Devontae C. Baxter, PhD

Using the TNG-Cluster simulation, we investigate how observational incompleteness affects protocluster identification when the Universe was less than 3 billion years old. We find that surveys targeting only the most massive or actively star-forming galaxies fail to locate the densest protocluster regions in over 50% of cases, often offsetting their true positions by more than 1 proper Mpc. Our results demonstrate how selection incompleteness biases both protocluster detection and inferred trends in galaxy properties with local galaxy density.

We performed a multi-cycle Keck/DEIMOS observational campaign targeting satellite galaxy candidates in galaxy groups when the Universe was roughly 7 billion years old. We established the first spectroscopic measurement of the quiescent satellite fraction down to 10^9.5 M_☉ at this cosmic epoch, constrained the environmental quenching timescale, and demonstrated that the doubling of group quenching timescales over cosmic time is consistent with starvation, or the gradual shutdown of star formation as galaxies exhaust their existing gas after fresh gas supply is cut off.

  We combine observations of 14 galaxy clusters from the GOGREEN and GCLASS surveys with infall histories from the TNG-300 cosmological simulation to constrain the timescales over which infalling satellite galaxies quench upon entering massive clusters at z ≳ 1. We find that for galaxies that enter as star-forming, the quenching timescale varies with stellar mass and matches the time required to exhaust their remaining cold gas after fresh accretion is cut off. These results establish gas “starvation” as the dominant environmental quenching mechanism for massive, star-forming galaxies accreted into clusters when the Universe was half its current age.

Characterizing the environments and star formation activity of faint dwarf galaxies usually requires expensive spectroscopic follow-up of candidates identified in deep photometric surveys. We overcome this barrier by combining a statistical background subtraction technique with a neural-network classifier to separate star-forming and quiescent galaxies using photometry alone. This framework enables the first machine-learning-based measurement of the satellite quenched fraction for low-mass dwarfs beyond the Local Group. Applying this method to SDSS galaxy groups (z < 0.1), we reveal a prominent upturn in the quenched fraction below 10⁹ M_☉. This result is consistent with trends seen in the Local Group and demonstrates that low-mass dwarfs are likely rapidly quenched by highly efficient quenching mechanisms such as ram-pressure stripping.