Materials Science • CHESS
Synchrotron experiments: enabling autonomous AI-guided beamlines
AI-guided synchrotron experiments adapt measurements as data are collected, allowing researchers to focus on the most informative regions of a sample instead of following a fixed scan. At the CHESS synchrotron, NSDF connects beamline measurements, AI services, persistent storage, and live visualization into a unified workflow that enables scientists to monitor, steer, and analyze experiments in near real time while preserving a complete record of the experiment.
The Scientific Challenge
Modern synchrotron experiments generate measurements continuously, but selecting the next location to sample requires coordinating instruments, data reduction, AI models, visualization, and storage across distributed systems. This coordination must occur reliably within a user facility while preserving existing beamline operations and supporting long-running experimental campaigns.
How NSDF Helps
NSDF provides the digital backbone that connects beamline measurements with AI-driven decision making through an event-based workflow. It captures reduced measurements, coordinates communication between distributed services, stores the evolving experiment state, and provides live dashboards for monitoring progress. The platform also preserves experiment history, allowing researchers to review, replay, and analyze experiments long after beamtime ends.
Scientific Impact
The integrated workflow supports sustained autonomous experimentation under real beamline conditions. During the CHESS campaign, the system completed more than 12 hours of continuous autonomous operation and collected over 500 measurements, demonstrating that AI-guided experimental workflows can reliably support large-scale synchrotron experiments while providing scientists with live insight into the evolving experiment.
Key Outcomes
- Autonomous Experimentation — AI continuously selects the next measurement location during beamtime.
- Live Visualization — Scientists monitor measurements, predictions, and uncertainty through interactive dashboards.
- Persistent Experiment Record — Every measurement and AI recommendation is stored for replay and future analysis.
- Operational at Scale — The workflow runs autonomously for more than 12 hours, collecting 500+ measurements during a real CHESS beamtime campaign.