Machine Learning Assisted Data Labeling for Dipole-Allowed and Dipole-Forbidden Transitions in Ytterbium

In this research project we will use supervised machine learning to classifying dipole-allowed and dipole-forbidden electron transitions in Ytterbium atoms. Data sets for electron-impact excitation often involve cross-sections calculated between large numbers of target states, typically in the region of several hundred to over 1000. These target states might not always possess a spectroscopic label, depending on the software used for the collision calculation. Even if the exact spectroscopic label is unknown, it is often useful to be able to classify these transitions are either dipole-allowed or dipole-forbidden. High accuracy labeling of these transitions, particularly with a computational method, speeds up research and could reduce human error associated with their classification.

To begin, we calculated electron collision cross sections with neutral Ytterbium using the Dirac B-Spline atomic R-matrix code [1,2,3]. These calculations were performed on the Frontera supercomputer at Texas Advanced Computing Center.  We will now use a Neural Network architecture to classify cross sections representing dipole-allowed and forbidden transitions. We are currently acquiring a sufficient amount of quality data in order to train our classification model. Our ultimate goal is to develop a general, atom-independent model that can accurately determine transitions types based on cross-sectional data alone.

 

 

 

[1] O. Zatsarinny and K. Bartschat, Phys. Rev. A 77 (2008) 062701

[2] O. Zatsarinny, https://github.com/zatsaroi

[3] K. R. Hamilton, K. Bartschat and O. Zatsarinny, Atoms 9 (2021) 47