[en] We develop a purely mathematical tool to recover some of the information lost in the nonlinear collapse of large-scale structure. From a set of 141 simulations of dark matter density fields, we construct a nonlinear Wiener filter in order to separate Gaussian and non-Gaussian structure in wavelet space. We find that the non-Gaussian power is dominant at smaller scales, as expected from the theory of structure formation, while the Gaussian counterpart is damped by an order of magnitude on small scales. We find that it is possible to increase the Fisher information by a factor of three before reaching the translinear plateau, an effect comparable to other techniques like the linear reconstruction of the density field.

[en] We use the cosmo-OWLS suite of cosmological hydrodynamical simulations, which includes different galactic feedback models, to predict the cross-correlation signal between weak gravitational lensing and the thermal Sunyaev-Zeldovich (tSZ) y-parameter. The predictions are compared to the recent detection reported by van Waerbeke and collaborators. The simulations reproduce the weak lensing-tSZ cross-correlation, ξ_y_κ(θ), well. The uncertainty arising from different possible feedback models appears to be important on small scales only (0θ ∼< 1 arcmin), while the amplitude of the correlation on all scales is sensitive to cosmological parameters that control the growth rate of structure (such as σ_8, Ω_m and Ω_b). This study confirms our previous claim (in Ma et al.) that a significant proportion of the signal originates from the diffuse gas component in low-mass (M_h_a_l_o ∼< 10"1"4 M_⊙) clusters as well as from the region beyond the virial radius. We estimate that approximately 20% of the detected signal comes from low-mass clusters, which corresponds to about 30% of the baryon density of the Universe. The simulations also suggest that more than half of the baryons in the Universe are in the form of diffuse gas outside halos (∼> 5 times the virial radius) which is not hot or dense enough to produce a significant tSZ signal or be observed by X-ray experiments. Finally, we show that future high-resolution tSZ-lensing cross-correlation observations will serve as a powerful tool for discriminating between different galactic feedback models