2009MNRAS.395.1319J

We propose a novel technique to refine the modelling of galaxy cluster mass distribution using gravitational lensing. The idea is to combine the strengths of both `parametric' and `non-parametric' methods to improve the quality of the fit. We develop a multiscale model that allows sharper contrast in regions of higher density where the number of constraints is generally higher. Our model consists of (i) a multiscale grid of radial basis functions with physically motivated profiles and (ii) a list of galaxy-scale potentials at the location of the cluster member galaxies. This arrangement of potentials of different sizes allows us to reach a high resolution for the model with a minimum number of parameters. We apply our model to the well-studied cluster Abell 1689. We estimate the quality of our mass reconstruction with a Bayesian Monte Carlo Markov Chain sampler. For a selected subset of multiple images, we manage to halve the errors between the positions of predicted and observed images compared to previous studies. This is due to the flexibility of multiscale models at intermediate scale between cluster and galaxy scale. The software developed for this paper is part of the public lenstool package which can be found at http://www.oamp.fr/cosmology/lenstool.