TY - JOUR TI - Insights on galaxy evolution from studies of the multiphase interstellar medium DO - https://doi.org/doi:10.7282/t3-b1d7-2j50 PY - 2019 AB - Modern galaxy formation and evolution models are able to match the statistical properties of galaxy populations across most of cosmic history. However, detailed observations and sophisticated analytical methods are still needed to test theoretical predictions in extreme scenarios, where the complex interplay of gas accretion, star formation, galaxy interactions, feedback, and other physical processes can have compounded impacts on interstellar medium properties. I discuss results from new far-infrared and submillimeter observations of star-forming galaxies in four massive galaxy clusters at redshifts 0.3 ≲ z ≲ 1.1. Despite being surrounded by hot intracluster plasma, some cluster members are able to retain their cold gas reservoirs, and have high dust-obscured star formation rates. I find that the prevalence of star-forming cluster galaxies increases with increasing redshift. I also present results from integral field spectroscopy, centered on the Paα line, targeting a sample of rare, low-redshift, compact starbursts that strongly resemble z ~ 3 Lyman break galaxies (LBGs). H2 ro-vibrational emission is measured for the first time in either distant LBGs or their low-z analogs. Warm molecular gas and ionized gas properties suggest that star formation feedback is the dominant excitation mechanism in these systems. I also find that, compared to typical nearby star-forming galaxies, LBG analogs are characterized by high velocity dispersions and low ordered-to-disordered velocity ratios. LBG analogs deviate from kinematic scaling relations such as the stellar mass and baryonic Tully-Fisher relations; their anomalously low kinematic support may be attributed to their small physical sizes. Finally, I optimize a deep convolutional neural network (CNN) to predict the gas-phase metallicities of typical low-z star-forming galaxies from three-band optical imaging. The trained CNN is not only able to accurately estimate metallicity, but also can reconstruct the empirical mass-metallicity relation with zero additional scatter. These results imply that multi-color morphological features are important for understanding the connection between galaxies' stellar mass assembly and chemical enrichment histories. KW - Physics and Astronomy KW - Extragalactic astronomy KW - Galaxies -- Evolution LA - English ER -