Cosmic Voids in Generated N-body Simulations

A Generative Adversarial Network (GAN) was used to investigate the statistics and properties of voids in $\Lambda$CDM universes. A total of 15,000 2D images, extracted from N-body simulations, were used to train the GAN, which was then used to generate 15,000 novel 2D images. The total number of voids and the distribution of void sizes is similar in both sets of images. However, the generated images yield somewhat fewer small voids than do the simulated images. In addition, the generated images yield far fewer voids with central density contrast $\sim$ -1. Because the generated images yield fewer of the emptiest voids, the distribution of mean interior density contrast is systematically higher for the generated voids than it is for the simulated voids. The mean radial underdensity profiles of the largest voids are similar in both sets of images, but systematic differences are apparent. On small scales (r 0.5$r_v$), the underdensity profiles of the voids in the generated images exceed those of the voids in the simulated images. The discrepancies between the void properties in the two sets of images are likely attributable to systematic difficulties faced by deep learning techniques, such as neural networks struggling to capture absolute patterns in the data.