Deep-learning-empowered programmable topolectrical circuits

Area of Science:

• Condensed Matter Physics
• Quantum Computing
• Materials Science

Background:

• Existing topolectrical circuits lack full programmability and inverse design capabilities.
• Bridging theoretical modeling with practical realization remains a challenge.

Purpose of the Study:

• To develop a deep-learning-empowered programmable topolectrical circuit platform.
• To enable flexible physical modeling, inverse state design, and hardware verification.
• To explore advanced physical phenomena and novel applications.

Main Methods:

• Implementing a system with continuously tunable on-site and off-site Hamiltonian terms.
• Utilizing physics-graph-informed generative models for inverse design.
• Employing flexible control and adiabatic path engineering for experimental observation.

Main Results:

• Experimental observation of boundary states in higher-order topological systems without global symmetry.
• Demonstration of adiabatic phase transitions and flat-band characteristics (Landau levels).
• Achieving position-controllable Anderson localization and demonstrating probabilistic information encryption and product anti-counterfeiting.

Conclusions:

• The developed platform establishes a new paradigm for on-demand inverse design by synergizing deep learning and programmable hardware.
• This approach bridges fundamental physics with information technologies.
• The system offers versatile capabilities for exploring complex physical models and developing innovative applications.