From Self-Driving Laboratories to Connected Experimentation in the Physical Sciences: Lessons from the Critical Materials Value Chain
Self-driving laboratories close the loop between prediction, synthesis, and measurement on individual platforms, yet most physical-science research outside user facilities proceeds through multistage workflows distributed across separate laboratories, with manual handoffs of samples, protocols, and data. This review assesses how autonomous experimentation can extend across such workflows, using the critical minerals and materials value chain for energy technologies as a use case. The chain is a demanding test: dilute, heterogeneous feedstocks require characterization, separations, and materials synthesis performed in different laboratories with different instruments and safety bases. For capabilities demonstrated across U.S. Department of Energy national laboratories and the international literature, the review distinguishes closed-loop optimization from fixed-design screening, reports acceleration with explicit denominators, and identifies where human intervention remains. It examines what prevents single-site capabilities from compounding across laboratories: missing interlaboratory evidence for key measurands, unqualified handoffs of physical samples, and unassigned decision rights. These constraints are not specific to critical materials; they arise wherever bench-scale physical science spans sites. Building on published federated-experimentation frameworks, the review outlines a qualification profile for cross-site handoffs and closes with research needs in fundamental science, computing and verification infrastructure, and process development, each paired with a milestone and measure of success.
Authors
- Long Hao Qi (ORCID: https://orcid.org/0000-0002-1213-2946)
- Phillip Halstenberg (ORCID: https://orcid.org/0000-0002-6030-4503)
- Elias Nakouzi (ORCID: https://orcid.org/0000-0003-0036-8326)
- Zachery Crandall (ORCID: https://orcid.org/0000-0003-3161-9378)
- Thomas J. Tarka (ORCID: https://orcid.org/0000-0002-6105-4366)
- Nancy Washton (ORCID: https://orcid.org/0000-0002-9643-6794)
- Craig A. Bridges (ORCID: https://orcid.org/0000-0002-3543-463X)
- Danny Pérez (ORCID: https://orcid.org/0000-0003-3028-5249)
- Xuan Cao (ORCID: https://orcid.org/0000-0002-7211-6999)
- Lun An (ORCID: https://orcid.org/0000-0003-2146-1490)
- Noah H. Paulson (ORCID: https://orcid.org/0000-0002-3548-9120)
- Zachary P. Tener (ORCID: https://orcid.org/0000-0003-1407-0379)
- Prashant Singh (ORCID: https://orcid.org/0000-0001-9648-2275)
- Christopher Sobecki
- Lindsay E. Roy
- Mason Sage
- Maxim Ziatdinov (ORCID: https://orcid.org/0000-0003-2570-4592)
- Hongkyu Yoon
- Ben Mintz
- Ping Yang
- Andrew Ritchhart
- Hongyou Fan
- Grant E. Johnson
- Michael Whittaker
- Sara Allec
- Guangping Xu
- Scott McCall
- Jennifer Bauer
- Sheng Dai
Institutions
- National Energy Technology Laboratory (US)
Publication Details
- Journal
- ChemRxiv
- Published
- 2026-11-16
- DOI
- https://doi.org/10.26434/chemrxiv.15012308/v1
- Primary Topic
- Machine Learning in Materials Science
- Type
- preprint