Time-reversed stochastic inflation in the quantum well

Abstract: Time-reversed stochastic inflation solves the stochastic evolution of the inflationary universe backward in time, by counting the number of e-folds from the end of quantum diffusion towards some initial state. The point of view of observers attached to the end-of-inflation hypersurface is thus enforced. In this work, we exactly solve time-reversed stochastic inflation in a flat and bounded potential, the so-called quantum well. At given lifetime, the field behaviour is found to be either indistinguishable from the one obtained in a semi-infinite flat potential, or, subject to enhanced stochasticity where any memory of the initial state is erased. The derived distribution of curvature perturbations reduces to the semi-infinite result for small fluctuations while it develops exponential tails for the large ones. Such tails arise for both positive and negative values, and decay twice as fast as the one obtained in the standard forward stochastic inflation. These differences may have important consequences for tail-sensitive phenomena, such as primordial black hole formation.

• The paper introduces an exact time-reversed stochastic framework, deriving the reverse Fokker-Planck solution for inflation dynamics in a flat quantum well.
• It contrasts forward and reverse formulations, showing that reverse-time distributions exhibit exponentially sharper suppression of large curvature perturbations.
• The study reveals that narrow well widths lead to saturated diffusion that erases initial conditions, impacting predictions for tail-sensitive early universe phenomena.

Time-Reversed Stochastic Inflation in the Quantum Well

Overview

The paper "Time-reversed stochastic inflation in the quantum well" (2605.31323) develops a comprehensive analytical framework for stochastic inflation under time-reversed evolution in a flat, bounded potential—commonly referred to as the quantum well. The authors provide exact solutions for the reverse-time stochastic process, analyze the limiting behavior for different well widths, and derive the implications for the distribution of curvature perturbations, with particular attention to the emergence of exponential tails. These aspects are critically compared to the forward-time stochastic formalism, clarifying essential differences for tail-sensitive early Universe phenomena, such as primordial black hole (PBH) formation.

Time-Reversed Stochastic Formalism

Stochastic inflation describes inflaton dynamics as a Langevin process, driven by quantum fluctuations, particularly relevant in flat-potential regions where classical drift is negligible. The conventional ('forward') approach evolves the system from an initial state forward, determining the probability of reaching a particular final field configuration (or number of $e$-folds). Observationally relevant quantities, however, are tied to the end-of-inflation hypersurface, motivating a 'time-reversed' treatment where the statistical ensemble is conditioned on field realizations reaching the exit boundary at a fixed end-of-inflation time, and the backward evolution is studied.

For a field $\phi$ in a flat quantum well with one absorbing boundary (end-of-inflation) and one reflecting boundary, the authors construct the reverse-time Fokker-Planck equation, where the drift term is reconstructed via the forward solution (using Maruyama-Girsanov’s theorem). Importantly, the time-reversed process is conditioned on realizations surviving for a total lifetime $N_0$ (the number of $e$-folds traversed from entrance to exit), and the statistics are constructed by further marginalizing over all possible lifetimes, weighted by their first passage time probabilities.

Exact Solutions in the Quantum Well

The quantum well is specified by a potential constant within $[\phi_r, \phi_a]$ (reflective to absorbing boundary), with field excursions constrained by the well width $\Delta = \phi_a - \phi_r$. The Langevin process reduces to a pure diffusion with a constant coefficient $G = H_0/(2\pi)$. The forward transition probability is found via spectral methods (Fourier/Jacobi theta function expansion), allowing explicit determination of first passage time distributions.

The reverse transition probability is then constructed exactly (Eq. (3.23)), involving dimensionless rescaling in terms of the normalized field variable $\chi = (\phi-\phi_r)/(G\sqrt{N_0})$ and the normalized 'lifetime' variable $N_0$. The solution smoothly interpolates between two regimes:

• Large-width limit ($\Delta \gg G\sqrt{N_0}$): The quantum diffusion is indistinguishable from that of an unbounded (semi-infinite) flat potential. Here, trajectories are not significantly constrained by the reflecting wall, and the probability distribution matches that of the semi-infinite problem.
• Small-width limit ($\phi$0, "saturated diffusion"): The field explores the entire well multiple times within its lifetime, leading to a distribution that is uniform in time and nearly independent of the initial condition. The memory of the starting field value is erased.

Statistics of Curvature Perturbations

A central object is the curvature fluctuation $\phi$1, which, in the time-reversed formalism, is tied to the stochastic local time for a given realization. The authors derive the full probability distribution of $\phi$2 at fixed lifetime, and after marginalization over lifetimes, for arbitrary well width and sink position (the field value at which the trajectory is absorbed).

• Small fluctuations: For $\phi$3, the curvature distribution reduces to the semi-infinite potential result. Specifically, the scaled distribution matches a functional of $\phi$4, independent of explicit diffusion details beyond the overall width and sink location.
• Large fluctuations ("tails"): For $\phi$5, the distribution is dominated by the saturated diffusion regime and acquires symmetric (positive and negative) exponential tails. Crucially, these tails decay as $\phi$6, i.e., twice as fast as the positive tail obtained in the standard (forward) formalism.

This faster decay rate for large fluctuations in the time-reversed regime is robust and echoes analogous results previously found for the semi-infinite potential with drift [Blachier:2025iwk].

Comparison with Forward Formalism

A detailed comparison shows fundamental differences:

• Support and skewness: The forward distribution supports only positive curvature fluctuations (with a bump for negative fluctuations), while the time-reversed distribution is (almost) symmetric.
• Large deviation rate: The decay constant for the far tails in the reverse distribution is exactly double that in the forward one.
• Well width dependence: For $\phi$7 much larger than the Brownian scale, both forward and reverse results match the semi-infinite limit (where the forward formalism becomes singular); for smaller $\phi$8, the reverse process encodes stronger stochasticity, deeply washing out initial conditions' memory.

The distinctions are not merely academic but have practical significance for any tail-sensitive mechanism, e.g., PBH formation via rare large positive $\phi$9 perturbations, as rates are exponentially suppressed in the time-reversed statistics.

Implications and Future Prospects

These results prompt a reconsideration of stochastic inflationary predictions for PBH abundances and for the statistics of the rarest curvature excursions. The time-reversed approach, by enforcing conditioning relevant to physical (end-of-inflation) observers, yields distributions with fundamentally different rare-event structure. Consequently, prior forward-formalism estimates may overpredict tail-driven phenomena (e.g., PBH formation rates) by parametrically significant, exponential factors.

Further implications extend to how quantum diffusion erases initial conditions, notably in saturated quantum well scenarios where the reverse process provides a concrete model for complete loss of pre-existing information about the inflaton position. In inflationary models with more complex potential structure or multiple exits, the time-reversed stochastic approach outlined here provides a blueprint for physically relevant observable statistics.

Extensions would include treatment of potentials admitting a semi-classical limit (e.g., with significant drift), higher-dimensional field spaces, and refinement of the connections to the path-integral/Girsanov framework for more general (non-flat) stochastic dynamics. The authors emphasize the importance of clarifying which observer-centric statistics are most relevant for observable cosmological signatures and for theory-experiment confrontation in stochastic inflationary scenarios.

Conclusion

By solving time-reversed stochastic inflation in the quantum well exactly, this work rigorously demonstrates that observer-relevant statistics for inflationary curvature fluctuations differ both qualitatively and quantitatively from standard forward-in-time stochastic predictions. Exponential suppression of the probability for large curvature excursions is more severe in the reverse framework, directly affecting predictions for primordial structure formation mechanisms that are sensitive to distribution tails. The formalism and exact analytical results here provide robust tools for confronting stochastic inflationary models with cosmological observations and highlight the theoretical necessity of precise statistical conditioning in quantum field cosmology.