[Paper Review] BEAGLE-AGN I: Simultaneous constraints on the properties of gas in star-forming and AGN narrow-line regions in galaxies

A. Vidal-García, A. Plat+5 more|arXiv (Cornell University)|Nov 24, 2022

Astrophysical Phenomena and Observations4 citations

TL;DR

This paper extends the BEAGLE photoionization modeling framework to simultaneously constrain physical properties of gas in both star-forming H II regions and AGN-driven narrow-line regions (NLRs), demonstrating that degeneracies between ionization parameter and dust-to-metal mass ratio, and between ionizing spectrum slope and accretion luminosity, bias parameter recovery unless these are fixed. It shows that S/N(Hβ) ≥ 30 is required for unbiased NLR parameter retrieval in mixed systems, and validates models against 463 SDSS type-2 AGNs using [He ii]λ4686 to constrain ionization states.

ABSTRACT

We present the addition of nebular emission from the narrow-line regions (NLR) surrounding active galactic nuclei (AGN) to BEAGLE (BayEsian Analysis of GaLaxy sEds). Using a set of idealised spectra, we fit to a set of observables (emission-line ratios and fluxes) and test the retrieval of different physical parameters. We find that fitting to standard diagnostic-line ratios from Baldwin et al. (1981) plus [O II]3726,3729/[O III]5007, Hbeta/ Halpha, [O I]6300/[O II]3726,3729 and Halpha flux, degeneracies remain between dust-to-metal mass ratio and ionisation parameter in the NLR gas, and between slope of the ionizing radiation (characterising the emission from the accretion disc around the central black hole) and total accretion-disc luminosity. Since these degeneracies bias the retrieval of other parameters even at maximal signal-to-noise ratio (S/N), without additional observables, we suggest fixing the slope of the ionizing radiation and dust-to-metal mass ratios in both NLR and HII regions. We explore the S/N in Hbeta required for un-biased estimates of physical parameters, finding that S/N(Hbeta)~10 is sufficient to identify a NLR contribution, but that higher S/N is required for un-biased parameter retrieval (~20 for NLR-dominated systems, ~sim30 for objects with approximately-equal Hbeta contributions from NLR and HII regions). We also compare the predictions of our models for different line ratios to previously-published models and data. By adding [He II]4686-line measurements to a set of published line fluxes for a sample of 463 AGN NLR, we show that our models with $-4<$ionisation parameter in the NLR gas$<-1.5$ can account for the full range of observed AGN properties in the local Universe.

Motivation & Objective

• To extend the BEAGLE Bayesian fitting framework to include nebular emission from AGN narrow-line regions (NLRs), enabling simultaneous fitting of H II and NLR components.
• To identify and quantify degeneracies between key physical parameters—particularly ionization parameter and dust-to-metal mass ratio in NLRs, and spectral slope and accretion luminosity—underlying emission-line diagnostics.
• To determine the signal-to-noise ratio (S/N) threshold in Hβ required for unbiased recovery of NLR physical parameters in composite systems.
• To validate the extended BEAGLE-NLR model against a large sample of 463 type-2 AGNs from SDSS DR7 using [He ii]λ4686 flux measurements.
• To assess the performance of different photoionization models (F16, Pérez-Montero et al. 2019, Thomas et al. 2018a) in reproducing observed NLR line ratios and metallicity relations.

Proposed method

• Incorporated NLR emission models into the BEAGLE framework using ionization structure simulations from Hirschmann et al. (2019) and Feltre et al. (2016), with gas properties parameterized by ionization parameter, metallicity, dust-to-metal mass ratio, and ionizing spectrum slope.
• Generated idealized spectra for various combinations of NLR and H II region parameters, including [O ii]λ3726,λ3729 / [O iii]λ5007, Hβ / Hα, [O i]λ6300 / [O ii], and Hα flux as observables.
• Performed Bayesian inference using BEAGLE to retrieve posterior distributions of physical parameters from these idealized spectra, assessing bias and degeneracy structure.
• Defined S/N thresholds for Hβ by simulating signal degradation and measuring bias in recovered parameters, finding S/N(Hβ) ≈ 20–30 is needed for unbiased results.
• Measured [He ii]λ4686 fluxes in 463 type-2 AGNs from SDSS DR7 to compare with model predictions and constrain ionization parameter distributions.
• Constructed a grid of N/O abundance scaling factors to test model sensitivity to nitrogen abundance, and compared model predictions to empirical calibrations (Dors 2021) in 12+log(O/H)–P–R23 space.

Experimental results

Research questions

• RQ1 What degeneracies impair the simultaneous recovery of NLR physical parameters when fitting emission-line ratios and fluxes?
• RQ2 What minimum signal-to-noise ratio in Hβ is required to retrieve unbiased physical parameters in NLR-dominated and mixed NLR/H II systems?
• RQ3 Can the extended BEAGLE-NLR model reproduce the observed distribution of emission-line ratios in a large sample of 463 type-2 AGNs from SDSS?
• RQ4 How do different photoionization models (F16, Pérez-Montero et al. 2019, Thomas et al. 2018a) compare in fitting observed NLR properties, especially with respect to ionization parameter and metallicity?
• RQ5 To what extent does including [He ii]λ4686 improve constraints on the ionization state of NLR gas in observed AGNs?

Key findings

• Degeneracies between ionization parameter and dust-to-metal mass ratio in NLRs, and between ionizing spectrum slope and accretion luminosity, bias parameter recovery even at high S/N, necessitating prior fixing of these parameters.
• S/N(Hβ) ≈ 10 is sufficient to detect a NLR contribution, but S/N(Hβ) ≈ 20–30 is required for unbiased retrieval of physical parameters in NLR-dominated and mixed systems, respectively.
• The inclusion of [He ii]λ4686 flux measurements in 463 SDSS type-2 AGNs shows that models with −4 < log U_s^NLR < −1.5 can reproduce the full range of observed AGN properties in the local universe.
• F16 and Pérez-Montero et al. 2019 models sample the observed NLR parameter space well, while Thomas et al. 2018a models may underestimate log U_s^NLR due to fixed nitrogen abundances.
• Model predictions agree well with the Dors 2021 empirical calibration in the observed metallicity range, but models curve upward relative to the calibration at high metallicities, while observations scatter above the plane—suggesting models can capture real data dispersion.
• The difference between models with varying α_pl (ionizing spectrum slope) is small at low metallicities, which enhances the potential for constraining gas properties in high-redshift galaxies with JWST.