A dust-enshrouded tidal disruption event with a resolved radio jet in a galaxy merger

See allHide authors and affiliations

Science  03 Aug 2018:
Vol. 361, Issue 6401, pp. 482-485
DOI: 10.1126/science.aao4669

An expanding radio jet from a destroyed star

If a star gets too close to a supermassive black hole, it gets ripped apart in a tidal disruption event (TDE). Mattila et al. discovered a transient source in the merging galaxy pair Arp 299, which they interpret as a TDE. The optical light is hidden by dust, but the TDE generated copious infrared emission. Radio observations reveal that a relativistic jet was produced as material fell onto the black hole, with the jet expanding over several years. The results elucidate how jets form around supermassive black holes and suggest that many TDEs may be missed by optical surveys.

Science, this issue p. 482


Tidal disruption events (TDEs) are transient flares produced when a star is ripped apart by the gravitational field of a supermassive black hole (SMBH). We have observed a transient source in the western nucleus of the merging galaxy pair Arp 299 that radiated >1.5 × 1052 erg at infrared and radio wavelengths but was not luminous at optical or x-ray wavelengths. We interpret this as a TDE with much of its emission reradiated at infrared wavelengths by dust. Efficient reprocessing by dense gas and dust may explain the difference between theoretical predictions and observed luminosities of TDEs. The radio observations resolve an expanding and decelerating jet, probing the jet formation and evolution around a SMBH.

The tidal disruption of stars by supermassive black holes (SMBHs) in the nuclei of galaxies was predicted theoretically 30 years ago (1, 2). In a tidal disruption event (TDE), roughly half of the star’s mass is ejected, whereas the other half is accreted onto the SMBH, generating a bright flare that is normally detected at x-ray, ultraviolet (UV), and optical wavelengths (35). TDEs are also expected to produce radio transients, lasting from months to years and including the formation of a relativistic jet, if a fraction of the accretion power is channeled into a relativistic outflow (6). TDEs provide a means of probing central black holes in quiescent galaxies and testing scenarios of accretion onto SMBHs and jet formation (3, 6).

On 2005 January 30, we discovered a bright transient in the near-infrared (IR) (7) coincident with the western nucleus B1 (Fig. 1) of the nearby [44.8 Mpc (7)] luminous infrared galaxy (LIRG) Arp 299. In galaxy mergers like that of Arp 299, large amounts of gas fall into the central regions, triggering a starburst. The long-term radio variability (8) and the IR spectral energy distribution (SED) (9) indicate a very high core-collapse supernova (SN) rate of ~0.3 year−1 within the B1 nucleus. The B1 region also harbors a Compton-thick active galactic nucleus (AGN) that has been seen directly only in hard x-rays (10). This is consistent with an extremely high visual extinction AV of ~460 magnitudes through an almost edge-on AGN torus (11). Galaxy mergers like that of Arp 299 are expected to have TDE rates several orders of magnitude higher than in field galaxies, albeit for relatively short periods of time [~3 × 105 year (12)].

Fig. 1 The transient Arp 299-B AT1 and its host galaxy Arp 299.

(A) A color-composite optical image from the HST, with high-resolution, 12.5 by 13 arcsec size near-IR 2.2-μm images [insets (B) and (C)] showing the brightening of the B1 nucleus (7). (D) The evolution of the radio morphology as imaged with VLBI at 8.4 GHz [7 × 7 milli-arcsec (mas) region with the 8.4-GHz peak position in 2005, right ascension (RA) = 11h28m30.9875529s, declination (Dec) = 58°33′40′′.783601 (J2000.0), indicated by the dotted lines]. The VLBI images are aligned with an astrometric precision better than 50 μas. The initially unresolved radio source develops into a resolved jet structure a few years after the explosion, with the center of the radio emission moving westward with time (7). The radio beam size for each epoch is indicated in the lower-right corner.

The transient source (henceforth Arp 299-B AT1) was discovered as part of a near-IR (2.2 μm) survey for highly obscured supernovae (SNe) in starburst galaxies (13). Over the following years it became luminous at IR and radio wavelengths, but was much fainter at optical wavelengths (7), implying substantial extinction by interstellar dust in Arp 299. Our follow-up observations show that the nuclear outburst had a peak brightness comparable to that of the entire galaxy nucleus at both near-IR and radio wavelengths (Fig. 1) (7). Based on the energetics and multiwavelength behavior of Arp 299-B AT1 over a decade of observations (Figs. 1 to 3), two broad scenarios to explain its origin are plausible: (i) an event unrelated to the SMBH, such as an extremely energetic SN, or a gamma-ray burst; or (ii) accretion-induced SMBH variability, such as an AGN flare, or a TDE.

Fig. 3 Infrared properties of Arp 299-B AT1.

(A) Evolution of the observed IR spectral energy distribution (points) shown together with blackbody fits between 136 and 4207 days after the first IR detection on 2004 Dec. 21.6 (7). Over this period, the blackbody temperature decreased from ~1050 to ~750 K, while the blackbody radius increased from 0.04 to 0.13 pc. (B) The evolution of the integrated blackbody luminosity (blue circles) and cumulative radiated energy (red squares). The observed radiated energy by day 4207 was about 1.5 × 1052 erg.

High–angular resolution [100 milli-arcsec (mas)], adaptive optics–assisted, near-IR imaging observations from the Gemini-North telescope (7) show that Arp 299-B AT1 remained stationary and coincident (within 37 mas, corresponding to ~8 pc projected distance) with the near-IR K-band nucleus, as seen in pre-outburst imaging from the Hubble Space Telescope (HST) (Fig. 1). Radio observations obtained with very-long-baseline interferometry (VLBI) constrain its position with milli-arcsec angular precision (7). Pre-discovery Very Long Baseline Array (VLBA) observations showed several compact sources at 2.3 GHz within the central few parsecs of the B1 nucleus, but no counterparts at higher frequencies (14). A new compact radio source was detected on 2005 July 17 at 8.4 GHz with the VLBA (7, 14). The coincidence of the near-IR and VLBI positions, together with the appearance of the VLBI source soon after the near-IR detection and their subsequent evolution (see below) (7), point to a common origin for both.

High–angular resolution radio observations of Arp 299-B AT1 with VLBI show that the initially unresolved radio source developed a prominent extended, jet-like structure, which became evident in images taken from 2011 onwards (Fig. 1) (7). The measured average apparent expansion speed of the forward shock of the jet is (0.25 ± 0.03) c between 2005 and 2015 (7), where c is the speed of light. The radio morphology, evolution, and expansion velocity of Arp 299-B AT1 rule out a SN origin. Similarly, a gamma-ray burst is inconsistent with both the observed peak flux density and time to reach that peak at radio wavelengths (15). Therefore, the most likely explanation is that Arp 299-B AT1 is linked to an accretion event onto the SMBH. The persistent 2.3-GHz radio emission most likely corresponds to the quiescent AGN core (7).

The multifrequency radio light curves of Arp 299-B AT1 (Fig. 2) are well reproduced by a model (16) of a jet powered by accretion of part of a tidally disrupted star onto a SMBH (7). The jet initially moves at relativistic speeds ~0.995 c, but after a distance of less than ~1017 cm (corresponding to ~760 days after the burst), it has already decelerated substantially to ~0.22 c, in agreement with expectations for TDE-driven jets (6). The apparent speed of the jet indicated by the VLBI observations, together with the nondetection of the counterjet and the modeling of the radio light curves, constrains the jet viewing angle, θobs, to be within a narrow range: 25° to 35° (7). If the jet had been launched by a preexisting AGN, its viewing angle with respect to our line of sight should have been close to 90°, as the AGN torus is seen almost edge-on (11), and a counterjet should have also been detected (7). However, a radio jet associated with a TDE does not necessarily have to be perpendicular to the preexisting AGN accretion disc (17). We therefore identify the observed radio jet as being launched by a TDE. No direct imaging has previously shown an expanding jet in a TDE, and its likely presence has been inferred on the basis of unresolved radio observations only in the cases of ASASSN-14li (3, 18, 19), IGR J12580+0134 (20), and Swift J164449.3+573451 (hereafter Sw J1644+57) (21). Our VLBI observations show a resolved, expanding radio jet in a TDE, in accordance with theoretical expectations (6).

Fig. 2 Radio properties of Arp 299-B AT1.

(A) Radio luminosity evolution of Arp 299-B AT1 at 1.7 (circles), 5.0 (pentagons), and 8.4 GHz (squares) spanning more than 12.1 years of observations, along with modeled radio light curves, using hydrodynamic and radiative simulations for a TDE-launched jet (16). The day zero corresponds to 2004 Dec. 21.6. (B) Intrinsic (beaming-corrected) jet kinetic energy, EK, versus outflow speed (19) [Γβ, where Γ = (1 – β2)–1/2 is the bulk Lorentz factor of the outflow and β = v/c], from radio observations of gamma-ray bursts (GRBs), supernovae (SNe), low-luminosity active galactic nuclei (LLAGN), and TDEs (4, 16, 1921, 32). The large circles show, from right to left, the inferred loci for Arp 299-B AT1 at four different epochs in the observer’s frame: just after the jet is launched by the TDE, and ~1, ~12, and ~760 days thereafter. For the LLAGN sample, we have assumed a constant jet kinetic power over 10 years. The triangles indicate upper limits for the expansion speed of IGR J1258+01 (20) and Sw J1644+57 (21).

The intrinsic (i.e., beaming-corrected) kinetic energy of the jet required to reproduce the radio light curves (Fig. 2) is (1.8 ± 0.9) × 1051 erg (7), similar to the case of the relativistic TDE Sw J1644+57 (16). The rise of the radio emission at high frequencies in less than about 200 days, and the substantial delay in the rise of the lower-frequency radio emission (7) implies the existence of substantial external absorption, consistent with the jet being embedded in the very dense nuclear medium of the AGN, which has a constant thermal electron number density of ~4 × 104 cm−3 up to a distance 6.3 × 1017 cm from the central engine (7).

Observations from the ground and the Spitzer Space Telescope show that the IR SED of Arp 299-B AT1 and its evolution from 2005 until 2016 can be explained by a single blackbody component (Fig. 3). The blackbody radius expands from 0.04 to 0.13 pc between May 2005 and January 2012, while its temperature cools from ~1050 to ~750 K. The size, temperature, and peak luminosity (6 × 1043 erg s−1) of the IR-emitting region agree well with both theoretical predictions and observations of thermal emission from warm dust surrounding TDEs (22, 23). Therefore, the IR SED and its evolution are consistent with absorption and reradiation of the UV and optical light from Arp 299-B AT1 by local dust.

We modeled the IR SED of the pre- and post-outburst (734 days after the first IR detection) components of Arp 299-B1 using radiative transfer models for the emission from a starburst within the galaxy, and from a dusty torus, as predicted by the standard unified model for AGN, including also the effect of dust in the polar regions of the torus (Fig. 4) (24). The model luminosities of the starburst and AGN dusty torus components remain constant within the uncertainties, whereas the luminosity of the polar dust component is found to increase by a factor of ~15 after the outburst, and the corresponding polar dust temperature increases from 500 to 900 K. Therefore, the observed IR SED of Arp 299-B AT1 can be most plausibly explained by reradiation by optically thick dust clouds in the polar regions of the torus, which suffer from a relatively low foreground extinction within Arp 299-B1 (7).

Fig. 4 Model for the observed properties of Arp 299-B AT1.

Best-fit models for the spectral energy distribution of the Arp 299 nucleus B at (A) pre-outburst and (B) post-outburst (734 days after the first IR detection). The models include a starburst component (dashed line), an active galactic nucleus (AGN) dusty torus (dotted line), and a polar dust component (thick solid line) (7). The sum of these components is shown as a thin solid line. In (B), most of the model parameters were fixed, while the temperature of the polar dust varied from 500 K in the pre-outburst case to 900 K in the post-outburst case. This yields a covering factor of the polar dust of 23 to 78%, implying that the total radiated energy is ~(1.9 to 6.5) × 1052 erg. (C) Schematic diagram (not to scale) showing the geometry of the emitting and absorbing regions (7). The TDE generates prominent x-ray, ultraviolet, and optical emission. However, the direct line of sight to the central black hole is obscured by the dusty torus, which is opaque from soft x-rays to infrared wavelengths. The polar dust reradiates in the infrared a fraction of the total energy emitted by the event. The transient radio emission originates from a relativistic jet launched by the tidal disruption of a star.

Integrating the luminosity of Arp 299-B AT1 over the period 2005 to 2016 (Fig. 3) yields a total radiated energy of about 1.5 × 1052 erg. However, a large fraction of the total energy emitted by the transient can be expected to be scattered, absorbed, and reradiated at substantially longer IR wavelengths by the dusty torus. We estimate that the fraction of energy that heated the polar dust was in the range of 23 to 78% (7). Thus, the total radiated energy of Arp 299-B AT1 was (1.9 to 6.5) × 1052 erg, which requires a disruption of a star with a mass of about 1.9 to 6.5 solar masses (Embedded Image), assuming a standard accreted fraction and radiative efficiency (7). Stars in this mass range can be disrupted by the ~2 × 107 Embedded Image SMBH in Arp 299-B1 (10, 25). The kinetic energy of the jet is expected to be about 1% of the total rest mass energy (6), which agrees well with our estimated kinetic energy for the radio jet of Arp 299-B AT1 (7).

In addition to Arp 299-B AT1, the only other TDE candidates (although with debated nature) to have an observed radiated energy on the order of 1052 erg are ASASSN-15lh (26, 27) and possibly transients similar to PS1-10adi (28). The high energy of ASASSN-15lh was originally proposed to be the result of an energetic SN (26), but was later explained as a tidal disruption by a high mass (7.6 × 108Embedded Image), rapidly rotating black hole (27). In the case of PS1-10adi, the large radiated energy was proposed to arise from the interaction of either an expanding TDE, or SN ejecta, with the dense medium in the nuclear environment (28). Arp 299-B AT1 was most plausibly the result of the disruption of a star more massive than about 2Embedded Image in a very dense medium. The soft x-ray photons produced by the event were efficiently reprocessed into UV and optical photons by the dense gas, and further to IR wavelengths by dust in the nuclear environment. Efficient reprocessing of the energy might thus resolve the outstanding problem of observed luminosities of optically detected TDEs being generally lower than predicted (29).

The case of Arp 299-B AT1 suggests that recently formed massive stars are being accreted onto the SMBH in such environments, resulting in TDEs injecting large amounts of energy into their surroundings. However, events similar to Arp 299-B AT1 may remain hidden within dusty and dense environments, and would not be detectable by optical, UV, or soft x-ray observations. The recent discovery of another TDE candidate in the nucleus of the luminous infrared galaxy IRAS F01004-2237 (30) yields further support for an enhanced rate of TDEs in such galaxies, which could be missed as a result of dust extinction. Such TDEs from relatively massive, newly formed stars might provide a large radiative feedback, especially at higher redshifts where galaxy mergers and LIRGs are more common (31).

Supplementary Materials

Materials and Methods

Figs. S1 to S7

Tables S1 to S8

References (3399)

References and Notes

  1. Materials and methods are available as supplementary materials.
  2. A. Becker. HOTPANTS: High Order Transform of PSF ANd Template Subtraction. Astrophysics Source Code Library, ascl:1505.004 (2015).
Acknowledgments: We thank A. Fabian, T. Hovatta, A. Levan, K. Nilsson, and C. Ricci for useful discussions. We also thank the anonymous referees for many insightful comments that improved the manuscript. Our findings are based mainly on observations obtained with the Spitzer Space Telescope, the European VLBI Network, the Very Long Baseline Array and Very Large Array, the Nordic Optical Telescope (NOT), and the Gemini Observatory. The Spitzer Space Telescope is operated by the Jet Propulsion Laboratory, California Institute of Technology, under a contract with NASA. The European VLBI Network is a joint facility of independent European, African, Asian, and North American radio astronomy institutes. The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. The Nordic Optical Telescope is operated by the Nordic Optical Telescope Scientific Association at the Observatorio del Roque de los Muchachos, La Palma, Spain, of the Instituto de Astrofisica de Canarias. The Gemini Observatory is operated by the Association of Universities for Research in Astronomy, Inc., under a cooperative agreement with the NSF on behalf of the Gemini partnership: the National Science Foundation (United States), the National Research Council (Canada), CONICYT (Chile), Ministerio de Ciencia, Tecnología e Innovación Productiva (Argentina), and Ministério da Ciência, Tecnologia e Inovação (Brazil). Figure 1 image credit: NASA, ESA, the Hubble Heritage Team (STScI/AURA)-ESA/Hubble Collaboration and A. Evans (University of Virginia, Charlottesville/NRAO/Stony Brook University). Funding: S.M. acknowledges financial support from the Academy of Finland (project: 8120503). The research leading to these results has received funding from the European Commission Seventh Framework Programme (FP/2007-2013) under grant agreement numbers 227290, 283393 (RadioNet3) and 60725 (HELP). A.A., M.P.-T., N.R.-O., and R.H.-I. acknowledge support from the Spanish MINECO through grants AYA2012-38491-C02-02 and AYA2015-63939-C2-1-P. P.G.J. acknowledges support from European Research Council Consolidator Grant 647208. C.R.-C. acknowledges support by the Ministry of Economy, Development, and Tourism's Millennium Science Initiative through grant IC120009, awarded to The Millennium Institute of Astrophysics, MAS, Chile, and from CONICYT through FONDECYT grant 3150238 and China-CONICYT fund CAS160313. P.M. and M.Á.A. acknowledge support from the ERC research grant CAMAP-250276, and partial support from the Spanish MINECO grant AYA2015-66889-C2-1P and the local Valencia government grant PROMETEO-II-2014-069. M.F. acknowledges support from a Science Foundation Ireland–Royal Society University Research Fellowship. D.L.C. acknowledges support from grants ST/G001901/1, ST/J001368/1, ST/K001051/1, and ST/N000838/1. P.V. acknowledges support from the National Research Foundation of South Africa. J.H. acknowledges financial support from the Finnish Cultural Foundation and the Vilho, Yrjö and Kalle Väisälä Foundation. J.K. acknowledges financial support from the Academy of Finland (grant 311438). Author contributions: S.M. and M.P.-T. co-led the writing of the manuscript, the data analysis, and physical interpretation. A.E. modeled the IR SED and contributed to the physical interpretation and text. P.M. and M.Á.A. modeled the radio light curves and contributed to the physical interpretation and text. M.F. analyzed the HST data and contributed to the physical interpretation and text. E.K. contributed to the observations and analysis of the near-IR data, the physical interpretation, and text. A.A., C.R.-C., and I.M.-V. contributed to the analysis and interpretation of the radio data, and text. E.V., M.B., R.H.-I., N.R.-O., R.J.B., and K.W. contributed to the analysis and interpretation of the radio data. T.H. and S.T. analyzed the x-ray data. P.G.J. and S.J.S. contributed to the physical interpretation and text. P.L. and C.F. contributed to the physical interpretation. A.A.-H., W.P.S.M., R.K., and P.V. contributed to the analysis and physical interpretation of the infrared data. J.H., T.K., and T.R. contributed to the observations and analysis of the near-IR data. D.L.C., J.K., K.N., R.G., S.R., N.A.W., and G.Ö. contributed data. All coauthors contributed with comments to the text. Competing interests: We declare that none of the authors have any competing interests. Data and materials availability: The raw observations used in this publication are available from the Spitzer Heritage Archive at (Proposal IDs: 32, 108, 60142, 80105, 90031, 90157, 10086, 11076), from the EVN data archive at (proposal IDs EP063, EP068, EP075, EP087, GP053), the NRAO data archive at (proposal IDs: BPU027, BP202, AC0749), the NOT data archive at, the Gemini Observatory Archive at (programs: GN-2008B-Q-32, GN-2009A-Q-12, GN-2009B-Q-23, GN-2010A-Q-40, GN-2011A-Q-48 and GN-2011B-Q-73), the Hubble Legacy Archive at, the Chandra Data Archive at (OBSIDs 1641, 6227, 15077 and 15619), the XMM-Newton Science Archive at (ObsId 0679381101), the Isaac Newton Group Archive at, and the United Kingdom Infrared Telescope Archive at The radiative transfer models are part of the CYGNUS project with the model grids available at The results of the hydrodynamic and radiative simulations used for modeling the radio light curves, and the data and Python code used to produce Fig. 2A and fig. S6 are available at The Python code used for determining the allowed values for the viewing angle of the radio jet and producing fig. S7 is available at Full details of all data and software used in this paper are given in the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article