Near band edge photoluminescence of ZnO nanowires: Optimization via surface engineering

Zinc oxide nanowire arrays (ZnO NWs) have been extensively studied for their versatile applications in optical and optoelectronic devices, such as light-emitting diodes (LEDs), photodetectors, optical modulator waveguides, and photoelectrochemical anodes,^1–9 for ZnO NWs exhibit fascinating features, i.e., large exciton binding energy, high transparency in the visible range, and intense near band edge emission.^10–15 Among various techniques that have been used to fabricate ZnO NWs, the wet-chemical approach becomes the most widely adopted method for its simplicity, as well as its capability to synthesize well oriented single crystalline ZnO NWs over various types of substrates in a scalable manner.^16–19 However, ZnO NWs grown by the wet-chemical process usually feature a high density of defects, such as oxygen vacancies and Zn interstitials, which suppress near band edge emission (NBE) and contribute to broad defect state emission (DE) in the photoluminescence (PL) spectra.^3,4,17,20,21 Such defect emissions greatly limit ZnO NWs' applications towards UV laser diodes or UV-blue light emitters. Various post-processing treatments have been applied aiming at enhancing the NBE while suppressing the DE in ZnO NWs' PL, with the belief that such treatments may remove most of the defect states and improve the crystal qualities of ZnO NWs.^22–25 Contrary to the common understanding, defects do not always affect the NBE from ZnO NWs adversely. Here, we demonstrate that surface oxygen vacancies introduced by low pressure oxygen plasma/sputtering significantly improve the NBE and suppress the DE, through the manipulation of ZnO NW surface band bending.

The ZnO NW arrays studied in this work are synthesized using a seed-mediated hydrothermal method that is slightly modified from the one developed by Greene et al.^17,18 On average, the nanowires have a diameter of $40–50$ nm and a length of 500 nm (supplementary material). In our previous studies, we demonstrated the effective enhancement of ZnO NWs' NBE and suppression of DE through post-processing of sequential 500 °C O₂ annealing and low pressure O₂ plasma treatment.^16,19,26 In Fig. 1, we present the room temperature PL spectra of ZnO NWs processed under various post-processing techniques, and the samples are named accordingly (supplementary material). The PL from ZnO NWs is excited at a wavelength of 280 nm (4.4 eV) and is collected after passing through a 355 nm long pass filter. Specifically, sample ZnO-ag is the “as-grown” ZnO NWs from the hydrothermal method; sample ZnO-A is as-grown ZnO NWs annealed at 500 °C under O₂ flow (A for 500 °C O₂ annealing); sample ZnO-P is as-grown ZnO NWs treated with low pressure O₂ plasma at room temperature (P for O₂ plasma, 20 W, 5 min, chamber pressure p < 25 mTorr); and sample ZnO-AP is as-grown samples treated with 500 °C O₂ annealing (A), followed by low pressure O₂ plasma (P). According to Fig. 1, sample ZnO-ag exhibits a broad defect emission (DE) peak typical of the wet-chemical fabricated ZnO NWs,^3,17,20 in addition to a weak NBE peak at 3.27 eV (378 nm) that corresponds to the radiative recombination of excitons.^10,21 The annealing treatment alone does not bring significant enhancement to the ZnO NWs' NBE. On the other hand, low pressure O₂ plasma treatment significantly improves the NBE at 3.27 eV and suppresses the DE, which is clearly observed by comparing the sample pairs of ZnO-ag/ZnO-P and ZnO-A/ZnO-AP. In particular, through the sequential combination of annealing and plasma treatments, DE is mostly eliminated from the ZnO NWs' PL, leaving an intense and symmetric NBE peak at 3.27 eV.

According to the prevailing hypothesis in the literature, it is believed that the low pressure O₂ plasma treatment efficiently eliminates the defect states in ZnO NWs, so that the NBE-dominated PL emission could be achieved. However, as suggested by further X-ray spectroscopy studies, the actual effect of the low pressure O₂ plasma is quite contrary. X-ray photoelectron spectroscopy (XPS) studies for freshly treated ZnO NWs reveal asymmetric O 1s peaks, which can be deconvoluted into two Gaussian peaks [Peak I in blue and Peak II in red, Fig. 2(a)]. Peak I (∼529.8 eV) is attributed to oxygen atoms from the ZnO crystal lattice (O_L), and Peak II (∼531.4 eV) is attributed to oxygen atoms in the vicinity of oxygen vacancies (O_V).^27–29 From the O_V/O_L ratios deduced from the XPS measurement [Fig. 2(b)], it appears that the surface of low pressure O₂ plasma treated samples (ZnO-P and ZnO-AP) features a significant higher fraction of oxygen vacancies. The increased amount of oxygen vacancy defects in samples ZnO-P and ZnO-AP is also observed in the X-ray absorption near edge structure (XANES) from the Zn K-edge. When compared to other samples, Zn K-edge XANES from the O₂ plasma treated samples (ZnO-P and ZnO-AP) features increased white lines [Fig. 2(c)], which suggests that the Zn atoms have lower oxidation states and correspondingly fewer oxygen neighbors after O₂ plasma treatment (i.e., more oxygen vacancies). Absorption in the white line region corresponds to the core electron transitions to Zn d states. Therefore, its intensity is determined by the electron density of available d-states, which directly correlates with the oxidation state.³⁰ It should be noted that, although XANES is a bulk sensitive technique, a large fraction of material locates at the surface due to the high surface-to-volume ratio of ZnO NWs, so that the surface oxygen vacancy may have a significant impact on the overall XANES spectral features. Therefore, we may conclude that, instead of eliminating defect states as previously believed, the low pressure O₂ plasma treatment is more of a sputtering process that preferentially removes the lighter element (oxygen) from the ZnO NW surface, leaving the surface rich of oxygen vacancies and partially reduced.

It is clear that the PL enhancement of ZnO NWs is highly correlated with the O₂ sputtering induced Zn-rich surface. One evidence is that ZnO NWs' NBE can only be enhanced when low pressure O₂ plasma treatment comes at last. In other words, no NBE enhancement will be observed if O₂ plasma is performed prior to O₂ annealing (supplementary material). In addition, not all O₂ plasma treatments are able to enhance NBE, and only those treatments conducted at a relatively low chamber pressure (p < 25 mTorr) which hold the sputtering capability enhance NBE. If conducted at a higher chamber pressure (e.g., 100 mTorr), the plasma ions would have less kinetic energy for sputtering and the NBE improvement is no longer observed (supplementary material). Moreover, other means of surface reduction are also capable of improving the ZnO NWs' PL. For H₂ plasma treated ZnO NWs, NBE is enhanced in a similar fashion to sample ZnO-P (supplementary material). Another example is that, according to Zhu et al., NBE from ZnO NWs can also be improved after annealing in H₂.^3,20,21 It is worth mentioning that although H₂ treatments can reduce the ZnO surface, they are much more aggressive in terms of their reducing power when compared to low pressure O₂ plasma. For instance, the H₂ plasma performs slow but steady chemical etching against ZnO NWs. When operating at 20 W, the H₂ plasma treatment shortens the ZnO NWs' length by 20% in 5 min (supplementary material). Besides, the H₂ plasma process brings a violet-blue band (⁠$2.59–3.27$ eV) to ZnO NWs' PL spectrum, which is likely due to Zn interstitial defects.^3,5,20 As such, the low pressure O₂ plasma technique remains the preferred method to create a mildly reduced ZnO surface for the improved NBE, with the advantage of not damaging the ZnO NWs or introducing much more excessive defects to the nanowires.

To understand how a partially reduced surface helps to improve the NBE from ZnO, we probe the surface potentials by low-energy electron microscopy (LEEM) and the work functions by ultraviolet photoelectron spectroscopy (UPS) for samples before and after O₂ plasma treatment.^31,32 For these measurements, we use ZnO thin films deposited by atomic layer deposition (ALD) in lieu of ZnO NWs to fulfill the sample requirements of the smooth surface. The deposition is conducted at low temperature (150 °C) to mimic the hydrothermal growth of ZnO NWs. The room temperature PL spectrum from the ALD film is very similar to that of the ZnO NWs and follows the same trend of NBE improvement after post-processing, making the thin film sample a good representation of the NWs (supplementary material). The surface potential is determined by measuring electron reflectivity vs. start voltage, i.e., the incident electron energy with respect to the sample surface [Fig. 3(a)], and corresponds to the transition between the low energy, direct reflection regime (high reflectivity) and the higher energy, elastic back scattering regime (low reflectivity). The surface potential of ZnO-A (⁠$−UZnO−A$⁠) is 0.3 eV lower than that of ZnO-AP (⁠$−UZnO−AP$⁠), with the same shift observed for ZnO-ag and ZnO-P pair. According to UPS, the valence band maximum (VBM) of ZnO-AP is also deeper than that of ZnO-A, which supports the LEEM results (supplementary material). The shift of surface VBM is very likely to be induced by surface dipoles formed at the Zn-rich surface after low pressure O₂ plasma treatment. However, the work function of ZnO NWs is not altered after the plasma treatment (supplementary material). While the surface VBM is deeper, the Fermi level is still pinned by the bulk of ZnO so the band will bend downwards (accumulation type for n-type ZnO) after the treatment, thus confining the minority carriers (holes) to the core of ZnO NWs [Figs. 3(b) and 3(c)]. Since most of the defects are located at the ZnO NWs' surface, the confinement of holes will prevent them from reaching or being trapped by the surface defects. The mechanism also explains why sample ZnO-AP has much better NBE than ZnO-P. Since O₂ annealing helps in removing the bulk lattice defects, sample ZnO-A has fewer bulk defects than ZnO-ag, better PL is achieved in ZnO NWs with combined treatment (ZnO-AP), for much fewer carriers will be trapped in the bulk (core) of a ZnO nanowire. The surface potential induced hole confinement, and PL modulation is not an isolated observation. Previously, we had reported that the PL of ZnO NWs could be modulated by electrochemical means. In that work, it was found that the ZnO NWs' NBE could be greatly intensified by swiping the electrode potential across a ZnO/electrolyte interface from anodic (oxidative) to cathodic (reducing), which was also due to the formation of downward band bending and the confinement of holes to the core of ZnO NWs.²⁶ 

In conclusion, we have shown that the improvement of ZnO NWs' band edge photoluminescence is not related to the removal of defects. For ZnO nanostructures with high surface-to-volume ratios, selective introduction of surface defects such as oxygen vacancies may help to confine minority carriers within the bulk through the decrease of work function, which subsequently improves the NBE efficiently. By further creating a defect-free nanowire core through O₂ annealing, NBE of ZnO NWs grown by the wet-chemical process can be significantly improved, which will be very useful in the field of UV laser devices. The O₂ plasma treatment has promising applications in creating the accumulation region near the oxides' surface on a large scale, and the carrier confinement mechanism we propose here should shed light on designing post-processing treatments for many other semiconductor devices.

See supplementary material for experimental details; SEM images; PL spectra of ZnO NWs and ALD ZnO films; and UPS spectra of ZnO samples.

This research used resources of the Center for Functional Nanomaterials, and beamlines in the National Synchrotron Light Source II, including beamline 8-ID ISS (Inner Shell Spectroscopy) and beamline 21-ID ESM (XPEEM/LEEM Spectro-Microscopy Endstation). All of these are DOE Office of Science User Facilities operated for the DOE Office of Science by Brookhaven National Laboratory under Contract No. DE-SC00122704.