The hunt for new pulsating ultraluminous X-ray sources: A clustering approach

¹ Department of Electronics, Information and Bioengineering, Politecnico di Milano, via G. Ponzio, 34, 20133 Milan, Italy
² INAF–Osservatorio Astronomico di Roma, via Frascati 33, 00078 Monte Porzio Catone, Italy
³ Dipartimento di Fisica, Università degli Studi di Roma “Tor Vergata”, via della Ricerca Scientifica 1, 00133 Roma, Italy
⁴ INAF, Istituto di Astrofisica Spaziale e Fisica Cosmica, Via Alfonso Corti 12, 20133 Milan, Italy
⁵ Institut d’Estudis Espacials de Catalunya (IEEC), Edifici RDIT, Campus UPC, 08860 Castelldefels (Barcelona), Spain
⁶ Institute of Space Sciences (ICE, CSIC), Campus UAB, Carrer de Magrans, 08193 Barcelona, Spain
⁷ Ciela – Montreal Institute for Astrophysical Data Analysis and Machine Learning, Montréal, Canada

^★ Corresponding author: nicolooreste.pinciroli@polimi.it

Received: 13 January 2025
Accepted: 19 July 2025

Abstract

Context. The discovery of fast and variable coherent signals in a handful of ultraluminous X-ray sources (ULXs) points to the presence of super-Eddington accreting neutron stars, drastically altering our understanding of the ULX class. Our capability of discovering pulsations in ULXs is limited, among other issues, by poor statistics. However, catalogues and archives of high-energy missions, such as XMM-Newton, Chandra, and Swift, contain information that is often overlooked, but could otherwise be used to identify new candidate pulsating ULXs (PULXs).

Aims. The goal of this research is to single out candidate PULXs among ULXs that have not shown pulsations due to an unfavourable combination of factors (low statistics, low pulsed fraction, etc.).

Methods. We applied an artificial intelligence approach to an updated database of ULXs detected by XMM-Newton. The sample counts 640 sources for a total of ~1800 observations, 95 of which are those of known PULXs. We first used an unsupervised clustering algorithm to sort out sources with similar characteristics into two clusters. Then, the sample of known PULX observations was used to set the separation threshold between the two clusters and to identify the one containing the new candidate PULXs.

Results. We found that only a few criteria are needed to assign the membership of an observation to one of the two clusters. Moreover, the best result in terms of the capability of assigning all the known PULXs in one of the two clusters was obtained when the maximum observed flux for each source is included in the clustering algorithm. The cluster of new candidate PULXs counts 85 unique sources for 355 observations, with ~85% of these new candidates having multiple observations. A preliminary timing analysis found no new pulsations for these candidates.

Conclusions. This work presents a sample of new candidate PULXs observed by XMM-Newton, the properties of which are similar (in a multidimensional phase space) to those of the known PULXs, despite the absence of pulsations in their light curves. While this result is a clear example of the predictive power of non-traditional, AI-based methods, it also highlights the need for high-statistics observational data to reveal coherent signals from the sources in this sample and thereby validate the robustness of the approach.