At the forefront of a quiet revolution in materials science, researchers at Argonne National Laboratory have developed an artificial intelligence-guided framework that fundamentally changes how X-ray spectroscopy experiments are conducted. The breakthrough, demonstrated at the Advanced Photon Source (APS) — a U.S. Department of Energy (DOE) Office of Science user facility — targets one of the most widely used analytical techniques in the physical sciences and dramatically reduces the time and human judgment required to execute it.
The technique in question is X-ray absorption near-edge structure (XANES) spectroscopy, a powerful method for probing the hidden chemistry of materials central to modern industry and energy technology, including batteries, catalysts, and superconducting materials. Historically, running a XANES experiment required researchers to make dozens or even hundreds of manual decisions about measurement locations and exposure durations across a range of X-ray energy levels — a process as tedious as it was error-prone.
To understand why this matters, consider how XANES functions at a fundamental level. Scientists direct X-ray beams of progressively increasing energy at a target material. When the beam energy reaches a threshold sufficient to dislodge a tightly bound electron from an atom, the material exhibits a sharp spike in X-ray absorption — a phenomenon known as the absorption edge. By monitoring how absorption evolves before, during, and after this edge, researchers can observe in precise detail how a specific element behaves chemically: how a metallic catalyst interacts with surrounding compounds, or how the charge state of a battery's active material shifts during repeated cycling.
The challenge has always been one of efficiency and expertise.
"XANES is incredibly powerful, but until now, scientists had to make dozens or even hundreds of choices about where to measure and how long to measure at each X-ray energy level," said Shelly Kelly, an APS physicist and group leader.Certain energy regions are information-dense and demand intensive sampling, while others contribute negligibly to the final chemical picture. Determining the optimal allocation of measurements across these regions has historically required specialized knowledge and significant trial and error.
The Argonne team's AI-driven solution eliminates this guesswork entirely. An algorithm automatically identifies the absorption edge location, distinguishes chemically rich spectral regions from those offering diminishing returns, and selects only the most analytically valuable measurement points. The outcome is striking: the new approach reduces the number of required measurements by as much as 80%, with no degradation in accuracy. Experimental duration is compressed substantially, enabling researchers to observe rapid chemical transformations as they unfold in real time.
"Our AI method measures only where needed," said Ming Du, a computational scientist and lead author on the research paper. "It's smarter, faster and more efficient, and it lets researchers focus on the big picture."This shift is not merely a matter of speed. By reducing the total X-ray exposure a sample receives, the method also lowers the risk of beam-induced sample damage — a persistent concern in high-intensity synchrotron experiments.
Perhaps the most consequential aspect of this work is the system's capacity for real-time, AI-directed experimental decision-making. By comparing a sample's evolving spectral signature against known reference states — for instance, a fully charged electrode versus a fully discharged one — the algorithm can inform researchers of the current chemical progress, confirm when sufficient data has been gathered, and signal when it is appropriate to conclude the measurement.
"It's not just speeding up the measurement," Kelly noted. "It's making decisions during the experiment — decisions a human used to make."
The implications extend well beyond incremental efficiency gains. This capability points toward a future in which X-ray beamlines operate with a meaningful degree of autonomy, capable of tracking and responding to complex chemical dynamics without continuous human intervention.
"This brings us closer to intelligent X-ray stations that make the most of every photon," said Mathew Cherukara, a computational scientist and group leader at APS. "Argonne plans to continue developing AI-driven tools for next-generation X-ray science, especially as the upgraded APS delivers beams up to 500 times brighter than before."
Cherukara also offered broader context for the work's significance within the current scientific landscape.
"There is a lot of hype around AI today in the media," he acknowledged. "Yet there is no question that AI can help researchers at APS and other light sources make breakthroughs in advanced chemical processes critical to American industry."The Argonne team's results serve as a concrete demonstration that such confidence is grounded in measurable, reproducible outcomes rather than speculation.
The method was validated across multiple beamlines at the APS, specifically beamlines 25-ID-C, 20-BM, and 10-ID. The research was funded by the DOE Office of Science, Office of Basic Energy Sciences, and was published in npj Computational Materials. Contributing authors include Mark Wolfman and Chengjun Sun, alongside Du, Kelly, and Cherukara.
The Advanced Photon Source ranks among the world's most productive X-ray light source facilities, serving a broad research community spanning materials science, chemistry, condensed matter physics, environmental science, and applied engineering. Each year, the facility supports more than 5,000 researchers who collectively produce over 2,000 publications and resolve more vital biological protein structures than users of any other comparable X-ray research installation. As the APS continues its major upgrade program — delivering beams of unprecedented brightness — the integration of AI-driven experimental tools positions the facility to extract even greater scientific value from each research session.
Argonne National Laboratory is managed by UChicago Argonne, LLC for the U.S. Department of Energy's Office of Science, and conducts leading-edge basic and applied research across virtually every scientific discipline in pursuit of solutions to critical national challenges. The U.S. Department of Energy's Office of Science remains the single largest supporter of basic research in the physical sciences in the United States. This research used resources of the Advanced Photon Source, a U.S. DOE Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357.
