Public Abstract

DE-SC0026053: Neuromorphic Computing Circuit Primitive Testbeds

Award Status: Active

Institution: University Of Delaware, Newark, DE
UEI: T72NHKM259N3
DUNS: 059007500

Most Recent Award Date: 09/15/2025
Number of Support Periods: 1
PM: Pino, Robinson

Current Budget Period: 06/01/2025 - 05/31/2026
Current Project Period: 06/01/2025 - 05/31/2030
PI: Zeng, Yuping

Supplement Budget Period: N/A

Public Abstract

Title: Neuromorphic computing circuit primitive testbeds Team: Yiran Chen (Duke University), Yuping Zeng (University of Delaware), Gina Adam (George Washington University), James Ang (Pacific Northwest National Laboratory) Analog neuromorphic circuit primitive testbeds are needed as the foundational building blocks of neuromorphic networks and systems. Neuromorphic computing aims to use insights from the brain organization and function to develop neuro-inspired computing technologies that are efficient and scalable for next generation scientific computing and artificial intelligence for scientific discovery. However, current digital technologies have limited bio-inspiration. A realistic neuromorphic computing circuit architecture needs to draw its foundations from the brain’s analog behaviors and it requires to accelerate the design of key analog computational components that underpin the biological neural systems. This project will research, design, model, simulate, prototype, improve, and characterize biologically plausible neuromorphic circuit primitives. Existing transistor technologies (CMOS) and emerging device technologies (X), e.g. ferroelectric field effect transistors (FeFETs) and oxide-based memristors (ReRAM) will be utilized to prototype key neuronal primitives and synaptic interconnects towards the realization of CMOS-only and CMOS+X neuromorphic testbeds. Three academic partners will develop and compare neuromorphic circuit primitives based on these technologies. Dr. Chen at Duke University has expertise in CMOS designs, Dr. Zeng at Univ. of Delaware is an expert in FeFET technology and Dr. Adam at George Washington University has experience in CMOS/memristor integration. These investigations across three technology prongs will ensure that these primitives will be validated for biologically-realistic functionality. The most promising primitives will be optimized, and the resulting testbeds will be documented and deployed at the Pacific Northwest National Laboratory in Dr. Ang’s team. The data obtained will be shared with the broader community. Shared tape-outs will be used to achieve an accelerated cycle of circuit design → prototyping → characterization → lessons learned for rapid innovation of neuromorphic primitives. The work is driven by six objectives that tackle the scientific and technological challenges necessary to push the boundaries of neuromorphic computing circuits. Objective 1 – Neuronal primitive testbeds explores the realization of neuronal primitive testbeds, via the implementation of Izhikevich neuronal models that capture essential spiking, adaptation, and reset dynamics in analog circuitry. Objective 2 – Synaptic interconnects investigates the synaptic primitives with different plasticity behaviors, such as spiking timing dependent plasticity and short-term plasticity, to support neuronal connectivity. Objective 3 – Testbeds validation and documentation will focus on the development of the necessary testing, evaluation and characterization equipment and simulation tools, software stack and abstraction methodologies for the primitives obtained in Objectives 1 and 2. The documentation will be prepared and maintained and the resulting measurement, performance and design data will be disseminated to the community. Objective 4 – Connectivity between primitive testbeds will explore how the resulting primitive testbeds from prior Objectives can be connected towards future demonstration of neuromorphic functionality in an integrated fashion. Objective 5 – Technology investigations will explore the latest advancements in CMOS, FeFET and memristor technologies. The memristor and FeFET devices will be optimized to realize desired threshold voltage and memory window based on the requirements from the CMOS circuitry. Issues related to the in-house integration of these devices onto foundry-sourced CMOS chips and wafers will be tackled. Objective 6 – Project management will support the scientific endeavors of the project. Shared tape-outs are planned periodically for chips and preplanarized wafers. Dissemination events will be planned to disseminate the results and the obtained data to the scientific community. These testbed capabilities will set the foundations towards the integration of the primitives into functional networks. This project directly addresses DOE’s mission in advancing computing for science via the development of neuromorphic computing architectures that can support future low power scientific computing and accelerate artificial intelligence applications for scientific discovery.

Title: Neuromorphic computing circuit primitive testbeds

Team: Yiran Chen (Duke University), Yuping Zeng (University of Delaware), Gina Adam (George Washington University), James Ang (Pacific Northwest National Laboratory)

Analog neuromorphic circuit primitive testbeds are needed as the foundational building blocks of neuromorphic networks and systems. Neuromorphic computing aims to use insights from the brain organization and function to develop neuro-inspired computing technologies that are efficient and scalable for next generation scientific computing and artificial intelligence for scientific discovery. However, current digital technologies have limited bio-inspiration. A realistic neuromorphic computing circuit architecture needs to draw its foundations from the brain’s analog behaviors and it requires to accelerate the design of key analog computational components that underpin the biological neural systems.

This project will research, design, model, simulate, prototype, improve, and characterize biologically plausible neuromorphic circuit primitives. Existing transistor technologies (CMOS) and emerging device technologies (X), e.g. ferroelectric field effect transistors (FeFETs) and oxide-based memristors (ReRAM) will be utilized to prototype key neuronal primitives and synaptic interconnects towards the realization of CMOS-only and CMOS+X neuromorphic testbeds. Three academic partners will develop and compare neuromorphic circuit primitives based on these technologies. Dr. Chen at Duke University has expertise in CMOS designs, Dr. Zeng at Univ. of Delaware is an expert in FeFET technology and Dr. Adam at George Washington University has experience in CMOS/memristor integration. These investigations across three technology prongs will ensure that these primitives will be validated for biologically-realistic functionality. The most promising primitives will be optimized, and the resulting testbeds will be documented and deployed at the Pacific Northwest National Laboratory in Dr. Ang’s team. The data obtained will be shared with the broader community. Shared tape-outs will be used to achieve an accelerated cycle of circuit design → prototyping → characterization → lessons learned for rapid innovation of neuromorphic primitives.

The work is driven by six objectives that tackle the scientific and technological challenges necessary to push the boundaries of neuromorphic computing circuits. Objective 1 – Neuronal primitive testbeds explores the realization of neuronal primitive testbeds, via the implementation of Izhikevich neuronal models that capture essential spiking, adaptation, and reset dynamics in analog circuitry. Objective 2 – Synaptic interconnects investigates the synaptic primitives with different plasticity behaviors, such as spiking timing dependent plasticity and short-term plasticity, to support neuronal connectivity. Objective 3 – Testbeds validation and documentation will focus on the development of the necessary testing, evaluation and characterization equipment and simulation tools, software stack and abstraction methodologies for the primitives obtained in Objectives 1 and 2. The documentation will be prepared and maintained and the resulting measurement, performance and design data will be disseminated to the community. Objective 4 – Connectivity between primitive testbeds will explore how the resulting primitive testbeds from prior Objectives can be connected towards future demonstration of neuromorphic functionality in an integrated fashion. Objective 5 – Technology investigations will explore the latest advancements in CMOS, FeFET and memristor technologies. The memristor and FeFET devices will be optimized to realize desired threshold voltage and memory window based on the requirements from the CMOS circuitry. Issues related to the in-house integration of these devices onto foundry-sourced CMOS chips and wafers will be tackled. Objective 6 – Project management will support the scientific endeavors of the project. Shared tape-outs are planned periodically for chips and preplanarized wafers. Dissemination events will be planned to disseminate the results and the obtained data to the scientific community. These testbed capabilities will set the foundations towards the integration of the primitives into functional networks. This project directly addresses DOE’s mission in advancing computing for science via the development of neuromorphic computing architectures that can support future low power scientific computing and accelerate artificial intelligence applications for scientific discovery.