A Machine Learning Approach to Integral Field Unit Spectroscopy Observations: II. HII Region LineRatios

TL;DR

This paper develops a neural network to estimate emission-line ratios in HII regions from spectral data, demonstrating high accuracy, reduced computational cost, and effectiveness in low signal-to-noise conditions, advancing spectral analysis methods.

Contribution

The paper introduces a neural network model for estimating emission-line ratios in HII regions, improving efficiency and accuracy over traditional fitting methods.

Findings

Neural network accurately estimates line ratios with tight constraints.

Model reduces computational costs by two orders of magnitude.

Effective in low signal-to-noise regimes and on real observational data.

Abstract

In the first paper of this series (Rhea et al. 2020), we demonstrated that neural networks can robustly and efficiently estimate kinematic parameters for optical emission-line spectra taken by SITELLE at the Canada-France-Hawaii Telescope. This paper expands upon this notion by developing an artificial neural network to estimate the line ratios of strong emission-lines present in the SN1, SN2, and SN3 filters of SITELLE. We construct a set of 50,000 synthetic spectra using line ratios taken from the Mexican Million Model database replicating Hii regions. Residual analysis of the network on the test set reveals the network's ability to apply tight constraints to the line ratios. We verified the network's efficacy by constructing an activation map, checking the [N ii] doublet fixed ratio, and applying a standard k-fold cross-correlation. Additionally, we apply the network to SITELLE observation of M33; the residuals between the algorithm's estimates and values calculated using standard fitting methods show general agreement. Moreover, the neural network reduces the computational costs by two orders of magnitude. Although standard fitting routines do consistently well depending on the signal-to-noise ratio of the spectral features, the neural network can also excel at predictions in the low signal-to-noise regime within the controlled environment of the training set as well as on observed data when the source spectral properties are well constrained by models. These results reinforce the power of machine learning in spectral analysis.