Artificial intelligence and quasar absorption system modelling; application to fundamental constants at high redshift

TL;DR

This paper introduces an AI-based automated method for modeling complex quasar absorption systems to improve measurements of fundamental constants at high redshift, emphasizing the importance of comprehensive modeling and large data samples.

Contribution

The paper presents a novel AI-driven, fully automated modeling system built on VPFIT for analyzing quasar spectra, reducing human bias and enhancing reproducibility.

Findings

Including thermal and turbulent components improves model accuracy.

Model non-uniqueness affects the precision of fundamental constant measurements.

Large sample sizes are essential for constraining variations in alpha.

Abstract

Exploring the possibility that fundamental constants of Nature might vary temporally or spatially constitutes one of the key science drivers for the European Southern Observatory's ESPRESSO spectrograph on the VLT and for the HIRES spectrograph on the ELT. High-resolution spectra of quasar absorption systems permit accurate measurements of fundamental constants out to high redshifts. The quality of new data demands completely objective and reproducible methods. We have developed a new fully automated Artificial Intelligence-based method capable of deriving optimal models of even the most complex absorption systems known. The AI structure is built around VPFIT, a well-developed and extensively-tested non-linear least-squares code. The new method forms a sophisticated parallelised system, eliminating human decision-making and hence bias. Here we describe the workings of such a system and apply it to synthetic spectra, in doing so establishing methods of importance for future analyses of VLT and ELT data. The results show that modelling line broadening for high-redshift absorption components should include both thermal and turbulent components. Failing to do so means it is easy to derive the wrong model and hence incorrect parameter estimates. We also argue that model non-uniqueness can be significant, such that it is not feasible to expect to derive an unambiguous estimate of the fine structure constant alpha from one or a small number of measurements. No matter how optimal the modelling method, it is a fundamental requirement to use a large sample of measurements to meaningfully constrain temporal or spatial alpha variation.