Expansion schemes for gravitational clustering: computing two-point and three-point functions

TL;DR

This paper explores various perturbative expansion schemes for studying gravitational clustering, focusing on two- and three-point correlation functions, comparing methods, and assessing their effectiveness against simulations and standard theory.

Contribution

It introduces and compares different expansion schemes, including resummations, for gravitational clustering, providing explicit one-loop results and analyzing their advantages and limitations.

Findings

Resummation schemes modify the scaling of diagrams.

Standard perturbation theory is recovered from the path integral.

Two-loop diagrams are likely needed for smaller scales.

Abstract

We describe various expansion schemes that can be used to study gravitational clustering. Obtained from the equations of motion or their path-integral formulation, they provide several perturbative expansions that are organized in different fashion or involve different partial resummations. We focus on the two-point and three-point correlation functions, but these methods also apply to all higher-order correlation and response functions. We present the general formalism, which holds for the gravitational dynamics as well as for similar models, such as the Zeldovich dynamics, that obey similar hydrodynamical equations of motion with a quadratic nonlinearity. We give our explicit analytical results up to one-loop order for the simpler Zeldovich dynamics. For the gravitational dynamics, we compare our one-loop numerical results with numerical simulations. We check that the standard perturbation theory is recovered from the path integral by expanding over Feynman's diagrams. However, the latter expansion is organized in a different fashion and it contains some UV divergences that cancel out as we sum all diagrams of a given order. Resummation schemes modify the scaling of tree and one-loop diagrams, which exhibit the same scaling over the linear power spectrum (contrary to the standard expansion). However, they do not significantly improve over standard perturbation theory for the bispectrum, unless one uses accurate two-point functions (e.g. a fit to the nonlinear power spectrum from simulations). Extending the range of validity to smaller scales, to reach the range described by phenomenological models, seems to require at least two-loop diagrams.