Precisely Measuring the Cosmic Reionization History from IGM Damping Wings Towards Quasars

TL;DR

This paper presents a Bayesian method to analyze quasar spectra, accurately measuring the IGM neutral fraction during reionization by modeling the entire spectrum and accounting for various sources of error.

Contribution

It introduces a generative probabilistic model and uses dimensionality reduction to improve the measurement of the IGM neutral fraction from quasar damping wings.

Findings

A six-parameter PCA model best describes quasar spectra.

The method reliably constrains the IGM neutral fraction and quasar lifetime.

The approach achieves high precision in modeling the reionization epoch.

Abstract

We introduce a new approach for analyzing the IGM damping wings imprinted on the proximity zones of quasars in the epoch of reionization (EoR). Whereas past work has typically forgone the additional constraining power afforded by the blue side continuum ($1216\,\r{A} \lesssim \lambda \lesssim 1280\,\r{A}$) and/or opted not to model the large correlated IGM transmission fluctuations in the proximity zone ($\lambda \lesssim 1216\,\r{A}$), we construct a generative probabilistic model for the entire spectrum accounting for all sources of error - the stochasticity induced by patchy reionization, the impact of the quasar's ionizing radiation on the IGM, the unknown intrinsic spectrum of the quasar, and spectral noise. This principled Bayesian method allows us to marginalize out nuisance parameters associated with the quasar's radiation and its unknown intrinsic spectrum to precisely measure the IGM neutral fraction, $\langle x_{\rm HI}\rangle$. A key element of our analysis is the use of dimensionality reduction (DR) to describe the intrinsic quasar spectrum via a small number of nuisance parameters. Using a large sample of $15,559$ SDSS/BOSS quasars at $z \gtrsim 2.15$ we trained and quantified the performance of six distinct DR methods, and find that a six parameter PCA model (five coefficients plus a normalization) performs best, with complex machine learning approaches providing no advantage. By conducting statistical inference on 100 realistic mock EoR quasar spectra, we demonstrate the reliability of the credibility contours that we obtain on $\langle x_{\rm HI}\rangle$ and the quasar lifetime, $t_{\rm Q}$. The new method introduced here will transform IGM damping wings into a precision probe of reionization, on the same solid methodological and statistical footing as other precision cosmological measurements.