Yu Dian Lim

AI-Designed Photonics Gratings with Experimental Verification

Yu Dian Lim [1], Chuan Seng Tan [1,2]

Abstract

Artificial Intelligence (AI) software based on transformer model is developed to automatically design gratings for possible integrations in ion traps to perform optical addressing on ions. From the user-defined (x,z) coordinates and full-width half-maximum (FWHM) values, the AI software can automatically generate the Graphic Design System (GDS) layout of the grating that shoots light towards the pre-defined (x,z) coordinates with built-in finite-difference time-domain (FDTD) simulation for performance verification. Based on the FDTD verification, AI-design gratings produced grating-to-free-space light that shoots towards the provided (x,z) target with < 2 micron deviations. For most attempts, the FWHM of FDTD simulation has < 2 micron deviations from the user-defined FWHM. The AI-designed gratings were successfully taped out and capable of producing output light for possible optical addressing of trapped ions.

Recognizing Beam Profiles from Silicon Photonics Gratings using Transformer Model

Yu Dian Lim [1], Hong Yu Li [2], Simon Chun Kiat Goh [3], Xiangyu Wang [2], Peng Zhao [1], Chuan Seng Tan [1,2]

Abstract

Over the past decade, there has been extensive work in developing integrated silicon photonics (SiPh) gratings for the optical addressing of trapped ion qubits in the ion trap quantum computing community. However, when viewing beam profiles from infrared (IR) cameras, it is often difficult to determine the corresponding heights where the beam profiles are located. In this work, we developed transformer models to recognize the corresponding height categories of beam profiles of light from SiPh gratings. The model is trained using two techniques: (1) input patches, and (2) input sequence. For model trained with input patches, the model achieved recognition accuracy of 0.938. Meanwhile, model trained with input sequence shows lower accuracy of 0.895. However, when repeating the model-training 150 cycles, model trained with input patches shows inconsistent accuracy ranges between 0.445 to 0.959, while model trained with input sequence exhibit higher accuracy values between 0.789 to 0.936. The obtained outcomes can be expanded to various applications, including auto-focusing of light beam and auto-adjustment of z-axis stage to acquire desired beam profiles.