Automation of etch pit analyses on solid-state nuclear track detectors with machine learning for laser-driven ion...

Area of Science:

• Nuclear physics
• Materials science
• Data science

Background:

• Solid-state nuclear track detectors (SSNTDs) are crucial for ion detection in laser-driven ion acceleration.
• Manual analysis of SSNTDs is time-consuming due to chemical etching, scanning, and visual identification of millions of ion pits.

Purpose of the Study:

• To enhance the efficiency and automation of SSNTD analysis for laser-driven ion acceleration experiments.
• To develop a machine learning approach for rapid and accurate ion etch pit identification.

Main Methods:

• Utilized machine learning algorithms to distinguish ion etch pits from noise in microscopic images.
• Trained and validated models using data from calibration experiments with known ion energies and from actual laser experiments.
• Processed over 10,000 microscopic images from SSNTDs.

Main Results:

• Achieved highly accurate etch-pit detection in calibration experiments.
• Successfully identified ion etch pits in noisy data from laser-driven acceleration experiments with >95% precision.
• Identified approximately 10^5 ion etch pits, demonstrating the method's effectiveness.

Conclusions:

• Machine learning significantly reduces the time and effort required for SSNTD analysis.
• Automated analysis of millions of images is feasible within a day using current computing power.
• This method offers a precise and efficient solution for ion diagnostics in laser-driven acceleration.