Introduction

Cuprous oxide (Cu2O), a p-type semiconductor with a direct band gap of 2.1 eV, has long been considered a promising material for low-cost solar-energy conversion and photocatalysis1,2,3,4,5. Its advantages include a high absorption coefficient, a suitable band-gap width, chemical stability, non-toxicity and abundant reserves. Nitrogen doping in Cu2O is an important research topic because of its tremendous potential for overcoming the major disadvantage of Cu2O - its high resistance. Moreover, recent research has revealed that in addition to conductivity improvement, nitrogen-doped Cu2O, hereafter referred to as Cu2O:N, exhibits enhanced light absorption below the band gap, most likely because of the introduction of an intermediate band (IB) located ~0.7 eV above the valence band maximum (VBM)6,7. Its advantageous band gap and IB level have made Cu2O:N an excellent candidate material for IB solar cells8. Its enhanced subband absorption in combination with its excellent visible-light absorption is also an outstanding advantage for photocatalysis because most inorganic photocatalysts suffer from poor activity or even inactivity under visible-light illumination, including the extensively studied TiO2 and perovskite compounds9,10,11. However, several other groups have observed no enhancement in subband absorption even in heavily doped Cu2O:N films12,13 and Nakano et al. even observed a band-gap-widening effect upon nitrogen doping14.

Through the application of first-principles calculations, several theoretical investigations of Cu2O:N have also been performed. Li et al. claimed that nitrogen impurities in Cu2O induce a marked widening of the band gap when oxygen vacancies are present15, which could account for the experimentally observed optical band-gap widening of Cu2O:N prepared via the sputtering technique14. Conversely, Zhao et al. reported a theoretical prediction that nitrogen doping should slightly widen the band gap, causing the formation of an IB in the gap located at ~0.9 eV above the VBM16.

In general, there is still some controversy regarding the effects of nitrogen doping on Cu2O and no comprehensive understanding has yet been reached. It is well known that impurities at different sites in the lattice have distinct effects on the electrical and optical properties of a material9,10. However, interstitial nitrogen (Ni) in Cu2O has long been ignored in previous experimental and theoretical studies, which have focused only on nitrogen impurities substituted at oxygen sites (NO)6,7,12,13,15,16,17,18,19. In this work, we found that even in the lightly doped samples, a considerable number of Ni formed in Cu2O, along with NO and oxygen vacancies (VO). In the course of annealing, migrating Ni reacted with VO forming more NO, thereby altering the corresponding contents of these point defects and resulting in a significant change of the optical and electrical properties of the material.

Results

Cu2O and Cu2O:N films were obtained via the post-oxidation of Cu (111) films that were initially deposited on c-plane Al2O3 buffered with a 400 nm thick, semi-insulating ZnO film20. Nitrogen doping was achieved by introducing a nitrogen plasma through a radio-frequency (RF) plasma gun during the oxidation process. It was found that a micro-zone phase separation occurs when fabricating Cu2O films at high temperatures, while oxidizing at 300°C resulted in single oriented Cu2O (111) films. So that this recipe was adopted for all samples used in this study. Figure 1(a) presents typical XRD θ-2θ scans for the samples. Only one peak corresponding to the Cu2O (111) plane can be observed, in addition to the ZnO (0002) and Al2O3 (0006) signals; these results are consistent with the in situ reflection high-energy electron diffraction (RHEED) observations, as shown in Figure 1(b). Thus no traces of CuO, Cu or Cu3N phases were found, suggesting high quality Cu2O with diluted nitrogen in the doped samples. The clear RHEED patterns also indicate the fine crystallinity of the films. Atomic force microscopy images (not shown) revealed a uniform surface with a root-mean-square roughness of ~10 nm in a scan area of 10 μm × 10 μm.

Figure 1
figure 1

(a) Typical XRD θ-2θ scans of the samples and (b) in situ RHEED patterns of undoped Cu2O sample, which are similar to those of Cu2O:N.

The resistivities of the as-grown Cu2O and Cu2O:N samples were approximately 340 Ωcm and 100 Ωcm, respectively, measured by Hall effect. The carrier density and mobility of the Cu2O were 3.5 × 1014 cm−3 and 53 cm2V−1s−1, respectively. Interestingly, the Cu2O:N sample demonstrated ambiguous p-type conduction behavior (Fig. S1 in the supplementary information), suggesting a low value for the Hall coefficient and/or the existence of hole traps. Annealing were applied to activate nitrogen related acceptors and/or remove the hole traps17,19.

Figure 2(a) shows room temperature values for the hole density, mobility and resistivity of the Cu2O and Cu2O:N samples after annealing at 750°C for 10 min as functions of the nitrogen plasma power. A power of 0 W represents undoped Cu2O. It is clear that the hole density increased with the power of the nitrogen plasma (likely to be proportional to the doping level) and the mobilities of Cu2O:N were all smaller than those of Cu2O (also consistent with the hypothesis of high nitrogen incorporation). The hole concentrations of Cu2O:N were in the range of 1016 cm−3, i.e. two orders of magnitude higher than that of Cu2O12,17,18,19. Figure 2(b) presents the room temperature Hall data for Cu2O:N synthesized using nitrogen plasma power of 200 W, as a function of annealing temperature. The lowest resistivity and the highest hole density were obtained upon annealing at 750°C. When the sample was annealed at 800°C, although the mobility was higher because of the improved crystal quality, the hole density decreased, leading to reduced conductivity. A possible explanation is that annealing at such a high temperature may also induce the out diffusion of nitrogen in addition to the acceptor activation. Note that the annealing conditions can be further optimized for better conductivity; nevertheless, in our study, the results indicated that the samples annealed at 750°C were already suitable for the study of nitrogen-related defects on behalf of the prominent evolution in electrical properties. Furthermore, the temperature-dependent hole concentration was also measured to determine the activation energy of the dominant acceptors in Cu2O:N, which was calculated to be 121 meV (supplementary information). This level is attributed to extrinsic NO acceptors, because the only reasonable alternative of copper vacancies (VCu) exhibits a much deeper level21,22,23. Figure 3 presents the nitrogen concentration versus depth profiles of the Cu2O:N films doped with 200 W nitrogen plasma as measured by secondary ion mass spectroscopy (SIMS), which confirms: (i) nitrogen incorporation into our films and (ii) its gradual out diffusion during anneals.

Figure 2
figure 2

Hole density, hole mobility and resistivity of (a) Cu2O:N samples annealed at 750°C as functions of the nitrogen plasma power (a power of 0 W represents an undoped sample) and (b) Cu2O:N samples doped with a nitrogen plasma power of 200 W as functions of the annealing temperature.

Figure 3
figure 3

Nitrogen concentration versus depth profiles in Cu2O:N samples.

Discussion

The data in Fig. 2 and Fig. 3 suggest that nitrogen, likely in favor of NO, behaves as an efficient acceptor in Cu2O:N. Logically, to enhance the conductivity of Cu2O:N, we simply need to increase the doping level and/or optimize the annealing process to balance the activation and outward-diffusion processes. However, the microscopic picture of the doping mechanism is still not clarified. NO is likely not the only defect type introduced by nitrogen doping. As is well known, when aliovalent substitutions occur in ionic compounds, a charge-compensation mechanism is required to ensure that the overall charge remains neutral. For Cu2O:N with ionized NO acceptors, the compensation can be achieved, e.g. through the formation of anion vacancies (VO) or less likely, cation interstitials (Cui) though their formation energy is relatively high21,22.

Photoluminescence (PL) is a non-destructive technique for the investigation of defects in semiconductors. The PL spectrum of Cu2O has been adequately researched in the literature and the luminescence related to excitons, VO and VCu has been well established24,25,26. However, previous research has not produced sufficient results with respect to PL studies of Cu2O:N. Ishizuka et al. have reported that Cu2O:N exhibits no PL peak prior to hydrogen or crown-ether cyanide treatment, indicating the existence of non-radiative recombination centers in the polycrystalline Cu2O18,19.

Figure 4(a) presents the PL spectra of the Cu2O samples, with arrows indicating positions related to the excitons, VO and VCu. The presence of VO is due to the oxygen-lean growth conditions. After high-temperature annealing, the exciton luminescence was remarkably enhanced, whereas the intensity of the VO signature did not change significantly, suggesting an improvement in crystal quality after annealing. Notably, the formation of VCu was suppressed because of the oxygen-lean conditions during growth, enabling a strong exciton luminescence25,26. The shape of the exciton luminescence peak can be well interpreted in terms of phonon-assisted transitions, as discussed in Ref. 26.

Figure 4
figure 4

Photoluminescence spectra of (a) Cu2O samples, (b) Cu2O:N samples doped at 200 W and (c) enlarged portion of (b) in the vicinity of the exciton emission.

Figure 4(b) presents the spectra of the Cu2O:N films while the spectrum of the Cu2O sample annealed at 750°C is also included as a reference. The most distinguishable features are the suppressed exciton luminescence and intense VO–related luminescence of the Cu2O:N films after annealing, supporting our hypothesis that incorporation of nitrogen is associated with the VO generation. Concurrently it has been demonstrated that the presence of VO has no responsibility for damping the exciton luminescence25,26, suggesting other traps to be involved and specifically Ni and NO may be considered as prime candidates for non-radiative recombination mechanisms.

Indeed, the exciton luminescence was slightly enhanced (see Fig. 4(c)), whereas the intensity of the VO signal, in contrast, decreased with increasing annealing temperature. Taking into account the fact that PL intensities are proportional to defect contents, the evolution of the VO signature suggests that the annealing processes somehow reduced the VO content in the Cu2O:N samples. It should be mentioned that the as-grown Cu2O:N sample is likely to contain the highest density of VO; however, this fact was not directly reflected by the PL intensity because of the poorer crystallinity of the as-grown sample. Considering that the annealing process played a limited role in repairing the VO in the undoped Cu2O (see Fig. 4(a) and Ref. 26), this trend of decreased VO concentration in Cu2O:N should be attributed to the evolution of defects not available in the undoped films. A reasonable explanation is the refilling of VO by interstitial atoms, likely Ni, which move faster at higher temperatures and thus can more easily occupy the vacancy sites. Interstitial oxygen (Oi) is another possible candidate to react with VO; however, its formation is disfavored under oxygen-lean growth conditions21,22. Thus, the difficulty encountered in judging the conduction type of the as-grown Cu2O:N based on the Hall measurements is the natural result of the presence of less NO and more Ni in the film.

Further evidence for Ni is observed in Figure 4(c), the enlarged spectra of the Cu2O:N films in the vicinity of the exciton level. The as-grown Cu2O:N sample exhibited a unique signature located at ~652 nm (1.90 eV). This signal did not originate from the β band in nonstoichiometric Cu2O, despite the very similar energy level (1.91 eV) because the β band created by exciton-defect complexes would not be affected by the annealing process27,28. It cannot be related to NO; if it were, it would not have disappeared after annealing. Therefore, this 652 nm peak is most likely attributable to the emission related to Ni. The absence of this signature in the annealed samples is the result of the diffusion of nitrogen out of the material and the transformation of Ni into NO. This interpretation is reasonable, considering the variation in the VO signature and the electrical properties.

It is worth noting that for all our samples, no signature could be observed at approximately 680 nm (Fig. 4(b)), where Ishizuka et al. observed a lone peak in the PL spectrum of Cu2O:N after passivation treatment18,19. Park et al. have also reported the observation of this 680 nm luminescence signature in Cu2O films without intentional doping29, indicating that the origin of this feature in polycrystalline Cu2O thin films remains uncertain.

In order to extend our interpretation, we applied X-ray photoelectron spectroscopy/X-ray Auger electron spectroscopy (XPS/XAES) to probe the defects30,31,32. A comparative study was conducted using four samples – two as-grown and two 750°C annealed samples of Cu2O and Cu2O:N, labeled as Cu2O, Cu2O AN, Cu2O:N and Cu2O:N AN, respectively. Notably, the confirmation of the nitrogen atoms' locations from their chemical states or bonding status is impractical because no signal related to nitrogen could be detected because of the low doping level (~1018 cm−3). Instead, we relied on the detection of other defects to interpret the formation and evolution of nitrogen-related defects by means of the defects' interactions.

Figure 5 presents the high-resolution XPS spectra of the (a) Cu2p and (b) O1s core levels. As shown in Fig. 5(a), the Cu-related peaks exhibit a symmetric shape with no satellite structure at approximately 943 eV, ruling out the presence of Cu2+ in these films. The fitting of the Cu2p3/2 peaks indicates a binding energy of 932.6 eV, corresponding to that of Cu+ in Cu2O. The presence of metallic Cu clusters may also be ruled out accounting for: (i) the Cu2p peaks exhibit strong symmetry and the full widths at half maximum (FWHM) for all different samples are nearly identical (1.43–1.49 eV); (ii) the XRD results show no evidence of Cu (Fig. 1); and (iii) the samples exhibited good p-type conductivity. Further, the signatures of the O1s peaks at 530.4 eV are consistent with O-state in Cu2O too. Notably, a much more pronounced variation in the FWHM values (1.15–1.58 eV, see Fig. 5(b)) indicates a trend that may be correlated with the presence of VO based on literature interpretations of oxygen deficient samples30,31.

Figure 5
figure 5

XPS (a) Cu 2p spectra and (b) normalized O 1s spectra of Cu2O and Cu2O:N samples.

As XAES is much more sensitive to the nature of the neighboring bonds and/or point defects than XPS30,32, the evolution of Cu LMM Auger lines was measured and Fig. 6 shows typical Cu-related signatures deconvoluted with five different components. Notably, three peaks located at 913.5 eV, 919.5 eV and 921.4 eV are known not to be sensitive to the chemical environment33,34,35. The dominating peaks in Fig. 6 at 916.8 eV (labeled as peak 1) and 918.3 eV (labeled as peak 2) are typically attribute to Cu2O and Cu, respectively, though peak 2 can also be observed in nonstoichiometric Cu2O33,34. In fact, the appearance of peak 2 in Fig. 6 is intriguing since we have ruled out the presence of Cui or Cu cluster in our films. An alternative identification of peak 2 may be to assign it to VO, since strong localized unpaired electrons of copper atoms neighboring VO might influence its effective charge state readily detectable by XAES21,22,30. In another words, the signal of Cu next to VO (-O-Cu-VO-), i.e. Cu dangling bond, is distinguishable from Cu1+ (-O-Cu-O-), assuming that one VO would lead to an “aggregation” of four Cu atoms (Fig. S3 in the supplementary information).

Figure 6
figure 6

Cu LMM spectra demonstrated with 5 different components for the Cu2O (left column) and Cu2O:N (right column).

Adopting this identification, further conclusions may be done considering the evolution of the peaks intensities in Fig. 6. Notably, the annealing results in much more pronounced decrease of the VO content (in terms of the integrated intensity ratio of peak1 and peak 2) in Cu2O:N, comparing to that in Cu2O, which is consistent with the PL results, altogether suggesting a critical role of Ni to interact with VO in the course of annealing. Indeed, during annealing, the supply of atmospheric oxygen is insufficient to change VO concentration drastically, while the presence of Ni in Cu2O:N makes the major contribution to interact with VO forming NO.

The higher ratio of the annealed Cu2O:N sample indicates the lower VO content, i.e., fewer Cu dangling bonds, compared with the annealed Cu2O sample; this finding reflects the important influence of the growth conditions, as the nitrogen plasma provided additional total anions. One may argue that this result seems to be in conflict with the stronger VO-related luminescence of Cu2O:N, as shown in Figure 5(b). We attribute this phenomenon to the passivation effect of nitrogen36, which arises because Ni can also move to defect sites other than VO. This passivation effect is further evidence of significant Ni formation induced by doping. Ishizuka et al. also observed an enhanced PL intensity after nitrogen doping18.

Notably, the fitted FWHMs of peak 1 for the Cu2O:N samples (Table 1) are slightly broadened compared with those of the undoped Cu2O samples. Fixing the FWHMs of peak 1 yielded very similar fitting results; however, it was found that allowing this broadening improved the fit considerably, yielding a smaller standard deviation. This broadening is interpreted as a sensitive indication of Cu-N bonding, as supported by the Cu LMM spectrum of a Cu3N thin film, which exhibits a similar shape to that of Cu2O, also with a predominant peak at 916.8 eV (supplementary information).

Table 1 Summary of the Cu LMM peak-fitting parameters

In conclusion, Cu2O and Cu2O:N thin films were synthesized through the oxidation of Cu films. NO acceptors effectively enhance the conductivity of such films, with an activation energy of approximately 121 meV. Probing the defects in the films using a combination of PL and XPS/XAES revealed the formation of Ni and the enhancement of the VO content in the prepared Cu2O:N thin films. The evolution of the optical and electrical properties of these films can be consistently explained in terms of the Ni interacting with VO and the resultant transformation into NO in the course of annealing.

Methods

The samples were fabricated in an ultra-high-vacuum chamber with a background pressure within the lower range of 10−7 Pa. A conventional Knudsen cell was used to evaporate metallic Cu (6N). Cu (111) films were initially deposited on c-plane Al2O3 buffered with a 400 nm thick semi-insulating ZnO (0001) film. Cu2O films were obtained via post-oxidation using oxygen plasma and Cu2O:N films were obtained under similar conditions through co-treatment with oxygen and nitrogen plasmas. Specifically, the oxygen plasma was used at a fixed O2 flux of 1 sccm (the partial pressure was maintained within the lower range of 10−3 Pa) and discharged via RF power. The doping level was controlled by varying the RF power and the N2 flux. The thickness of the prepared films was approximately 150 nm. The oxidation/doping temperature was 300°C for all samples. The typical oxidation time was approximately 90 minutes. The oxidation process was monitored by means of in situ reflection high-energy electron diffraction (RHEED) observations. Once the RHEED pattern indicated complete oxidation of Cu into Cu2O, the plasma exposure was terminated. Shorter or longer treatment times would result in undesired phases of Cu and CuO, respectively. More details can be found in our previous report20.

Several samples were subjected to post-annealing in the same chamber under an oxygen-gas atmosphere (also at ~10−3 Pa) at various temperatures. The temperature ramping rate was 30°C min−1. Once the set temperature was reached, that temperature was maintained for 10 min and the sample was then cooled at the same rate of 30°C min−1.

XRD measurements were performed using a Mac Science M18XHF diffractometer with a Cu Kα radiation source (λ = 1.5406 Å) at 40 kV and 50 mA (2 kW).

The van der Pauw configuration was used for the Hall measurements. Room-temperature tests were performed using an Ecopia HMS-3000 system and the results were double-checked using a home-made Hall system. The Hall coefficient was extracted from the trace curve of the transverse voltage recorded using an alternating magnetic field and a constant current (supplementary information). The temperature-dependent hole density was recorded using a Lakeshore model 7604 Hall system. The acceptor level was extracted using the following equation:

where NA, ND, Nv, EA, k and T are the acceptor density, donor density, effective density of states in the valence band, acceptor level, Boltzmann constant and absolute temperature, respectively, and

where mh and h are the hole effective mass and the Plank constant, respectively.

For the photoluminescence measurements, the samples were excited using a 532 nm laser at 1 mW focused to a spot of approximately 2 μm. The spectra were measured at room temperature in a back-scattering geometry. The integration time was 10 seconds in all cases.

SIMS measurements were performed using a Cameca IMS 7f microanalyzer.

XPS measurements were performed using a PHI-5300 ESCA Spectrometer with a monochromatic Al kα X-ray source at 1486.6 eV. All data were collected at a nominal depth of 6 nm after Ar-ion sputtering and the damage effect proved to be negligible (supplementary information). Spectral analysis was performed using XPSPeak version 4.1. Shirley background subtraction was adopted and the fitting results were determined based on the minimum standard deviation.