Doctoral Dissertations
Date of Award
5-2024
Degree Type
Dissertation
Degree Name
Doctor of Philosophy
Major
Biomedical Engineering
Major Professor
Stephen A. Sarles
Committee Members
Xiaopeng Zhao, Subhadeep Chakraborty, Catherine Schuman, Joseph Najem
Abstract
There is growing demand for scalable and efficient computing paradigms due to the increasing prevalence of artificial intelligence in everyday life. Traditional computing based on the Von Neumann architecture suffers from a bottleneck in performance when large amounts of data must be written to and retrieved from memory. At the same time, the collective trends in transistor density and performance previously dubbed “Moore’s Law” are showing signs of decline. These factors favor the development of new architectures to increase efficiency and prevent the Von Neumann bottleneck. One approach to developing suitable new architectures in this direction is neuromorphic computing. Due to low power consumption and massively parallel processing, human brains are extremely efficient. Studying the underlying structures and principles that make this possible is a promising route for developing new devices and materials to meet the needs of modern computation. One such device category following neuromorphic principles is the artificial synapse. They mimic biological synapses, which are the connection points between neurons. While most artificial synapses are made from solid state metal oxides, silicon, and silver, they can also be made from artificial biological membranes. These devices are called biomolecular synapses (BSs). BSs can incorporate membrane active molecules from biology to impart new behaviors. Theoretically this implies that any biological synapse behavior can be emulated by a BS.
Herein, I characterize BS electrical properties; model BS dynamic behavior; and use BSs as artificial synapses in a time-domain machine learning application called reservoir computing. Chapters 2 and 4 demonstrate that two different voltage-gated ion channels, monazomycin and alamethicin, impart a form of synaptic plasticity called sensory adaptation to BSs that is rare in other artificial synapses. Chapter 3 shows that monazomycin enables multiple forms of synaptic plasticity to occur concurrently. Chapter 5 focuses on using BSs in a reservoir computing framework and shows through simulation and experiment that sensory adaptation improves accuracy and reduces error. I introduce a novel gust classification task for use in aerospace engineering. In Chapter 6, BSs are interfaced with high performance organic electrochemical transistors that enable short term memory and neuromorphic paired-pulse facilitation and depression.
Recommended Citation
Maraj, Joshua, "Characterization, Modeling, and Applications of Biomolecular Artificial Synapses. " PhD diss., University of Tennessee, 2024.
https://trace.tennessee.edu/utk_graddiss/10140
Included in
Artificial Intelligence and Robotics Commons, Biology and Biomimetic Materials Commons, Biomaterials Commons, Biomechanics and Biotransport Commons, Signal Processing Commons, Statistical, Nonlinear, and Soft Matter Physics Commons