1. Introduction

As a wide bandgap semiconductor with a large exciton binding energy of 60 meV, ZnO material has been considered as a potential candidate for fabricating low-cost, high-efficiency ultraviolet light-emitting diodes and lasers [1,2]. However, these device applications have encountered great difficulty since stable and controllable p-type doping remains problematic. In spite of extreme difficulty and high controversy, a few outstanding experiments have shown that the p-typeness is not unachievable [1–4]. One of the current inconclusive issue is the actual form of the shallow acceptors since theoretical calculations do not support the shallowness of the simple substituting impurity [5,6]. On the other hand, the notorious self-compensation from intrinsic defects remains of a high level, indicative of low crystalline completeness and stoichiometry [7,8].

Zinc interstitials (Zn_i) and oxygen vacancies (V_O) are two frequently considered point defects that contributes to the self-compensation to the alleged acceptor-doped ZnO material [9]. The Vo and Zn_i can be easily formed due to their low formation energy of 0.02 and 1.73 eV, respectively, when the Fermi level equals the valence band maximum energy [10,11]. At the same condition, the acceptor-like defects in ZnO such as zinc vacancies (V_Zn) and oxygen interstitials (O_i) have a much higher formation energy of 5.46 and 5.88 eV, respectively [12]. Moreover, the calculated transition energy of 0/1 + for Zn_i is quite shallow, at the value of around 30 meV below the conduction band minimum energy [13]. While for V_O, the transition energy is much deeper (> 0.5 eV) [14]. Therefore, Zn_i and V_O could both act as compensating centers for acceptors. But furthermore, the Zn_i could efficiently contribute electrons to the conduction band and thus be the source of background n-type conduction.

However, the concrete form of the Zn_i in the material is still controversial. The isolated Zn_i has been proven as a fast diffuser, which cannot survive in ZnO material above 170 K [15]. Admitting this property of Zn_i, the room-temperature stabilization mechanisms of Zn_i in ZnO have been proposed. Look et al have conducted a calculation involving nitrogen, and have found that the isolated Zn_i can be stabilized in ZnO lattice by binding to a substituting nitrogen at oxygen site (N_O), forming the Zn_i-N_O complex [16,17]. Kim and Park have suggested that the isolated Zn_i could be stabilized in presence of a high concentration of V_O, forming the Zn_i-V_O complex [18]. Gluba et al have concluded that isolated Zn_i tend to form small clusters (Zn_i-Zn_i) [19]. Regardless of the proposals, isolated Zn_i requires some additional mechanism to stably exist in the ZnO lattice at room temperature.

Nevertheless, considering that a thermal annealing is an indispensable post-growth process to activate the acceptors for realizing p-typeness in ZnO [20], the stability, form, and properties of the Zn_i related complex could change accordingly. Look et al have found that due to the instability of the N_O acceptor and the diffusive nature of the Zn_i, the concentration of Zn_i-N_O would change with the annealing temperature [17]. Such changes would lead to the variation of the properties of the films. Therefore, the investigation on the thermal evolution of the form and properties of the Zn_i related complex is demanding before we can fully understand the nature and control the concentration of the complex, which are prerequisites for suppressing the compensating donors in acceptor-doped ZnO material.

However, such an investigation cannot be valid if the samples are of low crystalline quality. Commonly, in order to pursue higher doping concentration of potential acceptor-like dopants, materials are prepared under a relatively low temperature, especially for the near-equilibrium growth techniques like metal-organic chemical vapor deposition (MOCVD) [21–23]. In this case, the fabricated thin films are actually a combination of tightly arranged column nano-crystals. Such morphology naturally results in a huge number of alleged “grain boundaries”, where high density of defects and impurities co-exists due to the low crystalline quality of the boundaries. In fact, p-type regions can only be found near the grain boundaries. While the inner grain region, although exhibiting higher crystalline quality, is almost un-doped [24]. Therefore, most of the measured optical and electrical properties of the “thin films” are actually the ones of the grain boundaries. In other words, the discussions related to the properties of the acceptor-like dopants and compensating donors are meaningless unless the crystalline quality of the thin film samples being improved.

In order to balance the dopant solubility and the crystalline quality, we have developed a method. We have employed a 2-micron-thick high-quality ZnO film on sapphire (marked as ZnO template) as the substrate for the sample growth. After proper optimization, the crystalline quality of the films grown on the templates can be significantly improved while maintaining efficient nitrogen incorporation as compared to the samples grown on bare sapphire. Utilizing these films as samples and combining the characterizations from Raman spectroscopy, Hall, capacitance-voltage (C-V), and X-ray photoelectron spectroscopy (XPS), we have investigated the thermal evolution of Zn_i related compensating donors via changing the processing temperature of the post-annealing.

2. Experimental details

2.1. Sample growth and post-growth thermal treatment

Five samples were employed in this paper including an as-grown sample and four annealed samples. The as-grown sample was grown by a home-built MOCVD apparatus on the abovementioned template substrate. The template is of high crystalline quality with a ω-scan width of 266 arc sec from X-ray diffraction measurement. The surface roughness is less than 5 nm from the observation of an atomic force microscope [25,26]. Prior to the ZnO film growth, the template was pre-treated in a N₂O atmosphere at 1000 °C for 5 min in order to create nucleation sites and growth steps. After that, in situ growth of a piece of N-doped ZnO film (sample A) was done at 470 °C for 30 min. 20 SCCM (standard cubic centimeter per minute) of di-methyl zinc (DMZn) were diluted and carried by high-purity N₂ gas into the growth chamber. N₂O and NH₃ were used as the oxidizing and nitrogen doping sources at the flow rates of 1000 SCCM and 20 SCCM, respectively. A low temperature plasma system was employed to enhance the dissociation of N₂O and NH₃. The chamber pressure was controlled at 0.2 atm (20 kPa).

After growth, the sample A was cut into a few 5 mm × 5 mm pieces. Each piece of the sample A was transferred into the processing chamber of an Ecopia Rapid Thermal Processing System (RTP-1200), and a 5-min rapid thermal process (RTP) in a N₂O (10 SCCM) and N₂ (100 SCCM) mixed atmosphere was carried out at 600 °C (sample B), 700 °C (sample C), 800 °C (sample D), and 900 °C (sample E), respectively. Annealing below 600 °C cannot effectively eliminate incorporated isolated interstitial hydrogen, while above 1000°C, the as-grown ZnO film would begin to dissociate and form a very rough surface. Therefore, the RTP temperature was selected between 600 °C and 900 °C in a step of 100 °C. The mixed ambience of N₂O and N₂ is to provide a nitrogen-rich condition that protects nitrogen from being desorbed.

2.2. Characterizing methods

To compare and characterize the crystalline and surface quality of the samples, we have utilized high resolution X-ray diffraction (HRXRD, Philips X'pert Pro diffractometer, copper-target Kα₂ X-ray at the wavelength of 0.15405 nm) and an atomic force microscope (AFM, Nanoscope IIIa, Digital Instruments, Inc.). The vibrational properties of the samples were recorded by a JOBIN YVON HR800 UV Raman system in the backscattering geometry excited by an argon-ion laser (λ = 514.5 nm) at room temperature. The electrical properties were obtained by a Keithley Hall system and a high-frequency C-V system. The Hall test is under van der Pauw’s configuration with indium as contacts. The applied magnetic field is 2 kGs and the applied constant current is 100 μA. The frequency and the amplitude of the alternating current signal for the C-V measurement is 1 MHz and 50 mV, respectively. Mercury probes with different areas were utilized as the contacts for the C-V measurement. The chemical configuration of the elements was determined by XPS with an Al Kα X-ray monochromatic source at 1486.6 eV. Argon etching was performed to eliminate the surface contaminations before collecting the signals. All the spectra were calibrated to the contaminated C 1s line at 284.8 eV.

3. Results and discussion

3.1. The morphological and crystalline quality

The crystalline and morphological quality of the as-grown sample has been checked by the AFM surface scan and the X-ray rocking curve (XRC). Figures 1(a) and 1(b) firstly compare the typical morphologies of the samples grown on the bare c-plane sapphire and the template, respectively. Clear distinction can be found. The sample grown on a bare sapphire is quite rough with very small grains. While the sample grown on the template exhibits a smooth surface with much larger grains. Such a conspicuous difference could be due to the different lattice matching level. Since the other growth conditions are similar, meaning that the adatom adsorption ability and the chemical reaction rate are similar, the rough surface for the sample on bare sapphire should be due to the huge lattice mismatch (~18%) between the (0001) sapphire and the (0002) ZnO crystalline surfaces. As the initial nucleation sites grow larger, the lateral growth rate is significantly reduced because further lateral adhesion of the adatoms to the nucleation sites becomes much harder, requiring additional energy to overcome larger strain. In this case, adatoms would tend to make a new nucleation site, rather than growing on an existing one. This could be the reason why the grain size is quite small for the sample on bare sapphire. On the contrary, due to the high quality of the template and the proper pre-treatment to the template, the growth of ZnO on the template is nearly homo-epitaxial. Therefore, the resulted film sample is much smoother and denser [27].

 

Fig. 1 The AFM images of the nitrogen-doped ZnO films grown on (a) the bare sapphire, and (b) the ZnO template. The area of the scan is 2.5 × 2.5 μm². The z-scale bar is 100 nm. (c) The ZnO (0002) XRC of the as-grown NH₃-doped sample and the samples annealed ranging from 600 to 900 °C. (d) The extracted FWHM values of the XRC for all the samples.

Download Full Size | PDF

Regarding the crystalline quality, Fig. 1(c) shows the rocking curves of the ZnO (0002) diffraction peaks for all the as-grown and annealed samples. The full width at half maximum (FWHM) values are in the order of 0.1 degree. These values are slightly higher than that of the template (~0.074 degree, or 266 arc sec) [26], but are one order of magnitude lower than that of the sample grown on a bare sapphire (XRC measured on the nitrogen-doped samples in Ref [21], giving the FWHMs of around 1 degree), indicative of significantly improved crystalline quality for the samples grown on the templates. Besides, when we plot the FWHM values of all samples in Fig. 1(d), the trend shows that the FWHM value increases firstly from 0.123 to 0.140 degree as the annealing temperature increases to 700 °C, and then decreases to 0.120 degree as the annealing temperature further increases to 900 °C. This result indicates that the crystalline quality degrades first and improves again as increasing the annealing temperature. The sample annealed at 700 °C has a relatively lowest crystalline quality. Combining the AFM and XRC results, it can be concluded that the samples on the templates provide a valid platform for further investigation on the properties of the films.

3.2. The forms of zinc interstitials and their thermal evolution

In order to understand the trend of the crystalline quality as a function of the annealing temperature, Raman backscattering spectroscopy has been utilized since the intensity of the characteristic vibrational modes could reflect the crystalline quality of the samples. The recorded spectra are shown in Fig. 2(a). Regarding the assignment, the modes at 416, 576, and 750 cm⁻¹ are attributed to the sapphire substrate, which are denoted as asterisks (*) in the figure. Besides, the modes at 99, 332, 378, 437 cm⁻¹ are assigned to the ZnO-lattice-related E₂(L), E₂(M), A₁(TO), and E₂(H) modes, respectively [28]. Fig. 2(b) draws the extracted intensity of the E₂(L) and E₂(H) modes. As can be seen from the trend, the intensity of the two modes simultaneously increases as a function of the annealing temperature. Since the E₂(L) and E₂(H) modes can represent the lattice quality of Zn sub-lattice and O sub-lattice, respectively [29,30], the monotonous increase of the intensity of the two E₂ modes indicates that the quality of the ZnO lattice increases as the annealing temperature. In consequence, some other reasons should be responsible for the wider width of the XRC for samples annealed below 700 °C.

 

Fig. 2 (a) The Raman backscattering spectra (50-800 cm⁻¹) of the as-grown NH₃-doped sample and the samples annealed ranging from 600 to 900 °C. The extracted integral intensity of (b) the E₂(L) and E₂(H) modes, and (c) the four AMs for all the samples.

Download Full Size | PDF

Beside the modes mentioned above, additional modes (AMs) at 275, 510, 581, and 644 cm⁻¹ are observed for all samples. The mode at 275 cm⁻¹ has been observed in the past, on the contrary, its origin has been argued for a period of time. The mode was firstly reported in the Raman spectra on nitrogen-doped ZnO material. The intensity of the mode shows a direct proportion to the nitrogen concentration in the sample, so the authors have ascribed the mode to localized vibrational mode (LVM) from the N_O acceptors [31]. However, Bundersmann et al found the same mode in ZnO material doped with gallium [32]. Wang et al further found the mode in ZnO samples doped with manganese [33]. All these prior papers have demonstrated that the mode at 275 cm⁻¹ is not related to a specific impurity.

Nevertheless, Wang et al have discussed that although the mode intensity increases with the impurity concentration, the origin is not related to the impurity itself, but is a result of the changed environment around zinc atoms due to the nitrogen incorporation [33]. Artús et al have pointed out that the emergence of the mode should be related to the change of the ZnO lattice by forming a kind of zinc-related complex defects induced by additional impurity doping [34]. In 2013, Gluba et al have designed an experiment to investigate the origin [19]. A detailed investigation of the influence of ⁶⁴Zn and ⁶⁸Zn isotopes on the mode has been presented. Combining the isotopic shift and the line-width of the vibrational mode, the results have suggested that the mode is caused by Zn-Zn vibrations from small zinc interstitial clusters containing approximately 3 – 9 zinc atoms. Considering the above researching progress, the mode at υ = 275 cm⁻¹ are hereby attributed to the zinc interstitial related LVM, particularly the Zn_i small clusters.

Concerning the other three AMs, they appear simultaneously with the 275 cm⁻¹ mode. And moreover, their intensity trends as a function of the annealing temperature change in accordance with that of the 275 cm⁻¹ mode, as illustrated in Fig. 2(c). In fact, reviewing the Raman spectra in the literatures [31–34], the other three AMs always co-exist with the 275 cm⁻¹ mode. Therefore, we also attributed the modes at 510, 581, and 644 cm⁻¹ to the vibrations related to Zn_i clusters. One thing should be mentioned here that the mode at 581 cm⁻¹ could be overlapped with the phonon mode at 577 cm⁻¹, which is commonly ascribed to a resonant enhancement of the E₁ and A₁(LO) phonons in ZnO [35]. The origin of the Fröhlich scattering is attributed to the presence of localized states in the optical band gap of ZnO. Thus, it can be used as a benchmark to determine the quality of the ZnO material. However, the A₁(LO) mode is usually very weak in ZnO due to the virtual compensation of the two scattering terms given by the deformation potential and the electro-optic coupling coefficient with results for the LO phonons from its macroscopic electric field [36]. Therefore, we do not take the A₁(LO) mode into account when we extract the integral intensity of the 581 cm⁻¹ mode.

Utilizing the intensity of the AMs as fingerprints, we can obtain the thermal evolution for the Zn_i clusters. As can be seen from Fig. 2(c), the concentration of Zn_i clusters increases initially, and reaches a maximum for the 700 °C-annealed sample. For the samples annealed at higher temperatures (800 and 900 °C), the concentration of the Zn_i clusters starts to gradually decrease. The variation of the concentration of the Zn_i clusters indicates the transformation of the zinc atoms. For the samples annealed at 600 and 700 °C, the Zn atoms aggregate to form the Zn_i clusters. Considering that the quality of both the Zn and O sub-lattices improves as the annealing temperature increases, the Zn atoms forming the Zn_i clusters cannot come from the lattice zinc, but from the existing Zn_i defects in the as-grown sample.

As already mentioned in the Introduction, the isolated Zn_i is regarded as a fast diffuser with a low migration barrier, so the isolated Zn_i cannot be the form of the Zn_i related defects in the as-grown sample. Due to that, our samples are NH₃ doped and according to a previous literature [17], it is highly possible that the part of the Zn_i related defects in the as-grown sample are in the form of Zn_i-N_O. The experiments by Look et al have demonstrated that the thermal stability of the Zn_i-N_O complex is quite poor. Only 50% of the Zn_i-N_O complexes remain after 500 °C annealing, and almost no such complexes could survive after an annealing at 950 °C [16]. The reason could be due to that the desorption rate of the N_O drastically increases as the annealing temperature increases. The desorption of the N_O in the Zn_i-N_O complexes would naturally leave plenty of isolated Zn_i atoms in the material. Our results show that during the annealing process under moderate temperature (600 – 700 °C), the isolated Zn_i atoms could aggregate to form Zn_i clusters. In fact, part of the Zn_i related defects are in the form of Zn_i clusters for the as-grown sample as shown by Raman. The stability of Zn_i small clusters over that of the isolated Zn_i and Zn_i atoms binding extrinsic defects in ZnO lattice is related to the participation of the O 2p orbitals, which hybridize with the Zn_i-Zn_i bonding (σ) and anti-bonding (σ*) orbitals, making the charge occupying the σ* be transferred to the conduction band [19]. This mechanism not only lowers the total energy of the Zn_i-Zn_i complex but also makes it act as a shallow donor. Of course, our results also show that as the annealing temperature further increases above 700 °C, the Zn_i clusters become less stable. As increasing the annealing temperature, more and more clusters begin to decompose, possibly to isolated Zn_i first, and then desorb from the film.

From the above discussion, it is not difficult to understand the consistent trend between the XRC FWHM values [Fig. 1(d)] and the Raman mode intensity at 275 cm⁻¹ [Fig. 2(c)] because the formation of additional Zn_i clusters should be responsible for the deteriorated crystalline quality. The formation of the Zn_i clusters is established on the assumption that the N_O acceptors should desorb after annealing. In order to verify this, XPS of nitrogen 1s lines for the as-grown and two annealed samples (at low temperature 600 °C and high temperature 900 °C) have been measured and the spectra are shown in Figs. 3(a)–3(c). The nitrogen signals can be detected for all samples, demonstrating that nitrogen dopants have been incorporated in the samples with high crystalline quality. However, the signals are quite weak due to the very limited solid solubility of nitrogen in well-ordered ZnO lattice [37]. In spite of weak signals, differences between the three samples can still be found.

 

Fig. 3 The N 1s XPS fine scan (390 – 406 eV) of (a) the as-grown NH₃-doped sample, (b) the 600 °C-annealed sample, and (c) the 900 °C-annealed sample. (d) The Zn L₃M₄₅M₄₅ XPS fine scan (480 – 525 eV) for the as-grown, 600 °C-annealed, and 900 °C-annealed samples. The inset shows the intensity ratio between the Zn_i-related peak and the Zn L₃M₄₅M₄₅ peak.

Download Full Size | PDF

For the as-grown sample, the spectrum could be deconvoluted into two components located at 397 and 400 eV, respectively. The 397-eV-peak is ascribed to nitrogen at anion site, and particularly here the N_O acceptors [38]. While the 400-eV-peak is ascribed to nitrogen binding additional hydrogen, i.e., the N-H bonds [39]. The formation of the N-H bonds is quite easy to understand because (1) hydrogen is ubiquitous in the samples grown by MOCVD [40]; and (2) the NH₃ is employed as the nitrogen source, which provides naturally abundant N-H bonds. For the annealed samples as shown in Fig. 3(b) and 3(c), the 397-eV-component can hardly be deconvoluted, which is direct evidence that the N_O defects desorb from the annealed samples. Although the N_O component has been eliminated in the annealed samples, nitrogen in the form of N-H can even survive after 900 °C annealing. The prominent thermal stability of the N-H bonds could be due to the formation of the NH_x complex at zinc site [(NH_x)_Zn]. Bang et al have shown that zinc-sited nitrogen (N_Zn) binding additional hydrogen could reduce the repulsive interactions between the N_Zn and V_Zn, and thus lower the formation energy of the N_Zn, In their calculation, the formation energy of the zinc-sited ammonia molecule [(NH₃)_Zn] can be 2 meV lower than that of V_Zn [41]. Therefore, it is reasonable to say that the residual nitrogen in the annealed samples could be in the form of (NH_x)_Zn.

3.3. The influence of Zn_i related donors on the electrical properties

As the Zn_i-N_O and the Zn_i small clusters are both shallow donors, their thermal evolution would change the electrical properties of the samples. Hall test and C-V measurement have been thus employed to investigate this. Actually, either the as-grown sample or the annealed samples are highly compensated. Their resistance is much higher than that of the ZnO template. Besides, the thickness of the grown film (~200 nm) is less than 1/10 of the template and 10⁻⁴ of the distance between indium contacts. In this case, the measured Hall results of the samples are similar to that of the template, giving no useful information on the electrical properties of the samples. Therefore, we have adopted the C-V measurement to characterize the electrical properties.

Figure 4(a) shows the C-V curves for all the samples. The C-V curve for the template is also shown for comparison. It can be clearly seen that donors are the majority in the as-grown sample while acceptors are major in all the annealed samples. Utilizing the relationship of dC⁻²/dV versus bias the net doping concentration |N_D-N_A| [42], the net doping concentration for all the samples are shown in Fig. 4(b). The result shows that the net doping concentration (N_D-N_A) of the as-grown sample is 2.8 × 10¹⁴ cm⁻³. After annealing at 600 °C, acceptors become the majority. As increasing the annealing temperature, the net doping concentration (N_A-N_D) slightly increases from 4.0 × 10¹⁵ to 6.6 × 10¹⁶ cm⁻³. This result indicates that the concentration of the acceptors increases while that of the donors decreases as the annealing temperature increases.

 

Fig. 4 (a) The C-V curves for the as-grown NH₃-doped sample and the samples annealed ranging from 600 to 900 °C. The up-right inset shows the C-V curves for a p-type silicon wafer and the n-type ZnO template for assisting the judgement of the conduction type of the samples. The up-left inset draws the schematic for the C-V measurement setup. (b) The calculated net doping concentration values [(N_D-N_A) (for donor majority) and (N_A-N_D) (for acceptor majority)] for all the samples.

Download Full Size | PDF

With regard to the donors, although the concentration of the Zn_i small clusters has a slight increase from the as-grown sample to the 700 °C-annealed one. The total amount of the Zn_i-related donors (including the Zn_i-N_O and the Zn_i clusters) should decrease with increasing the annealing temperature [16,17]. Fig. 3(d) shows the XPS fine scan at the range around Zn Auger line. A shoulder could be deconvoluted from the main peak of the Zn L₃M₄₅M₄₅ Auger peak. This shoulder has been suggested to have some relation with the zinc atoms at interstitial state [43]. The inset of Fig. 3(d) shows the extracted intensity ratio of the shoulder over the main Auger peak. The monotonous decreasing trend demonstrates experimentally that the amount of the Zn_i related species decreases as increasing the annealing temperature. Moreover, isolated hydrogen interstitials would desorb from the material when the sample is annealed above 600 °C [20]. This could be another reason for the reduced concentration of donors.

In fact, as the annealing temperature increases, the desorption and diffusion of oxygen-sited nitrogen atoms not only leave plenty of isolated Zn_i, but also form some oxygen vacancies. Therefore, it can be inferred that the concentration of the oxygen vacancies might increase as increasing the annealing temperature. However, due to the limited solubility of nitrogen, the newly formed oxygen vacancies from the desorption and diffusion of oxygen-sited nitrogen atoms may not have a large amount. As a result, the possible change of the concentration of oxygen vacancies may not have an obvious effect on the C-V measured net doping concentration as what the zinc interstitial related donors have done.

With regard to the acceptors, the desorption of the N_O would decrease the total acceptor concentration. However, seen from another aspect, the desorption of the N_O has resulted in the isolation and aggregation of the Zn_i. The isolation of the Zn_i atoms would eventually lead to the out-diffusion of them. The bonds between the isolated Zn_i seem to enhance the ability of donating electrons. In fact, due to the combination of the isolated Zn_i atoms, the bonds between the Zn_i atoms would consume some valence electrons. As a result, the donating ability of the Zn_i clusters should be comparable to that of the Zn_i-N_O complexes. In this case, the net effect of the desorption of the N_O acceptors is more like a suppression of shallow donors, which is beneficial to realizing net p-typeness. Moreover, as shown by the nitrogen 1s XPS in Fig. 3, the (NH_x)_Zn complexes are the main forms of nitrogen in the annealed samples. If x equals 4, the (NH₄)_Zn complex could act as a shallow acceptor. Early papers reporting the NH₃-doped ZnO materials for realizing the p-typeness have not considered such a possibility [44–46]. Meanwhile, the V_Zn related green band emission has been found to emerge for the 900 °C-annealed sample (not shown) [47,48]. According to a previous paper, the aggregation of V_Zn to form V_Zn clusters occur at 1000 °C in the presence nitrogen in the ZnO material [49,50]. Provided the presence of the large amount of V_Zn and the V_Zn clusters being proven as shallow [51], it is reasonable to infer that the V_Zn clusters could be another source of free holes.

4. Summary and conclusion

In this work, we have successfully fabricated nitrogen-doped ZnO film on the ZnO template substrate. As compared to the samples grown on bare sapphires, the crystalline quality as well as the morphological smoothness have been improved significantly. On the platform of the high-quality material, we have extensively investigated the thermal evolution of the forms, concentration, and electrical properties of the Zn_i related defects. The process has been schematically summarized in Fig. 5. As can be seen, the Zn_i-N_O complex and the Zn_i small clusters are two forms of the Zn_i related defects in the as-grown sample. As performing annealing on the as-grown sample, the desorption of the N_O defects would naturally leave plenty of isolated Zn_i atoms. Some of the isolated Zn_i atoms would diffuse out of the film, and the others tend to aggregate to form Zn_i small clusters. Further annealing at higher temperatures would cause the Zn_i small clusters to decompose into isolated Zn_i atoms again, and finally desorb from the sample. Electrical characterizations have shown that the acceptors increase and donors decrease with increasing annealing temperature. The suppression of the donors is related to the desorption of the interstitial hydrogen and the Zn_i related donors. While the enhancement of the acceptors is possibly due to the formation of the acceptor-like (NH₄)_Zn and V_Zn clusters. A slight net acceptor majority has been realized for the annealed samples.

 

Fig. 5 The schematic summary of the thermal evolution of the Zn_i-related donors in nitrogen-doped ZnO film grown on the ZnO template.

Download Full Size | PDF

As the Zn_i related defects are quite notorious as compensating donors to p-type-doped ZnO material, a clear understanding of their properties, as shown by this paper, is an indispensable prerequisite for further controlling and suppressing them. Meanwhile, the doping of the NH₃ molecule could be beneficial to introducing shallow acceptors because the ammonia molecule at zinc site binding an additional hydrogen [(NH₄)_Zn] is energetically favorable while shallow.