Prediction of Dust Extinction and specific Star Formation Rate of High-z Galaxies with Machine Learning

Abstract

Dust can greatly affect observations, especially at high redshifts. Reddening and extinction are two effects that must be taken into effect, as they can significantly change the galaxy’s spectral energy distribution, which is used to model a host of properties such as star formation rate. In this project, machines are trained to predict the dust extinction and specific star formation rate of high redshift (z ≥ 6.5) galaxies. The models are then applied to the high redshift galaxy candidates in the early data release from James Webb Space Telescope. We find that one band models and three band models perform comparably on simulated data, where generally the percent error of the predictions from the models fall below 15%. When tested on observed JWST data, the models performed much poorer, with percent errors of over 100% of fitted values. Between the one and three band model, the three band models appear to be more robust and are able to predict a wider range of values on observed JWST data.

Date of Award	1 Aug 2023
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Christopher Conselice (Supervisor) & Gary Fuller (Supervisor)