TRACING RAM-PRESSURE STRIPPING WITH WARM MOLECULAR HYDROGEN EMISSION

We obtained data for each galaxy in all four (3.6, 4.5, 5.7, and 8 μm) IRAC channels (Fazio et al. 2004). IRAC imaging of 97073 and NGC 1427A was carried out as part of our Spitzer GTO program. IRAC data for NGC 4522 were obtained from the Spitzer Science Archive, observed as part of program 30945 (PI: J. Kenney). We constructed IRAC mosaics of the galaxy with MOPEX (version 18.3.1) using the Basic Calibrated Data (BCD) from the Spitzer pipeline (version S18.7) as input and following the Spitzer Science Center (SSC) IRAC reduction cookbook. To reduce banding artifacts, corrected BCDs were used as input for the construction of 8 μm mosaics. Bad pixel rejection was achieved through dithering, median filtering, and masking of known bad pixels using pixel masks provided by the SSC.

We also constructed 8 μm excess images for each galaxy to study its star-forming properties using the same method outlined in Paper 1. Excess 8 μm flux in galaxies is often produced by warm dust rich in aromatic hydrocarbons, which is associated with star-formation regions (Calzetti et al. 2007). Our technique subtracts the stellar continuum from the 8 μm image by subtracting a resolution-matched, flux-scaled, and well-registered 3.6 μm image. We verify the effectiveness of our method by ensuring that the foreground stars in our field are adequately subtracted in the final 8 μm excess image.

We obtained HST Advanced Camera for Surveys (ACS) archival data for NGC 1427A (Proposal ID: 9689; PI: M. Gregg) and NGC 4522 (Proposal ID: 9773; PI: J. Kenney). We used the F660N (narrowband Hα) and F775W (SDSS I-band) data for NGC 1427A, and F435W (B-band) and F814 (I-band) data for NGC 4522. We obtained the pipeline-reduced multidrizzled data for our analysis. To construct the Hα image for NGC 1427A, we used the F660N and F775W filter data. The pipeline-reduced data for each filter were already properly registered as part of the pipeline processing. We simply scaled the F775W data and subtracted it from the F660N data to remove the stellar contribution from the narrowband image. We verified the success of our subtraction by ensuring that the foreground stars in the field were adequately subtracted. Due to the poor signal to noise of the subtracted data, we binned and smoothed the data to improve the visibility of Hα emitting regions. Because of the finite width of the narrowband filter, there will be some contamination from [N ii] lines that are close to the Hα wavelength.

Our IRS observations used both the short–low (SL) and long–low (LL) IRS modules that together span 5.3–38.0 μm (Houck et al. 2004). The complete wavelength range allows us to observe several ground vibrational state H₂ rotational lines, specifically the ν = 0–0 S(0) through S(7) transitions, that can be used to characterize the thermodynamic properties of warm H₂. The spectral mapping techniques and data analysis methods are discussed in Paper 1. We use the most recent IRS pipeline-reduced BCD files (version S18.7) and CUBISM (version 1.7; Smith et al. 2007a) to construct our spectral cubes. Bad pixels are eliminated visually using the same method described in Paper 1, and the initial background subtraction is carried out by subtracting the background measured with the outrigger pointings. However, additional background subtraction was required for some observations because the outrigger pointing background subtraction did not remove all of the flux due to background gradients. In these cases, we generated a background spectrum for each spectral map by averaging the spectra in source-free regions. This background was then subtracted, and the error of the background spectrum was included in the final spectral map flux uncertainties. For the case of NGC 4522, we supplemented our analyses with archival LL spectrograph data from another Spitzer program (Proposal ID: 50819; PI: J. Kenney). We use the same methods of background subtraction and bad pixel rejection for analyzing the archival data set. The coverage maps for the two different sets of observations for NGC 4522 are quite different. Our IRS observations cover the entire galaxy, but the archival observations are deeper and focus on two specific locations where ram-pressure is thought to be stripping significant amounts of gas. The archival observations have approximately two to three times more integration time per sky pixel than our observations (∼90 s versus 250 s for LL2, and 150 s versus 250 s for LL1) and use a similar mapping strategy. The IRS coverage maps for 97073, NGC 4522, and NGC 1427A are shown in Figures 1–3, respectively. The coverage map for ESO 137-001 has already been presented in Paper 1.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The spectra presented in this paper are first extracted using a custom program that sums the spectra and root-sum-squares the errors within a specified region for each order from the SL and LL spectrographs. The code then stitches the spectra from all orders together without rescaling. The flux values within overlapping wavelength regions are interleaved. The stitched spectrum is then shifted to rest wavelengths using the optical redshift for the galaxy found in the NASA Extragalactic Database (NED). The line fluxes are derived by fitting the extracted spectrum using PAHFIT (Smith et al. 2007b). This routine fits for all H₂ and fine structure lines, aromatic features, and dust continuum within the wavelengths probed by IRS. We typically do not include fitting for extinction except for the case of NGC 4522, where there is a significant amount of dust, as evidenced by the 10 μm silicate absorption feature in the spectra. Strong extinction can significantly change the fluxes of S(1) through S(7) lines. The dust extinction fitting is discussed in further detail below. The H₂ line fluxes and 1σ errors used in our analyses of the H₂ population are the values reported by the PAHFIT routine. To determine the contribution of PAH complexes to the spectrum, we used PAHFIT's pahfit_main_feature_power routine to accurately determine the errors of correlated features. The H₂, PAH feature, and fine structure line fluxes are presented in Tables 1–3, respectively.

Two different types of H₂ images are generated for our analyses. The first does not employ any continuum subtraction. It is generated by summing the rest wavelength spectra within three wavelength bins centered on the 17.035 μm H₂ (0–0 J = 3–1 S(1)) transition. The second type of H₂ image is a continuum-subtracted version of the first. The continuum is estimated by generating two adjacent wavelength bins of the same size as the one already generated for the H₂ image, and then averaging the flux between the two background bins. This method works well for fairly strong H₂ features without significant 17 μm aromatic features, but it is more difficult to detect faint H₂ emission in the continuum-subtracted images due to the additional noise introduced by the subtraction.

We carry out an additional analysis step for NGC 4522 to combine our LL and archival data sets. For generating the 17 μm images, we co-added the two images generated for each set of observations using the following method. First, we registered and interpolated the image generated from the deeper observations to match the pixel grid of the shallower observations. Second, in regions of coverage overlap, we calculated a weighted average of the two different data sets using the inverse variance of each data set as weights. Third, in areas without any overlap, we kept the original value associated with the data set that covered the area. For the generation of spectra, if the chosen extraction region fell entirely within the region where the deeper observations were carried out, we extracted the spectra from just the deeper observations. Otherwise, we use the data from our shallower map.

We also construct 24 μm images that mimic what would be observed by the Spitzer MIPS imager. We compute a 24 μm flux with the IRS data for each sky pixel by using the MIPS 24 μm transmission curve as weights.

3.1.1. CGCG 97-073

97073 is a galaxy within A1367 that has a measured optical redshift of 0.024267, as reported by the NASA Extragalactic Database (NED). A1367, itself, has a redshift of 0.022, also obtained from NED, which we use to calculate the approximate luminosity distance to the galaxy of 96 Mpc. The galaxy is located at a projected distance of 640 kpc from the cluster center and has a 50 kpc long collimated Hα tail that stretches directly north of the galaxy (Gavazzi et al. 2001). In Figure 4, we show images at IRAC 3.6 and 8 μm, Hα (obtained from the GOLDMine database), and rest 17.035 μm (H₂ 0–0 S(1) transition) wavelengths. Moreover, using an arrow, we show the approximate location and orientation of the faint Hα tail, seen by Gavazzi et al. (2001) in their deep Hα images of this galaxy. The IRS slit positioning captures only a small portion of the Hα tail. The side of the galaxy thought to be interacting with the ram-pressure wind is the southernmost edge. From a comparison of the Hα, 8 μm, and 17 μm images, it is clear that the 17 μm emission does not fully match the observed Hα and 8 μm emission, even after considering the effects of spatial resolution. The overabundance of 17 μm emission at the tail end of the galaxy suggests the presence of warm H₂.

Zoom In Zoom Out Reset image size

Upon closer inspection, we detect significant emission of warm H₂ within the galaxy and in a portion of its tail in the extracted spectra. We show the extracted spectra within the full galaxy extraction region and tail extraction region in Figure 5. Within the full galaxy extraction region, H₂ lines in addition to a few fine structure lines dominate the spectrum. There are significant detections of the ground state rotational H₂ S(0) to S(3) lines. In the tail extraction region, the strongest lines are H₂ lines with the S(1) line being the strongest. What is particularly striking within this region is the large ratio of the H₂ line relative to continuum emission. It is clear that a significant fraction of the flux in the spectrum arises from H₂ line emission from the tail region. The H₂ emission of this galaxy is as dramatic as was observed in ESO 137-001, which is discussed at length in Paper 1. Bright fine structure lines, such as [Ne ii], [Ne iii], [S iii], and [Si ii], are also observed in both extraction regions.

Zoom In Zoom Out Reset image size

We calculate the column density, temperature, and total mass of the H₂ gas in the two regions using the methods outlined in Paper 1. To calculate the column density associated with each line transition, we use solid angles, Ω, of 5.16 × 10⁻⁸ and 1.30 × 10⁻⁸ sr for the galaxy and tail extraction regions, respectively. The excitation diagrams of the H₂ ground-state rotational transitions for the two regions are shown in Figure 6. The column densities for both regions cannot be fit with a single temperature model, because there appear to be multiple temperature components. This was also the case for ESO 137-001. We adopted the same two-temperature-component model discussed in Paper 1. The model (a hot and a warm component; Equation (4) in Paper 1) fits adequately for both extraction regions. The fit results are tabulated in Table 4. For the whole galaxy, we determine that the warm component has a temperature of $130^{+6}_{-12}$ K and a column density of $6.6^{+2.0}_{-0.8}\times 10^{19}$ cm⁻². For the hot component, we find a temperature of $593^{+233}_{-115}$ K and a column density of $2.6^{+3.1}_{-1.5}\times 10^{17}$ cm⁻².

Zoom In Zoom Out Reset image size

Table 4. Measured H₂ Properties and Gas Masses for Galaxies

Galaxy	Region	Component	Tex	Ntot	Σ	H2 Mass
			(K)	(c m−2)	(M☉ pc−2)	(M☉)
CGCG 97073	Full	Warm	$130^{+6}_{-12}$	6.$6^{+2.0}_{-0.8}$ × 1019	1.$1^{+0.3}_{-0.1}$	4.$7^{+1.4}_{-0.5}$ × 108
		Hot	$593^{+233}_{-115}$	2.$6^{+3.1}_{-1.5}$ × 1017	4.$2^{+4.9}_{-2.4}$ × 10−3	1.$8^{+2.2}_{-1.0}$ × 106
	Tail	Warm	$114^{+17}_{-13}$	1.$2^{+0.6}_{-0.4}$ × 1020	1.$9^{+0.9}_{-0.6}$	2.$1^{+1.0}_{-0.7}$ × 108
		Hot	$480^{+314}_{-44}$	6.$5^{+3.4}_{-5.0}$ × 1017	1.$0^{+0.5}_{-0.8}$ × 10−2	1.$2^{+0.6}_{-0.9}$ × 106
NGC 4522	Central	Warm	$116^{+13}_{-14}$	1.$9^{+1.0}_{-0.5}$ × 1020	3.$0^{+1.6}_{-0.8}$	3.$7^{+2.1}_{-1.0}$ × 107
		Hot	$423^{+123}_{-42}$	1.$3^{+0.8}_{-0.9}$ × 1018	2.$1^{+1.3}_{-1.4}$ × 10−2	2.$7^{+1.6}_{-1.7}$ × 105
	NE	Warm	141 ± 2	4.0 ± 0.3 × 1019	0.64 ± 0.04	4.8 ± 0.3 × 106
	SW	Warm	127 ± 2	6.$4^{+0.6}_{-0.5}$ × 1019	1.0 ± 0.1	8.$7^{+0.8}_{-0.7}$ × 106
NGC 1427A	Galactic	Warm	<91	>1.2 × 1020	>1.8	>5.9 × 107
		Warma	<104	>6.6 × 1019	>1.1	>3.4 × 107
	High IR Flux	Warm	<91	>1.7 × 1020	>2.8	>4.1 × 107
		Warma	<103	>1.0 × 1020	>1.6	>2.4 × 107

Note. ^aThis assumes an OPR of 1.5.

Download table as:  ASCIITypeset image

The errors were estimated using Monte Carlo simulations. Because there were no additional degrees of freedom in the fit, given that there are an equal number of data points as there are fit parameters, a new set of best-fit model parameters was derived from each trial run. This assumed that each flux data point was Gaussian distributed with a standard deviation that was equal to its 1σ error and that the data can only be fit by a two-temperature model. On the occasion where the fit failed to converge for a given trial (i.e., the two-temperature model was not a good fit), that trial run was discarded. This occurs when the ln (N/g) data points are not monotonically decreasing as a function of transition temperature, which is most likely unphysical. We carried out 10,000 successful trials to determine the distribution of 1σ errors in the model parameters. The number of trials was sufficiently large to ensure the convergence of the error estimates. For all error estimations in this paper, we follow the same method.

For the tail region, we derived temperatures for the warm and hot component of $114^{+17}_{-13}$ K and $480^{+314}_{-44}$ K, respectively. These components had corresponding column densities of $1.2^{+0.6}_{-0.4}\times 10^{20}$ cm⁻² and $6.5^{+3.4}_{-5.0}\times 10^{17}$ cm⁻². The values of these fits are consistent with those derived for the full galaxy. In both cases, the warm component mass dominates over that of the hot component by a significant margin. The column densities and temperatures for H₂ in these two regions are also similar to the values obtained for the tail in ESO 137-001, suggesting that there is similar physics exciting the H₂ gas in both galaxies.

We measure a total warm H₂ mass of $4.7^{+1.4}_{-0.5}\times 10^8$ M_☉ and $2.1^{+1.0}_{-0.7}\times 10^8$ M_☉ within the full galaxy and tail regions, respectively. This indicates that a significant fraction of the total galaxy warm H₂ is found within the tail region. We compare the warm H₂ mass measurement with other estimates of gas masses measured for this galaxy, such as cold H₂ and H i masses. In particular, we compare the warm-to-cold H₂ to look for any peculiarities. We determine the cold H₂ mass of this galaxy by using the H₂ measurement obtained from Boselli et al. (1994), who carried out CO observations of the galaxy; we applied the appropriate distance and X_CO corrections to account for changes in their assumed values. Boselli et al. (1994) assumed a distance of 65 Mpc to this galaxy and used an X_CO value of 2.3 × 10²⁰ cm⁻²/K km s⁻¹. For our work, we adopt an X_CO of 2.8 × 10²⁰ cm⁻²/K km s⁻¹ from Bloemen et al. (1986), giving an H₂ mass of 1.7 × 10⁹ M_☉. This yields a warm-to-total H₂ mass fraction of 0.21. We assume that the aperture correction effects between the CO measurements and our H₂ measurements are not significant. This fraction is higher than the highest mass fraction value reported by Roussel et al. (2007) in their warm H₂ survey of Spitzer SINGS galaxies with published CO measurements. This is also a factor of two higher than the fraction of 0.1 reported by Rigopoulou et al. (2002) for starburst galaxies in their warm H₂ survey. There has been some suggestion in Stephan's Quintet that shock-heated H₂ may have an anomalously high fraction of warm-to-cold H₂ gas (Guillard et al. 2012a), which also may be the case here. We also compare the warm H₂ -to-total gas fraction of the galaxy. We adopt the H i gas mass of this galaxy of 2.0 × 10⁹ M_☉ measured by Scott et al. (2010). We obtain a fraction of 0.11, which is a significant fraction of the total gas within the galaxy.

We test to see if star-formation is the source of the H₂ excitation by comparing the 24 μm image with the continuum-subtracted H₂ image. Figure 7 presents our results. The 24 μm emission (red) is mainly confined to the part of the galaxy directly opposing the tail, which is presumably the location where the ICM wind hits the galaxy. The emission appears to be mainly confined to the location where most of the star formation is occurring. The blue image is the 8 μm aromatic emission; the green image represents the continuum-subtracted H₂ S(1) transition image. The H₂ emission is clearly offset from the 24 μm emission toward the direction of the tail.

Zoom In Zoom Out Reset image size

3.1.2. NGC 4522

NGC 4522 is an infalling spiral located within the Virgo cluster and is at a projected separation of 1.1 Mpc from the cluster center. This galaxy has a large radial velocity of 1250 km s⁻¹ with respect to the cluster (obtained from NED), which suggests that a large portion of its motion is along our line of sight. We adopt a distance to Virgo of 16.5 Mpc (Mei et al. 2007). We present the Spitzer IRAC, IRS, and GOLDMine Hα images of NGC 4522 in Figure 8. The 3.6 μm image of the galaxy shows a typical unperturbed spiral. However, the Hα, 8, and rest 17 μm images reveal something completely different. Their emission is truncated compared to the radial extent of the stellar disk of the galaxy. We also see extraplanar tail-like features in the southwest quadrant of the galaxy at 8 μm and 17 μm, which is a clear signature of dust and possibly gas stripping. The undisturbed nature of the stellar disk indicates that ram-pressure and not tidal forces must be at play. However, this explanation is somewhat complicated by the fact that the galaxy is approximately 1 Mpc away from the Virgo cluster center and the traditional view is that the galaxy velocity and the density of the ICM at that distance cannot be sufficient to produce significant ram-pressure. However, it is likely that the relative velocity of NGC 4522 within the ICM is much higher than expected either because the ICM is dynamic (i.e., it is moving relative to the cluster mean velocity), and/or the galaxy is traveling at high speeds and is unbound from the cluster (Vollmer et al. 2006).

Zoom In Zoom Out Reset image size

We carry out a more careful analysis by extracting spectra from key regions within the galaxy, as shown in Figure 8. In Figure 9, we present the spectra extracted from the central extraction region shown in Figure 8. We detect warm H₂ in a tail-like feature that extends approximately 4 kpc in length from the midplane of the disk along the southwest (SW) extraction region. The central extraction region was defined to obtain most of the infrared emission originating from the galaxy. In it, we have significant detections of the ground-rotational level H₂ S(0) through S(3) lines. We also observe multiple fine structure lines and significant emission by dust, which is likely associated with star formation. We will explore the meaning of this observation in greater detail in Section 3.2. We also extract spectra from the northeast (NE) and SW regions of the 17 μm emission (see Figure 8 for the exact location of the extraction regions), where the effects of ram-pressure are most obvious. Figure 10 shows the spectra extracted from these two regions. The SW region spectrum was extracted from the higher signal-to-noise archival data set. Due to the poorly constrained and unphysical value of the extinction parameter obtained by PAHFIT, we fixed the optical depth of the 9.7 μm silicate absorption feature to that measured for the central extraction region for the fits of the SW and NE regions. In both regions, we detect both the H₂ S(0) and S(1) lines. The SW region has a detection of the S(7) line, while the NE region has a detection of the S(3) line. The S(7) line detection is not particularly convincing because of the increasing noise in that region, so we do not include it in our model fits. It is clear that the SW region, which contains most of the observed extraplanar 17 μm emission, contains a significant amount of warm H₂. We also observe aromatic features, fine-structure lines, and dust continuum, although the relative ratio of the H₂ and dust emission appears smaller when compared to the central extraction region. We investigate the implications of this difference in further detail in Section 4.2.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To understand the physical conditions of the H₂ present in the galaxy and its tail, we fit an excitation model to the H₂ line fluxes obtained from the three different spectral extraction regions, using the same methods used for 97073. The solid angles for the central, NE, and SW regions are 4.58 × 10⁻⁸, 2.75 × 10⁻⁸, and 3.10 × 10⁻⁸ sr, respectively. The excitation diagram and the best-fit models are shown in Figure 11. The results of these fits are also given in Table 4. For the central region we fit a two-temperature model, which yields a warm component with a temperature of $116^{+13}_{-14}$ K and column density of $1.9^{+1.0}_{-0.5}\times 10^{20}$ cm⁻², and a hot component with a temperature of $423^{+123}_{-42}$ K and a column density of $1.3^{+0.8}_{-0.9}\times 10^{18}$ cm⁻². For the NE region, we only fit a single-temperature model because we only have significant detections of two lines. We obtain a temperature of 141 ± 2 K and a column density of 4.0 ± 0.3 × 10¹⁹ cm⁻². For the SW region, we also fit a single temperature model due to only three significant detections of H₂ lines. We fit the S(0) and S(1) lines in this case. We obtain a temperature of 127 ± 2 K and a column density of $6.4^{+0.6}_{-0.5}\times 10^{19}$ cm⁻². The warm components of all three regions have largely consistent temperatures. We calculate a total warm H₂ mass of $3.7^{+2.1}_{-1.0}\times 10^7$ M_☉, 4.8 ± 0.3 × 10⁶ M_☉, and $8.7^{+0.8}_{-0.7}\times 10^6$ M_☉ for the central, NE, and SW regions, respectively. This is an order of magnitude less warm H₂ mass than was measured for 97073. We also compare the central region warm H₂ value with the measured cold gas mass values to see if the warm-to-cold gas ratio is abnormal in any way. This galaxy has a H i mass of 3.8 × 10⁸ M_☉ (Kenney et al. 2004), and a cold H₂ mass of 3.2 × 10⁸M_☉, derived using CO measurements (Smith & Madden 1997) corrected to our adopted distance to the galaxy and X_CO. This corresponds to a warm-to-cold H₂ fraction of 0.15 and a warm-to-total gas fraction of 0.06. This is largely consistent with what is observed in the SINGS sample with warm and cold H₂ measurements (Roussel et al. 2007).

Zoom In Zoom Out Reset image size

We looked for kinematic signatures of the warm H₂ gas being stripped permanently from the galaxy. The galaxy has a high enough recessional velocity with respect to the cluster average that, if the stripped gas were to be permanently lost to the cluster, then the velocity shift of a spectral line should be observable by the IRS low resolution spectrographs. We chose the 17.035 μm S(1) line for this measurement. We use two different extraction regions along the H₂ tail detected in the SW region: one close to the galactic disk and one farthest away along the tail where a reliable line detection can be obtained. The extraction regions are shown in the 17 μm image in Figure 8 as cyan boxes. The continuum-subtracted line profiles from each of these regions are shown in Figure 12. There is a clear blueward shift of the S(1) line as one moves from just outside the disk of the galaxy to the farthest reaches of the detected tail.

Zoom In Zoom Out Reset image size

We carry out a Gaussian fit to the line profile of each extraction region to quantify the shift in velocity. The fit velocity and FWHM parameters are v = 720 ± 170 and Δv = 2190 ± 280 km s⁻¹ for the region close to the disk. For the region along the tail, we obtain fit values of v = −740 ± 260 and Δv = 1510 ± 490 km s⁻¹. This gives a relative velocity difference of ∼1460 ± 310 km s⁻¹. The relative shift in velocity is in the right direction and magnitude if the gas is lost permanently from the cluster because the measured relative velocity of the galaxy with respect to the cluster mean is 1250 km s⁻¹. To further check the validity of the shift, we investigate the properties of the IRS spectrograph used for these detections. The spectral resolutions of the low-resolution spectrographs are not fixed and vary with wavelength. The spectral resolution at the 17 μm line with the LL2 spectrograph is approximately 100. This corresponds to a velocity width of an unresolved line of ∼3000 km s⁻¹, with two pixels per resolution element. With these spectrograph characteristics and our achieved signal to noise, one should be able to detect a shift of approximately a quarter of a line width, which corresponds to a velocity of 750 km s⁻¹. This value is slightly larger than the wavelength uncertainty in the calibration of 600 km s⁻¹. We conclude that we may have detected a velocity shift. No other mid-IR line, such as H₂ S(0) or [Si ii] 34.8 μm, could be identified in the extraction region farthest from the galaxy to confirm this result. If this tentative result is correct, some of the stripped gas is being permanently lost to the ICM. Other measurements that might confirm this interesting result are inconclusive. Kenney et al. (2004) also see a blueward shift in their Very Large Array (VLA) H i measurements of the stripped extraplanar gas in this galaxy. However, the blueshift is only modest, at the level of 10–20 km s⁻¹, and does not extend as far from the galactic plane as our measurement. CO observations were carried out by Vollmer et al. (2008), but they are even less extended than the H i map.

3.1.3. NGC 1427A

NGC 1427A is a dwarf irregular galaxy within the Fornax cluster with arc-like stellar morphology. The galaxy has a moderate line-of-sight velocity with respect to the cluster mean of 650 km s⁻¹ (reported by NED). We adopt a distance of 20.0 Mpc to the cluster measured using the surface-brightness fluctuation technique (Blakeslee et al. 2009). The galaxy is at a projected separation of 130 kpc from the cluster center. We present the HST Hα, Spitzer 3.6 μm, 8 μm, and rest 28.2 μm images in Figure 13. The 28.2 μm image is presented instead of a 17 μm image, because no significant emission was detected at 17 μm. The 28.2 μm image wavelength is centered on the H₂ S(0) line and also contains dust continuum emission associated with star formation. We did not subtract the dust continuum due to the low signal-to-noise detection of the S(0) line in each spatial element within the spectral cube (spaxel). The HST Hα image has been smoothed to better reveal the H ii regions within the galaxy. The arc-like star-formation morphology seen at 8 μm and Hα and filamentary star-forming regions seen at the western side of the galaxy at 8 μm cannot be explained purely by a tidal mechanism and must require ram-pressure stripping. A tidal mechanism would affect the entire galaxy, but star formation is restricted to the periphery of the galaxy where the interaction between the ISM and the ICM is the greatest. The bulk of the 28.2 μm emission is mainly centered on the part of the star-forming arc on the south side of the galaxy in an area where there is comparatively much more star formation than the rest of the galaxy. Other sources of 28.2 μm emission within the galaxy seem to coincide with the peaks in 8 μm emission.

Zoom In Zoom Out Reset image size

We search for signatures of warm H₂ within this galaxy by extracting spectra from two different regions; the dimensions and locations of the extraction regions are shown in Figure 13. We extract spectra from a region that covers most of the galaxy and one that focuses primarily on the high IR flux region within the galaxy. The extracted spectra for each region are shown in Figure 14. The S(0) line is clearly detected in both regions, but there is no hint of the S(1) line; this galaxy is the only one in our sample without a firm detection of the S(1) line. In comparison, the S(1) line is the strongest H₂ line in units of flux density observed in the other three galaxies. Because the signal-to-noise ratios of the spectra for this galaxy are low, it is worth putting this result in context. The 3σ upper limit to the S(1) to S(0) line ratio is unusually low with a value ⩽0.4. The typical values of the line flux ratio measured in other galaxies in our sample fall in the 2–6 range, so this upper limit is a factor of five less than the lowest ratio we measure in our sample.

Zoom In Zoom Out Reset image size

Given that our sample is relatively limited, it may not be unusual to find systems with such a low line ratio in a larger sample with measured H₂ rotational line fluxes, such as the SINGS galaxies (Roussel et al. 2007). The SINGS sample has a median line flux ratio of 3.1, which falls well within the range of values observed in the remaining three galaxies in our sample. However, there are three SINGS systems with unusually low S(1)/S(0) ratios (i.e., <1.0): NGC 24, NGC 1705, and NGC 4552. All three systems have low ortho-to-para ratios (0.5 < OPR < 1.5) with warm gas component temperatures (78 K < T < 127 K) on the low end of the observed distribution in all sample galaxies. Two of these galaxies, NGC 24 and NGC 1705, are dwarfs and also have comparably low S(1)/S(0) ratios of 0.3–0.4. The fact that we do not see the S(1) line in NGC 1427A can mean two things. (1) The warm H₂ within NGC 1427A is cooler than the gas found in the other three galaxies in our sample; and/or (2) the OPR for the H₂ gas is lower than our assumed value of three. Typically, the critical densities of the S(1) and S(0) transitions are low enough that they are thermalized in most galaxies. However, if the gas density is sufficiently low, the even and odd states of H₂ remain decoupled and the OPR of the H₂ gas is essentially set to the value of the gas when it initially formed. Roussel et al. (2007) suggest in their work that this may explain the low observed OPR values in the three galaxies in their sample.

We construct an H₂ excitation diagram, Figure 15, for the two regions from which the spectra have been extracted. The solid angles of the galactic and high IR flux regions are 7.96 × 10⁻⁸ and 3.71 × 10⁻⁸ sr, respectively. The column density values shown in the diagram assume a nominal OPR of three. We can place constraints on the total warm H₂ mass detected within the two extraction regions by using the detection of the S(0) and the 3σ detection limit for the S(1) line. To explore the effects of the OPR, we compute H₂ mass estimates for corresponding values of 1.5 and 3 using a single temperature model. The results of these fits are given in Table 4. The fit within the galactic extraction region yields a temperature upper limit of 91 K and a H₂ column density lower limit of 1.2 × 10²⁰ cm⁻². Similarly, the high IR flux region is best fit by a temperature of <91 K and an H₂ column density of >1.7 × 10²⁰ cm⁻². The temperature is a few tens of Kelvin cooler than the warm component temperature observed in the other three galaxies in our sample, but is consistent with the values derived from galaxies exhibiting low S(1)/S(0) ratios in the SINGS sample. This yields a H₂ mass lower limit of 4.7 × 10⁷ M_☉ and 5.9 × 10⁷ M_☉ for the high IR flux and galactic regions, respectively. These results indicate that most of the H₂ mass is concentrated within the high IR flux region, where there is significant star formation. For an OPR of 1.5, we obtain an upper limit to the temperature of 104 K and a lower limit to the column density of 6.6 × 10¹⁹ cm⁻² for the galactic region. This in turn translates into a lower limit on the warm H₂ gas mass of 3.4 × 10⁷ M_☉, which is a 40% reduction in mass compared to the OPR = 3 case. For the high IR flux region, we obtain the corresponding temperature upper limit and column density lower limit of 103 K and 1.0 × 10²⁰ cm⁻². This yields a warm H₂ mass upper limit of 2.4 × 10⁷ M_☉. There is a possibility that even lower OPR values may apply to the H₂ gas within NGC 1427A. This would push the upper limit of the temperature even higher than what has been found with our analysis, while reducing the lower limit on the total mass.

Zoom In Zoom Out Reset image size

Without a firm detection of the S(1), S(2), and S(3) lines, it is impossible to constrain the OPR ratio. We will conservatively assume that the lower limit to the H₂ mass of NGC 1427A is 3.4 × 10⁷ M_☉. To set a reasonable upper bound on the warm H₂ mass, we compute the mass using the lowest temperature of warm H₂ gas observed in the SINGS sample of 78K. It is worth noting that this temperature is observed in NGC 1705, which is one of the dwarf low OPR galaxies discussed previously. This temperature choice yields a value of 1.3 × 10⁸ M_☉. We compare this with the published value of H i mass of 3.3 × 10⁹ M_☉ (Koribalski et al. 2004). The ratio of warm H₂ to H i lies in the range ∼0.01–0.04. This limit is lower than what has been observed in 97073 and NGC 4522, suggesting that there is nothing unusual about the warm H₂ emission observed in NGC 1427A, and it may just be produced by star-forming activities. In the discussion, additional evidence involving the H₂–PAH flux ratio further substantiates this conclusion.

In this section, we study the star-forming properties of the galaxies within our full sample and present the results of individual galaxies in the sections below. All galaxies show significant star-forming activities and marked asymmetries in their distribution of star-forming regions. We will not discuss the star-forming properties of ESO 137-001 because they have already been discussed in detail in Paper 1. To understand star formation better in these ram-pressure stripped galaxies, we present the 8 μm excess images in Figure 16 and IRS-derived 24 μm images in Figure 17. Even though both bands trace star formation, their utility to our analysis varies. The 8 μm emission offers a high-resolution view of the morphology of star-forming activities, but it is a poor indicator of the star-formation rate. The 24 μm emission has poorer spatial resolution compared to 8 μm, but it is well calibrated to provide accurate measures of star-formation rates within the galaxies in question. Therefore, these bandpasses offer a complementary view of star-forming activities.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Thus, we have two goals. The first is to probe any effects of the ICM on star formation by comparing the morphology at 8 μm with the understanding of the stripping process from the H₂ behavior. The second is to see if the overall level of star formation in these galaxies is suppressed. This second issue has been examined for other cluster galaxies by Tyler et al. (2013). They find that the majority of star-forming galaxies in dense clusters fall close to the field galaxy SFR/mass relation. That is, if one takes the field relation and the width of the field distribution around this relation (FWHM ∼ 1.2 dex; Brinchmann et al. 2004), only a small minority of cluster galaxies fall significantly below expectations for field galaxies of similar mass.

3.2.1. CGCG 97-073

We confirm previous observations that the majority of the star formation is confined to one side of the galaxy, specifically the side opposite the Hα tail (Gavazzi et al. 2001). The arc-shaped star-forming region observed by Gavazzi et al. (2001) at Hα (see Figure 4) is also seen in our 8 μm excess image of the galaxy. In fact, the Hα and 8 μm emission are virtually identical with the exception of the 50 kpc tail and a northern spur observed at Hα. Significant 24 μm emission is also detected from this galaxy. The 24 μm emission is offset toward the direction with significant star formation and the emission is centered on the area where the greatest amount of emission is seen in both Hα and 8 μm. Our observations strengthen the original hypothesis proposed by Gavazzi et al. (2001) that the galaxy is interacting with the ICM on its southern edge and that ram-pressure is compressing the molecular gas on that edge and promoting increased star formation along an arc-like star-forming region. This is further supported by the lack of a similar feature at 3.6 μm. No obvious extraplanar star-forming regions or dust trails are seen at 8 μm like those observed in ESO 137-001. However, it is important to note that the galaxy and its orbit may be highly inclined to our line of sight due to its asymmetric morphology and high line-of-sight velocity, with respect to the cluster mean (v_gal = 680 km s⁻¹) compared to the cluster velocity dispersion of 891 km s⁻¹ determined by Cortese et al. (2004).

We measure the MIPS 24 μm flux for the two spectral extraction regions for the galaxy. We use the same techniques discussed in Paper 1 to calculate the equivalent MIPS 24 μm signal from the extracted IRS spectra. However, we use a slightly different error estimation technique to quantify the error in the flux measurement because our previous error estimation technique was far too conservative. The random errors derived from the CUBISM generated spectra do not reflect the fluctuations present in the IR continuum due to systematic errors that arise from the reduction. We attempt to properly quantify the error by following these steps: (1) we focus on the 20–30 μm spectral range, because that is roughly the MIPS 24 μm bandpass, and we remove any spectral lines present, namely, the 28.2 μm S(0) line; (2) we carry out a polynomial fit to the spectrum and subtract the fit from the spectrum to remove the smooth component associated with the continuum; and (3) we estimate the scatter about the mean by computing the standard deviation of the subtracted spectrum. We take the error for each flux data point within the 20–30 μm range to be the scatter computed in step three. The calculation of the MIPS 24 μm flux involves taking the weighted average of the 20–30 μm spectrum using the MIPS relative response curve as weights. We calculate the error in the 24 μm flux value accordingly, using the relative response curve weights as given in the following equation:

$$\begin{equation} \sigma _{f24} = \left[\frac{1}{\sum _{j}c_j}\sum _i (c_i \sigma _{f,i})^2\right]^\frac{1}{2}, \end{equation} \tag{ 1 }$$

where σ_f24 is the error in the 24 μm flux, c_i is the MIPS relative response value computed at each wavelength value in the IRS spectrum, and σ_{f, i} is the flux error for each IRS data point, which we take to be the scatter value computed previously. We apply this method to all of the galaxies in our sample.

We measure 24 μm fluxes of 18.8 ± 0.3 mJy and 0.30 ± 0.09 mJy for the galaxy and tail extraction regions of 97073, respectively. To determine the star-formation rate (SFR) of the galaxy, we use both the Hα and 24 μm flux measurements to account for obscured and unobscured star formation. Iglesias-Páramo et al. (2002) measure an Hα flux of 1.55 × 10¹³ erg s⁻¹ cm⁻². It is worth noting that this measurement also includes the [N ii] lines because the Hα flux was measured through narrowband photometry. However, for irregular galaxies the [N ii]/Hα ratio is very low (<0.1; Kennicutt & Kent 1983), so we do not correct for this. Using the calibration derived by Calzetti et al. (2010), we derive an SFR of 1.2 M_☉ yr⁻¹. Next, we determine if the galaxy's SFR is unusual in any way for its stellar mass. We estimate the stellar mass using the stellar mass-to-light ratio from the K-band magnitude and B−V color of the galaxy, along with the Bell et al. (2003) mass-to-light ratio relation. 97073 has a K-band apparent magnitude of 12.99 and a B−V color of 0.43 (Gavazzi & Boselli 1996). This translates to a mass-to-light ratio of 0.71 and a corresponding stellar mass of 8.7 × 10⁹M_☉. We calculate a specific star-formation rate (SSFR), log(SFR/M_⋆ [yr⁻¹]), of −9.9. This falls within the normal star-forming sequence for field galaxies (Tyler et al. 2013). This could mean two things: (1) the cluster environment has not yet had time to quench the star formation of this galaxy; (2) ram-pressure has induced star formation that has increased the SSFR of the galaxy as evidenced by an arc-like star-forming region observed in the galaxy. For a further discussion on this topic see Section 4.1.

3.2.2. NGC 4522

Like ESO 137-001, NGC 4522 is a spectacular example of how ram-pressure can strip dust and induce extraplanar star formation. In Figure 16, the 8 μm excess image shows evidence for significant dust stripping. The 8 μm dust disk is much smaller than the stellar disk, hinting that dust at larger radii has already been blown off due to ram-pressure stripping. This is consistent with the outside-in picture of how ram-pressure first strips the least dense gas at larger radii, as it is not as well-bound and then moves inward as the ram-pressure strength increases. On the western side of the galaxy, there is clear evidence for dust being blown out of the disk. The most spectacular feature is the one on the southwest end of the galaxy, where the largest amount of extraplanar excess 8 μm emission is observed. The excess 8 μm emission extends approximately 3.5 kpc from the mid-plane of the galactic disk, which is similar to the distribution of extraplanar warm H₂ emission. In addition to the diffuse dust emission at this location, we also see large clumps, presumably associated with star-forming regions, as they are also seen at Hα. The diffuse extraplanar dust emission is also visible at 24 μm, but we do not observe the clumps seen at 8 μm and Hα. This may be due to a combination of reduced resolution and sensitivity at 24 μm. This observation clearly shows that dust is also significantly affected by ram-pressure and can exist some distance from the plane of the galaxy. Similar behavior was also observed in the case of ESO 137-001, where 8 μm excess sources were seen past the tidal radius of the galaxy, although the dust seen here does not extend that far. The morphology of the dust also reveals the turbulent nature of ram-pressure stripping.

We search for optical counterparts for the stellar-continuum subtracted 8 μm emission with a particular focus on the areas with extraplanar dust. We present the color composite of HST F435W, F814W, and Spitzer 8 μm excess emission in Figure 18. The large dusty clumps located immediately west of the galaxy show some blue star forming knots associated with them, but the dust emission is slightly offset to the west. There is a single 8μm source located at the western edge of the frame that has no optical counterpart. It is unclear if the source is associated with the galaxy, although it could possibly be an extraplanar star-forming region. Searches within a number of catalogues (NED, SIMBAD, and VIZIER) yielded no matches to this source.

Zoom In Zoom Out Reset image size

We measure the 24 μm flux within the central, NE, and SW regions and obtain 102.8 ± 1, 6.6 ± 0.6, and 8.0 ± 0.3 mJy, respectively. Using the measured Hα flux of 4.2 × 10⁻¹³ erg cm⁻² s⁻¹ (Gavazzi et al. 2006) and the 24 μm flux within the central extraction region, we obtain a star-formation rate for the galaxy of 0.12 M_☉ yr⁻¹. The extraction apertures for the Hα and 24 μm measurements do not match exactly, but they should contain most of the flux observed at these two wavelengths. We crosscheck the measured star-formation rate using 100 μm Herschel Space Telescope photometry of this galaxy (Davies et al. 2012). Using the 100 μm flux–SFR relation from Rieke et al. (2009), we convert a 100 μm flux density of 5.12 Jy to a SFR of 0.16 M_☉ yr⁻¹, which is consistent with our original measurement. The 100 μm derived SFR is slightly larger because the Herschel measurement covers the entire galaxy, whereas we only consider the central extraction region. To be consistent with other galaxies in our sample, we only use the Hα/24 μm derived SFR for further analyses. We measure the SSFR for this galaxy to determine if the star-formation rate is similar to that for a typical star-forming galaxy. We calculate the stellar mass in the same manner described previoulsy. This galaxy has a K-band magnitude of 9.94 (Devereux et al. 2009) and a B−V color of 0.59 (Schroeder & Visvanathan 1996), which yields a stellar mass-to-light ratio of 0.75 and a stellar mass of 4.4 × 10⁹ M_☉. This corresponds to a log(SSFR) of −10.6. This galaxy falls below the typical range of field SSFR values for its stellar mass (Tyler et al. 2013). This may indicate that its SFR has been suppressed through environmental processes, but it is worth noting that its SFR is not significantly lower than the observed range of typical SFR values for field galaxies. All evidence suggests that the galaxy has already experienced significant ram-pressure stripping. It is likely that this galaxy has already passed through the densest part of the ICM and is on its way out. This hypothesis is investigated further in Section 4.1.

3.2.3. NGC 1427A

The fine-structure line flux ratios can be used to determine the excitation mechanism for the gas, and can provide clues about the hardness of the incident radiation field. Different excitation mechanisms produce different ionization ratios and therefore fine-structure lines emitted from photodissociation regions in galaxies can be used to delineate harder excitation sources (e.g., AGN) from softer ones (e.g., star formation). Dale et al. (2006) in their Spitzer IRS study of Seyfert, LINERS, H ii nuclei, and extranuclear H ii regions, found that they could isolate each type of source in [Ne iii] 15.56 μm/[Ne ii] 12.81 μm and [S iii] 33.48 μm/[Si ii] 34.82 μm space. We carry out a similar analysis to see if there are any unusual signatures in our derived fine structure line fluxes, which may either be associated with a shock or excitation by the hard cluster X-ray radiation field. For example, if the main excitation mechanism of the gas is X-ray irradiation from the ICM, one might expect more AGN-like line ratios. The fine structure line ratios for the spectral extraction regions for all our galaxies are shown in Figure 19.

Zoom In Zoom Out Reset image size

We find that the ionization state of the gas for three of our sample galaxies, 97073, NGC 4522, and ESO 137-001, falls within the star-forming region locus of the [Ne iii]/[Ne ii] versus [S iii]/[Si ii] plot. However, NGC 1427A shows a rather unusual ionization state that is similar to Seyferts and LINERs, which may be due to its close proximity to the center of the Fornax clusters X-ray emission. In galaxies where we see excess warm H₂ emission (i.e., 97073, NGC 4522, and ESO 137-001), we do not see any unusual fine-structure line ratios; there is no indication that these lines are excited by the hard X-ray emission within the cluster, even in the tail regions of these galaxies. In these cases, the observed lines are most likely emitted within PDRs associated with star-forming regions, and not associated with the unusually excited molecular hydrogen.

The unusual [Si ii] 33.5 μm/[Si ii] 34.8 μm line ratio observed in NGC 1427A could be entirely due to the faintness of the [Si ii] 33.5 μm line. One way to determine if the line flux is peculiar is to compare it to a higher excitation line of the same species, the 18.7 μm [Si ii] line. The typical [Si ii] 18.7 μm-to-[Si ii] 33.5 μm ratio for the SINGS sample is 0.82 with a standard deviation of 0.27 (Dale et al. 2006). In the case of NGC 1427A, the ratio is much higher (>2). It is also worth noting that the [Si ii] 18.7 μm emission is not centered on, but is slightly offset inward from, the brightest star-forming region. This suggests that the faintness of the [Si ii] 33.5 μm is likely real and is not likely a signal-to-noise issue. This could be potentially explained by two different effects: a hard UV radiation field and/or increased electron densities. While the former seems likely due to a galaxy's proximity to the center of Fornax where the density of the hot ICM is the greatest, one can compare the [Ne iii]/[Ne ii] ratio, which is a good indicator of the hardness of radiation, with galaxies with known hard UV fields, such as blue compact dwarfs (BCDs). For a select sample of low metallicity BCDs, Wu et al. (2008) find typical ratios of ∼3–5, which are consistent with the lower limit set by our measurements. The lower [S iii]/[Si ii] ratio can also be explained by increased electron densities. A ratio of >2 is consistent with densities >10³ cm⁻³ (Rubin 1989). This galaxy already exhibits significant compression through ram-pressure, which could lead to the enhanced densities. Based on the available information, it is not possible to distinguish between these two scenarios.

The blended feature of [O iv] 25.9 μm and [Fe ii] 26.0 μm has been detected in most spectra with the exception of those from NGC 1427A. The weakness of the higher ionization [Ne iii] 15.6 μm line makes [Fe ii] the most likely line detected in the blended feature. As discussed in Cluver et al. (2010), the [Si ii]/[Ne ii] and the [Fe ii]/[Ne ii] line ratios can be used as a diagnostic for shocks. For our sample, typical values of [Si ii]/[Ne ii] fall within a narrow range of 0.87–1.35, whereas the [Fe ii]/[Ne ii] ratio ranges from 0.05 to 0.20, which are similar to values observed by Cluver et al. (2010). Typical SINGS galaxies that are not experiencing ram-pressure stripping have [Fe ii]/[Ne ii] ratios lower than 0.05 (Wong et al. 2014). Although the elevated values suggest the presence of shocks, they cannot be explained entirely by either dissociative J-shock (Hollenbach & McKee 1989) or fast shock (Allen et al. 2008) models. J-shock models for density ranges of 10³–10⁴ cm⁻³ for a broad range of velocities produce [Si ii]/[Fe ii] ∼ 1. Fast shock models on the other hand produce large [Si ii]/[Ne ii] ∼10 and [Fe ii]/[Ne ii] ∼1 ratios (Cluver et al. 2010). A solution to this mismatch is the depletion of either Si or Fe through shocks, which has been posited by both Cluver et al. (2010) and Wong et al. (2014).

To determine the effects of ram-pressure on star formation, we place our measurements of the star-forming properties of the galaxies in a broader context by comparing with other studies of cluster and field galaxies. It is clear that ram-pressure is having a significant effect on the star-forming properties of the galaxies simply from the observed morphological differences compared to undisturbed galaxies, such as displaced dust disks, Hα trails, and arc-like star-forming regions. To test whether the overall star-forming rates are also affected by ram pressure, we compare to measurements of other field galaxy samples. We present the specific star-formation rates as a function of stellar mass in Figure 20 for all four galaxies in our sample using Hα and 24 μm measurements to estimate the SFR and stellar mass, which were presented in Section 3.2.

Zoom In Zoom Out Reset image size

In order to determine if ram-pressure stripping has significantly affected these galaxies, we compare their SSFRs with typical values for field galaxies. While many measurements of the field galaxy star-formation rates exist, we compare with Tyler et al. (2013) measurements due to the similarity of their methods with ours. Tyler et al. (2013) also employ Hα and 24 μm flux measurements of galaxies to determine SFRs, and use these to estimate the typical star-formation rates of local field galaxies as a function of their stellar mass. We present our results of the comparison in Figure 20, where it is seen that three out of four of our galaxies, 97073, NGC 1427A, and ESO 137-001, lie within the 1σ spread of star-formation values of typical field galaxies. The four galaxies in our sample are, as an ensemble, consistent with being drawn from the field distribution (i.e., one galaxy, NGC 4522, just beyond 1σ is expected and seen). For reference, we also compare with other star formation studies of field galaxies that have been done with UV and/or Hα measurements (e.g., Brinchmann et al. 2004; Salim et al. 2007). UV and Hα measurements, which often rely on an accurate determination of extinction and the intrinsic UV colors of the underlying stellar population, are not directly sensitive to heavily obscured star formation, which can significantly influence estimates of SFR. This can yield significant systematic offsets in SFR measurements obtained using different methods as seen in Figure 20, where the Salim et al. (2007) relation is also plotted (see Boquien et al. 2012 and Elbaz et al. 2007 for more details). However, using both Hα and 24 μm measurements directly accounts for both unobscured and obscured star formation as we have done in our case. In conclusion, 97073, NGC 1427A, and ESO 137-001 show signs of healthy star formation, while NGC 4522 shows, at most, a small amount of suppression.

This means that ram-pressure has not significantly altered the star-forming rates of these galaxies. However, there is an added possibility that ram-pressure induced star formation may compensate for any overall global reduction in star formation, as some galaxies show signs of ram-pressure enhanced star formation on the windward side of the galaxy. There is evidence observed in Virgo galaxies, including NGC 4522, that there is an enhancement in the molecular gas fraction and a possible increase of the star-formation efficiency in the windward side of the galaxies (Vollmer et al. 2012), which together might enhance the star-formation rate. However, in this same work, there are signs that stripped gas, having lost the gravitational confinement of the disk, can have lower star-formation efficiency. Vollmer et al. (2012) explain that without the gravitational potential of the disk, shocks, as observed in this work, can increase the thermal and turbulent pressure of the gas and significantly decrease its star-formation efficiency as it disperses when blown out of the galaxy. This might explain why even though we observe warm H₂ emission where molecular gas has been stripped, we do not see a commensurate increase in star-formation activity in those regions.

H₂ is typically excited within photodissociation regions through UV fluorescence in star-forming regions. This explains most of the warm H₂ emission detected in the nearby SINGS galaxy sample (Roussel et al. 2007), which we use as a reference. However, our goal is to detect anomalous H₂ excitation that is likely associated with ram-pressure stripping. Unfortunately, it is impossible to determine the source of H₂ excitation from just the lowest ground state rotational transition line ratios because they have low enough critical densities that they are typically thermalized (Roussel et al. 2007). To disentangle the excitation mechanism for the H₂ emission, we use two different metrics that use additional information to determine if the H₂ emission is anomalously higher than what is produced by just star formation.

The first metric, presented in Paper 1, is the ratio of MIPS 24 μm flux, a proxy for star formation, and the sum of the line fluxes of the ground-state rotational H₂ lines S(0) through S(3), a proxy for total H₂ line emission. We apply our metric to all of the SINGS galaxies in the Roussel et al. (2007) sample and show the results in Figure 21. The error bars shown only include the error in the H₂ line fluxes and not the 24 μm fluxes because 24 μm fluxes typically have very small errors. Almost all the galaxies have a ratio of less than 0.03, which we define as the threshold for anomalous H₂ emission. Roussel et al. (2007) highlight NGC 4450 and NGC 4579, two galaxies that are significant outliers in this relation, because their H₂ excitation cannot be explained by star formation or supernova remnants alone. They evoke either X-ray irradiation or shocks through cloud collisions as a possible explanation for the anomalous values.

Zoom In Zoom Out Reset image size

We calculate the values for the metric for all of our galaxies and plot the values in Figure 21. H₂ fluxes are obtained from line fits of the extraction regions explored in this paper and Paper 1. If the detections of the S(1), S(2), and/or S(3) lines are not significant, we do not include them in the H₂ line flux sum. The error bars show both the errors in the 24 μm fluxes and the root-sum-squared flux for the ratio that includes both the error in the H₂ line flux and the 24 μm flux. The values obtained from regions that represent the galaxies as a whole are plotted with larger symbols, while the other regions of interest are plotted with smaller symbols. The individual symbols for each galaxy are explained in the figure captions. It is clear that for 97073, ESO137-001, and NGC 4522 the extraction regions where stripping is occurring or one sees a tail have metric values that are outliers and therefore have significantly enhanced warm H₂ emission that cannot be explained simply by star formation. The tail region of 97073, represented by the small red triangular symbol, is definitely an extreme example because it has significantly more warm H₂ excess emission than any other galaxy in the sample. In fact, the offset H₂ emission seen in 97073 strongly suggests that H₂ is being dissociated from the head of the galaxy, where significant star formation is occurring and reforming some distance behind the area of intense star formation in an excited state. Furthermore, a significant fraction of its H₂ mass is in the warm state. In terms of anomalous metric values, the far tail region of ESO 137-001 comes in as a distant second, followed by the SW region of NGC 4522. If one looks at metric values for the galactic extraction regions themselves, ESO 137-001 and NGC 4522 fall within the normal values representative of star-forming galaxies in the SINGS sample. On the other hand, 97073 has enhanced emission even within the galaxy. NGC 1427A does not have any enhanced H₂ emission within either the galaxy or the high IR flux regions. In fact, it has the lowest values of the metric compared to all other galaxies and exhibits unusual H₂ emission characteristics where only the lowest energy rotational line S(0) is detected.

The second metric more directly tracks star-formation powered photodissociation regions (PDR) as being the location of the warm H₂ emission. In recent work of galaxies with anomalous H₂ emission, the H₂ line flux versus 7.7 μm PAH flux ratio has been an excellent discriminator of anomalous H₂ emission (Appleton et al. 2006; Roussel et al. 2007; Ogle et al. 2010; Guillard et al. 2012b). Ogle et al. (2010) empirically show, through a comparison of SINGS galaxies, that their molecular hydrogen emission galaxies (MOHEGs) are powered either by shocks or cosmic rays because their H₂-PAH flux ratios fall in the 0.04–4 range. Guillard et al. (2012b), in Figure 7 of their work, show the large range of galaxies with anomalous H₂ emission and confirm that star formation powered H₂ emission cannot explain flux ratios greater than 0.04. They further support this claim by running PDR models using existing code (Le Petit et al. 2006), which were found to be consistent with other models of H₂ emission in PDRs (Kaufman et al. 2006). We present the flux comparison for our measurements in Figure 22. It is worth noting that the far tail region of ESO 137-001 where this ratio would be the highest in that system is not included due to the lack of short wavelength IRS coverage. We also plot the ratios for star forming SINGS galaxies using the 7.9 μm and H₂ line flux measurements carried out by Roussel et al. (2007). It is clear that the star-forming SINGS galaxies form a narrowband in H₂/PAH flux ratio with typical values slightly less than 0.01. The results of our measurement are consistent with our conclusions from the use of our first metric. The regions that are experiencing stripping fall above the 0.04 flux ratio cut, with the exception of NGC 1427A, which is not thought to be experiencing significant stripping. 97073 remains the most dramatic example, falling well above the relation with almost 60 times the flux ratio of star-forming SINGS galaxies. NGC 4522 shows modestly elevated flux ratios with almost 10 times the SINGS average. This is slightly higher than similar measurements made by Wong et al. (2014) of the same galaxy, where they find an increase of 3–5 times the SINGS average. The difference may be attributed to different extraction regions and our inclusion of the S(3) to the H₂ flux measurement. Nevertheless, two conclusions can be drawn from this comparison: (1) star formation itself is insufficient for exciting the H₂, and (2) we observe a sequence of galaxies with varying levels of H₂ excitation, which may be due to the current evolutionary state of the galaxy; this is discussed further.

Zoom In Zoom Out Reset image size

As discussed in the introduction, a number of potential excitation mechanisms for H₂ beyond star formation were reviewed. To get a better understanding of potential excitation mechanisms, we can estimate the timescale of warm H₂ emission if there was no in situ energy source. Using the ideal gas law for a diatomic molecule, the timescale τ can be approximated to be

$$\begin{eqnarray} \tau & \sim & \frac{5}{2}N_{\rm H_2} k T_{\rm H_2}/L_{\rm H_2}, \end{eqnarray} \tag{ 2 }$$

where N, T, and L are the number, temperature, and luminosity of H₂ molecules, respectively. For our estimation, we only consider the warm component of the H₂ because it has been characterized in the tail regions of all three galaxies with anomalously high H₂ emission. We take $L_{H_2}$ to be the sum of the luminosities of the S(0) and S(1) lines. We obtain an $L_{H_2}$ of 3.2 × 10⁶ L_☉, 5.6 × 10⁶ L_☉, and 4.6 × 10⁵ L_☉ for the Full Tail region of ESO137-001, the Tail region of CGCG97-073, and both the NE and SW tail regions of NGC 4522, respectively. The timescales of these regions correspond to 4, 8, and 8 kyr. One can also estimate the true timescale of the H₂ emitting regions if the geometry of a galaxy's orbit and its velocities are known. ESO 137-001 is the best candidate for this measurement because its orbit is along the plane of the sky. Taking a conservative galaxy velocity of 1600 km s⁻¹ and a tail length of 20 kpc (Sivanandam et al. 2010), we obtain a timescale of its H₂ tail of 12 Myr. A similar estimation of NGC 4522 can be made, but its orbit is not along the plane of the sky. With its projected tail length of 4 kpc and a galaxy velocity estimate of 1000 km s⁻¹ along the plane of the sky, we obtain a timescale of 4 Myr. It is clear from our sample that the true timescales of these warm H₂ tails are much longer than the radiative timescales of the warm gas. This means that there must an in-situ source of energy that continuously heats the gas.

Of the several potential excitation sources, such as MHD waves, conduction, cosmic ray heating, and X-ray dissociation regions (XDRs), shock excitation seems to be the most likely explanation. We can make a morphological argument to support this view. In most cases, it is clear that the central regions of these galaxies do not show anomalous H₂–PAH flux ratio, whereas their stripped H₂ trails show significant enhancement of H₂ emission. This is especially seen in 97073 in Figure 7, where there is a clear offset between the regions with significant star formation where H₂ does exist and those where warm H₂ is observed. The cluster-centric sources such as MHD waves and cosmic ray heating should not discriminate where in the galaxy they deposit their energy. Moreover, these processes generally yield hotter (T ∼ 300 K; Johnstone et al. 2007) H₂ temperatures. For similar reasons, conduction and XDRs cannot adequately explain the morphology. Of course, if the observed excitation is indeed due to turbulent shocks, which are induced by ram-pressure stripping, there needs to be sufficient kinetic energy dissipated to account for the observed H₂ luminosity.

To calculate the amount of energy extracted from the ram-pressure stripping process, we present a toy model to estimate the loss in kinetic energy of the galaxy from ram-pressure drag as it falls into a galaxy cluster. This energy would then be injected into the stripped ISM of the galaxy and the ICM either through turbulent or viscous processes as the galaxy travels through the cluster (see Roediger & Brüggen (2008) for a discussion). Ram-pressure, P_ram, introduces a drag force that decelerates the galaxy:

$$\begin{eqnarray} a_{{\rm drag}} & = & -\frac{P_{{\rm ram}}\: A_{{\rm ISM}}}{M_{{\rm gal}}} \end{eqnarray} \tag{ 3 }$$

$$\begin{eqnarray} & = & -\frac{\rho _{{\rm ICM}}\:v_{{\rm gal}}^2\:A_{{\rm ISM}}}{M_{{\rm gal}}}, \end{eqnarray} \tag{ 4 }$$

where a_drag is the deceleration experienced by the galaxy, A_ISM is the galaxy ISM's cross-sectional area, M_gal is the mass of the galaxy, v_gal is its orbital velocity, and finally, ρ_ICM is the density of the ICM at the galaxy position. If we assume the ram-pressure is not changing in relative short timescales, and that the galaxy is traveling face-on to the ICM wind, we can make the following simplification to determine loss of kinetic energy:

$$\begin{eqnarray} \frac{{dE}}{dt} & = & \frac{d}{dt}(M_{{\rm gal}}\: a_{{\rm drag}}\: s) \end{eqnarray} \tag{ 5 }$$

$$\begin{eqnarray} & = & M_{{\rm gal}}\: a_{{\rm drag}} \frac{ds}{dt} \end{eqnarray} \tag{ 6 }$$

$$\begin{eqnarray} & = & -P_{{\rm ram}}\: A_{{\rm ISM}}\: v_{{\rm gal}} \end{eqnarray} \tag{ 7 }$$

$$\begin{eqnarray} & = & -\rho _{{\rm ICM}}\: v_{{\rm gal}}^3\: A_{{\rm ISM}}, \end{eqnarray} \tag{ 8 }$$

where s is the distance traveled by the galaxy. This equation can be written in more convenient units as

$$\begin{eqnarray} \frac{{dE}}{dt} & = & -7.8\times 10^8\left(\frac{\rho _{{\rm ICM}}}{10^{-27}\:g\: {\rm cm}^{-3}}\right) \nonumber \\ & &\times \left(\frac{v_{{\rm gal}}}{1000\:{\rm km\:s}^{-1}}\right)^3\left(\frac{r_{{\rm gal,ISM}}}{10\:{\rm kpc}}\right)^2\:L_\odot. \end{eqnarray} \tag{ 9 }$$

A few observations can be made from this derivation. First, the energy loss is the strongest as an infalling galaxy approaches the cluster core where the ICM density is the highest and the infalling galaxy is traveling the fastest. Second, as the ISM of the galaxy is stripped away, the cross-sectional area decreases, thereby reducing the kinetic energy loss. We calculate the expected kinetic energy loss rate for the three galaxies with anomalous H₂ emission. The calculated energy loss rate for ESO 137-001 with values obtained from Paper 1, ρ_ICM = 2.2 × 10⁻²⁷g cm⁻³, v_gal = 1602 km s⁻¹, and r_{gal, ISM} = 3 kpc, is 6.4 × 10⁸ L_☉. This is a few orders of magnitude more than the observed H₂ luminosity. Moreover, this galaxy has an observed X-ray (Sun et al. 2006, 2010) and Hα (Sun et al. 2007) tails with combined luminosity in the 10⁷ L_☉ range. The combined power output of these tails is consistent with the maximum power produced by the dissipation of the galaxy's kinetic energy through ram-pressure stripping. A similar calculation for NGC 4522 for a parameter choice of ρ_ICM = 10⁻²⁸ g cm⁻³, v_gal = 1500 km s⁻¹ (Kenney et al. 2004), and r_{gal, ISM} = 2.4 kpc (estimated from the 8 μm image) yields a energy loss of 1.5 × 10⁷ L_☉. Finally for 97073, the calculated energy loss rate is 1.2 × 10⁹ L_☉. For this galaxy, ρ_ICM of 1.1 × 10⁻²⁷ g cm⁻³ was obtained from Mohr et al. (1999) and v_gal was estimated to be $\sqrt{3}\sigma$ where σ, the cluster velocity dispersion, was 891 km s⁻¹ (Cortese et al. 2004). r_{gal, ISM} was estimated to be 6.1 kpc from the 8 μm image. All the observed H₂ luminosities can be explained by the energy loss by the ram-pressure drag experienced by the galaxy. What is also particularly interesting is that the calculated energy loss rate seems to track the magnitude of anomalous warm H₂ emission discussed in the previous section. This may be additional evidence for the dissipation of kinetic energy being the source of energy for the observed emission.

We place the results of our survey in a general context by attempting to answer the question of how excess warm H₂ emission relates to ram-pressure stripping. As discussed in Section 4.2, three of the four galaxies show enhanced H₂ emission and all three show evidence for a warm H₂ tail, while two of them, ESO 137-001 and NGC 4522, show strong evidence for extraplanar star formation. The fourth galaxy shows signs of ram-pressure induced star formation, but no enhanced warm H₂ emission or tail. If one arranges all galaxies in ascending order of their total gas mass, assuming ESO 137-001 has a H i mass close to the measured upper limit of 1 × 10⁹ M_☉ (Sun et al. 2007), we would have NGC 4522 with the least mass, followed by ESO 137-001, 97073, and NGC 1427A. In descending order of projected distance from their respective cluster centers we have: NGC 4522, 97073, ESO 137-001, and NGC 1427A.

This suggests a possible evolutionary sequence as an explanation for the observed differences in the properties of the four different galaxies. The first stage is represented by NGC 1427A. NGC 1427A is likely at the initial stages of being stripped by ram-pressure where the ICM does not have significant strength to strip out the molecular and H i gas, as evidenced by the extremely large H i mass measurement for this galaxy (Koribalski et al. 2004). This is plausible because, even though NGC 1427A is very close to the center of the Fornax cluster (only 130 kpc in projected separation), the cluster is very poor and has a fairly low ICM temperature of 1.6 keV (Ikebe et al. 2002). Furthermore, this galaxy has a large recessional velocity with respect to the cluster mean (∼650km⁻¹), and it is also possible that its actual radial separation from the cluster core may be much larger than what is observed. This galaxy was originally chosen to be part of the sample due to its prominent arc-like star forming regions (Georgiev et al. 2006). It also has an unusually large atomic gas reservoir. The galaxy is most likely experiencing some ram-pressure that is able to induce star formation by compressing gas to form stars where the ram-pressure is significant, but the pressure is not high enough to remove molecular gas or excite H₂ through the interaction.

The second stage is likely represented by 97073 where the interaction between the ICM and the galaxy's ISM becomes strong enough to dissociate the molecular gas and reform it in an excited state downstream, possibly through shocks, in addition to inducing star formation along the leading edge of the galaxy. 97073 is in a moderately rich cluster that has a relatively hot ICM with a temperature of 3.6 keV (Ikebe et al. 2002). The images of the galaxy at Hα, and 8 μm clearly show an arc of star formation occurring at the location directly opposite to the Hα tail. The 24 μm image also shows intense dust continuum emission that coincides with the star-forming arc. Furthermore, this galaxy's tail region exhibits the strongest warm H₂ excess of all the regions explored, while the galaxy itself exhibits an excess that is stronger than all other galactic regions considered. It is very likely that this galaxy is just beginning to experience strong ram-pressure. Its excess 8 μm emission is not truncated in any way, which suggests that there is sufficient molecular gas throughout the galaxy to promote star formation. It has a healthy amount of gas, both in molecular and atomic form, and it has a healthy star-formation rate. The current rate of stripping cannot be as significant as ESO 137-001 because we do not see large star-forming knots trailing behind the galaxy. Unfortunately, due to the alignment of the IRS slit, the true length of the warm H₂ tail was not determined.

The third stage is probably represented by ESO 137-001, where a significant fraction of atomic and molecular gas has been stripped from the galaxy and a significant fraction of molecular gas is within the observed Hα and X-ray tails as discussed in Paper 1. This galaxy is also fairly close in projected separation (∼280 kpc) to the core of A3627, which is the hottest cluster in our sample with an ICM temperature of 6 keV (Sun et al. 2006). The galaxy has an upper limit on its H i mass of 10⁹ M_☉ (Sun et al. 2007) and no published values for its cold molecular gas content. The galaxy still has a relatively healthy star-formation rate. It is clear from the 8 μm excess image that the emission is not as extended as the stellar disk and the galaxy has a trail of star-forming knots stretching approximately 12 kpc from the galaxy center. This trail of star-forming knots is mostly embedded within the >20 kpc long warm H₂ tail. The warm H₂ gas within the tail cannot be explained simply by star formation. The existence of long gaseous tails, the long trail of star-forming regions, and truncated dust emission suggest that this galaxy may be further along in its ram-pressure stripping phase than the previous two galaxies.

NGC 4522 possibly represents the last stage of this process. Observational signatures show this galaxy to be similar to ESO 137-001 in many ways. It has a smaller dust disk than its stellar disk. It has extraplanar dust emission albeit at a smaller scale, and similarly a ∼4 kpc long H₂ tail, which has a factor of 5–10 less mass than the H₂ tail in ESO 137-001. This galaxy is also fairly poor in gas, with comparable H i and molecular gas content, and the H i gas is known to be displaced in the direction of stripping (Kenney et al. 2004). All signs suggest that this galaxy is either in a similar phase to ESO 137-001 but is less affected by ram-pressure because this galaxy is in a much cooler cluster and not as dense a region of the ICM, or in a later phase where ram-pressure is becoming less significant. The morphologies of its dust and H₂ trails also suggest that it is on its way out of the cluster core. Vollmer et al. (2006) suggest from their dynamical modeling of VLA H i data that this galaxy must have passed through peak ram-pressure stripping 50 Myr ago. Spectroscopic observations corroborate this finding, where a K+A spectrum is seen in the outer stellar disk of the galaxy. Stellar population modeling suggests that the star formation in the outer stellar disk was quenched ∼100 Myr ago (Crowl & Kenney 2006). This most likely explains the suppressed star formation observed in this galaxy. This galaxy has a few peculiarities because it is approximately 1 Mpc away from Virgo's center and the traditional view is that ICM cannot be dense enough to ram-pressure strip the galaxy significantly in a fairly low temperature cluster (∼2.4 keV; White 2000) at that distance assuming a typical velocity for a galaxy, as discussed previously. However, in this case, the galaxy has a fairly large recessional velocity with respect to the cluster mean (∼1000 km s⁻¹). While it is clear from the multiple observational signatures that the galaxy is experiencing ram-pressure stripping, this may require that the ICM be dynamic, i.e., it is moving relative to the cluster mean, and its density is enhanced, or the galaxy is traveling at very high speeds and is unbound from the cluster (Vollmer et al. 2006).

Our study shows that warm H₂ heated by the interaction between the ICM and ISM is not uncommon in cluster galaxies experiencing significant ram-pressure stripping. In the most extreme cases of warm H₂ emission (i.e., ESO 137-001 and 97073), there are long 40–70 kpc tails observed at other wavelengths. There may be a strong association between the existence of an extraplanar star formation/Hα emission and a warm H₂ tail. However, a greater number of detections of H₂ tails is required to confirm this association. If one assumes such an association is true, the recent discovery of more than a dozen galaxies in Coma with extended Hα knots (Yagi et al. 2010) and UV trails (Smith et al. 2010) may suggest that warm H₂ tails may be present in all galaxies with significant ram-pressure stripping, and may be an indicator of direct stripping of molecular gas.