🔭 astrophysics

Planck Constraints on Turbulence in the Coma Cluster

By reanalyzing Planck Sunyaev-Zel'dovich surface brightness fluctuations of the Coma cluster using a simulation-based inference framework, this study constrains intracluster medium turbulence to have a large injection scale of approximately 540 kpc, a Kolmogorov-like slope, and a substantial 3D Mach number of 0.60, indicating significant non-thermal pressure support consistent with recent XRISM velocity measurements.

Original authors: Baptiste Sigal, Etienne Pointecouteau, Nicolas Clerc, Simon Dupourqué, Alexeï Molin, François Mernier, Rémi Adam, Tanguy Dusserre, François Pajot

Published 2026-08-03

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Baptiste Sigal, Etienne Pointecouteau, Nicolas Clerc, Simon Dupourqué, Alexeï Molin, François Mernier, Rémi Adam, Tanguy Dusserre, François Pajot

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe is filled with giant, invisible bubbles of hot gas, stretching millions of light-years across. These are galaxy clusters, the largest structures in the cosmos held together by gravity. Inside these bubbles, the gas isn't sitting still; it's churning, swirling, and crashing into itself like a massive, cosmic blender. This chaotic movement is called turbulence. Just like wind blowing through a forest or water rushing over rocks, this turbulence creates pressure. Scientists care deeply about this because if they ignore the "push" from this churning gas, they get the wrong answer when trying to weigh the entire cluster. It's like trying to guess the weight of a person by looking at a trampoline they're bouncing on, but forgetting that the trampoline springs are also pushing up. If you don't account for that extra push, you'll think the person is lighter than they really are.

To understand this cosmic churning, astronomers usually look at the gas in two ways. One way is to listen to the gas "sing" using X-ray telescopes, which can hear the speed of the gas molecules directly. But these telescopes can only hear a tiny part of the song at a time. The other way is to look at how the gas pressure ripples across the whole cluster, like watching the ripples on a pond after a stone is thrown in. This paper focuses on that second method, using a special map of the universe made by the Planck satellite. The researchers wanted to figure out how big the "stones" are that are causing the ripples, how fast the gas is moving, and how much of the cluster's total pressure comes from this wild, turbulent motion.

The team, led by Baptiste Sigal, took a fresh look at the Coma cluster, a massive, nearby galaxy cluster that is known to be a bit of a mess, with smaller groups of galaxies crashing into it. Instead of just guessing, they used a clever trick called "Simulation-Based Inference." Think of it like a video game where the computer generates millions of different versions of the Coma cluster, each with different levels of turbulence, different sizes of swirling storms, and different speeds. The computer then compares these fake maps to the real map from the Planck satellite. It keeps the versions that look the most like the real thing and throws away the ones that don't fit.

By running this massive digital experiment, the authors found that the turbulence in the Coma cluster is driven by huge events. They calculated that the "injection scale"—the size of the biggest swirling storms—is about 540 kiloparsecs (with a possible range of 200 to 990 kiloparsecs). To put that in perspective, that's a storm system roughly 1.7 million light-years across. They also found that the gas is moving at speeds between 357 and 1,095 kilometers per second. That is incredibly fast; a bullet travels at about 1 kilometer per second, so this gas is moving hundreds of times faster than a speeding bullet.

The study suggests that this wild motion provides a significant amount of extra pressure, accounting for about 17% of the total pressure in the cluster. This is a big deal because it confirms that if astronomers ignore this turbulence, they will underestimate the mass of the cluster by a noticeable amount. The results align well with other recent studies that used different methods, including direct measurements from the XRISM satellite, which adds confidence to the idea that the Coma cluster is indeed a very turbulent place.

However, the authors are careful to note that while they have a good handle on the size of the storms and the speed of the gas, pinning down the exact "slope" of the turbulence (how the energy changes from big swirls to tiny eddies) is still tricky. The data suggests the turbulence follows a pattern similar to what we see in water or air on Earth, but the evidence isn't strong enough to say that for sure yet. The paper concludes that while their method works beautifully for this giant, messy cluster, looking at smaller or more distant clusters will require even sharper eyes from future telescopes. For now, they have successfully mapped the invisible, churning heart of the Coma cluster, showing us that the universe is far more dynamic and energetic than a quiet, still picture would suggest.

Technical Summary: Planck Constraints on Turbulence in the Coma Cluster

Problem Statement
The intracluster medium (ICM) of galaxy clusters contains non-thermal pressure support generated by turbulent motions, which can bias hydrostatic mass estimates by 5–20%. Characterizing this turbulence is observationally challenging due to the non-linear relationships between observables (density, pressure, velocity) and the complexity of the underlying velocity field. While direct measurements of velocity via X-ray spectral line broadening (e.g., Hitomi, XRISM) have provided insights, they are often limited to small central regions and may suffer from sample variance or confusion with bulk flows during merger events. Indirect methods using thermodynamic fluctuations in X-ray or Sunyaev-Zel'dovich (SZ) emissions offer access to larger scales but require robust statistical frameworks to disentangle turbulence from other dynamical processes like mergers or gas sloshing. The Coma cluster, being dynamically disturbed and proximate, serves as a critical test case, yet previous constraints on its turbulence properties (injection scale, spectral slope, Mach number) have shown discrepancies between different studies and methodologies.

Methodology
This study performs a comprehensive reanalysis of SZ surface brightness fluctuations in the Coma cluster using Planck mission data. The authors utilize the Modified Internal Linear Combination Algorithm (MILCA) $y$ -map, which traces the integrated electron pressure along the line of sight.

Data Processing: The analysis focuses on a $4R_{500} \times 4R_{500}$ region centered on the cluster. A generalised Navarro-Frenk-White (gNFW) profile, projected as an elliptical 2D model, is fitted to the mean SZ emission to isolate fluctuations. Known point sources and the infalling subgroup NGC 4839 are masked to prevent bulk motion contamination from biasing the fluctuation statistics.
Power Spectrum Estimation: The 2D power spectrum of the residual $y$ -fluctuations is computed using a Mexican hat filtering method, which accounts for masked regions and provides robust error propagation.
Simulation-Based Inference (SBI): Due to the complexity of the likelihood function and the stochastic nature of the observable (single realization of a random field), a classical Bayesian approach is deemed insufficient. Instead, the authors employ an SBI framework using normalizing flows (specifically Sequential Neural Posterior Estimation, SNPE).
- Forward Modeling: They generate 200,000 synthetic $y$ -maps by simulating 3D pressure fluctuations modeled as a Gaussian random field with a power spectrum defined by an injection scale ( $l_{inj}$ ), a spectral slope ( $\alpha$ ), and a normalisation ( $\sigma$ ). These 3D fields are projected, convolved with the Planck beam, and contaminated with realistic noise simulations.
- Training: A neural network is trained to approximate the likelihood function $p(P_{2D} | \sigma, \alpha, l_{inj})$ directly from these synthetic observations.
- Validation: The training is rigorously validated using Simulation-Based Calibration (SBC) and Two-sample Analysis of Recovered Parameters (TARP) diagnostics to ensure the neural network is well-calibrated and does not underestimate uncertainties.
Physical Conversion: The inferred parameters of the pressure fluctuation power spectrum are converted into physical turbulence metrics (3D Mach number $M_{3D}$ , turbulent velocity dispersion $\sigma_{v,3D}$ , and non-thermal pressure fraction $P_{turb}/P_{tot}$ ) using scaling relations calibrated on hydrodynamical simulations (Zhuravleva et al. 2023).

Key Results
The study derives constraints on the turbulence properties of the Coma cluster's ICM:

Injection Scale: A large injection scale of $l_{inj} = 540^{+450}_{-200}$ kpc is inferred (using a Gaussian prior on the slope), consistent with turbulence driven by cluster-wide processes like mergers and accretion.
Spectral Slope: The spectral slope is constrained to $\alpha = 3.50^{+0.50}_{-0.46}$ under a Gaussian prior centered on the Kolmogorov value ( $11/3$ ). The authors note that constraining $\alpha$ remains difficult due to the limited dynamic range of the Planck data and the cluster's complex dynamical state, which may disrupt the turbulent cascade.
Mach Number and Velocity: The analysis yields a substantial 3D Mach number of $M_{3D} = 0.60^{+0.13}_{-0.09}$ . This corresponds to a turbulent velocity dispersion in the range $\sigma_{v,3D} = 357 - 1095$ km/s.
Non-Thermal Pressure: The turbulent pressure fraction is estimated at $P_{turb}/P_{tot} = 0.17^{+0.06}_{-0.04}$ , indicating that turbulence contributes significantly to the total pressure budget.

Significance and Claims
The paper claims that its primary contribution is the successful application of a simulation-based inference framework to Planck SZ data, representing the first machine learning-based analysis of Planck SZ fluctuations for turbulence characterization.

Consistency with Independent Methods: The derived Mach number and non-thermal pressure fraction are consistent with recent power spectrum analyses of XRISM data (Eckert et al. 2025) and previous SZ fluctuation studies (Khatri & Gaspari 2016), despite the different datasets and methodologies. This convergence suggests robustness in the constraints on $M_{3D}$ and $P_{turb}/P_{tot}$ .
Validation of Indirect Probes: The results support the scenario of significant turbulence in the Coma cluster, aligning with the complex interplay of dynamical processes (mergers, accretion) expected in such systems.
Methodological Advancement: The study demonstrates that SBI can effectively handle the sample variance and stochasticity inherent in single-observation astrophysical data, providing more reliable posterior distributions than traditional analytical likelihood approaches.
Future Prospects: The authors posit that their approach paves the way for systematic multi-probe studies combining SZ and X-ray data. They suggest that while Planck data is optimal for large, bright clusters like Coma, extending this methodology to more compact systems will require higher-resolution millimeter-wave observations (e.g., NIKA2/IRAM-30m, ACT/SPT) and future cross-validation with XRISM spectroscopic constraints.

The paper remains modest regarding the spectral slope, acknowledging that current datasets and turbulence modeling assumptions limit the ability to tightly constrain $\alpha$ , a challenge shared by recent direct velocity measurements.

More like this