Breeding gravitational lenses

Abstract

Gravitational lenses are a spectacular astrophysical phenomenon, a cosmic mirage caused by the gravitational deflection of light in which multiple images of a same background object can be seen. Their beauty is only exceeded by their usefulness, as the gravitational lens effect is a direct probe of the total mass of the deflecting object. Furthermore, since the image configuration arising from the gravitational lens effect depends on the exact gravitational potential of the deflector, it even holds the promise of learning about the distribution of the mass. In this presentation, a method for extracting the information encoded in the images and reconstructing the mass distribution is presented. Being a non‐parametric method, it avoids making a priori assumptions about the shape of the mass distribution. At the core of the procedure lies a genetic algorithm, an optimization strategy inspired by Darwin's principle of “survival of the fittest”. One only needs to specify a criterion to decide if one particular trial solution is deemed better than another, and the genetic algorithm will “breed” appropriate solutions to the problem. In a similar way, one can create a multi‐objective genetic algorithm, capable of optimizing several fitness criteria at the same time. This provides a very flexible way to incorporate all the available information in the gravitational lens system: not only the positions and shapes of the multiple images are used, but also the so‐called “null space”, i.e. the area in which no such images can be seen. The effectiveness of this approach is illustrated using simulated data, which allows one to compare the reconstruction to the true mass distribution.