FuSe-TG: Co-designing Novel Memristor Heterostructures for Brain Inspired Computers

脑启发计算机(BIC)引起了人们的极大兴趣，因为BIC潜在地可以执行传统数字计算机无法轻松执行的独特任务。因此，BIC有可能改变信息的处理和存储方式。BIC的关键电子元件是忆阻器，这是一种可以模拟大脑运行模式的设备。本项目旨在将二维(2D)半导体和传统的忆阻器材料相结合，开发一种新型的忆阻器。这种器件比目前最先进的忆阻器更节能、更可靠，而且可扩展。这一目标需要一个联合设计过程，其中包括一组具有不同专业知识的研究人员，如可扩展材料合成、原子模拟、原子级结构和特性表征、器件研究和电路建模。该项目通过在这一研究人员团队之间建立牢固的联系，并通过不断分享知识和实施反馈循环，在这一过程中及时调整规划和方法，从而使这种共同设计过程成为可能。通过利用物理、材料科学和电气工程的各种技术为BIC开发这种高效和可靠的关键硬件组件，可能会给BIC领域和广泛的计算领域带来革命性的变化。团队建设工作包括建立年度研讨会，允许教师团队成员和他们的学生会面并讨论该领域的初步数据和最新发展，以及不仅为直接参与该项目的学生提供劳动力培训，还为那些参与教师团队成员积极参与的当地社区各种外展项目的学生提供劳动力培训。该项目涉及的跨学科合作活动将有助于发展一支专注于未来半导体行业的多元化劳动力队伍，并促进来自当地社区和广泛的跨学科学术社区中代表性较低群体的学生成员。记忆电阻是BIC的关键组件，因为记忆电阻统一了信息处理和存储，模拟大脑。传统的过渡金属氧化物记忆阻器存在可靠性差、器件性能不均匀、集成度低、功耗高等问题。为了克服这些问题，基于原子薄2D材料和传统记忆电阻材料组成的异质结构的新型忆阻器将通过该团队成员的集体努力来开发。这些在新机制下工作的忆阻器异质结构将利用2D材料和传统的忆阻器材料来实现高效和可靠的电阻器件。这项提议旨在建立一个由材料科学家、物理学家和电气工程师组成的团队，共同设计用于BIC的新型忆阻器。合作设计过程涉及四个方面：1)从原子水平了解二维忆阻器的阻性开关的工作机制，并确定合适的忆阻器组件候选方案，包括电极、有源阻层和选择器；2)以可扩展的方式实验合成所有组件，并将这些组件集成到工作的忆阻器中；3)设计器件，以优化每个组件的功能，产生最节能和最健壮的单个忆阻器，并在忆阻器阵列中运行良好；4)神经元电路设计，以指导更高规模的忆阻器集成，以实现小规模BIC，着眼于可扩展的BIC。该项目将使团队成员之间建立紧密的联系，基于由2D材料和传统记忆电阻材料组成的异质结构，高效地共同设计下一代忆阻器。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Brain-inspired-computers (BICs) are of great interest, because BICs potentially can perform distinctive tasks that cannot be easily executed by conventional digital computers. Thus, BICs have the potential to transform how information is processed and stored. The key electronic component for BICs is a memristor- a device that can mimic the brain’s operation mode. This project aims to develop a new type of memristor by combining two-dimensional (2D) semiconductors and traditional memristor materials. Such a device is more energy efficient and reliable than current state-of-the-art memristors and is scalable. This goal requires a co-design process that involves a team of researchers with diverse expertise such as scalable material synthesis, atomistic simulations, atomic-scale structural and property characterization, device studies and circuitry modeling. This project enables such a co-design process by establishing a strong connection among this team of researchers and by constant knowledge sharing and implementation of feedback loops that can lead to timely adjustment of planning and methodology in the process. Developing such highly efficient and reliable key hardware components for BICs by utilizing various techniques in physics, materials sciences and electrical engineering can potentially revolutionize the field of BICs and the broad field of computing. The team building effort includes establishing an annual workshop that allows the faculty team members and their students to meet and discuss preliminary data and recent developments in the field, and work-force training not only for students directly involved in the project but also for those involved in various outreach programs in the local community the faculty team members actively engage in. The interdisciplinary teaming activities involved in this project will help develop a diverse workforce that focuses on future semiconducting industries and promote student members from underrepresented groups both in the local community and broad interdisciplinary academic communities.Memristors are key components for BICs as a memristor unifies information processing and storage, emulating the brain. The traditional transition metal oxide based memristor faces problems such as poor reliability, non-uniform device-to-device performance, low integration density, and high-power consumption. To overcome these problems, novel memristors based on heterostructures composed of atomically thin 2D materials and traditional memristor materials will be developed through a collective effort by members in this team. These memristor heterostructures operating under new mechanisms will take advantage of both 2D materials and traditional memristor materials to enable highly efficient and reliable resistive devices. This proposal aims to establish a team composed of materials scientists, physicists, and electrical engineers to co-design novel memristors for BICs. The co-design process involves four thrusts: 1) Atomic-level understanding of the operation mechanism of resistive switching of the 2D memristor and identifying suitable candidates for memristor components, including the electrodes, active resistive layer and the selector; 2) Experimental synthesis of all components in a scalable way and integration of these components into a working memristor; 3) Device design to optimize the functionalities of each component and yield the most energy-efficient and robust individual memristor that also performs well in a memristor array; 4) Neuronal circuit design to guide higher-scale integration of memristors to realize a small-scale BIC, with the view towards scalable BICs. This project will enable the team to develop a tight connection among the team members to efficiently co-design next generation memristors based on heterostructures composed of 2D materials and traditional memristor materials.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.