Speaker
Description
Constraining the Epoch of Reionization remains one of the pivotal tasks of modern cosmology, and next-generation telescopes are opening up the path to the first precision constraints on the timing of reionization derived from the Ly-alpha damping wing signature imprinted on the spectra of high-redshift quasars by the foreground neutral intergalactic medium (IGM). In the coming years, EUCLID will detect a number of high-redshift quasars never seen before, whose exquisite spectra collected by JWST are calling for powerful statistical methods to infer precision constraints on the IGM neutral hydrogen fraction as a function of redshift.
We developed a state-of-the-art Bayesian inference pipeline that allows us to disentangle the IGM damping wing from a quasar's unknown intrinsic spectrum and infer its lifetime as well as the neutral hydrogen (HI) column density in front it, directly translating into a constraint on the global IGM neutral fraction. We account for covariances across the full spectral range caused by IGM transmission fluctuations, quasar continuum reconstruction, and spectral noise. We discuss how simulation-based inference can be leveraged to overcome non-Gaussianities in the IGM transmission likelihood by training a normalizing flow as neural likelihood estimator. These developments are facilitated by our JAX-based inference pipeline, exploiting the latest machine learning (ML) infrastructure.
After marginalizing out nuisance parameters associated with the quasar continuum, we find that we can constrain the HI column density within the first 15 pMpc of each individual quasar to 0.1 - 0.9 dex and its lifetime to 0.2 - 1.0 dex. By applying our procedure to a set of mock observational spectra resembling the distribution of EUCLID quasars with realistic spectral noise, we show that our method applied to upcoming observational data can robustly constrain the evolution of the IGM neutral fraction at the < 5% level at all redshifts between 6 < z < 11.
