Soft-AE: Enabling accessible autonomous exploration of conducting polymer and nanocomposite systems

Developing efficient polymeric charge conductors for chemical separations and energy systems involves an increasingly multi-dimensional design problem with equal parts thermodynamic and non-equilibrium processing factors. At the same time, the burgeoning scope of candidate polymers and additives makes screening exercises burdensome and impractical. To help elucidate these formulation-processing-property relationships, we present Soft-AE, a multimodal autonomous experimentation (AE) and characterization platform with self-driving laboratory (SDL) capabilities.

We showcase multi-axis Bayesian campaigns in a model poly(ethylene oxide)-based system that explores Li⁺:EO ratio and silica filler content, resolved across temperature and relative humidity. Enabled by accessible embedded systems and rapid prototyping, we achieve navigation of nontrivial ionic conductivity landscapes with odd, difficult to predict extrema. We outline the limits of efficient exploration across a given design space with up to 32 samples, with strategically chosen boundary and sampling parameters and no physics-informed priors. Adding a human into the loop takes advantage of batched sampling as an input to Bayesian workflows, further expanding capacity for discovery.

In developing this system, we contribute to the ecosystem of accessible systems supporting novel soft-matter exploration, and opportunities to interface with AE/SDLs for greater aggregate impact in polymer materials science.