An interactive machine learning platform for analyzing multi-particle coincidence data from cold target recoil ion...

Area of Science:

• Atomic and molecular physics
• Computational physics
• Data science

Background:

• Modern momentum spectroscopy experiments generate high-dimensional multi-particle coincidence data.
• Analyzing this complex data requires advanced computational tools beyond traditional methods.

Purpose of the Study:

• To introduce SCULPT, a novel software platform for analyzing COLTRIMS data.
• To integrate machine learning with physics-informed analysis for uncovering latent patterns in complex datasets.
• To enhance the efficiency and scope of multi-dimensional data analysis in atomic and molecular physics.

Main Methods:

• Implementation of Uniform Manifold Approximation and Projection (UMAP) for non-linear dimensionality reduction.
• Development of an adaptive confidence scoring system for quantitative reliability assessment of clustering.
• Integration of interactive visualization, data filtering, and configurable molecular profiles within a web-based environment.

Main Results:

• SCULPT successfully analyzes photo-double-ionization data from D2O molecule dissociation using the COLTRIMS method.
• Distinct fragmentation channels and their correlations with physics parameters were revealed.
• The platform demonstrated its capability to identify rare events with non-linear correlations.

Conclusions:

• SCULPT significantly reduces the time required for complex multi-dimensional analyses in momentum spectroscopy.
• Its modular and web-based architecture makes it accessible to the broader physics community.
• The software facilitates on-the-fly discovery of rare events, improving experimental exploration and efficiency.