Extending the domain of validity of the Lagrangian approximation

TL;DR

This paper analyzes the convergence of Lagrangian Perturbation Theory (LPT) in cosmology, identifying its limitations and proposing a re-expansion method with controlled time steps to improve its predictive accuracy.

Contribution

It provides a formal structure for LPT series, explains convergence limitations, and introduces a re-expansion technique with a time step control strategy to extend LPT validity.

Findings

LPT convergence is limited by the analyticity of the exact solution.

Re-expanding LPT in multiple steps improves its predictive range.

A recipe for time step control enhances LPT re-expansion accuracy.

Abstract

We investigate convergence of Lagrangian Perturbation Theory (LPT) by analyzing the model problem of a spherical homogeneous top-hat in an Einstein-deSitter background cosmology. We derive the formal structure of the LPT series expansion, working to arbitrary order in the initial perturbation amplitude. The factors that regulate LPT convergence are identified by studying the exact, analytic solution expanded according to this formal structure. The key methodology is to complexify the exact solution, demonstrate that it is analytic and apply well-known convergence criteria for power series expansions of analytic functions.This analysis fully explains the previously reported observation that LPT fails to predict the evolution of an underdense, open region beyond a certain time. It also implies the existence of other examples, including overdense, closed regions, for which LPT predictions should also fail. We show that this is indeed the case by numerically computing the LPT expansion in these problematic cases. The formal limitations to the validity of LPT expansion are considerably more complicated than simply the first occurrence of orbit crossings as is often assumed. Evolution to a future time generically requires re-expanding the solution in overlapping domains that ultimately link the initial and final times, each domain subject to its own convergence criterion. We demonstrate that it is possible to handle all the problematic cases by taking multiple steps (LPT re-expansion). We characterize how the leading order numerical error for a solution generated by LPT re- expansion varies with the choice of Lagrangian order and of time step size. Convergence occurs when the Lagrangian order increases and/or the time step size decreases in a simple, well-defined manner. We develop a recipe for time step control for LPT re-expansion based on these results.