A 10¹⁰ SOLAR MASS FLOW OF MOLECULAR GAS IN THE A1835 BRIGHTEST CLUSTER GALAXY

The total CO (3–2) and CO (1–0) spectra are presented in Figure 1. The CO emission is centered within ∼100 km s⁻¹ of the nebular emission line redshift (Crawford et al. 1999). Each spectral profile was fitted with a single Gaussian after the continuum was subtracted. The emission integral at CO (1–0) is 3.6 ± 0.2 Jy km s⁻¹. The molecular gas mass was calculated as

$$\begin{equation} M_{\rm mol} = 1.05\times 10^4 X_{\rm CO}\left(\frac{1}{1+z}\right) \left(\frac{S_{\rm CO}\Delta v}{\rm Jy\,km\,s^{-1}}\right) \left(\frac{D_{\rm L}}{{\rm Mpc}}\right)^2 \,M_\odot. \end{equation} \tag{ 1 }$$

Zoom In Zoom Out Reset image size

This expression yields a total molecular gas mass of 4.9 ± 0.2 × 10¹⁰ M_☉. The conversion factor between CO and molecular gas, X_CO = 2 × 10²⁰ cm⁻²(K km s⁻¹)⁻¹, is the average Galactic value (Bolatto et al. 2013; Narayanan et al. 2012). The primary line at zero velocity has a Gaussian profile full-width at half maximum, FWHM = 130 ± 5 km s⁻¹, after correcting for instrumental broadening. This width is 5–6 times smaller than would correspond to the BCG's expected velocity dispersion ∼250–300 km s⁻¹. Molecular gas moving with a nearly isotropic velocity pattern cannot be supported against collapse at such low speeds. The gas may be supported instead by rotation in a disk projected nearly face-on. The observed velocity width would then represent gas speeds out of the disk's plane (Section 4.4).

The R-band and far-UV (FUV) images taken with the Hubble Space Telescope (O'Dea et al. 2010) are presented in Figure 2. The R-band image shows the BCG in relation to its molecular gas, hot atmosphere, and other neighboring galaxies. The box superposed on the image indicates the scale of the UV and CO (3–2) images presented in Figure 2. A Chandra X-ray image with a similar box superposed is shown in Figure 2. Most of the molecular gas lies within one arcsec (4 kpc) of the nucleus. The UV continuum emission is emerging from the sites of star formation proceeding at a rate upward of 100–180 M_☉ yr⁻¹ (M06; Egami et al. 2006; Donahue et al. 2011). No bright UV or X-ray point source is associated with a nuclear AGN. The CO (3–2) gas coincident with the UV emission is presumably fueling star formation. The CO and UV emissions are straddled by two bright and presumably rapidly cooling X-ray emission regions oriented to the northeast and southwest of the nucleus. No CO emission is detected toward the most rapidly cooling gas. Two X-ray cavities are located a few arcsec to the northwest and southeast of the CO (3–2) emission.

Zoom In Zoom Out Reset image size

Roughly half of the CO (3–2) flux is emerging from the inner half-arcsecond radius of the BCG and is unresolved. Assuming half of the central molecular gas mass and star formation lie within the same region, we find the surface densities of star formation and molecular gas to be log Σ_SFR = 0.87 M_☉ yr⁻¹ kpc⁻² and log μ_CO = 3.2 M_☉ pc⁻², respectively. Based on these values, the BCG lies with normal, circumnuclear starburst galaxies on the Schmidt–Kennicutt relation (Kennicutt 1998).

We present a grid of CO (3–2) emission spectra corresponding to the grid projected onto the CO (3–2) image in Figure 3. The mean line-of-sight velocities measured with Gaussian profile fits are indicated in each grid box. The size of each grid box corresponds approximately to the FWHM of the synthesized beam. Velocity differences of a few to a few tens of km s⁻¹ are observed across the central structure. No clear evidence for rotation is observed. If the CO (3–2) structure is a rotating disk, the small velocity gradients and narrow line width are consistent with it being nearly face-on.

Zoom In Zoom Out Reset image size

Figure 4 is similar to Figure 3, but with a coarser grid intended to increase the signal in the outer region of the central structure. The mean line-of-sight velocities of the emission features are indicated where significant CO (3–2) is detected in emission. The tongue of gas located 1.5 arcsec to the north has a broad line profile with velocities of −15 to  − 60 km s⁻¹. Likewise, the tongue extending to the west is traveling at a velocity of −70 km s⁻¹ with respect to the bulk of the gas. The gas in the east–southeast (bottom left) grid boxes has velocities similar in magnitude but opposite in sign (redshifted) with respect to the central emission. The north and west tongues of blueshifted gas are oriented roughly toward the northwest X-ray cavity. The redshifted gas to the southeast is oriented roughly toward the southeast X-ray cavity. The tongues of gas appear to be dynamically decoupled from the central structure. Below, we relate this gas to the more extended molecular gas seen in CO (1–0).

Zoom In Zoom Out Reset image size

We examine the molecular gas velocities on larger scales using the grid of spectra in CO (1–0) presented in Figure 5, which matches the resolution of the telescope configuration. The sky grid corresponds to spectra shown in the right panel. The contours represent CO emission, and their colors correspond to the color-coded velocity stripes superposed on the spectra. A two-dimensional Gaussian profile has been fitted to and subtracted from each channel in order to remove the central emission from the contour map.

Zoom In Zoom Out Reset image size

The CO (1–0) map reveals tongues of emission projecting roughly 10 kpc to the north–northwest and southeast of the nucleus. Their orientations are similar to the smaller, tongue-like features seen in the CO (3–2) image. Molecular gas traveling at +480 km s⁻¹ in the eastern box is redshifted with respect to the systemic velocity. A narrower, blueshifted gas velocity component is seen in the north and northwest boxes extending to velocities of −200 km s⁻¹. The redshifted gas contour north of the nucleus is significant only at the 2σ–3σ level. Present as a small bump in the nuclear spectrum at a velocity of 300 km s⁻¹, it is of marginal significance and will not be discussed further.

In summary, blueshifted gas lies exclusively to the north–northwest, whereas redshifted gas lies primarily to the east–southeast. No significant high-velocity gas is observed in the northeast, southwest, and west grid boxes, nor is it observed toward the nucleus. This pattern is consistent with a broad, bipolar flow of molecular gas, which we discuss in greater detail in Section 4.3. While this interpretation accounts best for the data in hand, it is not unique. The gas may in principle have accreted with some net angular momentum that placed it on nearly circular rather than radial orbits so that the gas is in nearly ordered motion about the BCG.

The integrated flux under the redshifted and blueshifted emission profile wings are 0.4  ±  0.2 and 0.27 ± 0.09 Jy km s⁻¹, respectively, giving a total flux integral of 0.7 ± 0.2 Jy km s⁻¹. They correspond to a molecular hydrogen mass of 1.0  ±  0.3 × 10¹⁰ M_☉. The accuracy of the integrated fluxes are sensitive to the continuum, particularly in the redshifted emission wing. A slightly higher mass is found in the redshifted component compared to the blueshifted component, indicating an asymmetric flow.

The high-speed molecular gas is observed in emission, so the ALMA observations alone are unable to discriminate between inflow and outflow. An inflow of molecular gas from the cooling flow would be a natural but problematic interpretation. Gas cooling from a steady accretion flow would fall inward, reaching its highest speeds in the nucleus (Lim et al. 2008). This is not observed. Instead, the high-velocity gas is projected away from the nucleus. Assuming the CO (3–2) and CO (1–0) lines track the same gas, higher line-of-sight velocities are observed 5–10 kpc toward the northwest and southeast of the nucleus. The gas to the north–northwest is blueshifted from velocities of a few tens of km s⁻¹ at radii of ∼3 kpc to ∼250 km s⁻¹ at a radius of ∼10 kpc. Likewise, the gas to the east–southeast is redshifted with velocities of ∼40–60 km s⁻¹ at ∼3 kpc, increasing to >300 km s⁻¹ at ∼10 kpc. The yellow wing in the nuclear spectrum at +250 km s⁻¹ strengthens as it is redshifted to higher velocities in the eastern grid box, indicating higher gas masses traveling at higher speeds.

Let us assume the gas projected 10 kpc from the nucleus with radial speeds lying between 250 and 480  km s⁻¹ began its descent at rest and fell radially to its current location. We estimate its initial radius using a Hernquist law by adopting an enclosed mass within a 20 kpc radius of 1.6 × 10¹² M_☉ (R. A. Main et al., in preparation), and an effective radius of ∼10 kpc. We find that this molecular gas would have achieved its radial speed had it fallen from an altitude of ∼20–30 kpc. If gas is flowing steadily onto the disk, we would expect to observe gas velocities toward the disk and nucleus lying between 600 and 800  km s⁻¹, but we do not. Molecular gas that began its journey with a significant initial velocity (imparted by turbulence or a donor galaxy) would be traveling faster. In principle, drag from the intracluster medium might slow infalling clouds more effectively closer to the cluster center. However, cloud parameters must then be finely tuned to allow the clouds to free fall at large radii while giving them terminal speeds on the order of 10 km s⁻¹ at smaller radii.

We are then left with the following two scenarios: the high-speed molecular gas cooled recently from the hot atmosphere in the past 10 Myr or so and has not yet arrived in the disk, or that it arrived provenance unknown and is supported against gravity by orbiting with a large velocity in the plane of the sky. The former interpretation implies an X-ray cooling rate of ∼1000 M_☉ yr⁻¹, which is inconsistent with an upper limit from X-ray spectroscopy of <140 M_☉ yr⁻¹ (Sanders et al. 2010). Neither scenario can be ruled out using the data at hand. However, the velocity patterns and flow rates implied by the CO (3–2) and CO (1–0) emissions are inconsistent with steady inflow and are more naturally interpreted in the context of a bipolar outflow of molecular gas.

Molecular outflows are common in ULIRGs, QSOs, and starburst galaxies (reviewed by Veilleux et al. 2005 and Fabian 2013). Driving mechanisms include radiation pressure on dust and mechanical winds powered by supernovae. Radiation from hot stars and AGNs will drive out gas when dM/dt × v_CO ≲ L_UV/c. The left-hand side is the product of the outflow rate and the gas velocity. The right-hand side is the sum of the UV luminosity from the AGN and stars divided by the speed of light. The FUV image shown in Figure 2 reveals no UV point source associated with the AGN. Stars are producing all of the UV flux. For a stellar UV luminosity L_FUV = 1.85 × 10⁴³ erg s⁻¹ (O'Dea et al. 2010), radiation pressure would be too feeble to drive an outflow rate of 200 M_☉ yr⁻¹ by more than three orders of magnitude.

The power input by core collapse supernovae, 4 × 10⁴³ erg s⁻¹ (M06), is comparable to the kinetic power of the outflow, E_K ≃ 10⁵⁸ erg, t_out ≃ 3 × 10⁷ yr, P_K ∼ 10⁴³ erg s⁻¹, and is therefore energetically significant. However, in order to power the flow by supernova explosions, most of their mechanical energy must couple to molecular gas driving bulk motion rather than thermal motion, which would be hard to understand. Furthermore, their spherical blast patterns and the work against gravity and the surrounding gas pressure would hinder a sustained and substantial bipolar flow over such large distances. Instead of driving a flow, supernova explosions may be thickening the disk and perhaps increasing the cross section between the molecular gas and the radio AGN, which can easily power the flow (Section 4.4).

The mechanical power of the jet estimated from the X-ray cavities, P_cav ≃ 10⁴⁵ erg s⁻¹, is by far the most potent power source. Although the jet momentum is insufficient to lift the gas, the kinetic energy of the cold flows is only ∼1% of the total energy output of the AGN. The molecular gas is projected along and behind the rising bubbles, providing circumstantial evidence connecting the bubbles to the molecular flow. Moreover, the molecular flow speeds are consistent with buoyancy speeds of cavities, which rise at a substantial fraction of the atmosphere's sound speed (Churazov et al. 2001). The atmospheric sound speed in A1835 is ∼1000 km s⁻¹.

This interpretation has its own problems. Despite ample AGN power, the bubbles must couple to the molecular gas and lift it out of the galaxy. By Archimedes' principle, they would be unable to lift more molecular gas than hot gas they displace, which is ∼3 × 10¹⁰ M_☉. In addition to displacing the hot plasma, rising X-ray bubbles draw metal-enriched X-ray plasma out from cluster centers at rates of tens to hundreds of solar masses per year (Simionescu et al. 2008; Werner et al. 2010; Kirkpatrick et al. 2009, 2011). A1835's AGN is lifting ∼4 × 10¹⁰ M_☉ of hot gas out along the bubble axis (C. C. Kirkpatrick, in preparation), a value that is a few times larger than the mass of the molecular outflow and close to our estimate of the amount of gas displaced by the bubbles. These estimates are uncomfortably close to the outflowing molecular gas mass and would imply surprisingly efficient coupling between the radio bubbles and both the hot, ∼4 × 10⁷ K, tenuous, 0.1 cm⁻³, volume-filling plasma and the ∼30 K molecular gas.

How the molecular gas couples to the bubbles is unclear. Ram pressure associated with simulated high-Eddington-ratio hydrodynamic jets is able to sweep away both the cold and warm phases of the interstellar medium (e.g., Wagner et al. 2012). Whether the molecular clouds in A1835 are being accelerated by jets or are being lifted in the updraft of the X-ray bubbles (e.g., Pope et al. 2010) is unclear. Observation suggests the latter. Molecular gas is more readily coupled to the hot gas when the density contrast is low. In Section 4.4, the density in the molecular disk is estimated to be ∼1000 times that of the hot phase. However, for turbulent velocities of ∼100 km s⁻¹, if the dynamical pressure of the molecular gas matched the pressure of the hot gas, it would only be 50 times as dense. The high level of turbulence maintained by rapid ongoing star formation may then help to explain how the tenuous hot gas is able to lift molecular gas.

The bubbles would be able to lift the mass more easily if the molecular hydrogen cooled out of hotter gas as it rises in the bubbles' wake (see, for example, Revaz et al. 2008). The mean plasma density and temperature in the central 20 kpc of the cluster are 0.1 cm⁻³ and T = 2.5 keV, respectively. The cooling time of this gas, assuming solar metal abundance, is 0.22 Gyr, which is longer than the time it would take to displace the gas to a projected distance of about 10 kpc. However, lower entropy gas at a temperature of 1 keV in local pressure equilibrium would have a density of 0.25 cm⁻³. Its cooling time would be only 2 × 10⁷ yr, which is comparable to the rise times of both the bubbles and molecular gas. Therefore, the ≲1 keV plasma lifted out of the center would have time to cool behind the cavities. This rapidly cooling plasma would have densities a few times larger than the ambient plasma, but it would be several orders of magnitude less dense than the molecular gas itself. It would therefore be much easier to lift and accelerate to the speeds observed. This mechanism, which would tend to draw the most rapidly cooling plasma out of the BCG, may help to explain the dearth of ≲1 keV plasma in A1835 and other clusters (Peterson et al. 2003).

The outflow rate is a poorly defined quantity. We estimate it by dividing the mass of the outflowing molecular gas by either the time for the bubbles to rise by buoyancy to their current projected distances or by the time required for the molecular gas to reach its current projected distance from the nucleus. A range of velocities and distances are observed, with timescales lying between 3–5 × 10⁷ yr. They imply an outflow rate of ∼200–300 M_☉ yr⁻¹, which is comparable to the BCG's mean star formation rate. The center of the galaxy would then be swept of its molecular gas in only a few hundred million years, starving the black hole and starburst of needed fuel. However, the fate of most of the gas is unclear. The one-dimensional outflow speeds are somewhat larger than the circular speed of the stars. If the molecular gas is flowing ballistically, most of it should return unless it evaporates into the hot atmosphere. If the molecular gas is coupled to the rising bubbles or continues to be accelerated by the AGN, it could travel further. However, if the molecular gas formed behind the bubbles in a cooling wake, it is unlikely to evaporate into the hot medium and would return in a circulation flow or "fountain" of molecular gas, similar to that inferred in NGC 1275 (Lim et al. 2008; Salomé et al. 2006, 2011). The impact of a molecular fountain on the star formation and AGN histories of BCGs and normal elliptical galaxies is not understood.

The dynamical state and high average density of the molecular gas in the central kiloparsec of the BCG have significant implications for this system. The lack of evidence for rotation in the molecular gas implies that, if the gas is rotationally supported, its rotation axis must be very close to our line of sight. At the same time, the full velocity width at half maximum for the molecular gas in this region is 130 km s⁻¹, corresponding to a line-of-sight velocity dispersion of σ_los ≃ 55 km s⁻¹ and a one-dimensional turbulent velocity v_T = σ_los. This suggests the gas lies in a face-on disk.

This high turbulent velocity is consistent with the disk being marginally gravitationally unstable; the Toomre Q parameter is

$$\begin{eqnarray} Q&=&{v_cv_T\over \pi G r\Sigma _g(r)}\approx 0.43 \left({v_c\over 400 \,{\rm km\,s^{-1}}}\right) \left({v_T\over 55 \,{\rm km\,s^{-1}}}\right)\nonumber \\ && \times \left({R_e\over 2\,\rm kpc}\right)^{-1} \left({\Sigma _g(R_e)\over 0.4 \,\rm g\,cm^{-2}}\right)^{-1}, \end{eqnarray} \tag{ 2 }$$

where we have scaled to the radius R_e enclosing half the CO (3–2) flux, which we assume to enclose half the mass. Based on A1835's mass profile (R. A. Main et al., in preparation; see Section 4.1), the circular speed at 10 kpc lies between 300 and 420  km s⁻¹.

We have therefore scaled the circular velocity at 2 kpc to a conservative value of 400 km s⁻¹. The circular velocity would have to exceed 940 km s⁻¹ to stabilize the gas disk, so the Toomre criterion is easily met. We noted in Section 3.2 that the BCG lies with starburst galaxies on the Schmidt–Kennicutt relation; galaxies on that relation have Q ≲ 1. Furthermore, A1835's disk has similar properties to those observed in vigorously star-forming galaxies at z ∼ 2 (e.g., Elmegreen & Elmegreen 2006; Newman et al. 2012).

For a gas-to-dust mass ratio of 100, the surface density (Σ_g = 1600 M_☉  pc⁻²) corresponds to an A_v ≈ 100. The weight per unit area under self-gravity or dynamical pressure of this gas is

$$\begin{equation} p_{\rm dyn}={\pi \over 2} G\Sigma _g^2\approx 1.7\times 10^{-8}\,{\rm dyne}\,\rm cm^{-2}. \end{equation} \tag{ 3 }$$

This is substantially higher than the thermal pressure of the hot gas.

Being marginally gravitationally stable implies that the disk scale height H = (v_T/v_c)r, or about 275 pc at R_e = 2 kpc. The mean density at that radius is then $\bar{\rho }_c=\Sigma _g/(2H)\approx 2.4\times 10^{-22} \,\rm g\,cm^{-3}$, or n_H ≈ 100. This mean density is somewhat higher than the mean density in massive Milky Way giant molecular clouds. The Toomre mass of molecular clouds is then M_T = H²Σ_g ≃ 1.4 × 10⁸ M_☉. The turbulent pressure of the cold gas is

$$\begin{equation} p_{\rm turb}=\bar{\rho }_c v_T^2\approx 0.7\times 10^{-8}{\,\rm dyne}\,\rm cm^{-2}, \end{equation} \tag{ 4 }$$

i.e., the turbulent motions provide enough pressure to support the disk in a marginally stable state.

Turbulence is believed to decay on a dynamical time. Maintaining the turbulence in A1835 would then require a turbulent power of

$$P_{\rm turb}={3 M_g v_T^2 \over 2 R_e / v_c} \simeq 3 \times 10^{43} \,{\rm erg\,s^{-1}}.$$

This is similar to the total luminosity supplied by supernovae, if the star formation rate is ∼200 M_☉ yr⁻¹, $L_{{\rm Sne}}=6\times 10^{43}(\dot{M}_*/200 \,M_\odot \,\rm yr^{-1}) \,{\rm erg\,s^{-1}}$, if the supernovae are well coupled to the molecular gas and if they do not radiate more than ∼50% of their energy away. We observe that Q ≈ 1, so we expect large-scale features such as spiral arms or bars in the disk to develop. These features will transport angular momentum efficiently inward, down to subparsec scales, e.g., Hopkins & Quataert (2010, 2011).

Molecular gas associated with starburst galaxies, ULIRGS, and QSOs is often attributed to wet mergers. The center of a rich cluster with a large velocity dispersion and a dearth of gas-rich donor galaxies is an unlikely location for a wet merger. Ram pressure experienced by a plunging, gas-rich donor galaxy would strip most of its atomic gas and much of its molecular gas before it reaches the BCG (Combes 2004; Roediger & Brüggen 2007; Kirkpatrick et al. 2009; Ruszkowski et al. 2012). Being dense and centrally concentrated, molecular gas is tightly bound and more resilient to stripping than atomic gas. Therefore, short of a direct collision, a plunging galaxy should retain much of its molecular gas (Young et al. 2011). Finally, the BCG's molecular gas mass exceeds by large factors that of most galaxies in clusters at its epoch. The likelihood that such a galaxy, if present, would hit the BCG directly and deposit its molecular gas at the low speeds observed seems remote.

The molecular gas in A1835 probably cooled from the hot atmosphere and settled into the BCG. Molecular gas masses of 10⁹–10¹⁰ M_☉ are prevalent in BCGs but only those centered in hot atmospheres whose central cooling times lie below ∼10⁹ yr (Edge 2001; Salomé & Combes 2003). BCGs in Coma-like clusters with long central cooling times are not gas rich. A1835 is an extreme example of this class of BCGs. Its cooling rate of ≲ 140 M_☉ yr⁻¹ (Sanders et al. 2010) would supply the molecular gas in a few hundred Myr, which is comparable to the age of the starburst (M06).