Influence of finite-temperature effects on CMB power spectrum

1. Problem Statement

The paper addresses the limitations of the standard $\Lambda$CDM cosmological model, specifically regarding the treatment of the Cosmological Constant (CC) and the potential role of quantum effects in the early Universe.

• The Core Issue: Standard cosmological models treat the CC as a constant vacuum energy density ($\Omega_\Lambda$). However, finite-temperature Quantum Field Theory (finite-T QFT) suggests that vacuum energy receives temperature-dependent corrections. Since the temperature of the Universe scales as $T \propto a^{-1}$ (where $a$ is the scale factor), these corrections introduce time-dependent terms into the CC.
• The Gap: Conventional hydrodynamic approaches and Boltzmann equations used in codes like CLASS do not account for these finite-T quantum loop corrections. The authors argue that these effects, though small, are non-negligible during the radiation-dominated era and around recombination, potentially acting as a form of "early dark energy."
• The Goal: To investigate whether incorporating finite-T quantum corrections to the CC improves the fit to Cosmic Microwave Background (CMB) data (specifically Planck 2018) compared to the standard $\Lambda$CDM model.

2. Methodology

The authors employ a hybrid approach combining theoretical modifications to cosmological equations with advanced statistical and machine learning techniques, diverging from the standard Bayesian inference used by the Planck collaboration.

A. Theoretical Framework

• Modified Cosmological Constant: Based on previous work (Refs [5–7]), the total CC is modeled as a sum of a constant term and temperature-dependent corrections:
  $$\Lambda_{tot} = \Lambda_1 + \Lambda_2 \left(\frac{a_0}{a}\right)^4 + \Lambda_3 \left(\frac{a_0}{a}\right)^2 + \dots$$
  • $\Lambda_1$: Standard vacuum energy.
  • $\Lambda_2$: Scales as $T^4$ (similar to radiation density but of quantum origin).
  • $\Lambda_3$: Scales as $T^2$ (similar to curvature density but of quantum origin).
• New Parameters: This introduces two new density parameters, $\Omega_{\Lambda 2}$ and $\Omega_{\Lambda 3}$, distinct from standard radiation ($\Omega_r$) and curvature ($\Omega_K$) because they arise from off-shell quantum loop corrections rather than on-shell particle densities or geometric curvature.
• Implementation: The authors modified the CLASS (Cosmic Linear Anisotropy Solving System) code to include these terms in the Hubble parameter equation.

B. Statistical and Machine Learning Approach

Instead of standard Markov Chain Monte Carlo (MCMC) Bayesian sampling, the authors use a deterministic, regression-based approach:

1. Data Generation: They performed brute-force scans using the modified CLASS (via its Python wrapper, CLASSy) to generate millions of theoretical CMB power spectra ($C_\ell$) across various parameter sets.
2. Target Variable: The "distance" (Euclidean distance) between the generated theoretical curve and the Planck 2018 theoretical reference spectrum.
3. Regression Modeling: They treated cosmological parameters as input features and the "distance" as the target variable. They applied quartic polynomial regression (degree 4) to map parameters to the distance metric.
4. Model Selection: They compared three models:
   • Standard $\Lambda$CDM with spatial curvature ($\Omega_K$) (7 parameters).
   • Finite-T Model with $\Omega_{\Lambda 3}$ only (7 parameters).
   • Finite-T Model with both $\Omega_{\Lambda 2}$ and $\Omega_{\Lambda 3}$ (8 parameters).
5. Validation: They utilized 10-fold cross-validation, bootstrap resampling (1,000 iterations), and metrics like $R^2$, Mean Squared Error (MSE), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC) to assess predictive accuracy and prevent overfitting.

3. Key Contributions

• Phenomenological Extension: Successfully extended the $\Lambda$CDM model to include finite-temperature quantum corrections as distinct, non-redundant density parameters ($\Omega_{\Lambda 2}, \Omega_{\Lambda 3}$).
• Methodological Innovation: Demonstrated the efficacy of using machine learning regression (specifically quartic regression) as a surrogate model for cosmological parameter inference, offering a computationally efficient alternative to expensive Bayesian sampling for initial model exploration.
• Non-Redundancy Proof: Theoretically and numerically verified that $\Omega_{\Lambda 2}$ and $\Omega_{\Lambda 3}$ are physically distinct from standard radiation and curvature parameters, despite sharing similar scale-factor dependencies ($a^{-4}$ and $a^{-2}$).

4. Key Results

The study compared the performance of the standard model against the finite-T models using the Planck 2018 data as a benchmark.

• Improved Fit Quality:

  • Distance Reduction: The standard $\Lambda$CDM model achieved a minimum distance of ~30.78. The 7-parameter finite-T model ($\Omega_{\Lambda 3}$) reduced this to ~28.12, and the 8-parameter model ($\Omega_{\Lambda 2} + \Omega_{\Lambda 3}$) achieved the lowest distance of ~26.83.
  • Regression Performance: The finite-T models showed significantly higher predictive accuracy.
    • $R^2$: Both finite-T models achieved training/testing $R^2$ values near 99.9% (vs. 99.1% for standard $\Lambda$CDM).
    • MSE: The test MSE for finite-T models was 0.08–0.11, roughly 7–10 times lower than the standard model (0.83).
  • Information Criteria: The AIC values for finite-T models were dramatically lower (e.g., -3138 for the 7-param finite-T model vs. +471 for standard $\Lambda$CDM), indicating a vastly superior fit that cannot be attributed solely to parameter count.

• Parameter Shifts:

  • The inclusion of $\Omega_{\Lambda 2}$ and $\Omega_{\Lambda 3}$ led to slight but systematic shifts in other parameters.
  • Hubble Constant ($h$): Shifted downward (e.g., from ~0.679 in standard to ~0.675 in the 8-param model).
  • Reionization Optical Depth ($\tau_{reio}$): Slightly increased.
  • $\Omega_{\Lambda 2}$ Value: The best-fit value was found to be negative and small ($\sim -2 \times 10^{-8}$). The authors discuss the "unnaturalness" of this magnitude, suggesting it may be an artifact of the renormalization scheme or require higher-order corrections to resolve.

• Robustness: Bootstrap resampling confirmed that the optimized parameters are stable and the results are not artifacts of specific data sampling.

5. Significance and Implications

• Physical Relevance: The results suggest that finite-temperature quantum gravitational effects are not negligible and may represent a form of early dark energy that improves the consistency between theoretical models and observational CMB data.
• Model Selection: The study argues that quantum corrections provide a more parsimonious explanation for cosmological dynamics than simply allowing for non-zero spatial curvature, as evidenced by the drastic improvement in AIC/BIC scores.
• Future Directions:
  • The authors acknowledge that their current approach is a "model-building" exercise using synthetic data scans rather than a full Bayesian likelihood analysis.
  • Future work should include full Bayesian parameter estimation (computationally intensive) and the inclusion of higher-order terms in the temperature expansion (e.g., linear $T$ terms) to address the small magnitude of the fitted $\Omega_{\Lambda 2}$.
  • The negative sign of $\Omega_{\Lambda 2}$ warrants further theoretical investigation regarding renormalization schemes.

In conclusion, the paper provides a compelling exploratory framework demonstrating that incorporating finite-temperature quantum corrections into the cosmological constant significantly enhances the fit to Planck 2018 data, offering a new avenue for resolving tensions in precision cosmology.