Schofield Parker, Bradicich Adelaide, Gurrola Rebeca M, Zhang Yuwei, Brown Timothy D, Pharr Matt, Shamberger Patrick J, Banerjee Sarbajit
Department of Chemistry, Texas A&M University, College Station, TX, 77843, USA.
Department of Materials Science and Engineering, Texas A&M University, College Station, TX, 77843, USA.
Adv Mater. 2023 Sep;35(37):e2205294. doi: 10.1002/adma.202205294. Epub 2022 Nov 29.
Future-generation neuromorphic computing seeks to overcome the limitations of von Neumann architectures by colocating logic and memory functions, thereby emulating the function of neurons and synapses in the human brain. Despite remarkable demonstrations of high-fidelity neuronal emulation, the predictive design of neuromorphic circuits starting from knowledge of material transformations remains challenging. VO is an attractive candidate since it manifests a near-room-temperature, discontinuous, and hysteretic metal-insulator transition. The transition provides a nonlinear dynamical response to input signals, as needed to construct neuronal circuit elements. Strategies for tuning the transformation characteristics of VO based on modification of material properties, interfacial structure, and field couplings, are discussed. Dynamical modulation of transformation characteristics through in situ processing is discussed as a means of imbuing synaptic function. Mechanistic understanding of site-selective modification; external, epitaxial, and chemical strain; defect dynamics; and interfacial field coupling in modifying local atomistic structure, the implications therein for electronic structure, and ultimately, the tuning of transformation characteristics, is emphasized. Opportunities are highlighted for inverse design and for using design principles related to thermodynamics and kinetics of electronic transitions learned from VO to inform the design of new Mott materials, as well as to go beyond energy-efficient computation to manifest intelligence.
下一代神经形态计算旨在通过将逻辑和存储功能置于同一位置来克服冯·诺依曼架构的局限性，从而模拟人脑中神经元和突触的功能。尽管已经有了高保真神经元模拟的显著成果，但从材料转变的知识出发对神经形态电路进行预测性设计仍然具有挑战性。氧化钒（VO）是一个有吸引力的候选材料，因为它表现出接近室温的、不连续的和滞后的金属-绝缘体转变。这种转变为构建神经元电路元件所需的输入信号提供了非线性动态响应。本文讨论了基于材料特性、界面结构和场耦合的修改来调整VO转变特性的策略。通过原位处理对转变特性进行动态调制被作为赋予突触功能的一种手段进行了讨论。强调了对位点选择性修饰、外部、外延和化学应变、缺陷动力学以及界面场耦合在修改局部原子结构方面的机理理解，以及其中对电子结构的影响，最终对转变特性的调整。文中突出了逆向设计的机会，以及利用从VO中学到的与电子跃迁的热力学和动力学相关的设计原则来指导新型莫特材料的设计，以及超越节能计算以实现智能的机会。
