High carrier mobility along the [111] orientation in Cu2O photoelectrodes

Preparation of the Cu₂O thin films

Si wafers of (100), (110) and (111) orientations without photoelectric function were used as the substrates for epitaxy. All wafers are phosphorus-doped with a resistivity of less than 0.001 Ω cm (Sil’tronix ST). To remove organic contamination on the surface, the wafers were sonicated in warm acetone, cold methanol and deionized water baths in sequence for 10 min each and blow-dried with nitrogen between each bath. Then 5% hydrofluoric acid solution (extreme caution and professional training should be exercised with this highly corrosive and volatile acid) was used to dissolve the native oxide layer and create a hydrogen-terminated surface. After 5 min dipping, the wafer was thoroughly rinsed with deionized water and installed on the electrodeposition set-up directly. The gold buffer layer was electrodeposited in a modified epitaxial method previously designed for epitaxial lift-off procedures¹³. Before immersing the substrates into electrolyte containing 0.06 mM HAuCl₄, 1 mM KCl, 1 mM H₂SO₄ (for stabilizing the Au precursor) and 20 mM K₂SO₄ (for enhancing the electrolyte conductivity), the cathode was biased at −1.9 V versus a Ag/AgCl reference electrode. The electrolyte was saturated with N₂ by 1 h of bubbling and continuous bubbling during film growth. To guarantee the uniformity of the gold layer, the cathode was adjusted to face the electrolyte current with approximately 1.5 cm to the Pt coil counter electrode and 2 mm to the reference electrode. The growth was carried out in a 250-ml container, with stirring at 150 r.p.m. In a typical gold deposition, the polarization is maintained for 20 min. The Cu₂O epitaxy was developed from previous methods^4,33, and subsequently carried out in a buffered copper sulphate solution while using the Au epitaxial configuration. A 7.98 g quantity of Cu₂SO₄, 21.77 g K₂HPO₄ and 67.5 g lactic acid were dissolved in 250 ml H₂O and then the pH of the solution was adjusted to 12 by adding 2 M KOH solution, thereby yielding a final solution volume of 1 l. The Cu₂O layer was deposited by chronopotentiometry in a cylinder cell with an 8 cm diameter with continuous stirring (200 r.p.m.). The electrolyte was bubbled with N₂ throughout the growth. The cathodic deposition was carried out in a two-electrode configuration with a Pt counter electrode using a current density of −0.02 mA cm⁻². The parallel working electrode and counter electrode are separated by a distance of approximately 0.8 cm, and their respective areas have an approximate ratio of 1:5. The deposition duration varies, and all samples were rinsed with copious deionized water before drying. Polycrystalline Cu₂O with preferential orientations was electrodeposited in the above-mentioned buffered copper sulphate solution on Au/FTO substrates. A 300-nm layer of polycrystalline Au was sputtered onto FTO substrates cleaned by acetone, ethanol and detergent under sonication. Poly-Cu₂O(100), poly-Cu₂O(110) and poly-Cu₂O(111) were prepared in pH 9.0, 9.6 and 12.6 solutions, respectively. All pH values in the specific electrolytes were measured and tuned three times at 1-h intervals, resulting in a volume of 1 l. Electrodeposition was carried out with these solutions after overnight stirring using a constant −0.1 mA cm⁻² current density. Thin films for photocathodes had an electrodeposition duration of 100 min followed by a thorough rinse with copious deionized water and drying with an air gun.

Fabrication of the photocathodes

To examine the PEC performance of the photoelectrodes with various types of Cu₂O, atomic layer deposition was applied to build the buried p–n junction³. Directly after the epitaxial layers, Ga₂O₃ was deposited at a substrate temperature of 150 °C using bis(μ-dimethylamino)tetrakis(dimethylamino)digallium (98%; Strem Chemicals) as the precursor and deionized water as the oxidant. The Ga precursor was preheated to 125 °C to maintain a sufficient vapour pressure. In every deposition cycle, 0.5 s precursor pulsing time, 0.05 s water pulse and 15 s pumping time were programmed with a constant nitrogen carrier gas of 10 sccm. For all Cu₂O photocathodes, the deposition cycle number is set to 135, resulting in a Ga₂O₃ layer of 20 nm. A TiO₂ layer was then coated to protect the photocathode against electrolyte corrosion using tetrakis(dimethylamino)titanium (99.999%; Sigma) as the metal precursor and deionized water as the oxidant. The chamber temperature remained at 150 °C whereas the precursor was heated to 75 °C. Each cycle consisted of 0.1 s precursor pulsing time, 0.05 s water pulse and 15 s pumping time. A total of 340 cycles of TiO₂ deposition were applied, resulting in a TiO₂ layer of 20 nm. The polycrystalline devices for stability tests use 100 nm TiO₂ as a protection layer. A RuO_x hydrogen evolution catalyst was used to extract photo-generated electrons for water reduction. Briefly, the deposition was carried out in a 1.3 mM KRuO₄ solution at a current density of −10 μA cm⁻² under simulated one-sun illumination. Each catalyst decoration required 6 min, and a platinum wire was used as the counter electrode.

Materials characterization

The crystal information of the Cu₂O films was characterized by X-ray diffraction on an Empyrean system (PANalytical) with a PIXcel-1D detector and Cu Kα radiation. Diffraction patterns were recorded at a scan rate of 1° min⁻¹ with a step width of 0.02° using a two-bounce hybrid monochromator and parallel-plate collimator diffracted optics. High-resolution scanning electron microscopy was carried out on the Zeiss Merlin or the Gemini 800 with in-lens detectors and an Oxford Instruments EBSD detector. The accelerating voltage was set to 30 kV and a 50-µm aperture was used. For EBSD, the sample was tilted at 70° towards the detector, approximately 17 mm from the scanning electron microscope pole piece, with the EBSD detector situated approximately 20–25 mm from the sample surface. The EBSD was calibrated for Cu₂O, Au and Si patterns before the construction of each map to maximize the chance of successful identification and orientation measurement. All high-resolution cross-sectional transmission electron micrographs were collected with a Tecnai Osiris (FEI) microscope on electron-transparent samples prepared by focused ion beam sampling. Elemental composition was analysed by the Escalab 250Xi X-ray photoemission spectroscopy instrument. The bulk chemical analysis was carried out by combining X-ray photoelectron spectroscopy with argon ion gun etching with an ion energy of 1,000 eV.

PEC measurements

All PEC measurements were implemented in a three-electrode configuration with Cu₂O photocathodes as the working electrodes, platinum wires as the counter electrodes and Ag/AgCl reference electrodes. The pH 5.01 and pH 7.01 buffer electrolytes were prepared by tuning the 0.5 M Na₂SO₄, 0.1 M K₂HPO₄ and 0.1 M KH₂PO₄ phosphate solution. The pH 9.03 buffer was prepared by tuning the 0.5 M Na₂SO₄, 0.1 M Na₂CO₃ and 0.1 M NaHCO₃ carbonate solution with sulfuric acid. Potentiostats (SP-200 or SP-150e; Biologic) were used to acquire the photoresponse under chopped illumination from the LCS-100 solar simulator (class ABB; Newport) with a built-in AM1.5 G filter. Calibration was carried out across the 300–800 nm wavelength region with a certified silicon diode behind a KG3 filter. All linear-sweep voltammetry was carried out at a 10 mV s⁻¹ rate. Potentials versus the Ag/AgCl reference electrode were transformed to the reversible hydrogen electrode scale using the following equation:

Both stability tests and IPCEs were measured in the pH 5.0 buffer at 0.5 V versus the RHE. The IPCE was measured by comparing the wavelength-dependent photoresponse of the photoelectrodes with that of a silicon photodiode (FDS100-CAL, Thorlabs) using light from a 300-W xenon lamp passed through a monochromator (TLS300XU, Newport). Current readings were taken 5 s after each wavelength change. The PEC performance with sacrificial agent was obtained by scanning the photocathodes in 0.1 M europium(III) nitrate solution with simulated one-sun illumination. Details on calculation of the charge separation efficiencies are provided in Supplementary Discussion 4.

Gas quantification

The faradaic efficiency of the photocathode was measured in a gas-tight PEC one-room cell. An Ag/AgCl (KCl saturated) reference electrode was used, and a platinum foil was used as the counter electrode. Epoxy was used to mask exposed parts of the cathode. The solution in the cell was vigorously stirred at 400 r.p.m. and constantly purged with argon at a rate of 13 sccm, controlled by a mass flow controller. The gas outlet was connected to a safe bottle that prevents the entry of water vapour into the gas chromatograph. These samples were subsequently analysed by gas chromatography (GC9790plus) every 20 min with 10 min analysing time and 10 min gap. The average exhaust gas flow rate was 12.4 sccm calculated by a burette. Gas chromatography system calibration and faradaic efficiency calculations are described in Supplementary Discussion 5.

Measurements for electronic and optical properties of semiconductors

The electrochemical impedance measurements were carried out in the dark with a Biologic 150e potentiostat. The photocathodes were immersed in a 1 M sodium sulphate solution with a Ag/AgCl reference electrode and Pt wire to form a three-electrode configuration. The potential was varied in the stable region defined by the Pourbaix diagram with an a.c. amplitude of 10 mV superposed on the d.c. component. The space-charge capacitance of the semiconductor varied as a function of the applied potential according to the Mott–Schottky equation as shown below:

in which C is the interfacial capacitance, A is the electrode active area, ε₀ is the vacuum permittivity, ε = 7.5 (ref. ²⁰) is the relative dielectric constant, E is the applied potential, and E_fb is the flat band potential. The majority carrier concentrations (N_A)—that is, the hole concentrations—in the Cu₂O thin films of three orientations were estimated. I–V responses of the hole-only Cu₂O devices were recorded using a Keithley 2450 SourceMeter with the assistance of a d.c. probe station. The Cu₂O hole-only device was prepared by sandwiching the Cu₂O layer with Au and MoO_x/Ag layers. A 10-nm layer of MoO_x and 80 nm of Ag were evaporated onto the Cu₂O to prevent short-circuiting and obtain a suitable workfunction. The samples were kept in a dark environment under a vacuum at room temperature during the measurement. The nonlinear responses with characteristic slopes (ohmic region, n = 1 and Child’s region, n = 2) were acquired and analysed according to the Mott–Gurney law:

in which V is the applied bias, ε₀ is the vacuum permittivity, ε (=7.5) is the relative dielectric constant and L is the thickness of the Cu₂O layer (Supplementary Figs. 12–14 and Supplementary Tables 1 and 2). The voltages were determined by the trap densities (n_t)

in which e is the elementary charge and L is the sample thickness. The ultraviolet–visible reflectance spectra of Cu₂O were collected on a Shimadzu UV-3600 Plus double-beam spectrophotometer in single-beam mode and using an ISR-603 Integrating Sphere Attachment (integrating sphere: 60 mm in diameter). For the measurement of total and specular reflectance spectra, the samples were placed in the respective position on the integrating sphere. The Au-sandwiched SC-Cu₂O devices, for transfer length method measurements, were prepared by evaporating 100 nm of patterned Au layer on the Cu₂O/Au/Si thin films. The single-crystal Au and evaporated Au were wired to a Keithley 2450 SourceMeter using Ag paste. J–V curves were recorded by scanning between −0.1 V to 0.1 V in the dark at room temperature. Owing to the interest in out-of-plane properties of the single-crystal films, the transfer length method measurement was modified. In contrast to the conventional transfer length method, the variable of length is replaced with thickness (five for each crystal orientation). By extrapolating to the intersection point on the y-axis, contact resistivity can be achieved. Photoluminescence measurements were carried out on an IMA Vis hyperspectral microscope. A 405-nm, continuous-wave laser was beam-shaped to form a top-hat profile and was focused onto the back focal plane of an Olympus ×100 air objective lens, producing a wide-field, flat illumination profile. The laser light excited the sample, producing photoluminescence that was collected by the same objective lens. A dichroic beamsplitter removed the excitation light before the light was guided onto a volume Bragg grating that spectrally split the light onto a Hamamatsu ORCA-Flash4.0 V3 digital CMOS camera. The spectra were acquired using 180 W cm⁻² excitation with spectral calibration and dark subtraction.

Transient reflection spectroscopy

Transient reflection spectroscopy measurements were carried out on two set-ups. In the first set-up (ultraviolet–visible probe), the output of a Ti:sapphire amplifier system (Spectra Physics Solstice Ace) operating at 1 kHz and generating pulses of about 100 fs was split into pump and probe beam paths. The 400-nm pump pulses were created by sending the 800-nm fundamental beam of the Solstice Ace through a second-harmonic-generating β-barium borate crystal (Eksma Optics). The pump was blocked by a chopper wheel rotating at 500 Hz. For the probe light path, a mechanical stage (Thorlabs DDS300-E/M) was used to adjust the delay between the pump and the probe. The ultraviolet–visible broadband beam (330–700 nm) was generated by focusing the 800-nm fundamental beam from the mechanical stage onto a CaF₂ crystal (Eksma Optics, 5 mm) that was connected to a digital motion controller (Mercury C-863 DC Motor Controller). The reflected pulses were collected with a monochrome line scan camera (JAI SW-4000M-PMCL, spectrograph: Andor Shamrock SR-163) with collected data fed straight into the computer. In the second set-up (visible–near infrared probe), the output of a Ti:sapphire amplifier system (Spectra Physics Solstice Ace) operating at 1 kHz and generating pulses of about 100 fs was split into two beam paths (pump and probe). The 400-nm pump pulses were created by sending the 800-nm fundamental beam through a second-harmonic-generating β-barium borate crystal (Eksma Optics). The pump was blocked by a chopper wheel rotating at 500 Hz. The visible broadband beam (520–780 nm) was generated in a home-built noncollinear optical parametric amplifier and was sent to a computer-operated mechanical delay stage (Thorlabs DDS300-E/M) to adjust the pump–probe delay. The white light from the delay stage was split into two identical beams (probe and reference) by a 50/50 beamsplitter. The reference beam did not interact with the pump at the sample, which allows for correcting for any shot-to-shot fluctuations. The reflected probe and reference pulses were collected with a silicon dual-line array detector (Hamamatsu S8381-1024Q, spectrograph: Andor Shamrock SR-303i-B) driven and read out by a custom-built board (Stresing Entwicklungsbüro). Calculations for carrier diffusion length, pump fluence and carrier density are discussed in Supplementary Discussion 1.