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

Basic Information

• Paper ID: 2510.25912
• Title: Evaluation of Structural Properties and Defect Energetics in Al$_x$Ga$_{1-x}$N Alloys
• Authors: Farshid Reza, Beihan Chen, Miaomiao Jin (Pennsylvania State University)
• Classification: cond-mat.mtrl-sci (Condensed Matter Physics - Materials Science)
• Submission Date: October 29, 2025 to arXiv
• Paper Link: https://arxiv.org/abs/2510.25912v1

Abstract

This study systematically investigates structural and defect-related physical properties of Al$_x$Ga$_{1-x}$N alloys, a critical material for high-performance optoelectronic and power devices, using machine learning interatomic potentials (MLIP). The research first validates the accuracy of MLIP by reproducing the equation of state, lattice constants, elastic constants, and defect formation and migration energies of binary endpoint materials (GaN and AlN). Subsequently, the potential is applied to evaluate elastic constants of AlGaN alloys, revealing nonlinear relationships in alloying effects. The study finds that nitrogen Frenkel pair formation energies and nitrogen point defect migration barriers are highly sensitive to local chemical environments and migration pathways, while Ga and Al vacancy migration energies are relatively insensitive to alloy composition, though interstitial migration energies exhibit stronger composition dependence. These results provide quantitative insights into how alloying affects defect energetics in AlGaN.

Research Background and Motivation

Research Questions

Al$_x$Ga$_{1-x}$N alloys are widely employed in high-frequency, high-power electronic devices (such as HEMTs, LEDs, and RF amplifiers), yet the influence of alloy composition on defect energetics remains insufficiently explored. Irradiation and thermal activation introduce defects such as vacancies and interstitials into materials, degrading electronic properties and device reliability. Therefore, understanding the formation and migration mechanisms of defects at the atomic scale is crucial for predicting material performance.

Problem Significance

1. Device Performance: The wide bandgap and polarization effects of AlGaN make it suitable for extreme environments (e.g., space electronics), but irradiation damage severely compromises device reliability.
2. Materials Design: Understanding defect physics is essential for developing defect engineering strategies for radiation-resistant, high-performance devices.
3. Knowledge Gap: Compared to pure GaN and AlN, atomic-scale simulation studies of AlGaN alloys are severely lacking.

Limitations of Existing Methods

1. DFT Methods: Computationally expensive, limited to small systems, difficult to simulate large configuration spaces of disordered alloys.
2. Empirical Potentials: Empirical potentials such as Tersoff and Stillinger-Weber are scalable to large systems but lack sufficient accuracy, particularly for defect formation and migration energetics.
3. Lack of Generality: Existing empirical potentials are typically fitted to specific compositions and lack universality for compositionally disordered alloy systems.

Research Motivation

Machine learning interatomic potentials (MLIP) provide a breakthrough pathway, combining near-DFT accuracy with the computational efficiency of classical MD. This study leverages a previously developed AlGaN neural network potential to systematically investigate local chemical effects and composition-dependent defect behavior across the full composition range.

Core Contributions

1. First Systematic Study: First systematic investigation of defect energetics across the full composition range of AlGaN alloys using high-fidelity MLIP.
2. Method Validation: Comprehensive validation of MLIP's accuracy in reproducing equation of state, elastic constants, and defect formation and migration energies of GaN and AlN.
3. Nonlinear Alloying Effects: Revelation of nonlinear variation of elastic constants with composition.
4. Quantitative Defect Behavior Description:
   • Discovery that N Frenkel pair formation energy is highly sensitive to local environment, exhibiting bimodal distribution
   • Elucidation of stabilization mechanisms for low-energy N defect configurations in low-Al-content alloys
   • Determination of strong composition dependence of interstitial defect migration energies
5. Engineering Guidance: Provision of atomic-scale insights for controlling defect tolerance through composition gradients.

Methodology Details

Task Definition

Input: Atomic configurations of Al$_x$Ga$_{1-x}$N alloys (x = 0, 0.25, 0.50, 0.75, 1.0)

Output:

• Structural properties: lattice constants, elastic constants
• Defect properties: Frenkel pair formation energies, vacancy and interstitial migration barriers

Constraints: Wurtzite crystal structure, neutral defect states

Computational Framework

1. MLIP Potential Function

Employs the AlGaN neural network potential (NNP) developed by Huang et al., trained on extensive DFT data covering multiple configurations and compositions. Integrated into the LAMMPS molecular dynamics package through the DeepMD-kit framework.

2. Simulation Setup

• Supercell: 2880-atom wurtzite structure supercell
• Boundary Conditions: Three-dimensional periodic boundary conditions
• Temperature Control: Nosé-Hoover thermostat at 300 K
• Time Step: 1 fs

3. Structural Property Calculations

Equation of State (EOS):

• Systematic variation of supercell volume (uniform lattice parameter expansion/contraction)
• Fitting energy-volume data to Birch-Murnaghan model
• Extraction of equilibrium lattice parameters and minimum energy

Ordering Assessment:

• Monte Carlo molecular dynamics (MCMD) simulations at 300 K, 100,000 MC steps
• Al-Ga atomic exchanges permitted, total potential energy monitored
• Radial distribution function (RDF) calculation to detect short-range ordering

Elastic Constants: For wurtzite structure, independent elastic constants are C₁₁, C₁₂, C₁₃, C₃₃, C₄₄, calculated by deforming the simulation box along appropriate directions and determining stress tensor changes. Bulk modulus obtained via:

$B = \frac{Y}{2(1+v)}$

where Young's modulus $Y = \frac{(C_{11}-C_{12})(C_{11}+2C_{12})}{C_{11}+C_{12}}$ and Poisson's ratio $v = \frac{C_{12}}{C_{11}+C_{12}}$

4. Defect Property Calculations

Defect Types:

• Vacancies: V$_{Al}$, V$_{Ga}$, V$_N$
• Interstitials: Al$_i$ and Ga$_i$ (octahedral configuration), N$_i$ (split configuration)
• Frenkel Pairs: vacancy-interstitial combinations of the same species
• Schottky Defects: simultaneous removal of one cation and one anion (binary compounds only)

Formation Energy Calculation: $E_f = E_{def} - \frac{N_d}{N}E_{perf}$

where $E_{def}$ is the total energy of the defective supercell, $E_{perf}$ is the total energy of the perfect supercell, $N_d$ is the number of atoms in the defective supercell, and $N$ is the number of atoms in the perfect supercell.

Migration Barrier Calculation: Climbing Image Nudged Elastic Band (CI-NEB) method employed:

• Initial and final configurations constructed based on literature (interstitial mechanism for interstitials, nearest-neighbor hopping for vacancies)
• Intermediate images linearly interpolated, NEB algorithm relaxed to track minimum energy path
• Migration energy $E_m$ defined as energy difference between highest saddle point and initial state on the path

Technical Innovations

1. Statistical Sampling Strategy: 100 random atomic configurations generated for each alloy composition to capture local chemical variation effects.
2. Full Composition Coverage: Systematic investigation across complete composition range (x = 0, 0.25, 0.50, 0.75, 1.0).
3. Large Supercell: 2880-atom supercell minimizes finite-size effects and better captures local structural relaxation.
4. Multiple Defect Types: Simultaneous investigation of vacancies, interstitials, and Frenkel pairs provides complete defect picture.
5. Directional Analysis: Distinction between in-plane and out-of-plane migration paths captures wurtzite structure anisotropy.

Experimental Setup

Computational Parameters

• Software: LAMMPS + DeepMD-kit
• Potential Function: AlGaN NNP developed by Huang et al. (2023)
• Supercell Size: 2880 atoms
• Temperature: 300 K (dynamics simulations)
• Time Step: 1 fs
• MCMD Steps: 100,000 steps
• Statistical Samples: 100 random configurations per alloy composition

Comparison Data Sources

1. Experimental Data:
   • Lattice constants: Roder et al., Figge et al., Chen et al.
   • Elastic constants: Kim et al. (experimental), Shimada et al. (DFT)
2. DFT Calculations: Lei et al., Kyrtsos et al., Limpijumnong et al., Zhu et al.
3. Empirical Potentials: Zhu et al. (AlN Stillinger-Weber potential)

Validation Benchmarks

• Binary endpoints (GaN, AlN) EOS, lattice constants, elastic constants
• GaN and AlN Frenkel pair and Schottky defect formation energies
• Vacancy and interstitial migration barriers (in-plane and out-of-plane directions)

Experimental Results

Lattice Constant Validation

Trends:

• a-axis and c-axis lattice constants decrease monotonically with Al content
• Consistent with linear Vegard's law relationship
• Trends consistent with experimental data (Roder, Figge, Chen, etc.)

Systematic Deviation:

• MLIP predictions systematically overestimate experimental values by approximately 1-2%
• Attributed to inherent characteristics of PBE functional used in training data (PBE typically predicts larger lattice constants)
• Despite deviation, MLIP accurately captures relative variation trends

Elastic Properties

Table I Summary (units: GPa):

Nonlinear Alloying Effects:

1. C₁₁: Significant decrease at low Al content (softening effect), begins to increase at high Al content.
2. C₁₂ and C₁₃: Plateau at intermediate compositions.
3. C₃₃: Pronounced decrease at intermediate alloy compositions, indicating disruption of bonding along c-axis.
4. C₄₄: Monotonic increase across full composition range, shear resistance increases with Al content.
5. Bulk Modulus B: Nonmonotonic trend, initially decreases with Al addition, partially recovers at high Al content.

Physical Interpretation: Substitution of smaller, lighter Al atoms for Ga induces local lattice distortion and weakens the bonding network. However, stronger Al-N bonds (compared to Ga-N) enhance certain elastic properties at high Al content.

Frenkel Pair Formation Energy

Binary Compound Validation

Stable Separation Distances:

• Ga and Al Frenkel pairs: >5 Å (defect pairs annihilate below this distance)
• N Frenkel pairs: ~3 Å sufficient for stability

Formation Energy Comparison (units: eV):

Key Findings:

• N Frenkel pair formation energy in GaN significantly lower than Ga Frenkel pair
• N and Al Frenkel pair formation energies comparable in AlN
• Highly consistent with DFT data (error <0.4 eV)

Composition Dependence in Alloy Systems

Figure 3 Key Results:

1. Ga and Al Frenkel Pairs:
   • Weak composition dependence of average formation energy
   • Narrow energy distribution, small standard deviation
   • Attributed to similar atomic size and bonding environment
2. N Frenkel Pairs:
   • Average formation energy increases significantly with Al content (7.43 eV@GaN → ~10 eV@75%Al)
   • Standard deviation increases substantially (reflecting local environment diversity)
   • Attributed to stronger Al-N bonds

Figure 4 Energy Distribution Characteristics:

25% Al Alloy:

• N Frenkel pair distribution exhibits "fat" low-energy tail
• Lowest energy configuration: 6.57 eV (Al-rich environment around interstitial)
• Highest energy configuration: 8.96 eV (Ga-rich environment around interstitial)
• Energy difference: 2.39 eV

75% Al Alloy:

• N Frenkel pair distribution exhibits bimodal characteristics
• Reflects emergence of two classes of local defect environments
• High-energy peak corresponds to N atom fully coordinated by Al

Physical Mechanism:

• Low Al content: Isolated Al atoms locally soften lattice, reducing N defect formation energy
• High Al content: Strong Al-N bonds dominate, overall formation energy increases
• Nonlinear dependence provides guidance for defect engineering

Vacancy Migration Energy

Binary Compound Validation

Table III Summary (units: eV):

Anisotropy: Out-of-plane migration barriers universally exceed in-plane barriers (wurtzite structure bonding anisotropy).

Composition Dependence in Alloy Systems

Figures 5 and 6 Key Findings:

1. Cation Vacancies (V$_{Ga}$, V$_{Al}$):
   • Average migration energy slightly increases with Al content
   • Narrow energy distribution, insensitive to local chemical fluctuations
   • In-plane migration distribution narrower than out-of-plane
   • Consistent with DFT results from Warnick et al. (Al₀.₃Ga₀.₇N: 1.8 eV@V$_{Ga}$, 1.6 eV@V$_{Al}$)
2. N Vacancy (V$_N$):
   • Average migration energy peaks at 50% Al
   • Extremely wide energy distribution (particularly at high Al content)
   • Reflects maximum local configuration disorder
   • Existence of low-energy migration paths suggests preferred diffusion channels
   • Warnick DFT value (2.2 eV@30%Al) falls within distribution range of this study

Physical Interpretation:

• 25% Al: Most paths resemble Ga-rich environment
• 50% Al: Highly mixed local environments, large variation in Ga-N and Al-N bonds
• 75% Al: Al-rich environments dominate, average barrier slightly lower but distribution remains wide

Interstitial Migration Energy

Binary Compound Validation

Table IV Summary (units: eV, interstitial mechanism):

Note: Although MLIP does not explicitly model charge states, migration paths and relative energy scales are consistent with literature.

Composition Dependence in Alloy Systems

Figures 7 and 8 Key Findings:

1. Ga Interstitial:
   • Average migration barrier shows moderate composition variation
   • Energy distribution significantly broadens in alloys
   • Low-barrier tail appears at 75% Al (percolation diffusion channel)
   • Suggests enhanced Ga long-range transport in Al-rich alloys
2. Al Interstitial:
   • Migration barrier increases with Al content
   • Distribution shifts toward higher energy and broadens
   • Attributed to stiffer Al-N bonding network
3. N Interstitial:
   • GaN→25%Al: Average barrier increases ~0.4 eV
   • 25%→75%Al: Average barrier relatively stable
   • Distribution peak narrows (possibly restricted migration paths)
   • Low-barrier tail appears at 75% Al (similar to Ga$_i$)

Important Finding: Emergence of low-energy diffusion channels in Al-rich alloys may lead to enhanced localized diffusion, with important implications for irradiation damage evolution.

Related Work

GaN and AlN Research

1. DFT Studies:
   • Lei et al.: GaN interatomic potential comparison
   • Kyrtsos et al.: Carbon and intrinsic point defect migration in GaN
   • Limpijumnong & Van de Walle: Intrinsic defect diffusion in GaN
   • Zhu et al.: Atomic methods for AlN defects and migration
2. Empirical Potentials: Tersoff, Stillinger-Weber potentials used for thermal conductivity and interface studies

AlGaN Alloy Research (Limited)

1. MD Studies:
   • Interface thermal conductivity in different AlGaN systems (Huang, Luo, et al.)
   • F ion motion in 25% AlGaN (Yuan et al.)
   • Effect of Al on defect production under irradiation (Jin et al.)
2. DFT Studies:
   • Warnick et al.: Vacancy diffusion in 30% AlGaN under strain and electric field
   • Li et al.: Vacancy states in Al₆Ga₂₄N₃₀

Machine Learning Potentials

• NNP demonstrates strong predictive capability in various material systems (carbon, Ag₂S, GaN, Ga₂O₃, AlN)
• AlGaN NNP developed by Huang et al. (2023) forms the basis of this study

Advantages of This Work

1. First Full-Composition Study: Systematic coverage of complete AlGaN composition range (x=0-1)
2. Statistical Sampling: 100 random configurations capture local environment effects
3. Comprehensive Defect Types: Simultaneous investigation of formation and migration energies
4. Quantitative Local Effects: Revelation of low-energy defect configurations and migration channels

Conclusions and Discussion

Main Conclusions

1. Structural Properties:
   • Lattice constants follow monotonic Vegard's law variation
   • Elastic constants exhibit significant nonlinear alloying effects
   • C₄₄ increases monotonically, C₃₃ decreases at intermediate compositions
2. Frenkel Pair Formation Energy:
   • Cation Frenkel pairs show weak composition dependence
   • N Frenkel pairs highly sensitive to local Al/Ga coordination
   • Low-energy N defect configurations appear at low Al content (local Al enrichment stabilization effect)
   • High Al content exhibits bimodal distribution (two classes of local environments)
3. Vacancy Migration:
   • Cation vacancy migration relatively insensitive to composition
   • N vacancy migration peaks at 50% Al, distribution extremely wide
   • Preferred diffusion channels exist
4. Interstitial Migration:
   • Ga interstitial exhibits low-barrier channels in Al-rich alloys
   • Al interstitial barrier increases with Al content
   • N interstitial exhibits low-energy tail at 75% Al
5. Engineering Implications:
   • Composition gradients can modulate defect tolerance
   • Low-energy configurations preferentially form under irradiation
   • Affects subsequent defect clustering and migration dynamics

Limitations

1. Absence of Charge States: MLIP does not explicitly model defect charge states; charge effects are important in actual devices.
2. Temperature Limitation: Primarily conducted at 300 K; high-temperature behavior requires further investigation.
3. Training Data Dependence: PBE functional systematic deviation (lattice constant overestimation).
4. Defect Types: Does not address complex defect clusters, dislocations, and other extended defects.
5. Dynamics: Long-timescale defect evolution and clustering not investigated.
6. Strain Effects: External strain effects on defect energetics not systematically considered.

Future Directions

1. Extended Defects: Investigation of defect clusters and dislocation core structures.
2. Dynamics Evolution: Long-timescale MD simulations of irradiation damage cascades and defect clustering.
3. Temperature Dependence: Systematic investigation of defect behavior across wide temperature range.
4. Electric Field Effects: Combination with charge models to study defects under device operating conditions.
5. Interface Effects: Defects at AlGaN/GaN heterostructure interfaces.
6. Composition Gradients: Defect behavior in functionally graded materials.
7. Experimental Validation: Comparison with positron annihilation, deep-level transient spectroscopy, and other experimental techniques.

In-Depth Evaluation

Strengths

1. Methodological Innovation:
   • First systematic investigation of AlGaN full-composition defect energetics using MLIP
   • Statistical sampling strategy (100 configurations) effectively captures local environment effects
   • Large supercell (2880 atoms) minimizes finite-size effects
2. Experimental Sufficiency:
   • Comprehensive validation of binary endpoints (EOS, elastic, defect formation/migration)
   • Multiple composition points (x=0, 0.25, 0.5, 0.75, 1.0) systematically cover range
   • Multiple defect types (vacancies, interstitials, Frenkel pairs)
   • Directional analysis (in-plane/out-of-plane)
3. Result Convincingness:
   • Highly consistent with DFT and experimental data (error <10%)
   • Clear physical trends and mechanisms revealed
   • Rich quantitative data (tables + distribution histograms)
   • Discovery of new phenomena such as low-energy defect configurations and migration channels
4. Writing Clarity:
   • Clear structure, rigorous logic
   • Detailed method description enabling reproducibility
   • High-quality figures with substantial information content
   • In-depth physical interpretation
5. Academic Value:
   • Fills gap in AlGaN defect research
   • Provides quantitative guidance for defect engineering
   • Demonstrates MLIP application potential in complex alloy systems

Shortcomings

1. Method Limitations:
   • Lacks charge state modeling (Fermi level effects)
   • Single temperature point (300 K)
   • PBE functional systematic deviation not corrected
2. Experimental Setup:
   • Does not consider external conditions such as strain and electric field
   • Defect concentration effects (defect-defect interactions) not discussed
   • Only neutral defects (actual devices have complex charge states)
3. Analysis Depth:
   • Physical mechanisms of certain nonmonotonic trends could be discussed more deeply
   • Lacks direct comparison with experimental defect concentration and mobility
   • Geometric structure characteristics of low-energy migration channels insufficiently analyzed
4. Practical Applicability:
   • Lacks specific parameter recommendations for actual device design
   • Irradiation dose dependence not addressed
   • Defect recombination and annihilation dynamics not investigated

Impact

1. Academic Contribution:
   • Provides first systematic MLIP study of AlGaN defect physics
   • Reveals nonlinear alloying effects on defect energetics
   • Provides benchmark data and methodological framework for subsequent research
2. Practical Value:
   • Guides radiation-resistant AlGaN-based device design
   • Provides atomic-scale insights for composition optimization
   • Supports defect engineering strategy development
3. Reproducibility:
   • Detailed method description
   • Uses public software (LAMMPS, DeepMD-kit)
   • Potential function already published (Huang et al. 2023)
   • Complete computational parameters
4. Limitations:
   • Requires high-performance computing resources
   • Depends on specific MLIP potential function
   • Experimental validation still requires further work

Applicable Scenarios

1. Materials Design:
   • AlGaN-based HEMT, LED device composition optimization
   • Radiation-resistant material screening
   • Composition gradient structure design
2. Defect Engineering:
   • Annealing process optimization (using migration barrier data)
   • Doping strategy (based on defect formation energy)
   • Irradiation damage prediction
3. Basic Research:
   • Alloy defect physics mechanisms
   • MLIP method validation and development
   • Multiscale simulation input parameters
4. Inapplicable Scenarios:
   • Research requiring precise charge states
   • Extreme temperature behavior (>1000 K)
   • Complex defect clusters, dislocation networks
   • Optical transition and electronic excitation state properties

Key References

1. Methodological Basis:
   • Huang et al. (2023): AlGaN NNP development Phys. Chem. Chem. Phys. 25, 2349
   • Wang et al. (2018): DeepMD-kit framework Comput. Phys. Commun. 228, 178
2. Validation Benchmarks:
   • Lei et al. (2023): GaN interatomic potential comparison AIP Advances 13
   • Kyrtsos et al. (2016): GaN defect migration Phys. Rev. B 93, 245201
   • Zhu et al. (2023): AlN defect computational study J. Mater. Chem. A 11, 15482
3. Experimental Comparison:
   • Kim et al. (1996): Elastic constants Phys. Rev. B 53, 16310
   • Roder et al. (2005): GaN thermal expansion Phys. Rev. B 72, 085218
4. Prior AlGaN Work:
   • Warnick et al. (2011): 30% AlGaN vacancy diffusion Phys. Rev. B 84, 214109
   • Jin et al. (2025): AlGaN irradiation response Acta Mater. 289, 120891

Overall Assessment: This is a high-quality computational materials science paper that, for the first time, systematically investigates AlGaN alloy defect energetics using machine learning potentials. The methodology is rigorous, validation comprehensive, results abundant, and important physical laws and nonlinear alloying effects are revealed. Despite limitations such as absence of charge state modeling, this work provides valuable quantitative data and atomic-scale insights for AlGaN defect physics and device design, making significant contributions to the field. The research demonstrates the powerful capability of MLIP in complex alloy systems and establishes a methodological benchmark for subsequent investigations.