The SLACS strong lens sample, debiased

TL;DR

This paper presents a method to correct for selection biases in strong gravitational lensing studies, applying it to the SLACS sample to accurately infer galaxy mass distributions and density slopes.

Contribution

It introduces an empirical modeling approach to account for complex lens selection effects directly from data, improving bias correction in galaxy property inference.

Findings

Recovered the mass distribution of the SLACS lens parent population.

Quantified the bias in density slope measurements due to selection effects.

Showed most bias arises from prioritization based on velocity dispersion.

Abstract

Strong gravitational lensing observations can provide extremely valuable information on the structure of galaxies, but their interpretation is made difficult by selection effects, which, if not accounted for, introduce a bias between the properties of strong lens galaxies and those of the general population. A rigorous treatment of the strong lensing bias requires, in principle, to fully forward model the lens selection process. However, doing so for existing lens surveys is prohibitively difficult. With this work we propose a practical solution to the problem: using an empirical model to capture the most complex aspects of the lens finding process, and constraining it directly from the data together with the properties of the lens population. We applied this method to real data from the SLACS sample of strong lenses. Assuming a power-law density profile, we recovered the mass distribution of the parent population of galaxies from which the SLACS lenses were drawn. We found that early-type galaxies with a stellar mass of $\log{M_*/M_\odot}=11.3$ and average size have a median projected mass enclosed within a $5$~kpc aperture of $\log{M_5/M_\odot}=11.332\pm0.013$, and an average logarithmic density slope of $\gamma=1.99\pm0.03$. These values are respectively $0.02$~dex and $0.1$ lower than inferred when ignoring selection effects. According to our model, most of the bias is due to the prioritisation of SLACS follow-up observations based on the measured velocity dispersion. As a result, the strong lensing bias in $\gamma$ reduces to $\sim0.01$ when controlling for stellar velocity dispersion.