Nanoscale Thermal Transport in Al_xGa_1−xN using Machine Learned Interatomic Potential

This repository contains computational tools and workflows for studying phonon thermal transport in Al_xGa_1−xN using the Atomic Cluster Expansion (ACE) based machine-learned interatomic potential.

Overview

This project implements a comprehensive computational framework for:

• Force constant calculations using ACE potential at zero temperature
• Thermal conductivity calculations via Green-Kubo molecular dynamics (MD)
• TDEP based Force constant mapping using Virtual Crystal Approximation (VCA) for alloy systems
• Format conversion between different force constant representations (ShengBTE, TDEP)

Repository Structure

ace_mlip_algan/
├── data/                                   
│   ├── algan_ace_potential.yaml            # ACE potential file
│   └── all_algan_aimd_dft.gzip             # AIMD DFT training data
│
├── zero_kelvin_lattice_dynamics/           # Zero-temperature lattice dynamics
│   ├── calculate_force_constants.py        # Calculate FC2 and FC3 using ACE with Phono3py
│   └── vca_force_constants.ipynb           # Virtual Crystal Approximation for force constants
│
├── force_constant_mapping/                  # TDEP Force constant mapping and conversion
│   ├── map_force_constants_2nd_order.py     # Map 2nd-order force constants (FC2)
│   ├── map_force_constants_3rd_order.py     # Map 3rd-order force constants (FC3)
│   ├── utils_fc2.py                         
│   ├── utils_fc3.py                         
│   ├── vca_force_constants_2nd_order.ipynb   # VCA interpolation for FC2
│   ├── vca_force_constants_3rd_order.ipynb   # VCA interpolation for FC3
│   ├── vca_unit_cell.ipynb                 
│   ├── convert_shengbte_to_tdep_fc2.ipynb  
│   ├── convert_tdep_to_shengbte_fc3.ipynb  
│   └── count_nonzero_force_constants.ipynb  
│
└── green_kubo_md/                          # Green-Kubo MD calculations
    ├── calculate_thermal_conductivity.py   # Post-process heat flux data and compute thermal conductivity
    ├── lammps_green_kubo_script.in         # LAMMPS input script for GK-MD
    └── algan_ace_potential.yace            # ACE potential file 

Key Components

1. Zero-Temperature Lattice Dynamics (zero_kelvin_lattice_dynamics/)

Computes harmonic (FC2) and anharmonic (FC3) force constants using the ACE potential:

• Uses Phono3py for generating displacement supercells
• Calculates forces using PyACECalculator (ASE interface)
• Outputs force constants in HDF5 format

Main script: calculate_force_constants.py

2. Force Constant Mapping (force_constant_mapping/)

Tools for mapping and averaging force constants in alloy systems:

• VCA (Virtual Crystal Approximation): Maps and Combines force constants from AlN and GaN to create VCA-based Al_xGa_1−xN force constants
• Format conversion: Converts between ShengBTE and TDEP force constant formats
• Analysis: Counts and analyzes non-zero force constants

Key scripts:

• map_force_constants_2nd_order.py: Maps FC2 for different Al/Ga compositions
• map_force_constants_3rd_order.py: Maps FC3 for different Al/Ga compositions

3. Green-Kubo MD (green_kubo_md/)

Calculates thermal conductivity using equilibrium MD:

• LAMMPS simulation: Runs EMD with modified many-body heat flux calculation (as proposed by )
• Post-processing: Computes heat current autocorrelation function (HCACF) and thermal conductivity

Key files:

• lammps_green_kubo_script.in: LAMMPS input for Green-Kubo MD
• calculate_thermal_conductivity.py: Post-processes heat flux data and computes thermal conductivity

Dependencies

External Software

• LAMMPS: For molecular dynamics simulations (with PACE potential support)
• TDEP: For temperature-dependent effective potential and phonon calculations at finite temperature

Packages

• ase
• phonopy
• phono3py
• pyace

Usage

Mapping Force Constants for Alloys

cd force_constant_mapping
python map_force_constants_2nd_order.py  # For TDEP FC2
python map_force_constants_3rd_order.py  # For TDEP FC3

Modify the fraction_al variable in the scripts to set the Al composition.

Data Files

• ACE Potential: The trained ACE potential is provided in both YAML (data/algan_ace_potential.yaml) and YACE (green_kubo_md/algan_ace_potential.yace) formats
• Training Data: AIMD DFT data used for training is stored in data/all_algan_aimd_dft.gzip

Citation

Contact

For questions regarding the data or methodology, please contact the corresponding author or open an issue in this repository and tag @rfhari.