1. Introduction

The worldwide industrial, technological development and all the products and materials in modern society were largely reliant on fossil feedstocks. As a result, CO₂ levels in the atmosphere has increased tempestuously and it becomes clear that this has substantial consequences for our climate. To maintain the quality of our life and continue sustainable development, searching alternatives to fossil fuels is urgent. Solar power is by far the largest source of renewable energy. It is a clean and inexhaustible natural resource, but tends to be intermittent and unpredictable [1]. Collecting and storing solar energy in the form of energy-dense H₂ through solar water splitting is proposed to be one of the most ideal routes for developing clean and sustainable energy in the future [2]. As mentioned by Ardo et al. in a recent review, several large research programs around the globe have been implemented to clearly understand the requirements and challenges over the past decade, with the aim of accelerating the development of the science and technology of solar water splitting devices [3].

The configurations of solar water splitting systems were discussed in several excellent reviews [4,5]. If we only consider the practical solar water splitting devices and systems that generate H₂ via proton/electron-transfer redox reactions under electrochemical potential formed by non-thermal photovoltaic action, the solar water splitting systems can be divided into two major configurations: photoelectrochemical (PEC) cell, and photovoltaic/electrolyser (PV/EC). The comparison from several aspects of the two configurations has been described in depth in some review articles [6,7]. In brief, the PV/EC configuration contains at least two separate components, with the light absorption component physically separated from the water-splitting/electrolysis component. In this configuration, the key aspects are the open-circuit voltage (V_oc) of the PV and the H₂ and O₂ evolution overpotentials, which are influenced by hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) catalysts [7]. Although the PV/EC configuration has the advantage to achieve a high solar-to-hydrogen conversion efficiencies (η_STH) over 10% [8,9], it is still too costly to compete with traditional methods using fossil fuels [10]. In the case of PEC cells, direct photo-electrolysis is a more elegant and potentially cheaper approach because it allows the combination of solar absorbers and catalysts into a fully integrated system. Thus, PEC water splitting has simultaneous functions of light harvesting and electrolysis in a single device. Moreover, the overall system cost of PEC configuration can be substantially reduced through further increasing the photo-energy conversion efficiency [11].

Designing the advanced semiconductor materials and deep understanding the fundamental principles of photoelectrodes are essential for constructing highly efficient, stable and low-cost PEC cells for practical water splitting. During the past few years, several reviews have discussed the screening and designing of semiconductor materials used for photocathodes and photoanodes [2,12–14]. According to the arrangement of the atoms inside the crystals, one can classified these semiconductor materials for PEC photoelectrodes into two categories: poly-crystalline and single-crystalline. The former mainly includes metal oxides, nitrides, oxynitride and chalcopyrite and so on. For example, TiO₂, WO₃, a-Fe₂O₃, ZnO, Ta₃N₅, TaON and BiVO₄ are used as photoanodes, while Cu₂O, CuFeO₂ and CuInS₂ are used as photocathodes. Generally, the carrier mobility and carrier diffusion length are very limited in poly-crystalline semiconductor materials [15]. The band bending at the semiconductor/liquid junction determines the charge separation, which is extremely sensitive to the physical and chemical properties of the semiconductor–electrolyte interface [16]. Moreover, there are lots of grain boundary, lattice mismatches, defects or trap states in poly-crystalline semiconductors, so charge recombination occurs readily. One the other hand, the single-crystalline semiconductors are widely used in the PV community, including Si and III–V materials (for example, GaAs, InP, GaN and GaP). Many of them present high carrier mobility, leading to superior charge transport properties in the relevant photoelectronic devices. They are usually well crystallized with few defects and lattice mismatches, and therefore show low bulk-recombination. Solar water splitting devices made with single-crystalline semiconductors can employ a solid-state buried junction to separate the photogenerated charges, which is beneficial to achieve a high device efficiency [17]. The costs of single-crystalline substrates, especially for those containing rare elements and compounds such as InP, might be a concern [18]. However, Si, as the second most abundant element in Earth’s crust, is earth-abundant and low cost. It is the most successful material in photovoltaics and microelectronics industries with a vast knowledge base. Its bandgap (1.12 eV) is relatively well matched to the solar spectrum and its theoretical maximum photocurrent can reach up to 44 mA/cm² under AM1.5 G one Sun illumination. These properties render Si a promising material for photoelectrodes in a PEC water splitting system.

To date, there are three main limitations of utilizing Si as photoelectrodes in a PEC cell. Firstly, a large proportion of incident visible light are reflected on a planar Si/water interface [19]. The second is that Si suffers from photoanodic corrosion, because it is easily oxidized when in contact with aqueous electrolytes. The electrically insulating SiO₂ will then result in device failure [20,21]. Thirdly, the sluggish water splitting kinetics on Si surfaces is another major limiting factor. To overcome these limitations, remarkable progress has been made on the fundamental principles, device designing and the integration of cost-effective catalysts [22].

In this minireview, we provide an overview of the recent advances of the Si photoelectrodes in PEC water splitting cells. To differentiate from many recent review articles, the emphasis of this work is on the design of highly efficient and stable Si photoelectrodes. Si photocathodes and photoanodes will be respectively presented and discussed with a particular focus on several basic strategies including surface textures, protective layer, catalyst loading and system integration. This review does not intend to be exhaustive but rather aims to present a snapshot of a rapidly changing area in as concise a manner as possible.

2. Evaluation of PEC water splitting

2.1 PEC activity

The activity of a PEC cell for water splitting can be directly evaluated by the H₂ or O₂ gas evolution rate (mmol/s) or photocurrent density (mA/cm²). In order to obtain a true η_STH, these measurements should be done in a two-electrode configuration under one sun air mass 1.5 global (AM 1.5G) illumination without any sacrificial reagents or external bias (pH gradient or applied voltage) between two electrodes.

If the amount of evolved hydrogen in the PEC cells could be collected and measured, η_STH can be defined as the ratio of output chemical/electric energy to input solar energy according to Eq. (1) [12]:

Central to the essence of water splitting is the photovoltage. The thermodynamic potential required to split water into H₂ and O₂ is 1.23 V. In practice, however, a voltage above 1.8 V is often required because of the transportation and ohmic losses and overpotentials at both the cathode and anode [2]. Crystalline Si (c-Si) can theoretically provide a maximum photovoltage of 0.8 V from a single junction. Although c-Si does not have enough band gap energy to split water into H₂ and O₂ simultaneously, it is good enough for the proton reduction half reaction, since its conduction band edge position is well above the hydrogen evolution reaction potential for reduction of water to H₂ [25]. It is useful to calculate efficiency of a photoanode or photocathode separately from the other half of the water splitting reaction, to allow for optimization of the electrode materials independently. In this case a three-electrode configuration is commonly used, in which the current between the working electrode (WE) and counter electrode (CE) is recorded as a function of the WE potential against the reference electrode (RE) [26]. A standard reference electrode should be used and the WE potential can be converted into standard potential versus reversible hydrogen electrode (RHE). For example, if an Ag/AgCl (3M KCl) is used as reference electrode, the WE potential can be referred to the RHE, according to the following equation.:

For a Si photocathode, the energy conversion efficiencies (η) can be calculated as follows [27,28]:

A similar method can be used for defining the η f of a Si photoanode:

Drawn from the above discussion, one should maximize all three important parameters (J₀, V_op and FF) to obtain a highly efficient Si photoelectrode.

2.2 PEC stability

In addition to improving the efficiency, there is a great need to improve long-term stability of Si based photoelectrodes for PEC water splitting. When the surface of photoanode (photocathode) is exposed to an aqueous solution, the photogenerated holes in the valence band (electrons in the conduction band) can oxidize (reduce) either the water or itself. Whether the semiconductor is thermodynamically stable under PEC conditions should be determine by the alignment of reduction potential (ϕ^re) relative to ϕ(H⁺/H₂) for the photocathode, and oxidation potential ϕ^ox relative to ϕ(O₂/H₂O) for the photoanode, according to the early research by Gerischer [29]. The stability change of the photocathode and photoanode as the shift of its ϕ^re and ϕ^ox are shown in Figs. 1(a) and 1(b). For the photocathode, it is thermodynamically unstable when the ϕ^re placed below ϕ(H⁺/H₂), while the photoanode is thermodynamically unstable when ϕ^ox is placed above ϕ(O₂/H₂O). From the calculated ϕ^re and ϕ^ox of Si in Fig. 1(c), one can conclude that Si is prone to be corroded when in contact with an electrolyte under anodic conditions. However, corrosion of semiconductors is complicated, because the number of available electrons or holes at the surface that participate in electrochemical processes varies depending on the illumination, biasing conditions, surface/bulk electronic properties and the solution composition [30]. Si photocathode is stable under cathodic conditions in theory. However, an insulating Si oxide can still form on the surface of bare Si photocathode over extended periods of time due to the existence of oxygen in the solution [20]. According to the previous study reported by Seger et al., n⁺p-Si electrode with a thick insulating oxide layer can form a 3.1 eV barrier in the acidic electrolyte that severely blocks the electron transport [31].

 

Fig. 1 Stability change of the photocathode (a) as its ϕ^re shifts down from above the conduction band minimum (CBM) to below ϕ(H⁺/H₂) and of the photoanode (b) as its ϕ_ox shifts up from below the valence band maximum (VBM) to above ϕ(O₂/H₂O). (c) Calculated ϕ^re (black bars) and ϕ^ox (red bars) relative to the normal hydrogen electrode (NHE) and vacuum level for a series of semiconductors in solution at pH = 0, ambient temperature 298.15 K and pressure 1 bar. The water redox potentials ϕ(O₂/H₂O) and ϕ(H⁺/H₂) (dashed lines) and the valence (green columns) and conduction (blue columns) band edge positions at pH = 0 are also plotted. Reprinted with permission from [32]. Copyright (2012) American Chemical Society.

Download Full Size | PDF

Since Si can’t be used directly in PEC-based systems due to the corrosion of the photoelectrode materials, and several strategies for stabilizing semiconductor/electrolyte junctions have been developed to mitigate this drawback. For example, corrosion rate of bare Si photocathode can be reduced under illumination by the use of relevant catalysts which improves charge transfer kinetics at the solid/electrolyte interface, and consequently reduce the surface oxidation. However, this approach still has the fundamental issue that it does not resolve the problem of degradation in the dark, since there is no benefit of photovoltage to provide a stabilizing cathodic bias [33]. The most straightforward approach to overcome this limitation is to encapsulate the Si photoelectrode by a chemically stable coating that effectively blocks the surface from the harsh reaction conditions. This protection layer must be highly stable but also sufficient conductive and transparent to maintain high light harvesting efficiency [34]. Although many progresses have been made, developing a stable Si photoelectrode is significantly challenging under PEC conditions. In most cases, a long-term current density versus time (J–t) experiment or a repeated PEC experiment is carried out under light illumination and/or dark to study the stability of a photoelectrode.

3. Basic strategies for Si based photoelectrodes

3.1 Surface textures

The conduction band minimum and valence band maximum of c-Si are located at different positions in wave vector space, indicating that the band edge absorption of c-Si is indirect. Compared with the direct band gap materials, the optical properties of c-Si are relatively poor, thus at least a 50 μm-thick absorber is needed in order to achieve substantial optical absorption [35,36]. The 1.12 eV band gap of c-Si allows effective light absorption at wavelengths from the UV to the near-infrared. Its theoretical maximum photocurrent can reach up to 44 mA/cm² under AM1.5 G one Sun illumination. However, a problem with a planar Si/water interface is that about 25% of the incident visible light is reflected [19]. Similar to structures used in solid-state photovoltaic cells, surface textures can be introduced to the Si photoelectrode, so that light absorption can be enhanced and hence the η can be boosted [37]. Various surface textures on Si surface were introduced and analyzed, including Si nanowires array [38,39], Si nanopore array [19], Si nanohole array [40], Si nanopyramid array [41], Si microwires array [42–44], Si macropore array [45], and Si micro-pyramids (MPs) array [46]. Efficient light absorption from surface textures can be achieved due to that the internal reflection can be enhanced, the absorption wavelength range can be broadened, and the sensitivity to both incident angle and light polarization can be minimized [22,47]. In addition to enhanced light absorption, surface textures on Si photoelectrodes possess other advantages: (1) textured morphology can lead to significantly reduced costs of solar electricity generation by using less material, while the conversion efficiency of the devices is close to or improved than that of the planar one [48]. (2) surface textures also provide enlarged junction areas and more efficient collection of photogenerated minority carriers [49]. (3) materials that have large value of roughness factor should relax the constraint on the needed catalytic activity, because the photogenerated charge-carrier flux can be distributed over a high internal surface area [50]. (4) photoelectrodes with surface textures also result in a reduced overpotential because of a decreased local current density flowing through the electrode–electrolyte interface [19]. Despite these potential advantages above, many problems have to be addressed when surface textures were introduced. One problem is that the textured surface may produce impurities and defects, which can generate surface trap states that lower the η of devices in comparison to the theoretically estimated value [51]. Therefore, designing of a novel textured surface with superior anti-reflectance and low surface recombination loss is essential to achieve efficient solar water splitting. Another problem is that the textured surface makes the full coverage of the protective layer on Si wafer difficult, and therefore degrades the device stability. As such, for a Si wafer with open surface textures it is highly desirable to deposit a conformal and thin layer of protective layer, along with the enhanced light absorption and low surface recombination [41].

3.2 Protective layer

Due to the instability of Si toward corrosion, the surface of Si must be protected by a protective layer in contact with aqueous electrolytes. The extensive historical efforts on the development of protective layer to stabilize photoanodes and photocathodes for water splitting have been reviewed recently [52,53]. In addition to autologous stability, an effectively protective layer should be electrically conductive, optically transparent, and compact enough to prevent the semiconductor surface from directly contacting the electrolyte, and should efficiently suppress carrier recombination on the Si surface through surface state passivation [30,54]. Clearly, not all coatings will fulfill all of these requirements. Metals have been demonstrated as protective layers due to the simple fabrication process, outstanding intrinsic conductive properties and excellent stability in extreme electrolyte. Generally, noble metals are widely used, but some non-noble metallic elements may also be adopted depending on their working condition and device configuration. For example, Ti film was reported as a protective layer for photocathodes in acid electrolyte [31,55], while Ni film can be used as a protective layer for both the photocathodes and photoanodes in alkaline electrolyte [21,56,57]. However, such strategies will block light from reaching the Si substrate due to the parasitic light absorption of the metallic layers when a setup with front illumination is used. Transparent metal oxide semiconductors such as Al₂O₃ and TiO₂ are thus widely used as passivation layers for Si based solar cells [58,59]. They have proved to be effective for PEC electrodes. To date, TiO₂ is the most extensively used protective metallic oxide material for Si photoelectrodes. Bulk titanium dioxide (TiO₂) is chemically stable in pH 0−14 electrolytes for both HER and OER. The large bandgap of TiO₂ (~3 eV) limits the absorption to the small fraction of the solar spectrum in the ultraviolet, indicating its excellent optical transmittance. However, its conductivity is difficult to be controlled which is markedly different under different deposition conditions. In addition, the polycrystalline TiO₂ films should be thick enough to avoid the permeation of electrolyte, which, however, were found to inhibit charge transfer from the substrate to the electrolyte, resulting in poor PEC performance [52,60].

3.3 Catalyst loading

In 1969, a conventional statistical model for the charge-transfer kinetics of semiconductor electrochemistry under equilibrium or non-equilibrium steady state was proposed by Gerischer. Further study by Shreve and Lewis found that the conventional PEC model is in accord with all available experimental data [61]. In this model, the charge-transfer kinetics on a bare Si interface without catalysts depend on the change of the surface charge density, which is involved in processes such as charge storage at the surface, charge transfer to the solution and charge recombination [62]. Si is neither a good HER catalyst nor a good OER catalyst with extremely low exchange current density. The reaction kinetics of the HER (2-electron) and OER (4-electron) are normally multielectron processes, which are typically sluggish on Si/electrolyte interface [63]. A large overpotential can be observed for useful current values. To improve the reaction kinetics at the surface, a catalyst is normally deposited on top of Si photoelectrodes to improve the interfacial electron transfer kinetics, which may dramatically reduce the overpotential for HER or OER [64].

At present, platinum (Pt) is the most preferred catalyst for HER, while ruthenium (Ru) and iridium (Ir) oxides are more suited for OER, due to their high catalytic activity and durability under both acidic and alkaline conditions [27,65–67]. However, the low terrestrial abundance and high cost of aforementioned metal sources motivate the search for earth-abundant alternatives to realize commercial PEC water splitting [68]. In terms of HER, for example, transition metal dichalcogenides (TMDs) such as molybdenum sulfide (MoS₂) and tungsten disulfide (WS₂) have been identified as the most attractive candidate for HER catalysts in acidic solutions [69,70]. Extensive research efforts have been directed toward TMDs, since they have been the most promising earth-abundant alternatives to Pt for catalyzing the HER [71]. Bulk TMDs are usually poor electrochemical catalysts because the exposed basal planes are HER inert, while its edge sites show highly catalytic activity [72]. In order to increase the edge sites and thereby improve the HER activity, various engineering approaches have been investigated. One approach is to engineer TMDs material with high geometric nanostructures to expose high densities of catalytic active sites [73]. Doping foreign atoms into TMDs to form ternary structures also plays a pivotal role in improving the HER activity. These foreign atoms included non-metal atoms and metal ion [74,75]. Similarly, transition metal phosphides (TMPs), transition metal selenides (TMSs) and transition metal carbides (TMCs) have also been intensively studied, with some of these materials functioning as HER catalysts in both basic and acidic media [64]. Metal-based alloys are other highly studied earth-abundant HER catalysts, however, they quickly corrode in acidic environments. Interestingly, both high catalytic activity and stability have been obtained under alkaline conditions [76]. Zhang et al. demonstrated that NiMo is an effective HER catalyst in alkaline media, whose HER activity was comparable to Pt and superior to those for the state-of-the-art platinum-free electrocatalysts [77]. This metal-based alloys include CoMo, FeMo, NiFe and NiCo [64]. In the case of OER, many earth-abundant OER catalysts have been developed, which were found to exhibit comparable and even higher OER activity than RuO_x and IrO_x. These OER catalysts range from cobalt phosphate (CoPi) and Fe-, Ni-, or Co-based oxides, to Fe-, Ni- and Fe-based (oxy)hydroxides. Again they only worked in neutral and alkaline electrolytes [12,78]. Nevertheless exchange current densities of above-mentioned earth-abundant catalysts are still an order of magnitude lower than that of conventional noble metal based catalysts [50]. In general, high loadings are required for these non-noble HER catalysts to obtain low overpotentials. Therefore, there is an issue when these earth-abundant catalysts are integrated with Si photoelectrodes, due to the large parasitic optical absorption losses [79]. Additional strategies need to be developed to mitigate this deleterious trade-off between catalytic activity and optical transparency, such as using the back-illumination from the Si substrate and spatially and functionally decoupling the optical absorption and the catalytic activity [80].

3.4 Integration

An integrated Si photoelectrode system for solar-driven water splitting requires a Si absorber, a protective layer and a catalyst. Generally, the integrated system functions as several key processes: (1) absorption of solar photon energy resulting in charge carrier generation, (2) charge separation and transport to the electrode surface, and (3) the utilization of excited photo-carriers to drive catalytic reactions at the electrode surface [81]. Two factors are critical for constructing a highly efficient and stable Si photoelectrode. One is the interface between Si and catalyst, since it is essential to continuously transport photo-generated carriers from a photo-absorber to a solid/electrolyte interface, rather than recombination in the above processes. The other is the surface of catalyst, which should expose sufficiently active sites for HER or OER.

Designing surface textures on Si surface can enhance the light absorption, but there is an increased risk of inducing surface states. A protective layer has been shown to not only protect the Si surface from corrosion, but also passivate surface states and reduce recombination losses at the interface, however it should be thin or conductive enough to allow the charge transfer from the Si substrate to the solid/electrolyte interface. In addition, catalyst is normally needed on top of the protective layer to activate the PEC water splitting. In some cases, the catalyst overlayer can also serve as a corrosion protection layer to improve the stability of the Si photoelectrode [82]. However, the effective integration of electrocatalysts with Si absorbers is quite challenging. In addition to the synthetic difficulties of catalyst on Si photoelectrode, such as control over morphology, coverage and thickness, in many cases the Si photoelectrode and catalyst is chemical incompatibility, which can induce interfacial defects, states and recombination sites [83]. Thus, suitable band alignment and efficient charge transfer across the interfaces should be designed carefully. Let us take the Si photocathode as an example. In earlier studies, p-Si was studied as a photocathode by Koshida et al. [84]. In this case, the PEC performance of Si photoelectrodes rely on the band-bending at the Si/electrolyte junction to separate charges, which is extremely sensitive to the physical and chemical properties of the semiconductor/electrolyte interface [16]. Taking metal catalyst as an example, apart from catalyzing the PEC reactions, an impermeable catalyst overlayer on a Si surface forms a Schottky barrier which is controlled by the difference between the Fermi level of Si and the work function of the metal [27,85]. Therefore, the selection of the metal with suitable work function is also important to build an effective Si/electrolyte junction. However, an intimate Schottky contact generally generates a density of states within the semiconductor band gap, causing Fermi level pinning and thus decreasing band bending at the interface of Si/electrolyte [86]. To solve this problem, catalytic metal nanoparticles with a comparable size to the thickness of the space charge region can be incorporated. Meanwhile, a thin protective layer is needed to protect the Si surface from corrosion. The efficiency of the integrated catalyst/p-Si system is limited by the low photovoltage generated at the p-Si/electrolyte junction, since the valence band edge potential of the p-Si is not sufficiently positive with respect to the potential for reduction of water to H₂ [2]. To overcome this inherent limitation of a p-Si/liquid device, Lewis and associate fabricated a n⁺ thin layer on the p-Si substrate in order to increase the band bending at the n⁺p interface relative to the aqueous solution/p-Si interface [20]. The n⁺-layer provided a buried junction which decouples the band bending and photovoltage characteristics of the electrode from the energetics of the semiconductor/liquid contact.

Let’s assume that Pt is used as the catalyst for a thin TiO₂ layer protected n⁺p-Si photocathode for HER. In this case, the performance of a Si photoelectrode with buried junction upon frontside illumination can be directly modelled based on its performance as a Si PV cell coupled to the catalytic activity of the catalyst layer [9]. If Pt can be deposited as particles, which are discontinuous and smaller than the wavelength of incident photons, the layer will be optically “transparent”. On the other hand, new class of catalysts made from earth abundant materials (such as NiMo) have recently emerged to be strong candidate that can act as a cheap and efficient alternative to replace Pt in alkaline electrolyte. However, a high mass loading is needed to obtain sufficiently catalytic activity, resulting in considerable parasitic light absorption. Metals (such as Ti) have been employed as a protective layer for n⁺p-Si photocathodes. Unlike the previously mentioned metal/p-Si, the contact of a Ti layer with n⁺p-Si photocathode surface shows an ohmic behavior. It means that the charge transfer at the interface of Ti/n⁺p-Si will be very efficient. Similarly, its application has also been limited due to the parasitic light absorption/reflection. Recently, n-Si with a rear p⁺ emitter has been utilized as a photocathode due to a much higher carrier lifetime as well as a much better lifetime stability under illumination compared with p-Si [46,87]. Most importantly, it allows light illumination from the rear side of the device. Therefore the protective layer and catalysts at the front side can be sufficiently thick and high mass loading for designing more efficient and stable photocathode without concerning about any light shielding issues [80]. On this occasion, many state-of-the-art earth-abundant catalysts and protective layers could be successfully integrated onto the Si photoelectrodes. The efficiency and stability of such integrated Si photoelectrodes can be comparable or even better than the ones based on noble metal catalysts.

4. Si photocathode for HER

Efficient PEC devices for solar water splitting are constructed in either acidic or alkaline media, at which the concentration of charge carriers is the greatest [88]. Either protons or hydroxide ions can be transferred between the anolyte and catholyte to avoid a substantial increase in pH gradients, which may impede the water-splitting process [89]. The detailed mechanisms of HER in acidic and basic electrolyte are as follows [77,90]:

Observed from the Eqs. above, one can conclude that HER is a simple reaction that occurs easily and involves two steps: electron-coupled protons in the acidic electrolyte or electron-coupled water dissociation in basic electrolyte (the Volmer step for the formation of adsorbed hydrogen (H_ads)); and the subsequent combination of adsorbed hydrogen into molecular hydrogen (the Heyrovsky or Tafel step). Obviously, the mechanism for the HER is pH dependent: at low pH, the HER proceeds primarily by the reduction of protons; whereas at high pH, water is primarily reduced to produce hydroxide ions [91].

Figure 2 summarizes the reported η of various Si photocathodes for HER. Device structures and reported stability are collected from [20,41,46,79,82,92–113]. Concluded from the above published articles, most studies on Si photocathodes for HER are done under acid conditions, and Si is relatively stable under such conditions. Pt is reported as the most efficient and stable catalyst for HER in the acidic electrolyte. Up to now researchers have made considerable progress on the PEC efficiency and stability of photocathodes by the integration of Si, protective layer and Pt or other catalysts.

 

Fig. 2 Chart visualizing data on reported η of various Si photocathodes for HER. Device structures for the Si photocathodes with reported stability are noted. Their η with pure color are listed directly in the articles, while those with pure color and small checks are calculated from the J–V curves in the corresponding articles.

Download Full Size | PDF

In 2011, Boettcher et al. reported a high efficiency (9.6% ± 0.9%) for a planar n⁺p-Si photocathode with evaporated Pt catalyst, obtained under 100 mW/cm² W-halogen illumination. As pointed out by the authors, the photocurrent was overvalued due to the mismatch between the American Society for Testing and Materials AM1.5G and W-halogen lamp spectra. The actual efficiency is less than 8% under 100 mW/cm² AM1.5G illumination. In addition, such high efficiency decreased after 22 h of continuous operation since surface protective layer was missing (Figs. 3(a) and 3(b)) [20]. Seger et al. demonstrated that a sputtered 100 nm TiO₂ layer on top of a thin Ti metal layer may be used to protect a planar n⁺p-Si photocathode [114]. By using a Pt catalyst, the electrode achieved an onset potential (V_op) of 0.52 V_RHE and a stability of 72 h when illuminated by the red part (λ > 635 nm) of the AM 1.5 spectrum (Figs. 3(c) and 3(d)). The true η was unknown since the electrode was illuminated by the red part of the AM 1.5 spectrum. To cope with the limited light absorption of planar Si, in 2012, Oh et al. reported that nanowire (NW) arrays fabricated on p-Si surface via metal-catalyzed electroless etching can enhance the saturation photocurrent density obviously. When the surface of SiNW arrays were modified with a Pt catalyst, the Pt/SiNW photocathode exhibited a dramatically enhanced V_op of 0.42 V_RHE compared to 0.33 V_RHE of the planar Si. However, the report of stability was missing [38]. Choi et al. reported that a porous surface with a thickness of ~300 nm prepared on p-Si surface exhibited a better saturation photocurrent density and V_op than its bare counterpart. However, the porous surface turned out to have a significant negative effect on the long-term stability and make the fully covering of the protective layer on Si difficult. Stable photocurrent can only operate for 12 h, even if a conformal and thin Al₂O₃ layer deposited by ALD was used as the protection layer [54]. In view of this point, our group demonstrated an open pyramid-like surface nanostructure on normal multi-crystalline (mc) n⁺p-Si photocathode, which is capable of enhanced light absorption, low surface recombination, and being fully protected by thin protective layer (Figs. 4(a) and 4(b)). After coating a 4.5 nm Al₂O₃ layer (Fig. 4(c)), the pyramid electrode exhibits better PEC activity than that of the normal one (Fig. 4(d)). The photocurrent density of the Al₂O₃-coated n⁺p-Si photocathode with pyramid structure can keep stable at ~-27.5 mA/cm² throughout the 100 h run (Fig. 4(e)). Furthermore, its η can be up to 6.8% after Pt loading, due to the lowered surface light reflection, increased surface area and minority carrier life time on the electrode surface (Fig. 4(f)).

 

Fig. 3 (a) SEM image of the Pt layer deposited on the planar n⁺p-Si electrode, inset: cross-section of the electrode architecture; (b) J-E data of Pt/n⁺p-Si obtained in ultra-pure aqueous 0.5 M H₂SO₄. The data were collected as a single sweep from positive to negative potentials at 20 mV/s. Reprinted with permission from [20]. Copyright (2011) American Chemical Society. (c) 72 hours durability tests of in situ annealed of Pt/100 nm TiO₂/5 nm Ti/n⁺p-Si electrodes; (d) Cyclic voltammograms during the 72 h run. CVs were taken in 24 h intervals at 50 mV/s scan rate. The samples were irradiated with the red part of the AM1.5 spectra (λ>635 nm) in 1M HClO₄ and held at a potential of 0.3 V_RHE. Reprinted with permission from [114]. Copyright (2013) American Chemical Society.

Download Full Size | PDF

 

Fig. 4 SEM images of the surface of (a) normal-Si fabricated by normal etching process on a solar-cell production line and (b) pyramid-Si obtained by a following two-step MCCE. Inset: the corresponding enlarged images. (c) TEM image of the Al₂O₃-coated Si. The thickness of the Al₂O₃ layer was determined to be about 4.5 nm. (d) PEC J–V curves of normal and pyramid mc-Si n⁺-p photocathodes with Al₂O₃ protective layers. (e) J–t curve for the Al₂O₃-protected pyramid electrode at −0.8V_RHE. Inset: PEC J–V curves before and after a 100 h long time test. (f) The PEC J–V measurements for the Pt-impregnated and Al₂O₃-protected normal- and pyramid- n⁺-p-Si photocathodes. All samples were irradiated with 100 mW/cm² Xe lamp in a stirred solution containing 0.5 mol K₂SO₄ and H₂SO₄ (pH = 1) and scanned at 10 mV/s from left to right potentials. Reprinted from [41], with the permission of AIP Publishing.

Download Full Size | PDF

As mentioned above, p-type silicon is usually used, or a n⁺ thin layer is fabricated on p-type substrate to fulfill a photocathode. In this case, the PEC properties are mainly determined by the front electrode surface. Our group constructed a different Si photocathode, in which the np⁺ junction is formed by alloying aluminum with n-type silicon at the back side of the electrode. The PEC performance and stability of np⁺-Si with and without Al₂O₃ layer was much better than the normal n⁺p one. The η of the np⁺ photocathode enhances about doubly to 8.68% after Pt loading, compared with 4.51% for the corresponding normal n⁺p electrode. However, it is only stable for less than 73 h, due to the loss of Pt from the electrodes [46]. Despite the progress obtained, the efficiency of the current Si photocathode is still much lower than that (~18%) achieved by the corresponding photovoltaics. To date, there have been very few demonstrations of stable operation with efficiency higher than 10% in Si photocathodes, which are much simpler and cheaper than the amorphous Si/crystal Si heterojunction ones [28]. Remarkably, Cabán-Acevedo et al. obtained a η of around 10% by loading 5 nm Pt film on micro-pyramid n⁺pp⁺-Si, however, the stability was not reported [102]. Kast et al. inserted a interface layers consisting of Ti metal and highly doped F:SnO₂ between the protective layer TiO₂ and textured commercial n⁺p-Si photovoltaics. By controlling the oxide thickness and coupling the Ir catalyst, a high η of ~10.9% was obtained in base. However, only ~24 h stability was observed [93]. Recently, by fabricating a hydrophilic SiO₂ using oxygen plasma on micro-pyramid (MP) Si surface for the water diluted PtCl₆^2- containing solution, uniformly distributed small Pt nanoparticles (10–20 nm) were obtained using an electroless deposition method. After the SiO₂ beneath Pt is etched, our group reported a high η of 10.8% in TiO₂ covered Pt/n⁺p-Si photocathodes, together with a superior stability of over one week (Fig. 5) [94]. Such idea was further improved by adding isopropanol into the diluted PtCl₆^2- containing solution, which has been demonstrated to be a more powerful, simple, and fast method to produce a high density of Pt nanoparticles (5–10 nm in size) on Si surface. Thereofore Yin et al. achieved a 11.5% η and a stability of 7 days under continuous PEC-HER in Pt/n⁺np⁺-Si photocathodes stabilized using a 15 nm TiO₂ layer [92]. More recently, Mi and associate investigated the use of GaN nanowires with N-terminated surfaces grown directly on Si photocathode as a multi-functional passivation layer for Si photocathodes, which not only offers robust protection against corrosion but also significantly improves the reaction kinetics by reducing the charge carrier transfer resistance at the semiconductor/liquid junction. As a result, the as-prepared n⁺-GaN nanowires/n⁺p-Si photocathode showed an unprecedented long-term stability (>100 h) at a large current density (>35 mA/cm²) and the optimized Pt decorated n⁺-GaN nanowires/n⁺p-Si photocathode showed a η of 10.5% (Fig. 6) [95].

 

Fig. 5 (a) Schematic diagram of the TiO₂/Pt/n⁺p-Si photocathode. (b) Top-down SEM image and (c) cross-sectional TEM micrograph of the TiO₂ covered Pt/Si. (d) PEC J–V curves under chopped illumination of the Pt/n⁺p-Si (black) and TiO₂/Pt/n⁺p-Si (red). (e) Potential vs. time data at a constant current density of −10 mA/cm². The inset shows the detailed potential fluctuation during the first 5 h. (f) Consecutive J–V measurements of the TiO₂/Pt/n⁺p-Si photocathode before (0 h) and after (168 h, one week) PEC testing. All the photoelectrodes were measured in 1 M HClO₄ solution under simulated AM1.5G illumination. Reproduced from [94] with permission from The Royal Society of Chemistry.

Download Full Size | PDF

 

Fig. 6 (a) Schematic of n⁺-GaN nanowires/n⁺p-Si photocathode showing light absorption by the underlying Si wafer, electron transfer from Si wafer to GaN nanowires, and proton reduction on platinized GaN nanowires. (b) J–V curves of platinized n⁺-GaN nanowires/n⁺p-Si photocathode (red curve) and platinized n⁺-p-Si photocathode (black curve) under AM 1.5G one sun illumination and dark condition (blue curve). (c) PEC long term stability measurement for platinized n⁺-GaN nanowires/ n⁺-p-Si photocathode at 0 V_RHE. All the measurements were done in 0.5 M H₂SO₄. Reprinted with permission from [95]. Copyright (2018) American Chemical Society.

Download Full Size | PDF

Though Pt-group metals are the most outstanding HER catalysts, their prohibitive scarcity and cost impede the large-scale practical application. In the last few years, many new classes of HER catalysts composed of only cheap earth-abundant elements have been reported. TMDs nanomaterials have the potential to provide excellent catalytic activity in strong acids. Seger et al. formed a MoS_x/Ti/n⁺p-Si photocathode, in which amorphous MoS_x was electrodeposited on the electrode as a catalyst for HER and a thin interlayer of Ti as a conductive protective layer. A V_op of 0.33 V_RHE and J₀ of −16 mA/cm² were obtained under red light with AM1.5 cutoff (<635 nm), but only 1 h stability was proved [31]. Using a thin MoS₂ as both catalyst and surface protecting layer, MoS₂/Mo/n⁺p-Si electrode exhibited a negligible loss in performance after 100 h of operation, while a low saturation photocurrent density of ~-17 mA/cm² and a low η below 2% were observed due to the lack of surface texturing and the parasitic light absorption of Mo metal layer. Benck et al. demonstrated that the η of a Mo₃S₁₃/MoS₂/n⁺p-Si electrode can be up to ~3.1% due to the substantial enhancement of the catalytic activity of Mo₃S₁₃ catalyst clusters [82].

In order to increase the edge sites and thereby improve the HER activity, various engineering approaches have been investigated. Doping foreign atoms into TMDs to form ternary structures seems to play a pivotal role. For example, Cabán-Acevedo et al. integrated pyrite-type CoPS catalyst on n⁺pp⁺-Si MPs photocathode, achieving a J₀ of 26 mA/cm², V_op of 0.44 V_RHE and η of 4.7% under simulated 1 Sun irradiation (100 mW/cm²), while its stability was not reported [102]. Besides large V_op, high photocurrent is a prerequisite for highly efficient PEC devices. Both the light harvesting properties of Si and the optical transparency of catalysts and protective layer have significant impacts on the photocurrent. One the other hand, most reported earth-abundant catalysts suffer from strong light absorption or high reflection, which consequently reduces the light absorbed by Si and decreases the generated photocurrent. In this case, Ding et al. synthesized amorphous MoS_xCl_y catalyst via a low temperature chemical vapor deposition (CVD) reaction, which exhibited high electrocatalytic activity and high transparency with low optical losses compared with the MoS₂ catalyst. By integrating the MoS_xCl_y catalyst with n⁺pp⁺ Si MPs, a high J₀ of 43 mA/cm², V_op of 0.41 V_RHE and a η of ~6% were achieved, while the stability was reported to be 2 h due to the lack of an interface layer [99]. Our group explored how to integrate MoS₂ with Si photocathode efficiently. Vertically standing, conformal, and crystalline nano-MoS₂ films were prepared on a ∼2 nm Al₂O₃ protected n⁺p-Si photocathode, where Al₂O₃ not only passivates the Si surface but also protects the Si surface from the formation of interfacial layer during the MoS₂ fabrication. As a result, a V_op of 0.4 V_RHE and J₀ of 32.4 mA/cm² were obtained. In addition, the electrode exhibited long-term stability under 120 h of continuous HER. However, the fill factor (FF) is limited to 0.28 due to the increased resistance on the interfaces of MoS₂/Al₂O₃/n⁺p-Si and thus the η is limited to 3.6% [105]. Our group further demonstrated that an inserted Ti buffer layer played an essential role in protecting Si surface and reducing resistances of charge transfer on the electrode/electrolyte interface. An ultrathin and homogeneous Co doped WS₂ (Co-W-S) catalyst layer produced on Ti/n⁺p-Si surface reveals good PEC-HER performance and impressive operation durability for solar hydrogen production. Although J₀ was limited to −30.4 mA/cm² due to the parasitic light absorption of ∼8 nm Ti layer, a η of 4.0% and 6 days stability were finally achieved [103]. Recently, Mi and associate demonstrated the use of defect-free GaN nanowires as an ideal linker of planar silicon and molybdenum sulfides to produce a high-quality shell-core heterostructure. Both theoretical calculations and experimental results revealed that the unique electronic interaction and the excellent geometric-matching structure between GaN and MoS_x were highly favorable for charge carrier extraction. The integrated photocathode exhibits high catalytic activity and stability for PEC water splitting with a η of 5% and a stability of at least 10 h in 0.5 M H₂SO₄ under standard 1 sun illumination (Fig. 7) [100].

 

Fig. 7 (a), (b) Schematic illustration of the MoS_x@GaN NWs/Si heterostructure. GaN nanowire core covered with a uniform shell of MoS_x (light-green section) is vertically aligned on planar n⁺p junction Si (the left inset of (b) shows the unique electronic interaction of MoS₂/GaN while the right part signifies the outstanding geometric matching between MoS₂ and GaN). (c) J–V curves of MoS_x and MoS_x@GaN NWs/Si in 0.5 M H₂SO₄ under standard one-sun illumination. (d) Electrochemical impedance spectroscopy (EIS) analysis of MoS_x/Si and MoS_x@GaN NWs/Si. Inset graph is the magnification of EIS of MoS_x@GaN NWs/Si. Reproduced with permission from [100]. Copyright 2017, Nature publishing group.

Download Full Size | PDF

Generally, the FF is negatively correlated with the saturated photocurrent density, because increasing the catalyst loading may results in high FF, but also produce larger parasitic optical absorption losses and thus decrease the value of saturated photocurrent density. Such behavior is especially apparent for metallic alloy electrocatalysts such as Ni–Mo alloy, which is optically opaque and requires large mass loadings to achieve the comparable catalytic activity with Pt. Lewis and associate reported an n⁺p-Si microwire array coupled with NiMo catalyst and TiO₂ light-scattering nanoparticles. The best-performing NiMo/TiO₂/n⁺p-Si microwire photocathode exhibited a high FF of 0.48 and V_op of 0.42 V_RHE, while the η was limited to 2.9% due to the low J₀ of −14.3 mA/cm² under 1 Sun of simulated solar illumination. Furthermore, direct stability measurement was not performed because the NiMo electrocatalyst is known to degrade rapidly after several hours of operation in 1.0 M H₂SO₄ conditions [79]. Various options have been developed to mitigate this deleterious tradeoff between catalytic activity and optical transparency. For instance, Vijselaar et al. deconvoluted the contributions of catalytic activity and light absorption to the PEC performance using high-aspect-ratio Si microwires with a radial n⁺p junction by varying the fraction of catalyst coverage over the microwires, and the pitch between the microwires. The best-performing NiMo/Si microwire photocathode exhibited a near-ideal J₀ of −35.5 mA/cm², a V_op of 0.42 V_RHE and an FF of 0.62 in 0.1 M H₂SO₄ under AM 1.5G illumination, resulting in an ideal η of 10.8%. In terms of stability, however, only a constant current density of 10 mA/cm² for 72 h can be achieved in a buffered (pH 4.0) electrolyte of 0.2 M potassium hydrogen phthalate under AM 1.5G light [115]. To extend the stability of the Si photocathode, subsequent study by Vijselaar et al. showed the chemical resistivity and applicability of Ni-Si as an interlayer for the protection of an efficient Si microwire photocathode in alkaline electrolyte. Coupling NiMo as an active and stable HER catalyst, the Si photocathode with a η of 10.1% was stable in alkaline electrolyte, which maintained constant activity for 12 days of operation [96].

5. Si photoanode for OER

The detailed mechanisms of OER in acidic and basic electrolyte are as follows [116]:

OER in acidic electrolyte:

OER in basic electrolyte:

In contrast to HER, the OER is more complex and has more sluggish kinetics, because the generation of only one oxygen molecule involves transfer of 4 e⁻ and removal of 4 H⁺ from water. n-Si or p⁺n-Si can be utilized as a small band-gap photoanode. However, compared with Si photocathode, the oxidation of Si is much easier on the Si photoanode/water interface, where the holes excited by energetic incident photons will meet and react with Si surface atoms. This deleterious effect can substantially suppress the PEC process due to a thick insulating oxide layer blocking the charge transport. Over the years, significant attention has been devoted to obtaining sustained and efficient Si photoanode as a counterpart to the photocathode for the PEC water splitting. As shown in Fig. 8, the progress for the Si photoanodes was collected from [21,60,67,117–139].

 

Fig. 8 Chart visualizing data on reported η of various Si photoanodes for OER. Device structures for the Si photoanodes with reported stability are noted. The η of the Si photoanodes with pure color are listed directly in the articles, while those with pure color and small checks are calculated from the J–V curves in the corresponding articles.

Download Full Size | PDF

Noble materials are usually required for stable PEC-OER in acidic media to drive OER or to serve as protective coatings. As early as 1984, Contractor et al. demonstrated a meaningful stability for more than 4 days using a crystalline n-Si photoanode protected by a Pt thin film in 0.5 M H₂SO₄ (pH 0.3) [140]. Recently, Chorkendorff and associate deposited Pt islands on TiO₂ protected p⁺n-Si photoanode. A V_op of above 1.35 V_RHE and a stability of 60 h was observed in 1 M HClO₄ [34]. It means that although Pt is an excellent HER catalyst, it is not good enough for OER. In 2011, Chen et al. reported a Si photoanode based on semiconductor-insulator-metal (MIS) junctions consisting of n-Si, thin and high quality TiO₂ by atomic layer deposition, and a thin Ir film. This photoanode showed an extended lifetime of more than 24 h compared to less than 0.5 h for the unprotected one. The η was limited to 1% in 1 M H₂SO₄ due to the lack of buried junction from underlying photovoltaic substrate. Furthermore, the thickness of protecting TiO₂ layer needs to be controlled precisely to achieve stability without diminishing the performance of the photoanode [60]. Mei et al. reported on p⁺n-Si photoanodes protected and catalyzed by sputter-deposited Ir/IrO_x thin films without using an interfacial protective layer. The simple IrO_x/Ir/p⁺n-Si structures not only enabled a current density of 10 mA/cm² at 1.12 V_RHE (a η above 1%), but also showed a stability of at least 18 h in strongly acidic media (1 M H₂SO₄) [141]. However, due to the high cost and scarcity of noble metal catalysts, the aforementioned designs were not scalable. Unfortunately, the majority of earth-abundant catalysts for OER suffer from the deleterious degradation side reactions in the low pH region [142].

On the other hand, oxidation of a negatively charged OH^- is easier than oxidation of the water or acidic electrolytes, so electrocatalysts for OER exhibit much lower overpotentials in alkaline media than at neutral and low pH [88]. Thus, the development of cost-effective catalysts with high catalytic activity and stability in alkaline electrolytes for OER is highly desirable. Kenney et al. reported a stable Si photoanode, in which 2 nm Ni film on n-Si with native SiO₂ was shown to serve as both a protective layer and a catalyst. In 1 M KOH, the formation of a MIS charge-separation structure exhibited a relatively high photovoltage of ∼500 mV and a V_op of 1.07 V_RHE. However, the 2 nm Ni film was not able to completely protect the underlying Si photoanode. As a result, the performance of the device started to degrade after 24 h of operation in 1 M KOH electrolyte solution. It was observed that oxidation and hydroxylation of Ni species with higher oxidation states are formed at the Ni metal surface when Ni is in contact with the electrolyte at oxidative potentials, which reduce the overpotential for OER significantly. It was also found that the thickness of the Ni film is critical for achieving a high photovoltage: the thicker Ni suffers from low photovoltage due to the low Schottky barrier height formed by Ni/SiO_x/Si [21]. Based on this seminal work reported by Kenney and Dai, a great number of studies have been devoted to further improve the stability and efficiency of the Si photoanodes using earth abundant materials. In the case of stability, n type c-Si coated with electrically conductive, optically transparent, 10–100 nm thick TiO₂ protective layer exhibited enhanced stability (more than 100 h) under continuous operation for OER in alkaline electrolytes [67]. After this work, Shaner et al. reported that the NiCrO_x/TiO₂/p⁺n-Si microwire array photoanode enables the continuous water oxidation for >2200 h in 1.0 M KOH [137]. However, there is always a trade-off between light shielding and catalytic performance. As a result, the η was below 0.2% due to the limited current density. (Fig. 9)

 

Fig. 9 (a) Schematic of a structure that consists of a p⁺n-Si microwire-array conformally coated with a protective TiO₂ layer and subsequently coated with a NiCrO_x OER catalyst. (b) SEM image of a fully processed microwire array. (c) J–t curve for the NiCrO_x/TiO₂/p⁺n-Si microwire array photoanode under 1 Sun simulated illumination in 1.0 M KOH (pH 14). (d) Cyclic voltammograms taken at 10 h intervals throughout the duration of the stability test. Reproduced with permission from [137]. Copyright (2015) The Royal Society of Chemistry.

Download Full Size | PDF

As for the enhancement of photovoltage, the most straightforward way is to form a buried p–n homojunction by a doping/diffusion process, which provides higher photovoltage values. For example, freshly etched n-Si and p⁺n-Si photoanodes protected by a multifunctional layer of NiO_x yielded open-circuit voltages of 180 mV and 510 mV, respectively [121]. Despite the significant improvement in photovoltage, the doping/diffusion process generally requires high temperatures, and adds complexity to the formation of a functional photoelectrode. To deal with this problem, Digdaya et al. fabricated an MIS photoanode consisting of Ni/Pt/Al₂O₃/SiO_x/n-Si photoanode and demonstrated a long-term stability (more than 200 h) and a sufficiently high photocurrent and photovoltage after the activation of the metallic surface layer (η >1.5%) in 1 M KOH solution (Fig. 10). The high photovoltage was achieved due to the following factors: (1) the dual-oxide layer (Al₂O₃/SiO_x,) is used to engineer the semiconductor/insulator and insulator/metal interfaces separately, which has the advantages to unpin the Si Fermi level at the Si/insulator interface; (2) Pt/Ni metal layer is used to decouple the functionalities for a higher effective barrier height and a highly efficient catalyst for the composite photoanode; (3) dual metal overlayers as a simple protection route can cover the underlying Si photoanode completely, to achieve an operational stability of over 200 h [125].

 

Fig. 10 (a) Schematic of Ni/Pt/Al₂O₃/SiO_x/n-Si photoanode. (b) Representative J–V curves of the photoanode in contact with 1 M KOH under simulated solar illumination collected periodically during 200 h of operation. (c) J–t curve of the photoanode measured at a constant applied voltage of 1.7 V_RHE in 1 M KOH solution under simulated solar illumination. The inset shows current fluctuations due to bubble formation during the measurement. Reproduced with permission [125]. Copyright 2017, Nature publishing group.

Download Full Size | PDF

Zhou et al. demonstrated that the introduction of an ultrathin (2 nm), compositionally controlled interfacial cobalt oxide layer between the n-Si absorber and the protective, multifunctional NiO_x film can also increase the band bending at the interface. The stable NiO_x/CoO_x/SiO_x/n-Si photoanode exhibited a V_op of 0.99 V_RHE, a J₀ of ∼28 mA/cm² and a η of 2.1 ± 0.2%, when in contact with 1 M KOH [120]. Note that such performance was modestly better than that reported for a buried homojunction p⁺n-Si electrode that was freshly etched and directly sputtered with NiO_x [121]. Moreover, trace amounts of foreign atoms incorporated into the NiO structure drastically improve the performance of a photoanode. Mei et al. demonstrated that the PEC-OER performance of a NiO protected p⁺n-Si photoanode can be significantly improved by employing a pretreatment in Fe-containing KOH (10 mM Fe) [143]. The Fe treatment enabled p⁺n-Si photoanode to obtain a current density of 10 mA/cm² at 1.15 V_RHE and a stability of 300 h under red-light (38.6 mW/cm²) irradiation in 1 M KOH. After this interesting finding, in 2016, Yu et al. successfully fabricated a 6 nm thick NiFe alloy protective and catalytic layer on p⁺n-Si surface with a native SiO_x by a facile electrodeposition method. The NiFe/SiO_x/p⁺n-Si photoanode showed a low V_op of 0.89 V_RHE, a high J₀ of 30.7 mA/cm², a large photovoltage of 620 mV, a high η of 3.3% and a stability of 12 h in 1.0 M KOH under 1 sun illumination [118]. As proved by Yu et al., the excellent η is attributed to the high conductivity and electrocatalytic activity of the NiFe catalyst, and the large band-bending at p⁺n.

Indeed, designing high-quality interfaces is crucial for high performance PEC water splitting devices. Most recently, a facile integration between pn⁺-Si and NiFe-layered double hydroxide (LDH) nanosheet array by a partially activated Ni (Ni/NiOx) bridging layer was demonstrated by Guo et al. This NiFe-LDH/NiO_x/Ni/SiO_x/pn⁺-Si photoanode leads to PEC performance with a V_op of ∼0.78 V_RHE and a J₀ of ∼37 mA/cm², a η of 4.3 ± 0.2%, and retains good stability for 68 h in 1.0 M KOH, which is the highest OER activity so far reported for the Si photoanodes. (Fig. 11) This excellent PEC-OER performance was achieved due to a few key features: (1) the Ni metal layer with excellent stability is used as an effective protection and interface layer for buried junction-Si photoanode; (2) NiO_x has a high capacity of hole accumulation and strong bonding with the electrodeposited NiFe-LDH due to the similarity in material composition and structure, facilitating transfer of accumulated holes to the catalyst; (3) the rear p-n junction configuration allows back illumination from the Si photoanode, making NiFe-LDH sufficiently thick for more catalytically active sites without compromising Si light absorption [117].

 

Fig. 11 (a) Cross-sectional schematic and energy band alignment of the designed multicomponent NiFe-LDH/NiO_x/Ni/SiO_x/pn⁺-Si photoanode for water oxidation. (b) J−V behavior of NiFe-LDH/NiO_x/Ni, NiO_x/Ni and Ni coated SiO_x/pn⁺-Si photoanodes, and NiFe-LDH/NiO_x/Ni and NiO_x/Ni coated non-photoactive p⁺⁺-Si electrodes. Reprinted with permission from [117]. Copyright (2018) American Chemical Society.

Download Full Size | PDF

6. Conclusions and outlook

An ideal Si photoelectrode must meet the requirements of absorbing a large portion of the available solar flux and providing a sufficient driving force for the relevant PEC reactions. It must have efficient photoinduced carrier separation and transportation, fast reaction kinetics, and long-term stability, while ideally consisting of inexpensive, non-toxic and abundant elements. Based on the above objectives, the advances for the Si photoelectrodes in PEC water splitting cells over the past decades have been breath-taking. First, various surface textures on planar Si surface were introduced and analyzed to enhance the light absorption, and the MPs structure exhibits the best competitiveness due to the omnidirectional broadband light-trapping ability, relatively low surface recombination and low-cost. Moreover, this open surface textures allow the deposition of a conformal and thin layer of protective layer, so the stability of Si photocathode can be substantially improved. Second, various protective layers have been evaluated. Among them, TiO₂ has been wildly used for both Si photocathodes and photoanodes in a wide range of pH levels from acidic to alkaline, owing to its intrinsic chemical stability, excellent optical transparency and electrical conductivity. For Si photoanodes, the NiO_x can also be a promising candidate as a protective layer and an efficient OER catalyst in the high pH region. If one doesn’t consider the parasitic light absorption of protective layer, metal films may also be used for Si photoelectrodes due to their excellent conductivity. Ti film was reported for photocathodes in acid electrolyte, while Ni film can be used for both photocathodes and photoanodes in alkaline electrolyte. Third, noble metal such as Pt, Ru and Ir oxides are the best catalysts for Si photoelectrodes. However, to realize commercial PEC water splitting, alternatives of earth-abundant catalysts are most preferred. Recent studies have proved that metal-based alloys such as NiMo and NiFe on Si photoelectrodes were found to exhibit comparable and even higher PEC activity than the noble metal catalysts, in case of parasitic optical absorption losses of catalysts could be addressed. Last but not least, the integration of Si photoelectrode, protective layer and catalyst is critical for constructing a highly efficient and stable Si photoelectrode. The interface of the three parts and the surface of the integrated Si photoelectrode should be well designed to enable unhindered transmission of photo-generated carriers from Si to electrolyte. A rear p⁺ emitter on n-Si utilized as photocathode allows light illumination from the rear side of the device, which spatially and functionally decouples the optical absorption and the catalytic activity of the photocathode. Such strategies can be also applied to Si photoanode. Up to now, the highest reporting η is limited to 11.5% for Si photocathode, while that for Si photoanode is much lower (4.3%). Much efforts should be paid to boost the efficiency of the Si electrodes.

Although c-Si alone does not have enough photovoltage for unassisted solar water splitting, it is nearly ideal as the bottom light absorber in the PEC tandem device. To pair with the Si photoelectrode, various wide band gap semiconductor have been introduced to fabricate PEC tandem device, including BiVO₄ [144] and InGaN [145]. However, the design and performance of such tandem devices is limited due to the difficulty to pair the two semiconductors which should have the close saturated photocurrent and work near the maximum energy output. We proposed that the Si photoanode and cathode can be a pair, although the corresponding photovoltage is around 1.2 V which is not sufficient to drive unassisted solar water splitting. However, this can be solved through coupling external Si photovoltaic cells. Such mode takes both the advantages of the PV/EC and PEC modes, and we believe that high η_STH above 10% can be achieved.