Evaluation of Structural Properties and Defect Energetics in Al x Ga 1-x N Alloys
Farshid Reza, Beihan Chen, Miaomiao Jin · 2025
Read Evaluation of Structural Properties and Defect Energetics in Al x Ga 1-x N Alloys by Farshid Reza, Beihan Chen, Miaomiao Jin (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-25912.txt.
Introduction
Farshid Reza, Beihan Chen, and Miaomiao Jin study how mixing aluminium into gallium nitride changes its structure and defects. The alloy, written AlGaN, is key for light based and high power devices, but modeling a disordered mix with first principles methods is slow and limited. They instead use a machine learning interatomic potential, a fitted model that predicts forces between atoms quickly.
Reza and the team first check the model against known values for the pure end materials and for defect energies. They then map stiffness across mixes and find a nonlinear trend. Nitrogen pair defects and moving nitrogen defects depend strongly on local chemistry, while metal vacancies move more steadily across compositions, which guides device design.
Catalog record
| CREATOR | Farshid Reza, Beihan Chen, Miaomiao Jin |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-25912
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Evaluation of Structural Properties and Defect Energetics in Al x Ga 1-x N Alloys Farshid Reza, Beihan Chen, Miaomiao Jin 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. I Introduction Al x Ga 1-x N alloys are widely used in high-frequency and high-power electronic devices, including high-electron mobility transistors (HEMTs), light-emitting diodes (LEDs), and RF amplifiers, due to their wide bandgap and high electron mobility Dridi, Bouhafs, and Ruterana 2002 ; Mishra et al. 2008 ; Fletcher and Nirmal 2017 ; Piprek 2012 ; Çiçek and Badali 2024 . Their large bandgap and polarization effects also make them well suited for extreme environments, such as space-based electronics, where radiation resistance is critical Pearton et al. 2015 . However, various radiation exposures can still introduce defects, including vacancies, interstitials, and extended defect structures, that degrade electronic properties and device reliability Liu et al. 1998 ; Kucheyev et al. 2000 ; Islam et al. 2019 ; Sequeira et al. 2021 ; Kuboyama et al. 2011 . Understanding how such defects form and migrate at the atomic scale is critical for predicting material performance under operational stresses. Density functional theory (DFT) is widely used for calculating structural and defect properties with reasonable accuracy, such as formation energies and migration barriers Lei, Chen, and Ruterana 2023 ; Kyrtsos, Matsubara, and Bellotti 2016 . However, DFT is computationally expensive and limited to small systems, which restricts its ability to model disordered alloys. Classical molecular dynamics (MD) simulations, enabled by empirical interatomic potentials such as the Tersoff and Stillinger–Weber (SW) models, can scale to larger sizes Böckmann and Grubmüller 2002 ; Nord et al. 2003 , but often fail to deliver high-level accuracy, especially for defect formation and migration energetics Gao, Bylaska, and Weber 2004 ; Lei, Chen, and Ruterana 2023 ; Kyrtsos, Matsubara, and Bellotti 2016 . Moreover, these empirical force fields are typically fitted to specific material compositions and lack generalizability to alloy systems with compositional disorder. Although there has been significant work on GaN and AlN using both DFT and empirical MD, atomistic modeling of Al x Ga 1-x N alloys has received much less attention. Prior MD study on the alloy investigated thermal conductance at different AlGaN interfaces Huang et al. 2023 ; Luo et al. 2025 , interstitial and substitutional flourine ion movement in pure GaN and 25% AlGaN Yuan, Wang, and Chen 2008 , and Al effect on defect production during irradiation Jin et al. 2025 . Previous DFT studies on AlGaN include: vacancy diffusion in 30% AlGaN under the influence of strain and electric field Warnick et al. 2011 , and vacancy state near valence and conduction band in Al 6 Ga 24 N 30 structure Li et al. 2023 . Although empirical force fields such as Tersoff and Stillinger–Weber (SW) have been developed for AlGaN systems Zhou et al. 2013a ; Sun et al. 2021 ; Zhou et al. 2013b ; Karaaslan, Yapicioglu, and Sevik 2020 , they often lack the accuracy to capture the fine details of defect energetics. Consequently, the impact of alloy composition and local atomic fluctuations on defect formation and migration remains poorly understood. This knowledge gap is critical for predicting how AlGaN responds to radiation damage and thermal activation in real-world applications. Recent advances in machine learning interatomic potentials (MLIPs) offer a promising pathway to overcome the limitations of both DFT and empirical force fields by combining near-DFT accuracy with the computational efficiency of classical MD. Among them, neural network potentials (NNPs) have demonstrated strong predictive power across a wide range of material systems, such as carbon Wang et al. 2022 , Ag 2 S Balyakin and Sadovnikov 2022 , GaN Wu et al. 2024 , Ga 2 O 3 Li et al. 2020 , AlN Li et al. 2024 , and AlGaN alloys Huang et al. 2023 . Despite these advancements, the capability of these models to resolve local chemical effects and capture composition-dependent defect behavior across the full composition range of Al x Ga 1-x N alloys has yet to be utilized. In this work, we address this gap by using the MLIP to investigate the structural and defect-related physical properties of Al x Ga 1-x N alloys. The capability of the potential is first benchmarked against DFT and literature data for the binary endpoints GaN and AlN, reproducing the equation of state, elastic constants, and point defect formation and migration energies. We then apply the model to alloy systems to examine how those properties vary with composition and local environment. These results offer new insights into the defect physics of AlGaN alloys. II Computational Details All simulations in this study were conducted using the MLIP for AlGaN developed by Huang et al. Huang et al. 2023 . This NNP was trained on DFT data across a wide range of configurations and compositions. The potential was implemented in the LAMMPS molecular dynamics package Thompson et al. 2022 via integration with the DeepMD-kit framework Wang et al. 2018 , enabling the use of ML-based force evaluations within classical MD simulations. A 2880-atom supercell based on the wurtzite crystal structure was constructed for GaN, AlN, and Al x Ga 1-x N alloys with x=0.25 , 0.50 , and 0.75 . Periodic boundary conditions were applied in all directions. For static calculations, the supercells were modified based on the specific property being evaluated. For dynamics simulations, the temperature was controlled using a Nosé–Hoover thermostat at 300 K, unless otherwise noted, and the simulation timestep was set to 1 fs. To validate the ML potential, the equation of state (EOS) was computed for pure GaN and AlN. This was done by systematically varying the supercell volume through uniform expansion and contraction of lattice parameters. The energy–volume data were then fitted to the Birch-Murnaghan model to extract the equilibrium lattice parameters and minimum energy, which were compared with literature values to confirm consistency. To assess whether the Al and Ga atoms in the Al x Ga 1-x N alloys adopt any preferential ordering, Monte Carlo molecular dynamics (MCMD) simulations were performed for each alloy system. Simulations were carried out at 300