'Towards Precision Astrophysics for Galaxy Formation'

The existence of morphological classes and "fundamental planes" for galaxies implies that their evolution must be deeply regulated by physical laws. However, deciphering the relevant astrophysics and robustly testing our modern paradigm for galaxy formation in a cosmological context requires a new generation of physics-informed, data-driven and AI-accelerated techniques for Bayesian inference and causal discovery. First, I will show that galaxies can be understood as complex dynamical systems whose gaseous atmospheres respond to over-pressurization from supernovae and black holes. This model naturally explains why galaxies are so inefficient at forming stars and why galactic atmospheres undergo a "phase transition" from being predominantly cool and turbulent at early times to globally thermalized at late times -- as implied by observations and predicted by simulations. Then, I will introduce my novel automatically differentiable, GPU-parallelized framework ("sapphire") which enables unprecedentedly exact and rapid sensitivity analyses as well as population-level inference. These reveal that observations of galactic atmospheres may hold the key to breaking numerous modeling degeneracies. I will summarize how my group is leveraging this machinery to pioneer transformative cosmology-style "precision astrophysics" for galaxy formation so that we can definitively answer ~50-year-old questions about how galaxies self-regulate and reliably use them to learn about the fundamental physics of dark matter, dark energy and inflation. Finally, I will briefly describe my JWST discovery of a significant excess of linear, elongated galaxies at high-redshift and their tantalizing connection to the triaxial collapse of early dark matter filaments. This new class of "early blue ellipticals" offers a fresh opportunity for decoding the origin of galaxy complexity with an interdisciplinary mixture of theory, observations and astrostatistics.