Deep Learning the Morphology of Dark Matter Substructure

TL;DR

This paper demonstrates that deep learning, specifically convolutional neural networks, can reliably classify different types of dark matter substructure in simulated strong lensing images, aiding future dark matter research.

Contribution

First application of deep neural networks to classify dark matter substructure in simulated strong lensing images, showing high potential for future dark matter identification.

Findings

Neural networks can distinguish dark matter substructure types

Simulated images enable effective training of classification models

Deep learning will be crucial with upcoming large datasets

Abstract

Strong gravitational lensing is a promising probe of the substructure of dark matter halos. Deep learning methods have the potential to accurately identify images containing substructure, and differentiate WIMP dark matter from other well motivated models, including vortex substructure of dark matter condensates and superfluids. This is crucial in future efforts to identify the true nature of dark matter. We implement, for the first time, a classification approach to identifying dark matter substructure based on simulated strong lensing images with different substructure. Utilizing convolutional neural networks trained on sets of simulated images, we demonstrate the feasibility of deep neural networks to reliably distinguish among different types of dark matter substructure. With thousands of strong lensing images anticipated with the coming launch of LSST, we expect that supervised and unsupervised deep learning models will play a crucial role in determining the nature of dark matter.