CAREER: Automated Synthesis of Electromagnetic Devices for Nanophotonic and Radio Frequency Applications

对新技术的不断增长的性能和效率要求，如即将到来的5G蜂窝标准和下一代计算，将需要对电子硬件设计采取变革性的方法，一直到电磁学层面。通过计算机辅助设计技术，在电路，特别是数字逻辑电路的自动化和最优化方面取得了重大进展。事实上，数字合成或计算机自动设计数字电路的出现，使以前被认为不可能实现的硬件变成了现实，例如具有数十亿个晶体管的微处理器，其形状系数为几平方厘米，能够每秒执行数十亿次数学运算。目前，系统的底层电路组件与电磁(EM)模块分开设计，通常由不同的工程师甚至完全不同的机构设计。不幸的是，大多数电磁设备，如天线、微波设备，甚至光子集成电路中的纳米光子设备，都必须由人类工程师从头开始艰苦地手动设计。这一过程不仅繁琐和耗时，而且还不是最优的，而且由于可用于设计人类不可能探索的此类设备的巨大自由度，经常会留下显著的性能提升。该项目旨在通过开发一个在精神上类似于数字逻辑综合的EM器件自动综合平台来解决这些问题，这将为人类工程师节省大量手动设计这些器件的时间，并导致新的、非直观的结构。这将是第一个通用的综合平台，可以处理跨EM频谱的广泛类型的设备。这项研究将得到一项教育计划的补充，该计划包括发展一个新的研究生班，在多学科领域培训研究生和本科生，并激励K-12学生追求STEM的职业生涯。描述所有EM设备的麦克斯韦方程的解析解除了简单的玩具问题外，不存在。这大大复杂化了新EM设备的设计，并且需要启发式方法和许多耗时的手动参数扫描。尽管具有前所未有的性能和节省时间的潜力，但目前还不存在用于一般EM设备的自动化合成框架，这些设备可以以适度的计算能力运行。该项目将通过开发脚本语言来描述光学和射频范围内的任意EM设备，提出基于边界积分方程组的高速EM模拟方法以将计算时间和所需内存减少数个数量级，并利用能够实现几乎不需要人工干预的新设备的自动化优化算法，从而设计出这样一个自动化EM综合平台。该项目的长期目标是使即使是计算能力一般的非专家，如桌面工作站，也能在几分钟到几小时内快速高效地设计多种不同的电磁设备。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

The ever-increasing performance and efficiency demands on new technologies, such as the upcoming 5G cellular standard and next-generation computing, will necessitate a transformative approach towards electronic hardware design all the way down to the electromagnetics level. Significant progress has been made towards the automation and optimization of circuits, especially digital logic circuits, via computer-aided design techniques. In fact, the advent of digital synthesis, or the automated design of digital circuits by computers, has brought hardware into reality which was previously thought impossible, such as microprocessors with billions of transistors occupying form-factors of a few square-cm and capable of performing billions of mathematical operations per second. Presently, the underlying circuit components of a system are designed separately from the electromagnetic (EM) blocks, often by different engineers or even different institutions altogether. Unfortunately, most EM devices such as antennas, microwave devices, and even nanophotonic devices in photonic integrated circuits must be painstakingly designed manually from the ground up by a human engineer. This process is not only tedious and time-consuming, but also suboptimal and often leaves significant performance gains on the table due to the enormous degrees of freedom available for designing such devices which are impossible for a human to explore. This project aims to tackle these issues by developing an automated synthesis platform for EM devices, analogous in spirit to digital logic synthesis, which will save human engineers significant time spent designing these devices manually and lead to novel, non-intuitive structures. This will be the first generalized synthesis platform which can handle a wide class of devices across the EM spectrum. This research will be complemented by an educational plan which includes the development of a new graduate class, training graduate and undergraduate students in multi-disciplinary fields, and motivating K-12 students to pursue careers in STEM.Analytical solutions for Maxwell's equations, which describe all EM devices, do not exist except for simple toy problems. This significantly complicates the design of new EM devices and requires heuristic approaches and many time-consuming manual parameter sweeps. Despite the potential for unprecedented performance and time savings, there does not presently exist an automated synthesis framework for general EM devices which can operate with modest computing power. This project will lead to the design of such an automated EM synthesis platform by developing a scripting language to describe arbitrary EM devices in both the optical and radio-frequency regimes, advancing high-speed EM simulation methods based on boundary integral equations to decrease compute time and memory required by several orders of magnitude, and leveraging automated optimization algorithms capable of realizing new devices with little to no human intervention. The long-term goal of the project is to enable even the nonexpert with only modest computing capability, such as a desktop workstation, to design many types of different electromagnetic devices rapidly and efficiently in a matter of minutes to hours.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.