Evaluation of Structural Properties and Defect Energetics in Al$_x$Ga$_{1-x}$N Alloys

TL;DR

This work addresses how alloy composition affects defect energetics in AlGaN and develops a high-fidelity MLIP trained on DFT to model disordered alloys. The approach validates the potential against GaN/AlN properties and then applies it to alloy compositions at $x=0.25$, $x=0.5$, and $x=0.75$ to map Frenkel pair formation energies and migration barriers across local environments. The results show that nitrogen defects exhibit strong composition- and environment-dependent energetics, with low-energy Frenkel configurations and emergent diffusion channels in Al-rich alloys, while cation defect migration is comparatively robust. These findings enable defect-engineering strategies for AlGaN-based optoelectronic and power devices by linking composition, local chemistry, and defect dynamics.

Abstract

Al$_x$Ga$_{1-x}$N alloys are essential for high-performance optoelectronic and power devices, yet the role of composition on defect energetics remains underexplored, largely due to the limitations of first-principles methods in modeling disordered alloys. To address this, we employ a machine learning interatomic potential (MLIP) to investigate the structural and defect-related physical properties in Al$_x$Ga$_{1-x}$N. The MLIP is first validated by reproducing the equation of state, lattice constants, and elastic constants of the binary endpoints, GaN and AlN, as well as known defect formation and migration energies from density functional theory and empirical potentials. We then apply the MLIP to evaluate elastic constants of AlGaN alloys, which reveals a non-linear relation with alloying effect. Our results reveal that nitrogen Frenkel pair formation energies and the migration barriers for nitrogen point defects are highly sensitive to the local chemical environment and migration path. In contrast, Ga and Al vacancy migration energies remain relatively insensitive to alloy composition, whereas their interstitial migration energies exhibit stronger compositional dependence. These results provide quantitative insight into how alloying influences defect energetics in AlGaN, informing defect engineering strategies for improved material performance.

Figures (8)

• Figure 1: ML calculated lattice constant at 300 K for different Al$_x$Ga$_{1-x}$N alloys and comparison to previous experimental data by Roder et al. roder2005temperature, Figge et al. figge2009temperature, and Chen et al. chen2006high along the $a$-direction (a) and $c$-direction (b).
• Figure 2: Frenkel pair interaction energy as a function of Frenkel pair distance in GaN (a) and AlN (b).
• Figure 3: Frenkel pair formation energy as a function of Al percentage.
• Figure 4: Histogram of the Frenkel pair formation energies in 25% AlGaN (a), 50% AlGaN (b), and 75% AlGaN (c). As a reference, dashed lines indicate the corresponding Frenkel pair energy in the pure nitrides.
• Figure 5: In-plane average vacancy migration energy with literature data of 30% AlGaN warnick2011room (a) and out-of-plane (b) average vacancy migration energy as function of Al%.