We present measurements of the growth rate of cosmological structure from the modelling of the anisotropic galaxy clustering measured in the final data release of the VIPERS survey. The analysis is carried out in configuration space and based on measurements of the first two even multipole moments of the anisotropic galaxy auto-correlation function, in two redshift bins spanning the range 0.5 < z < 1.2. We provide robust and cosmology-independent corrections for the VIPERS angular selection function, allowing recovery of the underlying clustering amplitude at the percent level down to the Mpc scale. We discuss several improvements on the non-linear modelling of redshift-space distortions (RSD) and perform detailed tests of a variety of approaches against a set of realistic VIPERS-like mock realisations. This includes using novel fitting functions to describe the velocity divergence and density power spectra P θθ and P δθ that appear in RSD models. These tests show that we are able to measure the growth rate with negligible bias down to separations of 5 h −1 Mpc. Interestingly, the application to real data shows a weaker sensitivity to the details of non-linear RSD corrections compared to mock results. We obtain consistent values for the growth rate times the matter power spectrum normalisation parameter of f σ 8 = 0.55 ± 0.12 and 0.40 ± 0.11 at effective redshifts of z = 0.6 and z = 0.86 respectively. These results are in agreement with standard cosmology predictions assuming Einstein gravity in a ΛCDM background.

We present a joint likelihood analysis of the real-space power spectrum and bispectrum measured from a variety of halo and galaxy mock catalogs. A novel aspect of this work is the inclusion of nonlinear triangle configurations for the bispectrum, made possible by a complete next-to-leading order ("one-loop") description of galaxy bias, as is already common practice for the power spectrum. Based on the goodness of fit and the unbiasedness of the parameter posteriors, we accomplish a stringent validation of this model compared to the leading order ("tree-level") bispectrum. Using measurement uncertainties that correspond to an effective survey volume of 6 ðGpc=hÞ 3 , we determine that the one-loop corrections roughly double the applicable range of scales, from ∼0.17 h=Mpc (tree level) to ∼0.3 h=Mpc. This converts into a 1.5-2x improvement on constraints of the linear bias parameter at fixed cosmology, and a 1.5-2.4x shrinkage of uncertainties on the amplitude of fluctuations A s , which clearly demonstrates the benefit of extracting information from nonlinear scales despite having to marginalize over a larger number of bias parameters. Besides, our precise measurements of galaxy bias parameters up to fourth order allow for thorough comparisons to coevolution relations, showing excellent agreement for all contributions generated by the nonlocal action of gravity. Using these relations in the likelihood analysis does not compromise the model validity and is crucial for obtaining the quoted improvements on A s . We also analyzed the impact of higherderivative and scale-dependent stochastic terms, finding that for a subset of our tracers the former can boost the performance of the tree-level model with constraints on A s that are only slightly degraded compared to the one-loop model.