Beyond diagnostic-diagrams: A critical exploration on the classification of ionization processes

TL;DR

This work critically evaluates widely used optical emission-line diagnostics for identifying the dominant ionization sources in nearby galaxies, revealing substantial limitations of the classical BPT diagrams due to degeneracies with retired stellar populations and shocks. It introduces EW(Hα)-based diagnostics (WHaN, WHaD, WHaO) and a revised, cross-consistent scheme that significantly reduces misclassifications, achieving higher purity in separating star-forming, retired, and AGN-driven ionization. Using a large NMJ sample drawn from the NSA cross-matched with SDSS spectroscopy, and four independent multiwavelength AGN selections, the authors show that BPT-based classifications overestimate star formation and undercount AGNs, especially at low ionization levels, while EW(Hα)-based methods recover a much more physical separation. The proposed framework, which emphasizes EW(Hα) with complementary diagnostics, improves the reliability of population studies of galaxy ionization and supports a shift toward EW-based schemes for future work, potentially complemented by spatially resolved data to further unravel ionization sources.

Abstract

Optical emission-line diagnostic diagrams, such as the classical BPT, are widely used to identify ionisation mechanisms in galaxies but often suffer from degeneracies, especially when multiple sources coexist. We critically evaluate the effectiveness of these diagnostics in distinguishing star-forming galaxies, retired galaxies (RGs), and active galactic nuclei (AGNs), and propose refined methods to reduce misclassifications. Using a large sample of nearby galaxies from the NASA-Sloan Atlas cross-matched with SDSS spectroscopy, we define representative subsamples of late-type/star-forming galaxies, early-type/RGs, and multiwavelength-selected AGNs. Their distributions are analysed across traditional and modern diagnostics, including WHaN, WHaD, and the newly introduced WHaO diagram, which combine Hα equivalent width with [N II]/Hα, σ(Hα), and [O III]/[O II], respectively. Quantitative comparisons reveal that classical BPT diagrams overestimate star-forming galaxies by ~10% and misclassify up to 45% of AGNs and nearly all RGs. Diagnostics incorporating EW(Hα) achieve improved accuracy, reducing misclassifications to {\sim} 20\ % for AGNs and {\sim} 15\% for RGs. A revised classification scheme based on EW(Hα) thresholds and consistent WHaD/WHaO results yields the highest purity (8-25% misclassifications) and better reflects underlying physical conditions. Our analysis demonstrates that BPT-based methods fail to reliably separate ionisation mechanisms, particularly in galaxies hosting weak AGNs or evolved stellar populations. Updated EW(Hα)-based diagnostics offer a more accurate framework for studying galaxy ionisation and should replace traditional schemes in future population studies.

Figures (9)

• Figure 1: Comparison between the SFR provided by the MPA-JHU catalog and the one derived combining the infrared and ultraviolet photometry as described in the text. The black dots correspond to each galaxy in the NMJ sample (i.e., the full sample of galaxies analyzed in this article) and each successive grey contour represents the area encircling a 90%, 65% and a 15% of these points. The dashed-line represent the one-to-one relation.
• Figure 2: Distribution of the sub-samples of galaxies across different diagnostic diagram. Top panels: Classical BPT diagrams baldwin81, showing the distribution of [@series O iii O iii O iii]/H$\beta$ line ratio as a function of [@series N ii N ii N ii]/H$\alpha$ ratio ( left panel), [@series S ii S ii S ii]/H$\alpha$ ( middle panel) and [@series O i O i O i]/H$\alpha$ ( right panel). Solid and dot-dashed lines correspond to the demarcation lines proposed by K03 and K01 to distinguish between the different ionizing sources. Bottom panels: Diagrams comparing the distribution of Equivalent-width of H$\alpha$ (WH$\alpha$) as a function of (i) the [@series N ii N ii N ii]/H$\alpha$ ratio left panel, WHaN diagram cid-fernandes10, (ii) the H$\alpha$ velocity dispersion ($\sigma_{\rm H\alpha}$, middle panel), WHaD diagram whad, and (iii) the [@series O iii O iii O iii]/[@series O ii O ii O ii] line ratio ( right panel), proposed here as the WHaO diagram. In each panel the black dots correspond to the full NMJ sample and each successive grey contour represents the area encircling a 90%, 65% and a 15% of these points. The blue (red) contour represent the area that encircles 90% of the values corresponding to the late-type (early-type) subsamples of galaxies, as defined in the text. Finally, the location of the X-ray selected AGNs are shown as dark-blue stars. The D-parameter derived for a set of 2D KS-tests comparing the distributions of the different subsamples are included on top of each panel, using the nomenclature L/E when comparing LTGs vs. ETGs, E/A for ETGs vs. X-AGNs and L/A for LTGs vs. X-AGNs.
• Figure 3: Figure quantifying the differences found in the classification of the dominant ionization when using different diagnostics. From top-left to bottom-right, each panel comprises a heat-map showing the fraction of objects (color scale and values within each cell) assigned to each type of ionization by a different diagnostic diagram for a different sub-sample of galaxies, including late-type galaxies, early-type galaxies, X-ray selected AGNs (X-AGNs) , infrared selected AGNs (I-AGNs), UV-optically selected AGNs (O-AGNs), and radio selected AGNs (R-AGNs). Each heat-map columns correspond to the different ionizing types considered in this work, namely (i) ionization associated with recent star formation (SFGs), (ii) mixed or unknown ionization (Mix/Unk), (iii) ionziation usually found in non-starforming/retired galaxies (RGs), due to hot evolved stars binette94flor11, and/or low-velocity shocks dopita96 and (iv) ionization associated with AGNs and or shocks associated with galactic scale winds carlos20 (AGNs). On the other hand, each row corresponds to a different diagnostic scheme, including the use of (i) the classical diagram by baldwin81 that uses [@series O iii O iii O iii]/$\rm{H}\beta$ and [@series N ii N ii N ii]/$\rm{H}\alpha$ line ratios ( BPT-N2), (ii) the three diagrams by baldwin81 that use the [@series O iii O iii O iii]/$\rm{H}\beta$ vs. [@series N ii N ii N ii]/$\rm{H}\alpha$ [@series S ii S ii S ii]/$\rm{H}\alpha$ and [@series O i O i O i]/Ha line rations (BPT-all), (iii) the BPT-N2 diagram including a cut in the equivalent width of H$\alpha$ ( BPT-N2+WHa), as described in ARAA, (iv) the WHaN diagram that uses the [@series N ii N ii N ii]/$\rm{H}\alpha$ and the equivalent width of H$\alpha$ ( WHaN), (v) the diagram introduced by whad that uses [@series N ii N ii N ii]/$\rm{H}\alpha$ and the velocity dispersion of H$\alpha$ ( WHaD), (vi) the new proposed diagram that uses [@series O iii O iii O iii]/[@series O ii O ii O ii] and the equivalent width of H$\alpha$ ( WHaO), and three different combinations that use the WHaD and WHaO diagrams (vii) WHaDoO, (viii) WHaDoO and (ix) WHaD+O, described in the text.
• Figure 4: Fraction of objects assigned to each type of ionization by the different explored diagnostic schemes for the full NMJ sample analyzed along this study. Colors, labels and legends are the same as in Fig.\ref{['fig:heat']}.
• Figure 5: Distribution of the full sample of galaxies in the four set of properties used to select the candidates to AGNs employed in this study: Top-left panel: X-ray properties, showing the X-ray hardness ratio as a function of the X-ray luminosity. Top-right panel: infrared properties, showing the WISE $W1-W2$ color as a function of the WISE $W2$ magnitude. Bottom-left panel: UV-optical properties, showing $NUV-u$ color as a function of $u-g$ one. Bottom-right panel: radio properties, showing the ratio between the integrated and peak intensity at 1.4 GHz as a function of the integrated intensity. Each panel adopts the same symbols and color scheme: (i) solid circles correspond to the full sample of galaxies with measured properties, comprising 1390 objects for the X-ray panel, 541478 for the infrared one, 547928 for the UV-optical one, and 15839 for the radio one; (ii) contours represent the area that encircles 95% of the objects with ionization classified as star-forming (SFGs, blue), retired galaxies (RGs, red), and AGNs (purple) using out final classification scheme described in Sect. \ref{['sec:ana']}; (iii) dashed-lines show the demarcation lines described in Sect. \ref{['sec:ana']} to select the AGN candidates using the represented properties.