Oak Ridge National Laboratory scientists developed a deep learning model—a type of artificial intelligence that mimics human brain function—to analyze high-speed videos of plasma plumes during a process called pulsed laser deposition, or PLD.

The PLD technique uses powerful laser pulses to vaporize a target material, creating a cloud-like stream of atoms and particles—the plasma plume—which then settles onto a target surface to form ultrathin films. This method is crucial for creating advanced materials used in electronics and energy technologies.

Materials synthesis platforms that are designed for autonomous experimentation are capable of collecting multimodal diagnostic data that can be utilized for feedback to optimize material properties. Pulsed laser deposition (PLD) is emerging as a viable autonomous synthesis tool, and so the need arises to develop machine learning (ML) techniques that are capable of extracting information from in situ diagnostics. In the study, the researchers demonstrate that intensified-CCD image sequences of the plasma plume generated during PLD can be used for anomaly detection and the prediction of thin film growth kinetics.

The scientists developed multi-output (2 + 1)D convolutional neural network regression models that extract deep features from plume dynamics that not only correlate with the measured chamber pressure and incident laser energy, but more importantly, predict parameters of an auto-catalytic film growth model derived from in situ laser reflectivity experiments.

"We've taught AI to do what expert scientists have always done intuitively—assess the plasma plume to check if the color, shape, size and brightness look the same as they did the last time a good sample was made," said ORNL's Sumner Harris, the lead author of the study. "This not only automates quality control but also reveals unexpected insights into how these microscopic particles behave during film formation.”

This innovation builds on ORNL's previous development of an autonomous PLD system that accelerates materials discovery tenfold, promising to transform materials synthesis monitoring and further streamline the creation of next-generation materials.

Source:

Nature