Predict relative permittivity and centroid shift for inorganic materials
This package provides two dependent models to predict the relative permittivity and centroid shift of inorganic materials via the command-line.
To cite relative permittivity and centroid shift predictions, please reference the following work:
Zhuo. Y, Hariyani. S, You. S, Dorenbos. P, and Brgoch. J, Machine learning 5d-level centroid shift of Ce3+ inorganic phosphors, J. Appl. Phys. 2020, 128, 013104.
This package requires:
Note: The centroid shift prediction needs the relative permittivity value as one of the inputs. If you have it ready, you can jump to Section 2. Or, you can get a predicted relative permittivity value following Section 1.
You should create a .xlsx
file named c_pounds.xlsx
, in which the compositions that are of interest are listed in the first column with the header "Composition
".
There is one example of customized dataset in the repository:examples/c_pounds.xlsx
.
You can get compositional descriptors by:
python descriptor_generator.py
descriptor_generator.py
will automatically read elements.xlsx
and c_pounds.xlsx
to generate descriptors. After running, you will get a .xlsx
file named to_predict_relative_permittivity.xlsx
. In this file, the first column is your composition followed by 85 columns of descriptors.
You also need to append another 13 structural descriptors to the compositional descriptors:
- space group number
- unit cell volume (nm3)
- density (Mg/m3)
- a/b
- b/c
- c/a
- alpha/beta
- beta/gamma
- gamma/alpha
- existance of inversion center (exist:1; nonexist:0)
- existance of polar axis (exist:1; nonexist:0)
- volume per Z (nm3)
- volume per atom (nm3)
This information could be extracted from crystallographic information files (CIFs) and inorganic cystal databases. There is one example of the final to_predict_relative_permittivity.xlsx
file in the repository:examples/to_predict_relative_permittivity.xlsx
.
Before getting a prediction, you will need to:
- Prepare a customized dataset named after
to_predict_relative_permittivity.xlsx
to store the composition-structure-property relations of interest.
Then, you can predict the relative permittivity by:
python relative_permittivity_predictor.py
relative_permittivity_predictor.py
will automatically read relative_permittivity_training_set.xlsx
and to_predict_relative_permittivity.xlsx
to generate a prediction. You will then get a predicted_relative_permittivity.xlsx
file in the same directory, in which the predicted relative_permittivity is provided next to the corresponding composition.
You should create a .xlsx
file named to_predict_centroid_shift.xlsx
in the format as:
A | B | C | D | E | F | G | H | I |
---|---|---|---|---|---|---|---|---|
Composition | Relative permittivity | Avg. cation electronegativity | Avg. anion polarizability | Rm | DeltaR (Rm-RCe | Avg. bond length | Coord. no. | Condensation |
There is one example of customized dataset in the repository:examples/to_predict_centroid_shift.xlsx
.
Before getting a prediction, you will need to:
- Prepare a customized dataset named after
to_predict_centroid_shift.xlsx
to store the composition-structure-property relations of interest.
Then, you can predict the relative permittivity by:
python centroid_shift_predictor.py
centroid_shift_predictor.py
will automatically read centroid_shift_training_set.xlsx
and to_predict_centroid_shift.xlsx
to generate a prediction. You will then get a predicted_centroid_shift.xlsx
file in the same directory, in which the predicted centroid shift is provided next to the corresponding composition.
This software was created by Ya Zhuo who is advised by Prof. Jakoah Brgoch.