Composable generation strategy framework enabled bidirectional design on topological circuits

作     者：Xi Chen Jinyang Sun Xiumei Wang Maoxin Chen Qingyuan Lin Minggang Xia Xingping Zhou 

作者机构：College of Integrated Circuit Science and Engineering Portland Institute College of Electronic and Optical Engineering Department of Applied Physics School of Physics Institute of Quantum Information and Technology 

出 版 物：《Physical Review B》 (Phys. Rev. B)

年 卷 期：2024年第110卷第13期

页      面：134108页

核心收录：

基　　金：NUPTSF, (NY220119, NY221055) National Natural Science Foundation of China, NSFC, (11774278) National Natural Science Foundation of China, NSFC Chengdu Medical College, CMC, (61409220140) Chengdu Medical College, CMC 

主　　题：Edge states Electronic structure Topolectrical circuits Topological insulators Topological materials Language models Deep learning Machine learning Topological tools 

摘      要：Topological insulators show important properties, such as topological phase transitions and topological edge states. Although these properties and phenomena can be simulated by well-designed circuits, it remains a complex task to design such topological circuits due to the intricate physical principles and calculations involved. Therefore, achieving a framework that can automatically complete bidirectional design of topological circuits is very significant. Here, we propose an effective bidirectional collaborative design framework with strong task adaptability, which can automatically perceive inputs and generate outputs in arbitrary combinations of text and images. In the framework, a large language model (LLM) is connected to multimodal and different encoders, which involves building a shared multimodal space by bridging alignment in the diffusion process. For simplicity, a series of two-dimensional Su-Schrieffer-Heeger circuits is constructed with different structural parameters. The framework at first is applied to find the relationship between the structural information and topological features. Then the correctness of the results through experimental measurements can be verified by the automatically generated circuit diagram following the manufacture of a printed circuit board. The framework achieves good results in the reverse design of circuit structures and forward prediction of topological edge states, reaching an accuracy of 94%. The key feature of our framework is its ability to effectively learn the bidirectional mapping between circuit structure and topological impedance response across the spectrum. While recently LLMs have made exciting strides, as humans always communicate through various modalities, developing a framework capable of accepting and delivering content in more modalities becomes essential to human-level artificial intelligence. Overall, in this paper, we demonstrate the enormous potential of the proposed bidirectional deep learning fra