Regensburg 2019 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
MM: Fachverband Metall- und Materialphysik
MM 12: Poster session I
MM 12.13: Poster
Monday, April 1, 2019, 19:15–20:45, Poster C
icet - A Pythonic approach to cluster expansions — •Mattias Ångqvist1, William A. Muñoz1, J. Magnus Rahm1, Erik Fransson1, Paul Erhart1, Céline Durniak2, Piotr Rozyczko2, and Thomas Holm Rod2 — 1Chalmers University of Technology, Department of Physics, Gothenburg, Sweden — 2Data Management and Software Centre, European Spallation Source, Copenhagen, Denmark
Many materials exhibit some form of chemical ordering, which can have a crucial impact on their macroscopic properties. Here, atomic scale modeling based on the so-called alloy cluster expansion (CE) technique can yield very valuable information. In this contribution, we present the open-source icet package that provides an efficient implementation of this methodology. It takes advantage of state-of-the-art machine learning techniques to generate accurate and predictive models based on quantum mechanical calculations. The icet package features a Python interface that enables seamless integration with other Python libraries including for example SciPy or scikit-learn. Yet, all computationally demanding parts are written in C++ providing performance while maintaining portability. We demonstrate the application of icet by (1) studying chemical ordering and associated properties in a series of intermetallic clathrates as a function of composition and temperature and (2) by predicting the phase diagrams of bulk and surface alloys.