Inversion of the Lyman-α forest: 3D investigation of the intergalactic medium

We discuss the implementation of Bayesian inversion methods in order to recover the properties of the intergalactic medium from observations of the neutral hydrogen Lyman-$\alpha$ absorptions observed in the spectra of high redshift quasars (the so-called Lyman-$\alpha$ forest). We use two complementary schemes (i) a constrained Gaussian random field linear approach and (ii) a more general non-linear explicit Bayesian deconvolution method which offers in particular the possibility to constrain the parameters of the equation of state for the gas. While relying on prior assumption for the two-point correlation functions, we show how to recover, at least qualitatively, the 3D topology of the large scale structures in redshift space by inverting a suitable network of adjacent, low resolution lines of sight. We also discuss the inversion of single lines of sight observed at high spectral resolution. Our preliminary investigations suggest that the explicit Bayesian method can be used to derive quantitative information on the physical state of the gas. Redshift distortion is considered by simultaneous constrained reconstruction of the velocity and the density field in real space while assuming statistical correlation between the two fields. The method seems to work well in the strong prior regime where peculiar velocities are assumed to be the most likely realization in the density field. Finally, we investigate the effect of line of sight separation and number of lines of sight. Our analyses suggest that multiple low resolution lines of sight could be used to improve most likely velocity reconstruction on a high resolution line of sight.