A quantum-gravity-agnostic classification of nonsingular, continuous, globally hyperbolic past extensions of Friedmann-Lemaître-Robertson-Walker (FLRW) spacetimes admits only three regular behaviors: a bounce, an emergent (loitering) universe, and an asymptotically emergent universe; all require violation of the null energy condition near the point where null geodesics defocus. The authors of this classification note that, because general relativity is time-symmetric and the bounce is a stationary point of the expansion where quantum-gravity effects are expected to dominate, one can imagine the two sides of the bounce carrying opposite time directions — two identical expanding universes joined at the bounce — but they give only a qualitative analogy with the Hartle-Hawking no-boundary proposal and the Newtonian gravitational arrow of Barbour, Koslowski, and Mercati. The open problem is to provide, or rigorously rule out, a dynamical, microphysical account of this arrow inversion: a mechanism within a concrete regular bounce model that derives rather than imposes the thermodynamic arrow of time pointing away from the bounce on both sides, together with a demonstration of the dynamical and thermodynamic consistency of two opposite-arrow branches joined at a regular hypersurface. Whether the required low-entropy state at the defocusing region is generic or a disguised Past Hypothesis is a closely related question from the broader arrow-of-time program. The problem is open: existing results are restricted to minisuperspace quantum cosmology, Newtonian N-body systems, or symmetric bounce backgrounds where the arrow still enters through imposed initial conditions, while the semiclassical no-go theorems assume the energy conditions these bounces must violate.

Background

Cosmological solutions of general relativity (GR) are generically singular: Penrose's focusing theorem and its cosmological extensions show that, if gravity focuses null geodesics as encoded in the null energy condition (NEC — the requirement that the stress-energy tensor contracted with any null vector be non-negative) and certain global regularity assumptions hold, incomplete null geodesics and singularities are unavoidable. In FLRW spacetimes — the spatially homogeneous and isotropic cosmologies described by a scale factor a(t)a(t) — this means that extending a decelerating universe backwards in time generically terminates in a past spacelike singularity such as a Big Bang. Carballo-Rubio, Liberati, and Vellucci (2024) extended to cosmology a previously developed quantum-gravity-agnostic program for regular black holes: classify all nonsingular, continuous, globally hyperbolic (admitting a well-posed initial-value formulation, and so free of pathologies such as closed timelike curves) extensions of FLRW metrics into the past by analyzing where the expansions of radial null congruences (families of radial light rays) change sign (the defocusing point, where a congruence that was converging starts to expand). The classification is remarkably restrictive: only three regular past behaviors are possible — a bouncing universe (a(t)a(t) reaches a minimum and re-expands), an emergent universe (a(t)a(t) approaches a constant past value, i.e. a loitering phase), and an asymptotically emergent universe (a(t)a(t) decreases monotonically towards a nonzero constant in the far past). Concrete examples constructed by Easson and Lesnefsky show that all three classes violate the NEC for at least some time, so the region around the defocusing point is necessarily a non-classical regime of modified matter content or quantum gravity. In their conclusions, the authors raise a question about the arrow of time. The thermodynamic arrow of time is the observed global directionality of entropy growth, records, and retarded radiation; its standard explanation in cosmology requires a Past Hypothesis — the posited special low-entropy initial macrostate of the universe — whose own origin is unexplained. Since GR is time-symmetric (flipping the time coordinate on a globally hyperbolic solution yields another valid solution, and at a bounce where a˙=0\dot{a}=0 this flip can keep the metric at least C1C^1, fine-tuned to a¨=0\ddot{a}=0 to keep curvature tensors regular), the usual picture of a universe contracting to a minimum and re-expanding implicitly assumes the direction of time is the same before and after the bounce. Instead, the authors imagine that the two parts of spacetime separated by the defocusing point have opposite time directions — two identical expanding bouncing universes joined at the bounce — so that the arrow of time would invert there. Their supporting arguments are qualitative: near the would-be singularity the Hartle-Hawking no-boundary proposal (the wave function of the universe defined by a Euclidean path integral over compact geometries, in which time becomes imaginary and loses a privileged direction) suggests such an inversion is natural, and the scenario connects to the gravitational arrow of time identified by Barbour, Koslowski, and Mercati (BKM), who showed that in the Newtonian N-body problem with vanishing total energy and angular momentum, generic complete solutions divide at a 'Janus point' with shape complexity growing away from it in both halves, giving each solution 'a single past and two distinct futures'. Related two-arrow constructions exist in relativistic cosmology: Aguirre and Gratton built a nonsingular de Sitter-like bounce joining two inflating asymptotic regions with opposite thermodynamic arrows, and Carroll and Chen proposed a time-symmetric eternally inflating universe on ultra-large scales; Vilenkin showed that, under the null convergence condition, such two-region nonsingular universes need a compact Cauchy surface (a spacelike slice whose initial data determine the entire spacetime) and a bounce size not much exceeding the de Sitter horizon (the horizon scale set by a positive cosmological constant), and still require special boundary conditions at the bounce. Whether two regions with opposite-running arrows can coexist and interact consistently has been debated since Schulman's weakly coupled model systems and Zeh's skepticism. What none of these results provides is what the source paper explicitly lacks: a dynamical, microphysical account — within a concrete regular (NEC-violating) bounce model, ideally with quantum-gravitational input at the defocusing point — of how the arrow of time inverts across the bounce, as opposed to a qualitative possibility argument, an imposed boundary condition, or a Newtonian analogue.

Problem Statement

In the geodesically complete FLRW classification, the bouncing class describes a regular, globally hyperbolic spacetime in which null congruences defocus at a finite past affine distance, the scale factor reaches a nonzero minimum, and the null energy condition is violated for at least some time in the region around the defocusing point. General relativity is time-symmetric, and at the bounce, where a˙(t)=0\dot{a}(t)=0, the time direction can be flipped while keeping the metric components at least C1C^1 (fine-tuned to a¨(t)=0\ddot{a}(t)=0 at the bounce for regular curvature tensors). The source paper asks, only qualitatively, whether the two parts of spacetime separated by the defocusing point can have opposite time directions — two identical expanding bouncing universes joined at the bounce — so that the thermodynamic/cosmological arrow of time inverts there. The open problem is to settle this question with a dynamical, microphysical account. Such an account requires: within a concrete matter-field or quantum-gravity realization of a regular bounce (respecting the regularity, homogeneity, isotropy, and global hyperbolicity requisites of the classification, or stating precisely which are relaxed), deriving the orientation of the thermodynamic arrow of time on both sides of the bounce, including its gravitational and perturbative entropy content, as a consequence of the dynamics rather than of imposed boundary conditions; and establishing that the resulting two-branch structure (opposite time directions joined at a regular hypersurface) is dynamically and thermodynamically consistent, including the mutual isolation or interaction of the branches and the behavior of quantum-gravitational effects at the defocusing point that the source paper invokes as the physical cause of the inversion. Both aspects serve one objective — establishing whether the bounce can host an inversion of the arrow of time on a dynamical foundation. Whether the low-entropy (or low-complexity) state required at or near the defocusing region is typical within the model's solution space or measure, or whether assuming it amounts to a disguised Past Hypothesis as envisaged in the Hartle-Hawking and gravitational-arrow proposals, is a closely related question of the broader arrow-of-time program that a derive-rather-than-impose account naturally engages, but establishing the typicality of the low-entropy state is not required to resolve the problem. The problem is answered either by a constructive account of the inversion satisfying the above, or by a rigorous demonstration, within stated and physically justified assumptions covering the NEC-violating regime, that no such account is possible or that it necessarily requires a fine-tuned Past-Hypothesis-like condition.

References

  1. R. Carballo-Rubio, S. Liberati, V. Vellucci, 'Geodesically complete universes', Phys. Rev. D 110, 044055 (2024), arXiv:2404.13112, https://arxiv.org/abs/2404.13112, DOI 10.1103/PhysRevD.110.044055
  2. D. A. Easson, J. E. Lesnefsky, 'Eternal Universes', Phys. Rev. D 112, 063545 (2025), arXiv:2404.03016 [hep-th], https://arxiv.org/abs/2404.03016, DOI 10.1103/PhysRevD.112.063545
  3. J. B. Hartle, S. W. Hawking, 'Wave function of the Universe', Phys. Rev. D 28, 2960 (1983), DOI 10.1103/PhysRevD.28.2960, https://doi.org/10.1103/PhysRevD.28.2960
  4. J. Hartle, T. Hertog, 'Arrows of Time in the Bouncing Universes of the No-boundary Quantum State', Phys. Rev. D 85, 103524 (2012), arXiv:1104.1733, https://arxiv.org/abs/1104.1733, DOI 10.1103/PhysRevD.85.103524
  5. A. Aguirre, S. Gratton, 'Inflation without a beginning: a null boundary proposal', Phys. Rev. D 67, 083515 (2003), arXiv:gr-qc/0301042, https://arxiv.org/abs/gr-qc/0301042, DOI 10.1103/PhysRevD.67.083515
  6. S. M. Carroll, J. Chen, 'Spontaneous Inflation and the Origin of the Arrow of Time', arXiv:hep-th/0410270 (2004), https://arxiv.org/abs/hep-th/0410270
  7. A. Vilenkin, 'Arrows of time and the beginning of the universe', Phys. Rev. D 88, 043516 (2013), arXiv:1305.3836, https://arxiv.org/abs/1305.3836, DOI 10.1103/PhysRevD.88.043516
  8. J. Barbour, T. Koslowski, F. Mercati, 'Identification of a gravitational arrow of time', Phys. Rev. Lett. 113, 181101 (2014), arXiv:1409.0917, https://arxiv.org/abs/1409.0917, DOI 10.1103/PhysRevLett.113.181101
  9. H. D. Zeh, 'Comment on the "Janus Point" explanation of the arrow of time', arXiv:1601.02790 (2016), https://arxiv.org/abs/1601.02790
  10. J. Barbour, T. Koslowski, F. Mercati, 'Janus Points and Arrows of Time', arXiv:1604.03956 (2016), https://arxiv.org/abs/1604.03956
  11. L. S. Schulman, 'Opposite Thermodynamic Arrows of Time', Phys. Rev. Lett. 83, 5419 (1999), arXiv:cond-mat/9911101, https://arxiv.org/abs/cond-mat/9911101, DOI 10.1103/PhysRevLett.83.5419
  12. D. Linford, 'Big Bounce or Double Bang? A Reply to Craig and Sinclair on the Interpretation of Bounce Cosmologies', Erkenntnis (2020), arXiv:2006.07748, https://arxiv.org/abs/2006.07748, DOI 10.1007/s10670-020-00278-5
  13. J. M. Deutsch, A. Aguirre, 'State-to-State Cosmology: a new view on the cosmological arrow of time and the past hypothesis', Found. Phys. 52, 82 (2022), arXiv:2106.15692, https://arxiv.org/abs/2106.15692, DOI 10.1007/s10701-022-00597-3
  14. M. de Cesare, 'Arrows of time in bouncing cosmologies', Fundam. Theor. Phys. 216, 253 (2024), arXiv:2405.01380, https://arxiv.org/abs/2405.01380, DOI 10.1007/978-3-031-61860-4_12
  15. M. Bojowald, 'A Momentous Arrow of Time', in The Arrows of Time (Springer, 2012) 169-189, arXiv:0910.3200, https://arxiv.org/abs/0910.3200, DOI 10.1007/978-3-642-23259-6_9
  16. K. Song, J. Zhang, 'Constraints on Reversing the Thermodynamic Arrow of Time from Black Hole Thermodynamics, Wormholes, and Time-Symmetric Quantum Mechanics', arXiv:2512.03380 (2025), https://arxiv.org/abs/2512.03380
  17. J. Barbour, F. S. N. Lobo, M. I. R. Lourenço, 'Structural morphology and the gravitational arrow of time', arXiv:2607.27526 (2026), https://arxiv.org/abs/2607.27526
  18. S. Gryb, S. Friederich, 'An account of the arrow of time when scale is surplus', arXiv:2510.03041 (2025), https://arxiv.org/abs/2510.03041
  19. A. Babbar, Y. Sadki, A. Prakash, S. L. Sondhi, 'Classical Fractons: Local chaos, global broken ergodicity and an arrow of time', Phys. Rev. B 111, 245134 (2025), arXiv:2501.12445, https://arxiv.org/abs/2501.12445, DOI 10.1103/g2l1-s2vy
  20. A. Nayeri, 'Inflationary branch decoherence and the cosmological arrow of time', Phys. Rev. D 113, 126030 (2026), arXiv:2602.21263, https://arxiv.org/abs/2602.21263, DOI 10.1103/1k1w-c7vd
  21. T. Banks, 'Singularities, Entropy and the Arrow of Time, or Is CRT a Gauge Symmetry in Quantum Gravity?', arXiv:2607.03199 (2026), https://arxiv.org/abs/2607.03199