Arrows of time and the beginning of the universe

TL;DR

This paper explores two cosmological models with opposite thermodynamic arrows of time, analyzing their boundary conditions, spacetime structures, and implications for the universe's beginning, including conditions under which non-singular, inflating universes can exist.

Contribution

It demonstrates that non-singular universes with two inflating regions require compact Cauchy surfaces and specific boundary conditions, linking two different cosmological scenarios.

Findings

Non-singular universes with two inflating regions must have compact Cauchy surfaces.

The size of the universe at the bounce cannot greatly exceed the de Sitter horizon.

Singular initial data generally lead to inflating regions surrounded by singularities.

Abstract

I examine two cosmological scenarios in which the thermodynamic arrow of time points in opposite directions in the asymptotic past and future. The first scenario, suggested by Aguirre and Gratton, assumes that the two asymptotic regions are separated by a de Sitter-like bounce, with low-entropy boundary conditions imposed at the bounce. Such boundary conditions naturally arise from quantum cosmology with Hartle-Hawking wave function of the universe. The bounce hypersurface breaks de Sitter invariance and represents the beginning of the universe in this model. The second scenario, proposed by Carroll and Chen, assumes some generic initial conditions on an infinite spacelike Cauchy surface. They argue that the resulting spacetime will be non-singular, apart from black holes that could be formed as the initial data is evolved, and will exhibit eternal inflation in both time directions. Here I show, assuming the null convergence condition, that the Cauchy surface in a non-singular (apart from black holes) universe with two asymptotically inflating regions must necessarily be compact. I also argue that the size of the universe at the bounce between the two asymptotic regions cannot much exceed the de Sitter horizon. The spacetime structure is then very similar to that in the Aguirre-Gratton scenario and does require special boundary conditions at the bounce. If cosmological singularities are allowed, then an infinite Cauchy surface with `random' initial data will generally produce inflating regions in both time directions. These regions, however, will be surrounded by singularities and will have singularities in their past or future.