Department of Materials Science and Engineering Spring Seminar Series: Arvind Murugan

Description

Arvind Murugan, an associate professor at the University of Chicago's Department of Physics, will give a talk titled "Materials with Physically Learned Phase Diagrams" for the Department of Materials Science and Engineering.

Abstract:

Materials are typically designed to have one fixed behavior through an optimal choice of parameters at the time of synthesis. We propose an alternative learning framework for a learning material in which molecular interactions adapt via a molecular Hebbian principle: "get together, grow to like each other." In multicomponent systems, this Hebbian learning can position phase boundaries within a phase diagram so that the material can compute and deploy specific responses to distinct high-dimensional molecular inputs. In this way, these learned phase boundaries function as decision boundaries for high-dimensional inputs, offering an alternative route to neural computation without neurons in material systems. Our theory and experiments show that these ideas enable diverse molecular systems—such as self-assembly, dimerization networks and phase separation—to learn associations and correctly respond to complex chemical patterns.