Modelling baryonic feedback for survey cosmology

TL;DR

The paper addresses how baryonic processes, especially AGN feedback, alter the non-linear matter distribution and bias cosmological inferences from weak lensing. It surveys current modelling strategies for $P(k,z)$, including hydrodynamical simulations, halo-model based approaches (hmcode), and the baryonification method, along with data-analysis techniques such as PCA marginalization and emulation. It provides a synthesis of calibration strategies, observational probes (clusters, tSZ/kSZ, cross-correlations), and cross-validation needs, and offers concrete recommendations for the next decade. The work aims to enable robust cosmology with linear and non-linear scales by constraining baryons through multi-probe observations and fast, scalable modelling tools.

Abstract

Observational cosmology in the next decade will rely on probes of the distribution of matter in the redshift range between $0<z<3$ to elucidate the nature of dark matter and dark energy. In this redshift range, galaxy formation is known to have a significant impact on observables such as two-point correlations of galaxy shapes and positions, altering their amplitude and scale dependence beyond the expected statistical uncertainty of upcoming experiments at separations under 10 Mpc. Successful extraction of information in such a regime thus requires, at the very least, unbiased models for the impact of galaxy formation on the matter distribution, and can benefit from complementary observational priors. This work reviews the current state of the art in the modelling of baryons for cosmology, from numerical methods to approximate analytical prescriptions, and makes recommendations for studies in the next decade, including a discussion of potential probe combinations that can help constrain the role of baryons in cosmological studies. We focus, in particular, on the modelling of the matter power spectrum, $P(k,z)$, as a function of scale and redshift, and of the observables derived from this quantity. This work is the result of a workshop held at the University of Oxford in November of 2018.

Figures (4)

• Figure 1: Expected bias in the parameters of the equation of state of dark energy, $w=w_0+w_a(1-a)$ (red and yellow), where $a$ is the scale factor, and the sum of neutrino masses, $\Sigma m_\nu$ (blue), relative to the statistical uncertainty in each parameter that would result from ignoring the impact of baryons on the distribution of matter. The figure assumes the information is directly extracted from the matter power spectrum up to a given maximum wavenumber $k_{\rm max}$ in the redshift range $0<z<3$, and that the impact of baryons is given by the upper limit of the $1\sigma$ scatter of predictions available from different cosmological simulations Chisari18. The dashed line indicates where the bias is equal to the $1\sigma$ statistical uncertainty.
• Figure 2: Expected fractional gain in constraining power of the parameters of the equation of state of dark energy, $w=w_0+w_a(1-a)$ (red and yellow), and the sum of neutrino masses, $\Sigma m_\nu$ (blue), as a function of maximum wavenumber $k_{\rm max}$ included in the analysis. The results were produced by a Fisher forecast in the same set-up as for Figure \ref{['fig:bias']}, where the information is extracted directly from the matter power spectrum. The uncertainty in each parameter is normalized relative to the $k_{\rm max}=0.1\,h\,\mathrm{Mpc}^{-1}$ case. The results indicate there is a significant amount of cosmological information to be gained from going to non-linear scales.
• Figure 3: Fractional impact of baryons on the matter power spectrum at $z=0$ for all the simulations described in Section \ref{['sec:sims']} from which this quantity is available. The curves are collected from Chisari18, Huang19 and Marcel van Daalen (private communication). The small scale upturn is representative of star formation and gas cooling, while the suppression at scale of a few $h$ Mpc$^{-1}$ is due to feedback redistributing gas and dark matter in the simulation.
• Figure 4: Ratio of the power spectra with and without baryon effects from the baryonification model (coloured lines) and from several hydrodynamical simulations (coloured bands) at $z=0$. The baryonification parameters have been tuned to match the gas and stellar fractions of the corresponding simulations. See Ref. Schneider18 for more information about the comparison method.