The chemical composition of individual stars provide a fossil record of chemical evolution in a galaxy over cosmic time, encoding important galactic enrichment mechanisms, timescales, and nucleosynthetic pathways. Within the Milky Way, large spectroscopic surveys on modest telescopes have enabled precise chemical abundance measurements for millions of stars and have transformed our understanding of the formation and evolution of the Galaxy. Extending this progress beyond all but the nearest Milky Way satellite galaxies, however, is severely limited by the faintness and crowding of stars at these distances, which require large (6–10+ meter) telescopes and low-resolution spectrographs to achieve sufficient signal/noise (S/N). Recent advances in spectroscopic techniques have dramatically improved our ability to recover accurate and precise chemical abundances from low-resolution spectra. However, these techniques have not yet been extended to the extragalactic regime despite the wealth of data that has been collected in Local Group dwarf galaxies over the past decade. Moreover, with increasingly powerful and multiplexed spectroscopic facilities on the horizon, high-quality (albeit low-resolution) spectra will soon be accessible for millions of stars throughout the Local Group and out to several Mpc. These new datasets have the potential to reveal in unprecedented detail the stellar chemistry of the smallest and faintest galaxies in the Universe, which can in turn be used to investigate the physics of low-mass galaxy evolution. In this dissertation, I lay critical groundwork necessary to fully capitalize on the wealth of chemical information embedded in existing and future extragalactic resolved stellar spectra.

As new spectroscopic facilities are designed, commissioned, and begin amassing large datasets of resolved stellar spectra outside the Milky Way, it is imperative that we understand what chemical abundance information these observations contain. In Chapter 2 of this dissertation, I employ the Cramér-Rao Lower Bound (CRLB) to forecast the theoretical precision to which 41 existing and future spectrographs can measure chemical abundances in metal-poor, low-mass stars in Local Group galaxies. I demonstrate that even at low- to moderate-resolution, blue-optical spectroscopy with modest S/N enables the recovery of a dozen or more elements to a precision of <0.3 dex. Additionally, I find that high-resolution stellar spectra contain substantial chemical abundance information even at low S/N, which can be extracted via full-spectrum fitting techniques. Looking to the future, I show that with reasonable integration times JWST/NIRSpec and 30-m class telescopes can recover ~10 and 30 elements, respectively, throughout the Local Group and bulk metallicities ([Fe/H] and [Alpha/Fe]) for resolved stars out to several Mpc. This analysis is paired with the development and release of an open-source python package, Chem-I-Calc, that facilitates similar forecasts for additional spectrographs, stellar targets, and observing conditions relevant to the astronomical community at large.

In practice, achieving the precision forecasted by the CRLB is impeded by shortcomings of the stellar models used to analyze stellar spectra, which can introduce systematic biases and uncertainties into the measurement of chemical abundances, especially at low resolutions when absorption features are heavily blended. In Chapter 3, I perform a self-consistent analysis of archival Keck/HIRES spectra of low-metallicity stars in M15, which I convolve to lower resolutions in order to quantify the resolution dependence of systematics introduced by model-data mismatches. I demonstrate that systematic biases and uncertainties remain small (~0.1 dex) for 20 (9) elements down to R~10,000 (2500). This analysis illustrates the great promise of low-resolution spectroscopy for stellar chemical abundance measurements in extragalactic systems. As stellar models and spectroscopic fitting techniques improve, the viability of low-resolution resolved stellar spectroscopy will further expand.

Ultra-faint dwarf galaxies (UFDs) represent some of the oldest, lowest mass, most metal-poor, and dark-matter dominated systems, which makes them excellent laboratories to study stellar and galactic physics in the high-redshift Universe and at the faintest end of the galaxy luminosity function. In Chapter 4, I apply a one-zone analytic chemical evolution model within a hierarchical Bayesian framework to CaHK-based photometric metallicity measurements in the reionization-era UFD Eridanus II (Eri II). I present novel constraints on the underlying galaxy formation physics at z>7, finding that Eri II is well characterized by a short (~400 Myr) and inefficient burst of star formation with large supernova-driven gas outflows. The inferred scenario is consistent with the prevailing notion that low-mass galaxies struggle to both convert their gas content into stars and retain their gas reservoirs. Spectroscopic follow-up of stars in Eri II with JWST/NIRSpec will greatly improve the presented constraints. The framework introduced in this chapter can readily be applied to all UFDs throughout the Local Group, providing new insights into the underlying physics governing the evolution of the faintest galaxies in the early Universe.

As a whole, this dissertation seeks to address a range of current and anticipated challenges in the field of extragalactic stellar archaeology, from the robust measurement of low-resolution stellar chemical abundance patterns to the interpretation of low-mass galaxy chemical evolution. Though considerable work remains in preparation for the next decade of extragalactic stellar spectroscopic observations and chemical evolution studies, this dissertation represents a substantial contribution to the field on which future work can build.

This work aims to connect the disparate elements of `doing astronomy’ involved in the production of knowledge, highlighting the advantages of an interdisciplinary approach and emphasizing the importance of each component to the overall success of the field. Broadly dividing the work into ideas, tools, and people, each area will be exemplified to play a critical role in the pursuit of a scientific result. The science case used to illustrate each components’ contribution will be the detection of black holes in our galaxy. First, a method of black hole detection is demonstrated, relying on the theory of relativity and astronomical observations (Chapter 2). Second, a technology that makes such observations possible is described, with a novel variation making improvements on former versions (Chapter 3). Finally, social dynamics that limit the entry and progression (and by extension, the potential scientific contributions in the aforementioned areas) of minoritized groups in the field are analyzed in the context of graduate programs (Chapter 4).

1 - Ideas: Theory and Observations

Though stellar-mass black holes (BHs) are likely abundant in the Milky Way (N=10^8--10^9), only ~20 have been detected to date, all in accreting binary systems. Gravitational microlensing is a proposed technique to search for isolated BHs, which to date have not been detected. Two microlensing events, MACHO-1996-BLG-5 (M96-B5) and MACHO-1998-BLG-6 (M98-B6), initially observed near the lens-source minimum angular separation in 1996 and 1998, respectively, have long Einstein crossing times (>300 days), identifying the lenses as candidate black holes. Twenty years have elapsed since the time of lens-source closest approach for each of these events, indicating that if the lens and source are both luminous, and if their relative proper motion is sufficiently large, the two components should be spatially resolvable. We attempt to eliminate the possibility of a stellar lens for these events by: (1) using Keck near-infrared adaptive optics images to search for a potentially now-resolved, luminous lens; and (2) examining multi-band photometry of the source to search for flux contributions from a potentially unresolved, luminous lens.

We combine detection limits from NIRC2 images with light curve data to eliminate all non-BH lenses for relative lens-source proper motions above 0.81 mas/yr for M96-B5 and 2.48 mas/yr for M98-B6. Further, we use WFPC2 broadband images to eliminate the possibility of stellar lenses at any proper motion. We present the narrow range of non-BH possibilities allowed by our varied analyses. Finally, we suggest future observations that would constrain the remaining parameter space with the methods developed in this work.

2 - Tools: Optics and Instrumentation

`Imaka is a ground layer adaptive optics (GLAO) demonstrator on the University of Hawaii 2.2m telescope with a 24'x18' field-of-view, nearly an order of magnitude larger than previous AO instruments. In 15 nights of observing with natural guide star asterisms ~16' in diameter, we measure median AO-off and AO-on empirical full-widths at half-maximum (FWHM) of 0.95 and 0.64 in R-band, 0.81'' and 0.48'' in I-band, and 0.76'' and 0.44'' at 1 micron. This factor of 1.5-1.7 reduction in the size of the point spread function (PSF) results from correcting both the atmosphere and telescope tracking errors. The AO-on PSF is uniform out to field positions ~5' off-axis, with a typical standard deviation in the FWHM of 0.018''. Images exhibit variation in FWMM by 4.5% across the field, which has been applied as a correction to the aforementioned quantities. The AO-on PSF is also 10x more stable in time compared to the AO-off PSF.

In comparing the delivered image quality to proxy measurements, we find that in both AO-off and AO-on data, delivered image quality is correlated with `imaka's telemetry, with R-band correlation coefficients of 0.68 and 0.70, respectively. At the same wavelength, the data are correlated to DIMM and MASS seeing with coefficients of 0.45 and 0.55. Our results are an essential first step to implementing facility-class, wide-field GLAO on Maunakea telescopes, enabling new opportunities to study extended astronomical sources, such as deep galaxy fields, nearby galaxies or star clusters, at high angular resolution.

3 - People: Society and Culture

Despite decades of interest in promoting diversity and inclusion in the field and higher education more broadly, physics and astronomy departments in American universities remain predominantly white and male. While some STEM fields have made progress in recent years towards correcting the historic overrepresentation of white men, 75% of physics and astronomy PhDs awarded in 2019 went to this demographic that constitutes only 30% of the US population, higher than chemistry, math, computer science, and engineering. This work seeks to understand the barriers to creating a more diverse field by examining how its cultural beliefs can work to maintain inequity in graduate programs.

We use the lived experiences of 12 female and gender-non conforming students of color to construct a model outlining how physics and astronomy graduate programs in American predominantly white institutions maintain equity gaps between majoritized and minoritized students. This model connects participant’s observations of attitudes, policies, and behaviors that relate to two foundational cultural beliefs in Western physics and astronomy: 1) that the field functions as a meritocracy where only the fittest survive, and 2) that physicists comprise an objective, cultureless, and apolitical community, impervious to social influence. We contextualize our findings with literature to demonstrate the negative outcomes such systems create for minoritized students, including alienation from their departments, mental health problems, and lower retention in the field. In outlining this system, we emphasize the design of program structures as a critical point of intervention. To this end, we conclude the work with recommendations based on participant input.

Over the past decade, large photometric surveys, and more recently spectroscopic follow up, have elucidated a clearer picture of galaxy evolution since the epoch of peak star formation at z = 1-3. However, despite these advancements, low mass galaxies, which exhibit much higher number densities than their $Log M_{\star}>9.5$M$\odot$ counterparts, have remained challenging to follow up observationally with current instruments given their faintness and sizes. Despite challenging systematics, gravitational lensing induced magnification and flux amplification is one path towards studying the faint high redshift Universe. Another complimentary approach lies in utilizing the stellar fossil record to turn back the clock on local resolved stars and place constraints on their progenitors. Advances in magnification maps and deep photometry from the Hubble Frontier Fields (HFF) survey, as well as improvement in stellar models and distance measurements, enable us to bridge between the near and far fields to study star formation in low mass galaxies across cosmic time.

In this dissertation, I use rest-frame optical spectroscopy from the MOSFIRE Deep Evolution Field (MOSDEF) survey to investigate the star formation histories (SFHs) of different galaxy types, ranging from actively star-forming to quiescent at $1.4\leq~z\leq2.6$. SFHs are constrained using stellar continuum spectroscopy, specifically through a combination of Balmer absorption lines, the 4000$\,\Angstrom$ break, and the equivalent width of the H$\alpha$ emission line. I use Spectral Energy Distribution (SED) modeling to constrain the average SFHs for five galaxy types including quiescent, green valley, and post-starburst galaxies (transitional galaxies). I find that quiescent and transitional galaxies in the MOSDEF sample are dominated by an SFH with an average star-formation timescale of $\tau \sim 0.1-0.2$~Gyr, in contrast to galaxies in the low-redshift Universe which form their stars over a more extended time period ($\tau>1$~Gyr) on average. Finally, I compare the average properties of the galaxies in the transitional bins to investigate possible paths to quiescence, and speculate on the viability of a dusty post-starburst phase.

I then utilize the the stellar fossil record and stellar population synthesis modeling to reconstruct the evolution of rest-frame ultra-violet (UV) luminosities of the most massive Milky Way dwarf spheroidal satellite galaxy, Fornax, and its five globular clusters (GCs) across redshift. I find that (1) Fornax's (proto-)GCs can generate $10-100$ times more UV flux than the field population, despite comprising $<\sim 5\%$ of the stellar mass at the relevant redshifts; (2) due to their respective surface brightnesses, it is more likely that faint, compact sources in the HFF are GCs hosted by faint galaxies, than faint galaxies themselves. This may significantly complicate the construction of a galaxy UV luminosity function at $z>3$. (3) GC formation can introduce order-of-magnitude errors in abundance matching. I also find that some compact HFF objects are consistent with the reconstructed properties of Fornax's GCs at the same redshifts (e.g., surface brightness, star formation rate), suggesting we may already have detected proto-GCs in the early Universe.

Finally, I present results from a Keck/MOSFIRE survey of 39 low mass galaxies in the HFF. I first discuss the sub-structure around the progenitor of a Milky Way-mass galaxy in the Hubble Frontier Fields (HFF). Specifically, I study an $r_e = 40^{+70}_{-30}$pc, $M_{\star} \sim 10^{8.2} M_{\odot}$ rest-frame UV luminous ``clump'' at a projected distance of $\sim$100~pc from a $M_{\star} \sim 10^{9.8}$M$_{\odot}$ galaxy at $z = 2.36$ with a magnification $\mu = 5.21$. I measure the star formation history of the clump and galaxy by jointly modeling the broadband SED from HFF photometry and H$\alpha$ from MOSFIRE spectroscopy, and explore methodology to constrain whether it formed $in-situ$ or presents a $1:40$ stellar mass ratio accretion event.

I then explore the recent star formation histories and kinematics of the rest of the sample which consists of: $7 \le \log M_{\star} / M_{\odot} \le 10.2$ moderately gravitationally lensed ($\mu<20$), star-forming galaxies at $1.8 \le z \le 3.2$. I find that \lha \ star formation rates (SFRs) exceed UV SFRs by factors of $\sim3-10$, indicative of highly bursty star formation. Galaxy size ($r_e$) is a strong function of decreasing \B\ and \sigha \ (but not \siguv). This suggests that $r_e$ is likely set by the ratio of instantaneous ($<10$Myr) to recent star ($<100$Myr) formation in these low-mass galaxies. I compare integrated velocity dispersions ($\sigma$) with specific SFRs and find a strong correlation at fixed stellar mass. This is consistent with scenarios in which stellar feedback is driving changes to the potential of low-mass galaxies. Finally, I compare the dynamical masses of my galaxy sample with their inferred total baryonic masses using the Kennicutt-Schmidt (KS) relation and find that my inferred baryonic mass is un-physical, even when size underestimates are considered, suggesting that the galaxies in my sample are gas poor compared to local ones.

Binary stars are foundational to modern astrophysics. They underpin precision measurements of stellar structure, age, and composition; they provide stringent tests of general relativity; they make possible the study of faint and rare objects such as black holes and neutron stars; they are the progenitors of gravitational wave events. The components of binaries often interact, dramatically changing their evolution and giving rise to a bewildering zoo of astrophysical phenomenology. Binaries are not rare: about half of all Sun-like stars are binaries, and the binary fraction is even higher for massive stars. To understand stars -- especially massive stars -- it is thus critical to understand binaries. In the last decade, industrial-scale stellar surveys have provided exquisite spectroscopic, photometric, and astrometric data for hundreds of millions of stars in the Milky Way. The goal of this thesis has been to leverage this deluge of data toward new insight into the formation, evolution, and cosmic importance of binaries. On the one hand, binary population demographics represent the end state of the star formation process and thus inform theoretical star formation models. On the other, binary population demographics are an important input for stellar population models: stars in interacting binaries follow qualitatively different evolutionary histories from single stars of similar mass and composition. An improved census of the binary population thus has broad utility across astrophysics. Among other results, this thesis expands the total number of binaries known by more than a factor of 10. The methods developed here are expected to continue to bear fruit in the coming decades as panchromatic surveys look ever deeper into the Galaxy.