Exploring New Redshift Indicators for Radio-Powerful AGN

TL;DR

This paper presents a machine learning model to efficiently estimate redshifts of radio-powerful AGN, enabling the detection of early universe SMBHs more rapidly than traditional methods.

Contribution

The study develops and applies a novel ML model for redshift prediction of AGN, achieving accuracy comparable to prior ML applications and facilitating large-scale early universe SMBH studies.

Findings

Median prediction error of 0.1162 for WISE-detected AGN

Outlier fraction of 11.58% at significant prediction deviations

Prediction error of 0.2501 when applied to Stripe 82 data

Abstract

Active Galactic Nuclei (AGN) are relevant sources of radiation that might have helped reionising the Universe during its early epochs. The super-massive black holes (SMBHs) they host helped accreting material and emitting large amounts of energy into the medium. Recent studies have shown that, for epochs earlier than $z~{\sim}~5$, the number density of SMBHs is on the order of few hundreds per square degree. Latest observations place this value below $300$ SMBHs at $z~{\gtrsim}~6$ for the full sky. To overcome this gap, it is necessary to detect large numbers of sources at the earliest epochs. Given the large areas needed to detect such quantities, using traditional redshift determination techniques -- spectroscopic and photometric redshift -- is no longer an efficient task. Machine Learning (ML) might help obtaining precise redshift for large samples in a fraction of the time used by other methods. We have developed and implemented an ML model which can predict redshift values for WISE-detected AGN in the HETDEX Spring Field. We obtained a median prediction error of $\sigma_{z}^{N} = 1.48 \times (z_{\mathrm{Predicted}} - z_{{\mathrm{True}}}) / (1 + z_{\mathrm{True}}) = 0.1162$ and an outlier fraction of $\eta = 11.58 \%$ at $(z_{\mathrm{Predicted}} - z_{{\mathrm{True}}}) / (1 + z_{\mathrm{True}}) > 0.15$, in line with previous applications of ML to AGN. We also applied the model to data from the Stripe 82 area obtaining a prediction error of $\sigma_{z}^{N} = 0.2501$.