This title appears in the Scientific Report :
2023
Please use the identifier:
http://dx.doi.org/10.34734/FZJ-2023-02298 in citations.
Application of Machine-Learning Models and XAI in Materials Science Using Magnetic DFT Data
Application of Machine-Learning Models and XAI in Materials Science Using Magnetic DFT Data
The prediction of material properties using ab-initio simulations [1] is a research field of fundamental significance due to the multitude of materials applications in technology and research. Such simulations can require huge computational resources if carried out for many materials, and thus the p...
Saved in:
Personal Name(s): | Hilgers, Robin (Corresponding author) |
---|---|
Wortmann, Daniel / Blügel, Stefan | |
Contributing Institute: |
Quanten-Theorie der Materialien; PGI-1 Quanten-Theorie der Materialien; IAS-1 |
Imprint: |
2023
|
DOI: |
10.34734/FZJ-2023-02298 |
Conference: | Helmholtz.Ai, Hamburg (Germany), 2023-06-13 - 2023-06-14 |
Document Type: |
Poster |
Research Program: |
Helmholtz School for Data Science in Life, Earth and Energy (HDS LEE) Topological Matter |
Link: |
OpenAccess |
Publikationsportal JuSER |
The prediction of material properties using ab-initio simulations [1] is a research field of fundamental significance due to the multitude of materials applications in technology and research. Such simulations can require huge computational resources if carried out for many materials, and thus the possibility to employ machine learning methods offers potential benefit by restricting the materials search space.In our work, we demonstrate this idea for the particular field of ordered and disordered magnetic Heusler alloys. In detail, we focus on theirCurie-temperature, which often is the major quantity describing the fitness of a magnetic material for application. Thus, we studied the possibility to predict the Curie-temperature and thereby classifying the materials relevance for possible technical application using machine-learning algorithms.We compared the performance of regression models and classification models in order to predict the range of the Curie-temperature of given compounds to demonstrate the possibility to reduce the computational expanses of simulating this quantity. This work can be seen as a small-scale case study in which lightweight machine learning algorithms can add value to existing simulation data and eventually replace costly computational steps in layered calculation workflows in the future. In addition, we demonstrate, that such models can lead to interesting unbiased physical insight. Acknowledgement: This work was performed as part of the Helmholtz School for Data Science in Life, Earth and Energy (HDS-LEE) and received funding from the Helmholtz Association of German Research Centres.[1] R. Kovacik, P. Mavropoulos, S. Blüugel, Materials Cloud (2022) 10.24435/MATERIALSCLOUD:WW-PV |