HUBBLE SPACE TELESCOPE FAR-ULTRAVIOLET OBSERVATIONS OF BRIGHTEST CLUSTER GALAXIES: THE ROLE OF STAR FORMATION IN COOLING FLOWS AND BCG EVOLUTION

Observations were obtained with the solar blind channel (SBC) MAMA detector of the Advanced Camera for Surveys (ACS; Clampin et al. 2004) on the HST during cycle 11 (program 11230, PI: K. P. O'Dea). Each galaxy was observed in two long pass filters, the one containing the Lyα line, the other redward of this line to measure the continuum. The F140LP filter containing the Lyα line was used for all galaxies except the nearer BCGs, A11 and A1664, which were observed using the F125LP filter. The continuum filter chosen was F140LP for objects with redshift z < 0.11, F150LP for objects with redshift 0.11 < z < 0.19 (ZWCL8193, A11, and A1664) and the F165LP filter for the remaining objects with 0.19 < z < 0.31 (A1835, ZWCL348, RXJ 2129+00, and ZWCL3146). Observations were obtained using a three point position dither. The exposure time in each filter was 1170 s so that the observations in the two filters were approximately one HST orbit per galaxy. Observations were taken between 2008 March and 2009 February. The long pass filters, F125LP, F140LP, F150LP, and F165LP, have pivot wavelengths of 1438, 1527, 1611, and 1758 Å, respectively, and similar maximum wavelengths of 2000 Å but minimum or cutoff wavelengths of 1250, 1370, 1470, and 1650 Å, respectively. The pixel scale for the SBC is approximately 0034 × 0030 pixel⁻¹. The camera field of view is 346 × 308. These FUV observations are summarized in Table 1.

The ACS/SBC images were reduced with the ACS calibration pipeline producing calibrated drizzled images. Continuum images were shifted to the position of the line images and subtracted from the line images after multiplication by an adjusted corrective factor larger than 1 to take into account the additional continuum photons present in the line images. Our procedure was to increase the correction factor until regions of the image became negative. FUV and continuum subtracted Lyα images are shown in Figures 1–7.

At the same time optical images were observed with the WPFC2 camera on board HST using the broad band F606W filter for A1664 and ZWCL 8193. Visible broad band images observed with WFPC2 were available from the Hubble Legacy Archive for the remaining galaxies in either the F702W filter (A1835) or the F606W filter (ZWCL 348, ZWCL 3146, A11, and RXJ 2129+01). The broad band optical images are shown for comparison in Figures 1–7.

We have overlaid 3 μm continuum observations as contours in Figures 1–7 on the FUV continuum images. These images were taken with the IRAC camera on the Spitzer Space Telescope and are described by Egami et al. (2006) and Quillen et al. (2008). We find that the FUV and Lyα emission are located near the center of the BCGs as seen at 3 μm.

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

Zoom In Zoom Out Reset image size

Chandra X-ray Observatory observations with the Advanced CCD Imaging Spectrometer (ACIS) were available from the archive for four of the galaxies: A1664, A1835, ZWCL 3146, and RXJ 2129+00. Exposure times were 11, 22, 49, and 12 ks, respectively. The event files were binned to 1'' pixels and the resulting images smoothed with the ciao adaptive smoothing routine csmooth using the algorithm by Ebeling et al. (2006). Constant surface brightness X-ray contours are shown for these four galaxies in Figures 2, 3, 5, and 7 overlaid on the continuum subtracted Lyα images.

Where available, we selected high resolution Very Large Array (VLA) observations from the NRAO archive. For some sources we chose an additional data set in order to obtain a complementary lower resolution image. The NRAO AIPS package was used for the calibration, imaging, self-calibration, and deconvolution. The properties of the final images are given in Table 2. We detected a faint point source in all the BCGs. The flux densities of the point sources are given in Table 4. These high resolution observations are not sensitive to very diffuse emission.

Figures 1–7 have been centered at the location of the central radio sources as measured from VLA archival data at 5 or 8.5 GHz (with positions listed in Table 4). Coordinate errors measured from HST, Spitzer Space Telescope, and ACIS Chandra X-ray Observatory observations are of order an arcsecond. The FUV and Lyα images lack point sources that could be used to register the images at sub-arcsecond scales.

We find that all seven galaxies observed display extended emission in both FUV continuum and Lyα emission. The FUV continuum is patchy, as was true for A1795 and A2597 (O'Dea et al. 2004). As discussed in that work, the FUV continuum is likely associated with young stars in star clusters. The Lyα morphology contains both clumps and a more diffuse or filamentary component. The diffuse component is seen in Lyα but not in the FUV continuum, e.g., ZWCL 8193 (Figure 8). Diffuse or filamentary Lyα was also seen by O'Dea et al. (2004) in A1795 and A2597. The association between the Lyα and FUV continuum implies that the FUV continuum contributes to the ionization of the Lyα emitting gas.

Zoom In Zoom Out Reset image size

All of our BCGs display asymmetry in the FUV emission. In A11, the FUV continuum and Lyα emission are arranged in an extended clump cospatial with the visible nucleus, with a more diffuse component about 2'' west of the nucleus (also seen in the optical image). The main clump of emission is slightly offset from the center of the Spitzer IRAC 3 μm isophotes (see Figure 1). In A1664, three large clumps of FUV and Lyα trace the disturbed morphology of the host galaxy as observed in the optical. Additionally, there is a low surface brightness filament of Lyα emission extending ∼25 kpc to the south of the three bright clumps. This filament is not associated with any optical counterpart in the WFPC2 image or any FUV continuum emission. The 3 μm peak, cospatial with the galaxy's nucleus, is also cospatial with the dust lanes in the optical image (see Figure 2). A1835 has also been observed by Bildfell et al. (2008) who measure a size for the blue star-forming region of 19 ± 2 kpc, which is in good agreement with our measurement of ∼17 kpc for the size of the Lyα emission (Table 3) though McNamara et al. (2006) find blue colors on even larger scales. In A1835, the 3 μm contours are also not centered on the brightest regions seen in the FUV, Lyα, or visible band images (see Figure 3). For ZWCL 348, the visible and 3 μm emission peaks are nearly centered and the FUV emission peaks on the center of the galaxy. However, the visible band image shows that the galaxy is disturbed and the outer contours seen at 3 μm are not round (see Figure 4). The Lyα emission extends eastward from the nucleus much further than to the west. In ZWCL 3146, the FUV and Lyα emission are centered on the 3 μm contours (see Figure 5). For ZWCL 8193, there is a nuclear bulge in the optical and 3 μm images. However, FUV and Lyα emission are brighter north of the nucleus and have a spiral shape suggesting that a smaller galaxy has been recently disrupted in the outskirts of the BCG (see Figure 6). The host galaxy is an elliptical in a rich environment with several nearby dwarf satellites. The disturbed morphology of the host galaxy is suggestive of a recent or ongoing series of minor mergers. RXJ 2129+00 displays Lyα emission which extends on only the northeastern side of the galaxy. The offset between radio and Lyα peaks is small and so may be due to a registration error in the HST image (see Figure 7).

3.1.1. "Clumpiness" of FUV emission

We find that most of these BCGs display strong asymmetries or uneven distributions in their star formation as seen from the FUV continuum images (see Figure 9). For these galaxies, 1'' corresponds to 2–4 kpc (see Table 1) thus these asymmetries are on a scale of order 10–50 kpc. On smaller scales, the FUV morphology is generally more clumpy and filamentary than the associated Lyα component. For the purposes of this paper, we qualitatively define a "clump" as a compact region of emission a factor of ∼2 brighter than the surrounding lower surface brightness diffuse component. A11, 1664, ZWCL 8193, and RXJ 2129+00 may be described as "clump dominated" in which the majority (>50%) of the FUV flux is associated with compact (<2 kpc) bright clumps. For example, the majority of FUV emission in A11 is associated with three bright clumps, the largest of which (the northern-most clump) extends ∼0.6 kpc. The three bright clumps together contribute > ∼ 60% of the total FUV flux from the source. The FUV morphology of RXJ 2129+00 is almost entirely associated with three small (∼0.5 kpc) clumps and appears to lack a diffuse component. For A1835, ZWCL 348, and ZWCL 3146, the distinction between clumpy FUV emission and the diffuse component is less clear, and the flux seems to more gradually peak toward the center than, for example, ZWCL 8193. Note also that these galaxies are vastly more symmetric on >20 kpc scales than are the "clump-dominated" BCGs.

Zoom In Zoom Out Reset image size

The high star formation rates of the galaxies studied here compared to others in the ROSAT BCG sample may be related to the large scale X-ray structure. We will discuss this in more detail later. Comparisons between Lyα and existing Hα images (not shown) suggest that there are large variations in emission line ratio. This can be explained either by patchy extinction or with shocks, affecting the intrinsic line ratios.

McNamara (1997) classified the morphology of star-forming regions in cooling core BCGs (studied using mainly ground-based optical observations) into four classes—point, disk, lobe, and amorphous. McNamara (1997) noted that the disks were very rare and that the amorphous morphologies are the most common. All of the seven BCGs studied here fall in the amorphous class. Why are disks so rare? We note that the star formation occurs over a large spatial scale (7–28 kpc). One possibility is that the stars form before the gas can collapse into a disk. In addition, studies of the kinematics of the optical emission-line nebulae in cool core BCGs find that the gas motions are mostly turbulent with very little organized rotation (e.g., Heckman et al. 1989; Baum et al. 1992; Wilman et al. 2006, 2009). Thus, the lack of star-forming disks may reflect the lack of systematic rotation in the star-forming gas.

3.1.2. Radio Emission

3.1.3. X-ray Morphology

3.1.4. Comparison to CO and Hα observations

Pipino et al. (2009) used optical and NUV colors of BCGs to demonstrate that the UV light is produced by a young rather than old stellar population. We use the FUV continuum flux to estimate star formation rates in these galaxies. The continuum flux was first corrected for Galactic extinction. Extinction correction was done using Galactic extinction at the position of each BCG and the extinction law by Cardelli et al. (1989; as done in Table 5 by O'Dea et al. 2004 for A1795 and A2597). We then compared the count rate predicted for the observed filter by the STSDAS synthetic photometry package synphot for a spectrum produced by starburst99.⁶ (Leitherer et al. 1999; Vazquez & Leitherer 2005).

The UV continuum estimated star formation rates are listed in Table 5. We compare the UV continuum estimated star formation with those based on the limited aperture Hα fluxes by Crawford et al. (1999) (using 13 wide slit) or spectroscopic measurements from the Sloan Digital Sky Survey (SDSS) archive (using a 3'' diameter fiber) and those estimated from infrared observations with the Spitzer Space Telescope by Egami et al. (2006), Quillen et al. (2008), and O'Dea et al. (2008). Neither the UV continuum estimated nor Hα estimated star formation rates have been corrected for internal extinction. This table also lists molecular gas masses by Edge (2001) and Salome & Combes (2003).

Balmer decrements are available for most of the galaxies considered here and range from ∼3–5 (Crawford et al. 1999), see Table 5. Assuming an intrinsic Hα/Hβ theoretical line ratio of 2.86 ("case B" recombination; Osterbrock 1989), the observed Balmer decrement allows us to estimate the color excess associated with the internal extinction in the source, following the parameterization of Cardelli et al. (1989):

$$\begin{equation} E\left(B-V\right)_{\mathrm{H}\alpha / \mathrm{H}\beta } = \frac{2.5\times \log \left(2.86 / R_{\mathrm{obs}}\right)}{k\left(\lambda _\alpha \right) - k\left(\lambda _\beta \right)}, \end{equation} \tag{ 1 }$$

where R_obs = F(Hα)/F(Hβ) is the observed flux ratio, and the extinction curves at Hα and Hβ wavelengths are k(λ_α) ≈ 2.63 and k(λ_β) ≈ 3.71, respectively, as given by Cardelli et al. (1989). One caveat is that if processes other than recombination (e.g., shocks, cosmic ray heating) contribute to Hα, the intrinsic Hα/Hβ ratio would be higher than the theoretical "case B" value and the estimated extinctions would be upper limits. With this possibility in mind, the upper-limit intrinsic optical extinction is E(B − V) ∼ 0.6 for a Balmer decrement of 5, corresponding to an extinction in the FUV of A_FUV ∼ 5.5 and therefore a correction factor of 160 to the measured flux.

We note from Table 5 that IR star formation rates exceed those estimated in Hα and these exceed those estimated from the FUV continuum. This would be consistent with patchy but significant levels of extinction. The correction factor estimated for the UV photometry of 160 for a Balmer decrement of 5 is vastly higher than that required to make up the deficit of star formation rates estimated between the UV and (e.g.,) the IR. Large levels of extinction are also likely because of the high molecular gas content in these galaxies and the dust lanes seen in the optical images. Previous comparisons by Hu (1992) between Hα and Lyα flux suggested modest extinctions intrinsic to the cluster of order E(B − V) ≈ 0.09–0.25. This was done assuming an unabsorbed Lyα/Hα ratio of 13 for photoionization and collision models (Ferland & Osterbrock 1986). However, the BCGs considered by Hu (1992) were not chosen via their Hα luminosity. As Hα luminosity is correlated with both molecular gas mass and infrared luminosity (O'Dea et al. 2008), it is perhaps not surprising that the sample considered here would have higher estimated internal extinctions than the same considered by Hu (1992).

In short, FUV estimated star formation rates range from ∼3 to ∼14 times lower than those estimated from the Spitzer observations. Balmer decrements and molecular gas observations suggest that internal extinction could be extremely high in some regions. The discrepancy between the estimated star formation rates and the internal extinction correction factor from the Balmer decrement suggests that internal extinction is patchy. As the infrared estimated star formation rate is least sensitive to extinction, it can be considered the most accurate, and suggests that about 90% of the FUV continuum has been absorbed. When taking patchy extinction into account, the discrepancy between star formation rates estimated at different wavelengths may be reconciled. Moreover, even when accounting for internal extinction, a one-to-one correspondence between star formation rates measured in the UV and IR are not necessarily expected, as the associated star formation indicators in those wavelength regimes may probe different aspects of the star-forming region (e.g., Johnson et al. 2007).

O'Dea et al. (2008) found that in gas-rich star-forming BCGs the nominal gas depletion time scale is about 1 Gyr. Since star formation is not highly efficient, it seems likely that 1 Gyr is an upper limit to the lifetime of the star formation. Pipino et al. (2009) suggest that the star formation in BCGs is relative recent with ages less than 200 Myr. Thus, a typical star formation rate of 50 M_☉ yr⁻¹ would result in a total mass of less than 10¹⁰M_☉ which is a small fraction of the total stellar mass in a BCG (e.g., von der Linden et al. 2007) as also noted by, e.g., Pipino et al. (2009).

O'Dea et al. (2008) noticed a discrepancy between the infrared estimated star formation rates and the size of the star-forming regions. Here, we have confirmed that the star-forming regions are not large and remain under 30 kpc. This puts the galaxies somewhat off the Kennicutt relation (Kennicutt 1998) but (as shown by O'Dea et al. 2008 in their Figure 8) only by a modest factor of a few. While we confirm the discrepancy, we find that it is small enough that it could be explained by other systematic effects such as an overestimate of the H₂ mass. The CO to H₂ conversion factor is suspected to be uncertain by a factor of a few (e.g., Israel et al. 2006). Thus, there may be no significant discrepancy between the size of the star-forming regions as measured in the FUV and that predicted with the Kennicutt relation.

The fact that the Lyα morphology closely traces that of the underlying FUV continuum emission suggests that the latter provides at least a significant fraction of ionizing photons for the former, and that the nebula is ionized locally. Here, we explore this argument more quantitatively, in considering whether the observed FUV continuum is consistent with a sufficient number of hot stars required to ionize the nebula. This question can be addressed using simple arguments outlined by Zanstra (1931) and employed in similar contexts by (e.g., Baum & Heckman 1989; O'Dea et al. 2004). We assume a case B recombination scenario (Osterbrock 1989) in which the medium is optically thin to Balmer photons but optically thick to Lyman photons, and that in 10⁴ K gas, ∼45% of all Balmer photons will emerge from the nebula as Hα photons. We further assume that all ionizing photons that are emitted by the stars are absorbed by the gas. This makes our estimate a rough lower limit, as in reality, a significant fraction of ionizing photons will escape, increasing the number of required stars.

Given these assumptions, the number of ionizing photons required to power the nebula is can be estimated via its observed Hα luminosity using the Zanstra (1931) method, which relates the emission-line luminosity to Q_tot, the total number of photons with energies greater than 13.6 eV needed, per second, to ionize the nebula:

$$\begin{equation} Q_{\mathrm{tot}}=\frac{2.2 L_{\mathrm{H}\alpha }}{h \nu _\alpha } \mathrm{photons} \mathrm{\, s}^{-1}, \end{equation} \tag{ 2 }$$

where L_Hα is the total nebular Hα luminosity, h is Planck's constant, ν_α is the frequency of the Hα line, and 2.2 is the inverse of the 0.45 Balmer photon to Hα photon ratio assumed above.

For each galaxy in our sample (save A11, for which we do not have an Hα luminosity), we have calculated this value and scaled it in terms of a required star formation rate as predicted by the same starburst99 models used to calculate the star formation rates based on the FUV continuum. We present these results in Table 6. In every case, the ratio of the observed star formation rate to that required by the Zanstra method is of order unity, suggesting that the FUV continuum provides a significant fraction of ionizing photons for the nebula. Note that it is likely that stellar photoionization is not the only source of energy for cooling flow nebula (e.g., Voit & Donahue 1997). This result is consistent with the notion that the nebula is ionized locally by the young stars embedded within it, as supported by the similar morphologies of the Lyα emission and FUV continuum emission. O'Dea et al. (2004) undertook a similar exercise in their study of the cool core clusters A1795 and A2597, finding that there were enough hot stars present to within a factor of a few. They also estimated the degree to which a hidden quasar ionizing continuum would contribute ionizing photons, finding that a modest AGN luminosity could contribute ∼10% of what was required.

Table 6. Can Hot Stars Ionize the Nebula?

Source	Hα Luminosity	Qrequired	SFR(Qrequired)	$\left(\frac{\mathrm{SFR}_\mathrm{FUV}}{\mathrm{SFR}\left[\mathrm{Q}_\mathrm{required}\right]}\right)$
	(erg s−1 cm−2)	(photons s−1)	($\dot{\mathrm{\it M}}_\odot$ yr−1)
(1)	(2)	(3)	(4)	(5)
A11	...	...	...	...
A1664	6.38 × 1041	4.64 × 1053	2.12	2.17
A1835	1.68 × 1042	1.22 × 1054	5.58	2.09
ZWCL 348	1.95 × 1042	1.41 × 1054	6.47	0.94
ZWCL 3146	3.29 × 1042	2.39 × 1054	1.10	1.13
ZWCL 8193	2.79 × 1042	2.03 × 1054	9.29	0.58
RXJ 2129+00	3.33 × 1040	2.42 × 1052	1.11	8.1

Notes. Results of a photon-counting exercise designed to determine whether there are sufficient hot stars present (giving rise to the underlying FUV continuum) to ionize the emission-line nebula. We have used the Zanstra (1931) method relating the Hα emission-line luminosity with the number of ionizing photons required to give rise to that luminosity, assuming all ionizing photons are absorbed in 10⁴ K gas obeying a case B recombination scenario. (1) Source name; (2) Hα luminosities in erg s⁻¹ cm⁻², from Crawford et al. (1999) excepting ZWCL 348, whose Hα luminosity is from the SDSS archive; (3) number of ionizing photons per second, required to ionize the nebula, as estimated using the Zanstra (1931) method described in Section 3.3; (4) star formation rate required to power the nebula, estimated by comparing Q_required with a starburst99 model calculating the number of photons with energies greater than 13.6 eV for a 10⁷ yr old starburst assuming a continuous star formation rate of 1 M_☉ yr⁻¹ and a initial mass function with an upper mass cutoff of 100 M_☉ and slope α = 2.35; (5) comparing the value calculated in Column 4 with the "observed" star formation rate as estimated by comparing our FUV photometry with the same starburst99 model. A ratio greater than 1 indicates that enough ionizing photons are present from the young stellar population to ionize the nebula. Obviously, there are significant uncertainties associated with these estimates. As such, this is meant only as an order-of-magnitude exercise. That the ratios are all of order unity indicates that the FUV emission due to the young stellar population is of sufficient strength to power the emission-line nebula.

Download table as:  ASCIITypeset image