Diffuse Ionized Gas (DIG): Astrophysical Insights

• Diffuse Ionized Gas (DIG) is a faint, extended warm ionized medium in galaxies characterized by low density, enhanced forbidden-line ratios, and a significant contribution to Hα emission.
• Observational methods classify DIG using Hα equivalent width and emission-line ratios, with surveys showing that DIG may contribute 30–60% of galaxies’ total Hα flux.
• Numerical models and simulations reveal that DIG arises from multiple ionizing sources—leaked radiation from H II regions, HOLMES, and shocks—affecting nebular diagnostics and star-formation rate estimates.

Diffuse ionized gas (DIG) is the faint, extended warm ionized component of the interstellar medium outside classical H II regions, often identified with the warm ionized medium in the Milky Way. In galaxies it occupies disks, interarm regions, bulges, extraplanar layers, and stripped tails, and it contributes a substantial fraction of the nebular emission budget. Relative to classical H II regions, DIG is characterized by lower density, lower surface brightness, and line ratios that are often enhanced in low-ionization forbidden lines such as $[\mathrm{N\,II}]$, $[\mathrm{S\,II}]$, $[\mathrm{O\,I}]$, and sometimes $[\mathrm{O\,II}]$ relative to Balmer recombination lines. Because integrated galaxy spectra mix DIG and H II-region emission, DIG is central to the interpretation of diagnostic diagrams, star-formation rates, and gas-phase metallicities (Asari et al., 2020, Lacerda et al., 2017).

1. Physical definition and astrophysical environments

DIG is observed in multiple galactic environments: early-type galaxies, bulges of late-type galaxies, interarm regions of disks, and gas layers above and below galactic planes. In the Milky Way context it is commonly referred to as the warm ionized medium or diffuse ionized medium (Asari et al., 2020). In star-forming galaxies it is frequently described as warm gas with $T_e \sim 10^4$ K, lower density than classical H II regions, lower ionization parameter, lower emission-line equivalent widths, and lower surface brightness, while also exhibiting a harder ionizing spectrum in many circumstances (Mannucci et al., 2021).

The physical contrast with classical H II regions is not solely geometric. DIG tends to show enhanced collisionally excited lines relative to recombination lines, especially $[\mathrm{N\,II}]/\mathrm{H}\alpha$, $[\mathrm{S\,II}]/\mathrm{H}\alpha$, and often $[\mathrm{O\,I}]/\mathrm{H}\alpha$, while $[\mathrm{O\,III}]/\mathrm{H}\beta$ can behave non-universally depending on environment and ionizing source (Asari et al., 2020, Zhang et al., 2016). In the Milky Way disk, the warm ionized medium is described as having density $\sim 0.1$ and temperatures $[\mathrm{S\,II}]$0–$[\mathrm{S\,II}]$1 (Luisi et al., 2017). In ram-pressure stripped systems, DIG is described as warm ($[\mathrm{S\,II}]$2 K), low-density ($[\mathrm{S\,II}]$3) gas with vertical scale heights of $[\mathrm{S\,II}]$4–$[\mathrm{S\,II}]$5 kpc (Tomicic et al., 2020).

Survey-based estimates show that DIG is not a minor residual. In spiral galaxies it has been cited as contributing roughly 30–60% of the total H$[\mathrm{S\,II}]$6 emission (Asari et al., 2020). In the CALIFA analysis of 391 galaxies, DIG defined as hDIG+mDIG contributes on average 56% of the total H$[\mathrm{S\,II}]$7 flux, with strong morphology dependence: approximately $[\mathrm{S\,II}]$8% in E/S0, $[\mathrm{S\,II}]$9% in Sa–Sb, and $[\mathrm{O\,I}]$0% in later types for $[\mathrm{O\,I}]$1 (Lacerda et al., 2017). Other methodologies yield similarly broad but survey-dependent ranges, including 20%–90% of integrated disk flux in GASP galaxies and 40%–70% of total H$[\mathrm{O\,I}]$2 in the BETIS showcase sample (Tomicic et al., 2020, González-Díaz et al., 2023).

A simulation-based definition has also been proposed. In a high-resolution isolated Milky Way-like simulation, the ionized gas shows a bimodal electron-density distribution with peaks at $[\mathrm{O\,I}]$3 and $[\mathrm{O\,I}]$4, motivating a threshold $[\mathrm{O\,I}]$5 to separate DIG from H II-region gas (McClymont et al., 2024). This suggests that DIG may be physically distinct from classical H II regions rather than merely the low-surface-brightness tail of a continuous distribution.

2. Ionization sources and internal diversity

DIG is not a single ionization phenomenon. Proposed ionizing or heating mechanisms include leakage of ionizing photons from star-forming regions, hot low-mass evolved stars (HOLMES), shocks, turbulence dissipation, cosmic rays, dust-scattered light, old supernova remnants, magnetic reconnection, and turbulent mixing layers (Asari et al., 2020, Tomicic et al., 2020). A useful tripartite distinction identifies “leaking-DIG,” ionized by photons escaping from H II regions; “HOLMES-DIG,” ionized by old post-AGB-like stars and commonly associated with LIER-like line ratios; and shocked DIG (Mannucci et al., 2021).

In actively star-forming galaxies, several studies argue that leaked radiation from young stars is energetically dominant. PHANGS-MUSE data support a two-component picture in which an energetically dominant DIG component is powered by Lyman-continuum photons leaking from H II regions, while a harder but energetically weaker component from hot low-mass evolved stars is required to explain enhanced low-ionization ratios in central regions, flat or decreasing $[\mathrm{O\,I}]$6 with $[\mathrm{O\,I}]$7, and offsets into LI(N)ER-like regions of BPT space. In that analysis, HOLMES contribute about 2% of the galaxy-integrated H$[\mathrm{O\,I}]$8 emission but are fundamental contributors to $[\mathrm{O\,I}]$9 emission (Belfiore et al., 2021).

A different but partly convergent picture emerges from galaxy-scale simulation. In the 2024 isolated-galaxy calculation, DIG is primarily ionized by stars aged $[\mathrm{O\,II}]$0–$[\mathrm{O\,II}]$1 Myr that become directly exposed to low-density gas after H II regions have been cleared; leakage from recently formed stars younger than 5 Myr is only moderately important. The model attributes DIG line-ratio trends to increasing temperature and a hardening radiation field at lower $[\mathrm{O\,II}]$2, with the hardening driven by the age shift in the dominant ionizing population and by the harder intrinsic spectra of $[\mathrm{O\,II}]$3–$[\mathrm{O\,II}]$4 Myr stars in BPASS-like binary-evolution models (McClymont et al., 2024). This suggests that recent star formation alone can explain much of the DIG phenomenology in normal star-forming disks, though it does not invalidate the observational evidence for HOLMES in bulges and low-EW environments.

Environmental dependence is substantial. In stripped tails, DIG with strong $[\mathrm{O\,II}]$5 excess, high LIER/LINER incidence, and projected distances up to 10 kpc from star-forming regions is interpreted as at least partly ionized by processes other than star formation, probably mixing, shocks, and accretion of intracluster and interstellar gas (Tomicic et al., 2021). In two local analogues of high-redshift star-forming galaxies, DIG is interpreted as most likely dominated by photon leakage from H II regions with additional contributions from feedback-driven shocks (Lagos et al., 27 Jan 2026). The aggregate picture is therefore plural: leaked young-star radiation, intermediate-age exposed populations, HOLMES, and shocks all appear in the literature, but their relative roles vary strongly with galaxy type and location within galaxies.

3. Observational signatures and classification schemes

The most widely used practical DIG classifier in resolved optical spectroscopy is the H$[\mathrm{O\,II}]$6 equivalent width, $[\mathrm{O\,II}]$7, defined as

$[\mathrm{O\,II}]$8

Because it measures nebular emission relative to the local stellar continuum, it links line emission to the underlying stellar population and behaves as an intensive quantity under line-of-sight superposition (Lacerda et al., 2017).

In CALIFA-based work, three nebular regimes are defined by $[\mathrm{O\,II}]$9 (Lacerda et al., 2017, Asari et al., 2020):

Regime	Criterion	Interpretation
hDIG	$T_e \sim 10^4$0	HOLMES-dominated diffuse gas
mDIG	$T_e \sim 10^4$1	Mixed diffuse regime
SFc	$T_e \sim 10^4$2	Star-forming complexes

The $T_e \sim 10^4$3 threshold is physically linked to the “retired galaxy” interpretation and to the low-$T_e \sim 10^4$4 peak near $T_e \sim 10^4$5, whereas the $T_e \sim 10^4$6 threshold is empirical and marks the mode of the high-$T_e \sim 10^4$7 population (Lacerda et al., 2017). Low-$T_e \sim 10^4$8 zones have $T_e \sim 10^4$9 of order unity, supporting the idea that HOLMES can energetically power them (Lacerda et al., 2017).

This classifier maps cleanly onto excitation diagrams in many resolved data sets: SFc zones occupy the star-forming wing of the BPT plane, hDIG zones populate the LINER/LIER-like tip of the right wing, and mDIG forms the bridge between them (Lacerda et al., 2017). The same framework underlies later MaNGA DIG-correction work, which is formally restricted to spectra with observed $[\mathrm{N\,II}]/\mathrm{H}\alpha$0 because more weakly emitting spectra are already DIG-dominated (Asari et al., 2019).

Surface-brightness criteria remain common but are contested. A threshold such as $[\mathrm{N\,II}]/\mathrm{H}\alpha$1 has been used to select H II-dominated emission (Mannucci et al., 2021), yet CALIFA-based analysis argues that $[\mathrm{N\,II}]/\mathrm{H}\alpha$2 is conceptually flawed because it behaves as an extensive projected quantity and can misclassify bright bulge DIG as star-forming while missing faint outer-disk star-forming regions (Lacerda et al., 2017). The critique becomes even sharper in ram-pressure stripped systems: in GASP, neither a single H$[\mathrm{N\,II}]/\mathrm{H}\alpha$3 threshold, nor a single $[\mathrm{N\,II}]/\mathrm{H}\alpha$4 threshold, nor an $[\mathrm{N\,II}]/\mathrm{H}\alpha$5 threshold was found to separate DIG-dominated from non-DIG emission reliably at kpc resolution (Tomicic et al., 2020).

More elaborate classifications have therefore been developed. GASP derives a spaxel-level DIG fraction $[\mathrm{N\,II}]/\mathrm{H}\alpha$6 by combining attenuation-corrected $[\mathrm{N\,II}]/\mathrm{H}\alpha$7 with metallicity-corrected $[\mathrm{N\,II}]/\mathrm{H}\alpha$8, using

$[\mathrm{N\,II}]/\mathrm{H}\alpha$9

with galaxy-specific $[\mathrm{S\,II}]/\mathrm{H}\alpha$0 and $[\mathrm{S\,II}]/\mathrm{H}\alpha$1 (Tomicic et al., 2020). BETIS instead combines adaptive binning based on $[\mathrm{S\,II}]/\mathrm{H}\alpha$2, morphological H II-region masks from pyHIIextractor, and an additional residual H$[\mathrm{S\,II}]/\mathrm{H}\alpha$3 surface-brightness threshold at $[\mathrm{S\,II}]/\mathrm{H}\alpha$4 to define lower and upper DIG limits over a wide range of spatial resolutions (González-Díaz et al., 2023). These developments underscore that DIG identification is operational and data-model dependent rather than fixed by a single universal observable.

4. Spatial structure, disk–halo connection, and kinematics

DIG is both a planar and an extraplanar phenomenon. In several edge-on spirals it forms vertically extended layers with kpc-scale structure, while in face-on systems it fills interarm and outer-disk regions and often dominates low-surface-brightness projected area (Voigtländer et al., 2013, Lacerda et al., 2017). CALIFA finds hDIG prevalent in ellipticals, S0s, bulges, and extraplanar regions, while the SF/mDIG proportion grows from early- to late-type spirals and from inner to outer radii (Lacerda et al., 2017).

Kinematic and structural modeling of NGC 4666 provides a detailed eDIG example. Its DIG is best fit by two ionized components: a thin disk with scale height $[\mathrm{S\,II}]/\mathrm{H}\alpha$5 kpc and a thick disk with scale height $[\mathrm{S\,II}]/\mathrm{H}\alpha$6 kpc, with a thin/thick flux ratio of 2.5, inclination $[\mathrm{S\,II}]/\mathrm{H}\alpha$7, radial scale length $[\mathrm{S\,II}]/\mathrm{H}\alpha$8 kpc, truncation radius $[\mathrm{S\,II}]/\mathrm{H}\alpha$9 kpc, general linewidth $[\mathrm{O\,I}]/\mathrm{H}\alpha$0, and maximum rotation velocity $[\mathrm{O\,I}]/\mathrm{H}\alpha$1 (Voigtländer et al., 2013). The same study identifies minor-axis line splitting with a component blueshifted by about $[\mathrm{O\,I}]/\mathrm{H}\alpha$2, interpreted as outflowing ionized gas, and concludes that enhanced star formation both drives outflow and maintains a Reynolds-layer-like thick ionized disk (Voigtländer et al., 2013).

The morphology of DIG is not universal. In NGC 4013 and NGC 4302 the extraplanar DIG is dominated by a smooth, diffuse component, while high-resolution dust images reveal strongly filamentary absorption structures with no counterpart in the H$[\mathrm{O\,I}]/\mathrm{H}\alpha$3 morphology. The conclusion is that the thick-disk DIG and the dusty extraplanar filaments trace physically distinct phases of the thick-disk ISM, with the latter representing denser neutral material (Rueff et al., 2013).

Milky Way studies likewise show that DIG encodes both galactic structure and disk–halo transport. In the first Galactic quadrant, Green Bank Telescope radio recombination-line data reveal two dominant DIG velocity components centered around $[\mathrm{O\,I}]/\mathrm{H}\alpha$4 and $[\mathrm{O\,I}]/\mathrm{H}\alpha$5, with the higher-velocity component associated with W43 and the lower-velocity component interpreted either as gas at a different distance or as bar-driven streaming near W43 (Luisi et al., 2017). In the anti-center, a LAMOST sample of 17,821 DIG spectra shows that $[\mathrm{O\,I}]/\mathrm{H}\alpha$6 and $[\mathrm{O\,I}]/\mathrm{H}\alpha$7 are enhanced in the interarm region between the Local and Perseus arms near $[\mathrm{O\,I}]/\mathrm{H}\alpha$8 kpc, while $[\mathrm{O\,I}]/\mathrm{H}\alpha$9, $[\mathrm{O\,III}]/\mathrm{H}\beta$0, and $[\mathrm{O\,III}]/\mathrm{H}\beta$1 all increase with $[\mathrm{O\,III}]/\mathrm{H}\beta$2 (Wen et al., 2024).

Cluster environments add a further DIG regime. In stripped tails, DIG-dominated regions can lie at projected distances $[\mathrm{O\,III}]/\mathrm{H}\beta$3 kpc and up to 10 kpc from star-forming regions, with high $[\mathrm{O\,III}]/\mathrm{H}\beta$4 and a high fraction of LIER/LINER-like emission (Tomicic et al., 2021). This suggests that stripped tails probe ionization channels—mixing, shocks, and ICM/ISM interaction—that are harder to isolate in normal disks.

5. Consequences for diagnostic diagrams, metallicity, and star-formation rates

DIG directly affects optical diagnostic ratios. In MaNGA, DIG-dominated low-$[\mathrm{O\,III}]/\mathrm{H}\beta$5 regions display enhanced $[\mathrm{O\,III}]/\mathrm{H}\beta$6, $[\mathrm{O\,III}]/\mathrm{H}\beta$7, $[\mathrm{O\,III}]/\mathrm{H}\beta$8, and $[\mathrm{O\,III}]/\mathrm{H}\beta$9 at fixed metallicity, and DIG contamination moves H II regions toward composite or LI(N)ER-like areas of BPT space (Zhang et al., 2016). Standard H II-region photoionization grids fail to reproduce these DIG line ratios, and leaky H II-region models shift them only slightly, favoring a harder ionizing spectrum for DIG with LI(N)ER-like emission (Zhang et al., 2016).

The impact on metallicity diagnostics is strongly calibrator-dependent. In the MaNGA-based DIG-correction framework built from 1,409 star-forming galaxies, the DIG effect is negligible for O3N2 but reaches $\sim 0.1$0 dex at the high-metallicity end for N2, and the real effect is expected to be larger than the measured one because MaNGA’s $\sim 0.1$1kpc resolution still mixes H II regions and DIG (Asari et al., 2019). Applying line-by-line DIG corrections to SDSS galaxies modifies the inferred $\sim 0.1$2–$\sim 0.1$3–SFR relation, especially at high stellar mass, and with the N2 indicator the corrected relation yields oxygen abundance increasing with SFR at high mass, contrary to previous claims (Asari et al., 2019).

Earlier MaNGA work reached a closely related but more categorical conclusion about calibrators. Metallicities derived using N2O2 are described as optimal because they exhibit the smallest bias and error, whereas O3N2, $\sim 0.1$4, N2, and the Dopita et al. (2016) N2S2H$\sim 0.1$5 calibration can introduce biases in derived metallicity gradients as large as the gradient itself. The strong-line method IZI cannot be applied to DIG accurately because it contains only H II-region models (Zhang et al., 2016). In the same study, DIG-enhanced N2 can overestimate metallicity by roughly $\sim 0.1$6 dex in an illustrative case (Zhang et al., 2016).

DIG also biases star-formation rates when total H$\sim 0.1$7 is interpreted as massive-star emission. The 2020 review emphasizes that DIG can contribute a large fraction of total H$\sim 0.1$8, so treating all H$\sim 0.1$9 as star-formation emission distorts or overestimates the inferred SFR, especially when old stars ionize part of the gas (Asari et al., 2020). In the GASP analysis, removing DIG lowers SFR estimates by about 0.2 dex, although the shift is similar in stripped and control samples (Tomicic et al., 2020). DIG fractions in GASP disks span $[\mathrm{S\,II}]$00, and $[\mathrm{S\,II}]$01 anti-correlates with both sSFR and $[\mathrm{S\,II}]$02 (Tomicic et al., 2020).

The extent of DIG contamination in integrated spectra of bright star-forming galaxies remains debated. One line of argument emphasizes that DIG is a serious contaminant for metallicity, SFR, and AGN classification (Asari et al., 2020, Zhang et al., 2016). Another argues that in normal bright star-forming galaxies the effect on integrated spectra is smaller than previously claimed because much of the apparent discrepancy between local H II-region slit spectra and galaxy-integrated spectra is actually an aperture effect. Using MUSE data for 11 bright H II regions in three nearby galaxies, Mannucci et al. show that when large apertures are used while still masking DIG-dominated spaxels, $[\mathrm{S\,II}]$03 and $[\mathrm{S\,II}]$04 change little, whereas sulfur ratios shift strongly and the DIG-free large-aperture H II-region spectra move into agreement with SDSS and MaNGA galaxy loci (Mannucci et al., 2021). The strongest practical implication is that sulfur-based discrepancies do not by themselves prove DIG contamination; aperture mismatch and internal H II-region stratification must be considered first (Mannucci et al., 2021).

6. Numerical modeling, synthesis, and current debates

Radiation-hydrodynamic and post-processed radiative-transfer models now reproduce several canonical DIG properties, but they do so with differing physical emphases. Idealized radiation-hydrodynamic simulations of disk patches show that photoionization feedback can drive low levels of turbulence in the dense disk and supply thermal pressure support for an extended diffuse ionized layer, creating a natural coupling between the ionizing photon budget and the mass in different ionization phases (Vandenbroucke et al., 2019). In that framework, photoionization raises the gas temperature from $[\mathrm{S\,II}]$05 K to $[\mathrm{S\,II}]$06 K and lowers the mean molecular weight, thereby increasing pressure support enough to build a warm quasi-hydrostatic layer (Vandenbroucke et al., 2019).

Cosmic rays address a related but distinct problem: maintaining the large vertical scale heights of DIG. Post-processing of SILCC simulations shows that cosmic-ray feedback produces more extended gaseous disks and H$[\mathrm{S\,II}]$07 DIG scale heights of $[\mathrm{S\,II}]$08–$[\mathrm{S\,II}]$09 kpc, far larger than thermal-feedback-only models and closer to observed values. However, adding a fiducial cosmic-ray heating term increases temperature with height but fails to reproduce observed nitrogen and sulfur forbidden-line intensities; the required heating must affect gas over a broader density range, or the total ionizing luminosity must be fine-tuned so that the ionizing spectrum hardens appropriately with height (Vandenbroucke et al., 2018).

A newer isolated-galaxy simulation with on-the-fly radiative transfer and non-equilibrium thermochemistry reaches a different synthesis. It reproduces observed correlations of $[\mathrm{S\,II}]$10, $[\mathrm{S\,II}]$11, $[\mathrm{S\,II}]$12, and $[\mathrm{S\,II}]$13 with $[\mathrm{S\,II}]$14, and it attributes these trends to increasing temperature and a hardening radiation field with decreasing $[\mathrm{S\,II}]$15. In that model the DIG contributes only 26% of the intrinsic H$[\mathrm{S\,II}]$16 but 60% of the observed H$[\mathrm{S\,II}]$17, because about $[\mathrm{S\,II}]$18 of H$[\mathrm{S\,II}]$19 from H II regions is absorbed by dust versus only $[\mathrm{S\,II}]$20 for DIG emission (McClymont et al., 2024). This suggests that observational DIG prominence can be amplified by differential attenuation even when its intrinsic recombination budget is smaller.

Current observational synthesis remains two-sided rather than uniform. PHANGS-MUSE argues for “two DIGs,” with leaked H II-region radiation dominating the DIG H$[\mathrm{S\,II}]$21 energy budget and HOLMES required for harder line-ratio phenomena (Belfiore et al., 2021). The 2024 simulation instead argues that ongoing star formation, specifically stars aged $[\mathrm{S\,II}]$22–$[\mathrm{S\,II}]$23 Myr, can account for both the low-ionization and the $[\mathrm{S\,II}]$24 trends without invoking secondary ionization sources in normal star-forming disks (McClymont et al., 2024). A plausible synthesis is that the dominant ionization channel depends on galaxy type and environment: young-star leakage and exposed intermediate-age populations in actively star-forming disks, HOLMES in bulges and low-EW systems, and shocks or mixing in outflows and stripped tails (Mannucci et al., 2021, Tomicic et al., 2021).

The main conceptual convergence is that DIG is neither negligible nor homogeneous. It is a structurally extended, spectroscopically distinct, and environmentally sensitive component of the ionized interstellar medium. Its observational definition remains operational, its inferred ionization source depends on spatial scale and host environment, and its impact on nebular diagnostics ranges from secondary to dominant depending on whether the target is a bright star-forming disk, a bulge-dominated system, or a low-surface-brightness extraplanar or stripped-gas structure (Asari et al., 2020, Mannucci et al., 2021).