a stable constant decay of the photocurrent measured for the performed periodic cyclic voltammograms until more anodic potentials than the stability potential.
It has been previously reported for NiO the existence of electrochromic darkening caused by surface ion adsorption 47 or electrochemical formation of less transparent NiOOH, with more Ni 3+ in the surface, having lower optical band gap and higher absorption coefficient than NiO and Ni(OH)2 45 . Bumps and nanoflakes formation observed in Fig. 7 and related to the electrochemical modifications measured are plausible causes of absorption and scattering causing the transparency variation over 1000 h.
Resuming, in Fig. 8c and d one can observe the effects of two different degradation mechanisms. On one hand, there is the progressive variation of the slope, a self-passivation mechanism (observed in Fig. 6b too). This points out an increase of the electrochemical overpotential. It is noteworthy to mention simultaneous resistive losses in the outermost layers of the NiO film is plausible. On the other hand, a steady decrease of the current density value in the current saturation potentials, which as we show below, is associated to the photon absorption efficiency due to a loss of the optical quality of the ALD-NiO surface.
So    Fig. S. 8a-c). Analyzing them closely, we can see the protective layer is still present, and the alkaline electrolyte found a path in few specific spots to reach the silicon etching it. On these few spots, in one case a particle was present and in the other one a hole, both of them exactly in the center. So, these pinholes in the protective layer are attributed to extrinsic or unwanted fabrication defects and not to intrinsic instability of the protection film, and thus, should be eliminated with optimal fabrication.

CONCLUSIONS
NiO transparent, protective, conductive and catalytic films grown by ALD have been used to avoid corrosion on frontal illuminated silicon photoanodes in alkaline electrolytes and anodic potentials while performing the oxygen evolution reaction.
Layers grown at 100, 200 and 300 ºC with thicknesses in the range of 50 nm present a polycrystalline structure with a change in preferential orientation when increasing from 100 ºC to 200 and 300 ºC, correlated with conductivity diminishment and transparency slight increase.
The sample grown at 100ºC presents the best electrical conductivity and optical transparency compromise.
Stability measurements on best-performing 100 ºC NiO-protected photoanodes present degradation mechanisms. Current density decays following a logarithmic time dependency, pointing at the formation of a self-passivation layer, limiting the electrochemical efficiency.
This is attributed to the interaction of the NiO layer and the alkaline electrolyte under anodic working conditions, forming of higher nickel oxidation states at the electrode/electrolyte interface, resulting in an OER overpotential increase. Under illumination, a second degradation mechanism is identified. A steady decrease of the current density at the current saturation potentials region is observed, attributed to losses of optical efficiency.
In the case of the electrochemical efficiency degradation, the application of periodic cycling ranging from anodic to cathodic voltage conditions corroborates the existence of a partial recovery procedure. Likely, these cathodic scans promote reduced forms at the surface of the ALD NiO layer, reversing some oxidation processes occurring at anodic potentials in highly alkaline media.
Applying this methodology under 3 hours cycling periodicity, over 1000 h stability measurements were possible with over 10 mA·cm -2 photocurrents, an 85% of the initial photocurrent.
The few corrosion spots observed have been correlated to extrinsic origin and thus, NiO ALD was demonstrated as a successful technique for long-lasting protective layers in anodic and alkaline conditions, avoiding corrosion of Si photoelectrodes or other alternative semiconductors. Furthermore, the use of low deposition temperature (100 ºC) for NiO deposition would allow for temperature-sensitive photoabsorbers to be implemented in PEC anodic conditions.
Consequently, this procedure opens a promising strategy for driving stability in photoelectrochemical devices. This regenerative protocol switching to cathodic potentials represents less than 1% of the working time under anodic conditions. However, there is still room to fully understand regenerative pathway's and to enable very long stability, and thus should be studied also in other materials suffering from electrochemical degradation.

Author Contributions
The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.