Researchers at Argonne National Laboratory, as reported by The Register, have connected an agentic AI platform to the 26-ID hard X-ray nanoprobe beamline, allowing scientists to direct nanoscale imaging with natural-language requests. In an experimental run, the system was asked to locate the interface between two regions on an integrated circuit and did so without manual beam aiming. The work is part of SYNAPS-I, an Argonne collaboration with Lawrence Berkeley National Laboratory and other DOE labs under the Department of Energy's Genesis Mission.
The beamline relies on ptychography, a technique that collects multiple diffraction patterns as X-rays interact with a sample. Modern experiments can generate millions of images in a few hours, and conventional reconstruction and analysis can take days. Argonne reports that its AI system reconstructs images in real time as data arrives. Phase one of SYNAPS-I tested AI-based ptychographic processing; phase two added the conversational agentic layer.
The agentic system goes beyond faster processing. It can analyze results, flag additional areas of interest, and pursue those targets on its own. Argonne scientists say this could open the beamline to researchers who lack specialized operating knowledge. Computational scientist Mathew Cherukara called the real achievement a next-generation, AI-enhanced microscope for many scientific fields, and Tao Zhou said the chat interface could be a game changer for attracting new users.