Session: 06-04 Neuromorphic Computing

Paper Number: 171954

171954 - Memristive Emulsion-Based Compartmentalized Tissues Containing Voltage-Gated Peptides 

Neuromorphic computing aims to emulate the structure and efficiency of biological brains by integrating memory and computation within the same physical substrate. Unlike traditional von Neumann architectures—where processing and storage are physically separated, resulting in latency and energy inefficiency—biological neural systems perform massively parallel, distributed processing while consuming less than 20 W of power. In contrast, modern supercomputers can draw upwards of 40 MW. This disparity has motivated the development of materials and architectures that replicate the spatially distributed and memory-integrated nature of biological neural networks.

To emulate these functions at the molecular scale, we investigate voltage-gated droplet interface bilayer (DIB) networks formed from either the amphiphilic block copolymer (BCP) poly(butadiene)-b-poly(ethylene oxide) (PB-PEO) or the phospholipid diphytanoyl phosphatidylcholine (DPhPC) and doped with alamethicin—a voltage-sensitive pore-forming peptide. Prior studies have shown that single DIBs incorporating alamethicin exhibit memory resistance (memristance), characterized by nonlinear, hysteretic current-voltage (i-v) relationships and conductance switching at biologically relevant thresholds. However, such single-membrane systems are mechanically fragile, prone to failure, and require an ambient oil environment for operation, thus limiting their scalability and practical utility in neuromorphic circuits.

To overcome these limitations, we introduce a soft, multicompartmental architecture in which millions of monolayer-encased aqueous droplets (10–100 µm) can be generated using rapid emulsification and centrifugation. This process rapidly suspends droplets in oil, coats them with lipids or BCPs, and compresses them into jammed, shape-holding tissue-like networks that are stable in both air and aqueous environments. Importantly, biomimetic bilayers form at shared interfaces between adjacent droplets, creating a robust and reconfigurable network of voltage-responsive membranes.

While preliminary results suggest promising neuromorphic behavior in compartmentalized tissues, several key questions remain. These include how to intentionally control the “active volume” of signal-conducting regions, the degree of spatial variability in memristive behavior across millions of droplets, and how local differences in droplet compositions or arrangement affect the collective function of the network. The stability of the conductance response across repeated stimulation cycles and over time is also not yet fully characterized. We hypothesize that by distributing memristive functionality across a large network of membranes, the system can buffer local defects, maintaining global performance even when individual membranes vary or fail. Additionally, we propose that tuning droplet composition and tissue-geometry can be used to modulate local conductive and memristive properties, enabling spatial patterning of functional domains within the tissue. Such tunability would allow us to design soft materials with programmed regions of high or low activity, creating a structured platform for adaptive, spatially resolved information processing.

Preliminary measurements demonstrate that these alamethicin-doped droplet networks exhibit volatile memristance. Electrophysiological recordings reveal voltage thresholds for channel insertion and nonlinear i-v curves with pronounced hysteresis. Using planar multielectrode arrays, we plan to measure ionic signal propagation across the tissue, characterizing how spatial location, geometry, and electrode configuration influence conduction and activation. Comparisons to measurements obtained via wire-type electrodes inserted into the tissue should reveal differences in localized versus distributed stimulation and highlight spatial heterogeneity in membrane behavior.

To interpret signal transmission across multi-membrane networks, we are also developing a model that incorporates experimentally measured bilayer resistance, specific capacitance, and alamethicin channel conductance. This allows estimation of voltage differences across individual membranes and prediction of scaling behavior as the size of the network increases.

Together, this work aims to advance the design of neuromorphic materials that integrate memory, computation, and adaptive ion transport using soft, biomolecular materials similar to those in the brain. Ongoing efforts aim to quantify variability in memristive response across droplets, assess the stability of conductance states over time, and explore spatial patterning strategies to intentionally tune network-level functionality.

Presenting Author: McKayla Torbett-Dougherty University of Tennessee

Presenting Author Biography: I am a Ph.D. student in Biomedical Engineering at the University of Tennessee, Knoxville. My research focuses on the design and characterization of bioinspired and tissue-like materials for applications in neuromorphic computing and synthetic biology. I specialize in the development of model membranes and compartmentalized systems that mimic the structure and function of biological tissues.