Solid-state iontronic devices for advanced data processing

现代电子设备，如海量数据存储单元和逻辑处理单元，正变得越来越复杂，但功率越来越密集。不断推动提高集成化/小型化和减少能源足迹，需要在新材料、设备结构和计算架构中增加控制和功能。拟议的计划旨在应用我们最近对可充电离子电池系统中离子运动的研究知识，通过模仿人脑使用的高效神经系统，从一个新的角度解决电子挑战。人脑使用的高效神经系统可以在很小的外形尺寸中处理大量数据，能耗相当于现代灯泡使用的能耗。当神经元通过开启和关闭电压控制的离子通道来发射电脉冲时，离子控制是最自然的探索策略。锂离子是最小的金属离子，具有高离子迁移率，已经在充电电池应用中发挥了关键作用，发明者获得了2019年诺贝尔化学奖。PI的团队计划在一个截然不同的环境中使用锂离子。他们将开发方法，将这些离子电动员到指定位置，如接口或桥，以控制设备电导，即在现有设备上添加离子方面，如阻性或磁性随机存取存储器(ReRAM和MRAM)。这种离子运动可以对这些混合器件产生非易失性、时间依赖或脉冲依赖的控制。综合起来，PI的目标是创建一个与金属氧化物半导体(CMOS)兼容的互补的硬件神经形态计算平台，其中阻性开关突触适应输入触发脉冲，而磁隧道结神经元发射(翻转)来自突触的累积电流。这些构成了本提案的主要目标：离子MRAM(IMRAM)、具有专用离子通道的ReRAM，以及它们的CMOS集成，从而产生了一个硬件神经网络平台，该平台可以轻松地批量集成到更节能的神经形态计算架构中。与传统的计算机体系结构相比，这种变革性的技术将离子能量存储和现代集成电子技术的丰富知识联系在一起，实际上为现有设备增加了另一个控制维度，并允许在相同的空间和能源足迹下进行更智能的分布式集成。由于加拿大在人工智能算法的许多方面已经处于领先地位，该项目通过直接模拟人脑和神经系统的行为，加强了加拿大硬件部门的ICT部门。该项目的高度跨学科性质也极大地有利于加拿大的HQP培训，培养出非常适合材料科学、电子、纳米技术和可再生能源领域的各种学术和工业机会的新一代HQP。

Modern electronic devices, such as mass data storage units and logic processing units, are becoming increasingly sophisticated but increasingly power intensive. The constant push to enhance integration/miniaturization and reduce energy footprints requires additional controls and functionalities in novel materials, device structures, and computing architectures. The proposed program aims to apply knowledge from our recent research studies on ion motion in rechargeable ion battery systems to address electronics challenges from a new angle, by mimicking the highly efficient neuron systems used by the human brain which can process a tremendous amount of data in a small form factor with energy consumption equivalent to that used by a modern light bulb. Ionic control is the most natural strategy to explore as neurons fire electrical pulses by opening and closing voltage-controlled ionic channels. Lithium ions, the smallest metal ion with high ionic mobility, have already played a key role in rechargeable battery applications, with the inventors awarded the 2019 Nobel Prize in Chemistry. The PI's team plans to use lithium ions in a drastically different context. They will develop approaches to electrically mobilize these ions to designated locations such as interfaces or bridges to control the device conductance, that is, adding iontronic aspects to existing devices such as resistive or magnetic random access memories (ReRAM and MRAM). Such ionic motions can produce nonvolatile, time dependent or pulse dependent control over these hybrid devices. Adding together, the PI aims to create a complementary metal-oxide-semiconductor (CMOS) compatible, hardware neuromorphic computing platform where the resistive switching synapses adapt to input trigger pulses, and the magnetic tunnel junction neurons fire (flip) at the accumulated currents from the synapses. These form the main objectives of this proposal: ionic MRAM (iMRAM), ReRAM with dedicated ion channels, and their CMOS integration producing a hardware neural network platform that can be readily mass-integrated into more energy efficient neuromorphic computing architectures. Compared to classical computer architectures, this transformative technology links the wealth of knowledge from ionic energy storage and modern integrated electronics, practically adds another dimension of control to existing devices, and allows for smarter, distributed integration under the same space and energy footprints. With Canada already leading in many aspects of artificial intelligence algorisms, this program strengthens Canada's ICT sectors in the hardware departments by directly mimicking the behavior of the human brain and nervous system. The highly interdisciplinary nature of this program also greatly benefits HQP training for Canada, producing next-generation HQP well suited for diverse academic and industrial opportunities in materials science, electronics, nanotechnology, and renewable energy sectors.