https://github.com/wardlt/ternary-semiconductors-mhm
Scripts from a paper on discovering ternary semiconductors with machine learning and crystal structure prediction
https://github.com/wardlt/ternary-semiconductors-mhm
Last synced: 3 months ago
JSON representation
Scripts from a paper on discovering ternary semiconductors with machine learning and crystal structure prediction
- Host: GitHub
- URL: https://github.com/wardlt/ternary-semiconductors-mhm
- Owner: WardLT
- Created: 2018-10-26T20:29:26.000Z (over 7 years ago)
- Default Branch: master
- Last Pushed: 2018-10-26T20:47:00.000Z (over 7 years ago)
- Last Synced: 2025-03-05T22:44:49.187Z (over 1 year ago)
- Language: Jupyter Notebook
- Size: 35.2 KB
- Stars: 3
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Ternary Semiconductors from Machine Learning and Crystal Structure Prediction
This repository contains scripts and datasets needed to reproduce the machine learning results from a recent paper by Amsler et al.
## Contents
The key machine learning task performed in this manuscript is to identify the compositions in the Ba-As-S ternary system that are favorable for semiconductor applications.
The `make-deltae-model.in` and `make-bandgap-model.in` models employ [Magpie](https://bitbucket.org/wolverton/magpie) to train machine learning models on data from the [OQMD](http://oqmd.org/) to predict stability (via the formation energy) and the band gap energy fo the model. The training set for the model is available in [`./datasets/`](./datasets/) and the models are in [`./models/`](./models) directory.
The `scan-BaAs-system.in` script runs the models on the Ba-As-S system and computes the stability of each prediction with respect to the convex hull of the training set.
The Jupyter notebook `plot-BaAsS-results.ipynb` produces the plots seen in the paper.
## Installation
You must clone the repository using `git clone --recursive` to get the necessary Magpie source code. Then, follow the instructions in the Magpie documentation to compile Magpie.
Besides Magpie, you need to install a Python 3 version of Jupyter with the packages listed in `requirements.txt`
## Running the Scripts
Call `./run-all.bs` to execute all of the scripts in this study.