Unfolding the matter distribution using three-dimensional weak gravitational lensing

Combining redshift and galaxy shape information offers new exciting ways of exploiting the gravitational lensing effect for studying the large scales of the cosmos. One application is the three-dimensional (3D) reconstruction of the matter density distribution which is explored in this paper. We give a generalization of an already known minimum-variance estimator of the 3D matter density distribution that facilitates the combination of thin redshift slices of sources with samples of broad redshift distributions for an optimal reconstruction; sources can be given individual statistical weights. We show how, in principle, intrinsic alignments of source ellipticities or shear/intrinsic alignment correlations can be accommodated, albeit these effects are not the focus of this paper. We describe an efficient and fast way to implement the estimator on a contemporary desktop computer. Analytic estimates for the noise and biases in the reconstruction are given. Some regularization (Wiener filtering) of the estimator, adjustable by a tuning parameter, is necessary to increase the signal-to-noise ratio (S/N) to a sensible level and to suppress oscillations in radial direction. This, however, introduces as side effect a systematic shift and stretch of structures in radial direction. This bias can be expressed in terms of a radial point-spread function (PSF) comprising the limitations of the reconstruction due to given source shot noise and a lack of knowledge of the exact source redshifts. We conclude that a 3D mass-density reconstruction on galaxy cluster scales (similar to 1 Mpc) is feasible but, for foreseeable surveys, a map with a S/N greater than or similar to 3 threshold is limited to structures with M-200 greater than or similar to 1 x 10(14) or 7 x 10(14) M-circle dot h(-1), at low to moderate redshifts (z = 0.1 or 0.6). However, we find that a heavily smoothed full-sky map of the very large-scale density field may also be possible as the S/N of reconstructed modes increases towards larger scales. Future improvements of the method may be obtained by including higher order lensing information (flexion) which could also be implemented into our algorithm.