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

Over 90% of dark matter haloes in cosmological simulations are unresolved. This hinders the dynamic range of simulations and also produces systematic biases when modelling cosmological tracers. Current methods cannot accurately preserve the multi-dimensional assembly bias found in simulations. Here we aim to enhance the unresolved structural and dynamic properties of haloes. We have developed HALOSCOPE, a machine learning technique using multi-variate conditional probability distribution functions given the input from haloes' local environment. In this work, we use HALOSCOPE to enhance the properties (concentration, spin and two shape parameters) of unresolved dark matter haloes in a low-resolution simulation. The algorithm trained on a high-resolution simulation allows to recover the multi-dimensional halo assembly bias, i.e. the correlations of different combinations of halo properties with the large-scale environment, in addition to the mean and distribution of the halo properties. This is achieved by including the linear halo-by-halo bias and tidal anisotropy in the set of input training parameters. We also study how the halo assembly bias produces galaxy assembly bias and how resolution effects can propagate errors into galaxy clustering. For this purpose, we have generated catalogues of central galaxies using two implementations of the assembly bias in a halo occupation distribution model. The clustering of central model galaxies is improved by a factor of three at $0.009<k (h{\rm Mpc^{-1}})<0.6$, when the unresolved haloes are enhanced with HALOSCOPE. The method developed here can preserve the multi-dimensional halo assembly bias, using the local environment of haloes and can also improve the accuracy of catalogues produced with approximate methods, when many realisations are needed.
Comment: 16 pages, 5 main figures