OB-I Diagram in Galaxy Classification

• OB-I diagram is an optical diagnostic that uses log(EW(Hβ)) and log([O III]/Hβ) to classify galaxies into AGN-dominated, mixed, and pure star-forming systems.
• It employs only two strong, nearby emission lines to minimize dust sensitivity and overcome redshift limitations of traditional diagnostics like BPT.
• Empirical and semi-empirical boundaries validated across surveys show OB-I maintains robust classification up to z≈2.7 with clear separation of galaxy types.

The OB-I diagram is a two-parameter optical diagnostic for galaxy classification that compares the rest-frame equivalent width of H$\beta$ with the emission-line ratio $[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$. In the formulation introduced for galaxy spectroscopy, it is designed for use from the Local Universe to the “Cosmic Noon,” especially at $1.5 \lesssim z \lesssim 2.5$, where $\mathrm{H}\alpha$ and $[\mathrm{N\,II}]$ often move beyond ground-based optical coverage. Because it uses only two nearby, strong lines, the diagram complements BPT-like diagnostics while reducing sensitivity to dust and minimizing line-availability constraints (Santos et al., 23 Jul 2025).

1. Definition and diagnostic role

In its galaxy-classification sense, the OB-I diagram uses two observables derived from rest-frame optical spectra. The abscissa is the rest-frame equivalent width of H$\beta$, plotted as $\log_{10}(\mathrm{EW}(\mathrm{H}\beta))$, with EW in Å. The ordinate is the flux ratio $[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$, plotted as $\log_{10}([\mathrm{O\,III}]/\mathrm{H}\beta)$ (Santos et al., 23 Jul 2025).

The name “OB-I” is read “Oh-Bee-One” and reflects the use of ionised oxygen, $[\mathrm{O\,III}]$, and hydrogen beta, H$[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$0. The diagnostic was developed as a complementary optical classification scheme for “intermediate” redshifts, where traditional BPT classifications become difficult because the crucial $[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$1 emission line is out of range of ground-based optical spectographs. Its intended function is not to replace all multi-line diagnostics, but to provide a practical two-line scheme that can separate galaxies with a dominating AGN component in their emission from other systems across a wide redshift interval (Santos et al., 23 Jul 2025).

The paper’s central empirical claim is that, at $[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$2, the OB-I diagram clearly separates galaxies between two distinct types: one dominated by AGN and a second made up of a mixed population of SF galaxies and AGN activity. It further proposes a secondary division that partially separates this mixed population from a pure SF population. At higher redshifts, the majority of AGNs identified by other classification schemes are correctly recovered by the OB-I diagram, and the empirical AGN division requires no significant adjustment up to $[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$3 (Santos et al., 23 Jul 2025).

2. Axes, measurement conventions, and observational inputs

The OB-I diagram is defined from two measured quantities. The first is the rest-frame equivalent width of H$[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$4,

$[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$5

where $[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$6 is the integrated line flux and $[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$7 is the continuum flux density at the line. In the implementation described in the paper, $[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$8 is enforced for use on the x-axis, because some catalogues define emission as negative EW and the diagnostic uses $[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$9 (Santos et al., 23 Jul 2025).

The second observable is the line ratio

$1.5 \lesssim z \lesssim 2.5$0

with the y-axis defined as

$1.5 \lesssim z \lesssim 2.5$1

These two lines are close in wavelength, so the ratio is minimally affected by internal dust. The paper therefore notes that Galactic extinction should be corrected, while no additional Balmer-decrement correction is applied specifically to $1.5 \lesssim z \lesssim 2.5$2 for OB-I in the SDSS implementation (Santos et al., 23 Jul 2025).

The measurement strategy varies by survey but follows a common principle: obtain reliable H$1.5 \lesssim z \lesssim 2.5$3 and $1.5 \lesssim z \lesssim 2.5$4 line fluxes, correct H$1.5 \lesssim z \lesssim 2.5$5 for underlying stellar absorption, and compute rest-frame EW(H$1.5 \lesssim z \lesssim 2.5$6). In SDSS and JADES, the authors use FADO spectral synthesis, which fits the stellar population plus nebular continuum self-consistently. In LEGA-C and 3D-HST, catalogue EWs and fluxes are used with positive-EW selection. In MOSDEF, the full near-IR line set is available, EWs are measured, and the OB-I placement can be compared to redshift-dependent NII-BPT labels (Santos et al., 23 Jul 2025).

3. Empirical and semi-empirical boundaries

The paper defines an empirical AGN/SF boundary at low redshift by minimizing cross-contamination between NII-BPT SF and AGN subsets. In the OB-I plane, the demarcation is a three-parameter hyperbola:

$1.5 \lesssim z \lesssim 2.5$7

Above this line lies the AGN-dominated region; below it lies a region containing SF galaxies and systems with mixed SF and AGN excitation (Santos et al., 23 Jul 2025).

At low redshift, the contamination across this line is quantified directly. The paper reports that 13% of NII-BPT SF cross into the OB-I AGN side, while 18% of NII-BPT AGN cross into the OB-I SF side. These values define the empirical trade-off achieved by the fit (Santos et al., 23 Jul 2025).

A second, semi-empirical boundary is derived from comparison with H II region evolutionary models from Stasińska (2001). This boundary is intended to isolate a “pure SF” region:

$1.5 \lesssim z \lesssim 2.5$8

The interpretation given is threefold. Below the model-based envelope, galaxies are consistent with pure SF excitation. Between the pure-SF line and the empirical AGN line lies a mixed SF+AGN population, analogous to NII-BPT composites. Above the empirical AGN line, the emission is AGN-dominated (Santos et al., 23 Jul 2025).

This two-boundary formulation is central to the OB-I diagram’s interpretive use. It does not claim that every object can be classified unambiguously by OB-I alone. Rather, it partitions the plane into a pure-SF locus, an intermediate mixed zone, and an AGN-dominated regime.

4. Validation across surveys and redshift

The empirical basis of the OB-I diagram spans several datasets. The low-redshift anchor is FADO-SDSS DR7 with 161,087 galaxies, median $1.5 \lesssim z \lesssim 2.5$9, and $\mathrm{H}\alpha$0 on $\mathrm{H}\alpha$1, $\mathrm{H}\alpha$2, $\mathrm{H}\alpha$3, and $\mathrm{H}\alpha$4 for BPT comparison. Intermediate- and high-redshift validation uses LEGA-C, VANDELS, 3D-HST, MOSDEF, and FADO-reprocessed JADES spectra, with $\mathrm{H}\alpha$5 on H$\mathrm{H}\alpha$6 and $\mathrm{H}\alpha$7 and positive EW(H$\mathrm{H}\alpha$8) selection where applicable (Santos et al., 23 Jul 2025).

The low-redshift comparison with SII-BPT is particularly strong. Among SII-SF galaxies, 98% lie on the OB-I SF side; among SII-AGN, 96% lie on the OB-I AGN side. Conversely, among OB-I SF galaxies, 99% are SII-SF, while among OB-I AGN, 59% are SII-AGN. The remaining fraction is described as often reflecting weaker AGN signatures or compositeness (Santos et al., 23 Jul 2025).

The comparison with NII-BPT is more nuanced. The fitted low-$\mathrm{H}\alpha$9 boundary is explicitly chosen to minimize cross-contamination, but the authors also argue that OB-I reveals a known failure mode of NII-BPT at low metallicity, where AGNs can appear inside the NII-SF region. They report that 90% of OB-I AGNs inside NII-SF/Composite have sub-solar metallicity according to MPA-JHU values. This suggests that the OB-I projection captures a population that is systematically difficult for $[\mathrm{N\,II}]$0-based classification (Santos et al., 23 Jul 2025).

At higher redshifts, validation uses external AGN labels rather than a universal optical baseline. Across the aggregate sample at $[\mathrm{N\,II}]$1–2.7, 47 galaxies are externally flagged as AGN; of these, 22, approximately 47%, fall on the OB-I AGN side. The high-redshift sample contains 136 total OB-I AGNs. By survey, the paper reports LEGA-C with 24 externally flagged AGN of which 13 match the OB-I AGN region; MOSDEF with 3 flagged and 2 matched; and 3D-HST with 20 flagged and 7 matched. VANDELS and JADES contain no externally flagged AGNs in the validation set, while OB-I identifies 0 and 1 AGN respectively (Santos et al., 23 Jul 2025).

A concise summary of the principal validation figures is useful:

Comparison	Result	Context
SII-SF on OB-I SF side	98%	Low-$[\mathrm{N\,II}]$2 agreement
SII-AGN on OB-I AGN side	96%	Low-$[\mathrm{N\,II}]$3 agreement
NII-SF crossing into OB-I AGN	13%	Low-$[\mathrm{N\,II}]$4 contamination
NII-AGN crossing into OB-I SF	18%	Low-$[\mathrm{N\,II}]$5 contamination
Externally flagged high-$[\mathrm{N\,II}]$6 AGNs recovered	22/47 ($[\mathrm{N\,II}]$7)	$[\mathrm{N\,II}]$8–2.7

These figures establish the paper’s position that OB-I is a useful, but not exhaustive, classifier: it is especially effective for identifying galaxies with a dominating AGN component in their emission.

5. Physical interpretation and redshift behaviour

The physical separation in the OB-I plane follows from the different sensitivities of its two axes. The ordinate, $[\mathrm{N\,II}]$9, increases with ionization parameter, harder radiation fields, and shocks. AGN narrow-line regions therefore naturally raise this ratio, although low-metallicity starbursts can also yield high values. The abscissa, EW(H$\beta$0), traces the strength of line emission relative to the continuum; in SF galaxies it scales with sSFR, whereas in AGN a bright non-thermal or thermal continuum depresses EW even when lines are strong (Santos et al., 23 Jul 2025).

The combined effect produces the morphology described in the paper: AGN-dominated emission occupies high $\beta$1 and low $\beta$2, while SF galaxies extend to higher EW(H$\beta$3) at generally lower ratios, forming a “teardrop” locus. The theoretical support comes from H II region evolutionary tracks of Stasińska (2001), which move toward lower $\beta$4 and lower EW(H$\beta$5) as massive stars fade. At high EW(H$\beta$6), the maximal SF-only track matches the empirical AGN boundary; at lower EW, the model tracks fall below the empirical line, leaving a region for mixed SF+AGN systems (Santos et al., 23 Jul 2025).

A major claim of the paper is that the OB-I diagram is relatively resistant to the “cosmic shift” that affects many optical classification schemes. In this context, “cosmic shift” denotes the redshift-dependent displacement of galaxy loci in line-ratio diagrams due to changing ionization parameter, gas conditions, abundances, shocks, and stellar populations. The authors argue that OB-I mitigates this problem because it uses two nearby lines and incorporates EW(H$\beta$7), which partly normalizes line strength by the continuum. They report that no significant adjustment of the empirical AGN line is required up to $\beta$8 (Santos et al., 23 Jul 2025).

The paper does note observable redshift trends: EW(H$\beta$9) increases with redshift, with approximately 53% of the high-$\log_{10}(\mathrm{EW}(\mathrm{H}\beta))$0 sample consistent with Khostovan et al. (2016) EW evolution, and $\log_{10}(\mathrm{EW}(\mathrm{H}\beta))$1 also increases with redshift. Yet the boundary effectiveness remains sufficient for the authors to treat OB-I as comparatively stable across cosmic time (Santos et al., 23 Jul 2025).

6. Practical use, uncertainties, and limitations

The recommended workflow is straightforward. One measures $\log_{10}(\mathrm{EW}(\mathrm{H}\beta))$2 and $\log_{10}(\mathrm{EW}(\mathrm{H}\beta))$3 from flux-calibrated spectra, preferably with $\log_{10}(\mathrm{EW}(\mathrm{H}\beta))$4 in both lines; fits and subtracts the stellar continuum around H$\log_{10}(\mathrm{EW}(\mathrm{H}\beta))$5 to correct underlying absorption; computes rest-frame $\log_{10}(\mathrm{EW}(\mathrm{H}\beta))$6 and enforces $\log_{10}(\mathrm{EW}(\mathrm{H}\beta))$7; computes $\log_{10}(\mathrm{EW}(\mathrm{H}\beta))$8 and $\log_{10}(\mathrm{EW}(\mathrm{H}\beta))$9; and then applies the empirical AGN boundary together with the semi-empirical pure-SF boundary (Santos et al., 23 Jul 2025).

The paper also provides explicit uncertainty propagation. For EW,

$[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$0

For the line ratio,

$[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$1

Classification reliability is explicitly stated to degrade near the boundaries, and the uncertainty bands around the separating curves should therefore be considered when assigning objects to a regime (Santos et al., 23 Jul 2025).

Several limitations are emphasized. The mixed SF+AGN region is intrinsically ambiguous, so OB-I alone cannot determine the dominant contributor in every intermediate case. High $[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$2 with low EW can also arise from shocks or LINER-like mechanisms, making ancillary X-ray, IR, radio, or spatially resolved diagnostics valuable. Low-EW spectra are sensitive to continuum-placement and absorption-correction systematics. Aperture and selection effects can alter measured EW(H$[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$3) and bias the observed locus. Finally, the high-$[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$4 external AGN labels used for validation are themselves incomplete and contaminated, so OB-I’s reported recovery fractions are contingent on imperfect reference sets (Santos et al., 23 Jul 2025).

Within those constraints, the paper’s recommended domain is clear: $[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$5, optical rest-frame coverage of H$[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$6 and $[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$7, and $[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$8 in both lines. In that regime, OB-I functions as a practical, reproducible, two-line diagnostic for separating galaxies with a dominating AGN component from pure-SF and mixed systems.

7. Terminological ambiguity and other uses of “OB-I”

The expression “OB-I diagram” is not uniform across the literature. In the galaxy-classification paper, it denotes the two-line diagram built from EW(H$[\mathrm{O\,III}]\,\lambda5007/\mathrm{H}\beta$9) and $\log_{10}([\mathrm{O\,III}]/\mathrm{H}\beta)$0 (Santos et al., 23 Jul 2025). Other works included under similar wording use the term differently or do not use it at all.

In software specification, the paper “State Based Service Description” introduces I/O*-state transition diagrams (I/O*-STD) for object services. It does not use the term “OB-I (Object Behavior Interface) diagram” anywhere, but the details note that what many communities call an OB-I diagram can be realized by the paper’s service-specific I/O*-STD together with an initial-state predicate, with semantics based on I/O*-state machines, tagged threads, and stacks of suspended service invocations (Paech et al., 2014). This is a different usage domain from optical spectroscopy.

In business information modeling, “Merging Object and Process Diagrams for Business Information Modeling” likewise does not use the term “OB-I diagram.” Instead, it refers to “merging class diagrams and business process modeling” into “one overview diagram,” implemented in Topologos on the QOBJ paradigm, where Nodes, Circles, Stars, Dots, Gates, and Pilots integrate process and data views in a single executable conceptual model (0804.0366).

In stellar astrophysics, the phrase “OB instability diagram” refers not to the galaxy-classification OB-I plane but to the spectroscopic Hertzsprung–Russell diagram populated with OB stars and overlaid with instability strips for heat-driven pulsations. In that study, the diagram uses $\log_{10}([\mathrm{O\,III}]/\mathrm{H}\beta)$1 on the abscissa and $\log_{10}([\mathrm{O\,III}]/\mathrm{H}\beta)$2 on the ordinate, with $\log_{10}([\mathrm{O\,III}]/\mathrm{H}\beta)$3, and is applied to TESS, IACOB, and OWN data for 98 OB-type stars (Burssens et al., 2020). A separate paper on Auriga provides practical guidance to construct what its details call an “OB‑I Diagram” for OB associations, consisting of Galactic $\log_{10}([\mathrm{O\,III}]/\mathrm{H}\beta)$4–$\log_{10}([\mathrm{O\,III}]/\mathrm{H}\beta)$5 and $\log_{10}([\mathrm{O\,III}]/\mathrm{H}\beta)$6–distance visualizations of five high-confidence associations identified from Gaia-based OB-star samples (Quintana et al., 2023).

These usages are unrelated in subject matter and notation. In current astronomical usage, the best-defined term “OB-I diagram” is the galaxy-classification diagram based on EW(H$\log_{10}([\mathrm{O\,III}]/\mathrm{H}\beta)$7) and $\log_{10}([\mathrm{O\,III}]/\mathrm{H}\beta)$8 (Santos et al., 23 Jul 2025).