---
title: 'O32: Nebular Oxygen Excitation Ratio'
url: https://www.emergentmind.com/topics/o32
type: topic
---

# O32: Nebular Oxygen Excitation Ratio

Searching arXiv for recent and foundational papers on O32 to ground the article in published work.
O32 is a nebular oxygen excitation ratio that compares emission from doubly ionized oxygen to emission from singly ionized oxygen. In contemporary astrophysical literature it functions as an ionization-sensitive observable, a component of strong-line metallicity diagnostics, and an indirect probe of nebular optical depth, neutral-gas content, and Lyman continuum leakage. The exact convention is not universal: some studies define O32 as a linear ratio and others in logarithmic form; some use only [O III] $\lambda5007$ (or $\lambda5008$ in vacuum wavelengths) in the numerator, whereas others use the summed [O III] $\lambda\lambda4959,5007$ doublet [2311.18008] [2601.12413] [2605.30410].

## 1. Definition and notational conventions

In one common convention, O32 is the dust-corrected flux ratio
\[
\mathrm{O32}=\frac{F([\mathrm{O\,III}]\,\lambda\lambda4959,5007)}{F([\mathrm{O\,II}]\,\lambda3727)}
\]
or, when the [O II] doublet is resolved,
\[
\mathrm{O32}\equiv \frac{[\mathrm{O\,III}]\,\lambda5007+[\mathrm{O\,III}]\,\lambda4959}{[\mathrm{O\,II}]\,\lambda\lambda3727,3729}.
\]
This form is adopted in studies of Lyman continuum leakage and Green Pea galaxies [2311.18008] [2509.04567].

A second convention uses only the strongest [O III] line in the numerator,
\[
\mathrm{O}_{32}=\frac{[\mathrm{O\,III}]\,\lambda5007}{[\mathrm{O\,II}]\,\lambda3727},
\]
which appears in low-redshift compact star-forming galaxy work and related LyC-leaker searches [1805.09865] [1706.08769]. A third convention takes the base-10 logarithm of the ratio,
\[
\mathrm{O32}=\log\left(\frac{[\mathrm{O\,III}]\,\lambda5007}{[\mathrm{O\,II}]\,\lambda\lambda3726,3729}\right),
\]
or equivalently with the summed [O III] numerator depending on the calibration set [2601.12413] [1809.03612].

Vacuum-wavelength notation introduces a further bookkeeping distinction. In RUBIES, O32 is defined as
\[
\mathrm{O32}\equiv \frac{[\mathrm{O\,III}]\,\lambda5008}{[\mathrm{O\,II}]\,\lambda\lambda3727,3730},
\]
with an explicit reminder that some authors instead use
\[
\mathrm{O32}'\equiv \frac{[\mathrm{O\,III}]\,\lambda\lambda4960,5008}{[\mathrm{O\,II}]\,\lambda\lambda3727,3730},
\]
and that the two conventions can be rescaled using the fixed ratio $[{\rm O\,III}]\,\lambda5008/[{\rm O\,III}]\,\lambda4960=2.985$ [2605.30410].

The literature therefore treats “O32” as a family of closely related observables rather than a single invariant quantity. For technical use, the precise definition—linear versus logarithmic, single-line versus summed-line numerator, air versus vacuum wavelengths—must be specified.

## 2. Physical content and measurement

Across the literature, O32 is interpreted primarily as a tracer of the ionization parameter and, more broadly, of the excitation state of the ionized gas. High O32 typically indicates a high ionization parameter, a hard ionizing spectrum, and in some cases a density-bounded H II region in which the outer O$^+$ zone is truncated; low O32 tends to indicate lower ionization parameter, more extended low-ionization zones, more radiation-bounded conditions, or more complex effects from metallicity, density, geometry, and dust [2509.04567] [2311.18008] [1809.09637].

That interpretation is not one-to-one. O32 is also affected by metallicity, electron density, the hardness of the stellar or AGN ionizing continuum, and nebular geometry. Testing strong-line metallicity diagnostics at $z\sim2$ showed explicitly that O32 is primarily sensitive to the ionization parameter and that its metallicity sensitivity is largely indirect, arising from correlations between metallicity and ionization parameter in galaxy populations [1809.03612]. More recent high-redshift work similarly emphasizes that the mapping from O32 to ionization parameter is intrinsically degenerate when metallicity, density, and spectral hardness are allowed to vary over realistic ranges [2605.30410].

Measurement practice reflects the large wavelength baseline between [O II] and [O III]. In many studies, line fluxes are corrected for dust using Balmer decrements and an extinction law before O32 is constructed. Examples include LRS2 follow-up of Mg II-selected HETDEX galaxies, where [O II] and [O III] were fit with Gaussian components and dust corrected using $E(B-V)$ from Balmer decrements [2311.18008], and JWST-era metallicity work, where de-reddened [O II], [O III], H$\beta$, H$\alpha$, and [N II] fluxes are required before computing O32 and related composite indicators [2601.12413]. By contrast, some AGN-host analyses emphasize relative comparisons within a restricted sample and do not apply an explicit dust correction to O32 in the main analysis [2507.08298].

Spatially resolved work shows that reddening corrections can alter the inferred O32 structure qualitatively. In two strongly lensed galaxies at $z\simeq1.3$–1.4, significant spatial variation in the Balmer decrement translated into significant spatial variation in reddening-corrected O32 and R23, corresponding to spreads of a few tenths of a dex in ionization parameter and metallicity. In one case, using only a global reddening estimate could invert the ordering of regions by O32 [2006.11387]. This indicates that O32 is not merely an intrinsic excitation index; it is also a quantity whose scientific interpretation depends on how faithfully differential attenuation is treated.

## 3. O32 as a LyC-leakage diagnostic

O32 became prominent in LyC-leaker studies because density-bounded H II regions suppress [O II] relative to [O III], potentially driving the ratio to large values. This logic motivated low-redshift HST/COS campaigns targeting compact star-forming galaxies with high O32. In one such program, five galaxies at $z=0.2993$–0.4317 with O32 $\sim8$–27 all showed LyC detections, with $f_{\rm esc}({\rm LyC})$ in the range 2–72 per cent; the sample also showed a general increase of LyC escape fraction with increasing O32, but with large scatter [1805.09865]. An even more extreme LBT sample of compact star-forming galaxies with O32 ranging from 23 to 43 concluded that high O32 may not be a sufficient condition for LyC leakage and introduced He I $\lambda3889/\lambda6678$ and $\lambda7065/\lambda6678$ as additional diagnostics of density-boundedness [1706.08769].

Subsequent work made the non-uniqueness of O32 increasingly explicit. A COS study of eight compact star-forming galaxies with O32 $\simeq22$–39 found a diversity of Ly$\alpha$ properties and showed that there is no correlation between O32 and $f_{\rm esc}({\rm Ly}\alpha)$ [1910.12773]. A simulation-based analysis of reionization-era galaxies found that high-$f_{\rm esc}$ systems are biased toward high O32, but that the relation is far from deterministic: among simulated galaxies with ${\rm O32}>0$, 11% have $f_{\rm esc}>10\%$, whereas among those with ${\rm O32}<0$, only 1% do. The same work argued that high O32 is likely a necessary but not sufficient condition for high $f_{\rm esc}$ in isolated galaxies, because metallicity, ionization parameter, viewing angle, and ISM geometry all affect where a galaxy falls in the O32–$f_{\rm esc}$ plane [2005.01734].

Observational selection outside the classical high-O32 regime reaches a similar conclusion. A Mg II-selected HETDEX sample, selected only for bright Mg II and [O II] emission rather than O32, spans O32 from 0.74 to 8.51. Using Mg II-based “R” and O32-based photoionization methods, the study identified 7 and 5 LyC-leaker candidates at $>2\sigma$ significance, with predicted $f_{\rm esc}$ values ranging from 3 to 80%. Notably, the strongest inferred LyC emitter had an O32 value one order of magnitude lower than previous strong leakers, directly challenging the idea that very high O32 is required for large escape fractions [2311.18008].

A complementary Green Pea H I study adopted O32 $\ge 10$ as a likely-LyC-leaker threshold and found a far higher H I 21 cm detection rate for the 32 Green Peas with O32 $<10$ than for the 28 with O32 $>10$: $\approx53^{+16}_{-13}\%$ versus $7.1^{+9.4}_{-4.6}\%$. The same work found that the H I mass, H I-to-stellar-mass ratio, and H I depletion timescale are all lower in the O32 $>10$ subsample, interpreting the O32–LyC link as a consequence of H I depletion and density-bounded ionized regions [2509.04567]. Taken together, these studies indicate that O32 is an efficient biasing criterion for LyC-leaker searches, but not a sufficient or uniquely reliable predictor of LyC escape in individual systems.

## 4. O32 in strong-line metallicity diagnostics

O32 has a dual role in metallicity work. Used alone, it is often a poor stand-alone oxygen-abundance indicator because of its strong dependence on ionization parameter. In a sample of 16 galaxies at $z\sim2$ with direct auroral-line metallicities, O32-based metallicity calibrations showed large dispersions: $0.41\pm0.06$ dex for the Maiolino et al. calibration, $0.27\pm0.05$ dex for Jones et al., $0.33\pm0.06$ dex for Curti et al., and $0.28\pm0.05$ dex for the Bian et al. “local analog” calibration. The study concluded that O32 is one of the least precise diagnostics in that comparison and should mainly be treated as an ionization-parameter or excitation indicator rather than a precise metallicity proxy [1809.03612].

At the same time, O32 substantially improves metallicity estimates when it is used explicitly as an ionization correction alongside another abundance-sensitive ratio. A 2026 JWST-era study defines
\[
{\rm R}_{\rm u}\equiv {\rm R23}+\alpha_1{\rm O32}+\alpha_2{\rm N2O2},
\]
\[
\widehat{\rm R}_{\rm u}\equiv \widehat{\rm R}+\beta_1{\rm O32}+\beta_2{\rm N2O2},
\]
\[
{\rm O}_{\rm u}\equiv {\rm O3N2}+\gamma_1{\rm O32}+\gamma_2{\rm N2O2},
\]
and calibrates $12+\log({\rm O/H})$ as low-order polynomials in each composite indicator. Applied to a JWST sample with $T_e$-method abundances, the updated indicators raise adjusted $\mathbb{R}^2$ from $\mathbb{R}^2\lesssim0$ for classical one-dimensional indicators to $\mathbb{R}^2\gtrsim0.5$ for the full sample and to $\sim0.7$ at $z>2$ [2601.12413]. In this framework, O32 is not itself a metallicity scale; it is the term that corrects strong-line metallicity estimates for ionization-driven variance.

At the very metal-poor end, extreme-O32 dwarf galaxies motivated specialized R23–O32 calibrations. J2229+2725, with O32 $\sim53$ and $12+\log({\rm O/H})=7.085\pm0.031$, led to an updated relation
\[
12+\log\frac{\rm O}{\rm H} = 0.950\,\log({\rm R}_{23}-a_1{\rm O}_{32})+6.805,
\qquad
a_1=0.080-0.00078\,{\rm O}_{32},
\]
designed for the most metal-deficient galaxies [2104.08035]. J1046+4047, with O32 $\sim57$ and $12+\log({\rm O/H})=7.091\pm0.016$, further refined the correction to
\[
12+\log\frac{\rm O}{\rm H} = 0.917\,\log({\rm R}_{23}-a_2{\rm O}_{32})+6.804,
\]
\[
a_2=0.080-0.00078\,{\rm O}_{32}+0.00000095\,{\rm O}_{32}^2,
\]
with stated applicability for all galaxies with O32 $<60$ and $12+\log({\rm O/H})<7.65$ [2311.01799].

A major caveat comes from radiative SPH simulations of reionization-era galaxies. Applying the O32 diagnostic directly to synthetic spectra generated from Technicolor Dawn plus Cloudy produced metallicities biased by about 1 dex relative to the true mass-weighted gas-phase oxygen abundance. The paper argues that the bias arises because single-zone strong-line calibrations do not capture the multi-zone, high-ionization, spatially complex ISM of simulated high-redshift galaxies [2510.20021]. This suggests that O32-based abundance estimates at high redshift are only as reliable as the assumed relation between ionization structure and metallicity.

## 5. Redshift evolution, galaxy scaling relations, and reionization-era analogs

O32 is closely tied to the evolving excitation state of star-forming galaxies. In a sample of 227 extreme [O III] emitters at $1.3<z<2.4$, stacked spectra show a monotonic increase of O32 with [O III] $\lambda5007$ equivalent width: $1.4\pm0.2$ for EW$=0$–225 Å, $3.6\pm0.2$ for 225–450 Å, $4.8\pm0.3$ for 450–800 Å, and $9.1\pm0.5$ for 800–2500 Å. The most extreme individual systems reach O32 $=17\pm4$ and $22\pm3$. These values are linked to very young, metal-poor stellar populations, very high specific star formation rates, and large $\xi_{\rm ion}$, and the paper argues that only the ultra-extreme [O III] emitters routinely enter the O32 regime associated with significant LyC escape [1809.09637].

JWST/NIRSpec analyses deepen that picture by connecting O32 to gas density and star-formation-rate surface density. In CEERS galaxies at $z=2.7$–6.3, the sample was divided into low- and high-O32 bins with mean O32 values of $1.978\pm0.057$ and $6.055\pm0.174$. The high-O32 composite has $\langle n_e\rangle \simeq 500\,\mathrm{cm^{-3}}$, at least a factor of $\simeq5$ larger than the low-O32 bin, while the mean ionizing photon rate $\langle \log Q\rangle \approx 54.50$–54.51 shows no significant difference between the two bins. The same study found a highly significant correlation between ionization parameter and star-formation-rate surface density,
\[
\log U = (0.378\pm0.021)\,\log\left[\frac{\Sigma_{\rm SFR}}{M_\odot\,\mathrm{yr^{-1}\,kpc^{-2}}}\right] + (-2.655\pm0.011),
\]
and concluded that gas density may play a more central role than metallicity in modulating $U$ at these redshifts [2303.11397].

On larger samples, O32-based inferences show a strong cosmic trend. RUBIES used Cloudy photoionization models and O32 ratios for 434 galaxies at $3<z<9$ and compared them to SDSS, LEGA-C, and KBSS. The study found that $U$ increases with redshift and specific star formation rate and decreases with stellar mass; even at fixed stellar mass and sSFR, $U$ increases by a factor of $\sim4$ from $z=2$ to $z=6$. At the same time, it emphasized that O32 alone imposes a systematic uncertainty in $\log U$ of $\sim0.3$ dex at zero measurement uncertainty because a wide range of photoionization models predict the same O32 ratio without informative priors [2605.30410].

Intermediate-redshift dwarfs provide a lower-redshift checkpoint on this evolution. In HALO7D/DEEPWinds composites at $0.6<z<1.0$, O32 decreases with stellar mass from $0.16^{+0.03}_{-0.03}$ at $\log(M_\star/M_\odot)=8.34$ to $-0.85^{+0.18}_{-0.31}$ at $\log(M_\star/M_\odot)=10.73$. The same work argued that the typical $z\sim1$ dwarf has already moved close to the low-ionization, higher-metallicity regime characteristic of the local universe, while a tail of individual dwarfs still reaches high-O32, high-Ne3O2 conditions similar to $z>7$ systems [2301.07444]. This suggests that O32 traces both instantaneous excitation and long-term demographic evolution.

## 6. Beyond H II regions: AGN, circumgalactic nebulae, and systematic caveats

O32 is also used in AGN and CGM studies, where it serves as an ionization-level-sensitive parameter rather than a direct abundance or LyC diagnostic. In X-ray selected AGNs at $0.5<z<0.9$, O32 is defined as $\log([\mathrm{O\,III}]5007/[\mathrm{O\,II}]3727)$ and shows a positive correlation with X-ray luminosity in both obscured and unobscured subsamples. The fitted relations are
\[
\log(\mathrm{O32})=(0.20\pm0.08)\,[\log(L_{\rm X})-42.5]-0.28\pm0.08
\]
for obscured AGN and
\[
\log(\mathrm{O32})=(0.48\pm0.09)\,[\log(L_{\rm X})-42.5]-0.19\pm0.07
\]
for unobscured AGN, with the latter showing the stronger correlation. Low-excitation obscured AGN form a distinct low-O32 population, whereas high-excitation obscured and unobscured AGN have similar O32 distributions [2507.08298].

A contrasting application is the circumgalactic oxygen nebula “Oxyster” at $z=0.924$. There the integrated [O III] and [O II] luminosities imply a global O32 of order $0.16$, and the narrow-band map shows O32 $<1$ over the entire nebula region. Standard AGN light echoes usually occupy a higher-O32 regime, so the uniformly low ratio is central to the argument that Oxyster is not a straightforward high-z analog of Hanny’s Voorwerp. Photoionization and shock models can each reproduce parts of the line-ratio constraints, but the study concludes that no single standard model explains simultaneously the nebular luminosities, low O32, and non-detection of H$\beta$ [2504.11531].

Several caveats recur across all of these contexts. Selection effects matter: O32-selected LyC samples, Mg II-selected samples, Green Peas, lensed galaxies, AGN hosts, and circumgalactic nebulae probe different regions of parameter space [2311.18008] [2509.04567]. Dust corrections matter because [O II] and [O III] are far apart in wavelength; unresolved spatial reddening can bias both local O32 values and gradients [2006.11387]. Model dependence matters because O32 is sensitive to metallicity, density, spectral hardness, optical depth, and geometry, so empirical calibrations derived in one redshift or population regime may fail in another [1809.03612] [2510.20021] [2605.30410].

Taken together, the literature supports a narrow but robust characterization. O32 is a powerful excitation diagnostic and an indispensable component of modern nebular analysis. It is often highly informative about ionization parameter, neutral-gas structure, and the likelihood of LyC escape, but it is not a one-to-one tracer of any of those quantities. Its scientific value is greatest when it is embedded in a multi-line, physically constrained framework rather than interpreted in isolation.

Source: https://www.emergentmind.com/topics/o32