Search

We present GIGA-Lens: a gradient-informed, GPU-accelerated Bayesian framework for modeling strong gravitational lensing systems, implemented in TensorFlow and JAX. The three components, optimization using multi-start gradient descent, posterior covariance estimation with variational inference, and sampling via Hamiltonian Monte Carlo, all take advantage of gradient information through automatic differentiation and massive parallelization on graphics processing units (GPUs). We test our pipeline on a large set of simulated systems and demonstrate in detail its high level of performance. The average time to model a single system on four Nvidia A100 GPUs is 105 seconds. The robustness, speed, and scalability offered by this framework make it possible to model the large number of strong lenses found in current surveys and present a very promising prospect for the modeling of $\mathcal{O}(10^5)$ lensing systems expected to be discovered in the era of the Vera C. Rubin Observatory, Euclid, and the Nancy Grace Roman Space Telescope., Comment: 23 pages, 13 figures, 2 tables. Submitted to ApJ

Constraints on the mass distribution in high-redshift clusters of galaxies are not currently very strong. We aim to constrain the mass profile, M(r), and dynamical status of the $z \sim 0.8$ LCDCS 0504 cluster of galaxies characterized by prominent giant gravitational arcs near its center. Our analysis is based on deep X-ray, optical, and infrared imaging, as well as optical spectroscopy. We model the mass distribution of the cluster with three different mass density profiles, whose parameters are constrained by the strong lensing features of the inner cluster region, by the X-ray emission from the intra-cluster medium, and by the kinematics of 71 cluster members. We obtain consistent M(r) determinations from three methods (dispersion-kurtosis, caustics and MAMPOSSt), out to the cluster virial radius and beyond. The mass profile inferred by the strong lensing analysis in the central cluster region is slightly above, but still consistent with, the kinematics estimate. On the other hand, the X-ray based M(r) is significantly below both the kinematics and strong lensing estimates. Theoretical predictions from $\Lambda$CDM cosmology for the concentration--mass relation are in agreement with our observational results, when taking into account the uncertainties in both the observational and theoretical estimates. There appears to be a central deficit in the intra-cluster gas mass fraction compared to nearby clusters. Despite the relaxed appearance of this cluster, the determinations of its mass profile by different probes show substantial discrepancies, the origin of which remains to be determined. The extension of a similar dynamical analysis to other clusters of the DAFT/FADA survey will allow to shed light on the possible systematics that affect the determination of mass profiles of high-z clusters, possibly related to our incomplete understanding of intracluster baryon physics., Comment: 13 pages, 14 figures, submitted to A&A, abstract shortened

In addition to detector fringing, the VIMOS IFU suffers from additional, not reproducible fringing affecting some pseudo-slits. It can severely decrease the sensitivity of the instrument. Its origin has not been clearly determined. Here, we present the analysis of calibration data obtained in October 2006. We show that fringing originates either from inside the mask at the pseudo-slits level or from an imperfect positioning of the masks themselves. The lack of reproducibility of the fringes may be caused by the imperfect fixation of the prisms in the pseudo-slits or a loose positioning of the masks.

Wireless LAN (WLAN) technologies such as WiFi or Bluetooth have become popular in many countries. The need for an efficient design of them becomes a central problem. Automatic transmitter placement provides cost savings when compared to the traditional human process of site planning. The performance of a LAN is strongly influenced by the access point's characteristics. That's why the amount of them, their location and their emission power have to be determined with care during the planning stage of the network. The planning problem can be exposed in different ways according to the optimisation goals that have been chosen. The traditional approach is to compute a coverage map that satisfies a simple constraint on the signal level based on a minimum threshold. Some more elaborated models introduce constraints of quality of service (as throughput or interference level) on the receiver locations. The evaluation of a solution (an access point configuration) relies on a propagation simulation that computes a coverage map. This simulation is often an expensive computational operation. The goal of the optimisation algorithm is to determine a solution that satisfies to the coverage constraints with a minimum number of propagation simulations. We propose a deterministic algorithm that relies on the multi-resolution Fourier domain parflow (FDPF) propagation model proposed by Gorce and Ubeda (2001). The originality of this algorithm is to exploit the multiresolution specificity of the FDPF method in order to reduce the computation cost of propagation simulations. Its aim is to find out an access point distribution that provides homogeneous coverage. We work on a two dimensional building floor representation and the variables are the x-axis and y-axis coordinates of the transmitters (access points). We consider that the number of sources to plan and their emission power is fixed. In the next part of this paper, a short overview of the planning problem is provided. In the third part, we define the problem we work on and the characteristics of the FDPF propagation model. In the fourth part, details about our deterministic algorithm are provided and in the fifth part, results are given. We end this paper by a short discussion and enlargement on our work.

In this paper a multi-resolution FDPF approach has been proposed. More precisely, a new nonregular pyramidal structure which fits the particular arrangement of the indoor environment has been presented. The use of a full-space discrete simulator instead of classical ray-tracing techniques is a challenge due to the inherent high computation requests. However, we demonstrate that the use of a multi-resolution approach allows one to perform some calculus in a pre-processing phase, independently of source locations.

The Hubble Frontier Fields (HFF) program combines the capabilities of the Hubble Space Telescope (HST) with the gravitational lensing of massive galaxy clusters to probe the distant Universe to an unprecedented depth. Here, we present the results of the first combined HST and Spitzer observations of the cluster Abell 2744. We combine the full near-infrared data with ancillary optical images to search for gravitationally lensed high-redshift (z > 6) galaxies. We report the detection of 15 I814-dropout candidates at z ~ 6-7 and one Y105-dropout at z ~ 8 in a total survey area of 1.43 arcmin^2 in the source plane. The predictions of our lens model allow us to also identify five multiply-imaged systems lying at redshifts between z ~ 6 and z ~ 8. Thanks to constraints from the mass distribution in the cluster, we were able to estimate the effective survey volume corrected for completeness and magnification effects. This was in turn used to estimate the rest-frame ultraviolet luminosity function (LF) at z ~ 6-8. Our LF results are generally in agreement with the most recent blank field estimates, confirming the feasibility of surveys through lensing clusters. Although based on a shallower observations than what will be achieved in the final dataset including the full ACS observations, the LF presented here extends down to Muv ~ -18.5 at z ~ 7 with one identified object at Muv ~ -15 thanks to the highly-magnified survey areas. This early study forecasts the power of using massive galaxy clusters as cosmic telescopes and its complementarity to blank fields., Comment: 9 pages, 7 figures, 3 tables. accepted for publication in ApJ