⚛️ high-energy theory

UV artefacts in ultra-slow-roll models of inflation

This paper demonstrates that analytical Hubble-flow parametrisations of ultra-slow-roll inflation often conceal sharp, unphysical UV features in the underlying scalar potential, which can be systematically removed via Fourier filtering to reveal that such models typically respect Wands duality and questioning the robustness of these simple parametrisations for modeling transient non-attractor phases.

Original authors: Gerald Barnert, Laura Iacconi, Hooshyar Assadullahi, Kazuya Koyama, David Wands

Published 2026-06-29

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Gerald Barnert, Laura Iacconi, Hooshyar Assadullahi, Kazuya Koyama, David Wands

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Building a Universe on a Slippery Slope

Imagine the early universe as a ball rolling down a hill. This ball is the "inflaton," and the hill is the "potential energy" of the universe. In standard inflation, the hill is very gentle and smooth, allowing the ball to roll slowly and steadily. This creates a calm, predictable universe.

However, to explain certain things we see today—like tiny black holes formed right after the Big Bang or specific ripples in space-time—scientists sometimes need the ball to do something weird. They need the ball to hit a flat plateau where it slows down almost to a stop, gaining speed in a specific way, before rolling down again. This is called "Ultra-Slow-Roll" (USR).

The Problem: Two Ways to Draw the Hill

The paper investigates two different ways scientists try to model this weird "stop-and-go" behavior:

The "Direct Control" Method (Hubble-flow Parametrisation): Instead of drawing the whole hill first, scientists just draw a graph of how fast the ball should be slowing down at every moment. They say, "At this exact time, the ball must be at a standstill." They then work backward to figure out what the hill must look like to make that happen.
The "Physical Model" Method (Analytical Potentials): Scientists start with a specific, smooth mathematical formula for the hill (like a polynomial or a bump) and see if the ball naturally does the stop-and-go dance on its own.

The Discovery: Hidden Spikes in the "Direct Control" Method

The authors of this paper found a hidden flaw in the "Direct Control" method.

The Analogy of the Smooth Road vs. The Bumpy Road:
Imagine you are a city planner.

Method A (Physical Model): You design a smooth, winding road. You check the math, and it works perfectly. The road is smooth everywhere.
Method B (Direct Control): You tell the engineers, "I want the car to be at exactly 0 mph at mile marker 10, and then instantly accelerate to 60 mph at mile marker 11." You don't care how the road looks, you just want the speed to match your graph.

The engineers (the math) build a road to satisfy your speed requirement. But to make the car go from 0 to 60 instantly to match your graph, they have to build a tiny, invisible, razor-sharp spike in the road right at mile marker 11.

To the naked eye, the road looks smooth. But if you zoom in with a microscope, you see a jagged, dangerous spike that shouldn't be there.

What the Paper Found:
When the scientists used the "Direct Control" method (Hubble-flow) to model the universe's stop-and-go phase, the math forced the creation of these sharp, invisible spikes in the higher-order derivatives of the hill (the "curvature" of the hill).

In the "Physical Model" method, the hill is naturally smooth.
In the "Direct Control" method, the hill looks smooth from a distance, but it is actually covered in microscopic, jagged spikes at the moment the ball transitions from the flat plateau back to the steep slope.

The Experiment: Smoothing Out the Road

To prove these spikes were real and not just a trick of the math, the authors used a "UV Filter."

The Analogy of the Noise-Canceling Headphones:
Imagine the jagged spikes are like high-pitched static noise in a song. The "UV Filter" is like noise-canceling headphones that remove all the high-pitched frequencies, leaving only the smooth, low notes.

They applied this filter to both types of models:

On the Physical Models: The filter barely changed anything because the road was already smooth. The song sounded the same.
On the Direct Control Models: The filter smoothed out the jagged spikes. Suddenly, the "road" changed shape significantly. The ball didn't roll the same way anymore.

The Consequence: Breaking the Rules of the Universe

The paper found that these hidden spikes break a fundamental rule of physics called Wands Duality.

The Analogy of a Mirror:
Think of the universe's behavior as a reflection in a mirror. There is a symmetry where the "slowing down" phase and the "speeding up" phase should look like perfect reflections of each other.

Physical Models: The mirror is clear. The reflection is perfect. The symmetry holds.
Direct Control Models (with spikes): The jagged spikes act like cracks in the mirror. The reflection is distorted. The symmetry is broken.

When the authors smoothed out the spikes (using the filter), the mirror became clear again, and the symmetry was restored. This proved that the "Direct Control" method was creating artificial, unphysical artifacts (the spikes) that distorted the results.

The Conclusion: Don't Trust the Shortcut

The main takeaway is that the "Direct Control" method (Hubble-flow parametrisation) is not as "model-independent" or safe as scientists thought.

While it is a very convenient shortcut to calculate how the universe behaves, it secretly forces the universe to have sharp, unnatural spikes in its geometry to make the math work. These spikes:

Are not found in real physical models.
Break fundamental symmetries (Wands duality).
Change the predicted patterns of gravitational waves and black holes.

In short: If you use the shortcut method to predict the universe's behavior, you might be predicting a universe full of invisible, jagged spikes that don't actually exist. To get the right answer, you need to build the hill from the ground up (using physical potentials) rather than just forcing the speed to match a graph.

Technical Summary: UV Artefacts in Ultra-Slow-Roll Models of Inflation

Problem Statement
The generation of primordial black holes (PBHs) and scalar-induced gravitational waves often requires a transient non-attractor phase in inflation, such as ultra-slow roll (USR), where the curvature perturbation $\zeta$ grows on super-horizon scales. A widely used approach to model these scenarios is the "Hubble-flow parametrisation," where the second slow-roll parameter (or its redefinition $\eta(N)$ ) is specified as a simple analytical function of the number of e-folds, $N$ , rather than being derived from a specific microphysical potential $V(\phi)$ . While this method offers phenomenological flexibility and direct control over the power spectrum amplitude, the authors investigate whether this approach introduces unphysical ultraviolet (UV) artefacts that compromise the physical consistency of the underlying inflationary potential. Specifically, the paper questions whether the smooth $\eta(N)$ profiles typically assumed in literature correspond to smooth potentials or if they hide sharp, localized features in higher-order derivatives of the potential.

Methodology
The authors employ a comparative analysis between two classes of models:

Hubble-flow-derived potentials: They reconstruct the scalar field potential $V(\phi)$ and its derivatives from a specific analytical $\eta(N)$ parametrisation (Eq. 2.1) used in previous literature to model transitions between slow-roll (SR), USR, and final attractor phases.
Analytical potentials: They analyze four explicit microphysical models (non-exponential, exponential, polynomial, and Gaussian bump) known to produce transient non-attractor phases.

To characterize and mitigate potential UV issues, the authors implement a UV-filtering procedure based on the Discrete Fourier Transform (DFT).

They discretize the field space $\phi$ and compute the power spectrum of the third derivative of the potential, $V_{\phi\phi\phi}$ .
They apply a Gaussian window function in Fourier space to suppress high-frequency modes (UV power) in field space.
They systematically apply this filter to both the reconstructed Hubble-flow potentials and the explicit analytical potentials to observe the impact on background dynamics and linear perturbations.

Key observables analyzed include:

The shape of $V(\phi)$ and its derivatives ( $V_{\phi\phi}$ , $V_{\phi\phi\phi}$ ).
The evolution of slow-roll parameters $\eta(N)$ and $\epsilon_1(N)$ .
The Mukhanov–Sasaki mass term $\mu^2$ and its time derivative, specifically testing for Wands duality invariance (WDI). WDI holds if the perturbation evolution is invariant under homogeneous background transformations between the non-attractor and attractor eras, a condition linked to the behavior of $V_{\phi\phi\phi}$ during the transition.
The primordial scalar power spectrum $P_\zeta(k)$ .

Key Contributions and Results

Hidden UV Artefacts in Reconstructed Potentials:
The reconstruction of potentials from smooth, analytical $\eta(N)$ functions reveals that even slow transitions in $\eta(N)$ hide sharp, localized spikes in the third derivative of the potential, $V_{\phi\phi\phi}$ , specifically at the transition from the USR phase back to the final attractor phase.
- In the "Case II" (repulsive) and "Case III" (attractive) models, $V_{\phi\phi\phi}$ exhibits distinct spikes.
- In "Case I" (Wands duality), the spike is replaced by a step, but the derivative remains significantly different from smooth analytical potentials.
- These features are absent in typical analytical potentials (e.g., inflection points or Gaussian bumps), which display smooth $V_{\phi\phi\phi}$ and red-tilted UV power spectra.
Impact of UV Filtering on Background Dynamics:
Applying the UV filter to suppress high-frequency modes in $V_{\phi\phi\phi}$ effectively removes the sharp spikes in the reconstructed potentials.
- This smoothing significantly alters the background evolution: the duration of the non-attractor phase is reduced, the peak value of $\eta(N)$ decreases, and inflation often ends earlier.
- This demonstrates that the sustained non-attractor dynamics in Hubble-flow models rely on a delicate fine-tuning of the potential's higher-order derivatives, which is disrupted by the removal of UV features.
Wands Duality Invariance (WDI):
- Hubble-flow models: In Cases II and III, the sharp spikes in $V_{\phi\phi\phi}$ cause the Mukhanov–Sasaki mass to vary during the second transition, breaking WDI. This results in a power spectrum with a complex peak structure featuring two consecutive breaks in power-law behavior.
- Effect of Filtering: When UV power is suppressed, the spikes in $V_{\phi\phi\phi}$ are smoothed, and WDI is approximately restored for Cases II and III. Consequently, the secondary break in the power spectrum disappears, leaving a simpler broken-power-law structure.
- Analytical models: Explicit microphysical models naturally satisfy WDI (or approximate it closely) because their potentials are smooth. Their power spectra typically show only one characteristic scale (one break), and this structure remains stable under UV filtering unless the filter is so aggressive that it flattens the potential entirely, ending inflation prematurely.
Sensitivity of Observables:
The paper finds that the amplitude and position of the peak in $P_\zeta(k)$ are highly sensitive to the UV filtering for both Hubble-flow and analytical models. This sensitivity reflects the extreme fine-tuning required to maintain a sufficiently long non-attractor phase for PBH formation, regardless of the modeling approach.

Significance and Claims
The authors conclude that the Hubble-flow parametrisation, while computationally convenient, is not a model-independent procedure. Specifying a simple analytical form for $\eta(N)$ implicitly enforces an underlying potential with specific, atypical, and sharp localized features in its higher-order derivatives. These features are "spurious UV artefacts" that do not appear in standard microphysical models.

The presence of these artefacts has significant consequences:

They can break symmetries like Wands duality invariance, leading to predictions for the power spectrum (e.g., double breaks) that may not be robust.
They likely affect higher-order observables, such as non-Gaussianity and loop contributions, which depend on higher derivatives of the potential.

Therefore, the paper argues that simple analytical Hubble-flow parametrisations may not be robust tools for modeling inflation with transient non-attractor phases, as their predictions are fundamentally tied to the specific, often unphysical, UV structure of the reconstructed potentials.