Related Experiment Video
Updated: Aug 5, 2026

Achieving Efficient Fragment Screening at XChem Facility at Diamond Light Source
Published on: May 29, 2021
Virtual commissioning of the Korea Light Source Hard X-ray Nanoprobe beamline with multilingual natural language
Minho Seo1, Hyeong Joo Choi1, Dasuel Ham1
14GSR Research Division, Pohang Accelerator Laboratory, POSTECH, Pohang, Gyeongbuk 37673, Republic of Korea.
Abstract:
Commissioning of a synchrotron hard X-ray nanoprobe beamline traditionally requires months of iterative alignment after hardware installation, during which operational knowledge accumulates but remains inaccessible to non-specialist users. To address this, we present a browser-based virtual commissioning platform for the Korea Light Source ID10 Hard X-ray Nanoprobe beamline (first light 2029) that allows beamline scientists to design, test and refine alignment procedures, scan plans and experimental workflows years before the first photon arrives. Specifically, the platform integrates a Monte Carlo ray-tracing engine, a standard Experimental Physics and Industrial Control System (EPICS)/Bluesky control stack, and a multilingual natural-language interface within a single deployable package, which we name HANBIT (Hybrid Agent-driven Natural-language Beamline Interactive Toolkit). Users can interactively explore parameter trade-offs, such as the effect of the secondary source aperture on beam size versus photon flux. The Monte Carlo engine reproduces the overall Shadow4 beam-profile shape and is validated against SPECTRA undulator spectra, source size and divergence. The natural language processing (NLP) agent achieves 98.2% automated action-identification accuracy across 228 test cases in Korean, English and Japanese, whereas expert review of the same responses yields an acceptance rate of 67.3%. We identify this 30.9 percentage-point gap as a central finding: automated accuracy does not guarantee operational acceptability, and closing it is the key challenge for deployment-grade natural-language beamline control. We further validate the zero-change hardware transition strategy that the beamline pursues on three real hardware subsystems, confirming that at the validated device layers the control code and scan plans operate unchanged on the real devices; the integration of the remaining parts, such as high-rate area detectors, which awaits the detector hardware, is also discussed.

