Max-Planck-Institut für Eisenforschung, Düsseldorf, Germany
Rapid prototyping and up-scaling atomistic workflows with pyiron
Ab-initio thermodynamics enables application-specific alloy design. This is achieved by addressing three levels of complexity: (1) the technical complexity of coupling simulation codes and methods developed in different communities, (2) the chemical complexity of iterating over the periodic table and (3) the thermodynamic complexity of identifying the dominant contributions to the phase stability.
To address these challenges, we developed the pyiron framework (https://pyiron.org/) which enables the development of workflows by combining the individual components – pyiron objects – like building blocks. Starting with rapid prototyping in a jupyter notebook, pyiron supports the user through the whole simulation lifecycle including the up-scaling of their simulation workflows to thousands of compute nodes and finally the publication of the workflow.
To highlight the impact of the pyiron framework on our research, we present three parameter studies and the unique insights we derived from these: starting with the uncertainty propagation of plane-wave density functional theory, over the fitting of interatomic machine learning potentials, up to the prediction of thermodynamic properties like the melting temperature.