Effects of high temperature annealing on defects and luminescence properties in H implanted ZnO

After 600 °C annealing, a band of voids is observed at the peak of the as implanted H profile, as illustrated in the XTEM image in figure 1(a) and the SIMS results in figure 1(b). The as implanted SIMS H profile peaks at ∼670 nm, which is close to the SRIM simulated projected range R_p≈660 nm [9]. Significant outdiffusion of implanted H is observed upon annealing, with the peak H concentration decreasing to ∼0.2–0.4% of the initial concentration after 800 °C annealing [7]. SIMS measurements also show that upon annealing, Li redistributes to a profile that matches the shape of the H profile. The resemblance of the Li and H profiles suggests that (Li_Zn–OH) complexes form during thermal annealing, however the trapping of H and interstitial Li at the same acceptor-like defects cannot be ruled out [7]. The XTEM image in figure 1(a) also shows that the voids in the 600 °C samples are round and irregular in shape, in an obvious contrast to the sharp faceted voids observed in the 800 °C samples [7]. By sampling a defect free region in the XTEM images taken along [1  $\overline{1}$ 0 0] (400 nm × 400 nm, with the centre at R_p) for each sample annealed at 600 °C and 800 °C, respectively, statistical analysis of the empty volume density and the voids sizes is carried out. By measuring 75 voids for each sample, the voids in the 600 °C sample are found to have larger sizes of ∼14 ± 9 nm, compared to ∼10 ± 5 nm in the 800 °C sample. Furthermore, by assuming the same thickness for both the XTEM foils, from an empty volume density measurement (total area of voids/total volume of the sampled area), the empty volume that is bound by the voids in the R_p region is estimated to be approximately three times larger in the 600 °C sample than in the 800 °C sample.

Zoom In Zoom Out Reset image size

The shrinkage in void size and a reduction in empty volume at >600 °C, as revealed by comparing figure 1(a) and the previous studies [7], corroborates the positron annihilation spectroscopy (PAS) data that exhibits a drop in the vacancy concentration in the H implanted region at temperatures >700 °C [10]. In hydrothermally grown ZnO samples implanted with Zn, upon annealing, mobile Zn interstitials (Zn_i) are observed to be released rapidly from the implanted region and 'kick out' substitutional Li from Zn sites (Li_Zn) [11]. This produces highly mobile Li interstitials (Li_i) and results in a significant Li depleted region beyond R_p [11]. This was ascribed to being similar to the commonly known transient enhanced diffusion (TED) of B in Si [12]. The burst of Zn_i is reported to occur at temperatures ∼600–800 °C [11], and coincides with the temperature at which a decrease in void sizes is observed in the present work. As Zn_i are highly mobile even below RT [13], they are expected to participate in the annihilation of V_Zn. During annealing, the burst of excess Zn_i from the implanted layer further promotes the V_Zn annihilation process, thus causing a shrinkage in size of the voids in the present work, as well as the decrease in V_Zn concentration as reported by PAS studies [10]. However, in contrast to the Zn implanted samples, the SIMS result in figure 1(b) shows no depletion of Li in the implanted region. This could be due to the voids/bubbles formed in H implanted samples behaving as a strong sink for Zn_i, thus minimizing the impact of Zn_i on the Li profiles through TED as observed in Zn implanted samples [11]. In addition, it is also likely that Li atoms are decorating the voids and influencing their shapes, as fast diffusing metals are commonly known to accumulate around voids in Si [14–16]. For irradiated hydrothermally grown ZnO, PAS studies have also shown the trapping of Li in the vicinity of vacancy clusters after annealing [17]. Nevertheless, this is difficult to prove by TEM imaging from the monolayers of a light element such as Li decorating the internal surface of an empty void in a heavier matrix such as ZnO.

From the DRCLS measurement, for the as implanted sample as shown in figure 2(b), little to no emission is observed from the implanted region for E_B < 14 keV. The weak emission could be a result of a high density of non-radiative recombination centres formed by the H implantation. After annealing, figures 2(c) and (d) show a near band edge (NBE) emission that peaks at ∼3.3 eV and a broad deep level (DL) emission centred between ∼2.0–2.5 eV. However, the emission near the surface (at low E_B) is still quite weak. Thus, the implantation damage is still not completely recovered after annealing to 800°C for 1 h. The NBE emission at 80 K results from the recombination of free and bound excitons [18], while the chemical origin of DL emission still remains elusive [19]. The observed DL emission can often be fitted with 2 Gaussian components: (1) a lower energy component (1.9–2.1 eV) that has been attributed to Zn vacancy (V_Zn) and vacancy clusters [20], or to Li [21]; and (2) a high energy component (∼2.3–2.5 eV) that is assigned to O vacancies (V_O) [22]. In addition, a set of equally spaced peaks separated by ∼72 meV attributed to the longitudinal (LO) phonon replicas of the free excitons are also observable. For the 600 °C sample, there is also a broad band that peaks at ∼3.1 eV as illustrated in figure 2(c) for E_B ≈ 10–20 keV. Furthermore, in both the annealed samples, the NBE peaks are also red shifted with higher E_B, possibly suggesting the creation of new radiative recombination centres by H implantation and subsequent annealing. Self-absorption effects may also be a cause of the NBE red shift as the CL emission escape depth in the sample will increase with increasing CL generation at higher E_B [23, 24].

Zoom In Zoom Out Reset image size

The electron–hole pair excitation depth at each E_B is simulated with the Monte Carlo code CASINO to correlate the CL emission with respect to the implantation profile. The electron–hole pair generation (primary electron energy loss) is not linear over its maximum excitation depth U_m, but instead has a stretched Gaussian shape that peaks at ∼0.3 U_m. Therefore, in the present work, a corrected excitation depth U₀ that corresponds to the depth at which 70% of electron–hole pairs are generated is used instead. Figures 3 show the intensity of the NBE emission I_NBE and the integrated intensity of the DL emission I_D with respect to U₀. I_D (2.0 eV) and I_D (2.4 eV) are obtained by deconvoluting the DL emission into 2 Gaussian distributions that peak at 2.0 eV and 2.4 eV, respectively. A Gaussian distribution that peaks at 3.1 eV is also used to fit the anomalous broad band at ∼3.1 eV found in the 600 °C sample, with the NBE band, the 1LO and 2LO of the free excitons were included in the fitting process, as shown in the inset of figure 2(c). The emission that peaks at 2.0 eV, 2.4 eV and 3.1 eV will be hereafter referred to as red, green and violet emission, respectively.

Zoom In Zoom Out Reset image size

After implantation, the sample shows very low DRCLS emission from the surface to U₀ ≈ R_p as illustrated in figures 3(b), suggesting that the most defective layer in the implanted region is located at U₀ < R_p. After annealing, the emission from this defective layer does not completely recover to the one in the as grown sample as shown in figure 3(a), and results in a plateau region with low I_NBE from the surface to U₀ ≈ 500 nm. This suggests that the non-radiative recombination channels caused by the high dose H implantation are not completely annealed out even after 800 °C. Nonetheless, for the annealed samples, the DL emission does not appear to share the same depth profile trend as the NBE. For the red emission, I_D (2.0 eV) is found to increase with higher E_B, irrespective of the annealing temperature. For the 600 °C sample, there is also a small peak for I_D (2.0 eV) at U₀ ≈ 500 nm, showing a possible increase of the defect concentration related to red emission. On the other hand, for the green emission, I_D (2.4 eV) appears to be completely suppressed after 600 °C annealing, which could be a result of the passivation of deep level defects by the implanted H. This is supported by the SIMS data in figure 1(b), where no H is detected at depth >1000 nm with I_D (2.4 eV) increasing sharply. However, after 800 °C annealing, the green emission is significantly enhanced in the near surface region and becomes more intense than the red emission. For the violet emission that is only observable at E_B ≈ 10–20 keV after 600 °C annealing, I_D (3.1 eV) increases from low E_B, then peaks at 14 keV/U₀ ≈ 430 nm, and eventually decreases when DRCLS probes deeper into the sample. The I_D (3.1 eV) peak is located at approximately the same depth as the small peak of I_D (2.0 eV). This depth is also where the plateau region of near surface low I_NBE ends.

The surface enhancement of the green emission after 800 °C annealing has also been reported in electron irradiated melt-grown samples [25] and as-received hydrothermally grown samples [26]. This increase has been ascribed to either the dissociation of the non-radiative V_O–acceptor defects complexes [25], or to the loss of H near the surface, as re-annealing the samples in forming gas (5% H₂/95% N₂ or 5% H₂/95% Ar₂) quenches the green emission [26]. Both of these studies propose that the passivated defects involved with the green emission could be due to Cu, which has often been assigned as a source for green emission, especially when fine structures resembling longitudinal optical (LO) phonon replicas with a periodic energy spacing of 72 meV is observed on the high energy slope of the band [27, 28]. As CuH and CuH₂ complexes dissociate below 800 °C [29], which is the temperature where the green emission starts to re-appear, this may also explain the surface enhancement of the green emission after 800 °C annealing observed in the present work. Similar to the works by Allen et al [26] and Knutsen et al [25], in the PL spectrum that is collected at 14K for the 800°C sample as shown in figure 4, fine structures separated by ∼72 meV occur on the high energy slope of the DL emission. The SIMS measurement also shows an accumulation of Cu concentration (∼1 × 10¹⁸ cm⁻³) near the surface that decreases gradually to the background concentration (∼1 × 10¹⁵ cm⁻³) at a depth of ∼400 nm (see the online supporting information for the Cu concentration-depth profile measured by SIMS for the H implanted sample before and after implantation and annealing (stacks.iop.org/JPhysD/47/342001/mmedia)). Therefore, Cu may be one of the defects related to the green emission enhancement after 800 °C annealing, and is susceptible to passivation by H after 600 °C annealing.

Zoom In Zoom Out Reset image size

The broad violet band centred at 3.1 eV that appears after 600 °C annealing (in figure 2(c)) is uncommon, but when observed it has been attributed to (a) a donor to acceptor pair (DAP) [30], (b) shallow donor levels such as Zn_i [31], and (c) Li_Zn acceptor states located at ∼0.3 eV above the valence band [32]. Figure 3(c) shows that the violet emission peaks at U_O ≈ 500 nm, which is slightly shallower than the peak of the SIMS implanted H profiles at ∼600 nm in figure 1(b). As discussed previously, depth profiling with DRCLS cannot be extracted linearly with respect to E_B. The actual depth for the violet emission peak may be slightly different than the one derived from the 70% of the maximum CL generation depth. In the implanted layer, it has been shown that the region shallower than R_p is mostly filled with vacancies, while the interstitial-rich region is located at a depth deeper than R_p [33]. As discussed previously, the small peak of the red emission at U_O ≈ 500 nm in figure 3(c) appears to overlap with the depth at which the violet emission peaks. Despite still being controversial, the red emission is typically suggested to be related to the V_Zn and clusters thereof [20]. In addition, I_D (3.1 eV) also has a depth profile that is similar in shape and trend to the SIMS Li profile in the 600 °C annealed sample in figure 1(b). This may indicate a correlation between the violet emission, vacancies and Li, in general agreement to the study by Zhang et al that relates the violet emission to Li_Zn [32]. However, other sources mentioned earlier for the violet band may also not be ruled out here, such as the DAP and free to bound recombination, which may have a weak phonon coupling that makes the phonon replicas hard to be detected.