UC Berkeley

Metal-poor massive stars are central to a wide range of astrophysics. They are thought to give rise to exciting transients and explosive enrichment in the early Universe, power cosmic reionization, and to pump large amounts of kinetic energy into the interstellar medium, thereby shaping the evolution of galaxies. They are the most likely progenitors of gravitational wave events and dominate the evolution and spectral energy distribution (SEDs) of star-forming galaxies across cosmic time. Despite their importance, observing and studying populations of metal-poor massive stars is challenging. Luckily, the Local Group (z=0, our cosmic neighborhood) harbors dozens of metal-poor star-forming dwarf galaxies that meet the sub-SMC metallicity threshold and are home to extremely metalpoor massive stars.In this thesis, I will presents the results of HST and JWST photometric and Keck and MMT optical spectroscopic studies in three of these dwarf galaxies; Leo A (Z ∼ 5% Z⊙; D ∼ 800 kpc), Wolf–Lundmark–Melotte (WLM, Z ∼ 14% Z⊙; D ∼ 970 kpc) and Sextans A (Z ∼ 6% Z⊙; D ∼ 1.32 Mpc).In Leo A, the spectroscopic sample of eighteen massive stars comprises a significant fraction of all known metal-poor massive stars with high-quality optical spectra at sub-SMC metallicities. More than half the stars in the sample show emission lines not associated with surrounding H II regions, and half the stars can be considered isolated star candidates (not within ∼ 40 pc of the nearest H II region). I adapt the The Payne, a highly efficient, multi-dimensional interpolator, to train a neural net on the TLUSTY non-local thermodynamic equilibrium (non-LTE) models, creating an infrastructure to fit metal-poor massive stars rapidly. I show that there are early indications that near-ultraviolet/optical/nearinfrared photometry fitted using the BEAST can be used to reliably characterize massive main-sequence star properties relative to low-resolution optical spectroscopy. I present six spectroscopically identified Be stars, representing the first sub-SMC Be stars to have both photometric and spectroscopic analysis. I postulate that Be stars may contaminate parts of the core helium burning branch in the optical CMD.In WLM, I present the lowest metallicity massive contact binary candidate to date. Identified through optical HST time-series photometry, I supplement these light-curves with novel JWST and UV HST photometry, providing light-curve data across six bands. Using machine learning, I emulate the eclipsing binary modeling software (PHysics Of Eclipsing BinariEs; PHOEBE), integrating it within a nested sampling framework to simultaneously fit the multi-band light curves. The best fit model consists of two hot massive stars stars (T1 = 29800+2300 −1700 K, M1 = 16+2 −3 M⊙, and T2 = 18000+5000 −5000 K, M2 = 7 +5 −3 M⊙). The contact system candidate shows a mass-ratio of q = 0.41+0.24 −0.14, which provides an important observational data point, as theoretically predicted mass ratio ranges are sensitive to model assumptions that drive our current binary population models. We expect time series imaging from LSST, BlackGEM, etc., to uncover similar objects in nearby galaxies and aid us in constraining the population of massive metal-poor binaries.In Sextans A, I analyze bright stars (F475W <24) using panchromatic photometry and identify 369 massive star candidates (Mact,p50 > 8 and Loggp50 > 3.7). I compare the derived stellar parameters to existing spectral typing of a subset of the stars, based on existing low-resolution optical spectroscopy. Overall, I find good agreement, with outliers explainable by binarity, binary interaction, or low-quality spectra. I discuss the shortcomings of SED fitting when it comes to capturing binaries and binary products, and showcase the importance of UV photometry in studying massive stars accurately. I identify promising stripped star candidates and find an increasing fraction of OBe stars at low metallicity.