A Coating Strategy to Achieve Effective Local Charge Separation for Photocatalytic Coevolution

A coating strategy to achieve effective local charge separation for photocatalytic coevolution

Tianshuo Zhaoa,b, Rito Yanagia,b, Yijie Xua, Yulian Hea,b, Yuqi Songa, Meiqi Yanga,b, and Shu Hua,b,1

aDepartment of Chemical and Environmental Engineering, Yale University, New Haven, CT 06520; and bEnergy Sciences Institute, Yale University, West Haven, CT 06516

Edited by Alexis T. Bell, University of California, Berkeley, CA, and approved January 12, 2021 (received for review November 19, 2020) Semiconductors of narrow bandgaps and high quantum efficiency coevolution. Coevolution refers to the concurrent generation of have not been broadly utilized for photocatalytic coevolution of two products from the reduction and oxidation reactions in the H2 and O2 via water splitting. One prominent issue is to develop same photocatalyst sample space (21). These reactions can be effective protection strategies, which not only mitigate photocor- thermodynamically uphill. Only when their reverse reactions are rosion in an aqueous environment but also facilitate charge sepa- suppressed at the respective catalytic sites can the two products ration. Achieving local charge separation is especially challenging coexist rather than chemically recombine. For water-splitting re- when these reductive and oxidative sites are placed only nanome- actions, those reductive and oxidative sites at the photocatalyst– ters apart compared to two macroscopically separated electrodes liquid interfaces coevolve H2 and O2, or produce H2 and oxidize in a photoelectrochemical cell. Additionally, the driving force of redox mediators, or reduce redox mediators and evolve O ,con- charge separation, namely the energetic difference in the barrier 2 currently. Therefore, a coating for photocatalysts should simulta- heights across the two type of sites, is small. Herein, we used conformal coatings attached by nanoscale cocatalysts to transform neously allow reductive and oxidative charge transfer, from the two classes of tunable bandgap semiconductors, i.e., CdS and photocatalyst through the coating to the liquid; whereas the conventional coatings only enable single-direction charge transfer, GaInP2, into stable and efficient photocatalysts. We used hydrogen evolution and redox-mediator oxidation for model study, and either reductive or oxidative. further constructed a two-compartment solar fuel generator that Recent publications imply that crystal facets, cocatalysts, or

separated stoichiometric H2 and O2 products. Distinct from the local doping can cause charges to separate locally at the solid- single charge-transfer direction reported for conventional protec- state photoabsorbers (6, 22–24). Different chemical states of ENGINEERING tive coatings, the coating herein allows for concurrent injection of bifunctional cocatalysts were also postulated to facilitate charge photoexcited electrons and holes through the coating. The ener- separation (25, 26). However, the energetics under operational getic difference between reductive and oxidative catalytic sites conditions have not been quantified. Therefore, a mechanism was regulated by selectivity and local kinetics. Accordingly, the leading to effective charge separation needs to be generalized. charge separation behavior was validated using numerical simula- Our design principle is to separate charges at spatially disparate tions. Following this design principle, the CdS/TiO2/Rh@CrOx pho- sites of photocatalysts. It should be different from molecular tocatalysts evolved H2 while oxidizing reversible polysulfide redox photosynthesis, where the excited states are long-lived to favor μ · −1· −2 mediators at a maximum rate of 90.6 mol h cm by stacking charge accumulation and catalysis over charge recombination three panels. Powered by a solar cell, the redox-mediated solar (27). In contrast, irreversible hole scavengers cannot be regen- water-splitting reactor regenerated the polysulfide repeatedly erated to sustain continuous PC, despite that their fast kinetics and achieved solar-to-hydrogen efficiency of 1.7%. are often leveraged for competing against photooxidation (28). Without the chemical bias provided by the irreversible photocatalytic synthesis | charge separation | coatings | corrosion protection | reactor engineering Significance articulate photocatalysis (PC) is a promising platform for Particulate photocatalysis is a promising approach to solar fuels Psunlight-driven hydrogen production or CO2 reduction at scale (1–5). Although particulate photocatalysts have attained production at scale. Herein, we present a general design by nearly 100% quantum efficiency recently (6), the state-of-the-art using conformal coatings and attaching nanoscale cocatalysts solar-to-hydrogen (STH) conversion efficiency for oxides (5, 7), to achieve local charge separation and, at the same time, to stabilize photocatalysts that are easily photocorroded other- oxynitrides (8, 9), and oxysulfides (10), remains around 1% (11). wise. With spatial charge separation, the nanometer-spaced An alternative direction is to utilize narrow and tunable bandgap reductive and oxidative surface sites can coevolve to produce photoabsorbers, such as those II–VI and III–V semiconductors. H and O , or to produce H and oxidize redox mediators, or to They are known for strong optical absorption, high quantum 2 2 2 produce O and reduce redox mediators. This work investigates efficiency, and reasonable carrier lifetime (12–14). For example, 2 – – the charge separation strategy for the semiconductor/coating/ cadmium sulfide (CdS), and gallium indium phosphide (GaInP2) cocatalyst structure both by tuning barrier height energetics photoabsorbers have bandgaps of 2.42 and 1.87 eV, respectively. and by building numerical models. The use of GaInP2 as the top absorber in a photoelectrochemical water-splitting device showed 19% STH efficiency (15, 16). CdS- Author contributions: T.Z. and S.H. designed research; T.Z., R.Y., Y.X., Y.H., Y.S., and M.Y. and GaInP2-based photocatalysts, if utilized in a two-photosystem performed research; T.Z., R.Y., Y.X., Y.H., Y.S., and M.Y. analyzed data; and T.Z., R.Y., and scheme or redox-mediated water-splitting reactor, promise a leap S.H. wrote the paper. in STH efficiency (1, 17). However, these semiconductors suffer The authors declare no competing interest. from poor photostability in water, and their few-hour stability is This article is a PNAS Direct Submission. much shorter than the thousands of hours durability shown by Published under the PNAS license. oxides (5). 1To whom correspondence may be addressed. Email: [email protected]. Although coating-protected narrow bandgap semiconductors This article contains supporting information online at https://www.pnas.org/lookup/suppl/ have been broadly studied in photoelectrochemical devices doi:10.1073/pnas.2023552118/-/DCSupplemental. (18–20), they have not been investigated for photocatalytic Published February 8, 2021.

PNAS 2021 Vol. 118 No. 7 e2023552118 https://doi.org/10.1073/pnas.2023552118 | 1of8 Downloaded by guest on September 27, 2021 scavengers, the charge separation in the presence of reversible can coexist due to their slow chemical recombination: They redox mediators is often not effective. transport away and eventually get separated. The charge sepa- Herein, we develop a coating approach that stabilizes photo- ration and recombination occur within semiconductor photocatalysts and, in the meantime, facilitates spatial charge sepa- absorbers on the order of picoseconds to microseconds, while the ration during light-induced coevolution. As illustrated in Fig. 1A, timescale of catalysis, such as H2 evolution, is in microseconds to the semiconductor/coating/cocatalyst interface is fabricated by milliseconds. The rate of charge transfer between the photo- coating the semiconductor film with thin protective coatings, catalyst and liquid is far outcompeted by the rate of electron- followed by sparsely loading metal nanoparticles as cocatalysts hole recombination at a single catalytic site. Therefore, the on the coating surface. This simple two-step procedure trans- spatial separation of photoexcited electrons and holes should forms narrow-bandgap semiconductors, such as CdS powder suppress carrier recombination. The charges are accumulated or films and epitaxial GaInP2 layers, into stable and efficient pho- trapped at spatially disparate surface states of the photocatalysts, tocatalysts: Coevolution of hydrogen and reversible redox me- where they reside for much longer than the recombination life- − diators (A/A ) occurs at the cocatalyst sites and bare coating time of the photoabsorber. Therefore, efficient charge separa- surfaces, respectively. Those two types of sites are alternating tion schemes should be established for those reductive and along the liquid interface. Such a general scheme does not rely oxidative sites that are only nanometers apart. Otherwise, severe on specific morphology or crystal facets of the photocatalyst. We charge recombination will lead to low or even no photocatalytic therefore fabricated CdS/TiO2/Rh@CrOx particulate panels, activity. −1 −2 which continuously evolved H2 at 90.6 μmol·h ·cm , and Based on our previous study (29), efficient charge separation showed an internal quantum yield (IQY) of 44.3% at 438 nm in a can be realized in photocatalysts through the spatially varying 2− 2− solution of reversible polysulfide/sulfide (Sn /S ) redox cou- barrier heights along the same photocatalyst–liquid interface. It ples. TiO2 coating-stabilized GaInP2 panels also performed H2 is hypothesized that the introduction of cocatalysts leads to −1 −2 2− evolution at a rate of 144.7 μmol·h ·cm with the same Sn / spatial charge separation. The barrier height energetics of the − S2 redox system, further demonstrating the general strategy. cocatalyst-decorated semiconductor surface were shown to be The mechanism of local charge separation is investigated by different from those of the bare semiconductor surface. Fur- combining numerical modeling and potentiostatic experiments. thermore, the barrier height energetics and local kinetics are A porous CrOx shell controlled the charge-transfer kinetics and mutually dependent, and can vary with local chemical environ- selectivity of the Rh cocatalyst core: The Rh@CrOx core@shell ments during operation (29). cocatalyst reduced the local barrier height energetics under op- First, we built a semiconductor/coating/cocatalyst model in- erational microenvironment. In addition, redox-mediated pho- terface to investigate a steady-state distribution of the flux of tocatalytic coevolution can shuttle charge carriers, so that one photoexcited electrons and holes during coevolution (Fig. 1B). can construct a safe, H2 and O2-separated reactor (2). The two-dimensional model shows the distribution both inside the photocatalyst and along the photocatalyst–liquid interface. A Results and Discussion semiconductor particle, e.g., CdS, of 1 μm in radius and with a Design for Efficient Charge Separation and Product Coevolution. To bandgap of 2.4 eV was conformally encapsulated by a 3-nm-thick achieve photocatalytic coevolution of two products, both elec- TiO2 protective layer, the surface of which was periodically trons and holes need to reach the semiconductor surface and covered with 100-nm-wide sites of Rh cocatalyst nanoparticles transfer to the liquid before their recombination. The products for H2 evolution (Fig. 2 A and B). The detailed construction of the numerical model can be seen in Materials and Methods, and the determination of specific physical parameters will be discussed in following sections. To reproduce the coevolution condition, a Schottky barrier of 0.11-V barrier height was set at the Rh sites, and 0.60-V barrier height for the bare TiO2 coating sites for redox mediator oxidation. These values were obtained by potentiostatic measurements, as revealed in the later sections. Under 1-sun illumination, the simulated electron potential energy exhibited a gradient within the photocatalyst (Fig. 1B), indicating steady-state electron-hole separation. The electron potentials were calculated to be 5.0 and 4.5 eV vs. vacuum at the redox oxidation sites and the hydrogen evolution reaction (HER) sites, respectively. This potential gradient drove the charge separation, which then reinforced cocatalyst photodeposition, both of which were consistent with the design principle for coevolving reactions. Therefore, this semiconductor/ coating/cocatalyst design is about creating spatially varying sites for local charge separation and should be general to narrow bandgap semiconductors without leveraging their energetics of specific facets or morphology (6, 30).

Fabrication of CdS/TiO2/Rh@CrOx Photocatalytic Panels. Next, we showcase this design principle by fabricating photocatalysts with commercial CdS powders, TiO2 coatings, and Rh cocatalysts for the coevolution of H2 and reversible polysulfide redox mediators. CdS particles of 1-μm averaged particle size were drop-casted Materials and Methods Fig. 1. Coating-stabilized photocatalysts for coevolution reactions. (A) onto a piece of frosted glass ( ). The panel was encapsulated by 3-nm TiO via atomic-layer deposition Schematics of the coating-stabilized photocatalytic panel for coevolving H2 2 − A and oxidizing reversible redox couple A/A .(B) Simulated potential energy (ALD) (Fig. 2 ). Then, the ALD TiO2 was decorated with of the photocatalyst CB edges. The black and red arrows represent the local Rh@CrOx core-shell nanoparticles via photodeposition to form hole and electron transport directions, respectively. ∼100-nm-wide islands of nanoparticulate cocatalyst at the

2of8 | PNAS Zhao et al. https://doi.org/10.1073/pnas.2023552118 A coating strategy to achieve effective local charge separation for photocatalytic coevolution Downloaded by guest on September 27, 2021 CV characterization indicates that the polysulfide reduction was suppressed on the Rh@CrOx core-shell cocatalysts. The CrOx shell possibly prevents the polysulfide diffusion to the Rh sites, an analogy to the previous report of utilizing CrOx shells to reduce oxygen reduction (35). Excess CrOx photodeposition, how- – ever, densified the shell and hindered the transport of H2 or OH products, thus detrimental to the HER performance (SI Appen- dix, Fig. S1). The H2-evolution rate increased with Na2S concentration and reached a saturated value after the concentration was higher than 100 mM (SI Appendix, Fig. S2A). This suggests that the rate- determining step was no longer the SOR at high redox concentration. Besides, the CdS/TiO2/Rh@CrOx panel can also evolve 3−/4− H2 and oxidize [Fe(CN)6] redox mediators in a 50 mM K4Fe(CN)6 aqueous solution at pH 7.0 (SI Appendix, Fig. S2B), showing the versatility of this coating strategy. The gas evolution rate can be further improved by tuning the coating energetics to make facile charge transfer to a number of redox mediators in future studies (36, 37).

Energetic Studies of Local Catalytic Sites. Through light-intensity dependent open-circuit potential (OCP) measurements of CdS particle-assembled electrodes (Materials and Methods) (29), we investigated the barrier height energetics of the two types of Fig. 2. Charge separation of the photocatalytic panels. (A) Scanning elec- surface sites, i.e., CdS/TiO2 and CdS/TiO2/Rh@CrOx in contact tron microscopy (SEM) of the CdS/TiO2/Rh@CrOx particulate panel on a glass substrate and (B) atomic force microscopy (AFM) nanoscale topography of with the redox-mediator solution. The sum of the total reduction

Rh@CrOx cocatalysts on the panel. (C)H2-evolution rates of the CdS panel of rates and the total oxidation rates is zero for photocatalysts; bare surface, with 3-nm TiO2 coating, with TiO2/Rh, and with TiO2/Rh@CrOx whereas, in photoelectrocatalysis, the electrons and holes

in 100 mM Na2S solutions (pH 13.5). Schematics in C show the corresponding recombine at the liquid interface, resulting in a net photo- ENGINEERING charge transport directions. The rate was an average of at least three sam- cathodic or photoanodic current. Therefore, an OCP condition ples for each condition. for a photocatalyst-assembled electrode resembles the coevolution. The ohmic back contact for the electrodes established be- TiO –liquid interface (Fig. 2B and SI Appendix, Note 2). tween Ti and n-CdS allows for probing of the averaged 2 Materials According to the design principle, photoexcited electrons and electrochemical potentials of photoexcited electrons ( and Methods Φ holes simultaneously accumulate at the active sites of Rh co- ). The local barrier height B for electron injection at the surface HER and SOR site, respectively, can be obtained catalysts and bare TiO2 surfaces, respectively, to drive HER and Φ = E − E E reversible sulfide oxidation reaction (SOR) (Eq. 1) (31) along by B F c, where F is the Fermi level at equilibrium represented by OCP measured in the dark (OCP ) and Ec is the photocatalyst–liquid interface. Regeneration of the redox dark 2− the conduction band (CB) edge of the semiconductor. More mediator (Eq. 1), i.e., reducing polysulfide (Sn ) back to sulfide SI Appendix 2− detailed calculation is provided in , Note 1. (S ), is necessary to sustain overall water splitting, but the Fig. 3A shows that as the light intensity increased, the OCPs of charge separation scheme should be as efficient for their re- the CdS/TiO2 electrode measured in a 10 mM Na2S solution (pH versible kinetics as the hole-driven water oxidation. In this work, 12) became increasingly negative and reached a steady-state we mainly focus on the reversible SOR in alkaline pH, but it can be value of 0.34 V vs. RHE, denoted as OCPlight. The OCPlight 3− − 3−/4− extended to other redox mediators, e.g., IO /I ,[Fe(CN)6] , shifted to −0.09 V vs. RHE after loading the CdS/TiO2 electrode 3+/2+ – and Fe in various pH (32 34): with Rh@CrOx (Fig. 3B), suggesting sufficiently reducing po- + tentials for the Rh@CrOx sites to drive H reduction. The 2− + 2− nS + (2n − 2)h ⇌ Sn (2 ≤ n ≤ 8). [1] OCPlight of the CdS/TiO2/RhCrOx electrode was consistent with the CB edge position of the CdS/TiO2 sites, which was deter- Consistent with our design, the introduction of coatings and mined from the Mott–Schottky analysis independently (SI Ap- cocatalysts improved photocatalytic activity of the CdS panel. pendix, Fig. S3A and Note 1). Therefore, both sites are able to The commercial CdS powder film showed a 1-atm H2-evolution coevolve HER and SOR, according to their band edge positions −1 −2 SI Appendix C rate of only 2.2 μmol·h ·cm in a 100 mM Na2S solution ( , Fig. S3 ). (Fig. 2C). The rate increased by a factor of 1.3 and 5.8, after Purging the redox solution with H2 resembles the local coating 3 nm of TiO2 only and after loading 1 wt% Rh on the chemical environment during H2-evolving PC. In this case, – TiO2, respectively. Upon growth of CrOx shells on individual Rh multiple charge transfer pathways coexist at the photocatalyst nanoparticles, a significant boost for the activity to 50.4 liquid interface, e.g., a two-redox liquid junction (29). When the − − μmol·h 1·cm 2 was achieved, which is equivalent to a photocur- cocatalyst could either reduce the polysulfide redox mediator or −2 + rent density of 2.7 mA·cm . Introducing CrOx on the cocatalyst reduce H , the OCP is the potential at the detailed balance of forced the selectivity toward HER and away from polysulfide the kinetic rates for all the forward and backward charge-transfer rereduction. Cyclic voltammetry (CV) on a bare FTO/Rh elec- pathways. Unlike the almost unaltered OCPs of the CdS/TiO2 trode in the polysulfide solution shows the onset potential of the electrode after H2 purging (Fig. 3A), the OCPdark of CdS/TiO2/ reductive current was 0.21 V vs. RHE and a redox peak at Rh@CrOx approached 0 V vs. RHE, while the OCPlight around 0 V vs. RHE (SI Appendix, Fig. S1). These suggest that remained at −0.10 V vs. RHE (Fig. 3B). In other words, the bare Rh reduced polysulfide undesirably in addition to HER. potential of the Rh cocatalysts (inside CrOx shells) shifted up- 0 + Nevertheless, after CrOx was deposited, the redox reduction wards to align with E (H /H2). Meanwhile, the flat-band po- peak was reduced, while the HER current was mostly unaffected tential of CdS at the Rh@CrOx site measured by the OCPlight and dominant under applied potentials below 0 vs. RHE. This was consistent with the CdS CB band edge position, which stayed

Zhao et al. PNAS | 3of8 A coating strategy to achieve effective local charge separation for photocatalytic https://doi.org/10.1073/pnas.2023552118 coevolution Downloaded by guest on September 27, 2021 Appendix, Fig. S6A). Such high selectivity against O2 reduction 2− or Sn reduction is consistent with the unchanged OCPlight at approximately −0.10 V vs. RHE regardless of the gas environment (SI Appendix, Fig. S6B). The continuous H2 production eventually replaces the headspace with 1-atm H2. Essentially, the coherent trends in activity and energetic studies validated our design principle and highlighted the dual functionality of the Rh@CrOx cocatalysts, i.e., 1) improving charge-transfer kinetics and 2) creating barrier height asymmetries to facilitate local charge separation.

Modeling of Steady-State Charge Separation. Based on the result of energetic studies and geometric parameters estimated from Fig. 2 A and B and SI Appendix, Fig. S7, we refined the model to simulate the charge transport and dynamics of the CdS/TiO2/ Rh@CrOx photocatalysts. The materials properties used are discussed in SI Appendix, Note 2 and listed in SI Appendix, Table S2. We hypothesized that the HER and SOR sites of 100 nm wide each distributed periodically on the photocatalyst surface (17). Because the distance between adjacent Rh@CrOx sites was found to mostly vary from 300 to 350 nm, with some occasionally within 100 to 200 nm apart (SI Appendix, Fig. S7B), we chose a periodic spacing of 300 nm as a representative model. The effect Fig. 3. Local energetics of the photocatalytic panels. Light intensity- of cocatalyst sizes and spacing on charge separation desires

dependent OCPs of (A) CdS/TiO2 and (B) CdS/TiO2/Rh@CrOx particle- further investigations, not discussed in this paper. The charge- assembled photoelectrode in a 10 mM Na2S aqueous solution (pH 12.0) transfer rates for electrons and holes were comparable at both 0 = without (black dots) and with (red dots) H2 purging. E (Ag/AgCl) 0.20 V vs. HER and SOR sites, corresponding to nonselective contacts. NHE. (Inset) Illustration of the (A) bare CdS/TiO2 site and (B) CdS/TiO2/ Therefore, the same electron and hole surface recombination · −2 −1 Rh@CrOx site. The OCP measured at 0 mW cm represents the dark condi- velocities (SRVs) of 10 cm·s were modeled for both sites (40). tion, denoted as OCPdark.(C) Band diagram of the CdS photocatalyst be- The use of local carrier-selective contacts and single-atom or tween the SOR (Left) and HER (Right) sites under illumination along the clustered cocatalysts could also induce charge separation (41, liquid interface, but not into the semiconductor. Electron and hole quasi- 42), but is beyond the scope of this study. Fermi levels are shown by the red and black dashed lines, respectively. − − A uniform carrier generation rate of ∼1 × 1020 cm 3·s 1 was calculated for the CdS particle thin film under 1-sun illumination Materials and Methods − SI Appendix ( ). From the HER and SOR site at the at 0.10 V vs. RHE ( , Note 1). The barrier height for liquid interface into the CdS bulk, ∼200 nm or 20% of the ab- electron injection at those CdS/TiO2/Rh@CrOx sites was con- sorber volume was under an electric field (SI Appendix, Fig. sequently reduced to 0.11 V, much smaller than the CdS/bare A–C SI Appendix S8 ). Within this volume, either the electron or hole current TiO2 barrier height of 0.60 V ( , Note 1). It is rea- dominated the total currents (SI Appendix, Fig. S8D). Away from sonable that the Rh nanoparticles did not form buried Schottky the surface, electron diffusion current became comparable to junctions with TiO2 coatings, because they were photodeposited and eventually higher than the electron drift current (SI Ap- over a hydrous or electrolyte-wetted TiO2 surface (29, 38). pendix, Fig. S8E). This spatial distribution indicated a steady- Therefore, the CdS/TiO2/Rh@CrOx photocatalysts achieved a state electron accumulation at the Rh@CrOx site (SI Appendix, barrier height difference, or asymmetric barrier energetics (39). Fig. S8F). Essentially, this diffusion-controlled charge separation C As illustrated in Fig. 3 , the barrier height difference induced a is distinctive to the drift-dominated regime typically observed in gradient of electron and hole electrochemical potentials, or solid-state energy conversion (22): During coevolution, efficient quasi-Fermi levels, between the neighboring HER and SOR charge separation is possible even with a 0.5-eV energy barrier sites. Consequently, an electric field was established, which made difference between the two catalytic sites, an asymmetry much photogenerated electrons transport to the Rh@CrOx sites for H2 less than the 2.4-eV bandgap of the photoabsorber. evolution and made holes accumulate at the bare TiO2 surface to Too thick TiO coatings may reduce the hole injection rates, 2− 2− 2 oxidize the Sn /S redox mediator. The in-depth calculation for and thus diminish charge separation efficiency and photo- SI Appendix the band diagram is explained in , Note 1 and Fig. S4, catalytic activity. Experimentally, we observed that the H2 ac- and the various energetic parameters are summarized in SI Ap- tivity of the CdS photocatalyst decreased dramatically when the pendix , Table S1. TiO2 thickness reached 10 nm (SI Appendix, Fig. S9). This Only decorating Rh cocatalysts but without the CrOx shell did thickness-dependent behavior should not always occur but was not achieve sufficient barrier height difference. The OCPlight of attributed to the energetic mismatch during hole transport from CdS/TiO2/Rh electrodes, i.e., the Rh potential, was not raised to CdS to the redox mediator. Through >3-nm-thick TiO , the hole SI Appendix 2 more negative potentials even under H2 purging ( , hopping transport was slow compared to the direct tunneling Fig. S5). Due to the poor selectivity against polysulfide reduction transport (43), because the Ti3+-defect band for hole transport as discussed above, the uncoated Rh site could also transfer does not match with the energy level of CdS valence bands. In 2− 2− electrons to Sn /S , thus competing with the desirable pathway the modeling, different charge injection currents can be simu- + of H reduction in a two-redox liquid junction. The resulting lated by modulating the SRV at the 10-nm TiO2. The modeling 0.45-V barrier height of the Rh sites became comparable to that outcome is consistent with the measured activity: Reducing the of the CdS/TiO2 sites. Therefore, this lack of energetic asym- electron and hole SRVs by 10 times decreased the photocurrent, metry adversely affected the charge separation efficiency for the or the H2-evolution rate, to a third of the 3-nm TiO2-coated CdS/TiO2/Rh panel without the selective CrOx shell (Fig. 2C). photocatalyst (SI Appendix, Fig. S9). Throughout the thickness Besides, the selective Rh@CrOx cocatalysts ensure the high ac- variation study, the ratios of electron and hole SRVs were kept 8 tivities regardless in 1-atm Ar, air, or H2 atmospheres (SI constant. The calculated effective resistivity of ∼10 Ω for the

4of8 | PNAS Zhao et al. https://doi.org/10.1073/pnas.2023552118 A coating strategy to achieve effective local charge separation for photocatalytic coevolution Downloaded by guest on September 27, 2021 10-nm TiO2 layer also agrees with the literature report (43). The Redox-Mediated Photocatalytic Reactors. We constructed a redox- − low SRVs led to significant hole accumulation up to 1021 cm 3 at mediated solar-fuel generator as proof-of-concept, which allows the 10-nm TiO2 surface (SI Appendix, Fig. S10). Therefore, in- for H2 and O2 evolution at two separated reactor chambers. As sufficient hole transport limits the photocatalytic activity, which illustrated in Fig. 4A, a PC reactor was integrated with a is irrespective to the CdS carrier lifetime or radiative efficiency. photovoltaic-driven electrolysis (PVE) device. The reversible 2− 2− With sufficient charge-transfer rates, the barrier height asym- Sn /S redox mediators shuttle photogenerated hole charges metry emerged to determine the spatial separation and between the PC reactor and PVE device. An amorphous Si (a-Si) accumulation. solar cell of 1.7- to 2.0-eV bandgaps and the same area as the photocatalytic panel absorbs the light transmitted through the Improving Quantum Yields of CdS Photocatalysts. A single CdS CdS panel, to power the PVE device (45). The current–voltage characteristics of the solar cell are shown in SI Appendix, Fig. panel with 3-nm TiO2, 1 wt% of Rh, and 1 wt% Cr loaded, − −1 −2 2 produced H at a rate of 50.4 μmol·h ·cm in a 100 mM Na S S14. The membrane-electrode electrolysis device reduces Sn 2 2 2− solution (pH 13.5) (Materials and Methods). We measured an ions to regenerate S at the cathode and produces O2 and apparent quantum yield (AQY) of 24.6% at 438 nm for the same protons by oxidizing pure water at the anode. A cation exchange membrane (CEM) prevents the redox mediator and O2 cross- panel in the 100 mM Na2S solution. Considering 55.5% of the light absorbed at that wavelength (SI Appendix, Fig. S11), the over, while replenishing the redox solution with the protons needed for H2 evolution. IQY of the panel was calculated to be 44.3%. Stacking multiple B panel devices can absorb the incident light fully. With a three- Fig. 4 shows a separated production of stoichiometric H2 and O2 by alternating the PC and the PVE processes. The amount of panel stacking, the H2 evolution rate increased to 90.6 − − H evolved was directly measured by an online gas chromato- μmol·h 1·cm 1, and the AQY reached 44.2%, approaching the 2 graph (GC), while the amount of O was quantified based on the IQY and achieving nearly zero optical loss (SI Appendix, Fig. S12 2 charge passed and the 100% Faradaic efficiency of IrOx catalysts and Movie S1). (Materials and Methods). In both the PC and PVE steps, 22.2 C The numerical simulation also revealed the property that of equivalent charges have been transferred, resulting in a 2:1 limited the measured IQY to be 44.3%. Experimentally, we molar ratio of H2 and O2 evolved. For this demonstration, the obtained a CdS carrier lifetime of 0.1 ns (SI Appendix, Fig. μ · −1· −2 A 25.6 mol h cm rate of H2 evolution at 1 atm under 1-sun S13 ), which is short. Using the experimental lifetime, we in- illumination indicated STH conversion efficiency of 1.7%, troduced Shockley–Read–Hall recombination in the model −2 equivalent to a photocurrent of 1.4 mA·cm . The H2 rate of

Materials and Methods − − ENGINEERING ( ) to show that 44.0% of photoexcited 25.6 μmol·h 1·cm 2 was kept the same after two consecutive electrons were collected as photocurrents. Further variation of cycles as in the first cycle, and the same redox mediator solution the modeling parameter showed that the IQY could reach 90% without addition of new solutions also returned to the original as long as the carrier lifetime of commercial CdS powders was OCP after each PVE regeneration (SI Appendix, Note 4 and Fig. longer than 0.5 ns (SI Appendix, Fig. S13B). Such sensitivity to S15). Both experiments indicated complete regeneration and the carrier lifetime is consistent with the diffusion-controlled stability of the redox mediators. In addition, SI Appendix, Note 4 charge separation driven by a small barrier height asymmetry. also includes further discussion about the polysulfide redox Therefore, increasing the carrier lifetime by synthesizing higher couple and potential improvement of the PVE device. QY photoabsorbers is desirable (SI Appendix, Note 3) (39) and CdS photocatalysts are known to rapidly degrade due to can follow various reported strategies (44). photooxidation of lattice sulfides, which forms soluble sulfates or

Fig. 4. Redox-mediated water-splitting reactor. (A) Schematics of the reactor, which is constructed by connecting a redox-mediated photocatalysis (PC)

chamber with a photovoltaic-driven electrolysis (PVE) device. (B) Stoichiometric amount of H2 and O2 evolved with STH 1.7% from the CdS panel during three cycles of alternating PC and PVE processes, respectively. (C) Continuous H2 generation over 150 h by the CdS/3-nm TiO2/Rh@CrOx panel (red), compared to the ceased generation by the bare CdS panel without coatings and cocatalysts after 15 h (black).

Zhao et al. PNAS | 5of8 A coating strategy to achieve effective local charge separation for photocatalytic https://doi.org/10.1073/pnas.2023552118 coevolution Downloaded by guest on September 27, 2021 insulating sulfur (46). Typical CdS photocatalysts were reported into photocatalytic panels following the design principle. The effi- to evolve H2 that lasted for only a few hours (47, 48). We tested cient charge separation was created locally by asymmetries in bar- the stability of TiO2-stablized CdS panel by measuring the cu- rier heights between the neighboring HER and SOR sites. Each of mulative amount of H2 evolved in 50 mM Na2S solutions (pH the two sites were formed nanometers apart and distributed along 13.0) over time (SI Appendix, Fig. S16 and Note 5, and Movie the liquid–junction interface. The CdS panels showed a maximum μ · −1· −1 S2). As shown in Fig. 4C, the CdS/3-nm TiO2/Rh@CrOx panel H2 evolution rate of 90.6 mol h cm ,AQYof44.2%and150h 2− 2− produced 1-atm H2 for 150 h continuously. In comparison, the of stable operation in reversible Sn /S redox solutions. By bare CdS panel produced H2 at a much lower rate while the regenerating the redox mediator, a PC-PVE solar fuel generator activity dropped to almost zero after 10 h. The lack of TiO2 demonstrated multiple-cycle overall water splitting with STH of coating made the CdS powder film delaminate even during the 1.7%. The performance of the CdS panel and redox-mediated re- Rh photodeposition. The core-level X-ray photoelectron spec- actor set records for visible-light–driven particulate water splitting troscopy (XPS) spectra of Cd 3d and S 3p showed little changes (SI Appendix,TableS3). The current photocatalytic activity was after 150 h (SI Appendix, Fig. S17). The absence of SOx signature limited by the intrinsic properties of CdS powders, not by the suggested that the TiO2 coating eliminated the primary failure coating-stabilization approach. The redox-mediated coevolution mode of CdS photocorrosion. The H2-production activity de- also featured the H2 and O2 production separation. Future direc- creased by 20% and 30% after the first 100 h, and 150 h, re- tions include 1) improving carrier lifetime of semiconductors, 2) spectively, which was not due to the loss of CdS particles (SI employing other reversible redox mediators and O2-evolving pho- Appendix, Fig. S18 and Note 5) (10). tocatalysts to optimize the redox regeneration processes, and 3) improving hole transport through >10-nm coatings. Charge Separation Coating Applied to GaInP2 Photocatalytic Panels. The application of the conformal coating and cocatalyst attach- Materials and Methods ment method can be extended to constructing GaInP2 photo- Chemicals and Materials. CdS powders were purchased from Nanoshel. N-type = – × 17 −3 catalysts for coevolution reactions (Materials and Methods). With GaInP2 (100) with 500-nm thickness and ND 1 2 10 cm grown on GaAs (100) wafers was purchased from University Wafe. Sodium hexa- 3-nm TiO2 and Rh@CrOx deposition, the n-type epitaxial − − 1 2 chlororhodate (III) (Na3RhCl6), potassium chromate (K2CrO4, >99.0%), and GaInP2 film evolved 1-atm H2 at 144.7 μmol·h ·cm in the tetrakis (dimethylamido)titanium (IV) (TDMAT) (99.999%) were purchased Na2S solution, equivalent to a photocurrent density of 7.8 −2 from Sigma-Aldrich. Anhydrous methanol (99.0%) and anhydrous sodium mA·cm and 11 times of the as-grown GaInP2 (Fig. 5A). This sulfide (Na2S) were purchased from Alfa Aesar. Cadmium chloride (CdCl2, activity improvement has a similar trend to that of the CdS 99.99%) was purchased from Acros Organics. photocatalyst. The same energetic study as of the CdS photo-

catalysts measured barrier heights of 0.44 and 0.82 V for the Fabrication of CdS/TiO2 Panels. CdS powders were treated by CdCl2 before – GaInP2/TiO2/Rh@CrOx and bare GaInP2/TiO2 liquid interface, constructing photocatalytic panels. The CdCl2 surface modification improved respectively, yielding an energetic asymmetry of 0.38 V (SI Ap- the optoelectronic properties of CdS. CdS powders (5 mg/mL) and CdCl2 pendix, Figs. S19 and S20 and Note 1). This is a demonstration of (10 mM) were mixed in methanol and sonicated for 10 min. Then the powders were dried by filtering through filter paper and evaporating the residue coevolving GaInP2 photocatalysts, and the activity is >10 times SI methanol at 80 °C to obtain CdCl2-treated CdS powders. On a precleaned of the other reported photocatalysts in this bandgap range ( 2 Appendix (1 M HCl and deionized [DI] water) frosted glass substrate (1.8 × 2.5 cm ), , Table S3). We construct the band diagram between μ B 10 mg of CdCl2-treated CdS powders were added with 30 L of methanol to the adjacent HER and SOR sites (Fig. 5 ), showing favorable form a slurry. A uniform film of powders was made via the roll press method

band bending for charge separation and transport, which is with a glass rod. For the TiO2 coating, CdS photocatalyst panels were placed consistent with that of the CdS example. in the ALD chamber and maintained at 150 °C. H2O and TDMAT were pulsed alternatively into the chamber as the oxidizing and titanium precursors, Conclusions respectively. Once the desired number of cycles were reached, the ALD This work demonstrates a coating-stabilization strategy for con- process was terminated, and the TiO2-coated CdS panels were cooled down to room temperature for use. structing stable photocatalysts, i.e., using thin TiO2 and selective H2-evolving cocatalysts. Two classes of high quantum-efficiency photoabsorbers, specifically CdS and GaInP , have been fabricated Fabrication of GaInP2/TiO2 Panels. The commercial GaInP2 of 500 nm was 2 grown on lattice-matched n+-GaAs (100) wafers with electron concentration − of 1∼2 × 1017 cm 3. A multilayer metal contact was sputtered in the order of Ni/Au-Ge/Ag/Au with thicknesses of 5/50/50/30 nm on the backside of the n+- GaAs. The substrate was annealed at 350 °C for 1 min to form ohmic contacts. A 2-min treatment in HF buffer was applied to remove surface oxide.

Immediately following the HF treatment, the TiO2 ALD is performed in the identical way as described above. Finally, copper wires were soldered to the back contact by indium and encapsulated by epoxy.

Photodeposition of Rh@CrOx on CdS/TiO2. For both photodeposition and photocatalytic reactions, we utilized a home-built photoreactor, which consists of a closed-loop glass manifold connected to a GC (SRI 8610C #3) via automatic sampling valves. The GC quantified the amount of gas produced with a molecular sieve column (MS-13X) and a thermal conductivity detec- tor. Argon (Ar) was used as a carrier gas for the GC. The glass manifold was − strictly sealed and airtight to ensure the vacuum level of 1 × 10 2 torr. A top- irradiation reaction vessel containing the photocatalyst and reactant solution was connected to the glass manifold with circulating cooling water to

Fig. 5. GaInP2/TiO2/Rh@CrOx panel for photocatalytic coevolution. (A) Time keep the reaction temperature at 10 °C. The system was degassed and courses of H2 evolution from the GaInP2 panel of bare surface (black), with purged with Ar to remove air; and a background pressure (Ar + water vapor) 3-nm TiO2 coating (red), with TiO2/Rh (blue), and with TiO2/Rh@CrOx (green) was adjusted to 100 torr before each reaction. The irradiation at the sample ∼ · −2 in 100 mM Na2S solutions (pH 13.5). (B) Band diagram of the GaInP2 pho- was 340 mW cm generated from a 1,000-W mercury-xenon arc lamp with tocatalyst between the SOR (Left) and HER (Right) sites under illumination an optical cutoff filter (λ ≥ 395 nm).

along the liquid interface, but not into the semiconductor. Electron and hole For photodeposition of cocatalysts, a TiO2-coated CdS panel containing quasi-Fermi levels are shown by the red and black dashed lines, respectively. 10 mg of CdS powders was placed in a degassed mixture of 8 mL of DI water

6of8 | PNAS Zhao et al. https://doi.org/10.1073/pnas.2023552118 A coating strategy to achieve effective local charge separation for photocatalytic coevolution Downloaded by guest on September 27, 2021 and 2 mL of methanol with 1 wt% of Rh from Na3RhCl6 (373 μg in the 10-mL home-made acrylic frame. A CEM separated the two compartments. Pt on liquid mixture). During the photodeposition, the amount of H2 evolved was carbon cloth (the Fuel Cell Store) was used as cathodic catalysts. IrOx nano- monitored by the GC every 30 min. The deposition was stopped after 1 to particles were synthesized and electrodeposited on carbon paper following

1.5 h as the H2-evolution rate became steady state. Following the Rh de- a reported method (50). An a-Si solar panel with the same active area placed position, the CrOx shell was deposited in a similar approach, where 1 wt% of in tandem with the photoreactor powered the PVE cell. Cr from K2CrO4 (373 μg in the 10-mL liquid mixture) was added to the mixture of DI water and methanol. Photodepositon of CrOx typically lasted Faradaic Efficiency Measurements. Faradaic efficiency (FE) was measured in an 2.5 to 3 h depending on when the H2-evolution rate became steady. The airtight cell with a three-electrode setup and two additional ports for carrier 1 wt% of Rh and 1 wt% of Cr were found to be the optimal loading con- gas. The gas outlet was connected to the GC for analyzing the gas products. dition that yielded the highest H2-evolution rate. Counter electrode (Pt wire) was separated from the cell by a Nafion 117 membrane. Ar carrier gas was continuously purged through the cell to the

Electrodeposition of Rh@CrOx on GaInP2/TiO2. To deposit Rh, the GaInP2/TiO2 GC. The flow rate was kept constant (10 sccm) using a mass flow controller, photoelectrode was cyclically scanned three times from –0.3 to 0.1 V vs. Ag/ which was calibrated using a bubble meter. The FE was determined by the −1 AgCl in an N2-purged aqueous Na3RhCl6 solution (5 mg·mL ). Then the CrOx ratio between the produced O2 and the amount of passed charges on the −1 deposition was conducted in a 10 mg·mL K2CrO4 solution in DI water by oxygen evolution reaction catalysts. Following this method, we measured a applying a constant potential of –1.0 V vs. Ag/AgCl for 5 h under N2 purging. ∼100% FE for the IrOx/C catalyst under chronopotentiometry at 1 mA.

Photocatalytic Reactions in the Na2S Solution. For a typical reaction, a CdS/TiO2 Simulation of Photocatalyst Electrostatics. The Semiconductor Module in panel or GaInP2/TiO2 photoelectrode loaded with Rh@CrOx cocatalysts was COMSOL Multiphysics, version 5.4, was used to build a two-dimensional immersed in 10 mL of Na2S aqueous solution. An optimal activity was ob- model for the CdS/TiO2/Rh@CrOx structure. The CdS was modeled as a served when the Na2S concentration was 100 mM (pH 13.5), which was then semisphere of 1 μm in radius, which was conformally coated with a 3-nm used for all rate measurements. The effect of Na2S concentration is still TiO2 layer. The Rh/CrOx sites of 100 nm in length were distributed periodi- under investigation. The background pressure in the photoreactor was ad- cally on the surface. The length scale of Rh sites is such that their local barrier

justed to be close to 760 torr before the reaction began. A solar simulator heights are not influenced by CdS/TiO2 surroundings (51). The energetics (AAA-grade; Abet Technologies) with an AM1.5G filter was used as the il- were simulated by solving the Poisson’s equation, drift-diffusion current, lumination source. The 1-sun illumination intensity was calibrated by a and continuity equations under open-circuit conditions. The heterojunction

certified photodiode. boundary condition was used for the CdS/TiO2 interface. Schottky contacts were set for the TiO2– and Rh@CrOx–liquid sites with the barrier heights of ∼ 5 −1 Quantum Yield Measurements. Photocatalytic H2 evolution was performed 0.60 and 0.11 V, respectively. The absorption coefficient of CdS is 10 cm . twice with a 425-nm and a 450-nm long-pass filter (Edmund Optics) under Light can penetrate ∼1 μm deep in the CdS. As a simplification, a generation × 20 −3· −1 μ AM1.5G illumination, respectively. The corresponding H2 evolution rates rate of 1 10 cm s was introduced uniformly in the 1- m CdS to mimic ± 1-sun illumination. Shockley–Read–Hall recombination processes were imi- ENGINEERING were obtained as rH2 ,425 nm and rH2 ,450 nm. The rate at 438 12.5 nm was then determined by rH ,438 nm = rH ,425nm − rH ,450 nm. AQY was calculated according tated by setting electron trap states with 0.25 eV below the CB of CdS (52) 2 2 2 − =( × × )=( · = υ) for the energy level and 1 × 1010 cm 3 for the trap density in the model. For to the following: AQY 2 rH2 ,438 nm NA A I438nm h , where NA is × 18 −3· −1 Avogadro’s number, A is the illuminated area, I438 nm is light intensity at 438 the IQY simulation, the generation rate was reduced to 3 10 cm s , nm, h is Planck’s constant, and ν is the photon’s frequency at 438 nm. corresponding to the light intensity of 438 ± 12.5 nm in the solar spectra. – AM1.5G spectrum was used to calculate the I438 nm by integrating the light Both the electron and hole surface recombination velocities at the TiO2 intensity from 425 to 450 nm to count for spectral dependence. liquid Schottky junction were varied in the same way to achieve a thermionic current that matched with the experimental result. OCP Measurements. CdS powder-based photoelectrodes were made through the particle transfer method (49). Five hundred nanometers of Ti and Materials Characterization. The scanning electron microscopy (SEM), atomic 200 nm of Au were sputtered onto the CdS film as conductive back contacts. force microscopy (AFM), and transmission electron microscopy images were Carbon tape attached to a glass substrate was then used to peel off the CdS taken by a Hitachi SU8230 UHR system, a Cypher ES Environmental AFM particles covered by the metal layers. The electrode was then loaded into the system, and an FEI Tecnai Osiris system (200 kV) equipped with energy-

ALD tool for TiO2 deposition. Finally, the electrode was completed by con- dispersive X-ray spectroscopy, respectively. XPS measurements were con- tacting copper wires with the carbon tape and encapsulated by epoxy. The ducted on the PHI VersaProbeII. Diffused reflectance spectroscopy of CdS OCP measurements were conducted on a Bio-Logic S200 potentiostat with panel was obtained on a UV-visible spectrometer (UV-2600; SHIMADZU). Ag/AgCl used as a reference electrode and a carbon rod as a counter elec- Time-resolved photoluminescence was collected using time-correlated

trode. All OCPs of the CdS powder electrode and GaInP2 electrode were single-photon counting (PicoHarp 300). The optical excitation was pro- measured in 10 mM Na2S solutions. vided by a 480-nm pulsed laser, and the detection wavelength was centered at 550 nm with a 500-nm long-pass filter. Electrochemical Impedance Spectroscopy Measurements. The CdS powder-

based photoelectrode or GaInP2 photoelectrode was measured in the Data Availability. All study data are included in the article and/or supporting 3−/4− aqueous [Fe(CN)6] solution at pH 12 [50 mM K3Fe(CN)6 and 350 mM information. K4Fe(CN)6] in the complete dark. The impedance was measured from 10 kHz to 1 Hz and fitted with a Randles circuit to extract the capacitance of the ACKNOWLEDGMENTS. We thank the startup support from Yale Energy liquid junction. In the Mott−Schottky plot, the intercept to the x axis pro- Sciences Institute. We thank Dr. Cheng Hua and Jake Heinlein for automa- vided the flat-band potential, and the slope was used to calculate the tion of the photoreaction measurements, Dr. Joerg Nikolaus for the timely doping concentration of the semiconductor. help with the AFM measurements, Prof. Brudvig for supporting the UV-vis spectrometer, Junying Tang for preparing the IrOx nanoparticles, and Xianb- ing Miao for setting up the Faraday efficiency apparatus. R.Y. acknowledges PVE Device. Flex-stak PEM fuel cell (purchased from the Fuel Cell Store) was fellowship support from the Japan Student Services Organization. Y.X. utilized to assemble the PVE device. Both the anodic and cathodic com- thanks the Tsinghua University undergraduate overseas academic research partment were composed of a plastic end plate, a graphite plate, and a support program.

1. S. Chen, T. Takata, K. Domen, Particulate photocatalysts for overall water splitting. 6. T. Takata et al., Photocatalytic water splitting with a quantum efficiency of almost Nat. Rev. Mater. 2, 17050 (2017). unity. Nature 581, 411–414 (2020). 2. D. M. Fabian et al., Particle suspension reactors and materials for solar-driven water 7. Q. Wang et al., Particulate photocatalyst sheets based on carbon conductor layer for splitting. Energy Environ. Sci. 8, 2825–2850 (2015). efficient z-scheme pure-water splitting at ambient pressure. J. Am. Chem. Soc. 139, 3. B. AlOtaibi, S. Fan, D. Wang, J. Ye, Z. Mi, Wafer-level artificial photosynthesis for 1675–1683 (2017).

CO2 reduction into CH4 and CO using GaN nanowires. ACS Catal. 5,5342–5348 8. W.-P. Hsu, M. Mishra, W.-S. Liu, C.-Y. Su, T.-P. Perng, Fabrication of direct Z-scheme

(2015). Ta3N5-WO2.72 film heterojunction photocatalyst for enhanced hydrogen evolution. 4. F. E. Osterloh, B. A. Parkinson, Recent developments in solar water-splitting photo- Appl. Catal. B 201, 511–517 (2017). catalysis. MRS Bull. 36,17–22 (2011). 9. J.-H. Yang et al., Highly enhanced photocatalytic water-splitting activity of gallium 5. Y. Goto et al., A particulate photocatalyst water-splitting panel for large-scale solar zinc oxynitride derived from flux-assisted Zn/Ga layered double hydroxides. Ind. Eng. hydrogen generation. Joule 2, 509–520 (2018). Chem. Res. 57, 16264–16271 (2018).

Zhao et al. PNAS | 7of8 A coating strategy to achieve effective local charge separation for photocatalytic https://doi.org/10.1073/pnas.2023552118 coevolution Downloaded by guest on September 27, 2021 10. Q. Wang et al., Oxysulfide photocatalyst for visible-light-driven overall water split- 32. Y. Qi et al., Inhibiting competing reactions of iodate/iodide redox mediators by sur- ting. Nat. Mater. 18, 827–832 (2019). face modification of photocatalysts to enable Z-scheme overall water splitting. Appl. 11. Q. Wang, K. Domen, Particulate photocatalysts for light-driven water splitting: Catal. B 224, 579–585 (2018). Mechanisms, challenges, and design strategies. Chem. Rev. 120, 919–985 (2020). 33. Y. Qi et al., Redox-based visible-light-driven Z-scheme overall water splitting with 12. B. M. Kayes et al., “27.6% Conversion efficiency, a new record for single-junction solar apparent quantum efficiency exceeding 10%. Joule 2, 2393–2402 (2018). cells under 1 sun illumination” in 37th IEEE Photovoltaic Specialists Conference (IEEE, 34. Y. Zhao et al., A hdrogen farm strategy for scalable solar hydrogen production with 2011), pp. 000004–000008. particulate photocatalysts. Angew. Chem. 132, 9740–9745 (2020).

13. J. Wallentin et al., InP nanowire array solar cells achieving 13.8% efficiency by ex- 35. K. Maeda et al., Noble-metal/Cr2O3 core/shell nanoparticles as a cocatalyst for pho- ceeding the ray optics limit. Science 339, 1057–1060 (2013). tocatalytic overall water splitting. Angew. Chem. Int. Ed. Engl. 45, 7806–7809 (2006).

14. W. K. Metzger et al., Exceeding 20% efficiency with in situ group V doping in poly- 36. G. Siddiqi et al., Stable water oxidation in acid using manganese-modified TiO2 crystalline CdTe solar cells. Nat. Energy 4, 837–845 (2019). protective coatings. ACS Appl. Mater. Interfaces 10, 18805–18815 (2018). 15. J. L. Young et al., Direct solar-to-hydrogen conversion via inverted metamorphic 37. X. Chen, X. Shen, S. Shen, M. O. Reese, S. Hu, Stable CdTe photoanodes with ener- multi-junction semiconductor architectures. Nat. Energy 2, 17028 (2017). getics matching those of a coating intermediate band. ACS Energy Lett. 5, 1865–1871 16. W. H. Cheng et al., Monolithic photoelectrochemical device for direct water splitting (2020). – with 19% efficiency. ACS Energy Lett. 3, 1795 1800 (2018). 38. F. Lin, S. W. Boettcher, Adaptive semiconductor/electrocatalyst junctions in water- 17. A. T. Garcia-Esparza, K. Takanabe, A simplified theoretical guideline for overall water splitting photoanodes. Nat. Mater. 13,81–86 (2014). splitting using photocatalyst particles. J. Mater. Chem. A Mater. Energy Sustain. 4, 39. Z. Pan et al., Elucidating charge separation in particulate photocatalysts using nearly – 2894 2908 (2016). intrinsic semiconductors with small asymmetric band bending. Sustain. Energy Fuels 3, 18. S. Hu et al., Thin-film materials for the protection of semiconducting photoelectrodes 850–864 (2019). – in solar-fuel generators. J. Phys. Chem. C 119, 24201 24228 (2015). 40. J. Wang et al., Reducing surface recombination velocities at the electrical contacts will 19. L. Liu et al., A transparent CdS@TiO2 nanotextile photoanode with boosted photo- improve perovskite photovoltaics. ACS Energy Lett. 4, 222–227 (2019). – electrocatalytic efficiency and stability. Nanoscale 9, 15650 15657 (2017). 41. E. T. Roe, K. E. Egelhofer, M. C. Lonergan, Limits of contact selectivity/recombination 20. R. Wang, L. Wang, Y. Zhou, Z. Zou, Al-ZnO/CdS photoanode modified with a triple on the open-circuit voltage of a photovoltaic. ACS Appl. Energy Mater. 1, 1037–1046 functions conformal TiO film for enhanced photoelectrochemical efficiency and 2 (2018). stability. Appl. Catal. B 255, 117738 (2019). 42. C. Gao et al., Heterogeneous single-atom photocatalysts: Fundamentals and appli- 21. F. E. Osterloh, Photocatalysis versus photosynthesis: A sensitivity analysis of devices cations. Chem. Rev. 120, 12175–12216 (2020). for solar energy conversion and chemical transformations. ACS Energy Lett. 2, 43. A. G. Scheuermann, J. D. Prange, M. Gunji, C. E. D. Chidsey, P. C. McIntyre, Effects of 445–453 (2017). catalyst material and atomic layer deposited TiO oxide thickness on the water oxi- 22. F. A. Chowdhury, M. L. Trudeau, H. Guo, Z. Mi, A photochemical diode artificial 2 dation performance of metal–insulator–silicon anodes. Energy Environ. Sci. 6, 2487 photosynthesis system for unassisted high efficiency overall pure water splitting. Nat. (2013). Commun. 9, 1707 (2018). 44. W. D. Kim et al., Role of surface states in photocatalysis: Study of chlorine-passivated 23. B. W. Roehrich, R. Han, F. E. Osterloh, Hydrogen evolution with fluorescein-sensitized CdSe nanocrystals for photocatalytic hydrogen generation. Chem. Mater. 28, 962–968 Pt/SrTiO3 nanocrystal photocatalysts is limited by dye adsorption and regeneration. (2016). J. Photochem. Photobiol. Chem. 400, 112705 (2020). 45. S. Keene, R. Bala Chandran, S. Ardo, Calculations of theoretical efficiencies for 24. H. Huang et al., Oriented built-in electric field introduced by surface gradient diffusion doping for enhanced photocatalytic H evolution in CdS nanorods. Nano Lett. 17, electrochemically-mediated tandem solar water splitting as a function of bandgap 2 – 3803–3808 (2017). energies and redox shuttle potential. Energy Environ. Sci. 12, 261 272 (2019). 25. K. Maeda, R. Abe, K. Domen, Role and function of ruthenium species as promoters 46. X. Ning, G. Lu, Photocorrosion inhibition of CdS-based catalysts for photocatalytic – with TaON-based photocatalysts for oxygen evolution in two-step water splitting overall water splitting. Nanoscale 12, 1213 1223 (2020). under visible light. J. Phys. Chem. C 115, 3057–3064 (2011). 47. Z. Li et al., Biomimetic electron transport via multiredox shuttles from photosystem II 26. K. Maeda et al., Preparation of core-shell-structured nanoparticles (with a noble- to a photoelectrochemical cell for solar water splitting. Energy Environ. Sci. 10, – metal or metal oxide core and a chromia shell) and their application in water split- 765 771 (2017). ting by means of visible light. Chemistry 16, 7750–7759 (2010). 48. C. M. Wolff et al., All-in-one visible-light-driven water splitting by combining nano- – 27. L. Hammarström, Accumulative charge separation for solar fuels production: Cou- particulate and molecular co-catalysts on CdS nanorods. Nat. Energy 3, 862 869 pling light-induced single electron transfer to multielectron catalysis. Acc. Chem. Res. (2018). 48, 840–850 (2015). 49. T. Minegishi, N. Nishimura, J. Kubota, K. Domen, Photoelectrochemical properties of 28. Y.-J. Yuan, D. Chen, Z.-T. Yu, Z.-G. Zou, Cadmium sulfide-based nanomaterials for LaTiO2N electrodes prepared by particle transfer for sunlight-driven water splitting. photocatalytic hydrogen production. J. Mater. Chem. A Mater. Energy Sustain. 6, Chem. Sci. (Camb.) 4, 1120 (2013). 11606–11630 (2018). 50. Y. Zhao, E. A. Hernandez-Pagan, N. M. Vargas-Barbosa, J. L. Dysart, T. E. Mallouk, A 29. Z. Pan et al., Mutually-dependent kinetics and energetics of photocatalyst/co-catalyst/ high yield synthesis of ligand-free iridium oxide nanoparticles with high electro- two-redox liquid junctions. Energy Environ. Sci. 13, 162–173 (2020). catalytic activity. J. Phys. Chem. Lett. 2, 402–406 (2011). 30. R. Li et al., Spatial separation of photogenerated electrons and holes among 010 and 51. F. A. L. Laskowski et al., Nanoscale semiconductor/catalyst interfaces in photo- – 110 crystal facets of BiVO4. Nat. Commun. 4, 1432 (2013). electrochemistry. Nat. Mater. 19,69 76 (2020). 31. S. Zhang et al., Recent progress in polysulfide redox‐flow batteries. Batter. Supercaps 52. K. H. Nicholas, J. Woods, The evaluation of electron trapping parameters from con- 2, 627–637 (2019). ductivity glow curves in cadmium sulphide. Br. J. Appl. Phys. 15, 783–795 (1964).

8of8 | PNAS Zhao et al. https://doi.org/10.1073/pnas.2023552118 A coating strategy to achieve effective local charge separation for photocatalytic coevolution Downloaded by guest on September 27, 2021