Critical and controversial issues pertaining to the growth and properties of Cu₂O in the context of energy conversion Open Access

Cuprous oxide (Cu₂O) is a p-type metal-oxide semiconductor that has a fundamental, direct energy bandgap of 2.1 eV and a cubic crystal structure belonging to the Pn3m crystallographic space group with a lattice constant of a = 4.2696 Å. The native p-type conductivity of Cu₂O is related to the occurrence of copper vacancies, V_Cu, that give rise to acceptor-like states located at E_A ∼0.2 eV above the valence band maximum.¹ 

Cu₂O has been suggested to be suitable as a solar cell absorber for a long time due to the fact that it has a direct energy gap but also an absorption coefficient that can reach α = 10⁵ in the visible.² However, so far, device efficiencies have been limited to less than 10%, despite the fact that the theoretical efficiency of Cu₂O solar cells is ∼20%.³ 

A Cu₂O p-n junction solar cell with an efficiency of ∼2% was fabricated by Elfadill et al.⁴ using Cl-doped n-type Cu₂O and Na-doped p-type Cu₂O, while a Cu₂O heterojunction solar cell (HJSC), with an efficiency of 3.4% and a very small cell area of 8 × 10⁻³ cm², was fabricated by Ravindra et al.⁵ Others, like Markose et al.,⁶ fabricated an ITO/B:Cu₂O/SiO_x/n-Si/Ag HJSC that exhibited an open-circuit voltage of 370 mV, a short-circuit current density of 36.5 mA/cm², and an efficiency of 5.4%, but this was obtained on Si. Similarly, Liu et al.⁷ employed a variety of interfacial engineering and light management strategies to push the efficiency limit of Cu₂O/Si solar cells and obtained a power conversion efficiency of 9.54%. More importantly, an all-oxide HJSC based on p-type Cu₂O and n-type Ga₂O₃ was fabricated by Shibasaki et al.⁸ on glass and displayed an efficiency of 8.4% when the Cu₂O contained minimal Cu and CuO.

In contrast to solar cells, Cu₂O/CuO layers have been shown to have better photocatalytic performance compared to Cu₂O or CuO single-component materials.⁹ Cu₂O is, therefore, an active topic of investigation for the fabrication of solar cells and photocatalysis, i.e., water splitting,¹⁰ but also for CO₂ reduction,¹¹ which is important in view of the escalating climate crisis.^10,12 Furthermore, Cu₂O is interesting from a fundamental point of view, as excitons were observed for the first time in Cu₂O by Gross¹³ in 1956, after which it has been used as an archetype for understanding the novel properties of excitons.^14–16 Rydberg excitons with principal quantum numbers as large as n = 25 and giant wavefunction extensions in excess of 2 µm were observed by Kazimierczuk et al. in natural gem crystals of Cu₂O.^17,18 Excitons with principal quantum numbers up to n = 3 have been observed in other semiconductors such as GaAs,¹⁹ but excitons with principal quantum numbers as large as n = 25 have only been observed in Cu₂O. Very recently, Rydberg exciton-polaritons were also detected in a SiO₂/Ta₂O₅/Cu₂O/Ta₂O₅/SiO₂ Fabry–Pérot cavity in which the Cu₂O layer was cut from a naturally occurring crystal and attached with epoxy to Ta₂O₅/SiO₂.²⁰ It is desirable then to obtain high purity and good crystal quality Cu₂O layers, which are important not just from a fundamental but also a technological point of view.

Cu₂O has been deposited using many different methods such as molecular beam epitaxy (MBE), atomic layer deposition (ALD),²¹ pulsed laser deposition (PLD), electrodeposition (ELD),²² aerosol assisted chemical vapor deposition (AACVD),²³ successive ionic layer adsorption and reaction (SILAR),²⁴ sputtering,^8,25 or the thermal oxidation of copper.²⁶ The hole density of un-doped (u/d) Cu₂O is of the order of p ∼ 10¹⁵ cm⁻³,²⁷ and room temperature hole mobilities of µ_p = 100 cm²/Vs have been achieved,²⁸ while the mobility of holes in Cu₂O at 4.2 K has been shown to reach quite large values of μ_p = 1.8 × 10⁵ cm²/Vs.²⁹ In general, p-type doping of Cu₂O has been carried out using Si,³⁰ Mg,³¹ and N,³² while n-type doping is possible with F, Cl, and Br³³ but also with Al.³⁴ 

However, despite ongoing efforts, single crystal epitaxial Cu₂O is not easy to obtain even by MBE on MgO (100), which has a lattice mismatch of ∼1% with Cu₂O. In other words, the epitaxy of Cu₂O on MgO (100) is not trivial and differs from simple cube-on-cube epitaxy, thereby giving Cu₂O epi-layers with multiple orientations.³⁵ Consequently, it is necessary to anneal the Cu₂O layers in order to improve their crystallinity. It is useful to mention at this point that single crystal Cu₂O has been obtained in the past by Toth et al.²⁶ back in the 1960s via the thermal oxidation of metallic Cu foils between 1020 and 1050 °C, which resulted in polycrystalline Cu₂O that was annealed at even higher temperatures. This promoted recrystallization and the formation of single-crystals, but Cu₂O reacts with O₂ according to 2Cu₂O + O₂ → 4CuO at the surface or grain boundaries. The suppression of this reaction is critical in obtaining high efficiency solar cells, like the p-type Cu₂O/n-type Ga₂O₃ solar cell fabricated by reactive sputtering on glass by Shibasaki et al.⁸ This can be achieved by reducing the partial pressure of oxygen during cooling in order to prevent a transition across the phase boundary between Cu₂O and CuO, in accordance with Schmidt-Whitley et al.³⁶ However, the process of annealing Cu₂O layers at elevated temperatures is complicated due to thermal inter-diffusion and reactions that will occur between the Cu₂O and the underlying substrate of choice. It has been shown that sapphire, i.e., Al₂O₃, can be used as a barrier against the thermal diffusion of copper up to 750 °C.^37,38 Wagner et al.³⁹ obtained Cu₂O on a-Al₂O₃ by Vapor Phase Epitaxy (VPE) via melting Cu at 1080 °C. Others, like Ottosson and Carlsson,⁴⁰ have grown Cu₂O on r-Al₂O₃, but the Cu₂O will react with Al₂O₃ according to the reaction Cu₂O + Al₂O₃ → CuAlO₂ under N₂ at temperatures higher than 700 °C.⁴¹ Lower temperatures between 450 and 550 °C have been shown to yield lower crystallinity CuAlO₂,⁴² which is a p-type transparent conducting oxide (TCO) with an energy bandgap of 3.4 eV.⁴³ It is then necessary to anneal the Cu₂O layers on Al₂O₃ at intermediate temperatures to prevent thermal inter-diffusion and reactions.

Here we have deposited Cu₂O on m-, r-, and a-Al₂O₃ by reactive sputtering of Cu using Ar:O₂ with different ratios of 90:10, 75:25, or 50:50% v/v, followed by annealing under optimum conditions at 500 °C under Ar:H₂ at 10⁻¹ mbar. In the past, only a few, like Unutulmazsoy et al.,⁴⁴ annealed Cu₂O using Ar containing 5% v/v H₂ at 10⁻⁴ mbar and 300 °C. Here we have carried out a systematic investigation into the optimization of the annealing conditions over a wider temperature range and flows of H₂ in order to suppress the oxidation and reduction of the Cu₂O layers and subsequently probe the electronic band structure and gap states of this novel p-type metal-oxide semiconductor by ultrafast pump–probe spectroscopy (UPPS) in order to gain an understanding of the generation and recombination mechanisms that are important from an applied and fundamental point of view. The Cu₂O layers obtained in this fashion exhibited a detailed spectral structure and distinct peaks corresponding to the indigo, blue, and yellow direct gap transitions of Cu₂O as observed by UPPS at room temperature, but we also observed a transition related to the occurrence of states ∼0.4 eV below the conduction band minimum of Cu₂O. We discuss the controversial origin of these states, which are usually attributed to oxygen vacancy donor-like states, and suggest that their origin may instead be related to the formation of CuO/Cu₂O nanostructures.

Square samples (10 × 10 mm²) were cut from m-, r-, and a-Al₂O₃ and n-type Si (001), after which they were cleaned sequentially in trichloroethylene, methanol, and acetone and rinsed in isopropanol and de-ionized water at 20 °C. Subsequently, they were dried with nitrogen, followed by a dehydration bake at 120 °C. A Cu₂O layer was deposited by reactive sputtering of a Cu (99.999%) target with Ar:O₂ at 10⁻² mbar. The thickness of the Cu₂O layers was either 100 nm or 300 nm, and the Ar:O₂ ratio used during deposition was 90:10, 75:25, or 50:50% v/v. The Cu₂O layers were deposited on m-, r-, and a-Al₂O₃ at room temperature and were subsequently annealed under optimum conditions in a 1 in. hot wall, low pressure chemical vapor deposition reactor capable of reaching 1100 °C, which was fed by a manifold consisting of four mass flow controllers connected to Ar, NH₃, O₂, and H₂. The reactor was initially purged for 10 min, and then the temperature was ramped up to 500 °C at 30 °C/min. Upon reaching 500 °C, the Cu₂O layers were annealed for 1 h, after which cooling took place. The samples were removed only when the temperature fell below 100 °C. A constant flow of 50 ml/min of Ar and 50 ml/min of H₂ was maintained throughout the entire process.

The surface morphology of the Cu₂O thin films was investigated using JSM-7610F+ Field-Emission Scanning Electron Microscopy (FESEM). The samples were coated with carbon film using a JEOL-4X vacuum evaporator, and the thickness of the carbon coating did not exceed 200 Å. The composition of the Cu₂O was measured using an Oxford (AZtec Energy Advanced) Energy Dispersive X-ray Spectroscopy (EDS) system attached to the FESEM.

The Raman spectra were recorded in the backscattering geometry using a DILOR spectrometer equipped with an optical microscope and 100× objective along with an Ar+ laser operating at 514.532 and 487.986 nm that was used for excitation with a power of about 4.5 and 7.5 mW on sample for the green (514.532 nm) and blue (487.986 nm) laser lines, respectively. Energy calibration of the spectra was performed utilizing a neon lamp.

The structural properties of the Cu₂O films were also investigated by X-Ray Diffraction (XRD) in the Bragg–Brentano geometry using a two-cycle Rigaku Ultima+ powder x-ray diffractometer with CuKa radiation (λ = 1.540 56 Å) operating at 40 kV and 30 mA.

Finally, the properties of the Cu₂O layers were investigated by measuring the time evolution of the differential transmission (dT/T) on a ps time scale by ultrafast pump–probe spectroscopy (UPPS) using a pump of λ_PU = 400 or 260 nm and a probe that was varied between λ_PR = 450 and 750 nm. The measurements were carried out in a typical pump–probe optical setup, as described in detail elsewhere.⁴⁵ 

It is instructive to begin by describing the optimization of the growth and annealing conditions of Cu₂O. Initially, Cu₂O layers were deposited on n-type Si (001) and soda lime glass (SLG) by reactive sputtering of Cu under Ar:O₂ with a ratio of 90:10, 75:25, or 50:50% v/v and annealed at temperatures ranging from 300 to 600 °C under a flow of 100 ml/min Ar for 60 min at 10⁻¹ mbar. All of the Cu₂O layers obtained in this way exhibited small but clear and well resolved peaks in the XRD corresponding to both CuO and Cu₂O. Consequently, an identical set of Cu₂O layers was prepared and annealed between 300 and 600 °C under Ar:H₂. The content of H₂ was 10, 50, or 100% v/v at 10⁻¹ mbar, keeping all else equal. We find that the optimum annealing temperature for the Cu₂O layers is 500 °C. The Cu₂O layers exhibited clear and well resolved peaks belonging to the cubic crystal structure of Cu₂O, as shown in Fig. 1(a), but the addition of 10% H₂ was not adequate to suppress the oxidation of Cu₂O, i.e., 2Cu₂O + O₂ → 4CuO, as we also observed peaks belonging to the monoclinic crystal structure of CuO in Fig. 1(a). On the other hand, 100% H₂ leads to the reduction of Cu₂O and CuO to metallic Cu according to Cu₂O + H₂ → 2Cu + H₂O. In contrast, 50% H₂ was sufficient to prevent the oxidation as well as the reduction of Cu₂O, which has a cubic crystal structure and corresponding unit cell shown in Fig. 1(b). These findings are consistent with Kim et al.,⁴⁶ who showed that a large flow of H₂ will reduce CuO into metallic Cu without the formation of intermediate Cu₄O₃ or Cu₂O. However, only a few have investigated the effect of H₂ on the reduction of Cu₂O layers, such as Unutulmazsoy et al.,⁴⁴ who annealed Cu₂O using Ar containing 5% v/v H₂ at 10⁻⁴ mbar and 300 °C and argued that intersecting grain boundaries and the porosity of Cu₂O provide diffusion paths and traps for hydrogen. Consequently, the higher grain boundary density of the thinner Cu₂O layers results in a faster reduction that starts once a certain density of accumulated oxygen vacancies is reached at the periphery of the Cu₂O grains. Hence, a careful adjustment of the reducing conditions and flow of H₂ needs to be carried out when annealing Cu₂O.

After optimization, Cu₂O layers with a thickness of 100 and 300 nm were deposited on m-, r-, and a-Al₂O₃ by reactive sputtering under Ar:O₂ (90:10, 75:25, 50:50% v/v), which were annealed under 50 ml/min Ar: 50 ml/min H₂ at 500 °C. The Cu₂O layers on m-, r-, and a-Al₂O₃ were semi-transparent, had a light-yellow color, and consist of grains with a size of 89.62 ± 7.87 nm, as shown by the Scanning Electron Microscopy (SEM) image of the Cu₂O and corresponding histogram shown as an inset in Fig. 2(a). Evidently, the Cu₂O layers are nanostructured, and the protrusions on the surface with a diameter of d ∼1 µm were found to be slightly richer in Cu by ∼5 at. %. We find that the Cu₂O layers consist of Cu and O by EDS; Al was also detected due to the underlying Al₂O₃, as the information depth of EDS is about 1.5–2 µm.^47,48 The structural properties and composition of the Cu₂O layers were investigated further by Raman spectroscopy. The Raman spectra of the Cu₂O layers grown on m-, r-, and a-Al₂O₃, which were obtained with different Ar:O₂ contents, are shown in Fig. 2(b). The unit cell of Cu₂O contains six atoms, resulting in 18 phonon modes, 15 of which are optical lattice vibrations and 3 are acoustic vibrational modes. The vibrational mode symmetries at the center of the Brillouin zone (k = 0) are^49–51, A_2u ⊕ E_u ⊕ T_2g ⊕ 3T_1u ⊕ T_2u. A, E, and T symmetries correspond to one-, two-, and threefold-degenerate phonons, respectively. All symmetries represent optical phonons except for T_1u, which corresponds to the three acoustic vibrational modes. Phonons having T_2g symmetry are the only Raman active vibrational modes of Cu₂O, while the two remaining T_1u symmetry modes are infrared active optical modes, and the rest, i.e., phonons having A_2u, E_u, and T_2u symmetry, are non-active Raman and infrared modes, i.e., silent modes.⁵¹ According to group theory, only the peak due to the threefold-degenerate T_2g lattice vibrational mode should appear in the Raman spectra of a natural Cu₂O crystal. However, more peaks are usually observed in the Raman spectra of Cu₂O, despite the fact that they are either silent or infrared active phonon modes. This is due to^49–51 (a) the excitation conditions and resonance with excitonic states of Cu₂O, (b) the polarization and scattering geometry used, and (c) the surface treatment. The non-stoichiometry of Cu₂O also plays a key role in the emergence of the non-active Raman lattice modes in the spectra, inducing several intrinsic defects, i.e., anti-sites, interstitials, and vacancies in the lattice.^49,50 Point imperfections reduce the local symmetry of a perfect Cu₂O crystal, thereby relaxing the selection rules and activating Raman-forbidden vibrational modes.

The Cu₂O layers on m-, r-, and a-Al₂O₃ peaks due to the following vibrational modes are detected: 93.2 cm⁻¹ (T_2u),^50,51 108 cm⁻¹ (E_u),^50,51 148.6 cm⁻¹ [T_1u, Transverse Optical (TO) and Longitudinal Optical (LO) phonons],^50,51 217.5 cm⁻¹ (2 E_u),^50,52 516.7 cm⁻¹ (T_2g),^50–52 and 626.9 and 643.5 cm⁻¹ (T_1u, TO and LO phonons, respectively).^50,51 A multi-phonon Raman scattering mode appears at 413.7 cm⁻¹, consistent with the literature,^52,53 while at 196 cm⁻¹, a peak is observed due to local vibrations of Cu on O-sites with T_d point symmetry.^50,53 Moreover, the A_g Raman active mode of CuO appears at 302.1 cm⁻¹.^54–58 As evident, the Cu₂O phase is dominant, and the Raman active mode of CuO that occurs via the oxidation of Cu₂O is not observable. In addition, we find no evidence for the formation of CuAlO₂, suggesting that there is no reaction between Cu₂O and Al₂O₃. Moreover, we do not find any significant dependence of the properties of the Cu₂O layers on the different orientations, m-, r-, and a-, of the Al₂O₃ substrate. Nevertheless, the T_1u peak of Cu₂O at ∼150 cm⁻¹, which is detected in the Raman spectra of the Cu₂O layers deposited on Si, gains intensity as the O₂ content in the Ar:O₂ mixture increases and is attributed to the formation of intrinsic point imperfections in the crystal lattice of Cu₂O (see the supplementary material for more details).

The temporal evolution of the differential transmission through the Cu₂O layers was measured by UPPS. Before considering in detail the spectra obtained from the Cu₂O layers it is useful to describe the generation and recombination of photogenerated electron-holes during UPPS in direct energy bandgap semiconductors.

Initially, the pump, i.e., a short pulse of energetic photons with energy of E_PU = 3.1 or 4.77 eV corresponding to λ_PU = 400 or 260 nm, will result in the excitation of electrons from the valence band into the conduction band, as shown in Fig. 3. The photogenerated electrons will initially occupy empty states residing high above the conduction band minimum. These hot electrons will gradually lose energy via electron–phonon or electron–electron interactions and will occupy lower energy states closer to the conduction band minimum (E_C) as well as empty states related to crystallographic imperfections (E_I) that are located energetically in the energy gap of Cu₂O. A separate light beam is used to probe the temporal evolution of the occupancy of the above-mentioned states located at different energies by varying the wavelength between λ_PR = 450 and 750 nm and measuring the change in transmission or differential transmission (dT/T). If a state is occupied by an electron, then the photon with a specific λ_PR will not be absorbed, resulting in a positive increase in differential transmission at the specific wavelength corresponding to a particular energy state. In contrast, if a state is not occupied by a photoexcited electron, then a photon with a specific energy will be absorbed, thereby reducing transmission at a specific λ_PR. In essence, one may find the fundamental direct energy bandgap(s) of a semiconductor as well as the energetic location of states lying in the energy gap that are related to crystal imperfections, similar to deep level transient spectroscopy.

The variation of dT/T vs τ and λ_PR for the 100 nm Cu₂O layers on m-Al₂O₃ that were obtained using Ar:O₂ (90:10% v/v) is shown in Fig. 4(a). One may observe the occurrence of two major peaks corresponding to a positive and negative maximum in dT/T at λ_PR = 450 and 487 nm, i.e., 2.75 and 2.54 eV, respectively. We also observe these by increasing the thickness of the Cu₂O layer from 100 to 300 nm, as shown in Fig. 4(b) but the maximum at λ_PR = 450 nm is not so pronounced compared to that in Fig. 4(a).

Similarly, the variation of dT/T vs τ and λ_PR for the 100 nm Cu₂O layers on m-Al₂O₃ that were obtained using Ar:O₂ 75:25% v/v is shown in Fig. 5(a). One may observe the occurrence of the peaks that appear in Figs. 4(a) and 4(b), but also the emergence of a maximum at λ_PR = 570 nm and a broad one at λ_PR = 687 nm, i.e., 2.17 and 1.8 eV, respectively.

It should be pointed out that the maxima at 1.84 and 2.75 eV correspond to positive dT/T, indicating that these states are in fact occupied by electrons. In contrast, the maxima at 2.17 and 2.62 eV correspond to negative dT/T, implying absorption, and these conduction band minima states are in fact empty after excitation and the generation of excess electrons and holes. It appears then that photogenerated electrons occupy the highest conduction band and subsequently move into states related to point-imperfections that are energetically located in the energy gap of Cu₂O. No changes occurred by increasing the excitation energy from 1 to 10 µJ, as shown by the inset in Fig. 5(a). In addition, no significant changes occurred by increasing the thickness of the Cu₂O layer from 100 to 300 nm, as shown in Fig. 5(b).

The spectral structure obtained from Cu₂O on m-Al₂O₃ using Ar:O₂ 50:50% v/v is also similar and shown as an inset in Fig. 5(b). We find that the spectral structure obtained by UPPS is not dependent on the thickness of the Cu₂O and is also the same for m, r-, and a-Al₂O₃.

The four peaks observed at λ_PR = 450, 487, 570, and 687 nm correspond to 2.75, 2.54, 2.17, and 1.84 eV, respectively. In order to understand the origin of these optical transitions, it is necessary to consider the electronic structure of Cu₂O, as shown in Fig. 6(a). The top of the valence band and bottom of the conduction band are derived from states belonging to the same ion, so they have the same parity, and dipole transitions are forbidden. The two upmost valence bands, with Γ₇⁺ and Γ₈⁺ symmetry, are related to Cu_3d electrons. The lowest conduction band with Γ₆⁺ symmetry is related to Cu_4s electrons, and the next higher conduction band with Γ₈⁻ symmetry is related to O_2p electrons.

Excitation from the two valence bands to the two conduction bands results in a four-exciton series with energies of 2.17, 2.30, 2.62, and 2.75 eV,⁵⁹ which are usually observed at low temperatures and described as the yellow, green, blue, and indigo transitions, respectively, as shown in Fig. 6(a). Evidently, the peaks observed at 2.17, 2.54, and 2.75 eV are very close indeed to the corresponding yellow, blue, and indigo transitions of Cu₂O. These transitions are not related to transitions of photogenerated electrons from the conduction band minima into the continuum at higher energies inside the conduction band following excitation, as suggested by Shenje et al.;⁶⁰ otherwise, we would have observed a host of other energetic transitions. Similarly, they are not related to the excitation of electrons occupying states below the top of the valence E_V into empty holes near the top of the E_V. The peaks observed at 2.17, 2.54, and 2.75 eV are related to the corresponding yellow, blue, and indigo direct gap transitions of Cu₂O. Their observation at room temperature is due to the careful optimization of the annealing conditions under H₂, and the direct gap transitions observed suggest that the Cu₂O is not coherently strained to the underlying m-, r-, and a-Al₂O₃. This may be understood by considering that m-Al₂O₃ is ideally suited for the growth of cubic crystals such as Cu₂O as it has an oxygen terminated surface with a tetragonal, two-dimensional unit cell of 4.34 × 4.76 Ų, which is expected to result in Cu₂O under a biaxial tensile strain of 1.6% and 11.5% if epitaxial growth occurs, considering that the lattice constant of Cu₂O is a = 4.2696 Å. The tensile strain is expected to be even larger in the case of epitaxial Cu₂O on r-Al₂O₃, which also has a surface consisting of oxygen atoms arranged in a tetragonal two-dimensional unit cell of 5.12 × 4.76 Ų, and this in turn is expected to change the energy gap of Cu₂O, which is strongly dependent on strain.⁶¹ However, we do not observe any dependence of the direct gap transitions of the Cu₂O layers on orientation, i.e., m-, r-, and a-Al₂O₃, suggesting that the Cu₂O layers are bulk-relaxed, not coherently strained, due to the fact that they were deposited by reactive sputtering at room temperature.

In addition to the above direct gap transitions, we also observed a broad peak with a maximum at 1.84 eV in the Cu₂O layers deposited under Ar:O₂ 75:25 and 50:50% v/v. In other words, the broad peak appears to be dependent on the content of oxygen in the Cu₂O layers during deposition. It is reasonable then to suggest that it is related to a local density of states that is energetically located between ∼0.3 and 0.5 eV below the conduction band minimum inside the energy gap. This is consistent and in very good agreement with those who have proposed that oxygen vacancies, i.e., V_O, behave as donor-like states that reside ∼0.4 eV below the conduction band minimum at 2.2 eV.^18,62–66 The occurrence of V_O donor-like states that are positively charged results in surface band bending and depletion, as shown in Fig. 6(b). In fact, it has been suggested that the Fermi level (E_F) of Cu₂O is actually pinned at surface states related to V_O that reside energetically at ∼0.4 eV below the conduction band minimum. The energetic position of the Fermi level with respect to the conduction band edge at the surface and in the bulk governs the overall band bending. In the bulk, a flat band condition exists, provided that the surface depletion is smaller than the thickness of the Cu₂O layer. The surface depletion width is of the order of a few tens of nm’s in u/d Cu₂O as the static dielectric constant of Cu₂O is rather low, i.e., ε_R = 4. For completeness, it is instructive to mention that the persistent photoconductivity effect in Cu₂O is due to the interplay between negatively charged copper vacancies $VCu−$ and positively charged oxygen vacancies $VO+$ or $VO++$ that forms pairs, e.g., $VCu−$ and $VO+$ or $VCu−$ and $VO++$⁠, due to electrostatic interaction. According to Mittiga et al.,⁶⁷  $VO++$ reside energetically in the upper half of the energy bandgap of Cu₂O.

Evidently, many have proposed that V_O donor-like states that are positively charged influence the electrical and optical properties of Cu₂O,^18,62–66 but according to the theoretical calculations of Scanlon et al.,⁶⁸ V_O is not charged and is energetically located close to the valence band, so this is still a matter of controversy. It is then necessary to find alternative explanations for the origin of the maximum at 1.8 eV, which may be attributed to a transition between acceptor-like states related to V_Cu that reside energetically ∼0.2 eV above the valence band maximum and states in the conduction band. However, the V_Cu acts as acceptor-like states, not donor-like states, and does not lead to surface depletion or downward band bending as observed experimentally.^18,62–66 Hence, one cannot rule out the existence of states residing energetically ∼0.4 eV below the conduction band minima. We suggest that these states corresponding to the maximum at 1.8 eV may be related to the formation of Cu₂O/CuO nanostructures at the surface and/or grain boundaries. Despite the fact that we have not observed a significant content of CuO in the Raman spectra of Fig. 2, the Cu₂O layers may well contain traces of CuO leading to the formation of Cu₂O/CuO nanostructures that will act in essence as traps considering that the Cu₂O/CuO heterojunction has a straddled (type I) band line-up and conduction band discontinuity ΔE_C(Cu₂O/CuO) ∼1.0 eV as shown in Fig. 6(c). The size of these Cu₂O/CuO nanostructures is going to be of the order of a few tens of nm’s considering that the Cu₂O is nanostructured, i.e., consists of grains with average sizes of less than 100 nm. Consequently, one may expect quantization to occur inside the Cu₂O/CuO nanostructures, which in turn will give rise to a quantum confined level that will act as a trap. Considering that the size of the Cu₂O/CuO nanostructures will also vary, this in turn will lead to an energetic distribution of traps commensurate with the fact that we observed a broad transition with a maximum at 1.8 eV. However, one must also keep in mind that CuO has a monoclinic crystal structure with a lattice constant of 4.68 Å that is larger than the lattice constant of cubic Cu₂O, which is 4.2696 Å, and as such will give rise to a plethora of states related to crystallographic defects at the Cu₂O/CuO interface. The poor properties and quality of the CuO/Cu₂O interface are well known and clearly manifested when trying to grow Cu₂O via thermal oxidation of a Cu foil under O₂. Initially, the Cu reacts with O₂, leading to the formation of Cu₂O, but the latter reacts again with O₂, leading to the formation of a CuO layer on top of Cu₂O that will separate after reaching a certain thickness due to the lattice mismatch between the CuO and Cu₂O, which inevitably leads to a very high density of crystallographic defects at the CuO/Cu₂O heterojunction interface that may be responsible for the distribution of states ∼0.4 eV below the conduction band minima as shown in Fig. 6(c). This hypothesis is consistent with Živković and de Leeuw,⁶⁹ who showed that both oxygen and copper point imperfections give rise to states energetically located below the conduction band minimum inside the energy gap of CuO. It is also consistent with the fact that the all-oxide p-n junction solar cell of Shibasaki et al.⁸ displayed the highest efficiency of 8.4% when the Cu₂O contained minimal Cu and CuO.

The formation of CuO on top of Cu₂O at the surface is also consistent with the observations of surface depletion and downward band bending, which occur due to the band line-up and difference in the energy gaps as opposed to the electrostatic band-bending induced by positively charged donors.

Considering the above-mentioned arguments, we suggest that the formation of CuO/Cu₂O nanostructures in Cu₂O layers is especially important as Cu₂O layers react over time with ambient O₂, which means that it is necessary to passivate their surface with a suitable oxide that will be transparent like, for instance, the n-type Ga₂O₃ used by Shibasaki et al.⁸ This is of paramount importance not just for devices but also for the observation of Rydberg excitons in thin layers of Cu₂O as opposed to natural occurring crystals. In contrast to solar cells, Cu₂O/CuO layers have better photocatalytic performance compared to Cu₂O or CuO single-component materials.^9,70–73 This might seem perplexing at first, but it can be explained by an increase in the surface area and porosity of the Cu₂O/CuO structure due to the formation of voids at the Cu₂O/CuO interface, which allows the infiltration of liquids. This is very similar to the case of quantum dot sensitized solar cells, in which CdS/CdSe quantum dots are deposited on top of a scaffold of TiO₂ particles with high porosity and surface area that is critical for the infiltration of the liquid electrolytes.

We suggest in closing that suppressing the formation of crystallographic imperfections due to the oxidation of Cu₂O is critical in exploiting the properties of this novel metal-oxide semiconductor from a fundamental but also applied point of view in the context of energy conversion.

Cu₂O has been deposited on m-, r-, and a-Al₂O₃ by reactive sputtering of Cu using Ar:O₂ (90:10, 75:25, or 50:50% v/v), after which it was annealed under optimum conditions at 500 °C under Ar:H₂ (50:50% v/v) at 10⁻¹ mbar. The Cu₂O layers have a cubic crystal structure, consist of bulk-relaxed grains, and do not contain CuO, as observed by Raman spectroscopy. The Cu₂O layers exhibited a detailed spectral structure and distinct peaks at 2.75, 2.54, and 2.17 eV corresponding to the indigo, blue, and yellow direct gap transitions of Cu₂O as observed by ultrafast pump–probe spectroscopy at room temperature. However, we also observed a transition at 1.8 eV, which is related to a local density of states that is energetically located between ∼0.3 and 0.5 eV below the conduction band minimum inside the energy gap. This is usually attributed to positively charged donor-like oxygen vacancies, but alternative explanations must also be considered, like the formation of CuO/Cu₂O nanostructures at the surface and grain boundaries of the Cu₂O layers. It is then necessary to suppress completely the formation of CuO/Cu₂O nanostructures, which we identify as one of the most important challenges in attaining high efficiency solar cells, as well as the observation of Rydberg excitons in Cu₂O layers, not just natural occurring crystals.

The Raman spectra of the Cu2O layers deposited on n-type Si (001) with different oxygen contents are described for completeness in the supplementary material so that one may compare with the Cu2O layers deposited on m-, r-, and a-Al2O3.

The authors have no conflicts to disclose.

E. Prountzou: Raman, EDS, SEM, XRD, Writing. A. Ioannou: Growth/Annealing, XRD Investigation. D. Sapalidis Raman, EDS, SEM, XRD E. Pavlidou: SEM, EDS. M. Katsikini: Supervision of E. Prountzou and D. Sapalidis. A. Othonos: UPPS. M. Zervos: Conceptualization, Growth/Annealing, Supervision, Writing (Lead).

Eleni Prountzou: Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Visualization (equal); Writing – original draft (supporting). Andreas Ioannou: Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Visualization (equal). Dimitrios Sapalidis: Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Visualization (equal). Eleni Pavlidou: Formal analysis (equal); Investigation (equal); Methodology (equal). Maria Katsikini: Supervision (equal). Andreas Othonos: Investigation (equal); Methodology (equal); Validation (equal); Visualization (equal). Matthew Zervos: Conceptualization (lead); Data curation (equal); Formal analysis (lead); Investigation (lead); Methodology (lead); Project administration (lead); Resources (equal); Supervision (equal); Validation (lead); Visualization (lead); Writing – original draft (lead).

The data that support the findings of this study are available within the article and its supplementary material.