In searching for a robust statistical method to map large regions of almost empty space formed alongside the large-scale structure of our Universe, researchers of the Cosmostatistics Initiative (COIN) found inspiration in the field of criminology.
At very large scales, the matter in our Universe is distributed in filamentary structures, regions of high density holding galaxy clusters and superclusters. These structures live side by side with large regions of almost complete emptiness known as ‘voids’. This lattice structure can be easily recognized in both spectroscopic surveys and results from N-body simulations. Although detecting and studying high-density regions in these maps may be more appealing to the researcher from an astrophysical point of view, voids can also be quite informative from a cosmological perspective.
Given that the same underlying cosmology which generated large-scale structure is responsible for the formation of regions of low matter density, understanding the statistics and dynamics of cosmic voids can impose constraints on the parameters of cosmological models. This provides information on the evolution of matter arrangement throughout the history of the Universe as well as on the nature of dark energy, while avoiding the need to model complex gravitational effects. Even so, the detection of these voids, or their 2D projections called ‘troughs’, is a challenging task. The main difficulty being that there is no clear definition of what constitutes a void, or how much emptiness is enough to characterize them.
This data-intensive challenge is not very different from the one encountered in criminology when trying to map regions exhibiting high crime incidence. By analyzing the data of reported crimes, experts aim at identifying dangerous zones to optimize patrol routes along incident-dense regions (see Figure 1). During a few hot summer days in the shadow of Mont Blanc, the team of the 6th COIN Residency Program judged that there were enough similarities between these scenarios, as well as previous related work, to encourage an investigation of the method’s performance when faced with current astronomical data.
The publicly available implementation of an algorithm, called subspace-constrained mean shift, was adapted and optimized to estimate regions of high density in mass maps from the first-year data release of the Dark Energy Survey.
Moreover, the CRP6 research team performed tests of statistical consistency using simulations with different levels of noise, and compared results to previous research and alternative techniques, providing a clear picture of the method’s applicability. Results from both simulated and real data showed that the upgraded algorithm can consistently find connected filamentary structures that follow the same patterns of mass distribution (see Figure 2). The latter, in turn, enable the identification of troughs as underdense regions delimited by ridges. A desirable consequence of this approach is that the emerging troughs can have any shape, as opposed to methods that limit their morphology.
With the proposed tool, the team produced a catalog of troughs, which will be made publicly available for other groups to impose additional constraints on cosmological models.
Text by the CRP6 team
Reference: Moews et al., 2020 - MNRAS