Investigating Dark Matter Models with Lyman-alpha forest

The aim of this project is to constrain dark matter models using Lyman-a forest flux spectra. A new set of the hydrodynamical Sherwood-Relics simulations, combined with the increased sensitivity of high redshift quasar spectra observed by HIRES and UVES spectrographs, allows to investigate cold-plus-warm dark matter (C+WDM) models by accessing the relevant small scales. In particular, the aim is to determine the thermal relic mass, m_wdm, and its relative abundance, given by F_wdm= ?_wdm/?_dm, from the suppression in the power spectrum.
To investigate this and other thermal and cosmological parameters affecting the power spectrum, we will initially use a standard technique based on an emulator that performs linear interpolation to predict the flux spectra between the available models. This technique further compares models against the data within a Bayesian framework, where Monte Carlo Markov Chain (MCMC) algorithm is used to infer the set of parameters that minimizes the likelihood.
After using the standard method to analyse these models against the high signal-to-noise ratio data, the aim is to investigate alternative techniques to carry out the interpolation and the inference, and to check whether these could provide tighter constraints or/and speed up the analysis. For instance, using a neural network interpolator or another machine learning-based approach, such as simulation-based inference, as well as exploring other sampling algorithms.
Overall, the goal of this PhD project is find novel constraints on C+WDM models using data that allows us to access the non-linear scales, and updating the standard Bayesian inference set-up with modern sampling algorithms and machine learning techniques.