⚛️ high-energy theory

Deviations from Gaussian White Noise in Stochastic Inflation

This paper investigates how relaxing standard assumptions in stochastic inflation—specifically regarding the background metric, window function, and initial state—modifies the driving noise, demonstrating that while background and window deviations primarily affect amplitude or introduce color, deviations from the Bunch-Davies vacuum uniquely generate non-stationary, colored, and non-Gaussian noise.

Original authors: Zahra Ahmadi, Mahdiyar Noorbala

Published 2026-07-24

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Zahra Ahmadi, Mahdiyar Noorbala

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

Imagine the universe as a giant, expanding balloon. In the very first split second of its existence, this balloon didn't just grow; it inflated faster than the speed of light, stretching tiny, quantum jitters into massive, cosmic structures. This is the theory of Inflation. To understand how this happened, scientists use a mathematical tool called Stochastic Inflation. Think of this tool as a way to simplify a chaotic, quantum mess into a manageable story. It treats the universe's expansion like a hiker walking up a hill, where the "wind" pushing them is a random force called noise.

For decades, scientists have assumed this "wind" is perfectly simple: it's Gaussian White Noise. In everyday terms, imagine a static hiss on an old radio. It's random, it has no memory of what happened a second ago, and every frequency sounds the same. This assumption makes the math easy and gives us a clear picture of how galaxies formed. But what if the universe isn't that simple? What if the "wind" has a rhythm, a memory, or a weird color? This is the question Zahra Ahmadi and Mahdiyar Noorbala set out to answer. They wanted to see what happens to our cosmic story if we tweak the rules of the game—specifically, if the universe wasn't perfectly smooth, if the way we measure the "wind" wasn't a sharp cut, or if the universe didn't start in its most peaceful, quiet state.

The Cosmic Static: When the Rules Change

In the standard story of inflation, the universe is like a perfectly smooth, expanding stage (called de Sitter space), and the quantum fields dancing on it start in their most relaxed state, known as the Bunch-Davies vacuum. When we use a "sharp cutoff" to separate the tiny, fast-moving quantum bits from the big, slow-moving ones, the resulting noise is that perfect, memoryless white hiss. It's clean, predictable, and mathematically friendly.

However, Ahmadi and Noorbala decided to poke holes in this perfect picture. They asked: "What if we relax one of these rules?" They tested three specific scenarios to see how the "noise" changes, treating the universe like a toy model where they could swap out the ingredients.

1. The Wobbly Hill (Changing the Background)
First, they asked what happens if the universe isn't a perfectly smooth, static hill but a slightly bumpy, changing one (a quasi-de Sitter space). In this scenario, the "wind" doesn't change its character; it's still white noise. However, the strength of the wind varies over time. Imagine a radio that is still full of static hiss, but the volume knob is being turned up and down as you listen. The noise is still "white" (no memory, all frequencies equal), but it's non-stationary, meaning its intensity depends on when you listen. The authors found that even with a changing universe, the noise remains Gaussian and white, just with a time-dependent amplitude.

2. The Blurry Lens (Changing the Window Function)
Next, they looked at how we separate the "short" waves from the "long" waves. The standard method uses a sharp cutoff, like a knife slicing perfectly through a cake. But what if we use a smooth window, like a blurry lens that gradually fades out the short waves instead of chopping them off?
Here, the story gets interesting. The authors found that this "blur" introduces memory into the system. The noise is no longer white; it becomes colored. Imagine the radio static suddenly developing a rhythm or a specific tone. The noise at one moment is now connected to the noise a moment ago. The universe "remembers" its past few steps. They calculated that this colored noise has a specific "color profile" (a power spectrum) that depends on how wide that blurry lens is. If the cutoff is perfectly sharp, the memory vanishes, and the noise is white again. But any fuzziness brings the color back.

3. The Excited Start (Changing the Initial State)
Finally, they asked: "What if the universe didn't start in its calm, quiet vacuum state?" What if it started "excited," with some extra particles already dancing around?
This is where things get truly wild. The authors discovered that if the universe starts in a state that isn't the standard vacuum, two major things happen:

The noise loses its Gaussian nature: The randomness becomes "lumpy" and unpredictable in a way that standard statistics can't easily describe.
The noise becomes non-stationary and colored: The "wind" changes its rhythm and color over time in a complex way.

Crucially, they showed that to get this weird, non-white noise, the initial state must essentially be a sum of two-particle states. If the universe starts in a state that is just a different kind of vacuum (created by a mathematical trick called a Bogoliubov transformation), the noise stays white. It takes a specific kind of "excited" state—where pairs of particles are already present—to break the white noise rule. They even calculated the "instantaneous power spectrum" for this scenario, showing how the noise's frequency profile shifts and changes as time goes on.

The Takeaway

The authors didn't just find that the noise changes; they mapped out exactly how it changes based on which rule you break.

If you change the background (the shape of the universe), the noise stays white but gets louder or softer over time.
If you change the cutoff (how you measure the waves), the noise gets "colored" and gains memory, but it stays Gaussian.
If you change the initial state (how the universe starts), the noise gets colored, loses its memory of being stationary, and becomes non-Gaussian (truly weird).

They also checked the "backreaction"—a fancy way of asking, "Does all this extra energy mess up the universe's expansion?" They found that as long as the initial excitement isn't too crazy, the universe can still inflate for a good amount of time (about 15 e-folds) before the extra energy becomes a problem.

In short, the paper suggests that the clean, simple picture of "Gaussian White Noise" is a special case, not the universal rule. The universe might be noisier, more colorful, and more memory-filled than we thought, depending on the fine details of its birth and its shape. While the standard model works well for a "toy" universe, the real cosmos might be a much more complex symphony of noise.

Technical Summary: Deviations from Gaussian White Noise in Stochastic Inflation

Problem Statement
Stochastic inflation serves as a classical effective theory for studying scalar field perturbations on an inflationary spacetime, where subhorizon (UV) quantum modes act as a noise source for superhorizon (IR) modes. In the conventional formulation, pioneered by Starobinsky, this noise is assumed to be Gaussian and white (uncorrelated in time). This assumption relies on three specific conditions: a strictly de Sitter (dS) background, a sharp momentum cutoff (Heaviside step function) separating long and short modes, and the Bunch-Davies vacuum initial state. The lack of memory in white noise leads to Markovian dynamics, simplifying the formalism to standard Langevin and Fokker-Planck equations. However, the authors posit that relaxing these assumptions may alter the statistical properties of the noise, potentially introducing "color" (frequency dependence), non-stationarity, or non-Gaussianity. The paper aims to systematically investigate how deviations from these standard assumptions modify the noise statistics for a free scalar test field.

Methodology
The authors employ the standard stochastic inflation formalism, separating the field operator $\hat{\phi}$ into long ( $\hat{\phi}_l$ ) and short ( $\hat{\phi}_s$ ) modes using a window function $W(\kappa)$ , where $\kappa = k/k_\sigma(t)$ . The noise operators $\hat{\xi}_\phi$ and $\hat{\xi}_v$ are derived from the time derivative of the short modes crossing the cutoff. The statistical properties of the noise are determined by calculating the correlation functions (specifically the anticommutators) in a chosen quantum state $|\Psi\rangle$ .

The study proceeds by relaxing one assumption at a time:

Background Geometry: Comparing exact dS space with quasi-dS (slow-roll) backgrounds.
Window Function: Replacing the sharp cutoff ( $W=\theta(\kappa-1)$ ) with a piecewise linear "non-sharp" window function with finite width $2\delta$ .
Initial State: Considering states $|\Psi\rangle$ that deviate from the Bunch-Davies vacuum $|0\rangle$ , specifically expanding the state in terms of multi-particle excitations (focusing on two-particle states for perturbative corrections).

The authors utilize mathematical tools including mode function expansions, Wronskian conditions, and Green's methods for slowly varying Hubble parameters. They analyze the noise via its two-point correlators and power spectra (both stationary and instantaneous for non-stationary cases).

Key Results

Deviations from Exact de Sitter (Section 3):
- Switching from exact dS to a quasi-dS background (e.g., with a time-dependent Hubble parameter $H(N)$ ) while maintaining the sharp cutoff and Bunch-Davies vacuum results in a non-stationary Gaussian white noise.
- The noise retains its white character (delta-function correlation in time), but the amplitude becomes time-dependent, scaling with the instantaneous power spectrum of the field fluctuations, $P_{\xi} \propto H(N)^2$ . Gaussianity is preserved.
Deviations from the Sharp Cutoff (Section 4):
- Replacing the Heaviside step function with a smooth or piecewise linear window function introduces a finite width in momentum space.
- This modification generates colored noise. The noise correlator $\langle \xi(N_1)\xi(N_2) \rangle$ becomes a function of the time difference $\Delta N$ rather than a delta function, indicating the system acquires "memory" of past e-folds (specifically $\Delta N_\pm$ ).
- The power spectrum becomes frequency-dependent (non-constant), exhibiting a decay at high frequencies.
- Crucially, if the background remains exact dS and the state is the vacuum, the noise remains stationary and Gaussian, despite being colored.
Deviations from the Bunch-Davies Initial State (Section 5):
- Considering a general state $|\Psi\rangle$ as a superposition of the vacuum and excited states, the authors find that to produce a non-trivial deviation from white noise at the next-to-leading order, the state must contain a sum of two-particle states (specifically, terms involving $C_2$ in the expansion).
- States related to the vacuum by a Bogoliubov transformation (which preserve the vacuum structure in a different basis) do not produce colored noise; they only modify the amplitude of the white noise.
- A state with a significant two-particle component leads to non-stationary, colored, and non-Gaussian noise.
- The correlator depends on both $N_1$ and $N_2$ individually, not just their difference. Consequently, the spectral properties are described by an instantaneous power spectrum.
- The paper demonstrates that for a specific model with an exponentially decaying two-particle wavefunction, the noise power is enhanced or suppressed around the time the characteristic momentum scale crosses the horizon.

Significance and Conclusions
The paper establishes that the "Gaussian white noise" assumption in stochastic inflation is a fragile result of specific idealizations.

Memory and Non-Markovianity: Deviations from a sharp cutoff or a vacuum state introduce memory into the stochastic process. This implies that the coarse-grained field dynamics are non-Markovian, rendering standard Langevin equations insufficient and requiring more complex Fokker-Planck formulations.
Gaussianity: While background deviations and window function changes preserve Gaussianity, deviations from the Bunch-Davies vacuum (specifically via multi-particle states) inherently break Gaussianity.
Backreaction: The authors address the potential backreaction of non-vacuum states on the background spacetime. They show that for small deviations (parameterized by $\epsilon$ ), the energy density of the excited modes remains sub-dominant to the background energy density for a significant duration (approx. 15 e-folds) before potentially reaching super-Planckian scales, suggesting the formalism remains valid for a substantial period of inflation.

The authors conclude that while the standard white noise approximation is robust for simple scenarios, a comprehensive understanding of stochastic inflation in more realistic or excited scenarios requires accounting for colored, non-stationary, and potentially non-Gaussian noise. The paper restricts its analysis to free test fields, noting that interactions (self-interactions or gravitational coupling) would introduce further sources of non-Gaussianity and memory, which are left for future investigation.

The Cosmic Static: When the Rules Change

The Takeaway

More like this