POLYCYCLIC AROMATIC HYDROCARBONS, IONIZED GAS, AND MOLECULAR HYDROGEN IN BRIGHTEST CLUSTER GALAXIES OF COOL-CORE CLUSTERS OF GALAXIES

The Spitzer IRS observations took place in 2005 and 2006 (see Table 1), and the data were reprocessed in 2009 April (v18.7). Both short (SL) and long (LL) wavelength observations were obtained, at low spectral resolution, R ∼ 60–130. With two spectral orders each, we obtained a total of four spectral modules. We have sparse spectral maps of nine cool-core BCGs, though we analyze only the central region here. We used IRSCLEAN v1.7 to apply the bad pixel mask supplied by the Spitzer pipeline and to find additional rogue pixels using a WCLEAN formula with an aggressive level of 0.5 (suitable for relatively faint targets such as these). The LL pixels are 51 across, while the SL pixels are 18 across (Houck et al. 2004).

To cross-check our flux calibration of the IRS spectroscopy, we also analyzed photometry data from the Spitzer Infrared Array Camera (IRAC; Fazio et al. 2004) and the from the Multiband Imaging Photometer for Spitzer (MIPS; Rieke et al. 2004). We list the archival observations (known by their Astronomical Observation Requests or AORs) in Table 2. The aperture photometry is discussed in Section 2.3.

This section discusses our choices regarding the IRS spectral extraction process, with particular attention to the treatment of extended sources compared to point sources. We used the software package CUBISM v1.7⁷ (Smith et al. 2007a) to combine the exposures into a data cube or spectral map. To reduce noise near the ends of the slit, we trimmed the exposures in the cross-dispersion direction by 3%–5%. We confirmed that the off-target, paired observations for each spectrum were indeed source free. These blank sky spectra were used for background subtraction.

To remove rogue pixels which were not caught by IRSCLEAN, we composed bad pixel lists using CUBISM's autobadpix algorithms, on both the global and individual record levels. At the global level, we flagged any pixel which deviated by more than 2.5σ from the median level in at least 50% of its appearances in the cube. This method flags only a few pixels, but each of these pixels has a relatively large effect on the spectral map. At the record level, we conservatively flagged any pixel which is a 5σ outlier in at least 75% of its appearances in the cube. This method flags more pixels, but only in individual exposures. We also manually removed obvious rogue pixels.

The Spitzer pipeline is optimized for single-slit observations of point sources, including a correction for the light lost from the slit, which can be as much as 36% (Smith et al. 2007a, Figure 4). However, CUBISM includes an option to remove this slit-loss correction factor, in order to extract spectra of extended sources. To determine which targets to treat as point sources, we used CUBISM to create a map combining all SL1 wavelengths and averaged the two rows which covered the peak source emission. Using this light profile, we measured the FWHM along the slit (see Table 3). We did the same for the LL2. These modules were selected because their signal to noise is highest. Spitzer's point-spread function (PSF) has an average FWHM of 26 in the SL and 66 in the LL.⁸ Four of our nine galaxies have an FWHM in the SL1 of ≲ 55 and were considered to be point sources. The other five sources do not uniformly fill the Spitzer slit, but they are not well characterized as point sources.

For the point sources, we extracted spectra using an aperture that just spans the slit (2 pixels) and a length that captures most of the light along the slit (see Table 3). Spitzer data are calibrated for point sources and for this kind of aperture. Single-pointing software such as SMART⁹ (Higdon et al. 2004; Lebouteiller et al. 2010) uses a similar aperture with its "tapered column" extraction, increasing the length of the aperture as the PSF broadens with wavelength, and further optimizing the extraction by weighting each pixel by its signal to noise (Lebouteiller et al. 2010). In order to use the same software for all of our spectra, we used CUBISM to extract our point-source spectra, approximating the tapered column type of aperture. (For PKS 0745-19 SL2, we truncated the aperture to avoid a noisy region. For A1068 SL2, CUBISM spreads the light from those two pixels across three rows.) To check our procedure, we also extracted the point-source spectra using SMART and found agreement to within about 10%, sometimes to within 2%.

For the extended targets, we removed the pipeline slit-loss correction factor. Our apertures include much of the available light with good signal to noise. However, our sparse spectral maps do not cover the full extent of the source.

After extraction, any noisy edges were trimmed from each order. In the observed frame, the four low-resolution IRS modules span the following wavelength ranges: SL2: 5.2–7.6 μm, SL1: 7.5–14.5 μm, LL2: 14.3–20.6 μm, and LL1: 20.5–37.5 μm. Spectra extracted from CUBISM are reported in units of MJy sr⁻¹, so the size of the extraction aperture is used to convert spectra to units of flux density (mJy). We then corrected to the rest frame for each target (see Table 1) by dividing both the wavelength and fluxes by a factor of (1 + z).

The light collected by the narrow Spitzer slit and our sparse spectral maps represents only a portion of the MIR light. Therefore, to obtain meaningful luminosities, we rely on IRAC and MIPS photometry (see Table 4; as in Egami et al. 2006a). The broadband aperture diameters are approximately three times the source's FWHM measured using the IRS (or equal to the FWHM of Spitzer's PSF for the 70 and 160 μm points) with subtraction of background computed from a larger annulus. We adjusted the aperture size to exclude unrelated foreground or background sources and applied the suggested aperture corrections given by the MIPS Instrument Handbook. At 24 μm, the corrections are 1.17 (for apertures of 26'' and 30'') and 1.13 (for apertures 50''). For 70 and 160 μm, the corrections are 1.22 and 1.752, respectively. The photometric uncertainty is 5% for IRAC, and 10%, 20%, and 20% for the MIPS 24, 70, and 160 μm points, respectively.

Because the spectra within the cited IRAC and MIPS circular apertures may differ from the spectra obtained from within our smaller, rectangular IRS apertures, our analysis and conclusions rely most heavily on relative quantities, i.e., ratios, rather than absolute quantities. All correction factors for each module are listed in Table 3. The interested reader can recover the flux in the apertures listed in Table 3 by dividing the fluxes published here by this factor.

In three cases (Hydra A, A1795, and MS0735), we do not have complete IRAC and MIPS coverage in the IRS wavelength range and needed a robust, standalone scaling procedure. For this purpose, we developed a scaling procedure that did not rely on IRAC or MIPS photometry. We validated this procedure, with scaled IRS spectra for sources with IRAC and MIPS photometry, by comparing the resulting spectrophotometry to IRAC and MIPS photometry. We now describe our scaling procedure.

For the sources we identified as extended, we fit Gaussian, azimuthally symmetric light profiles (and neglected the fainter extended haloes) to estimate how much of the source was included within a circle of the same area as our rectangular aperture. The spatial profile from the SL1 map was used as the reference to determine the scale for the SL1 and SL2 spectra for light from outside the rectangular aperture. The profile from the LL2 map was used for both LL orders. For point sources, this scaling has already been performed by the pipeline (See Section 2.2).

An additional scale factor is needed to match the orders of the IRS spectrum. Flux mismatches are expected between the spectral orders extracted with CUBISM, because it is impossible to perform an exact tapered column extraction using CUBISM. Even tapered column extractions with SMART sometimes show mismatches. In our sample, three targets had mismatches between the first and second orders (SL1/SL2 or LL1/LL2), and three had mismatches between the SL and LL. Some spectra are plagued by noise or decreased signal at the module interface (e.g., 2A0335 near 7.1 μm). We selected the LL2 spectrum as the photometric reference point because of the good agreement between the LL spectrum and MIPS photometry in a similar aperture and because the LL is less vulnerable to slit loss due to pointing errors (because of its larger pixels; Smith et al. 2007a). As mentioned above, the LL2 has superior signal to noise to the LL1. We used low-order polynomial continuum fits to match the modules when major features did not interfere.

In three cases an additional overall factor was needed (Hydra A (1.70), A478 (1.55), and MS0735 (1.12)), probably because of extended halo light contained within the broadband aperture but not represented by our Gaussian profile. This procedure worked well, because agreement between the IRS spectra and broadband photometry was relatively good, almost always within 10%, and usually within 5% (see Figure 1). Therefore, while the IRS spectrum appears to exceed the IRAC value at 8 μm for Hydra A, the agreement between the integrated spectrum and the IRAC photometry point is actually excellent. Note that the comparison here is between the MIPS and IRAC photometry and the observer-frame IRS spectra integrated over appropriate bandpasses. These flux points should not be confused with the integrated IRS photometry in the rest-frame 24 μm MIPS bandpass reported in Table 5.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For analysis and plots, we combine the uncertainty calculated by the Spitzer Science Center pipeline in quadrature with a 15% systematic uncertainty to all fluxes to account for the uncertainty in scaling the IRS photometry relative to Spitzer broadband photometry. The systematic uncertainty dominates in almost all measurements. An additional 5% absolute photometric uncertainty is applied when making comparisons with data from other telescopes.¹⁰

We used the spectral-decomposition package PAHFIT v1.2 (Smith et al. 2007b) to make empirical fits to the IRS spectra and to facilitate direct comparison with results from other workers using the same method. The short-wavelength PAHFIT results are plotted in Figure 1 and the full-wavelength results are shown in Figure 2. PAHFIT fits the following components: a starlight continuum, several thermal dust continuum components, broad PAH emission bands, narrow atomic and molecular emission lines, and broad silicate absorption bands (Figure 1). We customized the list of fitted emission features to a limited set, excluding those that were very weak. In the case of 2A0335, small parts of the spectrum (below 5.3 μm and between 7.05 and 7.35 μm) were excluded from the PAHFIT analysis because noise in those parts of the spectrum hindered a successful fit.

Zoom In Zoom Out Reset image size

The smallest dust grains are the PAHs, composed of only a hundred atoms or so. These structures generate emission from C–H or C–C–C bending modes (Leger & Puget 1984; Boulanger et al. 1998; Van Kerckhoven et al. 2000), excited by the absorption of UV photons (Allamandola et al. 1985; Sloan et al. 1999). UV photons can heat these tiny grains stochastically, causing them to suddenly increase in temperature then cool (e.g., Li & Draine 2001). PAH features at 3.3, 6.2, 7.7, 8.6, and 11.3 μm in spectra are thought to be an excellent tracer of B stars or of relatively recent star formation (Peeters et al. 2004; Brandl et al. 2006; Förster Schreiber et al. 2004). The continuum IR luminosity, which traces star formation, and PAH luminosity are strongly correlated in normal star-forming galaxies (Wu et al. 2010). Studies of low-metallicity star-forming dwarf galaxies by Rosenberg et al. (2008) and of star-forming regions in irregular galaxies by Hunter & Kaufman (2007) show the PAH emission decreases as metallicity decreases, so metallicity is one factor that can lead to scatter in the correlation between PAH emission and IR luminosity from dust.

The brightest features from the PAHs are the complexes at 11.3 μm and 7.7 μm. The sum of those lines in our BCG sample is strongly correlated with the 24 μm continuum in both flux and luminosity (Figure 6). The correlation coefficient r = 0.93 for both. Excluding MS0735 does not affect the correlation. The relationship is very close to linear: L_{11.3 + 7.7}∝L^{0.90 ± 0.03}₂₄, where L₂₄ = νL_ν at 24 μm rest frame. (L_11.3∝L^{0.96 ± 0.05}_TIR.)

Zoom In Zoom Out Reset image size

The ratio of the 11.3 μm luminosity to L_TIR is ∼0.0039 ± 0.0020 (omitting MS0735 from the sample for lack of MIPS data), plotted in Figure 6(b), to compare to the mean of ∼0.0066^+0.0045 _{− 0.0042}L^{0.02 ± 0.03}_11.3 μm from 123 starburst-dominated galaxies from a 24 μm flux-limited sample of 330 galaxies in Wu et al. (2010). The best-fit power law relating L_11.3 μm and L^α_TIR is α = 1.05 ± 0.05, also similar to that seen for starburst galaxies (Wu et al. 2010). Given the uncertainties in converting from L₇₀ and L₂₄ to L_TIR, this comparison shows that these galaxies have PAH/IR-luminosity ratios that are only somewhat lower than normal star-forming galaxies, with the exception of A1068 and possibly A2597. The PAHs in A1835 are about four times brighter compared to L₂₄ than the nearly equally IR luminous A1068, so we detect significant intrinsic scatter in this ratio. There is no correlation in the ratio of PAH/IR luminosities to IR luminosity (see Figure 6(b)).

The 11.3 μm PAH luminosity is highly correlated with the 24 μm and the TIR luminosity, which makes sense if the dust and PAHs are heated by the same process. The 15 and 24 μm continuum luminosities are strongly correlated with each other (r = 0.96) and nearly linearly correlated (L₂₄∝L^{1.042 ± 0.006}₁₅), as expected since both quantities are usually produced by dust grains.

In contrast, 24 μm continuum and 11.3 μm PAH luminosities are not correlated with 6 μm continuum luminosity. In fact, when MS0735 is excluded, there is no correlation between 24 μm and 6 μm luminosities, r = 0.17. Similarly excluding MS0735, there is no correlation between the 6 μm and PAH flux at 11.3 μm (r = 0.23) or [Ne ii] (r = 0.004). (Including MS0735 in the tests increases the correlations to ∼0.6, under the 2σ threshold, but because the computed significance relies on the inclusion of a single source, it must be considered spurious.) The 6 μm light in these BCGs is produced primarily by old (cool) stars and therefore is a metric for the stellar mass. The lack of correlation between dust and stellar continua luminosities suggests that dust, PAH, and gas heating is not determined by cool stars. The systems where the dust luminosity well exceeds the 6 μm luminosity from stars, A1068 and A1835, exhibit higher PAH 7.7–11.3 μm ratios, consistent with the hypothesis that these are like starburst galaxies with levels of PAH ionization similar to those seen in starbursts (Figure 7). We will consider these ratios more fully in Section 5.4.

Zoom In Zoom Out Reset image size

For the rest of this discussion, we will assume that the long-wavelength IR continuum is powered primarily by obscured recent star formation. However, even though dust heating by evolved stars does not seem to dominate these systems at 24–70 μm, evolved stars may be the dominant source of heat for cooler dust emitting at longer wavelengths, and this dust therefore could contribute to emission at shorter wavelengths (see Section 4). Furthermore, processes of interest such as suprathermal electron heating and weak AGN may also supply energy to these systems. Since the observed global quantities are galaxy-wide averages, they suffer from the same interpretation ambiguity as high redshift, unresolved sources. Even if we could interpret these spectra in the context of star formation alone, it is impossible to unambiguously distinguish between a star formation episode of a single age and stellar mass and a time-averaged star formation history of "constant" star formation. Physically, signatures of star formation in normal galaxies include cold, dusty molecular gas, excess UV continuum, Hα, PAH emission, and infrared dust emission. We will discuss these data in a framework where obscured star formation is tracked by the infrared and PAH emission. However, we will show that star formation alone is inadequate to explain the full set of infrared spectral features in these systems.

The luminosities of forbidden lines of neon, which are channels for radiative cooling, are sensitive to the thermal energy input into the ionized gas. They therefore have also been shown to be good tracers of SFRs in normal star-forming galaxies. Ho & Keto (2007) showed that the sum of the fine structure lines of Ne ii (12.8 μm) and Ne iii (15.6 μm) correlates strongly with IR luminosity in normal star-forming galaxies over five orders of magnitude in luminosity. The sums of [Ne ii] and [Ne iii] luminosities in our BCG sample also correlate strongly with L₂₄ (r = 0.95, fluxes correlate with r = 0.90). [Ne ii] alone is just as correlated (r = 0.94, fluxes correlate with r = 0.91). The relationship, however, deviates even more from linearity than the PAH–IR-luminosity relationship, with $L([{\rm Ne\,{\mathsc{ii}}}])$ scaling as L^{0.58 ± 0.03}₂₄ (or L^{0.79 ± 0.04}_TIR). We suspect that while the dust and the PAHs are heated primarily by star formation, this lack of linearity in the [Ne ii]-IR correlation suggests that star formation may not be the sole process producing [Ne ii] emission.

The ratio of [Ne ii] to total infrared luminosity (L_TIR (W); Table 6) decreases somewhat with increasing IR luminosity (Figure 8). The Ho & Keto (2007) mean relationship between [Ne ii] and L_TIR for star-forming galaxies is $\log ({\rm [{\rm Ne\,{\mathsc{ii}}}]/L_{\rm TIR}}) = {-3.44 \pm 0.56}$, nearly independent of L_TIR. Excluding MS0735, the [Ne ii] luminosity in BCGs is about 1.6–12 times higher than the mean [Ne ii] luminosities of normal star-forming galaxies of similar infrared luminosities. The largest differences are found for the BCGs with lower IR luminosities (<10¹⁰ L_☉). MS0735+74, for which there is only an upper limit continuum estimate, is particularly bright in [Ne ii] compared to its infrared luminosity (≳ 0.009), >25 × the mean. The observed scatter of this ratio for normal star-forming galaxies in Ho & Keto (2007) is large, ±0.5 dex; nevertheless, the BCG ratios sit consistently on the high side of the scatter for normal star-forming galaxies, indicating that another process beyond star formation is also contributing to the heating of the ionized gas, particularly in the low-luminosity systems. In summary, the [Ne ii] luminosities seen in the low IR–luminosity BCGs exceed what would be expected from a star-forming galaxy with the same IR luminosity, but the two quantities are strongly correlated.

Zoom In Zoom Out Reset image size

The [Ne iii]/[Ne ii] ratio is not at all correlated with the mid-IR luminosity (Figure 9). It is possible that the [Ne iii]/[Ne ii] ratio indicates an approximate starburst age, with BCGs having the highest [Ne iii]/[Ne ii] also having the youngest starburst populations, but not necessarily the largest numbers of young stars (Thornley et al. 2000; Rigby & Rieke 2004; Snijders et al. 2007).

Zoom In Zoom Out Reset image size

Extremely luminous pure rotational H₂ lines, usually S(3) 9.67 μm, S(2) 12.28 μm, and S(1) 17.04 μm, are detected in all nine galaxies. S(0) 28.22 μm was not detected in any of these sources. Rotational transitions from S(5) to S(7) (5.51 μm) are seen in a majority of these spectra. While the luminosities of rotationally excited molecular hydrogen lines are correlated with IR luminosities of star-forming galaxies (Treyer et al. 2010) that is certainly not the case with our BCG sample. The line luminosities from rotational molecular hydrogen transitions from these BCGs are much greater than expected from the level of star formation heating the warm dust. H₂ emission is also uncorrelated with the continuum at 15 or 24 μm, r = 0.4–0.5.

Rotational emission from molecular hydrogen is commonly detected in ULIRGs (Higdon et al. 2006) and in star-forming normal galaxies (Roussel et al. 2007). In such galaxies, the luminosities of these lines are only about 4 × 10⁻⁴ of the total infrared power between 8 and 1000 μm. However, the ratio of H₂ luminosity to L₂₄ for the BCGs in this sample, ranging from 0.004 to 0.3, is about 5–100 times more than one would expect from a PDR. The most extreme object is MS0735, owing to its faint (and uncertain) IR continuum. The large H₂ luminosity from off-nuclear regions in the BCG NGC 1275 (Johnstone et al. 2007) led Ferland et al. (2008, 2009) to propose that much of the H₂ luminosity in BCGs located in X-ray cool-core clusters can be generated by cosmic ray heating or by non-radiative processes such as plasma waves.

BCGs are not the only galaxies to exhibit unusually large luminosities of rotational molecular hydrogen. Ogle et al. (2007) find the FR II radio galaxies have strong H₂ lines, but these galaxies are dissimilar to the BCGs in our sample. For example, 3C 326 exhibits high-ionization [Ne v] and [O iv] emission, indicating AGNs or LINER-like lines, very tiny SFRs (<0.1 M_☉ yr⁻¹), and the H₂ line transitions are primarily S(0) and S(1), indicative of cooler molecular gas than in our sample. These transitions are also seen in IRS mapping of the nearby group of galaxies, Stephan's Quintet, which exhibits bright H₂ (Cluver et al. 2010). Similarly, H₂ S(0) and S(1) emission lines have been reported from IRS mapping of edge-on spiral galaxies (Laine et al. 2010). An archival study of ULIRGs by Zakamska (2010) suggests that their H₂ emission is not associated with star formation. While many of these studies speculate that shocks might be a source of energy (e.g., Ogle et al. 2010) and might be quite common, the unifying thread to all of their discussions is that the molecular hydrogen rotational lines are surprisingly bright and their source of energy is still unidentified. The situation is not much different here, except the BCGs tend to also exhibit rotational lines characteristic of warmer molecular gas than the groups or radio galaxies (S(2), S(3), and S(7)).

The mid-IR luminosity is not significantly correlated with the summed luminosity of the molecular hydrogen lines, here represented by the sum of S(2) and S(3) lines, which were reliably detected in all nine systems (Figure 10(a)). While we showed in Section 5.2 that [Ne ii] emission is correlated with dust continuum emission, here we see that dust continuum is not significantly correlated with molecular hydrogen emission (for fluxes, r = 0.5; for luminosities r = 0.68). We plot the ratio of H₂ sum to mid-IR luminosity (Figure 10(b)). These ratios decrease for the systems with the highest mid-IR luminosities. This trend appears because the H₂ luminosities are limited in range (factor of 20) while the IR luminosity spans a large range (>1000). We interpret the trend to mean that the H₂ heat source is more important and in fact dominates the IR emission features in systems with low mid-IR luminosities.

Zoom In Zoom Out Reset image size

The rather insignificant correlation between H₂ and L₂₄ utterly vanishes once MS0735 is omitted from the sample. A similar effect happens when [Ne ii] flux is compared with H₂ flux: dropping MS0735 from the sample causes a very weak (less than 2σ) correlation to completely vanish. On the other hand, the luminosity of [Ne ii] is correlated with the luminosity of the H₂ S(2) + S(3) lines (r = 0.92). The presence or absence of MS0735 has little effect on the inferred strong correlations between Ne ii, PAHs, and mid-IR continuum flux and luminosity correlations. This correlation analysis suggests that while there is some relationship between the heat sources for the ionized gas and the dust, there appears to be a much weaker relationship between the heating sources for the molecular hydrogen and the dust. The ratio of molecular hydrogen to IR luminosity (Figure 10(b)) decreases with increasing IR luminosity, however, and suggests that the H₂ heating process becomes less important to the total luminosity budget as star formation increases.

The strong luminosity correlation between [Ne ii] and H₂ required further investigation, since the lack of correlations in the flux quantities suggested the luminosity correlation may be a simple "bigger is bigger" luminosity–luminosity comparison. (See Kennicutt 1990 for an infamous description of this type of error, involving a cigar.) Intriguingly, we find that the ratio of [Ne ii] to L₂₄ correlates even more strongly with the ratio of H₂ S(2)+S(3) summed luminosity to L₂₄ (r = 0.98). The best power-law fit to this relationship is H₂/IR ∼ ([Ne ii]/IR)^{1.49 ± 0.12} (Figure 11). The correlation of these ratios suggests that whatever process heats the molecular hydrogen is likely to be the culprit that boosts the forbidden line luminosity (heating the ionized gas) as well.

Zoom In Zoom Out Reset image size

To explore this idea further, we investigated how much more luminous the neon and molecular hydrogen lines are than one would expect from a star-forming galaxy, if the dust luminosity were a reliable indicator of the level of star formation. In Figure 12, we estimated the SFRs inferred from the IR continuum (the best fits to the Groves et al. 2008 models, which are consistent with Kennicutt 1998 estimates), from the [Ne ii]+[Ne iii] luminosity (Ho & Keto 2007), and from the H₂ luminosity (Treyer et al. 2010). For the latter estimate, we assumed that the Ne⁺/Ne = 0.75 and Ne⁺⁺/Ne = 0.15. The H₂-based SFR in Treyer et al. (2010) relies on the sum of the S(0), S(1), and S(2) transitions, which were not all detected in our systems. The sums plotted are based only on luminosities of the detected lines. These plots show that for most of the BCGs, while [Ne ii] is moderately overluminous for the inferred IR-based SFR (a factor of 2–5 above the upper end of the scatter exhibited by the galaxy sample of Ho & Keto (2007), and a factor of ∼10 over the mean), the H₂ luminosity is a factor of 5–15 overluminous based on the IR-based SFRs. The BCGs in A1068, A1835, and Hydra A have ratios typical of starbursts.

Zoom In Zoom Out Reset image size

It is interesting that the best fit for the points in Figure 11 is nearly linear. The slightly steeper than linear fit might be explained in the context of heating by suprathermal particles if the luminosity of the ionized gas ([Ne ii]) is limited by the finite column density of ionized gas, while the luminosity of the rotational line emission from the molecular gas is limited by the penetration depth of the suprathermal particles into the molecular gas, not the total column density of molecular hydrogen.

We defer a full discussion of the excitation analysis of the individual molecular hydrogen lines to a future paper. The more approximate dual temperature fit that we have done here, however, shows very similar trends to those seen in NGC 1275 filaments: the H₂ rotational line intensities cannot be fit by a single temperature. This trend is consistent with any model with a non-radiative energy source (Ferland et al. 2008).

In summary, as we examined correlations of continuum, PAH, and emission lines of [Ne ii] and H₂, and compared them to correlations and infrared line ratios in other types of galaxies, it emerged that a single heating process cannot explain the range of infrared properties we see in these BCGs. Star formation seems to play a role, albeit with varying levels of dominance, in producing the emission from these systems, but other processes unrelated to star formation must also contribute, particularly in systems with apparently low rates of star formation but high fluxes from rotationally excited transitions of molecular hydrogen.

If the dust in these BCGs spent much time in contact with the hot, X-ray emitting gas (or more generally, suprathermal electrons), one might expect the dust properties, such as its size distribution or ionization fraction, to be different from dust that has not undergone such a traumatic experience. PAH survival alone is problematic if suprathermal particles alone provide heat: radiation and collisions make PAH lifetimes in the harsh environment of the center of a cool-core cluster of galaxies quite short. Using order of magnitude cross sections from Voit (1992a), and 0.5 keV photon fluxes of about 10⁶ cm⁻² s⁻¹, we estimate lifetimes of order one million years. The damage from particle collisions may be even more dire. From the analysis of Micelotta et al. (2010), the lifetime of PAH molecules embedded in ∼1 keV gas with a density of ∼0.1 cm⁻³ is limited to hundreds of years by collisions with the hot electrons and ions. Any processing along these lines causes the PAHs and small grains to evaporate preferentially compared to large grains. The presence of PAHs requires the dusty gas to be shielded from the hot gas and its radiation.

As in other galaxies observed with the Spitzer IRS (Smith et al. 2007b; Kaneda et al. 2008), the fluxes and luminosities of the PAH complex at 7.7 μm and at 11.3 μm are strongly and linearly correlated. The 7.7 and 11.3 μm PAH complex luminosities are strongly correlated (r = 0.98, fluxes at r = 0.94) for all seven systems in which both are detected. 2A0335 and MS0735 lack PAH 7.7 μm detections.

The ratio of 7.7–11.3 μm PAH complexes is relatively insensitive to the sizes of the PAHs (Schutte et al. 1993), but sensitive to the ratio of ionized to neutral PAHs (Allamandola et al. 1989; Draine & Li 2007). For example, the ratio of the PAH complex at 7.7 to the PAH complex at 11.3 μm is lower in galaxy centers that are AGN-dominated compared with those which are H ii dominated (Smith et al. 2007b). The mean range for seven ratios of the PAH complexes at 7.7 and 11.3 μm in our sample is 2.7 ± 0.2 (Figure 13). The mean sample ratio is intermediate between that of H ii/starburst galaxies (4.2; Smith et al. 2007b) and diffuse Galactic emission (2–3.3; Sakon et al. 2004). The BCGs with classic indicators of starburst activity (A1068 and A1835) have ratios typical of starbursts; the others are closer to that of Galactic interstellar medium (ISM). In elliptical galaxies, the PAH 7.7/11.3 ratio is unusually weak (∼1–2; Kaneda et al. 2008); the lowest ratios seen in dusty ellipticals are lower than the ratios detected in our BCG sample. For BCGs with a low PAH 7.7 μm to 11.3 μm ratio, the incident spectrum on the PAHs may be dominated by evolved stars, compared to a harder, ionizing spectrum with contributions from young massive stars. Our result suggests that the radiation fields in BCGs are intermediate in their hardness between those of dusty elliptical galaxies and star-forming galaxies, which is consistent with what one might expect in the ISM of galaxies with enormous old stellar populations, together with small numbers of recently formed stars.

Zoom In Zoom Out Reset image size

The ratio of PAH complexes at 17 and 11.3 μm is thought to be regulated by PAH sizes with large PAHs contributing more to the longer wavelength complex (e.g., Draine & Li 2007). The broad PAH feature at 17 μm was detected with confidence in four BCGs (2A0335, A1068, Hydra A, and PKS0745), while the others have 3σ upper limits (Figure 14). With the single exception of A1068, the detections and upper limits are consistent with the ratio of PAH 17 μm to PAH 11.3 μm fluxes typical of normal star-forming galaxies (∼0.5; Smith et al. 2007b). The consistency of the observed ratios and limits in nearly all of these BCGs compared with those with normal galaxies suggests that the PAH size distribution may be normal. The exception of A1068, with a ratio >2, indicates that it may have deficit of small PAHs compared to large PAHs. This BCG also has a smaller PAH to L_TIR ratio. In the BCG of A1068, PAHs, and particularly the small PAHs, may have been preferentially destroyed by collisions with particles or photons.

Zoom In Zoom Out Reset image size

Since both [Ne iii]/[Ne ii] and the ratio of PAH 7.7 μm to 11.3 μm flux are related to the hardness of radiation, we compared these quantities to see whether they are correlated (Figure 15) and whether the BCG points cover similar parameter space as other galaxies. We see no correlation between the PAH ratios and the neon forbidden line ratios, but that is not surprising since the gas producing the forbidden lines is different from the gas hosting the PAHs; similarly, the photons ionizing neon are not the same photons setting the ionization level of PAHs. As shown in Figure 15, the seven BCGs with detections in all four quantities exhibit ratios rather similar to SINGS galaxies (Smith et al. 2007b). There are spiral galaxies in the SINGS sample in Figure 15 with very high [Ne iii]/[Ne ii] ratios (>10); the explanation may be that these galaxies (about 15% of the SINGS sample) are dominated by very recent star formation, and therefore hotter O stars (Thornley et al. 2000; Rigby & Rieke 2004; Snijders et al. 2007), compared to the BCGs. LINERs in the Smith et al. (2007b) sample do not exhibit [Ne iii]/[Ne ii] <0.7–0.8 together with low PAH 7.7/11.3 μm fractions (<2), so a few of the BCGs in our sample with low [Ne iii]/[Ne ii] ratios (A1795, A478, and PKS0745) also have lower PAH 7.7/11.3 μm ratios than seen in the SINGS sample, more similar to those of dusty ellipticals (Kaneda et al. 2008) or diffuse Galactic ISM (Sakon et al. 2004). The two systems with the highest IR and PAH luminosities of the sample (A1068 and A1835) also exhibit the largest PAH 7.7/11.3 μm ratios, indicating that the PAHs in the most luminous BCG systems are more ionized than those in the other BCGs.

Zoom In Zoom Out Reset image size

We arrive at the following conclusions about the PAHs and dust in these systems.

• 1.  
  The presence of PAHs, and the similarity of the PAH emission ratios to those in star-forming galaxies, means that these tiny grains must be protected from the ICM and shocks. PAHs are easily destroyed by ionizing UV and X-rays, by collisions with hot thermal particles, and by shocks.
• 2.  
  The emissivities of the PAHs and the ionized gas are correlated with the mid-IR continuum emitted by dust grains. Therefore, there is a common source of heat for these components, consistent with being star formation. However, the excess luminosity of [Ne ii] and H₂ in the less luminous systems suggests that another component is contributing to the heating of the ionized gas in addition to star formation.
• 3.  
  The emissivity of the PAHs is not correlated with the stellar (6 μm) IR continuum. Correlation is expected if the stars were the main agent heating the dust.

We speculate that the evolved stars in the BCGs are the main production source of dust. The dust, except in a couple of cases, shows little indication that it may have been processed by hot, X-ray emitting plasma, or the AGN. The source of the cool BCG gas may be the hot ICM, but it seems unlikely that the dust came with the gas. Dust, however, is an extremely effective coolant, and if mixed with the hot gas, could precipitate the cooling necessary to fuel the formation of dusty molecular star-forming clouds.

Synchrotron radiation from a jet could contribute some of the infrared continuum. Cleary et al. (2007) find in their survey of an unbiased set of 3CR and quasars that only the quasars have a nonthermal contribution at 15 μm of >20%. About half of the 15 μm emission from nucleus of M87 is synchrotron emission, based on analysis of its IRS spectrum (Buson et al. 2009). Of the nine sources in our sample, only Hydra A has a radio synchrotron source that is luminous enough to contribute more than ∼10% to the IRS continuum at 24 μm, if its power law extends unbroken into the infrared.

High-ionization lines in the infrared could indicate a buried AGN or the presence of very hot stars. [Ne v] 14.3 μm and 24.3 μm, with an ionization potential of 97.1 eV, would be an unambiguous AGN indicator (Voit 1992b). We did not detect this emission line in any of our sources at a level ≳ 10⁻¹⁷ W m⁻²), although a weak [Ne v]14.3 μm feature is detected in A1835 and Hydra A, and a suspicious (possibly spurious) feature appears near 24 μm in the spectrum of A1068, which is known to have some AGN contribution in the infrared (Quillen et al. 2008). The lack of [Ne v] does not rule out some AGN contribution, since this line can be faint compared to [Ne ii]. [O iv] at 25.9 μm would also be an unambiguous sign of AGN photoionization, since O iii has an ionization potential just above that of He i (to He ii), 54.4 eV, and is therefore rare in regions ionized by stars. However, [O iv] 25.9 μm is blended with [Fe ii] at 26 μm. We detected this blend unambiguously only in PKS0745-19 and Hydra A. Puzzlingly (if this blend is an indication of the presence of O iv), the [Ne iii]/[Ne ii] ratio in PKS0745-19 is one of the lowest in the sample. The [O iv]–[Fe ii] blend does not appear at all in the spectrum of A1068, however, nor does [Ne v] 14.3 μm. The lack of these features casts further doubt on the candidate [Ne v] feature in that spectrum.

The presence of high-ionization lines such as [S iv] (10.51 μm) and [Ne iii] indicates the presence of young hot stars in these galaxies. The ionization parameters consistent with the [Ne iii]/[Ne ii] ratios in the ionized gas are similar to those seen in star-forming galaxies. The energy required to ionize neon once to Ne ii is 21.6 eV, compared to 41 eV for reaching Ne iii. [Ne ii] was easily detected in every galaxy in our sample, and only MS0735+74 lacked detectable [Ne iii]. Although S iii to S iv has an ionization potential (34.8 eV) similar to Ne ii to Ne iii, it has a very low photoionization cross section. So a significant detection of [S iv] not only indicates hot stars, but also a high density of them. In contrast to [Ne iii], [S iv] was not detected in any of the systems. PAHFIT results for 2A0335+096, A1835, A1068, and Hydra A indicate faint formal detections at ∼3σ–7σ, but inspection of the fits and data causes us to regard these detections as extremely marginal.

Based on the lack of [Ne v] emission lines from the gas, we have no conclusive evidence in favor of the gas being photoionized by an AGN, consistent with conclusions based on spatially resolved optical emission-line studies of these and similar galaxies (e.g., Heckman et al. 1989).

The detection of spatially extended UV continuum in a number of cases is unambiguous evidence of the importance of recent star formation over AGN contributions in this sample of BCGs (McNamara & O'Connell 1993; Martel et al. 2002; O'Dea et al. 2004; Hicks & Mushotzky 2005; Hicks et al. 2010; Donahue et al. 2010). Since lack of evidence is not the same as evidence of lack, we keep in mind that some of the spectra may have contributions from a low-luminosity AGN since none of these spectra exclude the nucleus. However, Occam's Razor prefers the simplest explanation for heating these systems, and as such, no need for AGN excitation or heating is required by these observations.