FuSe-TG: Co-design of Attojoule Multifunction Semiconductor Electronics with Atomic Precision

由于人工智能（AI）、机器学习（ML）和集成数据和计算的分布式边缘应用程序的广泛采用推动了数据处理的大规模扩展，未来的半导体系统将面临巨大挑战。根据半导体研究公司的“2021年10年计划”，计算的能源效率必须提高100万倍。一个令人信服的解决方案是使半导体器件更有能力，即多功能。具有带隙的材料不一定使其成为半导体，直到存在杂质离子，即掺杂剂，以调整其电导率和其他特性。调动半导体中的掺杂离子打开了一扇具有几乎无限机会的窗口。未来的半导体很可能在飞行中以创纪录的低阿焦耳（10-18焦耳）能级的原子精度进行动态可重构，这类似于但优于生物突触系统。该资助旨在组建一个团队，为具有原子精度的阿焦耳半导体材料、多功能器件和未来超低功耗计算和内存的智能系统开发模拟和人工智能指导的协同设计框架，从而实现一个无处不在的人工智能社会。教育目标是建立一个跨学科联盟，培养下一代半导体网络劳动力，他们将通过创新地使用高端计算、量子计算和人工智能的先进网络基础设施来解决具有挑战性的材料-设备-系统协同设计问题。该资助旨在建立具有aJ能耗和原子精度的可重构多功能突触开关的颠覆性范例，从而为节能边缘计算创造一种新的工业方法。关键创新包括：(1)质子电化学离子突触具有确定性，实现了阿焦耳能量消耗下的高速切换；(2)传感器内计算系统，在Edge上处理无外部电源的图像；(3)基于第一性原理的多尺度模拟和人工智能指导下的Synaptic材料-器件-系统协同设计，为半导体的未来提供可推广的理性协同设计框架（FuSe）。底层软件套件被开发成CyberFuse培训模块，可通过CyberFuse门户访问。培训模块在课堂上进行试点，以支持双学位课程（物理、材料科学或电气工程博士学位，计算机科学或人工智能硕士学位），并在CyberFuse培训研讨会上教授，重点关注代表性不足的群体。此外，这项资助还为社区学院的学生提供就业途径，这些学生占全国本科生的三分之一以上。该资助还通过(1)南加州大学的女性科学与工程（WiSE）项目和(2)由南加州大学、麻省理工学院、斯坦福大学、CMU、TAMU和霍华德大学（历史最悠久、规模最大的黑人学院和大学之一）的教师共同监督的代表性不足的群体的本科研究扩大了参与范围。FuSe团队包括来自SLK美国、应用材料和IBM的工业合作伙伴，以加速技术转移和对研究和未来劳动力需求的反馈。该项目由半导体未来计划（FuSe）和传统黑人学院和大学本科计划（HBCU-UP）共同资助。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。

Future semiconducting systems will face formidable challenges due to massive expansion of data processing that is driven by the broad adoption of artificial intelligence (AI), machine learning (ML), and distributed Edge applications that integrate data and computing. It is imperative to increase the energy efficiency of computing million-fold, according to the 2021 Semiconductor Research Corporation decadal plan. A compelling solution is to make semiconductor devices more capable, i.e., multifunctional. A material with a bandgap does not necessarily make it a semiconductor until there are impurity ions, namely dopants, to tune its electrical conductance and other properties. Mobilizing the dopant ions in semiconductors opens a window with nearly infinite opportunities. It is likely that future semiconductors will be dynamically reconfigurable on the fly with atomic precision at record-low attoJoule (10-18 Joule) energy level, which resembles but outperforms biological synaptic systems. This grant is to forge a team for developing a simulation and AI guided co-design framework for attoJoule semiconductor materials with atomic precision, multifunctional devices, and smart systems for future ultralow-power computing and memory, thereby realizing a sustainable society with ubiquitous AI. The educational goal is to establish a cross-disciplinary coalition that trains a future generation of semiconductor cyberworkforce, who will solve challenging material-device-system co-design problems through innovative use of advanced cyberinfrastructure at the nexus of high-end computing, quantum computing and AI.The grant aims to establish a disruptive paradigm for reconfigurable multifunctional synaptic switching with aJ energy consumption and atomic precision, thereby creating a new industrial approach for energy-efficient Edge computing. Key innovations include: (1) Protonic electrochemical ionic synapse that is deterministic and achieves high-speed switching with attoJoule energy consumption; (2) In-sensor computing systems to process images without external power supply at Edge; (3) Synaptic material-device-system co-design guided by first principles-based multiscale simulation and AI, thus providing a generalizable rational co-design framework for the future of semiconductors (FuSe). The underlying software suite is developed into a CyberFuse training module that is accessible through a CyberFuSe portal. The training modules are piloted in classrooms to support a dual-degree program (Ph.D. in physics, materials science or electrical engineering with MS in computer science or AI) and taught in CyberFuse training workshops with a strong focus on underrepresented groups. In addition, the grant provides career pathways to community college students, who constitute more than one third of the nation’s undergraduate students. The grant also broadens participation through (1) USC’s Women in Science and Engineering (WiSE) program and (2) undergraduate research by underrepresented groups jointly supervised by faculty from USC, MIT, Stanford, CMU, TAMU and Howard - one of the oldest and largest historically black colleges and universities. The FuSe team includes industrial partners from SLK America, Applied Materials and IBM for accelerated technology transfer and feedback on research and future workforce needs.This project is jointly funded by the Future of Semiconductors (FuSe) program and by the Historically Black Colleges and Universities Undergraduate Program (HBCU-UP).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.