Deep-Learning-Empowered Programmable Topolectrical Circuits

Hao Jia, Shanglin Yang, Jiajun He, Shuo Liu, Haoxiang Chen, Ce Shang, Shaojie Ma, Peng Han, Ching Hua Lee, Zhen Gao, Yun Lai, Tie Jun Cui · 2025

Read Deep-Learning-Empowered Programmable Topolectrical Circuits by Hao Jia, Shanglin Yang, Jiajun He, Shuo Liu, Haoxiang Chen, Ce Shang, Shaojie Ma, Peng Han, Ching Hua Lee, Zhen Gao, Yun Lai, Tie Jun Cui (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-24463.txt.

Introduction

Hao Jia leads a team of twelve authors on electric circuits and physics. Topolectrical circuits use wired boards to mimic quantum materials. Such boards let labs test exotic states with cheap, safe parts. Past boards were hard to retune and poor at guessing new states. The team add deep learning and full tuning to one platform.

Each site term and link term on the board tunes on its own. A physics-aware AI design tool proposes states to build and test. The board showed edge states, slow phase shifts, and flat energy bands. It also placed Anderson localization, wave trapping by disorder, at chosen spots. That control beats the random trapping seen in older setups. They even hid full text messages with the trapping patterns.

Catalog record

CREATORHao Jia, Shanglin Yang, Jiajun He, Shuo Liu, Haoxiang Chen, Ce Shang, Shaojie Ma, Peng Han, Ching Hua Lee, Zhen Gao, Yun Lai, Tie Jun Cui
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE41 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-24463

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Deep-Learning-Empowered Programmable Topolectrical Circuits Hao Jia, Shanglin Yang, Jiajun He, Shuo Liu, Haoxiang Chen, Ce Shang, Shaojie Ma, Peng Han, Ching Hua Lee, Zhen Gao, Yun Lai, Tie Jun Cui Abstract Topolectrical circuits provide a versatile platform for exploring and simulating modern physical models. However, existing approaches suffer from incomplete programmability and ineffective feature prediction and control mechanisms, hindering the investigation of physical phenomena on an integrated platform and limiting their translation into practical applications. Here, we present a deep-learning-empowered programmable topolectrical circuits (DLPTCs) platform for physical modeling and analysis. By integrating fully independent, continuous tuning of both on-site and off-site terms of the lattice Hamiltonian, physics-graph-informed inverse state design, and immediate hardware verification, our system bridges the gap between theoretical modeling and practical realization. Through flexible control and adiabatic path engineering, we experimentally observe the boundary states without global symmetry in higher-order topological systems, their adiabatic phase transitions, and the flat-band-like characteristic corresponding to Landau levels in the circuit. Incorporating a physics‑graph‑informed mechanism with a generative AI model for physics exploration, we realize arbitrary, position-controllable on-board Anderson localization, surpassing conventional random localization. Utilizing this unique capability with high‑fidelity hardware implementation, we further demonstrate a compelling cryptographic application: hash-based probabilistic information encryption by leveraging Anderson localization with extensive disorder configurations, enabling secure delivery of full ASCII messages. Main The eternal endeavor of human beings is to search for universal tools that facilitate the comprehension and elucidation of our world through a broad spectrum of disciplines, including mathematical equations, scientific laws, theoretical frameworks, and experimental approaches. With such tools, scientists are empowered to forecast outcomes, test hypotheses, and continually expand our grasp of the universe. Quantum simulators Feynman 2018 , a concept pioneered by Richard Feynman in 1982, propose Hamiltonian simulation as a solution for emulating physical complexity in a programmable manner. Quantum simulators have been realized on several experimental platforms, including systems of ultra-cold quantum gases Cooper et al. 2019 , trapped ions Shao et al. 2024 , photonic systems Bogaerts et al. 2020 ; Ozawa et al. 2019 ; On et al. 2024 ; Dai et al. 2024 ; Koh et al. 2022 ; Koh et al. 2024 , plasmonic systems You et al. 2021 , and superconducting circuits Wu et al. 2021 , which are particularly adept at simulating topological physics with a high degree of reconfigurability. Fig. 1 : The conception of DLPTCs ‘HeTu’. a, The schematic of DLPTC and integration with different components. Physics-graph information is embedded into the deep learning framework for physics phenomenon prediction and generation. The generated parameters are loaded into the driving controller and used to drive the DLPTC. b, By controlling the onsite and offsite terms, this setup facilitates various complex systems, including higher-order topological lattices, flat-band systems for Landau levels, and Anderson localizations, with further potential for exploration beyond these areas. c, The on-site and off-site terms, corresponding to the vortex and edge values of the graph, are fully programmable through voltage control. d, With the DLPTC, we demonstrate an ASCII-based message encryption system for practical application. To date, a significant challenge of programmable Hamiltonian simulators is the observation of multiple complex phenomena, such as topological phase transitions Xiao et al. 2010 ; Qi and Zhang 2011 ; Benalcazar et al. 2017 , Landau levels Rechtsman et al. 2013 ; Yang et al. 2017 ; Barsukova et al. 2024 , and Anderson localization Segev et al. 2013 ; Yang et al. 2024a . So far, achieving precise control over both the on-site potential and the off-site hopping strength across a multitude of samples remains a daunting task. In response to this limitation, we propose fully programmable topolectrical circuits (PTCs) that allow for independent control over the on-site and off-site terms within separate degrees of freedom. By granting full control over the circuit Laplacian, our platform enables the simulation of a diverse range of Hamiltonian systems, thereby facilitating a more straightforward investigation of a wide range of physical phenomena and mechanisms Lee et al. 2018a ; Hadad et al. 2018 ; Wu et al. 2022 . Building on this capability, we integrate an artificial intelligence (AI) module into the circuitry to create deep learning-empowered programmable topolectrical circuits (DLPTCs). This integration not only provides effective feature prediction mechanisms and promotes the investigation of physics, but also bridges the gap between emerging physical phenomena and their practical applications. Recent advances in deep learning have proven invaluable for feature extraction and prediction in complex physical systems Carleo et al. 2019 ; Raissi et al. 2020 ; Karniadakis et al. 2021 . However, traditional neural network architectures, including fully connected neural networks (FCNNs) LeCun et al. 2015 , convolution neural networks (CNNs) Gu et al. 2018 , along with their derivatives He et al. 2016 , often struggle to achieve satisfactory performance due to the inherent high disorder and intricacies of these systems. To overcome this bottleneck, we have proposed physics-graph-informed convolution neural networks (PGI-CNNs) and physics-graph-informed generative model. By embedding the tight-binding information, fundamental physical laws Karniadakis et al. 2021 ; Shang et al. 2022 , and domain knowledge of circuit topology Shang et al. 2024 as intrinsic network priors, these models achieve superior predictive and generative performance in capturing the nuances of disordered systems, particularly in reliably forecasting and generating Anderson localization characteristics. Through flexible control and adiabatic path engineering, we experimentally observe the boundary states without global symmetry in higher-order topological systems, their adiabatic phase transitions, and the flat-band-like characteristic corresponding to Landau levels in the circuit. Incorporating a PGI mechanism with a generative AI model, we realize arbitrary, position-controllable on-board Anderson localization, surpassing conventional random localization. Among them, the PGI-CNN model is utilized for the analysis and understanding of the Anderson system to predict the localization characteristic. A generative PGI-diffusion and PGI-conditional variational autoencoder model has been developed for the generation of arbitrary localization, enabling a unique information encryption framework. We further developed a universal ASCII message encryption system that harnesses the unique capability with high‑fidelity hardware implementation to transmit information between two universities. These achievements underscore the far-reaching potential of programmable Hamiltonian simulators enhanced by deep learning, offering a versatile platform for both exploring physics and enabling innovative applications in interdisciplinary research. Results The prototype of the DLPTCs platform is hereby named ‘HeTu’ (‘River Chart’ in English, which originates from an ancient Chinese civilization legend, meaning a diagram consists of trigrams, points, and connections, and can be used to explore unknown things and reveal the mechanisms). By allowing researchers to dynamically design, simulate, and control the Hamiltonian with greater precision and artificial intelligence, this platform could provide groundbreaking insights into phases of matter and pave the way for the invention of innovative applications. The base hardware structure is a honeycomb-like lattice, characterized by off-site hopping E and on-site potential V strengths in a (N+1)^{2} -site distributed in triangular geometry, as the conception Figure 1 a shows. The sites are designated as A_{ijk} and B_{ijk} for the sublattices \mathrm{A} (\xi=-1) and \mathrm{B} (\xi=1) (shown in red and blue circles), with i,j , and k being the indices in the directions \mathbf{e}_{1}=(-\sqrt{3}/2,-1/2) , \mathbf{e}_{2}=(\sqrt{3}/2,-1/2) , and \mathbf{e}_{3}=(0,1) , following i+j+k+(\xi+1)/2=N , respectively. The couplings of such lattices (hopping) are described by the

Read the complete text