Search

Conventional design and optimization of Photonic Crystal Surface Emitting Lasers (PCSEL) usually requires expert knowledge in semiconductor physics and optimization algorithms, which is also known as the inverse design problem. However, with the trend towards automation and depersonalization of the entire integrated circuits (IC) industry, the conventional method, with the drawback of being relatively labor-intensive and sub-optimal, warrants further refinement. This technical dilemma remained until the emergence of Large Language Models (LLMs), such as OpenAI's ChatGPT and Google's Bard. This paper explores the possibility of applying LLMs to machine learning-based design and optimization of PCSELs. Specifically, we utilize GPT3.5 and GPT4. By simply having conversations, GPT assisted us with writing Finite Difference Time Domain (FDTD) simulation code and deep reinforcement learning code to acquire the optimized PCSEL solution, spanning from the proposition of ideas to the realization of algorithms. Given that GPT will perform better when given detailed and specific questions, we break down the PCSEL design problem into a series of sub-problems and converse with GPT by posing open-ended heuristic questions rather than definitive commands. This paper shows that LLMs, such as ChatGPT, can guide the nanophotonic design and optimization processes, on both the conceptual and technical level, and we propose new human-AI co-design strategies and show their practical implications. We achieve a significant milestone for the first step towards an automated end to end nanophotonic design and production pipeline., Comment: 14 pages