Autonomous screening of complex phase spaces using Bayesian optimization for SAXS measurements.

Area of Science:

• Materials Science
• Data Science
• Chemical Physics

Background:

• Modern ultrafast X-ray experiments generate vast datasets, exceeding current storage and hardware capacities.
• Efficient data handling and selective storage are critical for scientific progress in X-ray science.
• Small Angle X-ray Scattering (SAXS) experiments produce complex spectra requiring sophisticated analysis.

Purpose of the Study:

• To introduce Bayesian optimization as a method for efficient data handling in X-ray experiments.
• To apply Bayesian optimization for locating global features in SAXS spectra.
• To demonstrate the potential for autonomous experimental operation through data science integration.

Main Methods:

• Bayesian optimization algorithm applied to SAXS data.
• Evaluation of the algorithm on over 250 experimental data points.
• Focus on locating global spectral features in experiments involving supercritical CO2.

Main Results:

• The Bayesian optimization implementation proved versatile, robust, and computationally efficient.
• The algorithm converged rapidly, often within a few iterations, with minimal error.
• Successful identification of global features in SAXS spectra was achieved.

Conclusions:

• Bayesian optimization offers an effective solution for managing large datasets in ultrafast X-ray experiments.
• The method facilitates autonomous experimental operation, enhancing scientific discovery.
• This approach minimizes experimental costs and maximizes the value of generated data.