Overcoming Polarons to Achieve p-Type Conductivity in Ultrawide-Bandgap Oxides

A major shortcoming of ultrawide-bandgap semiconductors is their lack of bipolar doping. For some ultrawide-bandgap oxides, n-type conductivity has been demonstrated, but p-type conductivity is in general inhibited by a strong tendency to form self-trapped holes (small polarons). This problem especially afflicts Ga₂O₃, which is among the most promising UWBG oxides, but in which polaronic hole trapping causes acceptors to have ionization energies exceeding 1 eV. Related materials, such as LiGa₅O₈, also suffer from hole trapping. Recently, rutile germanium oxide (r-GeO₂), with a band gap near 4.7 eV, was found to break from this paradigm. Though calculations found holes trapped much less strongly, the predicted acceptor ionization energies are still relatively high (~0.4 eV), limiting p-type conductivity [1,2]. Since r-GeO₂ appears to be an outlier, perhaps due to its crystal structure, the properties of a set of rutile oxides are calculated and compared. Our hybrid density functional calculations here show that rutile TiO₂ and SnO₂ strongly trap holes at acceptor impurities, in agreement with prior studies. However, self-trapped holes are found to be unstable in r-SiO₂, a metastable polymorph of silica with an 8.5 eV band gap. Group-III acceptor ionization energies are also found to be lower in r-SiO₂ than in the other rutile oxides. Furthermore, acceptor dopants have sufficiently low formation energies such that compensation by donors (such as oxygen vacancies) could be avoided, at least under O-rich limit conditions. Based on the results [3], it appears that r-SiO₂ has the potential to exhibit the most efficient p-type conductivity when compared to other UWBG oxides.

[1] S. Chae, K. Mengle, K. Bushick, J. Lee, N. Sanders, Z. Deng, Z. Mi, P. F. P. Poudeu, H. Paik, J. T. Heron, and E. Kioupakis, Appl. Phys. Lett. 114, 102104 (2019).
[2] J. L. Lyons, J. Appl. Phys. 131, 025701 (2022).
[3] J. L. Lyons and A. Janotti, J. Phys.: Condens. Matt. 36, 085501 (2023).

This work was supported by the ONR/NRL 6.1 Basic Research Program.