Collaborative Research: RUI: Natural Bio-organic Resistive Random Access Memory Based Synaptic Devices

当今全球计算系统面临的两个基本挑战是巨大的能源消耗和电子废物。同时解决这两个问题的一个潜在解决方案是“类脑”和“绿色”神经形态计算，具有节能操作和可生物降解的处理方式。神经形态计算系统需要能够模仿人类突触的硬件组件，这是生物神经网络的基本构建块，而来自活体或曾经活过的生物体（如植物、动物或微生物）的天然生物有机材料是可再生的、可持续的、生物相容性的、可生物降解的，并且在自然界中含量丰富。本研究将通过纳米加工和机器学习技术，推动基于天然生物有机材料的电阻性随机存储器的纳米级、超高密度和晶圆级制造的发展，并在“绿色”神经形态系统中实现基于生物有机材料的电阻性随机存储器的高精度和高效率神经网络。这个项目对美国和全球社会产生了巨大的影响，并提供了许多社会效益。基于电阻记忆的生物有机材料的神经形态系统是可拉伸、柔性和可穿戴电子产品在个人健康和生物医学应用中的理想选择，并解决了电子产品不可再生资源的过度开发和电子设备的处置带来的可持续和环境问题。该研究项目的跨学科性质涵盖了纳米技术、非易失性记忆、神经元和突触、神经形态计算系统和机器学习的理解和实践，为研究和教育的整合提供了一个完美的场所。少数民族、女性和高中生将被指导进行纳米技术和机器学习方面的研究。将开发一个基于虚拟现实的交互系统，在虚拟洁净室环境中提供电阻性记忆和突触装置制造的培训。将举办讲习班，以扩大传播和社区外展。该研究旨在解决阻碍基于电阻性记忆和人工突触装置的生物有机材料发展的技术挑战。这些挑战包括基于纳米级、高密度和可扩展的生物有机材料的电阻性记忆和突触器件的制造，以及将这些器件高精度和高效率地集成到神经网络中。在这个项目中，先进的纳米技术和纳米制造技术将被开发，以制造纳米尺度和高密度的生物有机材料为基础的电阻存储器的纳米尺寸的横条电极。机器学习算法将用于研究生物材料薄膜过程和性质、器件开关特性和突触行为之间的相关性。基于纳米级生物有机材料电阻记忆的突触结构将被开发，以模拟突触可塑性和突触效能。在神经网络中实现基于生物有机材料的电阻性记忆和突触装置，并通过利用连贯的硬件和软件协同设计来评估学习能力。该项目具有潜在的变革性，将在实现基于天然生物有机材料的电阻开关存储器和人工突触器件的纳米级、超高密度和晶圆级制造方面取得突破。该研究成果将通过精确的工艺优化加速器件开发，并在“绿色”神经形态计算系统的神经网络中使用时，建立对基于天然生物有机材料的电阻开关存储器和突触器件的基本理解。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Two essential challenges faced globally by computing systems today are tremendous energy consumption and electronic wastes. One potential solution to simultaneously address these two issues is by “brain-like” and “green” neuromorphic computing with energy-efficient operation and biodegradable disposals. Neuromorphic computing systems require hardware components capable of mimicking human synapse - the basic building block of biological neural networks, while natural bio-organic materials derived from living or once-living organisms such as plants, animals or microbial materials are renewable, sustainable, biocompatible, biodegradable, and abundant in nature. The proposed research will advance the development of nanoscale, ultrahigh-density and wafer-level manufacturing of natural bio-organic materials based resistive random access memory through nanofabrication and machine learning, and implementation of bio-organic materials based resistive memory in neural networks with high accuracy and efficiency for “green” neuromorphic systems. This project has great impacts on US and global societies and provides many societal benefits. The neuromorphic systems using bio-organic materials based resistive memory are desirable for stretchable, flexible and wearable electronics in personal health and biomedical applications, and address the sustainable and environmental issues brought by excessive exploitation of non-renewable resources for electronics and disposal of electronic devices. The interdisciplinary nature of this research project covers the understanding and practice in nanotechnology, non-volatile memory, neuron and synapse, neuromorphic computing systems and machine learning, which provide a perfect venue for integration of research and education. Minority, female and high school students will be mentored to perform research in nanotechnology and machine learning. A virtual reality based interactive system will be developed to provide trainings of resistive memory and synaptic device fabrication in a virtual cleanroom environment. Workshops will be organized for broadening dissemination and community outreach.The research aims to address technological challenges hampering the development of bio-organic materials based resistive memory and artificial synaptic devices. These challenges include the fabrication of nanoscale, high-density and scalable bio-organic materials based resistive memory and synaptic devices and incorporation of these devices in the neural network with high accuracy and efficiency. In this project, advanced nanotechnology and nanofabrication techniques will be developed to fabricate nanometer-sized crossbar electrodes for nanoscale and high-density bio-organic materials based resistive memory. Machine learning algorithms will be employed to study the correlation of biomaterial film process and property, device switching characteristics and synaptic behaviors. Synaptic architectures based on nanoscale bio-organic materials based resistive memory will be developed to emulate synaptic plasticity and synaptic efficacy. Implementation of bio-organic materials based resistive memory and synaptic devices in neural networks and evaluation of the learning capability will be carried out by leveraging a coherent hardware and software co-design. This project is potentially transformative and will achieve a breakthrough in the realization of nanoscale, ultrahigh-density and wafer-level manufacturing of resistive switching memory and artificial synaptic devices based on natural bio-organic materials. The research outcomes will expedite device development by accurate process optimization and establish a fundamental understanding of natural bio-organic materials based resistive switching memory and synaptic devices when used in the neural networks for “green” neuromorphic computing systems.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.