Design and Synthesis of Atomically Tunable Memristors

神经形态计算（NC）是一种受大脑启发的计算，最近作为解决当前基于独立处理和存储单元的计算中数据移动的冯·诺伊曼瓶颈的有希望的解决方案而出现。NC使信息能够在同一单元中进行处理和存储，有望为新兴的人工智能（AI）、机器学习、物联网等提供关键的计算硬件，并可能与量子计算融合，实现量子神经形态计算。研究人员试图通过使用人工版本的记忆电阻器来模拟生物大脑的运作。忆阻器被认为是继电阻器、电容器和电感器之后的第四个基本电路元件，它由一层薄薄的电介质（通常是氧化物）薄膜和一对电极夹在一起，以模仿生物大脑运作的功能。在忆阻器中，存储器、存储和计算的操作都集成在一个器件中。然而，随着数控领域的发展，在原子尺度上控制记忆电阻、开关速度、循环耐力以及其他性能标准的能力变得越来越重要，因为数控电路可能需要将不同性能的器件紧密集成在一起。特别是，具有可调谐特性的忆阻器对于模拟生物大脑运作至关重要。不幸的是，由于对超薄（sub- 3nm）氧化物薄膜的生长缺乏控制，对忆阻器参数的原子尺度控制尚未实现，从而导致缺陷，从而导致当前忆阻器的能量效率降低，器件不均匀性和低产量。提出的研究旨在通过基于超薄氧化物原子层堆栈（ALS）的原子控制合成原子可调谐忆阻器的协同集成，预测ALS物理性质的原子材料模拟/建模，以及材料和非原位器件的先进表征来解决这些挑战。该项目的成功可以对包括数控、人工智能、量子信息科学等在内的大量商业应用产生更广泛的影响。从科学上讲，材料研究中一个长期存在的问题是，几个原子层以原子精度堆叠是否可以提供先进电子产品所需的功能和大面积均匀性。在过去的五十年中，晶体管的尺寸不断缩小到目前的5纳米以下，以满足未来电子产品对原子精度功能调整的需求，这推动了这个问题。利用在PI之前的NSF支持下开发的真空ALD方法，拟议的研究旨在通过原子材料模拟/建模和材料和器件的原位和非原位高级表征指导下的原子控制氧化ALS合成的协同集成来解决挑战。具体来说，在两个拟议的目标中，Aim 1侧重于超薄氧化物ALS的设计和合成，Aim 2研究材料和器件层面的ALS特性。该研究的智力价值在于对超薄氧化ALS的物理性质有了新的基本理解，这对于实现未来电子和计算的原子可调谐记忆电阻器至关重要，并且在开发设计和合成超薄ALS的新方法以实现新功能方面具有变革性飞跃。基于ALS实现的原子可调谐忆阻器可以对包括神经形态和量子计算、人工智能、量子信息科学等在内的大量商业应用产生更广泛的影响。该项目强调前沿教育和前沿研究能力，这将吸引高质量的学生，特别是那些来自代表性不足群体的学生，在STEM领域从事职业，并将进一步扩大对培养独特和多样化的未来科学和工程劳动力的影响。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。

Neuromorphic computing (NC) is a brain-inspired computing that has recently emerged as a promising resolution to the von Neumann bottleneck of data movement in current computing based on separate processing and memory units. NC enables the information to be processed and stored in the same units and is expected to provide critical computing hardware for the emerging artificial intelligence (AI), machine learning, internet of things, etc., and may converge with quantum computing for quantum neuromorphic computing. Researchers have attempted to mimic biological brain operation by using artificial versions of the elements known as memristors. Memristor, regarded as the fourth fundamental circuit element next to resistor, capacitor, and inductor, consists of a thin dielectric (typically oxide) film of pinched hysteresis of resistance sandwiched with a pair of electrodes to mimic the functionality of biological brain operation. In memristors, the operations of both memory storage and computing are integrated in one device. However, as the field of NC evolves, the ability to control at the atomic scale the memristive resistance, switching speed, cycling endurance, among other performance criteria, becomes increasingly important as NC circuits potentially require devices with different performance capabilities closely integrated together. In particular, memristors with tunable properties are critical to mimic biological brain operation. Unfortunately, an atomic-scale control of the memristor parameters has not been achieved due to lack of control in growth of ultrathin (sub-3 nm) oxides films, resulting in defects that in turn lead to reduced energy efficiency, device non-uniformity and low yield in current memristors. The proposed research aims to address the challenges through a synergetic integration of atomically controlled synthesis of atomically tunable memristors based on ultrathin oxide atomic layer stacks (ALS), atomistic material simulation/modeling that predicts the physical properties of ALS, and advanced characterization both in situ on materials and ex situ on devices. The success of the project can have a broader impact on a large spectrum of commercial applications including NC, AI, quantum information science, etc. Scientifically, a long-standing question in material research is whether a few atomic layers stacked with an atomic precision can provide the functionality and large-area uniformity as required for advanced electronics. This question is driven by the continuous down-sizing of transistors in last five decades to currently sub-5 nm to meet the need in future electronics with functionality tuned with an atomic precision. Using an in vacuo ALD approach developed in PI’s prior NSF support, the proposed research aims to address the challenges through a synergetic integration of atomically controlled synthesis of oxide ALS guided by atomistic material simulation/modeling and advanced characterization both in situ and ex situ on materials and devices. Specifically in the two proposed aims, Aim 1 focuses on design and synthesis of ultrathin oxide ALS and Aim 2 investigates ALS properties at material and device levels. The intellectual merit of the proposed research is through a new fundamental understanding of physical properties of ultrathin oxide ALS that is crucial to achieving atomically tunable memristors for future electronics and computing and a transformative leap in developing novel approaches for design and synthesis of ultrathin ALS to enable new functionalities. The achieved atomically tunable memristors based on the ALS can have a broader impact on a large spectrum of commercial applications including neuromorphic and quantum computing, artificial intelligence, quantum information science, etc. The project emphasizes forefront education and the cutting-edge research capability which will attract high-quality students, especially those from underrepresented groups, to pursue careers in STEM fields and will further amplify the impacts towards producing a unique and diverse future workforce in science and engineering.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.