Flows of X-ray gas reveal the disruption of a star by a massive black hole
Jon M. Miller, Jelle S. Kaastra, M. Coleman Miller, Mark T. Reynolds, Gregory Brown, S. Bradley Cenko, Jeremy J. Drake, Suvi Gezari, James Guillochon, Kayhan Gultekin, Jimmy Irwin, Andrew Levan, Dipankar Maitra, W. Peter Maksym, Richard Mushotzky, Paul O'Brien, Frits Paerels, Jelle de Plaa, Enrico Ramirez-Ruiz, Tod Strohmayer, Nial Tanvir
aa r X i v : . [ a s t r o - ph . H E ] O c t Flows of X-ray gas reveal the disruptionof a star by a massive black hole
Jon M. Miller , Jelle S. Kaastra , , , M. Coleman Miller , Mark T. Reynolds ,Gregory Brown , S. Bradley Cenko , , Jeremy J. Drake , Suvi Gezari , JamesGuillochon , Kayhan Gultekin , Jimmy Irwin , Andrew Levan , DipankarMaitra , W. Peter Maksym , Richard Mushotzky , Paul O’Brien , FritsPaerels , Jelle de Plaa , Enrico Ramirez-Ruiz , Tod Strohmayer , Nial Tanvir April 2, 2018 Department of Astronomy, The University of Michigan, 1085 SouthUniversity Avenue, Ann Arbor, Michigan, 48103, USA. SRON Nether-lands Institute for Space Research, Sorbonnelaan 2, 3584 CA Utrecht,The Netherlands. Department of Physics and Astronomy, UniversiteitUtrecht, PO BOX 80000, 3508 TA Utrecht, The Netherlands. LeidenObservatory, Leiden University, PO BOX 9513, 2300 RA Leiden, TheNetherlands. Department of Astronomy, The University of Maryland,College Park, Maryland, 20742, USA. Department of Physics, Uni-versity of Warwick, Coventry, CV4 7AL, UK. Joint Space-ScienceInstitute, University of Maryland, College Park, MD, 02742, USA. Astrophysics Science Division, NASA Goddard Space Flight Cen-ter, MC 661, Greenbelt, Maryland, 20771, USA. Smithsonian Astro-physical Observatory, 60 Garden Street, Cambridge, Massachusetts,02138, USA. The Institute for Theory and Computation, Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge,Massachusetts, 02138, USA. Department of Physics and Astron-omy, University of Alabama, BOX 870324, Tuscaloosa, Alabama,35487, USA. Department of Physics and Astronomy, Wheaton Col-lege, Norton, Massachusetts, 02766, USA. Department of Physicsand Astronomy, University of Leicester, University Road, Leicester,LE1 7RH, UK. Columbia Astrophysics Laboratory and Departmentof Astronomy, Columbia University, 550 West 120th Street, New York,New York, 10027, USA. Department of Astronomy and Astrophysics,University of California, Santa Cruz, California, 95064, USA.
Tidal forces close to massive black holes can violently dis-rupt stars that make a close approach. These extreme eventsare discovered via bright X-ray and optical/UV flaresin galactic centers. Prior studies based on modeling decay-ing flux trends have been able to estimate broad properties,such as the mass accretion rate . Here we report the detec-tion of flows of highly ionized X-ray gas in high-resolutionX-ray spectra of a nearby tidal disruption event. Variabil-ity within the absorption-dominated spectra indicates thatthe gas is relatively close to the black hole. Narrow linewidths indicate that the gas does not stretch over a largerange of radii, giving a low volume filling factor. Modest out-flow speeds of few × km s − are observed, significantlybelow the escape speed from the radius set by variability. The gas flow is consistent with a rotating wind from the in-ner, super-Eddington region of a nascent accretion disk, orwith a filament of disrupted stellar gas near to the apocen-ter of an elliptical orbit. Flows of this sort are predicted byfundamental analytical theory and more recent numericalsimulations . ASASSN-14li was discovered in images obtained onNovember 22, 2014 (MJD 56983), at a visual magnitude ofV=16.5 by the All-Sky Automated Survey for Supernovae(ASAS-SN). Follow-up observations found the transient sourceto coincide with the center of the galaxy PGC 043234 (originallyZw VIII 211), to within 0.04 arc seconds . This galaxy lies ata red-shift of z = 0 . , or a luminosity distance of 90.3 Mpc(for H = 73 km s − , Ω matter = 0 . , Ω Λ = 0 . ), makingASASSN-14li the closest disruption event discovered in over 10years. The discovery magnitudes indicated a substantial flux in-crease over prior, archival optical images of this galaxy. Follow-up observations with the Swift
X-ray Telescope (XRT) es-tablished a new X-ray source at this location .Archival X-ray studies rule out the possibility that PGC043234 harbours a standard active galactic nucleus that couldproduce bright flaring. PGC 043234 is not detected in theROSAT All-Sky Survey . Utilizing the online interface to thedata, the background count rate for sources detected in the vicin-ity is .
002 counts s − arcmin − . With standard assump-tions (see Methods ), this rate corresponds to a luminosity of L ≃ . × erg s − , which is orders of magnitude below astandard active nucleus.Theory predicts that early tidal disruption event (TDE) evo-lution should be dominated by a bright, super-Eddington accre-tion phase, and be followed by a charactecteristic t − / declineas disrupted material interacts and accretes . Detections ofwinds integral to super-Eddington accretion have not been re-ported previously, but t − / flux decay trends in the UV (wheredisk emission from active nuclei typically peaks) are now a stan-dard signature of TDEs in the literature . Figure 1 shows the1ux decay of ASASSN-14li, as observed by Swift . A fit to theUVM2 data assuming an index of α = − / gives a disruptiondate of t ≃ ± (MJD). The V-band light is consistentwith a shallower t − / decay; this can indicate direct thermalemission from the disk, or reprocessed emission (see Meth-ods ).We triggered approved
XMM-Newton programs to studyASASSN-14li soon after its discovery. Although
XMM-Newton carries several instruments, the spectra from the two RGS unitsare the focus of this analysis. We were also granted a Direc-tor’s Discretionary Time observation with
Chandra , using itsLow Energy Transmission Grating spectrometer (LETG), pairedwith its High Resolution Camera for spectroscopy (HRC-S).The 18–35 ˚A X-ray spectra of ASASSN-14li are clearlythermal in origin, so we modeled the continuum with a singleblackbody, modified by interstellar absorption in PGC 043234and the Milky Way, and absorption from blue-shifted, ionizedgas local to the TDE. The self-consisent photoionization code“pion” was used to model the complex absorption spectra (seeTable 1, and Methods ).Assuming that the highest bolometric luminosity derived infits to the high-resolution spectra ( L = 3 . ± . × erg s − )corresponds to the Eddington limit, a black hole mass of . × M ⊙ is inferred. The blackbody emission measured in fitsto the time-averaged XMM-Newton spectrum gives an emittingarea of . × cm ; implying r = 1 . × cm for aspherical geometry. This is consistent with the innermost stablecircular orbit (ISCO) around an M ≃ . × M ⊙ blackhole. Modeling of the Swift light curves (see Figure 1) usinga self-consistent treatment of direct and reprocessed light froman elliptical accretion disk gives a mass in the range of M ≃ . − . × M ⊙ (please see the Methods ). In concert, thethermal spectrum, implied radii, and the run of emission fromX-rays to optical bands unambiguously signal the presence ofan accretion disk in ASASSN-14li.Figure 2 shows the best-fit model for the spectra obtainedin the long stare with the
XMM-Newton /RGS (see Table 1, and
Methods ). An F-test finds that photoionized X-ray absorptionis required in fits to these spectra at more than the 27 σ level ofconfidence, relative to a spectral model with no such absorption.The model captures the majority of the strong absorption lines,giving χ = 870 . for 563 degrees of freedom (see Table 1).The strongest lines in the spectrum coincide with ionized chargestates of N, O, S, Ar, and Ca. Only solar abundances are requiredto describe the spectra. The Chandra spectrum independentlyconfirms these results in broad terms, and requires absorption atmore than the 6 σ level of confidence.A hard lower limit on the radius of the absorbing gas is setby the the blackbody continuum. The best radius estimate likelycomes from variability time scales within the XMM-Newton longstare. Analysis of specific time segments within the long stare,as well as flux-selected segments, reveals that the absorptionvaries (see Table 1, and
Methods ). This sets a relevant limitof r ≤ cδt , or r ≤ × cm . While the column den-sity and ionization do not vary significantly, the blue-shift ofthe gas does. During the initial third of the observation, the blue-shift is larger, v shift = − ±
50 km s − , but falls to v shift = − − km s − in the final two-thirds. Shorter mon-itoring observations with XMM-Newton reveal evolution of theabsorbing gas, including changes in ionization and column den-sity, before and after the long stare (see Table 1, and
Methods ).Fundamental theoretical treatments of TDEs predict an ini-tial near-Eddington or super-Eddington phase ; this is confirmedin more recent theoretical studies . The high-resolution X-ray spectra were obtained within the predicted time frame forsuper-Eddington accretion, for our estimates of the black holemass . Although the ionization parameter of the observed gas ishigh, the ionizing photon distribution peaks at a low energy, andthe wind could be driven by radiation force. Such flows are natu-rally clumpy, and may be similar to the photospheres of novae .Given the strong evidence of an accretion disk in our observa-tions of ASASSN-14li, the X-ray outflow is best associated witha wind from the inner regions of a nascent, super-Eddington ac-cretion disk. The local escape speed at an absorption radius of r ≃ GM / c (appropriate for M ≃ M ⊙ ) exceeds theobserved outflow line-of-signt speed of the gas, but Keplerianrotation is not encoded in absorption, and projection effects arealso important. The small width of the absorption lines relativeto the escape velocity may also indicate a low filling factor, con-sistent with a clumpy outflow or shell.The existing observations show a general trend towardhigher outflow speeds with time. Corresponding changes in ion-ization and column density are more modest, and not clearlylinked to outflow speed . However, some recent work has pre-dicted higher outflow speeds in an initial super-Eddington diskregime, and lower outflow speeds in a subsequent thin diskregime . An observation in an earlier, more highly super-Eddington phase might have observed broader lines and higheroutflow speeds; future observations of new TDEs can test this.Figure 3 shows the time evolution of the blackbodytemperature measured in Swift /XRT monitoring observations.The temperature is remarkably constant, especially in con-trast to the optical/UV decline shown in Figure 1. Observa-tions of steady blackbody temperatures despite decaying multi-wavelength light curves in some TDEs has recently been ex-plained through winds . Evidence of winds in our data supportsthis picture.The low gas velocities may also be consistent with disruptedstellar gas on an elliptical orbit in a nascent disk, near apocen-ter. This picture naturally gives a low filling factor, resulting ina small total mass in absorbing gas (see Methods ). Recent nu-merical simulations predict that a fraction of the disrupted ma-terial in a TDE will circularize slowly , and that flows will befilamentary , while stellar gas that is more tightly bound canform an inner, Eddington-limited or super-Eddington disk morequickly.The highly ionized, blue-shifted gas discovered in our high-resolution X-ray spectra of ASASSN-14li confirms both funda-mental and very recent theoretical predictions for the structureand evolution of tidal disruption events. The field can now pro-ceed to pair high-resolution X-ray spectroscopy with an ever-increasing number of TDE detections to test models of accretion2isk formation and evolution, and to explore strong-field gravi-tation around massive black holes .
1. Bade, N., Komossa, S. & Dahlem, M. Detection of an extremely softX-ray outburst in the HII-like nucleus of NGC 5905.
Astron. Astrophys. , L35–L38 (1996).2. Komossa, S. & Greiner, J. Discovery of a giant and luminous X-rayoutburst from the optically inactive galaxy pair RX J1242.6-1119.
As-tron. Astrophys. , L45–L48 (1999).3. Esquej, P. et al.
Candidate tidal disruption events from the XMM-Newton slew survey.
Astron. Astrophys. , L49–L52 (2007).4. Cappelluti, N. et al.
A candidate tidal disruption event in the Galaxycluster Abell 3571.
Astron. Astrophys. , L9–L12 (2009).5. Gezari, S. et al.
UV/Optical Detections of Candidate Tidal DisruptionEvents by GALEX and CFHTLS.
Astrophys. J. , 944–969 (2008).6. Gezari, S. et al.
An ultraviolet-optical flare from the tidal disruption ofa helium-rich stellar core.
Nature , 217–220 (2012).7. Guillochon, J., Manukian, H. & Ramirez-Ruiz, E. PS1-10jh: The Dis-ruption of a Main-sequence Star of Near-solar Composition.
Astro-phys. J. , 23 (2014).8. Rees, M. J. Tidal disruption of stars by black holes of 10 to the 6th-10to the 8th solar masses in nearby galaxies.
Nature , 523–528 (1988).9. Strubbe, L. E. & Quataert, E. Optical flares from the tidal disruption ofstars by massive black holes.
Mon. Not. R. Astron. Soc. , 2070–2084(2009).10. Lodato, G., King, A. R. & Pringle, J. E. Stellar disruption by a su-permassive black hole: is the light curve really proportional to t − / ? Mon. Not. R. Astron. Soc. , 332–340 (2009).11. Lodato, G. & Rossi, E. M. Multiband light curves of tidal disruptionevents.
Mon. Not. R. Astron. Soc. , 359–367 (2011).12. Strubbe, L. E. & Quataert, E. Spectroscopic signatures of the tidal dis-ruption of stars by massive black holes.
Mon. Not. R. Astron. Soc. ,168–180 (2011).13. Shiokawa, H., Krolik, J. H., Cheng, R. M., Piran, T. & Noble, S. C.General Relativistic Hydrodynamic Simulation of Accretion Flow froma Stellar Tidal Disruption. arxiv:1501.04365 (2015).14. Miller, M. C. Disk Winds as an Explanation for Slowly Evolving Tem-peratures in Tidal Disruption Events. arxiv:1502.03284 (2015).15. Jose, J. et al.
ASAS-SN Discovery of an Unusual Nuclear Transient inPGC 043234.
The Astronomer’s Telegram , 1 (2014).16. Gehrels, N. et al.
The Swift Gamma-Ray Burst Mission.
Astrophys. J. , 1005–1020 (2004).17. Burrows, D. N. et al.
The Swift X-Ray Telescope.
Space Science Re-views , 165–195 (2005).18. Voges, W. et al.
The ROSAT all-sky survey bright source catalogue.
Astron. Astrophys. , 389–405 (1999).19. Phinney, E. S. Manifestations of a Massive Black Hole in the GalacticCenter. In Morris, M. (ed.)
The Center of the Galaxy , vol. 136 of
IAUSymposium , 543 (1989).20. Kaastra, J. S., Mewe, R. & Nieuwenhuijzen, H. SPEX: a new code forspectral analysis of X and UV spectra. In Yamashita, K. & Watanabe,T. (eds.)
UV and X-ray Spectroscopy of Astrophysical and LaboratoryPlasmas , 411–414 (1996).21. Loeb, A. & Ulmer, A. Optical Appearance of the Debris of a Star Dis-rupted by a Massive Black Hole.
Astrophys. J. , 573–578 (1997).22. Piran, T., Svirski, G., Krolik, J., Cheng, R. M. & Shiokawa, H. Disk For-mation Versus Disk Accretion - What Powers Tidal Disruption Events?
Astrophys. J. , 164 (2015).23. Shaviv, N. J. The theory of steady-state super-Eddington winds and itsapplication to novae.
Mon. Not. R. Astron. Soc. , 126–146 (2001).24. Ram´ırez, J. M. Kinematics from spectral lines for AGN outflows basedon time-independent radiation-driven wind theory.
Rev. Mexicana As-tron. Astrofis. , 385–399 (2011).25. Holoien, T. W.-S. et al. ASASSN-14ae: a tidal disruption event at 200Mpc.
Mon. Not. R. Astron. Soc. , 3263–3277 (2014). 26. Guillochon, J. & Ramirez-Ruiz, E. A Dark Year for Tidal DisruptionEvents. arxiv:1501.05306 (2015).27. Stone, N. & Loeb, A. Observing Lense-Thirring Precession in TidalDisruption Flares.
Physical Review Letters , 061302 (2012).28. Kaastra, J. S., Mewe, R. & Raassen, T. New Results on X-ray Modelsand Atomic Data.
Highlights of Astronomy , 648 (2005). Acknowledgments
We thank
Chandra
Director Belinda Wilkesand the
Chandra team for accepting our request for Director’s Discre-tionary Time,
XMM-Newton
Director Norbert Schartel and the
XMM-Newton team for executing our approved target-of-opportunity pro-gram, and
Swift
Director Neil Gehrels and the
Swift team for moni-toring this important source. J.M.M. is supported by NASA funding,through
Chandra and
XMM-Newton guest observer programs. SRONis supported by NWO, the Netherlands Organization for ScientificResearch. JJD was supported by NASA Contract NAS8-03060 to theChandra X-ray Center. W.P.M. is grateful for support by the Univer-sity of Alabama Research Stimulation Program.
Author Contributions
J.M.M. led the
Chandra and
XMM-Newton data reduction and analysis, with contributions from J.S.K., J.J.D.,and J.P. M.R. led the
Swift data reduction and analysis, with help fromB.C., S.G, and R.M. M.C.M., E.R.-R., and J.G. provided theoreticalinsights. G.B., K.G., J.I., A.L., D.M., W.P.M., P.O., D.P., F.P., T.S.,and N.T. contributed to the discussion and interpretation.
Competing Interests
The authors declare that they have no com-peting financial interests.
Correspondence
Correspondence and requests for materialsshould be addressed to J.M.M. ([email protected]). able 1 | Modeling of the high-resolution X-ray spectra reveals ionized flows of gas.
Each spectrum was fit with a simple blackbodycontinuum, modified by photoionized absorption via the “pion” model, and interstellar absorption in the host galaxy PGC 043234 and theMilky Way. The fits were made using “SPEX” , minimizing a χ statistic. In all cases, σ errors are quoted. Where a parameter is quotedwith an asterisk, the listed parameter was not varied. X-ray fluxes and luminosities listed with the suffix “b” for broad were extrapolatedfrom the fitting band to the 1.24–124 ˚A band; those with the suffix “f” represent values for the 18–35 ˚A fitting band. Interstellar columndensities are separately measured for the Milky Way ( N H , MW ) at zero red-shift, and the host galaxy PGC 043234 ( N H , HG ) at red-shift of z = 0 . . These parameters were measured in the XMM-Newton long stare and then fixed in fits to other spectra. Variable parametersin the photoionization model are listed together; the negative v shift values indicate a blue shift relative to the host galaxy. Mission
XMM-Newton XMM-Newton XMM-Newton XMM-Newton Chandra XMM-Newton
ObsId 0694651201 0722480201 0722480201 0722480201 17566, 17567 0694651401comment monitoring long stare stare (low) stare (high) – monitoringStart (MJD) 56997.98 56999.54 56999.94 57000.0 56999.97, 57002.98 57023.52Duration (ks) 22 94 36 58 35, 45 23.6 F X , b ( − erg cm − s − ) . ± . . ± . . ± . . ± . . +0 . − . . ± . X , b ( erg s − ) . ± . . ± . . ± . . ± . . +0 . − . . ± . X , f ( − erg cm − s − ) . ± . . ± . . ± .
08 1 . ± .
08 1 . +0 . − . . ± . X , f ( erg s − ) . ± .
06 0 . ± .
03 0 . ± .
01 0 . ± .
01 0 . +0 . − . . ± . N H , MW (10 cm − ) . ± . N H , HG (10 cm − ) . ± . N H , TDE (10 cm − ) 0 . ± . . +0 . − . . +0 . − . . − . . +0 . − . . ± . log( ξ ) (erg cm s − ) . ± . . ± . . ± . . +0 . − . . +0 . − . . ± . v rms (km s − ) ±
30 110 +30 − +60 − ±
20 120 +40 − +60 − v shift (km s − ) − ± − ± − ± − − − +60 − − ± kT (eV) . ± .
09 51 . ± . . ± . . ± . . ± . . ± . Norm ( cm ) . ± . . ± . . ± . . ± . . +0 . − . . ± . χ /ν igure 1 | The multi-wavelength light curves of ASASSN-14li clealy signal a tidal disruption event.
The light curves are based onmonitoring observations with the
Swift satellite. The errors shown are the σ confidence limits on the flux in each band. Contributionsfrom the host galaxy have been subtracted (see Methods ). The UVM2 filter samples the UV light especially well. The gray band depictsthe t − / flux decay predicted by fundamental theory . The X-ray flux points carry relatively large errors; a representative error isshown. Fits to the decay curve are described in the main text and in the M ethods. igure 2 | The high-resolution X-ray spectra of ASASSN-14li reveal blue-shifted absorption lines.
Spectra from the long stare with
XMM-Newton and the combined
Chandra spectrum are shown.
XMM-Newton spectra from the RGS1 and RGS2 units are shown in blackand blue, respectively; the RGS2 unit is missing a detector in the 20–24 ˚A band. The best-fit photoionized absorption model for theoutflowing gas detected in each spectrum is shown in red (see
Methods ), and selected strong lines are indicated. Below each spectrum,the goodness-of-fit statistic ( ∆ χ ) is shown before (cyan) and after (black) modeling the absorbing gas. . . . O b s e r v e d T e m p [ k e V ] TIME [ MJD − 56980 ]
Figure 3 | The temperatue of blackbody continuum emission from ASASSN-14li is steady over time.
The temperature measuredin simple blackbody fits to
Swift /XRT monitoring observations is plotted versus time. Errors are 1 σ confidence intervals. The temperatureis remarkably steady, contrasting strongly with the declining fluxes shown in Figure 1. Recent theory suggests that winds may serve tomaintain steady temperatures in some TDEs . ethods Estimates of prior black hole luminosity
Utilizing the ROSAT All-Sky Survey , the region aroundthe host galaxy, PGC 043234, was searched for point sources.No sources were found. Points in the vicinity of the host galaxywere examined to derive a background count rate of 0.002counts s − . Assuming the Milky Way column density along thisline of sight, and taking a typical Seyfert X-ray spectral index of Γ = 1 . , this count rate translates into L ≃ . × erg s − .This limit is orders of magnitude below a Seyfert or quasarluminosity. Optical/UV monitoring observations and data reduction
Swift monitors transient and variable sources via co-aligned X-ray (XRT: 0.3 - 10 keV) and UV-Optical (UVOT: 170-650 nm) telescopes. High-cadence monitoring of ASASSN-14liwith UVOT has continued in six bands: V, B, U, UVW1, UVM2,and UVW2 ( λ c ∼ , , , , ,
190 nm ).All observations were processed using the latest HEASOFTsuite and calibrations. Individual optical/UV exposures wereastrometrically corrected and sub-exposures in each filter weresummed. Source fluxes were then extracted from an aperture of3 ′′ radius, and background fluxes were extracted from a source-free region to the east of ASASSN-14li due to the presence ofa (blue) star lying 10 arc seconds to the South, using UVOT - MAGHIST .To estimate the host contamination, we have measured thehost flux in 3 ′′ aperture (matched to the aperture used for theUVOT photometry) in pre-outburst Sloan Digital Sky Survey(SDSS; ), 2 Micron All-Sky Survey (2MASS; ), and GALEX images. Extra caution was used to deblend the GALEX data,where large PSF resulted in contamination from the star ∼ ′′ tothe South. We estimated the uncertainty in each host flux byvarying the inclusion aperture from 2 ′′ to 4 ′′ .We then fit the host photometry to synthetic galaxy tem-plates using the Fitting and Assessment of Synthetic Templates(FAST; ) code. We employed stellar templates from the catalog, and allowed the star formation history, extinction law,and initial mass function to vary over the full range of pa-rameters allowed by the software. All best fit models hadstellar masses ≈ . M ⊙ , low ongoing star formation rates(SFR . − . M ⊙ yr − ), and modest line-of-sight extinction( A V . . mag).We took the resulting galaxy template spectra and integratedthese over each UVOT filter bandpass to estimate the host countrate. For the uncertainty in this value, we adopt either the root-mean-square spread of the resulting galaxy template models, or10% of the inferred count rate, whichever value was larger. Wethen subtracted these values from our measured (coincidence-loss corrected) photometry of the host plus transient, to isolatethe component due to TDE. For reference, our inferred countrates for each UVOT filter are: V = 5 . ± . s − , B =9 . ± . s − , U = 4 . ± . s − , UV W . ± . s − , UV M . ± . s − , and UV W . ± . s − .Figure 1 shows the host-subtracted optical and UV lightcurves ASASSN-14li. Fits to the UVOT/UVM2 light curve
The UVM2 filter provides the most robust trace of the massaccretion rate in a TDE like ASASSN-14li; it has negligibletransmission at optical wavelengths . Fits to the UVM2 lightcurve with a power-law of the form f ( t ) = f × ( t + t ) − α with a fixed index of α = − / imply a disruption date of t = 56980 ± (MJD). This model achieves a fair characteri-zation of the data; high fluxes between days 80-100 (in the unitsof Figure 1) result in a poor statistical fit ( χ /ν = 1 . , where ν = 54 degrees of freedom). If the light curve is fit with avariable index, a value of − . ± . is measured (90% confi-dence). This model achieves an improved fit ( χ /ν = 1 . , for ν = 53 degrees of freedom), but it does not tightly constrainthe disruption date, placing t in the MJD 56855–56920 range.That disruption window is adjacent to an interval wherein theASAS-SN monitoring did not detect the source , making it lessplausible than the fit with α = − / .The optical bands appear to have a shallower decay curvethan the UV bands. Recent theory predicts that optical lightproduced via thermal disk emission should show a decayconsistent with t − / ; this might also be due to reprocessing .The V-band data are consistent with this prediction, thoughthe data are of modest quality and a broad range of decays arepermitted. X-ray monitoring observations and data reduction
The
Swift /XRT is a charge-coupled device (CCD). In suchcameras, photon pile-up occurs when two or more photons landwithin a single detection box during a single frame time. Thiscauses flux distortions and spectral distortions to bright sources.Such distortions are effectively avoided by extracting eventsfrom an annular region, rather than from a circle at the centerof the telescope PSF. We therefore extracted source spectra fromannuli with an inner radius of 12 arc seconds (5 pixels), and anouter radius of 50 arc seconds. Background flux was measuredin annular region extending from 140 – 210 arc seconds.Standard redistribution matrices were used; an ancillary re-sponse file was created with the xrtmkarf tool utilizing a vi-gnetting corrected exposure map. The source spectra were re-binned to have 20 counts per bin with grppha . In all spectralfits, we adopted a lower spectral bound of 0.3 keV (36 ˚ A ). Theupper bound on spectral fits varied depending on the boundaryof the last bin with at least 20 counts; this was generally around1 keV (12 ˚ A ).The XRT spectra were fit with a model consisting of ab-sorption in the Milky Way of a blackbody emitted at the red-shift of the TDE, i.e., pha(zashift(bbodyrad)) , where N H ≡ × cm − and z ≡ . . The evolution of thebest-fit temperature of this blackbody component is displayed inFigure 3.The blackbody temperature values measured from the Swift /XRT are slightly higher (kT ≃ XMM-Newton and
Chandra . If an outflow com-ponent with fiducial parameters is included in the spectral modelanyway, the XRT temperatures are then in complete agreement8ith those measured using
XMM-Newton and
Chandra . Estimates of the black hole mass
Luminosity values inferred for the band over which the high-resolution spectra are actually fit, and a broader band are listedin Table 1. Taking the broader values as a proxy for a truebolometric fit, the highest implied soft X-ray luminosity is mea-sured in the last X MM-Newton monitoring observation, giving L ≃ . × erg s − . The Eddington luminosity for standardhydrogen-rich accretion is L Edd = 1 . × erg s − ( M/M ⊙ ) .This implies a black hole mass of M ≃ . × M ⊙ .Blackbody continua imply size scales, and - assuming thatoptically thick blackbody emission can only originate at radiilarger than the innermost stable circular orbit (ISCO) - thereforemasses. For a non-spinning Schwarzschild black hole, r ISCO =6 GM / c . The blackbody emission measured in fits to the time-averaged XMM-Newton “long stare” gives an emitting area of . × cm ; implying r = 1 . × cm for a sphericalgeometry. The actual geometry may be more disk-like, but theinner flow may be a thick disk that is better represented by aspherical geometry. If the black hole powering ASASSN-14liis not spinning, this size implies a black hole mass of M ≃ . × M ⊙ .We also estimated the mass of the black hole at the heartof ASASSN-14li by fitting the host-subtracted light curves (seeFigure 1) using the Monte Carlo software TDEFit . This soft-ware assumes that emission is produced within an elliptical ac-cretion disk where the mass accretion rate follows the fallbackrate onto the black hole with a viscous delay . This emissionis then partly reprocessed into the UV/optical by an opticallythick layer . Super-Eddington accretion is treated by presum-ing a fitted fraction of the Eddington excess is converted intolight that is reprocessed by the same optically thick layer. Thisexcess can be produced either with an unbound wind , orwith the energy deposited by shocks in the circularization pro-cess .The software performs a maximum-likelihood analysis todetermine the combinations of parameters that reproduce theobserved light curves. We utilize the ASASSN, UVOT, andXRT data in our light-curve fitting; the most-likely models pro-duce good fits to all bands simultaneously. Within the contextof this TDE model, a black hole mass of 0.4–1.2 × M ⊙ (1 σ )is dervived. Spectroscopic observations, data reduction, and analysis
Table 1 lists the observation identification number (ObsId),start time, and duration of all of the
XMM-Newton and
Chandra observations considered in our work.The
XMM-Newton data were reduced using the standard Sci-ence Analysis System (SAS version 13.5.0) tools and the latestcalibration files. The “rgsproc” routine was used to generatespectral files from the source, background spectral files, and in-strument response files. The spectra from the RGS-1 and RGS-2units were fit jointly. Prior to fitting models, all
XMM-Newton spectra were binned by a factor of five for clarity and sensitivity.The
Chandra data were reduced using the standard Chandra Interactive Analysis of Observations (CIAO version 4.7) suite,and the latest associated calibration files. Instrument responsefiles were constructed using the “fullgarf” and “mkgrmf” rou-tines. The first-order spectra from each observation were com-bined using the tool “add grating orders”, and spectra from eachobservation were then added using “add grating spectra”.The spectra were analyzed using the “SPEX” suite version2.06 . The fitting procedure minimized a χ statistic. The spec-tra are most sensitive in the 18–35 ˚A band, and all fits wererestricted to this range. Within SPEX, absorption from the inter-stellar medium in the Milky Way was modeled using the model“hot”; a separate “hot” component was included to allow forISM absorption within PGC 043234 at its known red-shift (usingthe “reds” component in SPEX). The photoionized outflow wasmodeled using the “pion” component within the SPEX suite.Pion includes numerous lines from intermediate chargestates that are lacking in similar astrophysics packages. Thefits explored in this analysis varied the gas column density( N H , TDE ), the gas ionization parameter ( ξ , where ξ = L/nr ,and L is luminosity, n is the hydrogen number density, and r is the distance between the ionizing source and absorbing gas),the rms velocity of the gas ( v rms ), and the bulk shift of the gasrelative to the source, in the source frame ( v shift ).Spectra from segments within the “long stare” made with XMM-Newton were made by using the SAS tool “tabgtigen” tocreate “good time interval” files to isolate periods within thelight curves of the RGS data.The
Chandra /LETG spectra were dispersed onto the HRC,which has a relatively high instrumental background. Fitting thespectra only in the 18–35 ˚A band served to limit the contribu-tions of background. Nevertheless, the
Chandra spectra are lesssensitive than the best
XMM-Newton spectra of ASASSN-14li(see Figure 2). Prior to fitting, spectra from the two exposureswere added and then binned by a factor of three.Figure 2 includes plots of the ∆ χ goodness-of-fit statisticas a function of wavelength, before and after including pion tomodel the ionized absortpion. There is weak evidence of emis-sion lines in the spectra, perhaps with a P-Cygni profile (see be-low). The best-fit models for the high-resolution spectra predictone absorption line at 34.5 ˚A (H-like C VI) that is not observed;small variations to abundances could resolve this disparity.Blue-shifts as small as
200 km s − are measured in the XMM-Newton /RGS using the pion model. According to the
XMM-Newton
User’s Handbook, available through the mis-sion website, the absolute accuracy of the first-order wave-length scale is 6 m ˚A. At 18 ˚A, this corresponds to a veloc-ity of
100 km s − ; at 35 ˚A, this corresponds to a velocity of
51 km s − . The model predicts numerous lines across the 18–35 ˚A band that are clearly detected; especially with this lever-age, the small shifts we have measured with XMM-Newton arerobust. In particular, the difference in blue-shift between the lowand high flux phases of the long stare, − ±
50 km s − ver-sus − − km s − , is greater than absolute calibration uncer-tainties. Differences observed in the outflow velocities between XMM-Newton observations are as large, or larger, and also ro-bust.9he lower sensitivity of the
Chandra spectra is evident inthe relatively poor constraints achieved on the column densityof the ionized X-ray outflow (see Table 1). Similarly, the rel-atively high outflow velocity measured in the
Chandra spectra,should be viewed with a degree of caution. The outflow velocitychanges from ≃
500 km s − to just − ±
130 km s − , forinstance, when the binning factor is increased from three to five.We have found no reports in the literature of a systematic wave-length offset between contemporaneous high-resolution spectraobtained with XMM-Newton and
Chandra .The small number of high-resolution spectra complicatesefforts to discern trends. The velocity width of the absorbinggas is fairly constant over time, but there is a general trendtoward higher blue-shifts. There is no clear trend in columndensity or ionization parameter with time.
Diffuse gas mass, outflow rates, and filling factors
There is no a priori constraint on the density of the absorbinggas. Taking the maximum radius implied by variability within
XMM-Newton long stare, r ≤ × cm , and manipulatingthe ionization parameter equation ( ξ = Ln − r − , where L isthe luminosity, n is the number density, and r is the absorbingradius), we can derive an estimate of the density: n ≃ × cm − . Even assuming a uniformly-filled sphere out to aradius of r = 3 × cm , a total mass of M ≃ × g isimplied, or approximately . M ⊙ .The true gas mass within r is likely to be orders of mag-nitude lower, owing to clumping and a very low volume fill-ing factor. Using the measured value of N H , TDE and assum-ing n ≃ × cm − , N H , TDE = n ∆ r gives a value of ∆ r ≃ . × cm . The filling factor can be estimated via ∆ r/r ≃ . . The total mass enclosed out to a distance r isthen reduced accordingly, down to × − M ⊙ , assuming auniform density within r . This is a small value, plausible eitherfor a clumpy wind or gas within a filament executing an ellipticalorbit.Formally, the mass outflow rate in ASASSN-14li can beadapted from the case where the density is known, and writtenas: ˙ M out = µm p Ω LvC v ξ − ,where µ is the mean atomic weight ( µ = 1 . is typical), m P is the mass of the proton, Ω is the covering factor( ≤ Ω ≤ π ), L is the ionizing luminosity, v is the outflowvelocity, C v is the line-of-sight global filling factor, and ξ isthe ionization parameter. Using the values obtained in fitsto the XMM-Newton long stare (see Table 1), for instance, ˙ M out ≃ . × Ω C v g s − . Taking the value of C v derivedabove, an outflow rate of ˙ M out ≃ . × Ω g s − results.The kinetic power in the outflow is given by L kin = 0 . M v ;using the same values assumed to estimate the mass outflowrate, L kin ≃ . × erg s − . Emission from the diffuse outflow
We synthesized a plausible wind emission spectrum by cou-pling the pion and “hyd” models within SPEX. They hyd codeenables spectra to be constructed based on the output of hy- drodynamical simulations. As inputs, it requires the electrontemperature and ion concentrations for a gas; these were takenfrom our fits with pion. We included the resulting emissioncomponent in experimental fits to the
XMM-Newton long stare.The best-fit model gives an emission measure of . ± . × cm − , a red-shift (relative to the host) of +350 − km s − ,and an ionization parameter of log ( ξ ) = 4 . ± . .Via an F-test, the emission component is only required atthe σ level; however, it has some compelling properties. Com-bined with the blue-shifted absorption spectrum, the red-shiftedemission gives P-Cygni profiles. For the gas density of n ≃ × cm − derived previously, the emission measure gives aradius of ≃ cm , comparable to the size scale inferred fromabsorption variability.The strongest lines predicted by the emission model includeHe-like O VII, and H-like charge states of C, N, and O. Thismodel does not account for other emission line-like features inthe spectra, which are more likely to be artifacts from spectralbinning, or calibration or modeling errors. Emission featuresin the O K-edge region may be real, but caution is warranted.Other features are more easily discounted as they differ betweenthe RGS-1 and RGS-2 spectra. Code Availability
All of the data reduction and spectroscopic fitting routinesand packages used in this work are publicly available.The light curve modeling package,
TDEFit , is propri-etary at this time owing to ongoing code development; a publicrelease is planned within the coming year.
29. Voges, W. et al.
The ROSAT all-sky survey bright source catalogue.
Astron. Astrophys. , 389–405 (1999).30. Gehrels, N. et al.
The Swift Gamma-Ray Burst Mission.
Astrophys. J. , 1005–1020 (2004).31. Ahn, C. P. et al.
The Tenth Data Release of the Sloan Digital Sky Survey:First Spectroscopic Data from the SDSS-III Apache Point ObservatoryGalactic Evolution Experiment.
Astrophys. J. Suppl. Ser. , 17 (2014).32. Skrutskie, M. F. et al.
The Two Micron All Sky Survey (2MASS).
As-tron. J. , 1163–1183 (2006).33. Martin, D. C. et al.
The Galaxy Evolution Explorer: A Space UltravioletSurvey Mission.
Astrophys. J. , L1–L6 (2005).34. Kriek, M. et al.
An Ultra-Deep Near-Infrared Spectrum of a CompactQuiescent Galaxy at z = 2.2.
Astrophys. J. , 221–231 (2009).35. Bruzual, G. & Charlot, S. Stellar population synthesis at the resolutionof 2003.
Mon. Not. R. Astron. Soc. , 1000–1028 (2003).36. Poole, T. S. et al.
Photometric calibration of the Swift ultraviolet/opticaltelescope.
Mon. Not. R. Astron. Soc. , 627–645 (2008).37. Breeveld, A. A. et al.
An Updated Ultraviolet Calibration for theSwift/UVOT. In McEnery, J. E., Racusin, J. L. & Gehrels, N. (eds.)
American Institute of Physics Conference Series , vol. 1358 of
AmericanInstitute of Physics Conference Series , 373–376 (2011). .38. Jose, J. et al.
ASAS-SN Discovery of an Unusual Nuclear Transient inPGC 043234.
The Astronomer’s Telegram , 1 (2014).39. Lodato, G. & Rossi, E. M. Multiband light curves of tidal disruptionevents.
Mon. Not. R. Astron. Soc. , 359–367 (2011).40. Guillochon, J., Manukian, H. & Ramirez-Ruiz, E. Ps1-10jh: The dis-ruption of a main-sequence star of near-solar composition.
Astrophys. J. , 23 (2014).41. Burrows, D. N. et al.
The Swift X-Ray Telescope.
Space Science Re-views , 165–195 (2005).
2. Guillochon, J. & Ramirez-Ruiz, E. Hydrodynamical Simulations to De-termine the Feeding Rate of Black Holes by the Tidal Disruption ofStars: The Importance of the Impact Parameter and Stellar Structure.
Astrophys. J. , 25 (2013).43. Guillochon, J. & Ramirez-Ruiz, E. A Dark Year for Tidal DisruptionEvents.
ArXiv e-prints (2015).44. Loeb, A. & Ulmer, A. Optical appearance of the debris of a star dis-rupted by a massive black hole.
Astrophys. J. , 573 (1997).45. Strubbe, L. E. & Quataert, E. Optical flares from the tidal disruptionof stars by massive black holes.
Mon. Not. R. Astron. Soc. , 2070(2009).46. Vink´o, J. et al.
A Luminous, Fast Rising UV-transient Discovered byROTSE: A Tidal Disruption Event?
Astrophys. J. , 12 (2015).47. Shiokawa, H., Krolik, J. H., Cheng, R. M., Piran, T. & Noble, S. C.General Relativistic Hydrodynamic Simulation of Accretion Flow froma Stellar Tidal Disruption.
Astrophys. J. , 85 (2015).48. Piran, T., Svirski, G., Krolik, J., Cheng, R. M. & Shiokawa, H. Diskformation versus disk accretion - what powers tidal disruption events?
Astrophys. J. , 164 (2015).49. Kaastra, J. S., Mewe, R. & Nieuwenhuijzen, H. SPEX: a new code forspectral analysis of X and UV spectra. In Yamashita, K. & Watanabe,T. (eds.)
UV and X-ray Spectroscopy of Astrophysical and LaboratoryPlasmas , 411–414 (1996)., 411–414 (1996).