Strong Gravitational Lenses in the Era of Wide-Field Surveys

Abstract

Gravitational lensing is the deflection of a light beam due to the distortion of space-time caused by a massive object. Strong gravitational lensing occurs when this deflection is sufficient to produce multiple images of the same background source as viewed by an observer. The applications of such strong lens systems are widespread, from studying the initial mass function to measurements of the Hubble Constant. In the coming years, strong lens science will undergo a revolution with key data releases from the Legacy Survey of Space and Time (LSST) and Euclid Wide Survey (EWS) which are each expected to identify ∼100 000 lensed systems, increasing the number of known candidates by two orders of magnitude. In this thesis, I describe the opportunities that this data will bring.

Typical searches for strong lenses have to date focussed on visible, sub-millimetre and radio wavelengths. However, the advent of large format, sensitive, Near-Infrared (NIR) detectors as seen in Euclid and the James Webb Space Telescope (JWST) will enable searches in high-resolution NIR surveys. I demonstrate that JWST will identify lensed galaxies at higher redshift than ever seen before and that Euclid will allow the NIR lens population to be studied at scale.

The vast data volume from wide-field surveys such as Euclid and LSST presents a false positive problem, whereby high-scoring samples from lens classifiers are dominated by false positives. I show that the current performance of strong lens classifiers can be improved by combining these into an ensemble. Moreover, I demonstrate that even the state-of-the-art classifiers will still produce heavily contaminated or incomplete samples of lenses when applied to wide-field surveys. Given the vast majority of the lenses identified in such surveys will not receive spectroscopic confirmation, analysis of the complete lens candidate datasets will require the possibility of contamination to be accounted for. I demonstrate that cosmological parameter inference can still be conducted even in the presence of such contamination. To do so requires accepting a known contamination rate and knowledge of the probability that each system is a lens but, as I demonstrate, such data can be obtained from current lens classifiers.

The coming years will be a truly exciting time for strong lens science; this thesis aims to confront some of the challenges we will face.

Date of Award	2025
Original language	English
Awarding Institution	University of Oxford
Supervisor	Aprajita Verma (Supervisor)