Skip to content

USCCACS/ALLM

Repository files navigation

Active Learning for Accelerated Design of Layered Materials

Bassman, L., Rajak, P., et al., npj Computational Materials 4, 74 (2018).

Hetero-structures made from vertically stacked monolayers of transition metal dichalcogenides hold great potential for thermoelectric devices of the future. Discovery of the optimal layered material for specific applications necessitates the estimation of key material properties, however, screening of properties via brute force ab initio calculations of the entire material structure space exceeds the limits of current computing resources. Moreover, the functional dependence of material properties on the structures is often complicated, making simplistic statistical procedures for prediction difficult to employ without large amounts of data collection. This repository includes Gaussian process regression code (predict_maxval.py, predict_structure.py) for prediction of band gap, conduction band minimum dispersion curve, valance band maximum dispersion curve, and Thermoelectic EFF vs. dopant concentration curve. It also includes Bayesian optimization code (Bayesian_opt.py) to predict the best thermoelectric material using the fewest structure calculations. Data sets for all codes are also included. Desciptions of these codes and datasets, as well as how to run the code, are given below. Python code (snl_prep.py) is also included for automatically generating a structure file for a given layered material and uploading it to the Materials Project Database, along with bash scripts (create-[X]layers.sh) for automatically generating and uploading all unique 2-, 3-, and 4-layered materials using 'snl_prep.py'.


1 predict_maxval.py:

A Gaussian Process Regression model to predict band gap.

predict_maxval.py: creates a gaussian process regression model using x% of data from 3-layer-band_gap.txt. After buiding the model predcits the band gap of the remaining (1-x%) data as test set.

Dataset
3-layer-band_gap.txt CBM and VBM 3-layer heretero-structure
xaxisvalue.txt Contains the value of the wave vector

Input Paramaters:
inputfile_name="3-layer-band_gap.txt" * ###file name of the input data *
train_test_split=0.60 * ###split between training and test set*
Nrun = 1

To run the program: predict_maxval.py

2 predict_structure.py:

A Gaussian Process Regression model to predict conduction band maxima, valance band minima and Thermoelectic EFF function.

predict_structure.py build a gaussian process (GP) regression model for condunction band minima and valance band maximum. It takes 3-layer-band_structure.txt as inout data and split in into training and test set. GP model is build using training set and the images of prediced condunction band minima and valance band maximum of the test set is saved in folder Bandstructure.

Dataset
"3-layer-band_structure.txt Band gap for 3-layer heretero-structure

Input Paramaters:
inputfile_name="3-layer-band_structure.txt" * ###file name of the input data *
train_test_split=0.60 * ###split between training and test set*
Nrun = 1

To run the program: predict_structure.py

3 Bayesian_opt.py:

An active learning model, based on Bayesian Optimization, to discover material with optimal property witn minimum structure evalulation

Dataset
N_doped_EFF_max.txt EFF (Electronic fitness function) maximum value for n-doped 3-layer heretero-structure
P_doped_EFF_max.txt EFF (Electronic fitness function) maximum value for p-doped 3-layer heretero-structure
3-layer-band_gap.txt Maximum Band gap for 3-layer heretero-structure

Input Paramaters:
Inside the code Bayesian_opt.py, we have
Nruns = 1 * ### total number of Bayesian Optimization runs*
train_test_split=0.10 * ### initial sampled data in a given Bayesian Optimization run*

To find n-doped 3-ayer hetero-structure with optimal EFF value
Run python3.6 Bayesian_opt.py N_doped_EFF_max
To find p-doped 3-ayer hetero-structure with optimal EFF value
Run python3.6 Bayesian_opt.py P_doped_EFF_max
To find maximum band gap
Run python3.6 Bayesian_opt.py 3-layer-band_gap.txt

4 snl_prep.py:

Given a layered crystal structure as a string, the python script generates a structure file and uploads it to the Materials Project Database

For example, for the 4-layered heterostructure MoTe2-MoS2-WS2-WTe2, type:

python snl_prep.py -s MoTe2-MoS2-WS2-WTe2 -d .

The -s parameter is the structure of the layers encoded in a string. It must be formatted by the above way, but it could be any number of layers instead of just 4. The -d parameter specifies the base directory in which the newly created structure file should be placed, either absolute or relative to the pwd.

5 create-[X]layers.sh:

Bash code for automatically generating all unique 2-, 3-, and 4-layered TMDC heterostructures and uploading them to MP Database by repeated use of the above python code, snl_prep.py. snl_prep.py is assumed to be in the same directory as the bash script

To run, type one of the following:

bash create-2layers.sh
bash create-3layers.sh
bash create-4layers.sh