Simulation and modeling of the electronic structure of GaAs damage clusters

Abstract

In an effort to build a stronger microscopic foundation for radiation damage models in gallium arsenide (GaAs), the electronic properties of radiation-induced damage clusters are studied with atomistic simulations. Molecular dynamics simulations are used to access the time and length scales required for direct simulation of a collision cascade, and density functional theory simulations are used to calculate the electronic properties of isolated damaged clusters that are extracted from these cascades. To study the physical properties of clusters, we analyze the statistics of a randomly-generated ensemble of damage clusters because no single cluster adequately represents this class of defects. The electronic properties of damage clusters are accurately described by a classical model of the electrical charging of a semiconducting sphere embedded in an uniform dielectric. The effective band gap of the cluster depends on the degree of internal structural damage, and the gap closes to form a metal in the high-damage limit. We estimate the Fermi level of this metallic state, which corresponds to high-energy amorphous GaAs, to be 0.46 +/- 0.07 eV above the valence band edge of crystalline GaAs.

Figures (5)

• Figure 1: Electronic density of states of crystalline GaAs (gray shaded region) compared to amorphous GaAs structures. The structure in (a) is formed by annealing at 2500 K. The structure in (b) is the first structure with As-As and Ga-Ga bonded minimized by pairwise atomic rearrangements and further annealing at 800 K. The Fermi level of the amorphous structures is aligned to the center of the crystalline GaAs band gap.
• Figure 2: Artificial damage clusters fit to a capacitive model of their defect levels, $\Delta_{x+0.5}^{x-0.5}$. (a) The correspondence between $x$ and the fit value, $\tilde{x}$, defined in Eq. (\ref{['xfit']}). (b) The distribution of $\mu_0$ values from the optimized models and a Gaussian fit.
• Figure 3: Artificial damage clusters fit to a metallic sphere model of their defect levels, $\Delta_{x+0.5}^{x-0.5}$. The correspondence (a) between $x$ and the fit value, $\tilde{x}$, defined in Eq. (\ref{['xfit']}), and (b) between geometric and electrostatic radii. (c) The variation of $E_F$ with cluster size plotted against the globally optimized value (dashed line).
• Figure 4: Radial distribution of excess charge in two damage clusters with one electron added to each. We compare the charge density of the defect state (gray shaded region) to the total electronic charge rearrangement between the charged and neutral clusters (line) and the geometric radius (vertical solid line) to the electrostatic radius (vertical dotted line). The electrostatic radius is defined here as the minimum radius containing the total excess charge of magnitude $\epsilon^{-1}$ that remains after screening from bulk GaAs.
• Figure 5: 10 GaAs damage clusters extracted from MD simulations, with (h-j) further annealed after extraction ((h) is the annealed version of (g)). Structures are visualized in QuteMolqutemol, with Ga and As as black and white spheres. Only atoms that deviate from their ideal crystalline positions by more than $5 \%$ of the lattice constant are shown. Total number of defective atoms in the cluster are given, with $+/-$ denoting the deviation from the crystal stoichiometry. The fit to the semiconducting sphere model in Eq. (\ref{['model3']}) as well as the RMS error in $\Delta_{Q+1}^{Q}$ is given for each cluster. Transitions between stable charge states, $\Delta_{Q'}^{Q}$, are shown along with the $Q$ value of the energy intervals for the model (M), vertical (V), and adiabatic (A) calculations of cluster energies. $Q' = Q+1$ for all model and vertical charge states, and only the $Q=0$ charge state is labeled.