Ultrafast and Energy-efficient Anti-ferromagnetic Electric-field-controlled Memory Devices

由于人工智能和自主系统的数据密集型计算任务的增长，对计算系统中新的片上存储器和数据存储解决方案的需求正在快速增长。计算系统的性能越来越取决于从系统的内存组件中读取或写入数据的速度和能效，而不是执行逻辑操作的速度。这一趋势正在推动半导体行业从以逻辑为中心的计算架构向以内存为中心的计算架构的范式转变，在这种架构中，数据的存储和处理紧密集成以提高计算效率。这种以内存为中心的计算范式在速度、开关能量、耐用性和制造过程方面对内存设备提出了新的要求。然而，目前的内存解决方案不能充分扩展以满足所有这些需求，并且由于泄漏和/或刷新要求而受到待机功耗的影响。该项目的重点是开发一种新型的双端磁存储器件，称为反铁磁压控存储器，以满足以存储器为中心的计算应用的要求。除了技术和科学方面的影响，该项目还将影响西北大学本科生和研究生的教育，包括女性和代表性不足的少数民族。学生们将在物理学、材料科学和电气工程的交叉点研究、开发和实施最先进的纳米制造和新兴存储设备的高速测量。由于该项目的跨学科性质，他们还将参与与西北大学和其他大学的工业界和其他部门的合作。该项目还利用并促进了西北大学正在进行的推广工作，使公众参与到科学的进步中来。本项目在磁性随机存取存储器的器件结构中包含两个创新元素：(i)使用具有单轴磁各向异性的金属反铁磁层进行数据存储。这与现有磁存储器中传统使用的铁磁自由层不同，并提供了许多优点：首先，由于反铁磁体的总磁矩为零，因此所得到的器件对外部磁场具有鲁棒性，不需要任何磁屏蔽。其次，它消除了位对位偶极子相互作用，这可能会导致缩放挑战，并增加铁磁存储阵列的错误率。第三，使用反铁磁共振可以实现更快的开关，由于反铁磁材料的大内置交换场，其频率可以达到太赫兹范围。电压控制磁各向异性将用于节能写入数据。具体来说，一个非常短的电压脉冲将被用来启动谐振动力学，并将被定时以导致电压诱导的N？el向量。开关发生时不需要电流，从而减少写入能量达到每比特1焦耳范围。电压控制磁各向异性效应以前只在铁磁器件中得到证实。因此，它在反铁磁体中的演示和应用是本提案的关键智力优点之一。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。

There is a fast-growing demand for new on-chip memory and data storage solutions in computing systems, fueled by the growth of data-intensive computing tasks for artificial intelligence and autonomous systems. Increasingly, the performance of computing systems is determined by the speed and energy efficiency of reading data from or writing it into the memory components of the system, rather than the speed with which logic operations can be performed. This trend is driving a paradigm shift of the semiconductor industry from logic-centric to memory-centric computing architectures, where storage and processing of data are closely integrated to increase computational efficiency. This memory-centric computing paradigm puts new requirements on the memory devices in terms of speed, switching energy, endurance, and manufacturing processes. However, present memory solutions do not scale adequately to address all of these demands and suffer from standby power dissipation due to leakage and/or refresh requirements. This project is focused on the development of a new type of two-terminal magnetic memory device, referred to as antiferromagnetic voltage-controlled memory, to address the requirements of memory-centric computing applications. In addition to its technical and scientific impact, the project will impact the education of students at undergraduate and graduate levels, including women and underrepresented minorities at Northwestern. The students will study, develop, and implement state-of-the-art nano-fabrication and high-speed measurements for emerging memory devices, working at the intersection of physics, material science, and electrical engineering. Due to the interdisciplinary nature of the project, they will also take part in collaborations with industry and other departments at Northwestern and other universities. The project also leverages and contributes to on-going outreach efforts at Northwestern, engaging the general public in the advancement of science.The proposed project incorporates two innovative elements in the device structure of magnetic random-access memory: (i) It uses metallic antiferromagnetic layers with uniaxial magnetic anisotropy for data storage. This is different from the conventionally used ferromagnetic free layers in existing magnetic memory, and provides a number of advantages: First, due to the zero overall magnetic moment of the antiferromagnet, the resulting device is robust against external magnetic fields, without any requirements for magnetic shielding. Second, it eliminates bit-to-bit dipole interactions, which may result in scaling challenges and increased error rates in ferromagnetic memory arrays with tight pitch. Third, much faster switching can be achieved using antiferromagnetic resonance, which can have frequencies up to the Terahertz range due to the large built-in exchange field of the antiferromagnetic material. (ii) Voltage-Controlled Magnetic Anisotropy will be used for energy-efficient writing of data. Specifically, a very short voltage pulse will be used to initiate the resonant dynamics and will be timed to result in a voltage-induced complete switching of the N?el vector. The switching occurs without the need for electric current, thus reducing write energy to reach the atto-Joule per bit range. The voltage-controlled magnetic anisotropy effect has previously only been demonstrated in the case of ferromagnetic devices. Hence, its demonstration and utilization in the case of antiferromagnets is one of the key intellectual merits of the present proposal.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.