TaW-tce-kmc

This repository contains the code and data necessary to reproduce our paper on arXiv.

Download the code & set up the Python environment

Download using git:

git clone https://github.com/MUEXLY/TaW-tce-kmc.git
cd TaW-tce-kmc/
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Benchmark the KRA relation on our dataset:

python kra.py

This will create the figure figures/kra.pdf.

Train a tce model on our dataset:

python training.py

This will create two files:

• a parity plot at figures/barriers.pdf
• a serialized tce model at TaWX.pkl

Visualize the cross-validation within the training:

python cross-val.py

This will create the figure figures/cross-val.pdf.

Visualize the fitted coefficients:

python coeffs-vis.py

This will create two files:

• a figure with two-body coefficients at figures/two-bodies.pdf
• a figure with three-body coefficients at figures/three-bodies.pdf

Predict the pure Ta and pure W barriers:

python pure-barriers.py

Run the KMC simulations in parallel:

python kmc.py

Important: This script is very expensive! On 64 cores, this parallelized KMC simulation takes up about ~10 GB of memory and runs for ~6 hours. If you have access to a supercomputing resource, consider submitting this as a batch job instead, of which we've provided a template for systems using SLURM:

sbatch kmc.slurm

This will create a lot of files:

• xyz coordinate files with pattern results/Ta*W*_*K/*.xyz, which can be visualized in OVITO
• Log files with pattern results/Ta*W*_*K/log.jsonl

Compute and visualize the distributions of visited barriers

python hop-histograms.py

This step will also take quite a bit, but only runs on a single processor.

This will create a grid of histograms of visited barriers figures/barrier-distributions.pdf.

Compute and store the mean-squared-displacement (MSD) curves:

python msd.py

This will also create a lot of files:

• MSD curves with pattern results/Ta*W*_*K/msd.png
• MSD fits with pattern results/Ta*W*_*K/fit.json

Compute the resulting diffusitivites and activation energies:

python diffusivity.py

This will create two files:

• A figure of diffusivities over compositions figures/diffusivities.pdf
• A figure of activation energies over compositions figures/activation-energies.pdf

Compute the excess Ta near the vacancy throughout the simulations:

python local-comp.py

This will create a figure visualizing the excess Ta figures/excess-ta.pdf.