OPTICAL IFU OBSERVATIONS OF THE BRIGHTEST CLUSTER GALAXY NGC 4696: THE CASE FOR A MINOR MERGER AND SHOCK-EXCITED FILAMENTS

The data were reduced using the WiFeS IRAF data reduction package (Dopita et al. 2010), which is based on and utilizes the Gemini IRAF package.² The procedure began by treating each frame as a two-dimensional image. The detector bias signals were removed in the standard way, using the overscan regions of the raw frames. Then the sky frame was scaled and subtracted from the bias-subtracted object frame.

The spectrograph optically divides the field of view into slices along the long axis of the field. These "slitlets" are stacked end-to-end and dispersed across the detector to form a series of two-dimensional slit spectra from adjacent slices on the sky. In processing, the two-dimensional spectrum from each slice was extracted from the bias- and sky-subtracted frame and placed in an individual FITS image extension. Standard methods for processing two-dimensional spectra were then applied individually within each slice as described below.

The flat-field frame is formed from exposures of a quartz–iodine lamp. Within each slice, the spectra were normalized, using a spline function fit to the spectral profiles. The shape of the lamp illumination in the spatial direction along the slices was corrected using twilight sky frames. An additional adjustment was made to maintain the relative signal levels between the slices. The final flat-field frame then retained only pixel-to-pixel variations in the spectral direction and the slit transmission profile in the spatial direction. These effects were subsequently removed from the science data by flat-fielding. Each slice of the galaxy frames was corrected with the corresponding two-dimensional slice from the flat-field frame.

The individual two-dimensional spectra were then transformed to a common spatial and spectral coordinate grid. Ne–Ar arc lamp frames were used to calibrate the spectral scales of the slices. The IRAF identify task was used to match the emission features in the arc lamp spectra, and fitcoords was used to apply the spectral coordinate solution to the science data, slice by slice. The average rms error in the fit to the arc-line positions is 0.2 Å for the red cube and 0.1 Å for the blue cube. For the spatial alignment, an occulting wire is available at an intermediate focus within the instrument. This provides a convenient spatial registration when the aperture is illuminated by the flat-field lamp. An image of the wire is seen near the center of each slitlet and this establishes the origin of the spatial coordinates in each slice. We assume a standard spatial scale of 05 pixel⁻¹ along the slices. The data were resampled onto a common rectilinear grid by the transform task and the two-dimensional slices were then stacked together to form a three-dimensional data cube. The ends of the spectra were trimmed from the reduced data cubes to remove the regions of low sensitivity in the grating+dichroic transmission profile. Approximately 1050 Å and 550 Å were removed from the short wavelength end of the red and blue cubes, respectively, and approximately 420 Å and 160 Å from the long wavelength end of the cubes. This still provided ∼500 Å of overlap between the cubes. The trimmed regions do not extend to within 400 Å of the emission lines measured.

Telluric absorption features were corrected in the red galaxy spectra using an integrated spectrum from a reduced smooth-spectrum standard star data cube. We normalized the spectrum to remove the continuum and obtain a spectrum containing only atmospheric absorption features. This was compared to the absorption features seen in the galaxy spectra, scaled and shifted as required, and used to correct the galaxy data using the IRAF telluric task. After removing the telluric features from a standard star data cube, an integrated flux-standard spectrum was extracted and the standard IRAF flux-calibration tasks used to calibrate the spectra in the galaxy data cubes. The spectra across the whole field are corrected for the shape of the spectral transmission profile as measured at the position of the standard star (approximately the center of the field). The flux calibration was performed without correcting for atmospheric extinction in the individual data frames. The variation in continuum flux among the frames gives an indication of the uncertainties in the flux calibration: there is a 1σ variation of ∼11%, attributable to the changing air mass (the galaxy altitude was between 1 and 1.4 airmasses during the observations) and fluctuations in sky transparency.

Each science frame was individually reduced by using the procedures described above. The dithered galaxy data cubes were then spatially registered using the continuum centroid of the galaxy nucleus. This was measured for both the red and blue cubes, using an image formed by summing several spectral planes in a line-free region near the center of the spectrum. The cubes were then offset as required and median-combined with the IRAF task imcombine. Cosmic rays and detector defects were removed in this process.

Measuring the flux of the extended line-emission regions surrounding NGC 4696 is complicated by the strong stellar continuum. This dominates the spectrum in most regions and must be removed to extract accurate emission-line fluxes. This can be done using the galaxy itself, rather than attempting to generate a theoretical stellar continuum template.

Since the distribution of line emission around NGC 4696 is asymmetric (see Section 3.3) and the filaments do not extend beyond approximately 5''–10'' from the galaxy center to the north and east of the nucleus, we were able to extract a spectrum from a region of the inner galaxy that is free from line emission and use this to subtract the stellar component from the spectra at other locations. The representative galaxy spectrum is constructed from a 3'' diameter aperture placed ∼8'' to the northeast of the continuum center, where no line emission is detected in either the red or blue data cubes, as close to the galaxy core as possible. For each of the strong spectral emission features of interest, this spectrum was scaled to match the flux level of the continuum in neighboring spectral regions (using the average level in two bands approximately 35 Å wide bracketing the line or complex of lines and centered within 50 Å of the feature), before it was subtracted at each spatial pixel. An estimate of the error introduced in this process is made in the following section.

After subtracting the galaxy stellar continuum, we extracted the flux of the emission lines that are detected at better than the 3σ level. This corresponds to a line surface brightness ≳4 × 10⁻¹⁷ erg s⁻¹ cm⁻² arcsec⁻²  in the red data cube and ≳6 × 10⁻¹⁷ erg s⁻¹ cm⁻² arcsec⁻²  in the blue for a line width of 300 km s⁻¹. The following lines are detected and fit: [O ii] λλ3726, 29 (spectrally unresolved), [O iii] λλ4959, 5007, [N i] λλ5198, 200 (spectrally unresolved), [O i] 6300Å, [N ii] λλ6548, 83, [S ii] λλ6716, 31 as well as Balmer hydrogen recombination lines Hα, Hβ, and Hγ.

We fit single Gaussian line profiles to the emission lines of interest in the continuum-subtracted spectra using the IRAF task ngaussfit and measured line fluxes, the centroid positions, and widths at each spatial pixel having sufficient signal to allow this to be done. Hα and the lines of the [N ii] λλ6548, 83 doublet were fit together as three Gaussian components, and the lines of each of the [S ii] λλ6716, 31 and [O i] λλ6300, 63 doublets were also fit simultaneously. All the other detected lines were fit individually.

Radial velocities are given relative to the velocity of the galaxy with respect to the local standard of rest. The velocities have been corrected for the motion of the earth with respect to the local standard of rest; the magnitude of the correction required was less than 9.5 km s⁻¹ for all observations. The instrumental resolution has been removed from the fitted line widths by subtracting in quadrature the average value of 97 km s⁻¹ FWHM that was measured from an arc lamp calibration image. The results derived from the data cube and the emission-line fits are discussed in the following sections.

Uncertainties in the relative line flux measurements were estimated using the noise properties of the spectra after continuum subtraction. For a given emission line, the 1σ error in the flux density per pixel was taken to be the standard deviation in a line-free region adjacent to the feature. These were combined in quadrature to form the uncertainties in the integrated line flux.

Figure 1 shows the integrated spectrum from a 5'' diameter circular aperture centered on the core of NGC 4696, prior to subtracting the galaxy continuum. Forbidden lines of N ii, O ii, and S ii, and the Hα recombination line are prominent, but the spectrum is dominated by a stellar continuum component that is typical of an early-type galaxy. The strength of the continuum increases toward longer wavelengths and exhibits absorption features characteristic of evolved stellar populations: e.g., Ca K and H lines (at 3934 Å and 3969 Å, respectively), G band (CH⁺; 4300 Å), blended Mg i b (5175 Å), Na i D blended doublet (5893 Å), and the broad TiO feature (7126 Å). In addition, the step in the average continuum intensity across the 4000 Åbreak (D₄₀₀₀ = 1.9) is typical of values in elliptical galaxies, including those of the low line-luminosity BCGs in the ROSAT "Brightest Cluster Sample" of Crawford et al. (1999).

Zoom In Zoom Out Reset image size

The galaxy spectrum shows no evidence for a blue stellar component that would indicate that significant star formation is occurring. This is in contrast to the blue and UV continuum excesses and large star formation rates estimated in high-luminosity BCGs: for example, those in clusters Zw 3146 (e.g., Hicks & Mushotzky 2005), A1835 (McNamara et al. 2006), and A1068 (Wise et al. 2004). It is consistent, though, with the general properties of the BGC population, which show reduced signs of star formation in less luminous BCGs (e.g., Allen et al. 1992; Crawford et al. 1999; Egami et al. 2006b), such as NGC 4696 itself.

We compare the radial surface brightness profile of the galaxy continuum with a Hernquist density model of the form

$$\begin{equation} \rho (r) = \frac{M_{t}}{2\pi }\frac{a}{r}\frac{1}{(a+r)^{3}}, \end{equation} \tag{ 1 }$$

following Lim et al. (2008), where M_t is the total galaxy mass and a is a characteristic scale radius. We assume a constant I-band mass-to-light ratio of M/L = 4.7, taken from the correlation between stellar velocity dispersion and I-band M/L presented by Cappellari et al. (2006).

A simple minimum least-squares criterion is used to compare the model with the measured galaxy surface brightness at 8000 Å, averaged azimuthally between position angles 35° and 125° (i.e., in the quadrant of the WiFeS field where the dust absorption features are least prominent) and at radii from 2'' to 175. The best fit is obtained with a total mass M_t = (4.0 ± 0.5) × 10¹¹ M_☉ and scale radius a = 3.5 kpc. The measured and model radial surface brightness profiles are shown in Figure 2.

Zoom In Zoom Out Reset image size

Laine et al. (2003) fit the I-band surface brightness profile with a "Nuker profile" (a broken power law that has a turnover radius of 14) in the central 10'' to fit the "core" seen in the profile data. We find that the Hernquist profile provides a better fit to the profile shape for our data beyond r ∼ 10''.

Our estimated total stellar mass is different to the galaxy mass of (11.6 ±  0.1) × 10¹¹ M_☉ presented by Rafferty et al. (2006) from R-band imaging. The procedure used by the authors to determine this mass is not described in sufficient detail to understand the nature of this discrepancy.

The stellar continuum is relatively redder closer to the galaxy core, as also observed by de Jong et al. (1990). This may be either the effect of dust extinction increasing toward the galactic center or else the result of a greater mean metallicity in the stars.

Figure 3 presents emission-line spectra in the region of the filaments where the line emission is strongest, before and after continuum subtraction. The spectra are integrated over a 3'' diameter circular aperture centered approximately 3'' northwest of the nucleus (ΔX = 0.2, ΔY = 30 in the reference frame of the images in Figure 4). The reference galaxy spectrum is shown, scaled to match the data in each panel, together with the total spectrum and the continuum-subtracted result, illustrating the strongest emission-line detections. These spectra have not been corrected for reddening.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 1 lists integrated, continuum-subtracted, and extinction-corrected line flux measurements in the same 3'' aperture, obtained from the Gaussian line profile fits. This table shows the emission features detected above the 3σ sensitivity level in the aperture. We correct the line flux measurements for extinction using the reddening law of Osterbrock (1989): the relative extinction at the wavelength of Hα is A_Hα/A_V = 0.744 and at the wavelength of Hβ is A_Hβ/A_V = 1.14. In addition, we adopt a value of R = A_V/E(B − V) = 3.1. To determine the total extinction, we assume that the intrinsic emitted flux ratio F(Hα)/F(Hβ) = 3.1. This is somewhat higher than the classical Case B recombination value, but is appropriate for active galaxies, including LINERs (Veilleux & Osterbrock 1987), in which there is a collisional excitation component in the Hα flux. From the dust maps of Schlegel et al. (1998), the reddening contribution from our Galaxy in the direction of NGC 4696 is A_V ∼ 0.34.

The resulting extinction measured in the 3'' diameter aperture at the line peak is A_V = 0.74 ± 0.07 mag. The error is obtained from the uncertainty in the measured Hα/Hβ flux ratio integrated over the aperture. The measurements represented in Table 1 have been corrected for extinction of this magnitude. The measured Hα line flux in the aperture is (4.0 ± 0.4) × 10⁻¹⁵ erg s⁻¹ cm⁻², and the extinction-corrected result is (6.7 ± 0.9) × 10⁻¹⁵ erg s⁻¹ cm⁻². The error here is determined from the uncertainty in the flux calibration and the extinction. Line fluxes for the other detected emission lines are presented as ratios to the Hα flux in Table 1. Estimated uncertainties in the flux ratios, derived as described in Section 2.3, are also shown in the table.

The top-right panel of Figure 3 indicates possible detections of the [Ne iii] λ3869, Hζ λ3889, and blended [Ne iii] λ3967  + H λ3970  lines, which are close to or below the 3σ detection threshold and could not be fit in individual pixels.

Figure 4 presents maps of the emission-line properties derived from the Gaussian profile fits for three strong emission lines in the spectra: the line flux in the top panel (no reddening corrections have been applied), the velocity of the line centroid in the middle panel, and the derived velocity dispersion in the bottom panel. The cross marks the position of the galaxy nucleus, at the peak of the continuum emission surface brightness distribution.

The distribution of the detected line emission from the filaments of NGC 4696 is illustrated in Figure 5, which shows a projection along the wavelength axis of the data cube over the [N ii] λλ6548, 6583 and Hα lines. The structures detected here can be matched to the features seen in narrowband imaging by Crawford et al. (2005). The detected filaments extend to the east, south, and west of the galaxy center. Some filaments have a generally radial orientation with respect to the galaxy, but there are curved arcs, bars, and filaments in the structures that are extended at orientations that are much more nearly tangential.

Zoom In Zoom Out Reset image size

The most prominent feature is the bright filament that spirals from the nucleus to the northwest, and in the outer regions curves around the galaxy core to the east. This is the structure that was identified in early imaging of the gas and dust by Fabian et al. (1982), Jorgensen et al. (1983), Nørgaard-Nielsen & Jørgensen (1984), and Sparks et al. (1989). The surface brightness of the filament decreases with distance from the galaxy nucleus along its length (other than at a brighter knot that appears at the intersection with another filament). The outer length of this main filament, to the south of the galaxy core, coincides closely with the prominent dust lane that is seen clearly in the HST (Laine et al. 2003) and B − I color (Crawford et al. 2005) images of the galaxy. In the HST images, this lane of dust appears to be part of a continuous structure that spirals around the galaxy center at small radii to the northeast and with increasing radius to the west and south, passing through the brightest regions of the line emission. Lanes of absorption also coincide with outer filaments to the south and west of the main filament. The association of the line emission and dust absorption features indicates that the observed system of gas and dust is inclined toward us, is on the near side of the galaxy, and has an intrinsically one-sided spiral distribution about the galaxy.

The peak in surface brightness of the extended emission occurs approximately 3'' (0.6 kpc) to the northwest of the galaxy center. This peak occurs inside the curve of the dust lane. The spatial association of the main filament with the dust lane is very clear in the outer extent of the filaments. Close to the nucleus, however, the main dust spiral departs from the ridge line of the line emission along the filament. As seen in Figure 6, the filament of optical emission bends toward the nucleus from the northwest, whereas the dust spiral continues around to the north of the core and finally turns toward the nucleus from the northeast. If the ridge of emission tracks a shock front, as proposed below, this might indicate that the shock front is progressively propagating through the dust lane as the dusty gas spirals in toward the nucleus.

Zoom In Zoom Out Reset image size

The morphology of the emission from weaker lines such as [O i] and [N i] is consistent with that of the strong emission lines (hence are not pictured in Figure 4). To the degree that our spatial resolution reveals, it appears that all the optical forbidden and hydrogen recombination emission lines originate in the same gas.

The integral-field data permit us to determine the total emission-line luminosity of the filaments, which is useful to constrain both the mass and energy budgets. The flux measurements from the emission-line fits are summed over all the pixels where the value is greater than the 3σ detection limit to obtain a total line flux.

The total flux detected in Hα is (5.7 ± 0.6) × 10⁻¹⁴ erg s⁻¹ cm⁻²  prior to correcting for extinction. The reddening in the observations is estimated from measured Hα/Hβ lineflux ratios as described in Section 3.2. Since Hβ is relatively faint, we do not detect it over the same spatial extent as we measure the total Hα flux. This restricts our ability to map the spatial variation of the extinction. In spatial pixels where both lines are detected above the 5σ detection limit, the extinction correction is determined from the Hα/Hβ line ratio measured in that pixel. Everywhere else, we assume an average extinction level of A_V = 0.4 mag. This value is determined from the integrated line fluxes over the region where both Hα and Hβ are detected. The line luminosity is then derived by summing the individually extinction-corrected fluxes from each pixel.

This results in an extinction-corrected, integrated Hα flux of (7.9 ± 0.9) × 10⁻¹⁴ erg s⁻¹ cm⁻². At a distance of 44 Mpc, the Hα line luminosity is then (1.8 ± 0.2) × 10⁴⁰ erg s⁻¹. This is consistent with the lower-limit estimate of Crawford et al. (2005) from narrowband imaging (saturation in their images allowed only a lower limit to be obtained): L(Hα)>1.5 × 10⁴⁰ erg s⁻¹. The total [N ii] λ6583 luminosity from our data is (3.7 ±  0.4) × 10⁴⁰ erg s⁻¹, similar to the value of ∼ 5 × 10⁴⁰ erg s⁻¹ measured by Fabian et al. (1982) for the [N ii] λ6583  flux. In earlier narrowband work, Sparks et al. (1989) estimated an approximate combined Hα + [N ii] λ6548+6583 luminosity of 9 × 10⁴⁰ erg s⁻¹ and given the inherent difficulties in calibrating such measurements and accurately extracting only the line emission, this is in reasonable agreement with our result.

An estimate of the mass of emitting ionized hydrogen can be made using the Hα emission-line luminosity, L(Hα) (Osterbrock 1989):

$$\begin{equation} M_{\rm H^{+}} = \frac{L({\rm H}\alpha)\,m_{p}}{n_{e}\,\alpha _{\rm {H}\alpha }^{\rm eff}\,h\,\nu _{{\rm H}\alpha }}, \end{equation} \tag{ 2 }$$

where we adopt the effective recombination coefficient for Hα line emission α^eff_Hα = 1.17 × 10⁻¹³ cm³ s⁻¹ from Osterbrock (1989) for Case B emission at T = 10⁴ K. Then,

$$\begin{equation} M_{\rm H^{+}} \approx 4.3\times 10^{5} \left(\frac{100\,{\rm cm}^{-3}}{n_{e}}\right)\,\hbox{$M_{\odot }$}. \end{equation} \tag{ 3 }$$

Thus, assuming an average electron density of ∼100 cm⁻³, we find a total mass of ionized hydrogen in the filaments of $M_{{\rm H}^{+}} \sim (4.3\pm 0.5)\times 10^{5}$ M_☉.

The mass and density can be used to define a volume filling factor ε for the ionized gas in the filaments. The volume filling factor for material in a structure is defined as the fraction of the encompassing volume that is filled by material at the estimated density. An assumption about the geometry of the emitting region is required. We can calculate a filling factor for the ionized material within a spherical volume that encompasses the filaments:

$$\begin{equation} \varepsilon = \frac{M_{{\rm H}^{+}} / \rho }{\mbox{$\frac{4}{3}$}\,\pi R^{3}}, \end{equation} \tag{ 4 }$$

where $M_{{\rm H}^{+}}$ is the mass of ionized gas derived from the Hα flux, ρ is the mass density of the emitting gas, and R is the radius of a sphere encompassing the emitting region. Assuming an electron density n_e ∼ 100 cm⁻³, R ∼ 2.9 kpc, and the ionized gas mass determined above, the volume filling factor is ε ≈ 10⁻⁶. This is comparable to values found for other filament systems (e.g., Ogrean et al. 2010; Hatch et al. 2007) and is consistent with the Hα-emitting gas being only a small fraction of the emitting material, located in a thin layer on the surface or within the volume of a larger mass of gas.

This larger volume is unlikely to be spherical, so a more appropriate estimate might be obtained within a more constrained region of the filaments. We assume the emitting material from the 3'' diameter aperture used in Section 3.2 and Table 1, at the line peak, is distributed over a spherical volume of radius R = 15 ∼ 9.7 × 10²⁰ cm. From Equation (2), the estimated emitting mass of ionized gas in this aperture is $M_{{\rm H}^{+}} = (3.6\pm 0.5)\times 10^{4}$ and the volume filling factor is ε ≈ 10⁻⁴, with n_e ∼ 100 cm⁻³. This is still small, indicating a clumpy or filamentary distribution, but may more accurately reflect the volume fraction of the main filament that is occupied by ionized gas.

The total gas mass is an important parameter, but is difficult to accurately ascertain. We discuss below various estimates from the literature of the masses of the gas and dust in the filaments. These provide an indication of the quantities of material present. Sparks et al. (1989), for example, studied the dust absorption in optical imaging of the galaxy and calculate a mass of dust in the main dust lane >2 × 10⁷ M_☉. This would then imply, using a Galactic gas-to-dust mass ratio, that the atomic component of the filaments has M ∼ 2 × 10⁹ M_☉.

Using the color gradient in the stellar continuum, de Jong et al. (1990) estimate the mass of dust present in the galaxy and dust lane from IRAS 100 μm observations, by assuming that the gradient is entirely due to reddening rather than metallicity changes, and calculating the mass of dust required to produce it. They obtain a total dust mass of ∼(0.2–5) × 10⁶ M_☉, with approximately 10% of this mass located in the dust lane and the rest distributed over the central ∼10 kpc of the galaxy. Assuming a Galactic gas-to-dust ratio, this implies a total gas mass on the order of M ∼ 3 × 10⁸ M_☉.

Such estimates of dust mass cannot fully take into account the total gas mass, if the dust clouds become optically thick and therefore cool. The mass in BCG line-emitting filaments is believed to be dominated by molecular material in such cloud cores. Edge (2001) has measured CO emission in a sample of luminous cool-core galaxies that is consistent with the presence of 10⁹–10^11.5 M_☉  of molecular gas at T ∼ 40 K in these galaxies. Using Figure 10 of their paper, along with the inferred mass deposition rate for NGC 4696 of 20 M_☉ yr⁻¹ from Fabian et al. (1982) then implies a molecular gas mass of M ∼ 2 × 10⁹ M_☉  in NGC 4696, in close agreement with the value inferred by Sparks et al. (1989) from the dust obscuration. In NGC 4696, Johnstone et al. (2007) derive a mass of warm H₂ of 1.3 × 10⁵  M_☉, from emission within their Spitzer spectrograph aperture and estimate that approximately 10 times more mass may exist at cooler temperatures.

As previously, we adopt an average reddening of A_V ∼ 0.4 along the extent of the main filament. Assuming a Galactic gas-to-dust ratio of N_H/E(B − V) ∼ 5.8 × 10²¹ cm⁻² mag⁻¹ (Bohlin et al. 1978), a length of 32'', and width of 2'' over which the emission in the filament is measured, we can make a rough estimate of the mass of material obscuring the line-emission region of M ∼ 10⁷ M_☉. This estimate is a lower limit for total mass because the ionized gas cannot be located entirely behind a dusty foreground screen, but is more likely mixed in with it. Over the total area of the line-emitting filaments, using the same average reddening, the resulting limit on the total mass is M > 8 × 10⁷ M_☉.

Although these mass estimates are approximate and vary over a large range, they clearly indicate that there is a large mass of material in the extended filaments, of order a few times 10⁹ M_☉. It is most likely that this mass is largely in a cool, dusty, and dense molecular phase.

There have been a number of long-slit spectroscopic studies of the kinematics of the gas in NGC 4696. Studies by Sparks et al. (1997) and de Jong et al. (1990) revealed smoothly varying line-of-sight velocities with values changing by ∼400 km s⁻¹ in the main filament. In addition, Sparks et al. (1989) detect sodium absorption features in continuum-subtracted long-slit spectra that have similar kinematics to the emission-line gas along both position angles observed. The measured line-of-sight velocities are between 0 and 450 km s⁻¹. There is a cold, neutral component that is associated with the ionized filaments. They comment that this would not be expected in a cooling-flow model, but is consistent with a merger.

Our integral-field spectroscopy greatly improves our ability to map the velocity dispersions and kinematics of the line-emitting gas. The middle panels of Figure 4 show the line-of-sight velocity of the emitting material measured from the Doppler shift of the Gaussian line profile centroids, relative to the galaxy. The bottom panel shows the line-of-sight velocity dispersion from the fitted profile widths, after correcting for the instrumental width (σ = 41 km s⁻¹).

The velocity dispersion detected across the main filament is relatively uniform, with most of the material having velocity dispersions of ∼120 ± 10 km s⁻¹  (see Figure 4). Ridges of higher dispersion (160–200 km s⁻¹) appear in the maps, but the dispersions in these regions are consistent with "beam-smearing" caused by our finite spatial resolution, which results in an increase in apparent velocity dispersion in regions of a steep spatial gradient in line-of-sight velocity. The measured velocity dispersions from the Hα and [N ii] lines are effectively identical.

The velocity dispersions in the main filament are significantly lower than the rotational/bulk velocities of the gas: without correcting for inclination, the ratio of rotation to turbulent velocity ranges from 1 to 2 in this structure. The central stellar velocity dispersion in the galaxy is also a factor of two larger than the line-of-sight velocity dispersion of the line emission. The velocity dispersion is much larger than the thermal broadening (∼7 km s⁻¹ at 10⁴ K for hydrogen) and is therefore either generated by shocks or by the superposition of large-scale random turbulent motions.

The main filament has broader lines and higher line-of-sight velocities than the outer filament arm to the south, in which the average velocity dispersion ranges from ∼60 km s⁻¹ to ∼90 km s⁻¹ and the line-of-sight velocity from ∼100 km s⁻¹ to ∼150 km s⁻¹. This outer filament also has lower surface brightness.

The velocity distribution of the line-emitting gas shows an ordered structure with a component of rotation in the nebula. The velocity maps derived from the line centroid positions for other ions are essentially identical to the [N ii] and Hα velocity maps shown in Figure 4, indicating that the hydrogen recombination and forbidden lines in the optical spectrum originate from the same gas.

The emission from the ionized gas is largely redshifted to the south of the galaxy center and blueshifted to the north. There is a smooth variation in the line-of-sight velocity along the principal filament that traces the dust spiral. The velocity increases inward along the filament as it bends around the galaxy to the south of the core and then decreases to become blueshifted as the filament curves to the north past the galaxy nucleus. As a result, the kinematics of the gas in this filament resemble orbital motions: if the material is infalling, then the filament is inclined so that the outermost end is nearest to us and the material recedes along the filament to the south of the core, before turning around on the far side of the galaxy core onto an approaching trajectory to the north of the nucleus.

At the outer end of the main filament at a radius of 13'' (2.7 kpc), the average line-of-sight velocity of the emitting gas is 180 km s⁻¹. This velocity increases to 280 km s⁻¹  at a radius of 65 (1.4 kpc) in the filament to the southwest. The maximum blueshifted velocity is 150 km s⁻¹  at a radius of 7'' (1.5 kpc) to the north of the core. There is a total velocity range of 430 km s⁻¹  in the measured line centroids, similar to the value measured by Sparks et al. (1997).

Assuming a Keplerian orbit with radius r = 1.5 kpc and circular velocity v = 300 km s⁻¹, the enclosed mass would be 3.1 × 10¹⁰ M_☉. The mass profile derived from the continuum surface brightness profile of the galaxy in Section 3.1 indicates that there is a stellar mass of 3.6 × 10¹⁰  M_☉ inside this radius, in excellent agreement with our kinematic estimate, and showing that the contribution of dark matter to the total mass in this central region must be fairly small. The orbital timescale implied by material at a galactic radius of 1.5 kpc traveling at velocity ∼300 km s⁻¹  is ∼3 × 10⁷ yr. Taking the total gas mass (from Section 3.4) as 2 × 10⁹ M_☉, the kinetic energy of the gas would be 1/2Mv² ∼ 1 × 10⁵⁷ erg. If this energy is radiated away by shocks on the timescale of a single orbit, the luminosity produced would be 2 × 10⁴² erg s⁻¹. The Hα line luminosity from our line measurements is (1.8 ± 0.2) × 10⁴⁰ erg s⁻¹, and from the shock models presented below, this is about 0.7% of the total luminosity. Thus, our estimated total shock luminosity is ∼2.6 × 10⁴² erg s⁻¹, in good agreement with our kinematic estimate of this quantity.

Previous long-slit spectroscopy of NGC 4696 has shown that the extended filaments produce emission with a high ratio of [N ii] to Hα flux (Lewis et al. 2003). We confirm this result. Figure 7 shows that the ratio of [N ii]/Hα is roughly constant at 2.0 ± 0.1 with remarkably little variation in the value (∼5%) across the inner filaments of the galaxy. This average value of the [N ii]/Hα flux ratio and the Hα luminosity of the system reported in Section 3.4 are consistent with other BCG emission systems of similar luminosity (Crawford et al. 1999). These results show that the characteristic line ratios, such as the high [N ii] flux relative to Hα and dominance of the other low-ionization species, are not simply nuclear phenomena confined to the galactic core, but persist over the extent of the detected emission in the filaments. In addition, the uniformity of the emission-line flux ratios across the inner filaments implies that whatever mechanism excites this spectrum is acting over the full extent (25'' or 5 kpc) of the detected filaments.

Zoom In Zoom Out Reset image size

This finding contrasts with results from other BCG filament systems. For example, Hatch et al. (2006) placed a spectrograph slit along several of the filaments in NGC 1275 and found the [N ii]/Hα ratio decreases along the filaments with distance from the galaxy. The authors suggest that this may be the result of either decreasing metallicity or changing excitation, such as greater specific star formation activity closer to the nucleus. Unlike NGC 4696, NGC 1275 is a very line-luminous system among BCGs and there is clear evidence of star-forming clouds associated with the extended filaments.

Edwards et al. (2009) compare optical integral-field unit (IFU) observations of the extended emission in a sample of cooling-flow and non-cooling-flow BCGs. They find diverse emission properties and morphologies among their sample, including a range of behavior in the variation of the [N ii]/Hα line flux ratios. In general, they attribute lower [N ii]/Hα ratios to regions where young stars contribute most of the ionizing flux. They conclude that the ionization mechanisms that produce the optical line emission are not uniform among the sample and represent a complex interplay between an AGN, cooling cluster gas, and star formation that may be triggered by AGN outbursts and/or tidal interactions with other galaxies. In IFU observations of a sample of other BCGs with a range of Hα luminosities, Hatch et al. (2007) find different degrees of [N ii]/Hα variation in the observed systems. For example, in RXJ 0821+0752 there is a distinct correlation between the low [N ii]/Hα and Hα flux, whereas in the BCG of A2390 there is less variation, but in general higher [N ii]/Hα ratios are associated with dust lanes and lower ratios with knots of blue emission. This may also reflect the trend among BCGs to display weaker forbidden emission in high-luminosity systems with abundant star formation (as evidenced by blue knots). Wilman et al. (2006) found little variation in the flux ratios over the measured extent of nebulae surrounding four high-luminosity systems. If, as suspected, such systems are dominated by more widespread regions of star formation, the emission-line ratios may be simply consistent with photoionization by hot stars. On this hypothesis, they suggested that in lower-luminosity systems the flux ratios might show more spatial variation with higher [N ii]/Hα ratios away from the less-abundant star-forming clumps.

NGC 4696 may represent the low-luminosity extreme of this model, as there are no detectable star-forming regions associated with the filaments. In this case, the underlying low-ionization excitation mechanism dominates everywhere. M87, low line-luminosity BCG of the Virgo Cluster, may be a similar example, though in this case observations resolve the core of the galaxy and the emission-line ratios associated with the extended emission differ from those of the nucleus. The extended filaments produce a typically LINER-type spectrum (Ford & Butcher 1979), found by Dopita et al. (1997) to be consistent with shock models, whereas the higher-ionization emission from the core is consistent with excitation by photoionization (Sabra et al. 2003).

The [O iii]/Hβ ratio is a powerful diagnostic of the excitation of the gas. This ratio is also measurable in the central regions of the filaments in our data, though the signal-to-noise ratio is much reduced as a result of the overall weaker line emission. We find little evidence of a variation in the value over the inner filament. However, there is evidence for a slight increase in the ratio at the galaxy core, where the value may be enhanced by radiation from the (weak) AGN in this system. This effect can be seen in the line-flux maps of Figure 4, as the peak of the [O iii] emission is closer to the galaxy core than the peak of the hydrogen lines. However, over most of the extent of the filaments, the line ratios, including [O iii]/Hβ, do not indicate a dominant contribution from AGN photoionization. The main filament uniformly produces a low-ionization (LINER-like) emission spectrum (see Section 3.7). We do not detect He i λ5876, which is sometimes seen in more Hα-luminous BCG systems (e.g., Sabra et al. 2000).

The intrinsic velocity dispersion in the ionized gas causes the [O ii] λλ3729, 26 doublet to be unresolved in our spectra. However, our wavelength range includes another density-sensitive pair of collisionally excited lines: [S ii] λλ6716, 31. Figure 7 maps the ratio of the flux in the 6731 Å component to that in the 6716 Å line. The average value of the [S ii] doublet flux ratio over the central filament is 1.5 ± 0.1, which is consistent with densities close to or within the low-density limit (≲300 cm⁻³). Again the [S ii] flux ratio is uniform over the brightest regions in the filaments, with a 1σ variation of 0.2 in the value.

The most commonly used optical diagnostic line-flux diagrams were developed by Baldwin et al. (1981) and Veilleux & Osterbrock (1987). These employ strong lines that are close in wavelength to decrease the effect of reddening on the ratios. Such diagnostic diagrams are often used in classifying galaxies and identifying AGN versus starburst-dominated systems by distinguishing different ionization mechanisms (e.g., Kewley et al. 2006).

In Figure 8, we plot the flux ratios from Table 1 at the emission-line peak on the common diagnostic diagrams. This figure shows the regions occupied by AGNs, H ii regions, and LINERs, as defined by the classification scheme of Kewley et al. (2006), developed from their earlier work (Kewley et al. 2001). The typical line ratios from Table 1 for NGC 4696 are plotted along with the integrated line ratios measured by Crawford et al. (1999) for a sample of X-ray-selected BCGs. The relative sizes of the points represent the relative Hα luminosities measured in that study. These demonstrate clearly the LINER-like properties of the filament spectrum of NGC 4696, typical of a low Hα-luminosity system. Thus, any conclusions that apply to the mode of excitation of NGC 4696 may be applicable to other low-luminosity BCGs.

Zoom In Zoom Out Reset image size

We have run a grid of radiative shock models using the photoionization and shock modeling code MAPPINGS III, an updated version of the code described in Sutherland & Dopita (1993). The motivation for these simulations is a model in which the filaments are comprised of clouds of cool gas that are remnants of a recently accreted, gas-rich, neighboring dwarf galaxy. These clouds are falling into the hot interstellar medium (ISM) of NGC 4696, which is compressed and heated by the ram pressure of the cloud's motion. The increased external pressure drives shocks into the cooler, dense cloud gas. The cooling time in the dense cloud material is short, so the cloud shocks are radiative and the observed optical line emission is produced in the cooling region behind the shocks. Figure 9 shows a simple schematic of this model and is further discussed in Section 4.2.

Zoom In Zoom Out Reset image size

In the one-dimensional plane-parallel shock models, the pre-shock medium is taken to be the warm atomic component of the infalling gas, rather than the dense molecular medium that is visible in the dust absorption. The reason for this choice is as follows. The total surface brightness of a radiative shock of velocity v propagating into a medium of pre-shock density ρ scales as the mechanical energy flux through the shock S = 1/2ρv³. For a system of shocks propagating into a fractal medium with a wide range of temperatures, the driving (ram) pressure P = ρv² = constant. Thus v ∝ ρ^−1/2 and therefore S ∝ ρ^−1/2. High-density regions do not significantly contribute to the total shock luminosity, not only because of this density-dependent factor, but also because the fractional filling factor of the denser cloud gas is small, reducing the total area that can radiate. Furthermore, in shocks propagating into molecular gas, the dominant luminosity is in collisionally excited H₂ infrared emission, not in atomic line emission. Indeed infrared emission lines from H₂ gas are also seen, and their luminosity is correlated with the optical line emission (Edge et al. 2002; Johnstone et al. 2007), suggesting that these are produced by a common shock mechanism.

The ionization state of the material prior to processing by the shocks (the pre-ionization) and the shock velocity are treated as independent parameters in the models. The shock velocity is varied between 80 and 220 km s⁻¹  in intervals of 20 km s⁻¹, and the pre-ionization from neutral (i.e., the fraction of ionized hydrogen is n_H ii/n_H = 0) to fully ionized (n_H ii/n_H = 1), with the fraction of hydrogen that is ionized in the pre-shock material increasing in steps of 0.1. A transverse magnetic field that is consistent with equipartition of thermal and magnetic field energy is assumed (B = 5 μG for n = 10 cm⁻³; B ∝ n^−1/2). No external radiation field was added.

The gas abundances in Table 2 are taken from those presented by Asplund et al. (2009), with typical dust depletion factors, since dust physics is not treated explicitly in these models. Observations of the Centaurus Cluster show that the ICM gas abundance in the vicinity of the galaxy and filaments is Z/Z_☉ ∼ 2 (Graham et al. 2006; Sanders & Fabian 2006) and this overall metallicity is used in the models. However, we also find that an enhanced nitrogen abundance, by a factor of two, is required to better fit the observations. Nitrogen is made in the more massive asymptotic giant branch (AGB) stars with hot-bottom burning (5 < M/M_☉ < 8), so possibly this nitrogen enhancement is associated with relatively less high-mass star formation and an older stellar population in the dense cluster environment.

With this abundance set, the shock model is sought that provides a least-squares best fit to the values of the following line flux ratios: [O iii] λ5007/Hβ, [N i] λ5198+5200/Hβ, [O i] λ6300/Hα, [N ii] λ6583/Hα, and [S ii] λ6731/Hα. The individual line ratios from the grid of models are shown on the line flux ratio diagrams in Figures 10–15, along with the line ratios measured in the filaments of NGC 4696. The range of parameter space plotted here is much narrower than the normal range displayed on these types of diagnostic diagrams. All except the [N i] λ5198+5200/Hβ ratio are consistent with shock models with velocities in the range v_s = 180–200 km s⁻¹. We have determined the best-fit model spectrum by minimizing the rms fit to all these ratios. The resulting best-fit model has a shock velocity v_s = 180 km s⁻¹. The best-fit model also has complete pre-ionization of H, though it can be seen from the diagnostic diagrams in Figures 10–15 that the results are not very sensitive to the pre-ionization of the high-velocity models. However, at shock velocities of 100–200 km s⁻¹, the pre-shock gas will be photoionized by the upstream radiation flux from the shocked gas, so the most highly ionized model is appropriate for these velocities. The best-fit shock spectrum (at optical wavelengths) is given in Table 3.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Insufficient observed strength of the [O iii] λ4363 line has been used to argue against shock models as a possible exciting mechanism in the filaments of BCGs (e.g., Sabra et al. 2000; Voit & Donahue 1997). We do not detect the [O iii] line in the spectra of NGC 4696, however, this is consistent with the predictions of the shock models described here. We can place an upper limit on the line strength of ≲6 × 10⁻¹⁷ erg s⁻¹ cm⁻² arcsec⁻²  and therefore the ratio of fluxes [O iii]/Hβ ≲ 0.35, at the emission-line peak. From the model spectrum in Table 3, the expected [O iii] λ4364 flux is approximately 10% of the Hβ line strength, and so below the detectable level and consistent with the data.

The only other emission line that is produced in the shock model spectrum at a level that may be detectable in our data is [Ne iii] λ3869; [Ne iii]/Hβ = 0.34 from Table 3. However, as indicated above, at the blue end of the spectrum the ratio of detectable line flux to the flux in Hβ at the line-emission peak is ≳0.35, so the predicted [Ne iii] strength lies at the edge of our sensitivity limit. There is evidence for a marginal detection of this line at the emission peak (see Figure 3), with an integrated flux of ≈ (8 ± 5) × 10⁻¹⁶ erg s⁻¹ cm⁻²  (after extinction correction), but the fluxes in individual pixels are below the limit of detection.

Though consistent in other measures, the shock models overpredict the strength of the [N i]  λ5200 doublet by a factor of ∼1.4, the effect of which is seen in Figure 13. Measurement errors may contribute to this discrepancy: the [N i] line flux is close to the detection limit in the blue spectra, and the line lies on the wing of the deep Mg i absorption feature, which could introduce systematic errors in the galaxy continuum subtraction and line fitting that are not represented in the uncertainties given in Table 1. Additionally, in the models the flux of the [N i] line is very sensitive to the boundary conditions that terminate the simulations and so is strongly affected by the selected model configuration.

In some sense, depending on projection effects, the velocity dispersion from the emission-line widths reflects the shock velocity. The line flux ratios also depend on the shock velocity and should therefore correlate with the velocity dispersion. So if the line ratios are uniform, then the velocity dispersion is also expected to be uniform. This is indeed observed to be the case in the main filament.

We conclude then that radiative shock models having super-solar abundances, and shock velocities of around v_s = 180 km s⁻¹  provide a good fit to the observed emission-line spectrum of NGC 4696 across the main filament. We will now examine how we can use this cloud-shock model to further constrain the parameters of the infalling emission-line gas.

In our model, the infalling motion of the gas clouds generates the ram pressure that drives radiative shocks into the cloud material. The cloud velocity and the properties of the hot, ionized gas in the galaxy halo (ISM) determine the magnitude of the driving pressure that is obtained.

Figure 9 outlines the geometry of the model and the components referred to in the following discussion. As a cool gas cloud falls through the ambient medium, a bow wave, or shock if supersonic, forms in the ISM ahead of the cloud. The properties of the ambient ISM gas are denoted with the subscript 1, and v₁ is the velocity of the ISM relative to the cloud (i.e., having magnitude equal to the infall velocity). Immediately behind the bow wave the gas has pressure, density, and temperature P₂, ρ₂, T₂, and the velocity relative to the bow wave is v₂. If v₁ is greater than the sound speed in the ISM c_s,1 = (γ p₁/ρ₁)^1/2, the bow wave is a shock, and the Mach number is M₁ = v₁/c_s,1 > 1. In this case, the pressure will increase across the shock front. If the motion is subsonic v₁ < c_s,1, so M₁ < 1, and the pressure is constant across the bow wave p₂ = p₁.

We determine the maximum pressure that acts on the cloud (due to its motion through the ISM) as the stagnation pressure at the leading surface of the cloud where the flow comes to rest. The Bernoulli equation along a streamline traversing the post-bow-wave region, assuming an adiabatic, compressible flow behind the bow wave, gives

$$\frac{1}{2} v_{s}^2 + h_{s} = \hbox{Streamline Constant} = \frac{1}{2} v_2^2 + h_2,$$

where v_s = 0 is the velocity at the stagnation point, and h is the specific enthalpy,

$$h = \frac{\gamma }{\gamma -1}\frac{p}{\rho } = \frac{c_{s}^2}{\gamma -1}.$$

From this, the stagnation pressure can be expressed as

$$\begin{equation} p_s = p_2 \, \left[ 1 + \frac{\gamma -1}{2} M_2^2 \right]^{\gamma /(\gamma -1)} \end{equation} \tag{ 5 }$$

in terms of the conditions in the flow behind the bow wave. We compare the infall velocity of the cloud to the ambient sound speed in the ISM to estimate the Mach number of the bow wave (M₁) and hence the properties of the gas behind it.

Graham et al. (2006) present density, pressure, and temperature profiles of the cluster gas, derived from X-ray observations obtained with Chandra, XMM-Newton, and ROSAT. We use these to estimate the pressure and sound speed in the halo of NGC 4696 and constrain the pressure driving the shocks in the filament clouds. At a radius of ∼1 kpc (5''), which corresponds to the inner region of the filaments, the X-ray-emitting plasma has a temperature of T₁ = (8 ± 3) × 10⁶ K and hydrogen particle density n_H,1 = 0.13 ± 0.04 cm⁻³. The sound speed in this gas is then c_s,1 = 440 ± 50 km s⁻¹  and the pressure is p₁/k = (2.5 ± 0.5) × 10⁶ K cm⁻³.

From our IFU observations of the inner filament, we assume an infall velocity of the clouds in the filament of v₁ ∼ 300 km s⁻¹, approximately the maximum line-of-sight velocity of the emission in the filament. The Mach number of the flow is then M₁ ∼ 0.7. In this case, the flow is in the transonic regime and a weak bow shock forms. For a transonic shock M₁ ≈ 1 ≈ M₂, and the pressure remains approximately constant across the shock: p₁ ≈ p₂ ≈ ρ₁ v²₁/γ. From Equation (5), the expression for p_s becomes

$$\begin{eqnarray} p_s &\approx & \frac{\rho _1 v_1^2}{\gamma } \left[\frac{\gamma +1}{2} \right]^{\gamma /(\gamma -1)} \\ &\approx & 1.23 \, \rho _1 v_1^2 = 2.05 \, p_1 \quad \hbox{for} \quad \gamma = \frac{5}{3}, \end{eqnarray} \tag{ 6 }$$

so the static pressure of the X-ray-emitting ISM is enhanced by a factor of approximately two at the surface of the cloud.

At the radius of the inner filaments, the mass density of the ISM is ρ₁ = 1.4 m_H n_H,1 ∼ 3.0 × 10⁻²⁵ cm⁻³. With a bow shock velocity of v₁ ∼ 300 km s⁻¹, the resulting stagnation pressure is p_s/k = 2.4 × 10⁶ K cm⁻³. This post-bow-shock pressure drives the cloud shock by providing an equivalent pressure to the gas behind the cloud shock. We assume this pressure remains approximately constant across the post-cloud-shock region of the cloud (denoted by the subscript 3 in Figure 9).

For a strong shock, the pressure in the post-cloud-shock region p₃ is related to the pre-shock cloud gas properties (p_c, ρ_c) as follows:

$$\begin{eqnarray} p_{3} &=& \frac{2}{\gamma +1}\rho _{c}v_{s}^{2} \\ &=& \frac{2}{\gamma +1} \mu \,m_{{\rm H}}\, n_{{\rm H},c}\, v_{s}^{2}, \end{eqnarray} \tag{ 8 }$$

and the pre-shock hydrogen particle density in the cloud is then

$$\begin{eqnarray} n_{{\rm H},c} &\approx & \frac{\gamma +1}{2} \frac{p_{s}}{ \mu \,m_{{\rm H}}\,v_{s}^{2} }, \\ &\approx & 1.76\times 10^{9} \, p_{s} \end{eqnarray} \tag{ 10 }$$

for γ = 5/3 and the best-fit shock-model velocity v_s = 180 km s⁻¹. Hydrogen is fully ionized in the pre-shock cloud gas, so μ = 1.4. From these results, we can estimate the density of the ionized cloud material. For the inner filaments, we then find a hydrogen density n_H,c ≈ 1 cm⁻³.

The pre-shock density in these models cannot be constrained by the [S ii] λ6716/6731 ratio, as the observations are consistent with the low-density limit (n_e < 300 cm⁻³  in the recombination zone of the shock, with T_e ∼ 6000 K). This prevents a direct measurement of the ram pressure associated with the shock. A somewhat less accurate estimate can be obtained from the Hα surface brightness, since this scales as the total emitted shock surface luminosity. The best-fit shock model has emitted surface luminosity in Hα:

$$\begin{equation} S_{\rm H\alpha } = \frac{S_{\rm H\alpha }}{S_{\rm total}} \rho v^{3} = 9.6 \times 10^{-5} \left[ \frac{n_{e,c}}{{\rm cm^{-3}}} \right]\hbox{\, erg\,s$^{-1}$\,cm$^{-2}$}, \end{equation} \tag{ 12 }$$

where n_e,c is the electron density in the pre-shock cloud material.

The average Hα surface brightness in the main filament measured in the 3'' diameter aperture around the region of the line peak is I_Hα ∼ 9.4 × 10⁻¹⁶ erg s⁻¹ cm⁻² arcsec⁻². The measured surface brightness varies across the filaments, with a typical value in the outer region of the main filament of ∼1 × 10⁻¹⁶ erg s⁻¹ cm⁻² arcsec⁻². This corresponds to a measured surface luminosity from the inner filament of

$$\begin{equation} S_{{\rm H}\alpha } = I_{{\rm H}\alpha } \frac{4\pi D^2}{\phi }\left(\frac{206264.8}{D}\right)^2 = \frac{5.35\times 10^{11}I_{{\rm H}\alpha }}{\phi }, \end{equation} \tag{ 13 }$$

where ϕ is the filling fraction of shock emission inside the 3'' aperture. This fraction is highly uncertain. However, if we assume ϕ = 1, we obtain S_Hα ∼ 5.2 × 10⁻⁴ erg s⁻¹ cm⁻²  in the central filaments and ∼5.5 × 10⁻⁵ erg s⁻¹ cm⁻²  in the outer region. Using the relation between the density and surface luminosity in Equation (12), we find a pre-shock density, n_e,c ∼ 5 cm⁻³ near the nucleus, and n_e,c ∼ 0.6 cm⁻³ in the outer regions. In the ionized pre-shock medium, the equivalent hydrogen densities are ∼4 cm⁻³ and ∼0.5 cm⁻³. There are uncertainties associated with these estimates, most notably in the filling factor of the shock surfaces in the surface luminosity estimates. However, the result is consistent with the density estimates of ∼1 cm⁻³ derived through the bow-shock/cloud-shock pressure arguments presented previously. The radiative shock model appropriate to the nuclear region with v_s = 180 km s⁻¹, n₀ ∼ 4 cm⁻³, and equipartition magnetic field B = 5 μG produces a mean density in the [S ii]-emitting zone of the shock of 190 cm⁻³, consistent with the [S ii] λ6716/6731 flux ratio being observed at its low-density limit. These results provide support for the feasibility of the model as a mechanism for generating shock-excited emission-line spectra in the extended filaments of NGC 4696.

The increased pressure in the bow shock region will lead to an enhancement of the X-ray emissivity in the vicinity of filaments. This can provide a natural explanation of the correlation between the X-ray and the optical emission observed by Crawford et al. (2005). These authors measure the 0.1–10 keV flux in the filament 22'' to the southeast of the nucleus to be ∼7 × 10⁻¹⁵ erg s⁻¹ cm⁻² arcsec⁻². From our data, the Hα surface brightness in this outer filament is ∼5 × 10⁻¹⁷ erg s⁻¹ cm⁻² arcsec⁻², so the Hα flux in the region is approximately 0.7% of the soft X-ray flux. The total surface brightness in the cloud shocks, from the model, is about 150 times the Hα surface brightness. Combining this with the measured ratio of X-ray to Hα flux, the ratio of the X-ray to total cloud-shock surface brightness is estimated to be S_X/S_c ∼ 0.9. Thus, the energy dissipation in the X-ray plasma is comparable to that of the cloud shocks. This ratio might be lower if there are faster (non-radiative) shocks in the infalling gas, as these could provide soft X-ray luminosity associated with the infalling gas, rather than the shocked ISM. Nonetheless, it is clear that the gas-dynamic drag of the hot ISM is an important factor in determining the accretion timescale of the infalling gas toward the nucleus.

Although LINER emission in galaxies is frequently a nuclear phenomenon, there are a range of galactic environments that host extended, extra-nuclear, LINER-like emission regions. Other than cool-core cluster BCGs, these include galactic winds, AGN jets, and interacting or merging galaxies.

A link between LINER-like emission and shock excitation was proposed in early studies of LINER galaxies and has since been extensively discussed and debated in the literature (e.g., Dopita 1994). Indeed, from early investigations of LINER spectra, Koski & Osterbrock (1976), Fosbury et al. (1978), and Heckman (1980) concluded that shock heating is responsible for producing the low-ionization spectrum in LINERs. More recently, Filippenko (2003) noted that despite a lack of strong evidence for shock heating in galactic nuclei, shocks are likely to dominate in cases of spatially extended LINER-like ionization. A connection between LINER emission and shock excitation has been proposed in models of galaxy "superwinds," e.g., in NGC 839 (Rich et al. 2010) and NGC 3079 (Filippenko & Sargent 1992), and in interactions and mergers (e.g., Delgado & Perez 1996).

Most interesting in the context of this paper are recent findings that increasingly point to shock excitation of LINER-like emission in the tidal regions of galaxy mergers and interactions. Ogle et al. (2007) present Spitzer observations of the luminous infrared emission associated with the double galaxy system 3C 326. The northern galaxy of the pair produces strong H₂ line emission and a LINER-type optical spectrum. The authors investigate several heating mechanisms for the IR emission and conclude that the molecular hydrogen spectrum may be powered by accretion shocks that arise in inflows induced by tidal interactions between the two galaxies. Monreal-Ibero et al. (2006, 2010) investigate the ionization of the extended emission regions in a sample of luminous infrared galaxies (LIRGs) using integral-field spectroscopy. They find that the emission in isolated systems is generally consistent with ionization by radiation from young stars. However, large fractions of the emission spectra from the extended regions in interacting pairs and more advanced mergers are consistent with ionization by shocks with velocities of 150–500 km s⁻¹. In their model, the shocks are generated by tidally induced, high-velocity flows that can be a result of the merging process (e.g., McDowell et al. 2003). They suggest that tidal forces may be central to the origin of shocks that ionize the gas in the extended regions and produce the low-ionization spectrum. The most luminous BCGs do produce IR luminosities that place them in the class of LIRGs (L_IR > 10¹¹ L_☉; e.g., Egami et al. 2006b). While NGC 4696 is not such an IR-luminous system: L_IR = 2.3 × 10⁹ L_☉ (Kaneda et al. 2007), results like those discussed above support the suggestion that infall and merger activity provides a mechanism for generating galaxy-wide LINER-like emission through intermediate-velocity shock excitation.

Lastly, we note the observations presented by Edwards et al. (2009) of the BCG system in the Ophiuchus cluster, which make an interesting comparison with our results for NGC 4696. In Ophiuchus, the BCG does not produce detectable line emission, but is accompanied by a nearby extended object that emits strong low-ionization optical lines. This object is ∼1.7 kpc from the center of the BCG on the sky and is separated by 600 km s⁻¹  in radial velocity. The red emission-line ratios fall in the AGN or LINER regions of the diagnostic diagrams, but the [O iii]/Hβ ratio is not available to distinguish between these classes. The authors conclude that it is most likely that the emission is produced within a large gas cloud falling toward the BCG through the galaxy halo. It is possible that the emission in this structure is excited by the ram pressure of its motion through the hot atmosphere of the BCG, as in the model we have discussed here. The Ophiuchus system may be in the early stages of a process of minor merger like the one that has occurred in NGC 4696. Further study is needed to better understand the nature of the interaction.