Accelerated (Re)Design of Sustainable Formulations with The Autonomous Formulation Lab

Industrial liquid formulations from electronics to nanoparticle coatings to drug delivery vehicles are often
strikingly complex, with large numbers of components (10-100), complex multistep processing,
and a wide variety of design requirements for a functional, sustainable, regulatory compliant product. This complexity often
precludes physics-informed mapping between component fractions, processing, structure,
leaving most formulation design to empirical trial-and-error or design of experiments strategies.
I will discuss recent efforts by the Autonomous Formulation Laboratory (AFL) team at NIST to
use machine learning driven, highly automated characterization to rapidly and intelligently map
formulation phase space using structural characterization tools such as small-angle x-ray and
neutron scattering (SAXS/SANS) together with secondary measurements, e.g. optical imaging,
UV-vis-NIR and capillary viscometry. Our initial studies have resulted in an order of magnitude
reduction in the time needed to map a model phase diagram and rapid cross-learning between model petroleum-based and biobased formulations. Other systems of interest from our personal care, biopharmaceutical, and alternative energy industrial collaborators will also be highlighted, including recent results on multimodal cooperative active learning.