Abstract:Accurate mock halo catalogues are indispensable data products for developing and validating cosmological inference pipelines. A major challenge in generating mock catalogues is modelling the halo or galaxy bias, which is the mapping from matter density to dark matter halos or observable galaxies. To this end, $N$-body codes produce state-of-the-art catalogues. However, generating large numbers of these $N$-body simulations for big volumes, especially if magnetohydrodynamics are included, requires significant computational time. We introduce and benchmark a differentiable and physics-informed neural network that can generate mock halo catalogues of comparable quality to those obtained from full $N$-body codes. The model design is computationally efficient for the training procedure and the production of large mock catalogue suites. We present a neural network, relying only on $18$ to $34$ trainable parameters, that produces halo catalogues from dark matter overdensity fields. The reduction in network weights was realised through incorporating symmetries motivated by first principles into our model architecture. We trained our model using dark-matter-only $N$-body simulations across different resolutions, redshifts, and mass bins. We validated the final mock catalogues by comparing them to $N$-body halo catalogues using different $N$-point correlation functions. Our model produces mock halo catalogues consistent with the reference simulations, showing that this novel network is a promising way to generate mock data for upcoming wide-field surveys due to its computational efficiency. Moreover, we find that the network can be trained on approximate overdensity fields to reduce the computational cost further. We also present how the trained network parameters can be interpreted to give insights into the physics of structure formation. Finally, we discuss the current limitations of our model as well as more general requirements and pitfalls of approximate halo mock generation that became evident from this study.

FLORAH: A generative model for halo assembly histories

FLORAH: a generative model for halo assembly histories

First Light and Reionisation Epoch Simulations (FLARES) XVII: Learning the galaxy-halo connection at high redshifts

Predicting dark matter halo masses from simulated galaxy images and environments

Generating Dark Matter Halo Merger Trees

A Hierarchy of Normalizing Flows for Modelling the Galaxy-Halo Relationship

Characterizing ultra-high-redshift dark matter halo demographics and assembly histories with the gureft simulations

Evaluating the galaxy formation histories predicted by a neural network in pure dark matter simulations

Teaching dark matter simulations to speak the halo language

Exploring the halo-galaxy connection with probabilistic approaches

CHARM: Creating Halos with Auto-Regressive Multi-stage networks

Leaving No Branches Behind: Predicting Baryonic Properties of Galaxies from Merger Trees

FLAMINGO: Calibrating large cosmological hydrodynamical simulations with machine learning

Characterising ultra-high-redshift dark matter halo demographics and assembly histories with the GUREFT simulations

The multi-dimensional halo assembly bias can be preserved when enhancing halo properties with HALOSCOPE

MAHGIC: a Model Adapter for the Halo–Galaxy Inter-Connection

How to Empirically Model Star Formation in Dark Matter Haloes – I. Inferences about Central Galaxies from Numerical Simulations

Universal Merger Histories of Dark-Matter Haloes

PineTree : A generative, fast, and differentiable halo model for wide-field galaxy surveys

A conditional abundance matching method of extending simulated halo merger trees to resolve low-mass progenitors and subhalos

A Conditional Abundance Matching Method of Extending Simulated Halo Merger Trees to Resolve Low-Mass Progenitors and Sub-halos