Star formation characteristics of CNN-identified post-mergers in the Ultraviolet Near Infrared Optical Northern Survey (UNIONS)

Abstract

The importance of the post-merger epoch in galaxy evolution has been well-documented, but post-mergers are notoriously difficult to identify. While the features induced by mergers can sometimes be distinctive, they are frequently missed by visual inspection. In addition, visual classification efforts are highly inefficient because of the inherent rarity of post-mergers (~1 per cent in the low-redshift Universe), and non-parametric statistical merger selection methods do not account for the diversity of post-mergers or the environments in which they appear. To address these issues, we deploy a convolutional neural network (CNN) which has been trained and evaluated on realistic mock observations of simulated galaxies from the IllustrisTNG simulations, to galaxy images from the Canada France Imaging Survey (CFIS), which is part of the Ultraviolet Near Infrared Optical Northern Survey (UNIONS). We present the characteristics of the galaxies with the highest CNN-predicted post-merger certainties, as well as a visually confirmed subset of 699 post-mergers. We find that post-mergers with high CNN merger probabilities (p(x)>0.8) have an average star formation rate that is 0.1 dex higher than a mass- and redshift-matched control sample. The SFR enhancement is even greater in the visually confirmed post-merger sample, a factor of two higher than the control sample.

The ALMaQUEST Survey X: What powers merger induced star formation?

“…In simulations, star formation tends to peak when the interacting galaxies are either close to pericentric passage or at the moment of coalescence (Lotz et al 2008;Scudder et al 2012;Hani et al 2020). Such results are supported by observational evidence, which has consistently demonstrated that global star formation rate values are greatest for interactions with equal mass ratios and small projected separations (Nikolic et al 2004;Lin et al 2007;Ellison et al 2008;Woods et al 2010;Scudder et al 2012;Ellison et al 2013;Bickley et al 2022).…”

Galaxy mergers can rapidly shut down star formation

“…Application of the CNN to the complete CFIS r-band coverage (without our additional cuts in stellar mass or spectral S/N) yields 2000 galaxies with a post-merger probability of at least 0.75 (Bickley et al 2022). However, due to the intrinsic rarity of postmergers, Bayes theorem predicts that even a classifier with accuracy as high as 90 percent will still only be 6 percent pure at the default decision threshold of 0.5 (Bickley et al 2021;Bottrell et al 2022).…”

“…The third challenge can be addressed with the development of machine learning methods, in particular through the application of convolutional neural networks (CNN), to identify galaxy mergers (Ackermann et al 2018;Ferreira et al 2020). Using a bespoke CNN that is tailored to the specifics of CFIS (Bickley et al 2021), with subsequent visual confirmation, we have identified a highly pure sample of 699 post-mergers in the CFIS 2nd data release (Bickley et al 2022). By combining the CFIS post-merger catalog with extant post-starburst classifications available in the literature, in the work presented here we make the first statistical assessment of the frequency of rapid quenching in recently coalesced mergers.…”

SDSS-IV MaNGA: Unveiling Galaxy Interaction by Merger Stages with Machine Learning

“…Recently, machine learning (ML) has been applied to derive various physical parameters of galaxies (e.g., Masters et al 2015;Krakowski et al 2016;D'Isanto & Polsterer 2018;Bonjean et al 2019;Davidzon et al 2019;Hemmati et al 2019;Chang et al 2021) and improves on linear combinations through nonlinear activations (e.g., Ackermann et al 2018;Walmsley et al 2019;Ferreira et al 2020;Bickley et al 2021;Bickley et al 2022;Ferreira et al 2022). In particular, classification by ML (e.g., Banerji et al 2010;Huertas-Company et al 2015;Domínguez Sánchez et al 2018;Bottrell et al 2019;Pearson et al 2019;Barchi et al 2020;Chang et al 2021) can avoid time-consuming visual inspections and will be helpful for the visual classification of galaxy−galaxy interactions from the forthcoming large surveys.…”

“…Ferreira et al (2020) achieve 0.90 accuracy to classify major mergers and measure galaxy mergers in all five CANDELS fields using CNN trained with simulated galaxies from the IllustrisTNG simulation and separate star-forming galaxies from post-mergers in a following work (Ferreira et al 2022). Bickley et al (2022) deployed a CNN and evaluated mock observations of simulated galaxies from the IllustrisTNG simulations to identify post-mergers. Bottrell et al (2022) examine both the morphological and kinematic features of merger remnants from the TNG100 and show that the stellar kinematic data have few contributions.…”