CAREER: High-Performance Ferroelectric Memory for In-Memory Computing

随着智能设备的广泛部署，各种形式的数据，如照片、视频和传感器数据，正在以前所未有的速度产生。从数据中提取信息以进行智能决策需要巨大的计算能力。根据摩尔定律，计算机的计算能力一直在稳步提高。然而，随着技术进步接近物理极限，技术进步正在减少。在这方面，能够克服现有计算硬件挑战的替代计算范式是非常可取的。内存计算（IMC）是一种很有前途的方法，它通过直接在内存中执行计算而无需在计算单元和内存之间传输数据来显示高性能和能效，这是传统计算机的一个关键瓶颈。然而，为了充分利用IMC，紧凑、节能、高性能的嵌入式非易失性存储器技术至关重要。提出的研究旨在利用最近发现的铁电HfO2来开发这种存储技术。铁电存储器由于其电场驱动的写入机制，具有优越的能量效率。本研究旨在解决铁电存储器中存在的问题。对于基于电容的铁电随机存取存储器，探讨了准无损极化传感器件的设计。对于铁电场效应晶体管，采用后端兼容金属氧化物通道和双端口结构，具有优异的耐用性、无读干扰特性和最小的电荷捕获。本文还探讨了所提出器件的内存计算应用。与研究工作紧密结合的是5年的教育活动。通过向K-12和社区大学的学生和教师提供讲座和实践经验，目标是促进兴奋并吸引他们进入半导体行业的人才管道。拟议的研究将涉及本科生和研究生，特别是那些代表性不足的群体。模糊学科界限的课程开发是未来硬件开发的方向。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。

With the wide deployment of smart devices, various forms of data, such as photos, videos, and sensor data, are being generated at an unprecedented rate. Extracting information from the data for intelligent decision making requires enormous computing power. Following the Moore’s law, the computing power has been steadily improved. However, technology gain is diminishing as it is approaching the physical limits. In this regard, alternative computing paradigms that can overcome the challenges of existing computing hardware are highly desirable. In-memory computing (IMC) is one promising approach that shows high performance and energy efficiency by performing computation directly inside the memory without data transfer between the computing unit and memory, a key bottleneck in conventional computers. However, to fully exploit the IMC, a compact, energy-efficient, and high performance embedded nonvolatile memory technology is critical. The proposed research aims to develop such a memory technology by leveraging recently discovered ferroelectric HfO2.Ferroelectric memory due to its electric field driven write mechanism, exhibits superior energy efficiency. The proposed research aims at addressing the remaining issues in ferroelectric memory. For capacitor-based ferroelectric random access memory, device design that enables quasi-nondestructive sensing of polarization are explored. For ferroelectric field effect transistor, back-end-of-line compatible metal-oxide channel and dual-port structure are adopted for excellent endurance, read disturb free feature, and minimal charge trapping. In-memory computing applications of proposed devices are also explored. Tightly coupled with the research efforts are 5-year education activities. Through the lectures and hands-on experience offered to K-12 and community college students and teachers, the goal is to promote excitement and attract them into the talent pipeline for semiconductor industry. The proposed research will involve undergraduate and graduate students, especially those in under-represented groups. Course development that blurs the boundaries between different disciplines is proposed for hardware development in the future.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.