On November 13, 2024, Dr. Stephan Petersen (GTT-Technologies) gave a lecture at NIMS as the 163rd Structural Materials Seminar.
Date: Wednesday, November 13, 2024, 13:30-15:00
Location: Conference Room, 5th Floor, Advanced Structural Materials Building, NIMS Sengen Site
Title: Thermodynamic genome for multicomponent materials: CALPHAD, ab initio, and machine learning
Speaker: Dr. Stephan Petersen(GTT-Technologies, ManagingDirector) https://gtt-technologies.de/
Orgnazer: Machiko Ode（Structural Thermodynamics Group）
Dr. Stephan Petersen(front row, third from right)
Abstract:
CALPHAD databases are the state-of-the-art for thermodynamic modelling of inorganic materials (metals, ceramics, slags, salts). However, the CALPHAD methodology is still a very manual process and requires the existence and human evaluation of large amounts of experimental data. Commercial CALPHAD databases therefore cover “only” few tens of elements with a focus on selected applications (such as steels, metallurgy or non-oxide ceramics). Ab initio databases on the other hand, such as materialsproject.org or oqmd.org, are not limited by the existence of experimental data and therefore cover larger parts of chemical space. The use of these databases for thermodynamic modelling is however restricted by both the temperature range (usually 0K), limited accuracy with respect to phase equilibria, and the difficulty in describing solution phase thermodynamics. In this presentation we demonstrate that machine learning techniques can be used to bridge the gap between CALPHAD and ab initio databases. This is especially important when modeling the processing of new functional materials or the recovery of minority elements during the recycling of complex end-of-life products.