Novel photonic devices for optical neural networks

该项目将从开发单个像素作为光学神经网络系统的基本构建块开始。像素将由一个光栅结构组成，支持作为反射器的导模共振。通过在光电可调材料（如氧化铟锡（ITO））中实现光栅，反射率可以被电控制。特别的挑战是使用我们的内部溅射系统来控制材料参数，并设计器件结构，使光束的振幅或相位都可以调制。在这里，学生将受益于为传感应用开发的导模共振的现有工作；例如，我们现在能够检测到Dn<10-5的折射率变化，这意味着同样小的电光诱导折射率变化将导致光束的明显调制，为低功耗和高速运行开辟了前景。该项目结合了设计和模拟，材料研究和制造以及结构和光学表征，这让学生对负责他们研究的所有相关元素感到满意。该学生将成为一个充满活力的跨学科团队的一员，并接触到一系列相关的研究领域，提供令人兴奋的材料，设备（包括建模），纳米制造和应用方面的交叉受精机会。他们还将通过我们与当地中小企业Optalysys的定期会议，接触到工业现实。Optalysys正在将类似的光学神经网络系统商业化。

The project will start by developing a single pixel as the fundamental building block of an optical neural network system. The pixel will consist of a grating structure supporting a guided mode resonance that acts as a reflector. By realising the grating in an electro-optically tunable material such as Indium Tin Oxide (ITO), the reflectivity can be controlled electrically. The particular challenge is to control the material parameters using our in-house sputtering system and design the device structure such that either the amplitude or the phase of the beam can be modulated. Here, the student will benefit from the body of existing work on guided mode resonances developed for sensing applications; for example, we are now able to detect refractive index changes of Dn<10-5, which means that similarly small electro-optically induced index changes will lead to a noticeable modulation of the beam, opening the prospect for low-power and high-speed operation.The project incorporates design and simulation, materials research and fabrication as well as structural and optical characterisation, which gives the student the satisfaction of being in charge of all relevant elements of their research. The student will be part of a vibrant interdisciplinary team and be exposed to a range of related research areas offering exciting cross-fertilisation opportunities regarding materials, devices (incl. modelling), nanofabrication and applications. They will also be exposed to industrial reality through our regular meetings with Optalysys, a local SME who are commercialising a similar optical neural network system.