Autonomous Experimentation

Soft materials often comprise high-dimensional compositional and processing parameter landscapes that can be burdensome to thoroughly explore given realistic experimental resource constraints. To help accelerate the mapping and utilization of structure-property-function relationships in soft materials, we are developing Soft-AE, a high-throughput, multimodal autonomous experimentation platform assembled with a combination of budget open-source and custom-prototyped hardware. Driven by a central Python-based interface, the system integrates liquid handling, microscopy, thermal processing, patterned optical fields, and electrochemical measurements, and couples this to flexible/tailorable implementation of machine learning algorithms for autonomous decision-making.

The ability to flexibly trade off exploration and exploitation in a soft materials context broadens the potential window for both fundamental and applied discoveries. We integrate open-source machine learning algorithms such as gpCAM (https://camera.lbl.gov/projects/gpcam) into our workflows, developing best practices and theory-aided methods to improve AE campaign outcomes and navigate exploration-exploitation tradeoffs.

In addition to its potential research benefits, we envision Soft-AE to become an effective and accessible platform for training the scientific workforce, providing valuable experience in interfacing with and developing AE. It is designed to be constructed and maintained on a limited budget, ensuring wide and flexible deployment in a variety of research environments.

In tandem with Membrane Separations, we are also interested in applying high-throughput and autonomous experimentation to explore polymer and composite formulations for efficient charge transport.