Unusual quasars from the Sloan Digital Sky Survey selected by means of Kohonen self-organising maps

TL;DR

This study uses self-organising maps and visual inspection of SDSS spectra to identify and catalog 1005 unusual quasars, revealing their properties and differences from normal quasars.

Contribution

It introduces a novel combination of machine learning and visual inspection to efficiently select and analyze unusual quasar spectra from SDSS data.

Findings

Unusual quasars are generally more luminous than normal quasars.

BAL quasars and strong iron emitters have lower radio luminosities.

Strong BALs tend to avoid the most radio-luminous quasars.

Abstract

We exploit the spectral archive of the Sloan Digital Sky Survey (SDSS) Data Release 7 to select unusual quasar spectra. The selection method is based on a combination of the power of self-organising maps and the visual inspection of a huge number of spectra. Self-organising maps were applied to nearly 10^5 spectra classified as quasars by the SDSS pipeline. Particular attention was paid to minimise possible contamination by rare peculiar stellar spectral types. We present a catalogue of 1005 quasars with unusual spectra. This large sample provides a useful resource for both studying properties and relations of/between different types of unusual quasars and selecting particularly interesting objects. The spectra are grouped into six types. All these types turn out to be on average more luminous than comparison samples of normal quasars after a statistical correction is made for intrinsic reddening. Both the unusual broad absorption line (BAL) quasars and the strong iron emitters have significantly lower radio luminosities than normal quasars. We also confirm that strong BALs avoid the most radio-luminous quasars. Finally, we create a sample of quasars similar to the two "mysterious" objects discovered by Hall et al. (2002) and briefly discuss the quasar properties and possible explanations of their highly peculiar spectra. (Abstract modified to match the arXiv format)