UV artefacts in ultra-slow-roll models of inflation

Background and Motivation

Cosmological inflation involves a canonical scalar field (the inflaton) driving accelerated expansion. Standard single-field slow-roll (SR) dynamics are sufficient to explain the observed nearly scale-invariant power spectrum P zeta(k) on large scales. However, achieving significantly enhanced primordial perturbations—such as those required for the formation of primordial black holes (PBHs) or generating observable scalar-induced gravitational waves—requires a transient non-attractor phase. This departure from the standard SR attractor dynamics is often modeled using specific inflationary models, such as ultra-slow-roll (USR).

Methodology: Hubble-Flow Parametrisation

The paper investigates modeling this transient non-attractor phase using a widely employed Hubble-flow parametrisation. This approach allows for model-independent studies by specifying the time evolution of the second slow-roll parameter, eta(N), rather than deriving the background from a specific V(phi potential.

The authors reconstruct the underlying inflationary potential V(phi from this eta(N) parametrization. They then apply a UV-filtering procedure based on discrete Fourier transform (DFT) to systematically suppress high-frequency modes in field space, comparing this process against applying the same filtering technique to standard analytical potentials.

Core Findings: Spurious UV Artefacts

The investigation reveals a critical physical consistency issue inherent in the the Hubble-flow method. The authors demonstrate that even slow transitions in the slow-roll parameters can hide sharp, localised spikes in higher-order derivatives of the potential at the transition from ultra-slow-roll to slow-roll. These sharp features are typically not found in standard analytical potentials.

The appearance of these spikes is a direct consequence of the the imposed background dynamics: "The reconstructed potentials and their derivatives are represented... We see that in all three cases, V phi phi increases smoothly as the field approaches the first transition into the non-attractor phase, reaching a maximum value at phi about phi in. Although the potentials look very smooth, cases II and III have step-like features in V phi phi near the second transition, i.e. at N out."

Impact of UV Filtering

The application of the UV filter (DFT) allows researchers to assess how these sharp features affect the physical observables. The filter effectively removes these high-frequency components: "The characteristic width of the sharp feature in case II is smaller than in Case III, which corresponds to a spectrum with support further into the UV. The colored lines are produced by applying the UV filter... In Fig 4 we show the effect that the UV filter has on the potential and its derivatives V phi phi and V phi phi phi. Interestingly, at the level of the potential nearly all filters produce very similar results... [but] in order to assess the impact that a progressive removal of UV power has on the background evolution, we solve the background dynamics by using the UV filtered potentials."

A significant consequence of this filtering is related to Wands duality invariance (WDI). The authors find that UV-filtered models typically respect Wands duality invariance as the field evolves back from ultra-slow roll to slow roll.

Conclusion and Implications

The findings lead to a critical questioning of the methodology itself. The paper concludes: Our results thereby question the robustness of simple analytical Hubble-flow parametrisation for modeling inflationary models with a transient non-attractor phase.

Furthermore, the study suggests that these spurious UV effects may extend beyond linear perturbation theory: Beyond linear perturbation theory, the introduction of spurious UV effects might affect other observables, such as non-Gaussianity and loop contributions.

The core scientific literature presented here revolves around advanced quantum field theory in curved spacetime, specifically modeling extreme dynamics during the early universe (Inflation) to predict observable quantities like non-Gaussianity and the abundance of Primordial Black Holes (PBHs).

To improve AI systems using this material, we must move beyond simple data fitting and build Physics-Informed Generative and Inference Engines. The system must not just learn correlations; it must respect fundamental physical constraints (like energy conservation, causality, and quantum field theory renormalization group flow).

Here are the specific improvements and capabilities:

Improvement: Develop a Hamiltonian Flow Generator Network (HFGN) that generates viable inflationary potential functions V(phi). This network will be trained not just on general cosmological data, but specifically on the required functional forms derived from complex theoretical calculations found in the bibliography (e.g., ultra slow-roll dynamics, transient constant-roll segments, inflection points).

What the Improved AI System Can Do:

• Generate Targeted Potentials: The system can generate novel V(phi) models that guarantee specific physical features required for PBH formation (e.g., a sharp dip or a sudden localized increase in the slow-roll parameter epsilon).

• Constraint Enforcement: It automatically enforces theoretical constraints, such as ensuring that the generated potential maintains symmetries (like duality invariance, per [47]) or satisfies specific power spectrum normalization conditions (P zeta about 2 times 10-9).

• Parameterization Efficiency: Instead of requiring thousands of free parameters for a complex potential, the HFGN will parameterize V(phi) using minimal, physically meaningful variables (e.g., the curvature and width of a dip) while ensuring the resulting dynamics remain mathematically stable.

• Inverse Problem Solving with Quantum Corrections: Given observed spectra (e.g., a measured bispectrum or non-Gaussianity level f NL), the ABIE can perform a high-dimensional Bayesian inference to constrain the underlying inflationary potential V(phi). Crucially, it integrates one-loop corrections (per [44], [45], [46]) directly into the likelihood function, allowing it to distinguish between standard slow-roll predictions and those requiring quantum field theory refinements.

• PBH Abundance Prediction: The system can calculate the probability distribution for PBH formation (P PBH) across a range of potential models. It will pinpoint the specific combination of potential features (e.g., depth of the dip, sharpness of transition) that maximizes the predicted PBH abundance while remaining consistent with current CMB data, thereby solving the highly constrained inverse problem posed by [51] and [52].

• Separating Signal Sources: It can decompose observed anisotropies into distinct components (e.g., separating primordial adiabatic perturbations from non-Gaussian contributions or secondary gravitational wave signals), which is vital for identifying the source of inflation's energy budget.

• Modeling Sharp Transitions: The system will accurately simulate the evolution of cosmological perturbations across extremely sharp changes in the slow-roll parameters (the 'sudden transition' scenario). It avoids numerical instability common in standard solvers by implementing specialized methods like those derived from the separate-universe approach [62].

• Non-Gaussianity Mapping: Instead of relying on simplified approximations, the SDSS will compute higher-order correlation functions (bispectrum, trispectrum) directly from the field equations. This allows for mapping how specific inflationary dynamics (like ultra slow-roll or punctuated inflation) translate into measurable non-Gaussian signatures, enabling precise discrimination between competing models.

• Optimizing Computational Flow: By recognizing the underlying symmetries and conservation laws inherent in the physics (e.g., duality invariance), the SDSS automatically reduces the dimensionality of the simulation space, drastically accelerating computation time without sacrificing physical accuracy.

Within single-field inflation, primordial black hole and scalar-induced gravitational wave production from enhanced primordial perturbations typically requires a transient non-attractor phase, such as ultra-slow roll. We investigate the physical consistency of modeling such scenarios through analytical Hubble-flow parametrisation. By reconstructing the underlying scalar field potential, we show that even slow transitions in the slow-roll parameters can hide sharp, localised spikes in higher-order derivatives of the potential at the transition from ultra-slow-roll to slow-roll. These are typically not found in analytic potentials. To evaluate the impact of these structures, we implement a UV-filtering procedure based on discrete Fourier transform to systematically suppress high frequency modes in field space in both classes of models. We find that the filter effectively removes sharp features in Hubble-flow-derived potentials. As a consequence, we show that UV-filtered models typically respect Wands duality invariance as the field evolves back from ultra-slow roll to slow roll. Beyond linear perturbation theory, the introduction of spurious UV effects might affect other observables, such as non-Gaussianity and loop contributions. Our results thereby question the robustness of simple analytical Hubble-flow parametrisation for modeling inflationary models with a transient non-attractor phase.