LETTER

doi:10.1038/nature17969

Cold, clumpy accretion onto an active supermassive

black hole
GrantR.Tremblay1,2, J.B.RaymondOonk3,4, FranoiseCombes5, PhilippeSalom5, Christopher P.ODea6,7, StefiA.Baum6,8,
G.MarkVoit9, MeganDonahue9, BrianR.McNamara10, TimothyA.Davis2,11, MichaelA.McDonald12, AlastairC.Edge13,
TracyE.Clarke14, RobertoGalvn-Madrid2,15, MalcolmN.Bremer16, LouiseO.V.Edwards1, AndrewC.Fabian17, StephenHamer5,
YuanLi18, AnalleMaury19, HelenR.Russell17, AliceC.Quillen20, C.MeganUrry1, JeremyS.Sanders21 & MichaelW.Wise3

Supermassive black holes in galaxy centres can grow by the

accretion of gas, liberating energy that might regulate star
formation on galaxy-wide scales13. The nature of the gaseous fuel
reservoirs that power black hole growth is nevertheless largely
unconstrained by observations, and is instead routinely simplified
as a smooth, spherical inflow of very hot gas4. Recent theory57 and
simulations810 instead predict that accretion can be dominated by
a stochastic, clumpy distribution of very cold molecular cloudsa
departure from the hot mode accretion modelalthough
unambiguous observational support for this prediction remains
elusive. Here we report observations that reveal a cold, clumpy
accretion flow towards a supermassive black hole fuel reservoir
in the nucleus of the Abell 2597 Brightest Cluster Galaxy (BCG),
a nearby (redshift z=0.0821) giant elliptical galaxy surrounded
by a dense halo of hot plasma1113. Under the right conditions,
thermal instabilities produce a rain of cold clouds that fall towards
the galaxys centre14, sustaining star formation amid a kiloparsecscale molecular nebula that is found at its core15. The observations
show that these cold clouds also fuel black hole accretion, revealing
shadows cast by the molecular clouds as they move inward at about
300 kilometres per second towards the active supermassive black
hole, which serves as a bright backlight. Corroborating evidence
from prior observations16 of warmer atomic gas at extremely high
spatial resolution17, along with simple arguments based on geometry
and probability, indicate that these clouds are within the innermost
hundred parsecs of the black hole, and falling closer towards it.
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=21 rotational line of the carbon
monoxide (CO) molecule. Such CO(21) emission is used as a tracer of
cold (1030K) molecular hydrogen, which is vastly more abundant,
but not directly observable at these low temperatures.
The continuum-subtracted CO(21) images (Fig. 2) reveal that the
filamentary emission line nebula that spans the galaxys innermost
~30kpc (Fig. 1b) consists not only of warm ionized gas1820, but also
cold molecular gas. In projection, the optical emission line nebula is

co-spatial and morphologically matched with CO(21) emission

detected at a significance between 3 (in the outer filaments) and
20 (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(21) line flux corresponds to a molecular hydrogen
gas mass of MH 2 = (1.8 0.2) 109M , where M is the mass of the
Sun. The critical (minimum) density for CO(21) 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.
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(21) 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 +335kms1 relative to the systemic
(stellar) velocity of the galaxy, all within an angular (physical) region
of 0.7150.533 (1kpc0.8kpc).
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 2mJy) of the millimetre synchrotron continuum
source, which serves as a bright backlight (13.6mJy at rest-frame
230GHz). The synchrotron continuum is emitted by jets launched from
the accreting supermassive (~3108M; ref. 13) black hole in the galaxys 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 +300kms1 towards the black hole at its core.

Yale Center for Astronomy and Astrophysics, Yale University, 52 Hillhouse Avenue, New Haven, Connecticut 06511, USA. 2European Southern Observatory, Karl-Schwarzschild-Strasse 2, 85748
Garching bei Mnchen, Germany. 3ASTRON, Netherlands Institute for Radio Astronomy, PO Box 2, 7990 AA Dwingeloo, The Netherlands. 4Leiden Observatory, Leiden University, Niels
Borhweg 2, NL-2333 CA Leiden, The Netherlands. 5Laboratory for Studies of Radiation and Matter in Astrophysics and Atmospheres (LERMA), Observatoire de Paris, PSL Research University,
College de France, CNRS, Sorbonne University, Paris, France. 6Department of Physics and Astronomy, University of Manitoba, Winnipeg, Manitoba R3T 2N2, Canada. 7School of Physics and
Astronomy, Rochester Institute of Technology, 84 Lomb Memorial Drive, Rochester, New York 14623, USA. 8Chester F. Carlson Center for Imaging Science, Rochester Institute of Technology,
84 Lomb Memorial Drive, Rochester, New York 14623, USA. 9Physics and Astronomy Department, Michigan State University, East Lansing, Michigan 48824-2320, USA. 10Physics and Astronomy
Department, Waterloo University, 200 University Avenue West, Waterloo, Ontario N2L 2GL, Canada. 11School of Physics and Astronomy, Cardiff University, The Parade, Cardiff CF24 3AA, UK.
12
Kavli Institute for Astrophysics and Space Research, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, USA. 13Department of Physics,
Durham University, Durham DHL 3LE, UK. 14Naval Research Laboratory Remote Sensing Division, Code 7213, 4555 Overlook Avenue Southwest, Washington DC 20375, USA. 15Instituto de
Radioastronoma y Astrofsica, UNAM, Apartado Postal 3-72 (Xangari), 58089 Morelia, Michoacn, Mexico. 16H. W. Wills Physics Laboratory, University of Bristol, Tyndall Avenue, Bristol BS8
1TL, UK. 17Institute of Astronomy, Cambridge University, Madingly Road, Cambridge CB3 0HA, UK. 18Department of Astronomy, University of Michigan, 1085 South University Avenue, Ann
Arbor, Michigan 48109, USA. 19Laboratoire AIM-Paris-Saclay, CEA/DSM/Irfu CNRSUniversit Paris Diderot, CE-Saclay, F-91191 Gif-sur-Yvette, France. 20Department of Physics and Astronomy,
University of Rochester, Rochester, New York 14627, USA. 21Max-Planck-Institut fr Extraterrestrische Physik, 85748 Garching bei Mnchen, Germany.
2 1 8 | NAT U R E | VO L 5 3 4 | 9 J U N E 2 0 1 6

2016 Macmillan Publishers Limited. All rights reserved

LETTER RESEARCH
a

Rapidly cooling
X-ray atmosphere

7.5 kpc (5)

B: X-ray
Y: Optical
R: H

Abell 2597 (z = 0.082)

Brightest Cluster Galaxy
and surrounding environment

Buoyant X-ray bubbles

H contours

Radio
contours

30 kpc (20)

b Ly

8.4 GHz contours

c FUV cont. (unsharp mask)

8.4 GHz contours

0.1

0.5

0.9

ALMA integrated >3V CO(21) emission

(Jy per beam km s1)

Cool gas

Young stars

5 kpc (3.3)

Figure 1 | A multiwavelength view of the Abell 2597 BCG. a, Chandra

X-ray, HST and DSS optical, and Magellan H+[Nii] 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; H, Carnegie Observatory/Magellan/W.
Baade Telescope/University of Maryland/M. McDonald). b, HST image
of Ly emission associated with the ionized gas nebula13, with 8.4GHz
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 8.4GHz source are overlaid in red.

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 galaxys core.
Gaussian fits to the spectral absorption features reveal narrow linewidths of v  6kms1, 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 sizelinewidth relations25,26 for giant molecular clouds in the Milky Way, they should have diameters not larger
than ~40pc. If in virial equilibrium, molecular clouds this size would
have masses of the order of 105106 M, and if in rough pressure equilibrium with their ambient multiphase (103107K) environment13, they

Figure 2 | ALMA observation of continuum-subtracted CO(21)

emission in the Abell 2597 BCG. Emission is integrated from
600kms1 to +600kms1 relative to the galaxys systemic velocity.
Channels are binned to 40kms1. Only 3 emission is shown. 8.4GHz
VLA radio contours are overlaid in black, and H 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.

must have high column densities of the order of NH 2 10 22 10 24 cm2

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 0.1 CO(2 1) 0.3 .
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 1.3GHz and 5GHz radio
continuum source down to scales of 25pc, 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 (Hi) has previously been found in absorption against this parsecscale jet, corroborating prior reports of inflowing atomic gas at lower
spatial resolutions16. The inflow velocity of this gas matches that seen
9 J U N E 2 0 1 6 | VO L 5 3 4 | NAT U R E | 2 1 9

2016 Macmillan Publishers Limited. All rights reserved

RESEARCH LETTER
2

b
0.1

Total CO(21)
profile

15
0

10
Beam size
(1 kpc by
0.8 kpc)

0.1

0
25

c
Redshifted continuum absorption
at +240, +275, +335 km s1

12.5

12.5

Flux relative to average continuum (%)

Continuum
absorption
on nucleus

25
500

+500

+1,000

Velocity (km s1)

Figure 3 | Shadows cast by molecular clouds moving towards the

supermassive black hole. a, Continuum-subtracted ALMA CO(21)
spectrum extracted from a central 10kpc region. Brackets mark CO(21)
emission shown in b, where 8.4GHz 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(21) spectrum extracted from
this region co-spatial with the mm and radio core. Absorption lines are
indicated in red.

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 (v310kms1) component co-spatial with the
core that is absent just ~20pc to the northeast, where only a narrow
(v50kms1) Hi line is found at the same redshift. This effectively
requires the inflowing atomic gas to be confined within the innermost
~100pc 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(21)
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 +300kms1 that is conceivably available to
attenuate the continuum signal is confined to the innermost 2kpc about
the nucleus (Fig. 4a, b). The radial dependence of molecular cloud
volume number density within this region is uncertain, but probably
steeper than r1, and likely to be closer to r2 (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 r2, a
cloud 100pc from the black hole is ten times more likely to cross our
line of sight than a cloud at a galactocentric distance of 1kpc. 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 galaxys outskirts.
The data therefore serve as strong observational evidence for an
inward-moving, clumpy distribution of molecular clouds within a few

CO(21) at 300 km s1

2 kpc

VLBA 1.3 GHz (ref. 17)

50 pc

Molecular
2 absorption

0.01

250
0
+250 +500
Velocity (km s1)
mJy per beam

CO(21) from galaxy centre

+400
Flux density (mJy)

Flux density (mJy)

20

Velocity (km s1)

200
0
+200

CO(21) emission (Jy per beam km s1)

25

400

16 mas 7 mas

Jy per beam km s1 0.02

Core
0

4
Narrow H I
absorption
400

0 +250

Broad H I
6
absorption

0 +250
+600 400
Velocity (km s1)

+600

Figure 4 | Corroborating evidence that the inflowing molecular clouds

must be in close proximity to the black hole. a, CO(21) absorption
spectrum from Fig. 3, with a region of emission at about +300kms1
marked in yellow. b, Integrated CO(21) emission (colour coded) from
this region, showing that gas at about +300kms1 is confined to the
innermost 2kpc of the galaxy. c, 1.3GHz radio continuum source from an
archival VLBA observation17 with an extremely high physical resolution
of ~25pc by ~10pc. d, e, Plots of 1.3GHz radio continuum emission
revealing Hi 21cm absorption observed against this synchrotron jet.
The signal varies dramatically over scales of tens of parsecs.

hundred parsecs of an accreting supermassive black hole. The result

augments a small but growing set of known molecular absorption
systems2729 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 105106 M each). If they are
falling directly towards the black hole, rather than bound in a noncircular orbit that tightly winds around it, they could supply an upperlimit 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 cloudcloud 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.
Online Content Methods, along with any additional Extended Data display items and
Source Data, are available in the online version of the paper; references unique to
these sections appear only in the online paper.
Received 17 December 2015; accepted 22 March 2016.
1. McNamara, B. R. & Nulsen, P. E. J. Heating hot atmospheres with active
galactic nuclei. Annu. Rev. Astron. Astrophys. 45, 117175 (2007).
2. McNamara, B. R. & Nulsen, P. E. J. Mechanical feedback from active
galactic nuclei in galaxies, groups and clusters. New J. Phys. 14, 055023
(2012).
3. Fabian, A. C. Observational evidence of active galactic nuclei feedback.
Annu. Rev. Astron. Astrophys. 50, 455489 (2012).
4. Bondi, H. On spherically symmetrical accretion. Mon. Not. R. Astron. Soc. 112,
195204 (1952).
5. Pizzolato, F. & Soker, N. On the nature of feedback heating in cooling flow
clusters. Astrophys. J. 632, 821830 (2005).
6. Voit, G. M., Donahue, M., Bryan, G. L. & McDonald, M. Regulation of star
formation in giant galaxies by precipitation, feedback and conduction. Nature
519, 203206 (2015).
7. Voit, G. M., Bryan, G. L., OShea, B. W. & Donahue, M. Precipitation-regulated
star formation in galaxies. Astrophys. J. 808, L30 (2015).

2 2 0 | NAT U R E | VO L 5 3 4 | 9 J U N E 2 0 1 6

2016 Macmillan Publishers Limited. All rights reserved

LETTER RESEARCH
8. Sharma, P., McCourt, M., Quataert, E. & Parrish, I. J. Thermal instability and the
feedback regulation of hot haloes in clusters, groups, and galaxies. Mon. Not. R.
Astron. Soc. 420, 31743194 (2012).
9. Gaspari, M., Ruszkowski, M. & Oh, S. P. Chaotic cold accretion on to black holes.
Mon. Not. R. Astron. Soc. 432, 34013422 (2013).
10. Li, Y. & Bryan, G. L. Modeling active galactic nucleus feedback in cool-core
clusters: the formation of cold clumps. Astrophys. J. 789, 153 (2014).
11. McNamara, B. R. et al. Discovery of ghost cavities in the X-ray atmosphere of
Abell 2597. Astrophys. J. 562, L149L152 (2001).
12. Clarke, T. E., Sarazin, C. L., Blanton, E. L., Neumann, D. M. & Kassim, N. E.
Low-frequency radio observations of X-ray ghost bubbles in A2597: a history
of radio activity in the core. Astrophys. J. 625, 748753 (2005).
13. Tremblay, G. R. et al. Multiphase signatures of active galactic nucleus feedback
in Abell 2597. Mon. Not. R. Astron. Soc. 424, 10261041 (2012).
14. Gaspari, M., Brighenti, F. & Temi, P. Chaotic cold accretion on to black holes in
rotating atmospheres. Astron. Astrophys. 579, A62 (2015).
15. Tremblay, G. R. et al. Residual cooling and persistent star formation amid
active galactic nucleus feedback in Abell 2597. Mon. Not. R. Astron. Soc. 424,
10421060 (2012).
16. ODea, C. P., Baum, S. A. & Gallimore, J. F. Detection of extended H I absorption
toward PKS 2322123 in Abell 2597. Astrophys. J. 436, 669677 (1994).
17. Taylor, G. B., ODea, C. P., Peck, A. B. & Koekemoer, A. M. H I Absorption toward
the nucleus of the radio galaxy PKS 2322123 in A2597. Astrophys. J. 512,
L27L30 (1999).
18. ODea, C. P., Baum, S. A., Mack, J., Koekemoer, A. M. & Laor, A. Hubble Space
Telescope STIS far-ultraviolet observations of the central nebulae in the
cooling-core clusters A1795 and A2597. Astrophys. J. 612, 131151 (2004).
19. Oonk, J. B. R., Hatch, N. A., Jaffe, W., Bremer, M. N. & van Weeren, R. J.
Far-ultraviolet emission in the A2597 and A2204 brightest cluster galaxies.
Mon. Not. R. Astron. Soc. 414, 23092336 (2011).
20. Tremblay, G. R. et al. Far-ultraviolet morphology of star-forming filaments in
cool core brightest cluster galaxies. Mon. Not. R. Astron. Soc. 451, 37683800
(2015).
21. Salom, P. et al. A very extended molecular web around NGC 1275. Astron.
Astrophys. 531, A85 (2011).
22. Jaffe, W., Bremer, M. N. & Baker, K. H II and H2 in the envelopes of cooling flow
central galaxies. Mon. Not. R. Astron. Soc. 360, 748762 (2005).
23. Salom, P. et al. Cold molecular gas in the Perseus cluster core. Association
with X-ray cavity, H filaments and cooling flow. Astron. Astrophys. 454,
437445 (2006).
24. Smith, E. P., Heckman, T. M. & Illingworth, G. D. Stellar dynamics of powerful
radio galaxies. Astrophys. J. 356, 399415 (1990).
25. Larson, R. B. Turbulence and star formation in molecular clouds. Mon. Not. R.
Astron. Soc. 194, 809826 (1981).
26. Solomon, P. M., Rivolo, A. R., Barrett, J. & Yahil, A. Mass, luminosity, and line
width relations of Galactic molecular clouds. Astrophys. J. 319, 730741
(1987).

27. Wiklind, T. & Combes, F. Molecular absorption and its time variations in
Centaurus A. Astron. Astrophys. 324, 5164 (1997).
28. Espada, D. et al. Disentangling the circumnuclear environs of Centaurus A.
II. On the nature of the broad absorption line. Astrophys. J. 720, 666678 (2010).
29. David, L. P. et al. Molecular gas in the X-ray bright group NGC 5044 as revealed
by ALMA. Astrophys. J. 792, 94 (2014).
30. Urry, C. M. & Padovani, P. Unified schemes for radio-loud active galactic nuclei.
Publ. Astron. Soc. Pacif. 107, 803845 (1995).
Acknowledgements ALMA is a partnership of ESO (representing its member
states), NSF (USA) and NINS (Japan), together with NRC (Canada) and NSC
and ASIAA (Taiwan), in cooperation with the Republic of Chile. The Joint
ALMA Observatory is operated by ESO, AUI/NRAO and NAOJ. We are grateful
to the European ALMA Regional Centres, particularly those in Garching and
Manchester, for their dedicated end-to-end support of data associated with
this paper. We thank R. Larson for discussions. G.R.T. acknowledges support
from National Aeronautics and Space Administration (NASA) through Einstein
Postdoctoral Fellowship Award Number PF-150128, issued by the Chandra
X-ray Observatory Center, which is operated by the Smithsonian Astrophysical
Observatory for and on behalf of NASA under contract NAS8-03060.
F.C. acknowledges the European Research Council (ERC) for the Advanced
Grant Program no. 267399-Momentum. B.R.M. is supported by a grant from
the Natural Sciences and Engineering Research Council of Canada. T.A.D.
acknowledges support from a Science and Technology Facilities Council (STFC)
Ernest Rutherford Fellowship. A.C.E. acknowledges support from STFC grant
ST/L00075X/1. A.C.F. and H.R.R. acknowledge support from ERC Advanced
Grant Program no. 340442-Feedback. M.N.B. acknowledges funding from the
STFC. Basic research in radio astronomy at the Naval Research Laboratory is
supported by 6.1 Base funding.
Author Contributions G.R.T. was principal investigator on the original proposal,
performed the data analysis, and wrote the paper. J.B.R.O., T.A.D., R.G.M. and
A.M. were substantially involved in planning both scientific and technical
aspects of the proposal, while T.A.D. and R.G.M. contributed ALMA data
reduction and analysis expertise once the data were obtained. J.B.R.O.,
F.C. and P.S. invested substantial time in analysis of the data. Substantial
scientific feedback was also provided over many months by F.C., J.B.R.O., C.P.O.,
S.A.B., G.M.V., M.D., B.R.M., M.A.M., T.E.C., H.R., A.C.E. and A.C.F., while all other
co-authors discussed the results and commented on the manuscript.
Author Information This paper makes use of the following ALMA data:
ADS/JAO.ALMA#2012.1.00988.S. Reprints and permissions information
is available at www.nature.com/reprints. The authors declare no competing
financial interests. Readers are welcome to comment on the online version of
the paper. Correspondence and requests for materials should be addressed to
G.R.T. (grant.tremblay@yale.edu).

9 J U N E 2 0 1 6 | VO L 5 3 4 | NAT U R E | 2 2 1

2016 Macmillan Publishers Limited. All rights reserved

RESEARCH LETTER
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 12m
Array. ALMAs Band 6 heterodyne receivers were tuned to a frequency of 213GHz,
sensitive to the J=21 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 1,875MHz (per baseband) with a 0.488MHz
channel spacing, for a maximum spectral resolution of about 2km s1. One baseband
was centred on the CO(21) emission line, while the other three sampled the local
continuum. Maximum antenna baselines extended to ~1km, delivering an angular
resolution at 213GHz of ~0.7 within a ~28 primary beam (field of view). ALMA
observed the Abell 2597 BCG, located at RA 23h 25min 20s, dec. 12 07 38
(J2000), for a total of ~3h over three separate scheduling blocks executed between
17 and 19 November 2013. The planet Neptune and quasars J22582758 and
J23311556 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 galaxys own continuum, yielding a
~14% decrease in RMS noise to a final value of 0.16mJy per 0.7150.533 beam
per 40kms1 channel. There is effectively no difference in CO(21) morphology
between the self-calibrated and non-self-calibrated data cubes. Measurement sets were
imaged using natural visibility weighting and binning to either 5km s1, 10kms1,
or 40kms1, as indicated in the figure legends. The figures presented in this Letter
show only continuum-subtracted, pure CO(21) line emission. The rest-frame
230GHz continuum observation is dominated by a bright (13.6mJy) point source
associated with the AGN (detected at 400), serving as the bright backlight against
which the continuum absorption features presented in this Letter were observed. The
continuum data also features compact (~5kpc) extended emission at ~10 that
extends along the galaxys 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(21) line
velocities discussed in this Letter are set relative to 213.04685GHz, where observed
CO(21) emission peaks. This frequency corresponds to 12CO(21) (rest-frame
230.538001GHz) at a redshift of z=0.0821. This redshift is consistent, conservatively
within 60kms1, with every other available multiwavelength tracer of the galaxys
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 412km s1) Hi absorption component16 at the
optical emission and absorption line redshift. It is also consistent, within ~60km s1,
with a cross-correlation of emission and absorption lines using galaxy template
spectra17, as well as with all other published reports of the galaxys 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 galaxys stellar component. Without caveat or ambiguity,
the absorbing cold clouds are moving into the galaxy at roughly ~30060kms1.
Mass estimates. All molecular gas masses estimated in this letter adopt the
following relation32:

1.05 10 4
X CO

Mmol =
cm2
3.2

20
2 10 K km s1

1 S v D 2

L
CO
1 + z Jy km s1 Mpc M

where SCOv 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 CDM model wherein H0=70kms1
Mpc1, M=0.3, and =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=21020cm2 (K km s1)1 and a CO(21) to
CO(10) 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(21) spectrum extracted from an aperture containing all detected emission yields an emission integral of S COv = 4.2 0.4 Jy
km s1 with a line FWHM of 25214kms1, corresponding to a total molecular
hydrogen (H2) gas mass of M H 2 = (1.80 0.19) 109 M. This is very close to
the previously reported15 mass, based on an IRAM 30m CO(21) observation, of
(1.8 0.3)109 M. 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(21)/CO(10) flux ratio of 4 (rather than
3.2, as we use here). These differences are minor, particularly because nearly all of
the CO(21) 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 sizelinewidth relation for
molecular clouds in the Milky Way (namely, the ref. 26 fit);

v = (1.0 0.1)S 0.5 0.05 km s1

where v is the velocity linewidth of the cloud and S is the diameter of the cloud in
parsecs. A measured absorber linewidth of v6kms1 would then correspond
to a size of ~36pc. 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 tensrather than hundredsof parsecs in diameter).
The three clouds are separated from one another by ~4560kms1 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,

Mcloud

R cloudv2
20pc (6 km s1)2

1.7 105 M
1
1 2
G
4.302 103 pc M
(km s )

where Rcloud is the cloud radius (as roughly estimated above) and v is its velocity
dispersion (also as above).
CO(21) optical depths for the absorbers were estimated by assuming that:

Itotal = Icontinuume CO(21)

where Itotal and Icontinuum are the integrated intensities of the total (line plus continuum) and continuum-only signals, respectively, and CO(21) is the optical depth
of the CO(21) absorption feature.
The stellar velocity dispersion of the BCG33 is v=220 19kms1. Under the
assumption of an isothermal sphere, the circular velocity should be ~300 km s1
(that is, 2 v), 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:

r 2 NH

M 109M fc
10 22cm2
1 kpc
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 1kpc, and a column density of 1022cm2, then the total mass of molecular hydrogen contained within that shell would be roughly one billion solar masses.
A column density in excess of 1022cm2 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 1023cm2,
for example, this mass must be contained within a sphere of radius 300 pc, or else its
total mass would exceed the 1.8109 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 Pythons
MatPlotLib and Veusz, which is available at http://home.gna.org/veusz/.
31. Voit, G. M. & Donahue, M. A deep look at the emission-line nebula in Abell
2597. Astrophys. J. 486, 242 (1997).
32. Bolatto, A. D., Wolfire, M. & Leroy, A. K. The CO-to-H2 conversion factor. Annu.
Rev. Astron. Astrophys. 51, 207268 (2013).
33. McNamara, B. R. et al. A 1010 solar mass flow of molecular gas in the A1835
brightest cluster galaxy. Astrophys. J. 785, 44 (2014).
34. Russell, H. R. et al. Massive molecular gas flows in the A1664 brightest cluster
galaxy. Astrophys. J. 784, 78 (2014).

2016 Macmillan Publishers Limited. All rights reserved