Biomolecular Neuristors from Voltage-Activated Lipid Membranes

Spiking neural networks (SNNs) show promise as energy-efficient and fast computing devices for adaptive spatiotemporal data processing and classification, which is critical for applications requiring efficient acquisition, and processing of time-dependent data. Taking inspiration from their biological counterparts, the artificial neuristors in SNN hardware systems have been typically designed to generate dynamic voltage spikes -i.e., <i>action potentials- </i>when a cumulative stream of received electrical inputs crosses a critical threshold. However, unlike biological neurons, action potential generation in man-made systems is achieved using complex circuitry of many solid-state devices that are often non biocompatible. Examples of artificial neurons, and <i>neuristor</i> circuits, include those based on threshold-switching memristors^[1] and organic electrochemical transistors^[2].<br/><br/>We recently demonstrated the use of voltage-activated lipid membranes as artificial synapses for neuromorphic computing. Advantages of this approach include the available diversity of functional biomolecules, the low voltage and power requirements, and the potential for greater biocompatibility and scalability. To-date, we have demonstrated that voltage-activated biomembranes can exhibit tunable memory resistance^[3], memory capacitance^[4], and various forms of activity-dependent plasticity that provide specific computational advantage^[5].<br/><br/>Herein, we study the behaviors of biomolecular neuristors consisting of voltage-activated lipid membranes exposed to alamethicin (ALM), protamine and histidine peptides. Remarkably, two distinct types of action potentials—fast (~10 ms) and slow (~1s)—have been observed, with their characteristic speeds and spike shapes differing based on the composition of the device and the direction and amplitude of the applied current. We have conducted experiments at varying compositions and concentrations of species to identify the molecular mechanisms for spike generation and develop mathematical models able to capture these behaviors.<br/><br/>While earlier approaches depended on complex multi-device circuitry, a biomolecular neuristor achieves this within a single stimuli-responsive biomembrane. This simplifies the fabrication process by reducing complexity, material volume, and cost, all while maintaining its adaptability to various input types. Leveraging the characteristics of synthetic biomembranes, this work can help unlock the potential for creating compact, energy-efficient, and biocompatible neuristors, thus driving progress in the realm of neural interface technology and spiking neural networks (SNNs).<br/><br/>[1] M. D. Pickett, G. Medeiros-Ribeiro, R. S. Williams, Nature Materials 2013, 12, 114.<br/>[2] T. Sarkar, K. Lieberth, A. Pavlou, T. Frank, V. Mailaender, I. McCulloch, P. W. M. Blom, F. Torricelli, P. Gkoupidenis, Nature Electronics 2022, 5, 774.<br/>[3] J. S. Najem, G. J. Taylor, R. J. Weiss, M. S. Hasan, G. Rose, C. D. Schuman, A. Belianinov, C. P. Collier, S. A. Sarles, ACS Nano 2018, 12, 4702.<br/>[4] J. S. Najem, M. S. Hasan, R. S. Williams, R. J. Weiss, G. S. Rose, G. J. Taylor, S. A. Sarles, C. P. Collier, Nature communications 2019, 10, 1.