Sputtered Nickel Oxide Thin Films on n-Si(100)/SiO2 Surfaces for Photo-Electrochemical Oxygen Evolution Reaction (OER): Impact of Deposition Temperature on OER Performance and on Composition before and after OER

Magnetron sputtered nickel oxide thin films deposited on the native oxide of crystalline n-Si(100) wafers are studied in dependence of the substrate deposition temperature (600 °C, 400 °C, 200 °C, and room temperature) using X-ray and synchrotron excited photoemission spectroscopy as well as cyclicvoltammetry under illumination. We show that the substrate temperature during nickel oxide sputtering governs the composition of the pristine NiOx film and the OER performance. Two dedicated nickel oxide species are found with Ni corresponding to stoichiometric NiO while Ni indicates an oxygen rich NiOx (x>1) phase. With decreasing deposition temperature, the ratio of Ni/Ni in the pristine NiOx film increases. Information depth dependent synchrotron related photoemission spectroscopy further suggests that oxygen rich NiOx is found on top of the surface and at the grain boundaries. The OER onset potential improves from 1.55 V to 1.1 V in correlation to an increasing Ni/Ni ratio in the pristine NiOx film and an increasing emission from a nickel oxyhydroxide phase (h-NiOx) after photo-assisted cyclic-voltammetry in alkaline solution. Upon electrochemical treatment, a reconditioning process is observed with the formation of h-NiOx that consists of Ni(OH)2 and NiOOH, while NiOx disappears.


Introduction
For nickel oxide thin films, the catalytic performance depends on structural and chemical properties which influence the electronic behaviour concerning energy distribution, density of states, as well as conductivity. The ability to split water is strongly affected by crystal orientation as shown for nickel oxide single crystals [1][2][3][4] . In addition, defects, phase boundaries, kinks, and step-edges are considered as reaction centers 2, 5-6 . The importance of porosity and grain size of NiOx thin films for electrochemical efficiency has been shown as well 7 . Recently, the interface properties of nickel oxyhydroxides on semiconductor substrates in contact to the electrolyte in a photoelectrochemical cell have been studied [8][9] . It is generally assumed that the catalytic active phase changes according to NiOx → Ni(OH)2 → NiOOH during electrochemical cycling in the anodic potential range. However, no defined interface experiments have been conducted so far. Therefore, we have investigated the deposition of NiOx onto well-defined n-Si(100)/SiOx substrates with the aim to analyze the junction properties before and after electrochemical anodization.
During magnetron sputtering the substrate temperature is one of the parameters that control the properties of the deposited material. For both, polycrystalline or amorphous thin films, defect density, phase purity, and grain size correlate with deposition temperature. Concerning most practical 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 A c c e p t e d M a n u s c r i p t F o r R e v i e w O n l y 2 applications with regard to water-splitting, amorphous and defect-rich thin films often exhibit an improved performance 10 . At first, the Si/NiOx interface has been optimized with respect to the pretreatment of the Si substrate prior to the nickel oxide deposition 11 . Best results have been obtained with natively oxidized silicon. Referring to our experiment, the next step is the optimization of the NiOx thin film itself by varying the substrate temperature.

Defective nickel oxide
In general, NiO tends to be non-stoichiometric with a concentration ratio of nickel to oxygen below one. With increasing nickel deficiency, the term defective NiOx (x>1) is introduced to describe an oxygen rich nickel oxide phase with unclear stoichiometry. Until now, stoichiometric Ni2O3 could not be prepared as single-phase material 12 which makes a clear characterization difficult 13 . Moreover, oxygen in Ni(OH)2 has a similar binding energy than in defective NiOx with differences in the spectroscopic signature of both materials in XPS being very small [14][15] . In the 1970s, nickel oxide thin films were prepared for the first time by oxidizing metallic nickel at moderate temperatures in air. It turned out that at RT more bulk like NiOx (x>1) with a layer thickness of several monolayers (ML) was formed besides NiO whereas at 250 °C NiOx (x>1) was mainly concentrated at the surface in the thickness regime below one ML. For the respective water free process, a similar temperature dependency was obtained [15][16] . Later, it was discovered that defective NiOx cannot be prepared in the presence of water due to the formation of hydroxides and oxyhydroxides 17 . At present, it is commonly accepted that NiOx (x>1) approximating the stoichiometry of Ni2O3 can only be formed in a water free environment at temperatures below 250 °C.

Other oxidation states: hydroxides and oxyhydroxides
Much better characterized are the hydroxidic forms of nickel, namely nickel hydroxide Ni(OH)2 and nickel oxyhydroxide NiOOH [18][19][20][21] . Both species play a major role in Ni-based batteries where they are used as cathode materials. Furthermore, hydroxidic nickel exhibits strong electrocatalytic properties with a huge potential for water splitting 22 . Bode et al. published a comprehensive model of the charging and discharging processes in nickel hydroxide-based batteries 23 . They identified two pseudopolymorphous types of Ni(OH)2 denoted with α and β, respectively. Similarly, β-and γ-NiOOH were identified as main constituents of nickel oxyhydroxide. However, a deeper discussion on the structure and on transformational processes of all these species can be found in literature [23][24][25][26][27][28][29] .
Linking deposition parameters of the thin films to electronic properties is limited by experimental restrictions probing the solid/liquid interface due to the critical behavior of the liquid film under vacuum conditions. This can be overcome using ambient pressure X-ray photoelectron spectroscopy (XPS) as it has been shown e.g. for the Si/TiO2/H2O interface with operando XPS 30 . Here, we investigate the NiOx film with photoemission before (in situ approach) and after (ex situ approach) the electrochemical treatment in the anodic regime (labeled before and after EC). The electrochemical performance during cyclic voltammetry (CV) is related to deposition temperature-controlled composition changes of the pristine material before and after the electrochemical treatment as derived from XPS. Furthermore, depth resolved NiOx composition profiles are determined with synchrotron-based photoemission (S-XPS) experiments using a UHV based transport box (quasi in situ approach).
We find that magnetron sputtering of nickel oxide onto an n-Si/SiO2 photo-anode leads to a mixed NiOx phase with Ni 2+ and Ni 3+ valence states due to stoichiometric NiO and an oxygen rich defective NiOx phase that is approaching "Ni2O3" in its stoichiometry. With decreasing deposition temperature, the "Ni2O3" concentration increases in the neat nickel oxide layer while in the electrochemical application the OER onset potential decreases. XPS after photo-assisted CV shows that dependent on the "Ni2O3" concentration in the pristine film a hydroxylated h-NiOx phase consisting of Ni (OH)2 and NiOOH was  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 A c c e p t e d M a n u s c r i p t formed, while "Ni2O3" disappears. As h-NiOx is also found in the 600 °C sample with initially no "Ni2O3", it is assumed that the hydroxylation proceeds also to some extent into the NiO grains.

Sample preparation
All experiments were performed on (001) phosphor doped n-type silicon wafers with a resistivity of 0.2 Ωcm obtained from Micro Chemicals. The samples were cleaned in an ultrasonic bath for 10 min in acetone and isopropanol followed by a short rinse with MilliQ water. The wafers are covered with a native oxide layer with a thickness of 4 ± 1 Å acting as passivation layer. In the following this kind of substrate is called n-Si/SiOx photoanode. We found that good photoelectrochemical performance in combination with a RT deposited NiOx thin film is achieved with this procedure (see 31 ).
The as prepared substrates were dried under nitrogen flow, before being introduced within 5 min into the ultra-high vacuum (UHV) deposition and analysis system called Darmstadt Integrated System for Fundamental Research (DAISY-Fun). Thin NiOx layers were deposited on the silicon substrates via reactive magnetron sputtering under UHV conditions at a base pressure of 8·10 -8 mbar with 19.6 sccm Ar (99.999 % pure) and 0.4 sccm O2 (99.995 % pure) flow at 0.03 mbar, using a 99.99 % pure nickel target at 15 W. The substrate temperature was adjusted to RT, 200 °C, 400 °C, and 600 °C with a total deposition time of 600 s resulting in 5 ± 1 nm thick NiOx layers as deduced from the Ni2p (deposit) and Si2p (substrate) XPS core level spectra.

Sample analysis
After the NiOx deposition, the samples were transferred under UHV to the analysis chamber which is equipped with a SPECS Phoibos 150 analyzer and a SPECS Focus 500 X-ray source providing monochromatized AlKα excitation at 1486.74 eV for XPS. The work function was determined by applying a bias voltage of -3 V. In addition to the lab based measurements at the DAISY-Fun system (Technical University Darmstadt), the pristine samples were also analyzed at the undulator beamline U56-2/PGM-1 at BESSY II (Helmholtz-Zentrum Berlin) using the SoLiAS (Solid/ Liquid Analysis System) end station, which is equipped with the same analyzer. By varying the excitation energy, the information depth can be tuned from extreme surface sensitive (monolayer regime) to more bulk sensitive (about 10 monolayers) conditions. The samples were transferred from the lab system to the synchrotron and vice versa using a dedicated UHV transport chamber with a small getter pump driven with a battery in order to avoid contaminations.

Electrochemistry
After the UPS/XPS measurements, the samples were taken out of the UHV system to perform electrochemical (EC) measurements. The sample transfer under ambient conditions was done as fast as possible to avoid unwanted contamination or damage. For this purpose, an ohmic back contact based on a gallium/indium eutectic from Alfa Aesar consisting of 99.99 % pure metal [10] was prepared. Fixed on an etched copper foil, the samples were inserted into a commercially available electrochemical cell (PECC-2 from Zahner) filled with N2 bubbled 0.1 M KOH as electrolyte. A standard Ag/AgCl reference electrode was used in a three-electrode setup to control the potentials. All EC measurements were carried out using a Zennium (Zahner) potentiostat with a 625 nm (2.14 eV) LED light source operated at a power of 200 W/m 2 to excite the Si substrate. After the EC measurements, the samples were shortly rinsed with MilliQ water and subsequently transferred back into the DAISY-Fun system for post EC UPS/XPS analysis.

Spectra analysis
The XP spectra were background subtracted using the Shirley method. A sputtered Au foil served as reference sample in order to calibrate the spectrometer by making use of the Au4f7/2 transition at  .0 eV and the metal Fermi level located at 0 eV by definition. The XPS core level peaks were fitted using a pseudo-Voigt function with a fixed ratio of 0.2 between the Lorentzian and Gaussian distributions. As result, binding energy, line width and intensity are obtained and compared on a qualitative basis. Beyond that, intensity changes within the same binding energy region (i.e. O1s or Ni2p) were treated on a quantitative basis since cross section and kinetic energy are unchanged. A more detailed description of the used full widths at half maximum (FWHM) and the initial binding energies of the Ni2p3/2 and O1s core level components is given in 32 . The synchrotron-based measurements were treated similarly. Excitation energy dependent cross section variations are neglected since only the O1s region with neighboring species is considered. Additional information is provided in the discussion.

Analysis of pristine NiOx in dependence on deposition temperature
The XPS O1s and Ni2p spectra of nickel oxide thin films sputtered at 600 °C, 400 °C, 200 °C, and RT substrate temperature on n-Si/SiOx photoanodes are depicted in Figure 1. In the O1s spectra (see Figure 1a) the sample sputtered at 600 °C shows a single component indicating pure phase material, while a second component appears at higher binding energy with increasing intensity towards lower deposition temperatures. In parallel, the line shape of the Ni2p spectra in Figure 1b is most defined at 600 °C and broadens with reduced temperatures supporting the findings above. A fit of the Ni2p spectrum based on six single components given for the 600 °C sample represents pure NiO 33 with a residual standard deviation below one. The envelope of the resulting peak structure is used for further routines as described in the following. In addition, fit parameters of the selected peaks are given in Table 1 to make the result comparable with other sources 33-34 .  For 600 °C, the O1s and the Ni2p3/2 peak maxima are located at 530.1 eV and around 854.0 eV, respectively and fit well to stoichiometric NiO 33,35 . Using the relative sensitivity factors of the O1s and Ni2p3/2 transitions as tabulated in the manual of the SPECS analyzer (with RSF = 3.6, 26.8) a value of 0.85 ± 0.15 for the [Ni]/[O] concentration ratio is obtained supporting this assumption. With decreasing deposition temperature an additional Ni2p component becomes apparent by subtracting the intensity adopted line shape of the pure NiO film deposited at 600 °C (black line) from the spectra of the films deposited at lower temperatures (grey circles). The peak maximum of the additional low temperature component is shifted by 1.4 eV towards higher binding energy as compared to the NiO component 31- 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 A c c e p t e d M a n u s c r i p t 32 . The advantage of this difference spectrum approach is that complex details of the Ni2p line shape concerning the impact of multiplet structures 33,36 and final state effects [37][38] can be surmounted. The [Ni]/[O] concentration ratio obtained from the fits of the according peak structures related to the second Ni2p component as shown in Figure 1a and b (greenly colored) yields a value of 0.35 ± 0.15 for the RT sample which is less than the nominal stoichiometry of Ni2O3 but of the same magnitude. In contrast to 39 , the FWHM value of the second O1s peak at 531.0 eV is enlarged (see Table 1) although no carbon species are involved (see Figure 7). We assign the additional component to Ni 3+ forming an oxygen rich defective NiOx (x>1) amorphous phase or a likely dispersed (nanosized) mixture of NiO/Ni2O3 (composite phase). Surface related effects might have an impact on the stoichiometry as well. As a consequence, besides the existing NiO phase we name the additional phase "Ni2O3" in short to take the stoichiometric uncertainty into account. The O1s peak ratio of the emissions of the two nickel oxide phases can be determined and compared for the different sputtering temperatures. A clear trend is visible: Starting from the sample sputtered at RT the "Ni2O3" concentration is between 22 % and 23 %. This value continuously decreases with increasing sputtering temperature. Since the nickel oxide films have a thickness of about 5 nm, photoelectrons from O1s states excited with AlKα radiation are able to permeate the whole film as apparent in the O1s spectrum ( Figure 1a To obtain depth profiles of the phase composition of the pristine films, the samples were also investigated at the synchrotron facility BESSY II in Berlin. By varying the excitation or photon energy in the range from h = 580 to 1330 eV and additionally comparing to the lab-based measurements at 1486.7 eV (AlK), the electron escape depth  for the O1s core level is varied from 4 to 25 Å  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 A c c e p t e d M a n u s c r i p t according to the QUASES-IMPF-TPP2M code 40 using NiO as base material. Thus, the spectra recorded at 580 eV are extremely surface sensitive, while the spectra recorded at 1486.7 eV are more bulk sensitive.
Depth profiles based on the formal "Ni2O3"/NiO ratios derived from the two species of the measured O1s line were determined and are plotted over and has shown in Figure 2. In addition, best fits (solid lines) according to a basic layer model 41 as depicted on the right side of Figure 2 are included.

Figure 2:
Excitation energy dependent "Ni2O3"/NiO ratios over the electron escape depth  (photon energy) in pristine nickel oxide thin films sputtered at different substrate temperatures on n-Si/SiOx. The evaluation is based on peak fitting of the respective O1s spectra measured at the synchrotron. In addition, lab system measured values at 1486.7 eV (AlKα) are added. The "Ni2O3"/NiO ratios do not follow the exponential fits that are expected for the simple layer structure as sketched on the right-hand side. Speculative models with a more complex geometry of the "Ni2O3" (dark green) depth distribution in the NiOx films are given with "Ni2O3" still forming the topmost layer in the RT case. In parallel, "Ni2O3" fills the grain boundaries between the NiO grains which become larger with increasing deposition temperatures.
While for a simply layered structure where a thin "Ni2O3" overlayer covers the underlying NiO an exponential decay of the surface layer intensity I to the bulk intensity INiO with increasing photon energy is expected, the measured "Ni2O3"/NiO ratios differ significantly from this behaviour. Best fits (solid lines) according to the basic layer model (see formula on the right-hand side of Figure 2) result in a thickness d of 0.5 Å, 0.3 Å, and 0.1 Å for the RT, 400 °C, and 600 °C prepared sample which is consistent but seems to be too small whereas I* is not physically assignable and therefore neglected. A more complex structure of "Ni2O3" in the nickel oxide film must be considered as shown in Figure 2 with "Ni2O3" forming an overlayer and filling the NiO grain boundaries to some extent. Besides intermixing of both phases, it is assumed that the NiO grains are enlarged towards higher temperatures. However, further experimental techniques are required to directly measure the deposition temperature dependent grain structure of the NiOx films.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 A c c e p t e d M a n u s c r i p t The enhanced formation of "Ni2O3" under RT conditions is beneficial for the electrochemical performance as it will be shown later in this work. While the oxygen evolution reaction (OER) under illumination mainly takes place at the solid/liquid interface, the n-Si/SiOx/NiOx interface itself also has a strong impact on energetics and charge transfer. It must be considered as well when optimizing the whole device. In Figure 3 various XP spectra of the RT, 400 °C, and 600 °C prepared samples before EC are shown that are related to the energy level alignment of the n-Si/SiOx/NiOx anodes in the dark. Fundamental parameters such as work function, valence band maximum, and band bending show distinct correlations with temperature. a) b) c) Figure 3: In situ measured XP spectra of differently prepared NiOx films before EC: a) Secondary electron cut off edges measured with a bias voltage of 3 V to determine the work function, b) Si2p core level emissions from the substrate measured through the thin NiOx films with according fits to estimate the energetic alignment of the inner n-Si/SiOx/NiOx contact with respect to the Fermi level, and c) valence band regions with onset and according fit to obtain the valence band maximum VBmax (more details are given in the text).
As it can be clearly seen in Figure 3a, the work function jumps down in one step from Φ = 5.5 eV for the RT prepared sample to Φ = 4.7 eV for the samples prepared at higher temperatures following the same trend of the Ni2O3 concentration as determined with XPS. In accordance to the presumed model of a dedicated "Ni2O3" layer on top of the RT prepared NiOx film, the value of Φ = 5.5 eV is assigned to trivalent nickel while Φ = 4.7 eV stands for bivalent nickel oxide. This further confirms the image of a completely closed "Ni2O3" overlayer in the RT case since the spectral features of the NiO secondary electron cut off as visible for higher temperatures are fully suppressed. It is assumed that for higher preparation temperatures most parts of the NiO grains are forming the topmost layer thus ruling the work function then. In this case patches of NiO and "Ni2O3" coexist on top of the NiOx film with NiO related areas getting larger for elevated preparation temperatures (see Figure 2). In parallel, the valence band maxima of the NiOx films steadily shift to higher binding energies with increasing temperature. Values of VBmax = 0.6 (RT), 0.8 (400 °C), and 0.9 eV (600 °C) are obtained from the XPS measurements shown in Figure 3c. The deviation of the energetic position of the Si2p peaks from the 8 expected one in n-doped bulk silicon accounts for a band bending at the n-Si/SiO2 interface. As expected, it is directed upwards due to pronounced Fermi-level pinning that is induced by oxygen related interfacial gap-states located 0.6 eV above the valence band edge 42 . Following the shift of the Si2p peaks in Figure 3b the band bending in Si increases roughly from 0.4 eV, to 0.45 eV, and 0.5 eV for the RT, 400 °C, and 600 °C sample, respectively. By making use of fundamental silicon related parameters like band gap EG =1.12 eV, electron affinity χ = 4.05 eV and conduction band minimum EC-EF ≈ 0.1 eV which on its part is dependent on dopant species and concentration level and which both are known, the discontinuity of the vacuum level EVac in the interface region and thus the involved dipole drop over the whole n-Si/SiOx/NiOx structure was calculated. A pronounced change is observed when going from the RT to the high temperature samples with δ = 0.95 (RT), 0.1 (400 °C), and 0.05 eV (600 °C), respectively. Energy diagrams of the layered device structure deduced from the obtained results are given in Figure 4. As it can be clearly seen in Figure 4, the silicon valence band edge is well aligned with the topmost NiOx valence states for the RT prepared sample. There is almost no offset visible compared to the high temperature preparations at 400 °C and 600 °C. While the band diagram itself does not allow for the determination of the current potential (IE) performance like fill factor or maximum power point, VBmax gives a good estimate of the achievable photovoltage Uph as upper limit. Moreover, the band alignment depicted in Figure 4 only represents the energetic situation under open circuit conditions in the dark neglecting the solid/liquid interface. The total performance of an integrated water splitting device under operation combines the IE behavior of the buried junction with the catalytic performance of the topmost layer in contact to the electrolyte. Therefore, cyclic voltammetry (CV) under illumination was applied to prove the IE performance of the whole n-Si/SiOx/NiOx device structure in contact to the electrolyte.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 A c c e p t e d M a n u s c r i p t The pristine samples were cycled under illumination in 0.1 M KOH. In Figure 5 CV curves of the NiOx films on n-Si/SiOx under illumination are depicted for the OER reaction dependent on the deposition temperature. The OER onset potential clearly decreases from 1.55 to 1.1 V with lower temperatures, i.e. with increasing "Ni2O3" concentration in the pristine films whereas the photo-current density at high potentials is limited by the light intensity. In addition, the first CV cycles show a distinct spread indicating a conditioning process as stated elsewhere 43 , which is most pronounced for the RT prepared photo-anode. In order to analyze this conditioning process, the samples were investigated again after photo-CV by XPS ( Figure 6).  After the electrochemical characterization, the samples have been investigated again with XPS using AlKα radiation. In Figure 6b, the Ni2p spectra after EC are shown for different deposition temperatures. An additional peak structure referred to hydroxylated nickel oxide h-NiOx is found by subtracting the NiO line shape as given by the pristine film grown at 600 °C with a nearly pure NiO composition that was fitted in intensity to minimize the respective NiO peak at 853.9 eV.   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 A c c e p t e d M a n u s c r i p t The O1s spectra in Figure 6a exhibit two main peaks located at 529.4 eV and 531.3 eV that can be assigned to NiO and Ni(OH)2 in accordance to 44 . In addition, emissions related to NiOOH (O2 -) and NiOOH (OH -) can be fitted next to the Ni(OH)2 peak assuming equal intensities 31 . The obtained binding energy values of about 531.0 eV and 532.0 eV that are slightly varying with the preparation temperature of the NiOx films are different from the values claimed in 44 but they are in the same region. Furthermore, ex situ performed electrochemistry leads to an additional carbon contamination as shown in Figure 7 for the 20 °C sample. Before EC there is nearly no carbon visible wheras after EC distinct emissions in the C1s region appear at 285.2 eV, 286.0 eV, and 289.4 eV attributed to adventitious C (Cadv) and to functional groups like -COH or -COx 44 . This behaviour is similar for the other samples (200 °C, 400 °C, and 600 °C) and thus not shown here. Consequently, the broad line at 532.9 eV in the O1s spectra of Figure 6a is assigned to oxidized carbon (see Table 2). Although it is known that the valency of carbon in COx affects the quantitative analysis 45 , this aspect is less regarded as the result of the O1s peak fitting is not very straightforward. Correlations with the preparation temperature of the NiOx films are not consistent, especially for the samples prepared at higher temperatures. At first glance, the NiO peak intensities in the O1s spectra first increase from 20 °C to 200 °C before they are decreasing towards 600 °C while the Ni(OH)2 peak intensities are fluctuating. On closer inspection, the relative NiO concentration is lowest for the 20 °C sample (5 %) compared to the other samples (between 11 % and 19 %). The same applies for the NiO/Ni(OH)2 intensity ratio that is much lower for the RT sample (0.07) than for the samples prepared at higher temperatures (between 0.21 and 0.35). This is in accordance to the relative Ni(OH)2 concentration that is highest for the RT sample (71 %) compared to the other samples (between 52 % and 54 %). Even an assumed error of 10 % in determining the peak intensity by fitting does not change this significance. Moreover, the Ni(OH)2/NiOOH intensity ratio is about twice as high for the RT sample compared to the other samples (here NiOOH is equal to NiOOH + NiOOH). Altogether this means that for the RT sample the relative Ni(OH)2 concentration is highest whereas the ones for NiO and NiOOH both are lowest. However, the volatile behaviour of the peak intensities for the samples prepared at higher temperatures cannot be explained in terms of the preparation conditions. This may be attributed to the ex situ transfer of the samples back to the XPS analysis system after EC under ambient conditions leading to an additional water and carbon related contamination or to an unknown electrochemical mechanism that changes the NiOx film in contact with aqueous KOH depending on the initial state. Obviously, hydroxides and oxyhydroxides forming the outmost reactive layer under OER conditions are mostly involved. The fact that in Figure 6a Ni(OH)2 and NiOOH constituents are both present leads to the terminus hydroxylated nickel oxide h-NiOx using one line shape (green shaded) that combines both species as shown in Figure  6b. The attempt to resolve both components in the Ni2p spectra as described in Biesinger et al. 34 was not successful due to the complexity of the system consisting of NiO, Ni(OH)2, and NiOOH related species then.
Besides carbon and water related contamination that occurs upon the electrochemical treatment, the impact of residual iron on the formation of the nickel hydroxide films in the electrolyte has been discussed in the past. It was found that traces of iron in the KOH electrolyte (on sub-ppm level) increase the activity of the initial NiOx films by a factor of more than 30 46 . There is clear evidence that this improvement is due to the incorporation of Fe into the NiOx film forming active sites. This is in accordance to previous work with Ni0.9Fe0.1Ox being reported to be the most active OER electrocatalyst among various transition metal oxides based on Co, Fe, Ni, Ir, and Mn 43 . Obviously, small amounts of Fe in NiOx are sufficient to improve the OER performance due to the formation of layered Ni0.9Fe0.1OOH oxyhydroxides with nearly every Ni atom being electrochemically active. Furthermore, the effect of iron on phase transition or formation as well as possible doping of the host material during electrochemistry is widely discussed [47][48][49] and confirms that Fe even in low concentrations can not be neglected. However, the shift of the OER onset potential in the cyclic voltammograms of pure Ni (OH)2  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 A c c e p t e d M a n u s c r i p t  Figure 5 in the same manner by about 130 mV towards lower potentials, there must be another reason for the occurrence of the different shifts depending on the preparation temperature of the initial NiOx films. Since our material consists of at least two phases, namely NiO and "Ni2O3" we assume that besides structural and morphological aspects, the ratio of two species in the compound and thus the amount of defective "Ni2O3" is responsible for the improved OER performance shifting the onset potential accordingly to the left. The improved incorporation of Fe in "Ni2O3" cannot be ruled out and should be taken into account in the future as this would affect the OER performance as well. Concerning the XPS measurements of the Ni2p region in Figure 6b, the h-NiOx/NiO emission ratio clearly increases with decreasing NiOx deposition temperature, as the "Ni2O3"/NiO emission ratio in the pristine samples did. Since the h-NiOx/NiO ratio seems to be larger than the "Ni2O3"/NiO ratio was in the pristine samples, we conclude that not only "Ni2O3" reacts to h-NiOx, but the hydroxylation reaction also involves stoichiometric NiO and proceeds to some extent into the NiO grains. A good fitting of the O1s spectra in Figure 6a is achieved without any "Ni2O3" component that would be expected about 0.4 eV lower binding energy than Ni(OH)2. We assume that "Ni2O3" is fully consumed in the hydroxylation process.
Dealing with chemical reactions that may occure on the pristine NiOx film assuming a mixture of NiO and stoichiometric Ni2O3 at the surface and at the grain boundaries the following reaction paths of a stoichiometric transformation under OER conditions in alkaline aqueous solution are considered: (I) Ni2O3 + H2O  2NiOOH only with Ni2O3, (II) xNi2O3 + yNiO + (y+x)H2O  yNi(OH)2 + 2xNiOOH with Ni2O3 and NiO (x, y are positve integers), and (III) Ni2O3 + NiO + 3 H2O ↔ 3Ni(OH)2 + ½ O2 also with Ni2O3 and NiO. While reactions (I) and (II) take place without any oxygen evolution, reaction (III) is conducted by an additional oxygen evolution competing with the OER. Apparently, the availability of Ni2O3 in the NiO grain boundaries of the pristine NiOx film is advantagous with respect to the formation of OER beneficial h-NiOx which is composed of NiOOH and Ni(OH)2 and which could be cleary measured with XPS. Similar results are found in XPS recorderd valence band spectra of the differently prepared n-Si/SiO2/NiOx photoanodes after the EC treatment as shown in Figure 8.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 A c c e p t e d M a n u s c r i p t In Figure 8a XP spectra of the nickel oxide thin films in the valence band region after EC are presented. In addition, the valence band spectrum of the RT prepared sample before EC is shown. All spectra are normalized to the Ni3d shoulder located at about 1.7 eV. Corresponding to the deposition temperature dependent increase in h-NiOx which mainly consists of Ni(OH)2 and NiOOH as displayed in Figure 6 we find an emission at around 3.2 eV binding energy indicating the transformation of most likely Ni2O3 to Ni(OH)2. We attribute this species to the OER reaction since height and thus concentration of the nickel hydroxide species is highest for the most efficient RT prepared NiOx thin film. Moreover, secondary electron cut off edges of the RT sample before and after EC are pictured in Figure 8b together with the spectrum of the 600 °C sample after EC. Obviously, the conditioning process upon the EC treatment leads to a distinctive change of the work function especially for the RT prepared sample with a value of Φ = 4.1 eV. Regarding the 600 °C sample, a closer look to the secondary electron cut off unveils a step like behaviour which hints to an overlap of two slightly different materials consisting most probably of Ni(OH)2 and NiOOH that are arranged on the newly formed surface side by side. This effect is less pronounced on the RT sample leading to the assumption that the surface is predominantly covered by the material that is better suited for the OER, namely Ni(OH)2.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 A c c e p t e d M a n u s c r i p t Based on the XP valence band spectra in Figure 8, energy level diagrams of the RT sample before and after EC are depicted in Figure 9. The valence band maximum shifts from 0.6 to 0.8 eV after the EC treatment while the position of the silicon valence band is assumed to be unchanged. In addition, the work function significantly changes from 5.5. to 4.1 eV which confirms the formation of a new h-NiOx related surface phase. As a consequence, the resulting dipole drop over the whole n-Si/SiOx/NiOx device structure shifts from δ = 0.95 to -0.45 eV with a pronounced change of sign. The Ni(OH)2 related state is with 3.15 eV far below the topmost position of the silicon valence band at the n-Si/SiOx interface which is located roughly 0.6 eV below the Fermi level. Thus, it is unclear how photogenerated holes participate in the OER related process by reaching these states. It is assumed that the Ni(OH)2 emissions in the valence band spectra in Figure 8a describe the final point of a dynamic reaction chain resulting in the evolution of oxygen which is not accessible with the methods used in this work. The OER at the n-Si/SiOx/NiOx photo-anode under alkaline conditions in aqueous solution is by far more complex with respect to energy and time related issues. Apparently, the Ni(OH)2 peaks shift with increasing deposition temperature from 3.15 eV towards 3.55 eV. Although a large error in the range of several hundred meV is expected due to averaging and normalization, the observed shift of 0.4 eV correlates with the shift of the OER onset potential in Figure 5 which is nearly the same with about 0.35 V and which goes into the right direction.  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 A c c e p t e d M a n u s c r i p t