Double the axions, half the tension: multi-field early dark energy eases the Hubble tension

This paper demonstrates that multi-field axion-like early dark energy models, particularly those with two fields, significantly alleviate the strong constraints from Planck CMB data and reduce the Hubble tension to a $1.5\sigma$ residual, outperforming single-field scenarios by better fitting high-$\ell$ CMB data and modifying the pre-recombination history.

The Gist

The Big Problem: The Universe's Speedometer is Broken

Imagine the universe is a giant car, and the Hubble Constant ($H_0$) is its speedometer. It tells us how fast the universe is expanding right now.

For the last decade, physicists have been arguing about what the speedometer actually reads.

• Team A (The Local Crew): They look at nearby stars and supernovae (like checking the speedometer directly). They say, "The car is going 73.5 units per hour."
• Team B (The History Buffs): They look at the Cosmic Microwave Background (CMB)—the "baby photo" of the universe taken 13.8 billion years ago. Using the standard rules of physics (the $\Lambda$CDM model), they calculate what the speed should be today. They say, "Based on the baby photo, the car is only going 67.2 units per hour."

This is a huge disagreement (over 7 standard deviations, or "sigmas"). In the world of science, this is like two mechanics looking at the same car and one saying it's doing 70 mph while the other insists it's doing 40 mph. One of them is wrong, or the rules of the road (physics) are incomplete.

The Previous Attempt: The "One-Field" Fix

To fix this, scientists proposed a new theory called Early Dark Energy (EDE). Think of the universe's expansion as a river. Usually, the river flows smoothly. But EDE suggests that just before the "baby photo" was taken, a tiny, invisible dam opened up, releasing a burst of water that sped up the river slightly. This would make the "baby photo" look different, allowing the math to match the faster speed Team A sees today.

The simplest version of this theory used one single "axion" field (a hypothetical particle) to act as that dam.

• The Result: It worked well enough to lower the disagreement to about 2 or 3 "sigmas."
• The New Problem: Recently, the Planck satellite released a super-sharp, high-definition version of the "baby photo" (called NPIPE data). When scientists tried to fit the "One-Field" EDE theory to this new, high-definition data, it failed miserably. The theory made the baby photo look too blurry in specific ways. It was like trying to fit a square peg into a round hole; the new data rejected the single-field solution.

The New Solution: "Double the Axions"

This paper asks: What if we didn't use just one dam, but two?

The authors propose a Multi-Field EDE model. Instead of one axion particle doing all the work, they suggest there are two (or more) different axion fields working together, but slightly out of sync.

The Analogy: The Orchestra vs. The Soloist

• The One-Field Model (The Soloist): Imagine a violinist trying to fix a broken song by playing one loud, sharp note. It might fix the rhythm, but it sounds harsh and out of place with the rest of the orchestra (the CMB data). The new Planck data is like a critical conductor saying, "That note is too loud and in the wrong spot!"
• The Two-Field Model (The Duet): Now, imagine two violinists playing slightly different notes that blend together. Instead of one sharp spike in energy, they create a smoother, broader wave of energy injection.
  • One axion peaks slightly earlier.
  • The other peaks slightly later.
  • Together, they create a "smeared out" effect that is much gentler on the data.

What Did They Find?

1. The Tension Drops: By adding that second axion, the disagreement between the "Local Crew" (73.5) and the "History Buffs" (Planck data) drops from a massive 3.7 sigma down to just 1.5 sigma.
   • Translation: The gap is now so small that it could just be a statistical fluke (bad luck in the dice roll), rather than a fundamental error in our physics.
2. The "Sweet Spot" is Two: Adding a third or fourth axion didn't help much more. It's like adding a third violinist to the duet; the music doesn't get significantly better, but the complexity (and the number of free parameters) goes up. Nature seems to prefer the "duet."
3. Fixing the High-Definition Data: The second axion specifically fixed the part of the data that the single-field model broke (the high-resolution "high-$\ell$" data). It smoothed out the energy injection so the "baby photo" looked exactly right.

The Bottom Line

The paper argues that the failure of the "Early Dark Energy" theory wasn't because the idea was wrong, but because the simplest version (one field) was too rigid.

By doubling the axions, the theory becomes flexible enough to satisfy the strict new data from the Planck satellite while still solving the Hubble Tension. It suggests that the early universe didn't just get a single "kick" of energy, but perhaps a more complex, multi-step "push" that smoothed out the expansion history just enough to make the speedometer readings match.

In short: The universe's speedometer isn't broken; we just needed to realize the engine had two pistons firing, not one.

Technical Summary

1. Problem Statement

The paper addresses the Hubble tension, a significant discrepancy ($>7\sigma$) between the local measurement of the Hubble constant ($H_0 \approx 73.5$ km/s/Mpc from the Local Distance Network) and the value inferred from Cosmic Microwave Background (CMB) data assuming the standard $\Lambda$CDM model ($H_0 \approx 67.2$ km/s/Mpc).

While Early Dark Energy (EDE) models, specifically single-field axion-like models, have been proposed to resolve this by reducing the sound horizon prior to recombination, they face a critical challenge:

• The NPIPE Constraint: The latest Planck NPIPE CMB data, when combined with Baryon Acoustic Oscillation (BAO) data, strongly disfavors the simplest single-field EDE model.
• Residual Tension: In the single-field scenario, the tension with the local $H_0$ measurement remains high at 3.7$\sigma$, suggesting the single-field implementation is insufficient or that the model fails to fit high-$\ell$ CMB data.

The authors hypothesize that the failure is not inherent to the EDE framework itself but is a limitation of the single-field implementation. They propose that a multi-field axion sector could alleviate these constraints.

2. Methodology

The authors extend the standard axion-like EDE model to include $N_{ax}$ non-interacting pseudo-scalar fields ($\phi_i$) with distinct masses ($m_i$) and decay constants ($f_i$).

• Model Setup:

  • The potential for each field is given by $V_i(\phi_i) = m_i^2 f_i^2 [1 - \cos(\phi_i/f_i)]^n$, with the index fixed at $n=3$ (motivated by data insensitivity to $n \in [2,5]$ and ensuring faster-than-radiation dilution).
  • Initial conditions are fixed to $\theta_i = \phi_i/f_i = 2.8$ and $\dot{\theta}_i = 0$ to reduce parameter space.
  • The fields are characterized by phenomenological parameters: the critical redshift ($z_c^i$) where the field becomes dynamical and the fractional energy density ($f_{ax}^i$) at that redshift.
  • The study analyzes models with $N_{ax} = 1, 2, 3$ fields.

• Data Combinations:

  • P: Planck 2018 (low-$\ell$ TT/EE, PR3 lensing) + CamSpec PR4 (high-$\ell$ TT/TE/EE).
  • PD: P + PantheonPlus (SNeIa) + DESI DR2 (BAO).
  • PDH$_0$: PD + Gaussian prior on $H_0 = 73.50 \pm 0.81$ km/s/Mpc (Local Distance Network).

• Analysis Techniques:

  • Bayesian Analysis: Performed using MCMC (MontePython) to derive posterior distributions.
  • Frequentist Analysis: Profile Likelihood (PL) analysis using the Procoli package to mitigate prior volume effects, which can artificially shift $H_0$ in Bayesian analyses of EDE.
  • Model Comparison: Used the Akaike Information Criterion (AIC) to penalize model complexity (extra parameters) against goodness-of-fit ($\chi^2$).

3. Key Contributions

• Introduction of Multi-Field EDE: The paper introduces a "multi-axion" framework motivated by the string axiverse, demonstrating that multiple fields can smooth out the energy injection history compared to a single sharp peak.
• Resolution of NPIPE Tension: It demonstrates that the tight constraints from Planck NPIPE data on single-field EDE are significantly relaxed when a second field is introduced.
• Optimization of Field Count: The study establishes that adding a second field is the "sweet spot." Adding a third field provides negligible improvement, indicating that the framework is predictive and not arbitrarily flexible.

4. Key Results

A. Reduction of Hubble Tension

• Single Field ($N_{ax}=1$): Residual tension with $H_0^{DN}$ is 2.6$\sigma$ (Bayesian) / 2.3$\sigma$ (Profile Likelihood). The fit to high-$\ell$ NPIPE data worsens significantly.
• Two Fields ($N_{ax}=2$): Residual tension drops to 1.5$\sigma$ (Bayesian) / 1.3$\sigma$ (Profile Likelihood). This is statistically compatible with a fluctuation.
• Three Fields ($N_{ax}=3$): Residual tension is 1.3$\sigma$, showing no significant improvement over the 2-field model.

B. Model Preference (AIC)

• The 2-field model is favored over the 1-field model for the PDH$_0$ combination ($\Delta$AIC = -1.70). The improvement in $\chi^2$ ($|\Delta\chi^2| \approx 4$) justifies the two extra parameters ($z_c$ and $f_{ax}$).
• The 3-field model is not favored over the 2-field model ($\Delta$AIC = +1.76), as the marginal improvement in fit does not justify the additional complexity.

C. Physical Mechanism

• Smoothing Energy Injection: In the 1-field case, the EDE fraction peaks sharply around $\log_{10} z_c \sim 3.6$. In the 2-field case, the energy injection is spread over a wider redshift range:
  • Field 1 peaks at lower redshift ($\log_{10} z_c \sim 3.4$).
  • Field 2 peaks at higher redshift ($\log_{10} z_c \sim 3.9$).
  • The peak heights are lower ($f_{ax} \sim 0.08$ vs $0.125$), resulting in a smoother total EDE fraction.
• High-$\ell$ Fit: The second field specifically improves the fit to Planck NPIPE high-$\ell$ data (multiples $\ell \sim 800-1200$), which was the primary driver of the tension in the single-field case. This is linked to a decrease in the damping scale ($r_d^*$) by $\sim 1\sigma$.

D. Other Cosmological Parameters

• $S_8$ Tension: The addition of a second axion does not worsen the $S_8$ (clumpiness) tension; it remains at the $\sim 2\sigma$ level compared to DES-Y6 results.
• Primordial Nucleosynthesis: The model requires an increase in the baryon density ($\omega_b$) to fit the damping tail, which increases tension with the PRIMAT determination of light element abundances (from 2.0$\sigma$ to 2.8$\sigma$), though it remains consistent with PArthENoPE estimates.

5. Significance and Conclusion

The paper concludes that the difficulty EDE faces in reconciling with Planck NPIPE data is a limitation of the single-field approximation, not a failure of the EDE concept itself.

• Theoretical Implication: A complex pre-recombination expansion history involving multiple light fields (potentially from the string axiverse) is a viable solution to the Hubble tension.
• Predictive Power: The fact that the 2-field model is optimal and the 3-field model offers no gain suggests the framework is predictive. Future high-precision data (e.g., from Simons Observatory, CMB-S4, or DESI) could confirm or exclude this specific multi-field scenario.
• Broader Context: The results suggest that the Hubble tension may point toward a richer pre-recombination physics landscape that cannot be captured by simple single-field modifications to $\Lambda$CDM.