SDSS-IV MaNGA: The Nature of an Off-galaxy H_α Blob—A Multiwavelength View of Offset Cooling in a Merging Galaxy Group

The CFHT u-, g-, r, and i-band images are presented in Figures 4(a)—4(c), respectively. There are extended stellar halos surrounding the two galaxies, but we find no apparent optical counterpart directly from the images at the position of Totoro. The limiting magnitude³⁴ and the surface brightness of Totoro are listed in Table 3. Due to the absence of an optical counterpart at the position of Totoro, only upper limits can be placed on the surface brightnesses (i.e., limiting surface brightness of the observations).

Zoom In Zoom Out Reset image size

In Paper I, we used a multicomponent GALFIT (Peng et al. 2010) model to search for a stellar component of Totoro from a g-band image and found no substructure that is responsible for the blob. However, the residual of this complex parametric model still showed significant fluctuations that could hinder the detection of small scale substructure (see Figure 6 in Paper I). The ongoing interaction of the main system creates an extended stellar envelope and asymmetric structures that are challenging to the model. In this paper, we use three different approaches to further search for a potential stellar counterpart to Totoro from multiple-band (g, r, and i) images. All three methods are commonly used in the literature for background subtraction and for both compact and extended source detection.

First, we use the Python photometry tool sep (Barbary 2016)³⁵ to generate a "background" model of the image with a small background box size (5 pixels). The tool sep uses the same background algorithm as in SExtractor (Bertin & Arnouts 1996). We then subtract the "background" model from the image to increase the contrast around the main galaxies. This method has been used to separate galaxy stellar halos and the light from adjacent (background) objects (e.g., Huang et al. 2018; Rubin et al. 2018), and to identify faint, extended emission such as tidal tails (e.g., Mantha et al. 2019).

As a second approach, we subtract a blurred version of the image from the original one. The blurred version of the image is created by convolving the image with a circular Gaussian kernel (σ = 4 pixels). This procedure is part of the unsharp masking method in digital image processing³⁶ and can also increase the contrast of the image. Compared to the first approach, the Gaussian convolution makes it more sensitive to a low-threshold feature with a sharp edge. This method has been commonly used to identity H ii regions in galaxies (e.g., Rahman et al. 2011; Pan et al. 2020, in preparation) and to detect faint embedded spiral and bar features in early-type galaxies (e.g., Barazza et al. 2002; Kim et al. 2012). The method has been applied to IFS data for the later purpose as well by Gomes et al. (2016).

For our third method, we perform isophotal fitting using the Ellipse task in IRAF (Tody 1986, 1993). We first mask out all detected objects other than the main galaxy using sep. Then, we run Ellipse on the masked images, allowing the centroid and the shape of the isophote to vary. Using the resulting isophotal parameters, we create the corresponding 2D model using the bmodel task and subtract it from the input image. Compared to the GALFIT parametric model, the Ellipse one does not depend on the choice of model component and typically leads to smoother residuals. This approach has been routinely used to determine the morphology of elliptical and lenticular galaxies (e.g., Hao et al. 2006; Oh et al. 2017) and to search for low-surface-brightness tidal features in nearby galaxies (e.g., Tal et al. 2009; Gu et al. 2013).

The residual maps of these methods are shown in Figure 5. From left to right, the panels correspond to the analysis of the background, unsharp mask, and isophotal fitting methods; from top to bottom, we show the residual maps and annular residual profiles of the u, g, r, and i bands, respectively (u-band results will be discussed later in this section). The annular profiles are centered on the Hα peak of Totoro. We also explore different parameters adopted in these procedures (e.g., the size of the background box or the convolution kernel). The choice of these parameters within a reasonable range does not affect the results.

Zoom In Zoom Out Reset image size

Using these methods, we detect a large number of unresolved (point-like) sources on the g-, r- and i-band images, presumably a combination of globular clusters of the main system and background galaxies. The residual profiles are generally flat, fluctuating around the zero value. This is true for all gri bands and methods, suggesting that there is no sign of a distinctive stellar counterpart at the position of Totoro. Therefore, our new analyses confirm the previous result by using GALFIT in Paper I. There are two unresolved sources in the Totoro area (red circle in Figure 5), but we do not notice any increase or decrease in the number density around the Totoro area. The two sources are associated with neither Hα nor the CO peak. We will come back to the nature of these two unresolved sources later. On the residual maps from the unsharp mask method, we uncover a pair of "ripple"-like features close to the center of the main galaxy. These substructures remain the same when we vary the convolution kernel used, and they resemble the structure we saw on the GALFIT residual maps in Paper I. Such "ripples" often relate to recent galaxy interaction or the presence of dust (e.g., Colbert et al. 2001; Kim et al. 2012; Duc et al. 2015; Bílek et al. 2016). However, it is unclear whether there is any connection between them and Totoro.

Because there is no optical (g, r, and i) continuum counterpart found in Totoro, it is expected to be composed mostly of young stars if it is indeed a galaxy. In Paper I, we use the excitation state of optical lines to constrain the presence of young stars. However, the result is method (diagnostic diagram) dependent. The u-band luminosity of a galaxy is dominated by young stars of ages <1 Gyr; therefore, it is more sensitive to any recent star formation than any of the other broadband luminosities available (Moustakas et al. 2006; Prescott et al. 2009; Zhou et al. 2017) and is a more straightforward probe than optical emission-line diagnostics.

Similar to the gri bands, the u-band emission in the MaNGA hexagonal FoV is dominated by the dry merger (Satsuki and Mei) as shown in Figure 4, but the u-band data are less affected by the large stellar halos associated with the dry merger (Satsuki and Mei) than the redder bands, and therefore serve as a better probe for the underlying stellar component. The limiting magnitudes and surface brightness of the u-band image are 26.3 mag and 27.54 mag arcsec⁻², respectively (Table 3). The point source in the upper-left corner of the dry merger (Satsuki and Mei) is a foreground star according to the MaNGA spectrum.

Although visually there is no distinguishable u-band feature (i.e., recent star formation) associated with Totoro, to ensure that the u-band counterpart of Totoro is not embedded within the light of the dry merger (Satsuki and Mei), we subtract the photometric models for the merging system from the u-band image using the three different methods mentioned above. The residual maps and profiles are shown in the top row of Figure 5. The residual maps and profiles are not perfectly smooth, but we find no obvious evidence for a distinctive, extended u-band counterpart at Totoro. The two point sources seen in the g-, r-, and i-band residual maps are also seen in the u-band residual image of the isophotal fitting (and marginally seen in the background method as well). This confirms the results in Paper I that star formation alone cannot explain the excitation state of Totoro.

To gain some insight into the nature of the two unresolved objects in the Totoro area, their photometric redshifts are determined using the EAZY (Brammer et al. 2008) and PÉGASE 2.0 (Fioc & Rocca-Volmerange 1997) template fitting the aperture magnitudes measured from the isophotal fitting residual images using GAIA (Graphical Astronomy and Image Analysis Tool). The default EAZY template is generated from the PÉGASE 2.0 models using the Blanton & Roweis (2007) algorithm and then calibrated using semianalytic models, plus an additional young and dusty template. The PÉGASE 2.0 template is a library including ∼3000 models with a variety of star formation histories and with ages between 1 Myr and 20 Gyr, as described in detail in Grazian et al. (2006). Figure 6 shows the chi-squared of the fit for a given redshift using the EAZY and PÉGASE templates (left column of each panel) and the best-fit SED with the observed fluxes at the i, r, g, and u bands overlaid as red circles (right column of each panel). The results based on the EAZY and PÉGASE templates are presented in the top and bottom rows, respectively. The minimum chi-squared value indicates that the southern source (17^h15^m22111, +57°26'5628) is a background galaxy at z ∼0.40 and 0.37 using the EAZY and PÉGASE templates, respectively (Figure 6(a)). The redshift of our target is marked by a yellow dashed line in the figures. We notice a second minimum at z >0.03, close to the redshift of our target. However, at such low redshifts, the source would be resolved, not point like. The most plausible redshift of the northern source (17^h15^m22573, +57°26'7659) is 0.29 according to both templates (Figure 6(b)), but we cannot rule out other possibilities of z >0.4, in particular z ∼0.15, due to the shallower basin-shaped chi-square distribution. Nonetheless, the redshift of Totoro (yellow dashed line) is not associated with any local minimum of the chi-square values. We find poorer fits when using stellar templates, providing further support that the source is not a nearby object. The best-fitted SEDs from PÉGASE are exported to the SED-fitting code New-Hyperz³⁷ in order to derive the stellar mass (M_*) and star formation rate (SFR) of these two objects. The results suggest that the southern and northern sources are ∼3 × 10⁸ and ∼6 × 10⁸ ${M}_{\odot }$, and their specific SFR (sSFR=SFR/M_*) are $\leqslant$ 10⁻¹¹ yr⁻¹. We also estimate their M_* assuming they are at the same redshift as Totoro (z = 0.03) and yield ∼10⁶ and ∼10⁷ ${M}_{\odot }$ for the southern and northern sources, respectively. Given their point-like morphologies (< 1 kpc²), we expect to see distinctive stellar mass surface density (${{\rm{\Sigma }}}_{* }$) distributions in the MaNGA data if they are associated with Totoro. However, we find no such features. It should be noted that the above stellar masses may be subjected to non-negligible uncertainties due to the lack of (near-) infrared measurements.

Zoom In Zoom Out Reset image size

We use the web interface Marvin,³⁸ a tool to visualize and analyze MaNGA data (Cherinka et al. 2019), to search for any signatures in the spectra of these two sources by redshifting strong optical emission lines (e.g., Hα) based on the derived photometric redshifts. However, the emission lines are too faint to be seen by MaNGA. Given their likely high redshift and their positions being offset from the centroid of the Hα blob, these two point sources are unlikely to be associated with Totoro.

If Totoro is indeed a separate galaxy, it may possess another tidal tail on the other side of the blob (i.e., as opposed to the ones that connect Totoro and Satsuki), but such tail(s) would not be seen by MaNGA due to the limited FoV. Figure 7 shows the new wide-field Hα image taken from the SAO RAS 6 m telescope. The new Hα map has a sensitivity comparable to that of MaNGA, but the FoV is ∼10 times larger. For comparison, Figure 7(a) zooms in to the region of the MaNGA hexagonal FoV. The new Hα image globally resembles that of the MaNGA Hα in Figure 1(a), but shows finer structures thanks to higher spatial resolution.

Zoom In Zoom Out Reset image size

A zoom-out view of the region of interest is displayed in Figure 7(b). There are several Hα knots beyond the MaNGA FoV toward the east. These are background galaxies or foreground stars. It is clear that there is no hint of Hα emission extending beyond the MaNGA FoV at the opposite side of Totoro. Quantitatively, we can estimate the possible missing flux of MaNGA by comparing the Hα flux related to Totoro from the two observations. The total Hα luminosity of Totoro and the surrounding ∼15'' (∼9 kpc) region outside of the MaNGA FoV (blue circle in Figure 7(b)) is 6.5 × 10³⁹ erg s⁻¹. The total Hα luminosity of Totoro measured by MaNGA is 5.9 × 10³⁹ erg s⁻¹. Therefore, the possible missing flux of Totoro due to a limited MaNGA FoV is no larger than 10%. Accordingly, the scenario of a separate galaxy appears less likely unless the tidal tail(s) develop at only one side of a galaxy. Such cases are relatively rare, though not impossible (e.g., Arp 173, Arp188, and Arp 273), depending on the stage of the interaction and the projected orientation on the sky (e.g., Mihos 2004; Struck & Smith 2012). However, the low gas velocity provides additional support against a merger scenario (Figures 3(c) and 3(d)).

On the other hand, it is possible that Totoro is a completely disrupted dwarf galaxy. The averaged surface brightness of Totoro has an upper limit of ∼27 mag arcsec⁻² (Table 3). It would be classified as a dwarf low-surface-brightness galaxy (LSB; >23 mag arcsec⁻²) if it is a galaxy. In the next section, we compare the star formation and cold gas properties of Totoro with other galaxy populations, including LSBs. As a side note, the nearby galaxy NGC 6338 is encompassed by the wide-field Hα observation. Figure 7(c) shows the Hα image of NGC 6338. There are no strong filaments or bridge linking NGC 6338 and VII Zw 700, presumably because the two are very separate entities (with a projected separation of ∼42 kpc and ∼1400 km s⁻¹ difference in line-of-sight velocity), but there is a bit more Hα emission between galaxies than the rest of the regions in the plotted area. The Hα emission of NGC 6338 is characteristic of three previously reported filaments in the southeast and northwest quadrants. The Hα intensity contours of 0.035 and 0.09 × 10⁻¹⁶ erg s⁻¹ cm⁻² are overplotted to highlight the asymmetric filaments. We also refer the reader to Martel et al. (2004) and O'Sullivan et al. (2019) for higher-resolution Hα images of NGC 6338 and Gomes et al. (2016) for optical IFS data analysis of NGC 6338.

Another way to constrain the origin of Totoro using data in hand is to look into the question of whether Totoro shares similar gas and star formation properties with nearby galaxies. This could not be concluded in Paper I due to the lack of cold gas data. Figure 8 compares the star formation rate (SFR) and molecular gas mass (${M}_{{{\rm{H}}}_{2}}$) of Totoro with other galaxy populations. The plot resembles the Kennicutt–Schmidt relation (Kennicutt 1989) assuming that the molecular gas and star-forming regions coexist. The galaxy data include nearby LSBs (O'Neil et al. 2003; Matthews et al. 2005; Cao et al. 2017) and nearby star-forming (sSFR >10⁻¹¹ yr⁻¹) and quiescent (sSFR <10⁻¹¹ yr⁻¹) galaxies (Saintonge et al. 2017).

Zoom In Zoom Out Reset image size

The total Hα luminosity of Totoro (including the connecting arms; ∼48.1 kpc² area in total) is converted to SFR using the calibration of

$$\begin{eqnarray}&&\displaystyle \frac{\mathrm{SFR}}{[{M}_{\odot }\,{{\rm{yr}}}^{-1}]}=7.9\,\times {10}^{42}\,\displaystyle \frac{{L}_{{\rm{H}}\alpha }}{[\mathrm{erg}\,\ {{\rm{s}}}^{-1}]}\end{eqnarray} \tag{ 1 }$$

(Kennicutt 1998), where ${L}_{H\alpha }$ is the Hα luminosity. This yields an SFR of 0.047 ${M}_{\odot }$ yr⁻¹. Here we assume all of the Hα results from star formation, but this is unlikely to be the case as BPT diagnostics indicate a composite nature, LI(N)ER–H ii mix excitation, for Totoro (Paper I). Therefore, the derived SFR is an upper limit.

The total molecular gas mass of Totoro traced by CO is computed using

$$\begin{eqnarray}\begin{array}{rcl}\displaystyle \frac{{M}_{{{\rm{H}}}_{2}}}{[{M}_{\odot }]} & = & 1.05\,\times {10}^{4}\,\displaystyle \frac{{X}_{\mathrm{CO}}}{[2\,\times {10}^{20}\,{\mathrm{cm}}^{-2}{({\rm{K}}{\rm{km}}{{\rm{s}}}^{-1})}^{-1}]}\\ & & \times \displaystyle \frac{{S}_{\mathrm{CO}}{\rm{\Delta }}v}{[\mathrm{Jy}\,\ \mathrm{km}\,\ {{\rm{s}}}^{-1}]}\displaystyle \frac{{D}_{{\rm{L}}}^{2}}{[\mathrm{Mpc}]}\,{(1+z)}^{-1},\end{array}\end{eqnarray} \tag{ 2 }$$

where X_CO, ${S}_{{\rm{CO}}}{\rm{\Delta }}\nu$, ${D}_{{\rm{L}}}$ are the CO-to-H₂ conversion factor, integrated line flux density, and luminosity distance, respectively (Bolatto et al. 2013). We adopt a Galactic X_CO of 2 × 10²⁰ cm⁻² (K km s⁻¹)⁻¹ (Bolatto et al. 2013). The derived ${M}_{{{\rm{H}}}_{2}}$ ³⁹ of Totoro is (2.1 ± 0.1)×10⁸ ${M}_{\odot }$. Note that the uncertainty due to different methodologies to derive he Galactic X_CO is ∼30% (Bolatto et al. 2013).

In addition to CO, the gaseous column density ${N}_{{{\rm{H}}}_{2}}$ (and ${M}_{{{\rm{H}}}_{2}}$) also follows from the amount of extinction A_V, which can be derived from the MaNGA Hα and Hβ data. For reference, the mean A_V across Totoro is ∼0.5 mag. The conversion from extinction to H₂ column density depends on the medium (Bohlin et al. 1978; Evans et al. 2009):

$$\begin{eqnarray}&&\displaystyle \frac{N({{\rm{H}}}_{2})}{[{\mathrm{cm}}^{-2}]}=6.9\,\times {10}^{20}\,\displaystyle \frac{{A}_{{\rm{v}}}}{[\mathrm{mag}]}\,(\mathrm{molecular}\,\mathrm{cloud})\end{eqnarray} \tag{ 3 }$$

$$\begin{eqnarray}&&=9.4\,\times {10}^{20}\,\displaystyle \frac{{A}_{{\rm{v}}}}{[\mathrm{mag}]}\,(\mathrm{diffuse}\,\mathrm{ISM}).\end{eqnarray} \tag{ 4 }$$

These yield an ${M}_{{{\rm{H}}}_{2}}$ of ∼1.9 × 10⁸ and ∼2.5 × 10⁸ ${M}_{\odot }$ assuming a molecular-cloud- (Equation (3)) and diffuse-ISM-type (Equation (4)) medium, respectively. The three ${M}_{{{\rm{H}}}_{2}}$ derived using radio and optical measurements agree well with each other. The H₂ mass derived from CO is used for the discussion in the rest of the paper.

Although only upper limits could be achieved for many LSB objects, they largely follow the trend established by star-forming galaxies toward the lower end in both ${M}_{{{\rm{H}}}_{2}}$ and SFR axes, consistent with the finding of McGaugh et al. 2017. On the other hand, quiescent galaxies, mostly early types, have lower SFR for a given ${M}_{{{\rm{H}}}_{2}}$ and a lower CO detection rate than that of star-forming galaxies (see also Calette et al. 2018). Due to the low SFR, Totoro deviates from the SFR–${M}_{{{\rm{H}}}_{2}}$ relation formed by LSBs and star-forming galaxies and appears to overlap with quiescent galaxies.

The Hα emission in quiescent galaxies is dominated by LI(N)ER excitation (Hsieh et al. 2017; Pan et al. 2018). In the LI(N)ER regions of quiescent galaxies, the surface density of Hα luminosity (${{\rm{\Sigma }}}_{H\alpha }$) is found to be tightly correlated with the underlying ${{\rm{\Sigma }}}_{* }$ (Hsieh et al. 2017), in which the Hα are primarily powered by the hot, evolved stars rather than recent star formation. BPT diagnostics suggest that Totoro is powered by a composite (LI(N)ER–H ii mix) mechanism (Paper I). If Totoro is analogous to a LI(N)ER region in quiescent (early-type) galaxies, the average ${{\rm{\Sigma }}}_{{\rm{H}}\alpha }$ of Totoro corresponds to a ${{\rm{\Sigma }}}_{* }$ of ∼7 × 10⁸ ${M}_{\odot }$ kpc⁻² according to the "resolved LI(N)ER sequence" of quiescent galaxies reported by Hsieh et al. (2017). The predicted ${{\rm{\Sigma }}}_{* }$ is higher than the mass of Satsuki's stellar halo by a factor of 3–5; however, there is no distinct stellar counterpart at the location of Totoro. For this reason, in spite of the overlap in Figure 8, Totoro is not analogous to the LI(N)ER region in quiescent galaxies. Lastly, we should note that if the true global SFR of Totoro is considerably lower than the upper limit, Totoro would fall below the main cloud of data points of quiescent galaxies.

On the whole, Totoro is unlikely to be consistent with nearby normal star-forming, early-type, and low-surface-brightness galaxies in terms of the Kennicutt–Schmidt relation and the resolved LI(N)ER sequence. In addition, it is worth noting that the average A_V of Totoro corresponds to an SFR surface density (${{\rm{\Sigma }}}_{\mathrm{SFR}}$) at least six times higher than the observed upper limit based on the local A_V–${{\rm{\Sigma }}}_{\mathrm{SFR}}$ relation derived from MaNGA galaxies (Li et al. 2019). Generally speaking, the dust and star formation relation (if any) of Totoro is also dissimilar to that of star-forming regions in nearby galaxies even when the systematic dependencies on other physical properties (e.g., metallicity) are considered.

While LSBs have been studied for decades, recently, van Dokkum et al. (2015) have identified a new class of LSBs in the Coma cluster, UDGs. These UDGs have a surface brightness as low as >24.5 mag arcsec⁻², but their sizes are similar to those of L^* galaxies. It is not clear how UDGs were formed. One possible scenario is that UDGs are failed massive galaxies, which lost their gas at high redshift by ram pressure stripping or other effects after forming their first generation of stars (van Dokkum et al. 2015; Yozin & Bekki 2015). Our finding of a large molecular gas reservoir appears in direct conflict with this scenario.

On the other hand, several studies suggest that UDGs are extended dwarf galaxies. Some simulations predict them to be rapidly rotating (Amorisco & Loeb 2016). Di Cintio et al. (2017) suggest that the extended sizes of UDGs are the consequence of strong gas outflows driven by starbursts. However, the SFR of Totoro is extremely low and the system is not rotating. Altogether, there is no evidence in our data to support Totoro as a UDG.

In summary, the newly obtained CO, Hα, and u-band data allow us to better constrain the nature of Totoro; however, combining the results in Section 3.2, we argue that Totoro is unlikely to be a separate galaxy interacting with the dry merger (Satsuki and Mei). The reasons include the lack of stellar counterpart and tidal features and the different star formation, ionized and cold gas, and dust properties (i.e., ${M}_{{{\rm{H}}}_{2}}$–SFR, ${{\rm{\Sigma }}}_{* }$–${{\rm{\Sigma }}}_{H\alpha }$, and A_V–${{\rm{\Sigma }}}_{\mathrm{SFR}}$ relations and gas kinematics) from those of a variety of nearby galaxy populations (i.e., star-forming and quiescent galaxies, LSBs, and UDGs).

Radio. A possible origin of Totoro is gas that is photoionized by X-ray emission from a misaligned blazar at the core of Satsuki. A blazar is a subclass of radio-loud AGNs which is characterized by a one-sided jet structure (Urry & Padovani 1995). However, blazars invariably have bright compact radio cores, whereas no detection at 1.4 GHz is found for Satsuki (O'Sullivan et al. 2019; Wang et al. 2019), and there is only marginal detection at 5 GHz (Paper I). For these reasons, a blazar is a very unlikely scenario for Totoro.

Optical. The Hα equivalent width (EW) is <3 Å across the entire Hα-emitting region and the EW does not present a rise toward the center of Satsuki. The Hα velocity dispersion does not present a central peak either (Paper I). Moreover, the Hα luminosity of Satsuki follows the continuum emission, i.e., it does not present a central peak emission with stronger intensities that follows an ${r}^{-2}$ or lower decline from the center (Singh et al. 2013). Therefore, Satsuki presents a lack of optical characteristics of a strong AGN.

X-ray. A comprehensive study of the X-ray emission of the region based on Chandra and XMM-Newton data has been reported recently by O'Sullivan et al. (2019). Figure 9(a) shows the Chandra X-ray image of the NGC 6338 group taken from O'Sullivan et al. (2019). X-ray and galaxy velocity studies have shown the group to be a high-velocity near head-on merger (Wang et al. 2019), with the two group cores visible as bright clumps of X-ray emission with trailing X-ray tails. The northern and southern clumps are respectively associated with VII Zw 700 and NGC 6338. Figure 9(b) zooms in to the MaNGA observed region. The northern X-ray clump, around Satsuki, is dominated by a bar of X-ray emission extending roughly southeast–northwest. The clump and bar are centered somewhat to the north of the optical centroid of Satsuki, and the bar is made up of three X-ray knots whose intensities progressively decrease from the western end. The nuclei of Satsuki and Mei are not correlated with the brightest X-ray emission (see also Wang et al. 2019). Nonetheless, after subtracting the overall surface brightness distribution, there is a hint of excess X-ray emission at the position of the nuclei of Satsuki and Mei (see O'Sullivan et al. 2019 for the construction of the overall surface brightness distribution). X-ray luminosities of 4.17 × 10³⁹ erg s⁻¹ for the nucleus of Satsuki and $\leqslant$ 2.5 × 10³⁸ erg s⁻¹ for Mei are reported (O'Sullivan et al. 2019). The values suggest that there are no strong X-ray AGNs in Satsuki or Mei, or the AGNs are fading. For reference, the X-ray luminosity of Totoro is (4.4 ± 0.2) × 10⁴⁰ erg s⁻¹.

Zoom In Zoom Out Reset image size

All in all, the multiwavelength data show no direct evidence for an active ongoing AGN in Satsuki. However, it is known that the AGN luminosity can vary over timescales as short as 10⁵ yr. Thus, we still cannot rule out the possibility of a recent AGN outflow. AGN-driven extended outflows are detected in multiple phases, including ionized gas and many molecular lines (Roberts-Borsani 2020 and references therein). Even though the observed radial extent of AGN outflows is $\leqslant$ 1 kpc in most cases, galactic-scale outflows, as Totoro would be, given its extreme distance from Satsuki, are also reported (e.g., Leung et al. 2019; López-Cobá et al. 2019, 2020). However, the low gas velocities (Figures 3(c) and 3(d)) do not support the scenario of an energetic outflow. Nonetheless, we should note that although an AGN does not seem like a plausible mechanism for moving Hα or CO out of the galaxy core to the current position of Totoro, the potential cavities identified in Satsuki suggest that in the past (∼40 Myr ago), the AGN was indeed active and had non-negligible impact on its surroundings (O'Sullivan et al. 2019).

There have been studies showing that AGNs are able to ionize gas extending to large distance, such as the well-known ionized cloud Hanny's Voorwerp ∼20 kpc from its host galaxy IC 2497 (e.g., Husemann et al. 2008; Lintott et al. 2009; Husemann et al. 2010). It is believed that an interaction-triggered, currently fading AGN illuminated and ionized Hanny's Voorewerp (Józsa et al. 2009; Lintott et al. 2009; Keel et al. 2012). Diffuse ionized gas similar to Hanny's Voorewerp was also found ∼32 kpc north of the iconic interacting galaxies NGC 5194/5195 or M51 (Watkins et al. 2018). The low AGN luminosity and activity and the tidal history provide similarities between Satsuki, IC 2497 and M51.

However, there are also several differences between Hanny's Voorwerp and the M51 cloud and Totoro in terms of excitation state (Seyfert versus LINER–H ii, morphology (diffuse for Hanny's Voorwerp and M51's cloud versus centrally concentrated for Totoro), and the properties of the host galaxy (late type versus early type; see Paper I for the details). For these reasons, we argue that this scenario is unlikely.

A scenario we did not consider in Paper I is the cooling of the hot IGM or ICM. Observations of the central regions of some galaxy groups and clusters show strong X-ray emission, suggesting that the IGM and ICM are undergoing rapid cooling (e.g., Fabian 1994), and in some cases, ionized and molecular gas which are thought to be the product of that cooling (Babyk et al. 2018; Lakhchaura et al. 2018; Olivares et al. 2019; Russell et al. 2019).

O'Sullivan et al. (2019) show that the X-ray peak of the southern clump in Figure 9(a) is consistent with the optical centroid of NGC 6338. Three X-ray filaments are observed to extend from the galaxy center, following the same branching filamentary structure as the Hα gas (Figure 7(c)). These X-ray filaments are cooler than their surroundings and have very low gas entropies and short cooling times (O'Sullivan et al. 2019), strongly indicating that they are a locus of cooling from the IGM. Young X-ray cavities are also found in NGC 6338, suggesting recent AGN outbursts in this galaxy.

As shown in Figure 9(b), the peak of the Hα emission (contours) at the nucleus of Satsuki is not associated with the bar, but the position of Totoro corresponds to the brightest X-ray peak. This was first noted by O'Sullivan et al. (2019) who showed that this was also the coolest part of the X-ray bar, and concluded that the Hα was likely material cooled from the IGM, as in the filaments of NGC 6338. They also suggested that the offset between Satsuki and the center of the X-ray clump is evidence that ram-pressure forces caused by the (supersonic) motion of the galaxy are detaching the gas from the galaxy. In this scenario, the X-ray bar would once have been centered on Satsuki but has been pushed back to the north and perhaps along the line of sight, away from the galaxy core.

Figure 9(c) shows our CO data overlaid on the Chandra image. The molecular gas is well correlated with the western X-ray knot, and (as previously discussed) with the Hα. Figures 9(d)–(e) show the Chandra temperature maps from O'Sullivan et al. (2019) showing the cool gas around NGC 6338 and Satsuki, and the high-temperature gas between the two, which has been shock-heated by the group–group merger. Overlaid Hα and CO contours show that the ionized and molecular gas are located in the coolest part of the IGM, as in NGC 6338 and the centers of other groups and clusters (e.g., Salomé et al. 2006; Hamer et al. 2012).

Studies have shown that warm and cold gas in the brightest group and cluster galaxies (BCGs) are preferentially observed in systems where the cooling times lie below ∼1 Gyr (Edge 2001; Salomé & Combes 2003; Cavagnolo et al. 2008; Pulido et al. 2018; Rafferty et al. 2008). O'Sullivan et al. (2019) found that the cooling times are as short as <1 Gyr in both cores of Satsuki and NGC 6338. We can also estimate the cooling time around Totoro, i.e., the west end of the X-ray bar as a cylinder of radius 28 (1.5 kpc) and length 112 (5.9 kpc, corresponding to Region 2 in Figure 10(b) of O'Sullivan et al. 2019). The temperature, density, and luminosity of the X-ray gas are estimated via spectral fitting of the Chandra data. Using Equation (1) in O'Sullivan et al. (2019), we estimate the cooling time (t_cool) in the Totoro region to be ${2.2}_{-0.1}^{+0.2}$×10⁸ yr, well within the regime where warm and cold gas are expected.

If the warm and cold gas have indeed cooled from the hot gas, one would expect the hot gas to be significantly more massive than the cool/warm gas. The hot gas mass is derived from a radial deprojected profile (Figure 9 in O'Sullivan et al. 2019), centered on the middle of the X-ray bar. Within a radius of 6.5 kpc, the hot gas mass (M_X) is ${1.2}_{-0.24}^{+1.07}$×10⁹ ${M}_{\odot }$. The derived hot gas mass is indeed significantly higher than the total mass of warm (∼10⁵ ${M}_{\odot }$ from Hα) and cold gas (∼10⁸ ${M}_{\odot }$ from CO).

The ratio of the cold to hot gas of Totoro is ∼17%. Pulido et al. (2018) investigate the molecular gas properties of 55 central cluster galaxies. They found a strong correlation of hot and cold gas mass traced by X-ray and CO, suggesting that the hot and cold gas arise from the same ensemble of clouds. In other words, the cold gas is unlikely a result of external effects, such as merger or stripping from a plunging galaxy. The average fraction of cold to hot gas in their sample is ∼18%. The cold to hot gas mass ratio of Totoro is in good agreement with that of these central cluster galaxies (see also Olivares et al. 2019). The consistency provides support that a similar process to that in the central cluster galaxies has occurred in Satsuki and Totoro.

In addition, molecular mass has been found to be correlated with the amount of Hα gas expressed by ${L}_{{\rm{H}}\alpha }$ in the cooling-core galaxies of clusters (Edge 2001; Salomé & Combes 2003). In Figure 10, we show the H₂ gas mass versus ${L}_{{\rm{H}}\alpha }$ for data taken from the literatures. Totoro is overlaid with an orange square and falls at the low end of the correlation. The consistency of Totoro with the ${M}_{{{\rm{H}}}_{2}}$–${L}_{H\alpha }$ relationship of cooling systems again supports a similar process of cooling in the system. To put it another way, our multiwavelength data of Totoro show that its position on the mass (or luminosity) relations between cold (≲100 K; CO), warm (∼10⁴ K, Hα), and hot (>10⁷ K, X-ray) gas is in line with the gas content of systems with short cooling times. We summarize the physical properties of Totoro in Table 2.

Zoom In Zoom Out Reset image size

As a side note, the cold and warm gas in cooling systems often appear filamentary, but it is worth noting that the multiphase gas morphologies of our target are strikingly similar to the BCG of cluster A1991 reported by Hamer et al. 2012 (see their Figure 2). The Hα-emitting gas is relatively circular (blob like) and is spatially coincident with the most rapidly cooling region (X-ray peak) of the ICM. The peak in the Hα and X-ray gas lies roughly 11 kpc to the north of the BCG, and there are connecting arm structures between the peak and the secondary peak at the galactic nucleus. Moreover, the bulk of the molecular gas with a mass of ∼8 × 10⁸ ${M}_{\odot }$ is also found at the location of the cooling region. However, note that spatial resolution must play an important role in detecting the morphology of cooling gas. We cannot rule out that the morphologies of Totoro and the cooling gas in A1991 are more filamentary than a single peak.

Finally, the velocity fields of cooling gas in galaxy clusters, traced by CO and Hα, are characteristic of slow motions (projected velocity <400 km s⁻¹), narrow line widths (>250 km s⁻¹), and a lack of relaxed (e.g., rotating) structures (e.g., Salomé et al. 2008; Hamer et al. 2012; Olivares et al. 2019). The observed velocity structures, despite low velocity resolutions, of Totoro traced by CO and Hα (Figures 3(c) and 3(d)) agree with those of other cooling systems. In addition, the gas velocities of Totoro traced by CO and Hα are not exactly identical, but the differences are small, ∼35 km s⁻¹ at the main blob region and 60—90 km s⁻¹ at the connecting arms. This is also consistent with the finding by Olivares et al. (2019) that the velocity difference between CO and Hα gas is well below 100 km s⁻¹, providing an additional support for CO and Hα gas arising from the same bulk of clouds. The velocity difference may be related to the different velocity resolutions of CO and Hα observations and line-of-sight projection effect. We should note again that our CO and Hα observations suffer from low velocity resolutions; therefore, the velocity fields must be interpreted with caution. Nonetheless, the maps still provide a guide of the velocity resolution needed for future observation of this system. Future high velocity resolution CO and Hα observations are required to reveal the detailed gas kinematics of Totoro.

While most of the studies of cooling gas focus on BCGs in rich clusters, cold gas cooling from the hot X-ray medium is also observed at smaller scales in galaxy groups, such as NGC 5044, NGC 4638, and NGC5846 (David et al. 2014; Temi et al. 2018). In fact, the observed cool-core fractions for galaxy groups are slightly higher than those of galaxy clusters (O'Sullivan et al. 2017). Therefore, gas cooled from the IGM is not unique to Totoro.

In Section 3.1, we argue that the cold gas in Totoro is unlikely to be the primitive gas of Satsuki being stripped by ram pressure. Nonetheless, the X-ray tail (Figure 9(a)) and the offset between the center of the X-ray bar and the optical centroid of Satsuki are both evidence that the motion of the dry merger (Satsuki and Mei) is rapid enough to lead to stripping of the hot gas halo (O'Sullivan et al. 2019). The question then arises of where and when the ionized and molecular gas we observe was formed; has ram pressure or other effects changed its location? We might normally expect to see the most rapid cooling in or near the galaxy center. However, the offset of the X-ray bar means that the densest, coolest IGM gas is no longer located at the center of Satsuki. Ram pressure, by pushing the X-ray halo and bar away from the core of Satsuki, may therefore have caused a reduction in cooling in the galaxy center (O'Sullivan et al. 2019).

CO emission occurs in dense molecular clouds that are "self-shielding" from the ionizing effects of the surrounding environment. While some of the Hα emission likely comes from the outer layers of such molecular cloud complexes, observations show that in some cooling systems, the Hα emission is considerably more extended than the molecular gas (e.g., in NGC 5044; Schellenberger et al. 2020), which may suggest that it is associated with a less dense cooled gas component. Such low-density material would likely move with the surrounding IGM if ram pressure pushed it back from the galaxy. Dense molecular clouds would not be affected by ram pressure and might be expected to fall under gravity toward the center of Satsuki, unless they are connected to the IGM via magnetic fields (McCourt et al. 2015) or surrounding layers of neutral and ionized gas (Li et al. 2018). Even with such connections to the surrounding environment, it seems implausible that the molecular gas could have condensed out of the IGM into the core of Satsuki and then been uplifted. The correlation between the CO, Hα, and the coolest X-ray gas strongly suggests that the molecular and ionized gas is the product of cooling from the IGM at its current location, i.e., that cooling has occurred (and may be ongoing) in Totoro, well outside the center of Satsuki.

Last but not least, as suggested by O'Sullivan et al. (2019) and this work, we are witnessing a merger between two groups undergoing rapid radiative cooling. Further analysis on the multiwavelength phase of cooling gas in NGC 6338, from cold to hot gas as in this study, will be carried in a separate paper (E. O'Sullivan et al. 2020, in preparation).

Cooling gas can potentially serve as the fuel for an AGN and/or central star formation (O'Dea et al. 2008; Rafferty et al. 2008; Mittal et al. 2009; Hicks et al. 2010; Fogarty et al. 2017). In X-ray-bright groups, like our target, star formation in the central galaxy is generally weak even in systems known to be cooling. By contrast, as many as 85%–90% of group-dominant galaxies have radio AGNs; moreover, dominant galaxies with active or recently active radio jets are relatively common in X-ray-bright groups (Kolokythas et al. 2018, 2019). Galaxy-cluster-dominant galaxies, however, seem much more likely to have significant star formation.

Here, we consider whether the cooling gas would fuel star formation assuming the gas will fall back into Satsuki. McDonald et al. (2018) compare the cooling rate of the ICM/IGM to the observed SFR in the central galaxy for a sample of isolated ellipticals, groups, and clusters. They found that the cooling ICM/IGM is not providing the fuel for star formation in systems with cooling rate < 30 ${M}_{\odot }$ yr⁻¹, which are dominated by groups and isolated ellipticals. On the other hand, SFR increases with increasing cooling rate for rapidly cooling systems (>30 ${M}_{\odot }$ yr⁻¹), presumably due to an increase in either the cooling efficiency of the hot gas or the star formation efficiency of the cooled gas (see also Edge 2001; Salomé and Combes 2003; O'Dea et al. 2008). The cooling rate (M_X/t_cool) of Totoro is ∼1.2 × 10⁹ ${M}_{\odot }$/2.2 × 10⁸ yr ≈5 ${M}_{\odot }$ yr⁻¹. In this aspect, the cooling gas in our group-dominant, relatively low cooling rate system is less likely to significantly contribute to star formation.

Moreover, star formation can be suppressed by AGN feedback even in systems with short cooling times. The summed AGN jet power (P_cav) for both cavities associated with Satsuki from O'Sullivan et al. (2019) is (0.97—2.67)×10⁴¹ erg s⁻¹ (an estimate of heating; the actual value depends on the cavity age used: buoyant rise time, sonic expansion timescale, or refill time); the energy is comparable to the X-ray luminosity for the whole X-ray-emitting gas (∼3 × 10⁴¹ erg s⁻¹; an estimate of cooling). Rafferty et al. (2008) find a tendency for star-forming systems to have low P_cav/L_X ratios (< 1) and quiescent systems to have high P_cav/L_X ratios (>1), supporting the suppression of star formation by AGN feedback. However, this is not exclusively the case; star-forming cooling systems can have high P_cav/L_X ratios and vice versa. The estimated heating available from the cavities around Satsuki make it a borderline case, and the ongoing merger and stripping add to the complexities. The fate of Totoro may depend on how effectively energy from the cavities can heat their surroundings.

Finally, it is worth mentioning that Satsuki and Totoro may host a little star formation activity with an upper limit of < 0.059 and < 0.047 ${M}_{\odot }$ yr⁻¹, respectively, assuming all of the Hα fluxes come from star formation. It is unclear if the current star formation (if any) is related to the cooling and gas fueling processes. McDonald et al. (2018) attribute the low-level star formation in low cooling rate systems to recycling of gas lost by evolved stars, namely, the star formation is not related to cooling gas.