Non-parametric reconstruction of cluster mass distribution from strong lensing: modelling Abell 370

We describe a new non-parametric technique for reconstructing the mass distribution in galaxy clusters with strong lensing, i.e. from multiple images of background galaxies, The observed positions and redshifts of the images are considered as rigid constraints, and through the lens (ray-trace) equation they provide us with Linear constraint equations. These constraints confine the mass distribution to some allowed region, which is then found by linear programming. Within this allowed region we study in detail the mass distribution with minimum mass-to-light variation, and also some other distributions, such as the smoothest mass distribution. The method is applied to the extensively studied duster Abell 370, which hosts a giant luminous are and several other multiply imaged background galaxies. Our mass maps are constrained by the observed positions and redshifts (spectroscopic or model-inferred by previous authors) of the giant are and multiple-image systems, The reconstructed maps obtained for Abell 370 reveal a detailed mass distribution, with substructure quite different from the light distribution. The method predicts the bimodal nature of the cluster, and that the projected mass distribution is indeed elongated along the axis defined by the two dominant cD galaxies, However, the peaks in the mass distribution appear to be offset from the centres of the cDs. We also present an estimate fur the total mass of the central region of the cluster. This is in good agreement with previous mass determinations. The total mass of the central region is M=(2.0-2.7) x 10(14) M. h(50)(-1), depending on the solution chosen.