Comparative study on beryllium and magnesium as a co-doping element for ZnO:N
Su Yu-Quan1, †, , Chen Ming-Ming2, 3, †, , Su Long-Xing1, Zhu Yuan1, ‡, , Tang Zi-Kang1, 3
School of Physics and Engineering, Sun Yat-sen University, Guangzhou 510275, China
Faculty of Science, Jiangsu University, Zhenjiang 212013, China
Department of Physics, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong, China

 

† These authors contribute to this work equally.

‡ Corresponding author. E-mail: zhuy9@mail.sysu.edu.cn

Project supported by the National Key Basic Research Program of China (Grant No. 2011CB302000), the National Natural Science Foundation of China (Grant Nos. 51232009 and 51202299), the Fundamental Research Funds for the Central Universities, China (Grant No. 11lgpy16), the Natural Science Foundation for Jiangsu Provincial Higher Education, Institutions of China (Grant No. 15KJB510005), and the Talent Fund of Jiangsu University, China (Grant No. 15JDG042).

Abstract
Abstract

Stable nitrogen doping is an important issue in p-type ZnO research for device applications. In this paper, beryllium and magnesium are systematically compared as a dopant in ZnO to reveal their nitrogen-stabilizing ability. Secondary ion mass spectrum shows that Be and Mg can both enhance the stability of nitrogen in ZnO while Be has a better performance. Zn 2p and O 1s electron binding energies change in both MgZnO and BeZnO thin films. Donor-acceptor luminescence is observed in the BeZnO samples. We conclude that Be is a better co-doping element than Mg for p-type ZnO:N.

PACS: 61.72.uj;68.55.Ln;78.55.–m;82.80.Pv
1. Introduction

ZnO has wide potential applications in ultra-violet optoelectronic devices mainly because of its band gap of 3.37 eV and exciton binding energy of 60 meV.[13] Fabrication of p-type ZnO, however, is still a main obstacle in achieving ZnO-based electroluminescence devices. Previous studies have shown that nitrogen as a doping element can produce shallow acceptor energy level and thus become a promising dopant for high hole concentration, high stability and reproducible p-type ZnO.[4,5] Direct doping of nitrogen into ZnO thin film is found to be very difficult in non-equilibrium growing experiments performed by most researchers. Co-doping is an effective way to solve this problem. Lithium, magnesium, gallium or other elements have been doped into ZnO together with nitrogen to acquire a more intense interaction between nitrogen and the lattice.[610] As isoelectronic cations in ZnO, beryllium and Mg have also attracted attention.[11] Wei et al. fabricated MgZnO:N by molecular beam epitaxy (MBE) and detected p-type conductivity after annealing.[12] Sanmyo et al. succeeded in fabricating Be–N co-doped ZnO with weak p-type conductivity by sputtering.[13] These results suggest the nitrogen-stabilizing effect of Mg and Be, which are not in agreement with calculations.[14,15] Comparison between the two neighboring group IIA dopants in ZnO helps to reveal their nitrogen-stabilizing ability, and provides guidance for further p-type doping. Differences in fabricating method and condition impede the systematic comparison between previous reports. Here we use molecular beam epitaxy, a clean growing method to avoid other disturbance, to fabricate MgZnO and BeZnO thin films under the same condition. By comparing our samples, we find that Mg and Be can both greatly enhance the stability of nitrogen in ZnO, while the nitrogen in BeZnO:N thin film stays stable even after high-temperature annealing. The changes in binding energies of Zn 2p and O 1s electrons imply an increase of oxygen content in the lattice. Luminescence from a donor–acceptor pair (DAP) is observed in BeZnO thin film. We deduce that the change in anion/cation ratio by doping affects the chemical potential of nitrogen and leads to the enhanced nitrogen stability. By comparing our results, we find that Be is better than Mg in nitrogen-stabilization. BeZnO would be a promising starting material for p-type doping.

2. Sample fabrication

The thin crystal films were grown on sapphire (0001) substrates by plasma-assisted MBE. Zn (6N), Be (4N) and Mg (6N) were supplied by solid source effusion cells. The O and N atoms were supplied from O2 and N2 gas through RF plasma generators. A designed buffer was employed to relieve the strain as much as possible and thus high crystal quality was achieved. The details of the buffer technique can be found in our previous work.[16] The doping content was controlled by the source temperature, i.e., higher source temperature corresponds to a higher doping level. In the nitrogen-doped samples, the flow rate was kept constant. The MgZnO:N thin films are n-type or highly resistant. The BeZnO:N thin films are highly resistant or weakly p-type, with a hole concentration of ∼ 1 × 1015/cm3 and a mobility of 2 cm2·V−1·S−1. The electrical properties of the films are stable at least for three months.

3. Experimental results and discussion
3.1. Secondary ion mass spectrum measurement

Secondary ion mass spectrum (SIMS) is used to investigate the relative nitrogen density in as-grown and annealed ZnO, BeZnO, and MgZnO films. Figure 1 shows the SIMS results of the nitrogen-doped films before and after annealing at 500 °C in N2 atmosphere. Nitrogen behaves differently in three samples. In the pure ZnO, the nitrogen counts are rather low even before annealing. In the MgZnO alloy, the nitrogen signal is strong in the as-grown sample but declines after annealing. In the BeZnO alloy, the nitrogen signal, although not as intense as in MgZnO for as-grown samples, becomes stronger and more uniform after annealing. Actually, the annealed BeZnO alloy has a nitrogen content value comparable to the annealed MgZnO alloy, both in concentration and uniformity. Moreover, as an experimental fact, MgZnO takes the nitrogen-stabilizing effect only for an Mg content value larger than ∼ 10% while BeZnO can readily show the effect for a Be content value less than ∼ 2%. Be as a co-doping element may be more powerful in stabilizing N dopant. These can be in part explained from the aspect of bond energy. The standard formation enthalpy for Zn3N2 is −20 kJ·mol−1, which is much smaller than that for ZnO (−348.28 kJ·mol−1).[17] This leads to the poor solubility of nitrogen in ZnO. On the other hand, the formation enthalpy values for Mg3N2[18] and Be3N2[19] are −461.08 kJ·mol−1 and −570 kJ·mol−1, respectively. As a result, Mg and Be can help to ameliorate the stability of nitrogen in ZnO. Our SIMS measurement demonstrates that nitrogen prefers MgZnO than BeZnO in as-grown samples, while BeZnO:N is stabler than MgZnO:N after annealing. This is due to the stronger Be–N bond.

Fig. 1. Secondary ion mass spectra for as-grown and annealed ZnO, BeZnO, and MgZnO doped with nitrogen. The content values of Mg and Be are ∼ 10% and ∼ 1% respectively.

The high concentration of nitrogen in the as-grown MgZnO implies that nitrogen can be easily doped into the lattice during growth. But the nitrogen losses when the system process proceeds towards thermal equilibrium state. These results agree with those given by Gai et al., who predicted a higher formation energy of neutral defect NO in MgZnO than in ZnO, but succeeded to fabricate MgZnO:N by modifying chemical potentials during nonequilibrium growth.[6] On the other hand, nitrogen in BeZnO seems to be very stable even under thermal equilibrium state. However, this experimental fact is somewhat against the calculation which predicted that incorporating Be into ZnO:N can lower the ionization energy but disadvantage the stability of nitrogen.[15] To clarify the origin of this divergence, we reconsider the equation of formation energy used in calculation. The formation energy of neutral defect NO in ZnO is given by

where E(N,0) is the total energy of the supercell with neutral NO, E (host) is the energy of the original supercell, μO and μN are the chemical potentials of constituents O and N relative to the element gas energy, nO and nN are the numbers of O and N. The system is less stable with higher formation energy. Calculations indicate that (E(N,0)–E(host)) is higher in MgZnO[6] and BeZnO[15] than in pure ZnO. That induces a lower formation energy of NO in ZnO than in BeZnO and MgZnO, which is in contradiction with our experimental results. Therefore, the chemical potential, which is kept constant in most calculations, needs to be considered carefully.

The chemical potential is strongly affected by stoichiometry of the lattice. Our previous work indicates that the anion/cation ratio in BeZnO alloy may change from 1 to over 1.1 by increasing Be doping. The positron annihilation spectrum also evidences that the oxygen vacancies are obviously suppressed by Be doping.[20] But no data have ever addressed that point for MgZnO. Here, on that point, we provide more detailed spectrum evidence for both BeZnO and MgZnO: photoluminescence (PL) and x-ray photoelectron spectrum (XPS) measurements are carried out. We compare samples with and without nitrogen doping to exclude the influence of nitrogen. It is found that no change is caused by nitrogen doping in the spectrum region we are interested in. This is because the nitrogen concentration is rather low compared with those of Be and Mg.

3.2. X-ray photoelectron spectrum

Figure 2(a) shows XPS Zn 2p3/2 binding energy spectra from ZnO, MgZnO, and BeZnO. Both dopants can increase the binding energies. For Mg doping, the peak shifts to higher energy in Mg0.12Zn0.88O and then shifts back in Mg0.2Zn0.8O. The shift for BeZnO with a wide range of Be content is larger than for MgZnO. The variation of the Zn 2p binding energy is related to the ionicity of ZnO.[21,22] When zinc atoms are bonded to oxygen, its core level binding energy will increase. It means the binding energy of Zn 2p electron should increase with increasing oxygen content. Therefore, the ratio of oxygen atoms to metal atoms is increased by Mg or Be doping, and Be seems to be more efficient. This agrees with the oxygen content obtained from XPS in Table 1. In ZnO, oxygen vacancies are abundant and the crystal is not stoichiometric, i.e., there are more Zn atoms than O atoms. This situation changes in these alloys. Oxygen vacancies are suppressed by Mg and Be. Note that the PL and XPS results mentioned above are observed in BeZnO and MgZnO in the case without nitrogen doping. The advantages brought by Be or Mg doping may also be beneficial when using acceptor dopants other than nitrogen.

Table 1.

Oxygen content of ZnO films doped with different content values.

.

Modification of the O 1s binding energy by doping is also observed. As shown in Fig. 2(b), the O 1s peak is composed of at least two components. The main component located at approximately 530 eV originates from oxygen at the ideal lattice site. It shifts slightly with doping due to the difference among the 1s bonding energies of oxygen bonded to Zn, Mg, and Be. The shoulder located at about 532 eV has been reported in ZnO[23,24] and other oxides.[25,26] It originates from oxygen atoms near the deficient region of the lattice. The shoulder peaks are weak in ZnO and MgZnO but become remarkable in BeZnO. This shoulder peak should relate to oxygen atoms near the acceptor defects in our study.

Fig. 2. XPS results of Zn 2p 3/2 (a) and O 1s (b) electrons of ZnO with different dopants.
3.3. Photoluminescence spectrum

The PL spectra of ZnO, MgZnO, and BeZnO are shown in Fig. 3(a). Only luminescences from free exciton (FX), exciton bonded to neutral donor (D0X) and their replicas are observed in ZnO. In MgZnO, the broad band on the low energy side of D0X (denoted as Δ) becomes stronger in intensity. This implies the existence of a new peak such as luminescence from the donor–acceptor pair (DAP), which was also observed by Li et al.[27] As this peak overlaps with the replicas of D0X and FA, it is difficult to locate it precisely. This is the case for a wide range of Mg content (judging from the enlargement of band gap). In the BeZnO sample, the band gap is barely affected since the Be content is kept low (2%). But an intense new peak appears at 3.316 eV. To investigate the origin of this peak, we measure the temperature-dependent PL of BeZnO (Fig. 3(b)). When the temperature decreases from 77 K to 25 K, another peak appears at ∼ 3.3 eV and becomes dominant gradually. The position of the 3.3 eV peak blue-shifts with increasing laser power (not shown in the figure). We attribute this peak to the DAP luminescence peak. The energy of the photon resulting from radioactive recombination of DAP is given by[28]

where Eg is the band gap energy, are the binding energies of neutral donor and acceptor, respectively, rDA is the distance between the donor and the acceptor in the lattice. The DAP luminescence is contributed by the recombination of DAPs with different rDA values and thus different energies. When the exciting power increases, contributions from DAPs with smaller rDA values increase. Therefore, the peak blue-shifts. At higher temperature, the DAP recombination is quenched and the peak is replaced by luminescence from the transition of free electrons to acceptors (FA). This temperature-dependent evolution is a typical feature of DAP. The BeZnO sample used here is not intentionally doped with other elements, so the acceptor is attributed to being VZn.[27] ZnO always presents a dense n-type background. The existence of this acceptor may help to compensate for this.

Fig. 3. Photoluminescence spectra of (a) ZnO, MgZnO, and BeZnO at 77 K and (b) BeZnO at different temperatures.
3.4. Further discussion

All the above evidence points to the same conclusion: the anion/cation ratios in BeZnO and MgZnO alloys increase and the anion/cation ratio is much more remarkable in BeZnO. This conclusion may be a way out for the problematic calculations mentioned above, because most of the calculated values of (E(N,0)–E(host)) are obtained on an assumed lattice with anion/cation ratio r = 1, which is no longer valid when the doping elements significantly affect the anion/cation ratio and break the balance of intrinsic defects. We suggest that it is more reasonable for the calculation work to take the variation of anion/cation ratio into account. The degree of the variation may take the experimental data as a reference. Whether this modification will do any favors to the model is still an open question waiting for its answer.

4. Summary

In this paper, we fabricate high-quality MgZnO:N and BeZnO:N alloy films by MBE under a similar growth condition. The nitrogen stability is confirmed by SIMS in both alloys. To study the nitrogen-stabilizing mechanism, we examine the BeZnO and MgZnO alloys in more detail by PL and XPS. In PL spectra, an additional donor–acceptor pair irradiative emission is observed in BeZnO while in MgZnO the emission signal is not so intense. This new emission peak implies the decrease of O vacancy and (or) the increase of Zn vacancy. On the other hand, XPS indicates that the content values of oxygen are increased in both alloys, which accords well with the implication from the PL spectrum. These phenomena are all related to the doping-induced modification of host element atomic ratio, which is not considered in calculations. Also from our experiments, the N stabilization and O vacancy suppression are both more prominent in BeZnO than in MgZnO. Thus Be is a better co-doping element than Mg for p-type ZnO:N.

Reference
1Hwang D KOh M S 2007 J. Phys. D: Appl. Phys. 40 R387
2Tang Z KWong G K LYu PKawasaki MOhtomo AKoinuma HSegawa Y 1998 Appl. Phys. Lett. 72 3270
3Lu Y MLiang H WShen D ZZhang Z ZZhang J YZhao D XLiu Y CFan X W2006J. Lumin.119228
4Tsukazaki AOhtomo AOnuma TOhtani MMakino TSumiya MOhtani KChichibu S FFuke SSegawa YOhno HKoinuma HKawasaki M2005Nat. Mater.442
5Wang XLu Y MShen D ZZhang Z ZLi B HYao BZhang J YZhao D XFan X W2007J. Lumin.122165
6Gai Y QYao BWei Z PLi Y FLu Y MShen D ZZhang J YZhao D XFan X WLi JXia J B2008Appl. Phys. Lett.7562110
7Kumar MKim TKim SLee B 2006 Appl. Phys. Lett. 89 112103
8Zheng H MFang X YCai L X 2014 Chin. Phys. 23 126102
9Collado-Fregoso EBoufflet PFei ZGann EAshraf SLi ZMcneill C RDurrant J RHeeney M 2015 Chem. Mater. 27 7934
10Hou Q YQu L FZhao C W 2016 Acta Phys. Sin. 65 057401 (in Chinese)
11Wei Z PLu Y MShen D ZWu C XZhang Z ZZhao D XZhang J YFan X W2006J. Lumin.119551
12Wei Z PYao BWang X HZhang Z ZLu Y MShen D ZLi B HZhang J YZhao D XFan X WTang Z K 2007 J. Mater. Res. 22 2791
13Sanmyo MTomita YKobayashi K 2003 Chem. Mater. 15 819
14Li J BWei S HLi S SXia J B 2006 Phys. Rev. 74 81201
15Shi H LDuan Y 2008 Eur. Phys. J. 66 439
16Chen M MZhang Q LSu L XSu Y QCao J SZhu YWu T ZGui X CYang C LXiang RTang Z K 2012 Mater. Res. Bull. 47 2673
17Yamamoto T2002Thin Solid Films420–421100
18Weast R C1969Handbook of Chemistry and Physics49th edn.Boca ratonCRC Press163816–38
19Baron TSaminadayar KTatarenko SLugauer H JWaag ALandwehr G1998J. Crystal Growth184–185415
20Chen M MZhu YSu L XZhang Q LXiang RGui X CWu T ZCao X ZZhang PWang B YTang Z K 2013 J. Alloys Compd. 577 179
21Johansson L IJohansson H I PHåkansson K L 1993 Phys. Rev. 48 14520
22Porte LRoux LHanus J 1983 Phys. Rev. 28 3214
23Wei X QMan B YLiu MXue C SZhuang H ZYang C 2007 Physica B: Condens. Matter 388 145
24Jing L QXu Z LShang JSun X JCai W MGuo H C 2002 Mater. Sci. Eng. 332 356
25Fan J C CGoodenough J B 1977 J. Appl. Phys. 48 3524
26Koudelka MMonnier ASanchez JAugustynski J 1984 J. Mol. Catal. 25 295
27Li Y FYao BLu Y MWei Z PGai Y QZheng C JZhang Z ZLi B HShen D ZFan X WTang Z K 2007 Appl. Phys. Lett. 91 232113
28Klingshirn C F2012Semiconductor OpticsBerlin/HeidelbergSpringer