Research Paper on Atomic spectrum transition probability predictions using machine learning

Description

Title: Atomic spectrum transition probability predictions using machine learning

Abstract: The individual transition probabilities for each transition in the spectrum must be known in order to perform forward modeling of optical spectra with absolute radiometric intensities. In many cases, it becomes nearly impossible to obtain these transition probabilities—also known as Einstein A-coefficients—either theoretically or experimentally. The accuracy of theoretical models will be hampered by complicated electronic orbitals with higher order effects. Due to physical limitations and the enormous amount of measurements needed, experimental measurements can be prohibitively expensive and rarely complete. These restrictions make it impossible to make spectral predictions for many elemental transitions. In this study, we examine the effectiveness of using fully connected neural networks (FCNN), a type of machine learning model, to forecast Einstein A-coefficients using information from the NIST Atomic Spectra Database. The data-driven modeling workflow performs well but can still be less precise than theoretical calculations for simple elements where closed form quantum calculations are feasible. Deep learning was found to be more comparable to theoretical predictions, like Hartree-Fock, for more complex nuclei. The deep learning approach scales well with the number of transitions in a spectrum, unlike experiment or theory, particularly if the transition probabilities are spread across a wide range of values. It can also receive simultaneous training on theoretical and experimental values. Additionally, when training on multiple elements before testing, the model’s performance improves. The machine learning approach’s scalability makes it a potentially effective method for calculating transition probabilities in previously inaccessible areas of the spectral and thermal domains on a noticeably shorter timeline.

Keywords: atomic spectroscopy; deep learning; transition probability; neural network

Paper Quality: SCOPUS / Web of Science Level Research Paper

Subject: Physics

Subject Category: Atomic and Molecular Physics, and Optics

Writer Experience: 20+ Years

Plagiarism Report: Turnitin Plagiarism Report will be less than 10%

Restriction: Only one author may purchase a single paper. The paper will then indicate that it is out of stock.

What will I get after the purchase?

A turnitin plagiarism report of less than 10% in a pdf file and a full research paper in a word document.

In case you have any questions related to this research paper, please feel free to call/ WhatsApp on +919726999915