Introduction
What is AIPHAD?
In AIPHAD (Artificial Intelligence techniques for PHAse diagram) project, AI technique to investigate detailed phase diagram has been developed. This method suggests appropriate “experimental conditions” to minimize a number of experiments for completing phase diagram as much as possible.
In the AIPHAD library, PDC (Phase Diagram Constructionn) algorithm is used. Specifically, first, the information on the experimental conditions in the phase diagram where the phases are identified is given. Label propagation/label spreading methods are used as a phase estimation algorithm to predict the phase diagram where the phase has not yet been identified. Based on these results, the most uncertain points in the phase diagram are selected as a candidate for the next experimental condition. As the experiments are conducted at the proposed points and the phases are identified, the training data increases and PDC proposes new experimental conditions. This iterative process allows phase boundaries to be known and new phases to be found even when the number of experiments is limited.
Followings are the features of this library.
Phase diagram can be investigated for experimental conditions with arbitrary dimensions.
Phase diagram estimation algorithm and uncertainty score can be easily changed.
Multiple experimental conditions can be proposed.
Predicted phase diagram can be output as an image file.
Experimental conditions with high appearance probability of a specified phase can be proposed.
A web application for investigating phase diagram, AIPHAD https://aiphad.org, is also available. By using this web application, researchers who are not familiar with programming can easily use PDC.
Citation
When using PDC in AIPHAD, please cite the following reference:
Kei Terayama, Ryo Tamura, Yoshitaro Nose, Hidenori Hiramatsu, Hideo Hosono, Yasushi Okuno, and Koji Tsuda, “Efficient construction method for phase diagrams using uncertainty sampling”, Physical Review Materials 3, 033802 (2019). DOI: 10.1103/PhysRevMaterials.3.033802
Available from https://journals.aps.org/prmaterials/abstract/10.1103/PhysRevMaterials.3.033802 (open access).
Bibtex is as follows
@article{PhysRevMaterials.3.033802,
title = {Efficient construction method for phase diagrams using uncertainty sampling},
author = {Terayama, Kei and Tamura, Ryo and Nose, Yoshitaro and Hiramatsu, Hidenori and Hosono, Hideo and Okuno, Yasushi and Tsuda, Koji},
journal = {Phys. Rev. Mater.},
volume = {3},
issue = {3},
pages = {033802},
numpages = {8},
year = {2019},
month = {Mar},
publisher = {American Physical Society},
doi = {10.1103/PhysRevMaterials.3.033802},
url = {https://link.aps.org/doi/10.1103/PhysRevMaterials.3.033802}
}
When using the multiple proposal technique for PDC in AIPHAD, please cite the following reference:
Ryo Tamura, Guillaume Deffrennes, Kwangsik Han, Taichi Abe, Haruhiko Morito, Yasuyuki Nakamura, Masanobu Naito, Ryoji Katsube, Yoshitaro Nose, and Kei Terayama, “Machine-learning-based phase diagram construction for high-throughput batch experiments”, Science and Technology of Advanced Materials: Methods 2, 153-161 (2022). DOI: 10.1080/27660400.2022.2076548
Available from https://www.tandfonline.com/doi/full/10.1080/27660400.2022.2076548 (open access).
Bibtex is as follows
@article{doi:10.1080/27660400.2022.2076548,
author = {Ryo Tamura and Guillaume Deffrennes and Kwangsik Han and Taichi Abe and Haruhiko Morito and Yasuyuki Nakamura and Masanobu Naito and Ryoji Katsube and Yoshitaro Nose and Kei Terayama},
title = {Machine-Learning-Based phase diagram construction for high-throughput batch experiments},
journal = {Science and Technology of Advanced Materials: Methods},
volume = {2},
number = {1},
pages = {153-161},
year = {2022},
publisher = {Taylor & Francis},
doi = {10.1080/27660400.2022.2076548}
}
Main developers
ver. 1.0-
Ryo Tamura (Center for Basic Research on Materials, National Institute for Materials Science)
Kei Terayama (Graduate School of Medical Life Science, Yokohama City University)
Naotoshi Tominaga (AdvanceSoft Corporation)
Licence
The program package and the complete source code of this software are distributed under the MIT License.