Chinese scientists applied an AI agent to create an X-ray mirror

Researchers from China have successfully applied an autonomous AI agent to control the process of coating multilayer X-ray mirrors. The system independently developed the technological process and implemented it on a real physical setup, achieving the necessary product parameters.

During the experiment, the agent controlled the parameters of layer deposition, analyzing data in real-time. To achieve the target period of the mirror, the system required only two iterations. In the first cycle, the agent identified deviations in the layer thickness, after which it adjusted the equipment control algorithm, allowing it to obtain a structure with high accuracy corresponding to the design characteristics.

The work demonstrates the possibility of integrating generative models into complex production cycles that require high accuracy. The use of the agent made it possible to exclude manual adjustment of the coating parameters, which usually takes a significant amount of time from qualified personnel. The authors note that the accuracy of control achieved by the neural network is comparable to the results obtained by experienced engineers.

This approach opens the way to the full automation of R&D processes in materials science. The transition from theoretical design to direct control of physical equipment reduces the development cycle of new optical components, allowing optimization of parameters without human involvement in the control loop.