The new Herschel/PACS Point Source Catalogue

TL;DR

This paper presents a comprehensive, high-accuracy Herschel/PACS point source catalogue using a hybrid classical and machine learning approach, significantly improving source detection completeness and reliability over previous versions.

Contribution

It introduces a novel hybrid source extraction methodology combining classical and ML techniques, enhancing the completeness and purity of the Herschel/PACS catalogue.

Findings

Increased source detection completeness across all bands.

Higher catalogue purity with more matches to existing catalogues.

Photometric accuracy of about 1%.

Abstract

Herschel operated as an observatory, therefore it did not cover the whole sky, but still observed ~8% of it. The first version of an overall Herschel/PACS Point Source Catalogue was released in 2017. The data are still unique and are very important for research, especially because no new far-infrared mission is foreseen for at least the next decade. In the framework of the NEMESIS project, we revisited all the photometric observations obtained by the PACS instrument on-board of the Herschel space observatory. We aimed to build the most complete and most accurate Herschel/PACS catalogue to date. Our primary goal was to increase the number of real sources, and decrease the number of spurious sources identified on a strongly variable background. Our goal was to build a blind catalogue, meaning that source extraction is conducted without relying on prior detections at various wavelengths, allowing us to detect sources never catalogued before. We define a hybrid strategy that includes classical and ML source identification and characterisation methods, providing catalogues at much higher completeness levels than before. Quality assessment also involves ML techniques. Our source extraction methodology facilitates a systematic and impartial comparison of sensitivity levels across various Herschel fields, a task that was typically beyond the scope of individual programs. We created a high-reliability and a rejected source catalogue for each PACS passband, i.e., at 70, 100, and 160 {\mu}m. With the high-reliability catalogue, we managed to significantly increase the completeness in all bands. At the same time, while the number of high-reliability detections decreased, the number of sources matching with existing catalogues increased, suggesting that the purity is also higher than before. The photometric accuracy of our pipeline is ~1% based on comparison with the standard star models.