Systematics in lensing reconstruction: Dark matter rings in the sky?

TL;DR

This paper investigates the limitations of non-parametric lensing reconstruction methods, demonstrating how overfitting can produce artificial ring-like dark matter structures, with implications for interpreting observed rings in galaxy clusters.

Contribution

The study highlights how non-parametric lensing methods can generate spurious ring structures due to overfitting, affecting the interpretation of dark matter distributions in clusters.

Findings

Overfitting can produce artificial ring-like features in reconstructed maps.

Systematic effects in inversion algorithms can mimic real dark matter rings.

Mock data tests reveal limitations of current non-parametric methods.

Abstract

Non-parametric lensing methods are a useful way of reconstructing the lensing mass of a cluster without making assumptions about the way the mass is distributed in the cluster. These methods are particularly powerful in the case of galaxy clusters with a large number of constraints. The advantage of not assuming implicitly that the luminous matter follows the dark matter is particularly interesting in those cases where the cluster is in a non-relaxed dynamical state. On the other hand, non-parametric methods have several limitations that should be taken into account carefully. We explore some of these limitations and focus on their implications for the possible ring of dark matter around the galaxy cluster CL0024+17. We project three background galaxies through a mock cluster of known radial profile density and obtain a map for the arcs ($\theta$ map). We also calculate the shear field associated with the mock cluster across the whole field of view (3.3 arcmin). Combining the positions of the arcs and the two-direction shear, we perform an inversion of the lens equation using two separate methods, the biconjugate gradient, and the quadratic programming (QADP) to reconstruct the convergence map of the mock cluster. We explore the space of the solutions of the convergence map and compare the radial density profiles to the density profile of the mock cluster. When the inversion matrix algorithms are forced to find the exact solution, we encounter systematic effects resembling ring structures, that clearly depart from the original convergence map. Overfitting lensing data with a non-parametric method can produce ring-like structures similar to the alleged one in CL0024.