Data-driven approach for modeling the temporal and spectral evolution of kilonova systematic uncertainties

TL;DR

This paper introduces a data-driven method to model and mitigate time- and filter-dependent uncertainties in kilonova observations, improving parameter estimation and understanding of systematic errors.

Contribution

It presents a novel approach to account for systematic uncertainties in kilonova modeling, enhancing the accuracy of source parameter recovery from observational data.

Findings

Reliable recovery of ejecta mass consistent with previous studies

Systematic error below 1 magnitude between 1 to 5 days post-merger

Insights into temporal and spectral evolution of uncertainties

Abstract

Kilonovae, possible electromagnetic counterparts to neutron star mergers, provide important information about high-energy transient phenomena and, in principle, also allow us to obtain information about the source properties responsible for powering the kilonova. Unfortunately, numerous uncertainties exist in kilonova modeling that, at the current stage, hinder accurate predictions. Hence, one has to account for possible systematic modeling uncertainties when interpreting the observed transients. In this work, we provide a data-driven approach to account for time-dependent and filter-dependent uncertainties in kilonova models. Through a suite of tests, we find that the most reliable recovery of the source parameters and description of the observational data can be obtained through a combination of kilonova models with time- and filter-dependent systematic uncertainties. We apply our new method to analyze AT2017gfo. While recovering a total ejecta mass consistent with previous studies, our approach gives insights into the temporal and spectral evolution of the systematic uncertainties of this kilonova. We consistently find a systematic error below $1$ mag between $1$ to $5$ days after the merger. Our work addresses the need for early follow-up of kilonovae at earlier times, and improved modeling of the kilonova at later times, to reduce the uncertainties outside of this time window.