Clusters of galaxies discovered in Planck data thanks to artificial intelligence
For the first time, taking advantage of innovative artificial intelligence methods, an IAS team has discovered thousands of galaxy clusters in Planck satellite observations, published in a new catalogue.
Data from the Planck satellite have already been used three times in the collaboration to produce catalogues of galaxy clusters (ESZ-2009, PSZ1-2013, PSZ2-2015) by detecting the Sunyaev - Zeld'ovich effect (SZ, which traces the hot gas in galaxy clusters). The technique used was multi-frequency matched filtering applied to the six frequencies of Planck's High Frequency Instrument (HFI).
In this brand new analysis, the authors proceeded in two steps. First, an all-sky map of the SZ effect was created from the six HFI maps. Then, a neural network trained on known clusters was applied to remove pixels from the map with a very low probability of tracing the hot gas. A filter was then used to keep only the sets of pixels tracing galaxy clusters. The final catalogue thus obtained contains nearly 4000 candidate galaxy clusters, of which only 10% have a non-negligible probability of being false positives. This new catalogue is made available to the community within the IDOC SZ galaxy cluster database (IAS, szcluster-db.ias.u-psud.fr) and by the CDS (Strasbourg Data Centre).
This analysis is the result of numerous developments over several years, devoted to the production of Planck's galaxy cluster catalogues and their validation. This validation step, pooling statistical analyses and physical characteristics, had already been improved thanks to neural networks (Aghanim et al. 2015). Another artificial intelligence technique, based on decision trees, had also made it possible to characterise cluster galaxies by estimating their stellar mass and star formation rate (Bonjean et al. 2019). The next steps will be to use unsupervised artificial intelligence algorithms guided by the physics of galaxy clusters to take advantage of their ability to detect weak signals and separate the components at the same time. An R&D study (funded by CNES), as well as a thesis, are underway on the subject.
The corresponding article is published in A&A (Hurier, Aghanim, Douspis 2021).
Contacts : Nabila Aghanim and Marian Douspis
Figure 2 : Map of the hot gas seen in SZ for a part of the sky, cleaned by Artificial Intelligence. The black circles show the detections while the grey regions are masked due to point source contamination