Argonne AI reduces quantum circuit design time for nuclear physics

Researchers at Argonne National Laboratory are developing an artificial intelligence agent to automate the design of quantum circuits for nuclear physics calculations. The system evaluates workflows based on performance metrics such as gate count, circuit depth, and expected accuracy to optimize performance beyond simply achieving running circuits. This AI uses reinforcement learning to explore encoding schemes, algorithms, and hardware configurations to identify efficient quantum solutions for complex nuclear problems. This project directly addresses challenge areas within the DOE Genesis Mission and focuses on the discovery of AI-powered quantum algorithms for nuclear and hadronic systems, with the first phase of development initially targeting nuclear structure and scattering problems. The long-term goal is to accelerate the adoption of quantum computing in nuclear physics, tools that enable solutions to problems that are inaccessible through classical calculations and manual design.

The pursuit of quantum simulation of nuclear systems faces a critical bottleneck: designing efficient quantum workflows. Translating complex physics into executable circuits requires expertise across encoding methods, algorithms, and hardware, but this process is currently performed manually and tends to yield suboptimal results. The Argonne-led team is tackling this challenge using artificial intelligence agents designed to automate the creation of workflows and optimize circuits by learning how to explore the design space. The researchers report that the system learns to design quantum workflows by exploring encoding schemes, algorithm strategies, and hardware configurations. Phase I efforts will focus on nuclear structure and scattering problems and demonstrate AI’s ability to generate competitive circuits. Ultimately, the team envisions a widely applicable tool that will accelerate the adoption of quantum computing in nuclear physics, allowing researchers to address problems that currently cannot be solved with both classical computers and manual quantum circuit design.

This focus marks a shift from simply applying quantum computing to proactively discovering the most effective quantum solutions.