Testing one-loop galaxy bias: Power spectrum

We test the regime of validity of the one-loop galaxy bias for a wide variety of biased tracers. Our most stringent test asks the bias model to simultaneously match the galaxy-galaxy and galaxy-mass spectrum, using the measured nonlinear matter spectrum from the simulations to test the one-loop effects from the bias expansion alone. In addition, we investigate the relevance of short-range nonlocality and halo exclusion through higher-derivative and scale-dependent noise terms, as well as the impact of using coevolution relations to reduce the number of free fitting parameters. From comparing the validity and merit of these assumptions, we find that a four-parameter model (linear, quadratic, cubic nonlocal bias, and constant shot noise) with a fixed quadratic tidal bias provides a robust modeling choice for the auto power spectrum of the less massive halos in our set of samples and their galaxy populations [up to k(max) = 0.35 h=Mpc for a sample volume of 6 (Gpc/h(3)]. For the more biased tracers, it is most beneficial to include scale-dependent noise. This is also the preferred option when considering combinations of the auto and cross power spectrum, which might be relevant in joint studies of galaxy clustering and weak lensing. We also test the use of perturbation theory to account for matter loops through gRPT, EFT, and the hybrid approach RESPRESSO. While all these have similar performance, we find the latter to be the best in terms of validity and recovered mean posterior values, in accordance with it being based partially on simulations.