Main

We observed the Abell 2597 BCG (Fig. 1) with the Atacama Large Millimeter/submillimeter Array (ALMA), enabling us to create a three-dimensional map of both the location and motions of cold gas at uniquely high sensitivity and spatial resolution. The ALMA receivers were sensitive to emission from the J = 2–1 rotational line of the carbon monoxide (CO) molecule. Such CO(2–1) emission is used as a tracer of cold (∼10−30 K) molecular hydrogen, which is vastly more abundant, but not directly observable at these low temperatures.

Figure 1: A multiwavelength view of the Abell 2597 BCG.
figure 1

a, Chandra X-ray, HST and DSS optical, and Magellan Hα+[N II] emission is shown in blue, yellow and red, respectively. Arrows point to the thermally unstable hot atmosphere and buoyant bubbles that permeate it. (Image credits: X-ray, NASA/CXC/Michigan State University/G. Voit et al.; optical, NASA/STScI and DSS; Hα, Carnegie Observatory/Magellan/W. Baade Telescope/University of Maryland/M. McDonald). b, HST image of Lyα emission associated with the ionized gas nebula13, with 8.4 GHz radio contours overlaid in black. The filamentary ionized nebula consists of cooler gas that has precipitated from the hot X-ray bright halo shown in a. c, Unsharp mask of the HST far-ultraviolet (FUV) continuum image of the central regions of the nebula15. The FUV emission directly traces the locations of young stars in the nebula. Very Large Array (VLA) radio contours of the 8.4 GHz source are overlaid in red.

PowerPoint slide

Full size image

The continuum-subtracted CO(2–1) images (Fig. 2) reveal that the filamentary emission line nebula that spans the galaxy’s innermost ~30 kpc (Fig. 1b) consists not only of warm ionized gas18,19,20, but also cold molecular gas. In projection, the optical emission line nebula is co-spatial and morphologically matched with CO(2–1) emission detected at a significance between (in the outer filaments) and (in the nuclear region) above the background noise level. The warm ionized nebula is therefore likely to have a substantial molecular component, consistent with results for other similar galaxies21. The total measured CO(2–1) line flux corresponds to a molecular hydrogen gas mass of , where M⊙ is the mass of the Sun. The critical (minimum) density for CO(2–1) emission requires that the volume filling factor of this gas be very low, of the order of a few per cent. The projected spatial coincidence of both the warm ionized and cold molecular nebulae therefore supports the long-envisaged hypothesis that the ionized gas is merely the warm ‘skin’ surrounding far colder and more massive molecular cores22,23, whose outer regions are heated by intense radiation from the environment in which they reside. Rather than a monolithic, kiloparsec-scale slab of cold gas, we are more likely to be observing a projected superposition of many smaller, isolated clouds and filaments.

Figure 2: ALMA observation of continuum-subtracted CO(2–1) emission in the Abell 2597 BCG.
figure 2

Emission is integrated from −600 km s−1 to +600 km s−1 relative to the galaxy’s systemic velocity. Channels are binned to 40 km s−1. Only emission is shown. 8.4 GHz VLA radio contours are overlaid in black, and Hα contours outlining the rough boundary of the ionized nebula are shown in grey. The nebula is slightly larger than the grey contours suggest: emission outside this boundary is still part of a smooth, fainter distribution of cold gas, co-spatial with similarly faint emission in the optical.

PowerPoint slide

Full size image

The data unambiguously show that cold molecular gas is falling inward along a line of sight that intersects the galaxy centre. We know this because the ALMA beam that is co-spatial with the millimetre continuum source, the radio core, and the isophotal centre of the galaxy reveals strong, redshifted continuum absorption (Fig. 3b), found by extracting the CO(2–1) spectrum from this central beam. This reveals at least three deep and narrow absorption lines (Fig. 3c), with redshifted line centres at +240, +275, and +335 km s−1 relative to the systemic (stellar) velocity of the galaxy, all within an angular (physical) region of 0.715″ × 0.533″ (1 kpc × 0.8 kpc).

Figure 3: ‘Shadows’ cast by molecular clouds moving towards the supermassive black hole.
figure 3

a, Continuum-subtracted ALMA CO(2–1) spectrum extracted from a central 10 kpc region. Brackets mark CO(2–1) emission shown in b, where 8.4 GHz radio contours are overlaid. The central radio contours have been removed to aid viewing of the continuum absorption, seen as the blue/black spot of ‘negative’ emission (which is the radio and mm core, the centre of the galaxy and the location of the black hole). c, Continuum-subtracted CO(2–1) spectrum extracted from this region co-spatial with the mm and radio core. Absorption lines are indicated in red.

PowerPoint slide

Source data

Full size image

These absorption features arise from cold molecular clouds moving towards the centre of the galaxy, via either radial or inspiralling trajectories. They manifest as continuum absorption because they cast ‘shadows’ along the line of sight as the clouds eclipse or attenuate about ~20% (or about 2 mJy) of the millimetre synchrotron continuum source, which serves as a bright backlight (13.6 mJy at rest-frame 230 GHz). The synchrotron continuum is emitted by jets launched from the accreting supermassive (~3 × 108 M⊙; ref. 13) black hole in the galaxy’s active nucleus (Fig. 4). The absorbers must therefore be located somewhere between the observer and the galaxy centre, falling deeper into the galaxy at about +300 km s−1 towards the black hole at its core. This radial speed is roughly equal to the expected circular velocity24 in the nucleus, consistent either with a nearly radial orbit, or with highly non-circular motions in close proximity to the galaxy’s core.

Figure 4: Corroborating evidence that the inflowing molecular clouds must be in close proximity to the black hole.
figure 4

a, CO(2–1) absorption spectrum from Fig. 3, with a region of emission at about +300 km s−1 marked in yellow. b, Integrated CO(2–1) emission (colour coded) from this region, showing that gas at about +300 km s−1 is confined to the innermost 2 kpc of the galaxy. c, 1.3 GHz radio continuum source from an archival VLBA observation17 with an extremely high physical resolution of ~25 pc by ~10 pc. d, e, Plots of 1.3 GHz radio continuum emission revealing H I 21 cm absorption observed against this synchrotron jet. The signal varies dramatically over scales of tens of parsecs.

PowerPoint slide

Full size image

Gaussian fits to the spectral absorption features reveal narrow linewidths of  km s−1, which means the absorbers are more probably spatially compact, with sizes that span tens (rather than hundreds or thousands) of parsecs. The shapes of the absorption lines remain roughly the same regardless of how finely the spectra are binned, suggesting that the absorbers are probably coherent structures, rather than a superposition of many smaller absorbers unresolved in velocity space. If each absorption feature corresponds to one coherent cloud, and if those clouds roughly obey size–linewidth relations25,26 for giant molecular clouds in the Milky Way, they should have diameters not larger than ~40 pc. If in virial equilibrium, molecular clouds this size would have masses of the order of 105–106M⊙, and if in rough pressure equilibrium with their ambient multiphase (103–107 K) environment13, they must have high column densities of the order of so as to maintain pressure support. The thermal pressure in the core of Abell 2597 BCG is nearly 3,000 times13 greater than that for the Milky Way, however, which means the absorbing clouds may be much smaller.

The absorbers have optical depths in the range . The physical resolution of the ALMA data is larger than the synchrotron background source, which means that the optical depth is probably contaminated by an unresolved, additive superposition of both emission and absorption within the beam. Compact, dense cold clouds are nevertheless likely to be optically thick, which may mean they eclipse the continuum source with an optical depth of unity but a small covering factor of roughly 0.2. Especially when considering beam contamination by emission, the covering factor cannot be known with certainty, as this depends on the unknown geometry of the absorbing and emitting regions within the ALMA beam.

This geometry can be constrained, however, given existing Very Long Baseline Array (VLBA) radio observations at extremely high spatial resolution17. These data resolve the 1.3 GHz and 5 GHz radio continuum source down to scales of 25 pc, revealing a highly symmetric, 100-pc-scale jet about a bright radio core (Fig. 4c). Just as we have found in cold molecular gas, inflowing warmer atomic hydrogen gas (H I) has previously been found in absorption against this parsec-scale jet, corroborating prior reports of inflowing atomic gas at lower spatial resolutions16. The inflow velocity of this gas matches that seen in our ALMA data. Remarkably, both the optical depth and linewidth of the warm atomic absorption signal varies dramatically across the jet, with a broad (σv ≈ 310 km s−1) component co-spatial with the core that is absent just ~20 pc to the northeast, where only a narrow (σv ≈ 50 km s−1) H I line is found at the same redshift. This effectively requires the inflowing atomic gas to be confined within the innermost ~100 pc of the black hole, as gas further out would give rise to an unchanging absorption signal across the compact jet. The infall velocity is the same as that for the cold molecular clouds seen in CO(2–1) absorption, which means they most probably stem from the same spatial region, within tens of parsecs of the accreting black hole.

This is further supported by the ALMA data itself. In emission, all gas around approximately +300 km s−1 that is conceivably available to attenuate the continuum signal is confined to the innermost 2 kpc about the nucleus (Fig. 4a, b). The radial dependence of molecular cloud volume number density within this region is uncertain, but probably steeper than r−1, and likely to be closer to r−2 (Fig. 4b). This means that the chances of a random line of sight crossing will drop with increasing distance from the black hole. If the gas volume density goes as r−2, a cloud 100 pc from the black hole is ten times more likely to cross our line of sight than a cloud at a galactocentric distance of 1 kpc. It would be exceedingly unlikely for three such clouds to cross our line of sight to the black hole were they spread over several kiloparsecs throughout the galaxy’s outskirts.

The data therefore serve as strong observational evidence for an inward-moving, clumpy distribution of molecular clouds within a few hundred parsecs of an accreting supermassive black hole. The result augments a small but growing set of known molecular absorption systems27,28,29 whose black hole proximity is less well constrained. The infalling clouds in Abell 2597 BCG are probably a few to tens of parsecs across and therefore massive (perhaps 105–106M⊙ each). If they are falling directly towards the black hole, rather than bound in a non-circular orbit that tightly winds around it, they could supply an upper-limit accretion rate of the order of ~0.1 to a few solar masses per year, depending on the three-dimensional distribution of infalling clouds. If most of the clouds are instead locked in non-circular orbits around the black hole, the fuelling rate would depend on the gas angular momentum, and the local supply of torques that might lessen it. Simulations suggest9,10,14 that such torques may be plentiful, as they predict a stochastic ‘rain’ of thermal instabilities that condense from all directions around the black hole, promoting angular momentum cancellation via tidal stress and cloud–cloud collisions. Even highly elliptical cloud orbits should therefore be associated with significant inward radial motions. The clouds might fall onto the accretion disk itself, or into a clumpy rotating ring akin to the ‘torus’ invoked in AGN unification models30.

Methods

Observations, data reduction, and analysis

The new ALMA data presented in this paper were obtained in Cycle 1 with the use of 29 operational antennae in the 12 m Array. ALMA’s Band 6 heterodyne receivers were tuned to a frequency of 213 GHz, sensitive to the J = 2–1 rotational line transition of carbon monoxide at the redshift of the Abell 2597 BCG (z = 0.0821). The ALMA correlator, set to Frequency Division Mode (FDM), delivered a bandwidth of 1,875 MHz (per baseband) with a 0.488 MHz channel spacing, for a maximum spectral resolution of about 2 km s−1. One baseband was centred on the CO(2–1) emission line, while the other three sampled the local continuum. Maximum antenna baselines extended to ~1 km, delivering an angular resolution at 213 GHz of ~0.7″ within a ~28″ primary beam (field of view). ALMA observed the Abell 2597 BCG, located at RA 23 h 25 min 20 s, dec. −12° 07′ 38″ (J2000), for a total of ~3 h over three separate scheduling blocks executed between 17 and 19 November 2013. The planet Neptune and quasars J2258−2758 and J2331−1556 were used for amplitude, flux, and phase calibration. The data were reduced using CASA version 4.2 with calibration and imaging scripts kindly provided by the ALMA Regional Centers (ARCs) in both Garching, Germany and Manchester, UK. Beyond the standard application of the phase calibrator solution, we iteratively performed self-calibration of the data using the galaxy’s own continuum, yielding a ~14% decrease in RMS noise to a final value of 0.16 mJy per 0.715″ × 0.533″ beam per 40 km s−1 channel. There is effectively no difference in CO(2–1) morphology between the self-calibrated and non-self-calibrated data cubes. Measurement sets were imaged using ‘natural’ visibility weighting and binning to either 5 km s−1, 10 km s−1, or 40 km s−1, as indicated in the figure legends. The figures presented in this Letter show only continuum-subtracted, pure CO(2–1) line emission. The rest-frame 230 GHz continuum observation is dominated by a bright (13.6 mJy) point source associated with the AGN (detected at ), serving as the bright ‘backlight’ against which the continuum absorption features presented in this Letter were observed. The continuum data also features compact (~5 kpc) extended emission at ~10σ that extends along the galaxy’s dust lane, to be discussed in a forthcoming paper.

Adoption of a systemic velocity

Interpretation of gas motions relative to the stellar component of a galaxy requires adoption of a systemic (stellar) velocity to be used as a ‘zero point’ marking the transition from blue- to redshift. All CO(2–1) line velocities discussed in this Letter are set relative to 213.04685 GHz, where observed CO(2–1) emission peaks. This frequency corresponds to 12CO(2–1) (rest-frame 230.538001 GHz) at a redshift of z = 0.0821. This redshift is consistent, conservatively within ±60 km s−1, with every other available multiwavelength tracer of the galaxy’s systemic velocity, including prominent Ca II H, K, and G-band absorption features17 that directly trace the galaxy stellar component, the redshift of all optical emission lines31, as well as a broad (FWHM 412 km s−1) H I absorption component16 at the optical emission and absorption line redshift. It is also consistent, within ~60 km s−1, with a cross-correlation of emission and absorption lines using galaxy template spectra17, as well as with all other published reports of the galaxy’s systemic velocity (found, for example, within the HyperLeda database). We are therefore certain that the reported redshift of the absorption features discussed in this Letter indeed corresponds to real motion relative to the galaxy’s stellar component. Without caveat or ambiguity, the absorbing cold clouds are moving into the galaxy at roughly ~300 ± 60 km s−1.

Mass estimates

All molecular gas masses estimated in this letter adopt the following relation32:

where SCOΔv is the emission integral (effectively the total CO flux over the region of interest), z is the galaxy redshift (z = 0.0821), and DL its luminosity distance (373.3 Mpc), for which we assume a flat ΛCDM model wherein H0 = 70 km s−1 Mpc−1, ΩM = 0.3, and ΩΛ = 0.7. This mass estimate most critically relies on an assumption of the CO-to-H2 conversion factor32, XCO. In this Letter we assume the average Milky Way value of XCO = 2 × 1020 cm−2 (K km s−1)−1 and a CO(2–1) to CO(1–0) flux density ratio of 3.2. Other authors have provided extensive discussion of these assumptions as they pertain to cool core BCGs29,33,34. Scientific conclusions in this paper are largely insensitive to choice of XCO.

A single Gaussian fit to the CO(2–1) spectrum extracted from an aperture containing all detected emission yields an emission integral of Jy km s−1 with a line FWHM of 252 ± 14 km s−1, corresponding to a total molecular hydrogen (H2) gas mass of M⊙. This is very close to the previously reported15 mass, based on an IRAM 30m CO(2–1) observation, of (1.8 ± 0.3) × 109M⊙. This comparison is not one-to-one, as the mass from the IRAM 30m observation was computed from within a beam size of 11″ (rather than 28″ for the ALMA data), and used a CO(2–1)/CO(1–0) flux ratio of 4 (rather than 3.2, as we use here). These differences are minor, particularly because nearly all of the CO(2–1) emission detected by ALMA is found within the central 11″ size of the IRAM 30m beam. It is therefore safe to say that our ALMA observation has detected nearly all emission that was detected in the single-dish IRAM 30m observation, and that very little extended emission has been ‘resolved out’ by ALMA.

Estimating physical properties of the redshifted absorbing molecular gas

We have estimated a rough upper-limit size of the absorbing clouds assuming the widely adopted Larson et al.25 and Solomon et al.26 size–linewidth relation for molecular clouds in the Milky Way (namely, the ref. 26 fit);

where σv is the velocity linewidth of the cloud and S is the diameter of the cloud in parsecs. A measured absorber linewidth of σv ≈ 6 km s−1 would then correspond to a size of ~36 pc. As noted in the main text of the Letter, the thermal pressure in the Abell 2597 BCG is about 3,000 times higher than that for the Milky Way13, so it is likely that the above relation does not apply. A higher ambient pressure implies higher compression and therefore smaller cloud size, so the above estimate should, at best, be considered a very rough upper-limit. The main lesson to take away from this exercise is that the absorbing clouds are probably physically compact (that is, a few to tens—rather than hundreds—of parsecs in diameter).

The three clouds are separated from one another by ~45–60 km s−1 in velocity space, which means they are unlikely to be closely bound satellites of one another. Instead, it is more likely that they represent three random points along a radial distribution of clouds.

If the absorbers are in virial equilibrium, their masses Mcloud can be roughly estimated by applying the virial relation,

where Rcloud is the cloud radius (as roughly estimated above) and σv is its velocity dispersion (also as above).

CO(2–1) optical depths for the absorbers were estimated by assuming that:

where Itotal and Icontinuum are the integrated intensities of the total (line plus continuum) and continuum-only signals, respectively, and τCO(2–1) is the optical depth of the CO(2–1) absorption feature.

The stellar velocity dispersion of the BCG33 is σv = 220 ± 19 km s−1. Under the assumption of an isothermal sphere, the circular velocity should be ~300 km s−1 (that is, ), which is roughly the line of sight velocity of the absorption features. That the absorbers’ redshift is a significant fraction of the expected circular velocity means they could be on a nearly radial orbit (though their transverse velocity cannot be known with this single observation).

While not discussed in the main text, there is an additional simple argument that independently suggests that the inward moving molecular clouds must be in close proximity to the black hole. If our line of sight is representative, and therefore a ‘pencil beam’ sample of a three-dimensional spherical distribution of clouds, the total mass of cold gas contained within this distribution should go roughly as:

where fc is the covering factor and r is the radius of an imaginary thin spherical shell of molecular gas with column density NH. If such a shell had a covering factor of 1, a radius of 1 kpc, and a column density of 1022cm−2, then the total mass of molecular hydrogen contained within that shell would be roughly one billion solar masses. A column density in excess of 1022cm−2 requires this distribution to be contained within a sphere of radius <<1 kpc, lest the limit set by the total mass of molecular hydrogen in the galaxy be violated. If the characteristic column density is 1023 cm−2, for example, this mass must be contained within a sphere of radius 300 pc, or else its total mass would exceed the 1.8 × 109 M⊙ of cold gas present in the galaxy.

Codes, software, and data availability

Codes that we have written to both reduce and analyse the data presented in this Letter have been made publicly available at https://github.com/granttremblay/Tremblay_Nature_ALMA_Abell2597. Reduction of the data as well as some simple modelling (for example, fitting of Gaussians to lines) was performed using routines included in CASA version 4.2, available at https://casa.nrao.edu/. Plots were made using both Python’s MatPlotLib and Veusz, which is available at http://home.gna.org/veusz/.