Beyond Kaiser bias: Mildly non-linear two-point statistics of densities in distant spheres

We present simple parameter-free analytic bias functions for the two-point correlation of densities in spheres at large separation. These bias functions generalize the so-called Kaiser bias to the mildly non-linear regime for arbitrary density contrasts and grow as b(ρ) − b(1) ∝ (1 − ρ⁻¹³/²¹)ρ¹ + n/³ with b(1) = −4/21 − n/3 for a power-law initial spectrum with index n. We carry out the derivation in the context of large-deviation statistics while relying on the spherical collapse model. We use a logarithmic transformation that provides a saddle-point approximation that is valid for the whole range of densities and show its accuracy against the 30 Gpc cube state-of-the-art Horizon Run 4 simulation. Special configurations of two concentric spheres that allow us to identify peaks are employed to obtain the conditional bias and a proxy for the BBKS extremum correlation functions. These analytic bias functions should be used jointly with extended perturbation theory to predict two-point clustering statistics as they capture the non-linear regime of structure formation at the per cent level down to scales of about 10 Mpc h⁻¹ at redshift 0. Conversely, the joint statistics also provide us with optimal dark matter two-point correlation estimates that can be applied either universally to all spheres or to a restricted set of biased (over- or underdense) pairs. Based on a simple fiducial survey, we show that the variance of this estimator is reduced by five times relative to the traditional sample estimator for the two-point function. Extracting more information from correlations of different types of objects should prove essential in the context of upcoming surveys like Euclid, DESI and WFIRST.