AIP Advances (Sep 2011)

Gallium interstitial contributions to diffusion in gallium arsenide

  • Joseph T. Schick,
  • Caroline G. Morgan

DOI
https://doi.org/10.1063/1.3644937
Journal volume & issue
Vol. 1, no. 3
pp. 032161 – 032161-26

Abstract

Read online

A new diffusion path is identified for gallium interstitials, which involves lower barriers than the barriers for previously identified diffusion paths [K. Levasseur-Smith and N. Mousseau, J. Appl. Phys. 103, 113502 (2008), P. A. Schultz and O. A. von Lilienfeld, Modelling and Simulation in Materials Science and Engineering 17, 084007 (2009)] for the charge states which dominate diffusion over most of the available range of Fermi energies. This path passes through the ⟨110⟩ gallium-gallium split interstitial configuration, and has a particularly low diffusion barrier of 0.35 eV for diffusion in the neutral charge state. As a part of this work, the character of the charge states for the gallium interstitials which are most important for diffusion is investigated, and it is shown that the last electron bound to the neutral interstitial occupies a shallow hydrogenic bound state composed of conduction band states for the hexagonal interstitial and both tetrahedral interstitials. How to properly account for the contributions of such interstitials is discussed for density-functional calculations with a k-point mesh not including the conduction band edge point. Diffusion barriers for gallium interstitials are calculated in all the charge states which can be important for a Fermi level anywhere in the gap, q = 0, +1, +2, and +3, for diffusion via the ⟨110⟩ gallium-gallium split interstitial configuration and via the hexagonal interstitial configuration. The lowest activation enthalpies over most of the available range of Fermi energies are found to correspond to diffusion in the neutral or singly positive state via the ⟨110⟩ gallium-gallium split interstitial configuration. It is shown that several different charge states and diffusion paths contribute significantly for Fermi levels within 0.2 eV above the valence band edge, which may help to explain some of the difficulties [H. Bracht and S. Brotzmann, Phys. Rev. B 71, 115216 (2005)] which have been encountered in fitting experimental results for heavily p-type, Ga-rich gallium arsenide by simply extending a model for gallium interstitial diffusion which has been used for less p-doped material.