Abundance Discrepancy Factor (ADF)

• ADF is the observational measure quantifying the discrepancy between heavy-element abundances derived from recombination lines versus collisionally excited lines in nebulae.
• It serves as a key diagnostic tool in calibrating gas-phase metallicity scales and probing temperature, density, and chemical inhomogeneities in both H II regions and planetary nebulae.
• Practical applications of the ADF include testing models of temperature fluctuations and multiphase gas structures to better understand nebular astrophysics.

The Abundance Discrepancy Factor (ADF) is the observational measure of the mismatch between ionic abundances derived from heavy-element recombination lines (RLs, or ORLs in much planetary-nebula work) and those derived from collisionally excited lines (CELs) for the same ion. In ionized nebulae, RL-based abundances are systematically higher than CEL-based abundances, making the ADF a central quantity in nebular abundance analysis and in the calibration of gas-phase metallicity scales. In H II regions the discrepancy is usually modest, whereas in planetary nebulae it can become extreme, reaching factors of tens to hundreds in a small subset of objects (Esteban et al., 2016, Jones et al., 2015, Simpson et al., 2022).

1. Definition, notation, and scope

In H II-region studies, the ADF is commonly written as a logarithmic abundance offset,

$${\rm ADF}({\rm X}^{i}) = {\rm log}({\rm X}^{i}/{\rm H}^+)_{\rm RLs} - {\rm log}({\rm X}^{i}/{\rm H}^+)_{\rm CELs},$$

or equivalently

$$\mathrm{ADF(X^{i+}) \equiv log(X^{i+}/H^+)_{RLs} - log(X^{i+}/H^+)_{CELs}}.$$

A positive value means that the RL abundance exceeds the CEL abundance, which is the usual observational outcome (Esteban et al., 2016, Cipriano et al., 2016). In much planetary-nebula literature, the same comparison is written as a linear ratio,

$\mathrm{ADF}(X^{i+}) = \frac{(X^{i+}/H^+)_{\mathrm{ORLs}}{(X^{i+}/H^+)_{\mathrm{CELs}}},$

with the largest body of work focused on $\mathrm{O}^{2+}$ because both optical RLs and bright CELs are available for that ion [(Garcia-Rojas et al., 2013); (Ruiz-Escobedo et al., 2022)].

The best-studied reference ion is therefore $\mathrm{O}^{2+}$. Other ions with observational ADF determinations include $\mathrm{O}^{+}$, $\mathrm{C}^{2+}$, and $\mathrm{Ne}^{2+}$ in H II regions, and $\mathrm{O}^{+}$, $\mathrm{Ne}^{2+}$, and other heavy ions in planetary nebulae, although those measurements are generally sparser or more sensitive to aperture, atomic-data, and recombination-correction systematics [(Esteban et al., 2016); (Garcia-Rojas et al., 2013)].

The discrepancy is astrophysically important because the adopted oxygen abundance scale underpins chemical evolution models, stellar-yield inferences, luminosity-metallicity and mass-metallicity relations, strong-line calibrations, and even primordial helium work (Esteban et al., 2016). The ADF is therefore not a minor spectroscopic detail but a systematic uncertainty in nebular abundance analysis.

2. Observational determination

The ADF is constructed by deriving the same ionic abundance from two line classes with very different emissivity dependences. CELs are bright and observationally efficient, but their emissivities depend strongly on electron temperature. RLs are intrinsically faint but much less temperature-sensitive. In H II regions, heavy-element optical RLs typically have intensities of order $$\mathrm{ADF(X^{i+}) \equiv log(X^{i+}/H^+)_{RLs} - log(X^{i+}/H^+)_{CELs}}.$$0 to $$\mathrm{ADF(X^{i+}) \equiv log(X^{i+}/H^+)_{RLs} - log(X^{i+}/H^+)_{CELs}}.$$1 times that of H$$\mathrm{ADF(X^{i+}) \equiv log(X^{i+}/H^+)_{RLs} - log(X^{i+}/H^+)_{CELs}}.$$2, which is why high-S/N spectra from large telescopes or deep IFU observations are required (Esteban et al., 2016).

For oxygen and carbon, the classic optical RL diagnostics are O II multiplet 1 around $$\mathrm{ADF(X^{i+}) \equiv log(X^{i+}/H^+)_{RLs} - log(X^{i+}/H^+)_{CELs}}.$$3 Å and C II $$\mathrm{ADF(X^{i+}) \equiv log(X^{i+}/H^+)_{RLs} - log(X^{i+}/H^+)_{CELs}}.$$4. The canonical CEL comparison for $$\mathrm{ADF(X^{i+}) \equiv log(X^{i+}/H^+)_{RLs} - log(X^{i+}/H^+)_{CELs}}.$$5 uses the bright [O III] nebular lines and the [O III] auroral line to define the standard optical $$\mathrm{ADF(X^{i+}) \equiv log(X^{i+}/H^+)_{RLs} - log(X^{i+}/H^+)_{CELs}}.$$6 method [(Cipriano et al., 2016); (Cipriano et al., 2017); (Mesa-Delgado et al., 2010)]. For carbon, historical CEL abundances often came from ultraviolet transitions, which introduced additional aperture and reddening systematics in early ADF(C$$\mathrm{ADF(X^{i+}) \equiv log(X^{i+}/H^+)_{RLs} - log(X^{i+}/H^+)_{CELs}}.$$7) work (Cipriano et al., 2016).

Recent work has expanded the diagnostic basis. Far-infrared [O III] $$\mathrm{ADF(X^{i+}) \equiv log(X^{i+}/H^+)_{RLs} - log(X^{i+}/H^+)_{CELs}}.$$8 and $$\mathrm{ADF(X^{i+}) \equiv log(X^{i+}/H^+)_{RLs} - log(X^{i+}/H^+)_{CELs}}.$$9 lines provide a CEL-based oxygen abundance scale that is only weakly temperature-sensitive, and ultraviolet CEL ratios such as O III] $\mathrm{ADF}(X^{i+}) = \frac{(X^{i+}/H^+)_{\mathrm{ORLs}}{(X^{i+}/H^+)_{\mathrm{CELs}}},$0 and C III] $\mathrm{ADF}(X^{i+}) = \frac{(X^{i+}/H^+)_{\mathrm{ORLs}}{(X^{i+}/H^+)_{\mathrm{CELs}}},$1 allow internally consistent CEL determinations of $\mathrm{ADF}(X^{i+}) = \frac{(X^{i+}/H^+)_{\mathrm{ORLs}}{(X^{i+}/H^+)_{\mathrm{CELs}}},$2 (Chen et al., 2023, Kelly et al., 5 Aug 2025). Spatially resolved IFU and long-slit data have further turned the ADF from a single integrated number into a mapped quantity, enabling direct comparison with $\mathrm{ADF}(X^{i+}) = \frac{(X^{i+}/H^+)_{\mathrm{ORLs}}{(X^{i+}/H^+)_{\mathrm{CELs}}},$3, $\mathrm{ADF}(X^{i+}) = \frac{(X^{i+}/H^+)_{\mathrm{ORLs}}{(X^{i+}/H^+)_{\mathrm{CELs}}},$4, morphology, and kinematics [(Mesa-Delgado et al., 2012); (García-Rojas et al., 2021); (Gómez-Llanos et al., 2024); (Singh et al., 23 Mar 2026)].

3. H II regions

In the H II-region literature, $\mathrm{ADF}(X^{i+}) = \frac{(X^{i+}/H^+)_{\mathrm{ORLs}}{(X^{i+}/H^+)_{\mathrm{CELs}}},$5 is typically between 0.10 and 0.35 dex (Esteban et al., 2016). Representative values collected in the review literature include Orion, M8, 30 Dor, N66C, and NGC 5253, and the discrepancy is positive for multiple ions, not only $\mathrm{ADF}(X^{i+}) = \frac{(X^{i+}/H^+)_{\mathrm{ORLs}}{(X^{i+}/H^+)_{\mathrm{CELs}}},$6 (Esteban et al., 2016).

The Magellanic Clouds provide one of the clearest homogeneous extragalactic datasets. In the Large Magellanic Cloud, $\mathrm{ADF}(X^{i+}) = \frac{(X^{i+}/H^+)_{\mathrm{ORLs}}{(X^{i+}/H^+)_{\mathrm{CELs}}},$7 ranges from $\mathrm{ADF}(X^{i+}) = \frac{(X^{i+}/H^+)_{\mathrm{ORLs}}{(X^{i+}/H^+)_{\mathrm{CELs}}},$8 in 30 Doradus to $\mathrm{ADF}(X^{i+}) = \frac{(X^{i+}/H^+)_{\mathrm{ORLs}}{(X^{i+}/H^+)_{\mathrm{CELs}}},$9 in N44C, while in the Small Magellanic Cloud it ranges from $\mathrm{O}^{2+}$0 to $\mathrm{O}^{2+}$1. The LMC values cluster near $\mathrm{O}^{2+}$2 dex, whereas the SMC values cluster near $\mathrm{O}^{2+}$3 dex, with low internal dispersion in each galaxy (Cipriano et al., 2016). In the same dataset, ADF(C$\mathrm{O}^{2+}$4) is more uncertain and more dispersed because the carbon CEL abundances came from ultraviolet data, but the SMC values are again larger than the LMC ones (Cipriano et al., 2016).

A broader comparison against oxygen abundance suggests that $\mathrm{O}^{2+}$5 is not monotonic with metallicity. The distribution shows a minimum around

$\mathrm{O}^{2+}$6

with larger ADFs at both lower and higher metallicity, producing the “seagull shape” emphasized in the review literature. The low-metallicity rise appears clearer than the high-metallicity branch, which is strongly affected by dispersion and by the confounding role of low-ionization objects (Esteban et al., 2016). This metallicity dependence is one reason the Magellanic-Cloud measurements are important: the SMC falls precisely in the regime where elevated ADFs are observed (Cipriano et al., 2016).

Spatially resolved work shows that the H II-region ADF can be either fairly smooth or strongly environment-dependent. In long-slit data for M8 and M17, $\mathrm{O}^{2+}$7 remains fairly constant along the observed slit positions, with mean values around 0.37–0.40 dex; by contrast, the integrated spectrum of NGC 7635 yields a remarkably high value of about 0.59 dex (Mesa-Delgado et al., 2010). The first resolved ADF($\mathrm{O}^{2+}$8) map of an H II region, for M8 from SDSS-V/LVM, gives a global mean of $\mathrm{O}^{2+}$9 dex and radial variations between $\mathrm{O}^{2+}$0 and $\mathrm{O}^{2+}$1 in the Her 36-dominated central region (Singh et al., 23 Mar 2026). At still smaller scales, proplyd spectroscopy in Orion shows that dense ionized structures can strongly bias CEL diagnostics: once the high density of proplyd 177-341 is treated properly and the background is subtracted, the intrinsic $\mathrm{O}^{2+}$2 is about zero, whereas the surrounding nebular background retains the usual positive ADF (Mesa-Delgado et al., 2012).

The comparison with young stars is mixed and environment-dependent. In Orion, RL-based C/H and O/H are more consistent with solar and stellar abundances than CEL-based values (Esteban et al., 2016). In the Magellanic Clouds, by contrast, B-type stellar abundances agree better with nebular CEL-based oxygen abundances than with RL-based ones, and RL-based carbon can be too high relative to the stellar benchmark if a dust correction is included (Cipriano et al., 2016). The review literature summarizes this as a possible metallicity-dependent reversal: RL-based nebular metallicities appear to agree better with stellar metallicities in the high-abundance regime, whereas CEL-based metallicities agree better at low abundance (Esteban et al., 2016).

Far-infrared tests complicate any universal interpretation. In Mrk 71, the far-IR [O III] abundance agrees with the optical CEL abundance and not with the RL abundance, while the directly inferred $\mathrm{O}^{2+}$3 is consistent with zero. For that object, the long-standing temperature-fluctuation explanation of the ADF is ruled out (Chen et al., 2023).

4. Planetary nebulae

Planetary nebulae exhibit the full dynamic range of ADF phenomenology. Ordinary planetary nebulae typically have ADFs of order 2–3, but a minority reach values many times larger, extending to 2–3 orders of magnitude in extreme cases (Simpson et al., 2022). The contrast with H II regions is therefore fundamental: planetary nebulae include an extreme-ADF population not seen in normal H II-region samples (Esteban et al., 2016).

Not all PNe with unusual central stars show extreme discrepancies. In deep studies of [WC] and wels nebulae, $\mathrm{O}^{2+}$4 is moderate, in the range 1.2 to 4, and no correlation is found with surface brightness, diameter, excitation class, density, metallicity, N/O, or [WC] spectral type (Garcia-Rojas et al., 2013). A preliminary MIKE survey of a dozen [WC]/wels PNe likewise found oxygen ADFs always below 4 and no direct evidence for the cold, dense H-poor clumps proposed in some earlier interpretations (García-Rojas et al., 2011). Young high-density nebulae such as Vy 2-2, Hu 2-1, Vy 1-2, and IC 4997 occupy an intermediate regime, with $\mathrm{O}^{2+}$5 between $\mathrm{O}^{2+}$6 and $\mathrm{O}^{2+}$7 (Ruiz-Escobedo et al., 2022).

The extreme tail is represented by objects such as NGC 6153, NGC 6778, Hf 2-2, M 1-42, A30, and Sp 3. NGC 6153 has a characteristic ADF $\mathrm{O}^{2+}$8 and was used as a benchmark for explicit multiphase abundance analysis (Gómez-Llanos et al., 2024). NGC 6778 shows an integrated $\mathrm{O}^{2+}$9 of 17.9 and rises to about 40 close to the central star (Jones et al., 2015). Sp 3 has an extreme oxygen ADF of $\mathrm{O}^{+}$0 despite a 4.81 d central-binary period, which is inconsistent with the purported trend that longer-period post-common-envelope PNe should exhibit normal ADFs (Miszalski et al., 2019). Hf 2-2 and M 1-42 have long been treated as canonical high-ADF nebulae with values of about 70 and 20, respectively, and resolved studies find local $\mathrm{O}^{+}$1 values as high as $\mathrm{O}^{+}$2 in Hf 2-2 and $\mathrm{O}^{+}$3 in M 1-42 near the central sightlines (Castañeda-Carlos et al., 19 Sep 2025). In the born-again planetary nebula A30, the polar knots reach an ADF of about 700, whereas the equatorial knot J4 yields only $\mathrm{O}^{+}$4 and a conservative upper limit of 35, demonstrating strong internal segregation of the discrepancy (Simpson et al., 2022).

Spatially resolved IFU and echelle spectroscopy show that these extreme ADFs are usually centrally peaked and associated with distinct ORL-emitting plasma. MUSE maps of NGC 6778, M 1-42, and Hf 2-2 show that both $\mathrm{O}^{+}$5 and $\mathrm{O}^{+}$6 peak toward the center, where recombination-based temperatures are lowest and heavy-element ORLs are strongest (García-Rojas et al., 2021). UVES PV analyses of Hf 2-2 and M 1-42 confirm two plasma components: a normal nebular plasma that emits both forbidden and permitted lines, and an additional, denser, cooler, centrally concentrated component that emits the permitted lines of O I, C II, N II, O II, and Ne II (Castañeda-Carlos et al., 19 Sep 2025). A complementary UVES reanalysis of Hf 2-2, M 1-42, and NGC 6153 finds CEL temperatures near $\mathrm{O}^{+}$7 K, Balmer and Paschen jump temperatures below that, and O II and N II ORL temperatures $\mathrm{O}^{+}$8 K, with ORL-emitting regions close to the nebular center (Huang et al., 4 Dec 2025).

5. Physical interpretations and discriminants

The classical explanation is spatial temperature fluctuations, parameterized by $\mathrm{O}^{+}$9 in the Peimbert formalism. In this picture, RL abundances are closer to the true abundances because their emissivities are almost identical to that of H$\mathrm{C}^{2+}$0 over the usual H II-region temperature range, whereas CEL emissivities are strongly temperature-sensitive and therefore biased by unresolved thermal structure (Esteban et al., 2016). This framework remains central in both H II-region and PN work.

Other interpretations remain active. The review literature discusses the $\mathrm{C}^{2+}$1-distribution hypothesis, but also notes that this idea has been seriously questioned (Esteban et al., 2016). A chemically inhomogeneous scenario due to semi-ionized, dense, cool, metal-rich clumps interpreted as unmixed supernova ejecta has also been proposed for H II regions, with the distinct implication that RL and CEL abundances bracket the true value rather than tracing the same plasma (Esteban et al., 2016). A related idea invokes high-density clumps without abundance contrast (Esteban et al., 2016). In H II regions, very dense ionized structures such as proplyds show directly that collisional de-excitation can depress CEL abundances so strongly that the intrinsic ADF tends to zero once the correct density structure is used (Mesa-Delgado et al., 2012).

For planetary nebulae with high ADFs, the weight of the resolved evidence favors explicit multiphase gas. In NGC 6153, a tailored two-phase analysis separates a warm CEL-emitting phase from a cold RL-emitting phase and introduces the abundance contrast factor (ACF), defined as the abundance ratio between the two plasma components once the H$\mathrm{C}^{2+}$2 partition is accounted for; the ACF is, on average, 0.9 dex higher than the ADF (Gómez-Llanos et al., 2024). MUSE and UVES studies of NGC 6153, Hf 2-2, M 1-42, and NGC 6778 consistently find centrally concentrated ORL emission, lower temperatures from recombination diagnostics, and kinematic differences between CEL- and ORL-emitting zones, supporting a cold, metal-rich component embedded within the normal nebular plasma (García-Rojas et al., 2021, Gómez-Llanos et al., 2024, Huang et al., 4 Dec 2025). In Hf 2-2 and M 1-42, the additional plasma component contains masses of $\mathrm{C}^{2+}$3 and $\mathrm{C}^{2+}$4 ions at least as large as those in the normal nebular plasma, which the authors identify as the main reason those nebulae have larger ADFs than NGC 6153 (Castañeda-Carlos et al., 19 Sep 2025).

Not all fluctuation-based models reproduce the full phenomenology. Simulations with realistic continuous density and temperature fluctuations can reproduce filling factors in the observed range and may be consistent with H II-region ADFs, but they do not satisfactorily reproduce the ADFs of planetary nebulae, especially the extreme tail. In those models, negative density-temperature correlation can increase the optical ADF, but it also tends to drive the inferred filling factor above unity, which is not generally observed (Bergerud et al., 2019).

The empirical relation to central-star evolution is also significant. Extreme ADFs are increasingly associated with close binaries and with born-again systems. NGC 6778 strengthens the link between high ADFs and central-star binarity (Jones et al., 2015), while Sp 3 shows that extreme ADFs are not confined to the shortest post-common-envelope periods and that selection effects likely affect any simple period-ADF trend (Miszalski et al., 2019). The review literature therefore treats low-ionization, dynamically disturbed, or binary-shaped nebulae as potentially special environments in which the ADF is amplified (Esteban et al., 2016).

6. Abundance scales, ratios, and current status

The ADF has immediate consequences for the abundance scale itself. If RL and CEL methods differ systematically by 0.1–0.35 dex in H II regions, and by much more in extreme PNe, then absolute O/H and C/H abundances are method-dependent and require an explicit physical choice (Esteban et al., 2016). In some low-metallicity H II regions, stellar benchmarks and far-IR tests support the CEL scale over the RL scale (Cipriano et al., 2016, Chen et al., 2023). In Orion and some higher-metallicity environments, RLs are favored by the comparison with local stellar and solar abundances (Esteban et al., 2016). The review conclusion is therefore deliberately non-universal: the best abundance indicator may depend on metallicity regime and nebular environment (Esteban et al., 2016).

At the same time, some abundance ratios are much more stable than absolute abundances. New ultraviolet-plus-optical work finds a clear ADF for $\mathrm{C}^{2+}$5, with

$\mathrm{C}^{2+}$6

and shows that UV CEL-based and optical RL-based $\mathrm{C}^{2+}$7 agree to within about $\mathrm{C}^{2+}$8 dex and generally within $\mathrm{C}^{2+}$9 dex object by object (Kelly et al., 5 Aug 2025). Absolute C/H and O/H remain uncertain at the ADF level, but C/O is comparatively robust if carbon and oxygen suffer similar discrepancies (Kelly et al., 5 Aug 2025).

The present observational picture is therefore stratified by object class. In H II regions, the ADF is systematic but usually modest, often smooth on resolved scales, and plausibly linked to metallicity, temperature structure, density inhomogeneities, or combinations thereof (Esteban et al., 2016, Singh et al., 23 Mar 2026). In planetary nebulae, especially the high-ADF subset, the discrepancy is frequently spatially and kinematically segregated and is most naturally described as a multiphase problem involving a cold, metal-rich component (García-Rojas et al., 2021, Gómez-Llanos et al., 2024, Huang et al., 4 Dec 2025). The review literature explicitly warns against a single universal mechanism and argues that the “story may be more complex than a single universal mechanism” (Esteban et al., 2016).

Taken together, the ADF remains a major unsolved problem in ionized-nebula astrophysics, but it is no longer merely an integrated ORL/CEL mismatch. It is an empirical tracer of the thermal, chemical, and dynamical structure of nebulae, and it defines the conditions under which absolute nebular abundances can—or cannot—be placed on a common scale across local H II regions, planetary nebulae, and high-redshift galaxy spectroscopy (Esteban et al., 2016, Chen et al., 2023, Kelly et al., 5 Aug 2025).