My research focuses on understanding galaxy evolution, particularly the impact of metal production by stars, cosmological gas inflow, and galactic winds on chemical abundances. I am interested in both the gas-phase abundance that traces the relatively recent enrichment history of a galaxy and the stellar abundances that provide a fossil record of the abundance of a galaxy over its entire history.
Until recently, large spectroscopic surveys of galaxies only took single spectra of galaxies to measure their global properties. However, the chemical signatures of galaxy evolution are present in the resolved gas-phase and stellar structures of galaxies. The SDSS-IV MaNGA survey is using integral field spectroscopy to measure the gas-phase and stellar properties of a statistically meaningful sample of nearby galaxies at ~kpc resolution to understand the physics driving galaxy evolution. Unlocking the full statistical potential of MaNGA's massive inter-connected and multi-dimensional data set will require state of the art techniques and software tools.
I am using deep convolutional neural networks to perform non-linear dimensionality reduction on the MaNGA cubes and spectra. We are using this lower dimensional embedding to detect anomalies within galaxies that were not found by the standard analysis pipeline.
I am a core developer of Marvin, an ecosystem for facilitating exploration, access, visualization, and analysis of the MaNGA data. The three components of Marvin are a web app, a Python package, and an API. The web app serves as a portal for querying, exploring, and visualizing the MaNGA data. For quantitative analyses, the Python package provides robust tools to extract publication-quality results. The multi-tiered structure of the MaNGA data, where the 2-D maps of galaxy properties summarize the full 3-D spectral data, means that data is often best accessed on-the-fly via the API on an as-needed basis.
Stellar abundance patterns provide vital clues to the formation and enrichment history of the Milky Way and its stellar populations. The Apache Point Observatory Galactic Evolution Experiment (APOGEE), a sub-survey of the Sloan Digital Sky Survey III (SDSS-III), will measure the abundances of approximately 16 elements in about 100,000 stars across the Milky Way. Both the high dimensionality and vast quantity of stars in this data set represent a significant step forward relative to previous studies. To more efficiently utilize the wealth of information in the APOGEE data, we will apply principal component abundance analysis (PCAA), a principal component decomposition of stellar abundances relative to iron.
In anticipation of the APOGEE data set, we performed PCAA on a small but interesting sample of microlensed bulge dwarf stars to study the formation hisotry of the bulge. The first principal component has large contributions from oxygen and other alpha elements, which reflects the interplay between enrichment from Type II and Type Ia supernovae. The second principal component is dominated by a sodium-nickel correlation. The nucleosynthetic yields of these elements from Type II supernovae are sensitive to metallicity, so the second principal component suggests a rapid timescale of formation for the bulge. Furthermore, the microlensed bulge dwarfs follow a bimodal distribution in the first principal component, which matches the bimodality seen in [Fe/H] for these stars.
Currently, we are using PCAA on the results of a simple chemical evolution model, and we will apply this technique to a more sophisticated chemical evolution model to make detailed comparisons with the APOGEE survey.
with David Weinberg and Jennifer Johnson
I'm a member of the SDSS-III collaboration and the APOGEE survey. My
main contribution to the survey was writing the target selection
algorithm for the halo fields containing globular clusters. These
fields present unique challenges for target selection due to the
extremely high density of targets in the globular clusters and the
paucity of targets in the rest of the field.
The relation between galaxy stellar mass and gas-phase metallicity is a sensitive diagnostic of the main processes that drive galaxy evolution, namely inflow of relatively pristine gas from the intergalactic medium, metal production in stars, and metal ejection via galactic winds. Previous mass-metallicity relations based on strong line metallicity measurements suffer from large systematic uncertainties that impact the physical interpretation of the mass-metallicity relation. We stacked the spectra of about 200,000 star-forming galaxies from the SDSS to significantly enhance the signal-to-noise ratio of the weak [O III] 4363 and [O II] 7320, 7330 auroral lines required to measure metallicity with the more relibable direct method. The direct method mass-metallicity relation spans a wide range in stellar mass and extends an order of magnitude lower in stellar mass than strong line mass-metallicity relations. We investigated how deviations from the mean direct method mass-metallicity relation correlate with SFR and found a factor of two to three times stronger correlation with SFR than strong line mass-metallicity relations, which implies that galaxies require a longer timescale to re-enrich themselves following the accretion of pristine gas. The steep low mass slope and comparatively low normalization of the direct method mass-metallicity relation indicate that galactic winds efficiently eject metals from galaxies, particularly low mass galaxies.
with Paul Martini
Figure: This figure compares the mass-metallicity relations based on direct method metallicities from Andrews & Martini (2013) (black points and line) and based on metallicities estimated with less reliable strong line calibrations from the literature (colored lines). The mass-metallicity relation serves as a crucial constraint for galaxy evolution models, particularly for the amount of material that is ejected from galaxies via galactic winds.
Radiation pressure from the absorption and scattering of starlight by dust grains may be an important feedback mechanism in regulating star-forming galaxies. We compared the luminosity and flux of star clusters, star-forming subregions, normal star-forming galaxies, and starbursts to their dust Eddington limit to evaluate the importance of radiation pressure as a feedback mechanism. In particular, the linear LIR–L'HCN correlation provides evidence that galaxies may be regulated by radiation pressure feedback. We show that star-forming galaxies approach but do not dramatically exceed the Eddington limit, but many systems are significantly below the Eddington limit, perhaps due to the “intermittency” of star formation---the tendency for only a small number of subregions within a galaxy to be actively forming stars at any moment because of the time dependence of the feedback process and the luminosity evolution of the stellar population.
with Todd Thompson
Figure: The data points are literature measurements of galaxy IR luminosity, a proxy for star formation rate, and the HCN line luminosity, a tracer of dense gas mass. Galaxies approach the dust Eddington limit (gray shaded region bordered by the solid red lines; Andrews & Thompson 2011), which suggests that radiation pressure on dust could be the dominant feedback mechanism regulating star formation in these systems, especially in galaxies with high dust-to-gas ratios (dashed red line).