\texttt{SWIM}: Stochastic Warm Inflation Module to generate and analyse Warm Inflationary power spectrum

The paper introduces **SWIM** (Stochastic Warm Inflation Module), a C++/Python-based numerical platform that enables the generation of fully numerical scalar power spectra for any warm inflation model and integrates with Cobaya to facilitate efficient parameter estimation against cosmological data using machine learning techniques.

The Gist

The Cosmic Soup Stirrer: Explaining SWIM

Imagine you are trying to understand how the universe began. Most scientists use a theory called "Cold Inflation," which is like a sudden, massive expansion of space that happens in a frozen, empty void. In this version, the universe expands so fast that everything stays quiet and still, and only much later does "heat" (radiation) get created.

But there is another theory called "Warm Inflation." Imagine instead of a frozen void, the early universe was a thick, bubbling, energetic cosmic soup. As the universe expanded, the energy wasn't just sitting there; it was constantly being stirred and transferred between different ingredients. This "stirring" (called dissipation) creates heat continuously, meaning the universe was "warm" from the very start.

The Problem: The Math is a Messy Kitchen
While Warm Inflation sounds more realistic, it is a mathematical nightmare. In Cold Inflation, the math is like following a simple recipe: Add A, then B, then C. In Warm Inflation, because everything is bubbling and interacting, the math is like trying to predict the exact movement of every single bubble in a boiling pot of thick stew while someone is shaking the pot.

If you try to use "shortcuts" (simplified math) to predict what the universe looked like, you get the recipe wrong. You might think you’re making soup, but you’ve actually made a very inaccurate version of it. If you use these inaccurate recipes to compare your theory to real telescope data, your conclusions will be wrong.

The Solution: SWIM (The Ultimate Cosmic Chef)
This paper introduces a new software tool called SWIM (Stochastic Warm Inflation Module). Think of SWIM as a super-intelligent, high-speed robotic kitchen.

Here is what SWIM does:

1. The Simulator (The Bubble Tracker): Instead of using "shortcuts" or simplified recipes, SWIM uses "stochastic" math. This is a fancy way of saying it simulates the chaos. It tracks the random, jittery, bubbling nature of those cosmic particles. It doesn't guess; it simulates the actual "splashing" of the cosmic soup.
2. The Speed Demon (The Machine Learning Sous-Chef): Simulating every single bubble is incredibly slow. If you tried to do this for every possible version of the universe, your computer would melt before you finished. To fix this, the authors added Machine Learning (AI).
   • Imagine a chef who is learning a new recipe. At first, they cook every single dish from scratch (which is slow). But as they cook, they start to recognize patterns.
   • Soon, the AI "Sous-Chef" says, "Hey, I've seen this ingredient combination before; I know exactly how it will taste!"
   • The AI predicts the results instantly, only going back to the "slow" manual cooking when it encounters a brand-new, tricky ingredient. This makes the research thousands of times faster.
3. The Judge (The Data Matcher): Finally, SWIM takes these simulated "soups" and compares them to real data from telescopes (like the Planck satellite). It asks: "Does our simulated bubbling soup look like the actual leftover echoes of the Big Bang we see in the sky?"

Why does this matter?

Before SWIM, scientists had to choose between being accurate but too slow or fast but too inaccurate.

SWIM breaks that choice. It allows scientists to test very complex, realistic theories about the early universe against real-world data without waiting years for a computer to finish the math. It’s a complete toolkit that helps us figure out if our universe started as a frozen void or a warm, bubbling cauldron.

Technical Summary

Technical Summary: SWIM (Stochastic Warm Inflation Module)

1. Problem Statement

In the Warm Inflation (WI) paradigm, the inflaton field is coupled to a radiation bath, leading to continuous energy dissipation. This makes the dynamics significantly more complex than standard Cold Inflation (CI). The primary challenges addressed by this paper are:

• Inaccuracy of Analytical Solutions: Most analytical derivations of the scalar power spectrum in WI assume the dissipative coefficient ($\Upsilon$) is independent of temperature ($T$). However, realistic particle physics models (e.g., Minimal Warm Inflation, Warm Little Inflaton) feature temperature-dependent dissipation, rendering analytical solutions inaccurate.
• The $G(Q)$ Limitation: To correct analytical errors, researchers use a correction factor $G(Q)$, where $Q$ is the ratio of dissipative to Hubble friction. Current methodologies often assume $G(Q)$ is a universal function of $Q$ alone. This paper demonstrates that $G(Q)$ can depend heavily on other model parameters (like potential normalization $V_0$ or relativistic degrees of freedom $g_*$), leading to systematic biases in parameter estimation if a pre-computed $G(Q)$ is used.
• Computational Bottlenecks: While full numerical solutions of the stochastic perturbation equations are necessary for accuracy, they are computationally expensive for Markov Chain Monte Carlo (MCMC) sampling, as they require averaging over many stochastic realizations at every step.

2. Methodology

The authors present SWIM, a multi-module numerical platform written in C++ and Python. The methodology is divided into three distinct approaches:

• Module I: $G(Q)$ Calculation (Semi-Analytical): This module solves the background and stochastic perturbation equations to generate a numerical table of $G(Q)$ as a function of $Q$. It uses a parallelized root-finding algorithm to set initial conditions and solves the equations to find the ratio between the numerical and analytical power spectra.
• Module II: Semi-Analytical Parameter Estimation: This module integrates the $G(Q)$ factor into the scalar power spectrum formula, allowing it to be used with standard MCMC codes like Cobaya or COSMOMC via CAMB.
• Module III: Full Numerical Power Spectrum & Machine Learning (The Core Innovation): To solve the "parameter-dependent $G(Q)$" problem, SWIM can bypass $G(Q)$ entirely by feeding the full numerical power spectrum directly into MCMC. To overcome the massive computational cost, the authors implement Random Forest Regression (RFR).
  • On-the-fly Emulation: The RFR acts as a surrogate model. During the MCMC "burn-in" phase, it learns the mapping between model parameters and the power spectrum parameters ($A_s, n_s, \alpha_s, \beta_s$).
  • Reliability Criterion: The code uses the agreement between individual trees in the Random Forest to decide whether to use the fast emulator or invoke the expensive, high-precision stochastic solver.

3. Key Contributions

• A Complete Numerical Platform: Unlike existing codes (WarmSPy or WI2easy), SWIM is a "one-stop shop" that handles everything from background evolution to full Bayesian parameter inference against CMB data.
• Stochastic Solver: SWIM directly solves the standard stochastic perturbation equations (rather than the deterministic Fokker-Planck equations used by some competitors), providing a more direct treatment of the physics.
• Machine Learning Integration: The use of an RFR emulator to accelerate MCMC for stochastic processes is a significant methodological advancement in inflationary cosmology.
• Handling Complex Physics: The code accommodates any inflationary potential and any form of dissipative coefficient, including complex Effective Field Theory (EFT) models.

4. Results

• Validation of $G(Q)$: The authors compared SWIM with WI2easy and found that both the stochastic approach (SWIM) and the Fokker-Planck approach (WI2easy) yield equivalent results in the weak ($Q \ll 1$) and strong ($Q \gg 1$) dissipative regimes.
• Demonstration of Bias: The paper proves that for certain models (e.g., the EFT model from Ref [6]), $G(Q)$ varies with $V_0$ and $g_*$. Using a fixed $G(Q)$ would result in incorrect parameter constraints.
• Efficiency of the Emulator: Testing the RFR emulator against the full numerical solver showed that the emulator accurately reproduces the posterior distributions of model parameters, making high-precision parameter estimation computationally feasible.
• Performance: SWIM was shown to outperform WI2easy in runtime for most tested scenarios.

5. Significance

SWIM provides the cosmological community with the first robust tool capable of testing complex, realistic Warm Inflation models against high-precision observational data (like Planck, ACT, and SPT). By bridging the gap between complex particle physics models and cosmological likelihood analyses through machine learning, it enables a level of rigor in Warm Inflationary research that was previously computationally impossible.