Shallow carrier traps in hydrothermal ZnO crystals

Native and hydrogen-plasma induced shallow traps in hydrothermally grown ZnO crystals have been investigated by charge-based deep level transient spectroscopy (Q-DLTS), photoluminescence and cathodoluminescence microanalysis. The as-grown ZnO exhibits a trap state at 23 meV, while H-doped ZnO produced by plasma doping shows two levels at 22 meV and 11 meV below the conduction band. As-grown ZnO displays the expected thermal decay of bound excitons with increasing temperature from 7 K, while we observed an anomalous behaviour of the excitonic emission in H-doped ZnO, in which its intensity increases with increasing temperature in the range 140-300 K. Based on a multitude of optical results, a qualitative model is developed which explains the Y line structural defects, which act as an electron trap with an activation energy of 11 meV, being responsible for the anomalous temperature-dependent cathodoluminescence of H-doped ZnO.


Introduction
Nominally undoped ZnO exhibits n-type conductivity due to the existence of shallow intrinsic defects and the high solubility of extrinsic donor-like impurities such as hydrogen [1,2]. It has been established that hydrogen strongly affects the electronic properties of ZnO. Calculations based on density-functional theory (DFT) by Van de Walle et al. [3,4] indicated that interstitial hydrogen (H i ) and hydrogen trapped at an oxygen vacancy (H O ) can act as two shallow donors, which can be a cause of n-type conductivity in ZnO. These theoretical results have been supported by experiments, which showed that the n-type conductivity and the near-band-edge (NBE) luminescence efficiency of ZnO can be enhanced at the expense of defectrelated emissions via hydrogen incorporation [5][6][7][8][9][10]. Although the origins of defectrelated emissions in ZnO are still not thoroughly understood, the quenching effect suggests that hydrogen interacts with native defects in some way. There have been reports of a stable hydrogen-related complex formed by zinc vacancy (V Zn ) acceptor and two H atoms [11]. As a reactive and common impurity, understanding the interaction of hydrogen with native defects in ZnO is not only of fundamental interest but also of technological importance.  [12][13][14][15][16]. Deep traps below the conduction band minimum such as E 1 (0.12 eV), E 2 (0.10 eV) and E 3 (0. 29 -0.30 eV) and E 4 (0.53 eV) were reported for ZnO grown by different methods; these traps were assigned to significant impurities or point defects in the bulk [12,[16][17][18]. Surface traps in ZnO have previously been investigated by C-DLTS [14]; however, C-DLTS signals can be severely distorted as the trapped charge can be varied as a function of applied bias. While the conventional C-DLTS is a powerful technique to investigate the behaviour of traps, it is not capable of identifying shallow trap states due to carrier freeze-out at low temperatures and immeasurably small capacitance of ZnO junctions [19,20]. Electrically active shallow defects play a pivotal role in determining the electrical and optical properties of ZnO; however, controlling the behaviour of these defects remains a major challenge. Conversely, the charge-based deep level transient spectroscopy (Q-DLTS) used in this work has been specifically developed to facilitate probing of shallow trapping states and accordingly has been applied to the investigation of these defect centers in ZnO crystals. In contrast to C-DLTS, Q-DLTS is an isothermal technique in which the transient process of trapped charge (not capacitance) is measured after voltage stimulation. In this work, combined results from Q-DLTS and variable-temperature photoluminescence (PL) and cathodoluminescence (CL) spectroscopy allow quantitative evaluation of shallow level defects in the near surface region of the as-grown and H-doped ZnO crystals.

New Journal of Physics
4 Traps in the near-surface region of a semiconductor can be investigated through the use of Q-DLTS because the sensitivity of charge transient detection is enhanced for surface states in metal-semiconductor junctions [21]. In the Q-DLTS method, cyclic bias pulses are applied to a Schottky barrier junction to excite carrier traps; the electron occupation of the traps is monitored by measuring the associated charge transients as the junction returns to thermal equilibrium. The charge transient is measured at two times (t 1 and t 2 from the beginning of discharge) and the charge Q flowed through the circuit during the period t 1  t 2 is measured as a function of rate window . For electrons, the gate timed charge difference is [22,23]: [1] where Qo is the total charge trapped during the filling pulse. The thermal emission rate e n , according to Maxwell-Boltzmann statistics, can be expressed as [21,22]: [2] where  is the capture cross section, E a the activation energy, and  n is a constant associated with the electron effective mass. The equations for holes are analogous. A common experimental approach is to keep the ratio constant; this leads to the function being maximum when the rate window is equal to the emission rate of the trap, i.e. . The activation energy of a trap and its cross section can therefore be obtained from an Arrhenius plot of equation [2], whereas the trap density can be calculated from the maximum [24].
New Journal of Physics 5

Experimental details
Identical samples from a single a-plane ZnO wafer (grown hydrothermally by the MTI Corp., USA, 0.5 mm thick) were used in this study. The crystal was polished both sides to 1 nm surface roughness. The samples of 5 mm  5 mm square sections were cleaned in acetone and ethanol then rinsed in deionised water. One sample was exposed to radio-frequency plasma in hydrogen atmosphere (power 15 mW for 1 min with the sample kept at 200°C); this doping method leads to a near-surface hydrogen rich region as described in our previous work [25,26]. No changes to the crystal structure were detected by Raman spectroscopy or X-ray diffraction. The Hall-effect characterization revealed that the as-grown and H-doped crystals have a carrier concentration of 4.4  10 15 to 7.1  10 17 cm -3 , respectively. The as-grown sample was also plasma cleaned by mild oxygen plasma to improve the reliability of electrical contacts according to the procedure as described by other workers [27].

Shallow carrier traps investigated by Q-DLTS
A typical I-V characteristic of the ZnO/Au junction, showing a rectifying behaviour, is presented in Fig. 1. The forward bias was obtained with a positive bias applied to the Au electrode. Similar rectifying junctions on c-plane ZnO using Au electrodes have been reported previously [28]. The bulk resistivity of the crystal is 15.6 ·cm at room temperature. The sign of rectifying voltage (positive on the Au electrode) indicates the presence of n-type band bending with a depletion layer for electrons at the ZnO/Au interface. The relatively large reverse current, characteristic of Au Schottky contacts to ZnO, has been attributed to tunnelling conduction [14].  [30,31], with the coupling strength found to be four times greater than that of GaN [32]. While the overall features of the CL spectra are similar for the as-grown and H-doped ZnO at temperatures below 140 K, hydrogen has a striking effect on the FX and its phonon replicas: their intensities continues to rise in intensity with increasing temperature up to 320 K. This intriguing phenomenon is clearly seen in Figure 4(b), which shows the BX and FX-1LO peaks exhibiting an opposite dependence on temperature.
The enhancement of the NBE due to H doping is more evident in Fig 4( Al shallow donors [1]. Neutral-donor-bound exciton transition I 9 at 3.357 eV, which was recently verified by time-of-flight PL [36], appears as a shoulder with intensity about an order of magnitude weaker than those of I 4 and I 6 . The presence of these impurities in the as-grown crystal is unsurprising since they are commonly present during the hydrothermal growth. For H-doped ZnO, the dominant peak at 3.362 eV (labelled I 4a ) has previously attributed to an excited state of the H-related neutral donor bound exciton I 4 [33,37]. Analysis of the BX spectral region reveals that I 4a intensity increases in comparison with that of I 4 after hydrogen incorporation suggests that hydrogen dopants introduced by the plasma is in a different chemical state to those incorporated during the hydrothermal growth. Figure 5 ZnO samples [35], possibly due to strain, this energy is in excellent agreement with previously reported activation values of 10 -12 meV for the Y line [34,35]. The Y line recombination has been found in ion implanted ZnO crystals and attributed to extended structural defects [35]; our results here indicate that similar structural defects could be introduced in ZnO by hydrogen plasma.

Discussion
The dominant electron trap E 11 in the H-doped ZnO could be associated with either hydrogen dopants or Y-line structural defects induced by the plasma process.
However, the shallow E 11 trap is unlikely to be a H-induced donor state, which has a New Journal of Physics 12 much larger ionisation energy (E i (H) = 47 -53 meV [38,39]). The nature of the shallow E 11 strap is currently undetermined in the literature, with its activation energy being significantly smaller than that of the shallowest traps detected previously by C-DLTS in hydrothermal ZnO (55 meV) [40] or in ZnO thin films deposited by pulsed laser deposition (31 meV) [41]. trap, rather than the detachment of an entire exciton, as suggested by Wagner et al. [35]. Furthermore, the Q-DLTS peak intensity of the E 11 trap does not saturate with increasing pulse-width in the range of 0.1 to 5 ms, which is highly characteristic of traps associated with extended defects that can trap multiple charges [42]. This result further supports the contention that the E 11 trap is associated with the Y line structural defects.
In Q-DLTS, the maximum depth of the detectable trap volume is the width of the depletion layer [19]. From the standard calculation of the depletion layer at the applied bias, we can estimate that the E 11 trap arises from depths up to 60 nm. Based on the above discussion, the E 11 trap is therefore be attributable to Y line structural defects in the near-surface region that are induced by the plasma process. Our reported by other workers using thermal admittance spectroscopy (TAS) [40], but this trap state was not detectable by C-DLTS, probably due to carrier freeze-out at low temperatures. The slight increase in the cross section of the E 23 trap after the hydrogen plasma might be due to reduction in the number of competitive traps as the ZnO surface is passivated by hydrogen.

Summary
In summary, Q-DLTS has been applied successfully to ZnO crystals to reveal shallow traps at 11 and 23 meV below the conduction band edge, which have not been