A deep learning approach to infer galaxy cluster masses from Planck Compton-y parameter maps

We estimate over a thousands masses of PSZ2 Planck galaxy clusters using deep learning with a 15 percent bias with respect to the masses estimated by Planck.
Published in Astronomy
A deep learning approach to infer galaxy cluster masses from Planck Compton-y parameter maps
Like

Share this post

Choose a social network to share with, or copy the shortened URL to share elsewhere

This is a representation of how your post may appear on social media. The actual post will vary between social networks

Small density fluctuations in the Early Universe evolve to form the most massive and gravitationally-bound structures: galaxy clusters. The main component of galaxy clusters is dark matter (DM), which accounts for 85 percent of the total mass, though it cannot be directly observed. Therefore, the mass of a galaxy cluster is usually estimated by the dynamics of stars (5 percent of the total mass) or the thermal properties of the Intra Cluster Medium (ICM, around 10 percent of the total mass).

The ICM can be observed through the inverse Compton scattering (the Sunyaev-Zel’dovich, SZ effect) in millimeter wave-lengths. The Planck collaboration has detected and measured over a thousand clusters of galaxies via the SZ signal and estimated their masses, which are one of the best cosmological probes to constrain cosmological parameters of the Universe. However, these masses are supposedly biased-low estimated in order to alleviate the tension with the Cosmic Microwave background cosmological parameters. 

To address this issue, we generate almost 200 thousand SZ maps from The300 simulations (http://the300-project.org/)  that are statistically similar to the real Planck clusters in mass, redshift, signal intensity and noise levels. We further train a set of Convolutional Neural Networks (CNN) using the mock images aiming at predicting the masses of over one thousand SZ cluster maps provided by the Planck collaboration. Note that the CNN model does not rely upon assumptions on the ICM physics, but it receives input data directly from simulations.  We show that the CNN predicted masses are in agreement with Planck masses but with a slight bias which might be caused by SZ intensity and mass scaling relation parameters. 

From the simulation side, we have tested our model with different hydrodynamical codes that model the feedback effect of the supermassive Black Holes at the center of galaxies with variable levels of intensity. Our results suggest that our method is robust to modifications in the physics modeling of the gas and stellar component  implemented in the simulations.

From the observational side, the standard assumption of hydrostatic equilibrium (HE) used for estimating cluster masses suggests that a bias between 10-20% is guaranteed. However, the bias dependence with cluster mass found in our results also suggests that HE is not the only source of bias but there is also an additional bias due to the calibration of the SZ signal intensity-mass relation. Nevertheless, this small  bias makes the reconciliation with the CMB constraints even harder.

Schematic view of the model trained with simulated SZ cluster maps aiming at predicting the
mass of real Planck SZ cluster maps

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in

Follow the Topic

Astronomy, Cosmology and Space Sciences
Physical Sciences > Physics and Astronomy > Astronomy, Cosmology and Space Sciences

Related Collections

With collections, you can get published faster and increase your visibility.

Wind, water and dust on Mars

In this Collection, we bring together recent work, and invite further contributions, on the nature and characteristics of the Martian surface, the processes at play, and the environmental conditions both in the present-day and in the distant past.

Publishing Model: Hybrid

Deadline: Ongoing

Progress towards the Sustainable Development Goals

The year 2023 marks the mid-point of the 15-year period envisaged to achieve the Sustainable Development Goals, targets for global development adopted in September 2015 by all United Nations Member States.

Publishing Model: Hybrid

Deadline: Ongoing