Collaborative Research: SHF: Small: RUI: CMOS+X: Honey-ReRAM Enabled 3D Neuromorphic Accelerator

当前的计算系统正面临着巨大的挑战，包括极高的能源需求、不可再生材料的巨大消耗、电子废物带来的环境和健康问题，以及缺乏通过将互补金属氧化物半导体（CMOS）微芯片与新兴器件技术（标记为X）集成来提高系统性能的技术（“CMOS+X”）。该项目将通过探索一种创新技术来解决这些挑战，该技术将CMOS外围电路与一种新的存储技术——基于天然有机蜂蜜的电阻开关随机存取存储器（honey- reram）集成在一起，用于一种新的受大脑启发的神经形态计算系统。原型“CMOS + honey-ReRAM”计算系统有望在工程、社会科学、国家卫生和国防等许多高影响领域促进高性能、节能和可持续的内存计算能力。此外，拟议的教育活动为代表性不足的研究人员提供了独特的培训机会，包括女性，非洲裔美国人和美洲原住民学生。该项目旨在探索器件制造和系统集成技术的创新，以优化honey-ReRAM器件，建立CMOS电路与honey-ReRAM阵列3D集成的可行性，并原型化CMOS + honey-ReRAM支持的神经形态加速器，这是计算和人工智能中广泛设备中数据处理的基本神经网络硬件组件。蜂蜜- reram将具有高度可复制的记忆特性、热稳定性、长期可靠性、低成本以及可持续性。蜂蜜ReRAM还将采用环保的合成工艺和设备制造。本文将开发一种具有正式设计流程的新型三维（3D）架构和制造技术，以确保通过微芯片上的异构集成将CMOS电路与蜂蜜- reram阵列相结合的兼容性。此外，本研究将为CMOS与X集成中的不兼容问题提供解决方案，并有效加速CMOS+X技术的发展和应用，以实现系统级计算的改进。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。

Current computing systems are facing significant challenges including extremely high energy demand, tremendous consumption of nonrenewable materials, environmental and health issues by electronic waste, and lack of technologies to improve system performance by integration of complementary metal oxide semiconductor (CMOS) microchips with emerging device technologies (denoted X) – “CMOS+X”. This project will address these challenges by exploring an innovative technology to integrate CMOS periphery circuits with a novel memory technology - natural organic honey based resistive switching random access memory (honey-ReRAM) for a new brain-inspired neuromorphic computing system. The prototyped “CMOS + honey-ReRAM” computing system is promising to promote high performance, energy efficient, and sustainable in-memory computing capability for many high impact domains such as engineering, social science, national health, and defense. In addition, the proposed education activities offer unique training opportunities for underrepresented researchers including female, African American, and Native American students.This project targets to explore innovations in device fabrication and system integration technologies to optimize honey-ReRAM devices, establish the feasibility of 3D integration of CMOS circuits with honey-ReRAM arrays, and prototype a CMOS + honey-ReRAM enabled neuromorphic accelerator, an essential neural network hardware component for data processing in a vast range of devices in computing and artificial intelligence. The honey-ReRAM will have highly reproducible memory characteristics, thermal stability, long-term reliability, low-cost, as well as being sustainable. The honey ReRAM will also utilize an eco-friendly synthesis process and device manufacture. A novel three-dimensional (3D) architecture and fabrication technology with a formal design flow will be developed to ensure the compatibility of combining CMOS circuitry with honey-ReRAM arrays by a heterogeneous integration on the microchip. Furthermore, this research will provide a solution to the incompatibility problem in the integration between CMOS and X and effectively accelerate the development and application of CMOS+X technologies for system-level improvements in computing.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.