Stochastic Modeling of Star Formation Histories. III. Constraints from Physically Motivated Gaussian Processes

Abstract

Galaxy formation and evolution involve a variety of effectively stochastic processes that operate over different timescales. The extended regulator model provides an analytic framework for the resulting variability (or “burstiness”) in galaxy-wide star formation due to these processes. It does this by relating the variability in Fourier space to the effective timescales of stochastic gas inflow, equilibrium, and dynamical processes influencing giant molecular clouds' creation and destruction using the power spectral density (PSD) formalism. We use the connection between the PSD and autocovariance function for general stochastic processes to reformulate this model as an autocovariance function, which we use to model variability in galaxy star formation histories (SFHs) using physically motivated Gaussian processes in log star formation rate (SFR) space. Using stellar population synthesis models, we then explore how changes in model stochasticity can affect spectral signatures across galaxy populations with properties similar to the Milky Way and present-day dwarfs, as well as at higher redshifts. We find that, even at fixed scatter, perturbations to the stochasticity model (changing timescales vs. overall variability) leave unique spectral signatures across both idealized and more realistic galaxy populations. Distributions of spectral features including Hα and UV-based SFR indicators, Hδ and Ca H and K absorption-line strengths, D n (4000), and broadband colors provide testable predictions for galaxy populations from present and upcoming surveys with the Hubble Space Telescope, James Webb Space Telescope, and Nancy Grace Roman Space Telescope. The Gaussian process SFH framework provides a fast, flexible implementation of physical covariance models for the next generation of spectral energy distribution modeling tools. Code to reproduce our results can be found at https://github.com/kartheikiyer/GP-SFH.

Mutual information between galaxy properties and the initial predisposition

“…Another example is the result of [23] from a hydrodynamical zoom-in simulation of three galaxies that the stellar angular JCAP05(2024)111 momenta are regulated largely by the initial conditions rather than by any evolutionary effects. In their work, the angular momenta of protogalaxies at z = 200 were deliberately controlled to show that the initial conditions control the angular momenta, sizes and bulge fractions of the stellar components of the present galaxies, even though the star formation histories of individual galaxies are highly stochastic [57][58][59][60][61]. Our results naturally answer the question of what the initial condition should be.…”

“…When the low dimensional correlation structure from the galaxy properties was first detected by the observational study of [14], it was thought of as an observational challenge to the currently favored CDM paradigm according to which the galaxies evolve through hierarchical merging processes. Not to mention the stochastic nature of star formation histories [57][58][59][60][61], the hierarchical merging histories themselves are expected to be haphazardly differentamong the galaxies, nurturing them to have a variety of dissimilar properties. This conventional expectation based on the CDM paradigm made it very hard to envision the existence of such a simple correlation structure of the key properties of the galaxies with comparable masses.…”

“…Meanwhile, the smallest MI with τ is exhibited by the stellar property, M ⋆ /M t , which may be related to the stochastic nature of galaxy star formation. As shown by [57][58][59][60], the star formation histories are stochastically different among individual galaxies, which are caused not only by unknown physical factors in the stellar evolutions of galaxies but also by the numerical butterfly effect [see 61, for details].…”