THE STATE OF THE WARM AND COLD GAS IN THE EXTREME STARBURST AT THE CORE OF THE PHOENIX GALAXY CLUSTER (SPT-CLJ2344-4243)

Spectro-imaging observations of Phoenix A were obtained with the GMOS-S integral field unit (IFU) on 2012 November 19 and 2012 November 16 during dark time with photometric conditions and ∼06 V-band seeing (corresponding to a physical resolution of ∼4.0 kpc at z = 0.597, the redshift of our target). The observations employed the single slit mode, resulting in a field of view of 5'' ×7''. We used the R400 filter centered at 8626 Å which provides wavelength coverage from 5360–9600 Å at a spectral resolution of λ/Δλ ∼ 2000 (3.9 Å pixel⁻¹). To cover the galaxy, we used a small mosaic: a central pointing plus one pointing either side (separated by a full IFU width). The final field of view is therefore 9'' × 5''. The exposure time for the central pointing was 2.4 ks (split into two 1.2 ks exposures) whilst the two outer pointings each have 4.8 ks (split into four 1.2 ks exposures).

To reduce the data, we used the GMOS data reduction pipeline to extract and wavelength-calibrate the spectra of each IFU element. Variations in fiber-to-fiber response were removed using twilight flats, and the wavelength calibration employed a CuAr arc lamp. However, since the arc lamps were taken several days after the science observations, we applied a small zero-point correction to the wavelength solution in each fiber for each science exposure using the sky OH emission. Flux calibration was carried out using observations of the standard star LTT3218 using the same observational setup as the science observations. The collapsed one-dimensional (1D) spectrum of the full flux-calibrated data cube is shown in Figure 1.

Zoom In Zoom Out Reset image size

Each spatial pixel in the GMOS-S IFU cube corresponds to an independent spectrum covering 5360–9600 Å (rest-frame 3356–6011 Å). These single-pixel spectra were first fit with a single stellar population (SSP) in order to model the continuum emission and absorption lines. SSPs were generated using the publicly available Sed@ code¹² and made available by González Delgado et al. (2005). These models cover ranges of 0.001–0.04 in metallicity and 10⁶–10¹⁰ yr in age, and we apply a range of reddening models spanning 0 < E(B − V) < 3. We choose the combination of age, metallicity, and reddening that minimizes χ² over the full spectrum, with the normalization being a free parameter. The best-fit continuum spectrum was subtracted from each spectrum, leaving only emission lines. We note that this part of the analysis was performed in order to ensure accurate Balmer line fluxes (factoring in stellar absorption to the estimates of emission). However, we find that, given the strong Balmer emission in this system, we obtain very similar results whether we assume no Balmer absorption or perform a more complicated continuum modeling as described above.

Each of the resulting continuum-subtracted spectra was smoothed in wavelength space, with a smoothing scale of 8 spectral pixels, in order to improve the quality of emission-line fits. We performed simultaneous fits of the [O ii]λλ3726, 3729, [O iii]λλ4959, 5007, Hβ, Hγ, Hδ, [Ne iii]λ3869, and He ii λ4686 emission lines, requiring all emission lines to have identical redshifts and linewidths. This resulted in maps of emission-line fluxes and radial velocities as a function of position. Finally, we went back and fit the unsmoothed [O iii]λλ4959, 5007 lines, using the previously measured redshift and flux as priors, in order to determine the gas kinematics as a function of position. During this iteration, we allowed multiple kinematic components for each line. Measured velocity dispersions were corrected for the instrumental linewidth of 3.1 Å at the wavelength of the [O iii]λ5007 emission line.

The Hβ, Hγ, and Hδ lines were used to estimate the amount of reddening at each position. We assume case B recombination, using intrinsic values of Hγ/Hβ = 0.47 and Hδ/Hβ = 0.26 (Osterbrock 1989). The measured Balmer ratios, combined with the Cardelli et al. (1989) extinction curve, yield two independent maps of E(B − V), which have qualitatively similar magnitudes and morphologies. These two maps are averaged. The resulting map of E(B − V), shown in Figure 2, is used to correct all measured fluxes for intrinsic reddening, assuming a dust-screen model.

Zoom In Zoom Out Reset image size

We used the SMA¹³ (Ho et al. 2004) on Mauna Kea, Hawaii, to observe the core of the Phoenix cluster on 2013 August 13 and 2013 August 18. The five available array antennas were arranged in a compact configuration that gave baseline lengths from 6 to 68 m. The weather conditions were very good to excellent on these two dates, with τ_225 GHz of 0.10 and 0.05 measured by the atmospheric opacity monitor located at the nearby Caltech Submillimeter Observatory. The receivers were tuned to search for emission from the CO J = 3 − 2 line (rest frequency 345.796 GHz) at the cluster redshift of z = 0.597. The correlator was configured to process two intermediate frequency (IF) bands, each of width 1.968 GHz, centered at ±5 GHz and at ±7 GHz from an low oscillator (LO) frequency of 221.435 GHz. This setup resulted in velocity coverage of approximately −1240 to 4270 km s⁻¹ around z = 0.597 for the CO J = 3–2 line in the lower sideband (LSB) (aside from a small gap of about 32 MHz between the two IF bands). The channel spacing was 0.8125 MHz, corresponding to 1.13 km s⁻¹. The pointing center was chosen to be at the center of the BCG, $\alpha = 23^{{\rm h}}44^{{\rm m}}43\buildrel{\mathrm{s}}\over{.}96$, δ = −42°43'122 (J2000), and the ∼58'' (FWHM) field of view was set by the primary beam size of the individual array antennas. The basic observing loop comprised 2.5 minutes on J2258−279, 2.5 minutes on J2248−325, and 7.5 minutes on the Phoenix. The absolute flux scale was set with an accuracy of ∼10% using observations of Uranus at the start of each track. The passband shape was determined using observations of 3C84 at the end of each track. Time-dependent complex gains were derived from the frequent observations of J2258−279 (0.88 Jy), and the efficacy of the solutions were verified by application to J2248−325 (0.18 Jy). All of the calibration steps were performed with the IDL based MIR software. Imaging and deconvolution were done in MIRIAD with standard routines. The synthesized beam size with natural weight was 75 × 25 (51×17 kpc), position angle −6°; the substantial north–south elongation of the beam results from the low declination of the cluster. After combining the two tracks, the rms noise was 4 mJy beam⁻¹ in 200 km s⁻¹ velocity bins.

In the continuum map (combined upper sideband (USB) and LSB, effective frequency of 220.7 GHz) we detect a point source at $\alpha = 23^{{\rm h}}44^{{\rm m}}43\buildrel{\mathrm{s}}\over{.}89$, δ = −42°43'1229 (J2000), with a flux of 0.25 ± 0.03 mJy, compared to 79.2 mJy at 833 MHz from the SUMSS database (Mauch et al. 2003). In order to determine whether the SMA continuum flux is due to star formation or synchrotron emission, we consider archival ATCA 1.4, 2.0, and 2.9 GHz data (PI: R. Kale). These data, when combined with the aforementioned SUMSS data, suggests a low-frequency radio slope of α ∼ −1.35. At a frequency of 220 GHz, this would result in a flux of 0.04 mJy, or a factor of ∼6 lower than what we measure. On the contrary, if we extrapolate the best-fit blackbody spectrum to the far-infrared data, we intersect the SMA continuum emission almost exactly (see Figure 3). Thus, we conclude that the SMA continuum emission is not probing the radio-loud AGN, and so we will not consider it further in this work.

Zoom In Zoom Out Reset image size

In order to achieve maximum signal-to-noise, we heavily binned and smoothed each spectrum before measuring the line flux. As a result, we measure the total emission-line fluxes, even in cases where there are multiple velocity components. In Figure 4, we show emission-line maps for the [O ii]λλ3726, 3729, [O iii]λλ4959, 5007, Hβ, Hγ, Hδ, [Ne iii]λ3869, and He ii λ4686 emission lines compared to a rest-frame blue image. Overall, the [O ii] emission has a very similar morphology to the blue continuum, suggesting that this gas may be photoionized by young stars. The presence of extended, filamentary [O ii] extending >30 kpc from the center of Phoenix A is reminiscent of the complex Hα filaments in nearby systems such as NGC 1275 (e.g., Conselice et al. 2001; Fabian et al. 2003; Hatch et al. 2006) and Abell 1795 (Cowie et al. 1983; McDonald & Veilleux 2009). Luminous, extended [O iii] emission is not, however, typically found in the central galaxies of cool core clusters. The additional presence of extended, high-ionization lines such as [Ne iii] and He ii suggest that an additional, harder, ionization source may also be operating on large physical scales in this system. All five emission lines, along with the stellar continuum, share a common peak. This nucleus is exceptionally bright in He ii, suggesting that it is most likely ionized by the powerful central AGN (McDonald et al. 2012a; Ueda et al. 2013).

Zoom In Zoom Out Reset image size

We quote, in Table 1, the integrated fluxes of all five aforementioned emission lines, both in the central, nuclear region (06 × 06 box), as well as in our total GMOS-IFU field of view. The total, extinction-corrected [O ii] flux in this system is f_[O ii] = 5.0 ± 0.3 × 10⁻¹⁴ erg s⁻¹ cm⁻², which corresponds to a luminosity of L_[O ii] = 7.6 ± 0.5 × 10⁴³ erg s⁻¹. Following Kewley et al. (2004), the [O ii]-derived SFR, excluding the nuclear contribution, is 431 ± 26 M_☉ yr⁻¹, while following Kennicutt (1998) predicts a higher value of 918 ± 55 M_☉ yr⁻¹. Our estimate based on the UV luminosity (McDonald et al. 2013a) lies between these two values at 798 ± 42 M_☉ yr⁻¹.

The ratios of optical emission lines such as [N ii]/Hα, [S ii]/Hα, and [O i]/Hα, are commonly used to separate low-ionization processes such as stellar photoionization from high-ionization processes such as fast shocks (e.g., Baldwin et al. 1981; Veilleux & Osterbrock 1987; Kewley et al. 2006). Such diagnostics have been applied to the complex emission-line nebulae in cool core clusters (e.g., Voit & Donahue 1997; Crawford et al. 1999; Hatch et al. 2006; McDonald et al. 2012b), finding LINER-like line ratios which could be due to a combination of photoionization from young stars and slow shocks (McDonald et al. 2012b). Due to its higher redshift, we are unable to perform these diagnostics for Phoenix A, as most of the relevant lines are redshifted to the near-infrared. Instead, we consider the ratios of [O iii]/Hβ, [O ii]/Hβ, and He ii/Hβ in Figure 5. These maps show a plume of material to the northwest of the galaxy center with very high ionization ([O iii]/Hβ > 5, He ii/Hβ > 0.1). This high-ionization material is not seen in the [O ii]/Hβ map, suggesting that the [O ii] and [O iii] have different origins. Outside of the central region, the [O iii]/Hβ emission-line ratio drops to roughly unity, which is typical for young star-forming regions.

Zoom In Zoom Out Reset image size

In Figure 6, we compare the measured optical line ratios in 3×3 pixel regions to models of fast radiative shocks (Allen et al. 2008), photoionization from young stars (Kewley et al. 2001), and photoionization from AGN (Groves et al. 2004) based on three blue emission-line ratio diagnostic plots from Groves et al. (2004). We find that much of the high-ionization emission is inconsistent with stellar photoionization and fully consistent with both shocks and AGN. In the outer region of the galaxy, where we are unable to detect [He ii], the emission is consistent with all three models, including photoionization. Given the presence of a highly luminous type-2 QSO in the center of this galaxy (McDonald et al. 2012a; Ueda et al. 2013), it is reasonable to conclude that the warm gas in the central ∼5 kpc (red points) is photoionized by the AGN. The plume of high-ionization gas extending ∼15 kpc to the north of the nucleus (green points), may instead be heated by shocks. We will return to this discussion in the next section, where we will incorporate the kinematics into these diagnostics.

Zoom In Zoom Out Reset image size

Overall, the line ratio diagnostics presented in Figures 5 and 6 suggest that there may be two to three separate ionization mechanisms at work in Phoenix A, resulting in a localized peak at the galaxy nucleus, a highly ionized plume of material to the north of the galaxy center, and a more extended, low-ionization component with the same overall morphology as the bright UV emission (McDonald et al. 2013a).

Utilizing our deep near-UV imaging (McDonald et al. 2013a), we can also do pixel-by-pixel comparisons of the emission-line fluxes to the UV fluxes. The results of such an experiment are presented in Figure 7, where we compare the [O ii], [O iii], and He ii fluxes to the near-UV continuum (∼3000 Å) from HST data. These maps, once again, reveal a plume of highly ionized material to the north of the galaxy center, with high [O iii]/UV and He ii/UV ratios which are inconsistent with star-forming regions. In contrast, the [O ii]/UV ratio map is fairly smooth, with a mean level consistent with the expectation for a star-forming region (e.g., Kennicutt 1998; Kewley et al. 2004). This figure suggests that the low-ionization and high-ionization lines may have different origins, particularly along the plume of warm gas extending north of the galaxy center.

Zoom In Zoom Out Reset image size

In Figure 8, we compare the pixel-to-pixel emission-line surface brightnesses to the cospatial rest-frame far-UV surface brightness. We confirm that the [O ii] flux is strongly correlated with the UV surface brightness, with a ratio consistent with the expectations from (Kennicutt 1998). The downturn in [O ii] at low UV surface brightness is interpreted as an age gradient; older stars still produce significant UV luminosity but their spectra are not hard enough to ionize the warm gas. Comparing the pixel-to-pixel fluxes in various emission lines to the UV continuum, we find the strongest correlation between [O ii] and the UV surface brightness (r² = 0.93), with the high-ionization lines being only weakly correlated with the UV surface brightness (r² = 0.23 and 0.07 for [Ne iii] and He ii, respectively). There appears to be two separate sequences in the [O iii] versus UV plot, again suggesting two (or more) sources of ionization: one which produces Kennicutt-like ratios (lower [O iii]/UV) and one which appears over-ionized (high [O iii]/UV). This plot further supports the emerging picture thus far that the low-ionization lines (e.g., Hβ, [O ii]) are produced in star-forming regions, while the high-ionization lines (e.g., [O iii], [Ne iii], He ii) are produced by a secondary process (e.g., shocks, AGN photoionization) which is uncorrelated with the local UV background.

Zoom In Zoom Out Reset image size

The measurement of line redshifts and widths is complicated by the fact that, in a significant number of spectra, we find multiple velocity components. We address this by attempting to fit two distinct velocity components to the [O iii]λλ4959, 5007 doublet without any smoothing/binning of the spectra. We show, in Figure 9, the relative intensities in regions where we detect emission-line splitting. We find that the region to the northeast of the galaxy center, which we showed previously to have bright [O iii] and He ii emission relative to the Hβ and UV emission, has a substantial contribution to its flux from two kinematically distinct components. This suggests a very dynamic environment, as we will discuss in more detail in Section 4.

Zoom In Zoom Out Reset image size

The resulting radial velocity and velocity dispersion maps are shown in Figures 10 and 11, respectively, where all velocities are with respect to z = 0.597. In general, we observe a relatively smooth velocity gradient from the southeast (+700 km s⁻¹) to the northwest (−400 km s⁻¹), with the minimum (v ∼ 0 km s⁻¹) occurring near the center of the [O ii] emission. It is difficult to interpret the physical meaning of these kinematics, as they could result from rotating, infalling, or outflowing gas observed from different viewing angles.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The velocity width of the [O iii] emission lines, shown in Figure 11, are significantly broadened to the northeast of the emission peak, slightly offset from the direction of the highly ionized plume shown in Figure 7. Overall, the velocity dispersion is very high, with a median value of FWHM_[O iii] ∼ 700 km s⁻¹. This is significantly higher than typical low-z cool core clusters, which have emission-line nebulae with linewidths spanning the range 100 < FWHM < 600 km s⁻¹ (e.g., McDonald et al. 2012b). Given the exceptionally high SFR, combined with the presence of a powerful AGN, there are several possible origins for the broad linewidths (e.g., Type II supernovae (SNe II), radiation-driven winds, AGN-driven winds, etc.), so it is perhaps unsurprising that the linewidths are elevated throughout the interstellar medium (ISM) of the central galaxy. We find little correlation between the [O iii] linewidth and the [O iii]/Hβ line ratio, with the exception of the high-ionization peak at "region B," which has both high He ii/Hβ and high-velocity width (∼1000 km s⁻¹). If shocks were the only contributor to the ionization, one would expect these quantities to be correlated across the entire field of view, which we do not observe. The lack of a correlation suggests a much more complex environment, with likely multiple ionization sources and significant turbulence/mixing.

We find a relatively strong velocity gradient along the highly ionized plume (Figure 10, rightmost panel), with a projected change in velocity of ∼750 km s⁻¹ over a distance of only ∼10 kpc. This could be indicative of rotation, which would imply an enclosed mass of ∼10¹¹ M_☉ in the central ∼5 kpc. For comparison, this is an order of magnitude higher than the total mass in the central ∼5 kpc of M87, the central galaxy in the Virgo cluster (Gebhardt & Thomas 2009). An alternative possible explanation for this strong gradient is a high-velocity outflow in the direction of the highly ionized plume, centered on the AGN. This outflow velocity is typical of massive galactic winds (Veilleux et al. 2005), with the ionized wind in M82 having a deprojected outflow velocity of ∼600 km s⁻¹ (Shopbell & Bland-Hawthorn 1998). The warm, ionized gas appears to reach peak speeds along the highly ionized plume, with the extended low-ionization material spanning a relatively small range in velocity. The kinematics of this gas may be influenced by an AGN-driven outflow, or general bulk motions of the gas (infalling cool material, rotation, etc.).

In Figure 12, we present a 3.8σ detection of CO(3–2) in the core of the Phoenix cluster. While this detection significance is low, it is bolstered by the proximity to Phoenix A in both position (13) and velocity (+11 km s⁻¹). We measure a line flux of S_CO(3–2) = 5.3 ± 1.4 Jy km s⁻¹. Following Solomon et al. (1992), we calculate $L^{\prime }_{{\rm CO}(3\hbox{--}2)} = 3.25\times 10^7S_{{\rm CO}(3\hbox{--}2)}\Delta v\nu _{{\rm obs}}^{-2}D_L^2(1+z)^{-3}$ = 1.1 ± 0.3 × 10¹⁰ K km s⁻¹ pc². Assuming a (3–2)/(1–0) ratio of r₃₁ ∼ 0.5, which is typical of dusty starburst galaxies (Yao et al. 2003; Iono et al. 2009; Leech et al. 2010; Papadopoulos et al. 2012) and only slightly lower than that found for Abell 1835 (r₃₁ ∼ 0.85; Edge 2001), we infer a luminosity of $L^{\prime }_{{\rm CO}(1\hbox{--}0)} = 2.2 \pm 0.6$ × 10¹⁰ K km s⁻¹ pc².

Zoom In Zoom Out Reset image size

The inferred H₂ mass is, of course, very sensitive to our choice of the CO-to-H₂ conversion, α_CO, but the latter is quite uncertain (see Bolatto et al. 2013, for a review). Previous studies of cool core clusters have used the Galactic value of α_CO ∼ 4 M_☉ pc⁻² (K km s⁻¹)⁻¹ (e.g., Edge 2001; Salomé & Combes 2003), while studies of ultraluminous infrared galaxies (ULIRGs) and other starburst galaxies have found values an order of magnitude lower: α_CO ∼ 0.6 ± 0.2 M_☉ pc⁻² (K km s⁻¹)⁻¹ (e.g., Papadopoulos et al. 2012; Bolatto et al. 2013). While we have insufficient information to properly constrain α_CO, there are several lines of evidence that lead us to adopt the lower value.

• 1.  
  The average velocity dispersion for the 10⁴ K gas in Phoenix A is σ ∼ 350 km s⁻¹. Shetty et al. (2011) show that the amount of turbulence in a given giant molecular cloud strongly correlates with the value of α_CO, such that an increase in the turbulent velocities by a factor of 10 can yield a factor of >3 decrease in the value of α_CO. While the amount of turbulence in the H ii and H₂ are bound to be different, the fact remains that the 10⁴ K gas in Phoenix A has 1–2 orders of magnitude more velocity broadening than a typical disk galaxy.
• 2.  
  The typical star formation surface density in the central ∼10 kpc is ∼5 M_☉ yr⁻¹ kpc⁻². This number is typical of starburst galaxies (α_CO ≲ 1) and orders of magnitude higher than what is measured in the Galactic disk (α_CO ∼ 4).
• 3.  
  The metallicity of the ICM in the central 50 kpc is ∼1.5 Z_☉. Under the assumption that the cooling ICM is the source of the cold gas reservoir, this implies that α_CO should be lower than Galactic (Bolatto et al. 2013).
• 4.  
  The dust temperature in Phoenix A is 87 K. This is higher than in a typical star-forming galaxy and is consistent with a lower value of α_CO, assuming that the cloud temperature and dust temperature are related (α_CO∝(σT)⁻¹; Bolatto et al. 2013).

Based on these arguments, we choose to adopt a lower, starburst-like value for α_CO. In the interest of being conservative, we will assume α_CO = 1.0 for Phoenix A but we will consider the full range of realistic values (0.4–4.0) throughout the discussion. This choice of α_CO leads to an estimate of $M_{{\rm H_2}}$ = 2.2 ± 0.6 ×10¹⁰ M_☉. At a glance, this number may seem low compared to other strong cool core clusters such as Abell 1835 ($M_{{\rm H_2}}$ = 9.2 ×10¹⁰ M_☉; Edge 2001; McNamara et al. 2013) and Zw3146 ($M_{{\rm H_2}}$ = 8.2 ×10¹⁰ M_☉; Edge 2001), which have substantially lower SFRs than the central galaxy in the Phoenix cluster. However, if we instead assume a Galactic value of the CO-to-H₂ conversion (α_CO = 4), we arrive at a much higher estimate of the total molecular gas mass: ∼8.8 × 10¹⁰ M_☉. Assuming that the same value of α_CO characterizes all starbursts in cool cores, we conclude that Phoenix, Abell 1835, and Zw3146 all harbor similar, if uncertain, quantities of cold molecular gas at their centers.

At the current rate of star formation (∼800 M_☉ yr⁻¹; McDonald et al. 2013a), and assuming the starburst value for α_CO, the current supply of cold molecular gas would be exhausted in less than 30 Myr, provided there is no compensating source replenishing the cold gas reservoir. Given that the typical BCG stellar mass in a cluster the size of Phoenix is ∼8×10¹¹ M_☉ (Lidman et al. 2012), this starburst may provide ∼3% of the total BCG stellar mass (or more, depending on how long it has been ongoing and whether or not it is being replenished).

The spatial distribution of the CO(3–2) emission is shown in Figure 13. This map, which is overlaid on an HST F814W image, was generated by combining the −150, +50, and +250 km s⁻¹ channels. There appears to be a slight offset (∼1'', ∼7 kpc) between the peak of star formation and the peak of the CO(3–2) emission, which is larger than our absolute astrometric uncertainty (∼02) but similar in size to our centroiding uncertainty (FWHM/(S/N) ∼ 8''/4 ∼ 2''). The emission shown in Figure 13 may be marginally extended but not significantly more so than the beam (75 × 24). Future ALMA observations with improved spatial resolution will be able to properly quantify any spatial offset and determine the morphology of this cold gas reservoir.

Zoom In Zoom Out Reset image size

In McDonald et al. (2013a), we report a SFR in Phoenix A of ∼800 M_☉ yr⁻¹. For context, the next most rapidly star-forming BCG lies at the center of Abell 1835, with a SFR of ∼150 M_☉ yr⁻¹ (McNamara et al. 2006). The addition of new CO(3–2) observations provide further insights into just how extreme Phoenix A is. In the upper panel of Figure 14, we show the correlation between the IR luminosity and the CO(1–0) luminosity, for a variety of systems, including cool core BCGs (O'Dea et al. 2008), luminous infrared galaxies (Papadopoulos et al. 2012), ULIRGs (Gao & Solomon 1999; Klaas et al. 2001), hyperluminous infrared galaxies (HyLIRGs; Ivison et al. 2013), and high-redshift submillimeter galaxies (SMGs; Bothwell et al. 2013). By comparing luminosities rather than derived products such as $M_{{\rm H_2}}$ and M_dust, we avoid the large uncertainties in conversion coefficients such as α_CO. This figure shows a tight correlation over five decades in luminosity, with the central galaxy in the Phoenix cluster having similar IR and CO luminosity to HyLIRGs and SMGs.

Zoom In Zoom Out Reset image size

In the lower panel of Figure 14, we compare the H₂ mass to the total infrared luminosity for the same sample. Here, we have taken $M_{{\rm H_2}}$ directly from various authors, meaning that these data span a range of α_CO from 0.6–4.6. This plot shows that Phoenix A, with a SFR of ∼800 M_☉ yr⁻¹, lies on the "extreme starburst limit," assuming a starburst-like value of the CO-to-H₂ conversion (α_CO ≲ 1). This limit, corresponding to L_IR/$M_{{\rm H_2}}$ = 500 L_☉/M_☉, represents the IR luminosity at which radiation pressure is able to disperse the cold gas reservoir (see, e.g., Thompson et al. 2005; Ivison et al. 2011). The fact that Phoenix A lies right on this line implies an exceptionally vigorous starburst, to the point of nearly tearing itself apart. Given that the warm (10⁴ K) gas is more susceptible to the effects of radiation pressure, it is indeed unsurprising that the turbulent gas motions and ionization ratios are so high throughout this central galaxy, as it is likely being repeatedly shocked as it is driven outward into the ICM by the starburst.

In McDonald et al. (2012a), we reported the presence of a dusty AGN in the central galaxy of the Phoenix cluster based on the presence of a heavily obscured X-ray point source and an exceptionally high IR luminosity. This AGN was later confirmed to be a type-2 QSO by Ueda et al. (2013) based on combined Suzaku and Chandra observations. This is one of only a small number of type-2 QSO discovered in the cores of cooling flow clusters, with the others being IRAS 09104+4109 (e.g., O'Sullivan et al. 2012) and Cygnus-A (Djorgovski et al. 1991). With these new data, we can further confirm that the reddening peaks on the nucleus (Figure 2), there are high-ionization emission lines ([Ne iii], He ii) coincident with the galaxy center (Figure 4) and the emission-line ratios are consistent with the expectation for photoionization by an AGN in the presence of dust (Figure 6; Groves et al. 2004). These data seem to paint a fairly clear picture of a dust-enshrouded AGN, although it remains unclear how much of an effect it is having on the surrounding ISM and ICM. We address the possible influence of this AGN on the ISM in the next section and await deeper X-ray observations in order to determine whether it is influencing the larger-scale ICM.

In Figure 4, we show a highly ionized plume of warm gas extending north from the central galaxy in the Phoenix cluster. This plume, detected at He ii, has substantially higher [O iii]/Hβ and [Oiii]/UV ratios than the surrounding gas, suggesting a separate ionizing mechanism. In addition, we show an ∼800 km s⁻¹ velocity gradient along the base of this plume in Figure 10. Taken together, this evidence is consistent with the presence of a galactic wind (see, e.g., Veilleux et al. 2005). We propose that some combination of both mechanical (e.g., AGN jets, SNe ejecta) and radiation pressure from both the AGN and the starburst are acting to drive the cooling material out of the cluster core. As this gas interacts with the slower-moving cool ISM/ICM, it is shock-excited which, combined with photoionization from the AGN itself, yields high-ionization lines (e.g., [O iii], He ii; Figure 4), elevated high-ionization line ratios (e.g., [O iii]/Hβ > 5, He ii/Hβ > 0.1; Figure 5), and large velocity dispersions, all of which are observed in "region B" to the northeast of the galaxy center. The narrow opening angle of this high-velocity plume seems to suggest that the wind is more likely launched by radio jets. The powerful radio source (see Figure 3) in Phoenix A is certainly energetic enough to drive such a wind, and there is some evidence that radio jets can drive an ionized outflow in nearby clusters (e.g., Werner et al. 2011; Farage et al. 2012). However, confirmation of this hypothesis awaits radio observations with higher angular resolution and broader frequency coverage.

As shown in Figure 6, the range of observed line ratios can be explained by both shocks and AGN photoionization, with stellar photoionization being more important for the low-ionization lines. Our proposed explanation for all of the observations from X-ray to radio is that the Phoenix cluster is undergoing a short-lived phase of rapid cooling, leading to both a massive starburst and rapidly accreting AGN (∼60 M_☉ yr⁻¹; McDonald et al. 2012a) that are quickly consuming the accumulated cold gas reservoir. Radiation pressure from the starburst and the obscured type-2 QSO (Ueda et al. 2013), combined with the powerful radio jets, is most likely driving a large-scale wind to the north of the central galaxy, shocking the high-velocity material against the cooling ICM (region B), leading to a plume of high-ionization material. Simultaneously, the dust-obscured AGN is locally photoionizing the nucleus (region A) and likely contributing to the ionization in the highly ionized plume, while the star formation is providing lower levels of photoionization over the full area of the starburst.

With the addition of spatially resolved optical spectroscopy and deep submillimeter data, a more complete picture of the core of the Phoenix cluster is emerging. The spectral energy distribution of Phoenix A, spanning the far-UV through submillimeter (Figure 3), is reminiscent of a dusty starburst with an obscured, radio-loud type-2 QSO at its center. The starburst, which we suspect is fueled by the 2700 M_☉ yr⁻¹ cooling flow (McDonald et al. 2012a), is operating near maximum efficiency (L_IR/$M_{{\rm H_2}}$ ∼ 500 L_☉/M_☉), rapidly depleting the cold gas reservoir over timescales of ≲30 Myr while threatening to disperse the molecular gas on even shorter timescales due to the immense radiation pressure. The central AGN, perhaps in combination with radiation pressure from the starburst, appears to be driving an outflow to the north of the central galaxy, with the He ii/UV ratio peaking ∼15 kpc north of the AGN (Figure 7).

This extreme system is almost certainly short-lived. The molecular gas reservoir, at its current size, cannot sustain such a starburst for more than ∼30 Myr. At the same time, winds generated by the AGN and radiation pressure from the starburst may eject the cold gas on even shorter timescales, as evidenced by the high L_IR/$M_{{\rm H_2}}$ ratio (which includes only the starburst). However, despite the transient nature of this phenomenon, it is worth addressing whether the Phoenix cluster can be thought of as simply a scaled-up version of a typical cool core, or if it is in an altogether different class of objects.

In Figure 15, we show the SFR (measured from UV or IR continuum) versus the Hα luminosity for a sample of cool cores drawn from the literature (Crawford et al. 1999; McNamara et al. 2006; O'Dea et al. 2008; Hicks et al. 2010; McDonald et al. 2011b). The correlation between the Hα flux and the UV or IR flux has been used to argue that the complex emission-line nebulae observed in cool core BCGs is primarily ionized by young stars. The fact that these data fall along the relationship for photoionization (Kennicutt 1998) supports this scenario. Assuming Hα/Hβ = 2.85, we find that Phoenix A lies almost exactly along this line, with L_Hα ∼ 8×10⁴³ erg s⁻¹. While there is nearly an order of magnitude gap between the next most star-forming system (Abell 1835; McNamara et al. 2006), there appears to be little difference in the ionization source between the central galaxies in systems like Abell 1795 (SFR ∼ 10 M_☉ yr⁻¹; McDonald & Veilleux 2009), Abell 1835 (SFR ∼ 150 M_☉ yr⁻¹; McNamara et al. 2006), and the Phoenix cluster (SFR ∼ 800 M_☉ yr⁻¹); they all appear to have their low-ionization lines produced by photoionization by young stars.

Zoom In Zoom Out Reset image size

In Figure 16, we compare the Hα luminosity to the molecular gas mass, following Edge (2001) and Salomé & Combes (2003). We show a range of values for α_CO in Phoenix A: from a ULIRG-like value of 0.4 to a Galactic value of 4.0 (see Section 3.4). This figure demonstrates that Phoenix A appears to be using its molecular gas more efficiently, with a relatively small cold gas reservoir given the enormous SFR and Hα luminosity. While our choice of α_CO is highly uncertain, even assuming a Galactic value (similar to Edge (2001) and Salomé & Combes (2003)) leads to a value of $M_{{\rm H_2}}$ for Phoenix A that lies below the extrapolation from nearby systems.

Zoom In Zoom Out Reset image size

In general, the low-ionization optical ([O ii], Hβ) emission lines in Phoenix A appear to be consistent with a "scaling-up" of classical low-redshift systems such as Abell 1795 and Abell 1835. Similar to low-redshift systems, the low-ionization lines appear to be ionized by young stars, while the high-ionization lines are produced by some combination of AGN photoionization and shocks. The primary difference between Phoenix A and a scaled-up, low-redshift BCG is the efficiency with which the cooling flow is converted into stars, which sits at ∼30%, compared to the typical low-z value of a few percent. This is also reflected in the ratio of star formation to cold molecular gas, which is much higher for Phoenix A than for nearby BCGs (regardless of our choice of α_CO). It appears that this highly efficient cooling phase must be short-lived, based on the current mass of the cold gas reservoir, unless this reservoir is being replenished. The starburst will likely cease in ∼30 Myr, leaving a more typical, highly inefficient cooling flow. It is currently unclear whether this short-lived, highly efficient phase of cooling is ubiquitous to cooling flows at early times. If it is, one would expect to see post-starburst signatures in a large fraction of high-redshift clusters, as these should survive for ∼10⁸–10⁹ yr.