Bunch-Davies initial conditions and non-perturbative inflationary dynamics in Numerical Relativity

The Gist

The Big Picture: Simulating the Universe's "Baby Photos"

Imagine the universe as a giant, expanding balloon. Long ago, during a period called inflation, this balloon was inflating faster than the speed of light. During this time, tiny quantum jitters (random fluctuations) were stretched out and frozen into the fabric of space. These jitters eventually became the seeds for all the stars, galaxies, and clusters we see today.

For decades, scientists have tried to predict what these jitters looked like using mathematical approximations (perturbation theory). It's like trying to predict the weather by assuming the wind only blows gently and never changes direction. This works well for calm days, but if a massive storm hits (a "non-perturbative" event), the gentle math breaks down.

This paper introduces a new way to simulate the universe. Instead of using gentle math approximations, the authors built a full-scale, high-precision video game engine based on Einstein's General Relativity. They call this Numerical Relativity. It allows them to simulate the universe's early days with all the messy, chaotic, and violent interactions included, not just the smooth parts.

The Challenge: Setting the Stage

To start a simulation of the universe, you need to set the "initial conditions." In the real universe, these conditions come from the Bunch-Davies vacuum, which is essentially the "ground state" of quantum fields before they start fluctuating.

Think of it like this:

• The Old Way: Scientists would draw a few random waves on a piece of paper, hoping they looked right, and then start the simulation. But General Relativity has strict rules (called constraints) that say the geometry of space and the energy inside it must balance perfectly. If you just draw random waves, the math breaks immediately because the rules aren't satisfied.
• The New Way: The authors created a special tool (a Python code called STOIIC-GR) that acts like a "magic sculptor." It takes the quantum rules (the Bunch-Davies vacuum) and carves out a 3D landscape of space and energy that perfectly satisfies Einstein's rules right from the very first frame. It ensures the "stage" is set correctly before the "play" begins.

The Experiment: Three Different Stories

The team ran their simulation on three different types of "universes" (models of the inflaton field) to see how their engine handled different scenarios:

1. The Boring, Smooth Universe (Quadratic Potential):

   • The Analogy: A gentle, rolling hill.
   • The Result: The universe expands smoothly. The random jitters stay small and behave exactly as the old, gentle math predicted.
   • Why it matters: This proved their new engine works. If they can reproduce the known, simple results, they can trust it for the complex stuff.

2. The "Speed Bump" Universe (Inflection Point):

   • The Analogy: Imagine a car driving down a hill that suddenly hits a flat, slippery patch where it almost stops, then speeds up again.
   • The Result: The field slows down dramatically (Ultra Slow-Roll). The authors found that while the field itself barely moved, the geometry of space reacted strongly. The simulation showed that even in this tricky phase, the universe stayed stable, but the "bumps" in the universe grew larger than usual.

3. The "Whiplash" Universe (Strong Resonance):

   • The Analogy: Imagine a trampoline with a bumpy, oscillating surface. If you jump on it at the right rhythm, you might bounce so high you fly off, or get stuck in a dip.
   • The Result: This was the most chaotic scenario. The oscillations were so strong that the universe didn't just expand smoothly; it became bimodal. Some parts of the universe got stuck in a "false vacuum" (a local dip in the energy field) and expanded forever (eternal inflation), while other parts rolled down the hill successfully.
   • The Breakthrough: In this extreme case, the old gentle math completely failed. The authors had to use their full Numerical Relativity engine to see that the universe was splitting into different regions with different fates.

The "Gauge" Problem: Choosing Your Camera Angle

One of the hardest parts of simulating General Relativity is that space and time are flexible. You can look at the universe from different "camera angles" (gauges).

• The authors chose a Geodesic Gauge.
• The Analogy: Imagine taking a photo of a crowd. You could take a photo from a helicopter (looking down at everyone), or you could take a photo from the perspective of a person walking through the crowd.
• The authors used a "walker's perspective" (Geodesic/Synchronous gauge). They showed that even though this angle is tricky and can sometimes cause mathematical glitches (like the camera getting stuck), it works perfectly fine for the inflationary period they studied.

The Results: What Did They Learn?

1. Validation: When the universe is calm, their new super-computer simulation matches the old, simple math perfectly. This proves the new tool is accurate.
2. Non-Perturbative Discovery: When the universe gets wild (Strong Resonance), the old math fails. The new simulation reveals that the universe can split into regions where inflation never ends (eternal inflation) and regions where it succeeds.
3. The "Ruler" Problem: In a chaotic universe, you can't just measure "height" or "density" easily because the ruler itself is stretching and warping. The authors developed a new way to measure the "curvature" of the universe that doesn't depend on which camera angle you use. This allows them to measure the chaos accurately.

The Limitations (The "Fine Print")

The paper is honest about where the simulation hits a wall:

• Resolution Limits: In the most chaotic "Strong Resonance" model, tiny, sharp walls formed in the fabric of space (domain walls). The simulation grid wasn't fine enough to see these walls perfectly, causing some mathematical errors in the "momentum" rules.
• The Fix: They noted that if they used Adaptive Mesh Refinement (AMR)—which is like a camera that automatically zooms in on the messy parts and zooms out on the calm parts—they could fix this. Their code is ready for this, but they didn't use it in this specific paper to keep the focus on the initial setup.

Summary

This paper is a proof-of-concept. It says: "We have built a new, high-fidelity engine that can simulate the birth of the universe from the very first quantum moment, satisfying all of Einstein's strict rules. It works for simple cases, and it reveals new, wild behaviors in complex cases that old math couldn't see."

It paves the way for future simulations that don't rely on "gentle approximations" but instead watch the universe evolve with all its potential chaos and complexity.

Technical Summary

Technical Summary: Bunch-Davies Initial Conditions and Non-Perturbative Inflationary Dynamics in Numerical Relativity

Problem Statement
Current theoretical predictions for cosmological inflation rely heavily on perturbation theory and the assumption of small fluctuations. While analytical methods (e.g., in-in formalism, transport equations) and lattice cosmology codes have advanced, they face significant limitations. Analytical loop computations on curved spacetime are analytically challenging and often lack general treatments. Lattice approaches that include spatial gradients often neglect full spacetime geometry and stochastic noise, failing to provide a rigorous coarse-graining of quantum diffusion. Furthermore, previous Numerical Relativity (NR) simulations of inflation have largely relied on simplified initial conditions (e.g., conformal flatness) that do not strictly satisfy the General Relativity (GR) constraints or originate from the specific quantum vacuum states (Bunch-Davies) required for realistic cosmology. There is a need for a framework that can evolve fully non-linear, non-perturbative inhomogeneities starting from physically admissible, quantum-motivated initial data.

Methodology
The authors present a pipeline, STOIIC-GR (STOchastic Inflation Initial Conditions for GR), which bridges Quantum Field Theory on Curved Spacetime (QFTCS) and full Numerical Relativity.

1. Initial Value Problem: The authors solve the Hamiltonian and Momentum constraints of GR at linear order in perturbation theory. They derive a mapping between the gauge-invariant curvature perturbation $\mathcal{R}$ (and its time derivative $\dot{\mathcal{R}}$) and the full set of initial data required for the BSSN formulation of GR: the conformal metric $\tilde{\gamma}_{ij}$, traceless extrinsic curvature $\tilde{A}_{ij}$, trace of extrinsic curvature $K$, lapse $\alpha$, shift $\beta^i$, inflaton $\phi$, and its momentum $\Pi$.
2. Gauge Choice: To facilitate numerical evolution, the authors adopt a geodesic gauge (equivalent to synchronous gauge in perturbation theory), setting $\Psi_G = B_G = 0$. They demonstrate that while the comoving gauge is standard for analytical calculations, the geodesic gauge allows for straightforward algebraic inversion of the constraints to generate initial data.
3. Stochastic Initialization: The initial conditions are generated by treating the quantum operators as random variables (Gaussian stochastic modes) consistent with the Bunch-Davies vacuum. A single 3D random realization is drawn in Fourier space, ensuring that all derived quantities ($\phi$, $\Pi$, metric perturbations) are cross-correlated correctly according to the linearized constraints.
4. Numerical Evolution: The initial data is evolved using GRChombo (and verified with GRTeclyn), open-source codes solving the BSSN equations with Adaptive Mesh Refinement (AMR) capabilities. The simulations utilize periodic boundary conditions and a 1+log slicing with Gamma drivers, though the specific gauge choice simplifies the shift vector evolution.
5. Extraction: To interpret results in a gauge-invariant manner beyond the perturbative regime, the authors utilize the non-linear generalization of the comoving curvature perturbation, $\mathcal{R}_i$, defined on comoving hypersurfaces, rather than relying solely on linearized variables.

Key Contributions

• Realistic Initial Conditions: The paper provides the first method to generate NR initial data for inflation that strictly satisfies GR constraints to leading order and is sourced directly from the Bunch-Davies vacuum spectrum, including a complete spectrum of modes from slightly sub-Hubble to super-Hubble scales.
• Non-Perturbative Framework: It establishes a pipeline capable of evolving inflationary dynamics where perturbations become large, moving beyond the limitations of linear perturbation theory and the "separate universe" approximation.
• Validation and Diagnostics: The authors introduce rigorous diagnostic tools, including the monitoring of Hamiltonian and Momentum constraint violations ($|H|/[H]$ and $|M|/[M]$) and convergence tests, to ensure the physical validity of the simulations.

Results
The authors present three distinct examples to validate and demonstrate the framework:

1. Quadratic Inflation (Perturbative Regime):

   • A standard slow-roll model ($V \propto \phi^2$) was simulated.
   • Result: The simulation perfectly recovered the background Friedmann evolution and the linear power spectrum predictions. The non-linear curvature $\mathcal{R}_{NL}$ matched the linear $\mathcal{R}$, and constraint violations were suppressed to second-order perturbation levels, validating the code's accuracy in the perturbative limit.

2. Inflection Point Inflation (Ultra Slow-Roll):

   • A potential with an inflection point was used to model Ultra Slow-Roll (USR) dynamics, relevant for Primordial Black Hole (PBH) production.
   • Result: The simulation captured the resonant enhancement of the power spectrum. While the field contrast ($\delta\phi$) decreased during the USR phase (consistent with separate universe predictions), the curvature contrast ($\mathcal{R}$) remained significant. The results showed a departure from perturbativity (up to 30% difference from the background) but maintained good constraint satisfaction. The non-linear curvature $\mathcal{R}_{NL}$ remained close to the linear prediction, suggesting that even in USR, the geometry effectively tracks the perturbations.

3. Strong Resonance Model (Non-Perturbative Regime):

   • A monodromy potential with a strong oscillatory feature was simulated, leading to a highly non-linear scenario.
   • Result: The background solution lost its meaning as the system deviated significantly from the Friedmann solution. The perturbations became bimodal, with some regions undergoing eternal inflation (stuck in false vacua) while others rolled through.
   • Constraint Violation: While the Hamiltonian constraint remained satisfied, the Momentum constraint showed violations at late times. The authors attribute this to the formation of domain walls where the physical width of the defect becomes smaller than the grid resolution (a known issue in topological defect evolution). However, filtering out the violating UV modes confirmed that the large-scale non-perturbative dynamics (bubble formation) were robust and not artifacts of the numerical error.

Significance and Claims
The paper claims to pave the way for the first realistic, non-perturbative, and fully non-linear Numerical Relativity simulations of the early inflationary universe.

• Beyond Perturbation Theory: The work demonstrates that it is possible to sidestep perturbation theory to make predictions for post-inflationary states involving full GR dynamics, particularly in scenarios where non-linearities are strong (e.g., strong resonance, eternal inflation).
• Physical Admissibility: By satisfying the GR constraints at the initial slice, the simulations ensure that the spacetime is physical, addressing a major hurdle in previous NR cosmology attempts.
• Future Directions: The authors modestly note that while the current work uses a fixed initial time slice, the framework is designed to incorporate stochastic inflation (where modes enter the simulation at different times relative to horizon crossing) in future work. This would allow for a more rigorous treatment of quantum diffusion and potentially enable precision predictions for observables like primordial non-Gaussianities and non-linear sub-CMB scales.

The authors emphasize that their results are robust against the removal of constraint-violating UV modes, confirming that the observed non-perturbative phenomena (such as domain wall formation and eternal inflation regions) are physical features of the model rather than numerical artifacts.