Galaxies and haloes on graph neural networks: Deep generative modelling scalar and vector quantities for intrinsic alignment

• DOI: 10.1093/mnras/stac2083
• Publication Date: 2022-08-03T02:54:54Z
• Authors: Yesukhei Jagvaral; François Lanusse; Sukhdeep Singh; Rachel Mandelbaum; Siamak Ravanbakhsh; Duncan Campbell
• Keywords: Astrophysics of Galaxies (astro-ph.GA); 0103 physical sciences; FOS: Physical sciences; [SDU.ASTR.GA]Sciences of the Universe [physics]/Astrophysics [astro-ph]/Galactic Astrophysics [astro-ph.GA]; [SDU.ASTR.GA] Sciences of the Universe [physics]/Astrophysics [astro-ph]/Galactic Astrophysics [astro-ph.GA]; Astrophysics - Astrophysics of Galaxies; 01 natural sciences

ABSTRACT In order to prepare for the upcoming wide-field cosmological surveys, large simulations of Universe with realistic galaxy populations are required. particular, tendency galaxies naturally align towards overdensities, an effect called intrinsic alignments (IA), can be a major source systematics in weak lensing analysis. As details formation and evolution relevant IA cannot simulated practice on such volumes, we propose as alternative Deep Generative Model. This model is trained IllustrisTNG-100 simulation capable sampling orientations population so recover correct alignments. our approach, cosmic web set graphs, where graphs constructed each halo, signal those graphs. The generative implemented Adversarial Network architecture uses specifically designed Graph-Convolutional Networks sensitive relative 3D positions vertices. Given (sub)halo masses tidal fields, able learn predict scalar features dark matter subhalo shapes; more importantly, vector orientation axis ellipsoid complex 2D ellipticities. For correlations good quantitative agreement measured values from simulation, except at very small transition scales. ellipticities, all Additionally, capture dependence mass, morphological type, central/satellite type.