Automating Battery Workflows with AiiDA— Towards Autonomous Labs that are FAIR by Design

Scientific discovery can be significantly accelerated by the development of autonomous workflows, driving simulations and experiments via artificial intelligence algorithms to discover new materials or optimise materials properties. A preliminary requirement for autonomous labs, though, is the availability of automated workflows both for simulations and experiments. On the simulation side, in the past decades robust workflows have been developed with the aim of performing automated high-throughput simulations carried out by workflow engines. One example of such engine is AiiDA (https://www.aiida.net), that has historically managed advanced simulation workflows, storing the full provenance (i.e., all the simulation steps, their inputs and outputs, and their logical relationships), thus ensuring full reproducibility of complete simulation processes. In this talk, I will discuss how we successfully interfaced off-the-shelf battery cycling hardware with AiiDA, allowing us to control experiments, in combination with the experimental scheduler tomato [2]. Advanced features of AiiDA have also been exploited to control experimental workflows, such as live monitoring for early detection of failed experiments. This integration serves as a first crucial step toward fully autonomous optimisation of batteries. I will highlight how the use of AiiDA is beneficial in the context of autonomous laboratories, where the amount of generated data can be very large, but with the advantage of being fully automated. It is thus of paramount importance to ensure that such data is properly stored with all its metadata to make results reproducible, and semantically annotated to facilitate searches and reuse. To illustrate this point, I will discuss the complexities of managing open research data (ORD) from mixed experimental/simulation workflows across research data management (RDM) platforms, and I will highlight the efforts of our PREMISE project (https://ord-premise.org), funded by the Swiss ETH Board ORD program. The mail goal of PREMISE is to enable and facilitate ORD interoperability between experiments and simulations, laying the foundation of autonomous laboratories that produce data that is FAIR by design, removing most of the burden from researchers to make their data open. Finally, I will briefly discuss additional use cases of autonomous integration of experiments and simulations. Also in the context of batteries, I will mention the integration of AiiDA with the FINALES brokering software, developed as part of the European BIG-MAP project (https://www.big-map.eu). FINALES has been already successfully demonstrated to scale to experiments distributed at the international level, with fully automated FAIR data storage and with semantic annotations [3,4]. Additional use cases include the control of crystal growth and the characterisation of the Fermi surface of metals.<br/><br/>[1] P. Kraus et al., J. Mater. Chem. A 12, 10773 (2024), https://doi.org/10.1039/D3TA06889G<br/>[2] P. Kraus et al., Tomato, https://dgbowl.github.io/tomato<br/>[3] M. Vogler et al., Matter 6, 2647 (2023), https://doi.org/10.1016/j.matt.2023.07.016<br/>[4] M. Vogler et al., ChemRxiv (2024), https://10.26434/chemrxiv-2024-vfq1n