Evidence for Shock-heated Gas in the Taffy Galaxies and Bridge from Optical Emission-line IFU Spectroscopy

The flux calibration and pointing refinements and final cube construction of these data were performed in three steps following the methods described by Blanc et al. (2013): (i) Relative spectrophotometric calibration was performed, which is applied to all the fibers. Observations were made in a six-point dither pattern (including several fibers) of the standard star Hz 15 (HIP 21776). An algorithm was used to solve for the position of the star on the fibers and determine the spectrophotometric transformation from native (ADU) units across the spectrum to units of erg s⁻¹ cm⁻² Å⁻¹. The results were applied to all the fibers irrespective of throughput. This step resulted in a relative flux uncertainty across the band of ∼8%. (ii) Astrometry and absolute flux calibration, using a bootstrapping method, were used to effectively cross-correlate a reconstructed image of the galaxy derived by integrating the light from each fiber, with a calibrated images of Taffy from the Sloan Digital Sky Survey (SDSS; York et al. 2000) in the g and r bands, suitably convolved to the resolution of the VIRUS-P fiber system. This helped refine the astrometry and the assembly of the final cube from the individual observations of each field. The cross-correlation also allowed the spectrum in each fiber to be absolutely scaled to the SDSS band in question. The details of this procedure are given in Blanc et al. (2013). Tests performed in that paper show that the absolute spectrophotometric flux calibration has a typical accuracy of 15%–30%, after taking into account the uncertainties in SDSS calibration and the VIRUS-P relative spectrophotometric accuracy. (iii) A final flux-calibrated 3D spectral cube was created by combining all the various observational pointing frames into a single interpolated cube with resulting 2 × 2² spaxels (∼0.3 kpc arcsec⁻¹ based on the assumed distance of 62 Mpc). These processes were repeated for the red and blue channels, creating final flux-calibrated blue and red spectral cubes.

The processing and extraction of astrophysical information from the data cubes was done using a combination of IRAF/PyRAF, IDL, and Python routines. Before beginning our analysis, we smoothed the data cubes spatially, but not in the spectral direction, using a Gaussian kernel with a standard deviation of 1.47 pixels. This was done to boost the signal-to-noise ratio in areas that we were interested in, particularly the Taffy bridge region, which has relatively low signal-to-noise ratio compared to the galaxies. This spatial smoothing effectively reduces the noise in spectra from individual spaxels by a factor of ∼2. We used these spatially smoothed cubes for all of the analysis done in this work.

For fitting each individual spaxel in the IFU data we used the IDL software toolkit LZIFU (LaZy-IFU; Ho et al. 2016). LZIFU automates fitting multiple emission lines superimposed on a continuum for multiple spaxels in each channel and provides 2D maps of continuum and line fluxes, velocities, and velocity dispersion. It is capable of fitting emission lines that are superimposed on deep absorption features and also emission lines with multiple velocity components. Figures 1 and 2 show our fitting results for two individual spaxels that show these features in their spectra. Figure 1 shows a spaxel that has strong Hβ absorption and Hβ emission superimposed on the absorption trough. This spaxel lies very close to the center of Taffy-N. This absorption must be accounted for with the continuum fitting to get accurate emission-line fluxes, as well as an accurate line profile. The emission lines also show evidence for double-line profiles in many positions across the system. As an example, Figure 2 shows a spaxel that lies close to the edge of the extragalactic H ii region, which clearly shows two separate velocity components.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

LZIFU works by first fitting the continuum with a custom implementation of the PPXF code (Capellari & Emsellem 2004) within LZIFU and then fitting the emission lines after subtracting the continuum model. The continuum fitting uses a set of model templates that are fit to a combined blue+red channel spectrum. LZIFU also accounts for systematic errors in the models and nonstellar contributions to the continuum data by fitting a multiplicative polynomial simultaneously with the continuum models. The emission lines, along with residuals from sky line subtraction, are masked during the continuum fitting process. We also specified the following lines to be fit (and masked during continuum fitting)—Hβ, the [O iii] λλ4959, 5007 doublet, [O i] λ6300, [O i] λ6364, Hα, the [N ii] λλ6548, 6583 doublet, and the [S ii] λλ6716, 6731 doublet. The extinction-corrected emission-line fluxes for all the lines detected in different regions of the Taffy system (as defined in Figure 3) are tabulated in Table 1.

Zoom In Zoom Out Reset image size

We ran LZIFU on the entire IFU data cube for the Taffy galaxies, which is 58 × 58 and 59 × 59 spaxels with 2227 and 2350 spectral elements for the blue and red channels, respectively. The process of fitting the full cube was not straightforward for several reasons relating to the peculiar kinematics of the Taffy system. The LZIFU software was designed to work best with a galaxy showing slowly varying velocity centroids relatively close to the initial guess for the velocity of the system. In the Taffy system, the velocity range of the emission lines over the whole system was large, with the emission lines sometimes exhibiting complex behavior, in addition to occasionally being observed superimposed on deep Balmer absorption lines. Furthermore, in a large number of spaxels, we found multiple velocity components that did not always move together as a function of position. As a result, a single set of initial guesses for the various starting parameters did not work for the whole cube, but had to be adjusted spatially to achieve good fits, especially in specific regions showing double-line profiles. After an iterative process of fitting with different starting parameters tuned to particular regions, we were able to get a consistent set of smoothly varying results across the whole system. We provide the most relevant LZIFU parameters and their starting guesses in Table 2.

Previous observations of the Taffy system in the visible wavelength range had reported the detection of ionized gas within the galaxies and the extragalactic H ii region (Bushouse 1986, 1987). Because of the sensitivity of the VIRUS-P instrument to very faint diffuse emission, we report the presence of ionized gas throughout the Taffy system—both within the galaxies, the so-called extragalactic H ii region, and also in the bridge between the galaxies. We detect emission from many lines, including Hα, Hβ, the [O iii] λλ4959, 5007 doublet, the [N ii] λλ6548, 6583 doublet, and the [S ii] λλ6716, 6731 doublet lines (see Figures 1 and 2). We also detect strong emission from the atomic oxygen line [O i] λ6300 and sometimes the weaker [O i] λ6364 line in the galaxies and the extragalactic H ii region in the bridge. Although emission lines dominate in many of the locations across the system, in some cases Hβ emission is observed superimposed on a broad absorption trough indicative of a post-starburst population (as seen in Figure 1).

In Figure 3 we present some extracted spectra in several places in the system to provide an overview of the complexity of the kinematics in this system. The spectra show expanded views of the [O iii] λ5007 and Hα lines along the major axis of both galaxies and a sampling of the bridge. As with the previous H i investigations of Taffy by Condon et al. (1993) and Braine et al. (2003), the ionized gas spectra along the major axis of Taffy-N (N1–N5) show clear rotation from low to high velocities as one proceeds northward, whereas in Taffy-S (S1–S5) the rotation is also obvious, but in the opposite sense. This confirms the suggestion that the galaxies were counterrotating when they collided. Many of the line profiles are complex and contain multiple components. Of special note are the broad lines in the bridge (especially B1 and B2), as well as complex multicomponent structures in the northwest of both galaxies (N4, N5, S4, and S5). The nucleus of Taffy-S (S2) also shows very broad strong [O iii] emission and weaker broad wings in Hα (especially when a correction is made for Hα absorption; see Section 4.2). We also show polygons marking additional regions of interest on the SDSS image in Figure 3. These polygons show regions that are investigated in the emission-line diagnostic diagrams in Section 5. In the emission-line diagnostic diagrams, we investigate the excitation mechanisms for the western regions of Taffy-N and the bridge separately because these regions exhibit peculiar kinematics distinct from the rest of Taffy-N and the bridge (see the detailed discussion in Sections 4.1 and 4.2).

Figure 4(a) shows total Hα emission contours overlaid on an SDSS i-band image. The two galaxies and the area near the extragalactic H ii region can clearly be distinguished as regions with the brightest Hα emission, with extended emission spread between the galaxies. For Taffy-S, the brightest Hα lies on either side of the nucleus with fainter emission from the direction of the nucleus. It is noticeable that there is no obvious ionized gas associated with the faint southern part of the outer ring in Taffy-S, although there may be some associated with the northern part of the ring between the galaxies. Taffy-N has a very different distribution of ionized gas, with a strong concentration in the inner disk and fainter emission extending along the northwest major axis, where it appears to join with bridge material.

Figure 4(b) shows a map of the equivalent width of the Hβ absorption across the system. In contrast to the lack of emission lines in the southern ring of Taffy-S, strong stellar absorption is seen there, along the extreme northwesterly edge of the galaxy, and in an extended region northeast of the faint stellar ring. In Taffy-N the strongest absorption is seen at the southeastern end of the major axis in the region outside the main body of Hα emission, although it extends at lower equivalent width as far as the center of the galaxy. We will discuss the implications of this Hβ absorption in Section 6.3.

Zoom In Zoom Out Reset image size

We estimate the extinction caused by dust by examining the line ratio of the Balmer lines, Hα and Hβ, referred to as the Balmer decrement, and assume case B recombination. The color excess E(B − V) is given by

$$\begin{eqnarray}&&E(B-V)=1.97\,\ {\mathrm{log}}_{10}\left(\displaystyle \frac{{\left({\rm{H}}\alpha /{\rm{H}}\beta \right)}_{\mathrm{obs}}}{2.86}\right).\end{eqnarray} \tag{ 1 }$$

The extinction at wavelength λ is related to the color excess by

$$\begin{eqnarray}&&{A}_{\lambda }=k(\lambda )E(B-V).\end{eqnarray} \tag{ 2 }$$

We assume a reddening curve k(λ) of the form given by Calzetti et al. (2000). Adopting the same method for all the observed lines allows us to correct, spaxel by spaxel, the observed line fluxes for dust extinction to arrive at intrinsic line fluxes.

A map of the dust extinction at visual wavelengths, A_V, derived from the Balmer decrement is shown in Figure 5, and superimposed on the map are CO (1−0) contours from Gao et al. (2003) using data from the Berkeley-Illinois-Maryland-Association interferometer. It can be seen that the dust-extinction map follows reasonably well the CO (1−0) surface density map, with a peak at the center of Taffy-N (as expected since the H₂ surface density is very high there). A high value of extinction is also seen at the southern tip of Taffy-S. This extends beyond the CO column density contours, but we note that the VLA maps of H i in this region by Condon et al. (1993) show a peak column density there of 1.8 × 10²¹ cm⁻², which would imply an A_V of ∼1 mag integrated over an 18 × 18''² beam (Güver & Özel 2009).

Zoom In Zoom Out Reset image size

Interestingly, we observe a low value of A_V at the center of the Taffy-S and along the northern edge of the spiral arm of that galaxy. The latter result is consistent with a low H₂ column there, although the low A_V in the nucleus may imply different conditions in the gas excitation (deviations from the assumed case B recombination—perhaps due to a low-luminosity active galactic nucleus [LLAGN]) or a significantly reduced dust-to-gas ratio there.

Given that dust opacity measurements depend strongly on galaxy inclination (see, e.g., Driver et al. 2007; Unterborn & Ryden 2008), we caution that our dust-extinction estimates for both galaxies should be treated as lower limits and that actual A_V values could be much higher. For example, Gao et al. (2003) estimate A_V values could be higher than 10 mag for both galaxies. Our low A_V values for the Taffy galaxies could be because measuring the dust extinction from the Hα and Hβ lines involved in the Balmer decrement effectively only probes the effects of dust superficially (essentially a "skin" effect; e.g., Calzetti 2001). It also assumes a simplistic dust geometry—a screen of dust between the observer and the Balmer-line-emitting regions. Such an assumption might not be true for the complicated kinematics within the post-collision Taffy system.

Figure 6 shows the Hα emission channel maps integrated over channels of width 70 km s⁻¹. It is well known that the two galaxies are counterrotating (e.g., Vollmer et al. 2012), and the brightest ionized gas in the two systems reflects this. Ionized gas in Taffy-S is seen in the lowest velocity channel (3806–3876 km s⁻¹) in the northwest on the inside edge of the faint stellar ring and progresses in a southeasterly direction with increasing velocity, eventually showing a major component of emission in the southeast disk that fades away around 4780 km s⁻¹. Taffy-S also has some peculiar kinematics. For example, in the NW part of the disk, faint gas emission is seen over a wide range of velocities along the northern major axis even at the highest velocities. This would not be expected for gas in normal rotation. Taffy-N is even more peculiar. The main centroid of emission from the southeast disk appears at around 4016–4086 km s⁻¹ and progresses steadily toward the northwest, showing the counterrotation. In addition, there is a peculiar region of emission that appears at even lower velocities on the northwest extreme tip of Taffy-N and cannot be part of the normal rotation of the galaxy. Indeed, that structure appears to be part of the bridge, since as velocities increase it becomes more extended and eventually connects to the northwestern region of Taffy-S. In addition to this bridge feature, a second bridge component starts to appear between the galaxies at velocities of 4000 km s⁻¹, and at higher radial velocities it becomes quite strong in the region of the extragalactic H ii region. The emission bridges the two galaxies where it joins with emission that potentially is associated with the faint stellar ring in the northeastern part of Taffy-S. The connection between the galaxies disappears at velocities in excess of 4650 km s⁻¹.

Zoom In Zoom Out Reset image size

The faint stellar ring in the northern half of Taffy-S exhibits some peculiar emission. Features that can be associated with this ring can be seen most clearly appearing from velocities around 4000 km s⁻¹. Moving to higher velocities shows several clumps that appear to follow the ring from NW to SE. These clumps also appear to be surrounded by emission that blends with the emission from the bridge. These features hint that this ring was strongly influenced by the collision and shows a discernible transition from material that is clearly associated with Taffy-S to material clearly associated with the bridge. A similar argument could also apply to Taffy-N, although such a transition is much harder to observe in Taffy-N, which is highly inclined. It is clear that the velocity structure of the gas in both galaxies and the bridge is very complex, and so we will now explore the gas in terms of its spectral profiles—which allows us to more easily separate normal regular rotation in the galaxies from peculiar motions.

As we have shown previously, there are regions in the full data cube where the emission-line spectra show more than one component. Similar, double-line profiles were noticed in both the H i (Condon et al. 1993) spectra and spectra taken in the far-IR [C ii] and [O i] lines with Herschel (Peterson et al. 2018). Some regions of the bridge also show two components in the CO 1−0 molecular gas observations of Gao et al. (2003). Our observations (consistent with those of Gao et al. 2003; Peterson et al. 2018) show that the double-line profiles in the ionized gas are not just confined to the bridge but are also seen projected against parts of the galaxy disks, especially Taffy-N.

To explore the kinematics further, we performed line fitting in two distinct passes. First, we ran LZIFU, forcing it to fit only one component across the whole system. This worked well in regions where the lines were single valued but produced poor results in regions where the lines were double profiled. The output from LZIFU at this stage was a model data cube built from the model fits, as well as the best-fitting continuum cube. Also included were additional data products, including integrated line maps for each of the lines fitted.

Second, we ran LZIFU again, but this time we required it to fit two components for everything. In this case, the fitting worked well for the case of two components, but we found that it performed poorly in places where the profiles were singular. In this mode, since the software was fitting two separate line profiles, the output included two sets of data products (model line cubes, integrated line and velocity field maps), one set for each component—a low- and high-velocity component.

We were able to interrogate the model results to determine, spaxel by spaxel, the mean, standard deviation, and amplitudes for each of the single- and two-component fits. This allowed us to divide the results into three classes of kinematic behavior:

• 1.  
  Spectra consistent with a single Gaussian component, V_s.
• 2.  
  Spectra consistent with two components, V₁ and V₂, at different velocities (V₁ represents the lowest-velocity component, and V₂ the highest).
• 3.  
  Spectra consistent with two components with nearly the same mean velocity, but exhibiting both a narrow V_1n and broad V_2b component.

To qualify as two components, the two Gaussian line centers, V₁ and V₂, were required to differ by at least 35 km s⁻¹, or one-half of the velocity resolution at Hα wavelengths. In practice, the lines were generally farther apart than this and clearly separated. Similarly, for a second component to be considered broad, V_2b must have an FWHM of at least 1.5× that of V_1n. In a few cases, the LZIFU modeling failed to fit two components to cases where a single line profile would have been more appropriate. In this small number of cases, we inspected the profiles by eye to perform what we considered to be the most reasonable classification.

In general when there was doubt about the classification, we inspected the profiles by eye to confirm the classification. Overall the separation between single and two components seemed reasonable, although we realize that there may be some areas where the distinction is somewhat subjective. Less subjective methods of deciding whether to fit single or multiple components to optical IFU data have been explored by Hampton et al. (2017) using artificial neural networks. These methods, which currently rely on training sets using "expert" astronomer guidance, are encouraging for the future, especially since such methods can classify velocity profiles faster than a human, and with statistically similar outcomes. As IFU data become more common, and as the amount of data increases with time, such methods may eventually be needed to replace human classifications. In our paper, in most cases, the classification of two-component versus single Gaussian component was relatively unambiguous.

In order to explore the relationship between the regions of emission where a single component is most appropriate compared with a region with two components, we have created composite moment maps (intensity, mean velocity, and velocity dispersion) by combining those positions consistent with a single profile V_s with those consistent with one or other of the double-profile cases. The reason we combined the single-component data with the two velocity components separately was to look for continuity between the single-component gas and one or other of the double lines. For example, if the single-component data mapped smoothly into velocity field of one of the two double components, this might suggest that they are really part of one single dynamical system, whereas a sudden discontinuity would suggest no such regularity.

These "merged" single- and double-profile moment maps are shown in Figure 7. Figures 7(a)–(c) represent moment maps created by combining spaxels containing components V_s with V₁ and V_1n, whereas Figures 7(d)–(f) represent the combination of spaxels containing V_s with V₂ and V_2b. To make it clear where the different kinds of profiles fall in the maps, we indicated in the figure those regions enclosed within the red polygon that are consistent with Class-2 above (two components at different velocities). Those regions of the maps consistent with a single line, Class-1, are colored with a red background color. Finally, those regions where a narrow component and a broad component were present, Class-3, are shown with a green background color.

Zoom In Zoom Out Reset image size

Given that the kinematics are quite complex, we start by identifying those regions that may show regular galactic rotation. The simplest kinematics to understand are those of Taffy-S in the low-velocity component of Figure 7(b). Here the increasing isovelocity contours, going from yellow to dark blue, progress regularly in the double-line region, merging smoothly with the V_1n (green underlying color) and single-component V_s (red underlying color) contours. The velocity dispersion in the low-velocity component in Taffy-S is also low across most of its disk. Concentrating only on the low-velocity component for Taffy-S, it is clear that the velocities and dispersions shown in Figures 7(b) and (c) look like a somewhat warped but regular rotating disk. In contrast, Taffy-S is much more peculiar in the high-velocity component of Figures 7(e) and (f), where the galaxy shows only a small amount of obvious rotation, as well as exhibiting a high velocity dispersion in a large part of the disk. It also has a band of spectra classified as broad-line (Class-3 type; green underlying color) in the nuclear regions.

We now turn our attention to Taffy-N, which has more complex kinematics. This galaxy is quite edge-on and may be expected to show regular rotation along its major axis. Evidence of rotation is seen in the high-velocity component of Taffy-N in the southern part of the disk centered on the dense dust lane and nucleus. Figure 7(e) shows a clear rotation signature where the velocities (starting with the pale-blue contours in the southeast) increase along the major axis (dark-blue contours in the northwest of the inner disk). Although this apparent regular rotation in the high-velocity component is confined to the inner parts of the Taffy-N disk, the increasing trend in velocity shows a reversal toward the northwestern extended disk. This might be interpreted as a turnover in the rotation curve there.

Next, we consider those parts of the velocity field that cannot be considered normal and are most likely caused by the strongly collisional nature of the Taffy pair. We have already pointed out in the discussion of the channel maps that the northwestern part of Taffy-N has a peculiar low-velocity structure that is not part of the normal rotation. This can be seen in Figure 7(b) (low-velocity component), where much of the NW disk of Taffy-N shows very little rotation and also shows high velocity dispersion (Figure 7(c)). Here we see that the gas extends as a finger toward the southwest, where it forms a western bridge with Taffy-S. A second eastern bridge structure is seen associated with the extragalactic H ii region, which is strongest in the high-velocity component. The two structures are graphically emphasized in Figure 8. Much of the gas between the two galaxies is seen in the eastern high-velocity bridge component and shows a velocity gradient that extends between the two galaxies along the direction of the radio continuum bridge discovered by Condon et al. (1993). Some regions of the high-velocity component bridge material have a high velocity dispersion—as was noted in the spectra in Figure 3. We will show that much of the bridge gas in the high-velocity component has the excitation properties consistent with shocked gas.

Zoom In Zoom Out Reset image size

Several regions also show Class-3 spectra—which means that one component is broad. These regions (colored green in all the panels of Figure 7) are confined to positions along the minor axis of Taffy-S and to a small region at the northwest tip of the same galaxy. We show an example of these kind of spectra in Figure 9. Here we show how LZIFU has fit two components to the Hα profile from the nucleus of Taffy-S after correcting for weak Balmer absorption. This appears as a region of high velocity dispersion in Figures 7(c) and (f). The Hα line profile shown is an average over a 4 × 4 spaxel region centered on the X-ray hot spot in Taffy-S. The two components have an FWHM of 320 ± 70 km s⁻¹ and 205 ± 70 km s⁻¹, with a small offset in velocity between the two.

Zoom In Zoom Out Reset image size

We first explore the possibility that the gas is excited by shocks. We overplot predicted line ratios from shocks on Figures 10–12, taken from the MAPPINGS III library of models (Allen et al. 2008). The model line ratios are plotted for different shock velocities as solid colored lines. Based on the models of Vollmer et al. (2012), we assume that much of the gas in the bridge and throughout the galaxies is close to solar metallicity, since the gas has been stripped from the galaxies or has been excited in situ. For the shock models we therefore assume solar metallicity. The other parameters of the models include the pre-shock gas densities in the range 0.1 < n [cm⁻³] < 1000 (stepping by a factor of 10 each time) and an assumed constant magnetic field of B = 5 μG (this is close to the equipartition magnetic field strength of 8 μG measured by Condon et al. 1993 through radio continuum measurements). We have plotted only the line ratios for the shock itself while excluding the precursor component of the shock. For the moderate shock velocities that seem compatible with the Taffy excitation velocities, it may be reasonable to ignore the effect of a strong ionizing shock precursor, as we shall discuss later.

For the east-bridge points (red crosses), in the high-velocity component ("b" panels), it is clear that the east-bridge points fall relatively neatly between the solid lines for shock velocities 175 and 200 km s⁻¹ (green and orange lines). For the west-bridge points, in cases where the points deviate from the H ii region area they are consistent with the same shock velocity, e.g., the low-velocity components in all diagrams. Thus, the high-velocity east-bridge component and the low-velocity west-bridge component seem consistent with shock excitation for all the line diagnostic diagrams.

The situation is mixed for the disks of galaxies themselves. As Figures 10(b), 11(b), and 12(b) show, there are some points within the disks of both galaxies that fall in the composite region of the diagnostic diagram. Some of these points would be consistent with a mixture of H ii region and shocked gas excitation. The nucleus of Taffy-S is an ambiguous case, because it could be excited by shocks in a mild outflow, or may be gas excited by UV emission from a weak LLAGN.

The spreading of the points along lines of constant [O iii] λ5007/Hβ ratio in the high-velocity component in the bridge has a number of possible interpretations if we assume that shocks are involved. First, the spread might imply that the shocks are occurring in an ensemble of gas clouds with different pre-shock densities. Such a picture is consistent with our previous observations of the Taffy bridge (Peterson et al. 2012, 2018), where we have observed gas in many different excited phases, from H i (Condon et al. 1993) to warm molecular gas from the Spitzer IRS, along with the detection of [CI] and [C ii] emission (Peterson et al. 2018) and boosted values of [C ii]/far-IR and [C ii]/PAH ratios. The existence of a highly multiphase (and multidensity) medium is also very consistent with the detection of soft X-ray emission from the bridge. Thus, it might be expected that shocks moving through such a multiphase gas would encounter a range of pre-shock densities—which would spread the points along lines of constant shock velocity—as observed in Figures 10 and 11 especially.

An alternative explanation might be that some of the gas is of lower metallicity. As the models of Allen et al. (2008) show, reducing the metallicity of the shocked gas moves the points in the diagnostic diagrams to the left at roughly constant values of [O iii] λ5007/Hβ ratio. However, if the collisional models of Vollmer et al. (2012) are correct, the gas in the bridge should have come from many different places within the original pre-collisional disks, and deviations of factors of 100 in metallicity in the bridge seem unlikely. We conclude that it is much more likely that we are observing shocks within the bridge and parts of the galaxy disks that encounter clumps of material at different densities.

What is the effect of ignoring the possible influence of a hot shock precursor in the models? This is an effect where, in high-velocity shocks, the gas in the shock is so strongly heated that UV radiation from the shocked gas ionizes large amounts of pre-shocked gas upstream of the shock. We show in Figure 13 an example of the [N ii] line diagnostic diagram, the effect of including the shock and the shock precursor (Allen et al. 2008). As can be seen by comparison with Figure 10, the behavior when we include the shock precursor with velocities <300 km s⁻¹ is very similar to the case with no shock precursor, which fits our data well. At shock velocities >300 km s⁻¹ the models including the shock precursor diverge significantly from these data. Similar behavior is noted in the other diagnostic diagrams (not shown). This implies that the shock models between 100 and 300 km s⁻¹ fit the bridge data well regardless of whether the precursor is included. We note, however, that in dense gas (e.g., molecular gas known to also be present in the bridge) the velocity at which a shock precursor may become important will be much lower. Therefore, future modeling of the molecular shocks may have to take precursor activity into account.

Zoom In Zoom Out Reset image size

In a recent paper by Weilbacher et al. (2018), it was noted that the Antennae galaxies (NGC 4038/39), like the Taffy system, also exhibit significant diffuse ionized gas emission. Although several mechanisms were put forward to explain the emission, including the possibility of shocks, the authors favor an interpretation that much of the diffuse gas is ionized by Lyman continuum photons (hereafter Ly-C) "leaking" from large numbers of H ii regions found primarily in the disks of both galaxies, as well as H ii regions found within the "overlap region." Much (but not all) of the diffuse component was found close to massive star formation complexes in the system. Using multicolor HST imaging of the clusters, the authors were able to compare the luminosity of the Hα emission associated closely with a cluster with theoretical models of the Ly-C flux from the clusters to determine whether the clusters were "leaking" Ly-C photons into the surrounding gas. It was found that many of the H ii regions had nonzero escape fractions of Ly-C UV radiation, especially in the center of NGC 4038 and also in the "overlap" region. Thus, for the Antennae system, it was found that the excess diffuse emission within the system could be explained as gas excited by UV radiation escaping from the clusters.

Could some or all of the extended ionized gas emission we see in the Taffy come from similarly degraded UV light from H ii regions in the disks of the Taffy galaxies that might diffuse outward and ionize large parts of the Taffy bridge? We estimate (Table 1) that the amount of Hα emission coming from the bridge, after correcting for extinction, is similar to that from Taffy-S and roughly 50% of the emission from Taffy-N. In this case the majority of the escaping photons would have to come from the galaxies themselves, since the SFR in the Taffy bridge is very low. Unfortunately, unlike the case of the Antennae, we do not have multicolor high-resolution images of the individual star clusters, and this means that it is difficult to perform the same kind of test that was applied, for all individual H ii regions, by Weilbacher et al. (2018). As a result, we cannot completely rule out a significant contribution to the ionized medium in the Taffy coming from leaky H ii regions. Nevertheless, many other lines of evidence already suggest that shocks must be present in the Taffy bridge, and so we prefer the shock explanation for the excitation of the high-velocity component, rather than leaky H ii regions. Future observations will be needed to attempt to model the history of star formation in the clusters in the galaxies, which will allow us to estimate the fraction of UV emission that may escape into the surrounding gas. This is beyond the scope of the current paper.

Using the following relation from Kennicutt (1998), we estimate the SFR in the entire Taffy system and the bridge region (using the entire region defined as the bridge in Figure 3), including Hα emission from the extragalactic H ii region:

$$\begin{eqnarray}&&\psi [{M}_{\odot }\,{\mathrm{yr}}^{-1}]=7.9\times {10}^{-42}\,L({\rm{H}}\alpha )[\mathrm{erg}\,{{\rm{s}}}^{-1}].\end{eqnarray} \tag{ 3 }$$

We obtain L(Hα)_SF = (4.99 ± 0.54) × 10⁴¹ erg s⁻¹ and L(Hα)_SF;bridge = (1.02 ± 0.14) × 10⁴¹ erg s⁻¹ for the extinction-corrected Hα luminosity coming from star formation for the entire Taffy system and bridge, respectively, using the method described previously. This translates to SFRs of 3.94 ± 1.0 M_⊙ yr⁻¹ and 0.81 ± 0.22 M_⊙ yr⁻¹, respectively, for the entire Taffy system and the bridge. These SFRs agree well with our previous estimates derived from UV–far-IR SED fitting (Appleton et al. 2015) of 3.65  ±  0.03 M_⊙ yr⁻¹ and 0.69 ± 0.06 M_⊙ yr⁻¹, respectively, for the total system and the bridge.

Interestingly, recent observations with the Atacama Large Millimeter Array (ALMA) show dense filaments of molecular gas, in the Taffy bridge, with little star formation in them. These ALMA observations and the overall star formation properties will be discussed in a future paper (P. N. Appleton et al. 2019, in preparation).

The mass of ionized gas in the Taffy system and in the bridge assuming case B recombination (Macchetto et al. 1996; Kulkarni et al. 2014) is given by

$$\begin{eqnarray*}&&{M}_{\mathrm{ion}}=2.33\times {10}^{3}({L}_{{\rm{H}}\alpha }/{10}^{39})({10}^{3}/{n}_{e})\,{M}_{\odot },\end{eqnarray*}$$

where L_Hα is the extinction-corrected Hα luminosity in units of erg s⁻¹ and n_e is the electron density. For the bridge, from the lower limits to the ionized gas fractions from star formation (see above) we obtained L(Hα)_SF;bridge = (1.02 ± 0.14) × 10⁴¹ erg s⁻¹. We also have n_e = 200 cm⁻³, based on the ratio of the [S ii] lines. This then gives us M_ion = (1.19 ± 0.22) × 10⁶ M_⊙. This calculation is uncertain because the Hα emission originates from two different processes—H ii regions and very likely shocks. However, it does show that the ionized gas mass is an insignificant fraction (∼0.2%) of the total mass of gas in the bridge (∼7 × 10⁹ M_⊙, made up of a mix of H i and H₂). This is in agreement with the very low ionized gas fraction responsible for exciting the [C ii] 157.7 μm far-IR cooling line in the bridge (Peterson et al. 2018) determined from the upper limit to the detection of [N ii] 206 μm in the bridge. For the Taffy system as a whole, using the extinction-corrected Hα luminosity coming from star formation, L(Hα)_SF = (4.99 ± 0.54) × 10⁴¹ erg s⁻¹, we get M_ion = (5.8 ± 1.0) ×10⁶ M_⊙ for the mass of ionized gas in the entire Taffy system. Again, this is an insignificant fraction (∼0.8%) of the total gas mass in the Taffy system.

We detect Hβ absorption lines within many spaxels on the galaxies. The spectra of post-starburst galaxies are known to contain strong Balmer absorption lines owing to their stellar populations being dominated by A-type stars. Evidence of a post-starburst population is not uncommon in merging galaxies (e.g., Zabludoff et al. 1996; Yang et al. 2004, 2008), but attempts to measure the age of the stellar population are difficult, especially when only Hβ is observed (see, e.g., Worthey & Ottaviani 1997). Since our spectral coverage did not include other post-starburst indices, we can only provide preliminary results here. Further observations using full UV–optical SEDs, better absorption-line indices, and detailed modeling (e.g., French et al. 2016) will be needed to obtain a better estimate for the age of the population that is responsible for the Hβ absorption.

We measured the EW of the Hβ absorption line for each spaxel that contained either the galaxies or the bridge. The EW was measured using

$$\begin{eqnarray}&&W({\rm{H}}\beta )\,[\mathring{\rm A} ]=\displaystyle \frac{{\int }_{\mathrm{line}}{f}_{\lambda }d\lambda }{\left\langle {f}_{\lambda ;\mathrm{cont}}\right\rangle },\end{eqnarray} \tag{ 4 }$$

where the integral is done over the continuum-subtracted absorption-line fit and $\left\langle {f}_{\mathrm{cont}}\right\rangle$ is the average continuum value measured on either side of the Hβ line. LZIFU provides as output the fit to the stellar continuum and the nebular emission lines separately. We use the continuum fit cube to refit a Gaussian absorption line to the region centered on the Hβ absorption. The parameters from this fit then give the area within the absorption line, and the average continuum is measured in a band of ∼10 spectral elements on both sides of the line.

Figure 4(b) shows the measured EW map for the Taffy system. Relatively deep Hβ absorption (W(Hβ) > 10) occurs where there is little Hα emission. Two regions of high W(Hβ) lie in the northern and southern parts of the faint stellar ring that surrounds Taffy-S. Another region, in the southeastern part of Taffy-N, does extend into regions where there is some star formation and older stellar populations are probably present, with the deepest absorption lying outside of the main star formation disk.

It is generally true that values of W(Hβ) in the range of 10–20 W(Hβ) imply stellar evolutionary ages for the post-starburst populations of several hundred megayears, and this would imply that there is no connection between the post-starburst population and the current collision between the two galaxies (e.g., Worthey & Ottaviani 1997). From dynamical arguments it has been argued that the collision between the two Taffy galaxies is quite recent (approximately 25–30 Myr; Vollmer et al. 2012), and so the fact that the outer ring of Taffy-S shows an old population would lead to an apparent problem, if the stellar population of the ring were created in the collision. However, one solution might be that the stellar population of the ring is a stellar density wave containing a much older pre-collisional system that had undergone star formation a long time in the past. This is also the conclusion reached by Jarrett et al. (1999), who argue that the ring consists of stars from an old disk population from the pre-collision disk.

The fact that both galaxies contain evidence of post-starburst activity might imply that the galaxies underwent a high-speed encounter in the more distant past that triggered star formation but did not lead to immediate merger. If that is the case, then we are probably currently witnessing the second (probably final) collision before full merger. Broader wavelength coverage in the blue, to detect more absorption lines and better characterize the age of the post-starburst population, will be needed to test this idea.