Observing flow of He II with unsupervised machine learning

Area of Science:

• Fluid dynamics
• Neutron capture diagnostics
• Machine learning applications

Background:

• Understanding fluid flow around complex objects requires time-dependent velocity field analysis.
• Traditional particle tracking velocimetry (PTV) and particle imaging velocimetry (PIV) are limited in certain experimental setups.

Purpose of the Study:

• To develop a method for observing and analyzing fluid flow using excimer fluorescence.
• To overcome the limitations of PTV and PIV in capturing fluorescence data from neutron capture-induced excimers.

Main Methods:

• Utilized thermal gradients to induce fluid flow within a ~1 cm³ volume.
• Observed fluid flow by recording fluorescence of neutron capture-produced excimers.
• Applied an unsupervised machine learning algorithm to identify clusters of excimers.
• Tracked cluster centroids using a particle displacement determination algorithm adapted from PTV.

Main Results:

• Successfully identified and tracked ensembles of excimers (clusters) from their fluorescence.
• Enabled the analysis of fluid flow dynamics where individual excimer photon detection is improbable.
• Provided a new approach for structure function analysis in fluid dynamics.

Conclusions:

• The developed unsupervised machine learning approach is effective for analyzing fluid flow from excimer fluorescence.
• This method offers a viable alternative to PTV and PIV for specific fluid dynamics studies.
• The technique advances the capability to model and understand fluid flow around complex objects.