Biological & Soft Matter Seminar: From Re-examining the Physics of Action Potentials to Computing with Sound Pulses

Abstract:

Excitable cells generate a characteristic transient change in transmembrane potential that propagates along the cell membrane in response to suitable stimuli. These signals, known as action potentials (APs), are primarily associated with the behavioral activities of many organisms. The mechanism underlying an AP is widely regarded as electrical and is typically modeled by representing the cell membrane as an equivalent electric circuit. However, this theory relies on phenomenological equations that require numerous fitting parameters. Moreover, several experimental observations remain neither readily explained nor predicted by the electrical theory. In contrast, solitary longitudinal pulses propagating within lipid monolayers exhibit remarkable similarities to APs, and the theory of sound offers a compelling alternative that bridges some of these gaps without requiring any fitting parameters. After presenting the debate, we will propose that, given their similarities to APs, longitudinal pulses might be harnessed for material-based computing, mimicking biological or artificial neural algorithms. We demonstrate that longitudinal pulses carry more information than typically accounted for in neural models. For example, they propagate both digital and analog information about the stimulus across different thermodynamic dimensions. Additionally, we employ longitudinal pulses to construct a feedforward physical neural network capable of performing classification and regression tasks, and we are currently investigating the richness of the transformations produced by the system, as well as its ability to make predictions on real datasets. Our findings highlight the potential of harnessing nonlinear longitudinal pulses—prevalent in numerous materials—for computational purposes.