Scientists have developed an intelligent method for managing X-ray studies of chemical materials. The new technology allows tracking rapid molecular changes directly during the experiment.
American researchers have created an artificial intelligence-based algorithm that increased the speed of X-ray spectroscopy exactly fivefold. The introduction of the neural network reduced the number of mandatory sample structure measurements by 80 percent without loss of accuracy in the final results. This has enabled specialists to study in detail hidden chemical processes in batteries and catalysts in real time, protecting fragile materials from the damaging effects of radiation.
With the traditional approach, physicists had to manually adjust scanning parameters and randomly select energy points, leading to wasted time and missed intermediate reaction phases. The new adaptive algorithm independently calculates the exact position of the X-ray absorption edge and focuses the instrument on the most informative parts of the spectrum. The system can compare the current readings of the object under study with reference states and fully coordinate the experiment, replacing humans in routine data collection stages.
According to the online publication "World of Innovations News," the proposed method outperforms existing analogs by an order of magnitude in speed, but scientists will need another few years to turn the algorithm into a commonly accepted standard for synchrotron centers. Computational group leader Matthew Cherukara confirmed the real benefits of neural networks for industrially important chemical research. In turn, lead author Ming Du emphasized that the intelligent tool relieves physicists of mechanical work, while research team leader Shelley Kelly stated that the right to make key decisions during experiments has shifted from humans to machines.
