A python workflow definition for computational materials design

Area of Science:

• Computational Materials Science
• Scientific Workflow Management
• Software Interoperability

Background:

• Diverse workflow formats in computational materials science hinder interoperability and reproducibility.
• Existing Python-based Workflow Management Systems (WfMS) like AiiDA, jobflow, and pyiron share structural similarities in representing workflows as graphs.
• The need for a standardized exchange format to facilitate seamless workflow sharing and execution across different WfMS is critical.

Purpose of the Study:

• To introduce the Python Workflow Definition (PWD) as a novel workflow exchange format.
• To foster interoperability and reproducibility among Python-based WfMS (AiiDA, jobflow, pyiron).
• To promote the principles of Findable, Accessible, Interoperable, Reusable (FAIR) workflows in computational materials science.

Main Methods:

• Developed the Python Workflow Definition (PWD) comprising three components: conda environment for dependencies, Python module for workflow functions, and a JSON-formatted workflow graph.
• Implemented support for Directed Acyclic Graph (DAG)-based workflows in the initial PWD version.
• Created a PWD Python library enabling export and import of workflows between AiiDA, jobflow, and pyiron.

Main Results:

• Successfully demonstrated the export and import of DAG-based workflows between AiiDA, jobflow, and pyiron using the PWD format.
• The PWD facilitates the adjustment of input parameters and assignment of computing resources post-import.
• The PWD Python library effectively bridges the gap between the specified WfMS for workflow exchange.

Conclusions:

• The Python Workflow Definition (PWD) provides a standardized solution for sharing workflows across different Python-based WfMS.
• PWD significantly enhances workflow interoperability and reproducibility in computational materials science, aligning with FAIR data principles.
• The PWD framework, with its modular design and library support, is poised to streamline computational materials science research by simplifying workflow management and execution.