Classifying Broad Absorption Line Quasars: Metrics, Issues and a New Catalogue Constructed from SDSS DR5

TL;DR

This paper presents a new hybrid classification method for broad absorption line quasars using SDSS DR5 data, resulting in a more complete and robust catalog of 3552 BALQSOs, and discusses the challenges in defining and identifying BALQSOs.

Contribution

A novel hybrid classification scheme combining metrics, neural networks, and visual inspection to improve BALQSO catalog accuracy and robustness.

Findings

The new catalog contains 3552 BALQSOs, representing 12.5% of the sample.

No single metric yields a perfect BALQSO classification; multiple metrics are needed.

BALQSOs are confirmed to be redder than non-BALQSOs and show a signal-to-noise dependence in their fraction.

Abstract

We apply a recently developed method for classifying broad absorption line quasars (BALQSOs) to the latest QSO catalogue constructed from Data Release 5 of the Sloan Digital Sky Survey. Our new hybrid classification scheme combines the power of simple metrics, supervised neural networks and visual inspection. In our view the resulting BALQSO catalogue is both more complete and more robust than all previous BALQSO catalogues, containing 3552 sources selected from a parent sample of 28,421 QSOs in the redshift range 1.7<z<4.2. This equates to a raw BALQSO fraction of 12.5%. In the process of constructing a robust catalogue, we shed light on the main problems encountered when dealing with BALQSO classification, many of which arise due to the lack of a proper physical definition of what constitutes a BAL. This introduces some subjectivity in what is meant by the term BALQSO, and because of this, we also provide all of the meta-data used in constructing our catalogue, for every object in the parent QSO sample. This makes it easy to quickly isolate and explore sub-samples constructed with different metrics and techniques. By constructing composite QSO spectra from sub-samples classified according to the meta-data, we show that no single existing metric produces clean and robust BALQSO classifications. Rather, we demonstrate that a variety of complementary metrics are required at the moment to accomplish this task. Along the way, we confirm the finding that BALQSOs are redder than non-BALQSOs and that the raw BALQSO fraction displays an apparent trend with signal-to-noise, steadily increasing from 9% in low signal-to-noise data, up to 15%.