In the accelerated transition to sustainable energy, solar fuels provide a promising approach to store intermittent solar energy in energy-dense molecules that can usefully be consumed in the existing infrastructure. These processes rely on effective photoinduced charge separation and transport in suitable photoactive materials. In photocatalytic systems, research efforts primarily focus on surface and interface engineering to optimize charge separation efficiencies, owing to built-in electric fields residing at surfaces and typically short charge transporting distances in particles^7,8. By contrast, state-of-the-art photoelectrochemical (PEC) devices utilize solid-state junctions to efficiently separate electrons and holes, leaving charge transport as a vital issue in achieving high performance^5,9. In such systems, bulk carrier recombination processes are identified as the key performance bottlenecks, in which a substantial fraction of photoexcited carriers are lost before being injected into the surface catalysts⁶; such processes remain poorly understood but will be essential to push these technologies forwards.

Single-crystal thin films are widely applied in the semiconductor industry^10,11. The absence of grain boundaries and a substantial decrease in defect density compared with polycrystalline analogues offer distinctive optoelectronic properties. A single-crystal semiconductor is also an ideal platform for detailed spectroscopic studies owing to the ordered and well-defined nature of such structures. Nonetheless, most studies on thin-film semiconductors are carried out with polycrystalline materials owing to the demanding preparation processes for single crystals that often involve high vacuum levels, high temperatures and sophisticated instruments¹².

Here we demonstrate liquid-phase epitaxial growth of unique single-crystal Cu₂O (SC-Cu₂O) thin films under ambient conditions resulting in extremely high-quality crystals with tunable crystal orientations. These unique single-crystal films enable optoelectronic and PEC characterization in steady-state conditions and with a customized femtosecond transient reflection spectroscopy technique spanning the ultraviolet to infrared spectral regions in three crystal orientations. Our collective results show that the mobility, conductivity and carrier diffusion length along the [111] direction are much greater compared with those in other directions in Cu₂O, which are essential bulk charge transport properties for solar fuels, photovoltaics, transistors and detectors. Building on these fundamental insights, we fabricated polycrystalline Cu₂O with an extraordinarily pure (111) orientation and (111) terminating facets using a simple and low-cost method, achieving a Cu₂O PEC performance beyond that of the state-of-the-art.

The Cu₂O and Au proxy layers were grown by liquid-phase epitaxy at room temperature using the three-electrode configuration shown in Fig. 1a. To achieve high-quality films, the substrates need to be fixed opposite to the flow with controlled velocity (Methods). Thin Au buffer layers were first grown on degenerate Si substrates with three selected orientations, followed by deposition of Cu₂O (Supplementary Figs. 1 and 2). The out-of-plane orientations of each layer were determined by X-ray diffraction, as shown in Fig. 1b and Supplementary Fig. 3. Even with the intensity being viewed on the logarithmic scale, only the expected out-of-plane orientation is detected for each pattern, indicating the exceptional orientation purity of the samples. The in-plane orientations were confirmed by electron backscatter diffraction (EBSD) pole figures (Fig. 1c, with corresponding inverse pole figures and z-direction inverse pole figure maps in Supplementary Figs. 4 and 5). Distinct spots occur at the expected azimuthal angles for each given orientation owing to the Bragg condition being satisfied in films with in-plane order¹³. Samples with (100) and (110) orientations show SC-Cu₂O features over all areas, whereas the Cu₂O(111) film is shown to be twinned, with two Cu₂O(111) crystalline domains rotated 60° to each other. Cross-sectional transmission electron microscopy was implemented on focused ion beam-thinned lamellae to image the interfaces of the Cu₂O layer electrodeposited onto the Au buffer layer on a Si(111) substrate (Fig. 1d). The Au section is less transparent owing to the higher resistance against focused ion beam etching during the sampling process. The lattice spacing of each layer was calculated by measuring the distance of 27 lattice lines under the grey value condition (Supplementary Figs. 6–8) and is marked in the close-up images in Fig. 1e,f. Spacings of 0.24 and 0.23 nm were measured for Cu₂O and Au, respectively. The same cubic structure and small lattice mismatch of 4.3% favour successful epitaxy with moderate compressive in-plane strain. The coincident site lattices in which 4 unit meshes of Au coincide with 3 unit meshes of Si allowed the epitaxy at the Au/Si interface even although the nominal lattice mismatch calculated from Fig. 1e is −25.8% (ref. ¹³). Additionally, the epitaxial relation between layers on the SC-Cu₂O(100) sample was also confirmed by cross-sectional transmission electron microscopy (Supplementary Fig. 9). These results collectively provide direct evidence of epitaxial growth of high-quality, aligned Cu₂O on Si wafers using Au as the buffer layer.Fig. 1

a, The electrochemical epitaxy set-up with the three-electrode configuration for liquid-phase epitaxial growth of Cu₂O thin films. b,c, X-ray diffraction patterns (b) and pole figures generated from EBSD (c) for single-crystal epitaxial layers on Si substrates with reflections highlighted in circles. a.u., arbitrary units. d–f, A cross-sectional transmission electron micrograph showing Cu₂O(111)/Au(111)/Si(111) layers (d), with close-up images at the Cu₂O/Au (f) and Au/Si (e) interfaces, with lattice spacings in agreement with bulk references. Scale bars, 10 nm (d) and 2 nm (e,f). g, PEC responses of SC-Cu₂O photocathodes for solar hydrogen evolution under a simulated one-sun condition. The inset shows the layered structure of the Cu₂O photocathodes. h, Mott–Schottky plots of SC-Cu₂O thin films tested in pH 9 carbonate buffer solution. The carrier densities were calculated based on the slopes of the linear fitting represented by the dotted lines. i–k, Current density–voltage curves of hole-only devices with SC-Cu₂O(100) (i), SC-Cu₂O(110) (j) and SC-Cu₂O(111) (k).

Intriguing PEC behaviour was discovered by testing the photocathodes prepared using SC-Cu₂O thin films of different orientations (Fig. 1g) in the illustrated architecture (inset in Fig. 1g). Although all photoelectrodes exhibit similar photovoltages, determined from the same buried p–n junction³, the SC-Cu₂O(111) device delivers a much higher current density of 4.5 mA cm⁻² at 0 V versus the reversible hydrogen electrode (RHE), a factor of two more than those of the other orientations. The scalability of the epitaxial method was demonstrated by fabricating a wafer-size SC-Cu₂O(110) photocathode with an active area of 60 cm² (Supplementary Fig. 10). To understand the anisotropic PEC performance, the orientation-dependent out-of-plane carrier transport properties were investigated. Electrochemical impedance spectroscopy at the semiconductor/electrolyte interface was used to determine the carrier densities and flat band potentials by fitting the Mott–Schottky function (Fig. 1h for 1 kHz; similar results were seen for other frequencies (Supplementary Fig. 11 and Supplementary Table 3)). Although all films demonstrate p-type character and carrier densities of the order of 10¹⁷ cm⁻³, which is similar to that of thermally prepared Cu₂O thin films^14,15, no apparent carrier density anisotropy was observed. The same finding also applies to the orientation-dependent flat band potentials, which are 0.62, 0.60 and 0.63 V versus the RHE for SC-Cu₂O(100), SC-Cu₂O(111) and SC-Cu₂O(110), respectively.

To investigate the carrier mobility, hole-only devices were built with SC-Cu₂O sandwiched between Au and MoO_x/Ag layers (Methods) and space-charge-limited current analyses were carried out^16,17. Figure 1i–k shows the j–V relation of SC-Cu₂O films on a double logarithmic scale. The clear quadratic dependence of the current on the applied voltage is observed in the Child’s law region. From this region, we extract carrier mobilities of 1.29, 0.87 and 15.4 cm² V⁻¹ s⁻¹ for SC-Cu₂O(100), SC-Cu₂O(110) and SC-Cu₂O(111), respectively, with the Mott–Gurney law (Methods, Supplementary Figs. 12–14 and Supplementary Tables 1 and 2). The anisotropic hole mobilities are reasonable, as the effective hole masses are found to be highly orientation dependent for cubic Cu₂O (refs. ^18,19). The anisotropy in mobility correlates well with the anisotropic PEC performance, with the highest mobility along [111] being more than one order of magnitude higher than those of the other orientations, and the PEC current being a factor of two larger. Remarkably, the SC-Cu₂O hole mobility measured here for samples fabricated in ambient conditions is comparable to those of Cu₂O thin films prepared using methods involving high temperatures, vacuum processing or doping^15,20,21.

The orientation-dependent resistivities were also calculated from the linear ohmic region. The rapid rise of current with increased voltage in the (111)-oriented samples (Fig. 1k) yields the lowest resistivity of 4.7 × 10² Ω cm, which makes [111] the most conductive direction among all. We rule out anisotropic contact resistivities between Au and Cu₂O as a dominant influence on device performance as the contact resistivities of 39.2, 7.25 and 17.4 mΩ cm² for SC-Cu₂O(100), SC-Cu₂O(110) and SC-Cu₂O(111), respectively, determined by a modified transfer length method (Supplementary Fig. 15), did not follow device trends. These results are in good agreement with the findings of an anisotropic resistivity study on Cu₂O nanoparticles²². The in-gap trap states were also studied to help unveil the physical origin of the performance and electronic properties. The trap-filled limit voltage (V_TFL) is located at the transition where the average density of injected holes becomes comparable to the material hole density in thermal equilibrium, and excess injected holes push the quasi-Fermi level towards the valence band, completely ionizing the donor-type hole traps (see comparison in Supplementary Fig. 16). We found the lowest trap density of 2.18 × 10¹⁵ cm⁻³ for the SC-Cu₂O(111) (8.88 × 10¹⁵ cm⁻³ for SC-Cu₂O(110) and 8.22 × 10¹⁵ cm⁻³ for SC-Cu₂O(100)), which points to a less defective electronic structure comparable to those of a wide range of established and emerging photovoltaic-grade materials such as polycrystalline Si and hybrid perovskites^10–12. Combining the results above (summarized in Supplementary Table 4; photoluminescence in Supplementary Fig. 17), we find that anisotropic carrier mobility and trap behaviour play a vital role in orientation-dependent PEC performance.

To further understand the relationship between orientation and performance, the carrier dynamics of SC-Cu₂O films were examined using ultrafast transient reflection spectroscopy across a wide spectral range. Figure 2a–c and Supplementary Fig. 18 show transient reflection spectra and maps in the ultraviolet–visible region for 15-nm-thick Cu₂O thin films of three orientations under a 3.1-eV excitation pump with a fluence of about 110 µJ cm⁻², corresponding to a carrier density of about 6.5 × 10¹⁹ cm³ (see set-up illustration in Supplementary Fig. 18m). An antisymmetric spectral pattern comprising two peaks at about 2.70 and about 2.48 eV was observed for all samples, with an isosbestic point at about 2.58 eV where the transient reflection intensity does not change with pump–probe delay. Another noticeable feature in the response is the shoulder peak at the lower-energy side of the antisymmetric peak. The transient absorption spectrum of the system can be derived from the inverse Hilbert transform of the experimentally measured transient reflection spectrum²³, with the pair of antisymmetric peaks in the transient reflection spectra corresponding to a negative symmetric peak centred at about 2.58 eV in the calculated transient absorption spectra. The features at the lower-energy side correspond to a weak broadband response that exhibits a much more pronounced positive component (Supplementary Fig. 19). The negative signals of the derived transient absorption features represent the bleaching of the transition from the ground state to the excited state (that is, the ground-state bleach), and the positive signals correspond to the photoinduced absorption^24,25. As a result, we attribute the antisymmetric peaks in the transient reflection spectra to the bleach of the transition and assign the low-energy features to the excited-state absorption (ESA) from the first excited state to higher excited states. To accurately collect the weak ESA signal, the transient reflection was recorded using a visible continuum probe with a pre-split reference beam (Methods) under the same excitation conditions (Supplementary Fig. 18). An ESA signal ranging from 530 to 760 nm was observed for all orientations²⁴.Fig. 2

a–c, Transient reflection spectra of 15-nm Cu₂O films with crystal orientations of (100) (a), (110) (b) and (111) (c). Pump wavelength, 400 nm; fluence, 110 µJ cm⁻². d,e, Kinetics of the orientation-dependent photoinduced ground state bleach signals from the reproduced transient absorption spectra (450–490 nm for the (100) orientation, 460–500 nm for the (110) orientation, and 470–510 nm for the (111) orientation; Supplementary Fig. 19) (d), and the photoinduced absorption (600–650 nm; Supplementary Fig. 18) from the transient reflection spectra (e) of SC-Cu₂O. f, Second-stage and third-stage lifetime of SC-Cu₂O films with various thicknesses. PIA, photoinduced absorption; GSB, ground-state bleach.

Figure 2d shows the kinetics of the ground-state bleach of Cu₂O samples with different crystal orientations extracted from the negative peak in the reproduced transient absorption signal (Supplementary Fig. 19). The signal builds up in the pump duration (about 100 fs) and decays with three discernible stages. The first stage is an ultrafast decay on a sub-picosecond timescale, which we ascribe to carrier cooling processes due to carrier–phonon interactions²⁶. This stage is followed by a second-stage short-lived component with a lifetime of 1.27, 1.27 and 1.45 ps for the (100), (110) and (111) samples, respectively. This short-lived decay has been assigned to the deactivation of excitons^24,25. The (111) sample shows the most extended exciton lifetime, consistent with the kinetics of the ESA signal (Fig. 2e), for which the exciton decay lifetime decreases from about 3.6 ps for the (111) sample, to about 2.7 ps for the (110) sample, to about 0.8 ps for the (100) sample extracted from bi-exponential fitting of the ESA signals. The prolonged exciton lifetime for the (111) orientation is further confirmed in Cu₂O samples of increasing thicknesses (Fig. 2f; see Supplementary Figs. 20 and 21 for details), indicating a lower trap density in the (111) sample as suggested in the steady-state results²⁵. Finally, a long-lived component is observed with a lifetime of 20.3, 971.3 and 1,823.4 ps for the (100), (110) and (111) samples, respectively, which is related to carrier recombination processes^24–27 (the third-stage decay for the 15-nm (110) and (111) samples is enlarged in Supplementary Fig. 22 and lifetimes are marked in Fig. 2f). The longest recombination lifetime in the [111] orientation, combined with the highest hole conductivity, leads to the longest hole diffusion length of L_h⁽¹¹¹⁾ = 269.7 nm in the SC-Cu₂O(111) films, much longer than L_h⁽¹¹⁰⁾ = 46.8 nm and L_h⁽¹⁰⁰⁾ = 8.2 nm for the (110) and (100) crystal orientations, respectively. Given an effective polaron mass of 0.69m₀ and 0.99m₀ (m₀ is the mass of the free electron) for heavy holes and electrons, respectively²⁸, and assuming a similar scattering rate for the holes and electrons, the electron diffusion lengths were calculated as L_e⁽¹¹¹⁾ = 188.0 nm, L_e⁽¹¹⁰⁾ = 32.6 nm and L_e⁽¹⁰⁰⁾ = 5.7 nm. For optoelectronic devices in which thick crystals (>500 nm) are required for optimal light harvesting, a short diffusion length leads to limited currents owing to inefficient carrier extraction. The substantially longer carrier diffusion length in SC-Cu₂O(111) thus explains the increased current for that orientation and the resulting anisotropic PEC behaviour. Therefore, we expect Cu₂O(111) thin films would yield the best Cu₂O optoelectronic devices, benefiting from excellent carrier mobility and long carrier diffusion lengths.

To exploit these findings, we fabricated polycrystalline Cu₂O analogues (poly-Cu₂O) with preferential (100), (110) and (111) orientations, with the polycrystalline films allowing greater light in-coupling than the more reflective single-crystal samples to demonstrate maximum device performance (Supplementary Discussion 2 and Supplementary Figs. 23–28). Unlike previous performance-enhancing strategies that involve integrating extra layers or high-temperature processing^4,5, our method was developed to yield specific orientations by simply fine-tuning the pH of the electrolyte; a materials cost estimation for the pilot production of a 1-m² Cu₂O layer (Supplementary Discussion 3 and Supplementary Tables 5–7), together with the scaled deposition (Supplementary Fig. 10), shows that the electrochemical methods used in this work (that is, liquid-phase epitaxy and electrochemical deposition) are both low cost and scalable. The crystallographic information was acquired by X-ray diffraction analysis, which confirms that the samples demonstrate strong orientation preference (Supplementary Fig. 29). Figure 3a–c shows top-view scanning electron micrographs distinguishing the morphology of well-defined crystals in films of poly-Cu₂O(100) (three-sided and four-sided pyramids), poly-Cu₂O(110) (sloping planes) and poly-Cu₂O(111) (triangle planes). To visualize the degree of (100), (110) and (111) orientation of the Cu₂O in the polycrystalline films, inverse pole figures were constructed, on the basis of EBSD data (Supplementary Fig. 30), by plotting the orientations with respect to the sample normal (insets of Fig. 3a–c). The preferential distributions towards the (100), (110) and (111) corners show a substantial tendency for the grains to be aligned with the respective orientations, highlighting the particularly extreme orientation purity for the poly-Cu₂O(111) sample. The EBSD maps corroborate this observation (Supplementary Fig. 31).Fig. 3

a–c, Top-view scanning electron micrographs of Cu₂O polycrystalline thin films with the preferential (100) (a), (110) (b) and (111) (c) orientations and their corresponding inverse pole figures from EBSD. Scale bars, 1 μm. d, Current density–potential (J–E) responses of conventional and polycrystalline Cu₂O photocathodes with preferred orientations under simulated one-sun AM1.5 G illumination. e, Wavelength-dependent IPCE of the poly-Cu₂O(111) photocathode tested with an applied bias of 0.5 V versus the RHE. f, Carrier separation efficiency as a function of applied potential for poly-Cu₂O photocathodes with three crystal orientations. g, Stability test on SC-Cu₂O photocathodes with a 20-nm TiO₂ layer biased at various potentials to each deliver an initial current density of 2 mA cm⁻² (SC-Cu₂O(100) at 0.03 V versus the RHE, SC-Cu₂O(110) at 0.21 V versus the RHE, SC-Cu₂O(111) at 0.48 V versus the RHE) using chronoamperometry techniques. All measurements with SC-Cu₂O were ceased when the photoelectrodes lost 25% of their initial photocurrent. h, Stability test on poly-Cu₂O(111) with a 100-nm TiO₂ protection layer at the fixed bias of 0.5 V versus the RHE in pH 5 buffered electrolyte under simulated one-sun illumination (AM1.5 G). After 120 h of operation with H₂ gas quantification, the initial current density decreased by less than 15%.

The different poly-Cu₂O films, including the conventional electrodeposited polycrystalline film, were then used to prepare photoelectrodes that were tested under air mass 1.5 G (AM1.5 G), 100 mW cm⁻² simulated illumination (Methods). We find that poly-Cu₂O(111) delivers 7 mA cm⁻² current density at 0.5 V versus the RHE, the value at which Cu₂O photocathodes operate in most standalone solar water-splitting devices^3,5,29. This yields a photocurrent density that is a factor of seven higher than that in poly-Cu₂O(100) (Fig. 3d). All devices in the polycrystalline form achieved a higher current density than their single-crystal analogues owing to the rougher surfaces of polycrystalline films allowing better light in-coupling. The pH-dependent PEC performance was recorded in acidic, neutral and alkaline electrolytes showing slight variation (Supplementary Fig. 32). The wavelength-dependent incident-photon-to-current efficiencies (IPCEs) of poly-Cu₂O(111) stand out as the highest among all crystal orientations, with an average efficiency exceeding 50% in the bandgap at 0.5 V versus the RHE (Fig. 3e and Supplementary Fig. 33). The integrated product of the IPCE and the photon flux of AM1.5 G yields current densities of 1.46, 0.66 and 6.9 mA cm⁻², which correlate well with the PEC values. Furthermore, the quantification of charge separation efficiencies, η_sep, is detailed in Supplementary Discussion 4 (also see Methods and Supplementary Figs. 34 and 35). Using Eu³⁺ as an effective electron acceptor, the onset potential of poly-Cu₂O(111) indicates a built-in photovoltage of beyond 1 V (Fig. 3f). Most importantly, the exceedingly high η_sep observed in poly-Cu₂O(111) highlights a substantial advantage of electronic properties along the [111] orientation, excluding other factors concerning light in-coupling and electrocatalysis.

Finally, we investigated how the orientation dependence translates to device operational stability. Comparative stability tests were carried out first on the single-crystal (SC-Cu₂O) photocathodes with all three orientations by biasing and maintaining the photocathodes at certain potentials to deliver 2 mA cm⁻² until 25% of the initial current density was lost. SC-Cu₂O(111) maintained the longest stable operation (>55 h) whereas the SC-Cu₂O(110) device degraded most quickly (7 h; Fig. 3g), thus establishing that the (111) devices are also the most stable. To assess the longer-term stability with the highest-performing polycrystalline devices, we prepared poly-Cu₂O photocathodes with 100-nm TiO₂ overlayers tested in pH 5 buffered electrolyte under simulated one-sun illumination and continuous stirring. The poly-Cu₂O(111) photocathode delivered extremely stable operation, losing only 15% of its initial photocurrent after 120 h (Fig. 3h). The faradaic efficiency for the hydrogen evolution reaction remained at 100% with a slight decrease at the end, indicating that electrolyte penetration could have occurred (see Supplementary Discussion 5). Indeed, the extended stability test on the examined poly-Cu₂O(111) revealed an accelerated current decay in the first 20 h followed by protection layer detachment and degradation of current (Supplementary Fig. 36a). Nonetheless, compared to the other two poly-Cu₂O photocathodes (Supplementary Fig. 36b,c), poly-Cu₂O(111) demonstrates notably superior stability. Although we acknowledge the crucial role of the TiO₂ protection layer in stabilizing the photocathodes (Supplementary Fig. 37), it is apparent that the facets exert a substantial impact on the anisotropic stability. The crystal orientations and terminating facets do not inherently have to align, but this particular situation holds true for the poly-Cu₂O(111) photocathode, for which a high ratio of (111) facets are exposed as seen as in-plane triangles in a cubic system³⁰. Stability tests were carried out on fine-tuned bare Cu₂O thin films, simultaneously showing the (100) and (111) facets (Supplementary Fig. 38a). Under identical testing conditions, the (100) facets experienced considerably greater damage, whereas the (111) facets maintained their flat profile (Supplementary Fig. 38b,c). The exceptional PEC stability aligns with (111) facets being reported to exhibit the lowest surface energy^31,32. Nevertheless, a comprehensive and meticulous investigation is desirable to consider other potential factors with thorough materials and spectroscopic characterization. In summary, our poly-Cu₂O(111) photocathodes concurrently demonstrate remarkable PEC performance and stability due to inhibited bulk recombination along the [111] direction and extensive exposure of (111) facets. The performance of 7 mA cm⁻² current density at 0.5 V versus the RHE represents a 75% improvement compared with the state-of-the-art electrodeposited devices and 15% compared with nanowire systems even although nanowires have intrinsic light management benefits. These important technological advances validate our findings of the anisotropic optoelectronic and PEC properties in Cu₂O.

The powerful combination of unique SC-Cu₂O thin films with selected orientations, customized ultrafast spectroscopy and device characterization has resulted in a step change in understanding anisotropic bulk carrier transport properties. To exploit the findings of excellent carrier mobility and carrier diffusion length along [111] in Cu₂O, poly-Cu₂O photocathodes with an extraordinarily pure (111) orientation (optimizing bulk recombination, which otherwise limits performance) and (111) terminating facets (optimizing surface effects, which otherwise limit stability) were realized by a simple and low-cost method, delivering a PEC current density of 7 mA cm⁻² at 0.5 V versus the RHE and more than 120 h of stable operation.

These results enable a new level of precision in future design of high-performance Cu₂O devices. For example, nanowire Cu₂O photoelectrodes with an axis parallel to the (111) crystal orientation and optimal radius, considering efficient light absorption and charge collection based on specific carrier diffusion lengths, should be targeted. Further optimizing p–n junctions, applying textured substrates, passivation approaches and photonics, while retaining both bulk and surface directionality properties, will be critical steps for further improvements of performance and stability of these device structures. The demonstration of understanding and exploitation of the anisotropy in thin films presented here provides a widely applicable strategy to cope with the poor bulk carrier transport nature of oxides in their applications in photovoltaics, transistors and detectors, in addition to solar fuels.

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-024-07273-8.