使用改进型八木-宇田纳米等离子体天线控制纳米光子网络中的耦合强度
Controlling the Coupling Strengths in Nanophotonic Networks using Modified Yagi-Uda Nanoplasmonic Antennas
浏览论文内容
中文总结 AI 辅助
该研究提出用改进型八木-宇田纳米等离子体天线结合III-V族纳米线，构建可灵活调控节点间耦合权重的紧凑型纳米光子网络，可用于实现复杂连接的神经网络。
中文摘要 AI 辅助
光神经网络中的片上通信通常通过波导实现，但这会导致系统占地面积过大。一种紧凑的替代方案是在自由空间中纳米光电器件之间广播光信号，并利用亚波长纳米光子学定制光场分布。我们提出并模拟了纳米等离子体金属结构与III-V族纳米线发射器和接收器相结合的方案，以构建一种光网络，其中这些纳米结构的形状和位置可在纳米光电子节点之间产生不同的权重。利用时域有限差分（Finite Difference Time Domain）建模，我们研究了经实验验证的纳米线光电器件与纳米等离子体结构相结合的系统，这些结构可在为纳米线供电的电接触的同一制造步骤中制备。我们同时研究了单个纳米线/天线器件以及对应神经网络两层的网络。我们表明，使用八木-宇田（Yagi-Uda）天线可显著改变纳米线节点的定向通信。通过修改天线使其具有不对称引向器组件，可实现将光同时引导至多个不同角度方向。我们发现，根据天线组件的组合几何形状，可在两个七节点层之间的连接中实现高度可变的复杂权重分布。纳米线节点紧凑层中的可能权重分布可用于创建具有复杂连接的各种神经网络。这些概念可推广到其他类型的纳米级发射器/接收器系统。
英文摘要
On-chip communication in optical neural networks is commonly done via waveguides which results in large system footprints. A compact alternative is to broadcast light signals between nano-optoelectronic components in free space and tailor the light field distribution using sub-wavelength nanophotonics. We propose and simulate nanoplasmonic metal structures in combination with III-V nanowire emitters and receivers to create an optical network in which the shape and position of these nanostructures create varying weights between the nano-optoelectronic nodes. Using Finite Difference Time Domain modelling, we investigate systems of experimentally verified nanowire optoelectronics combined with nanoplasmonic structures that can be made in the same fabrication step as electrical contacts powering the nanowires. We investigate both individual nanowire/antenna devices as well as networks corresponding to two layers in a neural network. We show that directed communication from a nanowire node can be significantly altered using Yagi-Uda antennas. Modifying the antenna with an asymmetric director component enables the directing of light several different angular directions simultaneously. We find that highly variable complex weight distribution between the connections in two seven node layers can be achieved depending on the combined geometry of the antenna components. The possible weight distributions in compact layers of nanowire nodes could be used for creating a variety of neural networks with complex connectivity. The concepts can be generalized to other types of nanoscale emitter/receiver systems.
发表机构
- Lund University（隆德大学）
机构由 AI 辅助整理，请以论文原文为准。
