Statistical properties of inflationary saddles in Gaussian random landscapes

Abstract

Random, multifield functions can set generic expectations for landscape-style cosmologies. We consider the inflationary implications of a landscape defined by a Gaussian random function, which is perhaps the simplest such scenario. Many key properties of this landscape, including the distribution of saddles as a function of height in the potential, depend only on its dimensionality, N, and a single parameter, γ, which is set by the power spectrum of the random function. We show that for saddles with a single downhill direction the negative mass term grows smaller relative to the average mass as N increases, a result with potential implications for the η-problem in landscape scenarios. For some power spectra, Planck-scale saddles have η ∼ 1 and eternal, topological inflation would be common in these scenarios. Lower-lying saddles typically have large η, but the fraction of these saddles which would support inflation is computable, allowing us to identify which scenarios can deliver a universe that resembles ours. Finally, by drawing inferences about the relative viability of different multiverse proposals we also illustrate ways in which quantitative analyses of multiverse scenarios are feasible.