Publications related to Gravitational lens

Methods for strong gravitational lens detection and analysis using machine learning and high performance computing

The study of strong gravitational lenses is a relatively new scientific field in astronomy with many applications in cosmology. Its unique observables allow astronomers to trace dark matter, determine the expansion of the universe and study galaxy evolution. While strong lenses are relatively rare phenomena, recent and future improvements in observational instrumentation are expected to lead to a sharp increase in the number and quality of their observation. With that increase comes however an observational challenge from the sheer quantity of data the lenses are hidden in. Current classification and modelling techniques are simply not capable of handling the new data in a reasonable timeframe. As a consequence this thesis concerns itself with the development and improvement of numerical methods necessary to find and analyse these newly found lenses, borrowing from deep learning and high performance computing techniques. Deep learning, a field of machine learning, shows significant success in finding lenses in simulations. Performing significantly better than any other classification method, CNN-based lens finders reach almost the required classification efficiency and accuracy for a fully automated Euclid strong lensing pipeline. Its dependency on a varied and complete training set and the difficulty creating it make its immediate application to new surveys however more difficult. Incorporating high performance computing techniques and more concurrent algorithm design into lens analysis results in crucial speed-ups for strong lens analysis. Applied to Lenstool, a popular mass modelling tool for gravitational lenses, the resulting software can be run on modern supercomputers, using Graphics Processing Units (GPU) and CPUs simultaneously. Running it on state-of-the-art hardware makes it possible to reduce computation time to an acceptable level when mass modelling complex cluster lenses.

EPFL2020

Advanced gravitational lensing techniques for precision cosmology

Markus Rexroth

Gravitational lensing - the deflection of light by gravity - has greatly developed since its famous first observation in 1919, which validated Einstein's General Theory of Relativity. The strength of this effect does not depend on the nature of the mass which produces the gravitational field and thus it is a great tool to weigh both the visible and the invisible parts of the universe. Consequently gravitational lensing has become a pillar of observational cosmology over the last decades, and it is used to study the nature of Dark Matter and Dark Energy, the two mysterious quantities which dominate our universe, but are not yet understood by physics theory. The success of this endeavor rests on a thorough understanding of lensing theory and observations, including their systematics, the availability of a sufficient amount of precise data, and the development of efficient software to precisely measure these lensing effects in digital astronomical images. This thesis presents advanced techniques which can improve several of these areas. In the first part, we develop a new theoretical method to break the mass-sheet degeneracy, which prevents accurate mass determinations from lensing observations. The second part focuses on spectroscopic data from MUSE, a second-generation Integral-Field Spectrograph installed on one of the largest ground-based telescopes on earth. We present a pipeline which permits the efficient determination of the redshift of a source observed by MUSE. The redshift indicates the distance of the source from us and depends on the expansion of the universe. In addition, we use MUSE observations of a galaxy cluster to improve the determination of its total weight, including the dominant Dark Matter component. In the third part of this thesis, we investigate how we can accelerate the computation of these mass maps. In the era of big data and large surveys, computing efficiency is key to obtaining new scientific insights. We use High Performance Computing techniques like graphics card acceleration to improve the code performance and we develop a method which harnesses extra performance from using single precision without loosing the required accuracy. In the last section, we present first results from measuring flexion, a higher order lensing effect which could substantially increase the resolution of lensing mass maps and thus lead to a sharper view of structure in the universe.

EPFL2018

Gravitational lenses

Christel Vuissoz

The Hubble constant H0 is one of the most important parameters in cosmology, as it encodes the age of the Universe and is necessary for any distance determination at a cosmological scale. It is, however, only poorly constrained by traditional methods. The current favored value, H0 = 72±8 km s-1 Mpc-1, is provided by the HST Hubble constant Key Project (Freedman et al. 2001), which combines several Cepheid-calibrated distance indicators. This roughly 10% error nevertheless denotes only the statistical uncertainty in the determination of H0, while the possible systematical errors in the first step of the distance ladder (the distance to the Large Magellanic Cloud) may be of the same order of magnitude. Time delays between gravitationally lensed images of distant quasars can yield a more precise measurement of the Hubble constant, on a truly cosmic scale, and independently of any local distance calibrator. At the beginning of this thesis, time delays had been measured in only ten lensed systems, nine of which gave H0 estimates. However before 2004, no concerted and long term action has succeeded to apply the time delay method at a level of precision really competitive with other techniques. The major difficulties arise from the modeling of the lens mass distribution, and from the uncertainty on the time delay measurement itself, which was typically of about 10% in past monitoring programs. COSMOGRAIL (COSmological MOnitoring of GRAvItational Lenses) is an international collaboration initiated in April 2004 at the Laboratory of Astrophysics of EPFL, and which aims at measuring precise time delays for most known lensed quasars, in order to determine the Hubble constant down to an uncertainty of a few percent. This thesis took place at the beginning of COSMOGRAIL and consisted in setting up this large photometric monitoring. It addressed both issues of carrying out accurate photometry of faint blended sources and of obtaining well sampled light curves, in order to measure precise time delays. As part of the COSMOGRAIL project, I have been managing the monitoring of over twenty gravitationally lensed quasars with the 1-2m telescopes involved both in the Northern and Southern hemispheres, and organizing the data. The first crucial work of this thesis was then to develop an automated reduction pipeline able to produce an homogeneous data set from images acquired with very different telescopes. This pipeline was also needed to perform aperture photometry of all lensed quasars, in order to study their variability and define the monitoring priorities. The powerful MCS deconvolution algorithm (Magain, Courbin, & Sohy 1998) was greatly used in this work and allowed to highly improve the image resolution, with the aim of obtaining accurate photometric measurements of the individual quasar lensed images. I have finally tested and improved three different numerical techniques to determine time delays between the quasar components from their light curves. In this thesis, time delays have been determined in four systems. The first one was measured in the doubly imaged quasar SDSS J1650+4251, after two years of monitoring with the 1.5m telescope of Maidanak Observatory, in Uzbekistan. The quadruply lensed system RXS J1131–1231 was then studied and three time delays determined from 3-year observations with the Swiss Euler 1.2m telescope located at La Silla, in Chile. The photometric monitoring of the quadruple WFI J2033–4723 was also carried out with the Euler telescope, and data were then merged with those obtained by a second monitoring group, with the SMARTS 1.3m telescope at the Cerro Tololo Interamerican Observatory (CTIO), also located in Chile. Two time delays were measured in this system, after three years of observations, the close pair A1 – A2 remaining unresolved. Three time delays were determined in the quadruply imaged quasar HE0435–1223, after four years of optical monitoring with Euler, Mercator and Maidanak telescopes, to which photometric measurements by SMARTS 1.3m telescope were added. Euler and SMARTS merged data for the doubly imaged quasar QJ0158–4325 were also analysed, the size of the source accretion disk was measured, but we failed to determine a time delay due to the high amplitude of the microlensing variability in this system. The accuracies on time delay measurements reached in this thesis are of the order of 3-4% and show a clear improvement from the typical 10% uncertainties of past monitoring programs. These results were finally converted into estimates of the Hubble constant following different models of the lensing mass potential. The H0 mean value obtained when considering the individual determinations from twelve gravitationally lensed quasars with known time delays is H0 = 60 ± 7 km s-1 Mpc-1. This result is consistent with the current favored value, and above all promising, as including additional systems to this ensemble will surely provide tighter bounds on H0. In conclusion, the increasing number of time delay measurements and improvements in lens modeling should reduce the errors on the Hubble constant estimate provided by gravitational lensing. Conversely, the determination of more time delays should put further constraints on lens galaxy density profiles when using a prior on H0 from other studies.

EPFL2008

High Performance Computing for gravitational lens modeling: Single vs double precision on GPUs and CPUs

Jean-Paul Richard Kneib, Markus Rexroth

Strong gravitational lensing is a powerful probe of cosmology and the dark matter distribution. Efficient lensing software is already a necessity to fully use its potential and the performance demands will only increase with the upcoming generation of telescopes. In this paper, we present a proof-of-concept study on the impact of High Performance Computing techniques on a performance-critical part of the widely used lens modeling software LENSTOOL. We implement the algorithm once as a highly optimized CPU version and once with graphics card acceleration for a simple parametric lens model. In addition, we study the impact of finite machine precision on the lensing algorithm. While double precision is the default choice for scientific applications, we find that single precision can be sufficiently accurate for our purposes and lead to a big speedup. Therefore we develop and present a mixed precision algorithm which only uses double precision when necessary. We measure the performance of the different implementations and find that the use of High Performance Computing Techniques dramatically improves the code performance both on CPUs and GPUs. Compared to the current LENSTOOL implementation on 12 CPU cores, we obtain speedup factors of up to 170. We achieve this optimal performance by using our mixed precision algorithm on a high-end GPU which is common in modern supercomputers. We also show that these techniques reduce the energy consumption by up to 98%. Furthermore, we demonstrate that a highly competitive speedup can be reached with consumer GPUs. While they are an order of magnitude cheaper than the high-end graphics cards, they are rarely used for scientific computations due to their low double precision performance. However, our mixed precision algorithm unlocks their full potential. Consequently, the consumer GPU delivers a speedup which is only a factor of four lower than the best speedup achieved by a high-end GPU. (C) 2019 Elsevier B.V. All rights reserved.

ELSEVIER2020

Sparse Linear Inversion for Strong Gravitational Lenses Reconstruction, Component Separation from Morphological Component Analysis with Multiple Discrimination Criteria

Rémy Elie Joseph

Strong gravitational is a natural phenomenon that produces distorted images of distant galaxies due to massive galaxies or clusters, called deflectors or lenses, lying along the line of sight. Due to the mass of the deflector, and as a result of general relativity, the light coming from galaxies at the background follows a path that appears to us as bent, thus forming magnified and sheared images. The spectacular images of gravitational lens systems give us information about a wealth of cosmological and physical processes, from the expansion of the Universe to the nature of dark matter. The study of strong gravitational lenses requires being able to extract information from images of such systems. When faced with this task, many problems arise. First, there is the problem of finding them. So far, only a few hundreds of strong gravitational lens systems are known, but future surveys are expected to bring hundreds of thousands new systems to be found amongst billions of light sources. The second problem is to be able to identify lensed features and separate them from the light profile of the foreground deflector. The third problem is to be able to reconstruct the image of a lensed background source as if it had been unaffected by lensing. The last problem is the reconstruction of the deflector's mass distribution. In practice, the last two problems have to be solved as one. Recovering the light distribution of the source requires being able to invert the distortion by the gravitational lens, which depends on its mass distribution. In turn, the lens mass distribution is constrained by the distortions applied to the source's image. In this thesis I propose to address these problems with the tools provided by sparse inverse problem solving with a particular emphasis on the problem of separating lens and source luminosity profiles. The problem of separating two overlapping light profiles is known as deblending. I introduce three new techniques for the deblending of strong gravitational lenses that exploit different properties of lens systems and that each contribute to solving the problems of strong lens finding and modelling. The first technique allows the general subtraction of galaxies light profiles in large survey in order to facilitate the search for strong gravitational lenses. The second technique allows to separate lens and source light profiles based on their difference in colour and the sparsity of a galaxy's light profile in the wavelet. This technique allows in particular to remove most of the light from foreground galaxy clusters, which facilitates the identification of lensed galaxies in view of the modelling of the cluster's mass distribution. Finally, I show that lensing itself can be used to separate the unperturbed light profile of a lens galaxy from the distorted profile of a lensed source. This technique involves the joint reconstruction of the source in its own referential based again on sparsity. At this stage, the joint reconstruction and separation is performed at fixed mass model, but this work paves the way for the development of a free form lens mass distribution reconstruction technique in the near future, based on the combination of these separation and reconstruction techniques.

EPFL2018

Galaxy Scale Lenses In The Rcs2. I. First Catalog Of Candidate Strong Lenses

We present the first galaxy scale lens catalog from the second Red-Sequence Cluster Survey. The catalog contains 60 lensing system candidates comprised of Luminous Red Galaxy (LRG) lenses at 0.2 less than or similar to z less than or similar to 0.5 surrounded by blue arcs or apparent multiple images of background sources. The catalog is a valuable complement to previous galaxy-galaxy lens catalogs as it samples an intermediate lens redshift range and is composed of bright sources and lenses that allow easy follow-up for detailed analysis. Mass and mass-to-light ratio estimates reveal that the lens galaxies are massive ((M) over bar approximate to 5.5 x 10(11) [M-circle dot h(-1)]) and rich in dark matter ((M/L) over bar approximate to 14 [M-circle dot/L-circle dot, B-h]). Even though a slight increasing trend in the mass-to-light ratio is observed from z = 0.2 to z = 0.5, current redshift and light profile measurements do not allow stringent constraints on the mass-to-light ratio evolution of LRGs.

2012

Cosmology with Gravitational Lensing

Malte Tewes

EPFL2013

,

In the context of gravitational lensing, the density profile of lensing galaxies is often considered to be perfectly elliptical. Potential angular structures are generally ignored, except to explain flux ratios of point-like sources (i.e. flux ratio anomalies). Surprisingly, the impact of azimuthal structures on extended images of the source has not been characterised, nor has its impact on the H-0 inference. We address this task by creating mock images of a point source embedded in an extended source and lensed by an elliptical galaxy on which multipolar components are added to emulate boxy or discy isodensity contours. Modelling such images with a density profile free of angular structure allows us to explore the detectability of image deformation induced by the multipoles in the residual frame. Multipole deformations are almost always detectable for our highest signal-to-noise ratio (S/N) mock data. However, the detectability depends on the lens ellipticity and Einstein radius, on the S/N of the data, and on the specific lens modelling strategy. Multipoles also introduce small changes to the time-delays. We therefore quantify how undetected multipoles would impact H-0 inference. When no multipoles are detected in the residuals, the impact on H-0 for a given lens is in general less than a few km s(-1) Mpc(-1), but in the worstcase scenario, combining low S/N in the ring and large intrinsic boxyness or discyness, the bias on H-0 can reach 10-12 km s(-1) Mpc(-1). If we now look at the inference on H-0 from a population of lensing galaxies with a distribution of multipoles representative of what is found in the light profile of elliptical galaxies, we find a systematic bias on H-0 of less than 1%. A comparison of our mock systems to the state-of-the-art time-delay lens sample studied by the H0LiCOW and TDCOSMO collaborations indicates that multipoles are currently unlikely to be a source of substantial systematic bias on the inferred value of H-0 from time-delay lenses.

EDP SCIENCES S A2022

The purpose of this Thesis is to develop, test, and characterize different models attempting to tackle the problem of measurement of galaxy shapes applied in interferometric observations. Shape measurement is a tool for estimating the underlying shear due to weak gravitational lensing by large-scale foreground matter distributions. The recent enormous progress in radio interferometric imaging during the last years, with projects such as MeerKAT, ASKAP, and in future SKA, motivates the development of advanced algorithms that utilize radio observations for this task. In most of these algorithms, the main disadvantage is that the proposed models' parameters have to be estimated in advance, assuming certain prior knowledge on the form of the objects. Our motivation is to develop and evaluate frameworks that do not require significant prior information on this form to make accurate measurements.To achieve this goal, we move in two implementation directions. In the first one, we are based on an existing approach that applies an object decomposition procedure using a dictionary of shapelet functions. Our proposal goes some steps forward trying to estimate this decomposition's key size parameter from a model built on advanced regularization. More specifically, we create an over-redundant dictionary formed as the concatenation of several groups of orthogonal shapelet basis functions with different parameters, and we use structured sparsity penalties to estimate the size parameter of the best group. As an alternative development strategy, we form an algorithm that employs multi-resolution least-squares analysis, which attempts to identify the size value that minimizes the relative residual in the fitting.The second path, instead of making the shape estimation directly on radio interferometric data, performs image reconstruction from the visibilities and measures the ellipticity of the objects in the resulting images. For this purpose, we adopt a sparsity averaging analysis algorithm that has been developed in the past and restores with high precision the intensity image from the visibilities. We also implement a similar framework that uses the CLEAN algorithm for image reconstruction, which helps compare the results between the two options and evaluate the quality of the measurements achieved.All our models are tested using a collection of simulated data that include objects of different profile types, ellipticity, size, orientation, and position in the field of view. The visibilities generation is done using either a simulated Gaussian profile coverage or a realistic SKA-like one. Additionally, we present an initial study of the same algorithms on real objects from the COSMOS survey.

EPFL2021

A 7 deg(2) survey for galaxy-scale gravitational lenses with the HST imaging archive(star)

Frédéric Courbin, Georges Meylan, Ravindra Pawase

We present the results of a visual search for galaxy-scale gravitational lenses in similar to 7 deg(2) of Hubble Space Telescope (HST) images. The data set comprises the whole imaging data ever taken with the Advanced Camera for Surveys (ACS) in the filter F814W (I-band) up to 2011 August 31, i.e. 6.03 deg(2) excluding the field of the Cosmic Evolution Survey which has been the subject of a separate visual search. In addition, we have searched for lenses in the whole Wide Field Camera 3 (WFC3) near-IR imaging data set in all filters (1.01 deg(2)) up to the same date. Our primary goal is to provide a sample of lenses with a broad range of different morphologies and lens-source brightness contrast in order to estimate a lower limit to the number of galaxy-scale strong lenses in the future Euclid survey in its VIS band. Our criteria to select lenses are purely morphological as we do not use any colour or redshift information. The final candidate selection is very conservative hence leading to a nearly pure but incomplete sample. We find 49 new lens candidates: 40 in the ACS images and 9 in the WFC3 images. Out of these, 16 candidates are secure lenses owing to their striking morphology, 21 more are very good candidates and 12 more have morphologies compatible with gravitational lensing but also compatible with other astrophysical objects such as ring and chain galaxies or mergers. Interestingly, some lens galaxies include low surface brightness galaxies, compact groups and mergers. The imaging data set is heterogeneous in depth and spans a broad range of galactic latitudes. It is therefore insensitive to cosmic variance and allows us to estimate the number of galaxy-scale strong lenses on the sky for a putative survey depth, which is the main result of this work. Because of the incompleteness of the sample, the estimated lensing rates should be taken as lower limits. Using these, we anticipate that a 15 000 deg(2) space survey such as Euclid will find at least 60 000 galaxy-scale strong lenses down to a limiting AB magnitude of I = 24.5 (10 Sigma) or I = 25.8 (3 Sigma).

Wiley-Blackwell2014