GASP XXXV: Characteristics of the Diffuse Ionised Gas in Gas-stripped Galaxies

For this paper, as in Vulcani et al. (2018a), we consider galaxies with gas-stripping features (hereafter referred to as stripping galaxies) and compare them to undisturbed galaxies that do not show any sign of interactions or gas stripping on their Hα maps (referred to as control-sample galaxies). Galaxies that are tidally interacting with others, merging, lopsided, or cosmic-web-enhanced are not included in this paper (Vulcani et al. 2017, 2018b, 2018c, 2019a, 2020, 2021).

For each galaxy, B. M. Poggianti et al. (2021, in preparation) have designated a specific number indicating a stage of stripping referred to here as J-stage number (Table 1). The J-stage numbers classify non-stripped galaxies as 0, galaxies in early stages of stripping as 0.5, clear gas-stripped galaxies and jellyfishes as 1 and 2, respectively, and galaxies with a truncated Hα disk as 3. Truncated galaxies are galaxies where the gas disk was mostly stripped, except in the galactic center (Koopmann & Kenney 2004; Fritz et al. 2017). We will use this classification system in this paper to compare galaxies at different stripping phases.

Both stripping and control samples are drawn from Vulcani et al. (2018a) with some modifications. The stripping sample is composed of 41 galaxies. With respect to Vulcani et al. (2018a), we exclude JO149 and JO95 since we are not able to measure their effective galactocentric radii and orientation (Franchetto et al. 2020). Furthermore, we add JO93 (a member of the control sample in Vulcani et al. 2018a) because its Hα map indicates an initial phase of stripping.

The control sample includes 30 galaxies, 16 of which are cluster members and 14 are field galaxies. With respect to Vulcani et al. (2018a), we exclude P19482, as Vulcani et al. (2019b) showed that the galaxy is not undisturbed but undergoing cosmic-web enhancement. The control-sample galaxies by definition have no tails.

The galaxies were observed by MUSE, an IFU at ESO-VLT (the Very Large Telescope of the European Southern Observatory), covering the optical range of 4800–9300 Å. The data analysis of the observed cubes follows established procedures described in Poggianti et al. (2017). Briefly, the observed spectrum in each spaxel of the cube was corrected for the effect of foreground dust extinction of the Milky Way, with corresponding E_B−V values for each galaxy in the line of sight (LOS) on-sky measured by Schlegel et al. (1998), considering the recalibration introduced by Schlafly & Finkbeiner (2011) and assuming the extinction curve of Cardelli et al. (1989) with R_V = 3.1. To account for seeing effects, the data were smoothed and convolved in the spatial dimension using a 5 × 5 pixel kernel, which corresponds to 1'', or 0.7–1.3 kpc, depending on the galaxy's redshift.

The convolved cubes were analyzed with the spectrophotometric code SINOPSIS (Fritz et al. 2017) to fit the stellar component of the spectra, and then with KUBEVIZ (Fossati et al. 2016) to fit the gas emission lines in the stellar-continuum subtracted spectra. The emission lines used in this work are: Hβ, [O iii]λ5007, [O i]λ6300 (referred hereafter as [O i]), Hα, [N ii]λ6584, and [S ii]λ6713,6731 (referred hereafter as [S ii]).

Maps of Hα surface brightness corrected for attenuation, labeled as ΣH α, corr, were calculated assuming the intrinsic Balmer line ratio Hα/Hβ = 2.86, ionized gas temperature of T ≈ 10⁴ K and case B recombination (Hummer & Storey 1987; Osterbrock & Martel 1992). We applied a cut in signal-to-noise S/N ≥ 4 for all emission lines in our analysis and for our results.

The boundaries of stellar disks were estimated using an isophote in the continuum maps that is 1σ above the average sky background noise and the center of the galaxies is designated to be the centroid of the brightest central region in the continuum map (Gullieuszik et al. 2020). We define the spaxels (with the ionized gas emission) outside the stellar disks as part of the galactic tails. The disk inclination with respect to the line of sight and the disk diameter were estimated by Franchetto et al. (2020) from the I-band images obtained by convolving the MUSE data cube spectra.

It is important to note that a clear separation between DIG and H ii regions requires a resolution that is not achieved by our observations. Thus we can only probe regions where H ii complexes and DIG are mixed together in the line of sight and measure the DIG emission fraction instead. Therefore, in this paper we use the terms "dense gas" and DIG, not H ii regions and DIG.

Two of the main quantities of interest in our analysis are the gas-phase metallicity and ionization parameter. These quantities were derived from the emission line ratios, [O iii]λ5007/[S ii] versus [N ii]λ6583/[S ii], using the PYQZ code (v0.8.2 version, Dopita et al. 2013; Vogt et al. 2015). We assume solar metallicity to be $12+\mathrm{log}({\rm{O}}/{\rm{H}})=8.69$. Details on how we estimate the gas-phase metallicity and log(q) for the GASP sample are described by Franchetto et al. (2020). log(q) is defined as the logarithm of the ratio of the hydrogen-ionizing photon density (in cm⁻³) over the local hydrogen gas density (in cm⁻³) multiplied by the speed of light (Kewley et al. 2006; Nagao et al. 2006; Kewley et al. 2019). Since it measures the ratio of ionization intensity and gas density, it may be used as an indication for the ratio of radiation and gas pressures (Yeh & Matzner 2012). We note that PYQZ uses photoionization models of line ratios that assume ionization with photons from SF regions.

The Baldwin, Phillips, and Terlevich diagnostic diagrams (BPT; Baldwin et al. 1981; Kewley et al. 2006) of emission line ratios are used to infer the dominating ionization mechanism, such as SFs, AGNs, or LINERs ⁸ /LIERs ⁹ . In this work, we only use the BPT diagram based on the [O i]λ6300 line (hereafter BPT-[O i], shown in Figure 1), and we removed spaxels dominated by AGN ionization. The [O i]/Hα line ratio is sensitive to the hardness of the ionizing radiation field (Kewley et al. 2006) and LINER/LIER-like spaxels may potentially indicate that the origin of ionization across the galaxy disks and tails can be other than SF regions.

Zoom In Zoom Out Reset image size

To quantify the difference in ionization source and variation in gas emission among the galaxies, for each spaxel we measure the difference between the measured $\mathrm{log}[{\rm{O}}\,{\rm\small{}}{\rm{I}}]/{\rm{H}}\alpha$ and the corresponding value on the BPT-[O i] line that separates the SF and LINER/LIER region, ¹⁰ at the measured log[Oiii]/Hβ. This value is called Δ log[O i]/Hα and is shown in Figure 1. Negative Δ log[O i]/Hα values represent regions defined as SF, while positive values represent LINER/LIER-like regions. We note that negative Δ log[O i]/Hα does not necessarily mean that the ionization is coming only from SF regions, but that these account for the largest fraction. Similarly, LINER/LIER-like regions do not exclude a possibility of partial ionization by star formation. The individual BPT diagrams of spatially resolved data for 16 GASP galaxies were shown by Poggianti et al. (2019b).

In GASP XXXII, we estimated the fraction of Hα emission coming from DIG for all 71 galaxies studied in this work. The initial assumption was that the DIG exhibits lower H α_corr flux surface density (Σ Hα) and higher [S ii]/Hα line ratio than the dense gas (Blanc et al. 2009; Kaplan et al. 2016). We account for the radial gas-phase metallicity distribution across the galaxy, where we divided the [Sii]/Hα ratio by the gas metallicity value at each given radial bin. Then we fitted the spaxel-by-spaxel anti-correlation between the [S ii]/Hα ratio and the extinction-corrected ΣHα_corr for each individual galaxy. For this fitting, we designated the spaxels with the highest [Sii]/Hα ratio and lowest ΣHα_corr to have C_DIG = 1, while the data with the highest ΣHα_corr were assumed to have low C_DIG fraction. We then derived the relation between the DIG fraction and ΣHα_corr and created maps of C_DIG from the ΣHα_corr maps. Further details of the method can be found in GASP XXXII.

In principle, considering metallicity as a parameter in our method to estimate C_DIG could introduce a bias in our analysis of C_DIG versus metallicity. However, at a given radius, the metallicity correction affects the [S ii]/Hα of dense-gas and DIG spaxels equally, thus not changing the slope of the spaxel-by-spaxel anti-correlation in the diagrams. Furthermore, our method uses both [S ii]/Hα and Hα, thus mitigating strong effects of metallicity on C_DIG and lowering the scatter of the data. Thus, the C_DIG values in spaxels at similar radii are not significantly affected by variations in metallicity.

In what follows, to clearly contrast regions whose emission is generated by different mechanisms, we will consider dense-gas-dominated spaxels those with C_DIG ≤ 0.3, DIG-dominated spaxels those with C_DIG > 0.7, and disregard the spaxels with intermediate values.

We present the maps of C_DIG, Δ log[O i]/Hα, and log(q) values for control (A3376_B_0261) and stripped galaxies (JO175, JO85) in Figure 2, and all other galaxies in the figure set. We indicate the control and stripped samples with thin and thick line edges of panels, respectively. We will further quantify the trends of Δ log[O i]/Hα and log(q) as a function of galactocentric radius in Section 3.4.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (18 panels)

Tarball of figure set images

The maps of C_DIG (left panels of Figure 2 and the figures in the Appendix) reveal some interesting trends: in most cases, the dense gas is confined to the galaxy's core and spiral arms, and DIG emission is preferentially found between the arms, in the galaxy's outskirts, and in tails of stripped galaxies. The maps are lacking Δ log[O i]/Hα values (middle panels) in many spaxels compared to the C_DIG map due to a low S/N of [O i] emission line.

Comparing the maps of C_DIG and those of Δ log[O i]/Hα and log(q), we recover a tendency for the dense-gas-emission-dominated areas to have higher ionization parameters and lower Δ log[O i]/Hα than the DIG-dominated areas. Moving from galaxy cores to the outskirts and tails, the incidence of LINER/LIER-like spaxels (higher Δ log[O i]/Hα values) increases and log(q) decreases.

Figure 3 compares the spatially resolved Δ log[O i]/Hα for dense-gas-dominated spaxels (blue, filled histograms) and DIG-dominated spaxels (red, hatched histogram), for different regions (disks and tails) of the control and stripped galaxies. We remind readers that control sample galaxies have no tails by definition. The median values of the corresponding distributions are represented by the blue and red vertical lines for the dense-gas and DIG-dominated spaxels, respectively.

Zoom In Zoom Out Reset image size

While dense gas in disks of the control galaxies shows median Δ log[O i]/Hα ≈ − 0.5, disks and tails of the stripped galaxies have median Δ log[O i]/Hα of ≈ − 0.4 and ≈ − 0.3 respectively. In general, the DIG-dominated spaxels have ≈ 0.4 dex higher median Δ log[O i]/Hα than dense-gas areas. The DIG spaxels in the disks show more LINER-like spaxels (approximately half of the spaxels), while the stripped tails show the majority (83%) of spaxels with positive Δ log[O i]/Hα.

In star-forming galaxies, DIG can reach scale heights of 1 − 2 kpc (Reynolds 1984; Haffner et al. 2009; Bocchio et al. 2016; Tomičić et al. 2017), which is larger than the thickness of stellar disks (up to a few hundreds of pc), favoring the hypothesis that DIG could require ionization by sources other than star formation (i.e., older stellar populations or mixing of ISM and ICM). Detecting DIG at large distances from SF regions would validate that hypothesis.

In this section we consider only the DIG-dominated spaxels and characterize their projected distance ¹¹ from SF regions (designated by the BPT-[O i] diagram). For each individual non-SF spaxel, we measure the mean distance from the closest 10 SF spaxels. Figure 4 presents Δ log[O i]/Hα values of spaxels in all galaxies as a function of the projected distance from the SF spaxels.

Zoom In Zoom Out Reset image size

As in Figure 3, we show data for different types of galaxies (control versus stripped) and their different parts (disks versus tails). In general, we notice a positive trend between Δ log[O i]/Hα and distance. In the stripped tails and in some parts of the stripped galaxies' disks, there are regions that span large distances from the SF spaxels (>2 kpc and up to 10 kpc), and which have a mean Δ log[O i]/Hα value of ≈ 0.4. These spaxels, which exhibit high Δ log[O i]/Hα ≈ 0.4 and are distant from the SF regions, correspond to highly stripped galaxies (JO93, JW100, JO204, JO194, JO147, JO85, JO60, JO27), of which the edge-on JW100 with its long tail contributes the most (approximately half of those spaxels).

From these results, we conclude that DIG in stripped galaxies extends to very large distances from SF regions, which would make its ionization by SF regions alone less probable.

Next, we test if regions dominated by DIG and dense-gas emission in disks of both control and stripped galaxies are also characterized by different values of gas-phase metallicity, log(q), and Δ log[O i]/Hα, at each given galactocentric distance.

We divided the spaxels into annuli of different deprojected galactocentric radii (normalized by the effective radius of the disk; Franchetto et al. 2020). In each radial annulus, we measured the median of all quantities as well as differences between median values of DIG and dense-gas-dominated spaxels. We present in Figure 5 the absolute values (left panels) and the differences between median values of dense gas and DIG (right panels) as a function of radial annuli. The figure shows each galaxy separately as a data point, color-coded by its stellar mass (blue for ${\rm{log}}\ {M}_{* }\lt 10.3$ and orange for ${\rm{log}}\ {M}_{* }\geqslant 10.3$).

Zoom In Zoom Out Reset image size

For the gas-phase metallicity (upper panels), the absolute values decrease with radius, as expected from the fact that metallicity gradients are negative for the great majority of galaxies (Mingozzi et al. 2020, A. Franchetto et al. submitted). The low-mass galaxies have lower metallicities, following the known mass–metallicity relation (Franchetto et al. 2020). A small difference ( < 0.05 dex) between the DIG- and dense-gas-dominated spaxels is detected, but no variation of this difference with galactocentric radius is found. The scatter of the data ( ≈ 0.5 dex) and the uncertainty of individual data points ( ≈ 0.1 dex) are larger than the difference in metallicities between dense gas and DIG.

The ionization parameter (middle panels) does not show a clear trend with radius for low-mass galaxies but shows a clear decrease with radius for the high-mass galaxies. The difference in log(q) between DIG and dense-gas areas shows a negative offset, with DIG data having ≈ 0.1 dex lower values. Furthermore, the log(q) offset between DIG and dense gas increases with galactocentric radius.

Mean absolute values of Δ log[O i]/Hα (bottom panels) increase with radius for galaxies of all masses by ≈ 0.2 dex. At a given radius, the high-mass galaxies tend to have slightly lower values compared to the low-mass galaxies. The offset in Δ log[O i]/Hα indicates that DIG-dominated regions have 0.2–0.3 dex higher Δ log[O i]/Hα values compared to dense-gas-dominated regions and the mean value of this difference increases with radius.

These results indicate that metallicity is not affected by the DIG fraction. The median log(q) of total gas increases in the centers of high-mass galaxies, with DIG fraction playing a small role in log(q). The median Δ log[O i]/Hα of total gas is increasing toward the edges of all galaxies, with the DIG fraction potentially playing an important role.

We investigate in Figure 6 different physical properties such as the integrated fraction of DIG emission ¹² (C_DIG,int, GASP XXXII) as a function of J stage (left panel), median log(q) (middle panel), and Δ log[O i]/Hα (right panel). We estimated the C_DIG,int within the disks and tails of all galaxies (control-sample and stripped galaxies). In the case of median log(q), we use the median value of only the spaxels designated as SF on the BPT-[O i] diagram.

Zoom In Zoom Out Reset image size

As seen in GASP XXXII, the disks of stripped and control-sample galaxies cover a similar range in C_DIG,int (from 0.1 to 0.9), with a median value ¹³ ${C}_{\mathrm{DIG},\mathrm{int}}={0.44}^{{| }_{0.33}^{0.52}}$ for the entire sample. The tails of stripped galaxies have higher DIG fractions (median ${C}_{\mathrm{DIG},\mathrm{int}}={0.83}^{{| }_{0.69}^{0.99}}$) than the disks.

There is generally no systematic variation of the C_DIG,int of disks with J stage. Truncated disks (J = 3) show a low DIG fraction ( ≈ 0.2), probably due to a large fraction of gas being stripped away from the disk, and preservation of only the dense-gas, star-forming regions in their centers.

Focusing on the ionization parameter, we find that control galaxies have lower median values ($\mathrm{log}({\rm{q}})={6.93}^{{| }_{6.88}^{7.0}}$) compared to the stripped galaxies (median $\mathrm{log}({\rm{q}})={7.03}^{{| }_{6.94}^{7.17}}$). The disks and tails of the stripped sample also show the same median values. We note that the control sample spans a narrower range of log(q) than the disks and tails of stripped galaxies.

The gas in disks of control and stripped sample galaxies has mostly negative Δ log[O i]/Hα values (median $-{0.22}^{{| }_{-0.31}^{-0.12}}$ and $-{0.09}^{{| }_{-0.2}^{0.02}}$, respectively). The tails have higher values in Δ log[O i]/Hα than the disks with more than half of them having positive values (median ${\rm{\Delta }}\,{{\rm{log}}}_{10}$ [O i]/H$\alpha ={0.0}^{{| }_{-0.06}^{0.12}}$).

We also notice a weak correlation between Δ log[O i]/Hα and C_DIG,int across stripped galaxies (Pearson coefficient ρ = 0.33 ± 0.00), while the control galaxies show a slight anti-correlation (ρ = − 0.24 ± 0.2). This effect is mostly due to the tails of high C_DIG,int, which have slightly higher Δ log[O i]/Hα.

These results show that the gas-stripping process, in general, slightly increases the log(q) range and systematically increases the values of Δ log[O i]/Hα, especially in the tails.

To investigate if log(q) and Δ log[O i]/Hα vary with integrated galactic properties, we compare galaxy global values (medians of spaxels) of log(q) and Δ log[O i]/Hα as a function of stellar mass, SFR, specific SFR (sSFR = SFR/M_⋆), and other overall galaxy properties. The integrated values of log(q) and Δ log[O i]/Hα as a function of stellar mass are presented in Figures 7 and 8, respectively. The comparison with other properties (SFR and sSFR of disks, gas-phase metallicity, ratio of area of tails and disks) is shown in the Appendix.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We separate panels and compare the median values of data as follows: (1) disks of control and stripped galaxy samples (panels A and C), (2) disks and tails of the stripped galaxies (panels B and D), and (3) dense-gas-dominated (panels A and B) and DIG-dominated (C and D) areas in the galaxies. We show the difference between DIG and dense-gas data in panel E, and the difference between disks and tails of the stripped galaxies in panel F. From these comparisons, we observe a few major results and trends, as follows.

The low-mass galaxies (${\mathrm{log}}_{10}{M}_{* }\leqslant 10.3\,{M}_{\odot }$) are characterized by a rather constant value of log(q), while log(q) for the high-mass galaxies (${\mathrm{log}}_{10}{M}_{* }\gt 10.3\,{M}_{\odot }$) strongly correlate with stellar mass. DIG in disks and tails has 0.1 − 0.2 dex lower log(q) compared to the dense gas in the same environments. There is no clear difference in the evaluated gas properties in disks of control and stripped galaxies (panel E) and between the tails and disks of the stripped sample (panel F). The large scatter in the measured properties prevents a conclusive analysis of trends with stellar mass in panels E and F.

Considering Δ log[O i]/Hα of the dense gas, we observe a weak anti-correlation with stellar mass (Figure 8). Furthermore, this anti-correlation is stronger for low-mass galaxies than for high-mass ones. DIG areas show higher Δ log[O i]/Hα values ( ≈ 0.4 dex) compared to dense gas (as clearly seen in panel E), with the tails having larger offsets (by ≈ 0.1 dex) than the disks. We highlight the fact that all dense-gas data have Δ log[O i]/Hα values below 0, while most of the DIG show positive values. We also notice in panel D that galaxies of higher J-stage number have higher Δ log[O i]/Hα values for DIG-designated spaxels in the tails (magenta square symbols) compared to the rest of the data in other panels.

Regarding other galactic values (presented in the Appendix), log(q) correlates with SFR and gas-phase metallicity. We explain this as a signature of well-known correlations between the stellar mass, metallicity, and SFR. The dense gas shows a somewhat strong anti-correlation between Δ log[O i]/Hα and metallicity and DIG exhibits a weak correlation between Δ log[O i]/Hα and the spaxel fraction of tails versus disks. The former relation is most likely an effect of metallicity on the increase in [O i]/Hα ratio.

To conclude, these results indicate that the stellar mass is driving the increase in median log(q) across the galaxies more than the fraction of DIG. Moreover, DIG clearly shows high Δ log[O i]/Hα values compared to the dense gas, especially in the case of the stripped tails.

In the previous section we showed that gas-phase metallicity at a given galactocentric radius and at kpc spatial resolution does not vary with different DIG fractions in the disks of all galaxies (Figure 5). This would indicate two major conclusions about the distribution of the metallicity within the disks. First, DIG emission should not hinder the measurement of metallicity at a given radius in galaxies. Second, there is no difference in the metallicity radial gradient between areas dominated by dense gas or DIG.

Similar to our conclusions, Kumari et al. (2019) did not find significant differences in metallicities between spatially close-by Hii-DIG/LIER pairs, which are expected to be chemically homogeneous. However, their results may be affected by a non-SF source of ionization changing line ratios used for metallicity diagnostics.

On the other hand, Poetrodjojo et al. (2019) found tentative results for different metallicity gradients in DIG- and H ii-dominated spaxels of the M83 galaxy. Analyzing IFU data from MaNGA, Zhang et al. (2017) also showed that DIG can significantly bias the measurement of gas metallicity and metallicity gradients derived using strong emission lines. However, Zhang et al. (2017) used an incomplete method to separate dense gas from DIG, using only Σ Hα information. Lacerda et al. (2018) note that DIG in galaxy centers can exhibit bright ΣHα and low equivalent width, which could have biased the results of Zhang et al. (2017). Using data cubes from MaNGA and fiber spectroscopy from the Sloan Digital Sky Survey Data Release 7, Vale Asari et al. (2019) found that accounting for the DIG contribution leads to a change in metallicity of up to 0.1 dex, but can be negligible depending on the metallicity estimation method.

The major drawback of the methods of separating DIG and H ii-dominated spaxels used by Poetrodjojo et al. (2019) and Zhang et al. (2017), is that they did not fully account for the effects of radial metallicity gradients on the [S ii]/Hα ratio. On the other hand, a drawback of our method is that the low spatial resolution potentially can dilute the metallicity values of dense-gas spaxels, as discussed by Sánchez et al. (2014), Belfiore et al. (2017b), and Sánchez-Menguiano et al. (2018). This effect can lead to measuring similar metallicities between DIG and dense-gas spaxels in our sample.

Our results at small spatial scales ( ≈ 1 kpc) lead us to conclude that DIG exhibits lower log(q) than dense gas (Figures 5 and 7). Similar conclusions have been drawn by Zhang et al. (2017) and Mingozzi et al. (2020) from their spatially resolved observations of log(q) in nearby galaxies. This is to be expected if we assume that the diffuse gas resides further from hot, young stars that are the source of a large number of ionizing photons. The only exceptions at those spatial scales are DIG-dominated spaxels in the edges of disks and in the tails of gas-stripped galaxies, which show an increase in log(q).

Another interesting aspect of our results is that high-mass galaxies show an anti-correlation between log(q) and galactocentric radius (Figure 5). We ascribe this increase to ionization due to a large number of HOLMES and evolved stars (Flores-Fajardo et al. 2011; Zhang et al. 2017) at decreasing galactocentric radius. Similarly, Dopita et al. (2014) observed higher log(q) values in the center of galaxies, and concluded that it is due to higher ΣSFR and changes in the geometry of the molecular and ionized gas. Simpson et al. (2007) observed an increase in excitation of gas toward the galactic center, probably due to shocks. On the other hand, Mingozzi et al. (2020) observed a correlation between log(q) and galactocentric radius, in high-mass galaxies. Similarly, Sánchez et al. (2012) and Sánchez et al. (2015) noted a weak correlation between those parameters.

There is an ongoing discussion about the relation between the ionizing parameter in galaxies and other integrated galactic properties. At galactic scales, we do not observe a strong effect of gas stripping on log(q), but we do note an increase in the log(q) range of the stripped galaxies compared to the control sample (Figure 6). However, this is the first study where a strong correlation between log(q) and the stellar mass for galaxies with $\mathrm{log}{M}_{* }/{M}_{\odot }\gt 10.3$ (Figure 7) is observed. We ascribe this trend to a higher fraction of HOLMES and evolved stars within high-mass galaxies compared to low-mass galaxies. It is interesting that the tails also show an increase in log(q) with galaxy stellar mass, which may be a result of an interplay between gas and photon densities of the ISM. Integrated SFR and metallicities yield a positive correlation with log(q) probably due to their correlation with stellar mass. Similar to our observations, Dopita et al. (2014) observed a correlation between log(q) and SFR, while Poetrodjojo et al. (2018) did not.

Our results indicate an increase in Δ log[O i]/Hα toward the edges of galaxy disks (Figure 5) and a large fraction of DIG-dominated spaxels showing positive Δ log[O i]/Hα, designating them as LINER/LIER-like regions (Figure 3). In general, LINER/LIER-like features for DIG-dominated spaxels in the centers of non-stripped galaxies has been previously observed by Flores-Fajardo et al. (2011) and Zhang et al. (2017). Zhang et al. (2017) speculated that only ionization by evolved stars (older than 125 Myr) as a major ionization source for DIG can lead to LINER/LIER-like emission. The evolved stars and HOLMES may contribute to the increase in Δ log[O i]/Hα within the disks of our sample (as seen in Figures 5 and 3).

The DIG areas in the tails of highly stripped galaxies (higher J-stage numbers) systematically exhibit the highest Δ log[O i]/Hα values compared to the galactic disks (Figure 8). Δ log[O i]/Hα correlates with distance from the SF regions in the disks and most of the tails (Figure 4). We also found that DIG in stripped tails extends to very large projected distances (Figure 4), and is spanning kpc and up to 10 kpc scales, which is larger than typical scale heights of galactic disks. The DIG at those large distances contributes to the increase in Δ log[O i]/Hα and is showing LINER/LIER-like features. Those features are lacking in control galaxies. Previously, Poggianti et al. (2019b) observed that the tails of many gas-stripped galaxies in GASP exhibit LINER/LIER-like emission in BPT-[O i] diagrams, while the other diagnostic diagrams (based on NII or SII) suggest star formation or at least a composite origin for the ionization, which suggests that SF photons are partly the source of ionization.

In light of these results, we hypothesize that DIG in the edges and tails of stripped galaxies is ionized by more than one process (of which star formation may be one). Two competing candidates for that ionization are: (1) an older stellar population (older than 125 Myr and younger than 13 Gyr, as proposed by Flores-Fajardo et al. (2011) and Zhang et al. 2017) for ionization of DIG in galactic disks, or (2) mixing of hot–cold gas layers such as ICM and galactic ISM, as proposed by Slavin et al. (1993), Binette et al. (2009), and Poggianti et al. (2019a). The first option is less likely because the older stellar population would most likely form in SF knots of tails and would cover small distances from those SF regions. Although this candidate of ionizing DIG in tails should not be completely disregarded, we find the mixing of hot–cold gas layers and shocks more probable. This is also backed up by the findings of Poggianti et al. (2019a); Campitiello et al. (2021), and Müller et al. (2021), who found an excess in X-ray emission and the existence of magnetic fields in the tails of three of the jellyfish galaxies studied here, JO201, JW100, JO206. This X-ray excess may be produced by radiative cooling (Müller et al. 2021; Sparre et al. 2020) of the hot and highly magnetized ICM gas during the interplay with the galactic tail's ISM and accretion of the ICM onto the stripped ISM.

To conclude, we hypothesize that the gas-stripping process is creating galactic tails that are mixing with the ICM, thus ionizing the DIG at large distances from SF regions and systematically showing an increase in Δ log[O i]/Hα.