Harold J.
Fu
a,
Pakpoom
Buabthong
b,
Zachary Philip
Ifkovits
a,
Weilai
Yu
a,
Bruce S.
Brunschwig
*c and
Nathan S.
Lewis
*ac
aDivision of Chemistry and Chemical Engineering, 127-72, California Institute of Technology, Pasadena, CA 91125, USA. E-mail: nslewis@caltech.edu
bDivision of Engineering and Applied Sciences, California Institute of Technology, Pasadena, CA 91125, USA
cBeckman Institute Molecular Materials Resource Center, California Institute of Technology, Pasadena, CA 91125, USA. E-mail: bsb@caltech.edu
First published on 29th November 2021
Ni and NiOx-based protective thin films are shown to catalyze the oxidation of Si in the presence of O2 in strongly alkaline KOH(aq) even in the absence of illumination. The O2 in solution drove the open-circuit potential of the electrode to >0.4 V, which is positive of the Si passivation potential. The elevated electrochemical potential of the surface promoted formation of passive oxides on exposed Si regions of Si/Ni electrodes. Catalytic passivation of Si extended the durability of an np+-Si(100)/NiOx photoanode to >400 h while operating under simulated day/night cycles. In contrast, electrodes without a Ni(Ox) layer and/or without O2 in solution displayed direct etching of the Si and corrosion pitting during non-illuminated, simulated nighttime episodes of day/night cycling. The O2-derived catalyzed passivation of Si using thin films can be generalized to multiple phases of NiOx as well as to materials other than Ni. Relative to operation in aqueous alkaline conditions, decreasing the pH of the electrolyte decreased the dissolution rate of the protective oxide layer formed by the catalyzed passivation process, and consequently increased the durability of the photoanode, but yielded lower photoelectrode fill factors for water oxidation due to the relatively large kinetic overpotentials for the electrocatalyzed oxygen-evolution reaction at near-neutral pH.
Broader contextGenerating fuels from sunlight electrochemically allows for the storage and utilization of solar energy. Electrode stability is a key consideration for photoelectrochemical solar fuels devices. Protective coatings can enhance the durability of photoanodes under continuous illumination, but pinholes or other defects in the protective film can expose the underlying semiconducting light absorber to the electrolyte solution. Thus, materials that undergo corrosion by dissolution and pitting when in contact with the electrolyte will inevitably suffer catastrophic device failure. However, materials that form passive oxide layers can facilitate self-healing at pinholes and enable continued operation as photoanodes. We describe herein a catalytic passivation process that inhibits corrosive dissolution at exposed regions of Si photoanodes by inducing the formation of a self-limiting passivating oxide layer through the O2-derived oxidation of the Si surface catalyzed by metals or metal oxides. Understanding and generalizing this self-healing effect will inform device design so that solar fuels devices can operate with enhanced stability and efficiency when subjected to cycling between the presence and absence of sunlight. |
During O2 evolution, bare Si photoanodes grow an oxide (SiOx) passivation layer that resists corrosion and are thus resilient to pinholes in the protection layer.11,12 In the dark at open circuit, however, when the electrode is poised at the rest potential, the SiOx slowly dissolves in alkaline media, in which Si itself rapidly etches. This process can lead to electrode failure under day/night cycling conditions.6,13,14
The etch rate of SiOx is orders of magnitude slower than that of Si in alkaline solution, so the dissolution of Si in the absence of illumination can in principle be inhibited if the oxide layer can be maintained on the Si surface.15,16 Kinetic passivation of the Si thus occurs when the growth rate of SiOx exceeds its rate of dissolution. The addition of [Fe(CN)6]3− to a strongly alkaline solution has been shown to limit the degradation of a Si photoanode decorated with an array of Ni islands.17 The [Fe(CN)6]3− (E0 ∼ 0.36 V vs. the normal hydrogen electrode, NHE, for [Fe(CN)6]3−/4−)18 sets the surface potential of the Si to be ∼1.4 V vs. the reversible hydrogen electrode (RHE) in 1 M KOH(aq), driving the formation of oxide on exposed regions of the Si surface. Consequently, regions of the Si that are exposed to the electrolyte between the Ni islands, or even under a porous Ni oxy-hydroxy film, are inhibited from corrosion and dissolution due to the formation of the protective, passivating oxide (Scheme 1a). The durability of such interfaces is then limited by the rate of dissolution of the Si oxide in the electrolyte of interest.
Scheme 1 Schematic of two Si passivation mechanisms for a Si anode decorated with Ni(Ox) islands in alkaline electrolyte at open circuit in the dark. (a) [Fe(CN)6]3− is introduced as a protective electrolyte and acts as an oxidizing agent for Si.17 (b) Ni(Ox) uptakes O2 and catalyzes Si passivation. The electrode potential shifts positive, permitting holes from the Si valence band to react with the electrolyte and oxidize Si to SiO2. |
np+- and p+-Si electrodes decorated with arrays of 3–6 μm diameter Ni islands have been reported to exhibit extended stability (>24 h) in dark, open-circuit conditions and in the absence of protective electrolytes, even when >80% of the Si surface is exposed to the alkaline electrolyte.13,17 An understanding of the chemical and spatial details of the failure processes of Si photoanodes and of the fundamental reasons why unprotected Si photoanodes do not readily fail by corrosion could provide a basis for strategies to extend the operational lifetimes of photoanodes.
Herein we have investigated the mechanism by which Si can exhibit extended operation for water oxidation under day/night cycling in alkaline electrolytes as well as under other conditions where Si would be expected to rapidly etch and undergo pit corrosion. A hypothesis is that the presence of O2, along with an oxygen catalyst or a protective layer that can react with O2, produces a surface potential that drives formation of Si oxide. This process (Scheme 1b) is analogous to that of a protective one-electron redox-active oxidant in the electrolyte. We demonstrate that metal or metal oxide thin films such as Ni or NiOx, in the presence of O2, catalyze the dark, passive oxidation of Si in alkaline electrolytes. Such films can additionally induce local oxidation of exposed Si at pinholes or other film defects and thus lead to inherent defect tolerance of such interfaces. The catalyzed passivation of Si using Ni or NiOx thin films is thus compatible with previous investigations of Si/Ni-based photoanodes that were only evaluated under constant illumination. This behavior provides a strategy for extending the stability of photoelectrodes at exposed regions during periods when the surfaces would otherwise be subject to etching and/or associated corrosion at pinholes in protection layers.4,5
Fig. 1 Cyclic voltammograms of p+-Si in O2- (blue) and N2- (black) saturated, 1 M KOH(aq) at a 1 mV s−1 scan rate in the dark. |
The dissolution rate, rD, of p+-Si(100) in 1 M KOH(aq) at open circuit (∼−0.1 V vs. RHE) was rD > 102 nm h−1 at room temperature.17 Fig. S1 (ESI†) shows rD as a function of the electrode potential. rD at a given potential was determined by etching exposed Si on an electrode patterned with a thermal oxide mask in 1 M KOH(aq), followed by removing the mask using buffered oxide etch, and measuring the resulting trench height using atomic force microscopy (Fig. S1c, ESI†). The rate decreased to 71.6 nm h−1 at 0.2 V vs. RHE and decreased further to a minimum of 1.7 nm h−1 at 0.5 V vs. RHE, consistent with the formation of a slowly dissolving SiOx layer at these potentials.20 The dissolution rate increased monotonically as the potential was increased to >0.5 V vs. RHE. Over the entire potential range, rD(E) correlated closely with the current density at the potential, E, of interest (Fig. S1b, ESI†). At low (<2 V) applied voltages, the Si dissolution behavior is consistent with expectations for a surface covered by a potential-dependent, steady-state thickness of Si oxide.21
Fig. 2 displays the open-circuit etching behavior of p+-Si(100), of p+-Si(100) with a thin film of 5 nm Ni (p+-Si(100)/Ni), and of p+-Si(100) with 60 nm of NiOx (p+-Si(100)/NiOx) in strongly alkaline 1 M KOH(aq) saturated with 1 atm of either O2(g) or N2(g). Electrodes tested in N2-saturated solution were used as a control during which O2 was purged from the cell. Unless otherwise stated, all of the films on Si were deposited as continuous thin films. Bare p+-Si(100) electrodes, in O2- or N2-saturated solutions, had open-circuit potentials, Eoc, < EPP with average potentials of Eoc = −0.07 and −0.11 V vs. RHE, respectively (i.e. more than 0.2 V negative of EPP). This low Eoc implies the direct dissolution of Si at open circuit regardless of whether O2 is present in the solution. In N2-saturated solution, p+-Si(100)/Ni or p+-Si(100)/NiOx surfaces initially exhibited Eoc ∼ 0.6 vs. RHE and Eoc ∼ 1.0 V vs. RHE, respectively; however, within a few hours the Eoc of both surfaces decreased to <0 V vs. RHE (Fig. 2b). In contrast, in O2-saturated solution, p+-Si(100)/Ni or p+-Si(100)/NiOx surfaces exhibited Eoc > 0.60 and >0.85 V vs. RHE, respectively, throughout the duration of the experiment (Fig. 2a).
The resulting etched electrode surfaces were characterized using scanning electron microscopy (SEM). After 120 h at open circuit in the O2-saturated electrolyte, SEM images of the p+-Si(100)/Ni surface showed circular etch pits (Fig. 2c). In contrast, after only 20 h at open-circuit in the N2-saturated electrolyte (Fig. 2d, inset) p+-Si(100)/Ni surfaces displayed inverted pyramids that visibly undercut the Ni film. Further, after 120 h continued undercutting led to a merging of the Si etch pits with the complete delamination of the Ni film on p+-Si(100)/Ni surfaces under N2, as confirmed by energy-dispersive X-ray spectroscopy (EDS, Fig. S2, ESI†). Qualitatively similar open-circuit etching behavior was also observed for p+-Si(100)/NiOx electrodes in either O2- or N2-saturated solutions (Fig. S3, ESI†). The cross section of the sample saturated in O2 showed that the underlying Si remained intact (Fig. S3c, ESI†).
To further investigate the potential-dependent behavior of the etching of Si electrodes, a p+-Si(100)/Ni electrode was held at E = −0.1 V vs. RHE in O2-saturated 1 M KOH(aq). After 20 h, SEM images showed inverted pyramids undercutting the Ni film (Fig. S4, ESI†), similar to the behavior of a p+-Si(100)/Ni sample in a N2-saturated solution.
Fig. S5 (ESI†) shows Eoc as a function of the thickness of the Ni film on p+-Si(100)/Ni electrodes in O2- and N2-saturated 1 M KOH(aq). In O2-saturated solutions, all electrodes showed initial values of Eoc between 0.7 and 0.8 V vs. RHE. Eoc remained in this range for the first hour, but after 120 h, Eoc was 0.29, 0.60, 0.64, and 0.73 V vs. RHE for 3 nm, 5 nm, 30 nm, and 60 nm thick Ni films on p+-Si(100), respectively. Fig. S6 (ESI†) shows SEM images of these electrodes after 120 h at open circuit in O2-saturated 1 M KOH(aq). Pronounced and spatially dense pinholes were produced on the p+-Si(100)/Ni electrode that had 3 nm of Ni, whereas the pinhole density decreased with increasing Ni thickness. The p+-Si(100)/Ni electrode with 60 nm of Ni exhibited much less undercutting of Ni. Regardless of the Ni film thickness, in N2-saturated 1 M KOH(aq), Eoc for the p+-Si/Ni electrodes decreased to <EPP within a few h of immersion, with 3 nm, 5 nm, and 30 nm thick Ni films requiring 1.6 h, 2.3 h, and 3.3 h, respectively, to produce Eoc < 0.2 V vs. RHE (Fig. S5b, ESI†).
Fig. S7 (ESI†) shows the behavior in O2-saturated 1 M KOH(aq) of p+-Si(100) electrodes coated with discrete, patterned Ni islands (μNi). A p+-Si electrode was patterned with an array of 60 nm thick, 3 μm diameter Ni islands (μNi) with a 7 μm pitch, p+-Si(100)/(μNi). This electrode exhibited a relatively steady potential of Eoc = 0.6 V vs. RHE for ∼120 h. Despite ∼86% of the Si surface being exposed to the electrolyte, the electrode did not exhibit characteristics of direct Si etching in the absence of illumination. In fact, SEM images showed no signs of inverted pyramid formation after the 120 h experiment, and the μNi array remained intact with minor radial undercutting (∼300 nm) of the μNi islands. The etching of the exposed Si surface was similar to the behavior exhibited by a Si/μNi electrode in KOH(aq) that contained a passivating electrolyte.17
Fig. S8 (ESI†) compares the electrochemical behavior of p+-Si/NiOx electrodes in O2- or N2-saturated 1 M KOH(aq). The Eoc of electrodes was measured for 6 h in 1 M KOH(aq), followed by chronoamperometry (CA) at 0.3 V vs. RHE for 1 h, after which cyclic voltammetry (CV) was performed from Eoc to +0.4 to −0.4 V vs. Eoc at 10 mV s−1. This cycle was repeated three times. At a potential that is positive of EPP, 0.3 V vs. RHE, reductive current was observed in both O2-saturated and N2-saturated solutions (Fig. S8b, ESI†).
As-deposited Ni films exhibit a relatively high overpotential for the OER in 1 M KOH(aq), in contrast to Fe-doped, Ni oxy-hydroxy (NiFeOOH) films.24Fig. 3 compares the open-circuit potentials in 1 M KOH(aq) for p+-Si(100) electrodes coated with either Ni or NiFeOOH films. Ni was converted to NiFeOOH by cycling the electrode potential between 0.3 and 1.63 V vs. RHE at 40 mV s−1 in 1 M KOH(aq) to oxidize the Ni and incorporate residual Fe from the solution.25 In both O2- and N2-saturated 1 M KOH(aq), p+-Si(100)/NiFeOOH electrodes initially displayed Eoc = 0.96 V vs. NHE, but Eoc decreased within <2 h of operation and converged towards Eoc values characteristic of as-deposited Ni films in both O2- and N2-saturated solutions. For the O2-saturated solution, Eoc = 0.65 V vs. RHE after 83 h, whereas for the N2-saturated solution, Eoc = 0.07 V vs. RHE after 4.3 h. The shift in Eoc of the Ni film in O2- or N2-saturated solution can be ascribed to changes in the relative amounts of oxidized Ni on the surface. The Ni oxy-hydroxy film readily converts to Ni oxide on the electrode surface at Eoc,26 because the potential for converting NiII into NiIII is positive of Eoc, lying in a potential region of ∼1.3 V vs. RHE.27
After day/night cycling, SEM images revealed isotropically etched Si at pinholes in the photoelectrode (Fig. 4d). The images revealed that the NiOx film remained intact on the Si surface with some undercutting around the pinholes, as expected for Si/Ni electrodes in alkaline media.6 In the Si 2p X-ray photoelectron spectrum (XPS) of the photoelectrode surface that was subjected to day/night cycling, a weak SiO2 signal (102.6 eV binding energy), but no Si peak (∼99 eV), was present.33 This behavior confirms that the exposed regions of the surface were small areas of oxidized Si under pinholes, with the majority of the unoxidized Si being obscured by the NiOx film (Fig. S11, ESI†).
Under similar day/night cycling conditions, an np+-Si/TiO2/Ni photoanode in 1 M KOH(aq) did not exhibit catalyzed passivation despite the presence of a conformal layer of Ni on the TiO2 film (Fig. S12, ESI†). Instead, the electrode exhibited similar open-circuit behavior to that of p+-Si/TiO2 electrodes in the dark (Fig. S9, ESI†). Although the photoanode maintained performance for the first 216 h of the run, from 216 h to 312 h the photocurrent density decreased by ∼17%. Further, the bias sufficient to produce 10 mA cm−2 of photocurrent density under simulated 100 mW cm−2 illumination increased by 90 mV (Fig. S12c, ESI†), indicating substantial corrosion of the underlying np+-Si junction. Large (>20 μm) inverted pyramid etch pits that undercut the TiO2/Ni film were observed by SEM after 312 h of day/night cycling (Fig. S12d, ESI†).
Fig. S14 (ESI†) shows the Ni 2p3/2 XPS data for as-deposited, np+-Si/NiOx photoelectrodes relative to np+-Si/NiOx photoelectrodes after 20 h in O2-saturated 1 M KOH(aq) at open circuit, as well as after extended day/night stability cycling (Fig. 4). Prior to XPS measurements, the open-circuit potentials for the latter two photoelectrodes were Eoc ∼ 0.6 V vs. RHE and Eoc ∼ 1.2 V vs. RHE, respectively. The XPS emissions at binding energies of 854.0 ± 0.2 and 855.8 ± 0.2 eV can be attributed to NiO, Ni(OH)2, and NiO(OH).34–36 The photoelectrode held at open circuit displayed a similar Ni 2p3/2 spectrum to that of the as-deposited NiOx film, indicating that a mix of nickel oxides was present on both surfaces. In contrast, the XPS data for the photoelectrode after extended day/night cycling revealed an emission at 855.8 eV but no lower energy emission (854.0 eV), indicating an increase in higher oxidation states of Ni after cycling.
Fig. 6 compares the performance of an np+-Si(100)/Ni photoanode under simulated day/night cycling in ambient (∼0.2 atm O2) conditions in 0.5 M K-Bi to the behavior under conditions analogous to those used for the KOH solutions. The steady light-limited photocurrent density was >24 mA cm−2 throughout the duration of each 120 h experiment. In 1 M KOH(aq), the photoanode exhibited qualitatively similar Eocvs. time behavior to that of the np+-Si/NiOx photoanode shown in Fig. 4, as expected due to the oxidation of surficial Ni to NiOx. In contrast, in K-Bi, Eoc decreased from ∼1.4 V to ∼1.0 V vs. RHE, followed by a comparatively gradual decline to Eoc ∼ 0.7–0.8 V vs. RHE, possibly due to a gradual conversion of Ni(Fe)OOH to NiO/Ni(OH)2. The photoanode exhibited a higher fill factor in 1 M KOH(aq) than in 1.0 M K-Bi(aq), with a ∼140 mV negative shift in initial onset potential and a reduction in series resistance from 15 Ω to 7 Ω as the pH was increased. These differences in performance were consistent with the increased overpotentials associated with the OER as the pH was decreased from highly alkaline conditions to near-neutral pH values.4,37 In both solutions, the photoanodes exhibited relatively unchanged voltammetric behavior throughout the 120 h experiment, but substantial changes to the surface morphology were nevertheless evident. Specifically, after cycling in 1 M KOH(aq), the photoanode surface was covered with circular etch pits, whereas no etch pits were observed on the photoanode that had been cycled in 1.0 M K-Bi(aq).
A combination of a Ni-based thin film and O2 in solution was required to facilitate Si passivation. Without the Ni(Ox) thin film, the Eoc of Si electrodes was ∼−0.1 V vs. RHE and thus below the anodic threshold for Si oxidation of EPP = 0.17 V vs. RHE (Fig. 1). At this potential, the observed inverted pyramid etch pits indicate rapid etching of Si(100) (Fig. 2d and Fig. S4, ESI†). Without O2 in the electrolyte, a Ni(Ox) film can initially define the surface potential of the electrode, likely due to the NiO, Ni(OH)2, and NiO(OH) species on the surface, but within a few hours the Eoc declined to <EPP and continued decreasing toward the open-circuit potential of Si (Fig. 2b). This decline indicates that the surface potential of the electrode is ultimately defined by Si dissolution. Conversely, Si(100) passivation in the presence of both O2 and a Ni(Ox) thin film is evident as shown by circular etch pits caused by isotropic SiO2 etching (Fig. 2c) and evidence of SiO2 at pinholes via XPS after testing (Fig. S11, ESI†).
The surface potential of the Si electrode at open circuit will equilibrate with the solution redox potential only if the solution couple can react at the Si surface. In the case of the O2/OH− (E0 = 1.23 V vs. RHE) redox couple, bare Si does not provide a low-energy pathway between the redox species. Thus, for a p+-Si electrode, the surface potential does not equilibrate with an O2-saturated 1 M KOH solution at open circuit. However, when the surface potential is controlled by a catalyst on the surface (p+-Si/Ni(Ox)), the potential responds to the presence of the O2 in the KOH(aq) solution.
The surface potential of p+-Si/Ni(Ox) electrodes is controlled by three factors: the redox couples within the NiOx film; interactions with solution redox couples; and dissolution of the bare Si. The Ksp of Ni(OH)2 is ∼10−15, so in strongly alkaline electrolytes only trace amounts of Ni(aq)2+ are present and solution Ni redox couples are not important.38 For a p+-Si/Ni(Ox) electrode under illumination or under anodic conditions, the Ni surface became more oxidized (Fig. 5). However, without O2 in the electrolyte, the Ni species on the surface of Si/Ni(Ox) electrode did not prevent the Eoc from decreasing to a value close to that of a bare Si electrode, below EPP (Fig. 2). Thus, for a p+-Si/Ni(Ox) electrode, the Ni solid-state redox couples alone did not dominate the electrode potential relative to Si dissolution. For p+-Si/Ni(Ox) in an O2-rich solution, the interactions between the Ni(Ox) on the surface of the electrode and the solution O2/OH− couple define the surface potential of the Si electrode, likely involving the oxidation and reduction of Ni species on the electrode surface.
When O2 is not present in the solution, few or no solution redox species are present. The solution potential is therefore largely undefined and cannot buffer the electrode potential despite the presence of a Ni(Ox) catalytic layer. Rather, the surface potential of the electrode is largely defined by Si oxidative dissolution, and as a result, the Eoc of the electrode rapidly decreases to negative of EPP.
The above observations suggest a correlation between the catalytic activity of the overlayer and Si passivation (Fig. S8c and S10, ESI†). This expectation is supported by the passivation of Si that was observed for p+-Si/Pt films in O2-saturated solutions (Fig. S9a and b, ESI†). The data are consistent with a model in which Si/Pt and Si/Ni surfaces are in communication with the O2/OH− solution couple. However, this simple correlation is not observed for Si covered by Co. The behavior of the Si/Co surfaces is consistent with complete air oxidation of the Co layer resulting in an electrically nonconductive overlayer,39 that precludes effective redox-catalyzed communication with the dissolved O2 to maintain the surface potential of the Si positive of EPP.
The Si/TiO2 surface is also electrically nonconductive, and hence leads to etching of the Si.40 The Si/TiO2/Ni interface has a buried Si surface, so changes in solution potential affect the Ni but do not affect the buried Si/TiO2 junction nor affect the surface potential of the Si.41
Several strategies have potential to further inhibit the dissolution rate of SiOx and extend the lifetime of the oxide-coated photoanodes. In one approach, the photoanode could be held at less positive operating potentials during the day cycles. For instance, an np+-Si/Ni photoanode in 1 M KOH(aq) (Fig. 6b) approaches the light-limited photocurrent density at an applied potential of ∼1.2 V vs. RHE rather than at 1.6 V vs. RHE. Consequently, operating at a less positive potential during the day cycles could inhibit the dissolution of the SiOx (Fig. S1a, ESI†). The oxide grown electrochemically could potentially be post-processed to a denser, more slowly dissolving oxide by thermal treatment, or could potentially be converted chemically to more inert materials such as Si nitrides, oxynitrides, carbides, or oxycarbides.
Operating the photoelectrode at pH 9.5 can also substantially inhibit dissolution of SiOx (Fig. S15, ESI†), although the lower pH introduces performance limitations due to increases in polarization losses and increased kinetic overpotentials for the OER. Additionally, the oxidized Ni film is partially soluble in K-Bi, and the associated loss of catalyst might consequently deleteriously reduce the device lifetime.42 The work herein clearly provides a systematic understanding of the durability of photoelectrodes by identifying the medium-term failure modes and mechanisms and then rationally implementing steps to mitigate the corrosion processes of concern.
An np+-Si/NiOx photoanode undergoing day/night cycling exhibited stable performance for >408 h. The catalyzed passivation process reduced the rate of corrosion of exposed Si at pinholes in the NiOx film. Catalytic passivation of Si was observed with multiple phases of Ni–Ni, NiO, Ni(OH)2, and NiOOH – as well as with Pt. Although Si-based photoanodes are susceptible to eventual failure due to slow SiOx dissolution and subsequent undercutting of the protective film, this effect can be circumvented by leveraging the thermodynamic stability of SiO2 in pH 9.5 at the expense of an increased resistance and positive shifts in the onset potential. This work underscores the ability of well-established Si photoanode configurations to withstand rapid corrosion associated with patterns of diurnal insolation. Moreover, the work demonstrates the benefits of a systematic approach to increase the durability of photoelectrodes by identifying the dominant failure modes of the device and taking rational steps to mitigate such processes.
Ni, NiOx, Pt, and Co films were deposited onto np+-Si or p+-Si wafers using radio-frequency (RF) sputtering (AJA Orion sputterer). Prior to deposition, the p+-Si electrodes were cleaned and etched in BOE by the same process described above for n-type wafers. For deposition of Ni, Co, and Pt, 100 W was applied to the metal target under 20 sccm of Ar flow at a 5 mTorr chamber pressure until the desired thickness of material was obtained. NiOx films were deposited applying 120 W to the Ni target under 20 sccm Ar and 1 sccm O2 at 300 °C at a chamber pressure of 5 mTorr. The deposition rate was calibrated via profilometry (Bruker DektakXT Stylus profilometer) of the film that had been deposited onto a glass slide. Unless otherwise stated, the thickness of Ni, Co, and Pt were 5 nm whereas NiOx and Ni islands had thicknesses of 60 nm.
A carbide scribe was used to cleave off edges of samples from the wafers, to prevent shunting during electrochemical operation. In–Ga eutectic was scratched onto the back of each sample and attached to Sn-coated Cu wire using Ag paint. The wire was threaded through a glass rod and the edges of the sample were sealed to the glass using epoxy. The epoxy was cured overnight before testing the electrode the following day. Electrode areas were determined using ImageJ software in conjunction with an optical image of the electrode surface.
For long-term photoelectrochemical stability experiments, the working electrode was placed <1 cm from the liquid surface and the electrolyte was replenished as necessary to maintain the liquid level. Illumination was provided with an ELH-type tungsten-halogen lamp and was calibrated to an equivalent power density of 100 mW cm−2 using a calibrated Si photodiode (Thorlabs). Photoelectrochemical stability was evaluated by cycling the electrode between day and night intervals within a 24 h period. Day cycles involved 6 h of continuous water oxidation under 100 mW cm−2 of simulated solar illumination at either 1.63 V vs. RHE in KOH or at 1.73 V vs. RHE in K-Bi, to ensure that the light-limited photocurrent density was reached in each case. Prior to each day cycle, the photoelectrodes were cycled between 0.63 V and either 1.63 V or 1.73 V vs. RHE in KOH(aq) and K-Bi(aq), respectively, at 40 mV s−1. Night cycles involved holding the photoelectrode at open circuit in the dark for 18 h.
X-Ray photoelectron spectroscopy (XPS) was performed using a Kratos Axis Ultra system at a base pressure of 1 × 10−9 Torr. Samples were irradiated with a monochromatic Al Kα source using X-rays (1486.7 eV) at 450 W. A hemispherical analyzer oriented normal to the sample surface was used to maximize the depth sensitivity. High-resolution spectra were acquired at a resolution of 25 meV with a pass energy of 10 eV. CasaXPS computer software was used to analyze the XPS data. The spectra were first calibrated by referencing the C 1s peak position to 284.8 eV, followed by peak fitting for Ni 2p3/2 to Ni subspecies.34,36 For simplicity, satellite and Ni 2p1/2 peaks of each subspecies were not used in peak fitting.6 The same peak width was used in all samples for a particular species. The XPS data were measured ex situ, which could potentially confound surface composition and oxidation states relative to those in the electrochemical measurements.
Inductively coupled plasma mass spectrometry (ICP-MS) data were collected using an Agilent 8800 Triple Quadrupole ICP-MS system to measure the extent of Si dissolution in 0.5 M K-Bi either at open circuit or at 1.63 V vs. RHE. To ensure that Si dissolution did not result from the borosilicate glass in the electrochemical cell, reference electrode, or working electrode, polypropylene and Teflon were used instead of glass. ICP-MS samples were collected periodically by removing 1 mL of solution at set time intervals. Before analysis, samples were diluted by a factor of 10 with 2% nitric acid. ICP-MS concentration standards were made by serial dilutions of a known concentration standard (Sigma-Aldrich) with 2% nitric acid. ICP-MS measurements were conducted using hydrogen as a reaction gas to eliminate interference due to atmospheric nitrogen in the measurement of Si. The total amount of dissolved Si from the electrodes was then calculated and normalized to the geometric electrode area.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d1ee03040j |
This journal is © The Royal Society of Chemistry 2022 |