UC Berkeley
UC Berkeley Previously Published Works
Title
A direct localization of a fast radio burst and its host
Permalink
https://escholarship.org/uc/item/8131z4sx
Journal
Nature, 541(7635)
ISSN
0028-0836 1476-4687
Authors
Chatterjee, S.
Law, C. J
Wharton, R. S
et al.
Publication Date
2017-01-04
DOI
10.1038/nature20797
Peer reviewed
eScholarship.org Powered by the California Digital Library
University of California
58 | NATURE | VOL 541 | 5 JANUARY 2017
LETTER
doi:10.1038/nature20797
A direct localization of a fast radio burst and its host
S. Chatterjee
1
, C. J. Law
2
, R. S. Wharton
1
, S. Burke-Spolaor
3,4,5
, J. W. T. Hessels
6,7
, G. C. Bower
8
, J. M. Cordes
1
,
S. P. Tendulkar
9
, C. G. Bassa
6
, P. Demorest
3
, B. J. Butler
3
, A. Seymour
10
, P. Scholz
11
, M. W. Abruzzo
12
, S. Bogdanov
13
,
V. M. Kaspi
9
, A. Keimpema
14
, T. J. W. Lazio
15
, B. Marcote
14
, M. A. McLaughlin
4,5
, Z. Paragi
14
, S. M. Ransom
16
,
M. Rupen
11
, L. G. Spitler
17
& H. J. van Langevelde
14,18
Fast radio bursts
1,2
are astronomical radio flashes of unknown
physical nature with durations of milliseconds. Their dispersive
arrival times suggest an extragalactic origin and imply radio
luminosities that are orders of magnitude larger than those of all
known short-duration radio transients
3
. So far all fast radio bursts
have been detected with large single-dish telescopes with arcminute
localizations, and attempts to identify their counterparts (source
or host galaxy) have relied on the contemporaneous variability of
field sources
4
or the presence of peculiar field stars
5
or galaxies
4
.
These attempts have not resulted in an unambiguous association
6,7
with a host or multi-wavelength counterpart. Here we report the
subarcsecond localization of the fast radio burst FRB121102, the
only known repeating burst source
8–11
, using high-time-resolution
radio interferometric observations that directly image the bursts.
Our precise localization reveals that FRB121102 originates within
100 milliarcseconds of a faint 180-microJansky persistent radio
source with a continuum spectrum that is consistent with non-
thermal emission, and a faint (twenty-fifth magnitude) optical
counterpart. The flux density of the persistent radio source
varies by around ten per cent on day timescales, and very long
baseline radio interferometry yields an angular size of less than
1.7 milliarcseconds. Our observations are inconsistent with the
fast radio burst having a Galactic origin or its source being located
within a prominent star-forming galaxy. Instead, the source appears
to be co-located with a low-luminosity active galactic nucleus or
a previously unknown type of extragalactic source. Localization
and identification of a host or counterpart has been essential to
understanding the origins and physics of other kinds of transient
events, including gamma-ray bursts
12,13
and tidal disruption
events
14
. However, if other fast radio bursts have similarly faint
radio and optical counterparts, our findings imply that direct
subarcsecond localizations may be the only way to provide reliable
associations.
The repetition of bursts from FRB121102
9,10
enabled a targeted
interferometric localization campaign with the Karl G. Jansky Very
Large Array (VLA) in concert with single-dish observations using
the 305-m William E. Gordon Telescope at the Arecibo Observatory.
We searched for bursts in VLA data with 5-ms sampling using both
beam-forming and imaging techniques
15
(see Methods). In over
83 h of VLA observations distributed over six months, we detected
nine bursts from FRB121102 in the 2.5–3.5-GHz band with signal-
to-noise ratios ranging from 10 to 150, all at a consistent sky position.
These bursts were initially detected with real-time de-dispersed imag-
ing and confirmed by a beam-formed search (Fig. 1). From these
detections, the average J2000 position of the burst source is right
ascension α = 05 h31 min58.70s, declination δ = + 33°08′ 52.5″ , with
a 1σ uncertainty of about 0.1″ , consistent with the Arecibo localization
9
but with three orders of magnitude better precision. The dispersion
measure (DM) for each burst is consistent with the previously reported
value
9
of 558.1 ± 3.3 pccm
−3
, with comparable uncertainties. Three
bursts detected at the VLA (2.5–3.5 GHz) had simultaneous coverage
at Arecibo (1.1–1.7 GHz). After accounting for dispersion delay and
light travel time, one burst is detected at both telescopes (Extended
Data Table 1), but the other two show no emission in the Arecibo
band, implying frequency structure at scales of approximately 1 GHz.
This finding provides new constraints on the broadband burst spectra,
which previously have shown highly variable structure across the
Arecibo band
8–10
.
Radio images at 3GHz produced by integrating the VLA fast- sampled
data reveal a continuum source within 0.1″ of the burst position,
which we refer to hereafter as the persistent source. A cumulative
3-GHz image (root-mean-square (r.m.s.) of σ ≈ 2 μ Jy per beam; Fig. 2)
shows 68 other sources within a 5′ radius, with a median flux density of
26 μ Jy. Given the agreement between the positions of the detected bursts
and the continuum counterpart, we estimate a probability of chance
coincidence of less than 10
−5
. The persistent source is detected in
follow-up VLA observations over the entire frequency range from
1 GHz to 26 GHz. The radio spectrum is broadly consistent with
non-thermal emission, although with deviations from a single
power-law spectrum. Imaging at 3 GHz over the campaign shows that
the persistent source exhibits around 10% variability on day timescales
(Fig. 2, Extended Data Table 2). Variability in faint radio sources is
common
6,7
; of the 69 sources within a 5′ radius, nine (including the
persistent counterpart) were apparently variable (see Methods). There
is no correlation between VLA detections of bursts from FRB121102
and the flux density of the counterpart at that epoch (Fig. 2, Methods).
Observations with the European Very Long Baseline Interferometry
(VLBI) Network and the Very Long Baseline Array detect the persistent
source and limit its size to less than 1.7milliarcseconds (see Methods).
The lower limit on the brightness temperature is 8 × 10
6
K. The source
has an integrated flux density that is consistent with that inferred at
lower resolution in contemporaneous VLA imaging, indicating the
absence of any detectable flux density on spatial scales larger than a
few milliarcseconds.
1
Cornell Center for Astrophysics and Planetary Science and Department of Astronomy, Cornell University, Ithaca, New York 14853, USA.
2
Department of Astronomy and Radio Astronomy Lab,
University of California, Berkeley, California 94720, USA.
3
National Radio Astronomy Observatory, Socorro, New Mexico 87801, USA.
4
Department of Physics and Astronomy, West Virginia
University, Morgantown, West Virginia 26506, USA.
5
Center for Gravitational Waves and Cosmology, West Virginia University, Chestnut Ridge Research Building, Morgantown, West Virginia 26505,
USA.
6
ASTRON, Netherlands Institute for Radio Astronomy, Postbus 2, 7990 AA Dwingeloo, The Netherlands.
7
Anton Pannekoek Institute for Astronomy, University of Amsterdam, Science Park
904, 1098 XH Amsterdam, The Netherlands.
8
Academia Sinica Institute of Astronomy and Astrophysics, 645 North A’ohoku Place, Hilo, Hawaii 96720, USA.
9
Department of Physics and McGill
Space Institute, McGill University, 3600 University Street, Montreal, Quebec H3A 2T8, Canada.
10
Arecibo Observatory, HC3 Box 53995, Arecibo, Puerto Rico 00612, USA.
11
National Research
Council of Canada, Herzberg Astronomy and Astrophysics, Dominion Radio Astrophysical Observatory, PO Box 248, Penticton, British Columbia V2A 6J9, Canada.
12
Haverford College,
370 Lancaster Avenue, Haverford, Pennsylvania 19041, USA.
13
Columbia Astrophysics Laboratory, Columbia University, New York, New York 10027, USA.
14
Joint Institute for VLBI ERIC,
Postbus 2, 7990 AA Dwingeloo, The Netherlands.
15
Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California 91109, USA.
16
National Radio Astronomy Observatory,
Charlottesville, Virginia 22903, USA.
17
Max-Planck-Institut für Radioastronomie, Auf dem Hügel 69, Bonn D-53121, Germany.
18
Sterrewacht Leiden, Leiden University, Postbus 9513,
2300 RA Leiden, The Netherlands.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Letter
reSeArCH
5 JANUARY 2017 | VOL 541 | NATURE | 59
We have searched for counterparts at submillimetre, infrared,
optical and X-ray wavelengths using archival data and a series of new
observations. A coincident unresolved optical source is detected in archi-
val 2014 Keck data (R-band AB magnitude of 24.9 ± 0.1) and in recently
obtained Gemini data (r-band AB magnitude of 25.1 ± 0.1; Fig. 2),
with a chance coincidence probability of less than 3.5 × 10
−4
(see Methods). The source is undetected in archival infrared
observations, in ALMA 230-GHz observations, and in XMM-Newton
and Chandra X-ray imaging (see Methods). The spectral energy
distribution of the persistent source is compared in Fig. 3 to some
example spectra for known source types, none of which matches our
observations well.
The observations reported here corroborate the strong arguments
10
against a Galactic location for the source. As argued previously, stellar
radio flares can exhibit swept-frequency radio bursts on subsecond
timescales
16
, but they do not strictly adhere to the ν
−2
dispersion
law (where ν is the frequency) seen for FRB121102
9,10
, nor are they
expected to show constant apparent DM. The source of the sizable DM
excess—three times the Galactic maximum predicted by the NE2001
electron-density model
17
—is not revealed as a H region, a super-
nova remnant or a pulsar-wind nebula in our Galaxy, which would
appear extended at radio, infrared or Hα
10
wavelengths at our localized
position. Spitzer mid-infrared limits constrain substellar objects with
temperatures of more than 900 K to be at distances of 70 pc or greater,
and the Gemini detection sets a minimum distance of about 1 kpc and
100 kpc for stars with effective temperatures greater than 3,000 K and
5,000 K, respectively. These limits rule out Galactic stars that could
plausibly account for the DM excess and produce the radio continuum
counterpart. We conclude that FRB121102 and its persistent counter-
part do not correspond to any known class of Galactic source.
The simplest interpretation is that the burst source resides in a host
galaxy that also contains the persistent radio counterpart. If so, the
DM of the burst source has contributions from the electron density in
the Milky Way disk (DM
NE2001
) and halo (DM
halo
)
17
, the intergalactic
medium (DM
IGM
)
18
and the host galaxy (DM
host
); we estimate DM
IGM
= DM− DM
NE2001
− DM
halo
− DM
host
≈ 340 pccm
−3
− DM
host
, with
DM
NE2001
= 188 pccm
−3
and DM
halo
≈ 30 pccm
−3
. The maximum
redshift of the fast radio burst, for DM
host
= 0, is z
FRB
≈ 0.32, which
corresponds to a maximum luminosity distance of 1.7 Gpc. Variance
in the mapping of DM to redshift
19
(σ
z
= σ
DM
(dz/dDM) ≈ 0.1) could
increase the upper bound to z ≈ 0.42. Alternatively, a sizable host-galaxy
contribution could imply a low redshift and a negligible contribution
from the intergalactic medium, although no such galaxy is apparent.
Hereafter we adopt z
FRB
≲ 0.32.
The faint optical detection and the non-detection at 230 GHz with
ALMA imply a low star-formation rate within any host galaxy. For our
ALMA 3σ upper limit of 51 μ Jy and a submillimetre spectral index of 4,
we estimate the star-formation rate
20
to be less than (0.06–19)M
⊙
yr
−1
(where M
⊙
is the mass of the Sun) for redshifts z ranging from 0.01
to 0.32 (luminosity distances of 43 Mpc to 1.7 Gpc), respectively. The
implied absolute magnitude of approximately − 16 a t z = 0.32 is similar
to that of the Small Magellanic Cloud, whose mass of around 10
9
M
⊙
would correspond to an upper limit on the mass of the host galaxy.
The compactness of the persistent radio source (less than about 8 pc
for z ≲ 0.32) implies that it does not correspond to emission from an
extended galaxy or a star-forming region
21
, although our brightness
temperature limits do not require the emission to be coherent. Its size
and spectrum appear consistent with a low-luminosity active galactic
nucleus (AGN), but X-ray limits do not support this interpretation.
Young extragalactic supernova remnants
22
can have brightness
temperatures in excess of 10
7
K, but they typically have simple
power-law spectra and exhibit stronger variability.
The burst source and persistent source have a projected separation of
less than about 500 pc assuming z ≲ 0.32. There are three broad inter-
pretations of their relationship. First, they may be unrelated objects
harboured in a host galaxy, such as a neutron star (or other compact
object) and an AGN. Alternatively, the two objects may interact, for
example, producing repeated bursts from a neutron star very close to
an AGN
3,23,24
. A third possibility is that they are a single source. This
possibility could involve unprecedented bursts from an AGN
25
along
with persistent synchrotron radiation; or persistent emission might
comprise high-rate bursts too weak to detect individually, with bright
detectable bursts forming a long tail of the amplitude distribution.
In this interpretation, the difficulty in establishing any periodicity
in the observed bursts
9,10
may result from irregular beaming from a
rotating compact object or extreme spin or orbital dynamics. The Crab
pulsar and some millisecond pulsars display bimodality
26,27
in giant and
regular pulses. However, they show well-defined periodicities and have
steep spectra that are inconsistent with the spectrum of the persistent
source, which extends to at least 25 GHz. Magnetars show broad spectra
that extend beyond 100 GHz in a few cases, but differ from the roll-off
of the spectrum of the persistent source.
–0.05 0.00 0.05 0.10 0.15 0.20
Time offset (s)
Right ascension offset (arcsec)
Right ascension offset (arcmin)
2,600
2,800
3,000
3,200
3,400
Frequency (MHz)
0
100
200
300
400
Signal-to-noise ratio
De-dispersed time series
–5 0 5 10 15 20 25 30
35
Signal-to-noise ratio
Pulse spectrum
b
ac
4
3
2
1
0
–1
–2
–3
–4
43210–1–2–3–4
10
5
0
–5
–10
10 50–5 –10
Declination offset (arcsec) Declination offset (arcmin)
Figure 1 | VLA detection of FRB121102. a, A 5-ms dispersion-corrected
dirty image showing a burst from FRB121102 at MJD 57633.67986367
(2016 September 2). The approximate localization uncertainty from
previous Arecibo detections
9
(3′ beam full-width at half-maximum
(FWHM)) is shown with overlapping circles. b, A zoomed-in portion of a,
deconvolved and re-centred on the detection, showing the approximately
0.1″ localization of the burst. c, Time–frequency data extracted from
phased VLA visibilities at the burst location shows the ν
−2
dispersive
sweep of the burst. The solid black lines illustrate the expected sweep for
DM = 558 pccm
−3
. The de-dispersed lightcurve and spectra are projected
to the upper and right panels, respectively. In all panels, the colour scale
indicates the flux density.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Letter
reSeArCH
60 | NATURE | VOL 541 | 5 JANUARY 2017
All things considered, we cannot favour any one of these interpre-
tations. Future comparison of spectra from the persistent source and
from individual bursts could rule out the ‘single source’ interpreta-
tion. The proximity of the two sources and their physical relationship
can be probed by detecting a burst in VLBI observations or by using
interstellar scintillations, which can resolve separations of less than
one milliarcsecond.
If other fast radio bursts are similar to FRB121102, then our dis-
covery implies that direct subarcsecond localizations of bursts are so
far the only secure way to find associations. The unremarkable nature
of the counterparts to FRB121102 suggests that efforts to identify the
counterparts of other fast radio bursts in large error boxes will be dif-
ficult and, given the lack of correlation between the variability of the
persistent source and the bursts, rapid post-fast-radio-burst follow-up
imaging in general may not be fruitful.
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 1 November; accepted 16 November 2016.
1. Lorimer, D. R., Bailes, M., McLaughlin, M. A., Narkevic, D. J. & Crawford, F.
A bright millisecond radio burst of extragalactic origin. Science 318, 777–780
(2007).
2. Thornton, D. et al. A population of fast radio bursts at cosmological distances.
Science 341, 53–56 (2013).
3. Cordes, J. M. & Wasserman, I. Supergiant pulses from extragalactic neutron
stars. Mon. Not. R. Astron. Soc. 457, 232–257 (2016).
4. Keane, E. F. et al. The host galaxy of a fast radio burst. Nature 530, 453–456
(2016).
5. Loeb, A., Shvartzvald, Y. & Maoz, D. Fast radio bursts may originate from
nearby aring stars. Mon. Not. R. Astron. Soc. 439, L46–L50 (2014).
6. Williams, P. K. G. & Berger, E. No precise localization for FRB150418: claimed
radio transient is AGN Variability. Astrophys. J. 821, L22 (2016).
7. Vedantham, H. K. et al. On associating fast radio bursts with afterglows.
Astrophys. J. 824, L9 (2016).
8. Spitler, L. G. et al. Fast radio burst discovered in the Arecibo pulsar ALFA
survey. Astrophys. J. 790, 101 (2014).
9. Spitler, L. G. et al. A repeating fast radio burst. Nature 531, 202–205 (2016).
10. Scholz, P. et al. The repeating fast radio burst FRB121102: multi-wavelength
observations and additional bursts. Astrophys. J. (in the press); preprint at
https://arxiv.org/abs/1603.08880.
11. Petro, E. et al. A survey of FRB elds: limits on repeatability. Mon. Not. R.
Astron. Soc. 454, 457–462 (2015).
12. Metzger, M. R. et al. Spectral constraints on the redshift of the optical
counterpart to the γ -ray burst of 8 May 1997. Nature 387, 878–880 (1997).
13. Bloom, J. S., Kulkarni, S. R. & Djorgovski, S. G. The observed oset distribution
of gamma-ray bursts from their host galaxies: a robust clue to the nature of
the progenitors. Astron. J 123, 1111–1148 (2002).
14. Gezari, S. et al. Ultraviolet detection of the tidal disruption of a star by a
supermassive black hole. Astrophys. J. 653, L25–L28 (2006).
15. Law, C. J. et al. A millisecond interferometric search for fast radio bursts with
the Very Large Array. Astrophys. J. 807, 16 (2015).
16. Maoz, D. et al. Fast radio bursts: the observational case for a Galactic origin.
Mon. Not. R. Astron. Soc. 454, 2183–2189 (2015).
17. Cordes, J. M. & Lazio, T. J. W. NE2001.I. A new model for the galactic
distribution of free electrons and its uctuations. Preprint at http://arxiv.org/
abs/astro-ph/0207156 (2002).
18. Inoue, S. Probing the cosmic reionization history and local environment of
gamma-ray bursts through radio dispersion. Mon. Not. R. Astron. Soc. 348,
999–1008 (2004).
19. McQuinn, M. Locating the “missing” baryons with extragalactic dispersion
measure estimates. Astrophys. J. 780, L33 (2014).
5 h 32 min 12 s
5 10 15 20 25 30 35
125
150
175
200
225
250
Flux density (μJy)
Observed
Average
FRB detection
125 130 135 140 145 150
MJD – 57,500 (days)
a
b
Declination
Right ascension
33° 12′
33° 9′
33° 6′
5 h 31 min 36
s
5 h 31 min 48 s5 h 32 min 0 s
Figure 2 | Radio and optical images of the FRB121102 field. a, VLA
image at 3 GHz with a combination of array configurations. The image
resolution is 2″ and the r.m.s. is σ = 2 μ Jy per beam. The Arecibo detection
9
uncertainty regions (3′ beam FWHM) are indicated with overlapping
white circles. The radio counterpart of the bursts detected at the VLA is
highlighted by a 20″ white square within the overlap region. The colour
scale indicates the observed flux density. Inset, Gemini r-band image of
the 20″ square shows an optical counterpart (r
AB
= 25.1 ± 0.1mag), as
identified by the 5″ bars. b, The light curve of the persistent radio source
coincident with FRB121102 over the course of the VLA campaign,
indicating variability on timescales shorter than 1day. Error bars are 1σ.
The average flux density of the source of about 180 μ Jy is marked in grey,
and the epochs at which bursts were detected at the VLA are indicated
(red triangles). The variability of the persistent radio counterpart is
uncorrelated with the detection of bursts (see Methods).
10
9
10
10
10
11
10
12
10
13
10
14
10
15
10
16
10
17
10
18
Frequency (Hz)
10
–18
10
–17
10
–16
10
–15
10
–14
10
–13
QF
Q
(erg cm
–2
s
–1
)
VLA
ALMA
GLIMPSE
UKIDSS
XMM-Newton
and Chandra
Gemini i-band
Keck R-band
Gemini r-band
Henize 2-10
Radio-loud AGN
Crab nebula
Radio counterpart
Optical counterpart
10
–5
10
–4
10
–3
10
–2
10
–1
10
0
10
1
10
2
10
3
10
4
Energy (eV)
Figure 3 | Broadband spectral energy distribution of the counterpart.
Detections of the persistent radio source (blue circles), the optical
counterpart (red and orange squares) and 5σ upper limits at various
frequency bands (arrows) are shown; see Methods for details. Spectral
energy distributions of other radio point sources are scaled to match the
radio flux density at 10 GHz and overlaid for comparison: low-luminosity
AGN in Henize 2-10, a star-forming dwarf galaxy
28
placed at 25 Mpc
(blue); radio-loud AGN QSO2128− 123
29
scaled by 10
−4.3
to simulate a
lower-luminosity AGN and placed at 3 Gpc (yellow); and the Crab nebula
30
at 4 Mpc (red). F
ν
is the flux density and νF
ν
is the flux density weighted by
photon energy.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Letter
reSeArCH
5 JANUARY 2017 | VOL 541 | NATURE | 61
20. Carilli, C. L. & Yun, M. S. The radio-to-submillimeter spectral index as a redshift
indicator. Astrophys. J. 513, L13–L16 (1999).
21. Condon, J. J. Radio emission from normal galaxies. Annu. Rev. Astron.
Astrophys. 30, 575–611 (1992).
22. Lonsdale, C. J., Diamond, P. J., Thrall, H., Smith, H. E. & Lonsdale, C. J. VLBI
images of 49 radio supernovae in Arp220. Astrophys. J. 647, 185–193 (2006).
23. Pen, U.-L. & Connor, L. Local circumnuclear magnetar solution to extragalactic
fast radio bursts. Astrophys. J. 807, 179 (2015).
24. Lyubarsky, Y. A model for fast extragalactic radio bursts. Mon. Not. R. Astron.
Soc. 442, L9–L13 (2014).
25. Romero, G. E., del Valle, M. V. & Vieyro, F. L. Mechanism for fast radio bursts.
Phys. Rev. D 93, 023001 (2016).
26. Lundgren, S. C. et al. Giant pulses from the crab pulsar: a joint radio and
gamma-ray study. Astrophys. J. 453, 433–445 (1995).
27. Kinkhabwala, A. & Thorsett, S. E. Multifrequency observations of giant radio
pulses from the millisecond pulsar B1937+ 21. Astrophys. J. 535, 365–372
(2000).
28. Reines, A. E. & Deller, A. T. Parsec-scale radio emission from the low-luminosity
active galactic nucleus in the dwarf starburst galaxy Henize 2-10. Astrophys. J.
750, L24 (2012).
29. Elvis, M. et al. Atlas of quasar energy distributions. Astrophys. J. Suppl. Ser. 95,
1–68 (1994).
30. Bühler, R. & Blandford, R. The surprising Crab pulsar and its nebula: a review.
Rep. Prog. Phys. 77, 066901 (2014).
Supplementary Information is available in the online version of the paper.
Acknowledgements The National Radio Astronomy Observatory is a facility
of the National Science Foundation operated under cooperative agreement
by Associated Universities. We thank the staff at the NRAO for their continued
support of these observations, especially with scheduling and computational
infrastructure. The Arecibo Observatory is operated by SRI International
under a cooperative agreement with the National Science Foundation
(AST-1100968), and in alliance with Ana G. Méndez-Universidad Metropolitana
and the Universities Space Research Association. We thank the staff at Arecibo
for their support and dedication that enabled these observations. Further
acknowledgements of telescope facilities and funding agencies are included as
Supplementary Information.
Author Contributions S.C. was principal investigator of the localization
campaign described here. C.J.L. and S.B.-S. are principal investigators of the
realfast project and performed the analysis that achieved the first VLA burst
detections. S.C., C.J.L., R.S.W., S.B.-S., G.C.B., B.B. and P.D. performed detailed
analysis of the VLA data. S.B.-S. and B.B. led the analysis of the VLA multi-
band spectral data. J.W.T.H. was principal investigator of the EVN observations,
which were analysed by Z.P. and B.M. G.C.B. was principal investigator of the
VLBA observations, and led their analysis. J.W.T.H., A.S. and L.G.S. led the
execution and analysis of the parallel Arecibo observing campaign. P.D. led
the commissioning of fast-sampled VLA observing modes. S.C. was principal
investigator of the ALMA observations. P.S. was principal investigator of the
X-ray observations, and performed the X-ray analysis, along with S.B. S.P.T. was
principal investigator of the Gemini observations, and along with C.G.B. led the
analysis of Keck, Gemini and archival UKIDSS and GLIMPSE data. S.C. and C.J.L.
led the writing of the manuscript, with substantial contributions from J.M.C. and
J.W.T.H. All authors contributed substantially to the interpretation of the analysis
results and to the final version of the manuscript.
Author Information 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
S.C. (shami.chatterjee@cornell.edu).
Reviewer Information Nature thanks H. Falcke and G. Hallinan for their
contribution to the peer review of this work.
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
