# Spontaneous solar water splitting with decoupling of light absorption and electrocatalysis using silicon back-buried junction

Aug 6, 2020

### Design of the BBJ-PEC cell

Decoupling the optical and catalytic interfaces in PEC water splitting has recently been explored with different approaches18,24,25. Partial decoupling of light absorption and electrocatalysis have also been demonstrated by spatially selective deposition of electrocatalysts on the surface of nanostructures with high aspect ratios18,25, but such designs are complex in nature and challenging to integrate into conventional PEC devices. To address these issues, we adopted a BBJ device design, which was recently demonstrated to achieve a record solar-to-current conversion efficiency in a photovoltaic (PV) cell26. Though the BBJ design has made tremendous progress for PV applications, employing such cells for PEC water splitting is not straightforward.

Figure 1a displays a schematic diagram of a unit BBJ-PEC cell (see “Methods” for complete fabrication details). Specifically, the front side of the BBJ device is optimized for light harvesting without any catalyst shadowing effect, whereas the back side of the device plays the vital role of electrocatalysis, enabling the comprehensive light-harvesting surface to be exposed on the front side completely (Fig. 1b).

Enhancement of the optical characteristics is realized by texturing the Si surface with micropyramidal structures (top panel of Fig. 1c), as well as applying an AR layer of 60 nm SiNx (bottom panel of Fig. 1c) after the deposition of passivation layer to both the front and back surfaces of the cell, ensuring broadband omnidirectional light-harvesting with minimal reflection. The overall contribution of photon management scheme can be observed from the light-harvesting capability of the device, as demonstrated in the surface absorbance spectra (Fig. 1d), measured by ultraviolet–visible spectrophotometry in the air. The corresponding equation is shown below:

$$A_{mathrm{air}}left( lambda right) = 100{mathrm{% }} – R_{mathrm{air}}left( lambda right) – T_{mathrm{air}}left( lambda right),$$

(1)

in which Aair(λ), Rair(λ), and Tair(λ) are the light absorbance, reflectance, and transmittance of the BBJ-PEC cell, respectively. The absorbance spectra of the BBJ-PEC cell demonstrates over 95% absorption of the solar spectrum from 300 to 1100 nm, which has not been realized using a conventional PEC design to date.

Photo-excited carrier recombination is another significant loss factor limiting PEC efficiency, and hence an optimized electrical design is vital for avoiding excessive recombination and improving the minority carrier collection (i.e., holes in the n-type Si used in this study). To achieve this goal, we fabricated the BBJ by doping boron (B) and phosphorous (P) via local ion-implantation in an interdigitated dopant pattern (Fig. 1e) on the back (catalysis) surface of the device to fabricate a p+-Si emitter (hole collection bar with ~400 µm wide) and a n+-Si back surface field (BSF acts as an electron collection bar with ~200 µm wide) to improve minority carrier collection. In this manner, we were able to produce a cascade-type band structure to enhance the separation/transport efficiency of photo-excited carriers in the photoelectrode26. Additionally, a thin layer of implanted P (dopant thickness of 70 nm) was formed as the front-surface field (FSF) (shown in Fig. 1a) of the entire micropyramidal surface (where the recombination velocity is high) to reflect the holes down to the hole collector at the back of the PEC cell, thus reducing the surface recombination. After ion-implantation, we deposited a 40-nm SiO2 passivation layer before the AR coating to saturate dangling bonds on both front and back surfaces. Supplementary Fig. 1 shows the cross-sessional images of SiNx/SiO2/Si structure obtained by scanning transmission electron microscopy (STEM) with energy dispersive X-ray spectroscopy/electron energy loss spectroscopy (EDX/EELS) mapping, illustrating the intact passivation and AR layers.

We then deposited the interdigitated metal contacts/catalysts (Pt on n+-Si for HER and Ni on p+-Si for OER) on the catalysis reaction surface at the back side of the cell, which aligns with the interdigitated dopant pattern of the BBJ to collect the photo-excited electrons and holes, thus facilitating the PEC water splitting. We inserted all the contact grids/catalysts at the back of the PEC device for two reasons. First, it can maximize the light absorption by eliminating shadowing losses caused by the catalyst/metal contacts (i.e., the optical design). Second, it brings the catalysts in close contact with the buried junction located at the catalysis reaction surface of the PEC cell in order to efficiently extract the photo-excited carriers simultaneously for the PEC water splitting (i.e., the electrical/catalyst design). The overall electrical management scheme aims to reduce contact resistance and carrier recombination, thus facilitating HER and OER at the back (catalysis reaction) surface of BBJ-PEC cells. Moreover, inserting uniform interdigitated junction/metal grid contacts (Fig. 1f) on the entire catalysis surface shortens the path of the photo-excited carriers to the contacts by creating more junction areas, allowing the charge carriers to be efficiently separated, transported, and collected by the back grids. Meanwhile, the metal grids (up to 70% coverage) also serve as back mirror for internal photon reflection to promote the long-wavelength absorption by the photoelectrodes, as shown in Fig. 1b.

### Photoelectrochemical characterization

Before characterizing the PEC water-splitting behavior of our design, we measured the PV properties of the BBJ cell under one sun AM 1.5 G illumination in the air (Fig. 1g). The advantages obtained from the BBJ design were evidenced by a short-circuit current density (JSC) of 40.98 mA cm−2, an open-circuit voltage (VOC) of 0.62 V, a fill-factor (FF) of 74.46%, and thus a PV conversion efficiency (Eff) of 18.92%. The external quantum efficiency (EQE) and current loss analysis (Supplementary Fig. 2) of the BBJ cell exhibits outstanding optical and electrical performance, with maximum absorption in the AM 1.5 G solar spectrum, as envisioned by integrating the EQE spectral response over the photon flux in the AM 1.5 G solar spectrum. The current loss analysis demonstrates the limited parasitic losses due to the adoption of the decoupled light harvesting and electrocatalysis management scheme and the BBJ device architecture.

In addition to these advantages over the conventional vertical junction PEC cell structure, the BBJ design with a horizontal array of p–n junctions at entire back surface can benefit the PEC cell assembly by allowing the simultaneous interconnection of a series of three BBJ cells (3S-BBJ) to achieve a VOC up to 1.83 V (a detailed description of the fabrication process is provided in the “Methods” section), as shown in Fig. 1g (blue curve). Most importantly, the improved VOC of 1.83 V by the 3S-BBJ device is higher than the required thermodynamic potential for unassisted photoelectrolysis (~1.23 eV) plus the catalysis overpotential requirements, enabling wireless water splitting27.

In the BBJ device, both electron and hole collectors were patterned and aligned with respect to the local interdigitated p–n junctions for facilitating water splitting, which successfully improve carrier separation, collection, and electrocatalysis reaction. The local p- and n-junctions are connected to their respective carrier collection bars and centralized at opposite ends of the cell (as shown in Fig. 1e, the electron and hole collection bars are marked by green and purple dashed boxes, respectively). Pt and Ni were deposited by sputtering on at the opposite ends of the cell electrons and holes collection area as the HER and OER electrocatalysts, respectively. Figure 2a shows the schematic and band profile of the BBJ photocathode and photoanode. Supplementary Fig. 3 shows the fabrication schematics of the isolated packing process for BBJ photocathode and photoanode (see “Methods” and Supplementary Fig. 3 for fabrication details), which makes HER and OER work individually from the interdigital BBJs. In order to demonstrate the unassisted water-splitting capability, it is essential to optimize and validate the performance of a unit BBJ-PEC cell for HER and OER half-cell measurements. We analyzed the PEC performance of a unit BBJ photoelectrode using linear sweep voltammetry (LSV) with a standard three-electrode setup in a quartz cell under simulated AM 1.5 G one sun illumination (see “Methods”). The behavior of the photoelectrode was evaluated using standard performance metrics, with potentials calculated relative to the reversible hydrogen electrode (RHE)28. It should be noted that the performance of the BBJ-PEC photoelectrode evaluated by this method is only for half-cell metrics and not the overall water-splitting efficiency. We carried out the HER and OER measurements under different pH conditions for a unit BBJ-PEC cell, as it is important for a photoelectrode to perform under a wide range of pH conditions to achieve the balance between the pH requirements of two working electrodes1,29. The HER and OER performances in 1 M H2SO4 (pH ~0), 1 M Na2SO4 (pH ~7), and 1 M KOH (pH ~14) electrolytes are shown in Fig. 2b, c and summarized in Supplementary Table 1. Under the best performing conditions, a unit BBJ photoelectrode achieved a saturation current density for HER ((J_{{mathrm{H}}^ + /{mathrm{H}}_{2}})) of 40.51 mA cm−2 and for OER ((J_{{mathrm{O}}_2/{mathrm{H}}_2{mathrm{O}}})) of 28.68 mA cm−2 at the equilibrium water reduction and oxidation potential, respectively. It should be noted that the non-ideal feature in the onset region of the OER is due to the Ni oxidation reaction (more details on the non-ideal feature of Ni oxidation reaction can be found in the Supplementary Information and Supplementary Fig. 4). In the wide range pH conditions for HER and OER, the BBJ photoelectrode demonstrates the excellent performance, outperforming reported Si-based PEC cell for HER thus far (Supplementary Table 2).

The temperature and angle-of-incidence (AOI) of sunlight on a PEC cell will change through the course of a day and season. Hence it is crucial to analyze temperature- and AOI-dependent PEC performances for real-world deployment. In general, the energy bandgap (Eg) of the semiconductor decreases with temperature, which is determined by the Green’s law (more details on the temperature dependence of Eg and the corresponding factor can be found in the Supplementary Information and Supplementary Fig. 5)30,31,32, resulting in the variation of PEC performance under fluctuations of temperature. Similarly, the AOI significantly affects surface geometry-related reflection/scattering. To understand how these variables may affect the PEC performance, we measured the PEC performance of BBJ photocathode at various temperatures (293–323 K) under AM 1.5 G illumination and varying the AOI from 90° to −90° (relative to a normal vector of the light-harvesting plane). The (J_{{mathrm{H}}^ + /{mathrm{H}}_2}) and |VOS − E0| (where the VOS is the onset potential measured at the current density of 1 mA cm−2, E0 is the equilibrium water reduction potential) results are shown in Fig. 2d, e and summarized in the Supplementary Table 3.

The increasing temperature induces bandgap shrinkage in semiconductors and promotes more thermally excited carriers33. Therefore the semiconductor responds to extra absorption in longer wavelength regions, leading to a slightly increased (J_{{mathrm{H}}^ + /{mathrm{H}}_2}) with temperature. Meanwhile, the increase in temperature can cause the reduction of VOC and |VOS − E0| (more detailed discussion is provided in the Supplementary Information). Supplementary Fig. 6 shows the change in the PEC performances as a function of temperature and the corresponding results are summarized in Supplementary Table 4. On the other hand, maximum light absorption should be warranted to maximize the PEC performance. As discussed earlier, the light-harvesting characteristics were optimized and as a result, a broad range of light acceptance was demonstrated with the degradation of <14% in (J_{{mathrm{H}}^ + /{mathrm{H}}_2}) over an AOI of 60° (−30° to 30°). Furthermore, over a wide AOI of 180° (−90° to 90°), |VOS − E0| degradation is <8%, which has not previously been achieved. The broad range AOI of the BBJ photoelectrode is due to the dedicated decoupling scheme with pyramidal surface and AR layer.

### Comprehensive analysis of current losses

To gain more insight into the spatial and functional decoupling design of the photoelectrode’s light harvesting and electrocatalysis capabilities, we analyzed the overall solar-to-hydrogen conversion from end-to-end photon-to-fuel pathways and the associated losses by employing theoretical current loss analysis of a unit BBJ-PEC device for HER at pH 0.

There are three fundamental steps of the PEC water splitting for the current loss analysis. First, photo-excitation generates the charge carriers. Second, photo-excited charge carriers are separated and transport to the catalysis reaction surfaces. Third, the charge carriers drive catalytic water reduction or oxidation at the surfaces. In the comprehensive analysis, the current losses can be grouped into the optical current loss (JOpt) from the first step, and the internal current loss (JInt) from the last two steps. Furthermore, the JOpt is separated into two components: surface reflection loss (JS) and transmission loss (JT). For the sake of clarity, the loss sources of JS and JT are depicted in Fig. 3a by the area colored with orange and yellow, respectively. The corresponding formulas are expressed as below:

$$J_{mathrm{S}} = frac{q}{{hc}}int_{300}^{1100} lambda cdot {Phi}(lambda ) cdot R_{mathrm{liq}}left( lambda right) cdot {mathrm{d}}lambda,$$

(2)

$$J_{mathrm{T}} = frac{q}{{hc}}int_{700}^{1100} lambda cdot {Phi}(lambda ) cdot T_{mathrm{liq}}left( lambda right) cdot {mathrm{d}}lambda,$$

(3)

in which q is the charge of an electron, h is Planck’s constant, c is the speed of light in vacuum, λ is the photon wavelength, Φ(λ) is the AM 1.5 G solar spectrum (300 to 1100 nm), and Rliq(λ) and Tliq(λ) are the reflectance and transmittance of the BBJ photoelectrode measured in liquid electrolyte, respectively (as shown in Supplementary Fig. 7). The JS is due to back scattered photons at the surface of the PEC cell and in the electrolyte, which was calculated to be 5.44 mA cm−2 according to Eq. (2). As compared to the literature, the minimal JS loss that occurred herein can be attributed to the unique optical and catalytic interface decoupling scheme employed, enabling the ideal optimization of the light harvesting component, which cannot be achieved using the conventional PEC photoelectrode design. With Eq. (3), we calculated the JT by optical transmission through the body and the interdigitated electrode gaps of the BBJ device to be as low as 0.67 mA cm−2, which can be attributed to the long-wavelength photons (700–1100 nm) with higher transmission than the short wavelength region. The low JT value indicates the reflective electrode/catalyst at the back of the photoelectrode (up to 70% coverage) can effectively serve as a mirror to reflect unabsorbed light back to the device.

To further analyze the JInt, we carried out light absorbance of a unit BBJ cell measured in liquid electrolyte (Aliq(λ)) and the incident photon-to-current efficiency (IPCE) of the half-cell, which can be written as:

$$A_{{mathrm{liq}}}left( lambda right) = left( {1 – R_{{mathrm{liq}}}left( lambda right) – T_{{mathrm{liq}}}left( lambda right)} right) times 100{mathrm{% }}.$$

(4)

$$begin{array}{*{20}{l}} {{mathrm{IPCE}}left( lambda right)} hfill & = hfill & {frac{{{mathrm{Number}},{mathrm{of}},{mathrm{collected}},{mathrm{carriers}}}}{{{mathrm{Number}},{mathrm{of}},{mathrm{incident}},{mathrm{photons}},{mathrm{shining}},{mathrm{on}},{mathrm{the}},{mathrm{cells}}}}} hfill \ {} hfill & = hfill & {frac{{1240 times J_{{mathrm{H}}^ + /{mathrm{H}}_2}left( {{mathrm{mA}}{/}{mathrm{cm}}^2} right)}}{{lambda left( {{mathrm{nm}}} right) times P_{mathrm{in}}left( {{mathrm{mW}}{/}{mathrm{cm}}^2} right)}} times 100% } hfill end{array}.$$

(5)

The IPCE describes the maximum possible efficiency with which incoming photons can produce hydrogen from water, with an implicit assumption that all electrons are used for the HER (and holes for the OER instead of other byproducts or corrosion, assuming that the Faradaic efficiency is 100%). Therefore, the JInt can be calculated by the difference between Aliq(λ) and IPCE(λ) as shown below:

$$J_{{mathrm{Int}}} = J_{mathrm{E}} + J_{mathrm{C}} = frac{q}{{hc}}int_{300}^{1100} lambda cdot {Phi}left( lambda right) cdot left( {A_{{mathrm{liq}}}left( lambda right) – {mathrm{IPCE}}left( lambda right)} right) cdot {mathrm{d}}lambda.$$

(6)

It should be noted that the JInt can be distributed into two parts: electrical loss (JE) and catalytic loss (JC), which can be understood by the absorbed photon’s failure to generate photon-excited carriers and the photon-excited carriers recombining before the catalysis reaction, resulting in JE, and photon-excited carriers accommodating at the catalyst–electrolyte interface but failing to participate in the surface catalysis reaction, resulting in JC. Based on Eq. (6), we calculated JInt to be 3.43 mA cm−2.

To decouple JE and JC in order to provide an overall dynamic picture of the photo-excited carriers in the bulk of the photoelectrode and in the catalyst/electrolyte interface, we compared the internal quantum efficiency (IQE, measured in air without electrolyte to understand the inherent EQE without considering the catalytic capability in the electrolyte) and the absorbed photon-to-current efficiency (APCE), as shown in Fig. 3b. The IQE is defined as the number of photo-generated carriers contributing to the photocurrent divided by the number of absorbed photons in the materials measured in air, while the APCE is defined by the ratio of the number of photo-generated carriers contributing to the (J_{{mathrm{H}}^ + /{mathrm{H}}_2})for HER to the number of absorbed photons by the BBJ half-cell measured in the electrolyte:

$$begin{array}{*{20}{l}} {{mathrm{QE}}left( lambda right)} hfill & = hfill & {frac{{{mathrm{Number}},{mathrm{of}},{mathrm{collected}},{mathrm{photo}} {hbox{-}} {mathrm{generated}},{mathrm{carriers}},{mathrm{for}},{mathrm{dry}},{mathrm{cell}}}}{{{mathrm{Number}},{mathrm{of}},{mathrm{absorbed}},{mathrm{photons}}}}} hfill \ {} hfill & = hfill & {frac{{{mathrm{EQE}}left( lambda right)}}{{A_{{mathrm{air}}}left( lambda right)}} times 100% } hfill end{array}.$$

(7)

$$begin{array}{*{20}{l}} {{mathrm{APCE}}left( lambda right)} hfill & = hfill & {frac{{{mathrm{Number}},{mathrm{of}},{mathrm{collected}},{mathrm{photo}} {hbox{-}} {mathrm{generated}},{mathrm{carriers}},{mathrm{for}},{mathrm{electrolysis}}}}{{{mathrm{Number}},{mathrm{of}},{mathrm{absorbed}},{mathrm{photons}}}}} hfill \ {} hfill & = hfill & {frac{{{mathrm{IPCE}}left( lambda right)}}{{A_{{mathrm{liq}}}left( lambda right)}} times 100 % } hfill end{array}.$$

(8)

The IQE determines the efficiency of charge separation and transport to the solid–liquid interface. The change in APCE (related to the efficiency of charge separation and transport to the solid–liquid interface and the efficiency of the catalysis reaction at the solid/electrolyte interface) could point to dynamic behavior of photo-excited carriers in both the bulk of general PEC cell and the catalyst–electrolyte interface, since bulk recombination and inactive catalytic reaction can consume the photo-generated carriers. Therefore, the difference in IQE and APCE can account for the JC arising from the imbalanced kinetics of the catalysis reaction, in which an electron or a hole transfers and accommodates at the catalyst–electrolyte interface but fails to participate in the surface catalysis reaction. Thus, JC can be obtained by integrating Eq. (9).

$$J_{mathrm{C}} = frac{q}{{hc}}int_{300}^{1100} lambda cdot {Phi}left( lambda right) cdot left( {{mathrm{IQE}}left( lambda right) – {mathrm{APCE}}left( lambda right)} right) cdot {mathrm{d}}lambda.$$

(9)

For the sake of clarity, the loss sources of JC are depicted in Fig. 3b by the red colored region and calculated to be 1.67 mA cm−2. According to Eqs. (6) and (9), a JE of 1.76 mA cm−2 can calculated by the difference between JInt and JC.

The result of current loss analysis is summarized in Fig. 3c. The optimized optical and electrical designs with minimal current losses achieved herein can be attributed to the spatial decoupling of light absorption and catalytic activity, and the excellent electrical design of the BBJ-PEC cell. Using the current loss analysis, we can physically monitor the dynamics of photo-excited carriers in the bulk of the photoelectrode and at the catalyst/electrolyte interface. It is worth noting that this comprehensive analysis of the current losses is based on the general semiconductor half-cell, which is extendable to any semiconductor photoelectrode.

### Unassisted solar water splitting

The potential requirement to drive wireless water electrolysis is ~1.7–1.8 V34. To implement wireless water-splitting, we fabricated monolithic devices by integrating three BBJ-PEC unit cells in series via tin soldering, in which p+-Si and n+-Si were bonded together (p+nn+-p+nn+-p+nn+) with the Al foil. The advantage of BBJ-PEC cells is that the junction can make series connection laterally on the same surface to remove the shadowing effect. Hence, the maximized light harvesting was obtained after series the connection, which cannot be achieved by conventional vertical junction devices. The schematic and the corresponding band profile for the fully integrated 3S-BBJ-PEC cell is represented in Fig. 4a (see “Methods” for fabrication details; the photograph and fabrication schematic of the isolated packing process for 3S-BBJ devices can be found in Supplementary Fig. 8). The JV characteristic of the 3S-BBJ cell measured in air demonstrates a VOC of 1.83 V (Fig. 1g), which is well-above the voltage requirement for unassisted water-splitting with the given electrocatalysts.

Finally, we employed a 3S-BBJ-PEC cell in a photocathode configuration for unassisted water splitting in a two-electrode measurement setup (see “Methods”) by connecting the 3S-BBJ-PEC cell to the dark anode (Pt). The PEC reaction was carried out under 1 M H2SO4 with pH ~0. Bubble formation was visible immediately after illuminating the device under one sun AM 1.5 G simulator, demonstrating spontaneous water splitting without the application of an external bias. Figure 4b and Supplementary Table 5 show the performance of a wireless 3S-BBJ-PEC cell for the HER in 1 M H2SO4 (red line), which shows the unassisted current density of 12.60 mA cm−2. We also demonstrate the time-resolved saturated (J_{{mathrm{H}}^ + /{mathrm{H}}_2}) and |VOS − E0| with chopped illumination in Fig. 4c, d, showing a stable STH conversion over four cycles. As expected, the results show that the 3S-BBJ cell has ~3 times the |VOS −  E0| and ~1/3 the (J_{{mathrm{H}}^ + /{mathrm{H}}_2}) compared to a BBJ unit cell configuration without a series connection. The STH efficiency for the 3S-BBJ photoelectrode without applying an external bias (Vapp = 0 V) can be calculated using Eq. (10):

$${mathrm{STH}}(%) = frac{{J_{{mathrm{H}}^ + /{mathrm{H}}_2} times (1.23 – V_{{mathrm{app}}})}}{{P_{{mathrm{in}}}}},$$

(10)

in which Pin is the incident optical power shining on the cell. With the unassisted (J_{{mathrm{H}}^ + /{mathrm{H}}_2}) of 12.60 mA cm−2, an STH efficiency of 15.62% was achieved, which is the highest efficiency ever reported for Si-based unassisted PEC water splitting. In order to simulate real-world solar hydrogen production, we employed water obtained from the Red Sea as an electrolyte (~pH 8.1), and in outdoor midday (12:00 p.m.) illumination conditions to drive a series connected 3S-BBJ cell, yielding an STH efficiency of 11.33% (green line in Fig. 4b and Supplementary Table 5).

### Unassisted water-splitting stability

In general, it is well known that technologically matured semiconductors such as Si-based photoelectrodes undergo severe corrosion/oxidation during the water-splitting process3335. Despite the demonstration of high half-cell efficiencies and improved stabilities under various electrolytic conditions, the PEC stability during the overall water-splitting is still vital and challenging. In literature, the stability issues has been partially addressed by depositing compact and continuous metal oxide overlayers such as TiO2 using atomic layer deposition (ALD)34,35. The deposition of protective metal-oxides has been shown to be successful under nominal current densities to extend the device lifetime, and hence we adopted the surface protection of ALD-TiO2 for unassisted water-splitting. To improve the PEC stability, we carried out the ALD deposition of TiOx (~140 nm) surface protection layers directly on the BBJ-contact pads before the sputter deposition of Pt (5 nm) for HER purpose. More details on the ALD-TiOx is provided in the “Methods” section. The resultant 3S-BBJ-PEC cell under unassisted water-splitting conditions demonstrates 40 h of stability (Supplementary Fig. 9), then the current density decreases due to the corrosion.

Gases evolved from the reactor during the stability measurements were collected with an airtight syringe and analyzed immediately using gas chromatography (GC) to confirm the as-claimed efficiencies (see “Methods”). The top panel of Fig. 4e shows the current density-time ((J_{{mathrm{H}}^ + /{mathrm{H}}_2}) − T) curve measured during the gas analysis and the middle panel of Fig. 4e displays the thermal conductivity detector (TCD) signal from the GC measurements. The quantity of generated H2 as a function of time can be calculated by two different ways: One is integrating the (J_{{mathrm{H}}^ + /{mathrm{H}}_2}) by the time (red line in the bottom panel of Fig. 4e) and the other is calculating the signals from TCD (blue dots in the bottom panel of Fig. 4e, and the details are shown in Supplementary Fig. 10). These two calculated results demonstrate consistent increase in H2 amount in the reactor headspace during the overall water-splitting reactions, which shows that for 20 min of continued operation, the gas evolution density was ~80 μg cm−2 under one sun illumination. More importantly, the quantity of gases obtained from the GC measurements matches well with the quantity of integrated electrons-to-H2 carriers, with the Faradaic efficiency of >90% for the initial period of 15 min.

Since the photoelectrode was corroded after operation for 40 h, further transmission electron microscopy (TEM) analysis was carried out to understand the failure mechanism on the surface protection of TiOx/Pt compact layer. The TEM result shows (Supplementary Fig. 11) the photocorrosion and etching of TiOx/Pt layer that leaded to the cracks in the compact ALD layers and failed to protect the underlying surface. The severe corrosion of 3S-BBJ-PEC could be due to the high current density (~280–300 mA) at the electrodes/electrolytes interface (~1 cm2). It should be noted that the total current from the 3S-BBJ-PEC should pass through the limited contact/electrocatalyst in this configuration as shown in Fig. 1a. To address the issue, we employed Pt-sputtered (5 nm)/Ti foil (0.01 mm) on the electron collection bar (see “Methods” for fabrication details) before immersing the 3S-BBJ-PEC cell in the electrolyte (1 M H2SO4). This idea is similar to the recent surface protection strategy demonstrated by Varadhan et al.36, in which a thin Ni-foil (0.125 mm) directly attached to the electrolysis surface acts as both the OER catalyst and the surface protection layer to extend the device lifetime. Herein with Pt/Ti treatment, the resultant 3S-BBJ-PEC cell under two-electrode configuration with 1 M H2SO4 as the electrolyte exhibits the initial unassisted current density ((J_{{mathrm{H}}^ + /{mathrm{H}}_2})) of 10.5 mA cm−2 with the STH efficiency of ~13% (in Supplementary Fig. 12). More importantly, the 3S-BBJ photocathode becomes stabilized with <10% drop in (J_{{mathrm{H}}^ + /{mathrm{H}}_2}) over the period of 100 h, demonstrating highly efficient and stable PEC water-splitting reaction.