Design And Performance Analysis Of The Photonic Crystal Nanobeam Cavity Sensor Based On The Neural Network

The photonic crystal nanobeam cavity has attracted extensive attention due to its high quality factor,small mode volume,compact footprint,and easy integration.In the nanobeam cavity sensor phylogeny,various structures emerge one after another.However,the design process for these structures requires massive computing resources inevitably.In recent years,the rapid development of artificial neural networks has brought new vitality to the design of nanophotonic devices.By learning the nonlinear relationship between the device structures and the corresponding optical response,the massive consumption of computing resources caused by traditional numerical methods can be overcome,and the dilemma of falling into non-global optimal solution caused by traditional inverse design and optimization algorithms can be solved.Therefore,this research aims at improving the design and optimization efficiency of the photonic crystal nanobeam cavity,and designing nanobeam sensors with high performances.This research proposed to apply the artificial neural networks in the design of nanobeam cavity sensors.By fitting the band structure and transmission spectrum of the nanobeam with the geometric structure parameters through the neural network,several high-performance nanobeam sensors are designed.The main research of this paper include the following contents:Firstly,an ultrabroad bandgap elliptical hole dielectric mode nanobeam refractive index sensor is designed with the assistance of trained artificial neural networks.By introducing quadratically tapered elliptical holes for the major and minor axes into the silicon waveguide,the dielectric mode nanobeam cavity is achieved.And the cell structures of the cavity are obtained through the prediction by the forward prediction network and the optimization by the inverse design network.For the trained forward prediction and inverse design neural networks,the mean square errors are 3.0×10^-4 and 1.8×10^-2,respectively.With the help of an artificial neural network,it takes only 0.73 seconds to complete the design and optimization of the cavity sensor,proving the great improvement in the design efficiency.In terms of cavity performance and for the designed cavity,the highest quality factor of 1.43×10⁷ and the lowest mode volume of 1.81(λ_res/n_si)³ can be obtained.In terms of the sensing performance,the sensitivity for the refractive index is 230 nm/RIU,and the footprint for the structure is 7.5×0.7μm².According to the simulation results,this structure shows superiorities of high sensitivity and compact footprint in the field of refractive index sensing.Secondly,with the help of the artificial neural network,the design of the nanobeam cavity with the coexistence of air and dielectric mode based on the band structure and the transmission spectrum is chosen.And a slotted elliptical hole nanobeam cavity with the coexistence of air and dielectric mode for simultaneous sensing of the refractive index and temperature is designed.For the design based on the band structure,the analysis and the optimization of the band structure for each unit cell can be carried out by the artificial neural network,so the structure which can confine two different modes at the same time can be obtained.The mean square errors for the forward prediction and inverse design neural networks are 5.1×10^-4 and 1.4×10^-2,respectively.For the design based on the transmission spectrum,the complexity caused by following the deterministic design method can be avoided.And the inverse design task can be realized by inputting the transmission spectrum with the characteristics of the dual-mode cavity.The mean square errors for the forward prediction and the inverse design neural networks are 2.3×10^-3 and 6.2×10^-3,respectively.We choose the structure designed based on the band structure to carry out the performance discussion.In terms of cavity performance and under the condition of ensuring the transmittance,the highest quality factor of 9.34×10⁴ and 1.55×10⁵ can be achieved for the air mode and the dielectric mode,respectively.In terms of sensing performance,the refractive index sensitivity for the air mode is 406 nm/RIU,and the temperature sensitivity is 40 pm/K,whereas the refractive index sensitivity for the dielectric mode is 520 nm/RIU,and the temperature sensitivity is 27 pm/K.The maximum anti-external interference ranges for the refractive index and temperature are 6.93×10^-6 and 0.09.According to the simulation results,this structure shows superiorities in sensitivity and anti-interference ability in the field of refractive index and temperature dual-parameter sensing.In summary,this paper focuses on the design and performance analysis of photonic crystal nanobeam sensors based on artificial neural networks.In our work,artificial neural network models for forward prediction and inverse design between the nanobeam structures and the optical response of band structure and transmission spectrum are established.Based on the models,nanobeam cavities are designed and optimized.A single-parameter refractive index sensor and a dual-parameter temperature and refractive index sensor are designed.Their sensing performances are further analyzed.The present work has a certain significance in the nanobeam sensor field,and is also of significance for further research on the application of artificial neural networks in other micro-nano optical devices design field.