The Atacama Cosmology Telescope: Probing new signatures of ultralight axions with gravitational lensing

This paper presents the strongest constraints to date on ultralight axions in the mass range $10^{-26}$ to $10^{-24.5}$ eV using gravitational lensing data from Planck, ACT, and SPT-3G, finding that such axions can constitute at most a few percent of dark matter while noting a tentative $2.1\sigma$ preference for their existence at $10^{-24.5}$ eV that warrants further investigation.

The Gist

Imagine the universe is a giant, invisible ocean filled with "dark matter." For decades, scientists have assumed this ocean is made of heavy, slow-moving particles, like cold, dense rocks floating in water. This is the standard model, called Cold Dark Matter (CDM).

However, a new theory suggests some of this dark matter might be made of Ultralight Axions (ULAs). Think of these not as rocks, but as ghostly, wavy ripples that stretch across entire galaxies. Because they are so light and wave-like, they don't clump together easily; instead, they smooth things out, acting like a cosmic "anti-clumping" force.

This paper is a report from the Atacama Cosmology Telescope (ACT) and its partners, who acted like cosmic detectives. They looked at the "fossilized light" from the Big Bang (the Cosmic Microwave Background) to see if these ghostly ripples are actually there.

Here is what they found, broken down simply:

1. The Mystery: The "S8 Tension"

Scientists have been arguing about how "clumpy" the universe is.

• The Big Bang View: Looking at the early universe, everything seems fairly smooth.
• The Galaxy View: Looking at galaxies today, they seem clumpier than expected.
  This disagreement is called the S8 tension. One way to fix this is if those ghostly axion ripples exist. If they do, they would smooth out the clumps in the early universe just enough to make the two views match.

2. The Investigation: Using Gravity as a Lens

The team didn't just look at the light; they looked at how that light was bent by gravity (gravitational lensing).

• The Analogy: Imagine looking at a streetlight through a wavy glass window. The distortion tells you about the shape of the glass.
• The Reality: The "glass" is the dark matter in the universe. By measuring how the light from the Big Bang is distorted, the team could map out where the dark matter is and how it is clumping.

They used a super-advanced computer model (a "simulation-calibrated nonlinear model") to predict what the universe would look like if it contained these ghostly axion waves. They compared these predictions against real data from three major telescopes: Planck, ACT, and SPT-3G.

3. The Results: How Much Axion is There?

The team tested different "weights" (masses) for these axions to see which ones fit the data best.

• The "Too Light" Ripples: For axions with a mass around $10^{-26}$ eV (extremely light), the data says they can make up less than 1.5% of all dark matter. They are likely not the main ingredient.
• The "Medium" Ripples: For axions with a mass around $10^{-25}$ eV, the limit is higher: they can make up less than 9% of dark matter.
• The "Heavier" Ripples (The Curious Case): For axions with a mass around $10^{-24.5}$ eV, the data showed a slight hint (about a 2.1 sigma preference) that they might exist and make up about 5% of dark matter.
  • What does this mean? It's like hearing a faint noise in a quiet room. It's not loud enough to be a shout (a confirmed discovery), but it's louder than the background silence. The team thinks this "noise" might be caused by a few specific data points that look a bit higher than expected, combined with the fact that these specific axions actually increase the clumping in certain ways.

4. The Conclusion: A "Maybe," Not a "Yes"

The authors are careful not to overhype the results.

• They confirmed that ultralight axions are not the main component of dark matter (they aren't 100% of it).
• They set the strictest limits yet on how much of these particles can exist in the universe.
• They found a faint signal suggesting a small amount of axions might exist, but they warn that this signal is driven by just a few data points.

The Bottom Line:
The universe is mostly made of the "cold rock" dark matter we already know. There might be a tiny splash of "ghostly wave" dark matter (axions) mixed in—perhaps 5% or less—but the evidence isn't strong enough yet to say for sure. The team needs more data and better simulations to confirm if that faint signal is a real discovery or just a trick of the light.

Technical Summary

Technical Summary: Probing New Signatures of Ultralight Axions with Gravitational Lensing

Problem Statement
Ultralight axions (ULAs) are well-motivated dark matter (DM) candidates arising from extensions to the Standard Model, particularly within the string axiverse scenario. While ULAs with masses $m_a \lesssim 10^{-27}$ eV are strongly constrained by Cosmic Microwave Background (CMB) temperature and polarization data, the mass range $10^{-26} \text{ eV} \leq m_a \leq 10^{-24.5} \text{ eV}$ remains less explored. In this regime, ULAs exhibit scale-dependent suppression of the matter power spectrum, potentially alleviating the $S_8$ tension (the discrepancy between CMB and weak lensing inferences of matter clustering). However, previous constraints in this mass range have been limited by the lack of accurate modeling of nonlinear clustering effects, which are crucial for interpreting high-resolution CMB lensing data.

Methodology
The authors analyze recent CMB gravitational lensing measurements from the Atacama Cosmology Telescope (ACT), Planck, and the South Pole Telescope (SPT-3G), combined with Baryon Acoustic Oscillation (BAO) data from DESI DR2. The analysis employs a state-of-the-art, simulation-calibrated nonlinear clustering model for ULAs, specifically the axionHMcode framework.

Key methodological components include:

• Nonlinear Modeling: Unlike linear theory, which predicts only suppression of power, the simulation-calibrated model accounts for the formation of solitonic cores and interference fringes in collapsed structures. This allows for the prediction of a potential enhancement in the matter power spectrum on semi-linear scales ($k \sim 0.1 - 1 \, h/\text{Mpc}$) for specific axion masses and fractions.
• Emulation: Due to the computational expense of solving the Schrödinger-Poisson system for ULAs, a neural network emulator (axionEmu) was trained on the axionHMcode outputs to rapidly evaluate matter and CMB angular power spectra.
• Statistical Analysis: The authors perform Markov Chain Monte Carlo (MCMC) sampling using the Metropolis-Hastings algorithm and nested sampling (via dynesty) to derive posterior distributions for the axion mass ($m_a$) and axion fraction ($\Omega_a/\Omega_d$). Two data combinations are tested: "AP" (ACT + Planck, $L_{\text{max}} \approx 1250$) and "APS" (ACT + Planck + SPT-3G, $L_{\text{max}} \approx 2700$).

Key Results
The study derives the strongest constraints to date on ULAs in the mass range $10^{-26} \text{ eV} \leq m_a \leq 10^{-24.5} \text{ eV}$:

• Constraints on Axion Fraction:
  • For $m_a = 10^{-26} \text{ eV}$, ULAs constitute less than 1.5% of the total DM (95% confidence level).
  • For $m_a = 10^{-25} \text{ eV}$, ULAs constitute less than 9% of the total DM (95% confidence level).
• Preference for Non-Zero Density: A slight preference for a non-zero axion density is identified at $m_a = 10^{-24.5} \text{ eV}$ with a significance of 2.1$\sigma$. The best-fit axion fraction for this mass is approximately 4.6–5.2%.
• Data Drivers: The preference for non-zero axion density at $10^{-24.5} \text{ eV}$ is driven by a combination of factors:
  1. Specific data points in the ACT and SPT-3G lensing bandpowers around multipole $L \approx 1000$ lie above the best-fit $\Lambda$CDM prediction.
  2. The nonlinear model predicts an enhancement in the matter power spectrum (and consequently the lensing signal) for small axion fractions ($\sim 5\%$) at this mass, which aligns with the observed residuals.
• Comparison with Other Probes: The inclusion of CMB lensing data improves constraints on the axion fraction by 72% for $m_a = 10^{-26} \text{ eV}$ compared to primary CMB data alone. For $m_a = 10^{-25} \text{ eV}$, the constraints are 68% tighter than those derived from galaxy power spectrum and bispectrum measurements (BOSS).

Significance and Claims
The paper claims to provide the most stringent limits on ULAs in the $10^{-26} - 10^{-24.5}$ eV mass range by incorporating a simulation-calibrated nonlinear model into CMB lensing analyses. The authors emphasize that while a mild preference ($2.1\sigma$) for ULAs at $10^{-24.5}$ eV exists, it is largely driven by a few data points. Consequently, the authors conclude that further investigation of nonlinear ULA physics is required to confirm or rule out this signal definitively. The work complements recent studies on nonlinear ULA effects by utilizing a broader dataset (ACT + Planck + SPT-3G) and demonstrates that accurate modeling of the nonlinear regime is essential for interpreting high-multipole CMB lensing data in the context of beyond-$\Lambda$CDM cosmologies.